Coop, M. R. & Atkinson, J. H. (1993). G&&-chnique 43, No.

1, 53-67

The mechanics of cemented carbonate sands

M. R. COOP* and J. H. ATKINSON*

The behaviour of artifkially cemented carbonate Le comportement d’un sable carbonat& artificielle-
sand was investigated in triaxial tests at confining ment cimentk a 6tt BtudiC 1 l’aide d’essais tri-
pressures of up to 9 MPa. The results show that an axiaux h pression de confinement pouvant
important effect of cementing is a reduction in spe- atteindre 9 MPa. Les rbultats obtenus montrent
cific volume resulting from the increase in fines que la cimentation se traduit par une diminution du
content. This influences both the stress-strain volume sp6cifique due P une augmentation du
behaviour and the peak strength at strains beyond nombre de particules fines. Ceci modifie le com-
those required to fracture the cement bonding. portement contrainte-d&formation ainsi que la r4
Comparisons between the behaviour of cemented sistance ultime pour des d&formations sup&ieures
and uncemented soils should, therefore, be carried P celles provoquant la rupture des liaisons du
out on samples with the same gradings. For ciment. L’ittude comparative du comportement de
cemented samples it is possible to identify a yield sols cimentb et de sols non-cimentb a done dB
curve outside the state boundary surface of the &e &ali&e sur des khantillons de m&me granulo-
uncemented soil. A framework for the behaviour m&rie. Pour des 6chantillons cimenti, une courbe
has been defined which depends on the relative de rupture situ&e en dehors de l’enveloppe limite
magnitudes of the confining pressure and cement d’un sol non-cimentb a Ct& mise en i?vidence. Une
bond strength. The behaviour of a natural cal- classiication de comportement a &t& klaborbe en
care&e agreed well with this framework which is fonction des amplitudes relatives de la pression de
also likely to be applicable to other cemented soils. confinement et de la r6sistance des liaisons du
ciment. Le comportement d’une calcarbnite natu-
relle est tout P fait en accord avec cette classi-
KEYWORDS: calcareoaa soils; compressibility; Iabor- fication qui apparait applicable B d’autres sols
atory teats; sands; shear strength. cimentCs.

INTRODUCTION cemented calcarenites. Although the latter tend to

When in 1982 the North Rankin platform was occur at greater depths, cementing coincides at
installed off North West Australia, unexpectedly least partly with deposition and is controlled by
low driving resistances indicated that the piles the chemistry of the current depositional environ-
might not provide the design shaft friction capac- ment. Consequently, test data obtained from
ity (King & Lodge, 1988). This prompted further naturally cemented carbonate sediments are fre-
investigation of the mechanical behaviour of car- quently highly scattered, and in order to examine
bonate soils (Jewell & Andrews, 1988; Jewel1 & the basic behaviour of calcarenites it is conve-
Khorshid, 1988). Most of this research has con- nient to test artificially cemented carbonate sands.
centrated on the bioclastic sediments formed by A comprehensive review of the engineering
the deposition of skeletal bodies which predomi- geology of carbonate soils and rocks has been
nate in the continental shelf environments and are given by Fookes (1988); surveys of the current
commonly cemented with calcium carbonate knowledge of their mechanical behaviour have
deposited at the grain contracts and in the void been made by Semple (1988) and Poulos (1988).
spaces. Recent research into the strength and stress-
In common with naturally cemented carbonate strain behaviour of carbonate soils has concen-
sediments, those at North Rankin (predomin- trated on triaxial tests on uncemented sands
antly calcarenites) are highly variable in grading (Coop, 1990), on sands artificially cemented with
and degree of cementing, ranging from loose, Portland cement (Allman & Poulos, 1988) and
completely uncemented sediments to strongly gypsum plaster (Boey & Carter 1988) and on
naturally cemented calcarenites (Carter, Johnston,
Discussion on this Paper closes 1 July 1993; for further Fahey, Chapman, Novello & Kaggwa, 1988;
details see p. ii. Airey 8c Fahey, 1991). Following Leroueil &
* The City University, London. Vaughan (1990), the influence of cementing on
53

Downloaded by [ University Of Western Australia] on [17/12/22]. Copyright © ICE Publishing, all rights reserved.
 54 COOP AND ATKINSON

