Geological perspectives on consolidation of clay-rich marine sediments

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 18, NO. B4, 2197, doi:1.129/21jb652, 23 Geological perspectives on consolidation of clay-rich marine sediments D. E. Karig Department of Earth and Atmospheric Sciences, Cornell University, Ithaca, New York, USA M. V. S. Ask Division of Rock Mechanics, Luleå University of Technology, Luleå, Sweden Received 29 May 21; revised 8 November 22; accepted 29 January 23; published 12 April 23. [1] Experimental consolidation of uncemented clay-rich marine sediments provides information concerning their stress history. A main finding is that some of the well-known behavior of soft sediment deformation in geotechnical applications cannot validly be extrapolated to sediments that have been subjected to higher stresses and longer times of geologic conditions. This study confirms that the yield stress of the uncemented sediment accurately reflects its previous maximum consolidation state. Furthermore, we have identified a new phase of post-yield strain that is associated with higher values of the modified compression index (the slope of the porosity versus the logarithm of effective vertical stress) than that of elastic deformation, but with much lower values than that for primary consolidation. This post-yield behavior is a linear, non-elastic deformation, and is termed tertiary consolidation. Yield stress appears independent of creep time or strain rate, whereas the tertiary-primary consolidation transition is sensitive to these parameters. During post-yield creep (secondary consolidation) the slope of the porosity versus the logarithm of time curve, or the secondary consolidation index, is generally assumed constant. However, this is not valid for claystones at effective vertical stresses above about 1 MPa, where the secondary consolidation index increases with stress. At a given effective vertical stress, the secondary consolidation index also increases with creep times greater than about 1 5 s(28 h). INDEX TERMS: 512 Physical Properties of Rocks: Plasticity, diffusion, and creep; 5114 Physical Properties of Rocks: Permeability and porosity; 5199 Physical Properties of Rocks: General or miscellaneous; KEYWORDS: sediment consolidation, marine clay, sediment creep Citation: Karig, D. E., and M. V. S. Ask, Geological perspectives on consolidation of clay-rich marine sediments, J. Geophys. Res., 18(B4), 2197, doi:1.129/21jb652, Introduction [2] Consolidation of sediments in the geological regime differs from most testing in geotechnical engineering laboratories in several respects. Not only does effective vertical stress (s v ) increase to much higher levels in sediments but the natural rate of consolidation in these materials is much slower. Lithification of sand- and clay-rich sediments is dominated by mechanical consolidation in the upper several km of a basin. Geologists generally use consolidation tests to constrain a present or past in situ stress state for sediments between depths of 1 m and about 3 km, which corresponds to a range of s v between 1 MPa and 3 MPa assuming hydrostatic pore fluid pressures. [3] Strain in sediments generally, and compactive strain in particular, is a function of both stress and time. Because of the intimate association of time and stress in compactive strain, the entire process of time dependent strain is termed consolidation in this paper. Primary consolidation occurs Copyright 23 by the American Geophysical Union /3/21JB652$9. during continuous increase of stress or strain such that the internal pore pressure is just being dissipated (Figure 1). Secondary consolidation here refers to the time dependent strain following pore pressure dissipation at a constant effective stress and is a type of what geologists refer to as creep. [4] Our studies show that a third type of consolidation occurs at stresses past the previous maximum consolidation state, which has not previously been recognized. Laboratory tests show that an increase of stress following a period of secondary consolidation results in a compactive phase during which the material shows a smaller compressibility than during primary consolidation, following which the compressive response rejoins the primary consolidation curve. This initial compactive phase is usually implied to be an elastic phenomenon [e.g., Imai, 1995] and to be part of simple recompression. The return to primary consolidation is generally interpreted as a yield and apparent preconsolidation stress. The degree to which this yield stress differs from the original consolidation stress and how it is affected by various factors has been a subject of ongoing concern in the geotechnical engineering literature [e.g., EPM 7-1

2 EPM 7-2 KARIG AND ASK: GEOLOGICAL PERSPECTIVES ON CONSOLIDATION OF CLAYS equivalents, with significantly larger compressibilities for the disaggregated versions. These tests were run to explore these differences and how they might relate to secondary consolidation during natural sedimentation. [7] Additional objectives were to: (1) investigate the relationship between the modified consolidation index (C ah ) and stress level, especially at stresses of several to tens of MPa, as occur in many geological situations; (2) test whether secondary consolidation is linear with respect to the logarithm of time at higher stresses, especially as time increases beyond that generally explored in laboratory experiments; and (3) generate porosity-effective vertical stress (h-s v ) curves for comparison with in situ porositydepth curves. 2. Background [8] Sediment strain (e) has often been assumed [e.g., Mesri and Choi, 1985] to be a function of stress (s) and time (t) in the form: Figure 1. Schematic view of strain as a function of the logarithm of stress. Primary consolidation occurs along sections A-B and F-G. Secondary consolidation is time dependent strain that takes place at a constant stress (section B-C). Unloading and reloading of a sample (i.e., sections C-D and D-E, respectively) is an elastic phenomenon, and the modified compression index (C ch s) isthe same for unloading and reconsolidation until the previous maximum consolidation state is reached. Tertiary consolidation (section E-F) occurs when the stress on a sample exceeds the previous maximum consolidation state and continues until primary consolidation is resumed (section F-G). Bjerrum, 1967; Leroueil, 1996]. Our studies show that this initial post-yield consolidation has a higher compressibility than does the elastic strain and must be treated separately. We have termed this phenomenon tertiary consolidation and describe it in this paper (Figure 1). [5] Consolidation is further complicated by a process variously termed cementation, structuration, bonding, or diagenesis [e.g., Bjerrum, 1973]. This process imparts a component of strength to the sediment in addition to that due to intergranular friction and to an indeterminate extent opposes the effects of secondary consolidation. Effects of cementation are most clearly shown on differential stress versus effective mean stress (s-s m ) plots by a rise in s above the bilinear curve observed in tests on uncemented sediments [Leroueil and Vaughan, 199; Karig, 1993]. [6] In this report consolidation tests were run on sediments both with and without cementation, but unless noted the data and discussion pertain to uncemented samples. The first investigations began with one-dimensional (uniaxial) compression tests on undisturbed and uncemented marine sediment cores. Objectives of these earlier tests included yield stresses (i.e., estimated effective in situ stress) and typical geotechnical parameters such as compressibility (C c ). Additional testing followed the recognition of different compressibilities between natural and their disaggregated de ¼ ð@e=@sþds þ ð@e=@tþdt ð1þ The terminology used to describe the stress and time components of this strain differs between geotechnical engineers and geologists. Following geotechnical terminology, compactive strain with reduction of porosity is termed consolidation. Most workers define primary consolidation as the strain associated with the dissipation of excess pore water pressure during application of a discrete increase of stress. The end of primary consolidation (EOP) is usually marked by a break in slope on a strain versus logarithm of time plot [e.g., Lambe and Whitman, 1969]. On the other hand, the geotechnical terminology describing subsequent time components of compactive strain is confusing, and in some respects contradicts geological terminology. Time dependent compactive strain occurring during primary consolidation has been termed secondary consolidation [Leonards and Deschamps, 1995], whereas that continuing after pore pressure dissipation is termed secondary compression by most geotechnical engineers. However, compression is a term associated with stress in geological literature. Because of the intimate association of time and stress components in compactive strain, the entire process of time dependent strain is termed consolidation in this paper. [9] Laboratory studies have suggested that there is an approximate linear relationship between some parameter of strain and the logarithm of effective stress during primary consolidation (Figure 1). Most geotechnical engineers describe this linearity using void ratio (e) versus the logarithm of effective vertical stress (log s v ) plots, on which the slope is termed the compression index (C c log s v ). However, over the wider stress range that is more appropriate to geologic settings, porosity (h)-log s v plots appear to be more nearly linear [Karig and Hou, 1992]. In this paper the slope is termed the modified compression index (C ch log s v ), which is analogous to the geotechnical usage of C c. Equation 1 leads to the Bjerrum plot of e-log s v [Bjerrum, 1973], on which consolidation at various constant strain rates or creep for similar times describe parallel straight lines with slope C c [Bjerrum, 1967]. The results of this study indicate that C c or C ch is

3 KARIG AND ASK: GEOLOGICAL PERSPECTIVES ON CONSOLIDATION OF CLAYS EPM 7-3 constant neither with stress or time at higher stresses or longer times. [1] It has also been generally accepted that, at stresses above yield, secondary consolidative strain is linear with respect to log t, at least in the lower stress regime of soils mechanics (1 MPa). The secondary consolidation index (C a ) is assumed to be constant not only with respect to time but also to be independent of s v (C a t). Here again porosity instead of void ratio is used in this study (C ah t). With this convention equation (1) can be written as: dh ds v þ Þdt ð2þ The origin of time (t) is taken at the time of application of an increase in s v. This is justified by the decrease in porosity at a given s v as the strain rate during consolidation decreases, which indicates that primary and secondary consolidation occur simultaneously. In the tests of this study that employed step increases of s v, these steps were large enough to absorb effects of secondary consolidation from previous steps during the primary consolidation phase. C ch and C ah can be reformulated in terms to obtain equations for porosity change in terms of observable parameters: or dh ¼ C ch =sv ds v þ C ah=t dt ð3þ dh=dt ¼ C ch =sv ds v =dt þ C ah=t Rates of natural consolidation are controlled by the rate of sedimentation and pore pressure dissipation. Most basins show near-hydrostatic pore pressures in the upper few km [e.g., Powley, 199]. Thus, constant rates of deposition at the sediment surface would lead to consolidation at a constant rate of stress increase, with the strain rate for a sediment element decreasing with increasing time and stress as it becomes successively more dewatered and deeply buried. Even for consolidation due to the most rapidly deposited sediments, initial strain rates are far slower than those generated in the laboratory. A sedimentation rate of 1 km/my would produce an initial strain rate of less than 1 14 /s in a clay and that rate decreases with depth. At such rates Equation 4 would suggest that secondary consolidation is far advanced as consolidation proceeds. 3. Test Program 3.1. Materials [11] The sediments used were all clay-rich marine sediments collected at depths below the sediment-water interface of a few to over 2 m. Two suites of samples were deep ocean sediments from Ocean Drilling Program (ODP) drill cores representing quite dissimilar environments. Sediments from a deep borehole in the Gulf of Mexico and glaciomarine sediment from Norway were used as highly disturbed or disaggregated material for primary and/or secondary consolidation tests. [12] The first set of samples used for this study came from drill cores from ODP Site 88 in the Nankai Trough, an oceanic trench south of Japan and from sediment depths ð4þ from about 11 m [Karig, 1993]. These Late Miocene silty mudstones were hemipelagically deposited in the oceanic Shikoku Basin and consist of 3 4% clay minerals, with clays dominated by illite (Figure 2 of Karig [1993]). The carbonate content was very low. Samples had initial porosities between 31 and 33%. Other physical properties are listed in Karig [1993]. [13] A second suite of samples was obtained from Core D-3R from ODP Site 897 at a depth of 619 m below the seafloor. This site was located on the Atlantic abyssal plain off Portugal. The sampled core was a mudstone of middle Eocene age, and, in contrast to the Nankai mudstone, has an estimated 45% carbonate content [Shipboard Scientific Party, 1994]. Smear slide analysis indicates that most of this carbonate is in the form of nannofossils, primarily coccoliths, and that there is a very low ratio of silt to clay, most of which was illitic [Alonso et al., 1996]. Bulk densities of these subsamples, measured in our laboratory from sample volumes and wet weight, varied between 2.18 to 2.22 g/cm 3, with a strong mode near 2.2 g/cm 3. With a grain density of 2.78 g/cm 3 [Shipboard Scientific Party, 1994] these samples have a calculated porosity of 33%. [14] In addition, a non-calcareous silty clay from 2385 m below the seafloor in the Pennzoil Pathfinder well in the northern Gulf of Mexico was used, as disaggregated material, for primary consolidation and creep tests. The Pathfinder well was a commercial borehole onto which a large research project was piggybacked [Hart et al., 1995; Flemings et al., 1994]. This project provided in situ porosity and values, with which the experimental data could be compared. [15] Finally, two samples of Glava clay from Vaernes, Norway were used as material for creep tests. This glaciomarine silty clay is naturally quite highly cemented (i.e., sensitive) and its properties have been well studied [e.g., Senneset et al., 1989]. The samples used in this study were from a depth of 12.5 m with a porosity of 47%, but because the samples were tested at stress levels many times greater than the preconsolidation stress, the properties of the undisturbed samples were completely overprinted. [16] Disaggregated and slurried sediment samples were created from remnants of the cores from which the undisturbed samples were taken. These disaggregated samples were assumed to mimic the original character of the sediment during deposition, but a comment concerning this assumption and on the method of disaggregation is warranted. Disaggregation of the natural sediment samples to a state that resembled the sediment at the time of deposition could only be approximated because of difficulties in regaining the original grain size distribution and because of possible diagenetic changes that might have occurred after sedimentation. The latter was considered to be limited to minor smectite-illite transformation in the samples from the Nankai Trough [Underwood et al., 1993]. [17] After drying at 15 C, residue from the cores was crushed, first in a jaw crusher and then in a roller mill to pieces about.5 mm or smaller. The pieces were then placed in a ceramic shatter box, in which the material could be milled. The time in the shatter box was adjusted such that the mean grain size was near 3 mm, which entailed destruction of clay aggregates and produced mostly single clay crystals. Results were checked optically with smear slides. s v

