Compressibility and consolidation properties of Santos soft clay near Barnabé Island

This is an Open Access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Compressibility and consolidation properties of Santos soft clay near Barnabé Island Vitor Nascimento Aguiar1# , Maurício do Espírito Santo Andrade2 , Ian Schumann Marques Martins3 , Jean Pierre Paul Rémy4 , Paulo Eduardo Lima de Santa Maria5 

The Santos harbor is the largest and most important port complex in Latin America. There are currently several expansion projects alongside the Santos harbor channel. A multipurpose terminal covering an area of 800,000 m 2 was built on Barnabé Island, on the left bank of the Santos harbor channel ( Figure 1). An instrumented pilot embankment was built in 2007 to provide field compressibility, consolidation and shear strength data of the soft clay foundation deposit (Rémy et al. 2011). A comprehensive laboratory test program on high-quality samples as well as in situ geotechnical investigation were also carried out.
The purpose of this paper is to present the characterization test results and the compressibility and consolidation parameters obtained from one-dimensional consolidation tests of the subsoil samples taken in the pilot embankment area before its construction. Figure 2 shows the boreholes location for standard penetration tests (SPM) and for taking undisturbed samples (SRA) in the pilot embankment area. The undisturbed samples taken from borehole SRA203 were sent to the Soil Rheology Laboratory of the Federal University of Rio de Janeiro, where they were tested. The samples taken from boreholes SRA201 and SRA202 were tested in another laboratory and are not presented herein.

Sampling, transportation and storage
To take good-quality samples, ABNT (1997) and "Technical specification for taking undisturbed samples" (Aguiar, 2008) were followed. 10 cm-inner diameter and 70 cm-long thin-wall fixed piston samplers were used. The samples were taken by a team trained by two of the authors when taking the first samples. Figure 3 shows the position of the twelve undisturbed samples taken from borehole SRA203 along the borehole SPM203 profile.   The borehole SPM203 showed a first layer of very soft clay 1.60 m thick with N = 0. According to the Massad (2009) genetic classification, this layer is a Mangrove clay. It must be recognized that N = 0 terminology comprises a wide range of soft clay consistencies, that goes from P/30, meaning a 30 cm penetration of the SPT sampler under the hammer's weight, to 0/XXX, where XXX means the penetration in centimeters of the SPT sampler under the rod set weight only. For instance, in Figure 3, at 1.0 m depth the symbol 0/153 means that the SPT sampler penetrated 153 cm under the rod set weight. Such consistency range is usually found in Santos Mangrove clay layers. Underlying this clay layer, there is a 3.30 m-thick sand layer with N between 3 and 6. The first undisturbed sample was taken 10 cm below the bottom of this sand layer (Figure 3).
The sample ends were covered with PVC film and aluminum foil and sealed with paraffin wax. The sampler tip was protected against bumps with a 10 cm-high PVC rigid ring. The samples were shipped in vertical position into wood containers with the tip downwards (ASTM, 2007). After sampling, the samplers were placed into the wood containers, stored in a room protected from the sun, where people circulation was not allowed. After taking the last sample, the containers were sent to the laboratory and stored in a humid room. Table 1 shows grain-size distribution (ABNT, 1995), liquid limit (w L ), plastic limit (w P ), plasticity index (I P ), specific gravity (G s ) and organic matter content (OM) obtained for the indicated segments of each SRA203 sample (second column of Table 1). Water content (w), unit weight of soil (γ), natural void ratio (e 0 ) and degree of saturation (S r ) are the average values obtained from the undisturbed consolidation test specimens belonging to each sample segment. Figure 3 shows the subsoil profile according to tactilevisual examination of the SPT samples from borehole SPM203 only. Figure 3 also shows the laboratory characterization test results carried out on SRA203 samples and w, γ and e 0 values are plotted for all undisturbed consolidation test specimens. The depths indicated are not the depths at the top and bottom of the undisturbed samples as shown in Figure 3, but rather the depths of segments of the samples where the undisturbed consolidation test specimens were trimmed. Samples SRA203(1), SRA203(2) and SRA203(3) test results revealed that the subsoil profile corresponding to their depths should be better described as "silty clayey sand". According to sample SRA203(11) test results, the subsoil profile at its depth should be better described as "clayey sand".

