ARTICLE IN PRESS

International Journal of Rock Mechanics & Mining Sciences 42 (2005) 431439

www.elsevier.com/locate/ijrmms

Technical Note

Relation between the Mogi and the Coulomb failure criteria

Adel M. Al-Ajmia,b, Robert W. Zimmermana,,1
a
Department of Earth Science & Engineering, Imperial College, Prince Consort Road, London SW7 2AZ, UK
b
Department of Petroleum & Chemical Engineering, Sultan Qaboos University, Oman
Accepted 5 November 2004
Available online 13 January 2005

1. Introduction commonly used power-law form. It is then shown that

The rock mechanics literature is rich with a number of
shear failure criteria that have been developed. Among
these criteria, the MohrCoulomb criterion is the most
commonly used in practice. There are two major
components of this criterion. The rst is the assumption
that, at failure, the major principal stress s1 is a linearly
increasing function of the minor principal stress, s3 : The
second aspect of this criterion is the assumption that the
value of the intermediate principal stress, s2 ; has no
inuence on the rock strength.
In recent years, much evidence has been accumulating
to suggest that the intermediate principal stress does
indeed have an inuence on rock strength [1,2,4,5,8,9].
Based on some polyaxial compression tests, Mogi [2]
suggested a failure criterion that considered the inu-
ence of s2 on rock failure. Nevertheless, the usefulness
of the Mogi failure criterion remains the subject of
debate [8,9]. In particular, the power-law form of the
Mogi criterion has been criticised because its two
parameters cannot be related to the standard parameters
of the Coulomb failure law, such as the cohesion and the
angle of internal friction.
In this paper, we rst review and dene the Coulomb,
Mohr and Mogi failure criteria. We then examine
published data from eight rocks, and show that a linear
form of the Mogi criterion does essentially as good a job
of representing polyaxial failure data as does the more
Corresponding author. Tel.: +44 2075947412;
fax: +44 2075947344.
E-mail address: r.w.zimmerman@ic.ac.uk (R.W. Zimmerman).
1
Department of Land and Water Resources, Royal Institute of
Technology (KTH), Teknikringen 72, SE-100 44 Stockholm, Sweden.
the two parameters appearing in the linear form of
Mogis criterion can be simply and uniquely related to
the two parameters that appear in the Coulomb
criterion. Although this identication is mathematically
quite simple, it does not seem to have been previously
presented in the literature. In contrast, there is no
unambiguous way to relate the Coulomb parameters to
the two parameters appearing in Mogis power-law
criterion [8]. Finally, the possibility of estimating the
Mogi strength parameters from conventional triaxial
test data is examined.
2. Coulomb, Mohr and Mogi failure criteria
In 1776, Coulomb introduced the simplest and most
important failure criterion. He suggested that rock
failure in compression takes place when the shear stress,
t; that is developed on a specic plane reaches a value
that is sufcient to overcome both the natural cohesion
of the rock plus the frictional force that opposes motion
along the failure plane. The criterion can be written as
t c sn tan f; (1)
where sn is the normal stress acting on the failure plane, c
is the cohesion of the material and f is the angle of
internal friction. As the sign of t only affects the direction
of sliding, Eq. (1) should be written in terms of jtj; but for
simplicity we will omit the absolute value sign.
As criteria (1) will always rst be satised on a plane
that lies in the direction of s2 ; the value of s2 will not
inuence sn or t; and so this failure criterion implicitly
assumes that s2 has no effect on failure. Alternatively,

1365-1609/$ - see front matter r 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijrmms.2004.11.004
 ARTICLE IN PRESS
432 A.M. Al-Ajmi, R.W. Zimmerman / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 431439

this criterion can be interpreted as being intended to internal friction by [3]

apply only to situations in which s2 s3 :
The Coulomb criterion can also be written in terms of q 1 sin f=1  sin f;
the maximum and minimum principal stresses, as C o 2c cos f=1  sin f: 3
follows:
The Coulomb criterion can also be expressed in terms
s1 C o qs3 ; (2) of the maximum shear stress, tmax ; and the effective
mean stress, sm,2 [3, p. 98]:
where the parameters q and Co (the uniaxial compressive
strength) can be related to the cohesion and the angle of tmax c cos f sin fsm;2 ; (4)

