SOIL LIQUEFACTION DURING EARTHQUAKES PP) SSS) R. W. BOULANGER ee EARTHQUAKE ENGINEERING RESEARCH INSTITUTESOIL LIQUEFACTION DURING EARTHQUAKESMNO-1 MNO-2 MNO-3 MNO-4 MNO-5 MNO-6 MNO-7 MNO-8 MNO-9 ORIGINAL MONOGRAPH SERIES Engineering Monographs on Earthquake Criteria, Structural Design, and Strong Motion Records Coordinating Editor: Mihran S, Agbabian Reading and Interpreting Strong Motion Accelerograms, by Donald E. Hudson, 1979 Dynamics of Structures, A Primer, by Anil K. Chopra, 1982 (out of print) Earthquake Spectra and Design, by Nathan M. Newmark and William J. Hall, 1982 Earthquake Design Criteria, by George W. Housner and Paul C. Jennings, 1982 Ground Motions and Soil Liquefaction During Earthquakes, by H. Bolton Seed and I. M. Idriss, 1983 Seismic Design Codes and Procedures, by Glen V. Berg, 1983 (out of print) An Introduction to the Seismicity of the United States, by S. T. Algermissen, 1983 SECOND MONOGRAPH SERIES Engineering Monographs on Miscellaneous Earthquake Engineering Topics Seismic Design with Supplemental Energy Dissipation Devices, by Robert D. Hanson and Tsu T. Soong, 2001 Fundamentals of Seismic Protection for Bridges, by Mark Yashinsky and M. J. Karshenas, 2003 MNO-10 Seismic Hazard and Risk Analysis, by Robin K. McGuire, 2004 MNO-12 Soil Liquefaction during Earthquakes, by I. M. Idriss and R. W. MNO-11 Boulanger, 2008 SECOND EDITION Earthquake Dynamics of Structures, A Primer, by Anil K. Chopra, 2005 (first edition: MNO-2)SOIL LIQUEFACTION DURING EARTHQUAKES by I. M. IDRISS Department of Civil and Environmental Engineering, University of California at Davis and R. W. BOULANGER Department of Civil and Environmental Engineering, University of California at Davis This monograph was sponsored by the Earthquake Engineering Research Institute with support from the Federal Emergency Management Agency Rex EARTHQUAKE ENGINEERING RESEARCH INSTITUTE MNO-12©2008 Earthquake Engineering Research Institute (EERI), Oakland, California, USA. All rights reserved. No part of this book may be reproduced in any form or by any means without the prior written permission of the publisher. The publication of this book was supported by FEMA/U.S. De- partment of Homeland Security under grant #EMW-2004-CA-0297,. EERI is a nonprofit corporation. The objective of EERI is to re- duce earthquake risk by advancing the science and practice of earth- quake engineering; by improving the understanding of the impact of earthquakes on the physical, social, economic, political, and cultural environment; and by advocating comprehensive and realistic mea- sures for reducing the harmful effects of earthquakes. Any opinions, findings, conclusions, or recommendations expressed herein are the authors’ and do not necessarily reflect the views of FEMA or EERI. Copies of this publication may be ordered from: EERI, 499 14th Street, Suite 320, Oakland, CA 94612-1934; tel: (510) 451-0905; fax: (510) 451-5411; e-mail: [email protected]; web site: http://www.ceri.org. Printed in the United States of America. ISBN #978-1-932884-36-4 EERI Publication No. MNO-12 Technical Editor: Douglas Becker Production Coordinator: Eloise Gilland Layout and Production: ICC Macmillan Inc. Printing: Lynx Communication Group, Inc.Dedicated to the memory of Professor H. Bolton Seed— truly the father of geotechnical earthquake engineering: a teacher, an advisor, a mentor, an esteemed colleague, and a true friend.TABLE OF CONTENTS Foreword . xi Preface .. xiii Symbols and Acronyms xv 1 Soil Liquefaction During Earthquakes. . 1.1 Effects of Soil Liquefaction 1.2. Development of Engineering Procedures: for Assessing and Mitigating Liquefaction . . . 6 1.3 Purpose and Scope of This Monograph . 7 2 Fundamentals of Liquefaction Behavior ... . 11 2.1 Monotonic Loading of Saturated Sands .. . wees HT 2.2 Cyclic Loading Behavior of Saturated Sands ...... 20 2.3. Laboratory Testing of Field Samples and the Effects of Sampling Disturbance............... 46 2.4 Field Processes Not Replicated in Laboratory Tests .. 52 3 Triggering of Liquefaction .............0 0.000 eee 59 3.1 Liquefaction Susceptibility of Soil Deposits Gao 6 rast si 6 59 3.2 Analysis Framework for Developing Liquefaction Triggering Correlations ................00005 63 3.3. Simplified Procedure for Estimating Earthquake- Induced Stresses... 1... ....-..0 0000 e eee ee 66 3.4 In-Situ Tests as Indices for Liquefaction Characteristics... 2.60.2... eee eee 70 3.5 Overburden Correction of in-Situ Test Results ..... 84 3.6 Magnitude Scaling Factor ...............0004 89 3.7. Overburden Correction Factor, Ky ............. 94 3.8 Static Shear Stress Correction Factor, Ky ......... 96 3.9 Development of Liquefaction Triggering Correlations from Case Histories .............. 98 vii3.10 SPT and CPT Correlations for Triggering of Liquefaction in Clean Sands. . . . 3.11 SPT and CPT Correlations for Triggering of Liquefaction in Silty Sands ....... 00.00. eee ee 3.12 Probabilistic SPT and CPT Correlations for Triggering of Liquefaction . 3.13 ¥;,-Based Correlations for Triggering of Liquefaction 6.0.0... 000... e eee 3.14 Liquefaction Triggering Analyses—-Examples and Discussion. ..... . Samara aoe 4 Consequences of Liquefaction 4.1 General Considerations . 4.2 Instability and the Residual Shear Strength of of Liquefied Soil . . an 4.3 Lateral Spreading Deformations ewesusmexesmes 4.4 — Post-liquefaction Reconsolidation Sotilement 45 Examples of Lateral Displacement and Settlement Calculations aie 4.6 Margin of Safety . 5 Mitigation of Liquefaction Hazards... 5.1 Fvaluating and Selecting Possible Mitigation Strategies... 2.2... 00.00. e ee eee 5.2. Methods of Ground Improvement .............. 5.3 General Design and Construction Considerations . gaasmun ren 6 Cyelie Softening in Clays and Plastic Sits... 6.1 Behavior of Saturated Clays and Plastic Silts during Earthquakes . E - 6.2 Relating Monotonic and Cyclic Undrained Shear Strengths . 63. Number of Equivalent Uniform Loading Cycles ami MEP. cs sos iss ee vee 6.4 Static Shear Stress Correction Factor . ; 6.5 Estimating CRRs ....... Sees 6.6 Factors of Safety and Choice of Reference Stress . 6.7 — Transition from Sand. Like to ) Clay. Like Behavior in Fine-Grained Soils............ . 6.8 Consequences of Cyclic Softening in Clays and Plastic Silts .... . viii 167 - 181 185 . 187 .- 191 - 193 196 - 200 201 2057 Concluding Remarks ...... 0... es gypeey arn E ae 209 References ............ Appendix A: Example of SPT-Based Liquefaction Triggering Analysis for a Single Boring .. .. Appendix B: Example of CPT-Based Liquefaction Tigering - 238 Analysis for a Single Sounding . . 240 Appendix C: Example of SPT-Based Calculation of Lateral Displacement Index (LDI) and I-D Reconsolidation ase sea ee pers 3082 Settlement ..............FOREWORD The original seven EERI monographs were published between 1979 and 1983 and grew out ofa seminar series on earthquake engineering organized by EERI and presented in several cities. The monographs covered the basic aspects of earthquake engineering in some detail, including seismicity, strong motion records, earthquake spectra, liq- uefaction, dynamics, design criteria, and codes. The themes were fundamental and focused, and the content was thorough and gener- ally non-controversial. These monographs filled a gap in available documents and were highly acclaimed. This monograph is intended to update a subject area covered in the original series. The 1982 monograph entitled Ground Motions and Soil Liquefaction During Earthquakes, authored by H. Bolton Seed and I.M. Idriss, became a classic text used around the world, primarily for the prediction of liquefaction considering local soils conditions and seismicity. It was so popular and well-used that requests for up- dates have been persistent for fifteen years. The material on ground motion covered in the 1982 monograph has become a major field of research and is now well covered in the literature. For example, the influence of soil conditions on local ground motion is now completely accepted and is incorporated into building codes. The use of response spectra to characterize seismic demand at a site is universal. Attenuation relationships not only for peak ground acceleration but also for spectral ordinates are in their third generation, including the recently completed next generation attenuations (NGA) developed under the coordination of the Pacific Earthquake Engineering Research Center. This monograph therefore focuses on liquefaction and covers liquefaction triggering analysis, consequences and mitigation of lig- uefaction, and includes an important chapter on cyclic softening of saturated clays not covered in the 1982 monograph. The material has xibeen painstakingly collated and edited and has been thoroughly re- viewed by a large group of experts, including Jonathan Bray, Steven Kramer, James Mitchell, Jonathan Stewart, Bruce Kutter, Lelio Mejia, Yoshi Moriwaki, and Dan Wilson. Witiam T. HoLmes Cram, EERI MonocrapH ComMMITTEE April 2008 xiiPREFACE The 1982 monograph, Ground Motions and Soil Liquefaction during Earthquakes, by H. Bolton Seed and I. M. Idriss, met the need at that time for a simple representation of the essential elements of both earthquake ground motions and soil liquefaction that could be readily understood by engineers who have no particular familiarity with the field. When we set out to update that monograph, it quickly became clear that a single monograph could no longer provide sufficient cov- erage of both earthquake ground motions and soil liquefaction. The body of knowledge about these topics has grown so considerably that the current need, which is markedly different from 26 years ago, is for a thorough synthesis of the last 26 years of progress into one accessible resource for students, practicing engineers, and other professionals. Fulfilling such a need for both ground motions and soil liquefaction was impractical within the format of a single monograph, and therefore we chose to focus this monograph on soil liquefaction. A draft of this monograph was sent to Drs. Jonathan D. Bray, Steven L. Kramer, Bruce L. Kutter, Lelio H. Mejia, James K. Mitchell, Yoshi Moriwaki, Jonathan P. Stewart, and Daniel W. Wilson for their review and comments. Dr. Wilson also checked the equations and the figures for correctness and consistency. The comments and suggestions we received were extensive, de- tailed, comprehensive, and on target. We believe that, in addressing these comments and incorporating many of the suggestions, the qual- ity of the monograph has been significantly enhanced. We are very grateful to these colleagues for generously giving of their time and for providing such valuable and thorough input. xiiiIt is hoped that this monograph on soil liquefaction will serve its intended purpose and, hence, prove useful to readers as a resource in understanding and addressing soil liquefaction problems in teaching, research, and engineering practice. xiv 1. M. Ipriss, Santa, Fe, New Mexico. and R. W. Boutanorr, Davis, CALIFORNIA July 2008It is hoped that this monograph on soil liquefaction will serve its intended purpose and, hence, prove useful to readers as a resource in understanding and addressing soil liquefaction problems in teaching, research, and engineering practice. xiv I. M. Ipriss, Santa, Fe, New Mexico and R. W. Boutanore, Davis, Catirornia July 2008My Trp IcD ICU height of free face for a lateral spread hydraulic gradient soil behavior type index from CPT data relative dilatancy index isotropically consolidated drained isotropically consolidated undrained correction factor for the effects of an initial static shear stress ratio coefficient of lateral earth pressure at rest overburden correction factor distance from the free face in a lateral spread lateral displacement lateral displacement index liquidity index large penetrometer test earthquake magnitude critical-state stress ratio in triaxial compression magnitude scaling factor number, number of blows, or number of loading cycles number of equivalent uniform loading cycles number of equivalent uniform loading cycles required to trigger liquefaction SPT blow count corrected to ER = 60% and an effective overburden stress of | atm equivalent clean sand (1Vj)¢o for computing the CRR equivalent clean sand (1Vj)69 for computing residual shear strength National Center for Earthquake Engineering Research National Research Council National Science Foundation overconsolidation ratio mean total stress mean effective stress mean effective stress at consolidation. atmospheric pressure xviSASW probability of liquefaction plasticity index deviator stress cyclic deviator stress cone tip resistance cone tip resistance corrected to an effective overburden stress of | atm normalized overburden corrected cone tip resistance normalized cone tip resistance for soil classification, or parameter in /gp and & relationships quasi-steady-state line shear stress reduction coefficient excess pore water pressure ratio limiting residual value of excess pore water pressure ratio reference stress level value of s/o}, for OCR = | undrained shear strength at critical state undrained shear resistance at quasi-steady state undrained shear strength remolded undrained shear strength residual shear strength of liquefied soil in the field sensitivity spectral analysis of surface waves standard penetration test settlement due to postliquefaction one-dimensional reconsolidation initial pore water pressure shear wave velocity shear wave velocity corrected to an effective overburden stress of | atm natural water content maximum depth initial static shear stress ratio excess pore water pressure xvii&a by ¢ Yer Vim Ynax &R axial strain volumetric strain effective friction angle critical-state effective friction angle total unit weight, or shear strain limiting value of shear strain maximum shear strain relative state parameter, or relative state parameter index major principal total stress minor principal total stress vertical total stress effective stress (e.g., o/, = vertical effective stress) effective stress at consolidation (e.g., oj, = vertical effective stress at consolidation) effective preconsolidation stress (c.g., Opp = vertical effective preconsolidation stress) cyclic shear stress maximum shear stress static shear stress1 SOIL LIQUEFACTION DURING EARTHQUAKES 1.1 Effects of Soil Liquefaction One of the most dramatic causes of damage to structures dur- ing earthquakes is the occurrence of liquefaction in saturated sand deposits. Loose sand tends to contract under the cyclic loading im- posed by earthquake shaking, which can transfer normal stress from the sand matrix onto the pore water if the soil is saturated and largely unable to drain during shaking. The result is a reduction in the effec- tive confining stress within the soil and an associated loss of strength and stiffness that contributes to deformations of the soil deposit. A common manifestation of liquefaction is the formation of sand boils or mud spouts at the ground surface by seepage of water through ground cracks or, in some cases, by the development of quicksand- like conditions over substantial areas. Figure 1 shows sand boiling along a fissure in the ground caused by an earthquake, and Figure 2 shows a sand boil after the liquefaction-induced boiling has ceased. The damage from liquefaction is seldom, however, due to the sand boils themselves, but rather due to the loss of strength and stiffness in the soils that have liquefied and the associated ground deformations that ensue. Some of the most dramatic illustrations of liquefaction-related damage to civil infrastructure were observed after the 1964 Niigata, Japan earthquake and 1964 Prince William Sound, Alaska earthquake, which helped to identify liquefaction as a major problem in earthquake engineering. For example, the loss of shear strength and stiffness in liquefied sands during the 1964 Niigata earthquake resulted in dramatic bearing failures beneath buildings (Figure 3), the floating of buried tanks and other buoyant structures (Figure 4), and the collapse of bridges, including the Showa Bridge (Figure 5).Figure 1. Sand boiling caused by liquefaction of underlying sediments during the 1978 Miyagi-ken-Oki, Japan earthquake (original source unknown). Figure 2. Sand boil after liquefaction-induced boiling from the 1989 Loma Prieta, California earthquake has ceased. 