pith. sign in

arxiv: 2606.01763 · v1 · pith:7YDN53IOnew · submitted 2026-06-01 · ❄️ cond-mat.mtrl-sci

Polaron Transport in TiO₂ from Machine Learning Molecular Dynamics

Christian S. Ahart , Denan Li , Jochen Blumberger , Shi Liu This is my paper

Pith reviewed 2026-06-28 14:03 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci
keywords polaron transporttitanium dioxidemachine learning molecular dynamicsrutileanatasecharge mobilityactivation energyadiabatic polaron
0
0 comments X

The pith

Machine learning molecular dynamics shows that electron polarons in rutile TiO2 localize on one Ti atom and hop only along the [001] direction with a 39 meV barrier.

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

The authors develop a machine learning framework to run first-principles molecular dynamics simulations of polaron motion in titanium dioxide over timescales three orders of magnitude longer than standard methods permit. They report that an added electron in the rutile phase forms a small polaron mostly on a single titanium atom, moves only along one crystal axis, and exhibits low activation energy and mobility values that match experimental data. In the anatase phase a hole forms a polaron on a single oxygen atom and must jump to second-nearest neighbors because of weak orbital overlap, producing a much higher barrier and lower mobility. These concrete transport parameters matter for photoelectrochemical devices because slow and anisotropic charge flow limits the efficiency of TiO2 as a catalyst. The work supplies finite-temperature, atomistic pictures of small-polaron behavior that earlier static or short-time calculations could not reach.

Core claim

DeepPolaron extends the accessible timescale of adiabatic polaron molecular dynamics by three orders of magnitude while retaining first-principles accuracy. Applied to TiO2, the excess electron in rutile relaxes to a polaron localized predominantly on a single Ti atom whose hops are restricted to the [001] direction, giving an activation energy of 39 meV and room-temperature mobility of 4.4 × 10^{-2} cm²/Vs. The hole polaron in anatase localizes on a single O atom and, owing to poor O 2p overlap with nearest neighbors, transports mainly to second-nearest neighbors with a 139 meV barrier and mobility of 1.4 × 10^{-3} cm²/Vs.

What carries the argument

DeepPolaron, a machine learning framework trained on short DFT trajectories that reproduces adiabatic polaron localization and long-time hopping dynamics.

If this is right

  • Electron polarons in rutile TiO2 remain localized on one Ti atom and hop exclusively along the [001] direction.
  • The activation energy for this process is 39 meV, producing a room-temperature mobility of 4.4 × 10^{-2} cm²/Vs.
  • Hole polarons in anatase localize on one O atom and hop primarily to second-nearest neighbors because of weak orbital overlap.
  • The corresponding activation energy is 139 meV and the room-temperature mobility is 1.4 × 10^{-3} cm²/Vs.
  • The same machine-learning approach transfers directly to other small-polaron materials and to interfacial charge-transfer processes.

Where Pith is reading between the lines

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

  • Crystal orientation will matter when TiO2 is used in thin-film or nanostructured devices because transport is strongly anisotropic in rutile.
  • The method opens the possibility of simulating polaron dynamics at defects or surfaces that were previously inaccessible.
  • The large mobility contrast between rutile and anatase supplies a concrete criterion for choosing which phase to employ in a given photocatalyst architecture.

Load-bearing premise

A machine learning model trained on short DFT trajectories reproduces adiabatic polaron localization and long-time hopping dynamics with virtually negligible loss in accuracy over the extended timescales accessed by the method.

What would settle it

An experimental observation of isotropic electron hopping in rutile TiO2 or a room-temperature mobility differing substantially from 4.4 × 10^{-2} cm²/Vs would falsify the reported transport picture.

Figures

Figures reproduced from arXiv: 2606.01763 by Christian S. Ahart, Denan Li, Jochen Blumberger, Shi Liu.

