Predicting electronic structures at any length scale with machine learning - npj Computational Materials

日本 ニュース ニュース

Predicting electronic structures at any length scale with machine learning - npj Computational Materials
日本 最新ニュース,日本 見出し
  • 📰 physorg_com
  • ⏱ Reading Time:
  • 69 sec. here
  • 3 min. at publisher
  • 📊 Quality Score:
  • News: 31%
  • Publisher: 55%

npj Computational Materials - Predicting electronic structures at any length scale with machine learning

, slight deviations between the radial distribution functions of different system sizes can be seen, most notably for the cells containing a stacking fault. Overall, these deviations are small in magnitude, especially for the unperturbed cells, and generally, all radial distribution functions agree very well up to the cutoff radius from which information is incorporated into the bispectrum descriptors.

This analysis hence provides evidence that training, inference test, and the ultra-large simulation cells possess, on average, the same local environments. It indicates that our MALA predictions of the electronic structure and energy are based on interpolations on observed data. Therefore our models can be expected to be accurate at ultra-large scales far exceeding those for which reference data exists.

At system sizes where DFT benchmarks are still available, we demonstrate that our ML model is capable of reproducing energies and electronic densities of extended systems at virtually no loss in accuracy while outperforming other ML models that are based solely on energy. Furthermore, we demonstrate that our ML workflow enables predicting the electronic structure for systems with more than 100,000 atoms at a very low computational cost.

We expect our ML model to set standards in a number of ways. Using our ML model either directly or in conjunction with other ML workflows, such as ML interatomic potentials for pre-sampling of atomic configurations, will enable first-principles modeling of materials without finite-size errors. Combined with Monte-Carlo sampling and atomic forces from automatic differentiation, our ML model can replace ML interatomic potentials and yield thermodynamic observables at much higher accuracy.

このニュースをすぐに読めるように要約しました。ニュースに興味がある場合は、ここで全文を読むことができます。 続きを読む:

physorg_com /  🏆 388. in US

日本 最新ニュース, 日本 見出し

Similar News:他のニュース ソースから収集した、これに似たニュース記事を読むこともできます。

Train derails outside rural Wisconsin village, authorities say no hazardous materials leakedTrain derails outside rural Wisconsin village, authorities say no hazardous materials leakedInvestigators are looking into what caused a freight train to derail outside a small village in south-central Wisconsin.
続きを読む »

Affidavit: Englewood police find bomb-making materials, suspect says he was making homemade fireworksAffidavit: Englewood police find bomb-making materials, suspect says he was making homemade fireworksA man said he was trying to create homemade fireworks after police found bomb-making materials in an Englewood garage on Thursday.
続きを読む »

A Jacksonville convicted sex offender indicted for producing, distributing and possessing child sexual abuse materialsA Jacksonville convicted sex offender indicted for producing, distributing and possessing child sexual abuse materialsA convicted sex offender from Jacksonville has been indicted for committing multiple child exploitation offenses. Kevin John Merritt, 61, has been indicted on charges of producing, distributing, and possessing child sexual abuse materials.
続きを読む »

China's export curbs on raw materials are concerning, Yellen says on Beijing trip | CNN BusinessChina's export curbs on raw materials are concerning, Yellen says on Beijing trip | CNN BusinessBeijing’s newly unveiled export controls on two strategic raw materials critical to the global chipmaking industry are a “concern,” US Treasury Secretary Janet Yellen said on the first day of her trip to China, where she met the country’s premier in an effort to stabilize bilateral relations after a particularly difficult period.
続きを読む »

Tenba's Expanded Axis V2 Bag Series Includes Stylish Slings and BackpacksTenba's Expanded Axis V2 Bag Series Includes Stylish Slings and BackpacksTenba's expanded Axis V2 bag collection features three new mid-size backpacks and a pair of stylish slings.
続きを読む »

Machine Gun Kelly Fulfills Fan’s Wish by Punching Him in the FaceMachine Gun Kelly Fulfills Fan’s Wish by Punching Him in the Face“I love you,” MGK assured the fan after the punch.
続きを読む »



Render Time: 2025-03-01 04:36:57