(Tom Schanz) Experimental Unsaturated Soil Mechani

springer proceedings in physics 112
springer proceedings in physics

91 The Dense Interstellar Medium 103 Computer Simulation Studies
in Galaxies in Condensed-Matter Physics XVII
Editors: S. Pfalzner, C. Kramer, Editors: D.P. Landau, S.P. Lewis,
C. Straubmeier, and A. Heithausen and H.-B. Schüttler
92 Beyond the Standard Model 2003 104 Complex Computing-Networks
Editor: H.V. Klapdor-Kleingrothaus Brain-like and Wave-oriented
Electrodynamic Algorithms
93 ISSMGE Editors: I.C. Göknar and L. Sevgi
Experimental Studies
Editor: T. Schanz 105 Computer Simulation Studies
in Condensed-Matter Physics XVIII
94 ISSMGE Editors: D.P. Landau, S.P. Lewis,
Numerical and Theoretical Approaches and H.-B. Schüttler
Editor: T. Schanz
106 Modern Trends in Geomechanics
95 Computer Simulation Studies Editors: W. Wu and H.S. Yu
in Condensed-Matter Physics XVI
Editors: D.P. Landau, S.P. Lewis, 107 Microscopy of Semiconducting Materials
and H.-B. Schüttler Proceedings of the 14th Conference,
April 11–14, 2005, Oxford, UK
96 Electromagnetics in a Complex World Editors: A.G. Cullis and J.L. Hutchison
Editors: I.M. Pinto, V. Galdi,
and L.B. Felsen 108 Hadron Collider Physics 2005
Proceedings of the 1st Hadron
97 Fields, Networks, Collider Physics Symposium,
Computational Methods and Systems Les Diablerets, Switzerland,
in Modern Electrodynamics July 4–9, 2005
A Tribute to Leopold B. Felsen Editors: M. Campanelli, A. Clark,
Editors: P. Russer and M. Mongiardo and X. Wu
98 Particle Physics and the Universe 109 Progress in Turbulence 2
Proceedings of the 9th Adriatic Meeting, Proceedings of the iTi Conference
Sept. 2003, Dubrovnik in Turbulence 2005
Editors: J. Trampetić and J. Wess Editors: M. Oberlack et al.
99 Cosmic Explosions 110 Nonequilibrium Carrier Dynamics
On the 10th Anniversary of SN1993J in Semiconductors
(IAU Colloquium 192) Proceedings
Editors: J. M. Marcaide and K. W. Weiler of the 14th International Conference, July
100 Lasers in the Conservation of Artworks 25-29, 2005, Chicago, USA
LACONA V Proceedings, Editors: M. Saraniti, U. Ravaioli
Osnabrück, Germany, Sept. 15–18, 2003 111 Vibration Problems ICOVP 2005
Editors: K. Dickmann, C. Fotakis, Editors: E. Inan, A. Kiris
and J.F. Asmus
112 Experimental Unsaturated
101 Progress in Turbulence Soil Mechanics
Editors: J. Peinke, A. Kittel, S. Barth, Editor: T. Schanz
and M. Oberlack
113 Theoretical and Numerical
102 Adaptive Optics Unsaturated Soil Mechanics
for Industry and Medicine Editor: T. Schanz
Proceedings
of the 4th International Workshop
Editor: U. Wittrock
Volumes 66–90 are listed at the end of the book.

T. Schanz (Ed.)
Experimental Unsaturated
Soil Mechanics
With 299 Figures and 67 Tables
123
Professor Dr. Ing. habil. Tom Schanz
Bauhaus-Universität Weimar
Laboratory of Soil Mechanics
Coudraystrasse 11c
99421 Weimar, Germany
tom.schanz@bauing.uni-weimar.de
ISSN 0930-8989
ISBN 978-3-540-69872-2 Springer Berlin Heidelberg New York
Library of Congress Control Number: 2007920611
This work is subject to copyright. All rights are reserved, whether the whole or part of the material is
concerned, specif ically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting,
reproduction on microf ilm or in any other way, and storage in data banks. Duplication of this publication or
parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its
current version, and permission for use must always be obtained from Springer-Verlag. Violations are liable
to prosecution under the German Copyright Law.
Springer is a part of Springer Science+Business Media
springer.com
© Springer-Verlag Berlin Heidelberg 2007
The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply,
even in the absence of a specif ic statement, that such names are exempt from the relevant protective laws and
regulations and therefore free for general use.
Typesetting: Integra Software Services Pvt. Ltd., Pondicherry, India
Cover design: WMX-Design, Heidelberg
Printed on acid-free paper SPIN: 11661139 42/Integra 5 4 3 2 1 0

Foreword
The event is a continuation of the series of International Conferences on Un-

saturated Soils in Germany. The first International Conference was held dur-
ing September 2003 in Bauhaus-University Weimar, Weimar, Germany. The
current event is the second one in the series entitled “Mechanics of Unsat-
urated Soils.” The primary objective of the Conference has been to discuss
and understand unsaturated soil behaviour such that engineered activities are
made better with times in terms of judgement and quality. We all realise by
now that in addition to the knowledge on the classical concepts, it becomes
an enormous challenging task to adapt convincing new concepts and present
them in such a way that it could be used in engineering practices. During
the last six years or so (2001–2007), scientific research works were exten-
sively taken up by five scientific research teams from five German universities,
whose scientific leaders are Wolfgang Ehlers (Universität Stuttgart), Jens En-
gel (HTW Dresden), Rainer Helmig and Holger Claas (Universität Stuttgart),
Tom Schanz (Bauhaus Universität Weimar), Christos Vrettos, Helmut Meiss-
ner and Andreas Becker (Universität Kaiserslautern). The research studies
involved theoretical and numerical approaches along with experimental stud-
ies on unsaturated soils. These two volumes present recent research findings
obtained within this collaboration by the above research groups along with
excellent contributions from several research groups throughout the World.
The experimental studies reported herein primarily focussed on the role
of microstructure and fabric for the complex coupled hydro-mechanical be-
haviour of cohesive frictional materials. Several papers considered the rele-
vance of temperature affecting the constitutive behaviour of clays. A careful
reader may recognise that in both the topics there is an ambiguity with re-
gard to the conclusions derived. Common features of state of the art theoret-
ical and numerical approaches, including TPM (theory of porous media) and
mixture theory, intend to describe the complex multi-field problems of fully
coupled thermo-hydraulic-mechanical-chemical initial-boundary value prob-
lems. Additional important field of research includes optimization of numer-
ical schemes to gain better computational performance. Applications include
VI Foreword
highly toxic waste disposals, slope stability problems and contaminants trans-
port in porous media. Some major significant contributions from the invited
and keynote speakers are also included.
I would like to extend my deep sense of appreciation as the editor and
the Head of the organizing committee, to many persons who have contributed
either directly or indirectly to organize the International conference and to
finalize these lecture notes. I would like to congratulate the authors for their
very interesting presentations and the reported results and advances in the
topics of the conference. I would like to thank all of those who promoted the
conference in their respective home countries. These two volumes would have
been not possible without financial support by the German Research Founda-
tion (DFG, Deutsche Forschungsgemeinschaft) through grant FOR 440/2. We
gratefully acknowledge the support of ISSMGE, especially TC6 “Unsaturated
Soils” with its chairman Eduardo Alonso. I appreciate the effort of the mem-
bers of the Technical committee and reviewers, who have spent their time to
select the valuable contributions and to suggest the changes improving the
presentation of the submitted papers. Finally, I wish to convey my thanks
to all the keynote and invited speakers, authors, and delegates attending the
conference.
I would like to express my deep sense of gratitude for the outstanding
work performed by those involved in the technical and administrative organi-
zation of these proceedings. Special thanks go to Yvonne Lins. Typesetting of
the proceedings was done by Venelin Chernogorov (alias Wily, Sofia Univer-
sity, Bulgaria, wily@fmi.uni-sofia.bg) in cooperation with Maria Datcheva.
Last but not least we appreciate the fruitful cooperation with Springer pub-
lishers, especially the guidance provided by Thomas Ditzinger.
Weimar, Tom Schanz

March 2007
Contents
Part I Microstructure and Fabric
Influence of Relative Density and Clay Fraction on Soils

Collapse
Khelifa Abbeche, Farid Hammoud, and Tahar Ayadat . . . . . . . . . . . . . . . . . 3
Microstructure Features in the Behaviour of Engineered

Barriers for Nuclear Waste Disposal
Pierre Delage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Microstructure of Gypsiferous Crust and Its Importance to
Unsaturated Soil Behaviour
Ghazi Mokdad, Omran Alshihabi, and Leo Stroosnijder . . . . . . . . . . . . . . . . 33
Fabric Changes in Compacted London Clay Due to Variations
in Applied Stress and Suction
Rafael Monroy, Lidija Zdravkovic, and Andrew Ridley . . . . . . . . . . . . . . . . 41
Microstructure of a Lime Stabilised Compacted Silt

Giacomo Russo, Sebastiana Dal Vecchio, and Giuseppe Mascolo . . . . . . . . 49
Part II Measuring Suction
Errors in Total Suction Measurements

Setianto Samingan Agus and Tom Schanz . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Application of a Dew Point Method to Obtain the Soil Water
Characteristic
Gaylon S. Campbell, David M. Smith, and Brody L. Teare . . . . . . . . . . . . . 71
VIII Contents
A Comparative Study of Soil Suction Measurement Using

Two Different High-Range Psychrometers
Rafaela Cardoso, Enrique Romero, Analice Lima, and Alessio Ferrari . . 79
Determination of the Soil Water Retention Curve with

Tensiometers
Sérgio Lourenço, Domenico Gallipoli, David Toll, Fred Evans, and
Gabriela Medero . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
Tensiometer Development for High Suction Analysis in

Laboratory Lysimeters
Cláudio Fernando Mahler and Abdoul Aziz Diene . . . . . . . . . . . . . . . . . . . . 103
Part III Strength and Dilatancy
Dilatancy of Coarse Granular Aggregates

Eduardo E. Alonso, Enrique F. Ortega Iturralde, and Enrique
E. Romero . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
A Laboratory Investigation into the Effect of Water Content
on the CBR of a Subgrade soil
Samuel Innocent Kofi Ampadu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
Shear Strength Affected by Suction Tension in Unsaturated
Fine Grained Soils?
Carola Bönsch and Christof Lempp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
Shear Strength Behaviour of Unsaturated Silty Soil

Ali R. Estabragh and Akbar A. Javadi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
Experimental Investigation on the Time Dependent Behaviour
of a Multiphase Chalk
Grégoire Priol, Vincenzo De Gennaro, Pierre Delage, and Thibaut
Servant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Testing Unsaturated Soil for Plane Strain Conditions: A New
Double Wall Biaxial Device
Tom Schanz and Jamal Alabdullah . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
Influence of State Variables on the Shear Behaviour of an
Unsaturated Clay
Viktoria Schwarz, Andreas Becker, and Christos Vrettos . . . . . . . . . . . . . . 179
Effect of Capillary and Cemented Bonds on the Strength of

Unsaturated Sands
Fabien Soulié, Moulay Saïd El Youssoufi, Jean-Yves Delenne, and
Christian Saix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
Contents IX
Determining the Shear Strength of Unsaturated Silt

Shulin Sun and Huifang Xu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Factors Affecting Tensile Strength Measurement and Modified
Tensile Strength Measuring Apparatus for Soil
Surendra Bahadur Tamrakar, Toshiyuki Mitachi, and Yasuo Toyosawa . . 207
The Tensile Strength of Compacted Clays as Affected by

Suction and Soil Structure
Rainer M. Zeh and Karl Josef Witt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
Part IV Temperature Effects
Modified Isochoric Cell for Temperature Controlled Swelling

Pressure Tests
Yulian Firmana Arifin and Tom Schanz . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Some Aspects of the Effect of the Temperature on the
Behaviour of Unsaturated Sandy Clay
Moulay Smaine Ghembaza, Said Taïbi, and Jean-Marie Fleureau . . . . . . . 243
Influence of Temperature on the Water Retention Curve of
Soils. Modelling and Experiments
Simon Salager, Moulay Saïd El Youssoufi, and Christian Saix . . . . . . . . . 251
Thermo-Hydro-Mechanical Behaviour of Compacted

Bentonite
Abbass Tavallali, Anh-Minh Tang, and Yu-Jun Cui . . . . . . . . . . . . . . . . . . . 259
Retention Curves of Two Bentonites at High Temperature
María Victoria Villar and Roberto Gómez-Espina . . . . . . . . . . . . . . . . . . . . 267
Part V Volumetric Behaviour – Expansive Materials
Experimental Study on Shrinkage Behaviour and Prediction

of Shrinkage Magnitudes of Residual Soils
Sarita Dhawan, Anil Kumar Mishra, and Sudhakar M . Rao . . . . . . . . . . . . 277
Assessment of Swelling Deformation of Unsaturated Kaolinite

Clay
Markus Dobrowolsky and Christos Vrettos . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
Suction and Collapse of Lumpy Spoilheaps in Northwestern
Bohemia
Vladislava Herbstová, Jan Boháč, and Ivo Herle . . . . . . . . . . . . . . . . . . . . . 293
X Contents
Oedometer Creep Tests of a Partially Saturated Kaolinite

Clay
Piotr Kierzkowski . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
Analysis of the Expansive Clay Hydration under Low

Hydraulic Gradient
Marcelo Sánchez, María Victoria Villar, Antonio Lloret,
and Antonio Gens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
Moisture Effects on Argillaceous Rocks

Chun-Liang Zhang and Tilmann Rothfuchs . . . . . . . . . . . . . . . . . . . . . . . . . . 319
Part VI Retention Behaviour
Results from Suction Controlled Laboratory Tests on

Unsaturated Bentonite – Verification of a Model
Ann Dueck . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
Variation of Degree of Saturation in Unsaturated Silty Soil

Ali R. Estabragh and Akbar A. Javadi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
Mechanical Behaviour of Compacted Scaly Clay During
Cyclic Controlled-Suction Testing
Camillo Airò Farulla, Alessio Ferrari, and Enrique Romero . . . . . . . . . . . 345
Prediction of Soil–Water Characteristic Curve Based on Soil

Index Properties
Navid Ganjian, Yadollah Pashang Pisheh, and Seyed Majdeddin Mir
Mohammad Hosseini . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
Water Balance and Effectiveness of Mineral Landfill Covers –
Results of Large Lysimeter Test-Fields
Wolf Ulrich Henken-Mellies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
A Retention Curve Prediction for Unsaturated Clay Soils

Mehrez Jamei, H. Guiras, and N. Mokni . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
Unsaturated-Zone Leaching and Saturated-Zone Mixing

Model in Heterogeneous Layers
Samuel S. Lee . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387
Prediction of SWCC for Coarse Soils Considering Pore Size

Changes
Xu Li and Limin Zhang . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
Contents XI
The Influence of the Pore Fluid on Desiccation of a

Deformable Porous Material
Hervé Péron, Liangbo Hu, Tomasz Hueckel, and Lyesse Laloui . . . . . . . . . 413
Determination of the Soil Water Retention Curve and the

Unsaturated Hydraulic Conductivity from the Particle Size
Distribution
Alexander Scheuermann and Andreas Bieberstein . . . . . . . . . . . . . . . . . . . . 421
Part VII Field Applications
Earthquake-Induced Mudflow Mechanism from a Viewpoint

of Unsaturated Soil Dynamics
Motoki Kazama and Toshiyasu Unno . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 437
Plate-Load Tests on an Unsaturated Lean Clay

Juan Carlos Rojas, Luis Mauricio Salinas, and Claudia Sejas . . . . . . . . . . 445
Selfsealing Barriers of Clay/Mineral Mixtures. The SB

Project at the Mont Terri Rock Laboratory
Tilmann Rothfuchs, Rüdiger Miehe, Norbert Jockwer, and Chun-Liang
Zhang . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453
Preferential Water Movement in Homogeneous Soils
Alexander Scheuermann and Andreas Bieberstein . . . . . . . . . . . . . . . . . . . . 461
Compaction Properties of Agricultural Soils
Anh-Minh Tang, Yu-Jun Cui, Javad Eslami, and Pauline
Défossez-Berthoud . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 475
Bearing Capacity of Model Footings in Unsaturated Soils
Sai K. Vanapalli and Fathi M.O. Mohamed . . . . . . . . . . . . . . . . . . . . . . . . . . 483
Influence of Soil Suction on Trench Stability
Valerie Whenham, Monika De Vos, Christian Legrand, Robert Charlier,
Jan Maertens, and Jean-Claude Verbrugge . . . . . . . . . . . . . . . . . . . . . . . . . . 495
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503
Part I
Microstructure and Fabric

Influence of Relative Density and Clay Fraction
on Soils Collapse
Khelifa Abbeche1 , Farid Hammoud1 , and Tahar Ayadat2

1
Civil Engineering Department, LARHYA, University of Batna, Avenue Chahid
Boukhlouf, Batna 05000, Algeria abbechek@yahoo.fr, hammoud_farid@yahoo.fr
2
Civil Engineering Department, University of Concordia, Sir George Williams
Campus, 1515 St. Catherine West, EV002. 139, Montreal Quebec, Canada
H3G-2W1 tayadat@yahoo.com
Summary. Most of the works conducted to investigate the parameters likely to

govern the behavior of collapsible soils, have been dedicated to initial dry density,
water content, degree of saturation and applied load. On the other hand, relatively
few works have been oriented towards the study of the influence of clay fraction and
relative density on the collapse of this type of soils. The experimental study, about
reconstituted soils, presented in this paper, aims at highlighting the effect of relative
density and clay fraction on the rate and magnitude of collapse.
Key words: collapsible soils, relative density, clay fraction, consolidation, oedome-
ter test, inundation
1 Introduction
Rapid growth in arid and semiarid regions has brought increased focus on
volume change characteristics of moisture sensitive unsaturated soils. One
such type of concern is collapsible soil. The latter is likely to undergo a rear-
rangement of its grains, and a loss of cementation, upon wetting, resulting in
substantial and rapid settlement under relatively low loads. Collapsible soils
occur as naturally debris flows, rapid alluvial depositions, and wind-blown
deposits. These soils are typically silt and sand size with a small amount of
clay. The collapse phenomenon is also likely to occur in the case of compacted
fills, used in man made structures such as earth dams, road subgrades, and
embankments, which are also placed in an unsaturated state.
Some investigations were aimed at developing identification and stabiliza-
tion methods (e.g. Jennings and Knight 1975, Ayadat and Belouahri 1996,
Ayadat et al. 1998, Abbeche et al. 2005, etc). On the other hand, some other
researchers (e.g. Ganeshan 1982, Ayadat et al. 1998, Cui and Magnan 2001,
etc) have concentrated their works on the collapse mechanism.
4 K. Abbeche et al.
Most of the work carried out on the parameters that govern the collapse
behavior, has focused on the initial dry density, the water content, the degree
of saturation and the applied load. Few studies have been conducted regarding
the influence of the relative density and the clay fraction on the collapse of
soils. The main interest of this paper lies in the study of the effect of these
parameters on the soil collapse.
2 Materials, Equipment and Testing Procedure

In the present study eleven reconstituted samples composed of sand and clay
in different proportions were tested. The proportions were chosen in order to
obtain samples which meet the collapse criteria proposed by Lutenegger and
Saber (1988) which distinguish between collapsible soils and non collapsible
soils.
The sand used is washed river sand, whose characteristics are summarized
as follows:
• sand equivalent, Es = 70%
• grain size distribution situated between 0.08 mm and 2.0 mm, of which
3.8% < 0.08 mm
• coefficient of uniformity, Cu = 2.5
• coefficient of curvature, Cc = 0.56
The clay used has the following characteristics:
• liquidity limit, wL = 41.50%
• plasticity limit, wP = 28.90%
• specific gravity of soil solids, Gs = 2.7
• percentage of particles finer than 2 μm (clay fraction), CF = 32%
The reconstituted soils have the geotechnical characteristics given in
Table 1 and the grain size distribution curves presented in Fig. 1.
Table 1. Geotechnical characteristics of soils

Soil S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11
Sand (%) 100 90 80 70 60 50 40 30 20 10 0
%PF < 80 μm 0 10 20 30 40 50 60 70 80 90 100
emax 0.765 0.773 0.817 0.828 0.846 0.913 0.955 1.007 1.035 1.184 1.472
emin 0.458 0.452 0.436 0.408 0.403 0.396 0.383 0.367 0.350 0.333 0.327
wL (%) – 13.7 15.1 16.1 19.3 21.4 26.8 34.0 34.3 40.6 41.5
wP (%) – 8.4 10.9 11.4 11.9 13.6 15.2 22.0 22.6 23.7 28.9
wopt (%) 9.64 7.86 7.94 8.29 8.64 11.08 13.01 14.89 16.48 18.66 22.14
%PF < 80 μm percentage of particles finer than 80 μm
Influence of Relative Density and Clay Fraction on Soils Collapse 5
Fig. 1. Grain size distribution curves of reconstituted soils used
The soils studied was compacted at a given water content and a dry unit
weight in a standard oedometric mold, in one layer, due to the small height of
the ring (20 mm). The equipment used for the compaction procedure, which
was made at the laboratory, is composed of a disk having a diameter slightly
smaller than that of the ring, which is fixed to a stem of guidance and a disk
shaped weight. The weight, having a mass of 152 g, sliding along the stem and
that falls from a 15 cm height, comes to strike the disk, thus compacting the
material in the oedometer ring.
The initial stage of sample preparation involves mixing the two compo-
nents so as to obtain a well homogenized soil. Then it is absolutely necessary
that meticulous preparation of a sample should be carried out to ensure a near
perfect fit in the oedometer ring. This is achieved by first kneading the soil
evenly in the oedometric mold and then compacting it with a certain energy
(number of weight drops) following the procedure described above. The sam-
ple was then struck off level with the top of the ring with a rigid steel blade,
to get a plane surface. After weighed, the specimen was put back in position
in the oedometer so as to carry out the compressibility test described by Jen-
nings and Knight (1975), for determining the collapse potential. The loading
stage of a sample, at an initial water content and an initial unit weight, took
place incrementally up to a pressure of 0.2 MPa and then flooded with water
and left for 24 hours, and the consolidation test is carried on to its normal
maximum loading limit. The sample settlements, before inundation and after,
at different time intervals had been measured. Figure2 shows the stages of the
collapse potential test. The collapse potential Cp is then defined as follows:
ΔeC
CP (%) = × 100 (1)
1 + e0
6 K. Abbeche et al.
Fig. 2. Typical collapse potential one dimensional consolidation test
where ΔeC is the change in void ratio of the sample upon flooding and e0 is
the initial void ratio before loading.
The tests were conducted on soils at different water contents and relative
densities. The relative density is defined as
emax − e0
Dr = (2)
emax − emin
The retained values of initial water content and relative density are:
• relative density, Dr : 10%, 30% and 50%
• initial water content, w0 : 2%, 4% and 6%
Each soil sample was tested at the values of Dr given above and for each
value, the initial water content was varied three times (i.e. 2%, 4% and 6%).
As explained below, soils S2, S3, S4 and S5 were also tested at a relative
density of 70%. Therefore, in all, 111 tests were carried out.
3 Results and Discussion

The results obtained clearly indicate that the collapses of the different soils
are in keeping with the classification of Jennings and Knight (1975) and agree
with the known properties of natural collapsible soils. For soils S1 to S11
the collapse potential Cp was found to vary from 0.16% to 12.55% for a water
content w0 = 2%, from 0.13% to 5.73% for a water content w0 = 4% and from
0.045% to 3.63% for a water content w0 = 6%. These results correspond to
the categories ranging from “no problem” to “severe trouble” (Table 2). On the
other hand, when the water content increases, the collapse potential decreases
or even can be equal to zero above a certain value. It is also noted that, at
a given water content, the collapse potential decreases with the increase of
the initial unit weight. These results are in agreement with those obtained
Table 2. Relation of collapse potential to the severity of foundation problem
CP (%) Severity of problem

0–1 No problem
1–5 Moderate trouble
5–10 Trouble
10–20 Severe trouble
> 20 Very severe trouble
by Lawton et al. (1989) and Ayadat et al. (1998). Consequently, it is noticed

that the artificially prepared soils possess a behavior analogous to those met
in situ, which justifies the tests program adopted.
The influence of the relative density on the collapse potential of soils is
presented in Figs 3a, 3b and 3c. It can be seen that CP decreases when
the relative density increases, whatever the soil type and the water content
considered. It is also noticed that for a given relative density, the collapse
potential decreases when the water content increases. For Dr = 50% and
w0 = 6%, Figure 3c shows that except soils S2, S3, S4 and S5, all tested soils
have a value of CP < 1 and are therefore not susceptible to collapse. It is also
noticed that the initial water content (i.e. 6%) is close to the values of the
optimum water content wopt of the collapsible soils studied (i.e. S2, S3, S4
(a) (b)
(c)
Fig. 3. Changes of collapse potential versus clay fraction, (a) for w0 = 2%, (b) for
w0 = 4%, (c) for w0 = 6%
8 K. Abbeche et al.
(a) (b)
(c) (d)
Fig. 4. Collapse potential variation versus relative density, (a) for soil S2, (b) for
soil S3, (c) for soil S4, (d) for soil S5
and S5) given in Table 1. Consequently, it was judged useful to test these four
soils with a more important relative density (Dr = 70%). The results obtained
are presented in Figs 4a, 4b, 4c and 4d. It can be noticed from these figures
that, for a water content w0 = 6%, the collapse potential is negligible when
the relative density is superior to 65%. It is also noticed that the more the
soils are loose and have low initial water content and the more the increase of
the collapse potential is important.
Figures 3a, 3b and 3c, which illustrate also the influence of the clay fraction
on CP , show that the collapse of soils depends on the clay content present in
their structure, which confirms the observation made by Lawton et al. (1992).
From these curves, it is clear that the collapse potential is negligible when the
clay percentage is greater than 30%. Below 5%, a collapse settlement, which
remains small, is likely to take place, while maximum collapse is reached for
about 15%. This result is in keeping with the interval established by Lawton
et al. (1992) who indicated that maximum collapse potential, for the natural
soils studied, is obtained when of the clay fraction is situated in the range
10% and 40%. The classification proposed in Table 3 allows identifying the
soils that are subjected to the collapse phenomenon.
Table 3. Proposed classification of collapsible soils
Clay Fraction CF (% < 2 μm) Probability of collapse

5% < CF < 15% High probability
15% < CF < 30% Probability of collapse
CF > 30% No collapse
4 Conclusions
This paper has presented results of one-dimensional oedometer collapse tests

using the procedure described by Jennings and Knight (1975). The main con-
clusions that can be drawn on the basis of the test results are as follows:
• It is possible to assess the soil collapsibility in the laboratory by means of
the relative density, on the one hand, and the clay fraction, on the other
hand.
• It is estimated that a soil is not subjected to collapse if its relative density
is superior to 65% with initial water content close to the optimum water
content.
• A soil is not susceptible to collapse if its clay fraction is greater than 30%.
References
Abbeche K, Mokrani L, Boumekik A (2005) Contribution à l’identification des sols
effondrables, Revue Française de Géotechnique, 110:85–90
Ayadat T, Belouahri B (1996) Influence du coefficient d’uniformité sur l’amplitude
et le taux de l’affaissement des sols, Revue Française de Géotechnique, 76:25–34
Ayadat T, Belouahri B, Ait Ammar R (1998) La migration des particules fines
comme approche d’explication du mécanisme de l’effondrement des sols, Revue
Française de Géotechnique, 83:1–9
Ayadat T, Ouali S (1999) Identification des sols affaissables basée sur les limites
d’Atterberg, Revue Française de Géotechnique, 86:53–56
Cui YJ, Magnan JP (2001) Affaissements locaux dus à l’infiltration d’eau en géomé-
canique environnemental, Hermes, pp. 139–164
Ganeshan V (1982) Strength and collapse characteristics of compacted residual soils.
Master Thesis, Asian Institute of Technology, Bangkok
Jennings JE, Knight K (1975) The additional settlement of foundation due to col-
lapse of sandy soils on wetting, Proc 4th Inter Conf on Soil Mechanics and
Foundation Engineering, 1:316–319
Lawton EC, Fragaszy RJ, James HH (1989) Collapse of compacted clayey sand, J
Geotechnical Engineering, ASCE 155(9):1252–1267
Lawton EC, Fragaszy RJ, Hetherington MD (1992) Review of wetting-induced col-
lapse in compacted soil, J Geotechnical Engineering 118(9):1376–1394
Lutenegger AJ, Saber RT (1988) Determination of collapse potential of soils,
Geotechnical Testing J 11(3):173–178
Microstructure Features in the Behaviour
of Engineered Barriers for Nuclear Waste
Disposal
Pierre Delage
Ecole Nationale des Ponts et Chaussées, Paris (CERMES, Institut Navier), 6–8 av.
B. Pascal, F–77455 Marne–la–Vallée cedex 2, France
pierre.delage@enpc.fr
Summary. Engineered barriers made up of bricks of compacted swelling clays are

considered as potential barriers for the isolation of high activity nuclear waste at
great depth. To better understand their coupled hydro-mechanical response, the mi-
crostructure of engineered barriers has been studied by various authors by using
most often scanning electron microscope observations and mercury intrusion pore
size distribution measurements (MIP). These studies confirmed that the microstruc-
ture of compacted bentonites was made up of aggregates with two classes of pores:
inter-aggregates pores and intra-aggregate pores. Further investigation conducted
by using X-Ray diffractometry at low angles conducted more recently provided
deeper insight into the hydration mechanisms that occur during hydration. This
paper presents some results obtained by various authors by using these techniques.
It shows how the hydration mechanisms occurring at the level of the clay particles
inside the aggregates help interpreting existing MIP data. Some conclusions about
the water retention properties of compacted bentonites with or without swelling al-
lowed are given. Some consequences on the water transfer properties of compacted
bentonites are also drawn. In both cases, the clogging of inter-aggregate pores due
to particle exfoliation has a significant effect.
Key words: engineered barrier, bentonite, hydration, microstructure, radioactive

waste, swelling, coupling
1 Introduction
Isolation of high activity nuclear waste at great depth is designed according
the concept of multibarrier system aimed at safely isolating the waste from
the biosphere. In this system, the placement of engineered barriers made up of
compacted swelling clay between the waste canister and the geological barrier
(host rock) as shown in Fig. 1 is considered as a possible option.
In this option, bricks of clay are statically compacted at given water con-
tent, often to unit masses as high as 1.8–2 Mg/m3 . Bricks are placed all around
12 Pierre Delage
Fig. 1. Typical scheme of a deep geological repository for nuclear waste disposal
(Gens and Olivella 2001)
the canister as shown in the figure, in order to fill the annular space between
the canister and the excavation. Engineered barriers are submitted to com-
plex thermo-hydro-chemo-mechanical actions due to the heat emitted by the
nuclear waste, to infiltration by pore water coming from the host rock, to the
movement of radionuclides through the barrier and to the emission of gas due
to the anaerobic degradation of the canister, among other things. A typical
feature that characterises the condition of engineered barriers is that they are
hydrated with very small swelling permitted in the volume between canister
and host rock. We will see later on that this particular situation is interesting
in terms of microstructure changes.
For these reasons, intense investigation has been carried out in the past
decades on the behaviour of engineered barriers. Researches clearly showed
that microstructure features played an essential role. Some effects of mi-
crostructure on the behaviour of engineered barriers are described in this
paper.
2 The Microstructure of Compacted Soils

2.1 Low Plasticity Soils
Following the initial Lambe’s model (1958) of the microstructure of compacted

soils that was based on the double layer theory, experimental observations of
the microstructure of compacted soils based on scanning electron microscopy
(SEM) and mercury intrusion porosimetry (MIP) afterwards provided a more
realistic description of the real microstructure. Pioneering works in this field
were provided by Barden and Sides (1970), Diamond (1970), Sridharan et al.
Microstructure Features in the Behaviour of Engineered Barriers 13
(1971). In mercury intrusion porosimetry, mercury, a non wetting fluid, is pro-

gressively intruded in a porous medium by increasing its pressure. As shown
by Laplace’s equation (1), the applied pressure can be related to an entrance
pore radius as follows:

1 1
pHg = σ cos θ + (1)
r1 r2
where r1 and r2 are the radius of curvature of the interface. Under the hy-
pothesis of a cylindrical pore of radius r, the air-water interface is spheri-
cal and r1 = r2 = r. Laplace’s relation between the mercury pressure and
the corresponding radius reduces to the following equation (sometimes called
Washburn’s equation, from Washburn 1921):
2σ cos θ
pHg = (2)
r
The interfacial capillary parameters of the mercury/solid phase/vacuum
system are the interfacial tension σ = 0.484 N/m and the contact angle θ =
141.3◦ (Diamond 1970). Mercury intrusion tests are performed by running a
progressive mercury pressure increase from vacuum up to 200 MPa. Tests have
to be performed on dehydrated samples. Most often, dehydration is made by
freeze-drying, where the sample is firstly quickly frozen by rapid immersion
in cooled liquid nitrogen and then sublimated in a freeze-dryer (Sridharan
et al. 1971, Tovey and Wong 1973, Gillott 1973, Delage and Pellerin 1984). To
avoid any boiling when plunging samples, liquid nitrogen needs to be cooled
to its freezing point (−210◦ C) by applying vacuum. Also, samples have to be
small enough to allow high speed freezing, that leads to a cryptocrystalline
structure with no volumetric expansion.
Ahmed et al. (1974) presented the pore size distribution (PSD) curves of
three samples of grundite compacted on the dry side, the wet side and at
optimum water content, as shown in Fig. 2.
The aggregate structure already suspected in soils compacted on the dry
side of the optimum (Barden and Sides 1970, Diamond 1970) was confirmed by
the bimodal shape of the PSD curve. This curve defines two pore populations,
namely the inter-aggregate and the intra-aggregate populations respectively.
In the figure, the entrance pore diameter of the inter-aggregate pores is defined
by the inflection point at 30 μm, which is quite large. Above the corresponding
pressure, mercury fills all the inter-aggregate porosity, finally surrounding the
aggregates themselves. Mercury starts to penetrate the aggregates at a pres-
sure that corresponds to the entrance diameter of 0.05 μm. On each curve,
the degree of saturation of the compacted sample is also plotted, showing the
inflection point corresponds to porous volume that was full of water. This con-
firms that the inter-aggregate pores in the dry sample are full of air. Results
similar to that of Ahmed et al. (1974) have been obtained by Delage et al.
(1996) and Romero et al. (1999), among others.
14 Pierre Delage
Fig. 2. PSD curves of a grundite statically compacted at various points of the

Proctor curve (Ahmed et al. 1974)
A SEM photo of a dry compacted silt of low plasticity (Jossigny silt) is

presented in Fig. 3. To ensure a well defined observation surface, samples have
been freeze-fractured just before being sublimated. In this technique (Delage
et al. 1982), the observation plane is defined by the fracture ice that works
like an impregnation resin. It does not follow any possible weakness path in
the microstructure and directly cross the different microstructure levels.
In Figure 3, the aggregates made up silt grains (10 to 20 μm in diame-
ters) are clearly apparent together with the inter-aggregate pores. In the dry
state, the clay fraction of Jossigny silt is not very apparent because platelets
are closely stuck to the silt grains. As commented previously, large inter-
aggregates pores are full of air and hence physico-chemical effects involving
clay water interactions take place inside the aggregates. Based on this photo,
the question of the location of air water menisci is interesting and indeed can
Fig. 3. SEM micrograph of a compacted Jossigny silt, dry side (Delage et al. 1996)
Fig. 4. SEM micrograph of a compacted Jossigny silt, wet side (Delage et al. 1996)
be further elucidated by using an environmental scanning electron microscope

in which observation is carried out on wet samples and where the ambient rela-
tive humidity can be increased. It is probable in Fig. 3 that menisci are located
inside the aggregates and that the connecting links between aggregates, that
govern the overall macroscopical resistance, are made up of hydrated clay
links.
A significantly different microstructure is observed on the wet side, as
shown in Fig. 4. The microstructure is more of a matrix type with dominant
appearance of clay platelets that constitute a matrix containing the silt grains.
Neither aggregates nor the corresponding pore population can be observed.
In other words, even in low plasticity soils, hydration makes the clay particles
become much more voluminous and visually apparent. The unimodal PSD
curve of Fig. 2 (also observed in the Jossigny silt) is compatible with the
matrix microstructure.
Going back to MIP investigation, an interesting feature relating mi-
crostructure changes when volume is decreased has been obtained from the
results of Sridharan et al. (1971), as presented in Fig. 5 (Delage and Graham
1995). The figure, that compares samples statically compacted at w = 21% at
various void ratios included between 0.85 and 0.56, shows that the decrease in
volume at constant water content only concerns the largest pores that exist
prior to compression. With the pore volume saturated by water in smallest
pores at w = 21% being represented on the figure, the figure also shows that
compression affected the pores full of air. The PSD of saturated pores smaller
than 0.2 μm in diameter is remarkably unaffected by the decrease in volume.
Note that a similar mechanism had been observed in a natural sensitive clay
of moderate plasticity by Delage and Lefebvre (1984). It appears here that
compression at void ratio smaller than 0.56 would concern a quasi-saturated
microstructure with build-up and subsequent dissipation of pore pressures.
16 Pierre Delage
Fig. 5. Effect of increased compaction on the pore size distribution of a compacted

soil (Delage and Graham 1995, after Sridharan et al. 1971)
2.2 Compacted Plastic Clays, Engineered Barriers
In common geotechnical engineering, compaction studies rather concerned low

plasticity soils able to constitute earthworks like embankments or earthdams,
with little attention given to plastic soils. This tendency has changed in the
past decade with increased attention given to the behaviour of engineered
barriers made up of heavily compacted swelling clays considered as possi-
ble barrier to be used in the multibarrier concept of isolating at great depth
high activity nuclear waste. Investigations carried out on engineered barri-
ers progressively showed the importance of microstructure effects for a sound
understanding of the coupled hydro-mechanical behaviour of engineered bar-
riers. In this context, the pioneering works conducted on low plasticity clay
described above have been extended to high plasticity compacted clays.
Microstructure units
Swelling clays (that include bentonites, a term used in practice) are made
up a significant amount of smectites, the active clay mineral responsible for
swelling when hydrated. As compared to less plastic clays like kaolinite, the
specificity of smectites (or montmorillonites) is related to the mineralogical
composition of the elementary layer, depicted in Fig. 6. Whereas kaolinite
is made up of the juxtaposition of one tetrahedral and of one octahedral
layer (see the figure where the composition and structure of these layers are
represented), a octahedral layer is located between two tetrahedral layer in
smectites. Whereas two adjacent kaolinite particles will be strongly linked
together because of possible hydrogen bridges acting between the oxygen and
hydroxyls ions on top and bottom side of the layer, no such strong link exists
between the two montmorillonite planes where oxygen atoms are located. As
Fig. 6. Structure of clay minerals (from Mitchell and Soga 2005). Left: kaolinite;
Right: montmorillonite
shown schematically in the figure, exchangeable cations with their hydrating

water molecules shells can penetrate this space due to the electrical attraction
caused by the electrical deficiency that is typical of clay minerals.
Note that in the case of illite that has the same structure as montmoril-
lonite, the stability of the stacking of various elementary layers is strongly
ensured by potassium ions K+ . In other words, the main difference between
kaolinite and illite on one side and montmorillonite on the other side is re-
lated to the stability of the clay platelet made up of stacks of elementary
layers stuck together. The various terms used to describe clay microstructure
are now recalled:
• The elementary layer, represented in Fig. 6, is defined for montmorillonite
by a 9.6 Å (96 nm) thickness in the direction perpendicular to the plane
(called d(001) ). This level can be observed indirectly by standard X-Ray
diffraction analysis. Direct observation can also be conducted by transmis-
sion electron microscope (see for example Ben Rhaïem et al. 1985).
• The clay platelet or particle (crystal or quasicrystal are somewhat found
in the literature, see Aylmore and Quirk 1962), made up the stacking of
elementary layers, with a number comprised between several (in the case of
clay suspensions) to several hundreds in dry states. This important feature
will be described further on in more detail. Clay particles can be observed
using scanning electron microscope techniques.
• The aggregate made up of clays particles aggregated together, and some-
what similar with those previously observed. Aggregates can be observed
18 Pierre Delage
using scanning electron microscope and detected by MIP, as discussed

above.
The microstructure of compacted bentonite
Figure 7 presents a SEM photo of a compacted Kunigel clay (Cui et al. 2002)
obtained on a freeze fractured and freeze-dried sample. Kunigel clay (Komine
and Ogata 1994) is a Japanese clay made up of 64% Na smectite with a
liquid limit wL equal to 474 (wP = 27) and a specific surface of 687 m2 /g.
The sample observed in the figure was statically compacted at a unit mass of
2 Mg/m3 and a water content of 8%, giving rise to an as-compacted suction
of 57 MPa.
Fig. 7. SEM photo of a compacted Kunigel clay (Cui et al. 2002)
In spite of the high compaction density, aggregates made up of aggregated

particles of clays are clearly apparent. Aggregates are separated by inter-
aggregates pores that appear to be significantly large with a diameter up
to 2–5 μm. Aggregates are composed of clay platelets made up of stacks of
elementary layers as described in Fig. 6. Advanced investigation based on low
angle X-Ray diffractometry on similar compacted bentonites by Saiyouri et al.
(2000) that will be discussed further on provide interesting further precisions
on the structures of the aggregates and on their changes during hydration and
swelling. Observations carried out using an ESEM have also been published
by Lloret et al. (2003) and Agus and Schanz (2005), among others.
MIP investigation conducted by various researchers (including Romero et
al. 1999, Cui et al. 2002, Lloret et al. 2003, Agus and Schanz 2005 and De-
lage et al. 2006) showed PSD curves similar to that of Ahmed et al. (1974)
showed in Fig. 2, evidencing the inter-aggregate and the intra-aggregate pore
Fig. 8. Intruded pore volume in compacted Almeria clay (Lloret et al. 2003)
populations (see Fig. 8). By comparing the intruded mercury volume at the
maximum mercury pressure to the total pore volume, as suggested in De-
lage and Lefebvre (1984), it is possible to check whether or not the sample
porosity has been entirely filled by mercury. This appeared to be the case in
low plasticity soils as for instance in the grundite of Ahmed et al. (1974) or
the silt of Delage et al. (1996). This is no longer true in the case of plastic
soils because of the presence of montmorillonite minerals and of the very thin
pores that exist inside the aggregates. As shown in Fig. 6, the clay platelets
are not stable with changes in water content and they may become quite thin
at higher water contents. The fineness of the platelets gives rise to thin pores
that are much thinner than the smaller entrance diameter accessible in MIP,
close to 6 nm (60 Å).
This point is illustrated in the PSD curve of compacted Almeria clay pre-
sented in Fig. 8 (Lloret et al. 2003). In this figure, the total intra-aggregate
void ratio appears to be much larger than the intruded one, due to an impor-
tant amount of pores of very small entrance diameter. Interestingly, the value
of this unintruded porosity is the same at the two densities considered in the
figure (dry unit mass of 1.55 and 1.73 Mg/m3 respectively, water content of
13%).
Observation of Fig. 8 also shows that the entrance inter-aggregate diameter
is in the order of 18 μm whereas the intra-aggregate entrance diameter is close
to 0.012 μm. Note also that the two curves are similar in shape for the pores
that are smaller than 1 μm. A presentation of the same curves as done in
Fig. 5 would also show more clearly that compression from a void ratio equal
to 0.57 down to 0.47 only affected the largest existing inter-aggregate pores
with no effect on the intra-aggregate porosity. In other word, no compression
of the aggregates occurred between these two void ratios.
20 Pierre Delage
3 Effect of Changes in Water Content

Swelling soils are known to be sensitive to changes in water content with
swelling when hydrated and shrinkage when dried. A common explanation is
generally based on the ability of water molecules that hydrate exchangeable
cations to penetrate in the interlayer planar spaces inside the platelets, as
mentioned in Fig. 6. The most commonly used theory of soil swelling is based
on the diffuse double layer theory and is based on the hypothesis of regularly
scattered individual clay layers, as suggested by Bolt (1956). Although hy-
dration mechanisms appear not to respect this hypothesis, this theory gave
some interesting results, mainly in the case of Na+ montmorillonite (see for
instance Sridharan and Jayadeva 1982, Tripathy et al. 2004, Schanz and Tri-
pathy 2005). The calculations of repulsion due to electrical interaction is very
dependent of the valence of the cations, with significant differences between
Na+ and Ca++ for example. Indeed, the behaviour of swelling soils also depend
of the valence of the dominant exchangeable cation with swelling properties
enhanced with Na+ , but in a different fashion from what the DDL theory pre-
dicts. No more considerations on DDL approaches will be made in this paper,
and interested readers are invited to read the papers mentioned previously.
Figure 9 shows the PSD curve of two compacted MX 80 bentonites (MX 80
is a Wyoming montmorillonite with wL = 520 and S = 700 m2 /g) compacted
at the same moderate density (1.32 Mg/m3 ) at two different water contents
(12.5 and 28.5%) with suction respectively comprised between 2 and 7 MPa
at 12.5% and higher than 50 MPa at 28.5% (Delage et al. 2006). The figure
Fig. 9. Effect of initial water content on two compacted MX 80 bentonite samples

at the same density (Delage et al. 2006)
shows that, whereas the two PSD curves are similar in shape – with however a
smaller entrance pore radius in the wettest sample – the unintruded porosity
is twice larger for the wettest sample at low suction.
The observation of the effect of drying and swelling on a heavily compacted
(ρd = 2 Mg/m3 ) sand bentonite mixture (50–50) made up of Calcigel bentonite
(50–60% in montmorillonite) at a water content of 11% has been carried out by
Agus and Schanz (2005). Sample was afterwards equilibrated using the vapour
equilibrium technique under a suction of 22.7 MPa that gave an equilibrated
value of water content equal to 9%. The swollen sample was obtained by lefting
a compacted specimen above distilled water in a desiccator for 180 days and
a final water content of 19% was reached. Results are presented in Fig. 10
in terms of PSD curves. As compared to the bimodal PSD curve of the as-
compacted sample that defines an inter-aggregate entrance pore diameter close
to 10 μm and an intra-aggregate entrance pore diameter close to 0.12 μm, the
PSD curve after swelling shows two points:
• A new pore population around 1 μm appears, at a diameter smaller than
that of the inter-aggregate porosity and larger than that of the intra-
aggregate porosity. The corresponding pore volume corresponds to 44% of
the total pore volume.
• A large amount of porosity is created below the limit of MIP, defining an
unintruded porosity approximately equal to the intruded one. This feature
is compatible with that observed on samples at higher water content in
Fig. 9, showing that samples under low suction have a large very small
porosity.
Fig. 10. Effect of wetting and drying on a heavily compacted Calcigel – sand mixture
(Agus and Schanz 2005)
22 Pierre Delage
Fig. 11. SEM observation of Kunigel clay at zero suction (Cui et al. 2002)
As already observed by Ahmed et al. (1974) on samples compacted dry of

optimum with bimodal porosity, there is little difference between the freeze-
dried as-compacted sample (suction 22.7 MPa) and the oven-dried sample
(suction estimated at 1 GPa according to Fredlund and Rahardjo 1993). Note
also that the unintruded porosities detected in the two curves are similar. This
shows that the suction of 22.7 MPa is high enough to already mobilize all the
shrinkage potential of both the aggregates themselves and the assemblage of
aggregates.
Figure 11 (Cui et al. 2002) presents two SEM photos carried out on sam-
ples of compacted Kunigel clay (described in the as-compacted state in Fig. 7)
after hydration and swelling under zero suction. Aggregates appeared to be
much less apparent in the various SEM observations conducted, as seen in
the figure. Typically, the microstructure rather appears to be of a honeycomb
type (Collins and McGown 1974) with clear presence of rather thin clay par-
ticle that form the walls of pores with a diameter of various micrometers.
Similar microstructure also typical of clay gels have been observed in swollen
bentonite (Marcial et al. 2002). The question arises as whether these pores
could correspond or not to the 1 μm entrance diameter observed in the PSD
curve of the swollen sample of Fig. 10.
4 Hydration Mechanisms in Smectites

Detailed investigations carried out on the hydration of dry smectites (Mooney
et al. 1952, Calvet 1972, Prost 1975, Sposito and Prost 1982, Bird 1984, among
others) showed that hydration initially occurred in the hydrophilic sites of the
clay, i.e. on the mineral surface and around the exchangeable cations in the
inter-layer spaces inside the particles. This first stage of hydration is called ho-
mogeneous hydration and it occurs by the ordered and progressive placement
of water molecules along the clay surface, layer after layer up to 4 layers. De-
tailed observation using transmission and scanning electron microscope also
showed that smectite particles change with changes in water content, with a
number of stacked clay layers that decreases when the clay is hydrated (Tessier
1990).
More recently, investigations using X-ray diffraction at low angles con-
ducted on samples of compacted MX 80 and FoCa 7 clays hydrated at various
suctions (Saiyouri et al. 1998, 2000, 2004) have been carried out. These re-
sults, based on Pons’ probabilistic model (Pons et al. 1981), provided essential
information on the hydration mechanism occurring at the particle level in com-
pacted clays similar to that considered here. Results obtained on compacted
MX 80 are presented in Fig. 12 a and b.
Starting from a high value of suction, hydration occurs in an organised
manner with the progressive placement of 1, 2, 3 and 4 layers of water
molecules along the clay monolayers in the interlayer spaces inside the par-
ticle. The corresponding distances between the clay monolayers (see Fig. 13)
start from 9.6 Å with no water (see also the clay layer structure represented in
Fig. 6) and are equal to 12.6, 15.6, 18.6 and 21.6 Å with 1, 2, 3 and 4 layers of
adsorbed water molecules, respectively. Note that the basal spacing 2d is ob-
tained by subtracting from the interlayer distance the clay thickness (9.6 Å),
giving 2d = 3, 6, 9 and 12 Å for 1, 2, 3 and 4 water layers, respectively.
Fig. 12. Effect of a suction decrease (hydration) on the MX 80 particles (after

Saiyouri et al. 1998, 2004): a) increase in the number of adsorbed water layers; b)
decrease in the number of stacked layers per particle
24 Pierre Delage
Fig. 13. Schematic distance between two layers of montmorillonite (Saiyouri et al.
2000)
As seen in Fig. 12a, the placement of water molecules is a function of

suction, with only one layer above 50 MPa (d = 1.5 Å), a second water layer
(d = 3 Å) when passing from 50 to 7 MPa, a third layer between 7 MPa and
60k̇Pa (d = 4.5 Å), and a fourth layer below 60 kPa (d = 6 Å). Similar results
were obtained on FoCa clay by the same authors and on other bentonites by
Delvaux et al. (1992). Simultaneously, the number of clay layers per particle
decreases with suction (Fig. 12b) starting with thick particles made up of
350 clay monolayers above a 50 MPa suction with a number of layers pro-
gressively decreasing down to 150 layers at 7 MPa and 10 layers at 3 MPa.
Hence, the swelling mechanism starting from high suction is the result of two
combined phenomena:
i) the progressive insertion of successive layers of water molecules in the
interlayer spaces inside he particles;
ii) the subdivision of particles into thinner ones that are made up of a smaller
number of stacked layers. Note that the observation of the compacted
Kunigel clay submitted to zero suction in Fig. 11 where thin clay particles
are observed is compatible with this mechanism.
In terms of changes in pore morphology inside aggregates in zones that
cannot be investigated by mercury intrusion, these mechanisms have the fol-
lowing consequences :
i) the enlargement of the planar spaces located inside the clay particles at
given distances that are function of the suction applied;
ii) the progressive creation inside the saturated aggregates of larger inter-
particles pores that are not necessarily planar.
By considering at various given suctions the value of the specific surface
of the clay, the number of layers of adsorbed water along the clay layers and
the number of clay layers stacked in a particle, Saiyouri et al. (2000, 2004)
Fig. 14. Water distribution in MX 80 compacted sample as a function of suction.

W is the water content, V1 to V4 are the water amounts in the interlayer spaces
containing 1 to 4 water layers (Saiyouri et al. 2004)
were able to distinguish, in the total water content of the clay, the interlayer
water from the interparticle water, as seen in Fig. 14. Comparison beetwen
the V4 points (that correspond to the totality of the interlayer water) and the
points representing the total water content gives an indication of quantity of
the interparticle water with respect to suction. The figure shows that a signif-
icant amount of interparticle water starts to appear during hydration when
suction goes down below 60–70 MPa. This means that above 70 MPa, swelling
is mainly due to the interlayer water. At 1 MPa, 22 g/100 g of water are lo-
cated between layers whereas approximately 13 g/100 g of water are located
in interparticle pores, with distances between particle larger than 21.6 Å. At
a suction of 10 kPa, close to the fully swollen state, the maximum distance is
estimated close to 150 Å. The authors also stated that these interparticle dis-
tances were large enough to allow for the development of diffuse double layers
close to particle surfaces inside the aggregates. In other words, the DDL the-
ory could be able to play a significant role in swelling at suctions below 5 MPa
where it could account for approximately 25% of the total amount of absorbed
water. Conversely, DDL phenomena could account for approximately 70% of
the volume of absorbed water under a suction of 10 kPa, i.e. in the swollen
state.
Considering the change in aggregate microstructure with suction, it can
be said that aggregates in a drier sample are made up of a smaller number of
thicker particles whereas aggregates in a wetter sample are made up of a larger
number of thinner particles, giving rise to a larger proportion of inter-particle
porosity. As derived from Fig. 12, suction changes have a well defined effect
on the intra-aggregate porosity. The values of suction at which changes in
the number of adsorbed water layers occur apparently do not depend on the
26 Pierre Delage
bentonite considered. These data will help interpreting the very thin porosity
that is detected in pore size distribution curves below 3.5 nm.
The data of Fig. 12 also suggest that the threshold values of suction (or
water potential) evidenced in the figure correspond to energetic barriers that
separate the various levels of water molecules adsorbed in the inter-layers
spaces. Physically, this correspond to a quite strong adsorption links for the
first layer (water potential higher than 50 MPa) and to a weak one for the
third layer (water potential smaller than 7 MPa). These values are apparently
independent of the smectite considered (Saiyouri et al. 2004, Delvaux et al.
1992).
5 Application to Compacted Bentonites

5.1 About the Intra-Aggregate Porosity
In an attempt to extrapolate pore size distribution curves in the range of pores

smaller than 3.5 nm in radius, it appears useful to deduce from previous data
the dimensions that correspond to the two intra-aggregates pore populations
that have been investigated by X-Ray low angle diffractometry, as follows:
• Interparticle pores radii deduced by Saiyouri et al. (1998, 2004). In a first
approximation, it seems reasonable to approximate these pores as cylindri-
cal pores as in standard mercury intrusion analysis. Interparticle distances
estimated between 20 and 100 Å corresponds in Fig. 15 to pore radii of 10
and 50 Å, i.e. 1 to 5 nm.
Planar interlayer pores located inside the particles. In this case, r2 in
equation (1) is infinite and Laplace’s law writes:
σ cos θ
pHg = (3)
r
Consequently, the x scale of the pore size distribution curves in zones where
planar pores are suspected to exist should be divided by two. As commented
earlier, the adsorption of 1, 2, 3 and 4 water layers correspond to d distances
of 1.5, 3, 4.5 and 6 Å, respectively.
These approximate estimations can be used to better understand the pore
morphology at dimensions smaller than 3.5 nm, as shown in Fig. 15. They
help to interpret the trend shown by the pore size distribution curves in this
zone.
In the figure, the representation of the range of size of interparticle pore
estimated using XRD at low angle size (radii between 10 and 50 Å) is reason-
ably compatible with the order of magnitude of the intra-aggregate porosity
given by the PSD curve.
The morphology inside the aggregate is now depending of the sample suc-
tion in the way defined in Fig. 12. In this regard, the interlayer distances
Fig. 15. Integration of XRD low angle data in PSD
corresponding to 3 layers (4.5 Å) and 1 layer (1.5 Å) of water molecules have

been plotted in an extended logarithmic x axis in the PSD graph of Fig. 15.
The higher value of the very thin porosity (r < 3.7 nm) observed in the
looser sample at w = 28.5% can be related to a low initial value of suction
(approximately 4 MPa) that corresponds to a higher number of thinner par-
ticles that contain between 10 and 100 clay layers. The particle division at
this suction is compatible with the development of a significant inter-particle
porosity, corresponding to the high value of very thin porosity of the sample.
The smaller amount of very thin porosity observed in the sample at w = 12.5%
can be related to a higher suction ( > 50 MPa) characterised by less numer-
ous thicker particles (350 layers) that have only one layer of adsorbed water
molecules.
Figure 14 can also be used to interpret some aspects of the change in mi-
crostructure observed in Fig. 10 in swollen Calcigel clay subsequently to suc-
tion reduction to zero. As suggested by the mechanism illustrated in Fig. 12,
the zero suction hydrated state is composed of thin particles with a significant
amount of interparticle pores. Observation of the corresponding PSD curve in
Fig. 10 indeed shows a considerable pore volume of small dimensions suggested
by the shape of the curve at pore diameters smaller than 10 nm (100 Å). Be-
sides the change observed at a diameter of 1 μm, the swollen volume at small
pores in Fig. 10 corresponds to the two kinds of swelling mechanisms iden-
tified in Fig. 14, i.e. inter and intra-particle mechanisms. In this figure, the
disappearance of the inter-aggregate porosity can be related to the filling of
these pores by the exfoliation of the aggregates limiting the inter-aggregate
pore. Actually, the exfoliation process can also be interpreted with similar
mechanisms of reduction in thickness and separation of particles in thinner
particles. This process helps filling the inter-aggregates pores with a gel hon-
eycomb structure similar to that observed in Fig. 11.
28 Pierre Delage
5.2 On the Retention Properties of the Confined Engineered

Barrier
As commented earlier, very little expansion is allowed to the engineered barrier

submitted to hydration between the canister and the wall in the host rock.
This situation has been suspected to have some effects on the retention and
transfer properties of the engineered barrier. A specific experimental device
aimed at determining the retention properties of the engineering barrier with
no volume changes allowed has been proposed by Yahia-Aissa et al. (2001). In
this device, compacted clay samples are inserted in rigid cells that are specially
designed to allow vapour exchanges. The water retention properties with no
volume changes have been determined by placing the cells in desiccators with
saturated saline solutions at constant temperature.
The data in Fig. 16 show firstly the retention properties obtained in a
standard manner in desiccators with volume changes allowed (see Delage et al.
1998 for more details). The graph gathers experimental data from a compacted
FoCa 7 bentonite and that of FoCa 7 powder. There is very little difference
between the two data set, showing negligible influence of the sample density
and indicating the predominant effects of physico-chemical actions, since no
particular pore configuration can be invoked to justify any capillary effect in
the case of the powder. The threshold values of suction (50 and 7 MPa) derived
from the analysis of Fig. 12 are also reported in the y log axis. Starting from
an initial suction of 110 MPa, it is possible to indicate on the water retention
curve the number of intra-particle water layers, defining three zone with 1, 2
Fig. 16. Water retention properties of FoCa 7 engineered barrier (Yahia-Aissa et

al. 2001)
Fig. 17. Effect of volume constraint on the retention properties of FoCa 7 barrier
and 3 layers. Simultaneously, as indicated in the figure, the number of layers

in the clay particles decreases from 300 down to 10.
The effect of volume constraint is described in Fig. 17. One can observe
that the constant volume condition does not allow anymore the adsorption of
third layer, whereas the adsorption of the two first layers of water seems to
occur, like in the free swelling condition, from the initial state (s = 110 MPa)
down to a suction of 7 MPa. As suggested previously, this seems to indicate
that the adsorption of the two first layers that occurs with a number of clay
layers decreasing from 350 down to 150 in the case of compacted FoCa 7 clay
(Saiyouri et al. 1998) corresponds to a significantly higher level of energy than
the third layer, when also the number of layers per particle goes down to 10
with significant creation of inter-particle pores. In other words, the main effect
of volume constraint is to prevent the separation of particles into thinner ones
and the corresponding development of inter-particle voids. It is also likely that
the creation of extra particles when passing from 350 layers to 150 clay layers
progressively clog the initial inter-aggregate porosity.
6 Concluding Remarks
Microstructure studies of engineered barriers have been developed in various

research groups in Europe in the last decade to better understand the cou-
pled hydro-mechanical behaviour of compacted bentonites used as engineered
barriers to isolate high activity nuclear waste at great depth. The examina-
tion of some results published in the literature and their interpretation based
on the careful description of the process of hydration of smectites allowed
30 Pierre Delage
to complete the understanding of the changes in microstructure that occur

in compacted bentonite during hydration. The description of the changes in
shape of the smectite particles with suction allow to describe more precisely
what occurs inside the aggregates of the compacted bentonite. As compared
with compacted low plastic soils in which an analysis in terms of aggregates
made up of stable particles is sufficient, the precise description of the change
in shape with suction of the clay particles inside the aggregates give significant
complementary information.
Some consequences have been drawn in this paper in terms of retention
properties. Other important consequences also affect transfer properties, as
suggested by Loiseau et al. (2002) who also showed that the changes in water
permeability of a compacted bentonite during infiltration was not as expected
in standard unsaturated soil mechanics analysis. Similar mechanisms also help
understanding some ageing effect observed in compacted bentonite (Delage
et al. 2006).
Obviously, the high sensitivity of the smectite microstructure to changes
in water content invites to some precautions when using standard approaches
of describing the retention and transfer phenomena in unsaturated compacted
bentonites. This is particularly true in numerical calculations aimed at pre-
dicting the rate of infiltration of the engineered barriers.
References
Agus SS, Schanz T (2005) Effect of shrinking and swelling on microstructures and
fabric of a compacted bentonite-sand mixture, In: Proc Int Conf on Problematic
Soils, Cyprus 2:543–550
Ahmed S, Lovell CW, Diamonds S (1974) Pore sizes and strength of compacted clay,
ASCE J Geotechnical Eng 100:407–425
Aylmore LAG, Quirk JP (1962) The structural status of clay systems, In: Proc 9th
Nat Conf on Clays and Clay Minerals. Pergamon, Oxford, pp 104–130
Barden L, Sides GR (1970) Engineering behaviour and structure of compacted clay,
ASCE J Soil Mech Found Div 96:1171–1200
Ben Rhaïem H, Pons CH, Tessier D (1985) Factors affecting the microstructure of
smectites: role of cations and history of applied stresses, In: Schultz et al. (eds)
Proc Int Clay Conf, Denver, The Clay Mineralogical Soc, pp 292–297
Bird P (1984) Hydration phase diagrams and friction of montmorillonite under lab-
oratory and geologic conditions with implications for shale compaction, slope
stability and strength of fault gauge, Tectonophysics 107:235–260
Bolt GH (1956) Physico-chemical analysis of the compressibility of pure clays,
Géotechnique 6:86–93
Calvet R (1972) Adsorption de l’eau sur les argiles; Etude de l’hydratation de la
montmorillonite, Bull Soc Chimique de France 8:3097–3104
Cui YJ, Loiseau C, Delage P (2002) Microstructure changes of a confined swelling
soil due to suction controlled hydration, In: Proc 3rd Int Conf on Unsaturated
Soils, UNSAT’2002. Recife, Brazil, Balkema 2:593–598
Collins K, McGown A (1974) The form and function of microfabric features in a

variety of natural soils, Géotechnique 24(2):233–254
Delage P, Pellerin FM (1984) Influence de la lyophilisation sur la structure d’une
argile sensible du Québec, Clay Minerals 19:151–160
Delage P, Tessier D, Marcel-Audiguier M (1982) Use of the Cryoscan apparatus
for observation of freeze-fractured planes of a sensitive Quebec clay in scanning
electron microscopy, Can Geotech J 19(1):111–114
Delage P, Audiguier M, Cui YJ, Howat MD (1996) Microstructure of a compacted
silt, Can Geotech J 33:150–158
Delage P, Graham J (1995) The mechanical behaviour of unsaturated soils, In:
Alonso EE, Delage P (eds) Proc 1st Int Conf on Unsaturated soils Paris, Balkema
3:1223–1256
Delage P, Lefebvre G (1984) Study of the structure of a sensitive Champlain clay
and its evolution during consolidation, Can Geotech J 21(1):21–35
Delage P, Howat M, Cui YJ (1998) The relationship between suction and swelling
properties in a heavily compacted unsaturated clay, Eng Geology 50(1–2):31–48
Delage P, Marcial D, Cui YJ, Ruiz X (2006) Ageing effects in a compacted bentonite:
a microstructure approach, Géotechnique 56(5):291–304
Delvaux B, Tessier D, Herbillon A, Burtin G, Jaunet AM, Vielvoye L (1992) Mor-
phology, texture and microstructure of halloysiticsoil clays as related to weath-
ering and exchangeable cations, Clays and Clay Min 40(4):446–456
Diamond S (1970) Pore size distribution in clays, Clays and Clay Min 18:7–23
Diamond S (1971) Microstructure and pore structure of impact compacted clays,
Clays and Clay Minerals 19:239–241
Fredlund DG, Rahardjo H (1993) Soil mechanics for unsaturated soils. Wiley,
New-York
Gens A, Olivella S (2001) Clay barriers in radioactive waste disposal, Revue
Française de Génie Civil 5(6):845–856
Gillott JE (1973) Methods of sample preparation for microstructureal analysis of
soil, In: Soil Microscopy, Proc 4th Int Working Meeting on Soil Micromorphology,
Kingston, pp 143–164
Komine H, Ogata N (1994) Experimental study on swelling characteristics of com-
pacted bentonites, Can Geotech J 31:478–490
Lambe TW (1958) The structure of compacted soil, J Soil Mech Found Div
84(SM2):1–34
Loiseau C, Cui YJ, Delage P (2002) The gradient effect on the water flow through a
compacted swelling soil, In: Proc. 3rd Int Conf Unsaturated Soils, UNSAT’2002
Recife, Brazil, Balkema 1:395–400
Lloret A, Villar MV, Sanchez M, Gens A, Pintado X, Alonso EE (2003) Mechanical
behaviour of heavily compacted bentonite under high suction changes, Géotech-
nique 53(1):27–40
Marcial D (2003) Comportement hydromécanique et microstructural des matériaux
de barrières ouvragées. PhD Thesis, Ecole des Ponts, Paris
Marcial D, Delage P, Cui YJ (2002) On the high stress compression of bentonites,
Can Geotech J 39:1–9
Mitchell JK, Soga K (2005) Fundamentals of soil behaviour. John Wiley, New-York
Mooney RW, Keenan AC, Wood LA (1952) Adsorption of water vapor by montmo-
rillonite. II. Effect of exchangeable ions and lattice swelling as measured from
X-ray diffraction, J Amer Chem Soc 74:1371–1374
32 Pierre Delage
Nalezny CL, Li MC (1967) Effects of soil moisture and thixotropic hardening on the
swell behaviour of compacted expansive soils, Highway Research Record No 209.
Highway Research Board, Washington DC
Pons CH, Rousseaux F, Tchoubar D (1981) Utilisation du rayonnement synchrotron
en diffusion aux petits angles pour l’étude du gonflement des smectites, Clay
Minerals 16:23–42
Prost R (1975) Interactions between adsorbed water molecules and the structure of
clay minerals: hydration of smectites. In: Bailey SW (ed) Proc Int Clay Conf
1975: Mexico City, Mexico. Applied Pub, Wilmette IL:351–359
Romero E, Gens A, Lloret A (1999) Water permeability, water retention and mi-
crostructure of unsaturated compacted Boom clay, Eng Geology 54:117–127
Saiyouri N, Hicher PY, Tessier D (1998) Microstructural analysis of highly com-
pacted clay swelling, In: Proc. 2nd Int Conf on Unsaturated Soils Academic
publishers, Beijing, China 1:119–124
Saiyouri N, Hicher PY, Tessier D (2000) Microstructural approach and transfer
water modelling in highly compacted unsaturated swelling clays, Mech Cohesive
Frictional Mater 5:41–60
Saiyouri N, Tessier D, Hicher PY (2004) Experimental study of swelling in unsatu-
rated compacted clays, Clay Minerals 39:469–479
Schanz T, Tripathy S (2005) Soil-water characteristic curves of clays from physico-
chemical concepts, In: Proc Int Conf on Problematic Soils. Famagusta, Cyprus
1:219–228
Sposito G, Prost R (1982) Structure of water adsorbed on smectites, Chem Rev
USA 82:552–573
Sridharan A, Altschaeffl AG, Diamond S (1971) Pore size distribution studies, ASCE
J of the Soil Mech Found Div 97:771–787
Sridharan A, Jayadeva MS (1982) Double layer theory and compressibility of clays,
Géotechnique 32(2):133–144
Tessier D (1990) In: Decarreau A (ed) Matériaux Argileux, Structure, Propriétés et
Applications. Soc. Fr. Minéralogie et Cristallographie, Paris, France 1:387–445
Tovey NK, Wong KY (1973) The preparation of soils and other geological materi-
als for the scanning electron microscope, In: Proc Int Symp on Soil Structure,
Gothenburg, Sweden, pp 176–183
Tripathy S, Sridharan A, Schanz T (2004) Swelling pressures of compacted ben-
tonites from diffuse layer theory, Can Geotech J 41:437–450
Washburn EW (1921) A method of determining the distribution of pore sizes in a
porous material, Proc of the National Academy of Science 7:115
Yahia-Aissa M, Delage P, Cui YJ (2001) In: Adachi, Fukue (eds) Suction-water
relationship in swelling clays, Clay Science for Engineering, IS-Shizuoka Int Symp
on Suction, Swelling, Permeability and Structure of Clays, Balkema, pp 65–68
Microstructure of Gypsiferous Crust and Its
Importance to Unsaturated Soil Behaviour
Ghazi Mokdad1 , Omran Alshihabi1 , and Leo Stroosnijder2

1
Faculty of Agriculture, Damascus University, Syria ghazi57@scs-net.org,
omran-alshihabi@hotmail.com
2
Department of Irrigation and Soil & Water Conservation, Wageningen
Agricultural University, The Netherland Leo.Stroosnijder@wur.nl
Summary. Formation of gypsiferous crust in agriculture lands deters seedling emer-

gence, which constitutes a major problem around the world. The objective of this
study is to investigate the micro-morphology of gypsiferous crust and its influence
on unsaturated soil behavior. Samples containing 47% of gypsum were taken from
Granada (North Syria), then left to dry under natural conditions. Undistributed soil
samples (8 × 8 cm) of the crust and underlying soil material were taken in tins and
impregnated with polyester resin. Thin sections were prepared, and examined with a
petrographic microscope (Olympus) in plane and polarized light. The microstructure
of these samples showed the distribution of soil materials and the types of pores. The
crust consisted of clay, silt, calcite and gypsum attributed to the mechanical impact
of raindrops. The main effect of raindrops on the soil surface layer was clogging the
pores by means of the mentioned materials, which was the dominant mechanisms
of crust formation. According to the structure of these soils, the infiltration rate
reached a very low value and a remarkable increase in stiffness was observed.
Key words: undistributed soil, unsaturated soil, microstructure, gypsiferous crust,

infiltration rate
Introduction
Gypsiferous soils Make up 22% of Syrian agricultural soil, with gypsum con-
tent of 1–80% (Ilaiwi 1983). A crust forms when these soils are exposed to
rainfall or irrigation (Stroosnijder 1995). This crust hinders seed germination,
and possible recultivation and planting. This study was initiated to describe
the micromorphology of gypsiferous crust.
Gypsum crusts have been reported from many (semi-) arid regions. Their
geographic distribution closely coincides with the areas receiving less than
250 mm rainfall per year (Watson 1982). Extensive gypsum crusts have been
described in Middle East countries, where they are a major limitation for
crop production because water infiltration rate, seedling emergence and crop
34 G. Mokdad et al.
growth are largely controlled by the thickness and gypsum content of the
crust (Nafie 1989). The severity of the mechanical hindrance that gypsum
crusts oppose to crop development can be assessed by measuring penetra-
tion resistance. Morphological and micro morphological study of gypsiferous
soils from the Middle Euphrates Basin (Syria) have been studied (Habib and
Robert 1992). The effect of moisture and gypsum content on the penetration
resistance of gypsiferous horizons has been studied by Poch and Verplancke
(1997) but gypsum crusts developing on the soil surface have not been given
yet the same attention (Jafarzadeh 1991). The objective of the present paper
is to investigate the microstructure of gypsiferous crust in the soil surface.
Materials and Methods

Soils: The soils chosen for this study from Granada which are located in
North-East Syria (35◦ 57 9 N – 38◦ 48 4 E and 290 m above the sea). The
mean annual rainfall is 250 mm, distributed in winter season with high rain-
fall intensity (47 mm/h) and long period of drought. The soils were silty loam
(gypsiferous soil). Soil samples are taken from the top layer (0–20 cm) of un-
cultivated area, The samples were characterized for particle size distribution
using hydrometer method (Gee and Bauder 1986), calcium carbonate content
using the volumetric calcimeter method (Nelson 1986), and organic matter
content by wet combustion (Nelson and Somers 1986). Results are presented
in Table 1. Gypsum content in the soil is high (47%). The organic matter
content in the soil was low (1.2%). Other undistributed samples are taken,
packed with boxes and then coated by cotton to maintain the crust surface
with no alterations.
Thin sections preparation: Undistributed soil samples (8 × 8 cm) of the
crust and underlying soil material were packed with boxes and coated with
cotton to maintain the crust surface with no alterations. To study the micro
morphology of the soil crust in Wageningen university (Holland). Thin sections
preparation were made according to the method of (FitzPatrick 1993), the
samples were oven dried (40–50◦ C) for 48 hours to remove hydroscopic water.
Then the samples were taken in tins and impregnated with polyester resin, and
it is included at vacuum for four weeks, to be given the cohesion. After that
samples were cut, polished, trimmed with oil used, and lastly covered and in
examined with a petrographic microscope (Olympus) in plane and polarized
light. The micro morphological description used the terminology of Bullock
et al. (1985).
The determination of the mineralogical gypsum crust was carried out
with X-ray diffraction technique. For that a Philips PW 1710/1820 diffraction
equipped with a co X-ray tube was used (λ = 1.7889 Å).
Microstructure of Gypsiferous Crust 35
Table 1. Physical and chemical analysis of Granada soil
Organic
Symbol USDA pH ECs CaCO3 Gypsum
Granulo. Analysis matter
soil Texture.T 1:2.5 dS/m % %
OM%
Clay Silt Sand
G Si.L 7.8 4.2 1.2 13.0 47.0
35.4 55.9 8.5
Results and Discussion
Crust Formation
The crust consists of two layers (layer 1 and layer 2, Fig. 1), which are parallel
oriented to each other. Each layer displays a textural fining upward towards
the soil surface
Layer 1 is situated at the soil surface. The upper part is 100 μm thick
(Fig. 2) densely packed and it consists of a mixture of clay and silt sized calcite
particles (< 10 μm). Occasional gypsum crystals of similar sizes are present.
The microstructure is compact grain microstructure, macro voids (> 10 μm)
are absent.
The lower part (Fig. 3) is built up of denticular gypsum crystals (100–
300 μm) and micro aggregates consisting of a mixture of clay and calcite crys-
tals (< 10 μm). Occasionally calcite crystals ranging in size from 10–150 μm
are present. Voughs (50–150 μm) frequently occur showing that the lower part
of layer 1 is more porous than the upper part.
Fig. 1. General overview of the crust

36 G. Mokdad et al.
Fig. 2. Detail of the upper part of layer 1
Fig. 3. Detail of the lower part of layer 1
Layer 2 is situated below layer 1.

The upper part (Fig. 4) is similar to the upper part of layer 1 regarding the
grain size distribution, mineralogy, density and microstructure. The thickness
of the upper part ranges from 150 to 250 μm.
The lower part (Fig. 5) is built up of denticulated gypsum crystals smaller
sized than those of layer 1 (50–100 μm). In addition micro aggregates (50–
100 μm) consisting of clay and fine (< 10 μm) calcite crystals are present.
Occasionally calcite crystals (50–100 μm) occur. Few voughs (< 50 μm) are
present, the microstructure is compact grainy. Layer 2 is finer textured, sig-
Fig. 4. Details of the upper part of layer 2
Fig. 5. Detail of the lower part of layer 2
nificant less porous than layer 1. Some plant tissues are present. The boundary
between layer 1 and layer 2 is within 50 μm. The total thickness of the crust
is approximately 2–3 mm.
The Soil Groundmass
As it is shown in Fig. 6, the groundmass consists of gypsum crystals and

gypsum aggregates up to 1 cm. Occasionally calcite crystals occur. The fine
material (< 10 μm) consists of clay and calcite particles. The distribution
38 G. Mokdad et al.
Fig. 6. Detail of the soil groundmass
pattern of the gypsum crystals and the fine material is random. Fragments of
crusts are frequently found. They reflect previous stages of crust formation.
The results of X-ray diffraction analysis of undistributed samples of gyp-
sum crusts are presented very much gypsum on the crust, much calcite, mod-
erate amounts of quartz and few clay minerals.
The two distinguished layers within the observed crust represent two differ-
ent stages of surface crust formation. Lenticular gypsum crystals are produced
under semiarid conditions by gypsum weathering.
Eswaran et al. (1981) reported the occurrence of many very fine, lenticular
crystals of gypsum, very closely packed. A petrogypsic horizon surface crust
is governed by a complex sequence of soil particle detachment and transport
processes at the soil surface. First particle detachment is achieved by various
mechanisms:
• Disaggregation by entrapped air compression upon moistening of aggre-
gates.
• Disaggregation by rain drop impact and/or flow turbulence.
• Micro cracking by shrinking and swelling.
• Physicochemical dispersion, resulting in the detachment of aggregates
which individually greatly vary.
The textural fining upward within one layer in the crust is typical for water
deposition: first sedimentation of the coarse particles continuously followed by
deposition of the finer particles.
Both layers depend on particle transport and sedimentation modes, which
determine two main morphological types, i.e structural crust and deposition
crust (Chen et al. 1980). Compaction due to raindrop impact may play a great
part in soil crusting. McIntyre (1958) attributed skin seal formation to this
process.
Capillary Barrier
Considering the porosity of the crust, it is condensed but especially the fine
textured, dense packed upper part of both layers could hinder water conduc-
tivity.
Current polarized light one insert two polarized filters in the microscope.
All lights waves are observed, but a crystalline mineral placed between the
polarizer’s results in birefringence.
The loss of porosity in the upper part of layer one and two and the smallest
volume of it is considered as a clear incidence of the rainfall energy. This
energy is reduced the volume of macroporisity and elongated porosity. The
loss porosity by clogging it with disintegration of aggregate soils to smaller and
finer units. As gypsum and calcite (Poch and Verplancke 1997). As a result
the crust strength will be increased and the infiltration are decreased. Also for
vughs are presented in the lower part of layer one and two in microstructure
crust have irregular shape, and some of it are spherical and it is correlated
with escaping of compressed air under the surface, and the soil ponding with
rainfall, and the fine grains these factors are limited the air from escaping
and substituted with water infiltration which limited. Microstructure of crust
formation reduced the porosity from 30 to 90% (West et al. 1990).
Conclusions
Morphological descriptions of gypsiferous crusts are a useful tool to help ex-
plain crust behavior and to provide direct evidence of processes that have been
important to the development of the crust. Pore size, shape, and amount in the
crust as compared to uncrusted soil, layers of micromass accumulation and it
consist of gypsum, calcite, silt and clay mineral. Additionally, these character-
istics when coupled with crust thickness, low porosity and crust strength may
help to explain the decline infiltration and seedling emergence in gypsiferous
crust behavior observed.
References
Bullock P et al. (eds) (1985) Handbook for thin section description. WAINE Re-
search Publications. Albrighton, Wolverhampton, UK
Chen Y, Tarchitzky J, Brouwer J, Morin J, Banin A (1980) Scanning electron mi-
croscope observations on soil crusts and their formation, Soil Sci 130:45–55
Eswaran H, Ilawi M, Osman A (1981) Mineralogy and micromorphology of arid-
sols. In: Proc 3rd Int Soil Classification Workshop, ACSAD/SS/P17, Damascus,
Syria:153–174
40 G. Mokdad et al.
FitzPatrick EA (1993) Soil microscopy and micromorphology. ohn Wiley & Sons,
New York
Gal M, Arcan L, Shainberg I, Keren R (1984) Effect of exchangeable sodium and
phosphogypsum on crust structure-scanning electron microscope observations,
Soil Sci Soc Am J 48:872–878
Gafarzadeh AA, Zinck JA (2000) Penetration resistance of gypsum crust from lab-
oratory experiments. Thesis, Wyecollege, University of London, UK
Gee GW, Bauder JW (1986) Particle size analysis. In: Klute A (ed) Methods of
soil analysis, Part 1, 2nd ed. Agron Monogr No 9, ASA and SSSA, Madison,
WI:383–409
Gooderham PT (1973) Soil physical conditions and plant growth. PhD Thesis, Uni-
versity of Reading, UK
Habib H, Robert M (1992) Morphological study of som gypsiferous soils from the
Middle Euphrates Basin (Syria)
Ilaiwi M (1983) Contribution to the knowledge of the soils of Syria. PhD Thesis,
University of Ghent, Belgium
Jafarzadeh AA (1991) Experimental studies of gypsum migration and deposition in
soil profiles. PhD Thesis, Wye College, University of London, UK
McIntyre DS (1958) Permability measurements of soil crusts formed from raindrop
impacts, Soil Sci 85:185–189
Nafie FAA (1989) The properties of highly gypsiferous soils and their significance
for land management. PhD Thesis, Wye College, University of London, UK
Nelson RE, Somers LE (1986) Total carbon, organic carbon and organic matter. In:
Page AL, Miller Rh, Keeny DR (eds) Methods of Soil Analysis, Part 2, 2nd ed.
Agron Monogr No 9, ASA and SSSA, Madison, WI:539–577
Nelson RE (1986) Carbonate and gypsum. In: Page AL, Miller Rh, Keeny DR (eds)
Methods of Soil Analysis, Part 2, 2nd ed. Agron Monogr No 9, ASA and SSSA,
Madison, WI:181–196
Poch RM, Verplancke H (1997) Penetration resistance of gypsiferous horizons, Eur
J Soil Sci 48:535
Stroosnijder L (1995) Crust formation, crust properties and crust control. Lecture 2.
In processes and models in soil and water conservation. K200–507. Department of
irrigation and soiul & water conservation, Wageningen Agricultural University,
The Netherland
Watson A (1982) The origin, nature and distribution of gypsum crusts in deserts.
PhD Thesis, University of Oxford, UK
West LT, Bradford JM, Norton LD (1990) Crust morphology and infiltrability in
surface soils from the southeast and Midwest U.S.A. In: Douglas LA (ed) Soil
micromorphology: A basic and applied science. Elsevier, Amesterdam, Develop-
ments in soil science 19:107–113
Fabric Changes in Compacted London
Clay Due to Variations in Applied Stress
and Suction
Rafael Monroy1 , Lidija Zdravkovic1 , and Andrew Ridley2

1
Imperial College London, UK
rafel.monroy@imperial.ac.uk, l.zdravkovic@imperial.ac.uk
2
Geotechnical Observations Ltd, UK andrew@geo-observations.com
Summary. In this study, the evolution of fabric in compacted London Clay has
been studied in detail. Samples were statically compacted to the same initial condi-
tions and taken along a number of stress paths – involving both wetting and loading
– to different final conditions. Wetting and loading was conducted using both con-
ventional and osmotic oedometers (the latter allowed the suction to be varied in
a controlled manner). The study of micro-fabric involved a combination of quanti-
tative measurements of the pore space, using Mercury Intrusion Porosimetry, with
qualitative observations of the overall arrangement of particles, using Environmental
Scanning Electron Microscopy. The study showed important changes to take place in
the soil micro-fabric during wetting, whereas the effect of loading was seen to be less
severe. The results suggest that changes in the overall micro-fabric of a compacted
clayey soil are strongly stress path dependent.
Key words: fabric, compaction, clay, suction, oedometer
Introduction
Fabric plays a fundamental role in defining the hydro-mechanical response of

a soil. It is now widely recognised that this parameter is even more relevant
when partly saturated soils are being considered (Alonso et al. 1987, Cui et al.
2002), particularly so in those that contain a significant amount of expansive
clay minerals. Current elasto-plastic models, such as the one proposed by
Gens and Alonso (1992), try to include explicitly the effect of fabric on the
overall soil behaviour by considering the interactions that take place between
the micro and macro structural levels. This requires an understanding of the
initial soil fabric, as well as its evolution along loading-unloading and wetting-
drying paths. While many studies have looked at the initial fabric of a partly
saturated soil, both in the intact and compacted states, less is known about the
overall effect of changes in applied stress and matrix suction on the subsequent
42 R. Monroy et al.
evolution of micro-fabric (Al-Mukhtar 1995, Cui et al. 2002, Romero et al.

2005). The aim of this paper is to present experimental evidence on some of
the changes that take place in the micro-fabric of compacted London Clay
along a number of different stress paths.
The material selected for the study consisted of brown, weathered London
Clay, extracted from a number of boreholes in north London, at a depth of
4–6 m. The liquid limit and plasticity index correspond to 83% and 54%,
respectively.
The soil was initially dried at 70◦ C, mechanically grinded, and made to
pass through a 425 μm (ASTM No. 40) aperture, before being mixed with the
correct proportion of distilled water. The powder and water were thoroughly
mixed and sealed in a plastic bag, after all the air had been removed with a
vacuum pump. This was stored in a laboratory under controlled temperature
for a minimum hydration period of three months. A comprehensive account
of the method of sample preparation can be found in Monroy (2006).
A total of eight samples were statically compacted – to the same initial
conditions, corresponding to dry of optimum in a Proctor plot – directly into
an oedometer ring. These conditions corresponded to a moisture content of
23.61%, dry density of 1.384 Mg/m3 , void ratio of 0.952, degree of saturation
of 67.1%, and matrix suction of 1000 kPa. The numbers quoted above rep-
resent average values measured in 87 samples compacted to the same initial
conditions, as explained in Monroy (2006).
The samples were taken along different stress paths, as shown in Fig. 1:
some of the samples were wetted under a nominal load of 7 kPa and allowed to
swell and to reach equilibrium under different values of matrix suction (mta–7,
mta–8, mta–9, mta–2); a further sample was wetted under confined conditions
of constant volume (mta–5); another sample was loaded at a constant moisture
content to a maximum vertical stress of 550 kPa (mta–3); and a final sample
was loaded and then wetted at a constant high stress of 220 kPa (mta–6).
Figure 1 shows the estimated location of the LC yield line (Alonso et al.
1987), as suggested in Monroy (2006).
A combination of conventional and osmotic oedometers was used to load
and wet the samples. The latter set of equipment allowed the independent
measurement and control of the matrix suction by means of the IC suction
probe (Ridley and Burland 1993) and an osmotic system (Dineen and Burland
1995).
There were three reasons for the selection of the stress paths shown in
Fig. 1:
1. To assess in detail the effect of a reduction in suction under a nominal
load on the progressive change in the soil micro-fabric.
Fabric Changes in Compacted London Clay 43
mta − 1
1000
LC
mta − 3
Suction (kPa)
470 mta − 7
150 mta − 8
40 mta − 9
0 mta − 2 mta − 5 mta − 6
7 130 220 550
Vertical stress (kPa)
Fig. 1. Stress paths followed by the compacted samples
2. To compare the effect of unconfined versus confined hydration on the final

soil fabric.
3. To determine the effect of stress path on the final fabric of samples un-
dergoing similar levels of compressive strains.
After reaching equilibrium, the samples were unloaded, removed from the
oedometer, and sealed in the way described above. They were then taken to the
Universidad Politécnica de Cataluña in Barcelona, where a detailed study of
their fabric was carried out using both Mercury Intrusion Porosimetry (MIP)
and Environmental Scanning Electron Microscopy (ESEM).
Results
Due to limitations in space, only those results derived from the MIP are
presented in this paper. The evolution of the pore space during unconfined
swelling (mta–2, mta–7, mta–8, mta–9) is presented in Fig. 2, in the form of a
pore size density function versus an equivalent pore diameter. This distribu-
tion is obtained by taking the derivative of the pore size distribution obtained
directly from the MIP, as explained in Romero et al. (2005). The fabric of
44 R. Monroy et al.
mta-1 (s=1000 kPa)

1,730 mta-7 (s= 470 kPa)
1 mta-8 (s=150 kPa)
mta-9 (s=40 kPa)
Pore Size Density Function
mta-2 (s=0 kPa)

0.8
17,440
0.6 110
0.4
0.2
0
1 10 100 1000 10000 100000 1000000
Equivalent pore diameter (nm)
Fig. 2. Evolution of fabric along a free-swelling path
the initial as-compacted sample (mta–1) was found to have two dominant
pore sizes, with peaks centred at around 110 nm and 17,440 nm. This pore
distribution would correspond to a soil composed of large aggregations of clay
particles separated by well defined macro-voids. The larger porosity would
represent the inter-aggregate porosity, whereas the smaller pore size would
refer to the intra-aggregate porosity (Alonso et al. 1987).
Figure 2 shows how suction decrements to 470 kPa (mta–7) and 150 kPa
(mta–8) did not result in significant variations in the distribution of the pore
space. It is only when the suction was reduced to 40 kPa (mta–9) and 0 kPa
(mta–2) that a significant change in the shape of the pore size density function
was observable. The final fabric of the saturated sample presents a single
dominant pore size, centred at around 1,730 nm. This fabric is characteristic
of a soil consisting of a matrix of elementary particle arrangements (consisting
of groupings of clay minerals arranged in a semi-parallel configuration), in
which there is a single intra-matrix pore space (Alonso et al. 1987).
Figure 3 presents a comparison between the pore size density functions
displayed by samples mta–2 and mta–5 (corresponding to free and confined
hydration, respectively). Both distributions have similar shape: there is a dom-
inant pore size in both cases, which in sample mta–5 is centred at around
840 nm. The decrease in the dominant pore size in sample mta–5 with respect
to mta–2 would be due to the effect of the confining stress in the former, which
was measured to be 130 kPa.
1,730
1
mta-1 (as-compacted)
0.8 mta-2 (free-swelling)

840
mta-5 (confined hydration)
0.6
0.4
0.2
0
1 10 100 1000 10000 100000 1000000
Fig. 3. Evolution of fabric during free-swelling and confined (constant volume)

hydration
The results on samples mta–3 (loaded to 550 kPa) and mta–6 (loaded to
220 kPa and saturated under a constant load) are presented in Fig. 4. The
1
mta-1 (as-compacted)
mta-3 (loaded to 550 kPa)
mta-6 (loaded to 220 kPa and wetted)
0.8 14,730
670
0.6
125
0.4
0.2
0
1 10 100 1000 10000 100000 1000000
Fig. 4. Evolution of fabric along loading and wetting paths

46 R. Monroy et al.
distribution of the pore space in sample mta–3 remained virtually unchanged,

despite the loading having continued beyond the yield stress: there is clear
indication of an inter-aggregate pore size centred at around 14,730 nm, and
an intra-aggregate porosity at around 125 nm. This result is consistent with
earlier findings Delage and Lefebvre (1984), Lapierre et al. (1990), Al-Mukhtar
et al. (1996), and suggests that the loading process affected mainly the larger
inter-aggregate porosity, leaving the aggregations practically unchanged. Wet-
ting of sample mta–6, on the other hand, resulted in the characteristic evo-
lution from an aggregate into a matrix type fabric: the pore space in sample
mta–6 can be seen to have been dominated by a single pore space centred at
around 670 nm.
Samples mta–3 and mta–6 underwent similar degrees of compressive
strains, although the mechanism behind these was different in each case.
Whereas sample mta–3 experienced collapse on loading, sample mta–6 un-
derwent collapse on wetting. This difference in the stress path determined in
an unequivocal way the final soil fabric, as Fig. 4 clearly demonstrates.
Comparison of the dominant pore sizes in samples mta–2, mta–5 and mta–6
(Figs. 3 and 4) appears to indicate that an increase in load in the fully satu-
rated state was accompanied by a reduction in the dominant entrance size of
the intra-matrix pore space (1,730, 840 and 670 nm, respectively).
Table 1 summarises total volumetric strains, εvol , from each of the above
experiments, with positive sign indicating compression.
Table 1. Volumetric strains

Test mta–7 mta–8 mta–9 mta–2 mta–5 mta–3 mta–6
εvol −1.0 −4.0 −7.0 −9.0 0.0 2.0 1.0
(%) (4.0)
Discussion and Conclusions

The research presented here was aimed at providing further evidence on the
changes that take place in the fabric of a compacted, partly saturated clayey
soil, when subjected to changes in stress and suction. In order to do so, samples
were compacted to similar initial conditions and were subsequently loaded and
wetted along different stress paths. A number of important findings have been
reported:
• The initial as-compacted fabric consisted of well defined aggregations, sep-
arated by large, inter-aggregate voids.
• Hydration to full saturation, both under confined and unconfined condi-
tions, resulted in the homogenisation of the pore space: the aggregations
swelled and eventually occupied the entirety of the inter-aggregate void

space. This is reflected in the shape of the pore size density function de-
rived from the MIP, which shows a single dominant pore size.
• The most significant changes in the soil fabric are believed to have taken
place as the suction was reduced to low values. This was at least the case
when the soil was hydrated under a nominal load.
• Loading at constant moisture content beyond the yield stress resulted in
a (minimal) reduction in the size of the inter-aggregate porosity, with no
significant visible modification in the overall distribution of the pore space.
• The evolution of fabric was found to be strongly stress path dependent: two
samples undergoing similar amounts of compression, but which followed
different paths, showed marked differences in their distribution of pore
space.
Some of the above results are in close agreement with those reported in the
literature in the past, and appear to indicate that fabric is a property that
can experience significant changes during a test, particularly so when the
suction is reduced. Fabric is known to play an important role in determining
the behaviour of partly saturated expansive soils, and it is felt that future
experimental research should play closer attention to this important topic.
Acknowledgements
The first author would like to acknowledge the financial support from the
Royal Academy of Engineering (UK), the Soil Mechanics section at Imperial
College London, and Geotechnical Observations Ltd.
References
Al-Mukhtar M (1995) Macroscopic behaviour and macro structural properties of a
kaolinite clay under controlled mechanical and hydraulic state, In: Proc 1st Int
Conf on Unsaturated soils, Paris, France 1:3–9
Al-Mukhtar M, Belanteur N, Tessier D, Vanapalli SK (1996) The fabric of a clay soil
under controlled mechanical and hydraulic stress states, Applied Clay Science
11:99–115
Alonso EE, Gens A, Hight DW (1987) Special problem soils. General Report. In:
Proc 9th Eur Conf on Soil Mechanics and Foundation Engineering, Dublin, Ire-
land 3:1087–1146
Cui YJ, Loiseau C, Delage P (2002) Micro-structure changes of a confined swelling
soil due to suction controlled hydration. In: Proc 3rd Int Conf on Unsaturated
Soils, Recife, Brazil 2:593–598
Delage P, Lefebvre G (1984) Study of the structure of a sensitive Champlain clay
and of its evolution during consolidation, Can Geotech J 21:21–35
Dineen K, Burland JB (1995) A new approach to osmotically controlled oedometer
testing. In: Proc 1st Int Conf of Unsaturated Soils, Paris, France 2:459–465
48 R. Monroy et al.
Gens A, Alonso EE (1992) A framework for the behaviour of unsaturated expansive

clays, Can Geotech J 29:1013–1032
Lapierre C, Leroueil S, Locat J (1990) Mercury intrusion and permeability of Loui-
seville clay, Can Geotech J 27:761–773
Monroy R (2006) The influence of load and suction changes on the volumetric be-
haviour of compacted London Clay. PhD Thesis, University of London (Imperial
College), London, UK
Ridley AM, Burland JB (1993) A new instrument for the measurement of soil mois-
ture suction, Geotechnique 43(2):321–324
Romero E, Hoffmann C, Castellanos E, Suriol J, Lloret A (2005) Microstructural
changes of compacted bentonite induced by hydro-mechanical actions. Advances
in understanding engineering clay barriers. In: Alonso EE, Ledesma E (eds) Proc
Int Symp on Large Scale Field Testing in Granite, Sitges, Barcelona (Spain).
A.A. Balkema Pub pp 193–202
Microstructure of a Lime Stabilised Compacted
Silt
Giacomo Russo, Sebastiana Dal Vecchio, and Giuseppe Mascolo
Facoltà di Ingegneria, Università di Cassino, Via G. Di Biasio 43, 03043 Cassino,

Italy giarusso@unicas.it, s.dalvecchio@unicas.it, mascolo@unicas.it
Summary. Lime stabilisation of fine-grained soils is a well tested technique, tra-

ditionally used to improve clay subgrades for road pavements. The lime treated
soils show a deep modification of the physical and mechanical properties (e.g., grain
size distribution, plasticity, compressibility, shear strength). A deeper insight into
the above mentioned behaviour has been achieved by investigating the changes of
microstructure taking place after the addition of lime. In the paper, results from
Mercury Intrusion Porosimetry (MIP) on a lime stabilised dynamically compacted
natural silty soil are reported. Those results highlight the time dependency of mod-
ifications induced by lime on the microstructure, which is strongly related to the
initial moulding water content of the sample.
Key words: lime stabilisation, compaction, microstructure
1 Introduction
Lime stabilisation of fine grained soils is widely recognized as a suitable ground
improvement technique, since it allows the reuse of soils normally inadequate
to be involved in earthworks (Croce and Russo 2002).
The effectiveness of the improvement is mainly due to the mineralogical
and physical properties of natural soil, but also the treatment procedures
play an important role. Therefore the lime stabilisation practice is based on
some laboratory analyses concerning relevant physical and chemical properties
of the soil, such as grain size distribution, Atterberg limits, organic matter,
sulphate content, in order to state the suitability of lime stabilisation, and
others tests in order to set up the treatment procedure (lime percent by weight,
initial water content, curing time, type and energy of compaction) in order to
gain the expected mechanical properties of the improved soil.
When lime is added to soil, two main phenomena take place: the cation
exchange and the pozzolanic reactions (Bell 1996). These two mechanisms, re-
ferred to as modification and stabilisation of treated soils respectively (Rogers
and Glendinning 1996), develop at different interval on the time scale. The
50 G. Russo et al.
cation exchange between calcium ions and clayey minerals takes place very
rapidly inducing the flocculation of the fine-grained fraction. On a macroscopic
scale, this structural modification results in a different grain size distribution
of the treated soil, characterized by a reduction of the finer fraction. As a
further consequence of the cation exchange, the plasticity index of the treated
soil is considerably reduced. The development of the pozzolanic reactions, with
the formation of stable compounds as hydrated calcium silicates and hydrated
alumina silicates, is slower than the cation exchange on the time scale, and it
is responsible for the improvement of the mechanical properties of the treated
soil (Diamond and Kinter 1965, Sherwood 1993). The considerable increase
of shear strength is strongly dependent on curing time (Bell 1996, Croce and
Russo 2003a). Otherwise the reduction of the soil compressibility measured
after the addition of lime seems to be less influenced by the curing time (Russo
and Croce 2001).
The treated soil is compacted and stabilised and both the improvement
techniques change the physical state of the natural soil and his mechanical
behaviour. Furthermore, the unsaturated state of the treated soil plays a rel-
evant role in affecting his mechanical behaviour. The time dependency of the
mechanical improvement of lime stabilised soils represents an additive com-
plexity in the analysis of the effectiveness of this technique. The time scale
over which the reactions take place is not well established, and there is not a
wide agreement about the exact sequence of the chemical reactions (Board-
man et al. 2001). From the practical point of view, it is relevant to appreciate
the longevity of lime stabilised soils with respect to the mineralogy of the soil
and to the treatment parameters.
Microstructural investigations allow a deeper insight in the complex of
phenomena taking place after the addition of lime, as already observed for soils
stabilised by cement (Ye 2003). In this perspective, in this paper some results
of an experimental analysis on remoulded natural and lime stabilised specimen
of an alluvial silty soil are reported. In order to highlight the time–dependency
of the microstructure of a lime treated soil, Mercury Intrusion Porosimetry
(MIP) on both natural and lime stabilized samples compacted at different
initial water content and cured at different times have been performed. Other
relevant parameters influencing the effectiveness of lime stabilisation (e.g. type
and percent by weight of lime, compaction energy) were fixed in order to
reduce the complexity of the data analysis.
2 Experimental Procedures
Laboratory tests were performed on remoulded samples of an alluvial silty soil.

