You are on page 1of 251

SOIL STRUCTURE: IN SITU

PROPERTIES AND BEHAVIOR

K. Arulanandan

SOIL STRUCTURE:
In Situ Properties and Behavior

K. Arulanandan
Department of Civil and Environmental Engineering
University of California, Davis

July 2003

Foreword
The sadness of realizing Arul has left us was my predominant feeling as I started
writing this foreword. But as I kept writing, trying to bring back the most memorable
moments of my acquaintance with him, a smile comes over and my feelings change to
happiness to have met him.
Kandiah Arulanandan, or Arul as he was known to us, was my friend despite the
age difference. Arul had the ability to adjust his age towards the younger side. My first
impression of him when I came to UC Davis as a young Assistant Professor was that he
never knocks on the door before he enters. Later I understood that this was part of his
infusing personality. He literally forced me to deal with soil mechanics, for which I am
grateful to this day. I still remember him storming into my office during the first years of
our acquaintance waving a bunch of papers on soil fabric studies and experiments he
wanted me to work on. And I happily did, until days later I was finding a paper missing
which Arul had it but of course he forgot to give it to me at first place, costing both of us
many lost hours. He was not very well organized; the few times his office was neat and
orderly was when he had just moved into a new office when the Civil Engineering
department was changing buildings, and the faculty offices. Yet, within the mess of
papers in his office, Arul was the only one to know where a paper was hiding. The order
he had was of a different kind than most of us know, rooted in his head with the visionary
ideas he lived by. In his personal life with his students and friends, Arul was very
generous and the main images that come to my mind are those of laughing heartily
together on the social occasions we so often had. Arul may have appeared to do things
spontaneously, and this may have been true in many cases, but he also had plans of great
depth for his field, his students and his colleagues. The outgrowth of this book is perhaps
the greatest of his personal plans; pity he is not around to see it materializing, unless he
is.
Arul was a person with a vision for modern methods of numerical analysis in
geomechanics, an even more remarkable trait realizing that he was a traditionally trained
geotechnical engineer with no immediate grasp of advanced numerical techniques,
although he was very adept in soil mechanics theories of the Cambridge school of
thought. The VELACS project realized at UC Davis under the directorship of Arul was
perhaps the largest gathering of renowned researchers in geomechanics with the goal of
validating constitutive and numerical soil modeling by true predictions of centrifuge
experiments considered as boundary value problems. The late Cliff Astill of NSF who
funded VELACS has said that upon this accomplishment Arul had moved on to being a
statesman.
The true technical strength of Arul, though, was on soil characterization by
electrical methods, which constitutes the core aspect of this book and is a testimony of
Aruls contribution that will stay with us and future generations of geotechnical
engineers, as long as these issues are relevant to the profession. It was not only the
theoretical aspects that lead to the refinement of existing theories and development of
new ones that Arul with his students and colleagues worked on, but also the practical
application of such findings that Arul was so keen to pursue by working on a probe and
applying it in the field.

iii

But before I close this brief foreword for my friend, I shall refer to the other big
testimony of Aruls contribution: the excellent students he had supported and guided
throughout his years as a Professor and advisor. He had an innate ability to choose the
best of them. In the Civil Engineering Department of UC Davis there is no other faculty
member, past or present, with so many of his PhD advisees becoming Professors at USA
Universities. I was personally benefited by being able to also teach these students upon
the request of Arul, and eventually learn from them. All of them are now my friends. And
I am sure that I express their feelings when I say: Arul, we miss you.
Yannis F. Dafalias
Davis, January 2008.

iv

Preface
Soil Structure: In Situ Properties and Behavior is designed as a text for
undergraduate and graduate students in Soil Mechanics. An intensive effort was made to
correlate soil structure to soil properties. We have then used these relations to help
redirect soil properties using nondestructive in situ methods. Examples and problems are
included to illustrate the application of these methods. This text has been successfully
used in ECI 174 and ECI 283 classes at the University of California, Davis. Soil
Structure: In Situ Properties And Behavior has been principally written for students;
however, geotechnical engineers, technical geologists, environmental engineers, and
instrumentation specialists should also find it valuable in improving their nondestructive
in situ soil testing capabilities.
The focus of this text is on the quantification of soil structure, the establishment
of property-structure relationships, the mechanisms causing the behavior of soils, and the
properties that must be determined to analyze the influence of local site characteristics
and environmental factors on soil behavior. The book is divided into seven chapters.
Chapter 1 describes the need for nondestructive soil and site characterization, and it
briefly compares conventional methods and electrical methods. Chapter 2 describes the
need for a method to assess the factors that determine soil structure: mineralogy, pore
fluid composition, stress history, particle orientation, cementation, water content, and
density. Chapter 3 is devoted to quantifying soil structures and clay properties. This
chapter derives equations characterizing the electrical response of aggregate systems by
extending the MaxwellWagner and Fricke equations. Here, microscopic approach
predicts electrical dispersion characteristics of soil containing clay minerals. A
macroscopic approach, known as the Three-Element Model, is also presented. Chapter 4
explores the relationship between the electrical dispersion characteristics of soil and
various engineering properties. Chapter 5 gives an overview of the equipment used to
nondestructively determine electrical dispersion data in situ. Chapter 6 discusses
characterization of non-clay minerals. The chapter presents theoretical development of
the formation factor tensor and anisotropy index along with an example evaluating the
formation factor and anisotropy index of Monterey 0 sand using theoretical equations.
Chapter 7 describes the successful use of electrical methods to predict two complex soil
behaviors. Erosion potential was predicted using a combination of dielectric dispersion
behavior and Sodium Absorption Ratio (SAR), while frost heave potential was evaluated
using the dielectric constant and c parameter (from the ThreeElement Model.) Appendix
A presents example applications using the formation factor method of nondestructive soil
characterization and SUMDES program for site response analysis. Derivations for double
layer equations and the three element model are given in Appendix B. Sample laboratory
experiments for understanding the methods presented are given in Appendix C.
The book is an outgrowth of an accidental observation I made in 1960 when I was
the senior materials engineer for the Gold Coast government. During this period, several
Russian geologists visited the Gold Coast to explore for oil. One method used to detect
hydrocarbons is to determine the electrical conductivity of a soil. During the
measurements a change in the frequency of measurement showed a change in the

electrical conductivity. This accidental observation led to the non-destructive


characterization of soils to quantify soil structure.
I pursued this idea at UC Berkeley as my Ph.D. thesis under the direction of
Professor J. K. Mitchell, to whom I owe a great deal of gratitude. The late Professor K. S.
Cole, head of the Neurological Division of the National institute of Health, who studied
the reactivity of nerves and tissues to electrical impulses, also helped and impressed me
tremendously during my time at UC Berkeley. I am also greatly indebted to Professor K.
L. Babcock of the Department of Soil Chemistry, UC Berkeley, who was on my
dissertation committee. During this time, Professor K. S. Spiegler continuously supported
my work and is responsible for introducing me to the application of the three element
electrical network model; I thank him for all his assistance. I also thank the late Professor
H. B. Seed for all his continued support throughout the years.
I would like to thank my fellow faculty members at UC Davis, especially
Professors George Tchobanoglous, Dan Chang, Yannis Dafalias, and the late Professor
Ray Krone, for their encouragement and assistance. I am also indebted to all of my
former students whose contributions made this book possible. I owe special gratitude to
the following students: Prof. Abdulla, Prof. Amin Alizadeh, Prof. A. Anandarajah, Dr. K.
Arulmoli, C. Baker, R. Basu, T. Blacic, G. Cross, Ernie Gillogly, R. Heinzen, Julie
Hiscox, Prof. B. Kutter, John Livingston, Prof. N. Meegoda, Prof. K. Muraleetharan, Dr.
H. Rashidi, C. Reyes, P. Ribbon, Prof. S. Sargunam, R. Sharlin, Dr. K. Sivathasan, Dr.
Scott Smith, J. Sybico Jr., Dr. C. Yogachandran. Special thanks go to Garrett Broughton
for his continued help during the preparation of this book.
I am grateful to Prof. A. Anandarajah, Prof. Malcolm Bolton, Prof. James
Mitchell, and Prof. Stein Sture for their assistance in reviewing the book. I am also
indebted to Prof. Ron Scott, Prof. Andrew Schofield, Prof. Leonard Herrmann, and Prof.
Yannis Dafalias. Their assistance is greatly appreciated. I also gratefully thank Dr.
Clifford Astill for giving me the opportunity to develop some of the material presented in
this book.
Finally I owe my deepest gratitude to my wife Rajes, for her patience,
understanding, encouragement, and love during the preparation of the book.
K. Arulanandan
April 2002
Dept. of Civil and Environmental Engineering
University of California, Davis

vi

CONTENTS

1. NEED FOR SOIL AND SITE CHARACTERIZATION

1.1 Importance of Soil and Site Characterization

1.2 Historical Development of Soil Classification

1.3 Development of In Situ Methods

1.4 New Developments in Non-Destructive Methods of In Situ Soil and Site


Characterization
1.5 Problems

2. SOIL STRUCTURE

9
10

12

2.1 Introduction

12

2.2 Mineralogy

13

2.3 Particle Size and Shape

17

2.4 Particle-Solution Interface Characteristics

18

2.5 Fabric

22

2.6 Conclusions

26

2.7 Problems

27

3. ELECTRICAL DISPERSION AND SOIL STRUCTURE

28

3.1 Introduction

28

3.2 Electrical Response Characteristics of Clay Minerals

29

3.3 Microscopic Model of the Electrical Dispersion Characteristics of Soils

34

3.4 Soil Structure Effects on Electrical Dispersion

37

3.5 Macroscopic Approach - The Three-Element Model

42

3.6 Intercluster Void Ratio as a Function of Three-Element Model


Parameters

45

3.7 Three-Element Model Optimization

46

3.8 Summary

56

vii

3.9 Problems

57

4. ELECTRICAL DISPERSION AND ENGINEERING PROPERTIES

61

4.1 Introduction

61

4.2 Electrical Dispersion Method of Soil Characterization

61

4.3 The Magnitude of Dielectric Dispersion and Swell Potential

65

4.4 The Magnitude of Dielectric Dispersion, 0 , and Compression Index

66

4.5 Critical State Friction Angle

67

4.6 The Magnitude of Dielectric Dispersion, 0 , and


Cation Exchange Capacity

68

4.7 and Porosity

69

4.8 Specific Surface Area

72

4.9 The Slope of the Unload-Reload Line,

74

4.10 Hydraulic Conductivity of Clays Using the Electrical Dispersion


Method

76

4.11 Summary

80

4.12 Problems

80

5. EQUIPMENT FOR MEASURING ELECTRICAL DISPERSION OF SOIL

88

5.1 Laboratory Equipment


5.2 In Situ Measurement Equipment

88
109

6. FORMATION FACTOR, SOIL STRUCTURE, ENGINEERING PROPERTIES


AND BEHAVIOR

113

6.1 Introduction

113

6.2 Formation Factor

114

6.3 Soil Properties using the Formation Factor Method

125

6.4 In Situ Formation Factor Probe

136

6.5 State-Parameter-Dependent Sand Models

140

6.6 Problems

143

viii

7. SOIL BEHAVIOR: EROSION IN RELATION TO FAILURE OF EARTH DAMS


AND FROST HEAVE

147

7.1 Erosion

147

7.2 Frost Heave

163

7.3 Problems

167

APPENDIX A EXAMPLE APPLICATIONS OF THE FORMATION FACTOR


METHOD

170

A.1 Evaluation of Site Responses and Deformations of the Instrumented


Wildlife Site, California

171

A.2 Before-the-event Prediction of the Response of Treasure Island


Instrumented Site to Expected Earthquakes

APPENDIX B DERIVATIONS
B.1 Double-Layer Theory

APPENDIX C LABORATORY EXERCISES

197

229
229

232

C.1 Dielectric Dispersion

232

C.2 Three-Element Model

234

C.3 Dielectric Cone Probe

236

C.4 Erosion

237

ix

Chapter 1
Need for Soil and Site Characterization
1.1 IMPORTANCE OF SOIL AND SITE CHARACTERIZATION
Over the past eight thousand years, we humans have built structures on the earth ranging
in size and complexity from simple mud huts to towering skyscrapers. Except for airplanes in
flight and ships in the ocean, virtually all our infrastructure systems are built on or in soil. So the
question must be asked, how can we reliably predict soil behaviors due to natural hazards, such
as earthquakes, erosion, frost heave, swelling, and liquefaction, and man-made hazards, such as
contaminant migration. To understand soil behavior, proper soil characterization is critical.
1.2 HISTORICAL DEVELOPMENT OF SOIL CLASSIFICATION
Up until the 19th century, every problem in soil mechanics was referred to a geologist.
The suitability of a site for constructing a large structure was determined by an extensive
geological survey. The geologist or soil engineer would classify the soil at the site into broad
categories. Each soil type was assigned a particular range of strength values based on the bearing
capacity of the soil. When a structure designed on the basis of this classification failed, the
failure was attributed to an inaccurate geological survey.
In 1910, an engineer named Karl Terzaghi started investigating the causes of such
failures. For two years of hard labor, he collected data pertaining to a number of case histories
of foundation failures. The geological surveys carried out at sites where failure had occurred
were found to be correct and adequate. Terzaghi concluded that it was the nature of the surveys
themselves that led to the failure of foundations which were thought to be safe. The methods
used to determine soil stability were somehow lacking. Thus, the actual causes of the failures
were unknown.
In 1916, Terzaghi realized that the cause of the failures was improper soil classification.
According to Terzaghi (1957):
After I arrived in Istanbul, late in the fall of 1916, the first topic of my
contemplation was the dismal failure of my attempt to get useful information out
of my collection of case records. Going over my data once more, I realized, quite
suddenly and surprisingly, the classification of the materials whose performance
was observed was, from an engineering point of view, inadequate, although for
geological purposes it was quite satisfactory.
The subsurface materials
encountered at the sites of construction operations were divided into categories
such as coarse and fine sand, or soft and stiff clay, each of which included
minerals of widely different engineering properties. Hence, I concluded, this time
correctly, that engineering geology cannot possibly become a reliable tool in the
hands of earth works engineers unless and until we acquire the capacity to assign
to each material of the earth numerical values which make it impossible to
mistake it for another with significantly differing engineering properties.
1

Atterberg made an apparent improvement in soil classification, which Terzaghi


introduced into geotechnical engineering. In 1908, Atterberg worked out a soil classification
scheme based on particle size fractions in decimal multiples of 2 micrometers and 6
micrometers. The "clay fraction" was defined as the percentage by weight of particles smaller
than 2 micrometers. However, Atterberg realized that particle size alone was insufficient to
classify cohesive soils and decided that a measure of their plasticity provided the additional
criterion required. The Atterberg Limits are water contents at certain stages in soil behavior:
Liquid limit (LL or wL) marks the lower limit of viscous flow
Plastic limit (PL or wP) marks the lower limit of the plastic state
Plasticity index (PI) identifies the range of water contents over which the soil is
plastic: PI = LL - PL
With some modification in nomenclature, geotechnical engineering has used this system for the
last 80 years.
The nature of most systems of soil classification, as well as their great diversity, led to a
critical review of the problem (Casagrande, 1948) and the proposal of the Unified Soil
Classification System, adopted in 1952 by the U.S. Corps of Engineers and Bureau of
Reclamation, and subsequently by many other organizations (USBR, 1963). Under the Unified
Soil Classification System (USCS), the soil is first separated into different grain size categories.
Clay minerals are distinguished from non-clay minerals based on particle diameter with an
arbitrary demarcation (2 m). The soil passing the Number 40 sieve is removed from the rest of
the soil. Classification is then based on Atterberg Limits tests performed on the remolded clay
fraction of a soil sample passing the Number 40 sieve.
In a series of five papers, Lambe and Martin (1953-1957) reported compositional data
and their relationships to engineering properties for a large number of soils. Their studies found
that the plasticity characteristics of natural soils are less than would be predicted from the
percentage of clay minerals in the soil and the corresponding properties of pure clay minerals. In
Figure 1.1, soil activity is plotted against the amount of two specific clay minerals present in the
soil. Activity of soil is defined as the plasticity index divided by the percent of clay-sized
particles ( 2 m).
As can be seen in Figure 1.1, natural soils plot below the range of activity that would be
expected based on the type and amount of clay minerals present in the soil and the type of ions
present in the pore fluid. This discrepancy could be due to the fact that the percentage of clay
minerals is likely to be considerably more or less than the percentage of clay size particles (2
m) as shown, for example, by the data in Figure 1.2 (Lambe and Martin, 1953-1957, and Basu
and Arulanandan, 1973). Aggregation, cementation, and interstratification of clay minerals have
been cited as the reasons for clay mineral contents in excess of the clay size content.

Figure 1.1
Activity vs. Clay Content
(After Lambe and Martin, 1953-1957)
For these reasons, Atterberg Limits are often not reliable for properly estimating soil
properties. Furthermore, for two soil samples with the same gradation, both containing a small
percentage of fines (e.g., 5%), it would not be possible to determine Atterberg Limits for either
sample because there are not enough fines to test. However, if the fines in one sample contained
clay minerals, and the fines in the other contained non-clay minerals such as silica flour, the
mechanical properties of the two soils would be different. Lambe and Martin (1953-1957) found
that Atterberg Limits could not always be used to accurately predict compressibility. For soils
indigenous to St. Louis and elsewhere, they found that "very large changes in volume can occur
with changes in load or changes in moisture conditions, and that these changes are partially
reversible. The Atterberg Limits of the St. Louis soil (Sample 19, [L.L.] = 58, [P.L.] = 22) do
not suggest such unusual compressibility characteristics" (Lambe and Martin, 1953-1957).

Figure 1.2
Relation Between Clay Content and Clay Size
for Several Soils (After Lambe and Martin, 1953-1957, and
Basu and Arulanandan, 1973)
The parameters used to identify swell potential, activity, and percent finer than 2 m, are
not adequate to evaluate percent swell for some natural soils. Seed, et al., (1962) used activity
and percent finer than 2 m to evaluate the swell potential of a soil. Their results were valid for
the artificial soils they tested. A study of both natural and artificial soils evaluated the ability of
the activity index and the percent clay size to predict swell potential (Basu and Arulanandan,
1973). Table 1.1 tabulates the composition, percent clay size, activity, and percent swell for
each sample. Figure 1.3 plots the percent swell for the samples on the chart developed by Seed,
et al., (1962). The percentage of total swell predicted for samples 1PB, 3SC, and Marysville Red
(MR) is significantly lower than the measured value. Thus, while the predicted values are
reasonable for artificial soils, the predictions are inconsistent for some natural soils.

Table 1.1
Composition and Swell Properties of Natural and Artificial
Soils (After Basu and Arulanandan, 1973)

Figure 1.3
Classification Chart for Swelling Potential
(after Seed, et al., 1962)
The soil samples from Table 1.1 are also plotted in Skempton's (1953) activity chart,
Figure 1.4. The three solid lines represent the three clay minerals, montmorillonite, illite, and
kaolinite, from highest to lowest activity. These lines were developed by mixing quartz sand
with varying percentages of the three clay minerals. Comparison of the samples to the solid
lines shows that activity does not consistently follow mineralogy. For example, Soil 15 with
15% montmorillonite is expected to plot near the solid line representing montmorillonite with an
activity of 7.2. Soil 15 actually plots below the line representing kaolinite (activity = 0.38),
suggesting an entirely different mineralogy. Note that samples 1PB, 3SC, 11, 9, 3, and 8 show
similar deviations. Thus, activity cannot be used to predict soil mineralogy.

Figure 1.4
Clay Fraction and Plasticity Index of Some Natural Soils in
Relation to the Activity Chart of Skempton (1953)
In the construction of a section of road in Marysville, California, the subgrade consisted
of a silty soil with the following properties: LL = 46, PI = 11, % < 2 m = 20, % < 1 m = 14,
% Sand Size = 10, % Silt Size = 70, and % Clay Size = 20 (Arulanandan, 1974). Three different
methods were used to predict the swell potential: Skempton's Activity (1953), Holtz and Gibbs
percent of free swell (1956), and a chart developed by Seed, et al. (1962). All of these methods
predicted that the soil would have a low swell potential. However, during construction, the
desired density could not be obtained regardless of the compactive effort, due to the expansive
nature of the soil. To determine the actual swell potential of the soil, free swell tests were
conducted in the laboratory. Numerous samples of the soil were compacted. They were then
subjected to swell at various water contents and dry densities, in -inch-deep by 4-inchdiameter, Teflon-lined rings with porous plugs. The free swell of the samples varied up to 30%.
This swelling suggested that the soil had a high swell potential (Arulanandan, 1974).
From the previous examples, several points can be made about some of the standard
methods used in conventional geotechnical analysis: (1) gradation analysis alone is insufficient
to characterize soils unless the mineralogy of the fine particles is taken into consideration;
(2) Atterberg Limits alone are insufficient for predicting the volume change characteristics of all
natural soils (Lambe and Martin, 1953-1957); and (3) the percent of particles less than 2 m in
size and the activity criteria are insufficient for predicting swell potential of all natural soils.
It has also been shown that numerous dams designed to withstand erosion based on the
Unified Soil Classification System have failed (Arulanandan and Perry, 1983). Exxon developed
methods for evaluating frost heave at a cost of 75 million dollars during Alaskan pipeline
construction. These methods, which used particle size criteria and properties such as gradation

and plasticity characteristics, have been shown to be inapplicable to soils in Canadas Mackenzie
Valley (Arulanandan and Shinde, 1981). For sands placed at the same porosity by different
methods, it reportedly takes significantly different stress ratios to cause a loss of bearing strength
due to high pore water pressure (liquefaction). This difference is a function of soil particle
arrangement (the soil fabric) (Arulanandan and Muraleetharan, 1988). It is well known that
hydrogeologic properties such as permeability depend on the in situ structure of soils. Therefore,
a non-destructive method of quantifying soil structure is necessary.
In spite of the significant observations reported by Lambe and Martin and others,
engineers have continued to use Atterberg Limits and particle size as the principal parameters to
classify soils and predict soil behavior. Swell, erosion, frost heave, and liquefaction potentials
have been evaluated using particle size, relative density, Atterberg Limits, and activity, with
serious consequences (Lambe and Martin, 1953-1957; Arulanandan, 1974; Basu and
Arulanandan, 1973; Muraleetharan and Arulanandan, 1986; Arulanandan and Perry, 1983;
Arulanandan and Shinde, 1981; Arulanandan, et al., 1986). Why do procedures based on grain
size or the properties of remolded materials fail? The answer may be they do not take into
consideration the characteristics of intact materials (soil structure) as found in nature.
1.3 DEVELOPMENT OF IN SITU METHODS
The methods thus far described exhibit severe limitation in soil classification. Other
laboratory procedures for determining soil composition, including x-ray diffraction and
differential thermal analysis, are destructive and cannot be used in situ to assess the intact state
of the soil. Additionally, while these methods may be used qualitatively to describe soil
composition, they cannot be used quantitatively to evaluate the potential behaviors of soils.
Since the 1930s, engineers have been using the Standard Penetration Test (SPT) and
Cone Penetration Test (CPT) as the primary methods of characterizing soils in situ. The SPT is
used because it permits rapid, economic evaluation of gross ground conditions in boreholes. The
CPT is used to obtain quantitative data on the thickness and consistency of soft layers and to
estimate the bearing capacity of piles in deep sand layers. Unfortunately, CPTs and SPTs cannot
identify the type of clay within a broad group called "clay." Standard field methods cannot
evaluate the soil mineralogy, pore fluid composition, particle orientation, mineral-solution
interface characteristics, total porosity, and inter- and intracluster void ratios in clays. Thus,
there is a need for in situ field tests that can be used to evaluate these characteristics to
accurately and rapidly determine soil properties and behavior at a reasonable cost.
Without disturbing the soil, current methods of site characterization cannot evaluate in
situ any of the following engineering soil properties and behavior: mineralogy (i.e., clay type and
amount), permeability, porosity, compressibility, strength, swell, erosion, frost heave, and
liquefaction potential.
Figure 1.5 shows the relationship between cyclic stress ratio and number of cycles
needed to cause initial liquefaction for samples of Monterey 0 sand prepared by different
methods at the same relative density. In Figure 1.5, dc is the amplitude of deviatoric cyclic stress
and o' is the initial confining pressure. Also given in Figure 1.5 are the measured electrical
anisotropy index (A) values. The figure shows that, for the same cyclic stress ratio, samples with
a higher value of A need fewer cycles to cause initial liquefaction. This implies that, for a given
number of cycles and given cyclic stress ratio, higher anisotropy produces a larger increase in
pore-water pressure. This effect cannot be quantified using traditional approaches.

Figure 1.5
Cyclic Stress Ratio versus Number of Cycles to cause initial
Liquefaction for Monterey 0 Sand Samples Prepared by Different Methods
(after Mulilis, et al., 1977)
1.4 NEW DEVELOPMENTS IN NON-DESTRUCTIVE METHODS OF IN SITU SOIL
AND SITE CHARACTERIZATION
During the last thirty years, methods have been developed to quantify of soil composition
and fabric by non-destructively measuring the electrical properties of a soil in its intact state.
These methods, namely the Dielectric Dispersion Method and the Formation Factor Method, are
described and detailed in Chapters 3, 4, 5, and 6 and used throughout this book to quantify soil
structure and to evaluate soil properties and behavior. By using the dielectric dispersion
behavior of clay minerals, the engineer can characterize the intact state of the soil nondestructively, thus evaluating, in situ, the engineering properties and soil behavior.
Measurement of electrical properties is a non-destructive process. This process can be carried
out in either field or laboratory on "undisturbed" samples, thereby preserving the intact structure
of the soil. Electrical soil characterization allows for in situ soil classification. This approach
can be used to evaluate and remediate a number of natural and manmade hazards, such as swell,
erosion, frost heave, and liquefaction.
1.5 PROBLEMS
1. Discuss the traditional approaches available for studying the relationship between soil
composition and engineering properties.
2. Why is it difficult to use the knowledge of soil composition to quantitatively predict
engineering properties?

REFERENCES
Arulanandan, K., (1974). Properties and Behavior of Marysville Soil, University of California
at Davis Report, Davis, CA.
Arulanandan, K., and Muraleetharan, K. K. (1988). "Level Ground Soil-Liquefaction Analysis
Using In Situ Properties: I," Journal of Geotechnical Engineering, ASCE, Vol. 114, No. 7, July.
Arulanandan, K., and Perry, E. B. (1983). Erosion in Relation to Filter Design Criteria in Earth
Dams Journal of Geotechnical Engineering, ASCE, Vol. 109, No. 5, pp. 682-697.
Arulanandan, K., and Shinde, S. B. (1981). Frost Heave Susceptibility Criteria for Soils,
University of California at Davis Report.
Arulanandan, K., Yogachandran, C., and Meegoda, N. J., (1986). Comparison of the SPT,
CPT, SV and Electrical Methods of Evaluating Earthquake Induced Liquefaction Susceptibility
in Ying Kou City During the Haicheng Earthquake, Proceedings of In Situ 86, ASCE,
Blacksburg, VA, pp. 389-415.
Atterberg, A. (1908). Studien auf dem Gebiet der Bodenkunde (Studies in the field of soil
science), Landw. Versuchsanstait, 69.
Basu, R., and Arulanandan, K. (1973). A New Approach for the Identification of Swell
Potential of Soils, Proceedings of the Third International Conference on Expansive Soils, Haifa,
Israel, Vol. 1, pp. 1-11.
Casagrande, A (1948). Classification and Identification of Soils, Transactions, ASCE, Vol.
113, pp. 901-991.
Holtz, W. G., and Gibbs, H. J. (1956).
Transactions, ASCE, Vol. 121, pp. 641-677.

Engineering Properties of Expansive Clays,

Lambe, T. W., and Martin, R. T. (1953-1957). Composition and Engineering Properties of


Soil, Proceedings of the Highway Research Board, I-1953, II-1954, III-1955, IV-1956, and V1957.
Mulilis, J.P., Arulanandan, K., Mitchell, J.K., Chan, C.K., and Seed, H.B. (1977). Effects of
sample preparation on sand liquefaction. Journal of Geotechnical Engineering Division, ASCE,
103(2), 91-108.
Muraleetharan, K. K., and Arulanandan, K. (1986). Site Characterization in Foundation
Engineering with Reference to Compressibility and Swell, Asian Region Symposium on
Geotechnical Problems and Practices in Foundation Engineering, Colombo, Sri Lanka, pp. 234244.

10

Seed, H. B., Woodward, R. J., and Lundgren, R. (1962). Prediction of Swelling Potential for
Compacted Clays, Journal of Soil Mechanics and Foundations Division, ASCE, Vol. 88, No.
SM3, pp. 53-87.
Skempton, A. W., (1953). The Colloidal Activity of Clays, Proceedings of the Third
International Conference on Soil Mechanics and Foundation Engineering, Zurich, Vol. 1, pp. 5761.
Terzaghi, K., (1957). Opening Session Address, Proceedings of the Fourth International
Conference on Soil Mechanics and Foundation Engineering, London, 3, 55-58.
U. S. Bureau of Reclamation, (1963). Earth Manual, 1st ed., revised, Washington, DC.

11

Chapter 2
Soil Structure
2.1 INTRODUCTION
Soil structure encompasses soil mineralogy, particle-solution interface characteristics,
and fabric or arrangement of particles. Many of the factors that make up the soil structure are
interdependent. The structure of a soil is also influenced by environmental factors such as
temperature and confining stress. Together, soil structure and environmental factors determine
the engineering properties and behavior of soils.
To characterize a soil adequately by any method, it is necessary to get a handle on the
seven factors that determine soil structure. These factors are mineralogy, pore fluid composition,
stress history, particle orientation, cementation, water content, and density. The importance of
each factor is briefly described below.
2.1.1 Mineralogy
Mineralogy refers to the composition of the individual particles in an element of soil.
The nature and arrangement of the atoms in a soil particle significantly affect its physicalchemical properties (Mitchell, 1993).
2.1.2 Pore Fluid Composition
The spaces, or pores, between soil particles are filled with pore fluid, composed of air
and/or water. This portion of the soil plays a vital role in site characterization. The composition
of the pore fluid strongly influences the nature of the mineral-solution interface. It follows that it
will also affect the interaction of particles with one another. Pore fluid composition is an
important factor when evaluating such concepts as swell potential, erosion potential, frost heave
potential, and contaminant migration.
2.1.3 Stress History
The history of a site will obviously play a role in site characterization. An
overconsolidated soil will exhibit different characteristics than a normally consolidated soil. In
most cases, soil deposits have been in place for many years. During this period they develop a
stress history that can influence the overall behavior. It is important to retain this history if a
proper evaluation is to be performed. Removing a sample from the ground significantly changes
the effects of the different levels of stress exerted on the soil over time. It is unlikely that a
laboratory experiment will be able to reconstitute or exactly simulate field conditions. Thus,
proper site characterization requires completely undisturbed soil samples or non-destructive, in
situ methods of characterization.

12

2.1.4 Particle Orientation


The nature of soils is such that particles are oriented differently depending on which
direction is being considered. Some soil properties, such as permeability, shear strength,
compressibility, swell potential, and erosion potential, have different values when measured in
the vertical and horizontal directions; i.e. all soils are anisotropic by nature. It is necessary to
account for this fact when quantifying soil structure. The term "soil fabric" refers to the
orientation and distribution of particles in a soil mass. A soil will behave differently depending
on the type of fabric it displays.
2.1.5 Cementation
In some instances, soil will exhibit a feature known as cementation. Cementation occurs
when soil particles are bonded to each other by a substance such as calcium carbonate or iron
oxide. The strength of the bonding depends upon the cementing material and the environment
that the soil deposit is in. For instance, in the case of pollutant migration, contaminants may
cement the soil particles together, and the cementation may persist as long as the contaminant is
present. However, in the absence of this pollutant, the soil may lack the cementation effect.
Therefore, when evaluating cementation, it is important to understand the subsurface
environment.
2.1.6 Water Content and Density
The soil mass may be considered a particulate system. Hence, it is inherently multiphase.
The three distinct phases are solid, liquid, and gas. There are a number of mathematical
expressions that relate the various phases. Two of the most important are water content and
density. The water content is the mass of the water divided by the mass of the solid in any soil
element. The density is defined as the mass of the solid divided by the volume of the soil
element. These two factors play an important role in evaluating soil behavior.
This chapter discusses the major components of soil structure: mineralogy, particle size
and shape, particle-solution interface characteristics, and fabric.
2.2 MINERALOGY
Mineralogy is the controlling factor determining the sizes, shapes, and surface
characteristics of the particles in a soil. It also influences interactions with the fluid phase.
Together, these factors influence hydraulic conductivity, swelling, compression, and strength.
Thus, mineralogy is fundamental to the understanding of geotechnical properties, even though
mineralogical determinations are not made for many geotechnical investigations. Instead, other
characteristics that reflect both composition and engineering properties, such as Atterberg limits
and grain size distribution, are determined.
Since about 1980, environmental issues, especially those related to the safe disposal and
containment of hazardous and nuclear waste, the clean-up of contaminated sites, and the
protection of groundwater, have assumed a major role in civil engineering practice. This boom
in environmental engineering has required an increasing focus on soil composition and its

13

relation to the long-term physical and chemical properties controlling soil behavior under altered
and extreme environmental conditions.
2.2.1 Types of Minerals in Soils
Silicates are, by far, the most common minerals in soils. Indeed, silicates make up over
90% of the Earths crust. Other minerals found in soils include carbonates, sulfates, and native
metals.
For the purposes of soil mechanics, the types of minerals found can be divided into two
broad groups: clays and non-clays. All clays are silicates, as are the vast bulk of non-clays.
What differentiates them is that non-clay minerals are only very weakly cohesive and are
considered granular soils. Their properties and behavior depend primarily on particle size,
particle shape, surface texture, size distribution, and fabric. The behavior of a soil consisting of a
substantial quantity of clay, however, depends largely on the type and amount of clay minerals
present, as well as mineral-solution interface characteristics and fabric.
Because of their crystal structure and composition, clay minerals exhibit a negative
surface charge. This property of clay minerals dominates their behavior when pore fluid is
present. Non-clay minerals lack this surface charge, and so their behavior can be explained in
terms of particle size, shape, and packing arrangement.
2.2.2 Clay Minerals
The clay minerals in soils belong to the mineral family called phyllosilicates. Clays have
small particle sizes, and their unit cells ordinarily have a residual negative charge that is
balanced by adsorbing cations from solution. The net negative charge is the result of
isomorphous substitution in the unit cells and has an important effect on soil properties such as
swelling and erosion.
Clay structures are made up of combinations of two simple structural units: the silicon
tetrahedron and the aluminum or magnesium octahedron, as shown in Figure 2.1. These
structural units may repeat in number to form silica sheets, alumina sheets (also called gibbsite
sheets), or magnesium sheets (also called brucite sheets). Different clay mineral groups are
characterized by the different stacking arrangements of these sheets, as illustrated in Figure 2.2.
Not only may the stacking arrangements vary between the clay minerals, but also the type of
chemical bonding between the clay sheets. As shown in Figure 2.2, illite minerals have
potassium ions complexed between the silica sheets, whereas in montmorillonite minerals, the
silica sheets are separated by thin layers of water molecules. Mitchell (1993) provides detailed
descriptions of the structure of many clay minerals.
Grouping the clay minerals according to crystal structure and stacking sequence of the
layers is convenient, because members of the same group have generally similar engineering
properties. Two-sheet minerals such as kaolinite consist of a silica sheet and an octahedral sheet
(e.g., gibbsite or brucite). Three-sheet minerals such as illite and montmorillonite consist of
either two or three octahedral sheets between two silica sheets.

14

Figure 2.1
Arrangement of the Basic Structural Units of Clay Minerals

Figure 2.2
Arrangement of Unit Layers to Form Different Clay Minerals

15

The combination of an octahedral sheet with one silica sheet (for two-sheet minerals) or
of an octahedral sheet with two silica sheets (in the case of three-sheet minerals) are called unit
layers. Most clay minerals consist of many unit layers stacked upon each other in a parallel
fashion as shown in Figure 2.2.
As mentioned earlier, isomorphous substitution is an important factor in the structure and
properties of clay minerals. In an ideal gibbsite sheet, only 2/3 of the octahedral positions are
filled, and all cations are Al3+. In an ideal brucite sheet, all octahedral positions are filled with
Mg2+, and in an ideal silica sheet, all cations are Si4+. However, due to a similarity of ion size,
some octahedral and some tetrahedral positions can be occupied by other cations. When this
substitution occurs without changing the crystal structure, it is known as isomorphous
substitution. Isomorphous substitution causes a net negative charge on clay minerals because of
the lower valence cations replacing those of higher valence. Common examples of isomorphous
substitution include aluminum in place of silicon, magnesium in place of aluminum, and ferrous
iron in place of magnesium.
Isomorphous substitution in all of the clay minerals (with the possible exception of
kaolinites) gives clay particles a net negative charge. To preserve electrical neutrality,
surrounding cations (from the pore fluid for example) are attracted and held between the unit
layers (as shown in Figure 2.2 for illite) and on the surface edges of the clay particles. Many of
these cations are exchangeable because they may be replaced by different cations. The quantity
of exchangeable cations is termed the cation exchange capacity of a particular clay mineral, and
is usually expressed as milliequivalents per 100 grams of dry clay.
The types of bonding between unit layers of three-sheet minerals (bonding between the
silica sheets) influence clay properties such as swelling and erosion. For example, the water
layers between the unit layers of montmorillonite lead to weak bonding. This influences
engineering properties such as strength, swelling, and erosion.
2.2.3 Non-Clay Minerals
Non-clay soils are virtually cohesionless and have special characteristics determined
primarily by particle size, shape, surface texture, size distribution, and packing. The mineral
composition determines hardness, cleavage, and resistance to physical and chemical breakdown.
While some carbonates and sulfate minerals are soluble within a project's time frame, in most
cases the non-clay minerals can be considered relatively inert.
Bulky, non-clay particles constitute gravel, sand, and most of the silt fraction in a soil.
Most soils results from the weathering or breakdown of preexisting rocks and soils. Thus, the
predominant mineral constituents in any soil are those abundant in the source material and either
are highly resistant to weathering, abrasion, and impact or are the weathering products of
minerals such as feldspar and olivine. The non-clay fraction of soil is mostly rock fragments or
grains of common rock-forming minerals.
Quartz is the most abundant mineral in most soils, with small amounts of feldspar and
mica present. Quartz belongs to the group of silicates called tectosilicates. The structure of
tectosilicates consists of silica tetrahedra arranged in three-dimensional frameworks with all four
oxygens in each tetrahedron shared with neighboring tetrahedra. This tetrahedral structure is
very stable and strongly bonded. The result is that quartz has no cleavage planes, no weakly
bonded ions in the structure, and a high hardness (resistance to scratching). These factors
explain the persistence of quartz in soil.
16

2.3 PARTICLE SIZE AND SHAPE


The range of particle sizes in soils used by convention are:
gravel
> 5 mm
sand
0.074 to 5 mm
silt
0.002 to 0.074 mm
clay
< 0.002 mm
These divisions by particle size are arbitrary, but convenient. The term clay is ambiguous
because it refers to both a particle size and a mineral type. In size, it refers to all constituents of
a soil smaller than a particular size, usually 0.002 mm (2 m) for engineering applications.
However, not all clay particles are smaller than 0.002 mm, and not all non-clay particles are
coarser than 0.002 mm. The definition of clay as a mineral type is discussed in Section 2.2.2.
Particle shape provides one of the main distinctions between clay and non-clay minerals
in soils. Due to the nature of their crystal structures, non-clay minerals tend to form bulky,
roughly spherical particles, while clays tend to form plate or even needle-shaped particles. One
exception is the silicate mineral muscovite (mica), which is a non-clay for the purposes of soil
mechanics, yet tends to form thin platy particles. Bulky particles have a small surface area
compared to their mass, resulting in a low specific surface area (SSA). The SSA is defined as
the surface area of the particle divided by its mass. Bulky particles are roughly spherical; thus:
SSA =

4r2
4

r3

3
6
=
r D

(1)

The SSA for bulky particles is on the order of 1 m2/g. Platy particles are long, thin boxes of
length L, width b, and height h as illustrated in Figure 2.3.

Figure 2.3
Schematic Representation of a Platy Particle

The SSA for platy particles is then

17

SSA =

2hb + 2hL + 2bL


bhL

(2)

The SSA for platy particles can be up to 600 m2/g. Size and SSA are usually inversely related as
shown in Figure 2.4.

