pith. sign in

arxiv: 1907.06321 · v1 · pith:5YLNR5VVnew · submitted 2019-07-15 · 🧮 math.NA · cs.NA· math-ph· math.MP· math.OC

Gradient Flow Based Discretized Kohn-Sham Density Functional Theory

Xiaoying Dai , Qiao Wang , Aihui Zhou This is my paper

Pith reviewed 2026-05-24 21:45 UTC · model grok-4.3

classification 🧮 math.NA cs.NAmath-phmath.MPmath.OC
keywords Kohn-Sham density functional theorygradient flowmidpoint discretizationorthogonality preserving iterationlocal minimizernumerical analysisdensity functional theory
0
0 comments X

The pith

A gradient flow model for Kohn-Sham density functional theory has critical points that are local minimizers of the total energy, and its midpoint discretization yields an orthogonality-preserving iteration that converges to such minimizers.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper introduces a gradient flow formulation whose equilibria coincide with local minimizers of the Kohn-Sham total energy. It then discretizes the continuous flow in time via a midpoint rule and extracts an iteration that automatically keeps the orbitals orthogonal. Under the paper's stated assumptions on the energy functional, the iterates are shown to converge to a local minimizer. The construction therefore supplies both a variational justification for gradient-flow ideas and a concrete algorithm that respects the orthogonality constraint without extra projection steps. Numerical tests are reported to illustrate the behavior.

Core claim

We prove that the critical point of the gradient flow based model can be a local minimizer of the Kohn-Sham total energy. Based on the midpoint scheme, the orthogonality preserving iteration scheme produces approximations that are orthogonal and convergent to a local minimizer under reasonable assumptions.

What carries the argument

Midpoint discretization of the gradient flow, which produces an orthogonality-preserving iteration scheme for the Kohn-Sham energy.

If this is right

  • Critical points of the continuous gradient flow are local minimizers of the Kohn-Sham energy.
  • The midpoint scheme approximates those critical points to any desired accuracy for small enough time steps.
  • The derived iteration automatically maintains orthogonality of the approximate orbitals at every step.
  • Under the stated assumptions the iteration converges to a local minimizer of the total energy.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same discretization strategy could be applied to other constrained variational problems that arise in electronic structure calculations.
  • Automatic orthogonality preservation may allow existing DFT codes to drop separate orthogonalization routines.
  • The convergence theory supplies a principled way to select time-step sizes in related flow-based solvers.

Load-bearing premise

The energy functional or discretization parameters satisfy reasonable assumptions that guarantee convergence of the midpoint iteration to a local minimizer.

What would settle it

An explicit example, analytic or numerical, in which the midpoint iterates either lose orthogonality or fail to approach a local minimizer while still obeying the paper's assumptions on the functional.

Figures

Figures reproduced from arXiv: 1907.06321 by Aihui Zhou, Qiao Wang, Xiaoying Dai.

Figure 1
Figure 1. Figure 1 [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 6
Figure 6. Figure 6 [PITH_FULL_IMAGE:figures/full_fig_p030_6.png] view at source ↗
Figure 6
Figure 6. Figure 6 [PITH_FULL_IMAGE:figures/full_fig_p031_6.png] view at source ↗
Figure 6
Figure 6. Figure 6 [PITH_FULL_IMAGE:figures/full_fig_p032_6.png] view at source ↗
Figure 6
Figure 6. Figure 6 [PITH_FULL_IMAGE:figures/full_fig_p033_6.png] view at source ↗
Figure 6
Figure 6. Figure 6 [PITH_FULL_IMAGE:figures/full_fig_p034_6.png] view at source ↗
Figure 6
Figure 6. Figure 6 [PITH_FULL_IMAGE:figures/full_fig_p035_6.png] view at source ↗
read the original abstract

In this paper, we propose and analyze a gradient flow based Kohn-Sham density functional theory. First, we prove that the critical point of the gradient flow based model can be a local minimizer of the Kohn-Sham total energy. Then we apply a midpoint scheme to carry out the temporal discretization. It is shown that the critical point of the Kohn-Sham energy can be well-approximated by the scheme. In particular, based on the midpoint scheme, we design an orthogonality preserving iteration scheme to minimize the Kohn-Sham energy and show that the orthogonality preserving iteration scheme produces approximations that are orthogonal and convergent to a local minimizer under reasonable assumptions. Finally, we report numerical experiments that support our theory.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes a gradient flow formulation for Kohn-Sham density functional theory. It proves that critical points of the continuous flow are local minimizers of the KS total energy, applies a midpoint temporal discretization to obtain an orthogonality-preserving iteration scheme, shows that this scheme produces orthogonal approximations convergent to a local minimizer under reasonable assumptions on the energy functional and discretization parameters, and presents numerical experiments supporting the theory.