Figure 1
Figure 1. Figure 1: Polaron structure in rutile and anatase. Excess spin density for charged ground [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Polaron hopping in rutile and anatase. Excess spin density for charged ground [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: (a, b) DFT-MD and (c, d) DeepPolaron-MD of the electron polaron in rutile. (a, [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Training of DeepPolaron for the electron polaron in rutile. Parity plots showing a [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: (a, b) DFT-MD and (c, d) DeepPolaron-MD of the hole polaron in anatase. (a, [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Training of DeepPolaron for the hole polaron in anatase. Parity plots showing a [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Rate constant and mobility as a function of temperature. (a) Rate constants cal [PITH_FULL_IMAGE:figures/full_fig_p015_7.png] view at source ↗
read the original abstract

Transition metal oxides have attracted much attention as photo(electrochemical)-catalysts but practical applications are typically hampered by their low and anisotropic charge mobility. A deep understanding of excess charge carrier transport in these materials requires a dynamical treatment of nuclear motion that goes well beyond standard approaches. Here we introduce DeepPolaron, a machine learning framework boosting the accessible time scale of first principles molecular dynamics of adiabatic polaron transport by three orders of magnitude at a virtually negligible loss in accuracy. We apply our method to excess electron and hole transport in titanium dioxide rutile and anatase. We find that the excess electron in rutile relaxes to a polaron predominantly localized on a single Ti atom with hopping occurring only along the [001] direction, associated with an activation energy of 39 meV and a room temperature mobility of $4.4 \times 10^{-2}$ cm$^2$/Vs in good agreement with experiment. In contrast the hole polaron in anatase is localized on a single O atom, and due to poor O 2p orbital overlap with first nearest neighbors charge transport occurs primarily to second nearest neighbors, with a large activation energy of 139 meV resulting in a small room temperature mobility of $1.4 \times 10^{-3}$ cm$^2$/Vs. This work provides a finite temperature first-principles characterization of small polaron transport in rutile and anatase, with a methodology that is directly transferable to other small polaron forming materials and interfacial charge-transfer processes.

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 paper introduces DeepPolaron, a machine learning framework that extends first-principles molecular dynamics timescales for adiabatic polaron transport by three orders of magnitude with claimed negligible accuracy loss. Applied to TiO2, it reports that the excess electron in rutile relaxes to a polaron localized predominantly on a single Ti atom with hopping only along the [001] direction (activation energy 39 meV, room-temperature mobility 4.4 × 10^{-2} cm²/Vs, in agreement with experiment). In contrast, the hole polaron in anatase is localized on a single O atom with transport primarily to second-nearest neighbors due to poor orbital overlap (activation energy 139 meV, mobility 1.4 × 10^{-3} cm²/Vs). The work claims a finite-temperature first-principles characterization of small polaron transport and a transferable methodology.

Significance. If the ML model transferability holds, the work provides a dynamical, finite-temperature characterization of anisotropic small-polaron transport in the two TiO2 polymorphs that is directly relevant to photoelectrochemical applications. The reported directional selectivity, activation energies, and mobilities constitute falsifiable predictions that can be tested against experiment or higher-level calculations. The introduction of a scalable MLMD framework for rare-event polaron hopping is a methodological contribution that could be applied to other small-polaron materials.

major comments (2)
  1. [Methods] Methods section (DeepPolaron training and validation): The central claim that the model reproduces adiabatic localization and long-time hopping 'with virtually negligible loss in accuracy' over three orders of magnitude in time is load-bearing for all quantitative results (39 meV and 139 meV barriers, reported mobilities). No direct side-by-side comparison of hop frequencies, mean-squared displacements, or diffusion constants between MLMD trajectories and extended DFT runs on any overlapping timescale is presented, leaving the activation energies and mobilities dependent on an untested extrapolation from short-trajectory training data.
  2. [Results] Results section (rutile electron and anatase hole paragraphs): The reported mobilities and activation energies are given without error bars, standard deviations from multiple independent trajectories, or sensitivity analysis to training-set composition or model hyperparameters. This is load-bearing because the headline numbers (4.4 × 10^{-2} and 1.4 × 10^{-3} cm²/Vs) are compared directly to experiment; absence of uncertainty quantification prevents assessment of whether the agreement is statistically meaningful.
minor comments (2)
  1. [Abstract] Abstract: The phrase 'good agreement with experiment' for the rutile electron mobility should specify the experimental reference values and temperature range being compared.
  2. Notation: The definition of the polaron localization (single Ti vs. single O) and the distinction between first- and second-nearest-neighbor hops would benefit from an explicit equation or figure panel showing the relevant Ti–Ti or O–O distances and orbital overlaps.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their detailed and constructive report. We address each major comment below and outline revisions that will strengthen the manuscript.

read point-by-point responses

  1. Referee: [Methods] Methods section (DeepPolaron training and validation): The central claim that the model reproduces adiabatic localization and long-time hopping 'with virtually negligible loss in accuracy' over three orders of magnitude in time is load-bearing for all quantitative results (39 meV and 139 meV barriers, reported mobilities). No direct side-by-side comparison of hop frequencies, mean-squared displacements, or diffusion constants between MLMD trajectories and extended DFT runs on any overlapping timescale is presented, leaving the activation energies and mobilities dependent on an untested extrapolation from short-trajectory training data.

