# Distance-insensitive Graph Neural Networks and multi-task learning for accurate prediction of adsorption energies on alloy nanoclusters

https://mdr.nims.go.jp/datasets/bddf1083-b445-4541-a756-89fa937aebdf

## File

- [Otsuka_ComputMatSci_2026.pdf](https://mdr.nims.go.jp/filesets/19b68c4c-22db-43bc-849e-eceb9439141b/download) ([Detail](https://mdr.nims.go.jp/filesets/19b68c4c-22db-43bc-849e-eceb9439141b.md))

## Id

bddf1083-b445-4541-a756-89fa937aebdf

## Local identifier



## Visibility

open_to_public

## State

published

## Created at

2026-08-18T03:58:02.149107Z

## Updated at

2026-08-18T04:40:15.948303Z

## Published at

2026-08-18T07:29:18.591079Z

## Doi



## First published url

https://doi.org/10.1016/j.commatsci.2026.114987

## Date published

2026-08-14

## Recorded date published

2026-9

## Resource type

journal_article

## Manuscript type

vor

## Collection



## Title

- title: Distance-insensitive Graph Neural Networks and multi-task learning for accurate
    prediction of adsorption energies on alloy nanoclusters
  title_type: original
  lang: en

## Description

- description: "Developing a methodology that enables accurate yet computationally
    efficient\r\nprediction of adsorption energies is pivotal for accelerating the
    discovery of high-\r\nperformance alloy catalysts. However, current data-driven
    approaches face significant\r\nchallenges, particularly the scarcity of high-accuracy
    distance-insensitive machine-\r\nlearning models suitable for screening unknown
    structures and the limited structural\r\ndiversity of available adsorption energy
    datasets. To address these challenges, we\r\ndeveloped an enhanced distance-insensitive
    graph neural network model, named\r\nBond-type Embedded Orbital Graph Convolutional
    Neural Network (BE-OGCNN), that\r\nintegrates orbital interaction features to
    maximize expressivity without geometric\r\ndependency. In addition, we employed
    a multi-task learning framework using d-band\r\ncenter and total energy as auxiliary
    tasks. This strategy overcomes data scarcity by\r\neffectively exploiting abundant
    bulk crystal data that was previously underutilized for\r\nsurface property prediction.
    Our model achieved a mean absolute error of 0.042 eV on\r\n44-atom alloy clusters,
    demonstrating accuracy comparable to that of a state-of-the-art\r\ndistance-sensitive
    model. Moreover, the multi-task approach successfully improved\r\nprediction accuracy
    on larger 85-atom clusters, suggesting high potential of our\r\nframework for
    rapid and reliable screening of realistic catalyst nanoparticles."
  description_type: abstract
  lang: und

## Creator

- name: Koki Otsuka
  role: author
- name: Anh Khoa Augustin Lu
  role: author
  orcid: https://orcid.org/0000-0003-4702-0933
- name: Koji Shimizu
  role: author
  orcid: https://orcid.org/0000-0001-5622-9582
- name: Satoshi Watanabe
  role: author
  orcid: https://orcid.org/0000-0002-8069-6938

## Contact agent



## Publisher

organization: Elsevier BV

## Managing organization



## Keyword

- subject: Adsorption energy prediction
  schema: not_defined
- subject: Graph neural networks
  schema: not_defined
- subject: Multi-task learning
  schema: not_defined
- subject: Alloy nanoclusters
  schema: not_defined
- subject: Catalyst screening
  schema: not_defined

## Rights

- identifier: https://creativecommons.org/licenses/by/4.0/

## Other identifier(s)



## Data origin



## Embargo



## Journal

- title: Computational Materials Science
  issn: '09270256'
  volume: '274'
  article_number: '114987'

## Conference



## Related item



## Funding

- identifier: 24K01284
  funder_name: Japan Society for the Promotion of Science
- identifier: JPMJSC21E2
  funder_name: Strategic International Collaborative Research Program

## Instrument



## Instrument operator



## Instrument managing organization



## Measurement method



## Specimen



## Chemical composition



## Structure for specimen



## Structural feature for specimen



## Specific property for specimen



## Process for specimen treatment



## Computational method



## Energy level/transition state



## Software



## Custom property



## Fileset

- id: 19b68c4c-22db-43bc-849e-eceb9439141b
  filename: Otsuka_ComputMatSci_2026.pdf
  content_type: application/pdf
  size: 1212402
  md5: 5b0e254685512e0fa3a8ba82df105dc4

## Thumbnail

fileset_id: 19b68c4c-22db-43bc-849e-eceb9439141b
filename: Otsuka_ComputMatSci_2026.pdf