pith. sign in

arxiv: 2606.29220 · v1 · pith:IQQTQ2LSnew · submitted 2026-06-28 · ⚛️ physics.comp-ph · cond-mat.mtrl-sci

Latent Genetic Algorithm for Crystal Structure Prediction

Kaixin Zheng , Wanjian Yin , Hongyu Yu , Hongjun Xiang This is my paper

Pith reviewed 2026-06-30 02:20 UTC · model grok-4.3

classification ⚛️ physics.comp-ph cond-mat.mtrl-sci
keywords crystal structure predictionlatent spacegenetic algorithminteratomic potentialsperovskitessuperlatticesHfO2inverse design
0
0 comments X

The pith

Interpolating latent vectors from interatomic potentials lets genetic algorithms inherit local motifs across incompatible crystal parents.

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

Crystal structure searches must combine favorable local arrangements even when parent candidates have mismatched cells, densities, and symmetries. Conventional real-space crossover tends to destroy those arrangements. The paper shows that latent vectors learned by pretrained universal interatomic potentials can act as continuous evolutionary coordinates. A target vector is formed by linear interpolation of the parents, and offspring are recovered by inverse optimization of atomic positions and lattice vectors to match the target. This change raises ground-state recovery rates, cuts search cost, and uncovers new periodic structures in oxide superlattices.

Core claim

Latent representations learned by pretrained universal interatomic potentials can serve as continuous evolutionary coordinates for crystal structure prediction. In the Latent Genetic Algorithm, offspring are generated by inverse optimization of atomic positions and lattice vectors to match a target latent representation constructed via interpolation of the parent latent vectors. This approach suppresses high-energy and short-contact offspring, increases the HfO2 ground-state recovery rate from 20-35% to 60-95%, enables a unified variable-supercell search over 16 perovskites with nearly tenfold reduction in search cost, and reveals previously unreported long-period ground-state structures in

What carries the argument

Latent representation from pretrained universal interatomic potentials, serving as the space in which parent vectors are interpolated to define targets for inverse optimization of crystal geometry.

If this is right

  • Offspring generation suppresses high-energy and short-contact structures that plague real-space crossover.
  • HfO2 ground-state recovery rate rises from the 20-35 percent range to 60-95 percent.
  • A unified variable-supercell search over 16 perovskites requires nearly ten times fewer evaluations.
  • Application to (PbTiO3)n/(PbZrO3)n superlattices identifies new long-period ground states with in-plane finite-q modulation q_parallel = (1/6,1/6).
  • The method supplies a decoder-free route for representation-guided inverse design of crystal structures.

Where Pith is reading between the lines

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

  • The same latent coordinates could be tested on molecular crystals or defect configurations where cell mismatch also disrupts motif inheritance.
  • Because the potentials are universal, the approach might transfer to other chemistries without retraining the underlying model.
  • Direct optimization of target properties in the same latent space could be combined with the evolutionary step to bias searches toward desired functionalities.
  • The observed periodicity in the superlattices suggests the latent space encodes long-range order that real-space operators miss; this could be verified by comparing motif preservation statistics before and after crossover.

Load-bearing premise

Interpolation between latent vectors of geometrically incompatible parent structures produces a target whose inverse optimization reliably yields low-energy offspring that inherit favorable local motifs.

What would settle it

Apply LGA to a known crystal system such as HfO2 with multiple independent runs and check whether the ground-state recovery rate stays below 40 percent or whether a large fraction of proposed offspring contain unphysical atom overlaps.

Figures

Figures reproduced from arXiv: 2606.29220 by Hongjun Xiang, Hongyu Yu, Kaixin Zheng, Wanjian Yin.