    Authors: We agree that direct comparison of long-time observables on overlapping timescales would provide stronger evidence for the extrapolation. The current validation focuses on short-trajectory agreement in forces, energies, and instantaneous polaron localization. Extended DFT-MD runs capable of sampling rare hops remain computationally prohibitive, which motivates the ML approach. In the revision we will add short-time mean-squared displacement comparisons between MLMD and DFT where overlap exists, and explicitly discuss the limits of validation for rare-event statistics. revision: yes

  2. Referee: [Results] Results section (rutile electron and anatase hole paragraphs): The reported mobilities and activation energies are given without error bars, standard deviations from multiple independent trajectories, or sensitivity analysis to training-set composition or model hyperparameters. This is load-bearing because the headline numbers (4.4 × 10^{-2} and 1.4 × 10^{-3} cm²/Vs) are compared directly to experiment; absence of uncertainty quantification prevents assessment of whether the agreement is statistically meaningful.

    Authors: We acknowledge that the absence of uncertainty estimates weakens the comparison to experiment. In the revised manuscript we will report standard deviations obtained from multiple independent MLMD trajectories for both activation energies and mobilities, together with a brief sensitivity analysis to training-set size and key hyperparameters. revision: yes

Circularity Check

0 steps flagged

No circularity: polaron localization, hop directions, barriers and mobilities are extracted from MLMD trajectories and compared to external experiment

full rationale

The paper trains an ML potential on short DFT trajectories and then runs long MLMD to observe adiabatic localization and rare-event hops. Activation energies (39 meV, 139 meV) and mobilities are computed directly from the resulting trajectories (hop frequencies, diffusion constants) rather than being fitted parameters or self-referenced quantities. Results are validated against independent experimental values rather than against the training data itself. No self-citation is load-bearing for the central claims, no ansatz is smuggled, and no uniqueness theorem is invoked. The derivation chain therefore remains independent of its own inputs.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 1 invented entities

Abstract-only view limits visibility into exact parameters; main additions are the new ML framework and the domain assumption of adiabatic dynamics. No explicit free parameters or invented physical entities beyond the method itself are described.

free parameters (1)
  • DeepPolaron model parameters
    Hyperparameters and training details of the machine learning potential are fitted to DFT data but not quantified in the abstract.
axioms (1)
  • domain assumption Polaron transport is adiabatic
    The framework is explicitly for adiabatic polaron transport as stated in the abstract.
invented entities (1)
  • DeepPolaron no independent evidence
    purpose: Machine learning framework to accelerate first-principles MD of polaron transport
    New method introduced to extend simulation timescales by three orders of magnitude.

pith-pipeline@v0.9.1-grok · 5814 in / 1477 out tokens · 40623 ms · 2026-06-28T14:03:46.523337+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

300 extracted references · 264 canonical work pages · 25 internal anchors

  1. [1]

    Accelerating Materials Screening with Machine Learning , year = 2018, file =

    2018

  2. [2]

    A Linear Response Approach to the Calculation of the Effective Interaction Parameters in the

  3. [3]

    Smeagol-1.2.Pdf , file =

  4. [4]

    Popelier, Paul , doi =

  5. [5]

    Structural- and Optical-Properties Analysis of Single Crystalline Hematite (

    Abdul Rashid, Nur Maisarah and Haw, Choonyian and Chiu, Weesiong and Khanis, Noor Hamizah and Rohaizad, Aliff and Khiew, Poisim and Abdul Rahman, Saadah , year = 2016, journal =. Structural- and Optical-Properties Analysis of Single Crystalline Hematite (. doi:10.1039/c6ce00573j , abstract =

    work page doi:10.1039/c6ce00573j 2016

  6. [6]

    Principles of Dynamic Nuclear Polarisation , author =. Rep. Prog. Phys. , volume =. doi:10.1088/0034-4885/41/3/002 , abstract =

    work page doi:10.1088/0034-4885/41/3/002

  7. [7]

