TL 637 r %

_ X Draft
..... Translation
July 1977 637

- iL
I C THERMAL CONDUCTIVITY OF SOILS:

o O. Johansen

SEP 12 1977

q=
•CORPS OF ENGINEERS, U.S. ARMY
"COLD REGIONS RESEARCH AND ENGINEERING LABORATORY
HANOVER, NEW HAMPSHIRE

""i
 Unclassified
S EC
U R ITY C L ASSI FI
C AT ION O F T1HIS P A G E("' On D ataEntered)

RFPORT OCCUMENTATION PAGE BRE COMPLETIFOR

I. REPORT mUMSCIR 2. GOVY ACCESSION NO 3.fEC1011ENT* CATALOO NUMBER

Draft Translation 637 2k' '' "'

41 1TITL&g(4id lu&*ie). -.. r
-:- o ?
PwiEP& 4 PERIOD COVKERE

ERMA CZONDUCTIVITY OF SOILS~ _ _ _ _ _ _ _ _ _

.. PERFORMING ORO. REPORT NUM8K.

T_ U.TH.R S. ON"RACT OR GARNA NUTM[R(,)

JJ Johansen
P EIF'ORMING ORGANIZATION NAME AND ADDRESS 10. PROGRAM £LEMNT1. PROJECT, TASK
AREA, A ORK UNIT NuMaiR$
U.S. Army Cold Regions Research and .-

Engineering Laboratory
Hanover, New Hampshire
I I. CONTROLL.ING OPICE NAME AND ADDRESS .1 . ReowVOWT0Av
/iJul1.r'0

14. MONITORING AOENCY N AME 6 ADORCS(It diflftert Iom Ce.illing 0ics) S. SECURITY CL'X . (of t piw t)

sol AUSIFICATION/ OOWNRAOING

IS. 0ISTRIBUTION STAT EMENT (of thl. R)erot)

Approved for public release; distribution unlimited.

17, 0IUTRIdUTION STATEMENT (.tho bettract ontord In Sleok 20, It diffren trm Repopte)

IS. SUPPLEMINTARY NOTES

/• / ' 'j .,
S.:. tI. KEY WOROS (ContInue an ?over*@ side it n.t laff aid Identify by btlok number)

THERM'AL CONDUCTIVITY SOILS--MOISTURE CONTENT

HEAT TRANSFER

20. AGSTRACT CrI•,n•u. on rover" Oldi It n•w.cae•. ind identify by block iu•.be)

The aim of this investigation has been to create a mathematical model for
calculating thermal conductivity of soils with ordinary soil parametirs as
~J~A input data. One part of this work has been devoted to literature studies on
heat-transfer mechanisms in moist materials. These studies have made it pos-
sible to give bounds for the different domains where the various mechanisms
have an appreciable influence on the total heat transfer.

DO 1473 ' 101TION

•" " OF I NOV 66 IS ONSOLgrg Unclaseaf Led fAjEWtDt.Etrd .
•'•' II~SEURlITY CLAR!IFCTlllON orI TmiI PAGE[ (KMVld
DIn- i@reed)
 DRAFT TRANSLATION 637

Nils Whila Soctinn

ENGLISH TITLE: THERMAL CONDUCTIVITY OF SOILS DOC •El
secitll
.INANNOINCrD E)
JUS[II IAI ION .........
• .................................... ..........

FOREIGN TITLE: VARMELEDNINGSEVNE AV JORDARTER OB /AYALAUL COOES

. . . . . . .. .. ........
.SPECI.

AUTHOR: 0. Johansen

1 •,. SOURCE: Trondheim, Group for Thermal Analysis of Frost in the

Ground, Institute for Kjoleteknikk, 1975, 231p.

IAK

Translated by Rusetta Stone, Nashua, N.H. for U.S. Army Cold Regions
Research and Engineering Laboratory, 1977, 2 9 1p.
d "'

NOTICE
The contents of this publication have been translated as presented in the
original text. No attempt has been made to verify the accuracy of any
statement contained herein. This translation is published with a minimum
of copy editing and graphics preparation in order to expedite the dissemination
of information. Requests for additional copies of this document should be
addressed to the Defense Documentation Center, Cameron Station, Alexandria,
Virginia 22314.
',.
 TRANSLATOR'S
NOTES: To the translation of "THERMAL CONDUCTIVITY OF SOILS",
by Oistein Johansen.

The following general remarks apply.

1. The author makes no distinction between sections (arabic or

"normal" numerals) and sub-sections (A, B, C, etc.). Thus,
references of the type "discussed in the previous section"
may refer to either section or subsection.

2. Certain words or expressions used by the author may not have a

direct, one-to-one correspondence in English. In such cases,
alternate translations are given in parenthesis.

3. Expressions .of the type "Hazens formel" can mean either "Hazen's
formula" or "Hazens' formula". The translation may contain some
errors in this respect.

"4. The original is notoriously inconsistent in the use of sub-

scripts, both in formulas and in the text. A major effort
would be required to correct for this and the translation fol-
lows the original in this respect.

5. The author was at one point associated with "Institutt for

Kjuleteknikk" at NTH (Norwegian Technical University). In at
least one figure caption, the original presents the translation:
"Institute for Refrigeration Engineering", which seems to be a
misnomer. This group is clearly active in the study of thermal
phenomena associated with frost penetration in and prevention
of frost damage to various kinds of structures in a cold (artic)

I climate, The least of their worries would appear to be enginee-

ring or design of refrigeration systems, since nature already
has taken care of that aspect. "Institute for Cold Technology"
is proposed as a more descriptive translation.
6. About 20 percent of the sentences in the original contain
a word or phrase equivalent to "however". A concentrated
effort has been made to eliminate as many of these as possible,
particularly when they occur in two or more consequtive aenten-
ces.

A number of foot-notes have been included in the translation to

clarify certain items. The following comments were deemed too
long for inclusion as foot-notes:

1. In chapter I, Eq 31, the meaning of d 1 0 is not entirely clear.

From the comments following Eq 30 it appears that d1 0 means
"the diameter that 10 percent of the particles are less than".
The line below Eq 31 ends with "gjennomgang", which means "pas-
sing through" or "penetration"t. That would indicate that the
subscript 10 for d some-how relates to the percentage of water
"passing through" or "penetrating" the material. Another pos-
sible interpretation of "gjennomgang" is "intersection", i.e.
the point where the diameter distribution intersects the 10 per-
cent level,

2. Water content is denoted by upper and lower case w in diffe-

rent parts of the report. The translation may retain some of
these inconsistencies.

3. Eqs 27 and 28 of chapter II are based on the following two re-

sistor networks

Eq 27:

L•vv•._•pv•._--e '-4I

(~1b

~r
 Eq 28: -A" ---

Heat flow 4

5. The appendix treating heat flow in cylindrical bodies contains

certain ambigous notations, In the translation, lower case x
without subscript is used only in Eq 6, while capital X has been
used for the parameter defined by Eq 8. Also, expressions of the
type beix 2 have been changed to the better defined form (beix )z
(see Eq 14).

4-.4
 THERMAL CONDUCTIVITY OF SOILS *

OF
Siviling. 0istein Johansen
Institutt for kjoleteknikk
703 4-Trondheim-NTH
1975

We

I~i

• With summary and figure captions in Engliuh.

 THERMAL CONDUCTIVITY OF SOILS

0istein Johansen
FOREWORD

This report will be submitted as a Ph.D. thesis. The work on

which it is based was begun in the fall of 1968 in the form of a
M.Sc. thesis at the Institute for Cold Technology, Norwegian
Institute of Technology (NTH), where I performed theoretical and
experimental studies of heat convection in coarse fill materials.
The National Road Laboratory took an interest in this work and
funded an extension of the study.

At that time, a more. extensive research program was initiated

under the direction of the "Committee for Frost in the Ground",
with economic support from the Norwegian Scientific Research
Council and the Highway Authority. Some of the work was performed
by a group at NTH, with participation by the Institute for Cold
Technology. This group concentrated on generating theoretical
methods for analysis of thermal conditions in structures subjected
to frost. The work included a study of thermal parameters for
soil materials, which since that time has been my field of acti-
vity. This program has thus provided the funds for performing the
investigations on which this report is based.

Cooperation within the research team at NTH has been of great

value to the investigation and has contributed toward a goal-oriented
study. I also wish to stress the importance of taking part in
meetings and conferences arranged by the Committee for Frost on

S~iii
the G:'Qund "or the final form of the report. Work on presenta-
tions and papers has taken considerable time but also provided

valuaile experience in arranging the material.

The professional support given by my co-workers at the Institute

has also been of great importance for the results presented here.
In particular, I wish to thank Mr. Per Erling Frivik, M. Sc.,
who has directed the work of the group, as well as other
co-workers who have performed much of the practical work in
connection with the development of test equipment and who have
given valuable help in performing the experiments.

Special thanks are extended to my thesis supervisor, Professor

Gustav Lorentzen, Sc.D., for advice and encouragement during my

work on the thesis.

The report was written while I was employed at the National Road
Laboratory. I wish to thaak the management and colleagues here
for the consideration they have shown despite the fact that this
work has taken more time than originally estimated.

Oslo in June, 1975 0istein Johansen

iv
 INDEX
P age
Foreword iii
Index v
Symbols viii
1. Latin letters viii
2. Numbers independent of units x
3. Greek letters xi
4. Mathematical symbols xiii
5. Often used indexes (subscripts) xiii
6. Conversion factors xiv
Bibliography xv
INTRODUCTION 1
CHAPTER I: HEAT TRANSPORT IN MOIST SOIL 3
1. Thermal Radiation 4
"2. Combined Moisture and Heat Transport 11
A. Water vapor diffusion 11
B. Capillary water transport 14
C. Contribution of vapor diffusion
to thermal conductivity 17
3, Convection 20
A. Heat Transport 20
B. Liquid transport in poro.us materials- 23
C. Forced convection 26
D. Free convection 31
4. Heat Transport in Freezing Soils 38
5. The Effects of Various Heat Transport
Mechanisms 42
REFERENCES 47
CHAPTER II: THEORIES FOR HEAT CONDUCTION IN COMPOSITE
MATERIALS 50
1. Exact Theory 50
"2. Evaluation of Limits 54
A. Limits according to Hashin-Shtrikman 54
B. Limits based on point correlations 57
C. Other evaluations of limits 62
D. conclusion 65
3. Mathematical Models 67
A. Introduction 67
B. "Exact" calculations 68
C. Approximate methods based on elec-
trical field theory 70
D. Approximate calculations based on
elements connected in series or shunt 76
E. The geometric mean model 80
F. Conclusion 83
REFERENCES 85

V
 Page
CHAPTER III: SOIL PARAMETERS 88
1. Thermal Conductivity of Soil Generating
Materials 88
A. Soil generating minerals 88
B. Quartz content in loose deposits
C. Thermal conductivity for the compo-
nents 10
2. Volumetric Composition and Texture 104
A. Volume ratios in soil materials 104
B. The importance of texture 107
C. Unfrozen water 113
3. Soil Material Classification 118
A. Geotechnical analysis 118
B. Supplemental analysis 123
REFERENCES 125
CHAPTER IV: EXPERIMENTAL METHODS FOR DETERMINING
CONDUCTIVITY IN SOIL MATERIALS 127
1. Thermal Conductivity Measurements
in Materials Containing Moisture 127
2. Stationary Methods 131
A. Kersten's cylindrical test unit 131
B. Planar test unit 136
3. Thermal Conduction Probes 143
A. Influence of finite size and varying 143
thermal properties
B. The effects of moisture transport 149
C. Experimental arrangement 152
4. Other Transient Methods 157
A. Linear temperature rise 157
B. Constant thermal flux 161
C. Periodic variation of temperature 165
D. Conclusion 173
REFERENCES 174

CHAPTER V: EXPERIMENTAL INVESTIGATIONS OF THERMAL CONDUCTIVITY

IN SOIL MATERIALS 177
I. Dry Soil Materials 178
A. Introduction 178
B. Measurements performed by Smith 180
C. Investigation performed by the author 181
0. Empirical relations 189
E. Stabilized materials 193
F. Conclusion 197

2. Saturated Mineral Soil Materials 199

A. Introduction 199
B. Measurements performed by Kersten 200
C. Ocean Sediments 202
D. Conclusion 204

: vi ]4
 Page
3. Moist Mineral Oils 205
A. Introduction 205
B. Study by Kersten 207
C. Investigation of Norwegian soil
matefials 214
D. Conc us on 217
4. Soil Materials Containing Humus 218
A. Introduction 218
B. Measurements on peat and bog 219
REFERENCES 222
CHAPTER VI: METHODS FOR COMPUTING THERMAL CONDUCTIVITY OF
SOIL MATERIALS 224
1. Existing Methods 224
A. Analytical methods 224
B. Empirical methods 230
C. Conclusion 246
2. New Methods for Determining Thermal
Conductivity in Soil Materials 248
A. Introduction 248
B. Mathematical model for mineral soils 248
3. Effects of Variations in Soil Parameters 254
SA. Effects on conductivity calculations 254
B. Conclusions 266
4. Effects of Conductivity Variations ... 270
A. Introduction 270
B. Effects of conductivity, variations 271
C. Conclusion 275
REFERENCES 276
SUMMARY 278
POST-SCRIPT 284
APPENDIX 286
k), REFERENCES 291

vii
 SYMBOLS

i. Latin letters

A attenuation ratio
a X/cp
X thermal diffusivity m2 /s
A Xe/(Cp)f thermal diffusivity, included in
M Rayleigh's number m2 /s
a,bc,d constants in Kersten's equations -

c V Ecp thermal capacity, by volume J/m3 K

c specific heat J/kgK
c clay content
Da diffusion coefficient for water m2 /s
vapor in air
Dev isothermal water vapor diffusivity m2/s

Dev isothermal water diffusivity m2 /s

DT v thermal water vapor diffusivity m2 /sk

d diameter (of particles) m, m

dekv equivalent diameter m

E electrical field strength v/m
F frost quantity h°C
Fi, i 1,2''.
1 factor in Fricke's equation -

f form factor, in expressions for -

specific surface area
G1,G2 factors depending on average
cell geometry for components "l"
and "2". Used in Fricke's
equations.
q gravity acceleration M2 /s

viii I'

A
gi' i = a,b,c form factor. Used in Fricke's
equations.
h thickness
i potential gradient m/m
K hydraulic conductivity m/s

K " k/W permeability m/s (cm/s)

k permeability m2

L latent heat of water J/kg, Wh/kg

1 length m

n porosity

nj, i 1 ,2".. relative volume for component "i"

ov relative volume of organic
materials
P total pressure N/m2

p partial pressure N/m2

N/r 2
p saturation pressure
q quartz content
q heat flux W/m2
ql water flux kg/m2s

qv water vapor flux kg/m 2 s

R gas constant m2 /.-k

r vaporization heat J/kg

r radius m

S specific surface area m2 /m 3

Sr degree of saturation
T temperature K, °C
t time s

v specific volume M3/kg

ix
V flow velocity m/s
V5 , V1 0 water content which gives 5 (10)
mm penetration (sinking)of cone in
consistency measurements
(relative to dry weight)
w relative water content by dry
weight
wv relative water volume
ww relative water content by wet
(total) weight
w25 1 wl0 0 water content which causes flowing
for 25 (100) blows in Lassagrande's
flow limit test-set (ratio by
weight)
wL (-W 2 5 ) flow limit
wu unfrozen water ratio by weight
(water/dry material)

x - r Fw/a angular velocity

x'y'z space coordinates m

2. Numbers independent of units

Fo - at/r 2 Fourier's number

Ke u(- X)
(XI - o) Kersten's number. Normalized
conductivity.
Nu Ae/X°e Nusselt's number; free
convection.
NUr - •rd/X2 Nusselt's number; radiation
?am- dov"(cp)f/xf Peclat's number, modified
Ra - •ghk4T Rayleigh's number, free convection
in porous materials
amy

Rak critical Rayleigh's number

I
 3. Greek letters

expansion coefficient, by volume Oc-I

-XZ/Xl conductivity ratio
constants (coefficients) in Kersten's
equations
factors (coefficients) in equation
for unfrozen water content
factors (coefficients) in empirical
equation
r
M. iat transfer coefficient for radia- 2
" -•-X1 normalized conductivity W/
B rate of temperature rise °C/s
B dimension in cubical unit cell
Yd dry density kg/m3
"Ys specific weight
Ym specific weight of mineral compo-
nent kglm3

Y specific weight of organic compo-

nent kg/m
denotes differences or small
V;11 variations
.. gap opening (width)
Semission factor (coefficient)
C dielectric constant As/V4)
n dynamic viscosity Ns/r 2

, moisture content by volume (in

Chapter I)
Ai, i - l,2' integrals containing point corre-
lations (Brown's equation).
thermal conductivity W/mK
AS, i - 1,2e'. thermal conductivity of component
TF'TEI'is the unit for absolute dielectric constant. The para-
meters el and c2 used in Eq. 20 of Chapter I1 are relative
xi
constants, i.e. pure numbers (Translator's note).

thermal conductivity of saturated

material W/MK
X0 thermal conductivity of dry
material

e effective thermal conductivity,

e.g. including contributions from
convection
X0 steady state conductivity
X*,X•,upper and lower Hashin-Shtrikman-
limits W/IMK
X subscr. other subscripts (indexes) are
explained in the text where they
occurr. "

v chinematic viscosity m2 /s
frost depth m
mathematical symbol for product,

p
a ai - a1 a 2 .... ap
i-I

0 density
mathematical symbol for sum, e.g.

D ai - a& + +.... +ap

radiation constant for black

bodies W/m2K4
I potential N/m2 ,m

V
0 relative humidity
dimension of cubical unit cell
Oi(r) function which assumes values 0 or
1 if the (position) vector r falls
outside or inside component "i",
respectively

S"' • - •--• • --- i .... "" ' , ",• • -- • '-• • = • -mxii• ,'
I,

-- 4
 moisture potential (suction) N/m2 ,m
angular velocity rad/s

4. Mathematical symbols

V gradient operator, e.g.: 1)

VT - (aT/ax)t + (aT/3y)3 + (aT/3z)t.
V2 Laplace's operator, e.g
V2 T- *aT/ ax2 + •2T/ay 2 + a 2 T/az 2

-• denotes vector quantity

average by volume± e.g.:
•oj(3' * t0r()
10 dv

5. Often used indexes (subscripts)

a effective, e.g. effective conductivity X., which

A/ includes contributions from convection, etc.

ill ,f liquid,
liquid (cp) thermal capacitance by volume for a
e.g. f.

i - l,2,''" denotes component. In composite materials, the

components are numbered according to increasing
conductivity: Xi < Xz <....< too-
.< . Xp

i I a,bc subscripts for form factors in Fricke's equation.

v volume, e.g. water content by volume Wv.
0 when occurring as a superscript for conductivity
X, dry material is indicated.
1 when occurring an a superscript for conductivity
X, saturated material is indicated.

i) t, • and J denote unit vectors along s, y and z-axes, respectively

(Translator's note).

xiii
6. Conversion factors

I kcal - 4187 Ws (J) w 1,163 Wh

I kcal/mh 0 C - 1,163 W/mK

xii

xiv S
 BIBLIOGRAPHY

P. Adivarahan et al: Heat Transfer in Porous Rocks. Society of

Petroleum Engineers Journal, 1962, pp. 290-296.
L. Alberta et al: The Influence of Low Moisture Content on the
19i, u tivities of a Granular Substance. Brit. J. Appl. Phys.
pp. 951-955. 17 S
M. Ambramowitz and I. Stegum. Ed.: Handbook of Mathematical Functions.
Nat. Bur. of Standards, Appl. Math. Ser., New York, 1969, p. 374.
"Analyseforskrifter. Statens vegvesen, 1966.
1. Jordarter. 111. Spesifikk vekt - vanninnhold - romvekt.

Analyseforskrifter. Statens vegveuen, 1966, 1 Jordarter, 141.

Mekaniske egenskaper.
Analyseforskrifter. Statens vegvesen, 1966, 1 Jordarter, 161.
Komprimering, p. 5.
D.M. Anderson and A.R. Tice: Predicting unfrozen water contents in
frozen roils from surface area measurements. Highway Res. Rec.
No. 393, 1972, pp. 12-18.
E. Angen: Analyse av vanninnhold i vegkonstruksjoner. Intern
rapport. SINTEP, NTH, Trondheim, 1973.
W.B. Argo and J.M. Smith: Heat Transfer in Packed Beds. Chemical
Engineering Progress, 49 (8), 1953, pp. 443-451.
/" B
L. Barden at al: The Col1a se.Mechan4sm in Partly Saturated Soil.
Eng naering Geology, 7 (1), 1973, pp. 49-60.

Beck: Transiený Determination of Thermal Properties.

Nuclear Engineering and Design,.3, 1966, pp.
373-381.
M. Beran: Use of the Vibrational Approach to Determine Bounds
for the Effective Permittivity in Random Media.
Nuovo Cimente. 38, (2), 1965, pp. 771-782.
M.J. deran: Application of Statistical Theories to Heterogeneous
Materials, Phys. Stat. Sol. (a), 6, 1971, p. 365.

E. Berge and 0. Aarvold: Some physical properties of grass silage.

Meldinger fra Norges Landbruksh~gskole. 52,
(1), 1974.
G.S.G. Beveridge and D.P. Haughey: Axial Heat Transfer in Packed
Beds. Stagnant Beds Between 20 and 750 0 C. Int.
J. Heat Mass, 14, 1971, pp. 1093-1113.
F. Birch and H. Clark: The Thermal Conductivity of Rocks. American
Journal of Science, 238, (5), 1940, pp. 529-558.

xv
J.H. Blackwell: A. Transient Heat Flow Method for Determining
the Thermal Constants. J..Appl. Phys. 25, (2),
1954, pp. 137-144.
J.H. Blackwell: The Axial-Flow Error in the Thermal Conductivity
Probe, Can. J. Phys. 34, 1956, pp. 412-417.
C.A. Bower and J.0. Goertzen: Surface area of soils and clays by
an Equilibrium Ethylene Glycol Method. Soil
Science, 87, 1959, pp. 289-292.
E. Brendeng and P.E. Frivik: New Development in Design of Equipment
for Measur•.a Thermal Conductivity and Heat Flow.
Institutt for kjoleteknikk, Trondheim, NTH, 1973.
W.F. Brown: Solid Mixture Permittivities. The Journal of
Chemical Physics, 23, (8), 1955, pp. 1514-1517.

W.F. Brown, Jr.: Dielectric Constants, Permeabilities, and Conducti-

vities of Random Media. Trans. of the Soc. of
Rheology 9, (1), 1965, pp. 357-380.
D.A.G. Bruggeman: Dielektrizitatskonstanten und Leitfahigkeiten von
Vielkristallen der nichtregularen Systeme. Ann.
Physik 5, (25), 1936, pp. 645-664.
Burke and Plummer: Referert i Chemie Ing.-Techn., 25, 1953, p. 262.
Bygg. Huvuddel lB. Almanna Grunder: Avd. 17:4, pp. 306-308.
AB Byggmastarens Forlag Stockholm 1972.
C.J.F. Hottcher (1938): Referert av D. Polder (1946). See ref. (36).

C
J.W. Cary: water Flux in Moist Soil: Thermal Versus Suction
Gradients. Soil Science, 100, (3), lq55,
pp. 168-175.
H.S. Carslaw and J.C. Jaeger: Conduction of Heat in Solids. Sec.
ed. Oxford Press, 1959, p. 345.
J.C. Champoussin: Determination Simultance des caracteristiques
termocinetiques des solides. Int. J. Heat.
Mass, 15, 1972, pp. 1407-1418.
S.C. Cheng and R.I. Vachon: The Prediction of the Thernal Conductivity
of Two and Three Phase Solid Heterogeneous Mixtures.
Int. J. Heat Mass. Transfer. 12, 1969, pp. 249-264.
S.C. Cheng and R.I. Vachon: A Technique for Predicting the Thermal
Conductivity of Suspensions, Emolusions and Porous
Materials. Int. J. Heat Mass. Transfer, 13, 1970,
pp. 537-546.
xvi
C. Codegone et al: Resonant Thermal Waves in Insulating Slabs.
1IR. 4
Corn.x..ssion 2. (Trondheiin) 1966. (Saertrykk)
S.R. Coriell and J.L. Jackson: Bounds on Transport Coefficients of
Two-Phase Materials. J. Appl. Phys. 39, (10),
1968, pp. 4733-4736.
P.B. Corson: Correlation functions for predicting properties
of heterogeneous materials. I. Experimental
nmeasurements of spatial correlation functions in
multiphase solids, J. Appl. Phys. 45, (7),
1974, pp. 3159-3164.
P.B. Corson: Correlation functions ... 1. Empirical construction
of spatial correlation functions for two-phase
solids. J. Appl. Phys. 45, (7), 1974, pp. 3165-3170.
P.B. Corson: Correlation functions ... IV. Effective thermal
conductivity of two-phase solids. J. Appl. Phys.
45, (7), 1974, pp. 3180-3182.
D. Croney and J.D. Coleman: Pore Pressure and Suction in Soil. Fra
"Pore Pressure and Suction in Soils". Butterworths.
London 1961, pp. 31-37.
J.A. Currie: Gaseous Diffusion in Porous Media. Part 2 Dry
sc
Uanylar telfs,
pp.
Bri jurnal of Applied
3 ?h

J.A. Currie: Gaseous Diffusion in Porous Media. Part 3 - Wet

Granular Materials. British Journal of Applied
Physics, 12, 1961, pp. 275-281.

W.E.A. DAvies: The theory of composite dielectrics. J. Phys.

D: Appl. Phys. 4, 1971, pp. 318-327.
F. De POnte and P.E. Frivik: Automatic Control of Guarded Hot Plate
Apparatuses. Institutt for kjoleteknikk, Trondheim,
NTH, 1972.
D.A. de Vries: Het warm=tgeleidingsvermogen van grond. Med.
Landbouwhogeschool. Wageningen, 52, 1952.
D.A. de Vries: Simultaneous Transfer of Heat and Moisture in
Porous 39, (5), Transactions,
Union, Media. 1958. American Geophysical

D.A. de Vries and A.J. Peck3 On the Cylindrical Probe Method of

Measuring Thermal Conductivity. Aust. J. Physics,
11, 1958, pp. 255-271.

xvii
R.G. Deissler and J.S. Boegli: An Investigation of Effective Thermal
Conductivities of Powders in Various Gases.
Transactions of the ASME, October 1058, pp. 1417-
1425.
P.B. Desphande and J.R. Cooper: Thermal Conductivity of Two-Phase
Systems. J. Heat Transfer. Trans. of the ASME,
May 1972, pp. 249-264.

E
J.W. Elder: Steady Free Convection in a Porous MediutL Heated
from Below. J. Fluid Mech. 1967, 27, pp. 29-48.
M.A. z0lsayed: Bounds for effective thermal, electrical, and
magnetic properties of heterogeneous materials
using higher order statistical information. J.
Math. Physics, 15, (ii), 1974, pp. 2001-2015.
I.M. Eltantawy and P.W. Arnold% Reappraisal of Ethylene Glycol
Mono-.Ethyl Ether (RGME) Method for Surface area
Estimation of Clays. Journal-Soil Science, 24,
1973, pp. 232-238.
S. ERgun: Fl~asigkeitsstr6mung durch Fallk5rpers~ulen
(Festhetten). Aeferert i Chemie Ing.-Techn,, 2ý,
1953, p,. 262.

O.T. Farouki: Physical Properties of Granular Materials with

Reference to Thermal Resiscivity. H{ighway Res.
Rec. 128, 1966., pp. 25-43.
T.P. Fidelle and H.S. Kirk: A study of unidirectional Versus Tri-
directional Heat Flux Models. AIChE Journal, 17,
(6), 1971, pp. 1427-1434.
A.K. Fleming: Applications of a Computer Program to Freezing
Processes. Proceedings of the XXt Internacional
Congress of Refrigeration. Washington, D.C. 1971.
P.E. Frivik, Institutt for kj•leteknikk, NTH, 1974. Upubliserte
resultater.
P.E. Frivik et al: Sondeapparat. Rapport fra Grupper for termisk
analyse av frost i jord. Institutt for kj~letek-
nikk, Trondheim, NTH, 1975.

xviii
 G

A. Gemant: The Thermal Conductivity of Soils. Journal of

Applied Physics, 21, 1950, pp. 750-752.
X. Gotoh: Thermal Conductivity of Two-Phase Heterogeneous
Substances, Int, J. Heat Mass Transfer, 14, 1971,
pp. 645-646.
C.G. Gurr et al: Movement of Water in Soil Due to a Temperature
Gradient. Soil Science 74 (5) 1952, pp. 335-345.

Z. Hashin and S. Shtrikman: A Variational Approach to the Theory

of the Effective Magnetic Permeability of Multi-
phase Materials. J. Appl. Phys.' 33, (10), 1962,
pp. 3125-3131.
P.. Hookstra et al: Measuring the Thermal Properties of Cylindrical
Specimens by use of Sinusoldal Temperature Waves.
Cold Regions Research and Engineering Laboratory,
Technical Report 244, Oct. 1973.
A. Holmes: Principles of Physical Geology. Nelson, London
1965, p. 84.
P..C. Hooper and F.R. Lepper: Transient Heat Flow Apparatus for the
Determination of Thermal Conductivities. Heatinq,
Piping & Air Conditioning. Aug. 1950, pp. 129-134.
A..Horai: Thermal Conductivity of Rock-Forming Minerals.
,, J. Geophys. Res. 76, (5), 1971, pp. 1278-1308.
E. Hogberg: Vattenhaltens innverkan pi densitet og kompressibilitet
hos packade jordarter. Byggforskningen. Rarport
R8, 1972, p. 87.

•i, I, J

J.L. Jackoon and S.R. Coriell: Transport Coefficients of Composite

Materials, J. Appl. Phys., 39, (5), 1968,
pp. 2349-2354.
Y. Jame and D.I. Norum: Phase Composition of a Partially Frozen
Soil. Department of Agricultural Engineering,
Univ. of Saskatchewan, Saskatoon.

*II.
xix
-----------------------------------
1
0. Johansen: Unders~kelse av varmetransporten i kult og pukklag.
Det store ekuamensarbeid. Instituitt for kjoletek-
nikk, NTH, Trondheim, Host 1968.
0. JOhansen: Innflytelse av fri konveksjon i grovkornede
vegbyggings-materialer. Institutt for kjoietek-
nikk, NTH, Trondheim, 1970.
W.D. Joshua and E. De Jong: Soil Moisture Movement under Temperature
Gradients. Can. J. Soil. Scie. 53, 1973, pp. 49-57.

P.W. Kasameyer et al: Layers of High Thermal Conductivity in the

Worth Atlantic. J. Geophys. Res. 77 (17), 1972,
pp. 3162-3167.
K. Katayama et al: Thermal Properties of Wet Porous Material Near
Freezing POint. ASHRAE Journal April 1973, pp. 56-61.
M.S. Kersten: Thermal Properties of Soils, Univ. of Minnesota,
Eng. Experiment Station, bull. 28, June 1949.
W.T. Kierkus et al: Radial-Axial Transient Heat Conduction in a
Region Bounded Internally by a Circular Cylinder
of Finite Length. Dept. of Mechanical Engineering,
The University of Calgary, Calgary, Alberta.
(Preprint)
A. Knutson and R.-Saetersdal: Teleregistering 1967/68. Veglaboratoriet.
Intern rapport. Oslo 1969, p. 8.
A. Knutson: Theory and Experience Regarding Frost Penetration
and Frost Heaving, Symp. on Frost Action on Roads.
I. Paris 1973, pp. 223-233.
A.F. Knutson: Upubliserte resultater, 1973.
0. Krischer: Die Leitfahigkeit des Erdbodens. Beiheft zum Gesundheits-
Ingenieur. Reihe I, Heft 33, Manchen 1934.
0. Krischer and H. ESdorn: WArmeleitung und Dampdiffusion in feuchten
Gaern. VDI-Forach.-Heft 402, Berlin 1940.
0. Krischer and H. Esdorn. Einfaches Kurzzeitsverfahren zur gleich-
zeitigen Bestimmung der Wirmeleitnahl, der Wlrme-
kapazitht und der W~rmeeindrin'zahl fester Stoffe.
VDI-Forschungsheft, 450, 1955, pp. 28-39.

D. Kunii and J.M. Smith: Heat Transfer Characteristics of Porous

Rocks. A.I. Ch. E. Journal, 6 (1), 1969, pp. 71-78.

xx
 R.J. Kunze et al: Factors Important in the Calculation of
Hydraulic Conductivity. Soil Sci. Soc. Amer.
Proc., 32, 1968, pp. 760-765.

1
E.R. Lapwood: Convection of A Fluid in a Porous Medium.
Proc. Cambr. Soc. 44, 1948, pp. 508-521.
P.F. Low et al: Some Thermodynamic Relationship for Soill at or
Below the Freezing Point: Freezing Po nt
Depression and Heat Capacity. Water Resources
Research, 4, 1968, pp. 379-394.
A.V. Luikov: Analytical Heat Diffusion Theory. Academic Press,
New York, 1968, p. 312.

M
M.W. Makowski and K. Mochlinski: An Evaluation of Two Rapid Methods
of Assessing the Thermal Relativity of Soil.
Proceedings. Institution of Electrical Engineers,
103, A (11), 1956, pp. 443-470.
S. Masamune and J.M. Smith: Thermal Conductivity of Beds of Spherical
Particles, Ind. and Eng. Chem. Fundamentals, 2, (2).
"Mc Gaw: Thermal Conductivity of Compacted Sand/Ice Mixtures.
,' Highway Res. Rec. 215, 1968, pp. 35-47.
R. McGaw: Heat Conduction in Saturated Granular Materials,
Highway Res. Board. Spec. Rep. No. 103, 1969,
Pp. 114-131.
R.E. Meredith and C.W. Tobias: Resistance to Potential Flow through
a Cubical Array of Spheres. J. Appl. Phys. 31,
(7), pp. 1270-1273.
M.N. Miller: Bounds for Effective Electrical, Thermal and
Magnetic Properties of Heterogeneous Materials,
J. Math. Physics, 10, (11), 1969, pp. 1988-2004.
A. Missenard: Conductivitd Thermique. Editions Eyrolles, Paris
1965, p. 356.

N
Y. Nakano and J. Brown: Mathematical Modeling and Validation of the
Thermal Regimes in Tundra Soils. Arctic and Alpine
Research 4, (1), 1972, p. 19-38.

xxi
w4
 O
K. Ofuchi and D. Kunii: Heat-Transfer Characteristics of Packed
Beds with Stagnant Fluids, Int. J. Heat Mass
Transfer, 8, 1965, pp. 749-757.

E. Palm et al: On steady convection in a porous medium. Preprint

series. Matematisk Inatitutt. Universitetet i
Oslo, 1971.

F.J. Pettijohn: Sedimentary rocks. Second edition. Harper &

Brothers, New York 1957, p. 115.
J.R. Philip and D.A. de Vries: Moisture Movement in Porous Materials
under Temperature Gradients. Transactions
Amercan eophys cal Union, 38, 195# pp. 222-
232.

D. Polder and J.H. van Santer: The Effective Permeability of Mixture

of Solids. Physica, 12, (5), 1948, pp. 257-271,
S. Prager; Improved Variational Bounds on Some Bulk Properties
of a Two-Phase Random Medium, J. Chem. Phys. 50,
(10), 1969, pp. 4305-4312.

R
E.H. Ratcliffe: The Thermal Conductivities of Ocean Sediments.
J. Geophys. Res. 65 (5), 1960, pp. 1535-1541.
Lord Rayleigh (1892). Referert av Meredith (1969) (18).
J.A. Reynolds and J.M. Hough: Formulae for Dielectric Constant of
Mixtures. Proc. Phys. Soc. (London) B70, 1957,
pp. 769-775. (Referert H. Fricke (1924).
R.L. Rollins et al: Movement of Water in Soil due to a Temperature
Gradient. Highway Research Board Proc. 33, 1954,
pp. 492-508.
C.W. Rose: Water Transport in Soil with a Daily Temperature
Wave. I. Theory and Experiment. Aust. J. Soil
Res., G. 1968, pp. 31-44.

S
E. Saare and C.G. Wenner: V~rmeledningstal hos olida jordarter.
Statens n&mnd f6r byggnadsforskning. Handlingar
no. 31, 1957.

xxii
14 J.H. Sass et al: Thermal Conductivity of Rocks from Measurements
on Fragments. J. Geophys. Res. 76. (4), pp.
3391-3401.

A.E. Scheidegger: The Physics of Flow through Porous Media. Univ.

of Toronto Press, 1957.
E. Schmidt: Einfirung in die Technische Thermodynamik. Springer
Verlag, Berlin 1963, p. 394.
R. Selmer-Olsen: Om norske jordarters variasJon i korngradering og
plastisitet. Norges Geologiske Undersokelse no.
186, 1954, p. 40 and fig. 30.
R. Selmer-Olsen: En regional unders0kelse av norske kvartaere
leirers finfraksjon basert pa DTA. Rapport sendt
til NTNF. Geologisk Institutt, NTH, Trondheim
1961, p. 29 f.f..
R. Selmer-Olsen: IngeniOrgeologi. Del 1. Generell geologi. Tapir
forlag. Trondheim 1971, pp. 8, 12 and 175.
R. Selmer-Olsent IngeniOrgeologi. Del II. L~smasser. Kompendium,
Geologisk Institutt. NTH, 1973.
W.A. Sinclair et al: Soil Thermal Resistivity. Typical Field Values
and Calculating Formulas, part IV. In: Soil
Thermal Characteristics in Relation to Underground
Power Cables. AlEE Transaction paper No. 60-785,
pp. 71-94.
A.G. Shashov et al: Thermophysical Properties of Thermally Insulating
Materials in the Cryogenic Temperature Region. Int.
J. Heat Mass. 15, 1972, pp. 2385-2390.
I. Szanto and J. Aquirre Puente: Etudes des caracteristiques termiques
des milieux poreux fins humides. Proc. XIIIth
Id. Intern Congr. of Refrigeration. Washington, D.C.
1k, 1971, Vol. I, pp. 751-757.
Sv. Skaven-Haug: Frostfundamenters dimensJonering. Frysevarme og
ajordvarme. Frost i jord, (3) 1971, pp. 9-27.
"Sv. Skaven-Haug: The Design of Frost Foundations. Norwegian Geo-
tecnnical Institute. Publ. 90,.Oslo 1971.
Sv. Skaven-Haug: Romforhold i jordmaterialsr. Meddelelser fra det
norske myrselskap. 70, (4) 1972, pp. 89-102.
Y., .W.O. Smith and H.G. Byers: The Thermal Conductivity of Dry Soils
of Certain of the Great Soil Group. Am. Soc.
Soil Sci. Proc. 3, 1938, pp. 13-19.
WO. Smith: Thermal Conductivities in Moist Soils. Soil Sci.
Soc. Am. Proc. 4, 1939, pp. 32-40.

Xx~ii
 (
W.O. Smith: The Thermal Conductivity of Dry Soil. Soil
Science, 53, 1942, pp. 435-459.
H. Sveian: Norske jordarters kvartsinnhold. Hovedoppgave i
Ingeni~rgeologi, Geologisk Institutt, NTH.
H~st 1972.
H. Sveian: Under~kelse av kvartsinnhold i losmasser ved hJelp
av DTA. Utf~rt for Institutt for kj~leteknikk,
Trondheim, 1973.
1. Szanto and J. Auirre Puente: Etude des caract~ristiques des
miyieux poreux fins humides.

K. Terzaghi and R.B. Peck: Soil Mechanics. John Wiley & Sons,
New York, 1948, p. 32.
J. V. Thue: Red 7sert fundamentdybde, - termiske problemer. Del I.
".nstitutt for husbyggingsteknikk. Trondheim, NTH,
p. 82.
A.R. Tics et al: The prediction of Unfrozen Water Contents in
Frozen Soils from Liquid Limit Determinations.
Symposium on Frost Action on Roads. 1. OECD.
Paris 1973, pp. 329-344.
G.T. Tsao: Thermal Conductivity of Two-Phase Materials. Ind.
Eng. Chem. 53, (5), 1961, pp. 395-397.

U,V,W
E.M.F. van der Held and F.G. van Drunen: A Method of Measuring
the Thermal Conductivity. Physica 15, (10), 1949,
pp. 865-881.
M. van Rooyen.and H.F. Winterkorn: Structural and Textural Influences
on Thermal Conductivity of Soils. Highway Research
Board. Proceedings. Washington, D.C. 1959, pp.
576-621.
L. van Zee and C.L. Babcock: A Method for the Measurement of Thermal
Diffusivity of Molten Glass. J. Am. Ceram. Soc.,
34 (8) 1951.
Vegnormaler. Statens vegvesen. Vegbygging. Overbygning med
bituminest dekke. 1970. Kapittel VI. Avan...3.

xxiv
K. Wakao and K. Kato: Effective Thermal Conductivity of Packed
Beds. Journal of Chemical Engineering of Japan.
2, (1), 1969, pp. 24-33.
N. Wakao et al: View Factor between two Hemispheres in Contact
and Radiation Heat-Transfer Coefficient in Packed
Beds. Int. J. Heat Mass Transfer, 12, 1969, pp.
118-120.

N. Wakao: Effect of Radiating Gas on Effective Thermal Conductivity

of Packed Beds. Chemical Engineering Science, 28,
1973, pp. 1117-1118.
G.K.H. Walther: Zur matematischen Rehandlung der Stationaren Warme-
leitung in stuckweise homogen Medien. Kernforschungs-
anlage Julich. 657-MA, 1970.
JE. Warren and H.S. Price: Flow in Heterogeneous Porous Media,
Society of Petroleum Engineers Journal, Sept. 1961,
pp. 153-169.
A. Watzinger et al: Unders~kelse av masseutskiftningsmaterialer
for vegog jernbanebygging. Medd. fra Vegdirektoren
1938, no. 6.
A. Watzinger et al: Unders~kelse av masseutskiftningsmaterialer for
vegog jernbanebygging. Medd. fra Vegdirektoren
1941, no. 6,7 and 8.
A. Watzinger st al: Masseutskiftningsmaterialer for teleforbygning
pa veg og jernbane. Norges Geotekniske Institutt.
Oslo 1965, p. 106.
0. Wiener (1912): Referert av W.F. Brown (1965).
P.J. Williams: Unfrozen Water Content of F oz.n ~i a soil
Moisture Suction. Geotechnfque, pp.
231-246.
W. Woodside: Probe for Thermal Conductivity Measurement of Dry
and Moist Materials. Heating, Piping & Air Condi-
tioning, Sept. 1958, pp. 163-170.
W. Woodside and J.B. Cliffs: Heat and Moisture Transfer in Closed
Systems of Two Granular Materials. Soils Science,
87, 1959, pp. 166-173.
W. Woodside and C.M. A de Bruyn: Heat Transfer in a Moist Clay.
Soil Science, 87, 1959, pp. 166-173.
W. Woodside and J.H. Messmar: Thermal Conductivity of Porous
Media. I. Unconsolidated Sand. J. Appl. Phys.,
32, (9) 1961, pp. 1688-1699.

1x
, i
 x'Y, z

S. Yagi and D. Kunii: Studies on Effective Thermal Conductivities

in Packed Beds. A.I. Ch. E. Journal, 3, (3),
1957, pp. 373-381.

P. Zehner and E.U. Schlunder: Einfluss der Warmestrahlung und

des Druckes auf den Wirmetransport in nicht durch-
gestr~mten Sch(ttungen. Chemie-Ing.-Techn. 44
(23), 1972, pp. 1303-1308.

xxvi
 INTRODUCTION

Knowledge of heat related parameters for minerals is necessary

in connection with any analysis.of temperature distributions in
the ground. This is particularly important when analysing struc-
tures which are subject to frost or have to be protected against
thawing (structures in permafrost). Artificial freezing of soil
or crushed rock in connection with tunneling and control of tem-
perature distribution around buried LNG storage tanks are other
areas where thermal analysis is applied and for which knowledge
of thermal properties of minerals is crucial.

When such analysis is to be performed, there is often information

available about the kinds of minerals involved but only seldom are
thermal parameters known for the entire structure. Thus, in most
cases one has to determine these parameters from the more or less
complete information available about the minerals, In short, this
is the starting point for this study and the emphaslis is placed on
finding methods for determining thermal properties of the minerals.

Methods which have been mostly used previously to treat this pro-
blem have been based on empirical correlation between a limited
V set of mineral parameters and results from systematic heat transfer
measurements with selected minerals. This has given results with
relatively wide tolerance limits. These methods mpsy have been
satisfactory for approximate determination of frost penetration.
' However, while using modern computer programs for finding tempera-
ture distribution in the ground, researchers have felt a need for
improved accuracy of thermal parameters.

A possible approach for narrowing the tolerance limits would be to

utilize more mineral parameters when correlating with experimental
results. This would certainly require a very comprehensive expe-
rimental data base with wide variation span in types of minerals
and choice of parameters. Before such an experimental program is

a -
4 undertaken, one has to clarify the most fundamental relations
between mineral parameters and thermal properties, in order to
arrive at the most conclusive test program possible.

In the study at hand it has been found that it is more useful to

emphasize clarification of basic relations through theoretical
studies of, among other things, heat transport processes in moist
materials and heat conductivity of composite minerals. This ana-
lysis has resulted in a method for reduction of experimental data
which has made it possible to utilize already existing test data
for developing a complete method for determining thermal proper-
ties of soil materials. At the same time, these methods will allow
more efficient planning of future test programs.

Still, the extensive measurements which are required to supplement

existing results and verify their validity for Norwegian conditions
will place great demands on our capacity for testing. To meet these
requirements, it has been necessary to develop new and more effi-
cient methods of measuring thermal properties of soils. However,
heat conduction measurements in moist materials pose very special
requirements on test methods, due to moisture transport in the
presence of temperature gradients. Thus, one must try to find me-
thods which combine the requirements on short test periods and small
temperature gradients. As part of this effort, several types of
test gear for measuring conduction have been developed and built.
These may be a valuable supplement to more conventional, statio-
nary test equipment.

To arrive at a method for determining thermal conduction properties

in minerals, which can be used at several levels, it was necessary
to clarify basic relations between conventional data for soil ma-
terials and the more fundamental parameters which determine ther-
mal conduction properties. Through these studies, an effort has
been made to arrive at relations which can be utilized for indirect
determination of parameters which in many cases are left out during

2
4/
geotechnical analysis. A mathematical model developed along these
lines will consequently give increased accurary in proportion to
the information available on the minerals involved. NThen using
mathematical models it is important to know the relation between
accuracy and level of information. Part of this report is there-
fore devoted to a study of this topic. The effects of uncertain-
ties in thermal conductivity values on thermal computations is also
discussed,

These introductory remarks are intended to give an overview of the

approach taken for the study at hand. It will be evident that this
study has an interdiciplinary character, in that areas which tradi-
tionally fall outside the field of thermal technology have been
treated. For experts in those areas, the material may appear ele-
mentary and in part superficial. However, the main intent is to
include all major aspects in developing the mathematical model.

CHAPTER I

HEAT TRANSPORT IN MOIST SOIL.

For computing temperature distribution in and around structures

placed on the ground one normally uses ordinary differential equa-
tions for heat conduction, together with boundary conditions (39).
Heat transport in moist and porous materials can, however, be in-
fluenced by a number of different mechanisms, in addition to heat
conduction. If these phenomena can be incorporated in an equiva-
lent heat conductivity for the porous material, one can without
significant difficulty use ruwerical methods designed to handle
nonlinear problems involving towperature dependent thermal para-
meters. Problems arise only when the heat transport is controlled
by potentials other than the temperature gradient, so that the mass
distribution becomes a function of time.

3
S.
. . . . . . . . ... . ... . . . . . . . . .. . . .
 •5

" M. 11 L ,r .• . r

i rur
%a ý- jw If
3, 4 t nlg

The different mechanisms which can contribute to heat transport

in a moist sol material, are illustrated in Figure
C . Thtse

mechanisms will be discussed in the following, in order to clari-

fy some of the limitations inherent in a mathematical model on
heat conduction only.

1. THERMAL RADIATION
The share of thermal radiation in the effective heat conductivity
increases with increased pore area and increasing temperature.
This can be illustrated by the simple case of radiation across a
gap between two parallel.i surfaces at temperaturei T1 and T2 .
If the emissivity of both surfaces is c, the transferred heat flucv

can be expressed as
q12 " a (T 4 _ T24) (W/m2 ) 1

where c is the radiation constant for a black surface.

1)
(a a 5.77 4 10-8 W/m2 K4 ) . The resulting contribution r to
the effective heat conductivity across the slot is given by
hr " q 1 2 d/(Tl - T2 ) (W/mK) 2
where 6 is the width of the gap.
IT iKhere stands for degrees Melvin (Translator's Note).
4
If T1 - T 2 is small (which is aften the case for narrow gaps)
Eqs 1 and 2 give
X
r 2/c 4a- I T3 6 (W/mK) 3

One can also define a heat conduction coefficient for radiation

across a gap

r - 4o T3 (W/m2 K) 4

Expressions 3 and 4 cannot readily be applied to the complicated

geometry of e.g. a particle fill. Still, similar expressions have
been used as a basis for several studies of radiation effects in
particle fills. (1,2,3). Wakao and Kato (4,5) have instead used
the radiation transfer between two half spheres touching each
other, where each sphere is assumed to have a constant surface
temperature. This led to a heat transfer coefficient based on the
areas of the spheres
"•r 4a T 3 (W/m2 K)
- 0.284-
S2'I7

Wakao and Kato used this heat transfer coefficient for numerical
analysis of regular lattices of identical spheres (4,5,6). Two
packing configurations were studied, cubic lattice with n equi-
valent porosity n = 0.476 and orthorhombic lattice with an equivalent
porosity n a 0.395. Figure 2 shows the two packing configurations
1)
and the assumed heat flow is indicated by arrows.

