An experimental study of the potential for fault reactivation during changes in gas and 1 pore-water pressure.

: The injection of CO 2 into a depleted reservoir will alter the pore pressure, which if 8 sufficiently perturbed could result in fault reactivation. This paper presents an experimental 9 study of fault reactivation potential in fully saturated kaolinite and Ball Clay fault gouges. 10 Clear differences were observed in fault reactivation pressure when water was injected, with 11 the addition of mica/illite in Ball Clay seen to reduce the pressure necessary for reactivation. 12 Slip occurred once pore-pressure within the gouge was sufficient to overcome the normal 13 stress acting on the fault. During gas injection localised dilatant pathways are formed with 14 approximately only 15 % of the fault observing an elevated gas pressure. This localisation is 15 insufficient to overcome normal stress and so reactivation is not initiated. Therefore faults 16 are more likely to conduct gas than to reactivate. The Mohr approach of assessing fault 17 reactivity potential gave mixed results. Hydro-mechanical coupling, saturation state, 18 mineralogical composition and time-dependent features of the clay require inclusion in this 19 approach otherwise experiments that are predicted to be stable result in fault reactivation. presents results from an experimental study aimed at evaluating fault reactivation potential within the laboratory in two fault gouges. The current study represents the second stage of a three-part investigation of the potential for fault reactivation during the sequestration of carbon dioxide. The three parts of the study were; 1) the role of stress history on fault flow properties, as reported in Cuss et al. (2016); 2) quantification of fault reactivation potential as a result of elevated pore pressure (the current study); and 3) the role of stress history on fault reactivation. The scenario being investigated is for a static boundary condition for stress acting on a fault with an increase in pore pressure initiating fault reactivation; therefore directly simulating an increase in pore pressure in response to the injection of CO 2 during sequestration. of

establish maximum pore pressure perturbations that could be employed during carbon 144 sequestration. 145 Previous experimental work at the British Geological Survey (BGS) on fracture 146 transmissivity in Opalinus clay (Cuss et al., 2011;2014 a,b ) and kaolinite gouge (Sathar et al.,147 2012) showed that hydraulic flow is a complex, focused, transient property that is dependent 148 upon stress history, normal stress, shear displacement, fracture topology, fluid composition, 149 and clay swelling characteristics. The current experimental program aimed to extend this 150 knowledge by investigating the potential for fault reactivation by elevating pore pressure 151 within gouge filled discontinuities.  Figure 1). The thrust blocks of the 189 apparatus were made with a contact area of 60 mm × 60 mm. The lower thrust block was longer than the top one so that the contact area of the experimental discontinuity could be 191 maintained constant throughout the test.

192
As shown in Figure 1 et al. (2015) showed that the gas entry pressure of kaolinite gouge was in excess of 5 230 MPa, therefore a starting pressure of 2 MPa would not result in gas flow within the gouge.

231
The injection syringe pump was switched to constant flow rate operation and delivered 10 ml  Gas entry-pressure was determined using the methodology described in Cuss et al. (2015), by 251 comparing the pressure predicted from Boyle's law with the observed gas pressure. Using the 252 ideal gas law it is possible to determine the mass flux into the clay gouge. A departure is seen 253 between predicted and observed once gas starts to enter the clay; from this the gas entry 254 pressure is then derived.