2Figure 3. Tilting of apartment buildings caused by the 1964 Niigata earthquake (photo: National Information Service for Earthquake Engineering, EERC, University of California, Berkeley). Figure 4, Floating of buried tank in liquefied ground, caused by the 1964 Niigata earthquake (photo: Kawasumi 1968).Figure 5. Showa Bridge collapse, caused by liquefaction and lateral spreading during the 1964 Niigata earthquake (photo: National Information Service for Earthquake Engineering, EERC, University of California, Berkeley). A number of other major earthquakes have since provided simi- lar and additional observations related to liquefaction. Among these earthquakes are the 1971 San Fernando and the 1989 Loma Prieta earthquakes in California, the 1995 Kobe earthquake in Japan, the 1999 Kocaeli earthquake in Turkey, and the 1999 Chi-Chi earthquake in Taiwan. The slide that formed in the upstream shell of the Lower San Fernando Dam during the 1971 San Fernando earthquake left the dam 4Figure 6. Slide in the upstream shell of the Lower San Fernando Dam after the 1971 San Fernando earthquake (photo: California Department of Water Resources). with no morc than about | m of freeboard against release of the reser- voir and necessitated the evacuation of about 80,000 people who were living downstream of the dam. The dramatic extent of the slope failure is shown by a photo (Figure 6) taken after the reservoir had been drawn down, with the distorted pavement of the former road across the dam crestmarking the extent of the slide movements. This near-catastrophe marked a major change in embankment dam engineering throughout the world and has been the subject of numerous research efforts. The 1995 Kobe earthquake caused pervasive liquefaction throughout the reclaimed lands and manmade islands in the Kobe region, which is home to one of the largest container port facilities in the world, Liquefaction in these fills caused extensive damage to quay walls around the port facilities and associated damage to the cranes and other supporting facilities. For example, Figure 7 shows quay wall displacements of a few meters, graben formation behind the quay wall, and damaged and collapsed cranes in the background. The pervasiveness of liquefaction-induced damage caused an almost complete loss of functionality for the ports around Kobe, and the re- sulting economic loss was far higher than the direct costs of repairing the physical damage.Figure 7. Quay wall deformation, graben formation, and crane collapses at Port Island after the 1995 Kobe earthquake. 1.2 Development of Engineering Procedures for Assessing and Mitigating Liquefaction To assess and mitigate potential liquefaction hazards at a specific site, several questions must be addressed: + Will liquefaction be “triggered” by the design ground motions? * What will be the consequences for the structure or facility? * Whatare the options for mitigating the potential consequences? The development of engineering procedures that address these ques- tions has involved the synthesis of theoretical as well as empirical considerations that have emerged over the last several decades, For example, in evaluating the potential for triggering liquefac- tion, an essential component is to identify an appropriate means of measuring, or estimating, the soil’s resistance to liquefaction dur- ing seismic (cyclic) loading. In principle, the cyclic behavior of a soil could be determined by obtaining high-quality field samples and then testing them in an appropriate laboratory device. Experi- ence has shown that sand samples obtained by conventional sampling techniques are sufficiently disturbed to render the resulting measure- ments of cyclic strength unreliable in most situations. More reliable sampling techniques are available that lessen this disturbance to ac- ceptable levels, but only at great expense. Consequently, recourse is 6generally sought by using in-situ tests—standard penetration tests (SPTs), cone penetration tests (CPTs), Becker penetration tests(BPTs), and shear wave velocity (V,) measurements—as indices for estimat- ing the liquefaction resistance of sands and other cohesionless soils. Cohesive sediments (e.g., clays and plastic silts) can also de- velop significant strains that result in ground deformations during earthquake loading, particularly where (a) the sediments are soft and sensitive, (b) there is a significant driving shear stress (e.g., a slope or foundation load), and (c) the shaking is sufficiently strong. The difference in shear strength characteristics between cohesionless and cohesive soils, however, affects the choice and outcome of engineer- ing procedures for evaluating a soil’s response to seismic loading (e.g., clays can be sampled and tested with reasonable confidence and expense). For this reason, it is preferable to use the term “liquefac- tion” to describe the behavior of cohesionless soils (gravels, sands, and very-low-plasticity silts) and the term “cyclic softening” to de- scribe the behavior of clays and plastic silts. Criteria and procedures for evaluating the potential for cyclic softening in cohesive soils are described in Section 6 of this monograph. The development of analytical procedures for assessing lique- faction triggering has relied on empirical data to provide the. link between liquefaction resistance and various in-situ test indices. This development can be described by the following steps: + Establishing a framework of analysis that is well founded in the fundamentals of soil mechanics and physics * Collecting case histories that represent a range of observed liq- uefaction characteristics, including cases in which liquefaction did not occur + Interpreting the case histories by using the established analy- sis framework, from which semi-empirical relationships that distinguish between the occurrence and nonoccurrence of liq- uefaction can be derived As with assessment, the development of engineering procedures for evaluating the potential consequences of liquefaction and designing mitigation strategies has involved a synthesis of theoretical and em- pirical considerations. 1.3 Purpose and Scope of This Monograph The purpose of this monograph is markedly different from the purpose of the 1982 monograph, Ground Motions and Soil 7Liquefaction during Earthquakes. At the time of the earlier work, there was a need for a simple representation of the essential elements of liquefaction that could be readily understood by engineers who have no particular familiarity with it. Since then, the body of litera- ture and knowledge about liquefaction has grown considerably, and the evaluation of liquefaction effects has become common in con- sulting practice. The current need is for a thorough synthesis of the progress from the last 25 years into one accessible resource for stu- dents, practicing engineers, and other professionals. Accordingly, the remaining sections of this monograph present the following topics: + Fundamentals of liquefaction behavior. Fundamental aspects of liquefaction behavior are summarized, to provide a frame- work for a common understanding of the development and limitations of various engineering analytical procedures. Monotonic and cyclic loading behaviors of saturated sands are described with reference to critical-state soil mechanics con- cepts, which are particularly valuable for organizing obser- vations pertaining to the combined effects of relative density and confining stress. The laboratory testing of field samples is discussed, to illustrate some additional fundamental fea- tures of soil behavior and the challenges posed by the effects of sampling disturbance. Field processes that are not repli- cated in laboratory tests but can be extremely important in how geotechnical structures perform are also described. Triggering of liquefaction. Methods for evaluating the poten- tial for liquefaction triggering are described and illustrated. This section discusses the liquefaction susceptibility of dif- ferent sedimentary deposits, the analytical framework for site- specific liquefaction triggering evaluations, important features of the in-situ tests used as indices for liquefaction character- istics, liquefaction triggering correlations for sands and silty sands, and examples of liquefaction triggering analyses. Consequences of liquefaction. The potential consequences of liquefaction are discussed, with an emphasis on three of the more common consequences that are of concern in engineering practice: (a) the residual strength of liquefied soil and the po- tential for slope instability, (b) lateral spreading of gently slop- ing or nearly level ground profiles, and (c) post-liquefaction settlement that is due to liquefaction beneath gently sloping or nearly level ground profiles. Examples of lateral spreading andpost-liquefaction settlement analyses are presented, followed by a discussion of the margin of safety in engineering practice. Mitigation of liquefaction hazards. The evaluation and selec- tion of possible mitigation strategies are discussed, followed by an overview of many of the more common methods for ground improvement. General design and construction considerations are also discussed. Cyclic softening of clays and plastic silts. The potential for cyclic softening of cohesive fine-grained soils during earth- quake shaking is discussed, and engineering procedures for evaluating their potential performance are presented. Low- plasticity fine-grained soils can transition from behavior that is more like sands to behavior that is more like clays, and simple index criteria for deciding how best to evaluate these types of soils are discussed. The potential consequences of cyclic soft- ening, and the factors affecting those consequences, are also discussed.2 FUNDAMENTALS OF LIQUEFACTION BEHAVIOR This section reviews the fundamental aspects of soil liquefaction behavior that are particularly important in understanding the devel- opment and limitations of the various engineering procedures that are presented in subsequent sections and provides a framework for guid- ing practical judgments and design decisions in many situations. The principal features of the response of saturated sand to drained and undrained monotonic and cyclic loading are described first. Then the issues involved in laboratory testing of field samples, and particularly the effects of sampling disturbance, are presented. Last, field pro- cesses that are not replicated in laboratory tests but can be extremely important in the way geotechnical structures perform are described. 2.1 Monotonic Loading of Saturated Sands The stress-strain response of sand to monotonic or cyclic load- ing is strongly dependent on the sand’s relative density (Dj), effec- tive confining stress, stress history, mode of deposition, and several other factors. The concepts of critical-state soil mechanics are particu- larly valuable for organizing observations pertaining to the combined effects of Dr and confining stress on the material response in labora- tory element tests (e.g., Schofield and Wroth 1968). The term “crit- ical state” refers to the conditions that exist in sand when it is being sheared continuously and no further changes in volume or stress are occurring, and it is described by the critical-state line (CSL), which represents all possible combinations of void ratio and confining stress at the critical state. The term “steady state” refers to the critical-state condition with the additional requirement of a steady rate of defor- mation. Inasmuch as the steady state and critical state are essentially synonymous, the term “critical state” is used in this monograph. 11Loose of critical Undrained path Undrained path Void ratio, € Dense of critical Chitical state line (or steady state line) Mean effective stress, p' Figure 8. The stress paths for monotonic drained loading with constant p’ and undrained loading (constant volume shearing) of saturated loose-of-critical and dense-of-critical sands. Figure 8 illustrates the paths of saturated sands to drained and undrained monotonic loading for initial states that are “loose of crit- ical” and for those that are “dense of critical.” The drained paths are shown for a constant mean effective stress (p’) loading condition, and the undrained paths occur at constant volume (or void ratio). Examples of drained and undrained responses for sands prepared loose and dense of critical are presented subsequently to illustrate these and other aspects of their behavior. Drained Loading The response of sand to drained monotonic loading is illustrated by the results from isotropically consolidated drained (ICD) triaxial compression tests on Sacramento River sand, as shown in Figure 9 (Lee and Seed 1967). The symbol P, shown in these plots represents atmospheric pressure, which is equal to 1.03 kg/cm?, 1.06 tsf, or 101 kPa, and 4, is the effective confining stress used to consolidate the specimen before shearing. The principal stress ratio at failure is related to the effective friction angle (¢’) as follows: (g). = tan? (54 $) ay 12‘Le ard S000 1967, os ie wr Send (F087 “b+ Principal stress rat, oo Principal stra ratio, o's Volumeti stain 36) @ Axis! strain oo, o ‘Axial strain (%) Figure 9. Monotonic loading response of dense (Dg = 100%) and loose (Dx = 38%) specimens of Sacramento River sand in drained triaxial compression tests, as shown in graphs (a) and (b) respectively (after Lee and Seed 1967, with permission from ASCE). with the peak principal stress ratio corresponding to the peak effective friction angle ($,,) and the residual principal stress ratio at large strains corresponding to the critical-state or constant-volume effective friction angle (#%,). The Dp = 100% specimens in Figure 9a were dilatant (loosened) during drained shear with effective confining stresses, 04, < 10 atm (approximately 1,000 kPa). The dilation can be conceptually viewed as arising from the need for the sand particles, which are densely packed, to ride up over each other (creating more void space between them) in order to shear past each other. Atoj, > 19 atm, the Dz = 100% specimens became contractive (densified) during drained shear. At these very high confining stresses, shear deformations can be accommodated through particle breakage and particle rearrangement without the particles necessarily having to roll or rotate up over each other. 