Behaviour of MH silts with varying plasticity indices

118 Dominic E. L. Ong BE(Hons), PhD, PEng, MIEM CPEng, MIEAust Associate Professor and Director, Research Centre for Sustainable Technologies, Faculty of Engineering, Science and Computing, Swinburne University of Technology Sarawak Campus, Kuching, Malaysia (corresponding author: elong@swinburne.edu.my) (Orcid:0000-0001-8604-8176) Retnamony G. Robinson BEng(Hons), MSc, PhD Professor, Civil Engineering Department, Indian Institute of Technology Madras, Chennai, India

It is important for geotechnical engineers to understand the intrinsic and mechanical behaviour of silt as it is presently recognised that there exist gaps in understanding its fundamental behaviour.The behaviour of kaolin samples with varying clay and silt contents was investigated in the present study.This study characterised the samples by their corresponding ranges of plasticity index (P I )namely, P I £ 13% and P I > 13%.This outcome is achieved by interpreting the results of Atterberg limit tests, particle size analysis, isotropically consolidated undrained triaxial tests and oedometer tests carried out on the kaolin samples with P I between 7 and 16%.The results obtained in this study show good reliability when compared to 65 sets of past significant experimental studies derived from 33 established past research papers.

Introduction
Based on particle size and plasticity characteristics, soils are classified as sand, silt and clay.The behaviours of sand and clay are unique.However, the behaviour of silts lies in between the behaviours of clays and sands.Some silts behave more like sand, and some others more like clay depending on their basic characteristics.It is essential to identify silts behaving as sand-like or clay-like.Boulanger and Idriss (2006) found that sand-like (behaving more fundamentally as sand) and clay-like soils (behaving more fundamentally as clay) have some fundamental differences in terms of stress-strain behaviour, compressibility and the slope of the critical-state line (CSL) in the e-ln p 0 space against the slope of the normally consolidated line (NCL).The major observations of Boulanger and Idriss (2004) include the following.
(a) Sand-like materials exhibit different slopes for the CSL and NCL, while clay-like materials show a similar slope for the CSL and NCL in the e-ln p 0 space.(b) The effective stress paths of sand-like materials in undrained monotonic shearing show an initially contractive response (positive pore pressure increments) followed by a transition to an incrementally dilative response (decreases in pore pressure).(c) Sand-like materials have little compressibility such that their void ratio does not change significantly as the effective consolidation stress is increased, while clay-like materials are relatively more compressible.
In a subsequent study, Boulanger and Idriss (2006) proposed that materials with P I less than 7% would have a sand-like behaviour, while materials with P I greater than or equal to 7% would exhibit a clay-like behaviour.In addition, it was also described that finegrained soils with P I values between 3 and 6% may exhibit a transitional behaviour.As a result, an appropriate procedure for examining the liquefaction resistance of soils was developed based on the behaviour of the materials.Boulanger and Idriss (2006) stated that the Atterberg limits tests are more reliable than particle size analysis, particularly for clays (i.e.<2 mm), in order to correlate to the stress-strain characteristics of soils and also to differentiate between clay-like and sand-like behaviours.
It is also vital to note that Ferreira and Bica (2006) reported that transitional soils with particle size distributions between those of clean sands and plastic clays develop their own respective unique NCL and CSL and are not in accordance with the critical-state framework described for either sands or clays.In a research study, Nocilla et al. (2006) recommended identifying a new framework to describe the behaviour of silts in the transitional form.Such a framework has yet to be identified or developed due to the behaviour of silt being perhaps more complex than those of sand and clay.
Much research has been reported on low-plasticity silt (ML), with a liquid limit (L L ) lower than 50%namely, by Wang and Luna (2012), Boulanger and Idriss (2006), Nocilla et al. (2006) and Hyde et al. (2006).They dealt with the characterisation and/or mechanical properties of ML.However, it was observed that very limited studies have been carried out to understand the behaviour of high-plasticity silts (MH) with a liquid limit higher than 50%.In this research, MH silts with L L that ranges from 52 to 64% were tested, and the results are reported.Nocilla et al. (2006) investigated the behaviour of Italian silt with a plasticity index (P I ) of 12% and found that as the clay contents reduce, the behaviour of silt changes from a clay-like behaviour to a transitional form demonstrating behaviours between those of clays and clean sands.For this situation, the silt in the transitional form demonstrates neither a unique NCL behaviour nor any unique CSL behaviour.Hyde et al. (2006) reported that the mechanical properties of low-plasticity silt with P I of 6% were sand-like.Furthermore, Wang and Luna (2012) characterised the lowplasticity Mississippi River Valley (MRV) silt with P I of 6% by using triaxial compression tests with different overconsolidation ratios (O CR ) of 1, 2 and 8 and reported that no unique critical state could be observed.The CSL in the void ratio (e) against the logarithm of mean effective normal stress (p 0 ) space was not parallel to the normal consolidation curve, thus possibly indicating a sand-like behaviour.Ladd et al. (1977) as cited by Wang and Luna (2012) stated that deviator stress of some clays can be normalised by the effective consolidation pressure.The deviator stresses of the MRV silts with different O CR values were normalised by the effective consolidation pressures, after which the results indicated that the MRV silts had a unique clay-like behaviour.The method identified by Boulanger and Idriss (2006) for distinguishing between sand-like and clay-like behaviours has worked well on both low-plasticity (ML) silts presented by Nocilla et al. (2006) and Wang and Luna (2012).Nonetheless, Wang and Luna (2012) likewise reported that the work of Boulanger and Idriss (2006) did not capture the effect of O CR on silt behaviour.
As such, it is envisaged that the work of Boulanger and Idriss (2006) can be further refined to accommodate a larger variety of soils that include high-plasticity silts (MH).In the present study, the authors carried out a series of experiments on four types of kaolin samples with different particle size distributions and plasticity values.The results are analysed along with the data presented in the literature so as to identify the clay-like and sand-like behaviours.

