{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,6,25]],"date-time":"2025-06-25T09:40:03Z","timestamp":1750844403749,"version":"3.41.0"},"reference-count":49,"publisher":"IOP Publishing","issue":"2","license":[{"start":{"date-parts":[[2025,6,25]],"date-time":"2025-06-25T00:00:00Z","timestamp":1750809600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"},{"start":{"date-parts":[[2025,6,25]],"date-time":"2025-06-25T00:00:00Z","timestamp":1750809600000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/iopscience.iop.org\/info\/page\/text-and-data-mining"}],"funder":[{"name":"European Research Council","award":["MOPPEX 833460"],"award-info":[{"award-number":["MOPPEX 833460"]}]},{"name":"Leading Innovation and Entrepreneurship Team of Zhejiang Province of China","award":["2023R01008"],"award-info":[{"award-number":["2023R01008"]}]}],"content-domain":{"domain":["iopscience.iop.org"],"crossmark-restriction":false},"short-container-title":["Mach. Learn.: Sci. Technol."],"published-print":{"date-parts":[[2025,6,30]]},"abstract":"<jats:title>Abstract<\/jats:title>\n               <jats:p>Computational astrochemical models are essential for helping us interpret and understand the observations of different astrophysical environments. In the age of high-resolution telescopes such as JWST and ALMA, the substructure of many objects can be resolved, raising the need for astrochemical modeling at these smaller scales, meaning that the simulations of these objects need to include both the physics and chemistry to accurately model the observations. The computational cost of the simulations coupling both the three-dimensional hydrodynamics and chemistry is enormous, creating an opportunity for surrogate models that can effectively substitute the chemical solver. In this work we present surrogate models that can replace the original chemical code, namely Latent Augmented Neural Ordinary Differential Equations. We train these surrogate architectures on three datasets of increasing physical complexity, with the last dataset derived directly from a three-dimensional simulation of a molecular cloud using a photodissociation region (PDR) code, <jats:monospace>3D-PDR<\/jats:monospace>. We show that these surrogate models can provide speedup and reproduce the original observable column density maps of the dataset. This enables the rapid inference of the chemistry (on the GPU), allowing for the faster statistical inference of observations or increasing the resolution in hydrodynamical simulations of astrophysical environments.<\/jats:p>","DOI":"10.1088\/2632-2153\/ade4ee","type":"journal-article","created":{"date-parts":[[2025,6,16]],"date-time":"2025-06-16T22:50:36Z","timestamp":1750114236000},"page":"025069","update-policy":"https:\/\/doi.org\/10.1088\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["NeuralPDR: neural differential equations as surrogate models for photodissociation regions"],"prefix":"10.1088","volume":"6","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-4346-5858","authenticated-orcid":true,"given":"Gijs","family":"Vermari\u00ebn","sequence":"first","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0003-2733-4580","authenticated-orcid":false,"given":"Thomas G","family":"Bisbas","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0001-8504-8844","authenticated-orcid":false,"given":"Serena","family":"Viti","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0009-0005-4074-5116","authenticated-orcid":true,"given":"Yue","family":"Zhao","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4955-0279","authenticated-orcid":false,"given":"Xuefei","family":"Tang","sequence":"additional","affiliation":[]},{"given":"Rahul","family":"Ravichandran","sequence":"additional","affiliation":[]}],"member":"266","published-online":{"date-parts":[[2025,6,25]]},"reference":[{"article-title":"xLSTM: extended long short-term memory","year":"2024","author":"Beck","key":"mlstade4eebib1"},{"key":"mlstade4eebib2","doi-asserted-by":"publisher","first-page":"2100","DOI":"10.1111\/j.1365-2966.2012.22077.x","article-title":"3D-PDR: a new three-dimensional astrochemistry code for treating photodissociation regions","volume":"427","author":"Bisbas","year":"2012","journal-title":"Mon. Not. R. Astron. Soc."