    The Journal of Chemical Physics , volume =

    Adamo, Carlo and Barone, Vincenzo , year = 1999, journal =. Toward Reliable Density Functional Methods without Adjustable Parameters:. doi:10.1063/1.478522 , abstract =

    work page doi:10.1063/1.478522 1999

  8. [8]

    Adams, George F and Bartlett, Rodney J and Pljrvls, George D , year = 1982, journal =. On the

    1982

  9. [9]

    Density Functional Theory Based Calculation of Small-Polaron Mobility in Hematite , author =. Phys. Rev. B - Condens. Matter Mater. Phys. , volume =. doi:10.1103/PhysRevB.89.245115 , abstract =

    work page doi:10.1103/physrevb.89.245115

  10. [10]

    and Curtarolo, Stefano and Buongiorno Nardelli, Marco , year = 2015, month = jan, journal =

    Agapito, Luis A. and Curtarolo, Stefano and Buongiorno Nardelli, Marco , year = 2015, month = jan, journal =. Reformulation of. doi:10.1103/PhysRevX.5.011006 , urldate =

    work page doi:10.1103/physrevx.5.011006 2015

  11. [11]

    , author =

    Finding Reaction Paths Using the Potential Energy as Reaction Coordinate. , author =. J. Chem. Phys. , volume =. doi:10.1063/1.2834930 , abstract =. 18345872 , eprinttype =

    work page doi:10.1063/1.2834930

  12. [12]

    and Rosso, Kevin M

    Ahart, Christian S. and Rosso, Kevin M. and Blumberger, Jochen , year = 2022, month = mar, journal =. Electron and. doi:10.1021/jacs.1c13507 , urldate =

    work page doi:10.1021/jacs.1c13507 2022

  13. [13]

    and Chulkov, Sergey K

    Ahart, Christian S. and Chulkov, Sergey K. and Cucinotta, Clotilde S. , year = 2024, month = aug, journal =. Enabling. doi:10.1021/acs.jctc.4c00371 , urldate =

    work page doi:10.1021/acs.jctc.4c00371 2024

  14. [14]

    and Rosso, Kevin M

    Ahart, Christian S. and Rosso, Kevin M. and Blumberger, Jochen , year = 2022, month = jul, journal =. Implementation and. doi:10.1021/acs.jctc.2c00284 , urldate =

    work page doi:10.1021/acs.jctc.2c00284 2022

  15. [15]

    Polaronic Structure of Excess Electrons and Holes for a Series of Bulk Iron Oxides , author =. Phys. Chem. Chem. Phys. , volume =. doi:10.1039/c9cp06482f , abstract =

    work page doi:10.1039/c9cp06482f

  16. [16]

    Thesis of

    Ahart, Christian , langid =. Thesis of

  17. [17]

    Supplementary

    Ahart, Christian S and Blumberger, Jochen and Rosso, Kevin M , pages =. Supplementary. doi:10.1039/cXCP00000x/neering , file =

    work page doi:10.1039/cxcp00000x/neering

  18. [18]

    Npj Comput

    Revealing and Exploiting Hierarchical Material Structure through Complex Atomic Networks , author =. Npj Comput. Mater. , volume =. doi:10.1038/s41524-017-0035-x , abstract =

    work page doi:10.1038/s41524-017-0035-x

  19. [19]

    Albanese, Davide and Visintainer, Roberto and Merler, Stefano , eprint =. Mlpy :

  20. [20]

    Learning

    Alber, Maximilian and Sch, Kristof T and Sven, D , year = 2015, eprint =. Learning

    2015

  21. [21]

    and Tew, David P

    Alborzpour, Jonathan P. and Tew, David P. and Habershon, Scott , year = 2016, journal =. Efficient and Accurate Evaluation of Potential Energy Matrix Elements for Quantum Dynamics Using. doi:10.1063/1.4964902 , abstract =

    work page doi:10.1063/1.4964902 2016

  22. [22]

    Electron Transport in Pure and Substituted Iron Oxyhydroxides by Small-Polaron Migration , author =. J. Chem. Phys. , volume =. doi:10.1063/1.4882065 , abstract =

    work page doi:10.1063/1.4882065

  23. [23]

