AIMNet2: A Neural Network Potential to Meet your Neutral, Charged, Organic, and Elemental-Organic Needs. Anstine, D., Zubatyuk, R., & Isayev, O. 2023. ![AIMNet2: A Neural Network Potential to Meet your Neutral, Charged, Organic, and Elemental-Organic Needs [link]](https://bibbase.org/img/filetypes/link.svg "link")
Paper doi abstract bibtex Machine learned interatomic potentials (MLIPs) are reshaping computational chemistry practices because of their ability to drastically exceed the accuracy-length/time scale tradeoff. Despite this attraction, the benefits of such efficiency are only impactful when an MLIP uniquely enables insight into a target system or is broadly transferable outside of the training dataset, where models achieving the latter are seldom reported. In this work, we present the 2nd generation of our atoms-in-molecules neural network potential (AIMNet2), which is applicable to species composed of up to 14 chemical elements in both neutral and charged states, making it a valuable model for modeling the majority of non-metallic compounds. Using an exhaustive dataset of 20 million hybrid quantum chemical calculations, AIMNet2 combines ML-parameterized short-range and physics-based long-range terms to attain generalizability that reaches from simple organics to diverse molecules with “exotic” element-organic bonding. We show that AIMNet2 outperforms semi-empirical GFN-xTB and is on par with reference density functional theory for interaction energy contributions, conformer search tasks, torsion rotation profiles, and molecular-to-macromolecular geometry optimization. Overall, the demonstrated chemical coverage and computational efficiency of AIMNet2 is a significant step toward providing access to MLIPs that avoid the crucial limitation of curating additional quantum chemical data and retraining with each new application.
@misc{anstine_aimnet2_2023,
title = {{AIMNet}2: A Neural Network Potential to Meet your Neutral, Charged, Organic, and Elemental-Organic Needs},
url = {https://chemrxiv.org/engage/chemrxiv/article-details/6525b39e8bab5d2055123f75},
doi = {10.26434/chemrxiv-2023-296ch},
shorttitle = {{AIMNet}2},
abstract = {Machine learned interatomic potentials ({MLIPs}) are reshaping computational chemistry practices because of their ability to drastically exceed the accuracy-length/time scale tradeoff. Despite this attraction, the benefits of such efficiency are only impactful when an {MLIP} uniquely enables insight into a target system or is broadly transferable outside of the training dataset, where models achieving the latter are seldom reported. In this work, we present the 2nd generation of our atoms-in-molecules neural network potential ({AIMNet}2), which is applicable to species composed of up to 14 chemical elements in both neutral and charged states, making it a valuable model for modeling the majority of non-metallic compounds. Using an exhaustive dataset of 20 million hybrid quantum chemical calculations, {AIMNet}2 combines {ML}-parameterized short-range and physics-based long-range terms to attain generalizability that reaches from simple organics to diverse molecules with “exotic” element-organic bonding. We show that {AIMNet}2 outperforms semi-empirical {GFN}-{xTB} and is on par with reference density functional theory for interaction energy contributions, conformer search tasks, torsion rotation profiles, and molecular-to-macromolecular geometry optimization. Overall, the demonstrated chemical coverage and computational efficiency of {AIMNet}2 is a significant step toward providing access to {MLIPs} that avoid the crucial limitation of curating additional quantum chemical data and retraining with each new application.},
publisher = {{ChemRxiv}},
author = {Anstine, Dylan and Zubatyuk, Roman and Isayev, Olexandr},
urldate = {2023-10-13},
date = {2023-10-12},
year = {2023},
langid = {english},
file = {Anstine et al. - 2023 - AIMNet2 A Neural Network Potential to Meet your N.pdf:C\:\\Users\\aritra\\Zotero\\storage\\79TD5B22\\Anstine et al. - 2023 - AIMNet2 A Neural Network Potential to Meet your N.pdf:application/pdf},
}
Downloads: 0
{"_id":"BiL7zAMubJcPpqdR7","bibbaseid":"anstine-zubatyuk-isayev-aimnet2aneuralnetworkpotentialtomeetyourneutralchargedorganicandelementalorganicneeds-2023","author_short":["Anstine, D.","Zubatyuk, R.","Isayev, O."],"bibdata":{"bibtype":"misc","type":"misc","title":"AIMNet2: A Neural Network Potential to Meet your Neutral, Charged, Organic, and Elemental-Organic Needs","url":"https://chemrxiv.org/engage/chemrxiv/article-details/6525b39e8bab5d2055123f75","doi":"10.26434/chemrxiv-2023-296ch","shorttitle":"AIMNet2","abstract":"Machine learned interatomic potentials (MLIPs) are reshaping computational chemistry practices because of their ability to drastically exceed the accuracy-length/time scale tradeoff. Despite this attraction, the benefits of such efficiency are only impactful when an MLIP uniquely enables insight