Material description
Four types of commercially available kaolin powder with varying clay contentsnamely KM20, KM25, KM35 and KM55were  utilised to set up the reconstituted kaolin samples.The kaolin was mined at depths between 4•5 and 6•0 m below the ground.Aluminium silicate is the predominant chemical constituent of kaolin.Reconstituted kaolin samples were found to be ideal in this study on the grounds that kaolin (a) has appropriate ranges of particle sizes (silt and clay fractions), (b) is relatively less expensive than obtaining undisturbed soil samples from the field and (c) tests can be consistently repeated with confidence.

Soil properties
British standard BS 1377-2:1990 (BSI, 1990a) was adopted to determine the index properties such as L L by using the Casagrande method as well as plastic limit (P L ) and particle size distribution.Both sieve and hydrometer analyses were carried out to determine the percentages of sand, silt and clay.
The L L , P L and P I of the kaolin samples are summarised in Table 1.They are found to be directly proportional to the clay contents of the kaolin samples as shown in Figure 2. The reconstituted kaolin samples are classified as high-plasticity silt (MH) in accordance with ASTM D 2487 (ASTM, 2000).
The activity (A c ) of fine-grained soils, the ratio of P I to the percentage clay content, is the amount of water that is attracted to the surfaces of the soil particles.The amount of water attracted is largely influenced by the amount of clay that is present in the soil (Lambe and Whitman, 1969;Skempton, 1953).Skempton (1953) stated that A c provides a convenient value for assessing the particular minerals found in clay.In other words, before X-ray diffraction tests are done on the samples, soils with different minerals, such as kaolin, can be differentiated by their respective A c values since different minerals are characterised by their unique A c values.Table 1 shows that the A c of the reconstituted kaolin samples are 0•66, 0•55, 0•77 and 0•62 for KM20, KM25, KM35 and KM55, respectively, which are considered reasonable as they are generally close to the A c value given by Skempton
Strips of filter paper were attached on the sides of the specimens for radial drainage as per specifications by Head (1998) so as to accelerate the consolidation during the consolidation phase.The specimens were saturated by applying back pressures until Skempton's pore pressure parameter (B) value of at least 0•98 was achieved.The CIU tests were carried out at different initial effective Effective confining pressure, p 0 c : kPa confining pressures of 100, 150, 200, 250, 300 and 400 kPa.As the time taken for consolidation was shorter than 2 h for all specimens, a minimum of 2 h was considered in order to calculate the required shearing rates as suggested by Head (1998).As such, a shearing rate of 0•07 mm/min was adopted.A similar shearing rate was also reported by Pillai et al. (2011) for tests on kaolin.
During the consolidation phase of the triaxial test, the volume change of the sample was continuously monitored with time.
Once the consolidation was over, the sample was sheared under undrained conditions.The axial load, axial deformation and the pore pressures were recorded using a load cell, a displacement transducer and a pore pressure transducer, respectively.The tests were continued until axial strain of 20% was reached.The corrected length was used for computing the strains during the shearing stage.