13,Intiaty loose or dense (tar ste * 0.85: iniiany dense: (Cnt)yaua® 0.7510 0.79 # (ri) new roe ICD triaxial compression tests ‘on Hostun sand (Desrues et a. 1996) Global (average) void ratio. ‘Shear zone (local) void ratio. int done nna” 05910069 05 1 ° 10 20 30 40 60 Global axial strain, (%) Figure 10. Global and local void ratios within triaxial sand specimens of loose and dense Hostun sand from x-ray computed tomography (adapted from Desrues et al. 1996 and Frost and Jang 2000, with permission from ASCE). The Dg = 38% speciinens in Figure 9b were dilatant (loosened) during drained shear with 04, of 1 atm, showed only minor volume changes with oj, of 1.9 atm, and were contractive (densified) with 04, > 4.4 atm. The ultimate or critical-state values for the principal effective stress ratio were all trending toward a consistent value of roughly 3.3 for the Dr = 38% and for the Dg = 100% specimens, regardless of the initial consolidation stress. This principal effective stress ratio corresponds to a critical-state friction angle (¢/,,) of about 32° for this sand Nonuniformities of strain within laboratory test specimens com- plicate the experimental determination of the CSL (e-p'), as illus- trated in Figure 10, which shows global and local void ratios in loose and dense sand specimens sheared in drained triaxial compression (Desrues et al. 1996). Shear bands occur within dense sand specimens during drained loading and result in post-peak strain softening: once a shear band forms, it becomes weaker than the surrounding soil, and thus any further deformation is concentrated on the already-formed shear band. The void ratios within the shear band are substantially different from void ratios outside the shear band. Shear bands do 14not tend to form in loose (contracting) sand during drained loading, because the soil strain-hardens as it continually gets denser during drained shear, in such a way that the onset of shearing on one plane means that it gets stronger than other potential shear planes, and thus any further deformation shifts away from the current shear plane. One consequence of shear band localizations in dense sands during drained shear is that the global void ratio (the average over the en- tire specimen) does not correspond to the critical-state void ratio. Instead, only the void ratio within the shear band corresponds to the critical state, and measuring the void ratio within a shear band re- quires very sophisticated measurements like x-ray tomography (e.g., Desrues et al. 1996) or digital image correlation techniques (e.g., Finno and Rechenmacher 2003). Undrained Loading The response of sand to undrained monotonic loading is illus- trated by the results from isotropically consolidated undrained (ICU) triaxial compression tests on saturated Toyoura sand, as shown in Figure 11 (Ishihara 1993). Volumetric strains are approximately zero during undrained shearing of saturated sand, and thus the void ratio of the specimens stays constant during undrained loading (local changes in the void ratio can occur, but the average for the entire specimen is constant). The deviator stress q and mean principal total and effective stresses p and p’ are defined as 9 =01 - 03 =01- 04 (2) ato to: 1 7 3 6) » a Fos +05 BEBE paw (4) The intermediate and minor principal effective stresses are equal in a triaxial compression test. As for notation, the effective stresses at the time of consolidation (before undrained loading) are identified by a subscript c, so the minor principal effective consolidation stress is a4, the major principal effective consolidation stress is o{,, and the mean effective consolidation stress is p/. The ratio ofg /p’ at critical state in triaxial compression is defined by the parameter M:, which is related to the effective friction angle 15i ¥ , i vee C : a ay in pil atecie sos pram) a i, : , we st str 08) pn iia eons Sess (atm) fo i; 7 | an pcp ate soem) Figure 11. Monotonic loading response of saturated Toyoura sand in ICU triaxial compression tests (after Ishihara 1993): (a) Dr = 16%, (b) Dp = 38%, (c) Dz = 64%. 4, as follows: _ (4) _ &sin (9, Me= (5) =F 7) ® sin (Hy) = go © The Dr = 16% specimens in Figure 11a were tested with oj, = 0.1 ~ 1.0 atm (10-100 kPa). For of, = 0.6 atm and 03, = 1.0 atm, the specimens showed some post-peak strain softening, followed by some strain hardening toward the constant volume shearing condition 16(or critical state). These specimens were initially contractive in their tendencies (i.e., they would have contracted if drained) and thus devel- oped positive excess pore water pressures (Au), with corresponding decreases in p’ during the initial loading and post-peak softening por- tions of loading response. Then, at strains of 10-15%, the specimens became incrementally dilative in their tendencies, and thus the pore pressures began to incrementally decrease, and p’ began to incremen- tally increase with increasing strains. The two specimens at 4, of only 0.1 and 0.2 atm showed strain-hardening behavior through aimost the full range of imposed strains. This strain-hardening behavior is due to their dilatant tendencies, which result in increasing p’ (decreasing u) during undrained shear loading. Most importantly, the four specimens tended toward the same critical-state strength at large strains, despite having been initially consolidated at very different confining stresses. The Dz = 38% specimens in Figure 11b were tested with of, = 1.0-30 atm (100-3,000 kPa). The specimens that had the lower con- solidation stresses exhibited a strain-hardening response (i.e., increas- ing p’ due to dilatant tendency) during undrained loading, while the specimens at the higher consolidation stresses exhibited some post- peak strain softening (i.e., decreasing p’ due to contractive tendency). The four specimens had the same critical-state strength at large strains, which was about 18 times larger than the critical-state strength for the Dr = 16% specimens. The Dr = 64% specimens in Figure 1 1c were also tested with 04, = 1.0-30 atm. The four specimens showed strain-hardening re- sponses and again reached the same critical-state strength at large strains. For these specimens, the critical-state strength was almost 3 times greater than for the Dp = 38% specimens. The transition from incrementally contractive tendencies (p' de- creasing) to incrementally dilative tendencies (p' increasing) during undrained shearing is called “phase transformation” (Ishihara et al. 1975). Phase transformation for the tests shown in Figure 12 occurs at points P and Q, which are also points of minimum shear resistance. The various states of stress and void ratio at this phase transformation point during monotonic undrained loading define what is called the “quasi-steady-state” line (QSSL), and the corresponding shear resis- tance is called the quasi-steady-state strength. The quasi-steady-state strength can be substantially smaller than the critical-state strength and is usually reached at strains equal to a few percent. The direction of shear loading can also have a large effect on the sand’s undrained stress-strain behavior. For example, Figure 13 1740 T T T T T T Ishihara (1993): Toyoura sand, icu-Te 4 Dy = 18% ‘Ac 05¢= 5.0.atm Deviator stress, q (atm) 00 0 5 10 1 2 2 30 95 @) Axial stain, «, (36) 40 r T T T T JM. =1.24 4 ‘Ac o'y= 50atm Deviator stress, q (atm) B: 0’ = 1.0 atm oo 0 1 2 3 4 5 6 (b) ‘Mean principal effective stress, p' (atm) 0.94 ot 092 4 ° g & oooh 4 3 ‘Steady or enical s stato ne (CSL) ost sy 4 Quasisteady state fne (QSSL) oes Lt ° 1 2 34 5 6 © ‘Mean principal effective stress, p' (atm) Figure 12. Quasi-steady-state (QSS) and steady-state behavior of very loose Toyoura sand in ICU triaxial compression tests (after Ishihara 1993). 1820 T Vaid et al. (1998) =o ba, and b held constant during shear a = | | E20 & fear] & | 3 4s. 5 10 of, p——_|_ oo 10 (o;-05)/2 (atm) ° L at L 0 02 05 O76 1 125 45 475 2 p'=(c', +0', +6',)3 (atm) Figure 13. Undrained stress-strain and stress-path responses of Fraser River sand tested along different stress paths in a torsional hollow cylinder (after Vaid et al. 1998). shows the results of undrained loading tests on Fraser River sand ina special laboratory device that allows the deviator stress to be applied at different inclinations from the vertical (Vaid and Eliadorani 1998). The shear resistance of the sand when loaded horizontally (a, = 90°, similar to triaxial extension loading) was less than half of the shear resistance when the sand was loaded vertically (a7. = 0°, similarto tri- axial compression loading), even though the specimens had the same initial relative density (void ratio) and consolidation stress, These results illustrate how the direction of loading and fabric anisotropy can have a strong effect on the undrained stress-strain response of sands and the corresponding values of QSS strengths. These data also raise questions about the uniqueness of a CSL, since these spec- imens did not reach the same shear resistance at large strains. These 19and other experimental results (¢.g., Finno and Rechenmacher 1997, Riemer and Seed 1997) suggest that the CSL may depend on factors that include the initial state, fabric, consolidation stress history, and loading path 2.2 Cyclic Loading Behavior of Saturated Sands Drained Cyclic Loading Drained cyclic loading with shear stress reversals can cause a net contraction (densification) of sand over a wide range of relative densities. This is why vibration is effective in compacting dry sand to a high relative density. The progressive densification of a sand specimen subjected to strain-controlled, drained, cyclic loading is shown in Figure 14 (Youd 0.58 Youd (1972): Drained simple shear, (Ottawa 64M, gag = 0.792, Og = 0484, 056 + oe= 484Pa yo 1 © 0.54 0.52 0.50 Void ratio 0.48 0.46 0.44 042 040 08 04 0 04 08 Shear displacement (mm) Figure 14. Void ratio versus cyclic shear displacement, showing densification of a sand specimen with successive cycles of drained simple shear loading (after Youd 1972, with permission from ASCE). 20Hussin, J. D., and Ali, S., 1987. Soil Improvement at the Trident Sub- marine Facility, Soil improvement-a ten year update; Geotechnical Special Publication No. 12, J. P. Welsh, ed., ASCE, 215-31 Hynes, M. E., and Olsen, R., 1998. Influence of confining stress on liquefaction resistance, in Proceedings, International Sympo- sium on the Physics and Mechanics of Liquefaction, Balkema, Rotterdam, pp. 145-52. Hyodo, M., Yamamoto, Y., and Sugiyama, M., 1994. Undrained cyclic shear behavior of normally consolidated clay subjected to initial static shear stress, Soils and Foundations, Japanese Society of Soil Mechanics and Foundation Engineering 34(4), 1-11. Tai, S., and Koizumi, K., 1986. Estimation of earthquake induced excess pore water pressure for gravel drains, in Proceedings, 7th Japan Earthquake Engineering Symposium, pp. 679-84. Idriss, I. M., 1985. Evaluating seismic risk in engineering practice, in Proceedings, 11th International Conference on Soil Mechanics and Foundation Engineering, San Francisco, Balkema, Rotterdam, pp. 265-320. Idriss, I. M., 1999. An update to the Seed-Idriss simplified proce- dure for evaluating liquefaction potential, in Proceedings, TRB Workshop on New Approaches to Liquefaction, Publication No. FHWA-RD-99-165, Federal Highway Administration, January. Idriss, I. M., and Boulanger, R. W., 2003a. Estimating K, for use in evaluating cyclic resistance of sloping ground, in Proceedings, 8th US-Japan Workshop on Earthquake Resistant Design of Lifeline Facilities and Countermeasures against Liquefaction, Hamada, O’Rourke, and Bardet, eds., Report MCEER-03-0003, MCEER, SUNY Buffalo, NY, pp. 449-68. Idriss, I. M., and Boulanger, R. W., 2003b. Relating K, and K, to SPT blow count and to CPT tip resistance for use in evaluating liquefaction potential, in Proceedings of the 2003 Dam Safety Conference, ASDSO, September 7-10, Minneapolis, MN. Idriss, I. M., and Boulanger, R. W., 2004. Semi-empirical procedures for evaluating liquefaction potential during earthquakes, in Proceedings, 11th International Conference on Soil Dynamics and Earthquake Engineering, and 3rd International Conference on Earthquake Geotechnical Engineering, D. Doolin et al., eds., Stallion Press, Vol. 1, pp. 32-56. Idriss, I. M., and Boulanger, R. W., 2006. Semi-empirical procedures for evaluating liquefaction potential during earthquakes, J Soil Dynamics and Earthquake Eng. 26, 115-30. 219— Compression curve Equivalent volume change ‘of grain structure due to cyclic stain applications ‘during drained loading au—>} Excess poe water ponerse dag onraed ding ow ve Effective stress Figure 15. Mechanism of pore pressure generation during cyclic loading. B to point C). In the limit, the cyclic loading would completely break apart the load-bearing contacts among sand particles in such a way that the sand skeleton would be carrying zero normal stress (o’ = 0), and the pore water would be carrying the entire normal stress (u = 0). The cyclic undrained loading behavior is illustrated in Figure 16 by the results from an anisotropically consolidated undrained (ACU) cyclic triaxial test on medium-dense clean sand with uniform sinu- soidal stress loading. The excess pore water pressure (Au) generated during undrained cyclic triaxial loading is normalized by the minor effective consolidation stress (04,); this ratio is called the excess pore water pressure ratio (ry): Au n=s %e a”) The standard cyclic triaxial test keeps the minor principal total stress constant, so the maximum possible value for 7, is 1.0 (or 100%), which occurs when Au = 04, and of = 0. 22oa [aeoahoe eae iciertear ery a | A. —_ Es a stasis eee 2 18. ee ‘ "oo 7 l i Tt s Figure 16. Response of Sacramento River sand to undrained cyclic triaxial loading (test from Boulanger and Truman 1996). For standard cyclic simple shear tests, r, is instead computed on the basis of the vertical effective consolidation stress (o;,.): Au a (8) Fe vu The maximum possible value for r, is again 1.0 when the total vertical stress is held constant, as in a standard cyclic simple shear test. The r, = 1.0 condition is often called “initial liquefaction.” ‘There are, however, advantages to using the more explicit phrase “excess pore pressure ratio of 100%,” because the term “liquefaction” has also been used in the literature to describe other specific field and laboratory conditions, as discussed subsequently in this monograph. It is also worth noting that r, values above 100% can develop if the mean total stress increases, such as can occur under the more general loading conditions produced in the field, in centrifuge models, or in numerical models. In such cases, it is more useful to focus directly on the values of effective stress rather than on excess pore pressure ratios, Several features of the behavior in Figure 16 are worth not- ing. The r, increased progressively throughout cyclic loading until Tu 1.0 was reached after about 27 cycles of loading. The axial strains (eq) remained relatively small (a fraction of 1%) until p’ ap- proached zero and r, approached 100%, after which the axial strains increased to about 2% in less than 2 additional cycles of loading. 23Axial strains would have increased very rapidly with continued cyclic loading, although this particular test was stopped after reaching 3% strain. The corresponding stress-strain response shows rapid softening as p’ approached zero, with the hysteretic loops taking on an inverted s-shape. The stress path—that is, q /(2p’) versus p’/p{—moved pro- gressively toward the origin during cyclic loading until it stabilized with repeating loops emanating from the origin. The interconnections among the various plots in Figure 16 can be illustrated by considering the positions of the points A and B. Point A corresponds to the time when r, = 100% (i.e., p’ = 0), which oc- curs only when q = 0 (i.e., the specimen is under an isotropic state of stress). Point A occurs as the specimen is unloaded to g = 0, at which time the strains are only slightly smaller than the peak during that par- ticular loading cycle. The specimen stiffness is very small at point A, and the subsequent application of shear stress in the positive direction results in the rapid growth of strains in that direction. As the shear stress increases toward Point B, the specimen progressively stiffens, with r,, decreasing and p’ increasing. Point B, therefore, corresponds to a local minimum in 7, (0.77) and to the largest axial strain in that direction of loading. The specimen is stable under this applied peak shear stress, which reflects the fact that it is dense of critical. This accumulation of limited strains, after r, = 100% has temporarily oc- curred, has been called “cyclic mobility” behavior (Casagrande 1976, Castro 1975) or “cyclic ratcheting” behavior (Castro 2008, personal communication). The inverted s-shaped stress-strain behavior that develops as r, nears 100% (i.e., p’ nears zero) arises because the specimen alternates between having incrementally dilative tendencies during shear load- ing and incrementally contractive tendencies during unloading. This is directly comparable to the drained cyclic loading response shown in Figure 14, in which the specimen alternated between incremental dilation and incremental contraction during loading and unloading, re- spectively. For undrained conditions, however, the tendency of sand to dilate increases p’ and hence increases tangent stiffness, while the ten- dency to contract decreases p’ and hence decreases tangent stiffness. Figure 16 also illustrates that r, = 100% is a temporary con- dition that occurs only under isotropic states of stress (i.e. zero shear stress) and that r, = 100% can be generated in sands that are dense of critical (i.e., they would have a dilative tendency under monotonic drained loading). Liquefaction of a dense-of-critical