Soil Dynamics and Earthquake Engineering 91 (2016) 72–86

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

journal homepage: www.elsevier.com/locate/soildyn

Non-conventional lab and field methods for measuring frequency-

dependent low-strain parameters of soil dynamic behavior
C.G. Lai a,n, A.G. Özcebe b
a
Department of Civil Engineering and Architecture, University of Pavia, Italy
b
Department of Civil and Environmental Engineering, Politecnico di Milano, Italy

art ic l e i nf o a b s t r a c t

Article history: The response of geomaterials to low-strain dynamic loading is of foremost importance in a variety of
Received 31 January 2016 situations in soil dynamics (e.g. foundation design of vibrating machines, assessment of vibrational
Received in revised form impact induced by railway, subway or roadway traffic) and earthquake geotechnical engineering (e.g.
20 August 2016
ground response analysis at stable sites). Yet, the current methods to determine the low-strain para-
Accepted 4 September 2016
meters of soil dynamic behavior are characterized by a number of shortcomings and inconsistencies.
Available online 22 September 2016
Even the very same definition of these parameters is imprecise and in a way contradictory. This has
Keywords: relevant implications when they are used to solve boundary-value problems, particularly through ad-
Dynamic soil properties vanced numerical modeling, as it will be shown in the paper. The article, after a critical review of two
Cross-hole seismic testing
widespread field and laboratory experimental tests, introduces non-conventional techniques to carry out
Resonant column test
and interpret the same experiments to measure the low-strain parameters of soil dynamic behavior. The
Viscoelasticity theory
Kramers-Kronig relations preliminary results obtained from these emerging methods are encouraging despite a few technical
Physical causality difficulties that still need to be overcome.
Frequency-dependent soil properties & 2016 Elsevier Ltd. All rights reserved.

1. Introduction
The expression “dynamic properties of geomaterials” or the more
popular “dynamic soil properties” denotes the set of mechanical
parameters controlling the response of soils and rocks to dynamic
loading. Though very popular in the earthquake geotechnical en-
gineering community, it is a rather unfortunate expression as al-
ready noted by Kramer [38] at page 184 of his monograph entitled
“Geotechnical Earthquake Engineering” where he states that “…the
properties themselves are not dynamic (indeed, they apply to a host
of nondynamic problems)…”. Perhaps we could go even further and
claim that the word “properties” is the most inappropriate in the
sentence. Is the undrained shear strength a soil property or rather a
soil parameter describing a certain hydro-mechanical behavior or,
even better, a soil parameter of the Tresca constitutive model ?
The description of soil response to any type of “loading” (e.g.
mechanical, hydraulic, electric, magnetic, thermal, etc.) is often
carried out through the adoption of a constitutive model linking a
set of response functions to the perturbation of one or more ex-
ternal fields. If of interest is the mechanical response, than the
components of the strain tensor may be the response function to
n
Corresponding author.
E-mail addresses: carlo.lai@unipv.it (C.G. Lai),
aliguney.ozcebe@polimi.it (A.G. Özcebe).
the perturbation of the existing stress field (e.g. the lithostatic
stresses). The constitutive model coincides in this case with the
stress-strain law. However, we could use different stress-strain
laws to describe the (phenomenological) mechanical response of
soils to static and dynamic loading being the elastic, viscoelastic or
elasto-visco-plastic only a few examples. Of course, different
constitutive models will depend on different constitutive para-
meters and these models may still be used to describe the same
physical phenomenon. For instance, the speed of propagation of a
mechanical perturbation in a soil deposit is a quantity that can be
measured independently from any constitutive assumption (al-
though the result of the measurement may depend on the fre-
quency of the monochromatic wave used to carry out the test).
However, the nearly automatic association of this physical quantity
with the stiffness of the medium implies in reality a selection of a
specific constitutive model, in this case that of a one-constituent,
isotropic, linear elastic solid whose constitutive parameters are for
instance the shear and oedometric moduli.
Actually, the very same idea of measuring the “stiffness” of a soil
specimen does not make sense without specifying the speed of
application of the loading. Let us consider for instance the Young
modulus: depending on the degree of saturation and hydraulic
conductivity of the soil specimen, if the load is applied very fast,
the corresponding uniaxial deformation may be very small and
vice versa. Even after the application of the loading up to its final
value, the strain may not remain constant but increase with time

http://dx.doi.org/10.1016/j.soildyn.2016.09.007
0267-7261/& 2016 Elsevier Ltd. All rights reserved.
 C.G. Lai, A.G. Özcebe / Soil Dynamics and Earthquake Engineering 91 (2016) 72–86
(viscous behavior). Only in a purely elastic material, at the equi-
librium, the mechanical response is independent from the speed of
application of the loading as stated by the Clapeyron’s theorem.
Regardless on whether the load is applied instantaneously or
quasi-statically, the Young modulus will be the same because the
strain will not change once the load has reached its final value.
However, it cannot be taken for granted that soil response is
purely elastic particularly when it is subjected to time-varying
loading, even though at low-strain levels.
Thus, while it seems natural to determine the elastic shear and
oedometric moduli from the measurement of the shear and
compression wave velocities VS and VP respectively, indeed this is
only one possible interpretation of wave measurements in soil
deposits, perhaps the most widespread. For example, if choosing a
poro-elastic constitutive model [9,10], the same physical quan-
tities VS and VP may be used to determine the porosity of the
medium [22]. Furthermore, in measuring the speed of propagation
VS and VP it is customary to assume these quantities to be fre-
quency-independent. Depending on the specific applications, this
may be a reasonable approximation. However, if not VS and VP but
rather VS(ω) and VP(ω) are measured in case of frequency-de-
pendence, then the use of a viscoelastic constitutive model would
allow to determine the damping ratio spectra DS(ω) and DP(ω), as
it will be shown later in the paper.
From the aforementioned considerations, it appears that de-
spite “dynamic soil properties” has become a rather popular idiom
in geotechnical engineering, it is actually ambiguous, since the
word “properties” suggests something intrinsic to soils when in
reality it is not so. Perhaps a more appropriate wording to use
would be “parameters” of specific constitutive models. There are
consequences of using this terminology that go beyond semantics.
They may be at the origin of a biased habit of mind which imposes
unnecessary restrictions when interpreting the results of experi-
mental measurements. This situation is not restricted only to ex-
perimental soil mechanics.
2. Dynamic response of soils in brief
Describing the mechanical behavior of soils is difficult because
soils are multi-phase, particulate materials composed of small
solid particles interacting with one (or more) fluid phases. From a
phenomenological point of view the response of geomaterials to
mechanical perturbations is in general non-linear, anisotropic,
elasto-visco-plastic with strong stress and/or strain history de-
pendence. Furthermore, due to their particulate nature, soils have
pronounced microstructural properties, which give rise to phe-
nomena that are difficult to explain with classical continuum
mechanics. Dilatancy, that is the coupling between shear and vo-
lumetric deformation, is a distinctive example. Other features
contributing to the complexity of soil behavior are the interacting
phenomena exhibited by soils due to perturbation of external
fields. Piezo-electric and chemo-mechanical interaction is an ex-
ample that has been recently exploited for subsurface exploration
[68].
Fortunately, despite the complexity of their behavior, geoma-
terials do not always exhibit all their features with the same de-
gree of importance. Depending on the nature of the problem, its
intrinsic spatial and temporal scales, the strain levels involved and
the dominant external fields, many of the features manifested by
soils may be regarded as second or even higher order effects. Thus,
they may be neglected in constitutive modeling without appreci-
able changes in the results of the analyses.
Focusing on the response to dynamic loading, a considerable
amount of experimental research has been carried out over the
past 40–50 years. These studies have been performed using
73
various laboratory techniques including resonant column, cyclic
triaxial and cyclic torsional shear apparata as well as bender ele-
ment devices. The outcome from these tests helped identifying the
main factors affecting the dynamic behavior of soils. Among them,
the magnitude of the deviatoric strain tensor plays a crucial role in
strain-controlled tests. This quantity is a measure of the level of
shear strains induced in the soil during dynamic loading.
For uniaxial loading, Vucetic [80] recognized the existence of
three distinct types of soil behaviors depending on the size of the
induced shear strain. For strains below the linear cyclic threshold
shear strain soils tend to exhibit a linear response under both
static-monotonic and dynamic loading when the phenomenon of
energy dissipation cannot be neglected even though it is small in
magnitude. At shear strains exceeding the linear cyclic threshold
but below the volumetric cyclic threshold shear strain, soils undergo
a limited yet non-negligible mechanical degradation. The response
in this region is weakly non-linear. The hysteretic loops become
larger however they remain stable with the increase in the num-
ber of cycles. In this region dilatancy phenomena, in drained tests,
are minor if not insignificant.
With the increase of strain magnitude, the strain point tres-
passes the volumetric cyclic threshold shear strain and enters into
the strongly non-linear region where the hysteretic loops become
unstable as the number of cycles increases and soils undergo se-
vere stiffness and strength degradation. Furthermore, irrecover-
able volume changes occur in drained tests and pore-water pres-
sure develops during undrained loading. A three-dimensional,
conceptual representation of these regions of soil behavior is
shown in Fig. 1. The dependence of the response from the mag-
nitude of the deviatoric strain tensor is illustrated through the
concepts of linear and volumetric cyclic threshold strain boundary
surfaces in the principal strain space.
3. Importance of low-strain parameters of soil dynamic
behavior
The focus of this article is on experimental measurement of
constitutive parameters controlling the response of soils to low-
strains dynamic loading. This is a relevant subject in soil dynamics
and earthquake geotechnical engineering since several are the
situations in which geomaterials are strained at levels below the
linear cyclic threshold shear strain. Examples in soil dynamics
include the assessment of vibrational impact induced by the rail-
way, subway or roadway traffic (Fig. 2). The transmission of
ground-borne vibrations from the source (e.g. an underground
train) to the receiver (e.g. a person placed inside a building) under
free-field conditions may be modeled using a linear soil con-
stitutive model due to the low-strain level induced in the medium
by the passage of mechanical waves..
Earthquake ground response analysis is another example
where the assumption of soil linearity may be an acceptable ap-
proximation when modeling complex 3D geological structures
such as deep alluvial basins in advanced numerical simulations
[72]. In general, the low-strain parameters of soil dynamic beha-
vior control the amplification or de-amplification of ground mo-
tion at stable sites. These parameters play an important role also in
case of soil nonlinear behavior as for instance in linear equivalent
site response analysis where the normalized modulus reduction
curves need to be scaled to the value of the low-strain elastic
moduli.
