Aseismic Transient Slip on the Gofar Transform Fault, East Pacific Rise

Home , Seismometer

Downloaded by guest on October 1, 2021 3,adfutzn eoiysrcueo oa 5 onl suggest jointly (5) Gofar on structure velocity zone distribution fault seismicity and fine-scale (3), the of observations OBS-based data. of teleseismic uncertainties using transform location exist large Charlie-Gibbs although offset (9), the Ridge that Mid-Atlantic as transforms the such on ridges, reported also slow-spreading every are quasi-periodically ruptures rupture (M Repeated that magnitude a patches moment trans- where distinctive 6.5 Ridge Ocean, two Blanco Pacific separates northeast the zone the along earthquake of Similar observed form 8). are (2, 1992 patterns since rupture Gofar after- on locations and events centroid foreshock relative M6.0+ the of 2008 and 1) the (Fig. along- (3) from distributions 20-km shock inferred to is a 15- pattern on by rupture y separated M 6 segments (EPR), to Gofar strike 5 Rise the every Pacific On mechanism quasi-periodically East (3–7). repeat source interaction the modes the possible slip of seismic and transform on and patterns light aseismic rupture between shed earthquake Recent that OTF near- 2). of data experiments provide (1, seismic ridges deployment fast-spreading [M7]) field with (OBS) associated 7 OTFs seismometer magnitude along mod- bottom to only (up ocean have potential earthquakes defined magnitudes observed thermally erate largest large the area the rupture despite and seismically, A transients aseismic faults transform oceanic zone fault relation and the swarms, candidates verify earthquake properties. hydromechanical to slip, provides fault experiments and aseismic geodesy between faults seafloor transform future for oceanic on M swarms transform. in aseis- Gofar transients episodic of the prevalence mic possible on the suggests swarms zones thus model seismic Our barrier of the migration along observed migrates slip speeds slow at episod- modeled arise transients The aseismic ically. where barriers rupture in transition, and results propagation brittle–ductile rupture seismic the along-strike stabilizes seismic to effectively down by the possible circulation and In supported porosity hydrothermal fluid-filled faults. enhanced strengthening, indicates transform that dilatancy imaging sequences oceanic on earthquake strong slip of model, we transient study Here aseismic unclear. modeling remain and numerical interaction a their seismic and present between slip partitioning aseismic the under- and govern the However, that the Ocean. mechanisms and Pacific Rise lying northeast Pacific magni- the East in the repeating Ridge of transform Blanco of Gofar the occurrence on as swarms, local averaged globally the (M) large tude and a deficit, including behavior, seismic seismic slip of combination aseismic unique review and for a (received display 2020 faults 19, transform March Oceanic approved and NY, Palisades, University, Columbia of 2019) Observatory 6, Earth August Lamont-Doherty Ekstrom, Goran by Edited 02467 MA Hill, Chestnut College, Boston Sciences, Environmental 02543; and MA Hole, Woods Institution, Oceanographic a aigLiu Yajing fault, Rise transform Pacific Gofar East the on slip transient Aseismic www.pnas.org/cgi/doi/10.1073/pnas.1913625117 eateto at n lntr cecs cilUiest,Montr University, McGill Sciences, Planetary and Earth of Department h essetrpuebrirtruhmlil yls recent cycles, multiple through barrier rupture persistent The 5)o hi cuuae moment accumulated their of release (∼15%) (OTFs) percentage faults small transform a oceanic globally, veraged ∼ 0t 0 /,saitmoal orltdwt the with correlated spatiotemporally m/h, 600 to ∼10 a,b,1 atqae ihaudn oehcsadseismic and foreshocks abundant with earthquakes 6 efe .McGuire J. Jeffrey , | esi swarms seismic ∼ | atqaerpuesegmentation rupture earthquake utr are oe hths active host that zones barrier rupture 6 0k are oe() The (8). zone barrier ∼10-km b,c n akD Behn D. Mark and , c atqaeSineCne,U elgclSre,MfetFed A905 and 94035; CA Field, Moffett Survey, Geological US Center, Science Earthquake ∼ 0k barrier ∼10-km earthquakes 6 4y(,7). (2, y ∼14 | w . to 6.0 ) a,Q,Cnd 3 0E8; H3A Canada QC, eal, ´ b,d e ioeesdee hntelwrlmto cai crustal oceanic of limit likely lower the mantle, the by uppermost defined than the seismicity deeper to kilometers depth few in a extending 4.5) experienced to zone (M0 barrier central from the distinct shock, are zone M the barrier in the those of properties material the that oie 5.Hwvr uhlresaeatrainwudproduce would alternation large-scale such or depths However, peridotite crustal (5). altered at tolite) be 90% alternation to or can mineral 50 8% of Gofar (e.g., to fraction 1.5 on volume of anomaly large porosity a fluid-filled velocity enhanced the large either on by observed This achieved as (5). transition brittle–ductile fault extend the not Gofar to may way it the and of of clear extent all less depth is The zones (12). low-velocity mechanical fracturing continental a intense to to conti- due attributed along zone usually damage is common and also faults is transform are to nental reduction which (up velocity Significant (11), seismic reduction. zone velocity ∼50%) also barrier seismic the are the with values in consistent drop earthquakes mainshock stress coupled for static fully estimated lower the in Systematically con- zone (4). in fault (5), area intact crust nearly oceanic a entire to the trast dam- through several-kilometer-wide revealed extending a zone zone in age velocities barrier P-wave the in reduction across line refraction doi:10.1073/pnas.1913625117/-/DCSupplemental at online information supporting contains article This 1 aapout a eacse nOe cec rmwr at Framework Science Seismology Open ZD, on for accessed code Institutions be can Research network products Data Incorporated center, the seismome- data bottom at ocean (IRIS) archived Gofar 2008 are the experiment from ter data waveform Seismic deposition: Data the under Published Submission.y Direct PNAS a is article This interest.y competing paper. no y the declare wrote authors M.D.B. The and J.J.M., Y.L., and Y.L. and research; data; performed analyzed J.J.M. J.J.M. and Y.L. research; designed Y.L. contributions: Author owo orsodnemyb drse.Eal [email protected] Email: addressed. be may correspondence whom To wrs rdcinta a etse ihftr seafloor microearthquake future with migrating tested be drive experiments. can prevalence geodesy that that a prediction events shows a model swarms, slip Our slow Rise. Pacific of trans- Gofar East the fault, on form observed earth- rupture partition the slip physics-based reproduce and simultaneously a segmentation can to model cycle dilatancy quake find frictional We coupling propagation. rupture that large prohibits earthquake but small permits nucleation, These that explanation ruptures. an earthquake require large barriers separate Recent in that unresolved. microseismicity zones barrier remain abundant document fault experiments same on seafloor the modes slip of different parts to conditions different lead geological that the mechanisms physical But and behavior. aseismic of and combination unique seismic a through slip faults transform Oceanic Significance b ∼ NSlicense.y PNAS eateto elg n epyis od Hole Woods Geophysics, and Geology of Department utr oe.Pirt h 2008 the to Prior zones. rupture 6 600 o https://www.fdsn.org/networks/detail/ZD stem(0.A active-source An (10). isotherm C . y 0 epniie troc- serpentinized >70% https://www.pnas.org/lookup/suppl/ NSLts Articles Latest PNAS 200foreshocks ∼22,000 d eateto Earth of Department https://osf.io/rw9bq/ M 0t 20% to ∼10 w . main- 6.0 | f7 of 1 2007/ .y .