CIU tests at six different initial effective confining pressures of 100, 150, 200, 250, 300 and 400 kPa were carried out for each type of kaolin sample with varying clay-silt contents (KM20, KM25, KM35 and KM55).Therefore, a total of 24 CIU tests were performed.The samples were labelled based on the type of kaolin and the confining pressure in the triaxial test.For example, KM20-100 means 100 kPa confining pressure was applied on the sample KM20.The conducted tests are summarised in Table 2.

Sample preparation
Reconstituted and undisturbed samples Burland (1990) demonstrated that the reconstituted clay should be mixed at higher water contents from 1•00 to 1•50 times its liquid limit (L L ) and preferably consolidated one-dimensionally to obtain the intrinsic properties of the soils.As there is no proper guideline for reconstituting kaolin silt, the mixing ratio as proposed by Burland (1990) has been adopted.A comparative mixing ratio was likewise embraced in the work of Pillai et al. (2011) for kaolin to prepare normally consolidated reconstituted samples.
In a different work by Hyodo et al. (1994), triaxial compression and extension tests were conducted on both normally consolidated undisturbed and reconstituted marine clays with a liquid limit of 124•2%, a plastic limit of 51•4% and a plasticity index of 72•8%.The monotonic axial load was applied at an axial strain rate of 0•1%/min.The test outcomes demonstrated that undisturbed and reconstituted samples behaved similarly in terms of deviator stresses and stress paths.Hyodo et al. (1994) added that the ageing effects such as chemical bonding or secondary compression for undisturbed samples were eliminated during the application of effective confining stress to the normally consolidated state during the consolidation stage.Therefore, the work of Hyodo et al. (1994) guaranteed that the behaviour of normally consolidated undisturbed samples can be decently depicted by reconstituted samples.

Sample preparation
The use of reconstituted kaolin samples prepared by consolidation of the soil-water mixture in the form of slurry state is a common approach in the laboratory testing of soils (Pillai et al., 2011;Wang and Luna, 2012).Firstly, the kaolin powder was mixed with distilled water at a water content of 1•50 times its liquid limit.
The water-soil mixture was mixed thoroughly in an automatic soil mixer to form the slurry.For triaxial tests, 38 mm dia.samples were used.The slurry was directly poured into the 38 mm dia.sampling tube after applying silicone grease on the inner surfaces of the tubes to reduce side friction.Subsequently, the sample was gradually loaded to a vertical consolidation pressure of 100 kPa.The vertical consolidation pressure was achieved by the application of a dead weight of 11•5 kg on the sample through a guide rod.
After the consolidation stage, the kaolin samples were extruded using a universal extruder.The reconstituted kaolin samples were then trimmed to 38 mm diameter and 76 mm height.
For the 1D consolidation tests, the slurry was carefully poured into a 60 mm dia.and 20 mm high 1D consolidation ring along with the collar so that the thickness of the sample was 30 mm.The slurry was initially consolidated to a consolidation pressure of 6•25 kPa.Once the consolidation under 6•25 kPa was complete, the sample was carefully trimmed to a thickness of 20 mm.The sample was then subjected to a 1D consolidation test with a load increment ratio of 1•0.

