🔬 condensed matter

Comparison of different exact generalized Langevin equations with a non-linear potential of mean force and an observable-dependent mass and friction

This paper utilizes the Mori-Zwanzig projection formalism to analyze and compare four distinct exact generalized Langevin equations for a scalar observable with a non-linear potential and observable-dependent mass and friction, highlighting that including the effective kinetic energy in the potential is advantageous for observables satisfying Wick's theorem as it ensures the correct distribution even without friction or orthogonal force contributions.

Original authors: Benjamin J. A. Héry, Lucas Tepper, Roland R. Netz

Published 2026-06-30

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Benjamin J. A. Héry, Lucas Tepper, Roland R. Netz

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to describe the movement of a single, complex dancer (let's call her "A") in a crowded, chaotic ballroom filled with thousands of other dancers. You can't track every single person in the room, so you want a simple set of rules to predict how "A" moves based only on her own position and speed.

This paper is about finding the perfect set of rules (mathematical equations) to describe that dancer's motion, even though she is constantly bumping into and being pushed by the invisible crowd around her.

Here is the breakdown of the paper's findings using simple analogies:

1. The Problem: The "Ghost" Forces

In physics, when you zoom out to look at just one part of a complex system (like a protein folding or a dancer moving), the rest of the system acts like a "ghost." It creates two main effects on our dancer:

Friction: The crowd slows her down (like moving through water).
Random Kicks: The crowd bumps into her unpredictably (like a random breeze).

For decades, scientists have used a tool called the Mori-Zwanzig formalism to write down equations for this. It's like a universal recipe that says: "Acceleration = Force from a hill + Friction from the crowd + Random kicks."

2. The Four Different Recipes

The authors of this paper took that universal recipe and cooked up four slightly different versions of the equation. They are all "exact" (mathematically perfect), but they differ in how they handle the ingredients:

The "Fixed Mass" Version: Imagine the dancer always weighs the same, no matter where she is on the dance floor.
The "Variable Mass" Version: Imagine the dancer feels heavier or lighter depending on where she is standing (maybe the floor gets sticky in some spots).
The "Energy-Inclusive" Version: This version is special because it includes the dancer's own kinetic energy (her speed) directly into the "hill" she is rolling down.

The paper compares these four versions to see which one tells the truest story about the dancer's life.

3. The Big Discovery: The "Perfect Map"

The most important finding is about the Variable Mass versions that include the dancer's own energy (called the IKE-GLE and DKE-GLE in the paper).

The Analogy: Imagine you are trying to draw a map of a hilly landscape.
- The "Old" maps (the other equations) tell you where the hills are, but they require you to constantly add "correction notes" about the wind and the random bumps to get the right picture.
- The "New" maps (the IKE-GLE and DKE-GLE) are so accurate that the hills themselves already contain the correct shape of the landscape. You don't need the "correction notes" (the random kicks and friction) to know where the dancer is likely to be. The map itself is perfect.

Why does this matter?
If you want to simulate how a protein folds (like the dancer moving), using these "Energy-Inclusive" equations means you get the correct distribution of positions and speeds just by looking at the "hill" (the potential energy). You don't have to rely on the messy, random noise to get the right answer. It's like having a GPS that knows the route perfectly without needing to ask for directions every five seconds.

4. The Real-World Test: The Villin Protein

To prove their point, the authors tested these equations on a real-world example: a small protein called Villin folding up.

They took data from super-computer simulations of this protein.
They tried to fit the four different equations to the data.
The Result: Even though the protein's movement wasn't perfectly "mathematical" (it didn't follow a simple bell-curve rule perfectly), the Variable Mass + Energy-Inclusive equations (IKE-GLE) were the best at predicting the protein's behavior. They matched the computer simulation data almost perfectly, whereas the simpler "Fixed Mass" models failed to capture the full picture.

5. A Warning About "Randomness"

The paper also points out a common misconception. Many scientists assume that the "random kicks" from the crowd and the "friction" are perfectly linked by a simple rule (called the Second Moment Relation).

The Paper's Claim: This simple link does not always exist. It's like assuming that every time the wind blows harder, the rain gets heavier by the exact same amount. Sometimes, the wind blows hard, but the rain stays light. The authors show that for these complex systems, you cannot assume this simple link exists unless very specific, rare conditions are met.

Summary

This paper is a guidebook for physicists who want to model complex systems (like proteins or polymers). It says:

There are four ways to write the rules of motion.
The best way is to include the object's own speed into the definition of the "landscape" it moves through.
This method gives you the most accurate picture of the system's behavior without needing to guess about the random noise.
Don't assume the random noise and friction are always perfectly linked; they often aren't.

In short: If you want to predict how a complex system moves, don't just look at the friction; look at how the system's own energy shapes the path it travels.

Technical Summary: Comparison of Exact Generalized Langevin Equations with Non-Linear Potentials and Observable-Dependent Mass

Problem Statement
The Mori-Zwanzig projection formalism provides an exact framework for deriving Generalized Langevin Equations (GLEs) for arbitrary coarse-grained reaction coordinates in complex systems. While the resulting equations universally consist of a Markovian force, a non-Markovian friction term, and an orthogonal (random) force, the specific form of these terms depends heavily on the choice of projection operators. A critical challenge arises when dealing with observables where the effective mass and friction are dependent on the observable itself, and the potential of mean force is non-linear. Furthermore, the relationship between the memory kernel and the orthogonal force (the Second Moment Relation, or SMR) is often assumed to hold but is not guaranteed in general cases. This paper addresses the need to systematically compare different exact GLE formulations to determine their properties, specifically regarding the inclusion of kinetic energy in the effective potential and the dependence of friction kernels on the observable.

