The internal fine structure of a fault zone （e.g., width, depth, and velocity contrast） directly controls the accumulation and release of fault stress, governing earthquake nucleation and rupture propagation. The Qujiang fault zone, situated at the southeastern margin of the Sichuan-Yunnan block, represents the seismogenic structure responsible for the 1970 M_S7.7 Tonghai earthquake. Accurate delineation of the boundaries and width of the low-velocity zone （LVZ） within this fault is critical for regional seismic hazard assessment. Currently, no unified criteria exist for imaging fault-zone structures using teleseismic travel-time delays from dense seismic arrays: some studies define fault boundaries as locations marked by abrupt increases in travel-time delays, whereas others adopt the average travel-time delay outside the fault zone as a fixed threshold. Using waveform modeling based on data from a dense linear array deployed across the Qujiang fault zone and the QSSP-CGFD3D teleseismic hybrid simulation method, this study assesses the applicability of these identification criteria and performs high-resolution characterization of the Qujiang fault zone structure.

We deployed an about 26-km-long dense linear array of 179 SmartSolo short-period seismometers across the Qujiang fault zone, with a station spacing of 100 m within 5 km on both sides of the fault and 200 m elsewhere. Over the observation period from August to October 2023, we selected 12 high-signal-to-noise-ratio teleseismic events （8 from the south, 4 from the north） with epicentral distances ranging from 30° to 90°. After filtering waveform data to a passband of 0.05–1.5 Hz, we picked the first peak arrival times of direct P-waves. Relative travel-time delays were calculated relative to the central station by subtracting theoretical arrival times derived from the AK135 velocity model and topographic delays （assuming a near-surface velocity of 2 km/s）. Two boundary-identification criteria were subsequently applied: Criterion 1 defines boundaries at locations of abrupt increases in travel-time delays, while Criterion 2 uses the average travel-time delay outside the fault zone （−0.06 s） as a fixed threshold. To evaluate these two criteria, we performed three-dimensional wavefield numerical simulations using the QSSP-CGFD3D teleseismic hybrid approach, by comparing synthetic travel-time delays from diverse fault-zone models （e.g., a 2-km-deep zone with 30% velocity reduction） with observational data.

Observed travel-time delay patterns indicate that the southwestern block of the fault zone exhibits lower average seismic velocities than the northeastern block, with multiple small-scale shallow low-velocity zones likely developed within the uppermost crust. For structural delineation, when applying Criterion 1 （the abrupt-increase method） to all teleseismic events collectively, only the southwestern boundary （1.25 km） can be identified. By contrast, separate processing of south- and north-originated teleseismic events enables accurate determination of LVZ boundaries at about 1.25 km southwest and about 0.58 km northeast of the surface fault trace, yielding an LVZ width of approximately 1.8 km. Boundaries derived from Criterion 2 （the fixed-threshold method）are offset outward on both sides （1.35 km southwest and 1.19 km northeast）, leading to an overestimated width of 2.5 km. Numerical simulation results verify that Criterion 1 provides more accurate constraints on the true fault-zone boundaries and maintains good stability even for gradational fault interfaces with a 10%–30% velocity-reduction transition layer. Additionally, we tested 25 fault-zone models with varying depths （1–5 km） and velocity reductions （20%–60%） to constrain fault-zone depth and velocity anomalies. Our results demonstrate that models with （3 km depth, 20% velocity reduction）, （2 km depth, 30% velocity reduction）, and （1 km depth, 50% velocity reduction） all fit the observational data well, revealing a trade-off effect of travel-time information in resolving fault-zone depth and velocity anomalies.

This study confirms that the two LVZ boundaries of the Qujiang fault zone lie approximately 1.25 km southwest and 0.58 km northeast of the surface fault trace, corresponding to an overall LVZ width of about 1.8 km. For practical applications of teleseismic travel-time delays in fault-zone detection, we recommend determining fault boundaries separately using abrupt travel-time-increase locations from seismic events on either side of the fault zone. This approach avoids one-sided masking effects induced by uneven source distribution and width overestimation associated with the fixed-threshold method. Given the pronounced trade-off between fault-zone depth and velocity reduction, future work should integrate teleseismic full-waveform information to enable quantitative high-resolution characterization of fine fault-zone structures.