Moreau-Yosida-based Kohn-Sham Inversion for Periodic Systems

This paper establishes a theoretical and numerical framework for density-potential inversion in periodic systems using Moreau-Yosida-regularized density-functional theory, leveraging lower semicontinuity of the kinetic-energy functional and proximal mappings to recover exchange-correlation potentials for Kohn-Sham and Gross-Pitaevskii equations.

The Gist

Imagine you are a detective trying to solve a mystery, but you only have the footprints left behind, not the person who made them. In the world of quantum physics, the "footprints" are the electron density (where electrons are likely to be found), and the "person" is the potential (the invisible forces and energy fields that push and pull those electrons into place).

This paper is about building a better, more reliable magnifying glass to reconstruct the invisible forces just by looking at the footprints.

Here is a breakdown of the paper's ideas using everyday analogies:

1. The Big Problem: The "Reverse Engineering" Puzzle

In chemistry and materials science, scientists use a tool called Density-Functional Theory (DFT) to predict how atoms behave. Usually, they start with the forces (the potential) and try to calculate where the electrons will go (the density). This is like knowing the wind speed and direction and predicting where a leaf will land.

However, sometimes scientists know exactly where the leaf landed (the density) because they measured it in a lab or from a super-accurate computer simulation, and they want to figure out what the wind was doing. This is called "inversion."

• The Challenge: This is incredibly hard. Small errors in the leaf's position can lead to wild guesses about the wind. It's like trying to guess the exact shape of a hidden mold just by looking at the cast it made; if the cast is slightly blurry, your guess about the mold could be completely wrong.

2. The Solution: The "Smoothie" Filter (Moreau–Yosida Regularization)

The authors introduce a mathematical technique called Moreau–Yosida regularization.

• The Analogy: Imagine trying to find the bottom of a very bumpy, rocky valley (the true physical system). If you try to walk straight to the bottom, you might get stuck on a small rock or fall into a hole.
• The Trick: Instead of looking at the bumpy ground directly, the authors put a thick, soft blanket over the rocks. This "blanket" smooths out the bumps, making the valley look like a gentle, easy-to-walk slope.
• The Process: They solve the problem on this smooth, easy slope first. Then, they slowly pull the blanket away layer by layer. As the blanket gets thinner, the smooth slope gets closer and closer to the real, bumpy ground. By watching how the solution changes as the blanket disappears, they can accurately reconstruct the original, complex shape of the valley.

3. The "Proximal Point": The Best Guess

To pull the blanket away, the math uses something called a proximal mapping.

• The Analogy: Think of this as a "smart guesser." If you give it a blurry photo of a face (the density), it doesn't just guess randomly. It finds the "closest possible" clear face that fits the blurry photo while also obeying the laws of physics.
• The paper proves that this "smart guesser" is very stable. If you give it a slightly blurry photo, it won't hallucinate a completely different face; it will just make a small, predictable adjustment. This stability is crucial because real-world data is never perfect.

4. The Playground: Periodic Systems (The Infinite Floor)

Most previous attempts to do this were like trying to solve the puzzle on a small, isolated island (like a single molecule). This paper focuses on periodic systems, which are like an infinite floor made of identical tiles (like a crystal or a solid metal).

• The Innovation: The authors built their mathematical "floor" using a special type of grid called Homogeneous Sobolev spaces.
• Why it matters: In an infinite crystal, the rules are different than on an island. The authors proved that their "smart guesser" (the proximal mapping) works perfectly on this infinite, tiled floor. They showed that the math holds up even when you are dealing with the repeating patterns of solid materials like silicon or salt.

5. The Results: Does the Magnifying Glass Work?

The team tested their method on two types of puzzles:

1. The Simple Test (Gross–Pitaevskii Equation): A simplified 1D model. They showed that as they pulled the "blanket" away (made the math more precise), their reconstructed forces matched the true forces almost perfectly.
2. The Real-World Test (Bulk Materials): They applied this to real 3D materials like Silicon, Gallium Arsenide, Sodium Chloride, and Potassium Chloride.
   • They took the "footprints" (electron density) from a standard computer simulation.
   • They ran their inversion algorithm.
   • The Result: They successfully reconstructed the "forces" (specifically the exchange-correlation potential) that created those footprints. The reconstructed forces looked almost identical to the ones used in the original simulation.

Summary of the Paper's Claims

• What they did: They created a rigorous mathematical framework to reverse-engineer the forces in solid materials just by looking at electron density.
• How they did it: They used a "smoothing" technique (Moreau–Yosida) that turns a hard, jagged math problem into a smooth one, solves it, and then refines the answer by slowly removing the smoothing.
• What they proved: They mathematically proved that this method is stable and works for infinite, repeating crystal structures (periodic systems).
• What they showed: They ran computer simulations on real materials (like salt and silicon) and demonstrated that their method can accurately recover the hidden forces, even when the input data has small imperfections.

In short: They built a reliable, mathematically sound "reverse-engineering machine" that can tell us exactly what invisible forces are shaping the electrons in solid materials, simply by looking at where the electrons are.