Methodology
The authors employ the Mori-Zwanzig operator formalism to derive and compare four distinct exact GLEs for a scalar observable $A(t)$ that depends on positional degrees of freedom. The derivation utilizes a dual-projection scheme involving pairs of projection operators ( $P_1, Q_1$ and $P_2, Q_2$ ).

The study focuses on the specific case where the velocity of the observable, $\dot{A}$ , satisfies the conditional Wick's theorem (Eq. 21). Four combinations of projection operators are utilized to generate four distinct GLEs:

I-GLE: Uses $P_1 = P_{LV}$ (Linear Velocity) and $P_2 = P_M$ (Mori). Features an observable-independent friction kernel.
D-GLE: Uses $P_1 = P_2 = P_{LV}$ . Features an observable-dependent friction kernel.
IKE-GLE: Uses $P_1 = P_{KE}$ (Kinetic Energy) and $P_2 = P_M$ . Features an observable-independent friction kernel and includes the effective kinetic energy in the potential.
DKE-GLE: Uses $P_1 = P_{KE}$ and $P_2 = P_{LV}$ . Features an observable-dependent friction kernel and includes the effective kinetic energy in the potential.

The authors define the effective mass $M_{eff}(A)$ and various effective potentials ( $U_{eff}$ and $U^{KE}_{eff}$ ) for each formulation. They rigorously analyze the conditions under which the Second Moment Relation (SMR) holds between the memory kernel and the orthogonal force. Finally, the theoretical frameworks are applied to molecular dynamics (MD) simulation data of the Villin protein folding, using the fraction of native contacts as the reaction coordinate. The authors compare the performance of the I-GLE, IKE-GLE, and a Constant-Mass GLE (CM-GLE) against the MD data.

Key Contributions and Results

Derivation of Four Exact GLEs: The paper systematically derives four exact GLEs. A primary distinction is identified in the definition of the effective potential:
- The I-GLE and D-GLE utilize a potential $U_{eff}(A)$ that depends only on position.
- The IKE-GLE and DKE-GLE utilize a potential $U^{KE}_{eff}(A, \dot{A})$ that explicitly includes the effective kinetic energy term ( $\frac{1}{2}M_{eff}(A)\dot{A}^2$ ).
Advantage of Kinetic Energy Inclusion: The authors demonstrate that for observables satisfying Wick's theorem, the IKE-GLE and DKE-GLE formulations reproduce the exact joint distribution of the observable $A$ and its velocity $\dot{A}$ solely through the Boltzmann factor of the potential $U^{KE}_{eff}$ . This occurs without requiring contributions from the non-Markovian friction or orthogonal forces. This is presented as a significant advantage for simulating complex data, as it ensures the correct equilibrium distribution is recovered naturally.
Validity of the Second Moment Relation (SMR): The study clarifies that the SMR (where the memory kernel is proportional to the autocorrelation of the orthogonal force) does not hold in general for these GLEs. The SMR is only valid under restrictive conditions, such as when the conditional average of the orthogonal force is zero or satisfies a specific partial differential equation. This finding challenges interpretations that attribute SMR breaking solely to non-harmonic potentials, suggesting instead that it is a consequence of the specific projection operator choices.
Observable-Dependent Friction: The paper confirms that friction kernels can be either independent (I-GLE, IKE-GLE) or dependent (D-GLE, DKE-GLE) on the observable $A$ , depending on the projection operators chosen.
Application to Protein Folding (Villin):
- The authors apply these GLEs to the folding dynamics of Villin. They note that the fraction of native contacts reaction coordinate does not strictly satisfy Wick's theorem (Eq. 21) because it involves a sum of multiple bilinear combinations ( $N_d > 1$ ).
- Despite this violation, numerical results show that the I-GLE and IKE-GLE accurately approximate the joint distribution $P(A, \dot{A})$ and the effective mass $M_{eff}(A)$ extracted from MD data.
- The Constant-Mass GLE (CM-GLE) fails to reproduce the joint distribution, only capturing the marginal distribution of $A$ .
- The memory kernels $\Gamma(t)$ extracted from MD data using all three GLEs (I-GLE, IKE-GLE, CM-GLE) are found to be remarkably similar.
- Analysis of the orthogonal forces reveals that while the CM-GLE yields an approximately Gaussian random force distribution, the I-GLE and IKE-GLE exhibit substantial non-Gaussian characteristics, reflecting the strong observable dependence of the mass.

Significance and Claims
The paper claims that the inclusion of the effective kinetic energy in the potential (as seen in the IKE-GLE and DKE-GLE) is particularly advantageous for analyzing complex data. When the observable satisfies Wick's theorem, these formulations guarantee the correct joint equilibrium distribution of position and velocity without relying on the friction or noise terms.

Regarding the application to protein folding, the authors modestly claim that while the Villin reaction coordinate does not strictly satisfy the theoretical prerequisites (Wick's theorem), the I-GLE and IKE-GLE formulations still serve as "good approximations" for describing folding dynamics. They successfully reproduce key kinetic quantities (potential of mean force, configuration-dependent mass, and memory kernels) and the joint distribution $P(A, \dot{A})$ observed in MD simulations. The work suggests that these exact GLE formulations are robust tools for modeling systems with non-linear potentials and observable-dependent masses, even when strict theoretical conditions are only approximately met.