soil behaviour can be understood with reference in yield stress as the cement content increased.
to the behaviour of the same material The bonding resulted in increased strength and
uncemented or destructured. For carbonate stiffness, both of which increased with increasing
sands, however, destructuring by mechanical cement content. Boey & Carter (1988) have also
remoulding leads to particle breakage, and so the compared the behaviour of uncemented and arti-
approach used here was to compare the behav- ficially cemented carbonate sands, using gypsum
iour of artificially cemented material with that of plaster to create the bonding. For a given plaster
the same material uncemented. Data from tests content they obtained a well-defined peak during
on a naturally cemented calcarenite were then shearing at low confining stresses which resulted
examined in the context of the behaviour from cohesion in the soil. This effect diminished
observed for the artificially cemented sand. with increasing confining pressures up to a
The essential features of the behaviour of maximum of 500 kPa; it was not determined
uncemented carbonate sands are relatively high whether or not the behaviour would return to
friction angles and compressibilities (Semple, that of the uncemented soil at still higher
1988; Coop, 1990). The angularity of the particles stresses.
gives rise to high voids ratios, while their delicate In both of these studies direct comparisons
nature causes large volume changes during com- between the behaviour of cemented and
pression and shearing. Both Semple (1988) and uncemented samples are complicated by two
Coop (1990) have emphasized that the basic factors. First, the bonded soils have lower water
behaviour of uncemented carbonate sediments is contents than the uncemented soils due to the
comparable to that of other granular materials at void spaces being filled by cement. Second, the
similar high voids ratios. presence of the cement may have other effects on
Coop (1990) provided a framework for the fun- the underlying behaviour of the soil, for example
damental mechanics of uncemented carbonate by changing its grading. In the work described in
sands. He showed that on compression to high this Paper, the separate effects of cementing and
stresses there was a normal compression line, and grading changes were examined by conducting
the large plastic volumetric strains that gave rise tests on a carbonate sand artificially cemented
to this were a result of particle breakage. During using gypsum casting plaster and on uncemented
shearing the large plastic volumetric strains that mixtures of the same sand and gypsum powder.
again resulted from particle degradation were
found to continue to high axial strains, making
critical states difficult to define. Coop (1990) MATERIALS TESTED
showed that critical state lines do exist for car- The material used in the tests described in this
bonate soils. For the particular soil he tested Paper was a carbonate sand from Dog’s Bay in
these were straight in deviatoric stress-mean Ireland, used by Evans (1987), Houlsby, Evans &
normal effective stress (q’:p’) space, and lay parallel Sweeney (1988) and Coop (1990) for research into
to the normal compression line in specific the behaviour of piles in carbonate soils and
volume-log mean effective stress (u : In p’) space. rocks. It is a biogenic carbonate sand consisting
Vaughan, Maccarini & Mokhtar (1988) and predominantly of foraminifera and mollusc shells.
Leroueil & Vaughan (1990) have developed a For the work described here it was mixed with
framework for the behaviour of bonded soils. gypsum plaster (3.33 parts sand to 1 part plaster
They found that these materials can reach states by weight). The untreated Dog’s Bay sand is a
outside the normal compression line for the poorly graded medium sand: as all the plaster
equivalent uncemented or destructured soil, but passed the 63 pm sieve, the effect of its addition
continued loading beyond yield results in rela- was to increase the fines content from 1% to
tively rapid compression as the state moves back 24%.
towards the normal compression line for The tests were carried out on the sand-plaster
destructured soil. mixture either uncemented or cemented; the
The behaviour of the naturally cemented cal- cemented samples were made by adding distilled
carenites from the North West shelf of Australia water (1 part water to 2.47 parts sand-plaster
has been investigated by Carter et al. (1988) and mixture by weight). Evans (1987) has demon-
Airey & Fahey (1991). They found that under iso- strated gypsum plaster to be the most suitable
tropic compression well-defined yield points artificial cementing agent for this soil: for a given
could be identified which were attributable to the ratio of cement matrix to soil particles, similar
breakage of the cemented bonds. Allman & unconfined compressive strengths to those of
Poulos (1988) used Portland cement to create natural calcarenites were obtained. Thin section
artificial bonding, and found for isotropic loading analyses by Price (1988) showed that for natural
that the principal effect was a reduction in the calcarenites light cementing appeared only
initial specific volume of the soil and an increase around the fringes of the particles and at particle

Downloaded by [ University Of Western Australia] on [17/12/22]. Copyright © ICE Publishing, all rights reserved.
 MECHANICS OF CEMENTED CARBONATE SANDS 55

contacts. Heavier cementing causes the void pressure stress path triaxial cell. Details of the
spaces to be filled with a cement matrix. Also initial states and loading paths for each test are
through inspection of thin sections, Huang and given in Table 1.
Airey (1991) have shown that the use of plaster as Cemented samples of either sand and gypsum
an artificial cement produces a similar type of or solid gypsum were prepared by tamping mix-
structure to that which occurs naturally. This has tures of carbonate sand, plaster and water into
been confirmed by the Authors’ thin sections of greased split moulds, in which they were allowed
both the natural and artificial soils. to cure for 24 h at 20°C. After removal from the
The cementing process in the natural soils mould the samples were allowed to cure for a
deposits calcium carbonate out of the pore fluid further week, or until they had reached constant
at grain contacts. This coincides with sedimenta- weight. Unconfined compression tests were gener-
tion. The procedure chosen for mixing sand with ally carried out on dry samples, but otherwise all
wet plaster and allowing it to cure at atmospheric cemented samples were saturated by immersion
pressure is a reasonable approximation to the in a bath of silicon oil under vacuum. Back pres-
natural process, which is not one of lithification sures of up to 500 kPa ensured B values
through deep burial, but rather takes place at low (Skempton, 1954) of at least 0.95. Uncemented
confining stresses. samples of sand with either gypsum or calcium
With gypsum plaster as the cement, the pres- carbonate fines were prepared by de-airing the
ence of water as the pore fluid reduces the mixture by immersion in silicon oil under
bonding strength. To maintain a strength consis- vacuum, followed by pluviation through silicon
tent in dry and saturated samples, a light silicon oil according to procedures described by Coop
oil of similar viscosity to water was used as the (1990). Samples of Rankin calcarenite were gener-
pore fluid for both cemented and uncemented ally cut to 166 mm long from the 83 mm dia. core
saturated samples. and then saturated by immersion in a water bath
Gypsum plaster is an hydrated calcium sul- under vacuum.
phate: its mineralogy and chemistry therefore For the artificially cemented samples and the
differ from those of natural cemented carbonate uncemented sand-gypsum mixtures, the initial
sands for which both the soil particles and the specific volumes were calculated from the initial
cementing material are predominantly calcium dry weights of material used and the dimensions
carbonate. To distinguish effects due to the of the samples. Initial moisture contents of the
change in mineralogy on addition of the gypsum natural calcarenites were measured from off-cuts
from those due to the increase in fines content, saturated in the same way as the sample. These
two further types of test were conducted. Two initial values were then averaged with those after
cemented samples of pure gypsum plaster with no testing, accounting for the volume strains mea-
sand were tested, then a series of three tests were sured during the test.
made on samples of Dog’s Bay sand with calcium Most of the samples were initially isotropically
carbonate rather than gypsum fines added, but in compressed (Table I), although some were ini-
the same proportion. These fines were created by tially one-dimensionally (K,) compressed using
crushing the sand mechanically until it passed a the cell pressure controller to maintain zero
63 pm sieve. radial strain. They were sheared in triaxial com-
Tests were carried out on samples of natural pression either drained or undrained. The drained
calcarenite cored from the foundations of the tests were carried out with either constant effec-
North Rankin platform. Descriptions of the local tive confining pressure or constant mean effective
geology and the drilling and coring procedures stress.
have been given by Apthorpe, Garstone & Turner The test data reported are the deviator stress
(1988) and Renfrey, Waterton & Van Goudoever q’ = (0: - a:), the mean stress p’ = (1/3)(a: + 2~:)
(1988). To minimize variation of the natural and the specific volume u = 1 + e, where e is the
samples, they were all taken over a depth interval voids ratio. Strains are given as the volumetric
of only 5 m from one borehole (B2-6). strain E, = (E, + 2q) and the shear strain E, =
(2/3)(&, - E,). In all cases these are natural strains
E, calculated from the measured linear strains E,
EQUIPMENT AND EXPERIMENTAL
E” = - ln(1 -El) (1)
PROCEDURES
The computer-controlled triaxial cells used
were the same as those described by Coop (1990).
Tests with confining pressures of up to 700 kPa BEHAVIOUR OF THE UNCEMENTED SOIL
were carried out in a standard hydraulic triaxial Isotropic and one-dimensional compression
cell (Bishop & Wesley, 1975); tests at higher con- The isotropic compression behaviour of
fining pressures were carried out in a new high uncemented samples of the sand-plaster mixture

