Approximate normalizations for approximate density functionals

This paper demonstrates that violating the standard electron number normalization in approximate density functional calculations can significantly improve energy accuracy, providing explicit one-dimensional corrections and generalizing them to arbitrary cavities via Weyl asymptotics.

The Gist

The Big Idea: Breaking the Rules to Win the Game

Imagine you are trying to bake the perfect cake. You have a recipe (a mathematical formula) that tells you exactly how much flour, sugar, and eggs to use. In the world of quantum physics, this "recipe" is called Density Functional Theory (DFT). It's the most popular tool scientists use to predict how atoms and molecules behave.

For 60 years, everyone has believed one "Golden Rule" about this recipe: The amount of "electron dough" in your cake must exactly equal the number of electrons you are trying to model. If you are baking a cake for 10 people, you must use exactly 10 cups of dough. If you use 9.9 or 10.1, the recipe is considered broken.

This paper says: "Actually, sometimes the best cake comes from using a slightly wrong amount of dough."

The authors discovered that if you intentionally use a slightly incorrect amount of electrons in your calculation (a "fake" number), the final result for the energy of the system becomes much more accurate than if you strictly followed the rule.

---

The Analogy: The Noisy Crowd in a Stadium

To understand why this works, let's use an analogy of a stadium filled with people (electrons).

The Standard Approach (The "Exact" Count):
Imagine you are trying to estimate the noise level of a crowd. The standard way is to count every single person and assume they are all standing perfectly still in their assigned seats. You do the math based on the exact number of people ($N$).

• The Problem: In reality, people wiggle, shift, and move. If you force your math to assume they are perfectly still and exactly $N$ people, your prediction of the noise level will be off. The "wiggles" create errors that the simple math can't handle.

The Authors' Approach (The "Approximate" Count):
The authors say, "Let's pretend there are slightly more (or fewer) people in the stadium than there actually are."

• Imagine you tell your math, "Okay, let's assume there are $N + 0.5$ people."
• By adding this tiny "ghost" person, the math smooths out the wiggles. It averages out the chaos.
• The Result: Even though you are lying about the number of people, your prediction of the total noise level (the energy) is actually closer to the truth than if you had tried to count the real people perfectly.

Why Does This Happen? (The "Edge" Problem)

The paper explains that the errors in these calculations usually happen at the edges (like the walls of a box or the surface of an atom).

• The Wall Effect: Think of a crowd of people in a small room. In the middle of the room, everyone is comfortable. But right against the walls, people get squished and confused.
• The Math Glitch: The standard formulas are great at describing the middle of the room but terrible at the walls. They get the "bulk" right but the "edge" wrong.
• The Fix: By slightly changing the total number of electrons (the normalization), the authors are essentially telling the math to "ignore the messy edge details" and focus on the smooth, average behavior of the crowd. It's like blurring the edges of a photo to make the center look sharper.

Real-World Examples from the Paper

The authors tested this on several scenarios:

1. The Particle in a Box: Imagine a tiny ball bouncing in a box.

   • Standard Math: Predicts the energy with a big error (sometimes off by 50%!).
   • Their Fix: They added a tiny fraction of an electron (0.5) to the count. Suddenly, the prediction was almost perfect.

2. Atoms (The Noble Gases): They looked at real atoms like Neon and Argon.

   • Standard Math: Underestimates the energy significantly.
   • Their Fix: By adjusting the electron count based on the size of the atom, they reduced the error from huge percentages down to tiny fractions of a percent.

3. The "Ghost" Electron: In some cases, they found that the best result came from adding a "ghost" electron that doesn't actually exist. It's a mathematical trick, not a physical reality, but it makes the numbers work better.

Why Should You Care?

This isn't just a math puzzle; it changes how we do science.

• Better Drug Design: If we can calculate the energy of molecules more accurately, we can design better medicines faster.
• New Materials: Engineers can design stronger, lighter materials for cars and planes without needing super-expensive supercomputers.
• The "Density-Corrected" Revolution: There is a popular method called "Density-Corrected DFT" where scientists try to fix errors by using the exact electron density. This paper shows that fixing the number of electrons is actually easier and often better than trying to find the exact density.

The Takeaway

For decades, scientists thought, "To get the right answer, I must follow the rules perfectly."

This paper says, "Sometimes, to get the right answer, you have to break the rules just a little bit." By intentionally using a slightly "wrong" number of electrons, they smooth out the messy mathematical errors at the edges, resulting in a much more accurate picture of the universe.

It's a bit like tuning a radio: sometimes, you have to turn the dial slightly off the station to get the clearest signal.

Technical Summary

1. Problem Statement

In standard Density Functional Theory (DFT), a fundamental axiom is that the electron density $n(\mathbf{r})$ must integrate to the exact number of electrons $N$ in the system:
$$\int d\mathbf{r} \, n(\mathbf{r}) = N$$
This constraint is applied even when using approximate exchange-correlation functionals (e.g., Thomas-Fermi, LDA). The authors challenge this dogma, proposing that for approximate functionals, enforcing the exact normalization $N$ is often suboptimal. They argue that minimizing the energy functional under the constraint $\int n(\mathbf{r}) = N$ yields less accurate energies than minimizing it under a slightly modified normalization $\int n(\mathbf{r}) = \tilde{N}$, where $\tilde{N} = N + \Delta N$.

