3OURNAL

OF GEOPttYSICAL
RESEARCtt VOL. 73, No. 14, 3UL¾15, 1968

Brittle-Ductile Transition in Rocks

JAMES D. BYERLEE

U. $. GeologicalSurvey,Silver Sp•ng, Maryland 2'09'10

The deformationalcharacteristics
of two limestones,
one gabbro,and one dunirehave been
investigated as a function of confining pressure.It was found that friction of these rocks and
friction of granite and serpentinitestudiedelsewhereare nearly identical and that the brittle-
ductile transitionpressureis simply the pressureat which the stressrequiredto form a fault
is equal to the stressrequiredto causeslidingon the fault. The transitionpressureis higher
in extension than it is in compression.This difference occursbecausethe frictional shear stress
requiredto causeslidingis determinednot by confiningpressurebut by the principalstresses
and the angleof the fault. For the samefrictionalshearstresson a fault surface,the con-
fining pressureis muchhigher in extensionthan it is in compression.

INTRODUCTION Mogi was that the coefficient of friction of

yon Kdrmdn [1911], Robertson [19'55], rocksis independentof confiningpressure.The
Handin and Hater [1957], Paterson [1958], coefficientof friction is found ,by dividing the
Heard [19'60], Mogi [1966], and othersfound shear stressrequired to causesliding by the
that at low confiningpressuremany rocks are normal stress across the surfaces. Maurer
brittle. That is, when the differentia.1stressis [1965], Handin and Stearns [1964], Raleigh
sufficientlyhigh, a fault is formed, and after and Paterson[1965], and Byeflee [1967a] have
faulting the compressivestrength is decreased. shown that for rocks the coefficient of friction
At high confiningpressure,however,the same dependson confiningpressureor normal stress,
rocksmay be ductile.That is, they may fault or and henceMott's basicassumptionis incorrect.
otherwise deform without loss of compressive Byeflee [1967a] studiedthe frictional,char-
strength. acteristicsof granite and found that at 10 kb
A possiblephysicalexplanationfor the phe- confining pressure the rock deforms without
nomenonwas given by Orowan [1960]. Ite sug- loss of compressivestrength, •oecauseat this
gestedthat, at high confiningpressure,friction pressurethe strength of the rock at faulting
may increaseto suchan extent that it requires is equal to the strength after faulting.
as much stress to overcome friction as it does The frictional hypothesis for the brittle-
to cause faulting; hence, strength does not ductile transition is attractive, but previously
drop after faulting. Maurer [1965] studiedthe sufficient data on the friction of rocks were not
friction of rocksand foundthat, with the excep- availableto test the generalityof this theory.
tion .of shale, friction does not vary signifi- This lack of frictional data is unusual consider-
cantly with rock type. FollowingOrowan, he ing the vast amount of researchthat has been
suggested that the .brittle-ductiletransitionmay done on the mechanical behavior of rocks.
occur when friction along the fracture surface Frictioncanb.edetermined very simply,bycon-
exceedsthe shear strength of the rock. Un- tinuing deformationafter faulting.If the angle
fortunatelyhis experimentswere not performed of the fault and the stresses
requiredto cause
at sufficientlyhigh confiningpressureto test movement are known, friction can be deter-
this hypothesiscritically. mined b.y a simple calculation.Unfortunately,
Mogi [1966] examinedmost of the published in only a few cases do the data exist in
data on the fracture and yield strength of the published literature. There are two reasons
rocks and concluded that at least for the sili- for the a•sence of these data. The first is that
cate rocks the frictional hypothesiswas valid many investigatorsuse copperas the iacketing
but for the weaker carbonate rocks the 'brittle- material, and, when faulting occurs,the jacket
ductile transition pressurewas not determined is broken. The ,brokeniacket allows the con-
by friction. The basic assumptionmade by fining pressure medium to enter the rock so
4741
 4742 JAMES D. BYERLEE

TABLE 1. Description of Materials

Both limestones
may containdolomiteor other impurities.

Av. Grain
Diameter,
Rock Density Porosity mm Modal Analysis*

Solenhofen limestone 2.663 0.048 0.01 99 Ca?