Downloaded by [ University Of Western Australia] on [17/12/22]. Copyright © ICE Publishing, all rights reserved.
 56 COOP AND ATKINSON

Table 1. Summary of triaxial tests

Test* Sample Isotropic p’ after Type of Depth of

diameter : mm (I) or K, consolidation: kPa shear? sample: m
compression
UNC2 60 I 300 D
UNC6 60 I 450 D
UNC7 60 I 300 U
UNC8 60 I 140 D
UNCA 60 I 5500 P
UNCB 60 I 3000 P
UNCC 60 I 8500
UNCD 60 I 1700 P
UNCE 60 K0 8360 P
UNCF 60 43 11 300 P
UNCG 60 K0 2740 U
UNCH 60 K0 2860 P
UNCI 60 Kll 9720 P
UNCJ 60 K0 790 U
CEMl 38 I 300 U
CEM2 38 I 150 D
CEM3 38 I 450 D
CEM4 38 I 300 D
CEMS 38 I 550 D
CEMA 38 I 2520
CEMB 38 I 2112
CEMC 38 I 5500 P
CEMD 38 I 1500 P
CEME 38 I 3000 P
CEMF 38 I 900 P
CEMG 38 I 1500 U
CEMH 60 K0 7180 P
CEMI 60 K0 12 120 P
RAN1 83 I 7200 P 137.1
RAN2 83 I 1670 - 135.6
RAN3 83 I 2280 P 134.9
RAN4 83 I 2200 P 134.6
RAN5 83 I 1000 P 134.4
RAN6 38 I 450 P 132.6
FIN1 60 I 5500 P
FIN2 38 I 250 P
FIN3 60 I 1500 P
PLAl 60 I 8800 P
PLA2 60 I 8800 P
* CEM = artificially cemented, UNC = uncemented sand-plaster mixture, RAN = Rankin cal-
carenite, FIN = uncemented with calcium carbonate fines, PLA = gypsum plaster only.
t D = drained, U = undrained, P = constant p’ drained.

is shown in Fig. 1, which plots the specific quence of its uniform grading. The isotropic
volumes to a logarithmic scale (Butterfield, 1979). normal compression line for the mixed samples
Fig. 1 also shows the compression line given by shown in Fig. 1 is given by
Coop (1990) for Dog’s Bay sand without added
lnv = N* - L*lnp’ (2)
gypsum fines, together with an average line for
samples of the untreated soil set up similarly. where 1* = 0.115 and N* = 1.54. These are
The data shown in Fig. 1 indicate that the smaller than the values for untreated Dog’s Bay
addition of gypsum fines resulted in samples of sand, confirming the need for tests on
considerably smaller specific volume. This con- uncemented sand and gypsum mixtures with the
firms the observation that the high initial specific correct gradings if comparisons are to be made
volume of Dog’s Bay sand is partly a conse- with tests on cemented samples. The data shown

Downloaded by [ University Of Western Australia] on [17/12/22]. Copyright © ICE Publishing, all rights reserved.
 MECHANICS OF CEMENTED CARBONATE SANDS 51

compression line, but while the effect of adding

the gypsum fines was to change the value of A$
and 1*, the K, value of 0.51 was the same as that
measured by Coop (1990) for untreated Dog’s
Bay sand.