CIU tests
Figures 3(a), 3(b) and 3(c) show the stress-strain behaviour, stress paths and pore pressure-strain behaviour of the reconstituted kaolin samples, respectively.For clarity, separate plots of these three parameters for the KM20, KM25, KM35 and KM55 samples can be found in Figures 4, 5, 6 and 7, respectively.
When the clay contents decreased (from KM55 to KM20), the deviator stresses at failure (q f ) increased as shown in Figure 3(a).The mean effective normal stress (p 0 ) and deviator stress (q) can be expressed as Equations 1 and 2a.Equation 2a can then be written in incremental form as shown in Equation 2b.For the CIU tests, the change in the total confining pressure (Ds 3 ) equals zero.Hence, Equation 2b can be rewritten as Equations 2c and 2d as the changes in deviator stress (Dq) are defined as negative and positive for strain softening and hardening, respectively.
From Equations 2c and 2d, it is obvious that samples are expected to exhibit strain softening only if (a) the change in the total vertical stress (Ds 1 ) or (b) the summation of the changes in the effective vertical stress ðDs 0 1 Þ and excess pore water pressure (Du) is negative.In other words, the kaolin samples will strain soften, due to the negative value of Du (dilation) and/or negative value of Ds 0 1 (axial unloading).
Similarly, the kaolin samples will exhibit strain hardening if (a) the change in Ds 1 or (b) the summation of Ds 0 1 and excess pore water pressure (Du) is positive.The kaolin samples will strain harden due to the positive values in Du (contraction) and/or positive value in Ds 0 1 (compression) For softening For hardening The effective stress paths in Figure 3(b) were plotted in the Cambridge stress space with Equation 1 in the abscissa and Equation 2a in the ordinate.A similar approach was also adopted by Wang and Luna (2012) to understand the stress paths of silts.
The effective stress paths of KM20 and KM25 with the lower P I   3 summarises the changes in effective vertical stresses and pore pressures.As shown in Figure 3(a), KM20 exhibited strain softening at s 0 3 values of 100, 150 and 200 kPa due to the development of negative values of Du (dilation).This is because of having negative values in both Du and Ds 0 1 as well as the summation of Du and Ds 0 1 being similarly negative, resulting in the change in deviator stress (Dq) also being a negative value (see Equation 2c).Theoretically, KM20 is expected to exhibit strain softening at the aforementioned confining pressures, as observed in Figure 3 6 and 7, the behaviours of strain softening and strain hardening are very obvious.Ishihara et al. (1975) defined the phase transformation point as the point at which the stress path turns its direction in p 0 -q space.They also observed that for cyclic triaxial tests, it is necessary for a sample to go at least once through this critical value in order to be taken to a completely liquefied state.In this sense, the phase transformation point may be considered as a threshold at which the behaviour of sand as a solid is lost and transformed into that of a liquefied state.

