Ultra slow sub-logarithmic diffusion of a sluggish random walker subject to resetting with memory

This paper presents an exact solution for a Brownian particle with algebraically decaying diffusion subject to memory-dependent resetting, revealing that the interplay of these mechanisms leads to ultra-slow sub-logarithmic diffusion characterized by a bimodal, non-Gaussian position distribution.

The Gist

Imagine you are watching a very tired, very forgetful explorer wandering through a vast, featureless desert. This explorer is our "sluggish random walker."

This paper is about figuring out exactly where this explorer will be after a long time, given two very specific rules that make their journey incredibly slow and strange.

The Two Rules of the Game

1. The "Heavy Boots" Rule (Sluggishness)
Imagine the explorer's boots get heavier the further they walk from their starting camp.

• Near the camp, they can jog easily.
• A mile out, they have to shuffle.
• Ten miles out, they are practically crawling because the sand feels like thick glue.
• The Science: In physics terms, the "diffusion coefficient" (how fast they move) drops as they get further away. The paper calls this $D(x) \sim |x|^{-\alpha}$. The further they go, the slower they get.

2. The "Nostalgic Reset" Rule (Memory)
Now, imagine the explorer has a strange habit. Every now and then, a bell rings. When it rings, they don't go back to the camp. Instead, they close their eyes, pick a random moment from their entire past history, and instantly teleport back to the spot they were standing at that exact moment.

• If they spent a lot of time sitting near a specific rock, they are more likely to land there again because they were there for a long time.
• The Science: This is "resetting with memory." Unlike a standard reset where you always go back to zero, this one sends you back to any place you've ever been, weighted by how much time you spent there.

The Big Surprise: Ultra-Slow Motion

Usually, if you have a walker that gets tired (Rule 1) and a walker that keeps getting sent back to old spots (Rule 2), you might expect them to get stuck in one place or move very slowly.

The authors of this paper solved the math to see exactly what happens when you combine both rules.

The Result: The explorer moves at a speed that is almost comically slow.

• Normal walking: Distance grows with time ($t$).
• Tired walking: Distance grows with the square root of time ($\sqrt{t}$).
• Tired + Nostalgic walking: Distance grows with the logarithm of time.

To put this in perspective:
If you walk for 100 years, you might be 10 miles away.
If you walk for 10,000 years (100 times longer), you might only be 12 miles away.
If you walk for 1,000,000 years, you might only be 14 miles away.

The paper calls this "Ultra slow sub-logarithmic diffusion." It's like the explorer is stuck in a time warp where moving forward requires an impossible amount of effort, and their nostalgia keeps pulling them back to the places they've already been, preventing them from ever truly exploring new territory.

The Shape of the Journey

You might think that because they are so slow, they would just sit in a tight little pile at the center. But that's not what happens.

• The Shape: If you took a snapshot of where all these explorers are at a very late time, the distribution wouldn't look like a normal bell curve (Gaussian). Instead, it looks like a dumbbell or a bathtub.
• Why? There is a "hole" in the middle. The explorer is actually repelled from the starting point because the "heavy boots" rule makes it hard to stay there, but the "nostalgic reset" keeps pulling them back to the places they visited just a bit further out.
• The Tails: The distribution has "non-Gaussian tails," meaning there are a few explorers who manage to get very far out, but they are very rare.

The "Magic" Math Trick

The authors didn't just guess this; they found an exact mathematical formula that describes the explorer's position at any time, not just at the end.

They used a clever trick involving "spectral decomposition" (breaking the problem down into simple vibrating waves, like plucking a guitar string) to solve the complex equations. They found that even though the math is incredibly complicated, the final shape of the explorer's location follows a very specific, predictable pattern that depends on how "heavy" the boots are (the $\alpha$ parameter).

Why Should We Care?

This isn't just about imaginary tired explorers. This model helps us understand real-world things:

1. Animal Behavior: Many animals (like monkeys or elk) don't just wander randomly. They remember where they found food or water and tend to revisit those spots. This model explains why their movement patterns look "sluggish" and why they stay within a specific "home range" rather than wandering off into infinity.
2. Search Strategies: If you are looking for something (like a lost hiker or a robot searching for a signal), knowing that "memory" combined with "difficulty of movement" creates ultra-slow progress is crucial. It tells us that if an agent keeps revisiting old spots, it might take an eternity to find something new.

The Bottom Line

The paper tells us that when you combine getting tired the further you go with constantly revisiting your past, you create a system that moves at a snail's pace. The explorer never truly settles down, but they also never really get anywhere fast. They are trapped in a slow-motion dance between their past and their present, moving only as fast as the logarithm of time allows.

Technical Summary

1. Problem Statement

The paper investigates the stochastic dynamics of a "sluggish" random walker in $d$-dimensional space subject to stochastic resetting with memory. The model combines two distinct physical mechanisms that typically slow down diffusion:

1. Sluggish Diffusion: The particle undergoes Brownian motion with a space-dependent diffusion coefficient $D(x) = |x|^{-\alpha}$ (where $\alpha > 0$). As the particle moves further from the origin, the diffusion rate decays algebraically, making the walker increasingly "sluggish."
2. Memory-Dependent Resetting: Instead of resetting to a fixed point (e.g., the origin), the particle resets to a previously visited position. The reset occurs at a constant rate $r$. When a reset happens at time $t$, the particle chooses a random past time $t' \in [0, t]$ uniformly (probability density $1/t$) and returns to the position $x(t')$ it occupied at that time. This mechanism favors revisiting frequently visited regions, inducing localization.

The central question is: How does the combination of spatially decaying diffusion and memory-based resetting affect the long-time behavior of the walker's position distribution and typical displacement?