Failure at critical states

Critical states for uncemented samples with
both gypsum and calcium carbonate fines are
shown in Fig. 2(a). All the data points fall close to
NCL with
a unique straight line given by q’ = Mp’, with a
value M = 1.65 corresponding to a critical state
friction angle &, = 40” which is the same as
given by Coop (1990) for the soil without added
fmes. This demonstrates that neither the presence
- Gypsum fines of the gypsum rather than calcium carbonate
----- Calcium carbonate fines fines nor the use of the silicon oil pore fluid had a
---- Average line Dog’s Bay sand without fines
significant effect on the soil friction.
Figure 2(b) shows the critical states in In II: In p’
!- space, and compares them with data from tests
IO0 loo0 10 000 on cemented samples discussed below. Again, the
p’: kPa data for samples with gypsum and calcium car-
bonate fines are indistinguishable, and a critical
Fig. 1. Isotropic compression of samples with and
without fines state line can be identified parallel to the isotropic
and K, compression lines, given by
lno = I-* - 1*lnp’ (3)
in Fig. 1 for samples to which calcium carbonate
fines had been added are very close to those with where r* = 1.41. Fig. 2(b) also shows that the
gypsum fines, indicating that the change in critical state lines for the treated and untreated
grading rather than in mineral composition influ- soils are closer to each other than are the iso-
ences the behaviour of the soil. Other data for tropic compression lines in Fig 1; one effect of
one-dimensional compression defined a K, adding the fines has been to reduce the separation
normal compression line parallel to the isotropic of the normal compression and critical state lines.

20 000 3

z
=
)

Uncemented (gypsum fines)

Uncemented (calcium carbonate

- CSL Dog’s Bay sand (no fines)

----- CSL cemented
a 1
0 10000 20 000 100 1000 loo00
p’: kPa P’: kPa
(a) (b)

Fig. 2. Ultimate failure of uncemented and cemented samples

Downloaded by [ University Of Western Australia] on [17/12/22]. Copyright © ICE Publishing, all rights reserved.
 58 COOP AND ATKINSON

3 soil is particle crushing. In Fig. 4(a) there do not

- Gypsum fines
---- Calcium carbonate fines
appear to be separate yield points for the break-
ing of the cemented bonds and the onset of par-
State boundary
ticle crushing. Coop (1990) showed that particle
2 with no fines breakage of the uncemented soil was a gradual
process, with increasing stress p’ giving rise to a
8
j&
curved first loading line and a less distinct yield.
D For the cemented soil it appears that the cement
matrix initially bears the confining stress, prevent-
ing particle breakage. When the matrix fractures
load is shed onto the individual particles, and so
the onset of particle crushing coincides with yield-
ing of the cement bonds.
The higher yield stress of the cemented soil
Normal than of the uncemented material is also apparent
comoression
from Fig. 4(a). As Leroueil & Vaughan (1990)
have observed for a variety of weak rocks, the
Fig. 3. State paths for uncemented samples cementing allows the soil to reach states outside
the normal compression line of the destructured
(i.e. uncemented) soil. Continued loading beyond
Behaviour during shear yield causes the soil state to converge with the
Figure 3 shows the state paths for drained and normal compression line for destructured soil, but
undrained triaxial compression tests carried out even at the highest stresses reached the specific
on uncemented samples with either gypsum or volume of the initially cemented soil is signifi-
calcium carbonate fines. The data have been nor- cantly higher than that of the uncemented soil.
malized with respect to the equivalent critical Vaughan et al. (1988) have suggested that even
state pressure pi, given by after the cement matrix has yielded, considerable
pi, = exp [(r* - lnu)/l*] additional compression would be required to
establish the same fabric as for the uncemented
Normalizing with respect to &, is particularly soil.
useful for the cemented samples for which a Data from one-dimensional compression tests
normal compression line is not easily defined. To are shown in Fig. 4(b) and (c). Again the soil
avoid data scatter arising from the difficulty in reaches states clearly outside the normal com-
measuring specific volumes accurately, the state pression line of the uncemented soil. Initially the
of each sample after primary consolidation response is extremely stiff, and in q’ : p’ space the
beyond yield was assumed to lie on the appropri- stress path moves quickly to states outside the
ate normal compression line. critical state line of the uncemented soil. The
The state boundary surface shown in Fig. 3 is stress ratios reached are similar to those mea-
of similar shape to that found by Coop (1990) for sured for the peak states of standard drained tri-
the untreated Dog’s Bay sand. It has a well- axial tests at similar stress levels. Addis (1987) has
defined peak in the state boundary surface which presented similar data for the one-dimensional
Chandler (1985) predicted for soils with deform- compression of chalk. Yield occurred at a small
able particles. Below the state boundary surface axial strain of about 0.2%, accompanied by the
the directions of the normalized state paths stress path moving back towards the K, com-
depend on the type of test. They are similar to pression path of the uncemented soil. Again,
those for soil without added fines. The state path however, the behaviour of initially cemented soil
for the undrained test (UNC7) shows the charac- does not return to that of the uncemented soil
teristic ‘8’ shape observed by Coop (1990). within the stresses reached during this test.