The parameters of soil dynamic behavior at low-strains are
relevant also in problems of dynamic soil-foundation-structure
interaction. Foundations of vibrating machines (e.g. compressors,
turbo-generators, etc.), wind turbines, radio antennas, structures
exposed to sea wave motion (e.g. wharves, off-shore platforms) are
74 C.G. Lai, A.G. Özcebe / Soil Dynamics and Earthquake Engineering 91 (2016) 72–86
Fig. 1. Outline of soil mechanical response to multi-axial dynamic loading [23].
Fig. 2. Schematic representation of the vibrational impact induced by the passage
of a train in an underground tunnel [39].
examples of systems whose response to dynamic loading is con-
trolled by these parameters. These are the low-strain stiffness,
specifically the elastic, initial tangent shear modulus and material-
damping ratio. However, these parameters are in conflict to each
other since elastic modulus and damping ratio are oxymora. They
refer to two different and incompatible constitutive models. Fur-
thermore, it was already said that in soils, stiffness is not an ab-
solute “property” since its value, in a given mode of deformation,
may depend on the rate of loading. The same applies to energy
dissipation and thus to material-damping ratio. The situation is
further complicated by the fact that stiffness and energy dissipa-
tion are not independent parameters despite in the current prac-
tice of experimental soil dynamics, they are measured separately
and using different procedures.
4. Current practice to measure low-strain stiffness and mate-
rial-damping ratio
In the current practice, the low-strain stiffness in soils is typi-
cally measured through in-situ geophysical seismic methods such
as the borehole cross-hole and down-hole tests, SCPT, VSP, SDMT,
SASW, MASW, ReMi, PS-suspension logging, seismic tomography
just to mention the most widespread techniques. Geophysical
seismic methods are particularly suitable to determine the low-
strain stiffness in geomaterials because the level of strain induced
by the passage of seismic waves is very low ( <10−5), certainly
below the linear cyclic threshold shear strain. In addition, some of
these tests like SASW and MASW are carried out directly from the
ground surface without the need to drill boreholes or insert special
probes like in SCPT or SDMT.
The physical quantity that is measured when using seismic
geophysical methods is the speed of propagation of mechanical
body waves or, alternatively, surface Rayleigh or Love waves as in
MASW testing. Technical standards exists for the cross-hole [3]
and down-hole [4] seismic testing which specify protocols for the
execution and interpretation of results of these geophysical
methods to determine the low-strain stiffness in geomaterials.
In-situ seismic methods can also be used to estimate the low-
strain material-damping ratio. Techniques have been developed
for SASW and MASW testing based on measuring the spatial at-
tenuation of surface Rayleigh waves [23,57,65,7]. Attempts have
also been made to measure damping ratio from the attenuation of
body P and S waves in cross-hole, down-hole and SCPT tests
([34,50,55,58,14,25]). However, the results obtained so far with
these methods have only been moderately successful mainly due
to the difficulties of separating geometric and intrinsic attenuation.
The low-strain stiffness and material-damping ratio have been
measured through laboratory tests as well since the early days of
experimental soil dynamics, mainly using the resonant column
apparatus. After the pioneering work of Japanese engineers Ishi-
moto and Iida [32] and Iida [30], the resonant column test became
popular in the United States and worldwide during the 60′s and
70′s by the work of Hall and Richart [26], Drnevich et al. [18],
Hardin and Black [27], Richart et al. [64], Hardin and Drnevich [28]
just to name a few contributors.
Other laboratory tests are used to determine the low-strain
parameters of soil dynamic behavior including the cyclic triaxial
and cyclic simple shear tests [31]. About 25–30 years ago, wave
propagation methods have started to be used also in laboratory
 C.G. Lai, A.G. Özcebe / Soil Dynamics and Earthquake Engineering 91 (2016) 72–86
specimens in a sort of miniaturized geophysical seismic tests
[20,78]. The key component of these experiments are the bender
elements, thin piezo-ceramic plates that can be used both as a
wave source or emitter and receiver. The shear wave velocity is
typically measured within the triaxial apparatus after placing one
bender element at the top and one at the bottom of the soil spe-
cimen. These tests allow investigating how variables such as
confining stress, porosity and anisotropy influence the low-strain
stiffness in soils however, due to the size of the specimen, mea-
surements are made at very large frequencies (order of kHz). This
raises questions upon the applicability of the results obtained with
these techniques in solving problems of earthquake geotechnical
engineering where the frequencies of interest are orders of mag-
nitude lower.
A further discussion of in-situ and laboratory tests to measure
the low-strain stiffness and damping ratio in soils is beyond the
scope of this paper. Still, the next two Sections will briefly sum-
marize the most salient aspects of two specific and widespread
experiments: the in-situ cross-hole seismic method and the la-
boratory resonant column test. This does not imply any loss of
generality with respect to other methods and it lays down the
groundwork for the introduction of non-conventional methods to
measure the low-strain parameters of soil dynamic behavior. In
fact, these methods are based on a fundamentally different way to
perform and interpret these two standard techniques commonly
used in geotechnical site characterization.
4.1. In-situ cross-hole seismic testing
Cross-hole seismic testing is a borehole seismic technique used
for measuring the shear and compression wave velocity profiles in
a 1D-soil stratigraphy. It was introduced for geotechnical char-
acterization of soil deposits in the late 70′s [29,8,83] and it is still
one of the most popular in-situ seismic test worldwide. The P and
S wave velocity profiles are determined from measurements at
different depths of the travel times of P and S waves propagating
along sub-horizontal paths through points separated by a known
distance (Fig. 3).
The execution of the cross-hole test requires drilling a mini-
mum of two boreholes within which to insert a source of me-
chanical energy and a receiver for the measurement of the travel
time. However, according to the ASTM D4428-14 [3] standard, the
preferred method requires drilling three in-line boreholes to
eliminate the travel time error associated with the identification of
the instant of trigger at the source. The spacing between the
Fig. 3. Typical configuration of cross-hole seismic testing.
75
boreholes varies between 3.0 and 5.0 m depending on the me-
chanical impedance of the soil. During travel time measurements,
source and receivers are located at the same depth so that the
velocity of propagation of P and S waves is measured within the
soil layers along horizontal paths. Thus, in cross-hole testing the
propagation of seismic waves is postulated to occur along straight
paths in laterally homogeneous, layered soil deposits. The P and S
wave travel time measurements are repeated after simultaneously
moving source and receivers at different depths.
Due to sensitivity of the measured P and S wave velocities from
possible errors in estimating the distance between receivers, par-
ticularly at depths greater than 10–15 m, a deviation survey to as-
sess the verticality of the boreholes and precisely define the dis-
tance along depth between the boreholes is mandatory.
The two most common techniques adopted to estimate the
travel time of P and S waves from the source to the receivers are
the visual and cross-correlation methods. With the visual method,
the travel time is estimated from the difference at the two re-
ceivers of the arrival times of P and S waves (Fig. 3). Crucial in this
method is therefore the identification (picking) of the first arrivals
of both P and S phases on the seismograms detected at the two
receivers.
The cross-correlation technique provides more objective results
than those given by the visual method, however it is reliable only
with signals of very good quality. To facilitate the picking of S wave
first arrival, a 180° reversed polarity source is used whenever
possible.
Usually the receivers are three-dimensional velocity transdu-
cers or geophones that are able to detect the particle motion in the
vertical direction and in two horizontal directions that are mu-
tually perpendicular. These geophones have therefore the cap-
abilities of detecting the particle motion associated to three types
of waves namely P, SV and SH waves (soils are anisotropic materials
and in general VSV ≠ VSH ). However, to improve the signal-to-noise
ratio, best measurements are made using specialized sources,
namely sources capable of generating impulses rich in either P, SV
or SH energy.
Knowing travel paths and travel times of P and S waves, their
speed of propagation can be easily computed. If ∆L is the distance
between the two receivers (Fig. 3), ∆tP and ∆tS the travel times of P
and S waves respectively, then VP and VS are obtained from the
following relations:
ΔL ΔL
VP = , VS =
ΔtP ΔtS (1)
It is important to point out that the speed of propagation of P
and S waves calculated with Eq. (1), is inherently assumed fre-
quency-independent. Thus, in the conventional interpretation of
cross-hole seismic testing, P and S waves are assumed to be non-
dispersive. Actually, in dissipative media like soils high frequencies
travel faster than low frequencies and VS (or VP ) computed with Eq.
(1) corresponds to the speed of propagation of the wave front
whose energy is carried mostly by high frequencies. These travel
faster because material response gets “stiffer” for higher loading
rates. Compare, as an example, the drained versus the undrained
response of saturated geomaterials. Seismologists are aware of this
phenomenon and take it into account through the Kolsky at-
tenuation-dispersion model [1,24] which will be discussed in
more detail in the sequel.
From Eq. (1), the low-strain oedometric and shear elastic
moduli M and G can be calculated from the following standard
equations of elastodynamics:
M = ρ⋅VP2, G = ρ⋅VS2 (2)
where ρ is the mass density of the soil.
76 C.G. Lai, A.G. Özcebe / Soil Dynamics and Earthquake Engineering 91 (2016) 72–86
4.2. Torsional-resonant column apparatus
The torsional-resonant column apparatus is a laboratory device
that can be used to measure both the low-strain stiffness and
material-damping ratio of soils. The ASTM D4015-07 [2] standard
provides instructions for experimentally determining these para-
meters of soil dynamic behavior for solid, cylindrical specimens in
undisturbed as well as remolded conditions. Sometimes hollow
specimens are used to obtain a more uniform distribution of shear
strain amplitude along the cross-section of the sample. In the
conventional resonant column test, low-strain stiffness and ma-
terial-damping ratio are measured using different procedures and
interpretation methods as briefly reviewed below.
4.2.1. Low-strain shear modulus
Prior to the dynamic phase, the soil specimen is consolidated
following a procedure similar to that used in a standard triaxial
test. Next, the specimen is subjected to a harmonic torsional ex-
citation by an electromagnetic driving system [19]. Impulsive ex-
citations can also be adopted when the device is used to determine
the low-strain damping ratio [73]. A typical configuration of the
resonant column apparatus is shown in Fig. 4 with the soil spe-
cimen fixed at the base and free to rotate at the top where the
driving torque is applied.
The frequency and amplitude of torsional loading can be in-
dependently controlled when the specimen is set into harmonic
oscillation. The objective of the test is to determine the first re-
sonant frequency of the soil specimen. This is achieved by setting
the sample to low-amplitude, torsional oscillations of gradually
increasing frequency. The response of the specimen is tracked
through an accelerometer attached at the top of the apparatus
(Fig. 4).
