Power-Law+Peak Model Overview

• Power-Law+Peak Model is a framework that combines power-law scaling with a localized peak to capture finite-scale deviations from pure scale invariance.
• It employs diverse mathematical realizations—from broken power laws to exponential rollovers—to accurately reflect empirical data across materials, cosmology, and liquids.
• Its cross-disciplinary applications, ranging from glass vibrational spectra to gravitational-wave sensitivity, highlight its utility in modeling complex physical phenomena.

Searching arXiv for recent and relevant papers on “Power-Law+Peak” formulations and closely related usages across fields. “Power-Law+Peak Model” denotes a family of constructions in which a power-law component is combined with a localized peak, rollover, cutoff, or slowly varying correction in order to describe data or effective physics that are not adequately represented by a single scale-free law. Across the literature, the phrase is used both explicitly and implicitly. In some settings it refers to an intrinsic spectrum with a low-frequency power-law tail and a finite-frequency peak, as in the nonphononic vibrational density of states of glasses (Moriel et al., 2023). In others it refers to a broken power law with a peak scale, a power law with an upper rollover, a power-law family whose envelope has a unique sensitivity minimum, or an effective power law plus smooth correction over a narrow physically sampled interval (Li et al., 2024). The common structure is not a single universal formula, but a shared modeling strategy: retain power-law asymptotics where scale-free behavior is observed or expected, while introducing a peak-like feature to capture finite-scale organization, finite-size effects, detector response, or local effective physics.

1. Conceptual definition and recurrent mathematical forms

The defining characteristic of a Power-Law+Peak formulation is the coexistence of a scale-free sector and a distinguished scale. The power-law sector is used to represent asymptotic behavior, tail statistics, or effective scaling. The peak sector may take the form of a finite-frequency maximum, a Gaussian bump, a rollover, a broken-slope turnover, or a slowly varying additive term that makes a local power-law approximation viable over the relevant interval.

Several distinct mathematical realizations appear in the literature. In glasses, the nonphononic vibrational density of states is represented as an intrinsic spectrum ${\cal D}_{\rm G}(\omega)$ with a universal low-frequency tail

${\cal D}_{\rm G}(\omega)\sim A_{\rm g}\,\omega^4$

and a peak at finite $\omega_{\rm p}$ (Moriel et al., 2023). In scalar-induced gravitational-wave studies, the primordial curvature spectrum is modeled by a broken power law with a peak at $k_\ast$,

$$\mathcal{P}_{\mathcal R}(k) = A\,\frac{\alpha+\beta}{\beta\left(k/k_\ast\right)^{-\alpha}+\alpha\left(k/k_\ast\right)^{\beta}},$$

which yields distinct infrared and ultraviolet power-law tails around a single peak scale (Li et al., 2024). In solar-flare statistics, the relevant form is a power law with an upper rollover,

$$P_{\rm plr}(S)=B\,S^{-\gamma_{\rm plr}}e^{-S/\sigma},$$

used to model a downward bend at large event size (Wheatland, 2010).

Other works instantiate the same logic without using the label explicitly. For dense viscous Buckingham liquids, the repulsive core is approximated over the first-shell range by an extended inverse power law,

$\upsilon_{eIPL}(r) = Ar^{-n} + B + Cr,$

that is, an effective power law plus a slowly varying linear correction (Veldhorst et al., 2011). In self-similar BAO modeling, a Gaussian bump is imposed on a power-law correlation function,

$$\xi_{\rm IC}(r)=\left(\frac{r_0}{r}\right)^{n+3} \left[1 + A_{\rm bump}\, e^{-(r-r_{\rm bao})^2/2\sigma_{\rm bao}^2}\right],$$

producing a deliberately simplified power-law-plus-peak toy model for nonlinear peak evolution (Orban et al., 2011).

This variety suggests that “Power-Law+Peak Model” is best understood as a structural class rather than a unique formalism. A plausible implication is that the term functions as a cross-disciplinary shorthand for models in which asymptotic scaling and finite-scale organization must be represented simultaneously.

