{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,10]],"date-time":"2026-10-10T23:48:42Z","timestamp":1791676122197,"version":"4.3.4"},"reference-count":47,"publisher":"Springer Science and Business Media LLC","issue":"1","license":[{"start":{"date-parts":[[2026,6,19]],"date-time":"2026-06-19T00:00:00Z","timestamp":1781827200000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2026,10,10]],"date-time":"2026-10-10T00:00:00Z","timestamp":1791590400000},"content-version":"vor","delay-in-days":113,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"funder":[{"DOI":"10.13039\/501100000266","name":"Engineering and Physical Sciences Research Council","doi-asserted-by":"publisher","award":["EP\/T022167\/1"],"award-info":[{"award-number":["EP\/T022167\/1"]}],"id":[{"id":"10.13039\/501100000266","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100000266","name":"Engineering and Physical Sciences Research Council","doi-asserted-by":"publisher","award":["EP\/W030489\/1 and EP\/X035891\/1"],"award-info":[{"award-number":["EP\/W030489\/1 and EP\/X035891\/1"]}],"id":[{"id":"10.13039\/501100000266","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100000266","name":"Engineering and Physical Sciences Research Council","doi-asserted-by":"publisher","award":["EP\/T022167\/1"],"award-info":[{"award-number":["EP\/T022167\/1"]}],"id":[{"id":"10.13039\/501100000266","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100014013","name":"UK Research and Innovation","doi-asserted-by":"publisher","award":["UKRI2710"],"award-info":[{"award-number":["UKRI2710"]}],"id":[{"id":"10.13039\/100014013","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["npj Comput Mater"],"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>\n                    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 \u2018hypersensitivity\u2019 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                    <jats:italic>N<\/jats:italic>\n                    -electron system that depends on the properties of the (\n                    <jats:italic>N<\/jats:italic>\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.\n                  <\/jats:p>","DOI":"10.1038\/s41524-026-02191-x","type":"journal-article","created":{"date-parts":[[2026,6,19]],"date-time":"2026-06-19T20:56:21Z","timestamp":1781902581000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Addressing hypersensitivity in density functional theory for reliable machine learning"],"prefix":"10.1038","volume":"12","author":[{"ORCID":"https:\/\/orcid.org\/0009-0000-3480-8890","authenticated-orcid":false,"given":"L.","family":"Arnstein","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-1190-3830","authenticated-orcid":false,"given":"J.","family":"Wetherell","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-0025-2497","authenticated-orcid":false,"given":"R.","family":"Lawrence","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4314-4093","authenticated-orcid":false,"given":"P. J.","family":"Hasnip","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2256-6860","authenticated-orcid":false,"given":"M. J. P.","family":"Hodgson","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"297","published-online":{"date-parts":[[2026,6,19]]},"reference":[{"key":"2191_CR1","doi-asserted-by":"publisher","first-page":"18A301","DOI":"10.1063\/1.4869598","volume":"140","author":"AD Becke","year":"2014","unstructured":"Becke, A. D. Perspective: fifty years of density-functional theory in chemical physics. J. Chem. Phys. 140, 18A301 (2014).","journal-title":"J. Chem. Phys."},{"key":"2191_CR2","doi-asserted-by":"publisher","first-page":"B864","DOI":"10.1103\/PhysRev.136.B864","volume":"136","author":"P Hohenberg","year":"1964","unstructured":"Hohenberg, P. & Kohn, W. Inhomogeneous electron gas. Phys. Rev. 136, B864\u2013B871 (1964).","journal-title":"Phys. Rev."