The shear strains are measured by means of proximity trans-
ducers (Fig. 4). The amplification function is then computed
through a spectrum analyzer and the lowest frequency ω0 at which
Fig. 4. Layout of the resonant column instrumentation to measure the low-strain elastic modulus and material-damping ratio of soils [73].
resonance is attained, represents the fundamental frequency of the
specimen for torsional oscillations. The resonance frequency can
be linked to the shear wave velocity VS of the soil specimen via the
following equation [64]:
ω0h ⎛ ω h⎞ J
tan ⎜ 0 ⎟ =
VS ⎝ VS ⎠ J0 (3)
where h is the height of the specimen, J0 is the rotational moment
of inertia of the mass mounted at the top of the specimen and
J = ρh·I is the mass polar moment of inertia of the sample having a
mass density ρ and area polar moment of inertia of the cross
section I . Eq. (3) is obtained by searching the first natural fre-
quency of vibration of an elastic, fixed-free, circular cylinder un-
dergoing free, harmonic torsional oscillations [31]. From Eq. (3),
the low-strain shear modulus is calculated using Eq. (2).
4.2.2. Low-strain shear-damping ratio
As mentioned earlier, the torsional-resonant column test can
also be used to determine the low-strain shear-damping ratio of
soils. The two most frequently used procedures are the so-called
free-vibration decay and half-power bandwidth methods. In the
free-vibration decay technique, material damping is obtained from
the amplitude decay of torsional oscillations exhibited by a soil
specimen set into free vibration. Damping ratio is determined
from the ratio, in logarithmic scale, of peak rotation amplitudes
measured over a number of consecutive cycles (logarithmic
decrements).
An alternative approach is the half-power bandwidth forced-
vibration method that exploits the dependency on damping ratio
of the shape of the frequency response curve obtained when
measuring the low-strain shear modulus (Fig. 4). It can be shown
that, for small values of damping ratio, the following result holds
[38]:
 C.G. Lai, A.G. Özcebe / Soil Dynamics and Earthquake Engineering 91 (2016) 72–86
ω2 − ω1
DS =
ω1 + ω2
where ω1 and ω2 are the angular frequencies at which the am-
plitudes of response equal 1/ 2 times the amplitude at the re-
sonance frequency (maximum). The mean value of time-average
rate of energy dissipation at frequencies ω1 and ω2 caused by
damping is in essence equal to one-half the peak average power
input and this is the motivation for the name given to the method
[13]. Recently Wu [84] derived a formula to determine material
damping ratio from the half-power bandwidth technique, which is
also valid for large values of damping.
4.3. Inconsistencies of current methods to measure VS and DS
The review, in the previous Sections, of the techniques for
measuring the low-strain parameters of soil dynamic behavior has
been limited only to the in-situ cross-hole seismic testing and the
laboratory resonant column method. This, however, does not im-
ply a loss of generality in the current discussion on measuring the
low-strain stiffness and material-damping ratio of soils since other
conventional field and laboratory tests are based on the same
principles and similar interpretation procedures. Thus, the short-
comings listed below can be assumed to hold for general con-
ventional methods used to measure the low-strain parameters of
soil dynamic behavior:
 VS and DS are measured independently using different experi-
mental procedures and interpretation methods sometimes in-
consistent to each other. For instance, with the resonant column
apparatus, VS is measured assuming soil behaving as a linear
elastic material, however the same equipment is used to de-
termine for the same soil specimen damping ratio which is a
normalized measure of energy dissipation due to inelastic
phenomena. This means that VS and DS are soil parameters
assumed to belong to incompatible constitutive models. Instead,
a consistent, unified framework of soil behavior should be used
to interpret the results of a dynamic experiment, in this case the
laboratory resonant column test.
 Current methods to determine low-strain stiffness and materi-
al-damping ratio of soils completely disregard the fact that
these parameters of soil dynamic behavior are coupled and they
exert an influence to each other. In some cases, this coupling
effect may be neglected, in some other situations it cannot be
neglected. In the present experimental practice however, they
are always treated as if they were two disjoint parameters.
 Frequency-dependence of low-strain shear modulus and ma-
terial-damping ratio of soils is usually overlooked in conven-
tional field and laboratory tests. As a result, rate-independent
(i.e. hysteretic) damping ratio is often invoked a-priori in several
situations of earthquake geotechnical engineering like in linear
or linear-equivalent ground response analyses. Even when other
approaches are used where the frequency-dependence of stiff-
ness and damping ratio is taken into account (e.g. [85,6,53,60]),
conventional field and laboratory tests measure these para-
meters at different frequency ranges. For instance, in cross-hole
seismic testing, phase velocity measurements of shear waves
are made at frequencies that fall in the range of hundreds of
hertz if not higher, depending on the source. In the resonant
column device, the low-strain shear modulus and damping ratio
are measured at lower frequencies, on the order of 15–150 Hz.
In light of the aforementioned considerations, it appears that a
fundamentally different approach is needed to measure the low-
strain parameters of soil dynamic behavior. This approach should
be based on assuming the same constitutive model to determine
77
all the parameters required to describe soil response to low am-
(4) plitude, uni-axial or multi-axial excitations. Furthermore, the new
methodology should be feasible from an experimental point of
view.
Two applications of this approach to experimental soil dy-
namics will be presented in the forthcoming Sections with re-
ference to non-conventional ways to perform and interpret the
laboratory resonant column test and the in-situ cross-hole seismic
method. Presentation of these non-conventional tests to measure
VS and DS should however be preceded by a Section dedicated to
the theoretical background needed to frame soil dynamic behavior
at low-strain levels.
5. Modeling dynamic response of soils at low-strains: an ac-
count of viscoelasticity theory
5.1. Constitutive relations in time and frequency domain
The simplest formal theory that can be used to describe the
mechanical response of dissipative materials to low-amplitude
dynamic strains is linear viscoelasticity. Despite its simplicity, the
theory of viscoelasticity has proved to be effective in describing
phenomena of wave propagation in materials like soils and rocks
at low-strain levels [31]. Experimental evidence shows in fact that
geomaterials subjected to low-strain dynamic loading exhibit both
the ability to store strain energy and to dissipate strain energy
over a finite time interval. Both these phenomena can quite ac-
curately be described by the theory of linear viscoelasticity. Its
formulation is based on the validity of a) the small strain as-
sumption, b) the time-translation invariance postulate which states
that material response is assumed independent of any time shift,
c) the inheritance postulate which states that at any point of the
material, the current value of the Cauchy stress tensor is only a
function of the current value of the strain tensor and of its past
strain history, and d) the fading memory assumption, which states
that the current state of stress depends more strongly on the re-
cent rather than on the distant strain history. With these as-
sumptions, the Riesz representation theorem [12] guarantees the
existence of a unique constitutive relation between the Cauchy
stress tensor and the strain history via the following linear func-
tional:
t
σ(t ) = ∫−∞ G(t − τ ): ddε(ττ ) dτ (5)
where G(t) is a fourth order tensor-valued function called the re-
laxation tensor function of the material, σ ( t ) and ε( t ) are the time-
dependent Cauchy stress and small strain tensors respectively. Eq.
(5) is sometimes called the Boltzmann’s equation. The relaxation
tensor function G(t) has 81 components, however, only 21 are
independent in a general viscoelastic anisotropic material. These
components represent the stress response of a material subjected
to a strain history specified as a Heaviside or step function. Al-
ternatively to the relaxation function, the response of a viscoelastic
material could be described in terms of the creep tensor function
J(t) whose components represent the strain response of a material
subjected to a stress history specified as a Heaviside function.
For an isotropic, linear, viscoelastic material, the relaxation
tensor function has only two independent components which can
be for instance the shear GS(t) and bulk GB(t) relaxation functions,
respectively. These functions are sufficient to completely char-
acterize the mechanical response of the material and they are the
analogous of the shear and bulk moduli of linear elasticity with the
important difference that the relaxation functions are time-
dependent.
Often the constitutive relationships of viscoelastic materials are
78 C.G. Lai, A.G. Özcebe / Soil Dynamics and Earthquake Engineering 91 (2016) 72–86
given a physical interpretation in terms of rheological models
formed by various combinations of elastic springs and viscous
dashpots. One such combinations is the Kelvin-Voigt model in
which a linear spring and a dashpot are combined in parallel. Al-
though the use of rheological models to formulate viscoelastic
stress-strain laws may be useful, by no means this is necessary.
Worse than that, it imposes unnecessary restrictions on the cor-
responding stress-strain law. Far more general viscoelastic models
may be constructed which do not admit a representation in terms
of networks of a discrete number of springs and dashpots such as
models characterized by a continuous spectrum of relaxation times.
Once GS(t) and GB(t) have been specified, Eq. (5) may be used to
compute the stress response of the material to a prescribed strain
history, although this operation may not be trivial. If however, the
strain history is specified as a harmonic function of time like
ε(t ) = ε0 ·eiωt , then integral-differential Eq. (5) reduces to the fol-
lowing algebraic equation:
σ(t ) = G*( ω): ε 0eiωt (6)
where G*(ω) = G (1)(ω) + i·G (2)(ω) is the fourth order complex-valued
tensor modulus with real and imaginary parts G ( 1) and G ( 2) re-
spectively. For isotropic viscoelastic materials, the mechanical re-
sponse in the frequency domain is completely described by the
shear GS*( ω) and bulk GB*( ω) complex moduli.
5.2. Material damping ratio
One important feature predicted by the constitutive relation of
a viscoelastic material undergoing a steady state harmonic oscil-
lation is that stresses and strains are in general out of phase (this is
reflected by the complex nature of the constitutive parameters).
The amount by which the stress lags behind the strain is measured
by the argument or the imaginary part of the complex modulus.