2. Vibrational spectra of glasses: intrinsic power-law tail plus peak

The clearest explicit use of the Power-Law+Peak picture appears in the theory of the boson peak in glasses. The central object is the nonphononic vibrational density of states,

$${\cal D}_{\rm G}(\omega) \equiv {\cal D}(\omega) - A_{\rm D}\,\omega^2,$$

obtained by subtracting the Debye contribution from the total vibrational density of states. In this formulation, ${\cal D}_{\rm G}(\omega)$ itself features both a universal low-frequency power-law tail and an intrinsic peak, and the conventional boson peak in ${\cal D}_{\rm G}(\omega)\sim A_{\rm g}\,\omega^4$0 is a ratio-based manifestation of that underlying structure (Moriel et al., 2023).

The low-frequency sector follows

${\cal D}_{\rm G}(\omega)\sim A_{\rm g}\,\omega^4$1

while the peak occurs at a finite frequency ${\cal D}_{\rm G}(\omega)\sim A_{\rm g}\,\omega^4$2. The coefficient ${\cal D}_{\rm G}(\omega)\sim A_{\rm g}\,\omega^4$3 depends strongly on disorder or thermal history, whereas the peak frequency and peak magnitude vary only mildly under annealing. In the reanalysis of Raman data on ${\cal D}_{\rm G}(\omega)\sim A_{\rm g}\,\omega^4$4 glasses, the peak frequency shifts upward by about 20% and the peak height by about 11% from the as-quenched to the most annealed sample, while the low-frequency tail drops by about an order of magnitude over the same range (Moriel et al., 2023). Large-scale atomistic simulations reproduce the same qualitative separation: strong suppression of the ${\cal D}_{\rm G}(\omega)\sim A_{\rm g}\,\omega^4$5 tail with slower quenching, but only modest changes in the peak.

The mean-field explanation is formulated in terms of interacting quasi-localized excitations with Hamiltonian

${\cal D}_{\rm G}(\omega)\sim A_{\rm g}\,\omega^4$6

Within this framework, the peak position and height scale mainly with the internal-stress parameter ${\cal D}_{\rm G}(\omega)\sim A_{\rm g}\,\omega^4$7, while the tail amplitude is mainly controlled by the interaction scale ${\cal D}_{\rm G}(\omega)\sim A_{\rm g}\,\omega^4$8 in the small-interaction regime (Moriel et al., 2023). The same quasi-localized nonphononic vibrations generate both the ${\cal D}_{\rm G}(\omega)\sim A_{\rm g}\,\omega^4$9 tail and the peak; modes near the peak involve many coupled quasi-localized excitations rather than a separate species of excitation.

A related but distinct route to a peak-plus-power-law structure is provided by heterogeneous elasticity theory with long-ranged power-law correlated disorder. There the disorder correlator decays algebraically, the phonon self-energy acquires a logarithmic factor, and the resulting density of states exhibits a boson peak in the reduced DOS together with a low-frequency $\omega_{\rm p}$0 regime in 3D and a logarithmic correction of the form $\omega_{\rm p}$1 around the boson-peak region (Cui et al., 2020). The commonality with the nonphononic-VDoS picture is structural: both combine low-frequency power-law scaling with a finite-frequency excess peak, but the microscopic interpretation differs.

3. Effective interactions and hidden scale invariance in dense liquids

In dense-liquid theory, the Power-Law+Peak idea appears as an effective description of the pair potential over the narrow distance interval sampled by the first coordination shell. For the modified Buckingham liquid, the microscopic interaction has an exponential repulsive term and an attractive $\omega_{\rm p}$2 term,

$\omega_{\rm p}$3

yet the physically sampled repulsive region is well approximated by the extended IPL form

$\omega_{\rm p}$4

near the first peak of the radial distribution function (Veldhorst et al., 2011).

The significance of this approximation is tied to strong virial–potential energy correlations,

$\omega_{\rm p}$5

with strongly correlating liquids defined by $\omega_{\rm p}$6, and to the slope parameter

$\omega_{\rm p}$7

For strongly correlating liquids, $\omega_{\rm p}$8, and for a pure IPL potential one has $\omega_{\rm p}$9 exactly (Veldhorst et al., 2011). The paper argues that the IPL part governs scaling and near-perfect $k_\ast$0 correlation, whereas the linear term contributes little to fluctuations at constant volume because its contributions nearly cancel when intermolecular distances rearrange.

