Luke Watson, Thomas Pope, Raphael M Jay, Ambar Banerjee, Philippe Wernet, Thomas J Penfold
{"title":"A Δ-learning strategy for interpretation of spectroscopic observables.","authors":"Luke Watson, Thomas Pope, Raphael M Jay, Ambar Banerjee, Philippe Wernet, Thomas J Penfold","doi":"10.1063/4.0000215","DOIUrl":null,"url":null,"abstract":"<p><p>Accurate computations of experimental observables are essential for interpreting the high information content held within x-ray spectra. However, for complicated systems this can be difficult, a challenge compounded when dynamics becomes important owing to the large number of calculations required to capture the time-evolving observable. While machine learning architectures have been shown to represent a promising approach for rapidly predicting spectral lineshapes, achieving simultaneously accurate and sufficiently comprehensive training data is challenging. Herein, we introduce Δ-learning for x-ray spectroscopy. Instead of directly learning the structure-spectrum relationship, the Δ-model learns the structure dependent difference between a higher and lower level of theory. Consequently, once developed these models can be used to translate spectral shapes obtained from lower levels of theory to mimic those corresponding to higher levels of theory. Ultimately, this achieves accurate simulations with a much reduced computational burden as only the lower level of theory is computed, while the model can instantaneously transform this to a spectrum equivalent to a higher level of theory. Our present model, demonstrated herein, learns the difference between TDDFT(BLYP) and TDDFT(B3LYP) spectra. Its effectiveness is illustrated using simulations of Rh L<sub>3</sub>-edge spectra tracking the C-H activation of octane by a cyclopentadienyl rhodium carbonyl complex.</p>","PeriodicalId":48683,"journal":{"name":"Structural Dynamics-Us","volume":null,"pages":null},"PeriodicalIF":2.3000,"publicationDate":"2023-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10629969/pdf/","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Structural Dynamics-Us","FirstCategoryId":"101","ListUrlMain":"https://doi.org/10.1063/4.0000215","RegionNum":2,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"2023/11/1 0:00:00","PubModel":"eCollection","JCR":"Q3","JCRName":"CHEMISTRY, PHYSICAL","Score":null,"Total":0}
引用次数: 0
Abstract
Accurate computations of experimental observables are essential for interpreting the high information content held within x-ray spectra. However, for complicated systems this can be difficult, a challenge compounded when dynamics becomes important owing to the large number of calculations required to capture the time-evolving observable. While machine learning architectures have been shown to represent a promising approach for rapidly predicting spectral lineshapes, achieving simultaneously accurate and sufficiently comprehensive training data is challenging. Herein, we introduce Δ-learning for x-ray spectroscopy. Instead of directly learning the structure-spectrum relationship, the Δ-model learns the structure dependent difference between a higher and lower level of theory. Consequently, once developed these models can be used to translate spectral shapes obtained from lower levels of theory to mimic those corresponding to higher levels of theory. Ultimately, this achieves accurate simulations with a much reduced computational burden as only the lower level of theory is computed, while the model can instantaneously transform this to a spectrum equivalent to a higher level of theory. Our present model, demonstrated herein, learns the difference between TDDFT(BLYP) and TDDFT(B3LYP) spectra. Its effectiveness is illustrated using simulations of Rh L3-edge spectra tracking the C-H activation of octane by a cyclopentadienyl rhodium carbonyl complex.
Structural Dynamics-UsCHEMISTRY, PHYSICALPHYSICS, ATOMIC, MOLECU-PHYSICS, ATOMIC, MOLECULAR & CHEMICAL
CiteScore
5.50
自引率
3.60%
发文量
24
审稿时长
16 weeks
期刊介绍:
Structural Dynamics focuses on the recent developments in experimental and theoretical methods and techniques that allow a visualization of the electronic and geometric structural changes in real time of chemical, biological, and condensed-matter systems. The community of scientists and engineers working on structural dynamics in such diverse systems often use similar instrumentation and methods.
The journal welcomes articles dealing with fundamental problems of electronic and structural dynamics that are tackled by new methods, such as:
Time-resolved X-ray and electron diffraction and scattering,
Coherent diffractive imaging,
Time-resolved X-ray spectroscopies (absorption, emission, resonant inelastic scattering, etc.),
Time-resolved electron energy loss spectroscopy (EELS) and electron microscopy,
Time-resolved photoelectron spectroscopies (UPS, XPS, ARPES, etc.),
Multidimensional spectroscopies in the infrared, the visible and the ultraviolet,
Nonlinear spectroscopies in the VUV, the soft and the hard X-ray domains,
Theory and computational methods and algorithms for the analysis and description of structuraldynamics and their associated experimental signals.
These new methods are enabled by new instrumentation, such as:
X-ray free electron lasers, which provide flux, coherence, and time resolution,
New sources of ultrashort electron pulses,
New sources of ultrashort vacuum ultraviolet (VUV) to hard X-ray pulses, such as high-harmonic generation (HHG) sources or plasma-based sources,
New sources of ultrashort infrared and terahertz (THz) radiation,
New detectors for X-rays and electrons,
New sample handling and delivery schemes,
New computational capabilities.