Thus, the latter is a measure of the amount of energy dissipated by
the material during cycles of harmonic oscillations. It can be
shown that the shape of the stress-strain loop predicted by the
theory, for a general viscoelastic material undergoing harmonic
oscillations, is elliptical [61] which compares fairly well with ex-
perimental stress-strain hysteretic loops obtained in soils at low
strains [16], below the linear cyclic threshold shear strain. The area
enclosed by the ellipse for uni-axial shear deformation is equal to:
2 linear viscoelasticity are formally identical to those associated
ΔWSdissip(ω) = π⋅G(2)S ⋅ εS 0 (7)
Although Eq. (7) has been obtained for the simple case of shear,
uni-axial oscillation, it can be easily generalized to the case of
multi-axial loading. At the microscopic level, different mechan-
isms have been proposed to explain the process of energy dis-
sipation occurring at very low strains in soils subjected to dynamic
oscillations [9,10,71,33,82,47,67]). These studies indicate that an
interactive combination of several individual mechanisms is re-
sponsible for most of the phenomena macroscopically called en-
ergy dissipation. For coarse-grained soils, the two mechanisms that
have been postulated are frictional losses among soil particles and
fluid flow losses due to the relative movement between the solid
and fluid phases. In fine-grained soils, energy dissipation at low
strain is controlled by electromagnetic interactions among water
dipoles and microscopic solid particles. For practical applications, a
dimensionless definition of the amount of energy (per unit volume)
dissipated by a viscoelastic material undergoing harmonic oscil-
lations has actually proved to be more effective. In soil dynamics,
this definition is represented by material-damping ratio:
ΔWSdissip( ω)
DS ( ω) =
4π⋅W Smax( ω) (8)
where WSmax is the maximum stored strain energy per unit volume,
during a cycle of harmonic loading. Although Eq. (8) is written
with reference to shear deformation, it is actually valid also for
bulk and oedometric loading. However, one of the problems as-
sociated with this definition of damping ratio is that the maximum
stored energy per unit volume of a harmonically excited viscoe-
lastic material depends not only on G( 1)S but also on G( 2)S as well as
on their derivatives with respect to frequency [59]. Thus, Eq. (8)
would lead to a rather complicated relation for damping ratio if
this is to be expressed as a function of viscoelastic constitutive
parameters. The problem is solved if in Eq. (8), WSmax is replaced by
W Save , the average stored energy over a cycle of harmonic oscilla-
tion. In fact, it can be shown that the following result holds [76]:
1 2
W Save(ω) = ⋅G(1)S ⋅ εS 0
4 (9)
If damping ratio in Eq. (8) is re-defined by replacing in the
denominator 4π·WSmax with 8π·W Save , substitution of Eqs. (7) and (9)
in Eq. (8) yields:
G(2)S
DS (ω) =
2⋅G(1)S (10)
When the energy losses in the material are small (a situation
occurring for strains below the linear cyclic threshold shear
strain), damping ratio defined by Eq. (8) and that involving W Save
yield the same results. However, the latter has the advantage of
linking damping ratio to viscoelastic material functions and of
being independent from the magnitude of the energy losses.
For damping ratio defined by Eq. (10), it can be shown that at
very low (ω → 0) and very large (ω → ∞) frequencies, a viscoelastic
material behaves as an elastic solid and thus the limiting values of
damping ratio are both zero [54]. Therefore, the spectrum of
damping ratio is nonzero and positive-definite over a closed and
bounded set of frequencies; thus, its integral is finite. Experi-
mental evidence is in this regard more problematic since no
general conclusion can be drawn to date concerning the low-strain
dynamic behavior of soils and rocks at the two ends of the
damping spectrum.
5.3. Relation between complex moduli and wave parameters
It can be shown that the Fourier-transformed field equations of
with linear elasticity except that the elastic shear and bulk moduli
are replaced by the frequency-dependent shear G*S ( ω) and bulk
G*B( ω) complex moduli [12]. Eq. (6) is an example of this result. The
analogy between the field equations of linear elasticity and vis-
coelasticity forms the basis of the “elastic-viscoelastic correspon-
dence principle” [63]. This principle states that elastic solutions to
steady state boundary-value problems can be converted into vis-
coelastic solutions for identical boundary conditions by replacing
the elastic shear and bulk moduli G and B with the corresponding
complex moduli G*S ( ω) and G*B( ω). Although the validity of the
correspondence principle is restricted to problems where the
prescribed boundary conditions are time-invariant, its exploitation
in elastodynamics is quite advantageous. If applied to Navier’s
equations of motion of linear elastodynamics, this principle yields
a pair of wave equations governing the propagation of P and S
viscoelastic waves. Their speeds of propagation are complex-va-
lued and related to constitutive parameters G*S ( ω) and G*B( ω) of the
medium through the following pair of equations:
4
GB*( ω) + 3 GS*( ω) GS*( ω)
V P*( ω) = ; V S*( ω) =
ρ ρ (11)
These relationships simultaneously define phase velocity and
 C.G. Lai, A.G. Özcebe / Soil Dynamics and Earthquake Engineering 91 (2016) 72–86
attenuation of monochromatic bulk viscoelastic waves as shown,
for S waves, by the following formulas [44]:
⎧ Re2 V S* + Im2 V S*
⎪ V ( ω) = ( ) ( )
⎪
⎪
S
( )
Re V S*
⎨ elapsed where d is the distance between the source and the ob-
⎪ ( ) ( )
Re V S* ⋅Im V S*
⎪ DS ( ω) =
⎪
⎩ Re2 V S* − Im2 V S*
( ) ( )
Analogous relations hold for longitudinal P waves. In Eq. (12)
( ) ( )
Re V S* and Im V S* denote the real and imaginary part respectively,
of the complex-valued shear wave velocity defined by Eq. (11).
Eq. (12) suggests two important remarks. The first is the fre-
quency-dependence of phase velocity and damping ratio inherited
by constitutive parameters G*S ( ω) and G*B( ω) through Eqs. (11) and
(12). Thus a viscoelastic medium is inherently dispersive, that is the
speed of propagation of viscoelastic waves is necessarily fre-
quency-dependent [24]. The second remark is that phase velocity
VS(ω) and damping ratio DS(ω) are a pair of material functions
alternative to the real and imaginary parts of the complex modulus
G*S ( ω). The same applies also to VP(ω) and DP(ω) with respect to
4
G*P(ω) = G*B(ω) + ·G*S (ω).
3
5.4. Causality constraint and Kramers-Kronig equations
For each mode of deformation (say e.g. shear), only one mate-
rial function is required in the time domain to completely specify
the mechanical response of a viscoelastic material (say for instance
the shear relaxation function). Thus in the frequency domain the
real and imaginary parts GS( 1) and GS( 2) of the complex shear
modulus G*S ( ω) cannot be independent since in the time domain
the shear relaxation function is real-valued and it would not be
acceptable that more information would be needed in the fre-
quency domain to fully describe the mechanical response of a
viscoelastic material. As a matter of fact, the relationship between
GS( 1) and GS( 2) can be easily derived using the Fourier integral
theorem. For the shear complex modulus the result is [12]:
GS (2)(τ )⋅ω2
2 ∞
GS (1)(ω) = GS (e) +
π
⋅ ∫0 τ⋅(ω2 − τ 2)
⋅dτ
where GS( e) = GS( t → ∞) is known as the equilibrium response of
the shear relaxation function GS(t). Eq. (13) is one form of the
Kramers-Kronig relation which states that the real and the ima-
ginary part of the complex shear G*S ( ω) (or bulk) modulus are
Hilbert transforms pairs.
From Eqs. (11) and (12) some kind of relationship is also ex-
pected to exist between the two functions VS(ω) and DS(ω) for
exactly the same reason that the real and imaginary parts of the
complex modulus are related through Eq. (13). This relation does
exist and it represents another form of the Kramers-Kronig
equation, perhaps the most important for applications in earth-
quake geotechnical engineering and soil dynamics. It is written
below [54]:
∞ 4 ⎡ DS( τ ) ⎤
VS2( ω) + ω2⋅ ∫0 ⋅⎢ ⎥ ⋅ VS2( τ ) dτ = GS(e )⋅
π ⎣ τ⋅(τ 2 − ω2) ⎦ 1 + 1 + 4D 2
Formally, Eq. (14) is a Fredholm, singular, integral equation of 2nd
kind with Cauchy kernel. It establishes an integral relation between
material functions VS(ω) and DS(ω) and as a result, they cannot be
assigned independently despite in the usual practice they are
determined separately using different experimental procedures
and often ignoring their frequency-dependence as discussed in
previous Sections.
It can also be demonstrated that Eqs. (13) or (14) constitutes
79
the necessary and sufficient conditions for material functions
G*S ( ω) or VS(ω) and DS(ω) to satisfy the physical principle of caus-
ality which states that a disturbance originated at a point of a
viscoelastic medium (source) is not allowed to arrive at a different
point of the same medium (observer) before the time d/c has
server and c is the speed of propagation of the disturbance in the
medium (VS or VP).
(12)
Full exploitation of Eqs. (13) and (14) is the essence of “non-
conventional methods” for measuring the low-strain parameters of
soil dynamic behavior, as it will be shown in the remaining Sec-
tions of this article.
5.5. Solutions of Kramers-Kronig relation
In seismology, a well-known, particular solution of Eq. (14) is
that obtained under the assumption that damping ratio DS(ω) is
frequency-independent over the seismic bandwidth (i.e. 0.001–
100 Hz). For shear waves, this is given by the following relation
also known as the Kolsky attenuation-dispersion model [1]:
VS ( ωref )
VS ( ω) =
⎡
⎣1 +
2DS
π
ln( )⎤⎦ ωref
ω (15)
where ωref denotes a reference angular frequency usually assumed
equal to 2π . A similar equation has also been proposed for P waves.
Eq. (15) is a dispersion relation and predicts a shear wave velocity
that increases monotonically with damping ratio for a specified
frequency. On the other hand, for a particular value of damping
ratio, Eq. (15) predicts an asymptotic increase of VS(ω) with fre-
quency (high frequency body waves travel faster).
Eq. (15) is often postulated a-priori not only in seismology but
also in soil dynamics owing to the rate-independence of damping
ratio exhibited by many soils (and rocks) at low-strains in the
seismic bandwidth [49,51,70,74,75,81]. Frequency or rate-in-
dependent damping ratio is also termed hysteretic, a word used in
physics to denote memory effects processes that are scale in-
dependent [79]. However, other studies are of opposite sign and
seem to demonstrate that in geomaterials damping ratio is sen-
(13)
sitive to the loading frequency even in the seismic band and at
low-strain [15,17,37,46,48,5,52,66,86].
Thus, the experimental results are controversial and they do
not allow to draw a definitive conclusion upon the presumed
hysteretic nature of damping ratio in geomaterials at low-strain.
Actually, this way of posing the problem is conceptually incorrect.
The frequency-dependence or independence of damping ratio in
soils should not be a matter to be postulated a-priori. It is as if
instead of measuring the relaxation function GS(t) of a viscoelastic
material (whatever it is), a particular functional form of GS(t) is
specified in advance when interpreting a set of experimental re-
sults. Indeed, the damping ratio spectrum can be viewed as the
material function that needs to be measured to fully characterize
the response of soils subjected to low-strain dynamic excitations.