    , year = 2015, journal =

    Alexandrov, Vitaly and Rosso, Kevin M. , year = 2015, journal =. Ab Initio Modeling of. doi:10.1039/c5cp00921a , abstract =

    work page doi:10.1039/c5cp00921a 2015

  24. [24]

    Silicon-Doped Hafnium Oxide Anti-Ferroelectric Thin Films for Energy Storage , author =. J. Appl. Phys. , volume =. doi:10.1063/1.4989908 , urldate =

    work page doi:10.1063/1.4989908

  25. [25]

    Theory and

    Almeshal, Abdelkareem Mohammed , langid =. Theory and

  26. [26]

    and Ananth, Nandini and Angulo, Gonzalo and Astumian, Raymond Dean and Beniwal, Vijay and Blumberger, Jochen and Bolhuis, Peter G

    Althorpe, Stuart C. and Ananth, Nandini and Angulo, Gonzalo and Astumian, Raymond Dean and Beniwal, Vijay and Blumberger, Jochen and Bolhuis, Peter G. and Ensing, Bernd and Glowacki, David R. and Habershon, Scott and. Non-Adiabatic Reactions:. Faraday Discuss. , volume =. doi:10.1039/c6fd90078j , abstract =

    work page doi:10.1039/c6fd90078j

  27. [27]

    and Rosasco, Lorenzo and Lawrence, Neil D

    Alvarez, Mauricio A. and Rosasco, Lorenzo and Lawrence, Neil D. , year = 2011, eprint =. Kernels for. doi:10.1561/2200000036 , abstract =

    work page doi:10.1561/2200000036 2011

  28. [28]

    Optical Properties of Rutile near Its Fundamental Band Gap , author =. Phys. Rev. B , volume =. doi:10.1103/physrevb.51.6842 , urldate =

    work page doi:10.1103/physrevb.51.6842

  29. [29]

    Brief Report:

    Anderson, Don O N L O N L and Anderson, Orson L , year = 1970, journal =. Brief Report:. doi:10.1029/JB075i017p03494 , abstract =

    work page doi:10.1029/jb075i017p03494 1970

  30. [30]

    and Lundqvist, B

    Andersson, Y and Langreth, D. and Lundqvist, B. , year = 1996, journal =. Van Der. doi:10.1103/PhysRevLett.76.102 , abstract =

    work page doi:10.1103/physrevlett.76.102 1996

  31. [31]

    F\"orster interaction induced phase shift in a pair state interferometer

    Nuclear Magnetic Resonance Spectra from a Crystal Rotated at High Speed , author =. Nature , volume =. doi:10.1038/1821659a0 , abstract =. arXiv , isbn =:1203.1851v1 , pages =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1038/1821659a0

  32. [32]

    Nuclear Magnetic Resonance Spectra of Rapidly-Rotated Solids Containing Reorienting Molecular Groups , author =. J. Phys. Chem. Solid St. Phys. , volume =. doi:10.1088/0022-3719/4/3/014 , file =

    work page doi:10.1088/0022-3719/4/3/014

  33. [33]

    Andrew, E. R. , year = 1981, month = mar, journal =. Magic. doi:10.1098/rsta.1981.0032 , urldate =

    work page doi:10.1098/rsta.1981.0032 1981

  34. [34]

    Anisimov, V. I. and Solovyev, I. V. and Korotin, M. A. and Czy. Density-Functional Theory and. Phys. Rev. B , volume =. doi:10.1103/PhysRevB.48.16929 , urldate =

    work page doi:10.1103/physrevb.48.16929

  35. [35]

    and Zaanen, Jan and Andersen, Ole K

    Anisimov, Vladimir I. and Zaanen, Jan and Andersen, Ole K. , year = 1991, journal =. Band Theory and. doi:10.1103/PhysRevB.44.943 , abstract =

    work page doi:10.1103/physrevb.44.943 1991

  36. [36]

    and Aryasetiawan, F

    Anisimov, Vladimir I. and Aryasetiawan, F. and Lichtenstein, A. I. , year = 1997, journal =. First-Principles Calculations of the Electronic Structure and Spectra of Strongly Correlated Systems:. doi:10.1088/0953-8984/9/4/002 , abstract =

    work page doi:10.1088/0953-8984/9/4/002 1997

  37. [37]