400 Dunham Dolomite 300 Solenhofen Limestone

a = 82.554 MPa
a = 86.602 MPa
300 b = 0.4788
b = 0.4123
r2 =0.9814
oct (MPa)

oct (MPa)
200 r2 = 0.9495
200

100
100

0 0
0 200 400 600 800 0 100 200 300 400

(a) m,2 (MPa) (b) m,2 (MPa)

200 150
Mizuho Trachyte Shirahama Sandstone
a = 39.868 MPa a = 14.867 MPa
150 b = 0.4417 b = 0.532
oct (MPa)

oct (MPa)

100
r2 = 0.9595 r2 = 0.9789
100
50
50

0 0
0 50 100 150 200 250 300 350 0 50 100 150 200

(c) m,2 (MPa) (d) m,2 (MPa)

700
KTB Amphibolite 160
600 140 Marble
a = 40.099 MPa
a= 9.1557 MPa
500 b = 0.6364 120
oct (MPa)

oct (MPa)

b =0.6373
400 r2 = 0.9865 100 r2 = 0.9789
300 80
60
200
40
100 20
0 0
0 200 400 600 800 1000 0 50 100 150 200 250
(e) m,2 (MPa) (f) m,2 (MPa)

120 600 Westerly Granite

Yuubari Shale
100 a = 24.069 MPa 500 a = 30.186 MPa
b = 0.4276 b = 0.7116
oct (MPa)

oct (MPa)

80 2 400 r2 = 0.9939
r = 0.9431
60 300
40 200
20 100
0 0
0 50 100 150 200 0 200 400 600 800

(g) m,2 (MPa) (h) m,2 (MPa)

Fig. 1. Linear Mogi failure criterion based on polyaxial test data: (a) Dunham dolomite, (b) Solenhofen limestone, (c) Mizuho trachyte, (d)
Shirahama sandstone, (e) KTB amphibolite, (f) Marble, (g) Yuubari shale and (h) Westerly granite.
 ARTICLE IN PRESS
A.M. Al-Ajmi, R.W. Zimmerman / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 431439 433

where mediate principal stress does indeed have an impact on

rock strength, and that brittle fracture occurs along a
tmax s1  s3 =2;
plane striking in the s2 direction. These results agree
sm;2 s1 s3 =2: 5 with subsequent observations of other researchers [4,5].
From this form of the Coulomb failure criterion, we can Since the fracture plane strikes in the s2 direction, Mogi
conclude that (a) the mean normal stress contributing to concluded that the mean normal stress that opposes the
the creation of a failure plane is sm,2, and (b) there is creation of the fracture plane is sm,2, as dened in Eq.
predicted to be a linear relationship between the (5), rather than the octahedral normal stress, soct ; which
maximum shear stress and the effective mean stress at is given by
failure.
soct s1 s2 s3 =3: (7)
Mohr suggested that at failure, the normal and shear
stresses across the failure plane are related by [3, p. 99] (However, Mogi veried experimentally that rock yield,
t f sn ; (6) which occurs throughout the entire volume prior to total
failure, is a function of soct :) Mogi suggested a new
where f is some function that is assumed to be obtained
failure criterion expressed as
experimentally. A linear form of Mohrs criterion is
equivalent to the Coulomb criterion. Consequently, a toct f sm;2 ; (8)
linear failure criterion such as Eq. (1) is often known as
the MohrCoulomb criterion. From the use of the where toct is the octahedral shear stress, dened by
s1 2s3 Mohrs circle to estimate the failure envelope
given by Eq. (6), Mohrs assumption implies that the q
1
fracture plane strikes in the s2 direction. This assump- toct s1  s2 2 s2  s3 2 s3  s1 2 (9)
3
tion has been veried experimentally by Mogi [2], and
can be interpreted as justifying the extension of a two- and f is some monotonically increasing function. As the
dimensional failure criterion into three dimensions. distortional strain energy is proportional to the octahe-
Mogi [2] conducted the rst extensive polyaxial dral shear stress [3, p. 125], this criterion is equivalent to
compressive tests in rocks. He noted that the inter- asserting that failure will occur when the distortional
strain energy reaches some critical value that increases
monotonically with sm,2.
Table 1 We consider assumption (8) to be the crux of Mogis
Mogi (linear model) strength parameters from polyaxial test data
contribution, and the specic algebraic form of the
Rock type a (MPa) b r2 function f to be a somewhat secondary matter of
convenience. Although this function has usually been
Dunham dolomite 82.55 0.48 0.981
taken to be a power law, below we will investigate the
Solenhofen limestone 86.60 0.41 0.950
Mizuho trachyte 39.87 0.44 0.960 implications of taking it to be a linear function. It is
Shirahama sandstone 14.87 0.53 0.979 important to note, however, that assumption (8) is non-
Yuubari shale 24.07 0.43 0.943 trivial, since it is not a priori obvious that failure, which
KTB amphibolite 40.10 0.64 0.987 depends on three variables s1 ; s2 ; s3 can be represented
Marble 9.16 0.64 0.979
in a two-dimensional mathematical sub-space, such as
Westerly granite 30.19 0.71 0.994
toct ; sm;2 or toct ; soct :

