Off-Fault Damage Zone in Rupture Mechanics

• Off-fault damage zone is the region surrounding a fault where brittle and inelastic deformations occur, dissipating elastic strain energy and influencing rupture propagation.
• Integrated studies using laboratory tests, field mapping, seismology, and numerical simulations reveal that damage zone width and structure vary dramatically from millimeter-scale microfractures to kilometer-scale distributed faults.
• Dynamic rupture models incorporating off-fault damage demonstrate its significant impact on seismic radiation, hydrological response, and energy partitioning, emphasizing the need for multiscale representations.

Searching arXiv for recent and foundational papers on off-fault damage zones in dynamic rupture, laboratory characterization, and related modeling. An off-fault damage zone is the region of rock surrounding a principal fault trace or slipping fault core that undergoes distributed brittle deformation or inelastic deformation during rupture, rather than slip being confined to the principal surface itself. Across laboratory, field, seismological, and numerical studies, it includes microfractures and secondary cracks in wall rock, distributed strain around complex fault junctions, and, in some fault zones, multiscale fracture networks that can themselves host dynamic rupture. Its significance follows from energy partitioning: part of the elastic strain energy released by shear failure is dissipated as off-fault damage, altering rupture dynamics, slip-weakening behavior, seismic radiation, and near-fault elastic and hydraulic properties (Aben et al., 2020, Klinger et al., 2018, Palgunadi et al., 2023).

1. Definition, geometry, and structural expression

The off-fault damage zone is distinct from the principal slip surface and from the much narrower on-fault process zone. In 2D plane-strain rupture mechanics, on-fault slip is accommodated within a process zone of size

$$R_0=\frac{9\pi}{32(1-\nu)}\frac{D_c\mu}{(\mu_s-\mu_d)(-\sigma_{yy})},$$

whereas the damage zone width $W_d$ is defined as the distance from the fault trace to the locus where a chosen damage threshold is first exceeded (Ribes et al., 2022). Outside $W_d$ the rock remains essentially elastic; inside $W_d$ irreversible deformation has consumed part of the elastic strain energy radiated by the rupture front (Ribes et al., 2022).

Its geometric expression depends on scale and setting. In continental surface rupture, off-fault damage zones can span hundreds to thousands of meters, as in the Kaikoura triple-junction area, where the Papatea fault shows a distributed damage zone width of $\sim2$ km, in contrast to tens of meters on the Jordan and Kekerengu faults (Klinger et al., 2018). In laboratory granite experiments, the same phenomenon appears as a millimetre-scale zone of microfracturing: after quasi-static failure the damage zone width is around $10$ mm, and after dynamic failure around $20$ mm (Aben et al., 2020). At seismogenic depth, aftershock clustering reveals finite-thickness planar patches with mean thickness $T=4\sigma_3\approx290$ m, suggesting that deformation is volumetric rather than confined to an infinitely thin core (Ouillon et al., 2010).

In explicit multiscale fault-zone models, the damage zone is volumetric and networked rather than a uniform halo. One 3D representation uses a $10$ km (strike) $\times1$ km (normal) $W_d$0 km (depth) region around an $W_d$1 km $W_d$2 km listric main fault, containing $W_d$3 explicit fractures with radii $W_d$4 between $W_d$5 m and $W_d$6 m, following a power-law size distribution $W_d$7 (Palgunadi et al., 2024). In the related multiscale study, two conjugate fracture families strike on average N120°E and N20°E, each with a $W_d$8 Fisher-distributed scatter in strike and dip; fracture lengths follow $W_d$9 between $W_d$0 m and $W_d$1 m; and fracture density decays away from the fault core as

$W_d$2

with $W_d$3 and $W_d$4 (Palgunadi et al., 2023). About $W_d$5 of fractures connect to more than one neighbor, a structural property that directly conditions rupture transfer (Palgunadi et al., 2023).

Near-surface field characterization also shows that “damage zone” is not internally homogeneous. In the Vado di Corno Fault Zone, a high-strain damage zone forms a very slow layer with $W_d$6–$W_d$7 km/s and thickness $W_d$8–$W_d$9 m in shallow seismic tomography, while field mapping places high-strain damage-zone lenses $W_d$0–$W_d$1 m outward from the principal fault surface (Fondriest et al., 8 Jan 2026). This suggests that the off-fault damage zone is better regarded as a mechanically heterogeneous envelope composed of units with different fracture intensity, sealing state, and compliance, rather than as a single material class.