EARTH, ATMOSPHERIC, AND PLANETARY SCIENCES A normal stress (normal stress minus pore pressure) is enhanced. −4.4˚ This dilatancy-strengthening mechanism has been proposed to be effective in inhibiting dynamic deformation in landslides, glacier tills, and fault gouges (17–19) and more recently in numerical models generating episodic slow slip events in subduc- M6 rupture tion zones (20, 21). Based on the aforementioned observational −4.6˚ “Barrier” evidence, we investigate the feasibility of dilatancy strengthening M6 rupture as a mechanism for the persistent rupture barrier behavior on the Gofar transform and its implications for the oceanic transform 010KM 20 30 fault slip budget. −106.4˚ −106.2˚ −106˚ −105.8˚ −105.6˚ OTF Earthquake Sequence Model Coupled with Dilatancy dilatancydilatancy dilatancy B (ab) We develop a three-dimensional (3D) numerical model based on 0 the Gofar transform, including OTF-speciﬁc loading and bound- 0.04 5 0.02 ary conditions, to simulate earthquake rupture sequences on 0 0 the western Gofar fault segment governed by laboratory-derived

Depth (km) 10 −40 −30 −20 −10 0 10 20 30 40 rate- and state-dependent friction laws (22). In the rate–state Along−strike (km) framework, the friction stability parameter a − b dictates to the first order whether slip has the potential to grow into a seismic C a−b a dc (mm) (MPa) 0 0 0 0 event (a − b < 0, velocity weakening [VW]) or will remain stable (a − b > 0, velocity strengthening [VS]) (23, 24) (Materials 5 5 5 5 and Methods). In our model, a − b is mapped on the OTF by Depth (km) converting temperature-dependent gabbro gouge rate–state fric- 10 10 10 10 0 0.02 0.04 0 0.02 0.04 0 20 40 0 20 40 60 tion data (25) using a western Gofar thermal structure calculated with hydrothermal cooling and visco-plastic rheology (10). Seis- Fig. 1. Gofar transform fault seismicity, M6 rupture segments, and earth- mic slip can be achieved when the down-dip dimension of the quake cycle model setup. (A) Seismicity was recorded and located in August VW segment exceeds a characteristic nucleation size defined by to December 2008 (black dots), surrounding an Mw 6 mainshock on Septem- ber 18, 2008. Yellow, foreshocks on September 10 to 12; red, aftershocks on the friction, normal stress, and pore pressure conditions on the September 18 to 20; cyan, swarm seismicity on December 7 to 8. White tri- fault (Materials and Methods). In the model presented here, the angles, seismometer; white stars, seismometer colocated with strong-motion above instability condition is always satisfied such that the entire sensor. White ellipses show the estimated rupture extents of the 2008 Mw 6 ∼4-km (1 to 5 km in depth) VW zone ruptures in a single earth- (west) and 2007 Mw 6.2 (east), with gray diamonds denoting respective cen- quake without an effective stabilizing mechanism (SI Appendix, troid locations. Image credit: modified from McGuire et al. (3). (B) Rate–state Fig. S1). friction stability parameter (a − b) mapped on the transform fault repre- The stable–unstable competition dictated by the VS–VW prop- senting western Gofar. Area within two white lines (friction stability transi- erties can be used to design spatial distributions of a − b, along tion, a − b = 0) has velocity-weakening friction (a − b < 0); outside is velocity strengthening (a − b > 0). (C) Depth distributions of friction parameters with other model parameters, such that earthquake ruptures are a − b, a, characteristic slip distance dc, and effective normal stress σ¯. mainly confined within the VW zones. For example, in numerical models consisting of homogeneous, alternating VS/VW segments, a larger VS zone, higher effective normal stress, and larger a − b values are all favorable conditions for inhibiting the prop- a gravity anomaly much greater than that observed across the agation of seismic ruptures that originated from the neighboring Gofar barrier zone, whereas the estimated porosity change would VW segments (26). For Gofar, we find that when a 10-km along- produce a gravity signature below the measurement sensitivity strike VS zone is embedded between two 20-km VW zones, M6 (5). Therefore, enhanced fluid-filled porosity is likely the primary ruptures in the VW zones do not propagate into the central VS mechanism for the seismic velocity reduction in the barrier zone zone, resulting in a rupture barrier (SI Appendix, Fig. S2; all other on Gofar. parameters the same as in the SI Appendix, Fig. S1 case). How- Precise seismicity relocation on western Gofar identified a ever, such a crustal-scale VS zone is unlikely because fault zone ◦ ∼600-m fault trace offset at the boundary between the 2008 Mw temperature there is largely lower than ∼500 C required for 6.0 rupture and the barrier zone (Fig. 1), which spatially coin- a − b > 0 based on laboratory gabbro friction data (25) and per- cides with a ∼500-m-deep seafloor depression (4). Although a vasive serpentinization appears unlikely as discussed above. The fault trace offset <1 km is not likely to be sufficient to stop a occurrences of ∼20,000 foreshocks in the barrier zone to the 2008 Mw 6.0 dynamic rupture based on observations from continen- Mw 6.0 also suggest that the entire oceanic crust cannot simply slip tal transform faults (13, 14), such local complexity in the fault aseismically under VS conditions. geometry provides favorable conditions for normal faulting and More complex spatial heterogeneities, such as random or fluid infiltration, in a manner similar to that of a pull-apart basin stochastic distributions of a − b, can also be introduced to gen- (15). Hydrothermal circulation is also likely responsible for low- erate small earthquakes on VW patches embedded in a largely ering the fault zone temperature and hence increasing the depth VS, creeping background. Such a VW–VS mélange can work of the brittle–ductile transition into the upper mantle, where effectively to stop M6 ruptures from adjacent segments, as best pervasive serpentinization is suggested to favor aseismic slip illustrated by the arrest of Parkfield earthquake ruptures at and promote abundant earthquake swarms, as recently observed the creeping section along the central San Andreas Fault (27). along the Blanco Ridge transform (7, 16). The persistent supply Within the San Andreas Fault creeping section, VW friction of excess fluids can result in large fluid-filled porosity and high appears to be limited to small areas along the streaks of micro- pore pressure, both of which can influence fault slip behavior. seismicity (28), therefore leaving the majority of the fault area at In fluid-saturated granular materials, porosity evolves as shear VS. By contrast, the western Gofar barrier zone is spectacularly begins (17, 18). If the shear sliding rate, and consequently the seismically active, generating thousands of microearthquakes per rate of creating new void space by rotating and fracturing grains, week in a short 10-km fault segment (3). The abundant and is more rapid than ambient fluids can flow into the new pore widespread distribution of microseismicity in the western Gofar space, local pore pressure is temporally reduced and effective barrier zone and its adjacency to the M6 mainshock zone strongly

