for molecular and amorphous systems*

GW has become the method of choice for the calculation of valence photoemission spectra [1]. To apply GW also to deep core excitations as measured by X-ray photoelectron spectroscopy, we recently advanced the GW methodology and our implementation by combining exact numeric algorithms in the real frequency domain [2] with partial self-consistency [3] and relativistic corrections [3,4]. We benchmarked our core-level GW implementation for 65 1s core excitations, for which we find that GW reproduces absolute molecular 1s excitations within 0.3 eV of experiment and relative binding energies with average deviations smaller than 0.2 eV [3]. We then combined GW with machine learning (ML). We computed 16,000 C1s and O1s excitations of organic molecules with GW and used them to train a Kernel Ridge Regression (KRR) ML model. The KRR-ML model predicts molecular core-level energies within 0.1 eV of the GW reference data. We used these models as starting point to develop GW-ML schemes for amorphous carbon materials, which show promise as electrode material for biomedical devices.

[1] D. Golze et. al. Front. Chem, 7 (2019), 377
[2] D. Golze et. al. JCTC, 14 (2018), 4856
[3] D. Golze et al. JPCL, 11 (2020), 1840
[4] L. Keller et al. JCP, 153 (2020), 114110