Oak Hall limestone 2.748 0.003 0.1 99 Ca?
Nahant gabbro 3.084 0.001 5.2 40 Py, 20 S, 15 01,
10 Arno, 10 Mi, 3 0
Spruce Pine dunite 3.262 0.002 0.5 96 01, 3 S, 1 0

* Ca, calcite;S, serpentine;

O, oxides,Py, pyroxene;O1,olivine;An, anorthite;Mi, mica.

that the effective confining pressure on the cell outside the pressurevessel.Strain rate of
sample becomeszero. The other reason is that 2.4 X 10-•/see was maintained constant
most investigations have been made to deter- throughout the experiments. Axial displace-
mine the stressat faulting, and, when faulting ment of the piston was measuredwith a San-
occurs, the experiments are terminated. born DCDT 500 transducer attached to the
The present investigationdeterminesfriction piston. The confiningpressurewas measured
for a number of rocks to determine whether, with a manganincoil situatedinsidethe pres-
as in granite,the brittle-ductiletransitionpres- sure vessel.To keep the pressureconstantas
sure is simply the pressureat which the stress the piston advancedinto the pressurevessel,
required to form a fault is equal to the stress the volumeof the hydraulicsystemwas kept
required to cause sliding on the fault. constant.This was done by automatic switch-
ing of an electric motor that advancedor with-
EXPERIMENTAL METHOD
drew a pistonfrom an auxiliary pressurevessel
Differential stress was measured as a func- connectedhydraulicallyto the main pressure
tion of strain for four rocks at room tempera- vessel. The direction of the motor was con-
ture and at confining pressuresup to about trolled by a logic discriminator circuit,which
5 kb. Porosity, grain size, and modal analysis
I i I
of the rocks studied are given in Table 1.
Preciselyground cylindersof the rocks, 3.8 cm
long and 1.58 cm in diameter,were jacketed in
a gum rubber tube with a wall thickness of
3.17 mm. The ends were sealed with a wire
4.66
clamped to hardened steel end plugs. It was
found that after deformation the specimens
$,55
fell to pieceson removal of the rubber jacket.
To prevent this, a copper cylinder with a wall
/•2.20
thicknessof 0.13 mm was placedover the speci- -• - •- - --I,59
men before finally jacketing in rubber. This
copper cylinder gave the specimen a slight
mechanical strength, so that the rock stayed
intact and permitted accurate measurementof
the angle of the fault surface after completion
of an experiment. i I I i

The pressurevesselused in the experiments

øo 5 i0 15 20
Sire in, (percent)
was described by Brace [1964]. Axial force
Fig. l. Stress-straincurves for Oak Hall lime-
was applied to the piston with a ball screw
stone.Uncorrectedfor changein length or cross-
driven by an electricmotor througha reduction sectional area. Numbers at the ends of the curves
gear box. The force was measuredwith a load are the confiningpressures.
 BRITTLE-DUCTILE TRANSITION IN ROCKS 4743

determined automatically whether the pressure

had to be increasedor decreased.The pressure
could be maintained constant within --+5 bars
with this device.
The output from the differential transformer
measuring displacementwas fed directly int•o
one axis of a Mosely model 136 XYY recorder.
The manganincoil and force cellwere connected
to bridge circuits, the output from which was
fed into the other two axis of the recorder.
The axial force F measuredby the load cell
is given by

where P is the confiningpressure,A• and A,

are the area of the piston and the rock, respec-
tively, A• is the differentialstressin the speci-
men, and • is the frictional force at the 0 ring
through which the piston moves. Before the Fig. 3. Stress-strain curves for Nahant gabbro.
piston comes into contact with the specimen, Uncorrected for changein length or cross-sectional
the force measuredby the load cell is the sum area. Numbers at the ends of the curves are the
of the first two terms. If the confiningpressure confining pressures.
and velocity of the piston is maintained con-
stant throughout the experiment, these two The differential force could 'be measured with
terms remain constant. The differential stress an accuracy of _1%, and the axial displace-
times the cross-sectionalarea of the specimen ment had an accuracy of better than 0.1 mm.
was found simply by subtractingthe force re- Fault angles could not be measured with an
quired to advance the piston against the con- accuracy of better than _2 ø because of the
fining pressuremedium from the total force re- irregularityof the faults, particularlythe faults
corded. formed at low confiningpressure.
EXPERIMENTAL RESULTS

