An inversion formula for the 2-body interaction given the correlation functions

This paper proves that for a classical gas at infinite volume, the pair interaction potential can be uniquely reconstructed and convergently expanded in terms of its truncated correlation functions of all orders.

The Gist

Imagine you are a detective trying to figure out the rules of a game just by watching the players move around.

In the world of physics, the "players" are atoms or molecules, and the "game" is how they interact with each other. Usually, scientists know the rules (the interaction potential—how much energy it takes for two atoms to get close) and they calculate how the atoms will behave (the correlation functions—how likely you are to find two atoms at a certain distance).

But in the real world, we can't easily measure the invisible "rules" (the forces between atoms). We can measure how the atoms are arranged (the correlation functions) through experiments or computer simulations.

The Big Question: If we only see the results (the arrangement of atoms), can we work backward to discover the rules (the forces) that caused them?

This paper is a mathematical "recipe book" that says: Yes, we can. And not just a rough guess, but a precise, step-by-step formula to reconstruct the rules, provided we have enough data.

The Detective's Dilemma: The "First Guess" Trap

Imagine you are trying to guess the recipe for a soup just by tasting it.

• The Old Way (Iterative Boltzmann Inversion): You take a guess at the recipe, make the soup, taste it, see how it differs from the original, tweak the recipe, make it again, and repeat. You keep doing this loop until the soup tastes right.

  • The Problem: This is slow. It's like trying to find your way out of a maze by bumping into walls. Also, you might get the taste right but mess up the texture (other thermodynamic properties).

• The New Way (This Paper): Instead of guessing and checking, you use a "magic decoder ring." You look at the soup's ingredients (the data) and use a specific mathematical formula to instantly write down the recipe.

  • The Catch: The formula is complex. It doesn't just look at the soup's main flavor (pair correlations); it looks at the soup's flavor, the texture, the smell, and how all the ingredients interact in groups of three, four, or more.

The "Magic Decoder Ring" Explained

The authors (Frommer, Kuna, and Tsagkarogiannis) developed a formula that acts like a reverse-engineering machine.

1. The Input: You feed the machine data about how particles are clustered together. But not just pairs! You need to know how groups of 2, 3, 4, and even 100 particles behave together. Think of this as knowing not just who is standing next to whom, but how entire crowds form, how sub-groups break off, and how the whole room vibrates.
2. The Process: The paper uses a sophisticated mathematical toolkit called Ruelle Calculus.
   • Analogy: Imagine the particle interactions are a giant, tangled ball of yarn. The "correlation functions" are the shape of the ball. The authors found a way to untangle the yarn by looking at the shape of the ball from every possible angle. They use a special type of "mathematical expansion" (like peeling an onion layer by layer) to separate the signal from the noise.
3. The Output: The machine spits out the exact interaction potential—the mathematical description of the force between two particles.

Why is this a Big Deal?

The paper proves that this "reverse-engineering" formula actually works and doesn't blow up into nonsense (mathematical convergence).

• The Onion Metaphor:
  • Layer 1 (The Simple Guess): If you only look at pairs of particles, you get a rough idea of the force. It's like guessing the weather by looking at a single cloud.
  • Layer 2 (The Correction): But clouds interact with other clouds. The paper shows that to get the exact force, you need to add a correction term that accounts for how a third particle influences the pair.
  • Layer 3 (The Full Picture): Then you add a correction for groups of four, and so on. The paper proves that if you keep adding these layers (using data from groups of all sizes), the formula converges to the true answer.

The "Hard Core" Problem

One tricky part of the puzzle is "hard-core" particles (like billiard balls). They can't occupy the same space. If two particles are too close, the force is infinite.

• The paper handles this by saying, "Okay, we can't look at particles that are touching, but if we look at particles that are just a tiny bit apart, we can still figure out the rules." They prove that as long as the particles aren't perfectly crushed together, the math holds up.

Real-World Impact

Why should you care?

