JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 102, NO.

B3, PAGES 5243-5250, MARCH 10, 1997

Stress wave attenuation in shock-damaged rock

Cangli Liu and Thomas J. Ahrens
Lindhurst Laboratory of Experimental Geophysics,Seismological
Laboratory
California Institute of Technology,Pasadena

Abstract. The velocity and attenuationof ultrasonicstresswavesin gabbroic

rock samples(San Marcos,California)subjectedto shockloadingin the 2 GPa
rangewerestudied. From P wavevelocitymeasurementswe determinedthe damage
parameterDp and crackdensitye of the samplesandrelatedtheseto the attenuation
coefficient
(qualityfactor)underdynamic
strainsof 2 x 10-? andat a frequency
of
2 MHz using the ultrasonic pulse-echomethod. A fit to the data yields the P wave
spatialattenuationcoefficient
at a frequency
of 2 MHz, ap(Dp) = 1.1+ 28.2Dp
(decibelsper centimeter).Fromthe relationbetweenthe attenuationcoefficientand
quality
factor,
thequality
factor
Qisgiven
byQ-• - 0.011(1
+ 25.6Dp)(1
- Dp)•/•'.
Using O'Connell-Budianskytheory relating crack density to velocity, the parameter
in Walsh's theory was determined based on experimental data. An approximate
method is also proposedto estimate the averagehalf-length of cracksbased on the
attenuation measurements.

Introduction
d lnA(L)
-d-• - A(L)
I dA(L) da (3)
dL - a-•-L•.
Experimental measurementsof attenuation of ultra-
sonic waves in various rocks have been carried out since
From publishedresultsspecifyingthe attenuation co-
the 1940s using different techniquesover a wide fre- efficient[GordonandDavis, 1968;Brennanand Stacey,
quencyrange [e.g., Born, 1941; Nur and Simmons, 1977; Toksoz et al., 1979; Winklet et al., 1979; John-
1969; $petzler and Anderson, 1968; Johnston et al., ston et al., 1980; Jackson,1993],it can be concluded
1978; Toksozet al., 1979; Winklet and Nut, 1982; Mur- that da/dL -0. Then (3) can be written as
phy, 1984]. Typical methodsemployedincludethe
ultrasonicresonantbar [Johnstonand Toksoz,1980],
the risetime of ultrasonicpulses[Gladwinand Stacey, dlnA(L)
1974],the pulse-echo technique[Papadakis et al., 1973] • - - dL ' (4)
and an improvedpulse-echotechnique[Winklet and Two methods can be used to evaluate crack density
Plona, 1982]. The commonlyusedparametersfor de- in rocks,one is suggestedby O'Connell and Budiansky
scribing attenuation are the attenuation coefficient a [1974,1977]who haveestablished the relationbetween
and the quality factor Q. These quantities are related the crack density and stress wave velocity; the other
by usedby Gradyand Kipp [1987]and Rubin and Ahrens
I aC
[1991]is the damageparameterdefinedas
•- •f, (1)
C •
where C is wave velocity and f is frequency.
For a plane wave propagating in a medium, the am- D- 1- (•00)
, (5)
plitude of stressis given by where C is the wavevelocityof the rock with cracksand
Co is the intrinsic wave velocity of the rock. Because
A(L,t) - Aoe
-•L ei(•L-•t), (2) (C/Co)• is theratioof elastic
moduliof rocksbefore
whereL is propagationdistance,cvis angularfrequency, and after the damage,D describesthe relative modulus
k iswavenumber,andt istime. Heretheei(kL-•t) term changeof rocks.
Because stress wave attenuation results from the oc-
representsa propagating wave and the attenuation is
determined
bytheAoe-• term.Let A(L) beAoe-•, currence of cracks, the relationship between attenua-
then a can be calculated from tion coefficientand damageparameter has an important
role in understandingthe propagation of stresswavesin
damagedrocks. The presentstudy presentsthe first ex-
Copyright 1997 by the American GeophysicalUnion.
perimental data describingstresswave attenuation in
Paper number 96JB03891. damagedrocks. Attenuation coefficientsand damage
0148-0227/97/ 96JB-03891$09.00 parameters(and crackdensities)in a seriesof damaged
5243
5244 LIU AND AHRENS: ATTENUATION IN SHOCKED ROCK

