---
title: Scaled Brownian Motion (sBm)
url: https://www.emergentmind.com/topics/scaled-brownian-motion-sbm
type: topic
---

# Scaled Brownian Motion (sBm)

Scaled Brownian motion (sBm) is a fundamental class of Gaussian, non-stationary stochastic processes characterized by a power-law time-dependent diffusivity. Emerging as a paradigm for modeling anomalous diffusion in complex, heterogeneous, or actively driven environments, sBm differs sharply from ordinary Brownian motion due to its temporal non-stationarity, weak ergodicity breaking, and a breadth of non-equilibrium behaviors under external constraints, resetting, or heterogeneity. The mathematical tractability and minimal parameterization of sBm make it a ubiquitous tool for analyzing experimental single-particle tracking, active matter, and systems with evolving transport properties.

## 1. Mathematical Definition and Formulation

Scaled Brownian motion $\chi_\beta(t)$ is a Gaussian process defined by the stochastic differential equation (SDE)
\[
\frac{d}{dt}\chi_\beta(t) = \sqrt{2\,\beta\,K_{\beta}\,t^{\beta-1}}\,\xi(t)
\]
where $\xi(t)$ is Gaussian white noise, $\beta>0$ controls the scaling, and $K_\beta$ is the generalized diffusion constant. The instantaneous variance of the noise term manifests a power-law scaling, defining the time-dependent diffusivity
\[
D_\beta(t) = K_\beta\, t^{\beta-1}
\]
and leading to a mean squared displacement (MSD)
\[
\langle \chi_\beta^2(t)\rangle = 2 K_\beta\,\frac{t^\beta}{\beta}
\]
For the canonical one-dimensional formulation, the Fokker–Planck equation is
\[
\frac{\partial P(x,t)}{\partial t} = \frac{\partial^2}{\partial x^2}\left[ D_\beta(t) P(x,t) \right]
\]
The increment covariance is time-inhomogeneous:
\[
\langle \chi_\beta(t)\chi_\beta(s)\rangle = 2 K_\beta [ \min(t,s) ]^{\beta }
\]
The parameter range $\beta < 1$ describes subdiffusive processes, $\beta = 1$ is normal diffusion, and $\beta > 1$ indicates superdiffusion.

## 2. Physical Interpretation and Key Properties

The central feature of sBm is its explicit breaking of time-translation invariance. Unlike fractional Brownian motion (fBm), which is stationary in increments but with long-range correlations, sBm presents **Gaussian, Markovian but non-stationary** increments. Its defining attributes are:

- The MSD scales as $t^\beta$.
- The propagator is Gaussian: $P(x,t) = \left[4\pi K_\beta t^\beta\right]^{-1/2} \exp\left(-x^2 / (4 K_\beta t^\beta)\right)$.
- Two-time (auto)correlations depend on both arguments, not merely their difference.
- The process is weakly non-ergodic: long-time ensemble and time-averaged observables diverge, but trajectory-to-trajectory scatter of time-averaged MSD diminishes as the observation time increases [1405.2193], [1507.02450], [1501.04810].
- sBm can be interpreted as the effective mean-field limit of a cloud of non-ergodic subdiffusive continuous time random walkers (CTRW) [1311.3455].

Such dynamics physically describe systems with continuously aging or dynamically reorganizing environments, where the “mobility” of the medium varies as a power-law in time, e.g., cooling granular gases, active matter with evolving persistence, or time-dependent viscoelastic parameters.

## 3. Ergodicity, Time-Averaged Observables, and Aging

A hallmark of sBm is **weak ergodicity breaking**. For a trajectory $x(t)$ over duration $T$, the time-averaged MSD for lag $\Delta \ll T$ is
\[
\overline{\delta^2(\Delta)} = \frac{1}{T - \Delta} \int_0^{T-\Delta} [x(t'+\Delta) - x(t')]^2\,dt'
\]
and satisfies
\[
\left\langle \overline{\delta^2(\Delta)} \right\rangle \sim 2 K_\beta \frac{\Delta}{T^{1-\beta}}
\]
rather than $\propto \Delta^\beta$. The ergodicity-breaking parameter is
\[
\mathrm{EB} = \frac{\langle [\overline{\delta^2(\Delta)}]^2 \rangle - \langle \overline{\delta^2(\Delta)}\rangle^2}{\langle \overline{\delta^2(\Delta)} \rangle^2}
\]
with asymptotic regimes (for $\Delta/T \ll 1$) [1507.02450]:
- $0<\beta<1/2$: $\mathrm{EB} \sim (\Delta/T)^{2\beta}$
- $\beta=1/2$: $\mathrm{EB} \sim (\Delta/T) \ln(T/\Delta)$
- $\beta>1/2$: $\mathrm{EB} \sim (\Delta/T)$

Aging, i.e., commencing observation after a time $t_a > 0$, modulates both ensemble and time-averaged observables via scaling prefactors, leading to factorized law in the strong and weak aging regimes [1501.04810].

## 4. Confinement, Underdamped Formulation, and Non-Stationary Effects

Embedding sBm in confining potentials, e.g., harmonic traps, yields fundamentally different behavior compared to both fBm and CTRW [1405.2193], [1501.04810]. The MSD under trapping is
\[
\langle x^2(t) \rangle = 2 K_\beta t^\beta e^{-2k t} M(\beta, 1+\beta, 2k t)
\]
with $M$ the Kummer’s function. For long $t$, the MSD scales as $t^{\beta-1}$: subdiffusive ($\beta < 1$) cases decay to zero, superdiffusive ($\beta > 1$) cases diverge—no stationary plateau is ever attained except for $\beta=1$.