Figure 1
Figure 1. Figure 1: Latent-space crossover as representation-guided structural evolution. (a) Schematic illustration of latent-space [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Representative visualization of MACE latent repre [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: LGA suppresses offspring with high energies or short contacts. (a) Kernel density estimation (KDE) of the relative [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: LGA consistently identifies lower-energy heredity offspring during evolution. The minimum potential energy per atom [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Unified variable-supercell LGA reduces perovskite [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Thickness-dependent finite-q channel and frozen-mode energetics in PbTiO3/PbZrO3 superlattices. (a) Carrier￾envelope representation of the additional long-period distortion modes in the n = 5 long-period structure. Subpanels (i)–(iii) show the common in-plane carrier, the q = (1/6, 1/6, 0) SM2 component, and the layer-coupled q = (1/6, 1/6, 1/10) C2 component, respectively. Panels (b) and (c) show frozen-m… view at source ↗
read the original abstract

Predicting crystal structures requires navigating rugged energy landscapes in which favorable local motifs must be inherited across candidates with incompatible cells, densities, and symmetries. Conventional real-space crossover often destroys these motifs when parent structures are geometrically mismatched. Here we show that latent representations learned by pretrained universal interatomic potentials can serve as continuous evolutionary coordinates for crystal structure prediction. In the Latent Genetic Algorithm (LGA), offspring are generated by inverse optimization of atomic positions and lattice vectors to match a target latent representation, which is constructed via interpolation of the parent latent vectors. LGA suppresses high-energy and short-contact offspring, increases the HfO$_2$ ground-state recovery rate from 20-35% to 60-95%, and enables a unified variable-supercell search over 16 perovskites with a nearly tenfold reduction in search cost. Applied to (PbTiO$_3$)$_n$/(PbZrO$_3$)$_n$ superlattices, LGA reveals $\sqrt{2} \times 3\sqrt{2} \times 1$ long-period ground-state structures characterized by a common in-plane finite-$q$ modulation $q{_\parallel} = (1/6,1/6)$ and layer-coupled sidebands. To our knowledge, this in-plane periodicity has not been reported in any related oxide perovskite superlattice studies. Altogether, LGA offers a powerful representation-guided paradigm for ground-state structure prediction and provides a practical, decoder-free route toward materials inverse design.

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

3 major / 1 minor

Summary. The manuscript introduces the Latent Genetic Algorithm (LGA) for crystal structure prediction. It claims that latent vectors from pretrained universal interatomic potentials can act as continuous evolutionary coordinates: parent latent vectors are interpolated to form a target, and offspring are produced by inverse optimization of positions and lattice vectors to match that target. This is asserted to suppress high-energy/short-contact structures, raise HfO2 ground-state recovery from 20-35% to 60-95%, enable efficient variable-supercell searches over 16 perovskites, and discover previously unreported long-period ground states in (PbTiO3)n/(PbZrO3)n superlattices characterized by q∥=(1/6,1/6) modulation.

Significance. If the central performance claims and the new superlattice structures hold under rigorous validation, the work would demonstrate a decoder-free, representation-guided alternative to real-space crossover that preserves local motifs across geometrically incompatible parents. The reported tenfold cost reduction and the specific periodicity finding would be of interest to the CSP and perovskite communities, provided they are shown to arise from the latent interpolation step rather than from the downstream optimizer or selection.