Figure 2.4
General Relationship Between Specific Surface Area (SSA) and Particle Size
Although clay size is defined in soil mechanics as smaller than 2 m, true clay properties result
from the high SSA of their plate-shaped particles. For example, quartz can be ground to a
powder of less than 0.002 mm size. Despite the fact that the quartz flour is of clay size, it has no
cohesion or plastic properties. The particles, though small, are still roughly spherical and bulky
and do not behave as clay particles. Atterberg Limits and gradation cannot be used to define soil
adequately because behavior depends on mineral composition, particle shape, and mineralsolution interface characteristics. Because of their large SSA, surface forces dominate in clays,
while gravity forces dominate in non-clays.
2.4 PARTICLE-SOLUTION INTERFACE CHARACTERISTICS
The termination of a solid particle at a surface or phase boundary produces unsatisfied
bonds. This is because of the absence of atoms that would normally be bonded to the atoms at
the surface. The unsatisfied bonding forces may be balanced in one of several ways (Mitchell,
1993):
1) attraction and adsorption of ions or molecules from the adjacent phase
2) cohesion to the surface of another particle of the same substance

18

3) solid-state adjustments of the structure beneath the surface


Each unsatisfied bonding force is large compared to the mass of atoms and molecules, but is
insignificant compared to the mass of a grain of sand. However, as discussed in the previous
section, the ratio of surface area to mass greatly increases with decreasing particle size. For
particles smaller than 1 or 2 m, surface forces can exert a distinct influence on the behavior of
the particle. Because clay minerals have thin, plate-shaped particles with high SSAs, they are
dominated by surface forces. Surface forces in clays include adsorption of water, electrostatic
forces, and surface conduction effects.
2.4.1 Water Adsorption
There is considerable evidence that water is attracted to the surfaces of soil particles,
especially to clays. It has been shown that the adsorbed water behaves differently from free
water. For clays with high SSAs, the particle surface influences a significant portion of the
surrounding water. For example, a clay with an SSA of 100 m2/g will have an average waterlayer thickness of about 100 Angstroms at a water content of 100% (Mitchell, 1993). Granular
particles (low SSA) will have a much smaller zone of influence. The adsorption of water onto a
clay particle releases energy. This energy causes large repulsive forces at short distances (less
than about 20 Angstroms) between particles.
2.4.2 Electrostatic Forces - Double Layer Theory
If a clay is placed in an electrolyte solution (e.g., water with dissolved salts such as NaCl
or CaCl2), the negative surface charge on the clay particle attracts cations from the solution. This
results in an increased concentration of cations near the clay particle surface. Two forces affect
the cations in this range. One is the attractive force of the negatively charged clay particle. The
other is a repulsive force due to mutual repulsion of the higher concentration of cations near the
clay surface, which makes the cations want to diffuse away. These opposing forces lead to a
distribution of cations near the clay particle surfaces, as shown in Figure 2.5.
The charged clay surface and the cloud of ions next to it are together known as the
diffuse double layer. The double layer controls the interaction between clay particles. Repulsion
of the plates is caused by the overlapping of their double layers. Van der Waals forces cause
attraction between particles at short distances. Thus, the net force acting on these particles is the
difference between the attractive and repulsive forces. A thick double layer increases repulsion
and may make the distance between plates too far for the van der Waals forces to act. Several
theories have been developed to describe the distribution of ions near a charged plate, but the
Guoy-Chapman (Gouy, 1910; Chapman, 1913) theory is the most popular.

19

Figure 2.5
The Diffuse Double Layer
In developing an equation for the double layer thickness, which is the region occupied by
the diffuse layer, the following idealizing assumptions were made:
1) One clay particle suspended in an electrolyte solution is considered.
2) The particle surface is a plate that is large relative to the thickness of the
double layer (no edge effects).
3) Charge on the particle surface is distributed uniformly.
4) The electrolyte solution has a uniform dielectric constant.
5) The negative charge on the clay plate is neutralized by excess ions of opposite sign.
6) All ions are taken as point charges and there are no interactions between them.
The thickness of the double layer for dilute suspension of particles can then be expressed as:
1 = o DkT
(3)
K
2no e 2 2
where:
1/K = double layer thickness, Angstrom
k = Boltzman's constant, 1.38 X 10-23 JK-1
T = absolute temperature, K
0 = permittivity of a vacuum, 8.85 X 10-12 C2/Jm
D = dielectric constant of the medium
v = valence of the cation
20

e = electron charge, 1.602 X 10-19 coulomb


no = bulk concentration of salt, ions/m3
The full derivation of the double layer thickness is presented in Appendix B.
EXAMPLE (from Mitchell, 1993)
Consider a smectite clay with a specific surface area of 800 m2/g, a cation exchange capacity of
0.83 meq/g, constant surface charge, and a solution of 0.83 X 10-4 M NaCl in water. The
dielectric constant of water is 80 at 290K. A concentration of 0.83 X 10-4 M NaCl corresponds
to 0.083 mmoles/liter and 0.083 moles/m3. Thus, using Avogadro's number, 6.02 X 1023
ions/mole:
no = (0.083 moles/ m3)(6.02 X 1023 ions/mole) = 5.0 X 1022 ions/m3
At a temperature of 290K:
kT = (1.38 X 10-23 JK-1)(290K) = 4.0 X 10-21J
and the double layer thickness is:
1

K=

(8.85X10
(2)(5.0X10

12

22

C 2 J 1 m 1 )(80)(4.0X1021 J)

ions/ m )(1.602X10
3

19

C ) (1)
2

= 3.33X10 m = 333A

The equation for the double layer thickness quantitatively indicates the influence of
compositional factors. Increasing the dielectric constant increases the double layer thickness.
Increasing the temperature should have the same effect, but the higher temperature reduces the
dielectric constant, so the net effect is a reduction in double layer thickness. A decrease in the
salt concentration increases the double layer thickness, while an increase in the bulk
concentration decreases in the double layer thickness. Cation exchange from a higher valence
ion to a lower valence ion increases the double layer thickness, because a weaker attraction to
the mineral surface means more ions are needed to neutralize the charge on the clay plate.
The composition and type of pore fluid present in the soil greatly affects the surface
forces acting on the soil particles. Therefore, it is necessary to quantify electrolyte to evaluate
soil behavior. The bulk concentration of electrolyte in the pore fluid can be obtained by
extracting some pore fluid from the soil and measuring the electrical conductivity. The following
equation may be used to estimate the concentration of salts in the fluid:

c 10

(4)

where is the conductivity in /cm, and c is the concentration of pore fluid in normality (N).
Using special electrodes, we can determine the concentrations of the major cations in soil pore
fluid, Na+, Ca2+, and Mg2+. From these concentrations, we can obtain the sodium adsorption
ratio (SAR):

21

SAR =

Na +
Ca 2 + + Mg2 +
2

(5)

Here Na+, Ca2+ and Mg2+ are the concentrations of the respective cations in the solution
expressed in milliequivalents per liter. The SAR measures the amount of Na+ in the solution
phase. Because sodium is of lower valence than either calcium or magnesium, a lower SAR
value corresponds to a thinner double layer (i.e., fewer low-valence cations compared to highvalence cations). SAR and conductivity are good indications of the composition and type of
pore fluid.
2.4.3 Density of Surface Charge
The density of charge on the surface of clay particles can be derived from the exchange
capacity per gram and the SSA.
2.5 FABRIC
The term fabric refers to the arrangement of particles, groups of particles, and pore
spaces in a soil. Figure 2.6 illustrates the two major types of clay soil fabrics.

Figure 2.6
Dispersed (a) and Flocculated (b) Soil Structure
There are two extremes: dispersed (or deflocculated) and flocculated. In dispersed soil, the clay
particles are oriented nearly parallel to each other, as shown in Figure 2.6a. In flocculated soil,
the platy clay particles are oriented in all different directions, as shown in Figure 2.6b. Two
particles may have edge to face or edge to edge associations. Sometimes a few soil particles
may have face to face associations. These small groups are called aggregates.
Many researchers have proposed the concept of soil particles existing in clusters (Olsen,
1962; Yong and Sheeran, 1973). Clusters are groups of particles or aggregates forming larger
fabric units. Soil particles, especially clays, form clusters. These clusters can contain millions of
particles. Dispersed structure results in only a few dozen particles in a floc (cluster), while
22

flocculated structure can result in hundreds of particles making up each floc. Pictures of the two
types of clusters are shown in Figure 2.7.

Figure 2.7
Dispersed and Flocculated (Clay) Clusters
If the overlapping double layers between two clay particles are thick, then the repulsion between
the two similarly charged layers may be sufficient to prevent the particles from getting close
enough for van der Waals' forces to operate. This means that the clay particles cannot form
clusters, and they remain in a completely dispersed or deflocculated state.
Electron microscope studies have shown that soils in which clay particles form clusters
exhibit the two kinds of void ratio shown in Figure 2.8: intracluster and intercluster. Intracluster
void ratio (ec) characterizes voids between particles inside clay flocs. Intercluster void ratio (ep)
describes the voids between the clusters. The clustering of particles in a floc controls several
properties. Permeability is highly dependent on intercluster void ratio, as flow rate is
proportional to the fourth power of the radius of the pore. Because intracluster pores are very
small compared to intercluster pores, the bulk of the flow moves between clusters. A dispersed
system has a low intercluster void ratio, so permeability is lower by one or two orders of
magnitude. Flocculated systems tend to have a higher intercluster void ratio and thus a higher
permeability.

23

Figure 2.8
Intracluster (eC) and Intercluster (eP) Void Ratios

Figure 2.9
Sodium Adsorption Ratio (SAR), Total Cation Concentration, and the Demarcation
between Deflocculated and Flocculated Soil Structures (after Rowell, 1963, and Aitchison
and Wook, 1965)

24

The effect of SAR on soil fabric can be seen by considering the following. A soil with a
constant total salt concentration will have different permeabilities when the SAR is changed. As
SAR increases, permeability will be roughly the same for a while. It will then drop off sharply
by about three orders of magnitude (10% to 15%) at a certain SAR. If SAR is varied for soils
with other total salt concentrations, still locating the sharp change in permeability, a curve can be
determined. The curve connecting the SAR values when the large change in permeability occurs
forms the demarcation line between flocculated and dispersed fabric. Figure 2.9 shows this
demarcation line for montmorillonite.
The change in fabric type is due to the SAR's effect on the double layer thickness. At
high SARs, the double layer is thick, and repulsive forces dominate between clay particles. This
results in a dispersed fabric and minimal cluster formation. At low SARs, repulsion between
particles is reduced, resulting in a flocculated fabric and the grouping of particles into clusters.

Figure 2.10
Compression of Dual Porosity Soils
Olsen's cluster model went beyond merely theorizing the existence of dual porosity in
clays. He hypothetically assumed that when soils with large water contents are compressed, the
void ratio of the clusters (intracluster void ratio) does not change very much, as shown by the flat
portion of Figure 2.10. Initial compression is the result of decreasing the intercluster void ratio
causing the clusters to become more tightly packed. When the intercluster pore space becomes
comparable to that of a system of closely packed spheres, the clusters themselves start to
25

compress. At this point, the intracluster void ratio starts to decrease along with the intercluster
void ratio, although the intercluster void ratio now decreases less rapidly than before. When the
soil is unloaded, the increase in total void ratio is mainly due to swelling of the clusters. Thus,
the slope of the swell line , as shown in Figure 2.11, relates to the intracluster void ratio, ec.

Figure 2.11
Typical Consolidation Curve
High ec means large swelling, and low ec means low swelling. The swelling line is
elastic, so the recompression will follow the same slope as the rebound. However, the virgin
compression line with slope represents plastic deformation, resulting in a change in void ratio
that cannot be entirely recouped when the soil is unloaded along the swelling line of slope
(<). Thus, the decrease in the intracluster void ratio occurring along the line is not wholly
recoverable when the soil is unloaded. The elastic rebound and recompression line characterize
ec, and the normal plastic compression line characterizes ep.
In summary, flocculated soils, characterized by low SAR and extensive cluster formation,
have low swell potential but are highly permeable. Conversely, dispersed or deflocculated soils,
characterized by high SAR and minimal cluster formation, have high swell and erosion potential
with low permeability.
2.6 CONCLUSIONS
Heterogeneous media, such as the three-phase soil-mass system, have a geometrical
structure that is not always regular. Interpreting the properties of such media is of fundamental
practical importance in understanding their behavior. Thus, it is worthwhile to characterize such
media by non-destructive methods. Examples of such media are clay-water-electrolyte systems,
granular soils, certain porous water- or oil-bearing rocks, ion-exchange columns with interstitial
pore space filled with aqueous electrolyte solutions, and even certain molecular filter
membranes. One approach to nondestructive characterization is to use electrical response
characteristics to quantify soil structure. This approach is based on evaluating the electrical
response characteristics of soil placed in an alternating electrical field. Chapter 3 presents the
background of the electrical methods.

26

2.7 PROBLEMS
1. Describe the electrical double layer that forms around clay particles. Why do the balancing
cations diffuse away from the particle surfaces in an aqueous environment?
2. Illustrate the effects of changes in each of the following on the thickness of the double layer.
Give reasons for the effects shown. (a) ionic concentration; (b) valence; (c) temperature; (d)
dielectric constant; (e) ionic size; (f) pH; (g) anion adsorption.
3. Discuss the effects of changes in each of the variables given in Question 2 on the swell and
swelling pressure of given clay.
4. Define Specific Surface Area (SSA) and calculate the SSA of a sphere, a flat plate, and a
circular plate (disk).
REFERENCES
Aitchison, G. D., and Wook, C. C. (1965). "Some Interactions of Compaction, Permeability, and
Post-Construction Deflocculation Affecting the Probability of Piping Failure in Small Earth
Dams," The Sixth International Conference on Soil Mechanics and Foundation Engineering,
Montreal, Canada, Vol. 2, p. 442.
Chapman, D. L., (1913). A Contribution to the Theory of Electrocapillarity, Philosophical
Magazine, Vol. 25, No. 6, pp. 475-481.
Gouy, G. (1910). Sur la Constitution de la Charge Electrique a la Surface d'un Electrolyte,
Anniue Physique, Paris, Serie 4, Vol. 9, pp. 457-468.
Mitchell, J. K., (1993). Fundamentals of Soil Behavior, John Wiley and Sons. New York.
Olsen, H. W., (1962). Hydraulic Flow through Saturated Clay, Proceedings of the Ninth
National Conference on Clays and Clay Minerals, pp. 131-161.
Rowell, D. L. (1963). Effect of Electrolyte Concentration on the Swelling of Oriented
Aggregates of Montmorillonite, Soil Science, Vol. 96, p. 368.
Yong, R. N., and Sheeran D. E., (1973). Fabric Unit Interaction and Soil Behavior,
Proceedings of the International Symposium on Soil Structure, Gothenburg, Sweden, pp. 176183.

27

Chapter 3
Electrical Dispersion and
Soil Structure
3.1 INTRODUCTION
The engineering properties of soil depend to a large extent on the following factors,
which influence soil structure:
1.
2.
3.
4.
5.

Mineralogy
Pore fluid composition
Fabric
Porosity
Stress history.

This is shown by the works of Lambe (1958, 1960), Lambe and Martin (1953-1957), Rosenquist
(1955, 1958), Bolt (1956), Mitchell (1956), Seed and Chan (1959), Martin (1960), Olsen (1961),
and many others. Different testing methods have demonstrated the fundamental dependence of
engineering properties on the above factors. However, no single method can quantitatively
evaluate all the factors making up the total structure of a soil. Most current methods for
determining mineralogy and soil fabric involve sample destruction. The search for a method to
evaluate these soil characteristics without destroying the sample has led to investigations into the
usefulness of soil electrical properties for characterization.
The extension of the Maxwell, Wagner, and Fricke work on the conductivity of aggregate
systems incorporates the conductivity of the soil particles associated with surface charges and the
multi-directional orientation of particles (Arulanandan and Yogachandran, 2000). In the
resulting equations, a microscopic approach predicts electrical dispersion characteristics of soils
containing clay minerals.
Numerous studies have evaluated the applicability of the radio-frequency dielectric
behavior and conductivity dispersion behavior of clay minerals to soil characterization.
Specifically, the studies have explained the extent to which a single set of these electrical
measurements can characterize the mineralogy, mineral-solution interface characteristics, fabric,
inter- and intracluster void ratios, and total void ratio of an intact sample for the purpose of
predicting engineering properties and explaining soil behavior (Arulanandan and Mitra, 1970;
Arulanandan and Smith, 1973; Smith and Arulanandan, 1981; Arulanandan et al., 1973; Basu
and Arulanandan, 1973; Alizadeh, 1975; Arulanandan, 1987; Arulanandan, K. and Arulanandan,
S., 1985; Arulanandan, 1991; Arulanandan and Yogachandran, 2000; and Arulanandan et al.,
1994). These studies showed that the structural parameters that control soil behavior also control
electrical dispersion.
A macroscopic approach will also be presented. This approach utilizes an electrical soil
model in which circuits of resistors and capacitors represent the different possible current paths
through a saturated soil-water-electrolyte system. This model is known as the Three-Element

28

Electrical Model. This chapter shows how a three-element electrical model can be used to
quantify soil structure.
3.2 ELECTRICAL RESPONSE CHARACTERISTICS OF CLAY MINERALS
Applying an electrical field to a material produces a response. This response can be
evaluated quantitatively in terms of resistance, R, and capacitance, C, irrespective of whether the
system represents a series or parallel combination of resistance and capacitance. The choice of
circuitry representation is arbitrary and not indicative of any mechanism. The material can be
solid, liquid, or a solid-liquid mixture (soils).
The values of capacitance and resistance can be converted into dielectric constant, ',
and conductivity, . The dielectric constant is defined as C/Co where Co is the capacitance of a
unit capacitor with a vacuum between the plates. The dielectric constant and conductivity can be
obtained by employing the following equations for a cylindrical specimen with two electrodes
placed on the ends, as shown in Figure 3.1.

Figure 3.1
Parallel Plate Electrodes on Opposite Ends of a Soil Sample

' = Cd/A a
= d/RA

(1a)
(1b)

where,
d = length of the specimen (cm)
A = cross sectional area of electrodes (cm2)
a = dielectric constant of vacuum (= 8.85 x 10 -14 farad/cm)

29

These values of electrical response characteristics, measured in terms of conductivity and


dielectric constant, depend only on the structure and composition of the material, provided the
applied voltage is in such a range that the C and R responses measured are independent of the
voltage.
3.2.1 Polarization
As seen in Chapter 2, fine grained soil may be considered an assemblage of particles
where the negative surface charges and positively charged counter-ions associated with the
particle surfaces form a double layer. When an alternating electric field is applied, we may
think of the charges as moving back and forth with a certain amplitude. The amplitude will vary
with the type of charge (i.e., ion valence, Z), the degree of association of the charge with particle
surfaces, the degree (angles) of particle orientation, and the temperature of the system. Some of
the charges (ions) that are free to drift through the material and discharge at the electrodes
produce a direct current conductivity. Other charges, bound to the particle surfaces, are not free
to drift from one electrode to the other, but they are able to oscillate back and forth under the
action of an alternating current field. The oscillation of the bound charges is generally thought to
produce polarization. This oscillation of the bound positive charges produces a current called the
polarization current, and the material is said to be in a polarized condition. The number of
polarizable charges multiplied by the distance moved per unit volume is called the polarizability
of the medium. The structure of the material and the frequency of the electric field determine the
magnitude of the polarizability. This magnitude can be expressed in terms of a directly
measurable constant: the dielectric constant, ', in the electrical frequency range (Arulanandan
and Mitchell, 1968; Mitchell and Arulanandan, 1968). The dielectric constant actually measures
the ability of a material to store electrical potential energy under the influence of an electric field.
Thus, the dielectric constant expresses the amount of charge present in a material and, hence, the
materials magnitude of polarizability. The model chosen to describe polarization dictates the
relationship between dielectric constant and polarizability.
There are many types of polarization depending on the time required for the polarization
process. In the case of an alternating electric field, the frequency dictates the time allowed for
polarization to occur. Thus, the type of polarization developed is a function of the frequency.
Examples of proposed polarization processes are dipole rotation (Debye, 1929), interfacial
polarization due to inhomogeneity (Wagner, 1914), ion atmosphere distortion (Schwarz, 1962),
and coupling of flows (Madden and Marshall, 1959).
3.2.2 Relaxation and Electrical Dispersion
Voltage applied to a system for a certain length of time produces a polarization current.
as discussed earlier. When the voltage is turned off, this polarization current decays with time in
an exponential manner. Vacquier, et al., (1957) observed an example of this process in claywater-electrolyte systems as shown in Figure 3.2.

30

Figure 3.2
Decay of Polarization Voltage
The sudden removal of an electric field in equilibrium with a dielectric material produces
dielectric relaxation, or the exponential decay, with time, of the magnitude of induced
polarization. The relaxation time is defined as the time in which the polarization reduces to 1/e
(about 0.37) times its original value, where e is the natural logarithmic base. The relaxation
mechanism depends on the process causing polarization. Different relaxation mechanisms
display different relaxation times.
When measuring over a range of frequencies, the dielectric constant, ', can
vary, as shown in Figure 3.3. The fall of dielectric constant with increasing frequency is called
anomalous dispersion. The dielectric constant falls off with rising frequency because the
polarization can no longer change fast enough to reach equilibrium with the electric field (Hill,
1969). This lag in attaining equilibrium is again referred to as relaxation. For a particular
dielectric material, several regions of anomalous dispersion may be observed over the frequency
range from zero to the microwave range (1011 Hz). At the lower end of the frequency spectrum,
there may be sufficient time between electric field alternations for many relaxation mechanisms
to contribute to the dielectric constant. However, as the frequency increases and the time
between alternation decreases, it may become impossible for certain relaxation mechanisms to
operate. As polarization mechanisms cease to function, the dielectric constant decreases.
As shown in Figures 3.3 and 3.4, there are three regions of electrical dispersion with
respect to frequency.

31

Figure 3.3
Dielectric Behavior of Clay-Water and Ion Exchange Resin Systems
(Arulanandan & Mitra, 1970)

Figure 3.4
Multiple Anomalous Dispersions in a Dielectric Material
These three dispersions occur in the electrical frequency range, the radio frequency range,
and the microwave range. Each dispersion is indicative of the different mechanisms involved in
the process. Electrical dispersion in the low frequency range (electrical frequency range) is due
to particle size and is a diffusion-controlled relaxation process (Schwarz et al,., 1962;
Arulanandan and Mitchell, 1968; Mitchell and Arulanandan, 1968). The electrical dispersion in

32

the microwave range is due to the Debye molecular rotation (Debye, 1929). The electrical
dispersion in the radio frequency range is due to the polarization of interfaces between mineral
particles and the solution (i.e., polarization of the double layer) (Okonski, 1960).
The Maxwell-Wagner, Maxwell-Wagner-Fricke, and extension of the Maxwell-WagnerFricke microscopic theoretical models (Arulanandan, et al., 1994) deal with this polarization of
interfaces between mineral particles and pore fluid. Measurements of electrical dispersion due to
the Maxwell-Wagner relaxation mechanism may potentially be used to gain information on the
structure of a soil without destroying the sample. For this reason, a number of studies have been
conducted on the radio frequency response characteristics of soils (Arulanandan and Mitra, 1970;
Arulanandan, et al., 1973; Arulanandan and Smith, 1973; and Arulanandan et al., 1974).
3.2.3 Mechanism of Radio Frequency Electrical Dispersion of Soils
When the conductivity and dielectric constant of a liquid such as water or of an
electrolyte solution are measured as a function of frequency in the radio frequency range, it is
found that 'and do not vary (see Figure 3.5).

Figure 3.5
Variation of Dielectric Constant and Conductivity with Change in Frequency
For 0.01 N NaCl Solution
The electrical properties do not change with frequency because substances such as water
and electrolyte solutions are considered homogeneous systems. In a homogeneous system the
current density, which is proportional to the ratio of conductivity to dielectric constant, does not
vary from point to point. However, when a heterogeneous system is considered, such as a
saturated clay-water-electrolyte system, current density varies from point to point, as the ratio of
conductivity to dielectric constant is different for each component. Such a system exhibits a
dispersion in the radio frequency range due to a Maxwell-Wagner relaxation mechanism.
Charges accumulate at the interface between the clay particle and the surrounding solution.
Because this buildup of charges takes time, as the frequency is increased there is less time for the

33

charges to accumulate at the interface. This decreases the ability of the system to store electrical
potential energy and thus decreases the dielectric constant. When the frequency reaches a certain
value, there is not enough time for any charges to accumulate at the interface. At this point, the
dielectric constant becomes independent of frequency.
3.3 MICROSCOPIC MODEL OF THE ELECTRICAL DISPERSION
CHARACTERISTICS OF SOILS
The electrical conductance of a soil particle depends on the mineralogy and the adsorbed
pore fluid. The electrical conductivity of the particle may be defined as the sum of two
conductivities. The first conductivity is due to the charged carriers in the soil particle, and the
second conductivity is due to the existence of the double layer in the interface between the clay
particle and the surrounding pore fluid (surface conductance). For dry particles, the double layer
is not present, and the conductivity of the particle is due to the charges in the particle. For
saturated systems, the double layer is formed, and the particle conductance is due to the double
layer and the charges on the particle surface. Weiler and Chaussidon (1968) investigated the
surface conductivity of spherical and ellipsoid particles based on Okonskis (1960) theoretical
approach. For the ellipsoidal particle, Weiler and Chaussidon indicated that the effective
conductivity of the charged particle consists of two components, one arising from the surface
conductance, and the other from the conductivity of the ellipsoid, which depends on the axis
under consideration (Weiler and Chaussidon, 1968). In highly hydrated systems, such as
saturated clay minerals, it is reasonable to consider the ellipsoid conductivity negligible when
compared to the particles surface conductivity.
As was shown in Chapter 2, clay minerals have thin, plate-shaped particles. Oblate
spheroids with R>>1 would appropriately model these particles. Extremely oblate spheroids may
be considered as discs or plates. Weiler and Chaussidon (1968) give the surface conductance, S
(in mhos), of such an oblate spheroid (R>>1) as:

S =

a S
4

(2)

where s is the surface conductivity of the oblate spheroid in mho/cm. For a circular disc (an
extreme case of an oblate spheroid) the specific surface area, SSA, can be obtained as:

SSA=

2b2 + 2ba
b2aS

(3)

where s is the density of the particle, b is the radius of the disc, and a is the thickness of the
disc. Equation (3) can be rewritten as:
SSA =

2
1
1+

Sa
R

(4)

34

where R = b/a. Using equations (2) and (3) and eliminating a from both equations, an expression
for S is obtained as (Arulanandan and Yogachandran, 2000):
S =

1 + S

2(SSA) S
R

(5)

The surface conductance of the clay mineral can be obtained using Equation (5). Because the
conductivity due to the charges within the particle can be neglected, the conductivity of the
particle, 2 , will be approximately equal to s.
The conductivity of the particle, 2, approximately equal to S , may be determined by
fitting electrical dispersion data using the three element model (as described in Section 3.8).
Figures 3.6, 3.7, and 3.8 show this fitting procedure for different montmorillonite/sand mixtures.
Once 2 is determined, S may be calculated by using Equation 5. Table 3.1 shows values of S
calculated for the montmorillonite/sand mixtures. These values of s agree with values of
surface conductance obtained by other researchers (Cremers, et al., 1966; Weiler and
Chaussidon, 1968; Schwan, et al., 1962). For example, Weiler and Chaussidon (1968) calculated
values of

s on the order of 10-9 mho, which agree with the values given in Table 3.1.
Dielectric and Conductivity Dispersion Curve
200

0.005

dielectric constant

180

dielectric constant

160

0.0045

conductivity

0.004

140
120

0.0035

100

0.003

80
0.0025

60
40
1

10

conductivity, mho/cm

three element model optimization

0.002
100

freqency, MHz

Figure 3.6
Optimization and Experimental Dielectric Dispersions of a 40% Montmorillonite, 60%
Sand Mixture

35

0.0028
three element model optimization

0.0026

dielectric constant

0.0024

conductivity

0.0022
0.002
0.0018
0.0016
0.0014

conductivity, mho/cm

dielectric constant

Dielectric and Conductivity Dispersion Curve


200
180
160
140
120
100
80
60
40
20
0

0.0012
1

10

0.001
100

freqency, MHz

Figure 3.7
Optimization and Experimental Dielectric Dispersions of a 30%
Montmorillonite, 70% Sand Mixture

0.0028
three element model optimization
dielectric constant
conductivity

0.0026
0.0024
0.0022
0.002
0.0018
0.0016
0.0014

conductivity, mho/cm

dielectric constant

Dielectric and Conductivity Dispersion Curve


200
180
160
140
120
100
80
60
40
20
0

0.0012
1

10

0.001
100

freqency, MHz

Figure 3.8
Optimization and Experimental Dielectric Dispersions of a 20% Montmorillonite, 80%
Sand Mixture

36

Table 3.1
Observed and Optimized Electrical Parameters for Given Conditions of Particle Shape,
Orientation, Porosity, and Specific Surface Area Used for Calculation of Surface
Conductance

Soil Type

Conductivity

Conductivity

Axial

Specific

Surface

of Pore Fluid

of particle

Ratio

Porosity

Surface

1 (mho/cm)

2 (mho/cm)

(optimized)

(optimized)

(assumed)

(measured)

Area
2
SSA (m /g)

Conductance
s (mho)
(Calculated)

Influence of Clay Content on Surface Conductance and Dielectric Dispersion


40% Montmorillonite + Sand

0.00050

0.00490

40

0.746

241

1.21E-09

30% Montmorillonite + Sand

0.00039

0.00300

30

0.709

181

9.96E-10

20% Montmorillonite + Sand

0.00041

0.00285

20

0.629

121

1.37E-09

3.4 SOIL STRUCTURE EFFECTS ON ELECTRICAL DISPERSION


This section demonstrates how various parameters of soil structure affect electrical
dispersion. These structural parameters include clay type (mineralogy), water content, pore fluid
composition, and fabric. This section focuses on qualitative effects; Section 3.8 will discuss
using the Three-Element Model to extrapolate quantitative effects.
3.4.1 Clay Type
Figure 3.9 shows the electrical dispersion behavior of montmorillonite, illite, and
kaolinite soils. It is known that the cation-exchange capacity of these clays increases in the
sequence kaolinite < illite < montmorillonite. This sequence is also clearly seen in the electrical
dispersion behavior shown in Figure 3.9. Clay mineralogy is the main factor determining the
radio-frequency electrical dispersion properties, and it strongly influences many other properties
of the clay-water-electrolyte systems. The electrical dispersion results may be interpreted to
mean that the higher the water-adsorbing capacity of the clay, the higher the electrical dispersion.
3.4.2 Clay Percentage
Samples of illite-grundite were mixed with 63%, 72%, and 81% sand. Figure 3.10 shows
electrical dispersion characteristics of the consolidated samples. An increase in percentage of
sand (1) decreased the dispersion, and (2) shifted the dispersion curves downward. These results
are to be expected, since the water associated with the solids decreases with increased sand
content. It will be shown later that both water content and sand content control dispersion
characteristics, but that added sand significantly lowers the average dielectric constant of the
soil cluster, r.

37

300

0.0035
0.003

250

illite

200

kaolinite

0.0025
0.002

3 element
optimization

150

0.0015
100
0.001
50

Conductivity, mho/cm

Dielectric Constant

montmorillonite

0.0005

0
1

10

0
100

Frequency, MHz

Figure 3.9
Effect of Mineral Type on Electrical Dispersion
(Arulanandan and Mitra, 1970)

Figure 3.10
Effect of Clay Percentage on Electrical Dispersion of Illite-Grundite
(Arulanandan and Smith, 1973)
3.4.3 Water Content
Three samples of illite-grundite were made homoionic to sodium and consolidated to
water contents of 78%, 52%, and 48.6%. Figure 3.11 shows the electrical dispersion
characteristics obtained. Increasing consolidation pressure reduces the water content. Thus the

38

relative transport of electric current through the solution path becomes less. At the same time the
particles are brought closer together, thus increasing the tortuosity of the current path. The
reduction in the pore water phase (i.e., the phase with larger dielectric constant) and a change in
the tortuosity shift the electrical dispersion characteristics downward.

Figure 3.11
Dielectric and Conductivity Dispersions for Illite Consolidated to
Different Water Contents (Arulanandan and Smith, 1973)
3.4.4 Particle Orientation
Electrical response characteristics were obtained on a sample of kaolinite hydrite UF that
was prepared by consolidating from a slurry under 1 kg/cm3. Cylindrical specimens were cut
with their longitudinal axes both in the direction of consolidation and at right angles to the
direction of consolidation. The resulting electrical dispersions, shown in Figure 3.12, are
significantly different. When the current flows along the long axis of the oriented particles, i.e.,
parallel to the 1 plane, obstruction in the current path is less than with the perpendicular flow.
This suggests that dispersion characteristics will change with the direction of the current path
through the soil since compositional and geometrical parameters are known to vary in the 1 and
3 directions, it is likely that they contribute to the variation in dispersion characteristics in the
same way. Later, we will quantify these parameters by applying the Three-Element Model to the
experimental results.
3.4.5 Flocculated and Dispersed Soils
Figure 3.13 shows the dispersion curves for two illite-solution aggregates. In preparing
of the flocculated soil, the clay was initially treated with 0.1 N NaCl. This relatively high
concentration of NaCl causes flocculation of the solid particles. After consolidation to 1 kg/cm2
the sample was leached with 0.01 N NaCl solution. The second illite sample was treated with
0.01 N sodium oxalate and then consolidated at 1 kg/cm2 to produce a sample with dispersed
structure.
The electrical dispersion curves are distinctly different, with the flocculated sample
showing smaller dispersion. This suggests that less water is associated with solid in the
flocculated than in the dispersed illite. It also suggests that the particles have a higher

39

tendency toward parallel orientation in the dispersed sample, increasing in the tortuosity of the
path (parallel to the loading direction) of the electric current through the solution.

Figure 3.12
Effect of Soil Fabric (Particle Orientation) on the Electrical Dispersion of a Kaolinite
Sample as Shown by Measurement in Different Directions (Arulanandan and Smith, 1973)

Figure 3.13
Effect of Flocculated and Dispersed Structure on the Electrical Dispersion of an Illite Soil
(Arulanandan and Smith, 1973)

40

3.4.6 Electrolyte Concentration


Samples of kaolinite hydrite R were made homoionic to sodium and consolidated to a
water content of 44% at different electrolyte concentrations. Figure 3.14 shows electrical
dispersions that decrease with an increase in electrolyte concentration. One explanation is that
increased electrolyte concentration decreases the thickness of the double layer and thus the
volume available for charge accumulation.

Figure 3.14
Effect of Electrolyte (Pore Fluid) Concentration on the Electrical Dispersion of an IlliteSand Mixture (Arulanandan and Smith, 1973)

Figure 3.15
Effect of Pore Fluid Cation Size on the Electrical Dispersion of an Illite-Sand Mixture
(Arulanandan and Smith, 1973)

41

3.4.7 Electrolyte Type


Figure 3.15 shows the dielectric dispersion characteristics of illite-grundite prepared with
equivalent pore fluid concentrations (normalities) but with different pore fluid cations. It is well
known that a change in ion type alters the thickness of the double layer. Figure 3.15 indicates
that a lithium clay with a thicker double layer exhibits a dispersion with larger measured
dielectric constant than a sodium clay.
3.5 MACROSCOPIC APPROACHTHE THREE-ELEMENT MODEL
Electrical current traveling through a porous media can be considered to have three
possible paths (Sauer, et. al., 1955):
(1) through solution and conducting particles in series,
(2) through particles in contact with each other, and
(3) through the solution alone.
Figure 3.16(A) shows schematic representation of the three pathways. Each pathway may be
represented by its respective fraction of the total cross-section. Figure 3.16(B) shows these
fractions where (a) represents the fractional cross-section through Path (1), (b) represents the
fractional cross-section through Path (2), and (c) represents the fractional cross-section through
Path (3). By their proportions, (a) + (b) + (c) = 1 or 100%. Note that in Path (1), or (a), the
current travels through both the particles and solution in series. The (d) parameter represents the
fractional path length through the conducting solid. Therefore, (1-d) is the fractional path length
through the solution phase as shown in Figure 3.16(B). To model the material, each fraction
may be further represented as a resistor and capacitor in parallel, resulting in what is shown in
Figure 3.16(C). The representations shown in Figure 3.16 are what is known as the ThreeElement Model.
Sachs and Spiegler (1964) have upgraded the Three-Element Model for the radio
frequency range. As shown in Figure 3.16(C), the model represents each liquid or solid particle
zone with a parallel circuit of a resistor and capacitor. The electrical network can be analyzed by
formulating complex impedances for each current path in terms of relevant compositional and
geometrical parameters, and summing them. After evaluating the complex admittance (reciprocal
of impedance) of the entire circuit, the real and imaginary parts of the formulation are separated.
The equations relating the models apparent dielectric constant, ', and apparent conductivity,
, to frequency depend on the systems compositional and geometrical parameters:
th =

k 2 k 2
2 2
a
r s + s r + 2 2 r s + r s
1 d
d (1 d )S 1 d
d
d

+ b + c
r
s

th =

k k 2 k 2k
2k
2k
a
r s + r s + 2 2 s r + r s
1 d
d(1 d)S 1 d
d
d

(6a)

+ bk + ck (6b)
r
s

42

where
2

k


k
S s + r + 2 2 s + r
1 d d
1 d d

(6c)

and in which is the capacitance of a unit capacitor in a vacuum (i.e., 8.85 x 10-14 F/cm), and
= 2 f is the angular frequency.

Figure 3.16
Model of Heterogeneous Porous Plug (after Sachs and Spiegler, 1964)

These theoretical equations can be fitted to experimental electrical dispersion data for
saturated fine-grained soils by computer optimization of the geometrical (a, b, c, d) and
compositional ( r, s, kr, ks) parameters. By systematically varying the structure-determining
factors, Arulanandan and Smith (1973) have evaluated the ability of the parameters to
characterize the composition and heterogeneity of the soil system. The c parameter was found to
measure the size and distribution (tortuosity) of the intercluster pores and the intercluster void

43

ratio. The c parameter is very sensitive to intercluster void ratio and intercluster pores and is
independent of compositional variations. Model parameter b measures the contact area between
the clusters. This parameter is always small for untreated fine-grained soils. The term r is
mainly an average measure of the type of clay mineral present in the soil and the intracluster
void ratio. The distinctive r values obtained for different wet minerals relate to each minerals
characteristic water retention. The r parameter is thus referred to as the dielectric constant of
the wet soil cluster.
We have formulated theoretical dispersion curves in terms of model parameters (i.e.,
Equations 6a, b, c). Next, a method must be found to extract the values of the model parameters
from experimental data. In other words, a combination of compositional ( r, s, kr, ks) and
geometrical (a, b, c, d) parameter values must be obtained for optimal fit to experimental
dispersion curves. If dielectric constant and conductivity are measured at n frequencies, then 2n
equations must be solved to get a best-fitting combination of the unknowns. A digital computer
optimization program has been developed to solve such a system of equations (Arulanandan and
Mitra, 1970; Muat, 1968). The optimization program is used to determine the values of the
geometrical and compositional parameters of the model, such that the electrical dispersion curves
calculated from the theoretical equations (6) fit the empirical results within acceptable error
limits. By setting s equal to 80, the dielectric constant of water, and noting that a = 1-(b+c),
the number of unknowns in equations (6) is reduced to six.
At this point, it is reasonable to question whether choosing six parameters to fit
dispersion curves can produce physically meaningful results. It is true that six varying unknowns
can describe just about any set of curves; however, closer scrutiny reveals that only three
independent geometrical parameters are genuine unknowns. The conductivities kr and ks of the
solid and interstitial solution, respectively, and the dielectric constant of the solid particles or
clusters, r, must have values that are reasonable from the point of view of physical science. The
conductivity of the interstitial solution should be of the same order as the conductivity of the soil
pore fluid extract. The conductivity of the solid should be of the same order as the
isoconductivity value. The dielectric constant of the cluster of particles should be between 4.5
(the dielectric constant of a dry silicate material) and 80 (the dielectric constant of pure water),
depending on the water content. Although the latter limitations were not written into the
computer program, the model results are consistent with these requirements.
Example of the Use of the Three-Element Model for Quantifying Clay Mineral Structures
The electrical dispersion data for illite soil was obtained using a Boonton Rx meter as
described in Chapter 5. Figure 3.17 shows the dispersion data. A simplex optimization program
was used to obtain the geometrical and compositional data as shown in Table 3.2.

