Phenomenological Theory of Aftershocks

• Phenomenological aftershock theory is a framework that characterizes post-mainshock seismicity as a staged relaxation process, integrating Omori’s law and modern source deactivation dynamics.
• It employs analytical models such as logistic equations, KPP dynamics, and proper-time constructs to describe temporal decay and spatiotemporal migration in seismic events.
• The theory reconciles empirical observations with mechanical insights by combining statistical, diffusion-based, and energy-constrained approaches to aftershock modeling.

Searching arXiv for recent and foundational papers on phenomenological aftershock theory. arxiv_search(query="aftershock phenomenology Omori stretched exponential deactivation coefficient proper time fluctuating diffusivity", max_results=10) arxiv_search(query="all:aftershock phenomenology Omori stretched exponential deactivation coefficient proper time fluctuating diffusivity", max_results=10) arxiv_search(query="id:(Mignan, 2015) OR id:(Guglielmi et al., 2020) OR id:(Mathey et al., 2024) OR id:(Abe et al., 2023)", max_results=10) Phenomenological theory of aftershocks is the body of empirical and semi-empirical descriptions that treats post-mainshock seismicity as the observable relaxation of an earthquake source or of a coupled fault system. Its historical core is Omori’s hyperbolic decay law, but modern formulations extend far beyond a single rate curve: they include source-deactivation equations, logistic and Kolmogorov–Petrovsky–Piskunov (KPP) dynamics, stretched-exponential relaxation, deformation-controlled scaling, afterslip-driven triggering, complexity-based aging and memory, and source-centered proper-time constructions (Mignan, 2015, Guglielmi et al., 2020, Guglielmi, 2023).

1. Classical empirical laws and their status

The classical starting point is Omori’s law,

$n(t)=\frac{k}{c+t},$

where $n(t)$ is the aftershock frequency at elapsed time $t$, and the later Hirano–Utsu generalization,

$n(t)=\frac{k}{(c+t)^p}.$

In the older Omori notation, the same phenomenology also appears as $n(t)\propto (t+t_{\min})^{-1}$, while modern model-comparison work often uses the simplified power-law form $n(t)\propto t^{-\alpha}$ for $t\ge t_{\min}$ (Mignan, 2015). The power-law description has long been treated as one of the foundational empirical regularities of seismology, second only to Gutenberg–Richter, and it underlies standard stochastic constructions such as ETAS (Mignan, 2015).

Within the phenomenological literature, however, this classical picture is not taken as settled. Two criticisms recur. First, the short-time offset $c$ or $t_{\min}$ is described as ad hoc or “unphysical.” Second, for $\alpha\le 1$, the power-law tail must be truncated at some $n(t)$0 if the total number of events is to remain finite (Mignan, 2015). A related critique is that the Hirano–Utsu law predicts

$n(t)$1

whereas real sequences should asymptote to nonzero background seismicity in source-centered formulations (Guglielmi, 2023). A further source-centered objection is that, when $n(t)$2, the parameter $n(t)$3 loses a fixed physical dimension, so the law functions primarily as a fitting formula rather than as a direct law of source evolution (Guglielmi, 2023).

This tension between empirical adequacy and physical interpretation defines the modern phenomenological field. Omori and Hirano–Utsu remain indispensable descriptive laws, but many recent works treat them as special regimes, fitting surrogates, or projections of deeper source-state dynamics rather than as final statements of aftershock physics (Guglielmi et al., 2024, Guglielmi et al., 2024).

2. Source deactivation, proper time, and staged relaxation

A major modern line of work reformulates aftershock decay as the evolution of a nonstationary source characterized by a deactivation coefficient. Its minimal equation is

$n(t)$4

where $n(t)$5 is aftershock frequency and $n(t)$6 is the source deactivation coefficient (Guglielmi et al., 2020, Guglielmi, 2021). For constant $n(t)$7, the solution is

$n(t)$8

which is exactly the hyperbolic Omori law up to reparameterization (Guglielmi, 2021). The same framework generalizes to

$n(t)$9

so departures from strict Omori behavior are interpreted as time dependence of the source state rather than immediate failure of the underlying dynamical picture (Guglielmi et al., 2020).

The same literature introduces the logistic extension

$t$0

with equilibrium background rate $t$1 (Guglielmi et al., 2020, Guglielmi, 2023). In this view, Omori decay is the early-time asymptotic regime selected by $t$2, whereas late-time behavior is controlled by approach to background seismicity rather than by indefinite decay toward zero (Guglielmi, 2021). The practical inverse problem is posed through

$t$3

with smoothing used to regularize noisy aftershock-rate estimates (Guglielmi, 2023).