The physical characteristics (grain size distribution, specific weight, plastic
limit, liquid limit) of the natural soil were determined and standard Proctor
tests were performed. Lime treated samples were prepared by hand mixing
the oven dried soil with 3% quicklime powder and distilled water, allowing the
Microstructure of a Lime Stabilised Compacted Silt 51
quicklime to hydrate for 24 hours. The selected percent by weight of quicklime

allowed the development of pozzolanic reactions (Rogers et al. 1997). The
samples were then compacted at three different water contents, namely dry
of optimum (wdry ), optimum (wopt ) and wet of optimum (wwet ). The initial
water contents for samples compacted dry and wet of optimum were selected
in order to obtain very similar densities.
Samples were then dehydrated by freeze-drying (Delage and Pellerin 1984),
that is rapid freezing in liquid nitrogen (boiling point −196◦ C) followed by
sublimation in a true vacuum. Freezing was accelerated using small pieces of
soil (1–2 g in weight). In such condition the 9% volume expansion of the water
does not take place, preserving the microstructure of the soil.
The dehydrated lime stabilised samples were cured for fixed time intervals
under controlled conditions before performing MIP tests. The repeatability of
the procedure was carefully verified performing several MIP test on samples
taken from the same specimen. The curing intervals of time selected were
t = 0, 7, 28 days, traditionally considered in order to evaluate the effectiveness
of stabilisation. Only for the optimum stabilised samples curing times of t = 0,
7, 14, 21, 28 days were considered. Since the duration of the freeze drying
procedure takes about 24 hours, all the time intervals are nominal.
3 Results
3.1 Physical Properties
In Table 1 the main physical characteristics of both natural and lime treated
samples are reported. The plasticity of lime stabilised sample has not been
reported since the plasticity limit wP was not determinable.
The grain size distribution of both natural and lime stabilised samples is
reported in Fig. 1a. The addition of 3% of quicklime changes the grain size
distribution, shifting and rotating the distribution towards an increase of the
medium-grained fraction of the sample. This change takes place immediately
after the addition of lime, as evidenced by the two distributions determined
by dry sieving and sedimentation respectively after the addition of lime (t =
0 days) and after a curing time of 7 days. The compaction curves of lime-
treated samples (Fig. 1b) are shifted with respect to those of natural samples,
showing an increase of the optimum water content and a reduction of the
Table 1. Physical properties of the alluvial silty soil
γs wL IP wopt γdmax
[kN/m3 ] [%] [%] [%] [kN/m3 ]
natural 26.4 25.0 8.0 14.5 18.6
3% lime 26.1 24.0 — 17.5 17.3
52 G. Russo et al.
100 20
natural
80 19
dry unit weight (kN/m)

3
percentage passing
3% lime
60 18
natural
40 17
Sr = 1
3% lime - t = 0
20 16
3% lime - t = 7
0 15
0.001 0.010 0.100 1.000 10.000 100.000 5 10 15 20 25
particle size (mm) (a) w (%) (b)
Fig. 1. Grain size distributions (a) and compaction characteristics (b) of natural
and stabilized samples
maximum dry weight. The shape of the curve remains unchanged. Stabilised
samples show an increase of the optimum water content wopt restricted to 3%
with respect to natural samples, while the reduction of maximum dry density
is about 1.0 ÷ 1.3 kN/m3 after the addition of lime. The latter results are in
good agreement with those reported in literature for other fine-grained soils
(Croce and Russo 2003b).
3.2 Pore Size Distributions
The influence of initial water content of the natural samples has been firstly in-
vestigated. As expected, the cumulative intruded volume of mercury is lower
for the optimum water content sample, characterized by the highest den-
sity (Fig. 2a). The percent porosity of the sample is about 21.4%. The other
two samples show a similar porosity after compaction. The higher frequency
of pore size of the sample compacted at optimum water content take place
around an average entry values of 1 μm (Fig. 2b). The pore size distribution
is unimodal. Both dry and wet samples are characterized by higher frequency
0.25 0.014
natural dry
dry 0.012
0.2 opt
opt
intruded volume (ml/g
0.01
dV/dlogD (ml/g
wet wet
0.15
0.008
0.1 0.006
0.004
0.05
natural 0.002
0 0
1000 100 10 1 0.1 0.01 0.001 100 10 1 0.1 0.01
pore diameter (Pm) (a) pore diameter (Pm) (b)
Fig. 2. Cumulative volume intruded (a) and pore size distribution (b) of natural
samples
0.25 0.014
natural natural
3%_t=0 0.012 3%_t=0
0.2
3%_t=7 3%_t=7
0.01
dV/dlogD (ml/g
3%_t=14 3%_t=14
0.15 3%_t=21 0.008 3%_t=21
3%_t=28 3%_t=28
0.1 0.006
0.004 wopt
0.05
wopt 0.002
0 0
1000 100 10 1 0.1 0.01 0.001 100 10 1 0.1 0.01
Fig. 3. Cumulative volume intruded (a) and pore size distribution (b) of optimum
stabilised samples
of larger pores than the optimum sample. The dry sample show a pore size
distribution slightly bimodal. The curves are qualitatively similar to those
obtained by Delage et al. (1996) on different soils.
In Figure 3a the cumulative volumes of mercury intruded for the sta-
bilised optimum samples are reported. Each curve corresponds to different
curing time of the sample. A relevant increase of the intrusion is shown by the
t = 0 days stabilised sample, corresponding to an increase of 33% of the nat-
ural sample percent porosity. This result is in agreement with the change of
the grain size distribution and the shift of the Proctor curve induced by lime.
The increase of porosity diminishes gradually for longer curing time, and the
sample cured for 28 day shows only a porosity raise of 7% with respect to nat-
ural sample. At curing time of t = 0 days, the pore size distribution (Fig. 3b)
shows a new peak not present in the natural sample curve around average
entry values of 15 μm and a slight increase of the porosity contained between
6 and 0.3 μm. This peak disappear at curing time higher than 21 days, while
the increase of porosity around 1 μm becomes stable and the curve unimodal
again. The intra-aggregate porosity (average entrance entry values smaller
than 0.3 μm) seems to be not affected by the addition of lime.
The behaviour shown by the wet stabilised samples is similar to the op-
timum stabilised samples. The percent porosity at curing time of t = 0 is
16% higher than the natural wet sample porosity, and it reduces for longer
curing time (Fig. 4a). The pore size distribution becomes bimodal after the
addition of lime (t = 0), but the new peak detected around 10 μm disappears
after a curing time of 7 days, and the curve becomes unimodal again (Fig. 4b).
The samples show a distribution characterised by higher frequency of pores
contained between 3 μm and 0.3 μm.
Slight differences in cumulative volume of mercury intruded are shown by
dry stabilised samples cured for different times (Fig. 5a). Negligible variations
of percent porosity are detected in function of the curing time. Dry stabilised
samples show a bimodal distribution also after the addition of lime with a
reduction in the average entry values from 10 to 3 μm (Fig. 5b). Samples
54 G. Russo et al.
0.25 0.014
natural natural
0.012
0.2 3%_t=0 3%_t=0
0.01
dV/dlogD (ml/g
3%_t=7 3%_t=7
0.15 0.008
3%_t=28 3%_t=28
0.1 0.006
wwet
0.004
0.05
wwet 0.002
0 0
1000 100 10 1 0.1 0.01 0.001 100 10 1 0.1 0.01
Fig. 4. Cumulative volume intruded (a) and pore size distribution (b) of wet sta-
bilised samples
0.25 0.014
natural natural
0.012
0.2 3%_t=0 3%_t=0
0.01
3%_t=7
dV/dlogD (ml/g
3%_t=7
0.15
3%_t=28 0.008 3%_t=28
0.1 0.006 wdry

0.004
0.05
wdry 0.002
0 0
1000 100 10 1 0.1 0.01 0.001 100 10 1 0.1 0.01
Fig. 5. Cumulative volume intruded (a) and pore size distribution (b) of dry sta-
bilised samples
cured 7 days and more are characterised by a unimodal distribution, with a

marked increase of diameters contained between 3 and 0.3 μm.
Variations in porosity induced by lime addition depend on the initial water
contents of samples. In order to highlight the time-dependency of this vari-
ations, in Fig. 6 the curves representing cumulative volumes of mercury in-
truded in stabilised samples at different initial water content and equal curing
time (t = 0, 28) are reported. It can be seen that lime affects the microstruc-
ture of the soil immediately after his addition. The more relevant effects are
mainly detected in stabilised optimum samples, that show an increase of poros-
ity (Fig. 6a) and, differently from the natural samples, a bimodal distribution
of pore sizes (Fig. 6b). The less effective changes, at the same curing time,
pertain to dry stabilised samples. Representing the intruded volume of all the
three samples (namely dry, optimum and wet) cured for 28 days, a very slight
difference in percent porosity (Fig. 6c) and a quite similar distribution of pore
sizes (Fig. 6d) can be noted.
0.25 0.014
dry
dry 0.012
0.2
opt
opt 0.01
dV/dlogD (ml/g
wet
0.15 wet 0.008
3% lime
0.1 0.006
t=0
0.004
0.05 3% lime
t=0 0.002
0
0
1000 100 10 1 0.1 0.01 0.001
100 10 1 0.1 0.01
0.25 0.014
dry
dry 0.012
0.2 opt
opt
0.01
dlogV/dD (ml/g
wet
0.15 wet
0.008
3% lime
0.1 0.006
t = 28
0.004
0.05 3% lime
t = 28 0.002
0 0
1000 100 10 1 0.1 0.01 0.001 100 10 1 0.1 0.01
pore diameter (Pm) (c) pore diameter (Pm) (d)
Fig. 6. Cumulative volume intruded (a) and pore size distribution (b) of stabilised
samples cured for t = 0 and cumulative volume intruded (c) and pore size distribu-
tion (d) of stabilised samples cured for t = 28
4 Conclusions
A study of the microstructure induced by compaction and lime stabilisation

on an alluvial silt was carried out by means of mercury intrusion porosimetry
on freeze-dried samples.
The moulding water content plays a relevant role on the microstructure
of compacted natural samples. The lesser porosity pertains to samples com-
pacted at optimum water content, in agreement with results reported in lit-
erature (Delage et al. 1996) on different soils.
The time-dependency of the stabilised samples microstructure was clearly
detected by means of MIP results. Addition of lime mainly affects the inter-
aggregate porosity. A relevant modification of porosity for samples stabilised
at 3% by weight and compacted at optimum water content of quicklime takes
place immediately after the addition of lime. In particular, an increase of
pores of relatively large diameter is detected. A subsequent reduction of this
effect occurs for longer curing time of the stabilised samples, bringing back the
pore size distribution towards the distribution of the natural samples. Similar
behaviour is shown by the wet stabilised samples, while the microstructure
of dry samples is less affected by the addition of lime. After a curing time
of 28 days the stabilised samples show a very similar porosity and pore size
distribution, irrespective of the moulding water content.
56 G. Russo et al.
Cation exchange induced by lime takes place immediately after the addi-
tion of lime, producing the flocculation of grains and increasing the porosity
of the stabilised sample. This effect is relevant for optimum and wet moulding
water contents but decays for longer curing time, probably due to changes
of the pH of pore water (Eades and Grim 1966). The similarity of pore size
distribution of samples compacted at different water content and cured for
28 days may suggest that the development of pozzolanic reactions and the
consequent cementation of the aggregates contributes to reduce porosity in-
duced by flocculation and increases the frequency of pores of small diameters.
In the long term, this effect may be more valuable on the microstructure than
the moulding water content.
References
Bell FG (1996) Lime stabilization of clay minerals and soils, Engineering Geology
42:223–237
Boardman DI, Glendinning S, Rogers CDF (2001) Development of stabilisation and
solidification in lime-clay mixes, Géotechnique 50:533–543
Croce P, Russo G (2002) Reimpiego dei terreni di scavo mediante stabilizzazione
a calce. In: XXI Convegno Nazionale di Geotecnica, L’Aquila, Patron Editore,
pp 11–216
Croce P, Russo G (2003a) Experimental investigation on lime stabilized soils. In:
Proc XIII European Conference of Soil Mechanics and Geotechnical Engineering,
Prague
Croce P, Russo G (2003b) Soil-water characteristic curves of lime stabilised soil. In:
Vermeer, Schweiger, Karstunen, Cudny (eds) Proc Int Workshop on Geotechnics
of Soft Soils – Theory and Practice. Noordwijkerhout
Delage P, Audiguier M, Cui YJ, Howat MD (1996) Microstructure of a compacted
silt, Canadian Geotechnical Journal 33:150–158
Delage P, Pellerin FM (1984) Influence de la lyophilisation sur la structure d’une
argile sensible du Québec, Clay Minerals 19:151–160
Diamond S, Kinter EB (1965) Mechanisms of soil-lime stabilization: an interpretative
review, Highway Research Record 92:83-96
Eades JL, Grim RE (1966) A quick test to determine lime requirements of lime
stabilization, Highway Research Record 139:61–72
Rogers CDF, Glendinning S (1996) Modification of clay soils using lime. In: Rogers
CDF, Glendinning S, Dixon N (eds) Lime Stabilisation, Thomas Telford, Lon-
don, pp 99–114
Rogers CDF, Glendinning S, Roff TEJ (1997) Modification of clay soils for construc-
tion expediency, Geotechnical Engineering 125:1–8
Russo G, Croce P (2001) Indagini sperimentali sui terreni stabilizzati a calce. In:
Incontro Annuale dei Ricercatori di Geotecnica – IARG 2001, Milano
Sherwood PT (1993) Soil stabilization with Cement and Lime: State-of-Art Review.
HMSO, London
Ye G (2003) Experimental study and numerical simulation of the development of the
microstructure and permeability of cementitious materials. Ph.D. thesis, Delft
University
Part II
Measuring Suction
Errors in Total Suction Measurements
Setianto Samingan Agus1 and Tom Schanz2

1
Civil and Environmental Engineering Consultants, Singapore
agus72indo@yahoo.com
2
Laboratory of Soil Mechanics, Bauhaus-University Weimar, Germany
tom.schanz@bauing.uni-weimar.de
Summary. Suction as one of the stress state variables of unsaturated soils is usu-
ally either imposed to the soil or measured. Measurements of suction therefore be-
come commonly practiced when dealing with unsaturated soil problems. Unlike the
measurement of matric suction, whereby one can arrive at a sufficient accuracy, the
accuracy in total suction measurements is questionable. Total suction measurements
involve equilibration of sensors used with the relative humidity of the vapour space
in or above the soils. Many factors have been cited to cause inaccuracy in the to-
tal suction measurements such as temperature gradient and the inaccuracy of the
sensors used to measure total suction. In this paper, the total suction measurement
inaccuracy of several sensors commonly used is discussed. It is found that the total
suction measurements can only be performed to lowest suction of 1000 kPa when
sufficient accuracy is to be obtained.
Key words: total suction, measurement error, accuracy, suction characteristic

curve, temperature fluctuation, bentonite–sand mixture
Introduction
Suction is one of the stress state variables of unsaturated soils and is defined
as the ratio of the partial energy of water vapor in the soil voids to that of
free pure water (Aitchison 1965). Suction or total suction consists of capillary
component or matric suction and osmotic component. In clays, capillary force
can present with sorptive forces due to the presence of ions to form matric
component of suction while osmotic component is associated with salt con-
centration of the free pore fluids. At completely dry conditions (i.e., oven-dry
conditions), suction in the soil is 1,000,000 kPa (Croney and Coleman 1961
and Fredlund and Rahardjo 1993).
The relationship between suction and water content (either gravimetric
or volumetric basis) in the soil following drying or wetting process is termed
as soil-water characteristic curve (Fredlund and Rahardjo 1993). Soil-water
characteristics curve can be used as a tool to derive other unsaturated soil
60 Setianto Samingan Agus and Tom Schanz
properties such as shear strength, compressibility, and coefficient of perme-

ability provided the soil properties at saturated state are known.
Another relationship between suction and water content can be established
when soil is mixed with water to reach several targeted water contents and
the measurement of suction is performed for the soil at those water contents.
Such a relationship is called suction characteristic curve. The two terms – the
soil water characteristic curve and suction characteristic curve should not be
mixed up. The soil-water characteristic curve is usually determined for the soil
starting from saturated conditions whereas this condition is not necessary in
the determination of the suction characteristic curve. In the case of clay soils,
the effects of initial fabric are more pronounced in the soil-water characteristic
curve but it is not clearly seen in the suction characteristic curve (Agus 2005).
Matric and total suctions are commonly measured while osmotic suction
is normally inferred from the two values in spite of the availability of tech-
niques to measure it. Many suction measurement methods are available and
the methods can be broadly grouped into two main categories – direct method
and indirect method.
The direct method of suction measurement refers to direct measurement of
negative pore-water pressure in the soil, e.g., using tensiometer and null-type
axis-translation techniques. Since the techniques essentially need the sensor
used to be in intimate contact with the soil, matric suction component is
measured. As long as the good contact between the soil and the sensor used
is warranted, one can arrive at sufficient accuracy using this technique as far
as the engineering accuracy is concerned. However, it is nearly impossible
to maintain good contact between the soil and the sensor at high suctions.
Moreover, cavitation poses another drawback to the technique. Thus, the di-
rect method is not applicable for the measurement of high suctions.
Indirect methods can be considered as alternatives. The indirect meth-
ods of suction measurement are applicable for both matric and total suction
measurements. For total suction measurement, only the indirect technique is
applicable. It is called the indirect method for measuring total suction be-
cause it requires the measurement of other parameters such as water content
or relative humidity to be performed for inferring the measured total suction.
Generally, it is more difficult to measure total suction than matric suction
as far as the accuracy is concerned, since the error involved is generally greater.
Although the accuracy of most sensors used in the total suction measurement
in terms of the measured parameters (i.e., water content, relative humidity,
etc.) is quantifiable, the error likely involved in the measurement in terms
of total suction value has not been mentioned in most literature on total
suction measurements. Arifin et al. (2006) discussed errors in the measurement
of total suction of bentonite–sand mixtures using different types of sensor.
However, the extent to which the errors affect the measurement results was
not comprehensively discussed in the paper.
This paper focuses on the indirect measurement of total suction using com-
monly used sensors. The main discussion brought out in this paper is the error
Errors in Total Suction Measurements 61
associated with several total suction measurement sensors; namely, filter pa-
per, psychrometer, capacitance sensor, warmed-head capacitance sensor, and
chilled-mirror hygrometer. An interesting discussion on the extent to which
the measurement error affects the results is presented at the end of the paper.
Total Suction Measurement Techniques

Various methods have been used to indirectly measure total suction of soils.
Filter paper technique is among the available methods whereby filter paper
water content is used to infer total suction of the soil. Other sensors like psy-
chrometer and capacitance sensor are among the famous methods for measur-
ing suction together with chilled-mirror hygrometer which has recently been
introduced in geotechnical engineering. The sensors measure relative humid-
ity of the vapour in the soil pores. The relative humidity is converted to total
suction using Kelvin’s equation (Fredlund and Rahardjo 1993):
−RT
st = ln(RH) (1)
Mw (1/ρw )
where R is the universal gas constant (i.e., 8.31432 J/mol/K), T is the absolute
measured temperature in degrees K, Mw is the molecular weight of water
(i.e., 18.016 kg/kmol), ρw is the unit weight of water in kg/m3 as a function
of temperature, and RH is the measured relative humidity defined as uv /uvo
where uv is the partial pressure of pore-water vapour in the specimen and uvo
is the saturation pressure of water vapor over a flat surface of water at the
same temperature.
When a filter paper is placed without contact with a soil specimen in a
closed container, the vapour space above the soil specimen acts as a true
semi-permeable membrane which is only permeable to water vapour but not
to ions from the pore-water. Therefore, in the (non-contact) filter paper tech-
nique, total suction is measured. The calibration curve for the filter paper
technique for total suction measurement is normally established using vapour
equilibrium technique (e.g., Chandler and Gutierrez 1986 and Leong et al.
2002). Ridley and Wray (1996) found that the filter paper technique is insen-
sitive when used for measuring low total suctions due to possible vapour and
temperature non-equilibrium during measurement.
The psychrometer technique has been widely used by many researchers
for measuring total suction (see Ridley and Wray (1996) for literature re-
view in this technique). There are many positive aspects, but few drawbacks
associated with the accuracy of psychrometers for measuring total suction.
The psychrometer technique can be used to measure total suction as high as
8500 kPa. The measurement of low total suctions (i.e., lower than 1000 kPa)
using this technique requires an accurate means to maintain temperature at
a constant value. Ridley and Wray (1996) noted that inaccuracies of the psy-
chrometer technique may be caused by the insensitivity of psychrometer due
to temperature effects. On the other hand, variability in the electromotive

force generated induces error in high suction range. Furthermore, there ex-
ists a possibility of deterioration in the sensitivity of psychrometer due to
corrosion problem (Hamilton et al. 1981 and Zerhouni 1995). Unlike the filter
paper method, no reports of hysteresis of the psychrometer technique for total
suction measurement have been found in literature.
The capacitance sensor has been used for measuring high total suction
by Albrecht et al. (2003). The capacitance of the sensor changes as the thin
polymer film within absorbs water from its surrounding. Polymer capacitance
technology is recently used in this type of sensors as the technology provides
high reliability to the relative humidity measurement, insignificant hysteresis,
and insensitivity to temperature fluctuation where measurement is conducted
(Wiederhold 1997). Arifin et al. (2006) used similar sensor and found that the
sensor also works at high temperature.
The warmed-head capacitance sensor generally has better performance
than the capacitance sensor without warming function. When measurement
is performed in an environment where relative humidity and temperature
change very rapidly, condensation occurs resulting in overestimation of rela-
tive humidity or underestimation of total suction. In this case, the warming
function in the warmed-head capacitance sensor plays its role in minimizing
the effects of condensation.
The chilled-mirror hygrometer technique has been used to measure suction
by Leong et al. (2003), Albrecht et al. (2003) and Schanz et al. (2004). The
device measures dew point and temperature of the head space or essentially
the relative humidity of the head space above the specimen. The description of
the chilled-mirror hygrometer can be found in Leong et al. (2003). The equip-
ment has a small chamber where the soil specimen is placed. The temperature
of the soil specimen, the vapour space, and the measuring device is controlled
and monitored very accurately using an infra-red thermometer. Considering
the way the control during measurement is performed, the technique can be
considered to be the most accurate method for measuring total suction to
date (Leong et al. 2003 and Agus 2005).
Total Suction Measurement Errors
This section discusses errors in the total suction measurement using the filter
paper technique (for Whatman 42 filter paper), the psychrometer (thermocou-
ple type PCT-55 with HR-33T voltmeter from Wescor Inc.), the capacitance
sensor (PSF102 series from BC component), the warmed-head capacitance
sensor (Humicap HMP 243 from Vaisala Oyj), and the chilled-mirror hy-
grometer (type 3TE from Decagon Devises Inc.). The first two devices have
long been used for measuring suction whereas the other three devices were
introduced in geotechnical engineering as total suction measuring techniques
not longer than 15 years ago.
Since the measurement of low total suction is more difficult than the mea-
surement of high total suction, the error is often related to how low total
suction can be measured by the sensors. Errors in the total suction measure-
ments can generally be divided into four categories:
1. Errors which are induced by inaccuracy of the equipment (or sensor) used.
This type of errors exists in all sensors.
2. Errors due to isothermal non-equilibrium between the vapour space, the
soil, and the sensor used (or temperature gradient). This type of error
can be found in the filter paper (when the size of vapour space is not
minimized) and in the total suction measurement using the psychrometer,
the capacitance sensor, and the warmed-head capacitance sensor.
3. Errors due to excessive condensation of water vapour on the measuring
device as mentioned as a result of rapid temperature and relative humidity
fluctuations. The excessive condensation causes a drop in the temperature
of water vapour in the vapour space and therefore this type of error can
also be regarded as an isothermal non-equilibrium in the measuring sys-
tem. This type of error can be found in the filter paper, psychrometer
technique and in the capacitance sensor.
4. Errors due to handling of the sensor and procedures adopted by operator.
This type of error is most pronounced in the filter paper technique.
Errors due to inaccuracy of the devices and temperature gradient can be
quantified and are discussed further in the following sections while errors due
to excessive water vapour condensation on the measuring devices and the
operator-related errors cannot be simply quantified.
Errors Due to the Inaccuracy of the Measuring Devices
The error of total suction measurement due to the inaccuracy of the measur-
ing devices which measures relative humidity of the vapour space in soil is
calculated from ratio of the first derivative of Equation (1) with respect to
relative humidity to the measured total suction:
100ΔRH
Error(%) = % (2)
RH ln(RH)
For the filter paper technique, the inaccuracy of the device used is in
essence the accuracy of the weighing balance used to measure the weight of wet
and dry filter papers to obtain the water content of the filter paper. Consider-
ing the ASTM calibration curve for Whatman 42 filter papers (ASTM 1997)
and 0.0002 g accuracy of the weighing balance (i.e., the one with 0.0001 g least
count), an accuracy of 10.5% and 2% of measured total suction is obtained
for the standard-sized 55-mm diameter filter paper (i.e., with an average dry
mass of 0.065 g) water content ranging from 0% to 45.3% and 45.3% to 90%,
respectively. Since water content of the filter paper is related to suction of
the soil, the accuracy of the filter paper technique in terms of total suction is
a function of the soil-water characteristic curve or the suction characteristic
curve of the soil. Assuming that the 45.3% filter paper water content corre-
sponds to a total suction of 100 kPa, the error of 10.5% and 2% total suction
is expected for the measured total suction ranging from 1,000,000 to 100 kPa
and from 100 to 0 kPa, respectively.
For Wescor thermocouple psychrometer type PCT-55 with HR-33T volt-
meter, the accuracy is equal to 2% reading (in microvolt) (Wescor 1986). Since
total suction is linearly related to the psychrometer reading (with a factor of
0.75 μV/bar), the error in the measured total suction is also equal to 2%.
However, the device exhibits a maximum noise of 0.1 μV. The error due to
the noise is greater than that due to the reading inaccuracy and one can work
out that this error can be quantified using the following expression with st in
kPa.
1333.33
Error(%) = (3)
st
Based on Equation (3), it is clear that the percentage of error reduces as
measured total suction increases.
For the capacitance sensor PSF102 series, the accuracy is a function of
measured sensor capacitance and the value is equal to 0.4 pF/% measured
RH (BC Components 1999). Since the capacitance of the sensor is also a
function of the measured RH, the accuracy depends on the calibration curve
obtained for the sensor. Arifin et al. (2006) noted that several capacitance
sensors of PSF 102 series exhibit slight variability in the calibration curve for
each individual sensor. The typical calibration curve for the sensors used by
Arifin et al. (2006) follows the equation below which is different from that of
the sensor of the same type obtained by Albrecht et al. (2003).
RH(%) = −2442.12 + 19804Cp − 49542Cp2 + 38978Cp3 (4)
where Cp is the capacitance of the sensor in nF.

The error due to the inaccuracy of the capacitance sensor can be deter-
mined by taking the first derivative of Equation (4) with respect to Cp and
substituting it to Equation (2).
The warmed-head capacitance sensor Humicap HMP 243 essentially also
measures the capacitance of the sensor. However, since the sensor normally
comes with a digital display which can show directly the measured RH value,
only the accuracy in terms of measured RH is of concern. Like the capacitance
sensor PSF 102 from BC Components, the accuracy of the warmed-head ca-
pacitance sensor is also a function of the measured RH, which follows the
equation below (Vaisala 2002).
RH Error(%) = (0.5 + 0.025RH)% (5)
Equation (5) can be directly substituted to Equation (2) to obtain the

error in the measured total suction for this sensor.
The chilled-mirror hygrometer type 3TE is a very accurate device for mea-
suring total suction. Leong et al. (2003) found that the device can measure
total suction of as low as 150 kPa. However, the value was determined by con-
sidering the readability of the device which is 0.1% RH instead of its accuracy.
The actual accuracy of the device is equal to 0.3% RH as described by the
manufacturer although Agus (2005) found that the device has as low as 0.2%
RH accuracy when used to measure relative humidity above standard salt so-
lutions. The 0.3% RH accuracy can be substituted to Equation (2) to obtain
the accuracy of the device in terms of measured suction.
Figure 1 shows a graphical representation of total suction measurement
error for different sensors assessed in this paper. It is seen that by considering
30% total suction error due to the inaccuracy of the sensors only, the lowest
total suction that can be measured using the psychrometer is about 50 kPa,
the capacitance sensor is 150,000 kPa, the warmed-head capacitance sensor is
15,000 kPa, and the chilled-mirror hygrometer is 1500 kPa. The error of the
total suction measurement using filter paper is much lower than that of the
other four sensors.
10000
Filter paper
Psychrometer
Capacitance sensor
Warmed head cap. sensor
Chilled-mirror hygrometer
1000 30% error
100% error
100
Error (%)
10
0.1
10 100 1000 10000 100000 1000000
Measured suction (kPa)
Fig. 1. Error in the total suction measurement due to the inaccuracy of the sensor
used only
Errors Due to Temperature Gradient
The total suction measurement errors due to temperature gradient depend on

how an isothermal equilibrium between the specimen, the vapour space, and
the sensor used in the measurement can be maintained. In a closed system, the
amount of water vapor is constant and thus the partial pressure of water vapor
(uv ) is also constant. A difference in temperature (or a temperature gradient)
between the specimen and the vapour space causes movement of water result-
ing in a change in the saturated water vapour (uvo ) while the change in uv is
zero. A change in the total suction (or the error in total suction measurement)
due to temperature gradient can be estimated by rearranging Equation (1)
and taking its first derivative with respect to temperature (Croney et al. 1952
and Agus and Schanz, in press).
d(ln st ) 1 1 d(uvo )
= − (6)
dT T uvo ln(RH) dT
The change in total suction or the error in total suction measurement

due to different values of temperature gradient is shown in Fig. 2 (Agus and
Schanz, in press). If the error is limited, for instance, to 30% of the measured
value, the measurement of total suction will be limited to a value of about
850 kPa assuming that a 0.5◦ C temperature gradient exists in the system.
10000
1000
10% error
30% error
100% error
Error (kPa)
500% error
100 Temperature gradient

(or fluctuations):
2 deg C
1 deg C
0.5 deg C
0.2 deg C
10
0.1 deg C
0.02 deg C
1
10 100 1000 10000 100000 1000000
Measured suction (kPa)

Fig. 2. Error in the total suction measurement due to temperature gradient only
(Agus and Schanz, in press)
The temperature gradient of 0.5◦ C can likely be maintained in a temperature-

controlled laboratory. Based on Equation (6), the accurate measurement of
total suction to as low as 100 kPa with 30% error requires a means to control
temperature within ±0.04◦ C.
Effects of Total Suction Measurement Errors
To illustrate the effects of total suction measurement errors, a suction charac-

teristic curve plotted as bentonite water content versus total suction relation-
ship that has been obtained for loose bentonite–sand mixtures is considered.
The properties of the materials can be found in Agus (2005). The suction char-
acteristic curve presented was measured using the chilled-mirror hygrometer
type 3TE. The procedures adopted were described in Agus (2005). The data
obtained have been best-fitted using a simple power function for simplicity
(Fig. 3). The possible envelopes where the measured total suction data may
be obtained using each sensor discussed above have been computed based on
the error due to sensor accuracy and that due to temperature gradient and
also drawn in Fig. 3. For the filter paper, psychrometer, capacitance sensor,
and warmed-head capacitance sensor, a temperature gradient of 0.5◦ C has
been considered. The error due to temperature gradient does not exist in the
chilled-mirror hygrometer. It must be noted that the errors due to condensa-
tion problems and operator-related factors are not considered. These errors,
although cannot be simply quantified cannot be ignored and may govern the
overall accuracy of the total suction measurement.
It can be seen from Fig. 3(a) that for the filter paper technique, the
measurement of total suction is accurate to a lowest total suction of about
1000 kPa. More erratic measurement data can be expected at lower range of
total suction as the temperature gradient effects become more pronounced.
Similar trend is seen for the psychrometer (Fig. 3(b)).
Generally, the capacitance sensor has poor performance almost in the en-
tire range of total suction (Fig. 3(c)). A slightly better performance may be
expected when measuring total suction of greater than 200,000 kPa using the
capacitance sensor (Fig. 3(c)) and greater than 20,000 kPa using the warmed-
head capacitance sensor (Fig. 3(d)) considering errors due to both the sensor
accuracy and temperature gradient.
The chilled-mirror hygrometer has the best performance. The sensor can
be used to measure total suction as low as 1000 kPa with sufficient accuracy
(Fig. 3(e)). It is depicted in the figure that some of the data are located outside
the envelopes for the chilled-mirror hygrometer technique. The reason for this
may be possible errors in the measurement of water content of the specimens.
Since the chilled-mirror hygrometer is the most accurate means of measuring
total suction, the sensor can be regarded as a benchmark for assessing the
accuracy of other sensors for measuring total suction greater than 1000 kPa.
Fig. 3. Suction characteristic curve of loose bentonite–sand mixtures and the enve-
lope of possible measured data for: (a) filter paper; (b) psychrometer; (c) capacitance
sensor; (d) warmed-head capacitance sensor; and (e) chilled-mirror hygrometer
Considering the possible significant errors involved in the total suction

measurement, it is recommended to measure matric suction instead of total
suction, for suction less than 1000 kPa. When significant amount of contam-
inant and/or salts are expected to exist in the soil pore-water, there will be
significant difference between matric and total suction. Therefore, for suc-
tion less than 1000 kPa, measurement of osmotic suction is recommended and
total suction can be computed by adding up the measured matric and osmotic
suctions.
Conclusions
A total suction measurement error analysis has been presented and the fol-
lowing conclusions can be drawn:
1. Measurement of low total suction measurement involves great errors. Two
quantifiable errors are noted; namely, errors due to the measuring sensor
accuracy and errors due to temperature gradient.
2. The chilled-mirror hygrometer possesses the best accuracy and, therefore,
can be regarded as a benchmark for assessing other sensors.
3. Total suction measurement can only be performed in the range of suction
greater than 1000 kPa, if sufficient accuracy is to be obtained.
4. For suction less than 1000 kPa, matric and osmotic suction measurements
are recommended to be performed and total suction is computed from the
two values.
References
Agus SS (2005) An experimental study on hydro-mechanical characteristics of
compacted bentonite–sand mixtures. PhD Thesis, Bauhaus-University Weimar,
Germany
Agus SS, Schanz T (In press) Discussion of paper “Free energy of water-suction-in
filter papers” by R. Bulut and W.K. Wray, Geotechnical Testing Journal, 29(6)
Aitchison GD (1965) Moisture equilibria and moisture changes in soils beneath cov-
ered areas. In: Aitchison GD (ed) A symposium in Print, Australia, Butterworth
Albrecht BA, Benson CH, Beuermann S (2003) Polymer capacitance sensors for
measuring gas humidity in drier soils, Geotechnical Testing Journal, 26(1):3–11
Arifin YF, Agus SS, Schanz T (2006) Temperature Effects on Suction Characteristic
Curve of Bentonite-Sand Mixtures. In: Miller GA, Zapata CE, Houston SL,
Fredlund DG (eds) Proc 4th Int Conf on Unsaturated Soils, 2006, Carefree,
Arizona, USA:1314–1325
ASTM (1997) Annual Book of Standards. Volumes 04.08 and 04.09, Soil and rock,
ASTM International, West Conshohocken, PA
BC Components (1999) Humidity Sensor Brochure
Croney D, Coleman JD, Bridge P (1952) The suction of moisture held in soil and
other porous materials, Road Research Technical Paper, 24
Croney D, Coleman JD (1961) Pore pressure and suction in soil. In: Proceedings of
Conference on Pore Pressure and Suction in Soils, London, Butterworths:31–37
Chandler RJ, Gutierrez CI (1986) The filter paper method of suction measurement,
Geotechnique 36:265–268
Fredlund DG, Rahardjo H (1993) Soil mechanics for unsaturated soils. John Willey
& Sons, Canada
Hamilton JM, Daniel DE, Olson RE (1981) Measurement of hydraulic conductivity

of partially saturated soils. In: Zimmie TF, Riggs CO (eds) Permeability and
Groundwater Contaminant Transport, ASTM Special Technical Publication 746,
ASTM:182–196
Leong EC, He L, Rahardjo H (2002) Factors affecting the filter paper method for
total and matric suction measurements, Geotechnical Testing Journal, 25(3)321–
332
Leong EC, Tripathy S, Rahardjo H (2003) Total suction measurement of unsaturated
soils with a device using the chilled-mirror hygrometer technique, Geotechnique
53(2)173–182
Ridley AM, Wray WK (1996) Suction measurement: a review of current theory and
practices. In: Alonso EE, Delage P (eds) Proc 1st Int Conf on Unsaturated Soils
(UNSAT 95), Paris, 1995, Balkema, Rotterdam:1293–1322
Schanz T, Agus SS, Tscheschlok G (2004) Determination of hydro-mechanical prop-
erties of Trisoplast . Research Report Bo-015/03. Laboratory of Soil Mechanics,
Bauhaus-University Weimar, Weimar, Germany
Vaisala (2002) HMP240 series transmitters. User’s guide. Vaisala Oyj., Helsinki,
Finland
Wescor Inc. (1986) HR 33T Dew point microvoltmeter: Instruction/service manual.
Wescor Inc., Utah
Wiederhold PR (1997) Water vapour measurement—methods and instrumentation.
Marcel Dekker, NY
Zerhouni (1995) Triaxial testing using psychrometers (In French). In: Alonso EE,
Delage P (eds) Proc 1st Int Conf on Unsaturated Soils (UNSAT 95), Paris, 1995,
Balkema, Rotterdam:673–678
Application of a Dew Point Method to Obtain
the Soil Water Characteristic
Gaylon S. Campbell, David M. Smith, and Brody L. Teare
Decagon Devices, Inc., 950 NE Nelson Ct., Pullman WA. USA

gaylon@decagon.com, daves@decagon.com, brody@decagon.com
Summary. The dew point method, which determines suction by measuring sample
temperature and dew point temperature of air equilibrated with the sample, provides
a quick and accurate method for measuring suction in the range 1 to 400 MPa (pF 4
to pF 6.6). We present a procedure, using the dew point method, to prepare samples
in the relevant suction range, measure suction and water content, and obtain an
equation describing the soil water characteristic (SWC). Characteristics for four soil
samples ranging in texture from sand to bentonite are presented. All show a linear
relationship when water content is plotted vs. logarithm of suction. The slope of the
characteristic is easily obtained for use in expansive soil identification.
Key words: soil water characteristic, chilled mirror dew point, expansive soil iden-
tification
Introduction
The soil water characteristic (SWC) relates total suction to water content
for a particular soil, and is of fundamental importance to unsaturated soil
mechanics. In spite of its importance, few methods exist for quickly and accu-
rately determining the SWC. Traditionally axis translation methods (pressure
plates) have been used to equilibrate samples at preset matric suctions, and
then the water content of the equilibrated samples is determined by oven dry-
ing. The assumption is made that samples equilibrate in the axis translation
apparatus at the preset pressures. Both models and measurement confirm
that this is a bad assumption at suctions drier than 200 to 300 kPa (Campbell
1988, Gee et al. 2002). Beyond these suctions the hydraulic conductivity of
both the sample and sample-plate contact become so low that even weeks are
not sufficient for equilibrium. Outflow decreases rapidly with time because of
the decreasing hydraulic conductivity giving an appearance of equilibrium,
but in fact the sample stays far from it.
Tensiometers provide another popular method to measure matric suction.
Most are limited to the range 0 to 80 kPa (Lu and Likos 2004), though special
72 G. S. Campbell et al.
designs and pretreatments allow some units to reach 1500 kPa. Such units are
not widely available. A tensiometer can be used, along with oven drying, to
determine the wet end of the SWC. A different method is required for the dry
end of the curve.
Vapor pressure methods are ideal for measuring suction of samples drier
than the axis translation or tensiometer range. Suction is determined by equi-
librating a sample with a head space, measuring the head space vapor pressure
and the sample saturation vapor pressure, and then using the Kelvin equation
to compute total suction:

R(T + 273.16) p
ψt = ln . (1)
Vw po
Here R is the gas constant (8.3143 J/mol), T is Celsius temperature, Vw
is the molar volume of water (1.8 × 10−5 m3 mol−1 ), p is the equilibrium
vapor pressure in the head space and po is the saturation vapor pressure at
sample temperature. The units of ψt are J m−3 or N m−2 or Pa. The head
space vapor pressure can be measured by determining wet bulb depression,
as with the thermocouple psychrometer (Rawlins and Campbell 1986) or the
dew point temperature as with the dew point potentiometer (Gee et al. 1992).
Because the vapor gap between the sample and the sensor is a perfect semi
permeable membrane, vapor pressure methods measure total suction. Axis
translation methods and tensiometers measure matric suction. The difference
between these is the osmotic suction. To determine matric suction from a
vapor pressure measurement one needs to have an independent measurement
of osmotic suction. Methods for making this measurement are discussed else-
where (Rawlins and Campbell 1986, Leong et al. 2003).
The Dew Point Method

The Dew Point Potentiameter (Model WP4, Decagon Devices, Inc., Pullman,
WA 99163 USA) determines total suction by measuring the dew point tem-
perature of the head space above a sample. This is done by cooling a mir-
ror whose reflectance is carefully monitored by an optical sensor. The mirror
temperature is controlled to maintain a constant thickness of dew. The mirror
temperature is measured with a thermocouple. Sample temperature is moni-
tored by infrared thermometry. The vapor pressure of the air is the saturation
vapor pressure at dew point temperature. Saturation vapor pressure is only a
function of temperature:

17.502T
po (T ) = 0.611 exp (2)
240.97 + T
Substituting Td in Eq. (2) to get vapor pressure (saturation vapor pressure
at dew point temperature) and Ts to get saturation vapor pressure at sample
temperature, and substituting these into Eq. (1) gives
Application of a Dew Point Method 73

R(T + 273.15)17.502 Td Ts
ψt = − (3)
Vw Td + 240.97 Ts + 240.97
Note that Eqs. (1) and (3) give negative values. The water potential, or
potential energy of the water with reference to pure free water is negative. The
term suction has an implied negative, so the suction values we report here are
positive values. Equation (3) shows two things. First, suction measurement ac-
curacy depends mainly on a correct measurement of temperature difference,
not on knowing the absolute temperatures. Second, suction is roughly propor-
tional to the difference between dew point and sample temperature. A rough
sensitivity analysis using Eq. (3) shows that 0.1◦ C difference between sample
and dew point temperature is approximately equal to 1 MPa. The WP4 has
a temperature resolution of approximately 0.01◦ C giving a water potential
resolution of 0.1 MPa.
Performance Verification
The dew point measurement is a primary measurement in the sense that its
prediction of suction comes from fundamental thermodynamic relationships
and measurements of temperature. The temperature measurements, of course,
need to be accurate for the results to be accurate. Factory calibration of the
instrument involves setting the sensitivity of the thermopile using a black
body radiator with an embedded temperature standard, and making minor
adjustments to the thermocouple sensitivity using unsaturated salt standards
of known suction. To verify the performance of the WP4 over its measure-
ment range we tested a standard unit using saturated salts of known suction
(Greenspan 1977). These were independent of the solutions used for setting
the thermocouple sensitivity. Readings for one unsaturated salt are also shown
to indicate performance at low suction. The results are shown in Table 1.
Table 1. WP4 measurements on standard salt solutions. The mean and standard
deviation for three readings on each salt are shown. Measurements are at 25◦ C
Suction WP4 Mean ± SD

(MPa) (MPa)
0.5 mol KCl 2.22 2.19 ± 0.08
Sat. KCl 23.37 23.41 ± 0.11
Sat. NaCl 39.09 39.16 ± 0.07
Sat. Mg(NO3 )2 87.76 88.62 ± 0.04
Sat. MgCl2 153.75 155.28 ± 0.05
Sat. LiCl 300.20 302.06 ± 0.80
The measurements agree with the standards to within 1% or better for

all of the saturated salts indicating excellent accuracy over that entire range.
The resolution is 0.1 MPa, which is 5% of the reading at 2 MPa and 10% of
the reading at 1 MPa. We point this out to make sure the reader is aware
of the limitations of vapor pressure methods on the wet end. They are not a
substitute for axis translation and tensiometer measurements. For the SWC
data we will present we will avoid using values wetter than 1 MPa.
Sample Preparation
Correct sample preparation methods are critical for obtaining good SWCs.
The following methods have given reliable results in our laboratory. It should
be pointed out that a wetting characteristic will be different from a dry-
ing characteristic because of moisture hysteresis. The methods here are for a
wetting characteristic. Previous work has shown that, at the suction values
covered by vapor pressure methods, bulk density has little if any effect on
suction (Campbell and Gardner 1971). All measurements are therefore made
on soil that has been air dried and sieved through a 2 mm sieve.
The range of water content for the SWC is strongly dependent on the clay
content of the sample, as shown in Table 2. This table gives silt and clay
fractions, 1.5 MPa water contents and air dry soil water contents for repre-
sentative members of the 12 soil texture classes. It is important to point out
that these are average values from many soils. They can serve as a guide, but
many things in addition to texture can influence these values. Our objective
is to obtain a set of samples with water contents roughly ranging between the
values shown in the table.
The procedure for preparing samples is as follows. Begin with air dry soil
(soil that has been exposed to the air for several days so that it is in moisture
Table 2. Representative silt and clay fractions and typical water content values at
1.5 MPa and air dry suction for the 12 soil texture classes
Silt Clay w1.5 wad

Texture
(g/g) (g/g) (g/g) (g/g)
Sand 0.05 0.03 0.008 0.003
Loamy sand 0.12 0.07 0.015 0.005
Sandy loam 0.25 0.10 0.045 0.015
Sandy clay loam 0.13 0.27 0.143 0.048
Loam 0.40 0.18 0.106 0.035
Sandy clay 0.07 0.40 0.204 0.068
Silt loam 0.65 0.15 0.098 0.033
Silt 0.87 0.07 0.075 0.025
Clay loam 0.34 0.34 0.174 0.058
Silty clay loam 0.58 0.33 0.166 0.055
Silty clay 0.45 0.45 0.204 0.068
Clay 0.20 0.60 0.234 0.078
equilibrium with the air). The soil water content is defined as the mass of
water divided by the mass of oven dry soil. This definition can be used to
obtain the following equation for determining the mass of water to add to a
given mass of air dry soil to get the desired water content:
w − wad
Mw = Mad (4)
1 + wad
where Mad is the mass of air dry soil, w is the desired final water content, and
wad is the air dry water content of the soil (from Table 2). To prepare samples,
determine the soil texture by feel or particle size analysis and use Table 2 to
find the approximate air dry and 1.5 MPa water contents. Weigh out five to
ten 50 g samples of sieved, air dry soil into containers which can be sealed
with a lid. Use Eq. (4) to compute amounts of water to add to the samples to
produce a set of samples with water contents spanning the range from air dry
to a little wetter than 1.5 MPa. Thoroughly mix the water with the samples,
seal them and allow them to stand for 24 hours or longer to equilibrate. At the
end of the equilibration period a sub sample of the soil is placed in a sample
cup and its suction determined with the WP4. Discard samples with suctions
wetter than 1 MPa. Determine the water content of all remaining samples by
oven drying. If the stainless steel cups are used in the WP4, the same sample
on which water potential was determined can be weighed, oven dried at 105◦ C,
and reweighed to determine the water content. If polyethylene cups are used,
a separate sub sample is dried using standard drying tins. The water content
is computed as the mass of water (change in sample mass on drying) divided
by the mass of the oven dry soil (dry mass of the sample and container minus
the mass of the container). Extreme care is necessary, especially in coarse
textured soils, to obtain accurate water content values. Once this initial set
of samples is run, additional samples can be prepared with appropriate water
contents to fill in where data are sparse.
An alternative approach, useful for high clay samples that can not be
mixed with small quantities of water to give a uniform water content is to
place approximately 5 g samples of air dry soil in WP4 sample cups, suspend
the cups over water to take up water through vapor absorption, and remove
the cups at intervals to get a range of suctions. Again, the cups need to be
sealed and allowed to equilibrate for 24 hours or more before readings are
taken.
Results
Data from samples measured using the WP4 with the technique just described
are shown in Fig. 1. Soils from sand to clay are shown. Here the base 10
logarithm of the suction is plotted as a function of water content. Each data set
except the sand contains data from 2 or 3 independent sample runs, indicating
excellent agreement among samples. No systematic deviation from a straight
3
Dune Sand
Bentonite
2.5
Missouri
Texas
2
New Mexico
Log 10 (MPa)
1.5
0.5
0
0 0.05 0.1 0.15 0.2 0.25
Water Content (g/g)
Fig. 1. Soil Water Characteristics for five soils. Each data set is for 2 or 3 samples.
Lines are from linear regression
line is seen in any of the data, so the SWC model for the range of suctions
shown in Fig. 1 is
ψt = 10aw+b (5)
where a and b are constants for the particular soil. Table 3 gives the values
of a and b for the soils in Fig. 1. The slope, a can be used as an index to
identify expansive soils (McKeen 1992) (note that McKeen used pF, which is
the base 10 logarithm of the suction in cm of water; these units are no longer
recommended). According to McKeen’s classification, only the Missouri and
Bentonite samples are expansive. The Missouri soil is low to medium, and the
Bentonite is called “special case,” beyond high.
Table 3. Coefficients for Eq. (5) for the five soils in Fig. 1
Soil Texture Slope a Intercept b r2

Dune Sand Sand −224 2.94 0.999
New Mexico Clay loam −25.8 3.14 0.997
Missouri Silty clay loam −22.3 3.02 0.997
Texas Clay −14.2 3.10 0.993
Bentonite Clay −5.2 2.46 0.98
Conclusions
The dew point instrument is a primary method for measuring total soil suc-
tion in the range 1 to 400 MPa with an accuracy of about 1%. A measurement
takes less than 5 minutes. This makes it a useful device for rapid determina-
tion of soil water characteristics in this range. Moisture characteristics can
be efficiently obtained from a texture-based estimate of the range of water
contents needed to cover the range of desired suctions. Plotting water content
vs. logarithm of suction gives a linear relationship, the slope of which can be
useful in detecting expansive soils.
References
Campbell GS (1988) Soil water potential measurement: an overview, Irrig Sci 9:265–
273
Campbell GS, Gardner WH (1971) Psychrometric measurement of soil water poten-
tial: temperature and bulk density effects, Soil Sci Soc Am Proc 35:8–12
Gee GW, Campbell MD, Campbell GS, Campbell JH (1992) Rapid measurement
of low soil water potentials using a water activity meter. Soil Sci Soc Am J
56:1068–1070
Gee GW, Ward AL, Zhang ZF, Campbell GS, J Mathison (2002) The influence of
hydraulic nonequilibrium on pressure plate data, Vadose Zone J 1:172–178
Greenspan L (1977) Humidity fixed points of binary saturated aqueous solutions, J
Res Nat Bureau Stds – A. Physics and chemistry 81A:89–96
Leong E–C, Tripathy S, Rahardjo H (2003) Total suction measurement of unsatu-
rated soils with a device using the chilled-mirror dew-point technique, Geotech-
nique 53:173–182
Lu N, Likos WJ (2004) Unsaturated Soil Mechanics. Wiley, Hoboken, NJ
McKeen RG (1992) A model for predicting expansive soil behavior. In: 7th Int Conf
on Expansive Soils, Dallas, 1:1–6
Rawlins SL, Campbell GS (1986) Water potential: thermocouple psychrometry. In:
Methods of Soil Analysis, Part 1, 2nd Ed. Agron Monogr 9:597–618, Madison,
WI
A Comparative Study of Soil Suction
Measurement Using Two Different
High-Range Psychrometers
Rafaela Cardoso1 , Enrique Romero2 , Analice Lima3 , and Alessio Ferrari4

1
Instituto Superior Técnico, Portugal rafaela@civil.ist.utl.pt
2
Universitat Politècnica de Catalunya, Spain enrique.romero-morales@upc.edu
3
Universitat Politècnica de Catalunya, Spain analice.lima@ig.com.br
4
Università degli Studi di Palermo, Italy ferrari@diseg.unipa.it
Summary. An extensive experimental study was performed to compare the mea-

surement capabilities within different ranges of two total suction measurement equip-
ment: SMI transistor psychrometers and a chilled-mirror dew-point psychrometer
(WP4 Dewpoint PotentiaMeter). The equipment were used in a systematic way
to determine the relative humidity of the environment surrounding different com-
pacted clayey soils along drying paths and covering a wide total suction range (0.1
to 70 MPa). The equipment description and the calibration protocols followed are
presented, as well as suggestions regarding the improvement of their performance.
The water retention curves of two different compacted clayey soils are presented
and commented by taking into account the corrections proposed for the readings.
A possible explanation for differences in the measurements observed between both
instruments in the high suction range is presented in terms of the hydraulic paths
undergone by the soils during the measurement period.
Key words: total suction, transistor psychrometer, chilled-mirror dew-point psy-

chrometer, calibration, water retention curve
1 Introduction
The determination of the water retention properties in a wide suction range of
natural and compacted soils is a fundamental issue in many geoenvironmental
and geotechnical applications, such as engineered barriers and liners, ground-
atmosphere interactions and compacted fills used in man made structures –
earth dams, road subgrades and embankments.
This paper focuses on a performance comparative study of two high-
range psychrometers for total suction measurement, which have been widely
used and their reliability widely accepted. One of these equipment is the
SMI transistor psychrometer (Soil Mechanics Instrumentation, Adelaide, Aus-
tralia), which has been used by several authors (e.g. Dimos 1991, Woodburn
80 R. Cardoso et al.
et al. 1993, Truong and Holden 1995, Woodburn and Lucas 1995, Mata
et al. 2002, Boso et al. 2004). The other one is a chilled-mirror dew-point
psychrometer (WP4, Decagon Devices, Inc., Pullman, Washington, USA,
www.decagon.com). The latter equipment has been developed in recent years
to accurately determine the relative humidity in a wide range and involving
a reduced time of reading. This equipment or a similar one has been used by
Loiseau (2001), Leong et al. (2003), Tang and Cui (2005), and Thakur and
Singh (2005).
To undertake such a study, two compacted clayey soils were used that
are able to hold moisture in a wide range of total suctions (0.1 to 70 MPa).
The drying branch of the water retention curves of these soils – com-
pacted/destructured argillite and compacted scaly clay – were determined
using both equipment. Careful and well-posed protocols were followed to cali-
brate the psychrometers and perform the measurements, which are discussed
within the context of the determination of the water retention curves. Based
on the results, suggestions for the calibration protocols and corrections for the
readings are made. A possible explanation for differences in the measurements
observed between both instruments in the high suction range is also discussed.
2 Equipment Used for Total Suction Measurement
A picture of the equipment used in this research is presented in Fig. 1. The

working principle of the psychrometers is based on the measurement of the
relative humidity of the air inside a small sealed chamber. At equilibrium,
the relative humidity of the air in the chamber is the same as the relative
humidity of the soil. The soil total suction, ψ, the equivalent suction derived
a) SMI transistor psychrometers b) Chilled-mirror dew-point psychrometer,

WP4
Fig. 1. Equipment used to measure total suction
A Comparative Study of Soil Suction Measurement 81
from the vapour pressure, p, in equilibrium with the soil water, relative to
the saturation vapour pressure, po , at the same temperature in equilibrium
with free pure water, is related to the relative humidity, RH = p/po , by the
psychrometric law (Fredlund and Rahardjo 1993):
RT ρw
ψ=− ln(RH) (1)
Mw
where R is the gas constant (8.314 J/(mol K)), T is the absolute temperature,
Mw is the molecular mass of water (18.016 kg/kmol) and ρw is the density of
pure water (998 kg/m3 at 293 K).
The psychrometers are known to have some variables and factors that
can affect the measurement. One of these variables is temperature, which
can induce fluctuations in relative humidity readings due to the temperature
dependence of po . Readings should be taken at a constant temperature of
the measurement environment in thermal equilibrium with the soil sample.
In addition, hysteresis effect, change of calibration with time and equalisation
period are also factors that affect the reliability of the readings.
The devices are presented schematically in Fig. 2. As shown in Fig. 2a,
the SMI psychrometer probe consists of two bulbs, which act as ‘wet’ and
‘dry’ thermometers that are placed inside a sealed chamber in equilibrium
with the soil sample (cylinder 12 mm high and 10 mm in diameter). A drop of
distilled water with specified dimensions is used to wet the ‘wet’ thermometer.
The rate of evaporation of the drop is inversely proportional to the quantity
of vapour present in the atmosphere of the chamber: the drier the air, the
quicker the evaporation. Evaporation from the ‘wet’ bulb lowers the tempera-
ture so that this thermometer shows a lower temperature than the one of the
‘dry’ bulb thermometer. The psychrometer measures indirectly the relative
humidity by the difference in temperature between the ‘dry’ and the ‘wet’
bulbs – the output voltage can be related to the relative humidity (or total
suction) through a suitable and extended calibration. The probe is installed
Fig. 2. Schema of a) SMI transistor psychrometer, and b) chilled-mirror dew-point

psychrometer (Leong et al. 2003)
in a thermally insulated bath to minimise temperature induced fluctuations.

The standard equilibration period is one hour.
The chilled-mirror dew-point psychrometer measures the temperature
at which condensation first appears (dew-point temperature). As shown in
Fig. 2b, a soil sample in equilibrium with the surrounding air is placed in a
housing chamber containing a mirror and a detector of condensation on the
mirror. The temperature of the mirror is precisely controlled by a thermo-
electric (Peltier) cooler. The sensing of the condensation point is done with a
photoelectric cell, which detects the presence of condensed water on the mirror
through the reduction in the mirror’s reflection. A thermocouple attached to
the mirror records the temperature at which condensation occurs. The relative
humidity is computed from the difference between the dew-point temperature
of the air and the temperature of the soil sample, which is measured with
an infrared thermometer. A fan is included in the sealed compartment to
speed up the equalisation period between the sample and the surrounding
air. The total suction measuring time is around 5 minutes. The device also
has a temperature controller to set the temperature of the sample at which
relative humidity measurement is to be made. It is assumed that since both
dew-point and sample surface temperatures are simultaneously measured, the
need for complete thermal equilibrium is eliminated (Decagon Devices, Inc.
2003). However, temperature gradients between the measurement environ-
ment and the sample, may result in unstable readings, and more accurate
results are obtained by controlling the temperature difference between the
sample and the chamber block – usually within 0.1◦ C of each other. More
details are described in Decagon Devices, Inc. (2003).
Table 1 presents the comparison of both equipment concerning suction
range, output, accuracy, measurement time and calibration. Calibration issues
will be discussed further in this paper.
Table 1. Specifications of the psychrometers used
Equipment SMI Psychrometer Chilled-mirror dew-point WP4

Suction range 1 to 70 MPa (*) 1 to 60 MPa (max. 300 MPa)
Output reading Voltage, suction (logger) Suction and temperature
Accuracy < ±0.05 pF ± 0.1 MPa from 1 to 10 MPa and
± 0.01 pF (repeatability) ± 1% from 10 to 60 MPa
Measuring time Usually 1 hour 3 to 10 minutes
Calibration Multiple point calibration Single point calibration
Sample geom. φ = 15 mm, h = 12 mm Sample cup: φ = 37 mm, h = 7 mm
(*) disconnecting the probe from the standard logger (Woodburn and Lucas 1995)
3 Calibration Protocols
The calibrations of the psychrometers were performed in a controlled-tempera-
ture room (22 ± 1◦ C) at a relative humidity of 40 ± 5% by comparing the
relative humidity measured to the one applied with standard salt solutions,
and covering a wide total suction range (2–73 MPa). The following sections
describe the calibration procedures adopted for each psychrometer.
3.1 SMI Psychrometer Calibration
The calibration of the SMI psychrometers – relationship between millivolts

output and known relative humidity value – was performed adopting the stan-
dard procedure described by Dimos (1991) and Woodburn et al. (1993). The
psychrometers were first equilibrated at zero total suction for one hour with
distilled water and the output adjusted to the initial offset. Afterwards, the
different outputs were recorded following a single step calibration of one-hour
stabilisation period – each total suction step was applied after equalisation at
zero total suction. Increasing suction steps were followed to avoid hysteresis
effects. The operation range of the SMI was extended in the high total suction
domain by disconnecting the equipment from the standard logger and using
a millivoltmeter instead, as suggested by Woodburn and Lucas (1995) and
Mata et al. (2002). The solutions used to apply the relative humidity were
prepared with KCl, NaCl and NaBr·2H2 O, covering a suction range from 2 to
73 MPa (Romero 2001, OIML R 121 1996). To fit the measured points, linear
relationships were used. Figure 3 shows the calibration for one-hour equilibra-
tion time using a standard drop size. As observed, two linear relationships are
necessary to best fit the results, one defined for the low total suction range
(from 0 to 20 MPa) and the other for the high one (greater than 20 MPa).
Fig. 3. Calibration of SMI psychrometers (one hour reading time and standard drop
size)
Fig. 4. Time evolution of output for the same psychrometer for different values of
total suction applied using a standard drop size. a) Results from the present study,
b) Results from Mata et al. (2002)
The time period adopted for equilibration has a major role on the calibra-
tion at high suctions since the readings are not constant in time, as illustrated
in Figs 4a and 4b for a standard drop size. As observed in these figures, ini-
tially the outputs increase due to the influence of the measuring environment
that maintains the relative humidity of the laboratory. At ψ < 40 MPa, the
outputs tend to rapidly level off without peak at the relative humidity in
equilibrium with the soil sample. On the contrary, at ψ > 40 MPa, the output
at the quasi-equilibrated state with the soil sample is rapidly achieved after
a peak in the transient evolution. In addition, the output evolution show a
continuous decreasing shift after the quasi-equilibrated state, in which lev-
elling out of the output does not occur when applying high total suctions
( > 40 MPa). This fact makes the selection of the final reading difficult and
implies smaller voltages for the higher suctions after one hour, which explains
the bilinear calibration relationship plotted in Fig. 3.
A reasonable explanation for the changes in the readings with time af-
ter equalisation can be the fact that the measuring chamber, where the soil
sample is placed, is not completely closed. As a consequence, the measuring
environment is affected by the relative humidity of the laboratory. At elevated
soil total suctions, the water drop evaporates continuously demanded initially
by the total suction of the soil and later by the influence of the relative humid-
ity of the laboratory. An observation consistent with this statement is that
the mass of the high-suction soil samples slightly increase along the reading
period, indicating some small water transfer to the soil, and that the water
drop evaporates. In fact, at ultimate conditions, the water drop can dry com-
pletely, and the voltage reading should return to the reading associated with
zero total suction, i.e., no temperature difference between bulbs.
In order to amend the aforementioned problem, two procedures have been
suggested to make readings in the high-suction range. A way to obtain a
more constant output from the instrument as the water drop evaporates is to
increase the size of the standard drop. Woodburn and Lucas (1995) reported
a flatter slope in the time evolution using 1.5 water drop size and one hour
measuring time. Another possibility is to reduce the time for equilibrium and
maintain the standard drop size as suggested by Mata et al. (2002), who
proposed to change the measurement time according to the total suction range
considered: 60 minutes for 1 ≤ ψ ≤ 33 MPa; 30 minutes for 33 < ψ ≤ 57 MPa
and 20 minutes for 57 < ψ ≤ 84 MPa. In this way, the linear relationship of the
low-suction range could be extrapolated throughout the entire suction range,
as shown in Fig. 5. The reduction in the measurement time depends on the
prediction of the soil total suction before doing the measurement, which is not
always easy to assume. In this paper, the bi-linear calibration with standard
drop size presented in Fig. 3 is preferred to measure soil suction and perform
the comparison with the chilled-mirror dew-point psychrometer.
Besides drop size and measurement time, temperature, hysteretic be-
haviour, change in calibration with time and gap size, are other factors that
affect the results of transistor psychrometers, as extensively studied by Truong
and Holden (1995).
3.2 WP4 Psychrometer Calibration

As suggested by the manufacturer, the single-point calibration of the WP4 is
performed with dilute KCl or NaCl solutions (0.5 m) applying a total suction
of 2.2 MPa at 20◦ C. The calibration slope is fixed during factory calibration
and the user only adjusts the zero offset in the low-suction range (Decagon
Devices, Inc. 2003). This single point will be named calibration point and
indicated with sub-script ‘cal’ from now on. However, in addition to offset
shift, sensitivity drifts over time are expected to occur, which make the single-
point calibration inappropriate.
Fig. 5. Linear relationship for total suction applied and SMI psychrometer output
for standard drop size and reduced measurement time. a) Results from the present
study, b) Results from Mata et al. (2002)
To verify the calibration throughout the measurement range, in-dependent

offset calibrations were performed at three different total suction levels, ψcal =
2.2, 19.4 and 38.2 MPa (22◦ C). Once the offset had been set, the verification of
the calibration was carried out for each ψcal by checking the WP4 reading with
different total suctions. Partially saturated solutions of NaCl were used along
the verification programme. The relationship between the relative humidity,
RH, and the molality, m, of the solute (mol of NaCl/kg of pure water, with
molecular mass of the salt 58.443 kg/kmol) at 22◦ C is given by the following
empirical expression (Horvath 1985, Romero 2001):
RH = 1 − 0.035m − χm(m − 3),
χ = 1.7147 × 10−3 for m ≥ 3mol/kg , (2)
−3
χ = 1.1421 × 10 for m < 3mol/kg .
Table 2 summarises the results of this verification programme.
Table 2. Results of the verification programme for different ψcal
Measured values for different ψcal (22◦ C)

◦
Applied total suction (22 C)
2.2 MPa 19.4 MPa 38.2 MPa
6.0 MPa (NaCl 1.3 mol/kg) 5.5 4.4 4.8
10.5 MPa (NaCl 2.2 mol/kg) 9.8 8.7 9.1
19.4 MPa (NaCl 3.7 mol/kg) 20.4 19.5 19.7
38.2 MPa (NaCl 6.1 mol/kg) 37.4 37.5 38.1
The readings performed by adopting ψcal = 2.2 MPa display some shift
when compared to the applied suctions, especially in the high-suction range.
In a similar way, deviations in the readings performed in the low-suction
range are detected for ψcal = 19.4 and 38.2 MPa. This situation highlights
the importance of carrying out the offset calibration nearby the expected
measurement range. Figure 6 shows the calibration results for different ψcal .
The sensitivities, b, of the calibrations, i.e., slope of the linear relationships
shown in Fig. 6, are plotted in Fig. 7 for different ψcal . Results were fitted to
the expression indicated in the figure (Fig. 7). In order to correct the WP4
psychrometer readings, the following expression can be used:

1
ψ = ψWP4 − (ψWP4 − ψcal ) 1 − (3)
b
where ψ is the soil total suction, ψcal is the calibration point, ψWP4 is the
measured value, and b is the sensitivity presented in Fig. 7. As observed in
Eq. (3), maximum corrections are obtained for points distant from the cali-
bration point. According to Fig. 7, the calibration point ψcal for b = 1 – no
need for correction according to Eq. (3) – is 2.7 MPa, which is close to the
value suggested by the manufacturer (ψcal = 2.2 MPa).
Fig. 6. Total suction readings for different ψcal
Fig. 7. Variation of the calibration sensitivity, b, for different ψcal
4 Water Retention Curves
Two different compacted clayey soils were used to study the performance of the
psychrometers. Main properties of these soils – destructured argillite (Soil A)
and compacted scaly clay (Soil B) – are presented in Table 3.
Only the drying branch of the retention curves of these soils was deter-
mined starting from the as-compacted water content. In this way, total suction
Table 3. Properties of the different soil samples (average values)
γd γs
Sample W (%) w (%) PI (%)
(kN/m3 ) L (kN/m3 )
A. Compacted destructured argillite 12.4 17.2 37 16 27.4
B. Compacted scaly clay 15.3 17.4 58 30 27.7
determinations followed the same suction increase path used in the calibration
of SMI psychrometers, and hysteresis effects (both of the soil and the equip-
ment) were minimised. The suction range was limited to the equipment used,
previously presented in Table 1. WP4 psychrometer readings were corrected
according to Eq. (3).
Measured values using the two equipment are plotted in Figs 8 and 9 for
soils A and B, respectively. Water retention results were fitted to a modified
form of van Genuchten (1980) expression for water content, w, as a function
of total suction, ψ (Romero and Vaunat 2000)
m
1
w = wmax C(ψ − ψ0 ) · , (4)
1 + (α(ψ − ψ0 ))n

ln 1 + ψ−ψa
0
C(ψ − ψ0 ) = 1 − (5)
ln 2
where wmax is the water content at saturation for total suction ψ0 . Parameters
n, m and α are the same as used in van Genuchten’s expression. The parame-
ter α is mainly associated (inversely) with the air-entry value of the soil. The
original van Genuchten’s expression is not adequate to fit retention curve data
for clayey soils. For this reason, the expression was modified by a correction
function Eq. (5), which makes the curve tend to a linear relationship between
100
SMI -Drying
WP4-Drying
Curve SMI (drying)
Curve WP4 (drying)
Total suction (MPa)
10
0
0 5 10 15 20 25
water content (%)
Fig. 8. Water retention curves of soil A (drying branch)

100
SMI -Drying
WP4-Drying
Curve SMI (drying)
Curve WP4 (drying)
Total suction (MPa)

10
0
0 5 10 15 20 25
water content (%)
Fig. 9. Water retention curves of Soil B (drying branch)
the logarithm of ψ and w in the high-suction zone, with intersection param-

eter, a, at w = 0. Fitted parameters for the two soils are listed in Table 4.
Table 4. Constants for the definition of the water retention curves for the soils
(drying branch)
Soil Soil A Soil B

Equipment SMI WP4 SMI WP4
wmax (%) 23.5 23.5 21.0 21.0
ψ0 (MPa) 0.1 0.1 0.1 0.1
a (MPa) 200 600 200 300
α (MPa−1 ) 2.913 3.400 0.990 0.990
m 0.14 0.13 0.10 0.10
n 2.5 2.5 2.5 2.5
Comparing the retention curves for each soil obtained with the two psy-
chrometers, it can be observed that a good agreement is found in the total
suction range 0.5 to 7 MPa. Above this upper value the readings measured
with WP4 psychrometer are systematically higher than the SMI ones and the
difference between instrument measurements become larger for the highest
total suctions.
The equipment were carefully calibrated and verified, so that differences
in the high-suction range cannot be solely attributed to errors in the measure-
ments. Nevertheless, the differences can be explained in terms of the hydraulic
paths undergone by the soils during the measurement time. To help in this
interpretation, Figure 10 shows the expected behaviour of the samples inside
the measurement chamber of SMI and WP4 psychrometers.
Just after closing the chambers that house the samples, the relative hu-
midity that prevails in the psychrometer chambers corresponds to the relative
humidity of the laboratory (HR0 = 40%). In the case of SMI psychrometer,
the water drop immediately evaporates increasing the relative humidity of the
chamber to HR1 > HR0 , as shown schematically in Fig. 10. The sample at
a lower relative humidity HRsoil undergoes some wetting before reaching the
equalisation state at HReq SMI , which is the state finally measured by the
SMI psychrometer. In other words, it is expected that the soil housed inside
the SMI chamber undergoes some wetting induced by the evaporation of the
drop. On the contrary, the sample placed inside the equalisation chamber of
the WP4 psychrometer undergoes some drying before reaching equalisation
HReq WP4 . The differences between the two psychrometers are readily ob-
served in Fig. 11, in which the hydraulic paths 2–3 followed by the samples
during equalisation in the measurement chambers are schematically plotted.
The sample installed in the WP4 undergoes some small drying along the main
drying curve during equalisation in the measurement chamber (path 2–3). On
the other hand, the sample installed in the SMI measuring chamber undergoes
some small wetting along a scanning wetting path 2–3. As a consequence, the
total suctions measured and the final water contents are different, displaying
slightly higher suctions and slightly lower water contents the sample equili-
brated in the WP4 measuring chamber.
Fig. 10. Equalisation process in the measurement chamber of SMI and WP4 psy-
chrometers. High-suction range for HRSOIL > HR0
Fig. 11. Hydraulic paths followed by the samples during equalisation in the mea-
surement chambers of SMI and WP4 psychrometers
5 Conclusions
This paper presents the equipment description and the calibration protocols
followed for two high-range psychrometers, namely the SMI transistor psy-
chrometer and the WP4 chilled-mirror dew-point psychrometer.
Regarding the improvement of the performance of the SMI psychrometer
for measuring in the high-suction range (20 to 70 MPa), three different proce-
dures were suggested: a) to use one-hour measurement time with the standard
drop size; b) to reduce the measurement time with standard drop size in ac-
cordance to the total suction measured; and c) to use one-hour measurement
time with 1.5 standard drop size. Procedures b) and c) reduce the shift in
the time evolution of the output, since they minimise the interaction with the
relative humidity of the laboratory –the measuring chamber is not a closed
system and vapour is lost inducing the reduction of the drop size. Protocols
b) and c) require a single linear calibration relationship along the measur-
ment range. However, protocol b) can only be followed if the total suction
is assumed before the reading. In the paper, procedure a) was used, which
required a bi-linear calibration relationship.
Concerning the WP4 psychrometer, the manufacturer’s calibration proto-
col that uses a single-point was found to be inappropriate. The calibration
slope is fixed during factory calibration and the user only adjusts the zero off-
set in the low-suction range. A verification programme using different single-
point calibrations was performed and contrasted with partially saturated salt
solutions that applied different relative humidities. It was observed that some
shifting in the readings occurred for points distant from the single calibra-
tion point. An expression to correct the readings was suggested based on the
verification study performed.
The drying branches of the water retention curves of two different com-
pacted clayey soils were determined by taking into account the corrections
proposed for the different psychrometers. A good agreement in the equipment
readings was observed, particularly in the low total suction range from 1 to
7 MPa. On the contrary, in the high-suction range (7 to 70 MPa) differences
between the readings of both psychrometers were observed, which increased
with the total suction of the soil. The paper presented a possible explanation
for these differences in terms of the different hydraulic paths undergone by the
soils along the measurement period, i.e., along the equalisation period inside
each equipment chamber. Samples installed inside the WP4 chamber undergo
some small drying, whereas the samples in the SMI psychrometric chamber
experience some wetting. As a consequence, the total suctions measured and
the final water contents are slightly different. WP4 total suction readings are
slightly higher since the samples follow their main drying curves, whereas SMI
readings are below these main drying curves.
Finally, the WP4 can be a promising alternative – it presents a reduced
measuring time- in the total suction range where the readings do not present
relevant differences with the SMI (1–7 MPa). However, this suction range can
be different for other type of soils, and further research is needed to clearly
state this range. The differences obtained in the high-suction range had shown
the need for further studies in this range to increase the reliability of both
psychrometers.
Acknowledgements
The authors would like to thank Dr Francesc Ferrer for lending to UPC the
WP4 equipment and also for his useful comments regarding its calibration
and use.
The first author acknowledges the financial support given by the Por-
tuguese Foundation for Science and Technology, FCT (Ref. POCTI/ECM/
59320/2004). The third author acknowledges the financial support given by
ESV EURIDICE GIE Belgium contract (Ref. EUR-04-248).
References
Boso M, Romero E, Tarantino A (2004) The use of different measurement techniques
to determine water retention curves. In: T. Schanz (ed) 93 Springer Proceedings
in Physics. Unsaturated Soils: Experimental Studies. Springer, Berlin 1:169–181
Fredlund DG, Rahardjo H (1993) Soil Mechanics for Unsaturated Soils. John Wiley
& Sons, New York
Decagon Devices, Inc. (2003) WP4 Water Dewpoint Potentiometer. Operator’s Man-
ual, Version 2.2. Decagon Devices, Inc., Pullman (www.decagon.com)
Dimos A (1991) Measurement of soil suction using transistor psycrometer. Internal

Report IR/91–3, Special Research Section, Materials Technology Department,
Vic Roads
Horvath AL (1985) Handbook of aqueous electrolyte solutions: physical properties,
estimation and correlation methods. Ellis Horword Limited, John Wiley & Sons
New York
Leong E-C, Tripathy S, Rahardjo H (2003) Total suction measurement of unsatu-
rated soils with a device using the chilled-mirror dew-point technique, Géotech-
nique 53(2):173–182
Loiseau C (2001) Transferts d’eau et couplages hydromécaniques dans les barriers
ouvragées. PhD thesis, École Nationale des Ponts et Chaussées, Paris
Mata C, Romero E, Ledesma A (2002) Hydro-chemical effects on water retention in
bentonite-sand mixtures. In: Jucá JFT, de Campos TMP, Marinho FAM (eds)
Proc 3rd Int Conf on Unsaturated Soils, Recife, Brazil. A.A. Balkema Lisse, The
Netherlands 1:283–288
OIML R 121 (1996) Organisation Internationale de Métrologie Légale. International
Recommendation. The scale of relative humidity of air certified against saturated
salt solutions. Grande Imprimerie de Troyes, Troyes, France
Romero E (2001) Controlled-suction techniques. In: Gehling WY, Schnaid F (eds)
4th National Brazilian Symposium on Unsaturated Soils. ABMS, Brazil 535–542
Romero E, Vaunat J (2000) Retention curves of deformable clays. Experimental
Evidence and Theoretical Approaches in Unsaturated Soils. In: Tarantino A,
Mancuso C (eds) Proc Int Workshop on Unsaturated Soils, Trento, Italy. A.A.
Balkema, Rotterdam 91–108
Tang A-M, Cui Y-J (2005) Controlling suction by the vapour equilibrium technique
at different temperatures and its application in determining the water retention
properties of MX80 clay, Can Geot J 42:287–296
Thakur VKS, Singh DN (2005) Swelling and suction in clay minerals. Advanced
Experimental Unsaturated Soil Mechanics. In: Tarantino A, Romero E, Cui YJ
(eds) Proc Int Symp (Experus 2005), Trento, Italy. A.A. Balkema, Rotterdam
27–31
Truong HVP, Holden JC (1995) Soil Suction Measurement with Transistor Psy-
chrometer. In: Alonso EE, Delage P (eds) Proc 1st Int Conf on Unsaturated
Soils (UNSAT 95), Paris. A.A. Balkema, Rotterdam 2:659–665
van Genuchten MTh (1980) A closed-form equation for predicting the hydraulic
conductivity of unsaturated soils, Soil Sci Soc Am J, 49(2):143–159
Woodburn JA, Holden JC, Peter P (1993) In: Houston SL, Wray WK (eds) Unsat-
urated Soils. Geotechnical Special Publication No. 39. ASCE, New York 91–102
Woodburn JA, Lucas B (1995) New Approaches to the Laboratory and Field Mea-
surement of Soil Suction. In: Alonso EE, Delage P (eds) Proc 1st Int Conf on
Unsaturated Soils (UNSAT 95), Paris. A.A. Balkema, Rotterdam 2:667–671
Determination of the Soil Water Retention
Curve with Tensiometers
Sérgio Lourenço1 , Domenico Gallipoli1 , David Toll1 , Fred Evans2 , and

Gabriela Medero3
1
School of Engineering, Durham University, South Road, Durham DH1 3LE, UK
s.d.n.lourenco@durham.ac.uk, domenico.gallipoli@durham.ac.uk,
d.g.toll@durham.ac.uk
2
Wykeham Farrance Ltd, Chiltern House, Unit 4B, Knaves Beech Business Centre
Loudwater, High Wycombe, Bucks HP10 9QY, UK Fred.Evans@wfi.co.uk
3
Heriot Watt University, School of the Built Environment, Riccarton, Edinburgh
EH14 4AS, UK G.Medero@hw.ac.uk
Summary. An alternative technique for the determination of the soil water reten-
tion curve has recently been proposed whereby a tensiometer is used to measure
soil suction and a balance to record the water content variations. The soil water
retention curve is obtained by drying the soil either continuously or by stages (i.e.
each drying stage is followed by an equalization period). Initial results from tests
on compacted soil suggest that the relatively fast evaporation rate during contin-
uous drying affects the water retention curve whereas the stage drying procedure
provides more accurate results. Factors such as sample geometry and tensiometer
position (relative to the sample) are also likely to affect the response obtained during
continuous drying. These are the object of future investigation.
Key words: tensiometers, soil water retention curve
Introduction
Recently, high capacity tensiometers have emerged as an alternative instru-

ment for the determination of the soil water retention curve because they
provide fast measurements and determine directly the water tensile stress un-
like instruments based on the axis translation.
Cunningham (2000), Toker et al. (2004) and Boso et al. (2003) determined
the soil water retention curve by using an electronic balance to record the pro-
gressive decrease of water content in a sample left to dry to the atmosphere
while using a tensiometer to measure the corresponding increase of suction.
Boso et al. (2003) presented a comparison between stage drying and contin-
uous drying for a sample of reconstituted clayey silt. The evaporation rate
96 S. Lourenço et al.
during continuous drying was slowed down by wrapping the sample in a geo-
textile. The results revealed no differences between the soil water retention
curves determined using the two procedures. Cunningham (2000) investigated
the influence of the evaporation rate for the continuous drying procedure ap-
plied to samples of reconstituted silty clay. In particular, he compared the soil
water retention curves obtained by drying continuously the sample either to
the atmosphere or inside a controlled humidity chamber. Similar results were
obtained from the two procedures suggesting that the evaporation rate had
little or no influence on the resulting soil water retention curve. These two
studies also confirmed that tensiometers could be used to determine the soil
water retention curve in a significantly shorter period of time in comparison
to other conventional testing techniques (see also Toker et al. (2004)).
The stage drying procedure is expected to yield the most accurate results
as the suction is measured after the sample has achieved equalization. On the
other hand, continuous drying tends to be faster and simpler but it is likely
to introduce inaccuracies due to the lack of equalization through the sample.
These inaccuracies depend on factors such as sample size, shape, extent of
surface exposed to drying and tensiometer position relative to the sample.
Clearly, if the suction is measured on the surface of relatively large samples, the
continuous drying procedure is expected to yield an incorrect water retention
curve. Further study is therefore required to confirm the results obtained by
Boso et al. (2003) and Cunningham (2000).
This paper presents some initial data from an experimental investigation
on the particular drying procedure used (i.e. stage drying and continuous
drying) for the determination of the soil water retention curve.
1 Experimental Set up
A sandy clay of intermediate plasticity with PL = 19.7%, LL = 43.3% and

grain size distribution shown in Fig. 1 (Mendes 2006) has been used in this
study. All samples were compacted according to the standard Proctor test at
an initial water content of 25% (optimum water content is 18%).
The sample is enclosed in a metallic ring with a diameter of 100 mm and
height of 30 mm and sealed by two (top and bottom) end plates, which include
fittings to house tensiometers. The tensiometers used in this work have a
nominal air entry value of 15 bar but they are capable of measuring suctions
up to 1600 kPa (see Lourenço et al. (2006)). The water content decrease by the
sample was measured by a digital balance logged into an acquisition system
via a RS 232 interface.
For the case of stage drying, the soil water retention curve was determined
by a sequence of independent drying phases and suction measurements on
the same sample. For each stage, the soil was left to dry by evaporation to
the atmosphere and, following an equalization period, the suction and sample
mass were measured.
Determination of the Soil Water Retention Curve with Tensiometers 97
Fig. 1. Grain size distribution of the sandy clay soil (Mendes 2006)
The detailed procedure was as follows:

1) Drying. The sample is dried by removing the top plate of the measuring
cell and allowing the pore water to evaporate to the air for a set period of
time. Drying was accelerated by a fan located above the sample;
2) Equalization. The top plate is placed back to seal the sample allowing
water redistribution within the soil mass;
3) Suction measurement. The tensiometer is inserted through the bottom
plate in contact with the sample (a good contact is ensured by the own
weight of the sample) and suction is continuously read;
4) Mass measurement. After suction becomes constant indicating equal-
ization, the total weight of cell, sample and tensiometer is recorded.
Steps 1) to 4) are repeated for the number of data points required to define
the entire soil water retention curve. The above sequence is illustrated for one
selected data point in Fig. 3.
For the continuous drying procedure, a sample with the same dimensions
and initial conditions as for the stage drying was placed on the balance to-
gether with the metallic ring as shown in Fig. 2. A tensiometer was gently
pushed into the top surface of the sample to a depth of approximately 3 mm.
The entire length of the tensiometer cable was also supported to minimize
any influence of its stiffness on the mass measurements. Pore water was left to
evaporate through the exposed top surface of the sample while the decrease
of water content and increase of suction were continuously recorded by the
connected PC. The initial sample conditions and a summary of the key results
for all six tests performed are shown in Table 1.
Fig. 2. Experimental set up for the continuous drying tests
Table 1. Initial and final physical indexes and summary of results
Test no. Drying procedure ei wi [%] wf [%] ΔMw [g] smax [kPa] Δt [h]
1 Stages 0.59 24.35 16.88 29.3 568.6 148.8
2 0.55 24.80 17.76 33.5 485.5 114.5
3 0.54 25.25 14.70 42.5 995.0 171.9
4 Continuous 0.52 24.70 18.37 26.0 376.0 25.1
5 0.56 24.76 17.25 30.0 745.2 32.5
6 0.55 24.17 17.23 27.9 584.6 28.44
ei initial void ratio, wi initial water content, wf final water content, ΔMw
mass of water evaporated, smax maximum suction of water retention curve,
Δt test duration
2 Results
During all tests the soil significantly reduced in volume but such shrinkage
was not measured. Therefore the corresponding variation of degree of satu-
ration could not be calculated and the results are presented only in terms of
gravimetric water content.
The entire test sequence for the definition of the soil water retention curve
by using stage drying is shown in Fig. 3, which includes information of both
gravimetric water content and suction for each of the nine drying stages. As
expected, the suction equalizes at increasing values as the gravimetric water
content decreases.
A similar set of information for a test performed by continuous drying is
shown in Fig. 4, where suction and the gravimetric water content are plotted
against the elapsed time. Inspection of Fig. 4 indicates that the decrease of
the gravimetric water content was linear (R2 = 0.9993) with respect to time
confirming that the influence of the stiffness of the tensiometer cable on the
mass measurement of the balance can be considered negligible. The analysis
of Table 1 also indicates that the tests using continuous drying were about
five times faster than the tests using the stage drying, with the entire water
retention curve obtained in less than two days.
The soil water retention curves for all six tests performed are shown in
Fig. 5. All curves show a similar pattern but they tend to diverge for increasing
values of suction. Some curves in Fig. 5 end at lower values of suction due to
premature cavitation of the tensiometer.
Discussion
Data presented in the paper revealed a partial agreement with Boso et al.
(2003) and Cunningham (2000) results, which suggested that the drying rate
had no influence on the soil water retention curve. Inspection of Fig. 5 indicates
that the curves obtained by continuous drying are slightly displaced upwards
with respect to the curves obtained by stage drying. In other words, the suction
measured by continuous drying is higher than that measured by stage drying
at the same water content. This may be explained by the inhomogeneous
distribution of suction and water content through the sample in the former
case, with faster drying of the sample surface with respect to the core. The
Fig. 3. Time sequence for the stage drying test 3, with the sequence of steps followed
for each drying stage illustrated
Fig. 4. Time sequence for the continuous drying test 5
Fig. 5. All soil water retention curves
use of stage drying avoids such shortcoming and is therefore expected to yield
more reliable results.
One limitation of both techniques is that volumetric measurements cannot
be easily integrated in the experimental set up, preventing the determination
of degree of saturation during the test (Toker et al. 2004).
a) 0 b) 0
pore water pressure
pore water pressure

-200 -50
-100
-400
(kPa)
(kPa)
-150
-600
-200
-800
-250
-1000
-300
0 10 20 30 40 0 1 2 3
time (h) time (h)
Fig. 6. Possible sources of error in the determination of the soil water retention curve
by the stage drying procedure: a) poor contact of the tensiometer–soil interface and,
b) insufficient equalization time
In the course of this study it was found that the following sources of errors
had also a possible effect on the accurate measurement of suction and water
content when using the stage drying procedure:
• When testing shrinkable soils, special care should be taken to ensure that
the tensiometer remains in contact with the sample at all times during
equalization of suction inside the measuring cell. In a preliminary test
(Fig. 6a), the suction reading did not stabilize because the tensiometer
was fixed to the top plate of the measuring cell and was, therefore, unable
to follow the downward displacement of the soil sample as this shrunk.
This problem was solved by moving the tensiometer to the bottom plate
so that the sample’s own weight was sufficient to ensure good contact of
the soil with the probe.
• The gravimetric water content determination was affected by the low dry
mass of the samples (approximately 400 g). For instance, an error of 2 g
in the measurement of the mass of the dry sample would be enough to
introduce an error in the water content of about 0.5%. It is possible that
the scatter between the stage drying curves in Fig. 5 is due to the loss of
little quantities of soil while opening and closing the measuring cell to let
the sample dry out at the different water contents.
• For stage drying it is important to ensure achievement of equilibrium at the
end of each stage. A suction rebound is sometimes observed after placing
the tensiometer in contact with the soil, indicating lack of equalization
(Fig. 6b). Hence, in order to avoid erroneous readings, enough time must
be allowed for the stabilization of suction at a constant value.
Conclusions
The paper investigates the use of tensiometers for the determination of the
soil water retention curve of a compacted sandy clay. The soil water retention
curves determined by continuous drying show higher suctions than the curves
obtained by stage drying at the same water content. The former procedure
is probably affected by incomplete equalization of the sample whereas the
latter is expected to yield the most reliable results. Possible factors affecting
the accuracy of continuous drying are also discussed and areas of further
investigation are identified.
References
Boso M, Romero E, Tarantino A (2003) The use of different suction measurement
techniques to determine water retention curves. In: Schanz T (ed) Unsaturated
soils: Experimental studies, Springer-Verlag Berlin, pp 171–181
Cunningham M (2000) The mechanical behaviour of a reconstituted unsaturated
soil. Ph.D. Thesis, Imperial College of Science, Technology and Medicine, London
Lourenço SDN, Gallipoli D, Toll DG, Evans FD (2006) Development of a commer-
cial tensiometer for triaxial testing of unsaturated soils, Geotechnical Special
Publication No. 147, ASCE, Reston, Vol. 2:1875–1886
Mendes J (2006) Personal communication
Toker N, Germaine J, Sjoblom K, Culligan P (2004) A new technique for rapid
measurement of continuous soil moisture characteristic curves, Géotechnique
54(3):179–186
Tensiometer Development for High Suction
Analysis in Laboratory Lysimeters
Cláudio Fernando Mahler and Abdoul Aziz Diene
COPPE/UFRJ – Cidade Universitário, Centro de Tecnologia, Programa de

Engenharia Civil, Caixa Postal 68506, CEP 21945-970 RJ/RJ, Brazil
cfmahler@acd.ufrj.br, abdoulazizdiene@yahoo.com
Summary. This paper presents results when using a tensiometer designed in the
Soil Mechanics Laboratory at COPPE/UFRJ, and several other instruments. This
new instrument measures suction values up to 1500 kPa. In more typical tensiome-
ters, the cavitation of the system hinders the measurement of suction with values over
100 kPa. Ridley and Burland (1993) designed a new model of tensiometer to mea-
sure suction of more than 100 kPa. Mahler et al. (2002) presented a new tensiometer
with which suction values up to 350 kPa were measured. This paper introduces an
updated model of a tensiometer, using a variation of the equipment proposed by
Mahler et al. (2002). The updated model worked extremely well and cost little to
build. The equipment used allows suction to be measured relatively quickly and, as
previously stated, is economical to produce (about US$300). To date, the range of
suction levels that can be measured reaches 1500 kPa with no difficulty. The tests
were carried out in two boxes of 160 × 50 × 60 cm under laboratory conditions. The
equipment used included a 15 m.c.a. ceramic block pressure sensor, de-aerated wa-
ter and a special acrylic tube specifically designed for this instrument. The results
were compared to two simple automated tensiometers and equivalent tensiometers.
In general very good results were obtained. The main final remarks so far are as
follows:
• the high bubble air entry of the ceramic block inhibits the presence of air bubbles,
but the response time is slower for suction values of more than 200 kPa;
• the saturation process used for the ceramic stone worked very well;
• as expected, the position of the equivalence tensiometer influences the value
measured;
• the mini-lysimeter system proved to be a very good alternative for laboratory
tests and for the development of instruments that measure suction;
• the new instrument presented herein proved to be a good and an economical
alternative for measuring matrix suction in the soil.
Key words: suction, tensiometer, unsaturated soil, mini-lysimeter

104 Cláudio Fernando Mahler and Abdoul Aziz Diene
1 Introduction
Tabor (1979) demonstrated theoretically that water resistance traction is
around −500 MPa. On this hypothesis, Ridley (1993) and Ridley and Bur-
land (1993) introduced the first equipment capable of measuring stress in soil
of more than 1 MPa.
König et al. (1994) used a tensiometer for pore pressure in centrifugal
Druck PDCR–81. This apparatus consists of an instrumented silicon diagram
tied in an internal glass cylinder and connected at the porous elements by
steel external box protector conform Fig. 1. Using a large-scale measuring
device, with a porous stone of 15 bar of air entry, saturated at a pressure of
2000 kPa, Ridley (1993) was able to measure suctions of more or less 1370 kPa.
He reported that the ample external strain of the diaphragm with broad neg-
ative pressure could compromise the integrity of the connection between the
diaphragm and the glass cylinder support. Consequently, water is free to pene-
trate the cavities of the device, causing the suspect result of pressure measured
for both negative and positive cases.
Ridley and Burland (1995) presented another prototype based on the prin-
ciple of maximising the sustainable stress in tensiometers to measure high
suctions in soil. This tensiometer makes it possible to measure onsite suctions
of 1500 kPa. The equipment consists of an integrated strain-gauge diaphragm
and a sealed ceramic porous element with a value of 15 bar of air entry, as
shown in Fig. 2.
Guan and Fredlund (1997), likewise, introduced a tensiometer, which mea-
sured matrix suctions in soils of over 1250 kPa when saturated under a six-
cycle pressure from 12000 kPa to −85 kPa (Guan and Fredlund 1997). The
ceramic stone of high air entry value is sealed halfway in a box detachable
from the device. Assembled in the water, the other half of the box acts as a
compression element adjusting and sealing the commercial transducer in the
water reservoir as showed in Fig. 3.
Fig. 1. Sketch of the tensiometer proposed by König et al. (1994)

Tensiometer Development for High Suction Analysis 105
Fig. 2. Sketch of the Tensiometer proposed by Ridley (1995)
Fig. 3. Sketch of the Tensiometer proposed by Guan and Fredlund (1997)
Pacheco (2001) developed a low-cost tensiometer as shown in Fig. 4a and

b, with measurements to even more than 3 atm without cavitation and con-
cluded that: “the reduced water volume used in the transducer and porous
stone interface of the new instrument and the saturation process inhibited
the formation of air bubbles in the system, and permitted suction measure-
ment up to 3 atm, with a time lag of a few seconds up to 1.5 atm.”
Take and Bolton (2002) wanting to overcome the high loss of “Druck
PDCR–81”, conceived new more robust instruments for measuring negative
pore pressure in centrifuge. They developed three prototypes introducing some
format particularities for each. In common, the prototypes were fitted with an
Entran (2000) EPB transducer with 7 bar air entry value and a porous stone
of 15 bar, as shown in Fig. 5.
Fig. 4. Sketch of the First New Tensiometer (Pacheco 2001, Mahler et al. 2002)
Fig. 5. Take and Bolton (2002) prototypes

Fig. 6. Tarantino and Mongiovi (2002)
Tarantino and Mongiovi (2002) introduced a new tensiometer similar to

the Ridley and Burland (1993) prototype, but with several modifications to the
diaphragm diameter and thickness, to the water reservoir size and closing of
the annular opening between the porous stone and tensiometer body (Fig. 6).
This new equipment measured suction up to 1000 kPa for a period of time of
more than 15 days and reached the maximum negative pressure of 2 MPa.
A problem common to all these items of equipment is the cavitation factor
that is expressed when the transducer reading is interrupted, even momentar-
ily. From the various models that explain water cavitation undergoing traction
strain, the most acceptable is that proposed by Harvey et al. (1944): “it is sup-
posed that cavitation arises from the non-dissolved gas nuclei in the space of
the reservoir walls instead of in the free cavities of liquid;” this due to the
fact that the spherical nucleus of free gas is generally unstable and tends to
move into the liquid. On the contrary, the gas nucleus in the cavities of the
container walls can remain indissoluble even under high water pressure. When
the pressure decreases to negative values, these nuclei can expand and pos-
sibly cause cavitation. This process is controlled by gases spreading through
the gas-liquid boundary and furtively moving into the solid-liquid-gas junction
determined by moving forward and back (inward contact angle). Physicians

using glass or steel in a Berthelot tube determine the stress tests for the water.
This tube is initially almost completely filled with water and the remaining
volume consists of a mixture of steam and air. The tube is heated to expand
the liquid content and force the air into the solution. After cooling, the liquid
sticks to the walls of the tube and undergoes gradual increasing stress until
breaking at the start of cavitation.
In order to inhibit the cavitation in the equipment, various solutions were
proposed by a number of authors, which was summarised in procedures ap-
plied to assembly or saturation, or their calibration. these proposals gave rise
to some conclusions on how to prevent cavitation. Ridley (1993) assumed that
the maximum tension sustainable by the tensiometer is also a function of the
air entry value of the porous element. “If the difference in pressure between the
water volume reservoir and the measured pressure in the soil exceeds the air
entry value, the air can be drawn into the water reservoir; a variation of exter-
nal pressures that operate in the device will result in expansion or contraction
of the air bubbles and the measured pressure is potentially detached from re-
ality.” Marinho and Chandler (1995) assigned the use of a small water volume
in the suction measurement system as responsible for the non-occurrence of
cavitation, and that the minimum volume possible is limited by the trans-
ducer diaphragm. The pre-pressurisation of water was considered essential for
the saturation of tensiometers by Ridley and Burland (1993) and Guan and
Fredlund (1997).
Some differences however do exist between the two pre-pressurisation
methodologies.
Guan and Fredlund (1997) apply several pre-pressurisation cycles first un-
der a vacuum of −85 kPa followed by higher positive pressures until 12 000 kPa.
In fact, it was suggested that the rupture tension is initially affected by the
number of cycles and magnitude of applied positive pressure (Guan and Fred-
lund 1997). Ridley and Burland (1993) affirmed that the pre-pressurisation
procedure is less important. These authors proposed that a tensiometer be
saturated by applying constant pressure of 4000 kPa maintained for at least
24 hours. Tarantino et al. (2000) made suction measurements higher than the
air entry value of the porous element without cavitation and concluded that
“cavitation may occur before the soil-equipment system equalizes, causing the
interruption of the test and pre-saturation of the instrument; the knowledge
of conditions that address cavitation is, therefore, essentially based on opti-
mising the instrument’s design and on determining a suitable experimental
process.”
Cavitation does not occur if the system is free from cavitation nuclei,
which in short means the use of pure clean “de-aerated” water, extremely
clean smooth surfaces, undergoing the vacuum system, cyclical application
of positive and negative pressures and pre-pressurising the system at high
pressures to dissolve the free air.
2 Description of the Equipment – Tensiometer

A traditional tensiometer usually consists of a porous stone in contact with the
soil, a body containing the porous stone, de-aerated water and a transducer (in
this case). The soil matrix suction is measured by excitation of the transducer.
The problems referring to cavitation that generally occur at approximately
80 kPa are very common to the traditional tensiometers mentioned above.
Such problems were solved in this new form of building tensiometers due to:
• the use of porous stone with a high air entry value;
• the use of “de-aerated” water throughout the saturation process and as-
sembly;
• the use of acrylic in constructing the tensiometer because of its smooth
surface that prevents the formation of micro air bubbles during the satu-
ration process of the system;
• the process of assembly, saturation and calibration of the prototype.
The equipment presented here consists only of these three elements –
porous stone, transducer and acrylic body.
This is extremely simple equipment. It uses the stone and transducer in
its original dimension. So this tensiometer can be tailor-made for each case.
Figures 7 and 8 presents respectively photographs of the prototypes de-
signed and used in this research and the acrylic capsule design with different
dimensions for each tensiometer used.
In the development of the current tensiometer prototypes, the following
porous stone and transducers were used (Table 1).
All prototypes presented satisfactory behaviour, the time-lag variation in
seconds, not significantly enough to negatively influence its use in laboratory
or field.
Fig. 7. Example of developed tensiometers – Tense EPX and Tense EPXO (Diene
2004)
Fig. 8. Design of the acrylic capsule prototype (Diene 2004)
Table 1. Dimensions and characteristics of the components of tensiometers (Diene

2004)
Tense- Tense- Tense- Tense-

Tensiometer EPX.1 ASH.1 EPXO.1 ASH.2
Diam. Thickn. Diam. Thickn. Diam. Thickn. Diam. Thickn.
Stone test test test test
Diam./Thickn. (mm) 15.85 7.55 15.6 7.05 29 7.20 29 7.20
Pressure (kPa) 1500 1500 500 500
In this research, other equipment for control and monitoring, such as TDR
and equitensiometers were used together with the tensiometers.
The equivalence tensiometer or equitensiometer consists of a Theta probe
and acrylic body, as shown in Fig. 9.
The equitensiometer sensor consists of a Theta probe built in a special
projected porous material (Fig. 9). The water content of this material enters
into equilibrium with the matrix potential of the soil involved, where it is
detected by the Theta probe, when absorbed. This operation is based on the
equivalence of the matrix potential between the soil material and instrument’s
body.
Fig. 9. Equivalence Tensiometer – Equitensiometer EQ2 (Diene 2004)
In principle, the EQ2 instrument response varies in a band from 0 to

−1000 kPa. The best precision occurs between −100 kPa and −1000 kPa, with
± 5% error. Its reading precision in suctions from 0 to −100 kPa is ± 10 kPa.
Calibration and Saturation Process
The saturation and calibration process comprised the following actions and
equipment (Fig. 10):
1. Check the inclusion of the porous stone in the acrylic body by applying
three water pressure cycles, to check the water tightness of the contact
(porous stone-acrylic);
2. The porous stone was saturated by applying a vacuum in the calibration
chamber, which has no water content inside it for longer than 15 hours.
Later, “air bubble free” water is introduced in the chamber (maintaining
the vacuum application) until the water covers the porous element. The
vacuum in the system should be maintained for another two hours;
3. Transfer the assembled set to the calibration/saturation chamber and ap-
ply pressure cycles varying from zero to 600 kPa, following the value of
the bubbling pressure of the porous stone used to generate a water flow
into the stone and remove any bubbles;
4. Install the pressure transducer in the system by screwing it to the acrylic
body in “air bubble free” water and re-apply a vacuum to the system for
three to four hours;
5. Calibrate the prototype using pots of mercury in three loading and un-
loading stages, with water pressure values until it reaches the calibration
Fig. 10. Calibration/saturation chamber and porous stone saturation chamber

(Diene 2004)
linear curve (these water pressure values vary according to the installed
prototype).
3 Results
Figures 11 and 12 showed some of the results obtained in lysimeter laboratory
tests (Diene 2004), compared with other automatic tensiometers and different
suction measuring systems such as equitensiometers.
The Tense-ASH1 and 2 tensiometers behaved in a similar manner until
reaching suction values of over −800 kPa, at which point a loss of pressure
in the Tense-ASH2 tensiometer was observed. This loss of pressure is due
to the fact that this tensiometer, with a porous stone of 500 kPa, reached
the air entry pressure, and the system achieved the air entry value limit.
Consequently, bubble air nuclei formation causes cavitation (Fig. 11). In this
case, cavitation might also be called maximum water-ceramic adhesion stress.
The Tense-EPXO1 tensiometer measured soil suction values with sufficient
precision and continuously until reaching higher values than air entry value
of 500 kPa of the porous stone attached to it.
The Tense-EPX1 tensiometer inserted to 30 cm in depth measured quite
similar suction values to those measured by the Tense-EPXO1 installed at a
depth of 15 cm. The tensiometer measured suction values of up to −1465 kPa
and when correlated to the Tense-EPXO inserted at a depth of 15 cm, showed
a difference of decreasing pressure of 15 to 25 kPa (Fig. 12).
PERFORMANCE OF TENSE ASH 1,ASH 2 &

EQ2A15
0,00
-200,00
SUCTION (kPa)
-400,00
-600,00 EQ2-A15
-800,00 TENSE-ASH1
-1000,00 TENSE-ASH2
-1200,00
-1400,00
-1600,00
20/mai
29/04 14:00:22
15/05 15:40:00
25/5/2003 16:01
30/5/2003 16:01
18/6/2003 20:38
22/6/2003 14:38
jul/14
6/jul
10/jul
18/jul
22/jul
27/jul
6/jun
11/jun
15/jun
27/06
30/06
TIME ( Days)
Fig. 11. Results of high suction tensiometers correlated with the equivalence ten-
siometer tested in lysimeters (Tank A) (Diene 2004)
PERFORMANCE OF TENSE EPXO1, EPX1 & EQ2-B15
0
-200
SUCTION (kPa)
-400 EQ2-B15
-600
-800 TENSE-EPXO1
-1000 TENSE-EPX1
-1200
-1400
-1600
21/05 14:06:28
27/5/2003 16:01
18/6/2003 20:38
23/6/2003 02:38
9/5/2003 14:41
5/8/2003 06:18
16/mai
28/06
jul/13
17/jul
21/jul
27/jul
3/jul
8/jul
10/jun
14/jun
3/jun
1/ago
TIME (DAYS)
Fig. 12. Results of high suctions tensiometers correlated to the equivalence ten-
siometer tested in the lysimeter (Tank B) (Diene 2004)
4 Final Comments
The following conclusions can be made from the developed prototypes and
their results:
1. With the manufacturing progress of porous stones and transducers there

are no longer problems in developing tensiometers to measure high suction
values;
2. The acrylic should be as smooth as possible, in order to prevent the ac-
commodation of any micro-bubbles in irregularities of the walls;
3. The variation band of suction measurement to be achieved will be de-
termined by the (air entry pressure) porous stone used and transducer
capacity;
4. The obtained results show that the developed prototype may be employed
in onsite measurements or in the laboratory with its adjustment, for ex-
ample, to the cells of triaxial tests or odometer;
5. This is robust equipment. It is able to use the porous stone and transducer
separately in other installations due to the easy handling of the proposed
procedure.
Acknowledgements
The authors thank the National Research Council (CNPq), CAPES, Foun-
dation Volkswagen and the Institute for Environmental System Research of
the University of Osnabrück, Professor M. Matthies, for their constant sup-
port and interchange. They thank Ms. Elvyn Marshall for the revision of the
English text.
References
Diene AA (2004) Tensiometers development for high suction measurement in labo-
ratory lysimeters. MSc Thesis, Coppe-UFRJ, 173 pp (in Portuguese)
r
Entran Pressure Sensors (2000) EPX Miniature Threaded Pressure Sensors, www.
entran.com
Guan Y, Fredlund DG (1997) Direct measurement of high soil suction. In: Simpósio
solos não saturados brasileiro, 3, Rio de Janeiro, 2:543–550
Harvey EN, Barnes DK, McElroy WD, Whiteley AH, Pease DC, Cooper KW (1944)
Bubble formation in animals. I. Physical factors, J Cell Comp Physiol 24(1):1–22.
König D, Jessberger HL, Bolton MD, Phillips R, Bagge G, Renzi R, Garnier J (1994)
Pore pressure measurement during centrifuge model test: experience of five lab-
oratories. In: Leung, Lee, Tan (eds) Centrifuge’94. Balkema, Rotterdam:101–108
Mahler CF, Pacheco AC, Souza HG (2002) Development of an automatic tensiometer
in laboratory using a Mini-Lysimeter. In: 3rd Int Conf on Unsaturated Soils,
REcife, Brazil
Marinho FAM, Chandler RJ (1995) Cavitation and the Direct Measurement of Soil
Suction. In: First Int Conf on Unsaturated Soils, Paris, France 2:623–630
Pacheco AC (2001) Development of a new tensiometer to measure suctions higher
than one Atm. MSc Thesis, Coppe-UFRJ 93 pp (in Portuguese)
Ridley AM (1993) The measurement of soil moisture suction. PhD Thesis, University
of London, Civil Eng Dept
Ridley AM (1995) Strength–suction–moisture content relationship for kaolin under

normal atmospheric conditions. In: First Int Conf on Unsaturated Soils Unsat’95,
Paris 2:645–651
Ridley AM, Burland JB (1993) A new instrument for measuring soil moisture suc-
tion. In: Technical Note, Geotechnique 43(2):321–324
Ridley AM, Burland JB (1995) Mesasurement of suction in materials which swell,
Appl Mech Rev 48(10):727–732
Tabor D (1979) Gases, liquids and solids, 2nd ed. Cambridge University Press
Take WA, Bolton MD (2002) A Negative Devices for the Measurements of Negative
Pore Water Pressure in Centrifuge Models, Physical Modeling On Geotecnics:
89–94
Tarantino A, Bosco G, Mongiovi L (2000) Response of the IC Tensiometer With
Respect to Cavitation. Unsaturated Soils for Asia:309–313
Tarantino A, Mongiovi L (2002) Design and Construction of a Tensiometer for Direct
Measurement of Matric Suction. Unsaturated Soils:319–334
Part III
Strength and Dilatancy

Dilatancy of Coarse Granular Aggregates
Eduardo E. Alonso, Enrique F. Ortega Iturralde, and Enrique E. Romero
Department of Geotechnical Engineering and Geosciences, Building D2,

Technical University of Catalunya, Jordi Girona 1–3, 08034 Barcelona, Spain
eduardo.alonso@upc.edu, enrique.francisco.ortega@upc.edu,
enrique.romero-morales@upc.edu
Summary. Coarse granular materials such as rockfill have been described in recent
years by means of strain and work hardening plasticity models Oldecop and Alonso
(2001), Chávez and Alonso (2003), Alonso et al. (2005) which require a proper
definition of the flow rule. The paper presents experimental information on dila-
tancy rates obtained in suction controlled triaxial tests performed on large diameter
(25 cm) specimens of compacted coarse gravels of hard limestone. Tests involved con-
fining stresses in the range 0.3–1 MPa and Relative Humidity in the range 10–100%.
Unloading-reloading during the tests provided information on the elastic response.
Volumetric and deviatoric plastic strains rates were determined from records of total
vertical and volumetric deformations. In general, a steady (critical) state was not
found. Plots have been prepared showing the variation of the plastic flow direction,
d, with a number of variables: vertical deformation, stress ratio, η, confining stress,
p, plastic work, W p , suction, s, and some derived normalized variables. It was found
that the common framework of dilatancy being a function of the current stress ratio
and the limiting stress ratio is far from being appropriate to describe the behaviour
of the tested aggregates. It was found that relationships d = f (ηW p /p, s), provide
a good description of test results. The paper discusses the rationale behind this
selection.
Key words: dilatancy, granular material, triaxial test, suction, relative humidity
Background
Information on dilatancy during shear is a fundamental contribution towards
understanding stress-strain behaviour. It is also required to build elastoplastic
constitutive models. Coarse granular aggregates differ from granular soils in
one fundamental aspect: They are able to fracture under stress states com-
mon in engineering practice. The breakage of particles results in an enhanced
compressibility under confining stress. The breakage of particles is not only
controlled by the confining stress but also by the prevailing Relative Humidity
(RH) or, alternatively, the total suction.
120 E. Alonso et al.
Oldecop and Alonso (2001) investigated the compressibility of rockfill by

performing large scale oedometer tests under controlled RH. The material
tested, a compacted gravel of quartzitic shale (Pancrudo shale), exhibited
a substantial compressibility under K0 conditions. Irreversible strains dom-
inated the total strains ever since the application of the first loading incre-
ments. It was also shown that increasingly dry atmospheres resulted in a sub-
stantial reduction of compressibility. Oldecop and Alonso (2001) explained this
macroscopic behaviour from the perspective of fracture mechanics concepts
applied to the mechanisms of crack propagation in individual rock particles.
Increasing suction retards the propagation of cracking in stressed particles and
leads to an overall stiffer response. Likewise, a suction reduction accelerates
the fracture propagation and this may lead, under appropriate stress condi-
tions, to rapid compression phenomena, identified as collapse. No capillary
action is invoked in this explanation.
Because of these effects, it turns out that modeling rockfill materials in
hardening plasticity requires a “cap” to account for the relevant irreversible
deformations in compression. The model developed by Chávez and Alonso
(2003) is an example. It was based on a series of large scale suction controlled
triaxial tests performed on compacted specimens of Pancrudo shale.
A fundamental reference for materials which dissipate energy (in compres-
sion and shear) in a frictional mode is Cam Clay (Schofield and Wroth 1968).
In the original associated model, dilatancy is simply expressed as:
ε̇pv
d= =M −η (1)
ε̇ps
where ε̇pv , ε̇ps are the plastic rates of volumetric and deviatoric strains, εpv =
εp1 + 2εp3 , εps = 2/3(εp1 − εp3 ) and η = q/p ; q = σ1 − σ3 ; p = σ1 + 2σ3 .
If equation (1) is integrated, the plastic potential of Cam Clay is obtained.
If the plastic flow is associated, integration of (1) leads to the yield function
(in triaxial stress states). Equation (1) is plotted in Fig. 1. For a stress ratio
η = 0, (isotropic compression) Cam Clay predicts a finite dilatancy (d = M ).
This is not the behaviour observed in soils since isotropic loading leads mainly
to volumetric deformations, which implies d → ∞. On the other hand, in Cam
Clay the rate of change of d with η is 1. Lagioia et al. (1996), aware of these
limitations, proposed a more flexible dilatancy law:

αM
d = μ(M − η) +1 (2)
η
where α and μ are parameters. When η tends to zero, d tends to infinity and
the plastic potential becomes rounded at its apex. The shape of the dilatancy
relationship given by equation (2) is also indicated in Fig. 1 for a particular
set of parameters. The authors showed that a few popular elastoplastic models
(Modified Cam Clay – Roscoe and Burland 1968; Sinfonietta Classica – Nova
1988, 1991; the single hardening models of Kim and Lade 1988 and Lade and
Dilatancy of Coarse Granular Aggregates 121
Fig. 1. Dilatancy in Cam Clay and Lagioia et al. (1996) models
Kim 1988) are well reproduced by particular combinations of parameters of

equation (2).
Rowe (1962) analyzed the dilatancy of granular materials and the effect
of applied stresses. He reached the conclusion, later supported by De Josselin
de Jong (1976), that the dilatancy parameter d could be written as:

σ1 1
d=1− (3)
σ3 Kcr
where Kcr = tan2 (45 + φcr /2) and φcr is the critical state angle. The principal
stress ratio may be expressed as σ1 /σ3 = (3 + 2η)/(3 − η). Therefore, Rowe’s
equation, as well as equations (1) and (2), relates dilatancy with the current
stress ratio, η, and the limiting value, M (or φcr ).
In none of the equations (1), (2) and (3) there is an explicit considera-
tion on the density of the material, or, alternatively, on the confining stress.
Different initial densities are implicitly considered as different materials.
Wan and Guo (1998) modified Rowe’s expression and introduced a pa-
α
rameter, (e/ecrit ) where e is the current void ratio, ecrit is the corresponding
void ratio at critical state and α is a constant, to take density into account.
Accepting this proposal, Chávez (2004) and Chávez and Alonso (2003) de-
scribed the results of suction controlled triaxial tests on Pancrudo shale and
examined the effect of suction on dilatancy. Some results are reproduced in
Fig. 2.
Note that, when D = 1 − d = 1, a zero dilatancy (critical state conditions)
are reached. Tests represented in Fig. 2 did not reach critical state conditions
Fig. 2. Dilatancy observed in four triaxial tests performed at 0.1, 0.3, 0.5 and
0.8 MPa at a constant relative humidity of RH = 36% (Chávez 2004)
in all cases but they were close enough. Results were interpreted with the help
of a modified Wan and Guo’s equation, introducing suction as an additional
controlling variable and taking critical state conditions as a reference.
Li and Dafalias (2000) argue that classical dilatancy theories do not include
the static and kinematic constraints at particle contacts which in practical
terms imply that density and stress levels effects are not properly included.
To overcome these limitations they formulate dilatancy in terms of a state
parameter ψ which expresses the difference between the current void ratio
and the critical state void ratio:
η
d = d1 emψ − (4)
M
where d1 , m, and M are material parameters. Chiu and Ng (2003) adopt this
formulation for the “shear” yield of their constitutive model for unsaturated
soils simply recognizing that material parameters (and the critical state line)
depend on suction. Suction now becomes an additional state parameter. They
define an additional law for the “cap” yield function of their model, making
sure that d = ∞ when η = 0. Cecconi et al. (2002) also criticize, on the
basis of experimental observations, the one-to-one relationship between d and
η implied by the classic models of dilatancy. Their study focuses on grain
crushing of pyroclastic soils and its implications. They interpret that grain
crushing leads to a progressive evolution of material parameters such as M
and m in equation (2). In their elastoplastic model degradation induced by
particle crushing occurs as plastic strains accumulate.
Dilatancy data in connection with suction controlled tests in soils has
been reported in several papers. Some results of the triaxial tests reported
by Cui and Delage (1996) on compacted Jossigny silt are shown in Fig. 3.
Fig. 3. Triaxial tests by Cui and Delage (1996) on compacted Jossigny silt. Measured
dilatancy for a constant η = 1 value. Effect of confining stress and suction
For a constant η = 1 value, a significant effect of confining stress and suction

was measured. Dilatancy decreases as the mean stress increases and suction
decreases. Later, Cui and Delage (1998) proposed a simple dilatancy equation
d = μη + α (5)
where α depends on confining cell pressure. Equation (5), however, fails to

reproduce the infinite d values associated with isotropic loading. Parameter μ
has found to depend linearly on suction.
Ng and Zhou (2005) performed suction controlled direct shear tests on a
compacted residual soil from granite and found that the maximum dilatancy
depended strongly on suction and density. Further suction controlled tests
on expansive clay were reported by Zhan and Ng (2006). Direct shear tests
are not very useful to derive general relationships for dilatancy, but they also
indicate that suction tends to increase negative dilatancy rates.
Cattoni et al. (2005) have recently reported the results of suction con-
trolled triaxial tests on a silty sand. In their analysis the confining stress p
also includes suction effects through a Bishop type of effective stress. Their
dilatancy vs. η plots show also the role of p to inhibit dilatancy. They also
found that dilatancy in bonded soils only develop significantly once the peak
strength is reached.
Plastic flow rules are therefore more complicated than suggested by equa-
tions (1), (2), (3) or (4). In this paper, dilatancy measurements performed
on large scale suction triaxial tests performed on a hard limestone gravel
are reported. Pancrudo shale (Fig. 2), tested previously, was a low quality
rockfill, currently being used to build the shoulders of an earth and rockfill
dam. There was an interest, however, in testing a harder material, of very
low porosity, probably not inter-connected. This material is generally used as
ballast in railway construction and maintenance. Capillary effects on this rock
are unimportant and therefore suction induced effects are related to particle
breakage and crack propagation phenomena.
Suction Controlled Triaxial Tests on Hard Limestone

Gravel
The raw material is a gravel having sizes in the range 40–1.2 mm. Figure 4
shows a photograph of one of the specimens tested. Some identification rock
parameters are given in Table 1.
Tests were performed in a double wall triaxial cell described in Chávez
(2004) and Chávez et al. (2005). Relative humidity inside the specimens was
maintained by means of a flow of moist air. Relative humidity was measured
by means of a capacitive hygrometer located in the circuit of moist air at a
point close to the specimen. Volume change was measured by local transducers
by means of three rings located at three different horizontal sections. The
instantaneous sample volume was calculated fitting a smooth interpolation
through measurements. Independently, volume change was measured by a
differential pressure transducer monitoring the excess water level inside the
Fig. 4. Limestone gravel tested
Table 1. Identification parameters of limestone gravel
Unconfined compression
Unit weigth Specific gravity of solids Void ratio
strength
17.73 kN/m3 2.7 0.544–0.548 150 MPa
Fig. 5. Stress paths and plastic strain increment vectors in triaxial tests performed
at RH = 50%
inner chamber. Specimens were compacted in six layers under an energy of

compaction equivalent to the Standard Proctor. Neoprene membranes had a
special inner coating to reduce friction at the particle-membrane contacts.
Vertical deformations and vertical loading were also measured inside the cell.
Figure 5 shows the stress paths followed and the estimated nonlinear
strength envelope for the tests performed at RH = 50%. Plastic strains de-
velop in compacted gravel specimens during the entire stress path. Plastic
strain vectors have been plotted in Fig. 5, during the deviatoric loading. It
can be noticed that (negative) dilatancy develops only in the proximity of
limiting strength conditions.
Figure 6 shows the stress-strain-volumetric behaviour of tests performed
at a relative humidity of 50% and varying confining stress. Additional results
for RH = 10% are shown in Fig. 7. At some stage during these tests specimens
were wetted and the RH was increased to 100% (without changing the vertical
deformation rate). The volumetric deformation rate changed in the sense of
increasing the compression rate (collapse) and reducing the dilatancy rate. As
a result, the deviatoric stress decreased suddenly but it later recovered again
as the imposed vertical deformation increased. The final limiting strength
corresponds to saturated conditions (RH = 100%).
Figures 6 and 7 show that specimens compress volumetrically during most
of the loading curve. Dilatancy develops when the specimen is close to limiting
conditions. It is worth noting also that triaxial tests on gravel specimens show
a ductile behaviour, even if the “soil” has been well compacted. There are
at least two reasons for this behaviour: the continuous breakage of particles
during deformation and the relative dimensions of specimen size and par-
Fig. 6. Suction controlled triaxial tests on compacted limestone gravel. RH = 50%.