44

Dielectric and Conductivity Dispersion Curve


0.0029

80

0.0028
75

0.0026

70

65

optimized results

0.0025

Dielectric Constant

0.0024

Conductivity

0.0023

60

0.0022

conductivity, mho/cm

dielectric constant

0.0027

0.0021
55
0.002
50
1

10

0.0019
1000

100

freqency, MHz

Figure 3.17
Dielectric and Conductivity Dispersion Characteristics of Saturated Illite-Grundite
Table 3.2
Optimized Three-Element Model Parameters of Illite-Grundite, Figure 3.10

Geometrical parameter
Geometrical parameter
Geometrical parameter
Geometrical parameter
Dielectric constant of the cluster
Conductivity of the cluster (mhos/cm)
Conductivity of the pore fluid (mhos/cm)
Dielectric constant of the pore fluid
Water content of the sample

a
b
c
d
er
kr
ks
es
78.00%

0.59
0.02
0.39
0.91
31
0.0036
0.00106
78

The compositional parameter, r, defining the dielectric constant of the cluster is used to
evaluate inter- and intracluster void ratios of the soil as described below.
3.6 INTERCLUSTER VOID RATIO AS A FUNCTION OF THREE-ELEMENT MODEL
PARAMETERS

An approximate expression has been proposed (Arulanandan, 1980) to quantify the


intracluster void ratio ec .. This expression uses the dielectric constant of the cluster, r, and a
simple mixing rule as follows:
r = n c S + (1 n c ) p

(7)

45

where nc is the intracluster porosity, defined as the ratio of the intracluster void to the volume of
the cluster. Assuming s = 80, p = particle 4.5 (dielectric constant of dry silicate mineral),
equation (7) becomes
r = 74.5n c + 4.5

(8)

Thus, using the Three-Element Model to determine r of a clay-water-electrolyte system, the


value of nc can be quantified. Thus, the intracluster void ratio,
ec =

nc
1 nc

(9)

can be calculated. The total void ratio of the soil, e, can be used to estimate the intercluster void
ratio, eP, because the total void ratio is the sum of the inter- and intracluster void ratios as
follows:
e = e P + eC

(10)

Thus, the three-element electrical model can be used to quantify the inter- and intracluster
void ratios. These ratios are necessary to evaluate hydraulic conductivity using the KozenyCarman equation and the swell index of clays due to unloading (as explained Section 3.10).
The inter- and intracluster void ratios for the illite soil sample shown in Figure 3.10 are as
follows:
Intercluster void ratio, e p = 1.74
Intracluster void ratio, ec = 0.41
Total void ratio, e = 2.15
3.7 THREE-ELEMENT MODEL OPTIMIZATION

This section demonstrates optimizing the Three-Element Model parameters to fit


experimental dielectric constant and conductivity results. At first it may seem that fitting six
parameters to two electrical dispersion curves (dielectric constant and conductivity vs.
frequency) could result in many possible solution combinations. However, constraints,
interrelationships, and methods of approximating the parameters must be taken into account.
Ultimately, the optimized parameters must be reasonable from a physical point of view.
3.7.1 Determining Initial Values

Ultimately, optimization is most conveniently carried out by an optimization program


that will be demonstrated later, but accurate optimization relies on educated calculations of the
starting parameters.
The a Parameter

The a parameter is most easily dealt with in the following equation:

46

a = 1 b c

The b Parameter

Unless soils are cemented, the b parameter is almost always approximately zero. This is
because current does not generally move all the way through a sample without moving through
intercluster pore fluid. Thus, under most conditions, the a parameter is very generally
approximated by:
a = 1 c
The S Parameter

S is the dielectric constant of the pore fluid. In all natural cases without contamination,
the pore fluid will be water and S will be approximately 80, the dielectric constant of water.
When the pore fluid is contaminated, it is necessary to adjust S by calculating or measuring the
dielectric constant of the pore fluid solution or mixture.
The r Parameter

r is the dielectric constant of the soil clusters or, more accurately, the average cluster
dielectric constant. As shown in Olsens cluster model (1962), the clusters contain pore fluid
(the intracluster porosity) and soil particles. Therefore, r is usually somewhere between the
dielectric constant of the soil particles, approximated by 4.5, and the dielectric constant of water,
80. r may be approximated using the intracluster porosity, nc and a simple mixing rule proposed
by Arulanandan (1980):
r = n c 80 + (1 n c ) 4.5

The obvious question arises, how does one obtain nc? Refer to the relationship between and
0 shown in next Chapter (Figure 4.7). Given the electrical dispersion experimental data,
0 may be approximated, and thus may be determined. Typically the slope of the rebound
line is 1/10 to 1/5 . A decent first approximation of for starting an optimization is
therefore = 1/8 . Also refer to the relationship between and the intracluster void ratio
shown in Chapter 4 (Figure 4.14). Once is estimated, using its relationship with ec/eT (the
intracluster void ratio divided by the total void ratio), the ec/eT ratio may be approximated. The
total void ratio may be determined by water content, w, and specific gravity of the soil, Gs
( 2.7 for non-organic clays):
eT = wGs
or by using the vs. porosity, n, relationship shown in next chapter (Figure 4.7). Knowing the
total void ratio eT, the porosity of the sample may be calculated:
e
n= T
1 + eT

47

Given eC/eT and eT, eC is easily calculated. Finally, the intracluster porosity, nC, may be obtained
by:
e
nC = C
1 + eC
The kS Parameter

The conductivity of the pore fluid, kS, should be roughly the same order of magnitude as
the conductivity of the pore fluid extract. However, the optimized value for kS is typically
different from that of the pore fluid extract. It is usually lower than the conductivity of the
extracted pore fluid.
The kr Parameter

The conductivity of the soil clusters, kr, may be directly measured by determining a soils
isoconductance value. This may be determined by measuring the conductivity of the saturated
sample and plotting sample conductivity vs. pore fluid conductivity, as shown in Figures 3.18
and 3.19. The isoconductance value, which approximates kr, is where sample conductivity
equals pore fluid conductivity. A Three-Element Model optimization will determine kr if it is not
known in advance.

Figure 3.18
Laboratory Determination of the Isoconductance Value

48

Figure 3.19
Determination of Isoconductance Value of an Illite Soil (Arulanandan and Mitra, 1970)

Figure 3.20 shows the direct relationship between kr and specific surface area, SSA.
Specific surface area may be estimated using 0 by the following correlation (Meegoda, 1985).

Figure 3.20
Magnitude of Dielectric Dispersion vs. Specific Surface Area

49

The d Parameter

The d parameter has a very critical influence on the electrical dispersion calculated by the
Three-Element Model. An initial guess for d is made by trial and error after c and r are first
approximated, as seen in the example problem that follows.
The c Parameter

The c parameter represents the pure-water electrical path. This is the primary pathway
responsible for the hydraulic conductivity of soils, as discussed earlier in this chapter. Smith and
Arulanandan (1981) have shown a clear relationship between the c parameter and the hydraulic
conductivity of several soils. c may be approximated initially by the following relationship
between c and eP/eT (Arulanandan, 1997).

Figure 3.21
Non-Dimensional Void Ratio, eP/eT vs. the Three-Element Model c Parameter
3.7.2 Fine Tuning Initial Guesses

Once the initial guesses for the parameters are made, it helps to plot the dielectric
constant and conductivity data vs. frequency. These are plotted over another set of curves
representing the Three-Element Model dielectric constant and conductivity curves obtained using
the initial guess parameters. Use a spreadsheet so changes made in the initial guess parameters
are reflected as they are made. Slightly change the initial guesses by trial and error to provide a
closer fit to the experimental data. Start by manipulating only the parameters kr, ks, and d. Keep
the a parameter a function of c and b, and try to keep changes to r and c to a minimum. Keep the
b parameter approximately zero for uncemented soils.
Once the parameters have been more closely fitted to the experimental data, a computer
program may be used to optimize the parameters.

50

3.7.3 Computer Optimization

The computer optimization program is included on the disks enclosed with this book.
The program runs in DOS mode for PCs. Keep all files under the same directory (optimize on
the disk). The computer program optimizes based on ranges specified for each Three-Element
Model parameter. The ranges are based on initial guesses for each parameter. Use a minimum of
11 frequency-dielectric constant and frequency-conductivity conductivity data points to define
the dispersion curves. If the dispersions are not smooth, it is very helpful to smooth over the
points and create new curves to enter into the program. If there are not enough points to use in
the program, interpolate to create an adequate number of points.
3.7.4 Running the Program

The optimization program works in two steps: 1) the input step, and 2) the optimization
step.
The Input Step

1) To run the input step, input under the program directory (optimize on the disk).
2) The prompt will ask for a file name. Keep the name to less than 8 digits as usual for DOS
applications. Do not type a file extension.
3) The prompt will ask for comments. Make any relevant comments to help identify the soil or
input file.
4) The prompt will ask for the S. Enter 80, which is the appropriate dielectric constant of water.
5) The prompt will ask for the number of data points. Type in the number of points to be
optimized. (This number should be at least 11.)
6) The program will begin prompting for frequencies and relevant dielectric constants
and conductivities one by one. Conductivities units are in mho/cm x 10-3. For
example, 0.0005 mho/cm is entered as 0.5.
7) The program will prompt for a number of generations. Use 500.
8) The program will prompt for a number of individuals. Use the number of data points.
9) The program will prompt for a lower limit for b, and then for an upper limit. Use low
values such as 0.0001 and 0.01 for uncemented soils.
10) The program will prompt for a lower limit for c, and then for an upper limit. Use a
range with the approximated initial guess as the median.
11) The program will prompt for a lower limit for d, and then for an upper limit. Give

51

the d parameter plenty of range for the computer program to work with. The initial
guess based on trial and error will indicate whether d is on the low (0.5) or on the
high (0.999) end.
12) The program will prompt for a lower limit for r, and then for an upper limit.
Typically it is useful to start with limits of 0.1 above and below the initial guess.
13) The program will prompt for a lower limit for kr, and then for an upper limit. The
limits should be expressed as mho/cm. kr is usually larger than ks for clay systems.
Give the optimization program a fair amount of range to work with for kr.
14) The program will prompt for a lower limit for ks, and then for an upper limit. The
limits should be expressed as mho/cm. A decent upper limit for ks is slightly above
the measured pore fluid conductivity.
The Optimization Step

1) Type opt under the appropriate directory (optimize on the disk).


2) Enter the input file name without an extension.
3) The computer will prompt for an iseed number. Use 0 at first. If this produces few
iterations and high error (greater than about 0.3), use 100 for iseed.
4) The optimization program will optimize giving columns of b, c, d, r, kr, ks and the error.
The last row with the minimum error is the row of optimized parameters.
Adjusting

1) If any of the optimized parameters are at the very edge of the initially prescribed ranges, edit
the input file to slightly change the ranges for the parameter and then reoptimize. Repeat if
necessary until the error is minimized and none of the optimized parameters occur at the ends
of the chosen ranges.

2) To change the input file, under the appropriate directory prompt (optimize) type edit
filename.in, where filename is the name of the input file. Use the curser to change the
parameters. The parameter ranges are listed from top to bottom in the following order: b, c,
d, r, kr, kS.
3) Lastly, plot the Three-Element Model optimized results over the experimental data to check
the optimization. Slightly manipulate the parameters by hand to achieve an even better fit if
needed.

52

3.7.5 Example Problem

This section describes the optimization procedure and results for the Marysville soil
electrical dispersion shown in Figure 4.3. Table 3.3 contains electrical dispersion data extracted
from Figure 4.3.
Table 3.3
Electrical Dispersion Data for Marysville Soil
(Extracted from Figure 4.3)
frequency
MHz
3
4
6
8
10
15*
20
25*
30
35*
40

76
74.5
69.6
63.6
59.5
55.2
50.9
49.4
47.9
47.1
46.4

mho/cm
0.000380
0.000410
0.000420
0.000440
0.000460
0.000495
0.000530
0.000565
0.000600
0.000625
0.000650

* interpolated values

Based on Figure 4.3, 0 is approximately 83 40 = 43. Based on an of 40, the


porosity is approximately 0.59 (taken from Figure 4.11). Thus the void ratio, eT, is calculated as
1.439. Figure 4.7 shows that for a 0 of 43, is approximately 0.21. Dividing this value by 8
gives an approximate value (slope of the swelling line) of 0.026. Given and Figure 4.14,
eC/eT is estimated as 0.7; therefore, eP/eT is approximately 0.3.
Looking at Figure 3.22, a first approximation for c is 0.12. eC is calculated given eT and
eC/eT. eC = 1.439(0.7) = 1.007. nC is calculated as eC/(1+eC) = 0.502. The first approximation
for er is then nC(80) + (1+nC)4.5 = 42.
Using er = 42, c = 0.12, es = 80, b = 0, and a = 1- b c, manipulate the Three-Element
Model parameters kr, ks and d by trial and error to give an approximate fit to the data as shown in
Figure 3.22.

53

Dielectric and Conductivity Dispersion Curve

90

0.0007

dielectric constant,

0.0006

70

0.00055

60

0.0005
50
0.00045
40
0.0004
30

0.00035

three element model


20

0.0003

10

0.00025

0
1

10

conductivity, mho/cm

0.00065

80

0.0002
1000

100

freqency, MHz

Three Element Model Parameters


S

79

0.0000

0.120

0.800

42.0

kr
kS
(mho/cm) (mho/cm)
0.000580 0.000200

a
0.880

Figure 3.22
First-Guess Approximations of Three-Element Model Parameters

Therefore, the following ranges are specified for the parameters in this example.
b
c
d
r

0.00001
0.07
0.75
30

to
to
to
to

0.01
0.22
0.999
50

kr

0.0004

to

0.0007

mho/cm

kS

0.0001

to

0.0003

mho/cm

The input program is run using the data and the parameter ranges, and the data is optimized using
the opt command.
The results using iseed = 0 produce few iterations and high error (over 0.4). Therefore,
the opt command is reissued, and 100 is used for iseed. The following results are obtained:

54

b
c
d
r

0.00001
0.07
0.75
34.06

kr

0.000588

kS
error

0.000188
0.3051

This error is still high. Notice that c and d are on the borders of their parameter. On the
command line, enter edit mary.in where mary is the specified filename, and change the c and
d ranges to:
c
d

0.05
0.6

to
to

0.2
0.85

Save the new input file, T.


The optimization program is rerun with the opt command (iseed = 100). The new
optimized parameters are:
b
c
d
r

0.00001
0.05
0.6
30.41

kr

0.000553

kS
error

0.000252
0.19

Notice the error is reduced, but c and d still fall on the borders of the specified ranges, and r is
now bordering on the earlier specified minimum of 30. Change the ranges in the input file to:
c
d
r

0.04
0.5
25

to
to
to

0.19
0.8
45

Save the new input file and rerun the opt command, again using iseed = 100. The new results
are:
b
c
d
r

0.000322
0.05157
0.5961
29.08

kr

0.000588

kS
error

0.00026
0.2

55

Note that the results are now all within the specified ranges, and the error is no longer reduced by
optimization. These Three-Element Model data are now plotted against the electrical dispersion
data in Figure 3.23.
Dielectric and Conductivity Dispersion Curve

90

0.0006

70

0.00055

60

0.0005
50
0.00045
40
0.0004
30

three element model

0.00035

20

0.0003

10

0.00025

0
1

10

conductivity, mho/cm

0.00065

80

dielectric constant,

0.0007

0.0002
1000

100

freqency, MHz

Three Element Model Parameters


S

79

0.0003

0.052

0.596

29.1

kr
kS
(mho/cm) (mho/cm)
0.000588 0.000260

a
0.948

Figure 3.23
Electrical Dispersion Data and Three-Element Model Optimization
of Marysville Soil

Because the optimization fits the experimental data reasonably well, the results can be accepted.
Minor hand modifications of the results are justified if necessary.
3.8 SUMMARY

We can use nondestructive technology to accurately classify clay and non-clay soils. The
composition and structure of intact soil, determined by the non-destructive radio frequency
electrical dispersion technique, provides a method to classify intact soils either as clay minerals
or non-clay minerals based on the magnitude of dielectric dispersion. The magnitude of
dielectric dispersion depends on mineralogy, with montmorillonite showing the largest
magnitude of dielectric dispersion (around 250-300), and kaolinite showing a magnitude of
dielectric dispersion of around 10-15. Sand exhibits no dielectric dispersion; i.e., zero magnitude
of dielectric dispersion. Mixtures of non-clay and clay minerals exhibit magnitudes of dielectric
dispersions between 0 and 300.

56

3.9 PROBLEMS

1. Explain the following terms:


a) dielectric constant
b) polarizability
c) dielectric relaxation
d) anomalous dispersion
e) current density
f) dielectric dispersion
2. Explain why clay minerals show a decrease in dielectric constant with an increase in
frequency, whereas the dielectric constant of non-clay minerals is independent of frequency.
REFERENCES

Alizadeh, A. (1975) Amount and Type of Clay and Pore Fluid Influences on the Critical Shear
Stress and Swelling of Cohesive Soils, Thesis submitted in partial satisfaction of the
requirements for the degree of Doctor of Philosophy in Engineering, University of California,
Davis.
Arulanandan, K. (1980). ECI 283 class notes, Department of Civil and Environmental
Engineering, University of California, Davis.
Arulanandan, K. (1987). Non-Destructive Characterization of Particulate Systems for Soil
Classification and In Situ Prediction of Soil Properties and Soil Performance. Keynote Lecture,
Proceedings of the International Conference in Geotechnical Engineering, Calgary, Canada; A.A.
Balkema, Rotterdam.
Arulanandan, K. (1991). Dielectric Method for the Prediction of Porosity of Saturated Soils,
Journal of Geotechnical Engineering, ASCE, Vol. 117, No. 2, February, pp. 319-330.
Arulanandan, K. (1997). ECI 174 class notes, Department of Civil and Environmental
Engineering, University of California, Davis.
Arulanandan, K., and Arulanandan, S. (1985). Dielectric Methods and Apparatus for In Situ
Prediction of Porosity, Specific Surface Area (i.e., Soil Type), and for Detection of Hydrocarbon,
Hazardous Waste Materials, and the Degree of Melting Ice and to Predict In Situ Stress-Strain
Behavior. Patent Serial No. 709,592. Regents of the University of California.
Arulanandan, K., Basu, R., and Scharlin, R. J. (1973). Significance of the Magnitude of
Dielectric Dispersion in Soil Technology, Highway Research Board, Highway Research
Record, No. 426, pp. 23-32.
Arulanandan, K., and Mitchell, J. K. (1968). Low Frequency Dielectric Dispersion of ClayWater-Electrolyte Systems, 4th Annual Meeting of the Clay Mineral Society, Denver,
Colorado.

57

Arulanandan, K., and Mitra, S. K. (1970). Soil Characterization by Use of Electrical Network,
Proceedings of the Fourth Asilomar Conference on Circuits and Systems, November 1970, pp.
480-485.
Arulanandan, K., and Smith, S. S. (1973). Electrical Dispersion in Relation to Soil Structure,
Journal of the Soil Mechanics and Foundations Division, ASCE, Vol. 99, No. SM12, December,
pp. 1113-1131.
Arulanandan, K., Smith, S. S., and Spiegler, K. S. (1974). Radiofrequency Properties of
Polyelectrolyte Systems, Polyelectrolytes, D. Reidel Publishing Company, Dordrecht, Holland,
pp. 301-321.
Arulanandan, K., Yogachandran, C., and Rashidi, H. (1994). Dielectric Conductivity Methods
of Soil Characterization, Geophysical Characterization of Sites, Volume Prepared by ISSMFE
Technical Committee #10 for the XIII International Conference on Soil Mechanics and
Foundation Engineering, New Delhi, India, pp. 81-90.
Arulanandan, K., and Yogachandran, C. (2000). Dielectric Method for Non-Destructive
Characterization of Soil Composition, GeoDenver 2000, ASCE Geo-Institute, ASCE Special
Publication.
Basu, R., and Arulanandan, K. (1973). A New Approach for the Identification of Swell
Potential of Soils, Proceedings of the Third International Conference on Expansive Soils, Haifa,
Israel, Vol. 1, pp. 1-11.
Bolt, G. H. (1956). Physico-Chemical Analysis of the Compressibility of Pure Clays,
Geotechnique, London, England, Vol. 6, pp. 86-93.
Cremers, A., Van Loon, J., and Laudelout, H. (1966). "Geometry Effects for Specific Electrical
Conductance in Clays and Soil." Fourteenth National Conference on Clays and Clay Minerals,
pp. 149-162.
Debye, P. (1929). Polar Molecules, New York, reprinted by Dover Publications, Inc., New
York.
Hill, N. E. (1969). Dielectric Properties and Molecule Behavior, Van Nostrand Reinhold,
London.
Lambe, T. W., and Martin, R. T. (1953-1957). Composition and Engineering Properties of
Soil, Proceedings of the Highway Research Board, I-1953, II-1954, III-1955, IV-1956, and V1957.
Lambe, T. W. (1958). The Structure of Compacted Clay, Journal of Soil Mechanics and
Foundations Division, ASCE, Vol. 84, No. SM2, Proceeding Paper 1654, May, pp. 1654-4-165434.

58

Lambe, T. W. (1960). A Mechanistic Picture of the Shear Strength of Clays, Proceedings:


Conference on Shear Strength of Cohesive Soils, ASCE, Boulder Colorado, pp. 555-580.
Madden, T. R., and Marshall, D. J. (1959). Induced Polarization Study of the Causes and
Magnitudes in Geological Materials; Final Reports for Atomic Energy Commission,
Unpublished.
Martin, R. T. (1960). Adsorbed Water on Clay: A Review, Proceedings: Ninth National Clay
Conference, Purdue University, Lafayette, Indiana, October.
Meegoda (1985) Fundamental Characterization of Soils for the Development of an Expression
for Permeability for Application in In Situ Testing Thesis submitted in partial satisfaction of the
requirements for the degree of Doctor of Philosophy in Engineering, University of California,
Davis.
Mitchell, J. K. (1956). The Importance of Structure to the Engineering Behavior of Clay,
Thesis presented to the Massachusetts Institute of Technology at Cambridge, Mass., in partial
fulfillment of the requirements for the degree of Doctor of Philosophy.
Mitchell, J. K., and Arulanandan, K. (1968). Electrical Dispersion in Relation to Soil
Structure, Proc. ASCE, Vol. 94, SM2, pp. 447-472.
Muat, R. W. (1968). Driving-Point Characteristic Approximation with RC Network,
Submitted in partial satisfaction of the requirements for the degree of Master of Science,
University of California, Davis.
Okonski, C. T. (1960). Electrical Properties of Macromolecules, V. Theory of Ionic
Polarization in Polyelectrolytes, Journal of Physical Chemistry, Vol. 64, pp. 605-619.
Olsen, H. W. (1961). Hydraulic Flow Through Saturated Clays, Thesis presented to
Massachusetts Institute of Technology at Cambridge, Mass., in partial fulfillment of the
requirements for the degree of Doctor of Science.
Olsen, H. W., (1962). Hydraulic Flow through Saturated Clay, Proceedings of the Ninth
National Conference on Clays and Clay Minerals, pp. 131-161.
Rosenquist, I. Th. (1955) Investigations on the Clay-Electrolyte-Water Systems, Norwegian
Geotechnical Institute, Publication 9
Rosenquist, I. Th. (1958). Physico-Chemical Properties of Soils: Soil Water Systems, Journal
of the Soil Mechanics and Foundations Division, ASCE, Vol. 85, No. SM2, Proceedings Paper
2000, April, pp. 31-53.

59

Sachs, S. B., and Spiegler, K. S. (1964). Radiofrequency Measurements of Porous Conductive


Plugs, Ion-Exchange Resin-Solution System, Journal of Physical Chemistry, Vol. 68, pp. 12141222.
Sauer, M. C., Southwick, P. F., Spiegler, K. S., and Wyllie, M. R. J. (1955). Electrical
Conductance of Porous Plugs, Ind. and Eng. Chemistry, Vol. 47, pp. 2187-2193.
Schwan, H. P., Schwarz, G., Maczuk, J., and Pauly, H. (1962). On the Low Frequency
Dielectric Dispersion of Colloidal Particles in Electrolyte Solutions, Vol. 60, pp. 2626-2642.
Schwarz, G. (1962). A Theory of the Low-Frequency Dielectric Dispersion of Colloidal
Particles in Electrolyte Solution, Journal of Physical Chemistry, Vol. 66, pp. 2626-2642.
Seed, H. B., and Chan, C. K. (1959). Structure and Strength Characteristics of Compacted
Clays, Journal of the Soil Mechanics and Foundations Division, ASCE, Vol. 85, No. SM5,
Proceedings Paper 2216, October, pp. 87-128.
Smith, S. S., and Arulanandan, K. (1981). Relationship of Electrical Dispersion to Soil
Properties, Journal of Geotechnical Engineering, ASCE, Vol. 107, No. GT5, May. pp. 591-604.
Vacquier, et al. (1957). Prospecting for Ground Water by Induced Electrical Polarization,
Geophysics, Vol. 22.
Wagner, K. W. (1914). Archives of Electrotechnology, Vol. 2, p. 371.
Weiler, R. A., and Chaussidon, J. (1968). Surface Conductivity and Dielectrical Properties of
Montmorillonite Gels, Clays and Clay Minerals, Vol. 16, pp. 147-155, Pergamon Press.

60

Chapter 4
Electrical Dispersion and Engineering
Properties
4.1 INTRODUCTION
The previous chapter demonstrates how the factors controlling the structure of soils
influence electrical dispersion behavior. Because the electrical dispersion and structure of soils
are influenced by the same factors, it follows that electrical dispersion can be used to quantify
soil structure. This chapter shows how the electrical dispersion method can be used to quantify
soil structure to evaluate engineering properties. Thus, the magnitude of dielectric dispersion can
be correlated with
The cation exchange capacity, CEC
Specific surface area, SSA

Critical state friction angle,


Compression index due to loading,
Swell index due to unloading,
Swell potential
Total porosity, n
Intercluster porosity, np
Intracluster porosity, nc
Hydraulic conductivity, kh

4.2 ELECTRICAL DISPERSION METHOD OF SOIL CHARACTERIZATION


The dielectric dispersion method of non-destructively characterizing intact soil can be
used to differentiate clay minerals from non-clay minerals. The method is rapid, and only two
measurements of dielectric constant (one at a high frequency of about 50 MHz, and the other at a
low frequency of about 2 MHz) are required. Figure 4.1 shows experimental and theoretical
dielectric dispersion relationships for various soils.

61

Figure 4.1
Experimental and Theoretical Dielectric Dispersions of Clay
(Montmorillonite, Illite, and Kaolinite) and Non-Clay (Silica Flour) Samples (Arulanandan
and Yogachandran, 2000)
The dielectric constant of clay minerals varies with frequency, while the dielectric
constant of non-clay minerals is independent of frequency. Thus, the magnitude of dielectric
dispersion can be used to determine whether a soil contains clay minerals.
The magnitude of dielectric dispersion, 0 , for a natural soil consisting of clay and nonclay minerals is the difference in the dielectric constants measured at about 2 MHz and 50 MHz.
These magnitudes are shown in Figure 4.2 for a hypothetical clay and in Figure 4.3 for a
Marysville soil.
As seen in Figure 4.3, the magnitude of dielectric dispersion, 0 , of the Marysville soil
is roughly 0 = 83-42 = 41. As seen in Figure 4.1 different soils may exhibit different
magnitudes of dielectric dispersion. The magnitude of dielectric dispersion depends on the
mineralogy. Montmorillonite shows the largest magnitude of dielectric dispersion, followed by
illite and kaolinite.
The magnitude of dielectric dispersion, 0 , is a parameter that can be used to generally
characterize intact soils. Because the dielectric constant of non-clay minerals is independent of
frequency, the dispersion can be used to determine whether a soil contains clay minerals. 0
can thus be used to measure the type and amount of clay minerals. Figure 4.4 illustrates the
effect of clay type and amount on the magnitude of dielectric dispersion.

62

ff

2 MHZ

50 MHz

Figure 4.2
The Magnitude of Dielectric Dispersion, 0

Figure 4.3
Electrical Dispersion and the Magnitude of Dielectric Dispersion, 0 ,
of a Marysville Soil (Arulanandan and Yogachandran, 2000)
In preparing the samples used in Figure 4.4, montmorillonite, illite and kaolinite were
mixed with different percentages of sand. Sand, by itself, demonstrates no electrical dispersion
due to lack of surface charge. The addition of sand lowers the magnitude of dielectric dispersion
of clay soils.
As described earlier, we know pore fluid concentration affects electrical dispersion
(Figure 3.14.) Figure 4.5 shows the influence of clay type and amount and pore fluid
composition on the magnitude of dielectric dispersion.

63

Figure 4.4
The Magnitude of Dielectric Dispersion vs. Clay Type and Percent.
(Alizadeh, 1975)

Figure 4.5
The Magnitude of Dielectric Dispersion vs. Clay Type, Clay Amount, and Pore Fluid
Concentration (Arulanandan, et. al., 1994)

64

4.3 THE MAGNITUDE OF DIELECTRIC DISPERSION AND SWELL POTENTIAL


Terzaghi (1931) and later Komornik and David (1969) presented convincing evidence to
show that both mechanical and physico-chemical factors control the swelling potential of a soil.
Seed, et al., (1962) have showed that the swelling potential of soil is mainly controlled by the
type and amount of clay minerals. These factors also control the magnitudes of dielectric
dispersion (Arulanandan and Smith, 1973). Table 4.1 gives compositional properties and the
percentage of swell for several soils (Basu and Arulanandan, 1973). Table 4.1 also gives the
magnitude of dielectric dispersion determined for these soils at saturation. Each of the soils listed
was consolidated from a slurry under a pressure of 1 kg/cm2. A cylindrical specimen
approximately 1 inch high and 1.4 inch in diameter was allowed to dry at 65F and at 50 percent
relative humidity. Each dried sample was then confined laterally and allowed to swell freely
with no surcharge. Water was introduced by placing the soil on a porous stone that remained
saturated with water during the test. Percentage swell given in Table 4.1 is the total swell
expressed as a percentage of the initial height of the sample.
Table 4.1
Mineral Composition, 0 , and Swell % of Natural and Prepared Soil Samples (Basu and
Arulanandan, 1973)

Figure 4.6 shows a correlation between the total percentage of swell and the magnitude of
dielectric dispersion, 0 , measured on samples at the end of the swell test.

65

Figure 4.6
Magnitude of Dielectric Dispersion vs. Total Swell Percent for Natural and Prepared Soil
Samples (Muraleetharan and Arulanandan, 1986)
A good correlation between the percentage of swell and 0 exists for artificial and
natural soils. This relationship can be used to evaluate the swell potential of soils once the
magnitudes of dielectric dispersion of saturated samples are known.
4.4 THE MAGNITUDE OF DIELECTRIC DISPERSION, 0 , AND
COMPRESSION INDEX
It has been shown that the type and amount of clay mineral significantly influences 0 .
The values of 0 have been shown to increase in the sequence kaolinite < illite <
montmorillonite. The compression indices of these soils also increase in this sequence. As shown
in Figures 3.10, 4.4, and 4.5, both the magnitude of dielectric dispersion and the compression
index increase with an increase in the percentage of clay in sand-clay mixtures (Arulanandan, et
al.,, 1973; Alizadeh, 1975; Arulanandan and Smith, 1973). Olson and Mesri (1970) have shown
that the compression index of kaolinite decreases when the electrolyte concentration is increased
from 0.0001 N Sodium to 1.0 N Sodium. Results presented in Figures 3.14 and 4.5 (Arulanandan
and Smith, 1973) show that 0 also decreases with increasing electrolyte concentration.
Since the clay mineral type and amount and pore fluid composition influence the
compression index and the magnitude of dielectric dispersion ( 0 ), a correlation between 0
and (slope of the consolidation line in e -ln p space) is expected. Figure 4.7 shows the
relationship between 0 and . This relationship was established mainly for clay minerals.

66

Figure 4.7
Slope of the Virgin Consolidation Line, , vs. Magnitude of Dielectric Dispersion, 0
(Arulanandan, et al., 1983)
Mitchell (1960) and Olson and Mitronovas (1962) investigated the influence of particle
size on compressibility characteristics. Their results show that particle size influences the
compressibility of sand-clay mixtures. The influence of particle size on was investigated
using three soils consisting of 3.3%, 5%, and 10% montmorillonite mixed with 96.7%, 95% and
90% Monterey 10/30' sand of particle size 0.3 mm. Figure 4.7 plots the measured values of
and 0 for these soils. The relationship obtained for the montmorillonite-sand mixtures is
lower than that obtained for clay minerals, probably because of the mechanical interaction
between the larger particles. The values of increase with an increased amount of clay mineral.
This is possibly due to the reduction in mechanical interaction between particles. The values of
and 0 were also measured for silt size silica flour (particle size = 0.01 mm) mixed with
3.3% and 5% montmorillonite, respectively. Figure 4.7 shows that the measured values of and
0 for these mixtures follow the relationship established for clay minerals. Thus, the
relationship established for clay minerals is valid for mixtures of clay and non-clay minerals in
which the non-clay minerals have particle sizes less than 0.01 mm. For soils with particle sizes
greater than 0.01 mm, different relationships similar to those developed for the Monterey sand
and montmorillonite mixtures shown in Figure 4.7 must be established.
4.5 CRITICAL STATE FRICTION ANGLE
Triaxial tests were conducted on samples to determine the slope of the critical state line
in p'-q space, M. Electrical dispersion tests were conducted on the same samples. The following
results show the correlation between M and the magnitude of dielectric dispersion.

67

M vs. Magnitude of Dielectric Dispersion

Slope of the , M

1.3
1.2
1.1
1
0.9
0.8
0.7
0

20

40

60

80

Magnitude of Dielectric Dispersion, o

Figure 4.8
Slope of the Critical State Line in p'-q Space, M, vs. the Magnitude of Dielectric Dispersion
Figure 4.8 shows the relationship for clays with 0 varying between 10 and 60. To calculate
the critical state friction angle, , the following formula can be used:
6 sin
M=
(1)
3 sin
4.6 THE MAGNITUDE OF DIELECTRIC DISPERSION, 0 ,
AND CATION EXCHANGE CAPACITY
The cation exchange capacity (CEC) of soils is important for use in both agriculture and
engineering. In agriculture, CEC is important when considering saline and alkali soil
reclamation, selection of soil fertility practices, and prediction of drainage water. In engineering,
CEC is important when considering electro-osmotic dewatering efficiency and erodability. CEC
is also used in soil classification.
This section details the evaluation of soil CEC using the magnitude of dielectric
dispersion in both field and laboratory. Conventional methods for determining soil CEC are time
consuming. Furthermore, conventional analytical procedures do not necessarily reflect field
conditions: They may alter the pH and ionic strength of the pore fluid, which directly influence
the CEC.
Recalling Figure 4.4, 0 varies with both clay type and clay percent. Montmorillonite
clay minerals produce the largest magnitude of dielectric dispersion, followed by illite and then
kaolinite. As shown in Figure 3.9, the percentage of clay mineral also influences the magnitude
of dielectric dispersion, 0 .

68

The CEC of montmorillonite, illite, and kaolinite were determined to be 95, 29.9, and 7.5
meq/100g, respectively, at a pH of 7. CECs of the various sand-clay mixtures shown in Figure
4.4 are calculated based on the relative mixture percentages, assuming sand has zero CEC. This
data shows a trend relating CEC to 0 as illustrated in Figure 4.9. However, the slope of the
correlation line is different when natural soils are considered.

Figure 4.9
Magnitude of Dielectric Dispersion vs. Cation Exchange Capacity
for Natural and Artificial Soils (Fernando, et al., 1977)
The reason for this discrepancy between artificial and natural soils is potentially due to
field conditions, such as the presence of organic matter, iron oxides, carbonates, or silica, which
act as cements or relatively inelastic bridges between charged colloidal particles. These variables
do not occur in the artificial soil samples produced in the lab. Ultimately CEC may then be
obtained for field- or laboratory-prepared samples based on 0 .
4.7 AND POROSITY
Figure 4.10 depicts another important dielectric index, , the dielectric constant at which
the dielectric dispersion levels off.

69

Figure 4.10
and 0

is usually defined at high frequencies above 50 MHz. However, most soils exhibit a leveling
off of dielectric dispersion around 50 MHz. Therefore, this frequency is often used for the
determination.
The formation factor, under very high frequencies, i.e., as , reduces to the
following when measurements are made in the direction of loading (Arulanandan, 1991):

P
1 P
1 n
(FV ) = 1 +
+
( 2 1 ) A 1 + ( 2 1 ) A
n
1+
a
b
1
1

(2)

and reduces to the following when measurements are made at right angles to the direction of
loading:

1 P
1 + P
1 n
(FH ) = 1 +
+
( )
2n ( 2 1 )
Aa 1 + 2 1 Ab
1+
1
1

(3)

where:
n = porosity
1 = dielectric constant of the solution
2 = dielectric constant of the particle
P = an orientation parameter
Aa, Ab = shape factors

For transversely isotropic soils (Fricke, 1924):

70

Aa + 2Ab = 1

(4)

For spherical particles, Aa = Ab = 1/3. By assuming the particles are spherical, and using the
relationship developed for the average formation factor,
F + 2FH
F= V
3

(5)

21
1
1 n

F = 1 +
+
(
(
2 1 )
2 1 )
3n
1 +
1 +

3
3

(6)

with some manipulation, we obtain:

which simplifies further to:


F = 1 +

1 n 31

n 21 + 2

(7)

Arulanandan (1991) also showed that at high frequencies (such as approximately 50 MHz):

(FV ) =

1 2
( V ) 2

(8)

(FH ) =

1 2
( H ) 2

(9)

where V and H are measured dielectric constants of the soil-particle/pore-fluid mixture in the
vertical and horizontal directions, respectively. Assuming an isotropic condition, the following
equation is obtained for the average formation factor at high frequencies:

(F ) =

1 2
2

(10)

where = (V ) = ( H )
By setting this equation for (F ) equal to 1 +

1 n 31
as derived earlier, we obtain:

n 21 + 2

71

1+

1 n 31
=

n 21 + 2

1 2
2

(11)

and through further algebraic reductions we can eventually solve for porosity alone, obtaining:
n=

1
1 2 2
+ 1

+
3
3

2
1

(12)

By plotting porosity vs. , using 1 = 80 for water and 2= 4.5 for the soil particles. we obtain
the relationships shown in Figure 4.11.

Figure 4.11

vs. Porosity: Theoretical Plot and Experimental Data


(after Arulanandan, 1991)

Note that in Figure 4.11, the theoretical plot shows calculations using axial ratios, R, of
1.0 (spherical particles), 0.1, and 0.001. The vs. porosity relationship shown in Figure 4.11
illustrates that axial ratio plays a relatively minor role in porosity prediction using .
4.8 SPECIFIC SURFACE AREA

Section 3.3 demonstrated the following equation:


1

s =
1 + s
2(SSA ) s R
where S = the surface conductance,

(13)

72

s = the soil particle density,


R = the soil axial ratio, and
s = the conductivity of the soil particle

The conductivity of the soil particle, s, can be replaced by kr, one of the Three-Element Model
parameters. As shown in Section 3.3, because the conductivity due to charges within the particle
can be neglected, the conductivity of the particle, k2, is approximately equal to s. The surface
conductance becomes equivalent to:

s =
(14)
1 + k r
2(SSA ) s R
where SSA = the specific surface area, area/mass.
The axial ratios, R, of pure montmorillonite and kaolinite are about 100 and 10,
respectively (Lambe and Whitman, 1969). Therefore the 1/R term may usually be neglected for
clay minerals. As discussed in this chapter, the surface conductance of clay minerals obtained by
many researchers is on the order of 1 x 10-9 mho (Cremers, et al., 1966; Weiler and Chaussidan,
1968; Schwan, et al., 1962; Fricke and Curtis, 1937). The specific surface area can therefore be
evaluated by rearranging the above equation, neglecting the 1/R term:

SSA =
kr
(15)
2 s s
Figure 4.12 illustrates the relationship between kr and SSA for several soils.
kr vs. Specific Surface Area
35
30% montmorillonite, 70% Snow Cal

25
20
experimental

marysville red

kr (x10 ) mho/cm

30

15

5% montmorillonite
95% snow cal

illite
ssa calculated based on
measured hydraulic
conductivity

yolo loam

10
kaolinite

5
0
0

20

40

60

80

100

120

140

160

180

Specific Surface Area, SSA, m /g

Figure 4.12
Kr vs. Specific Surface Area, SSA
*calculated SSA evaluated using measured hydraulic conductivity and the modified Kozeny-Carman
equation as detailed in Section 4.10

Thus, once kr is known, the specific surface area can be evaluated.

73

4.9 THE SLOPE OF THE UNLOAD-RELOAD LINE,

Many researchers have investigated the swelling characteristics of saturated clays due to the
removal of external load. Mechanical models, such as the one used by Terzaghi (1929), have
assumed that swelling results from elastic rebound of bent particles. Physico-chemical models,
such as the one used by Bolt (1956), have assumed that osmotic repulsive forces cause swelling.
Although it was possible to explain the mechanisms controlling swelling using the above
concepts, the models were not very successful owing to the complicated structural arrangements
of particles in clays.
Olsen (1961) used the concept of clusters in fine-grained soils, discussed by Michaels and
Lin (1954), and Quirk (1959), in his study of hydraulic flow through saturated clays. Olsen
concluded that discrepancies between the measured and calculated permeabilities are mainly due
to unequal pore sizes due to grouping of clay particles in clusters. Many have observed the
existence of primary particle aggregation using electron microscopes (Quigley and Thompson,
1966). Figure 4.13 shows the variation of intra- (ec) and intercluster (eP) void ratios obtained
using the Three-Element Model, with total void ratio using electrical dispersion data for snow cal
(95%) + montmorillonite (5%). The results, corresponding to vertical and horizontal
measurements, are identical and are very similar to the ones predicted by Olsen (1961), Meegoda
(1983), and Abdullah (1983).

Figure 4.13
Variation of Intracluster (ec) and Intercluster (eP) Void Ratio with Total Void Ratio (e) for
95% Snow Cal + 5% Montmorillonite (after Meegoda, 1983)

It has been shown that the swelling of fine-grained soils is caused by swelling of clusters (Smith
and Arulanandan, 1981), and the decrease in intercluster pores during compression is irreversible
(Meegoda, 1983). If the ratio of intracluster to total void ratio is large for a given soil, the elastic
compression due to an increase in external load would be high. Consequently, swelling would
also be high when the load was removed. Assuming this mechanism of swelling, the ratio of ec/e
is correlated with , the slope of the swelling line, as shown in Figure 4.14.

74

Figure 4.14
Slope of the Swelling Line, , vs. Intracluster/Total Void Ratios, ec/e
(Arulanandan, 1987)

In order to use this relationship for , we must be able to quantify ec/e. Recalling from Equation
(7):
r = n C S + (1 n C ) P
(16)
we may obtain nC, the intracluster porosity, if r is known. The value of r is obtained through
optimizing the Three-Element Model to fit electrical dispersion data of a given soil sample. The
optimization procedure was described in Chapter 3. Recall that S is 80 for pore water, and P
is typically about 4.5 for dry soil particles. ec is obtained by using the following volumetric
relationship:
n
(17)
ec = c
1 n c
The total void ratio may be calculated using the known water content of the sample:
e = wG s
(18)
where w is the water content in decimal form and Gs is the specific gravity of the soil, typically
about 2.7. An alternative method for determining the void ratio using the electrical parameter
was detailed in Section 4.7. Knowing ec and e, one can use the relationship in Figure 4.14 to
determine the slope of the swelling line, , for a given soil.

