The diffusion equation for non-Markovian Gaussian stochastic processes

This paper derives an exact, closed non-Markovian diffusion equation for the probability density of particle displacements driven by arbitrary Gaussian velocity processes by constructing a systematic hierarchy of equations based on Wick's theorem, which generalizes the Fokker-Planck description while preserving Gaussianity only in the infinite-order limit.

The Gist

Imagine you are watching a drunk person walking down a street. In the old, classic way of thinking about this (called the "Markovian" view), we assume the person has no memory. Every step they take is completely random and independent of the last one. If they stumble left, it doesn't change the odds of stumbling right next time. This is the "Fokker-Planck" equation, a famous rule that has described Brownian motion (the jittery movement of particles) for over a century.

However, in the real world, things often have memory. If that drunk person just stumbled left, they might be off-balance for a few seconds, making their next step more likely to be a recovery to the right. Their current movement is "connected" to their past. This is called a non-Markovian process.

This paper by Taloni, Pagnini, and Chechkin tackles a very specific, tricky problem: How do we write the exact mathematical rules for how a particle moves when it has memory, but its speed is still "Gaussian" (meaning it follows a nice, bell-curve distribution of speeds)?

Here is the breakdown of their discovery using simple analogies:

1. The Problem with Old Rules

The authors point out that previous attempts to describe this "memory-filled" movement (specifically the "Zwanzig-Balescu" and "Batchelor-Hänggi" equations) were like trying to describe a complex symphony by only listening to the first two notes.

• They worked okay for simple, short-term predictions.
• But they failed to capture the full "shape" of the movement over time. They couldn't perfectly predict the complex patterns of where the particle would be after many steps. They were approximations, not the exact truth.

2. The New Tool: "Wick's Theorem" as a Puzzle

To solve this, the authors used a mathematical tool called Wick's theorem.

• The Analogy: Imagine you have a long string of beads, where each bead represents a moment in time. You want to know how the whole string behaves. Wick's theorem says you don't need to look at the whole string at once. Instead, you can break the string down into pairs of beads.
• If you have 4 beads, you can pair them up in different ways (1-2 and 3-4, or 1-3 and 2-4, etc.).
• The authors realized that the complex movement of the particle is just the sum of all these possible "pairings" of past and present moments.

3. The "Connected" vs. "Disconnected" Clusters

The paper introduces a clever way to organize these pairings, borrowing a concept from quantum physics (Feynman diagrams).

• Disconnected Diagrams: Imagine a group of people at a party where some people are talking in one corner and others in another, but the two groups never interact. In the math, these are "disconnected."
• Connected Diagrams: Imagine a chain where everyone is holding hands in a single line. This is "connected."
• The authors found that to get the exact equation, you have to focus only on the "connected" chains. If you ignore the disconnected parts, you get a cleaner, more accurate picture of how the memory flows through time.

4. The Result: An Infinite Tower of Equations

The authors derived a new, exact equation (Equation 16 in the paper).

• The Old Way: Was like a flat, single-story house. It worked for simple cases but couldn't handle complex floors.
• The New Way: Is an infinite skyscraper.
  • The bottom floor (the first term) looks like the old, familiar equations.
  • But to get the perfect, exact answer, you have to add up an infinite number of higher floors.
  • Each new floor adds a layer of "memory" correction.
  • Crucial Point: The paper states that if you stop at any finite number of floors (truncate the series), the math loses its "Gaussian" nature (the bell curve shape gets distorted). You only get the perfect Gaussian shape back if you include the entire infinite tower.

5. What This Means for Real Physics

The authors tested their new "infinite tower" equation on two famous scenarios:

• The Ornstein-Uhlenbeck Process: This is the standard model for a particle with friction and memory. Their equation works perfectly here, recovering the known results but showing exactly how the memory terms stack up.
• Fractional Brownian Motion: This is a type of movement with very long-range memory (like a particle that "remembers" what happened hours ago). The authors showed that their equation correctly describes this movement, whereas previous equations (like the Batchelor-Hänggi one) gave the wrong answer.

Summary

In short, the paper says: "We found the exact recipe for how a particle moves when it has memory. Previous recipes were missing ingredients. Our new recipe uses a 'pairing' method to organize the memory, but to get the perfect result, you have to include an infinite number of terms. If you cut the recipe short, the math breaks."

They didn't invent a new drug or a new engine; they simply fixed the fundamental math that describes how things move when they remember their past.

Technical Summary

Technical Summary: The Diffusion Equation for Non-Markovian Gaussian Stochastic Processes

Problem Statement
The paper addresses the derivation of an exact evolution equation for the probability density function (PDF) of particle displacements, $P(x, t)$, generated by arbitrary Gaussian velocity processes. While the classical Fokker–Planck equation (FPE) and the Langevin description provide a robust framework for Markovian, stationary Gaussian processes (specifically the Ornstein–Uhlenbeck process), they fail to capture the exact dynamics of non-Markovian systems where velocity correlations are finite and non-exponential.

