Experimental evidence of seismic ruptures initiated by aseismic slip

Faults accommodate tectonic loading via both rapid seismic ruptures and slow slip events (SSEs). SSEs are now recognized in subduction and strike‑slip settings and can span a wide range of spatial scales, sometimes triggering large earthquakes and at other times recurring without leading to major events. A central open question is how aseismic slip interacts with and influences frictionally locked (coupled) regions, and what fault properties control the transition between slow and seismic slip. Conventional models often attribute SSEs to velocity‑strengthening regions that should not rupture seismically, yet geometric complexity alone can induce slow slip, and observations show that zones capable of seismic slip can also host slow slip. This study tests the hypothesis that a localized slow‑slipping heterogeneity within a fault can act as a nucleation center for seismic rupture, thereby modifying the stick‑slip cycle and increasing event frequency. Using a controlled laboratory interface, the work explores how a granular patch (emulating a slow‑slip zone) interacts with adjacent locked regions and how normal load distribution governs the extent and role of aseismic slip in rupture initiation.

Prior studies document the continuum between earthquakes and slow slip, including tremor and low‑frequency earthquakes, and show SSEs in both subduction and crustal faults. Modeling and observations indicate that fault geometrical complexity can spontaneously produce fast and slow slip and that seismically active regions can also exhibit slow slip. Laboratory experiments have identified slow ruptures under quasi‑static loading or under low normal stress/high pore pressure, and demonstrated that interfacial heterogeneities can confine ruptures, alter stick‑slip frequency, and host creep fronts. Fracture mechanics has been successfully applied to frictional rupture dynamics, relating rupture initiation to energy release and fracture energy. Despite this, the boundary between slow‑slipping and seismogenic zones and the mechanisms by which aseismic slip triggers seismic rupture remain insufficiently constrained, motivating controlled laboratory investigation.

A quasi‑1D frictional interface was built from two PMMA blocks (150 × 90 × 10 mm) whose contacting faces have 1 μm r.m.s. roughness. A semi‑elliptical recess was machined into each face to form a central eye‑shaped hole (≈30 × 6 × 10 mm). The hole could be left empty or filled with a 2D granular layer composed of nylon cylinders (length equal to block thickness; diameters 0.4, 0.7, 0.9, 1.3 mm in approximate volume ratios 5, 10, 35, 50%) to emulate a slow‑slip heterogeneity. To force interfacial slip within the granular pile, 1.3 mm cylinders were glued into grooves on the hole surfaces to roughen them. The granular packing fraction was varied by the number of cylinders inserted, producing different loading contrasts. Material properties: PMMA Young’s modulus E ≈ 3 GPa (low strain rate) to 5.6 GPa (high strain rate), Poisson’s ratio ν = 0.3, Rayleigh wave speed CR ≈ 1237 m/s; nylon E = 1.4 GPa, ν = 0.4. Loading and measurements: Normal load was applied via a manual press by displacement control, typically FN ≈ 3000 N (range 2800–3200 N), with additional tests at 750–3000 N. Shear was applied by translating the lower block at 20 μm/s. Forces Fy (normal) and Fs (shear) were recorded at 315 Hz. Ten strain‑gauge rosettes on the upper block (8 above solid‑solid sections, 1 above the central patch, spacing ≈10 mm) measured 2D strain tensors continuously at 315 Hz and during rapid events at 4 MHz. Interfacial slip tracking: A sub‑pixel particle‑tracking method (image correlation with Gaussian peak fitting) tracked positions of patterned cylinder faces and painted patterns on block faces at 100 fps, achieving ≈8 μm displacement resolution. Slip was measured at 4 positions along solid‑solid sections and 3 positions above the hole by subtracting top and bottom displacements at the same x location, with corrections for sample rotation. Quantities and definitions: Normal stress distributions along the interface were inferred from strain gauges to define σpatch (above the hole) and σsolid (above solid‑solid sections). A loading contrast Cl was defined from the difference between averaged σpatch and σsolid normalized by the interface‑averaged σy, yielding values from ≈−1 (empty hole) up to ≈2.5 (dense granular patch). Interfacial total slip ssolid(t), spatch(t) and cumulated inter‑event slip Ssolid(t), Spatch(t) were computed, excluding slip during rapid events. A normalized cumulated inter‑event slip S(I)(t) = s(I)(t)/s0(I)(t) quantified coupling, with converged values S taken over the last events. Rupture detection and nucleation: High‑frequency shear strain drops Δεxy(x,t) were used to detect rupture passage and pick nucleation locations (first departure from baseline), and to estimate rupture speeds from first‑arrival times across gauges. The slipping patch extent was inferred from low‑frequency shear‑strain evolution: linear loading indicates locked behavior, while sublinear (saturating) εxy(t) indicates local slow slip. For each inter‑event period, gauges exhibiting sublinear evolution were counted to estimate the slipping patch length ⟨lslip⟩ as a function of loading contrast. Controls: Experiments without a hole (flat solids), with an empty hole, and with granular patches at fixed FN but varying granular density were compared to isolate effects of unloading vs. compositional heterogeneity on stick‑slip frequency.