BEHAVIOUR OF THE ARTIFICIALLY Failure at critical states

CEMENTED SOIL The cemented samples did not always reach such
Isotropic and one-dimensional compression well-defined critical states as the uncemented
The data for isotropic compression of cemented samples, and the end of the test states are shown
soils shown in Fig. 4(a) have more distinct yield in Fig. 2. The data all fall close to a unique criti-
points than those for the uncemented sand. It has cal state line with gradient M = 1.50 (4:, = 37”).
been confirmed by observation of thin sections Fig. 2(a) shows the critical states of the samples
that, as for the uncemented soil, the principal made from cemented gypsum plaster only; these
mechanism of plastic volumetric strain for this fall close to the critical state line for the cemented

Downloaded by [ University Of Western Australia] on [17/12/22]. Copyright © ICE Publishing, all rights reserved.
 MECHANICS OF CEMENTED CARBONATE SANDS 59

Uncemented
‘, (from Fig. 1)

I I 1 I
1000 10000 1000 10000
p’: kPa p’: kPa
(b)

5000 10 000 15ooa

p’: kPa
(c)

Fig. 4. Compression of cemented samples: (a) isotropic; (b) one-dimensional; (c) one-
dimensional

samples. The lower friction angle for the there are no clear differences between the two sets
cemented samples compared with the value of data. A single critical state line has been drawn
4’ = 40” for uncemented samples with or without to represent both cemented and uncemented
gypsum fmes may result from cement adhering to samples of sand and plaster.
the sand particles after yield. Thus, while the min-
eralogy of the gypsum tines does not affect the
volumetric behaviour of the soil, it reduces the Behaviour during shear
friction angle by about 3”, but only when the soil For the high-pressure tests, the stress paths
has been cemented. shown in Fig. 5(c) do not have discernible peaks,
Figure 2(b) shows the end of test states in as the cemented bonds yielded during the initial
In D: In p’ space, and compares them with those compression stage of the test and appeared to
from the tests on uncemented samples of sand have little further influence during shearing. At
with gypsum or calcium carbonate fines. At low lower stresses uncemented carbonate soils reach
stress levels some of the cemented samples had peak states above the critical state line as a result
not yet reached stable states when the tests were of dilation, as shown by Coop (1990), but cement-
ended; the two samples at the lowest confining ing results in even higher peak shear stresses.
pressures also failed on well-defined planes, Cementing also gives the soil a substantial uncon-
making these data points less reliable. Otherwise fined compressive strength, a component of

Downloaded by [ University Of Western Australia] on [17/12/22]. Copyright © ICE Publishing, all rights reserved.
 60 COOP AND ATKINSON

Fig. 2

800 1600 800 1600

p’: kPa p’: kPa
(a) 63
20 000

16000

9
b

Figs 5(a) and

p’: kPa
w

Fig. 5. Stress paths for tests OIIcemented samples; prefix for all tests is CEM

strength that may be regarded as cohesive. This a mean line for seven tests. These, and the lowest
acts in addition to that resulting from friction and pressure triaxial tests CEM2 and CEM4, reached
Vaughan et al. (1988) observed that for cemented peak states at very low strains, followed by rapid
soils the peak states do not correspond to the strain softening as shear planes formed. As the
maximum rates of dilation, as they do for confining stress increases, the ratio of the peak
uncemented soils. At low confining stresses most strength to the critical state strength decreases,
of the peak strength results from the cohesive and the axial strain at the peak increases substan-
cementing component but, as discussed above, at tially, corresponding to a transition from shear
higher stresses the soil becomes purely frictional. plane failure to a barrelling mode.
This change in shearing behaviour is illustrated The high strains at the peak states for tests at
particularly well by the stress paths for the intermediate pressures (3, 4, 5, F and D) clearly
drained tests shown in Fig. 5(a). The drained could not be sustained by an intact cemented
unconfined compression tests are represented by fabric; closer examination of the stress-strain

Downloaded by [ University Of Western Australia] on [17/12/22]. Copyright © ICE Publishing, all rights reserved.
 MECHANICS OF CEMENTED CARBONATE SANDS 61

2000 2m
Critical /
r state ,
line ,/
/
a Yield points
/I

I
a” 0
4 Yield points bf loo0
b r’
. CEM3 / 8.8 ‘1 l
0 CEMD
/I \
I I
1 2 /I \

0
0
//
1000
I 1, 2000
p’: kPa

Fig. 7. Stress states at yield

conventional drained test (CEM3) yield is better

identified from the discontinuity in the stress-
strain curve. In this test elastic volumetric strains
. occurred before yield because of the increase in p’
. and, while yield again corresponds to the onset of
1.5L
plastic volumetric strains, the overall rate of com-
Fig. 6. Identilkation of yield points for cemented pression decreases. Yield points for tests on
samples cemented samples are shown in Fig. 7.
Coop (1990) showed that the stress-strain
behaviour of uncemented carbonate sand
data reveals clear breaks at axial strains of 0.3- depends principally on the current mean normal
0.7%, which represent the onset of yielding of the effective stress, the type of shearing, and the state
cemented structure. For these tests the cementing of the soil relative to the critical state line (i.e. on
may continue to have some influence on the the current value of p’/p:,). Normalization of
stress-strain behaviour even after initial yield of stress-strain curves using axes q’/p’ and VE,
the bonds, and the transition from cemented to resulted in unique curves for a given test type and
uncemented modes of failure is not so well initial state. For weak rocks, Leroueil & Vaughan
defined. (1990) have emphasized that the stress-strain
Coop (1990) showed that uncemented samples behaviour will depend on the position of the
were much stiffer in undrained loading than in initial state of the soil relative to the yield locus of
drained loading. The same feature is seen for the the bonding, as shown in Fig. 8.
cemented soil shown in Fig. 5(b), where the The first class of behaviour occurs where the
undrained test CEMG shows a higher peak state sample has passed its yield point during isotropic
than the constant p’ test CEMF, and at very compression; subsequent shearing should pro-
much lower strains. From these two tests it duce behaviour similar to that of an initially
appears that peak state is affected by the direc- uncemented soil, with no yield point. The second
tion of the stress path and the drainage condi- class occurs at intermediate confining stresses so
tions as well as the confining pressure. that although the cemented bonds are intact at
For tests that did not show yield during com- the start of the test, they yield during shearing and
pression, the initial stress-strain behaviour can be the peak state is governed by the frictional behav-
taken as elastic and yield points identified which iour of the now uncemented soil. The stress-
may correspond to the start of breakage of the strain curve for this type of test might be expected
cemented bonds. Typical stress-strain and volu- to show a distinct yield point after an initial
metric strain curves corresponding to the early elastic section. For the third class the sample is
stages of shearing of two drained tests are shown sheared at low confining stresses relative to the
in Fig. 6: yield points have been marked. For the strength of the cementing. A peak state occurs at
constant p’ test (CEMD) yield is readily identified low strains well outside the state boundary
by the onset of plastic volumetric strains. For the surface of the uncemented soil.