2 of 7 | www.pnas.org/cgi/doi/10.1073/pnas.1913625117 Liu et al. Downloaded by guest on October 1, 2021 Downloaded by guest on October 1, 2021 i tal. et Liu in arrows Black zone. high-dilatancy porosity central and the coefficient within dilatancy and between ratio the by represented events w simcsi rnins Q n Q eoetoearthquakes. two denote EQ2 and EQ1 transients; slip aseismic two V pres- Pore (5). zone barrier rupture Gofar the for porosity inferred bulk param- as second the The 31). is 30, (19, eter gouges simulated and natural arrest; veloc- on and sliding stabilization in slip between for increase is tendency given greater a hence and for ity porosity in increase larger oi ie r nesimcsi (V slip interseismic are lines solid 2. Fig. higher A steps. velocity shear with the changes is dila- parameter key first coefficient two The dilatancy setting. are nondimensional the model There our from (5). in zone inferred parameters barrier tancy values the in porosity anomaly plausible P-wave the stop the effectively given can fault dilatancy ruptures the whether end; determine reflect to each spatially us allows on to chosen km is along- 15 an M dilatancy Such and of 1B). center distribution (Fig. modified strike the not of at are segments elsewhere km properties along-strike 10 three fault, on the dilatancy introduced studies, previous is in strengthening predictions numerical and vations here observation barrier rupture Gofar. on large the with the VW–VS inconsistent the through from mixture, propagate still nucleated would zone demon- rupture seismogenic parameter VW numerically earthquake stability As frictional an a (29). such distribution, under Fault (29), Jacinto Anza Anza for San the strated on the depths of locking proposed geodetic section layer and seismogenic seismicity the reconcile of the to bottom in the distribution at VW–VS zone stochastic transition bar- the to the analogous VS proportionality material, be This (3). would sporadic VW ruptures large by stops of repeatedly zone only predominance rier this separated Despite are m patches. and VW–VS predominate a patches such if suggest, pl hst osrc oe ossetwt hs edobser- field these with consistent model a construct to Thus = ∼ omlzdmxmmsi aelog rate slip maximum Normalized (B) cm/y. 14 ,which 1A), (Fig. Gofar western on patterns rupture 6.0 χ uuaiesi nmlil atqaecce tdph4k.Black km. 4 depth at cycles earthquake multiple in slip Cumulative (A) budget. slip seismic vs. aseismic and pattern rupture earthquake Modeled s o5k)udrdfeetdltnyconditions, dilatancy different under km) 5 to (−5 zone dilatancy high central the within and cycles earthquake modeled the over averaged , 10 −5 and 10 −4

0.1 and 0.01 between varies which φ, Seismic CD Slip (m) AB esrdi al ldn experiments sliding fault in measured

s 1.5 2.5 max 0.2 0.4 0.6 0.8 2 3 0 1 10 4 2 040 20 0 -20 -40 < ´ lnei rsn,ta h VW the that present, is elange -4 10 6 V pl hc esrsporosity measures which , lte vr . .Rddse ie r oesi lp(V slip coseismic are lines dashed Red y. 0.5 every plotted ) SSE2 2a, 2b 10 -3 10 EQ2 (V Along-strike (km) / orsod oa to corresponds max 10 /V C -2 and pl ntefutfrtesm eida in as period same the for fault the on ) SSE1 D =0.1 =0.05 =0.02 =0.01 on otesmlto aesonin shown case simulation the to point rcino simctasetslip transient aseismic of Fraction (D) /φ. φ = EQ1 10 0.05, -1 Transient