Deviatoric-vertical strain curves (left) and volumetric-vertical strain curves (right)
Fig. 7. Suction controlled triaxial tests on compacted limestone gravel. RH = 10%.

Deviatoric stress-vertical strain curves (left) and volumetric-vertical strain curves
(right). RH was increased to 100% at an advanced state of deformation
ticle dimensions which prevent strain localization. Note also that dilatancy
maintains relatively high rates at the end of the test, at large vertical strains.
In practically no case of the set of triaxial tests performed a condition close to
critical state could be found. This is illustrated in Fig. 8, which shows the void
ratio-mean stress paths of the three tests performed at RH = 50%. The plot
shows the volumetric compression observed during most of the shear stage
and the rapid development of dilatancy at the end of the test. However, the
tests are far from reaching a constant volume condition. Therefore, no critical
Fig. 8. Suction controlled triaxial tests on compacted limestone gravel at RH = 50%.

Specific volume-mean stress during triaxial shearing
state condition could be defined in this case. These results were essentially
repeated when other tests under different RH’s (10%, 100%) were performed.
In the more compressible Pancrudo shale, in which the elongated particles
broke more easily, critical state conditions could be found in triaxial tests
performed in the same equipment (see Fig. 2, Chávez 2004). The harder nature
of the limestone particles is apparently the reason for the change in behaviour.
Presumably, if tested in much larger equipment so that the ratio of specimen
diameter to particle size increases dramatically, the limestone gravels would
exhibit a different behaviour. This result prevents the reference to critical
state when describing the dilatancy of the tested gravel.
Dilatancy
Consider first simple dilatancy plots in which d is represented against the
axial deformation. Figure 9 shows such a plot for two series of tests performed
under RH = 50% and 10%, respectively. In the second series of tests, RH
was increased to 100% when the vertical deformation was high (more than
8%). The zero reference for the axial deformation is the beginning of the
deviatoric load application. As expected, dilatancy starts at a high positive
value and decreases rapidly as deformation increases. However, negative values
are reached after a large accumulation of axial strains. The confining stress
controls also the dilatancy rate: low confining stresses lead to higher dilatancy
rates.
Fig. 9. Dilatancy measured in triaxial tests on compacted limestone gravel at RH =

50% (left) and RH = 10% (RH was increased to 100% at an advanced state of
deformation) (right)
For a more fundamental appraisal, following traditional understanding,

plots of dilatancy rate d in terms of the stress ratio η/M have been prepared
for the set of suction controlled triaxial tests performed (Fig. 10). Due to
the curvature of the strength envelope, the value of M varies with the par-
ticular test considered. Dilatancy rate, having initially a positive high value,
decreases with stress ratio and reaches a zero value for stress states close to
shear strength in most cases. The strong effect of confining stress dominates
the plot. Increasing confining stress decreases the (positive) dilation rate. In
comparison, the effect of RH for this hard gravel is less marked.
However, this type of plot is not very useful to interpret the negative di-
latancy stage because dilation accumulates at an essentially constant η value.
This is especially apparent when one considers the normalized η/M value in
Fig. 10. Dilatancy is still active at the end of tests but it increases at a con-
stant η/M = 1 value. All tests performed, irrespective of confining stress and
suction, are superimposed at the end of the plot shown in Fig. 10.
An attempt has been made to introduce a state parameter, the void ratio,
to explain the dilatancy rates observed. However, the initial void ratio of
specimens tested, before the application of the deviatoric loading, was very
similar because of the low isotropic compressibility of the material. During the
test (Figure 11 for RH=50%) void ratio decreases first and then increases. For
a given e (and confining stress), two dilatancy rates (positive and negative)
may be found. If the difficulties to find a critical state in this material are
added, it is concluded that void ratio does not help in this case to model the
measured results.
In a coarse aggregate, deformations integrate two phenomena: the break-
age of contacts and the particles themselves and the re-arrangement of the
granular structure by means of rotations and displacements of particles. The
Fig. 10. Dilatancy rate, d, vs. stress ratio, η, and normalized stress ratio, η/M , for
the set of triaxial tests performed on limestone gravel
positive dilatancy rates measured during the loading process are an indication
of the dominant effect of contact and particle breakage during the increase of
stress ratio. On the other hand, the dominant effect of confining stress and ver-
tical deformation in all cases suggests that the work input into the specimen
may be a suitable variable to integrate the stress and deformation processes
taking place in the gravel. The plastic work spent during the tests, W p , has
Fig. 11. Dilatancy rate, d, vs. current void ratio and confining stress for triaxial
tests performed at RH = 50%
been calculated along the stress paths. Dilatancy was initially plotted against
W p for the different tests performed, but the effect of the current p value
was still significant. Further regularization is achieved if the plastic work is
normalized with respect to the mean stress (Fig. 12).
Fig. 12. Dilatancy rate, d, vs. normalized plastic work, W p /p, and confining stress
for triaxial tests performed at RH = 100% and RH = 50%
However, dilatancy cannot be made dependent only on the plastic work.

In fact, in isotropic paths the accumulation of plastic work will lead towards
a continuous reduction of dilatancy, if the plots in Fig. 12 are accepted as a
general framework. It is clear that under isotropic loading, irrespective of the
work input into the specimen, the dilatancy rate should maintain an infinite
value (only plastic volumetric deformations are induced).
These considerations led to the proposal of dilatancy relations having the
form
d = f (ηW p /p, s) (6)
where s is the total suction. Dilatancy has been plotted against the variable
ηW p /p in Fig. 13 for triaxial tests performed at RH = 50% and RH = 10%.
All the tests performed seem now to plot into a common dilatancy relation-
ship. If only isotropic stress paths are applied, η = 0, and irrespective of the
work input into the aggregate, (positive) dilatancy maintains an infinite value.
When η increases, dilatancy tends towards negative values but it is necessary
to reach relatively high plastic work input values to really induce negative
dilatancy. The effect of confining stress, in the sense of inhibiting dilatancy as
it increases, is also included in the dimensionless parameter ηW p /p.
Plots in Fig. 13 suggest that empirical dilatancy equations of the type:
2
β
d= α+ − β2 (7)
(ηW p /p)2
where α and β are constant parameters may reproduce the measured dila-
tancy. A comparison of calculations using equation (7) and measurements is
given in Fig. 14 for two series of triaxial tests performed at RH = 100% and
50%, respectively.
Fig. 13. Dilatancy rate, d, vs. normalized variable, ηW p /p, for triaxial tests per-
formed at RH = 50% and RH = 10%
Fig. 14. Comparison of measured dilatancy and prediction through equation (7)
for triaxial tests at RH = 100% and 50%, respectively
The effect of suction on dilatancy is small in the hard limestone tested.

Tests performed at the same confining stress but different suctions are com-
pared in Fig. 15. Despite the limited influence, when suction increases (neg-
ative) dilatancy also increases. However, for the extreme range of suctions
tested (RH = 10% is equivalent to a total suction of 320 MPa) dilatancy
maintains similar values.
The result that negative dilatancy tends to be higher (in absolute terms)
when suction is high has been found when shearing unsaturated soils (Cui
and Delage 1996, Ng and Zhou 2005, Cattoni et al. 2005). A similar result
was also found by Chávez and Alonso (2003) and Chávez (2004) when testing
specimens of compacted gravel of Pancrudo shale. The explanation for this
Fig. 15. Effect of Relative Humidity on dilatancy rate for two series of triaxial tests
(σ3 = 0.3 and 1.0 MPa, respectively)
Fig. 16. Variation of measured dilatancy rates with Relative Humidity for a constant
value of the controlling parameter ηW p /p
behaviour, in the case of granular aggregates, is that low RH’s makes stronger
particles which, in turn, favour dilation. However, in the case of the hard
aggregates tested, small variations in specimen density or in the particular
arrangement of grains may be as significant as the applied suction. The effect
of suction on the hard limestone gravel tested may be observed in more detail
in Fig. 16, which provides the variation of d with suction for a constant value
of the controlling variable ηW p /p. With the exception of the test performed
at σ3 = 0.5 MPa, the data indicates an increase of (negative) dilation rate
with increasing suction. Equation (7) may be used to take this small influence
into account simply by making α and β functions of suction.
Summary and Conclusions
Large diameter suction controlled triaxial tests have been performed on com-
pacted specimens of limestone gravel. The paper has focused on the develop-
ment of dilatancy during shearing.
Several conclusions are reached:
Dilatancy was still active at the end of the tests performed, when axial
deformations were high (in excess of 15%). Critical state conditions could not
be found.
The common framework of dilatancy being a function of the current stress
ratio and the limiting stress ratio is far from being appropriate to describe
the behaviour of the tested aggregates. In particular, most of the observed
dilation takes place at constant stress ratio, very close or at the limiting state.
For this reason, the variable η has limited discriminating capacity to describe
dilatancy.
It has been found that the plastic work input into the specimens as well as
the average stress should be introduced to describe the measured dilatancy for
the range of stress ratios and confining pressures applied in the tests. Plastic
work may explain the amount of contact and particle breakages that occur
within the soil, a phenomena which is believed to control, to a large extent,
dilatancy.
Dilatancy may be described by a variable which combines stress ratio,
plastic work and mean stress. Relationships of the form
d = f (ηW p /p, s)
provide a reasonably good description of the tests performed.

Suction has a limited effect on the measured dilatancy. In any case, its
effect, although noticeable and, in general terms consistent with observations
in other suction controlled tests in soils and weaker gravels, is substantially
smaller than the rest of variables mentioned.
Acknowledgements
The authors wish to acknowledge the financial support provided by the Con-
sejo Nacional de Ciencia y Tecnologia (CONACYT) of México to the second
author during the performance of the research described in this paper.
References
Alonso EE, Olivella S, Pinyol NM (2005) A review of Beliche Dam, Géotechnique
55(4):267–285
Cattoni E, Cecconi M, Jommi C (2005) Soil dilatancy and suction: some remarks
on their mutual effects on the shear strength of granular soils. Proceedings of
the 11th International Conference on Computers Methods and Advances in Ge-
omechanics, Torino, Italy:19–26
Cecconi M, DeSimone A, Tamagnni C, Viggiani MBG (2002) A constitutive model
for granular materials with grain crushing and its application to a pyroclastic soil,
International Journal for Numerical and Analytical Methods in Geomechanics
26:1531–1560
Chávez C, Alonso EE (2003) A constitutive model for crushed granular aggregates
which include suction effects, Soils and Foundations 43(4):215–227
Chávez C (2004) Estudio del comportamiento triaxial de materiales granulares de
tamaño medio; con énfasis en la influencia de la succión. Tesis doctoral Univer-
sidad Politécnica de Cataluña, España
Chávez C, Romero E, Alonso EE (2005) Volume change measurement of partially

saturated rockfill in triaxial test, Proceedings International Symposium on Ad-
vanced Experimental Unsaturated Soil Mechanics, Trento, Italy. Taylor & Fran-
cis Group, London:93–98
Chiu CF, Ng CWW (2003) A state-dependent elasto-plastic model for saturated
and unsaturated soils, Géotechnique 53(9):809–829
Cui YJ, Delage P (1996) Yielding and plastic behaviour of an unsaturated compacted
silt, Géotechinique 46(2):291–311
Cui YJ, Delage P (1998) Plastic flow of an unsaturated compacted silt, Proceedings
of de 2nd international conference on unsaturated soils. Beijing, China vol. 1:467–
472
De Josselin de Jong G. (1976) Rowe’s stress–dilatancy relation based on friction,
Kim MK, Lade PV (1988) Single Hardening constitutive model for frictional mate-
rials. I. Plastic potential function, Computers and Geotechnics 5:307–324
Lade PV, Kim MK (1988) Single hardening constitutive model for frictional mate-
rials. II. Yield criterion and plastic work contours, Computers and Geotechnics
6:13–29
Lagioia R, Puzrin AM, Potts DM (1996) A new Versatile expression for yield and
plastic potential surfaces, Computers and Geotechnics 19(3):171–191
Li XS, Dafalias YF (2000) Dilatancy for cohesionless soils, Géotechnique 50(4):449–
460
Ng CWW, Zhou RZB (2005) Effects ofsoil suction on dilatancy of an unsaturated
soil, Proceedings 16th International Conference on Soil Mechanics and Geotech-
nical Engineering. Asaka, Japan, Millpress, Rotterdam, vol. 2:559–562
Nova R (1991) A note on sand liquefaction and soil stability, Conf Constitutive Laws
for Engineering Materials. ASME Press, Tucson, AZ
Nova R (1988) Sinfonietta classica: an exercise on classical soil modeling, Proc Int
Symp Constitutive Eq for Granular Non-Cohesive Soils. Balkema-Rotterdam,
Cleveland:501–520
Oldecop L, Alonso EE (2001) A model for rockfill compressibility, Géotechnique
51(2):127–140
Oldecop L (2000) Compresibilidad de escolleras influencia de la humedad. Tesis
doctoral Universidad Politécnica de Cataluña, España
Roscoe KH, Burland JB (1968) On the generalised stress strain behaviour of “Wet”
clay. Engineering plasticity, Cambridge University Press, Cambridge:535–609
Rowe PW (1962) The stress-dilatancy relation for static equilibrium of an essembly
of particles in contact. Proc Roy Soc London A269:500–527
Schofield AN, Wroth CP (1968) Critical state soil mechanics. McGraw Hill, London
Wan RG, Guo PJ (1998) A simple constitutive model for granular soil: Modified
stress–dilatancy approach, Computers and Geotechnics 22(2):109–133
Zhan TLT, Ng CWW (2006) Shear strength characteristics of an unsaturated ex-
pansive clay, Canadian Geotechnical Journal 43:751–763
A Laboratory Investigation into the Effect of
Water Content on the CBR of a Subgrade soil
Samuel Innocent Kofi Ampadu
Kwame Nkrumah University of Science and Technology, Kumasi, Ghana

skampadu.soe@knust.edu.gh, sikampadu@yahoo.co.uk
Summary. Residual soils, most of which are lateritic serve as subgrade and even
sub-base and base layers for road and highway pavements in the subregion. However,
the material is known to undergo substantial strength reduction when they become
saturated with water. An understanding of the dependence of the CBR strength of
local soils on water content will contribute towards better design and maintenance
practices. Samples of soil from a study site were prepared by laboratory compaction
at the optimum water content using different levels of compaction to obtain samples
at different densities. The remoulded samples were then subjected to different levels
of wetting in a water tank and different degrees of drying in the laboratory and the
CBR determined. The variation of the CBR with the water content is presented and
discussed and related to the matric suction.
Key words: CBR, subgrade, matric suction, water content, remoulded sample,
compaction
Introduction
The road pavement structure in many developing countries consists of rela-

tively thin sub-base and base layers made of lateritic gravel with very thin wa-
ter proofing surfacing, founded on a residual soil subgrade. The residual soils
which are the decomposition products of the local geology occur at various
degrees of laterization (Charman 1995) and are known to undergo substantial
strength loss on soaking (Ampadu 2006).
The most common parameter used to evaluate pavement layer strength
is the California Bearing Ratio (CBR). Even though the CBR is not a fun-
damental soil property, its significance lies in the fact that it is the basis of
the CBR Method of Pavement design which is still, by far, the most popular
pavement design method used in developing countries. The details of the CBR
test are covered in ASTM D 1883-91. The CBR value is influenced by the wa-
ter content and the dry density as well as the texture of the soil. Normally,
the CBR test in the laboratory is conducted on test samples prepared at the
138 Samuel Innocent Kofi Ampadu
dry density and water content likely to be achieved in the field. Whereas the
field dry density can be fairly well predicted the difficulty is to determine
the stable moisture content at which to conduct the test. The local practice
which is also used in many other countries is to use the 4-day soaked CBR. In
other countries like the UK the design CBR is the CBR corresponding to the
equilibrium water content. Recently, there have been attempts to interpret
the results of CBR test in terms of concepts of unsaturated soil mechanics
(Sanchez-Leal 2002).
The influence of the moulding water content on the CBR of local lateritic
soils has been studied by Hammond (1970) This investigation seeks to con-
tribute further towards a better understanding of the effect of water content
on the CBR of local soils. Samples of a subgrade material were prepared by
compaction using three different compactive efforts and then subjected to
different conditions of drying and wetting. The CBR corresponding to each
condition was determined. The soil-water characteristics were also estimated.
The results of the CBR variation with the water content for the different dry
densities are presented and discussed and modelled in terms of the matric
suction.
Methodology
Bulk samples of decomposed granite were obtained from depths of 0.3 m to

about 0.6 m in a trial pit. The samples were air-dried for two days and the
index properties determined in accordance with BS 1377: Part 2:1990 using the
wet sieving method for the grading analysis. The compaction characteristics
were also obtained by ASTM D 1557-91 commonly referred to as the Modified
AASHTO compaction using 55 blows of the rammer.
Three different test series were conducted. For Test Series 1, using the
optimum moisture content (OMC) obtained from the compaction test, eight
samples were prepared in the CBR mould by ASTM D 1557-91 except that
5 blows of the rammer per layer were used. After preparation, the first sample,
designated “OMC”, was covered in a black polythene bag to prevent moisture
loss and stored for 24 hours to ensure moisture equilibriation. The next four
samples, designated “wetting” were soaked in water for 1, 2, 3 and 4 days
respectively while the last batch of three samples, designated “drying” was
subjected to various degrees of drying by allowing the samples to dry out for
4, 6 and 8 days respectively. After each period of soaking and of drying, the
samples were covered in black polythene bags for 24 hours to prevent further
moisture loss and to equilibrate the moisture distribution within the sample.
After that each sample was subjected to the CBR test in accordance with
ASTM D 1883-91. After each test, the water contents at the top, middle
and bottom of the specimen were determined. For Test Series 2 and Test
Series 3 the same procedure was repeated except that 20 blows and 55 blows
respectively of the rammer per layer were used in preparing the samples.
Effect of Water Content on the CBR 139
Discussion of Results
The index properties of the soil used for the investigation are shown in Table 1.
In this study gravel is defined as particles larger than 2 mm. The soil used in
this study may be described as a sandy clay of medium plasticity.
Table 1. Summary of Index Properties
Gravel (%) Sand (%) Silt (%) Clay (%) LL PI Gs OMC (%) MDD (Mg/m3 )
9 28 14 49 49 24 2.65 16.8 1.80
The summaries of results are shown in Tables 2 to 4 for Test Series 1 to 3,

respectively. The initial states of the samples for the CBR tests are plotted on
the compaction characteristics in Fig. 1. The dry densities refer to the values
achieved after sample preparation which means that any effects of volumet-
ric changes arising from drying and from wetting have been neglected. The
mean dry densities achieved were 1.355, 1.649 and 1.706 Mg/m3 with stan-
dard deviations of 0.013, 0.047 and 0.028 Mg/m3 for Test Series 1, 2 and 3
corresponding to 75%, 92% and 95% level of compaction respectively based
on the maximum dry density (MDD). The scatter in the dry density values
may be due to variations in the moulding water content during compaction. It
may be noted that in Test Series 3 the same 55 blows of the rammer achieved
only a 95% level of compaction. This observation is consistent with the results
of studies (Ampadu 1997) which show that at the optimum moisture content,
this material gives lower MDD values when fresh samples are used for com-
paction than when samples are reused for each test point. The results of the
water content distribution across the sample during testing showed that on
the average the difference in water content across the sample was less than
6% and for Test Series 2 and 3, drying appeared to produce a more uniformly
distributed water content across the sample than wetting. The water content
values in the Tables are the average values across the samples.
The variation of the CBR with water content for all three test series are
shown in Fig. 2. The similar values of CBR at higher water contents for both
Series 2 and Series 3 may be due to similar dry density values. The results
show that as the soil dries from the OMC condition, initially there is a rapid
increase in CBR but this slows down as the water content reduces further
especially for lower density samples. The rate of reduction in the CBR during
wetting is relatively slower. In fact the rate of change in CBR per percentage
change in water content was 3 to 7 times larger for drying than for wetting
from the OMC. The practical implication of this is that there is tremendous
benefit in terms of increase in CBR in ensuring that the subgrade is well
drained to achieve equilibrium water content below the optimum value. The
results also show that on soaking from the OMC condition, the CBR of the
Table 2. Summary of Test Results for Test Series 1
Average Water Dry Density Degree of Sat-

Test Condition CBR
Content (%) (Mg/m3 ) uration (%)
9.56 1.377 27.4 34
Drying 11.73 1.371 33.3 33
13.37 1.362 37.4 25
OMC 16.58 1.348 45.5 8
26.54 1.346 72.6 0.8
26.81 1.348 73.6 0.6
Wetting
27.03 1.354 74.8 0.4
28.88 1.338 78.0 0.3

Test Condition CBR
10.51 1.622 43.9 95
Drying 12.66 1.579 49.5 79
13.54 1.643 58.5 52
OMC 16.09 1.595 64.5 28
18.30 1.669 82.5 18
19.35 1.684 89.4 16
Wetting
20.22 1.703 96.4 16
21.26 1.694 99.8 15

Test Condition CBR
12.10 1.697 57.1 94
Drying 13.73 1.674 62.4 79
14.89 1.670 67.2 59
OMC 16.45 1.743 83.8 28
18.22 1.733 91.2 20
18.30 1.719 89.5 18
Wetting
18.59 1.726 92.0 17
21.63 1.688 100 15
subgrade drops and the relative reduction in CBR is between 46% and 98%
for dry densities ranging between 1.71 and 1.36 Mg/m3 .
The logarithm of the CBR is plotted against water content in Fig. 3.
The results quoted in Croney and Croney (1998) for soil B showing a linear
variation of log CBR with water content over the likely field water content are
superimposed. The limited results from this study do not show such a linear
Fig. 1. Compaction characteristics and initial states of test samples
Fig. 2. Variation of CBR with water content for constant dry densities
relationship and suggest that hysteresis effect from wetting or drying may be
important in defining the relationship.
The soil-water characteristics of the same material obtained from a parallel
test programme using the filter paper method (ASTM D-5298-03) are shown
in Fig. 4 for dry densities of 1.79 and 1.51 Mg/m3 . A linear relationship may be
assumed within the range of saturation investigated with an average air entry
Fig. 3. Variation of log CBR with water content
Fig. 4. Soil-Water characteristics for different densities
pressure of about 420 kPa. Based on this and interpolating and extrapolating
for dry densities used in this study, the corresponding matric suction values,
(ua − uw ), for all the test points were estimated. These are plotted against
the unsaturated CBR, CBRu on a log-log scale in Fig. 5. The results suggest
a linear log-log model which can be expressed as
Fig. 5. CBR-matric suction relationship
n
ua − uw
CBRu = CBRs ×
ue
where CBRs is the soaked CBR, ue is the air-entry value and n is a constant
which depends on the suction and the dry density.
For this study, n was of the order of 1.4 and about 0.5 for the lower and
for the higher dry densities respectively, and constant for suction values up to
about 15,000 kPa.
Conclusion
From the laboratory CBR test results on a subgrade material at different water
contents for three different dry densities, it may be concluded that the rate of
change in CBR per percentage change in water content during drying from the
OMC was 3 to 7 times larger than during wetting from OMC. Soaking from
the OMC condition, leads to a relative reduction in CBR of between 46% and
98% for dry densities ranging between 1.71 and 1.36 Mg/m3 . A linear log-log
relationship between CBR and matric suction is suggested for matric suction
values of up to about 15,000 kPa.
References
Ampadu SIK (1997) The compaction characteristics of residual soils, In: Proc 3rd
Int Conf on Structural Eng Analysis and Modeling, Accra, 7–11 September 1997
Ampadu SIK (2006) The loss of strength of an unsaturated local soil on soaking,
Geotechnical Symposium, University of Rome, March 16-17, 2006 (in print)
ASTM D 1557-91 Standard Test Method for Laboratory Compaction Characteristics
of Soil Using Modified Effort. Annual Book of ASTM Standards
ASTM D 1883-91 Standard Test Method for CBR of laboratory-compacted soils.
Annual Book of ASTM Standards
ASTM D 5298-03 Standard Test Method for Measurement of Soil Potential (Suction)
Using Filter Paper. Annual Book of ASTM Standards
BS 1377–1990 British Standard Methods of tests for Soils for civil engineering pur-
poses, British Standards Institution
Charman JH (1995) Laterite in road pavements. CIRIA Special Publication 47.
Transport Research Laboratory, Department of Transport
Croney D, Croney P (1998) Design and Performance of Road Pavements, 3rd Edn.
McGraw Hill, pp 137–138
Hammond AA (1970) A study of some laterite gravels from the Kumasi district.
Building and Road Research Institute, Project Report SM.5
Sanchez-Leal FJ (2002) Interpretation of CBR test results under the shear strength
concept of unsaturated soil mechanics. In: Proc 3rd Int Conf on unsaturated
soils (USAT 2002), Recife, pp 663–668
Shear Strength Affected by Suction Tension
in Unsaturated Fine Grained Soils?
Carola Bönsch and Christof Lempp
Martin-Luther-University, D-06099 Halle (S.), Germany

carola.boensch@geo.uni-halle.de, christof.lempp@geo.uni-halle.de
Summary. Suction tension acts within the gas and liquid filled grain-pore-system
in unsaturated fine grained soils and should partly control the shear strength. The
development of pore pressure in a completely water saturated grain structure will
reduce the shear resistance, whereas the suction tension within an unsaturated soil
will cause additional grain-to-grain forces. These are changing with varying degree
of water saturation and thus may influence the shear strength probably by favouring
the formation of coarse grained aggregates.
An experimental approach was started to investigate this expected relation in
overconsolidated Tertiary clay and in Quaternary loess (silt) using triaxial and di-
rect shear test results and comparing them with the results of suction tension mea-
surements obtained by using pressure membrane apparatus. Based on the common
classifications, one should argue that shear velocity will control the shear strength
due to pore pressure effects in the shear zone. However, the antagonistic effects of
overpressure of the pore fluid and compression of grains due to suction tension act
on a micro-scale in unsaturated fine grained soils and influence the shear behaviour
differently.
Effects of suction tension are noticeable in triaxial and direct shear tests in un-
saturated fine grained soils. Quantification is difficult because other stress influences
like load level act more effectively.
Key words: suction tension, water saturation, shear strength, loess, clay
1 Expected Effects of Suction on Strength

Due to the interaction between the solid, fluid and gaseous phases, additional
forces are effective in unsaturated pore systems. An unsaturated soil has the
tendency to receive water and increase its saturation level. If the soil is more or
less saturated, a certain force is required to remove water from the pore space
and decrease the saturation level of the soil. These effects can be explained
by the idea of suction tension.
Suction tension consists of both capillary and adsorptive parts. The capil-
lary part depends on the pore size distribution. The adsorptive part of suction
146 Carola Bönsch and Christof Lempp
tension is a function of the mineralogical composition of the solid part, as well

as the chemical composition of the pore fluid (Bönsch 2006).
Normally, the water content or the saturation level is plotted against the
logarithm of the tension in the pF curve. This, originally agricultural idea of
suction, can be used because the natural water content and the connected
suction tension of an unsaturated soil normally range in the same dimensions.
Water content-pF curves of fine grained soils can be simplified approxi-
mated by using a logarithmic regression of the following the form:
y = a + b · ln(x) (1)
The amount of the variable a represents approximately the saturation

water content. The term b · ln(x) describes the form of curve. More pre-
cisely, the flatter the curve, the smaller b · ln(x) is. The term b · ln(x) has
the range −4 < b < 0. y is the water content at suction x. The regres-
sion equations and some statistical parameters of the regression are given in
Table 2. Using the inverse function of that regression one can estimate the
suction tension corresponding to certain water content.
What are the explicit influences of the suction tension on strength and
stiffness of the unsaturated soil? If the effective friction angle and the effective
cohesion as well as the relation between suction tension and water content of
a soil are known, the shear strength could be specified using the theory of
Fredlund and Xing (1994). The shear stress is a function of net normal stress
(σn − ua ) and the suction s (Vanapilli et al. 1996). It can be written as:

θw − θr
τ = c + (σn − ua ) · tan ϕ + s · (tan ϕ ) · (2)
θs − θr
where θw are the volumetric water content during shearing, θs the volumetric
water content in case of saturation and θr the residual water content. There is
still no agreement to find in literature what water content that residual water
content is. It is not possible to measure the pore air pressure ua directly so
one can calculate it from the effective stress equation (Bishop 1959). This uses
the normal stress, the corresponding shear stress and the suction curve of the
soil:

c
ua = − − (u a − u w ) · χ − σ n (3)
tan ϕ
where the term (ua − uw ) is
(ua − uw ) = ψm . (4)
The suction tension s, as the equivalent of the matrix potential Ψm , could

not be measured directly. Therefore after knowing the water content, s can
be calculated from the inverse suction curve. The weighting factor χ depends
on pore air and pore water pressures as well as saturation level (Vanapilli et
al. 1996). Theoretically it is 0 in a saturated soil and 1 in a dry soil:
Shear Strength Affected by Suction Tension 147
θw − θr
χ= . (5)
θs − θr
Since the suction pressure s is equivalent to pore water pressure and nega-
tive, it can be added to the normal stress σn . Due to the fact that the suction
level of residual water content is unknown it will be neglected. Using the as-
sumption that the residual water content θr goes to 0, then χ is equivalent
to the saturation degree sR . Naturally the χ value will be smaller then sR ,
but its linear influence will not change the general trend of effective suction
neither it influences whether the calculated effective suction will be a pore
water pressure.
Actually the partial pressure of pore water and air are disregarded in
reference to the very slow shear velocity. In the assumption that the principle
of effective stresses is applicable also for unsaturated soils, and there is no
correlation between cohesion and suction tension, then the shear strength τ
using Eqs. (2), (3) and (5) can also be written as

c
τ = c + σn − − − s · s R − σ n · tan ϕ + s · tan ϕ · sR . (6)
tan ϕ
Then Eq. (6) can be transformed into
τ /2 − c − σn · tan ϕ = s · sR · tan ϕ . (7)
Finally the suction tension s can be estimated:
τ /2 − c − σn · tan ϕ
s= . (8)
tan ϕ · sR
One can expect that the shear strength will increase with an increasing
suction pressure, but theoretically there could be an influence from the shear
velocity. If it is too fast the reallocation of pore water is detained and there
could be a local increase of water content in the shear splice, which decreases
the suction based part of shear strength. Otherwise a brittle behaviour in-
creases the strength at fast shear tests, and creeping reduces the strength at
very slow shear tests. The real dimension of these influences are unknown as
long as the suction tension in the shear splice can not be measured. These
effects are similarly described by Junge et al. (2000) for tensile tests. As a
matter of fact the measured shear strength of an unsaturated soil is normally
smaller than determined by Eq. (2) (Behrens and Neumann 2002, Schanz et
al. 2004). Contrary to their opinion that neither suction tension nor shear
velocity play roles for shear strength, we make two assumptions. First, there
is a false estimation of the suction in shear splice, and second, in dependence
on velocity deficient consideration of brittle behaviour or creeping processes.
2 Characterisation of the Tested Materials

Different fine grained soils were tested from locations around Halle/Saale.
As example for clayey soils a slightly overconsolidated Tertiary clay from
Table 1. Characteristic properties (median of grain size distribution M , coefficient

of unconformity U , particle density ρs , porosity n, content of organic matter corg ,
total carbonate cC and dolomite cDol , saturated water permeability kf and specific
surface S)
M U= ρs n corg cC cDol kf S
Sample:
[μm] d60 /d10 [g/cm3 ] [%] [%] [%] [%] [m/s] [m2 /g]
Silt FB 25 20.1 2.65 32.3 2.81 1.54 0.39 1.6 × 10−8 13.5
Silt PL 29 3.5 2.65 32.9 4.89 6.79 1.33 4.0 × 10−8 13.2
Loess SM 32 5.4 2.71 38.1 0.28 20.7 7.45 5.9 × 10−7 14.2
Clay SP 1.2 — 2.64 37.4 7.38 unverifiable 2.2 × 10−9 27.1
Schiepzig (SP) was used. The silty soils were located in the north or north
west from Halle: in an open pit near the village Ploetz (PL), the open pit
Fuchsberg (FB) and the archaeological excavation site “Erdwerk Salzmuende”
(SM). For a comparative overview over all characteristic properties of the
different soils see Table 1. The material SM is a loess, the other PL and FB
are leached loess-loams. The clay (SP) is emerged from a rhyolitic rock in situ
as a result of intensive weathering.
3 Suction Measurements Results
To determine the suction tension in the range of 6 to 1,500 kN/m2 a pressure

membrane apparatus was used. As expected water retention curves of the silty
soils are steeper than of the clay. The regression results of the pF curve are
given in Table 2. If the residual water content is defined as the adsorbed water
with a suction pressure of 10,000 kN/m2 (Lu and Likos 2004), the residual
water content can be estimated. With the assumption that the regression
function is valid for higher suction pressures too, the residual water content
in clay is about 19.5% and in silt it range from 0.1 to 2.1%.
Table 2. Regression results of the pF curves: equation, correlation coefficient and

residual water content θr (assumption: the suction of adsorbed residual water is
10,000 kN/m2 )
Soil pF function Correlation coefficient θr [%]

Clay (SP) y = 30.78 − 1.23 ln x 0.99 17.5
Silt (FB) y = 22.69 − 2.23 ln x 0.98 2.1
Silt (PL) y = 30.18 − 3.26 ln x 0.98 0.1
4 Shear Test Results

The shear velocity in triaxial and direct shear tests was 0.005 mm/min. The
influence of effective suction tension on maximal shear strength is more impor-
tant in direct shear tests as shown in Fig. 1. Higher effective suction pressures
are linked with higher maximal shear strengths.
600
500
shear strength [kN/m²]
400
300
200
100
0
0 200 400 600 800 1000
effective suction pressure [kN/m²]
silt (FB, SM, PL) direct shear tests clay (SP) direct shear tests
silt (FB, SM) triaxial tests clay (SP) triaxial tests
Fig. 1. Maximal shear strength versus effective suction pressure
4.1 Results of Triaxial Shear Tests
Estimating the shear strength parameters five single tests with cell pressures
of σ3 = 25, 50, 100, 200 and 400 kN/m2 with one soil in similar saturation
level were realized. All samples were isotropically consolidated at the same
cell pressure as during shearing until there were no significant deformation.
This consolidation took about half an hour to two hours in silt, and few hours
in clay.
Generally friction angle decreases with increasing water saturation (see
Table 3). There is no clear correlation between water saturation and effective
cohesion. Effective suction tension correlates with water saturation degree as
shown in Fig. 2. If the water saturation level is low, effective suction pressure
increases.
There are two different fracture mechanisms in unsaturated fine grained
soils. While in the clay a certain sloped plane is developed and the upper block
slides along the lower, in the silt a more or less symmetric system of crossed
shear planes, with two main directions, is formed (see Fig. 3). If the saturation
level is near the complete saturation during the triaxial test, samples of both
soil types show a barrel-shaped body structure without macroscopically visible

cracks.
Table 3. Average saturation degrees sR , total and effective shear parameters of

triaxial tests of unsaturated fine grained soils (shear velocity 0.005 mm/min)
Sample sR [%] φ [◦ ] cu [kN/m2 ] φ [◦ ] c [kN/m2 ]

Clay (SP) 69 17.46 27.81 17.46 27.81
78 1.78 21.20 7.65 19.19
Silt (FB) 41 35.37 64.03 35.37 64.03
45 33.84 0.00 35.07 11.24
53 31.29 4.90 38.60 11.54
Silt (SM) 37 40.45 23.46 40.45 23.46
41 35.46 21.08 35.42 21.36
88 17.11 10.45 39.85 7.15
1,0
y = -0,0124Ln(x) + 0,9222
0,8
water saturation
0,6
0,4 y = -0,0517Ln(x) + 0,7922
0,2
0,0
0 200 400 600 800 1000
effective suction tension [kN/m²]
silt (FB, SM) clay (SP)
Fig. 2. Water saturation degree versus effective suction pressure (calculated after
equation (8)) of three different soils in triaxial shear tests
A) B) C) D)
Fig. 3. Geometry of shear splices and sample deformation in triaxial shear tests
with 25 kN/m2 cell pressure: A) clay (SP) with 66% water saturation; B) clay (SP)
water saturated (100 kN/m2 cell pressure); C) silt (FB) with 39% water saturation,
D) silt (FB) water saturated (200 kN/m2 cell pressure)
4.2 Results of Direct Shear Tests
The cohesion and friction angle both seem to be dependent on saturation level.
This should be an effect of the enforced shear splice, while in the triaxial tests
the deformation can develop free. Even in case of very slow shear velocities a
pore pressure partly can be effective in the shear plane, because the enforced
movement causes turbulent water flow nearest to the shear plane. To calculate
the effective suction after Eq. (8), the average saturation degree of the sample
is used. But its influence is overlaid by dynamical forces of the water in the
nearest area around the shear plane. As a result, the effective suction pres-
sure does not correlate with water saturation. The effective suction pressure
depends on the water allocation in the sample, more precisely, in the shear
splice.
Table 4. Average water saturation sR and effective shear parameters of direct shear
tests of unsaturated fine grained soils (shear velocity 0.005 mm/min)
Sample sR [%] φ [◦ ] c [kN/m2 ]

Clay (SP) 69 23.88 5.60
Silt (FB) 29 23.61 34.52
69 26.73 9.73
Silt (PL) 42 29.17 17.32
63 26.09 8.13
Loess (SM) 26 31.91 13.14
53 29.19 14.72
70 27.64 4.66
Is there an influence of shear velocity? The clay is used to discuss

the influence of shear velocity between 0.002 and 0.15 mm/min, because of its
very low water permeability, the largest influence of pore water pressure is to
be expected. In statistical tests the water saturation degree of the clay ranges
from 70 to 85%.
On the one hand, the effective suction pressure decreases with higher shear
velocities with a logarithmic correlation. On the other hand the maximal
shear strength increases with increasing shear velocity until 0.03 mm/min. It
is almost constant at higher shear velocities. Maximal shear strength increases
with increasing suction pressure in statistical tests with the same velocity and
vertical loads but different saturation degrees.
If the effective suction tension is bigger at low shear velocities, there must
be another reason why the shear strength is lower at low velocities. Creeping
processes and reduction of void ratio by compaction of the soil could explain
these effects.
5 Conclusions
Suction tension is noticeable in unsaturated fine grained soils. Its effects can
be measured in mechanical tests (triaxial compression tests and direct shear
tests) by evaluating an effective suction pressure during shearing. This is based
on theoretical assumptions as well as on the knowledge of the degree of satura-
tion. The suction tension effect is quantified more likely in triaxial compression
tests than in direct shear tests. That is why in the latter one, smaller samples
are tested usually, and a distinct orientation of the shear splice is enforced.
Even at normal stresses of 50 to 200 kPa one can differentiate an increas-
ing trend of suction tension as a distinct effect at the same stress level. Either
way, the quantification is difficult due to various parameters influencing the
strength. Consequently, the effect of suction tension on the strength of un-
saturated soils may be even relevant at low stresses and median saturation
degrees around 50% (40–70%). Low shear velocities apparently reduce the
suction tension effects due to creeping with pore space reduction and increase
of saturation degree. Clay and silt differ in their strength behaviour and in
their relations between effective suction tension and saturation degree.
References
Behrens W, Neumann M (2002) Untersuchungsergebnisse zu einigen mechanischen
Eigenschaften von TRISOPLAST. Müll und Abfall 2:86–89
Bishop AW (1959) The principle of effective stress. Teknisk Ukeblad 106/39:859–863
Bönsch C (2006) Weathering of partial saturated soft rock. In: Von Cotthem A,
Charlier P, Thimus JF, Tshibangu JP (eds) Multiphysics coupling and long
term behaviour in rock mechanics. Taylor and Francis London, pp 397–402
Fredlund DG, Xing A (1994) Equations for the soil water characteristic curve, Can
Geotech J 31:521–532
Junge T, Gräsle W, Bensel G, Horn R (2000) Einfluss des Porenwasserdrucks auf
die Zugfestigkeit von Bodenproben. J Plant Nutrition Soil Sci 163(1):21–26
Lu N, Likos WJ (2004) Unsaturated Soil Mechanics. John Wiley and Sons, New
York
Schanz T, Agus SS, Tscheschlok G (2004) Hydraulischmechanische Eigenschaften
einer polymerverbesserten Sand-Bentonit-Mischung beim Einsatz im De-
poniebau, Geotechnik 27(4):344–355
Schick P (2004) Scherfestigkeit durch Kapillarität in unzementierten ungesättigten
bindigen Böden, Bautechnik 81(1):90–104
Vanapilli SK, Fredlund DG, Pufahl DE, Clifton AW (1996) A model for prediction
of shear strength with respect to soil suction, Can Geotech J 33:379–392
Shear Strength Behaviour of Unsaturated Silty
Soil
Ali R. Estabragh1 and Akbar A. Javadi2

1
Faculty of Soil and Water Engineering, University of Tehran, Karaj 31587-77871,
Iran a.estabragh@gmail.com
2
School of Engineering, Computer Science and Mathematics, University of
Exeter, EX4 4QF, UK a.a.javadi@exeter.ac.uk
Summary. This paper presents results of an experimental study on shear strength

behaviour of an unsaturated silty soil. A comprehensive set of laboratory exper-
iments have been undertaken in a double-walled triaxial cell (and a conventional
cell for saturated samples) on samples of a compacted silty soil. In the experiments
the soil samples were subjected to isotropic consolidation followed by unloading and
subsequent anisotropic reloading (shearing) under constant suctions. Volume change
and shear strength data for the samples were monitored continuously during the ex-
periments. In this paper, the results of the experimental study will be presented and
the effects of suction on the shear strength and volume change of unsaturated soil
will be discussed in detail.
Key words: unsaturated soils, suction, shear strength, overconsolidation ratio, vol-
ume change
1 Introduction
There are vast areas of the world, particularly in tropical and subtropi-
cal regions, where the soils are generally unsaturated. In addition, even in
many temperate regions, the soil above the water table remains unsaturated.
Geotechnical engineers need to be able to assess the shear strength of unsat-
urated soils for safe and cost effective design of structures founded on unsat-
urated soils.
The first attempt to explain the shear behaviour of unsaturated soil was
presented by Bishop (1959). The use of Bishop’s equation for unsaturated
soils was criticised by Jennings and Burland (1962) and Burland (1964). They
pointed out that Bishop’s equation, while appearing to explain shear strength
behaviour, could not fully explain volume change behaviour. The importance
of separating the stress state variables (σ − ua ) and (s = ua − uw ) has been
emphasized (where σ is total stress s is suction ua is pore air pressure and uw
is pore water pressure) by many researchers such as Fredlund et al. (1978),
154 Ali R. Estabragh and Akbar A. Javadi
Matyas and Radhakrishna (1968) and Wheeler and Karube (1995). Fredlund
et al. (1978) and Fredlund (1979) put forward concepts for unsaturated soil
based on independent stress state variables, giving the shear strength rela-
tionship as:
τ = c + (σ − ua ) tan φ + (ua − uw ) tan φb (1)
where τ is shear strength, c is the effective cohesion, φ is the angle of friction
and φb is the angle of friction for changes in ua − uw . In this method φb is as-
sumed to be constant for all values of matric suction (ua − uw ). This approach
has been widely used in interpreting shearing behaviour of unsaturated soils.
Fredlund et al. (1978) suggested that φb could be assumed to be equal to the
effective stress angle of friction measured in saturated condition (φ ). This
would suggest that it is constant for all values of matric suction. The use of
a linear relationship between ua − uw and φb (i.e., a constant value of φb )
was shown to be in error by Escario and Saez (1986). This non-linearity was
confirmed by Fredlund et al. (1978) who assumed φb varied as a function of
suction. In what follows the results of an experimental study on the shearing
behaviour of an unsaturated soil will be presented and the effect of suction
on the shear strength of unsaturated soils will be discussed.
2 Test Procedure and Program
2.1 Preparation of Compacted Specimens
The soil used in the testing program was a silty soil with low plasticity. The
soil comprised 5% sand, 90% silt and 5% clay and had a liquid limit of 29%
and plasticity index of 19%. The results of the standard proctor compaction
test indicated a maximum dry density of 1.74 Mg/m3 at an optimum water
content of 14.5%. Compacted specimens were prepared using a compaction
mould designed specially for static compaction. The purpose of using static
compaction as opposed to dynamic compaction was to obtain a more homoge-
nous specimen in terms of density and shear strength throughout the volume
of the specimen. The soil mass was mixed at a water content of 10% (4% less
than optimum from standard compaction test). The specimens were 76 mm
high and 38 mm in diameter. Compaction was done in nine layers with each
layer being compacted to a maximum vertical total stress of 1600 kPa. The
specimens produced by this method were found to be very uniform and the
interface between the layers was barely apparent.
2.2 Test Procedure and Program
A series of triaxial tests were conducted using a conventional triaxial cell for
saturated and a modified cell for unsaturated soil testing Estabragh et al.
(2004). Both triaxial cells were connected to GDS controllers for applying the
Shear Strength Behaviour of Unsaturated Silty Soil 155
cell pressures, pore water pressure (uw ) and axial strain. All the experimental
data were recorded continuously by a computer. The triaxial test program
consisted of several stages as follows:
3.2.1 Equalisation
After setting the apparatus, the first step of each test was equalisation. At
this stage by applying the required air and water pressures, the specimen was
brought to the desired suction.
3.2.2 Ramp consolidation and unloading
After the equalisation stage each specimen was consolidated isotropically to

the selected value of mean net stress (p ) by ramp consolidation. The sample
was then unloaded isotropically to a predefined lower value of mean net stress.
Suction was maintained constant during loading and unloading.
3.2.3 Shearing
After the unloading stage, each specimen, with pre-defined overconsolidation

ratio (OCR), was sheared by applying an axial load at a constant compressive
strain rate Estabragh and Javadi (2006).
Typical results of the triaxial tests are presented in Figs. 1 and 2. The triaxial
shear tests were performed under controlled suctions (0, 100, 200 or 300 kPa)
and constant cell pressures (50, 100, 200, 300 or 400 kPa). Each series of tests
was carried out at a specified overconsolidation ratio (OCR = 11, 5.5, 2.75,
1.38 and 1.1). Compression of a specimen during shearing is expressed using
a negative sign, and a positive sign is used for dilation of the specimen in
the graphs of volumetric strain versus axial strain. Figs. 1a and 2a show that
in the shear tests the deviator stress increased with increasing cell pressure
until a peak value and then remained nearly constant. Figs. 1b and 2b show
that, generally, the volume of the samples decreased during shearing except
for the samples with high overconsolidation ratios. For tests with σ3 = 50
and 100 kPa at suctions of 100 and 200 kPa the deviator stress first increased
and then slightly decreased. During shearing, the volume of these samples in-
creased after a slight initial contraction. However, it is concluded from these
results that the stiffness of the soil specimens generally increased with in-
creasing confining pressure and decreasing the value of OCR. In the tests
conducted on specimens with high OCR values of 11 and 5.5, the specimens
exhibited a relatively brittle behaviour and during shearing, a slight increase
.
1000
σ 3 = 400 kPa
σ3 =
Devatior Stress (kPa)
800 300
kPa
600
400
200 σ 3 = 100 kPa σ 3 = 200 kPa

σ 3 = 50 kPa
0
0 5 10 15 20
Axial Strain(%)
(a) Deviator stress with axial strain
2
1 σ 3 = 50 kPa
0
Volume Strain(%)
-1 σ 3 = 100 kPa
-2
-3
-4 σ 3 = 200 kPa
-5
-6
-7 σ 3 = 300 kPa
-8
-9
0 5 10 15 20
Axial Strain(%)
(b) Volumetric strain-axial strain curves
Fig. 1. (a) Stress-strain curves, (b) Volumetric strain-axial strain curves for samples
tested at s = 100 kPa under various cell pressures
in total volume was observed after an initial compression. This behaviour can
be attributed to the influence of matric suction on the stiffness, brittleness
and dilatancy of the soil specimens, especially at low confining pressures. The
exhibition of post softening behaviour is commonly observed in heavily over
consolidated specimens.
Figure 3 shows typical results of variation of maximum deviator stress
with suction at constant cell pressure. This figure shows that the deviator
Devatior Stress(kPa) 1000
800
600
400
200
0
0 5 10 15 20
Axial Strain(%)
(a) Deviator stress-axial strain curves
3
2 σ 3 = 50 kPa σ 3 = 100 kPa
1
Volume Strain(%)
0
σ 3 = 200 kPa σ 3 = 300 kPa
-1
-2
-3
-4
σ 3 = 400 kPa
-5
-6
-7
0 5 10 15 20
Axial Strain(%)
(b) Volumetric strain-axial strain curves
Fig. 2. (a) Stress-strain curves, (b) Volumetric strain-axial strain curves for samples
tested at s = 200 kPa under various cell pressures
stress increased with increasing suction. Also, Figs. 1 and 2 show that the
brittleness of the soil decreased with increasing confining pressure. It can
be concluded that the strength of soil specimens increases with increasing
suction but its variation is not linear. The increase in cell pressure causes
a progressive evolution from dilatancy to compression behaviour until the

dilatancy completely disappears.
500
q (shear stress), kPa
400
300
200
0 100 200 300
s (Suction), kPa
Fig. 3. Variation of maximum deviator stress with suction
100
c (Cohesion), kPa
50
’
0
0 100 200 300
s (Suction), kPa
Fig. 4. Variation of c with s

Figure 4 indicates a similar pattern of increase in cohesion intercept with

increase in matric suction at the peak shear stress for all specimens tested
under different confining pressures.
4 Conclusion
The experimental program on the saturated and unsaturated specimens pro-
vided some information on the shear strength behaviour of unsaturated com-
pacted silty soils. The following conclusions are drawn.
Both confining pressure and suction affect the shear strength behaviour of
unsaturated soil.
Softening and increasing the stiffness (hardening) of specimens for a given
cell pressure depend on the value of suction.
Dilatancy or compression may occur during shearing; suction and OCR
are important factors affecting dilatancy and compression.
The cohesion of soil is a function of suction and the relationship is non-
linear.
References
Bishop AW (1959) The principle of effective stress. Tek Ukeblad 39:859–863
Burland JB (1964) Effective stress in partly saturated soils, Géotechnique 14:64–68
Escario V, Saez J (1986) The shear strength of partly saturated soil, Geotechnique
36(3):453–356
Estabragh AR, Javadi AA, Boot JC (2004) Effect of compaction pressure on con-
solidation behaviour of unsaturated soil, Can Geotech J 41(3):540–550
Estabragh AR, Javadi AA (2006) Yielding of unsaturated silty soil under anisotropic
condition. In: Miller GA, Zapata CZ, Houston SL, Fredlund DG (eds) Proc 4th
Int Conf on Unsaturated Soils, Arizona, USA 1:1259–1266
Fredlund DG, Rahardjo H, Gan JKM (1978) The shear strength of unsaturated
soils, Can Geotech J 15:313–321
Fredlund DG (1979) Appropriate concepts and technology for unsaturated soil, Can
Geotech Eng J 16:121–139
Jennings JEB, Burland JB (1962) Limitation to the use of effective stress in partially
saturated soils, Géotechnique 12(2):125–144
Matyas EL, Radhakrishna HS (1968) Volume change characteristics of partially
saturated soils, Géotechnique 18(4):432–448
Wheeler SJ, Karube D (1995) State of the art report: Constitutive modellind in un-
saturated soils. In: Alonso EE, Delage P (eds) Proc 1st Int Conf on Unsaturated
Soils. Paris, France 3:1323–1356
Experimental Investigation on the Time
Dependent Behaviour of a Multiphase Chalk
Grégoire Priol1 , Vincenzo De Gennaro2 , Pierre Delage2 , and Thibaut

Servant2
1
Geotechnical Department, NTNU, Trondheim, Norway
gregoire.priol@ntnu.no
2
Ecole Nationale des Ponts et Chaussées, Paris (CERMES, Institut Navier), 6–8
av. B. Pascal, F–77455 Marne–la–Vallée cedex 2, France
vincenzo.degennaro@enpc.fr, pierre.delage@enpc.fr,
servantt@eleves.enpc.fr
Summary. A series of high pressure oedometer compression tests were carried out
on samples of Lixhe chalk (Belgium) containing various pore fluids: water, air (dry
sample), oil (Soltrol) and oil/water under a controlled suction of 200 kPa. Strain
curves with respect to time of tests carried out under a constant load were exam-
ined in a purpose of investigating the time dependent behaviour of chalk. A simple
empirical exponential law able to account for the combined effects of stress and
suction is proposed. Experiments show that it provides satisfactory results to pre-
dict a feature of multiphase chalk behaviour that is important in the study of the
behaviour of oilfields.
Key words: chalk, multiphase, creep, water weakening, suction, collapse
Introduction
Investigations carried out on multiphase chalk (i.e. chalk that contains vari-
ous non miscible pore fluids) as for example oil reservoir chalk (with gas, oil
and water) or unsaturated chalk (with air and water) have evidenced a typ-
ical water weakening effect under water infiltration. Water weakening leads
to: (i) instantaneous deformation (collapse) under constant load, (ii) strength
reduction and (iii) increasing compressibility (see among others: Masson 1973,
Schroeder et al. 1998, Gutierrez et al. 2000, Lord et al. 2002, De Gennaro et al.
2004). Research on oilfield chalks has shown that important physico-chemical
interactions exist between water, oil and chalk involving collapse and time
dependent behaviour (e.g. Newman 1983, Hellmann et al. 2002, Risnes et
al. 2005). Delage et al. (1996) first suggested the reliability of the mechanical
framework commonly used for unsaturated soils to study the coupled mechan-
ical response of multiphase chalks provided oil is considered as a non wetting
162 G. Priol et al.
fluid (like air in unsaturated soils). Instantaneous collapse during water injec-
tion in oil reservoir chalk was thus considered as a typical case of collapse due
to wetting under constant load (Pasachalk 2001, 2004).
In this paper, results from oedometer tests performed on a chalk saturated
with various pore fluids (oil, water and air) are presented. Particular atten-
tion has been paid to the time dependent behaviour of chalk that has been
investigated by running creep tests. The effect of suction on the time depen-
dent behaviour of chalk is investigated and the relevance of a simple empirical
creep law accounting for suction effects is demonstrated.
Tested Material and Experimental Techniques

The chalk specimens used in this study come from blocks of an outcrop chalk
extracted near Lixhe (CBR quarry, Belgium). This chalk originates from the
same geological unit as some oil reservoir chalks of the North Sea (Ekofisk).
Lixhe chalk is composed of almost pure calcite (with less than 1% of silica)
with an average porosity n ∼ = 43%. Oedometer tests were carried out on
chalk specimens that were saturated with water and/or Soltrol 170 (Phillips
Petroleum Company). Soltrol is a non polar organic liquid not miscible to
water used as an oil analog. Oil–water–chalk interaction is believed to be
mainly dependent on the capillary effects in the water menisci that are located
in the small intergranular spaces close to grain contacts. By defining the two
pressures uo and uw as the oil and water pressures respectively, the oil/water
suction (called suction in the following) is defined as s = uo − uw (to compare
to the air–water suction in unsaturated soil s = ua − uw , being ua the air
pressure).
Four standard step loading oedometer tests have been performed on sam-
ples of 38 mm diameter and 20 mm high with total tests duration ranging from
six months to more than one year. Tests were respectively performed on a wa-
ter saturated sample, an oil saturated sample, a dry sample and the last test
was carried out under controlled suction using the osmotic technique (Kassiff
and Ben Shalom 1971, Delage et al. 1992, Dineen and Burland 1995). The
dry sample was obtained after a heating period of 24 hours at 105◦ C. Based
on the retention properties of Lixhe chalk a suction of 200 kPa was selected,
corresponding to a degree of water saturation Srw of about 30%.
Void Ratio–Time Relationship

In saturated soils, compression strains in oedometer tests are due to both pore
pressure dissipation and creep (respectively primary and secondary consolida-
tion). In a multiphase chalk with two pore fluids, compression affects the pore
pressures and strain stabilisation corresponds with suction equilibration un-
der the applied mean net stress (σv = σv − uo ). In saturated soils, there is still
Investigation on the Time Dependent Behaviour of a Multiphase Chalk 163
1 Stage at 14.5 MPa 1E-007

0.72 Sample water saturated
Settlement
σ = 5,6 MPa Rate
NORMALIZED VOID RATIO e/e 0

Model settlement
0.68 0.96 Model rate
σ = 6,6 MPa 1E-008
e
Creep rate (s-1)

VOID RATIO,
0.64
σ = 8,6 MPa
0.92
e
0.6
σ = 10,7 MPa 1E-009
0.56 0.88
σ = 14,8 MPa
0.52 1E-010
1000 10000 100000 1000000 10000000 0 20 40 60 80 100
Temps (s) TIME (days)
TIME (s )
(a) (b)
Fig. 1. (a) Consolidation curves under a suction equal to 200 kPa; (b) Analysis of
the void ratio-time relationship during consolidation
no agreement how to distinguish the effect of viscous strain to that of excess

pore pressure dissipation during primary consolidation (Mesri and Choi 1985,
Leroueil 2006). Results from compression tests on two chalks (De Gennaro et
al. 2005, Priol 2005) showed that neither Casagrande’s nor Taylor’s methods
were helpful to identify any primary consolidation. Indeed, it seems likely that
the low compressibility of the soil skeleton due to intergranular bonding ef-
fect together with the relatively high chalk permeability (about 10−8 m/s) are
sufficient to prevent excess pore pressure generation (given that the amount
of the applied total load instantaneously transferred to pore water pressure
decreases when the drained bulk modulus increases, Biot 1941).
Indeed, the results of the suction controlled oedometer test (s = 200 kPa)
presented in Fig. 1a show that, from the early beginning of the test, the curves
giving the time dependent strain look very similar to creep curves. Hence, the
time dependent volumetric strain of chalk should be essentially due to creep
with the solid skeleton continuously settling under zero excess pore pressure.
This result should obviously be confirmed by pore pressure measurements
during initial loading.
The examination of a typical consolidation curve is presented in Fig. 1b in
terms of normalised void ratio e/eo plotted as a function of the elapsed time
in a linear scale. This allows defining the following evolution law:
e
= βi tαi (1)
eoi
where eoi is the void ratio at the beginning of loading stage i. βi represents the
instantaneous strain whereas αi controls the shape of the curve in Fig. 2, the
164 G. Priol et al.
Fig. 2. (a) Compression curves of Lixhe chalk at various saturation states; (b)
Influence of stress level and suction on the creep coefficient α
higher αi the larger the time dependent strain. Typical values of αi are be-
tween 0 (no time dependency) and 0.02. Note that equation (1) bears strong
similitude with the common interpretation in term of creep coefficient Cαe
when using a logarithmic time scale (Cαe = Δe/Δ log t). As shown in Fig. 1b,
the interpolation given by equation (1) satisfactorily fits the experimental
data. This was verified, with minor exceptions, for all the analysed consolida-
tion curves.
The creep strain rate evolution law presented in dotted line in Fig. 1b has
been obtained by deriving equation (1) with respect to time. Good agreement
is als observed with experimental data. This allows identifying the creep rate
at each creep phase (right axis in Fig. 1b).
Stress-Void Ratio Relationship
In the absence of any apparent pore pressure dissipation phase, it was neces-
sary to adopt a strain criterion to define the stress strain curves. In the case
of isochronous time dependent behaviour (i.e. only dependent of the loading
duration tc ), the strain after a loading duration tc can be adopted (Bjerrum
1967). In the alternative case of isotach behaviour (i.e. only dependent on
the strain rate ε̇), the strain adopted corresponds to a given creep rate (ε̇c )
(Šuklje 1957 in Leroueil 2006). This second approach was adopted here with
ε̇c = 10−10 s−1 . This value was chosen based on the accuracy limit of the
LVDT used to measure the vertical displacement. The values of e determined
at each loading stage with this approach correspond to given couples of αi
and βi values.
The compression curves obtained are presented in Fig. 2a. Like in unsat-
urated soils, the higher the suction (i.e. the lower the degree of saturation
in water Srw ), the higher the yield stress, in good agreement with the wa-
ter weakening effect. In contrast to unsaturated soils, the plastic compression
modulus appears to be slightly suction dependent. Note however that it is
smaller in the case of the dry sample. This shows that the fluid chemistry
and the chalk wettability to fluids (oil, water, air) influence the mechanical
behaviour.
Creep Evolution
Since relation (1) is valid for all the saturation states, the αi and βi coeffi-
cients have been assessed at all the loading stages of the four tests of Fig. 2a.
The values of the creep coefficients αi are plotted with respect to the corre-
sponding applied vertical stress in Fig. 2b. A careful examination of the plots
(between 5 and 10 MPa) shows that they are bilinear with a change in slope
that approximately corresponds with the yield stress observed in Fig. 2a (not
shown here in more detail for lack of space). Above 5 MPa, αi increases with
increased stress and decreased suction, starting from a fully water saturated
state (s = 0 and Srw = 100%), passing through a partially saturated state
(s = 200 kPa and Srw = 30%) up to a dry state. This is again consistent with
the water weakening effects discussed in the previous sections.
Relatively few creep tests on unsaturated materials have been presented
in the literature. Similar effects of suction and stress on creep were also doc-
umented from oedometer tests carried out on a partially saturated rockfill
(Alonso et al. 2005). However, the generalization of these findings to all pos-
sible stress states is an open issue.
The effect of water injection in a chalk sample saturated with oil under
a constant vertical stress of 19.8 MPa is shown in Fig. 2b. This situation
is somewhat similar to that prevailing during waterflooding, a technique of
enhanced oil recovery used in the North Sea. The sample was injected under
a 10 kPa water pressure and it exhibited a collapse that occurred within a few
hours. A more pronounced creep characterised by a αi coefficient of 0.0164 was
then observed. Starting from the oil saturated plot of Fig. 2b with a αi value of
0.0047 at point A, the path followed in this diagram leads to point B (0.0164).
It is observed that point B is located close to the s = 200 kPa line. This is
related to the residual saturation in oil (Sro = 30%) that probably has an
effect on the macroscopic creep behaviour of the chalk through the physics of
the intergrains bonds. In other words, there is apparently a difference between
two samples at zero suction between a fully saturated sample and a sample
with a residual oil content. Hysteresis effects may also play a role in this
difference.
166 G. Priol et al.
Conclusion
The experimental data presented in this paper were aimed at showing some
effects of suction on the compressibility and on the time dependent behaviour
of a sample used as a model of a multiphase oil reservoir chalk. Some simi-
larities with the behaviour of unsaturated soils have been observed, the non
wetting fluid in chalk being oil as compared to air in unsaturated soils. Chalk
strength is increased by increasing the suction level, i.e. by decreasing the
water content. Time dependent strains have been related to the changes in
water saturation by means of the oil–water suction. Some aspects of the wa-
ter weakening effect in chalk, including time effects, have been considered in a
more comprehensive framework based on the use of the oil-water suction as an
independent stress variable capable to account for the fluid-rock interactions.
Acknowledgements
The work carried out was conducted in relation with the European Project
Pasachalk 2 (EC contract ENK6–2000–00089) in the framework of the first
author’s PhD thesis that was supported by a grant from Ecole des ponts, Paris.
The first author is now in NTNU Trondheim with the support of the “Marie
Curie” RTN “AMGISS” network (MRTN–CT–2004–512120). The second and
third authors wish to acknowledge the support of the EU via the “Marie Curie”
RTN “MUSE” network (MRTN–CT–2004–506861).
References
Alonso EE, Olivella S, Pinyol NM (2005) A review of Beliche dam. Géotechnique
55(4):267–285
Biot MA (1941) General theory of three-dimensional consolidation, J Appl Phys
12:155–164
Bjerrum L (1967) Engineering geology of Norwegian normally-consolidated marine
clays as related to settlement of buildings, Géotechnique 17:81–118
De Gennaro V, Delage P, Cui YJ, Schroeder Ch, Collin F (2003) Time-dependent
behaviour of oil reservoir chalk: a multiphase approach, Soils and Foundations
43(4):131–148
De Gennaro V, Delage P, Priol G, Collin F, Cui YJ (2004) On the collapse behaviour
of oil reservoir chalk, Géotechnique 54(6):415–420
De Gennaro V, Delage P, Sorgi C, Priol G, Collin F (2005) Time dependent com-
pression behaviour of two chalks, In: Proc 11th Int Conf Int Ass of Comp Meth
Adv Geomech (IACMAG), Torino, Italy, June 2005:34–42
Delage P, Suraj De Silva GPR, Vicol T (1992) Suction controlled testing of non
saturated soils with an osmotic consolidometer. In: 7th Int Conf Expansive Soils,
Dallas:206–211
Delage P, Schroeder C, Cui YJ (1996) Subsidence and capillary effects in chalks.

In: EUROCK’96, Prediction and performance on rock mechanics and rock engi-
neering, Turin, Italy 2:1291–1298
Dineen K, Burland JB (1995) A new approach to osmotically controlled oedometer
testing. In: Proc 1st Int Conf on Unsaturated Soils UNSAT95, Paris:459–465
Gutierrez M, Oino LE, Hoeg K (2000) The effect of fluid content on the mechanical
behaviour of the fractures in chalk, Rock mech Rocks eng 33(2):93–117
Hellmann R, Renders PJN, Gratier J-P, Guiguet R (2002) Experimental pressure so-
lution compaction of chalk in aqueous solutions. Part1. Deformation behaviour
and chemistry. In: Hellmann, Wood (eds) Water-Rock Interactions, Ore De-
posits, and Environmental Geochemistry: A Tribute to David A. Crerar; The
Geochemical Society, Special Publication No 7:129–152
Kassiff G, Ben Shalom A (1971) Experimental relationship between swell pressure
and suction, Géotechnique 21:245–255
Leroueil S (2006) The isotache approach. Where are we 50 years after its develop-
ment by Prof. Sukjle? In: Proc 13th Danube-European Conf on Geotechnical
Engineering, Ljubljana 2:55–88
Lord JA, Clayton CRI, Mortimore RN (2002) Engineering in chalk. CIRIA Eds
Masson (1973) La craie. Bulletin des Laboratories des Ponts et Chaussées, Special
V190 pp
Mesri G, Choi YK (1985) The uniqueness of the End of primary (EOP) void ratio-
effective stress relationship. In: Proc 11th Int Conf on Soil Mech and Found Eng,
San Fransisco 2:587–590
Newman GH (1983) The effect of water chemistry on the laboratory compression
and permeability characteristics of some North Sea chalks, J Petroleum Eng
976–980
Pasachalk1 (2001) Mechanical Behaviour of PArtially and Multiphase SAturated
CHALKs Fluid-skeleton Interaction: Main Factor of Chalk Oil Reservoirs Com-
paction and Related Subsidence. Publishable Final report, European Joule III
contract No JOF3CT970033
Pasachalk2 (2004) Mechanical Behaviour of PArtially and Multiphase SAturated
CHALKs Fluid-skeleton Interaction: Main Factor of Chalk Oil Reservoirs Com-
paction and Related Subsidence. Part 2. Publishable Final report, Energy, En-
vironment and Sustainable Development Programme, contract No ENK6–2000–
00089
Priol G (2005) Comportement mécanique différé et mouillabilité d’une craie pétro-
lifère. ENPC, PhD Thesis
Risnes R, Madland MV, Hole M, Kwabiah NK (2005) Water weakening of chalk-
Mechanical effects of water-glycol mixtures, J Petroleum Sci Eng 48:21–36
Schroeder Ch, Bois AP, Maury V, Halle G (1998) Water/chalk (or collapsible soil)
interaction: Part II. Results of tests performed in laboratory on Lixhe chalk to
calibrate water/chalk models. SPE/ISRM (SPE 47587) Eurock’98, Trondheim
Šuklje L (1957) The analysis of the consolidation process by the isotache method.
In: Proc 4th Int Conf on Soil Mech and Found Engng, London 1:200–206
Testing Unsaturated Soil for Plane Strain
Conditions: A New Double Wall Biaxial Device
Tom Schanz and Jamal Alabdullah

tom.schanz@uni-weimar.de, jamal.alabdullah@uni-weimar.de
Summary. In this paper a new experimental device for unsaturated soil testing is
presented. The new Double Wall Biaxial Cell was designed to study the behavior of
unsaturated granular material under plane strain conditions. The required calibra-
tion of the new device is presented in detail. Saturated soil samples were tested in
the new device under isotropic consolidation conditions. The volume changes of the
specimen were measured by two alternative methods to check the reliability of the
new device. To verify the independent two stress state variables concept proposed by
Fredlund and Morgenstern, Null-tests were performed on sand-kaolin mixture spec-
imen. Axis translation technique was used for creating the desired matric suction.
The details of the new device and the test procedures are described. The obtained
result are presented and discussed in detail.
Key words: plane strain, biaxial, double wall cell, granular material
Introduction
For a saturated soil, one stress state variable is enough to describe the behavior
of two phases (solid and fluid/or gas) of soil mass. The stress state variable σ
is defined as σ = σ − uw , where σ is the effective stress, σ is the total stress,
and uw is the pore water pressure. The concept of the effective stress may be
the most important principle in soil mechanics.
Unsaturated soil is normally considered as a three-phase system, i.e., solid,
gas (air), fluid (water). In 1977 Fredlund and Morgenstern added the contrac-
tile skin as a fourth phase and these four phases were used by these two
authors in the stress analysis of unsaturated soil on the basis of continuum
mechanics.
Several authors tried to define a single stress state variable for unsaturated
soil but soil properties were involved in the proposed equations (Corney et al.
1958, Bishop 1959, Richards 1966, Aitchison 1965, 1973). State variables used
to describe the state of the stress have to be independent of soil properties
(Fung 1977, Fredlund and Morgenstern 1977).
170 Tom Schanz and Jamal Alabdullah
In 1961 Bishop and Donald performed a set of triaxial shear tests on

unsaturated silt where the total stress (i.e., cell pressure) σ3 , pore-air pressure
ua , and pore-water pressure uw were varied by equal amount keeping (σ3 −ua )
and (ua − uw ) constant. The result of these tests substantiate the concept of
using (σ3 − ua ) and (ua − uw ) as valid stress state variables for this type of
test.
Fredlund and Morgenstern (1977) discussed the stress state variables con-
trolling the behavior of unsaturated soil, stated that the stress state variables
have to be created from the individual forces components acting on the solid,
water and air phases and the air-water contractile skin. The proposed vari-
ables are the net stress (σ − ua ) and the matric suction (ua − uw ), where σ is
the total stress. In the same work, Fredlund and Morgenstern validated this
concept about the stress state variables by a set of null tests. In these tests it
has been proven that changing the air, water, and the total pressures causes no
changes in the state of the soil as long as the state variables remain constant.
Tarantino et al. (2000) investigated these variables using a new laboratory
apparatus. Using this apparatus, unsaturated kaolin samples were maintained
under a constant state of strain and water content. Keeping the water pressure
in the negative range, changes in net stress and matric suction were measure
in response to applied changes in pore air pressure. The results confirmed the
use of the net stress and the matric suction as stress state variables.
In practice, most common geotechnical problems involve unsaturated soils,
such as embankments, earth dams, highways. The state of deformation in
several structures (e.g., long retaining walls, strip foundations, embankments)
is plane strain state. In fact, unsaturated soil testing equipment normally
consist of odometer, shear box, or triaxial cell. There is no equipment (known
to authors) to test unsaturated soil specimen in biaxial state. All the biaxial
devices found in literature were designed to test saturated or dry soils (Duncan
and Seed 1966, Desrues 1984, Vardoulakis and Graf 1985, Drescher et al. 1990,
Alshibli et al. 2004).
The double wall cell concept is an important concept in unsaturated soil
testing. Several equipment were designed according to the double wall concept.
First double wall cell was presented by Bishop and Donald in 1961. Later,
several researchers presented different designs of the double wall cell (Wheeler
1986, 1988, Cui and Delage 1996, Geiser et al. 2000, Toyota et al. 2001, Ng
et al. 2002, Aversa and Nicotera 2002, Yin 2003, Sivakumar et al. 2004).
The New Biaxial Double Wall Cell
The new device consists of double-wall cell made of high quality acrylic where
the outer cell is reinforced with metal bands, two immovable walls to attain
the plane strain conditions, some accessories such as special mould for sample
preparation, special rectangular membrane, and metal plates at the top and
the base platen to insure the tightening of the membrane by the O-ring.
Testing Unsaturated Soil for Plane Strain Conditions 171
Loading Piston
Outer cell
Inner cell
Ceramic disks
Porous disks
Top cap
Porous disks Specimen
Ceramic disks Ball bearing
Base pedestal
Cell Pressure
C.D (top)
C.D (bottom) P.D (top)
P.D (bottom)
Fig. 1. Double wall biaxial cell
In order to reduce the friction between the base and the soil sample (i.e.,
end restraint) ball bearings are provided at the base of the sample (Vardoulakis
and Goldscheider 1981, Han and Vardoulakis 1991, Alshibli et al. 2004). Teflon
sheets with the thickness of 0.25 mm were used to reduce the friction between
the membrane and the metal side walls. Figure 1 shows schematic diagram
and a photograph of the new double wall biaxial cell.
The sample height is 12 cm and the cross section of the sample is 10 cm ×
4 cm. The top pressure pad and the base carry metal plates, each of these
metal plates are provided with two ceramic disks and two porous disks.
In the current equipment the ceramic disks have an air entry value of
100 kPa. The diameter of these disks is 1.5 cm and the thickness is 0.5 cm.
It is possible to change these metal plates with other plates provided with
different ceramic disks, giving the possibility to test the soil in a wide range
of matric suction. The porous disks facilitate the application of air pressure,
whereas the water pressure is applied through the fully saturated ceramic
disks inducing the desired matric suction in saturated soil sample. Both of
the water pressure uw and air pressure ua can be independently controlled at
the top and the bottom of the specimen, so suction gradient can be applied
during the test. Layout of the new system is shown in Fig. 2. The dimensions
of the inner and outer cell are summarized in Table 1.
Table 1. The dimensions of the inner and outer cell

Diameter (cm) Height (cm) Wall thickness (cm)
Inner cell 17.5 27 0.5
Outer cell 27.0 47 0.8
Fig. 2. Layout of the new double wall biaxial cell system
Calibration
Even though the confining pressure is supposed to be the same in the inner
and outer cell, several factors may affect, resulting in volume changes of the
inner cell. These factors are the flexibility of fittings, time lag in the equal-
ization of the pressures between the inner and outer cell (Leong et al. 2003),
compressibility of the water, temperature fluctuation, absorption of the water
by the acrylic wall of the inner cell (Wheeler 1988, Sivakumar et al. 2004),
and the creep effect of the acrylic inner cell. Another correction should also
be considered due to the piston movement inside the inner cell during the
compression stage of the test (Sivakumar et al. 2004, Ng et al. 2002).
Inner Cell Expansion Due to the Applied Pressure
The cell volume changes of the inner cell due to the applied pressure was deter-
mined by filling the two cells with de-aired water and a dummy metal sample
was used. Pressure was applied stepwise. The immediate volume change was
recorded five minutes after applying pressure. The calibration curve is pre-
sented in the Fig. 3. In the calibration of the inner cell, the pressure was
chosen to cover relatively wide range of pressure (0–600 kPa). The hysteresis
of the system was examined by performing a test in which the cell pressure
was increased from zero to 600 kPa, then the pressure was reduced to zero
following the same loading path. Figure 3 shows good repeatability of the
inner cell volume change during the loading and unloading process, and no
hysteresis was observed.
Calibration of the Inner Cell Using a Saturated Soil Specimen
The reliability of the system was checked by testing a saturated sample in

isotropic consolidation test. For a saturated sample, the volume change of the
Pressure (kPa)
0 100 200 300 400 500 600 700
0
Volume Change (cm3 )

load
2 unload
4
Fig. 3. Calibration of the inner cell (loading and unloading)
Pressure (kPa)
0 100 200 300 400 500 600 700
0
Method A
of the specimen (cm 3)
2
Volume change
Method B
Fig. 4. Volume changes of saturated soil sample measured by two methods A, B
specimen during the test (i.e., total volume change) is equal to the volume of
the water drained out of the specimen.
A saturated Hostun sand (Schanz 1998, Flavigny et al. 1990) specimen was
prepared, the initial void ratio e0 was 0.72. The cell pressure was applied simul-
taneously to the outer and inner cell (isotropic consolidation conditions). The
volume changes of specimen were measured using two methods. In method A
the double wall cell technique was used (i.e., by measuring volume changes of
the inner cell). Whereas in method B the volume of the water expelled out
of the specimen was measured using a burette with accuracy of 0.1 cm3 . Fig-
ure 4 shows the volume changes measured by the two methods. No significant
differences were observed between both method A and B.
Verification of Two Independent Stress State Variables Concept
Two null tests were conducted for a equalized soil sample under specific stress
state. The test is termed null, because the expected result is to measure no
volume changes due to the change in the values of the individual pressures.
It is difficult to measure zero volume change over a long period of time, so it

is expected to measure a slight volume changes due to several factors such as
imperfect testing procedure, air diffusion through the ceramic disk, water loss
of the specimen through the evaporation or diffusion, secondary consolidation
(Fredlund and Morgenstern 1977).
Null tests were performed on Hostun sand-kaolin mixture specimen, the
mixture consists of 10% kaolin and 90% sand. The initial void ratio was 0.56,
coefficient of uniformity 1.95, and d50 = 0.35 mm. The specimen was com-
pacted in the special mold, then it was saturated using back pressure method.
The back pressure value at the end of saturation stage was 60 kPa and cell
pressure was 70 kPa. The axis translation technique was used to create the
matric suction in the soil specimen. The specimen was equalized under these
pressure: cell pressure of 80 kPa, air pressure of 70 kPa, and water pressure
of 60 kPa. Thus the net stress (σ − ua ) = 10 kPa, and the matric suction
(ua − uw ) = 10 kPa.
Equilibrium was reached after four days (i.e., when neither volume changes
of the specimen nor water volume changes were observed), then the individual
pressures were changed with the same increment so that the stress state vari-
ables remain always constant and equal to their initial values. Air pressure was
applied through the porous disks, while water pressure was applied through
ceramic disks with high entry value of 100 kPa. Both of water and air pressure
were applied at the top and the bottom of the specimen. Table 2 shows the
different pressure values and the corresponding stress state variables used in
the tests.
Table 2. Pressures values (kPa) in the null-tests
σ ua uw (σ − ua ) (ua − uw )
Test 1 100 90 80 10 10
Test 2 250 240 230 10 10
Figure 5 shows the total volume changes and the water volume changes
measured in a period of two days.
To check the effect of the possible creep of the inner cell, two tests without
samples were conducted at constant cell pressure (100, 250 kPa), and the
volume changes of the inner cell were recorded for two days.
Volume change versus time plots show that the volume of the specimen
remain almost constant in the period of two days after taking into considera-
tion the volume change due to the creep of the inner cell. The volume changes
due to the creep of the inner cell were subtracted from the apparent volume
changes.
The maximum measured overall volume change (i.e., volumetric strain εv )
was 0.07%. The water volume changes which were measured as water leaving
(a)
(b)
Fig. 5. (a) Volume change versus elapsed time in null test No. 1; (b) Volume change
versus elapsed time in null test No. 2
the specimen during the test, could be contributed to the diffusion of the air to
the water phase of the specimen (Fredlund and Morgenstern 1977). Figure 5
shows the measured volume changes due to the creep of the inner cell. The
creep of the inner cell in this double wall cell system could be demonstrated by
taking some factors into consideration such as the expansion of the tubing, the
absorption of the water by the inner acrylic cell walls, and the compressibility
of the walls of the inner cell due to the compressive stress applied at both
sides of the wall.
To evaluate the accuracy of the volume change measurement, the corrected
volumetric strain εv can be given by: εv = εvm ± β, where εv is the corrected
volumetric strain, εvm is the measured volume change, and β is the absolute
error. The value of β = 0.1% was proposed by Geiser et al. (2000) as an

accepted absolute error in measuring the volumetric strain in triaxial tests on
unsaturated soils Leong et al. (2003).
Conclusions
In this paper, the new developed double wall biaxial cell is presented with
the required calibration tests. The calibration tests using saturated soil spec-
imen show that the system is capable to measure the volume changes of the
specimen according to the double wall cell technique. The observed error in
measuring the volume changes of the saturates soil specimen was 0.01–0.07%.
The maximum measured volumetric strain of the specimen in the null tests
on sand-kaolin mixture specimen was about 0.07%. Regarding to the accepted
absolute error proposed by Geiser et al. (2000), the measured values in the
new device could be accepted.
The null tests results shown in Fig. 5, strongly support the use of the
net stress and the matric suction as two independent stress state variables
to describe the behavior of unsaturated soil. The significant amount of creep
shown in Fig. 5 underlines that the creep of the inner cell in a double wall
system should be taken into consideration in the analysis of the tests results
regarding the volumetric strain.
Acknowledgements
The present work was carried out in the frame of the DFG-research project
Mechanik teilgesättigter Böden (Mechanics of unsaturated soils). An acknowl-
edgement is given to the research assistantship provided by the Bauhaus-Uni-
versität Weimar Germany through the research grant from Deutsche For-
schungsgemeinschaft (DFG), FOR 444/2.
References
Aitchison GD (1973) The quantitative description of the stress deformation behavior
of expansive soils. In: Proc 3rd Int Conf Expansive Soils (Haifa, Israel) 2:79–82
Aitchison GD (ed) (1965) Moisture Equilibria and moisture changes in soils beneath
covered areas. A Symp in print. Butterworth, Australia, 278 pp
Alshibli KA, Godbold DL, Hoffman K (2004) The Louisiana Plane Strain Apparatus
for Soil Testing, Geotech Test J 27(4):1–10
Aversa S, Nicotera MV (2002) A triaxial and oedometer apparatus for testing un-
saturated soils, Geotech Test J 25(1):3–15
Bishop AW (1959) The principle of effective stress (Lecture delivered in Oslo, Nor-
way, in 1955), Teknisk Ukeblad, 106(39):859–863
Bishop AW, Donald IB (1961) The experimental study of partly saturated soils in
the triaxial apparatus, Proc 5th Int Conf on Soil Mechanics and Foundation
Engineering, Paris 1:13–21
Corney D, Coleman JD, Black WPM (1958) Movement and distribution of water in
soil in relation to highway design and performance. Water and Its Conduction
in Soils, Highway Res Board, Special Report, Washington, DC, No. 40:226–252
Cui YJ, Delage P (1996) Yielding and plastic behaviour of an unsaturated silt,
Geotechnique 46(2):291–311
Drescher A, Vardoulakis I, Han C (1990) A biaxial apparatus for testing soil, Geotech
Test J 13:226–234
Duncan J, Seed HB (1966) Strength variation along failure surfaces in clay, Proc
Am Soc Civ Engrs 94(SM6):81–104
Desrues J (1984) La localization de la deformation dans les materiuax granulaires,
PhD Thesis, UJF–INP6 Grenoble
Flavigny E, Desrues J, Palayer B (1990) Note technique: Le sable d’Hostun RF, Rev
Fran C Geotech 53:67–70
Fredlund DG, Morgenstern NR (1977) Stress state variables for unsaturated soils,
J Soil Mech Found Div 103:447–466
Fung YC (1977) Afirst course in continuum mechanics, 2nd ed. Prentice-Hall, En-
glewood Cliffs, NJ 340 pp
Geiser F, Laloui L, Vulliet L (2000) On the volume measurement in unsaturated
triaxial test, Proc Asian Conf Unsat Soil:669–679
Han C, Vardoulakis IG (1991) Plane-strain compression experiments on water-
saturated fine-grained sand, Geotechnique 41(1):49–78
Leong EC, Agus SS, Rhardjo H (2003) Volume change measurement of soil specimen
in triaxial test, Geotech Test J 27(1):47–66
Mokni M, Desrues J (1998) Strain localization measurements in undrained plane-
strain biaxial tests on hostun RF sand, Mechanics of Cohesive-Frictional Mate-
rials 4:419–441
Ng CWW, Zhan LT, Cui YJ (2002) A new simple system for measuring volume
changes in unsaturated soils, Can Geotech J 39:757–764
Richards BG (1966) The significance of moisture flow and equilibria in unsaturated
soils in relation to the design of engineering structures built on shallow foun-
dations in Australia. Symp on permeability and capillary, Amer Soc Testing
Material, Atlantic City, NJ
Schanz T (1998) Zur Modellierung des mechanischen Verhatltens von Reibungsma-
terialien, Mitteilung 45, Institut für Geotechnik, Universität Stuttgart
Sivakumar R, Sivakumar V, Blatz J, Vimalan J (2004) Twin-Cell Stress Path Ap-
paratus for Testing Unsaturated Soils, Geotech Test J 29(2):175–197
Tarantino A, Mongiovi L, Bosco G (2000) An experimental investigation on the inde-
pendent isotropic stress variables for unsaturated soils, Geotechnique 50(3):275–
282
Toyota H, Sakai N, Nishimura T (2001) Effect of stress history due to unsaturation
and drainage conditions on shear properties of unsaturated cohesive soil, Soils
and Foundations 41(1):13–24
Vardoulakis I, Goldscheider M (1981) Biaxial apparatus for testing shear bands in
soils, Proc 10th Int Conf Soil Mech Fndn Engng, Stockholm 4:819–824
Vardoulakis I, Graf B (1985) Calibration of constitutive models for granular mate-
rials using data from biaxial experiment, Geotechnique 35(3):299–317
Wheeler SJ (1986) The stress-strain behavior of soils containing gas bubbles. PhD
Thesis, Oxford University, Oxford, U.K.
Wheeler SJ (1988) The undrained shear strength of soils containing a large gas
bubbles, Geotechnique 28(3):399–413
Yin JH (2003) Double cell triaxial system for continuous measurement of volume
change of an unsaturated or saturated soil specimen in triaxial testing, Geotech
Test J 26(3):353–358
Yin ZZ (ed) (1998) Settlement and consolidation of soil mass. China Electric Pub-
lication House (in Chinese)
Influence of State Variables on the Shear
Behaviour of an Unsaturated Clay
Viktoria Schwarz, Andreas Becker, and Christos Vrettos
Division of Soil Mechanics and Foundation Engineering, Technical University of

Kaiserslautern, Kaiserslautern, Germany
vschwarz@rhrk.uni-kl.de, becker@rhrk.uni-kl.de, vrettos@rhrk.uni-kl.de
Summary. Suction is one important parameter affecting the shear behaviour of

partially saturated soils. State variables such as void ratio and degree of saturation
control the development of suction. A series of deviatoric triaxial tests on kaolinite
clay with direct measurement of suction are reported. An appropriate elasto-plastic
constitutive for this type of soil is presented and its parameters are determined from
the test results.
Key words: clay, constitutive model, shear behaviour, triaxial tests
1 Introduction
The shear behaviour of partially saturated cohesive soils is strongly affected
by the value of suction. Work done in the past on this subject refers mainly to
triaxial tests with controlled value of suction (Blatz and Graham 2003). The
constitutive laws derived from these tests are based on extended Cam-clay
models whereby the suction is explicitly included as an independent stress
component (Alonso et al. 1990, Wheeler and Sivakumar 1995).
An alternative formulation is adopted in the present study. Following
Bishop’s equation (Bishop 1959) the variation of effective stresses with suction
is given by:
σ = (σ − ua ) − (ua − uw ) · χ . (1)
In this equation σ denotes the effective stresses, σ = the total stresses,
ua = the pore air pressure, uw = the pore water pressure (negative valued)
and χ = a factor that depends on soil type and degree of saturation. An appro-
priate elasto-plastic constitutive soil model for clays can thus be formulated
directly in terms of effective stresses for different values of the degree of satu-
ration. The model in its general form has been previously applied to describe
the behaviour of unsaturated cohesionless soils, cf. Weckbecker (1993) and
Becker (2002). A survey of models for unsaturated soils considering Bishop’s
stress representation is given by Khalili et al. (2004).
180 V. Schwarz et al.
A series of triaxial tests have been carried out for this purpose whereby
suction is directly measured during the tests and effective stresses are deter-
mined by equation (1).
In the following, the testing procedure is described, the proposed consti-
tutive relation is presented and its appropriateness for modeling the material
behaviour is shown.
2 Triaxial Tests
A series of triaxial tests using a kaolinite clay of high plasticity have been
carried out. The material properties are summarized in Table 1.
Table 1. Material parameters of the kaolinite clay
Clay minerals Kaolinite > 40%

Illite < 20%
Smectite < 5%
Liquid limit wL = 54.0%
Plastic limit wP = 21.3%
Water absorption value after 4 h wA = 73.5%
The triaxial apparatus developed especially for the study as well as the
specimen preparation technique is described in detail by Schwarz et al. (2006).
Shear behaviour has been investigated in dependency of the initial void
ratio e0 and the degree of saturation Sr at distinct values 0.68/0.78/0.83
and 0.85/0.9/1.0, respectively. This range has been chosen in order to avoid
high initial suction values and the application of the time-consuming axis-
translation-method.
Specimens were loaded up to the specified consolidation stress and then
consolidated until the measured suction value reached an almost constant
level. The consolidation stresses are defined in terms of the first stress invariant
Iσ as follows
Iσ = σ11 + σ22 + σ33 (2)
where σ11 , σ22 and σ33 are the principal stresses.
Tests at two values Iσ equal to 300 kPa and 900 kPa are reported herein. At
the low stress level only compression tests have been conducted while at the
high stress level both compression and extension tests were possible. All tests
were performed under drained conditions with a strain rate of 0.01 mm/min
for compression and 0.005 mm/min for extension, respectively.
Influence of State Variables on the Shear Behaviour 181
3 Constitutive Relation
A conventional elasto-plastic model is adopted here that is based on a proposal
by Stutz (1973). The yield surface has the general form
−m
1 IIIs
f = IIs − A · Iσ
2
1−B 3 =0 (3)
IIs2
where
IIs = sij · sji (4)
and
IIIs = sil · slj · sji (5)
with sij denoting the deviatoric stresses.
A, B and m are parameters to be determined from appropriate triaxial
tests. Parameter A is a function of e0 and Iσ . B is a constant while m is given
by the following relation:
ζ
m= (6)
ln 1 − √B6
where ζ is a constant.
The stress-strain-relationship within the yield surface during unloading
and reloading is assumed to be nonlinear elastic. Triaxial tests not reported
herein have shown that the associated values of bulk modulus K as well as
shear modulus G depend on the stress state, the void ratio, and the degree of
saturation.
The parameter A strongly depends on the initial void ratio e0 , while the
dependency on Iσ is neglected at the present stage. The following linear rela-
tion is assumed:
A = κ · e0 + ξ . (7)
The tests reported here refer to Sr = 0.9 and Iσ = 900 kPa. For this set of
values we obtain κ = −0.31 and ξ = 0.47. The influence of Sr on κ and ξ is
part of the ongoing investigation. The tests further show that the effect of e0
on the parameter B is very weak and can be neglected. Here, B = −0.88 and
ζ = 0.12, that yields from equation (6) a value m = 0.39.
Figure 1 shows the quality of the approximation via the constitutive model
described above by comparing results from five single tests.
For computing the plastic strain increments dεP ij we assume an associated
flow rule
dεPij = λ{h} (8)
with λ denoting the unspecified (positive) scalar multiplier, and

∂f ∂f δij
{h} = − −D , (9)
∂sij ∂sii 3
Fig. 1. Test results at failure and prediction by constitutive law in the deviatoric
plane for Iσ = 900 kPa, Sr = 0.9 at three values of the void ratio e0
sij , sii are the deviatoric stress components and δij is Kronecker’s Delta.
During hardening λ is assumed to be independent of e0
T
∂f
∂σ [C] · {dε}
λ= T T (10)
∂f
∂σ [C]{h} − ∂e ∂f
p {h}
where epij is the plastic deviatoric strain.