Critical state
The deviator stresses at failure (q f ) can be identified when either of the following conditions are achieved: (a) maximum principal stress ratio ðs 0 1 =s 0 3 Þ and (b) 'critical-state' conditionthat is, constant deviator stress and pore pressure.The corresponding values of strain and pore pressure are axial strain at failure (e f ) and pore pressure at failure (u f ), respectively (Head, 1998).The work of Wang and Luna (2012) showed that the criterion of maximum principal stress ratio always gives a consistent estimation of the effective angle of shearing resistance.Hence, in the present study, the maximum principal stress ratio method has been used to identify q f .The outcome of projecting the effective stress paths in Figure 3(b) onto the semilogarithmic plot of v-ln p 0 is shown in Figure 8.As drainage was not allowed during the shearing stage of the sample, the effective stress paths in v-ln p 0 were expected only to move horizontally since undrained shearing occurred at constant specific volume.It is observed that the CSLs and the NCLs are not parallel as shown in Figure 8 for the kaolin with lower clay contents (KM20 and KM25) due to the P I values of KM20 and KM25 being lower at 7 and 11%, respectively, compared to other samples, which have P I values greater than 13% (KM35 (P I = 16%) and KM55 (P I = 15%)).Therefore, it is defined that KM20 and KM25 behave as sand-like materials.A similar observation of non-parallel CSL and NCL curves was also reported by Wang and Luna (2012) in their study of triaxial compression test on the MRV silt with P I of 6%.Furthermore, it is interesting to observe that as the P I values of the kaolin samples increased starting from KM20 (P I = 7%) and KM25 (P I = 11%) to KM35 (P I = 16%) and KM55 (P I = 15%), the gradients of the respective CSLs seem to be getting more and more parallel to the gradients of their corresponding NCLs, indicating that the clay-like behaviour starts to be more prominent As shown in Figure 9 for better clarity, the normalised deviator stress of KM20 and KM25 cannot be effectively bound by the effective consolidation pressures ðp 0 c Þ into a narrower single band, particularly at 100 kPa for p 0 c , which indicates that they are not clay-like materials.Comparatively, KM35 and KM55 can be better normalised by p 0 c into a single band and have shown a unique behaviour as clay-like materials.
Based on this important observation, since KM35 and KM55 have higher P I values and their NCLs and CSLs are much more parallel (evident in Figure 8), it is thus postulated that KM35 and KM55 demonstrate clay-like behaviours.In view of this, a P I value of equal to or greater than 13% can be postulated as the boundary when claylike behaviour types of soils would start to dominate.Similar results were also reported by Wang and Luna (2012) and Boulanger and Idriss (2006) in their works, which showed sand-like materials exhibiting non-parallel NCL and CSL behaviours.
Table 4 summarises the results of the CIU tests performed on the reconstituted kaolin soils.Based on the mean effective normal stress at failure ðp 0 f Þ and q f from Table 4, the CSLs of reconstituted kaolin soil samples are then determined, as shown in Figure 10.The slope of the CSL increased when the clay contents of the reconstituted kaolin soils reduced.The slope of the CSL is M, and the effective angle of shearing resistance friction (f 0 ) can be back-calculated using the following equation.
A similar approach was also used by Wang and Luna (2012) to find the f 0 value.The M and f 0 values of the kaolin soils are summarised in Table 4.
The relationship between the clay contents and effective angles of shearing resistance is indirectly proportional as shown in Figure 11.The correlation between clay contents (C L %) and effective angle of shearing resistance (f 0 ) for the kaolin is with coefficient of determination (R 2 ) equal to 0•931.Similarly, Nocilla and Coop (2008) conducted CIU triaxial tests on alluvial sediments from the floodplain of Po River in Italy with C L % ranging from 16 to 25% and P I ranging from 13 to 16%.It was also found that the back-calculated f 0 showed a similar inverse correlation with C L %.However, the observed scatters in their data set could be due to the presence and the variation in silt and sand contents.

One-dimensional consolidation tests
The compressibility of the reconstituted kaolin was determined using 1D consolidation tests.The semilogarithmic e − log s 0 v (effective vertical stress) relationships obtained from the 1D consolidation testing are plotted in Figure 12.The initial void ratio is directly proportional to the L L of the soils.If the L L is higher, the initial void ratio will also tend to be higher due to the higher water content present within the void of the samples.The results of the 1D consolidation tests are summarised in Table 5.Furthermore, the gradient of compression line (l) and the gradient of swelling line (k) show a directly proportional relationship with the C L % of the samples, as shown in Figure 13.

