A machine learning route between band mapping and band structure

Abstract

Electronic band structure (BS) and crystal structure are the two complementary identifiers of solid state materials. While convenient instruments and reconstruction algorithms have made large, empirical, crystal structure databases possible, extracting quasiparticle dispersion (closely related to BS) from photoemission band mapping data is currently limited by the available computational methods. To cope with the growing size and scale of photoemission data, we develop a pipeline including probabilistic machine learning and the associated data processing, optimization and evaluation methods for band structure reconstruction, leveraging theoretical calculations. The pipeline reconstructs all 14 valence bands of a semiconductor and shows excellent performance on benchmarks and other materials datasets. The reconstruction uncovers previously inaccessible momentum-space structural information on both global and local scales, while realizing a path towards integration with materials science databases. Our approach illustrates the potential of combining machine learning and domain knowledge for scalable feature extraction in multidimensional data.

Review of photocathodes for electron beam sources in particle accelerators

“…5 These two important and widely used models of the photoemission process can be further extended with the help of simulation methods such as Monte-Carlo, [54][55][56][57] density functional theory (DFT) [58][59][60][61][62] and machine learning. [63][64][65][66][67][68] DFT, for example, is used to calculate energy states based on electron densities and their energies. The calculations show different scenarios of covalent binding interactions between different atoms and the resulting work functions.…”

Observation of ultrafast interfacial Meitner-Auger energy transfer in a Van der Waals heterostructure

“…Figure 2a-d shows energyresolved photoemission signals along the K 0 À K cut of the Brillouin zone, at selected time delays. The band mappings are contrastenhanced using a multidimensional extension of the contrast limited adaptive histogram equalization (MCLAHE) 30,31 for better visualization of the band structure. The momentum distributions above E F within the first 400 fs reveal that the excited states are localized in three different types of valleys: the Dirac cones of graphene at its K points (K Gr ) and the K and Q valleys of WSe 2 (K W Se 2 ,Q WSe 2 ), as shown in Fig.…”

Transfer Learning Application of Self-supervised Learning in ARPES

“…For example, a deep layer of convolutional neural network (ConvNet) is trained to denoise ARPES data [28]. Afterwards, there are efforts to obtain how the bandstructure calculation based on the ARPES data [29,30], where reference [30] provides additional feature of obtaining the result even through a noisy data (simulated noise). There is also work in automation of spatial domain assignment with a smaller subset of data over a predetermined area, where subsequent position measurement is calculated with Gaussian process to give the possible highest amount of information [31].…”