Two-Phase Decay Models

• Two-Phase Decay Models are mathematical frameworks that define dynamics with an initial rapid decay followed by a slower, enduring tail, marked by a clear switching point.
• They utilize sum-form and piecewise formulations, combining exponential and power-law behaviors to model diverse phenomena from hydrodynamics to quantum decay.
• Parameter estimation often involves nonlinear least squares and information criteria like AIC and R², demonstrating superior fit compared to single-phase alternative models.

A two-phase decay model describes a class of dynamical processes in which the decay of an observable—such as a correlation, probability amplitude, memory signal, or physical quantity—proceeds via two distinct regimes characterized by separate functional forms, typically an initial fast decay followed by a slower, longer-lasting tail. Such models are prevalent in the mathematical description of phenomena across statistical physics, hydrodynamics, quantum mechanics, information diffusion, collective memory, and relaxation theory. The essential defining property is the existence of two temporally separated dynamical phases, each governed by a distinct decay law, with a crossover (or "switching point") at which the slower process overtakes the fast initial decay.

1. Mathematical Framework and Model Classes

Two-phase decay models admit diverse mathematical realizations but universally incorporate two decay components, typically additive or piecewise-defined. The most canonical examples are:

• Sum-form models: The observable $S(t)$ is modeled as a sum of an exponential and a power law:

$$S(t) = C_1 e^{-\beta t} + C_2 t^{-\alpha},\quad t \ge 0,$$

where $C_1$, $C_2$ are amplitudes, $\beta$ is the exponential decay rate, and $\alpha$ is the power-law exponent (Igarashi et al., 2022).

• Piecewise models: The system exhibits initially a pure exponential decay up to a switching time $t_s$, beyond which power-law decay dominates:

$$S(t) = \begin{cases} C_1 e^{-\beta t} & t \le t_s \ C_2 t^{-\alpha} & t > t_s \end{cases}$$

The switching time $t_s$ is typically defined as the time at which the power law overtakes the exponential, or where the two terms become equal in magnitude.

The formalism is also extended to more general scenarios—such as systems exhibiting two exponentials, models with both algebraic and exponential spatial decay (e.g. in Kosterlitz–Thouless or coupled XY models), or cases with regularity-loss leading to two separate frequency-dependent dissipation mechanisms (Ueda et al., 2014, Bighin et al., 2019).

2. Key Dynamical Mechanisms and Switch Criteria

The origin of the two-phase behavior varies by context:

• In collective memory and sociotechnical dynamics (Igarashi et al., 2022), the fast exponential reflects temporary bursts of interest, whereas the slow power-law reflects sustained, network-mediated attention.
• In fluid dynamics and hydrodynamic flows (Wang et al., 2020, Huang et al., 2023, Saito, 2021), the two phases arise from linear versus nonlinear mechanisms such as low-frequency algebraic decay (diffusive or heat-like) and high-frequency exponential damping due to viscosity, surface tension, or drag.
• In quantum decay (Peshkin et al., 2017, Golubkova et al., 2016), early-time survival probability is quadratic (Zeno regime), which gives way to an intermediate exponential governed by resonant pole(s), and finally crosses over at late times to a power-law tail due to the continuum spectrum.

A universal feature is the dynamical switching point, which may be computed, for the sum-of-terms model, from

$$C_1 e^{-\beta t_s} = C_2 t_s^{-\alpha} \implies t_s = \frac{1}{\beta} \ln\left(\frac{C_1}{C_2} t_s^{\alpha}\right),$$

or, algorithmically, by searching for the first $t$ at which the dominance between the two terms in $S(t)$ changes (Igarashi et al., 2022).

3. Parameter Estimation and Fitting Procedures

For empirical two-phase decay models, parameter inference proceeds via (i) normalization and transformation (such as log-scaling to precondition the fitting surface), (ii) selection of appropriate functionals (sum-of-squares residuals in log–log space), and (iii) optimization algorithms (nonlinear least squares) to extract the amplitudes and decay rates.

Typical model selection involves the Akaike Information Criterion (AIC) alongside the coefficient of determination $R^2$. Comparative tables demonstrate that two-phase models frequently offer superior $R^2$ and lower AIC than single-phase alternatives such as bi-exponential or shifted power-law (Igarashi et al., 2022).

Category	$R^2$ (proposed)	AIC (proposed)
Earthquake	0.786	–311.1
Deaths	0.733	–245.4
Aviation	0.846	–436.9
Mass murder	0.798	–280.9
Terror	0.757	–262.1

Empirically inferred median values have shown universality in the fast exponential rate ($\beta \approx 0.4$), slow tail exponent ($\alpha$ typically $0.2$–$0.5$), and switching times clustering at $t_s \approx 10$–11 days in large-scale collective memory data (Igarashi et al., 2022).