The stiffness matrix [C] is defined in terms of the elastic parameters K
and G. D is the dilatation parameter defined by
dIε
D= 0.5
(11)
dIIep
with Iε = ε11 + ε22 + ε33 and IIep denoting the plastic component of IIe =
e211 + e222 + e233 with eii being the deviatoric strain components.
The test results reported herein show that the dilatation parameter D
is almost independent of the degree of saturation and the stress state, as
demonstrated in Fig. 2. It can be seen that there is a tendency of increase
with increasing void ratio.
Influence of State Variables on the Shear Behaviour 183
Fig. 2. Variation of the dilatation parameter D with initial void ratio at various
combinations of degree of saturation and stress state: (1) Sr = 0.85, Iσ = 300 kPa;
(2) Sr = 1, Iσ = 300 kPa; (3) Sr = 0.9, Iσ = 900 kPa; (4) Sr = 0.85, Iσ = 900 kPa
4 Conclusions
A constitutive model capable of describing the nonlinear elasto-plastic stress-
strain-relationship for partially saturated cohesive soils is presented. Volume
change behaviour during deviatoric loading is also included in the present
formulation while prediction of collapse is not possible. The effects of initial
void ratio and degree of saturation are assessed on the basis of triaxial tests.
Further tests are in progress to investigate the influence of degree of satu-
ration and overconsolidation ratio on the parameters defining the yield surface
as well as the dilatation parameter.
Acknowledgment
The presented investigations have been carried out as a part within the re-
search group “Mechanics of partially saturated soils” funded by the German
Science Foundation (DFG). This support is greatly appreciated.
References
Alonso EE, Gens A, Josa A (1990) A constitutive model for partially saturated soils,
Becker A (2002) Stoffmodell und numerisches Modell für zyklisch beanspruchte,
teilgesättigte Sande. Dissertation, FG Bodenmechanik & Grundbau, TU Kaiser-
slautern
Bishop AW (1959) The principle of effective stress, Tecknish Ukebland 106(39):

859–863
Blatz JA, Graham J (2003) Elastic-plastic modelling of unsaturated soil using results
from a new triaxial test with controlled suction, Géotechnique 53(1):113–122
Khalili N, Geiser F, Blight GE (2004) Effective stress in unsaturated soils: Review
with new evidence, Int J Geomech, 4(2):115–126
Schwarz V, Becker A, Vrettos C (2006) An initial study on the viscous behaviour of
a partially saturated kaolinite clay based on triaxial tests. In: Miller GA, Zapata
CE, Houston SL, Fredlund DG (eds) Proc. 4th Int Conf on Unsaturated Soils,
Arizona, ASCE, Geotechnical special publication No. 147, pp 1811–1820
Stutz P (1973) Comportement elasto-plastique des milieux granulaires. In: Sawczuk
A (ed) Foundations and Plasticity. Noordhoff, Leyden, pp 37–49
Weckbecker W (1993) Beitrag zur Berechnung oberflächennaher Tunnel. Disserta-
tion, Fachgebiet Bodenmechanik & Grundbau, TU Kaiserslautern
Wheeler SJ, Sivakumar V (1995) An elasto-plastic critical state framework for un-
saturated soil, Géotechnique 45(1):35–53
Effect of Capillary and Cemented Bonds
on the Strength of Unsaturated Sands
Fabien Soulié, Moulay Saïd El Youssoufi, Jean-Yves Delenne,

and Christian Saix
Laboratoire de Mécanique et Génie Civil, UMR UMII-CNRS 5508, Université

Montpellier II, cc048, place E. Bataillon, 34095 Montpellier cedex 5, France
soulie@lmgc.univ-montp2.fr, elyous@lmgc.univ-montp2.fr,
delenne@lmgc.univ-montp2.fr, saix@lmgc.univ-montp2.fr
Summary. The cohesive interactions between grains play a prevailing role in the
mechanical behaviour of unsaturated granular materials such as fine sands. These
interactions are generally bonds of various natures that evolve according to the
surrounding hygrothermic conditions. We study the case where the liquid present in
the material is a water solution saturated with sodium chloride. The bonds are then
of capillary type and the cohesive interactions are mainly attractive. In this case,
the mechanical strength in an unconfined compression test is relatively low. At low
relative humidity, the phase change of water involves a crystallization of salt at the
contact points between grains generating thus bonds of solid type. The mechanical
strength of the material is thus enhanced. An experimental study of the variation
of the mechanical strength during the crystallization of salt allowed us to show two
distinct cohesive regimes: capillary and cemented. The transition between these two
regimes does not seem to be correlated with the mass of the crystallized salt, but
rather with the residual degree of saturation. An analysis of these results is proposed
by comparison with numerical simulations based on the discrete element approach.
Key words: capillarity, cementation, evaporation, compression test
1 Introduction
In the case of granular soils such as sands, the interstitial liquid can be at
the origin of the macroscopic cohesion of material. For low contents of liquid,
liquid bridges are formed between grains. The hydrous field corresponding to
these contents of liquid is often described as “pendular”. The presence of liq-
uid bridges between grains generates attractive capillary forces that appear
at the macroscopic scale by a cohesion which is nevertheless relatively weak
(Soulié et al. 2006). Under the effects of the surrounding hygrothermic con-
ditions, water can evaporate, thus modifying the concentration in solutes. In
cases of extreme concentration, these solutes crystallize, which leads to the
186 F. Soulié et al.
formation of local cemented bonds. This phenomenon creates an improved

cohesion at the macroscopic scale, directly related to the cementation degree.
The macroscopic mechanical behaviour of granular materials thus results from
local phenomena that develop at a local scale. The knowledge of the evolu-
tion of the cohesive bonds from a liquid state (liquid bridge) to a solid state
(cemented bond) constitutes a major stake for a better understanding of the
mechanical behaviour of these materials.
At the local scale, the study of the formation of liquid bridges and capillary
forces generated between grains gave rise to an extensive literature and several
models of capillary cohesion were proposed (Soulié et al. 2006, Mikami et al.
1998, Willett et al. 2000). The case of cemented bonds was also considered
and models of mechanical behaviour of cemented bonds were proposed at the
local scale (Delenne et al. 2004).
The objective of this work is to study the mechanical behaviour of a cohe-
sive bond the nature of which evolves according to the environmental condi-
tions. If it is possible to follow, at a local scale, the evolution of a cohesive bond
from the liquid state to a solid state during the progressive evaporation of sol-
vent (Farber et al. 2003), it is particularly difficult to study in a same time the
evolution of its mechanical behaviour. For this reason, in a first approach, we
were interested in the evolution of the macroscopic cohesion of a wet granular
material during the evaporation of solvent (water) and thus during the crys-
tallization of the solute (sodium chloride). In this first approach, cohesion will
be characterized by the compressive strength of cylindrical granular samples.
2 Experiments: Preparation of the Samples and Protocol

of Test
The samples are prepared with Ventoux sand. The sand is first washed and
dried, and then filtered to keep only grains the diameters of which are in the
range 0.4 mm to 0.8 mm (Fig. 1a). This granular material is closely mixed
with an aqueous solution including water and a solute that can crystallize
during the evaporation of water. The aqueous solution used is saturated with
sodium chloride (NaCl) and named in the following saturated brine. This
solution contains 35.6 g of NaCl for 100 g of water. The sensitivity of this
Sample
Support
(a) (b)
Fig. 1. (a) Ventoux sand, (b) sample just after preparation

Capillary and Cemented Bonds in Unsaturated Sands 187
maximum concentration to temperature is low (Kaufmann 1960, Langer and

Offermann 1982). The initial gravimetric content of saturated brine of the
granular samples is obtained by mixing in a tight box the suitable mass of
saturated brine and mass of dry granular material. The samples (diameter
25 mm and height 17 mm, Fig. 1b) are prepared in a cylindrical mould.
Hence, each sample is initially a three-phase material including:
• a solid phase (denoted by S) composed of grains (sand),
• a liquid phase (denoted by L) composed of water and dissolved NaCl of
mass mLNaCl .
• a gaseous phase composed of air and water vapor with negligible mass
compared to the two other phases.
In this initial state, the samples are particularly delicate to handle since
their cohesion and mechanical strength are solely ensured by weak capillary
forces between grains. Because of this brittleness, particular experimental pre-
cautions were necessary in order to avoid damage to the samples.
According to hygrothermic conditions of the atmosphere, water can evapo-
rate. This leads to a thermodynamic disequilibrium of the solution that results
in an excess of NaCl in an already saturated solution. Thermodynamic equilib-
rium is recovered through the crystallization of NaCl and thus the formation
of solid NaCl deposits of a total mass mSNaCl .
Three series of samples are prepared respectively for three liquid contents
of saturated brine: 3%, 5% and 7%. These liquid contents are selected in the
hydrous field corresponding to the “pendular state”, for which liquid bridges
between grains ensure the mechanical integrity of the sample. The samples are
weighed just after preparation (the corresponding mass is noted mi ), then put
to dry under conditions controlled in temperature T = 20◦ C and in relative
humidity RH = 43%.
The samples are taken at various durations varying from 15 minutes to
20 hours, then weighed (mass mf ), before being submitted to an unconfined
compression test until rupture. The difference in mass (mf − mi ) enables us
know the mass of evaporated water and thus the mass of crystallized NaCl.
The unconfined compression test is carried out using a “low capacity” press
allowing to test samples with low dimensions and with forces up to 50 N with
an accuracy of 0.01 N.
3 Results and Discussions

The mass of “retained” water in the sample, even after a long time of drying,
is proportional to the initial content of saturated brine, and thus to the initial
mass of water. For each of the three initial contents of saturated brine, only
93% of the mass of water initially introduced can evaporate. The remainder
“is trapped” during the crystallization of NaCl.
The observation of the samples makes it possible to highlight three different

localizations of the NaCl deposits:
• a crystallization at the surface of samples, which corresponds to the phe-
nomenon of efflorescence,
• at the surface of the grains, isolated crystals are formed and growth but
they do not directly contribute to the creation of cohesive bonds between
grains,
• at the contacts between grains, crystals are mobilized in the formation of
solid bridges, and then directly contribute to intergranular cohesion.
The tests were carried out on 60 samples. The “low capacity” press allows
us to obtain the evolution of the compressive force according to the axial
displacement.
Figure 2 presents typical curves of the compressive force versus the axial
displacement, for samples with an initial content of saturated brine equals
to 5%, for a short time of drying (15 min) and a long time of drying (20 h).
The curve obtained for short times of drying shows a continuous evolution
of the compressive force with the axial displacement. The deformation of the
sample is progressive during the test, the behaviour appears to be of ductile
plastic type. The values obtained for the compressive force are relatively low
and correspond to a cohesion of capillary type. In the case of a long time
of drying, we can observe a peak force with a significant intensity and for
a weak axial displacement. The behaviour of this type of samples (cemented
ones) is of fragile type: a clear rupture of the sample appears at the same time
as the peak force during the compression test. In both cases, the maximum
compressive force can be defined without ambiguity. All the samples have the
same geometrical configuration, and the maximum compressive force can so
be used to characterize the mechanical strength of the granular materials.
Figure 3 presents a synthesis of the 60 tests carried out for the contents
of saturated brine 3%, 5%, and 7%. This figure shows the evolution of the
maximal force according to the mass of crystallized NaCl. We can note that,
whatever the initial content of saturated brine, the obtained results have the
1.2 30
Compressive force (N)
Compressive force (N)
1 25
0.8 20
0.6 15
0.4 10
0.2 5
0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
(a) Axial displacement (mm) (b) Axial displacement (mm)
Fig. 2. Compressive force versus axial displacement for samples with an initial
content of saturated brine equals to 5%. Duration of drying: (a) 15 min, (b) 20 h
Fig. 3. Maximal compressive forces versus mass of crystallized NaCl for the three
initial contents of saturated brine
same feature. At the initial state, the mechanical strength of the granular
samples results from the only phenomenon of capillary cohesion, which gets a
low cohesion to the material and thus a low rupture strength. Independently
of the initial brine content, the rupture strength remains low in spite of the
increasing mass of crystallized NaCl. When it reached a critical value, different
from the initial brine content, the rupture strength increases considerably
reaching values from 30 to 35 times higher than the initial strength.
We can note that the mechanical strength of the granular samples does
not depend on the quantity of crystallized NaCl. Indeed, Figure 3 shows,
for a same mass of crystallized NaCl, the maximal compressive force is not
necessary the same, it also depends on the initial content of saturated brine.
In order to limit the influence of the initial content of saturated brine, we
define a crystallization index ic :
mSNaCl
ic = (1)
mNaCl
where mNaCl is the NaCl mass present in the sample (mNaCl = mSNaCl +
mLNaCl ). This index is equal to 0 if NaCl is completely dissolved and equals
to 1 if all the NaCl is crystallized. The variation of the maximal compressive
force according to the crystallization index is presented in Fig. 4 for the three
initial contents of saturated brine. This figure clearly highlights the existence
of two zones:
• the first zone for which the maximal force seems to be constant and thus
does not depend on the mass of crystallized NaCl,
• the second zone for which the maximal force sharply increases for ic ≥ 0.85.
40
35
Maximal force (N) 30
25
20
15
10
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Crystallization index ic
Fig. 4. Maximal compressive force versus crystallization index for the three initial
contents of saturated brine (w = 3%, 5% and 7%)
It is remarkable that these zones are well defined whatever the initial content
of saturated brine. They correspond to two distinct modes: the capillary mode
(first zone) and the cemented mode (second zone). It is also interesting to note
that the limit value of the crystallization index (i.e. ic = 1) is never reached.
This should be related to the limit value of the mass of water being able to
evaporate, which corresponds to 93% of the total mass of water introduced
into the sample.
Observations at the local scale of the evolution of bridges showed that a
change of nature, from liquid to solid, is a slow complex process. NaCl crys-
tals are initially formed at the liquid-gas interface on grains; these crystals
enlarge and support the growth of other crystals and then form a crystalline
area which preferentially grows along the liquid-gas interface. This evolution
of the cohesive bridges is illustrated in Fig. 5. This type of evolution was also
observed for other types of aqueous solutions frequently used by pharmaceu-
tical industry (Farber et al. 2003). Solid cohesion is effective only when the
connectivity of the crystallized deposit is established between the two grains,
which explains the importance of the capillary zone (first zone in Fig. 4). This
importance is also due to the existence of the phenomena of efflorescence and
Fig. 5. Schematic representation of the evolution of a bridge of saturated brine

isolated crystallization which form deposits of crystallized NaCl that do not

directly contribute to intergranular cohesion.
4 Numerical Approach
A first numerical study is carried out for a better understanding of the be-
haviour of unsaturated sands during drying. To achieve this, a 2D numerical
code, based on the Discrete Element Method initially developed by Cundall
and Strack (1979), is used to simulate an unconfined compression test on a
2D polydisperse granular sample the configuration of which is similar to the
experimental samples. In addition to the traditional contact and friction in-
teractions between grains, explicit models, formulated at the local scale, allow
us to account for the limiting behaviours observed in Sect. 3: respectively, the
model of capillary cohesion (Soulié et al. 2006) and of cemented cohesion (De-
lenne et al. 2004). In the central zone of the samples, the cohesive interactions
are of the capillary type, while in the peripheral zone, cohesion is governed by
the model of cementation. The thickness of the peripheral zone increases with
time, it follows the crystallization front that progresses from the periphery
towards the heart of the sample. The rupture thresholds of cemented bonds
increases with time, thus representing the hardening of former bonds.
Figure 6 shows the evolution of the normalized maximal force versus the
crystallization index both for the experimental case and for the numerical
case. The normalized maximal force corresponds to the obtained maximal
macroscopic force divided by the maximal macroscopic force at the initial state
(just after the preparation of the sample, when the integrity of the sample is
only ensured by capillary cohesion). Even if some discrepancies exist, this first
Fig. 6. Numerical results versus experimental results

numerical study in a 2D configuration supplies results in good agreement with

experimentally obtained data. Such a kind of approach should be extended to
other configurations, especially to 3D situations.
5 Conclusion
An experimental study made it possible to estimate the effect of the crys-
tallization of a solute (NaCl) on the mechanical resistance of an unsaturated
sand. The results obtained clearly showed the existence of two markedly dis-
tinct modes: a capillary mode and a cemented mode. In order to characterize
the dependence between mechanical resistance and crystallization, the impor-
tance of the crystallization index ic in the evolution of the maximal compres-
sive force was highlighted.
The influence of the crystallization index was underlined by the success-
ful comparison of experimental results and numerical results supplied by a
series of 2D simulations. A thorough study should allow us to complete this
analysis in order to specify laws of macroscopic behaviour including physico-
chemical parameters. The contribution of the discrete approaches is partic-
ularly promising to study the scale-up from the physical phenomena at the
microscopic scale to the observable behaviour at the macroscopic scale. The
objective is to propose a local law of cohesion in the case of a cohesive bond
the nature of which evolves according to the environmental conditions. The
central idea is to generalize the methodologies employed in the cases of capil-
larity (Soulié et al. 2006, Willett et al. 2000) and cementation (Delenne et al.
2004, Jiang et al. 2006) to a unified representation of the cohesive bond inter-
polating between these two limits with internal variables that evolve according
to environmental conditions.
References
Cundall PA, Strack ODL (1979) A discrete numerical model for granular assemblies,
Delenne JY, El Youssoufi MS, Cherblanc F, Bénet JC (2004) Mechanical behaviour
and failure of cohesive granular materials, Int J Numer Anal Meth Geomech
28:1577–1594
Farber L, Tardos GI, Michaels JN (2003) Evolution and structure of drying material
bridges of pharmaceutical excipients: Studies on a microscope slide, Chem Eng
Sci 58:4515–4525
Jiang MJ, Yu HS, Harris D (2006) Bond rolling resistance and its effect on yielding of
bonded granulates by DEM analyses, Int J Numer Anal Meth Geomech 30:723–
761
Kaufmann DW, editor (1960) Sodium chloride, the production and properties of salt
and brine. Reinhold Publishing Corporation, New York
Langer H, Offermann H (1982) On the solubility of sodium chloride in water, J Cryst

Growth 60:389–392
Mikami T, Kamiya H, Horio M (1998) Numerical simulation of cohesive powder
behavior in a fluidized bed, Chem Eng Sci 53(10):1927–1940
Soulié F, Cherblanc F, El Youssoufi MS, Saix C (2006) Influence of liquid bridges on
the mechanical behaviour of polydisperse granular materials, Int J Numer Anal
Meth Geomech 30:213–228
Willett CD, Adams MJ, Johnson SA, Seville JPK (2000) Capillary bridges between
two spherical bodies, Langmuir 16:9396–9405
Determining the Shear Strength
of Unsaturated Silt
Shulin Sun1 and Huifang Xu2

1
Department of Geological and Geotechnical Engineering, College of Civil
Engineering, Hohai University, 1 Xikang Road, Nanjing 210098, China
slsun@hhu.edu.cn
2
Department of Geology and Geophysics, University of Wisconsin-Madison,
Madison, WI 53706, USA hfxu@geology.wisc.edu
Summary. In order to determine shear strength of fine-grained unsaturated soils,

a new design of a shear box for a modified direct shear device is described where
a desired matric suctions is added easily with a hanging column. Simple principles
were employed to develop the direct shear testing device so that it would be robust,
easy to use, and readily assembled and disassembled for testing and maintenance.
Tests were conducted with sandy silts under several net normal stresses (σn ua ) for
different desired matric suction (ua − uw ). The results show that unsaturated shear
strength and friction angle (φb ) is consistent with results from prediction procedure
(Fredlund and Barbour 1996) and estimated friction angle (φb ) based on a soil-water
characteristic curve (SWCC) of the sandy silt, respectively.
Key words: unsaturated soil, matric suction, direct shear, shear strength, shear
box
Introduction
Shear strength of unsaturated soils can be determined by using both labo-
ratory tests and empirical methods. Some experimental methods were con-
ducted by using both the triaxial equipment modified to allow the control
or measurement of pore air pressure (ua ) and pore water pressure (uw ) (Ho
and Fredlund 1982, Toll 1990, Rassam and Williams 1999), and the modi-
fied direct shear devices with suction controls (Escario and Saez 1986, Gan
et al. 1988, Nishimura and Fredlund 2001). The experimental studies were of
time consuming and difficult to perform although the experiment results were
found to be consistent with the shear strength equation proposed by Fredlund
et al. (1978) as follows:
τf = c + (σn − ua ) tan φ + (ua − uw ) tan φb (1)
where
196 Shulin Sun and Huifang Xu
• τf is the shear strength of an unsaturated soil;

• c is the effective cohesion of the soil;
• φ is the effective angle of shearing resistance for a saturated soil;
• φb is the angle of internal friction with respect to the matric suction;
• σn is the total stress;
• ua is the pore air pressure;
• uw is the pore water pressure;
• (ua − uw ) is the matric suction;
• (σn − ua ) is the net normal stress.
Due to the above reason, several semi-empirical shear strength functions
were proposed to predict the shear strength of unsaturated soils (Vanapalli
et al. 1996, Fredlund and Barbour 1996, Oberg and Sallfors 1997, Khalili and
Khabbaz 1997, Bao et al. 1998). Rohm and Vilar (1995) proposed a hyper-
bolic relationship to fit matric suction and the experimental strength data.
Abramento and Carvalho (1989) used an experimental function as a curve-
fitting technique for their experimental data. Fredlund and Barbour (1996)
provided a means of estimating the shear strength function from the soil-water
characteristic curve by using the saturated shear strength parameters as the
starting values. These empirical procedures or/and approaches described in
the literatures for predicting the shear strength of unsaturated soils, based
on limited number of experimental data, may or may not be suitable for all
types of soils. This paper is to describe a simple experimental technique to
determine the shear strength of unsaturated, fine-grained soils by using mod-
ified direct shear testing device that was conventionally used for measuring
the shear strength of saturated soils.
A modified direct shear testing device described in this paper is different
from those described in literatures (Escario 1980, Ho and Fredlund 1982, Gan
et al. 1988, Escario and Juca 1989, Wheerler and Sivakumar 1992, Vanapalli et
al. 1996). Gan et al. (1988) proposed control of matric suction with variation
of pore air pressure (ua ) and a pore water pressure (uw ) as zero. During a
shear testing with Gan et al. (1988) method, a matric suction (ua − uw ) was
positive. Cares must be taken to avoid contamination a high air-entry porous
disk with vacuum grease and to ensure there were no air leaks in a ceramic
disk (Vanapalli et al. 1996). In this paper, the modified direct shear testing
device is easy to measure the shear strength of unsaturated soils with different
pore water pressures (uw ). A principle that applies desired matric suction to
saturated soil specimen is based on the method in ASTM D 6836–03 (2004)
and leak-free pressure plate extractor for measuring the SWCC proposed by
Wang and Benson (2004), and the matric suction (ua −uw ) is negative pressure
during shear testing.
Determining the Shear Strength of Unsaturated Silt 197
Modified Direct Shear Testing Device

A modified direct shear testing device for measuring shear strength of unsat-
urated soils is similar to direct shear testing for measuring shear strength of
saturated soils. It consists of shear loading device, normal loading device, shear
force measurement device, and horizontal and vertical displacement measure-
ment device (Fig. 1). The shear loading device is applied to the specimen
with sufficient capacity and controls to deform the specimen at a uniform
rate of displacement with an electric motor. A shear force is determined by a
shear force sensor. The normal loading device is capable of maintaining con-
stant force for the entire testing duration, and a normal force is applied by a
lever loading yoke which is activated by fixed weights. Horizontal and vertical
displacement is determined by using displacement transformers.
Shear box. There is a difference between an unsaturated direct shear testing
apparatus and a saturated direct shear testing device. The shear box used in
unsaturated direct shear testing is able to maintain desired matric suction in
the soil specimen during shear testing. It is made of bronze with an outflow
port at the bottom for drainage and control of a matric suction (Fig. 2). The
Fig. 1. (a) Flexible boundary direct shear test setup. (b) A negative-pressure un-
saturated direct shear testing apparatus
Fig. 2. A Photograph of the unsaturated shear box
shear box is divided vertically by a horizontal plane into two halves of equal
thickness which are fitted together with alignment screws. The shear box is
also fitted with gap screws, which control the space (gap) between the top and
bottom halves of the shear box. The bottom half consists of a Buna-N square
O-ring, a ceramic disk, and upper and lower chambers arranged with screws.
The Buna-N square O-ring (Wang and Benson 2003), is placed at about the
intermediate edge of the ceramic disk between the upper chamber sidewall
and the lower chamber base.
Hanging column Adding a matric suction. A saturated specimen adds
a suction with hanging column, which is conducted following method A in
ASTM D 6836–03 (2004), Standard Test Methods for Determination of the
Soil Water Characteristic Curve for Adsorption Using a Hanging Column.
The shear box is attached to a small-bore horizontal tube for adding desired
matric suction directly by applying a vacuum to the water in the horizon-
tal measurement tube using water reservoirs set at different elevations before
shear testing (Fig. 1(b)). During the shear testing, the pore air pressure re-
mains at atmospheric pressure and the small-bore horizontal tube measures
outflow (Fig. 1(b)).
Shear Device, Loading, Shear force measuring, Vertical displacement, and
Horizontal displacement as specified in Test Method ASTM D 6836–03 (2004).
Procedures
Before placing the saturated specimen in the shear box, a ceramic disk is
saturated by flooding the base of the shear box with some water. The ceramic
disk is maintained in water until the removal of air in the ceramic disk, and
the water is flushed out through a water drain tube below the ceramic disk.
The shear box is dried by using a sponge except the saturated ceramic disk
and the water drain tube. Petrolatum is used at the interface of shear surfaces
to ensure that the water did not leak out and friction force at the contacted
interface between upper and lower parts of the shear box is set zero.
The saturated soil specimen is placed carefully into the shear box, and then
before the desired matric suction applied to the specimen, the horizontal tube
is connected with the water drain tube and reading of a air-water interface in
the horizontal tube on the scale is recorded after the horizontal tube is poured
with small amount of water.
A matric suction ranges provided by the hanging column, is between 0
and 35 kPa. The desired matric suction (ua − uw ) is added to the specimen by
moving a lower reservoir and changing water level showed on a manometer,
and the specimen was gradually desaturated with a outflow of water in the
specimen through the water drain tube. A reading of the air-water interface
is taken when the outflow of water in the specimen ceases. A desired vertical
loading does not apply to the specimen until the air-water interface forwarded
along the horizontal tube is stable for 1 hr.
When vertical loading is applied to the unsaturated specimen, the water
in specimen begins to flow out and the air-water interface moves forward.
It may take for about 1 hr that the water stops flowing out, and reading of
the interface is taken. After readings from sensors that measure shear force,
horizontal and vertical displacement, were checked out, a shear test began
with a displacement rate of 0.0096 mm/min. Each unsaturated shear testing
takes about 14 ∼ 15 hr with a matric suction and a vertical loading values for
each time.
Testing Experiment
Saturated specimen preparation. The 150.6 g dry weight of Boardman soils was
taken for the test (Table 1). According to calculation, the water content of a
saturated soil specimen is 29.7%, and the weight of water adds to the dry soil
specimen for testing is 44.7 g. A saturated soil specimen is made by mixing
the dry soil specimen with 44.7 g of water.
Shear box preparation. A ceramic disk at the bottom of a shear box is
saturated by filling water into the lower part of the shear box and removing
air in the ceramic disk, and a water drain tube connected to the ceramic disk
Table 1. Properties of Boarsman Soil for Tests
Specific Dry unit Water content Weight

Type of Dry weight Plasticity
gravity weight γd of saturated of water
a soil Ws (g) ωP (%)
Gs (kN/m3 ) soil (%) added (g)
Sandy
150.6 2.65 14.52 23 29.7 44.7
silt
is filled with water. An upper part of the shear box is fit to the lower part
with screws after petrolatum is used on the contact surface between the upper
and the lower. The saturated soil specimen is then put carefully into the shear
box.
Adding a desired matric suction. The water drain tube is connected with a
horizontal tube filled with water after the shear box is put on a shear machine.
Desired matric suction is added to the saturated soil specimen by decreasing
a highness of a lower reservoir, and the soil specimen is gradually changed
into unsaturated soil specimen with an outflow of water along the water drain
tube. The desired matric suction is successfully applied in an hour later after
the outflow is ceased. A soil water characteristic curve (SWCC) of the sandy
silt was measured and is shown in Fig. 7.
Applied a normal stress. A desired normal stress is applied to the unsat-
urated soil specimen by using fixed weight while the water was out flowing
along the water drain tube. It needs to wait about 1 hr for applying the normal
stress.
Shear testing. A shear force sensor, two horizontal and vertical displace-
ment sensors are checked, and screws fitted to the shear box is taken off before
shear testing began.
Testing results. The unsaturated direct shear testing is conducted by using
four net normal stresses with three matric suctions. The variation of shear
strength with net normal stress at different matric suction was displayed in
Fig. 3. The x-axis represents the net normal stresses and the y-axis represents
the unsaturated shear strength. All the experimental data is shown by using
different symbols. Figure 3 shows the straight line with a slope of about 36.9◦ ,
which is independent of the matric suction.
Fig. 3. Shear strength versus net normal stress with different matric suction
Fig. 4. Shear strength versus matric suction with different net normal stress
Figure 4 presents shear strength in terms of the variation with applied

matric suction at different normal stress. The x-axis represents the matric
suction, and the y-axis represents the shear strength. The variation in shear
strength with respect to matric suction is linear at lower matric suction range,
and nonlinear at higher matric suction range. The curves with different normal
stress show same trend. The slope of the lines is about 36.6◦ , and the slope
of the nonlinear is less than 36.6◦ . The shear strength measured is compara-
ble with results from prediction procedure by Fredlund and Barbour (1996).
Figure 5 indicates the tendency of measured shear strength basically fits the
predicted values.
Fig. 5. Comparison between measured and predicted variation of shear strength

versus matric suction
Measured frictional angle φb . According to the results described above

(Fig. 3 and Fig. 4), the parameters for the shear strength of unsaturated
sandy silt are listed in Table 2 where net normal stress (σn − ua ) and matric
suction (ua − uw ) are applied to the unsaturated soil specimens, and the shear
strength (τf ) is calculated by using shear force, which is measured with the
sensor (Fig. 1) during testing, divided by the shear area. The shear strength
(τf ) equals saturated shear strength (τs ) when applied desired matric suction
is zero. Unsaturated shear strength (τψ ) is equal to shear strength (τf ) minus
saturated shear strength (τs ). The cohesive strength (c) taken from Fig. 3 is
intercept with the τf axis (Fig. 3). The friction angle (φb ) with respect to
matric suction is calculated with the equation (2). An average of the effective
friction angle (φ ) for the saturated soil is 36.6◦ . Figure 6 indicates the vari-
ation of φb with the matric suction. It should be noted that the φb decreases
gradually as the increasing of the desired matric suction, and the φb is less
than the φ of the saturated soil.

τf − [c + (σn − ua ) tan φ ]
φb = arctan (2)
ua − uw
Table 2. Measured parameter of the shear strength for the unsaturated sandy silts
Normal stress Suction Shear

(σn − ua ) (ua − uw ) stress τf τsat τψ c tan φ φ tan φb φb
(kPa) (kPa) (kPa)
12.3 0 16.17 16.17 0.00 6.70 0.77 37.59
12.3 11 22.76 16.17 6.58 6.70 0.77 37.59 0.60 30.89
Test 1 12.3 22 32.08 16.17 15.90 6.70 0.77 37.59 0.72 35.86
12.3 28 33.50 16.17 17.33 6.70 0.77 37.59 0.62 31.75
12.3 35 38.56 16.17 22.38 6.70 0.77 37.59 0.64 32.60
24.6 0 24.14 24.14 0.00 6.70 0.71 35.33
24.6 11 30.95 24.14 6.81 6.70 0.71 35.33 0.62 31.75
Test 2 24.6 22 39.52 24.14 15.38 6.70 0.71 35.33 0.70 34.96
24.6 28 42.58 24.14 18.44 6.70 0.71 35.33 0.66 33.37
24.6 35 47.75 24.14 23.60 6.70 0.71 35.33 0.67 34.00
49.2 0 43.89 43.89 0.00 6.70 0.76 37.09
49.2 11 50.80 43.89 6.91 6.70 0.76 37.09 0.63 32.14
Test 3 49.2 22 58.03 43.89 14.14 6.70 0.76 37.09 0.64 32.73
49.2 28 64.19 43.89 20.30 6.70 0.76 37.09 0.73 35.94
49.2 35 68.13 43.89 24.24 6.70 0.76 37.09 0.69 34.71
98.4 0 79.60 79.60 0.00 6.70 0.74 36.53
Test 4 98.4 22 95.47 79.60 15.87 6.70 0.74 36.53 0.72 35.80
98.4 35 103.97 79.60 24.37 6.70 0.74 36.53 0.70 34.85
Fig. 6. Measured and estimated friction angles φb
Fig. 7. The SWCC of the sandy silt
Estimated friction angle φb . The friction angle φb was estimated from the
soil-water characteristic curve (SWCC). According to the estimation the shear
strength function provided by Fredlund and Barbour (1996), the estimated
formula of φb is followed:
k
θ − θ
tan φ
r
φb = arctan
θs − θr
or
φb = arctan (Θ)k tan φ (3)
where ψ = matric suction applied to soil specimens; θ = the volumetric

water content at any suction (ua − uw ); θr = the volumetric water content
at residual conditions; θs = the volumetric water content at saturation, Θ =
(θ − θr )/(θs − θr ), the normalized volumetric water content; k = 1, 2, and 3.
The SWCC of the sandy silt was measured with a hanging column ap-
paratus. A saturated specimen was prepared and suction (ψ was applied to
remove the water from the soil. The change in the volumetric water content
was monitored with increasing suction. Then the suction was slowly decreased
and the increase in volumetric water content was recorded. The Brooks-Corey
and van Genuchten models were fit to the SWCC data using a least squares
optimization scheme with Solver in Excel. The SWCC is shown in Fig. 7, and
the results of the friction angle φb estimated are listed in Table 3 illustrated
in Fig. 6. Figure 6 indicates the variation of the friction angle φb with matric
suction at different k = 1, 2, 3, which is consistent with the measured φb , but
the estimated friction angle is smaller than the measured angle.
Table 3. Estimated parameter of the shear strength for the unsaturated sandy silts
k=1 k=2 k=3

ψ (kPa) θ θr θs Θ
tan φ b
φb
tan φ b
φb
tan φb φb
0 0.360 0.025 0.36 1 0.74 36.64 0.74 36.64 0.74 36.64
11 0.349 0.025 0.36 0.97 0.72 35.73 0.70 34.83 0.67 33.94
22 0.333 0.025 0.36 0.92 0.68 34.37 0.63 32.16 0.58 30.03
28 0.324 0.025 0.36 0.89 0.66 33.58 0.59 30.65 0.53 27.87
35 0.313 0.025 0.36 0.86 0.64 32.60 0.55 28.80 0.47 25.30
The shear strength contribution due to matric suction (τψ was plotted in
Fig. 5, which was equal to (ua − uw ) tan φb from experimental results (see
Eqn. (1)), or [(ua − uw ){(Θ)(tan φ )}] from the semi-empirical prediction pro-
cedure(see Eqn. (3)).The variation of shear strength with respect to different
matric suction is nonlinear for tested suction range.
This paper describes a new device for determining the shear strength of un-
saturated fine-grained soils. Simple principles were employed to develop the
direct shear testing device so that it would be robust, and easy to use, and
readily assembled and disassembled for testing and maintenance. A testing
of a sample was conducted with sandy silts under the net normal stresses
(σn − ua ) of 12.3 kPa, 24.6 kPa, 49.2 kPa, and 98.4 kPa for matric suctions
(ua − uw ) of 0 kPa, 11 kPa, 22 kPa, 28 kPa, and 35 kPa. The measured results
show that the unsaturated shear strength and the friction angle (φb ) with
respect to the matric suction are consistent with the results from prediction
procedure (Fredlund and Barbour 1996) and estimated friction angle (φb )
based on SWCC of the sandy silt, respectively.
Acknowledgments
The authors thank the support provided by colleague of the Geo-Engineering
Program and Geological Engineering Program, Department of Civil & En-
vironmental Engineering, University of Wisconsin-Madison, and for the help
of Mr. X. Wang in conducting laboratory testing, and for thoughtful and
insightful comments of Professor Craig H. Benson.
References
ASTM D 3080–03 (2003) Standard Test Method for Direct Shear Test of Soils Under
Consolidated Drained Conditions, Annual Book of ASTM Standards, ASTM
ASTM D 6836–03 (2004) Standard Test Methods for Determination of the Soil Wa-
ter Characteristic Curve for Adsorption Using a Hanging Column, Pressure Ex-
tractor, Chilled Mirror Hygrometer, and/or Centrifuge, Annual Book of ASTM
Standards, ASTM
Abramento M, Carvalho CS (1989) Geotechnical Parameters for the Study of Nat-
ural Slopes Instabilization at ‘Serra do Mar’ Brazil. In: Proc 12th Int Conf Soil
Mechanics and Foundations Engineering. Rio de Janeiro 3:1599–1602
Bao CG, Gong B, Zan L (1998) Properties of Unsaturated Soils and Slope Stability
of Expansive Soils, Key Note Lecture, UNSAT 98, 2nd Int Conf Unsaturated
Soils, Beijing
Escario V (1980) Suction controlled penetration and shear tests. In: Proc 4th Int
Conf on Expansive ASCE Publication. Denver 2:781–797
Escario V, Juca JFT (1989) Strength and deformation of partly saturated soils.
In: Proc 12th Int Conf on Soil Mechanics and Foundation Engineering. A. A.
Balkema, Bookfield, MA 2:43–46
Escario V, Saez J (1986) The shear strength of partly saturated soils, Geotechnique
36(3):453–456
Fredlund DG, Morgenstern NR, Widger RA (1978) The shear strength of unsatu-
rated soils, Can Geotech J 15(3):313–321
Fredlund MD, Barbour SL (1996) The Relationship of the Unsaturated Soil Shear
Strength Function to the Soil-Water Characteristic Curve, Can Geotech J
33(3):440–448
Fredlund DG, Morgenstern NR (1977) Stress State Variables for Unsaturated Soils,
ASCE J Geotech Eng Div 103(GT5):447–466
Gan JKM, Fredlund DG, Rahardjo H (1988) Determination of the Shear Strength
Parameters of an Unsaturated Soil Using the Direct Shear Test, Can Geotech J
25:500–510
Ho DYF, Fredlund DG (1982) Increase in shear strength due to suction for two
Hong Kong soils. In: Proc ASCE Geotech Conf, Engineering and Construction on
Tropical and residual soils. Honolulu, Hawai, U.S.A., January 11–15 pp 263–295
Khalili N, Khabbaz MH (1997) A unique relationship for χ for the determination of

the shear strength of unsaturated soils, Geotechnique 48(5):681–687
Nishimura T, Fredlund DG (2001) Failure envelope of a desiccated unsaturated silty
soil. In: Proc 15th Int Conf Soil Mech Geotech Eng. Istanbul 1:615–618
Oberg A-L, Sallfors G. (1997) Determination of shear strength parameters of un-
saturated silts and sands based on the water retention curve, Geotech Test J
20:40–48
Rassam DW, Williams DJ (1999) A relationship describing the shear strength of
unsaturated soils, Can Geotech J 36:363–368
Rohm SA, Vilar OM (1995) Shear strength of an unsaturated sandy soil. In: Alonso
EE, Delage P (eds) Proc 1st Int Conf Unsaturated Soils. Paris 1:189–193
Toll DG (1990) A framework for unsaturated soil behavior, Geotechnique 40(1):31–
44
Vanapalli SK, Fredlund DG, Pufahl DE, Clifton AW (1996) Model for the prediction
of shear strength with respect to soil suction, Can Geotech J 33(3):379–392
Wang X, Benson CH (2004) Leak-free pressure plate extractor for measuring the
soil water characteristic curve, Geotech Testing J 27(2):163–172
Wang X, Benson CH (2003) Pressure Plate Extractor, United States Patent
No. 6,718,835
Wheerler SJ, Sivakumar V (1992) Development and application of a critical state
model for unsaturated soils, In: Houlsby GT, Schofield AN (eds) Proc Wroth
Mem Symp Predictive Soil Mechanics, ASCE, New York. Oxford
Factors Affecting Tensile Strength Measurement
and Modified Tensile Strength Measuring
Apparatus for Soil
Surendra Bahadur Tamrakar1 , Toshiyuki Mitachi2 , and Yasuo Toyosawa1

1
Japan National Institute of Occupational Safety and Health, Kiyose, Tokyo,
Japan tamrakar@s.jniosh.go.jp, toyosawa@s.jniosh.go.jp
2
Graduate School of Engineering, Hokkaido University, Hokkaido, Japan
mitachi@eng.hokudai.ac.jp
Summary. In this paper tensile strength measuring apparatus developed by Tam-

rakar et al. (2005a,b) was used to measure the tensile strength of one dimensionally
consolidated saturated NSF-clay and statically compacted unsaturated mixtures of
NSF-clay, CFP-silt and Toyoura-sand. Tensile strength (qt ) obtained from the tensile
tests were compared with the unconfined compressive strength (qu ). It was observed
that the ratio qu /qt lied within the range of 2 to 3 for saturated NSF-clay and 4
to 16 for compacted mixtures. Effect of specimen thickness within the tensile mold,
number of compaction layers and tensile pulling rates on the tensile strength were
also examined. Comparing the specimen thickness within the tensile mold, it was
found that the specimen having 5 cm thickness gave the minimum value. Also, ten-
sile strength increased with the increase in the number of compaction layers. As in
other shear strength, increment in the tensile strength was observed beyond tensile
pulling rate of 0.34 mm/min. But below this pulling rate also, some increments were
observed.
Key words: tensile strength, tensile apparatus, compacted soil, tensile pulling rate
Introduction
Most of the vertical slopes get failed with the development of tensile crack on
the top of the slope. Also, many earth dams, embankments, pavements, etc.
where soil layers are compacted, are failed due to the development of tensile
cracks. Prediction of probable position and depth of tensile crack is necessary
to protect the property and loss of lives of workers at the construction site. In
order to explain the position and depth of tensile crack, an accurate measure-
ment of tensile strength of soil is necessary. Very few researches (e.g. Suzuki
et al. 1998, Yao et al. 2002, Ono et al. 2003) have been made to measure the
tensile strength of soils having lower tensile values. Recently, Nahlawi et al.
(2004) and Tamrakar et al. (2005a,b) have introduced a new tensile strength
208 S. B. Tamrakar et al.
measuring apparatus which measures the tensile strength directly. One devel-
oped by Nahlawi et al. (2004) could be mainly used for compacted clayey and
stiff soils only whereas the one developed by Tamrakar et al. (2005a,b) seems
to be easy to use and simple to handle and could be used for both compacted
unsaturated and highly saturated soils.
Tamrakar et al. (2005a,b) measured the maximum tensile strength of
Kanto loam around 50 ∼ 60% of water content and showed the ratio of un-
confined compression strength and tensile strength around 12.5 which varied
with the water content. They also showed the effect of amount of finer particles
and their size on tensile strength. Possible measurement of tensile strength for
saturated NSF clay was also shown.
In this paper, tensile apparatus (type-A tensile mold) developed by Tam-
rakar et al. (2005a,b) was used to measure the tensile strength of saturated
and unsaturated soils. Also, the effect of number of compaction layers, thick-
ness of the specimens and tensile pulling rate on the tensile strength was
studied. Unconfined compression tests were also performed to compare their
values with tensile strength.
Test Apparatus
Tensile test apparatus shown in Fig. 1 consists of horizontal platform upon

which apparatus box having two halves; fixed box and movable box, is placed.
Inside this box, two tensile molds are placed. The inner shape of this mold is
like “C” structure and it holds the specimen. Two molds are screwed to the
apparatus boxes separately. One box of the apparatus is fixed to the horizontal
platform while the other box can move freely on the horizontal platform. To
Fig. 1. New tensile strength measuring apparatus (inset : tensile molds) (Tamrakar
et al. 2005a,b)
Factors Affecting TSM and Modified TSM Apparatus for Soil 209
reduce the friction, linear sliding roller is placed between the movable box
and platform. Movable box is pulled away in horizontal direction until the
soil specimen fails in tension with tensile crack appearing at the middle of the
specimen where two halves of the mold is attached. Load cell placed between
the movable box and motor axis measures the tensile load. This tensile load
divided by the area of the tensile crack perpendicular to horizontal pulling
direction, gives the tensile stress. These molds can be easily changed as they
are connected to the main apparatus by the screws only. The total surface
area of this mold is 38.5 cm2 . The minimum width at the constricted section
of this mold is 3 cm and the depth is 5 cm.
The apparatus box along with the mold and platform can be completely
separated from the motor for preparing the specimen before the test. Com-
pacted soil specimen is prepared within this mold by direct static compression.
Once the specimen is ready within the mold for the test, then it is connected
to motor shaft. Between the motor shaft and movable apparatus box, there
are some attachments where load cell is kept.
Materials and Specimen Preparation
Kanto loam, NSF-clay and the mixtures of NSF-clay, CFP-silt and Toyoura-
sand were taken as test materials. NSF-clay is commercially available clay
which consists of Pyrophylite, CFP-silt (100) is crushed form of Silica sand
and Toyoura-sand is also commercially available standard Japanese sand.
Grain size distribution curves and index properties for these soils are shown in
Fig. 2 and Table 1. Now onwards, NSF-clay, CFP-silt and Toyoura-sand are
NSF-clay silt (CFP-100)

Toyoura sand
100
percent passing (%)
80
60
40
20
0
0.001 0.010 0.100 1.000
particle size (mm)
Fig. 2. Grain size distribution curves

Table 1. Properties of test materials
Materials Density Dry density wL wP

soil solid max min
g/cm3 g/cm3 g/cm3 (%) (%)
NSF-clay 2.78 55.1 30.6
Silt (CFP-100) 2.66 1.59 1.17
Toyoura sand 2.64 1.65 1.34
represented by clay, silt and sand, respectively. For saturated test specimen,
consolidated clay specimens were used where as mixtures of clay, silt and sand
in different proportions were used for unsaturated compacted specimens. Ta-
bles 2 and 3 show the proportions and test conditions for different mixtures.
Before preparing the specimens, at first, tensile molds were fixed into the
apparatus box and screwing was done between the movable box and apparatus
horizontal plate so that movable box would be fixed. To reduce the friction
between the specimen and the inner wall of the tensile mold, thin film of grease
was applied over its inner surfaces. After the insertion of the consolidated
specimen in case of saturated specimens or after the completion of compaction
in case of compacted specimens into the tensile mold, load cell is set up towards
the pulling side of mold box. Finally, the screws which are earlier fixed to
prevent the movement of movable box of the apparatus are un-screwed.
Table 2. Test materials for mixtures of clay, silt and sand
Dry density NSF-clay CFP-silt Toyoura-sand qu qt qu /qt

(g/cm3 ) % % % (kPa) (kPa)
25 — 75 21.1 3.2 6.5
40 — 60 55.9 6.6 8.4
1.50 50 — 50 74.4 7.8 9.5
60 — 40 79.1 8.6 9.2
75 — 25 131.3 12.1 10.9
25 75 — 64.3 6.9 9.3
40 60 — 100.0 8.4 11.9
1.50 50 50 — 97.8 8.5 11.5
60 40 — 132.2 10.3 12.9
75 25 — 182.2 11.7 15.5
— 25 75 6.6 1.4 4.6
— 40 60 12.9 2.2 6.0
1.40 — 50 50 16.4 2.7 6.1
— 60 40 18.7 3.0 6.2
— 70 30 26.7 3.9 6.9
Table 3. Mixing proportions and testing conditions for clay ∼ sand mixtures
Mixing w Controlled Testing

Specimens ratio dry density compaction dry density Conditions
(by wt.) (%) (g/cm3 ) stress (kPa) (g/cm3 )
clay ∼ sand 3:1 10.0 1.50 No. of layers(1)

clay ∼ sand 3:1 10.0 1.50 Thickness(2)
clay ∼ sand 1:3 10.0 200 1.54
Pulling rate(a)
clay ∼ sand 3:1 10.0 200 1.26
clay ∼ sand 3:1 10.0 1.50
Pulling rate(b)
clay ∼ sand 1:3 10.0 1.50
(1)
one, two and four layer-compaction, (2) 1.25, 2.5, 3.75 and 5 cm
(a)
0.01, 0.09 and 0.34 and 0.88 mm/min, (b) 0.17, 0.34 and 0.88 mm/min
Saturated specimens were prepared by pre-consolidating the slurry of the

clay material in a separate, special consolidation mold. Two-way drainage with
one dimensional consolidation was done. Once the consolidation was over,
consolidation mold with consolidated specimen was placed over the tensile
mold. Centering of molds was done by guiding support screws which were
attached to the fixed portion of the apparatus box. By pushing the shaft of
the consolidation mold slowly, consolidated specimen was allowed to insert
into the tensile mold. Once the full depth (5 cm) insertion was completed,
then the specimen was cut and its upper surface was trimmed.
In case of compacted specimens, at first, materials were thoroughly mixed
with required distilled water and kept in an air tight plastic bag and sealed
so that water was uniformly distributed throughout the materials. Specimens
were prepared either under constant stress or under constant dry density
conditions. In both the conditions, compacted specimens were prepared by
directly and statically compressing the prerequisite amount of soil kept within
the tensile mold of the apparatus, using bellophragm cylinder. Collar was
generally placed over the tensile mold to prevent falling out of soil from the
mold. Specimens were compacted keeping the dry density, water content and
thickness of the specimens constant. Thickness, number of compaction layers
and tensile pulling rates were varied depending upon the test conditions.
In case of unconfined compression test, saturated specimens were prepared
by pre-consolidating the clay specimens in an ordinary consolidometer where
as unsaturated compacted specimens were prepared in a normal splitting mold
either under constant stress or constant dry density condition. Generally, one
layer compaction was done. But to see the effect of numbers of layer of com-
paction, some specimens were prepared with one, two, three, four, five and
ten layers.
Test Conditions
Tests in which specimen thickness was maintained at 5 cm with one-layer com-
paction and pulled under 0.34 mm/min tensile pulling rate, were considered
as reference tests. All the tests of saturated specimens are reference tests. Sat-
urated specimens of clay were prepared under 100, 200 and 300 kPa. Water
content, density, degree of saturation, etc. is shown in Table 4.
Table 4. Saturated sol specimens conditions and test results
Specimen type NSF-clay

Preconsolidation pressure (kPa) 100 200 200 300 300
Soil unit weight (kN/m3 ) 27.23 27.23 27.23 27.23 27.23
Water content w (%) 55.51 45.03 46.65 45.61 42.84
Degree of saturation Sr (%)∗ 98.40 97.20 96.70 98.75 96.00
qt (kPa) 13.54 18.34 20.64 25.35 27.04
qu (kPa) 29.81 52.81 52.81 77.12 77.12
qu /qt 2.20 2.88 2.56 3.04 2.85
∗
for tensile test
In case of unsaturated specimens, test conditions were changed depending

upon the type of tests. Mixtures of clay, silt and sand specimens were prepared
by mixing them in different proportions as shown in Tables 2 and 3. Speci-
mens were compacted keeping their dry density, water content and specimen
thickness constant. In case of tests where the effect of specimen thickness was
studied, specimens were prepared by one-layer static compaction and they
were pulled with 0.34 mm/min. In this case specimen thickness varied from
1.25 to 5 cm. Similarly, where the effect of number of layers of compaction
was studied, specimens were prepared with one-layer, two-layers, three-layers
and four-layers of static compactions, keeping the overall specimen thickness
to be around 5 cm and pulling them under 0.34 mm/min. In case of the tests
where the effect of tensile pulling rate was studied, specimens were prepared
by one-layer static compaction with specimen thickness of 5 cm. In this case,
tensile pulling rate was varied from 0.09 to 1.75 mm/min.
Unconfined compression test for saturated specimens were prepared by
trimming the pre-consolidated specimens whereas unsaturated compacted
specimens were prepared using ordinary splitting mold. Other than those for
the effect of number of compacted layers, all the compacted specimens were
prepared by one-layer static compaction using bellophragm cylinder. Other
conditions such as dry density and water content were kept same as those for
tensile compacted specimens. Compacted layers were prepared with 1, 2, 3,
4, 5 and 10 layers. Compaction time allowed for each layer was around one
minute. The height and depth of specimen for both saturated and unsaturated
cases were 10 cm and 5 cm, respectively. Unconfined compression tests were

conducted at constant displacement rate of 0.1 mm/min.
Results and Discussions

Figure 3 shows the photographs before (a) and after (b) the tensile failure tests
for clay ∼ sand (3:1) and clay ∼ sand (1:3), respectively. Clear and straight
tensile crack (failure line) could be seen. In Figure 3 (c) and (d), failure
planes (tensile crack plane) after the tests were shown. Clear and smooth
failure surfaces could be seen.
Tensile stress ∼ displacement curves obtained for the saturated specimens
of clay (pre-consolidated under 100, 200 and 300 kPa) are shown in Fig. 4. As
shown, with the increase in consolidation pressure, there is increase in tensile
strength. Small variation in the test result might have occurred either dur-
ing the transferring of the consolidated specimen from the pre-consolidometer
to tensile mold or during the trimming of the specimen surface. In addition,
variation in the water content (degree of saturation) or the friction between
the inner wall of the consolidation mold and slurry material during the pre-
consolidation might also have affected small change in their strength. Hence-
forth, proper attention must be paid during the specimen preparation.
Stress ∼ displacement curves for clay ∼ silt ∼ sand (1:1:1) obtained
from tensile test and unconfined compression tests were shown in Figs. 5
(a) and (b), respectively. Clear peak for tensile stress as well as unconfined
compressive stress could be seen. Here, tensile stress measured was shown in
negative value. From now onwards, tensile strength values would be shown as
positive values.
Table 4 shows the tensile strength and unconfined compression strengths
obtained for saturated specimens. Increase in both tensile and unconfined
compression strengths with the increase in the consolidation pressure could
(c)
(a) (b)
(d)
Fig. 3. Photographs before and after the tensile failure. (a) and (c) clay ∼ sand
mixture (3:1) and (b) and (d) clay ∼ sand mixture (1:3)
100 kPa (Sr=98.4%) 150 kPa (Sr=97.21%)

200 kPa (Sr=97.2%) 200 kPa (Sr=96.7%)
300 kPa (Sr=98.75%) 300 kPa (Sr=96%)
tensile stress (kPa) 0
-5
-10
-15
-20
-25
-30
0.0 0.4 0.8 1.2 1.6 2.0
displacement (mm)
Fig. 4. Tensile stress ∼ displacement curves for consolidated NSF-clay

tensile stress (kPa)
0 60
UC stress (kPa)
-2 40
-4 20
-6 0
0.0 0.1 0.2 0.3 0.4 0.5 0.0 1.0 2.0 3.0 4.0
(a) displacement (mm) (b) displacment (mm)
Fig. 5. Stress ∼ displacement curves for clay ∼ silt ∼ sand (1:1:1) mixture (a)
tensile test and (b) unconfined compression test
be seen. It was observed that the ratio of qu /qt for saturated clay varied from
2 ∼ 3. Another type of NSF-clay used by Tamrakar et al. (2005a,b) had shown
the average ratio as 6. All the test specimens shown in Table 4 have more than
93% of degree of saturation. Degree of saturation (Sr ) shown in the Table 4
was calculated by using unit wet of soil solid, total weight of the specimen
inside the tensile mold and the water content of the specimen after failure. As
it is difficult to measure the area of the specimen directly, the total area of the
specimen was considered to be same as that of the tensile mold. Thickness of
the specimen was measured once the trimming was done after transferring the
consolidated specimen into the tensile mold. It was assumed that the specimen
tightly fits into the mold. Ratio of qu and qt for clay ∼ silt ∼ sand (1:1:1)
specimen was found to be 8.6. In Table 2, tensile and unconfined compression
strengths measured for different mixtures are shown. It could be seen that
the ratio of qu /qt for the unsaturated mixtures varied from 4 to 16. Tamrakar
et al. (2005a,b) had also measured the similar ratio of qu /qt for Kanto loam
which varied from 10 to 13 (Tamrakar et al. 2005a,b).
Effects of specimen thickness are shown in Fig. 6. Here, tensile strength
values for different specimen thicknesses are shown. Decrease in the tensile
strength with the increase in the specimen thickness could be seen. Specimen
having 5 cm thickness, i.e., the thickness of tensile mold, showed the mini-
mum tensile strength. Here, specimens were compacted only once for all the
thicknesses. Therefore, uniformity in the density is more in case of specimen
which had the minimum thickness. Irrespective of specimen thickness, tensile
pulling for each test was done from the mid-height of the tensile mold. With
the change in the thickness, resultant pulling direction might have changed,
hence affecting the strength.
The effect of number of layers of compaction to tensile strength is shown
in Fig. 7. During this test, 5 cm thick specimens were prepared by compact-
ing predetermined amount of specimen within the tensile mold in one-layer,
two-layer, three-layer and four-layer. Clay-sand mixture (3:1, w = 10% and
rd = 1.5 g/cm3 ) showed the increment in tensile strength with the increase
in the number of compaction layers. One-layer compaction gave the mini-
mum value than those obtained for two, three and four-layer compaction.
With the increase in number of compaction layers, more uniformity of density
through out the specimen takes place and this will increase the strength of
the specimen. Similar test result was seen in case of unconfined compression
test shown in Fig. 8 where test specimens were prepared by statically com-
pacting the same amount of clay-sand mixture (1:3, w ∼ 10%, compaction
pressure 50, 100 and 200 kPa) with 1, 2, 3, 4, 5 and 10 layers. As shown in
Fig. 8, at the beginning, sudden increase in qu strength was seen but the rate
of increment decreased with the increase in the number of layers. As obvious,
1.25 2.5 3.75 5 21

18
tesnile stress (kPa)
21 15
qt (kPa)
18 12
15
12 9
9 6
6 3
3
0 0
0.0 0.1 0.2 0.3 0.4 1 2 3 4 5 6
(a) displacement (mm) (b) thickness (cm)
Fig. 6. Effect of thickness on qt (a) stress ∼ displacement curves and (b) tensile
strength
16
14
12
10
qt (kPa)
8
6
4
2
0
0 1 2 3 4 5
number of layers
Fig. 7. Effect of number of compaction layers on qt for clay ∼ sand mixture (3:1)
50 kPa 100 kPa 200 kPa

50
40
30
qu (kPa)
20
10
0
0 1 2 3 4 5 6 7 8 9 10 11
number of layers
Fig. 8. Unconfined compression test results for clay ∼ sand mixture (3:1)
with the increase in number of layers of compaction, more uniformly dense

specimens were obtained which makes the bonding between the soil particles
more strong and hence, strength is increased.
Figure 9 shows the tensile strength test results of different soils conducted
at different pulling rates which varied from 0.01 to 1.75 mm/min. In Fig. 9(a)
clay-sand mixtures (1:3 and 3:1) prepared under 200 kPa were shown where
as in Fig. 9(b) same soil specimens prepared under constant dry unit were
shown. Comparing the tensile strength of each specimen with respect to tensile
pulling rate, variation in the strength with the change in the pulling rate
could be observed for all types of soil specimens. It was observed that the
tensile strength measured in the range of 0.1 to 0.34 mm/min pulling rate
gave the minimum value. In contrary to this, tensile strength measured at the
higher and lower tensile pulling rates than 0.1 mm/min showed larger values.
clay~sand (1:3) clay~sand (3:1) clay~sand (1:3) clay~sand (3:1)

14 6
12
qt (kPa)
qt (kPa)
10 5
8
6 4
4
2 3
0.1 1 0.001 0.01 0.1 1
(a) tensile pulling rate (mm/min) (b) tensile pulling rate (mm/min)
Fig. 9. Effect of tensile pulling rate on tensile strength under controlled: (a) pres-
sure; (b) dry density
This difference in the tensile strength might have occurred while pulling the
specimens where soil particles are attached to each other either by suction or
internal bonding or cohesion. Tamrakar et al. (2005a,b) had discussed about
the relationship between the tensile strength and suction to some extent.
Effect of suction at higher and lower tensile pulling rates are yet to be studied.
Increment in the tensile strength at higher tensile pulling rate is obvious as
in any other shear strength. One reason for the increment at very low pulling
rates might be due the dryness on the surface of the specimen as it takes very
long time. Other reasons are yet to study.
Conclusions
From the tests conducted for saturated and statically compacted unsaturated
soils, following points could be concluded:
1. Stress-displacement curves obtained for both saturated and compacted
unsaturated specimens show the possibility of measuring tensile strength
with the apparatus used here.
2. Ratio of unconfined compression strength and tensile strength (qu /qt ) for
NSF-clay was found to be 2 ∼ 3. For unsaturated soils this ratio varied
from 4 to 16 depending upon the type and water content of the soil.
3. From the test results of clay-sand mixture (3:1), it was found that 5 cm
thick specimen gave the minimum tensile strength in comparison to other
specimens having thickness smaller than 5 cm. Therefore, it is recom-
mended to pull the specimen from its mid-height.
4. Effect of number of layers of compaction (one-layer, two-layer and four-
layer) on tensile strength was also studied. It was found that with the
increase in the number of compaction layers, there was increase in tensile
strength. Similar result was obtained in case of unconfined compression
test also. Increment in the strength might have occurred due to increase
the uniformity of density distribution throughout the specimen.
5. Effect of tensile pulling rate for variety of unsaturated soil specimens was
carried out. It was found that with the increase in the tensile pulling rate
beyond 0.34 m/min, tensile strength also increases. But below this pulling
rate also, some increments were observed.
Acknowledgements
This research is partially carried out under the Health and Labor Sciences
Research Grants of Ministry of Health, Labor and Welfare, Japan.
References
Nahlawi H, Chakrabarti S, Kodikara J (2004) A direct tensile strength testing
method for unsaturated geomaterials, Geotech Test J 27(4):356–361
Ono N, Mochizuki A, Kurosaki H, Ueno K (2003) Trial tests with compressive
and tensile strength measuring apparatus. In: 58th annual meeting of Japanese
Society of Civil Engineers:337–338 (in Japanese)
Suzuki T, Umei T, Sunaga F (1998) A research on the tensile strength of cement
treated soils. In: 53rd annual meeting of Japanese Society of Civil Engineers:600–
601 (in Japanese)
Tamrakar SB, Toyosawa Y, Mitachi T, Itoh K (2005) Tensile strength of compacted
and saturated soils using newly developed tensile strength measuring apparatus,
Soils and Foundations 45(6):103–111
Tamrakar SB, Mitachi T, Toyosawa Y, Itoh K (2005) Development of a New Soil
Tensile Strength Test Apparatus. In: Geo-Frontiers 2005, Site Characterization
and Modeling (GSP 138), ASCE
Yao S, Masui T, Ito A (2002) The relationship between tensile strength and the state
of water in Kaolin clay. In: 47th symposium on Geotechnical symposium:127–132
(in Japanese)
The Tensile Strength of Compacted Clays
as Affected by Suction and Soil Structure
Rainer M. Zeh1 and Karl Josef Witt2

1
Geotechnisches Institut AG, Hochstrasse 48, CH-4002 Basel, Switzerland
rainer.zeh@geo-online.com
2
Bauhaus-University Weimar, Geotechnical Engineering, Department of Civil
Engineering, Coudraystr. 11c, D-99423 Weimar, Germany
kj.witt@bauing.uni-weimar.de
Summary. This paper describes the influence of soil structure and soil suction on
the tensile strength of a compacted clay. Laboratory tensile strength tests were car-
ried out on a medium plastic clay. The soil samples were prepared by compacting a
soil-water mixture of predetermined water content. The soil samples were either wet-
ted or dried to different water contents corresponding to several suction levels. The
test results showed an increase in the tensile strength with higher suction values and
that samples compacted at different compaction states reach different magnitudes
of tensile strength values.
Key words: direct tensile strength tests, different compaction states, SWCC, ten-
sile strength, E-modulus, influence of soil structure and suction
Introduction
The tensile strength of soils is usually not taken into account when solving
typical geotechnical problems. It plays an important role in connection with
the examination of compacted clay cracking (Morris et al. 1992) in landfill
liners or in cores of dam embankments, for example.
The tensile strength has relatively often been analysed in literature (e.g.
Tang and Graham 2000, Snyder and Miller 1985, Ajaz and Parry 1975, Satya-
narayana and Rao 1972, Farrell et al. 1967), but the initial states, such as
density or soil structure, were often unequal. In addition, the testing methods
also varied. The tensile strength of soils can be measured by direct or indirect
testing devices, such as bending tests, punch tests, centrifuge or triaxial cell
tests, etc.
In this paper, a method to measure the direct tensile strength of compacted
clays is described. Furthermore, the test results correspond with a number of
investigations (e.g. Nagaraj and Miura 2001, Vanapalli et al. 1999) which
220 Rainer M. Zeh and Karl Josef Witt
have shown that soil-water interaction and soil structure generally have a
great influence on soil properties.
Soil Structure
It is known that the engineering properties of fine grained (clayey) soils are
closely related to the soil-water interaction (e.g. Nagaraj and Miura 2001,
Mitchell 1993). A number of investigations indicated that the water (content)
absorbed by the clayey soils (clay aggregates and particles) correlates to other
soil properties like shear strength or compressibility.
Figure 1a shows a sketch with a typical structure of fine grained soils –
sand, silt, clay aggregates (clusters) and pores with different sizes. The clay
aggregates (e.g. Fig. 1b) consist of clay particles which are formed by face to
face (parallel), edge to edge (normal) or mixed orientations – size between 0.01
to 1 μm. The orientation depends on the clay type, water content, soil prepa-
ration, etc. Clays compacted wet of Proctor optimum have a dispersed struc-
ture (parallel), with small inter-aggregate pores, compacted dry of Proctor
a) c)
wet of Proctor optimum
Proctor optimum
b)
dry of Proctor optimum
Fig. 1. a) fine grained soil structure – 1: intra-aggregate pores, 2 (3): inter-aggregate

(large enclosed) pores, b) possible micro fabric of clays (modified from Nagaraj and
Miura 2001), c) orientation of clay particles
The Tensile Strength of Compacted Clays 221
optimum a meta-stable flocculated structure (normal), with larger and more

inter-aggregate pores, and compacted at Proctor optimum a mixture of both
structures (Fig. 1c). Therefore, the number and the size of pores as well as
aggregates are strongly influenced by the conditions during compaction.
Depending on the definition of pores (Cuisinier and Laloui 2004, Nagaraj
and Miura 2001, Diamond 1971), intra-aggregate pores (1 in Fig. 1a) have a
maximum diameter of 0.002 to 0.01 μm. The size of the inter-aggregate pores
is 0.01 to 0.1 for the small pores, large enclosed pores are up to 10 μm (2 and
3 in Fig. 1a).
The water trapped in intra-aggregate pores is mainly influenced by particle
surface forces. The inter-aggregate pore water is dominated by capillary forces.
In relation to shear or tensile strength of soils, the number and size of the
pores conduct difference magnitudes of strengths. In general, soils with only
small pores exhibit higher strengths than soils with large pores or soils with
different pores. The particle orientation also governs the obtainable strength.
Parallel soil fabrics (dispersed structures), as in the case of clays compacted
wet of Proctor optimum, have preponderant small pores and exhibit higher
shear or tensile strength values by increasing suction values.
Test Procedure and Soil Characteristics
Tensile Strength Tests
The samples were prepared in a Proctor mould with constant conditions for
each series (3 layers, 25 blows, standard hammer, Proctor optimum or 97%
wet/dry of Proctor optimum) which created a cylinder of 150×120 mm. There-
fore, the initial soil structure per series remained the same. The cylindrical
soil samples were cut in 3 slices, which had been individually prepared to hol-
low cylinders of 90 × 24 mm (inner diameter 8 mm) by carefully trimming the
slices and drilling a hole (Fig. 2). The hollow-cylindrical samples were stored
for about 48 h to ensure homogeneous conditions. The samples were slowly and
regularly air-dried or wetted (spray bottle) – between they were always stored
in waterproof bags – until the designated water content value (or correspond-
ing suction) was obtained (about 14 to 21 days). Finally, the samples were
weighed and coated in wax to detect the volume by dip-weighing. The centre
of the inner hole was later filled with a filter textile and a modified dowel was
glued in both sample ends with epoxy resin. Two small hooks drilled in the
dowels were used to apply the tensile forces in the sample.
The tensile strength tests were run in a modified triaxial apparatus by
measuring the tensile force and the strain. The samples were normally torn
apart with v = 0.001 mm/s until rupture occurred. Besides, the water content
at the rupture zone was measured to determine the suction level by using the
SWCC of the soil (Zeh 2006).
Soil Characteristics
In this paper, test results of a medium plasticity clay called Plessa (Pl) are
presented. The soil properties are shown in Table 1.
Table 1. Soil properties of clay Plessa (Pl)
Parameters Values
Clay/silt/sand content 46.4/52.8/0.8%
Liquid/plastic limit wl /wp 49.7%/21.3%
Plasticity index Ip 28.4%
Specific gravity ρs 2.676
Proctor density/water content at optimum 1.696 g/cm3 /17.7%
at wet (97%) of optimum 1.645 g/cm3 /21.5%
at dry (97%) of optimum 1.645 g/cm3 /14.3%

Shear strength ϕ 25.0◦

Cohesion c 16.8 kN/m2
To convert the water contents into suction values, shrinkage curves and
soil-water characteristic curves (SWCC) were necessary as described before
(Zeh and Witt 2005a,b). Figure 3 represents the SWCC of clay Plessa, Proctor
compacted (called p18–100o) with a fitting curve by Fredlund and Xing (1994).
The curves of samples compacted wet or dry of Proctor optimum (called p21–
97f and p14–97tr) also show strong similarities. The SWCCs were determined
by the well-known axis-translation technique and chilled-mirror hygrometer
measurements (Leong et al. 2003).
dried
wetted
Proctor compacted cut in slices trimmed in cylindrical sample torn apart
Fig. 2. Schema sample preparation

0,40
344
0,35
volumetric water content [m³/m³]
0,30
0,25
0,20
0,15 axis-trans.-techn.
chilled-mirror-hygrom.
0,10
Fitting Fredlund & Xing
137600
0,05 air-entry value
residual value
0,00
0,1 1 10 100 1000 10000 100000 1000000
suction [kN/m²]
Fig. 3. Soil-water characteristic curve of clay Plessa, compacted at Proctor opti-

mum, fitted with Fredlund and Xing (1994)’s equation
1000
p21-97f
900
p18-100o
800 p14-97tr
compacted wet of
Proctor optimum
700
tensile strength [kN/m²]
600
compacted at Proctor
500 optimum
400
300
200
100 compacted dry of

Proctor optimum
0
100 1000 10000 100000
suction [kN/m²]
Fig. 4. Tensile strength of clay Plessa in dependence on the soil structure and the
suction
Test Results
Figure 4 presents the results of the suction versus tensile strength relationships
of all samples tested in the study. As can be seen in Fig. 3 until suction
values of about 600 kN/m2 all samples show a similar development of the
tensile strength. Beyond this value, the samples compacted dry of Proctor
optimum have a much smaller increase in the tensile strength than the samples
compacted wet of optimum or at Proctor optimum. The samples compacted
wet of optimum and at Proctor optimum have also a similar tensile strength
increase up to a suction of about 1150 kN/m2 ; beyond this value, the “wet”
samples exhibit higher tensile strength values.
A comparison of the obtained tensile strength at a suction value of about
11,000 kN/m2 , as an example, represents very well their difference as well
as the big influence of the soil structure (and pore sizes) as described above.
The dry densities (determined by the tensile strength tests) as shown in Fig. 5
verify the smaller difference in tensile strength of the samples compacted wet of
optimum or at Proctor optimum and the greater one of the samples compacted
dry of Proctor optimum, respectively.
However, the influence of the soil structure is hardly visible in Fig. 6 which
represents the E-modulus in dependence on the suction. The values scatter
very distinctively. The E-modulus is calculated by strain and stresses during
the tensile strength tests – a secant modulus of 80% and 15% stresses or rather
strains at failure.
2,05
p21-97f
2,00 p18-100o
p14-97tr
1,95 Proctor curve
Sr=0,70
1,90
dry densitiy [g/cm³]
Sr=0,85
Sr=1,00
1,85
1,80
1,75
1,70
wet (97%) of Proctor optimum (21,5%, 1,65kg/cm³)

1,65 compacted at Proctor optimum (17,7%, 1,70kg/cm³)
dry (97%) of Proctor optimum (14,3%, 1,65kg/cm³)
1,60
0 2 4 6 8 10 12 14 16 18 20 22 24 26
gravimetric water content [%]
Fig. 5. Dry density (determined by the tensile strength tests) and Proctor curve of
clay Plessa
250
p21-97f compacted wet of
Proctor optimum
p18-100o
200 p14-97tr
E-Modulus [MN/m²]
150
compacted dry of compacted at Proctor

Proctor optimum optimum
100
50
0
100 1000 10000 100000
suction [kN/m²]
Fig. 6. E-modulus (clay Plessa) in dependence on the soil structure and the suction
Conclusions
The direct tensile testing method presented is very appropriate for examining
the tensile strength of compacted clays (or other fine grained soils) exposed to
different suctions (water contents). The test procedure is quite simple, whereas
the sample preparation needs some practice.
The results of the tensile strength tests clearly showed the effect of soil-
water interaction on the tensile strength and the E-modulus. Besides, the soil
structure has surprisingly an influence only on the tensile strength. These
results and tendencies described have to be verified on further soils and soil
structures, respectively.
References
Ajaz A, Parry RHG (1975) Stress-strain behaviour of two compacted clays in tension
and compression, Géotechnique 25(3):495–512
Cuisinier O, Laloui L (2004) Fabric Evolution of an Unsaturated Compacted Soil
During Hydromechanical Loading. In: Schanz (ed) Unsaturated Soils – Experi-
mental Studies, Vol I. Springer Proceedings in Physics 93, Springer, Berlin:147–
158
Diamond S (1971) Microstructure and pore structure of impact compacted clays,
Clays and Clay minerals 19:239–249
Farrell DA, Greacen EL, Larson WE (1967) The Effect of Water Content on Axial
Strain in a Loam Soil Under Tension and Compression, Soil Sci Soc Amer Proc
31:442–450
Fredlund DG, Xing A (1994) Equations for the soil-water characteristic curve, Can
Geotech J 31(3):521–532
Leong EC, Tripathy S, Rahardjo H (2003) Total suction measurement of unsatu-
rated soils with a device using chilled-mirror dew-point technique, Géotechnique
53(2):173–182
Mitchell JK (1993) Fundamental of soil behaviour. John Wiley & Sons, NY
Morris PH, Graham J, Williams DV (1992) Cracking of drying soils, Can Geotech
J, 29:263–277
Nagaraj TS, Miura N (2001) Soft clay behaviour – analysis and assessment. Balkema,
Rotterdam
Satyanarayana B, Rao KS (1972) Measurement of Tensile Strength of Compacted
Soil, Geotech Eng 3:61–66
Snyder VA, Miller RD (1985) A pneumatic fracture method measuring the tensile
strength of unsaturated soils, Soil Sci Soc Am J 49:1369–1374
Tang GX, Graham J (2000) A method for testing tensile strength in unsaturated
soils, Geotech Testing J 23(3):377–382
Vanapalli SK, Fredlund DG, Pufahl DE (1999) The influence of soil structure and
stress history on the soil-water characteristics of a compacted till, Géotechnique
49(2):143–159
Zeh RM, Witt KJ (2005a) A direct testing method for the tensile strength of com-
pacted clays. In: Int Symp Advanced Experimental Unsaturated Soil Mechanics,
Trento, Italy, June 2005
Zeh RM, Witt KJ (2005b) Suction-controlled Tensile Strength of Compacted Clays.
In: 16th ICSMGE Osaka, Japan, September 2005
Zeh RM (2006) Die Zugfestigkeit bindiger Böden als Kriterium der Rissgefährdung
mineralischer Oberflächenabdichtungen. Dissertation (PhD Thesis), Fakultät für
Bauingenieurwesen, Bauhaus-Universität Weimar (submitted)
Part IV
Temperature Effects
Modified Isochoric Cell for Temperature
Controlled Swelling Pressure Tests
Yulian Firmana Arifin and Tom Schanz

yulian-firmana.arifin@uni-weimar.de, tom.schanz@uni-weimar.de
Summary. This paper focuses on the effect of moderately high temperature on the
swelling pressure of compacted bentonite and bentonite-sand mixtures which are
among the material used as buffer material in high level waste repository. Modified
isochoric cell was used to measure the swelling pressure of the material at 80◦ C. Two
different methods were utilized in order to implement two transport mechanisms of
water molecules in compacted mixtures. The equipment set-up, verification, and
experimental programs adopted in this study are described. The results obtained
are presented and discussed.
Key words: swelling pressure, isochoric cell, temperature effects, bentonite–sand

mixtures
Introduction
Emplaced high level wastes will emit quite significant amount of heat which is
thought to change coupled hydro-mechanical behaviour of the buffer element
used in the waste disposal facility. Swelling pressure, which is a representation
of both hydraulic and mechanical behaviour of the material and indicates the
performance of the sealing material, is one of important behaviour that should
be investigated in case when increase in temperature in the waste repository
is of concern.
Pusch et al. (1990) stated that the increase in temperature of a bentonite
decreased the hydration force due to a reduction in the number of hydrates
in the smectite surface within bentonite and increased the osmotic pressure
in the molecular system. These led to reduction or increase in the swelling
pressure of compacted bentonite depending on the dominant factor happen in
the type of bentonite used. On the other hand, the increase in temperature
may cause the smectite mineral to become unstable and transform to more
stable silicate phases. The transformation processes are known as illitisation,
cementation, and chloritisation processes (Wersin et al., in press). These may
lead to reduction in swelling pressure of the compacted bentonite.
230 Yulian Firmana Arifin and Tom Schanz
Earlier investigations show that the increase in temperature induced in-

significant increase in the swelling pressure of bentonite (e.g., Pusch et al.
1990 for Na-bentonite, Cho et al. 2000 for Ca-bentonite). On the other hand,
the reverse is also true (e.g., Pusch et al. 1990 for Ca-bentonite, Lingnau et
al. 1996 for Na-bentonite, and Villar and Lloret 2004 for FEBEX bentonite
which is mainly Ca-Mg-bentonite). However, the temperature effects on the
swelling pressure of bentonite remain important in the assessment of long-term
performance of the whole repository construction. The investigation into the
temperature effects on the behaviour of bentonite provides insight into un-
derstanding the hydro-mechanical processes taking place in the bentonite in
relation to its use as barrier material. The results of such a research will
provide fundamental data concerning the parameters to be used in the mod-
elling since the long-term behaviour of the material are to be predicted using
available sophisticated numerical tools.
In this study, the effects of moderately high temperature on the swelling
pressure of a calcium-type bentonite were investigated. Modified isochoric cell
was used to measure swelling pressure of the material by means of constant
volume test. The elevated temperature up to 80◦ C was used in this study since
up to now most HLW repository concepts are designed based on a design cri-
terion that a maximum temperature of 90◦ C is considered for bentonite buffer
(SKB 1999). Two types of test; namely, one step swelling pressure test and
multi-step swelling pressure test, were performed to study the two mechanisms
of water molecules transports (i.e., in fluid phase and in vapour phase). The
equipment set-up, verification, and techniques adopted are described. The re-
sults obtained as well as the comparison with the swelling pressure measured
at room temperature are presented and discussed.
Equipment and Verification
The constant volume test performed in this study was conducted using iso-
choric cell developed at the Technical University of Barcelona (UPC) Spain.
The cell was used to measure swelling pressure of compacted bentonite and
bentonite–sand mixture from low to high density at room temperature by
Villar et al. (2001) and Agus and Schanz (2005a). The cell was modified by
installing a flexible wire heater attached to the outer wall of the cell. The
heater was controlled using a proportional integral differential (PID) temper-
ature with programmable ramp and soak temperature features. Two small
thermocouples were installed in a hole drilled through the top cap of the
cell for measuring the temperature of the specimen during experiment. The
swelling pressure was measured using a load cell connected to a handled meter.
Figure 1 shows the schematic diagram of the one-step swelling pressure
test. In this test, the cell was connected to a water column filled with distilled
water. A flexible wire temperature-controlled heater was also installed in the
water column to monitor the temperature of the water used in the permeation.
Modified Isochoric Cell for Temperature Controlled Swelling Pressure 231
Fig. 1. Schematic drawing of the one-step swelling pressure test
Fig. 2. Schematic drawing of the multi-step swelling pressure test
Figure 2 shows the schematic drawing of the multi-step swelling pressure

test. Vapour equilibrium technique (VET) was used in this test whereby the
total suction of the specimen was reduced by increasing the relative humidity
of the vapour in the specimen’s pore space. Similar method was also adopted
by Agus (2005) using the same cell in the tests at room temperature.
The multi-step swelling pressure test in this study was performed in a
climate chamber that could control temperature to ±0.1◦ C. The cell was con-
nected to two Erlenmeyer flasks filled with salt solution. A relative humidity
and temperature sensor were installed to one of the Erlenmeyer flasks to mon-
itor the relative humidity of the water vapour above the salt solution in the
flask that was used to induce relative humidity to the specimen. The sensor
(together with a temperature sensor) was connected to a digital data dis-
play connectable to a personal computer. The sensor used has an excellent
performance not only at room temperature but also at moderately high tem-
perature (Agus and Schanz 2005b, Arifin et al. 2006a).
Prior to use, the deformation of the cell ring and the load cell due to the
non-isothermal test conditions were studied and verified so that the effects can
be taken into account. Increasing temperature resulted in decreasing zero off-
set of the load cell calibration curve. However, the intercept of the calibration
curves, which was also used for calculating the swelling pressure, does not show
significant change. The coefficient of thermal expansion of the cell used in this
study was found about 1.58 × 10−5 /◦ C, which is closed to the coefficient of
linear thermal expansion of stainless steel (AISI, αr = 1.60 × 10−5 /◦ C). Small
deviation from the value is expected since the non-steel components of the cell
also affect the thermal behaviour of the cell as a whole. The change in void
ratio, Δe, due to the change in radial ring deformation can be approximated
using the following equation (Romero 1999)

(2αr − αs )ΔT
Δe = (1 + e0 ) (1)
1 + αs ΔT
where e0 is initial void ratio, αr is the coefficient of thermal expansion of the

ring, αs is volumetric thermal dilatation of the clay particle (αs ≈ 2.9 × 10−5 /
◦
C, Horseman and McEwen 1996), and ΔT is temperature change (in this
study ΔT = 60◦ C). By inserting all known parameters, equation (1) can be
simplified to be the following equation:
Δe = 0.0002(1 + e0 ) . (2)
The change in the dry density of the specimen tested due to the change
in void ratio can be computed using equation (3), which is derived from sub-
stituting equation (2) to the first derivative of the relationship between dry
density and void ratio.
Δρd = −0.0002ρd (3)
where Δρd is dry density change in Mg/m3 , and ρd is initial dry density of
specimen. Although the swelling pressure result is an exponential function of
dry density of specimen (Villar and Lloret 2004, Agus and Schanz 2005a), it
is clear that the swelling pressure change due to thermal deformation of the
ring and thermal dilatation of the clay particles is not significant provided
that corrections are applied to the specimen’s dry density.
Material Used
The materials used in this study were calcium-type bentonite, Calcigel, and
mixtures of Calcigel and quartz sand. The bentonite used has specific gravity
of 2.8, the liquid limit of 180%, and the plastic limit of 33%. The monmoril-
lonite content of the bentonite used is 50–60% with a specific surface area of
651 m2 /g (Agus 2005). The quartz sand used has specific gravity of 2.65 with
a specific surface area of 0.25 m2 /g. Two types of specimens were prepared
with different bentonite contents (on a dry mass basis); namely, 50B/50S
bentonite–sand mixture with a bentonite content of 50% and 100B for pure
bentonite. The sand was sieved through 2-mm sieve openings before mixing.
The specimen conditions used in this study are summarized in Table 1.
Table 1. Specimen conditions of one step swelling pressure test
w ρd w ρd
Specimen Specimen
(%) (Mg/m3 ) (%) (Mg/m3 )
100B–1 9 1.43 50B/50S–2 17 1.77
100B–2 9 1.26 50B/50S–3 20 1.66
100B–3 9 1.05 50B/50S–4 25 1.49
Heavily compacted 50B/50S prepared by Gesellschaft für Anlagen und

Reaktorsicherheit (GRS) were used in the one-step swelling pressure test (i.e.,
denoted as 50B/50S–1 in the subsequent sections) and multi-step swelling
pressure test. The specimens had an average initial dry density of 2.0 Mg/m3 ,
initial water content of 9%, and initial total suction of 22000 kPa (Agus and
Schanz 2006a).
Swelling Pressure Measurement

One-Step Swelling Pressure Test
The test commenced by placing the specimen in the cell. The temperature of
specimen was elevated by heating the cell using the temperature-controlled
heater. The temperature increment should be slow enough to avoid tempera-
ture difference between the cell and the specimen. Rapid temperature incre-
ment results in decreasing the relative humidity of the air space which is ex-
isted in the cell (e.g. porous disks) due to increase its saturation water vapour
(Arifin et al. 2006a, Agus and Schanz 2006b). Some amount of water from the
specimen will evaporate to balance this condition. Trial tests had been done
to determine the appropriate rate of temperature increase. Figure 3 shows
the difference temperature between the cell and the specimen at different rate
of temperature increase. For rapid temperature increase (i.e., 20◦ C/min) as
shown in dashed line, the specimen temperature reached 80◦ C several minutes
after the cell temperature had reached 80◦ C. From the trial tests performed,
it is found that the appropriate rate of temperature increase is 0.016◦ C/min
(or 1◦ C/hour) where the temperature of the cell and the specimen reached
80◦ C at the same time.
After the temperature of both the cell and the specimen had reached 80◦ C,
warmed distilled-deaired water was circulated through the bottom and the top
boundaries of specimen. Warmed water was used in this study with a reason
90
T cell delay
80 T specimen
Temperature (°C)
70 20°C/min
60
50
40 supplied with
water 20°C
30 1°C/hour
20
0 1 10 100 1000 10000
Elapsed time (minutes)
Fig. 3. Investigation of the rate of temperature and supplied water to the specimen
temperature
that the use fresh water decreases the temperature of the specimen as also
shown in Fig. 3. The decrease in the water temperature due to heat release in
the water circulation line was anticipated by increasing the water temperature
in the water supply column up to 86◦ C. During the test, the swelling pressure
of the specimen was measured using the handled voltmeter. The equilibrium
was reached after the swelling pressure readings stabilised for a significant
period of time.
Multi-Step Swelling Pressure Test
The heavily compacted 50B/50S specimens prepared by Gesellschaft für Anla-

gen und Reaktorsicherheit (GRS) were tested following the multi-step swelling
pressure test method. After the arrangement shown in Fig. 2 had been set up,
the specimen was placed in the cell and the temperature of the chamber was
gradually increased at a rate of 0.01◦ C/min which was much lower than the
rate used in the one-step swelling pressure test for anticipating temperature
gradient between the specimen, the cell and the salt solution. The molal and
saturated sodium chloride solutions were used to induce relative humidity
to the specimen. The use of molal and saturated sodium chloride solutions
for controlling suction (or relative humidity) at room temperature and mod-
erately high temperature (i.e., 80◦ C) was verified by Arifin et al. (2006b).
It was found that the technique is applicable provided that the temperature
gradient in the system is minimised. The swelling pressure development, speci-
men temperature, and relative humidity and temperature of salt solution were
periodically recorded. The results obtained were compared with those of the
earlier investigation performed at room temperature and reported by Agus
(2005).
Result and Discussion

Figures 4(a) and 4(b) show swelling pressure development versus elapsed time
plotted in semi-logarithmic scale performed with one-step swelling pressure
test for 100B and 50B/50S specimens, respectively. The swelling pressure
recorded after distilled water supply show rapid increase rapidly at earlier
stage of the test for both specimens. Generally, after the swelling pressure
maximum was reached (i.e., after about 100 hours duration), the swelling pres-
sure of the 100B specimen slightly decreased possibly due to meta-stable of
structure of the compacted pure bentonite at low water content, whereas the
swelling pressure of the 50B/50S specimen remained constant. According to
the figures, the swelling pressure development of the specimens tested at 80◦ C
was more rapid than that of the specimens tested at 20◦ C. Agus and Schanz
(2005a) found that the rate of swelling pressure development is a function of
the initial total suction of the compacted specimen tested. However, the in-
crease in temperature causes decrease in total suction in compacted bentonite
(Villar and Lloret 2004, Romero et al. 2000) and in compacted bentonite–
sand mixtures (Arifin et al. 2006a). Therefore, the faster of swelling pressure
development rate of specimen tested at 80◦ C may most probably be due to
preconsolidation effects induced by heating.
Figure 4 shows also that the magnitude of swelling pressure of the speci-
mens tested at 80◦ C was smaller than that of the specimens tested at 20◦ C.
The result agrees well with the results reported by Pusch et al. (1990), Villar
and Lloret (2004), Lingnau et al. (1996). According to Pusch et al. (1990),
the effects of temperature on the swelling pressure depend on the net of two
opposite effects (i.e., the increase in osmotic pressure and the decrease in
hydration force). The increase in the osmotic pressure leads to increase in
swelling pressure, whereas the decrease in the hydration force causes decrease
in swelling pressure. Since the number of flakes of Ca-bentonite is considerably
1800 3500
(a) (b)
1600 100B-1(20°C) 50B/50S-1(20°C)
3000
100B-2(20°C) 50B/50S-2(20°C)
1400
Swelling pressure (kPa) .
100B-3(20°C) 50B/50S-3(20°C)
2500
100B-1(80°C) 50B/50S-4(20°C)
1200
100B-2(80°C) 50B/50S-1(80°C)
1000 100B-3(80°C) 2000 50B/50S-2(80°C)
50B/50S-3(80°C)
800 1500 50B/50S-4(80°C)
600
1000
400
500
200
0 0
0.001 0.01 0.1 1 10 100 1000 0.001 0.01 0.1 1 10 100 1000
Elapsed time (hours) Elapsed time (hours)
Fig. 4. Development of swelling pressure as a function of time for: (a) 100B speci-
men, and (b) 50B/50S specimen (one step tests)
high (higher than that of Na-bentonite), the increase in temperature affects

the hydration force more significantly than the osmotic pressure. The effects
of the reduction in the hydration force are more pronounced than the in-
crease in the osmotic pressure. Therefore, the increase in temperature results
in reducing swelling pressure.
Figure 5 shows the development of swelling pressure at 80◦ C followed by
cooling to the room temperature (i.e., ±20◦ C) for the test at 80◦ C. According
to the figure, the decrease in swelling pressure due to increase in temperature is
reversible. The swelling pressure of the specimen tested increased and reached
an equilibrium value which is very close to the swelling pressure performed
at 20◦ C. Therefore, temperature-induced transformation of smectite to non-
expandable illite layers is not evident in the range of temperature and time
duration considered in this study. However, prolonged heating may induce
such transformation which requires further investigation.
Figure 6(a) shows the swelling pressure of the 100B and 50B/50S spec-
imens plotted against the mixture dry density. As expected and supported
by other’s findings such as Komine and Ogata (2003) and Agus and Schanz
(2005a), the swelling pressure of the specimens shows exponential function
against mixture dry density. As depicted in the figure, the curve of swelling
pressures obtained for the specimens tested at 80◦ C is located below the curve
of swelling pressure obtained for the specimens tested at 20◦ C. The reduction
of swelling pressure due to increase in temperature of pure bentonite spec-
imens appears to be higher than that of the 50B/50S specimens. Agus and
Schanz (2005a) found that the swelling pressure of compacted bentonite–sand
mixtures with different bentonite contents is a function of bentonite dry den-
sity. The bentonite dry density (ρbentonite
d ) is computed by assuming that the
soil void in the compacted bentonite–sand mixtures only belongs to the ben-
tonite (Komine and Ogata 2003):
3500
50B/50S-1 (20°C)
50B/50S-1 (80°C)
3000
Swelling pressure (kPa).
2500
Testing period Cooling period for the

2000 test at 80°C
1500
1000
500
0
0.001 0.01 0.1 1 10 100 1000
0.001
0 0.01 0.1 1 10 100 1000
Elapsed time (hours)
Fig. 5. Development of swelling pressure after tested at 80◦ C

3200 3500
(a) (b)
2800 100B (20°C) 100B (20°C)

3000
100B (80°C) 100B (80°C)

2400 50B/50S (20°C) 50B/50S (20°C)
2500
50B/50S (80°C) 50B/50S (80°C)
2000
2000
1600
20°C 1500
20°C
1200 20°C
1000
800
80°C 80°C
80°C
400 500
0 0
1 1.2 1.4 1.6 1.8 2 2.2 1 1.1 1.2 1.3 1.4 1.5 1.6
3
Dry density (Mg/m3) Bentonite dry density (Mg/m )
Fig. 6. (a) Swelling pressure as a function of dry density, and (b) swelling pressure
as a function of bentonite dry density
rbs ρd
ρbentonite = (4)
d
(1 + rbs ) − (ρd /Gsand
s )
where rbs is the bentonite–sand mass ratio, ρd is mixture dry density in

Mg/m3 , and Gsand
s is the specific gravity of sand.
Figure 6(b) shows the relationship between swelling pressure and bentonite
dry density for the compacted bentonite and compacted bentonite–sand mix-
tures. Allowing slight scatters in the data, the curve of swelling pressure for the
50B/50S specimens tested at 80◦ C coincides with that for the pure bentonite
specimens. It indicates that at 80◦ C the swelling pressure of bentonite–sand
mixtures also shows similar trend as the swelling pressure curve obtained at
20◦ C. The increase in temperature only affects the magnitude of the swelling
pressure which represents the effects of temperature on the bentonite mineral
in the mixture. Therefore, the presence of sand in the mixture is expected to
give benefit in reducing the temperature effects or the reduction in swelling
pressure due to elevated temperature.
Figure 7 shows the swelling pressure development of highly compacted
bentonite–sand mixture performed using VET at 80◦ C together with the evo-
lution of the relative humidity of salt solution used as measured using the
relative humidity. The good performance of the testing system is indicated by
the constant relative humidity throughout the test as shown in the figure.
According to Fig. 7, the swelling pressure tested using VET shows slow
development at earlier stage of test. The swelling pressure increased rapidly
after 50 hours and reached equilibrium after 200 hours. Ignoring the magni-
tude of the swelling pressure, the results indicate a delay when compared with
the development of swelling pressure when the specimens were tested using
water in its liquid phase (i.e., in the one-step swelling pressure test). The
delay was probably caused by the retarding transport mechanism of water
25
Swelling pressure (kPa)

As-prepared-1
20 As-prepared-2
oven-dried-1
15 oven-dried-2
10
00
100
90
RH (%)
80
70
60
1 10 100 1000
Elapsed time (hours)
Fig. 7. Swelling pressure development from the multi-step swelling pressure test
molecules in the vapour form which occurs when using VET in the test. The
water molecules first migrate into the open voids in the specimen and are sub-
sequently absorbed when in contact with the exposed surfaces of clay clusters
(Pusch 2001, Agus 2005). A water potential gradient (suction) that exists
between the inter-aggregate and intra-aggregate pore-water and the mixture
permeability at micro and macro scales control the mechanism of water move-
ment from the exposed surface of the clay clusters to the intra-cluster pores
(Agus and Schanz, in print). These postulations also seem to be true at an
elevated temperature of 80◦ C as indicated by the results presented herein.
Figure 8 shows the development of swelling pressure with decreasing suc-
tion for the as-prepared 50B/50S specimens (i.e., those prepared by GRS)
and oven dried 50B/50S specimen performed in this study. The oven-dried
specimen as referred to herein is basically the GRS-prepared specimen which
was oven-dried prior to test. The data of the swelling pressure development
obtained for the same material tested at room temperature (i.e., ±20◦ C) and
reported by Agus (2005) using two different methods (i.e., axis translation
technique (ATT) and vapour equilibrium technique (VET)) are also plotted
for comparison.
According to Fig. 8, the difference between the swelling pressures per-
formed at room temperature and the swelling pressure performed at 80◦ C
appears obscure since the development of swelling pressure is insignificant
up to a suction of about 1000 kPa. Agus (2005) states that the insignificant
swelling pressure development might be due to a delayed ‘true’ equilibrium
in the specimen by considering the water movement mechanism mentioned
above. In addition, the corresponding water content of the specimens at the
4000
one step swelling
As-prepared at 80°C
pressure test at 20°C
3500 (VET), see Fig.7.
oven-dried at 80°C

3000 (VET), see Fig.7.
one step As-prepared at 20°C
2500 swelling (VET)
pressure test at As-prepared at 20°C
2000 80°C (ATT)
oven-dried at 20°C
1500 (VET)
oven-dried at 20°C
1000 (ATT)
500
0
1 10 100 1000 10000 100000 1000000
Suction (kPa)
Fig. 8. Swelling pressure versus suction following wetting obtained using the multi-
step swelling pressure test
suction range applicable in this method is close to the shrinkage limit of the
material reported by Agus and Schanz (2006a) (i.e., 9.8%). The increase in
water content with decreasing suction at the range of suction used in VET
does not result in significant change in the void ratio of the specimen. As
a result the pressure required to keep the void ratio constant (or swelling
pressure) when the specimen is wetted is also low.
Conclusion
The swelling pressure test at a moderately high temperature using modified
isochoric cell has been presented. The equipment verification and the swelling
pressure test procedures (i.e., the one-step swelling pressure test and the multi-
step swelling pressure test) performed in this study have been described. The
results have been presented and compared to the swelling pressure results
obtained at room temperature performed in this study and previous study.
The increase in temperature results in reduction in swelling pressure. The
reduction of swelling pressure performed in this study is reversible. It seems
to be mainly caused by the reduction in the hydration force of the bentonite in
the specimens tested. Thus, for the bentonite–sand mixtures, the increase in
temperature affects only the bentonite. The presence of sand in the mixture
gives benefit as expected to reduce the temperature effects. The transport
mechanism of water molecules controlled the development of swelling pressure
of the heavily compacted bentonite–sand mixture specimens not only at 20◦ C
but also at 80◦ C. The increase in swelling pressure of the as-prepared specimen
and the oven-dried specimen performed using VET method was insignificant
which is due to the limitation of suction range applicable using the VET.
The water content corresponding to the suction values used is close to the
shrinkage limit of the specimen. Thus, further investigation using ATT should
be performed to investigate the behaviour of the specimen at the lower suction
range (i.e., < 1000 kPa).
Acknowledgment
The research is funded by Bmbf project: Experimentelle und numerische
Untersuchungen zum Langzeit-verhalten von Abschlussbauwerken: Thermo-
Hydraulisch-Mechanisch-Chemisch gekoppelte Systeme. FKZ: 02C1104.
References
Agus SS (2005) An experimental study on hydro-mechanical characteristics of com-
pacted bentonite–sand mixtures. Ph.D. Thesis, Bauhaus-University Weimar
Agus SS, Schanz T (2005a) Swelling pressure and total suction of compacted
bentonite–sand mixtures. In: Proc Int Conf on Problematic Soils, Cyprus,
pp 61–70
Agus SS, Schanz T (2005b) Comparison of four methods for measuring total suction.
Vadose Zone Journal. Special section: Soil Water Sensing. Soil Science Society
of America 4(4):1087–1095
Agus SS, Schanz T (2006a) Drying, wetting, and suction characteristic curves of a
bentonite–sand mixture. In: Proc 4th Int Conf on Unsaturated Soils, Arizona,
USA, pp 1405–1414
Agus SS, Schanz T (2006b) Discussion of paper “Free energy of water-suction-in filter
papers” by R. Bulut and W.K. Wray, Geotechnical Testing Journal, ASTM 29(6)
Agus SS, Schanz T (in print) Discussion of paper “Parameters affecting soil-water
characteristic curves of fine-grained soils” by Thakur VKS, Sreedeep S, Singh
DN, J Geotech and Geoenvir Engrg ASCE 132(11)
Arifin YF, Agus SS, Schanz T (2006a) Temperature Effects on Suction Characteristic
Curve of Bentonite-Sand Mixtures. In: Proc 4th Int Conf on Unsaturated Soils,
Arizona, USA, pp 1314–1325
Arifin YF, Agus SS, Schanz T (2006b) Drying-wetting behavior of compacted
bentonite–sand mixtures as also affected by temperature. In: Proc Int Conf on
Geo-Environmental Engineering 2006, Kyoto, Japan, pp 63–69
Cho WJ, Lee JO, Kang CH (2000) Influence of temperature elevation on the sealing
performance of the potential buffer material for a high-level radioactive waste
repository, Annals of Nuclear Energy 27:1271–1284
Horseman ST, McEwen TJ (1996) Thermal constraints on disposal of heat-emitting
waste in argillaceous rocks, Eng Geol 41:5–16
Komine H, Ogata N (2003) New equations for swelling characteristics of bentonite-
based buffer materials, Can Geotl J 40:460–475
Lingnau BE, Graham J, Yarechewski D, Tanaka N, Gray MN (1996) Effects of tem-
perature on strength and compressibility of sand-bentonite buffer, Engineering
Geology 41:103–115
Pusch R, Karlnland O, Hokmark H (1990) GMM-a general microstructural model

for qualitative and quantitative studies of smectite clays. SKB Technical Report
90–43, Stockholm, Sweden
Pusch R (2001) The microstructure of MX-80 clay with respect to its bulk physical
properties under different environmental conditions. SKB Technical Report TR–
Romero E (1999) Characterisation and thermo-hydro-mechanical behaviour of un-
saturated Boom-clay: an experimental study. Ph.D. Thesis, Universidad Politéc-
nica de Cataluña, Barcelona
Romero E, Gens A, Lloret A (2000) Temperature effects on water retention and
water permeability of an unsaturated clay. In: Rahardjo, Toll, Leong (eds) Proc
of Unsaturated Soils for Asia. Balkema, Rotterdam
SKB (1999) SR-97: Waste, repository design and site. SKB Technical Report TR–
Villar MV, Rivas P, Campos R, Lloret A, Romero E, Mariano A (2001) First report
on thermo-hydro-mechanical laboratory tests. Report 70–IMA–L–0–86
Villar MV, Lloret A (2004) Influence of temperature on the hydro-mechanical be-
haviour of a compacted bentonite, Applied Clay Science 26:337–350
Wersin P, Johnson LH, McKinley IG (in press) Performance of the bentonite barrier
at temperatures beyond 100◦ C: A critical review, Physic and Chemistry of the
Earth
Some Aspects of the Effect of the Temperature
on the Behaviour of Unsaturated Sandy Clay
Moulay Smaine Ghembaza1 , Said Taïbi2 , and Jean-Marie Fleureau3

1
Laboratory of Soil Mechanics and Structure, University Djillali Liabès, Faculty of
the Engineer, BP 89, 22000 Sidi Bel Abbès, Algeria ghembaza_moulay@yahoo.fr
2
Laboratory of Mechanics, Physics and Geosciences, University of Le Havre,
BP 540, 76058 Le Havre Cedex, France Said.Taibi@univ-lehavre.fr
3
Laboratory MSS-Mat, Ecole Centrale de Paris, 92295 Châtenay-Malabry cedex,
France Jean-Marie.Fleureau@ecp.fr
Summary. The paper describes experimental results on triaxial paths at high tem-
perature concerning an unsaturated clay (wL = 38%, IP = 19 and Cc = 0.23).
Unsaturated drained triaxial compression tests on normally consolidated samples
were carried out at constant temperatures varying from 22 to 80◦ C. Desaturation of
samples were carried out by the imposition of suctions using the osmotic method.
This required the development of a new triaxial cell equipped with a collar heating
and an osmotic pedestal which allows the circulation of an osmotic solution of Poly
Ethylen Glycol (PEG) 6000, in contact with the sample via a dialysis membrane.
The principal phenomena simultaneously related on the temperature and saturation
were highlighted, in particular the hardening of material on isotropic path, a secant
modulus which decreases with temperature and increases with suction, a light reduc-
tion in the maximum stress on deviator path at high temperature and the increase
in this maximum stress when suction increases.
Key words: unsaturated triaxial path, normally consolidated, temperature, suc-

tion, THM behaviour, osmotic method
1 Introduction
The study of the thermo-hydro-mechanical behaviour of saturated and unsat-
urated soils is a very complex problem, and it is often useful to uncouple the
effects of temperature, mechanical loading and negative pore water pressure
or suction. This problem was tackled by several authors, either by compar-
ing the soil mechanical behaviour at various temperatures under isothermal
conditions, or by varying the temperature under constant stress. The main
conclusions of these studies are as follows:
• On normally consolidated path, an increase in temperature results in a den-
sification of the soil. On oedometric and isotropic normally consolidated
244 M. S. Ghembaza et al.
paths, the temperature causes a reduction in the void ratio (Fleureau and
Kheirbek-Saoud 1992, Rahbaoui 1996) whereas, on an overconsolidated
path, the temperature causes a dilation Belanteur et al. (1997), Sultan et
al. (2002).
• In the case of overconsolidated soils, one observes that an increase in the
temperature erases the overconsolidation of the material (Rahbaoui 1996).
• An increase in temperature reduces the mechanical strength of soils (Sultan
et al. 2002, De Bruyn 1999, Belanteur et al. 1997). Other authors show
an increase in the mechanical strength with temperature (Tanaka et al.
1997). In this case, a detailed attention must be given to the type of
studied material, its initial state and its mode of preparation to clarify
these contradictions.
• A few authors show that temperature reduces the value of the slope M in
the plan [p ; q] plane (De Bruyn 1999). However, others show little or no
influence of the temperature on the value of M , and sometimes a slight
increase of this parameter (Tanaka et al. 1997).
Concerning the THM behaviour of the unsaturated soils, very few results
exist in the literature. This is due to the complexity of the experiments, their
duration and the number of parameters to be controlled (El Youssoufi et
al. 2002). Recent works show on the one hand, a reduction in suction when
the temperature increases for a given water content (Romero 1999) and, on
the other hand, an increase in strength with temperature for a given suction
(Jamin 2003).
This paper presents experimental results allowing the study of the effect
of the temperature on the behaviour of unsaturated clay on triaxial paths.
2 Materials and Experimental Devices
The material used is a mixture made up of kaolinite (90%), Hostun RF sand

(5%) and silica (5%). The percentages were elaborate after several granu-
lometry curves on various proportions. This combination has a continuous
granulometry curve. This mixture will be called “sandy clay.” The principal
characteristics are given in Table 1.
The study of unsaturated triaxial paths according to the temperature re-
quired the development of a triaxial cell of 3.5 MPa of confining pressure
(Fig. 1). The heating system is composed of a heating insulated collar with
an internal diameter of 146 mm and a 230 mm height, developing a power of
Table 1. Description of the material used
Material wL (%) IP (%) % < 80 μm % < 2 μm d10 μm d60 /d10

Sandy Clay 38 19 95 53 0.2 21
Effect of the Temperature on the Behaviour of Unsaturated Sandy Clay 245
Fig. 1. Unsaturated triaxial cell with controlled temperature and imposed negative
pore water pressure by the osmotic technique
1800 W. This collar comes to wrap an aluminum hollow roll. The imposition
of a given temperature to the sample is controlled by an electronic regulator
of temperature and a Pt100 probe. A thermocouple is used to measure the
imposed temperature. Regulation and measuring equipment was developed.
Like system of safety, thermostats were fixed in various places of the cell. To
impose or control pressures and volumes, three pressure-volume controllers
were used. The axial loading is carried out using a 25 kN loading frame. This
instrumentation is controlled via a HP station.
In addition, a special pedestal at the bottom of the sample was developed
to make it possible to impose negative pore water pressures using the circula-
tion of a PEG solution in contact with the sample via a dialysis membrane. A
brass plate, 2.5 mm in thickness and 35 mm in diameter, perforated by holes
of 1.5 mm in diameter, comes to cap the cell pedestal. This plate serves as
basis for the sample. Between the perforated plate and the sample, a disc of
dialysis membrane is placed over the perforated plate. In order to prolong the
lifespan of the dialysis membrane and to avoid its degradation by the bacteria
naturally present in the soil, a small quantity of benzoic acid, acting like an
anti-bacterial, was added to the PEG solution. In addition, the imposition of
a negative pore water pressure in the soil sample using the osmotic technique
is based on the assumption that the air pressure within the sample is equal
to the atmospheric pressure. This condition is ensured by putting the drain-
ing circuit at the top of the sample to the atmospheric pressure. To maintain
a constant concentration of the osmotic solution in the circuit, a circulation
between the chamber of the pedestal and an external tank is carried out by
means of a peristaltic pump of low flow.
3 Results and discussion