4 EPM 7-4 KARIG AND ASK: GEOLOGICAL PERSPECTIVES ON CONSOLIDATION OF CLAYS Longer milling was avoided in an attempt to minimize the comminution of silt grains, although this was only partially successful. In spite of the problems, the disaggregated material, especially that of the less silty clay from Site 897, was quite similar to the original sediment. [18] Slurries of this material were made using distilled water containing 2 ppt NaCl. This concentration of salt, together with the salt remaining in the dried sample, was calculated to approximate the in situ salinity of the undisturbed sediment Experimental Procedures [19] The test program included experiments done with one of several testing systems. Uniaxial strain tests (alsocalled reconsolidation tests) on undisturbed ODP samples were run in a triaxial cell mounted in a computer-controlled servo-hydraulic load frame [Karig, 1996]. Lateral displacements were monitored with an array of eight linear variable displacement transducers (LVDT s) mounted in a ring around the sample midsection. Lateral strains were kept to less than.1% with a pseudo-derivative algorithm that continuously sensed and modified confining pressure and axial load. Axial strains were measured over the central section of the sample with a pair of LVDT s and volume strains were calculated both by dimensional changes and by pore water volume changes. [2] Test samples from the ODP sites were cored from sections of drill core having a diameter of about 6 mm. Because of the need to have sub-samples with both vertical and horizontal axes for other studies, most test samples were cylinders with radii of 1 mm and lengths between 5 and 55 mm. Other test samples, without such constraints had radii of 15 mm and lengths of 6 to 65 mm. Samples were isolated from a silicon oil confining fluid with 1.18 mm thick latex jackets and all samples were drained from both ends through sintered titanium filter disks. [21] Most reconsolidation tests on undisturbed sediment samples in the triaxial cell were performed as constant rate of effective vertical stress increase (7 1 4 MPa/min). Stress holds demonstrated that very little excess pore pressure was developed at this rate. Note that excess pore pressure for these tests implies pressures exceeding atmospheric pressures, and not like in situ conditions where excess pore pressure implies higher than hydrostatic pressures. The stress holds did incorporate secondary consolidation, and longer stress holds served to provide data on secondary consolidation. [22] Consolidation tests were also made on disaggregated and slurried sediment. All slurries were consolidated, at least initially, in a 77 mm diameter oedometer [Karig and Hou, 1992], most in constant load steps of 24 h duration (usually with a load step ratio, s/s, of 1). Others were loaded at a constant rate of s v increase of.14 MPa/min. This oedometer was mounted in a servo-mechanical balance-beam constant load testing machine for constant loadstep tests and in the servo-hydraulic machine for constant rate of stress tests. The differences in h-s v relationships between tests using step load and constant stress rate tests were very small (Figure 2). [23] Some of the 77 mm diameter samples were subsampled for further consolidation in the triaxial cell in order to take advantage of the greater accuracy and flexibility of that system. These slurried samples were initially long enough such that their final length was at least 6 mm. Other samples were of such initial lengths that their length after consolidation was as short as possible (25 mm), in order to minimize friction between the sediment and the oedometer walls. In all tests cell walls were lubricated with MoS 2 and differences in length had insignificant effects on h-s v relations [Karig and Hou, 1992]. [24] Secondary and tertiary consolidation test phases on the ODP samples were incorporated into the consolidation tests during axial stress holds of up to 3 h (1 5 s). Three long ( s) secondary consolidation tests on disaggregated samples of clay from ODP Site 897 and one from the Pathfinder well were run at different stress levels in the 77 mm oedometer mounted in the balance beam machine. Two secondary consolidation tests on Glava clay were run at the Geotechnical Institute of the Norwegian Technical University; the lower stress test (1 MPa) in a standard oedometer and the higher stress test (1 MPa) in a special split ring oedometer mounted in a constant load hydraulic press [Senneset, 1989]. 4. Results 4.1. Consolidation of Undisturbed Natural Sediments [25] All tests on undisturbed calcareous claystone samples from ODP Site 897 showed a simple response, with a gradual but obvious yield condition separating an elastic regime from a post-yield regime that has a different but constant stress ratio and a linear h-log s v relationship (Figures 2, 3, and 8 of Karig [1996]). No test displayed effects of cementation, in their s/s m plots. The change in response with increasing s v that led to greater compressibility, a lower s/s m gradient, and a higher differential stress ratio (Figure 3a) was interpreted as a yield stress by analogy with many other geotechnical tests. Moreover, unload-reload cycles run at stresses above yield stresses produced nearly linear and reversible stress-strain relations with C ch values significantly lower than the post-yield C ch (Figure 4). [26] Yield stresses (s y ) for all ODP samples in this study were much lower than the in situ s v values that would be expected if in situ pore pressures were hydrostatic [Karig, 1993, 1996]. However, the interpretation that these low yield stresses do represent preconsolidation stresses was supported by other geological evidence, which indicated that in situ pore pressures were very high. Samples from ODP Site 88 were near a major shear zone, which was suspected to be a permeable corridor for high-pressure fluids, as evidenced by unique chemistry and sediment properties [Hill et al., 1993]. The zone from which the core at ODP Site 897 was recovered had anomalously high porosity [Sawyer et al., 1994]. Moreover, the underlying material was intensely fractured and veined, symptomatic of rising high-pressure fluids [Karig, 1996] Pre-Yield Elastic Reconsolidation of Undisturbed Sediment [27] Mechanical parameters that have been used to diagnose the elastic behavior during uniaxial strain tests are the modified compression index (C ch s v ) and stress ratio during uniaxial strain (K = s h /s v ).