75

4.10 HYDRAULIC CONDUCTIVITY OF CLAYS USING THE ELECTRICAL


DISPERSION METHOD

The following equations describe hydraulic flow rate through liquid-saturated porous media.
These incorporate Darcys law and the Kozeny-Carman equation (Kozeny, 1927; Carman, 1956).
q=

p dh
k
dl

1
k=
k0T 2

(19)

e3 1

2
1 + e S0

(20)

where:
q = Darcy velocity of the permeant,
= viscosity,
p = unit weight of the permeant,
k0 = pore shape factor 2.5 for uniform pore sizes,
T = tortuosity 2 for sands,
S0 = specific surface area (particle surface area per volume),
e = void ratio,
dh/dl = hydraulic pressure gradient (ft/ft), and
k = permeability
The inadequacy of the above equation in evaluating the permeability, k, of clays has long been
recognized. The equation fails to predict the magnitude and the porosity dependencies of the
hydraulic flow rate. The equation does not account for the chemical composition of clays, the
chemical composition of the permeant, or the stress history of the soil (Michaels and Lin, 1954).
Olsen (1961) analyzed the possible causes for the inadequacy of the Kozeny-Carman equation to
predict the hydraulic conductivity of clays and concluded that the deviation of predictions could
only be due to unequal pore sizes. Olsen proposed the cluster concept to explain the deviation.
To evaluate the hydraulic conductivity of clays, it is therefore necessary to quantify the
inter and intracluster void ratios, the total void ratio, the specific surface area, pore shape, and
tortuosity parameters. Then, the modified Kozeny-Carman equation can be used:
kh =

ep3

k o T 2S0 2 1 + e T

(21)

where:
ep = the intercluster void ratio
eT = the total void ratio
kh = the hydraulic conductivity (velocity units)
p = the unit weight of the permeant (pore fluid)
= the viscosity of the pore fluid
ko = the pore shape factor
T = the tortuosity
S0 = the volumetric specific surface area of the soil (surface area/volume)

76

The term p/ is a constant of about 9.74 x 105 for water. Although the viscosity of
water is temperature dependent, temperature has only a minor effect. The term 1/koT2 is a
structural parameter denoting the shape and arrangement of particles. The volumetric specific
surface area, S0, may be rewritten as S0 = SSA x s , where SSA is the specific surface area in
units of area/mass and s is density of soil particles (mass/volume)
. The resulting equation becomes:
3
ep
1
1
cm
5

10 8
k h = 9.74 10
*
(22)

2
2
2
k oT 1 + eT SSA s
s
using units of m2/g for SSA, and g/cm3 for s .
The above equation is valid for clays. The original Kozeny-Carman equation presented at the
beginning of this chapter is not valid for clay minerals. The existence of unequal pore sizes
(Olsen, 1961) has been shown to be the reason for the inapplicability of the Kozeny-Carman
equation for the prediction of hydraulic conductivity of clays. Flow rate is based on the fourth
power of the radius of the pore. Pores between clusters have a much larger radius than pores in
the clusters, so more flow can travel through the intercluster pore space. Thus, we need to
quantify the intercluster void ratio in a clay to determine hydraulic conductivity. Section 3.7
showed an approximate method of quantifying the intracluster porosity based on the following
expression:
r = n c (80) + (1 n c )4.5
(23)
where
r = a Three-Element Model parameter representing the dielectric constant of the clay cluster,
nc = the cluster porosity,
80 = the dielectric constant of water, and
4.5 = the dielectric constant of soil particles.
Section 3.7 describes the Three-Element Model. Optimizing the Three-Element Model to
electrical dispersion data requires determining the dielectric and conductivity dispersion
characterization of a clay soil over the radiofrequency range, 2 50 MHz, as detailed earlier in
this chapter.
The only parameter left to be determined is 1/koT2. The c parameter in the Three-Element
Model is a function of tortuosity and porosity. It is approximately equal to k0. If c (a ThreeElement Model parameter) is approximated as 1/F, where F is the formation factor (described in
detail in Chapter 5), then k0T2 can be approximated by:
koT2 = np = cT2

(24)

where np is the intercluster porosity. The final Modified Kozeny-Carman equation then becomes:
1
k h = 9.74 10
np
5

e 3
p

1+ e
T

10 8 (cm / s )
SSA 2 G 2
s

(25)

This equation can be used to evaluate hydraulic conductivity as shown in the example problem
that follows.

77

Example Problem for Hydraulic Conductivity

Given the following electrical dispersion curves (dielectric constant and conductivity) for
illite (Figure 4.15) and the given optimized Three-Element Model parameters Table 4.2,
calculate the hydraulic conductivity of the soil.
75

0.0017
0.0016

70

Dielectric Constant

0.0014

three element model


60

0.0013

dielectric constant

0.0012

conductivity

55

0.0011
0.001

50

Conductivity, mho/cm

0.0015
65

0.0009
45
0.0008
40
1

0.0007
100

10
Frequency, MHz

Figure 4.15
Electrical Dispersion of Na+ Illite, and Optimized Three-Element Model Parameters
(After Smith, 1971)
Table 4.2
Optimized Three-Element Model Parameters
s

79

0.002

0.36

0.65

21

kr
ks
mho/cm mho/cm
1.500E-03 8.00E-04

a
0.638

Water content = 62%


Solution:

void ratio, e = wGs = 0.62(2.7) = 1.674


r = n c (80) + (1 n c )4.5
4 .5
21 4.5
therefore, n c = r
=
= 0.2185
80 4.5 80 4.5
nc
0.2185
ec =
=
= 0.2796
1 n c 1 0.2185
e p = e e c = 1.674 0.2796 = 1.3944

78

np =

ep
1 + eT

1.3944
= 0.5215
1 + 1.674

m2
m2

0.0015(mho / cm)
= 87.27
SSA =
kr =
2 s s
g
2 1 10 9 (mho) 2.7 (g / cm 3 )
10000 cm 2

10 8 (cm / s )
SSA 2 G 2
s

1.3944 3

1
1

10 8 (cm / s )

k h = 9.74 10 5

2
2

0.5215 1 + 1.674 87.27 2.7


cm
calculated
k h = 3.4 10 7
s
cm
(Smith, 1971). Using the
The measured hydraulic conductivity for this soil is 1.1 10 7
s
above detailed procedure, the hydraulic conductivity of several soils has been calculated and is
compared to the measured values in Figure 4.16.

k h = 9.74 10 5

1
np

e 3
p

1+ e
T

Evaluation of Hydraulic Conductivity Using Electrical Dispersion Method

measured hydraulic conductivity, (cm/s)

1.00E-05

95% snow cal


+ 5% montmorillonite

1.00E-06
yolo loam
60% snow cal + 40% illite
Marrysville red, CA

1.00E-07

yolo loam
illite example

illite
Encapco material, Kleinfelder

1.00E-08
1.00E-08

1.00E-07

1.00E-06

1.00E-05

calculated hydraulic conductivity, (cm/s)

Figure 4.16
Calculated (Theoretical) and Measured Hydraulic Conductivity of Various Soils

79

4.11 SUMMARY

The magnitude of dielectric dispersion has been correlated with swell potential as shown
in Figure 4.6, compression index as shown in Figure 4.7, critical state friction angle as shown in
Figure 4.8, and cation exchange capacity as shown in Figure 4.9. Dielectric constant and
conductivity vary with frequency in the radio frequency range. This variation has been used to
quantify the soil structure using a Three-Element electrical network Model described by
Arulanandan and Mitra (1970), Arulanandan and Smith (1973), and Smith and Arulanandan
(1981). The parameter r, representing the dielectric constant of the soil cluster, has been used
to evaluate intercluster and intracluster porosities. The dielectric constant at about 50 MHz
relates to total porosity (Arulanandan, 1991). The Three-Element Model parameter kr,
representing the conductivity of the soil cluster, relates to the specific surface area of the cluster
as shown in Figure 4.12. The ratio of the intracluster void ratio to the total void ratio relates to
the swell index due to unloading, , as shown in Figure 4.14.
Thus, one can approach the characterization and classification of intact soils and obtain
the following properties based on electrical dispersion characteristics:
The cation exchange capacity, CEC
Specific surface area, SSA

Critical state friction angle,


Compression index due to loading,
Swell index due to unloading,
Swell potential
Total porosity, n
Intercluster porosity, np
Intracluster porosity, nc
Hydraulic conductivity, kh

4.12 PROBLEMS

1. a) Plot the dielectric constant and conductivity variation in the frequency using the Rx meter
results of the two soils shown in the following tables.
b) Determine o for each soil.
c) Obtain appropriate values for the geometrical parameters (a, b, c, d) and compositional
parameters (r, s, ks, kr). Use the optimization program to determine and o.
d) Estimate the porosity, specific surface area, hydraulic conductivity, compression index,
and total swell potential for both soils.

80

Rx Meter Test Results


Control Soil Sample
Freq
2
4
8
16
32
40

L1
1.382
1.382
1.382
1.382
1.382
1.382

L2
0.75
0.75
0.75
0.75
0.75
0.75

l
1.855
1.855
1.855
1.855
1.855
1.855

Cp
12.5
9.42
7.6
6.41
5.5
5.25

C p'
19.74
14.08
10.80
8.57
6.61
6.04

Rp
290
284
279
270
254
245

R p'
Cp0

164
12.6 2.35
161
25.1 2.35
157
50.3 2.35
153 100.5 2.35
141 201.1 2.35
135 251.3 2.35

C0
0.0900
0.0900
0.0900
0.0900
0.0900
0.0900

Ct
2.200
2.200
2.200
2.200
2.200
2.200

Rs
Cs

2.76E+02 11.61 122.85 0.003388


2.72E+02 8.48 89.77 0.003442
2.80E+02 6.69 70.78 0.003345
2.94E+02 5.72 60.55 0.003186
-2.84E+03 6.04 63.94 -0.00033
-1.51E+02 6.83 72.31 -0.00621

Ct
2.320
2.320
2.320
2.320
2.320
2.320

Rs
3.54E+02
3.63E+02
3.89E+02
6.40E+02
-2.58E+02
-6.43E+01

Zygro Soil Sample


Freq
2
4
8
16
32
40

L1
1.382
1.382
1.382
1.382
1.382
1.382

L2
0.66
0.66
0.66
0.66
0.66
0.66

l
2.084
2.084
2.084
2.084
2.084
2.084

Cp
C p'
11.3 20.1
8.9 14.38
7.25 10.98
6.3 8.55
5.45
6.6
5.15 5.91

Rp
336
329
325
310
292
286

R p'
Cp0

155
12.6 2.35
153
25.1 2.35
150
50.3 2.35
144 100.5 2.35
135 201.1 2.35
129 251.3 2.35

C0
0.0900
0.0900
0.0900
0.0900
0.0900
0.0900

Cs

9.48 100.33 0.002644


7.53 79.72 0.002581
5.95 63.01 0.002407
5.73 60.67 0.001463
6.60 64.13 -0.00362
6.95 73.59 -0.01456

2.
A
2d

Given w/c = 40%


SSA= 300 m2/g
Show that you are within the range of the effect of the double layer.

3. Given the electrical dispersion characteristics of a soil in Figure 4.17 below, and the optimized
parameters of a Three-Element Model used to characterize the soil, evaluate the following
properties:
a) Total porosity, intra-and intercluster void ratios
b) Soil SSA, compression index, and swell index
c) Swell potential
d) The hydraulic conductivity of the soil, given a specific gravity for the soil of 2.75

81

Theoretical Dispersion Curve Data

63.48308
63.12631
61.77649
60.83841
54.83918
48.95364
44.45255
41.27760
39.06803
37.51096
36.39006
35.56424
34.94210
34.46373
34.08906
33.79074
33.54971
33.35240
33.18898
33.05219
32.93661
32.83110
32.75351

mho/cm
0.0009084
0.0009116
0.0009237
0.0009322
0.0009861
0.0010390
0.0010794
0.0011079
0.0011278
0.0011418
0.0011519
0.0011593
0.0011646
0.0011692
0.0011725
0.0011752
0.0011774
0.0011792
0.0011806
0.0011819
0.0011829
0.0011838
0.0011845

Dielectric and Conductivity Dispersion Curve


70

0.0014

60

0.0012

50

0.0010

40

0.0008

30

0.0006

20

0.0004

10

0.0002

0
1

10

0.0000
100

Frequency, MHz

Figure 4.17

Optimized Parameters

78

b
c
d

0
0.197
0.843

18.21

kr
ks

1.25E-03
4.99E-04

82

Conductivity

Dielectric Constant

Frequency
MHz
1
2
4
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100

4. You are given the following soil SSAs:


Soil
Montmorillonite
Illite
Kaolinite
Sand

SSA(m2/g)
500
160
20
1

Figure 4.18 shows the relationship between the percentage of clay minerals and the magnitude of
dielectric dispersion, 0 .

Figure 4.18
Effect of Clay Type and Amount on Magnitude of Dielectric Dispersion

Develop a relationship between the magnitude of dielectric dispersion, 0 and Specific Surface
Area, SSA. Give the results in graphical form.
5. Determine the hydraulic conductivity of the soil with the following electrical dispersion data
(Figure 4.19). What are some possible causes for discrepancies between the actual permeability
behavior of clays and that predicted by the Kozeny-Carman equation? What is the most likely
cause?

83

Figure 4.19
REFERENCES

Abdullah (1983) A Fundamental Approach for the Characterization of Mixed Soils to Predict
Stress Strain Behavior In Situ. Dissertation presented to the University of California, Davis, in
partial fulfillment of the requirements for the degree of Doctor of Philosophy in Engineering.
Alizadeh, A. (1975) Amount and Type of Clay and Pore Fluid Influences on the Critical Shear
Stress and Swelling of Cohesive Soils, Thesis submitted in partial satisfaction of the
requirements for the degree of Doctor of Philosophy in Engineering, University of California,
Davis.
Arulanandan, K. (1987). Non-Destructive Characterization of Particulate Systems for Soil
Classification and In Situ Prediction of Soil Properties and Soil Performance. Keynote Lecture,
Proceedings of the International Conference in Geotechnical Engineering, Calgary, Canada; A.A.
Balkema, Rotterdam.
Arulanandan, K. (1991). Dielectric Method for the Prediction of Porosity of Saturated Soils,
Journal of Geotechnical Engineering, ASCE, Vol. 117, No. 2, February, pp. 319-330.
Arulanandan, K., Anandarajah, A., and Meegoda, N. J. (1983). Soil Characterization for NonDestructive In Situ Testing, Symposium Proceedings Part 2: The Interaction of Non-Nuclear
Munitions with Structures, U.S. Air Force Academy, Colorado, pp. 69-75.

84

Arulanandan, K., Basu, R., and Scharlin, R. J. (1973). Significance of the Magnitude of
Dielectric Dispersion in Soil Technology, Highway Research Board, Highway Research
Record, No. 426, pp. 23-32.
Arulanandan, K., and Mitra, S.K. (1970). Soil characterization by use of electrical network,
Proceedings of the 4th Asilomar Conference on Circuits and Systems, Nov., pp. 480-485.
Arulanandan, K., and Smith, S. S. (1973). Electrical Dispersion in Relation to Soil Structure,
Journal of the Soil Mechanics and Foundations Division, ASCE, Vol. 99, No. SM12, December,
pp. 1113-1131.
Arulanandan, K., Yogachandran, C., and Rashidi, H. (1994). Dielectric Conductivity Methods
of Soil Characterization, Geophysical Characterization of Sites, Volume Prepared by ISSMFE
Technical Committee #10 for the XIII International Conference on Soil Mechanics and
Foundation Engineering, New Delhi, India, pp. 81-90.
Arulanandan, K., and Yogachandran, C. (2000). Dielectric Method for Non-Destructive
Characterization of Soil Composition, GeoDenver 2000, ASCE Geo-Institute, ASCE Special
Publication.
Basu, R., and Arulanandan, K. (1973). A New Approach for the Identification of Swell
Potential of Soils, Proceedings of the Third International Conference on Expansive Soils, Haifa,
Israel, Vol. 1, pp. 1-11.
Bolt, G. H. (1956). Physico-Chemical Analysis of the Compressibility of Pure Clays,
Geotechnique, London, England, Vol. 6, pp. 86-93.
Carman, P. C. (1956) Flow of Gases Through Porous Media, Academic Press, New York.
Cremers, A., Van Loon, J., and Laudelout, H. (1966). "Geometry Effects for Specific Electrical
Conductance in Clays and Soil." Fourteenth National Conference on Clays and Clay Minerals,
pp. 149-162.
Fernando, M. J., Burau, R. G., and Arulanandan, K. (1977). A New Approach to Determination
of Cation Exchange Capacity, Soil Science Society of America Journal, Vol. 41, No. 4, JulyAugust, pp. 818-820.
Fricke, H. (1924). A Mathematical Treatment of the Electrical Conductivity and Capacity of
Dielectric Dispersive Systems. Physical Review, Vol. 24, pp. 575-587.
Fricke, H. and Curtis, H.J. (1937). The Dielectric Properties of Water-Dielectric Interphases.
Journal of Physical Chemistry, Vol. 41, pp. 729-745.
Komornik, A., and David, D. (1969). Prediction of Swelling Pressure of Clays, Journal of the
Soil Mechanics and Foundations Division, ASCE, Vol. 95, No. SM1, pp. 209-225.

85

Kozeny, J. (1927). Ueber Kapillare Leitung des Wassers im Boden, Wien, Akad. Wiss., Vol.
136, Part 2a, p. 271.
Lambe, T.W. and Whitman (1969). Soil Mechanics Wiley, New York.
Meegoda (1983) Prediction of In Situ Stress State using Electrical Method Thesis submitted to
the University of California, Davis, in partial fulfillment of the requirements for the degree of
Master of Science in Engineering.
Michaels, A. S. and Lin, C. S. (1954) The Permeability of Kaolinite, Industrial and
Engineering Chemistry, Vol. 46, pp. 1239-1246.
Mitchell, J.K. (1960). "The Application of Colloidal Theory to the Compressibility of Clays."
Proceedings: Seminar on Interparticle Forces in Clay-Water-Electrolyte Systems,
Commonwealth Scientific and Industrial Research Organization, pp. 292-297
Muraleetharan, K. K., and Arulanandan, K. (1986). Site Characterization in Foundation
Engineering with Reference to Compressibility and Swell, Asian Regional Symposium on
Geotechnical Problems and Practices in Foundation Engineering, Colombo, Sri Lanka, February
25-27.
Olsen, H. W. (1961). Hydraulic Flow Through Saturated Clays, Thesis presented to
Massachusetts Institute of Technology at Cambridge, Mass., in partial fulfillment of the
requirements for the degree of Doctor of Science.
Olson, R. E. and Mesri, G. (1970). Mechanisms Controlling Compressibility of Clays, Journal
of the Soil Mechanics and Foundations Division, ASCE, SM6, pp. 1863-1878.
Olson, R.E., and Mitronovas, F. (1962). "Shear Strength and Consolidation Characteristics of
Calcium and Magnesium Illite." Proceedings of the 9th National Conference on Clays and Clay
Minerals, pp. 185-209.
Quigley, R. M., and Thompson, C. D. (1966). The Fabric of Anisotropically Consolidated
Sensitive Marine Clay, Canadian Geotechnical Journal, Vol. 3, No. 2, pp. 61-73.
Schwan, H. P., Schwarz, G., Maczuk, J., and Pauly, H. (1962). On the Low Frequency
Dielectric Dispersion of Colloidal Particles in Electrolyte Solutions, Vol. 60, pp. 2626-2642.
Seed, H. B., Woodward, R. J. and Lundgren, R. (1962). Prediction of Swelling Potential for
Compacted Clays, Journal of the Soil Mechanics and Foundations Division, ASCE, Vol. 80,
No. SM3, pp. 53-87.
Smith, S. S. (1971). Soil Characterization by Radio Frequency Electrical Dispersion,
Dissertation presented to the University of California, Davis, in partial fulfillment of the
requirements for the degree of Doctor of Philosophy in Engineering.

86

Smith, S. S., and Arulanandan, K. (1981). Relationship of Electrical Dispersion to Soil


Properties, Journal of Geotechnical Engineering, ASCE, Vol. 107, No. GT5, May. pp. 591-604.
Terzaghi, K. (1929). Technish-Geologische Beschreibung der Bodenbeschoffenheit fr
Bautechnische Zwecke, Chapter IX, Part A, Ingeneiurgeologie, by K. A. Redlich, K. Terzaghi,
and R. Kempe, Julis Springer, Wien and Berlin.
Terzaghi, K. (1931). The Influence of Elasticity and Permeability on the Swelling of Two Phase
Systems, Colloid Chemistry, Vol. III, J. Alexander (ed.), Chemical Catalog Co. Inc., NY, pp.
63-68.
Weiler, R. A., and Chaussidon, J. (1968). Surface Conductivity and Dielectrical Properties of
Montmorillonite Gels, Clays and Clay Minerals, Vol. 16, pp. 147-155, Pergamon Press.

87

Chapter 5
Equipment for Measuring Electrical
Dispersion of Soil
5.1 LABORATORY EQUIPMENT
5.1.1 Boonton Rx Meter
Many electrical dispersion studies of soil have been made using an Rx Meter such
as a Type 250 Boonton Rx meter (Boonton Radio Corporation, Division of HewlettPackard). This meter employs a novel cell design and method of connecting the cell to
the electrical bridge which eliminates the impedance effects in the circuit that arise at
radio frequencies (Sachs and Spiegler, 1964). This procedure eliminates the influence of
the transmission line, the electrodes themselves, and the surroundings of the cell in
general. Note that this product is no longer in production (see Figure 5.1).

Figure 5.1
Boonton RX meter
The main features of the cell and its connections are: (1) the length of the
transmission line is the same in both positions; and (2) the geometry of the line is similar

88

at two different sample heights, as shown in Figure 5.2. Measuring sample resistance and
capacitance at two different sample heights lets us essentially subtract out the
impedances of the transmission line and cell surroundings. The length of sample, for the
purpose of calculating conductivity and dielectric constant, is the length of soil removed
(or the difference in sample heights). Figures 5.2 and 5.3 show the cell and the principle
of measuring at two different electrode lengths.

Figure 5.2
Radio Bridge Transmission Line at Two Different Heights

Figure 5.3
The Radio Bridge Sample Cell
Sachs and Spiegler (1964) evaluated the measured impedances by means of the
equivalent circuits shown in Figure 5.4.

89

Figure 5.4
The Circuits Representing the Sample at Two Different Sample Heights
(a and b - longer length sample; c and d - shorter length sample)
From the circuits in Figure 5.4, we may derive the following equations for the true
resistance, R, and capacitance, C, of the sample.

R=

R p 2 Cp Ct
R 'p C'p C t

2
1 + R 2 2 C C 2
p
t
p
1 + R 'p 2 C'p C t 2

1 + R 22 C C 2 1 + R ' 22 C ' C 2 l2
p
p
t
p
t
p

)]

)]

2

2
2 2
' 2 2 '
'
R p 1 + R p C p C t R p 1 + R p C p C t (l 1)

)]

2

2 2
2
' 2 2 '
'
R p 1 + R p C p C t R p 1 + R p C p C t l

+
2
2

2 2
2
'
2 '
1 + R p C p C t 1 + R p C p C t (l 1)

)]

(1)

)]

2
2
2

1 + R p 2 2 C p C t 2 1 + R 'p 2 C'p C t (l 1)
1

C=
2 2
2
2
R2 R p 1 + R 'p 2 C 'p C t R 'p 1 + R p 2 2 C p C t 2 l R

)]

(2)

Where, L1 = longer sample length; L2 = shorter sample length; l = L1 L2 , L1 > L2 ;


C t = stray capacitance of the system; R p = resistance of the sample at L1 ;

C p = capacitance of the sample at L1 ; R p = resistance of the sample at L2 ;

C p = capacitance of the sample at L2 ; = 2f (rad/s); and f = frequency in hertz.


5.1.2 Boonton 250 A RX Meter Example Problem
The following example problem using data from Sachs and Spiegler (1964)
demonstrates the use of the Rx meter for determining dielectric constants and
conductivities. The selected frequencies ( f ) are: 20, 25, 35, 40, 50, and 65 MHz.
90

Convert the frequencies to hertz by multiplying by 106, and then calculate the angular
frequencies using = 2f :

Therefore, the selected angular frequencies become:


Mhz
20
25
35
40
50
65

(rad/s)
125663706
157079633
219911486
251327412
314159265
408407045

L1 and L2 are the distances between the electrodes at two different sample heights: L1 =
5.00 cm and L2 = 2.90 cm.
L1
l=
= 1.724
L2
Measure the stray capacitance, Ct , of the cell with no sample included (air only between
the electrodes).

C t = 4.24 1012 farad


Measure the resistance RP and the capacitance CP of the sample at electrode distance L1
at all the chosen frequencies. Convert CP to farad units for subsequent calculations:
Mhz
20
25
35
40
50
65

RP (ohm) CP (picofarad)
560
15.98
490
14.09
393
11.91
358
11.13
306
9.72
240
8.4

Measure the resistance RP' and the capacitance CP' of the sample at electrode distance L2
at all the chosen frequencies. Convert CP' to farad units for subsequent calculations.
Mhz
20
25
35
40
50
65

RP' (ohm) CP' (picofarad)


333
24.8
280
20.29
220
16.39
197
14.89
158
11.93
120.5
7.75

Use Equations 1 and 2 to calculate the resistance R (ohm) and capacitance (farad) of the
sample:

91

Mhz
20
25
35
40
50
65

R (ohm)
541.7
503.0
417.4
391.1
368.7
303.6

C (picofarad)
11.46
10.59
8.44
7.70
6.25
5.49

Measured resistance and capacitance must be converted to dielectric constant and


conductivity using dimensions of the cell.
area of the electrodes, A = 3.14 cm2
dielectric constant of a vacuum, a = 8.85 1014
The cell constant, Co is calculated as:
Co =

farad
cm

Aa
= 5.5578 1014 farad
L1

The dielectric constants are calculated based on the cell constant as follows.
equation for conductivity is also shown below:
C
=
Co
=
Mhz
20
25
35
40
50
65

206.1
190.4
151.9
138.4
112.4
98.7

The

L1
RA

(mho/cm)
0.00294
0.00317
0.00382
0.00407
0.00432
0.00497

To account for fringe field effects, use a correction factor, . is calculated by placing
water in the cell at a known conductivity. Deviation of dielectric constant from 80 and
deviation of conductivity from the true conductivity is used to calibrate . Typically the
same may be used as approximate for both dielectric constant and conductivity
corrections. corrections are taken into account as follows:

corrected =

corrected =

92

was calculated for the test frequencies as:


Mhz
20
25
35
40
50
65

0.980
0.984
0.989
1.007
1.017
1.072

and the corrected dielectric constants and conductivities are:


corrected
corrected (mho/cm)
Mhz
20
210.3
0.00300
25
193.5
0.00322
35
153.5
0.00386
40
137.5
0.00404
50
110.5
0.00425
65
92.1
0.00464
Sachs and Speigler (1964) have optimized these results using the Three-Element Model.
Their results are shown in Figures 5.5 and 5.6.

dielectric constant,

0.00400
150.0
0.00300
100.0
0.00200

50.0

0.00100

0.0

conductivity,

0.00500

200.0

(mho/cm)

0.00600

250.0

0.00000
100

10

Frequency, MHz

Figure 5.5
Dielectric Constant and Conductivity Dispersion Curves
For Ion Exhange Resin Example

93

dielectric constant, e

0.006

500

0.005

400

0.004
three element model

300

0.003

200

0.002

100

0.001

0
1

conductivity, s (mho/cm)

Dielectric and Conductivity Dispersion Curve

600

0
100

10

Freqency, MHz

Three-Element Model Parameters


a

er

es

0.65

0.02

0.33

0.9

38

78

kr
ks
(mho/cm) (mho/cm)
0.0075 0.000001

Figure 5.6
Three-Element Model Optimization for the Ion Exhange Resin Example
5.1.3 Hewlett Packard Impedance Analyzer

The HP4191A and the subsequent, newer models of Hewlett Packard and Agilent
Technologies impedance analyzers are useful for determining dielectric constants and
conductivities of soil samples at many radio frequencies. The HP4191A has been used
for studying the electrical dispersion of soils by Anandarajah (1982), Meegoda (1983,
1985), Broughton (2001), and others.
To successfully use impedance analyzers (or other radio frequency bridges) for
measuring the dielectric constants and conductivities of soil samples, it is critical to
compensate for, or eliminate, the electrical effects of the lead lines and the soil cell. One
way to eliminate these effects is to measure resistance and capacitance using two separate
electrode distances but with a constant lead length. This method was described in the
section regarding the use of the Boonton RX meter. Under circumstances where the soil
cell design makes varying the sample height difficult, one must electrically compensate
for cell and transmission line losses. The following section demonstrates the procedures
for making transmission line and cell corrections when using an impedance analyzer for
electrical dispersion measurements (Broughton, 2001). In this example, the cell is a
self-built, non-commercial, parallel electrode device of constant lead length.
For use and calibration, the testing setup includes: 1) the Hewlett Packard
HP4191A RF Impedance Analyzer, 2) a binding post fixture 16093A with a coaxial cable
attached to the posts and a BNC connection at the free end of the cable, and 3) a parallel
94

electrode soil cell, with a BNC connection to attach to the transmission line. On the cell
end of the BNC connection, the line should be split. The inner line should attach to one
electrode, and the outer line to the other. The soil cell should have a porous stone and
drainage system at the bottom for consolidating the sample and collecting pore fluid for
conductivity measurement.
All impedance readings should be made in R-X mode. The DC bias switch on the
front of the fixture table should be turned off, and the electrical length of the fixture
should be entered as 0.34 cm for the 16093A test fixture.
5.1.4 Autocalibration of the HP4191A RF Impedance Analyzer

Before operating and calibrating the impedance analyzer, autocalibration should


be performed without the test fixture. To calibrate, use the following four terminations:
the zero reference termination, the open reference termination OS, the 50
reference termination, and the reference termination (male/male) couple.
a) It is a good idea to set the calibration frequency limits initially in the manner
specified by the operating manual. This is done in the following way: Press the
START FREQ key, then enter the initial frequency, and then press ENTER.
Press the STOP FREQUENCY key, then enter the final frequency, then press
ENTER.
b) Press the CALIBRATION button. Connect the zero reference termination to
the unknown tray using the reference termination couple. Press the START
button and wait for the impedance analyzer to go through the frequency cycle.
When finished, the display may read, Conn OS.
c) Remove the zero reference termination and install the open reference
termination OS. Press the START button and wait for the impedance analyzer
to go through the frequency cycle. When finished, the display may read, Conn
50 .
d) Remove the open reference termination OS and install the 50 reference
termination. Press the START button and wait for the impedance analyzer to go
through the frequency cycle. When finished, the display may read Cal End.
Remove the 50 reference termination. Press the CALIBRATION button.
The selected frequencies to be measured must be selected before the soil cell calibration
begins.
5.1.5 Equivalent Circuits of the Fixture, Transmission Line, and Parallel Electrode
Cell

Calibration of the soil cell entails measuring the open electrodes, the shorted
electrodes, and various water samples with different conductivities. An example
calibration follows this discussion to demonstrate the procedure. The R, X mode of

95

the impedance analyzer is used when taking all measurements. This mode outputs real
and imaginary resistances (impedance). Checks should be made after the calibration
using samples with known conductivities and dielectrics other than water. To short the
electrodes, use the outer shielding mesh of a coaxial cable to keep resistance to a
minimum.
An equivalent circuit for the test fixture, transmission line (coaxial cable), soil
cell, and soil sample is shown in Figure 5.7.

Figure 5.7
Equivalent Circuit for Test Fixture, Transmission Line (Coaxial Cable), Soil Cell,
and Soil Sample

An equivalent circuit, used for the calibration and measurement calculations, is


shown in Figure 5.8. The circuit provides impedances where resistors and inductors or
resistors and capacitors may be combined into single complex (number) impedance. The
equivalent circuit representing the transmission line may identically represent the fixture
circuit; however, for brevity it is left without impedances because the line will be
calibrated (in a different way than the cell) as outlined in the HP4191A Operation and
Service Manual.

Figure 5.8
Equivalent Circuit for Calibration and Measurement Calculations
5.1.6 Choosing the Calibrating Frequencies

The first step is to choose the frequencies. An input matrix mhz contains the
chosen frequencies in MHz. These frequencies are not truly variables once the calibration
is finished, since only these frequencies may be used in subsequent measurements. The
chosen frequencies must be within the frequency limits set during autocalibration of the
HP4191A. All calibration variables are 1-dimensional arrays with as many entries as

96

chosen frequencies, thus, if there are 12 chosen frequencies in the mhz matrix, the Z
matrix shown in Figure 5.8 will also have 12 entries, one for each frequency.
5.1.7 Calibrating the Test Fixture

Z1 and Z2 are the impedances of the fixture as seen in Figure 5.8. These
impedances are calculated as described for the test fixture in the HP4191A Operations
and Service Manual. Using the 10693A binding post text fixture, we obtain the following
equations for Z1 and Z2.
Z 1 = R + (2 (mhz )10 6 x1.8 x10 9 )i
1
Z2 =
(2 (mhz )10 6 x1.8 x10 12 )i

(3)
(4)

The resistance value R in Z1 varies with frequency and must be accounted for as
the R matrix. R values of the test fixture are given for different frequencies in the
Operations and Service Manual.
5.1.8 Calibration of the Transmission Line

All readings should be corrected for the impedances of the transmission line. The
next step is to calculate the transmission line impedances. The transmission line may be
calibrated by obtaining impedance readings when the transmission line is open and closed
(shorted) at the BNC connector end. The matrices transo and transs are the
impedance analyzer readings for the open and closed transmission line for the
corresponding calibration frequencies.
The transmission line measurements transo and transs must be corrected for the
text fixture impedances, given Z1 and Z2. The quantities transoa and transsa are the test
fixture-adjusted values of transo and transs given by the following equations.
transo.Z 2 Z 1 .Z 2
Z 2 + Z 1 transo
transs.Z 2 Z 1 .Z 2
transsa =
Z 2 + Z 1 transs

transoa =

(5)
(6)

5.1.9 Calibration of the Test Cell

The cell impedances x, y, and z must be determined to calculate the soil


impedance. Neglecting the test fixture and the transmission line corrections for the
moment, and assuming a short circuit across the electrodes, we obtain the circuit shown
in Figure 5.9. The quantity ccc (closed cell reading) can be calculated using the following
equation.

97

ccc = x +

y ( z + wire )
y + z + wire

(7)

The cell should be shorted with a short length of coaxial cable shielding across the
electrodes. The quantity wire is the impedance of the cable and should be measured
directly across the fixture at the binding posts. The wire impedances are measured
directly at all the desired frequencies and entered as the variable wire. The ccc reading
impedances should be measured for all frequencies. ccc will be corrected for fixture and
transmission line effects later. wire will be corrected for fixture effects only.

Figure 5.9
Equivalent Circuit for Closed Cell Reading

Assuming an open circuit across the electrodes of the cell (Figure 5.10), we obtain the
following equation for cco (open cell reading). cco should be measured for all
frequencies.
cco = x + y

(8)

Figure 5.10
Equivalent Circuit for Open Cell Reading

There are now two equations involving the three unknowns x, y, and z. A third equation
is needed to solve the system of equations. Using a substance of known impedance in the
sample cell can solve this problem. We therefore use water of known conductivity in the
cell. The impedance analyzer registered impedance for the desired frequencies are
entered as the read matrix. Later, the read matrix will be used for the soil samples.
When using water, we obtain the circuit shown in Figure 5.11. The quantity s given in the
following equation is the true impedance of the water.
z( y + s )
(9)
read = x +
y+z+s

98

Figure 5.11
Equivalent Circuit for Sample Cell Reading

Equations (7), (8), and (9) may be manipulated to solve for x, y and z; however, this
system is difficult to solve because the variables are complex numbers. Subtracting x
from both sides of all the equations, the equations become:
y ( z + wire )
y + z + wire
cco x = y
z( y + s )
read x =
y+z+s

ccc x =

(10)
(11)
(12)

Now new variables are introduced. Variables cccx, ccox, and readx are used to represent
ccc, cco and read, each with x subtracted. Variable x will be solved by trial and error
later. In the meantime, the x matrix is initially set at all zeros for all frequencies. The
equations are then easily solved for s, y and z.

y = ccox
(cccx. y + cccx.wire) y.wire
z=
y cccx
readx.( y + z ) y.z
s=
y readx

(13)
(14)
(15)

The readings ccc, cco, and read must be adjusted for the test fixture and the transmission
line, which have been neglected briefly while explaining the previous solution for y and z.
Text fixture corrections using Z1 and Z2 are made for ccc, cco, and read. The corrected
values are ccca, ccoa, and reada. The adjustments are made in the same fashion as the
transmission line readings transs and transo are corrected to yield transsa and transoa in
Equations (5) and (6)..
wire.Z 2 Z 1 .Z 2
Z 2 + Z 1 wire
ccc.Z 2 Z 1 .Z 2
ccca =
Z 2 + Z 1 ccc
wirea =

(16)
(17)

99

cco.Z 2 Z 1 .Z 2
Z 2 + Z 1 cco
read .Z 2 Z 1 .Z 2
reada =
Z 2 + Z 1 read

ccoa =

(18)
(19)

Transmission line corrections must now be made for ccca, ccoa, and reada.
These corrections are made as directed in the impedance analyzer Operation and Service
Manual (Appendix II, Pages 3-B and 3-C). The new transmission-line-corrected values
are calculated and named cccat, ccoat, and readat and are given by the following
equations. The transmission line corrections are made using two key variables zo and
tanh gl, also given below.
zo = transoa transsa
transsa
transoa
zo. tanh gl ccca
cccat = zo.
ccca. tanh gl zo
zo. tanh gl ccoa
ccoat = zo.
ccoa. tanh gl zo
zo. tanh gl reada
readat = zo.
reada. tanh gl zo
tanh gl =

(20)
(21)
(22)
(23)

(24)

To accurately solve for y and z, use the adjusted cell impedances cccat, ccoat, and
readat should be used in place of ccc, cco, and read, respectively. Variables y and z are
then solved in terms of cccatx, ccoatx, and readatx, which are the x-subtracted values of
cccat, ccoat, and readat.
Once all circuit variables are known, we can solve for s, the impedance of the
sample. Variable s is a complex number representing a resistor and capacitor in parallel.
The equivalent circuit is then:
1 1
(25)
= + jC
s r
where r is the resistance of the sample, C is the capacitance of the sample, j = 1 , and
= 2 (mhz ) . Once r and C are known, the sample conductivity and dielectric constant
may be calculated using proper inputs, including length between electrodes and electrode
area. The conductivity and dielectric constant are calculated as follows:

L
RA

(26)

CL
A a

(27)

100

where = the conductivity of the sample, mho/cm.


= the dielectric constant of the sample
C = the capacitance, farad
R = the resistance, ohm
L = the distance between electrodes, cm
A = the electrode area, cm2
a = the dielectric constant of a vacuum = 8.85(10-14) farad/cm
However, we have not solved for the true conductivity and dielectric constant of
the sample (water at this point), because we have not yet solved for x. At the moment, x is
set equal to zero for all the chosen frequencies. Variable x must be adjusted by trial and
error (both the real and the imaginary terms) for each frequency until the dielectric
constant and conductivity match the dielectric constant and the measured conductivity of
water. Water should have a constant dielectric constant (80) and constant conductivity
for all frequencies.
The true conductivity of the water should be measured
independently. Once x has been determined, the same procedure should be repeated at
other water conductivities over the range of values expected for the soil samples. One
can always go back and add a water sample for another higher or lower conductive x if a
soil samples conductivity is too high or too low compared to the calibration range.
Now the calibration is essentially complete. When a soil sample is to be
measured, the R and X readings from the HP4191A impedance analyzer should be
entered as the read matrix. The frequency array mhz must remain the same to represent
the calibrated frequencies. The samples dielectric constant is much more sensitive than
its conductivity to the different x values found at different water conductivities.
Therefore, when the sample conductivity and dielectric constant are calculated, the
calibration water used should be as close as possible to the conductivity of the sample.
Using the sample conductivity, it is possible to develop two dielectric constant and
conductivity curves for the surrounding higher- and lower-conductive water sample
values for x. Using the low-frequency conductivity of the sample and the two water
conductivities, it is possible to interpolate between the dielectric constant and
conductivity dispersion curves to achieve a more precise electrical dispersion, if needed.
5.1.10 HP4191A Example Problem

Step 1: Calibrating the Impedance Analyzer, the Test Fixture, the Transmission Line,
and the Test Cell
Use the HP4191A automatic calibration
Select desired frequency and range (mhz values).
Initially set x equal to zero for all frequencies.
Use the data for transo, transs, wire, cco, and ccc. Note the X values are imaginary
reactive terms of the impedance readings. The zero frequency readings (the first row in
each array) are placeholder remnants of the MATHCAD program from which the data
was taken, and are basically ignored.

101

mhz

R (ohm)

0
1.4
2
3
5
7
10
15
21
30
41
55
71
80
100

0
0.0005
0.001
0.002
0.003
0.004
0.005
0.007
0.008
0.012
0.013
0.015
0.017
0.018
0.02

R (ohm)
0
-150
-70
-20
-9
-3.5
-3
0
-0.1
0.2
0.25
0.3
0.35
0.36
0.4

wire

mhz
0
1.4
2
3
5
7
10
15
21
30
41
55
71
80
100

transo

R (ohm)
0
0.023
0.023
0.03
0.035
0.041
0.045
0.053
0.063
0.075
0.088
0.102
0.116
0.127
0.15

transs
X (ohm)
0
-7300
-5160
-3450
-2077
-1484
-1037
-691
-490.8
-339.9
-244
-175.79
-128.96
-110.06
-78.51

R (ohm)
0
0.063
0.075
0.095
0.123
0.147
0.175
0.216
0.261
0.319
0.387
0.4775
0.585
0.66
0.85

ccc
X (ohm)
0
0.179
0.237
0.331
0.543
0.761
1.073
1.595
2.224
3.165
4.312
5.778
7.452
8.339
10.51

R (ohm)
0
0.076
0.094
0.123
0.165
0.201
0.245
0.315
0.397
0.54
0.78
1.31
3
7
480

X (ohm)
0
0.765
1.062
1.549
2.539
3.529
5.007
7.466
10.424
14.895
20.45
27.75
36.53
41.74
54.31
cco

X (ohm)
0
1.879
2.646
3.917
6.479
9.047
12.93
19.525
27.79
41.31
60.96
95.74
170.91
267.4
-1940

R (ohm)
0
-50
0
10
5
4
3
3.2
1.35
1
0.96
0.8
0.72
0.77
0.87

X (ohm)
0
-5300
-3750
-2520
-1519
-1086
-759
-505.1
-359
-247
-175.54
-123.95
-87.95
-73.08
-47.76

Use tap water (conductivity = 0.518 mmho/cm) for initial calibration. Measure the tap
water impedance (R ohms and X ohms) for each frequency, and enter into the read
variable as shown below.