Once DS(ω) is specified, the frequency-dependent shear wave
velocity can be computed from Eq. (14). Of course, the role of
2⋅(1 + 4DS2) DS(ω) and VS(ω) as dependent and independent material functions
may be reversed. However, viscoelasticity theory imposes a con-
S (14)
straint on the DS(ω) and VS(ω) relation. Indeed it can be shown
that a constant damping ratio over the entire frequency range
ω∈]−∞ , + ∞[ would imply a frequency independent shear (or
compression) wave velocity and this violates the principle of
causality since no Hilbert transform pair may satisfy the Kramers-
Kronig Eq. (14) with a constant damping ratio [1]. Therefore, some
type of frequency-dependence should be admitted outside the
seismic band even for a hysteretic damping ratio.
Using the theory of singular, integral equations, Meza-Fajardo
80 C.G. Lai, A.G. Özcebe / Soil Dynamics and Earthquake Engineering 91 (2016) 72–86
and Lai [54] obtained an explicit, closed-form solution of Eq. (14)
both for the case when VS(ω) is prescribed and for the case when
the specified function is the damping ratio spectrum DS(ω). For the
former situation, the result is:
DS (ω) =
2ωVS(ω)
πVS(0)
∞
∫0 ( VS(0)
⋅
dτ
VS(τ ) τ 2 − ω2 ) of the cylinder remain plane and normal to the axis), after in-
⎡ 2ωVS(ω) 2 troducing a system of cylindrical coordinates { r , ϑ, z} with the
∞
⎢⎣ πVS(0) ∫0 ( VS(0)
⋅
dτ
VS(τ ) τ 2 − ω2 )⎤⎥⎦ −1
(16)
where VS( 0) = lim VS ( ω). Eq. (16) represents an explicit dispersion
ω→ 0
relation for arbitrary dissipative, viscoelastic materials like soils
and rocks subjected to low-strain excitations. By measuring the
frequency-dependence of shear wave velocity VS(ω), this equation
can be used to calculate the damping ratio spectrum DS(ω). Note
that this is entirely done from phase velocity measurements only.
6. Laboratory non-resonance column methods
As shown in previous Sections, in conventional resonant col-
umn testing, the low-strain shear modulus is determined by
linking the experimentally measured torsional resonance fre-
quency of a soil specimen to the analytical solution of the corre-
sponding elastic boundary value eigenproblem. Instead, damping
ratio is obtained from the free-vibration decay or half-power
bandwidth methods. Nevertheless, the same resonant column
apparatus may be actually used to experimentally determine the
frequency-dependent, shear complex modulus G*S ( ω) of a soil
specimen idealized as a linear viscoelastic solid. Shear wave ve-
locity and shear damping ratio can then be calculated from Eqs.
(11) and (12) which also allow investigating the frequency de-
pendence of these parameters.
The use of this approach has been initially proposed by Lai and
Rix [43], preliminarily realized by Lai et al. [42] and fully de-
monstrated by Rix and Meng [66] who also proposed the name of
non-resonance column method after the application of the proce-
dure to polymers by Read and Dean [62]. The essence of the
method is on solving a forced boundary value problem of elasto-
dynamics associated to harmonic, torsional oscillations followed
by the application of the elastic-viscoelastic correspondence
Fig. 5. Configuration of the non-resonance column apparatus [66].
principle. The configuration of the test as realized by Rix and Meng
[66] is shown in Fig. 5. A right-cylindrical soil specimen fixed at
the base is excited at the top by a harmonic driving torque T0·eiωt ..
Assuming the validity of the small-strain theory and of the
usual kinematics for torsion (i.e. cross sections normal to the axis
origin of the z -coordinate set at the top of the specimen, the
equation of motion governing the vibrations of the specimen,
idealized as a homogeneous elastic cylinder, can be written as
follows:
∂ 2uϑ 1 ∂uϑ 1 ∂ 2uϑ ∂ 2uϑ 1 ∂ 2u
2
+ + 2 2
+ 2
= 2 ⋅ 2ϑ
∂r r ∂r r ∂ϑ ∂z VS ∂t (17)
where uϑ ( r , z, t ) is the displacement component along the cir-
cumferential direction ϑ . Solution of Eq. (17) may be sought using
the method of separation of variables coupled with the application
of the boundary conditions, which are uϑ ( r , h, t )=0 and
M ( 0, t ) = T0eiωt − J0 ϑ̈( 0, t ) where M ( 0, t ) is the twisting moment at
the top of the specimen. The result is:
T0eiωt πR 4 ρω ⎛ ωh ⎞
= ⋅ ⋅ cot⎜ ⎟ − J0 ω2
ϑ( 0, t ) 2 VS ⎝ VS ⎠ (18)
where ϑ( 0, t ) is angle of twist at the top of the specimen, R is the
radius of the specimen and J0 is the mass moment of inertia of the
added mass placed on top of the specimen. Application to Eq. (18)
of the elastic-viscoelastic correspondence principle yields:
T0eiωt ⎛ ωh ⎞
πR 4 ρω
= ⋅ ⋅ cot⎜ ⎟ − J0 ω2
ϑ( 0, t ) 2 VS * ⎝ V S* ⎠ (19)
where V S* = GS*/ρ with G*S = G*S ( ω) being the complex shear
modulus. Eq. (19) suggests that experimental measurement of
ϑ( 0, t ) would allow G*S ( ω) to be determined (for example using the
Newton-Raphson method) if the amplitude of the applied torque
T0·eiωt , the geometry of the specimen and the characteristics of the
apparatus are known. Since the driving torque T0·eiωt and the angle
of twist ϑ( 0, t ) will be in general out of phase, the amplitude of
ϑ( 0, t ) will be in general complex-valued.
If the analysis is carried out for a sufficiently large number of
frequencies, it will then be possible to experimentally determine
the frequency dependence of the shear complex modulus G*S ( ω)
and correspondingly, via Eqs. (11) and (12), of VS(ω) and DS(ω).
Note that through the non-resonance column method the shear
wave velocity and material-damping ratio of a soil specimen are
measured simultaneously at the same frequency of excitation and
using the same experimental procedure thereby overcoming the
major drawbacks of the conventional resonant column test.
Fig. 6 shows the results obtained by Rix and Meng [66] using
the non-resonance method to determine the low-strain para-
meters of soil dynamic behavior of a remolded kaolin. In this work
the “shear modulus” was defined as the modulus of G*S ( ω) i.e.
G*S ( ω) . In Fig. 6 this parameter increases monotonically with fre-
quency whereas damping ratio varies over the frequency range of
0.01–30 Hz exhibiting a minimum at about 1 Hz..
Other contributions on the use of non-resonance methods to
determine the low-strain parameters of soil dynamic behavior
include the work of Khan et al. [36] who investigated the fre-
quency-dependence up to 150 Hz of two dry-sands specimens and
a sand–bentonite–mud sample at different confining pressures and
shear strain levels. The results were compared with those obtained
using the conventional resonant column apparatus. A good
agreement was found for the low-strain shear modulus however,
the conventional resonant column test did not allow measuring
 C.G. Lai, A.G. Özcebe / Soil Dynamics and Earthquake Engineering 91 (2016) 72–86
Fig. 6. Frequency-dependence of low-strain parameters of soil dynamic behavior determined with the non-resonance column test on a remolded kaolin (from [66]).
the possible dependence of this parameter on the frequency of
excitation. The authors also point out the importance of elim-
inating in the non-resonance method as well as in the conven-
tional test possible flexural modes that may be excited in the
specimen during the torsional oscillations. To solve this problem
they suggest using two accelerometers on the driving plate placed
in opposite directions.
Camacho-Tauta et al. [11] used non-resonance methods in
conjunction with standard resonant column and bender element
experiments to study rate-effects on the shear wave velocity of
three saturated sands in the frequency range 0.1–10 kHz. It was
found that VS(ω) had the tendency to increase with the increase of
the excitation frequency and this outcome was more pronounced
in coarse saturated sands. The paper also describes the develop-
ment of an automatic procedure to carry out, in a synergic fashion,
resonant column, bender element and non-resonance
experiments.
Despite the non-resonance column method is not yet widely
used to measure the low-strain parameters of soil dynamic be-
havior, it is superior with respect to the conventional test. First, it
allows determining shear wave velocity and shear damping ratio
simultaneously, at the same frequency of excitation and with ex-
actly the same experimental procedure. This is consistent with
viscoelasticity theory, the underlying constitutive model adopted
to interpret soil response to low-amplitude, dynamic excitations.
Secondly, the non-resonance method provides the natural mean
for investigating the frequency-dependence laws of VS and DS. This
is in contrast with the usual practice, where this aspect is in-
herently neglected. The taking into account of the frequency-de-
pendence of VS and DS has been shown to have practical im-
plications in problems of earthquake geotechnical engineering like
for instance in ground response analyses e.g. [53,60]. However, the
ability to actually perform these analyses is subject to the avail-
ability of an experimental procedure able to determine VS(ω) and
DS(ω). The non-resonance column method seems ideal to reach
this goal. Perhaps another possibility would be that of determining
VS(ω) and DS(ω) in bender element experiments from the ex-
ploitation of Eq. (16) after the pioneering work of Karl et al. [35] on
damping measurements using bender element tests.
7. Full waveform in-situ cross-hole seismic testing
The interdependency of VS(ω) and DS(ω) stated by the Kra-
mers-Kronig relation sets up the ground for developing a method
to determine one of the two material functions from the mea-
surement of the other. This is facilitated by the availability of the
solution of Kramers-Kronig relation represented for instance by
Eq. (16), which allows material-damping ratio spectrum DS(ω) to
be completely determined from the measurement of the
81
dispersion function VS(ω) of shear waves. This idea was suggested
by Lai and Özcebe [40] in connection with in-situ cross-hole
seismic testing.
Determination of damping ratio spectra from phase velocity
measurements of P and S waves using geophysical field testing is
particularly attractive given the difficulties and costs of measuring
damping through standard laboratory experiments. Although la-
boratory measurements using resonant column, torsional shear
and other techniques are well established, the effect of specimen
disturbance on the measured value of low-strain damping ratio
may not always be negligible. Furthermore, laboratory tests are
performed on small size specimens and this limits considerably
the applicability of the results to large volumes of soil.
Conversely, in-situ measurements provide the opportunity to
assess low-strain damping ratio free from the undesirable effects
of sample disturbance and at the same time encompassing a lar-
ger, more representative volume of soil. That being said, in the
current practice, borehole and other invasive seismic methods are
used almost exclusively to measure the speed of propagation of P
and S waves. Few attempts were made on attenuation measure-
ments and most of them in a research environment. Recent con-
tributions include the works of Hall and Bodare [25], Michaels [56]
and Crow et al. [14]. The major challenges outlined by these stu-
dies are the uncoupling of geometric and material attenuation as
well as to account for the influence of scattered waves.