256
A total of 28 tests were conducted during the current study, as shown in Figure 2 and Table 2; 257 of these, 22 were conducted on kaolinite and 6 were Ball Clay. All 28 tests are reported here 258 for their mechanical shear content, the initial stage of each test was identical for all tests.  with the same protocols irrespective of whether they were fault reactivation tests or not, or 264 whether they were gas or water injection. Tests on kaolinite gouge ranged in vertical stress 265 from 1.1 to 6.4 MPa, while for Ball Clay the range was 2.6 to 6.3 MPa. As shown in Figure   266 2a and b, good repeatability was seen during repeat testing at given vertical stresses for both 267 kaolinite and Ball Clay gouges. Figure 2c shows parameter identified was peak shear stress. As shown in Figure 2, all tests showed classic 276 elasto-plastic behaviour. Therefore the peak stress condition also describes the residual 277 strength of the gouge. Table 2 outlines the vertical and shear stress for the start, yield, and 278 peak shear stress conditions. 279 Figure 3 and Table 3 show the results for starting, yield, and peak shear stresses for all 280 experiments in the current study. As can be seen, the data describe linear relationships with 281 few outliers. Linear regression is shown in Figure 3 with the intercept set to zero; as shown in 282 is not able to sustain as high a shear stress as kaolinite. Therefore the addition of illite, quartz, 289 and possibly water content are resulting in a reduced strength compared with pure kaolinite.
290 Figure 4 shows the data for shear modulus; as shown in Table 2 tests ASR_BigCCS_19BC 291 and ASR_BigCCS_25Kg gave anomalously low and high shear moduli respectively. as the pore pressure between slip events also decreased. Therefore the gouge was undergoing 309 strain softening as a result of reactivation, with further slip events taking less energy to 310 initiate.
All 13 reactivation tests conducted resulted in slip of the critically stressed fault plane as a 312 result of elevated pore pressure, results are shown in Figure 6, vertical stress, with a value of R 2 of 0.91 ( Figure 6a, Table 3). This is reduced to 0.39 when 317 the intercept is set to zero, with this suggesting that reactivation in kaolinite gouge is rate is very small and rises gradually but then the rate of increase of the flow rate abruptly 333 increases. The pressure at which this occurs is identified as the gas entry pressure. Gas peak 334 pressure is simply the maximum gas pressure experienced. Gas breakthrough is the pressure 335 when gas was able to reach the outside of the top block, resulting in a reduction in gas pressure. Table 5 shows the gas entry and maximum gas pressure for all gas injection 337 experiments. Note that test ASR_BigCCS_22Kg was started from 2.5 MPa, which was 338 greater than the gas entry pressure.

339
The results for the fault reactivation tests conducted on kaolinite using gas as the injection 340 fluid markedly contrast with the results seen for water injection (Figure 5d-f, Table 2). Only  Figure 7 shows the results from the fault reactivation experiments using gas as the permeant.

351
No sensitivity to vertical stress was observed in gas entry pressure or the maximum gas 352 pressure achieved (Figure 7a). Only one experiment resulted in fault reactivation. As seen, 353 gas pressure was not able to achieve the level observed during water injection, except for one 354 test conducted at a low vertical stress of 1.13 MPa. However, this test did not show any signs 355 of fault reactivation. Figure 7b shows that no significant differences were apparent in shear 356 stress between tests conducted with gas or water injection. As plotted, the shear stress at gas 357 entry and that during reactivation with water entry perfectly correspond, clearly 358 demonstrating that mechanically there were no differences between the two types of test.

359
The current study successfully reproduced fault reactivation in the laboratory and allowed 361 differences to be noted between water and gas injection, as well as variations related to clay 362 gouge mineralogy.

363
The mechanical aspects of the current study produced well constrained data for two fault 364 gouges. Very good repeatability was seen for repeat tests conducted at near identical 365 boundary conditions. Well constrained linear relationships were noted for starting, yield and clay saturation, although in all tests the gouge was close to 100 % saturation. Figure 4 shows 383 that the results from this study correspond with Byerlee's law (Byerlee, 1978) and therefore 384 that the measured values are consistent with natural rocks.
The fault reactivation study was able to clearly identify reactivation. However, some 386 hydraulic injection tests resulted in single reactivation, whereas others resulted in multiple 387 slip-events (see Figure 5b). The cause for this is uncertain. One hypothesis may be that a 388 larger single slip event releases more energy than a smaller one. However, no variation in 389 shear stress reduction or magnitude in dilation was observed. In general, all slip events using Opalinus Clay was conductive during hydraulic flow and that deformation along a sheared 395 fracture was localised into zones of differing texture. It is possible that the initial pore 396 pressure distribution is similar to that described by Figure 8a, but as slip occurs the gouge is 397 modified resulting in parts becoming conductive, whilst other parts are self-sealed by the 398 shear movement. In tests that showed limited slip events it is possible that the gouge 399 contained conductive channels following shear that resulted in pore pressure dissipation and 400 pressure not increasing as expected. In tests that did show multiple slip-events, these channels 401 did not result in pore pressure dissipation and pressure continued to ramp, becoming 402 sufficient to cause further slip events. Data is not available to fully determine the reasons for 403 these observations.