1) There seem to be no "arrows" in Figure 2, but that is what

the text says. (Translator's note)

5
, •: - •-• .- • . ..- ... ,.. . , • • """"";w" - • . •---•• :-•::: ""• •:•" . 4•
 A'JBISK PAKY4NG n.Qovs OROPI5 noons
CAOiC ,ATT1CE

II
The results of their computations were presented in the form of
curves with a dimensionless N4usselt number for radiation N4ur as
a parameters
Nur " OrI"JIWO11V
dp/Xsof Iso CA.",i "m ".PkW
i;Ya 6
where ais defined by Eq 5, d~ is the particle diameter and s'
is the thermal c.onductivity of each particle. Figure 3 shows the
result obtained for orthorhombic configuration (n =0.395).

r 1.00i
-
'4

PIG, 3 Inntlytalsen &Y it.-Slime *4 1,~drjinuiYmnn Aw

Comparable results have been presented by Zehner and Schiunder (3).

Although their calculations were based on a different model, the
results do not differ significantly from those quoted here.
Even if soil conditions differ significantly from the comparatively
simple geometrical models on which Wakao's results are based,
in terms of granule shape and size as well as packing configuration,
these results can-give a hint as to the magnitude of the effects
of radiation. In addition, one may assume that the effect of
radiation is less when the particle sizes vary widely, as compared
to spheres of the same diameter, even if the average size is the
same in both cases. The following calculations can thus be
assumed to yield an upper bound for the effects of thermal radiation.

Wakao and Kato (5) have used the results shown in Figure 3 to
derive an empirical equation of the form
Xe - ÷ 1.3
+O Or ' dp (W/mK) 7
where Ae is the effective thermal conductivity of the material
when no radiation takes place. This equation was derived on the
basis of both packing configurations and is said to be approxi-
mately valid also outside the considered range of porosity
(n - 0.395 to 0.476).

Eq 7 and the results shown for Nur - 0 in Figure 3 were used for
calculating the following examples. Since the study of dry soil
is of interest, thermal properties for gas and solid material are
assumed to be those of air and granite, respectively. See Table 1.

Table 1. Conductivity for air and granite at different temperatures

W/mK.
0C o
Temperature,
Air (7) .39 3Z'
2..
Granite (8) -

7
S__,
, • | • .. . .*- ...... ,i "= . . ; ' ' • "• , " _ • T ••
It should be noted that the conductivity for air increases with
temperature, while the opposite occurs for granite. The same is
true for most minerals, except for some alkaline materials such
as gabbro, for which the conductivity is nearly constant or

decreases slightly with temperature in this range (8).

Figure 4 shows the results for different particle size, as

function of temperature. One finds that even for a particle diameter
of 20 mm, the effect of radiation amounts to no more than 10
percent at normal atmospheric temperatures, while in sand
(particle diameters below 2 mm) the effect is less than I percent
under the same conditions. In these calculations, the porosity
was assumed to be n a 0,395 (orthorhombic configuration), which
corresponds to a dry density for sand of about 1600 kg/m 3 . For
higher densities, the effect of radiation will of course be even
less and in moist mauarials it will be completely insignificant.

In short, this indicates that radiation will be responsible for a

negligible portion of the heat transport on moist soils, while
it can play some part in coarse, crushed stone materials.

20 -

0 o

50 0 so 110 ISO 20C

TFEMIP9ATUR, °C

Pro. i, !ni.ol v •r:n 6•e~nn.,ve

RIO3I~
N 11nI M1 7 ri• p.l , M, L

A@ C',8
 qv .T
Da 7p (kg/m 2 s)
S1)

R is the water vapor gas constant (461 J/kgK)

T is absolute temperature (K)
p is water vapor partial pressure (bar, 105 *1/m2 )
Da is water vapor diffusion coefficient in air (m2 /s)

Krisher and Rohnalter (9) have determined this diffusion co-

efficient to be
Da 2 T 2.3 2/S)
P0 273
whare Po is the total pressure (bar).

The water vapor pressure in a soil material will depend on both

temperature and relative humidity of the air in the pores, The
relative humidity also depends on how firmly bound the water is.
The binding energy can be described by a moisture potential
and the relative humidity I is given by

SF, (!/ 2) 9
3
0 is the density of water (kg/m )
R , gas constant of water vapor (461 J/kgK)
T , absolute temperature (K)

The dependence of 0 on water content depends on the soil taterial.

Figure 5 shows moisture potential curves for a number of selected
soil materials. The scale for relative humidity is derived from
Fq. 9 and superimposed on the scale for moisture potential.')

1) In the original, p is shown as a subscript to the gradient

symbol V . The same occurs in Eq 10, but the more conventional
notation has been chosen for the translation,
2) "Moisture pote.itial" could also be termed "suction".
(Translator's notes)

9
It is evident that the relative humidity falls off significantly
only for comparativ.?ly low water content, even for finer grain
soils.

If one assumes a certain relative humidity ý in the pore air, the

vapor pressure gradient can be expressed as a total differential
which is a function of the gradients for both humidity and tempe-
rature

aT

where e is the water content in volume ratio.

The first term1 ) in the right hand member of Eq 10 represents the

pressure gradient due to moisture gradients, while the second
term expresses the vapor pressure gradient due to temperature
variations.

As previously mantioned, Eq 8 (Fick's law) is valid for diffusion

of water in air. In soil materials, the diffusion rate will be re-
duced due to presence of water and particles. Experiments with
porous materials show that the diffusion rates are strongly reduced
as the porosity decreases in dry materials. Figure 6 shows results
from measurements by Currie on diffusion of hydrogen in dry and
moist soils (13, 14),

However, Philip and de Vries (15) have pointed out that water va-
por and gas diffusion in moist materials are different, since
water and water vapor interact.

1) Double prime on p stands for saturation pressure. See

notations. (Translator's note).

10
 2. COMBINED MOISTURE AND HEAT TRANSPORT
A. Water vapor diffusion

100N/Mr

JLE
-- LOAM 4 sO

- SILT, SILT " 900

100.

I-

1 i \

0 ,

0 2i
20 20 */ 40
VANNINNHOLD. VYKT
WATIERCONtN!, W?~iOmT

•. P:O, .•Fl Ork~pt,,/.

f*01ea.iv ru•l~rhot i orskJot•,ge

I'P

Partial pressure variations for water vapor in the pores of a

soil material will cause diffusion in the direction of falling
pressure. Water vapor diffusion in air is determined by Fick's law

1±J
 08

0.4[

POROSITY

A. )iffusjon j~r mae'iaer :a ord*z'o~tl

D~ a 00fftitfln'
Q8 n, if~ ~f ;Q~d'od?'0,4¶nar

'01

02

8. Dirfusjon I rukti±l lord. , .n~tl- iiru ror

nydragiv 2.o1vailendt '.rrA -ulterl-a~r.no
,otfvn of diffusiLon ýn:ý-erodm r

Diffweton of hiydrojen in adbod00 1,

 They claim that this permits water transport via water lenses in
the moist soil material, of the tame magnitude as that due to
vapor diffusion. The temperature dependent diffusion is also
governed by the temperature gradients in the air within the pores,
which is always greater than the average gradient. Together,
these effects result in a larger diffusion in moist soils than
K that obtained be considering diffusion rates for gases only.

Based on the previous discussion one can define two diffusion co-
efficients for vapor diffusion due to moisture and temperature
variations, D and D, respectively
qV/0Dv D v
* - ye - DT 7 (m3 /m 2 s) 11
"V V
where both DO W/o) and DT (m2 /sK) contain correction factors
V V
which account for the previously mentioned increase in diffusion
,1 rate, relative to the case of gas diffusion in comparable materials.

The moisture dependent diffusion rate for different water content

is very dependent on the shape of the water potential curve for
the soil material (Figure 5), i.e. of (a/•), c.f. Eq 10. This
means that the contribution due to moisture dependent diffusion
will be quite small except for the lowest values of water content
where the relative humidity falls significantly below * u 1.0.

On the other hand, the temperature dependent diffusion coefficient

DT will decrease when the relative humidity in the soil starts to
V
fall off due to low water content. It will also decrease when the
water content increases, since the pores are then to a large ex-
tent filled with water. The maximum for Dev 1) will occur for
relatively low water content, when there is still a low degree of
continuity in the pore water 0 (c.f. ref. 13)

1) From the context it appears that this is a misprint in the

original. It should probably read DTv. (Translator's note)
TV

13
 B. Capillary water transport.

If a soil material, where the humidity is equally distributed at

the outset, is subjected to a temperature gradient, the previously
mentioned vapor diffusion will give rise to a transport of mois-
ture in the direction of falling temperature. In a closed system,
this will result in a drying out of the warm side and an increase
in humidity on the cold side, as has been demonstrated by numerous
experiments (16,17, 17, 19, 20, 21). This will result in moisture
gradients, which form a potential for capillary water traneport
in a direction opposite to that of the vapor diffusion caused by
temperature. This water transport depends both on the moisture
potential of the soil and on its permeability, i.e. its hydraulic
conductivity K (m)I), for different moisture content. Darcy's
law for water transport in moist materials can be expressed as

ql/Vw -Vý (m/a) 12

K is hydraulic conductivity (m/s)1)

4 is the total potential, expressed in metres) which i1
common in this context.

If the effect of gravity is neglected, the potential gradient is

given by the moisture potential T' (here in metres) and the mois-
ture gradient '7e

Se (M/m) 13

Eqs 12 and 13 can be used to define a moisture diffusivity

De u K (m2 /s)

Darcy's law (Eq 12) then takes the form

ql/Ow " D. '7 1

1) The original gives two units for K, m (sength)and m/s (veloci-

ty). (Translator's note)
14
 The change in this moisture diffusivity due to varying water content
is highly dependent on the hydraulic conductivity K. Figure 7
shows its variation for some fine grain soils. It can be seen
that the diffusivity decreases sharply for low water contents,
where the continuity of the pore water is broken. It is also evi-
dent that this decrease occurs for higherwater content in the finer
grain materials.

l~~~GM ,evea &V M oen ýOrdartvr

A. Medium a&nd
0 10 20 Lo io I. t•LP| silt
.,:, YAN. INNHO4LO,
V61..*/ S• it |&l,111i1

* Figure 8 shows a partially qualitative summary of how the pro-

viously defined diffusivities change as function of water content.
The broken (dotted) lines indicate how the three diffusivities tend
to fall for a finer grain material. This partially qualitative
presentation shown a region for relatively low water contents
where the thermal vapor diffusivity dominates over the two mois- .4
ture dependent diffusivities. As indicated, this region will move

1) The original uses the word "skjematisk", translated by "schema-

tically" in Fig 8, which indicates an intention to show a
qualitative trend rather than to present quantitative num-
bers. (Translatorls note)
 ci i
k,V4-
I& li
AIU
, -

'3 2 0
3C 41: so
VAN'h44'O.O, ',4 *1.

PIG. 8, Porlg. at av do f u1rt '.l. ftuKItd.iiUsiY1oQt ,

dranw, Joh.'m.i ,teo k.

In experiments where moist soil materials are subjected to tempe-

ratues gradients, it has also been demonstrated that remarkably
large moisture gradients occur for comparatively low water con-
tents As an example, Figure 9 shows the results obtained in Smith's
classic experiments in 1939 (22). Somewhat simplified, Smith cal-
culated the moisture gradient as the difference in water content
between the cold and warm sides of the test container (which was
3/411 thick). The temperature gradients are in the order of 4c'/cm,
but vary some-what from experiment to experiment. The settlingI)
time is not given

j" J, "AImr'•on #;Adyte "

C..n. , s loan,

- "BP! I6 I, R DII'III

.AININ424LW. '?,1' -040

PUlgrl i en~l'l
to l over
l or "ierUt o

i nL ' Ii....

tiflo rllU rafdlenter ;I

at T1 4 J
/",l Ui, tml 1939).
I.to_ iinl Mai'. so ./0.pee
f from &.../
on.

1) "Oppholdstid" means "duration" i.e. the time during which

the moisture gradient were- slowed to build up. (Transla-
tor's note)
Evidently, also in this case the previously mentioned tendency

is present for translation of the critical region towards higher

water content for finer grain soil materials.

This effect is obviously very important in connection with experi-

ments designed to determine thermal conductivity of soils. This
will be discussed in more detail in the chapter on heat conduction
measurements (chapter IV).

C. Contribution of vapor diffusion to

thermal conductivity.

"The mentioned wapor diffusion will contribute to the heat trans-

port in the form of latent heat exchanged during vaporization
and condensation. The temperature dependent vapor diffusion will
be controlled by the same temperature field as the thermal con-
duction in the pore air. It is thus possible to include this con-
tribution to the heat transport in the apparent thermal conduc-
tivity. The vapor pressure gradient due to a temperature gradient
was given by Eq 10:

apff
7 tV 16
According to Pick's law, the amount of diffused vapor is given by
*i.Da 17"

q a "a.~ (kg/m2.) 17

while the amount of transported heat is

q a rqv (W/ma) 18'

where r is the latent heat due to vaporization of water, equiva-

lent to 2450 KJ/kg at 204C/
The contributior. to the air conductivity is

17
 = rDa "p"t

RT ýT

while the apparent air conductivity x' is given by

a Xv 20

where X is the conductivity of air.

Figure 10 shows how X varies as a function of temperature

0 a 1.0), as compared to the conductivities for air and water
at the same temperatures (23).

The rapid rise with increasing temperature is clearly due to the

fact that the water vapor pressure increases almost exponentially
with temperature.

T sa.nw~'UM11".nn
l . Med,'••
1" {01'i 'I i .. x v.1
'm •) igjIInl..nN"
ing:,soynen ni'L
4V ý-Url fD)
W/m k 10 o, vwf•i 'A). Appae

dwe90 VSPOuP diruianep (0)

6," at it) and ýjvi (,).

0 10 20 30
A

01'
o 20 1.0 60 4
?"WPWRA!UR

This phenomenon is a major contribution to the increase in con-

ductivity with increasing temperature for moist soil materials.
This is evident from measurements of thermal conductivity
performed by de Vries on sand at various temperatures. His re-
sults are reproduced in Figure 11. For both extremely high and
low saturation levels, where the thermal vapor diffusion is ii-

18
 mited, the dependence on temperature is wea, while this depen-
dence is strong for low to moderate degrees of saturation, where
the thermal diffusion is significant.

0*0 0.2 0., 01 01 " ,

YErNINGS0IRAO S.

• •.. Varfmoema~ncheye &V fvxti SAM. Vln

l rGrSx•Jeq.11ll

I .I u t I vUitV #.0 vo iasl 4a d 4 ; d &,f,',#rjd t c w•;je I'e IU r e

The moisture dependent vapor diffusion will also contribute to

the heat transport when latent heat is transferred. Phillip
and do Vries (15) have combined the contributions of vapor dif-
* fusion to heat transport in a differential equation for thermal
conductivity in soil

Cx ,T) •ee)
7 +rV7, 21

t C is the specific heat, by volume, of the soil material

3 K).
(J/M

x"e is the thermal conductivity of the soil material, in-

eluding the contribution from vapor diffusion (';/mK).

This equation differs from the conventional differential equa-

tion for thermal conduction, the Fourier equation

a2T
a•• 22.

a thermal diffusivity (ma /s)

_

19
 c is specific heat (J/kgK)

0 is density (kg/m 3 )

First of all, Eq 21 assumes a temperature dependent thermal con-

ductivity, while the Fourier equation usually presumes constant
thermal properties. In addition, Eq 21 contains a term which is
related to humidity. This creates obvious problems in connec-
tion with thermal conductivity measurements in moist soil mate-
rials, when transient methods are used and the result interpre-
tation is based on the simple Fourier equation. These problems
will be discussed in the chapter on heat conductivity measure-
ments (Chapter IV). In this context it should be pointed out
that the problem is simplified somewhat by the fact that the
last term in Eq 21 usually is negligible, except for extremely
low water content when the relative humidity in the soil starts
to decrease. As mentioned previously the water content will
also increase for more fine grain materials.

3. CONVECTION

A. Heat transport

Heat transport due to convection in porous materials can be of

two types, forced convection and free convection. The first
type is due to forced currents in the liquid which fills the
pore., while the second type is caused by temperature gradients
and upward migration of the liquid. Forced convection in soils
and fill materials can, for example, be associated with ground
water flow or wind forces on exposed fill much as railroad banks.
Free convection can occur due to upward heat flow in rock fill.

The total heat transport due to thermal flux in porous materials

can be ascribed to several phenomena (mechanisms). Except for
heat transport via currents in the liquid, heat conduction in
solid particles and in the liquid are the most significant. In

20
addition, convection between particles and liquid ma:, contribute,
as well as heat transport due to "turbulent" mixing of liquids
with different temperatures within the "pore strata". The lat-
ter effect is often termed "turbulent diftussion" or dispersion
and can be included as a velocity dependent constribution to the
thermal conductivity of the liquid (25).

For low velocities of flow, the temperature differences between

liquid and particles will be small and the convection between
these media can be neglected. The total heat transport can then
be described by the equation

, aT + (co) v VT tm 23

XS is the effective conductivity for the porous material,

including the contribution from dispersion (W/nmK)
(cp)f, (is) thermal capacity of the liquid (J/m 3 K)

(CO)M, (is) thermal capacity of the liquid-filled porous

material

Eq 23 is used for determining the equivalent (effective) thermal

conductivity in experiments with flowing liquids or gases, e.g.
in columns filled with spherical particles (25, 26, 27). Such
experiments have been made with the axial thermal flow both in
the same and the opposite direction of that of the liquid. In
both cases one has found a significant increase in the equivalent
conductivity X when the flow rate increases. This increase is
comparable to the previously mentioned contribution from turbulent
diffusion, see Fig 12

It appears as if no experiments have been performed with vertical

1) This sentence was almost unreadable in the original (page 17a

bottom), so this is a "guesstimate". (Translator's note)

21
 heat flow and planar horizontal liquid flow in porous materials,
similar to horizontal ground water flow in a soil layer. There
is still reason to assume that the previously mentioned increase
in the effective vertical conductivity also should occur in such
cases, provided that the flow rate is sufficiently high.

Yagi amd Kunii have formulated a partially empirical expression

for the effect of axial liquii flow on the radial thermal con-
ductivity in particle filled columns. They define the effective
conductivity as a sum

•e1 "e X * X 24t

2

where X0 is the conductivity for no liquid flow

X is the contribution due to "lateral mixing"

A
Goo

* 601

/- /ASCW'G RINGER'
* -• SANOSTEIN

t 0002STIROMId•NOSMASTIGMIT
000 .00rollk_0.00 2.0 IWO

A. aiZ A• a .a, f j f•wrij I, .9jdi

Fla. 0 ., :nntl-yto'so av st.re',nde vxv,•t .*ý,k-.v

h , t,.ow (21),

BESTLAVAILABLE COPY'
22
 The contribution x. is given by

xt PP( 25
Xf

a and 3 are dimension-less factors

Pe, modified Pecket's number

Pe m- d (co)±f v/xf

d is the average particle diameter (m)

p 3ý )
(O)f is the specific heat, by volume, of the liquird (J/
v is liquid flow rate in open cross-section (m/s)
x is thermal conductivity of the liquid (,V/mK)
if

Values for (a$ ) are determined empirically for various particle

configurations. For spheres and cylinders, I ) has been found
0.08 to 0.14 (26).
to be in the range

As shown by Eq 24, the contribution due tu liquid flow will in-

crease in proportion to average particle size and liquid flow rate.

B. Liquid transport in porous materials

For low flow rates, the liquid transport ii porous materials is

determined by Darey's law (27)

v - P 26

v is velocity (flow rate)1 in "open cross-section"

k is the permeability
n is the dynamic viscosity of the liquid

1) The original does sometimes not clearly define whether

"liquid transport" refers to flow rate (m3/s or kz/s) or velo-
city (m/s). I sincerely hope the distinction is evident to the
reader (Translator's note)

23
*@ 0 is the density of the liquid
P is the pressuire in the liquid
g is the gravitation (acceleration) constant

The permeability k depends on the particle size, distribution

and shape, in addition to the porosity of the material. The
frequently used Kozeny-Carman equation for permeability contains
porosity n, as well as specific surface area per unit volume, S:

k 3.
5S2(1-n)e (m ) 27

The specific surface area for particles with diameter d is

given by

S = 6/dr (ma/m 3
) a8

This results in the following expression for the permeability

of fills consisting of identical spheres:

k ....0 (m 2)
(i0()-n)
29

For a mixture of spheres of different diameters one can use an

equivalent diameter dp * From calculations of specific surface
area one finds that this average or equivalent diameter should
be the harmonic mean of the different diameters

d - j. m-J •i 30

This means that the equivalent diameter will be close to the

smallest diameter in the particle distribution. In empirical
formulas for the permeability of sand one can thus use d 1 0 or
d30(i.e. the diameter which 90 or 70 percent of the particles
exceed)1) as the equivalent diameter.

1) Literal translation:"the diameter that 10 or 30 percent of the

particles are less than". (Translator's note)

24 "
As an example, Hazen's formula for water permeability of loosely
packed sand consisting of particles of equal sizeI) reads (28)
2
K = 116 d (0.7 + 0.03t) (cm/s) 31

d 1 0 is the particle size in cm for 10 percent penetrationI)

t is tne temperature C'c)

K is defined from Darcy's law, with potentials defined in terms
of pressure height and static height:

v =
* 32

where i is the potential gradient in m/m (or cm/cm). Calculations

show that K, given by Hazen's formula, corresponds to the earlier
defined parameter k, according to the Kozeny-Carman equation, for
a porosity near n 2 0.45. This corresponds to a dry density for
sand of about 1,500 kg/m3 (y3 = 2,700 kg/m3).

Figure 13 shows the permeability It for different porosities, cal-

culated from the Kozeny-Carman equation, along with correspon-
ding values of k from Hazen's formula. The range of d1 o for com-
mon soil materials is also shown.

For higher velocities of flow, the velocity is no longer propor-

tional to the pressure gradient in the manner expressed by Drazy's
law. As an example, Ergun (29) gives the following empirical re-
lation between pressure gradient and flow velocity in particle fills
p/ 3150 v -n
aP
1 - 1 0n- ," + 1"75 3-3•
nL d 1-n 33
For low flow velocities, the second term in the right hand member
of Eq 33 will be much smaller that the first. The expression
will then correspond to the Kozeny-Carman equation. For high

1) See appendix to translation (Translator's note)

25
 velocities, the first term wvill be negligible in
comparison to
the second. The equation will then approach the form suggested
by Burke and Plummer (30), where the flow velocity is proportio-
nal to the square root of both the pressure gradient and the par-
ticle size.

d-

),,PTO. 13. PormestL•itat av jordarlef bitOfe"e tir•~ n

J C. Forced Convection

Based on the expression for permeability or Eq 33 one can esti-

mate the magnitude of the velocity in ground water currents in
soil materials. The potential gradient can for these cases be
expressed as a "terrain slope" h/l (m/m). In Figure 14, flow
velocities have been calculated for varying diameter (d andnd
three different potential gradients h/l.. These calculations were
4.. based on Darcy's law and Hazen's formula, with corrections due to
the second order term in Eq 33 for large flow velocities, Theme
correctionw give rise to the typical "knees" in the curves for
high velocities.

In the case of flowing ground water, the liquid flow will usually
be nearly perpendicular to the direction of the heat flow. The

266I
 contribution to heat flow'by the liquli can thor. be ascribed
principally to the previously mentioned dispersion. If one
assumes that the case of vertical heat flow in the presence of
(horizontal) planar liquid flow is analogous to the case of
radial heat flow in a column with axial flow of liquid, the con-
tribution due to dispersion -an be derived from the empirical
relations of Yagi and Kunil , established for the latter case,
see Eqs 24 and 25.

* Figure 15 shows results from calculations based on this assump-

tion and (liquid) flow velocities corresponding to a terrain
slope h/l a 1/10 (see figure 13).

For sandy materials one finds that the effect of ground water
flow is slight except in the case of very coarse sands, where
the dispersion effects may increase the conductivity by up to
20 percent. For still coarser materials, such as gravel or rock,
this effect may be of an entirely different magnitude. However,
ground water flow seldom occurs in coarse gravel, due to its ex-
cellent drainage properties.

m I

.I o b Pi

;I 00,0 saila

io.

1*l,"

PlO. ill, lI~aatIlr~.t.P I gruntivannlttvvom.o 1 rOrkJIkL±v.

4ONPr+imr I1• VIPl~rqrJ. •IPm t~ •p
?l
ow v U4
ol o oleWl0 8tM.
at2 t''d r o •Pe.l
of
q' padv40l•
%,t
" ea , .

27
 ioo

too istI 0

•t MANN M•/t
U tI /

I.*tll

,-4. Ott! "OA'tmSO -- •

%5,
1?, Ie.ativ andel hy dim.rsimonun til 4n verikal.
111n tMLfeznsofm I rJo ar•e .JqnnoIfstrpAaMt \v Yuhn.

The effect of •wind

li o oopen fill was treated in an experimental

study of heat transport in rock fill, performed at the Institute

for Cold Technology from 1968 to 1969. (31) For these experiments,
a crushed rock material was laid to about 50 cm thickness in a
refrigerated room at the laboratory, see Figure 16. A total of
150 thermoelements were placed within the rock layer, at 7 diffe-
rent depths (planes) and with 22 elements at each depth (plane),
in order to moiitor the temperature distribution. Below the rock
layer, 9 heat flux guages (meters) were installed for measuring
heat transpOrt. The temperatures at top and bottom were control-
led by mewis of heat exchanger plates, in which alcohol could
rnirculaty, as well as by the air temperature in the room, by means
of cool~ng devices. The primary objective was to study the effects
of free convection in the presence of vertical heat flow. Those
results will be treated in the following section. Here, experi-
ments relating to "wind loads" %rill be discussed.

The two fans located at one end of the rock bed gave an air velo-
city of about 2 m/s across the fill. The air temperature was

28
10C, while the temperature at the bottom of the fill ranged
from 2 to 4*C. Figure 17 shows the isotherm surface for 9*C,
obtained through temperature measurements in the fill after
stabilization. Temperature profiles along the center line cut
0of the rock fill are shown in Figure 18. Both these represen-
tations show that there may occurred a strong influx of room

tkis of tn v 6aei-rv.

rock
.S
1~.L L~t5ar%tVQ0! (IOAi oool t u5. ~

2.
..
yarrl4og.
i.r.tf
Neate.
Z ah
pa n@.
e ';.9
,
j -sd.
/k.mj1ror.

7,,? '.l'r~ blip ;.r~orr.i

;.-r.."orntod peso.,

air in the upper layers of the fill and that this influx helps
to maintain a temeprature close to that of the air in the room
deep down in the rock fill. The average effective conductivity
of the fill was determined to be 1.46 V/mK, as compared to 0.45
SW/mK for the same conditions but with no forced air current.
This illustrates that even small wind velocities result in large
changes in temperature distribution and heat transport in coarse
rock fill. For high wind velocities, the demonstrated air in-
flux can practically eliminate the thermal resistance, particu-
larly if the air can penetrate completely into the fill. In
the wintertime, the outer surfaces of such fill may be

29 " f
covered1) by ice or snow, but during the autipnnonths and in cold
periods with no snow cover this mechanism can result in extreme-
ly heavy loss of heat from the ground, The importance of cove.
ring top and side surfaces of coarse fills wvth some-what finer
gravel or crushed rock is obvious.

pJr. l. So rMpIsn NoP .)mo l fed

res.4k med vinio.&-

k Jenr Lnc p. overflatei Av an aSteiLrn2.Lnc

NONMrtf 1-0a;Pp.
1970) . pe of.a . ..
*Upfaid of a had of dpi~jh~d roah.

pon o(t if.ux

1) The origi.nal uses the word "1tette(t)"l meaning "tighten(ed)r"

or "caulk(sid)" i~e. ice, snow or finer grain materials can
prevent influx of cold air (Translator's note)

I3
30
 D. Free Convection

Free convection in a porous material is often associated with

vertical heat flow in a horizontal layer of a certain thickness.
The temperature gradient causes density variations in the liquid
in the pores, which in turn results in liquid flow (transport).
The flow pattern is often characterized by polygon cells. How-
ever, experiments with both liquid layers and liquid-filled
porous materials show that the liquid does not start to flow until
a certain critical temperature has been reached. To be more exact,
this means that the liquid is at rest (i.e. stable) in the steady
state for temperature gradients below a critical value. For gra-
dients above that limit, the heat transport through the layer
will increase in proportion to the square of the temperature
gradient. The critical temperature difference across a layer of
a certain thickness is inversely proportional to the thickness of
the layer and the permeability of the material. This means that
convection sets in for a lower temperature difference if the thick-
ness of the layer increases.

Analytical studies of these relations lead to definition of a

specific, dimension-less (scalar) parameter Ra, Rayleigh's number

for porous layers containing liquids
Ra a au. 37
m
T is the temperature difference
h is the thickness of the layer
g is the gravity (accelleration) constant
a is the temperature expansion cofficient of the liquid
v is the kinetic viscosity of the liquid
am is the temperature diffusivity for the porous material

am (c0)ff

k is the permeability of the porous material

31
 Lapwood (33) has calculated critical Rayleigh's numbers for
layers of liquid-filled porous materials for various boundary
conditions:

(1) Two impregnable and conductive boundary surfaces : Rak a 40

(2) Conductive boundary surfaces, the lower impregnable, the

upper f~ee liquid: Rak * 27

Experimental studies for heat transport for Rayleigh numbers above

the critical value show that the effective condiictivity
averaged over an area, increases in direct proportion to the
Rayleigh numbez. For such numbers below a certain limit Ra a Ra 1 ,
Edder (34) found, for conditions corresponding to(1) above:

Nu - Ra/40 , 40 <Ra <Rai 38

Nu corresponds to Nuesels number and gives the ratio between

effective conductivity and conductivity without convection.

For other boundary values the equation has the form

1)
Nu a Ra/Ra , Rec< Ra <aRa 39

When Rayleigh's number exceeds Ra,, Nu will be smaller than the

value given by the linear relation, see e.g. the discussion by
0 Palm, Weber and Kvernvoll in (35).

Am mentioned before, the experimental arrangement described in

the previous section was used to study the effect of free convec-
tion in rock fill in the presence of vertical heat flow. These
experiments are descirbed in an internal report at the Institute
for Cold Technology in 1970 (32). The experiments were conduc-
ted with dry crushed rock having particle sizes in the range from

1) Subscript c is used to denote "critical", while k was used

previously. (Translator's note)

32
 20 to 80 mm. The rock was rather loosely packed and had a dry
density of about 1,500 kg/m3. The layer was 0.48 m thick and
covered an area of 1.8 "2.2 m. The arrangement in the refrige-
rated room is shown in Figure 16, see the previous section,
During the free convection experiments, a cover plate was in-
stalled at a certain height above the layer to protect it from air
movements caused by the air cooler. As stated before, heat trans-
port through the layer was measured with 9 heat flow gauges even-
ly distributed across the area, while temperature distributions
were recorded in 7 horizontal cuts, spaced 8 cm apart and with 22
sensors in each. The latter allowed generation of temperature
profiles and isotherms within the rock layer.

A total of 9 experiments were made with the heat flow directed

upwards. These were separated in two groups, corresponding to
different boundary conditions. In the first group (5 experiments),
the rock layer was exposed to the air in the room, while in the
second group the upper surface was covered (i.e. impregnable upper
* boundary). The cover consisted of a plastic sheet with a layer of
sand on top. In both cases, the boundaries are partially conductive
(i.e. not ideal heat conductors). Among other things, this causes
.. large variations in temperature near the upper and lower boundaries
when convection occurs.

In addition to these experiments, one was performed with downward

heat flow, to determine the conductivity of the layer when no
convection occurs. The average temperature during this test was
about 30C, which was assumed to be representative for the tempera-
ture range used. The most important results from these 10 experi-
ments are summarized in Table I1.

33
Table II: Experiments with free convection in crushed rock (32).
6 m2
Permeability K a 0.97 *10"

Experiment Xc) 6T(CC) tm(OC) Ra (e(W/mK) Nu Ra/Rak

I a 10.8 - - 0.45 1.0 -

a c 2.6 5.8 8.26 0.50 1.10 0.32
2 d 4.7 3.3 15.82 0.46 1.02 0.61
2 a 9.0 - 5.5 31.99 0.55 1.20 1.25
2 c 11.5 -13.4 44.72 0.79 1.75 1.74
2 f 19.0 -28.5 87.40 1.13 2.50 3.32
3 a 7.3 - 1.5 25.10 0.43 0.98 0.63
3 b 12.8 - 3.2 45.06 0.52 1.16 1.13
3 c 17.4 - 8.5 65.02 0.70 1.5b 1.63
3 d 19.6 -16.9 77.40 0,80 1.77 1.95

x) Experiment 1: Down-ward heat flow

Experiment 2: Up-ward heat flow, upper surface exposed
Experiment 3: Up-ward heat flow, upper surface covered

The relation Nu a Nu (Ra) given by these experiments is shown in

Figure 19a. By interpolation, one finds that the critical Rayleigh
numbers for exposed and covered upper boundary are approximately
Ra a 26 and Ra x 41, respectively. This comes rather close to the
theoretical values obtained for corresponding boundary conditions.
However, the theory assumes the boundary surfaces to be ideal heat
conductors. Figure 19 also clearly shows Nu(Ra)to bealinear re-
lation when the Rayleigh number e~xceeds a certain value.

Figure 19b, where Nussels' number is plotted as a function of

Ra/Rak (the ratio between the Rayleigh number and its critical
value) clearly shows that Eq 39 holds for Ra> Rak, up to a certain
value for Ra.

For the material used, with a thickness of 0.48 m, the theoretical

i3

34

S,--
i• '.-•
, ••k -,...
• • m. .. '--?" '• r.j• •,-••.•, .,•.,•,• . . . ..... •-.:... A.
 values for the Rayleigh number correspond to temperature differences
of 7.8 and 11.60C for exposed and covered upper boundary, respec-
tively, with an average temperature of O°C. For thicker layers,
the critical temperature difference will be less, since Tk is in-
versely proportional to the thickness. The effect of this in the
thermal resistance of rock fill is outlined in Figure 20. The ma-
terial is assumed to be dry and to have the same permeability as
the rock used in the experiments.

It is clear that doubling the depth, e.g. from 0.5 m to 1.0 m, may
in the worst case result in the same thermal resistance, provided
that the temperature difference is above the critical limit in
both cases.

, I*IdIT ~~~. A S9- -

4W
* j'lOLeKL~P *0001&!g

A-1-

71 /"

L I -
/ - '"

t 0 ..... 4 04 r r i ...
,,g * u O SO ld 10

RIO, %9, !LtIgnipJor1,tiq rIIoi@,@I:a~i¶, Iv forl.ll-

* td t
• l L<
e, IIiO'rI..ad
fio
d~lI~iije * 1 ,J@ i
I-JP .
l•35
P (15J
.

35
 *

I-I

January, Whnthe cl~lri.ticalll tempel.ll~l/ratur diffrec iecedd

Thisill-%p
wil cas a lowrin ofte teprt ure' at te bondar be

tweenthe roc layrl~lf ill and• the groun belo itwhich-reduces

th e lty temperature differencesacross the fii

rockca ill,
Astiircs

eatihe
prgrses u-airs temperaturewilnrmallyntedt inhies creaserdue
Jt uayo
thenyearl var Al temperature difference abov
ition, cded,
tex -

attheexesofainrsd heat loetayrhi

loss from nrae sinificanlyia ayr
This will inatie ontrinbuoe towrd redraureiag thegroundarymper-
ontelyeis during a rather shoert peridureainidy wianter, wrhen
turee theaovalauerfl)adherod th ntul below which
iscoerotedanular

mrgesssnth air temperatureswl omlytn oices u

withs wsdi
This pariclecause aeloerieng
sie fthe tes mperaturey
cnthre 1-ard
ine atathe
range,in the eround boundarybe
temeras
To h eryvrain emperature differences above the critica-ii
anb vie
enrly byu using
progpresethce aize fill havin looweres
i thempe pemaincrease
pelrmeaheilityitan
ratewilnorell thendt thle rusheu
xese
th
rokusdi of
e anto inceatsed
te yarlyvaratin.
Atemeratre
Inacosompartive
iffrenc
uly
core
abve te
n citilaayer.
nrealitytemperature differences above critical
the l imitawillo
cal
ten aluewillthu
to ecrase elowthe ritcal alue bu
atteepneo nicesdhatls rmteudrliglyr
"Ti ili im otiuetwad euigteArudtmea
tueteo h au ntrl hc scoet th anuI
 found to be 0.12 - 10-6 m2, as opposed to 0.97 10- 6 m2 for crushed
rock (32). This means an increase in the critical temperature
difference from AT x 60 C for a rock layer that is 1 m thick to
4T a 488C for a macadam layer with the same thickness.

In materials saturated by water, convection will set in for signi-

ficantly lower temperature gradients. This is primarily due to the
higher density and specific heat of water, as well as its lower
viscosity (c.f. Eq 37, where the Rayleigh number is defined).
Table III compares actual values of the parameters which enter
the Rayleigh number for air and water

Table III: Parameters included in Rayleigh's number. Tempera-

ture x OC

XO c ~ 0 a
W
....
J/k$oC ..kg/.
.........
1.25.. 3.
m../s 0-... ...
Dry 1/273 0.45 1000 . .6 o'4 3.6 10o6
Sa~turated 2 1 1.50 4187 1000 0.36 10-6 1.792 10"6
with water
*) tmm 20aC

In a material having a permeability k a 0.12. 106 m2 , equal to

that of the macadam mentioned before, one finds that free convec-
tion in a 0.1 m thick layer will begin already when the temperature
difference is 0.6*C, if the average temperature is 200C.

The low critical temperature differences for water saturated ma-

terials mean that free convection in connection with up-ward
heat flow can occur in a saturated gravel with d1 0 as low as 1 Mm,
In a 1 m thick layer of such a material (k a 10" M2), the criti-
cal temperature difference is 7.3 C, for an average temperature of
200C, However, in practice one will find that temperature diffe-
rences of sufficient magnitude and duration only occur during the
winter months, when the temperature is too low for convection in
water to take place. In addition, materials as coarse as that
referred to here will mostly have a water content well below satu-
ration.

37
 .. hTEAT TRANSPORT IN FREEZING SOILS

The most common approximate methods for calculating frost penetra-

tion in the ground (35) are usually based on the assumption that
water in the pores freezes at the normal freezing point for water.
If the heat capacity of the soil is neglected and the initial tem-
peratLre is assumed to be O0C throughout the material, one arrives
at the simple Stephan's equation:

S(t) = 2F '37

where ý is the frost penetration (depth)

X is the thermal conductivity of frozen soil
w is water content by weight
yd is dry density
L is the latent (freezing) heat of water
F is the frost quantity (s C)

t
F t (T- To)dt
0
where To is the freezing point
t is the time at which freezing occurs

This simplified approach can be extended to stratified media and

be further refined by introducing a contribution from heat flowing
from (lower levels of) the ground up to the freezing zone (frost
border-line).

There are strong indications that this rather simplified approach

is justified when determining the approximate fill depth needed to
prevent freezing in frost-heave prone sub-strata, i.e. when one
can assume that the freezing zone w141 not extend down to the finer
graul• frost-heave prone materials (36).

38
 IL. more fine grain materials, the assumption that all of the pore
water freezes at the normal freezing point for water will no longer
be valid. As an example, Figure 21 shows the portion of un-frozen
water in some silt and clay materials at temperatures below the
(normal) freezing point. The finer the grain, the more water will,
as a rule, remain unfrozen at a certain temperature. This sub-
ject will be discussed in more detail in Chapter II. Only the
effect of this condition on heat transport will be discussed here.

In a soil material where water freezes over a range of temepratures,

heat conductivity in the partially frozen material can be described
by the following equation

Ydý s wW ~wif) -wUU '5

a, specific heat of mineral particles (N 0.8 kJ/kgK)

caw specific heat of water ( 4 .2 kJ/kgK)
Cis, specific heat of ice (2.1 kJ/kgK)
Wu$ unfrozen water ratio by weight (water/dry material)
Other symbols as in the preceding equation (Eq 37)

This (last) equation shows that release of latent heat can be

included in an apparent specific heat for the freezing soil mate-
rial. This specific heat will, however, be strongly dependent on
temperature, which makes the equation quite non-linear.

3i

20d

~3
YI10 2. Eksemnler at raiatt vann~rnhQ'.J i Yanr%1V1V
e•

O
4t f *
1-e g " -i
 120, ,-

10,4

40/

•,, .: ': .• .• . ..'

M)I. 22, Pri~itt vhumv 'lu WAvJOUR. &v .1r11k?-9Ftrr

ro!,Zajj .id th
0I aJ iLId lu .' P. 1 '

This property is clearly important when one tries to measure the

thermal conductivity of frozen soils, using transient (not statio-
nary) methods and must use the linear equatica, for thermal conduc-
tion. This will be discussed in more detail in Chapter IV, in con-
nection with experimental investigations of thermal conductivity
in soil materials,

The effect of the un-frozen water content upon frost penetration

can be illustrated by considering released heat during cooling
down of different soil materials to temperatures below the free-
zing point. Figure 22 shows calculated values for the same two
materials as in Figure 21. For comparison, released heat for the
case where all the porewater freezes at the normal freezing point
are shown in the same fi Iure. In all cases, the dry density is
assumed to be 1,400 kg/m , corresponding to a saturated water con-
tent of 36 percent by weight.

It is clear, that the amount of heat released at a certain tempe-

rature be reduced significantly for a soil material where the
water freezes only partially. The greatest reduction will natural-
ly be found in cases where the soil is cooled only to a temperature
close to the freezing point. This effect will clearly result in
an increased frost penetration, in comparison to the case when all
the water freezes at the normal freezing point, which is assumed
when deriving Stefan's equation (Eq 37).
40
Calculations using the simplified method derived by Skaven-Haug
(36) for determining the frost resistance of roads~for which
frost depths and frost quantities have been measured point in
the same direction. The discrepancy between calculated frost
resistance (i.e. the calculated frost quantity required to freeze
the ground down to the registered frost depths) and the measured
frost quantity was largest for the cases where frost had penetra-
ted into very clay-rich materials (37). This experience shows
that knowledge of un-frozen water content at temperatures below
the freezing point is an important condition for satisfactory
calculations of temperature distributions in freezing ground.

When fine grain materials freeze, still another specialized condi-

tion will be important. As frost penetrates the material, suction
is created at the frost border-line. Depending on the permeabili-
ty of the un-frozen material, this creates water transport up to-
wards the frost border-line, where water freezes and releases la-
tent heat.

It would be logical to assume, that determining the velocity of the

up-wards moving water would be a relatively simple matter when the
rate of frost penetration is known. However, measurements of water
content during spring and autumn below the sections of road where
the previously mentioned experiments were made (37) indicate that
frost heaves often can be due to local re-distribution of water.
Such local re-distribution occurs most often in clays where the
suction is strong near the frost border-line, while the permeabi-
lity is low. Under such conditions, the water flow velocity will
also be low. In somewhat coarser materials, where the permeabi-
lity is significantly higher, one can expect that a larger portion
of frost heaves can be related to up-ward flowing water.

Clarification of these matters require a comprehensive experimen-

tal investigation of soil samples which freeze under controlled
conditions, Such experiments should include determination of heat

41
as well as (liquid) mass balance versus time, through measurements
of both heat flow and water transport velocity.

The water flow towards the freezing zone referred to above will
clearly tend to reduce frost penetration and will to some degree
counteract the effects of un-frozen water content.

EFFECTS OF VARIOUS HEAT T7 ANSPORT MECHANISMS

SýHE

An evaluation of theoretical and experimental studies of heat trans-

port in moist, porous materials shows that the effects of various
mechanisms to a large extent depend on the particle size of the
material, the degree of saturation and the temperature level. This
will in the following be utilized in an effort to determine limits
for the effects of each of the different mechanisms. As mentioned
previously, the goal is to clarify limitations for mathematical
models which use pure heat conduction aspects to find temperature
distributions in and around structures placed on the ground.

The regions of pre-dominance (influence) of different heat trans-

port mechanisms in un-frozen soil materials are shown in Figure 23
as function of average (equivalent) particle size (d10) and degree
of saturation. Types of soil corresponding to certain ranges of
d10 are indicated in the same figure. Dotted lines show expected
limits of saturation level in base materials, sands and gravels
in a road structure. The basis for this presentation will be dis-
cussed in the following, with reference to previous descriptions
of the different mechanisms.

For fine xrain materials, such as silt and clay) there is a region
at the lower saturation levels where moisture content gradients
will cause vapor diffusion due to changes in relative humidity in
the pore air as function of water content. From Figure 5 in this
chapter one can conclude that large variations in relative humidity

42
 T piau tIoc$.t ~wt~jo onm

0' hea
of~ in '0i

FIG 23 Intmirlyt l~s-tamjorf or Va' r&mis

at~

BEST eAVIAL COPY

~. onc~l~r~I 2uti. Io 43ee

* ".4 Ia~i~dI. -momP

as function of water content only occur at low saturation rates,
where the moisture potential (suction) exceeds about 10 bar. For
silt this may correspond to water content of about 5 percent by
weight (Sru 10 percent), while it may correspond to a saturation
level exceeding 30 percent in a clay material.

For moderate saturation levels in fine grain materials, temperature

gradients will cause vapor diffusion in the pore air due to corre-
sponding gradients in the water vapor saturation pressure in the
pores. For low saturation levels, this process is slowed down by
the lower relative humidity in the pore air. For higher satura-
tion levels, the re-distribution of moisture caused by this process
is counteracted by capillary water transport in the opposite direc-
tion. The latt,4r is determined by the moisture potential gradients
and the hydraulic conductivity for the saturation level at hand.
The hydraulic conductivity is strongly dependent on water content
for low to moderate water contents as well as the grain size in the
material (c.f. Figure 7). This means that the region in which the
temperature dependent vapor diffusion can cause significant moisture
gradients will move towards higher saturation levels as the material
becomes finer, as shown by the region labeled "2" in Figure 23.

The temperature dependent vapor diffusion will also contribute to

the transport of latent heat when vaporisation or condensation
occurs. However, this contribution can be treated as an added ther-
mal conductivity in the pore air, since the vapor diffusion in the
pore are governed by the temperature field in the pores in the
same manner as pure heat conduction in air.

For c soil materials, two regions are indicated at com-

paratively low and high saturation levels, when the heat transport
is influencid by convection in water and air, respectively. The
effect of free convection in horizontal layers (of fill) with an
up-wards heat flow is particularly limited by the permeability.
For example, In a dryer or dry fill which is one metre deep, the

44
~.. ... i'
 permeability must be about 0.5 -10-6 m2 for convection to occur
at a temperature difference of 100C. Permeabilities of this high
magnitude will only be present in very coarse gravel materials or
crushed rock. By comparison, a permeability of 10-9 m2 is requir-
ed for free convection to take place at the same temperature dif-
ference in a material saturated with water and for an average tem-
perature of 20C. Such permeability values can be expected in
medium sand with equal particle size. However, up-ward heat flow
will primarily occur during the cold (frost) season when convection
in water can not take place.

Forced convection due to ground water flow may affect heat trans-
port in materials having permeabilities in the range just mention-
ed (see Figure 15). For axial liquid flow in particle filled
* columns having radial host conduction (flow) it has been demonstra-
ted that the flow rate (velocity) affects radial heat transport,
which may be explained by "lateral mixing" or dispersion. Corre-
sponding studies involving horizontal liquid flow in fills (layers)
having vertical heat conduction (flow) are not known (to the author).
S..However, such conditions will normally not occur in conventional
road structures due to flow within the coarser materials in the
* structure itself.

Water permeability will clearly decrease when the degree of satura-

tion falls below I. In the same manner, the permeability in air
will decrease when the degree of saturation exceeds 0. Since actual
permeability values for air and water in such systems are not avail-
able, the limiting curves for convection are rather approximateI;
for saturation levels between 0 and 1, but are intended to corre-
spond to permeabilities of 0.5 1C 6 and 10-9 m, respectively.

Heat transport in coarse materials can also be partially due to

thermal radiation between particles. At an average temperature
of 20aC, the contribution due to heat radiation may increase the

1.) In other words "guesstimates" based on "informed ignorance"

(Translator's note)

Ii

45
S~_ ____ ____ __-A
effective heat conductivity by 5 percent in dry gravel with a uni-
form particle size of 10 mm (see Figure 4). This contribution
increases in proportion to particle size. Reduced pore cross-section
due to a certain water content and an increase in conductivity due
to increased saturation levels will tend to reduce this contribu-
tion in moist materials. The boundary of the region where ther-
mal radiation is important was derived under certain assumptions
regarding the effe:t of saturation level on the effective pore
cross-section and the increase in conductivity due to increased
degree of saturation. The boundary corresponds to a contribution
to the conductivity, due to heat radiation, of 5 percent.

In principle, the effect of heat radiation can be included in the

effective or apparent conductivity of the porous material and thus
does not present a problem in numerical determination of tempera-
ture fields, provided that one tak1ýa tho temperuturq 4.,cendence of
the parameters into account. However, this requires knowledge of
the effect of radiation in such structures (systems).-

If one compares the indicated boundaries with the saturation levels

in sub-strata and upper layers, (Figure 23) one finds that heat
transport in un-frozen materials will be affected only slightly by
mechanisms other than thermal conduction, except for the coarsest
materials where heat transport due to convection and radiation may
become significant.