One-dimensional consolidation test procedure
Incremental loading one-dimensional consolidation tests were carried out on the twelve samples from borehole SRA203.
Long-term loading stages were run in selected tests to investigate secondary consolidation and stress relaxation. However, these results are outside the scope of this paper.
Specimens are identified by the acronym CPMX, where CP means specimen, M is the sample number and X the letter that denotes the order position of the specimen in sample M, for instance: CP6E is the fifth (letter E) specimen trimmed in sample SRA203(6). Tables A1 and A2 (see Appendix) indicate the depth interval from which each specimen was trimmed.
All tests were carried out on Bishop-type consolidation frames, with settlements being measured by 0.01 mm/division dial gages, under temperature of 20 ± 1 ºC. Temperature variations were daily monitored by a maximum and minimum thermometer.
Each consolidation test was performed using one out of two loading criteria: criterion A, based on the specimen vertical strain rate (ε), and criterion B, based on stage duration.
In criterion A, a new loading stage was applied whenever the vertical strain rate (ε) reached 10 -6 s -1 , calculated as: where: H : specimen height corresponding to reading of order i; H ∆ : settlement difference ( ) time elapsed between readings of orders i and i + 1.
For samples SRA203(4) to SRA203(10), it was observed that ε =10 -6 s -1 corresponded to the higher integer power of 10 after the "end of primary" consolidation, calculated by both Taylor (1942) ( t ) and Casagrande (log(t)) methods of coefficient of consolidation (c v ) determination for specimens whose drainage path is about 1 cm. A first series of consolidation tests ("series one") was performed on all twelve samples using criterion A. The following specimens were tested: -four undisturbed specimens from sample SRA203(1), -three undisturbed specimens and one remolded specimen from each out of samples SRA203(2) to SRA203(7), -two undisturbed specimens from each out of samples SRA203(8) to SRA203(12), totalizing thirty eight tests, being thirty two on undisturbed specimens and six on remolded specimens. Remolding was done prior to specimen trimming, by smashing an amount of sample inside a plastic bag. Consolidation tests on remolded specimens were carried out to check the quality of the undisturbed specimens by comparing their results.
All tests in series one underwent the following loading sequence up to 100 kPa: 3.13, 6.25, 12.5, 25, 50 and 100 kPa. From 100 kPa on, the loading and unloading sequences followed different patterns, as shown in Table 2. In some tests, a loading stage was selected to monitor secondary consolidation under a chosen overconsolidation ratio (OCR). In other tests, stress relaxation was observed by preventing specimen settlements. These stages were analyzed by Aguiar (2008) and Andrade (2009).
In tests with specimens CP7A, CP7B, CP7C and CP7D, influence of temperature was investigated in loading stages beyond 300 kPa. As σ' p values of these specimens are not greater than 160 kPa and the compression index (C c ) values were determined in the virgin compression curves immediately after σ' p , the C c values obtained were not affected by the loading stages in which temperature effects were investigated.
A second series of consolidation tests ("series two") was carried out using criterion B, in which four undisturbed specimens were tested from each out of samples SRA203(3) to SRA203(10), totalizing thirty two tests.
In criterion B, the loading stages lasted 24 hours and the loading sequence was 3.13, 6. 25, 12.5, 25, 50, 100, 150, 200, 300, 500 and 800 kPa, followed by unloading to 400 and 200 kPa, except for specimens CP8C, CP8D, CP8E and CP8F, which were unloaded to different stresses in order to investigate secondary consolidation under different OCR values, as shown in Table 3. These unloading stages were analyzed by Andrade (2009).
Stress increment ratio (Δσ/σ) < 1 in the loading sequence of series two was intended to determine σ' p with more accuracy and to define more clearly the compression curve.
Hence, seventy consolidation tests were run in the two test series, being sixty four on undisturbed specimens and six on remolded ones.

