Drastic Efficiency Improvement of Soft X-ray Angle-Resolved Photoemission Spectroscopy through Grid Removal Technology Based on Deep Prior Distribution

【Summary】

Dr. Kohei Yamagami, a researcher at the Japan Synchrotron Radiation Research Institute (JASRI), along with Dr. Yuichi Yokoyama, Researcher, and Mr. Tetsuro Nakamura, Engineer, Professor Itaru Shono and Mr. Yuta Sumitani (at the time, Master's student) of the University of Electro-Communications, and Professor Jin-ichiro Mizumaki of Kumamoto University, have achieved an environment for ultra-high efficiency μSX-ARPES measurements without compromising conventional energy resolution. This was done by integrating their independently developed "Deep-Prior based Denoising Method (DPDM)" into the micro-focused soft X-ray angle-resolved photoemission spectroscopy system (μSX-ARPES) operating at the soft X-ray solid spectroscopy beamline BL25SU of the large synchrotron radiation facility SPring-8. The specific achievements and demonstrations of this development research are as follows:

1. Developed a system that effectively removes periodic grid structures and spike structures present in ARPES data in approximately 30 seconds.

2. In measurements of the heavy fermion compound CeRu2Si2, it was demonstrated that statistically reliable ARPES data can be obtained in a short time of about 40 seconds when combined with DPDM. The total measurement time, including grid removal processing, was approximately 70 seconds, which means a success in reducing the time by more than 90% compared to the 2700 seconds required to obtain equivalent data by conventional measurement methods.

3. This efficiency improvement enables ultra-high resolution measurements that were previously difficult due to time constraints. Even with measurements at 6 meV, one of the highest energy resolutions ever achieved in SX-ARPES, it is expected to be feasible within practical measurement times.

These research results are expected to overcome the fundamental technical constraint of the time burden of SX-ARPES measurements and pave the way for the development of new soft X-ray spectroscopy measurement technologies such as ultra-high resolution measurements utilizing next-generation synchrotron radiation sources and 3D non-equilibrium electronic structure observation. A paper summarizing these research results was published open access on May 8th in the international scientific journal "Review of Scientific Instruments" issued by the American Institute of Physics.

【Background of Development Research】

Angle-resolved photoemission spectroscopy (ARPES) is a powerful experimental technique that directly visualizes the electronic structure of materials by measuring the energy of electrons in materials as a function of momentum. Among these, soft X-ray ARPES (SX-ARPES), which uses soft X-rays (SX, photon energy: ~800 eV) as excitation light, has sensitivity even within the material and features the ability to observe the electronic structure of materials decomposed into 3D momentum space. However, compared to vacuum ultraviolet (VUV, photon energy: ~40 eV) ARPES (VUV-ARPES), which is suitable for observing fine electronic structures, the photoionization scattering cross-section for photoelectron emission is more than an order of magnitude smaller, requiring a long measurement time. This often leads to problems such as temporal changes in samples like surface oxidation and degradation of resolution due to fluctuations in excitation light energy. Therefore, SX-ARPES users had to conduct experiments while bearing a significant burden.

The research group focused on the fixed mode of the electrostatic hemispherical photoelectron analyzer (analyzer) to acquire ARPES data with higher statistical accuracy (S/N ratio) faster while maintaining conventional energy resolution (Fig. 1). In this mode, only photoelectrons reaching the detection area of the multi-channel plate (MCP) are detected, and their kinetic energy and emission angle are measured. Compared to the swept mode, which can measure a wider kinetic energy range, spectra with a high S/N ratio can be acquired in a shorter time. However, due to metal mesh filters designed to block stray electrons and the aging of the detection unit, periodic grid structures and non-periodic spike structures are formed in the data (Fig. 1). These have been major obstacles to spectral analysis, making grid removal technology indispensable.

【Development Content and Results】

The research group constructed a system that integrates the "Deep-Prior based Denoising Method (DPDM)" into the micro-focused SX-ARPES system (μSX-ARPES) installed at the soft X-ray solid spectroscopy beamline BL25SU of SPring-8 (Fig. 2). DPDM is a training-free method that does not require a pre-trained dataset and utilizes the structural characteristics of a 4-layer U-shaped convolutional neural network. By preparing a dedicated PC for DPDM and operating the grid removal system via remote connection from other PCs, it was possible to achieve the aforementioned results.