The proceeding overview (discussion) has been limited to un-frozen

soil materials. In frozen, fine grain materials (frost prone),
water transport towards the freezing zone (boundary) will affect
the heat transport due to loss of latent heat. The mechanism
behind this effect, which includes coupled heat and moisture trans-
port, has not been clarified. However, thermal design of road struc-
tures is normally performed so that frost penetration will not occur.
Under these circumstances, these mechanisms can be disregarded.

46
S..
. . •. ....i.....•j ... . "' nm
Iq lua ~la l~m , ,,,• • lilii wh• .. ..... 4
 REFERENCES - CHAPTER I

1. W. B. Argo, J. M. Smith: Heat Transfer in Packed Beds.

Chemical Engineering Progress. 49, (8), 1953, pp. 443-451.
2. S. Yagi, D. Kunii: Studies on Effective Thermal Conductivities
in Packed Beds, A.I. Ch. E. Journal, 3, (3), 1957, pp. 373-381.
3. P. Zetner, E. U. Schlunder: Einfluss der W~rmestrahlung und des
Drucket auf den Wirmetransport in nicht duregestromten Schýttungen.
Chemie-Ing.-Tech, 44, (23), 1972, pp. 1303-1308.
4. N. Vakao, K. Mato: Effective Thermal Conductivity of Packed
Beds, Journal of Chemical Engineering of Japan, 2 (1), 1969,
jp. 24-33.

5. N. Wakao et a&: View Factor between two Hemispheres in Contact

and Radiation Heat-Transfer Coefficient in Packed Beds. Int. J.
Heat Mass Transfer, 12, 1969, pp. 118-120.
6. N. Wakao: Effect of Radiating Gas on Effective Thermal Conducti-
vity of Packed Beds. Chemical Engineering Science, 28, 1973,
pp. 1117-1118.
7. E. Schmidt: Einftirung in die Technische Thermodynakik, Springer
Verlag, Berlin 1963, p. 394.

8. F. Birch, H. Clark: The Thermal Conductivity of Rocks and its

Dependence upon Temperature and Composition. American Journal
of Science, 238, (8), 1940, pp. 529-558.
9. 0. Krischer, H. Esdorn: Wirmeleitung und Dampdiffusion in feuchten
Gutern. VDl-Forsch,-Heft 402, Berlin 1940.

10. C. W. Rose: Water Transport in Soil with a Daily Temperature Wave.

.1 Theory and Experiment, Aust. J. Soil Res., G. 1968, pp. 31-44.
11. C. G. Gurr et al: MOvement of Water in Soil due to a Temperature
Gradient. Soil Science 74 (5) 1952, pp. 335-345.
12. D. Croney, J. D. Coleman: Pore Pressure and Suction in Soil
Fra "Pore Pressure and Suction in Soils". Butterworths, LonLon
1961, pp. 31-37.
13. J. A. Currie: Gaseous Diffusion in Porous Med.La, Part 2 - Dry
Granular Materials. British Journal of Applied Physics, 11,
1960, pp. 318-324.
14. J. A. Currie: Gaseous Diffusion in Porous Media, Part 3 - Wet
granular Materials. British Journal of Applied Physics, 12, 1961,
pp. 275-281.

15. J. R. Philip, 0. A. de Vries: Moisture Movement in Porous Materials

under Temperature Gradients. Transactions American Geophysical
Union, 38, (2), 1957, pp. 222-232.

47
"----. --
---------------------------
,'-,-r---
16. W. 0. Smith: Thermal Conductivities in Moist Soils. Soil Sci.
Soc. Am. Proc, 4, 1939, pp. 32-40.
17. 0. Krischer: Die Leitflhigkeit des Erdbodens. Beiheft zum
Gesundheits-Ingenieur, Reihe I Heft 33, MUnchen (Munich) 1934.
18. C. G. Gurr et al: MOvement of Water in Soil due to a Temperature
Gradient. Soil Science 74, (5), 1952, pp. 335-345.
19. H. L. Rollins et al: Movement of Water in Soil due to a Tempe-
rature Gradient, Highway Research Board Proc. 33 1954, pp. 492-508.
20. J. W. Cary: Water Flux in Moist Soil: Thermal Versus function
Gradients. Soil Science. -CO, (3) 1965, pp. 168-175.
21. W. D. Joshua, E. De Jongi Soil Moisture Movement under Tempera-
ture Gradients, Can. J. Soil.-Scie., 53, 1973, pp. 49-57.
22. R. J. Kunze et al: Factors Important in the Calculation of
Hydraulic Conductivity, Soil Sci. Soc. Amer. Proc., 32, 1968,
pp. 760-765.
23. D. A. de Vries: Simultaneous Transfer of Heat and Moisture in
Porous Media. Transactions, American Geophysical Union, 39 (5)t
1958.
24. D. A. do Vries: Hat warmtgeleidingsvermogen van grond, Med.
Landbouwhogeschool, Wageningen, 52, 1952.
25. P. Adivarahan at al: Heat Transfer in Porous Rocks, Society
of Petroleum Engineers Journal 1962, pp. 290-296.

26. S. Yagi, D. Kunii: Studies on Effective Thermal Conductivity in

Packed Beds, A.I.Ch. E. Journal, 3 (3), 1957, pp. 373-381.
27. A. E. Scheidegger: The Physics of Flow through Porous Media.
Univ. of Toronto Press, 1957.
28. Bygg. Huvuddel 1B. Alminna Grunder: Avd. 17:4 pp. 306-308.
Ab Byggmlstarens Fdrlag Stockholm 1972.
29. S. Ergun: FLUsuigkeitsstrbmung durch Fnllkdrpersaulen, (Pest-
beatten) Referert i Chemie Ing.-Techn., 25 1953, p. 262.

30. Burke & Plummer: Referext i Chemie Ing.-Techn., 25, 1953, p. 262.
31. 0,. Johansen: Unders0kolse av varmetransporten i kult og pukklag.
Det store ekuamensarteid, Inst. for kjoleteknikk, NTH, Trondheim,
1970.
32. 0. Johansen: Innfl teise av fri konveksjon i grovkornede
vegbyggings-materialer. Institutt for Kj~leteknikk, NTH,
Trondheim, 1970.

48
-- ... ~-
33. .R:. Lapwood: Convection of a Fluid in a Porous Medium.
c3 Cambr. Soc. 44, 1948, pp. 508-521.
34. J. W. Elder: Steady Free Convection in a Porous Medium
Heated from Below. J. Fluid Mech., 1967, 27, pp. 29-48.
35. E. Palm et al: On steady convection in a porous medium.
Preprint series, Matematisk Institutt, Universitetet i
Oslo, 1971.
36. Sv. Skaven haug: The Design of Frost Foundations. Norwegian
Geotechnical Institute, Publ. 90, Oslo 1971.
37. A. F. Knutson: Upubliserte resultater, 1973.
38. E. Angen: Analyse av vanninnhold i vegkonstruksjoner, Intern
rapport, SINTEF, NTH, Tronheim, 1973.
39. A. K. Fleming: Applications of a Computer Program to Freezing
Processes. Proceedings of the XII International Congress of
Refrigeration, Washington D.C. 1971.

49
 CHAPTIM II

THEORIES FOR HEAT CONDUCTION IN COMPOSITE MATERIALS

The previous chapter described (discussed) various mechanisms for

heat transport in moist and frozen soil. It was found that, in
most practical cases, thermal conduction plays a dominant role in
the heat transport (transfer). This prompts a more detailed study
of the factors which determine thermal conductivity in composite
materials.

The following sections present an overview of the most important

relations that apply to derivation of theoretical models for deter-
mining thermal conductivity in such materials. Emphasis will be
placed on organizing the different methods in a set of major approa-
ches.

Many of the methods (to be) mentioned are developed for determining
transport coefficients which are analogous to thermal conductivity,
e.g. parameters such as electrical conductivity, dielectric constant,
magnetic permeability, permeabilty associated with liquid flow in
porous materials and diffusion coefficients. To simplify matters,
all these analogue transport coefficients will in the (following)
text be denoted by "conductivity", without specific mention of
the nature of the transport coefficient. That will be evident from
the list of references.

1. EXACT THEORY

The problem of determining conductivity in a composite material

can be formulated as follows: A material is h-- e, •a on a ma-
croscopic scale and isotropic. On a microscopic scale it con-
*ists of regions having different conductivities. How do these
regions together form a macroscopic conductivity?

5o
Ideally (in principle), this problem can be solved by inteqra-
ting the heat flow equation for steady state (stationary) condi-
tions over a region of the material, large enough to give a re-
presentative picture of the average microgeomnetrical conditions(i).

That is only possible in simple systems, e.g. regular configura-

tions of spheres. Solutions for such cases will be given in sec-
tion 3 of this chapter. For natural materials, where the micro-
geometry is very complicated, such calculations are impractical.

However, there exists a general solution, for which one treats

the microgeometric conditions in a general (idealized) form.

If the material is homogenous on a macroscopic scale, one can de-

scribe the microgeometry by means of a set of random parameters
(p-point correlations), The basic form for these correlations
is a function ti (r), defined by

i (•) u 1 if (the position vector) I is in component "i"

M() " 0 if I is outside component "ill

The volume average for the function oi (rj will correspond to the
relative volume (volume share) of the component "il" in the material,
provided that the integration is carried over a sufficiently large
volume in a macroscopically homogenous and isotropic material.
That relative volume also gives the probability that a point selec-
ted at random will fall in component "1'02).

1) "Correlations" is a literal translation. The context indicates

a joint probability density function.
2) From the context it appears that "components" refers to "regions"
or "cells" with certain thermal properties (Translator's notes).

51
In the same manner, a volume average of *the type <Pi ('r)1
S(r),.....
0 (' *)> expresses the probability that all the
points r1 through rp are located within the same component. Such
a volume average is called a p-point correlation. As an examp2e,
a two-point correlation will define the probability that both end-
-points of a line segment fall within the same component. In a
macroscopically homogenous and isotropic material, this probability
will be a function of the length of the line segment. Three-point
correlations will give the probability for the three corners of a
triangle to fall within the same component. This probability is
a function of the angle between two sides in the triangle and their
lengths.

With this kind of random characterization of the microgeometry of

a material as a starting point, W,. Brown (1955) (2) 'derived a
general expression for the conductivity of a two-component material.
The expression can be written under the form:

1 * 7.A A (30o 1

where %.-XI X is the macroscopic conductivity

XB
2 XI (conductivity of the components)

Ai are integrals containing the p-point correlations

n is the volume share (relative volume) of phase I

W.EA. Davies (1971)(3) has shown that A contains the three-point

correlation, while higher order correlations are included in the
4

expressions for other A.

Eq I shows that the macroscopic conductivity depends on the random

properties of the component microgeometry, as well as the conduc-

52
tivity of each component and their relative volumes. This may
appear to be a trivial result, but has nonetheless been over-
looked in many studies within this field: "The flow of papers
seeking an1) universal formula (for the conductivity of compo-
site materials) is unending"t. Quote from P.J. Beran (1971)(4).

For a geometry where all 0, Eq I shows that the conduc-

= 0/.
tivity is exactly given by

" _( -n) .

fDavies (3) has shown that such a geometry can be represented as

a system consisting of concentric spheres, where a nucleus (core)
of one material is surrounded by a spherical shell of the second
material. These spherical units (cells) may be assumed to occur
in all sizes, so that they together completely fill the given ma-
terial. The diameter ratio in each one equals the macroscopic
volume ratio.

A material which consists of dispersed spherical particles,

together forming a small percentage of the volume, will closely
approximate this geometry. Consequently, the conductivity of such
a material can be calculated exactly2) from Eq 2. The same is
true for a system where a small portion of the total volume con-
sists of spherical pores, dispersed in a material wi-th higher
conductivity. (In this case, the properties of the materials
change places in the equation.)

For other microgeometrics it will clearly be impossible to cal-

culate the conductivity exactly, since it is difficult to see
how all the p-point correlations involved could be determined.

1) Sorry about this one, but that is exactly what the original
says. (Translator's note)
2) Probably some-what overstated (Translator's note).

53
 However, in a more recent report, Brown has pointed out the possi-
bility of determining limits for the conductivity in a certain
system, based on knowledge of the p-point correlations up to a
given order (5). The more information one has of the statistical
properties of the microstructure in the material, the narrower
these limits are.

The evaluation of such li.mits will be discussed in the following

section.

2. EVALUATION OF LIMITS

A. Limits according to Hashin-Shtrikman

Tt is trivial to state that the conductivity of a composite mate-

rial must fall between the highest and lowest conductivity of the
components which make up the material. This is the best one can
do based on knowledge of the conductivity of each component only.
If, in addition, one knows the relative (share of) volume for each
material component, limits for the conductivity can be found from
the arithmetic and harmonic means of the conductivities for each
component

SXH X A 3

F Awhere
A E n
ixI (arithmetic mean)

E ni/Ai)"
E.( ( harmonic mean)

The proof for this (relation) is usually based on (the work by)
Wiener (1912)(6), so the limits could be called Wiener limits.
These limits are general and will apply even if the material is
not homogenous or isotropic.

54
 Using calculus of variations, Z. Hashin and S. Shtrikman (1962)(7)
developed limits (limiting values) for macroscopically homogenous
and isotropic materials, which represent a further refinement (im-
provement) of limits for the conductivity of a composite material.

For a two component material these limits can be expressed by the

following inequalities:
I >l 0•(- n) X-~2 X1( ' X2 > %I

2 X1 - ý2~ S

These expressions are identical to those which were previously

proven to hold exactly for a system of concentric spheres (Eq 2)
and do thus correspond to a geometry for which the integrals A
A are zero, which Davies also proved in his report (3).
A.

For a material composed of m components, the limits can be ex-

* pressed in terms of similar$ dimension-less, ratios between conduc-
tivities. The conductivities of the components (involved) are
arranged in order of increasing value so that:

A1-•< 0 400< Xi 1i + 1 <" 0' " I

The inequalities can then be written under the following general

form:

m 7

If lower and upper limits are denoted • and • , respectively,

while the sums in the right hand members of Eqs 6 and 7 are de-
noted B. and s*, the following expressions for the limits can
be derived:
55
.A
 I (1 + 2 )/(I -a
)) 9

As indicated previously, the Hashin-Shtrikman limits constitute

the best(possible) estimate of conductivity limits in a material
when information about its microgeometry is lacking. These li-
miting values also correspond to extreme microgeometries. The
distance between them can thus give an idea of the sensitivity
to variations in the microgeometry.

Figure 1 compares the Hashin-Shtrikman limits to those given by

Wiener, as function of porosity and for various ratios of conduc-
tivity (in the two material components). It is evident that the
Wiener limits give the best improvment at low conductivity ratios.
One also finds that the Hashin-Shtrikman limits are practically
identical (coincident) for conductivity ratios less than 3 and that
the spread between these limits is less than 10 percent for con-
ductivity ratios up to 10. Thus, for most applications these li-
mite will by themselves give an estimate of the conductivity
* which is sufficiently accurate.

For larger conductivity ratios, the improvement is insignificant

and the limits are too far apart in order to give a (reliable)
estimate of the conductivity. As indicated previously, the large
spread between limits is due to a considerable sensitivity to
variations in microgeometry in the case of large conductivity
ratios. This condition has significant practical implications.

As mentioned at the end of the previous section, the range between

these limits can only be improved by inclueing information about
the statistical properties of the microgeometry in the materials.
In recent years, a number of reports have been published which seek
Ii to improve these* limits on that basis, These reports will be dis-
cussed in the following section.

jME56N
 B. Limits based on point correlations

Using the previously mentioned statistical description of micro-

geometry, M. Beran (8) and W.F. Brown (5) reported, in the same
year, improved limits for the conductivity by utilizing three-point
correlations. W.T. Brown found it possible to calculate his li-
mits for a particular form of geometry where the components have
the same relative (share of) volume and identical geometry (sym-
metrical material). The calculations show a significant improve-
ment (narrowing) of the limits in comparison to the Hashin-
Shtrikman limits, particularly for low conductivities.

However, the requirement that the three-point correlations must

be known makes it difficult to find the limits for other types
of materials.

----------- --------------------- -- ... ,

,-----r -r............. r-•..........

x . ~.08

,0 ,I•"• i '..
.. •I•. LoI, EI0 ,, .,IId3'iB ,it,.'-,o
'0 0 2 04, "S ý"S'•1`0 '0 -. ' 10 0 0

'" .. •...
.... ~I t•
" Z ... -*' '

57dotit%,l

N
7OP
An article from 1974 by P.B. Corson (9) gives the results of ex-
perimental determination of three-point correlations in selected
alloys (lead-aluminum, iron-lead). From these experimental re-
sults, Corson derives empirical formulas which express the three-
point correlations for the investigated materials as a function of
the length of the three sides in the triangle and the relative
volume for the components.

These empirical formulas were then used to calculate values for

the limits in materials w4th similar microgeometry (11). The
expressions given by M. Berans for limits of conductivity were
used, mince the three-point correlations there are included in
generalized form (8).

For low conductivities, these calculations also show a significant

improvement, in comparison to the Hashin-Shtrikman limits. How-
ever, the method can not be expected to have wide-spread applica-
tions. The effort required for determining three-point correla-
tions in a given material is extensive and the calculations re-
quired for finding the limits using empirical three-point corre-
lation expressions are complicated.

Also M.1. Miller (1969)(12) used the general expressions previous-

ly published by Beran as a starting point. He points out that these
expressions have limited usefulness as long as the three-point cor-
relations are difficult to determine. However, by defining a par-
ticular type of materials which he calls cell materials he has
beer, able to express three-point correlations as a numerical value
for each component. These values vary in magnitude from 1/9 to
1/3 and have a simple geometrical interpretation. The number 1/9
represents spherical cells while 1/3 signifies flat (plate-shaped)
cells. All other geometries are characterized by values between
these limits, regardless of how complicated the geometry is.

Miller makes a distinction between symmetrical and asymmetrical

cell materials. In the first type, the two components have the

58
 same average geometry, while in the other types the component
geometries are different. The most important restriction for the
cell materials is that the conductivity of one cell must be sta-
tistically independent of the conductivities of all other cells.
In addition, the material must be homogenous and isotropic in a
macroscopic sense.

For an asymmetric cellular material, the Miller limits can be

expressed as

r -.

"* .( ~ -'
r'4.** ' .5'. -: : i r):I

where X /X I X2 '1

n is the relative volume (share) of component 1

GI, G2 are numerical values given by the call geometry

For a symmetrical cellular material, G! a Ga. In general, materials

are of thm asymmetrical type and in addition it is usually possible
to determine the cell geometry for only one of the materials.

In such cases, limits for the conductivity can be determined by

using G-values for the component with un-known geometry, which re-
sult in the highest upper limit and the smallest value for the lower
limit. From expressions for the limits one obtains the following
numerical values:

Granular materials (G2 known)

0I a 1/3 gives upper limit, Gý 1/9 gives lower limit

59
Porous materials (G1 known)

G2 = 1/9 gives upper limit, G2 = 1/3 gives lower limit

(Note: Index I always refers to the component having the lower

conductivity)

Figure 2 shows calculated values of the Miller limits, as function

of volume ratio for a number of conductivity ratios. Examples are
given for both granular materials containing spherical particles
(02 a 1/9) and porous materials having spherical pores (G1 = 1/9).
The Hashin-Shtrikman limits are shown for comparison and the cross-
hatched regions define the Miller limits.

One finds that the largest improvement is obtained for small con-
centrations of particles or pores, where both Miller limits asymp-
totically approach one of the Hashin-Slitrikman limits.

For low conductivity ratios, the improvement is also significant

at larger concentrations. Here the Miller limits indicate that in
the case of spherical particles only the lower half of the range
defined by the Hashin-Shtrikman limits should be considered, while
in the case of spherical pores one should consider the upper half
of that range.

Miller indicates that the variations in conductivity represented

by his limits for a given particle geometry is due to the many
possible packing configurations which may occur for a certain
geometry. To improve on (narrow) the limits one thus needs in-
formation about the packing configuration, in addition to the par-
ticle geometry. This kind of information will also be included in
point correlations of an order higher than 3.

M.A. Elsayed (1974)(13) has recently presented a report where he

has succeded in including five-point correlations in expressions
for the conductivity of materials with two iomponents.

f/
60
 , * -I | -t T " '

CAA.

" 0 01.0 " 0 0 2'1

02

0 00 . 00 000 0 0 0

B:SI AVA'LABL cO
1TY
., I Q•
C)• : '. Os 0 00 10 1.0 0• i o 0

k. 1

These expressions aloinclude factors which depend on the packing

configuration, like Miller's formulas.

On the other hand.,Elsayed has not been able to give a geometrical

interpretation of the factors which represent packing configuration
in the improved expressions for (conductivity) limits. This causes
practical difficulties when using h1s (expressions for these) li-
mite. However, he has calculated limits for two geometries which
he claims are typical for spherical and flat (plate-shaped) particles,
respectively. The results show a significant improvement relative
to the Miller limits, particularly for low conductivity ratios.

Further investigations along the lines developed by Elsayed may re-

sult in a geometrical interpretation of the packing factors and

61
 thus contribute to better accuracy for estimates of conductivity
in certain systems.

C. Other evaluations of limits

S.R. Corriell and J.L. Jackson (1968)(14,15) used a cubical ele-

mentary cell in a material with known microgeometry and an approach
based on calculus of variations. They proved, that approximate
upper and lower bounds for the conductivity can be obtained by
assuming planar isotherms and paralell heat flow-lines, respectively
(Figure 3).

"*":IY2•E•S:ONA". •.'•0Q•'•' Ž,YI"•" '•

*' . !'•:I•'cS¶•

.on

Si,,. • • ,, S ,. .- €

- ,. '**.,' -.-. ',,.,j ,.*.

The general expressions for these limits are given by the follow-
ing integrals:

I i Cdx dy/ I dz,'x (xIyIZ) 12

0 0 0
11 1
-•. 1l / (dz/ I Xx,y,z,)dx d,5' 13

62
 As illustrated by Figure 3, these expressions are obtained by
assuming that each elementary cell is divided into an infinite
number of filaments parallel tojor plates perpendicular to the
direction of heat flow, respectively. The relative volume (share)
of the lower conductivity component in each of these filaments or
plates can be regarded asrespectively, one and two-dimensional
porosity.

The way the calculation methods are structured, these "porosities"

can be re-arranged according to magnitude, one obtains a cumulative
distribution functions as the number of filaments or plates tends
toward infinity. Examples of such cumulative distributions are
shown in Figure 4, for the case of tightly packed face-centered
spheres.

-i /

"4.
A.1

I/

'b. . I ., '-b* S '' n ,

If the statistical distribution of these two "porosities" has been

determined for a given material, one can consequently determine
upper and lower limits for the conductivity in the material. For
cubic configurations of face-centered spheres with different par-
ticle concentrations, such calculations yield results which are
insignificantly better that the previously mentioned Wiener limits,
except for low particle concentrations. However, in materials with
other distributions for the two porosities one finds that the li-
mits (so derived) give a better result than the more general limits.

63
- 1,•-. • 1 . • .. : ., ...... -...
:.. .....-i .... .-..-- I,..-*w .- -1,..v .- q• . i---. i -m•
. 4.i,
•
In principle it may be possible to improve on the limits (further),
provided that one can account for the statistical distributions
of one and two-dimensional porosities. However, no such efforts
have been reported in the literature.

The problem of describing the microgeometry in actual materials

has led to a third type of formulation for limits, based on expe-
rimental (empirical) values of conductivity in a material having
a given composition by volume and certain conductivities for the
components. (S. Prager (1969)(16)). Empirical values of conduc-
tivities in a porous material, in which the cavities are saturated
by a certain gas, can be inserted in these formulas in order to
determine the conductivity for cases where the material is satura-
ted by a different gas or liquid. In a similar manner, measured
values of the dielectric constant for a material can be used to
determine limits for the thermal conductivity in the same material.

These limits are given by relatively complicated expressions which

will not be reproduced here (see reterence 16, FIs 27 and 28).

As an illustration the Prager limits were computed under the

(ficticious) assumption that the condcutivity has been experimen-
tallydetermined for two different conductivity ratios (X2/X 1 10
and a 125) but for the same porosity (n = 0.40). Figure 5a
shows the limits computed for X2/A, u 10, while Figure 5b shows
values computed for the other ratio.

For the case where the (net) conductivity is assumed known for
/ 125, the Prager limita fall outside the Hashin-Shtrikman
limits over part of the range. It is also evident, that the Prager
limits diverge rapidly for values of x 2/AI higher than that where
the conductivity is assumed to be known.

Figure 5c illustrates the case where the average conductivity has

been measured at two different conductivity ratios (for the same

64
porosity). The limits (bounds) for the conductivity are here
given as the lowest upper limit and the highest lower limit for
the two cases, respectively. As shown by the Figure (5c), the
resulting limits (bounds) constitue a very substantial improve-
ment relative to the Hashin-Shtrikman limits. For conductivity
ratios below 125, the coaductivity is defined within a narrow
band with a doviation from the average which is less than t1o per-
cent.

By determining the Prager limits from two experimental values,

in the manner illustrated by this example, one can thus find the
average conductivity ratio less than the largest value for which
the conductivity has been (experimentally) determined. However,
one important condition is that the microgeometry must be the same
for the two cases where the conductivity is measured. This condi-
tion is most easily met by saturating the sample with different
liquids, provided that the porosity stays the same.

D. Conclusion

Togethersthe different limits (bounds) for the (average) conduc-

tivity described in this section illustrate that the conductivity
can not be determined exactly unless one has complete information
about the microgeometrical conditions. However, all the different
formulations show that the ratio of conductivities for the two com-
ponents are of primary importance in this context. For low conduc-
tivity ratios one can find the (average) conductivity with suffi-
cient accuracy, based on volume ratios and conductivities for the
components, but without information about the microgeometry. When
the conductivity ratio Li large, even relatively detailed informa-
tion about the microgeometry, in the form of three-point correla-
tions, will still result in widely separated limits so that the
conductivity can not be determined with sufficient accuracy.

65
 (.tpdrt'~dM~a1 uo!ud givenl' It ',/X 2 , M zi 0,40

ko,,

5 4,070 4C 6

b)2 Ekprmne7 ed itvd X/%

Ez er.Ma
t l a ko ) ve & l \t : SI L14

7, . 0 0

!o~s.twoah
'e~e.vrd ed ýuei,! a! a nsa!prer

In
such casess onl _ exermetalivsiainswleuti

accptbl codciiy /lpeterdrcl ridrcl

by mans
f th Praer lmits

Bet/a/lbeCP

66
 3. MATHEMATICAL MODELS

A. Introduction

The basic theory for conductivity of composite materials, treated

in the previous section, showed that the conductivity depends on
three sets of parameters: conductivity of the components, their
relative volumes and the microgeometry of the system. The funda-
mental treatment involving limits (bounds) showed that the latter
parameter set plays an important role, particularly when the con-
ductivity of components- are different from each other. Even with
relatively extensive information about the microgeometry (know-
ledge of three-point correlations) one can in principle not deter-
mine the conductivity with sufficient theoretical accuracy for
these cases.
Itowever,a number of efforts to develop mathematical models
for
determining the conductivity of composite materials have been re-
ported in the literature. Such models are usually based on approxi-
mate calculations of the conductivity for idealized geometries. The
choice of approximation will most often cause the calculated con-
ductivity to deviate significantly from the conductivity in the
idealized geometry. At the same time, the choice of geometry will
in itself constitute an approximation relative to the microgeometry
of the particular material for which the model is created. Both
these factors will be of increased importance for large differences
in the conductivities of the components. This means that such equa-
tions at best can be justified as semi-empirical relations, which
by suitable choice of idealized geometry may be fitted to experimen-
tal values for certain groups of materials.

In the following section, different (mathematical) models will be

discussed, with reference to these conditions.

67
 B. "Exact" calculations

For simple geometries, such as regular configurations of spherical

particles, one can compute the conductivity analytically or nume-
rically with a high degree of accuracy. An analytic solution for
spherical particles in a cubic (gitter) configuration was presen-
ted by Rayleigh already in 1892(17). The calculations contain cer-
tain approximations which apply to spheres which are nearly or com-
plete2"- in contact with each other. An improved solution was later
developed by Meredith and Tobias (1960)(18). deVries (1952)(19)
extended Rayleigh's methods to other configurations of spheres.
A different analytical approach was used by Walther (1970)(20) to
compute conductivity for spheres arranged in cubical and tetrahedri-
cal configurations (gitters). Recently, computers have been uti-
lized for numerical determination of conductivity in regular con-
figurations of spherical particles (21, 22, 23). Such results are
available for tightly packed cubic and orthorhombic configurations
of spheres with volume ratios of 0.524 and 0.605, respectively.

A comparison between numerical results for spheres arranged in a

cubic configuration and the previously mentioned analytical solu-
tions indicate that the latter give lower (conductivity) values
than the numerical methods, except for very low particle concentra-
tions where all calculated valu~es coincide with the lower Hashin-
Shtrikman limit,, Among the analytical results, those calculated
dl by Walters give the best agreement.

Figure 6 compares Walters@ solutions for cubic(al) and tetrahedri-

cal configurations (packing) of spheres, for a conductivity ratio
of 100. Even if the values are some-what low for large particle
concentrations, these results give a good indication of how the
configuration affects conductivity for a certain porosity.

68
 17

,,• , 'I// I

f 00" j CA,t.. L ". 0

A '117 tio
Sof
gr~t
,t . 0 .
11Vt ~',i
I ., j
s.I w
.r
*.v ',~
V .i kul ukilIei ved
a II " a . 4ak,

Figure 7 shows numerically determined values for regular configu-

rations of spheres, drawn together with the Miller limits for si-
milar structures. The shaded region shows the expected variations
for granular materials with dispersed particles and a conductivity
ratio of 300. This region is derived by interpolation of experi-
mental data for such materials with conductivity ratios in the range
between 100 and 1000. These data were taken from Chen and Vachon
(1970)(24), as well as from measurements made by the author on dry
soil materials (see chapter V).

'-.. 4, 1'-,4-L'L.

o.lot

69
,
69.
",j
 These results show that the expected variations in conductivity
of granular materials with dispersed particles is considerably
less than the theoretical limits (derived) for spherical particles.
The numerically determined values all fall within the empirical
limits.

Based on these studies one can state that the conductivity of ma-
terials containing spherical particles can not be expressed as a
function of porosity and conductivity ratio only. In addition to
these parameters, the geometrical (packing) configuration is also
significant. However, the latter parameter has less importance
(influence) than indicated by theoretical limitsderived for sphe-
rical particles. However, these computed values are valid for a
rather narrow range of porosities. This condition, as well as the
relatively wide incertainties that may be expected, mean that an
empirical determination of conductivity will be much superior to
this (analytical) method in most cases.

C. Approximate methods based on electrical

field theory.

The conductivity of a composite material is defined by the follow-

ing equation:

.X<aT> 14

where <q, is the volume average of the heat flux

, Si <vT> is the volume average of the temperature gradient

These volume averages can be expressed in terms of volume averages

for each of the material components:

<q> n <q>1 + (1-n) <q> 15

70
 <VT> = n <vT>I + (1-n) <"T >a 16

Since each of the components is homogenous and isotropic, the

following also holds

cq> = X <c7T> 17

<q> 2 = xp <vT 18

By combining these equations one can express the conductivity of

the composite material in terms of the volume averages of tempera-
ture gradient in both components:

Xin + X2(1-n) <7T>2 / <7T>

X ... .. 19
n + (1-n) /Te
/ <1T>9

This expression gives exact values for conductivity if the ratio

between volume averages of temperature gradient is known.

A common method for determining this ratio is based on an analogy

with the elctrostatic case. If a spherical particle is insortet
into a homogenous electrostatic field, one obtains the electric
field inside the sphere from
31
EI - 3 ,, E120

E is the impressed homogenous field

CI is the dielectric constant of the impressed homogenous field1 )
C2 is the dielectric constant of the sphere

At beat, such a treatment of a single particle may be valid in a

* strongly dispersed systems, where the field in the continuous com-
ponent2) is not significantly affected (changed) by the (presence of)
the spheres, In such cases, the impressed field can be set equal
to the field in the continous component and the (internal) field
in the spheres can be found from Equation 20.
I), 2) see next page

71
14 Inserted in Equation 19, this results in the following expression
for the conductivity:

(X + 2)n 12
(-n)
(2. 2X 1 )n + 34XI2 ( -n)

This equation was originally derived by Maxwell (1881). Later,

Friche (25) developed a similarly based equation valid for ellips-
oid-shaped particles. In this equation, the ratio between tempe-
rature gradients is given by:

11
a,b,c I + (A 2 /X 1)gj

where gas gbO go are parameters which depend on the (geometri-

cal) shape of the ellipsoid. (ga 1

For spherical particles : ga = 9b g 11/3

discshaped particles: C a 11 gb go a 0

needle-shaped particles: g * gb a 1/2, gc a 0

As mentioned previously, these equations are only valid for the

'I case of (widely) dispersed particles. The expression for spheri-
cal particles is also identical to the lower Hashin-Shtrikman l-
mits, while their upper limit is found by considering (widely)
dispersed pores in a continous material (X 2 c<1 )" For other geo-
metries, the equation yields results between or on these limits.

In an effort to find a more accurate approximation for cases other

than that of (widely) dispersed particles, a different formulation
of the exact expression for conductivity has been used:

1) That is, t'e dielectric constant of the (homogenous) medium

surrounding the sphere. (Translator's note)
2) The material in which spheres are imbedded or "dispersed"
(Translatorts note)

72
 1 += 2 " 1 )" (1-n) < T, 2 / ev T> 23

The ratio between temperature gradients is in this case found by

assuming that a single sphere is inserted in a homogenous field
which is equal to the macroscopic (average) field. This gives

<VT>2 / <VT> 24

When inserted in Equation 23, this results in the following

equation

n + (1-n)
+-*-.-A 0 25

This equation is usually ascribed to Bbttcher (1945)(26). Later,

Polden and van Santen (1946)(27) developed a (corresponding) ex-
pression for particles with ellipsoidal shapes.

Not even this assumption will be valid for the case of concentra-
ted particles, since the temperature field around each particle
(then) must depend on neighbouring particles.

Many different efforts have been made to solve this problem.

Bruggeman (1935)(28) used Equation 23 to determine the increase
in conductivity due to a small addition of particles in a material.
This results in a separable differential equation, which after
integration yields the following expression for the conductivity

n_ 26
nY2 '
Also in this case it is in effect assumed that the field inside
the few added particles does not depend on adjacent particles.

To include interaction between adjacent particles one must obvious-

ly form a description of the geometry of the system, including the
packing configuration. Since the latter can take many forms for

73
 a given article geometry it is evident that a general expression
for conductivity in a material containing particles cannot be
formulated. This was also shown, by means of a study of limits,
in the previous main section.

Figure 8 compares (results derived from) the three previously men-

tioned equations, formulated for spherical particles, and the Miller
limits (bounds) for spheres having a conductivity ratio of 10.

,/ BEST AVAILABLE COPY

-- .•,.|,.Ia~..,,.-f -j

As shown, the equations by B6ttcher and Bruggeman (give values

which) fall between the (Miller) limits, while Maxwell's equation
gives values below the lower limit.

All the different equations mentioned here lend themselves to

adjustments (fitting) to experimental result by choice of geometry
(-related parameters) in the equations. However, one can not ex-
pect this (best fit) particle geometry to agree with the particle
shapes in the material at hand*

If two g-values are equal, the shape of the ellipsoid(s) can be

found from the third g-value (27). For this case, Frichets
equation can only be fitted to one point on an experimental curve

74
•... ... ...* .. l +r m •m,• i • • A.
which gives the conductivity as a function of conductivity ratio
for constant porosity. The result of sucha fitting to a numeri-
cally computed functional dependence of the conductivity for shperes
in an orthorombic (packing) configuration is shown in Figure 9.

It that the fitted curves fall below the curve for

can be seen,
orthorombic (packing) configuration at high conductivity ratios.
As mentioned, the particle geometry corresponds to more elonga-
ted ellipsoids which become less like spheres as the conductivity
ratio increases.

This discussion shows that equations of this type only have theo-
retical relevance for rather low particle concentrations, where
the theoretical limits for a given particle shape coincide. In
other cases, they may give acceptable results when the conductivity
ratio is low and the theoretical limits for the particle shape at
hand define a relatively narrow range of conductivities. For cases
other than these, the equations are only relevant as semi-empirical
(mathematical) models which can be fitted to experimental conduc-
tivity curves within limited ranges of conductivity ratios and po-
rosities, as well as for a certain choice of geometrical (form)
factors.

7,+

11 IIV

5,kLt olidl4
q Ij 6t Ol

"1!4.11 I,
PIL

J * 0 ! t) , u0 ol II 0 0• p L O ., '17

!'I.• ++, a.k+'In+ t ltornI ti

. ut~rwrul tfl,¶)4mrr9n• iv Iri2 +.j +'I.ti,,P Ut' L: V•t

715 "'dl
 D. Approximate calculations based on elements
connected in series or shunt

One group of (mathematical) models is based on a conceptual sub-

division of the heat transfer mechanism in a particle-filled ma-
terial into (lumped) thermal resistance elements. Examples of
such elements are given in Refs 29 through 32. This concept can
be pictured in terms of cubical models (cells), within which the
dispersed material forms a parallelepiped (see Figure 10).

+I ..- - -- , , . .I -.. ... . ................

,-

... ~~
', . .' ~I

-,V • "r .-. I' ,. . - .,t

The most commonly used model is shown in Figure 10 C. The conduc-

tivity of such a cubical elementary (unit) cell is according to the
basic assumption, usually calculated from the thermal resistance
of the vertical liquid layer (thickness A ) connected in shunt across
the series combination if resistences associated with the horizon-
tal liquid film (thickness 0) and the particle, see Figure 10(C).
This results in the following expression for the conductivity1)

S(1
11
-0) / X+
1 ( - ) /27

It is worth noting, that connecting (a resistance representing)

the horizontal liquid film is series with a shunt combination of
(resistances representing)
ticle gives adifferent the
result 1 ) vertical liquid layer and the par-

1) See appendix to the translation

76
 / /x (0-0) / + X2 ) 28

These equations represent lower and upper bounds for the conduc-
tivityrespectively, for this particular (type of) cell. These
bounds are identical to those given by Coriell and Jackson, dis-
cussed in the previous section. In this case, the difference be-
tween limits (bounds) is relatively small, but for other configu-
rations (e.g. that of Figure 10 D), the difference may be signifi-
cant.

The choice of (equivalent circuit) representation will, however,

in such cases be of minor importance, as long as the geometrical
conditions of the cell are selected based on a semi-empirical fit-
ting to experimental values.

During such a semi-empirical fitting process one will in most

cases find it neccessary to vary the geometry along with conducti-
vity ratio and porosity, in order to obtain a satisfactory fit to
the experimentally determined conductivity dependence. This is
illustrated in Figure 11, which shows an effort to fit Eq 27 to the
numerically computed conductivity variation in a material consisting
of spheres in an orthorhombic configuration. For geometrical con-
ditions which result in agreement at conductivity ratios of 10, 100
and 1000, one finds that the agreement is poor for other conductivity
ratios, except at the low end of the scale where the three (cell)
models give practically speaking the same values.

Kv t R4, ,

0J J

:,-- • ,... ,---

. • ._ ... .. . .. . .::. V, I f.. -_...--
r, ".... .... ...". . . . . . . .

77
 A variable geometry parameter) such as 0 in Figure 10p was used
by " ii eteal. (33,34,35). These reports give the geometry para-
meter in graphical form, as function of conductivity ratio, for
two porosities. The authors recommend linear interpolation for
other values of porosity. The cell geometry is selected on the
basis of approximate calculations of conductivity in a (packing)
configuration of npheres, using the principle of parallel heat flow
lines. By considering the number of contacting points in other
types of configurations, an effort was made (by the authors) to ex-
tend these results to other values of porosity.

The results of these studies are somewhat in doubt due to a number

of inconsistencies, On the other hand, they present a single model
which gives conductivties in good agreement with numerical calcu-
lations in regular configurations of spheres.

In other report-, efforts are made to fit the geometrical parameter

to conditions in configurations of spheres by considering only the
volumetric properties of the liquid layers near the contact points
in various configurations. This results in constant values for the
geometry parameter for given porosities, independent of the conduc-
tivity ratio (36). This gives less freedom in fitting the results
computed from the (mathematical) model to conductivity variations
in a given material,

SAs evident from the preceding. discussion (mathematical) models of

S~this kind will always be semi-empirical to a cetain extent, both
in terms of choice of model type and choice of approximate computa-
tion method, as well as in the choice of geometrical parameters.
The usefulness of a given model depends primarily on how well it can
• be fitted to the functional dependence of conductivity in a given
:• material. In cases where this (fit) is accomplished by means of
geometrical parameters that depend on conductivity ratio and porosi-
~ty, they (the models'4 offer in reality no advantage over purely em-

7 8a
pirical correlations derived from experimental data for the mate-
rial at hand.

A closely related class of (mathematical) models involves efforts

to base the calculations on a geometry which to a greater extent
is similar to the microgeometry of the actual material. The micro-
geometry is then characterized in terms of statistical (stochastic)
distributions of the one or two-dimensional porosities, as discussed
in the previous section in connection with the limits (bounds) gi-
ven by Coriell and Jackson.

A model of this kind was first presented by G.T. Tsao in 1961(37).

This model contains certain fundamental errors but efforts have
been made to correct these in models of the same type (38). How-
ever, a12 these models are based on an erroneous assumption concer-
ning the relation between distribution functions for one and two-
dimensional porosities.

Tsao's model has received considerable attention in the literature.

For example, S.C. Cheng and R.I. Vachon have published a series of
papers based on this model (39, 24). They claim that the stochastic
distribution of the dispersed medium justifies the assumption that
the one-dimensional porosity is normally distributed. This normal
distribution is then approximated by a parabolic distribution.
However, the theoretical foundation for this model is unacceptable
in several respects. In effect, the model used by Cheng and Vachon
can be reduced to an idealized case where the dispersed component is
represented by a parabola whose volume corresponds to the volume ratio
of particles in the material at hand,

Even if the theoretical foundation for the model is questionable,

comparisons between calculated and experimental values for a rela-
tively large number of materials show a fairly good agreement (24).
If one compares values computed from this model with expected con-
ductivity variations in dispersed granular materials one finds that
the model yields values within the same range.

79
 The agreement with experimental data gives this model a certain
credibi±)1y as a toll for computing conductivity in granular and
dispersed materials. However, the uncertainty becomes large for
relatively small to moderate porosities, as shown in the corre-
sponding figure. In addition, the model does not allow fitting
to moderate porosities, as shown in the corresponding figure.
In addition, the model does not allow fitting to experimental func-
tion and is thus worthless as a semi-empirical relation.

The theoretical weaknesses inherent in models of this kind can

obviously be compensated for by a correct application of theore-
tical methods developed by Coriell and Jackson for determining con-
ductivity of composite materials. However, this does not imply an
improved accuracy of these models, since one will only be able to
determine the conductivity within rather wide limits in materials
with certain (given) statistical distributions for the one and two-
dimensional porosities. To even determine these statistical distri-
butions will also be quite a complicated task. It therefore seems
more realistic to attack the problem more directly by means of em-
pirical investigations of the conductivity on materials at hand,
in order to develop practical values for the range of variation, in
conductivity for different groups of materials. Such an approach
will represent an empirical parallel to the purely theoretical de-
rivation of limits (bounds) described in section 2 of this chapter.

E. The geometric mean model

Warren and Price (1961)(41) presented results from a series of nu-

R imerical experiments which are closely related to empirical deter-
mination of limits for the conductivity in certain groups of mate-
rials. Warren used a three-dimensional model in the form of a
cubic lattice where each element was given a different conduCtivity
in relation to determined distribution functions. The location of

80
individual elements in the lattice was made random. For each such
configuration, the effective conductivity was calculated by nume-
rical solution of the heat flow equation in the three-dimensional
lattice. The configura'ions were changed a sufficient number of
times to give the most probable (expected) conductivity with rea-
sonable accuracy

For all the different types of continous distribution functions

that were selected for the conductivities of the elements, it was
demonstrated that the geometric mean of these conductivities forms
a good approximation to the expected conductivity:
p ni
ff ýi29

where ni is the relative number of elements having conductivity xi

In a discontinous distribution where each element is given one of

two conductivity values (corresponding to a two-component system),
the expected conductivity was found to nearly equal the geometric
mean for low conductivity ratios ( / <10). Rowever, for higher
ratios between the two conductivities, the expected conductivity
deviated significantly from this.

Standard deviations of the calculated conductivities could in prin-

ciple have given information about expected variations in conducti-
vity for a given distribution of element conductivities. However,
this parameter was not reported for the discontinuous distributions.
In addition, the standard deviations for contirous distributions
were dependent on the selected discrete approximations for these
distributions.

The geometric mean equation has also been suggested by other authors
as a model for conductivity calculations for two component materials.

1) A Monte Carlo type simulation (Translator's note)

81 i
•••••••••••••••••••••••
 For example, W. Woodside and J.H. Messmer (1961)(30) found that
this model gave good agreement with conductivity measurement in
water-saturated sand at various porosities. Also, J.H. Sass et.al.
(1971) recommended this equation for use in connection with water
saturated crushed rock (41).

An important property of the geometric mean equation is that it al-

ways yields values which fall between the Hashin-Shtrikman limits.
For low conductivity ratios where these limits are close together,
it will thus give a good approximation for the conductivity of com-
posite materials which can be considered macroscopically homogenous
and isotropic. Since it also gives a good fot to experimental values
for granular material with moderate conductivity ratios (X /X '10),
there are obvious reasons for favouring this comparatively simple
equation over other more complicated mathematical models described
previously.

Figure 12 compares four of the previously discussed equations de-

veloped for calculating conductivity in granular materials with the
geometric mean equation, for constant porosity (n a 0.40) and con-
* ductivity ratios up to / 20. The Hashin-Shtrikman limits are
also shown in Figure 12. Evidently, the deviation from the geome-
tric mean is less than 20 percent for these equations, while the
Hashin-Shtrikman limits deviate considerably more. The three equa-
tions which most closely agree with the geometric mean (Cheng and

F
&
Vachon, de Vries and Bruggeman) have been shown by many to give
good agreement with experimental conductivities,
vity ratios (23, 31, 42).
for low conducti-

82
 i. 4641

-W

v.v• • •
: .
." tuni;.'s
'. equa iu.
0 .,. 3*S. -

-hzV: Chg.wg and Vacho~n'sa l7Ztia j ta)

NS: Upper a-ILI -lower Mash ¶-±,r!h.,..a, ~ (7:

Fla.
. 2. S . 1-r..Ing melllmP maocellcr r'or ,-. *r.';
er ¢nir.. av 'drinlrevv'fl

.,e . fa'. ,Jt;t :s LC'.? uf ' *

F. Conclusion

This discussion of different types of models for calculating the

conductivity in composite materials shows that such models in ge-
neral tend to consist of semi-empirical equations which are only
applicable to the extent that they can be fitted to experimental
results for a given group of materials*

For low conductivity ratios (<10), where the different models give
almost identical values, it seems logical to prefer the simple
geometric mean model as a basis for conductivity estimateso

For large conductivity ratios, the conductivity of a certain

group of materials will always show relatively wide variations.

83 Best Available Copy

 In this case, any given model will give large uncertainties. For
this reason it seems more fruitful to work towards generating an
empirical data base for estimated variations in conductivity in
different groups of materials. Such investigations should re-
sult in a form of empirical parallel to the purely theoretical
limit considerations,

"li4

3'

i8
 REFERENCES -. CHAPTER II

I. K. H. Walter; Zur matematischen Behandlung der station'ren

Warmeleitung in stackweise homogenen Medien, Kernforschungs-
anlage JClich, JUI-657-NA, April 1970.
2. W. F. Brown: Solid Mixture Permittivities. The Journal of
Chemical Physics, 23, (8), 1955, pp. 1514-1517.
3. W.E.A. DAvies: The theory of composite dielectrics. J. Phys. D:
Appl. Phys. 4, 1971, pp. 318-327.
4. M. J. Beran: Application of Statistical Theori:s to Heterogeneous
Materials. Phys. Stat. Sol. (a), 6, 1971, p. 365.
5. W. Ft Brown, Jr.: Dielectric Constants, Permeabilities, and
Conductivities of Random Media, Trans. of the Soc. of Rheology 9,
(1), 1965, pp. 357-380.
6. 0. Wiener (1912): Referert av W.F. Brown (1965) See ref. 5.
7. Z. Hashin and S. Shtrikman: A Variational Approach to the
Theory of the Effective Magnetic Permeability of Multiphase
Materials, J. Appl. Phys., 33, (10), 1962, pp. 3125-3131.
8. M, Beran: Use of the Vibrational Approach to Determine
Bounds for the Effective Permittivity in Random Media. Nuovo
Cimente, 38, (2), 1965, pp. 771-782.
9. P. B. Corson: Correlation functions for predicting properties
of heterogeneous materials. 1. Experimental measurement of
spatial correlation functions in multiphase solids. J. Appl.
Phys., 45, (7), 1974, pp. 3159-3164.
10. P. B. Corsons Correlation functions .... II. Empirical construc-
tion of spatial correlation functions for two-phase solids. J.
Appl. Phys. 45, (7) 1974, pp. 3165-3170.
11. P. B, Corson: Correlation functions .... IV. Effective thermal
conductivity of two-phase solids. J. Appl. Phys., 45, (7),
1974, pp. 3180-3182.
12. M. N. Miller: Bounds for Effective Electrical, Thermal, and
Magnetic Properties of Heterogeneous Materials. J. Math Physics,
10, (11), 1969, pp. 1988-2004.
13. M. A. Elsayedi Bounds for effective thermal, electrical, and
magnetic properties of heterogeneous materials using higher
order statistical information. J. Math. Physics, 15, (11), 1974,
pp. 2001-2015.

