Extending Nonlocal Kinetic Energy Density Functionals to Isolated Systems via a Density-Functional-Dependent Kernel

This paper resolves the Blanc-Cances instability of the Wang-Teter nonlocal kinetic energy density functional in isolated systems by introducing a density-functional-dependent kernel, thereby achieving an order-of-magnitude accuracy improvement for single atoms while maintaining superior performance in bulk metals.

The Gist

The Big Picture: Simulating Matter Without the "Heavy Lifting"

Imagine you are trying to predict how a crowd of people will move through a city.

• The Old Way (Kohn-Sham DFT): You track every single person individually, calculating their exact path, who they bump into, and where they stop. It's incredibly accurate, but if you have a million people (electrons), it takes a supercomputer years to finish the calculation.
• The New Way (Orbital-Free DFT): Instead of tracking individuals, you just look at the "density" of the crowd. Where is it thick? Where is it thin? You use a set of rules (a formula) to guess how the crowd moves based on that density alone. This is super fast—like checking a traffic map instead of following every car.

The Problem: For decades, the "fast" method worked great for solid blocks of metal (like a steel beam), but it completely broke down when applied to isolated things, like a single atom floating in space. It was like a traffic algorithm that worked perfectly on a highway but told you a single parked car had infinite energy and would explode.

The Villain: The "Blanc-Cancès Instability"

The paper identifies a specific bug in the math used for these fast calculations, called the Blanc-Cancès instability.

The Analogy:
Imagine you are trying to calculate the average height of people in a room to set the height of a doorframe.

• In a crowded stadium (Bulk Systems): You take the average height of everyone. It's a stable, reliable number.
• In an empty room with one giant (Isolated Systems): The old math tried to calculate the "average" by pretending the room was infinite and the giant was spread out over the whole universe. This created a nonsense number. Because the math was based on a fake average, the calculation got confused. It started saying, "If I make this atom smaller, the energy goes down... and down... and down." It predicted the atom would collapse into nothingness with negative infinite energy.

This happened because the old formula used a rigid, fixed average (like a ruler that doesn't change length) even when the object being measured changed size.

The Solution: A "Smart Ruler"

The authors (Liang Sun and Mohan Chen) fixed this by inventing a density-functional-dependent kernel.

The Analogy:
Instead of using a rigid, fixed ruler (the old average), they built a smart, stretchy ruler.

• This new ruler looks at the crowd right now. If the crowd is dense, the ruler stretches one way. If the crowd is sparse, it stretches another.
• Crucially, this ruler obeys the laws of physics: if you shrink the whole system down, the ruler shrinks with it perfectly. It never gets confused.

By making the "average" depend on the actual shape of the electron cloud rather than a fake, fixed number, they stopped the math from exploding.

The Results: Fast, Stable, and Accurate

The team tested their new formula (called ext-WT) on 56 different elements, from Hydrogen to Zinc.

1. Stability: The "explosion" is gone. The math now behaves nicely for single atoms, just like it did for big blocks of metal.
2. Accuracy:
   • The old fast method (WT) was about 20 times less accurate than the gold standard for single atoms.
   • The new method (ext-WT) is 20 times more accurate than the old fast method. It is now almost as good as the slow, heavy-lifting method, but still runs at lightning speed.
3. Universality: It works for everything. Whether you are simulating a single atom in a vacuum or a chunk of aluminum metal, the same formula works perfectly.

Why This Matters

Think of this as upgrading the GPS in your car.

• Before: The GPS worked great on the highway (bulk metals) but would tell you to drive off a cliff if you tried to navigate a single driveway (isolated atoms).
• Now: The new GPS understands both the highway and the driveway. It's fast, it doesn't crash, and it gives you the right directions everywhere.

This breakthrough allows scientists to simulate complex materials—like designing new lightweight alloys for airplanes or understanding how matter behaves under extreme heat—much faster and more accurately than ever before, without needing to wait weeks for a computer to finish the job.

Technical Summary

1. Problem Statement

The paper addresses a critical limitation in Orbital-Free Density Functional Theory (OFDFT): the failure of nonlocal Kinetic Energy Density Functionals (KEDFs), specifically the Wang-Teter (WT) family, when applied to isolated systems (e.g., atoms, molecules).

• The BC Instability: While WT-like functionals perform well for bulk materials (uniform electron gas), they exhibit a "Blanc-Cancès (BC) instability" in isolated systems. This manifests as the total energy becoming unbounded from below, rendering calculations unstable.
• Root Cause: The authors identify the root cause as the use of an ill-defined average charge density ($\rho_{avg}$) in the kernel. In density-independent kernels, $\rho_{avg}$ is a rigid spatial average that does not scale correctly under uniform density scaling ($\rho_\sigma(r) = \sigma^3\rho_1(\sigma r)$).
• Consequences: This rigidity causes two fundamental violations:
  1. Violation of Scaling Laws: The functional fails to satisfy the exact scaling relation $T_s[\rho_\sigma] = \sigma^2 T_s[\rho_1]$.
  2. Negative Pauli Energy: The mismatch between the fixed $\rho_{avg}$ and the physical density $\rho(r)$ generates unphysical negative Pauli energy ($T_\theta$), contradicting the requirement that Pauli energy must be positive.