Figures 1 through 4 are typical, tracings of

the differential force axial displacementcurves
as displayedon the recorder.For clarity in the
figures,some of the experimentalresults have
not been shown. Differential stress and strain
were calculated by using the initial cross-

_
• •2.o4 sectionalarea and length of the specimen.For
large strains, the true stressin the specimen
is different from the value shown because the
.2 •1,22
• ,81 -
area of the specimen supporting the load
changesduring deformation.The dashedlines
,6 in the figures indicate where deformation oc-
57 curredwith a suddenreleaseof elasticenergy.
In all other cases the deformation occurred
stably.
00 '
5 ,
IO ,
•5 20,
Stroin,(perceni) The differential stressat fracture, or at 5%
strain if the specimenwas ductile, is plotted
Fig. 2. Stress-strain curves for Solenhofen lime-
stone. Uncorrected for change in length or cross-
in Figure 5 for all the experiments.The sig-
sectional area. Numbers at the ends of the curves nificanceof the solid line in Figure 5 will be
are the confining pressures. explainedbelow.There is a changein the cross-
4744 JAMES D. BYERLEE

strain curve.For this reason,the data near the

transition could not be used to calculate fric-
tion. More than one fault was also formed at
lower pressures,but the distanceof sliding
betweensuccessive faults was large. The initial
fault could be easily identified,and there was
no ambiguity in identifying the minimum on
the stress-strain curves. The number of ½aults
formed in the specimenwas equal to the num-
ber of downward breaks in the stress-strain
curves. It was assumed that these disconti-
nuities indicated faulting. This assumptionwas
•4 _.----- 76 checkedin the following way. When two con-
jugate faults were formed, displacementon

O0 5
-••,42
I0 15 2
the secondfault was calculated from the angle
of the fault and the axial displacement.This
distancewas found to be equal to the offset of
Strc]in,(percent)
the first fault. The angle that the faults made
Fig. 4. Stress-strain curves for Spruce Pine with the axis of the specimenincreasedwith
dunitc. Uncorrected for change in length or cross- an increasein confiningpressure,and near the
sectional area. Numbers at the end of the curves
brittle-ductile transition pressureit was about
are the confining pressures.
30øø

sectionalarea of the specimen,particularly I)ISCUSSIOI•I

in the brittle region[Brace et al., 1966], but
its precisevalue is dimcultto determine.The Stress-straincurves for Cabramurra serpenti-
changein areapreceding fractureis small,how- hire, in which deformationwas continuedafter
ever, and the error introducedby neglecting it faulting, have been publishedby Raleigh and
does not exceed3%. Paterson [1965]. The fricti.onal shear stress
After faultingthe differentialstressrequired and normal stress were calculated from their
to causefrictional slidingon the fault increases stress-straincurves and an average fault angle
with displacement.The normal and shear of 30ø. The results are plotted in Figure 6.
stresses • and • across the surfaces at the Frictional data for sliding on mated surfaces
minimum, indicatedby open circlesin Figures of granite [Byerice, 1967a, b] are also shown.
I through4, were calculatedby the equations It shouldbe noted that, on surfacesof gran-
ite, the differential. stress required to cause
• = (A•/2) sin 2a sliding decreasesslightly with displacement,
whereas,on surfacesof the rocksin this study,
•* : 2 + it increased with displacement.Also, in the
case of granite, sliding occurs by stick slip;
where .a is the angle that the fault surface that is, movementis accompanied by a sudden
makes with the axis of the specimen, A• is releaseof elastic energy, whereasmovementbe-
the differential stress,and p is the confining tween the surfaces of the rock studied here
pressure.There is a changein the cross-sec- generally took place smoothly. A necessary
tional area of the specimensupportingthe load condition for stick slip to occur is that the
as movement occurs on the fault surface. This frictional stress must decreasewith displace-
area changewas correctedfor useis calculating ment [Rabinowicz, 19'65], and this seemsto be
the stresses.The results ,from all the experi- borne out here. This study showsthat some
ments are plotted in Figure 6. rocks deform by stick slip and others do not,
Near the pressureof the brittle-ductile tran- and this may have important consequences if
sition, multiple faulting occurs and there is sudden movement on a pre-existing fault is
considerableuncertainty in the cross-sectional responsiblefor crustal earthquakes[Brace and
area of the specimenat any point on the stress- Byeflee, 1966]. The phenomenonof stick slip
 BRITTLE-DUCTILE TRANSITION IN ROCKS 4745