},{"key":"2191_CR3","doi-asserted-by":"publisher","first-page":"1253","DOI":"10.1103\/RevModPhys.71.1253","volume":"71","author":"W Kohn","year":"1999","unstructured":"Kohn, W. Nobel lecture: electronic structure of matter-wave functions and density functionals. Rev. Mod. Phys. 71, 1253 (1999).","journal-title":"Rev. Mod. Phys."},{"key":"2191_CR4","doi-asserted-by":"publisher","first-page":"6062","DOI":"10.1073\/pnas.76.12.6062","volume":"76","author":"M Levy","year":"1979","unstructured":"Levy, M. Universal variational functionals of electron densities, first-order density matrices, and natural spin-orbitals and solution of the v-representability problem. Proc. Natl. Acad. Sci. 76, 6062\u20136065 (1979).","journal-title":"Proc. Natl. Acad. Sci."},{"key":"2191_CR5","doi-asserted-by":"publisher","first-page":"243","DOI":"10.1002\/qua.560240302","volume":"24","author":"EH Lieb","year":"1983","unstructured":"Lieb, E. H. Density functionals for coulomb systems. Int. J. quantum Chem. 24, 243\u2013277 (1983).","journal-title":"Int. J. quantum Chem."},{"key":"2191_CR6","doi-asserted-by":"crossref","unstructured":"Levy, M. & Perdew, J. P. The constrained search formulation of density functional theory. In Density Functional Methods in Physics 11\u201330 (Springer, 1985).","DOI":"10.1007\/978-1-4757-0818-9_2"},{"key":"2191_CR7","doi-asserted-by":"crossref","unstructured":"Alharbi, F. H. Kinetic energy density functional: advances, challenges, and future directions. Electron. Struct. 8, 023001 (2026).","DOI":"10.1088\/2516-1075\/ae55ec"},{"key":"2191_CR8","doi-asserted-by":"publisher","first-page":"A1133","DOI":"10.1103\/PhysRev.140.A1133","volume":"140","author":"W Kohn","year":"1965","unstructured":"Kohn, W. & Sham, L. J. Self-consistent equations including exchange and correlation effects. Phys. Rev. 140, A1133\u2013A1138 (1965).","journal-title":"Phys. Rev."},{"key":"2191_CR9","doi-asserted-by":"publisher","first-page":"090901","DOI":"10.1063\/5.0150587","volume":"159","author":"J Wu","year":"2023","unstructured":"Wu, J., Pun, S.-M., Zheng, X. & Chen, G. Construct exchange-correlation functional via machine learning. J. Chem. Phys. 159, 090901 (2023).","journal-title":"J. Chem. Phys."},{"key":"2191_CR10","doi-asserted-by":"publisher","first-page":"024110","DOI":"10.1063\/5.0063629","volume":"156","author":"S Kumar","year":"2022","unstructured":"Kumar, S. et al. Accurate parameterization of the kinetic energy functional. J. Chem. Phys. 156, 024110 (2022).","journal-title":"J. Chem. Phys."},{"key":"2191_CR11","doi-asserted-by":"publisher","first-page":"497","DOI":"10.1002\/qua.560280846","volume":"28","author":"JP Perdew","year":"1985","unstructured":"Perdew, J. P. Density functional theory and the band gap problem. Int. J. Quantum Chem. 28, 497\u2013523 (1985).","journal-title":"Int. J. Quantum Chem."},{"key":"2191_CR12","doi-asserted-by":"publisher","first-page":"10159","DOI":"10.1103\/PhysRevB.37.10159","volume":"37","author":"RW Godby","year":"1988","unstructured":"Godby, R. W., Schl\u00fcter, M. & Sham, L. J. Self-energy operators and exchange-correlation potentials in semiconductors. Phys. Rev. B 37, 10159\u201310175 (1988).","journal-title":"Phys. Rev. B"},{"key":"2191_CR13","doi-asserted-by":"publisher","DOI":"10.1038\/srep02810","volume":"3","author":"G Pilania","year":"2013","unstructured":"Pilania, G., Wang, C., Jiang, X., Rajasekaran, S. & Ramprasad, R. Accelerating materials property predictions using machine learning. Sci. Rep. 3, 2810 (2013).","journal-title":"Sci. Rep."},{"key":"2191_CR14","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-017-01251-z","volume":"7","author":"B Kolb","year":"2017","unstructured":"Kolb, B., Lentz, L. C. & Kolpak, A. M. Discovering charge density functionals and structure-property relationships with PROPhet: a general framework for coupling machine learning and first-principles methods. Sci. Rep. 7, 1192 (2017).","journal-title":"Sci. Rep."