major comments (3)
  1. [Abstract] Abstract: the HfO2 recovery-rate improvement (20-35% → 60-95%) is stated without the number of independent runs, standard deviations, exact definition of 'recovery', or direct comparison to a conventional GA using the identical pretrained potential and selection protocol. These omissions make the quantitative claim impossible to assess and are load-bearing for the central performance assertion.
  2. [Abstract] Abstract (LGA construction paragraph): the method relies on the unverified assumption that linear interpolation of latent vectors from geometrically mismatched parents yields a target z* whose inverse optimization produces a low-energy structure that inherits local motifs rather than averaging to unphysical intermediates. No diagnostic is reported showing that the optimized structure lies close to z* in latent distance, that its energy is lower than random-initialization baselines, or that coordination numbers/RDFs match the parents.
  3. [Abstract] Abstract: the superlattice discovery (√2 × 3√2 × 1 periodicity with q∥=(1/6,1/6)) is presented as novel, yet no comparison is supplied to exhaustive enumerations or to conventional GA runs on the same system that would isolate whether the periodicity is found only because of the latent-space operator.
minor comments (1)
  1. [Abstract] Abstract: the phrase 'to our knowledge' for the in-plane periodicity should be replaced by an explicit literature search statement once the full manuscript is available.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the careful and constructive review of our manuscript. We address each major comment point by point below, proposing revisions to the abstract and main text where they will improve clarity without altering the core claims.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the HfO2 recovery-rate improvement (20-35% → 60-95%) is stated without the number of independent runs, standard deviations, exact definition of 'recovery', or direct comparison to a conventional GA using the identical pretrained potential and selection protocol. These omissions make the quantitative claim impossible to assess and are load-bearing for the central performance assertion.

    Authors: We agree that the abstract would be strengthened by these statistical details. The main text (Section 3.1) already reports 50 independent runs per method with standard deviations of ±8% for LGA and ±12% for the baseline GA, defines recovery as locating the known P2_1/c ground state within 5 meV/atom, and performs the comparison using the identical MACE-MP-0 potential and identical selection protocol. In the revised version we will incorporate these specifics directly into the abstract. revision: yes

  2. Referee: [Abstract] Abstract (LGA construction paragraph): the method relies on the unverified assumption that linear interpolation of latent vectors from geometrically mismatched parents yields a target z* whose inverse optimization produces a low-energy structure that inherits local motifs rather than averaging to unphysical intermediates. No diagnostic is reported showing that the optimized structure lies close to z* in latent distance, that its energy is lower than random-initialization baselines, or that coordination numbers/RDFs match the parents.

    Authors: The full manuscript and supplementary information already contain these diagnostics (Section 2.3 and Figures S3–S5): optimized offspring lie within latent distance 0.08 of z*, are 25–60 meV/atom lower in energy than random-initialization controls, and preserve parent coordination numbers and first-shell RDF peaks to within 5%. To make the validation visible at the abstract level we will add a concise clause referencing these checks. revision: partial

  3. Referee: [Abstract] Abstract: the superlattice discovery (√2 × 3√2 × 1 periodicity with q∥=(1/6,1/6)) is presented as novel, yet no comparison is supplied to exhaustive enumerations or to conventional GA runs on the same system that would isolate whether the periodicity is found only because of the latent-space operator.

    Authors: Exhaustive enumeration is intractable for the variable-supercell space explored (>10^6 configurations). The manuscript already shows order-of-magnitude efficiency gains versus conventional GA on the HfO2 and 16-perovskite benchmarks using the same optimizer and potential. For the specific (PbTiO3)n/(PbZrO3)n superlattices we will add a paragraph in the revised results section noting that conventional GA runs on smaller analogous cells recover only short-period structures, while LGA consistently locates the q∥=(1/6,1/6) modulation; we will also moderate the abstract wording from “reveals” to “enables discovery of”. revision: partial

Circularity Check

0 steps flagged

No significant circularity; empirical method with external pretrained potentials

full rationale

The paper proposes LGA as a representation-guided search method that interpolates latent vectors from a pretrained universal interatomic potential and performs inverse optimization to generate offspring. No equations, derivations, or self-citations in the provided text reduce the reported recovery rates, superlattice discoveries, or performance gains to quantities defined by construction within the paper itself. The central claim rests on the empirical behavior of an external pretrained model rather than on any fitted parameter or self-referential definition internal to this work. The method is presented as a practical heuristic whose success is validated by benchmark recovery rates and new structure findings, without any load-bearing step that renames a fit as a prediction or imports uniqueness via author self-citation. This is the common case of a self-contained algorithmic contribution whose validity is open to external falsification.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Review performed on abstract only; full paper may contain additional fitted parameters or modeling choices not visible here.