5 KARIG AND ASK: GEOLOGICAL PERSPECTIVES ON CONSOLIDATION OF CLAYS EPM 7-5 Figure 2. Porosity (h) versus the logarithm of effective vertical stress (log s v ) during uniaxial consolidation of several clay-rich sediments for various consolidation conditions. The effective vertical stress was increased by either step load or constant stress rate in these tests. End of primary consolidation (EOP) porosities are nearly identical for step load and constant rate of stress increase tests on samples from ODP Site 897 and the Pathfinder well. The slope, C ch, is greater and similar for primary consolidation of disaggregated sediment from several localities. Values of C ch for the reconsolidation of undisturbed and uncemented samples are much lower. [28] The elastic cycles in this study were run only at stresses greater than the initial yield stress, i.e., the samples had already been subjected to at least one reconsolidation test to stresses above the in situ yield stress so that the yield envelope of the sediment had been expanded to a higher stress state. Tests were made on both a natural sediment sample from ODP Site 897 (Figure 4) and on a disaggregated sample of the same sediment that had been consolidated and allowed to creep for 34 d (3 1 6 s). For the elastic cycles on undisturbed samples from ODP Site 897, C ch values lay between.6 and.1. [29] K values for stress paths preceding initial yield were neither constant nor reflective of uniaxial elastic strain because the tests began at an isotropic stress state not far from s v at yield. There was a marked change in K at yield as defined by stress-strain and other relationships, but more reliable elastic K values were derived from the post-yield elastic cycles (Figure 5). The four values of K for pre-yield elastic strain cycles run on the ODP Site 897 samples ranged between.44 and Post-Yield Consolidation of Undisturbed Sediment [3] As the undisturbed sediment samples were reconsolidated, the response passed from elastic to one with greater compressibility. This transition was interpreted as the yield stress, as earlier justified. The post-yield consolidation showed a linear h-log s v response for at least 2 MPa beyond this yield stress, for which a well-defined value of C ch could be determined. [31] The values of post-yield C ch for undisturbed samples from ODP Site 897 were between.25 and.29 (Figure 2), with a mode toward the higher end of that range. These values are far higher than the values for C ch during the elastic cycles. The post-yield value of K also changed, to a value of.55 ±.2 (Figure 5), which is significantly higher than values for elastic cycles (.44.48). [32] A simple response across yield was also obtained from tests on undisturbed and uncemented sediment samples from below the décollement at ODP Site 88 [Karig, 1993]. These tests demonstrated that post-yield C ch was about.15 (Figure 2) Consolidation of Disaggregated Sediment [33] The h-log s v curves for the disaggregated material from ODP Site 897 are linear to at least 1 MPa (Figure 2), with values for C ch between.11 and.118. Corresponding values for disaggregated silty clay from ODP Site 88 and Pathfinder clay were.16 and.164, respectively. All these values of C ch are far greater than those for post-yield consolidation of natural samples from the ODP cores (Figure 2). [34] For the disaggregated samples from ODP Site 897, K ranged between.57 and.61 (Figure 5), which is similar to or possibly slightly greater than that for post-yield

6 EPM 7-6 KARIG AND ASK: GEOLOGICAL PERSPECTIVES ON CONSOLIDATION OF CLAYS Figure 3. Reconsolidation (reloading) of an undisturbed sample from ODP Site 897. A, Effective horizontal stress (s h ) versus effective vertical stress (s v ) showing the stress ratio (K ), and B, Porosity (h) versus the logarithm of effective vertical stress (log s v ) showing the modified compression index (C ch ). Stress ratios during elastic, preyield reconsolidation begin at an isotropic condition but drop to near.4 before increasing to.54 after yield (inset of Figure 3A). Following a mechanical problem and reequilibration at s v = 7.6 MPa, K averaged.56 ±., to values of s v well beyond 14 MPa, where C ch begins to increase from its post-yield value of.31 ±.1 toward that of primary consolidation Secondary Consolidation Tests [36] Information concerning secondary consolidation was derived from three types of tests: (1) constant stress holds during the consolidation of natural samples; (2) constant stress holds during the consolidation of disaggregated samples; and (3) long duration creep tests. In addition to providing parameters concerning secondary consolidation, a few of these tests also involved subsequent increase of s v, which gave information concerning tertiary consolidation. [37] Periods of constant stress during consolidation that were longer than that necessary to dissipate pore pressure provided values of C ah. All secondary consolidation tests were made at stresses above the yield stress of the undisturbed sediment to avoid any transient effects of cementation on values of C ch and C ah [Mesri and Castro, 1987]. [38] A total of 23 constant stress hold tests that were of sufficient duration to develop an apparent constant C ah were run on samples from ODP Site 897, the Glava clay, and disaggregated core from the Pathfinder well (Figure 6). Twenty stress holds on the disaggregated material from ODP Site 897 and the Pathfinder well, as well as those on the Glava clay, represent secondary consolidation of sediment that was at or near a state of primary consolidation and gave values of C ah from.1 to.8. Although there is a rather large scatter in these values, there is a strong indication of an increase in C ah with increasing stress level (Figure 6). Three stress holds during post-yield consolidation of undisturbed samples from ODP Site 897 showed values of C ah that were significantly lower than those of samples for which there was no earlier secondary consolidation (Figure 6). These three values also increase with increasing s v. [39] Much longer duration creep tests included those done on disaggregated samples from ODP Site 897 (Figure 7) and the Pathfinder well, as well as the two initial tests on the Glava clay. C ah for the test on Glava clay at 1 MPa remained approximately constant for s (6 d) of creep, consolidation of the undisturbed sediment. The relationship between these two ranges of values will be discussed more fully in the section on tertiary consolidation. [35] Consolidation tests on disaggregated sediments with elastic unload-reload cycles showed that yield stresses (s y ) were approximately identical to preconsolidation (s p ) values as generated by previous maximum s v (Figure 5). As defined by plots of e v -s v, s h -s v, and s-s m, s y differed from s p by up to.5 MPa, but in no systematic way. The difference reflects primarily the difficulty in determining s y during gradual yielding of these sediments. This observation supports the contention that s y for the undisturbed and uncemented samples also represents s p. Figure 4. Unload-reload cycles at stresses above the yield stress of undisturbed sediment from ODP Site 897 during post-yield consolidation. These cycles involve elastic strain and have much lower C ch values than during post-yield consolidation, which is interpreted to be tertiary consolidation.