San Marcos gabbro sampleswere obtained by usingthe The ultrasonic experimental apparatus used in this
ultrasonicpulse-echomethod developedby Winklet and work is similar to that developedby Winklet and Plona
Plona[1982]. [1982]for attenuationcoefficient
measurements
(Figure
1). A positioningscrewis used to hold and tighten
Experimental Technique the whole assemblyto insure good contact between the
coupling surfaces. The transducer is located along the
The rock used in this work was San Marcos gab- center line of the assembly.
bro (SMG). Shockwave propertiesand shockeffects The thicknessof the coupling lucite buffer hx is cho-
in SMG havebeenstudiedpreviously[McQueenet al., sensuchthat the reflectingsignalsfrom surfacesA and
1967; Lange et al., 1984; Polanskey and Ahrens, 1990; B can be unambiguouslyidentified. The thicknessof the
Rubin and Ahrens, 1991; Ahrens and Rubin, 1993; He sampleis h2. AlsoCpxand Cp are P wavevelocitiesfor
andAhrens,1994]. The densityof SMG is 2.87x10s the lucite and sample, respectively. The characteristic
kg/ms, the intrinsicP wavevelocityis assumed
to be times (Figure 1) for the systemare
7.12 km/s (under 0.4 GPa confiningpressure)[Birch,
1960],and shearvelocityis 3.7 km/s (under 0.4 GPa 2hx
confiningpressure)[Simmons,1964].Fromthe two ve- = (s)
locities, the intrinsic shear, bulk and ¾oung'smoduli
of SMG are 39, 93, and 103 GPa, respectively. The 2h• 2h:•
intrinsic Poisson's ratio of SMG is 0.31. - c,,' (9)
Initially, a SMG target with dimensions200 x 200 x tpa= 2tpl. (10)
150 mm was impacted by a lead projectile at a velocity
of 1.2 km/s. The projectilehada diameterof 7 mm and Typical thicknessesand velocities of the buffer and
mass of 3 g. The crater shape and fracture details in samplesare hi = I cm, h2 = I cm, Cpl = 2.69 km/s,
the target weredescribed by Ahrensand Rubin[1993]. and Cp = 6.5 km/s. From (8) - (10) we havetel = 7.4
The recoveredtarget was cut into I cm cubes,suchthat pS, tp2 -- 10.5 ps, and tp3 -- 14.8 ps. From the simple
calculation above we conclude that the wave reflected
two surfacesof the sampleswere paralleled to the im-
pact surface of the target. These cubeswere polished from surface B arrives at the transducer just after the
until the variation of thicknessfor each sample was less signalfrom surfaceA. This ensuresthat the signalsare
than 0.03 mm. Because water was used when the sam- clearly distinguished.
ples were polished, the sampleswere placed in an oven
under normal pressureat 100øC for 24 hours before the
Positional screw
measurements. This procedure is similar to that used
by Polanskeyand Ahrens[1990].
The pressureof the shock wave induced by the im- Transducer
holder
Transducer
pacting in the target can be estimated using the power / (Panametrics,
Model1102)
decay relation a---- Cable hole
(a)
/ Lucite buffer
")
P- Po(ro ,' >
- ,'o, (6)
/ Sample
where r and ro are the radial distance and the equiva- / Lucite sample
lent radius of the projectile, respectively. This expres- • Alignment cylinder
sion is mainly based on experimental data. Po is the
shock wave pressure on the impact surface, and • is • Support

assumedto be 1.5 for nearly crack-freerocks[Ahrens

and O'Keefe,1977].The initial shockwavepressurePo
is calculatedto be 11 GPa using the impedance-match
method[Ahrens,1987]andthe equations
of stateof lead
and gabbro[Ahrens,1987;Ahrensand Johnson,1995] Time
I Buffer • Sample i Back
_. up

and the velocity of projectile. The shock wave pressure (b)

tp3
[•-- - - •- .... -• O•OCK
in the gabbro target is then estimated as tp2
tp 1
P- 2.27r-•'5, (7)
wherethe unit of P is GPa and r(centimeters)is _• 0.35 Distance
Surface A SurfaceB
c•.
From (7) we infer that P decaysalongthe centerline Figure 1. Attenuation measurementsystem. (a)
of the impact from 11 GPa at a radius of 0.35 cm to Sketchof experimentalarrangement.(b) Distanceand
about 0.1 GPa at a radius of 8 cm. time diagram.
 LIU AND AHRENS: ATTENUATION IN SHOCKED ROCK 5245