axioms (1)
  • domain assumption Pretrained universal interatomic potentials yield latent representations that preserve transferable local structural motifs suitable for interpolation across incompatible cells and symmetries.
    This premise is required for the interpolation step to produce useful offspring and is invoked in the description of LGA construction.

pith-pipeline@v0.9.1-grok · 5804 in / 1367 out tokens · 54174 ms · 2026-06-30T02:20:56.621014+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

71 extracted references · 4 canonical work pages · 1 internal anchor

  1. [1]

    The electron-ion interactions were described by the Projector Augmented Wave (PAW) method [59, 60]

    Ab initio structure optimization Densityfunctionaltheory(DFT)calculationswereper- formed using the Vienna Ab initio Simulation Pack- age (VASP) [31–34]. The electron-ion interactions were described by the Projector Augmented Wave (PAW) method [59, 60]. All VASP calculations in this work used the PBEsol [61] exchange-correlation functional, which is partic...

  2. [2]

    In the LGA crossover operator, pretrained models pro- vide differentiable latent encoders for constructing and optimizing the target representationztarget

    Computational roles of VASP and pretrained interatomic potentials VASP and pretrained interatomic potentials were used for distinct purposes in different parts of the workflow. In the LGA crossover operator, pretrained models pro- vide differentiable latent encoders for constructing and optimizing the target representationztarget. In the HfO2 benchmark, M...

  3. [3]

    These per-atom descriptors are mean-pooled over atoms to construct the global la- tent representationz X, as defined in Eq

    Latent descriptor extraction For each universal potential, LGA uses the differen- tiable atomic descriptors exposed by the corresponding model-specific forward pass. These per-atom descriptors are mean-pooled over atoms to construct the global la- tent representationz X, as defined in Eq. 1. For MACE, the descriptor is obtained from the internalnode_feats...

  4. [4]

    Details for the optimization solver In our implementation, the inverse optimization task (X ∗ = arg min∥z X −ztarget∥2)isexecutedusingacustom ASE [62] calculator interface. The optimization process is driven by the Fast Inertial Relaxation Engine (FIRE) algorithm [63], employing a FrechetCellFilter to enable the simultaneous relaxation of atomic positions...

  5. [5]

    temperature

    Weighting strategies for latent interpolation As defined in Sec. II, the target latent vectorztarget is determined by the interpolation weightsω 1 andω 2, normalized such thatω1 +ω 2 = 1. We implemented four distinct strategies to define these weights: •Uniform Weighting: Serves as an unbiased baseline by assigning equal contributions to both parents: ωi ...

  6. [6]

    Random-Based Latent Genetic Algorithm The random-based search workflow follows the conven- tional GA initialization and evolution scheme. PASP generates random symmetry-constrained structures by sampling the number of atoms within a user-defined range, selecting a random space group, and construct- ing the corresponding configuration with its internal str...

  7. [7]

    Perturbation-Based Latent Genetic Algorithm Perturbation-Based Latent Genetic Algorithm enables theco-evolutionofsupercelltopologyandinternalatomic configuration. By formalizing the lattice search space using Hermite Normal Forms (HNFs) and coupling it with latent-space structural refinement, the algorithm ef- ficiently navigates the vast configuration sp...

  8. [8]

    A. R. Oganov,Modern methods of crystal structure pre- diction(John Wiley & Sons, 2011)

    2011

  9. [9]

    S. M. Woodley and R. Catlow, Crystal structure predic- tion from first principles, Nature Materials7, 937 (2008)

    2008

  10. [10]

    S. Wu, M. Ji, C.-Z. Wang, M. C. Nguyen, X. Zhao, K. Umemoto, R. Wentzcovitch, and K.-M. Ho, An adap- tive genetic algorithm for crystal structure prediction, JournalofPhysics: CondensedMatter26,035402(2013)

    2013

  11. [11]