2. Energetics and constitutive mechanics

In rupture mechanics, the strain energy released by shear failure is partitioned into radiated seismic waves, frictional work on the slip surface, and breakdown work. Breakdown work contains the fracture energy $W_d$2 dissipated in the tip process zone, of which a fraction $W_d$3 is spent in creating off-fault damage (Aben et al., 2020). A corresponding dynamic balance used in multiscale rupture models is

$W_d$4

where $W_d$5 is the elastic energy release rate (Palgunadi et al., 2023).

A recurrent result is that off-fault damage is an active energy sink rather than a passive elastic buffer. Laboratory measurements in Lanhélan granite give $W_d$6 kJ/m$W_d$7 for quasi-static rupture and $W_d$8 kJ/m$W_d$9 for dynamic rupture (Aben et al., 2020). In the quasi-static run, total fracture energy on the principal slip zone was $\sim2$0 kJ/m$\sim2$1, so that $\sim2$2; in the dynamic runs, the additional $\sim2$3 kJ/m$\sim2$4 of off-fault energy implies a $\sim2$5 increase in total $\sim2$6 with rupture speed (Aben et al., 2020). In 2D FDEM simulations, the energy fraction associated with off-fault fracture energy rises with rupture length and depth: at $\sim2$7, $\sim2$8 of total released energy at $\sim2$9 km, $10$0 at $10$1 km, and $10$2 at $10$3 km (Okubo et al., 2019).

Several constitutive descriptions are used to represent this energy sink. In multiscale fracture-network rupture, both main fault and fractures obey a laboratory-based rate-and-state friction law with rapid velocity weakening, with effective normal stress

$10$4

and relative pre-stress ratio

$10$5

An optimally oriented fracture has $10$6; sustained slip generally requires $10$7, while truly favorable fractures have $10$8 (Palgunadi et al., 2023). In those simulations, the state-evolution distance is tied to fracture radius by $10$9, corresponding to $20$0 (Palgunadi et al., 2023).

Other frameworks distribute damage in the bulk. Micromechanical models define damage by growth of tensile wing cracks from pre-existing penny-shaped cracks, with

$20$1

and evolving elastic moduli $20$2 and $20$3 (Ferry et al., 2024). Continuum damage–breakage models use a scalar damage $20$4 and breakage variable $20$5, causing reductions in shear modulus $20$6 and bulk modulus $20$7 and producing either smooth distributed damage or localized shear bands once brittle failure is reached (Niu et al., 27 Mar 2025). Phase-field models instead use a continuous damage variable $20$8 and a regularization length $20$9, so that fault propagation and off-fault damage follow from a single energy functional (Fei et al., 2022, Fei et al., 9 Mar 2025).

3. Observation and quantification

Off-fault damage zones are quantified by different observables in different regimes, but several studies converge on measurable width, density, and property-reduction metrics.

In laboratory triaxial experiments on intact Lanhélan granite at $T=4\sigma_3\approx290$0 MPa confining pressure, time-resolved 3D $T=4\sigma_3\approx290$1-wave velocity tomography and post-mortem microstructures quantify off-fault damage during and after failure (Aben et al., 2020). Tomography reveals a narrow low-velocity band around the slip surface with up to $T=4\sigma_3\approx290$2 drop in $T=4\sigma_3\approx290$3, and SEM fracture maps show fracture densities decaying to background over $T=4\sigma_3\approx290$4 mm after quasi-static rupture and $T=4\sigma_3\approx290$5 mm after dynamic rupture (Aben et al., 2020). The conversion from $T=4\sigma_3\approx290$6 to crack density uses an effective-medium model for dry rock with penny-shaped cracks, fitting $T=4\sigma_3\approx290$7 and $T=4\sigma_3\approx290$8 in each voxel to recover the crack-density tensor components $T=4\sigma_3\approx290$9 and $10$0 (Aben et al., 2020).