85
14. J. L. Jackson, S. R. Coriell: Transport Coefficients of
Composite Materials. J. Appl. Phys. 39, (5), 1968, pp. 2349-2354.
15. S. R. Coriall and J. L. Jackson: Bounds on Transport Coeffi-
cients of Two-Phase Materials. J. Appl. Phys. 39, (10),
1968, pp. 4733-4736.
16. S. Prager: Improved Variationa'l Bounds on Some Bulk Properties
of a Two-Phase Random Medium. J. Chem. Phys. 50, (10),
1969, pp. 4305-4312.
17. Lord Rayleigh (1892). Referent av Meredith (1969) (18).
18. R. Z. Meredith and Array
C.W. of
Tobias: Resistance
through a Cubca- Spheres. J. Appl.to Phys.
Potential Flow
31, (7),
pp. 1270-1273.
19. D.A. Do Vries: Hot warmtegeleidingsvermogen van grond. Med.
Landbouwhogeschool. Wageningen, 52, 1952.
20. G.K.H. Walther: Zgr makehemAtischen Behandlung der Stationaren
Waireleitung in stuckwelso homogenen Median. aernforschungs-
anlage Juilich, 657-MA, 1970.
21. R.S. Deissler, J.S. Boegli: An Investigation of Effective
Thermal Conductivities of Powders in Various Gases. Transac-
tions of the ASME, October 1958, pp. 1417-1425.
22. N. Wakao, K. Kate: Effective Thermal Conductivities of Packed
Beds. Journal of Chemical Engineering of Japan, 2, (1),
1969, pp. 24-33.

23. T.P. Fidelle, R.S. Kirk: A study of Unidirectional Versus

Tridirectional Heat Flux Models. AlChE Journal, 17, (6), 1971,
pp. 1427-1434.
24. S.C. Chong, R.I. Vachon: A Technique for Predicting the Thermal
Conductivity of Suspensions, Emolusions and Porous Materials.
Int. J. Heat Mass Transfer, 13, 1970, pp. 537-546.
25. J.A. Reynolds, J.M. Hough: Formulae for Dielectric Constant
of Mixtures. Proc. Phys. Soc. (London) B70, 1957, Dp. 769-775.
(Referert H. Fricke (1924)
26. C.J.F. Bottcher (1938): Referet av D. Polder (1946). See ref.
(36).
27. D. Polder, J.H. van Santen: The Effective PNrmeability of
Mixture of Solids. Physmca, 12, (5), 1946, pp. 257-271.
28. D.A.G. Bruggemani Dielektrizititskonstanten und Leitf~higkeiten
von Vielkristallen der nichtreguliren Systems, Ann. Physik 5,
(25), 1936, pp. 645-664.

86
29. W.O. SMith: The Thermal Conductivity of Dry Soil. Soil
Science, 53, 1942, pp. 435-459.
30. W. Woodside, J.H. Messmer: Thermal Conductivity of Porous
Media. I. Unconsolidated Sand., J. Appl. Phys,, 32, (9),
1961, pp. 1688-1699.
31. R. Mc Caw: Heat Conduction in SAturated Granular Materials.
Highway Res. Board., Spec. Rep. no. 103, 1969, pp. 114-131.
32. G.S.G. Beveridge, D.P. Haughey: Axial Heat Transfer in
Packed Beds. Stagnant Beds Between 20 and 7500 C. Int. J.
Heat Mass, 14, 1971, pp. 1093-1113.
33. S. Yagi, D. Kunii: Studies on Effective Thermal Conductivities
in Packed Beds. A.I. Ch. E. Journal 3, (3), 1957, pp. 373-381.
34. D. Kunii, J.M. Smith: Heat Transfer Characteristics of Porous
Rocks. A.I. Ch. E. Journal, 6, (1), 1969, pp. 71-78.
35. K. Ofuchi, D. Kuniii Heat-Transfer Characteristics of
Packed Beds with Stagnant Fluids. Int. J. Heat Mass Transfer,
8, 1965, pp. 749-757.
36. S. Masamune, J.M. Smith: Thermal Conductivity of Beds of
Spherical Particles. Ind. and Engl. Chem. Fundamentals, 2,
1963, pp. 136-143.
37. G.T. Tsao: Thermal Conductivity of Two-Phase Materials. Ind.
Eng. Chem. 53, (5), 1961, pp. 395-397.
38. P.B. Desphande, J.R. Cooper: Thermal Conductivity of Two-Phase
Systems. J. Heat Transfer. Trans. of the ASME, May 1972, pp.
249-264.
39. S.C. Cheng, R.I, VAchon: The Prediction of the Thermal Conduc-
tivity of Two and Three Phase Solid Heterogeneous Mixtures,
Int. J. Heat Mass Transfer, 12, 1969, pp. 249-264.
40. J.E. Warren, H.S. Price: Flow in Heterogeneous Porous Media.
Society of Petroleum Engineers Journal, Sept. 1961, pp. 153-169.
41. J. H. Sass et al: Thermal Conductivity of Rocks from Measure-
ments on Frarents. J. Geophysical Research, 76, 014), 1971,
pp. 3391-3401.
42. K, Gotoht Thermal Conductivity of Two-Phase Heterogeneous
Substances. Int. J. Heat Mass Transfer, 14, 1971, pp. 645-646.

87
 CHAPTER III

Soil parameters

The previous analysis of heat transport mechanisms in soil mate-

rials showed that heat transfer in soils can be regarded as a pure
thermal conduction process controlled by temperature gradients.
During the discussion of methods for calculating thermal conducti-
vity in composite materials it was shown that the conductivity
in such a material will depend on three sets of parameters: the
conductivity of each component; their relative volumes and the micro-
geometry of the system. These sets of parameters are here collec-
tively denoted thermal conductivity parameters.

In a soil material, these thermal conductivity parameters are more

or less completely described (defined) by the mineralogical and
mechanical composition of the material, as well as information
about dry specific weight, water content and temperature level
(frozen, un-frozen). The different material characteristics from
which the fundamental parameters can be derived will be discussed
in the following.

1. THERMAL CONDUCTIVITY OF SOIL GENERATING

MATERIALS

As Soil generating minerals

The mineralogical composition of soil maaterials is the result of

processes such as weatherin, transportp sedimentation and possibly
chemical withering of solid rock. These processes contribute to
giving the soil a different average composition than that of solid
rock. For example, loes crush-resistant minerals such as feldspar
will be crushed at a more rapid rate during transport. This re-
sults in an enrichment of more resistant minerals in coarser depo-

8'
sits (sand, coarse silt). In addition, finely crushed feldspar
is particularly prone to chemical withering, which gives rise to
new types of minerals. These minerals will occur in the form of
rather small flakes, rich in Aluminum, while the content of the
more easily soluble elements Na and Ca is reduced. Hydromica
is such a residual product while different types of clay (kaolinite,
montmorrillonite) represent an even more complete change during
which all Na or Ca has been lost (1).

These processes will thus result in differences in average minera-

logical composition for loose deposits and bedrock, as well as
certain differences in the mineralogical composition due to par-
ticle size in the soil materials. These conditions are illustra-
ted in Table I and in Figure 1, respectively (2,3).

Table I Comparison between average mineralogical composition

in volcanic rocks and sediments (2).

Mineral Average, volcanic rocks Average, sediments

quarts 12 35
Feldspar 59.5 16
Mica 3.8 15
Miscellaneou 24.7 34

I WA

W".4 %A

89
In Norway, most older loose deposits were removed during the ice
age. Present soils were primarily deposited during the last mel-
ting period, between 8,500 and 12,0OO years ago, or in later times.
Thus, the previously mentioned (chemical) changes in the most fine
grain materials have had a short time in which to occur and Nor-
wegian clays in general contain small amounts of the typical clay
minerals, except for hydromica.

For these reasons, Norwegian soils primarily consist of minerals

normally occurring in the bedrock. Of these, three groups are in
general dominant in Norway: Quarts, feldspar and mica. Together,
these form more than 90 percent of common rocks such as granite and
gneiss. In addition, these rock materials dominate the bedrock in
large parts of Norway, except for streaks of sediments and a few
streaks where quarts-free rocks such as gabbro and amphibolite
are predominant. These rocks consist mostly of feldspar (plagioclase)
in addition to pyroxene (gabbro) or amphibolite. Such quarts-free
rocks can be found in the Cambrosilurian field stretching from
Bergen over Jotunheim1) in to southern Tr~ndelag2). (See map in
Figure 2)3) Large regions of predominantly gabbro can also be
2) 2) The
found in the Cambrosilurian field in Troms 2 ) and Finnmark )
Telemark2) formation also contains significant amounts of the same
"rock type.

Depending on ice movements and transport in rivers during the ice

age, these rock types may have given rise to significant contents
of minerals such an pyroxene and amphibolite also in soils where
such materials occur to a very limited extent in the bedrock. This
in general means that the mineralogical composition of loose depo-
sits will deviate from the mineralogical composition of the local
bedrock.

1) Mountain formation (Translator's note)

2) These names refer to "Fylken", county-type districts (Transla-
torts note).
3) Not present in the original (Translator's note)

90
The most common minerals occurring in the loose deposits will
thus, as indicatedp be quartz, feldspar, mica1 amphibolite and
pyroxene. The firstthree of these will normally dominate, while
in limited areas amphibolite and pyroxene may occur in signifi-
cant amounts, together with feldspar.

B. Quartz content in loose deposits

Of the minerals mentioned previously, quartz plays a particularly

important role in this context (as will be evident from next sec-
tion), due to its relatively high conductivity. This led to a
study by H. Sveian of the quartz content in lTorwegian loose depo-
sits (5, 6). During this study, the quartz content was measured
in a total of 131 samples from different types of deposits and
bedrock regions, by means of differential thermal analysis (DTA).
In some samples, the quartz content as function of particle size
was also investigated. This study supplements in many respects
the investigation made by Selmer-Olsen of the mineralogy of Nor-
wegian clays- (7), which includrd determination of quartz content
in about 160 clay samples from tho ?ntire country by means of DTA
(see Figure 1).

Table II shows the distribution of quartz content in the samples

investigated by Sveian. The results are arranged according to bed-
rock region. As can be seen, there are rather wide variations.
Together, the results indicate an average quartz cont3nt of 37.5
percent. The lowest quartz contents are found in samples for the
Bergen region and areas having Caledonian intrusives, as well as
in some samples from the Cambrosilurian field. This indicates, that
the lowest quartz content in general occurs fo: high incidence of
gabbro or other alkaline types of rock.

For a small number of selected samples, the quartz content was de-
termined for several fractions (by particle oize) of the samples.
Based on tnese experiments, one can plot quartz content versus

91
 0 0 0 0 C) 00O 0

0 0

o - tt a - M. C.~--t p u-O0 0 0 Q u
y4~
u to

Id 0

(Oi

u0

* 0 0)
hotot\N~O 00000000o 000010 9

-0 0

8 14
4
$4 0

tho I U

0 04 0 -t 0 4-

.0 5 k .

00 0 t
(00

00

(0
6 * 0(0 cc H .w 1

4J (0 0 0
N N I, co0

92~ 4
grain size for these materials. These curves are shown in Figure
3. Most of these samples were sand or gravel materials with low
content of silt or finer fractions. The results from these samples
are drawn as bold solid lines, while results from samples of moraine
and lake-shore sediments having a more even particle size distri-
bution are shown as broken lines. The thin solid lines define the
range of quartz content variations in more fine grained materials
found by Selmer and Olsen (3).
SIT ,6!! SAND

I!ý I

*OEN IVtOADR
40/ . ....
.
i ./ . .- ** /"**•, -- '- '" *- .•-•""- -

%, / • ,.,,: -"- '-•,.--',m, w

// -0',..., . . -•r:
_ . . . .. __ I.

0 0.002 0.006 0.02 0.06 0.2. mm 0.6 2.0 6.0

Jr.,
IL,*-d'L ;s, :'L n... I : , &f" * +,uVp,..j
L l? 1 . 'j• (S'a, ' ;j|I'j/• (a.'

As can be seen, all samples are characterized by a more or less

pronounced maximum in the quartz content. For the different sand
and gravel materials, the maximum occurs for particle sizes in
the range fine to average sand, while maxima fall somewhat lower
(coarse silt to fine sand) for samples having more even size di-
stributions.

All told, this study shows that the quartz content in Norwegian
loose deposits can vary between wide limits. These limits de-
pend partially on the particle size distribution within the ma-
terials and partially on variations in mineralogical composition
of the original (rock) material, The first of these "mechanisms"

93
 will cause fine grain materials with relatively high content of
materials finer than average silt (d< 0.02 mm) to, in general,
have lower quartz contents than coarser materials where silt and
finer fractions are scarce (d>0.06 mm). In regions where the
bed-rock contains little quartz, the second mechanism will come
into play and cause low quartz content for all fractions.

Fraction Average quartz Range of vari-

mm content, percent ation, percent
d>0.02 45 20- 60
0,002 < d < 0,02 15 0- 40
d< 0.002 0 0- 10

Figure 4 shows a comparison between measured and calculated quartz

contents (for the samples where particle size distributios are re-
ported), based on this "model". With a few obvious exceptions,
most of the results fall within +30 percent of calculated values.

*, / ,,-/
, 'o rV. / .. ,,
i , /

,• ; // /",
.a
20

1' 0i/ o it-:

Vi ' C' .1P LJ tY'I 101'

i 0 'h li .
41A111
I.
O-
1
1,/
'.N
.2 *I
r.IAt, t
f"A'. 'I.E ' . 1 1 I. '

ii /,Vri
!IVm. f . bV1' W- 1 '.
rra ,.u i,:,c ji :
l.'e .
l 4YJ,:ýQI 0:, J., 14t,ýJ fph aP ~j'P V

If higher accuracy is desired, one may perform direct measurements

00 I0 :]. 1,0 5T1-
of the quartz contents

94
 C. Thermal conductivity for the components

Measurement of the thermal conductivity made by Birch and Clarkes

(1940)for a number of rock materials over the temperature range 0
to 400C show that the conductivity in most materials decreases
with increasing temperature. Their results for pure quartz, cer-
tain granites and a few gabbro-type rocks are shown in Figure 5.
The horizontal scale is 1/T and the parameter is t a T - 273.
It is evident, that this representation results in a linear de-
pendence for the conductivity in most rock materials (8).

Of the different materials, quartz has both the hianest conductivity

and the largest variation with temperature, while the gabbro mine-
rals have the lowest conductivity and show small or no temperature
dependence. One should also note the large difference in quartz
for directions parallel and perpendicular to the optical axis.
When quartz occurs as mineral particles in a rock or soil material,
the optical axis$ will have random directions. This results in
an average conductivity for quartz of about 7.7 W/mK at 20*C (90)
and the variation with temperature shown as a thin line in Figure

N OR
lGTA In..i.

ii

-50 0 SC 100 2(1 C 400

TEMPE.IATV9

1) Probably refers to the line marked (Translator's note)

95
 Figure 6 shows how the conductivity of some other soil materials
vary with temperature. The contribution from the temperature do-
pendent vapor diffusion is included (aee chapter I). Little data
exist about the conductivity of organic materials which may be
present in the soil (plant residue, humus). However, Watzinger
(10) gave a value of 0.45 W/m°C for the solid organic components
in peat or bog* The temperature dependence is not known.
i,. ~~~W/mK<.
wimxg

tO

0,0 2
T 1,0.00421 0C.

TINP,'AT'.R4

AV

(.l'!The results of Figures 5 and 6 show that the temp~erature dependence

•r of moist soil materials is quite small within the normal range of
'•i• Boil temperatures. However, one important exception is the contri-
•; bution to conduction via pore air by the temperature dependent
vapor diffusion, which in reality is the most important factor in
it temperature dependence of conductivity in moist soil materials
(See Figure 11 of chapter I). The transition from frozen to un-
frozen material and the effects of variations in relative unfro-
zen water content will be discussed in the next section.

From Figure 5 it is evident that the conductivity of rock materials

relativelyo wide
can vary temertue
over epndnc ranges fori a mis
cndctiit temperature.
given silma96al This
variation, which depends in the mineralogical composition of the
 rocks, is much more significant than the temperature dependence
within the normal (actual) range of temperature for the different
types of rock. A very comprehensive study of thermal conductivity
of rock minerals was presented by Sorai in 1971 (11). Th.• measure-
ments were made at 250C, using a heat conduction probe in crushed
and water saturated powders consisting of the various minerals.
Horai calculated conductivity for the mineral particles from the
arithmetric mean of upper and lower Hashin-Shtrikman limits.
Sass (1971) has performed similar measurements of conductivity for
rock fragments and compared those results with measurements in the
original rocks1) under stationary conditions. He claims, that
equations based on geometric means give a more accurate base for
calculations than the average of the Hashin-Shtrikman limits.
However, the difference is no more than about 10 percent, on the
average, with the highest values given by the geometric mean equa-
tions (12).

K
:Table III shows the distribution of results for the most important
* rock-generating minerals. As can be seen, the variation is rela-
tively large for the same minerals However, both feldspar and mica
show relatively small variations around a common mean of about
2.0 W/mOC. For pyroxene-amphibolite the variations are more sig-
nificant, with means of about 4.4 and 3.5 W/moC, respectively.
quartz has a notably higher conductivity of 7.7 WV/mC. The latter
result is also in agreement with measurements by Birch and Clark
on pure quartz, provided that the following rule (formula) is used
for computing the conductivity of randomly oriented crystals

Variations for each material can be ascribed to differences of

composition within each type of mineral. Thus, a more accurate
determination of conductivity than the averages indicated here will
require detailed geological investigations. However, the varia-
1) That isethe rocks from which the fragments were made (Transla-
tor's note)

"97
 tions are lowest for the most common minerals. Thus, the errors
in conductivity estimates for a rock material, based on these
averages, should normally be relatively small.

Birch and Clark specified the mineralogical composition for most

of the eruptive (volcanic) and metamort rock types they studied.
Table IV summarizes these data and gives the conductivity for
each type at 25"C.