sand during cyclic loading results in limited strains (or cyclic ratcheting), 24Ishihara et al. (1991): ar Uincained axa £ Toyoura sand, & Dp = 16% > 8 10) £ 5 a 8 y Cyclic to 3S fi ‘monotonic loading © os! BT encionic / | eang 5 ae 2 4 6 6 10 49 14 16 18 20 Axial strain, ¢, (%) 18 = Monotonic 3 loading : i g tol £ le 3 8 09 a os Cyclic to Pelosi = / 0 0 02 04 06 08 10 12 14 ‘Mean principal effective stress, p' (atm) Figure 17. Undrained behavior of Toyoura sand in monotonic loading versus eyclic-to-monotonic loading in triaxial tests of loose-of-critical sand (after Ishihara et al. 1991). because the sand exhibits dilative behavior under subsequent mono- tonic loading. Undrained Cyclic Loading of Loose-of-Critical and Dense-of-Critical Sands The undrained cyclic loading response of dense-of-critical sand versus loose-of-critical sand is illustrated by the test results in Fig- ures 17 and 18. These two figures show the response of sand to undrained monotonic loading and to undrained cyclic loading fol- lowed by undrained monotonic loading. The initial states of these specimens (e and p’) can be plotted against the critical-state (steady- state) line for this sand (Figure 12), showing that one state is initially above the critical-state line while the other state is initially below that 25Ishihara et al. (1991): = 14 ene one, & Toyoura sand, g 16% 8 19) g Monotone : : Kei é 0.5] _» ptt | Frets hadng F St as sal stan 99) 4 g = 14 | fos a 0 02 04 06 08 10 12 ‘Mean principal effective stress, p' (atm) Figure 18. Undrained behavior of Toyoura sand in monotonic loading versus cyclic-to-monotonic loading in triaxial tests of dense-of-critical sand (after Ishihara et al. 1991). line. The specimens in both tests were anisotropically consolidated in such a way that the specimens were carrying an initial static shear stress before undrained loading (point B in Figure 17 and point B' in Figure 18). The specimens were then subjected to undrained cyclic loading that produced high excess pore pressures and axial strains slightly smaller than 2% (points C and C’). Note that the cyclic stress amplitude was smaller than the initial static shear stress, so the devia- tor stress (q) was never equal to zero; thus neither specimen reached a state of r,, = 100% (or p’ = 0). The loose-of-critical specimen in Figure 17 collapsed toward the critical state after cyclic loading, with its shear resistance becoming smaller than the initial static deviator stress. The postcyclic shear 26resistance from point C to point D is essentially the same as that for a monotonic undrained test for sand of the same density, indicating that the cyclic loading did not affect the specimen’s critical-state strength. The cyclic loading was, however, sufficient to “trigger” the collapse of the specimen and the development of large uncontrolled deformations (.e., “flow liquefaction”). The collapse occurred while the effective stresses remained greater than zero (i.¢., 7, < 100%) and the soil maintained a nonzero shear resistance. The dense-of-critical specimen in Figure 18 developed limited strains from the cyclic loading but always maintained enough shear resistance to remain stable. The excess pore pressures generated by the cyclic loading in fact moved the specimen’ state further away from the CSL in e-log(p’) space (i.e., reducing p’ while e remained constant). The specimen subsequently strain-hardened during the posteyclic undrained monotonic loading, during which its dilative tendencies increased p’ as it moved back toward the critical state. The postcyclic monotonic loading resistance is again similar to that for an undrained monotonic test on sand of the same density. Dependence of Cyclic Strength on the Number of Loading Cycles, Relative Density, and Confining Stress The resistance of sand to the triggering of liquefaction (.e., its cyclic strength) depends on several factors, including the number of loading cycles, relative density, confining stress, depositional method, fabric, prior stress-strain history, age, cementation, and other environ- mental factors. This section discusses in some detail the effects of the number of loading cycles, relative density, and confining stress. The other factors are discussed separately in subsequent sections. Liquefaction of saturated sands can be triggered by different combinations of uniform cyclic shear stress ratio (CSR), which is the uniform cyclic shear stress divided by the initial effective confining stress, and the number of loading cycles (N). A greater CSR will trigger liquefaction (e.g.,r, = 100% or acyclic shear strain, y = 3%) in fewer loading cycles, whereas a smaller CSR will require more loading cycles. This aspect of behavior is illustrated by the results of the shaking table tests by De Alba et al. (1976), as shown in Figure 19. Note that the CSR for shaking table tests or simple shear tests is defined as the cyclic shear stress (t~,) acting on horizontal planes divided by the vertical effective consolidation stress (¢/,,), csr = (9) ve 2704 tt T ‘Shaking table tots conaucted by Be Aiba ot (1975) A. Relative Density = 90% © Relative Donsty = 32% I Relative Density = 66% 03 x Relative Density = 56% Initia contig pressure 8a (88 Pa) 02 Cyclic stress ratio of 0.0 1 10 100 Number of cycles to cause initial liquefaction Figure 19. The CSR required to reach initial liquefaction (r, = 100%), from shaking table tests by De Alba et al. (1976). whereas the CSR for isotropically consolidated cyclic triaxial tests is defined as the maximum cyclic shear stress (gey-/2) divided by the isotropic consolidation stress (4), csr = 22 (10) 205, The CSR that is required to reach liquefaction in a specified num- ber of loading cycles may also be called the sand’s cyclic resistance ratio (CRR), for notational convenience. The relationship between the CRR and N, within the range of cycles of interest for earthquake engineering, can generally be approximated with a power function as CRR=a-N? (iy) where the parameters a and b are determined by regression against the experimental data. Note that the CRR versus N relationship plots as a straight line on a log-log plot, whereas it plots as a curve on the semilog plot format of Figure 19. The parameter 4 for clean sands is typically about 0.34, whereas the parameter a depends on a wide range of factors. The dependence of the CRR on N means that any reference to a sand’s CRR must specify NV. 280 os (ln & Sanayi 1298) 2 ‘Relative density, Da (6) yi stress to 3% strain in cycles Pa) ‘i : Rare CRE fo 3% sven in 10 cycles Etfective consakseton sos, 6'y (MPa) Figure 20. Cyclic triaxial test results for clean Fraser Delta sand, showing that cyclic stress and the CRR cause 3% shear strain in 10 uniform cycles at Da values of 31-72% and effective consolidation stresses of 50-400 kPa (original data from Vaid and Sivathayalan 1996). The CRR of sand increases with increasing relative density, as illustrated by the shaking table test results in Figure 19. The CRR of sand also depends on the effective confining stress, which reflects the fact that the tendency of sand to dilate or contract depends on confining stress. This aspect of behavior is illustrated by the ICU cyclic triaxial test results on Fraser Delta sand by Vaid and Sivathayalan (1996), shown in Figure 20. The plot at left shows the cyclic stress (qeyc/2) required to reach 3% shear strain in 10 cycles versus the effective consolidation stress for specimens at Dr values of 31, 40, 59, and 72%. This plot shows that cyclic strength increased with increasing consolidation stress for all values of Da, but the shape of these relationships ranged from being nearly linear at the lowest Da to being most strongly concave at the highest Dr. The plot at right shows the ratio of qcyc/2 divided by 04, that is required to cause 3% axial strain in 10 cycles (i.e., the CRR at 10 cycles) versus Dg for different consolidation stresses. This plot shows that the CRR increased with increasing Dp but that it also decreased as the effective consolidation stress was increased from 50 kPa to 400 kPa for Dg values greater than 30%. Seed (1983) introduced the overburden correction factor (Ko) as a way to represent the dependence of the CRR on consolidation stress, with K, defined as CRRz: Ko = ERR, (12) where CRR,, is the CRR of a soil under a specific value of effective consolidation stress o{, and CRR,/-1 is the CRR of the same soil 29120 Fraser Delta sand, Dx = 72%, 1'y,= 50, 100, 200, 400 KPa (Vaid & Sivathayatan 1996) 8 1 P man = 028 [| ORR =0242 7 40 Cyclic stress, qeye/2 (kPa), to 3% strain in 10 cycles 0 100 200 300 400 500 Effective consolidation stress, o'y, (kPa) Figure 21. The amount of cyclic stress required to cause 3% strain in 10 uniform cycles versus effective consolidation stress in ICU cyclic triaxial tests on Fraser Delta sand. when o/ = 1 atm (~100 kPa). The definition of the K, factor is illus- trated in Figure 21, which shows the cyclic stress required to reach 3% strain in 10 cycles versus the effective consolidation stress for Fraser Delta sand at Dg = 72%. The relationship between cyclic stress and consolidation stress is curved, thus its secant slope (which is the CRR) decreases with increasing consolidation stress. For example, the data in Figure 21 show that the CRR at an effective consolidation stress of 400 kPa is 18% smaller than at 100 kPa. This slight curvature is analogous to the slight curvature in the failure envelopes for drained shearing tests and the associated observation that peak (secant) effec- tive friction angles decrease with increasing confining stress. The K, factor varies with Dg and with the test device (ie., simple shear versus triaxial) and appears to be different for freshly reconstituted laboratory specimens as compared with tube samples of natural soils, as shown in Figure 22. ‘The dependence of the CRR on both Da and confining stress can be related to the sand’s “state” (position) with respect to its CSL, Been and Jefferies (1985) introduced the state parameter (£) as a measure of state, where & is the difference between the current void ratio (e) and the critical-state void ratio (e-;) for the current value of p’. Konrad (1988) later showed that the value of € can be normalized by the difference in the maximum and minimum void ratios (@mar — @min) to 30Vaid & Sivathayalan (1996) © Dg=31%, simple shear 2 D,=72%, simple shear © Dg=31%, triaxial Wh Dg=72%, triaxial } p Harder & Boulanger T Field samples summarized in Seed & Harder (1990) 0 2 4 6 8 10 Vertical effective stress, o',/P, Figure 22. Comparison of K, relationships with data from reconstituted Fraser Delta sand specimens and various field samples. arrive at a relative state parameter that provides improved correlations with the shear behavior of sand. Determining the state or relative state parameters requires detailed experimental testing to define a sand’s CSL, its minimum and maximum void ratios, and its in-situ void ratio. The natural heterogeneity of sand deposits, however, makes it impractical to perform sufficient experimental tests to define CSLs for all the different zones within a natural deposit. Boulanger (2003a) introduced an index for representing the rela- tive state parameter. This index, £,, is defined in Figure 23 and repre- sents the difference between the current Dz and an empirical critical state Dp (denoted Dp.es) for the same mean effective normal stress (p’). The empirical CSL and the expression for Dyes in Figure 23 were derived by using Bolton’s (1986) relative dilatancy index (Ip) The parameter Q in Figure 23 determines the stress at which the CSL curves sharply downward—which indicates the onset of significant particle crushing—and the parameter’s value depends on grain type, with Q ~ 10 for quartz and feldspar, 8 for limestone, 7 for anthracite, and 5.5 for chalk (Bolton 1986). The resulting &g parameter pro- vides a rational means for including the interaction between relative 3Critical state line from Ig relation (Bolton 1986) with Q=10 Relative density, Dz 0.01 of 1 10 100 Mean principal effective stress, p'/P, Figure 23. Definition of the relative state parameter index, &, (after Boulanger 2003a, with permission from ASCE). density and confining stress in the analytical framework that is used to evaluate liquefaction potential. The CRR of sand for a given test type can be expressed as an approximately unique function of €p, as illustrated in Figure 24 with data from Vaid and Sivathayalan (1996) for Fraser Delta sand. These results are for specimens prepared at Dg values of 31-72% under effective consolidation stresses of 50-400 kPa, with the CRR corre- sponding to the development of 3% shear strain in 10 uniform loading cycles. These data suggest that &g can reasonably represent the com- bined effects of Dz and oj, on the CRR. In Figure 25, the ICU cyclic triaxial test results for Fraser Delta sand illustrate the interrelations among Ky, &g, CRR, Dp, and con- solidation stress. The CRR versus De relationship (for a given con- solidation stress) curves upward with increasing Dr, which directly relates to the CRR versus &z relationship curving upward with de- creasing g. Increasing the consolidation stress from 100 kPa to some higher value results in the same decrease in &, for all values of Dr. This Aég changes the CRR by amounts that depend on the initial Dp, because of the curvature of the CRR versus &g relationship (which applies to all values of Dg and consolidation stresses). Thus, the 32Fraser Delta sand, ——| = 31, 40, 59, & 72% Oi o/P,= 05, 1,2, &4 —| (Vaid & Sivathayalan 1996) 1 Triaxial ‘Simple shear 1.00. 08 06 04 -0.2 0 0.2 Relative state parameter index, Ex. Figure 24. CRR versus &» for reconstituted specimens of Fraser Delta sand (Boulanger 2003a, with permission from ASCE). reduction in the CRR, as expressed through the K, factor, depends on the initial Dp. The K, relationships derived from the CRR versus & relationship (Figures 24 and 25) match the relationships derived di- rectly from the experimental results (Figures 20 and 22), as expected. Relating Cyclic Strength from Triaxial and Simple Shear Tests to Field Conditions The differences in the CRR measured in cyclic simple shear and cyclic triaxial tests are largely related to their different states of consolidation stress. A normally consolidated sand specimen that is one-dimensionally consolidated in a simple shear device will have K, values of 0.45-0.5, whereas K, would equal | if tested in an ICU cyclic triaxial test. Ishihara et al. (1977, 1985) performed cyclic tor- sional shear tests with different K, values and showed that the CRR for anisotropically consolidated specimens (K, % 1) can be approx- imately related to the CRR for isotropically consolidated specimens (Ko = I) as 1+2Ko 3 ERR = ( J eRR 3) 3304 ——-——+ . os 03 Boz J or Ohta 7 i o % 0 6 8 10 “oe 06 a a2 0 Op eR ° Cita lative sate ine (ne) for 9 2 4 é © as KO 06 4 80 100 os on 1 wo 0 2 6 o2/Ps oP, Figure 25. The relationships among CRR, relative density, effective confining stress, relative state parameter index, and Ks, illustrated by using the CRR versus €p relationship from cyclic triaxial tests on Fraser Delta sand. Consequently, increasing K, from 0.4 to 0.8 causes an increase of about 44% in the CRR. Similarly, the CRR from the simple shear test can be related to the CRR from the ICU triaxial test as 142 (Ko: CRRgs = oe) CRRoy (14) For normally consolidated sand, K, would be about 0.45-0.50, which results in CRRgs = (0.63 to 0.67) CRRrx (5) 34This relationship between the CRR for simple shear and triaxial tests is consistent with the range of results obtained or recommended by a number of researchers (Seed and Peacock 1971, Finn et al. 1971, Ishibashi and Sherif 1974, Castro 1975, Seed 1979). One- and two-directional cyclic simple shear tests and shaking table tests have further shown that adding a second direction of cyclic loading reduces the CRR by about 10-15%, as summarized by Pyke et al. (1974), Seed (1979), and Ishihara (1996). This adjustment is needed whenever unidirectional cyclic laboratory tests are being used to estimate the CRR of in-situ sand. For level ground conditions, earthquake loading is best approximated as two-directional simple