0.05 0.15 t 0.25 oawoefutrpuea rdce ihn rawa dila- weak a or no with predicted as ( effect rupture tancy whole-fault of a values to other with cases simulation tional (see Gofar western for eta are oea ucino h ereo dilatancy, of degree the of parameter function nondimensional a the by as represented zone barrier central earthquake past the westward in the to Gofar similar (8). on cycles other, observed each rupture of of months M sequence few the a patches that rupture within two such pen the phasing of remain similar cycles also seismic have fault the M the where modeled pattern of of eral ends pair eastern The and aseismic. western the at zones M a to As barrier (−5 side. zone either dilatancy central on the boundaries result, the diminishing approaching quickly with speeds zone rupture no-dilatancy 2 the mainly (Fig. within is contained y propagation into and 6 zones, separated no-dilatancy to to is strong-dilatancy 5 dilatancy every strong repeating A segments, of eastern zones and with western model the in porosity with cient case simulation a 2 Fig. Barrier 20) Rupture (19, fault entire the slip on of Methods). pressure and evolution pore (Materials fric- and spatiotemporal rate–state stress, the shear the obtain rate, to and to coupled fault is framework the fault, tion along the slip across frictional diffusion ﬂuid by controlled evolution, sure = 0.1 0.2 0.3 0 10 efrhrqatf h esi opigrto( ratio coupling seismic the quantify further We 10 and −4 -4 A 10 = o h aein case the for .Csimcrpueiiitsa h onaisof boundaries the at initiates rupture Coseismic B). and −4 10 ∼ B 10 = hc r nterneo aaeesestimated parameters of range the in are which , -3 10 vnso ahsd.Teohrtodilatancy two other The side. each on events 6 hw a shows EQ1 EQ2 Log / max −5 S1adSE eoetetmn n oain of locations and timing the denote SSE2 and SSE1 A. A 10 ; and ≥ SSE1 SSE2 2a, 2b ,erhuk rupture earthquake S1), Fig. Appendix, SI A (V 10 5 10 o addi- for S3, and S1 Figs. Appendix, SI and max 0yhsoyo lpadsi aefor rate slip and slip of history ∼10-y -2 (C B. 6 χ V t pl B. /V vrgdoe utpeerhuk cycles earthquake multiple over averaged lte vr eod lt odn rate loading Plate second. every plotted ) rcino lprlae uigseismic during released slip of Fraction ) pl 0 ) φ 10 10 12 14 16 18 20 0 = ∼ -1 o5k)bcmsarupture a becomes km) 5 to utrsflosagen- a follows ruptures 6 .05 Time (yr) ∼ NSLts Articles Latest PNAS utrsotnhap- often ruptures 6 n iaac coefﬁ- dilatancy and φ and /φ χ .I contrast In ). see 2C; (Fig. s ihnthe within ) | f7 of 3

EARTH, ATMOSPHERIC, AND PLANETARY SCIENCES Materials and Methods for the formulation of /φ). Here χs is 0 day Log (V/V ) A 10 pl deﬁned as the ratio of slip released seismically to the total slip 0 4 accumulated at the full loading rate (Vpl = 14 cm/y) in mul- 2 5 0 tiple modeled earthquake cycles (Materials and Methods). For 10 -2 example, for the case in Fig. 2A (/φ = 0.002), about 30% of B 0.43 day the total moment is released seismically in the barrier zone, as 0 4 a result of coseismic slip propagating into it prior to a com- 2 5 plete stop. Overall, the coupling ratio decreases with increasing 0 /φ or, equivalently, greater dilatancy effect. Our results thus 10 -2 conﬁrm the hypothesis that dilatancy strengthening can serve C 0.84 day 0 4 as an effective mechanism for arresting seismic slip and result 5 2 in rupture barriers to M ∼ 6 earthquakes as observed along 0 western Gofar. 10 -2 Depth (km) D 10.5 day Aseismic Slip Transients Driving Seismic Swarms 0 4 2 The central barrier zone does not slip simply at, or below, the 5 0 plate loading rate Vpl . Rather, this region experiences transient 10 -2 aseismic slip events, as illustrated in Fig. 2B, with excursions E 0.42 yr Cumulative slip (m) in slip rate one to three orders of magnitude above Vpl dur- 0 0.2 ing the intervening periods between the M ∼ 6 earthquakes. 5 Here we quantify the aseismic transient slip ratio χt as fault 10 0 slip that would exceed the geodetic detection threshold (several -40 -20 0 2040 times Vpl ) but is not associated with earthquakes (i.e., exclud- F Along-strike (km) ing earthquake nucleation and coseismic and postseismic phases) 1000 (Materials and Methods). In contrast to the seismic coupling ratio slow slip migration 800

χs , which monotonically decreases with an increasing dilatancy (m/hr) effect (Fig. 2C), χt is greatest for the choices of porosity and 600 dilatancy coefﬁcient pairs that result in an intermediate value /φ = 0.002 (Fig. 2D); a weaker dilatancy effect is not sufﬁcient 400 Day 10.5 Day 0 to stabilize seismic slip to aseismic transients whereas a stronger 200 effect leads to slip rates near or lower than Vpl (SI Appendix, Fig.

Migration speed 0 S3). Within the intermediate range of /φ, the average χt in the -5 0 5 10 15 central barrier zone can reach 15 to 25%, a significant fraction of Along-strike (km) the total OTF slip budget. The transient slip events nucleate at the peripherals of the Fig. 3. Modeled aseismic transient slip propagation (SSE1 in Fig. 2 A and locked M ∼ 6 rupture zones and propagate across the adjacent B). (A–D) Snapshots of normalized slip rate on the fault at times 0, 0.43, rupture barrier zones at a speed of 10 to 600 m/h within ∼5 to 0.84, and 10.5 d from a reference starting time t = 10.5 y as shown in Fig. 10 d. Propagation speeds average ∼100 m/h for the events in the 2B. Red/orange colors represent migrating transient slip at rates between ∼ 4 central barrier zone (Fig. 3) and ∼20 m/h for the events in the Vpl and 10 Vpl. Dark blue areas represent locked segments that previously ruptured in M ∼ 6 earthquakes. (E) A total of maximum 30 cm slip accu- west-end barrier zone (SI Appendix, Fig. S4). Both cases are sim- −4 mulated in 0.42 y (t = 10.26 to 10.68 y), defined as the duration of this ulated with = 10 and φ = 0.05. Fig. 4 shows the December 0.8 episode when Vmax > 10 Vpl.(F) Westward and eastward aseismic tran- 2008 swarm events, the most abundant swarm episode during sient slip front migration speeds during the active 10.5-d period shown in the 2008 Gofar OBS experiment, on the westernmost segment A–D. The central high-dilatancy zone is defined within along-strike −5 to of Gofar (cyan dots in Fig. 1A, longitude −106.2 to −106.3◦S). 5 km. Dilatancy coefficient = 10−4, porosity φ = 0.05. The seismicity migration envelope can be accurately described by the propagation of modeled aseismic slip transients with a range of dilatancy coefficients = (0.7 to 1.2) × 10−4 and a porosity of (34, 35), the maximum cumulative slip in each aseismic slip φ = 0.05. Note that this range of /φ = (1.4 to 2.4) × 10−3 corre- episode can reach ∼30 cm (Fig. 3E), with surface displace- sponds to the peak aseismic transient slip budget (χt ) as shown ments up to ∼10 cm (SI Appendix, Fig. S6), rendering a unique in Fig. 2D. test of the dilatancy hypothesis by future seafloor geodesy The aseismic slip propagation fronts in general follow a dif- experiments. fusional curve, with faster migration in the first 1 to 2 d followed by slower speeds for the remaining period. The average Discussion and Conclusions migration speed does not vary much for the range of dilatancy Hypotheses proposed to explain the partitioning between aseis- parameters that produce significant transient slip; clear migra- mic and seismic slip on oceanic transform faults essentially tion patterns are difficult to quantify for cases outside the peak involve either a single-mode or a multimode rupture scenario. χt range due to lower slip rates and smaller slip areas. We In the single-mode scenario, earthquakes persistently occur propose that aseismic transient slip may be the driving mech- on fully coupled segments of distinct frictional properties due anism for seismicity swarms on Gofar, such as inferred for to spatially variable lithology and degree of hydration (3, 8). continental strike-slip faults (32) and subduction zones (33), and In the multimode rupture scenario, fault segments transition possibly a manifestation of porosity wave migration within the between aseismic and seismic slip modes as either frictional rupture barrier. Although no transient slip events have been properties evolve over time due to changes in fault hydrome- detected geodetically along OTFs due to lack of instrumen- chanical conditions (36) or the dimensional ratio between tation, their appearance in our numerical model and similar the seismogenic zone and the characteristic nucleation zone migration pattern to the Gofar swarms suggest they may be (defined in the rate–state framework) leads to a mixture of a common mode of slip along oceanic transform faults and aseismic and seismic slip (22). The quasi-periodic occurrences a prevalent slip mechanism within OTF rupture barriers. Dis- of Mw 5 to 6 earthquakes with similar centroid locations tinct from the slow slip events observed in subduction zones on the Gofar-Discovery and Blanco Ridge transform faults