Field-scale geodetic mapping resolves much broader off-fault deformation. In the Kaikoura earthquake, optical image correlation combines SPOT6 and Pleiades imagery, with a high-resolution correlation at $10$1 m pixel and detection threshold $10$2 cm (Klinger et al., 2018). Swath profiles show maximum right-lateral slip of $10$3 m on the Kekerengu fault and total left-lateral slip of $10$4 m distributed over $10$5 km width on the Papatea fault (Klinger et al., 2018). There, damage zone width is defined where the displacement gradient is diffuse rather than localized (Klinger et al., 2018).

The supershear-transition literature uses the off-fault damage zone itself as a diagnostic. For the 2001 Kunlun earthquake, optical image correlation yields a mean fault-zone width of $10$6 m for most of the rupture, but over an $10$7 km stretch the width collapses to $10$8 m, coincident with the inferred sub-to-supershear transition (Jara et al., 2019). Early aftershock catalogs for the Izmit, Denali, and Craig earthquakes show corresponding along-strike gaps in cumulative seismic-moment density over a few-kilometer segment at the transition location (Jara et al., 2019).

Seismological and near-surface structural methods add further constraints. A Gaussian-mixture clustering analysis of the Mount Lewis aftershocks reconstructs anisotropic planar patches with mean length $10$9 km, mean width $\times1$0 km, and mean thickness $\times1$1 m, a thickness an order of magnitude larger than relative location errors (Ouillon et al., 2010). In the Vado di Corno Fault Zone, ultrasonic measurements at $\times1$2 MHz and field tomography at $\times1$3 Hz together show that the fault core is slower than some damage-zone units at laboratory scale, yet tomography resolves a very slow intensely fractured high-strain damage zone at shallow depth (Fondriest et al., 8 Jan 2026).

Context	Width or thickness	Diagnostic
Laboratory granite rupture	$\times1$4 mm quasi-static; $\times1$5 mm dynamic	Microfracture density and $\times1$6 drop (Aben et al., 2020)
Kaikoura Papatea fault	$\times1$7 km	Diffuse displacement gradient (Klinger et al., 2018)
Kunlun supershear study	$\times1$8 m to $\times1$9 m	Optical image correlation (Jara et al., 2019)
Mount Lewis aftershocks	$W_d$00 m thickness	Gaussian-kernel seismic clustering (Ouillon et al., 2010)
Vado di Corno HSDZ	$W_d$01–$W_d$02 m tomography; $W_d$03–$W_d$04 m field lenses	Seismic tomography and field mapping (Fondriest et al., 8 Jan 2026)

These measurements are not interchangeable. They sample different materials, frequencies, depths, and definitions of damage. A plausible implication is that “width” is operational rather than unique: it depends on whether the threshold is based on crack density, velocity reduction, inelastic strain, scalar damage, or diffuse displacement.

4. Rupture interaction, scaling, and dynamic regimes

Off-fault damage zones interact dynamically with rupture through branching, jumping, stress transfer, and energy competition. In multiscale fracture networks, rupture on one fracture imposes dynamic shear stress changes $W_d$05 and normal stress changes $W_d$06 on neighbors; if $W_d$07 raises the local shear above the peak strength while $W_d$08 unclamps neighboring fractures, neighboring fractures may nucleate (Palgunadi et al., 2023). Two transfer modes are distinguished: branching at direct intersections, and jumping by dynamic overstress at distance $W_d$09 m (Palgunadi et al., 2023).

Successful cascading rupture within a fracture-network damage zone requires a specific combination of structure, prestress, and frictional scaling. In the 3D listric-fault models, cascades arise if at least one fracture family is favorably oriented with $W_d$10 under ambient $W_d$11 azimuth $W_d$12–$W_d$13, while the conjugate set has $W_d$14; fracture connectivity and spacing $W_d$15 m ensure rupture transfer; and $W_d$16 scaling with fracture size provides sufficient energy for self-sustained rupture (Palgunadi et al., 2023). Under these conditions, cascades involving hundreds of fractures can release moment magnitudes up to $W_d$17 without engaging the main fault, and cascading rupture within the fracture network discourages rupture on the main fault (Palgunadi et al., 2023). In “pure cascade” cases near $W_d$18, the main fault remains dynamically quiescent except for small self-arresting patches, whereas at $W_d$19–$W_d$20 runaway rupture occurs on the main fault with limited off-fault fracture slip (Palgunadi et al., 2023).