Moreover, in some of these applications, damping is a-priori
assumed to be hysteretic (i.e. frequency-independent) and that
may not always be acceptable depending on the specific soil type.
In other cases, a specific constitutive model, like the Kelvin-Voigt
solid is enforced in the interpretation of attenuation measure-
ments [56]. This also corresponds to enforcing a particular rheo-
logical model of soil behavior and thus the application of an un-
necessary bias.
On the contrary, the method based on using Eq. (16), is free
from such biases since damping is obtained from the inversion of P
and S dispersion functions, which in turn are computed entirely
from phase velocity measurements. The algorithm is just based on
implementing Eq. (16). Applying the two-station method at the
layout of cross-hole seismic testing (Fig. 3) in a fashion similar to
that used with the SASW method when determining the phase
velocity of Rayleigh waves [23], the dispersion functions of S and P
waves can be determined from the following equations:
ω⋅ΔL ω⋅ΔL
VS ( ω) = , VP ( ω) =
(
arg GRS1 − R2) arg( G ) (20)
P
R1 − R2
where arg( G S
) is the (unwrapped) phase of the cross-power
R1 − R2
spectrum of the signals detected at the two receivers after filtering
the contribution of P-wave arrivals. An identical definition holds
( )
for arg GRP1 − R2 after interchanging the role of P and S waves.
82 C.G. Lai, A.G. Özcebe / Soil Dynamics and Earthquake Engineering 91 (2016) 72–86
Fig. 7. (Top) Dispersion curve of S waves obtained from cross-hole seismic data
measured at Pisa, Italy in the year 2000. (Bottom) Shear damping ratio spectrum
from the inversion of VS ( ω) .
Once VS(ω) and VP(ω) are computed for a sufficiently broad
frequency band, Eq. (16) allows to calculate damping ratio spectra
DS(ω) and DP(ω). With this procedure, no a-priori assumption is
made concerning specific rheological models of soil behavior or
frequency-dependence law of damping ratio. Indeed the reverse is
true since the shape of the damping ratio spectra is completely
determined by the shape of the dispersion curves of S and P waves,
as predicted by the Kramers-Kronig relation [41]. A distinctive
feature of the procedure is that unlike the conventional inter-
pretation of cross-hole seismic data based on picking the first ar-
rivals of P and S phases, the method exploiting the inversion of P
and S dispersion functions uses the complete waveforms of the
signals detected at the two receivers.
Fig. 7 shows the results obtained at Pisa, Italy after the appli-
cation of the procedure to cross-hole data acquired at a depth of
14 m from the ground surface in the year 2000. The numerical
integration in Eq. (16) was performed using a robust quadrature
scheme given the singularity of the integrand. To increase the
accuracy and stability of the algorithm, the domain of integration
was split into small sub-domains using an adaptive logarithmic
frequency scale. The shear wave velocity measured at the same
depth with the conventional interpretation of cross-hole data is
161.5 m/s which compares well with the experimental dispersion
curve. Due to limitations of the frequency bandwidth emitted at
the source, the usable frequency range to construct the dispersion
curve of S waves is restricted to 50–500 Hz (Fig. 7, top). This in
principle limits the ability to resolve the damping ratio spectrum
at frequencies below 50 Hz, which includes the seismic band. The
damping ratio shown in Fig. 7 (bottom) has been calculated by
enforcing the hysteretic assumption at frequencies below 50 Hz.
In Fig. 7 (bottom) the values of damping ratio compare well
with those obtained from independent measurements carried out
at about the same depth by Lo Presti et al. [51] through laboratory
tests. It should be remarked however that these lab measurements
were performed at strains on the order of 0.01%. Figs. 8 and 9
Fig. 8. Wrapped (top) and unwrapped (bottom) phase measured at a site in Italy.
illustrate another application of the method at a different site in
Italy in 2008. Specifically, Fig. 8 shows the wrapped (top) and
unwrapped (bottom) phase of the measured cross-power spec-
trum. The “up” and “down” reported in the legend refer to mea-
surements carried out by reversing the polarity. The resolvable
frequency range is limited in this case to 30–300 Hz. The corre-
sponding shear dispersion curve and damping ratio spectrum are
shown in Fig. 9. Given the narrowness of the experimental fre-
quency spectrum, it was necessary to extend the dispersion curve
beyond the measured bandwidth. This was done according to two
alternative hypotheses. The first, denoted as “log linear extra-
polation” considers a linear increase of VS with the logarithm of the
frequency and perhaps it corresponds to the “least-biased” as-
sumption considering that some type of frequency-dependence of
VS must be assumed to comply with the causality constraint. The
second assumption, denoted in the legend as “Shibuya (1995)-like
model”, refers to enforcing frequency-independence of damping
ratio in the seismic bandwidth when performing the inversion of
the shear dispersion curve. As discussed in previous sections, this
hypothesis is often invoked for rock-like materials in seismology
and it was also proposed in soil dynamics by Shibuya et al. [70].
To comply with the theory of viscoelasticity, damping ratio has
been set to approach zero at very low and very large frequencies
(purely elastic response). No independent measurement of
damping ratio is available at the site to compare the results shown
in Fig. 9 (right).
The loss of uniqueness in the inversion of the dispersion
function VS(ω) and thus of the calculated damping ratio spectrum
highlighted in Fig. 9, may be overcome by adopting broadband
seismic sources and receivers with an ultra-large dynamic range.
The results shown above are preliminary and efforts are un-
derway to develop a robust and reliable method to measure DS(ω)
and DP(ω) in cross-hole and down-hole seismic testing from the
inversion of VS(ω) and VP(ω).
8. Applications in computational soil dynamics
Computational seismology and soil dynamics have advanced
considerably in latest years allowing the solution of difficult pro-
blems like simulating ground shaking scenarios of 3D large geo-
logical structures using refined constitutive models for geoma-
terials. The latter require the introduction of a consistent set of
model parameters and failure to do so may yield bad con-
sequences like loss of stability and convergence of the algorithm,
violation of fundamental physical principles (e.g. causality) or just
simply inaccurate results. These situations are undesirable because
they nullify the efforts of using advanced numerical modeling. To
illustrate this, a couple of examples are shown in the following
drawn from two classical problems of soils dynamics and earth-
quake geotechnical engineering.
8.1. Lamb's problem
The problem of determining the displacement and stress fields
induced by a vertical, time-harmonic point load applied at the free
surface of a homogeneous, linear elastic half-space, was first
solved by Lamb [45]. For this reason this problem has become
known in the literature as the Lamb’s problem which can be
considered as the dynamic version of the Boussinesq’s problem,
another classical theoretical problem of linear elasticity. Applica-
tion of the elastic-viscoelastic correspondence principle to the
solution of Lamb’s problem allows extending this solution to a
viscoelastic half-space.
Fig. 10 shows the dispersion and attenuation curves of Rayleigh
waves in a homogeneous, viscoelastic half-space. They have been
 C.G. Lai, A.G. Özcebe / Soil Dynamics and Earthquake Engineering 91 (2016) 72–86 83

Fig. 9. Cross-hole seismic measurements at a site in Italy. (Left) Experimental shear dispersion curve. (Right) Shear damping ratio spectrum obtained from the inversion of
VS ( ω).

Fig. 10. Dispersion (Left) and attenuation (Right) curves of Rayleigh waves in a homogeneous, linear viscoelastic half space. The curves have been computed for different
damping ratio models.

calculated using different models of damping ratio. The ones in the

legend obey the Kramers-Kronig relation (Eq. (14)) and thus
physical causality. Specifically, the “hysteretic” model corresponds
to assuming a frequency-independent damping ratio spectrum
over the seismic bandwidth of [10 2–102] Hz. The magnitude was
assumed equal to 2% and 5% respectively. Instead, the “Rayleigh
damping”, is a model characterized by a parabolic-like shape with
an upward concavity, if frequency is plotted in logarithmic scale
[41]. In this particular example, a minimum damping ratio of 0.5%
was set at 1 Hz whereas damping maxima of about 4% were set at
0.05 Hz and 11 Hz respectively. The other input data of the half-
space are a shear wave velocity at 1 Hz of 300 m/s, a real-valued
Poisson’s ratio of 0.25 and a mass density of 1900 kg/m3. Both
dispersion and attenuation curves in Fig. 10 exhibit a rather strong
dependence on the assumed damping ratio model. The curves Fig. 11. Synthetic seismograms obtained from the solution of the Lamb’s problem
denoted as “non-causal 2%” correspond to the usual practice of in a homogeneous, linear viscoelastic half-space using different damping ratio
assuming rate-independent damping ratio over the whole fre- models.
quency band. As discussed in previous Sections, this model violate
physical causality. As expected, there is a difference between the
dispersion and attenuation curves calculated for causal and non-
spectrum is spread between 0 and 30 Hz. Fig. 11 shows the free
causal 2% hysteretic model.
surface response of the viscoelastic half space at 60 m from the
Fig. 11 illustrates the synthetic seismograms resulting from the
analytical solution of the Lamb’s problem (in the far-field) for a source and for the same damping ratio models used to compute
concentrated, vertical point load whose time variation is that of a the dispersion and attenuation curves of Fig. 10. Clearly, the
Ricker wavelet of unit peak amplitude and maximum of Fourier computed seismograms show a sensitivity on the adopted damp-
amplitude spectrum centered at 10 Hz. The energy of this ing ratio model.
84 C.G. Lai, A.G. Özcebe / Soil Dynamics and Earthquake Engineering 91 (2016) 72–86
Fig. 12. Amplification function between bedrock and soil outcrops at Mirandola
(Northern Italy) for different damping ratio models in linear viscoelastic, 1D ground
response analysis.
8.2. Causal ground response analysis
Standard computer programs to perform 1D linear and linear
equivalent ground response analyses are founded on viscoelasti-
city theory. These include all members of the family of programs
originated from the popular SHAKE [69]. In these codes, the im-
plemented shear wave velocity and shear damping ratio are fre-
quency-independent, uncoupled parameters. Ground response is
thus computed based on this assumption which implies violation
of physical causality.
Fig. 12 compares the performance of conventional, 1D non-
causal ground response analysis with a causal assessment of 1D
site amplification. The figure shows the absolute value of the
transfer function between the motion at the ground free surface
and that at the outcropping bedrock. The calculation has been
carried out with respect to the shear wave velocity profile mea-
sured at Mirandola, a town in Northern Italy, up to a depth of
about 120 m. The measured values of VS have been assigned at a
reference frequency of 300 Hz since they were obtained via a
standard cross-hole seismic test. The bedrock was considered to
have a frequency-independent shear wave velocity of 1000 m/s
and a constant damping ratio of 0.5%. The shear dispersion and
damping ratio functions of the soil layers were assumed according
to the four models adopted for the solution of Lamb’s problem
discussed in the previous Section..