3.1 Effect of Temperature on Pre-Consolidation Stress in Isotropic
Paths
To study the effect of temperature on the preconsolidation pressure, several

tests were carried out at four values of temperature (22, 40, 60 and 80◦ C).
Samples were first preconsolidated at isotropic effective stress p = 200 kPa,
then heated at different temperatures. At equilibrium, isotropic effective stress
was increased with a constant temperature. The preconsolidation stress pc is
evaluated as the stress value at the intersection of the two linear parts of
the compression curves (logarithm of mean effective stress versus void ratio).
One observes a quasi linear decrease of the preconsolidation pressure when
the temperature increases (Fig. 2). This is in agreement with the results from
the literature (Sultan et al. 2002, Cekerevac and Laloui 2004). Cekerevac and
Laloui (2004) suggested a semi-logarithmic expression to describe this decrease
σc (T ) = σc (To ){1 − γ log[T /To ]} . (1)
3.2 Effect of Temperature on Secant Modulus
E50 Secant modulus is defined as
240 p'c
Pre-consolidation stress (kPa)
200
160
120
T
20 40 60 80 100
Temperature (°C)
Fig. 2. Influence of temperature on preconsolidation pressure
qmax /2
E50% = . (2)
ε50%
where, qmax is the maximum deviatoric strength and ε50% is the axial strain
corresponding to 50% of maximum deviatoric strength.
Deviatoric loading tests on triaxial paths were carried out at three values
of temperatures (22, 60 and 80◦ C). Figure 3 shows the variation of secant
modulus versus effective confining stress. One observes that for a given tem-
perature, the secant modulus increases with effective confining stress. For a
given effective confining stress, the secant modulus decreases when the tem-
perature increases. The variation of secant modulus versus confining stress
follows Hertz’s law E = α(σ3 )n (Biarez and Taibi 1997), parameterized in
temperature. Table 2 summarizes the parameters α and n according to the
studied temperatures.
For unsaturated tests, several deviatoric loading tests were carried out
on unsaturated samples at both constant suction and temperature. Figure 4
shows the variation of secant modulus versus suction for two values of tem-
perature (22 and 80◦ C). The total confining stress was maintained equal to
1250 kPa. It is noted that the secant modulus increases with suction following
a bi-logarithmic law E = β(uw )m (Fleureau and Kheirbek-Saoud 1992). In
addition, for a given suction, secant moduli at T = 80◦ C are lower than those
50
E50
Secant Moduli E50 (MPa)
40
T= 22°C
T= 60°C
30 T= 80°C
20
σ'3
10
100 1000
Confining effective stress (kPa)
Fig. 3. Variation of secant modulus versus effective confining stress for three tem-
perature
Table 2. Parameters α and n according to the studied temperatures
Temperature (◦ C) n α
22 0.64 0.086
60 0.39 2.63
80 0.27 14
20
E50 T= 22°C
Secant Moduli E50 (MPa)
T= 80°C
10
9
8
7
6
-uw
5
1 2 3 4 5 6 7 8 910
Suction (MPa)
Fig. 4. Variation of secant modulus versus suction for two values of temperature
(22 and 80◦ C)
corresponding to T = 22◦ C. In all the cases, one notes a loss of rigidity when
the temperature increases.
3.3 Effect of Temperature on the Shear Strength
Figure 5 shows an example of results of unsaturated triaxial compression tests

on a normally consolidated sandy clay for two imposed suctions s = −uw (3
and 8.5 MPa) and two temperatures (22 and 80◦ C) in [p + s, q] plan. It is
noticed that the envelope of maximum strengths decreases when the temper-
ature increases and for a given temperature, maximum strength increases with
suction (Wiebe et al. 1998, Ghembaza 2004).
q (kPa)
1500 M = 1,Saturated Criteria Failure at T = 22°C
T = 22°C
1300 T = 80°C
1100 T = 22°C
s = 3 MPa
900 s = 8.5 MPa
700 T = 80°C
s = 3 MPa
500 s = 8.5 MPa
300
100 p'=p-(-uw) (kPa)

0 4000 8000 12000
Fig. 5. Unsaturated triaxial compression tests on a normally consolidated sandy

clay for two imposed suctions (3 and 8.5 MPa) and two temperatures (22 and 80◦ C)
in [p + s, q] plan
4 Conclusion
The results of laboratory study using a temperature controlled triaxial cell

have been presented. Thermal effects on the mechanical behaviour of unsat-
urated sandy clay were analyzed by comparing tests at various temperatures
and suctions. In the case of saturated soils, the preconsolidation pressure de-
crease when the temperature increases. This observation is in agreement with
the results from the literature. It is about a negative hardening. In the case
of unsaturated soils, the preconsolidation pressure increases with suction for
a given temperature. It is about a hydrous hardening. For a null suction, the
secant modulus follow a Hertz law E = α(σ 3)n with n decreasing when the
temperature increases. For a non-null suction, the secant modulus increases
with suction. For a given suction, this modulus decreases when temperature
increases. In the case of unsaturated soils, the maximum strength decreases
when the temperature increases, in spite of the hardening produced by the
negative pressure. This strength increases with the suction, at a given tem-
perature.
References
Belanteur N, Tachrifet S, Pakzad M (1997) Etude des comportements mécanique,
thermo-mécanique et hydromécanique des argiles gonflantes et non gonflantes
fortement compactéesm Revue Française de Géotechnique 78:31–51
Biarez J, Taibi S (1997) Les modules en mécanique des sols. Journée du CFMS ‘Les
essais de laboratoires’ 22 Janvier 1997
Cekerevac C, Laloui L (2004) Experimental study of thermal effects on the mechan-
ical behaviour of a clay, Int J Numer Meth Geomech 28:209–228
De Bruyn D (1999) Influence d’une élévation de température sur le comporte-
ment physique et mécanique de l’argile de boom dans le cadre de la probléma-
tique de galeries d’enfouissement de déchets radioactifs. PhD Thesis, Université
Catholique de Louvain
El Youssoufi MS, Jamin F, Saix C (2002) Quelques aspects du comportement hy-
dromécanique lors de la saturation de l’état sec, Bulletin du GFHN Milieux
poreux et transferts Hydriques 48:141–146
Fleureau JM, Kheirbek-Saoud S (1992) Strength of compacted soils in relation to
the negative pore pressure, Revue Française de Geotechnique 59:57–64
Ghembaza MS (2004) Etude du comportement thermo-hydro-mécanique des sols
argileux soumis à une pression interstitielle négative, Thèse de Doctorat, Uni-
versité du Havre, France
Jamin F (2003) Contribution a l’étude du transport de matière et de la rhéologie dans
les sols non saturés a différentes températures. Thèse de Doctorat, Université de
Montpellier II, France, 260 p
Rahbaoui A (1996) Modélisation du comportement Thermo-Mécanique des argiles
saturées : Application au stockage des dechéts radioactifs. Thèse de doctorat,
Université d’Orléans, 293 p
Romero EM (1999) Characterization and thermo-hydro-mechanical behaviour of
unsaturated boom clay and experimental study. PhD Thesis, Univ Polytech
Catalunya, Barcelona
Sultan N, Delage P, Cui YJ (2002) Temperature effects on the volume change be-
haviour of Boom clay, Engineering Geology 64:135–145
Tanaka N, Graham J, Crilly T (1997) Stress-strain behaviour of reconstituted illitic
clay at different temperatures, Engineering Geology 47:339–350
Wiebe B, Graham J, Tang GXM, Dixon D (1998) Influence of pressure, saturation,
and temperature on the behaviour of unsaturated sand-bentonite, Can Geotech
J 35:194–205
Influence of Temperature on the Water
Retention Curve of Soils.
Modelling and Experiments
Simon Salager, Moulay Saïd El Youssoufi, and Christian Saix
LMGC UMR CNRS 5508, Université Montpellier 2, Place Eugène Bataillon

34095 Monpellier Cedex 5, France
salager@lmgc.univ-montp2.fr, elyous@lmgc.univ-montp2.fr,
saix@lmgc.univ-montp2.fr
Summary. We investigate the influence of temperature on the retention curve of

soils. This curve represents the constitutive relation between water content w and
suction s, for a given temperature T and a given void ratio e. We present a model
based on the differential of suction as a function of T , w and e. When adjusted for
a retention curve obtained at a given temperature, this model enables to predict
this curve for any temperature. In parallel, we carried out experiments on a clayey
silty sand by using a pressure cell immersed in a thermostatic bath. The model was
validated by several tests on the clayey silty sand at 20 and 60◦ C. The application
of the model to data found in the literature confirms its predictive power for a wide
range of porous materials. These results allow us to plot the retention surface, from
experimental tests obtained at a given temperature and from modelling. It can be
considered as a generalization of the classical retention curve. Finally, we discuss
the influence of the void ratio variation during experiments on the curve predicted
by the model.
Key words: retention curve, thermo-hydric behaviour, pressure cell, soil, porous
media
1 Introduction
Water retention curve is the constitutive relation between suction s and water
content w or water volume fraction θ or degree of saturation Sr . This relation
is always done for a given temperature T and a given void ratio e. The general
shape of this curve depends particularly on the porosity, the pore-geometry
characteristics, the grain-size distribution, and the nature (granular or clayey)
of the material. Besides its capability to describe the hydraulic behaviour,
some authors used it to yield the coefficient of permeability (Fredlund et
al. 1994, Van Genuchten 1980) or to analyse shear strength for unsaturated
soils (Fredlund et al. 1995). Finally, water retention curve gives a conceptual
252 S. Salager et al.
framework to understand the hydraulic behaviour of an unsaturated porous

media at a given temperature. However, several environmental problems like
energy structures (Saix 1987, Laloui et al. 2003), nuclear waste storage in deep
clay (Olchitzky 2002), require to take into account the temperature variable
in the analysis of hydraulic behaviour. In particular, it is necessary to study
the temperature dependence of water retention curve.
This paper presents a contribution to the characterization of the thermo-
hydraulic behaviour of unsaturated porous media. From a water retention
curve obtained experimentally at a given temperature, a theoretical model
allows to predict this curve for other temperatures, reducing then extensively
experimental tests to get water retention curve at several temperatures.
The model was validated by the authors on experiments realized on a
ceramic and on a clayey silty sand. Its application to other data found in
literature confirms its predictive power for a wide range of porous media.
2 Constitutive Thermo-Hydraulic Model
Several mechanisms can be proposed to explain the influence of temperature

on the hydraulic behaviour: thermal expansion of the different phases, surface
tension of water (liquid phase consists of pure water only), and contact angle
variations (Bachmann and Van Der Ploeg 2002). Our model is based on the
development of the differential of suction s with respect to the three indepen-
dent variables: water volume fraction θ (volume of water per unit apparent
volume), temperature T and void ratio e. To avoid hysteresis problem, we
consider here only the drying path from a saturated state.
We can then express the suction differential by the relation:

∂s ∂s ∂s
ds = dθ + dT + de . (1)
∂θ T,e ∂T θ,e ∂e T,θ
This relation can be explained by expressing the different partial derivative

of s from physical relations. The water volume fraction can be written as a
function of the specific weight of solid phase ρ∗s , the specific weight of the
liquid phase ρ∗e , the water content w (mass of water per unit mass of dry soil),
and the void ratio:
ρ∗ w
θ= ∗ s . (2)
ρe (1 + e)
From relation (2), it is possible to express the differential dθ, by intro-
ducing the volumetric thermal expansion coefficient of water βe and the vol-
umetric thermal expansion coefficient of the solid phase βs . That leads to the
relation:
ρ∗s ρ∗ w ρ∗ w
dθ = dw + ∗ s (βe − βs )dT − ∗ s de . (3)
ρ∗e (1+ e) ρe (1 + e) ρe (1 + e)2
Influence of Temperature on the Water Retention Curve 253
Equation (2) also allows us to write:

∂s ρ∗e (1 + e) ∂s
= . (4)
∂θ T,e ρ∗s ∂w T,e
In the context of studies of water retention in porous media, it is common
to use Jurin’s law which express the suction as a function of the surface tension
of water σs , the mean pore radius r, and the contact angle φ.
2σs cos φ
s= . (5)
r
From equation (5), we can express the derivative of suction with respect
to temperature:
∂s s dσs dφ
= − s tan φ . (6)
∂T θ,e σs dT dT
Relation (6) takes into account the fact that the surface tension of wa-
ter and the contact angle are only dependent on temperature. Introducing
relations (3), (4), and (6) in equation (1), we obtain the final form of the
differential of suction:

s dσs dφ
ds = Fw dw + Fw w(βe − βs ) + − s tan φ dT
σs dT dT
(7)
∂s Fw w
+ − de
∂e T,θ (1 + e)
where Fw is the reciprocal

∂s of the slope of the retention curve for a given tem-
perature (Fw = ∂w T,e
). The right-hand side of the relation (7) expresses
three terms reflecting the respective contributions of the variations of water
content, temperature and void ratio, to the infinitesimal variations of suction.
The first and the second terms show strong coupling between suction, water
content and temperature. In the case of low deformable materials, the third
term become negligible in respect to the others. Its experimental determina-
tion during the thermo-hydro-mechanical path is not easy. In addition, the
infinitesimal variations of void ratio, are understood here without external
mechanical stress, the deformations are thus limited. In a first approach, this
term will be neglected. The relevance of this choice will be discussed there-
after. In the same way, the variation of the contact angle versus temperature
is very difficult to measure. One can found few results in literature about
this subject. Thus, in our experimental validation of the model, this variation
will not be taken into account. Futhermore, the volumetric thermal expan-
sion coefficient of the solid phase is assumed to be negligible, compared with
volumetric thermal expansion coefficient of water. Under these assumptions,
relation (7) can be written in the simplified form (8):

s dσs
ds = Fw dw + Fw wβe + dT . (8)
σs dT
Finally, from a retention curve obtained at a temperature T0 , the rela-

tion (8) allows to predict, for each suction value, the water content variation
as a function of the temperature variation. Thus, we obtain the water retention
curve for any temperature T .
3 Experimenal Validation
3.1 Experimental Device and Materials
For model validation, several tests were done with an experimental pressure
cell, regulated in temperature (Fig. 1). Suction is applied using the air over-
pressure method (Jamin 2003). The sample is placed in an airtight cell for
which the air pressure is regulated by pneumatic system. It is put on a porous
ceramic plate with an high air-entry pressure to allow the continuity of the
liquid phase between the sample and the hydraulic system. The airtight cell
is immersed in a thermostatic bath, where the temperature ranges from 20 to
70◦ C.
In the experimental part, we used two materials:
• The first is a ceramic, with a specific weight of 2738 kg.m−3 , a dry density
of 1870 kg.m−3 and a void ratio of 0.464. Before each loading, the samples
are saturated with chemically pure water. At saturation, water content is
16.9% at 20◦ C.
Fig. 1. General scheme of the pressure cell in its environment

• The second is a clayey silty sand with a specific weight of 2650 kg.m−3 . The
samples are compacted in order to arise a dry density of 1500 kg.m−3 . That
corresponds to a void ratio of 0.767. The saturated initial state corresponds
to a water content of 28.9% at 20◦ C.
3.2 Results and Discussion

Figure 2 shows the experimental water retention curve of the ceramic ob-
tained at 20 and 60◦ C. Figure 3 gives the same results for the clayey silty
sand. These curves are modelled from experimental data using the fitting
function proposed by Fredlund and Xing (1994). Retention curve at 20◦ C is
used as a reference for the modelling described in Sect. 2. For a temperature
of 60◦ C, results of modelling are shown in Fig. 2 for the ceramic and in Fig. 3
for the clayey silty sand. Retention curves, obtained from experiments and
modelling, are very similar in both cases. For the clayey silty sand, one can
note a divergence after a suction which corresponds to the transition between
funicular state and pendular state (Fig. 4). Indeed, the liquid phase is not
continuous anymore in the pendular state.
To complete the validation, we applied our model to data found in the
literature for two different clays:
• The first is a FoCa compacted clay (Imbert et al. 2005). Its major compo-
nent is an interstratified clay of 50% calcium beidellite and 50% Kaolinite.
It has a liquid limit of 112%, a plastic limit of 50% and a specific weight
of 2675 kg.m−3 .
• The second is an artificially prepared clay obtained from natural Boom
clay (Romero et al. 2001). It is a moderately swelling material (20–30%
kaolinite, 20–30% illite and 10–20% smectite) which has a liquid limit of
56%, a plastic limit of 29% and a specific weight of 2700 kg.m−3 .
Fig. 2. Water retention curves for the ceramic: experiments and modelling
Fig. 3. Water retention curves for the clayey silty sand: experiments and modelling
(1) (2) (3) (4)
Fig. 4. Different hydraulic states: (1) Hygroscopic state: the gas phase is contin-
uous but not the liquid phase; (2) Pendular state: continuity of each phase is not
ensured; (3) Funicular state: the liquid phase is continuous but not the gas phase,
(4) Saturated state
Fig. 5. Water retention curves for the FoCa compacted clay (Imbert et al. 2005):
experiments and modelling
Figure 5 shows the results obtained for the FoCa compacted clay; the
temperatures of the tests were 20 and 80◦ C. We can note on this figure that
experimental results at 80◦ C are in good agreement with the modelling results.
Figure 6 gives Romero’s results (Romero et al. 2001) on the Boom clay.
These tests were made for two different temperatures (22 and 80◦ C) and for
a dry unit weight of 16.7 kN.m−3 . In this case, our model is still in good
agreement with experimental data. Other results obtained by Romero et al.
(2001), in the same conditions but with a dry unit weight of 13.7 kN.m−3 , were
well predicted by the model. The hypothesis which could result in neglecting
the third term of the relation (7) with respect to the others seems acceptable.
In fact, although the variation of void ratio can be not negligible on the whole
test, it does not seem necessary to take into account explicitly this variation
in the calculation. Indeed, the determination of the reciprocal slope of the
water retention curve Fw integrates implicitly the contribution of the gradual
deformation of the material during the hydraulic path.
Comparison between modelling and experimental results, for four different
materials, demonstrates the capability of relation (8) to predict retention curve
for any temperature from a retention curve made at a given temperature.
Finally, these results allow us to plot a water retention surface w = w(s, T ).
This surface obtained from experimental tests at a given temperature and
from modelling, can provide a conceptual framework to understand thermo-
hydraulic behaviour of unsaturated porous media.
Fig. 6. Water retention curves Boom clay (Romero et al. 2001): experiments and
modelling
4 Conclusion
We investigated the influence of temperature on the water retention curve of
soils. We presented a model based on the differential of suction as a function
of temperature, water content and void ratio. When adjusted for a water
retention curve obtained at a given temperature, this model enables to predict
this curve for any temperature. In parallel, we carried out experiments on
a clayey silty sand and on a ceramic by using a pressure cell immersed in
a thermostatic bath. The model was validated by several tests on these two
materials at 60◦ C. The application of the model to data found in the literature
confirmed its predictive power for a wide range of porous materials. These
results allowed us to plot the water retention surface, from experimental tests
obtained at a given temperature and from modelling. It can be considered as
a generalization of the classical retention curve.
References
Bachmann J, Van Der Ploeg RR (2002) Soil Sci 165:468–478
Fredlund DG, Xing A (1994) Can Geotech J 31:521–53
Fredlund DG, Xing A, Huang S (1994) Can Geotech J 31:533–545
Fredlund DG, Xing A, Fredlund MD, Barbour SL (1995) Can Geotech J 32:440–448
Imbert C, Olchitzky E, Lassabatère T, Dangla P, Courtois A (2005) Engineering
Geology 81:269–283
Jamin F (2003) Contribution à l’étude du transport de matière et de la rhéologie
dans les sols non saturés à différentes températures. PhD Thesis , Université
Montpellier 2, Montpellier
Laloui L, Moreni M, Vulliet L (2003) Can Geotech J 40:388–402
Olchitzky E (2002) Couplage hydromécanique et perméabilité d’une argile gonflante
non saturée sous sollicitations hydriques et thermiques. PhD Thesis, Ecole Na-
tionale des Ponts et Chaussées, Paris
Romero E, Gens A, Lloret A (2001) Geotechnical and Geological Engineering
19:311–332
Saix C (1987) Contribution à l’étude des sols non saturés en température. Applica-
tion à l’exploitation thermique des sous-sols. PhD Thesis, Université Montpellier
2, Montpellier
Van Genuchten MT (1980) Soil Sci Soc Am J 44:892–898
Thermo-Hydro-Mechanical Behaviour
of Compacted Bentonite
Abbass Tavallali1 , Anh-Minh Tang2 , and Yu-Jun Cui2

1
Department of Civil Engineering, Katholieke Universiteit Leuven, Heverlee,
Belgium Abbass.Tavallali@bwk.kuleuven.be
2
Ecole Nationale des Ponts et Chaussées, CERMES, Paris, France
tang@cermes.enpc.fr, cui@cermes.enpc.fr
Summary. Like every human activity, the nuclear activities generate waste. Be-
cause of the particular characters of the nuclear waste, a special geological formation
and also plug material for storage of the high level nuclear waste is necessary.
The plug material plays an important role in confining the radioactive waste.
Therefore, one needs a material with the character of very low permeability and a
good capacity of holding the radionuclides. The compacted bentonite will thus be
employed as such plug material. So it is very important to evaluate the swelling
characters of compacted bentonite.
Here, it is presented an experimental study of thermo-hydro-mechanical be-
haviour of compacted bentonite of MX–80, based on tests carried out in an isotropic
high-pressure cell (60 MPa) in the range 20 and 80◦ C (±0.1◦ C). The results present
the deformation of compacted bentonite as a function of temperature, suction and
pressure.
Key words: nuclear waste, temperature effects, compacted bentonite, MX–80,

swelling soils, suction, laboratory tests
1 Introduction
Swelling soils such as bentonite have recently attracted increasing attention

as back-filling (buffer) materials for high-level nuclear waste (Swedish Nuclear
Fuel Supply Co./Division KBS 1983, Yong et al. 1985, Bucher and Mayor
1989).
High-density compacted bentonite (Fig. 1) in the form of blocks has also
been proposed as sealing material in high-level radioactive waste repositories,
as it provides very low permeability, high exchange capacity, sufficient thermal
conductivity, and adequate mechanical resistance (Lloret et al. 2003).
The function of buffer material is to create an impermeable zone around
the canister containing the high-level nuclear waste, since nuclear waste must
be kept separate from the surrounding environment (Komine and Ogata 1996).
260 A. Tavallali et al.
Fig. 1. Proposed engineered barrier made up of compacted bentonite clay for a

radioactive waste repository (Lloret et al. 2003)
The buffer materials must have swelling properties because cracks in the sur-
rounding rocks may appear and need to be filled up. Compacted bentonite
will therefore be used as buffer material (Komine and Ogata 1996). To develop
buffer materials using compacted bentonite, it is important to evaluate the
swelling characteristics of compacted bentonite.
2 Materials and Methods

In this study MX–80 is selected to evaluate the thermo-hydro-mechanical
behaviour of compacted bentonite. It is focused on thermal volume-change
of MX–80 at different states of suction and pressure. The characteristics of
MX–80 are presented in Table 1. MX–80 (Pusch 1982) is a Na+ –Ca+ 2 smectite
with a clay fraction (< 2 μm) of 85% and a montmorillonite fraction of about
80–90% of this fraction. The remaining silt fraction mainly consists of quartz
and feldspars, with some micas, sulphides, and oxides (Marcial et al. 2002).
The tests were performed in temperature-controlled high-pressure isotropic
cells (temperature and pressure varied between 20 and 80◦ C ± 0.1◦ C, and 0–
60 MPa, respectively), connected to GDS pressure volume controllers. Volume
changes were obtained from the GDS monitoring of the all round water ex-
changes (Sultan et al. 2000).
Initially MX–80 is in the powder shape. To dry the powder and to sta-
bilize the initial conditions, a desiccator containing saturated salt solutions
of K2 CO3 is used. The suction was controlled by imposing controlled humid-
ity conditions in desiccator containing saturated salt solutions. In the vapour
Thermo-Hydro-Mechanical Behaviour of Compacted Bentonite 261
phase method of controlling suction, saturated salt solutions are used to con-
trol the relative humidity of the atmosphere of desiccators in which the sam-
ple is placed (Marcial et al. 2002). This powder is dried under the suction of
113 MPa (Table 2). When the mass of the powder does not change during a
few days, it can be said that stabilization is finished. With this suction the
water content is about 10%.
For this study, disc-shaped samples with a diameter of 80 mm and a height
of 12 ± 2 mm were needed. The procedure of the compaction and then ma-
chining were used to produce the samples (Tavallali 2005).
In the compaction step, loading up to the 40 MPa is carried out stage
by stage (0.1–0.2–0.5–1–2–5–10–20–40 MPa). For each stage it is necessary to
wait until the volume of water which left GDS becomes constant. The duration
of this step is about six hours. In this part, suction in the sample changes
because of changing the volume of void under isotropic compression. Then
to reach the initial suction, the sample is placed in the desiccator containing
saturated salt solutions of K2 CO3 (relative humidity of 44% and the suction of
113 MPa). For machining it is necessary to be sure that the mass of compacted
sample is stabilized (similar way for powder of MX–80). By a special machine
in CERMES, it is possible to have a cylindrical sample with a diameter of
80 mm. Then by sawing manually every 12 mm, a disc of 80 mm in diameter
and 12 mm in height is produced. It is important to put the discs immediately
in the preceding desiccator to reach the initial suction of 113 MPa.
To evaluate the thermo-hydro-mechanical behaviour of unsaturated com-
pacted bentonite, having the samples with different suctions is necessary. With
different salt solutions it is possible to have different suctions. In humidifying
Table 1. Identification parameters of MX–80 (Marcial et al. 2002)
Mineralogy 85% Na–Ca smectite

Liquid limit, wL (%) 520
Plastic limit, wP (%) 62
Plasticity index, IP 458
Unit mass of the solids, ρs (Mg/m3 ) 2.65
Table 2. Salts used in the vapour phase method of controlling suction (Marcial et
al. 2002)
Salt Suction (MPa)

K2 SO4 4.2
KNO3 9
NaCl 38
NaNO2 57
Mg(NO3 )2 82
K2 CO3 113
Table 3. Test program of thermal isotropic compression

Initial conditions Loading
Test ρd Suction 1 2
(g/cm3) (MPa) P (MPa) T (◦ C) P (MPa) T (◦ C)
Iso 8∗ 1.421 39 0.1 25 50 25
Iso 9 1.203 9 0.1 25 0.1 80
Iso 10 1.422 39 0.1 25 5 25
Iso 11 1.205 9 0.1 20 50 20
Iso 12 1.206 9 0.1 25 0.1 80
Iso 13 1.421 39 0.1 25 5 25
∗
There is a cycle of 25–75–25◦ C at P = 5 MPa
the samples, suction change must be stage by stage. For example to decrease
the suction from 113 MPa to 9 MPa it is necessary to use salt solutions of
Mg(NO3 )2 , NaNo2 , NaCl, KNO3 (82, 57, 38, 9 MPa, respectively) stage by
stage. Changing the salt solutions only after mass stabilization of the ben-
tonite is possible.
Table 3 shows the test program carried out in this study. As in this study
the vapour phase method of controlling suction is used, when the tempera-
ture in a test is 25◦ C (the temperature of environment), the same salt solution
which is already used to humidify the sample is used, even in the fast increase
or decrease of temperature (approximately 10 h). It is supposed that in these
two cases suction does not change. But when it is plan to increase the tem-
perature and to stabilize it during 1 or 2 months, suction increases and thus
it is necessary to use another salt solution (that already has used to humidify
the sample). For example when initial suction is 9 MPa (by using salt solu-
tion KNO3 ) and test must carry out in 80◦ C, it is necessary to use the salt
solution K2 SO4 with the suction of 4.2 MPa. Because when the temperature
increases, suction increases and will be roughly equal to initial suction (9 MPa)
and thus the test carry out under constant suction. With this condition one
can evaluate the influence of the temperature on the behaviour of compacted
bentonite.
3 Test Results
The aim of this study is to evaluate the thermal volume-change behaviour

of the bentonite. The volume changes were measured by monitoring the ex-
changes of the water contained in the cell (Sultan 1997) using a pressure–
volume controller (GDS) of high capacity (60 MPa maximum pressure). Fig-
ure 2 shows the effect of heating on the thermal volumetric strain of the
bentonite. This figure shows the results for the tests Iso 8, Iso 9 and Iso 12.
Fig. 2. Temperature effect on the thermal volume-change for the tests Iso 8, Iso 9
and Iso 12
Figure 2 shows that heating the samples Iso 9 and Iso 12 initially produced a
small thermal dilation until 40◦ C. Then when the temperature exceeds 40◦ C,
a thermal contraction can be observed. Also this figure shows that volume
strain is very small and therefore temperature change of the environment can
easily influence this deformation. This problem exists for the test Iso 8 until
50◦ C. In spite of that, the thermal volume behaviour of the bentonite in test
Iso 8 is similar to the tests Iso 9 and Iso 12. The important point in the test
Iso 8 is the irreversible thermal volume strain behaviour of bentonite. In this
test when the temperature was decreased a very small thermal contraction
was continued.
Figure 3 shows the variation of the void ratio according to pressure. This
figure presents the results for the tests Iso 8, Iso 10, Iso 11 and Iso 13. A cycle
of heating–cooling on the sample Iso 8 is applied. In spite of that, Figure 3
shows that the variations of void ratio of the samples Iso 8, Iso 10 and Iso 13
under the applied pressures of 0.1 MPa to 5 MPa are roughly identical. Thus it
can be concluded that the thermal volume–change is smaller than the volume–
change by the pressure. These three samples (Iso 8, Iso 10 and Iso 13) have
the same suctions of 39 MPa but the suction of the sample Iso 11 is 9 MPa.
By comparing these four samples it can be concluded that the influence of
suction on the volume-change is considerable. The results show that the effect
of suction on the volume–change is larger than the effect of the temperature.
Fig. 3. Variation of void ratio in the function of pressure for the tests Iso 8 (cycle
of 25–75–25◦ C at P = 5 MPa), Iso 10, Iso 11 and Iso 13
4 Conclusion
Figures 2 and 3 give some of the thermo-hydro-mechanical behaviour of com-

pacted bentonite. They can be summarized as follows:
1. The samples under the heating have a small dilation until 40◦ C and then
contract and during cooling a small contraction continue, in agreement
with Sultan et al. (2000).
2. The cycle of the heating–cooling of the sample is irreversible, in agreement
with Sultan et al. (2000).
3. The thermal volume–change of bentonite under the effect of the temper-
ature is very small (approximately 1%).
4. The thermal volume–change in comparison to effects of the pressure is
negligible, in agreement with Campanella and Mitchell (1968) and Cui et
al. (2000).
5. By comparing the result of this study with the result of saturated clays
(Cui et al. 2000) it can be concluded that the thermo-hydro-mechanics
behaviour of unsaturated compacted bentonite is similar to the saturated
soil.
6. The thermal volumetric deformation in comparison with the effects of
suction is negligible.
7. When suction increases the initial void ratio and the slope of the e − log p
diagram decrease, in agreement with Lloret et al. (2003).
References
Bucher F, Mayor PA (1989) Medium-scale experiments on high-compacted ben-
tonite. In: 12th international Conference on Soil Mechanics and Foundation En-
gineering, Rio de Janeiro 1:583–585
Campanella RG, Mitchell JK (1968) Influence of temperature variations on soil
behaviour, J Soil Mech Found Div, ASCE 94:709–734
Cui YJ, Sultan N, Delage P (2000) A thermomechanical model for saturated clays,
Komine H, Ogata N (1996) Prediction for swelling characteristics of compacted
bentonite, Can Geotech J 33(1):11–22
Lloret A, Villar MV, Sànchez M, Gens A, Pintado X, Alonso EE (2003) Mechanical
behaviour of heavily compacted bentonite under suction changes, Géotechnique
53(1):27–40
Marcial D, Delage P, Cui YJ (2002) On the high stress compression of bentonites,
Pusch R (1982) Mineral-water interactions and their influence on the physical be-
haviour of highly compacted Na bentonite, Can Geotech J 29:1044–1059
Sultan N, Delage P, Cui YJ (2000) Comportement thermomécanique de l’argile de
Boom, CR Acad Sci Paris Série II b 328:457–463
Sultan N (1997) Étude du comportement thermo-mécanique de l’argile de Boom:
expériences et modélisation. Ph.D. Thesis École nationale des ponts et chaussées
Swedish Nuclear Fuel Supply Co./Division KBS (1983) Final storage of spent nu-
clear fuel-KBS-3, III barriers. Swedish Nuclear Fuel Supply Co./Division KBS
Technical Report pp 9:1–16:12
Tavallali A (2005) Comportement thermo-hydro-mécanique de la bentonite com-
pactée non saturée, Rapport de stage, École nationale des ponts et chaussées
Yong RN, Boonsinsuk P, Yiotis D (1985) Creep behaviour of a buffer material for
nuclear fuel waste vault, Can Geotech J 22:541–550
Retention Curves of Two Bentonites
at High Temperature
María Victoria Villar and Roberto Gómez-Espina
CIEMAT, Avd. Complutense 22, 28040 Madrid, Spain

mv.villar@ciemat.es, roberto.gomez@ciemat.es
Summary. Two methods were developed to determine the retention curve for a
fixed dry density and at high temperature. The materials used are the MX–80 and
the FEBEX bentonites. The retention capacity of the FEBEX bentonite is higher
than that of MX–80, i.e. for a given suction and dry density, the water content of
the FEBEX bentonite is higher. However, the difference between both bentonites
becomes smaller towards the lower suctions. In both bentonites the retention ca-
pacity has decreased as temperature increased, and the difference in the retention
capacity of the two bentonites became less important at high temperature.
Key words: bentonite, temperature, retention capacity, density, water retention

curve
1 Introduction
The work presented here has been done in the context of projects related to
the study of the bentonite engineered barrier in high-level radioactive waste
(HLW) repositories. Most probably the bentonite will be placed in the repos-
itory in the form of compacted blocks around the waste containers. Since the
barrier will be manufactured with the clay at its hygroscopic water content,
the bentonite will initially have degree of saturation lower than 100%. Due to
the heat produced by the radioactive decay of the wastes, the barrier will be
subjected to temperatures of up to around 100◦ C, depending on the particular
disposal concept.
One of the properties that have to be known to analyse and model the hy-
dration of the bentonite barrier is its water retention capacity. Previous inves-
tigations have shown that the retention curve (or water adsorption isotherm),
for the same initial conditions of the material, differs significantly depending
on the volume restriction imposed to the sample during the determination.
Since the state of the bentonite in the barrier will be mostly confined, the
methods developed to determine the retention curve must keep the volume of
268 María Victoria Villar and Roberto Gómez-Espina
the material constant (Villar and Lloret 2004). They must also try to repro-
duce the temperatures expected in the barrier.
The materials tested are two bentonites compacted to high densities: the
MX–80 and the FEBEX. Both have been selected by different countries as
reference materials for the sealing of HLW repositories.
2 Material
Two bentonites have been used in this investigation: the Spanish FEBEX
bentonite and the American MX–80 bentonite.
The FEBEX bentonite comes from the Cortijo de Archidona deposit
(Almería, Spain) and its characterisation can be found in ENRESA (1998,
2000), Villar (2002), Fernández (2003). The smectite content of the FEBEX
bentonite is higher than 90% (92 ± 3%) and it contains variable quantities of
quartz (2 ± 1%), plagioclase (2 ± 1%), K-felspar, calcite and opal-CT. The
CEC varies from 96 to 102 meq/100 g, and the major exchangeable cations
are Ca (35–42 meq/100 g), Mg (31–32 meq/100 g), Na (24–27 meq/100 g) and
K (2–3 meq/100 g). The hygroscopic water content in equilibrium with the
laboratory atmosphere is 13.7 ± 1.3%.
The MX–80 bentonite is extracted from Wyoming (USA). It is a worldwide
known material supplied in the form of powder homoionised to sodium. The
MX–80 bentonite consists mainly of montmorillonite (65–82%). It also con-
tains quartz (4–12%), feldspars (5–8%), and smaller quantities of cristobalite,
calcite and pyrite. The CEC is 74 meq/100 g, and the major exchangeable
cations are Na (61 meq/100 g), Ca (10 meq/100 g) and Mg (3 meq/100 g). The
hygroscopic water content at laboratory conditions is 9%.
3 Methodology
3.1 Cell Method
To determine the water retention curve at constant volume, avoiding the

swelling of the clay in wetting paths, special cells were designed (Villar 2002,
Villar and Lloret 2004). The cells consist of a stainless steel cylindrical body
with two perforated covers joined by bolts. Granulated clay is compacted
directly inside the cell ring at room temperature using static uniaxial com-
paction. The length of the specimen is 1.2 cm and its cross section area is
11.34 cm2 .
The cells are placed in desiccators with a sulphuric acid solution or with
a NaCl solution. There are temperature-dependent experimental relations be-
tween the concentration of the solution and its water activity (aw ). The cal-
culation of suction on the basis of relative humidity (RH = aw /100) is ac-
complished through Kelvin’s equation. For the cell method the suction is,
Retention Curves of Two Bentonites at High Temperature 269
therefore, imposed controlling the relative humidity. The cells allow the ex-
change of water in the vapour phase between the clay and the atmosphere
of the desiccators. Once the water content of the clay is stabilized (approxi-
mately after 2 to 3 months), the solution in the desiccators is changed in order
to apply a different suction. To determine the curve at different temperatures,
the desiccators are placed inside ovens.
3.2 Sensor/Cell Method
The drawback of the cell method is the duration of the test, because the time to
reach equilibrium at a given suction is very long. To overcome this drawback,
another method to determine the retention curve at constant volume and at
different temperatures has been developed (Villar et al. 2005, 2006). It consists
on mixing the clay with a given quantity of deionised water and compacting
this mixture to obtain a bentonite block whose relative humidity is measured
by means of a capacitive sensor while the block is kept inside a hermetic cell
made out of stainless steel.
The wetted clay is kept for two days in a plastic bag to allow homogeni-
sation of the moisture and afterwards, a block is compacted to the desired
density. The dimensions of the block are equal to the internal dimensions of
the cell, which are 7 cm diameter and 10 cm height. A hole is drilled in the
central, upper part of the block to insert the capacitive sensor and the cell is
closed. The external wall of the cell is covered with a silicone-rubber laminated
heater that fixes the temperature all over the cell. After measuring the RH
of the bentonite corresponding to the laboratory temperature, the tempera-
ture of the external heating mat is increased from 40 to 120◦ C in intervals of
20◦ C. Afterwards, the temperature is decreased according to the same pat-
tern. Each target temperature is kept for about two days. The accuracy of the
humidity sensor is ±1% over the range 0–90% RH and ±2% over the range
90–100% RH.
4 Results
4.1 FEBEX Bentonite
The sensor/cell method has been used to measure the change of suction of
the FEBEX bentonite compacted at nominal dry density of 1.60 g/cm3 during
heating/cooling. The average equilibrium values of suction measured for each
temperature are plotted in Fig. 1 and they show a decrease of the retention
capacity with temperature.
The retention curve at constant volume was determined for the bentonite
compacted to nominal dry density of 1.60 g/cm3 and temperatures of 20 and
80◦ C using the constant volume cells (Lloret et al. 2004, Villar and Lloret
2004). The initial water content of the samples was the hygroscopic one (about
140 14%
16%
120
19%
100 21%
Suction (MPa)
22%
80
60
40
20
0
20 40 60 80 100 120 140
Temperature (°C)
Fig. 1. Equilibrium suction measured during heating/cooling in blocks of FEBEX

bentonite compacted at 1.60 g/cm3 with different water contents (filled symbols:
cooling)
30
wetting 20°C
wetting 80°C
Water content (%)
25
cell/sensor 20°C
cell/sensor 80°C
20
15
10
0 20 40 60 80 100 120 140
Suction (MPa)
Fig. 2. Retention curves obtained with two methods for the FEBEX bentonite
compacted at 1.60 g/cm3
14%). The results obtained are shown in Fig. 2. A wetting path was followed
from the suction corresponding to the hygroscopic water content of the ben-
tonite (100–130 MPa) to 1 MPa. The results shown in Fig. 2 under the label
wetting are in fact the average of the results obtained in the wetting paths
followed in two cells. Figure 2 shows also some of the results obtained with
the cell/sensor method. A good agreement between the results obtained with
the two methods is observed, although at 80◦ C, the suctions measured with
the hygrometers for a given water content are higher than expected according
to the results obtained in cells.
4.2 MX–80 Bentonite
The sensor/cell method has been used to measure the change of suction of
MX–80 bentonite compacted at nominal dry density of 1.60 g/cm3 during
heating/cooling (Villar et al. 2006). The average equilibrium values of suction
measured for each temperature are plotted in Fig. 3 and show a decrease of
the retention capacity with temperature, as well as hysteresis between heating
and cooling.
The retention curves of MX–80 compacted at different densities were also
obtained in a previous investigation using the cell method (Villar 2005). Fig-
ure 4 shows the results obtained at 60◦ C with the cell method and with the
sensor/cell method for the bentonite compacted at dry density 1.60 g/cm3 .
120 w=11% w=13%

w=15% w=19%
100
Suction (MPa)
80
60
40
20
0
20 40 60 80 100 120 140
Temperature (°C)
Fig. 3. Equilibrium suction measured during heating/cooling in blocks of MX–

80 bentonite compacted at 1.60 g/cm3 with different water content (filled symbols:
cooling)
25
wetting after drying, 60°C
cell/sensor, 61°C
Water content (%)
20
15
10
5
0 20 40 60 80 100 120 140
Suction (MPa)
Fig. 4. Retention curves obtained at 60◦ C with two methods for the MX–80 ben-
tonite compacted at 1.60 g/cm3
5 Summary
Two methods have been used to determine the retention capacity at high tem-
perature of two bentonites compacted to high density. The results obtained
with both methods are coherent, although, for a given suction and temper-
ature, there is a trend to find a lower water content with the “cell/sensor
method” than with the cell method.
A decrease of retention capacity with temperature has been observed for
the two bentonites. As well, for a given temperature, the suction measured in
the sample is higher during cooling than when it was heated. This difference
is less important for the FEBEX bentonite.
Some of the results for the two bentonites are plotted together in Fig. 5
in the form of retention curves. Also, the retention capacity of the FEBEX
bentonite for a given temperature is higher than that of MX–80, i.e. for a
given suction, the water content of the FEBEX bentonite is higher. The dif-
ference tends to attenuate for the lower suctions and high temperatures. It
must be recalled that FEBEX bentonite has mainly divalent cations in the
exchange complex (though the quantity of Na is also important) and MX–80
is predominantly sodic.
The effect of dry density on the water retention capacity has been checked
in previous investigations (Villar and Lloret 2004, Villar et al. 2006). For
the two bentonites and in terms of water content the effect of dry density
is imperceptible for suctions higher than 10 MPa. For lower suctions, i.e. for
high RH and when the specimen is approaching saturation, the lower the dry
density of the bentonite, the higher the water content for a given suction.
140
MX-80, 41°C
120 MX-80, 101°C
100 FEBEX, 40°C
Suction (MPa)
FEBEX, 102°C
80
60
40
20
0
10 12 14 16 18 20 22 24
Water content (%)
140
MX-80, 26°C
120 MX-80, 81°C
FEBEX, 82°C
100
Suction (MPa)
FEBEX, 26°C
80
60
40
20
0
10 12 14 16 18 20 22 24
Water content (%)
Fig. 5. Retention curves at various temperatures for MX–80 and FEBEX bentonites
compacted to nominal dry density 1.60 g/cm3
Nevertheless, the influence of density seems to be less important for the MX–
80 than for the FEBEX bentonite (Villar et al. 2005).
References
ENRESA (1998) FEBEX. Bentonite: origin, properties and fabrication of blocks.
Publicación Técnica ENRESA 4/98, Madrid
ENRESA (2000) FEBEX Project. Full-scale engineered barriers experiment for a
deep geological repository for high level radioactive waste in crystalline host
rock. Final Report. Publicación Técnica ENRESA 1/2000, Madrid
Fernández AM (2003) Caracterización y modelización del agua intersticial en mate-

riales arcillosos: Estudio de la bentonita de Cortijo de Archidona. Ph.D. Thesis,
Universidad Autónoma de Madrid, Madrid (In Spanish)
Lloret A, Romero E, Villar MV (2004) FEBEX II Project. Final report on thermo-
hydro-mechanical laboratory tests. Publicación Técnica ENRESA 10/04, Madrid
Villar MV (2002) Thermo-hydro-mechanical characterisation of a bentonite from
Cabo de Gata. A study applied to the use of bentonite as sealing material in
high level radioactive waste repositories. Publicación Técnica ENRESA 01/2002,
Madrid
Villar MV (2005) MX–80 bentonite. Thermo-hydro-mechanical characterisation per-
formed at CIEMAT in the context of the Prototype Project. Informes Técnicos
CIEMAT 1053, Madrid
Villar MV, Lloret A (2004) Influence of temperature on the hydro-mechanical be-
haviour of a compacted bentonite, Applied Clay Science 26:337–350
Villar MV, Martín PL, Lloret A (2005) Determination of water retention curves
of two bentonites at high temperature. In: Tarantino A, Romero E, Cui YJ
(eds) Advanced experimental unsaturated soil mechanics. EXPERUS 2005. A.A.
Balkema Publishers, London pp 77–82
Villar MV, Gómez-Espina R, Martín PL (2006) Behaviour of MX–80 bentonite at
unsaturated conditions and under thermo-hydraulic gradient. Work performed
by CIEMAT in the context of the TBT project. Informes Técnicos CIEMAT
1081, Madrid
Part V
Volumetric Behaviour – Expansive Materials

Experimental Study on Shrinkage Behaviour
and Prediction of Shrinkage Magnitudes
of Residual Soils
Sarita Dhawan1 , Anil Kumar Mishra2 , and Sudhakar M. Rao1

1
Department of Civil Engineering, Indian Institute of Science, Bangalore 560012,
India sarita@civil.iisc.ernet.in, msrao@civil.iisc.ernet.in
2
Department of Bio-production and Environmental Sciences, Faculty of
Agriculture, Kyushu University, Fukuoka 812-8581, Japan
anilk.iisc@gmail.com
Summary. The present study examines the shrinkage behaviour of residually de-
rived black cotton (BC) soil and red soil compacted specimens that were subjected
to air-drying from the swollen state. The soil specimens were compacted at varying
dry density and moisture contents to simulate varied field conditions. The void ratio
and moisture content of the swollen specimens were monitored during the drying
process and relationship between them is analyzed. Shrinkage is represented as re-
duction in void ratio with decrease in water content of soil specimens. It is found
to occur in three distinct stages. Total shrinkage magnitude depends on the type
of clay mineral present. Variation in compaction conditions effect marginally total
shrinkage magnitudes of BC soil specimens but have relatively more effect on red soil
specimens. A linear relation is obtained between total shrinkage magnitude and vol-
umetric water content of soil specimens in swollen state and can be used to predict
the shrinkage magnitude of soils.
Key words: residual soils, clay mineralogy, compaction conditions, shrinkage, pre-
diction, volumetric water content
Introduction
Residually derived clay soils like black cotton (BC) soil and red soil in India
cover around 32% of the total landmass. These generally occur in the semi-
arid regions and are problematic to engineering structures because of their
tendency to heave during wet season and shrink during dry season. The mag-
nitude of these movements being dependent upon several factors still poses a
challenge for estimation.
Shrinkage is an important phenomenon in clayey soils. It occurs at lower
humidity and higher temperatures prevalent during the dry seasons. Methods
278 S. Dhawan et al.
to predict the shrinkage magnitude of soils are useful in anticipating shrink-

age settlements of soils. The present study examines the shrinkage behaviour
of two residual clay soils namely, BC soil and red soil upon drying of the
compacted specimens after swelling. The swollen specimens were allowed to
shrink until the mass of the dried specimens became nearly constant. Void
ratios and water content of the specimens were monitored during the drying
process. Relationships between void ratio and water content are analyzed to
evaluate shrinkage behaviour of the soil specimens. Based on the factors in-
fluencing the shrinkage, relation is developed to predict shrinkage magnitudes
of clay soils from the swollen state.

For the present study, BC soil was procured from Davangere district, Kar-
nataka and red soil from the site of Indian Institute of Science, Bangalore.
These soils were air-dried and pulverized to pass IS 425 μm sieve prior to
use. The BC soil mainly contains montmorillonite clay mineral (Shivananda
2003) while the red soil contains kaolinite mineral (Revanasiddappa 2000).
The physical properties of the soils were determined as per IS codes and are
listed in Table 1.
Table 1. Physical properties of residual soils
Liquid Plastic Shrinkage Plasticity Clay Silt Sand USCS

Soil
limit (%) limit (%) limit (%) Index (%) (%) (%) symbol
BC 84 23 10 61 55 29 16 CH
Red 39 19 15 20 24 16 60 CL
Compaction curves of the soils were determined by the Standard Proctor

method as per IS code. For the present study four compaction conditions were
chosen. These are:
1. Maximum Dry Density (MDD) & Optimum MoistureContent (OMC)
2. 95% MDD & OMC
3. MDD & 5% dry of OMC
4. 95% MDD & 5% wet of OMC
These conditions were chosen to bring out the significance of variations
in the initial compaction condition on the shrinkage behaviour of the soil
specimens and are listed in Table 2.
Determination of Shrinkage of Compacted Soil Specimens

The soils were mixed at the desired water content and were statically com-
pacted in rings of 75 mm diameter to 14 mm height at the desired dry density.
Experimental Study on Shrinkage Behaviour 279
Table 2. Compaction conditions of soil specimens
Compaction condition BC soil Red soil

3
MDD & OMC Dry density [Mg/m ] 1.42 1.73
Water content [%] 28 18
95% MDD & OMC Dry density [Mg/m3 ] 1.35 1.64
MDD & 5% dry of OMC Dry density [Mg/m3 ] 1.42 1.73
95% MDD & 5% dry of OMC Dry density [Mg/m3 ] 1.35 1.64
Details of experimental setup can be found in Sarita (2003). The specimen

assembly was placed in a tray, inundated with distilled water and allowed
to swell under 6.25 kPa pressure. After two days of inundation, water was
removed and the specimens were allowed to air-dry at 32 ± 3◦ C under the
applied pressure. During drying process, the mass, height and diameter of
each specimen were measured regularly. The reduction in volume and mass
are used to calculate the void ratio and water content of the specimen at any
stage of drying from the relations:
Δv
Δe = × (1 + e0 ) , (1)
V0
md − m s
w= × 100 (2)
ms
where V0 and e0 are the volume and void ratio of the specimen in the swollen
state, ms is dry mass of soil used for specimens, Δv and Δe are the change in
specimen volume and void ratio at any stage of drying, md and w are the mass
and water content of the specimen at any stage of drying. The drying process
was terminated when the mass of the specimens became nearly constant. The
time taken for complete air-drying of the specimens was about two months.
At each compaction condition of a soil, triplicate specimens were used and
their average was taken for calculations.
Shrinkage Behaviour of Compacted Specimens
Variation in void ratio with decrease in water content upon drying the swollen
BC soil specimens is shown in Fig. 1. The shrinkage of the specimens occur
in three stages. In stage one termed as Initial Shrinkage (IS), a small change
in void ratio occurs with reduction in water content. The IS region is followed
by the Primary Shrinkage (PS) region where the void ratio rapidly decreases
1.4
® MDD and OMC
£ MDD and 5% dry of OMC
1.2 p 95% MDD and OMC
¡ 95% MDD and 5% wet of OMC
1
Void ratio
0.8
0.6
0.4
0.2
0 5 10 15 20 25 30 35 40 45 50
Water content (%)
Fig. 1. Void ratio variation with water content during shrinkage of compacted BC
soil specimens after swelling
with reduction in water content in a linear manner. Majority of soil shrinkage

occurs during PS. After PS, the Residual Shrinkage (RS) sets in with marginal
changes in void ratio due to rearrangement of particles on reduction of water
content. The degree of saturation reduces in all the three stages of drying
implying that total volume decreases at a slower rate than volume of water
during the entire drying process.
The magnitude of Initial Shrinkage for the BC soil specimens is calculated
as:
(e0 − eIS )
IS (%) = × 100 (3)
(1 + e0 )
where e0 and eIS are void ratio of specimen in the swollen state and at end of
IS, respectively. The Primary shrinkage is defined as:
(eIS − ePS )
PS (%) = × 100 (4)
(1 + eIS )
where ePS is void ratio of the specimen at end of PS region.

The Residual shrinkage is defined as:
(ePS − eRS )
RS (%) = × 100 (5)
(1 + ePS )
where eRS is void ratio at end of RS and simply corresponds to the final void
ratio of specimen obtained during drying.
The swell and shrinkage stage parameters of the BC soil specimens are
summarized in Table 3. The PS dominates the Total shrinkage (TS) mag-
nitudes. The PS and TS magnitudes of the BC soil specimens vary within
Table 3. Shrinkage magnitudes of BC soil specimens
Compaction condition w0 Sr0 e0 eIS ePS eRS IS PS RS TS

[%] [%] [%] [%] [%] [%]
MDD & OMC 39.8 96.6 1.10 1.04 .48 .44 3.0 27.5 2.9 33.4
95% MDD & OMC 42.8 97.1 1.18 1.10 .51 .46 3.6 28.4 3.3 35.3
MDD & 5% dry of OMC 43.0 98.0 1.18 1.07 .47 .44 5.0 29.0 1.6 35.6
95% MDD & 5% wet of OMC 39.8 97.1 1.10 1.04 .42 .41 2.6 30.5 0.9 34.0
Sr0 Degree of saturation at end of swelling
the narrow limits. Interestingly, the void ratio of swollen specimens varies
marginally irrespective of initial compaction conditions.
The changes in void ratio with reduction in water content upon drying the
swollen red soil specimens are plotted in Fig. 2. The specimens shrink in three
stages same as observed for BC soil.
Same as for BC soil, there was continuous reduction in degree of saturation
during entire drying process. The magnitudes of IS, PS, RS were calculated
for the red soil specimens from Eqs (3)–(5) and are listed in Table 4 along
with swollen state parameters.
Data in Table 4 show that the PS magnitude ranges from 3.8 to 8.2% and
TS magnitudes from 6.9 to 11.8%. The trend of variations in PS and TS mag-
nitudes suggests that besides the swollen void ratio, the shrinkage magnitudes
of the red soil specimens are influenced by some other soil parameter as well.
0.75
® MDD and OMC

0.7 £ MDD and 5% dry of OMC
p 95% MDD and OMC
¡ 95% MDD and 5% wet of OMC
0.65
Void ratio
0.6
0.55
0.5
0.45
0.4
0 5 10 15 20 25
Water content (%)
Fig. 2. Void ratio with water content during shrinkage of compacted red soil spec-
imens after swelling
Table 4. Shrinkage magnitudes of red soil specimens
Compaction condition w0 Sr0 e0 eIS ePS eRS IS PS RS TS

[%] [%] [%] [%] [%] [%]
MDD & OMC 19.5 92.8 .56 .53 .47 .46 1.9 3.8 1.2 6.9
95% MDD & OMC 20.8 87.3 .64 .62 .55 .53 1.3 4.3. 1.3 6.9
MDD & 5% dry of OMC 22.7 85.9 .71 .68 .57 .56 1.6 6.3 0.6 8.5
95% MDD & 5% wet of OMC 23.3 92.3 .68 .64 .51 .48 2.1 8.2 1.5 11.8
Prediction of Total Shrinkage Magnitudes
Examination of the shrinkage behaviour of various soil specimens had revealed

that the void ratio and degree of saturation of swollen specimens play impor-
tant role in determining TS magnitudes. Soil specimen with larger e0 and
Sr0 experienced a larger magnitude of TS on drying. Such behaviour may be
anticipated as soils with high water content and void ratios experience large
magnitudes of shrinkage (Yong and Warkentin 1975, Chen 1988). Also the
TS exhibited by the soil specimens have a bearing with their Atterberg limits
and hence depend on the clay mineral present in soil. The BC soil specimens
with high Atterberg limits (wL = 84%, Ip = 61%) exhibit high shrinkage
magnitudes (33.4 to 35.6%).The red soil specimens with low Atterberg limits
(wL = 39%, Ip = 20%) exhibit shrinkage magnitudes ranging from 6.9 to
11.8%.
Total shrinkage magnitudes are analyzed as function of both swollen poros-
ity and degree of saturation. From volume-mass relations for soils, the product
of porosity and degree of saturation represent the volumetric water content
(θ) of the soil. Fig. 3 correlates the TS magnitudes with θ after swelling of
the soil specimens. A correlation coefficient of 0.997 is exhibited by the plot
in Fig. 3. The trend of result in Fig. 3 suggests that the TS magnitude of a
specimen from swollen to air-dried state may be predicted by knowing the θ
of the specimen after swelling.
The relation between TS and swollen θ of the soil specimens follow the
linear relation:
TS = 154.6 × θ − 45.6 . (6)
Use of Eq. (6) allows the prediction of total shrinkage of soils from the
swollen condition by air-drying. This equation is expected to be valid for soils
having liquid limits ranging from 40 to 90% and plasticity index ranging from
20 to 60%. These ranges of Atterberg limits are typical for clayey soils in arid
regions that are susceptible to swelling and shrinkage upon moisture content
fluctuations.
40
35
y = 154.61x - 45.56
Total Shrinkage (%) 30 R2 = 0.997
25
20
15
10
0
0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
Volumetric water content
Fig. 3. Relationship between Total shrinkage magnitudes with volumetric water

content of swollen specimens
Conclusions
The following major conclusions emerged from the study. Soils shrink in three
distinct stages, namely, Initial Shrinkage, Primary Shrinkage and Residual
Shrinkage. In the first and third stage, there was little reduction in void ra-
tio with reduction in water content compared to the second stage where the
void ratio decreased rapidly. The degree of saturation reduced in all the three
stages of drying. The swollen void ratio and degree of saturation of speci-
mens play an important role in controlling the total shrinkage magnitudes.
Total shrinkage magnitudes can be predicted using swollen volumetric water
contents by equation TS = 154.6 × θ − 45.6.
References
Chen FH (1988) Foundation on expansive soils. Elsevier, New York
IS: 2720-Part 4 (1985) Indian standard methods of test for soils: Grain size analysis.
BIS, New Delhi
IS: 2720-Part 5 (1985) Indian standard methods of test for soils: Determination of
liquid limit and plastic limit. BIS, New Delhi
shrinkage factors. BIS, New Delhi
water content-dry unit weight relation using light compaction. BIS, New Delhi
Revanasiddappa R (2000) Collapse behaviour of red soils of Bangalore district. Ph.D.
thesis, Indian Institute of Science
Sarita (2003) An experimental study of shrinkage behaviour and SWCC relations of
clays. MSc (Engg) thesis, Indian institute of Science
Shivananda P (2003) Role of lime-soil reactions and environmental factors in the

engineering behaviour of lime-amended expansive clays. Ph.D. thesis, Indian
Institute of Science
Yong RN, Warkentin BP (1975) Soil properties and behavior. Elsevier, New York
Assessment of Swelling Deformation
of Unsaturated Kaolinite Clay
Markus Dobrowolsky and Christos Vrettos
Division of Soil Mechanics and Foundation Engineering,

Technical University of Kaiserslautern, Kaiserslautern, Germany
dobrow@rhrk.uni-kl.de, vrettos@rhrk.uni-kl.de
Summary. The swelling behaviour of kaolinite clay is systematically investigated

by means of two types of oedometer devices suitably modified to measure the varia-
tion of swelling pressure in dependence of the vertical and volumetric strain. A wide
range of values for initial void ratio as well as initial and final degree of saturation
has been covered. Based on the experimental results approximate equations for the
swelling deformation are derived that are implemented in a FEM-Code. The appli-
cation to a typical situation in practice is demonstrated by the numerical solution
of the respective boundary value problem.
Key words: unsaturated soils, expansive soils, swelling, oedometer tests, numerical
modelling
1 Introduction
Swelling of unsaturated cohesive soils may lead to high pressures and defor-
mations of geotechnical structures. So far, publications on swelling of geoma-
terials refer to rocks (Wittke 2003) or to mudstone (Von Wolffersdorff and
Heidkamp 2005, Heidkamp and Katz 2004).
The aim of the investigation presented herein is the measurement of the
volume and pressure changes of swelling cohesive soils in dependence of the
moisture content and the void ratio. The influence of the constraints on
geotechnical structures due to swelling caused by the non-stationary mois-
ture diffusion in the ground is then analysed by the Finite-Element-Method
using an appropriate soil model.
In the following, the constitutive relation proposed for modelling the vol-
ume change is formulated, the developed oedometer apparatus are described
and representative test results are given. Numerical simulations of element
tests and of a boundary value problem are also presented.
286 Markus Dobrowolsky and Christos Vrettos
2 Constitutive Relation for Volume Change

The volume change behaviour due to swelling is described here in terms of
the first stress and strain invariants by the following relationship:
IεSW = Iσ,max
SW
(e0 , Sr,0 , Sr,∞ ) · f (Iσ ), Iσ,0 ≤ Iσ ≤ Iσ,max
SW
(1)
IεSW = 0, Iσ > SW
Iσ,max (2)
SW
where Iσ,max is the maximum swelling pressure under constant volume condi-
SW
tions, Iε is the swelling strain, Iσ is the actual stress, e0 is the initial void
ratio, Sr,0 is the degree of saturation at the begin of the swelling phase, and
SW
Sr,∞ the accordingly value at the stress level Iσ,max .
3 Test Material, Equipment, and Procedure

The material chosen for the tests is kaolinite clay of medium to high plasticity
and is described in detail by Dobrowolsky and Becker (2002) and Dobrowolsky
and Vrettos (2005). The swelling potential of this material is classified as low
to medium. The test material has been investigated in the past and allows
also the accurate measurement of the soil water suction at relatively low levels
using standard techniques.
Two new modified oedometer devices have been developed to determine
the parameters in the constitutive equation (1).
With the first oedometer, called oedometer A, it is possible to determine
separately the axial and radial pressure. This apparatus as well as test results
are described elsewhere (Dobrowolsky and Vrettos 2005).
The second oedometer device, called oedometer B, is designed for 100 mm
diameter specimens with a thickness of 50 mm. A rubber membrane is mounted
on the inner side of the oedometer allowing the control and the measurement
of radial pressure and deformation. The space between membrane and ring is
filled with glycerin. The axial pressure is recorded by a load cell. Thus, this
apparatus works similar to conventional triaxial equipment.
Results are reported here only for tests carried out in oedometer B. The
preparation technique of the soil sample and its optimization is described in
detail by Dobrowolsky (2003). Tests have been carried out for selected values
of the initial and the final degree of saturation, and of the initial void ratio.
The correlation between IεSW and IσSW is investigated. Through watering
SW
from Sr,0 to Sr,∞ under constant volume the maximum swell stress Iσ,max
SW
is imposed to the sample. An isotropic deformation Iε is then stepwise
allowed through control of vertical and radial strain. During this process the
actual swelling pressure IσSW is measured. This leads to swell pressure-strain
relationships as exemplarily shown in Figs 1 and 2.
Additional tests have been carried out in different devices to deter-
mine the characteristics of the water transport through the soil sample.
Assessment of Swelling Deformation of Unsaturated Kaolinite Clay 287
The soil–water suction ψ is determined by the pF value, where pF =

log[ψ/(1 cm water column)]. Up to a value of pF = 3 measurement is per-
formed by using a miniature tensiometer placed in the middle of the specimen
in the oedometer. The hydraulic relations for values pF > 3 are determined by
means of i) the axis – translation method in a pressure plate cell (3 ≤ pF ≤ 4)
and ii) the vacuum desiccator method (pF > 4). In the pressure plate cell tests
are performed by dewatering and subsequent watering allowing the identifi-
cation of hysteretic behaviour in the soil–water–characteristic–curve SWCC.
The vacuum desiccator method allows only dewatering and is used here only
for the sake of completeness since pF > 4 lies outside the range of the degree
of saturation investigated (Sr,0 ≥ 0.7). In addition to the SWCC the unsatu-
rated hydraulic conductivity Ku is needed to determine the water transport
characteristics of this particular soil type. The variation of Ku with degree of
saturation, void ratio, and temperature is taken from Wendling (2004).
4 Test Results
It can be shown that the swell pressure at a constant volume is of isotropic
nature, cf. Dobrowolsky and Vrettos (2005). The results presented in that
SW
paper lead to the following approximation for Iσ,max :
SW
Iσ,max = 3.29 · pa · exp(3.65 · Sr,∞ − 4.7 · e0 ) · (Sr,∞ − Sr,0 )(0.95·Sr,∞ −0.54) (3)
where pa is the atmospheric pressure. The above equation is valid for 0.68 ≤
e0 ≤ 0.88; 0.6 ≤ Sr,0 ≤ 1.0 and 0.78 ≤ Sr,∞ ≤ 1.0.
The experimentally determined variation of swelling strain with normal-
ized swelling pressure is approximated by equation (4) that is depicted in
Fig. 1 for the range of values investigated:

a Iσ
SW
Iε = b log SW
, Iσ,0 ≤ Iσ ≤ Iσ,max
SW
(4)
Iσ Iσ,max
SW
Iσ,max
SW
whereby a and b are functions of e0 , and of Iσ,max normalized to the imposed
isotropic initial pressure Iσ,0 = 27 kPa:
SW
Iσ,max
a = (−0.021 · e0 + 0.01) , (5)
Iσ,0
SW
Iσ,max
b = −0.096 + (0.257 · e0 − 0.131) . (6)
Iσ,0
A significant increase of the swell strain IεSW occurs with decreasing stress
Iσ . It can be seen that the increase of swell strain is stronger for lower values
of void ratio. Results from other tests not reported herein show that the
swelling potential of the soil sample increases with decreasing initial degree of
saturation.
5 Numerical Simulations
General Considerations
We assume that material used in the study is homogeneous and isotropic. A

nonlinear elastic – ideal plastic stress–strain–relationship is assumed for the
mechanical loading. Swell deformations are superimposed after completion
of initial loading. Investigations presented by the authors Dobrowolsky and
Vrettos (2005) show that the swell–shrinkage behaviour of the test material
at a specific swell pressure level is approximately linear elastic.
In the following, the plastic deformations are left-out for the sake of sim-
plicity. Thus, the total strain is:
dε = dεel + dεSW . (7)
The elastic relationship between the stress increment and the strain incre-
ment is given by

dσ el = Del,m · dεel + Del,SW · dεSW (8)
where [Del,m ] is the stiffness matrix for the mechanical elastic behaviour and
[Del,SW ] is the matrix for the elastic swelling behaviour.
The FEM-code ABAQUS Version 6–4 is used for the analysis. The fol-
lowing general relationship between maximum volumetric swelling strain and
degree of saturation is already built in:
1 SW
ΔεSW
jj,max = εv (S) − εSW I
v (S ) , j = 1, 2, 3 (9)
3
Fig. 1. Swelling strain IεSW vs. normalized swelling stress Iσ /Iσ,max

SW
for various
◦
values of void ratio e0 and Sr,0 = 0.7, Sr,∞ = 1.0, T = 20 C
whereby S and S I are the actual and initial degree of saturation, respectively.
The above relation describes unhindered swelling. During our tests the strain
is controlled and stress is measured. Thus, by setting Iσ = Iσ,0 and for small
values of Iσ,0 equations (3)–(4) approximately correspond to the relationship
given by equation (9). The value Iσ,0 = 27 kPa used in the tests is the lowest
practicable stress level.
The stiffness matrix [Del,m ] is obtained in terms of the modulus of elas-
ticity E and the Poisson’s ratio ν. For the modulus of elasticity we selected
a stress dependent relationship derived from the equation for the small strain
shear modulus used in dynamic analysis (Hardin 1978):
0.5
Iσ 1
E = C · f (e) · · pa · . (10)
3 · pa 2 · (1 + ν)
For Poisson’s ratio we set ν = 0.3. For small strains the constant C = 625.
A value C = 156 yields a good fitting of the test results and corresponds to
a reasonable approximation of the secant modulus at the strain levels under
consideration.
The swell-stress dependent swell modulus Q (bulk modulus) to be used in
[Del,SW ] is obtained from the test results, as exemplarily given in Fig. 1.
Simulation of Laboratory Tests
First, the appropriateness of the relationships given above is checked by sim-

ulation of representative laboratory tests carried out in oedometer A and B.
The simulation starts with initial conditions given by Sr,0 = 0.7 and
e0 = 0.78. At first, an equilibrium step is computed during which suction
as well as self-weight are imposed. In the next step the model is watered from
the bottom at constant volume to Sr,∞ = 1.0. After that, a prescribed defor-
mation is imposed stepwise: axial deformation in odoemeter A and volumetric
deformation in oedometer B, respectively. Figure 2 compares the results of the
tests and of the numerical simulation. The variation of the swell modulus Q
(bulk modulus) yields a good fitting of the curve.
Simulation of a Footing on Expansive Soil
The application to a typical situation in engineering practice is demonstrated

by the numerical solution of the boundary value problem of an axisymmetric
footing on an expansive soil. The footing with radius of 1.0 m and thickness
of 0.4 m is placed on the surface of a homogeneous soil with initial conditions
e0 = 0.68 and Sr,0 = 0.7. The size of the model is 5 m in half-width by 4 m
in depth. Underneath the base of the model a porous water-bearing layer is
assumed. The footing is loaded by force of 200 kPa after the equilibrium step
is completed and the self-weight of footing and soil is applied. In the next
step the groundwater rises within in the water-bearing stratum up to the
Fig. 2. Volumetric swell strains IεSW vs. actual stress Iσ normalized to Iσ,max
SW
for
e0 = 0.78, Sr,0 = 0.7, Sr,∞ = 1.0, T = 20◦ C. The solid lines describe equation (4),
the data points are for the numerical simulation
bottom of the expansive soil. This is simulated by setting suction equal to

0 at the interface between soil and water-bearing stratum. The suction-free
front propagates upwards through the soil and leads to its saturation and
consequently to swelling. The time the water from the water-bearing stratum
is allowed to penetrate the soil is assumed to last 150 days.
Figure 3 displays two snapshots of the vertical heave before and after
saturation, respectively. The vertical heave of the foundation amounts to 17 cm
while the vertical heave of the ground at the side of the mesh is 30 cm. The
differential heave of 13 cm due to swelling is noteworthy.
Fig. 3. Vertical heave before and after watering from the bottom
6 Conclusions
A wide range of tests have been performed to determine the water transport
characteristics in the soil and the associated swell pressure vs. swell strain
relationship of the cohesive material. The numerical simulations show the
capability of the proposed analytical expressions to describing the swell be-
haviour.
Acknowledgements
The investigation presented herein has been carried out within the frame of a
project funded by the German Science Foundation dealing with mechanics of
unsaturated soils. This support is greatly appreciated.
References
Dobrowolsky M, Becker A (2002) Adsorptions- und Desorptionsverhalten durch-
strömter bindiger Böden. 4. Workshop Teilgesättigte Böden, Schriftenreihe
Geotechnik, Bauhaus-Universität Weimar, Heft 8:71–76
Dobrowolsky M (2003) Volume change behavior of cohesive soils. Proc Int Conf
From Experimental Evidence towards Numerical Modeling of Unsaturated Soils,
Weimar I:227–239
Dobrowolsky M, Vrettos C (2005) Swell and shrink tests in modified oedometer
apparatuses, Advanced Experimental Unsaturated Soil Mechanics, Trento, Italy,
pp. 163–168
Von Wolffersdorff P-A, Heidkamp H (2005) Quellverhalten diagenetisch verfes-
tigter Tonsteine in den Einschnitten der Neubaustrecke Nürnberg – Ingolstadt
der Deutschen Bahn AG, Universität Kassel, Festschrift zum 60. Geburtstag
Prof. Kempfert, Schriftenreihe Geotechnik, Heft 18:35–52
Hardin BO (1978) The nature of stress-strain behavior of soils. Proc ASCE Spec
Conf on Earthq Eng Soil Dyn, Pasadena I:3–90
Heidkamp H, Katz C (2004) The swelling phenomenon of soils – Proposal of an
efficient continuum modelling approach. Proc ISRM Regional Symposium EU-
ROCK & 53rd Geomechanics Colloquium, Verlag Glückauf, pp. 743–748
Hibbit, Karlson & Sorensen, Inc. (2002) ABAQUS/Standard User’s Manual, Version
6–4, Pawtucket, RI, USA
Wittke M (2003) Begrenzung der Quelldrücke durch Selbstabdichtung beim Tun-
nelbau im anhydritführenden Gebirge, Dissertation, Fakultät für Bauingenieur-
wesen, TH Aachen
Wendling S (2004) Untersuchungen zur Entstehung von Austrockungsrissen in
mineralischen Deponieabdichtungen, Dissertation, Heft 10, Fachgebiet Boden-
mechanik und Grundbau, TU Kaiserslautern
Suction and Collapse of Lumpy Spoilheaps
in Northwestern Bohemia
Vladislava Herbstová1 , Jan Boháč1 , and Ivo Herle2

1
Department of Engineering Geology, Charles University, Albertov 6,
12843 Prague, Czech Republic herbsto1@natur.cuni.cz, bohac@natur.cuni.cz
2
Institut for Geotechnik, Technical University Dresden,
D-01062 Dresden, Germany ivo.herle@mailbox.tu-dresden.de
Summary. Large spoilheaps of overburden from open-cast coal mining in North-

western Bohemia offer a space for development. However, they exhibit unfavourable
features such as collapse potential caused by the lumpy structure with high over-
all void ratios. Suction oscillations contribute to the structure degradation by the
intervoids closure. The paper presents in-situ suction monitoring and laboratory
investigation of the collapse potential of spoilheaps.
Key words: lumpy structure, intervoids closure, suction, collapse potential
1 Introduction
An intensive open-cast coal mining in Northwestern Bohemia produces a large
amount of spoil. The brown coal deposit is typically in the depths of 80–200 m.
The overburden – mainly clays to stiff fissured claystones – are disintegrated
through mining into lumps or blocks ranging in size from centimetres up to
tenths of centimetres. The blocks are subsequently end-dumped without any
compaction to the bodies of spoilheaps, which reach up to 100 m in height but
typically a few tens of metres. The extent of clayfills in Northwestern Bohemia
reaches about 100 km2 (see Fig. 1). The spoilheaps are gradually reclaimed
and nowadays when there is a lack of building sites the areas of landfills offer
reasonable space for development.
2 Structure of the Spoilheap Soil

The freshly filled spoilheap material has the form of a lumpy soil. The total
porosity of such material consists of the primary voids of the clayey lumps sc.
intragranular porosity (reaching typically about n = 40% reported by Feda
(1998)) and the voids between the lumps sc. intergranular porosity. The total
porosity of such material, which reaches up to 70%, is given by
294 V. Herbstová et al.
Fig. 1. Extent of non-engineered spoilheaps within the Czech Republic
ntot = ni (1 − ne ) + ne (1)
where ne and ni are intergranular and intragranular porosities. The state of

freshly filled spoilheap is therefore granular (see Fig. 2 left) which is confirmed
by its high angle of repose (typically of about 40◦ ). The clayey blocks deposited
to the spoilheaps are overconsolidated not only by the current depths but also
by the geological history – considerable denudation. Hurník (1978) refers that
the thickness of denudated layers in the Miocene could reach 70 to 300 m.
The excavated overburden is filled in a dry way and therefore the intervoids
are typically air-dry while the intravoids of lumps exhibit almost full satura-
tion by its natural water content of about w = 34%. Due to the granular
nature of freshly filled material there is no suction within the intervoids and
some suction within the intravoids in the lumps, where the negative pore wa-
ter pressures developed first due to water-level sinking and subsequently due
to unloading by exploitation. Further suction within the lumps is affected by
the weather dependent moisture changes. The intervoids form the preferential
paths for surface water. So the freshly filled spoilheaps are highly permeable
until the structured soil degrades.
Fig. 2. Spoilheap just after filling (left) and about 10 years after filling (right)
Suction and Collapse of Lumpy Spoilheaps in Northwestern Bohemia 295
The initially lumpy structure of the soil degrades with the increasing load.
The intervoids close up due to plastic straining of relatively weak lumps.
In greater depths the interporosity vanishes. In lower depth wetting-drying
suction cycles are the main agent, which causes the structure degradation.
Through the structure degradation clayfill material changes back into more
or less homogeneous soil. The overall hydraulic conductivity decreases with
increasing degree of structure degradation. Figure 2 right shows the aged
spoilheap material of about 10 years after filling, where some lumps are already
degraded into finer mass.
3 Field Suction Measurement
The classic type of jet-fill tensiometers is used for monitoring in-situ suction
changes at two spoilheaps near Ústí nad Labem (at the site of a highway
construction) and at the site of mine Bílina (see Fig. 1). The lengths of ten-
siometers used are 0.5–2 metres.
Very low or zero suctions and even positive values of pore water pres-
sure have been measured at the spoilheap near Ústí nad Labem (Fig. 3),
which indicates the fully saturated conditions. The zero suctions correspond
to practically no collapse potential measured in the laboratory for the partic-
ular landfill (see Sect. 4.1).
At the spoilheap of the mine Bílina the tensiometers are installed in a
moderate slope. The measured suctions (Fig. 4) reached maximum values of
about 90 kPa in late autumn. It is however the limit of negative pore pressures
that can be measured by this type of gauges due to cavitation, e.g. Fredlund
and Rahardjo (1993). In late winter and spring the suction dropped to the
values close to zero even in the deeper gauges. The seasonally dependent cyclic
suction changes cause the disintegration of the clay lumps, which leads to the
intervoids closure.
Fig. 3. Suction field measurements at the site near Ústí nad Labem
Fig. 4. Suction field measurements at the site of Bílina mine
4 Collapse Potential
The high overall porosity of the freshly filled material causes an open struc-
ture which is potentially susceptible to collapse. Charles and Watts (1996)
gave three criteria for fills to qualify as particularly vulnerable to collapse
compression: the fill
• was not subject to a heavy compaction,
• was placed relatively dry and has never been fully saturated or above a
critical degree of saturation,
• is under relatively low stress.
At least in the early stages after landfilling the present clayey spoilheaps fulfill
all three criteria. Two sets of tests on undisturbed and lumpy samples have
been carried out to investigate the collapse vulnerability of spoilheaps.
4.1 Laboratory Tests on Undisturbed Samples
The collapse potential was investigated at the site of highway construction

near the town Ústí nad Labem (see Fig. 1). The highway is crossing the
reclaimed spoilheaps in overall length of about 12 km. The particular spoilheap
is about 40 years old and in a high degree of homogenization and interporosity
decrease but the collapse potential could still be expected. The soils were clays
of high plasticity, average wL = 59% (from 56 to 62%), wP = 28% (27–31%),
activity IA = 0.9 (0.77–1.03).
Double oedometer tests were carried out according to Abelev (1959), Cui et
al. (1998), Jennings and Knight (1957). Two “identical” specimens are placed
into the oedometer and during the test one specimen is maintained at the
original water content (referred to as “unsaturated” specimen, marked with
unsat in Fig. 5), the second one is inundated at low total vertical pressure
(“saturated” specimen, marked with sat in Fig. 5). Some curves needed to be
transposed (marked with trans in Fig. 5) to the same initial void ratios as
1.1
0.9
0.8
e (-)
0.7
J801 11.7-12m-trans
0.6 J801 11.7-12m-sat
J801 11.7-12m-unsat
0.5 J801 5.7-6,0m-trans
J801 5.7-6.0m-sat
J801 5.7-6.0m-unsat
0.4
10 100 1000
σax (kPa)
Fig. 5. Double oedometer tests on undisturbed samples
suggested by Jennings and Knight (1957). The divergence of the two stress-
void ratio curves indicates the magnitude of collapse due to wetting.
In total 44 pairs of oedometer specimens were evaluated in classic type
of oedometer rings (100 mm diameter, 30 mm hight). In total 24 specimens
exhibited some collapse, and only two of them were classified as moderately
collapsible according to ASTM (2003a). Twenty specimens exhibited a neg-
ative value of collapse index which indicated swelling. The standard double
oedometer tests did not confirmed any collapse potential.
Matric suction was determined using filter paper method according to
ASTM (2003b). Values mainly did not exceed suctions over 100 kPa, and were
possibly caused by unloading through sampling. For more detail see Herbstová
and Boháč (2005).
There was a scatter in the initial void ratios of undisturbed samples, which
could refer to some remaining nonhomogeneity of the aged spoilheap. There-
fore, more detailed field testing should be carried out to study collapse vulner-
ability in an appropriate scale to capture the nonhomogeneity of the clayfill
material.
4.2 Laboratory Tests on Lumpy Samples
The second set of tests was carried out on the artificially prepared compacted
and loose samples of lumpy clayfill material. The material used for testing was
the spoilheap soil from the Bílina mine (see Fig. 1) sampled about 4–5 years
after filling. The drying-wetting weathering processes already caused some
disintegration of bigger lumps into finer mass. In order to carry out the tests
in standard types of oedometer rings (d = 50–100 mm, h = 20–30 mm) the
spoilheap material with reduced grain size was used (Fig. 6 left). The soil was
sieved through a 5 mm sieve which could cause a difference between the initial
overall void ratio of the natural and sieved material but the main feature of
the lumpy structure was maintained.
Fig. 6. The lumpy material before the test (left) and after the test (right)
Figure 7 shows the results. The samples inundated at σax = 10 kPa are
marked with s. Two sets of the tests with samples of natural water content
(w = 34%) were carried out on loose and compacted (marked with c in Fig. 7)
soil. The collapsibility index was determined using
i = (eunsat − esat )/(1 + eunsat ) (2)
where e are void ratios natural moisture and inundated samples at each load-
ing step. The collapse potential was estimated with its maximum of i = 20.2%
at σax = 100 kPa for loose samples and i = 15.9% at σax = 200 kPa for ini-
tially compacted samples, which indicates severe degree of collapse potential
according to ASTM (2003a).
Collapse potential vanishes at vertical confining pressures of about σax =
1000 kPa, where saturated and natural moisture compression lines unite.
Above this pressure the slope of NCL of natural moisture samples changes and
so the character of the soil. In the region of σax ≥ 1000 kPa the interporosity
vanishes and the compressibility is controlled just by the intragranular voids.
2.8 056_32
2.6 056_33
056_34
2.4 056_35
2.2 056_06 c
2 056_18 c
056_28 s
e (-)
1.8 056_29 s
1.6 056_30 s
1.4
1.2
1
0.8
0.6
10 100 1000 10000
σax (kPa)
Fig. 7. Double oedometer tests on structured compacted and loose samples

2.4
056_06 c
2.2 056_18 c
056_28 s
2 056_29 s
1.8 056_30 s
P01 s200
1.6 P03 s200
e (-)
1.4
1.2
1
0.8
0.6
10 100 1000 10000
σax (kPa)
Fig. 8. Double oedometer tests on structured initially compacted samples and col-
lapse under wetting at 200 kPa
After the sample dismantling at the end of test no obvious structure was vis-
ible. The material was homogenized in both natural moisture and saturated
states (Fig. 6 right).
To confirm the collapse potential indicated by double oedometer tests the
standard method for estimating collapse potential according to ASTM (2003a)
was used. The samples prepared by compaction with natural water content
were inundated at σax = 200 kPa (tests no. P01 and P03 in Fig. 8). The
collapse potential estimated through the standard method fits well to the
double oedometer curves, which confirms the severe collapse vulnerability of
initially lumpy structured samples.
Severe collapse potential was indicated by the tests on model lumpy soil
contrary to negligible collapse potential estimated on undisturbed samples.
One possible explanation would be that the site, where undisturbed samples
were taken, is in a high degree of homogenization so the intervoids are already
closed up and the lumpy structure vanished in the whole profile due to suc-
tion cycles, overburden and possible collapse under wetting. There was also
no suction present (see Sect. 3). On the other hand, the significant scatter
in initial void ratios of undisturbed samples indicates some remaining inho-
mogeneity of the aged material. Field testing should be carried out to fully
exclude collapse risk.
5 Conclusions
Spoilheap material is highly collapse susceptible due to its open structure

mainly in the early stages after landfilling. There are two main agents causing
the degradation of the original open structure: overburden and drop of suction.
In greater depths the overburden causes plastic straining and decrease of the
total porosity of the material. However even under under low surcharge pres-
sures the lumpy material changes back into more homogeneous soil, and the
suction oscillations due to weather are considered responsible for it. The drop
of suction within intravoids causes the decrease of the strength of the lumps.
On wetting, the intravoid high suctions vanish, the lumps become softer and
the structure collapses.
At the site of Bílina mine the seasonally dependent suction oscillations
were recorded through standard type of tensiometers reaching in late autumn
its maximum of about 90 kPa. On the other hand at the site of highway
construction near Ústí nad Labem there has been no or negligible suction
through the whole monitored period. It corresponds well with the fact that
there was no collapse potential indicated by laboratory investigation at this
site.
Acknowledgements
The first author is grateful for the financial support by the research grants
GAUK 6/2006/R and GAUK 331/B-GEO/PřF. A part of the laboratory in-
vestigation was carried out at the Technical University, Dresden, during the
stay of the first author supported by the grant no. D-CZ 21/05-06 PPP of
DAAD-AV ČR.
References
Abelev YM (1959) Collapse of macro-porosity loamy (loess) soil, Research Inst NI-
IOPS, Moscow, Proc No 37:5–25
ASTM (2003a) Standard Test Method for Measurement of Collapse Potential of
Soils. ASTM D5333–03
ASTM (2003b) Standard Test Method for Measurement of Soil Potential (Suction)
Using Filter Paper. ASTM D5298–03
Charles JA, Watts KS (1996) The assesment of the collapse potential of fills and its
significance for building on fill, Proc Instn Civ Engng Geotech Engng 119:(1)15–
28
Cui YJ, Delage P, Alzoghbi P (1998) Microstructure and collapse behaviour of a loess
from Northern France. Proc 2nd Int Conf Unsat’98. Internat Acad Publishers,
Beijing Vol 1:31–36
Feda J (1998) Fragmentary clay – a difficult waste material, Eng Geol 51:77–88
Fredlund DG, Rahardjo H (1993) Soil Mechanics for Unsaturated Soils. John Wiley
and Sons Inc., New York
Herbstová V, Boháč J (2005) Laboratory investigation of the collapsibility of a
clayey spoil heap soil. In: Proc GEOPROB 2005, Int Conf on Problematic Soils.
Famagusta, Cyprus, Vol 2:715–722
Hurník S (1978) Reconstruction of the Thickness of Overlying Complex in North
Bohemian Brown-Coal Basin (Miocene), J Geol Min 23(3):265–276 (in Czech)
Jennings JE, Knight K (1957) The additional settlement of foundations due to a
collapse of structure of sandy subsoils on wetting. Proc 4th ICSMFE 316–319
Oedometer Creep Tests of a Partially Saturated
Kaolinite Clay
Piotr Kierzkowski
Division of Soil Mechanics and Foundation Engineering, Technical University of

Kaiserslautern, Kaiserslautern, Germany pkierz@rhrk.uni-kl.de
Summary. In the following paper an influence of several state parameters on one

dimensional time-dependent material behaviour of an unsaturated soil is presented.
For identification of parameter values, laboratory experiments were performed. Stan-
dard oedometric compression tests were carried out to investigate stress-strain-time
relationships. Apart from the degree of saturation the initial void ratio vary as pa-
rameters. For measuring the matrix suction a tensiometer is mounted inside the
specimen. A detailed description of the equipment used and the test procedures are
given in the paper. The tests reported here were carried out with a kaolinite to
determine parameter values, depending on different state variables. The characteris-
tic behaviour of unsaturated samples observed during consolidation and creeping is
shown in a series of figures. Parameters like Cc , Cs and Cα are used in constitutive
model that can simulate both consolidation processes as well as creep phenomenon
for partially saturated clay in a single consistent analysis.
Key words: kaolinite, oedometer, creep, partially saturated clay, suction
1 Introduction
Cohesive soils belong to geomaterials with a distinct viscous behaviour. They

are often used as mineral seal in landfills. Viscosity is connected with physic-
chemical properties of adhesive water and with the contact properties between
soil particles. Time-dependent behaviour is plays an important role during
creep of slopes or flowing of banks often leading to instabilities. The creep
settlement has been a research topic over the last four decades (Bjerrum 1967,
Berre and Iversen 1972, Leroueil et al. 1985, Yin 1999, Niemunis and Krieg
1996). In engineering practise, two hypotheses have been adopted for the es-
timation of the creep settlement. One assumes that secondary consolidation
occurs only after the primary consolidation is completed, while another one
assumes that creep occurs during the entire consolidation process. In partially
saturated soils suction dominates the behaviour and there is lack of experi-
mental evidence for this type of soil. A series of oedometer tests with direct
302 Piotr Kierzkowski
measurement of suction have been carried out to assess the characteristics

of secondary consolidation. The equipment developed for this purpose, the
testing procedure and selected results are described next.
2 Test Material
The material used for the tests is a kaolinite clay which is described in detail
by Dobrowolsky and Becker (2002).
The clay mineral fractions as well as the relevant soil properties are listed
in Table 1. Looking from the mineralogy site view clay minerals belong to the
layer silicates. The difference between them and normal silicates is that they
have a higher dispersion, sorption- and ion exchangeability. They consist of
tetrahedrons and octahedrons that are arranged in layers, corresponding to a
so-called sheetstructure.
Table 1. Clay fractions and soils properties
Kaolinite > 40%

Clay minerals Illite < 20%
Smectite < 5%
Liquid limit wL = 51.3%
Plastic limit wP = 22.4%
3 Test Equipment and Test Procedure
The creep behaviour of kaolinite was investigated during the secondary com-
pression settlement in an oedometer device, cf. Fig. 1. The cylindrical speci-
men has a diameter of 100 mm and a height of 19 mm. It is located in 2 mm
thick metal ring. The metal ring is brushed in the inner face with silicon fat
to reduce friction. Two porous stones are placed on the top and bottom of the
specimen.
Variation of suction is measured using UMS-tensiometer type T5-10, cf.
Fig. 2. Measurement range is between −85 kPa and 100 kPa. The applica-
tion of tensiometer measurement technique requires that the values of initial
void ratio e0 and degree of saturation Sr do not yield suction values lower
that −85 kPa. A different technique like the axis translation method is here
not applicable. To enforce this maximum level of suction the initial degree of
saturation was kept higher than 0.90 and the initial void ratio higher than
0.70. The tensiometer is installed through the bottom of the specimen in ver-
tical position, cf. Fig. 4.
Oedometer Creep Tests of a Partially Saturated Kaolinite Clay 303
Fig. 1. Oedometer Fig. 2. Tensiometer
stamp
porous stone
specimen oedometer ring
water outlet
tensiometer
Fig. 3. Oedometer with plexiglas cover Fig. 4. Oedometer with position of

tensiometer
Loading and unloading cycles as in conventional tests have been carried

out. Each test consisted of a total of 14 loading/unloading stages with the fol-
lowing sequence: 12 → 25 → 50 → 100 → 200 → 100 → 50 → 100 → 200 →
400 → 200 → 100 → 50 → 25 kPa. The duration of the creep stage depends
on the suction development and took in average three to four days. During
the test the settlement and suction changing was measured. The axial settle-
ments were recorded using the electronic displacement transducer. Accuracy
of measurement is 0.01 mm.
Both temperature and air moisture during the test were kept constant.
Both were measured and controlled with temperature- and moisture sensors.
To achieve constant air conditions the whole equipment was placed in special
plexiglass box, cf. Fig. 3. The air moisture during the test was about 90% in
order to prevent drying or swelling of the soil sample that is accompanied by
an unintentional suction change. Temperature was kept at 22◦ C.
4 Test Results
A typical test result in terms of variation of vertical strains with time is shown
in Fig. 5.
After completion of primary consolidation, secondary creep settlement is
observed at all stress levels. The beginning of creep is defined as the time
when the suction reaches a stable value, cf. Fig. 6. The suction responds
immediately after loading. It increases fast and then gradually reduces in
value. The variation of suction with time for the test is displayed in Fig. 7. It
can be seen that after a certain amount of time (approx. 1800 s) it reaches an
almost constant value.
In order to define appropriate values for the stress dependent deformation
moduli effective stresses σ are defined following Bishop (1959)
σ = σ − ua + χ(ua − uw ) (1)
where: ua – pore air pressure

uw – pore water pressure
χ – coefficient depending on degree of saturation
σ – total stress
The difference between pore air pressure and pore water pressure is de-
fined as suction. The effective stress in ground may thus be higher than the
total stress. Soil parameters like E-modulus or consolidation coefficient Cc are
evaluated with such defined stress.
time [s]
1 10 100 1000 10000 100000 1000000
0,0
-0,5
V = 12.5 kPa
-1,0
H 1 [%]
-1,5
-2,0 V= 25 kPa

axial strain
-2,5
-3,0 V = 50 kPa
-3,5 CB = 0,0024
-4,0
-4,5
Fig. 5. Variation of strains with time for e0 = 0.78, Sr0 = 0.95, T = 22◦ C
time [s]
1 10 100 1000 10000 100000 1000000

-1,0 80
-1,5 60
CB = 0,0024
pore water pressure [kPa]

H 1 [%]
-2,0 40
axial strain
-2,5 20
consolidation creep
-3,0 0
-3,5 LL2: 25 kPa -20
variation of suction
-4,0 -40
Fig. 6. The beginning of creep
10
pore water pressure [kPa]
0 V= 12.5 kPa
-5 V= 25 kPa
-10 V= 50 kPa
-15
1 10 100 1000 10000 100000 1000000
time [s]
Fig. 7. Variation of suction during creep
When the curve is plotted in the graph: void ratio during the one level
against logarithmic time scale, then it passes into the straight line after the
consolidation. The slope of this straight line is described with secondary com-
pression index Cα :
0,80
creep phase 1 (12.5kPa)

0,75
creep phase 2 (25kPa)
creep phase 3 (50kPa)

0,70
void ratio e [-]
0,65 Cc = 0,176
Cs = 0,035
0,60
0,55
0,50
1 10 100 1000
effective stress V´1 [kPa]
Fig. 8. Loading-unloading paths and creep phases
t2
e1 − e2 = Cα log . (2)
t1
This parameter can be replaced by Buisman-factor CB . There is the fol-
lowing relations between them:
Cα = CB (1 + e0 ). (3)
5 Conclusions
The results show that stress-strain behaviour in first loading range can be
described with a straight line (slope Cc ) well accurately. The consolidation
coefficient does not depend on stress level also.
Time-strain curves show a typical S-form of consolidations curve only in
saturation range. The specimen is partially saturated at the first loading levels
and air only is pushed away. Pore water stays inside. Decreasing of void ratio
leads to increasing of degree of saturation. Only at higher stress level, the
water is pushed away and the consolidations curve takes a classic S-form. The
end of consolidation tp was observed after 30 minutes, respectively.
The creep strain under σ = const in all tests increases linear with logarithm
of time. The creep index Cα is constant during the creep phase. Therefore
there is sufficient to carry out the creep test so long, till the straight line
forms (slope Cα ) in time-axial strain graph definitely.
According to the tests, 3–4 days (150 tp ) are enough to determine Cα

definitely. Longer tests duration brings no further information about creep
behaviour. The creep index has the same value for all loading levels. It does
not depend also on the stress.
References
Berre T, Iversen K (1972) Oedometer tests with different specimen heights on a clay
exhibiting large secondary compressionn, Geotechnique 22(1):53–70
Bjerrum L (1967) Engineering geology of Norwegian normally consolidated marine
clays as related to the settlements of buildings, Geotechnique 17(2):83–118
Buisman K (1936) Results of long-duration settlement tests. In: Proceedings 1st
International Conference on Soil Mechanics and Foundation Engineering, Cam-
bridge, Mass. Vol.1, pp. 103-107
Dobrowolsky M, Becker A (2002) Adsorptions- und Desorptionsverhalten durch-
strömter bindiger Böden. 4. Workshop Teilgesättigte Böden, Schriftenreihe
Geotechnik der Bauhaus-Universität Weimar, Heft 8, pp. 71–76
Krieg S (2000) Viskoses Bodenverhalten von Mudden, Seeton und Klei, Heft 150,
Veröffentlichung des Institutes für Bodenmechanik und Felsmechanik der Uni-
versität Fridericiana in Karlsruhe
Leroueil S, Kabbaj M, Tavenas F, Bouchard R (1985) Stress-strain-strain rate rela-
tion fot the compressibility of sensitive natural clays, Geotechnique 35(2):159–
180
Niemunis A, Krieg S (1996) Viscous behaviour of soil under oedometric conditions,
Can. Geotech. J. 33:159–168
Yin J-H (1999) Non-linear creep of soils in oedometer tests, Geotechnique 49(5):669–
707
Analysis of the Expansive Clay Hydration under
Low Hydraulic Gradient
Marcelo Sánchez1 , María Victoria Villar2 , Antonio Lloret3 , and Antonio

Gens3
1
Department of Civil Engineering, Strathclyde University, U.K.
marcelo.sanchez@strath.ac.uk
2
CIEMAT, Madrid, Spain mv.villar@ciemat.es
3
Technical University of Catalonia (UPC), Barcelona, Spain
Antonio.Lloret@upc.edu, antonio.gens@upc.edu
Summary. Engineered barriers made up of compacted expansive clays are a basic

component of the systems for the isolation of high-level radioactive waste (HLW).
The barrier is initially unsaturated and during hydration the hydraulic gradient
varies from high values (i.e. higher than 50000) at the beginning of the wetting to
very low values (practically zero when reaching stationary conditions). The values
of bentonite permeability are frequently obtained at the laboratory under high hy-
draulic gradients (i.e. higher than 15000), necessary to induce a measurable flow. A
drawback of this practice is that the applied hydraulic gradients are very far from
those (expected) in actual conditions. So, the experimental results obtained might
not be realistic in many cases, affecting the reliability of the numerical analysis. To
advance in the knowledge of expansive clays behaviour this work presents an exper-
imental and numerical study. The hydraulic permeability of FEBEX clay samples
has been measured under low injection pressures. This work presents the modelling
of infiltration laboratory tests designed to reproduce the thermo-hydro-mechanical
conditions of large-scale tests. The evolution of the tests is analysed and discussed
in detail using non-standard flow models which consider the presence of a threshold
gradient and thermo-osmotic effects.
Key words: expansive clays, thermo-hydraulic cell tests, numerical modelling, non-
standard flow models
1 Introduction
The long-term behaviour of low permeability (expansive) clays under low hy-
draulic gradient is not well-know, especially when the hydration takes place
under non-isothermal and confined conditions, which are the expected con-
ditions in high-level radioactive waste repositories (HLW). A trend to a very
low rate of hydration, compared to the expected or predicted one, has been
310 M. Sánchez et al.
observed in a number of large scale tests designed to study the behaviour of

clay barriers intended for the isolation of HLW. A virtually locking of the
hydration has been detected in the ‘FEBEX mock-up’, a large scale heating
test that is being carried at CIEMAT laboratories (FEBEX Project 2000).
An important part of the clay barrier is still unsaturated after eight years of
hydration.
Standard thermo-hydro-mechanical (THM) models under-predict the sat-
uration of the clay barrier (Sánchez and Gens 2005). Similar behaviour has
been detected in the ITT test, a large-scale in-situ experiment carried out by
AECL (Atomic Energy of Canada Limited). An unexpected very low satura-
tion was achieved at the end of the test. Modelling evidences that a conven-
tional THM formulation cannot match the experimentally measured values
(Thomas et al. 2003). The trend to a very low hydration rate observed at
advanced stages of these experiments, causes significant differences between
models predictions and measurements. This problem affects strongly the re-
liability of the long-term predictions, especially the one related to the time
required to reach the full saturation of the barrier, which is a key variable in
the design of engineered barriers.
This work presents an experimental and numerical study oriented towards
the identification of THM processes and phenomena that can explain the
unexpected clay behaviour introduced above. The behaviour of unsaturated
FEBEX bentonite in two infiltration tests, performed under thermal gradient
and at isothermal conditions reported in Villar and Gómez-Espina (2006) are
analysed herein. In particular, this work focuses on the study of the effect of
‘non-Darcian behaviour’ and ‘thermo-osmotic flow’ on the evolution of these
tests. In addition, a series of laboratory tests have been carried out to explore
the validity of Darcy’s law at low hydraulic gradients in samples of FEBREX
bentonite. A basic THM formulation (Olivella et al. 1994) has been extended
to consider these two non-standard flow models (i.e., no linearity of Darcy’s
law and thermo-osmotic flows). The infiltration tests are critically analysed
by comparing the experimental data and the model results obtained with the
extended formulation.
2 Basic THM Formulation
The clay behaviour during the infiltration tests is very complex due to the
THM processes and couplings that take place during the simultaneous heat-
ing and hydration. A full description of the THM formulation adopted in
the analysis is presented in Olivella et al. (1994). The formulation incorpo-
rates the main thermal, hydraulic and mechanical phenomena. The problem
is approached using a multi-phase, multi-species formulation that expresses
mathematically the main THM phenomena in terms of: i) balance equations,
ii) constitutive equations and iii) equilibrium restrictions.
Analysis of the Expansive Clay Hydration under Low Hydraulic Gradient 311
The following mass balance equations are considered in the approach: i)

water mass balance; ii) air mass balance and iii) solid mass balance. In ad-
dition the balance of energy and the balance of momentum have been con-
sidered. Two equilibrium restrictions have been considered: the psychometric
law, which establishes the concentration of water vapour in the gas phase;
and Henry’s law, which gives the concentration of air dissolved in the liquid
phase. Finally the constitutive models establish the link between the depen-
dent variables and the state variables. Concerning the hydraulic problem, the
generalized Darcy’s law relates the unsaturated flow with the fluid pressures
through the following expression:
qα = −Kα (∇Pα − ρα g) (1)
where Pl and Pg are liquid (l) and gas (g) phase pressures, respectively, ρα
is the phase density (α = l, g) and g is the gravity vector. The permeability
tensor, Kα depends on fluid viscosity, degree of saturation (relative perme-
ability law) and pore structure (intrinsic permeability). Figure 1a presents the
variation of intrinsic permeability with porosity adopted in this work.
The retention curve establishes the link between the degree of saturation
and suction. A modified van Genuchten law has been adopted (Fig. 1b):
1−λ
1 −λo λd
s o s
Sl= 1+ fd : fd = 1− (2)
Po Pd
where s is the suction, Po is the air entry value and λo is a model parameter.
The function fd is included to obtain more suitable values at high s.
It has been assumed that the diffusion flows are ruled by Fick’s law (i.e.
water vapour diffusion in gas phase and air diffusion in water) and that the
heat conduction is governed by Fourier’s law. To describe the mechanical
Fig. 1. FEBEX bentonite constitutive laws: a) variation of saturated permeability

with porosity, b) retention curve
behaviour the Barcelona Basic Model (BBM) has been adopted (Alonso et
al. 1990). All the laws and parameters used in the analysis are presented in
detail in Sánchez and Gens (2005).
The system of equations is solved in a coupled way. One unknown (or
state variable) is associated to each balance equation. Then, from the state
variables, the dependent variables are calculated using the constitutive equa-
tions or the equilibrium restrictions. In this way the finite element program
CODE_BRIGHT Olivella et al. (1996) has been developed. This code has been
used to perform the numerical analysis presented herein.
3 Infiltration Tests
The infiltration tests are being performed by CIEMAT in cylindrical cells,
whose internal diameter is 7 cm and inner length 40 cm (Fig. 2). They are
made of Teflon to prevent as much as possible lateral heat conduction, and
externally covered with steel semi-cylindrical pieces to avoid the deformation
of the cell by bentonite swelling. The material tested has been unsaturated
FEBEX bentonite, which is the clay used in the FEBEX Project to manufac-
ture the blocks of the FEBEX large scale tests.
The FEBEX clay was compacted with its hygroscopic water content
(around 14%) at an initial nominal dry density of 1.65 g/cm3 . Granitic water
is injected through the upper lid of the cell at a pressure of 1.2 MPa. In one
of the tests (GT40) the clay is being heated through the bottom surface at a
temperature of 100◦ C. The other test (I40) is being carried out at isothermal
conditions. The cells are instrumented with sensors of relative humidity and
Fig. 2. Experimental setup for the infiltration tests

temperature placed inside the clay at three different levels separated 10 cm.
The water intake and the relative humidity and temperature evolution at
different levels inside the clay are being measured online.
4 Numerical Modelling
The modelling has been focused on the analysis of the temperature and rela-
tive humidity. In relation to the water intake, only one analysis is presented,
due to problems occurred in its measurement. The study of these tests has
the additional interest to check whether the, already explained, problems in
the hydration rate observed in the FEBEX mock-up and ITT tests are also
present in these cells. The initial and boundary conditions of the model have
been imposed in order to be as close as possible to the experiments. An initial
suction of 120 MPa has been adopted, which corresponds approximately to
the relative humidity of 42% registered by the sensors inside the bentonite
just before hydration. An initially uniform temperature of 23◦ C has been as-
sumed. Initial hydrostatic stresses of 0.15 MPa have been adopted. Regarding
the boundary conditions a temperature of 100◦ C is imposed in the GT40 cell
at the contact between heater and bentonite (the bottom of the cell), while
a constant water pressure of 1.2 MPa is imposed at the other end of the cell
(upper part). The thermal boundary condition along the sample has been
adopted in order to adjust the temperature field, in that sense an external
temperature of 23◦ C has been fixed with a radiation coefficient of 1 (one).
Three main cases have been considered. Firstly, a ‘base case’ has been
adopted using the basic formulation presented in Sect. 2. The constitutive laws
adopted in the base case are the same one used in the analysis of the mock-up
test (Sánchez and Gens 2005). The second and third cases studied correspond
to the application of non-standard flow models developed to explain the slow
hydration observed in test GT40.
4.1 Base Case
The results of the isothermal cell (I40) measurements and modelling are shown
in Fig. 3. The evolution of relative humidity in different positions and also the
water intake is well captured by the model. Figure 4 presents the results of
the thermo-hydraulic cell (GT40). The thermal field is quite well reproduced
considering that the variations of temperature due to the changes in the lab-
oratory temperature are not taken into account in the simulations. In terms
of relative humidity, the model captures qualitatively the main trend of be-
haviour, that is, an increasing saturation in the zones close to the hydration
boundary and, in zones close to the heater, an initial wetting, due to the con-
densation of the water vapour coming form the bottom, and then a drying.
However, the model predicts a quick hydration after the drying, while the
experiment shows practically constant values of relative humidity in the three
measured positions. This is a trend similar to the one observed in the mock-
up test, with a kind of locking of the hydration of the clay. To explain this
unexpected behaviour, two non-standard flow models and the corresponding
numerical analyses are introduced in the following sections.
Fig. 3. Isothermal cell (I40): a) evolution of the relative humidity, b) water intake
(d is the distance from the heater)
Fig. 4. Thermo-hydraulic cell (GT40): a) evolution of temperature, b) relative

humidity (d is the distance from the heater)
Fig. 5. a) Threshold gradient phenomenon. b) Minimum hydraulic needed to get

measurable flows as a function of the dry density of the FEBEX bentonite
4.2 Non-Darcian Flow Behaviour
Experimental evidences show that under low hydraulic gradients (J), Darcy’s
simple relationship does not rule the liquid flow in some soils.
The strong clay-water interactions in clayed soils are suggested to explain
this non-Darcian flow behaviour. In this work the law adopted considers two
characteristic gradients (Fig. 5a): the threshold hydraulic gradient (Jo ) and
the critical hydraulic gradient (Jc ). Jo is the hydraulic gradient below which
no flow occurs. Jc is the hydraulic gradient below which flow occurs but it is
not Darcian. If the hydraulic gradient is higher than Jc , Darcy’s law applies
(Dixon et al. 1992). The average hydraulic gradient applied to obtain the
permeability of FEBEX bentonite shown in Fig. 1a was 15200. To clarify
the effect of hydraulic gradient on the value of hydraulic conductivity, the
hydraulic conductivity of FEBEX clay samples has been measured under low
hydraulic gradients and injection pressures
The hydraulic gradients applied range from 200 to 7200. No measurable
outflows have been obtained for certain hydraulic gradients depending on the
dry density of the bentonite. These values are shown in Fig. 5b and could
be regarded as threshold hydraulic gradients, since no flow has been obtained
below these gradients. The dispersion found when hydraulic gradients lower
than 1000 are applied could indicate that the critical gradient for this ben-
tonite would be around this value (Villar and Gómez-Espina 2006).
Figure 6a shows the results obtained using a threshold gradient model. The
base case results are also presented in dot lines. The following considerations
have been taken into account: i) a threshold gradient equal to 50 (this is a
common value found in the literature), ii) a critical gradient close to 1500;
and iii) a power law for the range of hydraulic gradient with non-Darcian’s
flow. This model can reproduce very well the trend to a very low hydration
observed in test GT40. On the other hand, the model underestimates the
hydration in the isothermal cell.
Fig. 6. Evolution of the relative humidity in the thermo-hydraulic cell (GT40): a)

threshold hydraulic gradient model, b) thermo-osmosis flow model (d is the distance
from the heater)
Fig. 7. Direct and coupled flow processes
4.3 Thermo-Osmosis Effects
The hydraulic gradient is the main force influencing the movement of water in
soils. It is, however, not the only one. Figure 7 presents all the different kinds
of flow (except the electrical component) that can occur in a porous media
and the corresponding gradient responsible for the movements. Generally, the
non-diagonal coefficients are relatively small and negligible compared to the
diagonal terms and the coupled processes can be ignored. However there are
certain problems in which, due to their particular conditions, the coupled
processes may play a more influential role. The thermal conditions imposed
on the thermo-hydraulic cell correspond to a practically constant thermal
gradient during the test. In contrast, the hydraulic gradient is very high at
the beginning of the test, but diminishes with the hydration of the barrier.
The liquid flows associated with these two gradients have opposite directions.
Generally, the advective flow (Darcy’s law) is the dominant flow. However,
at advanced stages of the test (when the hydraulic gradient becomes smaller),
it is possible that coupled phenomena (thermo-osmotic flow) could have a
noticeable effect on the behaviour of the system, causing a trend to slow
down the hydration in the zones close to the heaters. On the other hand, this
phenomenon would not influence the isothermal cell.
Fig. 6b shows the results of the modelling considering thermo-osmotic flow.
The base case results are also presented in dot lines. It is worthy to note the
lack of experimental data concerning the phenomenological coefficient asso-
ciated with thermo-osmotic effects. The thermo-osmotic constant adopted is
5.0 × 10−12 m2 /K/s and falls in the range of possible values found in the liter-
ature. A good agreement (qualitative) with the observations can be observed
in zones close to the heater. Near the hydration front the measured trend is
not well captured. The results of the isothermal cell do not change under this
hypothesis.
5 Conclusions
The analyses of two infiltration tests have been presented in this work. First
the results of a ‘base case’ using a ‘standard THM model’ have been pre-
sented. The results obtained with this model are not satisfactory, especially
when the comparisons with the thermo-hydraulic cell are analysed. Thresh-
old hydraulic gradient and thermo-osmotic effects have been included in the
formulation, these being two plausible phenomena that could explain the un-
expected trends observed in this kind of experiments. The results obtained are
very interesting since a physical interpretation for the unexpected behaviour
has been done. Each of these phenomena does not exclude the others and it
is possible that an explanation for the whole behaviour of the system would
require the combinations of some of them. These analyses show the need of
improving the ‘classical THM formulation’ when complex processes, as those
developed in the thermo-hydraulic cells, are studied.
References
Alonso E, Gens A, Josa A (1990) A constitutive model for partially saturated soils,
Géotechnique 40:405–430
FEBEX Project (2000) Full-scale engineered barriers experiment for a deep geolog-
ical repository for high level radioactive waste in crystalline host rock (Final
project report EUR 19612 EN EC Brussels)
Dixon D, Gray M, Hnatiw (1992) Critical gradients and pressures in dense swelling
clays, Can Geotech J 29:1113–1119
Olivella S, Carrera J, Gens A, Alonso EE (1994) Non-isothermal multiphase flow of
brine and gas through saline media, Transp in porous media 15:271–293
Olivella S, Gens A, Carrera J, Alonso EE (1996) Numerical formulation for a sim-

ulator (CODE-BRIGHT) for the coupled analysis of saline media, Engineering
Computations 13:87–112
Sánchez M, Gens A (2005) Final Report on THM modelling. FEBEX II (UPC-
Geomechanical Group, ENRESA Report: 70-UPC-L-5-015)
Thomas H, Cleall P, Chandler N, Dixon D, Mitchell H (2003) Water infiltration into
a large-scale in-situ experiment in an underground research laboratory, Géotech-
nique 53:207–224
Villar MV, Gómez-Espina R (2006) Deliverable 3.2.9: Progress report on laboratory
tests performed by CIEMAT (WP3.2 NF-PRO Report). Madrid, 47 pp
Moisture Effects on Argillaceous Rocks
Chun-Liang Zhang and Tilmann Rothfuchs
Gesellschaft für Anlagen- und Reaktorsicherheit (GRS),