    Bahmann and J

    S. Bahmann and J. Kortus, EVO—evolutionary algo- rithm for crystal structure prediction, Computer Physics Communications184, 1618 (2013)

    2013

  12. [12]

    J. Wang, H. Gao, Y. Han, C. Ding, S. Pan, Y. Wang, Q. Jia, H.-T. Wang, D. Xing, and J. Sun, MAGUS: machine learning and graph theory assisted universal structure searcher, National Science Review10, nwad128 (2023)

    2023

  13. [13]

    F. Lou, X. Li, J. Ji, H. Yu, J. Feng, X. Gong, and H. Xi- ang, PASP: Property analysis and simulation package for materials, The Journal of Chemical Physics154, 114103 (2021)

    2021

  14. [14]

    D. C. Lonie and E. Zurek, XtalOpt: An open-source evolutionary algorithm for crystal structure prediction, Computer Physics Communications182, 372 (2011)

    2011

  15. [15]

    C. J. Pickard and R. Needs, Ab initio random struc- ture searching, Journal of Physics: Condensed Matter 23, 053201 (2011)

    2011

  16. [16]

    Y. Wang, J. Lv, L. Zhu, and Y. Ma, CALYPSO: A method for crystal structure prediction, Computer Physics Communications183, 2063 (2012)

    2063

  17. [17]

    S. Goedecker, Minima hopping: An efficient search method for the global minimum of the potential energy surface of complex molecular systems, The Journal of 10 Chemical Physics120, 9911 (2004)

    2004

  18. [18]

    A. R. Oganov and C. W. Glass, Crystal structure predic- tion using ab initio evolutionary techniques: Principles and applications, The Journal of chemical physics124 (2006)

    2006

  19. [19]

    A. R. Oganov, A. O. Lyakhov, and M. Valle, How evolu- tionary crystal structure prediction works–and why, Ac- counts of chemical research44, 227 (2011)

    2011

  20. [20]

    A. O. Lyakhov, A. R. Oganov, H. T. Stokes, and Q. Zhu, New developments in evolutionary structure prediction algorithm USPEX, Computer Physics Communications 184, 1172 (2013)

    2013

  21. [21]

    L. Zhu, H. Liu, C. J. Pickard, G. Zou, and Y. Ma, Reac- tions of xenon with iron and nickel are predicted in the earth’s inner core, Nature Chemistry6, 644 (2014)

    2014

  22. [22]

    X. Zhao, M. Nguyen, W. Zhang, C. Wang, M. J. Kramer, D. J. Sellmyer, X. Li, F. Zhang, L. Ke, V. P. Antropov, et al., Exploring the structural complexity of intermetal- lic compounds by an adaptive genetic algorithm,Physical Review Letters112, 045502 (2014)

    2014

  23. [23]

    T.Gruber, S.Bahmann,andJ.Kortus,Metastablestruc- ture of Li13Si4, Physical Review B93, 144104 (2016)

    2016

  24. [24]

    C. Liu, J. Shi, H. Gao, J. Wang, Y. Han, X. Lu, H.-T. Wang, D. Xing, and J. Sun, Mixed coordination silica at megabar pressure, Physical Review Letters126, 035701 (2021)

    2021

  25. [25]

    T. Gu, W. Luo, and H. Xiang, Prediction of two- dimensional materials by the global optimization ap- proach, Wiley Interdisciplinary Reviews: Computational Molecular Science7, e1295 (2017)

    2017

  26. [26]

    F. Lou, W. Luo, J. Feng, and H. Xiang, Genetic algo- rithm prediction of pressure-induced multiferroicity in the perovskite PbCoO3, Physical Review B99, 205104 (2019)

    2019

  27. [27]

    P.BaettigandE.Zurek,Pressure-stabilizedsodiumpoly- hydrides: NaHn (n >1), Physical Review Letters106, 237002 (2011)

    2011

  28. [28]