has been studiedin detail, and the resultswill are steepthan is requiredif they are inclined
be presentedin a later paper. at a lower angle.For a fixed configuration of
Figure 6 showsthat at low pressuresthe the surfaces in contact, the shear stress re-
frictional shear stressincreasesrapidly as nor- quiredto causeslidingin this mannerincreases
mal stressincreases.Beyond2 kb, the frictional linearly as normal stressincreases.A stage is
stressrises less rapidly and there is a nearly reached,however,at which it is easierto slide
linear relationshipbetween shear and normal by breakingthroughthe asperitiesthan by lift-
stress.Byeflee [1967a, b] studied the friction ing over them.In the caseof granite,this point
of granite in detail and gave the followingex- seems to occur at a normal stress of about 2
planation for the relationshipbetween shear kb, and, beyond this point, the increase in
and normal stress for sliding. At low normal frictional shear stress with normal stress repre-
stresses,surfacesmay slide by lifting over the sents the change in strength of the material
irregularities on the surfaces.The shear stress with pressure.At normalstresses •etweenabout
required to do this is determinedby the angle 100 bars and 2 kb, the behavior is transitional.
made by the irregularities with the plane of That is, the most steeply inclined asperities
sliding. A larger shear stressis required to lift break, but the surfacesmay lift over the less
the surfacesover one another i,f the asperities steeply inclined ones. In this region there is

14 i i I I

12

10

•o•'•
•>/
ß ß -ß
• o/•**

o
øoj,,' ' oSolenhofen
Limestone
•./ •Oak Hall Limestone
/ "Nahant Gabbro
/ • SprucePine Dunile
• CabrarnurraSerpen•inite
(Ralel,
ghandPaterson,,
I 2 $ 4 ,5 6 ?
Confining
Pressure,
(kilobars)
Fig. 5. Differential stressversusconfiningpressureat fracture or 5% strain if the specimen
wasductile.Opensymbolsindicatebrittle behavior;closedsymbols,ductile.Solidline is the
boundarybetweenthe brittle and ductileregionsdeterminedfrom frictiondata (Figure6).
4746 JAMES D. BYERLEE

6 i i i i i i i

o Solenhofen Limestone
• Oak Hall Limestone
[] Nahant 6'(]bbro _

• Spruce Pine Dunlie

• CabramurraSerpentinite
ß Westerlygranite
o
o 3 5 6 7 8
)
NormalS•ress•(kilobars
Fig. 6. Shear stress versus normal stress for friction. Cabramurra serpentinite data from
Raleigh and Paterson [1965] and Westerly granite data from Byerlee [1967a, hi.

a nonlinear increase of the frictional shear stress the surfaceswere completelyinterlocked.From

with normal stress.
This explanation is cons;.stentwith the re-
suits presented here for other rock types.
The most remarkable feature of the friction
of rocks as shown in Figure 6, is that it is
almost independent of rock type. Maurer
[1965] also found this to 'be true at low
normal stresses.Byeflee [1967b] developed a
theory of friction basedon brittle fracture.The
theory predicts that, when contact is made
only at the tips of asperities,the coefi%ientof
friction, which is the ratio of shear to normal
stress, should be independent of the material
and have a value of 0.1. For geometrical situ-
ations that were close to the theoretical model
this prediction was found to be correct. The
theory could not, however, be extendedto a
general situation in which the irregularitieson
the present results, however, where there is
almost certainly a high degree of interlocking,
it seemsthat friction is almost independent of
the material.
It was found in the presentstudythat, near
the pressureof the brittle-ductiletransition,the
angle that the fault surfacesmade with the
axis of the specimenswas closeto 30ø for all the
rocks. The reasonfor this is not clear, but it
seemsto be generally true for rocks [Handin
and Hager, 1957]. Perhaps this phenomenon
is related in someway to the fact that friction
is nearly independentof mineralogyfor inter-
locked surfaces.
Before we discuss the brittle-ductile transi-
tion, we will first transform the line through
the points in Figure 6 into Art, p coordinates
by means of the equations
 BRITTLE-DUCTILE TRANSITION IN ROCKS 4747