• Designing New Materials: Imagine you want to design a new plastic that is super strong but lightweight. You can simulate how atoms should behave to get that strength. This paper gives you the tool to figure out exactly what the atomic forces need to be to create that material.
• Better Simulations: Currently, scientists spend millions of hours running computer simulations to guess the right forces. This formula could let them calculate the forces directly from the desired outcome, saving massive amounts of time and computing power.

In a Nutshell

This paper is a breakthrough in inverse statistical physics. It moves us from "guessing and checking" to "direct calculation." It tells us that if we have enough detailed data about how particles cluster together (from pairs to huge groups), we can mathematically reconstruct the invisible forces holding them together, layer by layer, with perfect precision.

It's like finally having the ability to look at a finished cake and instantly write down the exact recipe, including the oven temperature and mixing speed, just by analyzing the crumbs.

Technical Summary

1. Problem Statement

The paper addresses the inverse problem in statistical physics: given the correlation functions $\rho^{(n)}$ of a classical gas (described by a Gibbs measure) at equilibrium, can one uniquely determine the underlying interaction potential (specifically the $n$-body interactions) that generates these correlations?

• Context: In experimental and simulation data, direct measurement of interaction potentials is impossible. Instead, one observes correlation functions.
• Current Limitations:
  • Iterative Methods: Techniques like Iterative Boltzmann Inversion (IBI) or Inverse Monte Carlo (IMC) attempt to find a potential by iteratively matching correlation functions. However, these methods often lack rigorous convergence guarantees and may fail to reproduce other thermodynamic quantities (like pressure).
  • Series Inversion: Previous attempts to invert the relationship between pair correlation and pair potential as a power series (e.g., in terms of density) lacked a rigorous proof of the radius of convergence.
• Goal: The authors aim to derive a rigorous inversion formula that expresses the pair interaction potential $H(x_2)$ (and chemical potential $\mu$) as a convergent power series expansion in terms of the truncated correlation functions of all orders ($\rho_T^{(n)}$), specifically in the infinite volume limit.

2. Methodology

The authors employ a framework based on Ruelle calculus and the properties of Janossy densities to bridge the gap between correlation functions and the Hamiltonian.

A. Theoretical Framework

1. Gibbs Measures and GNZ Equation: The system is modeled as a point process $P$ satisfying the Georgii-Nguyen-Zessin (GNZ) equation, which relates the expectation of adding a particle to the system with the interaction energy $W$.
2. Janossy Densities: The authors utilize Janossy densities $j_\Lambda^{(n)}$, which represent the marginal probabilities of finding $n$ particles in a bounded volume $\Lambda$. These are related to correlation functions via an explicit inclusion-exclusion formula (Eq. 2.3).
3. Ruelle Calculus: They introduce an algebraic structure on functions defined on finite configurations. Key operations include:
   • The convolution product $\ast$.
   • The exponential map $\exp_\ast$, which relates the full correlation functions $\rho$ to the truncated (connected) correlation functions $\rho_T$ via $\rho = \exp_\ast \rho_T$.
4. Logarithmic Expansion: By taking the logarithm of the Janossy density ratio $\log(j_\Lambda^{(2)}/j_\Lambda^{(0)})$, they isolate the Hamiltonian $H(x_2)$ and the chemical potential $\mu$. The core idea is to expand this logarithm as a power series in terms of the truncated correlations.

B. Key Assumptions

To prove convergence, three specific assumptions are introduced:

• Assumption A (Thermodynamic Limit): Ensures that the ratio of Janossy densities converges to the exponential of the Hamiltonian as the volume $\Lambda \to \mathbb{R}^d$.
• Assumption B (Strong Brillinger Mixing): Imposes a bound on the truncated correlation functions, ensuring they decay sufficiently fast. This is crucial for the convergence of the series.
• Assumption C (Hard-Core/Repulsion): A technical condition ensuring that the pair correlation $\rho^{(2)}$ does not vanish relative to the product of densities at large distances, handling systems with hard-core repulsion.