Table 2
Power-law and parabolic Mogi criteria; all stresses are in MPa

Rock type Power law r2 Parabolic r2

Dunham dolomite toct 4:50s0:69

m;2
0.989 toct 0:0003s2m;2 0:69sm;2 43:9 0.986
Solenhofen limestone toct 8:08s0:57
m;2
0.945 toct 0:0003s2m;2 0:24sm;2 111:13 0.951
Mizuho trachyte toct 3:39s0:69
m;2
0.984 toct 0:0007s2m;2 0:70sm;2 17:21 0.976
Shirahama sandstone toct 1:84s0:79
m;2
0.983 toct 0:0012s2m;2 0:78sm;2 3:56 0.984
KTB amphibolite toct 1:76s0:86
m;2
0.994 toct 0:0001s2m;2 0:75sm;2 20:04 0.988
Marble toct 1:49s0:85
m;2
0.983 toct 0:0005s2m;2 0:74sm;2 4:81 0.980
Yuubari shale toct 2:77s0:69
m;2
0.944 toct 0:0002s2m;2 0:49sm;2 20:22 0.943
Westerly granite toct 1:54s0:89
m;2
0.997 toct 0:0002s2m;2 0:86sm;2 8:16 0.996
 ARTICLE IN PRESS
434 A.M. Al-Ajmi, R.W. Zimmerman / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 431439

3. Analysis of polyaxial failure data in toct  sm;2 space
Under polyaxial (also called true triaxial) compres-
sion, rock failure will be a function of three variables,
the three principal stresses. In order to try to verify that
failure can be described by a function in the two-
dimensional space (toct ; sm,2), we have searched the
literature for polyaxial failure data, and analysed these
data in the (toct ;sm,2) plane.
Eight such data sets have been located and analysed.
These rock types were Dunham dolomite, Solenhofen
limestone, Mizuho trachyte, coarse grained dense

400 Dunham Dolomite 300 Solenhofen Limestone

a = 58.32 MPa a = 103.95 MPa
b = 0.5454 b = 0.3497
300
r2 =0.9904 200 r2 = 0.9828
oct (MPa)

oct (MPa)
200
100
100

0 0
0 100 200 300 400 500 0 100 200 300 400
(a) m,2 (MPa) (b) m,2 (MPa)

200 150
Mizuho Trachyte Shirahama Sandstone
a = 30.374 MPa a = 32.946 MPa
150 b = 0.4904 b = 0.3933
r2 =0.9751 100 r2 = 0.9326
oct (MPa)
oct (MPa)

100
50
50

0 0
0 50 100 150 200 250 300 0 50 100 150 200
(c) m,2 (MPa) (d) m,2 (MPa)

600 80
KTB Amphibolite Marble
500 a = 26.303MPa a = 9.802 MPa
b = 0.6939 b = 0.5782
60
400 r2 = 0.9977 r2 = 0.9986
oct (MPa)

oct (MPa)

300 40
200
20
100

0 0
0 100 200 300 400 500 600 700 0 50 100 150

(e) m,2 (MPa) (f) m,2 (MPa)