A piezoelectrictransducer(Panametrics,Model 1102) is fiected from surfaceB propagate through the sample

used as the pulse generatorand the receiver. The trans- twice. Becauseboth wavespropagatethe samedistance
5052UA pulserreceiver.The in the buffer material, the attenuation can be explicitly
ducer'sdriveris a Panametrics
ultrasonicsignalsare recordedusinga digitaloscilloscopeevaluated.
(Gould 4074), and the typical samplingrate used in As in the previouspaper[ WinkletandPlona, 1982],
experimentsis 4.9 ns betweendata pointswith a signal it is assumedthat the reflectingand transmissioncoeffi-
voltageresolutionof 8 bits. cientsof the surfacesbetweenthe buffer and the sample
The strain inducedby the stresswave in the gab- are equal to those for plane wave incidence. Thus the
bro sampleis measuredusing a semiconductor strain change in amplitude with the frequency of the wave
gauge(Entran, Model ESU-025-1000).It hasa gauge results only from attenuation after correctionfor reflec-
factor of 155 and dimensions of 0.64 x 0.13 mm. The tion and transmission of the stress wave at the surfaces.

strain measurementsystemis shownin Figure 2. The Supposethat L/2 is samplethicknessand A(f) and
semiconductorstrain gaugeis bondedto the surfaceof B(f) are the frequency-dependentamplitudesof the
thegabbrosamples TAC 10•'M), pulse reflected from surfaces A and B of the sample,
usingepoxy(Devcon,
which solidifies in 24 hours under normal conditions. respectively. The attenuation coefficientobtained from
The signalfrom the strain gaugeis amplifiedand then thesetwo ultrasonicsignalsis expressed
as [ Winklet and
recordedusing the Gould oscilloscope. Plona, 1982]

Data Reduction
a(f)- •8.686
ln[B(f)
A(f)(1- R2)], (11)
P wave velocity and damage parameters . P
wave velocity is measuredby using two transducers. wherethe unit of a(f) is decibelsper centimeterwhen
One is an ultrasonicwave generatorand the other the the unit of L is centimeters. The constant results from
receiver.We first measurethe time durationfor a pulse the changein the attenuationunit (1.0 dB/cm - 8.686
to be generatedand receivedwithout an interposing Np/cm). Here R is the reflectioncoefficientfor the in-
sample between the two transducers. The time dura- terface between the coupling buffer and sample and is
tion with the sampleis then measured.The propagation given by

n - Cpp--
+Cpcp½ (12)
time of ultrasonicP wavesin each sampleis obtained
upon subtraction of the two time durations. Further
detailsare described
by RubinandAhrens[1991].The where Cp and p are the P wave velocityand the den-
damageparameter is obtained using the definition of sity of gabbrosamples,respectively.Cpcand pc are the
GradyandKipp [1987](equation(5)). P wavevelocityand densityof the couplingbuffer (lu-
Attenuation coefficient. The method used in cite), respectively.
Fromultrasonicmeasurements, Cpc
this work is similar to that givenby Winklet and Plona is 2.68 km/s and Pc is 1.19x10s kg/ms. Thesedata
[1982]. This methodis basedon the two stresswaves are similar to the results of Hartmann and Jarzynski
reflected from surfacesA and B of the sample, as shown [1972]( Cpc= 2.69km/s andPc= 1.19x10a kg/ma ).
in Figure 1. The stresswavesreflectedfrom surfaceA In the calculation of attenuation coefficients,a possible
do not propagate through the sample; the waves re- correctionof attenuationdue to wavespreadingwasnot