},{"key":"mlstade4eebib3","doi-asserted-by":"publisher","first-page":"729","DOI":"10.1093\/mnras\/stac3487","article-title":"PDFCHEM: a new fast method to determine ISM properties and infer environmental parameters using probability distributions","volume":"519","author":"Bisbas","year":"2023","journal-title":"Mon. Not. R. Astron. Soc."},{"year":"2023","author":"Bovino","key":"mlstade4eebib4"},{"article-title":"JAX: composable transformations of Python+NumPy programs 0.3.13","year":"2018","author":"Bradbury","key":"mlstade4eebib5"},{"key":"mlstade4eebib6","doi-asserted-by":"publisher","first-page":"5718","DOI":"10.1093\/mnras\/stac3512","article-title":"Neural networks: solving the chemistry of the interstellar medium","volume":"518","author":"Branca","year":"2022","journal-title":"Mon. Not. R. Astron. Soc."},{"key":"mlstade4eebib7","doi-asserted-by":"crossref","DOI":"10.1051\/0004-6361\/202449193","article-title":"Emulating the interstellar medium chemistry with neural operators","author":"Branca","year":"2024"},{"article-title":"Neural ordinary differential equations","year":"2019","author":"Chen","key":"mlstade4eebib8"},{"year":"2013","author":"Collette","key":"mlstade4eebib9"},{"article-title":"Efficient marginalization of discrete and structured latent variables via sparsity","year":"2020","author":"Correia","key":"mlstade4eebib10"},{"key":"mlstade4eebib11","doi-asserted-by":"publisher","DOI":"10.5281\/zenodo.8409685","article-title":"Scientific colour maps","author":"Crameri","year":"2023"},{"key":"mlstade4eebib12","doi-asserted-by":"publisher","first-page":"A117","DOI":"10.1051\/0004-6361\/201935973","article-title":"Incorporating astrochemistry into molecular line modelling via emulation","volume":"630","author":"de Mijolla","year":"2019","journal-title":"Astron. Astrophys."},{"key":"mlstade4eebib13","doi-asserted-by":"publisher","first-page":"595","DOI":"10.1086\/190513","article-title":"Photoelectric heating of interstellar gas","volume":"36","author":"Draine","year":"1978","journal-title":"Astrophys. J. Suppl. Ser."},{"article-title":"Augmented neural ODEs","year":"2019","author":"Dupont","key":"mlstade4eebib14"},{"key":"mlstade4eebib15","first-page":"pp 156","article-title":"Speed up training with variable length inputs by efficient batching strategies","author":"Ge","year":"2021"},{"key":"mlstade4eebib16","doi-asserted-by":"publisher","first-page":"42","DOI":"10.3847\/1538-4365\/aceaf9","article-title":"Implementation of Chemistry in the Athena++ Code","volume":"268","author":"Gong","year":"2023","journal-title":"Astrophys. J. Suppl. Ser."},{"key":"mlstade4eebib17","doi-asserted-by":"publisher","first-page":"759","DOI":"10.1086\/427976","article-title":"HEALPix: a framework for high-resolution discretization and fast analysis of data distributed on the sphere","volume":"622","author":"G\u00f3rski","year":"2005","journal-title":"Astrophys. J."},{"article-title":"Reducing the complexity of chemical networks via interpretable autoencoders","year":"2021","author":"Grassi","key":"mlstade4eebib18"},{"article-title":"Mapping synthetic observations to prestellar core models: an interpretable machine learning approach","year":"2025","author":"Grassi","key":"mlstade4eebib19"},{"article-title":"FFJORD: free-form continuous dynamics for scalable reversible generative models","year":"2018","author":"Grathwohl","key":"mlstade4eebib20"},{"key":"mlstade4eebib21","doi-asserted-by":"publisher","first-page":"2199","DOI":"10.1093\/mnras\/stab1347","article-title":"STARFORGE: toward a comprehensive numerical model of star cluster formation and feedback","volume":"506","author":"Grudi\u0107","year":"2021","journal-title":"Mon. Not. R. Astron. Soc."