.98

9I.
 t4

C)

~~~ 40''
'0v.0 0 0 00 0

O0 N 0 4N N N 0000000
. I.l

5.4*4),

Nu~l,4U 0 0 00 00 00 0 00 0

e4~~ VA ANOA~% A~OA A %Z

A 0 X's N UI a% N UIA I ON E% I N

99
 'pý-N N 4 N N - ý N N U -~

4- -
(1

U 04

0 UN \UND Lý

N 00
q4 ~' 0 .

0o a*

oc

00

.44J

4- 4

100
W~4

41

104

1.101
 Among other things, it is evident that rocks containing quartz
have small or no quantities of pyroxene, amphibolite or olivine
These can thus be approximately represented as a two component
system consisting of quartz and feldspar - mica with conductivi-
ties 7.7 W/mK and 2.0 W/mK (feldspar, etc.), respectively. The
conductivity of such two component system can be determined ap-
proximately from the geometric average equation

• • - ~,7.7 qq 2.01-
-q
7 7 32 0

where q is the relative quartz content.

Figure 7 shows results obtained by Birch and Clark for rock ma-
terials containing quartz, plotted versus the relative quartz con-
tent. The values obtained from Eq 2are drawn in for comparison.
The data indicate that knowledge of the quartz content will make
it possible to estimate the conductivity of quartz-bearing rocks
with errors not exceeding 20 percent.
j10

C''h
2C el

•, '• ,,,i T I INN

F1U 74 ) v-n i. , , :.-j 1.z

,-IiIW ."l V11.;6 1!1'.t' ll

102
A corresponding mathematical model could be used for eruptive
rocks not containing quartz. One component would be feldspar-
mica (conductivity 2.0 W/mK) while the other would be pyroxene,
amphibolite, etc., (with a conductivity of 4.0 W/mK). However,
the large variation in conductivity within the latter group will
result in relatively large uncertainties.

In loose deposits for which the quartz content is known, Eq 2

can be used for estimating the particle conductivity. In case
minerals such as pyroxene, amphibolite, etc. occur, some uncer-
tainties will enter the picture.

For loose deposits where the quartz content is un-known, one may
use the results obtained by Sveian to estimate quartz content
based on the grain size distribution in the material. In the pre-
vious section it was suggested that the following values may be
used

d <0.002 mm 0 percent
0.002,d 0.02 mm 15 percent
d 0.02 mm 45 percent

This forma the basis for sketching a triangular aiagram of the

form shown in Figure 8, which can be used to estimate quartz con-
tent when the grain size distribution is known.

PFl • 40 2,,
"0 -0

Yto, . Truitantuiagr-t rot, AnxL~lrUVI ballteMMUe160 fv kVntDr'.tm~h0.' I

0 103
As shown in Figure 4 of the previous chapter, this mathematical
model normally gives uncertainites for the quartz content which
are less than 30 percent. This means a some-what higher uncer-
tainty (about 40 percent) in the estimated particle conductivity,
when calculated as a weighted geometric mean o-f conductivities for
quartz and feldspar/mica. To this one must add the uncertainty
due to any occurance of minerals such as pyroxene, amphibolite,
etc. Overall, this method will therefore give an uncertainty in
the particle conductivity of up towards 50 percent of the esti-
mated value, ise. twice that which seems possible to obtain when
the quartz content is known.

2. VOLUMETRIC COMPOSITION AND TEXTURE

A, Volume ratios in soil materials

The volumetric composition of a soil material is usually specified

in terms of dry specific weight Yd and the ratio between water and
dry material contents, by weight, w. When using the MKSA system
(of units) it is more correct to use dry density d and correspon-
ding density conditions also for the water contenL. Ho7ever, to
avoid misunderatanding, the conventional concepts will be used in the
following.

These two parameters do not give a complete definition of volumetric

conditions in a soil material, The specific weight ys (=density) of
the mineral particles will also affect these conditions. For example,
the relative (share of the) volume for particles in a soil material
is given by
1-i Ya/ye 3
However, the specific weight of mineral particles will vary within
relatively narrow limits, with a mean of about 2700 kg/m 3 for

1) Except for a possible difference in units, the distinction

between Yd and Pd is unclear (Translator's note)

104
 the sand fraction and about 2800 kg/m 3 for the finest fractions
where the content of mica and heavier minerals usually is larger
(see Table V). For organic materials (plant residues, etc.), Y8
will vary between 1460 and 1: Ig/m 3 (13). In soils containing
both organic materials and minerals, the volume ratio between the
two types of material is given by the average specific weight
of solid components. Thus, the relative volume (share) of organic
miterials is given by
Y Ys 7
0v • Yd Y " Y0

0V is the relative volume of organic materials

Y is dry density (kg/r 3 )
Y8, Y•, Y. are specific weights for solid components,
mineral components and organic components,
reapectively

Skaven Haug (1972) discussed different methods for determiniiug

Y and comments on the heat loss method 1 ) (13), &mong uthers.
Y is determined from the ash oontent (ash/dry material) after
'heating the soil material. to a state of qlowing.

In the following, we will primarily discuss pure mineral soil ma-

terials. However, in the next chapter, thermal coductlvity
measurements on peat and bark will be discussed, along with *Aeasure-
ments on soils containing •inerals only.

The dcgree of saturation in a soil material, Sr, can be determined

from information about dry density Yd' specific weight of the solid
component, y., as well as the water content by weight, w. 5 r re-
presents the ratio between ice or water volume and the total volume
cf the porer,,

1) A literal translation which to not very especific. Hopefully,

the following sentonce Tives a better idea cf what is in.
volved (Tranmlato's note)

LOS
S•.-L-
 w Yd Ys 5

As mentioned previously, dry density •d and relative water content

by weight w are the only parameters normally specified. The un-
certainty about degree of saturation will in such cases depend on
the uncertainty in estimating the specific weight of the mineral
particles. That relation is given by

ASr Yd A6s
r Ys '*Yd Y's
where S and LY are the relative errors in Sr and Ys, respective-
ly. Figure 9 presents this relation in graphical form.

.,*' /

S"V 'K VCIV "'- n".aTlI- 3-t aC .. ,

.•a. t . • a-.U 1 - ' ,r'. o., rC* v I:'It I

.a.

For the most tightly packed materials, the relative error in de-
termining the degree of saturation can thus be 3 to 4 times greater
than the relative uncertainty in specific weight of the mineral
particles. If the latter uncertainty is assumed to be about _5
percent (corresponding to Y. 2700 + 150) the error in Sr can thus
amount to 15 - 20 percent in tightly packed gravel materials. The
corresponding error in determining the relative particle volume is
+5 percent i.e. the same as the specific weight uncertainty.

106
Higher accuracy in determining volume ratios in soil materials
requires knowledge about the specific weight of the mineral par-
ticles. This is normally obtained by means of a pyknometer (14).

To describe the volume ratios in a frozen soil material one must

also know the ratio between frozen and un-frozen water. This to-
pic will be discussed in section C.

B. The importance of texture

The texture of soil materials is often related to the grain size

diatribution but will clearly also be affected by the particle
shape. Together, these factors will affect the packing geometry
and the soil material density. A gravel material with widely di-
stributed particle sizes can thus be packed to a considerably
higher density than, for example, a fine sand with particles of
equal size. Figure 10 indicates the relation between particle
size distribution and maximum packing density (15).

,25 I

> 1.

1,5 I1,9

1 10, ,,, 10 , '0' lot

•'.d6 d jQdo

PI10 10. Aaktkim', paknir1Is? 5o:A r.n/jon av JAtr.lKOk.rri i *,@nd

107

-. . , , , - . ,- - - ,- ." ,.- ,• ,• ,._"7:.. '-%' '. .. "" • "'

However, in more fine grained materials there will exist a finely
porous system with a very large specific surface area. Large amounts
of water can be bound to this area, which gives these types of soils
special properties, both in terms of packing configuration and in
other respects (e.g. un-frozen water content). In addition, one
can expect that a certain water content in a coarse soil material
will mean different microgeometric conditions than the same water
content in a more fine grain' soil where larger amounts of the water
is bound to the surface of the particles. Such conditions indicate
that specific surface area may be at least as important a parameter
as particle size distribution.

The particle size distribution is normally determined by screening

of the coarser fractions (d> 0.06 mm) and by means of sludge ana-
lysis (hydrometer analysis) for more fine grain materials (16).
Both methods give uncertainties in the particle size distribution
due to variations in particle shapes. Particle size distributions
are normally represented as cumulative distribution curves in dia-
grams having a logarithmic scale for the particle size. These par-
ticle (size) distribution curves are often used for classification
of soil materials. For example, concepts such as gravel, sand,
silt and clay refer to certain well defined intervals of the grain
bIze scale in these diagrams (see Figure 11). Unfortunately, these
concepts are not uniquely defined, but Figure 11 shows the defini-
tions most common in Norway.

From curves of this type one can determine such parameters as the
size distribution coefficient Cu a d 6 0 /do 1) or the sorting number
S d7 5 /d 2 5 , which together with the median Md a d50 can be used
to obtain a numerical approximation to the particle size distribu-
tion or to represent this distribution as one point in a diagram
(17). This can also be achieved by separation into three fractions
and plotting the result in a triangular diagram as indicated in
Figure 8.
1) The subscripts for d indicate the percentage of particles which
are smaller than the diameter in question. Thus, 60 percent
of the particles have a diameterid 6 0 , etc. (Translator's note)

108
 $AND

lo00 1 o I jo 11. 4 a jiV )/4' wI

40,1 - -. q-.--j , . 4 4 ; . i. j

. 1.
I s

. .. ,......
...... ,," ..... . ', ,•
**, A -- u II4,*~ .,w

PI. .461iAJ u.1jQ ,JaJbrtk~t.'i LU rlkdr'i

JtIJ *i . SoilCO

* Legend for Figure 11:

Passert i % Passed (through screen), percent

SLeir Clayr
?in Fine
Middels Medium
cGrov Coarse
Grua Gravel
Stein Rock
U.S. Standard sikt US Standard screen (measure)
Best p! i % Residue on (screen), percent
Profil nr. Profile No
Dybde Depth
Kurve Curve
Betegnelse Notation
* Telegr. Degree of frost susceptibility

109
 In connection with thermal properties of soil materials it may
seem as if the particle size distribution curve plays a secondary
role as "parameter" for the more central factor of specific sur-
face area. For a certain particle shape (e.g. spherical), the
ratio between surface area and volume (specific surface area) is
inversely proportional to the particle size.
1...*. .-.. . ........
....

"S f/d 7

where f depends on the particle shape. For spheres, f * 6 and

for discs or cylinders one obtains the following expressions for
the ratio between surface area and volume.

Discs 1 (2 d/t

Cylinders sa'

The specific surface area is clearly strongly dependent on the

Sparticle shape. Fqr example, if the ratio diameter/thickness is
* d/t a 10 the factor f will be 24 for a disc, but 6 for a sphere
with the same diameter. If the d/t -ratio is 100, the factor f
will be as high as 204!

For a continous variation in particle size one finds that the

specific surface area per unit particle volume is strongly depen-
dent in the size distribution. As an example, if the cumulative
distribution function is linear between a smallest diameter d, and
a largest diameter d2, the specific surface area is given by the
following expression
~ d
d dln 2 (linear distribution) 8

110
while a logarithmic distribution gives
f1/

In e~•I( l/dI - I/d 2 ) 9

(logarithmic distribution)
Based on these expressions one may define an equivalent diameter
deqv i.e. the particle diameter which alone will result in the
same specific surface area as the given size distribution. For
the two types of distributions one obtains
d2.f di 1)
linear distribution deqv " i /1 1) 10
In d 2/dl
logarithmic distribution d 2 1 11
eqv T7d, 7/d
Figure 12 shows the ratio deqv/dI as function of d./d, for the
two distributions.

-ev di

! •%, • .1N5R C.'"lOL'NO

UI 1

5%

'3 I I I I , I , I ~ ~ i -I - "' -=P" - -- I",=

1) The original states d2 - d2 , which must be a "typo,' (Transla-

tar's note)
2) Note that £ different function is shown in the figure (Trans.-

111
"ini n li i" - .... .
 The logarithmic distribution will clearly always give lower
equivalent diameters than the linear distribution. For wide
size variations one also finds that the equivalent diameter is
close to the minimum diameter dI,

* Selmer-Olsen (1961) used an electron microscope to study the par-

ticle shapes in six clay materials and found that between 25 and
65 percent of particles smaller than 2 um had a width/thickness
ratio larger than 5 (18). There is some evidence that also the
size fraction 2 <d <20 um contains considerable numbers of thin
particles. Selmer-Olsen (18) estimated specific surface areas
for the six clay materials (d <2=m) from 16 m2 /g to 43 m2/g. In
terms of surface area per unit volume this corresponds to the
range from 4.50 107 to 1.2 e 108 m2 /m 3 . The equivalent sphere
diameter in 0.05 to 0.15 m. Sludge analysis in effect also implies
determination of the equivalent spherical particle diameter. How
these compare to results obtained from specific surface area con-
siderations has not been investigated*

At any rate, these results indicate that the specific surface area
of a soil material is largely detemined by the content of finer
fractions, with the clay content being the most important parameter.

Within the past few years, a method for determining specific sur-
face area has been developed based on adsorption of a monomolecu-
lar layer of ethylene glycol (19) or ethylene glycol and mono-
ethyl ether (20). These methods have been used by Anderson and
Tics (1972) in connection with a study of un-frozen water in soil
materials (21). The results from this study will be discussed
in the next section.

112 I
 C. Un-frozen water

The water in afine grain soil material with a water content be-
low full saturation will freeze at a temperature below the normal
freezing point for water. The decrease in freezing point is often
related to the moisture potential in the soil material at the ac-
tual degree of saturation. For example, Williams (19?3)(22)
gives the following relation
vT

6T * Eo12
&T is the decrease in freezing point (°C)
To is absolute temperature at the normal freezing point (K)
L is the latent heat of water (J/kg)
v is the specific volume of water (m3 /kg)
2
Sis the moisture potential (N/m )

Insersion of numerical values with ' expressed in bar (105 N/mr)

given

AT a 0.0815 * (bar) 13

Low, Anderson and Hoekstra (1968) point out that this equation is
valid only for soil materials with relatively high degrees of
saturation (23). Using thermodynamic theory they formulated a
more general relation between freezing point decrease, moisture
potential and latent heat. This complicated expression for AT
was also used to generate an extensive table from which either
of the parameters mentioned can be found when the other two are
known.

This relation between water content and freezing point decrease

in a soil material corresponds to the relation between un-frozen
water content and temperature in the material, e.g. when a sa-
turated sample freezes (24)o This makes it possible to deter-

113.3
mine the content of un-frozen water as function of temperature when
performing freezing experiments on samples with different water
content (degree of saturation). For relatively low water contents,
this method can be expected to give low accuracy. For this reason
calorimetric methods are mostly used to-day.

Anderson and Tice used such a method to determine un-frozen water

content in 11 different soils and mineral powders. The specific
surface area (determined by means of the ethylene glycol method)
varied from 6 to 800 ma/g for these materials. In general, the re-
sults could be fitted to an exponential function of the form (21)

wu ( AT )8 (percent by weight) 14

where wu is the relative amount of un-frozen water (by weight)

at are empirical parameters which depend on the specific
surface area
AT is temperature below the normal freezing point

In a 0.552 in S + 0.262 15

in (- 8) a 0.264 in S * 0.371 16

Since specific surface area is aparameter which is seldom known

for a given soil material, the same authors made an effort to re-
late their measured data on un-frozen water content to the flow (liquid)
limit for the soil materials. This parameter, which is a more com-
mon geotechnical concept# gives a measure of the consistence limit
when water is added, i.e. a limit for the water content where the
soil materials Rfter stirring are no longer plastic but "flowing".
This limit was earlier most often determined by means of Cassa-
grande's flow-limit test-set (25)p which also was used in the study
mentioned previously. However, a technique more widely used in
Norway today is the cone sinking method, whereby the consistence
limit i. defined by the water content which allows a 60 degree cone

I1
114
weighing 60 g to sink 10 mm (sometimes 5 mm) into the sample
(v 10 , v 5 ). Selmer Olsen (1954) show that the water content de-
fined in this manner agrees very well with previous definitions
of the flow limit (26).

Anderson, Tice and Banin found a linear correlation between the

flow limit at 25 blows 1 ) in the Cassegrande test-set (w 2 5 ) and the
un-frozen water content at -10C. In addition, they found a linear
relation between the flow limit after 100 blows (w 1 0O) and the un-
frozen water content at -20C. The latter consistence limit (w1 0 )
seems to approxmately equal V5 obtained in cone sinking experi-
ments while, as mentioned, wa 5 seems to agree well with v 1 0 . The
two relations are (27)

Wula 0.346 w2 5 - 3.0 17

w100 - 3.72 18
u2 R 0.338

where wul and wu2 are the un-frozen water contents at -1 and -2'C,
respectively.

The relation between un-frozen water content and temperature can

be written
log wu2 /W u9
wu •
W2w ul ((&T=
T
x .3010 11

where Wul and wu2 are given by the previous equations.

By plotting wu and wU2 in a log - log diagram for each of the two
temperatures one can determine the contours for un-frozen water
content by drawing straight lines between these values, due to the
exponential nature of the equation. This results in a graphical
method for determining the un-frozen water content at various
temperatures, as shown in Figure 13.

1)Literal translation of "slag". It seems that this test involves

beating on the sample in some manner. (Translator's note)

11 5
L_= . . .. .. . ........
 i 0c I I II . ..

800

•- ,, .. 2. .-•. • _•
42------------ _-

-0,1 -0,2 -0,/,, -0.6 -1,0 -2,0 -4,,0 -13,0 -10

TEMP•ERATUR• ,0

PlO Oz{3 ilak Mfto14J, , e•~ fitJ.i'

irl'. &V Jt'?Os, . vari*,ltt(*i ri~fll'~l u' fr

In many cases only the flow limit w2 5 (V1 o) :ls determined. However,
Selmer Olsen's consistence limit determinations for a number of
Norweg~an clay: show an avorage ratio of about 1.2 between ViO ad

rals studied by Anderson and Tice are also shown, as function of

wlOO. These data1 ) indicate that the same relation exists between
these two parameters.

1') Rings in Figure 14+ (Translator's note)

i
Consequently, if
only one of the consistence limits is given, the
other can be found from the relations

v 10 1.2 v 5 20

or w2 5 1•2 w1 0 0 21

This forms the basis for the thin un-frozen water contours in
Figure 13. If both consistence limits are given, the two scales
may be used as described previously.

At temperatures near the normal freezing point, the calculated

un-frozen water content may exceed the actual water content in
the soil material. Anderson and Tice recommend that the curve for
un-fromen water content in such cases is terminated at the latter
value. The corresponding temperature may then be considered as
a freezing point reduction. This condition indicates that water
content measured in relation to dry density is not entirely re-
levant at temperatures near 0 'C. If the soil material is nearly
saturated for the actual water content, the real freezing point
reduction may be considerably less than if the material were far
from saturation for the same water content. The empirical equa-
tion, which is based on water content by weight, gives no basis
for distinguishing between these conditions.

80
I ,

1(.0
W ,W,

>"20 j"
- VO . VS

0 ?0 40 60 1(
V S W'10' %

*I~1 V;I ;',*.:irr Ill J .. :, , ,.'*.**-,, ,

11"7
Until determinations of un-frozen water content in Norwegian soil
materials become available, the approach taken by Anderson and
Tice may be applied also to Norwegian conditions. Figure 14 also
indicates that the normal range of flow limits for Norwegian clays
is well covered by the materials studied by Anderson and Tics.

3. SOIL MATERIAL CLASSIFICATION

The previous discussion of various thermal conductivity parameters

gives a certain basis for generating classification requirements
for soil materials. The fundamental thermal conduction parameters
were earlier shown to be of three types:

Thermal conductance of the components X,, 2 40*.....n

Relative volume (shares) of the components n 1 , n 2 ..... nn
Microgeometric properties of the material

As shown, each one of these parameter groups pose special require-

ments on soil material classification. In practice, one will to
a large extent have to rely on geotechnical classification and samp-
ling normally performed in connection with road construction or
other types of in-ground work, It will thus be of interest to find
out to what extent this kind of classification covers the para-
metors which are important from a thermal point of view.

A. Geotechnical analysis

In connection with major road construction, samples are normally

taken of the sub-soil at every 20 or 100 metresalong the road-bed
and in some cases also perpendicular to its centerline. These samp-
les are primarily taken in order to determine stability properties
of the sub-soil. In fine grain materials such as silt or clay one

118
normally uses a sampling tube (diame.ter 54 mm) to extract "lndis-
turbed" samples. The samples are taken for esch metre (a metre
at a time) to a depth which depends in soil conditionsý. From 4;hese
samples one determines volumetric conditions (density, water con-
tent), consistence limits, screrning curvel) and various other
properties of importance to stdbility and compreasibilit'.

If the sub-soil consists of coarser materials, from coarse silt

through fine sand to sand and gravel, it will be difficult to ob-
tain undisturbed samples. In such cases, so-called representative
(stirred up) samples are taken, from which water content by weight
and grain size distribution curves are determined.

In connection with construction of large foundations (buildings,

etc.) the same procedures for sub-soil sampling as in road con-
struction are usually followed.

Figure 15 shows a sketch of a road bank, including the normally

used terminology (28).

Materials used in a road bank must meet certain requirements on

wear resistance (petrographic composition), stability, degree of
compression (grain size distribution) etc. To ensure that these
requirements are met, the materials used are subjected to certain
evaluations which also may serve for classification from a thermal
point of view.

1) Particle size distribution (Translator's note)

:1

, I119
 T _
... , .. . ...
• .... ;' '

0! *r d . f
01,J
.- • .s
- aj

U. . ,. i ai.•

" *a*•
:,;. .- . . .

V-.. 4 ,
- V!2 . U ;12 0 ner'- o

• ra, :•a r
¶ a *

Leen: Ovrbgn Roa paemn

_. ? ...... ?*-. -

:' a. J.• .
Proiline 5k i •sc av 3p r.'r. '., av e' n v e .,oV •' r bY • •• Upe-e1 . a n r iv surfac
ze . ~5 -'-

Of .z :O,2, D'J"p:,~.j'I .

Legend: Overbygning Road pavement

Vegd ekke Road finish
Vegfundamen t Road f'undation
Pro fillilnJe Upper surface
Slitelag Wear layer
Bindlag Binder layer
Baerelag Structural (support) layer
Ovre Upper
Forsterkningslag Reinforcement layer
Nedre Lower
Filterlag Drainage layer
Planeringslinj e Baseline

Compression properties are determined by means of packing experi-

ments (standard proctor), using a packing force defined by water
content. From these experiments one finds the maximum degree of
compression (packing) for a related "optimum" water content. The
results are presented in curve form, as shown in Figure 16. This
leads to requirements in dry density for the materials after pre-
1I)
paration. The requirements are 100 percent proctor in rein-
forcement layers (corresponding to maximum density in Figure 16)

1) Hopefully a standard term. (Translator's note)

120 Best Available Ccop.,

103 percent proctor in structural (support) layers and about 95
percent proctor in the drainage layer. The density of materials
in the road pavement are thus given within narrow limits (pro-
vided that regulations are followed) (28).

,Jott i '
1-W i c
]--. -'rr I.,,9'P ri tv'Ik" 1 c./]..

"k'i.o.
L .l. P (LIi LradI
W OW"
r' ''~~ I , L
Dv Lil~ll,
ýa*'Ll t m "', ,yi.•y 'J P
4,n
I -),
,T,/'r', r,)1tVL,'k soGm rjf J i•U+mti
ay vuxnninmnnoý. yu-1+r la,~Tn.,

For the structural (support) layer (mechanically stabilized) one

has comparatively strict requirement on grain size distribution,
defined in terms of limiting curves such as those shown in Figure
17. In the reinforcement layer it is required that the content
of fine material (d <2 =m) does not exceed 13 percent. The normal
variation in grain size is shown in Figure 17. (See Figure 11 for
legend. The bottom scale is labeled "Screen with mesh size.....

'.. m . .. .). .... .;

•° " ! t r+. . II- - - , . . . F' . . -'. . .

I ' I •• " /"'

S..... F-....
... *........ 0 . QA
... . .. _'"• ......
¢ - Jo,...

9
,,,.. .... ... T ." j .........
• -....... ,
.4 ,P • •_ _ • _ . ,

1211

I~1.3. 3.1. KI'I,iV '.UJ Io~ (Or,ifotJ I ,¶ItA PI..r J~t I'r d;5l 1..' ivrl+.+
ia|,JUr.,

121.

- .- ,- -. I
Requirements in wear resistance for the materials are checked
through fall tests (crush tests) and passing through a rod screen.
A petrographic determination of the rock fraction is also included
(30). However, this evaluation is primarily of a qualitative na-
ture. Artificially crushed rock materials are subjected to a more
quantative evaluation, including the percentage of different mine-
rals.

The results obtained from these analyses are listed in Table V,

where a distiction has been made between sub-soils containing
.fine grain and coarser materials, as well as materials which are
to be used for the road pavoment.

Table V: Classification of samples taken during road construction

Type of analysis

-'~~
94v.r-4~
10)r0

Pavement1 !aer7l 4J1

Type of material
Sub-soils Clays silt X x X x
Sandy soils x X
Pavement materials x-

1) Given by proctor

122
-.- I--,-;y*-~.'i
 B. Supplemental analysis

Bas,,d on the different thermal conduction parameters discussed in

the previous section it is possible to list the most important
parameters which must be determined to obtain as accurate a value
as possible for the conductivity. Figure 18 summarizes these pa-
rameters in a block diagram. The diagram is arranged in terms of
information levels, I, I, and III, corresponding to different
uncertainties in conductivity when determined on the basis of
parameters listed in each level. The first and lowest level only
requires knowledge of water content, dry density and mechanical
composition (screening curve). o2h highest level, III, requires
more detailed information about physical properties of the materi-
al.

1VARfl4_SD'4: NGSEVNE ,
{ WE4IWIA.,%. '• 'C,,t
0• SO,'.

T.'S ,VA NN
".'JF !PA .
It'CLE COI.NVO AC','N, ,EY ,'AF

.*Y F.RAtOG! .Fr t VANIN 7 QORNFORY • .-

I.'r

'775
T'S!vSR
'VTs!'I4N": .7 E
. :S S ..... i Q GANIS•
I !r~~r;
,"IS
LYTr:,3,1ENSr7,L~eu
c~

.tLAqTS CONV4 , 1 ' AVITY

"EC U4I0ANIC •,',T LI I uIT
W

;VA--1N _r"77 T I; ': <•,;

. EL T-vr.
VA
1 ~ I'i~JN
~vv ~WVA T
'VOAIE4 WN'j I :Jrq ••.''Iv ¢()4•p
(-¢:H
"":T 1."..
•,r .".,, .i;!,,v

Best Available Copy

123
 Together, T$. and Figure 18 give an overview of how compre-
hensive the • normally performed during road construction
are for the factors which determine thermal conductivity. One
finds that the tests listed for fine grain sub-soils are almost
* equivalent to level II, except for knowledge of quartz content
in the material. However, this parameter can be estimated from
the particle size distribution (screening curve) with sufficient
accuracy for this case. For coarser (sandy) sub-soils, informa-
tion about dry density is lacking. This makes results of common
analysis incomplete for all the levels shown in Figure 18. For
materials used in the upper levels, informption on water content
will be lacking. One must therefore make an estimate based on
experience from similar road bed designs and take into account
expected variations. The petrographic analysis, which is limited
to rock fractions and also is rather superficial, gives little
information about quartz content in the material at hand. For
"* that case, the sample analysis may thus at best give sufficient
information for the lowest level of conductivity determination.

** Supplemental studies required to reach level I1 will first of all

be the following: Determination of density in coarser oub-soil ma-
terials, more accurate estimates of water content in upper layers
and quartz content determination for sand and gravel materials
(e.g. by means of DTA).

* The uncertainty involved in determining the conductivity for the

three levels defined here will be discussed in detail in the final
chapter. It may be worth mentioning at this point, that the lowest
level (I) will in this rospect correspond to Kersten's empirical
materials. Kersten claims that correct use of his method will give
uncertainties below 25 percent. (Indications are that the error
limits are some-what wider than that.) Transition to level I1 will
give a significant narrowing of these error limits,

124
 REFERENCES - CHAPTER III

1. A. Holmes: Principles of Physics-Geology, Nelson, London

1965, p. 84.
2. F.J. Pettijohn: Sedimentary rocks, Second Edition, Harper &
Brothers, New York 1957, p. 115.
3. R. Selmer-Olseni Ingeni~rgeologi, Del II, Losmarser,
Kompendium, Geologisk Institutt, NTH, 1973.
4. R, Selmer-Olsen: Ingeniorgeologi, Del 1, Generell geologi, Tapir
forlag, Troniheim 1971, p. 175.
5. H. Sveian2 Norske jordarters kvartsinnhold. Hoveioppgave i
Ingeni~orgeologi, Geologisk Institutt, NTH Host 1972.
6. H. Sveian: Unjers~kelne av kvartsinnhold i losmasser ved hjelp
av DTA. Utftrt for Institutt for kjoleteknikk, Tronheim 1973.
7. R. Selmer-Olsen, En regional undersokelse av norske kvartare
larer. finfraksjon basert pa DTA. RApport senit til NTNF,
Geologisk Institutt, NTH, Tronheim 1961.
8. F. Birch, H. Clark: The Thermal Conductivity of Rocks.
American Journal of Science, 238, (8), 1940, pp. 529-558.
9. D.A. de Vries: Het warmtegeleidinguvermogen van grond. Med.
Landbouwhogenschool. Wageningen, 52, 1952, pp. 1-73.
10. A. Watzinger at al: Masseutskiftningsmaterialer for toleforbygning
pi veg og jernbane. Norges Geotekniske Institutt, Oslo 1965,
p. 106.
11. K. Horai: Thermal Conductivity of Rock-Forming Minerals. J.
Geophys. Res. 76, (5), 1971, pp. 1278-1308.
12. J.H. Sass et al: Thermal Conductivity of Rocks from Measure-
ments on Fragments. J. Geophys., Res. 76, (4), pp. 3391-3401.
13. S.Skaven-Haug: Romforhold i jordmaterialer. Meddelelser fra
dot norske myrselskap, 70, (4), 1972, pp. 89-102.
14. Analyseforskrifter, Statens vegveuen, Vegdirektoratet, 1966.
1. Jordarter, 111, Spesifikk vekt - vanninnhold - romvekt.
15. E. Hogberg: Vattenhaltens innverkan pi densitat og kompressi-
bilitet ho. packade jordarter. Byggforskingen. RApport R8,
1972, p. 87.

125
........................................... I
 16. R. Selmer-Olsen: Ingeni~rgeologi. Del I. Generell geologi.
Tapir forlag. Tronheim 1971, p. 8.

17. R. Selmer-Olsen: Ingeni~rgeologi. Del I. Generell geologi.

Tapir forlag. Tronheim 1971, p. 12.
18. R. Selmer-Olsent En regional unders0kelse av norske kvartare
leirers finfrakajon basert pa DTA. RApport senjt til NTWF.
Geologisk Institutt, NTH, Tronheim 1961, p. 29 f.f.
19. C.A. Bower, J.O. Goertzen: Surface area of soils and clays
by an Equilibrium Ethylene Glycol Method. Soil Science, 87,
1959, pp. 289-292.

20. I.M. Eltantawy, P.W. Arnod: Reappraisal of Ethylene Glycol

Mono-Ethyl Ether (EGME) Method for Surface area Estimation
of Clays. Journal-Soil Science, 24, 1973, pp. 232-238.

126
 CHAPTER IV
EXPERIMENTAL METHODS FOR DETERMINING THERMAL
CONDUCTIVITY IN SOIL MATERIALS

In this chapter, the emphasis will be placed on discussing the

most commonly used methods for measuring thermal conductivity
in soil materials. However, other pertinent approaches will al-
so be treated in the last section. The general influence of
moisture transport on experimental results will be discussed in
section 1 but this topic will also be dealt with in connection
with the various methods of measurement.

I THERMAL CONDUCTIVITY MEASUREMENTS IN IATERIALS

CONTAINING MOISTURE

Experimental determination of thermal conductivity is usually

based on the equation which differs conductivity

* q- VT
q is the thermal flux (l/ma)

VT is the temperature gradient ( 0 C/m)

The discussion of heat transport mechanisms in moist materials

(chapter I) showed that the heat transfer in a soil material con-
taining moisture is influenced by transport mechanisms other then
pure thermal conduction, for which this equation is valid. How-
ever, it was shown that these mechanisms could be included in a
temperature dependent "apparent" or "effective" thermal conductivi-
ty.

Iztroduction of a temperature dependent thermal conductivity

leads to a non-linear thermal conduction equation (Fourier's
equation). For constant thermal conductivity this equation has
the form:

127
 aT L V2 T2

c is the specific heat (J/kgK)

3
o is the density of the material (kg/m )

From this equation one defines a diffusivity for the material

. 2 3
For variable conductivity:

*T v( 7 T) 4
Scp

Most methods for experimental determination of thermal conducti-

vity are based on the linear equation (Eq 2). Use of the non-
linear equation will require detailed information about the tem-
perature distribution and a considerable effort for temperature
data reduction. •Towever, such methods have become feasible due
to availability of modern computers (1, 2>.

If one neglects the temperature range where freezing and thawing

occur in the soil materials, the temperature dependence of the
thermal parameters is so small that the errors due to use of the
linear equation can be reduced to negligible magnitudes, pro-
vided that the temperature variations are kept within reasonable
limits during the experiments.

In the temperature rango where phase changes water/ice take place,

the nonlinearities will be large, particularly as a result of wide
variations in the apparent thermal capacity per unit volume, so
that the linear equation becomes almost non-useable.

For the case where no water transport occurs during freezing, the
thermal conduction equation for the partially frozen zone in a soil
material can be written:

128
4
( CO + u YL) BT
-v(X T)
ST

wu is relative un-frozen water content by weight

L is latent heat of water, when freezing (J/kg)
3 K)
E co is the thermal capacity per unit volume (J/m

The latent heat when freezing can thus be included in the apparent
thermal capacity/volume, but this parameter will vary extremely
fast with temperature near the freezing point. An example is shown
in Figure 1. The curves apply for a silt-clay material having a
dry density of 14090 kg/m3 and a flow limit of 30 percent by weight
(see Figure 13 of chapter III).

I.

" .t , j j
TEPERA%~ I

." r , . .r I. jp.i t ' "

For such extreme nonlinearities in' the thermal parameters it is

obvious that the linear version of Fourier's equation can not be
used in conjunction with transient methods for determining con-
ductivi ty.

In stationary experiments (aT/at where only non-linearities

0),
in the thermal conductivity come into play, smaller errors can be

129
expected. The approach will then be to determine an average ther-
mal conductivity, based on heat flux and temperature difference.
This average conductivity can then be related to the average tem-
perature in the samples. Non-linearities in the conductivity due
to varying amounts of un-frozen water can be expected to give cer-
tain errors in this relation. If the temperature coefficient for
the conductivity can be assumed constant within the given tempe-
rature interval, the errors should be insignificant. However, this
will be true only if relatively small temperature differences are
used across the samples.

At temperatures above the freezing point, the different heat trans-

port mechanisms will cause difficulties of a different kind. The
nonlinearities in thermal parameters will, as mentioned earlier,
play a minor role. However, some of the heat tranport mechanisms
discussed previously will cause a certain mass transport which may
disturb the homogenity of samples. These mechanisms were detailed
in section 2 of chapter I.

In that section it was shown that the change in moisture distribu-

tion caused by a temperature gradient is largely dependent on the
water content in the material and its moisture related properties,
such as moisture potential (suction), hydraulic conductivity and
moisture diffusivity. It w~s indicated that large changes in the
moisture distribution can be expected in a "critical" region of low
to moderate water content (degree of saturation), where the moisture
diffusivity decreases rapidly while the thermal vapor diffusivity
lies close to its maximum value. It was also shown that the loca-
tion of this saturation region tends to move towards higher satu-
ration levels for more fine grain materials.

For saturation levels above this region, the moisture diffusivity

will increase sharply so that any tendency for change in moisture
content due to thermal vapor diffusion will readily be counter-
acted by capillary water transport in the opposite direction.

130
 One must clearly impose certain requirements on methods used for
measuring thermal conductivity within the region of critical sa-
turation. The goal must be to maintain changes in the moisture
distribution within certain acceptable limits. This could be
accomplished by lowering the temperature gradient or by reducing
the time during which samples are subjected to this gradient, or
by combining these approaches.

In addition, methods used should give an indication whether major

changes in the moisture distribution occur during the experiment.
The following section will describe commonly used methods and dis-
cuss these from this point of view.

2. STATIONARY METIHODS

Stationary methods usually employ one-dimensional heat flow (con-

duction perpendicular to a plate) or radial heat flow through a cy-
lindrical shell. Both methods have earlier been used for studying
thermal conductivity of soil materials (3, 4, 5, 6).

A. Kersten's cylindrical teiý unit.

A cylindrical test unit (arrangement) employing radial heat flow

* was used during an extensive experimental investigation of ther-
mal conductivities in soil materials, published in 1949 (6). A
sketch of the equipment is shown in Figure 2.

The cylindrical shape is supposed to allow a more efficient packing

of soil materials than for a traditional plate arrangement. However,
the geometry also has the advantage of permitting direct measurement
of the heat transfer through samples without use of special "heat
screens" required for one-sided planar units. An effort is made to
obtain the desired radial heat flow by separating the heaoing ole-
ments into one measurement section and two screening sections. The

131
 &VRE SKJERM • '

, -. . .,.
SENI. ALT i
vARFE,=LEMENT "•.•, l

NEU.'..O !, ,..,

SPRI~T --.
SPR!Ti
I nim*!.n
Xr 1 j•.. P.i

PLUGO FMI

-* Pt~~rrnon &p~r~
anip or r av o:yi.r var*mn"-

qOZ
1500w'" 50i Diameters 150 mm, 50 mm
SKJEM SEKSJON
•VRE
.I Cro,"ecton. upper screen
JORDPROVE
SENTRALT VARMEELEMENT
MALESEKSJON
Soil sample
:Center heating
S.. element
Measurement zone
. ...
NEDKJ0LT SPRIT 2Cooled alcohol
NEDRE SKJZRM SEKSJOX Crou..sectiong lower screen
SPRIT INTAK Alcohol inlet
BORD TBoard (table)
PLUGO FOR TOMNING :Drain valve

132
latter correspond to the ring-shaped screens used in planar arrange-
ments )4 One tries to minimize the temperature differences across
the gaps between sections by adjusting the (heating) power in the
screen sections. These adjustments are made manually.

Kersten does not report typical stabilization times, but his des-
cription of the procedure used at the end of each experiment gives
an indication of the conditions.

Temperatures at the inner and outer surfaces of the sample were re-
corded each time, as well as current and voltage for the center hea-
ting element. If the thermal conductivity calculated from these
values was found to be constant within error limits of one percent
during three to five hours, the final conductivity was computed as
the mean of results obtained for the final hours.

The stabilization periods should have been significantly longer, at

least 12 hours. For each sample placed in the test unit, conducti-
vity was determined for from four to six different temperatures of
20, 4, -4 and -304C, or 20, 4, -4, -18 and -30 0 C, in addition to a
repeat test at +20'C in the latter case (6.levels). The total time
that each sample remains in the test unit should thus amount to be-
tween 48 and 120 hours, for a stabilization time between 12 and 20
hours. Against thIs background, the completed test program is quite
impressive: A total of 240 samples were tested, utilizing two test
units of the type previously described.

The long time required for each experiment is perhaps the biggest
drawback with the stationary methods. This is also the reason why
non-stationary (transient) methods (particularly thermal conduction
probes) have come into use, despite the considerably higher accu-
racy obtained with stationary methods. However, the long duration

1) The word "screen" is probably adopted from similar structures

used to eliminate edge effects in electrostatic measurements.
(Translator's note)

133
of each experiment may also be a questionable factor due to the
previously mentioned moisture transport.

In this context it is also worth noting that the cylindrical geome-

try causes some-what higher temperature gradients on the warm side
than that obtained with planar geometry.

Kersten used a temperature difference between warm and cold surfa-

cas of about 5.69 (corresponding to 10 0 F). With the actual test
unit dimensions this implies a temperature gradient on the warm
side close to 1 0 C/cm. In a planar test unit with a sample thickness
equal to the width of the ring-shaped sample gap, the same tempera-
ture difference will result in a gradient of about 0.6 0 C/cm. For
larger diameter ratios, the difference will be more pronounced and
for thermal conductivity probes, the cylindrical geometry will de-
finitely be unsuitable (see the section on conductivity probes).

The thermally excited moisture transport will cause the sample to

become inhomogenous. The cold side will have an increased moisture
content while the warm side dries out. The measured conductivity
then represents the apparent conductivity for an inhomogenous mate-
rial and does not represent the real conductivity of the initially
homogenous material.

One can obtain an estimate for the magnitude of resulting errors

by assuming that the moisture content varies linearly with radius
in the cylindrical test unit. If one also assumes that the conduc-
tivity varies linearly with moisture content, one can derive the
following expression for the apparent conductivity after the mois-
ture has been re-distributed.
"r2/r - l2/XI
in r 2 /r 1

In r 2 /r 1 - in r 2 /r 1 - 1
/1
Sr
2 /rI is the ratio between outer and inner radii

X2/Al is the ratio between conductivities at the cold and

warm surfaces

134
Under the assumptions made, the conductivity obtained for even mois-
ture distribution is given by

Xm = 1 ý2 +7

Figure 3 shows calculated ratios of the apparent conductivity af-

ter moisture re-distribution and the conductivity with evenly
distributed moisture. Curves are given for three different ra-
, broken
dius ratios as well as for the planar case (r 2 /rI,-
curve).

0.5. -

0.2 -

1X x,

F1G . T.nnrln."y e.-j• av ine'• o r "t ,J l : av .:k.nlv:°'

r'I I- em

For relatively moderate distorsions of the moisture content,

corresponding to X2/X1 < 2, the errors will not exceed 10 per-
cent for each of the calculated :ases. One also finds, that the
errors increase with increasing radius ratio and that the cylindri-
cal geometry is always less favourable than the planar (plate test
unit).

Best Ava'ilable Copy

135
As mentioned previously, Kersten performed measurements for up to
six temperature levels on each sample. For a number of samples,
the final measurement was made at the same temperature as
the first (+20"C). A comparison of the results obtained in these
two measurements will illustrate the influence of moisture trans-
port in these experiments. Figure 4 shows such a comparison.

2.5,

2.0-• //
IX Y

z-, 1.5 ,-

"-~.2- 05- I /Y

S0.5 .
0 0 A- I I
I
0 0.5 %.0 I.5 1.0 2.5
F 6RSTE M;\L'NG. Wire
m

= * .-. S:•±~ni'j.nin'.~ r,,L.,,w..'n t.- m• iflrw ';e. 20C3C o.: run*'.e:-"

It can be seen, that the final (repeated) test with few exceptions
gave a lower value for the conductivity than the first measurement
at 20 0 C. However, the deviation is in the majority of cases less
than 5 percent.

B. Planar test unit

Experiments with moist soil in a planar (plate) test unit show con-

siderable- influence by moisture transport. Thermal conductivity

measurements performed on moist soil materials by Smith were dis-
cussed earlier (chapter I). A few years earlier (1934) comparable
results were reported by Krischer (4). ,'Noodside (1958) published
results from stationary measurements on moist sand and clay, using
a planar test unit. These results also indicate the importance

136 Best Available Copy

of moisture transport (7, 8). Common to all these experiments is
a relatively high average temperature in the samples, 200C or higher.

During his measurements on sand and clay, Woodside placed emphasis

on obtaining stationary conditions, including a stationary moisture
distribution. In the experiments with moist sand this required
stabilization periods up to 200 hours, while for clay stable con-
ditions were obtained in 163 hours. Figure 5 shows recorded heat
flow into and out from the sample in an experiment with clay con..
taining 18 percent water, by weight.

_ , 60
"A SA !-'LCW
. ' ,6 f
"•-'-• 1 ? 2. S ~'0 o2 • •0 'O ?
T'9 ! n':!,r'

*• 5. Va stz".rtl'd• min n,, u'. av p. . v•. fl i•-

vu ',,±_1.r::n;'o~'.:.'i

:' PI.l .'•.•.I

.• '-:.
. '•o ,'z_'a ,' ' ' ,• .. t-•
d ; : -•d.

As can be seen, there is a period of 5 to 10 hours after the initia-

tion of the experiment, during which conditions appear to approach
a steady state. After this time there is a significant reistri-
bution of moisture, which continually changes the balance.

One may also note the large thermal flux at the warm surface of the
sample during the early phase of the experiment. This indicates
that large temperature gradients occur at the warm sample surface
following'a step-wise increase in heater temperature at the start
of the experiment. Correspondingly large gradients will also
occur initially when experiments are performed in Kersten's circu-
lar (cylindrical) test unit. These large gradients will obviously
accelerate the moisture transport.
Best Available Cpy
 One could possibly reduce these gradients by using a "softer" ini-
tial temperature transient (step). However, this would lengthen
the stabilization period and allow the moisture transport more time
to re-distribute the moisture content.

As indicated in the introduction to this chapter, use of the linear

thermal conduction equation as a basis for thermal conduction ex-
perimerJs really requires that the material under test be homogenous.
If .ealistic values of conductivity are to be derived from experi-
ments where significant changes in moisture distribution have occur-
red, like in the experiments reported by Woodside, one has to per-
form detailed measurements of both temperature and moisture profiles
(distributions) at the end of the experiment. in such cases, conduc-
tivities can be calculated for thinner layers within the material
over which temperature gradients and moisture content can be expec-
ted to change by small amounts. In effect, this means utilizing
the non-linear version of Fourier's equation:

I:.C = T )

This will mean a very time-consuming procedure, not least due to

the long stabilization times required.

Another possibility is to arrange experiments so that the redistri-

f. bution of moisture is minimized, by reducing both experiment dura-
tion and temperature gradients as much as possible. (For very spe-
cial cases, where the conductivity is to be measured at particular-
ly high temperatures, this problem may have to be approached in a
different manner, e.g. by transient experiments and continous re-
cording of both temperature and moiature profiles.)

The principle indicated above was utilized for the development of

planar (plate-type) test units used for determining thermal con-
ductivity in soil materials at the Institute for Cold Technology
at NTH. This test equipment was constructed as part of a research

138
program named "Frost in the Ground", directed by NTNF and the
Highway Authoroty with participation of the "Research Group for
Thermal Analysis of Frost in the Ground" at the Institute for Cold
Technology.

The test equipment is designed for planar, one-sided heat conduc-

tion. Considerable emphasis is placed on obtaining planar, horizon-
tal isotherms within the field of measurements and to ensure the
best possible stability of temperature and heat flux.

The construction and function of this equipment has been described

in more detail by Brendeng, Frivik and de Ponte (9, 10), who have
also together been responsible for its development. Only the main
aspects on this equipment will be discussed here.

Figure 6 shows a sketch of one of these units, with major construc-

tion features indicated. The same figure contains details of the
heating plate and the power screen.

Power to the power screen is regulated automatically, by means of

a signal from the central measuring plate, so that any temperature
difference between power screen and measuring plate is eliminated.
This ensures that the entire power applied to the measuring plate
will pass through the sample. Thus, the thermal flux through the
sample can be recorded by direct measurement of current and voltage
applied to the measurement plate.

Figure 6 also shows that the heating plate is divided into a mea-
surement area and a screen area. The power applied to the screen
is automatically controlled by means of thermal sensors placed
across the gap between measurement and screen areas so that the
temperature difference across that gap is maintained at zero.

To further ensure planar isotherms, an additional heating element

(coaxial cable) is placed around the outer rim of the screen area.

139
 A. Cross-section of the heating plate
unit.
B, Top view of heating plate unit.
a. Central heating plate with hea-
ting foil 0.2 X0.2 m2
- -
.. .b, Ring-shaped screen with heating
SL ',foil

,,...c, • Gap between measurement and

'/ screen areas
. d. Power screen with heating foil
, e. Heating cable
-.,. • f. Insulation and spacers

"g. Temperature sensors (platinum

elements)

*, .

,- - \- \.. . . . .. r i i

. *' *i

Jt. Grdin scee. Hai

~ ?
ncalatop
s~"p---
*n " . X .

~14
C. Assembly sketch of plate unit.
h. Heating plate
±. Cold plate with pipe. for circulating alcohol and heating foil
j. Gradient screen. Heating cable at top.
k. Upper cold plate.
1. Mixing tank with magnetic valve and pump.
The power applied to this element is also controlled by thermal sen-
sors so that its temperature is maintained equal to that of the
central measurement area. The stainless steel sample holder con-
nects warm and cold plates thermally such that one obtains approxi-
mately the same temperature profile in the connecting member as in
the sample material (gradient screen). This prevents heat leakage
from the sides of the samples.

The design is shown in Figure 6 and Figure 7 shows the function of

the automatic controls.

As indicated, the temperature of a cold plate is controlled by a

regulated flow of the cooling medium (methanol) through a magnet
valve into a mixing tank. To ensure stable temperatures, a heating
foil is included with the cold plate and the power applied (to the
foil) is automatically regulated by temperature sensors in the cold
plate (feed-back). The reference temperature for the cooling medium
(set point) is a few degrees lower than the desired cold plate tem-
perature, in order to provide a nearly constant power consumption
in the heating fbilo

The power applied to the heating plate is controlled by means of

feed-back, by means of a temperature sensor in the heating plate.
A common adjustment (set-point) is used for the temperatures of
warm and cold platcs, along with a variable difference between these
temperatures. This makes it posslble to maintain a set temperature
difference while chainging the temperature level.

To obtain short stabilization times, the warm platp temperature is

regulated towards a cotistant Palue already at the beginning of each
experiment. However, this step-wise change in temperature is some-
what limited by the thermal capacity of the heating plate itself,
as well as the limited power capacity of the heater. Figure 8 shows
an example of how applied power and warm plate temperature may vary
from the start of an experiment and until stability (steady-state
condition) is reached. This also gives an indication of the degree

- -.
141.
1
of temperature stability that can be obtained with this test equip.
ment.

F let .. -. .. ..... . P t a

'F'!
- -r

7, Ot 20.7.

. ,. ... *V. -'t. ...... " ?•

.. rO.
tat en.r

-Z 4

-o .- 0 7'.
' ,2

2 40

i jt• 1'# I,1t) ,; r ! ;Q:'t-'I,'.,• . IJ f'.. P41

. ) ,Z
V pilO,

hV

1.42
In addition to the plate units described (3 units), equipment was
developed and built for transient measurements by means of thermal
conduction probes. This will be described in the following sec-
tion.

3. THFRMAL CONDUCTION PROBES!

As mentioned in the previous section, there is today a tendency

to abandon traditional stationary (steady state) methods in favour
of non-stationary (transient) methods. This is due partly to a
requirement for reduced experiment tizes, partly to a desire to mi-
nimize the influence of the previously mentioned moisture transport.
Of the different methods in question, the so called thermal conduc-
tion probe approach has seen the widest use for measuring thermal
conductivity in granular materials. This method will be discussed
in this section while other applicable methods are treated in sec-
tion 4.

A. Influence of finite size and varying thermal

properties

A thermal conduction probe can ideally be considered as a filament2)

of infinite length, surrounded by a homogenous medium of infinite
extent in all directions. It is assumed, that the temperature
field in that medium initially is homogenous (T = T0 ). Starting
at a time to, a constant power q (W/rm) is applied to the filament.
Based on the linear thermal conduction equation (Fourier's equa-
tion), the temperature field around the filament can be determined
(12).

1) The original use the worU "sonde", literal translation "soun-

ding device" or "probe", which usually refers to a passive sensor.
However, the device described here is active (Translator's note).
2) "Lin.ekilde" also implies an "infinitely thin" filament (Tranwla-
torts note).

143
 T - To q rEi r' )I~
4 at 9

a is thermal diffusivity (m2/s)

W X
Ei (x) - I S dx

The integral function Ei can be expanded in a power series:

- El (-x) * - y - in x + x - x2 + .... 10

y is Euler's constant a 0.5772

For 4 at/r 2 >1 one can thus write the equation describing the
temperature field in the approximate form
S~q
T- To -Y + lnt + In 4 a/r 2 ) 11

If the power to the filament is disconnected at time t', the

temperature field can be approximated by
Sq t
T - To 4nU- in -tI. 12

The increase intemperature, T2 -T 1, between times t and t

derived from Eqq 11, becomes
T2 - Ti - in t2/ti

The two latter expressions do not contain the thermal diffusivity

and the radius r, while the thermal conductivity can be solved
for explicity. Eq 13 is the expression normally used for cal-
culatinS the thermal conductivity in connection with experiments
utilizing thermal conductivity probes, Eq 12 may be valuable as
a check of the results obtained, particularly for measurements in
moist materials.
1) TUs integral defines -Ei(-x). The original is probably in
error. (Translator's note)
144
S.• . ,, - - e_- . m "l IIIHI'IIIr 1i" 1 nIP• t ' '••_.'_: _ " " . • . , ,•-o• • ._ , .. .. " ,• ''
Use of the thermal conduction probe was originally proposed in 1931
by Stalhane and Pyk (11). They developed an empirical method for
computing conductivity from the temperature rise in the probe.
The analytical foundation was later developed by Van der Held and
Van Drunen (1949)(12). In both cases, the method was used for mea-
suring conductivity of soil materials(13).

For measurements in liquids, the probe can be configured as a wire

or thin filament. Probes for use in soil materials are usually
made from thin tubes, typically filled by an insulating material
into which heating wires and temperature sensors are molded. The
geometry of the probe will then deviate significantly from the ideal
model on which the analytical solution is based.

Several researchers have investigated the effect of this deviation

from the ideal case (14, 15, 16, 17). De Vries and Peck (1958)
have for example treated the case of a probe having finite thick-
ness (diameter) but containing a filament1) (16). They find, that
Eq 13 can be used for determining thermal conductivity, provided
that 4 at/r 2>1, if the contact resistance between probe and the
material under test is small and the thermal capacity is not signi-
ficantly larger than that of the material under test. They also
claim that this condition is met for soils made up of min-rala,
while conditione are less favourable for organic soil materials.

Probes of finite extent (length) were first treated by Blackwell

(1953, 1956) (17, 18). He considered the case where a probe is im-
mersed in a homogenous material of infinite extent. With the aid
of various mathematical approximations he arrived at an expression
for the errors caused by axial heat flow in the probe. For a
particular type of probe he found that these urrors are less than
one percent if the length to radius ratio exceeds 30.

1) It seems implied that this (thin) filament carries the hea-

ting current (Translator's note).

145
Kierkus et.al. (1973) (19) have performed an analysis for a thin
filament of finite length, with some-what different boundary con-
ditions. The study was performed in conjunction with conductivity
measurements in liquids, when the probe consists of a thin metal
heating wire. Since the conductors required for supplying the
heating current have significantly larger cross-sections (than the
heater wire itself), the temperatures at the ends of the probe are
assumed constant during the entire experiment and equal to the ini-
tial temperatures* Using modern computer methods for numerical in-
tegration, the temperature rise in such heating wires of finite
extent was computed for different length/radius ratios and diffe-
rences in thermal properties between heating wire and the surroun-
ding medium. The results obtained (from these computations) give
a good illustration to the dependence of different parameters.
Some of these results are shown in Figure 9, where temperature rise
and time are given as normalized1) numbers.

Figure 9a shows the effect of finite length and different thermal

(heat) capacities for the case where the probe has a much higher
thermal conductivity than the surrounding medium ( Xi/X - 200. As
mentioned previously, this applies to conductivity measurements
with the heating wire placed in liquids. The dependence on the con-
ductivity ratio is given by Figure 9b for a given length/radius ra-
tio of 500.

These results show that an infinite long thinZ) cylinder has the
same response as an infinitely thin filament when the cylinder has
the same heat capacity as the surrounding medium. If the infinitely
long cylinder and the medium have different heat capacities, the
temperature rise will initially be steeper than for a filament but