7 KARIG AND ASK: GEOLOGICAL PERSPECTIVES ON CONSOLIDATION OF CLAYS EPM 7-7 Figure 5. Stress ratios for three samples from ODP Site 897, showing K, values for different consolidation conditions. Elastic strains produce significantly lower K values (.44 and.48) than do primary or tertiary consolidation, which appear to have approximately equal K values. For the two paths involving unload-reload paths, yield stresses (s y ) and previous maximum consolidation stresses (s p ) are approximately equal. The values of s y were derived from the relationships among all stress and strain test parameters. whereas that for the Glava clay test at 1 MPa showed a gradual but clear increase in magnitude after about s. [4] Creep tests on disaggregated material from ODP Site 897 (Figure 7) and the Pathfinder well were run at stress levels between 2.8 MPa and 19 MPa and for durations exceeding 1 6 s(12 d). Despite scatter in the porosities, traced to voltage instabilities in the recording system, all tests showed clear and large increases in C ah after about 1 5 s (Figure 7). During these tests C ah could be considered constant for at least 1 5 s, but C ah shows a large increase in the range of t between 1 5 s and s(1 35 d). There is some indication that C ah does not continue to increase for t greater than about s. For the three samples from ODP Site 897, C ah for this interval is.11 ±.2, about three times that for shorter t. The equivalent value of C ah for the Pathfinder sample was only.4, but again this was about three times the value for t less than 1 6 s(12 d). [41] Few and less reliable data were obtained concerning horizontal stresses during the creep phases of tests in this study, but there was evidence that s h changed little if at all during creep. Three creep periods of more than 75 h (3 1 5 s) during stress holds on the disaggregated sediment from ODP Site 897 showed both very small increases and decreases in effective horizontal stress (s h ), and these very likely reflected voltage drift in the sensor system. A s creep period, following primary consolidation at s v =.9 MPa, produced an increase of.2 MPa in s h. Creep of 1 7 s(116 d) in this sample at s v = 7.8 MPa led to a decrease in s h of less than.3 MPa and a reduction in K of less than.4. A creep phase in a second sample of the same material of more than 1 6 sats v = 4.88 MPa showed variations in s h of less than.3 MPa and in K of less than Tertiary Consolidation Tests [42] Data on tertiary consolidation are scanty, because even the longest periods of secondary consolidation in the laboratory resulted in too small porosity reductions to produce a post-yield deformation path that could resolve the difference between yield and the return to primary consolidation. Nevertheless, some data on tertiary consolidation can be derived from these tests because the initial C ch value for resumed loading after long stress holds approximates that of tertiary consolidation (Figure 1). [43] One reconsolidation test in the triaxial cell was made on a natural sample from ODP Site 897. This test reached

8 EPM 7-8 KARIG AND ASK: GEOLOGICAL PERSPECTIVES ON CONSOLIDATION OF CLAYS Figure 6. Variation of modified compression index (C ah ) with respect to effective vertical stress (s v ) for uncrept samples that had been consolidated only to end of primary consolidation (EOP), and for crept samples that had already undergone creep associated with natural sedimentation or experimentally subjected to significant creep. These data suggest that C ah increases with increasing s v for both crept and uncrept samples and that C ah is greater for uncrept than for crept samples. All C ah values are for time periods less than 1 5 s. yield at s v of about 1 MPa. The yield as marked by an increase in C ch to.3 that was approximately constant to about 1 MPa. At higher values of s v,c ch clearly increased, without any change in stress ratio (Figure 3). There was a clear tendency toward a primary consolidation response (C ch =.56) near the end of the test. The post-yield C ch value of.3 is our most direct evidence of tertiary consolidation. [44] Only two tests on disaggregated sediments provided reasonable estimates of C ch, for tertiary consolidation. The first estimate came from a consolidation test of material from ODP Site 897, following a 24 h stress hold at 1.64 MPa (Figure 2). This test was run at a rate that was slow enough to allow dissipation of excess pore pressure but fast enough to prevent much secondary consolidation. The second estimate came from a step load consolidation test of a sample from ODP Site 88. This sample was subjected to secondary consolidation at 4.76 MPa for 182 d (>1 7 s), followed by partial unloading and reloading to 19.4 MPa. The stress steps were small to better define the region at and just past yield and were of durations just greater than EOP. Resumption of stress increase after the stress holds in the two tests both produced a C ch value of.3. The stress ratios during both these tests changed at their yield stresses but then remained constant during further primary consolidation In Situ Porosity Versus Effective Vertical Stress (H-S v ) [45] Data on in situ h-s v relationships that can be compared with the results of laboratory consolidation are difficult to find. Lithologic variations, the paucity of pore fluid pressure data, and difficulties in reconstructing the stress history of the site from which the data were acquired all reduce the pool of reliable data. [46] One data set is available from the Pathfinder borehole [Hart et al., 1995]. This project provided data on porosity and s v from a clay-rich section of the borehole. Porosities, as determined from sonic well logs, were 5% or more lower than