    Anitha, V. C. and Banerjee, Arghya Narayan and Joo, Sang Woo , year = 2015, month = dec, journal =. Recent Developments in. doi:10.1007/s10853-015-9303-7 , urldate =

    work page doi:10.1007/s10853-015-9303-7 2015

  38. [38]

    Journal of Catalysis , volume =

    The Design and Development of Highly Reactive Titanium Oxide Photocatalysts Operating under Visible Light Irradiation , author =. Journal of Catalysis , volume =. doi:10.1016/S0021-9517(02)00104-5 , urldate =

    work page doi:10.1016/s0021-9517(02)00104-5

  39. [39]

    Hole Localization in

    Ansari, Narjes and Ulman, Kanchan and Camellone, Matteo Farnesi and Seriani, Nicola and Gebauer, Ralph and Piccinin, Simone , year = 2017, journal =. Hole Localization in. doi:10.1103/PhysRevMaterials.1.035404 , abstract =

    work page doi:10.1103/physrevmaterials.1.035404 2017

  40. [40]

    Antoniewicz, P. R. and Kleinman, Leonard , year = 1985, journal =. Kohn-. doi:10.1103/PhysRevB.31.6779 , abstract =

    work page doi:10.1103/physrevb.31.6779 1985

  41. [41]

    and Brinkman, W

    Appelbaum, Joel A. and Brinkman, W. F. , year = 1969, month = oct, journal =. Theory of. doi:10.1103/PhysRev.186.464 , urldate =

    work page doi:10.1103/physrev.186.464 1969

  42. [42]

    Arrigoni, Marco and Madsen, Georg K. H. , year = 2020, month = jan, journal =. A Comparative First-Principles Investigation on the Defect Chemistry of. doi:10.1063/1.5138902 , urldate =

    work page doi:10.1063/1.5138902 2020

  43. [43]

    J. Chem. Phys. , volume =. doi:10.1063/1.3617244 , urldate =

    work page doi:10.1063/1.3617244

  44. [44]

    Machine Learning methods for interatomic potentials: application to boron carbide

    Artrith, Nongnuch and Morawietz, Tobias and Behler, J. High-Dimensional Neural-Network Potentials for Multicomponent Systems:. Phys. Rev. B - Condens. Matter Mater. Phys. , volume =. doi:10.1103/PhysRevB.83.153101 , abstract =. arXiv , isbn =:1512.09110 , pages =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1103/physrevb.83.153101

  45. [45]

    and Taga, Y

    Asahi, R. and Taga, Y. and Mannstadt, W. and Freeman, A. J. , year = 2000, month = mar, journal =. Electronic and Optical Properties of Anatase. doi:10.1103/PhysRevB.61.7459 , urldate =

    work page doi:10.1103/physrevb.61.7459 2000

  46. [46]

    Atomic Charges of the Water Molecule and the Water Dimer , author =. J. Phys. Chem. A , volume =. doi:10.1021/jp980574e , abstract =

    work page doi:10.1021/jp980574e

  47. [47]

    Samaneh and Mohammadizadeh, M

    Ataei, S. Samaneh and Mohammadizadeh, M. Reza and Seriani, Nicola , year = 2016, month = apr, journal =. Ab. doi:10.1021/acs.jpcc.6b00019 , urldate =

    work page doi:10.1021/acs.jpcc.6b00019 2016

  48. [48]

    and Ren, Xinguo and Rinke, Patrick and Scheffler, Matthias , year = 2016, journal =

    Atalla, Viktor and Zhang, Igor Ying and Hofmann, Oliver T. and Ren, Xinguo and Rinke, Patrick and Scheffler, Matthias , year = 2016, journal =. Enforcing the Linear Behavior of the Total Energy with Hybrid Functionals:. doi:10.1103/PhysRevB.94.035140 , abstract =

    work page doi:10.1103/physrevb.94.035140 2016

  49. [49]

    Polarons in Crystalline and Non-Crystalline Materials , author =. Adv. Phys. , volume =. doi:10.1080/00018736900101267 , urldate =

    work page doi:10.1080/00018736900101267

  50. [50]

    A Classical Point Charge Model Study of System Size Dependence of Oxidation and Reorganization Free Energies in Aqueous Solution , author =. J. Phys. Chem. B , volume =. doi:10.1021/jp0748516 , abstract =

    work page doi:10.1021/jp0748516

  51. [51]