3. Key Contributions and Results

A. The Main Inversion Formula (Theorem 3.1)

The paper proves that under the stated assumptions, the chemical potential $\mu$ and the pair interaction energy $H(x_2)$ can be expressed as absolutely convergent series involving truncated correlation functions of all orders:

1. Chemical Potential:
   $$\mu = \log \rho + \sum_{k=1}^{\infty} \frac{(-1)^k}{k!} \int_{(\mathbb{R}^d)^k} \omega_1^{(1+k)}(0, y_k) dy_k$$
2. Pair Interaction Potential:
   $$H(x_2) = -\log\left(\frac{\rho^{(2)}(x_2)}{\rho^2}\right) - \sum_{k=1}^{\infty} \frac{(-1)^k}{k!} \int_{(\mathbb{R}^d)^k} \omega_2^{(2+k)}(x_2, y_k) dy_k$$

Here, $\omega_1$ and $\omega_2$ are recursively defined families of functions derived from the truncated correlations $\rho_T$.

• First Order Correction: The term for $k=1$ in the potential expansion corresponds to the standard "potential of mean force" approximation.
• Higher Order Corrections: The series includes terms involving 3-body, 4-body, etc., truncated correlations, providing a systematic way to improve accuracy beyond the mean-force approximation.

B. Recursive Definitions and Combinatorics

The paper provides explicit recursive formulas (Eqs. 3.7 and 3.8) to compute the coefficients $\omega^{(n)}$. These formulas involve combinatorial partitions of sets (Bell polynomials), linking the structure of the expansion to the graph-theoretic structure of cluster expansions.

• Lemma 3.2: Decomposes the expansion terms into contributions from single-particle effects ($\omega_1$) and genuine two-body effects ($\omega_2$).

C. Proof of Convergence

The most significant mathematical contribution is the rigorous proof of the radius of convergence for these series.

• The authors use exponential generating functions and Lambert W functions to bound the integrals of the recursive terms.
• They demonstrate that for sufficiently small particle density (controlled by the parameter $D_\rho$), the series converges absolutely. This resolves the "pending" status of the radius of convergence proof mentioned in previous literature.

4. Examples and Applications

The authors validate their assumptions and results on three distinct types of point processes:

1. Pair Interaction Systems: Standard Gibbs measures with stable pair potentials (e.g., Lennard-Jones type). They show that for low densities, the assumptions hold.
2. Kirkwood Closure Process: A specific model where correlations factorize. The authors derive explicit closed-form expressions for the $\omega$ terms, demonstrating the method's applicability to non-Gibbsian or simplified models.
3. Determinantal Point Processes: Models arising in random matrix theory and quantum systems. They show that under boundedness conditions on the kernel, the assumptions hold, extending the inversion formula to these quantum-like systems.

5. Significance and Impact

• Theoretical Rigor: This work provides the first rigorous proof of convergence for the inversion of correlation functions into interaction potentials using a power series expansion in the infinite volume limit.
• Beyond Iteration: Unlike iterative methods (IBI/IMC) which are heuristic and computationally expensive, this offers a direct, non-iterative analytical formula.
• Systematic Improvement: The formula naturally incorporates higher-order correlations. This suggests a strategy for designing effective coarse-grained models: one can truncate the series at a specific order (e.g., including 3-body terms) to match specific thermodynamic properties more accurately than the standard pair-potential approximation.
• Computational Physics: The results provide a theoretical foundation for "Inverse Henderson" problems, potentially leading to more efficient algorithms for recovering interaction potentials from simulation data by utilizing higher-order correlation data (up to order 4 or higher) rather than just pair correlations.

In summary, the paper establishes a mathematically sound bridge between microscopic interaction potentials and macroscopic correlation data, offering a convergent series expansion that improves upon existing approximation methods by systematically accounting for many-body effects.