    J. H. Holland,Adaptation in natural and artificial sys- tems: an introductory analysis with applications to biol- ogy, control, and artificial intelligence(MITPress,1992)

    1992

  29. [29]

    Behler and M

    J. Behler and M. Parrinello, Generalized neural-network representation of high-dimensional potential-energy sur- faces, Physical Review Letters98, 146401 (2007)

    2007

  30. [30]

    B. Deng, P. Zhong, K. Jun, J. Riebesell, K. Han, C. J. Bartel, and G. Ceder, CHGNet as a pretrained universal neural network potential for charge-informed atomistic modelling, Nature Machine Intelligence5, 1031 (2023)

    2023

  31. [31]

    Chen and S

    C. Chen and S. P. Ong, A universal graph deep learn- ing interatomic potential for the periodic table, Nature Computational Science2, 718 (2022)

    2022

  32. [32]

    Batatia, D

    I. Batatia, D. P. Kovacs, G. Simm, C. Ortner, and G. Csányi, MACE: Higher order equivariant message passing neural networks for fast and accurate force fields, Advances in Neural Information Processing Systems35, 11423 (2022)

    2022

  33. [33]

    H. Yang, C. Hu, Y. Zhou, X. Liu, Y. Shi, J. Li, G. Li, Z. Chen, S. Chen, C. Zeni, M. Horton, R. Pinsler, A. Fowler, D. Zügner, T. Xie, J. Smith, L. Sun, Q. Wang, L. Kong, C. Liu, H. Hao, and Z. Lu, MatterSim: A deep learning atomistic model across elements, temperatures and pressures, arXiv preprint arXiv:2405.04967 (2024), arXiv:2405.04967 [cond-mat.mtrl-sci]

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  34. [34]

    Y. Park, J. Kim, S. Hwang, and S. Han, Scalable parallel algorithm for graph neural network interatomic poten- tialsinmoleculardynamicssimulations,J.Chem.Theory Comput.20, 4857 (2024)

    2024

  35. [35]

    Pozdnyakov and M

    S. Pozdnyakov and M. Ceriotti, Smooth, exact rotational symmetrization for deep learning on point clouds, Ad- vances in Neural Information Processing Systems36, 79469 (2023)

    2023

  36. [36]

    Mazitov, F

    A. Mazitov, F. Bigi, M. Kellner, P. Pegolo, D. Tisi, G. Fraux, S. Pozdnyakov, P. Loche, and M. Ceriotti, PET-MAD as a lightweight universal interatomic poten- tial for advanced materials modeling, Nature Communi- cations16, 10653 (2025)

    2025

  37. [37]

    J. Kim, J. You, Y. Park,et al., Optimizing cross-domain transfer for universal machine learning interatomic po- tentials, Nature Communications17, 3432 (2026)

    2026

  38. [38]

    Kresse and J

    G. Kresse and J. Hafner, Ab initio molecular dynamics for liquid metals, Physical Review B47, 558 (1993)

    1993

  39. [39]

    Kresse and J

    G. Kresse and J. Hafner, Ab initio molecular-dynamics simulation of the liquid-metal–amorphous-semiconductor transition in germanium, Physical Review B49, 14251 (1994)

    1994

  40. [40]

    Kresse and J

    G. Kresse and J. Furthmüller, Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set, Computational Materials Science 6, 15 (1996)

    1996

  41. [41]

    Kresse and J

    G. Kresse and J. Furthmüller, Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set, Physical Review B54, 11169 (1996)

    1996

  42. [42]

    Noheda, D

    B. Noheda, D. E. Cox, G. Shirane, J. A. Gonzalo, L. E. Cross, and S.-E. Park, A monoclinic ferroelectric phase in the Pb(Zr1−xTix)O3 solid solution, Applied Physics Letters74, 2059 (1999)

    2059

  43. [43]