Technical Summary

Technical Summary: Moreau–Yosida-based Kohn–Sham Inversion for Periodic Systems

Problem Statement
Density-functional theory (DFT) is a cornerstone of electronic structure calculations, yet the development of accurate exchange-correlation (xc) functionals remains a significant challenge. A primary difficulty lies in the limited mathematical insight into the relationship between the exact universal density functional and widely used approximations. A critical step in improving these approximations is the "inverse Kohn–Sham (KS) problem": reconstructing the effective KS potential (specifically the xc potential) from a known ground-state density. While several inversion schemes exist, robust and efficient numerical methods for periodic systems (solid-state materials) remain an open problem. Existing approaches often lack a rigorous functional-analytic foundation or struggle with stability and convergence in periodic settings.

Methodology
The authors develop a mathematically rigorous framework for density-potential inversion in periodic systems based on Moreau–Yosida (MY) regularisation. The methodology proceeds through the following steps:

1. Functional-Analytic Framework: The theory is formulated within the dual of the periodic homogeneous Sobolev space, $H^{-1}_{\text{per,hom}}(\Omega)$, for densities, and its dual, $H^{1}_{\text{per,hom}}(\Omega)$, for potentials. This choice of space is crucial as it naturally handles the mean-field Hartree energy and the gauge freedom of potentials (defined up to a constant).
2. Moreau–Yosida Regularisation: The universal density functional (specifically the kinetic-energy constrained-search functional) is regularised using the MY envelope. This involves minimizing a functional of the form $F(\sigma) + \frac{1}{2\varepsilon}\|\sigma - \rho_{\text{gs}}\|^2_X$, where $\rho_{\text{gs}}$ is the target ground-state density and $\varepsilon > 0$ is a regularisation parameter.
3. Proximal Mapping: The minimiser of the regularised functional is the "proximal density" $\rho_\varepsilon$. The effective KS potential is recovered via a limiting procedure:
   $$v_{\text{KS}} = \lim_{\varepsilon \to 0^+} \frac{1}{\varepsilon} J(\rho_\varepsilon - \rho_{\text{gs}})$$
   where $J$ is the duality mapping between the density and potential spaces.
4. Theoretical Guarantees: A key analytical contribution is the proof that the kinetic-energy constrained-search functional is lower semicontinuous (l.s.c.) in the chosen topology. This ensures the existence of the proximal point and the validity of the inversion scheme.
5. Numerical Implementation: The proximal density is computed using three distinct approaches:
   • Direct Minimisation: Minimising the energy functional directly over orbitals (using BFGS or Newton methods).
   • Self-Consistent Field (SCF): Solving KS-like Euler-Lagrange equations iteratively.
   • Proximal Point Algorithm: An iterative scheme specifically designed for non-smooth convex optimisation.

Key Contributions
The paper makes three primary contributions:

1. General Inversion Scheme: The authors formulate a general MY-based inversion algorithm for densities in reflexive Banach spaces. They prove that the scheme relies only on the lower semicontinuity of the functional and the non-expansiveness of the proximal mapping, which bounds the propagation of errors in the input density.
2. Application to Periodic Systems: The framework is specialised to periodic homogeneous Sobolev spaces. The authors prove the lower semicontinuity of the kinetic-energy functional in this specific topology, thereby placing the Zhao-Morrison-Parr (ZMP) method on a firm functional-analytic footing for the first time in periodic settings. This allows for the determination of the exact xc potential as an element of minimal norm in the subdifferential.
3. Numerical Analysis: The paper provides a comprehensive numerical investigation of the inversion scheme using the periodic one-dimensional Gross–Pitaevskii equation and three-dimensional bulk materials (Silicon, Gallium Arsenide, Sodium Chloride, Potassium Chloride).

Results

• Convergence: Numerical experiments demonstrate that the proximal density $\rho_\varepsilon$ converges strongly to the ground-state density $\rho_{\text{gs}}$ as $\varepsilon \to 0$. The reconstructed potential converges to the reference xc potential.
• Algorithm Performance: For the Gross–Pitaevskii equation, direct minimisation using a full Newton method (implemented in DFTK) showed superior convergence compared to SCF and quasi-Newton approaches, particularly for small $\varepsilon$. The proximal point algorithm exhibited similar convergence properties to the PDE solver but incurred higher computational overhead.
• Bulk Materials: The method successfully recovered the xc potential for various bulk materials. The results indicate that the convergence rate depends on the smoothness of the target potential; resolving sharp features (e.g., near atomic nuclei) requires smaller values of $\varepsilon$.
• Guiding Functionals: The study investigated the impact of the guiding functional's composition (Hartree energy, external potential). Including the external potential in the guiding functional significantly improved the accuracy of the reconstructed potential.
• Stability: The inversion scheme was tested with perturbed input densities (simulated by Fourier truncation). The results showed that the accuracy of the reconstructed potential is bounded by the magnitude of the input error, confirming the stability of the regularised approach.

Significance
The paper claims that its primary significance lies in providing a mathematically rigorous and numerically stable framework for KS inversion in periodic systems. By establishing the lower semicontinuity of the kinetic-energy functional in periodic homogeneous Sobolev spaces, the authors validate the use of Moreau–Yosida regularisation for recovering exact exchange-correlation potentials. This work bridges the gap between abstract convex analysis and practical DFT applications, offering a robust tool for analyzing and improving density functionals. Furthermore, the numerical results suggest that the MY-based approach is a viable alternative to existing inversion schemes, particularly for solid-state systems where robustness against numerical noise and periodic boundary conditions is essential.