Theodor-Heuss-Strasse 4, 38122 Braunschweig, Germany
chun-liang.zhang@grs.de, tilmann.rothfuchs@grs.de
Summary. In view of the disposal of high level radioactive waste in argillaceous

rocks various laboratory experiments have been performed at GRS’s geotechnical
laboratory on samples drilled from the Opalinus clay at the Mont Terri Rock Labo-
ratory in Switzerland and the Callovo-Oxfordian argillite at the Bure site in France.
This paper focuses on experimental studies of moisture effects on the clay rocks
including the relationship between suction and water content, reaction on de- and
re-hydration, and the self-sealing potential of the examined materials. The results
indicate that the clay rocks react very sensible to a moisture change. Wetting and
drying caused significant swelling and shrinking. After re-saturation, highly damaged
samples exhibited a very strong reduction in gas-permeability from 10−16 –10−17 m2
to 10−21 –10−22 m2 , indicating a high self-sealing capacity of the studied argillaceous
rocks.
Key words: clay rock, moisture, swelling, shrinking, self-sealing
Introduction
Argillaceous formations are being investigated worldwide as host medium for

the disposal of high level radioactive waste, because of their favourite proper-
ties. Argillaceous rocks are highly consolidated and usually characterized by
very low permeabilities. Their high swelling capacity and visco-plasticity pro-
vide a high potential for sealing cracks, pre-existing ones as well as new ones
created by construction of underground repositories. Additionally, argillaceous
rocks also offer a high sorption capacity for most radionuclides. Because of the
very low permeability and low hydraulic gradients in argillaceous formations,
diffusion dominates the water and radionuclide transport. Consequently these
properties lead to a reliable long-term isolation of disposed radioactive waste.
During the construction and operation phase of a repository, the ventila-
tion of the openings by relatively dry air and the subsequent backfilling with
unsaturated materials will lead to evaporation of the pore water from the sur-
rounding rock. After closure of the repository, the previously de-hydrated rock
320 Chun-Liang Zhang and Tilmann Rothfuchs
mass will take-up water again from the saturated area. De- and re-hydration
may alter the rock properties because of the high sensitivity of the clay miner-
als in the rocks. This issue has been intensively investigated at GRS’s geotech-
nical laboratory in the frame of several national and international research
projects.
Studied Materials
For the laboratory experiments, a number of cores were drilled with com-
pressed air from the Opalinus clay at the Mont Terri Rock Laboratory in
Switzerland and from the Callovo-Oxfordian argillite at the Bure Under-
ground Laboratory in France. The sedimentary argillaceous rocks are over-
consolidated. The main mineral compositions of the rocks are clay minerals,
quartz, and carbonates (Pearson et al. 2003, Lebon and Ghoreychi 2000).
The other properties such as grain density, dry density and water content
were measured on the samples by drying at 105◦ C. The basic characters of
the clay rocks are summarized in Table 1. Both studied clay rocks have very
similar properties.
Table 1. Petrophysical properties of the studied argillaceous rocks
Opalinus clay Callovo-Oxfordian

Property
at Mont Terri argillite at Bure
Clay content (%) 58–76 40–45
Carbonate content (%) 6–24 20–30
Quartz content (%) 5–28 20–30
Water content (%) 6.7 7.7
Grain density (g/cm3 ) 2.71 2.70
Dry density (g/m3 ) 2.34 2.25
Porosity (%) 15.1 16.5
Water Retention
In highly-consolidated clay rocks, a very significant portion of the water is ad-

sorbed on mineral surfaces. The adsorbed water is so strongly bound that
it may not be able to participate in advective transport under normally-
encountered pressure gradients. However, the adsorbed water is able to move
out thermodynamically from the pores at high external suction. In contrast
to that, external water can also be taken-up by unsaturated clays. The pro-
cess of de- and re-hydration is controlled by the relation between suction and
Moisture Effects on Argillaceous Rocks 321
25
Opalinus clay at Mont Terri
20
volume free GRS-data, wetting path
Water content (%)

wetting GRS-data, drying path
15 UPC-data, drying path
UPC-data, wetting path
10
volume constraint
5 wetting
initial state drying
0
0 1 10 100 1000
Suction (MPa)
Fig. 1. Water retention curve of the Opalinus clay
water content (water retention curve). For the Opalinus clay, the water re-
tention curve was determined on unconstraint samples placed in desiccators
at different values of relative humidity adjusted by means of salt solutions.
The result is illustrated in Fig. 1 in comparison with other data on constraint
samples (Munoz et al. 2003). It is evident that the water content decreases as
the external suction is high and conversely increases at low suctions. It is also
obvious that the unconstraint samples took-up a significant amount of water
of up to w = 24% at null suction over 100 days, accompanied by expansion.
Swelling Pressure
As well known, swelling in clays is the physico-chemical process due to inter-

actions between water and particle surface, which cause adsorption of water
on the internal and external surfaces of clay minerals, forming electrostatic
double-layers. In highly-consolidated natural clay rocks, the double-layers are
usually overlapping in the narrow spaces between particles. The interparti-
cle water-films are thus assumed to carry the lithostatic stress (Rodwell et
al. 1999). The average disjoining pressure acting in the water-films is equiva-
lent to the swelling pressure, which is usually measured on volume-constraint
samples by wetting with liquid water or water vapor. In conventional tests,
the swelling pressures obtained on clay rocks are limited to below 2 MPa
(Lebon and Ghoreychi 2000, Zhang et al. 2004), much lower than the litho-
static stresses at the sampling depths. GRS (Zhang et al. 2004) has developed
a new method for the determination of the swelling pressure of argillaceous
rocks. Figure 2 shows schematically the test principle and a test result ob-
tained on a sample of the Callovo-Oxfordian argillite, which was taken at the
depth of 455 m below the surface, corresponding to an overburden stress of
about 11 MPa.
20 120
axial strain = const
relative humidity
axial fixed 100
air outlet 15 drying
(%)
80
Axial stress (MPa)

stress reaction
Relative humidity
10 60
sample wetting
40
5
Lithostatic stress of 11 MPa
at sampling depth of 455 m 20
Callovo-Oxfordian argillite at Bure
air inlet axial fixed

0 0
0 10 20 30
Time (day)
Fig. 2. Swelling pressure of an argillaceous sample under axially-fixed and laterally-

unconstraint conditions
The sample was first uniaxially loaded to 13 MPa and then the axial strain
was fixed. Under the axially-fixed and laterally-unconstraint conditions, the
sample was wetted by pumping water vapor to it for five days, then dried
by pumping dry air for one day, and finally wetted again for 20 days. In the
first wetting phase at 100% relative humidity, the axial stress was stabilized
at about 12 MPa. The subsequent drying however caused a very quick drop
of the stress tending to zero. Conversely, the following wetting led the axial
stress to rise again. At an elevated humidity of 100%, the axial stress reached
about 10 MPa and remained at this level over the long test duration. The
periodical variation of the stress is well correlated to the fluctuation of the
humidity which was in turn induced by temperature changes. Obviously, the
stress reaction or the build-up of swelling pressure is controlled by the amount
of the pore water adsorbed on the minerals. When fully re-hydrated, the
swelling pressure reaches its maximum, which is almost equal to the lithostatic
stress at the sampling depth. Additionally, it is also interesting to point out
that the swelling pressure σ1 was built-up in the fixed axial direction at the
laterally-unconstraint condition, σ2 = σ3 = 0. This means that this pressure
is probably not a scalar quantity and should be represented by a second-rank
tensor (Rodwell et al. 1999).
Swelling and Shrinking Strains

In addition to the swelling pressure, swelling and shrinking deformations of
the clay rocks have also been examined on samples under different loading
conditions. Figure 3 illustrates the measurements of a swelling strain test on
an Opalinus clay sample at an axial stress of σ1 = 1 MPa and null lateral
stress (σ2 = σ3 = 0). The sample was dried and wetted by circulating dry air
and water vapor at different temperatures between 24 and 58◦ C over 1 year.
axial load
1 10
extension
drying wetting drying
air outlet T=24°C T=58°C
T=24-38°C
RH=20-50% RH=75-95% RH=3% 8
Water content (%)

0
Strain (%)
6
axial strain
sample
radial strain
4
air -1 water content
intlet axial stress = 1 MPa
compression
2
Opalinus clay at Mont Terri

volumetric strain
-2 0
0 60 120 180 240 300 360
Time (day)
Fig. 3. Swelling/shrinking of an argillaceous sample at low axial load
It can be clearly seen that axial, radial and volumetric strains of the sample
varied significantly with the changes in water content which resulted from the
changes of the environmental temperature and humidity. The first drying at
T = 24–38◦ C and RH = 20–50% caused a rapid reduction of the water content
from 6.7% to 1.3%. In correlation with the water release, a large shrinkage
took place in all directions to a volumetric reduction of 1.6%. When the water
content remained constant, the strains were nearly not changed. Elevating the
humidity to RH = 75–95% increased the water content to 4.5–5.5%. Corre-
spondingly, the sample expanded so strongly that the pre-shrinkage is almost
fully recovered. Drying again at 58◦ C and the resulting humidity of 3% led to
a quick evaporation of the pore water and reduction of the water content to
w = 0.5%, resulting in a strong collapse of the pores leading to a volumetric
reduction of 1.75%.
Figure 4 presents the measurements of swelling/shrinking deformations
of other samples of the Callovo-Oxfordian argillite at a high axial stress of
15 MPa. After a long-term creep over 9 months, the saturated samples were
exposed to relatively dry air of RH = 24 ± 4%. Under the highly-dried condi-
tion, the pore water evaporated from the samples, leading to a rapid collapse
of the pores equal to a compaction of 0.3%. The slight swelling and shrinking
strains observed over the further test duration are well correlated with the
fluctuation of the water content.
The above experimental observations suggest that a) the argillaceous rocks
have a high swelling capacity; b) the swelling/shrinking deformations in the
clay rocks are dependent on two factors, namely the amount of water uptake
and release as well as the confining stress; and c) the swelling and shrinking
strains are reversible under the tested conditions.
In addition, the mechanical stiffness and strength of the argillaceous rocks
are also strongly determined by the water content, as shown in Fig. 5. The air-
dried sample exhibited a high compressive strength of ∼ 42 MPa and a failure
strain of ∼ 2%, about two times higher than that strength of 24.5 MPa and
1.6 8
Collovo-Oxfordian argillite at Bure
axial stress = 15 MPa

sample 1
1.4 6
Water content (%)

Axial strain (%)
1.2 sample 2 4
shrinking
1.0 2
water release
swelling
0.8 0
240 260 280 300 320 340
Time (day)
Fig. 4. Swelling/shrinking of an argillaceous sample at a high axial load
50
Callovo-Oxfordian argillite at Bure 41.7 MPa
40
air-dried sample
Axial stress (MPa)
w = 2.8%
30
24.5 MPa
20
saturated sample
w = 7.1%
10
0
0 0.5 1 1.5 2
Axial strain (%)
Fig. 5. Stress-strain curves of saturated and air-dried argillaceous samples
of 1.0%, respectively, of the saturated sample. The high strength of the dried
clay rocks is probably induced by the increased friction resistances between
the grains with very thin or no water-films. However, the large compressive
strains of the dried clay rocks are mainly due to the collapse of the de-saturated
pores.
It is also to be noted that drying often gives a rise to cracks in the clay
rocks, depending on the confining stress. The cracks are usually concentrated
along the bedding planes.
Self-Sealing
As an indicator for self-sealing, the gas permeability was measured on a num-

ber of samples. Figure 6 compares the gas-permeaiblities obtained on two
highly-damaged samples of the Callovo-Oxfordian argillite before and after
water injection. The permeabilities measured at a confining stress of 1.5 MPa
before the water injection decrease steadily from 4 · 10−16 –9 · 10−17 m2 to
1E-14
sample 1
1E-16 damaged
(m 2)
state
sample 2
Test conditions:
Gas permeability
1E-18 confining stress = 1.5 MPa
injection pressure = 0.75 bar
Callovo-Oxfordian argillite before water injection
at Bure
1E-20
re-sealed
Test conditions: state
1E-22 confining stress = 2.4 MPa
injection pressure = 20 bar
after water injection
1E-24
0 10 20 30 40 50 60 70 80
Time (day)
Fig. 6. Gas-permeabilities of highly-damaged argillaceous samples before and after

water injection
1 · 10−16 –1 · 10−17 m2 over two months. After that, synthetic pore water was
injected to both ends of the samples at 10–15 bars for more than 5 months,
but no water outflow was observed. Subsequently, gas was injected again over
another two months, during which very low permeabilities were recorded be-
tween 5·10−22 and 2·10−21 m2 . These values are about five orders of magnitude
lower than those before the wetting.
In fact, the water is so strongly bound on the surfaces of the clay particles
and so completely occupies the narrow pores that it can not be removed by the
gas (or water) injected under the applied pressure gradients. The observation
of the permeability reduction due to re-hydration suggests a high self-sealing
capacity of the clay rocks.
Conclusions
The experimental findings on the Opalinus clay and the Callovo-Oxfordian

argillite suggest that the argillaceous rocks react very sensible to moisture
change. In the tests, it was observed that:
• re-hydration produced significant swelling strains even at high confining
stresses and large swelling pressures carrying the lithostatic stresses;
• in contrast, de-hydration caused shrinking and stress reduction;
• swelling and shrinking strains are reversible in the tested conditions; and
• re-hydration by liquid water led to a large reduction in gas-permeability
of highly-damaged samples from 10−16 –10−17 m2 to 10−21 –10−22 m2 , indi-
cating a high self-sealing capacity of the studied argillaceous rocks.
Acknowledgements
The authors gratefully acknowledge the funding by the German Federal Min-
istry of Economics (BMWi) and by the Commission of the European Com-
munities (CEC) under a number of contracts. The support from ANDRA by
providing the samples is also gratefully acknowledged.
References
Pearson FJ, Arcos D, Bath A et al. (2003) Mont Terri Project – Geochemistry of
Water in the Opalinus Clay Formation at the Mont Terri Rock Laboratory
Lebon P, Ghoreychi M (2000) French Underground Research Laboratory of
Meuse/Haute-Marne: THM Aspects of Argillite Formation. In: EUROCK2000,
Aachen, 27–31 March 2000
Munoz JJ, Lioret A, Alonso E (2003) Characterization of hydraulic properties under
saturated and non saturated conditions, VE-Experiment
Rodwell WR, Harris AW, Horseman ST et al. (1999) Gas Migration and Two-Phase
Flow through Engineered and Geological Barriers for a Deep Repository for
Radioactive Waste. EC/NEA Status Report, EUR 19122 EN
Zhang CL, Rothfuchs T, Moog H et al. (2004) Thermo-Hydro-Mechanical and Geo-
chemical Behaviour of the Callovo-Oxfordian Argillite and the Opalinus Clay.
GRS-202, ISBN 3-931995-69-0
Part VI
Retention Behaviour
Results from Suction Controlled Laboratory
Tests on Unsaturated Bentonite – Verification
of a Model
Ann Dueck
Clay Technology AB, Ideon, SE-223 70 Lund, Sweden ad@claytech.se
Summary. Highly compacted bentonite has been proposed as buffer material in

the Swedish concept for the disposal of nuclear waste. A laboratory program with
focus on hydro-mechanical properties of unsaturated highly compacted bentonite
has been running and one part included changes in volume. The results could be
used to determine different moduli but are presented here as a verification of a model
covering the impact of confinement on the retention properties.
The tests were performed with suction control, which was achieved by the vapour
equilibrium technique. The applied relative humidity was approximately 85% in the
tests presented in this paper. The tests were done both as constant load tests with
a change in relative humidity and as tests with constant relative humidity and a
change in the applied load. The load was applied axially and the maximum constant
load corresponded to an axial stress of 20 MPa. Also the radial stress was measured.
All tests started from a compacted condition with a water content of 10% and a
void ratio of approximately 0.6.
The investigation showed that the model, which was mainly based on constant
volume tests, also seems to be valid after volume change since good agreement was
achieved between measured and calculated results in the presented examples.
Key words: model, bentonite, unsaturated, water retention, relative humidity
1 Background
Highly compacted bentonite is proposed as buffer material in the Swedish

concept for the disposal of nuclear waste. Although bentonite is expected to
become saturated, it is important to understand the saturation process. A
laboratory study with mainly constant volume tests was presented by Dueck
(2004, 2006b) and new tests involving volume change have now been made.
Test results and interpretations of different moduli (swelling/shrinkage
modulus and compression modulus) and different retention curves are pre-
sented by Dueck (2006a). However, in this paper examples from the test re-
sults have been used to show that a model proposed by Dueck (2004) and
330 Ann Dueck
Dueck and Börgesson (2006) is valid also after the volume change exemplified
here. The model describes the relationship between swelling pressure, suction,
void ratio and degree of saturation and was mainly based on results from
constant volume tests.
2 Material and Test Methods

2.1 Material
The material is a commercially available sodium bentonite with a quality

symbol MX-80 from American Colloid Co. The bentonite has montmorillonite
as the dominating smectite clay mineral.
The particle density ρs = 2.78 Mg/m3 and water density ρw = 1.0 Mg/m3
were used for evaluation of void ratio and degree of saturation from measured
variables (see e.g. Börgesson et al. 1988, 1995).
2.2 Determination of Retention Curves
The appearance of the retention curve depends to some extent on the initial
water content (e.g. Dueck 2004). The retention curve is represented by the
relative humidity RH vs. water content w (the mass of water divided by the
mass of solids). The retention curve determined under free swelling condition
and starting with an initial water content deviating from 0% is called the
specific retention curve. Specific retention curves with initial water contents
of 10%, 17% and 27% were determined with a method described by Wadsö et
al. (2004) using 10 g samples placed in pans hanging from lids in glass jars.
The sample weight is monitored by below balance weighing.
Measurements were carried out at seven different RH values (11%, 33%,
59%, 75%, 85%, 93%, 98%). These RH values were achieved using various
saturated salt solutions (LiCl, MgCl2 , NaBr, NaCl, KCl, K2 SO4 ) except in
the case of RH = 93% where an aqueous unsaturated NaCl solution was used.
The RH values were taken from Greenspan (1977) and the vapour pressure
above the unsaturated 2 molal NaCl solution was taken from Clarke and Glew
(1985). The jars were placed in laboratory room climate with a temperature
of 20 ± 0.5◦ C.
2.3 Determination of Volume Change Properties
The volume change properties were investigated in oedometers especially de-

signed for this purpose. With this equipment it is possible to apply constant
load and constant suction to a sample. Both the load and suction are mea-
sured together with the deformation of the sample. The applied load was in
the range 0 to 20 MPa. After the test, the water content and void ratio were
measured.
Suction Controlled Laboratory Tests on Unsaturated Bentonite 331
The equipment consists of a steel ring around the sample with filters on
both sides. A piston and a force transducer are placed vertically above the
sample. A piston and a force transducer are also placed radially through the
steel ring. The sample has a diameter of 35 mm and a height of approximately
14 mm.
The constant suction is generated by the vapour equilibrium technique,
where saturated salt solutions generate a specified relative humidity. The sam-
ples are exposed to the humidified air in a circulation system where air is
pumped from the box with the salt solution to the steel ring enclosing each
sample. The air passes in grooves above and below the filter stones.
In order to be able to use the relative humidity tabulated for the salt
it is necessary to avoid temperature gradients in the equipment including
the compartment with the salt solution. To avoid this problem, the relative
humidity adjacent to the sample was measured and this value was used for the
evaluation. Three saturated salt solutions were used (NaCl, KCl and K2 SO4 )
which generate different RH (75%, 85%, 98%) at 20◦ C. In this paper only the
results from tests with RH = 85% are analysed.
Powder samples with the initial water content 10% were compacted and
placed in each oedometer ring. The initial void ratio was approximately 0.6.
Two different test types were performed; test types 1 and 2. In both test
types the dependent (observed) variables were volume and water content. In
test type 1 the independent (the varied) variable was the applied load and in
test type 2 the independent variable was RH.
3 Results
3.1 Retention Curves
The test results from measurements of specific retention curves with free
swelling samples are shown in Fig. 1. Results from Dueck (2006a) with the
initial water content equal to 10% are presented together with results from
Dueck (2004) with initial water contents of 17% and 27%. From the initial
water content both drying and wetting takes place.
A continuous formulation of the specific retention curve was necessary
for the analysis. A mathematical relation according to Eq. (1) was previously
formulated by Kahr et al. (1990). Eq. (1) was used with different constants (a,
b) to coincide with the measured values representing the different initial water
contents; 10% (6.3, 0.16), 17% (7.25, 0.2) and 27% (7.5, 0.2). The constants
are only valid and were only used in ranges shown in Fig. 1.
w = (a − ln[−(10−6 · ρw · R · T /ωw ) · ln(RH/100)])/b (1)
In Eq. (1) w is the water content (%), T is the temperature (K), R the
universal gas constant (8.31432 J/(mol K)) and ωv the molecular mass of water
332 Ann Dueck
40
35
30
25
w(%)
20
15
10
5
0
0 20 40 RH (%) 60 80 100
Fig. 1. Retention curves. The points represent laboratory results from Dueck
(2006a) and Dueck (2004). Lines are derived from to Eq. (1). The initial water
contents are 10% (), 17% () and 27% (♦)
vapour (0.018 kg/mol). The relative humidity is defined as RH = 100 · (p/ps )

where p is the partial pressure of pore-water vapour and ps is the saturation
pressure of water vapour over a flat surface of pure water at the temperature T .
3.2 Volume Change Properties
Examples of test results are shown in Fig. 2. The chosen tests were performed
at RH ∼ 85%. Two tests are shown from each test type and both mean stress
(solid lines) and axial stress (broken lines) are plotted vs. void ratio. The
mean stress Pmean is calculated from axial Paxial and radial Pradial stresses
according to Eq. (2).
Pmean = (Paxial + 2 · Pradial )/3 (2)
0,85 0,85
0,8 2 0,8
2
0,75 0,75
0,7 0,7
void ratio
void ratio
0,65 0,65
4
1
0,6 0,6
1
3
0,55 0,55
2
0,5 0,5
1b
0,45 0,45
10 100 1000 10000 100000
10 100 1000 10000 100000
axial stress/mean stress (kPa) axial stress/mean stress (kPa)
Fig. 2. Axial stress (broken lines) and mean stress (solid lines) vs. void ratio at
RH = 85% from a. two tests from test type 1 (left figure) and b. two tests from test
type 2 (right figure). Initial state (1), large load applied (1b), swelling (1–2), loading
(2–3), unloading (3-4)
In test type 1 a small load (approximately 100 kPa) was initially applied
(point 1). The radial stress was initially zero. The air with controlled RH
was applied subsequently, which caused an increase in radial stress and a
swelling (2). After equilibrium the load was increased in steps (2–3) and the
deformation was continuously measured. Finally, one of the samples was un-
loaded (3-4).
In test type 2 the final load was initially applied (1) (in this case 100 kPa
and 20,000 kPa). In the latter case this caused the radial stress to increase and
the volume to decrease (1b). The air circulation was applied subsequently and
the sample was brought to equilibrium with the deformation measured (2).
4 Analysis
4.1 Model
The model used includes mainly two equations, equations (3) and (4), which
relate the variables stress, relative humidity, water content and void ratio.
It was proposed by Dueck (2004) and mainly based on tests with constant
volume conditions. The included relationships are further described by Dueck
and Börgesson (2006).
Equation 3 describes the relation between the mean stress P , the actual
relative humidity RHact in the sample and the relative humidity according to
the retention curve at actual water content RHret . The retention curve is de-
rived from Eq. (1) with the constants a and b corresponding to the appropriate
water content.
R · T · ρw RHret (w)
P (RHact , w) = − · ln (3)
ωv RHact
Equation (4) describes the relation between the mean stress and the degree
of saturation Sr at a specified void ratio e. The initial degree of saturation
of the unloaded sample Sr,ini is included. The swelling pressure at saturation
Pret can be calculated from the retention curve (5) which includes the void
ratio e and the water content at saturation ws (6).
Sr − Sr,ini
P (Sr , e) = · Pret (e) (4)
1 − Sr,ini
R · T · ρw RHret (ws )
Pret (e) = − · ln (5)
ωv 100
e · ρw
ws = 100 · (6)
ρs
4.2 Evaluation According to the Model
The evaluation was made in steps. From the measured variables RH and P
at a specific time the water content w, void ratio e and degree of saturation
334 Ann Dueck
Sr were calculated from Eqs. (3)–(6). Since the equations are not explicitly
expressed in the unknown variables, void ratio was assumed and iterations
made until agreement was achieved between the calculated and measured
stress P .
The calculated and measured results are compared in Fig. 3. The mean
stress is plotted vs. measured void ratios (filled symbols) and vs. calculated
void ratios (unfilled symbols). Good agreement is shown between the measured
and calculated results.
0,85 0,85
0,8 0,8
0,75 0,75
0,7 0,7
void ratio
void ratio
0,65 0,65
0,6 0,6
0,55
0,55
0,5
0,5
0,45
0,45 10 100 1000 10000 100000
10 100 1000 10000 100000
mean stress (kPa) mean stress (kPa)
Fig. 3. Mean stress vs. measured void ratio (filled symbols) and calculated from the
model (unfilled symbols). a. two tests from test type 1 (left figure) and b. two tests
from test type 2 (right figure)
In the calculations only loading was considered, not unloading. At loading

the retention curve according to Eq. (1) corresponding to the actual calculated
water content at the start of the loading was used. The constants a and b were
then interpolated from the values valid for initial water contents 17% and 27%.
Only small changes in water content were calculated with the model during
loading.
Good agreement was also shown between measured and calculated void
ratio from most of the tests performed at RH = 75% and 98%. Further
evaluation will be made.
5 Conclusions
Suction and stress controlled laboratory tests on swelling samples of com-

pacted unsaturated bentonite have been made. The resulting measured vol-
ume change was compared to volume change calculated from measured relative
humidity and mean stress with a model presented by Dueck and Börgesson
(2006). The model was mainly based on results from constant volume tests but
the presented results show good agreement between calculated and measured
void ratio also after volume change.
Acknowledgement
This paper is a result of work funded by the Swedish Nuclear Fuel and Waste
Management Company (SKB).
References
Börgesson L, Hökmark H, Karnland O (1988) Rheological properties of sodium
smectite clay, SKB Technical Report TR 88–30
Börgesson L, Johannesson L-E, Sandén T, Hernelind J (1995) Modelling of the
physical behaviour of water saturated clay barriers; laboratory tests, Material
models and finite element application, SKB Technical Report TR 95–20
Clarke ECW, Glew DN (1985) Evaluation of the Thermodynamic Functions for
Aqueous Sodium Chloride from Equilibrium and Calorimetric Measurements
below 154◦ C. J Phys Chem Ref Data 14(2):536–537
Dueck A (2004) Hydro-mechanical properties of a water unsaturated sodium ben-
tonite, laboratory study and theoretical interpretation. PhD Thesis, Lund Insti-
tute of Technology, Sweden
Dueck A (2006a) Hydro-mechanical properties of unsaturated MX–80, laboratory
study 2005–2006. Report in progress
Dueck A (2006b) Laboratory results from hydro-mechanical tests on a water unsat-
urated bentonite. Submitted for publication
Dueck A, Börgesson L (2006) Model suggested for an important part of the hydro-
mechanical behaviour of a water unsaturated bentonite. Submitted for publica-
tion
Greenspan L (1977) Humidity fixed points of binary saturated aqueous solutions,
Journal of research of the national Bureau of Standards, A. Physics and Chem-
istry, 81A(1):89–96
Kahr G, Kraehenbuehl F, Stoeckli HF, Mûller-Vonmoos M (1990) Study of the
water-bentonite system by vapour adsorption, immersion calorimetry and X-
ray techniques: II. Heats of immersion, swelling pressures and thermodynamic
properties, Clay Minerals 25:499–506
Wadsö L, Svennberg K, Dueck A (2004) An experimentally simple method for mea-
suring sorption isotherms, Drying Technology 22(10):2427–2440
Variation of Degree of Saturation
in Unsaturated Silty Soil
Ali R. Estabragh1 and Akbar A. Javadi2

1
Faculty of Soil and Water Engineering, University of Tehran, Karaj 31587-77871,
Iran a.estabragh@gmail.com
2
School of Engineering, Computer Science and Mathematics, University of
Exeter, EX4 4QF, UK a.a.javadi@exeter.ac.uk
Summary. This paper presents the results of an experimental study on variation

of degree of saturation of a compacted unsaturated silty soil during isotropic loading
and unloading. A series of laboratory experiments have been conducted in a double-
walled triaxial cell (and a conventional triaxial cell for saturated samples) on samples
of a compacted silty soil. In the experiments the soil samples were subjected to
isotropic consolidation followed by unloading under constant suction. The results
show that during consolidation a significant increase in degree of saturation was
observed. In contrast, during subsequent isotropic unloading, where only a very
small elastic component of swelling occurred, the changes of degree of saturation were
very modest. Comparison has been made between the experimental and theoretical
prediction using a relationship proposed in the literature.
Key words: unsaturated soils, suction, degree of saturation, consolidation, unload-

ing
1 Introduction
It is generally understood that unsaturated soils have a high suction by virtue

of their partial saturation and the resulting capillary actions, and its magni-
tude is related to the water content or degree of saturation.
Gallipoli et al. (2003) and Buisson and Wheeler (2000) indicated that the
relationship between the degree of saturation, Sr , and suction, s, for a given
soil is non-unique because the variation of the void ratio in deformable soils
results in changes in the void dimensions and also in changes in the connecting
passageway between them. This, in turn, causes corresponding variation in the
soil water characteristic curve. As changes in void ratio affect the soil water
characteristic curve, irreversible changes of degree of saturation will occur
during loading and unloading of mean net stress if there is an irreversible
change of void ratio.
Bishop and Blight (1961) used the concept of state surface in order to
represent the volume change of unsaturated soils. Matyas and Radhakrishna
(1968) found the state surface for a mixture of flint powder and kaolin. Their
tests involved either isotropic or anisotropic compression for examining the
validity of stress state variables that were suggested by Bishop and Blight
(1963). They plotted the results of the both tests series in terms of void ratio,
e, and Sr against s and mean net stress, p . Matyas and Radhakrishna (1968)
presented experimental data defining state surfaces which related e and Sr to
p and s.
Fredlund (1979) was the first to suggest the mathematical expression for
the state surfaces for void e and water content, w. Lloret and Alonso (1985)
also proposed equations for state surfaces for void ratio and degree of satura-
tion and their equation for degree of saturation is as follows:
Sr = a − tanh(bs)(c + dp ) (1)
where a, b, c and d are soil constants. They reported that the above equation
gives better predictions at low stress levels however it does not satisfy satu-
rated condition when suction approaches zero. In this paper, the variation of
degree of saturation during loading and unloading is studied and the applica-
tion of equation (1) in prediction of degree of saturation of unsaturated soils
is examined in the light of a comprehensive experimental study.
2 Experimental Study
2.1 Soil Properties
The soil used in the testing programme was a silty soil with low plasticity. The
soil comprised 5% sand, 90% silt and 5% clay and had a liquid limit of 29%
and plasticity index of 19%. The results of the standard proctor compaction
test indicated a maximum dry density of 1.74 Mg/m3 at an optimum water
content of 14.5%.
2.2 Test Procedure and Program
Compacted samples were prepared using a compaction mould designed specif-

ically for static compaction. The compaction was done in nine layers in a
compression frame at a fixed displacement rate of 1.5 mm/min to a maximum
vertical load of 400 kPa. The samples were 76 mm high and 38 mm in diam-
eter. All samples were compacted in an identical fashion in order to produce
the same initial soil fabric in each test. The samples produced by this method
were found to be very uniform.
Variation of Degree of Saturation in Unsaturated Silty Soil 339
2.3 Experimental Apparatus
A Bishop Wesely triaxial cell was modified to a double walled cell for measur-
ing the volume change of the samples. The pore air pressure was controlled
by applying an air pressure through a line from the top of the sample. Pore
water pressures were measured through a high air entry disk at the bottom
of the sample. The axis translation technique (Hilf 1956) that was adapted
for triaxial testing by Bishop and Donald (1961) was used for creating the
desired suction in the samples. Four GDS controller units were used to ap-
ply the pressures in the apparatus. All the experimental data were recorded
continuously by a computer.
2.4 Experimental Procedure
Equalization
The first stage of the tests involves equalization. The purpose of the equal-
ization tests was to create a desired suction in a sample by allowing the pore
air pressure and pore water pressure to equalize to the applied air pressure
and back pressure respectively. At this stage by applying the required air and
water pressures the samples were brought to the desired value of suction (zero,
100, 200 and 300 kPa).
Ramp Consolidation and Unloading
After the sample was equalized at a specified suction (zero, 100, 200 or
300 kPa) and mean net stress, it was loaded isotropically under the constant
suction (air back pressure and water back pressure were kept constant) to a
pre-selected value of mean net stress. The process of ramped consolidation
was used to limit the excess pore water pressure generated at the top face
of the sample (Estabragh and Javadi 2004). The sample was then unloaded
isotropically in ramp procedure to a predefined lower value of mean net stress.

A set of experiments were carried out on the samples of unsaturated silty soil.
The results of variation of degree of saturation during loading and unloading
at constant suctions of 0, 100, 200 and 300 kPa are shown in Fig. 1. Dur-
ing, isotropic loading when large plastic reductions in void ratio occurred, a
significant increase in the degree of saturation was observed. In contrast, dur-
ing subsequent isotropic unloading, when only a very small elastic swelling
occurred, the changes of degree of saturation were very small. The results
of irreversible changes of degree of saturation presented by Zakaria (1995),
Romero (1999) and Gallipoli et al. (2003) strongly support theses findings.
Degree of saturation 0.7

0.65
(a )
0.6
0.55
0.5
0.45
0.4
0.35
10 100 1000
Mean net stress, p (kPa)
0.5
Degree of saturation
(b)
0.45
0.4
0.35
0.3
10 100 1000
0.4
Degree of saturarion
(c)
0.35
0.3
0.25
10 100 1000
Fig. 1. Isotropic loading-unloading tests at: (a) s = 100 kPa; (b) s = 200 kPa; (c)
s = 300 kPa
As the specific volume, v, decreases the dimensions of voids and of connecting

passageways between voids tend to decrease, so that a higher value of suction
is required to produce a given degree of saturation. Figure 1 shows that the
main variation in degree of saturation occurred after the yield point as the
great proportion of deformation occurs after yielding. Estabragh and Javadi
(2004) performed a comprehensive set of isotropic compression tests on soil
0.7
0.6
0.5
0.4
0.3
(a )
0.2
300 350 400 450 500 550
0.4
0.35
0.3
(b)
0.25
300 350 400 450 500 550
0.4
0.35
0.3
(c)
0.25
300 350 400 450 500 550
Fig. 2. Experimental and predicted values of Sr on isotropic normal compression

lines at: (a) s = 100 kPa; (b) s = 200 kPa; (c) s = 300 kPa
samples at different values of suction (0,100,200 and 300 kPa). The analysis of
the experimental results indicated that, for the range of stresses considered,
the normal compression lines at constant suction follow a linear relationship
in the semi-logarithmic plane of e − ln p .
It was also shown that the variations of degree of saturation for post yield
stresses are on straight line similar to the normal compression lines.
The results were analyzed using the conventional state surface expression
for degree of saturation, suggested by Lloret and Alonso (1985) (equation (1)).
A value of 1 was used for the parameter a (in order to predict full saturation
at zero suction), and the values of b, c and d were found by fitting the exper-
imental isotropic normal compression lines at constant suctions of 100, 200
and 300 kPa. For each suction the values of Sr were predicted from equa-
tion (1). Figure 2 shows the experimental and predicted variations of degree
of saturation during post yield sections of isotropic loading at three different
values of suction 100, 200 and 300 kPa. It appears that equation (1) provides
a reasonable match at s = 200 and 300 kPa but the match is not so good
at s = 100 kPa. This behaviour is similar to that reported by Gallipoli et al.
(2003). They developed a simplified version of the equation that was proposed
by Van Genuchten (1980) for prediction the variation degree of saturation in
unsaturated soils during loading and unloading. They also examined appli-
cability of equation (1) for predicting the variation of degree of saturation
and concluded the accuracy of the predictions for some suctions were not
reasonable.
4 Conclusion
The results show that significant changes in Sr occurred during isotropic load-
ing, even though the suction was held constant This can be attributed to the
influence of volumetric strain as the main changes of Sr coincide with the post
yield sections of loading stages when large changes of v were occurring.
Corresponding simulations were also performed with a conventional state
surface expression for Sr . The results demonstrate that, where as a state
surface expression can match the observed variation of Sr for a particular
type of stress path (by selection of suitable parameter values); it is incapable
of representing the variation of Sr for the full range of stress paths. This
indicates that an alternative explicit form of state surface expression for Sr
might provide a better match to the isotropic normal compression behaviour
at different values of suction.
References
Bishop AW, Blight GE (1961) Strength and consolidation characteristics of com-
pacted soils. Ph.D Dissertation, University of London
Bishop AW, Blight GE (1963) Some aspects of effective stress in saturated and
unsaturated soils, Geotechnique 13(3):177–197
Bishop AW, Donald IB (1961) The experimental study of partly saturated soil in
the triaxial apparatus, Proc 5th Int Conf Soil Mech Paris 1:13–22
Buisson MSR, Wheeler SJ (2000) In: Tarantrno A, Mancuso C (eds) Inclusion of
hydraulic hysteresis in a new elasto-plastic framework for unsaturated soils. In
experimental evidence and theoretical approaches in unsaturated soils 109–119
Estabragh AR, Javadi AA (2004) Effect of compaction pressure on consolidation
behaviour of unsaturated soil, Can Geotech Eng J 41:540–550
Fredlund DG (1979) Appropriate concepts and technology for unsaturated soil, Can
Geotech Eng J 16:121–139
Gallipoli S, Wheeler SJ, Karstunen M (2003) Variation of degree of saturation in
deformable unsaturated soil, Geotechnique 53(1):105–112
Hilf JN (1956) An investigation of pore water pressure in compacted cohesive soils.
Technical Memorandum No. 654, Bureau of Reclamation, USDI, Denver, Col-
orado
Lloret A, Alonso EE (1985) State surface for partially saturated soils. Proc 11th Int
Conf Soil Mech Found Engng San Francisco 2:557–562
Matyas EL, Radhakrishna HS (1968) Volume change characteristics of partially
saturated soils, Geotechnique 18(4):432–448
Romero E (1999) Characterisation and thermo-hydromechanical behaviour of unsat-
urated boom clay: an experimental study. PhD Thesis, Universitat politecnical
de catalurya, Spain
Van Genuchten MT (1980) A closed-form equation for predicting the hydraulic con-
ductivity of unsaturated soil. Soil Science Soc Am J 44:892–898
Zakaria I (1995) Yielding of unsaturated soil. PhD Thesis, University of Sheffield,
UK
Mechanical Behaviour of Compacted Scaly Clay
During Cyclic Controlled-Suction Testing
Camillo Airò Farulla1 , Alessio Ferrari1 , and Enrique Romero2

1
Università degli Studi di Palermo, Italy
airo@diseg.unipa.it, ferrari@diseg.unipa.it
2
Universitat Politècnica de Catalunya, Spain
enrique.romero-morales@upc.edu
Summary. Scaly clays are stiff and highly fissured soils due to their complex geo-
logical history. Owing to these characteristics, they may be placed and compacted
as excavated, to obtain a material with adequate engineering properties, suitable as
core material in earth dams and in waste isolation fills.
With reference to these engineering applications, an exhaustive experimental
programme is developed to characterise the hydro-mechanical behaviour of this com-
pacted material under a variety of controlled-suction paths.
This paper presents selected test results, focusing on the interpretation of spe-
cific behavioural features of the volume change of this complex material; namely,
the strain accumulation and the volumetric irreversible features developed during
suction cycling at different stress levels.
Key words: compacted tectonised clay, wetting–drying cycle, controlled suction

oedometer, strain accumulation, elastic strain, microstructural elastic compression
index
Introduction
Scaly clays are strongly tectonised and are composed of small stiff clayey
fragments (scales), almost always angular, whose size ranges from a few mil-
limetres to some centimetres.
Owing to their structural characteristics and natural water content these
clays may be placed and compacted as excavated to obtain a material with
low permeability and appreciable shear strength and stiffness, suitable as con-
struction material in many Civil Engineering applications. In Sicily these ma-
terials were used, successfully, for the core of two dams built in the early
sixties and are often used for isolation of industrial and domestic waste banks
(Airò Farulla and Valore 1993).
346 C. A. Farulla et al.
In these cases because of construction management and environmental

fluctuations, compacted scaly clay fills may be subjected to cyclic and strong
drying and wetting paths during different stages of their design life.
Clayey soils undergo an increase in volume during water uptake, but also
experience an important amount of shrinkage on water removal, which give
rise to the accumulation of compression or expansion strains during suction
cycles.
Several studies and research programmes have been developed on this sub-
ject (Dif and Bluemel 1991, Day 1994, Alonso et al. 2005); however, very few
experimental studies have been reported in the literature with respect to liquid
water transfer in cyclic wetting and drying tests under controlled conditions
in the low suction range.
The objective of the research presented in this paper is focused on the
investigation of the volume change response of compacted scaly clay samples
subjected to several wetting and drying cycles in controlled suction oedome-
ters. Oedometer tests were carried out under different values of constant ver-
tical net stress, and wetting and drying cycles were performed varying applied
matric suction between 10 and 800 kPa.
The volume change response of the samples is discussed and interpreted
within the context of the elasto-plastic model proposed by Gens and Alonso
(1992) and Alonso et al. (1999) (BExM: Barcelona Expansive Model).
Experimental Programme
Tested Material
Tested samples were prepared using a scaly clay outcropping near Palermo.
The material is a kaolinitic-illitic clay with liquid limit wl = 58% and plasticity
index Ip = 30%. The specific gravity is Gs = 2.78.
The air-dried clay with a hygroscopic water content wh = 0.05 was disag-
gregated by a rubber pestle, and the fraction passing at n◦ 4 ASTM sieve was
selected. Distilled water was added and carefully mixed. After a curing time of
2 or 3 days, samples were compacted dynamically with a non standard Proc-
tor procedure to a target dry density. Compaction energy was approximately
0.71 MJ/m3 .
Some of the physical characteristics of the tested samples are collected in
Table 1.
Controlled-Suction Technique and Experimental Programme
The test programme included five cyclic wetting and drying tests at constant
net vertical stress, σv , carried out of by means of two controlled-suction oe-
dometers designed at the Geotechnical Laboratory of the UPC (Technical
Mechanical Behaviour of Compacted Scaly Clay 347
Table 1. Initial characteristics of the tested samples
Sample w0 γd0 [kN/m3 ] e0 S0

A1 0.156 17.17 0.59 0.73
A2 0.151 17.68 0.54 0.77
A3 0.151 17.59 0.55 0.76
A4 0.143 18.15 0.50 0.79
B2 0.150 18.25 0.50 0.84
University of Catalonia, Barcelona, Spain). Each oedometer is equipped with

a ceramic disk with 1.5 MPa nominal air entry value.
Suction was applied by means of the axis translation technique, keeping a
constant air pressure, ua = 1000 kPa, and varying the applied water pressure,
uw , according to the air overpressure technique (Romero 2001).
Due to the high values of the control air pressure and of the degree of
saturation of the tested samples, the translation of the air pressure to the
reference value was accomplished, under water undrained conditions, with
a very careful step by step procedure in order to minimise the volumetric
deformation of the soil skeleton and the variation of the initial matric suction
of the samples (Airò Farulla 2004, Airò Farulla and Ferrari 2005).
The displacements measured during the axis translation steps were in the
order of a few microns and the variations in the degree of saturation were
negligible.
After the reference air pressure translation, initial suction was applied
(s = 800 kPa), and, when samples reached equilibrium, the selected vertical
net stresses (100, 200, 300 and 400 kPa), were applied.
The first wetting path started at the end of the stress-induced consolidation
process by reducing the applied suction to 10 kPa. Subsequent drying-wetting
reversals in the suction interval of 10–800 kPa were then applied. Each equali-
sation step lasted until the rate of straining had reduced to a limit strain rate
equal to or lower than 0.1%/day.
To reduce test duration, suction was changed in a single step in the interval
10–800 kPa. Four to six cycles were applied for each vertical stress.
In order to assess the reliability of the adopted procedure, in test B2
each wetting and drying step was accomplished limiting suction increment to
200 kPa. Because this procedure resulted in a very large test length, only three
suction cycles were applied.
Water content and saturation degree variations were determined by mea-
suring water inflow or outflow with a burette with a resolution of 0.02 cm3 .
Displacements were measured with a micrometer with a resolution of 1 μm.
The tests were performed at a constant temperature of T = (20 ± 1)◦ C.
Stress-Path Followed
The different stress paths followed in the test programme are indicated in the
(σv , s) plane (Fig. 1).
Fig. 1. Stress paths followed and LC yield locus for the as-compacted material
The drying-wetting paths were selected taking into account the position
of the yield locus LC related to the as compacted material properties (Alonso
et al. 1990). To determine the LC yield curve position in the (σv , s) plane, a
complementary test programme was developed, including two constant suc-
tion loading and unloading oedometer tests (s = 200–600 kPa), and a loading-
unloading oedometer test on a compacted sample that was previously satu-
rated under constant volume (isochoric) conditions.
As shown in Fig. 1, applied vertical stress were lower or equal to the yield
stress for saturated conditions, in order to investigate the behaviour of the
soil in the “swelling region” bounded by the LC curve.
Test Results and Interpretation

Complementary Tests
Parameter values determined in the complementary tests are collected in Ta-
ble 2, where σv0 is the yield vertical net (or effective for the saturated sample)
stress; λ is the slope of the virgin compression line (e–ln σv plane) and κ is
the elastic (unloading-reloading) compression index (e–ln σv plane).
Table 2 reports, also, suction values (s) applied during the loading-
unloading cycles, water content (w0 ), void ratio (e0 ) and dry density (γd )
of the as-compacted samples.
Table 2. Initial characteristics of the tested sample and parameter values deter-
mined in the complementary tests
Sample w0 γd0 e0 s σv0 λ κ

[kN/m3 ] [kPa] [kPa]
LP10 0.150 18.33 0.49 0 400 0.083 0.034
LP8 0.155 17.66 0.54 200 480 0.079 0.030
LP9 0.160 17.97 0.52 600 540 0.065 0.020
The best fit of the experimental points on the s–λ(s) plane, according to

equation:
λ(s) = λ(0) r + (1 − r)e−βs (1)
allowed determining: r = 0.74 and β = 2.07 MPa−1 .
The yielding locus, LC, in the (σv0 , s) plane (Fig. 1) was determined by
fitting the experimental points according to equation:
∗
λ(0)−κ
σv0 σv0 λ(s)−κ
= (2)
σvc σvc
∗
where σv0 is the yield vertical effective stress for the saturated sample and σvc
a reference stress fixed equal to 180 kPa for the best fitting of the experimental
points. Even if data collected in Table 2 pointed out dependence with applied
suction, κ was supposed a constant parameter equal to 0.030.
Cyclic Wetting-Drying Tests
Wetting and drying volumetric deformations at equilibrium are plotted in

Fig. 2 for each applied vertical net stress. Compressive strains are considered
positive.
The first wetting path results in sample expansion for σv = 100–300 kPa,
while sample A4 (σv = 400 kPa) swells at the beginning of the equalization
step and then collapses – an expected behaviour due to the proximity to the
LC curve plotted in Fig. 1. Swelling and collapsing strains are quite similar,
so that final volumetric strain is almost nil (εv = 0.02%).
Successive suction increase from 10 kPa to 800 kPa induces a significant
shrinkage in all samples. Volumetric contraction strains along this path are
in the range of 2%, almost independent from applied vertical net stress. As
a consequence, in the first suction cycle a quite different material volumetric
behaviour in wetting and in drying may be detected for the different stresses
applied. However, as the subsequent wetting-drying cycles are applied the
differences between two successive wetting-drying volumetric strains become
smaller, and the soil tends, clearly, towards a reversible or “elastic” state.
This trend is better appreciated in Fig. 3 where the absolute values of the
Fig. 2. Volumetric strains in cyclic controlled suction oedometer tests
volumetric strains either in expansion or compression are plotted versus the

cycle number.
As the number of cycles increases, shrinkage strain intensity reduces and
swelling strain increases, so that their difference tends to zero. However,
shrinkage strains are greater than swellings and the material accumulates
irreversible compression before the final reversible condition is achieved.
Usually four cycles are sufficient to achieve the final reversible condition
(Fig. 3).
These results presented in Figs 2–3 provide data to investigate the abil-
ity of the BExM (Gens and Alonso 1992, Alonso et al. 1999) to model soil
behaviour against cyclic suction changes. Before this is done, however, some
simplifying hypotheses have to be formulated in the double structure approach
in order to facilitate parameter determination (Alonso et al. 2005). These are:
suction induced elastic deformations are considered only as microstructural
deformations, and any suction induced macrostructural elastic deformations
are ignored. During the application of drying and wetting cycles the SI (suc-
tion increase) and SD (suction decrease) yield surfaces are assumed always
activated and macrostructure undergoes only coupled plastic deformations.
According to these hypotheses the total volumetric deformation increment
(Δεv ) measured at the end of each wetting or drying step may be considered
as the sum of the elastic microstructural deformation increment indicated
with superscript m and the coupled plastic strain increment denoted with
Fig. 3. Evolution of the absolute values of the volumetric strains with the number
of controlled suction cycles
superscript c:
e p
Δεv = (Δεm c
v ) + (Δεv ) . (3)
The elastic components, extrapolated from the total volumetric deforma-
tions measured in the last applied cycles when irreversible swelling or shrink-
age strain increments were negligible, are represented in Fig. 4a versus the
vertical net stress.
Elastic volumetric strain induced by a change in suction from an initial
value si to a final value sf , can be expressed as (Alonso et al. 2005):

m e κm σ v + sf
(Δεv ) = ln (4)
1 + e0 σ v + si
where κm is the microstructure elastic (unloading-reloading) compression in-
dex.
Equation (4) was used to derive the microstructural elastic index, κm , for
each of the four vertical net stresses applied. The calculated values, repre-
sented in Fig. 4b versus vertical net stress, belong to a very narrow range
(0.011–0.013); the mean value, κm = 0.012, may be assumed as the constant
value of the loading-unloading compressibility index of the microstructure.
If the elastic components are subtracted from the total final strain incre-
ments given in Fig. 3, coupled plastic volumetric components may be obtained
and their evolution with the applied suction cycles could be investigated.
Fig. 4. Elastic volumetric strains (a) and microstructural elastic index (b) vs ver-
tical net stress
As an example in Fig. 5 plastic volumetric strain increments versus cycle

number are plotted for A2 and B2 samples.
Plastic strain increments both in drying and in wetting are positive. It
means that during wetting even if measured total volumetric strain incre-
ment is negative (swelling), macrostructure experiences a moderate plastic
compression. This plastic compression is associated with the invasion of the
macrostructure by the active microstructure (Alonso et al. 1999).
Fig. 5. Evolution of computed plastic volumetric strain increments with the number
of cycles
Conclusions
A systematic experimental programme was carried out to study the effect of
applying controlled-suction cycles using axis translation technique on com-
pacted scaly clay. Selected test results were presented, which showed strain
accumulation and volumetric irreversible features developed during suction
cycling at different stress levels.
Test results have been interpreted within the framework of an elastoplastic
constitutive model (BExM) described in Alonso et al. (1999), which uses a
double structure approach. Collected results enable to calculate the constant
loading-unloading compressibility index of the microstructure according to
this model.
As previously noted, test B2 was carried out in order to assess the influence
of the suction increment intensity on test results. Analysed results show that
very little differences may be detected between A2 and B2 test results. From
this point of view, cycling suction in a single step results in a reliable procedure
in wetting-drying tests, even if this procedure is not lacking in shortcomings
when parameter of the model are to be determined.
Owing to cyclic suction variations material accumulates shrinkage for all
the intensities of constant net vertical stress applied. However, a few cycles
are needed to achieve an “elastic” or reversible condition, in which total vol-
umetric strain increment in wetting and in drying are almost equal. During
wetting macrostructure experiences moderate plastic (coupled) compressions,
which are associated with the invasion of the macrostructure by the active
microstructure. Plastic volumetric strain increments at macroscopic level are
slightly masked by the dominant microstructural volumetric strains of the
active clay.
References
Airò Farulla C (2004) Comportamento idraulico e meccanico dell’argilla a scaglie
compattata del nucleo delle dighe Scanzano e Rossella. In: AGI, XXII Conv Naz
di Geotecnica, Palermo, 22–24 September, 445–452
Airò Farulla C, Ferrari A (2005) Controlled suction oedometric tests: analysis of
some experimental aspects. In: Tarantino, Romero, Cui (eds) Advanced Exper-
imental Unsaturated Soil Mechanics. Taylor & Francis Group, London, 43–48
Airò Farulla C, Jommi C (2005) Suction controlled wetting-drying cycles on a com-
pacted scaly clay. In: Proc Int Conf on Problematic Soils, 25–27 May, East Medit
Univ, Famagusta, N. Cyprus
Airò Farulla C, Valore C (1993) Some aspects of the mechanical behaviour of com-
pacted tectonized clays. In: Anagnastopulos et al. (eds) Geotech Eng Of Hard
Soils – Soft Rocks. Balkema, Rotterdam, 335–342
Alonso EE, Gens A, Josa A (1990) A constitutive model for partially saturated soil,
Alonso EE, Vaunat J, Gens A (1999) Modelling the mechanical behaviour of expan-
sive clays, Engineering Geology 54:173–183
Alonso EE, Romero E, Hoffmann C, Garcia-Escudero E (2005) Expansive bentonite-
sand mixtures in cyclic controlled-suction drying and wetting, Engineering Ge-
ology 81:213–236
Day RW (1994) Swell-shrink behaviour of compacted clay, J Geotech Eng
120(3):618–623
Dif AE, Bluemel WF (1991) Expansive soils under cyclic drying and wetting. Tech-
nical note, Geotech Testing J 14(1):96–102
Gens A, Alonso EE (1992) A framework for the behaviour of unsaturated expansive
clays, Can Geotechnical J 29:1013–1032
Romero E (2001) Controlled-suction technique. In: Gehling WYY, Schnaid F (eds)
Proc 4th Symp Brasil Solos Nao Saturados, Porto Algre, Brasil, 535–542
Prediction of Soil–Water Characteristic Curve
Based on Soil Index Properties
Navid Ganjian1 , Yadollah Pashang Pisheh2 , and Seyed Majdeddin Mir

Mohammad Hosseini2
1
University of Tehran, Iran nganjian@ut.ac.ir
2
AmirKabir University, Iran ypashangpisheh@aku.ac.ir, smmirhos@aku.ac.ir
Summary. The measurement of soil parameters for the unsaturated soil constitu-
tive models needs extensive laboratory tests. For most practical problems, it has
been found that approximate soil properties are adequate for analysis. Thus, em-
pirical procedures to evaluate unsaturated soil parameters would be valuable. The
soil–water characteristic curve (SWCC) can be used to estimate various parameters
used to describe unsaturated soil behavior. The SWCC is a relationship between
soil suction and some measure of the water content. It can be measured or predicted
based on soil index properties. Estimation based on index properties is highly desir-
able due to its simplicity and low cost.
In this paper, a new model for predicting the SWCC based on soil index prop-
erties and Van Genuchten equation is presented. The comparisons show that the
SWCC predicted by this model is in a good agreement with the experimental re-
sults.
Key words: unsaturated soils, soil–water characteristic curve (SWCC), matric suc-
tion, water content
1 Introduction
Unsaturated soil mechanics has received wide attention across geotechnical

communities because of its practical importance. The investigation of unsat-
urated soils behavior requires evaluating the particular behavior parameters
in these soils. A theoretical basis for unsaturated soil mechanics has been es-
tablished over the past three decades. The constitutive equations for volume
change, shear strength, and flow through unsaturated soil have become gen-
erally accepted in geotechnical engineering (Fredlund and Raharjo 1993). The
fundamental accepted principal in this theory is that the unsaturated soil be-
havior could not be described just by making use of one stress state variable;
in other words, both the net normal stress, (σ −ua ), where σ is the total stress
and ua is the pore-air pressure, and the matric suction, (ua − uw ), where uw
is the pore-water pressure, are generally required for the constitutive models.
356 N. Ganjian et al.
As a result, evaluation of suction is most essential to assess the unsaturated

soil behavior.
Suction evaluating in direct measure needs to time consuming and costly
laboratory tests, so estimation of suction with indirect methods based on
other parameters such as water content would be considered. Consequently,
the soil–water characteristic curve (SWCC) that defines degree of saturation
corresponding to particular suction in a soil is widely used to estimate un-
saturated soil properties. Through the practical applications of the soil–water
characteristic curve, prediction of shear strength, water storage, and perme-
ability coefficient may be pointed out. Figure 1 provides a qualitative illus-
tration of the benefits derived from using estimated unsaturated soil property
functions (Fredlund 2000). Estimates of the unsaturated soil property function
are shown to provide a significant increase in the accuracy of the engineered
designs, for a nominal increase at the soil investigation and testing stage. The
accuracy of the output from an analysis depends strongly upon the indepen-
dent variable being computed.
There are different methods to evaluate the soil–water characteristic curve.
However, these methods are divided to two general branches: laboratory and
estimating methods. The researchers, with taking account of the general form
of SWCC, have suggested some equations to approximate these curves. Such
equations basically consist of two or three constant evaluated by making use
of either suction laboratory results in various water contents or statistical
relations based on other soil properties. According to existing difficulties in
evaluation of these curves experimentally and noticeable variability, the esti-
mation of such parameters has widely been used by many of researchers.
In current study, an estimating model has been proposed for cohesive
soils to evaluate the constant parameters of equation suggested by Van
Genuchten (1980) based on soil index parameters. These parameters are, how-
ever, the plasticity index, PI, and the percentage passing 200# sieve, W. Using
Fig. 1. Qualitative representation of the benefits derived from using estimated

unsaturated soil property functions
Prediction of Soil–Water Characteristic Curve 357
statistical approach on the collection experimental results, some equations

have been proposed for estimation of the mentioned constants parameters
and corresponding SWCC.
2 Definitions
The behavior of unsaturated soils is strongly related to the pore size geometry
and the pore size distribution. The soil–water characteristic curve is defined as
the relationship between water content and suction for the soil (Williams et al.
1983). In soil science, volumetric water content is most commonly used, and
in most of researches the soil–water characteristic curve has been suggested
as a relationship between volumetric water content and matric suction. The
volumetric water content is the ratio of the water volume in the soil to the
total volume as below:
Vw Vw Vv S·e
θ= = · =S·n= (1)
Vt Vv Vt 1+e
where θ = volumetric water content, Vw = water volume within the soil, Vt =
total volume of soil, Vv = volume of pores in soil, S = degree of saturation,
n = porosity ratio, and e = void ratio.
The suction may be either the matric suction of the soil, ua − uw , or total
suction (i.e., matric plus osmotic suction).
Figure 2 shows a typical plot of a soil–water characteristic curve for a
silty soil, along with some of its key characteristics. The air-entry value of
the soil is the matric suction where air starts to enter the largest pores in the
soil. The residual water content is the water content where a large suction
change is required to remove additional water from the soil. The main curve
shown in this figure is a desorption curve. The adsorption curve differs from
Fig. 2. Typical soil–water characteristic for a silty soil

Fig. 3. Soil-water characteristic curves for sandy soil, silty soil, and clayey soil
the desorption curve as a result of hysterics. The end point of the adsorption
curve may differ from the starting point of the desorption curve because of air
entrapment in the soil. Both curves have a similar form; however, this paper
considers the desorption curve.
Typical soil–water characteristic curves for different soils are shown in
Fig. 3. The saturated water content, θS , and the air-entry value, (ua − uw )b ,
generally increase with the plasticity of the soil. Other factors such as stress
history also affect the shape of the soil–water characteristic curves.
3 Methods of Obtaining SWCC
The options for obtaining SWCC can be generally classified as:

a) Experimental methods, based on fitting a curve through a number of points
obtained doing direct tests measure the pore water pressure in the soil on
the different water content conditions. The development of a SWCC for a
particular soil from suction measurements can require several tests and can
take some time to obtain all the necessary data. In order to facilitate the
efficient determination of the SWCC, several mathematical models have
been developed to describe the SWCC for a particular soil from just a few
points. On the other hand, many application of the SWCC require that
it be differentiated or integrated and be continuous. Some of the fitted
functions that have been proposed are summarized in Table 1. Generally,
the equations in Table 1 have been validated for certain soils and ranges in
suctions. The process of fitting experimental suction data to one of the pro-
posed equations requires a minimum number of experimentally obtained
suction measurements, depending upon the number of unknown parame-
ters in the chosen function.
Table 1. Some equations proposed for obtaining Soil–Water Characteristic Curve
Reference Parameter Description Equation
Assouline ψ: capillary head θw = θL + (θs −

θL
) η
et al. (1998) ψL : capillary head corresponds to a very × 1 − exp −ξ ψ 1
− 1 (2)
ψL
low water content, at which the hydraulic
conductivity is negligible
θL : volumetric water content at capillary
head ψL
ξ, η: fitting parameters
θS
Fredlund θw : volumetric water content θw = C(ψ) (3)
lnc [e+(ψ/a)b ]
and Xing a: a soil parameter which is a function of

(1994) the air entry value of the soil in kPa ln(1+ψ/ψr ) (4)
C(ψ) = 1 − ln(1+106 /ψr )
b: a soil parameter which is a function of
the rate of water extraction from the soil,
once the air entry value has been exceeded
c: a soil parameter which is a function of
the suction at which residual water con-
tent occurs in kPa
θS −θr
McKee θr : residual volumetric water content θ w = θr + 1+exp((ψ−a)/b)
(5)
and Bumb a: curve-fitting parameter
(1987) b: curve-fitting parameter
Williams et A: curve-fitting parameter ln ψ = A + B ln θw (6)
al. (1983) B: curve-fitting parameter
θS −θr
Van θr : residual volumetric water content θ w = θr + (7)
[1+(ψ/a)b ]C
Genuchten a, b, c: soil parameters, as described for
(1980) equation 3
θS −θr
Van θr : residual volumetric water content θ w = θr + (8)
Genuchten a: a soil parameter which is a function of [1+(ψ/a)bm ](1−1/bm )
(1980) the air entry value of the soil in kPa
bm : a soil parameter which controls the
slope at the inflection point in the SWCC
Farrel and α: empirical constant ψ = a exp[α(θS − θw )] (9)
Larson (ua − uw )b : air-entry value
(1972)
Brooks θr : residual volumetric water content θw = θr + (θS − θr )(ab /ψ)bb (10)
and Corey ab : bubbling pressure in kPa
(1964) bb : pore size index
θS −θr
Gardner θr : as above θ w = θr + (11)
1+(ψ/a)b
(1958) a, b: soil parameters, as described for
equation 3
Among these equations, equations proposed by Fredlund and Xing, Eq. (3)
and Eq. (4), and Van Genuchten, Eq. (7), seem to be fit and more accurate
compared to others. So, they have been studied and compared in this paper.
b) Predictional methods, based on the GSD and other soil properties. These
methods can be divided in three main categories:
• The first of these approaches is based upon statistical estimation of
water contents at selected matric suction values. These water contents,
at each suction value, are correlated to soil properties. This process
requires a regression analysis followed by a curve fitting procedure.
This approach has been followed by several researchers including Visser
(1969), Gupta and Larson (1979), Reddi and Poduri (1997).
• The second approach includes those methods that correlate, by regres-
sion analysis, soil properties with the fitting parameters of an analytical
equation of SWCC. This statistical approach has been followed by

many researchers such as Williams et al. (1983), Tomasella and Hod-
nett (1998). Zapata (1999) proposed relations to predict the constants
of Fredlund and Xing equation based on soil index properties. The
proposed option of this paper classified in this category.
• The third approach includes the methods that estimate the SWCC us-
ing a physics-based conceptual model. It involve physical models based
upon the conversion of the GSD (textural information) into a pore-size
distribution, which in turn is related to a distribution of water contents
and associated pore pressures. This approach was followed by Arya and
Paris (1981), Fredlund et al. (1997).
4 Proposed Model for Predicting the SWCC

The studies done by various investigators show that equations proposed by
Fredlund and Xing and Van Genuchten predict the SWCC with more con-
formity compared to experimental results while they do not contain too con-
stants. Using Fredlund equations, the water content of soil always intends to
zero at matric suction equal to 106 kPa; which it was not confirmed in all soils
and it is dubitable. In contrast, Van Genuchten equation seems to predict
the residual water content of soils more accurately. So in current study, Van
Genuchten equation was selected as the base and its fitting parameters (a, b,
and c) were correlated with index properties of fine soils with plasticity index,
PI, greater than zero. Using these relations, one can predict the SWCC with-
out doing expensive and time consuming tests. Neglecting the residual water
content at high matric suctions compared with the water content of soil, basic
equation proposed by Van Genuchten can be rewritten as:
1
S= c . (12)
[1 + (ψ/a)b ]
To establish the relations, a database characterizing approximately 60 soils
with PI greater than 2 was assembled from knowledge-based program devel-
oped by Soilvision Systems, Ltd as PhD dissertation of Zapata (1999).
For the fine soils, the product of the percentage passing the 200# sieve, as
a decimal, was multiplied by the PI as a percentage, to form the weighted PI,
W.PI. This parameter selected as the main index soil property for correlation.
The reasoning behind this choice is as follows:
The equilibrium soil suction at a given degree of saturation was expected
to be proportional to the specific surface area of the soil. The PI is a fair
indicator of surface area and the use of PI alone can be considered. However,
a soil with a small percentage of highly active clay would have a high PI but
only a moderate specific surface area. Therefore, the weighted PI (W.PI) was
considered a better index of soil particle surface area for predicting Soil–Water
Characteristic Curve.
Using the described database, and fitting Van Genuchten equation for
SWCC to the experimental results for each soil, the parameters a, b, and
c, were correlated with the weighted PI as the main index parameter. The
relations found are as follows:
a = 0.0015(W.PI)3 + 0.1028(W.PI)2 + 0.5871(W.PI) + 11.813 (13)

b = 0.00011(W.PI) − 0.01358(W.PI) + 1.76987
2
(14)
−6
c = −5 × 10 (W.PI) − 0.00014(W.PI) + 0.14745 .
2
(15)
The W.PI parameter in equations (13)–(15) is defined as:
W.PI = Passing#200 × PI (16)
where:
Passing 200# = Material passing the 200# U.S. Standard Sieve expressed as
a decimal
PI = Plasticity Index of Soil (%)
The SWCC is obtained using proposed model for various soils with W.PI
between 1 and 40 are shown in Fig. 4. As it is expected, increasing W.PI, the
air-entry value of suction increased.
Fig. 4. Predicted SWCC based on W.PI
5 Comparison
To assess the accuracy of proposed model in estimating the soil-water charac-
teristic curve, some comparisons have been made between laboratory suction
tests results with those of mentioned model for various soils. Figs. 5–8 illus-
trate the results for four different types of soil. The soil characteristics and
Van Genuchten equation parameters which have been computed with pro-
posed model equations are shown in Table 2.
Table 2. The soil characteristics and computed parameters
Constant Passing Plasticity Dry

Reference parameters 200# index density Soil Type
c b a (W) (%PI) (g/cm3 )
Escario and
0.129 1.459 167 0.98 31 1.33 Madrid Gray Clay
Juca (1989)
Escario and
0.145 1.627 35 0.83 14 1.80 Red Silty Clay
Juca (1989)
Escario and
0.147 1.756 12.5 0.13 8 1.91 Madrid Clayey Sand
Juca (1989)
Zapata (1999) 0.138 1.447 187 0.92 35 1.14 Fountain Hills Clay
As it can be seen in Figs. 5 and 6, the results of proposed model are in

a good agreement with those of laboratory tests for clayey soils. The average
variance in these samples is about 6%.
For sand sample, Fig. 7, particularly in low suctions, the convergence of
results would decrease; however, as it will be discussed later this would be
expected respect to the soil type (i.e., very low fine contents and W.PI about
1). Furthermore, comparison between the results of proposed model with those
of laboratory tests for Fountain Hills clay, Fig. 8, represents an insignificant
convergence in the results; however, this may be explained with respect to the
100%
Exprimental Data
Van Genuchten
80%
Degree of Saturation
60%
40%
20%
0%
1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06
Soil Suction , kPa
Fig. 5. Comparison between soil-water characteristic curves resulted from proposed

model with laboratory results for Madrid Gray clay
100%
Exprimental Data
Van Genuchten
80%
60%
40%
20%
0%
1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06
Soil Suction , kPa

model with laboratory results for Red Silty clay
100%
Exprimental Data
Van Genuchten
80%
60%
40%
20%
0%
1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06
Soil Suction , kPa

model with laboratory results for Madrid Clayey sand
100%
Exprimental Data
Van Genuchten
80%
60%
40%
20%
0%
1.E-05 1.E-04 1.E-03 1.E-02 1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06
Soil Suction , kPa

model with laboratory results for Fountain Hills clay
very slight dry density about 1.14 g/cm3 and the high plasticity index in these
materials. Also, the distribution of laboratory test results could represent the
specific conditions of this soil and furthermore the high probability of errors in
the empirical methods. So, current approximating model for clayey soils may
not have adequate accuracy for particular soils with special characteristics.
The empirical test results for some of soils that have been assessed in this
research and the resulting curves of proposed model are shown in Figs. 9–12.
In these figures, the points are representative of tests results and the curves
are results of the soil-water characteristic curve for proposed equations in
assessed domain. It should be noted that in present study, a database (60 sets
of various soil types) have been used; though, according to impossibility and
nonnecessity of total data proposal, just 20 selective data from all are shown
in mentioned figures. As it can be seen in Fig. 9, for soils with low plasticity
index (i.e., silts) and W.PI < 3, the existing variation between the proposed
model and the suction tests results are considerable that this may be because
of the more effect of aggregate distribution on the soil-water characteristic
curve in these materials.
For cohesive soils with 3 < W.PI < 30, Figs. 10 and 11, the proposed model
predicts the soil-water characteristic curves precisely. Increasing more in W.PI,
Fig. 12, the convergence between results decreases in suctions with values
greater than 105 kPa. However, this may be the result of existing difficulties
in high suction measurements.
100%
wPI = 0.1
wPI = 3
80%
60%
40%
20%
0%
1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06
Soil Suction , kPa
Fig. 9. The empirical and soil-water characteristic curves results for soils with
0.1 < W.PI < 3
6 Conclusions
According to extensive practical usage of water retaining curves and existing
difficulties in direct evaluation of these curves using laboratory tests, in this
100%
wPI = 3
wPI = 10
80%
60%
40%
20%
0%
1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06
Soil Suction , kPa
3 < W.PI < 10
100%
wPI = 10
wPI = 30
80%
60%
40%
20%
0%
1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06
Soil Suction , kPa
10 < W.PI < 30
100%
wPI = 30
wPI = 50
80%
60%
40%
20%
0%
1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06
Soil Suction , kPa
30 < W.PI < 50
paper, based on some index soil parameters, an approximating model for fine
soils has been developed. The laboratory tests have been shown that for cohe-
sive soils, product of the plasticity index, PI, in the percentage passing 200#
sieve, W , is an appropriate parameter for estimating the soil-water characteris-
tic curves. Therefore, investigating the empirical data of more than 60 various
types of soils, some relationships have been developed to approximate the Van
Genuchten equation parameters based on W.PI.
The main results concluded from this research are listed below:
• Increasing the parameter of W.PI, the air-entry value will be enhanced,
and the soil tendency for water retaining will be increased.
• Respect to existing difficulties in the suction laboratory tests and the vari-
ability during tests, the approximating procedures to obtain soil–water
characteristic curves by making use of some relations, similar to those
represented in this paper, would be proper.
• Comparisons between resulted curves obtained from proposed relations
with those of empirical tests represent a good agreement in results; al-
though, this may not exactly be correct for some especial soils (problematic
soils).
• For soils with W.PI < 3, the effect of this parameter on soil–water char-
acteristic curves and consequently on the proposed model accuracy is in-
significant; thus, other parameters effects such as soil classification would
be more considerable.
Acknowledgements
The authors would like to thank Claudia E. Zapata for supplying experimental
data and for helpful suggestions.
References
Arya LM, Paris JF (1981) A Physicoempirical Model to Predict the Soil Moisture
Characteristic from Particle-Size Distribution and Bulk Density, Soil Sci Soc Am
J 45(6):1023-1030
Assouline S, Tessier D, and Bruand A (1998) A Conceptual Model of the Soil Water
Retention Curve, Water Resource Research 34(2):223–231
Brooks RH, Corey AT (1964) Hydraulic Properties of Porous Media, Hydrology
Papers, 3. Fort Collins, Colorado State University
Escario V, Juca J (1989) Strength and deformation of partly saturated soils. In:
Proc 12th Int Conf Soil Mech Found Eng, Rio de Janerio, 3:43–46
Farrell D, Larson W (1972) Modeling the Pore Structure of Porous Media, Water
Resource Research 8(3):699–706
Fredlund DG (2000) The implementation of Unsaturated Soil mechanics into
geotechnical engineering, Can Geotech J 37:963–986
Fredlund DG, Raharjo H (1993) Soil Mechanics for Unsaturated Soils. Wiley Inter-
science
Fredlund DG, Xing A (1994) Equations for the Soil–Water Characteristic Curve,
Can Geotech J 31(3):521–532
Fredlund M, Fredlund D, Wilson G (1997) Prediction of the Soil–Water Character-
istic Curve from Grain–Size Distribution and Volume–Mass Properties. In: Proc
3rd Brazilian Symp on Unsaturated Soils, Rio de Janerio.
Gardner WR (1958) Some Steady State Solutions of the Unsaturated Moisture Flow
Equation with Application to Evaporation from a Water-Table, Soil Science
85(4):228–232
Gupta SC, Larson WE (1979) Estimating Soil Water Retention Characteristics from
Particle Size Distribution, Organic Matter Percent and Bulk Density, Water
Resources Research 15(6):325–339
McKee CR, Bumb AC (1987) Flow-testing coalbed methane production wells in the
presence of water and gas. In: SPE Formation Evaluation, December, pp 599–608
Reddi LN, Poduri R (1997) Use of Liquid Limit State to Generalize Water Retention
Properties of Fine-Grained Soils, Geotechnique 47(5):1043–1049
Soil Vision User’s Guide (Version 1.2) (1997) [Computer Software], Saskatoon,
Saskatchewan, Canada, Soil Vision System, Ltd
Tomasella J, Hodnett MG (1998) Estimating Soil Water Retention Characteristics
from Limited Data in Brazilian Amazonia, Soil Science, 163(3):190–202
Vanapalli SK, Fredlund DG, Pufahl DE (1996) The Relationship Between the Soil–
Water Characteristic Curve and the Unsaturated Shear Strength of a Compacted
Glacial Till, Geotechnical Testing J 19(3):259–268
Van Genuchten MT (1980) A Closed-Form Equation for Predicting the Hydraulic
Conductivity of Unsaturated Soils, Soil Sci Soc Am J 44:892–898
Visser WC (1969) An Empirical Expression for the Desorption Curve, Water in the
Unsaturated Zone:329–335
Williams J et al. (1983) The Influence of Texture, Structure and Clay Mineralogy
on the Soil Moisture Characteristic, Australian J Soil Res 21:15–23
Zapata CE (1999) Uncertainty in Soil–Water Characteristic Curve and Impact on
Unsaturated Shear Strength Predictions. Ph.D. Thesis, Arizona State University,
Tempe, United States
Zapata CE, Houston WM, Houston SL, Walsh KD (2000) Soil-water characteristic
curve variability. In: Shackleford CD, Houston SL, Chang N-Y (eds) Advances
in Unsaturated Soils, Geotechnical Special Publication No. 99, Reston, ASCE:
84–124
Water Balance and Effectiveness of Mineral
Landfill Covers – Results of Large Lysimeter
Test-Fields
Wolf Ulrich Henken-Mellies
LGA – Geotechnical Institute, Tillystr. 2, D-90431 Nuremberg, Germany

wolf-ulrich.henken-mellies@lga.de
Summary. Landfills are provided with cover systems in order to separate the
waste from the environment and in order to prevent percolation of precipitation
and recharge of leachate. The layers of a landfill cover system (restoration profile,
drainage layer, compacted clay liner – CCL) have to be specifically designed for
long-lasting effectiveness. Special care has to be taken to prevent the CCL from
desiccation-cracking due to high matric suction. On a landfill in Northern Bavaria,
test fields of some 260 m2 each were installed with different profiles: The test fields
were built according to the recent German landfill ordinance, but with a thicker
restoration profile (2.0 m and 1.5 m thick restoration layer, respectively). Water
content and matric suction of the soil profiles was measured by FDR-probes and
tensiometers. In test field 1 reduction of soil water content during summer was de-
tected down to 1.8 m depth. In test field 2 (1.5 m thick restoration layer) decreasing
water content was measured in the CCL, whereas in test field 1 (2.0 m thick restora-
tion layer) the CCL was not affected by loss of water. In conclusion, at the given
landfill site only the 2.0 m thick restoration profile is able to effectively shield the
CCL from desiccation effects.
Key words: landfill cover, compacted clay liner, lysimeter test field, percolation,
matric suction, soil water content
Introduction
Upon completion solid waste landfills are provided with a final cover in order
to prevent the migration of contaminants. The cover has to meet two basic
requirements: It should be at least as tight as the lining system at the base
of the landfill and its ability to minimize percolation should not deteriorate
with time.
Landfill covers are exposed to highly variable hydraulic conditions due
to seasonally changing weather conditions and influences of vegetation. The
upper meters of the pedosphere are dominated by unsaturated conditions.
Key element of a landfill cover system is the sealing layer. Most commonly
370 Wolf Ulrich Henken-Mellies
compacted clay liners are used as sealing layers in Germany, either in com-
bination with a geomembrane or as the only sealing element. Since mineral
liners are quite common in landfill cover systems and their proper functioning
has to be relied upon for long term use, the problem of preventing desiccation
and shrinkage cracks of clayey mineral liners has been paid much attention.
The physical phenomena which lead to shrinkage cracks have been studied by
laboratory experiments on different soils. Theoretical approaches have been
developed, and evidence of desiccation or of the successful prevention of desic-
cation has been obtained by a number of field trials (Ramke et al. 2002, Witt
and Zeh 2004).
This paper presents a field testing program, and intermediate results of
ongoing research aimed at a better understanding of the performance of the
restoration profile and at specifying its properties required for the prevention
of desiccation of the compacted clay liner within the cover system of a sanitary
landfill under site specific conditions.

Field Site
The field site for this study is situated near the village of Aurach in Northern
Bavaria, some 60 km southwest of the city of Nuremberg in a gently hilly
region at an elevation of 500 m above sea level. Average annual precipitation
at the site is 750 mm, mean annual potential evapotranspiration amounts to
550 mm. Average annual temperature is 9◦ C. The lysimeter test fields were
erected on the slope of a municipal waste landfill, inclined at 14◦ , exposed
towards the south. Vegetation at the site is mainly grass.
Profiles of the Lysimeter Test-Fields

The test fields were placed inside large lysimeters (13 × 20 m) made of PEHD-
geomembranes and drainage composites, allowing collection of surface runoff,
lateral drainage and percolation.
The test fields consist of a three-component cover system including a soil
cover, a drainage layer and a compacted clay liner. The soil cover consists of
locally available sandy loam, but instead of compacting it, care was taken to
construct it with little compaction. Thus the soil cover should act as water
storage layer. The two profiles of test-field 1 and 2 differ in the thickness of
the subsoil layer of the restoration profile, which is 1.8 m in field 1 (thickness
of the entire restoration profile of 2.0 m), and 1.3 m in field 2 (thickness of
the entire restoration profile of 1.5 m). The drainage layer with a thickness of
0.3 m is made of gravel. The compacted clay liner consists of silt loam (25%
clay/50% silt/25% sand; plasticity index 21%). It was constructed in two lifts
of 0.25 m each, compacted to > 98% Proctor-density at water content 1–2%
below optimum.
Water Balance and Effectiveness of Mineral Landfill Covers 371
Fig. 1. Profiles of the lysimeter test fields (left: field 1; right: field 2) at Aurach
landfill
Table 1. Properties of soil layers of the test fields
Saturated hydraulic Plastic limit;

Layer Thickness Material
conductivity Liquid limit
Top soil and 2.0 m (#1) sandy loam 10−6 to 10−7 m/s
storage layer 1.5 m (#2)
Drainage layer 0.3 m coarse grave > 10−2 m/s
Compacted clay 0.5 m silt loam 5 × 10−10 to 20%; 41%
barrier 2 × 10−9 m/s
Data Collection
Both lysimeter test fields are equipped with tipping buckets to measure
• Surface Runoff,
• Drainage flow in the drainage layer above the compacted clay liner,
• Percolation through the compacted clay liner.
Matric suction of soil at different depths is measured by tensiometers (pres-
sure transducer tensiometer, type “T 6” by UMS, Munich). Volumetric water
content of soil at different depths is measured by FDR-probes (type “Theta

Probe ML2” by Delta-T devices, Cambridge). Temperature of soil at different
depths is recorded by thermocoupes. Precipitation and air temperature are
recorded at the test field site. Readings of all measuring devices are taken
automatically. The data are stored locally on PC and are transferred to the
office by GSM.
Precipitation and Flow Measurements from Lysimeter
Data from lysimeter 1 for the time period from May 2002 until December 2005
are plotted in Fig. 2. Note that the scale of precipitation on the left side is
four times the scale of flows on the right side of the plot. In Central Europe,
commonly the precipitation is distributed rather evenly over the year, so the
cumulative curve of precipitation (thin black line) shows an almost linear
increase.
Surface run-off (dashed line) is generally low. In the long run it plays no
important role. Among the measured flows, the discharge from the drainage
layer (thick grey curve) has the greatest share. Drainage occurs mainly during
winter months from November to April. During summer, the precipitation is
3000 750
Precipitation
2500 Drainage flow 625
Surface runoff
Cumulative precipitation [mm]
Percolation through CCL

Precipitation
Cumulative flows [mm]
2000 500
Drainage flow
1500 375
1000 250
500 125
Surface runoff
Percolation through CCL

0 0
Apr-02 Oct-02 Apr-03 Oct-03 Apr-04 Oct-04 Apr-05 Oct-05 Apr-06
Fig. 2. Precipitation and measured flows at lysimeter test-field 1

partly consumed by evapotranspiration directly and partly stored in the soil

layers of the restoration profile for later evapotranspiration, so that there is
no drainage flow for a period of several months every year.
Percolation through the compacted clay liner (bold solid curve) occurs
simultaneously with lateral drainage. It amounts to about one tenth of the
drainage flow. In the water balance of the landfill cover percolation through
the CCL amounts to 3% of the total precipitation.
In some other test fields where desiccation of the CCL or GCL took place,
the amount of seepage increased and the efficiency of the seals decreased with
time (e.g. Melchior 2002). Therefore it is important to take a closer look at the
processes which are going on within the layers of unsaturated soil, of which
the landfill cover system is composed.
Moisture Content and Matric Suction
Measurements of matric suction (tensiometer measurements) and volumetric

moisture content FDR-probe measurements) at different depths within the
soil profile are shown in Fig. 3 (for test field 1) and Fig. 4 (for test field 2).
The bold lines in both plots represent the probes, which are installed within
the CCL (at 2.4 m and 2.6 m in test field 1 and 1.9 m and 2.1 m in test field 2,
respectively). The thin lines represent the probes within the storage layer.
During winter season the moisture content of the entire soil profile is gen-
erally high (θ = 0.35 to 0.4 in the storage layer and 0.4 to 0.45 in the CCL)
and at the same time matric suction is low (generally below 100 hPa). During
summer evapotranspiration is usually higher than precipitation and the veg-
etation extracts water from the soil. The plots show the different response of
the soil to the relatively wet summer of 2002 in contrast to the dry summer
of 2003:
In 2002 only the tensiometer at 1.0 m depth recorded rapidly increasing
matric suction; at greater depths matric suction remained below 100 hPa dur-
ing summer and fall of 2002. Volumetric moisture content decreased in the
upper meter of the soil; below that it almost remained constant. In the dry
summer of 2003 the FDR-probes recorded a strong decline in soil water con-
tent even at 1.4 and 1.8 m depth. At the same time the tensiometers even at
the base of the storage layer recorded an increase of matric suction beyond
the measurement range of 700 hPa.
The bold lines represent the measurements within the CCL. In 2002 matric
suction and water content showed almost no change at all. During summer
of 2003 there was a slight increase in matric suction up to 150–200 hPa and
a 1% drop in water content. In comparison with the measurements in the
storage layer the diagram shows, that there is an effective hydraulic decoupling
between storage layer and CCL.
Figure 4 shows the progress lines of matric suction in the CCL of test
field 2 (upper plot) and volumetric water content in the storage layer and in
the CCL (lower plot). Again, no critical decrease in water content/increase in
800
Tensiometer-measurements
700
1.0 m
600
Matric Suction [hPa]
1.4 m
1.8 m
500
2.4 m (CCL)
400 2.6 m (CCL)
300
200
100
0
-100
0.6
Volumetric water content [m³/m³]
FDR-Probe measurements FDR-probe in CCL (2.4 m)

FDR-probe in CCL (2.6 m)
0.5
0.4
0.3
FDR-probes in storage layer

0.2 FDR 1 ( 0.2 m)
FDR 2 (0.6 m)
0.1 FDR 3 (1.0 m)
FDR 4 (1.4 m)
FDR 5 (1.8 m)
0
4/1/02 7/1/02 10/1/02 1/1/03 4/1/03 7/1/03 10/1/03 1/1/04
Fig. 3. Results of tensiometer measurements (upper plot) and FDR-probe measure-

ments (lower plot) at lysimeter test-field 1 for the years 2002 to 2003
matric suction occurred in 2002. But in summer of 2003 water content of the
CCL dropped below 0.4 m3 /m3 and at the same time matric suction increased
to about 500 hPa.
According to Witt and Zeh (2004) values of matric suction above 250 to
500 hPa are critical with respect to the formation of desiccation cracks in
CCLs. The mainly silty material with low to medium plasticity used in this
study is less susceptible to cracking than high plasticity clays. The ongoing
measurements have not shown any increases of percolation in lysimeter 2
compared to test field 1.
Conclusions
Landfill cover profiles commonly consist of a series of unsaturated soil layers,
which are exposed to the dynamics of weather and vegetation. Regulations of
EU and Germany recommend a compacted clay layer as sealing component
within the design of a landfill cover. In temperate humid climate like in Ger-
many there are periods (generally during winter months) where precipitation
800
700 Tensiometer-measurements
Matric Suction [hPa]
600
1.9 m (CCL)
500 2.1 m (CCL)
400
300
200
100
0
-100
0.6
volumetric water content [m³/m³]
FDR-Probe measurements FDR-probe in CCL (1.9 m)

0.5 FDR-probe in CCL (2.1 m)
0.4
0.3
FDR-probes in storage layer

0.2 FDR 1 (0.2 m)
FDR 2 (0.6 m)
0.1 FDR 3 (1.0 m)
FDR 4 (1.4 m)
0
4/1/02 7/1/02 10/1/02 1/1/03 4/1/03 7/1/03 10/1/03 1/1/04
Fig. 4. Results of tensiometer measurements (upper plot) and FDR-probe measure-

ments (lower plot) at lysimeter test-field 2 for the years 2002 to 2003
exceeds evapotranspiration. In these times low hydraulic conductivity of the

CCL is the determining factor for low percolation rates.
In contrast, there are periods during summer and fall, where evapotran-
spiration exceeds precipitation. The lysimeter measurements show that the
periods without drainage flow may last up to nine months. During these pe-
riods of time the CCL has to be protected against critical loss of water. This
task has to be fulfilled by the top soil or water storage layer. The in-situ-
measurements in the soil profiles show an increase in matric suction and at
the same time a decrease in water content of the storage layer down to its base
at 2.0 m and 1.5 m, respectively. The drainage layer acts as a barrier, which
inhibits the capillary rise of water from the CCL to the top soil. In test field 2
with a 1.5 m thick storage layer a reduction of water content of the CCL was
detected, which can be attributed to vapour transport through the drainage
layer. The temporary reduction of water content did not yet have an adverse
effect on the sealing effectiveness of the CCL.
The thickness of the top soil cover should be “> 1 m” according to German
and EU regulations. In many cases the actual thickness of the top soil cover
is even lower. The results reported here emphasize the importance of a well
designed landfill cover including a thick water storage layer in order to protect
the CCL against desiccation.
References
Melchior S (2002) Field studies and excavations of geosynthetic clay barriers in
landfill covers. In: Zanzinger et al. (eds) Clay Geosynthetic Barriers. Balkema,
Lisse, pp 321–330
Ramke H-G, Gartung E, Heibrock G, Lükewille W, Melchior S, Vielhaber B,
Bohne K, Maier-Harth U, Witt K-J (2002) Austrocknungsverhalten mineralis-
cher Abdichtungsschichten in Deponie-Oberflächenabdichtungssystemen. Höx-
teraner Berichte zu angewandten Umweltwissenschaften, Vol. 03, Höxter
Witt K-J, Zeh R (2004) Maßnahmen gegen Trockenrisse in mineralischen Abdich-
tungen. In: Kranert (ed) Stuttgarter Berichte zur Abfallwirtschaft 81:83–98
A Retention Curve Prediction for Unsaturated
Clay Soils
Mehrez Jamei, H. Guiras, and N. Mokni
Civil Engineering Laboratory, National Engineering School of Tunis (ENIT)

mehrez.jamei@enit.rnu.tn
Summary. In order to obtain the retention curve of unsaturated soil which is de-
fined as the relationship between water content and suction, numerous laboratory
tests have been done for many soils with different clay contents. The laboratory tests
concern different types of soils; the swelling and not swelling clays (natural clay with
high silt content and the bentonite clay). Two complementary experimental ways
are conducted. The first which leads to measure suction is based on the filter paper
method. The second with the dessicator allows to impose a high suction values.
The paper discusses the theoretical approaches for the grain-size distribution
and water retention curve’s relation. It discusses also their correlations according to
each type of tested soil. The experimental data are used to validate the theoretical
model selected in the study. The influence of the different parameters used by the
model is also studied.
All the parameters required to run the model can be obtained by performing
independent, common laboratory tests and are related to the physical parameters
like the initial dry density and the initial void ratio.
After some numerical tests, the direct procedure to obtain the soil-water char-
acteristic curve is improved.
Key words: water retention curve, grain-size distribution, laboratory tests, model,
prediction
1 Introduction
The measurement of soil parameters, such us the permeability and shear
strength functions, used to describe unsaturated soil behaviour can be ex-
pensive, difficult and time consuming. Hence, several models were proposed
to predict the hydro-mechanical behaviour of unsaturated soils using soil wa-
ter characteristic curve (SWCC).
SWCC can be easily estimated using grain size distribution curve. Several
methods have previously been proposed for the estimation of the SWCC (Arya
and Paris 1981, Haverkamp and Parlange 1986, Fredlund et al. 1997). In the
378 M. Jamei et al.
estimation’s methods the effect of swelling was not been addressed. However,
swelling compacted clays receive increasing attention as soils which lead to
many stability structures problem (shallow foundations, slope stability, etc.).
In special applications, some typical clays and bentonite are used as technical
buffer materials in geotechnical barriers in order to isolate waste (Börgesson
1985). Therefore, the knowledge of SWCC of swelling soils is of a special
interest.
The aim of this paper is to test the Arya and Paris’s model for the es-
timation of the SWCC on swelling soil. The shape of the estimated curve is
controlled by the grain size distribution and influenced by the density of the
soil. The influence of initial water content, soil structure and stress history,
on the SWCC curve was also studied by Vanapalli et al. (1999). It appears
that SWCC curve is influenced by initial water content and stress history for
the specimens compacted at dry and optimum conditions. However, this influ-
ence is not significantly for both high drying conditions (for the high suction
ranges) and wet of optimum conditions (for the specimens compacted with
wet of optimum conditions).
Unimodal and bimodal models proposed by Fredlund et al. (2000) were
used to best fit grain–size distribution data. The fitting method provides a
continuous fit of the entire grain size distribution curve including the coarse
and fine extremes (Fredlund et al. 2000).
The unimodal form provides methods for accurately representing uniform
and well-graded soils.
The equation for the unimodal curve is given as follows:
⎡ ⎡ ⎤
dr
⎤7
1 ⎢ ln 1 + d
⎦ ⎥
Pp (d) = ga ga gm ⎣1 − ⎣ ⎦ (1)
ln exp(1) + d dr
ln 1 + dm
where:
• Pp (d) = percentage by mass, of particles passing a particular size;
• ga = parameter designating the inflection point on the curve an is related
to initial breaking point on the curve;
• gm = parameter related to the steepest slope on the curve;
• gn = parameter related to the shape of the curve as it approaches the fines
region;
• dm = diameter of the minimum allowable size particle (mm);
• dr = parameter related to the amount of fines in a soil;
• d = diameter of any particle size under consideration (mm).
The five parameters in unimodal equation provide a closer fit than previous
two-parameter and log-normal equations (Sang II Hwang et al. 2002). It also
improves representation of silt and clay soils (Fredlund et al. 2000).
When the soils are gap-grated, it is necessary to consider the use of bimodal
equation witch is given as follows:
A Retention Curve Prediction for Unsaturated Clay Soils 379
⎧ ⎡ ⎤⎫
⎪ ⎪
⎨ 1−w ⎬
w ⎢ ⎥
Pp (d) = a nbi mbi + ⎣
kbi bi ⎪ ⎦
⎪
⎩ ln exp(1) + dbi
l
⎭
ln exp(1) + jdbi
⎧ ⎡ ⎫
⎪
⎨ ⎤7 ⎪
⎬
ln 1 + drbi
× 1−⎣ d ⎦
(2)
⎪
⎩ ln 1 + ddrbi ⎪
⎭
m
where:
• abi = parameter relates to the initial breaking point along the curve;
• nbi = parameter related to the steepest slope along the curve;
• mbi = parameter related to the shape of the curve;
• jbi = parameter related to the second breaking point of the curve;
• kbi = parameter related to the second steep slope along the curve;
• lbi = parameter related to the second shape along the curve;
• drbi = parameter related to the amount of fines in a soil.
1.1 Description of the Model
Arya and Paris (1981) presented the first physico-empirical methods to esti-
mate the SWCC. The basic information of the model is the grain size distribu-
tion curve. Volumetric water content was calculated basing on an estimation
of the pore sizes of the soil. pore radius estimation was based on the assump-
tion of spherical particles and cylindrical pores using the capillary theory. An
empirical factor is used to take into account uncertainties in the estimation.
The principle employed to deduce a distribution of pores from a grain
size distribution is as follows: the finer particles are the more their stacking
spares narrow voids, with each N classes of size of particles (N fractions), a
pore is associated (N pores on the whole). The fraction of size Ri is supposed
to contain spherical particles. The associated cylindrical pore is supposed to
follow the edge of the juxtaposed particles and its length is li = Ni Riα (α is
equal to 1 for a rectilinear pore and it is a parameter to be estimated for a
natural geometry).
2Ri
li = Ni Riα , α=1
The capillary model used is composed by N independent capillary tubes,

from which N couples of values of suction (S), and water content (W ) are
380 M. Jamei et al.
calculated. The parameter α is determined by calibration with the experimen-

tal data. It may be related to the physical soil parameter like plasticity index
(Ip ).
1.2 Formulation of the Model
The grain size distribution was divided into N fractions. Individual weight
fractions can then be calculated as follows
Pi = (gi+1 − gi )γt (3)
where:
• Pi = weight of individual fraction (kg);
• gi = function representing percent passing versus particle diameter (i =
1, 2, . . . , N );
• N = number of fractions into which grain-size distribution is divided;
• γt = total density of the soil sample (kg/m3 ).
The calculation of the void ratio is possible once the void ratio is known
Pi
Vvi = ×e (4)
Gs × γw
where
• e = void ratio;
• ρw = density of water;
• Gs = specific gravity of the soil.
The sum of all void ratios can be calculated as follows
'
n
Vv = Vvi . (5)
i=1
Hence volumetric water content is

(
i
Vvj
j=1
θvi = . (6)
VT
The pore radius is calculated according to the following equation:
12
(1−α)
4Ni
ri = Ri (7)
6
where:
• ri = mean pore radius;
• Ri = mean particle radius;

• Ni = the number of spherical particles in the i-th particle-range;
• α = an empirical factor and greater than 1.
Pore radius and soil water pressure:
2TS cos θ
si = (8)
ρw × g × ri
where:
• si = soil water pressure;
• Ts = surface tension of water;
• θ = contact angle;
• ρw = density of water;
• g = acceleration due to gravity, and
• ri = pore radius.
2 Test Program, Results and Analysis
The soils tested are sandy loam, silt loam, and bentonite reconstituted by
compaction in the laboratory. The physical and mineralogical properties of
tested samples are presented in Table 1.
Table 1. Physical and mineralogical characteristics of soil
Sample designation Sandy loam Silty loam Bentonite

Specific gravity (kN/m )
3
26.8 27.2 27.4
Liquid limit Wl (%) 28.5 51 375
Plasticity index Ip (%) 13.4 26 312
Initial water content Wi (%)
14.5 21 30.2
(Optimum Condition)
Dry density γd (%) (Optimum Condition) 18 16.4 12.2
Specific surface SST (m2 /g) — 124 504
2.1 Soil Water Characteristic Curve
Two complementary experimental ways are conducted. The first which leads
to measure suction is based on the filter paper method. The second with the
dessicator allows to impose a high suction values.
The filter paper method for determining soil water suction consists of plac-
ing a small piece of filter paper in direct contact with the soil. Equilibration of
moisture is allowed and the filter paper water content, through calibration, is
382 M. Jamei et al.
then related to the soil matric suction. With this method, it is assumed that
the salt can move into the filter paper along with the water provided there
is adequate moisture and that an osmotic gradient between the buried filter
paper and the soil moisture is absent or negligible.
The samples were compacted in a static way under the optimal conditions
to have a 30 mm height. We then cut out cubes of 30 mm on sides. The inserted
filter papers are initially wet. Once hydrous balance is reached (one month),
the samples are weighed, plunged in mercury to determine final volume.
The second method consists in placing small soil cubes of side of 20 mm in
a closed enclosure. The relative humidity of the system is imposed by a saline
solution. Seven desiccators was prepared, in which was poured one litter of
saturated salt solution, and then the samples are placed on the perforated
plate. The desiccators are then closed and preserved until hydrous equilibrium
is reached. The materials used are compacted in a static way to the Proctor
Optimum. Table 2 indicates the salt solution used and corresponding suction.
Table 2. Salt solution

Suction (kPa) Salt solution
2818 CuSO4 ,5H2 O
7079 KNO3
12600 ZnSO4 ,7H2 O
22400 KCl
33100 NaCl
56200 NaNO2
158500 CaCl2 ,6H2
2.2 Free-Swell Odometer System

These tests were carried out according to standard procedure on swelling soil
(XP P94–090–1). An odometric cell was used. The sample was initially putted
in contact with water. The vertical deformation is measured until stabilization
of swelling. The value of the vertical stress necessary to bring back the sample
to its initial height (vertical swelling stress) is then measured. The test was
performed on a sandy loam for various initial conditions (Fig. 1), and on
bentonite under Proctor optimum conditions (Fig. 2). Sandy loam is a non
swelling soil.
In the case of bentonite, the stabilization of final swelling is reached after
25 days (Fig. 1). The rate of final swelling exceeds 100%.
For the silt loam compacted under optimal condition, the stabilization of
swelling was reached after two days. For low water content the steady state
was reached after three days. The rate of final swelling is more important in
the case of low initial water content and high dry density. This result was
shown by several authors (Guiras 1996).
–140
–120
–100
dH/H0 (%) –80

–60
–40
–20
0
0 5 10 15 20 25 30
Time (day)
Table of symbols
Wi (%)
Symbol Δh/h0 (%)
γd (kN/m3 )
30.2 120
Proctor optimum 12.2
Fig. 1. Times versus swelling: bentonite
–18
–16
–14
–12
dh/h (%)
–10
–8
–6
–4
–2
0
0 2 4 6 8 10 12
Time (day)
Fig. 2. Times versus swelling: silt loam
3 Prediction of the Experimental Results
The model was tested on the studied soils. The comparisons between the ex-
perimental and predicted data are shown in Figs. 3–5. The estimation of a
SWCC from grain–size distribution has been attempted for all soil types. The
estimated still appear to be quite reasonable in the case of swelling soil (silt
384 M. Jamei et al.
30
25
20
W (%) 15 Modelisation
10 Experimental point
5
0
1 10 100 1000 10000 100000
Suction (kPa)
Fig. 3. Comparison of experimental and estimated SWCC for sandy loam
35
30
25
W (%)
20 Modelisation
15 Experimental point
10
5
0
1 10 100 1000 10000 100000
Suction (kPa)
Fig. 4. Comparison of experimental and estimated SWCC for silt loam
40
30
W (%)
Modelisation
20
Experimental point
10
0
1 10 100 1000 10000 100000
Suction (kPa)
Fig. 5. Comparison of experimental and estimated SWCC for bentonite
Table 3. Estimated α value

Texture α values
Bentonite (clay) 1.57
Silt loam 1.38
Sandy loam 1.60
Table 4. α values proposed by Arya et al. (1999)
Texture α values
Clay 1.16
Silt loam 1.15
Sandy loam 1.459
loam and bentonite) and for the non swelling soil like sandy loam. Neverthe-
less, an estimation of parameter α is necessary to define the natural length of
pores. The estimated α value’s are shown in Table 3.
The estimated values of α, are different compared to that ones proposed
by Arya et al. (1999) for soils with same textural classes (Table 4).
The estimated values performed on swelling soil are higher than the pre-
vious values performed on non swelling soil.
The result shows that the Arya and Paris’s model provide a good estima-
tion of SWCC of swelling soils if a reasonable estimation of parameter α is
made.
4 Conclusion
Estimation of the soil- water characteristic curve from grain size distribu-
tion was found to be reasonably reliable for non swelling soil (sand loam)
and for swelling soil (silt loam and bentonite). The prediction model of the
SWCC curve basing on the grain size distribution curve is well established
and it seems that the unimodal equation can be used with success for clay
and swelling clay.
The first analytical adjustment is an important step to previous well the
SWCC curve.
It seems also, that the Arya and Paris’s model predicts well the SWCC
curve not only for the sandy and silty soils but also for swelling clays. However,
the α parameter must be related to a physical parameter like the plasticity
index. Hence, it appears that the lack of complete experimental data to verify
the models for the swelling clays remains greatest task at hand.
References
Arya LM, Paris JF (1981) A physicoempirical model to predict soil moisture char-
acteristic from particle-size distribution and bulk density data, Soil Sci Soc Am
J 45:1023–1030
Arya LM, Leiji FJ, Van Genuchten MTh, Shouse PJ (1999) Scaling parameter to
predict the soil water characteristic from particle size distribution data, Soil Sci
Soc Am J 63:510–519
386 M. Jamei et al.
Fredlund MD, Wilson GW, Fredlund DG (1997) Prediction of the soil water char-
acteristic curve from grain-size distribution and volume mass properties. In: 3rd
Brazilian symposium on unsaturated soil, Rio de Janeiro, Brazil, April 22–25,
1977
Fredlund MD, Wilson GW, Fredlund DG (2000) An equation to represent grain size
distribution, Can Geotech J 37:817–827
Guiras H (1996) Déformabilité des sols argileux non saturés : études expérimentales
et application à la modélisation. Thèse de doctorat, Ecole Nationale Supérieure
de Géologie de Nancy
Vanapalli SK, Fredlund DG, Pufahl DE (1999) The influence of soil structure and
stress history on the soil–water characteristics of a compacted till, Geotechnique
49(2):143–159
Unsaturated-Zone Leaching
and Saturated-Zone Mixing Model
in Heterogeneous Layers
Samuel S. Lee
D2SI San Francisco Regional Office, Federal Energy Regulatory Commission,

San Francisco, CA, USA samuel.lee@ferc.gov
Summary. A screening level model was developed for simulation of pollutant mi-
gration through the unsaturated-zone and subsequent mixing within the saturated-
zone. This one-dimensional finite difference model simulates the transport processes
of liquid-phase advection, liquid- and vapor-phase dispersion, sorption, and decay of
the contaminant. The model estimates contaminant concentration in the saturated-
zone by using a simple mass balance technique for mixing of the unsaturated-zone
leachate with the groundwater. The model can be a useful tool in making prelim-
inary assessments of the potential impacts of contaminants in the subsurface. The
model can handle vertical heterogeneity of the soil columns and non-uniform initial
contaminant concentration. It was verified by comparing its simulation results to an
analytical solution and laboratory soil column experiments. In addition the model
was validated against laboratory experiments with three different soil sample sizes
of Ottawa quartz sand and 480 ppm saline water as a groundwater contamination
source.
Key words: unsaturated-zone, saturated-zone, leaching model, groundwater, mod-

elling, GUI
1 Introduction
Although several computer codes, such as VLEACH, VLEACHSM 2.0, EPACML, etc.
can incorporate the heterogeneity of the soil properties often many sites do not
have the sufficient degree of details of field-measured data to allow the use of
the existing models. Even when there is a site with a reasonable amount of field
data, a rigorous parameter estimation and calibration work is often necessary
before conducting a comprehensive simulation. This manuscript describes a
newly developed model titled Unsaturated-Zone Leaching and Saturated-Zone
Mixing Model. The model allows sufficient model simulations and estimation
of contaminant concentration with a smaller degree of site sampling, analysis,
and parameter estimation than the existing models. It accomplishes it by
388 Samuel S. Lee
employing a one-dimensional (1-D) finite difference scheme for solving the

unsaturated-zone transport equation and uses a mass-balanced technique to
estimate contaminant concentration in the saturated zone.
Ravi and Johnson (1993) developed a 1-D transport program, titled
VLEACH, which handles only vertical migration of pollutant in a homogeneous
soil column. Later, Lee (1996, 1999a), and Lee and In (2005) developed sev-
eral versions of VLEACHSM by adding the liquid-phase dispersion, decay terms
in the unsaturated zone, and the saturation zone mixing into VLEACH. This
Unsaturated-Zone Leaching and Saturated-Zone Mixing Model is further im-
proved by implementing the heterogeneous soil property and a Graphic User
Interface. In addition, the Unsaturated-Zone Leaching and Saturated-Zone
Mixing Model allows the specification of two different types (Dirichlet’s and
Cauchy’s) of boundary conditions at the top of the soil column. Using a simple
mass-balance technique, the saturated zone module estimates the concentra-
tion of contaminants by mixing of leachate from the unsaturated zone with
groundwater.
2 Governing Equation and Boundary and Initial

Conditions
2.1 Unsaturated Zone Transport
Considering the three equilibrium phases of pollutants in an unsaturated soil

column, its 1-D governing transport equation can be expressed as follows:

∂Cw ∂Ca ∂Cs ∂ ∂Cw ∂ ∂Ca
θw + θa + ρb = θw Cw + θa Ca
∂t ∂t ∂t ∂z ∂z ∂z ∂z
∂
− (qw Cw ) − μw θw Cw − μa θa Ca − μs ρb Cs (1)
∂z
where, Cw denotes the concentration of a contaminant in liquid (water)
phase (mg/L), Ca denotes the concentration of a contaminant in vapor (air)
phase (mg/L), Cs denotes the concentration of a contaminant in solid phase
(mg/kg), θa denotes the volumetric water content (volume of water/total vol-
ume) (m3 /m3 ), θa denotes the air-filled porosity (volume of air/total vol-
ume) (m3 /m3 ), qw denotes water flow velocity (recharge rate) (m/yr), Dw
denotes dispersion coefficient for the liquid phase contaminant in the pore
water (m2 /yr). Da denotes gaseous phase diffusion coefficient in the pore air
(m2 /yr), μw denotes first order decay rate of a contaminant in water phase
(l/yr), μa denotes first order decay rate of a contaminant in gaseous phase
(l/yr), μs denotes first order decay rate of a contaminant in solid phase (l/yr),
ρb denotes bulk density of the soil (gr/cm3 ), z denotes vertical coordinate
with positive being downward, and t denotes elapsed time. Several assump-
tions were considered in the development of (1) as follow: The total porosity
Unsaturated-Zone Leaching and Saturated-Zone Mixing Model 389
(n) equals the sum of the water filled porosity the air filled porosity. The air
flow velocity (qa ) is assumed to be zero. For simplicity, it is also assumed that
μw = μa = μs = μ. Instantaneous equilibrium (partitioning) of the contami-
nant among the phases according to the following linear relationships:
• Liquid-solid phase equilibrium is
Cs = Kd Cw . (2)
• Liquid-gas phase equilibrium is
Ca = HCw . (3)
In (2) and (3), Kd (ml/g) denotes the distribution coefficient between the
solid phase and liquid phase, and H (dimensionless) denotes the partition
coefficient between the air phase and water phase. Using the empirical rela-
tionship, Kd can be expressed as Kd = Koc ·foc , where Koc (ml/g) denotes the
organic carbon-water partition coefficient and foc (g/g) denotes the fraction
organic carbon of the soil.
The dimensionless form of the Henry’s partition coefficient, H, can be
determined from the more common form having the units of atmospheres-
cubic meters per mole (atm-m3 /mol) using the following equation
KH
H= (4)
RT
where KH (atm-m3 /mol) denotes the dimensional form of Henry’s Law con-
stant, R denotes the universal gas constant (R = 8.2 × 10−5 atm-m3 /mol K),
and T is the absolute temperature in Kelvin (◦ K = 273.16 + ◦ C).
The dispersion coefficient in the unsaturated zone is regarded as a linear
function of the pore water velocity as:

qw
Dw = αL (5)
θw
where αL is the longitudinal dispersivity (feet) of the unsaturated zone.