},{"key":"2191_CR15","doi-asserted-by":"publisher","DOI":"10.1038\/s41524-020-0310-0","volume":"6","author":"R Nagai","year":"2020","unstructured":"Nagai, R., Akashi, R. & Sugino, O. Completing density functional theory by machine learning hidden messages from molecules. npj Comput. Mater. 6, 43 (2020).","journal-title":"npj Comput. Mater."},{"key":"2191_CR16","doi-asserted-by":"publisher","first-page":"1385","DOI":"10.1126\/science.abj6511","volume":"374","author":"J Kirkpatrick","year":"2021","unstructured":"Kirkpatrick, J. et al. Pushing the frontiers of density functionals by solving the fractional electron problem. Science 374, 1385\u20131389 (2021).","journal-title":"Science"},{"key":"2191_CR17","unstructured":"Akashi, R., Sogal, M. & Burke, K. Can machines learn density functionals? Past, present, and future of ML in DFT. https:\/\/arxiv.org\/abs\/2503.01709 (2025)."},{"key":"2191_CR18","doi-asserted-by":"publisher","DOI":"10.1038\/s41524-019-0162-7","volume":"5","author":"A Chandrasekaran","year":"2019","unstructured":"Chandrasekaran, A. et al. Solving the electronic structure problem with machine learning. npj Comput. Mater. 5, 22 (2019).","journal-title":"npj Comput. Mater."},{"key":"2191_CR19","doi-asserted-by":"publisher","first-page":"022512","DOI":"10.1103\/PhysRevA.100.022512","volume":"100","author":"K Ryczko","year":"2019","unstructured":"Ryczko, K., Strubbe, D. A. & Tamblyn, I. Deep learning and density-functional theory. Phys. Rev. A 100, 022512 (2019).","journal-title":"Phys. Rev. A"},{"key":"2191_CR20","doi-asserted-by":"publisher","unstructured":"Snyder, J. C., Rupp, M., Hansen, K., M\u00fcller, K.-R. & Burke, K. Finding density functionals with machine learning. Phys. Rev. Lett. 108, https:\/\/doi.org\/10.1103\/PhysRevLett.108.253002 (2012).","DOI":"10.1103\/PhysRevLett.108.253002"},{"key":"2191_CR21","doi-asserted-by":"publisher","unstructured":"Snyder, J. C. et al. Orbital-free bond breaking via machine learning. J. Chem. Phys. 139, https:\/\/doi.org\/10.1063\/1.4834075 (2013).","DOI":"10.1063\/1.4834075"},{"key":"2191_CR22","doi-asserted-by":"publisher","first-page":"819","DOI":"10.1002\/qua.25040","volume":"116","author":"L Li","year":"2016","unstructured":"Li, L. et al. Understanding machine-learned density functionals. Int. J. Quantum Chem. 116, 819\u2013833 (2016).","journal-title":"Int. J. Quantum Chem."},{"key":"2191_CR23","doi-asserted-by":"publisher","unstructured":"Li, L., Baker, T. E., White, S. R. & Burke, K. Pure density functional for strong correlation and the thermodynamic limit from machine learning. Phys. Rev. B 94, https:\/\/doi.org\/10.1103\/PhysRevB.94.245129 (2016).","DOI":"10.1103\/PhysRevB.94.245129"},{"key":"2191_CR24","doi-asserted-by":"publisher","unstructured":"Meyer, R., Weichselbaum, M. & Hauser, A. W. Machine learning approaches toward orbital-free density functional theory: Simultaneous training on the kinetic energy density functional and its functional derivative. J. Chem. Theory Comput. 16, https:\/\/doi.org\/10.1021\/acs.jctc.0c00580 (2020).","DOI":"10.1021\/acs.jctc.0c00580"},{"key":"2191_CR25","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-020-19093-1","volume":"11","author":"M Bogojeski","year":"2020","unstructured":"Bogojeski, M., Vogt-Maranto, L., Tuckerman, M. E., M\u00fcller, K.-R. & Burke, K. Quantum chemical accuracy from density functional approximations via machine learning. Nat. Commun. 11, 5223 (2020).","journal-title":"Nat. Commun."},{"key":"2191_CR26","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-017-00839-3","volume":"8","author":"F Brockherde","year":"2017","unstructured":"Brockherde, F. et al. Bypassing the Kohn-Sham equations with machine learning. Nat. Commun. 8, 872 (2017).","journal-title":"Nat. Commun."},{"key":"2191_CR27","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1007\/s40192-024-00389-9","volume":"14","author":"P Ray","year":"2025","unstructured":"Ray, P., Choudhary, K. & Kalidindi, S. R. Lean CNNs for mapping electron charge density fields to material properties. Integrating Mater. Manuf. Innov. 