Downloaded by [ University Of Western Australia] on [17/12/22]. Copyright © ICE Publishing, all rights reserved.
 62 COOP AND ATKINSON

Fig. 9(c) shows behaviour similar to the ideal

behaviour shown in Fig. 8, except that first yield
of the cemented bonds is seen slightly before a
well-defined peak is reached. The peak probably
represents the complete breakdown of the cement
matrix; other studies (e.g. Airey & Fahey (1991))
have interpreted yield as a two-stage process.
Figure 9(b) shows the second class of stress-
strain response which, like the third, gives much
higher stiffnesses for the cemented than for the
uncemented soil. Yield of the cemented bonds for
test CEM3 is again followed by a slight peak in
the stress-strain curve which must result from
some continued influence of the cement at rela-
tively high strains.
Normalized stress paths for artificially
cemented samples are shown in Fig. 10, using p:,
as the normalizing parameter as discussed above.
The differences in the friction angles of the
cemented and uncemented soils have again been
accounted for by normalizing q’ with respect to
M. The paths are split into two groups: Fig. lo(a)
shows paths for samples that started from pre-
Strain - yield states; Fig. 10(b) shows paths for samples
that had yielded during isotropic compression.
Fig. 8. Idealised behaviour of cemented soils: (a) stress Paths that start at values of p’/pL, higher than - 3
paths; (b) stressstrain behaviour correspond to initial states outside the normal
compression line of the uncemented sand-gypsum
mixture. Even samples that started at lower
Figure 9 shows stress-strain and volumetric values of p’/p& reached states well outside the
strain data from tests on cemented soil, and data state boundary surface of the uncemented soil
from similar tests on samples of uncemented sand before the cemented bonds yielded during shear-
and gypsum plaster mixture. These correspond to ing.
the three classes of behaviour illustrated in Fig. 8. For tests CEMl and CEM2 carried out at the
In each case the data were obtained from simi- lowest confining pressures, the presence of the
lar types of test carried out at similar values of cemented bonds allowed the soil to reach
p’/p,,‘. Because of the small difference in friction unstable peak states outside the state boundary
angles between cemented and uncemented surface of the uncemented soil followed by rapid
samples, the q’/p’ axes have been normalized with strain softening. The other tests shown in Fig.
respect to M. 10(a) all showed yield of the cemented bonds
For uncemented soil, normally consolidated during shearing. Before yield each of the tests fol-
samples should give a unique normalized stress- lowed a straight path, indicating elastic behav-
strain curve. For the three tests shown in Fig. iour. The yield points correspond to the onset of
9(a), this appears to be the case. Superimposed on plastic straining and the start of significant
these data are those for the two initially cemented volume changes as the stress paths turn away
samples (CEMC and CEME) which started from the linear paths. Each of these tests (CEM3,
shearing from a post-yield state. No distinct yield 4, 5, D, F and G) reached a peak state after initial
points are seen in the stress-strain curves for yield with values q’jp’ in excess of M; this is
these tests, and the data are inseparable from believed to indicate continued influence of the
those of the uncemented samples despite the fact gypsum cement. Following the peak state the
that the states of the cemented samples were ini- paths appear to define a unique surface on which
tially still outside the normal compression line for each of them travels down towards the critical
uncemented samples. Data from test CEMF are state point. The stress-strain curves in Fig. 9
also shown in Fig. 9(a). Although this sample has show that this results from the tests reaching an
a higher initial value of p’/p& than CEME, it ultimate stress ratio at significantly smaller
started from a pre-yield state. The resulting strains than are required to bring the soil to a
stress-strain curve is much stiffer and a yield constant volume state.
point can be identified. The state paths for samples that had yielded
For the third class of test, sample CEM2 in during initial compression are shown in Fig.

Downloaded by [ University Of Western Australia] on [17/12/22]. Copyright © ICE Publishing, all rights reserved.
 CEMP

GEM3
UNCG
-c-
P- - ---_ ---UNC8
HH-------
/
1 /
/I
- -- Uncemented :
- Cemented
l Yield points
I

E,: %
0 20
Es: %
40 60
t.__F_T
0
, _Lj 20
E*: %
40

WI
(a) (W (c)

Fig. 9. Stress-strain and volumetric strain behaviour of cemented and uncemented samples: (a) high confining pressures; (h) intermediate confining pressures; (c) low confining
pressures

Downloaded by [ University Of Western Australia] on [17/12/22]. Copyright © ICE Publishing, all rights reserved.
 64 COOP AND ATKINSON

Critical state point

3 Yield points
- --- State boundary for
uncemented sand-plaster 0 Critical states
mixture (Fig. 3)

NCL

1000
(4
p’: kPa

Fig. 11. Isotropic compression of calcarenite

Shearing
Figure 12 shows the stress paths for constant p’
tests on natural calcarenite, with points marked
corresponding to the end of shearing and to yield.
The critical state line drawn through the end of

n I I
-; \E I
C
1 I
test points has gradient M = 1.59 (& = 39”); a
tentative yield curve has been drawn through the
“0 1 2 3 4 5 6 7 yield points.
P’lP’cs Figures 13 and 14 show stress-strain behaviour
K’)
and the stress paths normalized with respect to