500
Westerly Granite
a = 23.878 MPa
400 b = 0.7401
r2 = 0.9991
oct (MPa)

300

200

100

0
0 200 400 600
(g) m,2 (MPa)

Fig. 2. Linear Mogi failure criterion based on triaxial test data: (a) Dunham dolomite, (b) Solenhofen limestone, (c) Mizuho trachyte, (d) Shirahama
sandstone, (g) KTB amphibolite, (h) Marble and (g) Westerly granite.
 ARTICLE IN PRESS
A.M. Al-Ajmi, R.W. Zimmerman / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 431439 435

marble, Shirahama sandstone, Yuubari shale, KTB all be t quite well with a linear function (Fig. 1):
amphibolite and Westerly granite. The polyaxial data
for Dunham dolomite, Solenhofen limestone and toct a bsm;2 ; (10)
Mizuho trachyte are taken from Mogi [2], the polyaxial
where a is the intersection of the line with the toct -axis,
data for marble are from Michelis [6,7], the polyaxial
and b is its inclination. In Fig. 1, the polyaxial test data
data for Shirahama sandstone and Yuubari shale are
from Takahashi and Koide [4], and the polyaxial data Table 5
for Westerly granite is from Haimson and Chang [5]. Polyaxial test data for Dunham dolomite [2]
The polyaxial data for KTB amphibolite is taken from
the paper by Colmenares and Zoback [8], from tests s1 s2 s3 sm;2 Experimental Theoretical
(MPa) (MPa) (MPa) (MPa) toct (MPa) toct (MPa)
carried out by Chang and Haimson [9]. These sets of
data are shown in Tables 512. 257 0 0 129 121 128
Some of these data, such as those for the Westerly 400 25 25 213 177 174
granite, were available in numerical form in the 474 68 25 249 202 194
500 91 25 263 210 201
published papers. In most cases we digitised the data
553 135 25 289 227 216
from gures in the original papers. As an estimate of the 574 177 25 299 231 222
errors that may have been incurred in the digitisation 594 232 25 310 235 227
process, we can compare our values with those of 544 269 25 285 212 213
Colmenares and Zoback [8], who also digitised many of 488 45 45 267 209 204
562 100 45 303 232 224
the same data sets, and then displayed their results in
586 124 45 316 239 231
tabular form. In most cases our values agree with those 607 159 45 326 242 236
of [8] to three gures, and the largest discrepancies are 639 183 45 342 254 245
about 2%. Assuming as a worst case that the values 671 241 45 358 261 253
used in [8] are exact, our values have an average error 670 263 45 358 259 253
622 293 45 334 236 240
of less than 1%. We also point out that all available data
568 65 65 316 237 231
points from the original sources have been included in 636 113 65 351 259 250
our analysis. 642 152 65 353 254 251
We have plotted all the data points recorded in Tables 687 208 65 376 266 263
512 in the (toct ; sm,2) plane, and found that they could 684 259 65 374 258 263
725 306 65 395 273 274
700 390 65 383 259 267
624 85 85 354 254 252
682 126 85 384 272 268
Table 3
718 150 85 402 284 277
Mogi (linear model) strength parameters from triaxial test data
743 230 85 414 282 284
Rock type a (MPa) b r2 771 300 85 428 286 292
818 371 85 451 301 304
Dunham dolomite 58.32 0.55 0.990 798 440 85 442 291 299
Solenhofen limestone 103.95 0.35 0.983 679 105 105 392 271 272
Mizuho trachyte 30.37 0.49 0.975 776 165 105 441 303 299
Shirahama sandstone 32.95 0.39 0.933 784 202 105 445 300 301
KTB amphibolite 26.30 0.69 0.998 804 265 105 455 299 306
Marble 9.80 0.58 0.999 822 331 105 464 299 311
Westerly granite 23.88 0.74 0.999 839 351 105 472 305 316
820 411 105 463 293 311
863 266 105 484 326 322
724 125 125 424 282 290
823 186 125 474 315 317
Table 4 859 241 125 492 322 327
Coulomb parameters (Co, q), and the Mogi parameters (a, b) 862 288 125 493 316 327
calculated from Eq. (14) 893 359 125 509 322 336
942 411 125 533 338 349
Rock type Co (MPa) q a (MPa) b 918 458 125 522 325 343
887 510 125 506 311 334
Dunham dolomite 298.93 3.66 60.49 0.54 892 254 145 519 329 341
Solenhofen limestone 351.50 2.16 104.87 0.35 929 292 145 537 340 351
Mizuho trachyte 139.90 3.05 32.59 0.48 924 319 145 535 334 350
Shirahama sandstone 123.59 2.31 35.16 0.37 922 342 145 534 330 349
KTB amphibolite 220.35 6.44 27.92 0.69 1016 387 145 580 367 375
Marble 54.02 4.15 9.88 0.58 1003 404 145 574 359 371
Westerly granite 240.09 8.20 24.61 0.74 953 451 145 549 333 358
 ARTICLE IN PRESS
436 A.M. Al-Ajmi, R.W. Zimmerman / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 431439