shear loading, so the CRR from a unidirectional simple shear test would be reduced by 10% to represent in-situ conditions. Note that vertical shaking of saturated soils beneath a level ground surface has a negligible effect on the soil’s CRR. Consequently, the in-situ CRR for two-dimensional shaking would be estimated from ICU cyclic triaxial tests as 142 (Koei CRR feta = 0.9 (Eee ) CRRry (16) and from cyclic direct simple shear tests as 142 (Ko)fieta 1+2(Ko)ss For example, if the sand is normally consolidated in situ with an estimated K, of 0.5, and the CRR is determined by ICU cyclic triaxial tests, then the CRR would be multiplied by 0.60 (i.e., the product of 0.9 and 0.67) to represent in-situ conditions and bidirectional shaking. CRRyieta = 0.9 ( ) CRRss (17) Other Factors Affecting the Cyclic Strength of Sand Laboratory element tests have also shown that the CRR of satu- rated sands depends on depositional method, fabric, prior stress-strain history, age, cementation, and other environmental factors. Ladd (1974, 1977) and Mulilis et al. (1977) showed how speci- mens of the same sand prepared to the same Dz by different reconsti- tution techniques can result ina CRR that varies by almost a factor of 2 from highest to lowest (Figure 26). The effect of the sample prepa- ration method is attributed primarily to differences in the fabric of the resulting sand particle matrix. Overconsolidation has also been shown to increase the CRR beyond the amount that is due to associated increases in K,, but the 3505 TT 2 a — 2 = 0.3 i 3 2 02}- $ D high req vio. moist samples GZ) high ro. vib. dry samples 0.1} ® moist tamping ® piwieted - water a ® ‘ow treq. vib. dry samples © pluviatod— air (aftr Mis et 1977) 00 ran L oan 1 1 10 100 Number of cycles to cause r, = 100% and 42.5% axial strain Figure 26. Results of cyclic triaxial tests, showing the influence of the sample preparation method on the CRR of Monterey No. 0 sand at Dr = 50% (after Mulilis et al. 1977, with permission from ASCE). increase in the CRR has ranged from being relatively minor to as much as being proportional to the square root of the overconsolidation ratio (Lee and Focht 1975, Ishihara and Takatsu 1979, Finn 1981). The effects of strain history have been shown to be very signifi- cant and to range from a beneficial effect at small prestrain levels to a detrimental effect at large prestrain levels (Suzuki and Toki 1984, Seed et al. 1977, Finn et al. 1970). For example, cyclic triaxial test results by Singh et al. (1982) and Goto and Nishio (1988), as pre- sented later in Section 2.3, showed increases in the CRR of 30-100% from prior cyclic straining at strain levels that did not significantly change the density of the specimens. Cyclic straining at larger strain Ievels will produce associated volumetric strains, which may range from a net contraction to a net loosening. The effect of larger prestrain levels on sand behavior during subsequent undrained cyclic loading depends on the volumetric strains induced by the prestraining and their uniformity within the laboratory testing device. Effect of Static Shear Stresses on Cyclic Behavior The effect of an initial static shear stress on the undrained cyclic loading behavior of saturated sand is illustrated by the cyclic simple shear test results for Sacramento River sand shown in Figure 27. This 36te goat nh = = IH g we “4 oe ML — ‘Shear stain, (%) os ween ° ot enna 4] 7 Dasha oa ctor ine TTA se Bete os 4 “i i oof? : 4 Cycle number ie Figure 27. Response of Sacramento River sand to undrained cyclic simple shear loading with an initial static shear stress ratio of 0.32 (test from Boulanger et al. 1991), simple shear test specimen was consolidated with a static horizontal shear stress (z;) equal to 0.32 times the vertical effective consolidation stress, which is called an initial static shear stress ratio (a = t,/o/.) of 0.32. The applied CSR was about 0.28, which means that the horizon- tal shear stress was always positive (about 0.04-0.60). The induced ry increased quickly in the first few loading cycles and then increased slowly with each subsequent cycle of loading but was always less than 100%. The shear strain quickly reached 2-3% in the direction of the static shear stress, after which the specimen slowly accumulated additional shear strains with each additional loading cycle. The stress path (1j, versus /) moved over toward the failure line, after which it stabilized on an essentially repeating pattern. When the behavior for this specimen is compared with that in Figure 16 for a specimen with zero initial static shear stress (w = 0), it is clear that the presence of the static shear stress had strong effects on the pore pressure and shear strain generation behavior. The effect of a static shear stress ratio on the cyclic strength of clean sand at different Da values is illustrated by the simple shear test results for Ottawa sand in Figure 28a. Data are shown for the Teye/o4. causing 3% shear strain in 10 cycles of loading for specimens at Dp values of 50% and 68% under o;.. of 200 kPa (Vaid and Finn 1979). The Dz = 68% specimens exhibit a progressive increase 37030 : fe Saas — Bl tment cron re S 020) So. s = g § | S01 ——Joyeson 4] 2 ‘ ‘8 © 005] 8 9g °F sine sear tess wth oi =z S I iGammaatnte] | é ooo Static shear stress ratio, a Static shear stress ratio, « Figure 28. Variation of the CRR for 3% shear strain in 10 cycles with the initial static shear stress ratio, «. Graph (a) shows the effect of varying the relative density, and graph (b) shows the effect of varying the effective consolidation stress. in the cyclic strength with increasing initial static shear stress ratio, whereas the Dz = 50% specimens exhibit a lower cyclic strength with increasing initial static shear stress ratio. This effect of Dz is attributed to the differences in the dilative tendencies of the sand, with the sand becoming more strongly dilatant in shear as Dp is increased (at the same effective confining stress). The effect of a static shear stress ratio on the cyclic strength of sand at different effective confining stresses is illustrated by the cyclic triaxial tests on tailings sand in Figure 28b, Data are shown for specimens at a Dp of 70% consolidated at minor principal effective stresses (03,.) of 200 and 1,600 kPa (Vaid and Chern 1985). For triaxial tests, the static shear stress ratio is computed from the shear and normal stresses on the potential failure planes (i.e., planes inclined at 45 + ¢"/2 degrees from the horizontal, as used by Seed et al. 1975a). The specimens at 04, = 200 kPa exhibit a progressive increase in the cyclic strength with an increasing initial static shear stress ratio, whereas the specimens at 04, = 1,600 kPa exhibit a lower cyclic strength at high initial static shear stress ratios. This effect of 03, is similarly attributed to its influence on the dilative tendencies of the sand, with the sand being more strongly dilatant in shear at the lower effective confining stress (at the same Dp). ‘The generation of excess pore pressure and shear strains during undrained cyclic loading of saturated sand is significantly affected by the rotation of principal stress directions, which reverses the shear stress direction on certain planes. In cyclic triaxial tests with an initial 38static deviator (shear) stress, there is no rotation of principal stresses until the cyclic deviator stress exceeds the static deviator stress, after which there is a 90° rotation of the principal stress directions (and hence shear stress reversal on all possible planes). In cyclic simple shear or torsional shear tests, there is a continuous rotation of the principal stress directions during cyclic loading, regardless of the ini- tial static shear stress. The effect of this difference is evident in the shapes of the cyclic strength versus a relationships obtained in dif- ferent testing devices. For example, this is why the cyclic triaxial test results for 03, = 1,600 kPa in Figure 28b exhibit an initial increase in the cyclic strength up to about @ ~ 0.1, followed by a subsequent decrease in the cyclic strength at higher values of w. Seed (1983) developed the K, correction factor to represent the effects of an initial static shear stress ratio (a) on the cyclic strength. Kg is defined as the cyclic strength for some value of a, divided by the cyclic strength for w = 0: CRRq ~ CRRe=0 Numerous researchers have studied this phenomenon by using cyclic triaxial, cyclic simple shear, torsional shear, and torsional ring shear devices. These studies have shown that K,, depends on relative density and confining stress, as illustrated in Figure 28, which simply reflect. the state of the sand in relation to its critical state (e.g., Vaid and Chern 1985, Mohamad and Dobry 1986). In addition, Ky depends on the failure criteria that are used to define the CRR and depends somewhat on the laboratory test device, with simple shear preferred over triaxial loading because it more closely approximates the in- situ rotation of principal stress directions expected during earthquake shaking. A review of these studies is in Harder and Boulanger (1997). ‘The general trends in Kg data are illustrated in Figure 29, which shows simple shear test results for Sacramento River sand, along with the simple shear test results for Ottawa sand (Figure 28a). The data for Ottawa sand are at Dg values of 50% and 68% under o/,, of 200 kPa (Vaid and Finn 1979), and the data for Sacramento River sand are at Dp values of 35% and 55% under of, of 200 kPa (Boulanger et al, 1991). These results correspond to 10 cycles of loading and a failure criterion of 3% shear strain. The €g value for each set of test data is shown beside the corresponding K,, curve in Figure 29. There is a consistent progression in Kq values from less than 1.0 for the loosest sand (g = —0.16) to greater than 1.0 for the densest sand Ke (18) 39Cyclic triaxial tests with Dy =70%: 5: /Pq =2 & 16 (Vaid & Chern 1985) Simple shear tosts with 6 /P, = 2 4 55% (Boulanger et a. 1991) & 68% (Vaid & Finn 1971) 0 o4 02 03 04 Qa Figure 29. Effect of the relative state parameter index &, on the static shear stress ratio correction factor Ky (Boulanger 2003a). (Ex = —0.49), with the Ep values simply tracking the progression from Dr = 35%-68%, since all tests were at the same confining stress. The dependence of K, on the effective confining stress is also depicted in Figure 29 by the results of the cyclic triaxial tests on tail- ings sand by Vaid and Chern (1985) (Figure 28b). For determining Kq from these triaxial tests, the cyclic resistances were normalized by the mean effective consolidation stresses (Boulanger 2003a). The &p values and K,, relationships for these specimens are consistent with the trends obtained on the clean sands at the lower confining stress only. In particular, the Eg value for the Dr =70% tailings sand at 03, = 1,600 kPa is greater than the &g value for Dg = 35% Sacramento River sand at o/, = 200 kPa, and this correctly indicates that the tailings sand would be more contractive and hence have a smaller Kq value at high values of w. In this manner, the Eg index ap- pears to provide a reasonable means of accounting for the combined influence of relative density and confining stress on K, relationships for sands. 40os Average 06 wf 02 Excess pore pressure ratio Ty = AWS 4 Cycle ratio, Ne /Nyg Figure 30. Excess pore water pressure generation versus the number of loading cycles divided by the number of loading cycles to liquefaction (after Seed et al.1976, with permission from ASCE). Excess Pore Pressures at Different Factors of Safety against Liquefaction The generation of excess pore pressure during uniform cyclic undrained loading of saturated sands depends on the presence or absence of any initial static shear stress. Laboratory tests with « = 0 are considered most representative of the conditions that exist beneath level ground surfaces where the shear stresses on horizontal planes are zero. In this case, the generation of r,, can be related to the ratio Ne/ Nig, a8 shown in Figure 30, where N, is the number of equivalent uniform loading cycles, and Njg is the number of equivalent uniform loading cycles required to trigger liquefaction. For level ground conditions, the residual r,, that remains after cyclic loading of sands and gravels can also be related to the factor of safety against liquefaction, FSjig, as shown in Figure 31. The FSjig is determined as CRR FSiiq = Cop (19) and is related to the ratio N. / Njig through the power relation presented previously. Figure 31 shows residual r, values dropping to 0.1-0.65 with an FS fig of 1.2. The presence of sustained static shear stresses (# > 0), such as exist beneath sloping ground, affects both the rate of pore pressure generation and the magnitude of the residual pore pressure (i.e., the 4Gravel Evans 1987, Hynes 1988) ‘Sand (Tokimatsu & Yosnim 1983) Residual excess pore pressure ratio, r, 15 2 Factor of safety against liquefaction, FSyq Figure 31. Excess pore water pressure ratio versus FSjig under level ground conditions, from laboratory test data (after Marcuson etal. 1990). residual is what remains after cyclic loading has ceased). The limiting value of the residual r,, will be less than 100%, as shown by Ishihara and Nagase (1980) and Boulanger (1990), and is illustrated by the experimental results in Figure 27. The value of ry,lim decreases with increasing w, as shown in Figure 32 for cyclic simple shear tests on Fuji River sand (Ishihara and Nagase 1980). The rate of pore pressure generation in the presence of a sustained static shear stress ratio is also significantly different from that for level ground conditions (Finn 1981, Boulanger et al. 1991), as illustrated in Figure 33, and thus the final relationship between F'Sjjg and residual r, is significantly different from that derived for level ground conditions. Strain-Based Evaluations of Undrained Cyclic Loading Behavior The undrained cyclic loading behavior of saturated sands can also be evaluated via strain-based approaches. Experimental data show that the generation of excess pore pressures is more uniquely related to im- posed shear strains than to imposed shear stresses. In practical applica- tions, the use of a strain-based approach would then require estimating shear strains induced by ground shaking, which requires estimating both induced shear stresses and the soils’ shear moduli. Consequently, a stress-based approach has remained preferable in practice, and thus only a brief discussion of strain-based findings is presented here. 421.0 08 06 04 pressure £atiO, Fs Limiting value of residual pore 0.2 F Ishinara & Nagase (1980) Fuji river sand, Dg = 53-64%, o'ye = 196 kPa 00 0.00 0.05 0.10 0.15 020 0.25 se ‘Sustained static shear stress ratio, « Figure 32. Limiting residual pore pressure ratio versus a sustained static shear stress ratio in undrained cyclic simple shear tests (after Ishihara and Nagase 1980). os 02 Values Of fen aNd Nig determined at Yar OF 3%, ° 02 04 06 08 1 Cycle ratio, No/Nig Figure 33. Pore pressure as a fraction of the limiting residual value (Fy/ustin) versus the cycle ratio (Ne/Nug) for different sustained static shear stress ratios (@) (data from Boulanger etal. 1991). The stress-strain response of saturated sand to strain-controlled cyclic undrained loading in illustrated in Figure 34, which shows the results of an ICU triaxial test on loose Sacramento River sand (Seed and Lee 1966) and an ICU hollow-cylinder torsional shear test on 43os 7 : 06 - Sood & Lee 1860) I Figrono 199 [Ovonic na Oye (Ciironoyistertwson! sear, | ‘Sacramento River sand, A ‘Rold Bedford sand. Dyesen,p,=solen | Gynt ps t20 kPa | | 2 | g | © oh |g 07 q 8 3 fo | 3 £ 1 § & oo oo 2 5 -02-—} p| t = 02 4 L. o4 ‘03 02 07 0 01 02 09 ‘99 os 02 0 02 04 06 Ail strain, ©, ) ‘Shear strain (8) Figure 34. Stress-strain response of sand in strain-controlled cyclic undrained loading in a triaxial test (Seed and Lee 1966) and a torsional shear test (Figueroa et al. 1994) (both with permission from ASCE). medium-dense Reid Bedford sand (Figueroa etal. 1994). The imposed cyclic loading with an axial strain amplitude of 0.17% in the triaxial test caused the specimen to progressively soften (i.e., the mobilized shear stress decreased with each cycle of loading) as r,, progressively increased to 100% in about 16 cycles. The torsional shear test showed very similar stress-strain behavior, with r, reaching 100% after about 8-10 cycles of loading at a shear strain amplitude of 0.46%. After r, had reached 100% in each test, the specimens had nearly zero tangent stiffness in subsequent loading cycles. The negligible stiffness of the specimen after r,, reaches 100% is comparable to the very low stiffness that temporarily develops during stress-controlled loading after 7, has temporarily reached 100%. The difference is that the stress-controlled tests cause the strains to increase with each cycle of loading (exceeding the maximum value from the prior loading cycle), and this additional strain is what enables the specimen to transition to an incrementally dilative tendency. Dobry (NRC 1985) showed that the generation of 7, for a given number of cycles at different cyclic shear strain amplitudes falls within a relatively narrow band for a broad range of sand types, rela- tive densities, and consolidation stresses, as shown in Figure 35. The data in this figure show that 10 cycles at a shear strain amplitude of 44Dobry (1985) Cycle triaxial tests, _ Eight sands, Dy= 200 80%, 2 = 0.25 f0 1.9 atm, tO cycles. Excess pore waler pressure ratio, Au/o'y. 10° 10* 0 10° Amplitude of shear stain, 1. Figure 35. Excess pore water pressure generation versus shear strain amplitude in strain-controlled cyclic undrained triaxial tests on sand (NRC 1985). 