after a certain time the response will flatten out and coincide
with that of a filament.

1) The English text in Figure 9 uses the word "dimensionless".

2) The English text under Figure 9 uses the word "compact" (Trans-
lator's notes).

146
 - Lit
Lit

z_ LS- LI .r,00 /
;" | X~),II, C
;',2

fl, '0i

02 "
CII,,0

10' 10o I li i ol l0e ,107 0T ¶0. 10,.

0!wENSJONSL0% T! .0IA ,

a. Gitt lel~nl.nrevnereor-ho!4• Ai/ = 200 ogr

varierenle nvJer ,l ;.u or'no' o ,'I.arm -
ot;
kapasiet. Conjuc vi:,o ratio ',I/
t 200
anid v~ariable Lenzth-2-'id'Us ratto 01Id hea C.DaciCI.

"00

1
/

0.7sF

2000

0.2Si

10
L V0 a 0
e tsi t0e
0n 1en0.,Ie- 'o,Ü r". tt

'~c /oj.
VariLd.th.,dusr'n•.
eer~.~
i oi
.ýn C~rrr1o'.I
give .
:ion vartob.:c:;pac:'
co,/ ..?7I
n

S . rer.atur-t'e.nons av en kolr.a.ct .ytinle! met, ns-an

p?, overflt" ,,er..n' .ra Ldrj r.!r, oan, "
a compaoe spyi.¶Jr u-••h con 'i=:; aiir'nc. , "'-
(Xierk'.. IL.UJ 4L aJ.

The finite extent (length) of the probe will, after a certain time,
cause another bend in the temperature rise curve towards values be-
low those for the infinitely long probe. This bend will occur ear-
lier for lower values of C/C 1 . Thus, for different heat capacities
and finite length one may at best expect that the probe will have
the same response as a thin filament over a limited time interval,

14 Best Available Copy

1_47
such as that shown by the curve for C/CI = I0"2 and L/r = 5000.

For equal thermal capacities and finite length, the conductivity

ratio will determine how closely the response will follow that of
a filament, see Figure 9b. In principle, high conductivity in the
probe is not desirable, since a low conductivity brings the re-
sponse closer to the filament response.

Traditionally, probes for use in soil materials have been fabrica-

ted from thin wall tubing, e.g. stainless steel, filled by an in-
sulating (plastic) material in which heater wire and thermoele-
ments are placed1 ). This makes it difficult to apply the results
just discussed to probe measurements in soil materials. The bounda-
ry conditions at the ends of the probe vill also deviate from those
assumed in the analysis performed by Kierkus, which further compli-
cates matters.

Practical considerations dictate that the length/radius ratio of

probes should be limited, e.g. to about 200, which for a probe of
2 mm thickness (diameter) gives a length of 20 cm. The results de-
scribed here also indicate that the thermal properties of the probe
should be matched to those of the material under test as well as
possible, both in terms of thermal capacity and conductivity, in
order for satisfactory results to be obtained (c.f. Figure 9b).
However, the limited length/radius ratio will not guarantee that
the response is the same as that of a thin filament.

Another alternative would be to increase the length/radius ratio

significantly, e.g. by using the heater wire itzelf as a probe.
This would result in extreme temperature gradients near the probe,
which is not desirable for measurements in moist materials (c.f.
next section). However, this approach may be practical for measure-
ments in completely dry or saturated materials.

1) The original uses the word "tnnLatt', lIt ral translaticn

"1"baked in" (Translator's note).

148
Comparative investigations initiated by the Institute for Cold
Technology indicate that probes having the previously given di-
mensions (d = 2 mm, L = 20 cm) give up to 20 percent higher con-
ductivity values than steady state measurements performed in a
planar test unit. This applies both for dry and moist soil samples.
However, experiments performed with longer probes (d a 1 mm,
L = 100 cm) gave much better agreement with results obtained in the
planar test unit. Experiments using "probes" consisting of a hea-
ting wire (alone) also gave results in good agreement with planar
measurements in dry and saturated soils. Even if these studie6
are still going on, they still serve to support the analytical re-
sults discussed here, including the new requirements for length/radius
ratios developed here (20).

B. The effects of moisture transpovt

Heat conduction probes constructed in the form of heater wires appear

to be suited for measu'ements in dry or saturated soil materials.
However, special problems arise in moist soil materials, where the
large temperature gradients occurring near the probe can cause mois-
ture transport and a drying out of a "kernel" around the probe. By
taking the derivative of Eq 11 with respect to r one obtains the
following expression for the temperature gradient:

8T 14
ar 21r X

A reduction of the temperature gradient will thus require an in-

crease of the probe diameter and/or reduction of the applied power.
However, both these parameters have practical limitations. Low
power results in a small rise in temperature, which may cause un-
certainties in the conductivity measurements. A large probe dia-
meter leads to a longer probe in order to maintain a certain length/
radius This would require larger aample volumes and more
ratio.
work for preparation of samples in the laboratory.

149

~
. *"A -,
 If one wants to use a power which causes the temperature to rise
by IOC over a 10:1 time interval (e.g. from t = 1 min to t 2 = 10
min) one finds from Eq 13 that the ratio q/(2Hx ) should be very
close to 1. During the measurement period, the temperature gra-
dient near the probe then becomes:
3T 1
- = __ 15
ar r

A probe diameter of 2 mm would thus result in temperature gradients

of 10 0 C/cm. Even larger gradients will occur immediately after the
start of the experiment. During that period, the probe temperature
will normally rise by a larger amount than during the measurement
period itself.

The gradients which occur are significantly larger than those oc-
curring during stationary experiments in a planar test unit. It is
doubtful whether the much shorter measurement times (using the probe
method) can compensate for this difference.

Experiments in moist materials also show effects which can be re-

lated to drying out of the material near the probe. For example,
Woodside registered a steady decrease in measured thermal conducti-
vity during experiments with moist clay (21). The probe diameter
was in this case about 0.5 mm. A series of experiments conducted
at the Institute for Cold Technology at NTR1 indicates, however, that
this effect is moat pronounced for degrees of saturation within a
certain range. Figure 10 shows the temperature rise in experiments
with a silt material at different degrees of saturation.

As shown by Figure 10, the temperature rise changes significantly

beyond a certain time for the three cases having the highest de-
gree of saturation. This significant increase in the rise can be
L directly related to a drying out of the material near the probe,
since an increased (temperature) rise corresponds to reduced con-
ductivity (c.f. Eq 13). For the lower degrees of saturation there

I, 150
is no similar effect and the effect is also relatively small for
the. highest degree of saturation.

,- ". 20 C it

b I
16 cm rvtaC/.

C 0~c

0 I ! ,t ~ II IO 2
20
20 3oo00o
T
Tl0 PTTER STAR . SEKUNDER

FIG. .1.0, Innfl. * e f,ruktV•n- n n .. I aE , on,,.n krr

values of the teiperature gradient near the probe were

The quoted

calculated from E• 14,

A comparison of cases B and D, which pertain to nearly the same de-

gree of saturation, illustrate the dependence on gradient. These

results show that the effects of moisture transptrt are limited to

a relatively narrow range in terms of degree of saturation in moist

soil materials, for the probe design used. However, the previous-
ly mentioned comparative studies indicate that this in itself does
not guarantee that measured data are correct, as long as length/ra-
dius ratio for the probes are as limited as for the probes used.
For an increase in the length/radius ratio not to impair the prac-
tical usefulness of the probes, one may, as mentioned, have to re-
sort to probes consisting of heater wires only. This would result
in significantly higher gradients than in the experiments mentioned
here and much more severe problems in terms of moisture transport
and drying-out near the probe can be expected.

151
 C. Experimental arrangement

Based on experience from the investigations on the effects of mois-

ture transport, an experimental arrangement was developed at the
Institute for Cold Technology, utilizing six probes that can be ope-
rated simultaneously (22). This equipment was built before results
from the previously mentioned comparative investigation were avail-
able. A comprehensive series of experiments was conducted by the
author on dry soil materials, using this equipment. Later, a series
of experiments on six different soil materials at various degrees
of saturation was performed. The results from these experiments *ill
be discussed in the chapter on experimental investigations of con-
ductivity in soil materials.

The material samples were packed into cylirdrical steel containers

which were placed in a chamber where the air temperature was recor-
ded automatically, by means of a data logger and punched out on pa-
per tape. Figure 11s shows a sketch of the arrangement, while Fig-
ure 11b contains a simplified circuit diagram for the electrical
components.

Experiments with measuring the thermal conductivity of frozen soil

materials, using a heat conduction probe in the same set-up, have
to date not given satisfactory results. This may be due to the
previously mentioned non-linearity that occurs when water freezes,
as well as to the homogenity of the samples being disturbed near the
Y probe during the freezing process.

The actual measurement period in this test unit is only 30 minutes.

However, a stabilization period is required to obtain homogenous
temperature fields in the samples before measurements. For this
reason, the samples are usually prepared on the day before the ex-
periments and placed in the chamber at 4VC (or other solected tem-
perature) over nightq As mentioned earlier, the sample containers
are cylindrical and made of (sheet metal) steel. A diameter of 20 cm
was chosen, while the depth is 30 cm. This gives a sample volume
of about 7.5 litres, corresponding to sample weights between 12 and

152
15 kg. Preparation of six samples normally requires one man-day.
The next day, following the thermal conductivity measurements,
tests are made of the moisture content in the containers.

As mentioned before the results are coded on punched paper tape.

They can be decoded or printed out in the form of a table showing
temperature rise and currents supplied to the probes at given time
intervals. As an alternative, the paper tape can be fed into a com-
puter which calculates conductivities and plots the temperature rise.
An example of such a plot is shown In Figure 12, where also part of
the computer print-out is reproduced. The program was written by
Mr. K. Aflekt at the Institute for Cold Technology. Data reduction
by means of a computer results in considerable time savings, as com-
pared to manual reduction which involves an extensive plotting effort
following each experiment.

The equipment for probe measurement was developed by the author,

in cooperation with P.E. Frivik, M.Sc., Institute for Cold Techno-
logy. Mr. Frivik carried the main responsibility for the develop-
ment, including selection of components for registration and tempe-
rature control. Initial experiments were conducted by Mr. K. '1 itso"
Jonsen for his Master's Thesis in the fall of 1972. The series of
experiments with dry materials was performed by the author. Mr. Si-
gurdson of the Institute for Cold Technology has since assumed the
responsibility for sample preparation and conducting experiments
using this test equipment.

1I5
:53
 r14- ITr
A- r7rI I \I i <

IIr

IC 2?7ým i,,t

Fit J,
4
ýo~I ~ t Ag eorvaih pd a'niriin
i-r for '~tN~.; I. i

154
 Legend, u Ii.

A -Fan
B - Cooling olement
C - Oven
D - Pt element (Thermal sensor)
E - Probe
F - Material

a) Sketch of chamber with six sample containers and probes.

Termoolement : Thermal sensor

Hetetrad : Heater wire
Scnde : Probe
Oltronix : Power supply (trade-name)

c) Sketch of the probe with power wupply and racording equipment.

Translation of print-out, Figure 12.

*.. THERMAL CT:4DUCTIVITY - PROBE ,IEASURj•12T

i•XPERIMDT :o 5
PROBE No 4
INITIAL PROBE T*1,PERATURZ : 4.29"C

STARTING TZIPMATURE IN (Tr:ST) PERIOD : 13.80

FI:1AL TIP.ERATURE IIN (TEST) PERIOD : 20.01
TMIPMATURE 1IS3 IN (TEST) P2IOD : 1.21
AVMAGE CURRNT IN (TEST) PMIOD : -. 6508

CALCULATED CONIDUCTIVITY 2.097 ,/m•

155
I.. . . ....... ............ ......................

.......................................

*444*4
fl. ~ ~ .. 444
....... . *4...14
.. ......... **4**4*4

F44181. list

.1. 141 1

k I.

f'or b ViI t av I'cgt t7

(.-,7! vi . l

!0 t ,dt I i
hall ..*!t3 a t, v is t

BEST- AVAILABLE COPY

156
 4. OTT-I" TRAUISIE•T METHODS

From the previous discussion it is evident that the thermal conduc-

tion probe really does no", constitute a suitable solution if one
wants to reduce the effects of moisture transport. The method may
give short measurement times in comparison to traditional steady
state methods but the temperature gradients will be much larger.
The comparatively large sample volumes required are also inconvenient.
For these reasons it has become desirable to search for other methods
which can meet the combined requirements of small temperature gradients,
short measurement times and small sample volumes. '8ith regard to
sample preparation it may be suitable to take a cylindrical (sample)
geometry as a starting point. The types of test equipment with cy-
lindrical geometry that vrill be discussed here can also be realized
in planar configurations (c.ef. Rfs. 23, 24, 25, 26). However, only
test equipment utilizing cylindrcal geometry will be treated in the
following.

The various method which have been utilized are based on either
a programmed temperature variation on the cylindrical surfaces or
constant thermal flux. The latter case is most aprlicable for a
hollow cylinder when hoat is applied from its inside. As will be
evident in the following, thermal diffusivity can be derived from
data on temperature only, while computation of thermal conductivity
and heat capacity by volume always required knowledCe about both
thermal flu:: and temperature distribution.

A. Linear temperature rise

Cne of the simplest methods is based on a linear rise of tempera-

ture in the surface of the material under tesýt, when shaped as a
small (compact) cylinder. After a certain time, the temperature
difference between periphery and centerline in such a sample will
assume a constant value, while tLiperaturea throu.hout the sample
rise at the same rate as the surface temporature (i'iur'e 3).
This condition may be described as a quasi-statiýnary state.

157
If the temperature rise on the surface is given by

T (R, t) - To = Bt 16

the rise of temperature at each point within the sample, under

quasi-stationary conditions is given by

T (r,t) - To - -8 (R 2 - r 2) 17
4a

T is the initial temperature in the entire sample

0 is the rate of temperature rise (CC/s)
R is outer radius
r is arbitrary radius
a is thermal diffusivity (m I's)

The thermal diffusivity can be determined from meaaurcJ. t;r=peraturea

at the periphery and centerliae of the sample

a-. BR2

44T

AT is the temperature difference between periphery and cen- S

terline.

The maximum temperature gradient over the quasi-stationery period

is obtained by taking the derivative of eq 17.

TSR
- )R - 19
3r
2a

158 5.

"-A
 2C - 'C ,._

~~T1
. ' ... 2TC

Soo 100
Og0.5? 2CO 0
T!20, IS-K.i NT.F.

~'I
IIa'JrI'OIU•
[, 'i.al,
e.: I.rILt!. C p ip(rrI.:. -V
I- !'1 ( t .U1 P I P.&I
I.4~
in
".Wr~j r I.- it, )I if r' it C

The gradient can consequently be reduced by lowering the rate of

temperature rise and by reducing the sample diameter. The sample
diameter is also limited by the desire to reduce the time before
the quasi-stationary phase is reached. A lower limit for the dia-
meter of the sample is given by the requirement for accurate tem-
perature measurements and probleams associated with (precise) loca-
tion of the centerline temperature sensor.

For example, when considering a fine grain soil material within the
previously discussed critical range of saturation levels, the ther-
mal diffusivity can be estimated to 2-10-7 m2 /s.

If one requires a maximum temperature gradient below 1"C/cm and a

minimum temperature difference AT = I1C, one can use Eqs 18 and 19
to find the possible range for rate of temperature rise 8 with dif-
ferent sample diameters. According to Luikov (28) one can estimate
the times for quasi-stationary conditions from

FO M at/R2 - 0.2 20

159
 Fo is Fourier's number

The results of this example are summarized in Figure 14.

6C/sek
e. r I4000
sek
< 5/ 3000
Cr
2000 a. zT Io

cc L 4U: O
2 ,000

J -C 0 1 2 3 4 5 6 7 8!0.2 m

?1G. 14. .'illa'.te tc-neratutzti-'shrsl stik'-iete'- i en

5yjin.Ari5/,r•,'%'vv ne: ter.'n .' i*: ':A 'i i
a a 2 " i c. l .. e ,; 0," L.) T%. t2!nn
{ a s.asia tr:-.r-e :':a¢re .r ni " er , vi, rv', C.

-u,,,'e u
wit r: o .j;'.5j. j;:. a = 2 I(-/
' ,
!oLIfl .. ,..; ., 3 ?,C 2J. IL'*J'S.: (

One finds that the smallest radius which satisfies both conditions
is 2 cm. This leads to a quasi-stationary period of 400 seconds.
The rate of temperature rise which satisfies the requirements is
2 • 10-30C/sec, corresponding to about 7 0 C/hour.

This effect means that this method can be used for soil samples
taken with the 5-.4 mm diameter sampler normally used for geotech-
nical studies. Some trimming around the periphery -mill probably
be advantageous, but this is common practice in most normal studies
of samples of this kind.

One draw-back with this method is the relatively complicated auto-

matic temperature control circuitry required for each sample holder.
However, the small sample volumes represent a large advantage over
the probe method. Except for diameter, the sample volume also de-
pends on the length of each sample, which should be rather large to

6doO ele IsEG

160
avoid edge effects. The required length/diameter ratio can be
reduced by insulating the endI surfaces. If a length/diameter ra-
tio of 4 is selected, the r-oquired volume for the case mentioned
previously (dP 50 mm) is less than 0.5 liter. However, the re-
quirement for homogenity will preclude experiments with coarser
materials than sand in such a unit.

As mentioned previously, equipment of this type will only give in-

formation about thermal diffusivity. Szanto and Aguirre-Puente
(23) have proposed a modified version of this method, in which
also heat flux is recorded. In their test unit, the samples are
surrounded by a material with known thermal properties, e.g.
plexiglass. During the quasi-stationary phase, the thermal flux
through the plexiglaw pipe can be determined from temperature
measurements at both surfaces of the pipe.

When the thermal flux is knovrm one can determine both thermal capa-
city and conductivity for the sample material, as mell as diffusi-
vity. Aguirre-Puente give the following expressions for c0 and

co= q 21

O - q
22
411A T

Since this method is based on indirect measurement of the thermal

flux, accuracy will depend partly on the calibration of th. flux-
meter.

B, Constant thermal flux

(1972)
Shashkov et.al.Ahave developed a method in which the thermal flux
can be measured directly (29). The test samTle is bounded by two

161
concentric cylinders. The heating elements are attached to the
inside of the hollow cylinder. The outer cylindrical surface faces
a heat screen (guard) which is controlled by a signal f~rom a tempe-
rature sensor at the outside of the sample. In this manner, adiaba-
tic conditions (no exchange of heat) are ensured on the surface of
the samples. A sketch illustrating this method is shown in Figure 15.

II ,

, , SENTRAL VA,•M(,UOE
7/ -'SCN'Ut. HEAT SOUICE
ra * CONSTANT

AAT

/ , A IA-A11C GUARD

IBESTT AVAILABLE -COR~Y

' .'] . Ski.-se av apparat rur ¶r'i av or avterr vtr,ne-

j 162
 The method utilizes the principle of constant heat flux. However,
also in this case a quasi-stationary state will be reached after
some time where the temperature rises linearly (with time) and a
constant temperature difference AT exists between inner and outer
surfaces of the sample. In this phase, the temperatures are given
by 2
q R2 2Fo - 1 r2
T (r,t) - To = R 1 (1- 2
R - R 4 R2
22 33 2
R2 r R2
- 2 + 1 R 3
(in 2 2 i + )j
2
R R2 R1 R2 R1 4

RI, R., inner and outer sample radii

F , Fourier's number = at/R1 2
t , time after flux turn-on

The thermal parameters for the sample can be computed from the rate
of temperature rise a and the temperature difference 4T:

XR 2 R2 R2 2R2 I R
R2
2 (R 2 - k2 ) AT
1 2

2 2 2
a at (R2 R -2R 1I in

4 AT a2 25

) cP 8 q R2.. .
2 2

26

163
The temperature rise within the samples is determined by taking the
derivative of Eq 23 and setting r RI.

aT
-qbX
Dr R 227

where q is the thermal flux per unit area of the inner surface.
If the heat flux is measured per unit length of the cylinder (like
for the thermal conduction probe) one obtains

R 28
ar 2n R X
1

where q is the thermal flux per metre cylinder length. Figure 16

illustrates the temperature gradients that will occur for different
sample dimensions, if the temperature difference also in this case
is kept below 10C.
•00

• 00

20 1.
I , a

(I 2 • ' ' 1 "'

PS., 6.6, evmperatur.•trajcnt.er i at apoarat som sk~eadvt pi

rie. 15 VJi
vI r .i 1*i.2:or
o..i *ý yr
yir.lRMS.•0
Y~rl l't
t
l ,•
T '
u
•
. +l"p
PiJ 'VI.i.J
Pl
01. 1
~
l'.pl.l&i
JP
. L:,Itom iti
OJC lO tpl
1.|tU
:)I,1

The tAmneature gradients will evidently be much less than 1OC/cm,

except for very small sample dimensions.

164
As mentioned previously, this method has the advantage that the
thermal flux can be measured directly as current and voltage for
the heating elements. An axial heat screen similar to that used
in Kersten's cylindrical test unit should contribute to increased
accuracy in these measurements and ensure radial heat flow. The
fact that temperature sensors can be mounted directly on the walls
of the sample holder also represents an advantaged in comparison
to the previously described method where one of the temperature
sensors had to be centered within the sample itself. However, the
geometry is somewhat less favourable than the small (compact) cyl-
inders in terms of sample preparation. In addition, undisturbed
samples can not be used.

C. Periodic variation of temperature

This method is, like the preceding, based on a linear temperature

rise in the samples. In some cases, it may be desirable to avoid
a temperature rise with time during the experiment. This can be
accomplished by controlling the surface temperature of the samples
with a sinusoidal signal having a constant average value. Van Zee
and Babcock (1cj51) proposed such a method for measuring thermal
diffusivity in glass (30). The corresponding test unit can in
principle be constructed in the same manner as the unit developed
by Aguire-Puente, but using a different signal source. This is in-
dicated in Figure 17, which shows a test set-up of this type.

For such a stationary, sinusoidal surface temperature variation,

the temperature within the samples will also have a sinusoidal
variation arcund a -nstant .. ioan va.Lue, but wilh a curtain phase-
shift and attenuated amplitude. The thermal diffusivity can be
calculated either from the phase shift (radians) between the tem-
peratures in thp center and at the periphery or from the amplitudo
attenuation. OAppendix I summarizes the mathematical basis for this
method. For this reason, only the expressions for phaseshift and
attenuation are listed here.

165
 IkiNt -LA

b. 3ab~onwn Perio-lisk en .ra rvr sj ,

a. Limiter toprtrt g r Varmastr& .~v~,m.Ael ror

1, Ov r ., rýze . " ma
-mad

_T. Av.dy1
qjtl-550

.7 ~~au&WsyJndrislca iinn: fiter for Pili? av

n?r(v1rA r
IIl'Pi i'a

. 6C ,ma
It m.*1 r? r0 01
 Fhaseshift tan bei X 29
bor X

A A ['3 >ei )2+ (ber )r - 30

Attenuation

X -R wa 31

where w is the angular velocityI) of the temperature (radians/

sec), The relation between frequency (Hz or sec" 1 ) and angular
velocity is given by f = w/ 2n, while the cycle timne (period) to
is given by to - 2 n/w.

bei and ber are Kelvin functions, which are tabulated in mathe-
matical handbooks (e.g. (31), page 430).

Figure 18 shows the relation between X and phaseshift and attenu-

ation2• in graphical form.

-• .- O '4INGA J' *

- = - *4*~ ~or
Wt- V1NX1:E1.AS!!GHC!T raid./%

3 f IP•~ P~(Y'fA'NE
iGIl'F,'AA: DIF•'US WIVI

P1R
.If.RAOIUS
v
1AM•I.t$L
RADIUS

I 2 3 . 5 a '7 0

au il*-i1 O! O o,4d* Ie1,94ri ±t n.o.l iiIe,

1) Normally called "angular frequency" or "radian froquancy"

2) The English text in Figure 13 uses the word "damping" and
the ordinate scale for A is in Nepers. (Translator's notes)

167
 Such a diagram can be used for determining × from measured phase-
shift or amplitude attenuation. Thermal diffus.vity can then be
calculated from Eq 32, if period (frequency) and sample radius are
known. In cases where the attenuation is strcn:,, calculations
based on phaseshift will give the best results, as evident from
Figure 18.

Test unit dimensions may be selected based on requirements for short

measurement times and accurate temperature readings. As an example,
Figure 19 shows the relation between largest sample radius and pe-
riod of oscillation for attenuation values no smaller than A a 1/10,
for accurate temperature readings. As in the previous examples, the
diffusivity is assumed to be 2 .10-7 m2 /S.

2000

000, L ,,t

.1WWI

. )• j * . Xrvn ',-: A a/o.. 1, ~~9bt'd

,9f

c ' I

d~i,!;vr9~F . A.. a .. . i 'Jn' .

t U)1.1 Of U , a"

The sample radius can evidently not be much larger than about 2.5 cm
if one wants the period to be shorter than 10 minutes.

A prototype unit of this kind was built at the Institute for Cold
Technology in the spring of 1972, as part of the study reported here.
Figure 20 shows a sketch of the set-up, as well as a modified ver-

1.68
 -- - - -- -- - --

C,@*'0 04 4y

00*1

of *0kIo1-0 4 "f
II0,'

.. ..... ...

porlatilfl autoati

regiuin a tm~lqltre

ved tslakeilkuomatiSng
sgg~ectd mesiiod foi' auto'-
matio camYpalratdre Oontrol
b'ao~d om feed-bacOk.

F10, 20. prototyp av at. amparat rot' mfl~ir¶f l~v tri

Univel'Jiall of' Noi-w'y. rf, ondheim'i'I

TdoJimiuaL

'
169
sion with improved temperature control system. The sample holder
was a steel tube with inner diameter 60 mm, wall thickness 2 mm and
a length of 150 mm. A heating foil of type ESVA1) was wrapped a-
round the tube. To ensure constant average temperature in the sam-
ples, a 1/4 inch copper pipe was also wrapped around the outside,
separated from the sample holder by a layer of pipe insulation. In
addition, the sample holder was insulated from the environment by
means of pipe Insulation on the outside of the cooling pipe.

The periodic temperature variation was achieved by periodically va-

rying the power applied to the heating foil. The power was controlled
by means of a function generator which permits variation of frequen-
cy, amplitude and mean level (off-set). In turn, this signal con-
trols the power via a thyristor regulator which provides an AC po-
wer proportional to the control signal. The power level is varied
by means of a variable auto-transformer (Variac). Temperatures at
the periphery and center of the samples were sensed by copper-constan-
tan thermocouples and recorded om a potentiometer recorder having
a minimum full scale deflection of 100 uV, corresponding to 2.5 0 C.

~qgt nt
of the power level is required (via the auto-transformer) in order
to obtain a stable mean temperature in the samples. This balance
is reached when the power in the heating foil equals the heat loss
due to the cooling pipe. A better solution has been proposed, by
which the signal from the function generator is used to control the
surface temperature via a feed-back loop containing a proportional
integrating regulator, as well as the previously mentioned arrange-
ment with a thyristor control and auto-transformer. This will permit
automatic setting of the mean temperature by adjusting the signal
source (function generator).

If the periodic signal is replaced by a linearly rising (ramp)

signal, this test unit will also operate in tho same manner as the
unit described in the first part of this section. If the sample

holder is constructed for temperature measurements this unit will

I) Probably a brand name (Translator's note)

170
then correspond to the equipment developed by Aguirre-Puente. As
mentioned previously, such a test unit will yield information about
thermal diffusivity, thermal conductivity and thermal capacity in
the samples. This has also been achieved by Champoussin (1972) in
a planar test unit using periodic temperature variations (27). The
cylindrical configuration using sample holders made from a material
with known thermal properties can also be expected to yield this
(same) information.

A few preliminary experiments were conducted with the previously

described prototype in a silt type soil material at a relatively low
degree of saturation. The surface temperature amplitude was adjus-
ted to about 10C, while the frequency was varied in the range 10" 3!,z
to 10-4 Hz. Three experiments within this frequency range gave
approximately the same thermal diffusivity. This would indicate
that the measurements were not affected by moisture transport.

Based on an evaluation of the need for new test methods, as opposed

to the need for collecting more data on thermal conductivity in
Norwegian soil materials, further work on this unit was postponed.
Instead, the main experimental effort was -
ment and construction of the previously mentioned test unit where
six thermal conduction probes can be operated simultaneously. In
addition to this probe unit, three planar test units for stationary
measurements were built.

However, the new test method was adopted by P. Iloekstra at CRREL.

He developed the method further and conducted a series of experi-
-ents with un-frozen dry and saturated sand, as well as a sandy
clay material, at temperatures below the freezing point. Results
from these experiments and a description of the equipment used can
be found in a report from 1973 (32). In that report, the new me-
thod is recommended for replacing planar test unite when measuring
thermal parameter in moist soils when problems associated with mois-
ture transport can be expected. It is also claimed that the method

171
is suitable for measuring ti:erzmal Qiffusiv_.ty of fino grain mate-
rials within the freezing zone, despite the non-linearities which
occur. This is based in the possibility of using extremely small
temperature variations for the tests. For example, it is stated
that amplitudes as low as 0.20C can be used without loss of accu-
racy. (Thermistors were used as temperature sensors).

A closer look at the fundamental differential equation for heat

transport when thermal parameters are temperature dependent can to
some extent clarify this:

a (M(T) aT ) u co(T) aT 33
ax ax rt

By completing the derivative in the left hand member one obtains

OX aT +X ~2r' cp MT aT

ir the temperatutre gradient !s very low, one can prooabqyngic

the term marked (1) in Eq 34. The equation is then of the same
form as the linear differential equation which formed the starting
point for the mathematical treatment of the method.

The fact that the thermal conductivity and capacitance are tempera-
ture dependent will cause deviations from the linear equation. How-
ever, if the variations due to temperature are sufficiently small
one can possibly neglect also these effects. In this context, varia-
tions in the apparent heat capacitance will be most important.
Since this parameter may vary by more than a factor of 10 for a tem-
perature variation of only 0.5'C near the freezing point (0 to -1 0 C)
it would seem that this condition poses unrealistic requirements on
(sufficiently) small temperature variations.

172
A more complete mathematical analysis of the effects of these non-
linearities may give a better understanding (clarification) of
these matters.

D. Conclusion

This overview of transient methods for measuring thermal parameters

in soil materials shows that there are several solutions (methods)
which in many ways are more useful than (provide advantages over)
the conventional thermal conduction probe. The comparative studies
which have been initiated at the Institute for Cold Technology in-
dicate that (such) probes of common construction will not yield the
expected accuracy, even when no moisture transport effects are pre-
sent. However, for dry and saturated samples it appears as if probes
configured as heating wires can give good accuracy without unreaso-
nable requirements on sample volume, while in moist (not saturated)
materials this configuration is almost out of the question due to
moisture transport effects. On the other hand, extensive studies
will be required to establish whether the described methods invol-
ving periodic temperature variations will emerge as a more suitable
supplement to conventional steady state measurements in planar
test units.

173

- ~ 4 _j
4 REFERENCES - CHAPTER IV

1. E. Berge, 0. Aarvold: Some physical properties of grass silage.

Meldinger fra Norges Larilbrukshogskole, 53 (1) 1974.
2. Beck: Transient Determination of Thermal Properties. Nuclear
Lngineering and Design, 3, 1966, pp. 373-381.

3. 0. Krischers Die Leitfahigkeit des Erdbodens. Beiheft zum

Gesundhets-ingenieur. Reihe I. Heft 33, MUnchen (Munich) 1934.
4. W.O. Smith: Thermal Conductivity in Moist Soil Soil Science
Soc. Am. Proc. 4, 1939, pp. 32-40.
5. A. Watzinger at al: Unders~kelse av masmautskiftningsmaterialer
for veg-og jernbanebygging. Medd. fra Vegdirektoren 1938
no 6.
6. M.S. Kersten: Thermal Properties of Soils, Univ. of Minnesota,
Eng. Experiment Station, Bull. 28, June 1949.
7. W. Woodside, J.B. Cliffes Heat and Moisture Transfer on Closed
Systems of Two Granular Materials. Soils Science, 87, 1959, pp.
75-82.
8. W. Woodside, C.M. A de Bruyn: Heat Transfer in a Moist Clay.
Soil Science, 87, 1959, pp. 166-173.
9. F, Do Ponte, P.E. Privik: Automatic Control of Guarded Hot
Plate Apparatuses, Institutt for kjoleteknikk, Trondheim, NTH,
1972.

10. E. Brendeng, P.E. Privik: New Development in Design of Equip-

ment for Measuring Thermal Conductivity and Heat Flow,
Institutt for kj~leteknikk, Trondheim NTH, 1973.
11. E. Saare, CG. Wenner: Virmeldningstal hos olika jordarter.
Statens namnd fdr byggnadsforskning. Handlingar no 31, 1957.
(Ref. B. ST&lhane and S. Pyk 1931).
12. E.M.F. van der Held, F.G. van Drunen: A Method of Measuring
the Thermal Conductivity. Physica 15, (10) 1949, pp. 865-881.
13. D.A. do Vries: Het warmtegeleidingsvermogen van grond. Med.
Landbouhogeschool Wageningen, 52, 1952.
14. F.C. Hooper, FR. Lopper: Transient Heat Flow Apparatus for
the Determination of Thermal Conductivities. Heating,Piping &
1 Air Conditioning, Aug. 1950, pp. 129-134.
15. H.S. Carslaw, J.C. Jaeger: Conduction of Heat in Solids.
Sec. ad. Heat Oxford Press, 1959, p. 345.

174
16. D.A. de Vries, A.J. Peck: On the Cylindrical Probe Method of
Measuring Thermal Conductivity. Aust. J. Physics, 11, 1958,
pp. 255-271.
17. J.H. Blackwell: A Transient Heat Flow Method for Determining
the Thermal Constants. J. Appl. Phys. 25, (2), 1954, pp. 137-144.
18. J.H. Blackwell: The Axial-Flow Error in the Thermal Conductivity
Probe. Can. J. Phys. 34, 1956, pp. 412-417.
19. W.T. Kierkus et al: Radial-Axial Transient Heat Conduction in
a Region Bounded Internally by a Circular Cylinder of Finite
Length.
Dept. of Mechanical Engineering University of Calgary, Calgary,
Alberta (Preprint).
20. P.E. Privik: Institutt for kjoleteknikk, NTH 1974. Upubliserte
resultater.
21. W. Woodside: Probe for Thermal Conductivity Measurement of
Dry and Moist Materials. Heating, Piping & Air Conditioning,
Sept. 1958, pp. 163-170.
22. P.E. Privik et al: Sondeapparat. Rapport fra Gruppen for
termisk analyse av Frost i Jord. Institutt for kj~leteknikk,
Trondheim, NTH, 1975.
23. 0. Krischer, H. Esdorn: Einfaches Kurzzeitverfarhren zur
gleichzeitigen Bestimmung der Warmeleitzahl, der Warmekapazitat
und der Warmeeindringzahl fester Stoffe, VDI-Forschungsheft.
450, 1955, pp. 28-39.

24. K. Katayama et al: Thermal Properties of Wet Porous Material

Near Freezing Point. ASHRAE Journal April 1973, pp. 56-61.

25. C. Codegone et al: Desonant Thermal Waves in Insulating Slabs.

fIR, Commission 2 (Trondheim)1966. (Saertrykk)
26. J. C. Champoussin: Determination Simultance des caracteristiques
termocinetiques des solides. Int. J. Heat. Mass, 15, 1972,
pp. 1407-1418.
27. I. Szanto, J. Aquirre Puente: Etude des caraýtdristiques thermiques
d.s milieux por.ux fin% hutmi.dm.
Proceedings of the XIIIth International Congress of Refrigeration,
Washington D.C., 1971, vol. I, pp. 751-757.
28. A.V. Luikov: Analytical Heat Diffusion Theory. Academic Press,
New York 1968, p. 312.
29. A.G. Shashkov et al: Thermophysical Properties of Thermally
Insulating Materials in the Cryogenic Temperatuire Region. Int.
J. Heat Mass, 15, 1972, pp. 2385-2390.

175
30. L. van Zee, C.L. Babcock:A Method for the Measurement of
Thermal Diffusivity of Molten Glass. J. Am. Ceram. Soc. 34
(8) 1951.
31. M. Ambramowitz, I* Stegum: Ed. Handbook of Mathematical
Functions. Nat. Bur. of Standards. Appl. Math. Ser.,
New York 1969, p. 374.
32. P. Hoekstra et al: Measuring the Thermal Properties of
Cylindrical Specimens by use of Sinusoidal Temperature Waves.
Cold Regions Research and Engineering Laboratory, Technical
Report 244, Oct. 1973.

J~J

176
-A
 CHAPT2R V

EXPERIMENTAL INVESTIGATIONS OF THERMAL CONDUCTIVITY IN

SOIL MATERIALS

Previous chapters reviewed theoretical foundations for calculating

thermal conductivity in soil materials. The theoretical investiga-
tions showed that purely analytical methods only can be expected
to work for saturated materials in which conductivities of soil com-
ponents differ little from each other. In such materials, varia-
tions in the microstructure are of secondary importance for the
resulting thermal conductivity. On the other hand, for partially
saturated materials, the presence of air with low thermal conducti-
vity results in hiJh sensitivity to differences in microstructure
within the system. Since these properties are difficult to incor-
porate in a mathematical model for the thermal conductivity one
must in such cases resort to an emperical approach based on thermal
conductivity measurements on selected soil materials.

There is a relatively extensive amount of information on thermal

conductivity measurements in soil materials available in the lite-
rature. The first major study, presented by 0. Krisher in 193i.) (1),
was followed by a study of American soil materials by 7.0. S.nzith
(1930, 1939, 1942) (2, 3, 4) and studies of some iorwegian soil ma-
terials by '.atzinger, Kindem and Michelsen (1938, 1941) (5, 6). In
1949, M.S. IKersten presented an e:itensive series of experiments on
a number of soil materials from Alaska (7). This study is still
the most valuable reference for empirical studies of thermal conduc-
tivity in soil materials. Among later contributions, the work pub-
lished by 1. van Rooyen and H.F. Winterkorn in 1959 constitutes
the only systematic investigation that can be compared to :,ersten's
work in terms of extent (8).

In addition to these studies, a number of scattered single ex-

periments have been published (9, 10, 11, 12). The lack of experi-

i. -. ,
 mental results pertaining to Norwegian soil materials is now being
improved on by a series of experiments initiated at the Institute
for Cold Technology, NTH. This investigation is part of the acti-
vities within the committe for frost on the ground. Some of the
results will be discussed in the following section, which is devo-
ted to measurements on dry soil materials.

1. DRY SOIL MATERIALS

A. Introduction

In a dry soil material, the ratio between thermal conductivities

of particles ( X" ..5w/mK) and air in the pores (A * 0.024 w/mK)
is in most cases larger than 100. Based on limits (bounds) for
the thermal conductivity in composite materials derived (described)
previously, one find that this implies a very high sensitivity to
differences in the microstructure of the material. Figure 1 illu-
strates this by showing the Iashin-Shtrikman limits for a two com-
ponent system having a conductivity ratio of 100.

•,' 50 r
'I0

20

0
0 0.2 0,4 0.5 0.• •1.0
PART IIXKELV0LUW ANOEL
VOb.diMgF*ACTIO OiP PADICIII•|

8. On7, I11 1t itU r I;

178l.
:,~~':•• O , ' tl 0 •••:r~• :L.U"L.
For a relative particle volume 1-n = 0.60, the conductivity at the
upper limit is at least 7 times that at the lower limit. For lar-
ger conductivity ratios, the distance (spread) between limits will
be even larger (wider). As mentioned earlier, these limits repre-
sent extreme differences in microstructure. Even if such differen-
ces are not likely to occur in soil materials, the wide spread be-
tween these limits still indicates a high sensitivity to differen-
ces in the microstructure for dry soil materials.

The previoulsy mentioned conductivity calculations for regular

configurations of spheres can give an idea of the effect of par-
ticle conductivity on the (effective) conductivity of dry soil ma-
terials. Results obtained by Wahao and Kato for spheres arranged
(packed) in an orthorhombic configuration (n * 0.395) are reproduced
in Figure 2 (•3).

'A,

'001

GC
VWiC!V'TY "ANO

IVIO N i
a-lk biv ll U.14 I twa ~ v r

PIz O, j# , (•,,"%I
•k a,
l . IA. , to , I ":
'! . , '

For the ratios between component conductivities normally found in

dry soil materials one finds that the sensitivity due to variations
in particle conductivity is relatively small. As an example, for
a conductivity ratio of 100, the particle conductivity must increase
by more than a factor of 10 before the (effective) conductivity in-

179

i.
 creases by 50 percent. By comparison, when the conductivity ratio
is 10, the same relative increase in conductivity requires leas than
a three-fold increase in particle conductivity,

This implies that one can expect two characteristic properties of

the conductivity in dry soil materials:

1. Considerable sensitivity to variations in iictcstructure

2. Low sensitivity to variations in particle conductivity.

B. Measurements performed by Smith

W.O. Smith has presented (results from) thermal conductivity

measurements in dry soil materials, first in 1938 and later in 1942
(2, 4). Before the experiments, the soil materials were dried in
z
I an oven at 105*C for 24 hours. Results from these two series of
* experiments are reproduced in Figure 3, where conductivity is plot-
ted as a function of porosity. The latter is computed on the basis
*. of dry density and specific weight measured for the mineral par-
ticles.
0.1.

04

I S

0.2

•.+.0 * ' 4*

O e Slf*lM 4 *

0.1
0.2 0.3 0. 03.
. of 0.1 10

PT ,3 . 1 .l.I ur LIV Vayr',ij - n r"n

@A.! , ",V kii iiV o•r+t,' rIn I ..r
1)' Utt, l 50•
+,:l1 1'ý.* h'i
.,5,*"-1
! , "Y
+++ 1: 5+.I ,', Itt+" LP '+

18,
180
j
As can be seen, these results show a variation of about t20 per-
cent around a mean curve which can be drawn as a straight line in
the log/log diagram. Variations among the results can be ascrib-
ed to differences in mineralogic composition as well as different
textures (microstructure).

C. Investigation performed by the author

An investigation of thermal conductivity in dry soil materials

performed by the author at the Institute for Cold Technology in
the spring of 1973 can contribute to a clarification of the effects
of the latter two properties. The measurements were performed in
the probe test unit described in the previous chapter.

Measurements were made on 16 different materials, 11 of which were

natural soil materials while the remaining 5 were crushed rock ma-
terials. For each material, experiments were performed for three
or four packing densities. Specific weights were determined by (at)
the Institute (Department)of Geology, NTH, for all but a couple
of the materials. quartz content was determined by (at) the same
institute, using DTA 3quipment. In addition, particle size di-
stribution curves were obtained for all materials.

The experimental results are given in Table I, along with calcula-

ted values for porosity, as well as specific weight and quartz
content. Particle size distributions are shown in Figures 4, 5,
6 and 7. The symbols used for esch material when plotting the
following figures are also shown.

Figure 8 shows results for four natv-ral soil materialn, all having
quartz contents in the range 4i to 53 percent. For those four
materials one can expect the particle conductivities to be about
the same, about 4.0 W/mK. For each material, the results follow
a straight line in the log/loi. diagram rather closely. Variations
from material to iaterial are .i.rnitficantly larger than the varia-

181
i i i , i i.i
'1. 1V,r I 7

I d 10
ow nvr .I~
0

*Vr silt , . n d:wd

50 1 I 20

I ":ht7

?r720£roo 0'4o

TABLE 11

192 02ý

a. Old0 :.O '.27

TABLE I. THERMAL CON;DUCTIVITY NEASURZIENTS ON DRY SOIL -LITERIAL5.
A. NATURAL SOIL MATERIALS.

Translation of terms:
JORDART Soil material
KVARTSINNH. 0, Quartz content %
SPES. VEI,:T kg/m3 Spec. weight kg/m3
TORR R. VEKT Dry density
POROSITET Porosity
LEDN. EVNE \V/mK Conductivity W/mI,
Tyholt leire TyholtI) clay
(pulverisert) (pulverized)
Folio silt Folio )silt
pulveriser t pulverized
Klumpet Lumps
Vormsun silt Vormsund ) silt
Tyholt fincand Tyholt )1fine sand
Jessheim JessheimI
finsand fine sand
Skanir Skan~rI)
finsand fine sand
Boskorp BoskorpI )
finsand fine sand
Tiller sand Tiller1 )sand
Tyholt TyholtI)
fin grus fine gravel
0Otfold ostfoldI)
TrUS gravel
Troms grus Troms I)gravel

1) Locations where material samples were obtained (Translator's note)

_ __ _ _ _ _ '

183I
TABLE I. B. CRUSHED ROCK MATERIALS.

MATERIALE Material
KVARTSINNH. % quartz content %
SPEC. VEKT kg/m 3 Spec. weight %
TORR R. VEKT Dry density
POROSITET Porosity
LEDN. EVNE W/an Conductivity W/mK
Kunst kvarts Crushed quartz
Knust glimmerskifer Crushed mica shale
1
Crushed greenstone )
Xnust gronnstein
Oppland pukk Oppland macadamZ)
0atfold pukik ,0'stfold macadam2 )

1) A mineral related to gabbro and diabase (Translator's note)

2) see note under Table I. A.

?^•XLL 1. ANV!1¶ Dn'UAP'TC,

" ,,.1
A1?'q !h,\TS 'tH K ) 1'•,' II, KT
4q/N'*., V'J1,AI•Tr-T W I•/Am
tLIE;X ,V.-M

iit•• vrta .37 a

'' o 17)0 j ,0.."5) {'
100 65I ,31?
1900 0.H },•.

ai07
,4 ii.r ok I.for50 17'0 *,,,,
179C 0,.I,'
,. l j ,!0 ,, •" 0.•9

A,0-Ito n I 0 e 3

PIa

. 6. • O .s60 )2•
Io ,31K33)
KN 312680 1"920 1( 1'21
1 0 ,.,.1.

190C 3,:i1

184
 r140
01` W0411 aa to to h a 1I 1,4" 1 1"

03

I- Z

451410 Al.4)81

Fin
. f ikt~exury-ir Cor U-4s f Inkornsoivho J 2:,. Vr~a

UP1f n 1mL~
Ulf .1 8 1 ~i011 L.*
iona 1." J Ll
94hro j jpýA4."W'.

ff-M
)11 lvls ia e1l01,1 ___p~~ j&0 __1J ' dJ4

V1, O -f-l'-
UM10 4/jou to 30 I V 1 . . 1 1,
i -' -l T ** T - 8-

CY.

ACA

sO00,110 Al 4101

195
 ý014 IN
*di.. too ~
%.1... 3%)o 1A
L; q j

go 0

Q101 (0.01 11)0. .44 1 1 401 ....

.. I
ne. Ovate ,Agpl4 jisel1e k4104,P*4

108 4101"8 437PA

1

I!.6. qiktvkurvor for fire iruonat.riaaor Cra varnimo

IV Il44sa'8 I.% VOI100 10 0

to 14 1 I 'd- ~* I V. P

*300 ~r~rr7!
~ 4

14I

toI.
`7t~

0-. IOU
0.001, e w '0.149 .4 C4 1).. 441)

! L0AN
.0IJ80l KI'.PW
, ,ýN T fu

3
7 .¶iktakurvur for- remn knustfv bur~art or tri~ nannip ng. uAr-
* 1 ke.06 aIO
;.J J8Ift Af4djo jcpjb1 g ,,I IIIe

186
4 ~0 5'

0
0.2

0.0

0,2 0. 04 05 i05J.8 0.5 1.0

POROSITE!

L*dnir'I-&VIlet' nv rire ti'rre jorlartur tneJi ca.50

pro,

S 0.1" 0. 3 0.*l

.&!.enint-sevnen
Iz. av t~rr* j0:IILrter OK knust# btfrg:.r'tel,
p~eI.I.CI Sam~ Vi'nKSjonf ay pr-.1111tuIrl. (.'&J.14cIt C v t t a
of dr s fI 4
Puti 0,i1-n~h14 'O1# ...
ro : nrit4 1: JnDto I~ r
*4aturJige joJar4t4er, &Knustbo0FI!o
NlaturaiJ uoý..C in'
1-0~iid

S% 14VARTS

1 *.3

0.2~I0.

02 0.3 D.& 103 0.6 03.0

T
P400'
k 7ET

Fla, J.u. LaJInilnppsevn1W 4LV t~rrj fl.jj1gtjujia0r'mal<arts0I~nnhvit

gintyjdt Vei Is *o ate~yMb~la.I.
AVt.. can!., 1-Iit&II
)If 'Irv magjiO? .j J ut. .' qoI'le t ant 01tdtl.'.? '
&.14 .:S

K ~ ~ ~ ~~1.87-'-....--
tions around these straight lines for each material.

Differences between conductivities in these materials, which here

amount to about 20 percent around the mean curve, can mainly be
ascribed to differences in microstructure. However, no direct re-
lation seems to exist between particle size distribution and such
differences in microstructure. The four soil materials in Figure
8 vary in mechanical composition from silt to single grain size
fine sand and sand, without any traceable systematic variation in
condcutivity.

However, a comparison of results from natural soil materials and

crushed rocks show a clear tendency towards higher conductivities
for the crushed materials. This is made evident by Figure 9, where
all test results are shown together, using different plotting sym-
bols for crushed materials and natural soils.

The difference in particle form between crushed rocks and natural

soil materials evidently results in systems with significantly
different microstructures.

A more detailed scrutiny of individual results also show that dif-

ferences in microstructure havea significantly larger effect on the
(effective) conductivity than does the particle conductivity. Figure
10 shows results plotted with different symbols for different ran-
ges of quartz content. These sy;abols are defined next to Aigure 10.
There is no obvious systematit variation of conductivity as the
quartz content increases. In fact, the highest conductivity for a
certain porosity is obtained for the material that has the lowest
quartz conteint.

The low sensitivity to variations in particle conductivity makes

it possible to represent the conductivity of dry materials by means
of empirlcal relations containing only porosity as a variable.
The effects of microstructure can be accounted for by distinguishing
between two groups of materials, natural soils and crushed rocks.

188
As shown, the experiments give no basis for expecting variations
within each of these groups due to differences in microstructure.

D. Empirical relations

Chapter II reviewed a number of empirical relations which could be

fitted to measured results for dry soils by suitable choice of
"form factors". Purely empirical correlations between conductivity
and porosity could also be used. However, these empirical corre-
lations should be consistent in the sense that they satisfy the
physical bounds (limits) of conductivity for the material combina-
tion at hand:

n - 1.0
X X'2 n - 0.0

On this basis it may by suitable to use as a starting point a semi-

empirical equation which satisfies these conditions.

de Vries has used Maxwell-Fricke's equation to represent measured

data for soil materials (9). This equation has the form

*n + (1-n) F 2X
n + (1-n) F

The "form factor" F is given by

3 -
F -1 + ( - )gi 1 - 1,2,3

1* + g2 + 93 1.0

189
iA
Values of the factors g1' g. and g 3 formally refer to definite
geometric relations for ellipsoid-shaped particles. However,
as mentioned in chapter II, the equation is based on approxima-
tions which make it impossible to derive these geometry related
parameters from the particle shapes in an actual material. Thus,
g1' g. and g 3 should in practice be regarded as parameters avail-
able for fitting to empirical data.

de Vries arrived at his set of these parameters by fitting (the

equation) to measured values of diffusion in soil materials. This
case, which corresponds to the particle conductivity being zero,
results in a particularly simple form for the expressions for con-
ductivity and form factor F.

Based on this, de Vries selected g, = g 2 = 0.125 and g3 = 0.750.

This gave good agreement for water saturated soils but values which
are 25 percent too low for dry soil materials, according to de Vries.
Also, the semiempirical nature of the equation does not give rea-
sons to expect equally good fits over this wide range of conducti-
vity variations.

Figure 11a shows results obtained from Maxwell-Fricke's equation,

using the values for g 1 ' g 2 and g 3 suggested by de Vries, together
with experimental results for natural soil materials and crushed
rocks. Curves are drawn for conductivity ratios of 100 and 200,
respectively. Figure 11b shows curves fox the same conductivity
ratios but for gS a g 2 a 0*10, g 3 2 0.80, which results in a bet-
ter fit to the experimental data.

The latter set of g-values are thus more representative for the
conductivity of natural soils. If the conductivity ratio is set to
125, one obtains the following semiempirical expression for the
conductivity of dry, natural soil materials*

n + (1-n) 6.65
/0.024 -n + (1-n) .053 20%

190
 A0 -

XA -200

0.2 0.) 04 01 0 7 -1 1 'n 2 0 1.0O q (it o a 1

l'II4drmgI• Q' .i.rd. "'a c ,'r v :n ~ :*a u'

As mentioned in chap~ex Iii, the specific weights of soil mate-

rials
mean ofvary
Ys within relatively
= 2700 kg/r 3 . If narrow limits
this mean value(about+
and dry•%)density
around are
a

used to compute porosity, this uncertainty will to some extent

affect the error (confidence) limits for calculated conductivity,
iue to the relatively large uncertainty in the empirical relation.
Eq 3 can thus be transformed to a function of dry density without
significant change in the error limits. The semiempirical equation
for conductivity of dry natural soil materials will then be of the
form

0.35
y + 6_.7 20% S

where Y is dry density expressed in ka/m 3 .

F'igure 12 shows results from measurements on dry soil materials, "

compared with values computed from Eq 3. In addition to results
obtained for Norwegiavl soil materials by the author, data measured
by Smith on American soils are included. It is evident that most
results fall within error limits of +20 percent.

I/

20.04
.
 U.'!I

W/* .,,.. ,•. /'7

/ / /'K,

0.1. r*
l*-O2; / ," / 1. I'

0 '

t0 0.1 02 o0• 0& 0-1

b4iASUA10 CW-'1,vr, 'Y

F.
IQI ? 3•.3, ui n iknirlij t e .. UPI ! ± rLnin l.uVru .i ,l

For crushed materials it can be shown that a semiempirical equa-

tion of a different type gives a better fit to the rise in con-
ductivity. Figure 13 shows a comparison between results obtained
for crushed soil materials and Bruggeman's equation for (fill) ma-
terials having spherical particles. Data from this equation are
shown for two conductivity ratios, X 2/xlv 100 and 200. The con-
:. ductivity is given by the following relation

One finds that this equation gives a relatively good fit to the
,• experimental conductivity data for crushed dry materials, if a
conductivity ratio of 125 is selected. However, due to the
complicated structure of the equation it is more convenient to
use a purely empirical relation for the mean curve. The follow-
ing exponential function is found to give a good approximation to
the conductivity in dry crushed rocks, within the porosity range
considered

S0.039 ' n + 251 6

192
However, not consistent (with all
this equation is parameter val-
ues).. Thus, it should be regarded as an approximation to Br-ugge-
man's equation (Eq 5) which meets the physical boundary condi-
tions mentioned.

W/m10 K --

0.4 'l

120

0.1

0.2 0.3 0, 0,• O. 0,9 10

F•0I$ 'ET

*'P•
13., lI' • .mai•(l11fl,,-,',;, t'u, I ury11 r'. vriP l~arpil~ 1 Ii.'li
,tU•
I'. Ill. IIt ,'C'l' J
•a ".I l~t 'll. iii J llfl~'ltl. I M jllllli ~ ni

E. Stab.lilzed materilals

van Rooyern and Wi-nterkorn (1959) (3) and later Farouki. (1966) (11)
have published (results from) conductivity measurements on dried
soil materials. For some materials, these results show much
higher conductivities than even the "highest values obtained at the
S.........
.. . '.. . ..... .. . ,, 0. ..--0 -- '- S-- 0 -. O .- :i',s-
1 ,~.; -0, -- , -•-: • • .•
Institute for Cold Technology. Those measurements were performed
with thermal conduction probes having significantly smaller diame-
ters (about 1 mm) than those used in the previously described ex-
periments (about 2 mm). The experiments were performed at various
degrees of saturation and for different dry densities. Only the
measurements performed in dry materials will be discussed here.
For such materials, the small probe diameters used should not affect
the results.

193
van Rooyen found exceptionally high (conductivity) values for
materials with some-what special properties, where a certain
amount of clay was mixed into otherwise sandy materials rich in
quartz. These materials were used as fill around high voltage ca-
bles and high thermal conductivity was specified in order to pre-
vent overheating even when high powers were transmitted. The
materials were called "thermal sand". Figure 14 shows results
obtained for these materials, along with (results from) van Rooyen's
measurements on one type of quartz sand and crushed quartz, as
well as measurements on crushed quartz performed at the Institute
for Cold Technology.

As can be seen, van Rooyen's results for quartz sand and crushed
quartz are close to the results for crushed quartz describet ear-
lier, while his results for the two special materials show sig'nifi-
cantly higher conductivities. Of these materials, "thermal sand I"
had the largest clay content (6.0 percent particles < 5 um), while
"thermal sand II" contained 3.2 percent clay. Particle size distri-
"butions for these two materials are shown in Figure 15.

•, 0.... .. . , "-
2.0

1'o2
0.4

"0. a

0.1 0.? 0.3 0.4 0.5 0.5 01 1.0

PuaI•i t,

i,•]
,l, L~utninj:=•w~nvn ity "Vhvrnt). ,,nt" sl¢,,,un•.ikiiet rMtj

of
;,,,.aj .I.t .'t-'100
l 4rd '1''aLSV.V ,"'Iifrt td a I
(I 4Q r I '11111'1& O j¼ 4'I I- it
q~

I~i'Li

194 J
 I m. ', P I . ' 1 . T

, ,o 106 ". 1f
0 A. , 11,4 NI 1

' .....- I

J... .-I ... .... . ... ... I ., •... I,..

... .... .. ., - , -- i o

"* L . & ..- L". . ... ....... "

,, . . .-

r~'.a' LOLi~i
"tiliriD CU III

van Rooyen al.so investigated the effect of adding Itaoline clay

to sand containing 94 percent quartz and having one single par-
ticle size& For this mixture, the conductivity of dry material
* increased significantly when the clay content increased. The con-
ductivity increased from 0.54 '.1/inK for I percent kaolinite to
oT82 ',V/m1Ffor 3 percent kaolinitel while for 10 percent kaolinite
*. added the measured conductivity was as high as 1.15 .V/rnC. Dry
density is not given.

Farouki (1966) has also investigated the effect of various addi-

tions of kaolinite clay to sands with high quartz content (11)a
Also in his case, the thermal conductivity was measured by means
of thermal conduction proben, for various densities and water con-
tents The results for dry materials with a density of 2100 ktg/rn.3
showed a maximum conductivity for a kaolinite content of about 8
percent, while the conductivity decreased for further increase of
the kaolinite content. This maximum value was measured to be as
hsl/ae
high as 2.5 for a dry density of 2100 kg/w 3, while the conduc-
tivity was found to be 1.0 for the same density but with no
dres/mK
kaolinite added.

195
In all these cases, the conductivity was determined for materials
which were dried after kaolinite was added. Farouki also reports
ono test performed on a dry material into which dry kaolinite
was mixed. This mixture showed a significantly lower conductivity
than the corresponding materials which were mixed under wet con-
ditions and subsequently dried. Measured values for a dry density
of 2100 kg/m 3 were 1.25 and 2.5 W/mK, respectively.

"These results indicate that the extreme(unusually high) values of

conductivity found in dried samples containing kaolinite are due
to microstructural conditions.

If one assumes that the clay particles initially are in suspension

within the pore water in saturated samples, it is possible that
these particles concentrate near the points of contact between
sand particles during the gradual drying process. The plate-shaped
clay particles would then form efficient "thermal bridges" between
the larger (coarser) particles, as indicated by Figure 16.

1!
iiw

4
ItO 16- Z10 aoa v t, irustru Wurmn i toft%*t Rin met i*-
8"11011r~ ~i r
UV fi'd L , SIt~ Ia) ~
00 *i)P1J tP'"W. L., r.

Barden(1973) has recently showed,

by means of electron microscopy,
that such clay bridges exist in dried samples of soil materials ha-
ving relatively low clay contents (14). Then dry components are
mixed, the microstructure would obviously be different, with a more

196
random distribution of clay particles. The occurrence of an opti-
mum clay content can possibly be explained by assuming that in-
crease of the clay content above a certain limit will result in
the sand particles becoming separated by "cushions" of clay par-
ticles. This hypothesis is supported by "packing" experiments per-
formed by van Rooyen on samples with different clay contents. When
the latter increased beyond 8 percent, the densities decreased con-
tinously for the entire sample (mixture) and thus also for the en-
tire "sand skeleton'"2). Up to that limit, the density of the entire
mixture increased, while dry density based on the sand alone was
nearly constant. In this range one can thus assume that the sand
skeleton is independent of the added clay, since the clay particles