laboratory-determined values of core samples at equal effective stress conditions, probably because of the empirical constant used to convert sonic velocity to porosity, but the porosity-depth gradient, which is of primary importance, is considered reliable. The in situ h-log s v gradient from the well logs is nearly identical to that obtained from the EOP consolidation tests on disaggregated material from the Pathfinder core (Figure 8). [47] Porosity and s v data can also be derived from Deep Sea Drilling Project (DSDP) and ODP sites in clay-rich sediments of the deep ocean basins in which pore pressure can be reasonably assumed to be hydrostatic [Hamilton, 1976; Busch, 1989]. These data are largely restricted to depths below the seafloor of less than 1 km and to values of less than 1 MPa. Moreover, porosities are based on laboratory measurements on cores, corrected for stress reduction from in situ conditions. These corrections may be excessive [Karig and Hou, 1992] but both corrected and uncorrected porosities from Hamilton and Busch give C ch values near.2, which are even greater than the.16 value for the Pathfinder well. s v 5. Interpretation and Discussion [48] Despite the extensive literature on the consolidation of clays, there is still much to be learned, especially about

9 KARIG AND ASK: GEOLOGICAL PERSPECTIVES ON CONSOLIDATION OF CLAYS EPM 7-9 Figure 7. Porosity (h) versus the logarithm of time (log t) during secondary consolidation at different effective vertical stress (s v ) levels for three disaggregated sediments from ODP Site 897. The origin of time is taken at the initiation of the load step. In all tests, the modified compression index (C ah ) increases with logarithm of time beyond a time of about 1 5 s(28 h). aspects bearing on the geological conditions of high stress and very slow consolidation rates. Many aspects of geological consolidation seem to be reasonably well explained by geotechnical concepts, but the test results presented here indicate that some of these concepts require quantitative modification or serious reconsideration Tertiary Consolidation [49] Our test results indicate that post-yield consolidation for undisturbed samples of uncemented clay-rich sediments in the laboratory is tertiary consolidation. During reconsolidation tests on undisturbed, uncemented samples, tertiary consolidation is preceded by elastic strain, and the transition between these phases appears to be a yield stress with distinct increases in values of C ch and stress ratio (e.g., Figure 3). Moreover, the yield stress values obtained from these ODP sediments were repeatable, geologically reasonable, and could most reasonably considered to be preconsolidation stresses (s p ). This identification is supported by the similarity of s y and s p during reconsolidation of disaggregated sediments in the triaxial cell following consolidation in the 77-mm diameter oedometer, both with and without secondary consolidation. [5] This conclusion has been implied in the geotechnical literature [e.g., Bjerrum, 1967], but the relationship of this tertiary consolidation to yield and preconsolidation stresses, as indicated by this study, is quite different from that generally presented in the geotechnical literature. Djéran- Maigre et al. [1998] conducted consolidation tests on disaggregated and undisturbed samples of the La Bouzule clay, a highly consolidated illitic clay from the Paris basin. They obtained normal compressibility (C c =.48) for the disaggregated sample, whereas a lower value (C c =.19) than predicted was obtained for the undisturbed sample up to the load of 1 MPa, from which the samples presented similar C c values (Figure 2 of Djéran-Maigre et al. [1998]). Because no significant microstructural difference was noted between the two samples at 1 MPa, Djéran-Maigre et al. [1998] attributed the small C c value to the fact that the sample had been consolidated at 1 km depth. However, Figure 21 of Djéran-Maigre et al. [1998] offers an alternative interpretation, namely that the undisturbed sample yielded at about 1 MPa, followed by tertiary consolidation from 1 to 1 MPa, where primary consolidation was resumed. [51] Most, if not all, geotechnical consolidation tests identify only the transition from a single, elastic reconsolidation response to primary consolidation, which is termed a preconsolidation or apparent preconsolidation stress [e.g., Bjerrum, 1973; Wood, 199]. Differentiation of the yield stress from this transition stress is more obvious for more highly consolidated sediments than for most soils because of the greater difference in magnitude between these two stresses in the stiffer sediments. The differences in C ch from elastic strain to tertiary consolidation also become larger as the elastic moduli increase in the sediments. The different modes of geotechnical testing have also obscured this differentiation. Lateral stress is very seldom measured during oedometer tests, so the changes in stress ratio at yield cannot be recognized. Furthermore, the geotechnical tests that involve an increase of vertical stress after secondary consolidation bypass the condition that is here interpreted as elastic Model of Primary, Secondary, and Tertiary Consolidation [52] The results of the consolidation tests reported in this study support the general concept that some consolidation parameter remains constant with respect to log s v during primary consolidation, but, together with previous highstress tests [e.g., Karig and Hou, 1992], suggest that porosity produces a more nearly constant relationship than does void ratio or strain. [53] Equations (3) and (4) imply independence of the effects of time and stress on dh or dh/dt. Using void ratio (e), Bjerrum [1967, 1973] and others describe consolidation behavior with e-log s v plots on which consolidation paths at equal t or de/dt are parallel, quasi-straight lines that are equally spaced for each decade of t or of de/dt (Figure 1 of Bjerrum [1973]). If, as this test program has suggested, C ch