0.3-1.0% would generally trigger r, = 100%. The volumetric cyclic threshold shear strain (y,,) is the shear strain amplitude below which there is no potential for volumetric strains or pore water pressure gen- eration. The value of yi, for sand is typically about 0.01-0.02% (e.g., Ladd et al. 1989, Hsu and Vucetic 2004), as shown in Figure 35 Energy-Based Evaluations of Undrained Cyclic Loading Behavior The dissipated energy (or work) per unit volume of soil has also been shown to be useful in describing liquefaction behavior (e.g., Davis and Berrill 1978, Law et al. 1990, Cao and Law 1991, Figueroa etal. 1994). Experimental studies by these investigators have shown that the generation of excess pore water pressures in saturated sand or silt is strongly correlated with the dissipated energy during undrained cyclic loading. The subsequent application of energy-based approaches to the assessment of liquefaction in the field requires that the seismic loading be represented in terms of energy. One approach hasbeen the use of Arias intensity (the integral of acceleration squared over time) to describe the energy in ground motions at different depths in a soil profile, as in studies by Egan and Rosidi (1991) and Kayen and Mitchell (1997). These initial studies suggested that energy-based approaches had promise, but this approach has not advanced further, because of (1) potentially large variations in Arias intensity versus 45depth in a soil profile and (2) questions about relating Arias intensity at a specific depth to that at the ground surface 2.3 Laboratory Testing of Field Samples and the Effects of Sampling Disturbance The most direct approach to estimating the cyclic strength of sat- urated sands is to obtain high-quality field samples and then test the specimens in cyclic laboratory tests. This approach requires knowl- edge of the effects of sampling disturbance (i.c., how the process of sampling the soil affects its subsequent loading behavior) and an understanding of the tests’ limitations in approximating the cyclic loading that is induced in situ during an earthquake. Cyclic triaxial devices and cyclic simple shear devices are the two laboratory test devices that are well suited for cyclic strength testing of undisturbed field samples. The loading conditions in a sim- ple shear device approximate the one-dimensional consolidation con- ditions beneath level ground before an earthquake and the cyclic shear deformations that develop during vertical propagation of horizontal shear waves. The triaxial device commonly uses isotropic consoli- dation stresses, so the resulting CRR must be corrected for the dif- ferences in mean effective consolidation stresses (i.e., the correction for K, conditions, as presented in Section 2.2, must be used). Triax- ial devices and most simple shear devices are unidirectional in their loading, and thus the CRR must also be reduced by about 10% for the effects of two-directional shaking in the field. Laboratory test results are also affected by nonuniformities of stress and strain that arise from imperfect boundary conditions, mem- brane compliance, and other experimental limitations. For example, Gilbert (1984) studied the effect of cyclic loading on specimen uni- formity by subjecting Dg = 40% sand specimens to undrained cyclic loading and then dissecting the specimens in layers to mea- sure the variation in Dz over the height of the specimen. The results in Figure 36 include one specimen that was cyclically loaded to an axial strain of about 11% and later found to have experienced dra- matic loosening at its top. The effects of these experimental limita- tions may become more significant as the sand specimen becomes stronger, because the weakening effect of a stress concentration or interface slip can then become the controlling factor on the over- all measurement of cyclic resistance. For this reason, some concerns remain about whether laboratory devices fully represent the cyclic 46(Top) 1 —~ ~~ ' specimens ‘Specimen cyciically (] loaded to double-amplitude {axial strains of 22%, Layer numbers ETT beled (unstrained) [ kpecimens (source Gilbert 1984; 1 afer NRC 1985) 1 (Bottom) 8 = “20-18-16 -14 12-10 -8 6 -4 20 2 4 6 8 10 Relative density change (%) Figure 36. Dg values of various horizontal layers from cyclically loaded triaxial specimens of sand at an initial average Dz of 40% (after NRC 1985). stress-strain behavior of dense sands under in-situ loading conditions. Nonetheless, careful laboratory testing can provide a valuable mea- sure of a soil’s cyclic stress-strain behavior, if the samples are truly undisturbed. Sampling disturbance has, however, been established as an even greater concern when field samples are tested. This issue has been well illustrated by the comparison of laboratory measurements of the CRR for clean-sand samples that were obtained via high-quality tube sampling techniques versus the preferred “frozen sampling” tech- nique. In frozen sampling, the sand is frozen in situ, and then samples are obtained by coring the frozen ground. The frozen samples are transported to the laboratory, mounted in a testing device, thawed, and tested. The freezing and thawing processes must be largely uni- directional, so that the expansion of the pore water upon freezing and its contraction upon thawing are accommodated by pore water flow rather than by volumetric straining of the soil skeleton. Frozen sampling, if done correctly, imposes very small net volumetric strain on the sand from the in-situ condition to the tested condition. In con- trast, tube sampling invariably produces significant volumetric strains as well as partial or total destruction of the sand’s fabric. 47Yoshiniet ot (1994) Frozen samphng yur piston sam bo "Yost e a (1984) Frozen samping (© Hyeraule piston sampler 4 oo || & ccuble-tube core boner (ogeye/20'3e) f0 5% strain in 15 cycles (acy/20'se) 0 5% strain in 15 cycles (a) oo) ° ° L ° 0 2 9 ao 0 70 40 6080100 (Neg forthe in situ sand (0, (%) ofthe sample as tested 120 Figure 37. Comparison of CRRs from cyclic triaxial testing of samples obtained via tube sampling and frozen sampling techniques (after Yoshimi etal. 1994). Figure 37 compares CRR values from cyclic triaxial tests on field samples of clean sands obtained via frozen sampling, hydraulic piston tube sampling, and double-tube core barrel sampling; the test data are from Yoshimi et al. (1994). Figure 37a plots the CRR values against the corrected SPT blow count of (Nj )¢o that was measured in the same sand deposit, and Figure 44b plots the CRR values against the De of the sand specimens as tested in the laboratory. The frozen sampling results show the expected trend of the CRR increasing with increasing values of Dg or (Ni)60, whereas the conventional tube sampling results show a relatively low CRR even for the densest sand specimens. One major factor is that tube sampling caused the looser sands to densify and the denser sands to loosen over the course of sampling, transporting, handling, mounting, and reconsolidation. Another major factor is that tube sampling disturbs the sand fabric, which is believed to reduce the fabric’s stability against cyclic loading and to particularly reduce any increases in fabric stability that may have been acquired in situ because of the processes of aging, prior stress and strain loading, overconsolidation, and cementation. The experimental results of Singh et al. (1982) and Goto and Nishio (1988), as presented in Figures 38 and 39, illustrate both the effects of prior strain history and the efficacy of frozen sam- pling techniques. The cyclic triaxial tests by Singh et al. (1982) in 48Cre} Virgin sample Data trom Goto & Nishio (1988) © Sample with strain history qT {for sand at Dy=90% gob © Atera treeze - thaw cycle | zg Altera freeze-thaw eycle g © confining stress romoved) & 06 | g Data trom Singh tal (1982) = {or sand at Dy=48% & 04 & S oz 02 t 4 (o) ° o Leet 2 5 10 2 80 100 200 5 10 20 80 100 200 Number of cycles to r, = 100% ‘Number of cycles to ta = 5% Figure 38. Effects of a freeze-thaw cycle on the cyclic resistance of saturated clean sand in undrained cyclic triaxial tests: (a) data from Singh et al. (1982) for sand at De = 48%, and (b) data from Goto and Nishio (1988) for sand at Dg = 90% (after Yoshimi et al. 1994). Figure 38a included tests on identical reconstituted sand specimens at Dr = 48%, with one set having been subjected to a cyclic shear strain of relatively small amplitude. The cyclic strain history did not measurably change the specimens’ De but nonetheless increased the specimens’ CRR by 30-40%. The cyclic triaxial tests by Goto and Nishio (1988) on sand at Dz = 90% provides a similar comparison (Figure 38b), but with the “strain history” specimens having a far more extensive strain history (10,000 cycles with an axial strain am- plitude of 0.05%). Again, the specimens’ Dp was not significantly altered, but the CRR nonetheless increased by about 80-100%. The CRR’s increase with prior strain history is attributed to the sand fab- ric’s becoming more stable as a result of the cyclic straining, even without becoming measurably denser, and this increase is considered illustrative of how a natural sand deposit’s CRR can be affected by the stress and strain it experiences over its geologic life. The CRRs of the sand specimens with prior strain history were unaffected by the process of freezing and thawing, as shown in Fig- ure 38, which indicates that the sand’s fabric was not seriously disrupted by the freezing and thawing process. Disturbance during the frozen sampling process was further evaluated in the laboratory 49sts on 305 mn da. samplos (source Singh eta. 1982) oe ‘Samples with seismic history Bees Vege g og 3 % e oN sg i, 8 a ~ & $ Pe Test data for 74 mm oko, samples obtained ‘by coring from 308 mm ofa. rozen samples. © Wan cyoe stain bistory © Winout eyote strain history 10 100 ‘Number of unitorm loading cycles to r, =100% Figure 39. Cyclic resistance of samples obtained by freezing and coring sand at an initial Dz of 60% and effective confining stress of 55 kPa (after Yoshimi et al, 1994, data from Singh et al. 1982). by Singh et al. (1982) by testing samples cored from two types of frozen reconstituted sands: those with a prior strain history and those with no such history. For both types, the samples obtained by frozen sampling techniques produced the same CRR as obtained for spec- imens that had not gone through the freezing and coring processes (ie., reconstituted samples that were prepared directly in the labora- tory test device), as shown in Figure 39. These results illustrate why frozen sampling techniques are believed to be capable of obtaining clean-sand samples that are sufficiently undisturbed to properly reflect the in-situ properties of a sand deposit. The CRR of a dense sand sampled by frozen sampling and tube sampling techniques was also compared with the CRR obtained for specimens of the same sand reconstituted to the same Da in the lab- oratory, as shown in Figure 40 (after Yoshimi et al. 1984). Reconsti- tuted specimens were prepared by air pluviation and moist tamping. ‘The CRR values for the tube samples and the reconstituted specimens 5015 undisturbed samples frozen samples (FS) tube samples (TS) Test resus from Yoshimi et (1968) | 0, 98KPe a5}- Cyclic stress ratio ao . L aul 1 10 100 — reconstituted samples ‘moist tamping (MT) air pluviation (PA) 15 1.0}- 0.5 Cyclic stress ratio ~~) 4.45 Un S145 im? ooh a 1 10 100 Number of cycles Figure 40. CSRs required to cause 5% double-amplitude axial strain of in-situ frozen samples (FS), conventional “undisturbed” tube samples (TS), samples reconstituted by air pluviation (PA), and samples reconstituted by moist tamping (MT) (after Yoshimi et al. 1984). were all substantially smaller than those for the frozen samples, which shows that important characteristics of this sand, as they existed in situ, were destroyed or lost by the disturbance produced during tube sampling and were not recreated by the reconstitution of specimens in the laboratory. The greater CRR for the frozen samples is attributed to the in-situ environmental factors (e.g., age, cementation, and stress and strain history) that increase the sand’s cyclic strength while not significantly affecting its density. 51Environmental factors can further affect the in-situ CRR through their influence on the in-situ lateral stress (i.e., through K,). In most cases, the environmental factors of prior cyclie straining, overconsol- idation, and aging would be expected to increase the in-situ Ky, with a corresponding increase in the in-situ CRR. 2.4 Field Processes Not Replicated in Laboratory Tests The excess pore water pressures generated by an earthquake will dissipate over time as pore water seeps out of the zones of higher excess pore water head (Ah = Au/yy) to zones of lower excess pore water head. The resulting pore water seepage that takes place both during and after earthquake shaking can greatly complicate the spatial occurrence of strength loss and deformation in a soil deposit. Laboratory element tests (e.g., triaxial or simple shear tests) do not replicate these effects, which have important implications for design practice, as discussed below. The redistribution of pore water pressures within a layered soil profile is illustrated by the profile in Figure 41. A layer of loose sand 3-9 m deep is overlain by dense silty sand and underlain by stiff clay. The total and effective vertical stresses and the pore water pressures arc shown for the conditions before carthquake shaking and after earthquake shaking has triggered “liquefaction” in the loose sand (KP) (Kea) (KPa) au (a) 0 400” 200 "400 © 20040" z000 tan” 200 wa 8 g = + — sy 2 a, Gan (nl concn Bote shaking Tin oee sand Figure 41. Stresses and pore pressures in a level-ground soil profile before earthquake shaking and for the condition in which r, = 100% has been triggered in the loose sand but no significant seepage has occurred. 