Gradient of compression lines of NCL and CSL from CIU tests
The gradients of the compression lines of the reconstituted kaolins were back-calculated from the NCLs and CSLs of the CIU tests as shown in Figure 8.Both the gradients of the compression lines of NCLs and CSLs are directly proportional to the clay contents as shown in Table 6.Furthermore, the gradient of the compression lines of NCL (l NCL ) converges to the gradient of the compression lines of CSL (l CSL ) when the clay content of kaolin increases.This also indicates that the NCLs and CSLs of KM55 and KM35 are relatively more parallel than the NCLs and CSLs of KM20 and KM25.As shown in Figure 14 and Table 6, the ratios of the gradients of compression lines (l NCL /l CSL ) of KM20, KM25, KM35 and KM55 are 0•702, 0•806, 0.846 and 0•948, respectively.A value of l NCL /l CSL closer to 1•000 indicates that the NCL and CSL are more parallel.The R 2 of the l NCL /l CSL and Axial strain, ε a : % Normalised deviator stress, q/p' Distinguishing between clay-like and sand-like soil behaviours through plasticity index (P I ) A sand-like behaviour is observed from the reconstituted kaolin soils of KM20 and KM25, characterised by their lower P I values of 7 and 11%, respectively, as well as their CSL and NCL being non-parallel (see Figure 8).In the event that the method of Boulanger and Idriss (2006), which was developed for distinguishing clay-like (P I greater  than or equal to 7%) and sand-like (P I smaller than 7%) behaviours of silt based on P I , was applied to the high-plasticity silt (MH) of KM20 and KM25, then KM20 and KM25 would be 'incorrectly' classified as having a clay-like behaviour.
To evaluate further the reliability of the original work of Boulanger and Idriss (2006) on MH, various P I and L L data points obtained from well-established papers (see Table 7) are subsequently used to develop Figure 15 together with the classification rules for sand-like and clay-like behaviours as reported by Boulanger and Idriss (2004).As illustrated in Figure 15, apparently, some of the well-documented sand-like soils are lying slightly above the P I = 7% boundary suggested by Boulanger and Idriss (2006), which evidently demonstrates the conflict in trying to categorise the well-documented sand-like data to be within the P I = 7% boundary.Therefore, it is now proposed that the boundary for describing sand-like behaviour soils be raised from P I = 7% (Boulanger and Idriss, 2006) to P I = 13% so that the well-documented data points (see Table 7) for soils with a sand-like behaviour can be fulfilled, including that of KM20 (P I = 7%) and KM25 (P I = 11%), whose sand-like behaviour is clearly demonstrated in Figure 8.The boundary P I = 13% is also supported by Nocilla et al. (2006), who conducted reliable triaxial compression tests on Italian silts.In their work, the samples with clay contents of 4% (L L = 37% and P I = 13%) and 8% (L L = 34% and P I = 13%) had evidently demonstrated a sand-like behaviour as the effective stress paths showed an initially contractive response (due to increase in pore pressure) followed by a transition to an incrementally dilative response (due to decrease in pore pressure).
Finally, a threshold of P I greater than 13% is now suggested for soils behaving in a clay-like manner.This would then include the     As all the well-documented data points, including those from this study, can be simultaneously included within the newly postulated framework as shown in Figure 15, this study has been proven to be reliable, consistent and beneficial to engineers seeking a less expensive classification method for identifying silt behaviour.

Conclusions
Reconstituted kaolin samples of high-plasticity silts (MH) with varying clay contents were used to investigate the transitional behaviour of silts.Atterberg limit, dry sieving, hydrometer, CIU and oedometer tests were conducted on these reconstituted samples to characterise the behaviours of the samples in terms of clay-like or sand-like behaviours.
In the v-ln p 0 space derived from the CIU tests, the gradients of the respective CSLs seem to be approaching the same gradient of the NCLs when the P I of the reconstituted kaolins increased progressively from KM20 (P I = 7%), KM25 (P I = 11%), KM35 (P I = 16%) to KM55 (P I = 15%).KM35 and KM55 demonstrate a clay-like behaviour with P I values greater than 13%, while KM20 and KM25 exhibit a sand-like behaviour with P I values smaller than 13%.Therefore, with the support of well-documented soils sourced from the 33 papers summarised in Table 7 and the research work carried out in this paper, thresholds for (a) sand-like and (b) clay-like silts are thus proposed as (a) P I £ 13% and (b) P I > 13%, respectively.Based on the newly postulated framework, all the well-documented data points from Table 7 as well as the MH data points from this research can be simultaneously included.
Based on the mentioned important findings of this study, if the values of Atterberg limits are the only available data, silts can then be characterised in terms of sand-like or clay-like behaviours by using the newly proposed framework before an appropriate