• ----(A,/2) sin 2a linesin Figures7 and 8. The transitionpressure

in silicate rocks fell near this line, but, for
the weaker carbonaterocks,the transition took
O'n = 2 (-)cos2 place at much lower pressures.He inferred
If the arg].ea, which the fault surfacemakes from this that the friction hypothesisfor the
with the axis of the specimen,is 30•, the above brittle-ductile transition may be correct for
the silicates but friction could not be the con-
equationsreduce to
trolling mechanism for the carbonates. The
A• = 2.31• basicassumption made by Mogi was that the
coefficientof friction for rocksis independent
P = •n-- 0.58• of confiningpressure.The boundarybetween
The transformation was carried out numeri- the two regionsshouldthereforebe a straight
cally, and the resultis shownas a solidline in line passingthrough the origin.
The present work shows that his basic as-
Figure 5.
If the friction hypothesisfor the brittle- sumptionwas incorrect.The solidline in Fig-
ductile transition in rocks is valid, the rock ures7 and 8 is the frictionboundaryline be-
will be brittle if at any confiningpressurethe tweenthe two regionsfoundin this study,and
differential stress that a rock can support at
failure falls above the line; the rock will be 24.
ductile if the differential stress falls below the
line. In other words, the brittle-ductile transi- 2:2- -

tion pressureshouldbe the pressureat which

the stressrequired to form a fault is equal to 20- o _

the stressrequiredto causeslidingon the fault.

Figure 5 showsthat this statementseemsto be ,,-. 18- / -
true for the rocks studied here. The open ssqn- • o o /
bols to the left of the line represent brittle go 16- o /
-
behavior; the solid symbolsto the right of the -- /
• /
line represent ductile behavior. Slight depar- '-" 14-- / -
o/
tures can be accountedfor by the small de-
pendenceof friction on rock type. For example,
• o o /7
•) 12-
.• o// -
the transition pressure is slightly lower for
Oak Hall limestone and slightly higher for U) o o
Nahant gabbrothan the valuespredicted.From '•lo- oo oø// •/ -
Figure 6, however,it can be seenthat the fric- •- o o
tional shear stress of Oak Hall limestone is _ o o , -
slightly higher and that of Nahant gabbro, • 8 o ./•
which is slightly lower than the line used in
calculatingthe boundary between the brittle
and ductile regions.
o:o;O
Mogi [1966] has collectedfrom the published
literature the strength data for a wide variety
of rock types. These data are shownin Figure
7 for the silicate rocks and in Figure 8 for the o i 2 3 4 5 6

weaker carbonaterocks.The open circlesrepre- Confining Pressure (kilob•rs)

sent brittle behavior and the closed circles duc-
tile behavior. Mogi labeled the ordinate in his
figures axial strength. The original data were
checked,and it was found that axial strength
is, in the notationusedin this paper, differential
stress.Mogi divided the brittle and the ductile
regionsby a straight line, shownas the dashed
Fig. 7. Di•erenti•l stres•versusco•ning pres-
sure for silicate rocks [from •og•, 1966]. Solid
line is the boundary between the brittle and duc-
tile regions determined from friction data. Dashed
line i• the friction boundary suggestedby Mogi.
Open symbo]• indicate brittle behavior; half
closedsymbols,transitional; clo•ed symbols,duc-
tile.
4748 JAMES D. BYERLEE

where a is the angle that the fault surface

14
I • • • / /' makes with the maximum principal stress. In
extension,Heard [1960] found that for 8olen-
0 o /o hofen limestone near the transition pressure