This source-centered program defines the Omori epoch as the finite interval during which $t$4, so that Omori’s law holds exactly (Guglielmi, 2023). Several papers then interpret abrupt changes in $t$5 or in $t$6 as source bifurcations (Guglielmi, 2021, Guglielmi et al., 2024). Later developments replace ordinary time by the source’s proper time,

$t$7

and, in discrete form, by an event-count “underground clock” synchronized with world time through $t$8 (Guglielmi, 2 Oct 2025, Zotov et al., 6 Aug 2025). In the Tohoku case, this approach yields a three-phase relaxation picture: an initial phase with $t$9, a main phase with constant $n(t)=\frac{k}{(c+t)^p}.$0, and a recovery phase with irregular $n(t)=\frac{k}{(c+t)^p}.$1 (Guglielmi, 2 Oct 2025). The same Tohoku analysis reports that the first 800 aftershocks obey a linear synchronization law $n(t)=\frac{k}{(c+t)^p}.$2, while the later main phase is approximated by

$n(t)=\frac{k}{(c+t)^p}.$3

which reproduces the same constant deactivation coefficient in discrete proper time (Zotov et al., 6 Aug 2025).

A central implication of this school is that aftershock evolution is not a single smooth decay process. It is a staged relaxation of the source, and the empirical Omori law is exact only on the interval where the deactivation coefficient remains constant (Guglielmi, 2024).

3. Spatiotemporal migration and transport formulations

Temporal decay is only one part of the phenomenology. A second tradition models aftershocks as a spatiotemporal process with slow migration away from the main-shock region. The standard source equation used for this purpose is the KPP-type reaction–diffusion law

$n(t)=\frac{k}{(c+t)^p}.$4

where $n(t)=\frac{k}{(c+t)^p}.$5 is a diffusion coefficient (Guglielmi et al., 2020, Guglielmi et al., 2024). In the weak-diffusion limit, this reduces to the logistic or deactivation ODE. In the spatially extended case, it admits traveling-wave solutions $n(t)=\frac{k}{(c+t)^p}.$6, with front speed estimated as

$n(t)=\frac{k}{(c+t)^p}.$7

and in one exact example

$n(t)=\frac{k}{(c+t)^p}.$8

This framework was proposed to account simultaneously for Omori-law temporal decay and the observed slow outward spread of aftershock epicenters at speeds of several km/h (Guglielmi et al., 2020).

A more explicitly source-transport interpretation appears in work that combines nonlinear evolution equations with the Umov-Poynting idea of directed energy flux. There the epicentral zone of the main shock is treated as a “track detector,” and foreshocks and aftershocks are treated as marks left by a propagating energetic factor in a stressed-strained rock mass (Guglielmi et al., 2024). Using 138 main shocks with $n(t)=\frac{k}{(c+t)^p}.$9, foreshock, main-shock, and aftershock hypocentral depths all $n(t)\propto (t+t_{\min})^{-1}$0 km, and $n(t)\propto (t+t_{\min})^{-1}$1-day windows, that study analyzed 1960 foreshocks and 34,627 aftershocks (Guglielmi et al., 2024). In stacked event-order space, the mean foreshock radius obeys

$n(t)\propto (t+t_{\min})^{-1}$2

with correlation coefficient $n(t)\propto (t+t_{\min})^{-1}$3, while the mean aftershock radius obeys

$n(t)\propto (t+t_{\min})^{-1}$4

with correlation coefficient $n(t)\propto (t+t_{\min})^{-1}$5 (Guglielmi et al., 2024). The phenomenological interpretation is convergence of foreshocks toward the future rupture nucleus and divergence of aftershocks away from it. The aftershock intercept of about $n(t)\propto (t+t_{\min})^{-1}$6–$n(t)\propto (t+t_{\min})^{-1}$7 km is further linked to rupture-scale unloading through the empirical fault-length scaling

$n(t)\propto (t+t_{\min})^{-1}$8

(Guglielmi et al., 2024).

A more radical reinterpretation of migration appears in the claim that aftershock memory is not fundamentally in clock time but in accumulated deformation. There the central rescaling is

$n(t)\propto (t+t_{\min})^{-1}$9

and the correlation function is found to satisfy

$n(t)\propto t^{-\alpha}$0

while the same data collapse onto

$n(t)\propto t^{-\alpha}$1

when plotted against deformation lag (Mathey et al., 2024). In that framework, the characteristic timescale is entirely fixed by $n(t)\propto t^{-\alpha}$2, and aftershocks are described as a deformation-controlled, quasi-static phenomenon rather than as a relaxation process with an intrinsic clock (Mathey et al., 2024).