From Fig. 12 the implications of physical causality in 1D linear
ground response analyses are readily apparent. Of particular in-
terest is the difference in the amplification function between the
hysteretic casual 2% and the non-causal (i.e. conventional) 2%
damping ratio models. The peaks of the causal damping exhibit a
shift toward lower frequencies and the peak values are slightly
larger. It is also curious to note that higher damping does not al-
ways lead to smaller-amplification.
9. Concluding remarks
This article concerned with the experimental measurement of
the low-strain dynamic properties of soils. It attempted to in-
troduce a novel approach to the problem, which has been built
around three main ideas. First, the definition of the low-strain
parameters of soil dynamic behavior was revisited and re-
formulated within the framework of linear viscoelasticity. Sec-
ondly, a critical review of widespread in-situ and laboratory
methods to measure these parameters was performed high-
lighting major flaws and inconsistencies. Thirdly, the principles of
recently proposed, emerging methods that redefine the execution
and interpretation of the standard cross-hole seismic test and the
laboratory resonant column technique, were introduced and ap-
plied to few cases. These emerging methods offer the opportunity
to overcome some of the main limitations of the corresponding
parental tests.
The motivation for this subject originated from the observation
that the current, widespread approach to model the low-strain
dynamic behavior of soils is characterized by a number of incon-
gruities. For instance, elastic shear and Young moduli are mea-
sured and used in geotechnical dynamic analyses while material-
damping ratio is also measured and used in the same analyses
despite this is a parameter of inelastic behavior. It is not only a
problem of consistency between constitutive models. Indeed, it
also has practical implications. For instance, computational soil
dynamics and engineering seismology have experienced a tre-
mendous progress in recent years and as a result, the recourse to
high-performance, advanced numerical modeling to solve diverse
and complicated boundary-value problems has become relatively
frequent. This requires to introduce adequate constitutive models
for geomaterials with consistent model parameters which in turn
calls for proper experimental measurements.
Conventional field and laboratory methods provide the correct
parameters for standard calculations and methods of analyses.
However, they may fail to provide the analyst the right set of
model parameters for advanced calculations, even if restricting the
analysis to low-strain dynamic response. This happens since a full
characterization of a linear viscoelastic material requires two
material functions, namely VS(ω) and DS(ω), even for the simple
case of uni-axial loading. Yet, in the usual practice, it is customary
to neglect the frequency dependence of VS and DS as well as the
interdependency of these parameters as established by physical
causality through the Kramers-Kronig relation.
Problems may arise also with more advanced laboratory
methods such as bender element tests. In fact, with these tech-
niques soil parameters are measured at very large frequencies, far
above most practical applications, and the frequency-insensitivity
of soil constitutive parameters cannot be taken for granted, par-
ticularly when the rate of loading is on the order of kHz while in
earthquake geotechnical engineering the seismic bandwidth is
two orders of magnitude smaller.
In light of these considerations, the paper presented the main
features of a laboratory non-resonance method and of the in-situ
cross-hole seismic test based on full waveform inversion of the
recorded seismograms. Although these techniques seem to over-
come most of the deficiencies of the conventional versions of these
tests and the preliminary results are encouraging, there are still
technical difficulties that need to be overcome. Perhaps the most
important of these, shared by both experiments, concern with the
limited frequency bandwidth at which the measurements are
currently made. This prevents a broadband characterization of
material functions VS(ω) and thus DS(ω) which is the ultimate goal
of these tests’ upgrades. There are reasons to be cautiously opti-
mistic that these difficulties will be finally resolved with the ad-
vance of technological progress.
Acknowledgments
The authors would like to acknowledge Prof. M. Jamiolkowski
and SOLGEO Ltd for providing the cross-hole seismic data. The
authors are also grateful to Prof. F. Tatsuoka for kindly providing
his papers on strain rate-dependency of soils. Finally, a word of
appreciation goes to Prof. G.J. Rix for supplying the figures of the
article by Rix and Meng [66] and to an anonymous reviewer whose
comments helped to improve the manuscript.
 C.G. Lai, A.G. Özcebe / Soil Dynamics and Earthquake Engineering 91 (2016) 72–86
References
[1] Aki K, Richards PG. Quantitative seismology. Sausalito 2002. CA: University
Science Books.
[2] ASTM D4015-07. Standard test methods for modulus and damping of soils by
resonant-column method. West Conshohocken, PA: ASTM International; 2014.
〈www.astm.org〉.
[3] ASTM D4428-14. Standard test methods for crosshole seismic testing. West
Conshohocken, PA: ASTM International; 2014. 〈www.astm.org〉.
[4] ASTM D7400-14. Standard test methods for downhole seismic testing. West
Conshohocken, PA: ASTM International; 2014. 〈www.astm.org〉.
[5] Araei AA, Razeghi HR, Tabatabaei SH, Ghalandarzadeh A. Loading frequency
effect on stiffness, damping and cyclic strength of modeled rockfill materials.
Soil Dyn Earthq Eng 2012;33(1):1–18.
[6] Assimaki D, Kausel E. An equivalent linear algorithm with frequency and
pressure-dependent moduli and damping for the seismic analysis of deep
sites. Soil Dyn Earthq Eng 2002;22(9–12):959–65.
[7] Badsar SA, Schevenels M, Haegeman W, Degrande G. Determination of the
material damping ratio in the soil from SASW tests using the half-power
bandwidth method. Geophys J Int 2010;182(3):1493–508.
[8] Ballard RF Jr. Method for crosshole seismic testing. Geotechnical Engineering
Division, ASCE, Vol 102, No. GT12; December 1976. p. 1261–73.
[9] Biot MA. Theory of propagation of elastic waves in a fluid-saturated porous
solid. I. Low-frequency range. J Acoust Soc Am 1956;28(2):168–78.
[10] Biot MA. Theory of propagation of elastic waves in a fluid-saturated porous
solid. II. Higher frequency range. J Acoust Soc Am 1956;28(2):179–91.
[11] Camacho-Tauta JF, Reyes-Ortiz OJ, Santos JA. Evaluation of the frequency ef-
fects on the shear wave velocity of saturated sands. In: Proceedings of the 15th
World conference on earthquake engineering. Lisbon, Portugal; September,
2012. p.24-28.
[12] Christensen RM. Theory of viscoelasticity – an introduction. 2nd editionDover
Publ; 2010. p. 384.
[13] Clough RW, Penzien J. Dynamics of structures. 2nd edition. Singapore: Mc
Graw Hill Inc; 1993. p. 739.
[14] Crow H, Hunter JA, Motazedian D. Monofrequency in situ damping measure-
ments in ottawa area soft soils. Soil Dyn Earthq Eng 2011;31(12):1669–77.
[15] Darendeli MB. Development of a new family of normalized modulus reduction
and damping curves [Ph.D. dissertation]. Austin, Texas: University of Texas at
Austin; 2001.
[16] Dobry R. Damping in soils: its hysteretic nature and the linear approximation.
Massachusetts Institute of Technology; 1970. p. 82 [Research report R70-14].
[17] d’Onofrio A, Silvestri F, Vinale F. Strain rate dependent behavior of a natural
stiff clay. Soils Found 1999;39(2):69–82.
[18] Drnevich VP., Hall JR Jr., Richart FE. Jr. Effects of amplitude of vibration on the
shear modulus of sand. In: Proceedings of the International Symposium on
Wave Propagation and Dynamic Properties of Earth Materials 1967, Albu-
querque, N.M.
[19] Drnevich VP. Recent developments in resonant column testing. In: Proceed-
ings of richart commemorative lectures. Sponsored by Geotech. Eng. Div. in
conjunction with ASCE Convention, Detroit, Michigan; October 23, 1985.
[20] Dyvik R, Madshus C. Lab Measurements of Gmax using bender elements. In:
Proceedings of ASCE Annual Convention. Advances in the Art of Testing Soils
under Cyclic Conditions. Detroit, USA; 1985. p. l86–96.
[22] Foti S, Lai CG, Lancellotta R. Porosity of fluid-saturated porous media from
measured seismic wave velocities. Géotechnique 2002;52(5):359–73.
[23] Foti S, Lai CG, Rix GJ, Strobbia C. Surface wave methods for near-surface site
characterization. Boca Raton, Florida: CRC Press, Taylor & Francis Group; 2014.
p. 467.
[24] Futterman WI. Dispersive body waves. J Geophys Res 1962;67(No. 13):5279–
91.
[25] Hall L, Bodare A. Analyses of the cross-hole method for determining shear
wave velocities and damping ratios. Soil Dyn Earthq Eng 2000;20:167–75.
[26] Hall Jr. JR, Richart Jr. FE. Dissipation of elastic wave energy in granular soils. J
Soil Mech Found Eng 1963;89(SM6):27–56 [ASCE].
[27] Hardin BO, Black WL. Vibration modulus of normally consolidated clay. J. Soil
Mech Found Eng 1963;94(SM2):353–69 [ASCE].
[28] Hardin BO, Drnevich VP. Shear modulus and damping in soils: measurement
and parameter effects. J. Soil Mech Found Eng 1972;98(SM6):603–24 [ASCE].
[29] Hoar RJ, Stokoe KH II. Generation and measurement of shear waves in-situ.
Dynamic Geotechnical Testing, ASTM, STP 654; 1977. p. 3–29.
[30] Iida K. The velocity of elastic waves in sand. Bulletin of Earthquake Research
Institute; 1938. Tokyo Imperial University. Vol. 16. p. 131–44.
[31] Ishihara KG. Soil behavior in earthquake geotechnics. Oxford: Clarendon Press;
1996. p. 350.
[32] Ishimoto M, Iida K. Determination of elastic constants of soils by means of
vibration methods: Part 1, Young's Modulus. Bulletin of Earthquake Research
Institute; 1936. Tokyo Imperial University, Vol. 14. p. 632–57.
[33] Johnston DH, Toksöz MN, Timur A. Attenuation of Seismic Waves in Dry and
Saturated Rocks: II. Mechanisms. Geophysics 1979;44(4):691–711.
[34] Jongmans D. In-situ attenuation measurements in soils. Eng Geol 1990;29:99–
118.