into a target system or is broadly transferable outside of the training dataset, where models achieving the latter are seldom reported. In this work, we present the 2nd generation of our atoms-in-molecules neural network potential (AIMNet2), which is applicable to species composed of up to 14 chemical elements in both neutral and charged states, making it a valuable model for modeling the majority of non-metallic compounds. Using an exhaustive dataset of 20 million hybrid quantum chemical calculations, AIMNet2 combines ML-parameterized short-range and physics-based long-range terms to attain generalizability that reaches from simple organics to diverse molecules with “exotic” element-organic bonding. We show that AIMNet2 outperforms semi-empirical GFN-xTB and is on par with reference density functional theory for interaction energy contributions, conformer search tasks, torsion rotation profiles, and molecular-to-macromolecular geometry optimization. Overall, the demonstrated chemical coverage and computational efficiency of AIMNet2 is a significant step toward providing access to MLIPs that avoid the crucial limitation of curating additional quantum chemical data and retraining with each new application.","publisher":"ChemRxiv","author":[{"propositions":[],"lastnames":["Anstine"],"firstnames":["Dylan"],"suffixes":[]},{"propositions":[],"lastnames":["Zubatyuk"],"firstnames":["Roman"],"suffixes":[]},{"propositions":[],"lastnames":["Isayev"],"firstnames":["Olexandr"],"suffixes":[]}],"urldate":"2023-10-13","date":"2023-10-12","year":"2023","langid":"english","file":"Anstine et al. - 2023 - AIMNet2 A Neural Network Potential to Meet your N.pdf:C\\:\\\\Users\\\\aritra\\\\Zotero\\\\storage\\\\79TD5B22\\\\Anstine et al. - 2023 - AIMNet2 A Neural Network Potential to Meet your N.pdf:application/pdf","bibtex":"@misc{anstine_aimnet2_2023,\n\ttitle = {{AIMNet}2: A Neural Network Potential to Meet your Neutral, Charged, Organic, and Elemental-Organic Needs},\n\turl = {https://chemrxiv.org/engage/chemrxiv/article-details/6525b39e8bab5d2055123f75},\n\tdoi = {10.26434/chemrxiv-2023-296ch},\n\tshorttitle = {{AIMNet}2},\n\tabstract = {Machine learned interatomic potentials ({MLIPs}) are reshaping computational chemistry practices because of their ability to drastically exceed the accuracy-length/time scale tradeoff. Despite this attraction, the benefits of such efficiency are only impactful when an {MLIP} uniquely enables insight into a target system or is broadly transferable outside of the training dataset, where models achieving the latter are seldom reported. In this work, we present the 2nd generation of our atoms-in-molecules neural network potential ({AIMNet}2), which is applicable to species composed of up to 14 chemical elements in both neutral and charged states, making it a valuable model for modeling the majority of non-metallic compounds. Using an exhaustive dataset of 20 million hybrid quantum chemical calculations, {AIMNet}2 combines {ML}-parameterized short-range and physics-based long-range terms to attain generalizability that reaches from simple organics to diverse molecules with “exotic” element-organic bonding. We show that {AIMNet}2 outperforms semi-empirical {GFN}-{xTB} and is on par with reference density functional theory for interaction energy contributions, conformer search tasks, torsion rotation profiles, and molecular-to-macromolecular geometry optimization. Overall, the demonstrated chemical coverage and computational efficiency of {AIMNet}2 is a significant step toward providing access to {MLIPs} that avoid the crucial limitation of curating additional quantum chemical data and retraining with each new application.},\n\tpublisher = {{ChemRxiv}},\n\tauthor = {Anstine, Dylan and Zubatyuk, Roman and Isayev, Olexandr},\n\turldate = {2023-10-13},\n\tdate = {2023-10-12},\n\tyear = {2023},\n\tlangid = {english},\n\tfile = {Anstine et al. - 2023 - AIMNet2 A Neural Network Potential to Meet your N.pdf:C\\:\\\\Users\\\\aritra\\\\Zotero\\\\storage\\\\79TD5B22\\\\Anstine et al. - 2023 - AIMNet2 A Neural Network Potential to Meet your N.pdf:application/pdf},\n}\n\n","author_short":["Anstine, D.","Zubatyuk, R.","Isayev, O."],"key":"anstine_aimnet2_2023","id":"anstine_aimnet2_2023","bibbaseid":"anstine-zubatyuk-isayev-aimnet2aneuralnetworkpotentialtomeetyourneutralchargedorganicandelementalorganicneeds-2023","role":"author","urls":{"Paper":"https://chemrxiv.org/engage/chemrxiv/article-details/6525b39e8bab5d2055123f75"},"metadata":{"authorlinks":{}}},"bibtype":"misc","biburl":"https://bibbase.org/network/files/L6oHYoy8a5nTT8bpW","dataSources":["DwbKXqN9YyN8gkNab","Z9u3GzKyzYSLc7khn"],"keywords":[],"search_terms":["aimnet2","neural","network","potential","meet","neutral","charged","organic","elemental","organic","needs","anstine","zubatyuk","isayev"],"title":"AIMNet2: A Neural Network Potential to Meet your Neutral, Charged, Organic, and Elemental-Organic Needs","year":2023}