404
The results for hydraulic injection produced reliable data that showed a marked difference 405 between the two clay gouges. As shown in Figure 6, reactivation tended to occur when the 406 average pore pressure exceeded the yield strength of kaolinite, whereas in Ball Clay 407 reactivation occurred at a stress below the initial starting shear strength. This results in two 408 different reactivation envelopes as shown in Figure 6c. This clearly shows that mica/illite 409 and/or quartz reduces the stress at which a fault will reactivate. However, considering data in the effective mean stress versus differential stress space (Q-P) results in a well constrained 411 single reactivation envelope, as seen in Figure 6d. Effective mean stress (P) is defined simply 412 as the mean stress minus the effect of pore pressure, i.e. P = (( 1 +  2 +  3 )/3) -P f . The 413 differential stress (Q) is simply defined as the difference between the maximum and 414 minimum principal stresses, i.e. Q =  1 - 3. This suggests that in Q-P, mineralogy plays no 415 role in determining reactivation. This envelope suggests that reactivation will occur when 416 differential stress is 2.5 times the effective mean stress: These tests show a stress state that should be stable. Figure 10e shows an example of a test 510 where reactivation occurred at a pore pressure greater than predicted. Generally these results 511 are mixed. Some tests are successfully predicted, some under-estimated and some over-

512
estimated. An under-estimate of pore-pressure variation is acceptable, where an over-estimate 513 means that faults that are predicted to be stable would in fact slip. Figure 10f shows the 514 results for the single gas test that resulted in reactivation. As seen, the Mohr approach shows 515 that reactivation should have occurred at this gas pressure and that the approach would appear 516 valid. However, Figure 10g,h show that at least three tests, with possibly a fourth, were at a 517 stress condition where reactivation should have been observed. Therefore the localised nature 518 of gas pathway formation is not fully accounted for in the approach. Given the mixed results, 519 caution needs to be used when using the Mohr approach to determining fault reactivation 520 potential. Should a maximum pore pressure be restricted to 0.5 -0.75 of the pore pressure 521 predicted by the Mohr approach then this approach may be satisfactory.

522
The Mohr-Coulomb approach to predicting fault reactivation is used by many studies 523 reported, e.g. Cappa & Rutqvist, 2011, 2012Rinaldi & Rutqvist, 2013;Rinaldi et al., 2015. 524 The current study suggests that as a first approximation the approach is valid, although the 525 complete prediction of the pore-pressure is more complex. This may be due to artefacts of the 526 experimental set-up or be associated with complex coupling that occurs as a result of the 527 hydro-mechanical properties of the clay gouge that are not fully described by the simplified 528 approach presented here. It is clear that this is an area that requires further research in order to 529 fully appreciate the physics driving fault reactivation. The observations of the current study 530 also suggest that free-gas will not result in fault reactivation. However, it should be 531 acknowledged that the experimental geometry meant that gas was able to drain from the fault 532 gouge and that in nature sufficient quantities of gas may become present within faults to 533 initiate reactivation.
One limitation of the current study was not being able to inject super-critical CO 2 . Therefore 535 the emphasis of the study was on changes in pore-water pressure as a result of CO 2 injection 536 and should free-gas be present in the reservoir, the consequence of elevated gas pressure on 537 existing faults. The influence of super-critical CO 2 directly in contact with faults was not 538 investigated, nor was the influence of CO 2 should a gaseous phase form. The study was 539 conducted at low pressures compared with in situ stress states and further investigation is likely that mica/illite is responsible for the reduction in cohesion.

554
 Fault reactivation occurred at pressure related to the yield strength in kaolinite and at a 555 pressure less than the starting shear stress in Ball Clay. This shows that Ball Clay has a 556 much lower frictional strength than kaolinite. A single envelope was achieved for fault 557 reactivation potential when data were viewed in the differential (Q) versus effective mean stress (P) space; stating reactivation will occur when Q = 2.5 P. This suggests that 559 the Q-P representation is irrespective of mineralogy, at least for the range of conditions 560 tested in the current work.

561
 During gas injection, only one test showed reactivation and this occurred at a pressure 562 predicted by the Mohr approach. However, 3 further tests predicted to slip showed no 563 evidence of movement.

564
 Gas entry and maximum gas pressure showed no pressure sensitivity to vertical stress.

565
The gas entry pressure is dictated by the frictional properties of the clay gouge, which do 566 not significantly alter over the range of vertical stresses investigated. The maximum 567 pressure achieved is also related to the frictional properties and therefore also showed 568 little to no sensitivity to vertical stress over the limited stresses investigated.    stress (e) and dilation on the fault plane (f). As shown, only one slip event was identified. Gas 781 flow is seen to increase following slip, as seen by a reduction in gas gradient (d).

Gouge Supplier Geological information Location Composition
Kaolinite Imerys