Insightful classification of crystal structures using deep learning

Abstract

Computational methods that automatically extract knowledge from data are critical for enabling data-driven materials science. A reliable identification of lattice symmetry is a crucial first step for materials characterization and analytics. Current methods require a user-specified threshold, and are unable to detect average symmetries for defective structures. Here, we propose a machine learning-based approach to automatically classify structures by crystal symmetry. First, we represent crystals by calculating a diffraction image, then construct a deep learning neural network model for classification. Our approach is able to correctly classify a dataset comprising more than 100,000 simulated crystal structures, including heavily defective ones. The internal operations of the neural network are unraveled through attentive response maps, demonstrating that it uses the same landmarks a materials scientist would use, although never explicitly instructed to do so. Our study paves the way for crystal structure recognition of—possibly noisy and incomplete—three-dimensional structural data in big-data materials science.

Full text

Its licence does not allow us to show the full text. Read it at PubMed Central:

References 68

1. 1.

   Olson GB. Designing a new material world. Science. 2010;288:993-998.

   Not in our library10.1126/science.288.5468.993 (opens in a new tab)

2. 2.

   Curtarolo S, Morgan D, Persson K, Rodgers J, Ceder G. Predicting crystal structures with data mining of quantum calculations. Phys. Rev. Lett. 2003;91:135503.

   Not in our library10.1103/physrevlett.91.135503 (opens in a new tab)

3. 3.

   Fischer CC, Tibbetts KJ, Morgan D, Ceder G. Predicting crystal structure by merging data mining with quantum mechanics. Nat. Mater. 2006;5:641-646.

   Not in our library10.1038/nmat1691 (opens in a new tab)

4. 4.

   Nye, J. F. Physical Properties of Crystals: Their Representation by Tensors and Matrices, ser. Oxford science publications (Clarendon Press, Oxford, 1985).

   Not in our library

5. 5.

   Smith, W. F. & Hashemi, J. Foundations of Materials Science and Engineering, ser. McGraw-Hill Series in Materials Science and Engineering (McGraw-Hill, New York, 2004).

   Not in our library

6. 6.

   Hahn, T. International Tables for Crystallography. International Tables for Crystallography, Vol. A (International Union of Crystallography: Chester, England, 2006). http://it.iucr.org/Ab/.

   Not in our library

7. 7.

   Stokes HT, Hatch DM. FINDSYM: program for identifying the space group symmetry of a crystal. J. Appl. Crystallogr. 2005;38:237-238.

   Not in our library10.1107/s0021889804031528 (opens in a new tab)

8. 8.
   Structure validation in chemical crystallography · 2009

   No notice
9. 9.

   Grosse-Kunstleve RW. Algorithms for deriving crystallographic space group information. Acta Crystallogr. A. 1999;55:383-395.

   Not in our library10.1107/s0108767398010186 (opens in a new tab)

10. 10.

    Englert U. Symmetry relationships between crystal structures. Applications of crystallographic group theory in crystal chemistry. By Ulrich Müller”. Angew. Chem. Int. Ed. 2013;52:11 973-11 973.

    Not in our library10.1002/anie.201306902 (opens in a new tab)

11. 11.

    Atsushi T. Spglib, https://atztogo.github.io/spglib/ (2009).

    Not in our library

12. 12.

    Hicks D, et al. AFLOW-SYM: platform for the complete, automatic and self-consistent symmetry analysis of crystals. Acta Crystallogr. Sect. A. 2018;74:184-203.

    Not in our library10.1107/s2053273318003066 (opens in a new tab)

13. 13.

    NOMAD Laboratory. NOMAD. https://nomad-coe.eu (2015).

    Not in our library

14. 14.

    Ryan MP, Williams DE, Chater RJ, Hutton BM, McPhail DS. Why stainless steel corrodes. Nature. 2002;415:770-774.

    Not in our library10.1038/415770a (opens in a new tab)

15. 15.