In underdamped sBm (UDSBM), inertia and time-dependent friction further prevent reduction to the standard overdamped sBm in broad parameter regimes: persistent inertial effects dominate at long times for subdiffusive or ultraslow cases, invalidating the overdamped approximation [1609.07250].

Non-stationarity pervades all two-point and multi-time observables. Specifically, memory kernels in the telegrapher or Smoluchowski–Fokker–Planck equations contain explicit time arguments, resulting in observable consequences for ISF, kurtosis, and orientation correlations, especially in active systems [2401.03127].

## 5. Extensions: Resetting, Heterogeneity, Superstatistics, and Curved Spaces

### (a) Resetting
SBM under stochastic resetting (to the origin, with full or partial memory reset) exhibits a broad taxonomy, sensitive to the reset protocol [1812.05667], [1812.05664]:
- Renewal resetting (reinitializing diffusion clock) eventually yields a non-equilibrium steady state, with a stationary PDF governed by the interplay of exponent $\beta$ and reset rate (e.g., stretched exponential, compressed exponential, or Laplace law for $\beta=1$).
- Non-renewal resetting (resetting position only) typically destroys steady states: the process remains non-stationary and PDFs continue to evolve.
- Optimal search (mean first-passage time) as a function of reset rate displays distinct dependencies on $\beta$ and domain size, with a single minimum for Poisson resetting.

### (b) Superstatistical Heterogeneity
Real experimental systems often involve distributed $\beta$ and/or $K_\beta$ over sub-populations. The superstatistical sBm aggregates over a joint PDF $\pi(\beta,K_\beta)$, resulting in explicitly non-Gaussian propagators and new scaling regimes [2206.07820], [2403.20206]. In mixtures with random exponents, e.g., two-point or beta distributions, ensemble moments, TAMSD, and ergodicity breaking display crossovers and persistent fluctuations not found in pure sBm.

### (c) Ultraslow sBm and Limit Behavior
For $\beta\to 0$, the diffusivity $D(t)\sim 1/t$ produces “ultraslow” sBm (USBM), with logarithmic MSD growth,
\[
\langle x^2(t) \rangle \simeq 2 D_0 \tau_0 \ln \frac{t}{\tau_0}
\]
and even slower vanishing of the ergodicity-breaking parameter (as $1/\ln^2(t)$), situating USBM at a singular endpoint of the broader sBm family [1503.08125].

### (d) Manifolds and Active Extensions
On compact manifolds (e.g., spheres), sBm preserves Gaussianity and navigation-strategy independence; long-time distributions are always equilibrium, with time-dependent approach rates [2210.04414]. In active systems, e.g., active Brownian tracers with propulsion angle undergoing sBm, the velocity autocorrelator becomes a stretched exponential and the persistence time depends non-trivially on $\beta$, generating genuine anomalous transport at long scales [2401.03127].

## 6. Single-Trajectory Analysis, Spectral Methods, and Bayesian Inference

Recent developments emphasize characterization of sBm from short or single trajectories, common in biophysical experiments [1903.06673], [2201.00654]. Key results include:

- The single-trajectory power spectral density (PSD) is indistinguishable from ordinary Brownian motion at fixed $T$: $\mu(f,T)\sim T^{\beta-1}/f^2$ for high $f$, making $T$-dependent scaling essential for correct inference.
- PSD amplitude (“ageing”) allows discrimination of $\beta$; supplementing time-averaged MSD analysis.
- Bayesian inference combining sBm and fBm models robustly recovers anomalous exponents, model selection, and uncertainties for realistic trajectory lengths, provided careful marginalization of noise and correct priors.

| Observable            | sBm (fixed $\beta$)         | Superstatistical / Random Exponent        |
|-----------------------|-----------------------------|-------------------------------------------|
| MSD                   | $t^\beta$                   | $E[t^A]$, mixture or hypergeometric law   |
| PDF                   | Gaussian                    | Superstatistical (non-Gaussian)           |
| Ergodicity Breaking   | Vanishes for $T\to\infty$   | May saturate to finite value              |
| Resetting             | Non-stationary vs. steady   | Crossover behaviors                       |

## 7. Significance, Limitations, and Applications

sBm provides a minimal, analytically tractable framework for anomalous diffusion arising from non-stationary environments but lacks long-range time correlations present in fBm or heavy-tailed renewal traps intrinsic to non-Gaussian CTRW. Its essential role is twofold:

1. **Practical modeling:** sBm’s closed-form solution and two-parameter nature ($K_\beta$, $\beta$) streamline the fitting and interpretation of anomalous yet Gaussian-like single-particle trajectories in viscous, glassy, or active baths.
2. **Benchmark and diagnostic:** The contrast between ensemble and time-averaged statistics, as well as reset-dependent phenomena, sharply discriminates sBm from other paradigms in anomalous diffusion—especially under experimental constraints or additional heterogeneity.

However, sBm’s intrinsic non-stationarity under confinement or resetting, and its lack of physical stationarity in closed environments, limit its applicability in strictly thermalized systems [1405.2193]. In complex, aging, or actively sustained environments, sBm (and its generalizations) remain central to both qualitative and quantitative understanding of anomalous transport.

---

For detailed derivations, computational protocols, and extensive simulation validations, see [2401.03127], [1405.2193], [1311.3455], [1507.02450], [1501.04810], [1903.06673], [2206.07820], [2403.20206], [1609.07250], [1812.05667], [2210.04414].

Source: https://www.emergentmind.com/topics/scaled-brownian-motion-sbm