4. Physical and Mathematical Interpretations Across Domains

• Hydrodynamics and Relaxation: In two-phase fluid models, decay rates elucidate spectral separation into slowly-decaying (diffusive, heat-like) low-frequency modes and rapidly damped high-frequency components. The regularity-loss structure, for instance, yields a dissipation rate proportional to $|\xi|^{2p}/(1+|\xi|^2)^q$ in Fourier space, leading to algebraic decay for low $|\xi|$ and exponential suppression for high $|\xi|$—the prototype of two-phase time decay (Ueda et al., 2014, Wang et al., 2020, Saito, 2021, Huang et al., 2023).
• Quantum Mechanics: The survival probability of an unstable state in quantum theory exhibits: (i) quadratic decay at short times (Zeno), (ii) exponential decay (dominant resonance pole), (iii) power-law tail (branch point at threshold), and, for systems with several poles, possible multiple-exponential (two-phase) regimes with crossover times set by the inverse difference in decay rates of the resonances (Peshkin et al., 2017). In two-proton radioactivity, the competition between direct three-body emission and sequential decay via an intermediate resonance yields a dynamical transition quantified by $(E_T/E_r-1)/(\Gamma_r/E_r)$, sharply reflected in experimental energy-sharing correlations (Golubkova et al., 2016).
• Statistical Physics and Complex Systems: In coupled XY models, the "BKT-paired" phase demonstrates exponential decay of single-particle correlations and slow (power-law) decay of pair correlations over an intermediate temperature window, with switching governed by the separation of vortex proliferation in symmetric and antisymmetric phase channels (Bighin et al., 2019).

5. Generalizations and Theoretical Scope

The two-phase paradigm is not restricted to specific disciplines but captures a phenomenological motif: competitive or sequential dominance of two distinct decay mechanisms. Mathematical generalization includes models where the functional forms of decay are not limited to exponential or power-law but can involve arbitrary functional transitions provided a dynamical switching criterion is definable.

Key features enabling two-phase decay include:

• The coexistence of a transient-dominated channel (e.g., "burst-and-forget" agents, hydrodynamic high frequencies, or dominant quantum resonance) and a slower, often network- or interaction-moderated channel (e.g., long-term retention in networks, diffusive tails in hydrodynamics, quantum continuum branch cut).
• The presence of spectral or mode decomposition, resulting in a splitting of time decay rates between components associated with distinct spatial, physical, or dynamical subspaces (Wang et al., 2020, Ueda et al., 2014).
• The applicability of similar parameter estimation, rescaling, and switching-point detection methodologies to a wide range of post-peak activity time series, regardless of domain (Igarashi et al., 2022).

6. Illustrative Applications and Universal Patterns

The two-phase decay framework has found quantitative and predictive success in modeling:

• Online collective memory and digital attention dynamics (Igarashi et al., 2022): Wikipedia pageviews for significant events quantitatively fit by two-phase models with universal parameters across categories.
• Fluid flows and two-phase hydrodynamics (Huang et al., 2023, Wang et al., 2020, Saito, 2021): Time-decay of solutions decomposed into algebraic and exponential regimes; optimal $L^p$–$L^q$ time-decay rates match heat equation exponents under Hodge- or spectral decompositions.
• Quantum decay and nuclear fragmentation (Peshkin et al., 2017, Golubkova et al., 2016): Survival amplitude crossovers between nonexponential and exponential phases, and observable transition from direct to sequential decay in three-body nuclear disintegration.
• Statistical mechanics of coupled systems (Bighin et al., 2019): Phase diagrams with windows of exponential vs. algebraic decay in composite observables, emerging from symmetry-breaking or mode-coupling.

The emergence of universal parameter bands (e.g., $\beta \approx 0.4$, $\alpha \approx 0.3$, $t_s \approx 10$ days in collective memory) across event types suggests underlying genericity in two-phase decay processes (Igarashi et al., 2022).

7. Analytical Methods and Theoretical Import

Analysis of two-phase decay models leverages:

• Laplace/Fourier transform and contour-integration techniques to separate low- and high-frequency modes, predicting algebraic and exponential time scales (Saito, 2021, Ueda et al., 2014).
• Interpolation between parabolic, hyperbolic, and network-mediated regimes via spectral expansions and interactive energy functionals (Wang et al., 2020).
• Nonlinear least-squares fitting, spectral bootstrapping, and information-theoretic model selection to infer empirical decay law parameters (Igarashi et al., 2022).
• Renormalization group and dual-channel analysis of effective actions to identify separated unbinding transitions in statistical mechanical models (Bighin et al., 2019).

These models unify diverse fast–slow and regularity-loss behaviors under a rigorous framework and provide diagnostic tools for both theoretical and empirical discrimination between competing decay mechanisms. Their transferability across fields underlines the structural significance of two-phase decay in the dynamics of complex systems.