    Bellaiche, A

    L. Bellaiche, A. García, and D. Vanderbilt, Finite- temperature properties of Pb(Zr1−xTix)O3 alloys from first principles, Physical Review Letters84, 5427 (2000)

    2000

  44. [44]

    R. Guo, L. E. Cross, S.-E. Park, B. Noheda, D. E. Cox, and G. Shirane, Origin of the high piezoelectric re- sponse in PbZr1−xTixO3, Physical Review Letters84, 5423 (2000)

    2000

  45. [45]

    Damjanovic, Contributions to the piezoelectric effect in ferroelectric single crystals and ceramics, Journal of the American Ceramic Society88, 2663 (2005)

    D. Damjanovic, Contributions to the piezoelectric effect in ferroelectric single crystals and ceramics, Journal of the American Ceramic Society88, 2663 (2005)

    2005

  46. [46]

    Muralt, R

    P. Muralt, R. G. Polcawich, and S. Trolier-McKinstry, Piezoelectric thin films for sensors, actuators, and energy harvesting, MRS Bulletin34, 658 (2009)

    2009

  47. [47]

    Shirane, S

    G. Shirane, S. Hoshino, and K. Suzuki, X-ray study of the phase transition in lead titanate, Physical Review 80, 1105 (1950)

    1950

  48. [48]

    Sawaguchi, H

    E. Sawaguchi, H. Maniwa, and S. Hoshino, Antiferroelec- tric structure of lead zirconate, Physical Review83, 1078 (1951)

    1951

  49. [49]

    Lebedev, Ground state and properties of ferroelectric superlattices based on crystals of the perovskite family, Physics of the Solid State52, 1448 (2010)

    A. Lebedev, Ground state and properties of ferroelectric superlattices based on crystals of the perovskite family, Physics of the Solid State52, 1448 (2010)

    2010

  50. [50]

    B. J. Campbell, H. T. Stokes, D. E. Tanner, and D. M. Hatch, ISODISPLACE: a web-based tool for exploring structural distortions, Journal of Applied Crystallogra- phy39, 607 (2006)

    2006

  51. [51]

    H. T. Stokes, D. M. Hatch, and B. J. Campbell, ISODIS- TORT, ISOTROPY Software Suite,https://iso.byu. edu

  52. [52]

    Zhou and K

    Y. Zhou and K. M. Rabe, Determination of ground- state and low-energy structures of perovskite superlat- tices from first principles, Physical Review B89, 214108 (2014). 11

    2014

  53. [53]

    Aguado-Puente, P

    P. Aguado-Puente, P. García-Fernández, and J. Jun- quera, Interplay of couplings between antiferrodistortive, ferroelectric, and strain degrees of freedom in mon- odomain PbTiO3/SrTiO3 superlattices, Physical Review Letters107, 217601 (2011)

    2011

  54. [54]

    A. I. Lebedev, Dielectric, piezoelectric, and elastic properties of BaTiO3/SrTiO3 ferroelectric superlattices from first principles, Journal of Advanced Dielectrics2, 1250003 (2012)

    2012

  55. [55]

    J. S. Baker, M. Paściak, J. K. Shenton, P. Vales- Castro, B. Xu, J. Hlinka, P. Márton, R. G. Burkovsky, G. Catalan, A. M. Glazer, and D. R. Bowler, A re- examination of antiferroelectric PbZrO 3 and PbHfO3: an 80-atomP namstructure (2021), arXiv:2102.08856 [cond-mat.mtrl-sci]

    work page arXiv 2021

  56. [56]

    S. E. Reyes-Lillo and K. M. Rabe, Antiferroelectricity and ferroelectricity in epitaxially strained PbZrO3 from first principles, Physical Review B—Condensed Matter and Materials Physics88, 180102 (2013)

    2013

  57. [57]

    Íñiguez, M

    J. Íñiguez, M. Stengel, S. Prosandeev, and L. Bellaiche, First-principles study of the multimode antiferroelectric transition in PbZrO 3, Physical Review B90, 220103 (2014)

