MEGNet

Version: v2022.9.20 Added: 2022-11-14 Published: 2021-12-18 168k parameters Missing preds: 0 pip install megnet
Repo Paper Docs DOI Files PyPI PR Checkpoint
Predictions
Discovery

ML vs DFT Formation Energies

ML vs DFT Convex Hull Distance

Convex hull distance prediction errors projected onto elements

1 H 0.62
2 He 0.00
3 Li 0.22
4 Be 0.25
5 B 0.28
6 C 0.33
7 N 0.56
8 O 0.62
9 F 1.73
10 Ne 0.00
11 Na 0.26
12 Mg 0.23
13 Al 0.28
14 Si 0.35
15 P 0.33
16 S 0.50
17 Cl 1.33
18 Ar 0.00
19 K 0.30
20 Ca 0.21
21 Sc 0.22
22 Ti 0.37
23 V 0.53
24 Cr 0.31
25 Mn 0.33
26 Fe 0.34
27 Co 0.32
28 Ni 0.27
29 Cu 0.21
30 Zn 0.22
31 Ga 0.27
32 Ge 0.27
33 As 0.30
34 Se 0.42
35 Br 0.46
36 Kr 0.00
37 Rb 0.31
38 Sr 0.23
39 Y 0.25
40 Zr 0.28
41 Nb 0.47
42 Mo 0.30
43 Tc 0.26
44 Ru 0.27
45 Rh 0.25
46 Pd 0.26
47 Ag 0.21
48 Cd 0.20
49 In 0.27
50 Sn 0.26
51 Sb 0.25
52 Te 0.37
53 I 0.41
54 Xe 0.72
55 Cs 0.30
56 Ba 0.27
57 La 0.25
58 Ce 0.25
59 Pr 0.25
60 Nd 0.24
61 Pm 0.13
62 Sm 0.23
63 Eu 0.22
64 Gd 0.23
65 Tb 0.25
66 Dy 0.25
67 Ho 0.24
68 Er 0.23
69 Tm 0.23
70 Yb 0.19
71 Lu 0.22
72 Hf 0.29
73 Ta 0.51
74 W 0.23
75 Re 0.27
76 Os 0.27
77 Ir 0.27
78 Pt 0.27
79 Au 0.24
80 Hg 0.19
81 Tl 0.23
82 Pb 0.24
83 Bi 0.24
84 Po 0.00
85 At 0.00
86 Rn 0.00
87 Fr 0.00
88 Ra 0.00
89 Ac 0.12
90 Th 0.32
91 Pa 0.21
92 U 0.28
93 Np 0.30
94 Pu 0.47
95 Am 0.00
96 Cm 0.00
97 Bk 0.00
98 Cf 0.00
99 Es 0.00
100 Fm 0.00
101 Md 0.00
102 No 0.00
103 Lr 0.00
104 Rf 0.00
105 Db 0.00
106 Sg 0.00
107 Bh 0.00
108 Hs 0.00
109 Mt 0.00
110 Ds 0.00
111 Rg 0.00
112 Cn 0.00
113 Nh 0.00
114 Fl 0.00
115 Mc 0.00
116 Lv 0.00
117 Ts 0.00
118 Og 0.00
57-71 La-Lu Lanthanides
89-103 Ac-Lr Actinides

Model Authors

  1. Chi Chen  UC San Diego  
  2. Weike Ye  UC San Diego  
  3. Yunxing Zuo  UC San Diego  
  4. Chen Zheng  UC San Diego  
  5. Shyue Ping Ong  UC San Diego  

Model Info

  • Model Version v2022.9.20
  • Model Type GNN
  • Targets E
  • Openness OSOD
  • Train Task RS2RE
  • Test Task IS2E
  • Trained for Benchmark No

Training Set

Graphs of MP 2019: 133k structures

Description

MatErials Graph Network is another GNN for material properties of relaxed structure which showed that learned element embeddings encode periodic chemical trends and can be transfer-learned from large data sets (formation energies) to predictions on small data properties (band gaps, elastic moduli).

Training

Using pre-trained model released with paper. Was only trained on MP-crystals-2018.6.1 dataset available on Figshare.

Hyperparameters

  • graph_construction_radius: 4
  • max_neighbors: null

Dependencies