10 EPM 7-1 KARIG AND ASK: GEOLOGICAL PERSPECTIVES ON CONSOLIDATION OF CLAYS Figure 8. Comparison of experimental consolidation of a core sample from and in situ conditions in the Pathfinder well [this study; Hart et al., 1995]. The undisturbed natural sample showed moderate cementation on s/s m and other plots. Disaggregated material from the core sample was consolidated in stress steps, with EOP defining primary consolidation. The best fit to the EOP points has the same slope as the in situ porosity (h)-log s v curve [Hart et al., 1995]. Some of the difference in h between similar experimental and in situ stresses is quite likely due to the constants used in determining in situ h from sonic logs. One disaggregated subsample of the core was allowed to creep at a s v of 2.9 MPa for 1 7 s during which time the porosity decreased by only.7%. increases with time and also stress, the Bjerrum plot must be modified, with lines of equal t or dh/dt diverging with increasing log s v and becoming more widely separated for each decade of change in t or dh/dt (Figure 9). Thus neither C ch nor C ah would be constant, and C ch would be a function of both time and stress. Figure 9, constructed using data from tests on samples from ODP Site 897, suggests that the effects of increasing t and higher s v on C ch are modest even at geological conditions of consolidation. The assumption of independent time and stress effects still seems adequate for most problems in soil mechanics. [54] A Bjerrum-type plot is also a very useful construct with which to describe the behavior of an uncemented sediment during both geological and experimental consolidation. Two stress paths of particular interest for these consolidative phases are loading during geological sedimentation and the reloading in the laboratory under uniaxial strain conditions. [55] Figure 1 of Bjerrum [1973] implies primary consolidation during sedimentation and burial, followed by secondary consolidation at some constant burial stress. A more realistic assumption is that sediments consolidate so slowly, even during rapid deposition, that a very large fraction of secondary consolidation occurs simultaneously with burial. [56] This model of simultaneous primary and secondary consolidation during continuous sedimentation (Figure 9) was developed from the experimental consolidation of disaggregated sediment from ODP Site 897 but could be tested by a comparison of the experimental results with in situ porosity-stress data. Unfortunately only a few partial comparative data sets are available, and such a comparison involves the extrapolation of the experimental secondary consolidation data for many orders of magnitude of time. Even so, several useful insights can be gleaned from this exercise Comparison of H-S v Curves and C ch Values for in Situ and Laboratory Data [57] The only in situ stress data from ODP Site 897 comes from the reasonable assumption that the yield stress (s y ) of the undisturbed samples represented preconsolidation (s p ) as well as in situ (s v ) conditions. If so, h between in situ and primary consolidation at 1 MPa is about 7% (Figure 9), whereas the extrapolated h, assuming a constant value of C ah after 1 6 s, would be at least 1%. A similar comparison for the uncemented Nankai section shows a 16% difference between in situ and primary consolidation of the disaggregated version of that sediment [see Karig, 1993]. [58] These preliminary comparisons indicate that in situ porosity for uncemented clay-rich sections is less than that for primary consolidation at equal s v, but may be greater

11 KARIG AND ASK: GEOLOGICAL PERSPECTIVES ON CONSOLIDATION OF CLAYS EPM 7-11 Figure 9. Porosity (h) versus the logarithm of effective vertical stress (log s v ) relationships during geological and experimental consolidation of an uncemented clay-rich sediment as proposed using the results of tests on samples from ODP Site 897. The increase in C ah with t and with s v leads to C ch lines that increase in spacing with t and in slope with increasing s v. Primary and secondary consolidation proceed simultaneously during sedimentation and burial, at a progressively decreasing strain rate. Unloading during sampling and reloading to the in situ s v involve elastic strain, and there is a welldefined yield stress at that point. Continued reloading results in tertiary consolidation until the deformation path reaches the h-log s v curve representing the strain rate of the test. than that predicted for disaggregated samples after secondary consolidation to 1 12 s assuming a constant C ah after 1 7 s. This suggests that C ah does not continue to increase for t > 1 7 s. [59] Porosity differences between in situ and primary consolidation also reflect the degree of cementation; the greater the degree of cementation, the greater the porosity at equal s v. On the basis of data from the Pathfinder well, in which the clay-rich core sample appeared moderately cemented, the in situ value of C ch was nevertheless quite similar to that for primary consolidation (Figure 8). [6] Reconsolidation tests on both undisturbed and disaggregated samples from the three marine sites also provide cautionary suggestions for the interpretation of common laboratory tests. For uncemented sediments that, have been naturally consolidated to values of s v well above 1 MPa, laboratory reconsolidation tests resulted in a tripartite h-log s v relationship (Figure 9). Each of the three phases (elastic, tertiary consolidation and primary consolidation) must be correctly identified and characterized. The yield stress does not appear to be affected by strain rate effects but is highly sensitive to cementation. [61] Continued consolidation in the laboratory results in tertiary consolidation, the slope of which may be dependent on strain rate. The data here do not show any such effects, but this may have been due to the small range of loading rates. Neither is there evidence that C ch during tertiary consolidation is a function of the degree of secondary consolidation. Our results show that primary consolidation of disaggregated material results in far greater values of C ch than does tertiary consolidation in undisturbed samples of the natural material. However, there is no large difference in K values between the two post-yield mechanical states. This is not a function of different porosities or stresses because C ch for primary consolidation is constant over a very wide range of porosity and stress [e.g., Karig and Hou, 1992]. The transition from tertiary to primary consolidation is gradual and strain rate dependent, and it seems to be this transition that has been the focus of much attention in geotechnical studies [Leroueil, 1995; Mesri and Choi,