Wheatstone bridge

-i

i
....... •

Pulser-
receiver
•-- Trigger
Power
amplifier
I
Oscilloscope
I
Figure 2. Sketchof the strain measurementarrangement.
5246 LIU AND AHRENS' ATTENUATION IN SHOCKEDROCK

consideredbecausewe were unable to model explicitly Table 1. Experimental Results

the behavior of rocks with different crack densities.
In order to obtain attenuation coefficients, a fast X, cm Y, cm Cp, km/s Dr ap, dB/cm Q
Fouriertransform(FFT) is employedto calculatethe
4.93 1.89 6.02 0.285 8.2 13.0
frequencyspectrum of the two signals. Then the at-
4.93 4.36 6.31 0.215 7.3 15.8
tenuation coefficientversusfrequencyis obtained using 4.93 8.20 6.44 0.181 5.2 17.8
(11). The FFT subroutineusedis oneof subprograms 7.47 4.36 6.10 0.267 7.9 13.5
in the SeismicAnalysisCodedevelopedby Tull [1989]. 3.64 4.36 5.74 0.349 10.5 11.3
There were about 300- 400 data points in each win- 7.47 8.20 6.72 0.108 5.0 25.6
7.47 6.82 6.48 0.171 6.7 16.6
dow. In order to eliminate edgeeffects,the window was
3.64 6.82 6.09 0.269 8.8 13.5
extendedto 2048 points by repeating a few points near 2.35 5.58 5.87 0.319 9.4 12.0
the boundary of the signal. 3.64 5.58 6.34 0.207 7.5 16.2
Strain measurement. The strain induced by ul- 8.75 5.58 6.77 0.096 4.4 27.6
trasonic stresswaves is measuredby the semiconductor 6.20 5.58 6.44 0.181 5.2 17.8
7.47 5.58 6.49 0.168 5.7 18.8
strain gaugeshownin Figure 2 and is evaluatedusing
4.93 3.12 6.29 0.220 8.4 15.5
5R 6.20 3.12 6.46 0.176 7.2 17.4
7.47 3.12 6.29 0.220 7.7 15.5
e-/•Ro' (13) 8.75 3.12 6.46 0.177 5.9 18.1
3.64 8.20 6.56 0.151 4.6 20.3
where /• is the gauge factor of the strain gauge and 8.75 8.20 6.60 0.140 5.0 21.4
JR and Ro are the resistancechangeand the initial 3.64 3.12 5.65 0.365 12.0 11.0
resistanceof the gauge,respectively.
Becausea Wheatstone bridge is used, the relation
betweenthe resistanceandthe voltagechange[Kreuzer, The definitions of X and Y are given in Figure 8.
1988]is
In order to relate the attenuation coefficient to the

Ro- ( - 1) , (14)location of samplesin the initial target, a coordinate

systemis definedin Figure 8. Figure 9 showsthe ex-
where •V and V are the voltagechangeand the supply perimentalresultsof attenuationcoefficients versusthe
voltageacrossthe strain gauge,respectively.The strain radius from the impact point. The relation between
amplitudeis calculatedwith (13) and (14). attenuation coefficient and radial distance is found to
be
Experimental Results and Analysis ap= 48.9r-ø'•5, (17)
The P wave velocities and damage parameters for wherer(centimeters)is radial distance. Thus, as ex-
about 20 samplesare listed in Table I with their po- pected, the attenuation coefficientsof the samplesde-
sitionsin the initial target. A typical signal recorded creasewith increasingdistancesfrom the impact point.
for the attenuation measurementis shown in Figure 3. It must be noted that only the attenuation coefficients
Typical results of spectral analysisare shownin Fig- alongdirectionX (Figure 8) are measuredbecausewe
ure 4. Using (5) and (11), the attenuationcoefficients
for the sampleswith differentdamageparametershave
I I I I
beenevaluated(Table 1). Figure 5 showsthe attenu-
ation coefficientdependenceversusfrequencyfor sam- Reflected from Surface A
ples with different damageparametersfrom the FFT

/ -
0.5
calculation. From these results, we can seethat the at-
Reflected from Surface B
tenuation coefficientsincreasewith both frequency and I
I I • I
I
I I • I
damage parameter. I

