Jack Wetherell
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Jack Wetherell has a total of 24 co-authors. Here are the top co-authors by number of shared publications:
- Matthew J. P. Hodgson
Department of Physics, Durham University, South Road, Durham DH1 3LE, United Kingdom
10 shared publications R. W. Godby
Department of Physics, University of York, Heslington, York YO10 5DD, United Kingdom
3 shared publications- Abdul Razak Elmaslmane
Department of Physics, University of York, Heslington, York YO10 5DD, United Kingdom
1 shared publication Andrea Costamagna
1 shared publication- Andrea Costamagna
EPFL, Lausanne, Vaud, Switzerland
1 shared publication - Bradley Longstaff
Department of Mathematics, Imperial College London, London SW7 2AZ, United Kingdom
1 shared publication
Latest Publications
- Open Access
- 19 June 2026
- Published by Springer Nature in npj Computational Materials
- Vol. 12 (1)
Addressing hypersensitivity in density functional theory for reliable machine learning
In principle, machine learning (ML) can be used within density functional theory to obtain any electronic property of a time-independent many-body system from its ground-state electron density alone. However, when the target quantity (the output) is highly sensitive to small variations in the density (the input), accurately learning the mapping between them is highly challenging in practice. We identify this ‘hypersensitivity’ in two classes of density functionals that contain observables central to most electronic structure calculations: (1) any observable that varies subject to a global constant shift in the external potential (such as the total energy); (2) any observable of the N -electron system that depends on the properties of the ( N + 1)-electron system (such as the electron affinity). We demonstrate that the accuracy with which widely used ML models approximate density functionals in these two classes is significantly limited by the pathological nature of the mapping from the density to the observable, even when the reference level of the external potential is fixed to prevent constant shifts. In addition, we find that providing the external potential as input to the ML model resolves this issue, in some cases reducing the error by up to three orders of magnitude when predicting quantities in these two classes.
- 19 June 2026
- Published by Springer Nature in npj Computational Materials
- Vol. 12 (1)
Open AccessAddressing hypersensitivity in density functional theory for reliable machine learning
In principle, machine learning (ML) can be used within density functional theory to obtain any electronic property of a time-independent many-body system from its ground-state electron density alone. However, when the target quantity (the output) is highly sensitive to small variations in the density (the input), accurately learning the mapping between them is highly challenging in practice. We identify this ‘hypersensitivity’ in two classes of density functionals that contain observables central to most electronic structure calculations: (1) any observable that varies subject to a global constant shift in the external potential (such as the total energy); (2) any observable of the N -electron system that depends on the properties of the ( N + 1)-electron system (such as the electron affinity). We demonstrate that the accuracy with which widely used ML models approximate density functionals in these two classes is significantly limited by the pathological nature of the mapping from the density to the observable, even when the reference level of the external potential is fixed to prevent constant shifts. In addition, we find that providing the external potential as input to the ML model resolves this issue, in some cases reducing the error by up to three orders of magnitude when predicting quantities in these two classes.
- Open Access
- 15 December 2021
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 3 (1) , 015011
Machine learning the derivative discontinuity of density-functional theory
Machine learning is a powerful tool to design accurate, highly non-local, exchange-correlation functionals for density functional theory. So far, most of those machine learned functionals are trained for systems with an integer number of particles. As such, they are unable to reproduce some crucial and fundamental aspects, such as the explicit dependency of the functionals on the particle number or the infamous derivative discontinuity at integer particle numbers. Here we propose a solution to these problems by training a neural network as the universal functional of density-functional theory that (a) depends explicitly on the number of particles with a piece-wise linearity between the integer numbers and (b) reproduces the derivative discontinuity of the exchange-correlation energy. This is achieved by using an ensemble formalism, a training set containing fractional densities, and an explicitly discontinuous formulation.
- 15 December 2021
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 3 (1) , 015011
Open AccessMachine learning the derivative discontinuity of density-functional theory
Machine learning is a powerful tool to design accurate, highly non-local, exchange-correlation functionals for density functional theory. So far, most of those machine learned functionals are trained for systems with an integer number of particles. As such, they are unable to reproduce some crucial and fundamental aspects, such as the explicit dependency of the functionals on the particle number or the infamous derivative discontinuity at integer particle numbers. Here we propose a solution to these problems by training a neural network as the universal functional of density-functional theory that (a) depends explicitly on the number of particles with a piece-wise linearity between the integer numbers and (b) reproduces the derivative discontinuity of the exchange-correlation energy. This is achieved by using an ensemble formalism, a training set containing fractional densities, and an explicitly discontinuous formulation.