Fig. 10. State paths for cemented samples: (a) low con-
fining pressures; (b) high confining preswres; prefix for 12000 /
all tests is CEM End of test
Yield points
-- -- Yield locus

10(b). None of the tests had a distinct yield point,

and the paths indicate that if sufftciently high
confining stresses were applied, the state path Critical
might follow the surface for uncemented material. state /
line

BEHAVIOUR OF NATURAL CALCARENITE

Compression
Figure 11 shows the behaviour of the natural
calcarenite samples during isotropic compression.
/

\
Although all the samples were nominally of the

L!L
same category of calcarenite, there were consider-
able differences in the initial specific volumes.
Also shown in Fig. 11 are the end points from the
shearing stages. From these, a critical state line
has been identified parallel to a tentative com-
pression line for destructured soil. The natural 0
0 4000
samples again reached states well outside the sug- p’: kPa
gested normal compression line of destructured
material. Fig. 12. Critical states and yield of calcarenite

Downloaded by [ University Of Western Australia] on [17/12/22]. Copyright © ICE Publishing, all rights reserved.
 MECHANICS OF CEMENTED CARBONATE SANDS 65

pressures do not have clearly defined yield points

2 and, as the confining pressure was increased (i.e.
the value of p’/p& decreased), the state paths
tended towards a state boundary surface for
destructured material.

CONCLUSIONS
It has been shown that the behaviour of
cemented carbonate sand which is often thought
to be difficult to determine can be described by
conventional soil mechanics. The behaviour of
natural calcarenite was found to agree well with
Fig. 13. Stress-strain behaviour of calcarenite; prefix for the general framework developed from tests on
all tests is RAN artificially cemented carbonate sand, and this
general framework is also likely to be applicable
to other naturally cemented soils.
p:,. The patterns of behaviour are very similar to The principal effects of cementing are to reduce
those observed for artificially cemented samples. the specific volume by addition of cementitious
In Fig. 13, the normalized stress-strain curves for material into the void spaces and to introduce
tests that had yielded during initial compression well-defined yield points into the stress-strain
(RANl, RAN3 and RAN4) are inseparable. This responses under compression and shearing. For
was also found for the tests on artificially the degree of cementing used in this work, yield-
cemented samples (see Fig. 9(a)). In Fig. 14(b) the ing under both compressive and shear loads
state paths for samples with higher confining occurred at states outside the boundary surface of
the uncemented soil. Yielding is associated with
breakage of the cement bonds and, for these car-

r
I bonate soils, with the onset of particle fracture.
0 Critml state
l Yield points
At large strains, both cemented and
II uncemented samples reached ultimate or critical
states. The critical state friction angle for
cemented material (&, = 37”) was slightly lower
than that for uncemented material (&, = 40”)
possibly due to the influence of cement coating
the particles. Otherwise, the cemented and
uncemented samples reached essentially the same
critical state lines.
There are two consequences of the reductions
6 in specific volume due to cementing. First, in
,L L- order to compare the behaviour of cemented and
(a)
uncemented material it is necessary to have the
same gradings in both sets of samples. In the
work described this was achieved by adding
gypsum casting plaster to all samples but leaving
it unhydrated in tests on uncemented material.
Second, the state of the soil is moved away from
the normal compression line. For uncemented
carbonate sand, Coop (1990) identified the state
of the soil relative to the normal compression and
critical state lines as a major factor controlling
peak strength. Hence, cementing contributes to
increased peak strength by virtue of the reduction
4
0 I I in specific volume.
0 2 4 6 Comparisons between the behaviour of Dog’s
P’lP’cs Bay sand described by Coop (1990) and the same
@) material with the addition of uncemented gypsum
Fig. 14. State paths for calcarenite: (a) low confining plaster demonstrated the significance of grading
pressures; (b) high conthing pressures; prefix for all tests on the behaviour of granular soils in general. The
isRAN mineralogy of the gypsum as compared with

Downloaded by [ University Of Western Australia] on [17/12/22]. Copyright © ICE Publishing, all rights reserved.
 66 COOP AND ATKINSON