0
Figure6 showsthe attenuationcoefficient
dependence
•/ t t t
on the damageparameterfor the frequencyof 2 MHz \/ '
i
,
i
m
i
(the peak energy of the ultrasonicP wave is at • 2 -0.5
,.•..t
i
•..•- ..... l•J -

MHz). The data are fitted with T1 T2

ap = 1.1+ 28.2Dp, (15) I I I
-1
0 1 2 3 4 5
whereap is in decibelsper centimeter.
From(1), (5), and (15), the qualityfactorof the dam- Time(•s)
aged samplesis
Figure 3. Typical ultrasonicrecord. T1 and T2 are
Q-x_ 0.011(1
+ 25.6Dp)(1
- Dp)1/2, (16) the data periods(usedin FFT) of the signalsreflected
which is plotted in Figure 7. from surface A and surface B, respectively.
 LIU AND AHRENS: ATTENUATION IN SHOCKED ROCK 5;247

I I I

12

From Surface A _ 4
Eq.(15) /•
10 Present
data • • ,, __
,• 8
2 a,

• 6
1

0 0
0.1 0.2 0.3 0
0 2 4 6 8

Frequency (MHz) Dp
Figure 6. Relation betweendamagedeficit and at-
Figure 4. Typical spectral amplitude of signals. tenuation coefficient.

are mainly interested in the relation between atten- locities,Cp (damagedsamples)and CpO(intrinsicve-
uation coefficientsand damage parameters. In gen- locity), and crackdensityas
eral, the damage in samplesinduced by shockwavesis
anisotropic. This will of courseresult in anisotropy in
the attenuation of shock-damagedsamples. In order to
Cp)2_
(•pO (1- .)(1+-o)K
- (1+,)(1- (18)
investigate attenuation anisotropy,additional measure-
ments are required. This may be a direction of future
research.
K =1- 16(1-
Ko
-•)e
9(1- 2,) '
(19)
Figure 10 demonstratesthat the largest strain in the . 16e
SMG samples induced by ultrasonic waves is about = I (20)
"o 9 '
2 x 10-7 (Thesample
usedin thismeasurement
wasun-
damagedSMG.). We note that the shapeof the strain where, and -o are the effective and the intrinsic Pois-
gaugesignal is dissimilarto that recordedby the piezo- son'sratio, respectively,and K and Ko are the effective
electrictransducerin Figure 3. The main reasonfor this and the intrinsic bulk moduli, respectively. The crack
is that the time resolution of the semiconductor strain density is defined as
gaugeis poorerthan that of the piezoelectrictransducer
becauseof the strain gauge's large dimensions. How-
e-N<a s >, (21)
ever, the strain amplitude can be used to approximately where N is the number of cracks per unit volume and
evaluate the strain induced in the samples. a is the half-length of cracks.
For dry circular cracks, O'Connell and Budiansky O'Connell and Budiansky's relations are derived for
[1974]haveestablished the relationbetweenP waveve- penny-shaped,circularcracks. Although cracksin shock-
damagedrocksare not circular in detail, we adopt their
I I I

12
D =0.319 I I I

Dp=,0.285 P• • 50-

10

40- Present data -

•3o-
_ D =0.220 P Eq. (16)

20-
.......... D =0.108
p

i i i
10-
5 2 2.5 3 3
i i i
Frequency (MHz) 0.1 0.2 0.3 0

Figure 5. Experimental results of attenuation coef- D

P
ficient versus frequency for the samples with different
damage parameters. Figure 7. Quality factor versusdamage parameter.
5248 LIU AND AHRENS:ATTENUATIONIN SHOCKEDROCK
I I
Impact point

...............
•.........
Crater Strain peak

ß•:•z% .• • •:•

-,,.. ß.... ,..

0 3 6 9 12
Time ( I• s )

Figure 10. Typical strain profile.