    Duarte MJ, et al. Element-resolved corrosion analysis of stainless-type glass-forming steels. Science. 2013;341:372-376.

    Not in our library10.1126/science.1230081 (opens in a new tab)

16. 16.

    Gault B, Moody MP, Cairney JM, Ringer SP. Atom probe crystallography. Mater. Today. 2012;15:378-386.

    Not in our library10.1016/s1369-7021(12)70164-5 (opens in a new tab)

17. 17.
    Classification of crystal structure using a convolutional neural network · 2017

    Not checked
18. 18.
    Insightful classification of crystal structures using deep learning · 2018

    No notice
19. 19.

    Ghiringhelli LM, Vybiral J, Levchenko SV, Draxl C, Scheffler M. Big data of materials science: critical role of the descriptor. Phys. Rev. Lett. 2015;114:105503.

    Not in our library10.1103/physrevlett.114.105503 (opens in a new tab)

20. 20.

    Behler J, Parrinello M. Generalized neural network representation of high-dimensional potential-energy surfaces. Phys. Rev. Lett. 2007;98:1-4.

    Not in our library10.1103/physrevlett.98.146401 (opens in a new tab)

21. 21.

    Rupp M, Tkatchenko A, Müller KR, Lilienfeld V, Anatole O. Fast and accurate modeling of molecular atomization energies with machine learning. Phys. Rev. Lett. 2012;108:58301.

    Not in our library10.1103/physrevlett.108.058301 (opens in a new tab)

22. 22.

    Bartók AP, Kondor R, Csányi G. On representing chemical environments. Phys. Rev. B. 2013;87:184115.

    Not in our library10.1103/physrevb.87.184115 (opens in a new tab)

23. 23.
    Quantum-chemical insights from deep tensor neural networks · 2017

    No notice
24. 24.

    Huo, H. & Rupp, M. Unified representation for machine learning of molecules and crystals. Preprint at http://arxiv.org/abs/1704.06439 (2017).

    Not in our library

25. 25.

    Ward L, et al. Including crystal structure attributes in machine learning models of formation energies via Voronoi tessellations. Phys. Rev. B. 2017;96:024104.

    Not in our library10.1103/physrevb.96.024104 (opens in a new tab)

26. 26.
    Universal fragment descriptors for predicting properties of inorganic crystals · 2017

    No notice
27. 27.

    Zhu L, et al. A fingerprint based metric for measuring similarities of crystalline structures. J. Chem. Phys. 2016;144:034203.

    Not in our library10.1063/1.4940026 (opens in a new tab)

28. 28.

    Deringer VL, Csányi G. Machine learning based interatomic potential for amorphous carbon. Phys. Rev. B. 2017;95:094203.

    Not in our library10.1103/physrevb.95.094203 (opens in a new tab)

29. 29.

    Morawietz T, Singraber A, Dellago C, Behler J. How van der Waals interactions determine the unique properties of water. Proc. Natl. Acad. Sci. USA. 2016;113:8368-8373.

    Not in our library10.1073/pnas.1602375113 (opens in a new tab)

30. 30.

    Henke B, Gullikson E, Davis J. X-ray interactions: photoabsorption, scattering, transmission, and reflection at E = 50–30,000 eV, Z = 1–92. At. Data Nucl. Data Tables. 1993;54:181-342.

    Not in our library10.1006/adnd.1993.1013 (opens in a new tab)

31. 31.

    Friedrich W, Knipping P, Laue M. Interferenzerscheinungen bei Röntgenstrahlen. Ann. Phys. 1913;346:971-988.

    Not in our library10.1002/andp.19133461004 (opens in a new tab)

32. 32.

    THOMSON GP, REID A. Diffraction of cathode rays by a thin film. Nature. 1927;119:890-890.

    Not in our library10.1038/119890a0 (opens in a new tab)

33. 33.

    Wollan EO, Shull CG. The diffraction of neutrons by crystalline powders. Phys. Rev. 1948;73:830-841.