102

read
MHz
0
1.4
2
3
5
7
10
15
21
30
41
55
71
80
100

r
0
428.6
412.3
377.6
298.5
228
152.4
85.9
50.17
28
17.52
12.33
10.724
11.15
18.78

x
0
-84.1
-114.8
-156.5
-204.3
-215.4
-201.5
-161.1
-121.3
-80.3
-47.53
-17.73
11.35
28.56
83.06

Now use Equations (3) and (4) to obtain Z1 and Z2. Next make the text fixture corrections
for the measured variables transo, transs, wire, ccc, cco, and read using Equations (5),
(6), and (16)-(19). Next calculate the transmission line corrected values for ccca, ccoa,
and reada for all frequencies using Equations (20)-(24). Next calculate the x-subtracted
values for cccat, ccoat, and readat using the following equations. Note at this point,, for
x, resistive R and reactive X terms are set equal to zero for all frequencies. Also note that
s is the actual impedance of the sample in the test cell.
y (z + wirea )
y + z + wirea
ccoatx = ccoat x = y
z( y + s )
readatx = readat x =
y+z+s

cccatx = cccat x =

(28)
(29)
(30)

Now solve for y, z, and s, in that order, as follows:


y = ccoatx
(cccatx. y + cccatx.wirea ) y.wirea
z=
y cccatx
readatx.( y + z ) y.z
s=
y readatx

(31)
(32)
(33)

Knowing s, we can calculate the dielectric constants and conductivities of our water
sample at each of the measured frequencies using Equations (25)-(27). The calculated
values are given below. In these calculations a value of 9.2625 cm2 was used for the area
of the electrodes (A) as and the dielectric constant of the vacuum was set at 8.85 x 10-14
farad/cm.
103


NA
127.3
127.1
126.9
126.8
126.4
126.5
126.3
126.4
127.5
129.7
134.1
142.3
149.2
176.5

mhz
0
1.4
2
3
5
7
10
15
21
30
41
55
71
80
100

(mmho/cm)
NA
0.854
0.854
0.854
0.855
0.858
0.860
0.866
0.872
0.884
0.908
0.949
1.033
1.111
1.527

Note that the dielectric constant and conductivity, respectively, are not equal to the
expected values of 80 and 0.518 mmho/cm. We must still solve for x to complete the
calibration. Alter the real (R) and imaginary (X) terms of x until the resulting dielectric
constants and conductivities are 80 and 0.517 mmho/cm for all frequencies. This can be
done by trial and error, but after a few frequencies are completed, plotting R and X versus
frequency can aid in determining R and X values for subsequent frequencies. The
resulting x and sample readings are given below.

mhz

0
1.4
2
3
5
7
10
15
21
30
41
55
71
80
100

R (ohm)

X (ohm)

NA
-77.63
-105.80
-42.71
-2.70
16.45
13.81
14.93
13.90
7.54
3.70
1.67
0.91
0.50
1.05

NA
5479.3
3885.2
2647.5
1596.0
1143.4
793.7
525.4
374.8
260.3
193.3
149.0
123.1
114.3
106.0

water sample

(mmho/cm)
NA
NA
79.1
0.518
79.1
0.518
79.1
0.518
79.1
0.518
79.0
0.517
79.0
0.518
79.0
0.516
78.8
0.515
78.9
0.516
78.8
0.516
78.7
0.514
78.5
0.511
78.3
0.509
77.9
0.503

104

The values of x changes slightly with different sample conductivities. Therefore it is


beneficial to calibrate x using a range of water conductivities which are likely to be
encountered. A range of x values are shown below for conductivities ranging from 1.023
to 4.07 mmho/cm.
1.023
(mmho/cm)
MHz
0
1.4
2
3
5
7
10
15
21
30
41
55
71
80
100

r
0
-100.981
-122.353
-64.3907
-22.8181
-5.03689
-0.04354
6.248193
11.55807
7.487272
3.936412
2.043891
1.284339
0.922129
1.954983

x
0
5427.745
3850
2623
1593
1143.592
801.9285
536.3008
384.7344
266.5719
196.8831
150.8854
124.1292
115.1041
106.3844

2.10
(mmho/cm)
r
0
-122.081
-155.378
-85.327
-40.6249
-21.4822
-14.2653
-7.56388
1.850995
0.316321
-0.40596
-0.33262
0.688203
1.143266
4.581347

x
0
5841.939
4143.508
2827.552
1715.24
1231.729
865.2386
579.9084
419.6082
292.0062
214.4
160.4149
131.2158
120.6966
109.9255

3.02
(mmho/cm)
r
0
-140.136
-166.529
-93.1435
-45.432
-26.0896
-17.9498
-11.4037
-1.96793
-4.39382
-4.07537
-2.40844
-0.21745
1.444852
6.38364

x
0
5952.003
4223.887
2883.89
1748.461
1255.615
882.4597
591.5876
429.2124
299.139
219.9355
166.1044
133.8379
122.3232
110.2987

4.07
(mmho/cm)
r
0
-149.666
-176.251
-97.4894
-49.9551
-29.534
-20.5029
-15.2276
-5.22832
-9.08661
-8.70252
-5.62203
-2.17945
-0.23126
6.344455

x
0
6061.132
4302.076
2936.177
1781.464
1279.022
898.6246
602.8984
437.509
305.2653
224.5197
169.602
136.4989
124.8382
111.1607

Step 2: Determine Sample Dielectric Constants and Conductivities

Using Yolo Loam soil in the sample cell, the following values for R and X are
determined:
MHz
0
1.4
2
3
5
7
10
15
21
30
41
55
71
80
100

r
0
352.4
341.5
323.2
284.9
247.9
199.8
142.8
102.25
69.23
49.72
38.12
32.95
32.27
35.96

x
0
-56.6
-72
-93.9
-124.8
-141.9
-151
-143.3
-125.2
-98.07
-71.94
-47.19
-25.24
-14.43
7.76

Using x data from the 0.518 mmho/cm conductivity (water), we obtain the following
values for the dielectric constant and conductivity of the sample over the frequency
range:

105

mhz
0
1.4
2
3
5
7
10
15
21
30
41
55
71
80
100

NA
81.9
73.6
66.3
59.0
55.5
51.8
48.3
45.0
41.8
39.2
36.8
35.0
34.2
32.5

(mmho/cm)
NA
0.638
0.645
0.655
0.672
0.686
0.709
0.744
0.785
0.845
0.909
0.980
1.050
1.083
1.126

Noting the low-frequency sample conductivity (0.638 mmho/cm), by choosing the x


values from 0.518 mmho/cm water, we have chosen the closest calibration conductivity
(the next closest calibration water conductivity is 1.023 mmho/cm). However, to be even
more accurate, we may use the low frequency conductivity 0.638 mmho/cm to interpolate
between 0.518 and 1.023 mmho/cm. Using 0.518 mmho/cm, we obtained the dielectric
and conductivity data given above. Using 1.023 mmho/cm data for x we obtain the
dielectric and conductivity data given below:

mhz
0
1.4
2
3
5
7
10
15
21
30
41
55
71
80
100

NA
80.6
72.7
65.1
57.7
54.0
50.6
47.3
44.3
41.3
38.9
36.7
34.9
34.1
32.5

(mmho/cm)
NA
0.640
0.648
0.658
0.674
0.688
0.708
0.740
0.777
0.836
0.901
0.973
1.043
1.076
1.114

106

Using 0.638 as the chosen value of conductivity between 0.518 and 1.023 and
interpolating between the two sets of dielectric constant and conductivity data, we obtain
the final dielectric constant and conductivity values for the Yolo Loam sample. Note that
in this case, the interpolation using the two x values makes little difference in the final
values of the dielectric constant and conductivity.

90

0.0012

80

0.0011

70

0.001

60

0.0009

50

0.0008

40

0.0007

30
20

0.0006

10

0.0005

0
1

10

(mho/cm)

dielectric constant,

(mmho/cm)
NA
0.638
0.646
0.656
0.673
0.687
0.709
0.743
0.783
0.843
0.907
0.978
1.048
1.082
1.123

conductivity,

NA
81.6
73.4
66.1
58.7
55.1
51.5
48.1
44.8
41.7
39.1
36.8
35.0
34.2
32.5

mhz
0
1.4
2
3
5
7
10
15
21
30
41
55
71
80
100

0.0004
100

frequency, MHz

Figure 5.12
Final Electrical Dispersion Curves, Yolo Loam Example

.
5.1.11 Laboratory Dielectric Probe

107

Based on the relationship of dielectric dispersion to soil mineralogy and certain


engineering behaviors, a laboratory probe was developed which measures dielectric
constants at two frequencies. The laboratory probe operates at 2 MHz and 32 MHz.
These frequencies are separate enough to measure differences in dielectric constants in
soils with appreciable dielectric dispersions. These frequency differences make the
laboratory dielectric probe useful for soil classification. In addition, the magnitude of
dielectric dispersion correlates with swell potential as shown earlier. The laboratory
dielectric probe is shown in Figure 5.13.

Figure 5.13
Geoelectronics Laboratory Dielectric Probe

The soil cell, shown on the right side of Figure 5.13, contains two horizontally placed
electrodes. These electrodes are used to measure dielectric properties across the diameter
of the sample. The concave electrodes serve to contain the electrical field and
concentrate the field in the middle of the sample. This eliminates the sample disturbance
effect on the measurements. The laboratory dielectric probe is automatically operated,
and the dielectric constants are reported on the screen shown on the left.
5.1.12 Example Application of the Laboratory Dielectric Probe

Selected soil samples obtained in a borehole at the Rough and Ready Island site in
Stockton, California were tested at the University of California, Davis laboratory to
obtain the magnitude of dielectric dispersion, 0 . Identical samples were tested at the
Kleinfelder laboratory in Stockton, California to obtain the coefficient of permeability.

108

The permeability predicted using the magnitude of dielectric dispersion 0 compared


reasonably well with the measured permeability:
Table 5.1
Results of Dielectric Constant Measurements Using the Geoelectronics Laboratory
Dielectric Probe, and Calculated and Measured Permeabilities of Stockton Soil
Samples
Depth SSA Porosity Apparent Dielectric Dielectric Dielectric Evaluated Measured
Constant (2MHz) Constant Dispersion UC Davis Klienfelder
2
ft M /g
n
0

0
k (cm/sec) k (cm/sec)
7 59.4 0.47
60
33
27
9
34 0.56
57
36
21
3.71E-07 1.80E-07
14 57.2 0.59
69
43
26
2.93E-07 3.80E-07
22 110 0.38
77
27
50
29 50.6 0.41
52
29
23
34 87.3 0.31
61
22
39

Permeabilities were calculated using the modified Kozeny-Carman equation as described


in Chapter 4, Section 4.10.
5.2 IN SITU MEASUREMENT EQUIPMENT
5.2.1 Where Are We In Soil And Site Characterization Using The Cone
Penetrometer?

Before 1916 it was common engineering practice to classify soils as simply clays,
silts, and sands. In 1916 Terzaghi pointed out the need for distinguishing between the
different types of clays and silts. Terzaghis idea led to the development of soil
mechanics. Mechanical methods such as the CPT are used today to characterize soils in
situ. However, since these methods do not determine the type of silt or clay present, the
resulting site characterization must revert to pre-1916 practice.
Thus, a more fundamental approach to site characterization is required. A
fundamental approach for characterizing soils requires quantifying mineralogy, mineralsolution interface characteristics, and soil fabric in situ, and relating these microstructural characteristics to macro soil properties and behavior.
5.2.3 In-Situ Dielectric Cone Probe

Operating much like the laboratory dielectric probe developed by Geoelectronics


an in-situ dielectric cone probe has also been produced. See Figure 5.14.

109

Figure 5.14
Geoelectronics Ge 400 Dielectric Cone Probe

The dielectric cone probe produced by Geoelectronics quantifies the magnitude of


dielectric dispersion for soil layers as it is pushed into the ground. The tool is useful both
for characterizing soil and for estimating soil properties (such as specific surface area,
porosity, and swell potential) based on the electrical dispersion method.
The probe has the same outer diameter as a standard CPT cone and is pushed into
soil in the same manner as a CPT cone. The probe has two horizontally oriented, diskshaped electrodes (as seen in Figure 5.14 about 5 inches from the probe tip). These
electrodes are designed to measure the conductivity and dielectric constant of soil.
Dielectric constant and conductivity are measured at two frequencies, 2 and 32 MHz.
5.2.3 Example Application of the Dielectric Cone Probe

The dielectric cone probe was used to determine the critical state friction angle of
the soils at a location on Rough and Ready Island in Stockton, California. The dielectric
cone probe measurements and the parameters derived from these measurements are
shown in Table 5.2.

110

Table 5.2
Results of Dielectric Constant Measurements Using the Geoelectronics
Dielectric Cone Probe, and Calculated Specific Surface Area, Slope of the Critical
State Line in p'-q space, M, and Critical State Friction Angle,
Stockton, California Samples
Depth
ft
6
10
12
14
16
18
22

Vr
-4.64
-4.89
-4.81
-4.89
-5.11
-5.05
-4.98

Vc
2MHz
1.6
2.12
1.45
1.83
2.83
2.76
2.29
SSA
M2/g
22
66
46.2
11
17.6
0
17.6

Vr

44
54
62
60
42
39
51

-2.94
-3.62
-2.76
-2.54
-3.42
-3.35
-2.85
M
1.17
1.03
1.11
1.19
1.17
1.21
1.11

Vc
32MHz
5.69
5.37
5.48
5.08
4.76
4.76
4.87

34
24
41
55
34
39
43

10
30
21
5
8
0
8

Deg
29
26
28
30
29
30
28

REFERENCES

Anandarajah (1982) In Situ Prediction of Stress Strain Relationships of Clays using a


Bounding Surface Plasticity Model and Electrical Methods Dissertation presented to the
University of California, Davis, in partial fulfillment of the requirements for the degree of
Doctor of Philosophy in Engineering.
Meegoda (1983) Prediction of In Situ Stress State using Electrical Method Thesis
submitted to the University of California, Davis, in partial fulfillment of the requirements
for the degree of Master of Science in Engineering.
Meegoda (1985) Fundamental Characterization of Soils for the Development of an
Expression for Permeability for Application in In Situ Testing Thesis submitted in
partial satisfaction of the requirements for the degree of Doctor of Philosophy in
Engineering, University of California, Davis.

111

Broughton, G. P. (2001) Radio Frequency Dispersion Characteristics of Organic Soils


Thesis submitted to the University of California, Davis, in partial fulfillment of the
requirements for the degree of Master of Science in Engineering.
HP4191A LF Impedance Analyzer Operation and Service Manual
Sachs, S. B., and Spiegler, K. S. (1964). Radiofrequency Measurements of Porous
Conductive Plugs, Ion-Exchange Resin-Solution System, Journal of Physical Chemistry,
Vol. 68, pp. 1214-1222.

112

Chapter 6
Formation Factor, Soil Structure,
Engineering Properties and Behavior
6.1 INTRODUCTION
The formation factor, F, is the ratio of the conductivity of the fluid that saturates a sand
aggregate to the conductivity of the mixture. The formation factor has been shown to be
dependent on a particles shape, long axis of orientation, contact orientation, void ratio,
cementation, and degree of saturation (Dafalias and Arulanandan, 1979). Thus, it is a function of
the grain (particle shape) and aggregate (porosity and fabric anisotropy) characteristics, which
together define the soil structure.
The formation factor method is an electrical method that can be used to characterize soils
and to index their structure by taking into account the porosity, anisotropy and particle shape.
The electrical measurements required for this method can be carried out in situ, thus overcoming
the difficulty of retrieving undisturbed samples. The formation factor method is especially useful
in determining liquefaction potential of granular soils because both depend on grain and
aggregate characteristics. The mechanical properties relevant to analyzing soil liquefaction have
been correlated with appropriate electrical indices, as discussed in this chapter.
This chapter demonstrates that the average formation factor, F , for a given sand
is only a function of porosity and is independent of the orientation of particles for non-cemented
soils. The average formation factor is related to porosity and to the average shape factor, f , by
the expression F = n f . This chapter will also show that F and f are independent of particle
orientation. Arulanandan and Kutter (1978) have proposed an electrical index, A (the anisotropy
index), quantifying the anisotropy of particles.
The grain and aggregate characteristics of particles and stress conditions govern the
derived properties of non-clay minerals. It has been shown that F is a unique function of
porosity. The anisotropy index, A, quantifies particle orientation, and f indicates particle shape.
Thus F and A may be used to quantify the aggregate property, which is sensitive to sample
disturbance and needs to be measured in situ. Meanwhile, grain property (shape), which is
insensitive to sampling disturbances, can be determined on disturbed samples. Empirical
correlation of these parameters ( F , A, and f ) could be extremely useful in evaluating the
performance of sites with sand deposits during earthquakes. This chapter presents laboratory
correlations between electrical parameters and the following variables: cyclic stress ratio
required to cause liquefaction as a function of number of cycles of loading, shear wave velocity,
shear modulus, compression index, hydraulic conductivity, coefficient of earth pressure at rest,
critical state friction angle, and in situ porosity.

113

6.2 FORMATION FACTOR


The mechanical behavior of soils, as well as properties such as permeability, depend on
grain and aggregate characteristics (including the orientation of particles and particle contacts).
Both natural deposits and soil samples prepared in the laboratory are, as a result of preferred
grain orientation, anisotropic. The electrical method using the formation factor involves in situ
measurement of soil conductivity in two different directions and the conductivity of the pore
solution. The formation factor method of soil characterization is presented in later sections.
This section presents the theoretical development of the formation factor tensor and the
anisotropy index. The formation factor is shown theoretically to depend on the basic structural
features of a sand. Analytical expressions are derived relating the formation factor to porosity
and to properly defined parameters associated with the shape and orientation of particles for
transversely isotropic sands. Because F depends on particle orientation, it assumes different
values if measured along different directions of anisotropic sand samples.
This section also gives an example in which the formation factors and anisotropy index of
Monterey 0 sand are evaluated using the theoretical equations.
6.2.1 Extension of Maxwells Equation
Maxwell (1881) derived an expression for the conductivity of a heterogeneous media
consisting of spherical particles immersed in a solution. Maxwell assumed the particles were in
dilute suspensions, such that the electric field of one particle did not influence the electric field
of another. He obtained the following expression for the conductivity of the medium, , as a
function of the conductivity of the solution, 1, the conductivity of the particle, 2, and the
porosity, n:
21 + 2 2(1 n )(1 2 )
=
1
21 + 2 + (1 n )(1 2 )

(1)

Wagner (1914) extended Maxwells model to incorporate the influence of frequency of


an alternating current field on the electrical response. Since the publication of this work, the
mechanism causing the variation of dielectric constant, , and the conductivity, , with
frequency over the radio frequency range has been called the Maxwell-Wagner effect.
Fricke (1924) presented a more general treatment of suspensions, taking into account
both the effects of particle shape and of lower volume concentrations with all particles oriented
in one direction only. It has been shown recently that Frickes expression also applies to high
volume concentrations.
Dafalias and Arulanandan (1979, 1983) captured the effect of multiple orientations of
particles by using appropriate probability density functions. They developed the following
expressions as functions of porosity, complex conductivities, and shape factors (fV and fH), for
vertical and horizontal formation factors, FV and FH:
FV =

k1 k 2
1 n
fV
= 1+
n
kV k2

(2)

114

FH =

k1 k 2
1 n
fH
= 1+
n
kH k2

(3)

where:
fV =

1 P
P
+
k k1
k k1
1+ 2
Aa 1+ 2
Ab
k1
k1

1 P
1 + P
1

fH =
+
k 2 k1
2 k 2 k1
Aa 1 +
A b
1 + k
k1
1

(4)

(5)

Equations (3), (4), and (5) were originally developed for dilute suspensions and were
later shown to also apply to higher volume concentrations of particles (Dafalias and
Arulanandan, 1979; Arulanandan, 1991). In the above expressions, k1 and k2 are the complex
electrical conductivities of the solution and the particle, respectively. Variables kV and kH are the
complex electrical conductivities of the composite medium in the vertical and horizontal
directions. Complex conductivity is defined as:
k = + ja

(6)

where is the conductivity, is the apparent dielectric constant, a is the dielectric constant of
air, is the angular frequency and j is equal to 1 . is related to the frequency, f, by the
expression:
= 2f

(7)

Shape indices Ab and Aa for a spheroid are defined as:

1 1 R 2
2
R
ln
Ab =
2 1 R 2 2 1 R 2 1 + 1 R 2


+ 1 for 0 R < 1

(8a)

and:
R2

Ab =
tan 1 ( R 2 1) 1 for R > 1

2 R 2 1 R 2 1
1

(8b)

and Aa = 1-2Ab, where the axial ratio, R = b/a. Some values of shape indices for spherical cases
are given: for spherical particles, R = 1 and Ab = 1/3. For laminae shaped particles, R and

115

Ab 0; for infinite cylinders, R = 0 and Ab = 1/2. The orientation factor, P , in equations (4)
and (5) represents multi-directional particle orientations and is defined as:
/2

P = p() cos 2 d

(9)

Here, the probability density function, p( ), characterizes the distribution of the orientation of
semi axis a with respect to the vertical direction for 0 2 (Dafalias and Arulanandan,
1979), such that:
/2

p()d = 1

(10)

An explicit expression for kV can be obtained by rearranging equation (3a) as

kV = k1 + (k 2 k 1 )

(1 n)fV
n + (1 n)fV

(11a)

A similar expression for kH can also be obtained by rearranging equation (3b) as

kH = k1 + (k 2 k 1 )

(1 n)fH
n + (1 n)fH

(11b)

Equations (11a) and (11b) are expressions for complex electrical conductivities kV and kH, in
terms of the complex conductivities of the solution, k1; the particles, k2; porosity, n; the particle
orientation, P ; and the axial ratio, R. When the complex conductivities in equation (11a) are
replaced by the conductivities and dielectric constants in equation (6), the real part of equation
(11a) yields an expression for the variation of vertical conductivity, V, as a function of 1, 1,
2, 2, P , f, R, and the porosity, n. In this expression, 1 and 2 are the solution and the
particle conductivity; 1, and 2 are the dielectric constants of the solution and the particle; and
f is the frequency. The imaginary part of equation (11a) is an expression for the vertical
dielectric constant, V, as a function of 1, 1, 2, 2, P , R, f, and n. Similarly, expressions
for the horizontal conductivity, H, and the horizontal dielectric constant, H, can be obtained
using equations (11b) and (6). The dielectric constant of the solution, 1, is approximately equal
to 80 for water. The dielectric constant of dry clay particles, 2, is about 6.5.
6.2.2 Formation and Shape Factors

The formation factor is defined as the ratio of the conductivity of the electrolyte that
saturates a particulate media to the conductivity of the electrolyte/particulate mixture. The
formation factor is thus a non-dimensional parameter. Experimental studies conducted by
Archie (1942), Arulanandan and Kutter (1978), Arulmoli (1980, 1982), Jackson (1975), and
Wyllie and Gregory (1953), showed the formation factor to depend on particle shape, long axis

116

of orientation, contact orientation, void ratio, cementation, and degree of saturation.


Furthermore, for an anisotropic particulate system, the formation factor has been shown to vary
in different directions. The vertical and horizontal formation factors can be defined in the
following manner:

Fv = sol
v

Horizontal formation factor, FH = sol


H

Vertical formation factor,

(12a)
(12b)

where sol = conductivity of the solution (electrolyte); v = vertical conductivity of the


mixture; and H = horizontal conductivity of the mixture.
Extending the pioneering work of Maxwell (1881) on the electrical conduction through
heterogeneous media for spherical particles, as well as Frickes work (1924) for ellipsoids of
random orientation, Dafalias and Arulanandan (1979) derived equations that take into account
the orientation of particles with an assumption of transverse isotropy. Transverse isotropy is a
common feature for most natural soil deposits and laboratory prepared samples. Transverse
isotropy is characterized by an axis of rotational symmetry, coinciding with the direction of
sedimentation of natural deposits and with the vertical axis of laboratory samples. For most
natural deposits, the vertical axis can be considered the axis of rotational symmetry. The work of
Dafalias and Arulanandan (1979) can be summarized as follows:
Fv = 1+

1n
f
n v

(13)

FH = 1 +

1n
fH
n

(14)

where n = porosity, and


Vertical shape factor,

fv =

2S P (3S 1)
4S(1 S)

Horizontal shape factor, fH =


and, P =

S + 1 + P (3S 1)
4S(1 S)

p()cos2 d

(15)

(16)

(17)

where p( ) is the probability density function that characterizes the distribution of long axes
orientation of the particles with respect to the vertical axis for 0 /2, such that:

p( ) d = 1

(18)

and is the inclination of the particles long axis with the vertical.

117

If a, b, and c represent the semiaxes of an ellipsoid, then for a prolate spheroid (a > b = c)
the factor S, which depends only on the axial ratio (R = b/a) of the particles, is given by:
R2

1 1 R2

S=
ln
+ 1 ,
2
2
2

21 R 2 1 R
1+ 1 R

0R1

(19)

R1

(20)

Also, for an oblate spheroid (a < b = c):


S=

R2
tan 1 R 2 1 1 ,

2 1 R 2 R 2 1

Equations (13) and (14) assume no electric field interaction between particles and are
valid for very dilute suspensions. For a dense suspension, the interaction between particles can
be taken into consideration indirectly by using Bruggemans (1935) integration technique, which
yields the following equations:
Fv = n f v
FH = n fH

(21)
(22)

6.2.3 Formation Factor Tensor

Ohms law can be stated as

E i = ijJ j

(23)
where Ei is the electric field intensity in the direction i; Ji is the electric current density in the
direction j; and ij is the resistivity in the direction i due to a current density in direction j.
In matrix form, equation (23) becomes:
E1 11 12 13 J1
(24)
E 2 = 21 22 23 J2

3 31 32 33 J3
If 1, 2, and 3 represent principal directions of resistivities, the resistivity matrix becomes
a diagonal matrix. In addition, if transverse isotropy is assumed (such as with respect to a
horizontal bedding plane), the principal directions are vertical and horizontal, and the resistivity
matrix becomes:
v 0
0
ij = 0 H 0
(25)
0 0

H
Substituting the following relations in equation (25):
1
v =
v

(26)

118

1
H
and multiplying equation (25) by the conductivity of the solution, sol (a scalar), gives:
0
0
sol v

sol ij = 0
sol H
0

0
sol H
0
H =

(27)

(28)

or, in terms of formation factors:


0
Fv 0

(29)
sol ij = 0 FH 0 = Fij
0
0 FH
The first invariant of the formation factor tensor (Fij), which does not depend on the
orientation of the axes, is given by:
Fii = Fv + 2FH
(30)
Defining an average formation factor, F , as:
F
F = ii
(31)
3
it follows that F is also then independent of the orientation of the axes. Finally, from equations
(13), (14), and (31), it follows that:
1n
F =1+
(f + 2fH )
(32)
3n v
If an average shape factor, f , is defined as:
f + 2fH
f = v
(33)
3
then it follows that:
1n
F =1+
f
(34)
n
and, from equations (5), (6), and (23):
3S + 1
f =
(35)
6S(1 S)

Equation (35) shows that f is only a function of the shape of the particles. However, the
experimental results show that f is also a function of cementation (Wyllie and Gregory, 1953).
From equations (34) and (35), it is evident that f and F are independent of particle
orientations. This is an expected result, since F is shown to be an invariant of the formation
factor tensor. Extending equation (34) for dense suspensions gives:
F = n f
(36)
Hence, the average formation factor, F , is only a function of the porosity and shape of the
particles for uncemented soils.

119

6.2.4 Anisotropy Index

Consider a Cartesian coordinate system in which x and y represent the horizontal axes
and z represents the vertical axis. Assuming transverse isotropy, with the vertical axis as the axis
of rotational symmetry, and x, y, and z as the principal directions of conductivity, Ohms law can
be written as follows:

(37)
J = H
i+
j v
k
y
z
x
where J is the electrical current density; is the electric potential; and i, j, and k represent the
unit vectors in the directions x, y, and z, respectively.
For the case of constant conductivities, continuity of charges leads to the following
expression:
2 2
2

J = 0 = H
+
+ v
(38)
x 2 2 y
z 2

When H = v, i.e., in the isotropic case, equation (38) becomes Laplaces equation.
Introducing the coordinates u = x, v = y, and w = z, where = ( H/ v)1/2, and noting that:
2
z 2

= 2

2
w 2

(39)

Equation (38) becomes:


2 2 2
=0
H
+
+
u 2 v 2 w 2

(40)

That is, in (u, v, w) space, the potential in the anisotropic case satisfies Laplaces equation just as
it does in the (x, y, z) space for the isotropic case. Alternatively, as approaches 1, the
anisotropic case approaches the isotropic case in the limit. Hence, is the measure of anisotropy
for the electrical conduction through a suspended medium (Mousseau and Trump, 1967).
Arulanandan and Kutter (1978) proposed an electrical index, A, quantifying the
anisotropy of particles. It is defined in the following manner:
Fv
(41a)
A=
FH
sol
A=

sol

v
H

H
=
v
Also from equations (15), (16), (21), (22), and (41a):

or

(41b)

A=

A = n

(41c)

(41d)

120

where

(3S 1)(3P 1)
8S(1 S)

(41e)

With the introduction of key electrical indices complete, attention is now turned to predicting
values for these, as well as for related indices.
6.2.5 Predicting Formation Factors and Anisotropy Index

Using thin section studies (Mitchell, et al., 1976), Dafalias and Arulanandan (1979) have
predicted the vertical and horizontal formation factors and the anisotropy index for Monterey 0
Sand. Two pieces of information are necessary for the prediction: the shape factor, S, and the
orientation factor, p . Figure 6.1 shows typical results of these thin section studies, as well as
the predicted p values using equation (7) with known probability density p( ). The p( ) is
given in a discrete form as the percent frequency p of the long axis orientation with respect to
the vertical. This value was obtained using the histograms of Figures 6.1 and 6.2, where the sum
of p from = 0o to = 180o is equal to one.

Figure 6.1
Histograms of Particle Long Axis Orientations for Samples of Monterey 0 Sand Prepared
to 50% Relative Density by Different Methods (after Mitchell, et al., 1976)

121

Figure 6.2
Histograms of Particle Long Axis Orientations for Samples of Monterey 0 Sand Prepared
to 80% Relative Density by Different Methods (after Mitchell, et al., 1976)

Noting that cos2 = cos2( - ), equation (27) can be written in a discrete form as:
P =

= 85 O

2
(p + p ) cos
O

=5

(52)

where summation is carried out for nine 10o angular intervals from 0o to 90o, taking at the
middle of each interval, i.e. = 5o, 15o, ..... , 85o. Table 6.1 tabulates the values of
p +p from the histograms, the computed values of p from equation (32), and the average
azimuthal angles = cos 1 P .

122

Table 6.1
Tabulation of the Percent Frequency of Long Axes Orientation from Histograms of
Figures 6.1 and 6.2 and Calculation of the Orientation Factor and Average
Azimuthal Angle

Based on these thin section studies, Monterey 0 Sand particles were modeled as prolate
spheroids with R = 0.65. For R = 0.65, equation (29) yields S = 0.386. In Tables 6.2 through
6.4, the predicted vertical and horizontal formation factors, and the anisotropy index values
obtained using equations (11), (12), and (31d), respectively, are compared with the values
measured in the laboratory. For the moist tamped sample at n = 0.397, no corresponding
histogram was available, thus p was obtained by interpolating between the p at n = 0.38 and n
= 0.419. Close agreement between measured and predicted values confirms the validity of the
theoretical equations derived earlier.

123

Table 6.2
Prediction of Vertical Formation Factor Using Orientation Factor (P) Calculated from
Thin Section Studies and Comparison with Measured Values

Monterey 0
preparation
method
Dry
Pluviated
Moist tamped

Moist vibrated

Porosity
n
0.380
0.419
0.380
0.397
0.419
0.419

Orientation
factor
P
0.340
0.414
0.438
0.429
0.420
0.434

Vertical
shape
factor
fv
1.515
1.491
1.483
1.486
1.488
1.484

Vertical
formation
factor
Fv = n-fv
4.33
3.66
4.20
3.94
3.64
3.63

Vertical
formation
factor
measured
4.32
3.74
4.22
3.92
3.60
3.47

Percent
error
(%)
0.23
-2.14
-0.47
0.51
-1.11
4.60

Table 6.3
Prediction of Horizontal Formation Factor Using Orientation Factor (P) Calculated from
Thin Section Studies and Comparison with Measured Values

Monterey 0
preparation
method
Pluviated
Moist tamped
Moist vibrated

Porosity
n
0.380
0.397
0.419

Orientation
factor
P
0.340
0.429
0.434

Horizontal
shape
factor
FH
1.518
1.533
1.534

Horizontal
formation
factor
FH = n-fH
4.34
4.12
3.80

Horizontal
formation
factor
measured
4.18
3.97
3.60

Percent
error
(%)
3.83
3.78
5.56

Table 6.4
Prediction of Electrical Anisotropy Index Using Orientation Factor (P) Calculated from
Thin Section Studies and Comparison with Measured Values

Monterey 0
preparation
method
Pluviated
Moist tamped
Moist vibrated

Porosity
n
0.380
0.397
0.419

Orientation
factor
P
0.340
0.429
0.434

0.0017
0.0239
0.0252

Electrical
anisotropy
index
A = n
0.998
0.978
0.978

Electrical
anisotropy
index A
measured
1.016
0.994
0.982

Percent
error
(%)
-1.77
-1.60
-0.41

The agreement is much better for Fv than for Fh. This can be attributed to two factors.
First, the p is computed directly from the vertical, thin-section histograms. Its use to predict Fh
assumes ideal transversely isotropic symmetries that may not be satisfied. Second, the values of
Fh were measured initially in cubical samples. They were then corrected by a geometrical factor
124

to correspond to measurements of Fv in cylindrical samples, which may have introduced some


error for Fh.
6.3 SOIL PROPERTIES USING THE FORMATION FACTOR METHOD

This section describes the use of the formation factor, shape factor, and anisotropy index
for evaluating maximum shear modulus, slope of the virgin compression and swelling lines,
coefficient of earth pressure at rest, K0, hydraulic conductivity, porosity, critical state friction
angle, location of the critical state line, and state parameter.
6.3.1 Porosity

Section 6.2 showed porosity to be theoretically related to the average formation factor,
F . To illustrate this, horizontal and vertical formation factor measurements were made in the
laboratory on lucite balls and Monterey 0/30 sand samples prepared by three different methods.
Two 6-inch cubical cells, one with two 6-inch-square, platinum-coated, copper electrodes fixed
on two opposing vertical faces, and the other with electrodes on top and bottom faces, were used
for making horizontal and vertical electrical resistance measurements, respectively (Arulmoli, et
al., 1985). As Figures 6.3 and 6.4 show, the F vs. porosity relationships obtained for all three
methods of sample preparation were shown to be a function of porosity.

Figure 6.3
Vertical, Horizontal and Average Formation Factor versus Porosity
Relationships for Lucite Balls Prepared by Three Different Methods

125

Figure 6.4
Vertical, Horizontal and Average Formation Factor versus Porosity Relationships
for Monterey 0/30 Sand Prepared by Three Different Methods

Note that the samples were prepared in two separate cubical boxes and vibrated to different
porosities. Electrodes were placed in the horizontal direction for FV measurements in one box,
and in the vertical direction for FH measurements in the other box. As a result, the FV and FH
values plotted in Figures 6.3 and 6.4 are at different porosities. A unique relationship between
F and n was found to exist for five more uniform sands and a silty sand obtained from the
Revelstoke Dam site as shown in Figure 6.5.

Figure 6.5
Average Formation Factor versus Porosity Relationships Using Cubic Cells in the
Laboratory

126

The unique relationship between F and n can be used to evaluate the in situ porosity of
uncemented sands using an in situ method (see Figure 6.17) for measuring the formation factors.
6.3.2 In Situ Determination of Stress Ratio Required to Cause Liquefaction as a Function
of Number of Cycles of Loading and an Electrical Structure Index, A3 / F f

To correlate the cyclic stress ratio required to cause liquefaction in 10 cycles, c10 , with
an electrical parameter, laboratory electrical measurements were made on reconstituted samples
whose liquefaction characteristics have been extensively studied and reported by various
investigators. It has been observed that c10 increases with increasing F and f , and decreases
with increasing A, and further, that the combination of F , f , and A that best correlates with
stress ratio to cause liquefaction is A3 / F f . Table 6.5 summarizes the results, and Figure 6.6
shows the correlation curve between c10 and the electrical parameter, A3 / F f .

Table 6.5
Electrical Measurements and Cyclic Stress Ratios to Cause Initial Liquefaction for Sands
Tested

Sand type
Monterey 0

Method of
preparation
Dry pluviation
Wet pluviation
Moist tamping

Porosity
n

A3

0.416
0.395
0.416
0.416

3.72
3.94
3.72
3.72

0.401
0.406
0.380
0.366
0.433
0.401

3.87
3.44
3.74
3.90
3.20
3.48

A
1.029
1.026

1
F fm
0.196
0.185
0.194
0.174

0.225
0.306
0.250
0328

0.167
0.217
0.200
0.192
0.234
0.215

0.370
0.180
0.232
0.270
0.110
0.180

0.202
0.197
0.183

0.225
0.286
0.348

0.202
0.195
0.191
0.183

0.225
0.270
0.325
0.285

0.172

0.435

10

Investigator
Mulilis,
et al.

0.989
Ottawa C-109

Dry pluviation

1.01

Harder Jr.

Finn, et al.

1.01

Sierra
Diamond
Lawson's
Landing site

Moist tamping

Reid Bedford
site

Moist tamping

Wet pluviation

0.383
0.375
0.463

3.71
3.80
3.33

0.429
0.419
0.412
0.424

3.62
3.74
3.82
3.52

0.404

3.74

0.982
1.032

0.985

Arulmoli
Arulmoli

Townsend,
et al.

127

Figure 6.6
Laboratory Correlation Between Cyclic Stress Ratio Required to Cause Initial
Liquefaction in 10 Cycles and Electrical Parameter, A 3 / (F f )

Similar correlations have been established for the cyclic stress ratios that cause
liquefaction in 5, 10, 15, 30 and 50 cycles (Arulmoli, 1982). These are shown in Figure 6.7.

Figure 6.7
Laboratory Correlations Between Cyclic Stress Ratio to Cause Initial Liquefaction
and Electrical Parameter, A3 / F f

128

Using these correlations, the cyclic stress ratios were plotted against the number of
cycles, N, for various values of the electrical parameter A3 /( F f ) , as shown in Figure 6.8.

Figure 6.8
Correlation Between Cyclic Stress Ratio Required to Cause Initial Liquefaction and
Number of Cycles for Different Values of Electrical Index, A3 / F f

If A3 /( F f ) can be obtained for a sand deposit, then the relationship between the cyclic
stress ratio that causes liquefaction and number of cycles can be established using the results
shown in Figure 6.8.
Figure 6.9 shows the relationship between cyclic stress ratio and number of cycles that
cause initial liquefaction for the Monterey 0 sand prepared by different methods at the same
porosity. In Figure 6.9, dc is the amplitude of deviatoric cyclic stress and '0 is the initial
effective confining pressure. Also given in Figure 6.9 are the measured electrical anisotropy
index (A) values. It shows that for the same cyclic stress ratio, samples with a higher value of A
need fewer cycles to cause initial liquefaction. This implies that, for a given number of cycles
and given cyclic stress ratio, higher anisotropy produces a larger increase in pore-water pressure.
Since the same sand with the same relative density was considered, F and f were constants for
all three cases shown in Figure 6.9. Hence, an increase in A implies an increase in A3 /( F f )
and an increase in pore water pressure.
A decrease in porosity, on the other hand, reduces the amount of settlement occurring for
a given number of cycles and hence reduces the pore-water pressure generated in a given number
of cycles. Also, a decrease in porosity increases F and hence decreases the index A3 /( F f ) .
Concerning the shape of particles, rounded particles are known to be more susceptible to
pore pressure generation due to dynamic loadings than platy ones. Also, rounded particles have
a smaller f than platy particles and hence a larger value of A3 /( F f ) . These observations
qualitatively justify correlating the parameter A3 /( F f ) to properties associated with dynamic

129

responses. The power 3 in the index shows the increased influence of anisotropy on dynamic
properties.

Figure 6.9
Cyclic Stress Ratio versus Number of Cycles to Cause Initial Liquefaction for
Monterey 0 Sand Samples Prepared by Different Methods (after Mulilis, et al., 1977);
Measured Electrical Anisotropy Index Values are after Arulanandan and Kutter (1978)

Figure 6.10
Comparison of Relationship Between Cyclic Stress Ratio to Cause Initial Liquefaction and
Number of Cycles Obtained from Conventional Cyclic Tests and that Predicted from
Electrical Measurements on Undisturbed Sand Samples from the Niigata Earthquake Area

Correlations were performed to examine the validity of the electrical method to predict,
in situ, the relationship between the cyclic stress ratio needed to cause initial liquefaction and the
number of cycles for soils. These correlations compared laboratory cyclic tests on undisturbed
samples of sand from the Niigata earthquake area with predictions using electrical measurements
on the same sample. Undisturbed samples were obtained by freezing the soil. Figure 6.10
compares the results. The close agreement between the predicted and measured values justifies

130

using this methodology to predict, in situ, the relationship between cyclic stress ratio required to
cause initial liquefaction and the number of cycles of loading.
6.3.3 Residual Friction Angle

By measuring the formation factors (vertical and horizontal) and porosity of a sand in the
laboratory, one may obtain the average shape factor.
fV =

log FV
log n

(42)

fH =

log FH
log n

(43)

f + 2f H
f = V
3

(44)

Knowing f , the following correlation provides the slope of the critical state line (M) in
q - p space (Arulanandan, et al., 1994):

Figure 6.11
Relationship Between the Average Shape Factor and the Slope of the Critical State
Line (M) in q p Space

Moreover, from:
M=

(6 sin )

(3 sin )

(45)

One may, therefore, obtain the critical state friction angle, , using Equation (45).