Gabbro Target

Figure 8. The definitionof coordinatesystemrelative Sincey0is known,e versusDp canbe calculated

from
to impact point and crater. (23). The relationbetweenDp ande canbe fit by the
equation
Dp- 2.4e- 1.2e
2. (24)
equationsand assumethat their equationsare goodap-
proximationsto the present damagedsamples. More- From (15) and (24) we obtainedan approximateex-
over, in the present paper, the strain and strain rate pressionrelating a to e,
of the ultrasonic wavesin the samplesare low, and al-
though O'Connell andBudiansky'stheoryis for "static" ap = 1.1+ 67.7e(1- 0.5e). (25)
moduli, we believe that it applies in the present case.
Figure11 presentsthe experimentalresultsof atten-
The damageparameter is then written as uation versuscrack density and the calculated results
from the expressionsabove(crackdensityis calculated
by solving(24) for Dp).
Dp- 1- (Cpo)
Cp2-1- (l+y)(1-y0)K0'
(1- v)(1
+v0)K (22) Walsh[1966]
gaveanexpression forthe P wavequality
factor Q as
and uponsubstituting(19) and (20) into (22), the ex-
pressionfor Dp is Q_x_B (1
- Eo(X-y)
- 2v)eF(p,y), (26)
where Q-X is the reciprocalof the quality factor and
E and E0 are the effectiveand intrinsic Young's mod-
uli, respectively.F(p, •) is the functionof the internal
(23) friction coefficientp and the effectivePossion'sratio •.

15 15
I I I I

10
A
Present data

10-
Present
•
aata
• Fitted
'curve
-

I•]D
•E]
•'• _ (Eq.
(25))Walsh's
theory
Eq. (17)

I I I I I I
6 8 10 12 0.1 0.2 0.3
r(cm)

Figure 9. The relation between attenuation coeffi- Figure 11. The relationbetweenattenuationcoeffi-
cient and radial distance. cient and crack density.
 LIU AND AHRENS: ATTENUATION IN SHOCKED ROCK 5249

1.2
For simplicity we assumethat F(p,.) is approxi- I I

mately a constant for a given material. Then, from

the relation between E and K, the expressionfor the
quality factor is
Eq. (29)

K(1
-- 2•0)'
Q-•- reK0(1 y) (27)
andfrom(19), (20),and(27), Q-• (e) canbecalculated
provided the parameter F is determined. 0.6

Basedon the experimental resultsfor the quality fac-

tor and (19), (20), and (27), F is foundto be 0.55+0.05
(for each datum, Q and z are known and F was de- 0.4 I I
0 0.05 0.1 0.15
terminedusing(27)). Comparisons of attenuationco-
efiicients and quality factors between the experimen-
tal results(Table 1), (15) and (16) and Walsh'stheory Figure 13. Averagehalf-length of cracksversuscrack
(equation(27)) are shownin Figures11 and 12. The density.
calculatedresult from Walsh'stheory is in good agree-
ment with the experimentalresultsupon the fitting of
F, although the relation between Q or a and e is non- 7.1e
linear. a- 1.1
q-67.7e(1
- 0.5e)' (29)
Becauseattenuation coefficientsdepend on the sur- where a is in centimeters. The calculated values of a as
face area of cracks, the change of attenuation coeffi-
a function of e are shownin Figure 13.
cients with damage level reflects the changeof crack
surfacearea; therefore the average crack size can be
estimated from the attenuation coefficient. If we as- Conclusions
sumeapproximately that the attenuation coefficientde- Some20 samplesof San Marcos gabbrowere cut from
pends linearly on the averagecrack surfacearea, that a target in which the in situ P wave velocity variesfrom
is,ap O(a2 (a is the average
half-length
of cracks),
and 5.65 to 6.77 km/s at radii of 5 to 11 cm from the im-
that the crackdensityis proportionalto the averagevol- pact site. The shock wave pressurevaries from about
umeof cracks,that is, e cra s, the averagehalf-length 2 GPa at 5 cm to 0.6 kbar at 11 cm. The attenuation
of cracks can be written as
coefficientsof the samples are measured using ultra-
he sonic methods. We also measure the strain induced by
= --, (2s) the ultrasonic stresswave using semiconductorstrain
(•p
gauges.We use O'Connelland Budiansky's[1974]the-
where h is an undetermined constant. ory to evaluate the crack density versus the damage
From the data for SMG given by Ahrens and Ru- parameter. Walsh's[1966]theory is usedto calculate
bin[1993](e -- 0.01 when a • 0.4 mm) and (25), h the quality factor and attenuation coefficientversusthe
is determinedto be 7.1 dB. The averagehalf-length of damage parameter. The main results obtained in this
cracks is work are given,
1. We obtain a relation (equation(15)) betweenthe
P wave attenuation coefficientand damageparameter.
50 I I I
2. The relationship(equation(16)) betweenquality
factor and damageparameter is obtained basedon the
40 experimental results.
- '"'..•,
Present
data - 3. Based on O'Connell and Budiansky'stheory, the
30
relation betweenthe crack density and damage param-
Eqs. (16) and (24) eter can be expressed
approximatelyas (24).
4. The relation between attenuation coefficients and
20
the radius from the impact point is given (equation
(17)).
10 5. The parameterin the expression(equation(27))
Walsh's theory
(Eq. (26)) given by Walsh is determinedapproximately.Walsh's
0.0 I I I theory on the attenuation coefficientis in good agree-
0.0 0.05 0.10 0.15 0.20 ment with the experimental data.
6. The averagehalf-lengthof cracksis approximately
estimatedfrom the attenuation coefficient(equation
Figure 12. Quality factor versuscrack density. (29)).
5250 LIU AND AHRENS: ATTENUATION IN SHOCKED ROCK