Test results
Vertical strain (ε) versus vertical effective stress (σ' v ) (log) and void ratio (e) versus σ' v (log) compression curves of series one specimens were plotted with ε and e of each loading stage corresponding to: a) "end of primary" consolidation calculated by Taylor's (1942) method and b) vertical strain rate (ε) of 10 -6 s -1 .
ε versus σ' v (log) and e versus σ' v (log) compression curves of series two specimens were plotted with ε and e of each loading stage corresponding to: a) "end of primary" consolidation calculated by Taylor's (1942) method,   b) vertical strain rate (ε) of 10 -6 s -1 and c) end of 24 hours. Compression curves plotted in terms of void ratio (e) according to the criteria mentioned above were compared. Figure 4 shows an example of such comparison for specimen CP5E, from series two. The comparisons of all the other specimens were presented by Aguiar (2008) and Andrade (2009). As observed in all tests, the 24-hour compression curve lies below the ε = 10 -6 s -1 compression curve, which, in its turn, lies below the "end of primary" compression curve.
From 12.5 kPa on, c v values were determined by Taylor's (1942) method and plotted against the average σ' v values of the respective loading stages. ε versus σ' v (log) and e versus σ' v (log) compression curves corresponding to "end of primary", ε = 10 -6 s -1 and 24 hours, together with the c v (log) versus average σ' v (log) curves, of all seventy specimens were presented by Aguiar (2008) and Andrade (2009). Figure 5 shows the ε versus σ' v (log) curves corresponding to ε = 10 -6 s -1 and the c v (log) versus average σ' v (log) curves of all SRA203(6) specimens. Specimen CP6D was remolded. Figure 6 shows the ε versus σ' v (log) curves corresponding to 24 hours and the c v (log) versus average σ' v (log) curves of series two SRA203(8) specimens. The excellent repeatability of the results obtained for the SRA203(6) and SRA203(8) specimens ( Figures 5 and 6) was also observed in all SRA203(4) to SRA203(10) specimens, which belong to the SFL clay layer, as discussed further.   Figure 7 gathers typical e versus σ' v (log) curves corresponding to 24 hours and their respective c v (log) versus average σ' v (log) curves from one undisturbed specimen of each sample from SRA203(4) to SRA203(10). Compression curves for samples SRA203(1), SRA203(2), SRA203(3), SRA203(11) and SRA203(12) are not included since they are sand. Figure 7 shows that, for practical purposes, samples SRA203(4) to SRA203(10) can be assumed to belong to a single "homogeneous" clay layer.
The preconsolidation (yield) stress (σ' p ), compression index (C c ) and recompression index (C r ) were obtained for all specimens (series one and two) from e versus σ' v (log) curves corresponding to ε = 10 -6 s -1 as shown in Table A1 (see Appendix A), which also shows C c /(1+e 0 ) and C r /C c values. The same parameters, including the swelling index (C s ), were also obtained for series two specimens from e versus σ' v (log) curves corresponding to 24 hours as shown in Table A2 (see Appendix A), which also shows C c /(1+e 0 ) and C r /C c values. All C s values correspond to an OCR = 4. The σ' p values were obtained according to Silva (1970) method. The C r , C c and C s values were determined as shown in Figure 8. σ' v0 is the effective overburden stress. Figure 9 shows the profiles of σ' p , C c /(1+e 0 ) and C r /C c obtained from e versus σ' v (log) curves corresponding to ε = 10 -6 s -1 of all specimens (series one and two - Table A1), and from e versus σ' v (log) curves corresponding to 24 hours of all series two specimens (Table A2). Figure 10 shows the stratigraphy of the subsoil based on borehole SPM203 SPT samples, characterization tests and physical indexes of the undisturbed samples as well as on tactile-visual examination when trimming the consolidation test specimens.

Stratigraphy and characterization
Since no undisturbed samples were obtained from the mangrove clay layer, a unit weight of 13.0 kN/m 3 was assigned to it based on Massad (2009, pp. 106). The existence of this layer was confirmed in situ via SPT samples examination. As no undisturbed samples were obtained from the top sand layer, a unit weight of 20.0 kN/m 3 was assigned to it since this layer is sandier than SRA203(12) sample, which unit weight is 19.7 kN/m 3 (Table 1).
Between the top sand layer and the SFL clay layer, there is a transition layer composed by three sandy sublayers identified based on samples SRA203(1), SRA203(2) and SRA203(3) characterization tests and physical indexes.
Samples SRA203(4) to SRA203(10) characterization tests and physical indexes revealed that they belong to a single SFL clay layer according to the Massad (2009) genetic   classification. It is worth noting the increase of water content and plasticity from sample SRA203(3) to SRA203(4), as well as the decrease of water content and plasticity from sample SRA203(10) to SRA203(11). Samples SRA203(6) and SRA203(7) have water content, void ratio, liquid limit and clay content higher than the others. Presence of kaolinite, smectite and illite was identified along the SFL clay layer by X-ray diffraction.
Samples SRA203(11) and SRA203(12) characterization tests and physical indexes showed that they belong to sand layers below the SFL clay layer.
The unit weights shown in Figure 10 are the average values from undisturbed consolidation test specimens of each sample and the effective overburden stress (σ' v0 ) profile was estimated with these unit weights.

