Fig. 1

Schematic diagram of multilayer neural network"

Fig. 2

Comparative analysis of downstream task sample efficiency. The horizontal coordinate indicates the amount of downstream data required and the vertical coordinate indicates the convergence of the RMSE energy or force prediction[5]"

Fig. 3

DP-GEN automation process[21]"

Fig. 4

Diffusion trajectories of Na+ in (a) low-Na and (b) high-Na content C8S materials[23]"

Fig. 5

Li-S-Si system DP model training and inspection process[25]"

Fig. 6

Snapshots of inhomogeneous deposition at different temperatures (T) and generation rates (Rg). Rg = 0.5 ps-1 (a, d, g), 1.0 ps-1 (b, e, h) and 5.0 ps-1 (c, f, i); T= 400 K (a—c), 300 K (d—f), 200 K (g—i)[26]"

Fig. 7

Oxygen-producing reaction pathway diagram of CoO2 model based on enhanced sampling DPMD simulation[31]"

Fig. 8

Flowchart of DPMD-based solid-state NMR chemical shift calculation[33]"

Fig. 9

Li+ diffusion behavior in the Li10GeP2S12 lattice. (a) Arrhenius plot of diffusion coefficients simulated with AIMD and DPMD. (b) Dimensional contribution of Li-ion diffusion as a function of temperature inverse. (c) Color-filled plots of Li-ion probability density in Li10GeP2S12 at 1000 K and (d) 300 K[37]"

Fig. 10

Temperature dependence of conductivity (σ) and Haven's ratio. (a, b, c) Arrhenius plots of ionic conductivity of DPMD; (d, e, f) Haven's ratio as a function of temperature[39]"

Fig. 11

Intra-layer lithium ion motion trajectories along the Li3TiCl6 (a) [010] and (b) [110] view directions; (c) Li atom displacement evolution with time corresponding to (a) and (b); Schematic diagram of the inter-layer lithium ion motion trajectories along the Li3TiCl6 (d) [010] and (e) [110] view directions; (f) Corresponding to (d) and (e) Li atom displacement evolution with time[44]"

Fig. 12

(a) O-Ge-O angular distribution (decomposed as a product of Gaussian and sinusoidal functions); (b) Possible transitions of the oxygen environment of liquid/glass Ge atoms: 1—octahedral, 2—square pyramid, 3—triangular bipyramid, 4—transition structure, 5—tetrahedral[49]"

Fig. 13

Depth potential pre-training flow for amorphous Li-La-Zr-O system[53]"

Fig. 14

Automated workflow of acidity and redox potential. (a) Initialization; (b) DP potential function training process; (c) DPMD free energy calculation[11]"

Fig. 15

(a) MSD of Li ions with simulation time and diffusion coefficients obtained for different thicknesses of α-LiF coated regions. (b) Atom stacking morphology of LiF with 0.5 nm and 2.0 nm layer thicknesses. Regular Li-F tetrahedral numbers and P-S tetrahedra are also shown[72]"