The gas phase diffusion coefficient (Da ) in the porous medium is calcu-
lated by modifying the free air diffusion coefficient using the Millington model
(1959):
(n − θw )7/3
Da = Dair (6)
n2
where Dair denotes the diffusion coefficient of the contaminant in the free air.
By substituting equation (2) and (3) into (1), the governing transport
equation can be simplified as:

∂Cw ∂ 2 Cw ∂D ∂Cw ∂qw
θ =D + − q w − + μθ Cw (7)
∂t ∂t2 ∂z ∂z ∂z
390 Samuel S. Lee
where θ = θw + θa H + ρb Kd and D ≡ θw Dw + θa Da H.
The initial conditions used in this work to solve equation (7) are given in
the following equation

) Cs (z, 0)/Kd if Kd > 0,
Cw (z, t))t=0 = (8)
Cw (z, 0) if Kd = 0,
where Cs (z, 0) is the initial solid-phase concentration specified by the user.

When the distribution coefficient (Kd = Koc · foc ) is zero, liquid-phase con-
centration must be entered as an initial concentration to avoid the program
run-time error (division by zero).
The following assumptions are made in the conceptualization of the un-
saturated zone:
1. Linear isotherms describe the partitioning of the pollutant among the
liquid, vapor, and solid phases with instantaneous equilibrium.
2. Liquid phase dispersion and vapor phase diffusion are combined as one
term via the Henry’s equilibrium constant.
3. Vapor-phase advection is assumed negligible compared to the infiltrating
water velocity and omitted in this screening level model.
4. Contaminant decay is assumed by a first-order process using two con-
stants; one constant for the contaminant decay in the soil, and another
constant for the contaminant source reduction.
5. Flow and mixing processes in the capillary fringe is beyond the scope
of this work and thus it also assumes that the capillary fringe depth is
negligible with respect to the unsaturated column.
The most common types of boundary condition applied at the top of the
soil column are either the first type (Dirichlet’s) or the third type (Cauchy’s).
Equation (9) describes these boundary conditions.

) C0 (z = 0) exp(−γt) if t ≤ t0
)
Cw z=0 = (9)
0 if t > t0
or )
∂Cw ) qw C0 exp(−γt) if t ≤ t0
−D + qw Cw )) = (10)
∂z z=0 0 it t > t0
where C0 is the liquid phase solute concentration in the infiltration water,
γ is the decay rate (1/yr) of the solute source due to either degradation or
flushing by the infiltration, and t0 is the duration of solute release (yr) which
can be selected to simulate either “slug” or continuous input.
At the bottom of the soil column, the second type boundary condition
(Neuman’s) is commonly applied
∂Cw
=0 (z = ∞) (11)
∂z
In applying this boundary condition, equation (11) is actually implemented

at a finite column length (i.e., z
= ∞). To reduce the finite length effect,
dummy cells are added at the bottom of the soil column automatically in
the numerical calculation in the model. After evaluation of Cw (z, t), the total
contaminant mass (M ) per unit volume of the soil is calculated as:
M (z, t) = Ma + Mw + Ms = [θa H + θw + ρb Kd ]Cw = θCw . (12)
2.2 Saturated Zone Mixing
After estimating the liquid phase solute concentration (Cw ) at the bottom
of the soil column, the mixed concentration in the aquifer can be calculated
using a mass-balance technique as below (USEPA 1989, Summers et al. 1980):
Caq qaq Aaq + Cw qw Asoil

Cmix = (13)
qaq Aaq + qw Asoil
where Caq is the concentration of horizontal groundwater influx, qaq is the

Darcy velocity in the aquifer, Aaq is the cross-sectional aquifer area perpen-
dicular to the groundwater flow direction, and Asoil is the cross-sectional area
perpendicular to the vertical infiltration in the soil column. The aquifer area
(Aaq ) is determined by multiplying the horizontal width of the soil column
with the vertical solute penetration depth.
Procedure for the mixing calculation is different depending on the type of
soil column arrangement. In the case of the transverse (right angle) arrange-
ment, the mixing calculation is straight forward: simply apply equation (13)
at the each mixing element underneath the soil columns. For the parallel ar-
rangement case, however, the mixed concentration at the upgradient cell is
considered as an influx concentration to the next cell. The mixing concentra-
tion at the next cell is estimated by reapplying the equation (13) using the
two inflow concentrations.
The solute penetration depth is the mixing thickness of the contaminant in
the aquifer beneath the unsaturated soil column. An estimation of the plume
thickness in an aquifer can be made using the relationship below (USEPA
1990):
* −Lqw
Hd = 2αv L + B 1 − exp (14)
qaq B
where, Hd is the penetration depth (m), αv is the transverse (vertical) dis-
persivity (m) of the aquifer, L is the horizontal length dimension of the waste
(m), and B is the aquifer thickness (m). In equation (14) the first term rep-
resents the thickness of the plume due to vertical dispersion and the second
term represents that due to displacement from infiltration water. When im-
plementing this relationship, it is necessary to specify that in the event the
computed value of Hd is greater than B, the penetration thickness, Hd is set
equal to B.
392 Samuel S. Lee
3 Numerical Implementation
3.1 Unsaturated Zone Leaching
The governing solute transport equation (7) is solved using the finite difference
method. Differential equations dealing with liquid contaminant concentration
Cw as a function of time and depth are converted into the finite difference
equations dealing with the corresponding variable Cik centered on time be-
tween two time steps:
Cik+1 + Cik ∂Cw C k+1 − Cik

Cw → , ≈ i (15)
2 ∂t Δt
where Δt is the time increment, the subscript i refers to the discretized soil
column cell and the superscript k refers to the time level. The subscript w is
dropped for simplicity. Converting the other terms into finite difference form,
the governing equation can be written as:
(−Mi + Mi − Ni )Ci−1

k+1
+ (1 + 2Mi + Ni + Li )Cik+1 + (−Mi − Mi + Ni )Ci+1
k+1
= (Mi − Mi + Ni )Ci−1

k
+ (1 − 2Mi − Ni − Li )Cik + (Mi + Mi − Ni )Ci+1
k
(16)
where the dimensionless constants Mi , Mi , Ni , Ni , and Li are:
Δt 1 Δt 1 Di+1 − Di−1
Mi ≡ Di , Mi ≡ ,
2(Δz)2 θi 2(Δz)2 θi 4
(17)
Δt 1 Δt 1 Δt
Ni ≡ qi , Ni ≡ (qi+1 − qi−1 ) , Li ≡ μ.
4Δz θi 4Δz θi 2
Similarly, the finite difference form of the initial condition for the liquid
phase solute concentration is

(Cs1 )i /Kd if Kd > 0 ,
1
Ci = 1
2 < i < n − 1. (18)
(Cw )i if Kd = 0 ,
The finite difference forms of the top boundary conditions for the soil column
are:
• First Type Top Boundary Condition
Csk (z = 0)
Cik = exp[−γ(k − 1)Δt] ,
Kd k = 1, 2, . . . . (19)
Csk
= 0 if t ≤ t0 ,
Csk = 0 if t > t0 ,
• Third Type Top Boundary Condition

Ψ k+1 Ω Ψ qw C0
C1k+1 −
C2 = − C2k + C2k + exp(−γt) (20)
Φ Φ Φ Φ
where
D(2M + L + 1) qw DM
Φ = + , Ψ = ,
4(Δz)(M + N ) 2 4(Δz)(M + N )
D(2M + L − 1) qw
Ω = + ,
4(Δz)(M + N ) 2
and M , N , L were defined in equation (17).
The second type bottom boundary condition is used in this model as fol-
lows:
Cnk+1 − Cn−1
k+1
= 0. (21)
Δz
This work incorporates the development of a C++ computer code to solve
for the values of C1k+1 using the above finite difference form of simultaneous
equations by employment of the Thomas algorithm (Press et al. 1992).
3.2 Numerical Stability

Often, the efficiency of a numerical technique is limited due to the instability,
oscillation, and mass-balance problems. Several methods have been proposed
to determine the stability criteria of finite difference calculation (e.g., Fourier
expansion method, matrix method, and other, (Hirsch 1989)). The Fourier
expansion method, developed by von Neumann, relies on a Fourier decompo-
sition of the numerical solution in space while neglecting boundary conditions.
It provides necessary conditions for stability of constant coefficient problems
regardless of the type of boundary condition (Mitchel and Griffiths 1980). The
matrix method, however, analyzes the eigenvectors of the space-discretization
operator, including the boundary conditions, as a basis for the representation
of the spatial behavior of the solution (Hirsch 1989, Ames 1997). Based on
the von Neumann method, Crank–Nicolson scheme of finite difference equa-
tion can be optimize with:
2D
Δz < , (22)
qw
2θ(Δz)2
Δt < * . (23)
(2D)2 − (qw Δz)2
According to the stability criteria expressed in equations (22) and (23),
it is clear that the combined dispersion coefficient (D) must be greater than
zero. In natural soil conditions, it is rare to have a value of D as zero or
close to zero. Specifically, if there is downward infiltration in the unsaturated
zone, hydrodynamic dispersion of contaminant is inevitable. In addition, the
air diffusion coefficients of selected organic compounds must also be greater
than zero.
394 Samuel S. Lee
3.3 Saturated Zone Mixing
Based on the mass balance principle of equation (13), the mixed solute con-
centrations are estimated as:
qw1 L1 Cw1 + qaq H̃d1 Caq

Cmx1 = ,
qmx1 H̃d1
(24)
qw2 L2 Cw2 + qaq H̃d1 Caq + qmx1 H̃d1 Cmx1
Cmx2 = ,
qmx1 (H̃d1 + H̃d2 )
where Cmx(i) is the “Mixed Concentration in Groundwater” and H̃ is the

adjusted penetration depth for each column.
4 Program Execution
The hypothetical example with two soil columns of Press et al. (1992) is
simulated herein to demonstrate the impact of soil heterogeneity on the spread
on the contaminant in the soil.
4.1 Problem Descriptions
The example depicts two soil columns arranged perpendicular to the ground-
water flow direction (see Fig. 1). The unsaturated zone soil is divided into
four soil layers whose total porosity decreases (from 0.44 to 0.38) and water-
filled porosity increases (from 0.26 to 0.32) along with the depth shown in
Fig. 1. The bulk density is adjusted according to the total porosity change
(contribution from the water content change is disregarded). The soil column 1
has first type top boundary condition. The soil column 2 has third type top
boundary condition. This assumed set of parameters are derived based on a
filed geologic situation where the total porosity of soil decreases along with
the depth due to gravitational pressure while the soil becomes wetter (water-
filled porosity increases) along with the depth because water sinks down to the
lower layers. Recharge rate qw is kept constant (0.3048 m3 /yr/m2 ) in order to
keep water-filled porosity of each layer constant. Organic content foc is also
kept constant (0.005 g/g).
4.2 Results
Simulation of a homogeneous soil columns were conducted (Figs. 2 and 3)

in order to demonstrate the soil heterogeneity effect on the dispersion of the
contaminant. The uniform homogeneous soil properties that were used in the
simulations were bulk density of 1.6, total porosity of 0.4, and water filled
porosity of 0.3. In both Soil Column 1 and Soil Column 2, the effects of having
Fig. 1. Profile example problem. Soil column 1 is the first type of boundary condi-
tion. Soil column 2 is the third type boundary condition
four different layers instead of one vertically homogenous layer are obvious as
shown in Figs. 4 and 5. The liquid-phase contaminant leached deeper in the
heterogeneous soil after the same period of time (71 mg/L at −4.572 m in
Soil Column 1 at 10 years compared with 28 mg/L for the homogenous case).
That is because the total porosity values for the first (0 ∼ −1.524 m) and
second (−1.524 ∼ −4.572 m) layers are larger than the homogeneous value of
0.4. Also, since the total porosity and water-filled porosity values in the third
layer (−4.572 ∼ −10.058 m) were the same for both the homogenous and
the heterogeneous soils, the inclination of the contaminant profile seems to be
identical in both case. After 20 years, the contaminant penetrated deeper while
keeping a similar profile. After 30 years, the contaminant completely reached
the bottom of the unsaturated zone and mixed with the groundwater. Figure 6
depicts the simulation results of mixing the leachate with the ground water.
Figure 7 illustrates a comparison between predicted results of the Unsat-

urated-zone Leaching and Saturated-zone Mixing Model for simulation of a
homogeneous soil column with an analytical transport solution of Cleary and
396 Samuel S. Lee
Liquid-phase Contaminant Concentration (mg/L)

0 20 40 60 80 100
0
Depth from Ground Surface (ft)

5
10
15
20
25
30
35 at 10 years
at 20 years
40
at 30 years
45 at 40 years
50
Fig. 2. Simulation results of homogeneous soil column 1, first type of boundary

condition at 10, 20, 30, and 40 years of liquid-phase contaminant concentration
(mg/L) vs. depth from ground surface (ft)

0 20 40 60 80 100
0
5
10
15
20
25
30
35 at 10 years
at 20 years
40
at 30 years
45 at 40 years
50
Fig. 3. Simulation results of homogeneous soil column 2, third type of boundary

Ungs (1994). The close matched of the results indicates that the Unsaturated-
zone Leaching and Saturated-zone Mixing Model program works correctly in
homogeneous case. For the heterogeneous case, Unsaturated-zone Leaching
and Saturated-zone Mixing Model results showed reasonable match with col-
umn experiment data (Cleary and Ungs 1994), which is available through
Internet (http://www.vadose.net) (Lee 1999b).
0 20 40 60 80 100
0
at 10 years
Depth from Ground Surface (ft) at 20 years 1st Layer
10
at 30 years
at 40 years 2nd Layer
20
3rd Layer
30
4th Layer
40
50
Fig. 4. Simulation results of heterogeneous soil column 1, first type of boundary

0 20 40 60 80 100
0
1st Layer
10
2nd Layer
20
3rd Layer
30
4th Layer at 10 years

40 at 20 years
at 30 years
at 40 years
50
Fig. 5. Simulation results of heterogeneous soil column 2, third type of boundary

5 Summary
Unsaturated-zone Leaching and Saturated-zone Mixing Model was developed.
This computer code can handle vertical heterogeneity of soil. The Graphic
User Interface used in Unsaturated-zone Leaching and Saturated-zone Mixing
Model has made it easy to create the input data file and view the simulation
results. Additional benefits stems from the ability to handle the many addi-
tional parameters that are needed for vertically heterogeneous soil columns,
398 Samuel S. Lee
TIME (year)
0 50 100 150 200

0.59
Concentration (mg/L) 0.61
0.63
0.65
0.67
0.69
0.71
Soil Column 1 (mg/L) Soil Column 2 (mg/L)
0.73
Fig. 6. Simulation results of mixed concentration at the ground water
Fig. 7. Verification of Unsaturated-zone Leaching and Saturated-zone Mixing Model

with homogeneous soil column by comparing with an analytical solution (PRINCE)
modified from Cleary and Ungs (1994). This figure is not referred to in the text!
where a set of parameters is needed to describe the soil properties of each

layer.
Verification of the computer code resulted with good match between
Unsaturated-zone Leaching and Saturated-zone Mixing Model predictions and
an analytical solution (for homogeneous column) and the soil column exper-
iment (for heterogeneous soil). Development of Unsaturated-zone Leaching
and Saturated-zone Mixing Model was funded partially by the United State
Environmental Protection Agency.
References
Ames WF (1997) Numerical Methods for Partial Differential Equations, 2nd Ed.
Academic Press, New York
Cleary RW, Ungs MJ (1994) PRINCE: Princeton Analytical Models of Flow and
Transport, User Guide, Version 3.0, Waterloo Hydrogeologic Software. Waterloo,
Ontario, Canada
Hirsch C (1989) Numerical Computation of Internal and External Flows. Vol. 1.
Fundamentals of Numerical Discretization. John Wiley & Sons, New York
Lee S (1996) A Screening Level Model for Estimation of Vadose Zone Leaching and
Saturated Zone Mixing: VLEACHSM, North American Water and Environment
Congress, American Society of Civil Engineers
Lee S (1999a) A Modeling Study on Vadose Zone Leaching, Saturated Zone Mixing,
and Groundwater Flow in Heterogeneous Aquifer, Ph.D. Thesis, University of
Rhode Island
Lee S (1999b) http://www.vadose.net
Lee S, In H (2005) Simulating Groundwater Transport Process Using a Vertical Het-
erogeneity Model: A Case Study, Lecture Notes in Computer Science 3398:536–
544
Millington RJ (1959) Gas Diffusion in Porous Media, Science 130:100–102
Mitchel AR, Griffiths DF (1980) The Finite Difference Method in Partial Differential
Equations. John Wiley & Sons, New York
Press WH, Teukolsky SA, Vetterling WT, Flannery BP (1992) Numerical Recipes
in FORTRAN, 2nd Ed. Cambridge Univ. Press New York
Ravi V, Johnson JA (1993) VLEACH: An One-Dimensional Finite Difference Vadose
Zone Leaching Model, Developed for USEPA, RS Kerr Res Lab, Ada, OK
Summers K, Gherini S, Chen C (1980) Methodology to Evaluate the Potential for
Ground Water Contamination for Geothermal Fluid Releases, Prepared by Tetra
Tech Inc. for USEPA/IERL, Cincinnati, OH, EPA–600/7–80–117
USEPA (1989) Determining Soil Response Action Levels Based on Potential Con-
taminant Migration to Ground Water: A Compendium for Examples, Office of
Emergency and Remedial Response, Washington D.C., EPA/540/2–89/057
USEPA (1990) Background Document for EPA’s Composite Model for Landfills
(EPACML), Prepared by Woodward-Clyde Consultants for USEPA, Office of
Solid Waste, Washington D.C.
Prediction of SWCC for Coarse Soils
Considering Pore Size Changes
Xu Li and Limin Zhang
Department of Civil Engineering, The Hong Kong University of Science and

Technology, Clear Water Bay, Hong Kong lixu@ust.hk, cezhangl@ust.hk
Summary. The soil water characteristic curve (SWCC) of a soil is often predicted
from the pore size distribution (PSD) of the soil. Experiments show that the PSD
may change in the drying or wetting process. It is necessary to consider the pore-size
distribution changes in predicting SWCCs.
In this research, different grain-size proportions of completely decomposed gran-
ite were separated and then mixed artificially to obtain five soil specimens of different
grain-size distributions. These specimens vary from gravel to sand, and silty clay.
The SWCCs of these specimens were measured using the axis-translation technique
and a Dewpoint psychrometer; and the PSDs of these soil specimens were measured
using a mercury intrusion porosimetry (MIP) method. The PSDs obtained from
the MIP tests were used to predict the SWCCs, which were compared with the
experimental results further.
The PSDs of each soil sample at full saturation and at the end of the SWCC dry-
ing test were compared. The PSDs showed a considerable pore-size reduction after
soil drying. This phenomenon is referred to as ‘pore shrinkage’. A model proposed
by Simms and Yanful (2001) was adopted to consider the shrinkage influence in the
prediction of the drying SWCC from the PSDs of saturated and dried soil samples.
After applying this model, the predicted SWCCs were closer to the experimental
SWCCs, especially for fine soils.
For coarse soils, the pore volume with pore diameter larger than a specific value
that cannot be measured by MIP tests is large. This volume was estimated and
used to correct the predicted SWCC for each coarse soil. With this correction, the
predicted SWCCs for coarse soils were also close to the experimental SWCCs.
Key words: SWCC modelling, PSD, pore shrinkage
Introduction
The measurement of soil–water characteristic curve (SWCC) is often time
consuming. A traditional method of determining the SWCC using a pressure
plate is limited in a narrow suction range. A viable method is to predict SWCC
from the pore-size distribution (PSD) of soil.
402 Xu Li and Limin Zhang
The nitrogen adsorption and the mercury intrusion porosimetry (MIP) can
give PSD curve quantitatively. However, the MIP method has been widely
used as a reliable method to measure PSDs of a wide variety of porous solids
(Aung et al. 2001). The PSD as measured by MIP has often been advanced as
a parameter to help interpret geotechnical behaviors. It has been correlated
with saturated hydraulic conductivity (Lapierre et al. 1990), frost heave (Reed
et al. 1979), macroscopic volume change (Al-Mukhtar et al. 1996) and SWCC
(Prapaharan et al. 1985, Romero et al. 1998).
Based on the PSD from MIP tests, a SWCC can be estimated rapidly
in a much wider suction range comparing with the traditional method of
determining the SWCC using a pressure plate. But the methodology is still
not mature and satisfactory (Simms and Yanful 2004). For example, the PSD
changes with soil water content even for a same soil sample. Also, the SWCC
prediction method based on the PSD from MIP tests are seldom applied to
coarse soils.
Four coarse soils and one fine soil were used in this paper. To investigate
the changes in PSD during the soil drying process, a series of MIP tests had
been conducted. Also scanning electron microscopy (SEM) tests were used to
study the pore geometry of soils qualitatively. Based on the PSDs from the
MIP tests, the SWCCs for five soils were predicted with the consideration
of the changes of PSD during soil drying and the pore volumes with pore
diameters larger than a specific value, which could not be measured in the MIP
tests. The predicted SWCCs were further compared with the SWCCs from
laboratory tests, which were measured by a pressure plate and a Dewpoint
psychrometer.
Materials
Two unimodal soils, one fine soil with an average particle diameter d50 of
0.0185 mm and one coarse soil with d50 of 6 mm, were ‘brewed’ from a natural
completely decomposed granite (CDG), which was taken from a construction
site located at Beacon Hill, Hong Kong. The grain-size distribution (GSD)
curves and the grain density function curves are shown in Fig. 1. Firstly,
the natural soil was sieved into ten portions according to particle diameter,
namely < 0.063 mm, 0.063–0.125 mm, 0.125–0.185 mm, 0.185–0.6 mm, 0.6–
2 mm, 2–4.25 mm, 4.25–6 mm, 6–10 mm, 10–14 mm and 14–20 mm. The grain-
size distribution of the finest soil–particle portion, which had a maximum
soil–particle diameter of 0.063 mm, was measured by the hydrometer tests
(BS 1377, part 2. 1990). Secondly, a fine soil was mixed from the finest particle
portion with particle diameters smaller than 0.063 mm and the particle portion
with particle diameters in range of 0.063–0.125 mm by weight proportions of
0.72 and 0.28. Thirdly, a coarse soil was mixed from all the particle portions
following the design GSD (refer to Fig. 1).
Prediction of SWCC for Coarse Soils Considering Pore Size Changes 403
100 SM
90 GP
SW
Percent passing (%)

80 SP
GW
70 GDF (GP)
60 GDF (SM)
50
40
30
20
10
0
100 10 1 0.1 0.01 0.001 0.0001
Particle diameter(mm)
Fig. 1. Grain size distributions of design soils
Further, the fine soil and the coarse soil were mixed into three new soils
according to three weight proportions, namely 75% coarse soil mixed with
25% fine soil, 50% coarse soil mixed with 50% fine soil and 25% coarse soil
mixed with 75% fine soil.
Basic parameters of these soils were measured and are shown in Table 1.
Four types of soil samples were used in the experiments: compacted soil sample
(CS), first compacted and then saturated soil sample (SS), oven-dried soil
sample (OS) and powder sample (PS). The relative compaction (RC) value,
which is defined as the ratio of the dry density to the maximum dry density,
was chosen to indicate the soil compaction state.
Table 1. Detailed parameters of five design soils
Soil Maximum Optimum Grain size distribution Group symbols

type dry density water content Clay Silt Sand Gravel (ASTM D2487)
(g/cm3 ) (%) (%) (%) (%) (%)
1 1.97 9 3.9 3.1 16.7 76.3 GP
2 1.96 11 5.9 18.7 17.9 57.5 GW
3 1.90 13 7.9 34.3 19.1 38.7 SP
4 1.71 18 9.9 49.9 20.3 19.9 SW
5 1.55 21 11.9 65.5 21.5 1.1 SM
Experimental Methods
Scanning Electron Microscopy Tests
SEM was used to study the soil fabric for SM qualitatively. Based on the
photographs from the backscattered electron scanning images, the macro-
micro fabric at the soil surface could be recognized.
Mercury Intrusion Porosimetry Tests
MIP is routinely and effectively used to evaluate the PSDs of powder and
bulk materials with open and interconnected pore structures. Compacted and
saturated soil samples (SS) and the oven-dried samples (OS) of five soils (SM,
SW, SP, GW and GP) were used to study the changes in PSDs before and
after soil drying.
During a MIP test, a dehydrated sample was initially surrounded with
mercury at a specific low pressure. Then the surrounded mercury pressure
was increased in steps until it reached the system maximum capacity and the
intruded volumes of mercury were measured for each pressure increment.
The largest soil sample was limited to 15 cm3 in the MIP tests. To maintain
the homogeneity of soil samples, soil particles with diameters above 10 mm
were taken out from the soil when preparing the soil samples for the MIP
tests. With the assumption that the remaining part retained the void ratio
and PSD of the soil, the density of the soil sample used in the MIP tests was
corrected as,
(1 − p)ρ
ρc = (1)
1− Gpρ
s
where
ρ = initial soil density;
ρc = corrected soil density;
p = weight proportion of the soil particles with diameters larger than
10 mm;
Gs = gravimetric density of the soil solid.
Soil Water Characteristic Curve Tests
The SWCCs of five soils (SM, SW, SP, GW and GP) were measured in the
laboratory.
Three devices, a Fredlund SWCC device, a 5-Bar pressure plate extractor
and a Dewpoint psychrometer were used to measure the SWCCs. The Fred-
lund SWCC device was used to measure the wetting and drying SWCC in the
suction range of 0–500 kPa. It is similar to traditional one-dimensional pres-
sure plate extractors. The 5-Bar pressure plate extractor was used to measure
the drying SWCC in the suction range of 0–500 kPa. With a large sample
container, 300 mm in diameter and 150 mm in height, it is able to measure

the SWCCs of large size soil samples and is suitable for coarse soils. The
Dewpoint psychrometer was only suitable for measuring the SWCCs of small
soil samples or powder soil samples, because of the small sample container,
20 mm in diameter and 12 mm in height. The suction range of the Dewpoint
psychrometer was above 500 kPa with an accuracy of ±100 kPa.
Prediction of SWCC Based on the PSD Obtained from MIP Tests
The measured PSD from the MIP tests was used to predict the SWCC. As-
suming the pore geometry can be modeled as a bundle of cylindrical tubes,
the pore radius r can be calculated by using Jurin’s equation (Hillel 1980)
2Ts,water cos αwater 2Ts,mercury cos αmercury
r= = (2)
ψ P
where
ψ = matric suction;
P = mercury injected pressure;
r = entrance pore radius;
Ts,water = surface tension at the water–air interface;
Ts,mercury = surface tension at the mercury–air interface;
αwater = contact angle of water–air interface to solid;
αmercury = contact angle of mercury–air interface to solid.
In equation (2), the constants, air/water surface tension Ts,water and mer-
cury/air surface tension Ts,mercury decreased with temperature. For water in
contact with air, Ts,water = 74.2 mN/m2 at 10◦ C, 72.0 mN/m2 at 25◦ C, and
62.6 mN/m2 at 80◦ C. The surface tension of mercury Ts,mercury , 472 mN/m2
at 20◦ C, is much higher.
Using the symbols used by Simms and Yanful (2002), we define
∂w(ψ)
= vf (ψ) (3)
∂ψ
where
w(ψ) = total water content of the soil sample with a suction of ψ;
vf (ψ) = volume fraction of water corresponding to a suction range of dψ.
Transforming soil suction ψ and mercury intrusion pressure P to pore
radius r according to equation (2), one obtains
vf (ψ) = vf (r) = vf (P ) . (4)
That is to say, the volume fraction vf (ψ) by drainage of water when the
suction increases from ψ to ψ + dψ is equal to the volume fraction vf (P )
by mercury injection when the mercury intrusion pressure increases from P
to P + dP . Also the soil drying process corresponds to a mercury intrusion
process.
In the MIP test, the start mercury intrusion pressure P0 is non-zero. The
total water content in a soil sample at suction ψ can be written as
+ ψ + P0 + P
w(ψ) = w(0) − vf (ψ) = ws − vf (P ) − vf (P ) (5)
0 0 P0
where
ws = saturated water content of a soil sample;
P0 = started mercury intrusion pressure;
vf (P ) = volume fraction of injected mercury at the mercury intrusion
pressure of P .
Limited by the mercury injection pressure ranging from 9.3 kPa to 212 MPa,
only pores which have entrance diameters ranging from 0.006 μm to 142 μm
can be measured in the MIP tests. There may be a considerable volume of
pores with pore entrance diameters larger than 142 μm. For a coarse soil, this
volume may be large. This volume is also indicated by the water retention
capacity in a suction range of 0 to 2 kPa in the SWCC. If the air entry value
(AEV) is larger than 2 kPa, the volume will be 0 and can be ignored. This
volume can be estimated by the experimental data and used to correct the
predicted SWCC. This correction is referred to as ‘P0 correction.’
In the drying SWCC test, water is gradually removed from soil pores by
drainage of the liquid phase, accompanied by pore shrinkage. Earlier research
shows that soil shrinkage will change the PSD of the soil sample considerably
(Simms and Yanful 2001, 2002). Simms and Yanful (2001) offered a method
to model the PSD changes during the SWCC tests based on the assumption
that the pores shrunk elastically as water drained from the soil. In this paper,
this correction to SWCC prediction was exercised referred to as ‘shrinkage
correction.’
Results and Discussions

Influence of Water Content on the PSD
SEM photos of an SM SS (RC = 0.85) and an SM OS (RC = 0.85), which was

first saturated before it is oven-dried, are shown in Fig. 2. There are much
fewer macro pores in the OS. This indicates that most of the macro pores
disappear in the process of soil drying.
Under the same soil dry density, the saturated soil sample (freezing dehy-
dration method) and the oven dried soil sample were analyzed by MIP. The
PSD curves of the SM SS and the SM OS are shown in Fig. 3. The vertical axis
of the figure is the volume frequency, which is defined as the partial derivative
of the cumulative injected mercury volume with respect to log r. There were
much fewer macro pores (larger than 2 μm) in the SM OS than in the SM SS,
which is consistent with the SEM photos shown in Fig. 2. This indicates that
4 m
8 mm
a. Pore geometry of SM OS (RC = 0.85). b. Pore geometry of SM SS (RC = 0.85).

Scale: 800:1 Scale: 700:1
Fig. 2. Pore geometry of SM at different water contents
0.35
SS
0.30
OS
Volume frequency
0.25
0.20
0.15
0.10
0.05
0.00
0.001 0.01 0.1 1 10 100
Pore radius (um)
Fig. 3. Pore size distribution curves of SM SS and SM OS (RC = 0.85)
most macro pores with diameter larger than 2 μm disappeared in the drying
process of the SM soil sample. A possible reason for this is pore shrinkage.
The macro pores may shrink into smaller ones due to the drainage of water
in the drying process. At last, only few macro pores are left in the oven-dried
SM soil sample. Most of them are surrounded by the smaller pores and cannot
be displaced by mercury at low mercury intrusion pressures.
Similar phenomena are found in the comparison of the PSDs of the OS
and SS for GP, GW, SP and SW (refer to Fig. 4). General features between
the PSD of the SS and the PSD of the OS for a sandy soil or a gravel soil can
be summarized as follows:
(1) The PSDs of the SS and the OS are considerably different.
0.1 0.20
SS SS
0.08 OS 0.16 OS
Volume frequency
Volume frequency
0.06 0.12
0.04 0.08
0.02 0.04
0 0.00
0.001 0.01 0.1 1 10 100 0.001 0.01 0.1 1 10 100
Pore radius (um) Pore radius (um)
a. Pore size distribution curves for GP b. Pore size distribution curves for GW
0.25 0.40
OS SS
0.20 SS OS
0.30
Volume frequency
Volume frequency
0.15
0.20
0.10
0.05 0.10
0.00 0.00
0.001 0.01 0.1 1 10 100 0.001 0.01 0.1 1 10 100
Pore radius (um) Pore radius (um)
c. Pore size distribution curves for SP d. Pore size distribution curves for SW
Fig. 4. PSDs of saturated samples and oven-dried samples (RC = 0.85)
(2) Both the PSD of the SS and the OS can be divided into a macro part
and a micro part, which has a boundary pore radius of approximately
2 μm. The volume frequencies in the micro part show an obvious peak.
The volume frequencies in the macro part are flat without obvious peak.
(3) In the micro part, the peak volume frequency, fpeak , and the pore radius
at peak volume frequency, rp , for the SS are larger than those for the OS.
(4) In the micro part, the pores can be divided into three portions by two
boundary pore radii rl and ru . In the small pore-radius portion, in which
the pore radius is smaller than rl , the volume frequencies of the SS and
the OS are same. In the intermediate pore-radius portion, in which the
pore radius is larger than rl but smaller than ru , the volume frequencies of
the SS are smaller than these of the OS. In the large pore-radius portion,
in which the pore radius is larger than ru , the volume frequencies of SS
are larger than these of OS.
Features (3) and (4) may be partially induced by shrinkage. If large pores
shrunk during the soil drying process, the peak volume frequency and the pore
radius at peak volume frequency would decrease. The pore with pore radius
smaller than rl will not shrink into a smaller one. The upper boundary pore
diameter ru is an apparent radius. The volume of pores with diameter ru that
shrunk from larger pores is equal to the volume of pores with diameter ru
that will shrink into smaller ones. Values of the parameters, fpeak , rp , ru and
rl were listed in Table 2.
Table 2. Characteristic parameters of PSD for GP, GW, SP and SW
Type of soil sample rp (μm) fpeak rl (μm) ru (μm)

SS 1.3 0.080
GP, RC = 0.85 0.05 0.5
OS 0.5 0.042
SS 1.1 0.19
GW, RC = 0.85 0.09 0.7
OS 0.6 0.12
SS 0.5 0.22
SP, RC = 0.85 0.04 0.32
OS 0.23 0.17
SS 0.5 0.37
SW, RC = 0.85 0.04 0.35
OS 0.32 0.26
Notes:
fpeak = peak volume frequency in the micro part;
rp = pore radius at peak volume frequency in the micro part;
ru = boundary pore radius, above which the volume frequencies of SS were larger
than these of OS;
rl = minimum pore radius of pore shrinkage during soil drainage.
SWCC Prediction Based on the PSD Obtained from MIP Tests
The drying SWCCs of the five soils under the same RC value are put together
in Fig. 5. The water retention ability of the five soils increases with the fraction
of fine contents in soil.
When the soil suction decreases to 5 kPa, the degrees of saturation of gravel
soils, GP and GW, are 0.73 and 0.4. Also the degree of saturation for SW is
0.88 at suction of 5 kPa. That indicates that there are abundant pores with the
pore entrance diameter larger than 56 μm (refer to equation (2)). That is to
say the pore volume with pore entrance diameter larger than 142 μm for GP,
GW and SW, which cannot be measured in MIP tests, should be corrected
in the SWCC prediction process. In contrary, the degree of saturation for
SM and SP is about 0.99 when the soil suction decreases to 5 kPa. The pore
volume with pore entrance diameter larger than 142 μm for SM and SP can
be ignored in the SWCC prediction.
The predicted SWCCs of SM are shown in Fig. 6. After the considera-
tion of shrinkage influence (Simms and Yanful 2001), the predicted SWCC is
Degree of saturation 1.00 0,50
0.80 0,40
Volmetric WC (%)
0.60
0,30
0.40
0,20
0.20
0,10
0.00
0.1 1 10 100 1000 0,00
Suction (kPa) 0,1 1 10 100 1000
Suction (kPa)
GP GW SP SW SM
GP GW SP SW SM
a. Degree of saturation vs. suction b. Volumetric water content vs. suction

Fig. 5. SWCCs of five soils measured by the soil sample (RC = 0.85)
40%
Mass water content (%)
30%
20% Predicted SWCC

Corrected by shrinkage
10% 5 bar SWCC test
Dewpoint SWCC test
van Genuchten Fit
0%
0.1 1 10 100 1000 10000 100000
Suction (kPa)
Fig. 6. Predicted SWCCs from PSD for SM (RC = 0.85)
closer to the experiment SWCC. But the predicted AEV, 8 kPa, is less than
the AEV from the experiments, 15 kPa. This phenomenon is consistent with
Prapaharan et al. (1985) and Simms and Yanful (2001). Prapaharan et al.
(1985) gave a possible reason that the clay mineralogy adsorbed to the soil
surface and induced a smaller pore diameter at the soil surface in the SWCC
tests compared to that in MIP tests. This phenomenon may also be due to
the disturbance of the macro pore structures during sample preparation for
the MIP tests or due to the inaccuracy of the mercury-air contact angle at
the soil surface used in the calculation.
The predicted SWCC for SP is close to that from the experiment (refer
to Fig. 7(a)). But its shape does not fit the experimental data very well. Also
the similarity between the predicted SWCC and the SWCC obtained in the
experiment is improved by the shrinkage correction. The P0 correction is not
needed for SP, which is consistent with the experimental SWCCs (refer to
Fig. 5(b)).
0.30 0.40
Mass water content

Mass water content
0.30
0.20
0.20
0.10 0.10
0.00
0.00 0.1 10 1000 100000
1 100 10000 Suction (kPa)
Suction (kPa)
Predicted SWCC Corrected by shrinkage
Predicted Corrected Experiment Experiment Corrected by P0
a. Predicted SWCCs from PSD for SP b. Predicted SWCCs from PSD for SW
0.25 0.30
Mass water content
Mass water content
0.20
0.20
0.15
0.10 0.10
0.05
0.00 0.00
0.1 10 1000 100000 1 10 100 1000 10000
Suction (kPa) Suction (kPa)
Experiment Corrected by P0 Experiment Corrected by P0
c. Predicted SWCCs from PSD for GP d. Predicted SWCCs from PSD for GW
Fig. 7. Predicted SWCCs from PSDs for GP, GW, SP and SW (RC = 0.85)
The predicted SWCCs from the PSDs of SW, GP and GW are shown
in Figs. 7(b)–(d). Without the P0 correction, the predicted SWCCs are far
away from the experimental SWCCs. With the correction of the pore volume
with pore entrance diameters larger than 142 μm, the predicted SWCCs of
SW, GP and GW are close to the experimental data. The predicted SWCC
of GP is satisfactory. But the shape of the predicted SWCCs of SW and GW
are different from that obtained in the SWCC experiments. The influence of
shrinkage correction is rather limited for SW, GP and GW.

A series of MIP tests and SEM tests were conducted in this paper to inves-
tigate the changes of PSD during soil drying. Shrinkage phenomena occurred
during the drying process for all five CDG soils.
The PSDs obtained from the MIP tests were used to predict the SWCCs
of the five soils. The results demonstrated that this method was effective and
could give a drying SWCC rapidly in a much wide suction range. With the
P0 correction, the predicted SWCCs for the coarse soils were close to the
experimental SWCCs. With the shrinkage correction, the predicted SWCC
for SM fitted the experimental SWCC quite well.
Acknowledgements
This research was substantially supported by a grant from the NSFC/RGC

Joint Research Scheme between the National Natural Science Foundation of
China and the Research Grants Council of the Hong Kong SAR (Project
No. N-HKUST611/03). The authors are grateful to Prof. D. G. Fredlund for
constructive discussions during the course of the study.
References
Al-Mukhtar M, Belanteur N, Tessier D, Vanapalli SK (1996) The fabric of clay soil
under controlled mechanical and hydraulic stresses, Appl Clay Sci 11:185–197
Aung KK, Rahardjo H, Leong EC, Toll DG (2001) Relationship between porosimetry
measurement and Soil water characteristic curve for an unsaturated residual soil,
Geotech Geol Eng 19:401–416
Hillel D (1980) Fundamentals of soil physics. Academic Press Inc., London
Lapierre C, Leroueil S, Locat J (1990) Mercury intrusion and permeability of
Louisville clay, Can Geotech J 27:761–773
Prapaharan S, Altschaeffl AG, Dempsey BJ (1985) Moisture curve of compacted
clay: mercury intrusion method, J Geotech Geoenv Eng 111(9):1139–1143
Reed MA, Lovell CW, Altschaeffl AG, Wood LE (1979) Frost-heaving rate predicted
from pore-size distribution, Can Geotech J 16:463–472
Romero E, Gens A, Lloret A (1998) Water permeability, water retention and mi-
crostructure of unsaturated Boom clay, Eng Geol 54:117–127
Simms H, Yanful EK (2001) Measurement and estimation of pore shrinkage and
pore distribution in a clayey till during soil–water characteristic curve tests,
Can Geotech J 38(4):741–754
Simms H, Yanful EK (2002) Predicting soil–water characteristic curves of compacted
plastic soils from measured pore-size distributions, Geotechnique 52(4):269–278
Simms H, Yanful EK (2004) Discussion of the application of mercury intrusion
porosimetry for the investigation of soils, including an evaluation of its use to
estimate volume change in compacted clayey soils, Geotechnique 54(6):421–426
The Influence of the Pore Fluid on Desiccation
of a Deformable Porous Material
Hervé Péron1 , Liangbo Hu2 , Tomasz Hueckel2 , and Lyesse Laloui1

1
Soil Mechanics Laboratory, Ecole Polytechnique Fédérale de Lausanne – EPFL,
Station 18, CH-1015 Lausanne, Switzerland
herve.peron@epfl.ch, lyesse.laloui@epfl.ch
2
Department of Civil and Environmental Engineering, Duke University, Durham,
N.C. 27706, U.S.A. lh19@duke.edu, Hueckel@duke.edu
Summary. This paper aims at elucidating the influence of pore fluid properties
on the shrinkage strains of deformable materials subjected to drying. Results of
isothermal drying tests of two kinds of silts saturated with three different pore fluids
are presented. Results show different strain amounts during drying and the final void
ratios, depending on pore fluid. The rates of drying are consistent with saturated
vapor pressures, while their amount with the surface tension. The shrinkage limit
appears also to be controlled by a limitation in skeleton compressibility.
Key words: desiccation, shrinkage, pore fluid properties, unsaturated soil, cracks
Introduction
The study of deformation due to drying is fundamental in the understanding of
the patterns of stress generated by the constrained shrinkage and cracking in
soils. When drying shrinkage is constrained, reaction forces arising located at
the constraints generate stresses, which depend on the amount of shrinkage.
Our previous papers (Péron et al. 2006, Hu et al. 2006) revealed how the
development of such stress in the elastic domain leads to soil cracking.
Notably, several previous studies established that most of the desiccation
shrinkage develops when soil is saturated (Croney and Coleman 1954, Fleureau
et al. 1993). The state of full saturation when cracking initiates was hypoth-
esized by several authors (Konrad and Ayad 1997, Abu-Hejleh and Znidarcic
1995, Lloret et al. 1998). Recently, Péron et al. (2006) showed that cracks
occur at the end of saturation range or at the early stage of desaturation.
The goal in this work is to investigate a possible relationship between
the total amount of shrinkage, its rate and the mechanisms of liquid removal
and deformation. In particular, the role of the evaporation flux versus surface
tension, in addition to soil compressibility and transport properties are of in-
terest. Slow, constant humidity environment, (isothermal) air-drying of thin
414 H. Péron et al.
and long rectangular soil slabs was carried out. Two different soil types were
used: clayey silt, and a powder of rock. The soils were wetted with three dif-
ferent pore fluids, characterized by distinctly different surface tension, latent
heat, saturation vapor pressure and viscosity.
Desiccation Tests
Desiccation Tests Procedure
The silts employed were: a clayey silt (referred to as silt A), and a pow-
der of granite (silt B). They were saturated with three fluids: tap water,
and two alcohols: ethanol/water mixture (with a volume ratio of 50%/50%),
and ethylene-glycol/water mixture (65%/35%) with the properties shown in
Table 1. The two alcohols were used in mixtures with water to avoid floccula-
tion, which is especially acute in pure ethanol. Saturation vapor pressure and
latent heat were calculated as for pure mixtures.
Table 1. Properties of fluids used for material preparation
Ethylene-
Ethanol
Pure water glycol
solution
solution
Specific Latent Heat [MJ/m3 ] 2448 1479 1591
Saturation Vapor Pressure [Pa] 2337 1639 3253
Surface Tension [mN/m] 72.6 56.4a 30a
Viscosity [cm2 /s] 0.01 0.0423b 0.029b
Density [g/cm3 ] 1 1.071 0.92
a
Vasquez et al. (1995) J Chem Eng Data 40;
Won et al. (1981) J Chem Eng Data 26
b
Fernandez & Quigley (1988) Can Geotech J 25;
Hayduk & Malik (1971) J Chem Eng Data 16(2)
Two silts were used: a clayey silt (referred as silt A) and granite powder
(silt B). Particles greater than 90 μm (sandy fraction) have been removed.
In silt A, the clay fraction is 25%; clay minerals are illite (10% of the total
amount of mineral species), smectite (10%) and chlorite (5%). The liquid limit
is wL = 31.8%, plastic limit wP = 16.9%, and the unit weight of solid particles
is γs = 27.1 kN/m3 . Silt B has low clay content (less than 7%, mainly illite
and chlorite). The liquid limit is wL = 29.5%, the plastic limit wP = 20.1%,
and the unit weight of solid particles is γs = 27.7 kN/m3 .
The tests were performed in a climate chamber with a controlled relative
air humidity and temperature, of 40% and 18◦ C, respectively. Strains in three
The Inf luence of the Pore Fluid on Desiccation 415
directions were measured with calipers, which were fixed on the support. Av-
erage fluid content of the whole cake was recorded versus time by continually
weighing an identical cake, dried in the same time in identical conditions as the
one used for strain measurement. Details of the tests are reported elsewhere
(Péron et al. 2006, Hu et al. 2007).
Density changes (due to the fact that the more volatile fluid of the mixture
evaporates faster) were tested for the calculation of void ratio evolution during
drying, as well as for the fluid properties determination.
Desiccation Test Results
The main result is the relationship between the void ratio evolution and the
volumetric fluid content for both materials plotted in Fig. 1A and B. In these
graphs the thin dotted line denotes the theoretical void ratio evolution under
the assumption that the material remains water saturated at constant volume
of solids. It is noted that at the beginning of drying, all the curves do not
much deviate from this line. However, well before all fluid is evaporated, the
shrinkage stopped. The volumetric water content value at that point iden-
tified as the shrinkage limit tends to depend on the pore fluid. Void ratio
stabilizes finally, reaching a “limiting void ratio.” The curves obtained could
be conceptualized as piecewise bilinear, with the first part merging with the
saturation line, and a second part with void ratio equal to a constant “limiting
void ratio.”
For either silt, water involves the highest amount shrinkage and ethanol
the lowest. Ethylene-glycol induces intermediate values, closer to water in the
case of silt B, and closer to ethanol in the case of silt A. Shrinkage in the case
of silt A was much more significant for all three liquids, than in the case of
silt B. The values of the limiting void ratios exhibit also an apparently wide
scatter. In Figure 2 the volume loss of fluids for silt A and B has been plotted
versus time. The drying of silt with glycol is significantly slower. The drying
Fig. 1. Void ratio evolution with respect to volumetric fluid content

Fig. 2. Fluid loss with respect to time
rates are well correlated with vapor pressure of the permeating liquid: the
higher vapor pressure, the higher drying rate.
Discussion
First, it has to be outlined that the rates of fluid removal and void ratio change
(Fig. 1 and Fig. 2) follow the order of saturated vapour pressure of the fluids,
that is glycol – water – ethanol.
A hypothesis of air-water interface (or contractile skin) is described in
Fredlund and Rahardjo (1993). The surface tension associated with the con-
tractile skin results in a reaction force, which is likely to produce compression
of the soil structure. For the saturated range, Hu et al. (2007) showed that
evaporation at the external solid-fluid-air interface is the only driving force
of shrinkage. Capillary force exists only at this boundary, as there is no air
within the body of the sample. During drying the removal of the pore fluid
at the boundary generates an outward transport of fluid, and a consequent
fluid content decrease within the medium. That induces a negative pore pres-
sure gradient across the body. That also produces strain and the proportional
stress field that is equilibrated at the boundary by capillary forces. Hu et al.
(2007) performed a simple numerical simulation of the process in 1D using
Biot theory that captures the salient points of the process.
The limiting value of the void ratio change and the total strain variation
over the course of drying are left outside of the above model. Shrinkage of un-
constrained specimens ceases at a relatively well-defined configuration. The
mechanism of this cessation remains an open question. At least for the silt A
when wetted with water, the air entry value obtained from an independent
evaluation via the water retention curve coincides with the shrinkage limit (see
Péron et al. 2006). Nevertheless, in some cases, shrinkage limit may signifi-
cantly differ from the air entry value; see experimental results of Taibi (1992)
for remolded silt. Fleureau et al. (1993) hypothesise that for low saturation
state, the capillary forces between the particles are normal to a plane tangent
to the particle at the contact point and thus cannot result in a rearrangement
of the soil structure. Other hypotheses on the shrinkage limit, include a locking
value of compressive strain, air phase spreading due to either bulk pore water
(internal) cavitation or an external air/fluid interface instability followed by
fingering. In general, such hypotheses may be linked to the properties of the
solid skeleton and of the fluid. In what follows we explore some numerical
correlations between the limiting void ratio and void ratio variation during
drying and basic fluid properties (surface tension, vapor pressure, latent heat,
viscosity). Figures 3–6 present such correlations. The limiting void ratios and
void ratio variation values are clearly different for the two materials. They
also depend on the permeating fluid. Interestingly, the numerical values of
the characteristics of interest for the fluid employed do not follow the same
sequence.
For both soils, the value of limiting void ratio, as well as void ratio evolution
exhibit similar trends with respect to viscosity, surface tension, saturation
vapor pressure and latent heat. However, silt A is clearly more sensitive to
pore liquid than silt B. Indeed, the differences in the values of limiting void
Fig. 3. Limiting void ratio and void ratio variation with respect to viscosity
Fig. 4. Limiting void ratio and void ratio variation with respect to vapor pressure
Fig. 5. Limiting void ratio and void ratio variation with respect to latent heat
Fig. 6. Limiting void ratio and void ratio variation with respect to surface tension
ratio, as well as void ratio variation in silt B are not substantial. Visibly,
the most clear monotonic trend of the void ratio variables is seen for surface
tension (Fig. 6). The higher the surface tension the higher the total variation
of void ratio and hence the volumetric strain during drying and the lower
the limiting void ratio. Early research (Kingery and Francl 1954) assessed a
similar trend and showed the final shrinkage of kaolinite clay during drying
was a linear function of the surface tension of the pore fluid.
This observation supports the assumption that the shrinkage deformation
is controlled by external boundary menisci, which act as a stretched mem-
brane. It must be recognized however, that the capillary forces depend also
on the contact angle, which however, is not known at the grain scale.
Viscosity, and hence permeability show much less influence except for a
clear difference between the two alcohols and water in silt A (Fig. 3). As these
two variables control the fluid flux to the external surface they may have a
greater impact on the rate than the total or limiting strain.
Finally, it seems that the clayey character of silt A makes it much more
shrinking with water than it is with alcohol as the pore fluid. Earlier tests
of Sridharan and Venkatappa Rao (1973) indicate a dramatically lower com-
pressibility of clayey soils permeated with alcohols than with water. This is
attributed to the existence of highly deformable electric diffuse double layer
coatings around clay particles in water. Hence, it appears that difference in

response of soils to drying in terms of the total shrinkage and limiting void
ratio point also to their difference in generating the adsorbed water.
Conclusions and Open Issues

Drying of unconstrained slabs of silts has been tested experimentally. Most of
the shrinkage strain occurs in the saturated phase of drying with an evapo-
rative flux at the surface and a perimeter capillary force. Limiting void ratio
and void ratio variation correlate well with surface tension confirming a role
of surface capillary forces in the process. Clayey silt is more sensitive to pore
fluid properties than granite powder. Hence, clay water evaporation mech-
anism is of particular interest for further investigation. The shrinkage limit
appears also to be controlled by a limitation in skeleton compressibility.
Acknowledgment
This work is funded by a cooperation of the US NSF (grant # 0324543) and

Swiss NSF (grant 200021–101917).
References
Abu-Hejleh AN, Znidarcic D (1995) Desiccation theory for soft cohesive soils, J
Geotech Eng 121(6):492–502
Croney D, Coleman JD (1954) Soil structure in relation to soil suction (pF), J Soil
Sci 5:75–84
Fleureau JM, Kheirbek-Saoud S, Soemitro R, Taibi S (1993) Behavior of clayey soils
on drying-wetting paths, Can Geotech J 30(2):287–296
Fredlund DG, Rahardjo H (1993) Soil mechanics for unsaturated soils. Wiley-
Interscience Publication, John Wiley and Sons, Inc.
Hu L, Péron H, Hueckel T, Laloui L (2006) Numerical and phenomenological study
of desiccation of soil. In: Lu N, Hoyos LR, Reddi L (eds) Advances in Unsaturated
Soil, Seepage, and Environmental Geotechnics, GSP 148 ASCE:166–173
Hu L, Péron H, Hueckel T, Laloui L (2007) Drying shrinkage of deformable porous
media: mechanisms induced by the fluid removal. GeoDenver 2007, GSP ASCE
(to appear)
Kingery WD, Francl J (1954) Fundamental Study of Clay: XIII, Drying Behavior
and Plastic Properties, J Am Ceram Soc 37(12):596–602
Konrad JM, Ayad R (1997) An idealized framework for the analysis of cohesive soils
undergoing deesiccation, Can Geotech J 34:477–488
Lloret A, Ledesma A, Rodriguez R, Sanchez MJ, Olivella S, Suriol J (1998) Crack
initiation in drying soils. Unsaturated Soils, Pékin, International Academic Pub-
lishers, pp 497–502
Péron H, Laloui L, Hueckel T, Hu L (2006) Experimental study of desiccation of

soil, In: Proceedings of the Fourth International Conference on Unsaturated
Soils, UNSAT 2006: pp 1073–1084
Sridharan A, Venkatappa Rao G (1973) Mechanisms controlling compressibility of
clays and the role of the effective stress concept, Géotechnique, 23(3):359–382
Taibi S (1992) Comportement mécanique et hydraulique des sols partiellement sat-
urés. PhD. Thesis, Ecole Centrale de Paris
Determination of the Soil Water Retention
Curve and the Unsaturated Hydraulic
Conductivity from the Particle Size Distribution
Alexander Scheuermann and Andreas Bieberstein
Institute of Soil Mechanics and Rock Mechanics, Division of Embankment Dams

and Landfill Technology, University of Karlsruhe (TU)
alexander.scheuermann@ibf.uka.de,
andreas.bieberstein@ibf.uni-karlsruhe.de
Summary. Because of the complexity of the metrological determination of the soil

water retention curve (SWRC), so-called pedotransfer functions (PTF) have been
developed for several years. Mostly these PTF are based on a more or less sim-
ple regression analysis using a limited set of data. In such methods the SWRC is
predicted with data on the amount of soil components sometimes supplemented by
values regarding the density or the amount of organic materials. Only few PTF
deal directly with the particle size distribution. In many cases empirical factors are
necessary to obtain a prediction for the water retention curve. A new method for
determining the soil-hydraulic properties using the pore constriction distribution
of a soil has been developed, whereby the pore constriction distribution is derived
from the particle size distribution depending on the density of the soil. The contri-
bution will present the new pedotransfer method and shows results in comparison
to experimental investigations.
Key words: pedotransfer method, soil water retention curve, hydraulic conductiv-
ity, pore model
1 Introduction
Different methods have been developed to estimate the SWRC from the par-
ticle size distribution. For this purpose different approaches were used. By
making estimations using a simple model for the pore structure, the particle
size distribution can be separated into fractions of different sizes. Using this
information it is possible to incrementally estimate the SWRC for a prede-
fined bedding of particles within a specific size range (Arya and Paris 1981,
Fredlund et al. 2002). The analogy between the progression of the particle size
distribution and SWRC makes it possible to reach a direct solution using a
functional relationship between both curves (Haverkamp and Parlange 1986).
Based on the particle size distribution, newly developed methods calculate the
422 Alexander Scheuermann and Andreas Bieberstein
capillary and adsorptive bound water separately, which are then summarised
in order to receive the SWRC (Schick 2002, Aubertin et al. 2003). However, as
a rule, these kinds of solutions require additional empirical information which
naturally contains uncertainties. A numerical solution for the calculation of
the SWRC and the unsaturated hydraulic conductivity has been given by Ki-
tamura et al. (1998). For the calculation they use a statistical distribution for
the pore diameter and the pore inclination, which are both derived from the
particle size distribution (cf. Kitamura et al. 2000). On the basis of the consid-
eration of sphere packings Zou (2003, 2004) has developed an elegant solution
for the estimation of the SWRC because it is purely analytical. However, for
this reason this method can only be used at best for uniform soils.
In the following, a method will be briefly introduced, which uses the pore
constriction distribution for the calculation of the SWRC for the case of
drainage as well as the relationship for the unsaturated hydraulic conduc-
tivity. In this connection, the pore constriction distribution will be calculated
from the particle size distribution under consideration of the relative density
of the soil.
2 Calculation of the Pore Constriction Distribution
The structure of a pore system depending on the density is the most important
factor influencing the soil hydraulic parameters. Another relevant geotechnical
question regarding pore structures is the ability of materials to pass through
pores in a soil. In this connection the suitability of a soil for grouting purposes
(Schulze 1992) or questions concerning geotechnical filters (Wittmann 1980,
Witt 1986, Schuler 1997) should be mentioned. However, the direct measure-
ment of the pore size distribution is a challenging undertaking. For this reason
very early first approaches have been developed in order to estimate pore size
distributions from more easily measurable parameters like the particle size
distribution.
A procedure of this kind has been developed by Silveira (1965). It consid-
ers a particle group consisting of three touching spherical grains. With this
consideration he assumes from the beginning the densest condition of a soil
(hexagonal packing). In the space between the grains a spherical pore can be
subscribed which touches every three grains representing the pore constriction
of this group. As a basis for the selection of the grain sizes Silveira uses the
particle size distribution finer by weight, which has been criticised by Ziems
(1969) since the introduction of the procedure, because of the underestimation
of the finer part of the soil. For a similar procedure Ziems used the particle
size distribution finer by amount (quantity distribution).
Schuler (1997) has finally developed the procedure, which is the basis for
the calculation of the pore constriction distribution. As a basis for the selec-
tion of grains, he uses the particle size distribution finer by surface (surface
distribution). Additional modifications carried out by Schuler are as follows:
Determination of the SWRC from PSD 423
Fig. 1. Calculation of pore constrictions within a pore space consisting of four

particles a) for the lowest relative density D = 0 and b) for the highest relative
density D = 1
• Experimental investigations have shown that most of the pores are formed
by four particles (Witt 1986 (pore imprints) and Glantz 1997 (magnetic
resonance imaging)). Due to this observation instead of three particles
Schuler considers a group of four particles as shown in Fig. 1.
• When a pore constriction is formed by four particles, the probability that
the centre of each spherical grain is located in one intersection is very low,
which is verified by observations according to Witt. Due to these observa-
tions it is assumed that one particle is shifted in the third dimension, while
the radius of this particle is decreased in the two dimensional consideration

of Fig. 1 to rD = rD · (2/3)0.5 (cf. grey particles (D) in Fig. 1). In Figure 1
it is assumed that all particles have the same radius (rA = rB = rC = rD ).
• For a pore constellation (group of particles with specific positions) as
shown in Fig. 1 spaces with different sized areas are possible. If parti-
cles A and B touch each other, the densest packing of the four particles
will be obtained and the angle α as well as the area FP E of the space
between the particles then have their lowest values (cf. Fig. 1b)). As the
values for α increase also the area FP E of the space between the particles
grows until FP E reaches a maximum (cf. Fig. 1a)). Under this condition
the constellation reaches its loosest density. Consequently the constellation
of particles takes on different densities according to the area of the space
between the particles, which can be assigned to the loosest and densest
bedding of the particles. In order to take into account the bedding of the
soil for the calculation of the pore constrictions, an analogy is introduced
between the area FP E of the space between the particles and the porosity
n of the soil as shown in equation (1)
nmax − n FP E,max − FP E
D= = . (1)
nmax − nmin FP E,max − FP E,min
As mentioned before, the selection of the grains takes place using the
surface distribution. For this purpose, the particle size distribution is divided
into sections of equal percentage and for every section a mean diameter of
the particle is calculated. As a result every particle diameter has the same
probability of occurrence. Because of the consideration of four particles and
depending on the amount of sections k chosen for the calculation, a total
amount of Zk = k 4 constellations has to be considered. Finally, for each pore
constellation two pore constriction radii are subscribed and a mean value
characterising the pore is calculated.
The subscription of pores into the space between the particles is a purely
geometric problem. In order to improve the efficiency of the calculations, an
analytical method is used and integrated into a programme. Calculations with
k = 15 pore sections are possible, which is very important for the consideration
of well graded soils.
Figure 2 shows the result of a calculation of a pore constriction distribution
with a relative density of D = 0.8 in comparison to experimental results from
Witt (1986). The solid line on the right shows the particle size distribution,
the long dashed line shows the quantity distribution and the line with dashes
Fig. 2. Comparison between calculated pore constriction distribution and a mea-

sured distribution based on silicon pore imprints (according to Witt 1986)
and dots shows the surface distribution. It can be seen from the propagation of
the curves that by using the surface distribution the probability of occurrence
of finer particles is higher than by using the particle size distribution finer by
weight and lower by using the quantity distribution.
The bundle of lines on the left shows the pore constriction distributions
finer by amount. The dotted lines represent the results of the calculation. The
thick line represents the mean value of the pore constriction and both fine
lines represent the limits for maximum or minimum pore constrictions. The
experimental result according to Witt is shown by a dashed line with two dots.
Witt determined his distribution using silicon imprints of the pores. On the
one hand, it is a challenging procedure to assess pore size distributions, but,
on the other hand, it is a quite exact one. It can be very clearly recognised
that the experimental result is reproduced very well with the calculation,
which has already been verified by Schuler (1997). This comparison illustrates
the applicability of the introduced procedure for the calculation of the pore
constriction distribution depending on the bedding, at least for granular soils.
3 Derivation of Soil Hydraulic Parameters

In the following the required steps for the calculation of the soil hydraulic
parameters are presented. The pore constriction distribution constitutes the
necessary input data for this calculation, which must be carried out sepa-
rately. The following calculation examples are presented for a weak gravelly
sand, of which the particle size distributions, quantity distribution and surface
distribution as well as the calculated distributions for the pore constriction
are shown in Fig. 3.
3.1 Soil Water retention Curve (SWRC)
As already mentioned, the SWRC of a soil for the drainage condition is mainly
influenced geometrically by the pore constrictions of a soil. The retentive force
acting against the drainage of water out of the soil is the capillary force which
is activated by the greatest pore with a connection to the free air phase.
Consequently, with the information of the pore constriction distribution, it
is possible to determine the capillary forces as isotropic forces acting in a soil.
For this purpose the well known equation (2) for the description of the rise
of water in a capillary tube is used. In this connection, the so-called capillary
rise of water is inversely proportional to the radius of the pore constriction
rP E . For the case of drainage (completely wetted surface) it is feasible to
assume a value of δ = 0◦ for the contact angle (cf. Schubert 1982, Fredlund
and Rahardjo 1993). The surface tension σwa of water to the air phase is a
temperature-dependent property, which can be inferred from Table 1. Finally,
with the specific weight of water γw it is possible to calculate two values
for every pore constellation for the capillary rise of water (according to the
Fig. 3. Particle size distribution, quantity distribution and surface distribution

for a weak gravelly sand as well as pore constriction distributions for mean pore
constrictions for a relative density of D = 0.8 (with variation due to maximal and
minimal pore constrictions)
Table 1. Surface tension σwa of water to air depending on the temperature
Temperature (◦ C) 10 15 20 30
Surface tension σwa (N/m) 0.0724 0.0735 0.07275 0.0712
minimum and maximum pore constrictions) as well as a mean value, which is

set equal to prevailing capillary forces |ψP E | due to the a pore constriction:
2σwa cos δ
|ψP E | = (2)
γw rP E
The decrease in the water content during drainage is calculated from the
loss of water volume, which is set equal to the area loss of the drained pore
space for each pore constellation. The sum of the area of all pore spaces
equates to S = 100% saturation. Finally, by presetting the porosity n accord-
ing to the porosity used for the calculation of the pore constriction distribution
(cf. equation (1)), the analogy to the saturated volumetric water content is
established.
Without considering water-retention in the pore space, the soil would be

completely drained in the calculation. In fact, some water is stored in the
contact points or gussets of the particles, which leads to residual water in
every pore constellation. However, the water in the gussets of the particles
forms a corpus, which is approximated in the 2D-consideration. Furthermore,
it is unknown with which radius water is stored in the gussets. Due to these
uncertainties, the residual water is considered to be a variable. As an em-
pirical value it is feasible to predefine the residual water content as the sum
of all water stored in the gussets of the particles. Another possibility for a
predefinition of the water content is the use of a matric suction at which the
residual water content will be set (similar to the definition of the field capacity
between 60 and 310 hPa). Finally it is also possible to automatically calculate
the residual water content using the pore radii of the constrictions of each
pore constellation. If the smaller pore constriction of each pore constellation
is used for the calculation, an upper limit for the residual water content is
obtained and with the use of the smallest pore constriction of all pore constel-
lations, a lower limit is determined. So far adsorptive bound water has been
neglected in the calculation.
For a direct estimation of the residual water content or the water content
at saturation, data from literature can also be used (e.g. relative values for
the water contents by Luckner et al. 1989).
With the procedure presented discrete points of the SWRC are calculated,
which can be parameterised using appropriate models like Brooks and Corey
(1964) or van Genuchten (1980). Figure 4 shows the result of a calculation
(white squares) in comparison to a statically measured SWRC (grey circles)
for the weak gravelly sand presented in Fig. 3. Here the saturated water con-
tent θS was given by the assumed porosity, and the residual water content θr
was automatically calculated using the smallest pore constriction radius for
all pore constellations. Whereas the saturated water content is more or less a
given parameter, the residual water content represents a value which can be
estimated taking values from experience.
As can be seen in Fig. 4, there is a satisfactorily good match between the
measured and the calculated SWRC. In this connection, the calculation of the
SWRC was carried out as a prediction for a given porosity n with a resulting
relative density D (under consideration of nmax and nmin , cf. equation (1)).
It can be assumed that the accordance of both curves could be much better
with an adapted value for D.
3.2 Unsaturated Hydraulic Conductivity
There are different microscopic approaches for the determination of the un-
saturated hydraulic conductivity. Microscopic approaches are based on a pore
model and the hydraulic conductivity is calculated by summation of the con-
ductivities of individual capillaries. The most well known model based on
a microscopic approach is the model of Mualem (1976) which is the most
Fig. 4. Comparison between measured and calculated soil water retention curve
using the pore constriction distribution for the weak gravelly sand (cf. Fig. 3)
widespread model in combination with the parameterisation of van Genuchten

(1980).
For all microscopic approaches specific conditions are assumed which were
defined for the first time by Childs and Collis-George (1950):
• The pore size distribution of a soil can be derived from the soil water
retention curve.
• For individual pore fragments, the flow model of Hagen-Poiseuille for lam-
inar flow in channel-like capillaries applies (cf. equation (3)).
• Since in the equation of Hagen-Poiseuille the radius of the capillary is used
to the power of four, the effective flow resistance in a sequence of pores in
series is significantly influenced by the pore with the smallest diameter.
• The total conductivity of a porous media is determined only by sequences
of pores in series. Hence a lateral connectivity of pore sequences is ne-
glected.
In this connection, the negative effects of the last two assumptions should
counterbalance each other. All microscopic approaches use pore sequences
which are combined in a specific way. By the summation of water-bearing
pore sequences up to a defined water content or matric suction, it is possible
to calculate the unsaturated hydraulic conductivity.
From a statistical point of view, the pore constriction distribution is a
representative solution of an arbitrary cut through a soil and, since the pore
constriction distribution shows per definition a priori hydraulically active pore
openings, it is not necessary to combine them in the sense mentioned above.
This means that for the calculation of the hydraulic conductivity it is feasi-
ble to use the equation directly according to Hagen–Poiseuille, as shown in
equation (3).
Equation (3) can easily be transformed into a form for the definition of the
hydraulic conductivity according to Darcy. Since the term (Δp/l) corresponds
to the hydraulic gradient i, the flow through a capillary is defined only by the
form factor fHP and the dynamic viscosity η. By equitation of equation (3)
with the well-known Darcy equation Q = kf iA equation (4) is obtained, which
describes the saturated hydraulic conductivity for all capillaries of number m
considered as a bundle representative of a soil with a total surface area of AB
in a representative soil cut.
ΔV Δp 1
Q= = fHP (3)
Δt l η
Q discharge (m3 /s) l length of the capillary (m)

ΔV flow volume (m3 ) η dynamic viscosity
Δt time interval (s) (10−6 kNs/m2 ) (cf. Table 2)
Δp pressure difference at the fHP form factor (m4 ), which is
end of the capillary (kN/m2 ) (rP4 E π)/8 for circular cross
section
Table 2. Dynamic viscosity of water depending on the temperature
Temperature (◦ C) 10 15 20 30
Dynamic viscosity
1.306 1.138 1.002 0.798
η (10−6 kNs/m2 )
' 1 1
kf = fHP,m (4)
η AB
With equation (4) the saturated hydraulic conductivity of a soil considered

as a bundle of capillaries is well defined. However, two very important proper-
ties of a porous media have been neglected so far, namely the tortuosity of the
pore channels and the connectivity among them. In the well known equation
according to Mualem, both soil properties are combined in one parameter,
for which Mualem originally proposed the constant value 0.5 based on exper-
iments. However, more and more frequently this parameter is being used as a
fitting parameter, as a result of which the original physical meaning has been
lost. Nevertheless, both the tortuosity and the connectivity are dependent on
the water content, which has to be considered in the calculation.
The tortuosity T0 describes the relationship between the distance of two
points LP and the actual length of the pore channel LP e between these two
points. Most frequently the tortuosity is defined as shown in equation (5) de-
creasing the saturated hydraulic conductivity (cf. Bear 1972). For a saturated
condition it is feasible to use T0 = 2/3 as a constant value for the tortuosity.
2
LP
T0 = ≤1 (5)
LP e
When considering the connectivity it is more difficult, because as a topo-
logic property of a porous medium it must be determined based on microstruc-
tural investigations or by using the Euler–Poincaré characterisation (cf. Vogel
and Roth 1998). However the connectivity K can be considered in context
with the relationship between the specific surface of a porous medium and
its pore volume (Vasconcelos 1998). Based on this approach, it is possible
to assess a constant value for K assuming constant conditions regarding the
pore volume. For this a cubic arrangement of same sized particles with the
radius rP resulting in a maximum pore surface is considered. For such a pack-
ing of particles the pore constriction between the particles has a radius of
rP E = 0.4142rP . Finally the connectivity results from the relationship be-
tween the inner surface of the pore and the pore length, and a value for the
connectivity of about K = 2.6 is determined.
The tortuosity decreases the hydraulic conductivity, whereas the connec-
tivity has an increasing influence. However, as mentioned before, both are
dependent on the water content or the saturation of the soil, whereby the
influence of the saturation prevails over the progression of the unsaturated
hydraulic conductivity. As a result of this consideration, the factor fT K , as
shown in equation (6), takes into account the influence of tortuosity T0 and
connectivity K as the multiplicand for equation (4). A comparison between
the calculated curve and a curve determined using the model according to
Mualem/van Genuchten based on experimental data is given in Fig. 5.
θ
fT K = T0 KS 2 = 1.73S 2 with S= (6)
θS
As can be seen from Fig. 5, there is a satisfactorily good match between
the progression of the unsaturated hydraulic conductivity determined with
the model of Mualem/van Genuchten and the calculated curve based on the
pore constriction distribution. But even more astonishing is the direct match
of the saturated hydraulic conductivity calculated with the presented proce-
dure. Naturally the hydraulic conductivity depends on the density of the soil.
Concerning this additional factor, for the saturated hydraulic conductivity a
variance of 10−4 ≤ kf ≤ 8 × 10−4 m/s is obtained, which is a good estimation
of this soil property.
4 Summary and Conclusion

A new procedure for the determination of the soil hydraulic properties based
on the particle size distribution has been presented. It uses the pore constric-
Fig. 5. Progression of the calculated unsaturated hydraulic conductivity from the

pore constriction distribution in comparison to the curve determined with the pa-
rameterisation according to Mualem/van Genuchten under consideration of the mea-
sured saturated hydraulic conductivity of the weak gravelly sand
tion distribution, which has to be calculated separately taking the density of

the soil into consideration. Using the equation for the capillary rise of water,
the SWRC can be calculated very easily from the pore constriction distribu-
tion. The calculation of the unsaturated hydraulic conductivity is based on a
consideration of the soil as a bundle of capillaries. In this way it is possible
to use the equation according to Hagen-Poiseuille for the flow in capillary
tubes, in order to determine the hydraulic conductivity of the soil. However,
for this purpose the influence of tortuosity and connectivity must be taken
into account.
The comparisons between calculated and measured curves matched satis-
factorily well. The introduced procedure was also tested with other granular
materials up to silty materials and it was also compared with other pedotrans-
fer functions based on the particle size distribution (cf. Scheuermann 2005).
Thus this procedure has proved to be in constant accordance over the range
of tested materials.
One great advantage of the procedure presented is its independence from
empirical parameters. The shape of both curves of the soil hydraulic properties
is the result of the pore constriction distribution, which is directly influenced
by the assumed density of the soil. The saturated water content of the SWRC
is given by the porosity of the soil, and the residual water content can be au-
tomatically calculated or predefined using a measured residual water content
or an estimated matric suction.
Furthermore, this procedure also provides a possibility to determine the

imbibition SWRC. A concluding step in this connection could be the investi-
gation of the hysteresis of the soil hydraulic parameters.
References
Arya LM, Paris JF (1981) A physicoempirical model to predict the soil moisture
characteristic from particle-size distribution and bulk density data, Soil Sci Soc
Am J 45:1023–1030
Aubertin M, Mbonimpa M, Bussière, Chapui RP (2003) A model to predict the
water retention curve from basic geotechnical properties, Can Geotech J 40:
1104–1122
Bear J (1972) Dynamics of fluids in porous media, Dover Publications, Inc., New
York
Brooks RH, Corey AT (1964) Hydraulic properties of porous media, Hydrology
Papers, Colorado State University, Fort Collins, Colorado
Childs EC, Collis-George N (1950) The permeability of porous materials, Proc Roy
Soc A 201:392–405
Fredlund DG, Rahardjo H (1993) Soil mechanics for unsaturated soils, John Wiley &
Son, Inc
Fredlund MD, Wilson GW, Fredlund DG (2002) Use of the grain-size distribution
for estimation of the soil-water characteristic curve, Can Geotech J 39:1103–1117
Glantz R (1997) Porennetzwerke von Erdstoff-Filtern – Mathematisch-
Morphologische Beschreibung kernspintomographischer Aufnahmen. Ph.D
Thesis, Mitteilungen der Abteilung Erddammbau und Deponiebau am Institut
für Bodenmechanik und Felsmechanik der Universität Karlsruhe, Heft 9
Haverkamp R, Parlange J-Y (1986) Predicting the water retention curve from
particle-size distribution: 1. Sandy soils without organic matter, Soil Science
142(6):325–339
Kitamura R, Fukuhara S, Uemura K, Kisanuki G, Seyama M (1998) A numerical
model for seepage through unsaturated soil, Soils and Foundation 38(4):261–265
Kitamura R, Seyama M, Abe H (2000) Investigation of seepage behaviour through
unsaturated soil, Unsaturated Soils for Asia. Balkema, Rotterdam, pp 405–408
Luckner L, van Genuchten MTh, Nielsen DR (1989) A consistent set of paramet-
ric models for the two-phase flow of immiscible fluids in the subsurface, Water
Resour Res 25(10):2187–2193
Mualem Y (1976) Hysteretical models for prediction of the hydraulic conductivity
of unsaturated porous media, Water Resour Res 12(6):1248–1254
Scheuermann A (2005) Instationäre Durchfeuchtung quasi-homogener Deiche, Mit-
teilungen des Institutes für Bodenmechanik und Felsmechanik der Universität
Karlsruhe, Heft 164
Schick P (2002) Die pF-Kurve bindiger Böden bei grossen Wasserspannungen,
Bautechnik 79(12):842–849
Schubert H (1982) Kapillarität in porösen Feststoffsystemen. Springer-Verlag,
Berlin, Heidelberg, New York
Schuler U (1997) Bemessung von Erdstoff-Filtern unter besonderer Berücksichti-
gung der Parameterstreuung, Mitteilungen des Institutes für Bodenmechanik
und Felsmechanik der Universität Karlsruhe, Heft 143
Schulze B (1992) Injektionssohlen – Theoretische und experimentelle Untersuchun-

gen zur Erhöhung der Zuverlässigkeit, Mitteilungen des Institutes für Boden-
mechanik und Felsmechanik der Universität Karlsruhe, Heft 126
Silveira A (1965) An analysis of the problem of washing through in protective filters.
In: Proc 6th Int Conf on SMFE, Montreal
van Genuchten (1980) A closed-form equation for predicting the hydraulic conduc-
tivity of unsaturated soils, Soil Sci Soc Am J 44:892–898
Vasconcelos WL (1998) Connectivity in sol-gel silica glasses, Quimica Nova
21(4):514–516
Vogel HJ, Roth K (1998) A new approach for determining effective soil hydraulic
functions, Eur J Soil Sci 49:547–556
Witt KJ (1986) Filtrationsverhalten und Bemessung von Erdstoff-Filtern, Mitteilun-
gen des Institutes für Bodenmechanik und Felsmechanik der Universität Karl-
sruhe, Heft 104
Wittmann L (1980) Filtrations- und Transportphänomene in porösen Medien, Mit-
teilungen des Institutes für Bodenmechanik und Felsmechanik der Universität
Karlsruhe, Heft 86
Ziems J (1969) Beitrag zur Kontakterosion nichtbindiger Erdstoffe, Dissertation an
der Technischen Universität Dresden
Zou Y (2003) Ein physikalisches Modell der pF-Kurve für teilgesaittigte grobkörnige
Böden, Bautechnik 80(12):913–921
Zou Y (2004) Ein erweitertes physikalisches Modell der pF-Kurve für teilgesät-
tigte grobkörnige Böden auf primäre Entwässerung und skundäre Be- und En-
twässerung, Bautechnik 81(5):371–378
Part VII
Field Applications
Earthquake-Induced Mudflow Mechanism from
a Viewpoint of Unsaturated Soil Dynamics
Motoki Kazama and Toshiyasu Unno
Tohoku University, Graduate School of Engineering, 6-6-06, Aoba, Aramaki,

Aoba-ku, Sendai, Japan 980-8579
m-kazama@civil.tohoku.ac.jp, unno@soil1.civil.tohoku.ac.jp
Summary. This paper discusses the general liquefaction state of unsaturated soil
related to the mudflow type slope failures observed during earthquakes in areas
covered with volcanic ash sand deposits. It is found that the volume compressibility
of soil structure, the degree of saturation and the confining pressure are key factors
governing the liquefaction of unsaturated soils.
Key words: cyclic shear, suction change, air volume change, liquefaction, mudflow
1 Introduction
Japan has had many experiences of mudflow type slope failures involving
volcanic sandy soil during the several earthquakes. It has been suspected that
most of the soil of the failed slopes was under unsaturated condition. So far,
the unsaturated condition has been assumed tacitly safer against cyclic shear
because of the high compressibility of the pore air. Therefore, little attentions
has been paid to the shear strength reduction of unsaturated soils subjected to
cyclic shear in practical engineering. In facts, it is known that when the degree
of saturation decreases to 90%, the cyclic shear strength is double that of fully
saturated soil under ordinary testing conditions in the case of fine clean sand
Yoshimi et al. (1989). In this paper, firstly the necessity of the research on
unsaturated soil dynamics is explained. Secondary, the liquefaction state of
unsaturated soils as a three phase material is discussed based on the results
of cyclic triaxial test for fine clean sand, with pore air and water responses
taken into consideration.
1.1 Mudflow Type Slope Failure During Earthquakes
Mudflow type failures of artificial fill and natural slope have occurred dur-
ing several earthquakes around the world. Such failures occur especially in
regions covered with the volcanic ash sandy soil. Las Colinas landslide, which
438 Motoki Kazama and Toshiyasu Unno
Fig. 1. Mudflow type slope failure during the 2003 earthquake in Japan
occurred in El Salvador in 2001 earthquake and resulted in the deaths of over

500 people, is a typical example Konagai et al. (2002). Figure 1 shows an
example in Japan which occurred during a 2003 earthquake Uzuoka et al.
(2004). The collapsed portion was about 40 m wide and 80 m long, with a
depth of about 5 m and an averaged original slope angle of about 7 degrees.
It is estimated that the volume of the collapsed soil was about 8,100 m3 . The
mudflow destroyed some houses in its path, but fortunately, there was no loss
of life.
According to the investigation report, the flowed fill material was a pyro-
clastic sediment that was classified as a volcanic sandy soil with pumice, and
it was unsaturated. Therefore, in the first stage of the research, the authors
conducted laboratory shaking table tests to examine the change of the water
retention nature of this volcanic sand under an unsaturated condition Unno et
al. (2006). It was found that the apparent volumetric water content increased
by shaking when the water content was at a certain level. The level of water
content corresponded to an in-situ one. This observed behaviour is thought
to be attributable to the peculiarity of volcanic sandy soil with pumice, and
may be related to mudflow-type slope failure.
1.2 Why Is Unsaturated Soil Dynamics Important?
Past experiences implied that the water retention nature of volcanic sandy soil
plays an important role in mudflow type slope failures during earthquakes.
This can be explained by soil water characteristic curves. Figure 2 shows the
comparison of physical properties between an ordinary fine clean sand and
volcanic sandy soil. These two soils are actual soils we have used in experi-
ments.
As shown in the figure, because volcanic sand particles have many micro
cavities, its capacity to retain water is much larger than fine clean sand. It
Earthquake-Induced Mudflow from Unsaturated Soil Dynamics 439
Fig. 2. Comparison of properties between fine clean sand and volcanic sand
results in a high degree of saturation several meters from the free water ta-
ble. This is a clear indication of the need to study dynamics of unsaturated
soils. In the second stage of the research, using the volcanic sandy soil samples
from slope failure site, the authors performed cyclic triaxial tests under the
unsaturated condition Kazama et al. (2006). In this paper, the general lique-
faction state of unsaturated sandy soil is discussed, using a fine clean sand as
a representative sand.
2 Cyclic Triaxial Test for Unsaturated Sandy Soils

2.1 Physical Properties of Soils used in the study
As was mentioned in a previous section, for the under unsaturated condition,

the cyclic behavior of soils with a high water retention capacity is much more
important than that of clean sand from a practical engineering viewpoint.
However, the test results of fine clean sand are presented in this section to
explain the general liquefaction state for unsaturated soils. Figure 2 shows
the grain size distribution of the fine clean sand used in this study. The sand,
called Toyoura sand, is representative of fine clean sand in Japan. It has no
fine content, and density of soil particle is 2.643 g/cm3 . Its maximum and
minimum void ratios are 0.967 and 0.956, respectively. From Figure 2, air
entry value can be seen to be 4 to 5 kPa.
2.2 Testing Method
We have conducted cyclic triaxial test for unsaturated soils under the undrai-
ned condition and measured the pore air and water pressure responses. The
initial soil conditions before cyclic shear loading, such as dry density (Dr0 =
60% and 26%), the degree of saturation (0–100%), the confining pressure
(σnet0 = 20 kPa and 60 kPa in target value) and initial suction (0–11 kPa),
were the testing parameters.
Step motor
Cell press Displacement
gage gauge
Air water
pressure line tank
Cell press Load cell
controler
Solenoid valve
& Standard
Pore air press water head
transducer(in)
G-FRP cloth
High air
Specimen entry disk
AEV=200kPa
Water press
controller
:2-way valve Press gauge
(cell press)
Trap Water line Pore water

Pore air press water head gauge volume controller
transducer (out)
Pore water press transducer
Fig. 3. Cyclic triaxial test apparatus for unsaturated soils
Testing Apparatus
Figure 3 shows a testing apparatus used in this study. The specimens were
d = 50 mm in diameter and h = 100 mm in height. A grass fiber filter and a
ceramic disk with an AEV of 200 kPa were installed at the top and bottom
of specimen, respectively. The pore air pressure during cyclic shear was mea-
sured by the air pressure transducer attached directly above the specimen. A
solenoid valve is attached at the right after in order to avoid the effect of the
aerial compressibility in the pipe line. The volume of the inner pipe from top
of the specimen to the solenoid valve is 0.18 cm3 , which is small enough for
accuracy.
Specimen Preparation Method
The method for making the initial condition was as follows. To begin with,
75 cc of water was put into the mold. This corresponds to 95% saturation, or a
water content of about 25%. Next, dried sand was dropped through air into the
mold. In this condition, the sand absorbs the water, and the specimen becomes
a uniform moisture state. In the consolidation process, confining pressure is
applied step by step, as shown in Fig. 4. Because of the difficulty controlling
the air pressure to achieve the prescribed initial suction state, we controlled
the drained water volume. That is, when the target degree of saturation was
achieved during a consolidation step, the route of the pore water was closed.
After that, the air pressure was controlled to achieve a pore water pressure of
98 kPa ( = atmospheric pressure), as shown in Fig. 4.
Fig. 4. Specimen consolidation process for achieving initial suction state
Loading Condition
After making the initial isotropical stress state as described, cyclic axial strain
was applied under the undrained condition. The axial strain single amplitude
of the sinusoidal wave was 0.2, 0.4, 0.8, 1.2, 1.6, 2.0% with every ten cycles.
The concept of a strain controlled cyclic shear test is shown in the literature
by the first author Kazama et al. (2000). The loading frequency was 0.005 Hz,
which was slow enough allow the pore air and pore water response for clean
sand to be followed. This point was confirmed in a preliminary test.
Definition of Stress State Variables
It is well known that there are many definitions of effective stress for unsatu-
rated soils. For simplicity, in this study, Bishop’s proposed equation was used
to evaluate the effective stress Bishop et al. (1960). In the equation, the degree
of saturation at initial state was adopted as a suction parameter. Therefore,
Sr0
σm = (σc − ua ) + (ua − uw ) (1)
100
By using the definition above, the effective stress reduction ratio during

the cyclic shear can be determined as 1 − (σm /σm0 ), where σm0 is the initial
mean effective stress before cyclic shear. This index indicates the degree of
effective stress loss ranging from zero to unity, which corresponds that of the
initial state to a zero effective stress state due to cyclic shear.
3 Test Results and Discussion
3.1 Representative Test Results
Figure 5 shows an example of a stress strain relationship and an effective

stress path for unsaturated soil with a relatively low degree of saturation. It
is noteworthy that, even in the case where the degree of saturation is consid-
ered relatively small, the soil particle skeleton is degraded by the cyclic shear
Fig. 5. Test results (Dr0 = 60%, Sr0 = 85%, ua0 = 6.0 kPa, σnet0 = 17.2 kPa)
and reaches to the zero effective stress state, thereby causing a failure of the
microstructure and engendering the reduction of the soil shear strength.
Figure 6 shows the pore air and water pressure response during the cyclic
shear. In the figure, the difference between ua and uw , which is indicated
by shading in the figure, represents the suction. It is found that pore air
pressure gradually increased and reached to the initial mean confining stress
at around 3000 seconds. At this point, the net stress reaches zero. The pore
water pressure also gradually increased but with relatively large fluctuation
and reached the initial mean confining stress at around 7000 seconds. At this
point, it can be regarded that netstress and suction contribution to effective
stress were completely diminished.
Figure 7 shows the time histories of effective stress reduction ratio for
several specimens. As shown in the figure, when the same axial strain history
is applied, it is more difficult to reach the effective stress reduction ratio to
unity, when the degree of saturation is lower, and the initial confining stress
and relative density are higher. Consequently, the final effective stress after
cyclic shear of all cases can be written as a function of the initial degree of
saturation, as shown in Fig. 8. The liquefaction of unsaturated soils is affected
Fig. 6. Time histories of pore air pressure and pore water pressure during the cyclic
shear (Dr0 = 60%, Sr0 = 85%, ua0 = 6.0 kPa, σnet0 = 17.2 kPa)
Fig. 7. Time histories of effective stress reduction ratio (Dr0 = 60% and 26%)
Fig. 8. Effective stress reduction ratio versus the initial degree of saturation
not only by the volume compressibility of the soil structure, which is reflected
by dry density, but also by the degree of saturation and the initial confining
pressure.
3.2 Discussion on the Liquefaction State for Unsaturated Soils

Based on the test results and equation (1), it can be understood that a com-
plete liquefaction state for unsaturated soils is the condition in which both
pore air and water pressure are at the same pressure as the initial mean total
confining pressure. If we consider the volume change of pore air ΔVa between
the initial and final full liquefaction states, and if the pore air is assumed to
be an ideal gas, the following equation can be obtained.

ua0 Va0 = σm0 (Va0 − ΔVa ) (2)
where Va0 is the initial volume of pore air. The volume change of pore air
ΔVa represents the volume change of the soil particle structure required to
cause complete liquefaction. The relationship expressed in equation (2) can
be rewritten by the change of void ratio as follows:

ua0
e0 − eua =σm0
= 1− (1 − Sr0 /100)e0 . (3)
σm0
It can be understood that void ratio change required to cause liquefaction

is a function of the initial degree of saturation and confining stress.
4 Conclusions
1) Generally speaking, because volcanic sand has a higher water retention

capacity than ordinary fine clean sand, the unsaturated zone with a high
degree of saturation is thicker. This points to the necessity of further in-
vestigations into unsaturated soil dynamics in order to develop an under-
standing of the mechanism of the mudflow type slope failure.
2) The liquefaction state for unsaturated soil can be defined by the same
concept of effective stress used in saturated soil. That is, the liquefaction
state is the zero effective stress state.
3) For unsaturated soils subjected to cyclic shear under the undrained condi-
tion, both pore air and pore water pressure can reach the initial confining
stress, when the volume compressibility of the soil particle structure is high
enough.
4) The liquefaction of unsaturated soils is affected not only by the volume
compressibility of the soil structure but also by the degree of saturation
and initial confining pressure.
References
Bishop AW, Alpan I, Blight GE, Donald IB (1960) Factors controlling the strength
of partly saturated cohesive soils, Proc Colorad Conference 503–532
Kazama M, Takamura H, Unno T, Sento N, Uzuoka R (2006) Liquefaction mecha-
nism of unsaturated volcanic sandy soils, J of Geotechnical Engineering, JSCE
62(2):546–561 (in Japanese)
Kazama M, Yanagisawa E, Yamaguchi A (2000) Liquefaction resistance from a duc-
tility viewpoint, Soils and Foundations 40(6):47–60
Konagai K et al. (2002) Las Colinas Landslide Caused by the January 12, 2001 off
the Coast of El Salvador Earthquake, J of JAEE 2(1):1–12
Unno T, Kazama M, Uzuoka R, Sento N (2006) Change of moisture and suction
properties of volcanic sand induced by shaking disturbance, Soils and Founda-
tions 46(4):519–528
Uzuoka R, Sento N, Kazama M, Unno T (2004) Landslides during the earthquakes
on May 26 and July 26, 2003 in Miyagi, Japan, Soil and Foundations 45(4):149–
163
Yoshimi Y, Tanaka K and Tokimatsu K (1989) Liquefection resistence of a partially
saturated sand, Soils and Foundations, 29(3):157–162
Plate-Load Tests on an Unsaturated Lean Clay
Juan Carlos Rojas1 , Luis Mauricio Salinas2 , and Claudia Sejas2

1
Department of Geotechnical Engineering, University of Naples Federico II,
Naples, Italy rojasvidovic@unina.it
2
Laboratorio de Geotecnia, Universidad Mayor de San Simón, Cochabamba,
Bolivia msalinas@fcyt.umss.edu.bo, gtumss@fcyt.umss.edu.bo
Summary. Plate-load tests are a common field method to estimate soil bearing
pressure. Classic-result-test interpretations do not consider the matric suction effect.
This paper analyzes the matric suction effect in bearing pressure from plate-load
tests conducted in a lean clay soil. Seven plate-load tests were carried on in an
unsaturated lean clay deposit; two of them were performed in saturated condition
to verify the site homogeneity and test repeatability, five of them took into account
different matric-suction values. Matric suction along the plate influence zone was
monitored with four Jet Fill tensiometers, inserted at 0.1, 0.3, 0.6 and 0.9 m around
every test pit. The results show bearing pressure values are highly influenced by
matric suction changes.
Key words: plate-load test, bearing pressure, unsaturated soils, field tests
Introduction
The most reliable method to obtain the ultimate bearing capacity at a uniform
site is a load test. These kinds of tests are often used to predict the response
of buildings foundations and road pavements. If a footing is placed in the
capillary zone, the soil below is not fully saturated and a negative pore-water
pressure, uw , is present. However, in the design practice, the bearing capacity
of the soil is underestimated because of the actual higher level of stress is not
taken into consideration.
The aim of the research was to quantify the effect of matric suction on
the in-situ-bearing capacity obtained by means of a plate-load test. A com-
prehensive experimental testing program was carried out in a field site at
Cochabamba, Bolivia on lean clay (CL). Seven tests were conducted under
different matric suction values. Matric suction was measured using tensiome-
ters installed at different depths covering the zone of load influence.
446 J. C. Rojas et al.
Previous Studies
Previous studies considering plate-load tests on unsaturated soils have been
published; Table 1 shows a summary of the highlights of these projects. The
literature related with deals mainly with four variables: plate size, soil stiffness
variation, loading rate and ultimate load definition.
Table 1. Details of plate-load tests performed on unsaturated soils
Author Soil type Suction evaluation Plate base [m]

Steensen-Bach et al. (1987) sand tensiometers 0.1† , 0.2†
Larson (1997) silt w correlation 0.5∗ , 1∗ , 2∗
Conciani et al. (1998) collapsible clay tensiometers 0.8†
Rojas and Salinas (2002) lean clay w correlation 0.8†
Costa et al. (2003) clayey sand tensiometers 0.8†
Xu (2004) expansive clay w correlation 0.3∗
Mohamed and Vanapalli (2006) sand tensiometers 0.1∗ , 0.15∗
∗ †
square plates; circular plates
Site Characterization
The test site is located in the city of Cochabamba, Bolivia. Figures 1a and 1b
show the results of cone penetration tests, CPT. The soil profile is presented
in Fig. 1c to a depth of 3 m. The ground water table was at a depth of about
2.6 m during the cone penetration tests. Fig. 1c presents the average results
of characterization tests executed in samples gathered from 8 spread points,
at profundities of 1.5 m and 2 m.
Figure 2 shows the pressure-plate test data measured by Pardo and Sali-
nas (2006) and the drying branch of the soil-water retention curves, SWRC,
obtained following two methods: Fredlund et al. (1997) physico-empirical es-
timation and Zapata et al. (2000) correlation.
Field Experimental Program