4 of 7 | www.pnas.org/cgi/doi/10.1073/pnas.1913625117 Liu et al. Downloaded by guest on October 1, 2021 Downloaded by guest on October 1, 2021 i tal. et Liu reference. for m/h, 100 0.8 cases cases), at simulation from Black speed fronts slip porosity (3). migration using aseismic template diffusive linear modeled their show a lines from Solid shows to independently (M0 line relocated end. events not west dashed smaller the were of toward that detections interaction 2) matched-ﬁlter–based transform–ridge show Gofar dots the Gray near d M ∼10 relocated show dots Red fronts. tion 4. Fig. psd ntews n fGfr(i.4,simct propa- seismicity 4), is (Fig. speed Gofar swarm migration of 2008 swarm end December of west the median the range For a Gofar. on wide with western episode A m/h for 2,000 (7). found consistent seis- to also m/h found 100 bottom 370 Ridge from ocean of Blanco rates using migration the study swarm migration near recent Swarm underlying deployed A the 40). mometers behavior. into (7, creep slip insights layers aseismic useful fault mantle by provide upper driven thus or asperities characteristics crust manifestation small a the on be in to failure suggested brittle is of and (37–39) faults boundary consideration. into taken the be of must state stress zone the seismogenic that future suggests which rupture, earthquake an 2 2B (Fig. Fig. transients from aseismic M to the leads barrier and adjacent in the zones on neighboring loading slip increases consequently the slip postseismic zones ture in the and of nucleation Coseismic peripherals earthquake segments. bar- the promotes the around and in stresses event zone slip increases transient zone aseismic rier an from release energy M barrier the rupture and high-dilatancy, the zones in creep aseismic between ing rupture earthquake the extent OTFs’ what global patterns. oceanic to affects determine on mechanism to studies dilatancy needed high-precision Future are and location Gofar. structure, seismicity velocity along geometry, those zones fault than transform effect possi- barrier dilatancy indicates the weaker 5.5 and in magnitude strength local fault a higher more bly of play occurrence may the the where changes in trans- bathymetry roles as Ridge and Blanco important zones geometry the barrier For rupture fault transform. rupture form, seismic Gofar clearly persistent the stop for Gofar in observed can after result The dilatancy study and pattern. our propagation strong single-mode in that the modeled demonstrates with evolution consistent slip more OTF are 8) (7, case). one (green,

