Development of ultra-high efficiency soft X-ray angle-resolved photoemission spectroscopy equipped with deep prior-based denoising method

TL;DR

Soft X-ray ARPES provides bulk-sensitive three-dimensional electronic structure but is limited by low photoelectron yield from small cross-sections, leading to long acquisition times. The authors integrate a training-free, deep-prior denoising method (DPDM) with the μSX-ARPES system at SPring-8 BL25SU to achieve high-S/N images in about 1 minute per dataset. They demonstrate substantial noise removal of grid and spike artifacts with DPDM, enabling reliable EDC analysis at short accumulation times (e.g., $t_{acc} \approx 40$ s) and a total processing time of around $60$–$70$ s for each dataset. They report an energy resolution of $\Delta E_T = 51.6$ meV at $h\nu = 708$ eV in a gold sample and discuss the potential for further improvements toward sub-30 meV resolution with next-generation, fully coherent soft X-ray sources, as well as extending DPDM to VUV-ARPES and related imaging modalities.

Abstract

Soft X-ray angle resolved photoemission spectroscopy (SX-ARPES) is one of the most powerful spectroscopic techniques to visualize the three-dimensional bulk electronic structure in reciprocal lattice space. Compared with ARPES employing low-energy photon sources, the time burden imposed by a lower photoelectron yield, stemming from the photoionization cross-section, has been a persistent technical challenge. To address this challenge, we have developed a noise removal system by using the deep prior-based method and integrated it into the micro focused SX-ARPES (μSX-ARPES) system at BL25SU in SPring-8. Our implemented system effectively eliminates the grid and spike noise typically present in ARPES data acquired using the voltage Fixed-mode, within about 30 seconds. We demonstrate, through the μSX-ARPES measurements on a single crystal of CeRu2Si2, that data with sufficient statistical accuracy can be obtained in approximately 40 seconds. In addition, we present the potential of high signal-to-noise ratio ARPES measurement, achieving an energy resolution of 51.6 meV at an excitation energy of 708 eV in μSX-ARPES measurements on polycrystalline gold. Our developed system successfully reduces the time burden in SX-ARPES and paves the way for advancements in lower photoelectron yield measurements, such as those requiring higher energy resolution and three-dimensional nonequilibrium measurements.

Figures (5)

• Figure 1: (a) Photoionization cross-sections of Ce 4$f$, Cu 3$d$, and O 2$p$ electrons, digitized from Ref. yeh1985atomic. (b) Grid and spike noise commonly observed in ARPES image obtained by Fixed mode. (c) Overview of the DPDM system implementation. (d) Screenshot of the jupyter Notebook interface after running DPDM. The number of iterations, the degree of completion, and the noise processing time are indicated within the red frame. The ARPES image can be viewed during the denoising process, allowing for real-time monitoring. The optimal denoised data can be selected by referring to the Loss function shown below. (e) Loss function as a function of iterations. The inset ARPES images correspond to the iteration counts indicated by the red circles.
• Figure 2: The upper two figures present 3D surface plots of CeRu$_2$Si$_2$ (a) before and (b) after denoising. The data were acquired at $h\nu = 750$ eV and 77 K in Fixed mode with an accumulation time of 1000 seconds, which ensures a sufficient S/N ratio. (c) Comparison of EDC spectra at a detector angle of 4 degrees, before and after denoising, highlighting the removal of grid and spike noise.
• Figure 3: Relationship between accumulation time and raw kinetic energy-detector angle ($E_k$-$\theta$) images for CeRu$_2$Si$_2$ acquired in Fixed mode. (a-d) $E_k$-$\theta$ images with different accumulation time. (e-h) The same $E_k$-$\theta$ images as panels (a-d), but after denoising. (i) Raw $E_k$-$\theta$ image measured in Swept mode for comparison, with an accumulation time of 45 minutes. (j-k) Comparison of the accumulation time dependent EDC spectra (j) before and (k) after denoising. The EDC spectrum acquired in Swept mode is also shown for comparison.
• Figure 4: The binding energy - momentum ($E$-$k$) images of Mn$_3$Si$_2$Te$_6$ acquired in Fixed mode (a) without and (b) with denoising. (c) The same $E$-$k$ image, but obtained in Swept mode for comparison. The data were acquired at $h\nu = 500$ eV and a temperature below 30 K along the $\Gamma$-$M$ line. (d) EDC spectra at the $\Gamma$ and $M$ points. The filled circles indicate the peak clearly observed in Fixed mode, but not in Swept mode.
• Figure 5: Angle-integrated ARPES spectrum near the Fermi level of gold in Swept mode. The red solid line represents the fit of the Fermi-Dirac distribution, with a total energy resolution of 51.6 meV, to the data.