2. Methodology

The authors employ an analytical approach based on the Fokker-Planck equation and spectral decomposition.

• Governing Equation: They derive the Fokker-Planck equation for the position distribution $P_r(x, t)$ in one dimension (and later generalize to $d$ dimensions). The equation includes:

  • A diffusion term with the Itô interpretation: $\partial_x^2 [D(x) P(x, t)]$.
  • A loss term due to resetting: $-r P(x, t)$.
  • A gain term due to resetting from past positions: $\frac{r}{t} \int_0^t P(x, t') dt'$.
  • Note: The memory kernel allows the equation to close on the one-point function without requiring higher-order correlations.

• Separation of Variables: The authors use the ansatz $P_r(x, t) = \phi(x) f_r(t)$. This separates the problem into:

  • Time-dependent part: An ordinary differential equation (ODE) involving the resetting rate $r$ and an eigenvalue $\lambda$. The solution involves Kummer's confluent hypergeometric function $M(a, b, z)$.
  • Space-dependent part: An eigenvalue equation involving the diffusion coefficient $D(x) = |x|^{-\alpha}$. This leads to a Bessel differential equation. The solutions are expressed in terms of Bessel functions $J_{\pm \nu}$, where $\nu = 1/(\alpha + 2)$.

• Spectral Decomposition: The full solution is constructed as an integral over the eigenvalues $\lambda \geq 0$. The spectral density is determined by applying the initial condition (particle starts at the origin) and utilizing orthogonality relations of Bessel functions.

• Generalized Moments: To verify results and avoid the difficulty of computing the mean square displacement (which diverges or is hard to compute directly), the authors calculate generalized moments $\mu_{r, \alpha}(m, t) = \langle |x|^m \rangle$ by expanding the spectral solution in power series and matching coefficients.

3. Key Contributions and Results

A. Exact Position Distribution

The paper provides an exact analytical expression for the position distribution $P_{r, \alpha}(x, t)$ at all times $t$ and for arbitrary spatial dimensions $d \geq 1$.

• 1D Solution: The distribution is given by a spectral integral involving Bessel functions and Kummer's functions (Eq. 39/40).
• Higher Dimensions: The solution is generalized to $d$ dimensions (Eq. 82), showing that the radial dependence follows a similar structure with modified Bessel indices.

B. Ultra-Slow Sub-Logarithmic Diffusion

The most significant finding concerns the asymptotic behavior at late times ($t \to \infty$):

• Typical Displacement: The typical distance of the walker from the origin grows as:
  $$x_{\text{typ}} \sim [\ln(rt)]^{\frac{1}{\alpha+2}}$$
  This represents an ultra-slow, sub-logarithmic diffusion.
  • Without memory ($r=0$), the walker is subdiffusive: $x \sim t^{1/(\alpha+2)}$.
  • With memory but constant diffusion ($\alpha=0$), the walker is logarithmic: $x \sim \sqrt{\ln(rt)}$.
  • The combination yields a growth rate slower than any power law and slower than the square root of a logarithm.

C. Scaling Law and Bimodality

Despite the extreme slowing of dynamics, the system does not reach a static steady state. Instead, it approaches a scaling form:
$$P_{r, \alpha}(x, t) \approx \left( \frac{r}{\ln(rt)} \right)^\nu G_\alpha \left( \frac{x}{[\ln(rt)]^\nu} \right)$$
where $\nu = 1/(\alpha+2)$.

• Universality of the Scaling Function: Remarkably, the scaling function $G_\alpha(z)$ is identical to that of the sluggish walker without resetting ($r=0$).
• Shape: For any $\alpha > 0$, the scaling function is bimodal (non-monotonic) with a minimum at $x=0$ and non-Gaussian tails. This indicates that the particle is repelled from the origin (due to the vanishing diffusion coefficient) but prevented from escaping to infinity by the memory-based resetting.

D. Generalized Moments

The authors derive exact expressions for generalized moments $\langle |x|^m \rangle$ at all times.

• They show that the moment of order $m = \alpha + 2$ (and generally $n(\alpha+2)$) is exactly proportional to the moments of the standard memory-resetting model (where $\alpha=0$).
• Specifically, $\langle |x|^{\alpha+2} \rangle_{r, \alpha} \propto \langle x^2 \rangle_{r, 0}$. This establishes a deep mathematical link between the sluggish model and the standard model with constant diffusion.
• These analytical predictions were verified against numerical simulations (Monte Carlo), showing excellent agreement.

4. Significance and Implications

• Theoretical Physics: The paper solves a non-trivial non-equilibrium stochastic process exactly. It demonstrates that memory-based resetting can induce a new universality class of "ultra-slow" diffusion, distinct from standard subdiffusion or logarithmic diffusion.
• Animal Movement Modeling: The model serves as a robust theoretical framework for animal foraging behavior. It captures two key biological features:
  1. Site Fidelity: Animals tend to return to familiar, resource-rich areas (memory resetting).
  2. Exploration Cost: Moving away from a central home range becomes energetically or physically more difficult (sluggish diffusion).
     The result suggests that such animals exhibit extremely slow range expansion, effectively localizing their movement within a slowly expanding logarithmic envelope.
• Mathematical Insight: The work highlights the power of spectral methods in solving Fokker-Planck equations with non-local (memory) terms and space-dependent coefficients. It also reveals that the scaling function of the position distribution is robust against the introduction of memory resetting, changing only the time scale of the dynamics.

In summary, the paper establishes that the interplay between spatially decaying mobility and memory-driven returns creates a unique dynamical regime where the walker explores space at an ultra-slow sub-logarithmic rate, while maintaining a specific bimodal spatial distribution shape.