Table 6 Table 7
Polyaxial test data for Solenhofen limestone [2] Polyaxial test data for Mizuho trachyte [2]

s1 s2 s3 sm;2 Experimental Theoretical s1 s2 s3 sm;2 Experimental Theoretical

(MPa) (MPa) (MPa) (MPa) toct (MPa) toct (MPa) (MPa) (MPa) (MPa) (MPa) toct (MPa) toct (MPa)

395 20 20 208 177 177 100 0 0 50 47 55

415 52 20 217 179 180 193 15 15 104 84 81
413 91 20 217 171 180 253 30 30 142 105 100
455 165 20 237 181 187 300 45 45 173 120 115
459 203 20 240 180 188 314 55 45 180 125 118
464 231 20 242 181 189 326 71 45 186 127 121
442 40 40 241 190 188 333 96 45 189 126 123
455 40 40 248 196 191 349 142 45 197 127 127
486 80 40 263 201 196 361 214 45 203 129 130
496 113 40 268 200 198 365 289 45 205 137 131
526 190 40 283 203 203 351 332 45 198 140 128
542 267 40 291 205 206 339 60 60 200 132 128
534 312 40 287 202 204 352 91 60 206 131 131
472 60 60 266 194 197 383 142 60 221 137 139
516 87 60 288 209 205 396 191 60 228 138 142
535 100 60 298 215 208 404 229 60 232 141 144
529 111 60 295 210 207 400 271 60 230 140 143
573 162 60 316 222 215 383 331 60 222 142 139
551 196 60 305 207 211 365 75 75 220 137 138
556 271 60 308 203 212 400 114 75 238 145 147
529 80 80 305 212 210 417 153 75 246 146 151
569 125 80 324 221 217 438 229 75 257 149 156
580 150 80 330 221 219 439 300 75 257 150 156
641 205 80 361 241 230 424 343 75 250 149 153
592 221 80 336 216 221 451 391 75 263 165 159
674 280 80 377 247 236 419 100 100 260 150 158
659 294 80 369 239 233 460 137 100 280 162 168
648 373 80 364 232 231 489 186 100 294 167 175
678 448 80 379 246 237 494 274 100 297 161 176
522 382 100 311 175 183
513 411 100 307 176 181

are plotted as empty circles, and the solid line is their

best-tting linear model. The data for which s2 s3 ;
that is, the traditional triaxial test data, are shown as
large black circles. In each case the triaxial data lie
fairly close to the line that has been t to the
polyaxial data. The linear form provides a good
t to the data sets, with very high correlation
coefcients, r2. Based on the polyaxial test data, the
Mogi strength parameters, a and b, for the eight tested
rocks and the corresponding correlation coefcient is
given in Table 1. According to this goodness of t, a
linear model can be considered to be a reasonable
assumption.
Power-law and parabolic Mogi failure criteria have
been also been t to the data, with the results shown in
Table 2. From the correlation coefcient values, the
power-law and parabolic failure envelopes have almost
the same goodness of t, with a slight preference usually
observed for the power-law model. However, compar-
ison of the linear and power-law failure envelopes
reveals that, in general, the linear form does as good a
job as the power-law form in tting the data (see
Tables 1 and 2); the variance between the best-tting
equation and the data is due more to experimental
scatter than to any inherent difculty in tting the data
to a straight line.
If the Mogi assumption is valid, the strength
parameters a and b can be approximated from a set of
triaxial tests, as this is just a special case of a polyaxial
stress state. A good test of this assumption would be to
t the conventional triaxial data to a linear Mogi failure
criterion, and then see whether or not the polyaxial data
fall on this line. The triaxial test data are plotted in
Fig. 2, together with the best-tting linear models;
Yuubari shale is excluded, due to the lack of triaxial test
data for this rock. Tables 3 and 4 records the strength
parameters a and b that were obtained from the triaxial
test data, along with their associated r2 values. These
models are then used to calculate the predicted
value of the octahedral shear stress at failure, which
are then compared to the values determined
from the polyaxial failure data (i.e., the experimental
toct values). The predicted/theoretical and measured/
experimental toct values are shown in Tables 512. The
agreement is quite good, with correlation coefcients
in the range 0.950.99, and the mean relative error
 ARTICLE IN PRESS
A.M. Al-Ajmi, R.W. Zimmerman / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 431439 437