esi wr ciiyhsbe ogrcgie nplate on recognized long been has activity swarm Seismic trigger- stress two-way a suggest also results modeling Our Longitude -106.25 -106.15 -106.3 -106.2 -106.1 esi wr nDcme 08admdldaesi lpmigra- slip aseismic modeled and 2008 December in swarm Seismic × 10 φ −4 htntalaesi rnin lpeet precede events slip transient aseismic all not that = 3 4 4 4 4 348 346 344 342 340 338 bu,tocss,10 cases), two (blue, 5addltnycoefficient dilatancy and 0.05 ∼ nteculdsimgnczns Strain zones. seismogenic coupled the in 6 M w Day ofYear2008 . o65cnrlbrirzn (7), zone barrier central 6.5 to 6.0 −4 o3erhuksdtce within detected earthquakes 3 to ∼1 100 m/hr bak w ae) n 1.2 and cases), two (black, = 0.7 A × and 10 −4 .W note We B). ca,three (cyan, ∼ rup- 6 × 10 −4 al fadmningetrta h hrceitcnceto iedefined size nucleation characteristic the than as greater dimension a of fault yai”apoiain(45) approximation dynamic” (22). earthquakes a (W using zone zone nucleation 1) seismogenic the (Fig. of depth-dependent width effect. circulation be hydrothermal for to accounts σ that it (10) model convert thermal Gofar and (25) conditions when rate slip Reference (δ following cycles, slip earthquake OTF fault in dilatancy evolution simulate with coupled to (23), law (19), friction state-dependent and rate- derived Dilatancy. with Coupled Friction Rate–Station Methods and evaded Materials long has under- that better process to nucleation seismologists. us and earthquake allow swarms, the potential may stand seismic The earthquakes creep, large 43). pre- aseismic subsequently (42, cases between repeating several triggering earthquakes of in stress megathrust form reported large the been in ceding has sometimes slip as aseismic swarms, how microearthquakes, seismic understand to to boundary required plate leads earth- major is on work of budget Future type slip faults. the total for and implications sequences quake fundamental migration has space, effect swarm and dilatancy of no patterns to diffusional little are that observed. nor primar- such be exhibited serpentine, may By is layer by depths. mantle filled Blanco crustal the ily at in space migration pore swarm contrast, slip diffusional transient driving aseismic Gofar promoting the and in thereby effect zones, dilatancy barrier hence rupture and com- porosity Hydrothermal (10) enhance alternation to involvement. mineral bine may of extent fluid differences characteristics minor the of migration and above circulation degrees swarm The Blanco different (41). and suggest diffusion Gofar migra- the fluid to exhibiting in of comparable California southern swarm indicative more in Gofar indeed tion bursts The are seismic lines). of pattern those solid and 4, speed of (Fig. approx- migration majority curve an followed the diffusive sequence whereas swarm imately 2008 (7), December a Gofar December pattern demonstrate the the migration swarms in linear Blanco those than the roughly num- fewer Furthermore, event far episode. the are 2008 although episode (4), each S7) of bers Fig. the also Appendix, on are (SI those episodes epicenters than speeds swarm slower migration other magnitude with Several identified of fault. order transform an Blanco least at wk, 1 rcinsaiiyparameter temperature-dependent of stability measurement laboratory friction adopt We unity. of tor parameters friction state where at average on gates ¯ = h* h lsi eainbtenslip between relation elastic The time in evolving slip, aseismic abundant of presence The 0Maadtecaatrsi lpdistance slip characteristic the and MPa 50 = V τ θ γ (x = stesaevral.Efcienra stress normal Effective variable. state the is (b−a) , V ξ µbd 0 , h atqaencetso eoiywaeig( velocity-weakening a on nucleates earthquake The . t τ )= 2 c σ ¯ ¯ = − − dp d dt dt 4) where (44), h* σ θ f η Z V ∼ = = = 0 ∂δ 0 X = L 4k,udrwihnal l lpi eesdin released is slip all nearly which under km, 1.14 1 (σ − (x Z ∂ − 10 p , − 0 ∼20 t a, ξ W − −6 V t d , p) d p t a b, θ c p µ ) k [f − , /,adnmnlfito coefficient friction nominal and m/s, 0 , (x ssermdls and modulus, shear is ,sersrs (τ stress shear ), d 0 o10mhoe eido about of period a over m/h 100 to + b c + − r iecntn u ptal variable. spatially but constant time are β o abogueudrhydrothermal under gouge gabbro for a x ∼ δ 0 ln(V θ 1 , n ha stress shear and m saot35tmsta fthe of that times 3.5 about is km) 4 ξ ,0 / rmterelocated the from m/h <1,000 d dt , θ ξ /V 0 , ) 0 eipeettelaboratory- the implement We δ ) + (x NSLts Articles Latest PNAS 0 d b , ,adpr rsue(p) pressure pore and ), c ξ ln(V 0 is , t γ msc htthe that such mm ∼5 ) 0 τ σ ¯ − d θ/ sagoerclfac- geometrical a is ≡ olw h “quasi- the follows V σ pl c )] t − , dx n rate– and p, 0 a d ξ − | f 0 0 b = f7 of 5 < 0.6 [2] [1] [4] [3] 0)

EARTH, ATMOSPHERIC, AND PLANETARY SCIENCES 0.5 where XL = 80 km and Wd = 10 km are the along-strike and along-dip quantified as transient slip (χt ). Lower cutoff velocities, 10 Vpl and 0.3 dimensions of the modeled OTF, respectively. The stiffness matrix k(x − 10 Vpl, are also tested, and the overall slip budget dependence on dila- x0, ξ, ξ0) represents shear stress change at location (x, ξ) due to a unit tancy parameters as in Fig. 2 C and D and SI Appendix, Fig. S3 is not affected. 0 0 slip at location (x , ξ ), where a Fourier transform is applied along the χs, χa, and χt are all calculated as the ratio of slip released in respective strike (x) to account for translational symmetry and reduce computation phases to the total slip accumulated in the seismogenic zone (velocity weak- time. ening) over multiple OTF earthquake cycles, typically ∼10 cycles when the 2 In Eq. 3, tp ≡ ηβd /k is the characteristic diffusion timescale for fault coupling coefficients reach long-term steady-state values (22). SI Appendix, zone pore pressure p to reequilibrate with the ambient pressure p0 across a Fig. S3 shows the along-strike distribution of χs, χa, and χt slip ratios as the fault zone of finite thickness d (19). It is challenging to estimate tp because effect of dilatancy varies from weak to strong, = (0.2, 0.3, 0.5, 0.6, 0.7, 1, 2, fault zone thickness d and permeability k are largely unknown for Gofar. 5)×10−4 (light gray to dark lines), with a constant porosity value φ = 0.05. A plausible range of k ≈ 10−14 to 10−20 m2 for d ∼ 1 mm (fault core) to We can see a clear locking effect under strong dilatancy = 2 × 10−4 and 9 −3 −4 ∼1 m (damage zone) (46) results in a wide range of tp ≈ 10 φ to 10 φ 5 × 10 , the latter shown in red, in the central ∼10 km and ∼15 km at [s], where φ is fault zone porosity and β = φ(βf + βφ) is fault gouge bulk each of the two ridge–transform intersections. Aseismic slip is in general −4 −1 compressibility, defined with fluid compressibility βf = 5 × 10 MPa , complementary to the seismic slip budget, but the two modes of slip do not −2 −1 −3 solid compressibility βφ = 10 MPa , and fluid dynamic viscosity η = 10 necessarily add up to 1 in the three segments with strong dilatancy, because 3 0.8 Pa·s. At intermediate drainage conditions, for example, tp ≈ 10 φ, for φ = a significant portion is released when Vmax < 10 Vpl, which accounts for 0.01 to 0.1, as inferred for the rupture barrier zone of Gofar (5), tp ≈ 10 extremely slow creep during the interseismic loading period. to 100 s, which is comparable to the coseismic rupture timescale, mean- ing that any pore pressure change would nearly instantaneously return to Aseismic Transient Migration. Colored solid lines in Fig. 4 show the migra- its ambient level (effectively drained), equivalent to no dilatancy effect. tion fronts of aseismic transient events modeled with φ = 0.05 and = Furthermore, previous numerical studies have found that at intermediate 0.7 to 1.2 × 10−4 (a total of eight simulation cases). The origin of these ◦ drainage conditions, tp/(dc/Vpl) ≈ 0.01 to 0.1, the total coseismic slip and modeled diffusional fronts is set at day 339.8 and longitude 106.262 S, rupture depth range of modeled earthquakes are similar to their refer- assuming one longitudinal degree is 105 km, for a direct comparison to the ence values without dilatancy (22). Therefore, in this study we focus on Gofar swarm migration at the same spatial and temporal scales. For each 9 the nearly undrained condition of tp ≈ 10 φ, which maximizes the dilatancy snapshot, the east and west fronts of the bilateral slip migration are respec- effect; this is equivalent to assuming k = 10−20 m2 within d = 1 mm. With tively defined as the eastern- and western-most positions on the contour the above assumptions, the controlling parameters to pore pressure evo- line where slip velocity is a fraction (e.g., 0.8) of the maximum velocity Vmax lution (Eq. 3) are and φ, which enter the second term in Eq. 3 as their on the fault of that snapshot. Several fraction thresholds in the range of ratio /φ. 0.6 to 0.85 are tested and the migration front positions are not strongly Small perturbations are introduced into the initial conditions to allow affected (SI Appendix, Fig. S5). the system to evolve off steady state and into nonuniform slip mode. Specif- ically, at the beginning of each simulation, t = 0 y, we set state variable Data Availability. Seismic waveform data from the 2008 Gofar ocean bottom θ = dc/Vpl (steady state) everywhere on the fault, except for the segment seismometer experiment are archived at the Incorporated Research Insti- between −25 to −5 km along strike; θ = dc/(2.5Vpl) on that segment. V, tutions for Seismology (IRIS) data center, network code ZD, https://www. θ, and p evolutions become independent of the initial conditions after the fdsn.org/networks/detail/ZD 2007/. Data products can be accessed on Open first cycle of earthquakes. Science Framework at https://osf.io/rw9bq/.