131

6.3.4 Hydraulic Conductivity

Specific surface area of sands can be obtained from grain size distribution curves as those
shown in Figure 6.12 (Arulanandan and Muraleetharan, 1988).

Figure 6.12
Particle-Size Distribution Curves for Sierra Diamond and Monterey 0/30 Sands

The specific surface area, given as surface area per unit volume of soil particles, may be
calculated as follows.
W W W1
100 Wn 1
+ ......... +

6 1 + 2
d
d
d
1
2
n

SSA =
100

(46)

where W1, W2, . . . , etc., are the percentages passing; and di is the average diameter between Wi
and Wi-1.
Once we know the porosity (obtained electrically via the formation factor method) and
the specific surface area as defined above for sands, we may use the Kozeny-Carman equation
to calculate hydraulic conductivity. By solving for hydraulic conductivity and substituting in
known electrical parameters for porosity, we obtain the following electrical index, which is
directly proportional to hydraulic conductivity:
F3 / f
1 F

1 / f

(SSA )

(47)

Figure 6.13 shows the linear relationship between hydraulic conductivity and the electrical index
given in Equation (47) (Arulanandan and Muraleetharan, 1988).

132

Figure 6.13
Correlation between Vertical Hydraulic Conductivity of Sands and Silty Sands and an
Electrical Index
6.3.5 Shear Modulus

One way of predicting maximum shear modulus, Gmax, is by measuring the in situ shear
wave velocity, VS, and using:

G max = VS 2

(47)

in which is the mass density of the deposit at the depth of measurement. Investigations have
shown that the maximum shear modulus values for sands are strongly influenced by the
confining pressure and the void ratio (Hardin and Drenevich, 1970; Seed and Idriss, 1970). Seed
and Idriss (1970) gave a relationship between shear modulus, G, in pounds per square foot and
mean effective confining pressure, ' m , in pounds per square foot, as:
G = 1000 K 2 max (' m )1 2

(49)

In the above equation, K 2max depends largely on the void ratio, the age of the deposit and
in situ stresses (Seed and Idriss, 1970). A correlation between K 2max and an electrical parameter,
F /( A f )1 / 2 , was developed using measurements made in both the laboratory and the field
(Arulanandan, et al., 1982). Field shear-wave velocities were measured using crosshole seismic
methods; the electrical measurements in the field were made using an electrical probe, which is
described below. Figure 6.14 shows the correlation.

133

Figure 6.14
Correlation between K2 max and Electrical Parameter, F /( A f )1 / 2
6.3.6 Compression Index, CC, and Swelling Index, CS

Figure 6.15 correlates the compression index, CC, of sands (i.e., non-clay minerals) and
the electrical index, Af / F , which is a polynomial of degree 2 (Harvey, 1981).

Figure 6.15
Correlation Between Compression Index and Electrical Index Af / F for Sands
(after Harvey, 1981)

134

Since the initial void ratio influences the compression index of a sand, it follows that the average
formation factor ( F ) calculated using the initial porosity should be included in the correlation.
This correlation is valid up to a vertical effective stress of 5,500 lbs/sq ft. An approximate
estimate of the swelling index (CS) may be possible by assuming CS = (1 / 4)CC .
Electrical parameters can only be used to evaluate porosity, cyclic stress ratio required to
cause liquefaction, and shear wave velocity if in situ measurement of F and A is possible. Since
f can be obtained from the F - n relationship, independent of aggregate properties and thus
sample disturbance, it can be measured from laboratory tests of disturbed samples. F and A
depend on aggregate properties and should be measured in situ. Section 6.3.8 describes an
electrical formation factor probe used for measuring electrical parameters in situ.
6.3.7 The Coefficient of Lateral Earth Pressure at Rest, K0

Anisotropic constitutive models (Anandarajah, et al., 1984; Pietruszczak and Mroz, 1983)
show that the coefficient of lateral earth pressure at rest, K0, is a function of fabric anisotropy and
fabric friction. Since clay structure is viewed using the cluster concept, it is appropriate to use the
term fabric friction. It is impossible for frictionless material to develop any fabric anisotropy.
On the other hand, if there is no fabric anisotropy, (i.e., the fabric is isotropic), any isotropic
constitutive model should show that K0 is a function of the angle of internal friction, ' , which is
a measure of fabric friction. The anisotropy index, A, is an indirect measure of fabric anisotropy.
Anandarajah, et al., 1982, and Arulanandan, et al., 1982, quantified fabric friction in terms of
electrical parameters. Since K0 is a function of fabric anisotropy and fabric friction, it is possible
to develop a correlation between K0 and electrical parameters as shown in Figure 6.16.

Figure 6.16
The Relationship Between the Coefficient of Lateral Earth Pressure at Rest and the
Electrical Index, A4 f (Meegoda and Arulanandan, 1986)

Figure 6.16 is derived for clay soils. Figure 6.16 has been used successfully for sands
when obtaining K0 for predicting earthquake simulation of liquefaction-induced deformation.

135

However, further work is needed to show the correlation between K0 and the electrical
parameters for sands.
In developing Figure 6.16, when there is no friction or anisotropy, K0 is assumed as 1,
and the lower limit of K0 for fine-grained soils is assumed as 0.38. The higher power of A in the
electrical index A4 f demonstrates the heavy dependence of fabric anisotropy on K0.
6.3.8 Critical Void Ratio versus Mean Normal Stress

Relationships between critical void ratio, eC, and effective mean normal stress, p', for
Sacramento and Nevada sands were established based on triaxial test data and the average shape
factor, f , obtained using formation factor tests. Figure 6.21 shows the interpolated eC versus p'
relationships based on f variation for Niigata and Toyoura sands (Arulanandan, 1995).

Figure 6.21
Critical Void Ratio, ec, vs, Effective Mean Normal Stress, p', for Different Average Shape
Factors, f
6.4 IN SITU FORMATION FACTOR PROBE

The electrical formation factor probe shown in Figure 6.17 can be used to make electrical
measurements of soils in situ (Arulanandan, 1977).

136

Figure 6.17
Geoelectronics Electrical Formation Factor Probe with Control Unit and Cable

The probe is capable of making measurements at any depth below the water table. The
probe consists of three main parts.
1) The Main Body

A 4-foot-long, 3-inch-outside-diameter, steel tube houses electronics for the electrical


bridge. The tube also houses a vacuum-actuated sampler, which has electrodes for solution
conductivity measurements.
2) Probe Tip

A 12- to 18-inch-long replaceable steel tube (3 inch outside diameter and 1/16 inch thick)
is screw fitted to the above. The probe tip carries the electrodes for horizontal and vertical
measurements. Figure 6.18 shows a schematic view of the electrodes.

137

Figure 6.18
Schematic Views of In Situ Formation Factor Electrodes

The probe tip also carries a porous stone on its outside, about 4 inches from the end,
which is attached to one end of a thin metal tube. The other end of this tube is connected to the
water sampler through the pump.
3) Control Unit

A microprocessor control unit for transmitting the electrical signal and receiving the
measured electrical properties is connected to the electronics in the main body through a stiff
cable.
6.4.1 Operation

The probe tip is connected to the main body, and the assembly is dropped into a
previously bored hole to the desired depth. The tip is pushed about 8 to 9 inches into the soil
below the end of the borehole so that measurements can be made in the undisturbed region of
soil. The area ratio is 8.9%, well below the minimum Hvorslev (1949) recommends to achieve
good results. The setup is operated with a 12 V DC power source. The solution is pumped from
the saturated soil into the water sampler. The microprocessor varies the bridge impedance in
steps. The horizontal, vertical, and solution resistances are obtained to an accuracy of about 1%.
These results are used together with calibration curves for the vertical and horizontal electrodes
and the water sampler, obtained from laboratory measurements. Vertical measurements are
made at an angle. By constructing a Mohrs circle of formation factors (similar to a Mohrs circle
of stresses), the true vertical formation factor is obtained. The applicability of the electrical
probe was checked in the laboratory and in the field and is reviewed below in detail.
6.4.2 Verification of the Electrical Probe Using Chamber Test

To check the validity of the electrical probe in predicting the void and stress ratios
required to cause liquefaction, controlled laboratory tests were performed on samples with and
without confining pressure (Arulmoli et al., 1985).

138

Monterey 0/30 sand was pluviated through water into a cylindrical bucket, and
formation factor measurements were made with the probe.

Figure 6.19
Verification of Formation Factor Test in Chamber under Confining Pressure

The sample was then densified by vibration, and measurements were repeated. Average
formation factor values were obtained from these measurements using the following equation:
F=

(FV + 2FH )
3

These values were compared with the F - n relationship already obtained in the
laboratory. Figure 6.20 shows the results. The probe measurements appear to be within
acceptable accuracy range for sands. While measuring with the probe, due to the electrode
configuration, most of the electric potential drop occurs in the central portion of the sample,
where the disturbance due to pushing the probe is expected to be low. This may explain why the
predictions do not show a substantial variation in porosities.

139

Figure 6.20
Comparison of Laboratory Average Formation Factor Measurements Made in Cubic Cells
with Those Made with the Formation Factor Probe in Laboratory Chamber Tests
6.5 STATE-PARAMETER-DEPENDENT SAND MODELS
6.5.1 Introduction

Attempts have been made in recent years to take into account the state dependent
response of sand. Manzari and Dafalias (1997) introduced a sand model that defines a linear
dependence of phase transformation stress ratio on (Figure 6.22), such that when = 0 , the
phase transformation stress ratio becomes identical to the critical stress ratio. The model
guarantees satisfaction of the basic premises of critical state soil mechanics, and allows modeling
of both loose and dense sand behavior with a unique set of parameters. Li (1997) investigated
the response of sand at the ultimate stress ratio. He pointed out that the dilatancy is related not
only to the stress ratio, but also to the plastic volumetric strain. To model dilative hardening
correctly, care was taken to distinguish the ultimate stress ratio from the ultimate material state
(critical state), at which the dilatancy is equal to zero. The dependence of dilatancy on density
(proposed by Li for the ultimate stress ratio), is generalized to all stress ratios by using the
technique proposed by Manzari and Dafalias (Li, et al., 1999).
The loose and dense states of sand are distinguished in reference to the critical or steady
state. In this state, the effective mean normal stress p', the deviatoric q, and the volumetric strain
v , are all constants, while the deviatoric strain q continuously develops. Been and Jefferies
(1985) introduced a parameter, , called the state parameter, which measures the difference
between the current and critical void ratios at the same confining pressure. Figure 6.22 illustrates
the definition of the state parameter. When is positive, the material is in a loose state (looser
than the critical state). When is negative, the material is in a dense state (denser than the
critical state). When the critical state is reached, = 0 . Li and Dafalias (2000) identified

140

dilatancy, d, as the key variable that uniquely relates the contractive and dilative responses to
stress and material state: d = d vp d qp , where d vp and d qp are the increments of the plastic
volumetric and deviatoric strain, respectively.

Figure 6.22
Critical State Line and State Parameter,
6.5.2 In-Situ Determination of the Critical State Line and the State Parameter

To determine the state parameter, , we need to define a critical state line in the e- p
plane. It has been shown experimentally (e.g., Verdugo and Ishihara, 1996) that a straight line in
the e- p space can approximate the critical state for sand if a wide range of pressure is
considered. To describe this line, Li (1997) adapted the following relationship, which was
originally proposed by Wang, et al. (1990) for the virgin compression line of sand:
p
e C = e c c
pa

where ec and p c are the critical void ratio and associated critical mean normal stress,
respectively. p a is a reference pressure usually set to equal the atmospheric pressure (101 kPa)
for convenience; and c , and e are dimensionless material constants. Locating the critical
state line in e-p' space is critical for determining the state parameter, . It has been shown that
we can closely approximate the critical state line in the e-p' space by considering the relationship
between the average shape factor, f , and the position of the critical state line of sands based on
laboratory triaxial test data as shown in Figure 6.21. The plot shown in Figure 6.23 defines the
slope of the critical void ratio versus c for Sacramento sand:

141

Figure 6.23
Relationship between Critical Void Ratio and Mean Normal Stress for Sacramento Sand

Figure 6.24
In-Situ Critical Void Ratio Variation with Mean Normal Pressure for Sands at the
Instrumented Site, Port Island, Kobe

Figure 6.24 shows the in situ critical void ratio variation with the mean normal pressure
for sands at an instrumented site, Port Island, Kobe, Japan (Arulanandan, 1995). To determine

142

the model parameters c , , and e one has to transfer the critical state line to a line on

p
e c plane as shown in Figure 6.22.
pa
6.6 PROBLEMS

1. Describe in detail the non-destructive electrical method to determine the in situ porosity of
saturated non-clay minerals (sand) and clay minerals.
2. a) Given the average formation factor = 3.0, average shape factor = 1.3, and anisotropic
index = 1.1 of a saturated sand layer, determine whether the sand will boil.
b) Derive any equations used in solving part (a).
3. For a dilute suspension containing non-clay minerals, show that the following equation is
true for the formation factor, F:
1 n
F = 1 +
f ,
n
where n is the porosity and f is related to the shape factor S.
(3S 1)
4. Show that for concentrated suspensions F = n-f , where f =
.
6S (1 S )
5. Describe in detail how you can independently predict the formation factor using thin section
studies by considering orientation and shape parameters.
6. Discuss using the formation factor, shape factor, and anisotropy index in non-destructively
quantifying the structure of non-clay minerals to determine in situ properties.
7. Describe briefly the mechanism causing each of the following soil behaviors: (a) Swelling;
(b) Liquefaction. Discuss how to evaluate them using non-destructive soil characteristics.
REFERENCES

Anandarajah, A., Dafalias, Y. F., and Herrmann, L. R. (1984). Bounding Surface Plasticity
Model for Anisotropic Clays, Proceedings of the 5th ASCE-EMD Conference, University of
Wyoming, Laramie.
Anandarajah, A., Arulanandan, K., Dafalias, Y. F., and Herrmann, L. R. (1982). In-Situ
Determination of Stress-Strain Relationships of Clays,
International Symposium on
Constitutive Laws for Engineering Materials: Theory and Applications, Tucson, Arizona, Jan.
10-14.
Archie, G. E. (1942). The Electric Resistivity Log as an Aid in Determining Some Reservoir
Characteristics, Transcripts of the American Institute of Mining, Metallurgical and Petroleum
Engineers, Vol. 146, pp. 54-61.

143

Arulanandan, K. (1977). Method and Apparatus for Measuring In-Situ Density and Fabric of
Soils, Patent, Regents of the University of California.
Arulanandan, K. (1980). ECI 283 class notes, Department of Civil and Environmental
Engineering, University of California, Davis.
Arulanandan, K. (1987). Non-Destructive Characterization of Particulate Systems for Soil
Classification and In Situ Prediction of Soil Properties and Soil Performance. Keynote Lecture,
Proceedings of the International Conference in Geotechnical Engineering, Calgary, Canada; A.A.
Balkema, Rotterdam.
Arulanandan, K. (1991). Dielectric Method for the Prediction of Porosity of Saturated Soils,
Journal of Geotechnical Engineering, ASCE, Vol. 117, No. 2, February, pp. 319-330.
Arulanandan, K. (1995). UC Davis Report, Lecture, Kyoto University, Japan.
Arulanandan, K. (1997). ECI 174 class notes, Department of Civil and Environmental
Engineering, University of California, Davis.
Arulanandan, K., and Yogachandran, C. (2000). Dielectric Method for Non-Destructive
Characterization of Soil Composition, GeoDenver 2000, ASCE Geo-Institute, ASCE Special
Publication No. 110.
Arulanandan, K., and Kutter, B. (1978). A Directional Structure Index Related to Sand
Liquefaction, Proceedings of the Specialty Conference on Earthquake Engineering and Soil
Dynamics, ASCE, Pasadena, California, pp. 213-230.
Arulanandan, K., and Muraleetharan, K. K. (1988). Level Ground Soil Liquefaction Analysis
using In-Situ Properties: I and II, Journal of Geotechnical Engineering, ASCE, Vol. 114, No. 7.
Arulanandan, K., Arulmoli, K., Dafalias, Y. F., and Herrmann, L. R. (1982). In Situ
Characterization of Saturated Sands and Silts for the Prediction of Dynamic Shear Modulus and
Shear Wave Velocity, Department of Civil Engineering, University of California, Davis,
California, Report to the Air Force Office of Scientific Research, Grant No. AFOSR-82-0216.
Arulanandan, K., Yogachandran, C., and Rashidi, H. (1994). Dielectric Conductivity Methods
of Soil Characterization, Geophysical Characterization of Sites, Volume Prepared by ISSMFE
Technical Committee #10 for the XIII International Conference on Soil Mechanics and
Foundation Engineering, New Delhi, India, pp. 81-90.
Arulmoli, K. (1980). Sand structure characterization for in situ testing. Thesis submitted in
partial satisfaction of the requirements for the degree of Master of Science in Engineering,
University of California, Davis.

144

Arulmoli, K. (1982). Electrical Characterization of Sands for In Situ Prediction of Liquefaction


Potential, Thesis submitted in partial satisfaction of the requirements for the degree of Doctor of
Philosophy, University of California, Davis.
Arulmoli, K., Arulanandan, K., and Seed, H. B. (1985). New Method for Evaluating
Liquefaction Potential, Journal of the Geotechnical Engineering Division. ASCE, Vol. 111,
No. 1, pp. 95-114.
Been, K., and Jefferies, M.G. (1985). A state parameter for sands. Geotechnique, 35(2), 99102.
Bruggeman, D.A.G. (1935). Berechnung Verschiedenez Physikalischer Konstanten Von
Heterogenen Substanzen, Ann. Phys. Lpz., 24(5), 636.
Dafalias, Y. F., and Arulanandan, K. (1979). The Formation Factor Tensor in Relation to
Structural Characteristics of Anisotropic Granular Soils, Proceedings, Colloque International du
C.N.R.S., Euromech Colloquium 115, Villard-de-Laus, France.
Dafalias, Y. F., and Arulanandan, K. (1983). The Formation Factor Tensor in Relation to
Structural Characteristics of Anisotropic Granular Soils, Colloques Internationaux du CNRS,
pp. 183-197.
Fricke, H. (1924). A Mathematical Treatment of the Electrical Conductivity and Capacity of
Dielectric Dispersive Systems. Physical Review, Vol. 24, pp. 575-587.
Hardin, B. O., and Drenevich, V. P., (1970) Shear Modulus and Damping in Soils:
I. Measurement and Parameter Effects; II. Design Equations and Curves, Technical Reports
UKY 27-70-CE 2 and 3, College of Engineering, University of Kentucky, KY., July.
Harvey, S. J. (1981). Electrical Characterization of Sand Compressibility, Thesis submitted in
partial fulfillment of the requirements for the degree of Master of Science, University of
California, at Davis.
Hvorslev, M. J. (1949). Subsurface Exploration and Sampling of Soils for Civil Engineering
Purposes, Report of the Committee on Sampling and Testing, Soil Mechanics and Foundations
Division, ASCE.
Jackson, P. D. (1975). An Electrical Resistivity Method for Evaluating the In Situ Porosity of
Clean Marine Sands, Marine Geology, Vol. 1, No. 2, pp. 91-116.
Li, X. S. (1997). Modeling of Dilative Shear Failure, Journal of Geotechnical Engineering,
ASCE, Vol. 123, No. 7, pp. 609-616.
Li, X.S., Dafalias, Y.F., and Wang, Z.L. (1999). State-dependent dilatancy in critical-state
constitutive modeling of sand. Canadian Geotechnical Journal, 36(4), 599-611.

145

Manzari, M.T., and Dafalias, Y.F. (1997). A critical state two-surface plasticity model of
sands. Geotechnique, 47(2), 255-272.
Maxwell, J. C., (1881). A Treatise on Electricity and Magnetism, 2nd ed., Clarendon Press,
Oxford.
Meegoda, N. J. (1983). Prediction of In Situ Stress State Using Electrical Method, Thesis
submitted in partial satisfaction of the requirements for the degree of Master of Science in
Engineering, University of California, Davis.
Meegoda, N. J., and Arulanandan, K. (1986). Electrical Method of Predicting In Situ Stress
State of Normally Consolidated Clays, Proceedings of the In Situ 86, GT Division, ASCE,
Blacksburg, Virginia, June 23-25, pp. 794-808.
Mitchell, J. K., Chatoian, J. M., and Carpenter, G. C. (1976). The influence of fabric on the
liquefaction behavior of sand, Report, U.S. Army Corps of Engineering, Waterways Experiment
Station, Vicksburg, Mississippi.
Mousseau, R.J., and Trump, R.P. (1967). Measurement of electrical anisotropy of clay like
materials, Journal of Applied Physics, 38(11), 4375-4379.
Mulilis, J. P., et al. (1977). Effects of Sample Preparation on Sand Liquefaction, Journal of
the Geotechnical Engineering Division, ASCE, Vol. 103, No. 2, pp. 91-108.
Pietruszczak, St., and Mroz, Z. (1983) On Hardening Anisotropy of K0 Consolidated Clays,
International Journal for Numerical and Analytical Methods in Geomechanics, Vol. 7, pp. 1938.
Seed, H. B., and Idriss, I. M. (1970). Soil Moduli and Damping Factors for Dynamic Response
Analysis, Report No. EERC 70-10, University of California, Berkeley, California, December.
Verdugo, R., and Ishihara, K. (1996). The Steady State of Sandy Soils, Soils and Foundations,
Vol. 36, No. 2, pp. 81-91.
Wagner, K. W. (1914). Archives of Electrotechnology, Vol. 2, p. 371.
Wang, Z. L., Dafalias, Y. F., and Shen, C. K. (1990). Bounding Surface Hypoplasticity Model
for Sand, Journal of Engineering Mechanics, ASCE, Vol. 116, No. 5, May, pp. 983-1001.
Wyllie, M. R. J., and Gregory, A. R. (1953). Formation Factors of Unconsolidated Porous
Media: Influence of Particle Shape and Effect of Cementation, Petroleum Transcripts,
American Institute of Mining, Metallurgical and Petroleum Engineers, Vol. 198, pp. 103-109.

146

Chapter 7
Soil Behavior:
Erosion in Relation to Earth Dam Failure
and Frost Heave
7.1 EROSION
7.1.1 Introduction
One of the major concerns regarding the safety of embankment dams is the
problem of internal soil stability, especially when particles are subjected to drag forces
resulting from reservoir seepage. This problem is particularly important when transverse
cracking of the core occurs. There are various reasons why cracking occurs, such as
differential settlement or hydraulic fracture, all of which can cause serious internal
erosion resulting in catastrophic dam failure. Numerous cases of near failure and total
failure of dams, resulting from internal erosion, have been investigated and reported in
detail (Sherard, 1972a, b; Vargas and Hsu, 1970; Vaughan, 1976).
7.1.2 Classification of Impervious Core Materials with Respect to Dispersion and
Erodibility
Serious internal erosion of dams appears to have occurred with increasing
frequency during the past 40 years, based on the number of cases summarized by Ripley
(1978). These incidents have resulted in extensive studies to identify core materials that
have dispersive fines.
Investigators have used a variety of empirical tests to identify and classify soils
with respect to their dispersion characteristics. These include the crumb test (Emerson,
1954), the SCS dispersion ratio test (Decker and Dunnigan, 1976), and the pinhole test
(Sherard, et al., 1976). These three tests are used mainly to identify dispersive clays, and
even then are not suitable for all cases. Perry (1976) summarizes the reasons why these
tests are not able to identify erosion characteristics of some soils.
Figure 7.1 shows the classification chart for erosion proposed by the Bureau of
Reclamation (1962). The misuse of Figure 7.1 in deciding on the erodibility of a core
without site-specific testing could lead to problems. Plotted on this chart are the
plasticity characteristics of nondispersive core materials used in various dams
(Washington Co., Dam, Hills Creek Dam, Balderhead Dam, and Stockton Creek Dam)
that have shown piping failure through the core. Also plotted are the plasticity
characteristics of nondispersive core materials used in two dams (Cedar Springs Dam and
Yosemite Dam) that have shown stable performance with respect to internal erosion.
These results show that the classification system used is not able to predict piping
performance. The reason for this is that the system does not take into account important

147

erosion factors such as the chemistry of the soil and the composition of pore and eroding
fluid.

Figure 7.1
Soil Erodibility Based on Plasticity Chart versus Performance of Dams Constructed
of Nondispersive Cohesive Soils (PI > 6, % Clay < 12)
7.1.3 New Approach to Quantifying the Erodibility Characteristics of Soils
Any classification system used to identify the erodibility of soil should be based,
ideally, on some measure of the ease with which a soil can be eroded. The phenomenon
of surface erosion occurs when fluid flow induces strong shear stresses in the surfaces of
a cracked zone of core material, removing aggregates or flocs of particles. One of the
accepted approaches for evaluating the initiation and sustaining of particle motion or
aggregation involves calculating the shear stress caused by hydraulic flow. Any critical
shear designation should denote either: (1) a stress at which erosion begins, or (2) a stress
that would cause a particular erosion rate. Shields (1963) defined the critical shear stress,
C , as a measure of erosion resistance. Specifically, C is the value of the stress for zerosediment discharge obtained by extrapolating a graph of observed erosion rate vs. shear
stress as shown in Figure 7.2.
Values of critical shear stress have been measured directly on remolded and
undisturbed samples of soils using a rotating cylinder (Arulanandan et al., 1976b).
Figures 7.3, 7.4 and 7.5 show a photograph and schematics of a rotating cylinder
apparatus.

148

Figure 7.2
Definition of Critical Shear Stress, Rate of Change of Erosion Rate, and Rate of
Erosion

Figure 7.3
Photograph of a Rotating Cylinder Apparatus (Arulanandan and Krone, 1974)

149

Figure 7.4
Schematic of a Rotating Cylinder Apparatus and Soil Sample
(Arulanandan, Loganathan, and Krone, 1974)

Figure 7.5
Rotating Cylinder Apparatus Shear Stress Measurement
(Arulanandan et al., 1976b)
The literature reports numerous results showing the influence of clay type and
amount, pore and eroding fluid compositions, pH, temperature, and organic matter on C
(Arulanandan, et al., 1976b; Arulanandan and Heinzen, 1977; Karpoff, 1955;
Arulanandan, et al., 1973). The mechanism of soil erosion is a complex phenomenon
involving soil structure and the nature of the interaction between the pore and eroding

150

fluid at the surface. Because the critical shear stress, C , depends on these factors, it can
be used as a fundamental parameter to classify erodibility characteristics.
In cases where the rotating cylinder or flume apparatus is not available, we can
indirectly measure C from knowledge of the soil structure and the composition of the
pore and eroding fluid. A composite index that can characterize the soil type is the
magnitude of dielectric dispersion, 0 (Arulanandan and Heinzen, 1977). This index,
obtained from the measurement of electrical properties, has been used to propose a new
soil classification system as shown in Figure 7.6.

Figure 7.6
Effect of Clay Type and Amount on Magnitude of Dielectric Dispersion
(Alizadeh, 1975)
In conjunction with values of pore fluid concentration and the Sodium Adsorption
Ratio (SAR), with distilled water as the eroding fluid, 0 can be used to indirectly
measure C . This measurement is based on correlations established from extensive tests
on a wide variety of clay soils (Alizadeh, 1975; Arulanandan and Heinzen, 1977), as
shown in Figures 7.7 through 7.10.

151

Figure 7.7
Relationship between Critical Shear Stress, Sodium Adsorption Ratio, and
Dielectric Dispersion for Saturated Remolded Soil with Distilled Water as Eroding
Fluid and Soil Pore Fluid Concentration of 5 Milliequivalents/Liter
(Arulanandan and Perry, 1983)

Figure 7.8
Relationship between Critical Shear Stress, Sodium Adsorption Ratio, and
Dielectric Dispersion for Saturated Remolded Soil with Distilled Water as Eroding
Fluid and Soil Pore Fluid Concentration of 50 Milliequivalents/Liter
(Arulanandan and Perry, 1983)

152

Figure 7.9
Relationship between Critical Shear Stress, Sodium Adsorption Ratio, and
Dielectric Dispersion for Saturated Remolded Soil with Distilled Water as Eroding
Fluid and Soil Pore Fluid Concentration of 125 Milliequivalents/Liter
(Arulanandan and Perry, 1983)

Figure 7.10
Relationship between Critical Shear Stress, Sodium Adsorption Ratio, and
Dielectric Dispersion for Saturated Remolded Soil with Distilled Water as Eroding
Fluid and Soil Pore Fluid Concentration of 250 Milliequivalents/Liter
(Arulanandan and Perry, 1983)

153

7.1.4 Use of Critical Shear Stress to Examine Piping Failures of Dams


Sherard (1972a, b) has compiled an excellent catalogue of soil and pore fluid
composition data from dams where severe erosion problems occurred. Tables 7.1, 7.2,
and 7.3 give selected data for dispersive-cohesive, nondispersive-cohesive, and noncohesive soils obtained from these dams.
The magnitude of dielectric dispersion, 0 , has been estimated from the cation
exchange capacity values given by Sherard for the core material of these dams, using a
relationship developed by Fernando, et al. (1977), as shown in Figure 7.11.

Artificial soils were various mixtures of silica flour


and clays (montmorillonite, illite, and kaolinte).
Natural soils were eleven soils obtained from various
locations in California

Figure 7.11
Relationship between Magnitude of Dielectric Dispersion and Cation Exchange
Capacity for Artificial and Natural Soils
(Fernando et al., 1977)
We can now estimate the value of C for each dam from Figures 7.7 through
7.10, which apply to remolded saturated soil containing 40% clay. Tables 7.1 through
7.3 summarize the results.

154

Table 7.1
Properties of the Fines in Dispersive-Cohesive (PI > 6, % Clay > 12) Embankment
or Core Materials and their Erodibility Characteristics
(Arulanandan and Perry, 1983)

155

Table 7.1 Continued

Table 7.2
Properties of the Fines in Nondispersive-Cohesive (PI > 6, % Clay > 12)
Embankment or Core Materials and their Erodibility Characteristics
(Arulanandan and Perry, 1983)

Table 7.1 shows 19 dispersive-cohesive soils from dams that all experienced
piping failure. In each case the estimated critical shear stress is 4 dynes/cm2. Table 2
shows six nondispersive-cohesive soils from both stable and unstable dams. For the four
dams that experienced piping failure, the estimated critical shear stress is 4 dynes/cm2.

156

For the two stable dams, it is 9 dynes/cm2 and 13 dynes/cm2. Table 3 shows four
noncohesive soils from both stable and unstable dams. For the three dams that
experienced piping failure, the critical shear stress is 4 dynes/cm2. For the stable dam
it is 14 dynes/cm2.
Table 7.3
Properties of the Fines in Noncohesive (PI 6, % Clay 12) Embankment or Core
Materials and their Erodibility Characteristics
(Arulanandan and Perry, 1983)

One of the main points these data demonstrate is that core materials other than
dispersive clays can show low erosion resistance similar to dispersive clays. A good
example of this is the core material from the Balderhead Dam, a glacial till that,
according to Figure 7.1, has been considered to have high erosion resistance. It actually
has low erosion resistance, as suggested by the low value of critical shear stress (4
dynes/cm2). The core materials from the Stockton Creek, Washington County, and Hills
Creek Dams would also have been expected to have high erosion resistance according to
Figure 7.1. Again the low values of C ( 4 dynes/cm2) suggest a low resistance to
erosion. In the case of the Teton, Hyttejuret, Matahina, and Mill Site Dams, the core
materials are all suspected of being highly erodible due to their low plasticity. For the
first three this is confirmed by the low critical shear stress ( 4 dynes/cm2). The high
value (14 dynes/cm2) of C for the Mill Site Dam suggests high erosion resistance.
Based on this limited data, certain categories are proposed for the classification of
core materials with respect to their erosion resistance. These categories, defined in terms
of critical shear stress, are as follows:

157

1. Erodible soils, C 4 dynes/cm2. Core materials in this category warrant extensive,


detailed, filtration tests of the proposed filter material to investigate its capacity to
block migration of clay fines.
2. Moderately erodible fines, 4 dynes/cm2 < C < 9 dynes/cm2. Core materials in this
category also warrant filtration tests of the proposed filter material.
3. Erosion resistant soils, C 9 dynes/cm2. For core materials in this category, filter
gradation can be determined on the basis of simple filter design criteria.
In addition to the previous requirements, two other considerations are important in
evaluating dam stability. First, the concentration of eroding fluid (reservoir water for
existing dams and river water for proposed dams) can have a major effect on the
erodibility of core material. Second, attention should be paid to the permeability, crack
resistance, and segregation characteristics of the filter material (Arulanandan and Perry,
1983).
7.1.5 Summary and Conclusions
1. The erodibility characteristics of a clay core material have been quantified in terms of
a parameter called the critical shear stress, C , which is the stress required to initiate
erosion during hydraulic flow.
2. Critical shear stress is influenced by the clay type and amount, composition of pore
and eroding fluid, and soil structure. A highly erodible soil is defined as one where
C = 0. This quantitative definition allows us to classify the erodibility characteristics
of all soils without resorting to qualitative definitions for dispersive and
nondispersive clays. The value of C can be obtained either by a rotating cylinder
test or a flume test.
3. Based on erosion resistance in terms of C , core materials with significantly different
properties from dispersive clays can have low erosion resistance similar to dispersive
clays.
4. Using Atterberg limits to classify erosion characteristics has been shown to be
inadequate, especially when dealing with CL and CH materials. Even for very low
plasticity materials, such as sands and silts with low clay content, Atterberg Limits
classification does not always distinguish a nonerodible from an erodible material.
5. Performance of clay core dams has been examined with respect to piping. A
relationship between C and the piping behavior of various dams is shown. Core
materials with a value of C less than 4 dynes/cm2 have had previous piping failure.
7.1.6 A Brief Case Study: Teton Dam
On June 5, 1976, the Teton Dam, situated along the Teton River about three miles
northeast of Newdale Idaho, failed due to erosion of the core. Figure 7.12 shows a
photograph of the eroded dam. Needless to say, the failure of the dam was a disaster.
Fourteen people were killed, and the break caused an estimated $400 million to $1 billion

158

dollars in property damage (Independent Panel to Review Cause of Teton Dam Failure,
1976).

Figure 7.12: Failure of the Teton Dam


(Independent Panel to Review Cause of Teton Dam Failure, 1976)

Figure 7.13
Results of Rotating Cylinder Test on a Teton Dam Core Sample
(Arulanandan, et al., 1976a)

159

Figure 7.14
Results of Rotating Cylinder Test on a Teton Dam Core Sample
(Arulanandan, et al., 1976a)
Subsequent analysis of the failure and core soil revealed that the core material had
a low critical shear stress. Figures 7.13 and 7.14 show the results of rotating cylinder tests
on core samples of the Teton Dam (Arulanandan, et al., 1976a). The three tests show
critical shear stresses below 1 dyne/cm2
7.1.7 Modified Rotating Cylinder Apparatus
Moore and Marsh (1962) developed an original rotating cylinder test at the
University of Texas at Austin. Baker (Baker, 1996) modified the rotating cylinder device
to reduce friction in the system and to enable direct measurement of the shear stress
induced on the sample.
In essence, the modified rotating cylinder device consists of a stationary soil
sample with a rotating outer cylinder surrounding, but not touching, the soil sample. The
space between the cylinder and soil sample is filled with the chosen eroding fluid. As the
cylinder rotates, the eroding fluid produces shear stress around the sample. The rotational
speed of the cylinder controls the amount of shear. The shear stress may be measured
directly with the modified rotating cylinder device as follows. A floating air puck is
connected on top of the sample. If the soil sample were allowed to turn freely as the
shear stress was induced, friction would be minimal due to the air bearing. However, an
external force resists rotation, which aids in measuring the shear stress. A thin aluminum
beam is attached to the top center of the air puck. A stationary post is placed near the far
end of the beam to resist movement. By this mechanism, the sample is not allowed to
rotate with the eroding fluid and the outer cylinder. As the force is applied, the aluminum

160

beam bends. The amount of deflection is measured by a strain gauge and directly
correlated to shear stress induced on the sample via a gauge factor. Figure 7.15 shows a
schematic of the modified rotating cylinder device.

Figure 7.15
Schematic of the Modified Rotating Cylinder Device (Baker, 1996)
7.1.8 Calculating the Gauge Factor
A strain indicator can be modified via the gauge factor to display shear stress
directly. The gauge factor may be calculated as follows given the dimensions of the
aluminum beam and the placement of the strain gauge:

Figure 7.16
Strain Gauge Placement for Calculation of the Gauge factor (Baker, 1996)
Measure the radius of the sample and the length from the center of the sample to
the strain gauge represented by l 2 in Figure 7.16. Measure the length from the strain
gauge to the point where the stationary post touches the rod and force Fa is applied as
represented by l1 in Figure 7.16. Based on a free body diagram, the equation for shear
stress is:

161

Ebh (l1 + l 2) 68947.57 dynes / cm 2


2
shearstress =
(dynes / cm )

2
2
lb / in
12r l1

(gage factor)(shear stress)(10 8 )


=
strain cons tan t

(1)

(2)

where:
E = Youngs Modulus of the aluminum beam = 9.6 x 106 psi
b = the thickness of the aluminum beam
h = the height of the aluminum beam
r = the radius of the sample
strain constant = 2.065
The gauge factor may be calculated as:

2.065(12l1 r 2 )
lb / in 2
gage factor =

10 8 [bhE ](l1 + l 2) 68947.57 dynes / cm 2

(3)

When the gauge factor is entered into the strain indicator, the output can be
directly read as shear stress in dyne/cm2.
7.1.9 Measuring the Erosion Rate:
The erosion rate of a sample during a rotating cylinder test is measured as
follows:

e& =

M
At

(4)

where:
e& = the erosion rate
M = the change in sample mass due to a period of testing
t = the time elapsed during a period of testing
A = the surface area of the sample exposed to the eroding fluid
As shown in Figure 7.17 a rotating cylinder test consists of several trial periods
with different shear stresses. Each trial period results in a different erosion rate, which is
plotted versus shear stress. The critical shear stress, C , is determined as the shear stress
at which erosion begins or where the experimental erosion rate/shear stress line crosses
the shear stress axis as Figure 7.2 demonstrates. Below the critical shear stress, the
erosion rate is zero, and the soil will not erode. As seen in Figure 7.17, the critical shear
stress for the Yolo Loam sample is approximately between 0 and 1 dyne/cm2. This

162

indicates a very erodible soil, in combination with this particular pore fluid concentration,
pore fluid Sodium Adsorption Ratio, and eroding fluid (deionized water).

Erosion Rate (g/cm 2/min)

0.25
0.2
0.15
0.1
0.05
0
0

10

20

30

40

Shear Stress (dynes/cm 2)

Figure 7.17
Rotating Cylinder Test on Yolo Loam with a Pore Fluid Concentration of 50 meq/L
and an SAR of 45 with Deionized Water as the Eroding Fluid (Baker, 1996)

The shear stress may be controlled during a rotating cylinder test by varying the
speed of cylinder rotation. Faster rotation causes greater eroding fluid velocity around
the sample, producing larger shear stresses. The shear stress should be kept constant
during a test time period. Baker (1996) used one-minute test periods. The sample should
be weighed before and after the test period to determine the change in mass. The sample
surface area may be assumed constant during the testing.
7.1.10 Conclusion

The erodibility of consolidated soils depends on the swell potential of a soil, i.e.,
the soil mineralogy, pore fluid composition (SAR and concentration), and concentration
of the eroding fluid. The magnitude of dielectric dispersion depends on mineralogy (and
swell potential), and thus is a useful parameter for determining soil erodibility. The
critical shear stress, C , is a useful parameter defining soil erodibility. C may be
determined via a rotating cylinder test, or by combining the magnitude of dielectric
dispersion with knowledge of the SAR and pore fluid concentration.
7.2 FROST HEAVE
7.2.1 Introduction

When a soil freezes, ice layers or lenses form in the soil, as shown in Figure 7.18.
They range from a millimeter to several centimeters in thickness. The lenses are pure ice

163

and are free from large numbers of entrapped soil particles. As the soil freezes, the
ground surface may heave by as much as several centimeters, and the overall volume
increase can be many times greater than the 9% increase due to the increase in the
volume of water upon freezing.
Due to high heaving pressures, freezing soils under foundations and pavements
can lead to major problems. Uneven uplift can occur during freezing, and there can be a
loss of support when ice lenses melt, leaving water-filled voids in the soil. Clearly,
freezing soils can cause large amounts of damage to structures and roadways.

Figure 7.18
Ice Lenses and Frost Heave (Mitchell, 1993)

There are three necessary conditions for ice lens formation and frost heave:
1. A frost-susceptible soil
2. Freezing temperatures
3. A supply of water
Frost heave is therefore common in areas where winter temperatures are below freezing
and in soils where the water table or a pocket of water is close to the freezing front.
Almost any soil could be made to heave if the water supply and the freezing rate
were properly controlled. However, only certain soils in nature have freezing rates that
make them frost susceptible. Sands, gravels, and fat intact clays generally do not heave.
Soils with more than 3% of their particles smaller than 0.002 mm are potentially frost
susceptible.
7.2.2 Mechanisms of Frost Heave

The formation of ice lenses is a complex phenomenon involving interrelationships


between the phase change of water to ice, transport of the water to the ice lenses, and
unsteady heat flow in the freezing soil. There are four stages in the ice lens formation
cycle:
1. Nucleation of ice
2. Growth of ice lenses
3. Termination of ice growth
4. Heat and water flow between the end of the termination of ice and the start of
the nucleation of ice

164

The process of freezing and ice lens formation is complicated. For example, if a
homogeneous soil at a uniform water content and temperature, T0, above freezing is
subjected to a surface temperature, Tn, below freezing, temperatures will vary with depth
at some time t, as shown in Figure 7.19.