Table 8 Table 9
Polyaxial test data for Shirahama sandstone [4] Polyaxial test data for KTB amphibolite [8,9]

s1 s2 s3 sm;2 Experimental Theoretical s1 s2 s3 sm;2 Experimental Theoretical

(MPa) (MPa) (MPa) (MPa) toct (Mpa) toct (MPa) (MPa) (MPa) (MPa) (MPa) toct (MPa) toct (MPa)

94 9 5 50 41 52 165 0 0 83 78 84
97 15 5 51 41 53 346 79 0 173 148 146
88 29 5 47 35 51 291 149 0 146 119 127
109 44 5 57 43 55 347 197 0 174 142 147
94 65 5 50 37 52 267 229 0 134 118 119
109 12 8 59 47 56 410 30 30 220 179 179
129 27 8 69 53 60 479 60 30 255 205 203
132 41 8 70 53 61 599 100 30 315 253 245
135 50 8 72 53 61 652 200 30 341 262 263
127 79 8 67 49 59 571 249 30 301 222 235
147 15 15 81 62 65 637 298 30 334 248 258
157 29 15 86 64 67 702 60 60 381 303 291
165 62 15 90 63 68 750 88 60 405 319 307
162 82 15 89 60 68 766 103 60 413 323 313
159 88 15 87 59 67 745 155 60 403 303 306
168 97 15 92 63 69 816 199 60 438 329 330
178 20 20 99 74 72 888 249 60 474 354 355
183 30 20 102 75 73 828 299 60 444 321 334
173 41 20 97 68 71 887 347 60 474 343 355
185 50 20 103 72 73 954 399 60 507 369 378
177 57 20 98 67 72 815 449 60 438 308 330
197 68 20 109 75 76 868 100 100 484 362 362
194 82 20 107 72 75 959 164 100 530 391 394
193 97 20 106 71 75 1001 199 100 551 403 408
185 100 20 103 67 73 945 248 100 523 368 389
197 30 30 114 79 78 892 269 100 496 341 370
218 47 30 124 85 82 1048 300 100 574 408 425
224 69 30 127 84 83 1058 349 100 579 406 428
232 88 30 131 85 85 1155 442 100 628 439 462
229 109 30 130 82 84 1118 597 100 609 416 449
241 129 30 136 86 86 1147 150 150 649 470 476
227 150 30 128 81 83 1065 198 150 608 420 448
215 171 30 122 79 81 1112 199 150 631 442 464
224 40 40 132 87 85 1176 249 150 663 462 486
244 60 40 142 92 89 1431 298 150 791 572 575
252 70 40 146 93 90 1326 348 150 738 514 538
253 79 40 146 92 91 1169 399 150 660 434 484
252 100 40 146 89 90 1284 448 150 717 480 524
274 99 40 157 99 95 1265 498 150 708 466 517
265 118 40 153 93 93 1262 642 150 706 455 516
279 138 40 160 98 96
274 159 40 157 95 95
231 50 50 141 85 88