are deposited in the pores. When the clay content exceeds 8 per-
cent, the clay disturbs the sand skeleton and separates the cand
particles from each other.

In natural loose deposits with a more even particle distribution,

no similar concentration of clay particles can be expected, at least
no to the same degree. However, different additives used to sta-
bilize road building materials, etc (lime, cement) can possibly
cause effects on the thermal conductivity for low levels of satu-
ration similar to those shown to occur when kaolinite is added (to
sand).

F. Conclusion

To summarize, this investigation of rmeasured conductivity data

for clay soil materials shows that differences in microstructure
indeed gives large variations in conductivity for such soils, as
was indicated already in the introductory discussion. rhi:. is par-
ticularly evident in Figure 17, where the results given in Figure
14 are plotted in normalized form between the Hashin-Shtrikman li-
mits for materials having particle conductivities corresponding

1) One that gives maximum conductivity.

2) The sand particles are separated further from each other by the
clay "cushions" (Translator's notes).

197
 to the minerals with high quartz content for which the measurements
were made.

Based on these results and previously described measurements, a

range of expected values is also indicated (in Figure 17) for the
conductivity of dry soil materials. Even if this range is a small
portion of the (region between the) Hashin-Shtrikman limits, this
still implies a considerable variation in conductivity for a given
porosity. However, if one neglects rather specialized materials
of the type "thermal sand", the expected variations will be much
smaller. The "center of gravity" falls some-where between the low-
er boundary of the bounded region and the points representing
measured data for crushed quartz and quartz sand. It was shown
earlier that the "center of gravity" for natural soil materials
can be represented by an empirical relation between porosity and
conductivity, within error (confidence) limits of about +20 percent
(see Eq 3). A similar equation was found to represent results for
dry, crushed rocks, with about the same error limits (see Eq 6),
In keeping with the introductory discussion, particle conductivity
variations can be neglected in both relations.

300 -

, JOK • 7 .---

10

0
0 0: 4 g O6 l 01

F.' . A t, .*1taine vltvin ,iv •d'~ o -~{•,,•B o':(! ,...

e.' n'jw,/ 9ll j'rt~g •,,"4 jp $l,/J!.I.j• Q

••11 i'i ,',,.•*thI .'; ,

L'o'pa ld Co f'; f98

198
A"
 2. SATURATED MINERAL SOIL MATERIALS

A. Introduction

A saturated soil material can be assumed to consist of two or

three components, represented by mineral particles, water and/or
ice. However, the mineralogic composition can vary considerably
from particle to particle and is often related to particle size.
This distribution of minerals among particles can be accounted for
by an apparent particle conductivity which mainly depends on the
average mineral composition in the soil material. In addition, a
possible interdependence between mineral content (composition)
and particle form/size may affect the apparent particle conductivi-
ty between soil generating minerals indicates that such microstruc-
tural conditions will be of minor importance.

The previously described Hashin-Shtrikman limits for a system of

mineral particles, ice and water will give a rather narrow range of
variations for the conductivity. Figure 18 shows these limits for
a hypothetical, saturated clay material with varying relative con-
tent of frozen water. This example is valid for a porosity 04 50
percent and a particle conductivity of 3 W/mR.

-
; 11 0............

Por":Si r.el 0. ,50

A40C6. rOS$T VANNk PIrt iaf¢l~ ,ld t¢, ,'

0k llN,"
' @ vQM
Al~lf #.Wtl4
W.,4,MU WMA.- 1

Ila .li v ih tl- irlJ•, F..tn"v',•l r'r fill•• va rro tot. | 1ovi t-
I '+ ' ;• . 1W &,11.
+ L nL I
lt~lJ+ .'lu .l
Vio '.1 l'r I , luril

199
Evidently, the limits are particularly narrow when the material
is complotely saturated by iCe, but also for water saturated ma-
terials the limits form a basis for relatively accurate estimates
of the conductivity, without regard for differences in microstruc-
ture.

In chapter II, results were also presented which indicate that the
geometric mean of component conductivittes, calculated on the basis
of the relative volume for each component, will give a good appro-
ximation to the conductivity in such materials. As shown in the
figure values calculated from this model fall within the Hashin-
Shtrikman .limits over the entire range considered. The validity
of this model will be discussed in the following, based on results
from experimental investigations of conductivities in saturatoA soil
materials.

B. Measurements performed by Kersten

IKersten did not perform measurements on completely saturated samples.

However, based on three or four measurements for a given density
he extrapolated data for full saturation. (Pages 185 to 193 in
Kersten's report from 1949 (7).) These extrapolated results are
used in the following study of conductivity in saturated soil ma-
terials.

Figure 19 shows these extrapolated values for saturated materials,

plotted as function of relative particle volume. In this semiloga-
rithmic diagram, the previously mentioned geometric mean equation
gives a straight line between the conductivity of water for zero
relative particle volume and the particle conductivity when the
relative particle volume is unity, as indicated by the thin lines
in the figure.

1) The original does not specify which figure is referred to

here. It could be one on chapter II. (Translator's note)

200
 1/,

0 .. .. .. ... .. .

4t

508 183 I I I/r

Lr*ýlC..-I *s.Ift0 :1 11
of )at~e I I
r'? om
aCt..b~ 81; k",~O

NIext to Figure 19, the soil materials concerned are listed accor-
ding to decreasing quartz content, with the symbols used in the
figure. By comparing the plotted data ,.nd the table one finds that,
with few exceptions, the conductivities are grouped according to
increasing quartz content,

In chapter IV it waa proposed to use a geometric mean equation of'

* the followving form to estimate particle conductivity on the basis
* of quartz content in a soil material.
; , ......
. .... .... A
X2- 2 .0 1-q 7 .7 C' W/MK7
where q is the quartz content.
When the particls conductivity is known, one may also expect that
the conductivity of a saturated soil material can be estimated from
the geometric mean: 1-A
X M Xý Xn (W/MK) 8
where n is porosity
1-n is relative particle volume
X.iis conductivity of water (ice) (W/mK)
X2 is particle conductivity (W/mK)
Figure 20 shows a comparison between the conductivities extrapolated
by Kersten for saturated soil materials at 4C and values calculated
from Equ 7 and 8, for the values of relative particle volume and
H quartz content given by Karsten.

201
 A, A

,
, //"/ le//

0 Z/

) O• 40 4. 20 2• 20
MALT LIONIPdO!SV'4E

S')l•~iý)5 '•r,!i v • :1
genv,•u •). ! .",)I Vtl~, '.
lr. b•,s1 ! n,) " Or!
o

'As can be seen, the calculated values fall largely inside error
limits of between 15 and 20 percert of measured values. These
error limits contain both the measurement uncertainties and the
uncertainty of the mathematLcal model used.

Corresponding comparisons between measured and calculated values

for saturated, frozen soil materials will be distorted due to the
uncertainty in regards to the content of un-frozen water. How-
ever, 6or coarse soil materials containing no finer fractions than

sand, one can assume that all the water is frozen. For the four
soil materials studied by Kersten, which meet this requirement, one
obtains a difference of less than 10 percent between calculated
values and those extrapolated to full saturation by Kersten.

C. Ocean sediments

The effects of different quartz contents on the conductivity of

saturated materials was also pointed out by Kasameyer et.al. (1972),
in connection with studies of thermal conductivity of ocean sedi-
ments in the North Atlantic (15). Samples were taken from the
ocean floor by a research ship when crossing the North Atlantic.
All these experiments were performed with a thermal conduction
probe, partly on board the ship and partly in a laboratory.

202
Earlier investigations of thermal conductivity in ocean sediments
have resulted in rather stable values which mainly vary with water
content in the samples (Ratcliffe 1960 (16)). Similar values were
found during this investigation in the eastern and central parts
of the region covered, while measurements from the western part
showed significatly wider variations. A few samples gave thermal
conductivities more than twice as Iigh as for normal ocean sedi-
ments. The authors refer to geological surveys in the W/es- Atlan-
tic which revealed thick deposits of quartz sand and silt with high
quartz content also in particles smaller than 2 ,m.

0177
o i/ VAN

20 40 so 40 00
VAN4IN IHOLO AV VArT VV(

'00 O 0,60 C• 00 )0 0'0

, 9 L.VOLUdAIOIEL
KAI

. . Lf.t.ringn,,:v nen ivi h1u ,so I Lfunt.et' p ooto com. rwuik-J on

k ''''l ,iej.'t dirHr'r

'

~ ~~~
.1,1~~~ IIv. . e 11 ,.

Figure 21 shows results obtained by i:asameyer with samples from

the region mentioned, plotted versus water content relative to
wet weight. Since these samples were saturated, the results also
give an indication of how tightly packed the sediments wore. If
specific weights for the mineral particles are known, dry densi-
ty can be derived from this parameter by means of the relation

Yd Yw (1-Ww) 9
Y8 YW (1wW) + Ys ww

203 I
where ww is relative water content by wet weight

ys, yw are specific weights for mineral particles

and water, respectively

yd is dry density

For comparison, results from "normal" ocean sediments in the At-

lantic are included in Figure 21 (2atcliffe 1960). Theoretical
curves for the conductivities in materials with quartz contents
from 0 to 75 percent are also shown. The latter were derived from
the previously used mathematical model for particle conductivity
and conductivity in saturated materials (geometric mean equation).
The volume ratios 1 ) are based on the relation between dry density
relative water content by wet weight, given above (Eq 9), with S
specific weight for the mineral particles assumed to be 2700 kg/m3.

As can be seen (from Figure 21), points representing "normal" sedi-

ments fall close to the curve for 25 percent quartz, while the data
obtained from the western region are grouped between the curves for
25 and 75 percent quartz and show a much wider spread. The authors
give no quantitative values for the quartz content in these sedi-
ment,, but in view of the qualitative descriptions of the materials
as being sand or silt rich in quartz, the indicated values are not
improbable.

D. Conclusion

The analysis performed of thermal conductivity measurements on sa-

turated soil materials show that differences in mineral composi-
tion have significant effects on the conductivity. Comparisons
between experimental values and calculations based onknown quartz
content, as well as volumetric composition, show that the proposed
mathematical model for determining conductivity on saturated soil
materials give satisfactory agreement.

1) "Forhol" could here also mean "conditions" (Translator's note).

204
However, this mathematical model should also be compared against
experimental data from a representative selection of Norwegian
soil materials. To this end, measurements have been initiated
by the Institute for Cold Technology, which will cover a wide
variety of soil materials. This research program includes measure-
ments for varying degrees of saturation (as well as dry materials) as
will be described in the next section.

3. MOIST MINERAL SOILS

A. Introduction

A partially saturated material can be assumed to consist of three

or four components, particles, ice and/or water and air. For such
a system, the Hashin-Shtrikman limits will enclose a relatively
large region within which the conductivity may vary. This is illu-
strated by Figure 22, where the limits are calculated versus de-
gree of saturation in a hypothetical sand material with a porosity
of 0.30 and a particle conductivity of 4 W/mE./K.

2 0 2 0 %-4 '

J-1
.. iis i -h r ua do
iin ra ir iso v

02 0

,olli o 0, 3U;0 i k•Ii

• "! •.. .kO MI i"I'NI 0,30 , '. o II l'j! u

) ii aonio4j LoL(o

VU•
[,l', U I.••4eldltl•Pl'-)l[
lYII/tl!l••'ll'. {•I.|'

"1.1|ll"l 205Oi
As mentioned previously, the wide separation between the limits
is due to high sensitivity of the conductivity to variations in
microstructure. Since this effect can not be incorporated into
analytical models for conductivity calculation, due to the compli-
cated geometry within natural soil materials, one must here resort
to purely empirical methods for describing the conductivity depen-
dence on the degree of saturation.

The mathematical models proposed in the previous section mel~e it

possible to determine the conductivity for two extremes in the de-
gree of saturation, corresponding to values for completely dry and
completely saturated materials. For partially saturated materials
the conductivity will fall between these extreme values.

This condition forms a basis for defining a normalized conductivity,

which permits results from different materials to be presented in
a uniform manners This normalized conductivity is here given the
notation "Kersten's numbers Ke":

Ker " -XQ X0

X2 10

X is measured conductivity for a given combination of den-

sity and degree of saturation
0
.\ and X are conductivities for dry and saturated material,
respectively, for the same density

With this representation, the conductivity of dry material will

give Ke a 0, while the conductivity of saturated materials gives
Ke a 1. For other degrees of saturation one obtains values for
Ke between these limits. This representation is particularly sui-
table for experimental results which pertain to a wide range of
saturation levels for the same degree of packing. Values for dry
or saturated materials can then possibly be determined by extra-
polation.

206
 B. Study by Kersten

As mentioned earlier, the thermal conductivity experiments per-

formed byKgr nossiJ-
-Oiafost t~eriais-frvm thKe-p region in
Alaska remain the most extensive study of its kind. Measurements
were performed on a total of 19 different materials. Measured re-
sults for the ten natural soil materials which were studied in
most detail, with respect to variations in degree of saturation
and density, will be treated in the following. The mechanical
and mineralogical compositions of these materials are listed in
Table II of this chapter. Plotting symbols used are shown in the
same table.

Each of these soil materials was subjected to thermal conductivity

measurements at three or four different water conten.s for: ;',o
or three different densities. Conductivities were measured for at
least four different temperature levels$ two above and two below
the freezing point.

Results obtained at +40C and -40C will primarily be discussed in

the following.

Figure 23 shows results obtained by 'ersten at +4"C for two soil

materials, plotted on the previously mentioned normalized form
(Eq 10). Different densities are indicated by means of marks on
the plotting symbols. As can be seen, the results for each soil
material show nearly the same functional dependence, independent
of dry density, while the difference between the two soils is sig-
nificant.
One of these soil materials has relatively coarse particles while
the other is a fine grain material containing a total of 90 percent

of silt and clay fractions. This would indicate that this kind
of representation also may be used for studying the effects of
different textures on the conductivity of moist woils.

207
 06
;,! .

03 //
-

006 of 0203 0600 10

* PIG. 23. Xer'560n5 mA.,Linivr j'A to 0'4rromjor1irter pottet

14 ormallH rt :'or" du;1. !'¶A..K..j n av ,. - jn.

diiPoe u . ,,tu•ar .on.

Figure 24 shows results from Kersten's measurements on ten natural.

soil materials at +4 C, plotted in the same normalized form as
before. Results obtained for the coarser materials (no silt or
clay fractions) are shown in Figure 24a, while results for the fine
grain soils are given in Figure 24b. One finds that soils having
a certain content of silt and clay give data points which fall lo-
Ai; wer in the diagram. than points representing soil materials lacking
these fraction.

"+'.

010. of 00 O ./a*
0I t1

0 ---
-'6 .II 0 f; it,1,1.i..a.....a.If
0 .US ' 0, -" 0 3•
0

WOIWI I. IOIA
*I'•lj .rtlL,,.., r ' i20
.'.,.43'/l•I : ,'sU .'i P.t I
If4+ m ('P at-Ik 1, 1 .ýfft~~
W0,, 4 *1C,+

IP•II + 't- I~ P. ~t•.,l~ r n• d rl1le + Pll •+• ;•I+.+

, )• *+l•
208l,,• ,l l{~1 + i ;A~+1., . .f 1
For the first group - fine grain materials - the spread is rela-
tively small around a mean curve given by

Ke % log Sr + 1.0 11
Sr > 0.1

For the coarse materials, the spread is somewhat larger around

a mean curve given by

Ke d 0.7 log Sr + 1.0

Sr > 0.05 12

Kersten's results for frozen materials at -4*C are shown in Figure

25, using the same normalization as before but with a linear scale
for the degree of saturation.

• 06'*

0-

0 0.4 06 08 '0
METNIN050AA0 I

* '5 ur~tuns MI.inr~tjr iu.ý ~t'r,10n '01 Il~o;L

209M
lT IN$•d iC

209
 As evident from Figure 25, all the results fall near a common
mean curve (line), regardless of the previously used separation
according to texture. The mean can be represented by the simple
relation

Ke - Sr 13

with a variation somewhat less than +tKe a 0.1. This simple re-
lation follows directly from the nearly linear relation between
conductivity and degree of saturation shown by Karsten's data
from measurements on frozen soilso

The previous analysis of Karsten's experiemental results was li-

mited to data obtained at two temperature levels, +4*C and -4LC.
The conductivities of soil components are to some extent dependent
on temperature. While the conductivities of water and air increase
for increasing temperature, the conductivities of mineral particles
and ice will decrease. As discussed in chapter I and II, the effect
of water vapor diffusion on the apparent conductivity in the pore
air will contribute to an increasing thermal conductivity with in-
creasing temperature for moderate (mid-range) degrees of satura-
tion.

Figure 26 compares results obtained by Kersten at two temperature

levels above the freezing point and for varying degrees of satura-
tion. Data for coarse and fine grain materials are shown with dif-
* ferent plotting symbols.

A clear tendency can be seen for the largest increase in conducti-

vity to occur for moderate degrees of saturation. This region
appears to fall at some-what higher degrees of saturation for fine
grain than for coarse materials. For very low or very high satura-
tion levels, the differences between conductivities at the two
temperatures are insignificant. Even at mid-range, the difference
seldom exceeds 10 percent of the value obtained at *40C.

210
 G;0"VK0ftNET
J * FINKORNET

#0

09%

2 '. 1 9 to 20 40 6.O 80 iOiQ

.4CTNI NGOR AO
.(NH(I. Olt SAIWqATIQN

6, ZiIlnrt
1Z. lktfi Av tw"pUOLtull.Iti A WIinini!,swinui Il

Ilif&Iwilliati Of 4um Iid,11910 ,.,0 cion'tu.1, clif y 0,0 1'c0fr~afvil

*21
 !
flNKORNZr

EP ~

1 2 0 20 % 406*01

M!CTN164GS(RAO
Dt('Rq C)" 5AI.'ATION

Figure 27 shows a similar comparison of results obtained at two

temperature levels below the normal freezing point, Also here,
data for coarse and fine grain materials'are showrn separatel~y.

Also in this case there seems to be different trends for the two
groups of materials. However, the decreasing conductivity of air
with decreasing temperature seems to be the dominant factor at low
and moderate degrees of saturation. For higher degrees of satu-
ration, the increasing conductivities of mineral particles and
ice with decreasing temperatures become more and more-important.
In the fine grain materials, the latter effect is amplified by
the fact that the relative content of u~n-frozen water decreases
both with decreasing temperature and increasing degree of satura-
ti~on, For the lowest degrees of saturation, all the water may be
frozen at both temperature levels at hand, For higher degrees of
saturation, a significantly larger portion of the water will be
frozen at the lower temperature (-300 C) tihan at the higher (-.46C),
which contributes to an increase in conductivity with falling
temperatures, Also in this case one finds that the increase in
conductivity seldom exceeds 10 percent of the value at -4"C,

212
 .0,in~
Ket'u.
%(elit
e.07 wl st
I,,*JQ e I~* 'A l

C) I4,od l U nqiiiWl 27.5 '0.0 2A2.17 50.4 9.9 0.1 n0o

" "M4
"IL 00 100.0 00 'JA 2 47 77.2 20.61 . 3i.0 -

V Not I1,..v Wil~d 3. 97.0 11.0I 0, ?774 7.5 0.0 - e%

1
Ni,1* in.e
o f4, sl~in 0.0 07.0 3.( 0.0 2 11U 12 r) 10.0 7.0

Niti me
11nI0uUgq
berOl 10.2 17.0 21 W to -) 7.71 60 1 13.0 - 17.1 174
4
14 .Y
W rikilhall'O no 0.4 S3.61 2) 4 IRi T.111 51.3 11.4 16A)

0 'l*I.~Vq 1,11hlkifl tria 1.0 21.0 tK4. 1301 7.)n1 i.s 3i,1 19.5 71 '1
i t 1 1
jf IllOt~i1
1
IluVl IM 0.0 7.0 Vill) 11 5 2 10 04 1. 1

,'. ',i...
'lI.I1,10yn 0.0 91.2 43 U 2117 '.71 1141 - 3.? 2.2 21h.0

Legend for Table 11

Symboler Plotting symbols
Betegnelser Notation
Kornfordelning Particle size distribution
vekt%,b Percent by weight
Spec, vikto t/m3 Specific weight, metric tons
3
(1000 kg)/m
*Sammansettning Composition
Sandholdingp siltholding Contains sand, silt
mo Common name for fine grain~ soils
leiri; Contains clay
Grus Gravel
Leir Clay
rivar t Quartz
Feltspat Feldspar
Glimmer Mica
Pyrcoxer. Pyroxene
am fibo 1 Amphibolit e
Oli vi n Olivine
Leir minoraler Clay minerals

213
 C. Investigation of Norwegian soil materials.

A. Watzinger and E. Kindem (5) performed a series of thermal con-

ductivity measurements on a limited selection of Norwegian soil
materials in 1936/37. These measurements were made in a planar
horizontal test unit with the cold plate above and the warm plate
below the samples. Temperature differences across samples were
measured by means of thermocouples attached to both plates. The
mineral composition of the materials was not determined but scree-
ning curves are available for all mineral soil materials. Lack
of information about mineral composition and variations in degree
of packing from one experiment to another makes it difficult to
analyze the results.

Efforts to extrapolate these results to fully saturated conditions

indicate that all the investigated soils have particle conductivi-
ties in the 2.0 to 3.0 W/mK range (largest value for a sand material
lowest for clay and kvabbjord.According to the proposed method for
estimating particle conductivity, this corresponds to quartz con-
tents from 0 to 30 percent. The selected materials are thus not
very representative for the variations in quartz content that are
found in Norwegian loose deposits.

Against this background, the need for an extensive investigation

of thermal conductivities in Norwegian soils is quite obvious. As
mentioned earlier, such an investigation has begun at the Institute
for Cold Technology. The plan for this study calls for investi-
gating a total of 45 Norwegian soil materials, as well az a small
number of foreign soils (all toghether 4). The material library
is described in Table III (quartz content) and Figure 3 (mechani-
cal composition).

*RkvabbJord - translation unknown.

214
 Table III. Summary of quartz content in the mineral library
at the Institute for Cold Technology

SOIL NUIP4ER OF RANGE OF QUARTZ AVERAGE Q1JARTZ

MATERIAL SAMPLES CONTENT, P RCMIT CONTEIT, PMCENiT

CLAY 7 10 - a8 18.8
SILT 6 21 - 53 33.5
SANJD 13 10 - 100 44.9
GRAVEL 13 2 - 57 38.8
MACADAM 10 2 - 53 12.7

TOTALS 49 2 - 100

The mechanical compositions of all samples, except macadam, are

shown in Figure 28 in the form of an Md - So diagram 1," where
each material2) is indicated by a different plotting symbol. The
numbers correspond to those used in the material library,

As shown by Figure 28, the selected coil samples cover marine se-
diments and glacial flow materials rather well, while moraine clays
and moraine gravels (not separated into layers) are poorly repre-
sented. In terms of quartz content, the ran!;e of interest seoms
to be well covered for all types oi soil, porhaps with tho excep-
tion of quartz-poor loose deposits from regions characterized by
special petrographical conditions. lowever, the macadams include
a relatively large selection of materials with low quartz content.

Accrrint the plan, thermal conductivity measurements on these

material vwill be performed in a parallel fashion in several types
of test units (planar, probe and sinusoidal vseiation) at degrees
of saturation ranging from complotely dry to fully saturated. :low-
ever, for the coarsest materials it is planned to measure conduc-
tivity only for water contents occurinG after run-off, in addition

1) Refers to particle size distributions, see Figure 28.

2) Grus - gravel, Leire • clay (Translator's notes).

215
 '0

1s
20- I
180

- ..- 2 2

0.2 2 20 0 2 -Tim 2 20
M E0 AN Nid=d
PIG. ~ eri~iv a.~ jori- tf.ri~lt~r f rit~ r 1~b
v&, r.: u ctLr k. .,:tkvi~kk,;,TH{. ~ 'rI2 ru:
i YM1 ti]. .atrialbibll±otzk~t,.

216
4

to experimento with dry and saturated sample6.

These experiments, which according to plan will be completed at

the end of 1975, will give a very valuable data base for evalua-
ting the mathematical models described in this report.

D. Conclusion

The empirical relations derived for dry materials, the proposed

mathematical model for saturated materials and the empirical re-
lation developed here between the normalized conductivity (Kers-
ten's number) and the degree of saturation together form a new
base for calculating the conductivity of moist and frozen mineral
soil materials. From the definition of Karsten's number one can
compute the conductivity for a given combination of density and
degree of saturation from
X Xo + Ke (S ) 1 Xo)
r
14

Conductivities for completely dry and fully saturated materials

having certain densities can be computed from the empirical rela-
tions for dry materials and the mathematical model for saturat~ed
materials, respectively. The functional dependence Ke (Sr) 13
given by one of the three empirical equations derived previously.
The uncertainty inherent in this method depends, among other things,
on how accurately the properties of the material at hand are mnon.
These aspects will be treated in more detail in next chapter (VI),
where also other avaiable methods for calculating conductivity of
soil materials will be discussed.

217
 4. SOIL MATERIALS CO'ITAINI.NG 1iUIUS

A. Introduction

Previous discussions have been limited to purely mineral-based

soil materials. This section will discuss soils containing
humus. The humius content is in geotechnical literature given in
terms of glow loss. It is then assumed that organic materials are
completely burned when heated to a state of glowing, while mineral-
based materials have no glow loss. On this basis, the glow loss
is a measure of the relative mineral and organic content in a soil
material. These matters were discussed in more detail in the sec-
tion on volumetric conditions in chapter III.

Materials having glow losses between 1 and 6 percent are usually

classified as humus carrying, while materials with glow losses in
the 6 to 30 percent range are described with terms such as clay-
bound peat. Glow losses larger than 30 percent warrant the nota-
tion "peat". However, in the latter class of materials the mineral
content may constitute up to 50 percent of the dry volume.

Few systematic measurements have been made on such mixed materials.

Smith (1939) presented results from different typical ground pro-
files, where sa~mples containing considerable amounts of organic
material are included (3). Irowever, the information concerninr
these materials is incomplete and the measurements were not reli-
able due to considerable re-distribution of moisture during; the
experiments.

On the other hand, certain reliable results are available for pure-
ly organic materials such as pure peat and bog. These will be dis-
cussed in the following.

.Watzinger claims that the thermal conductivity of organic compo-

nents is 0.46 'N/mX (5). The same value was used by Misaenard for

218

.........
. .....
 the conductivity of hardwood (18)IJ, Zincc tee conductivitjy for
water is 0.57 W/mK (at about t46C), the previously mentioned
Hashin-Shtrikman limits will be much narrower for water-saturated
peat and the geometric mean equation can be used without major
errors. The same will be true for ice-filled materials, since
the conductivity of ice is 2.2 W/mK. (at about 0CC), so that the
conductivity ratio for the components will not exceed 5.

For dry materials, conditions are some-what complicated by the

strongly porous nature of the materials, which will give heat
transport by means of radiation an opportunity to come into play.
The very porous structure may also cause a relatively high, tem-
perature dependent water vapor diffusion resulting in re-distribu-
tion of the moisture content in materials with low to moderate
degrees of saturation. This may be of importance for heat trans-
port through peat layers, e.g. in permafrost regions, as pointed
out by Nakano and Brovm (1972) (17) and others.

B. Measurements on peat and bog

As mentioned earlier, Kersten performed measurements on one peat

material from Alaska (7). Watzinger performed measurements on one
bog material and one type of peat moss (3). The results of these
measurements are shown together in Figures 29 and 30, for frozen
and un-frozen samples, respectively.

i The materials are grouped according to dry density by means of

different plotting symbols. There is evidently little variation in
these measured results as function of density. However, closer
scrutiny reveals a tendency for the lowest densities to give the
lowest conductivities at low degrees of saturation, while the
opposite tendency occurs for high saturation levels, particular.
ly for frozen materials.

1) Literal translation "compact wood", probably referring to

dense variaties of wood, i.e. "hard wood" (Translator's note).

219
 /

0 1 @4 of
"4 IQ
I
PIO. 29.
.o@ -,7 . pA... ramsse.
m1411Uinsge . . tarv- ap .IO pJ.Out 00111.
funWLajon av IlnvlninCfrrA.Ion. '!jgartimoplee oil
fP'aaeii peat JZ& LOPIO illato~i to d 7l'0 o",
sdaturation.

S.... I w.... . ir, aiI .... . .. u ntu .r

0.1

",, 1 6AI-.... -

1.140
I 11--! i,PIA..
.. V .. . t .. . n:, . .. 1. L 1...

,I , roatta lima
0 1 vu
' 4 &Lou :j ,j. ae o'f ial ti'r1 1 on.

11A

The explanation to this can be found in the functional dependence

of the conductivity for dry and saturated materials, respectively.
Figure 31 shows theoretical conductivities for dry and saturated

4this
materials, computed from the geometric mean equation.
results in values which are some-what low for dry materials,
since the effects of thermal radiation are neglected.
As expected,

However,
the trend as function of porosity will be the same.

220
 iI

r,, 1.0-

Ole oil 000 a 'V

nj. PoJ-o6%1116 ,
Pl.3 1. IIfIammannm,, mlj l~oiJo Ivininwiic~vuri oir. n '1litet~of i~v

Of JrI.l and Jo1.Wer td..' I-121 O 4

.JL .OU tol bP 0.111

One finds, that within the normal rancte of-poroalties, the conduc-
tivity of' water saturated materials will vary less than 3 percent
around the mean. For frozen and s~aturated materials, the correspon-
ding variation is leus than +10 percent. In both cases, the conduc-
tivity tends to increase with increasing porosity since the organic
component has lower conductivity. The variations for dry materials
are larger percentage-wise and the trend is the opposite.

The resulting conductivity variation as a function of dry density

is on the other hand sa small that one can une a common mean curve
for all materials to obtain a good approximation for the conducti-
vity at a given degree of saturatiorn,

The cur'ves shown in Figures 29 and 30 can be expressed by the

formulas

-(un-±ozen Sr soils) 15

A (X1 /XO)Sr (frozen soils) 16

The following average values are proposed for conductivities

of dry and saturated materials

221
 Ao = 0.05 .v/.I:
= 0.55 W/mK (un-frozen)
= 1.30 .lmI (frozen)

REFERKICES, CHAPTER V

1. C. Krischer: Die Leitf~higkeit des Erdbodena. Beineft zum

Gesundheits-Ingenieur. Reihe, Heft 33, Munchen (Munich) 1934.
2. W. O. Smith, H. G. Byers: The Thermal Conductivity of Dry
Soils of Certain of the Great Soil Group. Am. Soc. Soil Sci,
Proc. 3, 1938, pp. 13-19.
3. W.O. Smith: Thermal Conductivity in Moist Soils. Soil Sci.
Soc. Am. Proc., 4, 1939, pp. 32-40.
4. W.O. Smith: The Thermal Conductivity of Dry Soil. Soil
Science, 53, 1942, pp. 435-459.
5. A. Watzinger at al: Undersokelse av masseutskiftningsmaterialer
for vegog Jernbancbygging. Iledd. fra Veglirektoren 1938, no. 6.
6. A. Watzinger et al: Unjersokelse av masseutskiftningsmaterialer
for vegog jerntanebygging. Medd. fra Vegdirektoren 1941, no.
6, 7 and 8.
7. M.S. Kersten: Thermal Properties of Soils. Univ. of Minnesota,
Inst. of Technology, Bull. no. 28, June 1949.
8. m. van Rooyen, H.F. WInterkorn: Structural and Textural
Influences on Thermal Conductivity of Soils. Highway Research
Board Proceedings, Washington D.C. 1959, pp. 576-621.
9. D.A. de Vries: Hot wirmtegeledingsvermogen van grond, Medd.
Landbouwhogeschool, Wageningen, 52, 1952.
10. E. Saare, C. G. Wenner: Varmeleiningstal ho. olika jordarter.
Statens nimnd for byggnadsforskning, Handlingar ar. 31, 1957.
11. O.T. Farouki: Physical Properties of Granular Materials with
Reference to Thermal Resistivity. Highway Res. Rec. 128, 1966,
pp. 25-43.
12. R. Mc Gaw: Thermal Conductivity of Compacted Sand/Ice Mixtures.
Highway Res. Rec. 215, 1968, pp. 35-47.

222
S• • ,-
 13. N. Wakao, K. Kato: Effective Thermal Conductivity of Packed
Beds. J. Chemical Engineering of Japan, 2, (1) 1969, pp. 24-33.
14. L, Barden at al: .The Collapse Mechanism in Partly Saturated
Soil. Engineering Geology, 7 (1), 1S73, pp. 49-60.
15. P.W. Kasameyer et al: Layers of High Thermal Conductivity in
the North Atlantic. J. Geophys. Res. 77 (17), 1972, pp. 3162-
3167.
16. E.H. Ratcliffe: The Thermal Conductiviites of Ocean Sediments.
J. Geophys. Res. 65, (5) 1960, pp. 1535-1541.
17. Y. Nakano, J. Brown: Mathematical Modeling and Validation
of the Thermal Regimes in Tundra Soils. Arctic and Alpine
Research (1) 1972, pp. 19-38.
18. A. Missenarii Conductivite Thermique. iditions Eyrolles,
"Paris 1965,"p. 356.

.1.
---. •..........

H 223
 METHODS FOR COMPUTING TIHE:!AL CCNDUCTIVITY
OF SOIL XATERIALS

The previous analysis of experimental studies of thermal conducti-

vity in :ninaa!-baied jsol matnria-s forms the basis for a new
mathematical model for computing thermal conductivity in moist and
frozen soil materials. The first section reviews existing mathe-
matical methods in the light of this analysis. The next section
presents the new mathematical model in more detail.

1, EXISTING METHOD3

A. Analytical methods

The important effects of the microstrudture on the thermal conducti-

vity in non-saturated soil materials was pointed out in the previ-
ous section. Such microstructural parameters are difficult to in-
corporate in a purely analytical mathematical model mithout s.g-
nificant simplifications, both in terms of geometry and computatio-
nal methods. Still, eft'orts in this diructi.on hav'e been made.

Gemant (1950) (1) used a very simplified model of a moi.t soil

.l.

material, where the soil particles were assumed to be uniform (equal

size) spheres, tightly packed in a cubical lattice. !oisture .1as
assumed to collect in meniscus-ehaped bodies around the contact
points between spheres. (See Figure 1)

The thermal conductivity was in this model computed by assuming pla-

nar isotherms. As shown in chapter II, this method yields an upper
bound for the conductivity, while assuming parallel heat flow results
in a lower 'ound. The calculations are further simplified by net-
lecting %he conductivity of air, as well as the contribution from a
meniscus arou.ind the "equator". In this manner, Gemant arrived at

224
4.
an analytical expression for conductivity, with di3tance Yo as
parameter. This parameter defines the distance from the center
of each sphere to the plane intersecting the lower "meridian" boun-
dary of the meniscus, normalized to the radius. The relation be-
tween relative water volume and Yo can be deri.ved from geometrical
considerations and is

1.33 wV R 11(0.33 - yo 2 + 0.67 yo3 )1

This equation is only valid when adjacent menisci are not in con-
tact with each other. The expression for conductivity has the form

-(f+g) (f-g YO o h +, h
X
a g (f-g) (f+g yo) fh2 h - h Yo

where f2 = X2 - Y2 ý1, g2 X2 - Xl, h2 -X2

0
is conductivity of water
is particle conductivity

This method has one obvious weakness as a general way of determining

conductivities in 3oil materials: The model does not allow for chan-
ges in the degree of packing (packitng configuration). The cubic lat-
tice gives a porosity of 0.476, which for dry sand corresponds to
a density of only 1400 kg/m3 .
VAFIMESTZ,•- TNINO
VE

*~~~Q~j
~J~I~iTN.O

225 •
 Gemant carried out a numerical example for ;ihich the particle con-
ductivity was assumed to be 4.2 '3/mi,. Figure 2 compares the func-
tional dependence of these results with Kersten's experimental va-
lues for a silt/sand mixture (Ramsey sandy loam), which has about
the same particle conductivity. The lowest density in this test
run was about 1600 kg/m3 and thus some-what higher than that given
by the mathematical model.

*,*, I= , , t-- t --

W/ni •ItUn
0. 04
II.
aA.Ii 6
(Qoytlll
2 1

I - -'j- n~mY.I ~s~,

9 '1-
10 -. Pe@ *,.

The calculated curve is obviously in poor agreement with the ex-

perimental results, both in terms of functional dependence and ab-
solute values, Particularly for low degrees of saturation, the
calculated values are significantly larger than corresponding ex-
perimental data, Inhi.. can partly be caused by the assumption that
water accumulates in isolated rnenizs~i not being valid for the lo-
t west degrees of saturation* For this case one may speculate that
a significant part of the water forms a film on the particle sur-
faces. This water will probably contribute little to the thermal
conductivity but will "rob" the meniscus of part of the total water
content,

M.ak'owslki and Mochlinsky (2) refer to a modiffied model, which

Geinant developed later, where this condition is taIten into account.
Their description of the modified model is incomplete, However,
it is evidently po3sible to fit this model also to different de-
grees of pack-ing, although the method for fitting, seemrs to be ~

226
piaical. .akozki an. uohlinszy have developed a relatively nim-
ple equation# based on curves fitted to data obtained from Gemant's
modified model. This equation will be treated in next section.

Alberts (1966) has proposed a similar model, where the soil par-
ticles are represented by tiohtly packed spheres in a cubic lattice,
while the water is partly spread out as a film on the surface of
each particle and partly contained in a meniscus around each con-
tact point (3). Part of the calculations were carriea out numeri-
cally but the results are of little value due to the very low con-
ductivity assumed for the particles (x a 1.0 W/mK), as well as some
approximations which are introduced based on this assumption about
the particle conductivity.

de Vries (1952) presented a method for calculating thermal conduc-

tivity in soil materials, based on the previously mentioned Maxwell-
Fricke's equation (4). The foundation for this equation is discus-
sod in chapter III. It should be pointed out here that this equa-
tion is derived under assumptions which are valid in a composite
material where ellipsoid-shaped particles occur in rather small
quantities (dispersed) in an otherwise continuous mo -um. For soil
deposits in which the particles are in mutual contaý: these asamgp-
tions are not valid.

The equation can be written in the form:

N
Z ni•iFj
S- i-o

L nipi
1-o
where ni i4 relative volume of coniponent "il
± is conductivity for component "i"

-.1

F .. + (!i/x 0 - J , j albc

227
 a ÷÷ c -= 1.0

The factor Ti gives the ratio bet-voen the average temperature

gradient in the dispersed component "i" and the continuous medium
"o", provided that the indicated assumptions are valid. For
more tightly packed materials, this factor can be considered
as a "form factor", where values for gat Sb and gc are obtained
by fitting to empirical data.

These matters were discussed in section 1 of mhapter V, where it

was shown that the equation can bo fitted to conductivity data for
dry natural soil materials by the following choice of g-values.

gl x a .10 , g3 .
C.80 (Dry soils) 5

For water saturated soil materials, de Vriea found a different

set of values

2 o.ia5 , 3= 0.75 (Saturated soils)6

This set of g-values corresponds to a long ellipsoid ,,rith a ratio

of about 5:1 between largest and smallest diameters. Such a par-
ticle geometry can hardly be considered typical for a sand mate-
rial. The set of g-values for dry materials imply a further
trend towards rod-shaped particles.

Figures 3a and b compare results obtained from de Vries' equation

for saturated soils to values calculated from the geometric mean
equation. rFigure 3a shows conditions for a constant conductivity
ratio y2,•x • 8 ) with relative particle volwue as the variable.
Figure 3b shows conditions with constant porosity (n a 0.35) and
varying particle conductivity (conductivity ratio).

228
-A
 . .4 4/, I

2... . ... .....

060
06 t
1
O.J2 0 O&O

"),.
4''-Ac

~
~r ia
4I~14or t I

pro444 4 v ,..

These graphs show that the two equations deviate from each other
by less than 5 percent within the ranges of particle conductivi-
ties and porosities that are typical (may occur) in saturated soil
materials.

Thus, the two equations can for practical purposes be considered

equivalent. Since the geometric mean equation is the simpler of
the two, it is most convenient to use for determining (calculating)
conductivity in saturated soils, as proposed earlier.

De Vries suggested that the equation also could be used for calcu-
lating conductivity in moist soil materials. He separated out se-
veral degreesof saturation ranges in which the three components,
particleswater and air, were treated differently with respect to
geometrical conditions. For low degrees of saturation, the partic-
les were considered to be enclosed in water, while the air was
considered to be the continuous medium. For high degrees of satu-

229
ration, the water was considered as the continuous medium, in
which either the grains of sand or air bubbles are contained in the
form of particlesI)

The affect of water vapor diffusion on the thermal conduction in

the air-filled pores was assumed to be zero in dry materials and
gradually increase to a maximum value at a given (low to moderate)
degree of saturation. The different regions were selected partly
by comparison with experimental data for moist sand, partly after
evaluating the natural ranges in which the different mechanisms are
effective. In reality, this amounts to fitting analytical results
to experimental values. The applicability to other soil materials
has not been investigated. In addition, the method is very cumber-
some and does not offer any advantages over the purely empirical
method proposed in the previous chapter for degrees of saturation
ranging between dry and fully saturated conditions.

This discussion has shown that analytical methods are useful only
if they can be fitted to experimental data. As will be evident
from the following section, empirical methods also have their li-
witations, due to the large number of variables which occur.

B. Empirical methods

Kersten (1949) developed his method for calculating thermal con-

ductivity in soil m&terials by fitting mathematical expressions to
the extensive experimental data base described in chapter V (5).
His model consists of two sets of equations for computing conduc-
tivity of frozen and un-frozen soil materials; one for coarse
grain materials having less than 10 percent silt and clay content
and one for fine grain soils containing more than 50 percent of

1) That is, either sand or air is considered as the "particles"

distributed in water, to conform with the previously defined
two-component model (Translator's note).

230
.A
 these fractions. When these equations were developed, results
obtained for two of the coarser materials (Northway sand and
fine sand) were not included due to the special mineral composi-
tion of these materials (Kersten, page 84), i.e. due to their low
quartz content. This gives an indication that the chosen separa-
tion (classification) according to texture in reality means a
grouping according to particle conductivity.

The equation for un-frozen materials was given the form

" log (w%+ 0,62Yd

1 0o° W/mK

where c and B are empirical factors which are different for coarse
and fine soil materials:

Fine grain soils 0.13 -0.029

Coarse grain soils 0.10 0.058

For frozen materials, the equation has a different form:

S= aiObyd + c10 dyd (W%) W/mK 8

a b a d
Fine grain soils 0.0014 1.4 0.012 0.05

Coarse grain soils 0.011 0.81 0.0046 0.91

Kersten presented these equations in a series of diagrams, as

shown in Figure 4.

231
..... I.....! I•
"ii "1 |..........................- .~"=-"..-.i -----. • , " -- f---.-.•--• .'= ,. ... : ,
-• • , , I If I I I II I I I . I
10'ky,mnl' .

- \\\~ G C/04E'20- IAAlt~

\ \
t-'a,'.A ' tO I

4'\ $q

4AL I'j* m-I" .a "

a) Lediningsevnj av ufrosnie jordarter. 'du:veg

22\ i' ftOVKOANI? 2.0 l~i4.*U

I 01

1..4
~111
0~ ~111' 1 i % 20 1 oAS 4
32~"US ' ~ tPK54 VANN

b)ni Lo
it-n av f vK".-dl-*f

r n kU 25nSvi v3iP v4 7 'e 2t

232PC~VN
The previous analysis of 1Kersten's experimental results showed
that the thermal cor.ductivity of saturated soil materials can
be determined from the geometric mean equation, when quartz con-
tent and volumetric compositions are known. Figure 5 shows va-
lues for saturated materials calculated from Kersten's equations,
compared with data obtained from the geometric mean equation.

i0 j"

nod,- • .;"

VCisW,IACI ý "411L1
0Bi ',/r #

I 11i
z . t
-'a 1.)6k v -'di vfCfoa fo UIo -09'1a

0 0.20 0.J.0 000 000 •0

PARTI FIJLV0.UMA NOCL

110. U.#!rd.
4. 14lo iic d envIdr kgr vtnnmvtt.d$ Joret&v.i. iom -

trio mean ekt

This comparison clearly shows that the two selected soil material
groups represent differences in particle conductivity. The equa-
tion for coarse materials gives an average particle conductivity
of about 5 "//mK, while the equation for fine gqrain soil gives
about 3 73/mlt. From the previously proposed mathematical model for
particle conductivity one finds that these values correspond to
quartz contents of 60 and 30 percent, respectively. An evaluation
of reported values for the quartz content gives a mean of 57 per-
cent for the five coarse grain soils considered. The correspon-
ding average for the four fine grain soil materials at hand is
uncertain due to incomplete information about quartz content for
two of these soils.

Sveian's study of quartz content in 4'orwegian loose deposits,

which was discussed in chapter 1II (6), indicates that the mean

233
------------------------------------------------------ i'--
value for quartz content in coarse grain soils that is obtained
here from Kersten's equations is not representative for Norwegian
conditions. Figure 6a shows a cumulative distribution curve based
on quartz contents in coarser materials determined by Sveian.

As can be seen, only 6 percent of the samples have a quartz con-

tent larger thar 60 percent. The median corresponds to a quartz
content of :j percent, which also is the arithmetic mean. This
fiP,-: o also shows that variations in quartz content are so large
within this group of soil materials that use of a "representative"
average value will mean large uncertainties in estimates of the
conductivity.

Selmer-0lsen'q studies of quartz contents for the fractions d• 21 m

and 2 <d <20 wm in Norwegian clays show a definite difference in
distribution between these two fractions (7). Results for appro-
ximately 160 samples are summarized as cumulative distribution cur-
ves in Figure 6b. The distribution curve for coarse soil materials
shown in Figure 6a can be considered representative for fractions
larger than 20 wm.

U,- . - me....... ............ .

. . 4.. Ipo. -.-. . .

- to 4 4 47i - 0

4 ,i i I lllS 0 ll)0 I l *

LO. 6, Kuwmua,•I V. orde.Ii&Ng Iv kvtrtsinnnhold 'ta vriaer av

Il§+rlk g•tQv•korrItd j; rd ,ar.t e&) , :t.k+J*..lT •"+ .

fIl' o l b) c. - .lr*
• . 1 j-1rQ,''+vd

Kersten defined fine grain soils as those containing at least

50 percent silt and sand fractions. Within the frame-work of

234
this definition, the relative contents of the fractionz d <2 irm,
2< d< 20 um and d )- 20 ' m can vary within aide limits, Since these
three fractions show such considerable differences in their dis-
stribution curves for quartz content, the definition for coarse
grain soils implies no definite classification according to quartz
content.

In chapter III, a method was proposed for estimating quartz content

on the basis of grain size distribution curves. This method is
based in the average quartz content given by distribution curves
for the three fractions. This resulted in the triangular diagram
shown in Figure 7.

16'ý'•7 'to rM, 40 2r

.P10
I Ic a.I 'w I:$~ 'I

If one, some-what arbitrarily,

replaces Kersten'3 border-line for
fine grain materials with a requirement that more than 30 percent
of the material shall be smaller than 20 ium, one obtains a region
in this diagram which represents fine grain materials.. ithin
this region, the quartz content varies from 0 to about 30 percent.
Since these numbers are based on mean values for each fraction, the
possible variation is some-what larger and can be egtimated to
range from 0 to 45 percent.

This corresponds to a possible variation in particle conductivity

from 2 to about 3.2 .,7/mK for fine grain soil materials. The dis-

235
"......A
tribution function obtained for coarse materials gave a possible
variation in particle conductivity from about 2.2 to 5 t/i/mK, if
one neglects the most extreme 10 percent of the results.

Based on these circumstances one can expect thermal conductivity

variations in water saturated Norwegian soils as shown in Figure 8.
Expected variations in dry density for the two groups of soil mate-
rials are also indicated, For coarse grain materials this range
3
is from 1500 to 2200 kg/m while fine grain materials may vary from
1200 to 1800 kg/m3 . For comparison, reoults obtained from Kersten's
equation are included.

w/mK

1.0 ' '2

,
,• "M+
wsm 'W,. ,
11 U11111t

a .II , tI i . • .
o 010 0. 0C Oe
O 1.0
AWN,11 FV•IUIUPl
N FOA114i6l,,

P-IN. S. Dellltained •ltl'n~l ln lPvntar Vari•mmvtodl ard iro o

J•POl• t',
tu,r
Keorstene lik injr s,,m.'e.nl ~knt or-do
.,vA lot Li1qtt•,V.n
veWhgde varL miijlc'ui1 ij o|¶ v'r • r:.ll r", ils'.iI'.

1d*?elon' dqual .o@ l oIompar#J r ci•I 0'4 oPp -

COJ ifsed of V,'L•G'ias e for SoralV apt foi:.,

As shown by Figure 8, Kersten's curves for coarse grain soils fall

close to the upper limit of the region representing Nrorwegian
coarse grain soil materials. Kersten's curve for fine grain ma-
terials cuts diagonally across the region representing fine grain
soils.

One also finds variations around mean curves for the two regions
of between 20 and 30 percent for coarse grain soils and from 10
to 15 percent for fine grain soil materials. The mean curves
represent particle conductivities of 3.5 and 2.6 1V/mK for coarse

236
- +•..,• • , ......--- --•" ',,'r ..........:w .... ... • ' - , ,_• • +...4M'I
and fine Grain soils, respectively. This corresponds to quartz
contents of 40 and 20 percent, respectively.

Figure 9 presents corresponding regions of variation for the con-

ductivity in frozen, saturated soil materials, as well as results
obtained from Iersten'i equations for such soils. Also in this
case, the latter curve falls in the upper portion of the variation
rogion for coarse grain materials. to"ever, for fine grain soils
IKersten's curve passes through the cross-sectioned region at a sig-
nificantly lower level than for the case of un-frozen materials
(Figure 8).

4.0 i

03V -

F K9. . I I.....

j r . '; ',99

1[,
0 ., 4' 1 ,- ' VO.. %4.JL11 ",

f' 1W1 td o l4' !'o m "d

jdýW 'I.vd fJtWI' ' lo Ut

, L hT I 1 , 4 . C

Thi3 li'vor poa~lt~ion ý(.ýf th.a C."rste;a cu.va) can be a.3.xed to be

caused by a certain average portion of un-frozen water in tha fino
Crin samples, due to the tomperatui a used during the experimento
(mean temperature -4*Ct teraperature difference 5.5'0), The varia-
tion' around mean curves through the two regiona are, aG in FiEure
8Ip bet'33en 20 and 30 percent for
.rain
coarse
.Oteria3 ,J;.r,|a~r
1•1. :I' ,4r,
+•gl.+1,+•
;]L l)"d+; PII"•t~'v•
• ,' ", '. fr.-A arnd from
10 to 15 percent for fine -rain aoile, H~owever, expectod varia-
tione in relative un-frozen water content will cause significant
widening of the range for fine grain soils, This is indicated in
Figure 9 by extending the actual reGion of variations in

237
particleI) conductivity with a broken line, representing an un-
frozen water content of 10 percent. (Nu = 0.10)

Variations in un-frozen water content in fine grain soil materials

was discussed in chapter III. There, a diagram was presented for
determining the variations in un-frozen water content at tempera-
tures below the normal freezing point when the flow-limit for the
material is known. This variation in un-frozen water content at
temperatures below the freezing point will, as mentioned earlier,
also affect the apparent thermal capacitance in soil materials
during freezing. This effect can be expected to be more important
for the temperature field distribution in freezing soil than the
related effect in thermal conductivity.

The previous discussion of Keratcn'I equatio'i was limited to con-

ditions in saturated soils. As indicated earlier, this condition
represents an extreme value for the conductivity variation for vary-
ing degrees of saturation. The other extreme, i.e. conductivity
for dry soil materials, can not be derived fron these equations
since they are only valid to limits of 3 and 7 percent water content
by weight, for coarse and fine grain soils, respectively. How-
ever, the empirically derived equation for variations in conducti-
vity of dry materials can be used to estimate the lower extreme
value. This permits extension of Kersten's equation, using the same
normalized representation as when his experimental results were
discussed. The results of such a treatznont of Kersten's empirical
equations are shown in Figure 10 for different densities.

1) This looks like a mistake. The variation discussed here occurs

in the over-all conductivity, not in the particle conductivity.
(Translator's note)

238
 IWA
I "",,, I //

0,04 01 0O110SO3 4 0% 081'0

?:a. 10 Kerstena Itkninger rramstut ,% iiotrmaiis,o~rt form, itim-

'Nhen presented in this form, Kersten's equations show certain vari-

ations due to dry density, while no siuch variations were found in
the results from his study (see Figure 27 of chapter V), On the
i•, other hand, the agreement with previously derived empirical re-
lt+! lations between degree of saturation and normalized conductivity
I is satisfactory) particularly
for fine grain materials.
I.-,
This discussion of Kersten's equations has thus shown that these •
equations can be regarded as a satisfactory empiric presentation
of the experimental data they were derived from. Htowever, these.,
data are not completely representative for Norwegian sails. P•ar-
ticularly for th~e group coarse grain soils the data show., a siqnifi-
cantly higher average quartz content than one can expect to find
in N~orwegian soils. T4his will in most cases result in too high
estimates for conductivity in Norwegian coarse grain soil mate-
rials. The large variation in quartz content which can. bo• ?xrec- "
ted for such materials will also Siveý large uncertainty -In c¢unduc-
tivity estimates, even if one mana•e s t o c o re n t 'a r '.'.e ,
average! quartz c-rt,.;nt.
For fpnedrin materials, the expected variation in un-frozen v

water content is a particularly "heavy load" on the accuracy of

the equationen while variations in quartz conoent in general will
be less important than for coarse materials, An improved method

_I 239
should thus include possibilities to account for varying quartz
content in coarse grain materials, while the effects of un-frozen
water should be included for fine grain soils. As mentioned be-
fore, it may also be desirable to change the method for classi-
fication of the two groups of soil.

As mentioned previously, Makowski and Mochlinsky (1956) developed

an equation for thermal conductivity in soils, based on a semi-
empirical mathematical model generated by Gemant (1951) (1, 2).
This equation was put in the same form as Karstents empirical
equations for un-frozen soil materials.
- (alog(W%) .. ~ )1 0 o.62yd (W/mK)

where Y is dry density expressed in metric tons/m3

Sand S are factors depending on clay content c (percent)

* (1.4241 - 0.00465 c)I0I

3 u(0.4192 - 0.00313 c)10"1

The clay content was said to affect the conductivity of soil par-
ticles. In clay-free materials, the particle conductivity was
assumed to be 5.84 W/mK while a certain clay content c (percent)
was assumed to reduce this value according to the following formu-
la.
X2 a 5.84 - 0.033c 10

According to this expression, pure clay would have a particle

conductivity of 2.51 W/mK. This corresponds to a quartz content
as high as 80 percent for clay-free materials and 15 percent for
pure clay, according to the previously proposed model for deter-
mining particle conductivity.

Figure 11 shows the conductivity for water saturated clay-free

240
 sand, and for pure clay determined from the equation given by Makowski & Mochlinsky.
Both these curves are close to that derived from the geometric mean
equation on the basis of assumed particle conductivities.

The curve for clay-free materials will consequently give signifi-

cantly higher values than those which can be expected for Norwegian
coarse grain materials, while the curve for pure clay is represen-
tative for fine grain materials.

' Figure 12 shows data for varying degrees of saturation, obtained

from the same equation and presented in normalized form for diffe-
rent densities. The derived relations between Kersten's number
and the degree of saturation are included for comparison.

One finds that the curves for clay-free materials on the average
follow the empirical relation for coarse grain materials, while
for pure clay the curve falls above the empirical relation for
fine grain materials.

:'ii.

AATK I LYOLIMA'NU1L

IP13.I , •I~ t n 1vdfln.r•os~o n*1' ott t, Y,'CKOwnki 04 MO,!r,'l nRK ,

241
 OW
0 ;.2 l . CIL
A .4
//o

S0 ', I. . "

0.1 0 ,3 C2 06 01 10
W 41t4I•'4IRA0

P!C . 1 I2 MlItOWtk Oiglochlinaisi. 1tkntlijgr tramti.lt pA norrAn1i-

Ler troi s, 0nMiV1n1tno Mell dt qinpjwleji 1JOktutnger CIar
`iovIk;rnfd9 og rilnk,)rnlde 4jWira,,'tlhJ.
hi'k; gAJ
N00.10e/Wqet'i d~q.aICO' iq JI. Mt:~l•Mda tipPueniMId9N0
tii Shi e•IPsilt.. 4at iOM4 for jOroi tjr j*,i1.1d

The method developed by Makowski and Mochlinsky is proposed as

an improvement to Kersten~s method, since the former offers a
possibility for continous transition between the two groups of
soil materials defined by Kersten. This "transition" is tied to
an expected variation of the particle conductivity for varying
clay content. As shown earlier in this chapter and by others,
variations in particle conductivity will primarily be due to dif-
ferences in mineral composition, particularly the quartz content.
In more fine grain materials it has indeed been shown that a cer-
tain correlation exists between quartz content and texture. How-
ever, this correlation can not be satisfactorily represented by
a two component model (clay, "no clay"). A three component model
is needed, in which the non-clay fraction is separated into two
components, d >20 wm and 2 <d <2C Nm.

The normalized presentation of Kersten's experimental results

showed different functional dependence between Kersten's number
and the degree of saturation for coarse and fine grain materials.
This difference is not present when Makowski and Mochlinaky's method
is used, see Figure 12,

The work by Makowski and Mochlinsky was directed towards conditions

when high voltage cables are buried in the ground. In that case,
there is a danger for overheating in the cable insulation where
the surrounding ground has unusually low thermal conductivity.

242
...
-------------------------------------.---
For this reason, they only treated un-frozen soils.

Van Rooyen and Winterkorn (1959) later presented a report from

work having the same purpose (8). They performed thermal conduc-
tivity, using thermal conduction probes, for a series of soils
and modified 1 ) materials (crushed quartz, fractions of sand with
high quartz content), all with relatively high quartz content.
Based on measurements performed for different combinations of den-
sities and water content, they derived an empirical method for de-
termining thermal resistivity of un-frozen soil materials, i.e.