Significance. If the proofs are rigorous, the work supplies a discretization method for KS-DFT that inherits orthogonality preservation and convergence guarantees from the continuous gradient flow, addressing a practical need in electronic structure computations. The explicit flagging of assumptions and the inclusion of both analysis and experiments are positive features.

major comments (2)
  1. [Proof of local minimizer property (likely §3)] The central claim that critical points of the gradient flow are local minimizers (abstract) rests on properties of the KS energy functional; the specific conditions (e.g., local convexity or coercivity assumptions) must be stated explicitly in the proof section, as they are load-bearing for the subsequent discretization analysis.
  2. [Convergence theorem for the iteration scheme (likely §5)] The convergence of the midpoint-based orthogonality-preserving iteration to a local minimizer (abstract, final paragraph) is asserted under 'reasonable assumptions' on the functional and discretization parameters; these assumptions need to be listed and their necessity justified with a concrete statement of the theorem, because they directly determine whether the scheme is practically useful.
minor comments (2)
  1. [Abstract] The abstract packs multiple distinct results into a single paragraph; splitting it would improve readability.
  2. [Introduction] Notation for the KS energy functional and the gradient flow should be introduced with a clear definition before its first use in the main text.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive report and positive assessment of the manuscript's contributions. We address each major comment below and will revise the manuscript to improve the explicitness of the stated assumptions.

read point-by-point responses

  1. Referee: [Proof of local minimizer property (likely §3)] The central claim that critical points of the gradient flow are local minimizers (abstract) rests on properties of the KS energy functional; the specific conditions (e.g., local convexity or coercivity assumptions) must be stated explicitly in the proof section, as they are load-bearing for the subsequent discretization analysis.

    Authors: We agree that the conditions on the KS energy functional need to be stated more explicitly. In the revised manuscript we will insert a new paragraph at the beginning of the proof in §3 that lists the precise assumptions (local convexity of the energy in a neighborhood of the critical point together with a coercivity condition ensuring the Hessian is positive definite on the tangent space). These assumptions will be cross-referenced in the discretization analysis so that their role is transparent. revision: yes

  2. Referee: [Convergence theorem for the iteration scheme (likely §5)] The convergence of the midpoint-based orthogonality-preserving iteration to a local minimizer (abstract, final paragraph) is asserted under 'reasonable assumptions' on the functional and discretization parameters; these assumptions need to be listed and their necessity justified with a concrete statement of the theorem, because they directly determine whether the scheme is practically useful.

    Authors: We accept that the phrase 'reasonable assumptions' is insufficiently precise. In the revised §5 we will replace the informal statement with a fully explicit theorem that enumerates every assumption on the energy functional (C^2 smoothness, uniform bounds on second derivatives near the minimizer) and on the discretization parameters (step-size restriction relative to the Lipschitz constant of the gradient). A short remark following the theorem will justify the necessity of each assumption by pointing to the corresponding step in the proof. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation is self-contained

full rationale

The paper's central claims consist of mathematical proofs: that critical points of the proposed gradient flow correspond to local minimizers of the Kohn-Sham energy, that the midpoint discretization approximates these points, and that the resulting orthogonality-preserving iteration converges to a local minimizer under explicitly stated assumptions on the functional and discretization parameters. These steps are established directly from the model definitions and standard analytic techniques rather than by fitting parameters, renaming known results, or reducing to self-citations whose content is itself unverified. The argument chain is therefore independent of its own outputs.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Analysis depends on standard variational and discretization assumptions typical of numerical DFT; no free parameters, new entities, or ad-hoc axioms are introduced in the abstract.