The core problem addressed is how to determine the optimal correction $\Delta N$ for approximate functionals to improve energetic accuracy, particularly in the semiclassical limit, without requiring the exact density.

2. Methodology

The authors employ semiclassical asymptotics to derive the correction term $\Delta N$. Their approach involves:

• Thomas-Fermi (TF) Approximation: Using the TF kinetic energy functional as the baseline approximate functional.
• WKB Theory (1D): In one dimension, they utilize the Wentzel-Kramers-Brillouin (WKB) approximation to relate energy levels to the density. By analyzing the remainder terms in the WKB quantization condition, they derive a correction to the particle number.
• Weyl Asymptotics (Higher Dimensions): For $d$-dimensional cavities and general potentials, they apply Weyl's law, which describes the asymptotic distribution of eigenvalues for the Laplacian. The total energy $E(N)$ is expanded in powers of $N$:
  $$E(N) = C_1 N^{1+2/d} + C_2 N^{1+1/d} + \dots$$
  The standard TF approximation captures only the leading term ($C_1$). The authors show that by shifting the normalization to $\tilde{N} = N + \Delta N$, the approximate energy $\tilde{E}(\tilde{N})$ can recover the second term ($C_2$) of the asymptotic expansion, significantly improving accuracy.
• General Formulation: They propose a universal form for the correction:
  $$\Delta N = B N^q$$
  where the exponent $q$ is universal ($q = (d-1)/d$ for cavities), and the coefficient $B$ depends on the geometry (e.g., surface-to-volume ratio) and the nature of the potential.

3. Key Contributions

• Violation of Normalization Principle: The paper demonstrates that violating the standard normalization constraint ($\int n = N$) in favor of an "approximate normalization" ($\int n = N + \Delta N$) yields better energy results than using the exact density with the same approximate functional.
• Derivation of $\Delta N$:
  • 1D: Derived $\Delta N = 1/2 - \nu$, where $\nu$ is the Maslov index (e.g., $\Delta N = 1/2$ for a particle in a box with hard walls).
  • Higher Dimensions: Derived $\Delta N$ using Weyl asymptotics. For a $d$-dimensional cavity, $\Delta N \propto N^{(d-1)/d}$.
  • Specific Cases: Provided explicit formulas for 2D/3D boxes, circular cavities, harmonic oscillators, and Coulomb potentials.
• Application to Realistic Systems:
  • Interacting Electrons: Applied the method to neutral atoms ($Z=N$) to recover the Scott correction (the $N^{5/3}$ term in the energy expansion) within the Thomas-Fermi framework.
  • Exchange Energy: Applied the concept to the Local Density Approximation for Exchange (LDA-X), showing that scaling the density by a factor $(1 + \Delta N/N)$ significantly reduces errors in exchange energy calculations for noble gases.
• Comparison with Density-Corrected DFT (DC-DFT): The authors show that their normalization correction is often more accurate than DC-DFT (which uses the exact density) and is computationally much cheaper to implement.

4. Key Results

• 1D Particle in a Box: The normalization-corrected TF energy ($\tilde{E}_{nc}$) with $\Delta N = 1/2$ outperforms both the standard TF energy and the TF energy evaluated on the exact density. The error drops from ~50% to negligible levels as $N$ increases.
• 3D Box: For a 3D box with incommensurate edges, the standard TF error is ~15% for $N=1000$. The normalization-corrected version reduces this to 0.5%.
• Circular Cavity (2D): For 19 electrons, the standard TF error is -26%, while the corrected version reduces the error to -2%.
• Noble Gases (Interacting Systems):
  • Total Energy: For heavy atoms like Radon (Rn), the standard TF error is -15%, while the corrected version (incorporating the Scott correction via $\Delta N$) reduces the error to 1.5%.
  • Exchange Energy: For noble gases, applying the normalization correction to LDA-X reduces the error from ~4-16% (standard LDA) to 0.1-5%, rivaling more sophisticated functionals like PBE.
• Universality: The exponent $q$ in $\Delta N = B N^q$ is found to be universal ($q = (d-1)/d$) across different cavity shapes and dimensions, while $B$ scales with geometric properties like the perimeter-to-area ratio.

5. Significance and Outlook

• Theoretical Insight: The work bridges the gap between semiclassical physics (Weyl/WKB) and practical DFT, showing that "approximate" functionals are mathematically consistent with "approximate" densities that do not integrate to the exact particle number.
• Practical Utility: The method offers a simple, low-cost way to improve the accuracy of orbital-free DFT and other approximate functionals without needing to solve for Kohn-Sham orbitals or use exact densities.
• Future Directions: The authors suggest that deriving a universal functional form for $\Delta N[v(\mathbf{r})]$ based on phase-space integrals could lead to a new generation of highly accurate, size-extensive functionals. They also highlight the potential for applying these corrections to periodic boundary conditions (solids) and the exchange-correlation functional in general Kohn-Sham DFT.

In summary, the paper overturns a long-held assumption in DFT, demonstrating that intentionally mis-normalizing an approximate density to match the asymptotic behavior of the energy spectrum yields superior energetic results, effectively capturing higher-order corrections (like the Scott correction) that standard approximations miss.