Sample quality and remolding effects on onedimensional compression curves
The comparisons between compression curves of remolded and undisturbed specimens ( Figure 5) highlighted the following remolding effects (Ladd, 1973): 1) Decreases the void ratio (or increases the strain) at any given σ' v value; 2) Makes it difficult to define the point of minimum radius, thus obscuring σ' p ; 3) Lowers the estimated value of σ' p ; 4) Increases the compressibility in the recompression region; 5) Decreases the compressibility in the virgin compression region. Coutinho (1976) and Martins (1983) have also observed that remolding turns the concave shape of the virgin compression curve into a straight line. As σ' v increases, structure of undisturbed specimens is destroyed, making their behavior approach to that of the remolded specimen. Thus, as σ' v increases, the compression curves of all specimens tend to merge into a single curve ( Figure 5).
Another remarkable feature of high-quality specimens is the abrupt fall of the c v versus σ' v (log) curves when σ' v straddles σ' p . Such fall may be of two orders of magnitude ( Figures 5, 6 and 7). This is not observed in the c v versus σ' v (log) curve of the remolded specimen ( Figure 5). The smaller c v values in the recompression region of the remolded specimen are due to the compressibility increase in the recompression region caused by remolding.
Although not shown herein, no difference at all was observed between the compression curve of the remolded specimen and those of the "undisturbed" specimens trimmed on sample SRA203(2), which is 69% sand (Aguiar, 2008).
Regarding footnote (b) in Table A1, C r values could not be obtained according to Figure 8 since remolding pushed σ' p to a value lower than σ' v0 . Table 4 shows the quality classification of series two specimens according to Lunne et al. (1997), Coutinho (2007) and Coutinho (2007) modified by Andrade (2009) criteria. Based on his experience with highly plastic soft clays, Coutinho (2007) proposed a modification of Lunne et al. (1997) criterion. Andrade (2009) observed the following shortcoming in both criteria: the quality assigned to the upper bound of a class does not coincide with the quality assigned to the lower bound of the immediately above class. Andrade (2009) was able to solve this shortcoming by subdividing the classes in such a way that on the borderline of two subsequent classes, the quality to be assigned is the common term of both classes (Table 5). For instance: for Δe/e 0 = 0.080 the quality to be assigned is "fair".
Only three specimens (CP5G, CP10D and CP10F) out of thirty two were classified below "Good to Fair" according to the Coutinho (2007) modified criterion ( Table 4).

Comparison between compressibility parameters
obtained from ε = 10 -6 s -1 and 24-hour compression curves Table 6 compares σ' p , C r and C c values obtained from ε = 10 -6 s -1 and 24-hour "e versus σ' v (log)" curves of series two specimens from the SFL clay layer.
The ratio between σ' p from the ε = 10 -6 s -1 compression curve, denoted by σ' p (10 -6 s -1 ), and σ' p from the 24-hour  curve, denoted by σ' p (24 h), is within 1.03 and 1.12, with an average of 1.08, which is among the rate effects described by Graham et al. (1983), Leroueil et al. (1985) and Crawford (1986). For the Santos soft clay studied herein, σ' p (10 -6 s -1 ) is 8% higher, on average, than σ' p (24 h). As shown by Leroueil et al. (1985), σ' p depends on the strain rate adopted to plot the one-dimensional compression curve "e versus σ' v (log)", σ' p being higher, the higher the strain rate. This phenomenon is associated with the squeezing out of the viscous adsorbed water layers surrounding clay particles (Terzaghi, 1941;Taylor 1942;Lambe & Whitman 1979, pp. 299). The higher the plasticity index, the greater the thickness of the adsorbed water layer, in the sense explained by Bjerrum (1972;, magnifying secondary compression. Being so, the higher the plasticity index, the wider the spacing expected between ε = constant normally consolidated one-dimensional compression lines (isotaches) in the e versus σ' v (log) plot. Therefore, the dependence of σ' p on the strain rate is expected to be higher, the higher the clay plasticity. This also suggests that there is a viscous  Terzaghi (1941), Taylor (1942) and Taylor (1948, pp. 245) (see also Lima, 1993;Garcia, 1996;Santa Maria, 2002;Aguiar, 2008;Andrade, 2009). Nevertheless, a detailed discussion on this subject is out of scope of this article and will be presented in another article where the long-term loading stages run to investigate secondary consolidation and stress relaxation will be shown.
The ratio between C c from the ε = 10 -6 s -1 compression curve, denoted by C c (10 -6 s -1 ), and C c from the 24-hour compression curve, denoted by C c (24 h), is within 0.94 and 1.08, with an average of 1.02. The ratio between C r from the ε = 10 -6 s -1 compression curve, denoted by C r (10 -6 s -1 ), and C r from the 24-hour compression curve, denoted by C r (24 h), is within 0.76 to 1.29, with an average of 1.02 (value of 0.59 not included). A practical conclusion is that it is possible to reduce the total duration of a consolidation test from ten to about three days by using the ε = 10 -6 s -1 loading criterion without changes in C c and C r values.