    Rutile, Anatase, Brookite and Titania Thin Film from

    Badalov, S V and Bocchini, A and Wilhelm, R and Kozub, A L and Gerstmann, U and Schmidt, W G , year = 2023, month = jul, journal =. Rutile, Anatase, Brookite and Titania Thin Film from. doi:10.1088/2053-1591/ace0fa , urldate =

    work page doi:10.1088/2053-1591/ace0fa 2023

  52. [52]

    npj Comput Mater , issn =

    Koopmans-Compliant Density Functional Framework for Polaron Self-Trapping in Titanate Oxides , author =. npj Comput Mater , issn =. doi:10.1038/s41524-026-02060-7 , urldate =

    work page doi:10.1038/s41524-026-02060-7

  53. [53]

    Bagrets, Alexei , year = 2013, month = jun, journal =. Spin-. doi:10.1021/ct4000263 , urldate =

    work page doi:10.1021/ct4000263 2013

  54. [54]

    Current-Induced Phonon Renormalization in Molecular Junctions , author =. Phys. Rev. B , volume =. doi:10.1103/PhysRevB.94.035411 , urldate =

    work page doi:10.1103/physrevb.94.035411

  55. [55]

    Thesis of

    Bailey, Steven , file =. Thesis of

  56. [56]

    Dynamic Nuclear Polarization at 9

    Bajaj, V S and Farrar, C T and Hornstein, M K and Mastovsky, I and Vieregg, J and Bryant, J and El. Dynamic Nuclear Polarization at 9. J. Magn. Reson. , volume =. doi:10.1016/S1090-7807(02)00192-1 , urldate =

    work page doi:10.1016/s1090-7807(02)00192-1

  57. [57]

    Bajaj, Vikram S and Hornstein, Melissa K and Kreischer, Kenneth E and Sirigiri, Jagadishwar R and Woskov, Paul P and Mak, Melody and Herzfeld, Judith and Temkin, Richard J and Griffin, Robert G , year = 2007, journal =. 250

    2007

  58. [58]

    and Nowotny, J

    Bak, T. and Nowotny, J. and Rekas, M. and Sorrell, C.C. , year = 2003, month = jul, journal =. Defect Chemistry and Semiconducting Properties of Titanium Dioxide:. doi:10.1016/S0022-3697(02)00481-X , urldate =

    work page doi:10.1016/s0022-3697(02)00481-x 2003

  59. [59]

    and Nowotny, M

    Bak, T. and Nowotny, M. K. and Sheppard, L. R. and Nowotny, J. , year = 2008, month = aug, journal =. Mobility of. doi:10.1021/jp801028j , urldate =

    work page doi:10.1021/jp801028j 2008

  60. [60]

    Oxide Chemistry and Catalysis , author =. J. Chem. Phys. , volume =. doi:10.1063/5.0021819 , abstract =

    work page doi:10.1063/5.0021819

  61. [61]

    Support Vector Machine Regression (

    Balabin, Roman M and Lomakina, Ekaterina I , year = 2011, journal =. Support Vector Machine Regression (. doi:10.1039/c1cp00051a , abstract =

    work page doi:10.1039/c1cp00051a 2011

  62. [62]

    J Mater Sci , volume =

    Electrical Conductivity in Non-Stoichiometric Titanium Dioxide at Elevated Temperatures , author =. J Mater Sci , volume =. doi:10.1007/BF00547436 , urldate =

    work page doi:10.1007/bf00547436

  63. [63]

    Discovery of the Soft Electronic Modes of the Trimeron Order in Magnetite , author =. Nat. Phys. , volume =. doi:10.1038/s41567-020-0823-y , urldate =

    work page doi:10.1038/s41567-020-0823-y

  64. [64]

    and Chiodo, L

    Baldini, E. and Chiodo, L. and Dominguez, A. and Palummo, M. and Moser, S. and. Strongly Bound Excitons in Anatase. Nat Commun , volume =. doi:10.1038/s41467-017-00016-6 , urldate =

    work page doi:10.1038/s41467-017-00016-6

  65. [65]

    and Schieber, M

    Balog, M. and Schieber, M. and Michman, M. and Patai, S. , year = 1977, month = mar, journal =. Chemical Vapor Deposition and Characterization of. doi:10.1016/0040-6090(77)90312-1 , urldate =

    work page doi:10.1016/0040-6090(77)90312-1 1977

  66. [66]