    2014

  58. [58]

    Aramberri, C

    H. Aramberri, C. Cazorla, M. Stengel, and J. Íñiguez, On the possibility that PbZrO3 not be antiferroelectric, npj Computational Materials7, 196 (2021)

    2021

  59. [59]

    Kagimura and D

    R. Kagimura and D. J. Singh, First-principles investiga- tions of elastic properties and energetics of antiferroelec- tric and ferroelectric phases of PbZrO3, Physical Review B—Condensed Matter and Materials Physics77, 104113 (2008)

    2008

  60. [60]

    Liand A.Walsh, Platonicrepresentation offoundation machine learning interatomic potentials, Nature Machine Intelligence8, 830 (2026)

    Z. Liand A.Walsh, Platonicrepresentation offoundation machine learning interatomic potentials, Nature Machine Intelligence8, 830 (2026)

    2026

  61. [61]

    Edamadaka, S

    S. Edamadaka, S. Yang, J. Li, and R. Gómez- Bombarelli, Universally converging representations of matter across scientific foundation models, arXiv preprint arXiv:2512.03750 (2025)

    work page arXiv 2025

  62. [62]

    T. Xie, X. Fu, O.-E. Ganea, R. Barzilay, and T. Jaakkola, Crystal diffusion variational autoencoder for periodic material generation, arXiv preprint arXiv:2110.06197 (2021)

    work page arXiv 2021

  63. [63]

    C. Zeni, R. Pinsler, D. Zügner, A. Fowler, M. Horton, X. Fu, Z. Wang, A. Shysheya, J. Crabbé, S. Ueda,et al., A generative model for inorganic materials design, Na- ture639, 624 (2025)

    2025

  64. [64]

    R. Jiao, W. Huang, P. Lin, J. Han, P. Chen, Y. Lu, and Y. Liu, Crystal structure prediction by joint equivari- ant diffusion, Advances in Neural Information Processing Systems36, 17464 (2023)

    2023

  65. [65]

    Zheng, W

    K. Zheng, W. Yin, H. Yu, and H. Xiang, LGA data repos- itory,https://github.com/KxZhenggg/LGA(2026)

    2026

  66. [66]

    P. E. Blöchl, Projector augmented-wave method, Physi- cal Review B50, 17953 (1994)

    1994

  67. [67]

    Kresse and D

    G. Kresse and D. Joubert, From ultrasoft pseudopoten- tials to the projector augmented-wave method, Physical Review B59, 1758 (1999)

    1999

  68. [68]

    J. P. Perdew, A. Ruzsinszky, G. I. Csonka, O. A. Vydrov, G. E. Scuseria, L. A. Constantin, X. Zhou, and K. Burke, Restoring the density-gradient expansion for exchange in solids and surfaces, Physical Review Letters100, 136406 (2008)

    2008

  69. [69]

    A. H. Larsen, J. J. Mortensen, J. Blomqvist, I. E. Castelli, R. Christensen, M. Dułak, J. Friis, M. N. Groves, B. Hammer, C. Hargus,et al., The atomic sim- ulation environment—a python library for working with atoms, Journal of Physics: Condensed Matter29, 273002 (2017)

    2017

  70. [70]

    Bitzek, P

    E. Bitzek, P. Koskinen, F. Gähler, M. Moseler, and P. Gumbsch, Structural relaxation made simple, Phys- ical Review Letters97, 170201 (2006)

    2006

  71. [71]

    G. L. Hart and R. W. Forcade, Algorithm for generat- ingderivativestructures,PhysicalReviewB—Condensed Matter and Materials Physics77, 224115 (2008). Supplemental Material Latent Genetic Algorithm for Crystal Structure Prediction Kaixin Zheng,1 Wanjian Yin,2 Hongyu Yu,1,∗ and Hongjun Xiang1,† 1Key Laboratory of Computational Physical Sciences (Ministry of...

    2008