ACKNOWLEDGMENTS. We thank Joan Gomberg, Ruth Harris, Steve OTF Slip Budget. We use the maximum fault slip rate V as a criterion max Hickman, Shane Detweiler, Mike Diggles, and two anonymous external to quantify seismic versus aseismic slip. Slip is deﬁned as seismic (χs) when reviewers for their thoughtful comments that helped to improve the 6 0.8 6 Vmax ≥ 5 mm/s ≈ 10 Vpl and aseismic (χa) when 10 Vpl < Vmax < 10 Vpl, manuscript. This study was supported by Natural Sciences and Engineer- and a subset of aseismic slip during periods not associated with earthquakes, ing Research Council of Canada Discovery Grants RGPIN/418338-2012 and i.e., excluding slip in earthquake nucleation and postseismic phases, is RGPIN-2018-05389; and NSF Grants OCE-10-61203 and OCE-18-33279.

1. M. S. Boettcher, T. H. Jordan, Earthquake scaling relations for mid-ocean ridge 15. P. M. Gregg, J. Lin, D. K. Smith, Segmentation of transform systems on the East Pacific transform faults. J. Geophys. Res. 109, B12302 (2004). Rise: Implications for earthquake processes at fast-slipping oceanic transform faults. 2. M. S. Boettcher, J. J. McGuire, Scaling relations for seismic cycles on mid-ocean ridge Geology 34, 289–292 (2006). transform faults. Geophys. Res. Lett. 36, L21301 (2009). 16. J. J. McGuire, The geology of earthquake swarms. Nat. Geosci. 12, 82–87 (2019). 3. J. McGuire et al., Variations in earthquake rupture properties along the Gofar 17. C. Marone, C. B. Raleigh, C. H. Scholz, Frictional behavior and constitutive modeling transform fault, East Pacific Rise. Nat. Geosci. 5, 336–341 (2012). of simulated fault gouge. J. Geophys. Res. 95, 7007–7025 (1990). 4. B. Froment et al., Imaging along-strike variations in mechanical properties of 18. P. L. Moore, N. R. Iverson, Slow episodic shear of granular materials regulated by the Gofar transform fault, East Pacific Rise. J. Geophys. Res. 119, 7175–7194 dilatant strengthening. Geology 30, 843–846 (2002). (2014). 19. P. Segall, J. R. Rice, Dilatancy, compaction, and slip instability of a fluid-infiltrated 5. E. Roland, D. Lizarralde, J. McGuire, J. Collins, Seismic velocity constraints on fault. J. Geophys. Res. 100, 22155–22171 (1995). the material properties that control earthquake behavior at the Quebrada- 20. Y. Liu, A. M. Rubin, Role of fault gouge dilatancy on aseismic deformation transients. Discovery-Gofar transform faults, East Pacific Rise. J. Geophys. Res. 117, B11102 J. Geophys. Res. 115, B10414 (2010). (2012). 21. P. Segall, A. M. Rubin, M. Bradley, J. R. Rice, Dilatant strengthening as a mechanism 6. M. Wofson-Schwehr, M. Boettcher, J. McGuire, J. Collins, The relationship between for slow slip events. J. Geophys. Res. 115, B12305 (2010). seismicity and fault structure on the Discovery transform fault, East Pacific Rise. 22. Y. J. Liu, J. J. McGuire, M. D. Behn, Frictional behavior of oceanic transform faults and Geochem. Geophys. Geosyst. 15, 3698–3712 (2014). its influence on earthquake characteristics. J. Geophys. Res. 117, B04315 (2012). 7. V. M. Kuna, J. L. Nabelek, J. Braunmiller, Mode of slip and crust-mantle interaction at 23. J. H. Dieterich, Modeling of rock friction 1. Experimental results and constitutive oceanic transform faults. Nat. Geosci. 12, 138–142 (2019). equations. J. Geophys. Res. 84, 2161–2168 (1979). 8. J. McGuire, Seismic cycles and earthquake predictability on East Pacific Rise transform 24. A. L. Ruina, Slip instability and state variable friction laws. J. Geophys. Res. 88, 10359– faults. Bull. Seismol. Soc. Am. 98, 1067–1084 (2008). 10370 (1983). 9. K. Aderhold, R. E. Abercrombie, The 2015 Mw 7.1 earthquake on the Charlie-Gibbs 25. C. He, Z. Wang, W. Yao, Frictional sliding of gabbro gouge under hydrothermal transform fault: Repeating earthquakes and multimodal slip on a slow oceanic conditions. Tectonophys. 445, 353–362 (2007). transform. Geophys. Res. Lett. 43, 6119–6128 (2016). 26. Y. Kaneko, J. Avouac, N. Lapusta, Towards inferring earthquake patterns from 10. E. Roland, M. Behn, G. Hirth, Thermal-mechanical behavior of oceanic transform geodetic observations of interseismic coupling. Nat. Geosci. 3, 363–369 (2010). faults: Implications for the spatial distribution of seismicity. Geochem. Geophys. 27. W. H. Bakun et al., Implications for prediction and hazard assessment from the 2004 Geosys. 11, Q07001 (2010). Parkfield earthquake. Nature 437, 969–974 (2005). 11. P. A. Moyer, M. S. Boettcher, J. J. McGuire, J. A. Collins, Spatial and temporal 28. A. M. Rubin, D. Gillard, J. L. Got, Streaks of microearthquakes along creeping faults. variations in earthquake stress drop on Gofar Transform Fault, East Pacific Rise: Nature 400, 635–641 (1999). Implications for fault strength. J. Geophys. Res. 123, 7722–7740 (2018). 29. J. Jiang, Y. Fialko, Reconciling seismicity and geodetic locking depths on the Anza 12. E. S. Cochran et al., Seismic and geodetic evidence for extensive, long-lived fault section of the San Jacinto Fault. Geophys. Res. Lett. 43, 10,663–10,671 (2016). damage zones. Geology 37, 315–318 (2009). 30. J. Samuelson, D. Elsworth, C. Marone, Shear-induced dilatancy of fluid saturated 13. S. G. Wesnousky, Seismological and structural evolution of strike-slip faults. Nature faults: Experiment and theory. J. Geophys. Res. 114, B12404 (2009). 335, 340–343 (1988). 31. J. Samuelson, D. Elsworth, C. Marone, Influence of dilatancy on the frictional consti- 14. G. P. Biasi, S. G. Wesnousky, Steps and gaps in ground ruptures: Empirical bounds on tutive behavior of a saturated fault zone under a variety of drainage conditions. J. rupture propagation. Bull. Seismol. Soc. Am. 106, 1110–1124 (2016). Geophys. Res. 116, B10406 (2011).