• The presence of a granular patch increases stick‑slip frequency beyond what is expected from a simple reduction in normal stress on solid‑solid sections. Period shortening in granular experiments exceeds that observed in no‑hole or empty‑hole experiments at comparable ⟨σsolid⟩.
• Differential inter‑event slip develops: the granular patch exhibits significant aseismic inter‑event slip while adjacent solid‑solid sections remain mostly locked. The differential Spatch − Ssolid increases with loading contrast and correlates with higher mean stick‑slip frequency ⟨1/ΔT⟩.
• Rupture dynamics: High‑frequency strain shows each force drop corresponds to a rupture spanning the entire interface, with rupture speeds ranging from ~500 to ~2500 m/s, indicating both sub‑Rayleigh and supershear events.
• Nucleation locations shift outward with increasing loading contrast: histograms of nucleation positions show that as contrast increases, ruptures initiate farther from the central patch, moving toward the outer corners. The mean nucleation distance ⟨dnuc⟩ increases with contrast, implying extension of a central slipping zone that does not host dynamic nucleation.
• Slipping zone growth: Low‑frequency strain indicates an expanding slowly slipping patch during inter‑event periods whose average length ⟨lslip⟩ increases with loading contrast and can exceed the hole length (30 mm), extending into solid‑solid sections.
• Initiation mechanism consistent with fracture mechanics: As ⟨lslip⟩ increases, the stress needed to destabilize the initial (aseismic) rupture decreases, qualitatively matching Griffith‑type scaling τc ∝ √(Gc E / l). Additionally, the normal stress measured at nucleation points decreases with contrast, suggesting a local reduction of effective fracture energy at the tip further lowers initiation thresholds.
• The granular patch acts as a nucleation seed by establishing an initial rupture (a creeping, decoupled zone) that destabilizes earlier, reducing the inter‑event loading time and increasing event frequency.
• Contrary to a simple corner stress‑concentration scenario, slow slip in the patch promotes creep of neighboring contacts and modifies the nucleation phase of the entire interface rather than triggering local dynamic failure at the patch edges.

The experiments directly address how aseismic slip interacts with locked regions to affect seismic rupture initiation. Introducing a granular heterogeneity creates a partially coupled, slowly slipping zone that expands with increased normal load on the patch. This central slow‑slip zone functions as an initial rupture: by lengthening ⟨lslip⟩, it lowers the critical stress for dynamic propagation, consistent with fracture‑mechanics initiation trends. Simultaneously, the normal stress at nucleation points decreases with contrast, implying a local reduction of effective fracture energy that further eases destabilization. Together, these effects shift nucleation away from the center and lead to more frequent, system‑spanning ruptures at lower shear stress levels, thereby shortening the stick‑slip period. The findings suggest that aseismic slip can actively initiate seismic events rather than being merely a byproduct or barrier to rupture. Two mechanisms may facilitate stable slip and contact creep near the patch: (1) time‑dependent aging vs. low‑rate loading leading to creep of contacts; and (2) shear‑induced dilatancy of the granular patch, locally reducing normal stress and promoting stable sliding. Unlike scenarios in which heterogeneities act as arresting barriers or evolve frictional properties through prior sliding, the dominant effect here is modification of the nucleation process by a growing creeping zone. Implications include the need to monitor the evolution and spatial extent of slow‑slip regions along natural faults, as their growth can precondition faults for earlier and more frequent dynamic ruptures. The laboratory framework demonstrates that off‑fault strain measurements and fracture‑mechanics concepts can be combined to infer destabilization conditions of creeping patches, advancing seismic hazard assessment strategies.

A localized slow‑slipping area embedded in a frictional interface serves as a nucleation center for seismic rupture. As the normal load carried by the granular patch increases, the slowly slipping zone extends beyond the heterogeneity into adjacent solid‑solid regions, lowering the stress threshold for dynamic rupture and increasing stick‑slip frequency. Rupture nucleation locations migrate outward, and the response follows fracture‑mechanics‑based initiation trends wherein a longer initial (aseismic) rupture requires lower critical stress for destabilization. These results show that compositional heterogeneity modifies rupture nucleation and the seismic cycle by promoting aseismic precursors that seed dynamic failure. Future directions include: higher‑resolution temporal measurements to resolve slow nucleation fronts; quantitative estimation of spatially varying fracture energy and residual stress within the creeping zone; acoustic emission measurements to link source properties to radiated waves; and modeling of slow‑slip zone growth and its destabilization length under varying loading conditions. Monitoring the evolution of slow‑slip regions on natural faults may improve forecasting of nucleation conditions and seismic hazard.

• Fracture energy Gc could not be directly quantified due to heterogeneous normal stress and limited spatial sampling near the slipping zone; residual stress distributions were not resolved, precluding a full energy‑balance analysis.
• The slipping patch length and local fracture energy could not be controlled independently in this setup, complicating separation of their contributions to initiation.
• Spatial resolution of slip tracking limits direct measurement of the precise slipping‑zone extent; inference relies partly on low‑frequency strain indicating sublinear loading.
• Temporal resolution was insufficient to fully characterize slow nucleation dynamics and creep‑front propagation.
• Laboratory materials (PMMA and nylon) and geometry provide analog behavior; scaling to natural faults involves uncertainties (e.g., fluid effects, temperature, roughness, true gouge behavior).