calcium carbonate fines did not influence the 8,) % axial and radial strains
behaviour significantly. E, linear strain
Other than to alter the specific volume and E”natural strain
hence the state of the soil relative to the critical E, =
2(&,- &J/2
state line, cementing has little influence on E, =
E, + 2Er
1* gradient of critical state and normal consoli-
strength except at small confining pressures rela-
dation lines in In o:ln p’ space
tive to the bond strength, although the influence o:> 0: axial and radial effective stress
of cementing on stiffness and stress-strain behav- 44, critical state friction angle
iour is apparent at higher confining pressures.
For cemented samples it is possible to identify
yield points which may occur during compression REFERENCES
and shearing: from these, a yield curve can be Addis, M. A. (1987). Mechanisms of sediment compaction
drawn which, with knowledge of the behaviour of responsible for soil subsidence. PhD thesis, University
uncemented or destructured material, may allow of London.
the classes of behaviour shown in Fig. 8 to be Allman, M. A. & Poulos, H. G. (1988). Stress-strain
distinguished. At low confining stresses shearing behaviour of an artificially cemented calcareous soil.
may result in yield at strengths higher than the Proc. Int. Co@ Calcareous Sediments, Perth 1,
51-60. Rotterdam: Balkema.
frictional failure envelope of the uncemented soil,
Airey, D. W. & Fahey, M. (1991). Cyclic response of
and continued loading leads to strain softening.
calcareous soil from the north west shelf of Aus-
At high confining stresses yield occurs during tralia. Gtotechnique 41, No. 1, 101-122.
compression, in which case the soil behaviour is Apthorpe, M., Garstone, J. SC Turner, G. J. (1988).
strain hardening and its strength is frictional. The Depositional setting and regional geology of North
data also showed that the transition from cohe- Rankin ‘A’ foundation sediments. Proc. Znt. Conf.
sive to frictional behaviour was not well-defined. Calcareous Sediments, Perth 2, 357-366.
At intermediate stress levels samples were Bishop, A. W. & Wesley, L. D. (1975). A hydraulic tri-
observed to yield during the initial stages of axial apparatus for controlled stress path testing.
GPotechnique 25, 657-670.
shearing. Their peak strengths were slightly above
Boey, C. F. & Carter, J. P. (1988). Mechanical testing of
the frictional failure envelope, indicating some artificially cemented carbonate soil. Proc. 5th Aust.
continued post-yield influence of the cementing. N.Z. Geomech. Conf, Sydney, 145-149.
Butterfield, R. (1979). A natural compression law for
soils. Giotechnique 29, No. 4. 469-480.
ACKNOWLEDGEMENTS Carter, J. P., Johnston, I. W., Fahey, M., Chapman,
This research was carried out partly within a G. A., Novello, E. A. & Kaggwa, W. S. (1988). Tri-
project at City University sponsored by BP Inter- axial testing of North Rankin calcarenite. Proc. Int.
national to investigate the behaviour of piles in Co@ Calcareous Sediments, Perth 2, 515-530.
Chandler, H. W. (1985). A plasticity theory without
carbonate soils. The Authors are grateful to
Drucker’s postulate suitable for granular materials.
members of the project steering committee for
J. Mech. Phys. Solids 33,215-226.
their discussions and comments, and particularly Coop, M. R. (1990). The mechanics of uncemented car-
to Professor G. T. Houlsby who provided the bonate sands. Giotechnique 40, No. 4,607-626.
samples of Dog’s Bay sand and M. Sweeney who Evans, K. M. (1987). A model study of the end bearing
arranged for the delivery of the core samples. The capacity of piles in layered carbonate soils. DPhil
Authors are also grateful for the comments and thesis, Oxford University.
suggestions made by Dr R. N. Taylor. Additional Fookes, P. G. (1988). The geology of carbonate soils
tests on uncemented samples and gypsum plaster and rocks and their engineering characterisation
and description. Proc. Int. Co@ Calcareous Sedi-
were carried out by Zhang Fuyou of the Xian
ments, Perth 2, 787-806.
Architectural Design Institute, China, and T.
Houlsby, G. T., Evans, K. M. & Sweeney, M. (1988).
Richter of Restock University, Germany. End bearing capacity of model piles in layered car-
bonate soils. Proc. Int. Conf Calcareous Sediments,
Perth 1, 209-214.
NOTATION Huang, J. T. & Airey, D. W. (1991). The manufacture of
e voids ratio cemented carbonate soils. Research report R631.
M = q’/p’ at the critical state University of Sydney.
N* value of In u for isotropically compressed sam- Jewell, R. J. & Andrews, D. C. (Eds) (1988). Proc. lnt.
ples on the NCL at p’ = 1.0 kPa Conf: Calcareous Sediments, Perth 1.
N*0 value of In o for one-dimensionally compressed Jewell, R. J. & Khorshid, M. S. (Eds) (1988). Proc. Int.
samples on the K, NCL at p’ = 1.0 kPa Conf Calcareous Sediments, Perth 2.
K, ut/u: for one-dimensional compression King, R. & Lodge, M. (1988). North West Shelf
p’ = (a; + 2a;)/3 development--the foundation engineering challenge.
q’ = (a: - 5:) Proc. Int. Conf: Calcareous Sediments, Perth 2, 333-
u specific volume 342.

Downloaded by [ University Of Western Australia] on [17/12/22]. Copyright © ICE Publishing, all rights reserved.
 MECHANICS OF CEMENTED CARBONATE SANDS 67

Leroueil, S. & Vaughan, P. R. (1990). The general and Conf. Calcareous Sediments, Perth 2,343-356.
congruent effects of structure in natural soils and Semple, R. (1988). State of the art report on engineering
weak rocks. Gtotechnique 40, No. 3,467-488. properties of carbonate soils. Proc. Int. Conf: Cal-
Poulos, H. G. (1988). The mechanics of calcareous sedi- careous Sediments, Perth 2,807-836.
ments. Proc. 5th Aust. N.Z. Geomech. Conf, Sydney. Skempton, A. W. (1954). The pore pressure parameters
Price, G. P. (1988). Fabric of calcareous sediments at A and B. Ghotechnique 4, No. 4, 143-147.
North Rankin ‘A’, North West Shelf. Proc. Int. Con& Vaughan, P. R., Maccarini, M. & Mokhtar, S. M.
Calcareous Sediments, Perth 2, 367-376. (1988). Indexing the engineering properties of
Renfrey, G. E., Waterton, C. A. & Van Goudoever, P. residual soils. Quart. J. Engng Geol. 21, No. 1,
(1988). Geotechnical data used for the design of the 69-84.
North Rankin ‘A’ platform foundation. Proc. Int.

Downloaded by [ University Of Western Australia] on [17/12/22]. Copyright © ICE Publishing, all rights reserved.