Designing explicit functionals for the charge density in terms of a potential

This paper proposes and validates a strategy for constructing explicit functionals that directly map Kohn-Sham potentials to charge densities in inhomogeneous materials using homogeneous electron gas data, successfully demonstrating improved accuracy through increasingly sophisticated approximations without solving the Kohn-Sham Schrödinger equation.

The Gist

The Big Picture: Predicting the Weather Without a Storm Chaser

Imagine you want to know exactly how the air is moving in a specific city (the charge density of a material). In the world of quantum physics, the standard way to do this is like hiring a team of storm chasers to run through every street, measuring wind speed, humidity, and pressure, then feeding all that data into a massive, complex computer simulation (solving the Kohn-Sham equation) to get the answer. It's accurate, but it takes a long time and requires a lot of computing power.

The authors of this paper asked a different question: Can we predict the weather just by looking at the map of the terrain (the potential) without sending out the storm chasers?

They wanted to create a "shortcut formula" (an explicit functional) that takes the shape of the landscape as input and instantly spits out the air movement as output, skipping the complex simulation entirely.

The Problem with Previous Shortcuts

Scientists have tried to write these shortcut formulas before, but they usually failed for complex, bumpy landscapes (like real materials with atoms).

• The Old Way (Thomas-Fermi): This was like saying, "If the ground is flat, the wind is flat." It works okay for a smooth meadow, but if you have a mountain range (like a solid block of helium), this simple guess is way off.
• The Taylor Expansion: This is like trying to predict the whole mountain range by looking at just one small hill and assuming the rest of the world looks exactly like that hill, just shifted slightly. It works for gentle slopes but fails miserably on steep cliffs.

The Solution: The "Connector" Strategy

The authors developed a new strategy called Connector Theory (COT). Here is how it works, using a metaphor:

Imagine you are trying to describe a very bumpy, unique road (your real material). You don't have a map of the whole road. However, you have a perfect, detailed map of a smooth, straight highway (the Homogeneous Electron Gas, or HEG).

1. The Connector: Instead of trying to guess the whole bumpy road from scratch, the authors ask: "If I were driving on my smooth highway, what speed would I need to drive to make the road feel exactly like this specific bumpy spot on my real road?"
2. The Calculation: They use a mathematical tool (the Lindhard function) to find that "speed" (the connector potential) for every single point on the road.
3. The Result: Once they know the "speed" for that spot, they just look up the traffic flow on their smooth highway map for that speed. Because the highway map is perfect, they instantly know the traffic flow for the bumpy spot.

By doing this for every point, they reconstruct the entire traffic flow (charge density) of the bumpy road without ever simulating the bumps directly.

How They Improved the Shortcut

The authors didn't stop at the first guess. They built a hierarchy of increasingly accurate shortcuts:

• Level 1 (Local Potential Approximation): They assumed the road at any point looks exactly like the smooth highway at that exact spot. This was a good start but missed the details of the bumps.
• Level 2 (Linear Response): They added a rule that says, "If the road ahead is steeper, the traffic changes slightly." This helped, but sometimes the math predicted "negative traffic," which is impossible.
• Level 3 (The Connector Fix): This is their big breakthrough. They realized that even if the road is bumpy, the average behavior of the smooth highway can still describe it perfectly if you pick the right "speed" (connector). This method automatically fixed the "negative traffic" errors and made the prediction much more accurate.
• Level 4 (The "Tweaked" Connector): They added a tiny bit of "engineering" (two adjustable numbers) to the formula. Think of this as calibrating a radio. Once they tuned these two numbers using one specific type of rock (cubic helium), the formula worked incredibly well, even when they tested it on rocks that were squished together or pulled apart.

The Results: Testing on "Solid Helium"

To test their idea, they used cubic helium. Why helium? Because it is the ultimate "stress test." It is a material where the electrons are packed very tightly around the atoms, creating a landscape that is extremely bumpy and uneven. It is the worst-case scenario for a shortcut formula.

• The Outcome: Their new "Connector" formulas were able to predict the electron density with high accuracy, even in these extreme conditions.
• The Efficiency: They achieved this without solving the heavy, slow equations that usually take hours. Their method is fast and simple.
• The Transferability: When they took the "tuned" formula (the one with the two adjustable numbers) and applied it to helium that was compressed or expanded, it still worked remarkably well. This suggests the formula is robust and not just a lucky guess for one specific shape.

Why This Matters (According to the Paper)

The paper claims this is a promising new route for materials science.

1. Speed: It allows scientists to calculate properties of materials much faster than current methods.
2. Simplicity: It avoids the need for complex, iterative loops that often get stuck or take forever to solve.
3. Design: Because the formula is "explicit" (a direct equation), it could theoretically be reversed. This means scientists could start with a desired property (like "I want a material that conducts electricity this way") and work backward to find the potential that creates it, essentially "inventing" materials on a computer.

In short, the authors found a way to use a simple, perfect model (the smooth highway) to accurately predict the behavior of a messy, complex reality (the bumpy road), using a clever "connector" to bridge the gap.

Technical Summary

Technical Summary: Designing Explicit Functionals for the Charge Density in Terms of a Potential

Problem Statement
The central challenge addressed in this work is the construction of explicit functionals that map the external (Kohn-Sham) potential directly to the charge density, $n(\mathbf{r}) = n(\mathbf{r}; [v_{KS}])$, without solving the Kohn-Sham Schrödinger equation. While Density Functional Theory (DFT) expresses the total energy as a functional of the density, the density itself is typically obtained by solving the Kohn-Sham equations self-consistently. This process is computationally expensive and obscures the direct relationship between the potential and the observable. Furthermore, inverting the problem to design materials with specific properties is difficult within the standard framework. The authors aim to bypass the solution of the Kohn-Sham equation by utilizing model data—specifically from the Homogeneous Electron Gas (HEG)—to build "potential functionals" that are explicit, computationally efficient, and systematically improvable.