The program was designed with the objective of determining the ultimate
bearing capacity of a clayey soil in saturated and unsaturated conditions.
Seven pits, 1.2 × 3.0 m2 , were excavated up to 1.4 m of depth in order to
assure a homogeneous clayey soil profile under the plate and to fulfil ASTM
criteria of free space next to the plate (i.e. 4 times the diameter of the plate).
Four tensiometers were installed in each pit at depths of 0.1, 0.3, 0.6 and 0.9 m
below the base of the pit.
Plate-Load Tests on an Unsaturated Lean Clay 447
Fig. 1. CPT test results and the first 3 m of soil profile
Fig. 2. Estimated soil-water retention curves and matric suctions measurements
Two out of seven excavations (i.e. pits S1 and S2) were filled with water at
a constant level and suction was measured; in such a manner the repeatability
and deviations of the soil conditions and properties in saturated conditions
were verified. It was considered saturation in the soil profile was reached when
the suction in the four tensiometers reached the zero value. The rest of the
excavations (i.e. pits U1 to U5) were also filled with a constant level of water;
nevertheless, it was taken away before a zero value was observed. This proce-
dure ended up with different suction values in the rest of the five pits.
Seven plate load tests were carried out using a 0.31-m-rigid-circular-steel
plate, following ASTM D 1194 standard, a world-wide-accepted procedure.
The counter load, 200 kN, was given by a Mercedes Benz truck, prepared by
Ap Van Den Berg for this kind of tests. During the execution of each test,
measurements of vertical strain were taken until a maximum deformation
of 0.03 mm/min for 3 consecutives measurement was read. The number of
load increments that was set for the tests was taken with the principle that
each increment of load should not be greater than 10% the maximum load
to be reached (i.e. 650 kPa due to manometer capacity of the hydraulic jack).
In addition, the tests were stopped when the maximum expected pressure
was reached or a deformation of 31 mm was observed (i.e. 10% of the plate
diameter).
The results obtained from CPT tests (Fig. 1) and plate-load tests S1 and S2
(Fig. 3), verified a reasonably repeatability of test results and a low variability
of the soil deposit in the selected site.
Figure 3 shows the results of the 7 plate-load tests. As it is expected in
this type of soils, none of them show a peak load, therefore, the peak load
Fig. 3. Pressure-settlement curves

has been estimated by extrapolation of the load-settlement curves using the

hyperbolic equation proposed by Kondner (1970).
For homogeneous saturated clay the bearing capacity is considered in-
dependent of the foundation breadth. Consoli et al. (1998) determined the
effect of footing size on bearing capacity of unsaturated low plasticity clay
(CL), and concluded that the effect of the size of the load area in the mea-
sured settlement and bearing capacity was shown to be negligible. Thus, the
stress-strain characteristics obtained by means bearing capacity do not need
corrections. However, the saturated approach assumes constant shear-strain
modulus, E, but his variability increases under unsaturated conditions, due
to matric suction variations with depth.
The ultimate bearing capacity, qP L , is also defined as the pressure level
that corresponds at predefined settlement or to the onset of permanent soil
deformation. The standard method ASTM D1194 consider as qP L value the
pressure corresponding to a settlement equal to 10% of plate diameter, d.
Another criterion is to estimate the qP L value as the stress defined by the
intersection of the tangents of the initial and final portions of the pressure-
settlement curve, named yield stress criteria (Steensen-Bach et al. 1987, Lar-
son 1997, Consoli et al. 1998, Costa et al. 2003, Xu 2004). Tests S1, S2 and
U1, which have the smallest values of suction, present a yield load, the other
tests were stopped before the yield load was reached. Table 2 summarizes the
bearing capacity values calculated following these criteria.
Additionally, for saturated conditions the theoretical ultimate bearing ca-
pacity, qu , of 115 kPa was calculated using Meyerhof’s method. The ratio be-
tween the experimental ultimate bearing capacity considering failure criteria,
qP Lfailure , and qu (Table 2), shows that, for saturated conditions, Meyerhof’s
Table 2. Bearing capacity values determined using different criteria
Test Suction Bearing capacity [kPa] qP Lunsat /qP Lsat qP Lfailure /qu
∗ † ∗ †
[kPa] Yield ASTM Failure Yield ASTM Failure
S1 0 122 244 450 1 1 1 3.9
S2 0 124 220 376 1 1 1 3.3
U1 10 197 316 552 1.6 1.4 1.5 4.8
U2 48 375 499 671 3.0 2.2 1.8 5.8
U3 56 670 802 961 5.4 3.5 2.5 8.3
U4 60 715 844 980 5.8 3.6 2.6 8.5
U5 63 800 968 1112 6.5 4.2 3.0 9.7
∗
yield stress criteria
†
pressure corresponding to 31 mm settlement
qP Lunsat /qP Lsat = ultimate unsaturated − saturated bearing capacity ratio
qP Lfailure /qu = field failure criteria and Meyerhof’s method ratio
method greatly under predicts the ultimate bearing capacity, similar results
were reported by Larson (1997).
Matric suction values measured before the tests started are summarized in
Table 3, together with the water content values measured on samples gathered
from similar depths. The matric suction variation during plate loading was
not registered, and is not considered in the analysis.
The field values of matric suction-water content, measured at 0.1 m depth,
follow approximately the pressure-plate data trend (Fig. 2). This fact is due
to the hysteresis of the SWRC. Fredlund et al. (1997) and Zapata et al. (2000)
methods provide a main drying curve. By the other hand, the procedure fol-
lowed by Pardo and Salinas (2006) to condition the undisturbed soil samples
before the test (i.e. addition of water on samples surface without reaching the
completely saturation) generated a drying scanning curve; then, this SWRC
reflects properly the stress history, age, and water changes of the sample.
Thus, in the cases were the water content is measured in the field to obtain
the matric suction by means of the main drying SWRC, the bearing capacity
can be overestimated compared to when the matric suction is determined by
measurements in situ.
A representative matric suction value must be defined. In previous works
where the matric suction profile was obtained, this value was establish as an
average matric suction corresponding to values registered in the stress bulb
zone, a depth beneath the plate assumed as 1 × d in Costa et al. (2003) and
1.5 × B in Mohamed and Vanapalli (2006). In this study, the matric suction
value registered at 0.1 m depth has been considered as representative of the
soil profile, considering the average value does not correspond to experimental
evidence and more studies have to be carried on. However, for this case, the
average in the stress bulb zone (i.e. 1 × d) is almost the same.
The increase of ultimate bearing capacity with matric suction is illustrated
in Fig. 4. Due to the scattered points, it is not possible to recognize a clear
trend. Mohamed and Vanapalli (2006) and Costa et al. (2003) established that
Table 3. Matric suction and water content values

Matric suction [kPa] Gravimetric water content [%]
Test at depth [m] at depth [m]
0.1 0.3 0.6 0.9 0.1 0.3 0.6

S1 0 0 0 0 27.4 24.5 23.2
S2 0 3 0 0 22.4 23.5 25.8
U1 10 4 13 0 22.8 21.0 22.6
U2 48 38 15 0 21.1 – –
U3 56 46 4 0 14.8 20.8 22.7
U4 60 53 6 4 15.6 12.1 20.9
U5 63 57 11 0 16.7 9.8 15.2
Fig. 4. Variation of bearing capacity at failure with respect to matric suction
the bearing capacity versus matric suction relationship is similar to the shear
strength behaviour of unsaturated soils (i.e. a linear increase up to the air
entry value and a nonlinear increase beyond the air entry value). Following
this criterion, a trend curve has been plotted (Fig. 4).
For the suction range covered in this work (i.e. 0 to 63 kPa), the different
criteria to define the ultimate bearing capacity, qLP , does not reflect in the
same proportion the matric suction effect (Table 2). For instance, the bearing
capacity of U4 test using yield stress criterion is six times higher than the sat-
urated conditions, four considering ASTM norm and three if failure condition
is assumed.
Conclusions
The following conclusions are applicable to the results of this experimental

program on a lean clay soil.
In order to know the magnitude of suction during a plate load test, it has
to be measured in the field. The estimation of the suction by means of the
water content of the field might increase its actual value.
For saturated conditions, the ultimate bearing capacity estimated by Mey-
erhof’s theoretical method provided lower values than the ones calculated with
the plate load tests.
The different criteria to define the ultimate bearing capacity, qLP , do not
reflect the matric suction effect in the same proportion. In the analyzed suction
range, the matric suction effect increase the maximum bearing capacity, from
plate-load test, in three to six times, depending on the criteria selected to
define the qLP .
The results suggest a considerable contribution of the matric suction to
the bearing capacity of the soil.
References
ASTM 1194-94 Standard test method for bearing capacity of soil for static load and
spread footings. 2003 Annual book of ASTM Standards 04.08.
Conciani W, Soares MM, Naime JM, Crestana S (1998) Plate load test with CT.
In: 2nd International conference on unsaturated soils, Beijing: 333–337
Consoli NC, Schnaid F, and Milititsky J (1998) Interpretation of plate load tests
on residual soil site, J Geotechnical and geoenvironmental engineering 124(9):
857–867
Costa YD, Cintra JC, Zornberg JC (2003) Influence of matric suction on the results
of plate load tests performed on a lateritic soil deposit, J Geotechnical Testing
26:219–227
Fredlund MD, Fredlund DG, Wilson GW (1997) Prediction of the soil-water char-
acteristic curve from grain-size distribution and volume-mass properties. 3rd
Brazilian symposium on unsaturated soils, Rio de Janeiro, Brazil
Kondner RL (1970) Hyperbolic stress-strain response: cohesive soils, J Soil Mechan-
ics and Foundations Division 93:283–310
Larson L (1997) Investigations and load tests in silty soils. Swedish Geotechnical
Institute, Report No. 54
Larson L (2001) Investigations and load tests in clay till. Swedish Geotechnical
Institute, Report No. 59
Mohamed FMO, Vanapalli SK (2006) Laboratory investigations for the measure-
ments of the bearing capacity of an unsaturated coarse-grained soil. Canadian
geotechnical conference, Vancouver 2006 (in print)
Pardo JM, Salinas LM (2006) Curva de retención de agua del suelo (SWRC): Análisis
de estimaciones y ajustes. IV Congreso de Ciencia del Suelo, Bolivia
Rojas JC, Salinas LM (2002) Bearing pressure and settlement for a lean clay in
saturated and unsaturated conditions. In: Proceedings of 3rd Unsaturated Soils
Conference, Recife, Vol. 2:703–708
Steensen-Bach JO, Foged N, Steenfelt JS (1987) Capillary induced stresses – fact or
fiction? In: Proceedings 9th European conference on soil mechanics and founda-
tion engineering: 83-89
Xu Y (2004) Bearing capacity of unsaturated expansive soils, J Geotechnical and
Geological Engineering 22:611–625
Zapata CE, Houston WN, Houston SL, Walsh KD (2000) Soil-water characteris-
tic curve variability. Advances in unsaturated geotechnical special publication,
No. 99
Selfsealing Barriers of Clay/Mineral Mixtures.
The SB Project at the Mont Terri Rock
Laboratory
Tilmann Rothfuchs, Rüdiger Miehe, Norbert Jockwer, and Chun-Liang

Zhang
GRS-Final Repository Research Division, Theodor-Heuss-Strasse 4,

D38122 Braunschweig, Germany
Tilmann.Rothfuchs@grs.de, Ruediger.Miehe@grs.de, Norbert.Jockwer@grs.de,
Chun-Liang.Zhang@grs.de
Summary. Clay/sand mixtures are currently being investigated at GRS’s geotech-

nical laboratory on their suitability for sealing disposal rooms containing gas gen-
erating radioactive waste. It has been found that moderately compacted mixtures
may represent a reasonable alternative to highly compacted bentonite envisaged as
buffer in various multi-barrier repository concepts. In contrast to highly compacted
buffers moderately compacted clay/sand mixtures exhibit a comparably low gas
entry/break through pressure in the saturated state while providing an adequate
sealing potential due to swelling of the clay minerals in consequence of water uptake
from the host rock. The evolution of high gas pressure in the repository near-field
due to corrosion of the waste containers will thus be avoided while possible mi-
gration of radionuclides from the waste matrix through the buffer will be diffusion
controlled. Besides laboratory investigations and mockup testing, full-scale exper-
iments are being conducted at the Mont Terri Underground Rock Laboratory to
test and demonstrate the advantageous sealing properties of moderately compacted
clay/sand-mixtures under representative in-situ conditions.
Key words: radioactive waste repository, barrier, clay/sand-mixture, swelling pres-

sure, permeability, gas entry pressure
1 Introduction
To completely seal radioactive waste from the biosphere various repository
concepts rely on a multiple engineered barrier system (EBS). Some of these
concepts involve highly compacted bentonite buffers between the waste con-
tainers and the host rock.
Considering these concepts in more detail, there is a concern about gas
pressure built-up in the near-field once formation water re-saturates the buffer
and leads to start-up of hydrogen gas generation by corrosion of the canisters.
454 T. Rothfuchs et al.
According to Rübel et al. (2003), clay formations like the Opalinus clay provide
enough water to completely corrode metallic high-level waste canisters. Up
to 481 m3 of hydrogen gas would be produced per canister by its complete
corrosion. Because of the very low permeability of the bentonite buffer after
re-saturation, gas migration is hindered and a higher gas pressure may evolve
in the repository near field. In case the least principal stress and the tensile
strength are exceeded, fracturing and disintegration of the buffer and/or the
host rock may take place.
Moderately compacted clay/sand mixtures have been found to represent a
suitable alternative to highly compacted bentonite buffer (Jockwer et al. 2000,
Miehe et al. 2003). After re-saturation, such mixtures exhibit a comparably
low gas entry/break through pressure while providing an adequate sealing
potential against intruding water due to swelling of the clay minerals after
water uptake from the host rock. By using optimized material mixtures, the
parameters of primary interest, such as permeability to gas and water, gas
entry pressure in the saturated state, and two phase-flow parameters can be
adjusted to the prevailing local conditions of the host rock in a way that the
buffer material acts as a gas vent thereby avoiding the generation of higher
gas pressure in the repository near field.
The most important properties the material mixtures must fulfil in com-
parison to those of the host rock are:
• Low permeability to water in the saturated state between 10−17 to
10−18 m2 in analogy to the permeability of the excavation disturbed zone
(EDZ) in the host rock.
• Gas entry pressure or break-through pressure lower than that of the host
rock.
• Higher permeability to gas than that of the host rock after break-through.
• Swelling pressure lower than the least principal stress in the host rock in
order not to disturb the integrity of the host rock.
The SB (Self-sealing Clay/Sand-Barriers) project started in summer 2003
comprises material characterization in the laboratory as well as the conduction
of confirmative mockup test in the laboratory and in-situ experiments at the
Mont Terri Rock Laboratory (MTRL) in Switzerland.
2 Laboratory Characterization of Moderately

Compacted Clay/Sand-Mixtures
On basis of the experiences gained in the preceding projects the laboratory
investigations within the SB-project were focused on samples with clay/sand
mixing ratios of 35/65, 50/50, and 70/30. The bentonite used is a commer-
cially available calcium-type, named Calcigel, produced from the company
Süd-Chemie AG in Germany. The sand is a quartz sand with grains less than
2 mm. The bentonite has a grain density of 2.7 g/cm3 and an initial water
Selfsealing Barriers of Clay/Mineral Mixtures 455
content of 8%, while the sand has a grain density of 2.67 g/cm3 and a wa-
ter content of 1%. Water content was determined by drying the materials in
an oven according to DIN 18121–1. The density of the mixtures achieved by
mixing bentonite and sand together in an electric mixer ranges from 1.40 to
1.88 g/cm3 .
The most interesting parameters such as permeability to gas and water,
swelling pressure, gas breakthrough pressure were measured in specially de-
veloped oedometer cells of 50 mm diameter and 50 mm.
The gas permeability in the initial dry state was determined by in-
jecting nitrogen gas. The determined data range between 1.23 × 10−13 m2
(35/65 clay/sand mixture) and 1.16 × 10−15 m2 (70/30 clay/sand mixture).
The water permeability was determined by injecting synthetic clay formation
water as well as fresh water and salt brines. The results for the different liq-
uids are compared in Fig. 1. It can be seen that the permeability to water of
the mixtures decreases with the clay content and with increasing density. The
mixtures with clay/sand ratios of 35/65 and 50/50 exhibit very low perme-
ability to clay water of less than 10−17 m2 , and thus meet the above mentioned
requirement of 10−17 –10−18 m2 for the seals. It is also interesting to see that
the permeability to fresh water is about one order of magnitude lower than
that to salt brine and clay water.
The gas break-through pressure of the investigated mixtures ranges be-
tween 0.2 and 2.8 MPa and increases clearly with the clay content. Fig-
ure 2 depicts as an example a measurement on a moderately compacted
50/50 clay/sand sample. The determined average gas permeability after gas
break-through amounts to about 10−17 m2 for 35clay/65sand samples and to
about 5 × 10−18 m2 for 50clay/50sand samples (see Table 1) and is considered
sufficiently high in comparison to the permeability of the intact Opalinus clay
which is in the order of 10−20 to 10−22 m2 .
Fig. 1. Permeability of clay/sand Fig. 2. Gas pressure and flow rate evolution
mixtures to different fluids as a func- during a gas break-through test of a moder-
tion of clay content ately compacted 50clay/50sand sample
The swelling pressures were measured after full re-saturation with syn-
thetic clay water. The average values of 0.2 to 0.4 MPa are significantly below
the upper limit of 2 MPa which represents the least principle stress at MTRL.
Average data measured on the moderately compacted clay/sand mixtures are
summarized in Table 1.
Table 1. Hydraulic data determined on moderately compacted clay/sand mixtures

in comparison to required material properties
Gas perme- Water perme- Gas break- Gas perme-

Clay/ Dry Poro- Swelling
ability at ability at full through ability after
sand density sity pressure
dry condi- saturation pressure gas break-
ratio (g/cm3 ) (%) 2 2 2 (MPa)
tions (m ) (m ) (MPa) through (m )
3 · 10−17 0.4
35/65 1.88 29.8 1 · 10−13 1 · 10−17 0.2
∼ 9 · 10−18 ∼ 1.1
1 · 10−18 0.4
50/50 1.67 37.7 5 · 10−14 5 · 10−18 0.3
∼ 4 · 10−18 ∼ 2.8
70/30 1.40 48.0 1 · 10−15 5 · 10−19 1 0.4
Requirements
1 · 10−17
high <2 high <2
∼ 1 · 10−18
In general, the investigated mixtures with bentonite contents higher than

30% and with installation densities around 1.9 g/cm3 meet the requirements
that are to be considered in the specific case of the MTRL.
3 In-Situ and Full-Scale Mockup Experiments

The SB in-situ experiments are specially designed to enable the investigation
of the sealing properties of moderately compacted clay/sand mixtures which
are most important with regard to the long-term evolution of the repository.
Four individual experiments will be conducted in boreholes of 0.31 m diameter
and 3 m depth drilled from the floor of a test niche of 5 m width, 4 m height
and 8 m length (Fig. 3). The lower part of the boreholes, the injection volume,
is filled with a porous material (e.g. sand). Above the injection volume a filter
frit is placed for ensuring a homogeneous distribution of the injected water
over the entire borehole cross section. The clay/sand-seal is installed in several
layers to a height of 1 m above the frit. At the top of the seal, a further filter
frit is installed for water and gas collection. The whole borehole will be sealed
against the ambient atmosphere by a gastight packer. The most upper part of
the test borehole is grouted for keeping the packer in place. No instruments
Fig. 3. Principle design of a SB-borehole sealing test
are installed in the SB seal itself to avoid any negative impact on the sealing
properties.
For re-saturation of the seal, synthetic clay formation water is injected
into the lower injection volume. The water flow through the seal is collected
in the upper frit by a further tube running back to the control valve panel
where water flow rate and pressure are controlled and measured.
For the design of the in-situ tests, scoping calculations were performed
using the FE code CODE-BRIGHT (UPC 2002) developed by the Technical
University of Catalonia in Barcelona. In the modelling, gas and water flow are
modelled according to Darcy’s law, the molecular diffusion of water vapour is
governed by Fick’s law, the solubility of air in water is controlled by Henry’s
law, and an elasto-plastic model (BBM) which is able to represent swelling
and shrinking features of unsaturated soils is applied for the description of
the mechanical behaviour of the mixtures. The parameters were established
on basis of preliminary laboratory tests described in (Rothfuchs et al. 2005).
Figure 4 illustrates the predicted evolution of water re-saturation of a 1 m
long moderately compacted seal of 35% clay and 65% sand in a test borehole.
According to the calculation full saturation may be reached at an injection
pressure of 1 MPa after about 170 days.
Before going in situ, full-scale mockup tests have been started in March
2005 at the GRS geotechnical laboratory, in order to test proper installation
methods for achieving the pre-determined installation density, to determine
the time needed to reach full saturation and to determine the permeability
to gas and water as well as swelling pressure, gas breakthrough pressure and
permeability after re-saturation under ideal laboratory conditions. The first
experiment under representative in-situ conditions was started at Mt. Terri
in October 2005.
120
35clay/65sand mixture
Distance to bottom (m): water injection pressure = 1 MPa
Degree of water saturation [%]

100
0
80 0.2
0.4
60
0.6
40
0.8
20
1
0
0 50 100 150 200
Time [day]
Fig. 4. Predicted evolution of water saturation in a moderately compacted

35clay/65sand seal of 1 m length
In both cases the most promising 35/65 clay/sand mixture has been
installed. The installation densities achieved amount to 1.94 g/cm3 in the
mockup and 1.91 g/cm3 in the in-situ experiment. These values are quite sim-
ilar to those of the moderately compacted laboratory samples (compare Ta-
ble 1). At the beginning of both experiments the permeability to gas of the dry
material was measured first. The determined values amount to 6.2 × 10−14 m2
for the mockup and to 3.71 × 10−14 m2 for the in situ experiment. These
values agree very well with the data determined at the small samples in the
laboratory (see Table 1).
By June 2006 about 23 litres of water have been injected in the mockup at
an injection pressure of 1.1 MPa while an amount of 22 litres has been injected
at an injection pressure of 0.4 MPa in the in-situ experiment. While the total
pressure evolution is measured at three different levels along the tube wall
and at top of the seal in the mockup it is measured in situ only at top of the
seal.
While the water pressure sensors mounted along the tube wall in the
mockup allow an assessment of the evolution of the seal saturation in the
mockup this is not possible in situ since no instruments are allowed in the
seal to avoid any bypassing of water along instrument cables. The swelling
pressure of the clay/sand mixture, however, which is a useful indicator of the
sealing properties, can be assessed in both experiments by observation of the
total pressure at the two pressure sensors mounted in both set ups at top of
the seal.
Figure 5 shows the pressure evolution in the mockup measured along the
tube wall. At the lower level, 14 cm into the seal, the pressure reached rather
quickly the level of the injection pressure of about 1.1 MPa. At the middle
level the actual value ranges around 1.05 MPa. No pressure increase can be
observed so far at the upper level at 86 cm in the seal. One of the sensors at
the seal top, however, shows a pressure increase of about 0.04 MPa (Fig. 6)
14,00
12,00 lower
Pressure [bar, abs.]

10,00
8,00
middle
6,00
4,00
upper
2,00
0,00
05
05
05
05
05
05
06
06
06
3.
5.
7.
9.
0.
2.
2.
6.
4.
.0
.0
.0
.0
.1
.1
.0
.0
.0
01
31
30
28
05
04
03
28
29
Fig. 5. Evolution of total pressure along the mockup tube wall
1,70 1,75
1,60
1,50 1,5
Pressure [bar, abs.]
Pressure [bar (abs.]
sensor sp 1
1,40
1,30 sensor sp 1 1,25
1,20
1,10 1
sensor sp 2
1,00
sensor sp 2
0,90 0,75
19 .05
19 .05
18 .05
17 06
19 .06
18 .06
18 .06
17 .06
17 06
06
03 05
01 05
31 05
29 06
06
04 05
30 05
28 05
28 06
7.
1.
6.
5.
2.
6.
3.
7.
9.
0.
2.
4.
5
1
4
.1
.1
.0
.0
.0
.0
.1
.0
.0
.0
.0
.0
.0
.0
.0
.0
.1
.1
.0
20
05
Fig. 6. Evolution of total pressure at Fig. 7. Evolution of total pressure at top

top of the mockup-seal of the in-situ seal
which is quite comparable to a similar pressure increase of about 0.05 MPa

observed in situ (Fig. 7).
Figures 6 and 7 indicate that the swelling pressures which will be reached
in the mockup and the first in-situ experiment will be in the same order of
magnitude as that determined on the laboratory test samples (compare the
value of 0.2 MPa Table 1). Thus, similar sealing properties as observed on
small samples in the laboratory can be expected in the in-situ experiment.
4 Conclusions
The preliminary results of the running SB mockup and in-situ experiment

at the MTRL indicate confirmation of the advantageous sealing properties of
moderately compacted clay/sand mixtures preliminarily determined on small
samples under ideal conditions in the laboratory. The time needed to reach
full saturation of the test seals in both the mockup and the in-situ experiment,
however, seems to exceed the predicted ones significantly. Further efforts will
be needed to clarify the observed discrepancies and to achieve appropriate
process understandding and model improvements.
Acknowledgements
The authors gratefully acknowledge the funding of the SB-project by the Ger-
man Federal Ministry of Economics and Technology (BMWi) under contract
No. 02E9894 and by the Commission of the European Communities (CEC)
under contract No. FI6W-CT-2003-508851.
References
Jockwer N, Miehe R, Müller-Lyda I (2000) Unter-suchungen zum diffusiven Trans-
port in Tonbarrieren und Tongesteinen. Gesellschaft für Anlagen- und Reaktor-
sicherheit (GRS) mbH, Köln, GRS-167
Miehe R, Kröhn P, Moog H (2003) Hydraulische Kennwerte tonhaltiger Mineral-
gemische zum Verschluss von Untertagedeponien (KENTON), Gesellschaft für
Anlagen- und Reaktorsicherheit (GRS) mbH, Köln, GRS-193
Rothfuchs T, Miehe R, Jockwer N, Zhang C-L (2005) Self-Sealing Barriers of
Clay/Mineral Mixtures in a Clay Repository – SB Experiment in the Mt. Terri
Rock Laboratory, Gesellschaft für Anlagen- und Reaktorsicherheit (GRS) mbH,
GRS-212
Rübel A, Nosek U, Müller-Lyda I, Kröhn P, Storck R (2003) Konzeptioneller Um-
gang mit Gasen im Endlager, Gesellschaft für Anlagen- und Reaktorsicherheit
(GRS) mbH, GRS-205
UPC (2002) CODE-B l media, USER’S GUIDE
Preferential Water Movement in Homogeneous
Soils
Alexander Scheuermann and Andreas Bieberstein
Institute of Soil Mechanics and Rock Mechanics, Division of Embankment Dams

and Landfill Technology, University of Karlsruhe (TU)
alexander.scheuermann@ibf.uka.de,
andreas.bieberstein@ibf.uni-karlsruhe.de
Summary. Preferential water movement is a frequently observed phenomenon in

soil science. Especially on natural slopes hydraulic behaviour can sometimes be
observed which arises from preferential water flow. If the soil consists of a more
or less homogeneous material, the preferential water movement can be caused only
by heterogeneously distributed water. A sprinkler irrigation test on a large scale
homogeneous dyke model and corresponding measurement results using different
techniques have shown on the one hand, that preferential water flow in the form
of fingering possibly occurred during the experiment. On the other hand, lateral
water movement due to heterogeneously distributed water content inside the dyke
body could be observed. Both phenomena were also investigated in simple laboratory
experiments. The contribution presents results from the above mentioned experiment
which are analysed and discussed.
Key words: preferential flow, fingering, capillary barrier effect, experimental in-
vestigation
1 Introduction
Preferential water flow is mainly seen to arise due to structural heterogeneities

of the soil. The random stratification – especially in granular soils – can cause
capillary barrier effects governed by the unsaturated hydraulic properties of
the underlying coarser soil (Stormont and Clifford 1999). Water movement
across the interface occurs in this connection when the matric suction at the
interface reaches the so-called breakthrough head of the underlying coarser soil
layer, defined as the matric suction when the coarser layer becomes conduc-
tive. The matric suction corresponds in this case approximately to the water
entry value ψWEV which can be estimated from the appropriate soil water
imbibition curve (Baker and Hillel 1990). In this sense the breakthrough of
water only depends on the soil hydraulic properties of the coarser material and
is independent of the properties of the overlying finer soil and the hydraulic
boundary conditions.
In a similar manner, preferential water flow in the form of fingering can
occur in stratified soils. The development of fingers at the interface of two
adjacent layers or at the transition from wet to underlying dry conditions can
also be considered as a kind of breakthrough. Fingering is well known as the
result of an unstable wetting front (Hill and Parlange 1972) caused by un-
favourable conditions above or below the wetting front (e.g. water infiltration
during overpressure in the air phase). According to Hillel and Baker (1988), a
concentrated water flow in the form of fingering can occur, if the unsaturated
hydraulic conductivity at the water entry value ku (ψWEV ) of the underlying
layer or area is higher than the prevailing infiltration rate from above. As is
generally known, water infiltrates into a soil much more easily if the soil is
in an initially wet condition. Thus in this case the development of fingers de-
pends on the soil hydraulic properties of the underlying layer and the initial
moisture condition as well as on the hydraulic boundary conditions.
Consequently both effects are based on the same physical property, namely
on the water entry value of the underlying soil. Hence it is not surprising that
both phenomena can occur similarly where there are stratified soil conditions
(Walter et al. 2000). Differences in the water content distribution can intensify
such hydraulic behaviour of soils (Stephens and Heermann 1988). But also in
homogeneous soils, preferential flow in the form of lateral water movement
can occur due to anisotropic water content distribution (McCord et al. 1991).
The triggering factor in this connection is the sudden change of the matric
suction at the water front.
2 Observations on the Full-Scale Dyke Model

Measurement results during a sprinkler irrigation test on a full-scale dyke
model have indicated that both phenomena possibly occurred simultaneously
during the experiment. In this connection, a newly developed moisture mea-
surement technique based on Time Domain Reflectometry (TDR) was applied
for the first time on such a scale in order to observe the spatially distributed
soil moisture content (Spatial-TDR). Additional laboratory experiments were
carried out in order to verify the observations on the dyke model. In the fol-
lowing the dyke model as well as the measuring technique Spatial-TDR will
be described. Later on the results from the experiment on the full-scale dyke
model will be presented and discussed in comparison to the results of the
laboratory experiments.
2.1 Design and Instrumentation of the Model
In order to investigate transient seepage through dykes, the full-scale dyke

model in Fig. 1 at the Federal Waterways and Research Institute in Karlsruhe
Preferential Water Movement in Homogeneous Soils 463
Fig. 1. Full-scale dyke model at the Federal Waterways and Research Institute in
Karlsruhe (upstream side on the right)
was made available for the experiment. The dyke is constructed homoge-
neously with uniform sand (grain size 0.2–2 mm) and is situated on a base
consisting of waterproof plastic sheets, so that water infiltrating into the dyke
body will flow to a drain at the toe of the downstream slope. The slope in-
clinations are 1:2.0 on the upstream side and 1:2.25 on the downstream side.
Both slopes of the dyke are covered with an approximately 20 cm thick top-
soil layer. The discharge of the water collected in the drain is recorded in the
monitoring and data collection container.
In order to assess the pore water pressures within the dyke body, the model
is equipped with piezometer gauges which are also distributed along the base
of the dyke. Furthermore, the condition of the capillary forces can be observed
using tensiometers vertically installed in one profile (cf. Fig. 2).
With the aim of observing water content changes within the dyke body,
a new monitoring system based on Time Domain Reflectometry (TDR) was
installed in the model. The advantage of this system is its high resolution
both in space and time which is required for monitoring transient water con-
tent changes. The system consists of 12 flat ribbon cables from 1 to 3 m in
length which are installed vertically inside the dyke (cf. Fig. 3). They are con-
nected on both sides with coaxial cables to a multiplexer and a TDR-device
in a box on the crest of the dyke. Using adapted programmes a conventional
PC, located in the container at the toe of the downstream slope, takes over
data collection and controlling of the measuring equipment (multiplexer and
TDR-device). With this system the data acquisition time for the whole cross
section is only 5 minutes. The changes in water content are determined with a
spatial accuracy of about 3 cm and an accuracy of ±2 Vol-% (volumetric water
content) compared to independent water content measurements. Below, the
TDR measuring method and the modifications for the spatial assessment of
the water content along the extended sensors will be described.
2.2 Spatial Time Domain Reflectometry (Spatial-TDR)
Time Domain Reflectometry (TDR) is a widely used method for the determi-
nation of the water content in soils. With the conventional method, a fast rise
voltage step is launched from a TDR-device into a sensor consisting of two
or three wires of non-insulated metallic rods buried into the soil (cf. Fig. 4
on the left). At the transition between coaxial cable and sensor a portion of
the signal is reflected. The remaining signal travels along the sensor until it is
completely reflected at the open end. An oscilloscope within the TDR-device
records the sum of the incident signal and the reflected signal, also called the
TDR trace, from which the travel time and the mean wave velocity in the
sensor can be determined.
The permittivity of the surrounding soil can be calculated from the wave
velocity. The permittivity of the soil is dependent on the dielectric properties
of the soil constituents: air (1), grain (3–8) and water (80). Therefore, it is
possible to determine the volumetric water content using appropriate calibra-
tion functions (e.g. Topp et al. 1980). As the use of non-insulated metallic
rods as sensors for TDR measurements in soils is limited to about 1 m, a flat
ribbon cable has been developed which can be used in soils over several meters
(see Fig. 4 on the right, cf. Hübner et al. (2005)).
The TDR trace contains far more information than the travel time of
the reflected signal, since the signal responds also to changes in the water
content along the sensor. In order to use this additional information, a three
step algorithm has been developed to reconstruct the soil moisture profile
from the signal response along the sensor (Schlaeger 2005). This new TDR
measurement system consisting of a recorded TDR trace with spatial analysis
waterproof
plastic sheet
Fig. 2. Positions of piezometer gauges and tensiometers within the dyke body
Fig. 3. Positions and lengths of the flat ribbon cables in the cross section of the
dyke and schematic description of the measuring system
Fig. 4. Left: Basic TDR setup and typical TDR trace (schematic), right: Flat ribbon
cable
has already proven its effectiveness in various applications (Hübner et al. 2005,
Scheuermann and Bieberstein 2006) and is called as Spatial-TDR.
2.3 Experimental Investigations
The aims of the investigations on the full-scale dyke model were to observe
water balance processes and to quantify the influence of the initial water con-
tent distribution inside the dyke on the transient seepage in the case of a flood
event (cf. Scheuermann and Bieberstein 2005). For this purpose, an extreme
precipitation event for Karlsruhe expected to occur once in every 100 years
(148 mm in three phases over 72 h) was simulated using sprinkler irrigation
equipment. Figure 5 shows the resulting hydrographs of the parameters mea-
sured. The grey columns in the graph at the top represent the amount of water
irrigated during the experiment and the line in the same graph shows the dis-
charge from the drain measured in the data collection container. The mean
saturation is given in the graph below. The small black dots are measured with
the Spatial-TDR measuring system and the grey rhombuses are determined
manually using a tube probe (IMKO), which is also based on TDR. The next
five graphs show the measurement of the matric potential at the tensiometers
saturation [%] seepage water [l/min]

10 0
irrigation [mm/h]
8 2
6 4
discharge of
4 6
2 8
0 10
50
Spatial-TDR
40
tube probe (IMKO)
mean
30
20
0,1
2,65 m
0,55 m
-0,2
-0,5
positions of tensiometers below surface

0,1
matric potential Ψm [ kPa] at z =
2,2 m
1,0 m
-0,2
-0,5
0,1
1,7 m
1,5 m
-0,2
-0,5
0,1
1,2 m
2,0 m
-0,2
-0,5
0,1
0,7 m
2,5 m
-0,2
-0,5
50
MQ5
25
0
water level above dyke basis [cm] at
50
MQ4
25
0
50
MQ3
25
0
50
MQ2
25
0
50
MQ1
25
0
00:00 12:00 00:00 12:00 00:00 12:00

29. May 2001 30. May 2001 31. May 2001
Fig. 5. Temporal evolution of irrigation, discharge, mean saturation, matric poten-

tial and water level above dyke base during the irrigation experiment (cf. Fig. 2)
(cf. Fig. 2). The graphs of the tensiometer measurements are arranged starting
with the flattest tensiometer at the top and finishing with the one installed
in the deepest position. The last five graphs give the measurements of the
water level above the dyke base with piezometer gauges. Here the graphs are
arranged starting with the piezometer near the drain at the top and finishing
with the piezometer on the upstream side (cf. Fig. 2). From the graphs of
Fig. 5 the temporal changes of the state variables within the dyke body can
be recognised very easily. The first phase of irrigation started at midnight
on May 29th. The corresponding saturation distribution is given in Fig. 6a).
The dotted lines represent the positions of the individual flat ribbon cables
together with the measured saturations at these locations (cf. colour bar in
Fig. 6). For a better visualisation, the single measurements were interpolated
over the observed area in the cross section. The dark colours show the wet
zones, whereas the light grey colours represent the more dry zones.
As expected, the mean saturation measured with Spatial-TDR shows the
first changes only a few hours after the beginning of the first irrigation phase
(cf. Fig. 5). The top tensiometer at a depth of 0.55 m registered the first
increase in the matric suction indicating the arrival of the water front approx-
imately 7 h after the start of the experiment. The closed water front, which
moves downwards, can be observed very well from the temporal evolution of
the matric potential measurements. However the first increase in the water
level above the dyke base can be recognised at almost every piezometer gauge
after approximately 9 h (see graphs for MQ1 to MQ5 in Fig. 5). At this point
in time the water content above the dyke base in the middle of the cross
section shows almost no change (cf. Fig. 6b)).
One possible explanation for the observation of a rise in the water table
near the dyke base without any obvious water content changes directly above
the base is preferential water flow in the form of fingering. This phenomenon
occurs for example when the moving wetting front reaches an unstable con-
dition, which means that the water front begins to develop concentrated flow
paths, the so-called fingers.
After the last irrigation phase (9:00 on May 31st) the saturation distribu-
tion of Fig. 6c) was measured. It can be seen that despite the high quantity
of irrigated water, no homogeneously distributed saturation was reached. In
fact the larger part of the water was stored near the surface in the slopes of
the dyke (cf. Fig. 6a) and c)). In these more saturated areas the sand has
a higher hydraulic conductivity, so that water can flow laterally to the wa-
terproof sealing at the base of the dyke due to capillary barrier effects. As
a consequence, an area in the middle of the cross section remained nearly
unchanged (cf. Fig. 6d)).
4
28.05.2001, 23:44
3 18 18 a)
19 23 24 16
2 22 23 23 23 23 22
18 22 23 24 23 23 23 23
1 22 23 24 22 21 20 20 22 24 22
22 24 23 22 22 20 20 19 21 22 24 23
0 44
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
4
29.05.2001, 09:15
3 28 34 b)
45 25 25 34
2 45 26 25 24 24 42
40 25 22 24 22 22 25 33
1 39 27 23 22 20 19 18 21 23 35
29 20 22 19 20 21 19 17 20 19 22 23
0 47
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
4
31.05.2001, 08:43
3 39 39 c)
43 33 38 38
2 44 38 30 30 34 44
38 35 28 28 26 27 34 44
1 35 35 29 26 23 23 23 25 31 38
34 34 29 27 27 25 22 21 25 24 34 36
0 66
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
difference in saturation
between a) and c) > 5 % d)
Fig. 6. Distribution of saturation during an irrigation test (148 mm in 72 h) on

the full-scale dyke model: a) at the beginning of the irrigation test, b) after the
first irrigation phase 9 h after beginning of the experiment, c) at the end of the
experiment and d) shows the difference in the saturation between a) and c) during
the test (area with changes > 5% highlighted in colour)
3 Laboratory Investigations
In order to verify the observations on the dyke model, additional labora-
tory experiments were carried out. The infiltration chamber used is 105 cm
in length, 62.5 cm in height and 3 cm in depth. At the base of the chamber,
water outlets are located in order to drain water out of the chamber. For the
experiments, soil formations of the same material as in the dyke model were
placed in the infiltration chamber. The initial condition of the soil within the
chamber was always totally dry. At the top of the soil formation, an infiltra-
tion system was set up to use drop irrigation of a rate much smaller than
the saturated hydraulic conductivity of the material. The infiltration process
was recorded with a digital camera. In the following two sections, experiments
concerning the phenomena observed are presented.
3.1 Fingering Experiment
For the fingering experiment, a soil formation was constructed in a cham-

ber with a thin horizontal top soil layer. At 11 mm/h the mean infiltration
rate during the experiment was approximately 65 times lower than the sat-
urated hydraulic conductivity. Figure 7 shows infiltration fronts at specific
mean infiltration rate iB = 11 mm/h

kf / i B = 65
30
60
90
120
150
62,5 cm
180
210
240
time steps
in [min]
270
kf = 2 10 -4 m/s 300
105 cm 3,0 cm
Fig. 7. Experiment on an infiltration chamber in order to investigate the fingering

effect: Water fronts at different time intervals
time intervals. The grey scale colouring shows the temporal evolution of the
infiltration. From the infiltration fronts, it can be seen very clearly that al-
ready after 60 minutes the first finger began to develop on the right side of
the chamber. Sixty minutes later the second finger grew out of the infiltration
front on the left side. As soon as a finger inclines to develop the water front
around the finger starts to decelerate. Similar experiments have shown that
almost saturated conditions prevail at the finger tip, whereas at the tail of
the finger saturation decreases. As soon as the velocity of the water in the
finger decreases, the water content increases and a growth of the finger as
well as an accelerated movement of the water front starts in the surrounding
material (Kawamoto and Miyazaki 1999). Both could also be observed in the
laboratory experiment presented here.
3.2 Experiment on Lateral Water Movement
The soil formation for the investigation on the phenomenon of lateral wa-
ter movements in inclined slopes was built with a similar slope inclination
as the slopes of the full-scale dyke model. As in the experiment described
above, the irrigation was set at 6.8 mm/h much lower than the saturated
hydraulic conductivity (approx. 100 times). From the temporal evolution of
the infiltration fronts, it can be seen very clearly that during the course of
infiltration, the water movement was relocated more and more towards the
slope, where the water flowed laterally in the direction of the drain. After
approximately 200 minutes, water started to flow primarily from the slope to
the base. Furthermore, a finger developed underneath the horizontal part of
the soil formation on the right side of the infiltration chamber. The disturbed
water fronts developed due to small heterogeneities which result from the way
the soil formation was constructed.
4 Analysis of Experimental Investigations

As already pointed out in the introduction, both the development of fingering
as well as the lateral movement of water due to the capillary barrier effect
are based on the same soil hydraulic property, namely the breakthrough head
which roughly corresponds to water entry value ψWEV of the soil. In this
connection, the water entry value is equal to the matric potential when the soil
becomes wet for the first time. However, physical investigations have proven
that in reality the water entry value depends also on the initial condition of
the soil (Wang et al. 2000). Also the daily observations have shown that soil
becomes wetted much more easily, if it has already been wetted with water.
A method, which uses the inflection point of the imbibition soil water re-
tention curve to determine an effective value for the water entry value, has
been given by Baker and Hillel (1990), cf. Fig. 9. With the corresponding
unsaturated hydraulic conductivity ku (ψWEV ) = 2.8 · 10−5 m/s ≈ 100 mm/h
top ground
time steps in [min]
surface
mean infiltration rate iB = 6,8 mm/h
kf = 2 10 -4 m/s
kf / i B = 106
drain
65 cm
40
80
120
160
200
400
240 280 320 360
105 cm 3,0 cm
Fig. 8. Experiment on an infiltration chamber in order to investigate lateral water

movements due to a capillary barrier effect: Water fronts at different time intervals
Fig. 9. Estimation of the effective water entry value ψWEV according to Baker
and Hillel (1990) from the imbibition soil water retention curve of the construction
material of the full-scale dyke model with experimental results and a van Genuchten
parameterisation
(estimated from the soil water retention curve given in Fig. 9 and the satu-
rated hydraulic conductivity kf = 2 · 10−4 m/s with the model by Mualem
(1976)) it is possible to estimate the possibility of the occurrence of preferen-
tial flow. Using the criterion according to Hillel and Baker (1988), as already
mentioned in the introduction, and comparing ku (ψWEV ) with the infiltration
rates of the experiments (5–6 mm/h on the full-scale dyke model, 11 mm/h for
the fingering experiment and 6.8 mm/h for the experiment on lateral water
movement) it is completely possible that both phenomena could occur during
the precipitation experiment on the full-scale dyke model.
5 Summary and Conclusion
A precipitation experiment on a homogeneously built full-scale dyke model at

the Federal Waterways and Research Institute in Karlsruhe has shown anoma-
lies, which indicate that preferential water movements in the form of fingering
and lateral water movement have possibly occurred. In this connection, a new
measuring system for the observation of spatially distributed water contents
has provided important information on the hydraulic situation within the dyke
body.
In order to verify the observations on the dyke model and to prove the
hypothesis of preferential flow, additional laboratory investigations were con-
ducted on an infiltration chamber. Both experiments showed clearly that un-
saturated infiltration processes in inclinational soil formations can lead to
preferential water movements in the form of fingering and lateral water move-
ment. The appearance of these kinds of preferential flow can be expected,
especially if the infiltration rate is lower than the unsaturated hydraulic con-
ductivity at the water entry value.
Whereas the occurrence of lateral water movement could be proven by the
moisture measurements using Spatial-TDR, some questions still remains re-
garding preferential flow due to fingering. The observations of the water level
above the dyke base during the experiment indicate the occurrence of finger-
ing. However, the investigations presented could not provide an undisputable
proof of fingering.
Against the background of slope failures induced by heavy or extreme pre-
cipitation events, the questions concerning water infiltration and water move-
ment processes in natural (hill) or artificial (deposit) slopes play a significant
role.
References
Baker RS, Hillel D (1990) Laboratory tests of a theory of fingering during infiltration
into layered soils, Soil Sci Soc Am J 54:20–30
Hill DE, Parlange J-Y (1972) Wetting front instability in layered soils, Soil Sci Soc
Am Proc, 36(5):697–702
Hillel D, Baker RS (1988) A descriptive theory of fingering during infiltration into
layered soils, Soil Science 146(1):51–56
Hübner C, Schlaeger S, Becker R, Scheuermann A, Brandelik A, Schaedel W, Schuh-
mann R (2005) Advanced measurement methods in Time Domain Reflectometry
for soil moisture determination. In: Kupfer K (ed) Electromagnetic Aquametry,
Springer, pp 317–347
Kawamoto K, Miyazaki T (1999) Fingering flow in homogeneous sandy soils under
continuous rainfall infiltration, Soils and Foundation 39(4):79–91
McCord JT, Stephens DB, Wilson JL (1991) Hysteresis and state-dependent
anisotropy in modelling unsaturated hillslope hydrologic processes, Water Re-
sour Res 27(7):1501–1518
Mualem Y (1976) Hysteretical models for prediction of the hydraulic conductivity
of unsaturated porous media, Water Resour Res 12(6):1248–1254
Scheuermann A, Bieberstein A (2005) Influence of the initial water content on the
transient seepage through dikes. In: International Symposium in Dam Safety
and Detection of Hidden Troubles of Dams and Dikes, Xi’an, China, November
1–3
Scheuermann A, Bieberstein A (2006) Monitoring of dams and dikes – water con-
tent determination using Time Domain Reflectometry (TDR). In: 13th Danube-
European Conference on Geotechnical Engineering, Ljubljana, Slovenia, May
29–31, ISBN 961.90043–8–8, pp 493–498
Schlaeger S (2005) A fast TDR-inversion technique for the reconstruction of spatial
soil moisture content, Hydrol Earth Sys Sci 2:971–1009
Stephens DB, Heermann S (1988) Dependence of anisotropy of saturation in a strat-
ified sand, Water Resour Res 24(5):770–778
Stormont JC, Clifford EA (1999) Capillary barrier effect of underlying coarser soil
layer, J Geotechnical and Geoenvironmental Engineering 125(8):641–648
Topp GC, Davis JL, Annan AP (1980) Electromagnetic determination of soil wa-
ter content: Measurements in coaxial transmission lines, Water Resour Res
16(3):574–582
Walter MT, Kim J-S, Steenhuis TS, Parlange J-Y, Heilig A, Braddock RD, Selker
JS, Boll J (2000) Funneled flow mechanisms in a sloping layered soil: Laboratory
investigation, Water Resour Res 36(4):841–849
Wang Z, Wu L, Wu QJ (2000) Water-entry value as an alternative indicator of soil
water-repellency and wettability, J Hydrology 231–232:76–83
Compaction Properties of Agricultural Soils
Anh-Minh Tang1 , Yu-Jun Cui1 , Javad Eslami1 , and Pauline

Défossez-Berthoud2
1
ENPC, 6–8 av. Blaise Pascal, F–77455 Marne–la–Vallée Cedex 2, France
tang@cermes.enpc.fr, cui@cermes.enpc.fr, eslami_javad@yahoo.com
2
INRA, rue Fernand Christ, F–02007 Laon cedex, France defossez@laon.inra.fr
Summary. The compaction of field soils due to repeated rolling of agricultural

vehicles is one of the main reasons for the agricultural soil degradation. A good
understanding of the compaction properties of these soils is essential for an optimum
organisation of agricultural activities, and therefore for environmental protection
in terms of nitrate migrations. In the present work, the compaction properties of
agricultural soils from four sites in France are studied after experimental data from
oedometer tests. In the oedometer tests, a quick loading procedure was applied to
simulate the loading of tire rolling. The soils that were initially in unsaturated state
were loaded under constant water content condition. The compaction properties of
these soils (i.e. the precompression vertical stress, compression index and swelling
index) were then determined. The effect of initial dry density and initial water
content on these properties is discussed. A possible effect of loading velocity on
the apparent compressibility was observed. The results are finally discussed in the
context of unsaturated soil mechanics.
Key words: compaction, agricultural soil, oedometer, quick loading, compressibil-

ity, dry density, water content
Introduction
Soil compaction induced by vehicle traffic is one of the major problems in mod-
ern agriculture. It is well-known that soil compaction increases soil strength
and decreases soil hydraulic conductivity, as a result, root penetration is re-
duced; water extraction becomes more difficult; plants growth is therefore
affected. From an economical point of view, this would result in the increase
of production cost (Hamza and Anderson 2005, Raper 2005, Chan et al. 2006).
Oedometer test is usually used to study the compaction properties of arable
soils (Arvidsson and Keller 2004). The main parameters determined from this
test are: (i) precompression stress, σp ; (ii) the slope of the normal consoli-
dation curve, λ; (iii) the slope of the unloading curve, κ. These parameters
are useful in the modelling of agricultural soils compaction induced by vehicle
476 Anh-Minh Tang et al.
traffic (Berli et al. 2003). As the conventional oedometer test is time consum-
ing, it is common practice to use pedotransfer functions (Imhoff et al. 2004) to
estimate the soil mechanical properties. Horn et al. (2005) used this method to
predict the mechanical strength of arable soils in Eastern and Western Europe
countries at various scales.
In the present work, oedometer tests are performed to study the com-
paction properties of four soils from France. The effects of dry density and
water content on the compressibility of soil are discussed.

Soil samples were taken from four sites in France: (1) Mons, La Somme; (2)
Epernay, Marne; (3) le Breuil, Nièvre; (4) Avignon, Vaucluse. The soils sam-
ples were taken from two different horizons: from the cultured horizon at
0–30 cm and the undisturbed horizon at 30–60 cm depth. The physical prop-
erties determined according to the French Standards (AFNOR) are presented
in Table 1. The specific gravity was determined using water pycnometer on soil
sieved at 2 mm; Atterberg limits were determined with soil sieved at 0.4 mm;
and the methylene blue absorption was measured with soil sieved at 0.5 mm.
The classification is based on the Atterberg limits.
Prior to oedometer test, the bloc of undisturbed soil from 30–60 cm depth
was wetted (by spraying) or dried (in air) to have the desired water content.
When the target water content value was reached, the soil bloc was put in a
hermetic box during 24 h for homogenisation of water distribution. Finally, the
soil sample (70 mm in diameter, 20 mm in height) was trimmed directly from
the soil bloc and inserted in the oedometer cell. The soils from 0–30 cm depth
were air dried and sieved at 2 mm. Prior to oedometer test, the soils were
wetted (by spraying) until the desired water content was reached. They were
Table 1. Physical properties of soils studied
Soil Mons Breuil Epernay Avignon

Specific gravity, G 2.62 2.56 2.68 2.71
Liquid limit, wL (%) 32 58 49 31
Plastic limit, wP (%) 22 51 29 20
Plasticity index, IP (%) 10 7 20 11
Methylene blue absorption 1.4 0.4 7.4 2.3
(g/100 g)
Grain size distribution (%):
– Clay (< 2 μm) 19 19 47 34
– Silt (2–50 μm) 75 23 33 51
– Sand (> 50 μm) 6 58 20 16
Classification Low plas- High plas- Low plas- Low plas-
ticity clay ticity silt ticity silt ticity clay
Compaction Properties of Agricultural Soils 477
then stocked in a hermetic box during 24 h for obtaining homogeneous water

distribution. Afterwards, the soils was compacted directly in the oedometer
cell to the desired dry density.
Vertical normal stresses of 15, 30, 50, 100, 200, 300, 600, and 800 kPa were
applied sequentially during loading stage. During unloading stage, the vertical
normal stress was decreased from 800 kPa to 600, 300, 200, 100, 50, 30, 15 kPa.
Each stress was applied for 5 min and the displacement (accuracy ±0.001 mm)
was read at the end of each step. At the end of test, the soil sample was taken
out of the oedometer cell and its dimensions were measured using a calliper
(accuracy ± 0.001 mm). Finally, its water content was determined by oven-
drying at 105◦ C during 24 h. These measurements allowed determining the
final void ratio and the degree of saturation. The initial void ratio is calculated
by back analysis using final void ratio and total displacement measured.
Experimental Results
The test program and the main results are presented in Table 2. In Figure 1,
the results of some test (void ratio and degree of saturation as a function of
vertical stress) are shown. It can be observed that the relationship e − log σv
in the unloading path is linear for all the tests. The swelling index is then
calculated from the unloading path as follows: κ = Δe/Δ ln σv (slopes in Fig. 1
divided by ln 10 = 2.3). In the tests where a clear elasto-plastic behaviour is
observed as test 12 (Fig. 1a) and test 39 (Fig. 1d), the compression index,
λ = Δe/Δ ln σv , is calculated from the three last points in the compression
curve. In test 16 (Fig. 1b) and test 31 (Fig. 1c) where a change of the slope can
be observed during the compression curve, the maximum value of the slope is
taken to calculate the compression index. The precompression stress (σp ) is
calculated as the interception of the compression line and the line that across
the initial point and that is parallel to the unloading line. All the parameters
obtained (σp , λ and κ) are shown in Table 2 with the void ratio (initial, ei ,
and final, ef ), the water content (initial, wi , and final, wf ) and the initial dry
density (ρi ).
In Figure 2, the precompression stress (σp ), the compression index (λ)
and the swelling index (κ) of all soils are drawn as functions of initial water
content (wi ) and mean initial dry density (ρi ). For all tests, a decrease of the
precompression stress can be observed when the initial water content increases
or when the initial dry density decreases. In addition, the swelling index (κ)
seems to be insensible to the initial dry density and the swelling index. In the
case of soils from Breuil (Fig. 2a) and Mons (Fig. 2d), it is observed that at
the same water content, looser soil samples (lower dry density) have higher
compression index, and that at the same dry density the compression index
increases with the water content increase. For the soil from Epernay, Fig. 2b,
the increase of water content reduces the compression index. In the case of soil
from Avignon (Fig. 2c), wetting induced an increase following by a decrease
Table 2. Test program and results (Br.: Breuil; Ep.: Epernay; Av.: Avignon; Mo:
Mons)
No Soil Depth ei ef wi wf ρi σp λ κ
(cm) (%) (%) (Mg/m3 ) (kPa)
1 Br. 0–30 1.49 0.85 24.7 24.7 1.03 34 0.225 0.012
2 Br. 0–30 1.43 0.81 18.9 18.2 1.05 45 0.235 0.010
3 Br. 0–30 1.49 0.81 25.3 22.7 1.03 24 0.224 0.012
4 Br. 0–30 1.08 0.73 23.3 22.5 1.23 65 0.153 0.009
5 Br. 0–30 1.16 0.79 24.7 24.7 1.19 67 0.163 0.012
6 Br. 0–30 1.09 0.80 17.9 18.8 1.22 106 0.133 0.009
7 Br. 0–30 0.93 0.76 20.2 19.1 1.33 165 0.111 0.008
8 Br. 0–30 0.99 0.77 24.2 24.2 1.29 129 0.135 0.011
9 Br. 0–30 1.01 0.74 25.8 25.2 1.27 114 0.148 0.010
10 Br. 30–60 1.41 0.83 18.1 17.4 1.06 50 0.259 0.007
11 Br. 30–60 1.24 0.72 24.7 23.8 1.14 35 0.181 0.010
12 Br. 30–60 1.30 0.84 16.3 13.5 1.11 95 0.222 0.006
13 Ep. 0–30 1.44 0.87 37.8 30.4 1.10 26 0.230 0.009
14 Ep. 0–30 1.64 0.84 32.5 28.6 1.02 26 0.358 0.016
15 Ep. 0–30 1.74 0.82 25.6 25.2 0.98 39 0.377 0.014
16 Ep. 0–30 1.27 0.88 37.1 29.7 1.18 29 0.146 0.015
17 Ep. 0–30 1.36 0.86 31.3 28.7 1.14 42 0.241 0.014
18 Ep. 0–30 1.41 0.78 25.6 25.0 1.11 73 0.304 0.010
19 Ep. 0–30 1.17 0.90 37.9 31.5 1.24 32 0.083 0.011
20 Ep. 0–30 1.10 0.82 30.4 28.4 1.28 56 0.133 0.012
21 Ep. 0–30 1.11 0.79 25.1 24.2 1.27 76 0.183 0.012
22 Ep. 30–60 1.16 0.90 32.5 30.6 1.24 60 0.108 0.016
23 Ep. 30–60 1.49 1.15 41.4 39.6 1.08 50 0.151 0.014
24 Ep. 30–60 1.08 0.90 30.2 29.6 1.29 83 0.093 0.013
25 Av. 0–30 1.36 0.66 16.8 16.4 1.15 37 0.247 0.011
26 Av. 0–30 1.24 0.59 22.4 19.5 1.21 19 0.258 0.014
27 Av. 0–30 1.04 0.64 28.7 20.4 1.33 15 0.103 0.015
28 Av. 0–30 1.13 0.63 16.5 16.2 1.27 77 0.228 0.010
29 Av. 0–30 1.10 0.58 22.1 19.5 1.29 26 0.229 0.014
30 Av. 0–30 0.91 0.60 28.5 20.2 1.42 20 0.096 0.014
31 Av. 0–30 1.06 0.64 16.8 16.4 1.32 98 0.212 0.009
32 Av. 0–30 0.98 0.59 21.9 19.6 1.37 28 0.154 0.013
33 Av. 0–30 0.88 0.63 28.2 19.6 1.44 23 0.068 0.013
34 Av. 30–60 0.76 0.60 19.2 18.4 1.54 99 0.089 0.012
35 Av. 30–60 0.80 0.60 23.0 19.2 1.51 77 0.086 0.012
36 Av. 30–60 0.79 0.61 21.0 19.8 1.51 45 0.083 0.014
37 Mo. 30–60 0.59 0.51 12.4 12.4 1.65 106 0.073 0.009
38 Mo. 30–60 0.79 0.63 19.5 19.5 1.46 112 0.109 0.008
39 Mo. 30–60 0.75 0.62 18.2 18.2 1.50 137 0.111 0.008
40 Mo. 30–60 0.78 0.60 29.7 20.5 1.47 68 0.105 0.012
41 Mo. 30–60 0.75 0.57 27.1 19.9 1.50 83 0.106 0.012
1.3 1.3
1.2 (a) 1.2
(b)
1.1 1.1
e (-)
e (-)
1 1
0.9 0.9
0.8 0.8
1 1
0.8 Test 12 0.8
Sr (-)
Sr (-)
0.6 0.6
Test 16
0.4 0.4
0.2 0.2
0 0
10 100 1000 10 100 1000

Vv (kPa) Vv (kPa)
(d)
1.1 0.76
1
0.9
(c) 0.72
0.68
e (-)
e (-)
0.8 0.64
0.7 0.6
0.6 0.56
1 1
0.8 0.8
Sr (-)
Sr (-)
0.6 0.6
0.4 0.4
Test 39
0.2 Test 31 0.2
0 0
10 100 1000 10 100 1000

Vv (kPa) Vv (kPa)
Fig. 1. Void ratio (e) and degree of saturation (Sr ) as a function of vertical stress
(σp ): (a) Breuil; (b) Epernay; (c) Avignon; (d) Mons
of compression index in case of loose soils (ρi = 1.25 − 1.35 Mg/m3 ). On the
contrary, wetting induced only a decrease of compression index in the case of
dense soils (ρi = 1.40 Mg/m3 ).
200 0.3 Vp, O (Ui= 1.05 Mg/m3)

Vp, O (Ui= 1.20 Mg/m3)
160
Vp (kPa)
O and N
0.2 Vp, O (Ui= 1.30 Mg/m3)
(a) 120
80
0.1
Vp, O (undisturbed)
N (Ui= 1.05 Mg/m3)
40
N (Ui= 1.20 Mg/m3)
0 0
N (Ui= 1.30 Mg/m3)
16 18 20 22 24 26 16 18 20 22 24 26 N (undisturbed)
w (%) w (%) Vp, O (Ui= 1.05 Mg/m3)
100 0.4
Vp, O (Ui= 1.15 Mg/m3)
80 0.3
Vp (kPa)
O and N
Vp, O (Ui= 1.25 Mg/m3)
(b) 60 0.2 Vp, O (undisturbed)

N (Ui= 1.05 Mg/m3)
40 0.1
N (Ui= 1.15 Mg/m3)
20 0 N (Ui= 1.25 Mg/m3)
24 28 32 36 40 44 24 28 32 36 40 44 N (undisturbed)
w (%) w (%)
100 0.3
Vp, O (Ui= 1.25 Mg/m3)
80 0.25
Vp, O (Ui= 1.35 Mg/m3)
Vp (kPa)
O and N
0.2
(c) 60
40
0.15
Vp, O (Ui= 1.40 Mg/m3)
Vp, O (undisturbed)
0.1
20 0.05 N (Ui= 1.25 Mg/m3)
0 0 N (Ui= 1.35 Mg/m3)
N (Ui= 1.40 Mg/m3)
16 20 24 28 32 16 20 24 28 32
N (undisturbed)
w (%) w (%)
140 0.12
120
Vp (kPa)
O and N
0.08
(d) 100
0.04
Vp, O (undisturbed)
N (undisturbed)
80
60 0
12 16 20 24 28 32 12 16 20 24 28 32
w (%) w (%)
Fig. 2. Precompression stress (σp ), compression index (λ) and swelling index (κ) as
a function of mean initial water content (wi ) and initial dry density (ρi ): (a) Breuil;
(b) Epernay; (c) Avignon; (d) Mons
Discussions
In the domain of unsaturated soil mechanics, it is well-known that the precom-
pression stress of looser soil is lower that that of denser soil. In addition, an
increase of water content (that corresponds to a decrease of suction) reduces
the soil strength or precompression stress (Alonso et al. 1990).
On the other hand, as the swelling index of soil depends on the stiffness
of soil grains, it is independent on the soil density and on the water content
in case of low plasticity soils. All these phenomena have been observed on the
agricultural soils.
On the contrary, the effect of water content on the compression index ob-
served in the present work is different from that found in the literature. After
Alonso et al. (1990), wetting softens the soil and increases then the compres-
sion index. Cui and Delage (1996) observed the same phenomenon on the
compacted Jossigny silt. Nevertheless, this phenomenon can be observed only
on the soils from Breuil and Mons. This contradiction can be explained by the
consolidation mechanism in these tests. Indeed, in geotechnical engineering, a
loading duration longer than 24 h is applied to simulate the stress generated
by buildings construction. But in the present work, 5 min was applied for each
loading stage in order to simulate the stress generated by rolling of agricul-
tural vehicles. In case of plastic soils, as its permeability is low, this duration
of 5 min may be not sufficient for water movement within the soil at high
saturation degree. In Figure 1b, for example, loading increased the degree of
saturation and the soil reached saturation state under 200 kPa vertical stress.
The change of the compression curve slope observed in this test can be then
explained by a partial consolidation of the soil.
In conclusion, wetting softens the soil and increases then the compression
index. But this compression index can be reduced also by wetting due to
the partial consolidation during short loading duration. This phenomenon
depends on the permeability (that is influenced by soil dry density and soil
plasticity) and the degree of saturation. The combination of these two trends
governs the effect of water content on the compression index.
Conclusions
Oedometer tests were performed on agricultural soils taken from four sites
in France. Rapid loading stage (5 min for each stage) was applied to study
the soil compaction due to agricultural vehicles traffic. The effect of dry den-
sity and water content on the compaction properties of soils was observed.
Generally, the soil is more compressible at lower dry density or higher water
content. Nevertheless, in case of plastic soils or dense soils, wetting reduces
the compression index. The consolidation mechanism was discussed to reveal
the effect of loading duration on the compression index.
Acknowledgements
The authors are grateful for the technical assistance of D. Boitez and F. Bor-
net. This work is part of the French National Projects ANR–ADD and Gessol2.
The financial support from the two projects is gratefully acknowledged.
References
Alonso EE, Gens A, Josa AA (1990) A constitutive model for partially saturated
soil, Géotechnique 40(3):405–430
Arvidsson J, Keller T (2004) Soil precompression stress. I. A survey of Swedish

arable soils, Soil & Tillage Research 77:85–95
Berli M, Kirby JM, Springman SM, Schulin R (2003) Modelling compaction of agri-
cultural subsoils by tracked heavy construction machinery under various mois-
ture conditions in Switzerland, Soil & Tillage Research 73:57–66
Chan KY, Oates A, Swan AD, Hayes RC, Dear BS, Peoples MB (2006) Agronomic
consequences of tractor wheel compaction on a clay soil, Soil & Tillage Research
89:13–21
Cui YJ, Delage P (1996) Yielding and plastic behaviour of an unsaturated compacted
silt, Géotechnique 46(2):291–311
Hamza MA, Anderson WK (2005) Soil compaction in cropping systems. A review
of the nature, causes and possible solutions, Soil & Tillage Research, 82:121–145
Horn R, Fleige H, Richter FH, Czyz EA, Dexter A, Diaz-Pereira E, Dumitru E,
Enarche R, Mayol F, Rajkai K, de la Rosa D, Simota C (2005) SIDASS project.
Part 5: Prediction of mechanical strength of arable soils and its effects on physical
properties at various map scales, Soil & Tillage Research 82:47–56
Imhoff S, Da Silva AP, Fallow D (2004) Susceptibility to compaction, load support
capacity, and soil compressibility of Hapludox. Soil Sci Soc Am J 68:17–24
Raper RL (2005) Agricultural traffic impacts on soil, J Terramechanics 42:259–280
Bearing Capacity of Model Footings
in Unsaturated Soils
Sai K. Vanapalli and Fathi M.O. Mohamed
Department of Civil Engineering, 161 Louis Pasteur Str., University of Ottawa,

Ottawa, K1N 6N5 Ontario, Canada
vanapall@eng.uottawa.ca, fmoha035@uottawa.ca
Summary. A simple technique is proposed to predict the variation of the bearing

capacity of an unsaturated soil with respect to matric suction. This technique is
based on extending conventional bearing capacity theory proposed by Terzaghi. The
proposed equation in this paper is presented as a functional relationship such that
the variation of the bearing capacity of an unsaturated soil with respect to matric
suction can be predicted. This technique is developed extending the concepts for
predicting the shear strength of unsaturated soils proposed by Vanapalli et al. (1996).
Using the approach presented in this paper, the bearing capacity of an unsaturated
soil can be predicted using the saturated shear strength parameters, c and φ and
the soil-water retention curve (SWRC).
Key words: bearing capacity, unsaturated soils, shear strength, matric suction,
soil–water retention curve, model footings
Introduction
The bearing capacity is one of the key parameters required in the design
of shallow foundations. Several approaches are available in the literature for
determination of the bearing capacity of soils based on the saturated shear
strength parameters (Terzaghi 1943, Meyerhof 1951). However, in several situ-
ations, shallow foundations are located above the ground water table where the
soil is typically in a state of unsaturated condition. Nevertheless, the bearing
capacity of soils is commonly determined assuming fully saturated conditions
ignoring the influence of capillary stresses or the matric suction. Due to this
reason, estimation of the bearing capacity of shallow foundations using the
conventional approaches may not be reliable leading to uneconomical designs.
Foundation designs for unsaturated soils are complex and require not only
the soil–structure interaction but also a fundamental understanding of soil be-
haviour that comprises the combined role of suction and cementation (Schnaid
et al. 1995). Several researchers carried out investigations on the bearing ca-
pacity of unsaturated soils (Broms 1963, Steensen-Bach et al. 1987, Oloo 1994,
484 S. K. Vanapalli and F. M. O. Mohamed
Miller and Muraleetharan 1998, Costa et al. 2003, Mohamed and Vanapalli
2006). All these studies have shown significant contribution of matric suction
to the bearing capacity of unsaturated soils. However, limited theoretical re-
search work is reported in the literature with respect to the interpretation of
the bearing capacity of unsaturated soils (Fredlund and Rahardjo 1993, Oloo
et al. 1997).
Based on the results presented in this study, a semi-empirical equation
is proposed to predict the variation of the bearing capacity of unsaturated
soils using the saturated shear strength parameters, c and φ and the SWRC.
The equation presented in this paper is developed extending the concepts for
predicting the shear strength of unsaturated soils proposed by Vanapalli et
al. (1996). The proposed equation is also extended for other model footings
studies reported in the literature that includes both coarse-grained and fine-
grained soils. The studies presented in this paper show that there is a good
comparison between the measured and predicted bearing capacity of model
footings.
Test Equipment
Figure 1 shows the details of equipment specially designed at the University
of Ottawa for determining the bearing capacity of coarse-grained soils using
model footings. All the key features of this equipment are summarized in
Mohamed and Vanapalli (2006). This equipment has special provisions to
achieve fully saturated and unsaturated conditions of the compacted sand in
the test tank. While the water table level in the test tank can be adjusted
to the desired level using drainage valves, the capillary tension (i.e., matric
suction) variation with respect to depth in the unsaturated soil zone below
the model footing can be measured using commercial Tensiometers.
Properties of the Tested Soil

The properties of the soil used in this study were determined in the geotech-
nical laboratory of the University of Ottawa. The soil was classified using
USCS as poorly graded sand (SP). The average void ratio after compaction
was 0.63 and the dry unit weight was 16.02 kN/m3 . The internal friction angle
was 35.3◦ from direct shear tests. Bolton (1986) studies on sands show that
the effective stress and soil density affect the rate of dilatancy of soils and
thereby influence the internal friction angle, φ . As the dilatancy of sands has
a significant effect on the φ value, it was suggested that the measured φ val-
ues can be modified as φ = (φ + 0.8 of dilatancy angle). Some investigators
have found that better comparisons can be provided between the measured
and computed bearing capacity values by using φ values about 10 to 15%
higher than the measured values (Steensen-Bach et al. 1987). Therefore, φ
Bearing Capacity of Model Footings in Unsaturated Soils 485
Fig. 1. University of Ottawa Bearing Capacity Equipment (UOBCE)
value which is 10% higher than the measured value (i.e., 39◦ ) was used in this
study.
Typical Experimental Results
The relationship between the applied stress and settlement of typical experi-
mental results on a 100 mm × 100 mm square model footing is shown in Fig. 2.
This relationship demonstrates that there is a significant increase in the bear-
ing capacity of the model footing due to the contribution of matric suction
in the range 0 to 6 kPa for the tested compacted, coarse-grained soil. The
analysis presented in this paper is based on the average matric suction value
in the proximity of the stress bulb. The procedure used for the determination
of the average suction value is detailed using Fig. 3.
The variation of the matric suction with respect to depth underneath the
model footing is non-linear as shown in Fig. 3 (left hand side). The variation
of matric suction in the capillary zone above the ground water table (GWT) is
typically hydrostatic (see right hand side of Fig. 3) for coarse-grained soils. The
experimental results supported this generally observed behavior and indirectly
900
6 kPa
800 e=0.63 4 kPa
S=58% e=0.64
700 S=78%
Applied stress, (kPa)
600
2 kPa
500 e=0.62
S=90%
400
300 Suction=0 kPa

e=0.63
200 S=100%
100
0
0 4 8 12 16 20 24 28
Settlement, (mm)
Fig. 2. The relationship between the applied stress versus settlement for 100 mm ×
100 mm square model footing
demonstrated that the Tensiometers performed well. As the significant soil

stresses are typically distributed over a depth of 1.5B (Poulos and Davis 1974
and Chen 1999) the matric suction value is considered as the average value of
(ua − uw )1 (matric suction close to the surface of the footing) and (ua − uw )2
(matric suction value at the bottom of the stress bulb) as shown in Fig. 3.
The Measured and the Predicted SWRC
Figure 4 shows the soil–water retention curve (drying curve) plotted as a rela-
tionship between the degree of saturation, S and the matric suction, (ua −uw )
using three different methods. Two direct methods were used for measuring
the SWRC’s. The first method constituted the measurement of the SWRC di-
rectly from the test tank. The SWRC was also measured using the Tempe cell
in the laboratory, which formed the second method. More details of the pro-
cedures used in the determination the SWRC are available in Mohamed and
Vanapalli (2006). The third method for estimation of the SWRC was based
on the procedure summarized in Vanapalli and Catana (2005). This proce-
dure uses one measured point (i.e., water content and matric suction) along
with data obtained from the grain size distribution curve. Figure 4 shows that
Tensiometer
B (ua-uw )1
(ua-uw )1 GSL
7 kPa Stress
bulb 1.5 B
4 kPa
A ct u (ua-uw )
2
e
ro f i l
al p
2 kPa
p
ro
at i c
f il e
r ost
H yd
GWT - Test 1
Fig. 3. Schematic to demonstrate the procedure used for determining the average
matric suction below the footing
Measured from Tempe cell

Measured from the test tank
Predicted using Vanapalli and Catan 2005
Degree of saturation, S (%)
100
75
50
Best Fit
25
0
0 2 4 6 8 10
Matric suction, (u a - uw ) kPa
Fig. 4. Measured SWRC from the Tempe cell and test tank along with the estimated
SWRC
there is a good agreement between the SWRC’s using all the three methods.
The objective of the determination of the SWRC was to understand its re-
lationship with the bearing capacity of unsaturated soils similar to the shear
strength of unsaturated soils.
Bearing Capacity of Unsaturated Soils
Terzaghi (1943) proposed an equation for computing the bearing capacity of

shallow strip footings extending Prandtl (1921) assumptions for the soil failure
mechanism. This equation is valid for strip footings resting in a homogenous
soil and subjected to vertical loading.
qu = c Nc + qNq + 0.5BγNγ (1)
where:
qu = ultimate bearing capacity, kPa
q = overburden pressure, kPa
c = effective cohesion, kPa
Nc , Nq , Nγ = bearing capacity factors due to cohesion, surcharge and unit
weight, respectively
γ = soil unit weight, kN/m3
B = footing width, m
Equation (1) can be written as given below for interpreting the bearing ca-
pacity of surface footings taking account of the influence of the shear strength
contribution due to matric suction for unsaturated soils:
qu = [c + (ua − uw ) tan φb ]Nc + 0.5BγNγ . (2)
Vanapalli et al. (1996) proposed a relationship for estimating the shear

strength contribution with respect to matric suction, tan φb using the SWRC
and the saturated shear strength parameters as tan φb = S κ tan φ , where S is
the degree of saturation. This term, S κ tan φ , takes account of the non-linear
variation of the shear strength of unsaturated soils using a fitting parameter,
κ. Extending the concepts of shear strength of unsaturated soils proposed
by Vanapalli et al. (1996), equation (2) can be used to predict the bearing
capacity of unsaturated soils which desaturate on application of matric suction
qu = [c + (ua − uw )S ψ tan φ ]Nc ξc + 0.5γBNγ ξγ (3)
where:
ψ = bearing capacity fitting parameter
ξc , ξγ = shape factors due to cohesion and unit weight (from Vesić 1973)
The fitting parameter, ψ used in equation (3) is referred as a bearing
capacity fitting parameter in the remainder of the paper. The bearing ca-
pacity contribution due to matric suction can be obtained from a part of
equation (3), which is equal to (ua − uw )S ψ tan φ . This is similar to predict-

ing the shear strength contribution due to matric suction using the expression
[(ua − uw )S κ tan φ ] from Vanapalli et al. (1996). Similar to prediction of the
shear strength of unsaturated soils, the contribution of matric suction to the
bearing capacity can be determined using the SWRC.
The philosophy of using the fitting parameter, ψ in the bearing capacity of
unsaturated soils is similar to using the fitting parameter, κ for predicting the
shear strength of unsaturated soils. Due to this reason, the limitation of using
equation (3) in the prediction of the bearing capacity of unsaturated soils is
similar to using the equation proposed by Vanapalli et al. (1996) for the pre-
diction of the shear strength of unsaturated soils. In other words, equation (3)
can only be used for predicting the bearing capacity when the experimental
results are available. To alleviate such a limitation for the shear strength of
unsaturated soils, Vanapalli and Fredlund (2000) and Garven and Vanapalli
(2006) provided a relationship between κ versus plasticity index, IP for pre-
dicting the shear strength of unsaturated soils. Such a relationship will be
useful to obtain fitting parameter value from the plasticity index, IP of the
soil.
In the present study, the same philosophy has been extended to propose a
relationship between the bearing capacity fitting parameter, ψ and plasticity
index, IP such that the bearing capacity of unsaturated soils can be predicted
without the experimental results.
Equation 3 will take the form as equation 4 for interpreting the experi-
mental results of the square model footing by including Vesić (1973) shape
factors. Close observation of this equation shows that the equation has been
modified to take account of the air-entry value, (ua − uw )b . Up to the air-
entry value, the contribution of matric suction to the bearing capacity is
equal to (ua − uw )b (tan φ − S ψ tan φ ). The form of this term is similar to
(ua − uw )b (tan φ − tan φb ), which was derived by Oloo (1994). The bear-
ing capacity contribution due to matric suction tan φb in this term has been
replaced with the term S ψ tan φ (see equation 4):
qu = [c + (ua − uw )b (1 − S ψ ) tan φ + (ua − uw )AVR S ψ tan φ ]

Nq B B
× Nc 1.0 + + 0.5BγNγ 1.0 − 0.4 (4)
Nc L L
where
(ua − uw )AVR = 12 [(ua − uw )1 + (ua − uw )2 ] as defined in Fig. 3,
(ua − uw )b = air entry value, kPa.
The bearing capacity factors due to cohesion, Nc , surcharge, Nq and unit
weight, Nγ were developed by several researchers (Terzaghi 1943, Meyerhof
1951, Vesić 1973, Kumbhokjar 1993). The bearing capacity factors of Nc and
Nq proposed by most of the investigators are approximately the same. For
this reason, the bearing capacity factors, Nc and Nq originally proposed by
Terzaghi (1943) using the limit equilibrium method were used in the analy-
sis. There is no general consensus with respect to the bearing capacity factor
due to unit weight, Nγ . The Nγ values proposed by various investigators are

significantly different (Terzaghi 1943, Meyerhof 1951, Vesić 1973). Kumbhok-
jar (1993) has undertaken an extensive study and proposed Nγ values based
on numerical analysis which are relatively higher in comparison to other Nγ
values reported in the earlier literature. This study also shows that bearing
capacity computations provide better comparisons with the measured bear-
ing capacity values using the proposed Nγ values. For this reason, the bearing
capacity factor, Nγ values proposed by Kumbhokjar (1993) are used in this
study.
Comparison between the Measured and Predicted

Bearing Capacity of Unsaturated Soils
Equation (4) is used in the prediction procedure to provide comparisons with

the measured values of bearing capacity for typical model square footings.
The model footings were subjected to static vertical loads in the UOBCE
both under saturated and unsaturated conditions. Different values of capillary
suction (i.e., matric suction) were achieved by varying the water table level in
the test tank (Fig. 1).
Figure 5 shows comparisons between the measured and predicted values of
the variation of bearing capacity with respect to matric suction for the com-
pacted coarse-grained sand studied. There is a reasonably good comparison
between the measured and predicted bearing capacity values. However, the
predicted values are slightly lower than the measured bearing capacity values.
These differences may be attributed to the assumption used in the procedure
Experimental Results
Proposed Equation (4)
Bearing capacity, qu (kPa)
1200
800
400
0
0 2 4 6 8 10
Matric suction, (u a - u w) (kPa)
Fig. 5. Comparison between the measured and predicted bearing capacity versus
matric suction for 100 mm × 100 mm footing
for the estimation of the average matric suction values below the footing,
which typically results in the lower contribution of matric suction compared
to the actual value. Similar trends were observed for the measured and pre-
dicted values of bearing capacity for four other soils data from the literature.
For this reason, the proposed prediction procedure may be summarized to
be conservative as the measured bearing capacity is slightly higher than the
predicted bearing capacity.
All the above observations are derived from the results of model footings
tested in a laboratory environment. More details of the analysis and compar-
isons between the measured and predicted bearing capacity of the other four
soils could not be provided in this paper due to space limitations.
12
Fitting Parameter,
= – 0.0031 IP2 + 0.34 I P + 1
Botkin Pit Silt

8
3 different sands
4 Glacial Till
0
0 5 10
15 20
Plasticity Index, I P
Fig. 6. Relationship between the bearing capacity fitting parameter, ψ and plasticity
index, Ip
The bearing capacity fitting parameter, ψ value was equal to 1 for all the
three sandy soils studied and higher values were required for the other fine-
grained soils. Based on results of the study undertaken through this research
program on five soils, a relationship was developed between the bearing capac-
ity fitting parameter, ψ and the plasticity index, IP (Fig. 6 and equation (5))
ψ = 1.0 + 0.34IP − 0.0031IP2 . (5)
Conclusions
In this paper, a simple technique is proposed for predicting the bearing ca-
pacity of unsaturated soils using the saturated shear strength parameters, c
and φ and the SWRC. The results of the study suggest that there is a good
comparison between the measured and predicted bearing capacity values. The
framework is based on studies undertaken on model footings and shows con-
siderable promise for extending it to field studies.
Acknowledgements
The authors thank the General Public Committee of Higher Education in
Libya (G.P.C.H.E) for providing the financial support to F.M.O. Mohamed
during this research program. The authors would like to acknowledge the
funding received from National Sciences and Engineering Research of Canada
(NSERC) for this project. Thanks also go to C. Catana, graduate student from
the Department of Civil Engineering, University of Ottawa for his assistance.
References
Bolton MD (1986) The Strength and Dilatancy of Sands, Geotechnique 36(1):65–78
Broms BB (1963) The Effect of Degree of Saturation on the Bearing Capacity of
Flexible Pavements, Highway Research Record 71:1–14
Chen FH (1999) Soil Engineering, Testing, Design and Remediation. CRC Press
LLC, N.W., Boca Raton, Florida, USA
Costa YD, Cintra JC, Zornberg JG (2003) Influence of Matric Suction on the Results
of Plate Load Tests Performed on a Lateritic Soil Deposit. Geotechnical Testing
Journal, 26(2):219–226
Fredlund DG, Rahardjo H (1993) Soil Mechanics for Unsaturated Soils, 1st Ed.
Wiley, New York
Garven E, Vanapalli SK (2006) Evaluation of Empirical Procedures for Predicting
the Shear Strength of Unsaturated Soils. Proceedings of the Fourth International
Conference on Unsaturated Soils, Carefree, Arizona, American Society of Civil
Engineers Geotechnical Special Publication 147(2):2570–2581
Kumbhokjar AS (1993) Numerical Evaluation of Terzaghi’s Nγ . Journal of Geotech-
nical Engineering, American Society of Civil Engineers 1999(3):598–607
Meyerhof GG (1951) The Ultimate Bearing Capacity of Foundations, Geotechnique
2:301–332
Miller GA, Muraleetharan KK (1998) In Situ Testing in Unsaturated Soil. Pro-
ceedings of the Second International Conference on Unsaturated Soils, Beijing,
China, Vol. 1:416–421
Mohamed FMO, Vanapalli SK (2006) Laboratory Investigations for the Measure-
ment of the Bearing Capacity of an Unsaturated Coarse-Grained Soil. 59th Cana-
dian Geotechnical Conference, BC, Vancouver (in print)
Oloo SY (1994) A Bearing Capacity Approach to the Design of Low Volume Traffic
Roads. Ph.D. Thesis, University of Saskatchewan, Canada
Oloo SY, Fredlund DG, Gan JK-M (1997) Bearing Capacity of Unpaved Roads.
Canadian Geotechnical Journal 34:398–407
Poulos HD, Davis EH (1974) Elastic Solutions for Soil and Rock Mechanics, John
Wiley and Sons, New York
Prandtl L (1921) Eindringunsfestigkeit und Festigkeit von Schneiden. Z. Angew.
Math. Mech. 1(4):15–20
Schnaid F, Consoli NC, Cudmani RO, Militistsky J (1995) Load-settlement response
of shallow foundations in structured unsaturated soils. Proceedings of the First
International Conference of Unsaturated Soils, Paris, France 999–1004
Steensen-Bach JO, Foged N, Steenfelt JS (1987) Capillary Induced Stresses – Fact

or Fiction? Nineth ECSMFE, Groundwater Effects in Geotechnical Engineering,
Dublin 83–89
Terzaghi K (1943) Theoretical Soil Mechanics. John Wiley and Sons, New York
Vanapalli SK, Fredlund DG, Pufahal DE, Clifton AW (1996) Model for the Pre-
diction of Shear Strength with respect to Soil Suction. Canadian Geotechnical
Journal 33:379–392
Vanapalli SK, Fredlund DG (2000) Comparison of Empirical Procedures to Pre-
dict the Shear Strength of Unsaturated Soils Using the Soil-Water Characteris-
tic Curve, Geo-Denver 2000. American Society of Civil Engineers, Geotechnical
Special Publication 99:195–209
Vanapalli SK, Catana MC (2005) Estimation of the Soil–Water Characteristic Curve
of Coarse-Grained Soils Using One Point Measurement and Simple Properties.
Proceedings of an International Symposium on Advanced Experimental Unsat-
urated Soil Mechanics 401–410
Vesić AS (1973) Analysis of Ultimate Loads of Shallow Foundations. Journal of the
Soil Mechanics and Foundation Division, ASCE 99(SM1):45–73
Influence of Soil Suction on Trench Stability
Valerie Whenham1 , Monika De Vos1 , Christian Legrand1 , Robert Charlier2 ,

Jan Maertens3 , and Jean-Claude Verbrugge4
1
Geotechnical Division, Belgian Building Research Institute (BBRI), Brussels,
Belgium valerie.whenham@bbri.be, monika.de.vos@bbri.be,
Christian.legrand@bbri.be
2
GeomaC Department University of Liège (ULg), Liège, Belgium
Robert.Charlier@ulg.ac.be
3
Department of Civil Engineering, Catholic University of Leuven (KUL), Leuven,
Belgium jan.maertens.bvba@skynet.be
4
Laboratoire Jacques Verdeyen Université Libre de Bruxelles (ULB), Brussels,
Belgium jverbrug@ulb.ac.be
Summary. A research project on the stability of temporary trenches in unsaturated

soil is carried out in Belgium. The main objective of the project is to evaluate the
seasonal variations of suction in the soil and to quantify the consequences of these
suction variations on trench stability. Within the framework of the research, a full
scale instrumented test trench with vertical sides has been excavated in June 2004
at the site of BBRI, characterized by quaternary loam (Limelette, Belgium), to
compare calculations with full-scale observations. As expected first failures occurred
during the winter, when suction in the soil was minimal. This paper presents the
details of this experiment.
Key words: temporary, trench, slope stability, suction, full-scale experiment
1 Introduction
Common methods applied in Belgium for designing trenches and evaluating
slope stability do not take into account the effects of suction when present
in silty and sandy unsaturated soils. The suction is one of the reasons that
steeply inclined slopes remain stable, while this stability can not be proven
by common design rules. Because of the large occurrence of unsaturated loam
and sand soils (possibly after water lowering) during excavations, a research
project on the stability of temporary trenches in unsaturated soil is carried out
in Belgium, with the financial support of the federal ministry of Economical
Affairs. The main objective of the project is to assess the influence of precipita-
tions on trench stability. In a first approach, it is assumed that the influence of
precipitations on slope stability can be related to the variations of the suction
in the soil. In order to evaluate this approach, theoretical predictions – based
496 V. Whenham et al.
on “easily applicable” measurement systems, soil characterization procedures

and calculation methods that can be promoted in the practice – were com-
pared with full-scale observations obtained from the trench test experiments
that are presented in this paper.
2 Geotechnical Context
2.1 Selection of a Site
Because of the large occurrence of unsaturated loam in Belgium, the site
of BBRI at Limelette where the subsoil exists out of overlying quaternary
loam (Ip = 10%) was selected to carry out the test trench experiments. An
extensive geotechnical investigation of the test site is described in Van Alboom
and Whenham (2003).
2.2 Preliminary Investigation

In order to evaluate the soil layering in the test area, eight Electrical Cone
Penetration Tests (CPT-E) were performed on the selected site. Further a
boring with undisturbed sampling was executed to define through laboratory
tests the physical and mechanical properties of the soil (Figs 1 and 2).
Gravimetric soil water content profiles were determined in order to char-
acterise the soil layering in terms of water content susceptibility, and a first
series of suction measuring devices were placed with the aim to assess the
range of suctions developed at different depths as well as the evolution of this
suction with the seasons (Fig. 3).
2.3 Additional Soil Characterization Tests

Additional laboratory tests were performed on undisturbed samples in or-
der to allow further interpretation of the experiment. Besides, unsaturated
Rf (%) qc (MPa)
0 1 2 3 4 5 0 4 8 12 16 20 24
0 0
2 2
4 4
z(m)
z(m)
6 6
8 8
10 10
Fig. 1. Average values of the CPT-E results for the test site at Limelette
Influence of Soil Suction on Trench Stability 497
100
80
60
40 1m depth
2m depth
3m depth
20 4m depth
5m depth
0
10 1 0,1 0,01 0,001
diameter [mm]
Fig. 2. Grain size distribution curve
gravimetric water content (%) suction (kPa)

12 14 16 18 20 22 24 26 28 0 -5 -10 -15 -20 -25 -30
0 0
-1 -1
depth (m)
depth (m)
-2 -2
-3 -3
-4 -4 Date: 17-06-04
Date : 23-03-04 Date: 15-04-04
-5 Date : 17-06-04 -5
Fig. 3. Gravimetric water content and suction profiles at Limelette (preliminary

investigation results)
specimens were tested using a modified triaxial apparatus. Tests were carried
out under constant water (CW) conditions, where the specimen was drained
with respect to the pore air pressure and undrained with respect to the pore
water pressure during shear. Typical derived values of friction angle and cohe-
sion from CU- and CW-triaxial tests are given in Table 1. Soil water retention
curves were determined in the laboratory on undisturbed soil samples using
the axis translation technique. Results are shown in Fig. 4.
Table 1. Typical results of CU and CW triaxial tests (sampling depth = 1.5 m)
Type of Applied suction Effective cohesion Effective friction

triaxial test s [kPa] c [kPa] angle ϕ [◦ ]
CU-test 0 0 32.2
CW-test 20 19.2 35.7
CW-test 50 18.1 35.8
50
Volumetric water content [%]

45
40
35
30
25
20
15
1.5m depth
10
2.5m depth
5
3.5m depth
0
10000 1000 100 10 1 0,1
Suction [kPa]
Fig. 4. Soil water retention curves determined on undisturbed soil samples – axis
translation technique
3 Test Trench Experiment
3.1 Objective of the Experiment
The objective of the experiment was to observe the influence of the seasonal
variations of suction in the soil on the stability of a full scale instrumented
trench, and to compare this influence in terms of soil movements and failures
with very simple calculation methods proposed in the literature (Fredlund and
Rahardjo 1993, Vanapalli and Fredlund 2000). This experiment took place
between June 2003 and June 2004.
3.2 Definition of the Geometry of the Test Trench
The geometry of the test trench was established based on the preliminary
soil investigation data, taking into account the goal that was to observe the
influence of suction degradation on slope stability. A length of 20 m was found
to be sufficient in order to avoid 3D effects. Besides, it was calculated that a
vertical 3 meters depth trench side would only be stable taking into account
the cohesion increase attributable to the suction levels measured in the sum-
mer, while suction levels measured in the winter would not be sufficient. It
was therefore decided to excavate a test-trench of 20 m length, with vertical
3 m depths sides.
3.3 Instrumentation
Based on the test-trench layout and in order to follow the suction and degree
of saturation in the soil, an extensive instrumentation was specified (Fig. 5).
Before the excavation was executed, reference tensiometers were placed all
around the excavation area, at depths varying from 0.5 m to 3.5 m. Further
Fig. 5. Test-Trench layout and instrumentation
instrumentation consisted in watermark sensors for suction measurements and

gypsum blocs for soil water content evaluation. Besides inclinometer casings
and reflectors were installed in order to monitor soil movements.
3.4 Observations
The test trench was excavated in June 2004. Weekly averaged rain measure-
ments obtained for the site as well as averaged suction results are presented in
Fig. 6. For shallow measurements, the influence of precipitations on the suc-
tion results can clearly be observed. However, below 1.5 m depth, the response
to rains in terms of suction variations is significantly “smoothed”. From Fig-
ure 6, it can also be observed that while it rained much more in the summer,
and in particular during the months of July and August, suctions were still
measured even in the upper layers due to the evaporation being much more
important than in the winter (from December to March).
Settlement and horizontal movements of the trench sides were monitored
from June 2004 (before excavation of the test trench) using reflectors and in-
clinometers. Results of these measurements are shown in Fig. 7. Few displace-
ments were measured, and the degradation of the trench side was observed to
be limited to very superficial erosion until first failures occurred.
First failures of the test-trench occurred in January 2005 after heavy rains.
These first failures – registered by webcams focussed on the trench sides – were
shown to be localized and to occur very suddenly. Further failures occurred
during the following days. This first experiment also illustrated the effects of
erosion on the inclined sides (see Fig. 8).
20
rain [l/m³.j]
15
10
5
0
-05
05
4
04
04
4
05
05
5
04
05
04
oct-0
déc-0
avr-0
juil-0
mai-
août-
août-
janv-
janv-
juin-
nov-
juin-
mars
0
suction [kPa]
-10
-20
-30
-40 Tensiometers - Average - 1m 21/01/05 & 11/02/05
Tensiometers - Average - 1,5m (Failures)
-50 Tensiometers - Average - 2,5m
Tensiometers - Average - 3,5m
-60
-05
4
5
4
04
04
05
05
5
4
mai-0
juil-0
déc-0
avr-0
juin-0
oct-0
juin-0
nov-0
août-
août-
janv-
janv-
Fig. 6. Rain measurements and Averaged suction measurements mars
10 displacement (mm)
horizontal displacement [mm]
0 2 4 6 8
8
0
6
2
4
depth (m)
2 4
0
6
16-05-04
10-06-04
05-07-04
30-07-04
24-08-04
18-09-04
13-10-04
07-11-04
02-12-04
27-12-04
21-01-05
15-02-05
Fig. 7. Inclinometer results (horizontal movements measured at a distance of 1 m

from the vertical trench sides): (a) movements measured at 0.5 m depth, (b) dis-
placement profiles measured after the first trench failures
4 Conclusions
This contribution has given a general overview of an experiment carried out

at the site of BBRI, within the framework of a national research project on
the influence of precipitations on trench stability. This experiment involved
the excavation and monitoring of a full scale instrumented test-trench. As
expected, it was observed that the test trench remained stable during the
first seven months, while first failures occurred in the winter when suction
measurements were minimal.
Fig. 8. Upper left: June 2004 – after excavation of the test trench; Upper right:
first localised failures (January 2005); Lower left: eroded inclined sides; Lower right:
generalized failures (February 2005)
References
BBRI – Research report (2005) Stabilité des talus: Méthodes de calcul avec prise
en compte du degré de saturation du sol, et déduction de règles pratiques pour
l’exécution des tranchées et fouilles temporaires, biennale 2003–2005
Fredlund, Rahardjo (1993) Soil mechanics for unsaturated soils, John Wiley & Sons,
USA 517 pp
Vanapalli, Fredlund (2000) Comparison of different procedures to predict unsatu-
rated soil shear strength. In: Shackleford C, Houston SL, Chang N-Y (eds) Ad-
vances in Unsaturated Soils, Geotechnical Special Publication No. 99. Reston,
American Society of Civil Engineers:195–209
Van Alboom, Whenham (2003) Soil investigation campaign at Limelette (Belgium):
Results. In: Proc Symp on screw piles in sand – design & recent developments,
May 7th 2003, Brussels
Index
Abbeche, Khelifa, 3 Farulla, Camillo Airò, 345

Agus, Setianto Samingan, 59 Ferrari, Alessio, 79, 345
Alabdullah, Jamal, 169 Fleureau, Jean-Marie, 243
Alonso, Eduardo, 119
Alshihabi, Omran, 33 Gómez-Espina, Roberto, 267
Ampadu, Samuel, 137 Gallipoli, Domenico, 95
Arifin, Yulian Firmana, 229 Ganjian, Navid, 355
Ayadat, Tahar, 3 Gens, Antonio, 309
Ghembaza, Moulay Smaine, 243
Bönsch, Carola, 145 Guiras, H., 377
Becker, Andreas, 179
Bieberstein, Andreas, 421, 461 Hammoud, Farid, 3
Boháč, Jan, 293 Henken-Mellies, Wolf Ulrich, 369
Herbstová, Vladislava, 293
Campbell, Gaylon, 71 Herle, Ivo, 293
Cardoso, Rafaela, 79 Hu, Liangbo, 413
Charlier, Robert, 495 Hueckel, Tomasz, 413
Cui, Yu-Jun, 259, 475
Iturralde, Enrique Ortega, 119
Défossez-Berthoud, Pauline, 475
Dal Vecchio, Sebastiana, 49 Jamei, Mehrez, 377
De Gennaro, Vincenzo, 161 Javadi, Akbar A., 153, 337
De Vos, Monika, 495 Jockwer, Norbert, 453
Delage, Pierre, 11, 161
Delenne, Jean-Yves, 185 Kazama, Motoki, 437
Dhawan, Sarita, 277 Kierzkowski, Piotr, 301
Diene, Abdoul Aziz, 103
Dobrowolsky, Markus, 285 Laloui, Lyesse, 413
Dueck, Ann, 329 Lee, Samuel, 387
Legrand, Christian, 495
El Youssoufi, Moulay Saïd, 185, 251 Lempp, Christof, 145
Eslami, Javad, 475 Li, Xu, 401
Estabragh, Ali R., 153, 337 Lima, Analice, 79
Evans, Fred, 95 Lloret, Antonio, 309
504 Index
Lourenço, Sérgio, 95 Schwarz, Viktoria, 179

Sejas, Claudia, 445
Maertens, Jan, 495 Servant, Thibaut, 161
Mahler, Cláudio Fernando, 103 Smith, David, 71
Mascolo, Giuseppe, 49 Soulié, Fabien, 185
Medero, Gabriela, 95 Stroosnijder, Leo, 33
Miehe, Rüdiger, 453 Sun, Shulin, 195
Mir Mohammad Hosseini, Seyed
Majdeddin, 355 Tamrakar, Surendra Bahadur, 207
Mishra, Anil K., 277 Tang, Anh-Minh, 259, 475
Mitachi, Toshiyuki, 207 Tavallali, Abbass, 259
Mohamed, Fathi, 483 Taïbi, Said, 243
Mokdad, Ghazi, 33 Teare, Brody, 71
Mokni, N., 377 Toll, David, 95
Monroy, Rafael, 41 Toyosawa, Yasuo, 207
Péron, Hervé, 413
Unno, Toshiyasu, 437
Pashang Pisheh, Yadollah, 355
Priol, Grégoire, 161
Vanapalli, Sai, 483
Rao, Sudhakar M, 277 Verbrugge, Jean-Claude, 495
Ridley, Andrew, 41 Villar, María Victoria, 267, 309
Rojas, Juan Carlos, 445 Vrettos, Christos, 179, 285
Romero, Enrique, 79, 119, 345
Rothfuchs, Tilmann, 319, 453 Whenham, Valerie, 495
Russo, Giacomo, 49 Witt, Karl Josef, 219
Sánchez, Marcelo, 309 Xu, Huifang, 195

Saix, Christian, 185, 251
Salager, Simon, 251 Zdravkovic, Lidija, 41
Salinas, Luis Mauricio, 445 Zeh, Rainer M., 219
Schanz, Tom, 59, 169, 229 Zhang, Chun-Liang, 319, 453
Scheuermann, Alexander, 421, 461 Zhang, Limin, 401
springer proceedings in physics
66 The Structure and Conformation 79 Nonlinear Dynamics
of Amphiphilic Membranes and Pattern Formation
Editors: R. Lipowsky, D. Richter, in Semiconductors and Devices
and K. Kremer Editor: F.-J. Niedernostheide
67 Nonlinearity with Disorder 80 Computer Simulation Studies
Editors: F. Abdullaev, A.R. Bishop, in Condensed-Matter Physics VIII
and S. Pnevmatikos Editors: D.P. Landau, K.K. Mon,
and H.-B. Schüttler
68 Time-Resolved Vibrational Spectroscopy
V 81 Materials and Measurements
Editor: H. Takahashi in Molecular Electronics
Editors: K. Kajimura and S. Kuroda
69 Evolution of Dynamical Structures
in Complex Systems 82 Computer Simulation Studies
Editors: R. Friedrich and A. Wunderlin in Condensed-Matter Physics IX
Editors: D.P. Landau, K.K. Mon,
70 Computational Approaches and H.-B. Schüttler
in Condensed-Matter Physics
Editors: S. Miyashita, M. Imada, 83 Computer Simulation Studies
and H. Takayama in Condensed-Matter Physics X
71 Amorphous and Crystalline and H.-B. Schüttler
Silicon Carbide IV
Editors: C.Y. Yang, M.M. Rahman, 84 Computer Simulation Studies
and G.L. Harris in Condensed-Matter Physics XI
Editors: D.P. Landau and H.-B. Schüttler
72 Computer Simulation Studies
in Condensed-Matter Physics IV 85 Computer Simulation Studies
Editors: D.P. Landau, K.K. Mon, in Condensed-Matter Physics XII
and H.-B. Schüttler Editors: D.P. Landau, S.P. Lewis,
73 Surface Science
Principles and Applications 86 Computer Simulation Studies
Editors: R.F. Howe, R.N: Lamb, in Condensed-Matter Physics XIII
and K. Wandelt Editors: D.P. Landau, S.P. Lewis,
74 Time-Resolved Vibrational Spectroscopy
VI 87 Proceedings
Editors: A. Lau, F. Siebert, and W. Werncke of the 25th International Conference
on the Physics of Semiconductors
75 Computer Simulation Studies Editors: N. Miura and T. Ando
in Condensed-Matter Physics V
Editors: D.P. Landau, K.K. Mon, 88 Starburst Galaxies
and H.-B. Schüttler Near and Far
Editors: L. Tacconi and D. Lutz
76 Computer Simulation Studies
in Condensed-Matter Physics VI 89 Computer Simulation Studies
Editors: D.P. Landau, K.K. Mon, in Condensed-Matter Physics XIV
and H.-B. Schüttler Editors: D.P. Landau, S.P. Lewis,
77 Quantum Optics VI
Editors: D.F. Walls and J.D. Harvey 90 Computer Simulation Studies
in Condensed-Matter Physics XV
78 Computer Simulation Studies Editors: D.P. Landau, S.P. Lewis,
in Condensed-Matter Physics VII and H.-B. Schüttler

(Tom Schanz) Experimental Unsaturated Soil Mechani

Uploaded by

Copyright:

Available Formats

(Tom Schanz) Experimental Unsaturated Soil Mechani

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Tom Schanz) Experimental Unsaturated Soil Mechani

Uploaded by

Copyright:

Available Formats

springer proceedings in physics 112

springer proceedings in physics

Volumes 66–90 are listed at the end of the book.

With 299 Figures and 67 Tables

Library of Congress Control Number: 2007920611

Printed on acid-free paper SPIN: 11661139 42/Integra 5 4 3 2 1 0

The event is a continuation of the series of International Conferences on Un-

Weimar, Tom Schanz

Part I Microstructure and Fabric

Inﬂuence of Relative Density and Clay Fraction on Soils

Microstructure Features in the Behaviour of Engineered

Microstructure of a Lime Stabilised Compacted Silt

Part II Measuring Suction

Errors in Total Suction Measurements

A Comparative Study of Soil Suction Measurement Using

Determination of the Soil Water Retention Curve with

Tensiometer Development for High Suction Analysis in

Part III Strength and Dilatancy

Dilatancy of Coarse Granular Aggregates

Shear Strength Behaviour of Unsaturated Silty Soil

Eﬀect of Capillary and Cemented Bonds on the Strength of

Determining the Shear Strength of Unsaturated Silt

The Tensile Strength of Compacted Clays as Aﬀected by

Part IV Temperature Eﬀects

Modiﬁed Isochoric Cell for Temperature Controlled Swelling

Thermo-Hydro-Mechanical Behaviour of Compacted

Part V Volumetric Behaviour – Expansive Materials

Experimental Study on Shrinkage Behaviour and Prediction

Assessment of Swelling Deformation of Unsaturated Kaolinite

Oedometer Creep Tests of a Partially Saturated Kaolinite

Analysis of the Expansive Clay Hydration under Low

Moisture Eﬀects on Argillaceous Rocks

Part VI Retention Behaviour

Results from Suction Controlled Laboratory Tests on

Variation of Degree of Saturation in Unsaturated Silty Soil

Prediction of Soil–Water Characteristic Curve Based on Soil

A Retention Curve Prediction for Unsaturated Clay Soils

Unsaturated-Zone Leaching and Saturated-Zone Mixing

Prediction of SWCC for Coarse Soils Considering Pore Size

The Inﬂuence of the Pore Fluid on Desiccation of a

Determination of the Soil Water Retention Curve and the

Part VII Field Applications

Earthquake-Induced Mudﬂow Mechanism from a Viewpoint

Plate-Load Tests on an Unsaturated Lean Clay

Selfsealing Barriers of Clay/Mineral Mixtures. The SB

Microstructure and Fabric

Khelifa Abbeche1 , Farid Hammoud1 , and Tahar Ayadat2

Summary. Most of the works conducted to investigate the parameters likely to

2 Materials, Equipment and Testing Procedure

Table 1. Geotechnical characteristics of soils

Fig. 1. Grain size distribution curves of reconstituted soils used

Fig. 2. Typical collapse potential one dimensional consolidation test

3 Results and Discussion

Table 2. Relation of collapse potential to the severity of foundation problem

CP (%) Severity of problem

by Lawton et al. (1989) and Ayadat et al. (1998). Consequently, it is noticed

Table 3. Proposed classiﬁcation of collapsible soils

Clay Fraction CF (% < 2 μm) Probability of collapse

This paper has presented results of one-dimensional oedometer collapse tests