4. MogiCoulomb criterion

(in absolute value) of about 4%. These results verify For conventional triaxial tests, s2 s3 ; and so from
that a Mogi failure criterion based on triaxial Eq. (9) the octahedral shear stress takes the form
test data correlates well with the polyaxial strength q
1
data. Therefore, failure under polyaxial stresses toct s1  s2 2 s2  s3 2 s3  s1 2
can in fact be predicted from triaxial test data using 3
p
the linear Mogi criterion. (Of course, it is known 2
s1  s3 : 11
that failure envelopes typically develop some concave- 3
downwards curvature at high conning stresses, so it is The linear Mogi criterion, Eq. (10), then reduces to
never advisable to extrapolate tted failure envelopes p
beyond the range of s3 values for which data is 2 b
s1  s3 a s1 s3 : (12)
available.) 3 2
 ARTICLE IN PRESS
438 A.M. Al-Ajmi, R.W. Zimmerman / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 431439

Table 10 Table 11
Polyaxial test data for coarse grained dense marble [6,7] Polyaxial test data for Westerly granite [5]

s1 s2 s3 sm;2 Experimental Theoretical s1 s2 s3 sm;2 Experimental Theoretical

(MPa) (MPa) (MPa) (MPa) toct (MPa) toct (MPa) (MPa) (MPa) (MPa) (MPa) toct (MPa) toct (MPa)

61 4 0 30 28 27 201 0 0 101 95 98
77 10 0 38 34 32 306 40 0 153 136 137
105 20 0 53 45 40 301 60 0 151 130 135
103 35 0 52 43 40 317 80 0 159 135 141
66 4 4 35 30 30 304 100 0 152 127 136
75 5 4 39 33 33 231 2 2 117 108 110
84 7 4 44 37 35 300 18 2 151 137 136
94 10 4 49 41 38 328 40 2 165 146 146
130 28 4 67 55 48 359 60 2 181 156 157
193 69 4 98 78 67 353 80 2 178 150 155
143 83 4 73 57 52 355 100 2 179 149 156
84 7 7 45 36 36 430 20 20 225 193 190
113 14 7 60 48 44 529 40 20 275 235 227
134 28 7 71 56 51 602 60 20 311 265 254
192 55 7 99 78 67 554 62 20 287 242 236
188 83 7 98 74 66 553 61 20 287 242 236
175 110 7 91 69 62 532 79 20 276 229 228
117 14 14 65 48 48 575 100 20 298 245 244
126 21 14 70 51 50 567 114 20 294 239 241
147 28 14 80 60 56 601 150 20 311 249 254
155 41 14 84 61 59 638 202 20 329 259 267
195 57 14 105 77 70 605 38 38 322 267 262
255 86 14 135 101 88 620 38 38 329 274 267
277 113 14 145 108 94 700 57 38 369 308 297
138 21 21 80 55 56 733 78 38 386 319 309
150 28 21 86 59 59 720 103 38 379 307 304
209 62 21 115 81 76 723 119 38 381 306 305
260 83 21 141 101 91 731 157 38 385 303 308
289 110 21 155 111 99 781 198 38 410 319 327
171 28 28 99 67 67 747 60 60 404 324 323
167 48 28 97 61 66 811 90 60 436 347 346
222 55 28 125 86 82 821 114 60 441 347 350
275 83 28 151 106 97 860 180 60 460 352 364
314 110 28 171 120 109 861 249 60 461 342 365
349 110 55 202 128 127 889 77 77 483 383 381
954 102 77 516 408 405
992 142 77 535 417 419
998 214 77 538 406 422
1005 310 77 541 394 424
1012 100 100 556 430 435
1103 165 100 602 458 469
Comparison with the form of the Coulomb law 1147 167 100 624 479 485
1155 216 100 628 472 488
given by Eqs. (4) and (5) shows that for triaxial data,
1195 259 100 648 483 503
the linear Mogi criterion coincides with the 1129 312 100 615 444 479
Coulomb criterion, if we make the following
identication:
p p
2 2 2 2
a c cos f; b sin f: (13)
3 3 As another test of the equivalence between the linear
Hence, for triaxial stress states (s2 s3 ) the linear Mogi Mogi criterion and the Coulomb criterion, we can make
criterion given by Eq. (10) is exactly equivalent to the the following comparison (Table 4). First, the para-
Coulomb criterion. Alternatively, using Eq. (3), the meters Co and q are determined by tting the triaxial test
parameters of the linear Mogi criterion can also be data to a Coulomb failure line. These values are then
identied with the Coulomb failure parameters (q, Co) used to calculate the MogiCoulomb strength para-
as follows: meters, a and b, using Eq. (14). If we compare these
p p values of a and b with the values found directly from
2 2 Co 2 2 q1 applying a linear Mogi model in toctsm,2 space (see
a ; b : (14)
3 q1 3 q1 Table 3), we nd extremely close agreement.
 ARTICLE IN PRESS
A.M. Al-Ajmi, R.W. Zimmerman / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 431439 439