},{"article-title":"Mamba: linear-time sequence modeling with selective state spaces","year":"2024","author":"Gu","key":"mlstade4eebib22"},{"key":"mlstade4eebib23","doi-asserted-by":"publisher","first-page":"2050","DOI":"10.1111\/j.1365-2966.2009.15598.x","article-title":"The relation between optical extinction and hydrogen column density in the Galaxy","volume":"400","author":"G\u00fcver","year":"2009","journal-title":"Mon. Not. R. Astron. Soc."},{"key":"mlstade4eebib24","doi-asserted-by":"publisher","first-page":"404","DOI":"10.1093\/mnras\/stad2814","article-title":"Understanding molecular abundances in star-forming regions using interpretable machine learning","volume":"526","author":"Heyl","year":"2023","journal-title":"Mon. Not. R. Astron. Soc."},{"key":"mlstade4eebib25","doi-asserted-by":"publisher","first-page":"A76","DOI":"10.1051\/0004-6361\/202140357","article-title":"Chemulator: fast, accurate thermochemistry for dynamical models through emulation","volume":"653","author":"Holdship","year":"2021","journal-title":"Astron. Astrophys."},{"key":"mlstade4eebib26","doi-asserted-by":"publisher","first-page":"90","DOI":"10.1109\/MCSE.2007.55","article-title":"Matplotlib: a 2D graphics environment","volume":"9","author":"Hunter","year":"2007","journal-title":"Comput. Sci. Eng."},{"article-title":"CODES: benchmarking coupled ODE surrogates","year":"2024","author":"Janssen","key":"mlstade4eebib27"},{"article-title":"On neural differential equations","year":"2022","author":"Kidger","key":"mlstade4eebib28"},{"key":"mlstade4eebib29","doi-asserted-by":"publisher","first-page":"233","DOI":"10.1002\/aic.690370209","article-title":"Nonlinear principal component analysis using autoassociative neural networks","volume":"37","author":"Kramer","year":"1991","journal-title":"AIChE J."},{"key":"mlstade4eebib30","doi-asserted-by":"publisher","first-page":"489","DOI":"10.1023\/B:BITN.0000046811.70614.38","article-title":"Singly diagonally implicit Runge\u2013Kutta methods with an explicit first stage","volume":"44","author":"Kv\u00e6rn\u00f8","year":"2004","journal-title":"BIT Numer. Math."},{"key":"mlstade4eebib31","doi-asserted-by":"publisher","DOI":"10.1098\/rspa.2021.0162","article-title":"Parameterized neural ordinary differential equations: applications to computational physics problems","volume":"477","author":"Lee","year":"2021","journal-title":"Proc. R. Soc. A"},{"article-title":"SGDR: stochastic gradient descent with warm restarts","year":"2016","author":"Loshchilov","key":"mlstade4eebib32"},{"article-title":"Decoupled weight decay regularization","year":"2017","author":"Loshchilov","key":"mlstade4eebib33"},{"key":"mlstade4eebib34","doi-asserted-by":"crossref","DOI":"10.3847\/1538-4357\/ad47a1","article-title":"MACE: a machine learning approach to chemistry emulation","author":"Maes","year":"2024"},{"article-title":"On the difficulty of training recurrent neural networks","year":"2013","author":"Pascanu","key":"mlstade4eebib35"},{"key":"mlstade4eebib36","doi-asserted-by":"publisher","first-page":"A74","DOI":"10.1051\/0004-6361\/202348244","article-title":"PDRs4All - III. JWST\u2019s NIR spectroscopic view of the Orion Bar","volume":"685","author":"Peeters","year":"2024","journal-title":"Astron. Astrophys."},{"key":"mlstade4eebib37","doi-asserted-by":"publisher","first-page":"4930","DOI":"10.1093\/mnras\/stae1432","article-title":"A fast neural emulator for interstellar chemistry","volume":"531","author":"Ramos","year":"2024","journal-title":"Mon. Not. R. Astron. Soc."},{"key":"mlstade4eebib38","doi-asserted-by":"publisher","first-page":"187","DOI":"10.1051\/0004-6361:20065918","article-title":"A photon dominated region code comparison study","volume":"467","author":"Rollig","year":"2007","journal-title":"Astron. Astrophys."},{"article-title":"Latent ODEs for irregularly-sampled time series","year":"2019","author":"Rubanova","key":"mlstade4eebib39"},{"key":"mlstade4eebib40","doi-asserted-by":"publisher","first-page":"1465","DOI":"10.1093\/mnras\/stz3563","article-title":"SILCC-Zoom: H2 and CO-dark gas in molecular clouds - the impact of feedback and magnetic fields","volume":"492","author":"Seifried","year":"2020","journal-title":"Mon. Not. R. Astron. Soc."