Enhancing transfer learning in angle-resolved photoemission spectroscopy (ARPES) with spatially-aware representations via graph convolution

Sugiarto, Hendrik Santoso [Calvin Institute of Technology, Jakarta (Indonesia)] (ORCID:0000000218319808); Ekahana, Sandy Adhitia [Paul Scherrer Institute (PSI), Villigen (Switzerland); Carnegie Mellon University, Pittsburgh, PA (United States); Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States). Advanced Light Source (ALS)] (ORCID:0000000216938082); Wijaya, Bryan Christofer [Calvin Institute of Technology, Jakarta (Indonesia)] (ORCID:0009000519450938); Winata, Genta Indra [Hong Kong University of Science and Technology (HKUST) (Hong Kong)]; Soh, Y. [Paul Scherrer Institute (PSI), Villigen (Switzerland)]; Aeppli, G. [Paul Scherrer Institute (PSI), Villigen (Switzerland); Ecole Polytechnique Federale Lausanne (EPFL) (Switzerland); Eidgenoessische Technische Hochschule (ETH), Zurich (Switzerland)]

doi:10.1088/2632-2153/ae3fe7

DOE OSTI · 3028174

Enhancing transfer learning in angle-resolved photoemission spectroscopy (ARPES) with spatially-aware representations via graph convolution

Sugiarto, Hendrik Santoso [Calvin Institute of Technology, Jakarta (Indonesia)] (ORCID:0000000218319808)·Ekahana, Sandy Adhitia [Paul Scherrer Institute (PSI), Villigen (Switzerland); Carnegie Mellon University, Pittsburgh, PA (United States); Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States). Advanced Light Source (ALS)] (ORCID:0000000216938082)·Wijaya, Bryan Christofer [Calvin Institute of Technology, Jakarta (Indonesia)] (ORCID:0009000519450938)·Winata, Genta Indra [Hong Kong University of Science and Technology (HKUST) (Hong Kong)]·Soh, Y. [Paul Scherrer Institute (PSI), Villigen (Switzerland)]·Aeppli, G. [Paul Scherrer Institute (PSI), Villigen (Switzerland); Ecole Polytechnique Federale Lausanne (EPFL) (Switzerland); Eidgenoessische Technische Hochschule (ETH), Zurich (Switzerland)]

View original source ↗Read source document ↗Export metadata ↓

Abstract

A recent application of machine learning has been to spatially-resolved angle-resolved photoemission spectroscopy (ARPES). Here we advance the state-of-the-art by applying representational learning to transform ARPES data into an embedding space of a pre-trained self-supervised learning model, thus enhancing the pipeline that improves the bandstructure classification and domain assignment/segmentation performance compared to a k-means clustering method. In the current iteration, the real-space information is entered into the domain assignment through the graph convolution method, which improves the transfer learning performance of the original self-supervised model. Lastly, an unsupervised automated tool is developed that incorporates these techniques to enable automatic domain assignment.

Explore related subjects

ARPES deep learning graph convolution representation learning self-supervised transfer learning

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Sugiarto, Hendrik Santoso [Calvin Institute of Technology, Jakarta (Indonesia)] (ORCID:0000000218319808), Ekahana, Sandy Adhitia [Paul Scherrer Institute (PSI), Villigen (Switzerland); Carnegie Mellon University, Pittsburgh, PA (United States); Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States). Advanced Light Source (ALS)] (ORCID:0000000216938082), Wijaya, Bryan Christofer [Calvin Institute of Technology, Jakarta (Indonesia)] (ORCID:0009000519450938), Winata, Genta Indra [Hong Kong University of Science and Technology (HKUST) (Hong Kong)], Soh, Y. [Paul Scherrer Institute (PSI), Villigen (Switzerland)], Aeppli, G. [Paul Scherrer Institute (PSI), Villigen (Switzerland); Ecole Polytechnique Federale Lausanne (EPFL) (Switzerland); Eidgenoessische Technische Hochschule (ETH), Zurich (Switzerland)]. 2026-02-23. Enhancing transfer learning in angle-resolved photoemission spectroscopy (ARPES) with spatially-aware representations via graph convolution. https://doi.org/10.1088/2632-2153%2Fae3fe7

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related reports

DOE OSTI2026

Evidence for strong electronic correlations in the bulk state of grey arsenic

Here, we investigate the electron band structure of grey arsenic, whose (111) face hosts the topological Shockley state. Interestingly, the bulk band close to the touching point with the surface state exhibits the characteristics of inelastic scattering. Moreover, the band structure analysis reveals linearity in the imaginary part of electron self-energy. These features are analogous to those observed in high-temperature superconductors and marginal Fermi liquid systems, respectively, where strong electronic correlations exist. Our results suggest that correlated many-body states can be connected by non-interacting topological states, providing a viable playground to explore the coupling between topological and correlated states via grey arsenic surface.

ARPES↗

DOE OSTI2025

Tuning the electronic properties of MgB 2 by substitution with Mn and C

We use high-resolution angle-resolved photoemission spectroscopy (ARPES) to investigate the electronic band structure of the superconductors Mg 0.995 Mn 0.005 B 2 and Mg(B 0.98 C 0.02 ) 2 , in comparison with pristine MgB 2 . Our experimental findings reveal modest changes in the size of the Fermi surface sheets following manganese or carbon substitution, with the most pronounced effect observed in the outermost σ band. Additionally, the dispersion kink, indicative of strong renormalization due to the interaction between conduction electrons and phonon modes near −0.065 eV, remains unaffected by chemical substitution. Given that the superconducting transition temperature decreases significantly more rapidly in Mn-substituted MgB 2 than in C-substituted MgB 2 , our ARPES results strongly suggest that the primary mechanism behind the pronounced T C suppression in Mn–MgB 2 is the disruption of Cooper pairs via Abrikosov–Gorkov spin-flip scattering caused by the local magnetic moments of Mn.

ARPES↗

DOE OSTI2025

Possible evidence of excitonic condensation in a topological insulator

The transient excitonic condensate is a nonequilibrium electron-hole Bardeen-Cooper-Schrieffer state in a photoexcited semiconductor and semimetal, where electron-hole pairs undergo a phase transition and condense into a single coherent quantum state. Despite numerous experimental works to realize the predicted excitonic condensation phase, experimental evidence still remains elusive. This is largely due to the absence of direct measurements of a material's transient momentum-dependent electronic structure and the excitonic state in the condensation regime. Here, using time and angle-resolved photoemission spectroscopy, we find direct evidence of a transient excitonic condensate in the spin-polarized spatially indirect excitonic topological states in Bi2Te3. Accompanying the formation of the excitonic topological states by photoexcitation, we reveal a splitting of the hole's and electron's quasi-equilibrium chemical potential followed by the band flattening and backbending of the transient topological surface state. Moreover, within the same momentum range, we report a reshaping of the bulk valence band in the form of a Mexican-hat-like Bogoliubov dispersion-hallmarks of the excitonic condensation, followed by the opening of an energy gap at the Fermi level. The fluence and temperature dependence of these renormalization effects are reminiscent of excitonic condensation within Bardeen-Cooper-Schrieffer (BCS)-like behavior. These results, together with theoretical simulation, point to the possible formation of a transient excitonic condensate and provide opportunities to manipulate topologically protected Bose condensates with light.

ARPES↗