52hPa) (kPa (kPa) au (kPa) 2 "10" 2000 100 a0 409” 2000 foo z Linuetecton ve ens i ‘upward soepape — 4 = Loose i Figure 42. Two consequences of upward pore water seepage for a layered soil profile during and after earthquake shaking, layer. This example further assumes that no excess pore pressures are generated in the overlying dense silty sand, that no significant pore water seepage has occurred, and that all soils weigh 20 KN/m?. The o/, has dropped to zero throughout the loose sand, while Au is now equal to o/,. atall depths. The Aw increases with depth because the vertical stresses increase with depth, and thus there is an upward hydraulic gradient (7 = Ah/AL) that initially is equal to unity throughout the loose sand layer. This condition of = 1 upward in the loose sand layer is analogous to the conditions that can initiate piping erosion (or quicksand conditions) during steady seepage problems. The conse- quence of this upward hydraulic gradient is seepage of pore water up through the soil profile, often concentrated along cracks and localized channels and producing sand and water boils at the ground surface. The complications that arise from the upward seepage of pore water during and after an earthquake are schematically illustrated in Figure 42, which is the same profile considered in Figure 41 but with the consequences of upward seepage included. The upward flow of pore water through the loose sand layer (ie., from point A toward point B) can increase the excess pore pressures within the overlying dense silty sand layer (which has a slightly lower permeability than the clean loose sand) during or after shaking. This may lead to lique- faction of the overlying dense silty sand, which otherwise may have had a sufficiently large CRR to have precluded liquefaction during 53Effective stresses reduced; Cracking ‘Sand loosened Sand with by outward high pore pressure bya igh pore pi Figure 43. Two mechanisms by which void redistribution contributes to instability after earthquake-induced liquefaction (NRC 1985, Whitman 1985). shaking if that sand had been perfectly undrained. A second possible complication of upward seepage can occur at even lower-permeability layers, such as the clay seam shown at a depth of 6 m (i.e., at point C). Upwardly seeping pore water can accumulate immediately below this interface, resulting in the formation of water films or water pockets under certain conditions. Note that the formation of a water film does not increase Aw, because the total w is already equal to the total overburden stress. Water films and water pockets may either slowly dissipate through the overlying layer or break through at concentrated cracks or piping channels that can extend up to the ground surface. The trapping of seeping pore water at interfaces between lower- and higher-permeability soils can result in local void ratio changes without necessarily forming water films. Whitman (1985) used the term “void redistribution” to describe this process, which empha- sizes the fact that zones can be getting denser locally while other zones are getting looser during the pore water pressure diffusion pro- cess, Figure 43 illustrates mechanisms by which void redistribution and upward seepage can contribute to deformations or instability of slopes. Figure 43a shows an infinite slope with liquefiable sand beneath a lower-permeability layer. Upward water seepage that is due to shaking-induced excess pore water pressures leads to a loosening (dilation) of sand at the top of the liquefied layer and densification (contraction) at the bottom. If the top of the layer becomes loose enough, then its critical-state strength can drop to a value less than that required for stability of the slope, after which deformations would develop along a localized shear plane at the top of the sand layer. Figure 43b is similar, but it also illustrates how the outward seep- age can weaken overlying layers by increasing their pore pressures 54and/or facilitating the formation of cracks. Intermixing of finer- and coarse-grained layered soils along a shear zone may also contribute to progressive strength losses as deformations increase (Byrne and Beaty 1997, Naesgaard and Byrne 2005). Redistribution of excess pore water pressures, and any associated void redistribution, may explain delays between the end of shaking and the time of failure that have been observed in several case histories (e.g., Elorza and Machado 1929, Akiba and Semba 1941, Kawakami and Asada 1966, Seed et al. 1975, Seed 1979, Marcuson et al. 1979, Hamada 1992, Mejia and Yeung 1995, Harder and Stewart 1996, Berrill et al. 1997). The major consequence of void redistribution is that the liquefied soil’s shear strength and stress-strain response do not solely depend on the pre-earthquake material properties and state (e.g., relative density and effective confining stress); rather, they can also reflect the re- sponse of the entire system. The effects of void redistribution are poorly understood and are not explicitly incorporated in current en- gineering practice. The formation of water films has been observed under level- ground conditions in physical model studies (Liu and Qiao 1984, Elgamal et al. 1989, Dobry and Liu 1992, Fiegel and Kutter 1994) and simple cylindrical column tests (Scott and Zuckerman 1972, Kokusho 1999, Kokusho and Kojima 2002). Figure 44 shows a wa- ter film that formed beneath a silt seam within a cylindrical col- umn of saturated sand as a consequence of void redistribution after impact-induced liquefaction in the sand (Kokusho 1999). These stud- ies have illustrated the importance of stratigraphy (layer thickness and sequence), permeability contrasts, and initial Dr on the thickness of water films. For sloping ground conditions, numerous physical model tests have identified shear strain localizations at the interface between a liquefied sand layer and an overlying lower-permeability layer (e.g., Arulanandan et al. 1993, Fiegel and Kutter 1994, Balakrishnan and Kutter 1999, Brandenberg et al. 2001, Malvick et al. 2002, Kulasingam et al. 2004), including cases in which the majority of de- formations were delayed until after shaking had ceased (e.g., Kokusho 1999, 2000; Kokusho and Kojima 2002; Malvick et al. 2004). For example, Figure 45 shows the postshaking deformed shape of a cen- trifuge model consisting of a uniform sand slope with embedded silt arcs and planes. The majority of the slope deformations were concen- trated in a localized shear zone in the sand immediately beneath the silt arc, and this localization formed after the shaking had ceased. A 55:LU) : sand layer = Silt seam Lower sand layer Water film Figure 44, A water film that formed beneath a silt seam in a cylindrical column of saturated sand after liquefaction (Kokusho 1999, with permission from ASCE). ‘Nevada sand | ooronsones St 0 100 200mm (model) rat 0-45 9.0m (prototype) Pore pressure transducer Figure 45. Localization of shear deformations along a lower-permeability interlayer within a saturated sand slope tested in a centrifuge (Malvick et al. 2008). 56dense array of pore pressure transducers showed that the onset of lo- calized shear deformation was caused by the upward seepage of pore water being impeded by the lower-permeability silt arc (Malvick et al. 2008). Void redistribution and localized shear deformations that are due to liquefaction-induced pore water seepage are complicated phenom- ena thatare difficult to quantitatively predict, fora variety ofreasons— such as heterogeneity of geologic deposits, formation of cracks and sand boils, and uncertainty in seismic shaking intensity and duration. However, the understanding of the basic void redistribution mecha- nism is increasing because of physical modeling studies and simple theoretical analyses. The results of physical modeling studies of liq- uefaction in layered soil profiles demonstrate that the residual shear strength of liquefied soil depends on all the factors that affect void redistribution and is therefore not uniquely related to pre-earthquake soil properties alone, as commonly represented by penetration resis- tances or by tests on “undisturbed” samples. The implications of these observations are addressed in Section 5.3 TRIGGERING OF LIQUEFACTION The assessment of potential liquefaction hazards involves two ques- tions: (1) will liquefaction be triggered by the earthquake ground motions under consideration and (2) what are the potential conse- quences of liquefaction having been triggered? This section addresses the procedures used to evaluate liquefaction triggering, including a discussion of geologic considerations, the analysis framework, in- situ testing, liquefaction triggering correlations, and an example of an analysis. 3.1 Liquefaction Susceptibility of Soil Deposits A study of local site geology is an essential part of characterizing the nature and possible extent of soils that are susceptible to liquefac- tion at a specific site, The extent or degree of liquefaction depends on the distribution of cohesionless sediments (gravels, sands, and very- low-plasticity silts) within the deposit and requires a sufficiently high water table for the sediments to be largely saturated. The most sus- ceptible sediments are fills and alluvial, fluvial, marine, deltaic, and wind-blown deposits. In addition, sediments are most susceptible to liquefaction when they are recently deposited, becoming more resis- tant as they become older. These factors are represented in the criteria recommended by Youd and Perkins (1978), as shown in Table 1, for identifying the likely presence of soils susceptible to liquefaction. These types of criteria are commonly used with geologic maps to produce regional maps of liquefaction hazards for planning or zoning purposes. Cohesive sediments (e.g., clays and plastic silts) can also de- velop significant strains and ground deformations during earthquake 59anof £19, MoT ua aiqeuren é[2907] eupiqag mop K9q— Moy AA, MOT aury é é ua peasdsoptay mop 9 Moy AIO, MOT any Uy, mop A9q MO} AIBA, Mor] aqqeue,, 1 [eDe|D, umouyug, us usin Way aiqeue,, $820] ano] A193, MOT a1BI=POW, us peoadsoptay sounq mol X19, MO] AIA, NOT ‘MOT peasdsapi ay, snyey, Mo] 19, MOr] a1RIOpOPY ua aqqene,, wintanyjod Avo A193, MOT a1BIOPO/ Wain aqqene,, eAeid pur outsnoe] Noy A19, MOT aTRIDPOY, ua, proadsapta\ vyjop wey pur vyj9q Mop 19 MOT AOA, NOT = peoidsapiay — sureyd pur sooviiay suey WO] 19, NOT MOT — aIvI9pOW, peodsopiqy sured pur wey jeranyyy Moy A19\, MO] oTeIDPOY, at ayqeesea £11290] e{dpoo}.| ao] 19, Mor] USI yaty Aaa, Jouurys sary Teyuounu0> dUaD0ISI9[q-Ad FHDD0JSIaJq —aUaD0]0H] saad QOS > ysodap ut yisodap yo ad, uonovjanby 0} aqndaasns oq pjnow ‘payeanyes wayar ‘syUDUIPaS SsaTUOISaYOS JLYJ POOYLPYLT SquaUIIpas ssajuorsoyoo Jo wonngqinysi, (AOSV Woy uorssturiad YIM “8161 subpiog pue pnog) Supyeys o1wstos Buoys SuLMp UoHoeonby] 0} sysodap j1os Jo AUyiqndaosng | 21qe], 60sso] AIA, sso] A19A, soy A109, sso] A19A, soy A19A, soy A109, MOT MOT MOT Moy AIA, MOT Moy a1eropoW, oye1OpO|A oyeIOPOAI MOT oyeIepOyAl u3IH oy Bry A219, Wat W3IH aH ayexopoy ystH ysry Aro, aqqenen, aqqeneA aqqeuren A112] aqqeurea A1[2007] peardsapia, peardsapia, aqqeirea 4112907] peordsapray Ty pasedutog, I1y pasediooup, Ty epypay a10yso10-, [euoosey AB10U9 AVAL MO[—YoRog ABr0u9 oavan YSI—youog ouuensy ued auoz [eISeO. 61loading—particularly where the sediments are soft and sensitive, there is a significant driving shear stress (e.g., slope or foundation load), and the shaking levels are sufficiently strong. However, because cohesive soils differ in shear strength characteristics from cohesion- less soils, different engineering procedures are required to evalu- ate how cohesive soils respond to seismic loading. For this reason, the term “liquefaction” is used in reference to the behavior of cohesionless soils (e.g., gravels, sands, and very-low-plasticity silts), whereas the term “cyclic softening” is used to describe the behav- ior of clays and plastic silts. Criteria and procedures for evaluating the potential for cyclic softening in cohesive soils are described in Section 6. Surface evidence of liquefaction has most commonly been asso- ciated with liquefaction occurring at depths of less than about 15 m. This is related to the fact that the shallower deposits are typically the youngest and therefore most susceptible to liquefaction. Lique- faction at greater depths beneath level ground surfaces may not be manifested at the ground surface and thus may go undetected. Liq- uefaction at greater depths is, however, a concern in many situations, including earth embankments constructed of looser material (e.g., hydraulic fill) or constructed over younger sediments. Historical records or geologic evidence of prior liquefaction at a site provides the most direct evidence that a soil deposit is susceptible to liquefaction, because soils that liquefy in one earthquake have often been observed to liquefy in subsequent earthquakes. Consequently, any evaluation of liquefaction hazards at a particular site would sig- nificantly benefit from a review of available historical records. Local sedimentary processes and the historical construction of fills are often clearly evident in aerial photos. For example, Figure 46 shows an aerial photo of Moss Landing, California taken in 1952 and another taken in 1987. Both photos provide a good view of how beach, eolian, fluvial, and estuarine depositional processes all come together to produce a rather complex depositional environment. In addition, detailed comparison of these two photos shows that areas of the shoreline have been modified by port development (areas of filling and dredging) or changed by the natural deposition and erosion processes. The areas of very recent natural deposits and loosely placed fills would be considered highly susceptible to liquefaction, and in fact they closely corresponded to the areas of more severe ground deformations caused by liquefaction during the 1989 Loma Prieta earthquake. 62Figure 46. Aerial photos of Moss Landing, California, showing the geologic setting; the photo at left is from 1952, and the photo at right is from 1987. When compared, these photos can identify shoreline movements, artificial fills, and recent construction (photos: USGS). 3.2 Analysis Framework for Developing Liquefaction Triggering Correlations Several approaches or frameworks have been proposed over the last 45 years for assessing the potential for triggering liquefaction. The most widely used approach has been the stress-based approach that compares the earthquake-induced cyclic stresses with the cyclic resistance of the soil. Strain-based and energy-based approaches are less common and are therefore not covered here. Earthquake-induced cyclic stresses beneath level-ground sites are attributed primarily to the effects of horizontal shaking. Figure 47 schematically illustrates the stresses and pore pressures acting on an element of soil beneath a level ground surface before and during hori- zontal shaking by an earthquake. Vertical shaking of this profile would produce additional transient changes in the total vertical stress, total horizontal stress, and pore pressure, but the vertical and horizontal effective stresses would be unaffected. This is why the effects of verti- cal shaking are not considered in the analysis of level-ground profiles. The induced horizontal cyclic stresses are normalized by the vertical effective consolidation stress (o,,,) to arrive at a CSR, which is then compared with the soil’s CRR, as shown in Figure 48. Liquefaction 63TTRTRTRTR TRISTE he -ED-~ afer t Consolidation stresses Undrained cyclic loading Bye Oye Uy y= Oye (Up * AU) = oye - AU (Ste +AOp) - (Uy +AU) Gp + (Aon - AU) = Koy She = Ko Wve oh Figure 47. Cyclic stresses on a soil element beneath level ground during horizontal shaking Peak cyclic shear Str0SS, ta, CSR and CRR Expected 2one of liquefaction yolc stress induced by Depth below ground surface Figure 48, The expected zone of liquefaction is determined by comparing the earthquake-induced eyclic stresses with the cyclic resistances of the soil. is expected at depths where the induced stresses exceed the cyclic resistances. The development of design procedures therefore requires meth- ods for estimating the earthquake-induced cyclic shear stresses (i.