},{"key":"mlstade4eebib41","doi-asserted-by":"publisher","first-page":"4797","DOI":"10.1093\/mnras\/stx2343","article-title":"SILCC-Zoom: the dynamic and chemical evolution of molecular clouds","volume":"472","author":"Seifried","year":"2017","journal-title":"Mon. Not. R. Astron. Soc."},{"article-title":"Speeding up astrochemical reaction networks with autoencoders and neural ODEs","year":"2023","author":"Sulzer","key":"mlstade4eebib42"},{"article-title":"Reduced order model for chemical kinetics: a case study with primordial chemical network","year":"2022","author":"Tang","key":"mlstade4eebib43"},{"key":"mlstade4eebib44","doi-asserted-by":"publisher","first-page":"770","DOI":"10.1016\/j.camwa.2011.06.002","article-title":"Runge\u2013Kutta pairs of order 5(4) satisfying only the first column simplifying assumption","volume":"62","author":"Tsitouras","year":"2011","journal-title":"Comput. Math. Appl."},{"key":"mlstade4eebib45","doi-asserted-by":"publisher","DOI":"10.5281\/zenodo.15699332","article-title":"uclchem\/neuralpdr: v0.2 (Zenodo)","author":"Vermari\u00ebn","year":"2025"},{"article-title":"3D-PDR Orion dataset and NeuralPDR: neural differential equations for photodissociation regions","year":"2024","author":"Vermari\u00ebn","key":"mlstade4eebib46"},{"key":"mlstade4eebib47","doi-asserted-by":"publisher","first-page":"247","DOI":"10.1146\/annurev-astro-052920-010254","article-title":"Photodissociation and x-ray-Dominated Regions","volume":"60","author":"Wolfire","year":"2022","journal-title":"Annu. Rev. Astron. Astrophys."},{"key":"mlstade4eebib48","doi-asserted-by":"publisher","first-page":"37","DOI":"10.3847\/1538-4357\/ad64c6","article-title":"Turbulent diffuse molecular media with nonideal magnetohydrodynamics and consistent thermochemistry: numerical simulations and dynamic characteristics","volume":"973","author":"Yue","year":"2024","journal-title":"Astrophys. J."},{"article-title":"Construction of smoothing splines by the generalized cross-validation method","year":"2022","author":"Zemlyanoy","key":"mlstade4eebib49"}],"container-title":["Machine Learning: Science and Technology"],"original-title":[],"link":[{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ade4ee","content-type":"text\/html","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ade4ee\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ade4ee","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ade4ee\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ade4ee\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ade4ee\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ade4ee\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"similarity-checking"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ade4ee\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,6,25]],"date-time":"2025-06-25T09:00:48Z","timestamp":1750842048000},"score":1,"resource":{"primary":{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ade4ee"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,6,25]]},"references-count":49,"journal-issue":{"issue":"2","published-online":{"date-parts":[[2025,6,25]]},"published-print":{"date-parts":[[2025,6,30]]}},"URL":"https:\/\/doi.org\/10.1088\/2632-2153\/ade4ee","relation":{},"ISSN":["2632-2153"],"issn-type":[{"type":"electronic","value":"2632-2153"}],"subject":[],"published":{"date-parts":[[2025,6,25]]},"assertion":[{"value":"NeuralPDR: neural differential equations as surrogate models for photodissociation regions","name":"article_title","label":"Article Title"},{"value":"Machine Learning: Science and Technology","name":"journal_title","label":"Journal Title"},{"value":"paper","name":"article_type","label":"Article Type"},{"value":"\u00a9 2025 The Author(s). Published by IOP Publishing Ltd","name":"copyright_information","label":"Copyright Information"},{"value":"2025-03-21","name":"date_received","label":"Date Received","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2025-06-16","name":"date_accepted","label":"Date Accepted","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2025-06-25","name":"date_epub","label":"Online publication date","group":{"name":"publication_dates","label":"Publication dates"}}]}}