    Baraff, G. A. and Appelbaum, Joel A. and Hamann, D. R. , year = 1977, month = jan, journal =. Self-. doi:10.1103/PhysRevLett.38.237 , urldate =

    work page doi:10.1103/physrevlett.38.237 1977

  67. [67]

    Contemporary Issues in Electron Transfer Research , author =. J. Phys. Chem. , volume =. doi:10.1021/jp9605663 , abstract =

    work page doi:10.1021/jp9605663

  68. [68]

    and Faulkner, Larry R

    Bard, Allen J. and Faulkner, Larry R. , year = 2001, edition =. Electrochemical Methods: Fundamentals and Applications , shorttitle =

    2001

  69. [69]

    , year = 1961, month = jan, journal =

    Bardeen, J. , year = 1961, month = jan, journal =. Tunnelling from a. doi:10.1103/PhysRevLett.6.57 , urldate =

    work page doi:10.1103/physrevlett.6.57 1961

  70. [70]

    Barkoutsos, Jerome F

    Barkoutsos, Panagiotis Kl and Gonthier, Jerome F. and Sokolov, Igor and Moll, Nikolaj and Salis, Gian and Fuhrer, Andreas and Ganzhorn, Marc and Egger, Daniel J. and Troyer, Matthias and Mezzacapo, Antonio and Filipp, Stefan and Tavernelli, Ivano , year = 2018, journal =. Quantum Algorithms for Electronic Structure Calculations:. doi:10.1103/PhysRevA.98.0...

    work page doi:10.1103/physreva.98.022322 2018

  71. [71]

    Barnett, Robert N. and H. Hydrogen. Nano Lett. , volume =. doi:10.1021/nl049054n , urldate =

    work page doi:10.1021/nl049054n

  72. [72]

    Low-Dimensional Semiconductor Structures:

  73. [73]

    Gaussian

    Bart. Gaussian. 1003.2817v1 , archiveprefix =

    Pith/arXiv arXiv

  74. [74]

    Gaussian Approximation Potentials: the accuracy of quantum mechanics, without the electrons

    Bart. Gaussian Approximation Potentials:. Phys. Rev. Lett. , volume =. doi:10.1103/PhysRevLett.104.136403 , abstract =. arXiv , isbn =:0910.1019 , pages =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1103/physrevlett.104.136403

  75. [75]

    Gaussian Approximation Potentials:

    Bart. Gaussian Approximation Potentials:. Int. J. Quantum Chem. , volume =. doi:10.1002/qua.24927 , abstract =. arXiv , isbn =:1502.01366 , pages =

    work page doi:10.1002/qua.24927

  76. [76]

    Bart. Machine. Phys. Rev. X , volume =. doi:10.1103/PhysRevX.8.041048 , abstract =. arXiv , keywords =:1805.01568 , pages =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1103/physrevx.8.041048

  77. [77]

    Machine-Learning Approach for One- and Two-Body Corrections to Density Functional Theory:

    Bart. Machine-Learning Approach for One- and Two-Body Corrections to Density Functional Theory:. Phys. Rev. B - Condens. Matter Mater. Phys. , volume =. doi:10.1103/PhysRevB.88.054104 , abstract =

    work page doi:10.1103/physrevb.88.054104

  78. [78]

    On Representing Chemical Environments , author =. Phys. Rev. B - Condens. Matter Mater. Phys. , volume =. doi:10.1103/PhysRevB.87.184115 , abstract =. arXiv , isbn =:1209.3140 , pages =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1103/physrevb.87.184115

  79. [79]

    Regularized

    Bart. Regularized. J. Chem. Phys. , volume =. doi:10.1063/1.5094646 , abstract =. arXiv , file =:1903.01007 , pages =

    work page doi:10.1063/1.5094646 1903

  80. [80]

    Stability of Non-Metal Dopants to Tune the Photo-Absorption of

    Basera, Pooja and Saini, Shikha and Arora, Ekta and Singh, Arunima and Kumar, Manish and Bhattacharya, Saswata , year = 2019, month = aug, journal =. Stability of Non-Metal Dopants to Tune the Photo-Absorption of. doi:10.1038/s41598-019-47710-7 , urldate =

    work page doi:10.1038/s41598-019-47710-7 2019

Showing first 80 references.