4. Competing temporal kernels, statistical discrimination, and memory

One of the sharpest empirical challenges to the Omori paradigm is the proposal that aftershock decay is better described by the exponential family, especially the stretched exponential

$n(t)\propto t^{-\alpha}$3

rather than by $n(t)\propto t^{-\alpha}$4 (Mignan, 2015). In a large comparison across Southern California, Northern California, and Taiwan catalogs, with nearest-neighbor, Reasenberg second-order moment, and Gardner–Knopoff window declustering, only sequences with more than 100 events above a completeness threshold were retained, and $n(t)\propto t^{-\alpha}$5 and model-specific $n(t)\propto t^{-\alpha}$6 were estimated by maximum likelihood and Kolmogorov–Smirnov selection (Mignan, 2015). Model selection used

$n(t)\propto t^{-\alpha}$7

Across all 245 tested sequences, no sequence was ultimately best fit by a power law once the comparison was performed on equal footing; the exponential family was always preferred, usually as stretched exponential and occasionally as pure exponential (Mignan, 2015). This is one of the strongest statistical arguments for replacing canonical Omori decay by a phenomenological relaxation law with a characteristic rate scale.

A different route to Omori–Utsu behavior comes from hierarchical fluctuating-diffusivity theory. There the load state $n(t)\propto t^{-\alpha}$8 diffuses with a diffusivity $n(t)\propto t^{-\alpha}$9 treated as a slow random variable. At fixed $t\ge t_{\min}$0, each Fourier mode relaxes exponentially, but averaging over a gamma distribution,

$t\ge t_{\min}$1

yields

$t\ge t_{\min}$2

which is an exact Omori–Utsu-type relaxation law (Abe et al., 2023). In the reduced description, the Green’s function depends on $t\ge t_{\min}$3, or equivalently on $t\ge t_{\min}$4, so the subsystem has temporal translational invariance in logarithmic time rather than in ordinary time (Abe et al., 2023). Aging follows directly from this logarithmic-time structure.

Complexity-based analyses push the memory claim further. In California aftershock sequences, the event–event correlation function shows aging collapse,

$t\ge t_{\min}$5

with $t\ge t_{\min}$6 and $t\ge t_{\min}$7 for the Landers aftershocks (Abe et al., 2012). The same study tests the scaling relation

$t\ge t_{\min}$8

required for a class of singular Markovian processes and finds strong violation: for Landers, $t\ge t_{\min}$9, $c$0, and reported sums are about $c$1–$c$2 (Abe et al., 2012). The phenomenological conclusion is non-Markovianity and long-range memory. This directly contradicts any view of aftershocks as a simple memoryless renewal process.

Taken together, these studies show that phenomenological aftershock theory is not divided only by preferred fitting curves. It is also divided by competing views of what the decay law encodes: scale-free rate decay, stretched relaxation, hierarchical heterogeneity, or explicitly non-Markovian memory (Mignan, 2015, Abe et al., 2023, Abe et al., 2012).

5. Mechanistic phenomenologies: afterslip, rheological transition, and causal proxies

Several phenomenological models try to connect statistical aftershock laws to simplified but interpretable source mechanics. One influential route links aftershocks to afterslip in a coupled creeping region. In a two-block model with a seismogenic block $c$3 elastically coupled to a velocity-strengthening block $c$4, the postseismic slip of $c$5 is logarithmic,

$c$6

so that

$c$7

With heterogeneous Coulomb thresholds, the aftershock rate becomes

$c$8

that is, proportional to afterslip rate up to a hazard-function correction (Lippiello et al., 2018). In this model, Omori decay with $c$9 follows from logarithmic afterslip, and a realistic Gutenberg–Richter exponent near $t_{\min}$0 is recovered when Coulomb stress thresholds are power-law distributed (Lippiello et al., 2018).

A spatially extended version of the same idea places the brittle seismogenic layer above a velocity-strengthening viscoelastic layer across the brittle–ductile transition (Petrillo et al., 2020). The upper layer is velocity weakening and heterogeneous; the lower layer relaxes by afterslip and reloads the brittle fault slowly. In large synthetic catalogs, this model reproduces several empirical regularities quantitatively: $t_{\min}$1, moment–area exponent $t_{\min}$2, Omori-Utsu exponent $t_{\min}$3, and productivity $t_{\min}$4 (Petrillo et al., 2020). It also yields a flatter foreshock magnitude distribution than the aftershock one, with $t_{\min}$5 and $t_{\min}$6 (Petrillo et al., 2020). The intended implication is that the rheological transition from a brittle, velocity-weakening fault to a ductile, velocity-strengthening substrate can itself generate the main phenomenological laws of aftershocks and foreshocks.