- Open Access
Exact exchange-correlation potentials for calculating the fundamental gap with a fixed number of electrons
Capturing the discontinuous shift by $\Delta$ in the exact exchange-correlation (xc) potential is the standard proposal for calculating the fundamental gap, $E_\mathrm{g}$, from the Kohn-Sham (KS) gap, $\varepsilon_\mathrm{g}$, within KS density functional theory (DFT), as $E_\mathrm{g} = \varepsilon_\mathrm{g} + \Delta$, yet this discontinuity is absent from existing approximations. The '$N$-centered' formulation of ensemble DFT artificially maintains a total electron number, $N$, in order to yield $E_\mathrm{g}$ not through a discontinuous shift in the xc potential but via the ensemble-weight derivative of the xc energy. Within the $N$-centered approach we calculate exact xc potentials for a one-dimensional finite system and show analytically that $\Delta$ can in fact be interpreted as a discontinuous shift in the exact $N$-centered ensemble xc potential, thereby extending to charged excitations an exact property of uncharged excitations. We show that applying the Levy-Zahariev 'shift-in-potential' procedure in this context relocates the discontinuous shift to the unimportant periphery of the system, so that the exact xc potential in effect is free of discontinuities and thus the inability of a local functional to capture discontinuous behavior is inconsequential.Open AccessExact exchange-correlation potentials for calculating the fundamental gap with a fixed number of electrons
Capturing the discontinuous shift by $\Delta$ in the exact exchange-correlation (xc) potential is the standard proposal for calculating the fundamental gap, $E_\mathrm{g}$, from the Kohn-Sham (KS) gap, $\varepsilon_\mathrm{g}$, within KS density functional theory (DFT), as $E_\mathrm{g} = \varepsilon_\mathrm{g} + \Delta$, yet this discontinuity is absent from existing approximations. The '$N$-centered' formulation of ensemble DFT artificially maintains a total electron number, $N$, in order to yield $E_\mathrm{g}$ not through a discontinuous shift in the xc potential but via the ensemble-weight derivative of the xc energy. Within the $N$-centered approach we calculate exact xc potentials for a one-dimensional finite system and show analytically that $\Delta$ can in fact be interpreted as a discontinuous shift in the exact $N$-centered ensemble xc potential, thereby extending to charged excitations an exact property of uncharged excitations. We show that applying the Levy-Zahariev 'shift-in-potential' procedure in this context relocates the discontinuous shift to the unimportant periphery of the system, so that the exact xc potential in effect is free of discontinuities and thus the inability of a local functional to capture discontinuous behavior is inconsequential. - 12 January 2021
- Published by American Physical Society (APS) in Physical Review A
- Vol. 103 (1) , 012806
Exact exchange-correlation potentials for calculating the fundamental gap with a fixed number of electrons
Capturing the discontinuous shift by in the exact exchange-correlation (xc) potential is the standard proposal for calculating the fundamental gap from the Kohn-Sham (KS) gap within KS density functional theory (DFT) as , yet this discontinuity is absent from existing approximations. The “-centered” formulation of ensemble DFT artificially maintains a total electron number in order to yield not through a discontinuous shift in the xc potential but via the ensemble-weight derivative of the xc energy. Within the -centered approach, we calculate exact xc potentials for a one-dimensional finite system and show analytically that can in fact be interpreted as a discontinuous shift in the exact -centered ensemble xc potential, thereby extending to charged excitations an exact property of uncharged excitations. We show that applying the Levy-Zahariev “shift-in-potential” procedure in this context relocates the discontinuous shift to the unimportant periphery of the system, so that the exact xc potential in effect is free of discontinuities and thus the inability of a local functional to capture discontinuous behavior is inconsequential.- 12 January 2021
- Published by American Physical Society (APS) in Physical Review A
- Vol. 103 (1) , 012806
Exact exchange-correlation potentials for calculating the fundamental gap with a fixed number of electrons
Capturing the discontinuous shift by in the exact exchange-correlation (xc) potential is the standard proposal for calculating the fundamental gap from the Kohn-Sham (KS) gap within KS density functional theory (DFT) as , yet this discontinuity is absent from existing approximations. The “-centered” formulation of ensemble DFT artificially maintains a total electron number in order to yield not through a discontinuous shift in the xc potential but via the ensemble-weight derivative of the xc energy. Within the -centered approach, we calculate exact xc potentials for a one-dimensional finite system and show analytically that can in fact be interpreted as a discontinuous shift in the exact -centered ensemble xc potential, thereby extending to charged excitations an exact property of uncharged excitations. We show that applying the Levy-Zahariev “shift-in-potential” procedure in this context relocates the discontinuous shift to the unimportant periphery of the system, so that the exact xc potential in effect is free of discontinuities and thus the inability of a local functional to capture discontinuous behavior is inconsequential.- 12 June 2020
- Published by Royal Society of Chemistry (RSC) in Faraday Discussions
- Vol. 224, 265-291
Insights into one-body density matrices using deep learning
Deep-learning constraints of the one-body reduced density matrix from its compressibility to enable efficient determination of key observables.