    Not in our library10.1103/physrev.73.830 (opens in a new tab)

34. 34.

    Setyawan W, Curtarolo S. High-throughput electronic band structure calculations: challenges and tools. Comput. Mater. Sci. 2010;49:299-312.

    Not in our library10.1016/j.commatsci.2010.05.010 (opens in a new tab)

35. 35.

    Looijenga-Vos A, Buerger MJ. International Tables for Crystallography. 2006;44-54.

    Not in our library

36. 36.

    De Graef, M. & McHenry, M. E. Structure of Materials: An Introduction to Crystallography, Diffraction and Symmetry (Cambridge University Press, Cambridge, UK, 2007).

    Not in our library

37. 37.

    Ashcroft, N. W. & Mermin, N. D. Solid State Physics (Cengage Learning, London, 2011).

    Not in our library

38. 38.

    Bengio Y. Learning deep architectures for AI. Found. Trends Mach. Learn. 2009;2:1-127.

    Not in our library10.1561/2200000006 (opens in a new tab)

39. 39.

    Schmidhuber J. Deep learning in neural networks: qn overview. Neural Netw. 2015;61:85-117.

    Not in our library10.1016/j.neunet.2014.09.003 (opens in a new tab)

40. 40.
    Deep learning · 2015

    No notice
41. 41.

    Hinton GE. Reducing the dimensionality of data with neural networks. Science. 2006;313:504-507.

    Not in our library10.1126/science.1127647 (opens in a new tab)

42. 42.

    Hinton GE, Osindero S, Teh YW. A fast learning algorithm for deep belief nets. Neural Comput. 2006;18:1527-1554.

    Not in our library10.1162/neco.2006.18.7.1527 (opens in a new tab)

43. 43.

    LeCun Y, et al. Backpropagation applied to handwritten zip code recognition. Neural Comput. 1989;1:541-551.

    Not in our library10.1162/neco.1989.1.4.541 (opens in a new tab)

44. 44.

    Lecun Y, Bottou L, Bengio Y, Haffner P. Gradient-based learning applied to document recognition. Proc. IEEE. 1998;86:2278-2324.

    Not in our library10.1109/5.726791 (opens in a new tab)

45. 45.

    Krizhevsky, A., Sutskever, I. & Hinton, G. E. ImageNetClassification with deep convolutional neural networks. in Advances in Neural Information Processing Systems 25 (eds Pereira, F., Burges, C. J. C., Bottou, L. & Weinberger, K. Q.) 1097–1105 (Curran Associates, New York, 2012).

    Not in our library

46. 46.

    Pàmies P. Auspicious machine learning. Nat. Biomed. Eng. 2017;1:0036.

    Not in our library10.1038/s41551-017-0036 (opens in a new tab)

47. 47.

    Curtarolo S, et al. AFLOWLIB.ORG: a distributed materials properties repository from high-throughput ab initio calculations. Comput. Mater. Sci. 2012;58:227-235.

    Not in our library10.1016/j.commatsci.2012.02.002 (opens in a new tab)

48. 48.

    Taylor RH, et al. A RESTful API for exchanging materials data in the AFLOWLIB.org consortium. Comput. Mater. Sci. 2014;93:178-192.

    Not in our library10.1016/j.commatsci.2014.05.014 (opens in a new tab)

49. 49.

    Mehl MJ, et al. The AFLOW Library of crystallographic prototypes. Comput. Mater. Sci. 2016;136:S1-S828.

    Not in our library10.1016/j.commatsci.2017.01.017 (opens in a new tab)

50. 50.

    Zeiler, M. D., Krishnan, D., Taylor, G. W. & Fergus, R. ImageNetClassification with deep convolutional neural networks. in 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (eds. Ma, B., Su, Y. & Jurie, F.) 2528–2535 (IEEE, San Fransisco, CA, 2010).

    Not in our library

51. 51.