Comparison between the SFL clay layer compressibility parameters obtained in this study and by Massad (2009)
Since the Massad (2009) compressibility parameters are interpreted as having been obtained from 24-hour compression curves, only the compressibility parameters obtained in the same way are considered for comparison purposes. Table 7 shows the ranges of SFL clay compressibility parameters presented by Massad (2009, Tables 5.1 and 5.2) and those obtained from series two specimens from samples SRA203(4) to SRA203(10), disregarding specimens CP5G, CP10D and CP10F, classified as "fair to poor" according to Coutinho (2007) modified criterion.
The σ' p , OCR and C r /C c obtained in this study are within the ranges presented by Massad (2009). However, the lower and upper bounds of the C c /(1+e 0 ) range in this study are higher than those presented by Massad (2009), with the average in this study being higher than the upper bound of the Massad (2009) range.
The series two specimens of samples SRA203(6) and SRA203 (7) showed C c /(1+e 0 ) within 0.60 and 0.68, whereas all the other series two specimens from samples SRA203(4) to SRA203(10) showed C c /(1+e 0 ) within 0.46 and 0.59 (average of 0.52). Nevertheless, even excluding samples SRA203(6) and SRA203 (7), the C c /(1+e 0 ) values are still higher than the Massad (2009) values. Since disturbance decreases the compressibility in the virgin compression domain, Massad (2009) specimens seem to be of poorer quality than the ones studied herein, which is corroborated by the straight shape of the virgin compression lines shown by Massad (2009, Figures 5.43, 5.45 and 5.46), a disturbance effect also discussed in section 4.2.
It must be pointed out that Santos soft clay compressibility data available in the literature were mainly obtained before the nineties, when sampling standards and testing procedures were different from the current ones.
Unfortunately, in civil engineering practice, even today, sampling and testing procedures do not usually receive due care recommended by current standards. The authors' intention is to highlight the importance of following rigorously the current standards as well as special technical specifications (see Ladd & DeGroot, 2003) in order to obtain better-quality results. Table 6. Ratio between σ' p and compressibility parameters from ε = 10 -6 s -1 and 24-hour compression curves of series two specimens from the SFL clay layer.   Figure 11 shows the c v average values profile in the recompression (between σ' v0 and σ' p ) and virgin compression domain of all undisturbed specimens. Except for sample SRA203(2), which is sand, for all specimens, c v values in the recompression domain are higher than those in the virgin compression domain. The sandy specimens, which do not belong to the SFL clay layer, showed smaller differences between c v values from the two domains than the SFL clay specimens.

Coefficient of consolidation
The SFL clay specimens showed c v values in the recompression domain within 3.0 x 10 -7 m 2 /s and 2.5 x 10 -6 m 2 /s. In the virgin compression domain, c v values are within 7.0 x 10 -9 m 2 /s and 5.0 x 10 -8 m 2 /s, the values between 1.0 x 10 -8 m 2 /s and 2.5 x 10 -8 m 2 /s being more frequent.

Conclusions
1) The stratigraphy of the Santos soft clay deposit near Barnabé Island follows the genetic pattern described by Massad (2009). 2) Following ABNT (1997) and additional cares in sampling, transportation, storage and specimen trimming (Aguiar, 2008;Andrade, 2009), high-quality one-dimensional consolidation test specimens were obtained. 3) Comparison between undisturbed and remolded specimen compression curves evidenced all the remolding effects described by Ladd (1973), Coutinho (1976) and Martins (1983). 4) In the authors' experience with highly plastic clays, incremental loading one-dimensional consolidation tests, which usually last ten days adopting 24-hour loading stages on double drained 20 mm-high specimens, are reduced to three days by starting a new loading stage whenever ε =10 -6 s -1 .

5)
Series two tests showed 24-hour "e versus σ' v (log)" curves displaced to the left of the ε = 10 -6 s -1 "e versus σ' v (log)" curves, keeping C r and C c average values unchanged. 6) For the Santos soft clay studied herein, σ' p from 24-hour compression curve is about 8% lower than σ' p from ε = 10 -6 s -1 compression curve, confirming that σ' p depends on strain rate. 7) SFL clay C c /(1+e 0 ) values of this study are higher than those presented by Massad (2009). Since disturbance decreases the compressibility in the virgin compression region, Massad (2009) specimens seem to be of poorer quality than the ones studied herein. 8) It is feasible to carry out a high-quality laboratory test program for design purposes following current standards rigorously.