Figure 2 .
Figure 2. Linear relationship between clay contents and L L , P L and P I of kaolin samples Figure 3. (a) Stress-strain, (b) stress paths and (c) pore pressure-strain behaviours of kaolinite samples subjected to CIU tests

Figure 3
Figure 3(c) shows the pore pressure behaviour of the kaolin samples.Table3summarises the changes in effective vertical stresses and pore pressures.As shown in Figure3(a), KM20 exhibited strain softening at s 0 3 values of 100, 150 and 200 kPa due to the development of negative values of Du (dilation).This is because of having negative values in both Du and Ds 0 1 as well as the summation of Du and Ds 0 1 being similarly negative, resulting in the change in deviator stress (Dq) also being a negative value (see Equation2c).Theoretically, KM20 is expected to exhibit strain softening at the aforementioned confining pressures, as observed in Figure3(a) or Figure4for better clarity.However, on the contrary, with KM20 tested under confining pressures s 0 3 of

Figure 4 .
Figure 3(c) shows the pore pressure behaviour of the kaolin samples.Table3summarises the changes in effective vertical stresses and pore pressures.As shown in Figure3(a), KM20 exhibited strain softening at s 0 3 values of 100, 150 and 200 kPa due to the development of negative values of Du (dilation).This is because of having negative values in both Du and Ds 0 1 as well as the summation of Du and Ds 0 1 being similarly negative, resulting in the change in deviator stress (Dq) also being a negative value (see Equation2c).Theoretically, KM20 is expected to exhibit strain softening at the aforementioned confining pressures, as observed in Figure3(a) or Figure4for better clarity.However, on the contrary, with KM20 tested under confining pressures s 0 3 of Figures 4(b), 5(b), 6(b) and 7(b), all samples exhibited the initial contraction before the phase transformation points.As shown in Table3, all samples exhibit post-peak dilation except for KM25-150, KM35-150, KM35-200, KM35-250 and KM55-150.Except for KM20-250, KM20-300, KM20-400, KM35-100 and KM35-150, all the samples exhibited postpeak strain softening.

Figure 5 .
Figure 5. (a) Stress-strain, (b) stress paths and (c) pore pressure-strain behaviours of KM25 kaolin samples subjected to CIU tests Figure 6.(a) Stress-strain, (b) stress paths and (c) pore pressure-strain behaviours of KM35 kaolin samples subjected to CIU tests

C3
well-documented soils sourced from the literature summarised in Table7, as well as KM35 (P I = 16%) and KM55 (P I = 15%) as included in Figure15.Indeed, Figure8evidently shows that the apparent clay-like behaviours of KM35 and KM55 have parallel CSLs and NCLs.

Figure 15 .
Figure 15.Boundaries for clay-like and sand-like behaviours (this study and others)

Table 1 .
Gradation, L L , P L and P I values of the kaolin samples L : % P L : % P I : %

Table 2 .
Summary of the conducted tests L , P L and P I

Table 3 .
Changes in effective vertical stresses and pore pressures 1 Figure 8. CSLs and isotropically NCLs of kaolin samples derived from CIU tests 127

Table 4 .
Results of CIU tests, slopes of CSLs and effective angles of shearing resistance

Table 7 .
Documented Atterberg limits for soils with sand-like and clay-like behaviours (continued on next page)