the inverse of thermal conductivity. Their equation has the form

r - A 1i0 + a (cm/oC4) 11

where A, B and s are empirical functions of dry density, mineral

composition and particle size distribution.

For completely water saturated materials, the thermal resistance

was set equal to s, which was expressed as a linear function of
dry density

s - sI - s 2 Yd 12

where sI and s. are linear functions of the quartz content, while

Y is dry density in metric tons/m3 .

This in principle agrees with what has been pointed out earlier:
The thermal conductivity of water saturated soil material depends
primarily on relative particle volume and particle conductivity,
while the sensitivity to differences in microstructure is small.
However, calculations using this equation to determine conductivity
of saturated materials with different quartz contents show that

1) "Kunstig" means "artificial" or "man-made". In this context

the term means "modified", e.g. by crushing or screening
(Translator's note).

243
van Rooyen's equation tend to overemphasize the effect of in-
creasing dry density, particularly for low quartz contents. This
is illustrated by Figure 13, which shows calculated conductivity
for three values of quartz content and data obtained from the geo-
metric mean equation for comparison.

Figure 14 shows how the conductivity varies with varying degree

of saturation and different densities, according to van Rooyents
equation. The data are presented in the normalized form used pre-
viously. For comparison, van Rooyen's measured results for sand
with high quartz content are also shown in Figure 14.

%-C

2.0

'.01-2.0
.40 '. ' i. 2..

-dJ' ýqIt .¶V ',c PI ZIt g,',

10

Oaa.

k~Ir tiaIllilld. if
Vatn il-Oql
fI'~~~ $if~,'n I),"9

Q4r. t
*ju1i.... ' 1 "'h' • uVIJ "'..

0.21 4

I, M / M'
%.0 0.' 02 04086 t0
244F Sf.

FIG. 14. van f oyen . v

ekninKjer ir -lO
tid mod '1301 kY týe t.rhiA tilt
pA n.)rmiiljiuirt :'OrM* su.. : r.. j l t.jrijIP i inr.ýr PI
244.I "
 The calculated curves evidently give poor agreement with measured
values for degrees of saturation above 20 percent.

For dry materials, the expression for thermal resistivity is

r = A+ S 13

The term A is given as an exponential. function of dry density, clay

content and specific surface area for the parts of the material
which have particles larger than those within the clay fraction.
Due to the structure of the exponent in A, the values for dry
materials are very sensitive to variations in clay content, as
shown in Table I.

* Table I. Effect of clay content on conductivity of dry soil,

calculated from van Rooyen's equation. Dry density
Ki 3
1400 kg/m . Specific surface area for fractions lar-
ger than clay: 2,000 cm 2 /cm 3 . 40 percent quartz.

Clay content, percent 0 5 10 20 50

Dry conductivity, W/mK 0.22 0.30 0.52 0.74 0.94

A corresponding variation is found for soils not containing clay

when the specific surface area for fractions larger than clay is
varied, see Table II.

Table II. Effect of specific surface area on the conductivity

for dry soil calculated from van Rooyen's equation.
Dry density: 1800 kg/m 3 . No clay content. 80 per-
cent quartz.

Specific surface 50 100 200 500 2000

2 3
area cm /cm
Dry conductivity 0.31 0.48 0.59 0.67 0.72
W/mK

245
 In both these cases, the conductivity variations are larger than
those found for dry materials during the study discussed earlier
in the last chapter. In addition, that study gave no basis for grou-
ping conductivities in soil materials according to differences in
texture, except for the proposed separation between natural soil
materials and crushed rocks.

The previous-analysis shows that van Rooyen's equation gives uncer-

tain values for conductivity in soils, despite the large number of
parameters involved and the complicated form of the equation.
Part of the explanation for this can be found in the specialized
choice of soil materials that were used for the experiments. Two
of the soil samples were so called "thermal sands" which have ab-
normally high conductivities both when dry and moist, as was shown
in the sections of the previous chapter discussing conductivity of
mineral soil materials.

C. Conclusion

The previous discussion of methods for computing conductivity of

soil materials has shown that analytical methods only can be used
to obtain reliable results if they are supplemented by empi-
rical relations.

Out of the purely empirical approaches, Kersten's method is shown

to give the best agreement with conductivities of natural soil
materials one can expect to find in Norway. The agreement is best
for fine grain soils, while Kersten's values for more nandy mate-
rials will be somewhat high in comparison with the expected range
of variation for Norwegian soil materials. This is mainly due to
the selection of soil samples on which Kereten's empirical equa-
tions were based. In particular, the coarse grain materials sho-
1wed considerably higher quartz contents than those to be expected
on corresponding Norwegian materials.

kI 246
Another draw-back inherent in the Kersten method is the mixture
of factors related to texture and mineral content implicit in
his equations. An improved method should treat the effects of
these factors separately, e.g. so that both coarse grain soils
having low quartz content and fine grain soil materials with
high quartz content can be correctly related.

The discussion of van Rooyen and Winterkorn's empirical method

showed that it is difficult to improve on the methods for calcu-
lating (determining) thermal conductivity of soil materials by
means of empirical methods. The number of variables is so great,
that the effect of each individual parameter on the conductivity
best can be separated out by a more fundamental analysis of physi-
cal relations. However, as mentioned previously, such an analysis
can not be successful without support by experimental investiga-
tions,

In the following, a new method for determining thermal conductivi-

ty will be developed on the basis of theoretical and empirical
studies described in this and previous chapters. These studies
also form a basis for estimating the error limits inherent in the
method, depending on how much information one has about the soil
material at hand. The latter is important in different applications
of the method, which pose different requirements on accuracy of
the results. The mathematical model will be designed accordingly
and with respect to requirements for studies of materials at hand.

A discussion of the effects of variations in thermal conductivity

on the development of temperature fields, e.g. in road banks and
subsoils, will also be included#

247
 2. NEV METHODS FOR DETERMINING THERMAL
CONDUCTIVITY IN SOIL MATERIALS

A. Introduction

The previous evaluation of methods for determining thermal conduc-

tivity in soil materials showed that existing methods result in
relatively wide error limits which can only be improved upon by
introducing mineral as well as texture dependent factors as para-
meters in the mathematical models. The foundation for such an
improved model was laid in the previous discussion of empirical
methods (chapter V), In this section, the emphasis will be placed
on a presentation of this mathematical model, by means of graphical
methods for determining the conductivity of soil materials. Rela-
tions between information level (in the sense of knowledge about
soil parameters at hand) and resulting error (confidence) limits
will be discussed in the next section. The effects of variations
in different soil parameters in the development of temperature
fields in freezing ground will also be treated.

B. Mathematical model for mineral soils

The foundation for this model was developed in chapter V, where

thermal conductivity measurements on dry, moist and saturated
soil materials were discussed. In the following, the main aspects
of the model will be summarized, with references to Table III,
where the different parts of the model are listed.

The mathematical model can be formulated in terms of a main equa-

tion (Eq(a) in Table III) which gives the soil material conducti-
vity for certain degrees of packing and saturation as a function
of conductivities for the soil material in dry and saturated
states for the degree of packing at hand, as well as values of the
normalized conductivity (Kersten's number) at the actual degree
of saturation.
These variables can be determined from information about the
texture of the soil material (screening curve and particle shape),
degrees of packing and saturation, as well as the mineral compo-
sition (quartz content) which determines the particle conductivity.

The conductivity of dry soils was found to be rather insensitive

to variations in particle conductivity. On the other hand, textu-
ral conditions tied to particle shape, etc. turned out to be of
major importance. However, the material studied gave no means for
differentiating between soil materials having different particle
size distributions, except for a special group of materials where
a limited amount of clay was mixed into an otherwise typical sand
material ("thermal sand"). Artificially crushed rocks were also
found to have conductivities which are significatly different from
those of natural soil materials. On this basis, two empirical re-
lations were developed between dry conductivity and porosity, one
for natural soils and one for crushed rock (Eqs (b) and (c) in
Table III). Due to the limited variation in specific weights of
soil materials, these relations can also be expressed as functions
of dry density.

For saturated soil materials it was found that the conductivity is

not very sensitive to variations in texture. Thus, the cornductivi-
ty can in that case be determined if volume ratios and particle
conductivity are known, without consideration for differences in
texture. The geometric mean equation was found to yield a good
approximation to the theoretical results obtained from the Hashin-
Shtrikman limits (Eqs (g) and (h) in Table III). A mathematical
model for the particle conductivity was developed on the basis of
certain trends for mineral composition in various soils, where
quartz, feldspar and mica are the dominant minerals. Of these,
quartz has the highest conductivity (7.7 W/mK) while feldspar and
mica have approximately equal conductivities (2.0 W/mK). Based on
this, particle conductivity could be determined from a two compo-
nent model, with quartz content as the main factor (Eq (i) in Tab-
le III). This model is valid for soil materials that are generated

249
 BEST AVAILABLE CORY
Tabpll 11Z. nrciningsrodo11 for Vkr-m.l9dni~evng.I V rv
,rnva.l4e joo r'ur.
"?L
th Trt calculatitg thermal conductivity of irnr soia.

- Ke . . . .......... . (a) Mcvedlikning.

,: inf qi.ation.

v ;,r,: iAL 2700 -0,9LA7 Yd Tý.rr led.,±rip yrt..

- ~* QQ' ,2 Dry con*ýuctivitýY.
L idT:F?• . . .. . ............... . ( )

0,7 106 $r +,0. ............. (d)

0..... N10 Sr + 100 . .. .. .,?v,.? (e) e ivi,'..

FINE
* -,, *t en - tl

::" ';,,4E T S 5r ..
t . .. . ............... (f

S......... . '• ..............

.;LLN- • - ,•..-.-\re

S.. . .
i•.'-'..• 2,2n .0, -nO z6 u ...... (h) : ' :a r' :J ' y

t• ~7,7q '2,01' ,, . F,:t

ot • e 1 :.r.. v.no

,31j -;. .. I.'t •.r. e •:,•t i v i ty

Sq ,0.20 GNOV 7 ,iq ,

COARSE 2

250
14
mainly from rocks containing quartz, such as granite and gneiss.
Soils originating from alkaline rocks, such as gabbro, will con-
tain small quantities of quartz, as well as considerable amounts
of other minerals whose conductivities fall between the two impor-
tant groups mentioned. Coarse grain materials with low quartz
contents should consequently be treated separately (Eq (J) in Tab-
le 11). For fine grain materials it is proposed that Eq (i) be
used also for low quartz contents, since minerals in the interme-
diate groups seem to occur in small quantities in the finest frac-
tions.

Empirical relations between the normalized conductivity (Kersten's

number) and degree of saturation were derived for the range from
dry to fully saturated, These relations were found to have low
sensitivity to variations in packing degree (density). However,
for un-frozen materials, a systematic difference was found between
materials containing no clay fractions and those having a certain
clay content (Eqs (d) and (e) in Table III). No similar difference
could be demonstrated for frozen materials (0q (f)).

This mathematical model was used to generate diagrams for graphical

determination of conductivity in mineral-based soils. These dia-
grams are shown in Figure 15, which also contains examples of their
use. The input parameters are also described in more detail. This
model can also be used to generate computer programs. Such pro-
grams are currently being propared at the Institute for Cold Tech-
nology, NTH.

251.
4 ~'i
'HE-MAL CZOJOUCtIvTY OF A'RU OIL

-b

W,,I,
.3

a CRUI
0 20 4.0 50 80 100 12 1.4. 16 le 2.0 01 02 0.4 0.6 10

allA T ICNTE4T CRY CESSITI DlrRLE OF ý,AT~JAI~ION

Fig. l5a. \'z~~.en~ av ufrosrie mndra-

ral~ke Jcra'crter. le;,rmat ocndt.-a
t~ivity of unj.roain mi~nfraz soilsa.

1. Yvartsinnhold 7. utzoen
~.Orov/firn ~ OZiS/~

.ov

v".k
4.
"L e.JKC-
n i~a
vonvk
f a,:.,
1.
t

4c
*PAVE :77:;.....
VIA V"O
25Ir .

!Z5 ...
 1.IE3'.kL. CCVl4D'.-CTVITY OF 'CkC7EN 9YILS

(WIT0* S0

2.0

-~10

- . 2

..
80
2040K0 r .4..

1.6 IIa
i010/f 20p 0 02 04 06 06
Q 20 0 B lU 1.2 1,1

enl~~i AV fvci"sne irne-

P!S. '5b.

BEST As aILABL
1 rtvtj Sf Inn h' . ~ ,wr o~

I-.~a of e~fJ

E~.n

C~*.0 AVEL

. .
* ,-.j j.~.....,..t.t ij.*2

Ex 2 b '-

-.It'v f~kAY

17-
e.-

253
 3. EFFECTS OF VARIATIONS IN SOIL
PARA14ETERS

A. Effects on conductivity calculations

The mathematical models developed in preceding sections make

it possible to determine thermal conductivity.of mineral-based
soils with normally required accuracy, provided that relevant
soil parameters are given (known). However, one usually has to
utilize data from more or less complete geotschnical investigations.
the parameters which are lacking will in such cases cause uncer-
tainties in the estimated conductivity, which can be predicted only
if one knows the statistical distribution of the parameter(s) at
hand, within the soil material studied. As an example, if a soil
matirial is classified in general terms, such as sand/gravel or
silt/clay, the uncertainty of the conductivity estimate will depend
on the statistical distribution of all relevant soil parameters in
these material groups. If only information about quartz content
is lackingg the statistical variation of quartz content in the soil
type at hand will determine the error (confidence) limits for the
calculated conductivity.

The two cases Just mentioned can be said to represent two informa-
tion levels, each having characteristic error limits for the con-
ductivity estimate. In chapter III, such levels of information
were given in a block diagram containing the different parameters
which affect conductivity in a soil material. This block diagram
is reproduced in Figure 16.

In addition to the three levels shown in Figure 16 one could in-

troduce a lowest level of information where one, as mentioned, only
has access to a broad classification of the soil materials, without
any of the relevant soil parameters being given. For example, in
connection with development of insulation designs for shallow
foundations (houses without basements) this may be the situation
(Thue 1973)(10).

254
 L.,

'- 1 tl e o -)1--"a jI pa"r--

t ") . .... a 'r '-...'.1- h.. " d. tI

If one has access to (results from) a sufficiently large number

of thermal conduction measurements performed in the field, one
can form estimates of expected mean values and variance (spread)
for the conductivity of soil materials at hand, rather than taking
the route via statistical distributions of individual soil para-
meters, Such data are summarized in an American study from 1960,
which consisted of thermal resistance measurements in soils in
connection with placing high voltage cables in the ground (Sinclair
1960) (11). The results from more than 900 measurements on sand
materials and 00 measurements on clay materials are shown in
Figure 17, in the form of cumulative distribution curves for the
conductivity.

* •tL'
IO•, I '.'a"%

101 '

Thermal Cohductivity
!P1 .7 , rCNuat v, tIOr' tl. r *e; '.4V
.h' r. wqevrin t1'tr sond to,

-. v.r.j. 255, .' r- '..* .

255
For sand materials, these results show a near normal distribution,
with a mean value of about 1.90 W/mK and a standard deviation of
0.60 W/mK. 90 percent of the results fall between the limits
0.90 and 2.9 W/mK. The results for clay materials deviate sig-
nificantly from a normal distribution. The mean value was deter-
mined to be 1.80 WN/mK and 50 percent of the results fell within
the limits 1.60 and 2.0 W/AK. However, the rest of the data showed
a wider variation, with the 90 percent limits at 1.20 and 3.20
W/mK.

These results can not without reservations be assumed relevant for

Norwegian conditions but they give a notion of error limits which
must be accepted if conductivity estimates are based on broad clas-
sifications of soil materials.

Since no results are available from field measurements on Norwegian

soils, the error limits for conductivity estimated on the lowest
level (of information) must be determined from information about
mean values and variations in relevant soil parameters. There are
few studies on which a determination of such distributions can be
based. Thus, one must resort to more arbitrary estimates.

Angen (1973) has published a table showing dry densityties for 19

different Norwegian clays (12). These values are used to estimate
a distribution having a mean of 1400 kg/m3 and a standard deviation
of 90 kg/m3 . He also includes a table over typical dry densities
in sand materials. The values vary from 1430 k9/ 3 for loosely pac-
ked, single grain size sand, to 2120 kg/m 3 for a morain material
with large variations in particle size. If the extreme values are
assumed to represent 95 percent of the cases, the standard devia-
tion will be 170 kg/m3 , around a mean calue of 1770 kg/m3 . These
estimates of dry density distributions are shown graphically in
Figure 18.
4A
";
I " /LAM

.1 I I'/ '
•0 r' . ...
" ... . . .....,. ... ..I.. .. !

12 1.4 .0 Ill 2.0 kymI 12 -%U1

0" ROMVY.T
Dry Dengity
plO. 18. KLX1ulativw fordoUrnur kv torr ronvvY•t for vleft fig
aaid dv&
r, pA .'!. v , .tV ,'.r•ggh
4 1(AV
j1 • )
O J 1.VO d
kmML ~tri d Lv tt r ' ".), 4
!,tl .. ,§ oiif d irpV 0,,

Angen also performed an analysis of water content in sub-soils

consisting of silt and clay, based on geotechnical investigations.
The variations in degree of saturation were small for these mate-
rials, with a mean of 90 percent and a standard deviation of 5
percent. For sandy materials, few measurements of water content
in free ground have been reported. The distribution must for this
case be a pure estimate (guess). It is proposed here to use a
mean value of 50 percent and a standard deviation of 20 percent.
This distribution corresponds to 95 percent of the cases having
degrees of saturation between 10 and 90 percent.

In connection with frost registrations performed by the National

Road Laboratory, moisture contents were measured when a total
of 15 road beds were excavated in the 0stland region (13). The
cumulative distribution for moisture contents in road beds (pave-
ments) shown in Figure 19 are based on these measurements. The
arithmetic mean is 5*.5 percent by weight. For a dry density of
1900 kg/m3 , the median corresponds to a degree of saturation of
30 percent. In the following, this will be considered as a re-
presentative average value, while the standard deviation for the
degree of saturation is estimated to be 10 percent. This distri-
bution thus implies that 95 percent of the cases will show degrees
of saturation between 10 and 50 percent, corresponding to from
about 1.5 to 8 percent by weight for a dry density of 1900 kg/m3 .

257
 .10

' 0
4~

20

* VANOIVNNPýO00 VE(T %,

Iplo 19. xKuistia~v rordw1itv, a' voriminnil

LJi -vrynr*

d'~
.9iJsu
IJr w t 01 )atvP IJji uj 1

258
Little data exist about the distribution of dry density in road
beds (pavements). It is proposed here to assume a distribution
having a mean of 1900 kg/m3 and a standard deviation of 75 kg/m3 ,
which implies that 95 percent of the cases fall between 1750 and
2050 kg/M3 .

SveianOs study of quartz content in Norwegian loose deposits can

form a basis for similar estimates of quartz content distributions
in coarse grain materials (6). These results were given in the
form of a cumulative distribution curve in Figure 5 in this chapter.
This curve can be approximated by a normal distribution with a
mean of 37 percent and a standard deviation of 15 percent.

In the case of fine grain materialsresults from quartz content

measurements are available for fractions of Norwegian clays (7).
These results are also shown as cumulative distribution curves in
Figure 5 in section I of this chapter. Based on these results, the
quartz content in fine grain materials was assumed to fall between
0 and 45 percent. This may be construed to imply a distribution
with a mean of 20 percent and a standard deviation 7.5 percent for
clay and clay/silt, while for silt the mean is assumed to be
30Z: 7.5 percent. For the first group, 95 percent of the cases will
then occur between 5 and 35 percent, while corresponding numbers
for silt are 15 and 45 percent.

The distributions just described are summarized in Table IV. Even

if most of these should be regarded as approximate estimates, they
serve to illustrate the principle behind conductivity error limit
evaluations, until more conclusive results become available.

259

' .. ,**,-*
... i. . ... ....
Table IV. Assumed (estimated) distributions for key soil para-
meters, to be used when calculating thermal conducti-
vity. Mean values and standard deviations.

Dry density Saturation Quartz content

kg/m 3 percent percent

Clay/silt 1400±90 90+5 20 +7.5

Silt 1600 100 90+5 30 +7.5
Sand, free ground 1770+ 170 50 +20 37 +15
Gravel, pavement 1900 75 30 +10 40 +10

Based on these distributions, error limits for the conductivity

can be determined by means of an error analysis using the proposed
mathematical model as a starting point. This error analysis is
performed in the normal manner, using partial derivatives and sums
of squares.

Table V shows the results of such an error analysis, performed for

three levels of information for both sand and clay. At the middle
level, all parameters are given except quartz content, while at
the third level dry density, degree of saturation and quartz con-
tent are all known,

The values in row six illustrate the effects of uncertainties in

each individual parameter. For example, one finds that the uncer-
tainty in dry conductivity is an insignificant part of the total
error. When passing to the level where dry density and degree of
saturation are known, one finds that the effect of uncertainties
in Kersten's number is about halvede At the same time, the effect
of uncertainties in saturated conductivity is reduced by about 50
percent. If also the quartz content is known, this effect is fur-
ther reduced. In all cases, there is a significant reduction of
relative errors in the conductivity estimates, from about ±40 per-
cent at the "lowest" level to only 12-14 percent at the "highest"

260 F
Table V. Error (confidence) limits in calculated conductivity
of soil materials. Un-frozen sand and clay.

Level 1 2 3
1. Texture Sand clay Sand clay Sand clay
2. Dry density 1770 1400 1770 1400 1770 1400
A d 340 180 -0 -0 -0 -0
3. quartz content 0.37 0.20 0.37 0.20 0.37 0.20
Aq 0.30 0.16 0.30 0.16 -0 -0
4. Deg. of satura- 0.50 0.90 0.50 0.90 0.50 0.90
tion A Sr 0.40 0.10 -0 -0 -0 -0
5. Conductivity 1.50 1.15 1.50 1.15 1.50 1.15
(x/;0) L o 0.056 0.005 0.03 0.004 0.03 0.004
6, 1
(x/aX ) X1 0.498 0.387 0.292 0.271 0.097 0.127
(0X/aKe) AKe 0.393 0.187 0.150 0.103 0.150 0.105

7. A 0.63 0.43 0.33 0.29 0.18 0.17

8. Relative error
percent 42.0 37.5 22.0 25.2 12.0 14.3

One also notes that the errors are about 25 percent in cases where
the quartz content is not known, while dry density and degree of
saturation are given. This level corresponds approximately to the
level of Kersten's empirical equations, for which the uncertainty
also was found to be t25 percent.

In this summaryp the calculations are based on average values for

dry density and degree of saturation. Table VI also shows how the
error limits vary with degree of saturation for the two highest
levels.

261
Table VI. Error (confidence) limits for different degrees of
saturation. Un-frozen sand.

[Degree of saturation 0.15 0.25 0.50 0.75

Dry density .. 1770. 1770 1770 / 1770
Quartz content 0.3 0.37
0 * 0.37.* 0.37
Error limits, percent 18 20 13.6 18 12 12.1 26.5
Not known S
Figure 20 shows calculated values plotted versus degree of satura-
tion.

From Table VI and Figure 20 one can see that the improvement in
error limits due to knowledge of the quartz content is largest
for the highest degrees of saturation.

As mentioned previously, the degree of saturation is relatively

low in road pavement materials, about 30 percent, while the dry
density is significantly higher than in the previous example, about
1900 kg/m.

Error limits for the conductivity were calculated for the varia-
tions in dry density and degree of saturation estimated previously
(see Table IV) and four different cases, see Table VI. For the two
first cases, dry density and degree of saturation are assumed to
deviate from their means by one a# while in the other two cases
these parameters are assumed to be known. For cases 1 and 3, the
quartz content deviates from the mean value by one • while it is
assumed known in cases 2 and 4.

262
 Y-
3
- I' "gm . - f.D - (KVAR-SNNHOL. 37%/,
-QUA KJ IGON r r,'NT

20r' - - KVARTSINNHO'.D JIU.ENT.

e ,i: 10 --

0 ,, I I

0 0.20 C.0 05r 0.8' 1.

M ET'NING SO AD

iJ | vrun
enlwvd
In W
Io n a •s U'
t re,'ii.i n,i
I~ J.
r t j ,I1i
.' :wr,
•
re.+ivr, iiv
W , P C. 0 "1
}0.
$ 40 1 t
,..rI
. I. ?' a t4C u a

Table VII. Error (confidence) limits for estimated conductivi-

ties in frozen gravel road bedI).

Case 4

Dry density 1900 1900 1900 1900

* .as
....e
S•d1I Y
. .. .. .......... ..
0 150 0 . ..O,
0
Deg,. of satura- 0.30 0.30 0.20 0.20
tion ASr 0:20 0.20 0 0
Quartz contont 0.40 Io0.40 0.40 0.40
Aq 0.20 0 0.20 0
Conductivity, 1.15- 1.15 1.15 1.15
frozen, W/m.K
Relative error 52.?7 47.3 2.1.2 :
percent 5 4 2

This example illustrates that Inowing the quartz content contribu-

tea little towards reducing the error limits of conductivity
estimates for road surface layers, particularly when dry sensity
and degree of saturation are not accurately known. By contrast,
one finds that the error limits are reduced to about one third
1) "overbygning" refers to the upper layers in a road, e.g. "pave-
ment" for concreteor asphalt covered roads and "bed" or "sur-
face layer" for gravel roads (Translator's note). /

263
 when data for degree of saturation and density can be given exact-
ly. Thus, it is more important to obtain reliable data on den-
sity and moisture content in road surface layers than to accurate-
ly determine the quartz content.

For these relatively low degrees of saturation, uncertainties in

the estimate of dry conductivity are rather important. As mention-
ed previously, investigations of conductivity in dry soil materials
gave cause for distinction between two types of texture - natural
soil materials and crushed rocks. The empirical relations derived
from these two groups are illustrated by Figure 21, where the range
of variation also is indicated.

2.0.

w.f .'"

?, •~~~..1.i• • .... .. - J;,~

f0 - - •. ,

Dry Density

If one considers the case where no information about particle shape

is available, the uncertainty in the estimate of dry density may
increase significantly, particularly for high densities. For ex-
3 the error limits will increase
ample, at a density of 2000 kg/m
from the +20-25 percent given previously to nearly 50 percent.
This increase in uncertainty will be most pronounced for low de-
grees of saturation, where errors in computed conductivity are

264
 dominated by the uncertainty in dry conductivity. This is also
shown in Figure 22, where error limits fq r the case of lacking par-
ticle shape information is compared to those obtained when all
relevant information about the gravel ma'terial at hand is available.

* For high packing densities, an uncertainty in the specific weight

estimate will also affect the error lirhits for conductivity esti-
mates. This is particularly evident f/or #ushed gravel, whore
dry conductivity is most dependent onpacking density. This is
illustrated by Figure 23, which shows' err~or limits when calculating
conductivity of crushed gravel with and without knowledge of speci-
fic weight. For natural gravel the. effect of uncertainty in spe-
cific weight is important only for iower degrees of saturation.

S'1' \ ,~

\I C

I- N

7! .13 i

I._ _ _ _ _ _ _ _ _
,
S.... . ..... . . .. ... ....... . ,•........ .. , ? - . , . •: ;=+.,+.•.++.•f•..._.•. . ..... ..... ...

S• ~

: ,: 265
 - '20 S, ..

0 I. '-. C 2' 22,,'0l

-,2 10

F'IG. .1 . tnn •.y•,e@.t - 1 AvXzrf.'#An

kJP til ;re,.orlkk V~lkt Vw,! I)#-
N, Minmq AV l.llil1 A~',
V I' ¥.; ,(Rilst iF.nv• ,

B. Conclusions
S1The in the previous examples show that
error analysis performed

uncertainties in various soil parameters are important to a vary-

ing degree, depending on degrees of saturation, density and type
of material. Without knowledge of any of the relevant soil para-
meters one can estimate the conductivity of different types of ma-
terials within error limits that vary from 40 to 50 percent. If
dry density and degree of saturation are known, these error limits
can just about be cut in half for most types of materials. Another
reduction of the error limits by a factor of two can be obtained
for materiala near saturation, provided that the quartz content
is known. For road surface materials, where the degree of satu-
ration normally is lowlinformation about the quartz content normal.-
ly gives a small improvement of the error limits, while information
concerning particle shape and specific weight can reduce the err'or
limits to about t15 percent at normal degrees of saturation.

In cases where one lacks relevant soil data one must resort to uaing
average values for the conductivity, with inherent uncertainties.

266
Such values are listed in Table IX. Average values and variations
for the soil parameters on which these conductivity calculations
are based are also included. In all cases, the variations repre-
sent one 0- limit in a normal distribution and will thus corre-
spond to 95 percent of possible cases (occurrences). This will
then also hold for the calculated ranges of conductivity variations.
The results are represented in terms of histograms in Figure 24.

At this point it is prudent to again remind the reader that the

data used for estimating variations in various soil parameters are
approximations. Thus, mean values and ranges of variation given
must be considered as highly uncertain and in some cases specula-
tive estimates. It is obviously highly desirable to obtain more
complete data in these areas. However, the average values and va-
riations for the conductivity of different types of materials pre-
sented here may serve as guide-lines when estimating average values
and variances (spreads). In the next section, these approximate
results will also be utilized in an investigation of the importance
of variations in actual soil parameters for thermal calculations
related to road designs subjected to frost.

267
 Free
ground Upper I

sub.-soil -

W r

0+0 0

+I+ 000 H +

SI I0 t0 I +
HS 9+
0 + C+

0, I+0

~Is I+- I++ 1+ 1+ 1+ L+i-

o~~N 40*** 02

C~0 0 .J I+ C.1+J

, C . .1 1
1+~0
1+ o+ o+ 10 1 !
I 03

I+ Ii+ I+- I+ Ii + I+- I+ I>

. ýj1 %J1j 0 0 0

0 0 0

ci.000 v0

268
* -A
 VA
FI

VA, c [:\ rq'F1 N([ ].\, bJU / rb•,L ' .'-•L ,.." ,.. v.v,' Y ',.'),/ <

0 0,5 10 15 20 2S 30 35

. 1' - Ir T*N. . V.

KNUýT GRUS FR.

• U-eH[OU. Fk,)

NATUPGRUS I __.
5",N / FILTERLAG FM.

f R.H
I Ft.

SANPO / iL1-- , -.....- _ _FI ... ...... .. ... .. I.... . . ..

F.R.
U~F.

S ILT

R,3 -n 24 f : ... .... .....' - 'I *d.

... --i
i.*l RE/SILT
C LAY/l~i
l41
FR.I-
uR
1~~

-c% Ode z. 14fi r . -

. L il
toL', ; V tt V), ," J;,cnl ;' t .. r to $ . E ( . *'.7 V)

v•L i.

269
, il
 4. EFFECTS OF CONDUCTIVITY VARIATIONS
ON TTIE1MAL CALCULATIONS

A. Introduction

Determining temperature distributions (fields) in freezing ground

has many applications. These include thermal analysis of road and
building foundations to ensure against frost damage, calculating
frost depth for buried water or sewage lines and determining tha-
wing depths for road or building foundations in permafrost regions.
Knowledge of thermal parameters of the soil is neccessary in all
these cases, regardless of the calculation method (mathematical
model) being used.

In principle, requirements oii the accuracy with which thermal soil

parameters are given will be different, depending on the accuracy
of the computational method (model). First of all, the thermal
soil parameters should be given with high enough accuracy to pre-
vent that the uncertainties in these parameters become the limi-
ting factors for the accuracy of the end result. On the other hand,
it does not make sense to require high accuracy for the thermal
parameters unless the computational method (model) can provide
results with a correspondingly high accuracy.

No quantitive basis for such a judgement exists today, but there

is reason to believe that uncertainties in the thermal parameters
do not constitute the limiting factor when using available approxi-
mate procedures for determining frost penetration in and under
road structures (Watzinger/Skave-Haug's method, (14, 15)), while
this often may be the case with modern numerical methods based on
the use of computers (16). In such canes it will thus be important
to try to reduce the uncertainties due to estimates of thermal pa-
rameters as much as possible, e.g. by performing the best possible
analysis of the soil materials which form the actual structure.

270 "U

A
Effects of variations in the soil parameters at hand or uncertain-
ties in estimated conductivities will in the following be trea-
ted for some' selected cases, limited to road foundations where
freezing of the ground is to be prevented.

B. Effects of conductivity variations

The error analysis performed in the previous sections showed

that the conductivity of materials in various layers of a road
structure can vary within relatively wide limits, due to varia-
tions in density, degree of saturation and mineral composition.
To determine the effects of such conductivity variations on ther-
mal calculations, the upper layer (pavement) thickness required
for prevention of freezing in the subsoil was determined (calcu-
lated) for different types of insulation (frost preventing)
layers. In these calculations, the conductivities of each layer
were varied within the limits established by the previous error
analysis. Variations in calculated thicknesses for the insula-
ting layers will thus form a basis for evaluating the effects of
conductivity uncertainties. The calculations were performed accor-
ding to the method given by Skaven-Haug, described in detail in
references (14) and (15).

The results of these calculations are shown in histogram form

by Figure 25. The average values and variations assumed for the
calculations are shown in the same figure. The calculations were
performed for a frost quantity of 30,000 h*CI) and three types of
frost preventing (frost resistant) layers.

For frost preventing layers consisting of gravel and bark, which

utilizes their cold accumulating ) properties, one finds that
the results are very sensitive to the conductivity of upper layers,

1) The unit for frost quantity is "Celsius degree-hours".

2) A strange expression, faithfully translated. (Translator's notes)

271
-rmi
while conductivities of sub-soil and the frost preventing layers
themselves have relatively small effect. The latter is particu-
larly true for bark layers. The large variations due to upper
layer conductivities indicates the importance of obtaining accu-
rate data for dry density and water content of materials used in
these layers when cold accumulating materials (layers) are used for
frost protection.

For structures utilizing insulation layers for frost protection

one finds that variations in sub-soil conductivity is the most
important parameter for determining insulation layer thickness.
However, tho approximations made may be responsible for this re-
sult. In the method used, a given temperature gradient in the
ground immediately below the insulation layers is used when deter-
mining the ground heat contribution to freezing resistance. The
heat loss from the sub-soil is obtained by multiplying this gradient
by the conductivity of the ground and the duration of the frost pe-
riod. Such a simplified approach can not be expected to give re-
liable results when the heat loss from the ground is as high as
in this case. Only a computation based on integration of the ther-
mal conduction equation, using a digital computer, can give the
required accuracy.

Thermal parameters for the sub-soil , except for quartz content,

are normally determined by geotechnical investigations in connec-
tion with stability calculations, etc. As mentioned previously,
the quartz content in fine grain soils can be estimated on the
basis of particle size distribution curves. The results obtain-
ed show the importance of utilizing such data also for the thermal
design.

For the calculation examples discussed here it was always assumed

that frost does not penetrate into the sub-soil. In cases where
this condition is not met, variations in un-frozen water content
will have a much larger effect on the calculations than the ex-
pected conductivity variation in the un-frozen sub-soil. To in-

272
 vestigato this case, calculations were performed for a structure
a
consisting of a 5 cm layer of insulation mater'ial (plates) and
* pavement (upper structure) identical to that assumed for earlier
* calculations. According to those calculations, this kind of struc-
* ture can withstand a frost quantity of about 30,000 hOC with no
frost penetrating the sub-soil,
L W A
P 30 000 hC
It,.,

I"
C

440'

:lot
is 0 64
ItOS )

It'ICI~~m4Wi4 -30 30 OS 2 4
P

NIORS. 'MA
I IiUU
"11 ,i S OM1091 02

It 1
VAMIa00
9.70 0
34it~s
A0 ol roL1 C. L

.,.A-Dj IRSSA AVU 6A POtxCisofAqtt4f ul

LONt
9h1'WNW 4I

*s oil 'A 4 J
4 4..~-t 0.0* ItOJ 0 & I'iIm4ij

AVO SLS -. I O NW mtývl

ti0ld VMIJN.II I I INAL

WVo

kp~v4 r.4i .' M" 1i...1

;A' !U* 0,) titleL

LSN.vai IuIM fro

i uil lild 1?A
:I At o~A 4.1 O OI AI
av on Sae Omrhd

IJAOCRM273'
 If the structure is subjected to a frost quantity of 40,000 h"C,
the sub-soil will consequently freeze to a certain depth, depending
on how large a portion of the available water actually freezes.
Figure 26 shows calculated values for the thickness of frozen 1
sub-soil, as a function of percentage un-frozen water in the sub-
soil, for this case. It is here assumed that all water "available
for freezing" has frozen at 0C.

IlOl60 . . . . . til

a.10

0 10. ...... . .. . . .-

" 0
O 10, ~~~......... 1•0
...... . ..
* kflQEy VANN V06.i
• • 1" .h
1 41 %U'1

Oro ea 1.0 4Ii ' l l'

Lia*. j I . I, i t t V.,

Similar calculations could be performed for structures using frost

prevention layers consisting of bark or gravel. For such cases
one can expect an even higher sensitivity to variations in the un-
frozen water content. However, like in the case of frost preven-
tion by means of an insulation layer2), the results will be so un-
certain that estimates of this type really have little value.

1) The original actually says "un-frozen", which appears to be

a Itypo",
2) It is not clear what kind of material this insulation layer is
assumed to consist of. Presumably, it has different (better
insulation?) properties than bark or gravel (Translator's notes).

274
:
For example, the effect of the temperature distribution in the
sub-soil on the un-frozen water content is uncertain. As men-
tioned befor% estimates of the contribution from (stored) soil
heat is another uncertain quantity. However, calculations of
the type indicated here should be performed as soon as experimen-
tal data are reduced and the computer program being developed by
the Institute for Cold Technology becomes operable, in order to
clarify the effects of under-designed frost prevention layers.

C. Conclusion

The previous analysis of effects of variations in pertinent para-

meters for road construction materials and sub-soil has demonstra-
ted that the accuracy of thermal analysis largely depends on how
accurately one can determine relevant material parameters for the
different layers. For the upper layers (pavement), determination
of water content is particularly important for the calculation ac-
curacy.

It is also apparent that the ratio of un-frozen water is an impor-

tant parameter in cases where the frost can penetrate into the sub-
soil, Such cases can be adequately treated only by means of ad-
vanced programs and modern digital computers. In general, this
also seems to be true for thermal analysis of structures in which
the frost prevention layers consists of insulation materials. The
normal method for finding the contrbution from heat (stored) in
the ground can not be considered satisfactory for this case.

275
 REFERENCES, CHAPTER VI.

I. A. Gemant: The Thermal Conductivity of Soils. Journal of

Applied Physics. 21, 1950, pp. 750-752.
2. M.W. Makowski and K. Mochlinski: An Evaluation of Two Rapid
Methods of Assessing the Thermal Resistivity of
Soil Proceedings. Institution of Electrical
Engineers. 103, A, (11), 1956, pp. 543 - 470.
3. L. Alberts et al: The Influence of Low Moisture Content on
the Conductivities of a Granular Substance. Brit.
J. Appl. Phys. 17, 1966, pp. 951-955.
4. D.A. de Vriae: Hot warmtgeleidingsvermogen van grond. ,led.
Landbouwhogeschool. Wageningen, 52, 1952.
5. M.S. Kersten. Thermal Properties of Soils. Univ. of Minnesota,
Eng. Experiment-station, Bull 28, June 1949.
6. H. Sveian: Unders~kelae av kvartsinnhold i lsmasser ved hjelp av
MTA. Institutt for Kjoleteknikk. NTH, Trondheim 1973.
7. R. Selmer-Olsenl En regional unders~kelse av norske kvartaere
lelrers finfraksjon basert pa DTA. Rapport sendt til
. .vNTNF. Geologisk Institutt, NTH Trondheim 1961.
S. .1 van Rooyen and H.F. Winterkorn: Structural and Textural
Influences on Thermal Conductivity of Soils.
Highway Res. Board Proceedings, Washington D.C. 1959,
pp. 576-621.
9. S. Skaven-haug. Romforhold i jordmaterialer. 'feddelelser fra
dot norske myrselskap, 70, (4), 1972, pp. 89-102.
10. J.V. Thues Redusert fwndamentdybde, - termiske problemer. Del
I. Institutt for husbyggingsteknikk. Trondheim, NTH,
p. 82.
11. W.A. Sinclair at ala Soil Thermal Resistivity. Typical Field
Values and Calculating Formulas, part IV. In: Soil
Thermal Characteristics in Relation to Underground
Power Cables. AIEE Transaction paper no. 60 - 785,
pp. 71-94.
12. S. Angen: Analyse av vanninnhold i vegkonstruksjoner. SINTEF
arbeidanotat. Trondheim 1973, p. 11.
13. A. Knutson and R. Saoterndal: Teleregistering 1967/68. Veg-
* laboratoriet. Intern rapport. Oslo 1969, p. 8.

276
14. S. Skaven-Haug: Frostfundamenters dimensjonering. Frysevarme
og jordvarme. Frost i jord, (3) 1911, pp. 9-27.
15. A.F. Knutson: Theory and Experience Regarding Frost Penetration
and Frost Heaving, Symp. on Frost Action on Roads, I,
Paris 1973, pp. 223-233.
16. A.K. Flemings Applications of a Computer Program to Freezing
Processes. Proceedings of the XII International
Congress of Refrigeration, Washington, D.C. 1971.

277 I
I
 THERMAL CONDUCTIVITY OF SOILS

SUMMARY

The aim of this investigation has been to create a

mathematical model for calculating thermal conductivity
of soils with ordinary soil parameters as input data.
One part of this work has been devoted to literature
studies tn hf.at-transfor mechanisms in moist materials.
These studies have made it possible to give bounds for
Fig. 23, I the different domains where the various mechanisms have
an appreciable influence on the total heat transfer.

With important exceptions such as freezing in frost

susceptible soils and convective heat transfer in coarse-

grained crushed rock, the heat transfer in road-building --

and subgrade materials can be treated as a heat conduction

problem.

Under these conditions, the thermal conductivity is

determined by three sets of parameters -- the volume
fraction of the components, their conductivities and the
microgeometry of the systom.

Various theories are suggested where these sets of para-

meters are combined in models for calculating thermal
conductivity of composite materials. However, only methods
where the thermal conductivity is given between upper and
lower bounds, referring to extreme microstructures, can

*As already translated in the Swedish text.

278
zY~-
 be said to give exact results because of the problem of
incorporating the microgeometry of natural materials in
an analytical solution of the heat conduction problem.
Such bounds as those developed by Hashin and Shtrikman
Fig. 1, I1 (1962) shows that the conductivity is very sensitive for
variations in microgeometry when the component-conductivities
are widely different, while the sensitivity is small when
the component-conductivities are of the same order of
magnitude. It follows that the conductivity may be
calculated analytically if the component-conductivities
are of the same order, while only empirical methods can
be used under other circumstances.

Results from thermal conductivity measurements on soils

are analysed on the basis of such fundamental theories
with the aim of deriving empirical relations to be used
in a model for calculating thermal conductivity of soils.
The analysis is centered around the extensive measurements
published by Kersten (1949), in addition to measurements
made at the Division of Refrigeration Engineering (Univer-
sity of Trondheim).

From this analysis the problem of deriving a mathematical

model consists of three parts. The conductivities of dry
soils and saturated soils are handled as two extremes
between which the conductivity must be found at a given
degree of saturation. The relation between conductivity
EQ. 10, V and degree of saturation is reduced to a normalized or
dimensionless form.

279
EQ. 5, V On the basis of the experimental results, two empirical
EQ. 6,V relations between conductivity of dry soils and porosity
were derived, with different relations for crushed rocks
and natural soils.

For saturated soils, the influence of differences in

Fig. 18, V microgeometry is found to have negligible influence, and
the thermal conductivity can be calculated with acceptable
accuracy, from volumetric compositiun and component-
conductivities. The problem of determining the conductivity
of soil particles was solved from an examination of certain
Tab. II,I•1 elementary aspects of rock mineralogy and recently published
data on thermal conductivity of rock forming minerals.
EQ. 2, 1II This resulted in a model where the particle conductivity
is derived from the quartz content of the soil.

The dimensionless reprewentation of the conductivity in

relation to the degree of saturation is calculated as a
ratio of the difference between the given conductivity and
the conductivity of dry soil, to the difference between
EQ. 10, V conducziviites rif saturated and dry soil. This ratio is
named the Kersten number. From Kersten's experimental
results empirical relations were derived between this
Kmrsten number and the degree of saturation, two relations
Fig. 24, V for unfrozen soils; one for finegrained soils and one for
Fig. 25, V coarsegrained soils, and one relation for frozen soils.
In the latter case no basis for discriminating between
soils of different texture was found. However, the
unfrozen water content, which is related to texture and

S280 t

A
 temperature below freezing, was found to have an appre-
ciable influence on the thermal conductivity. This influence
is incorporated by a correction in the saturated conduc-
tivity, on the basis of volume fractions of unfrozen water.

Combined with computed values for dry and saturated

soils, these normalized relations make it possible to
calculate the conductivities at various degrees of
Tab. I11, V1 saturation of a soil where quartz content and dry density
Fig. 15a, VI is given. This complete mathematical model is represented
Fig. 15b, V1 as nomograms where quartz content, dry density and degree
of saturation are input values in addition to informations
of grain shape (crushed v.s. naturals), grainsize distrib-
ution (finegrained v.s. coarsegrained), and volume fraction
of unfrozen water.

As informatioruof all relevant soil parameters are

available only occasionally, an analysis is performed of
the relationship between ordinary available geotechnical
data and the fundamental parameters determining thermal
conductivity. From this analysis relations are derived
for indirect determination of the fundamental parameters.
Such indirect relations will lead to a reduction in the
accuracy of the calculation of the thermal conductivity,
which however has to be accepted when incomplete soil
data are to be used.

2
281 I•,
 An extreme situation is present when no relevant
soil data is available. Under such circumstances average
thermal conductivities for the given soil group must
be used. Principally, such averages may be derived on
the basis of statistics on the distribution of the
relevant soil parameters for different soil groups. Such
data are scarce. However, in conjunction with an analysis
of expected variations in thermal conductivity of different
soil groups, such distributions are sketched partly on
the basis of available data.

From these approximate distributions and an error

analysis of the suggested mathematical model, averages
and expected variations in the thermal conductivity of a
Tab. ZX, VI set of soils and pavement materials are calculated, These
Fig. 24, VI averages represent a method for estimating conductivities
at the lowest level of information. The estimations can
only be improved by making use of more relevant information
of the soil parameters.

The improvement in the accuracy in the calculation of

thermal conductivity at different levels of information
is discussed for some cases. It is concluded.that infor-
mation on grainshape (crushed v.s. natural) in addition to
dry density and degree of saturation will result in an
appreciable improvement in the estimation of the thermal
conductivity of pavement materials with relatively small
degrees of saturation. Information on quartz content will

282

-VAN
 give similar improvements when dry density and degree
of saturation is given in subgrade materials with high
degrees of saturation.

The influence of variations in thermal conductivities

on thermal analysis is examined for some road construc-
tions with different types of frost-resistant layers.
Variations in conductivities of materials in the pavement
Fig. 25, VI are found to have much influence when frost accumulating
layers, such as bark or moist sand are used in the
construction, while variations in the conductivity in the
subgrade are of greatest importance when insulating
materials are used.

The results which are published in this report will be

followed up with an extensive experimental investigation
on thermal conductivity of Norwegian soils at the
Division of Refrigeration Engineering (University of
Trondheim). This investigation will make it possible to
verify the suggested mathematical model, and may eventually
result in some improvements. These investigations will
also give valuable information on the accuracy and validity
of various experimental methods for determining thermal
conductivity of soils.

283

-
 POST-SCRIPT

This report presents a new mathematical model for determining

thermal conductivity of mineral-based soil materials. Conven-
tional soil data and quartz content are used as input parame-
ters. The model is developed on the basis of a litterature
study of thermal transport mechanisms in moist, porous materials
and theories for thermal conductivity in composite materials an
well as extensive experimental results obtained by M.S. Kersten.
An error analysis is performed, which shows that conductivity
values within error limits of from !10 to 20 percent can be ob-
tained form this model if all relevant soil data are given.

The model is primarily developed for thermal analysis of structures

placed on the ground and subject to frost effects. A computer pro-
gram for determining temperature distributions (fields) in such
structures is under development at the Institute for Cold Techno-
logy, NTH as a part of the NTNF project "Frost in the Ground".
This computer program will include the model as a sub-routine with
soil data as input parameters.

A major experimental study of thermal conductivities in Norwegian

soil materials, which has been initiated at the same Inetitute,
may contribute material for improvements in the model. However,
initial results from this study indicated that the model is largely
acceptable for Norwegian soils. Certain changes and improvements
may still be required.

This report is primarily devoted to studies of conductivity in

mineral-based soils. In addition, certain experimental values for
peat and bog are included. Lack of results from systematic measure-
ments on other important materials, such as soils containing hu-
mus, concrete and asphalt has prevented treatment of such materials.
However, the experimental study at the Institute for Cold Techno-
logy will hopefully correct for this deficiency.

284
In order for the model to be used also in cases where one lacks
information about materials used in the structure, it is neces-
sary to have relevant data on expected mean values and variations
of soil parameters for various types of materials. In this report,
such values have been used to determine expected variations in
conductivity for different types of soils. However, in some ca-
see these values had to be estimated, due to lack of information
about water content and density in sub-soils and upper layers of
road structures. A study initiated by the Institute for Cold
Technology, NTH, may supply the required corrections for these da-
ta. When such results become available, a new error analysis
should be performed for the model, using revised mean values and
variations for the soil parameters.

This analysis should include a sensitivity analysis, using the com-

puter program mentioned previously, for a selection of typical road
structures utilizing different types of insulating layers. During
this analysis, thermal properties of the different materials with-
in the structure should be systematically varied within the limits
given by the previously mentioned error analysis, in order to cla-
rify the effects of uncertainties in soil material data.

This study is primarily concerned with thermal conductivities in

soil materials. To generate a computer program for determining
temperature distributions in the ground one must also know speci-
fic heat, as well as latent heat exchanged during freezing and tha-
wing. A calorimeter for determining theue parameters is being built
at the Institute for Cold Technology. The studies to be conducted
with this unit will form an important supplement to the investiga-
tions reported on here.

These concluding remarks show that the work reported here is part
of an extensive project, which is by no means finished by comple-
tion of this report. As indicated, several on-going invesgations
may form the basis for necessary improvements in the result given

285
here. These results will be reported in detail when the project
"Frost in the Ground" is concluded at the end of this year (1975),
e.g. in reports issued by the "Group for Analysis of Frost in the
Ground", Institute for Cold Technology, NTH.

APPENDIX. PERIODIC TEMPERATURE VARIATION

IN A HOMOGENOUS CYLINDER

In a homogenous cylinder of infinite length the thermal conduction

equation can be expressed in the form

ae8 1 le
ma +

where e isthe temperature

Sis thermal diffusivityt )
r is the radius
t is time

For the case at hand, O~r, t) is to be determined when the outer

surface temperature varies as a sinusoid:

;,
e (R,t) - eo sin (Wt) 2

where R is the radius of the cylinder

S0o is the amplitude
w is the angular velocity (frequency)

To solve this problem one can use results from text-books such
as Caralaw and Jaeger (1) and mathematl.cal handbooks (2), as well
as general control theory (3).

1) The original uses both 3(alpha) and a for this parameter

(Translator's note).

286
 From control theory it is known that a system subjected to an
applied, stationary and periodic forced input will give a eta-
tionary response which is periodic and damped (attenuated).
For the problem at hand this implies that the temperature vari-
ation on the axis of the cylinder (r m 0) can be written (4):

e(o,t) - A'8o sin (Wt + 3

where A ia the damping (attenuation) ratio

0 is the phaseshift

Attenuation and phaseshift can be determiihed from the transfer

function of the system. This function expresses the relation be-
tween the applied force function and the response in the Laplace
transform sense. In the case at hand, the transfer function be-
tween outer surface temperature and on-axis temperature is to be
determined. This is done by taking the Laplace transform of the
heat conduction equation and solving the resulting ordinary dif-
ferential equation for the proper boundary conditions.

* Carslaw and Jaeger (5) present a solution to this problem. They

give the following expression for the desired transfer function

G(s) - •os
S(R, s) 1 0-_(R V-3-7a)

where 4 is the Laplace transformed temperature

a is the Laplace variable
a is thermal diffusivity
Io ( ) represents the modified Bessel function of
zero order

In the theory for frequency response, the attenuation ratio is gi-

ven by the magnitude of the transfer function for a a i w Ui a
a •F-7-, while the phaseshift is given by the argument (phase
angle) of the transfer function for the same value of the variable
s(3). 237
 A - IG(iw)1

-argjG(iw)j

For s = i w, the transfer function becomes a complex quantity with

real and imaginary parts
G(iw) *X + iy
, 6

The absolute value (magnitude) is given by /x2"+ rthe argu-

ment is arctan (y/x)

For s a i w, the variable of the modified Bessel function takes

the form
R /777/ - R(l+i)

since (1 + i)2• 2

This complex number can also be expressec as

in/4 , Rwa
Xwe X 8
i8

where X is the magnitude (absolutevalue) of the number.

In mathematical handbooks (6) one finds that a modified Bessel

function of order zero with a variable of this form can be ex-
pressed as a complex number where real and imaginary parts are
represented by the Kelvin functions berX and beiX:

berX + i beiX - il1/4

1o (Xe 9
Magnitude 1 )and argument for the transfer function can thus be

1) The original says "real part", which is wrong (Translator's

note).288 288
expressed in terms of these two Kelvin functions

A - IG(iw) a I [(beiX) 2+ (berX) 10

= LG(iw) - -arctan beiX 11
berX
/
/

This follows from the rule that the absolute value (magnitude)
of a complex number l/z, z = Xe i, equals I/X, while its argu-
ment is-

The functions berX and beiX are tabulated in mathematical hand-

books (7). Figure I shows phaseshift and attenuation as function
of X x R /V-7 , calculated from these tables.

This representation ( Oand A as functions of X) makes it possible

to calculate thermal diffusivity in a homogenous material formed
as a cylinder, by determining attenuation or phaseshift for the
periodic temperature variation from the outer surface to the cen-
ter line of the sample.

The same method can be used for finding attenuation and phaseshift
between two arbitrary radii within such a sample. For this case
the transfer function can be expressed

GC,2(a) "----
ýW-•le "o(ri_ •,M ,
(8) 42(s) IC(r2 ý712

This results in the following expressions for phaseshift and

attenuation:

289 *

S......,, ' . ...

..... ... ,, *.. .... " •- • J .• ? •. : - • .. • . :• - " ''[ •- ... . . ,u• !• ,.*
 t• beix 1 beix 2 13
"*• - arotan .. _- arotan -

1,2 berxI berx2

2 2
(beix) + (berx

* (beix) + (beix)

AI,24*~. . - - Z
, i'
wher'e x, r, =w'. X2.. =r=
I~

.............. I--,
-.-........... .-.--

*g
4c *i.,__................ . .,+++

; .'I. ......... --..

+ .Wll4I
, M. - P- 1 16,. t. • 4p 6.1 .1

F1 l "I
29,0.
41 y
, i fl.o.,•
rL Ie oo
'. '@l :4l ,1 ; I I !.I •+
r, , I , i"jI

BESTAVAILABLE COPY

29 0 t
 REFERENCES

1. H.S. Carslaw and J.C. Jaeger: Conduction of Heat in Solids.

Oxford at the Claredon Press, 1959, 2, ed.

2. M. Ambramowitz and I. Stegum. Ed: Handbook of Mathematical

Functions, Nat. Bur. of Standards, Appl. Math. Ser.
New York 1969.
3. J.G. Balchent REguleringsteknikk. BInd 1. Tapirs forlag
1971, 5, oppl.
4. J.G. Balchen, op. cit: pp. 71-72.

5. H.S. Carslaw, op. cit: pp. 326-327.

6. M. Ambrarnowitz, op. cit: p. 379.
7. M. Ambramowitz, op. cit.: p. 430.

291