- 12 June 2020
- Published by Royal Society of Chemistry (RSC) in Faraday Discussions
- Vol. 224, 265-291
Insights into one-body density matrices using deep learning
Deep-learning constraints of the one-body reduced density matrix from its compressibility to enable efficient determination of key observables.
- Open Access
Insights into one-body density matrices using deep learning
The one-body reduced density matrix (1-RDM) of a many-body system at zero temperature gives direct access to many observables, such as the charge density, kinetic energy and occupation numbers. It would be desirable to express it as a simple functional of the density or of other local observables, but to date satisfactory approximations have not yet been found. Deep learning is the state-of the art approach to perform high dimensional regressions and classification tasks, and is becoming widely used in the condensed matter community to develop increasingly accurate density functionals. Autoencoders are deep learning models that perform efficient dimensionality reduction, allowing the distillation of data to its fundamental features needed to represent it. By training autoencoders on a large data-set of 1-RDMs from exactly solvable real-space model systems, and performing principal component analysis, the machine learns to what extent the data can be compressed and hence how it is constrained. We gain insight into these machine learned constraints and employ them to inform approximations to the 1-RDM as a functional of the charge density. We exploit known physical properties of the 1-RDM in the simplest possible cases to perform feature engineering, where we inform the structure of the models from known mathematical relations, allowing us to integrate existing understanding into the machine learning methods. By comparing various deep learning approaches we gain insight into what physical features of the density matrix are most amenable to machine learning, utilising both known and learned characteristics.
Open AccessInsights into one-body density matrices using deep learning
The one-body reduced density matrix (1-RDM) of a many-body system at zero temperature gives direct access to many observables, such as the charge density, kinetic energy and occupation numbers. It would be desirable to express it as a simple functional of the density or of other local observables, but to date satisfactory approximations have not yet been found. Deep learning is the state-of the art approach to perform high dimensional regressions and classification tasks, and is becoming widely used in the condensed matter community to develop increasingly accurate density functionals. Autoencoders are deep learning models that perform efficient dimensionality reduction, allowing the distillation of data to its fundamental features needed to represent it. By training autoencoders on a large data-set of 1-RDMs from exactly solvable real-space model systems, and performing principal component analysis, the machine learns to what extent the data can be compressed and hence how it is constrained. We gain insight into these machine learned constraints and employ them to inform approximations to the 1-RDM as a functional of the charge density. We exploit known physical properties of the 1-RDM in the simplest possible cases to perform feature engineering, where we inform the structure of the models from known mathematical relations, allowing us to integrate existing understanding into the machine learning methods. By comparing various deep learning approaches we gain insight into what physical features of the density matrix are most amenable to machine learning, utilising both known and learned characteristics.
- Open Access
- 2 March 2020
- Published by American Physical Society (APS) in Physical Review A
- Vol. 101 (3) , 032502
Accurate real-time evolution of electron densities and ground-state properties from generalized Kohn-Sham theory
The exact static and time-dependent Kohn-Sham (KS) exchange-correlation potential is extremely challenging to approximate as it is a local multiplicative potential that depends on the electron density everywhere in the system. The KS approach can be generalized by allowing part of the potential to be spatially nonlocal. We take this nonlocal part to be that of unrestricted Hartree-Fock theory. The additional local correlation potential in principle ensures that the single-particle density exactly equals the many-body density. In our case, the local correlation potential is predominantly nearsighted in its dependence on the density and hence an (adiabatic) local-density approximation to this potential yields accurate ground-state properties and real-time densities for one-dimensional test systems.
- 2 March 2020
- Published by American Physical Society (APS) in Physical Review A
- Vol. 101 (3) , 032502
Open AccessAccurate real-time evolution of electron densities and ground-state properties from generalized Kohn-Sham theory
The exact static and time-dependent Kohn-Sham (KS) exchange-correlation potential is extremely challenging to approximate as it is a local multiplicative potential that depends on the electron density everywhere in the system. The KS approach can be generalized by allowing part of the potential to be spatially nonlocal. We take this nonlocal part to be that of unrestricted Hartree-Fock theory. The additional local correlation potential in principle ensures that the single-particle density exactly equals the many-body density. In our case, the local correlation potential is predominantly nearsighted in its dependence on the density and hence an (adiabatic) local-density approximation to this potential yields accurate ground-state properties and real-time densities for one-dimensional test systems.