This yields an effective Power-Law+Peak-like description in the sense that the dominant repulsive contribution is represented locally by a power law, while the remaining smooth part over the first-neighbor range is approximately linear. The structural and dynamical consequences are tested along isomorphs in a Kob-Andersen binary Buckingham mixture. The reduced-coordinate radial distribution function $k_\ast$1 is nearly invariant along an isomorph, especially for AA pairs; the incoherent and coherent intermediate scattering functions collapse well in reduced units along isomorphs but differ strongly along isotherms (Veldhorst et al., 2011). The authors also construct a purely repulsive IPL model that mimics the viscous dynamics and temperature dependence of the intermediate scattering function of the Buckingham system.

This use of the Power-Law+Peak idea differs from explicit peak-in-spectrum models. The “peak” here is not a separate additive bump in an observable, but the physically narrow first-shell interval around the first peak of $k_\ast$2 within which a power law plus smooth correction becomes accurate. This suggests that in dense-liquid contexts the term can denote an effective local representation rather than a global parametric ansatz.

4. Cosmology and gravitational-wave theory: broken power laws, bumps, and sensitivity minima

Cosmological applications use Power-Law+Peak constructions in several technically distinct ways. One concerns primordial scalar spectra with a single characteristic scale. The broken power-law peak model

$k_\ast$3

has a peak amplitude $k_\ast$4, infrared scaling $k_\ast$5 for $k_\ast$6, and ultraviolet scaling $k_\ast$7 for $k_\ast$8 (Li et al., 2024). Its purpose is to provide an analytic near-peak approximation wide enough to cover many models. When inserted into the radiation-era SIGW convolution, the model yields analytic control over how both asymptotic power-law tails and the intermediate peak imprint features on the induced gravitational-wave spectrum. The paper identifies far-infrared regimes such as $k_\ast$9 when $$\mathcal{P}_{\mathcal R}(k) = A\,\frac{\alpha+\beta}{\beta\left(k/k_\ast\right)^{-\alpha}+\alpha\left(k/k_\ast\right)^{\beta}},$$0, the generic peaked-spectrum behavior $$\mathcal{P}_{\mathcal R}(k) = A\,\frac{\alpha+\beta}{\beta\left(k/k_\ast\right)^{-\alpha}+\alpha\left(k/k_\ast\right)^{\beta}},$$1, and ultraviolet decay typically governed by $$\mathcal{P}_{\mathcal R}(k) = A\,\frac{\alpha+\beta}{\beta\left(k/k_\ast\right)^{-\alpha}+\alpha\left(k/k_\ast\right)^{\beta}},$$2 (Li et al., 2024).

A second cosmological usage treats the peak as an explicit localized bump on a scale-free background. In self-similar BAO simulations, the baseline clustering model is $$\mathcal{P}_{\mathcal R}(k) = A\,\frac{\alpha+\beta}{\beta\left(k/k_\ast\right)^{-\alpha}+\alpha\left(k/k_\ast\right)^{\beta}},$$3, corresponding to

$$\mathcal{P}_{\mathcal R}(k) = A\,\frac{\alpha+\beta}{\beta\left(k/k_\ast\right)^{-\alpha}+\alpha\left(k/k_\ast\right)^{\beta}},$$4

and a Gaussian BAO-like bump is imposed in configuration space (Orban et al., 2011). Nonlinear evolution broadens and flattens the bump while approximately preserving its area. The bump width evolves diffusively according to

$$\mathcal{P}_{\mathcal R}(k) = A\,\frac{\alpha+\beta}{\beta\left(k/k_\ast\right)^{-\alpha}+\alpha\left(k/k_\ast\right)^{\beta}},$$5

and detectable peak shifts appear only for the steepest tested background, $$\mathcal{P}_{\mathcal R}(k) = A\,\frac{\alpha+\beta}{\beta\left(k/k_\ast\right)^{-\alpha}+\alpha\left(k/k_\ast\right)^{\beta}},$$6, following

$$\mathcal{P}_{\mathcal R}(k) = A\,\frac{\alpha+\beta}{\beta\left(k/k_\ast\right)^{-\alpha}+\alpha\left(k/k_\ast\right)^{\beta}},$$7

(Orban et al., 2011). Here the power law represents the broadband self-similar background, and the peak is a localized correlation excess.