14, 1\u201313 (2025).","journal-title":"Integrating Mater. Manuf. Innov."},{"key":"2191_CR28","doi-asserted-by":"publisher","DOI":"10.1038\/s41524-023-01115-3","volume":"9","author":"BG del Rio","year":"2023","unstructured":"del Rio, B. G., Phan, B. & Ramprasad, R. A deep learning framework to emulate density functional theory. npj Comput. Mater. 9, 158 (2023).","journal-title":"npj Comput. Mater."},{"key":"2191_CR29","unstructured":"Chen, F. et al. Towards universal material property prediction with deep learning and single-descriptor electronic density. https:\/\/arxiv.org\/abs\/2510.13207 (2025)."},{"key":"2191_CR30","doi-asserted-by":"publisher","first-page":"083802","DOI":"10.1103\/PhysRevMaterials.5.083802","volume":"5","author":"JR Moreno","year":"2021","unstructured":"Moreno, J. R., Flick, J. & Georges, A. Machine learning band gaps from the electron density. Phys. Rev. Mater. 5, 083802 (2021).","journal-title":"Phys. Rev. Mater."},{"key":"2191_CR31","doi-asserted-by":"publisher","unstructured":"Gedeon, J. et al. Machine learning the derivative discontinuity of density-functional theory. Mach. Learn. Sci. Technol. https:\/\/doi.org\/10.1088\/2632-2153\/ac3149 (2021).","DOI":"10.1088\/2632-2153\/ac3149"},{"key":"2191_CR32","doi-asserted-by":"publisher","unstructured":"Mills, K., Spanner, M. & Tamblyn, I. Deep learning and the schr\u00f6dinger equation. Phys. Rev. A 96, https:\/\/doi.org\/10.1103\/PhysRevA.96.042113 (2017).","DOI":"10.1103\/PhysRevA.96.042113"},{"key":"2191_CR33","doi-asserted-by":"crossref","unstructured":"Nigam, J., Smidt, T. & Dusson, G. Machine learning electronic structure and atomistic properties from the external potential. https:\/\/arxiv.org\/abs\/2602.15345 (2026).","DOI":"10.1063\/5.0332678"},{"key":"2191_CR34","doi-asserted-by":"publisher","first-page":"2339","DOI":"10.1103\/PhysRevA.29.2339","volume":"29","author":"B-G Englert","year":"1984","unstructured":"Englert, B.-G. & Schwinger, J. Statistical atom: some quantum improvements. Phys. Rev. A 29, 2339 (1984).","journal-title":"Phys. Rev. A"},{"key":"2191_CR35","doi-asserted-by":"publisher","first-page":"146404","DOI":"10.1103\/PhysRevLett.92.146404","volume":"92","author":"W Yang","year":"2004","unstructured":"Yang, W., Ayers, P. W. & Wu, Q. Potential functionals: dual to density functionals and solution to the v-representability problem. Phys. Rev. Lett. 92, 146404 (2004).","journal-title":"Phys. Rev. Lett."},{"key":"2191_CR36","doi-asserted-by":"publisher","first-page":"062505","DOI":"10.1103\/PhysRevA.88.062505","volume":"88","author":"A Cangi","year":"2013","unstructured":"Cangi, A., Gross, E. & Burke, K. Potential functionals versus density functionals. Phys. Rev. A At. Mol. Opt. Phys. 88, 062505 (2013).","journal-title":"Phys. Rev. A At. Mol. Opt. Phys."},{"key":"2191_CR37","doi-asserted-by":"publisher","first-page":"21135","DOI":"10.52202\/075280-0923","volume":"36","author":"G Kornowski","year":"2023","unstructured":"Kornowski, G., Hanneke, S. & Kontorovich, A. Near-optimal learning with average h\u00f6lder smoothness. Adv. Neural Inf. Process. Syst. 36, 21135\u201321157 (2023).","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"2191_CR38","first-page":"015050","volume":"5","author":"V Martinetto","year":"2024","unstructured":"Martinetto, V., Shah, K., Cangi, A. & Pribram-Jones, A. Inverting the Kohn-Sham equations with physics-informed machine learning. Mach. Learn.: Sci. Technol. 5, 015050 (2024).","journal-title":"Mach. Learn.: Sci. Technol."