7. Using semiconductorstrain gauges,the strain in- Kreuzer, M., Linearity and sensitivity error in the use of sin-
duced in the sample by the ultrasonic stresswave was gle strain gaugewith voltage-fedand current-fedcircuits,
measured to be about 2 x 10-7. Strain Gauge Transducer Tech., 1, 10-15, 1988.
Lange, M. A., T. J. Ahrens, and M. B. Boslough,Impact
cratering and spall fracture of gabbro, Icarus, 58, 383-
Acknowledgments. We thank T. Duffy, T. Mukerji, 395, 1984.
and D. R. Schmitt for numeroussuggestionsthat improved McQueen, R. J., S. P. Marsh, and J. N. Fritz, Hugoniot
the manuscript, and we thank G. Ravichandran for the use equation of state of twelve rocks, J. Geophys.Res., 72,
of the ultrasonic apparatus. This researchwas supportedby 4999-5036, 1967.
NASA under NAGW-1941 and Air ForceTechnicalApplica- Murphy, W. F., Acoustic measuresof partial gas saturation
tion Center, Contr. F19628-95-c-0115. Contribution 5631, in tight sandstones,J. Geophys.Res., 89, 11549-11559,
Division of Geologicaland Planetary Sciences,California In- 1984.
stitute of Technology. Nur, A., and G. Simmons, The effect of viscosityof a fluid
phaseon velocity in low porosityrocks,Earth Planet. Sci.
References Left., 7, 99-108, 1969.
O'Connell, R. J., and B. Budiansky, Seismic velocities in
Ahrens, T. J., Shock wave techniquesfor geophysicsand dry and saturated cracked solids, J. Geophys.Res., 79,
planetary physics, in Methods of Experimental Physics, 5412-5426, 1974.
edited by C. G. Sammis and T. L. Henyey, vol. 24, pp. O'Connell, R. J., and B. Budiansky, Viscoelasticproperties
185-235, Academic, San Diego, Calif., 1987. of fluid saturated cracked solids, J. Geophys.Res., 82,
Ahrens, T. J., and M. L. Johnson, Shock wave data for 5719-5735, 1977.
rocks, in Rock Physics and Phase Relations: A Handbook Papadakis,E. P., K. A. Fowler, and L. C. Lynnworth, Ultra-
of Physical Constants, A GU Ref. Shelf Set., vol.3, edited sonic attenuation by spectrum analysis of pulsesin buffer
by T. J. Ahrens, pp. 35-44, AGU, Washington, D.C., rods, J. Acoust. Soc. Am., 53, 1336-1341, 1973.
1995. Polanskey,C., and T. J. Ahrens, Impact spallation experi-
Ahrens, T. J., and J. D. O'Keefe, Equations of state and ments: Fracture patterns and spall velocities, Icarus, 87,
impact-induced shock-waveattenuation on the moon, in 140-155, 1990.
Impact and Explosion Cratering, edited by D. J. Roddy, Rubin, A.M., and T. J. Ahrens, Dynamic tensile failure
R. O. Pepin, and R. B. Merrill, pp. 639-656, Pergamon, induced velocity deficits in rock, Geopq.ys. Res. Left., 18,
Tarrytown, N.Y.,, 1977. 219-223, 1991.
Ahrens, T. J., and A. L. Rubin, Impact-induced tensional Simmons,G., Velocity of shearwavesin -:ccksto 10 kilobars,