Figure 7.19
Variation of Temperature with Depth in a Freezing Soil (after Mitchell, 1993)

The rate of heat flow at any point is given by k(dT/dz). If (dT/dz) at a point A is
greater than at a point B, then the temperature of the element will drop. When the water
changes to ice, it gives up its latent heat, thus slowing the rate the freezing front moves
through the soil.
Heave results from the formation of an ice lens at point A in Figure 7.19. The
energy needed to lift the overlying material is available because ice forms under a
condition of supercooling. One degree Celsius of supercooling is enough to lift 12.5 kg a
distance of 1 cm. As long as water can flow to a growing ice lens fast enough, the
volumetric heat and latent heat can produce a temporary steady-state condition so that
heaving will occur. For example, a silty clay can supply water at a rate adequate for a
heave rate of 1 mm/hr. In a sand, the volume of water in the pores is large, and the latent
heat raises the freezing temperature to the normal freezing point. Therefore, there is no
supercooling and no heave. In a clay, the hydraulic conductivity is so low that water
cannot be supplied fast enough for ice lens growth. Heave can only develop in clays if the
freezing rate is slowed beyond that found in nature. Silts and silty soils have the optimum
combinations of pore size, hydraulic conductivity, and freezing point depression so that
natural rates of freezing in the field can result in large heave.
7.2.3 Segregation Potential Method for Predicting Frost Heave Potential

165

The segregation potential method uses a laboratory freezing test to measure the
property of segregation potential. Segregation potential is defined as the ratio of the rate
of water migration to the overall temperature gradient in the frozen fringe (the zone
between the frost front and the segregation front where ice lenses grow). Water migration
and temperature gradient depend on soil structure and physico-chemical properties.
Therefore, using such an approach to quantify the frost susceptibility of soils will result
in accurate predictions of frost heave.
Although the segregation potential approach is relatively accurate, the freezing
test requires a long time to conduct. Also, unless factors such as cooling rate are kept
constant, the results will vary from laboratory to laboratory. To overcome the long
testing time and discrepancies between laboratories, Arulanandan and Shinde (1989)
developed a relationship between the segregation potential, the c parameter from the
Three-Element Model, and the specific surface area. Figure 7.20 shows this relationship.
Specific surface area can be used to quantify the mineralogy and the shape of the
particles, and the c parameter quantifies pore size distribution. By considering the rate of
water migration, temperature gradient, pore size distribution, mineralogy, and particle
shape, we can accurately predict frost heave potential.

Figure 7.20
Correlation between Specific Surface Area, the c Parameter of the Three-Element
Model, and Segregation Potential

In preparing the results shown in Figure 7.20, specific surface area of the soils
was predicted from the magnitude of dielectric dispersion using the relationship shown in
166

Figure 3.20. The c parameters were determined via Three-Element Model optimizations.
The soil mixtures used by Rieke, et al. (1983) during their segregation potential studies
were duplicated in the laboratory at the University of California, Davis, for electrical
dispersion tests. Rieke then performed segregation potential measurements on the soils
newly mixed at UC Davis (Rieke, 1972; Arulanandan and Shinde, 1989). The pressure
and the rate of cooling affect segregation potential. In this newer study, both the pressure
and the rate of cooling were kept constant to concentrate on pore size and specific surface
area effects (Arulanandan and Shinde, 1989).
7.3 PROBLEMS

1. How can the erodibility characteristics of clays be evaluated? Assuming that you have
no erosion test apparatus at your disposal, describe how you will evaluate the erosion
potential of soil. Discuss the mechanism of clay erosion. Explain the factors that
influence the erodibility characteristics.
2. Describe the mechanism causing frost heave, and how you would evaluate frost heave
potential.
REFERENCES

Alizadeh, A. (1975). Amount and Type of Clay and Pore Fluid Influences on Critical
Shear Stress and Swelling of Cohesive Soils, Dissertation, presented to the University of
California at Davis, CA, in partial fulfillment of the requirements for the degree of
Doctor of Philosophy.
Arulanandan, K., Gillogley, E., Rabbon, P., and Heinzen R. (1976a). Evaluation of
Erodibility Characteristics of Teton Dam Soils, Report to the Independent Panel to
Review Cause of the Teton Dam Failure, October.
Arulanandan, K., and Heinzen, R. T. (1977). Erosion and Solid Matter Transport in
Inland Waters Symposium, Proceedings: Symposium-United Nations Educational,
Scientific and Cultural Organization and IAHS, pp. 75-81.
Arulanandan, K., Loganathan, P., and Krone, R. B. (1975) Pore and Eroding Fluid
Influences on Surface Erosion of Soil, Journal of the Geotechnical Engineering
Division, Proceedings of the American Society of Civil Engineers, Vol. 101, No. GT1,
Jan., pp. 51-66.
Arulanandan, K., Loganathan, P., and Krone, R. B. (1976b) Closure to Pore and
Eroding Influence on the Surface Erosion of a Soil, Journal of Geotechnical
Engineering Division, ASCE, Vol. 102, No. GT8, Proceedings Paper 11078, Aug., pp.
883-884.

167

Arulanandan, K., and Perry, E. B. (1983). Erosion in Relation to Filter Design Criteria
in Earth Dams, Journal of Geotechnical Engineering, ASCE, Vol. 109, No. 5, May, pp.
682-698.
Arulanandan, K., Sargunam, A., Loganathan, P., Krone, R. B. (1973). Application of
Chemical and Electrical Parameters to Prediction of Erodibility, (Special Report 135 on
Soil Erosion), Proceedings: Conference Workshop on Soil Erosion, Highway Research
Board, Jan. 26, pp. 42-51.
Arulanandan, K., and Shinde, S. B. (1989). Frost Heave Susceptibility Criteria For
Soils, Report: Research Department, Esso Resources Canada, ERCL.RS.88.03.
Baker, C. T. (1996). In-Situ Evaluation of Critical Shear Stress and Porosity of Soils,
Thesis submitted in partial satisfaction of the requirements for the degree of Master of
Science, Department of Civil Engineering, University of California at Davis.
Decker, R. S., and Dunnigan, L. P. (1976). Development and Use of the SCS Dispersion
Test, American Society for Testing and Materials Symposium on Dispersive Clays,
Chicago, IL.
Emerson, W. W. (1954). The Determination of the Stability of Soil Crumbs, Journal of
Soil Saturation, Vol. 5, pp. 233-250.
Fernando, M. J., Burau, R. G., and Arulanandan, K. (1977). A New Approach to
Determination of Cation Exchange Capacity, Proceedings: Soil Science Society of
America.
Gibbs, H. J. (1962). A Study of Erosion and Tractive Force Characteristics in Relation
to Soil Mechanics Properties, Soils Engineering Report No. EM-643, U.S. Department
of the Interior, Bureau of Reclamation, Denver, CO.
Independent Panel to Review Cause of the Teton Dam Failure (1976). Report to U.S.
Department of the Interior and the State of Idaho on Failure of the Teton Dam.
Karpoff, K. P. (1955). The Use of Laboratory Tests to Develop Design Criteria for
Protective Filters, Proceedings: American Society of Testing and Materials, Vol. 55, pp.
1183-1193.
Mitchell, J. K., (1993). Fundamentals of Soil Behavior, John Wiley and Sons, New
York.
Moore, W.L., and Masch, F.D. (1962). Experiments on the Scour Resistance of
Cohesive Sediments, Journal of Geophysical Research, 67(4), 1437-1449.

168

Perry, E. B. (1976). Dispersion Characteristics of Soil From Teton Dam, Idaho, Report
to the Independent Panel to Review Cause of Teton Dam Failure.
Rieke, R. D. (1972). The Role of Specific Surface Area and Related Index Properties in
the Frost Susceptibility of Soils, Thesis submitted in partial satisfaction of the
requirements for the degree of Master of Science, Oregon State University, Corvallis.
Rieke, R. D., Vinson, and T. S., Mageau, D. W. (1983). The Role of Specific Surface
Area and Related Index Properties in the Frost Heave Susceptibility of Soils, Fourth
International Conference on Permafrost, National Research Council, Washington, D.C.
Ripley, C. F. (1978). Filter Zones for Earth Dams, Internal Communication-Water
Rights Branch, Ministry of the Environment, Victoria, British Columbia, Canada.
Sherard, J. L. (1972a). Embankment Dam Cracking, Embankment-Dam Engineering,
John Wiley and Sons, Inc., New York, NY, pp. 271-354.
Sherard, J. L. (1972b). Study of Piping Failure and Erosion Damage from Rain in Clay
Dams in Oklahoma and Mississippi, Soil Conservation Service, Washington D.C.,
March.
Sherard, J. L., Dunnigan, L. P., Decker, R. S., and Steel, E. F. (1976) Pinhole Test for
Identifying Dispersive Soils, Journal of Geotechnical Engineering Division, ASCE,
Vol. 102, No. GT1, Jan., pp. 69-85.
Shields, A. (1963). Anwendung der Aen lickei Kemechanik und der Tur bulentz
Porschung auf die Geschiebebewegung, (Berlin, 1936), translated by W. P. Off and
J. C. Van Uchelen, California Institute of Technology, Pasadena, CA.
Vargas, M., and Hsu, S. J. C. (1970). The Use of Vertical Core Drains in Brazilian
Dams, Proceedings, 10th ICOLD, Vol. 1, A36, R36, Montreal, Canada.
Vaughan, P. R.. (1976). Cracking of Embankment Cores and Design of Filters for their
Protection, Soueded Espanola de Mechanica del Suelo y Cimentaciones, No. 23, Sept.Oct., pp. 23-24.

169

Appendix A
Example Applications of the Formation
Factor Method

170

A.1 EVALUATION OF SITE RESPONSES AND DEFORMATIONS OF THE


INSTRUMENTED WILDLIFE SITE, CALIFORNIA (K. Arulanandan 1 M. ASCE and K.
(Siva) Sivathasan 2 S. M. ASCE)

A.1.1 Abstract
The observed dynamic response of the instrumented Wildlife site, California,
during the 1987 Superstition Hills earthquake was utilized to demonstrate the feasibility
of computer simulation of earthquake-induced site response and liquefaction-induced
deformations of a level ground site. A non-destructive in-situ electrical method was used
to obtain the initial state parameters and constitutive model constants representative of
the site. The measured shear wave velocity measurements were compared with the shear
wave velocity values evaluated using the electrical method. The dynamic response
analysis used the effective stress-based, nonlinear, fully coupled, finite element program
SUMDES with a reduced order bounding surface hypo-plasticity model to simulate the
stress-strain behavior of cohesive soils and modified reduced order bounding surface
hypo-plasticity model to simulate the stress-strain behavior of non-cohesive soils. The
results of the dynamic analysis showed that the acceleration, excess pore water pressure
response histories, and liquefaction-induced deformations agreed reasonably well with
the observed behavior during the Superstition Hills earthquake. The results of this study
show that computer simulation of earthquake effects of a level ground site is possible
using non-destructive in-situ testing to obtain the initial state and constitutive model
parameters representative of the site and a verified numerical procedure.
A.1.2. Introduction
Before-the-event evaluation of the response of our infrastructure systems
subjected to earthquakes is a desirable objective for the adoption of appropriate, safe and
economical retrofit measures of existing facilities and the safe and economical design of
new infrastructure facilities. Once a verified numerical procedure and in-situ properties
representative of the site are available, computer simulation of earthquake effects of any
site can be carried out, as demonstrated in a recent paper, Numerical Simulation of
Liquefaction-Induced Deformations (Arulanandan et al., 2000). The first objective of
this paper is to provide an additional example: an evaluation of the earthquake response
of the instrumented site (Wildlife site, California) where the pore pressure and
acceleration time responses and liquefaction-induced deformation were recorded during
the Superstition Hills earthquake of 1987. Wildlife site, California is the only
instrumented site where acceleration time responses, pore pressure time responses,
vertical and lateral deformations have been recorded (Holzer et al., 1989a). The second
objective of this paper is to demonstrate how the initial state parameters and constitutive
1
2

Professor, Dept. of Civ. and Envir. Engrg., Univ. of California, Davis, CA 95616.
Graduate Student, Dept. of Civ. and Envir. Engrg., Univ. of California, Davis, CA 95616.

171

model constants representative of the site were obtained using shear wave velocity and
in-situ formation factor measurements (ERTEC WESTERN, INC., 1982). In the earlier
paper Numerical Simulation of Liquefaction-Induced Deformations (Arulanandan et
al., 2000), the initial state parameters and constitutive model constants were obtained
using laboratory formation factor measurements on remolded samples and shear wave
velocity measurements.
A.1.3 Earthquake and Soil Characteristics at Instrumented Site
On November 24, 1987, at 05:15 PST (Pacific Standard Time) the Superstition
Hills, California earthquake struck the instrumented Wildlife site, adjacent to the Alamo
river in the Imperial valley, caused liquefaction, and developed lateral displacements
toward the river ranging from 0 to 230 mm. Extensive ground cracking indicative of
lateral spreading accompanied liquefaction at the array. The site is 23 km east of the
epicenter of the Elmore ranch earthquake and 31 km east-northeast of the epicenter of the
Superstition Hills earthquake. A location map of the liquefaction array and earthquake
epicenters, a stratigraphic cross-section of the array, and a schematic of instrument
deployment are shown in Figures A1a and A1b. A representative soil profile of the
Wildlife site is shown in Figures A2a and A2b.

172

173

The site had been instrumented with both accelerometers and electrical
piezometers and, for the first time, excess pore water pressure ratios of 100% were
measured in the field in a saturated silty sand site during an earthquake (Youd and
Bartlett, 1988; Holzer et al., 1989 a, b). Six pore water pressure transducers and two three
component force-balance accelerometers, one at the surface and one downhole, were
installed at the array. Instrumentation was installed in 1982. Shallow deposits at the array
consist of saturated, flood plain sediments that fill an old incised channel of the Alamo
River. The uppermost unit at the array is a 2.5 m thick flat lying silt bed that overlies the
unit that liquefied a 4.5 m thick silty sand. Beneath these floodplain deposits is a 5.0 m
thick silty clay unit, the uppermost unit of a dense and regionally extensive sedimentary
deposit. The Alamo River, a perennial stream because of drainage from irrigation,
currently occupies a 3.7 m deep channel 23 m east of the center of the array and controls
the water table depth at about 1.2 m. The soil properties of the site had been thoroughly
measured prior to the earthquake. Therefore, the 1987 Superstition Hills, California
earthquake records provide researchers with a unique opportunity to improve the
understanding of the mechanics of seismic pore pressure build up and liquefaction, as
well as the understanding the complex relation between pore pressure, water flow,
cracking and permanent displacements during lateral spreads.
A.1.4 Field Investigations
Seismic shear wave velocity tests and in-situ resistivity measurements were
performed at the Wildlife site on February 9, 10, 11 and 12, 1982. The field resistivity
measurements were obtained using a Geoelectronics Electrical Soil Probe Model GE-100
shown in Figures A3a and A3b. The electrical soil probe consists of a 3-inch diameter, 5
foot long tube with one foot hollow section at the end. The one foot hollow section is
similar to a conventional thin wall sampler, except that it contains three oriented
electrodes located on the surface of the inside wall. Aggregate soil resistivity is measured
in two directions between the three electrodes. In the probes solid section, an electric
pump draws fluid from the soil into an electrode cavity where water resistivity
measurements are automatically taken.
A drill rig is used to bore the hole to push the probe 9 inches into the undisturbed
soil at the borehole bottom. The operator then balances the down-hole resistivity bridge
from the surface control box. The bridge impedance is incrementally stepped by digital
electronic messages sent from the microprocessor housed in the instrument case or, if
necessary, the bridge is balanced manually. Separate readings from the horizontal
electrode set, the vertical electrode set, and the water cavity electrode set are then taken.
Crosshole shear wave velocity surveys were conducted at the field site. The
surveys were performed to determine the compressional (Pwave) and the shear wave
(Swave) velocities of the subsurface materials. During a crosshole survey, the seismic
waves were generated at a specific depth in a boring. The waves were detected by
geophones which were positioned at the same depth in adjacent borings. The result of a
crosshole survey is a profile of horizontally measured seismic velocities varying with
depth. Seismic waves were generated by striking the top of the drill rod in the source
boring with a sledge hammer while the drill bit was in contact with the formation. This
procedure generated relatively large amplitude, vertically oriented, S waves and much

174

lower level P waves in the horizontal travel path. Crosshole shear wave velocity
measurements are compared with the shear wave velocities evaluated using the electrical
method in Figure A3c.

175

A.1. 5 Response Studies Utilizing Wildlife Records


Several investigators have utilized records from the Wildlife site to analyze site
and pore-pressure responses during the 1987 earthquake. Here we only mention studies
by Zeghal and Elgamal (1994), Vucetic and Thilakaratne (Department of Civil
Engineering, University of California, Los Angeles, 1989), Dobry et al. (1989) and Gu et
al. (1994).
Vucetic and Thilakaratne (1989) and Dobry et al. (1989) performed site response
analyses using the computer program DESRAMOD. Dobry et al. (1989) reported results
of both simplified calculations and site response analyses using the program
DESRAMOD. Only excess pore water pressure predictions were reported in their paper.
Zeghal and Elgamal (1994) utilized the recorded surface and downhole (at 7.5 m
depth) acceleration, and the excess pore water pressures measured by P5 (at 2.9 m depth),
to estimate: (1) Average shear stress-strain history within the top 7.5 m layer of
sediments and (2) effective stress and strain paths at the elevation of piezometer P5 (2.9
m depth). A simple procedure was adopted to obtain direct estimates based on these three
seismic records exclusively. Seismic soil behavior was analyzed by investigating: (1) The
variation of shear stress as a function of average shear strain; (2) the variation of shear
stress and average shear strain as a function of effective vertical pressure (at P5, 2.9m
depth). As acceleration records are available only at the surface and down-hole stations,
linear interpolation was utilized to evaluate shear stress and strain histories within the top
7.5 m layer.
Gu et al. (1994) conducted the stress redistribution analysis under the fully
undrained condition to study the effects of strain-softening behavior of liquefied
materials and the re-consolidation analysis using Biots theory to consider the effects of
176

dissipation of excess-pore water pressures. In the stress redistribution (static) analyses, a


simplified undrained boundary-surface model (Gu el al., 1992) was used to simulate the
behavior of liquefiable soils and the hyperbolic strain-softening model, introduced by
Chan and Morgenstern (1989), was adopted to simulate the soil behavior during collapse
from its peak strength to steady-state strength. The collapse surface proposed by Sladen
et al. (1985) was embedded in the model as a triggering condition for the onset of strainsoftening behavior of liquefied materials. In the re-consolidation analyses, elastic models
were used. The same finite-element mesh for stress re-distribution analyses was used for
the re-consolidation analyses.
In the above procedure the response of soil is somehow predetermined. Beforeliquefaction and post-liquefaction are treated as two separate processes. In such a
procedure, liquefaction and post-liquefaction deformations are triggered by certain preset
criteria. The sensitivity of the triggering criteria to the analysis and the impact of the
uncertainties in the evolution of soil states, which the liquefaction triggering depends on,
have not been addressed in detail. Events before and after liquefaction are essentially two
divided stages of a single continuous process, and a unified procedure is preferable to
treat earthquake response including the post liquefaction deformation. A unified
procedure needs a unified constitutive model, in particular a model that can reproduce all
significant stress-strain responses of granular soils during the entire process of earthquake
motion. In addition, the response of the model should be loading-history dependent, i.e.,
not predetermined by the analysis.
In this paper, a unified, verified, effective stress-based, non-linear, fully coupled,
finite element method-based procedure is utilized to analyze the instrumented Wildlife
site. The constitutive model used in this analysis can simulate flow liquefaction as well as
cyclic mobility. In this model, there is no need to predetermine which soil is flow
liquefaction and which is not, no need to have a rather unnatural liquefaction triggering
mechanism, and no need to separately define the behavior of the material model for
before- and post-liquefaction events.
A.1.6 Verified Numerical Procedure: SUMDES
SUMDES is a computer program to perform dynamic response analyses of Sites
Under Multi-Directional Earthquake Shaking (Li et al., 1992). This procedure is
formulated on the basis of the effective stress principle, vectored motion, transient pore
fluid movement, and generalized material stiffness; therefore, it is capable of predicting
three-directional motions, the pore pressure buildup, and dissipation within the soil
deposits. The overall procedure consists of two computer programs: SUMDES and
TESTMODL. SUMDES performs the site response analyses; TESTMODL is used to
examine the responses of the built-in inelastic models with given model parameters and
loading conditions. This numerical procedure was verified during the VELACS project.
Numerous examples have been published that demonstrate the applicability of the
methodology, which utilizes the numerical procedure and the non-destructive in situ
electrical method of site characterization (Arulanandan et al., 1982; Arulanandan et al.,
1986; Arulanandan et al., 1989; Reyes et al., 1996; Arulanandan et al., 1996; Sivathasan
et al., 1998; Arulanandan and Sivathasan, 1998; Arulanandan et al., 1999).

177

A.1.7 Constitutive Models Used in the Analysis


1. Reduced Order Bounding Surface Hypo-plasticity Model:
The reduced order bounding surface hypo-plasticity model (Li et al., 1992; Li, 1996)
is a special version of the original bounding surface hypo-plasticity model (Wang et al.,
1990). The model has a smaller number of model parameters than the original model (due
to the special loading conditions imposed by site response analysis), but still retains the
capability to simulate liquefaction potential and pore water pressure generation. In the
reduced order bounding surface plasticity model, ten parameters are involved in the site
response analysis (see Table A1). Among the ten, three parameters (b, hp and Rp/Rf) are
either inactive in the given loading conditions or vary within a small range, so they can
be assumed empirically. Two parameters d and kr can be calibrated based on cyclic
strength data obtained from field measurement. Other parameters can be obtained from
established correlations between electrical parameters and required parameters.
2. Modified Reduced Order Bounding Surface Hypo-plasticity Model:
The reduced order hypo-plasticity model is modified according to a recent paper by
Li et al. (1999) to improve the capability in modeling the constitutive behavior at high
strains and most importantly, to incorporate the critical state concept in the sand
response. The modification consists of rendering the phase transformation line a function
of the state parameter , which measures the difference between the current and critical
void ratios at the same mean normal pressure, p, such that when the state parameter is
zero, the phase transformation line becomes identical to the critical state line in q p
space. This idea was originally proposed and applied in another constitutive model by
Manzari and Dafalias (1997). As a result of this modification, the dilatancy depends on
the state in a way which yields a zero value at critical state. This dependence allows a
realistic modeling of the response of a sand in either loose or dense state, or in the
transition from one state to another state. After the modification, dilatancy is not only a
function of stress state but also a function of internal material state. In this model, twelve
parameters (see Table A2) are involved in the site response analysis. Among the twelve,
d parameter can be calibrated based on cyclic strength data obtained from field
measurement as described below. Locating the critical state line in the e-p space is of
critical importance for the determination of the state parameter, . It has been shown that
the critical state line in e-p space can be closely approximated by considering the
relationship between the average shape factor and the position of the critical state line of
sand based on laboratory test data. For the Wildlife site sand, the location of the critical
state line in e-p space is shown in Figure A4 (Arulanandan, 1994; Arulanandan and
Sivathasan, 1995). Other parameters can be obtained from established correlations
between electrical parameters and required parameters (see Arulanandan et al., 2000).

178

A.1.8 Parameter Based on Cyclic Strength Data (d)


The cyclic strength is estimated based on the in situ electrical method (Arulmoli
et al., 1985). For example, cyclic strength curve for electrical parameter A 3 /(Ff ) = 0.26 is
shown on Figure A5. This curve can be simulated by running the utility program
TESTMODL from the program package of SUMDES for a single element response with
other calibrated parameters controlling the development of pore pressure under undrained
conditions:
The calibration procedure to find d is a process of trial and error:
1. Select cyclic stress ratio, with a specified mean normal confining stress;
2. Assume a value of d;
3. Run the utility program TESTMODL for the selected stress level and count
the number of cycles to cause liquefaction (see Figure A6);
4. Compare the results obtained from Step 3 against the given relationship of
stress ratio required to cause liquefaction versus number of cycles to cause
liquefaction. Adjust the d value if needed;
5. Repeat Steps 3 and 4 until satisfactory results are obtained.
Repeat the same procedure for two or more shear stress amplitudes, with a specified
mean normal stress.

179

A.1.9 Non-Destructive Electrical Method of In Situ Characterization


Considerable research, for example the VELACS project, has been undertaken to
verify numerical procedures using laboratory test results and known initial state
parameters. Difficulty in obtaining undisturbed samples and carrying out reliable

180

laboratory tests to avoid disturbance of the sample to calibrate constitutive models seem
to be a major obstacle in advancing the use of fully coupled, effective stress-based,
nonlinear, numerical procedures using elastoplastic constitutive models. The in-situ nondestructive site characterization is necessary to get the initial state parameters and
constitutive model constants representative of the site for the computer simulation of the
site to earthquakes.
It has been shown experimentally by numerous studies conducted by Archie
(1942), Arulanandan and Kutter (1978) and Wyllie and Gregory (1957) that a nondimensional electrical parameter, called the formation factor, is dependent upon the
particle shape, long axis orientation, contact orientation and size distribution, and also
cementation, degree of saturation, void ratio and anisotropy. The formation factor is thus
a function of the fabric of the soil and varies with stress to the extent that the fabric is
altered by the stress. An index, the formation factor F, is experimentally obtained as the
ratio of the conductivity of the pore fluid and the conductivity of the soil sample. Thus,
the formation factor is a non-dimensional parameter. It depends on sand structure,
especially on its elements associated with particle arrangement, particle shape, and
porosity. The dependence of the formation factor on void ratio, particle shape and long
axis orientation has been shown theoretically by Arulanandan and Dafalias (1979), and
Dafalias and Arulanandan (1979). The formation factor F has been shown to be a
valuable directional index parameter when the sand structure is anisotropic. Due to the
above facts, in-situ measurements of formation factor (F) provides a means to
characterize the structure of sands in the field and to obtain empirical correlations
between shape and the anisotropic state of sand characterized in terms of F and the
engineering behavior such as liquefaction characteristics of sands.
The formation factor can be used to quantify and predict the porosity, shape and
anisotropy of sand deposits. An average formation factor F , is given by
F=

( Fv + 2 Fh )
3

(A1)

where, Fv is the vertical formation factor and Fh is the horizontal formation factor.
It has been shown that F is independent of anisotropy caused by the orientation
of preferred particles and is a direct measure of porosity (n), for a given sand. For
practical purposes, an integration technique proposed by Bruggeman (1935) was used by
Dafalias and Arulanandan (1979) to derive an expression for the average formation
factor, F , as a function of porosity (n), and average shape factor ( f ), as
-f
F =n
(A2)
It has been shown both theoretically and experimentally that the shape factor is
directional and depends on porosity, gradation, the particles shape and orientation of the
particles (Dafalias and Arulanandan, 1979; Arulanandan and Kutter, 1978). An
anisotropy index (A), was introduced by Arulanandan and Kutter (1978) as

F
A = v
Fh

0.5

(A3)

181

The anisotropy of sand structure is due to the orientation of individual particles and the
contact orientation.
A.1.10 Method of Analysis
Samples were taken from the silty sand layer at the instrumented site at wildlife
site, California, and a correlation between the formation factor and porosity was
established (see Figure A7).

182

From the non-destructive electrical measurements, the average shape factor and
anisotropy index were calculated. The average formation factor was calculated using insitu non-destructive resistivity measurements in two different directions. Porosity was
obtained from the pre-established relationship (see Figure A7) between formation factor
and porosity using the calculated average formation factor. All the constitutive model
constants and initial state parameters were calculated using the average formation factor,
average shape factor and anisotropy index as detailed in Chapter 6 (also see Arulanandan
et al. 2000).
A.1.11 Comparison of Shear Wave Velocity Measurements
A correlation between K2max and the electrical parameter F /( A f ) 0.5 is given by
Arulanandan et al. (1983) and is used to obtain K2max. K2max depends on void ratio, strain
amplitude, geological age of the sand mass and in situ stresses. The mean effective
confining pressure, p, can be calculated using site configuration and unit weight of the
soil mass. The maximum shear modulus can be obtained using
Gmax = 1000 K 2 max ( p )

0.5

(A4)

where p and Gmax are given in pounds per square foot. Using known maximum shear
modulus and density of soil mass, one can obtain the shear wave velocity
Vs =

G max

(A5)

The comparison of the shear wave velocity values calculated using above
procedure with the measured crosshole shear wave velocity values as shown in Figure
A3c demonstrate a reasonable agreement.
A.1.12 Analysis Results

The numerical simulation of the Wildlife instrumented site, California was


performed by the computer program SUMDES. In the analysis, multi-directional shaking
was applied to the soil deposits. The input motions and their response spectrum are
shown in Figure A8. The reduced order bounding surface hypo-plasticity model was used
to simulate the cohesive soil behavior and the modified reduced order bounding surface
hypo-plasticity model was used to simulate the non-cohesive soil behavior. Calibrated
constitutive model constants and state parameters are provided in Tables A1 and A2.
From the specification for the instrumented array at Wildlife site, California,
accelerometers were placed at two different depths. At each depth three different
directional components of acceleration were reported. In the present analysis, all the
predictions and comparisons were made at the two different depths in three different
directions.
Recorded and evaluated accelerations at different depths in different directions are
presented in Figures A9a and A9b. Response spectra from analysis results and recorded

183

accelerations at the ground surface are compared in Figure A10. Note that recorded and
SUMDES results compare favorably.

184

Recorded and evaluated excess pore water pressures at different depths are shown
in Figures A11a and A11b, respectively. Field pore water pressure measurements were
reported for comparison with the evaluated pore water pressure during the earthquake
shaking at four different depths. The maximum pore water pressure ratio variation with
depth is shown in Figure A12. It shows that the zone of high pore water pressure ratio in
excess of 90% extended in the first 6.8m below water table. Pore pressure time histories,
in terms of pore-pressure ratio, recorded and evaluated are shown in Figures A13a and
A13b, respectively. Vertical settlement variation with time at the ground surface is shown

185

in Figure A14a. Reported average vertical settlement at the ground surface is about 5 cm.
Evaluated lateral displacement variation with depth is shown in Figure A14b. Reported
maximum lateral displacement 230 mm occurred at the waterfront (Dobry et al. 1989).
The observed lateral displacements were generally large near the waterfront and
decreased with distance inland. These results provide validation to the computational
analysis.

186

187

188

189

190

A14a

191

A14b

A.1.13 Conclusions

The pore pressure, acceleration and liquefaction-induced deformation time


histories were simulated using the nonlinear, fully coupled, effective stress based method
of analysis. The evaluated acceleration and the excess pore pressure time histories are in
close agreement with the recorded acceleration and the excess pore pressure time
histories at the Wildlife site. This analysis shows that liquefaction occurred in the silty
sand layer and agrees with observed liquefaction phenomenon at the Wildlife site,
California. Evaluated vertical settlement and lateral displacements are in close agreement
with the observed behaviors at the Wildlife site.
It has been demonstrated that it is possible to evaluate the dynamic response of a
level ground instrumented site using verified numerical procedure and input properties
such as initial state parameters (porosity, coefficient of earth pressure at rest, hydraulic
conductivity) and constitutive model constants (slope of the rebound line, friction angle,
maximum shear modulus, slope of the critical state line in ec-p space, hardening
parameters, etc.) representative of the site obtained by non-destructive in situ testing and
shear wave velocity measurements. The comparison of the crosshole shear wave velocity
values obtained by wave propagation method with the shear wave velocities obtained by
the electrical method are shown to be in close agreement.
A.1. 14 Acknowledgments

The authors are indebted to Dr. X. S. Li and Prof. Y.F. Dafalias for reviewing the
manuscript and providing valuable comments.

192

A.1.15 References

Archie, G. E. (1942). The Electrical Resistivity Log as an Aid in Determining Some


Reservoir Characteristics, Transactions, American Institute of Mining, Metallurgical
and Petroleum Engineers, Vol. 146, pp.54-61.
Arulanandan, K. (1994). In-Situ Soil Characterization, Location of Critical Void Ratio
Variation with Mean Normal Stress by In-Situ Testing, UC Davis Report and lecture
presented at the Kyoto University, Kyoto, Japan.
Arulanandan, K., Anandarajah, A., and Meegoda, N. J. (1983). Soil Characterization for
Non-Destructive In-Situ Testing, Symposium Proceedings part 2, The interaction of
Non-Nuclear Munitions with Structures, US Air Force Academy, Colorado, May 10 13.
Arulanandan, K., Arulmoli, K. and Dafalias, Y. F. (1982). In Situ Prediction of Dynamic
Pore Pressures in Sand Deposits, International Symposium on Numerical Models in
Geomechnics, Zurich, Switzerland, 13-17 September, pp. 359-367, A. A.
Balkema/Rotterdam.
Arulanandan , K., and Dafalias, Y. F. (1979). Significance of Formation Factor in Sand
Structure Characterization, Letters in application and Engineering Sciences, Vol. 17,
pp.109-112.
Arulanandan, K. and Kutter, B. L. (1978). Directional Structure Index Related to Sand
Liquefaction, Proceedings of the June 19-21, Specialty Conference on Earthquake
Engineering and Soil Dynamics, ASCE, Pasadena, California, pp. 213-229.
Arulanandan, K., Li, X. S., Paulino, G. H. and Sivathasan, K. (1996). Dynamic
Response of Saturated Level Ground Sites using Verified Numerical Procedure and InSitu Testing, Proceedings of the National Science Foundation and California
Transportation Dept. Sponsored Workshop/Conf. on Application of Numerical
Procedures in Geotech. Earthquake Engrg., Univ. of California, Davis.
Arulanandan, K., Li, X. S. and Sivathasan, K. (2000). Evaluation of Dynamic Site
Response and Deformation of Instrumented Site-Kobe, Journal of Geotechnical and
Geoenvironmental Engineering, ASCE, (in press).
Arulanandan, K., Muraleetharan, K. K., Dafalias, Y. F., Shinde, S. B., Kaliakin, V. N.
and Herrmann, L. R. (1989). Pore Pressure and Lateral Stresses Using In Situ
Properties, Proceedings of the Twelfth International Conference on Soil Mechanics and
Foundation
Engineering,
Rio
De
Janerio,
13-18
August,
A.
A.
Balkema/Rotterdam/Brookfield.
Arulanandan, K. and Sivathasan, K. (1995). In-Situ Prediction of Critical Void Ratio Vs
Mean Normal Pressure, Technical Report, University of California, Davis.

193

Arulanandan, K. and Sivathasan, K. (1998). "Evaluation of Site Response and


Deformation of Instrumented Bridge Sites Subjected to Large Magnitude Earthquakes,"
Preliminary Report to Dept. of Transportation, California.
Arulanandan, K., Yogachandran, C., Meegoda, N. J., Liu Ying and Shi Zhauji (1986).
Comparison of the SPT, CPT, SV and Electrical Methods of Evaluating Earthquake,
Proceedings of In Situ 86, GT Div., ASCE, June 23-25, Blacksburg, VA.
Arulmoli, K., Arulanandan, K. and Seed, H. B. (1985). New Method for Evaluating
Liquefaction Potential, Journal Geotechnical Engineering Division, ASCE, 111(1), 95114.
Bennett, M. J., McLaughlin, P. V., Sarmieto, John and Youd, T. L. (1984). Geotechnical
Investigation of Liquefaction Sites, Imperial Valley, California, Open File Report 84252, U. S. Geological Survey, Denver, Colorado, 1-103.
Bruggeman, D. A. G. (1935). "Berechung Verschiedenez Physikalischer Konstanten Von
Heterogenen Substanzen," Ann. Phys. Lpz. 5, Vol. 24, pp. 636.
Chang, C. S. and Misra, A. (1990). Packing Structure and Mechanical Properties of
Granulates, J. of Engineering Mechanics, ASCE, Vol. 116, No. 5, pp. 1077-1093.
Dafalias, Y. F. and Arulanandan, K. (1979). Electrical Characterization of Transversely
Isotropic Sands, Archives 0f Mechanics, Warsaw, 31(5), 723-739.
Dobry, R., Elgamal, A. W., Baziar, M. and Vucetic, M. (1989). Pore Pressure and
Acceleration Response of Wildlife Site During The 1987 Earthquake,
ERTEC WESTERN, INC. (1982). Investigation of: Correlation Between Electric
Resistivity Measurements and Crosshole Shear Wave Velocity Measurements,
Preliminary Report, The Earth Technology Corporation, 3777 Long Beach Boulevard,
Long Beach, CA 90807. December, ERTEC I. D. 82-236-(1-10)
Gu, W. H., Morgenstern, N. R. and Robertson, P. K. (1992). Progressive Failure of
Lower San Fernando Dam, J. Geotech. Engrg., ASCE, 119(2), 333-348.
Gu, W. H., Morgenstern, N. R. and Robertson, P. K. (1994). Postearthquake
Deformation Analysis of Wildlife Site, J. Geotech. Engrg., ASCE, Vol. 120, No. 2,
February.
Holzer, T. L., Youd, T. L. and Hank, T. C. (1989 a). Dynamic of Liquefaction During
the 1987 Superstition Hills, California Earthquake, Science, 244, pp. 1-116.
Holzer, T. L., Youd, T. L. and Bennett, M. J. (1989 b). In Situ Measurements of Pore
Pressure Build-Up During Liquefaction, Proceedings of the 20th Joint Meeting of the U.
S. Japan Cooperative Program in Natural Resources Panel on Wind and Seismic

194

Effects, Published by the National Institute of Standards and Technology, Gaithersburg,


MD 20899.
Li, X. S. (1996). Reduced-Order Sand Model for Ground Response Analysis, Journal
of Engineering Mechanics, ASCE, 122(9), pp. 872-881.
Li, X. S., Dafalias, Y. F. and Wang, Z. L. (1999). A Critical-State Hypo-plasticity Sand
Model with State Dependent Dilatancy, Canadian Geotechnical Journal (accepted).
Li, X. S., Wang, Z. L., and Shen, C. K. (1992). SUMDES: A Non Linear Procedure for
Response Analysis of Horizontal-Layered Sites Subjected to Multi-Directional
Earthquake Loading, Department of Civil and Environmental Engineering, University of
California at Davis.
Manzari, M. T. and Dafalias, Y. F. (1997). A Critical State Two-Surface Plasticity
Model for Sands, Geotechnique, Vol. 47, No. 2, pp. 255-272.
Reyes, C., Arulanandan, K., Steve Mahnke, Chad Baker and Sivathasan, K. (1996).
Fully Coupled Effective Stress Based Analysis to Investigate the Consequences of Soil
Liquefaction at Mosher Slough for FEMA Project, Report to Kleinfelder and Associates,
Stockton, California.
Sivathasan, K., Paulino, G. H., Li, X. S. and Arulanandan, K. (1998). Validation of Site
Characterization Method for the Study of Dynamic Pore Pressure Response,
Geotechnical Special Publication No. 75, Volume one, Geotechnical Earthquake
Engineering and Soil Dynamics III, ASCE, Seattle, Washington.
Sladen, J. A., DHollander, R. D. and Krahn, J. (1985) The Liquefaction of Sand, a
Collapse Surface Approach, Canadian Geotechnical Journal, 22(4):564-578.
Vucetic, M. and Thilakaratne, V. (1989). Liquefaction at the Wildlife Site-Effect of Soil
Stiffness in Seismic Responses, Proceedings of the 4th International Conference on Soil
Dynamics and Earthquake Engineering, Mexico City, Mexico.
Wang, Z.L., Dafalias, Y.F. and Shen, C.K. (1990). Bounding Surface Hypo-plasticity
Model for Sand, Journal of Engineering Mechanics, ASCE, Vol.116, No. 5, May,
pp983-1001.
Wyllie, M. R. J. and Gregory, A. R. (1953). Formation Factors of Unconsolidated
Porous Media, Influence of Particle Shape and Effect of Cementation, Petroleum
Transactions, American Institute of Mining, Metallurgical and Petroleum Engineers, Vol.
198, pp.103-109.
Youd, T. L. and Bartlett, S. F. (1988). US Case Histories of Liquefaction-Induced
Ground Displacement, First US-Japan Workshop on Liquefaction, Large Ground
Deformation and their Effect on Lifeline Facilities, Tokyo, Japan.

