-
Transformers as Neural Operators for Solutions of Differential Equations with Finite Regularity
Authors:
Benjamin Shih,
Ahmad Peyvan,
Zhongqiang Zhang,
George Em Karniadakis
Abstract:
Neural operator learning models have emerged as very effective surrogates in data-driven methods for partial differential equations (PDEs) across different applications from computational science and engineering. Such operator learning models not only predict particular instances of a physical or biological system in real-time but also forecast classes of solutions corresponding to a distribution…
▽ More
Neural operator learning models have emerged as very effective surrogates in data-driven methods for partial differential equations (PDEs) across different applications from computational science and engineering. Such operator learning models not only predict particular instances of a physical or biological system in real-time but also forecast classes of solutions corresponding to a distribution of initial and boundary conditions or forcing terms. % DeepONet is the first neural operator model and has been tested extensively for a broad class of solutions, including Riemann problems. Transformers have not been used in that capacity, and specifically, they have not been tested for solutions of PDEs with low regularity. %
In this work, we first establish the theoretical groundwork that transformers possess the universal approximation property as operator learning models.
We then apply transformers to forecast solutions of diverse dynamical systems with solutions of finite regularity for a plurality of initial conditions and forcing terms. In particular, we consider three examples: the Izhikevich neuron model, the tempered fractional-order Leaky Integrate-and-Fire (LIF) model, and the one-dimensional Euler equation Riemann problem. For the latter problem, we also compare with variants of DeepONet, and we find that transformers outperform DeepONet in accuracy but they are computationally more expensive.
△ Less
Submitted 29 May, 2024;
originally announced May 2024.
-
Comparative Study of Structural and Electronic Properties of Cu-based Multinary Semiconductors
Authors:
Yubo Zhang,
Xun Yuan,
Xiudong Sun,
Bi-Ching Shih,
Peihong Zhang,
Wenqing Zhang
Abstract:
We present a systematic and comparative study of the structural and electronic properties of Cu-based ternary and quaternary semiconductors using first-principles electronic structure approaches. The important role that Cu d electrons play in determining their properties is illustrated by comparing results calculated with different exchange correlation energy functionals. We show that systematic i…
▽ More
We present a systematic and comparative study of the structural and electronic properties of Cu-based ternary and quaternary semiconductors using first-principles electronic structure approaches. The important role that Cu d electrons play in determining their properties is illustrated by comparing results calculated with different exchange correlation energy functionals. We show that systematic improvement of the calculated anion displacement can be achieved by using the Heyd-Scuseria-Ernzerhof (HSE06) functional compared with the Perdew-Burke-Ernzerhof (PBE) functional. Quasiparticle band structures are then calculated within the G0W0 approximation using the crystal structures optimized within the HSE06 functional and starting from the PBE+U mean-field solution. Both the calculated quasiparticle band gaps and their systematic variation with chemical constituents agree very well with experiments. We also predict that the quasiparticle band gaps of the prototypical semiconductor Cu2ZnSnS4 in the kesterite (KS) phase is 1.65 eV and that of the stannite (ST) phase is 1.40 eV. These results are also consistent with available experimental values which vary from 1.45 to 1.6 eV.
△ Less
Submitted 23 June, 2011;
originally announced June 2011.
-
Prediction Of A Multi-Center Bonded Solid Boron Hydride for Hydrogen Storage
Authors:
T. A. Abtew,
Bi-ching Shih,
Pratibha Dev,
Vincent H. Crespi,
Peihong Zhang
Abstract:
An ideal material for on-board hydrogen storage must release hydrogen at practical temperature and pressure and also regenerate efficiently under similarly gentle conditions. Therefore, thermodynamically, the hydride material must lie within a narrow range near the hydrogenation/dehydrogenation phase boundary. Materials involving only conventional bonding mechanisms are unlikely to meet these re…
▽ More
An ideal material for on-board hydrogen storage must release hydrogen at practical temperature and pressure and also regenerate efficiently under similarly gentle conditions. Therefore, thermodynamically, the hydride material must lie within a narrow range near the hydrogenation/dehydrogenation phase boundary. Materials involving only conventional bonding mechanisms are unlikely to meet these requirements. In contrast, materials containing certain frustrated bonding are designed to be on the verge of frustration-induced phase transition, and they may be better suited for hydrogen storage. Here we propose a novel layered solid boron hydride and show its potential for hydrogen storage. The absence of soft phonon modes confirms the dynamical stability of the structure. Charging the structure significantly softens hydrogen-related phonon modes. Boron-related phonons, in contrast, are either hardened or not significantly affected by electron do**. These results suggest that electrochemical charging may facilitate hydrogen release while the underlying boron network remains intact for subsequent rehydrogenation.
△ Less
Submitted 1 March, 2010;
originally announced March 2010.