Table 12 Furthermore, we found that the numerical values of

Polyaxial test data for Yuubari shale [4]
the parameters that appear in the MogiCoulomb
s1 (MPa) s2 (MPa) s3 (MPa) criterion can be estimated from conventional triaxial
test data. Thus, this polyaxial failure criterion can be
161 25 25 applied even in the absence of polyaxial (true triaxial)
168 25 25
data. This offers a great advantage, as most laboratories
182 35 25
187 36 25 are equipped to conduct only traditional s2 s3 tests.
175 45 25 Finally, we showed that if the linear form of the Mogi
175 56 25 criterion is used, the parameters that appear in it can be
186 66 25 unambiguously related to the traditional parameters
200 77 25
194 79 25
appearing in the Coulomb failure law. The lack of such
196 85 25 a relationship for the parameters appearing in the
201 96 25 power-law Mogi criterion has been cited in [8] as a
194 100 25 major drawback to the use of that model.
186 114 25
197 124 25
183 133 25
228 50 50
239 50 50 Acknowledgments
245 50 50
257 69 50
261 90 50
Adel Al-Ajmi thanks Sultan Qaboos University of
266 100 50 Oman for nancial support of his Ph.D. studies.
260 110 50
260 122 50
285 129 50
266 148 50 References
256 159 50
[1] Haimson BC. The hydrofracturing stress measuring method and
recent eld results. Int J Rock Mech Min Sci 1978;15(4):16778.
[2] Mogi K. Fracture and ow of rocks under high triaxial
compression. J Geophys Res 1971;76(5):125569.
5. Summary and conclusions [3] Jaeger JC, Cook NGW. Fundamentals of rock mechanics, 3rd ed.
London: Chapman & Hall; 1979.
We have shown that linear Mogi criterion does a good [4] Takahashi M, Koide H. Effect of the intermediate principal stress
job in representing rock failure under polyaxial stress on strength and deformation behavior of sedimentary rocks at the
depth shallower than 2000 m. In: Maury V, Fourmaintraux D,
states. When s2 s3 ; the linear version of Mogis editors. Rock at great depth. Rotterdam: Balkema; 1989. p. 1926.
triaxial failure criterion reduces exactly to the Coulomb [5] Haimson BC, Chang C. A new true triaxial cell for testing
criterion. Hence, the linear Mogi criterion can be mechanical properties of rock, and its use to determine rock
thought of as a natural extension of the Coulomb strength and deformability of Westerly granite. Int J Rock Mech
criterion into three dimensions (i.e., polyaxial stress Min Sci 2000;37:28596.
[6] Michelis P. Polyaxial yielding of granular rock. J Eng Mech
space). As Mohrs extension of the Coulomb criterion 1985;111(8):104966.
into three dimensions is often referred to as the [7] Michelis P. True triaxial cyclic behavior of concrete and rock in
MohrCoulomb criterion, we propose that the linear compression. Int J Plast 1987;3(2):24970.
version of the Mogi criterion be known as the [8] Colmenares LB, Zoback MD. A statistical evaluation of intact
rock failure criteria constrained by polyaxial test data for ve
MogiCoulomb failure criterion. The classical Cou-
different rocks. Int J Rock Mech Min Sci 2002;39(6):695729.
lomb failure criterion can therefore be thought of as a [9] Chang C, Haimson B. True triaxial strength and deformability of
special case, which applies only when s2 s3 ; of the the German Continental deep drilling program (KTB) deep hole
more general linear Mogi failure criterion. amphibolite. J Geophys Res 2000;105:189999013.