failure in rock, J. Geophys.Res., 98, 1185-1203, 1993. J. Geophys.Res., 69, 1123-1130, 1964.
Birch, F., The velocity of compressionalwaves in rocks to Spetzler, H., and D. L. Anderson, The effect cf tempera-
10 kilobars, J. Geophys.Res., 65, 1083-1102, 1960. ture and partial melting on velocity and attenuation in a
Born, W. T., The attenuation constant of Earth materials, simple binary system, J. Geophys.Res., 75, 6051--6060•
Geophysics,6, 132-148, 1941. 1968.
Brennan, B. J., and F. D. Stacey, Frequency dependenceof Toksoz, M. N., D. H. Johnston, and A. Timur, Attenuation
elasticity of rock-test of seismicvelocity dispersion,Na- of seismicwaves in dry and saturated rocks: Laboratory
ture, 268, 220-222, 1977. measurements, Geophysics,•, 681-690, 1979.
Gladwin, M. T., and F. D. Stacey, Anelastic degradation Tull, J. E., Manual for the Seismic Analysis Code, Univ. of
of acousticpulsesin rocks, Phys. Earth Planet. Inter., $, Calif. Press, Berkeley, 1989.
332-334, 1974. Walsh, J. B., Seismic wave attenuation in rock due to fric-
Gordon, R. B., and L. A. Davis, Velocity and attenuation tion, J. Geophys.Res., 71, 2591-2599, 1966.
of seismicwaves in imperfectly elastic rock, J. Geophys. Winkler, K. W., and A. Nur, Seismic attenuation: Effects
Res., 7'3, 3917-3935, 1968. of pore fluids and frictional sliding, Geophysics,•{7, 1-15,
Grady, D. E., and M. E. Kipp, Dynamic rock fragmenta- 1982.
tion, in Fracture of Mechanics of Rock, edited by B. K. Winkler, K. W., and T. J. Plona, Techniquefor measuring
Atkinson, pp. 429-475, Academic,San Diego, Calif., 1987. ultrasonic velocity and attenuation spectra in rocksunder
Hartmann, B., and J. Jarzynski, Ultrasonic hysteresis ab- pressure,J. Geophys.Res., 87, 10,776-10,780, 1982.
sorption in polymers, J. Appl. Phys., ,•3, 4303-4312, 1972. Winkler, K. W., A. Nur, and M. Gladwin, Friction and seis-
He, H., and T. J. Ahrens, Mechanical properties of shock- mic attenuation in rocks, Nature, 227, 528-531, 1979.
damaged rocks, Int. J. Rock Mech. Min. Sci. Geomech.
Abstr., 31, 525-533, 1994. T. J. Ahrens and C. Liu, Lindhurst Laboratory of Exper-
Jackson, I., Progress in the experimental study of seismic imental Geophysics,SeismologicalLaboratory, 252-21, Cal-
wave attenuation, Annu. Rev. Earth Planet. Sci., 21, ifornia Institute of Technology,Pasadena,CA 91125. (e-
375-406, 1993. mail: tja@caltech.edu;liu@seismo.gps.caltech.edu)
Johnston, D. H., and M. N. Toksoz, Ultrasonic P and S wave
attenuation in dry and saturated rocks under pressure, J.
Geophys.Res., $5, 925-936, 1980.
Johnston, D. H., M. N. Toksoz, and A. Timur, Attenuation
of seismic waves in dry and saturated rocks: Theoretical (ReceivedJanuary 3, 1996; revisedDecember10, 1996;
models and mechanisms,J. Geophys.,ZZ, 691-711, 1978. acceptedDecember11, 1996.)