12 EPM 7-12 KARIG AND ASK: GEOLOGICAL PERSPECTIVES ON CONSOLIDATION OF CLAYS 1985]. From a geological perspective it is most important to realize that this transition does not mark a prior consolidation state. [62] The recognition of cementation is also critical in that the behavior of cemented sediments during reconsolidation tests is very different from that of uncemented samples. Cementation not only causes the yield stress to exceed that of an uncemented sediment previously consolidated to the same s v, but also results in a stiffer elastic response, as shown by increased E c values (or decreased swelling indices)[burland, 199]. Such a cemented sediment will also have a higher porosity at yield than will its uncemented equivalent [Leroueil and Vaughan, 199]. [63] If s v is increased beyond yield during a reconsolidation test, the behavior of the cemented and uncemented sediment also differs radically. As shown in this study, uncemented sediments undergo tertiary consolidation, followed by a transition to primary consolidation. This occurs with a constant stress ratio, giving rise to a linear s/s m relationship. On the other hand, cemented sediments show a post-yield decrease, or at least inflection, in s on s/s m plots, followed by a transition directly to primary consolidation (e.g., Figure 3 of Karig [1996]). This reflects rapid porosity loss related to a highly time-dependent collapse of structure [Leroueil and Vaughan, 199]. The two post-yield phenomena thus require different modes of interpretation Attributes of Secondary Consolidation and Stress Ratio During Creep [64] Measurement of s h during uniaxial consolidation tests proved to be extremely valuable in deciphering secondary consolidation, as well as providing insights concerning the behavior of the stress ratio during the natural consolidation of clay-rich sediments. [65] The debate over the values and progressive history of stress ratio over geological time, reviewed by Karig and Morgan [1994], pits a group that holds that clays in particular creep toward an isotropic stress state [e.g., Warpinski et al., 1985] against those who argue that sediments can support large differential stresses indefinitely [e.g., Evans, 1989]. A major difficulty in resolving this problem arises from the lack of control on the stress history at most study sites, which can radically affect the measured stress ratio. The results of this study indicate that the stress ratio remains nearly constant for secondary consolidation of at least 1 7 s, which adds some support for the maintenance of differential stress over geological time. Some of the best recent in situ stress ratio data, from boreholes on the Scotian shelf [Yassir and Bell, 1994], where strain appears to have been uniaxial or with slight lateral extension, produced stress ratio values in mudstones near.5 to a depth of nearly 4 km. [66] There is also some debate concerning the stress ratio during the secondary consolidation of soils [e.g., Schmertmann, 1983], but most workers think that s h increases with time [e.g., Leroueil and Marques, 1996], at a logarithmically decreasing rate [Mesri and Castro, 1987]. The data presented by Mesri and Castro [1987] indicate that the rate of increase in s h was such that the stress ratio would not approach 1. even over geological time, but the increase over 1 7 to 1 8 yr would certainly be large enough to be observed by in situ stress measurements. The tests of Mesri and Castro [1987] were at very low stress (<.5 MPa) on undisturbed clays, at creep stresses not much beyond their yield stresses. It might be questioned whether the effects of destructuration may not have had an effect, as was seen on cemented clays at higher stresses [Karig, 1993, Figure 3] Speculation on Physical Mechanisms Responsible for Secondary and Tertiary Consolidation [67] Results of the consolidation tests at higher stress levels offer a different, more geological perspective on the still poorly understood physical mechanisms responsible for secondary consolidation. The test results indicate that the amount of secondary consolidation over a given time increases with increasing s v, which seems counterintuitive if extrapolated to very high stresses and to porosities approaching zero. Thus, secondary consolidation does not appear to be a simple dissipative process like pore pressure reduction during primary consolidation, but rather to be more a process similar to pressure solution. [68] Pressure solution is a widely recognized geological process operative at moderate (diagenetic-anchimetamorphic) temperatures and pressures. Pressure solution transfers mineral mass from grain contact points under higher normal stress to areas of lower stress, which are often adjacent pore space [e.g., Rutter, 1983; Passchier and Trouw, 1996]. Effects of this intercrystalline diffusive mass transport include compactive strain in the direction of maximum effective stress and porosity reduction. Experiments indicate that the rate of pressure solution is only mildly dependent upon temperature but increases markedly with decreasing grain size and increasing stress [e.g., de Meer and Spiers, 1997]. Pressure solution occurs in a wide range of lithologies, most notably in carbonates and quartz sands but also in clay-rich sediments [Rutter, 1983]. Several mechanistic similarities between pressure solution and secondary consolidation support the speculation that the latter is a result of the former, at least at higher stresses. Most striking is the increase in both pressure solution and secondary consolidation with increasing stress. This has been experimentally documented for pressure solution of quartz sand [Elias and Hajash, 1992] and gypsum [de Meer and Spiers, 1997] and qualitatively for and clay [Rutter, 1983]. [69] Both processes decrease with time at constant stress, but with less than exponential decay. Replotted data for pressure solution creep on quartz and gypsum show curves of porosity versus the logarithm of creep time that are very similar to those of secondary consolidation (Figure 1). For pressure solution this decay is attributed to increased intergranular contact and concomitant reduction of intergranular stress. The same argument could explain why crept sediment has less additional creep than do sediments that begin creep at the end of primary consolidation. [7] Tertiary consolidation is shown in this study also to be non-reversible, but with a stiffer response than during primary consolidation, although it appears to be associated with the same stress ratio. It could be speculated that increased stress during tertiary consolidation results in modification of grain fabric until the effects of pressure solution are destroyed by grain slip and rotation, after which a primary consolidative response resumes. [71] No information concerning K during experimental pressure solution was found, but if pressure solution is a