[35] Karl L., Haegeman W., De Grande G. and Dooms D. Determination of the
material damping ratio with the bender element test. J Geotech Geoenviron
Eng. ASCE 2008; 134(12). p. 1743–56.
[36] Khan ZH, Cascante G, El Naggar MH, Lai CG. Measurement of frequency-
85
dependent dynamic properties of soils using the resonant-column device. J
Geotechn Geoenviron Eng 2008;134(9):1319–26 [ASCE].
[37] Kim DS., Stokoe KH., Hudson WR. Deformational characteristics of soils at
small to intermediate strains from cyclic tests. Research report 1177-3. Center
for Transportation Research, Bureau of Engineering Research. Austin, TX: The
University of Texas at Austin; 1991.
[38] Kramer SL. Geotechnical earthquake engineering. New Jersey: Prentice-Hall;
1996. p. 653.
[39] Lai CG, Callerio A, Faccioli E, Morelli V, Romani P. Prediction of railway-in-
duced ground vibrations in tunnels. J Vib Acoust 2005;127(5):503–14.
[40] Lai CG., Özcebe AG. In-Situ Measurement of damping ratio spectra from the
inversion of phase velocities of P and S waves in cross-hole seismic testing. In:
Miura S. et al. (Eds.) Proceedings of the 2nd international conference on
transportation geotechnics. IS-Hokkaido 2012; Hokkaido-Japan: p. 578–82.
[41] Lai CG, Özcebe AG. Causal damping ratio spectra and dispersion functions in
geomaterials from the exact solution of Kramers-Kronig equations of viscoe-
lasticity. In “Continuous Media with Microstructure 2”. Editors B. Albers and M.
Kuczma. Springer International Publishing Switzerland; 2016. p. 418. ISBN
978-3-319-28241-1.
[42] Lai CG, Pallara O, Lo Presti DCF, Turco E. Low-strain stiffness and material
damping ratio coupling in soils. In: Tatsuoka F, Shibuya S, Kuwano R, editors.
Advanced Laboratory Stress-Strain Testing of Geomaterials. Lisse: Balkema;
2001. p. 265–74.
[43] Lai CG, Rix GJ. Simultaneous inversion of rayleigh phase velocity and at-
tenuation for near-surface site characterization. [GIT-CEE/GEO-98-2]. School
of Civil and Environmental Engineering, Georgia Institute of Technology; 1998.
[44] Lai CG, Rix GJ. Solution of the Rayleigh Eigenproblem in Viscoelastic Media.
Bull Seismol Soc Am 2002;92(6):2297–309.
[45] Lamb, H. 1904. On the Propagation of tremors over the surface of an elastic
solid. Philosophical Transactions of the Royal Society of London. Series A,
Containing papers of a mathematical or physical character 203 (ArticleType:
research-article/Full publication date: 1904/Copyright © 1904 The Royal So-
ciety): p. 1-42.
[46] Leroueil S, Marques MES. State of the art: importance of strain rate and
temperature effects. Geotech Eng Meas Model Time Depend Soil Behav
1996:1–60.
[47] Leurer KC. Attenuation in fine-grained marine sediments: extension of the
biot-stoll model by the effective grain model (EGM). Geophysics 1997;62
(5):1465–79.
[48] Lin ML, Huang TH, You JC. The effects of frequency on damping properties of
sand. Soil Dyn Earthq Eng 1996;18:269–78.
[49] Ling HI, Callisto L, Leshchinsky D, Koseki J. Soil stress-strain behavior: mea-
surement, modeling and analysis. Springer: 2007. In: Proceedings of a col-
lection of papers of the geotechnical symposium in Rome; March 16–17, 2006.
ISBN: 9781402061462. p. 981.
[50] Liu HP, Warrick RE, Westerlund RE, Kayen RE. In-situ Measurement of Seismic
Shear-Wave Absorption in the San Francisco Holocene Bay Mud by the Pulse-
Broadening Method. Bull Seism Soc Am 1994;84(1):62–75.
[51] Lo Presti DCF, Pallara O. Damping ratio of soils from laboratory and in-situ
tests. In: Proceedings of the 14th international conference on soil mechanics
and foundation engineering. Hamburg; September; 1997. p. 6–12.
[52] Matesic L, Vucetic M. Strain-rate effects on soil secant shear modulus at small
cyclic strains. J Geotech Geoenviron Eng 2003;129:536–49.
[53] Meng J. Earthquake ground motion simulation with frequency-dependent soil
properties. Soil Dyn Earthq Eng 2007;27:234–41.
[54] Meza-Fajardo KC, Lai CG. Explicit causal relations between material-damping
ratio and phase velocity from exact solutions of the dispersion equations of
linear viscoelasticity. Geophys J Int 2007;171:1247–57.
[55] Michaels P. In-situ determination of soil stiffness and damping. J Geotech
Geoenviron Eng 1998;124(8):709–19.
[56] Michaels P. Comparison of Viscous Damping in Unsaturated Soils, Compres-
sion and Shear. In: Miller, GA Zapata, CE, Houston, SL, Fredlund DG (Eds.).
Unsaturated Soils 2006 (GSP 147). Proceedings of the 4th international con-
ference on unsaturated soils, Carefree, Arizona, USA; 2–6 April 2006. Reston,
VA.
[57] Misbah AS, Strobbia CL. Joint estimation of modal attenuation and velocity
from multichannel surface wave data. Geophysics 2014;79(3) EN25-EN38.
[58] Mok YJ, Sánchez-Salinero I, Stokoe KHII, Roesset JM. In situ damping mea-
surements by crosshole seismic method. In: Von Thun JL, editor. Earthquake
engineering and soil dynamics II – recent advances in ground motion eva-
luation. New York: Geotech. Spec. Publ, ASCE; 1988. p. 305–20 [No. 20].
[59] O’Connell RJ, Budiansky B. Measures of dissipation in viscoelastic media.
Geophys Res Lett 1978;5:5–8.
[60] Park D, Hashash YMA. Rate-dependent soil behavior in site response analysis.
Can Geotech J 2008;45:454–69.
[61] Pipkin AC. Lectures on Viscoelasticity Theory. 2nd edition. Berlin: Springer-
Verlag; 1986. p. 188.
[62] Read BE, Dean GD. The determination of dynamic properties of polymers and
composites. New York: John Wiley & Sons; 1978.
[63] Read WT. Stress analysis for compressible viscoelastic materials. J Appl Phys
1950;21:671–4.
[64] Richart Jr. FE, Woods RD, Hall JR. Vibrations of Soils and Foundations. Engle-
wood Cliffs, New Jersey: Prentice-Hall; 1970. p. 414.
[65] Rix GJ, Lai CG, Spang AW. In situ measurements of damping ratio using surface
waves. J Geotech Geoenviron Eng 2000;126(5):472–80 [ASCE].
[66] Rix GJ, Meng J. A non resonance method for measuring dynamic soil
86 C.G. Lai, A.G. Özcebe / Soil Dynamics and Earthquake Engineering 91 (2016) 72–86
properties. Geotech Test J 2005;28(1):1–8.
[67] Santamarina JC, Klein KA, Fam MA. Soils and waves. Chichester: Wiley; 2001.
[68] Schakel MD, Smeulders DMJ, Slob EC, Heller HKJ. Seismoelectric interface
response: experimental results and forward model. Geophysics 2011;76
(4):29–36.
[69] Schnabel PB, Lysmer J, Seed HB. SHAKE: A computer program for earthquake
response analysis of horizontally layered sites [Report No. EERC 72-12]. Ber-
keley, USA: Earthquake Engineering Research Center, University of California;
1972.
[70] Shibuya S, Mitachi T, Fukuda F, Degoshi T. Strain rate effects on shear modulus
and damping of normally consolidated clay. Geotech Test J 1995;18(3):365–75.
[71] Stoll RD. Acoustic Waves in saturated sediments. Physics of sound in marine
sediments, Plenum Press; 1974. p 19–39.
[72] Stupazzini M, Paolucci R, Igel H. Near-fault earthquake ground-motion simu-
lation in the grenoble valley by a high-performance spectral element code.
Bull Seismol Soc Am 2009;99:286–301.
[73] Tallavo F, Cascante G, Sadhu A, Pandey M. New analysis methodology for
dynamic soil characterization using free-decay response in resonant-column
testing. J Geotech Geoenviron Eng 2014;140(1):121–32.
[74] Tatsuoka F, Di Benedetto H, Enomoto T, Kawabe S, Kongkitkul W. Various
viscosity types of geomaterials in shear and their mathematical expression.
Soils Found 2008;48(1):41–60.
[75] Tatsuoka F. Rate effects on elastic and inelastic stress-strain behaviours of
geomaterials observed in experiments. In: Ling, H-I, Smyth, A, Betti, R, editors.
PoroMechanics IV. United States: DEStech Publication; 2009.
p. 43–56 [Burmister Lecture].
[76] Tschoegl NW. The phenomenological theory of linear viscoelastic behavior –
an introduction. Berlin: Springer-Verlag; 1989. p. 769.
[78] Viggiani G. Dynamic measurements of small strain stiffness of fine grained
soils in the triaxial apparatus. Experimental characterization and modelling of
soils and soft rocks. In: Proceedings of the Workshop in Napoli; 1991, p. 75-97.
[79] Visintin A. Differential models of hysteresis. Berlin: Springer-Verlag; 1994. p.
407.
[80] Vucetic M. Cyclic threshold shear strains in soils. J Geotech Eng
1994;120:2208–28 [ASCE].
[81] Wang YH, Santamarina JC. Attenuation in sand: an exploratory study on the
small-strain behavior and the influence of moisture condensation. Granul
Matter 2007;9:365–76.
[82] White JE. Underground sound: applications of seismic waves. Elsevier Science;
1983. p. 253.
[83] Woods RD. Measurement of dynamic soil properties. In: Proceedings of the
ASCE geotechnical engineering division specialty conference on earthquake
engineering and soil dynamics, Vol I: 19–21 June 1978. p. 91–178.
[84] Wu B. A correction of the half-power bandwidth method for estimating
damping. Arch Appl Mech 2015;85:315–20.
[85] Yoshida N, Kobayashi S, Suetomi I, Miura K. Equivalent linear method con-
sidering frequency dependent characteristics of stiffness and damping. Soil
Dyn Earthq Eng 2002;22:205–22.
[86] Zambelli C, di Prisco C, d’Onofrio A, Visone C, Santucci de Magistris F. De-
pendency of the mechanical behaviour of granular soils on loading frequency:
experimental results and constitutive modeling. In: Hoe I Ling, et al., editors.
Soil stress-strain behavior: measurement, modeling and analysis. Springer;
2007. p. 567–82.