    Zeiler, M. D. & Fergus, R. Visualizing and Understanding Convolutional Networks 818–833, 10.1007/978-3-319-10590-1_53 (2014).

    Not in our library

52. 52.
    On Pixel-Wise Explanations for Non-Linear Classifier Decisions by Layer-Wise Relevance Propagation · 2015

    No notice
53. 53.

    Kumar, D. & Menkovski, V. Understanding anatomy classification through visualization. In NIPS Machine Learning for Health, no. Nips 1–5 http://arxiv.org/abs/1611.06284 (2016).

    Not in our library

54. 54.

    Montavon G, Lapuschkin S, Binder A, Samek W, Müller KR. Explaining nonlinear classification decisions with deep Taylor decomposition. Pattern Recognit. 2017;65:211-222.

    Not in our library10.1016/j.patcog.2016.11.008 (opens in a new tab)

55. 55.

    Kumar, D., Wong, A. & Taylor, G. W. Explaining the unexplained: a class-enhanced attentive response (CLEAR) approach to understanding deep neural networks. In 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 1686–1694 (IEEE, Honolulu, HI, 2017).

    Not in our library

56. 56.
    A manifesto for reproducible science · 2017

    No notice
57. 57.

    Baker M. 1,500 scientists lift the lid on reproducibility. Nature. 2016;533:452-454.

    Not in our library10.1038/533452a (opens in a new tab)

58. 58.

    Ong SP, et al. Python Materials Genomics (pymatgen): a robust, open-source python library for materials analysis. Comput. Mater. Sci. 2013;68:314-319.

    Not in our library10.1016/j.commatsci.2012.10.028 (opens in a new tab)

59. 59.

    Sakurai, J. J. & Napolitano, J. Modern Quantum Mechanics (Addison-Wesley, Waltham, MA, 2011).

    Not in our library

60. 60.

    Bunge, H.-J. H. J. Texture Analysis in Materials Science: Mathematical Methods (Butterworths, London, 1982).

    Not in our library

61. 61.

    Britton T, et al. Tutorial: Crystal orientations and EBSD—or which way is up?”. Mater. Charact. 2016;117:113-126.

    Not in our library10.1016/j.matchar.2016.04.008 (opens in a new tab)

62. 62.

    Hantke MF, Ekeberg T, Maia FRNC. Condor: a simulation tool for flash X-ray imaging. J. Appl. Crystallogr. 2016;49:1356-1362.

    Not in our library10.1107/s1600576716009213 (opens in a new tab)

63. 63.

    Kingma, D. & Ba, J. Adam: a method for stochastic optimization. In International Conference on Learning Representations 1–13. Preprint at http://arxiv.org/abs/1412.6980 (2014).

    Not in our library

64. 64.

    Martin, A. et al. TensorFlow: large-scale machine learning on heterogeneous systems, https://www.tensorflow.org/ (2015).

    Not in our library

65. 65.

    Chollet, F. Keras, https://github.com/fchollet/keras (2015).

    Not in our library

66. 66.

    Nair, V. & Hinton, G. E. Rectified linear units improve restricted boltzmann machines. In Proceedings of the 27th International Conference on Machine Learning (ICML-10) (eds Fürnkranz, J. & Joachims, T.) 807–814 (Omnipress, Madison, WI, 2010).

    Not in our library

67. 67.

    Srivastava N, Hinton G, Krizhevsky A, Sutskever I, Salakhutdinov R. Dropout: a simple way to prevent neural networks from overfitting. J. Mach. Learn. Res. 2014;15:1929-1958.

    Not in our library

68. 68.

    Rumelhart DE, Hinton GE, Williams RJ. Learning representations by back-propagating errors. Nature. 1986;323:533-536.

    Not in our library10.1038/323533a0 (opens in a new tab)

Cited by papers we hold 2

• A review of the recent progress in battery informatics · 2022
  Not checked
• Data‐Driven Materials Science: Status, Challenges, and Perspectives · 2019
  Corrected

Questions about this paper

Reads

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.