},{"key":"2191_CR39","doi-asserted-by":"publisher","first-page":"241102","DOI":"10.1103\/PhysRevB.88.241102","volume":"88","author":"MJP Hodgson","year":"2013","unstructured":"Hodgson, M. J. P., Ramsden, J. D., Chapman, J. B. J., Lillystone, P. & Godby, R. W. Exact time-dependent density-functional potentials for strongly correlated tunneling electrons. Phys. Rev. B 88, 241102 (2013).","journal-title":"Phys. Rev. B"},{"key":"2191_CR40","doi-asserted-by":"publisher","first-page":"063411","DOI":"10.1103\/PhysRevA.72.063411","volume":"72","author":"A Gordon","year":"2005","unstructured":"Gordon, A., Santra, R. & K\u00e4rtner, F. X. Role of the coulomb singularity in high-order harmonic generation. Phys. Rev. A 72, 063411 (2005).","journal-title":"Phys. Rev. A"},{"key":"2191_CR41","first-page":"Article-ID-2015","volume":"472","author":"R Loudon","year":"2016","unstructured":"Loudon, R. One-dimensional hydrogen atom. Proc. R. Soc. A Math., Phys. Eng. Sci. 472, Article-ID-20150534 (2016).","journal-title":"Proc. R. Soc. A Math., Phys. Eng. Sci."},{"key":"2191_CR42","unstructured":"Huang, J. W., Roberts, S. J. & Calliess, J.-P. On the sample complexity of Lipschitz constant estimation. Trans. Mach. Learn. Res. https:\/\/openreview.net\/forum?id=UIalYAHdBH (2023)."},{"key":"2191_CR43","first-page":"117:1\u2013117:54","volume":"25","author":"Y Ashlagi","year":"2024","unstructured":"Ashlagi, Y., Gottlieb, L.-A. & Kontorovich, A. Functions with average smoothness: structure, algorithms, and learning. J. Mach. Learn. Res. 25, 117:1\u2013117:54 (2024).","journal-title":"J. Mach. Learn. Res."},{"key":"2191_CR44","doi-asserted-by":"publisher","unstructured":"Moreno, J. R., Carleo, G. & Georges, A. Deep learning the Hohenberg-Kohn maps of density functional theory. Phys. Rev. Lett. 125, https:\/\/doi.org\/10.1103\/PhysRevLett.125.076402 (2020).","DOI":"10.1103\/PhysRevLett.125.076402"},{"key":"2191_CR45","doi-asserted-by":"publisher","first-page":"1115","DOI":"10.1002\/qua.24939","volume":"115","author":"K Vu","year":"2015","unstructured":"Vu, K. et al. Understanding kernel ridge regression: common behaviors from simple functions to density functionals. Int. J. Quantum Chem. 115, 1115\u20131128 (2015).","journal-title":"Int. J. Quantum Chem."},{"key":"2191_CR46","unstructured":"Kanagawa, M., Hennig, P., Sejdinovic, D. & Sriperumbudur, B. K. Gaussian processes and kernel methods: a review on connections and equivalences. https:\/\/arxiv.org\/abs\/1807.02582 (2018)."},{"key":"2191_CR47","doi-asserted-by":"publisher","first-page":"3404","DOI":"10.1021\/ct400195d","volume":"9","author":"K Hansen","year":"2013","unstructured":"Hansen, K. et al. Assessment and validation of machine learning methods for predicting molecular atomization energies. J. Chem. Theory Comput. 9, 3404\u20133419 (2013).","journal-title":"J. Chem. Theory Comput."}],"container-title":["npj Computational Materials"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.nature.com\/articles\/s41524-026-02191-x","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.nature.com\/articles\/s41524-026-02191-x.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.nature.com\/articles\/s41524-026-02191-x.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,10,10]],"date-time":"2026-10-10T23:05:16Z","timestamp":1791673516000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.nature.com\/articles\/s41524-026-02191-x"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,6,19]]},"references-count":47,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2026,12]]}},"alternative-id":["2191"],"URL":"https:\/\/doi.org\/10.1038\/s41524-026-02191-x","relation":{},"ISSN":["2057-3960"],"issn-type":[{"value":"2057-3960","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,6,19]]},"assertion":[{"value":"18 February 2026","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"4 June 2026","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"19 June 2026","order":3,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"The authors declare no competing interests.","order":1,"name":"Ethics","label":"Competing interests","group":{"name":"EthicsHeading","label":"Ethics"}}],"article-number":"338"}}