195

Zeghal, M. and Ahmed-W. Elgamal (1994). Analysis of Site Liquefaction Using


Earthquake Records, Journal of Geotechnical Engineering, ASCE, Vol. 120, No. 6,
June.
A.1.16 Notation

The following symbols are used in this paper:


A
b
d
F

Fh
Fv
f

Gmax
Go
hp
hr
Ko
K2max
k
kr
m
n
n
q
Rp/Rf

ec
p

= anisotropy index
= parameter affecting the shape of the stress paths of the virgin shear
loading
= parameter which characterizes the rate of the effective mean normal stress
change caused by shear unloading
= average formation factor
= horizontal formation factor
= vertical formation factor
= average shape factor
= maximum shear modulus
= constant related to shear modulus
= parameter controlling the amount of shear strain increment due to the
change of the maximum effective mean normal stress in the loading
history
= parameter which characterizes the relationship between shear modulus
and shear strain magnitude
= coefficient of earth pressure at rest
= modulus related to maximum shear modulus
= coefficient of hydraulic conductivity
= parameter which characterizes the amount of the effective mean normal
stress change caused by shear loading
= model constant
= model constant
= porosity
= deviatoric stress
= ratio of stress ratio at phase transformation state and stress ratio at failure
state
= friction angle
= slope of the rebound line in e-ln(p) space
= constant
= state parameter
= critical void ratio
= pressure

196

A.2 BEFORE-THE-EVENT PREDICTION OF THE RESPONSE OF TREASURE


ISLAND INSTRUMENTED SITE TO EXPECTED EARTHQUAKES
A.2.1 Objectives

Before-the-event evaluation of the response of our infrastructure systems to


earthquakes is a desirable objective. By predicting infrastructure response, we provide the
basis for appropriate, safe and economical retrofit measures of existing facilities and safe
and economical design of new facilities. Once a verified numerical procedure and in-situ
properties representative of a site are available, computer simulation of earthquake effects
can be performed.
A.2.2 The Treasure Island Site

Treasure Island is a 400-acre (162-hectare) man-made island formed between


1936 and 1937 as the site of the 1939 Golden Gate International Exposition in
California's Bay Area. Since the 1940s, the island has served as a U.S. Naval installation.
Figure A15 shows the location of the Treasure Island in relation to the San Francisco Bay
Area and the major faults in the region.
The instrumentation site evaluated for this project is situated on a 1.25-acre
parcel of ground, which includes the island's fire station. The general stratigraphy from
surface to bedrock at this site was determined in a study by the Electric Power Research
Institute (EPRI), the California Division of Mines and Geology (CDMG), the United
States Geological Survey (USGS), and the University of California at Davis. At that time,
a 104 m borehole was completed for installation of a bedrock accelerometer. Field
classification, limited standard penetration testing, and soil sampling were carried out.
Approximately two thirds of the island was constructed on the sandy shoals adjacent to
Yerba Buena Island. The remaining portion of the island was built directly on Young Bay
Mud. The location of the fire station site suggests that the instrumentation site is located
above the sandy shoal area. The island was constructed by pumping sand fill behind a
perimeter rock starter dike and building subsequent levels of dike over previously placed
hydraulic fill (Lee and Praszker, 1969); this procedure is commonly known as the
upstream method of hydraulic filling. Sandy fill was dredged from the bay and placed, in
thicknesses ranging from 9.4 m to 11.6 m, to form an island 4.0 m above the mean lower
low water mark. Hydraulic fill at the fire station site is fine silty sand containing varying
amounts of clay and shells. Several feet of settlement have occurred in the loosely placed
fill and surcharged underlying sediments over the past 50 years. Most of this settlement
took place during the first seven years after construction (Lee and Praszker, 1969).

197

Figure A15
Major Causative Faults in Treasure Island (TI) Vicinity (modified from Housner
and Penzien, 1991), Probability of Magnitude 7+ Earthquake (Working Group on
California Earthquake Probabilities, 1990). Rectangle T1 Indicates Instrumented
Area

198

A.2.3 Soil Profile of Treasure Island Site

The first 13.6 m at the fire station site is composed of tan and grey sands. The
groundwater level was measured at approximately 1.4 m below the surface of this
material in August 1991. Beginning at the surface (see Figure A16), a thin layer of
dumped fill overlies 7.7 m of tan and grey hydraulic fills. Below these fills is a thin layer
of naturally deposited shoal material. This is underlain by a thin layer of sandy clay to
clayey sand, which transitions to a dark grey silty clay layer. The dark grey clay layer
shown between 13.6 m and 29.2 m in Figure A16 contains several shelly layers and is
part of the Young Bay Mud formation. The fine to coarse sand layer underlying the
Young Bay Mud is made up of four sublayers belonging to the San Antonio formation.
Small, uniformly sized shell particles in sublayer A at the fire station site suggest an
association with the aeolian parts of the Merritt sands. Sublayer B does not fit
descriptions of units found in the Posey sands, which often underlies the Merritt sands,
but is somewhat similar to a silty and clayey unit belonging to the Merritt sands. The
large shell fragments and many rounded pebbles in sublayer C mark a clear distinction
from the upper two layers (A and B). Sublayer C fits well into what Trask and Rolston
term the D-2 unit of the San Antonio formation. Trask and Rolston's San Antonio forms
the lowest portion of the larger San Antonio formation, suggesting an association with
sublayer D. The grey-green clay layer between 42.4 m and 75.6 m is part of the Old Bay
Clay. Between the Old Bay Clay and bedrock is 5.5 m of gravelly sand, followed by 10 m
of clay. These are parts of the Alameda formation.

199

Figure A16
Schematic Soil Profile with Correlations of Layers to Geological Formations.
A.2.4 Analysis

1. Initial and constitutive model constants were evaluated using the electrical parameters
F, f , and A . Using the values of porosity n, coefficient of earth pressure at rest Ko,
and shear modulus Gmax, and the relationships between the electrical parameters
F, f , and A and the above properties, electrical parameters F, f , and A were
evaluated.
2. Verified numerical code SUMDES was used to predict, before-the-event the site's
response to earthquake loading. Vertical heterogeneity observed at the lower sand
layer (29.2 m to 42.4 m) was analyzed; the influence of the heterogeneity on the
lateral deformation was demonstrated. The influence of earthquake magnitude on pore

200

pressures, lateral deformation, and settlement were shown using the Kern County and
Foster City Earthquakes (see No. 3).
3. Kern County CA 1952 07 21 0453 PDT Earthquake of Magnitude 7.5 was scaled for
the distance from the causative fault using scaling laws, and used in the analysis.
Similarly, the Foster City Earthquake of Magnitude 7.0 was used in the analysis.
A.2.5 Soil Properties

The following tables and figures (from Pass, 1994) were used to obtain the values
of porosity n, coefficient of earth pressure at rest Ko, and shear modulus Gmax for major
layers.

201

202

Table A3 Average layer properties from thin-walled tube specimens

203

Table A4 Crosshole test results

204

Table A4 (continued) Crosshole test results

Figure A17
Profiles of G0 Calculated Using Crosshole Shear Wave Velocities

205

Figure A18
Profiles of Lateral Stress Coefficient (K0) from Self-Boring Pressuremeter Testing
and Dilatometer Testing

206

A.2.6 Verified Numerical Procedure, SUMDES

Model Parameters

SUMDES is a computer program to perform dynamic response analyses of Sites


Under Multi-Directional Earthquake Shaking (Li, et. al, 1992). The basic assumptions
made in the formulation are listed below:
1. The site is horizontally layered and extends infinitely in horizontal directions.
2. The ground surface is free of stress.
3. Water flow is laminar and Darcys law holds.
4. The bottom boundary is impermeable.
5. Soil layers below the water table are fully saturated.
6. Waves travel along the vertical direction only.
The procedure is formulated on the basis of the effective stress principle, vectored
motion, transient pore fluid movement, and generalized material stiffness; therefore, it is
capable of predicting three-directional motions and pore pressure buildup and dissipation
within soil deposits. The overall procedure consists of two computer programs: SUMDES
and TESTMODL. SUMDES performs site response analyses; TESTMODL examines
the responses of the built-in inelastic models with given model parameters and loading
conditions. This numerical procedure was verified during the VELACS project.
Numerous examples have been published demonstrating the applicability of using the
numerical procedure SUMDES and a non-destructive, in situ, electrical method of site
characterization (Reyes, et al., 1996; Arulanandan, et al., 1996; Sivathasan, et al., 1998;
and Arulanandan and Sivathasan, 1998). Examples of the combining in situ, nondestructive, site characterization using the formation factor method, and SUMDES are
described in Section A.1.
The required model parameters are

1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.

Friction Angle
n Porosity
K0 Earth pressure coefficient at rest
k Coefficient of hydraulic conductivity
G0 Initial elastic shear modulus
The slope of the virgin compression line
The slope of the rebound line
hr The parameter that characterizes the relationship between shear modulus and
shear strain magnitude.
b A parameter affecting the shape of the stress path of the virgin shear loading
Rp/ Rf The ratio of the slope of the phase transformation line to the slope of the
failure line.
d characterizes the rate of the effective mean normal stress change caused by shear
unloading.
kr characterizes the amount of the effective mean normal stress change caused by
shear loading.

207

The Procedure to Calibrate Kr and d (Trial And Error)

Figure A19
Influences of d, Kr, b, and Rp/Rf

a. For a sandy soil with known A 3 / F f , select two or more shear stress amplitudes
first, with a specified mean normal confining pressure.
b. Assume a pair of parameters (kr and d).
c. Run utility program TESTMODL for the two or more stress levels. Count the number
of cycles required to cause liquefaction for each of the shear stress amplitudes.
d. Check the results obtained from (c) against the given relationship of stress ratio versus
number of cycles to liquefaction. Adjust the two parameters (kr and d) if needed.
e. Repeat steps (c) and (d) until satisfactory results are obtained.
Parameter Descriptions
- Friction angles

Using electrical measurements one can obtain f . With f , the slope of the
critical state line (M) in the q -p' space can be obtained as shown in the paper by
Arulanandan et al. 1994 (see Figure A20).
Moreover, from;

208

M =

(6 sin )
(3 sin )

the critical state friction angle () can be calculated.

Figure A20
Relationship Between the Average Shape Factor and the Slope of the Critical
State Line (M) in q p Space

n -Porosity
Using electrical measurements one can obtain porosity using an established
relationship between formation factor and porosity for a particular soil as shown in
Arulmoli et al.1985 (see Figure A21).

209

Figure A21
Vertical, Horizontal and Average Formation Factor versus Porosity
Relationships for Monterey 0/30 Sand Prepared by Three Different Methods

K0 -The earth pressure coefficient at rest


Using electrical measurements one can obtain K0 from pre-established correlation
between the coefficient of earth pressure at rest and the electrical index A4 f (Meegoda
and Arulanandan, 1986).

210

Figure A22
The Relationship Between the Coefficient of Lateral Earth Pressure at Rest
and the Electrical Index, A4 f (Meegoda and Arulanandan, 1986)

k -The coefficient of hydraulic conductivity


The specific surface area given as surface area per unit volume of soil particles
may be calculated as follows.
W W W1
100 Wn 1

+ ......... +
6 1 + 2
d
d
d
2
n

1
SSA =
100
where W1, W2, ..etc, are the percentages passing; and di is the average diameter
between Wi and Wi-1.
Arulanandan and Muraleetharan, 1988, have shown the linear relationship
between hydraulic conductivity and an electrical index above as shown in Figure A23.

211

Figure A23
Correlation between Vertical Hydraulic Conductivity of Sands and Silty
Sands and an Electrical Index

Go
This parameter is the modulus coefficient defining the initial (maximum) elastic
shear modulus Gmax using the following equation (Li, et al., 1992):

K2max depends on void ratio, strain amplitude geological age of the sand mass and in situ
stresses. P is the mean effective confining pressure, in pounds per square foot and Gmax is
given in pound per square foot. A correlation between K2max and the electrical parameter
F /( A f )0.5 is given by Arulanandan, et al. (1983) and is used to obtain K2max.

212

Figure A24
Correlation between K2 max and Electrical Parameter, F /( A f )1 / 2

A correlation between the compression index (CC) of sands (i.e., non-clay


minerals) and the electrical index Af / F , which is a polynomial of degree 2, is given in
Figure A25 (Harvey, 1981).

Figure A25
Correlation Between Compression Index and Electrical Index Af / F for
Sands (after Harvey, 1981)

213


The slope of the rebound line in e -In(p') space is defined by (Li, et al., 1992)

hr

The parameter, hr , characterizes the relationship between shear modulus and shear
strain magnitude. The value of hr can be determined based on the standard modulus
reduction curve (Li, et al., 1992) or the secant shear modulus measured at a control shear
strain. The parameter hr, so calibrated, is in general a function of density (Li, et al., 1999).
b

This is a parameter affecting the shape of the stress paths of the virgin shear
loading. A typical value of b is 2.0.
kr

The parameter, kr characterizes the amount of the effective mean normal stress
change caused by shear loading. kr value also affects the liquefaction resistance. Since do
is a constant kr is not an independent parameter (Li, et al., 1999). kr must satisfy the
calibrated value of do.
d

The parameter, d, characterizes the rate of the effective mean normal stress
change caused by shear unloading. This parameter is the major one controlling the
liquefaction resistance of a material.
Stress Ratio Required to Cause Liquefaction

If A3 /( F f ) can be obtained for a sand deposit, then the relationship between


cyclic stress ratio to cause liquefaction and number of cycles can be established using the
results shown in Figure A26.

214

Figure A26
Correlation Between Cyclic Stress Ratio Required to Cause Initial
Liquefaction and Number of Cycles for Different Values of Electrical Index,

A3 / (F f )

A.2.7 Treasure Island Site Calculations

Layer 1: Tan fill (0 4.6 m)


From Table A3,
Void Ratio, e = 0.68
Therefore, n = 0.40
From Figure A18,
The earth pressure coefficient at rest, K0 = 0.7
Using Figure A22, log A4 f = 0.14
A4 f = 1.38

(A6)

From Table A4,


Shear Modulus Gmax = 34.9 MPa (728900 pounds/sq. ft)
From Table A3,
Density = 1.92
At the center of the layer
' = 2.3*1.92*9.81
= 43.32 kPa (904.8 pounds/sq. ft)
Using,

215

K2max = 24.2

Using Figure A24,


F /( A f )0.5 = 2.6

(A7)

Using,
F = n f
F = 0.4 f

(A8)

Solving equations (A6), (A7), and (A8)


F = 2.82
f = 1.13
A = 1.08
Using Figure A20,
For log ( f ) = 0.05
log (M) = 0.04
M = 1.1
(6 sin )
Using M =
(3 sin )
= 27.60

Assuming SSA = 1 m2/g


F 3 / f

(1 F ) (SSA)
1 / f

=5

Using Figure A23,


Permeability k = 4 * 10-2 cm/s
Using Figure A25,
For Af / F = 0.43,
Compression index (CC) = 0.015
Therefore = 0.015
= 0.003
Using HR sub function of SUMDES,
hr = 0.46
Using A3 /( F f ) = 0.4
Using Figure A26,
Stress Ratio required to cause Liquefaction = 0.1 (for 10 cycles)
Using Calibration (TESTMODL),

216

kr = 0.5
d = 0.72
Similar calculations were done for other layers. The following table summarize the state
and model parameters.
Table A5 State and Model Parameters

A.2.8 Analysis Results

The numerical simulation of the instrumented site at Treasure Island was


performed by the computer program SUMDES. In the analysis, multidirectional shaking
was applied to the soil deposits. The input motion of the Kern County Earthquake and
evaluated responses of the pore pressures, lateral deformation, settlement, and surface
motions and response spectra are shown in Figures A27 though A33. The input motions
and evaluated responses for the Foster City Earthquake are shown in Figures A34
through A39.

217

Taft Kern County Earth Quake

Horizontal Direction 1

Horizontal Direction 2

Vertical
Figure A27
Kern County Earthquake Input Motions

218

Displacement (m)

Depth (m)

-0.50

0
0.00
-20
-40
-60
-80

0.50

1.00
With vertical
Hetrogeniety
Without Vertical
Hetrogeniety

Evaluated Lateral Displacement with Depth

Figure A28
Evaluated Lateral Displacement with Depth
Simulated Kern County Earthquake

219

Variation of Pore Water Pressure Ratio with


Depth

Excess Pore Water


Pressure Ratio

1.2
1

5 m Depth

0.8

10 m Depth

0.6

26.5 m Depth

0.4

36.5 m Depth
62 m Depth

0.2
0
0

20

40

60

Time (Sec)

Figure A29
Variation of Pore Water Pressure with Depth
Simulated Kern County Earthquake
Predicted Vertical Settlement
0.000

Settlement (m)

-0.005 0

10

20

30

40

50

-0.010
-0.015
-0.020
-0.025
-0.030
-0.035
-0.040

Time (Sec)

Figure A30
Predicted Vertical Settlement
Simulated Kern County Earthquake

220

Stress (Pa)

Stress - Strain Plot at 40 m Depth


( Liquefied Element)

60000
40000
20000
0
-0.05 -200000.00
-40000
-60000

0.05

0.10

0.15

Strain (%)
Figure A31
Stress-Strain Plot at 40 m Depth (Liquefied
Element)
Simulated Kern County Earthquake
Acceleration at the Surface N00E Component

Acc (m/s )

1.0
0.0
0

10

20

30

40

50

60

50

60

50

60

-1.0
Time (Sec)

Acceleration at the Surface N90E Component

Acc (m/s2)

2.0
1.0
0.0
-1.0 0

10

20

30

40

Time (Sec)

Acc (m/s )

Acceleration at the Surface UP Component

1.0
-4.0

10

20

30

40

Time (Sec)

Figure A32
Evaluated Surface Accelerations
Simulated Kern County Earthquake

221

Acc (m/s )

Response Spectrum for N00E Component Damping 5%


4
3
2
1
0
0.10

1.00

10.00

Period (Sec)

Response Spectrum for N90E Component Damping 5%

Acc (m/s

2)

6
4
2
0
0.10

1.00

10.00

Period (Sec)

Acc (m/s )

Response Spectrum for UP Component Damping 5%


20
15
10
5
0
0.10

1.00

10.00

Period (Sec)

Figure A33
Evaluated Surface Response Spectra
Simulated Kern County Earthquake

222

FOSTER CITY EARTHQUAKE

Horizontal Direction 1

Horizontal Direction 2

Vertical
Figure A34
Foster City Earthquake Input Motions

223

90
80
70
60
50
40
30
20
10
0
0

100

200

300

400

500

600

Time (Sec)

Figure A35
Excess Pore Water Pressure at a Depth of 5m
Simulated Foster City Earthquake

Excess Porewater
Pressure Ratio

Excess Porewater Pressure at 26.5m Depth

1.5
1.0
0.5
0.0
0

20

40

60

Time (Sec)

Excess Porewater Pressure at 10m Depth


Excess Porewater
Pressure Ratio

Excess Pore Water Pressure


(KPa)

Excess Pore Water Pressure Variation with Time (at 5m Depth)

1.5
1.0
0.5
0.0
0

20

40

60

Time (Sec)

Figure A36
Excess Pore Water Pressure Ratios at Different Depths
Simulated Foster City Earthquake

224

Excess Porewater
Pressure Ratio

Excess Porewater Pressure at 36.5m Depth

1.5
1.0
0.5
0.0
0

20

40

60

Time (Sec)

Excess Porewater
Pressure Ratio

Excess Porewater Pressure at 62m Depth

1.5
1.0
0.5
0.0
0

20

40

60

Time (Sec)

Figure A36 (Continued)


Excess Pore Water Pressure Ratios at Different Depths
Simulated Foster City Earthquake

225

Depth (m)

Predicted Lateral Displacement

-0.85

0
-10
-20
-30
-40
-50
-60
-70
-80

0.15

1.15

2.15

With Vertical
Hetrogeniety
Witout Vertical
Hetrogeniety

Displacement (m)
Figure A37
Predicted Lateral Displacements
Simulated Foster City Earthquake

Predicted Vertical Settlement

Settlement (m)

0.000
-0.001

10

20

30

40

50

60

-0.002
-0.003
-0.004
-0.005
-0.006
-0.007

Time (Sec)

Figure A38
Predicted Vertical Settlement
Simulated Foster City Earthquake

226

Stress Strain Plot of Liquefied Element


(at 40m Depth)

Stress (KPa)

100
50
0
0.00

-0.10

0.10

0.20

0.30

0.40

-50
-100
Strain (%)

Figure A39
Stress-Strain Plot for a Liquefied Element at 40 m Depth
Simulated Foster City Earthquake
A.2.9 Summary

Using the values of porosity n, Coefficient of earth pressure at rest Ko, and Shear
Modulus Gmax and the relationships between the electrical parameters F , f , and A and
the above properties, the electrical parameters F , f , and A were evaluated. Initial and
Constitutive model constants were evaluated using the calculated electrical parameters.
Verified numerical code SUMDES was used to predict before the event evaluation of the
response of the site to earthquake loading.
A.2.10 References

Arulanandan, K. and Sivathasan, K. (1998). "Evaluation of Site Response and


Deformation of Instrumented Bridge Sites Subjected to Large Magnitude Earthquakes,"
Preliminary Report to Dept. of Transportation, California.
Arulanandan, K., Anandarajah, A. and Meegoda, N. J. (1983) Soil Characterization for
Non- Destructive In Situ Testing, Symposium Proceedings Part 2: The Interaction of
Non-Nuclear Munitions with Structures, U.S. Air Force Academy, Colorado, pp. 69-75.

227

Arulanandan, K., and Muraleetharan, K. K. (1988) Level Ground Soil Liquefaction


Analysis using In-Situ Properties: I and II, Journal of Geotechnical Engineering, ASCE,
Vol. 114, No. 7.
Arulanandan, K., Li, X. S., Paulino, G. H. and Sivathasan, K. (1996). Dynamic
Response of Saturated Level Ground Sites using Verified Numerical Procedure and InSitu Testing, Proceedings of the National Science Foundation and California
Transportation Dept. Sponsored Workshop/Conf. on Application of Numerical
Procedures in Geotech. Earthquake Engrg., Univ. of California, Davis.
Arulanandan, K., Yogachandran, C., and Rashidi, H. (1994). Dielectric Conductivity
Methods of Soil Characterization, Geophysical Characterization of Sites, Volume
Prepared by ISSMFE Technical Committee #10 for the XIII International Conference on
Soil Mechanics and Foundation Engineering, New Delhi, India, pp. 81-90.
Arulmoli, K., Arulanandan, K., and Seed, H. B. (1985). New Method for Evaluating
Liquefaction Potential, Journal of the Geotechnical Engineering Division. ASCE, Vol.
111, No. 1, pp. 95-114.
Harvey, S. J. (1981) Electrical Characterization of Sand Compressibility, Thesis
presented in to the University of California, at Davis, California, in partial fulfillment of
the requirements for the degree of Master of Science.
Li, X. S., Dafalias, Y. F., and Wang, Z. L. (1999). A Critical-State Hypo-Plasticity Sand
Model with State Dependent Dilatancy, Canadian Geotechnical Journal(accepted).
Li, X. S., Wang, Z. L., and Shen, C. K. (1992). SUMDES: A Non Linear Procedure for
Response Analysis of Horizontal-Layered Sites Subjected to Multi-Directional
Earthquake Loading, Department of Civil and Environmental Engineering, University of
California at Davis.
Meegoda, N. J., and Arulanandan, K. (1986) Electrical Method of Predicting In Situ
Stress State of Normally Consolidated Clays, Proceedings of the In Situ 86, GT
Division, ASCE, Blacksburg, Virginia, June 23-25, pp. 794-808.
Pass, D.G. (1994). "Soil Characterization of the Deep Accelerometer Site at Treasure
Island, San Francisco, California," Thesis submitted to the University of New Hampshire
in partial fulfillment of the requirements for the degree of Master of Science.
Reyes, C., Arulanandan, K., Mahnke, S., Baker, C., and Sivathasan, K. (1996). Fully
coupled effective stress based analysis to investigate the consequences of soil
liquefaction at mosher slough for FEMA project.
Sivathasan, K., Paulino, G. H., Li, X. S., and Arulanandan, K. (1998).Validation of site
characterization method for the study of dynamic pore pressure response. Proc.,
Geotech. Earthquake Engrg. and Soil Dyn. III, ASCE, Reston, Va., 469481.

228

APPENDIX B
DERIVATIONS
B.1 DOUBLE-LAYER THEORY
B.1.1 Double Layer Equations
The concentration of ions (ions/m3) of type i, ni, in a force field at equilibrium is
given by the Boltzmann equation
E Ei
ni = ni 0 exp i 0

kT

(B1)

where the subscript 0 represents the reference state, taken to be at a large distance from
the surface, E is the potential energy, T is temperature (0K), and k is the Boltzmann
constant (1.38 x 10-23 J0K-1).
For ions in an electric field, the potential energy is
E = vi e

(B2)

where vi is the ionic valence, e is the electronic charge (1.602 x 10-19 coulomb), and is
the electrical potential at the point. Potential varies with distance from a charged surface
in the manner shown by Figure B1. In clays, is negative because of the negative
surface charge. The potential at the surface is designated as 0. As Ei0 = 0, because = 0
at a large distance from the surface,
Ei0 Ei = - vi e and the Boltzmann equation becomes
vi e
ni = ni 0 exp

kT

(B3)

The Poisson equation relates potential, charge, and distance. For one-dimensional case
d 2

=
2

dx

(B4)

where x is distance from the surface (m), is charge density (C/m3), and is the static
permittivity of the medium (C2J-1m-1). The charge density in the diffuse layer is
contributed by the ions so that
= e vini

(B5)

229

with ni expressed as ions per unit volume.


Substitution for ni from equation (3) gives
vi e

kT

= e vi ni 0 exp

(B6)

which when substituted into equation (B4) yields


d 2
e
v e
= vi ni 0 exp i

dx
kT

(B7)

Equation (B7) is the differential equation for electric double layer adjacent to a planer
surface. Its solution provides a basis for computation of electrical potential and ion
concentrations as a function of distance from the surface.
For the case of a single cation and anion species of equal valence, that is, i = 2 and
v+ = v - = v, n0+ = n0- = n0, equation (B7) simplifies to the Poisson-Boltzmann equation
d 2 2n0 ve
v e
=
sinh i

dx
kT

(B8)

Figure B1
Variation of Electrical Potential with distance from a Charged Surface
B.1.2 Single Diffuse Double Layer

230

Solutions are given in terms of dimensionless quantities


ve
y=

kT
ve 0
z=
kT

and

Potential functions

(B9)

= Kx Distance function

where
K2 =

2 n0 e 2 v 2
kT

(B10)

In terms of these variables, equation (B8) becomes


d2y
= sinh y
d 2

Boundary conditions for the first integration are that at = , y = 0, and


dy
y
= 2 sinh
d
2

(B11)
dy
= 0 . Thus,
d

(B12)

The boundary condition for the second integration is that at = 0 , y = z; i.e., = 0,


which leads to
ey/2 =

e z / 2 + 1 + (e z / 2 1)e
e z / 2 + 1 (e z / 2 1)e

(B13)

Equation (B13) describes a roughly exponential decay of potential with distance from the
surface.
If the surface potential is small (less than about 25 mV), ve/kT << 1, and
equation (B8) may approximated by
d 2
= K 2
dx 2

(B14)

= 0 e Kx

(B15)

where

and the potential decreases purely exponentially with distance. In this case, the center of
gravity of the diffuse charge is located at a distance x = 1/K from the surface.
Consequently, the quantity (1/K) is a measure of the thickness of the double layer.
From equation (B10),
1 = o DkT
(B16)
K
2no e 2 2

231

APPENDIX C
LABORATORY EXERCISES
C.1 DIELECTRIC DISPERSION
C.1.1 Objective
Use laboratory equipment to determine radio frequency alternating current (AC) response
characteristics for different soil samples. These electrical properties are then correlated to
standard geotechnical engineering properties.
C.1.2 Equipment and Materials

Soil samples
Test cell with opposing semi-circular electrodes
Impedance analyzer for use at frequencies of 2 MHz and 32 MHz
Moisture content cups and spatula for obtaining moisture content sample
Microwave oven and analytical scale for moisture content determination

C.1.3 Background
Traditional soil behavior analysis involves index testing of remolded laboratory samples,
mechanical testing of undisturbed field samples in the laboratory, or in-situ field testing,
typically using empirical correlations with standard penetration test (SPT) or cone penetration
test (CPT) results.
This laboratory exercise demonstrates the use of relatively simple radio frequency
measurements on soil samples from the field. Other equipment that have been developed for
direct AC response in the field include the Dielectric Conductivity Cone Probe and the
Formation Factor Probe. As described previously, various correlations have been developed to
obtain soil physical properties from the soil-radio frequency AC response (Arulanandan and
Smith, 1973; Arulanandan et al., 1994; and Basu and Arulanandan, 1973).
C.1.4 Procedure
Soil samples are prepared for testing as follows:
A quantity of sand is pluviated into the test cell.
A clay sample in a field sampling tube may be extruded into the test cell, or a clay
sample may be consolidated inside the test cell.
Using the dielectric laboratory probe (shown in Chapter 5, Section 5.1.11) or an
impedance analyzer, each sample is tested at AC frequencies of 2 MHz and 32 MHz. Specific
assumptions for the tests include:
All samples are saturated.

232

Specific gravity of the clay solids is 2.7.


The apparent dielectric constant, , at 32 MHz is equal to , i.e., the dielectric
constant where becomes independent of frequency.

The last assumption may not be valid for some soils. For example, heavy clays such as
illite and montmorillonite require a frequency of at least 55 MHz. Note that with an impedance
analyzer, higher frequencies may be used, but 32MHz will be used in this text to concur with the
laboratory dielectric probe frequency readings. With each soil in the test cell, testing involves
balancing electrical circuitry to obtain constant voltage readings, VC and VR. These values are
used with the calibration charts to obtain the apparent dielectric constant, . Porosity is
determined as described in this textbook. After each sample is tested, it is extruded from the test
cell, and a sample is obtained for moisture content determination. Based on routine soil
mechanics correlations, porosity is calculated and compared to that obtained by electrical testing.
C.1.5 Report
1. Sketch the test apparatus.
2. Report dielectric constant at different frequencies for each soil sample.
3. Determine porosity based on Figure 4.11, and determine the theoretical porosity based on
traditional soil mechanics.
4. Compare and discuss results.
C.1.6 Lab References
Arulanandan, K., and Smith, S. S. (1973). Electrical Dispersion in Relation to Soil Structure,
Journal of the Soil Mechanics and Foundations Division, ASCE, 99 (SM12), 1113-1133.
Arulanandan, K., Yogachandran, C., and Rashidi, H. (1994). Dielectric and Conductivity
Methods of Soil Characterization, Geophysical Characterization of Sites, XIII ICSMFE, New
Delhi, India, 81-90.
Basu, R., and Arulanandan, K., (1973). A New Approach to the Identification of Swell
Potential in Soils, Proceedings of Third International Conference on Expansive Soils, Haifa,
Israel, Vol. 1, 1-11.

233

C.2 THREE-ELEMENT MODEL


C.2.1 Objectives
1. To use laboratory equipment to determine radio-frequency alternating current (AC)
response characteristics for a sample of saturated clay. The dielectric constant and
conductivity are then plotted against frequency.
2. To use this data and specified soil parameters (geometrical and compositional) in an
optimization curve-fitting program for the Three-Element soil Model. Using this program
will demonstrate how the input parameters are selected, and what effect variation of input
parameters has on the output.
3. To correlate the output parameters with standard geotechnical engineering properties.
C.2.2 Equipment and Materials

Clay sample
Boonton RX Meter, Model 250
Moisture content cups and spatula for obtaining moisture content sample
Microwave oven and analytical scale for moisture content determination
Optimization software program

C.2.3 Background
What follows first is a brief discussion of the Three-Element Model. The Three-Element
Model addresses the mineral-solution interface characteristics of a soil-water-electrolyte system
by describing the electrical behavior of soil in terms of an electrical network model. The
geometrical parameters in the model are:
a = area of the cluster-solution path
b = area of the cluster contact path
c = area of the solution path
d = length of path in the solids solution phase
The area of b is very small and may be taken as about 0.0006 to 0.006. Since this is
negligible, we have a = 1 c. Further, d is a function of water content, so depending on water
content and type of clay (or dielectric dispersion), d will vary between 0.7 and 1.0. Arulanandan
and Smith (1973) note that the parameter, c, should depend on such factors as porosity, particle
shape, orientation, and cementation. As can be seen, this parameter becomes very important in
optimizing the Three-Element Model. The c parameter varies from about 0.03 to 0.2.
The compositional parameters in the model are:

234

s = dielectric constant of pore fluid (for water, s = 80 at about 20C)


r = dielectric constant of soil particles. This value should be between 4 (for dry
silicate material) and 80 (for water). For soil, typical values are about 20 to 35.
kr = conductivity of the conducting soil particles, which is dependent on the mineralsolution interface characteristics and the amount of water associated with the
particles (for montmorillonite, illite, and kaolinite, this value is about 0.005,
0.002, and 0.0002, respectively), [Arulanandan and Smith, 1973].
ks = conductivity of the pore fluid, which is a measure of the concentration of
dissolved salts and can be easily measured. For the soils tested, this is in the
range of 0.0003 to 0.005.
Based on this discussion of the Three-Element-Model parameters, the most important variables
should be c, d, r, and kr.
Nearly all properties of clay soils may be considered functions of clay mineralogy and
mineral-solution interface characteristics. If clay mineralogy characteristics are not known for a
specific soil, dielectric constant and conductivity readings may be obtained in the field.
Arulanandan and Smith (1973) recommend a minimum of about 12 to 14 readings to adequately
define the dielectric dispersion and conductivity curves for use in the optimization program.
However, this textbook notes that fewer readings are often adequate. The optimized curves can
then be used to provide and 0, which can then be used to calculate soil-specific parameters:
porosity (n), slope of the critical state line (M), critical state friction angle (crit), slope of the
isotropic compression line (), compression index (Cc), slope of the swell/rebound line (),
specific surface area (SSA), coefficient of permeability (k), and total swell potential.
C.2.4 Procedure
Procedures are discussed in Chapter 3, Section 3.7.
C.2.5 Report
1. Sketch the test apparatus.
2. Describe procedure for testing dielectric constant of saturated clay sample at different
frequencies.
3. Calculate dielectric constant and conductivity for the water sample.
4. Plot dielectric constant and conductivity against frequency for the clay and sand samples.
5. Specify what geometrical and compositional parameters were used as input for the
optimization program, and present the final output parameters.
6. Determine and 0 for the clay sample.
7. Estimate porosity, specific surface area, hydraulic conductivity, compression index, and total
swell potential for both soils.
C.2.6 Lab Reference
Arulanandan, K., and Smith, S. S. (1973). Electrical Dispersion in Relation to Soil Structure,
Journal of the Soil Mechanics and Foundations Division, ASCE, 99 (SM12), 1113-1133.

235

C.3 DIELECTRIC CONE PROBE


C.3.1 Objective
Use the Dielectric Cone Probe to determine radio-frequency alternating current (AC)
response characteristics for different soil samples. These electrical properties are then correlated
to standard geotechnical engineering properties.
C.3.2 Equipment and Materials

Dielectric Cone Probe


Soil samples
Large containers for soil testing (i.e., 5-gallon buckets)
Moisture content cups and spatula for obtaining moisture content sample
Microwave oven and analytical scale for moisture content determination

C.3.3 Background
Traditional soil behavior analysis involves index testing of remolded laboratory samples,
mechanical testing of undisturbed field samples in the laboratory, or in-situ field testing,
typically using empirical correlations with standard penetration test (SPT) or cone penetration
test (CPT) results. This laboratory exercise demonstrates the use of relatively simple radio
frequency apparatus on soil samples from the field. Other equipment that have been developed
for direct AC response in the field include the Dielectric Laboratory Probe and the Formation
Factor Probe. Various correlations have been developed to obtain soil physical properties from
the soil radio frequency AC response (Arulanandan and Smith, 1973; Arulanandan, et al., 1994;
and Basu and Arulanandan, 1973).
C.3.4 Procedure
Soil samples are prepared for testing as follows:
Sand is pluviated into a large container
A saturated clay sample is prepared in a similar container
Calibration charts for the dielectric probe are needed beforehand. Each sample is tested at AC
frequencies of 2 MHz and 32 MHz. Specific assumptions for the tests include:

All samples are saturated.


Specific gravity of the clay solids is 2.7.
The apparent dielectric constant, , at 32 MHz is equal to , i.e., the dielectric
constant where becomes independent of frequency.

This textbook notes that the last assumption may not be valid for some soils. For
example, heavy clays such as illite and montmorillonite need a frequency of at least 55 MHz.
With each soil in the test cell, testing involves balancing electrical circuitry to obtain constant

236

voltage readings, VC and VR. These values are used with the calibration charts to obtain the
apparent dielectric constant, . Porosity is determined based on the relationship developed in this
textbook.
After each soil is tested with the Dielectric Cone Probe, the probe is removed from the
test container and a representative soil sample is obtained for moisture content determination.
Based on routine soil mechanics correlations, porosity is calculated and compared to that
obtained by electrical testing.
C.3.5 Report
1. Sketch the test arrangement, including soil sample, probe and connections to control unit.
2. Report dielectric constant at different frequencies for each soil sample.
3. Determine the porosity of both samples based on , and determine the theoretical
porosity based on traditional soil mechanics.
4. Compare and discuss results.
5. Discuss applications in the field.

C.3.6 Lab References


Arulanandan, K., and Smith, S. S. (1973). Electrical Dispersion in Relation to Soil Structure,
Journal of the Soil Mechanics and Foundations Division, ASCE, 99 (SM12), 1113-1133.
Arulanandan, K., Yogachandran, C., and Rashidi, H. (1994). Dielectric and Conductivity
Methods of Soil Characterization, Geophysical Characterization of Sites, XIII ICSMFE, New
Delhi, India, 81-90.
Basu, R., and Arulanandan, K., (1973). A New Approach to the Identification of Swell
Potential in Soils, Proceedings of Third International Conference on Expansive Soils, Haifa,
Israel, Vol. 1, 1-11.
C. 4 EROSION
C.4.1 Objective
The objective of this laboratory experiment is to test the surface erosion potential of a
cohesive soil. The laboratory exercise will demonstrate the use of a rotating cylinder apparatus
to test erosion potential of a sample of remolded young bay mud. From the test, a value may be
obtained for the critical shear stress, c, defined as the stress at which erosion begins.
C.4.2 Equipment and Materials

Soil sample3-inch-diameter, cylindrical shaped, composed of remolded young bay


mud
Distilled water (5 to 10 liters, as needed)
Rotating cylinder test apparatus (maximum speed = 1,500 rpm)

237

Strain gauge and transducer


Stop watch
Compressed air source

C.4.3 Background
The subject of cohesive soil erodibility has been of concern in soil science and dam core
design for some time, as has the general erodibility of clayey soils. Various methods have been
developed in an attempt to evaluate the potential erodibility of a cohesive soil. The Rotating
Cylinder Test was originally proposed by Arulanandan, et al. (1975). The test method is
described below under Procedure. Some benefits of this test over the previous dispersion tests
are:
Quantitative test results are available.
The test can be used on any combination of clay soil and eroding water.
A critical shear stress can be identified below which surface erosion is absent or very
slow.
Because of the geometry of the test apparatus and the nature of clay soil erodibility, a
linear relationship has been demonstrated between applied shear stress (above critical
stress) and erosion rate.
The critical shear stress, c, is an important result from this test, and is defined as the
shear stress at which erosion begins. Arulanandan and Perry (1983) showed that this value could
be obtained by extrapolating a graph of observed erosion rate vs. shear stress as shown in Figure
7.17.
This value can then be used to quantify erodibility. In fact, Arulanandan and Perry
(1983) proposed some preliminary categories based on limited data for erodibility in terms of
critical shear stress, as follows:
Erodible soils, c 4 dynes/cm2
Moderately erodible soils, 4 dynes/cm2 c < 9 dynes/cm2
Erosion resistant, c 9 dynes/cm2
C.4.4 Procedure
1. Prior to mounting in the test apparatus, weigh the soil sample (a cylinder of cohesive soil,
approximately 3 inches in diameter and 3 inches high).
2. Measure the conductivity of the eroding fluid, and fill the transparent rotating cylinder half
full with the eroding fluid. Deionized water is a good eroding fluid to use for testing, as it
reasonably represents rainwater as well as natural waters found in freshwater impoundments.
3. Mount the soil sample concentrically inside the larger cylinder such that a -inch annular
space remains between sample and cylinder.
4. Rotate the cylinder at a specified speed (shear stress). After rotation for 2 minutes, remove
the eroding fluid (distilled water) and soil sample.
5. Replace the eroding fluid with clean fluid, if needed, and weigh the soil sample.
6. Repeat Steps 4 and 5 as necessary, calculating the erosion rate after each 2-minute run. The
erosion rate, e& is defined as:

238

e& =

M
At

Where, M is the change in mass (grams), A is the surface area of the sample exposed to the
eroding fluid, and t is the time of rotation.
Note: As the outer cylinder is rotated with the inner cylinder (soil sample) held stationary,
rotation is imparted to the fluid. This movement of the fluid, in turn, transmits a shear to the
surface of the inner cylinder (soil). Since the annular spacing between the soil sample and the
outer cylinder is constant, and since there are no abrupt changes in roughness on the eroding
surface, the stresses are believed to be nearly uniform at all points on the surface of the soil. The
soil sample is stationary, but is mounted on a compressed air bearing. Thus, the shear stress
transmitted to the surface results in a slight rotation of the supporting mandrel against a
mechanical load. The rotation is measured by a strain gauge and pressure transducer. A
schematic of the rotating cylinder test apparatus is shown in the text.
C.4.5 Report
1.
2.

3.
4.
5.

Sketch the test apparatus.


Tabulate the test data, including run number, speed (%), speed (rpm), spin time
(minutes), shear stress (dynes/cm2), post-spin mass (g), and erosion rate (g/cm2-sec and
g/cm2-min).
Graph erosion rate (g/cm2-min) vs. shear stress (dynes/cm2), and comment on the
erodibility of the test soil.
Read the paper by Arulanandan, et al. (1975), and provide a one page summary.
What parameters should be considered in preventing soil erosion and why?

C.4.6 Lab References


Arulanandan, K., Loganathan, P., and Krone, R. B. (1975). Pore and Eroding Fluid Influences
on Surface Erosion of Soil, Journal of the Geotechnical Engineering Division, ASCE, 101
(GT1), 51-66.
Arulanandan, K., and Perry, E. B. (1983). Erosion in Relation to Filter Design Criteria in Earth
Dams, Journal of Geotechnical Engineering, ASCE, 109(5), 682-698.

239

ISBN 978-0-615-18983-3

You might also like