axioms (2)
  • domain assumption Critical points of the gradient flow correspond to stationary points of the Kohn-Sham energy under the stated functional setting
    Invoked to prove that critical points are local minimizers.
  • domain assumption The midpoint scheme preserves the necessary structural properties (orthogonality) for convergence
    Required for the iteration scheme to remain well-defined and convergent.

pith-pipeline@v0.9.0 · 5659 in / 1357 out tokens · 22141 ms · 2026-05-24T21:45:15.222226+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

33 extracted references · 33 canonical work pages

  1. [1]

    Antony, K.N

    B.K. Antony, K.N. Joshipura, N.J. Mason, and J. Tennyson , R-matrix calculation of low- energy electron collisions with LiH , J. Phys. B: Atomic Molecular Opt. Phys., 37 (2004), pp. 1689–1697

    work page 2004

  2. [2]

    Bao and Q

    W. Bao and Q. Du , Computing the ground state solution of Bose–Einstein condensates by a 35 normalized gradient flow, SIAM J. Sci. Comput., 25 (2004), pp. 1674–1697

    work page 2004

  3. [3]

    Le Bris, Computational chemistry from the perspective of numerical analysis, Acta Numer- ica, 14 (2005), pp

    C. Le Bris, Computational chemistry from the perspective of numerical analysis, Acta Numer- ica, 14 (2005), pp. 363–444

    work page 2005

  4. [4]

    Y. Cai, L. Zhang, Z. Bai, and R. Li , On an eigenvector-dependent nonlinear eigenvalue problem, SIAM J. Matrix Anal. Appl., 39 (2018), pp. 1360–1382

    work page 2018

  5. [5]

    H. Chen, X. Dai, X. Gong, L. He, and A. Zhou , Adaptive finite element approximations for kohn-sham models, Multiscale Model. Simul., 12 (2013), pp. 1828–1869

    work page 2013

  6. [6]

    T. Chen, Y. Hua, and W. Yan , Global convergence of Oja’s subspace algorithm for principal component extraction, IEEE T. Neural. Networ., 9 (1998), pp. 58–67

    work page 1998

  7. [7]

    X. Dai, X. Gong, Z. Yang, D. Zhang, and A. Zhou , Finite volume discretizations for eigen- value problems with applications to electronic structure calculations , Multiscale Model. Sim., 9 (2011), pp. 208–240

    work page 2011

  8. [8]

    X. Dai, Z. Liu, L. Zhang, and A. Zhou , A conjugate gradient method for electronic structure calculations, SIAM J. Sci. Comput., 39 (2017), pp. A2702–A2740

    work page 2017

  9. [9]

    Dai and A

    X. Dai and A. Zhou , Three-scale finite element discretizations for quantum eigenvalue prob- lems, SIAM J. Numer. Anal., 46 (2008), pp. 295–324

    work page 2008

  10. [10]

    Edelman, T

    A. Edelman, T. A. Arias, and S. T. Smith , The geometry of algorithms with orthogonality constraints, SIAM J. Matrix Anal. Appl., 20 (1998), pp. 303–353

    work page 1998

  11. [11]

    Fournais, M

    S. Fournais, M. Hoffmann-Ostenhof, T. Hoffmann-Ostenhof, and T. Ø. Sørensen , Positivity of the spherically averaged atomic one-electron density , Math. Z., 259 (2008), pp. 123–130

    work page 2008

  12. [12]

    J. B. Francisco, J. M. Martınez, and L. Martınez , Globally convergent trust-region meth- ods for self-consistent field electronic structure calculations , J. Chem. Phys., 121 (2004), pp. 10863–10878

    work page 2004

  13. [13]

    Hahn, Stability of Motion , vol

    W. Hahn, Stability of Motion , vol. 138, Springer, 1967

    work page 1967

  14. [14]

    Helmke and J

    U. Helmke and J. B. Moore , Optimization and Dynamical Systems , Springer Science & Business Media, 2012

    work page 2012

  15. [15]

    Kobus, D

    J. Kobus, D. Moncrieff, and S. Wilson , Comparison of the polarizabilities and hyperpolar- izabilities obtained from finite basis set and finite difference Hartree-Fock calculations for diatomic molecules, J. Phys. B: Atomic, Molecular Opt. Phys., 34 (2001), pp. 5127–5143

    work page 2001

  16. [16]