- Open Access
Accurate real-time evolution of electron densities and ground-state properties from generalized Kohn-Sham theory
The exact static and time-dependent Kohn-Sham (KS) exchange-correlation (xc) potential is extremely challenging to approximate as it is a local multiplicative potential that depends on the electron density everywhere in the system. The KS approach can be generalised by allowing part of the potential to be spatially nonlocal. We take this nonlocal part to be that of unrestricted Hartree-Fock theory. The additional local correlation potential in principle ensures that the single-particle density exactly equals the many-body density. In our case, the local correlation potential is predominantly nearsighted in its dependence on the density and hence an (adiabatic) local density approximation to this potential yields accurate ground-state properties and real-time densities for one-dimensional test systems.
Open AccessAccurate real-time evolution of electron densities and ground-state properties from generalized Kohn-Sham theory
The exact static and time-dependent Kohn-Sham (KS) exchange-correlation (xc) potential is extremely challenging to approximate as it is a local multiplicative potential that depends on the electron density everywhere in the system. The KS approach can be generalised by allowing part of the potential to be spatially nonlocal. We take this nonlocal part to be that of unrestricted Hartree-Fock theory. The additional local correlation potential in principle ensures that the single-particle density exactly equals the many-body density. In our case, the local correlation potential is predominantly nearsighted in its dependence on the density and hence an (adiabatic) local density approximation to this potential yields accurate ground-state properties and real-time densities for one-dimensional test systems.
- 16 January 2019
- Published by American Physical Society (APS) in Physical Review B
- Vol. 99 (4) , 045129
Advantageous nearsightedness of many-body perturbation theory contrasted with Kohn-Sham density functional theory
For properties of interacting electron systems, Kohn-Sham (KS) theory is often favored over many-body perturbation theory (MBPT), owing to its low computational cost. However, the exact KS potential can be challenging to approximate, for example in the presence of localized subsystems where the exact potential is known to exhibit pathological features such as spatial steps. By modeling two electrons, each localized in a distinct potential well, we illustrate that the step feature has no counterpart in MBPTs (including Hartree-Fock and or hybrid methods involving Fock exchange because the spatial nonlocality of the self-energy renders such pathological behavior unnecessary. We present a quantitative illustration of the orbital-dependent nature of the nonlocal potential, and a numerical demonstration of Kohn's concept of the nearsightedness for self-energies, when two distant subsystems are combined, in contrast to the KS potential. These properties emphasize the value of self-energy-based approximations in developing future approaches within KS-like theories.- 16 January 2019
- Published by American Physical Society (APS) in Physical Review B
- Vol. 99 (4) , 045129
Advantageous nearsightedness of many-body perturbation theory contrasted with Kohn-Sham density functional theory
For properties of interacting electron systems, Kohn-Sham (KS) theory is often favored over many-body perturbation theory (MBPT), owing to its low computational cost. However, the exact KS potential can be challenging to approximate, for example in the presence of localized subsystems where the exact potential is known to exhibit pathological features such as spatial steps. By modeling two electrons, each localized in a distinct potential well, we illustrate that the step feature has no counterpart in MBPTs (including Hartree-Fock and or hybrid methods involving Fock exchange because the spatial nonlocality of the self-energy renders such pathological behavior unnecessary. We present a quantitative illustration of the orbital-dependent nature of the nonlocal potential, and a numerical demonstration of Kohn's concept of the nearsightedness for self-energies, when two distant subsystems are combined, in contrast to the KS potential. These properties emphasize the value of self-energy-based approximations in developing future approaches within KS-like theories.- Open Access
- 13 April 2018
- Published by American Physical Society (APS) in Physical Review Materials
- Vol. 2 (4) , 040801
Accuracy of electron densities obtained via Koopmans-compliant hybrid functionals
We evaluate the accuracy of electron densities and quasiparticle energy gaps given by hybrid functionals by directly comparing these to the exact quantities obtained from solving the many-electron Schrödinger equation. We determine the admixture of Hartree-Fock exchange to approximate exchange-correlation in our hybrid functional via one of several physically justified constraints, including the generalized Koopmans' theorem. We find that hybrid functionals yield strikingly accurate electron densities and gaps in both exchange-dominated and correlated systems. We also discuss the role of the screened Fock operator in the success of hybrid functionals.
- 13 April 2018
- Published by American Physical Society (APS) in Physical Review Materials
- Vol. 2 (4) , 040801
Open AccessAccuracy of electron densities obtained via Koopmans-compliant hybrid functionals
We evaluate the accuracy of electron densities and quasiparticle energy gaps given by hybrid functionals by directly comparing these to the exact quantities obtained from solving the many-electron Schrödinger equation. We determine the admixture of Hartree-Fock exchange to approximate exchange-correlation in our hybrid functional via one of several physically justified constraints, including the generalized Koopmans' theorem. We find that hybrid functionals yield strikingly accurate electron densities and gaps in both exchange-dominated and correlated systems. We also discuss the role of the screened Fock operator in the success of hybrid functionals.