72
[o oøoo o// • ß _ the angle is •bout 20ø. With this value the

• • o //•/•"/' a•boveequations reduce to

A• : 3.11r

p: o',•d- 2.75r
The transformationwas made numerically,and
a oo/
.' '
: . .. ß
' - the result is plotted as a solid line in Figure 9.
• •/I • - ß This line shouldmark the boundary between
brittle and ductile behavior in extension. The
open symbols in the figure represent brittle
oo ', k I k behavior; the solid symbols,ductile behavior.
The circles were plotted from the data pub-
Confin;ng Pressure (kilobors)
lishedby Heard [1960']; the triangles,from the
Fig. 8. Differen•i81 s•ressversus confiningpres-
sure for carbonate rocks [from Mos•, 1966]. Solid data publishedby Handin et al. [19'6.7].The
line is the bounds• between •he bri•fie 8rid brittle-ductile transition takes place at a pres-
ductile regions de•e•ined from friction dstmsure slightly below the predicted value, ,but,
Dsshed Hne is •he friction bounds• suggestedby considering the uncertainties in the strength,
Mogi. Open symbols indicste bri•le behavior;
h81f closedsymbols,•rsnsi•ion81; closedsymbols,
friction, and fault angle,the transition pressure
ductile. in extensionis consistentwith the friction hy-
pothesis.
the transition pressurefor both the silicates It may well be argued that the intermediate
and the carbonates falls close to this line. The stressmay have an effecton friction, but Byef-
frictional hypothesisseems,therefore, to be lee (1967a and more recent unpublished re-
generallytrue for mostrocktypes.Moqi [1900] suits) has shownthat, for granite in compres-
found that the transition pressurefor shale is sion, the frictional, shear stress at any given
very much higher, but this higher pressure normal stressis the same for sliding surfaces
seems to occur because the frictional shear inclined at 30 ø and 45 ø to the axis of the
stressrequired to causesliding on surfacesof specimen. For example, at a normal stress of
shale at any given normal stressis very much 8 kb, the shear stressrequired to causesliding
lower than it is for other rocks [Maurer, 190.5]. on surfaces of granite is 5.3 kb for both 30ø
For the weaker silicatesin Figure 7, the transi- and 45 ø. In both cases the intermediate stress

tion pressureseemsto be slightly higher than
the value predicted. The transition pressureis
also slightly higher for porous tuffs; it would
be of interest to find out whether
occurs becausethis rock type has lower fric-
tion than other rocks.
Heard [1900'] has found that the britfie-
ductile transition pressureis very much higher
in extensionthan it is in compression.Before it
can be determined whether the friction hy-
pothesis is valid for extension, friction must
first be convertedinto this new coordinatesys-
tem. In extension
is the confiningpressureP. If the angle is 45ø,
P is 2.7 kb, but it is 4.94 kb if the angle is 30ø.
This indicates that within the experimental
this relation error, the intermediate stress has no effect on
friction.
One question that still remains unanswered
is what determinesthe stressa rock will sup-
port at any given confining pressure. There
is at presentno satisfactorytheory to predict
the stress required to form a fault surface in
rocks in the brittle region [Brace and Byeflee,
1967]. In the ductile region,it is also not clear
just what determinesthe stressthat a rock will
support.In somerocks,particularlylimestones,
plastic deformation of the individual crystals
may occur during deformation; for fully plastic
2c• materials,
theyieldstress-and-strain
hardening
 BRITTLE-DUCTILE TRANSITION IN ROCKS 4749

o Heard
"Handin el'. [1967)

i I
o 2 4 6 8 10
Confining
Pressure,
[kilobars}
Fig. 9. Differential stressversusconfiningpressureat fracture or 5% strain for 8olenhofen
limestone in extension. Solid line is the boundary between the brittle and ductile regions
determinedfrom friction data. Open symbolsindicate brittle behavior; closedsymbols,ductile.