Methodology
The authors propose a framework that combines Taylor expansions of the density functional with Connector Theory (COT). The core strategy involves expanding the real, inhomogeneous system around a homogeneous reference system (the HEG).

1. First-Order Taylor Expansion: The density is expanded around a constant reference potential $v_{0\mathbf{r}}$. The zeroth-order term corresponds to the Thomas-Fermi approximation, while the first-order term utilizes the Lindhard density-density response function $\chi^h_0$ of the HEG.
2. Connector Theory (COT): To improve upon low-order expansions, the authors employ COT. This approach posits that the exact density at a point $\mathbf{r}$ in an inhomogeneous system equals the density of a homogeneous system with a specific "connector" potential $v^c_{\mathbf{r}}$. Instead of solving for the exact $v^c_{\mathbf{r}}$, the method approximates it using the response of the model system.
   • COT1: Uses a linear response approximation to determine the connector potential as a weighted average of the local potential. This implicitly includes higher-order corrections.
   • Optimal Termination (Cauchy-Lagrange): To capture second-order effects without explicitly calculating the complex second-order response function, the authors apply the Cauchy-Lagrange theorem. This involves evaluating the derivative (the response function) at a "midpoint" potential between the reference and the actual potential.
   • COT1-av and COT1-$\alpha$: Since the midpoint potential is inhomogeneous, the authors approximate the response function using COT again.
     • COT1-av: Assumes the second-order response is short-ranged, approximating the potential in the numerator as a simple average.
     • COT1-$\alpha$: Introduces a local weight factor $\alpha_{\mathbf{r}}$ (parameterized as $A(n(\mathbf{r}))^B$) to correct the imbalance introduced by the short-range approximation. These parameters are fitted to a benchmark system.

Key Contributions

• Explicit Potential Functionals: The paper demonstrates a systematic route to derive explicit expressions for the charge density as a functional of the Kohn-Sham potential, avoiding the need to solve the Kohn-Sham equations.
• Integration of Model Data: It successfully adapts the Connector Theory, previously used for exchange-correlation potentials, to the direct calculation of observables (density) using HEG data.
• Systematic Improvement Hierarchy: The authors establish a hierarchy of approximations:
  • Local Potential Approximation (LPA) $\approx$ Thomas-Fermi.
  • Linear Response (COT1).
  • Improved Termination (COT1-av) using Cauchy-Lagrange logic.
  • Parameterized Correction (COT1-$\alpha$).
• Transferability: The work investigates the transferability of the fitted parameters ($A$ and $B$) across different lattice constants (compressed and expanded structures) of cubic helium.

Results
The methods were tested on cubic helium, a prototypical strongly inhomogeneous material, under equilibrium, compressed, and expanded conditions.

• Qualitative Performance: The simple LPA (zeroth order) captures the general shape of the density but fails quantitatively, particularly underestimating density on atoms and overestimating it in low-density regions. First-order linear response (COT1) improves the result but can yield unphysical negative densities in low-density regions.
• Quantitative Improvement: The COT1-av approximation (using the midpoint derivative) significantly improves the density profile, ensuring positivity and better agreement with the benchmark. The COT1-$\alpha$ approximation, with two fitted parameters, achieves excellent agreement with the benchmark density, reducing errors to levels comparable to the Local Density Approximation (LDA) in high-density regions.
• Error Quantification:
  • The Mean Absolute Difference Error (MADE) decreases systematically through the hierarchy. COT1-$\alpha$ reduces the error on the atom to near zero and keeps errors below 5% in significant density regions.
  • The approximations show robust transferability; parameters fitted at equilibrium lattice constant ($a=8.016$ Bohr) perform well for compressed ($a=2.5-4.0$ Bohr) and expanded ($a=15$ Bohr) structures.
• Derived Observables:
  • Electron Number: LPA yields a large error in total electron count ($N_{el} \approx 2.26$ vs. 2). COT1-$\alpha$ reduces this error significantly ($N_{el} \approx 2.29$).
  • Kinetic Energy Functionals: The approximations show systematic improvement for orbital-free kinetic energy functionals (Thomas-Fermi-von Weizsacker and Perdew-Constantin). COT1-$\alpha$ reduces relative errors to $\approx 2-3\%$, demonstrating that the framework captures density gradient physics effectively without being fitted to kinetic energy data.
• Self-Consistency: The approximations behave well within a self-consistent Kohn-Sham loop, suggesting their potential use as preconditioners or for initial steps in iterative calculations.

Significance and Claims
The paper claims that designing explicit functionals for observables directly from the potential is a viable and promising route. By leveraging model data (HEG) and Connector Theory, the authors demonstrate that:

1. Accuracy vs. Cost: It is possible to obtain expressions that are relatively simple to calculate (scaling linearly with system size due to the short-range nature of the response function) while achieving accuracy that systematically improves upon the Thomas-Fermi approximation.
2. Systematic Control: The approach provides a controllable hierarchy of approximations, moving from simple local expressions to those incorporating non-local response effects via the connector concept.
3. Future Potential: While the current work uses the Kohn-Sham potential as input, the authors suggest this framework paves the way for designing functionals of the external potential (bypassing the Kohn-Sham system entirely) and for machine learning applications where transferable parameters can be learned from model systems.

The authors remain modest, acknowledging that while the results are promising for the specific test case of cubic helium, further work is needed to handle non-local pseudopotentials and to refine the treatment of low-density regions where linear response is inherently short-ranged.