A third mechanistic phenomenology comes from analog dynamics rather than geophysical constitutive modeling. In a multi-component lattice Boltzmann model of a sheared soft-glassy material, direct interventions allow parent events to be suppressed and causal descendants to be identified (Kumar et al., 2022). More than 1300 event–aftershock pairs were recovered, and the directly causal catalog obeys

$t_{\min}$7

with

$t_{\min}$8

(Kumar et al., 2022). The same model distinguishes two physical triggering classes: contact-triggered aftershocks with

$t_{\min}$9

and shear-wave-triggered aftershocks with

$\alpha\le 1$0

(Kumar et al., 2022). About 80% of first aftershocks lie within the shear-wave light cone, all lie within the sound-wave light cone, and the nearest-neighbor metric yields distinct peaks for triggered and independent pairs (Kumar et al., 2022). This does not prove that crustal aftershocks have the same microphysics, but it does show that Omori-like decay and standard declustering diagnostics can emerge from explicitly causal triggering in a controlled proxy system.

These models share a common feature: they do not derive aftershocks from one universal microscopic law. Instead, they translate slow post-mainshock loading, rheological contrast, and heterogeneous thresholds into discrete triggered sequences whose aggregate statistics resemble canonical phenomenology (Lippiello et al., 2018, Petrillo et al., 2020, Kumar et al., 2022).

6. Triggers, modulation, conservation constraints, and continuing controversies

Phenomenological theory also includes proposals in which aftershock sequences are modulated by triggers generated by the Earth system itself. One such proposal treats the crust after a giant main shock as a near-critical medium subject to exogenous forcing from the main shock’s own wavefield (Guglielmi et al., 2013). For the 2004 Sumatra–Andaman earthquake, the strongest aftershock, $\alpha\le 1$1, occurred about 3 h 20 min after the $\alpha\le 1$2 main shock. Using surface-wave speed $\alpha\le 1$3, the round-the-world return time is estimated as

$\alpha\le 1$4

and the aftershock is interpreted as likely triggered by a round-the-world seismic echo (Guglielmi et al., 2013). The same work argues for modulation of aftershocks by the Earth’s free spheroidal oscillation $\alpha\le 1$5, with frequency $\alpha\le 1$6 and period 54 min; the Sumatra aftershock spectrum shows enhancement in the 0.28–0.30 mHz band, the Tohoku spectrum has a maximum near 0.285 mHz, and global spectra from 1973–2010 show peaks near 0.307 and 0.309 mHz (Guglielmi et al., 2013). These claims are explicitly phenomenological and suggestive rather than definitive.

Another controversy concerns the formal consistency of indefinite Omori decay with energy conservation. If each aftershock consumes a finite portion of source energy and

$\alpha\le 1$7

is extrapolated indefinitely in a closed source, then the cumulative number

$\alpha\le 1$8

diverges logarithmically as $\alpha\le 1$9, implying infinite released energy (Guglielmi et al., 2024). Two phenomenological resolutions are proposed: first, Omori behavior is valid only during a finite Omori epoch; second, the source should be modeled as an open system, leading to the logistic equation

$n(t)$00

where the $n(t)$01 term represents effective energy inflow from the environment (Guglielmi et al., 2024). In this view, the paradox is not observational but conceptual: it is a signal that the hyperbolic law cannot be a literal all-time law of a closed relaxing source.

The contemporary field therefore remains plural. One line replaces Omori power laws by stretched exponentials on statistical grounds (Mignan, 2015). Another retains Omori but restricts it to a finite constant-deactivation epoch (Guglielmi, 2023). Another treats apparent time memory as deformation memory with timescale $n(t)$02 (Mathey et al., 2024). Others derive Omori-like decay from afterslip, from brittle–ductile coupling, or from hierarchical fluctuating diffusivity (Lippiello et al., 2018, Petrillo et al., 2020, Abe et al., 2023). Complexity analyses add aging and long-range memory, while trigger-based studies emphasize global echoes and normal-mode modulation (Abe et al., 2012, Guglielmi et al., 2013).

The resulting encyclopedic picture is not one of convergence to a single law, but of a broadening phenomenological program. Aftershocks are modeled variously as scale-free decay, stretched relaxation, staged source deactivation, migration in a diffusion-wave medium, deformation-controlled clustering, heterogeneous relaxation under fluctuating diffusivity, or delayed triggering by coupled creeping or ductile regions. What these approaches share is the claim that aftershock sequences are not exhausted by a single empirical curve: they are signatures of post-mainshock source evolution, and the phenomenological theory of aftershocks is the effort to infer that evolution from the observable rate, timing, and migration of seismicity.