A third usage arises in stochastic-background detector sensitivity. The Power-Law Integrated Sensitivity curve is the envelope of detectable power laws

$$\mathcal{P}_{\mathcal R}(k) = A\,\frac{\alpha+\beta}{\beta\left(k/k_\ast\right)^{-\alpha}+\alpha\left(k/k_\ast\right)^{\beta}},$$8

with fixed SNR threshold (Belgacem, 8 Mar 2025). The exact parametric envelope is shown to define a single-valued convex function $$\mathcal{P}_{\mathcal R}(k) = A\,\frac{\alpha+\beta}{\beta\left(k/k_\ast\right)^{-\alpha}+\alpha\left(k/k_\ast\right)^{\beta}},$$9 with a unique minimum at $$P_{\rm plr}(S)=B\,S^{-\gamma_{\rm plr}}e^{-S/\sigma},$$0. The peak sensitivity is

$$P_{\rm plr}(S)=B\,S^{-\gamma_{\rm plr}}e^{-S/\sigma},$$1

and the corresponding frequency is a weighted geometric mean of the observing band (Belgacem, 8 Mar 2025). In this setting the “peak” is not a feature of the source spectrum but of the detector network’s sensitivity envelope. That distinction is conceptually important: the same Power-Law+Peak language can refer either to a physical spectrum or to an observational threshold curve.

5. Statistical tail models, rollover models, and partial realizations

In statistical applications, Power-Law+Peak formulations often arise because a single power law fits only a tail, while the body or upper end requires a separate treatment. In distribution-feeder load forecasting, the daily peak-load variable $$P_{\rm plr}(S)=B\,S^{-\gamma_{\rm plr}}e^{-S/\sigma},$$2 is modeled only in the upper tail: $$P_{\rm plr}(S)=B\,S^{-\gamma_{\rm plr}}e^{-S/\sigma},$$3 with the lower-load region left unmodeled by the power law (Sangrody et al., 2017). The practical model is therefore piecewise: $$P_{\rm plr}(S)=B\,S^{-\gamma_{\rm plr}}e^{-S/\sigma},$$4 The threshold $$P_{\rm plr}(S)=B\,S^{-\gamma_{\rm plr}}e^{-S/\sigma},$$5 is selected by minimizing the KS distance, the exponent $$P_{\rm plr}(S)=B\,S^{-\gamma_{\rm plr}}e^{-S/\sigma},$$6 is estimated by MLE, an improved KS test with Monte Carlo simulation is used for goodness-of-fit, and bootstrap resampling provides 95% confidence intervals (Sangrody et al., 2017). The paper explicitly notes that it does not define a named Power-Law+Peak model; it provides only a tail-only power-law model for high daily peaks. This is a partial realization of the general idea rather than a full body-plus-tail composite distribution.

Solar-flare statistics provide a stronger example of a genuine departure from a single power law. For AR 11029, the background-subtracted GOES peak-flux distribution is modeled either by a simple power law or by a power law with an exponential rollover (Wheatland, 2010). Above the threshold $$P_{\rm plr}(S)=B\,S^{-\gamma_{\rm plr}}e^{-S/\sigma},$$7, Bayesian inference yields $$P_{\rm plr}(S)=B\,S^{-\gamma_{\rm plr}}e^{-S/\sigma},$$8 for the simple power law and

$$P_{\rm plr}(S)=B\,S^{-\gamma_{\rm plr}}e^{-S/\sigma},$$9

for the rollover model, with posterior odds ratio $\upsilon_{eIPL}(r) = Ar^{-n} + B + Cr,$0 assuming equal prior odds (Wheatland, 2010). The preferred interpretation is a finite magnetic free-energy reservoir, making arbitrarily large flares increasingly unlikely in a small active region.