rate is almost independentof confiningpres-
sure. These effectsare illustratedby the work
of Paterson[1964]. I-Ie showedthat, for copper,
the stress-straincurve at a pressureof 8 kb
does not differ by more than a few per cent
from the curveat atmosphericpressure.This is
clearly not true for limestone.Figures '1 and 2
show that the strain strengtheningrate in-
creasesremarkzbly with confiningpressure.
Cataclasis(that is crushingof the grains)
may be the controllingmechanismof deforma-
tion at room temperature over the confining
pressurerange investigated.At present, how-
ever, there is no theory that can be used to
predict either the stressrequired to deform a
rock in this manner or the effect confiningpres-
sure should have on the strain strengthening
rate. Clearly the problemrequiresfurther work.
The present experimentswere performed at
room temperature and at strain rates of 2.4 X
10-4/sec.Recent unpublishedresultsfor granite
show that, withim the experimentalerror, fric-
tion is not different at strain rates of 2.4 X
10-6/secand 2.4 X 10-4/see.The effectof strain
rate on the fricti.on of other rock types remains
to be determined, however, high temperature
and chemical environment may also have an
effect on .friction, and this possibility should
be investigated.
The present work shows that the 'brittle-
ductile transition pressure in rocks at room
temperature both in compressionand in exten-
sion is the pressure at which the stress re-
quired to form a fault surfaceis equal to the
stressrequired to causesliding on the fault.
Acknowledgments. The researchfacilities at the
MassachusettsInstitute of Technology were made
available through the courtesy of Professor W. F.
4750 JAMES D. BYERLEE
Brace. Professor Brace, Drs. L. Peselnick, C. B.
Raleigh,and E. C. Robertsonread the manuscript
and suggesteda number of improvements.
The investigation, which was carried out at the
Department of Geology and Geophysics,Massa-
chusettsInstitute of Technology,was supportedby
Air Force Cambridge Research Laboratories, Of-
fice of Aerospace Research, U.S. Air Force, Bed-
ford, Massachusetts,under contract AF 19(628)-
3298.Publication authorizedby the Director, U.S.
Geological Survey.
REFERENCES
Brace, W. F., Brittle fracture of rocks,in State o/
Stress in the Earth's Crust, edited by W. R.
Judd, pp. 110-178, American Elsevier, New
York, 1964.
Brace, W. F., and J. D. Byerlee, Stick slip as a
mechanismfor earthquakes,Science,153(3739),
990-992, 1956.
Brace, W. F., and J. D. Byerlee, Recent experi-
mental studies of brittle fracture of rocks, in
Failure and Breakage o/ Rock, edited by C.
Fairhurst, pp. 58-81, American Institute of
Mining, Metallurgical, and Petroleum Engineers,
New York, 1957.
Brace, W. F., B. W. Paulding, Jr., and C. Scholz,
Dilatancy in the fracture of crystalline rocks,
J. Geophys. Res., 71(16), 3939-3953, 1956.
Byerlee, J. D., Frictional characteristicsof granite
under high confining pressure,J. Geophys.Res.,
72(14), 3639-3648, 1967a.
Byerlee, J. D., Theory of friction based on brittle
fracture, J. Appl. Phys., $8(7), 2928-2934,1967b.
Handin, J., and R. V. Hager, Experimental defor-
mation of sedimentary rocks under confining
pressure: Tests at room temperature on dry
samples,Bull. Am. Assoc.Petrol. Geologists,41,
1-50, 1957.
Handin, J., H. C. Heard, and J. M. Magouirk,
Effect of the intermediate principal stresson the
failure of limestone,dolomite, and glassat differ-
ent temperaturesand strain rates, J. Geophys.
Res., 72(2), 611-640, 1957.
Handin, J., and D. W. Stearns,Sliding friction of
rock (abstract) Trans. Am. Geophys. Union,
45 (1), 103, 1964.
Heard, H. C., Transition from brittle fracture to
ductile flow in Solenhofen limestone as a func-
tion of temperature,confiningpressure,and in-
terstitial fluid pressure,
Rock Deformation,Geol.
$oc. Am. Mere., 79, 193-226, 1960.
Maurer, W. C., Shear failure of rock under com-
pression,$oc. Petrol. Engrs. J., 5, 167-175,1965.
Mogi, K., Pressuredependenceof rock strength
and transition from brittle ductile flow, Bull.
Earthquake Res. Inst. Tokyo Univ., •, 215-232,
1966.
Orowan,E., Mechanismof seismicfaulting, Rock
De]ormation, Geol. $oc. Am. Mem., 79, 323-
345, 1950.
Paterson, M. S., Experimental deformation and
faulting in Wombeyanmarble, Bull. Geol. Soc.
Am., 69, 465-476, 1958.
Paterson, M. S., Triaxial testing of materials at
pressures up to 10,000kg/cm2 (150,000lb./sq.
in.), J. Inst.'Engrs.,Australia,23-30,Jam-Feb.,
1964.
Rabinowicz,E., Friction and Wear o] Materials,
John Wiley, New York, 1965.
Raleigh,C. B., and M. S. Paterson,Experimental
deformation of serpentiniteand its tectonic im-
plications,J. Geophys.Res. 70(16), 3965-3925,
1965.
Robertson, E. C., Experimental study of the
strength of rocks, Bull. Geol. Soc. Am., 66, 1275-
1314, 1955.
yon Kirmin, Th., Festigkeits Versuche unter all
suitigcm Druck, Verhandl. deut. Ingr., 55, 1749-
1757, 1911.
(Received December 29, 1967.)