¢., the CSR) and the in-situ CRR. The earthquake-induced CSR is often estimated via the Seed-Idriss simplified procedure described in Section 3.3 64Site A: Surface evidence of "liquefaction" pr ree Site B: No surface evidence of "iquefaction” 717 Empirical relation between cyclic resistance and penetration resistance ‘Some measure of the earthquake-induced stresses Some in situ test indice for liquefaction resistance Figure 49. Schematic of the approach used to develop relationships between the in-situ CRR of sand and the results of in-situ tests. The in-situ CRR of sands could be evaluated on the basis of laboratory testing of field samples, but this would require the use of frozen sampling techniques if meaningful results are to be obtained, as discussed in Section 2.3. The cost of frozen sampling techniques is prohibitive for the vast majority of projects. Consequently, semi-empirical relationships are developed be- tween the in-situ CRR of sands and the results of in-situ tests, on the basis of compilations of case histories in which evidence of lique- faction has or has not been observed, as schematically illustrated in Figure 49. For each site, the earthquake-induced CSR and in-situ test results (¢.g., SPT and CPT results) are examined, and a critical layer is identified. A boundary line is then developed that separates the case histories in which “liquefaction” was observed (e.g., Site A in Figure 49) from case histories in which liquefaction was not observed (c.g., Site B in Figure 49). This boundary line is then adopted as pro- viding the relationship between in-situ CRR and the in-situ test index. The definition of “liquefaction” in this context refers to observa- tions, mainly from the ground surface, that are interpreted as meaning that sand at some depths must have developed high excess pore water pressures and significant strains (shear or volumetric). The observed effects may include soil and water boils, ground cracking, ground deformations (lateral or vertical), upheaval of buried structures, and settlement or failures of structural foundations. The interpretation of field observations is complicated by the fact that surface observations can be inconclusive in identifying the depths at which liquefaction 65probably occurred or in identifying the absence of liquefaction at depth. For example, sands may liquefy within localized pockets or depth intervals that are not extensive or shallow enough to produce any surface manifestations. The effectiveness of the resulting correlation also depends on the degree to which the in-situ test truly reflects the soil’s resistance to cyclic loading. For example, Seed (1979b) explained that both the penetration resistance from an SPT or a CPT and the CRR of a sand would be expected to respond similarly (i.e., increase or decrease) to changes in factors such as Dr, K,, age, cementation, or seismic history. This observation provides qualitative support for the expec- tation of a reasonable correlation between penetration resistance and CRR, although a more quantitative understanding of how these factors might affect the resulting correlation remains a desirable goal. Our understanding of the cyclic loading behavior of saturated sands, on the basis of experimental findings, indicates that the in-situ CRR of sand also depends on the duration of shaking (i.e., this is analogous to the number of loading cycles), the effective overburden stress (i.¢., o/. and K,), and the presence of sloping ground (ie., K,). The case history database is not extensive enough to empirically define these individual effects, in part because they are correlated with other factors and relationships in the analyses, so these effects are instead accounted for by using relationships that are derived from our fundamental understanding of sand behavior. Relationships for these three effects are discussed in Sections 3.6-3.8. The organization of case history data onto a single graph of CSR versus in-situ test measurement can then be achieved by adjusting the data points for each case history to a common reference con- dition (i.e., one duration, one overburden stress, and level-ground conditions). These adjustments are performed by using the same relationships that will be used in the final design procedure, and thus the final liquefaction correlation should strictly be used only in combination with the relationships that were used to process the case history database. 3.3 Simplified Procedure for Estimating Earthquake-Induced Stresses The shear stresses induced at any depth in a soil deposit with a level ground surface during an earthquake appear to be due primar- ily to the vertical propagation of horizontal shear waves. Analytical 66Maximum shear stress Camda) t (maa Figure 50. Schematic for determining maximum shear stress, Tmax, and the stress reduction coefficient, rg. procedures are available to calculate these stresses if the character- istics of the soils comprising this deposit and the input motions are known. Such information is not available for most of the “liquefaction/ no liquefaction” case histories that have been used to develop corre- lations on the basis of field observations. In addition, borings drilled for most projects seldom extend to the depths needed to define the soil profile in sufficient detail for site response studies. For these reasons, the simplified liquefaction evaluation procedure (Seed and Idriss 1971) for calculating the induced shear stresses, and hence the CSR, continues to be widely used. If the soil column above a depth z behaved as a rigid body (Fig- ure 50), then the maximum shear stress on the base of that column can be computed as the product of its mass and maximum horizontal surface acceleration: w a, Cmdr = nae = oun (20) where dmax is the maximum ground surface acceleration, y is the average total unit weight of the soil above depth z, and oy is the total vertical stress at depth z. In reality, the soil column behaves as a deformable body, and hence the maximum shear stress will differ from the value for a rigid body with the same maximum ground surface acceleration, The maximum shear stress for a deformable body can be determined from dynamic site response analyses, and the results compare to the rigid body case as follows: (Tinax)d = 1a (Tmax)r (21) 67where rq is a shear stress reduction coefficient. The variations of (Tmax) and (Tmax)a Will typically have the form shown in Figure 50, and thus the value of 7g will decrease from a value of | at the ground surface to lower values at large depths. One-dimensional dynamic site response analyses have been used to develop simplified expressions for rg. These analyses have shown that ry is particularly dependent on the earthquake ground motion characteristics (e.g., intensity and frequency content), the shear wave velocity profile of the site, and the nonlinear dynamic soil properties (Seed and Idriss 1971, Golesorkhi 1989, Idriss 1999, Cetin et al. 2004). Idriss (1999), in extending the work of Golesorkhi (1989), per- formed several hundred parametric site response analyses and con- cluded that, for the purpose of developing liquefaction evaluation procedures, the parameter ry could be adequately expressed as a func- tion of depth and earthquake magnitude (M). The following expres- sions were derived by using those results: ra = exp(ae(z) + B(z)M) (22) af i ete) =—1.012~1.126sin (TF 45.133) (23) af NZ A(z) = 0.106 + 0.118 sin (Fa + 5.142) (24) in which z is depth in meters, M is moment magnitude, and the ar- guments inside the sine terms are in radians. Equations 22-24 are mathematically applicable to a depth of z < 34 m. However, the un- certainty in ry increases with increasing depth, so these equations should actually be applied only for depths that are less than about 20 m. Liquefaction evaluations at greater depths often involve spe- cial conditions for which more detailed analyses can be justified. For these reasons, it is recommended that the CSR (or equivalent rg val- ues) at depths greater than about 20 m be based on site response studies—provided, however, that a high-quality site response calcu- lation can be completed for the site. Site response analyses for this purpose require sufficient subsurface characterization of the site and must account for variability in the possible input motions. Figure 51 shows plots of rg calculated by using the above recom- mended expressions for M values of 5.5, 6.5, 7.5, and 8. Also shown. in this figure is the average of the range published by Seed and Idriss (1971). The information in Figure 51 indicates that the average of that 68Stress reduction coefficient, ry 03 o4 os 08 o7 oa 09 40 oc | +/+} — g © Foaap figs Paniahod verage a 2 a by Seed & ldriss (1971) | $ | | 3 Bw + <__— 3 | | 2 16| —__; — - = t 3 | z | 5 20 +— } = 5 Magnitude: M=5% = M=6% = M=7% M=8 & 24 4 |. 28 Figure 51. Variations of the stress reduction coefficient ry with depth and earthquake magnitude (Idriss 1999). range is comparable with the curve calculated by using the revised expressions with M = 7.5 for depths up to about 14 m. Cetin et al (2004) proposed rg values that are a function of depth, earthquake magnitude, level of shaking, and average shear wave velocity over the top 12 m of the site, whereas Kishida (2008) proposed ry values that are a function of depth, a response spectra ratio (as a measure of frequency content), level of shaking, and shear wave velocity profile. The rg values by Kishida are in good agreement with those in Fig- ure 51 when the shear wave velocity profile corresponds to sandier soil deposits and are smaller than those in Figure 51 for depths greater than 6 m when the shear wave velocity profile corresponds to sites with predominantly soft fine-grained soils. The ry values by Cetin et al. reduce more quickly with depth and are significantly smaller than those by Kishida or Idriss. ‘The earthquake-induced cyclic stress time series involves nu- merous cycles at different strengths, with the damaging effects of the irregular time series depending on the number of cycles and the stress magnitudes for each cycle. Various studies have shown that an irreg- ular time series can be approximated by a uniform cyclic stress time 69series with an equivalent number of uniform cycles that depends on the uniform cyclic stress amplitude (as described in Section 3.5). Consequently, Seed and Idriss (1971) chose to represent earthquake-induced cyclic stresses by using a representative value (or equivalent uniform value) equal to 65% of the peak cyclic stress. ‘The corresponding earthquake-induced CSR is therefore computed as ve ‘ve CSR = 0.65 = 0,652 Sm, (25) 9, g The choice of 0.65 to represent a reference stress level is somewhat arbitrary, but it was selected in the beginning of the development of liquefaction evaluation procedures in 1966 and has been in use ever since. More importantly, the overall liquefaction evaluation procedure would be essentially unaffected by the choice of a different reference stress ratio, provided that the adjustment factors for the duration of shaking and the empirically derived liquefaction correlations were all derived for that reference stress (see Section 3.5), 3.4 In-Situ Tests as Indices for Liquefaction Characteristics The in-situ tests that have been most widely used as indices for evaluating liquefaction characteristics include the SPT, CPT, BPT, large penetrometer test (LPT), and shear wave velocity (V,) test. The SPT was used first in developing liquefaction correlations and was the most common in practice up through the 1990s. The CPT has a number of advantages, however, that have made it the primary site characterization tool in certain geologic settings. The BPT, LPT, and V, tests tend to be used in special situations and thus are used less often than the SPT and CPT in liquefaction evaluations. Each of these tests is discussed separately below, after which the complementary roles of site investigation techniques and the advantages of pairing techniques (c.g., CPT soundings and SPT borings) are discussed. SPT The SPT is a widely available sampling method that indicates a soil’s compactness or strength. The SPT measures the number of blows (N) by a 140-pound hammer falling freely through a height of 30 in. that are required to drive a standard split-spoon sampling tube (2 in. outside diameter, 1°/g in. inside diameter) to a 12-in. depth after an initial seating drive of 6 in. The thick walls of the split-spoon sampler 70make it rugged enough to use in a wide range of soil conditions. The soil samples are too disturbed for meaningful engineering property tests but are nonetheless suitable for index testing (c.g., gradations and Atterberg limits). The SPT blow count or “N’ value” is low in soft or loose soils and increases with increasing stiffness or strength of the soil, and this value can thus be used as an index of the soil’s in-situ strength or compactness. Because of its economical nature and ruggedness, this test is widely used to evaluate the spatial variability of a soil deposit, and consequently the NV value has been correlated with a wide range of engineering characteristics (e.g., Kulhawy and Mayne 1990). Satisfactory use of the SPT for liquefaction analyses requires that the apparatus and procedures conform with the ASTM D-6066 standard, The features of the procedure recommended by Seed et al (1985) for liquefaction evaluations, as summarized in Table 2, meet this standard and continue to be followed in practice. Among the most important variables is the amount of energy de- livered to the drill rod stem by each impact of the SPT hammer. The Table 2 Recommended features of SPT procedure for liquefaction evaluations (after Seed et al. 1985, with permission from ASCE). Feature Description Borehole Rotary borehole diameter of 4-5 in, with drilling mud for stability; the drilling mud should be kept thick enough, and the hole should always be full Special care is required when pulling rods out of the hole, to avoid suction. Drill bit Upward deflection of drilling mud (e.g,, tricone or bafiled drag bit) Sampler O.D.=2in 1. D. = 1.38 in. (constant; ive., no room for liners in barrel) Drill rods ‘Aor AW for depths < 50 ft NN, BW, or NW for greater depths Energy delivered to 2,520 in.-Ib. (i.e., 60% of theoretical maximum sampler of 140 Ibs. falling 30 in.) Blow count rate 30-40 blows per minute Penetration resistance Measured over a range of 6-18 in. of penetration count into the ground 7range of delivered energy can be 30-90% of the theoretical maximum energy (the 140-Ib hammer multiplied by its 30-in. drop height), de- pending on the amount of energy lost to frictional and mechanical resistances that depend on the type of equipment and its operating condition. The N value is essentially inversely proportional to the delivered energy (Schmertmann and Palacios 1979). In U.S. practice, the delivered energy is commonly about 55-60% of the theoreti- cal maximum energy (Kovacs et al. 1983), and therefore Seed et al (1984) recommended adopting Ngo as a standard. The value of Noo is computed as ERm Noo = Nm (26) where N,, is the measured blow count, ER, is the measured delivered energy ratio as a percentage, and N¢o is the blow count for an energy ratio of 60%. The ratio of ER,,/60 is also referred to as an energy ratio correction factor, Cy. The energy ratio is one of the most impor- tant variables in obtaining reliable Ngo values. Therefore, it is impor- tant that energy ratios be routinely measured as part of liquefaction evaluations. Additional correction factors may be needed to arrive at a more standardized value of Neo. The resulting relationship is given by Noo = CeCeCrCsNm 27) in which Cg is the energy ratio correction factor described above, Cz is a correction factor for borehole diameter, Cr is a correction factor for rod length, and Cs is a correction factor for a sampler that had room for liners but was used without the liners. Suggested ranges for each factor are in Table 3. The borehole diameter and sampler correction factors can be important in interpreting older borings, but for future applications it is recommended that appropriate standards be followed so that the Cy and Cs factors are unnecessary (i.e., each is equal to unity). The short rod correction factor Cp (Table 3) is intended to ac- count for how the energy transferred to the sampling rods is affected by rod length (e.g., Schmertmann and Palacios 1979). The hammer- to-anvil impact sends a compressive stress wave down the sampling rods, which then reflects from the sampler as a tension wave. This ten- sion wave returns to the anvil, where it causes the hammer to bounce off the anvil. Schmertmann and Palacios (1979) concluded that the energy imparted to the sampling rods during this primary impact 72