    X. Liu, X. Wang, Z. Wen, and Y. Yuan , On the convergence of the self-consistent field iteration in Kohn–Sham density functional theory , SIAM J. Matrix Anal. A., 35 (2014), pp. 546–558

    work page 2014

  17. [17]

    X. Liu, Z. Wen, X. Wang, M. Ulbrich, and Y. Yuan , On the analysis of the discretized Kohn–Sham density functional theory , SIAM J. Numer. Anal., 53 (2015), pp. 1758–1785

    work page 2015

  18. [18]

    Martin, Electronic Structure: Basic Theory and Practical Methods , Cambridge University Press, 2004

    R. Martin, Electronic Structure: Basic Theory and Practical Methods , Cambridge University Press, 2004

    work page 2004

  19. [19]

    R. G. Parr and W. Yang , Density-Functional Theory of Atoms and Molecules , OUP USA, 1994

    work page 1994

  20. [20]

    J. P. Perdew and A. Zunger, Self-interaction correction to density-functional approximations for many-electron systems, Phys. Rev. B, 23 (1981), pp. 5048–5079

    work page 1981

  21. [21]

    PHG, http://lsec.cc.ac.cn

    work page

  22. [22]

    Y Saad, J. R. Chelikowsky, and S.M. Suzanne , Numerical methods for electronic structure calculations of materials , SIAM Rev., 52 (2010), pp. 3–54

    work page 2010

  23. [23]

    Schneider, T

    R. Schneider, T. Rohwedder, A. Neelov, and J. Blauert , Direct minimization for calcu- lating invariant subspaces in density functional computations of the electronic structure , J. Comput. Math., (2009), pp. 360–387

    work page 2009

  24. [24]

    Thøgersen, J

    L. Thøgersen, J. Olsen, D. Yeager, P. Jørgensen, P. Sa lek, and T. Helgaker , The trustregion self-consistent field method: Towards a black-box optimization in Hartree-Fock and Kohn-Sham theories , J. Chem. Phys., 121 (2004), pp. 16–27

    work page 2004

  25. [25]

    Udriste, Convex Functions and Optimization Methods on Riemannian Manifolds , vol

    C. Udriste, Convex Functions and Optimization Methods on Riemannian Manifolds , vol. 297, Springer Science & Business Media, 1994

    work page 1994

  26. [26]

    Ulbrich, Z

    M. Ulbrich, Z. Wen, C. Yang, D. Klockner, and Z. Lu , A proximal gradient method for ensemble density functional theory , SIAM J. Sci. Comput., 37 (2015), pp. A1975–A2002

    work page 2015

  27. [27]

    Vecharynski, C

    E. Vecharynski, C. Yang, and J. E. Pask , A projected preconditioned conjugate gradient algorithm for computing many extreme eigenpairs of a Hermitian matrix , J. Comput. Phys., 290 (2015), pp. 73–89

    work page 2015

  28. [28]

    Wen and W

    Z. Wen and W. Yin, A feasible method for optimization with orthogonality constraints , Math. Program. Ser. A., 142 (2013), pp. 397–434

    work page 2013

  29. [29]

    W. Yan, U. Helmke, and J. B. Moore , Global analysis of Oja’s flow for neural networks , IEEE T. Neural. Networ., 5 (1994), pp. 674–683

    work page 1994

  30. [30]

    C. Yang, W. Gao, and J. C. Meza , On the convergence of the self-consistent field iteration 36 for a class of nonlinear eigenvalue problems , SIAM J. Matrix Anal. Appl., 30 (2009), pp. 1773–1788

    work page 2009

  31. [31]

    C. Yang, J. C. Meza, and L. Wang, A trust region direct constrained minimization algorithm for the Kohn–Sham equation , SIAM J. Sci. Comput., 29 (2007), pp. 1854–1875

    work page 2007

  32. [32]

    Zhang, J

    X. Zhang, J. Zhu, Z. Wen, and A. Zhou , Gradient type optimization methods for electronic structure calculations, SIAM J. Sci. Comput., 36 (2014), pp. C265–C289

    work page 2014

  33. [33]

    Z. Zhao, Z. Bai, and X. Jin , A riemannian newton algorithm for nonlinear eigenvalue prob- lems, SIAM J. Marix Anal. A., 36 (2015), pp. 752–774. 37

    work page 2015