Rupture speed leaves a direct signature in damage-zone width. Laboratory experiments show that for slips up to a few mm, the spatial extent of off-fault peak stresses, and hence damage width, is controlled by rupture speed $W_d$21, not by cumulative slip; the observed doubling from $W_d$22 to $W_d$23 mm implies $W_d$24 in the dynamic runs (Aben et al., 2020). Fracture-mechanics analysis of supershear transition predicts that damage-zone half-width scales approximately as

$W_d$25

so that width first broadens and then collapses as $W_d$26 because the dynamic stress intensity factor contracts (Jara et al., 2019). FDEM simulations of planar supershear rupture show a corresponding “damage gap” around the transition point, with minimal off-fault failure during the jump (Okubo et al., 2019).

Depth dependence is robust, but model-specific. In one micromechanical formulation,

$W_d$27

with $W_d$28, implying narrower yet denser damage at depth (Ferry et al., 2024). In 3D continuum damage–breakage simulations, the $W_d$29 $W_d$30 contour extends $W_d$31 km from the fault at $W_d$32 km depth but only $W_d$33 km at $W_d$34 km depth, summarized by a simple scaling argument $W_d$35 (Niu et al., 27 Mar 2025). This suggests that narrowing with depth is consistent across models, while the precise exponent is formulation-dependent.

5. Numerical representations and methodological issues

A large part of the literature concerns how to represent off-fault damage numerically without collapsing it into an imposed plastic halo or ignoring it entirely.

Combined finite–discrete element methods represent off-fault fractures explicitly on element interfaces. In HOSSedu, each potential interface can undergo tensile or shear cohesion softening and slip-weakening friction, allowing spontaneous off-fault crack networks to emerge from local stress concentrations (Okubo et al., 2019). This framework has been used for planar faults, kinks, step-overs, and rough faults; with off-fault damage included, a rupture can arrest at a compressional kink, jump a dilational step-over through relay-zone fracturing, or arrest early on a rough fault because cracking amplifies stress heterogeneity (Okubo et al., 2019). The Kaikoura rupture reconstruction uses the same 2D FDEM approach to compare alternative rupture paths through the Jordan–Kekerengu–Papatea triple junction, with the preferred scenario reproducing both narrow localization on Jordan/Kekerengu and broad distributed deformation on Papatea (Klinger et al., 2018).

Micromechanical continuum models resolve damage through internal variables rather than discrete cracks. In one formulation, wing-crack growth updates a scalar damage field $W_d$36 and therefore the homogenized elastic stiffness tensor at each time step, permitting three local regimes: purely elastic, sliding along microcracks, and tensile opening of penny and wing cracks (Ferry et al., 2024). A distinct 3D continuum damage–breakage model uses damage $W_d$37 and breakage $W_d$38 to distinguish smooth distributed damage from localized, mesh-independent shear bands, with observed band orientations in agreement with a linearized bifurcation analysis (Niu et al., 27 Mar 2025).

Phase-field models unify on-fault rupture and off-fault damage with a single diffuse variable. In the quasi-dynamic rate-and-state formulation, off-fault damage is the region where $W_d$39, and a practical thickness measure is obtained by choosing a threshold $W_d$40 and defining the extent normal to the main rupture where $W_d$41 (Fei et al., 2022). In the in-plane extension, damage zone thickness $W_d$42 is naturally of order the phase-field width $W_d$43, while the model also incorporates pressure-dependent shear strength and a revised propagation criterion that accounts explicitly for the coupling between shear strength and normal stress (Fei et al., 9 Mar 2025).

Diffuse-interface hyperbolic formulations take a broader continuum view. The GPR model introduces a damage concentration $W_d$44 and a solid volume fraction $W_d$45, allowing secondary cracks and complex fault geometries to be represented without mesh alignment or remeshing (Gabriel et al., 2020). In the dynamic rupture examples, the diffuse fault core is imposed with width $W_d$46 m and spontaneous off-fault shear cracks develop in the extensional quadrants, with a typical off-fault damage zone width $W_d$47–$W_d$48 m for $W_d$49 (Gabriel et al., 2020).