2. Methodology

To resolve the BC instability while preserving the computational efficiency of nonlocal functionals, the authors propose a rigorous reformulation of the kernel.

• Density-Functional-Dependent Kernel: Instead of a density-independent kernel $w(k_F^0, |r-r'|)$ based on a fixed average density, they introduce a kernel $w(k_F[\rho], r-r')$ where the Fermi wave vector $k_F$ depends on a functional of the density, $\zeta[\rho]$.
  $$k_F[\rho] = (3\pi^2 \zeta[\rho])^{1/3}$$
• Design Principles for $\zeta[\rho]$: The functional $\zeta[\rho]$ is constructed to satisfy three criteria:
  1. UEG Recovery: It must recover the average density $\rho_{avg}$ in the limit of a uniform electron gas.
  2. Scaling Covariance: It must scale correctly: $\zeta[\rho_\sigma] = \sigma^3 \zeta[\rho_1]$.
  3. Positivity Enforcement: It must ensure the Pauli energy remains non-negative by satisfying $\zeta[\rho] \geq \rho_c$, where $\rho_c$ is a derived characteristic density threshold.
• Proposed Ansatz: The authors propose a specific form for $\zeta[\rho]$ involving a ratio of density moments:
  $$\zeta[\rho] = \frac{\int \rho^{\kappa+1}(r) d^3r}{\int \rho^\kappa(r) d^3r}$$
  The parameter $\kappa$ is analytically determined (without empirical fitting) by matching the exact solution of the Hydrogen atom to a uniform density distribution while preserving the average nucleus-electron separation. This yields $\kappa \approx 0.832$.

3. Key Contributions

• Resolution of BC Instability: The paper provides a rigorous theoretical proof that replacing the rigid $\rho_{avg}$ with the scaling-covariant $\zeta[\rho]$ eliminates the BC instability, ensuring the total energy is bounded from below.
• Preservation of Formal Properties: The new ext-WT KEDF maintains the formal exactness of the original framework, including:
  • Correct adherence to the scaling law ($T[\rho_\sigma] = \sigma^2 T[\rho]$).
  • Preservation of the Lindhard linear response behavior for homogeneous systems.
  • Guarantee of positive Pauli energy.
• Computational Efficiency: The method retains the $O(N \ln N)$ computational scaling of the original WT functional. The additional terms introduced by the density-dependent kernel are efficiently evaluated using Fast Fourier Transforms (FFT), avoiding the cubic scaling or empirical tuning required by previous adaptations (like HC or LDAK-X).
• Parameter-Free Universality: Unlike previous nonlocal functionals that required system-specific empirical tuning, the ext-WT KEDF uses a single, analytically derived parameter ($\kappa$), making it universally applicable.

4. Results

The authors benchmarked the ext-WT KEDF against 56 single-atom systems (using bare Coulomb potentials, BLPS, and HQLPS) and bulk metals (Li, Mg, Al).

• Isolated Systems (Single Atoms):

  • Accuracy: The ext-WT KEDF achieved a Mean Absolute Relative Error (MARE) of 1.8% for total energies, representing a 20-fold improvement over the original WT KEDF (38.7%) and outperforming semilocal functionals like GE2 (7.4%) and LKT (7.6%).
  • Charge Density: It achieved the lowest Mean Absolute Error (MAE) for charge densities ($2.7 \times 10^{-4}$ a.u.).
  • Physical Features: The ext-WT KEDF correctly satisfied Kato's nuclear cusp condition and reproduced the positive-definite Pauli potential profiles, which were unphysical (negative or oscillatory) in WT and semilocal methods.
  • Stability: The total energy remained constant with respect to simulation cell size ($L$), confirming the elimination of the pathological $L$-dependence found in WT.

• Bulk Systems:

  • The ext-WT KEDF preserved the superior accuracy of the original WT method for bulk metals (Li, Mg, Al), outperforming semilocal functionals by an order of magnitude in predicting bulk moduli, equilibrium volumes, and energies.

5. Significance

This work bridges a long-standing gap in OFDFT, enabling the use of highly accurate nonlocal functionals for both isolated systems (molecules, atoms) and bulk materials within a single, unified framework.

• Theoretical Impact: It resolves the fundamental theoretical conflict between nonlocal kernels and scaling laws in finite systems, providing a rigorous path forward for nonlocal functional development.
• Practical Impact: By achieving near-linear scaling ($O(N \ln N)$) with high accuracy across diverse electronic environments, the ext-WT KEDF significantly expands the applicability of OFDFT to large-scale simulations of heterogeneous materials, liquid metals, and warm dense matter without the prohibitive cost of Kohn-Sham DFT ($O(N^3)$).
• Generalizability: The strategy of constructing density-functional-dependent kernels to enforce scaling and positivity offers a generalizable design principle for future kinetic energy functionals.