Comment on "Is there a density-gradient expansion for the electronic exchange energy?"
Abstract
Perdew et al. [arXiv:2610.07201 (2026)] dispute our finding [Phys. Rev. B 113, 115136 (2026)] that the separate exchange (bx) and correlation (bc) contributions to the q2 coefficient of the long-wavelength proper polarization function depend on the infrared regularization of the Coulomb interaction, although their sum bxc=bx+bc does not. We question their claim that the conventional exchange coefficient is uniquely determined. In Qian's [J. Math. Phys. 56, 111901 (2015)] derivation, which Perdew et al. cite in support of their claim of a unique exchange-gradient coefficient, an integral of the form ∫−abdx/x crosses a singularity and does not exist as an ordinary improper integral. Its finite Cauchy principal value (which is a particular way of regularizing the integral) is obtained only by excluding equal intervals on the two sides of the singularity; unequal exclusions give different answers. Independently, Sham's calculation with the screened interaction e−λ0 r/r gives an exchange-gradient coefficient different from the bare-Coulomb calculations of Svendsen--von Barth [Int. J. Quantum Chem. 56, 351 (1995)] and Engel--Vosko [Phys. Rev. B 42, 4940 (1990)]. The discrepancy remains when λ0>0 is so small that 1/λ0 is comparable to the size of the observable universe, making the two interactions experimentally indistinguishable on laboratory scales. Why should the exactly unscreened result be preferred, even if its calculation is accepted as valid? Our calculation shows that the ambiguity of the separate contributions cancels in the combined, regulator-independent coefficient bxc.
Related
- Spin-polarized low-density asymptotics of RPA-screened second-order exchange in the uniform electron gas
- Octupole deformation in even-even Ra isotopes from covariant density functional theory with localized exchange terms in a three-dimensional lattice space
- Comment on: Microscopic signatures of an imaginary charge density wave in a kagome metal
- Pressure Effects on CeMnSi: Evolution of Ce 4f and Mn 3d Electronic States and Negative Thermal Expansion
- Heavy-quark expansion at leading power for local QCD operators in gradient flow
BibTeX
@misc{benites2026comment,
title = {{Comment on "Is there a density-gradient expansion for the electronic exchange energy?"}},
author = {Mario Benites and Angel Rosado and Efstratios Manousakis},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.38943},
eprint = {2610.38943},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.38943},
}
RIS
TY - UNPB
TI - Comment on "Is there a density-gradient expansion for the electronic exchange energy?"
AU - Mario Benites
AU - Angel Rosado
AU - Efstratios Manousakis
PY - 2026
DA - 2026/10/28
PB - arXiv
DO - 10.48550/arxiv.2610.38943
UR - https://arxiv.org/abs/2610.38943
AB - Perdew et al. [arXiv:2610.07201 (2026)] dispute our finding [Phys. Rev. B 113, 115136 (2026)] that the separate exchange (b_x) and correlation (b_c) contributions to the q^2 coefficient of the long-wavelength proper polarization function depend on the infrared regularization of the Coulomb interaction, although their sum b_xc=b_x+b_c does not. We question their claim that the conventional exchange coefficient is uniquely determined. In Qian's [J. Math. Phys. 56, 111901 (2015)] derivation, which Perdew et al. cite in support of their claim of a unique exchange-gradient coefficient, an integral of the form ∫_−a^bdx/x crosses a singularity and does not exist as an ordinary improper integral. Its finite Cauchy principal value (which is a particular way of regularizing the integral) is obtained only by excluding equal intervals on the two sides of the singularity; unequal exclusions give different answers. Independently, Sham's calculation with the screened interaction e^−λ_0 r/r gives an exchange-gradient coefficient different from the bare-Coulomb calculations of Svendsen--von Barth [Int. J. Quantum Chem. 56, 351 (1995)] and Engel--Vosko [Phys. Rev. B 42, 4940 (1990)]. The discrepancy remains when λ_0>0 is so small that 1/λ_0 is comparable to the size of the observable universe, making the two interactions experimentally indistinguishable on laboratory scales. Why should the exactly unscreened result be preferred, even if its calculation is accepted as valid? Our calculation shows that the ambiguity of the separate contributions cancels in the combined, regulator-independent coefficient b_xc.
ER -