A more heuristic central-peak-plus-tail model appears in finance. There empirical return distributions are described by a generalized finite-sum Maxwell-Boltzmann-type construction,

$\upsilon_{eIPL}(r) = Ar^{-n} + B + Cr,$1

which the paper interprets as producing a unimodal, symmetric central region together with asymptotic power-law tails (Tuncay, 2020). For relative returns, the reported tail exponent is around $\upsilon_{eIPL}(r) = Ar^{-n} + B + Cr,$2, while the model is presented as a heuristic data-collapsing framework rather than a formal likelihood-based body-plus-tail fit (Tuncay, 2020).

These examples show that the statistical meaning of “peak” varies. It can denote an unmodeled empirical body below a tail threshold, an upper rollover due to finite-size effects, or an explicitly modeled central mode with heavy tails.

6. Domain-specific meanings, limits, and common misconceptions

The main source of ambiguity in the term is that “peak” does not have a uniform interpretation across fields. In glasses, the peak is an intrinsic maximum of the nonphononic VDoS (Moriel et al., 2023). In dense liquids, it is tied to the first-shell distance range around the first peak of the radial distribution function, where an eIPL approximation becomes accurate (Veldhorst et al., 2011). In BAO work, it is a Gaussian correlation bump on top of a power-law background (Orban et al., 2011). In flare statistics, it is effectively a rollover or cutoff at the large-event end (Wheatland, 2010). In PLS analysis, it is the unique minimum of a sensitivity envelope rather than a source feature (Belgacem, 8 Mar 2025). In feeder-load forecasting, the model does not supply a peak component at all; it models only the upper tail (Sangrody et al., 2017).

A common misconception is to treat all such models as variants of a single composite density with a central hump and power-law tails. The literature surveyed here does not support that identification. Some papers do use a genuine intrinsic “tail plus peak” spectrum (Moriel et al., 2023), some use a broken power law with a peak scale (Li et al., 2024), some use a power law plus smooth additive correction (Veldhorst et al., 2011), and some explicitly state that they do not define a full Power-Law+Peak model (Sangrody et al., 2017). Another misconception is to assume that the peak sector always represents a distinct physical species. The glass literature argues the opposite: the boson peak is not a separate class of excitations, but the peak-like continuation of the same quasi-localized nonphononic family that generates the $\upsilon_{eIPL}(r) = Ar^{-n} + B + Cr,$3 tail (Moriel et al., 2023).

The principal methodological limit is therefore semantic as much as technical. “Power-Law+Peak Model” is most precise when accompanied by an explicit statement of which of the following is meant: a broken-slope peak, a localized bump on a power-law background, a rollover, an effective local correction to a power law, or an observational sensitivity envelope with a unique minimum. Without that clarification, the label is structurally informative but mathematically underdetermined.

7. Broader significance

Despite the heterogeneity of implementations, the recurrence of Power-Law+Peak constructions across condensed matter, liquid-state theory, cosmology, astrophysics, and statistical modeling points to a common modeling problem: pure scale invariance is often approximately valid only outside a finite region where local structure, finite resources, detector weighting, or interaction-induced reconstruction becomes decisive. The Power-Law+Peak strategy preserves the analytic and interpretive advantages of power-law scaling while introducing the minimum additional structure needed to represent that finite-scale feature.

In dense liquids, this supports hidden scale invariance even when the microscopic repulsion is not an inverse power law (Veldhorst et al., 2011). In glasses, it unifies the $\upsilon_{eIPL}(r) = Ar^{-n} + B + Cr,$4 tail and the boson peak within a single nonphononic sector (Moriel et al., 2023). In gravitational-wave theory, it yields analytic control over peaked primordial spectra and sensitivity envelopes (Li et al., 2024). In solar physics, it operationalizes finite-energy departures from scale-free flare statistics (Wheatland, 2010). In probabilistic forecasting, it formalizes the distinction between a tail law and an unmodeled or differently modeled body (Sangrody et al., 2017).

This suggests a general editorial characterization: the Power-Law+Peak model is not one model but a recurrent research pattern for representing systems in which asymptotic scaling coexists with a physically consequential finite scale.