Existing approaches for non-Markovian Gaussian processes, such as the Zwanzig–Balescu equation (ZBE) and the Batchelor–Hänggi equation (BHE), rely on approximations. The authors note that while these equations apply to Gaussian stationary processes, they are not exact; specifically, they fail to reproduce Gaussian position moments beyond the second order. Furthermore, the BHE depends explicitly on the initial time $t_0$ and may violate positivity. The central problem is to derive a closed, exact diffusion equation that preserves the Gaussian nature of the process without assuming stationarity or Markovianity.

Methodology
The authors adopt a systematic approach based on the characteristic function of the position density, $\hat{P}(q, t)$, rather than starting from a specific Langevin equation. The methodology proceeds as follows:

1. Characteristic Function Expansion: The characteristic function is expanded in a Taylor series around $q=0$, relating it to the even moments of the displacement, $\langle x^{2n}(t) \rangle$.
2. Wick's Theorem Application: Since the velocity process $v(t)$ is assumed to be Gaussian, all statistical properties are determined by the two-time correlation function $\langle v(t_1)v(t_2) \rangle$. The authors utilize Wick's theorem to decompose higher-order velocity correlations into sums of products of two-point correlations.
3. Truncated Hierarchy: A moment-truncated characteristic function, $\hat{\rho}_N(q, t)$, is defined. The time evolution of this truncated function is expressed as a sum over all possible Wick pairings of velocity variables.
4. Diagrammatic Rearrangement: The authors introduce a diagrammatic representation of the Wick contractions. By regrouping these diagrams, they distinguish between "geometrically connected" Wick diagrams and disconnected ones. This step mirrors the linked-cluster theorem in quantum field theory, where only connected diagrams contribute to the logarithm of the generating functional.
5. Recursive Structure: The rearrangement leads to a hierarchical recurrence relation involving coefficients $R_k$, which represent the sum of all geometrically connected Wick diagrams of order $2k$.

Key Contributions and Results

1. Exact Evolution Equation: The primary result is the derivation of an exact, closed non-Markovian diffusion equation for $P(x, t)$:
   $$\frac{\partial}{\partial t}P(x, t) = \sum_{k=1}^{\infty} \int_0^t dt_1 \int_0^{t_1} dt_2 \cdots \int_0^{t_{2k-2}} dt_{2k-1} R_k \frac{\partial^{2k}}{\partial x^{2k}} P(x, t_{2k-1})$$
   Here, $R_k$ are kernels derived from the connected velocity correlation functions.

2. Generalization of Existing Models:

   • The first term of the series ($k=1$) recovers the Zwanzig–Balescu equation (ZBE).
   • The authors demonstrate that any finite truncation of this series destroys the Gaussian character of the solution. Only the full infinite series exactly preserves Gaussianity.
   • The equation is valid for both stationary and non-stationary kinetics, provided the single-time velocity probability density is Gaussian.

3. Combinatorial Analysis: The paper establishes a recursive relation for the number of connected Wick diagrams ($\bar{R}_k$), showing that the total number of Wick pairings $(2n-1)!!$ is the sum of contributions from connected diagrams of lower orders. This structure is analogous to the recurrence relations for connected Feynman diagrams in Quantum Electrodynamics (QED).

4. Applications to Specific Processes:

   • Ornstein–Uhlenbeck (OU) Process: In the overdamped limit ($\gamma \to \infty$), the equation correctly reduces to the standard Smoluchowski diffusion equation, with higher-order terms vanishing.
   • Fractional Brownian Motion (fBm): For fBm with Hurst exponent $1/2 < H < 1$, the leading term of the derived equation yields a fractional differential equation involving the Riemann–Liouville fractional derivative. The authors note this specific form has not been previously studied in the literature.

Significance and Claims
The paper claims to provide the first exact diffusion equation for the probability density of displacements generated by arbitrary non-Markovian Gaussian velocity processes. The significance lies in:

• Exactness: Unlike the ZBE or BHE, the derived equation (16) is exact and does not rely on approximations that fail to capture higher-order moments.
• Generality: It does not require the assumption of stationarity, making it applicable to a broader class of internal and external colored noises.
• Structural Insight: The derivation highlights that the preservation of Gaussianity in the position variable is an emergent property of the infinite-order limit of the hierarchy, governed by geometrically connected Wick contractions.

The authors explicitly state that for the antipersistent regime of fractional Brownian motion ($0 < H < 1/2$), the derivation of a leading-order equation remains an open problem due to the non-integrability of the velocity correlation kernel at coinciding times, which they intend to address in future work.