A central methodological issue is stress initialization. For 2D strike-slip rupture under plane strain, neglecting the out-of-plane stress or choosing $W_d$50 arbitrarily can move the model into a forbidden reverse-fault regime and lead to a six-fold underestimation of damage-zone width: in the studied cases, correct 3D stress ordering gives $W_d$51 m, whereas the incorrect setup gives $W_d$52 m (Ribes et al., 2022). This is not a minor technicality, because the same factor-of-six appears whether damage is diagnosed by $W_d$53, plastic strain $W_d$54, or scalar damage $W_d$55 (Ribes et al., 2022).

6. Seismic radiation, hydrology, and hazard implications

Off-fault damage zones affect not only rupture mechanics but also source spectra, near-field shaking, groundwater response, and rupture reconstruction.

Explicit 3D fracture-network simulations show that cascading ruptures in the damage zone generate distinct ground-motion signatures with high-frequency content, resembling source coda-wave-like signatures (Palgunadi et al., 2024). In the modeled cascading case, near-field corner frequency is $W_d$56 Hz with standard deviation $W_d$57 Hz, versus $W_d$58 Hz with standard deviation $W_d$59 Hz in non-cascading cases; the seismologically inferred stress drop is $W_d$60 MPa, an order of magnitude larger than the directly computed average off-fault $W_d$61 MPa (Palgunadi et al., 2024). Peak ground acceleration and peak ground velocity are roughly twice those of the non-cascading cases despite the smaller moment magnitude, and pseudo-spectral accelerations at short periods $W_d$62–$W_d$63 s are elevated (Palgunadi et al., 2024).

Other dynamic-rupture frameworks yield the same qualitative conclusion by different mechanisms. FDEM simulations show enhanced high-frequency radiation in the near field due to coseismic off-fault damage, with $W_d$64 Hz signals arriving after the main rupture and a reduction of seismic efficiency by up to $W_d$65 at $W_d$66 km depth when damage is included (Okubo et al., 2019). Continuum damage–breakage simulations show that rapidly evolving bulk moduli generate broadband energy above $W_d$67 Hz and nearly isotropic fault-normal to fault-parallel amplitude ratios in the $W_d$68–$W_d$69 Hz band, while also reducing rupture speed and increasing supershear transition distance (Niu et al., 27 Mar 2025).

The hydraulic consequences are likewise non-negligible. A 3D groundwater model coupled to dynamic rupture with inelastic off-fault deformation shows that inelastic dilation causes notable depressurization within $W_d$70 to $W_d$71 km off the fault at shallow depths $W_d$72 km, with modeled coseismic pore-pressure drops of approximately $W_d$73 to $W_d$74 MPa within $W_d$75 km of the fault (Hung et al., 30 Dec 2025). Near-fault water-level drawdowns reach $W_d$76–$W_d$77 m at $W_d$78 m, $W_d$79–$W_d$80 m at $W_d$81 km, and $W_d$82–$W_d$83 m at $W_d$84 km, whereas elastic-only poroelasticity gives only a few meters and opposite-sign changes across the fault (Hung et al., 30 Dec 2025). This indicates that off-fault damage can act as a transient pressure sink during the coseismic phase and accelerate postseismic recharge (Hung et al., 30 Dec 2025).

In porous-media flow modeling, the damage zone can also be represented as a distinct hydrogeologic layer rather than collapsed into the core. A mixed-dimensional Darcy model treats the damage zone as a lower-dimensional manifold $W_d$85 of thickness $W_d$86, with its own tangential permeability and Robin-type coupling to both rock matrix and fault core (Fumagalli et al., 2019). This captures pressure-drop localization at damage-zone interfaces, as well as flow focusing or diversion around low-permeability patches (Fumagalli et al., 2019).

For seismic hazard and rupture interpretation, the principal implication is that off-fault damage is diagnostically informative and dynamically consequential. Off-fault deformation “fingerprints” can distinguish competing rupture paths in complex systems such as Kaikoura (Klinger et al., 2018). Highly connected, favorably pre-stressed fracture networks can shadow or inhibit main-fault rupture while releasing off-fault seismic moment up to $W_d$87 (Palgunadi et al., 2023). Standard single-plane source models and conventional ground-motion prediction equations may therefore underpredict near-fault PGA and high-frequency content when multiscale fault-zone complexity is ignored (Palgunadi et al., 2024).