6 of 7 | www.pnas.org/cgi/doi/10.1073/pnas.1913625117 Liu et al. Downloaded by guest on October 1, 2021 Downloaded by guest on October 1, 2021 4 .Daet .Wn,T .Jms ietsi vn ntedee acdasubduction Cascadia deeper the on event slip silent A James, S. T. Wang, K. Dragert, and H. 2007 in 34. events slip slow Boso Salton The the Kimura, H. in Kimura, creep T. Matsuzawa, aseismic T. Hirose, by H. driven 33. swarms Earthquake McGuire, J. Lohman, R. 32. 6 .A enn .D ek,T .Tli,Tefitoa eairo iadt n antigorite and lizardite of behavior frictional The Tullis, E. T. Weeks, D. J. Reinen, A. L. 36. accompanied events slip slow Episodic Kasahara, K. Yamamizu, F. Hirose, H. Obara, K. 35. 8 .Hlkm,M rdisi eahuterhuk wrsidct frictional indicate swarms earthquake Megathrust Brudzinski, M. Holtkamp, S. faults. transform 38. on swarms Earthquake McGuire, J. J. Roland, E. 37. i tal. et Liu interface. swarm. earthquake accompanying the for process Lett. driving a as 2011 California. Trough, epniie xeiet,cntttv oes n mlctosfrntrlfaults. natural for implications Geophys. and Appl. Pure. models, constitutive experiments, - serpentinites zone. subduction (2004). Japan L23602 southwest in tremors non-volcanic by (2014). ruptures. earthquake large delimit which changes (2009). 1677–1690 7828 (2014). 2778–2785 41, Science 5512 (2001). 1525–1528 292, .Gohs Res. Geophys. J. 1–5 (1994). 317–358 143, –0(2007). 1–10 112, at lntSi Lett. Sci. Planet Earth epy.Rs Lett. Res. Geophys. epy.J Int. J. Geophys. epy.Res. Geophys. 234–243 390, 178, 31, 6 .Mzgci .Hrs,T hmmt,E uuaa nenlsrcueadper- and structure Internal Fukuyama, E. Shimamoto, T. Hirose, T. Mizoguchi, K. fault. 46. a on slip of complexity Spatiotemporal Rice, R. J. 45. friction. rate-and-state earthquake. and events Chile slip 2014 slow Episodic April Rubin, M. 1 A. the from 44. foreshocks Recognizing Lay, T. Brodsky, E. 43. seis- Ito Y. within earthquakes 42. of aseismic migration by Spatial triggered Abercrombie, R. swarms Shearer, seismic P. Chen, Modeling X. Ogata, Y. 41. McGuire, J. Llenos, A. 40. Wei S. 39. eblt fteNjm al,suhetJapan. southwest fault, Nojima the (2008). of meability (1993). (2008). B11414 Science earthquake. Tohoku-Oki diffusion. ﬂuid for (2012). Evidence B04301 California: Southern in clusters mic transients. Lett. Sci. Planet Earth psdcso lpeet nteJpnsbuto oebfr h 2011 the before zone subduction Japan the in events slip slow Episodic al., et h 02Balysamtigrdb neto-nue simcslip. aseismic injection-induced by triggered swarm Brawley 2012 The al., et 0–0 (2014). 700–702 344, at lntSi Lett. Sci. Planet Earth 1–2 (2015). 115–125 422, Tectonophys. 96 (2009). 59–69 281, 42 (2013). 14–26 600, NSLts Articles Latest PNAS .Gohs Res. Geophys. J. .Src.Geol. Struct. J. .Gohs Res. Geophys. J. .Gohs Res. Geophys. J. 9885–9907 98, 513–524 30, | f7 of 7 117, 113,

EARTH, ATMOSPHERIC, AND PLANETARY SCIENCES