Slim multi-scale convolutional autoencoder-based reduced-order models for interpretable features of a complex dynamical system
In recent years, data-driven deep learning models have gained significant importance in the analysis of turbulent dynamical systems. Within the context of reduced-order models, convolutional autoencoders (CAEs) pose a universally applicable alternative to conventional approaches. They can learn nonl...
Saved in:
| Published in: | APL machine learning Vol. 3; no. 1; pp. 016112 - 016112-15 |
|---|---|
| Main Authors: | , , , , , |
| Format: | Journal Article |
| Language: | English |
| Published: |
AIP Publishing LLC
01.03.2025
|
| ISSN: | 2770-9019, 2770-9019 |
| Online Access: | Get full text |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Abstract | In recent years, data-driven deep learning models have gained significant importance in the analysis of turbulent dynamical systems. Within the context of reduced-order models, convolutional autoencoders (CAEs) pose a universally applicable alternative to conventional approaches. They can learn nonlinear transformations directly from data, without prior knowledge of the system. However, the features generated by such models lack interpretability. Thus, the resulting model is a black-box that effectively reduces the complexity of the system but does not provide insights into the meaning of the latent features. To address this critical issue, we introduce a novel interpretable CAE approach for high-dimensional fluid flow data that maintains the reconstruction quality of conventional CAEs and allows for feature interpretation. Our method can be easily integrated into any existing CAE architecture with minor modifications of the training process. We compare our approach to Proper Orthogonal Decomposition (POD) and two existing methods for interpretable CAEs. We apply all methods to three different experimental turbulent Rayleigh–Bénard convection datasets with varying complexity. Our results show that the proposed method is lightweight, easy to train, and achieves relative reconstruction performance improvements of up to 6.4% over POD for 64 modes. The relative improvement increases to up to 229.8% as the number of modes decreases. In addition, our method delivers interpretable features similar to those of POD and is significantly less resource-intensive than existing CAE approaches, using less than 2% of the parameters. These approaches either trade interpretability for reconstruction performance or only provide interpretability to a limited extent. |
|---|---|
| AbstractList | In recent years, data-driven deep learning models have gained significant importance in the analysis of turbulent dynamical systems. Within the context of reduced-order models, convolutional autoencoders (CAEs) pose a universally applicable alternative to conventional approaches. They can learn nonlinear transformations directly from data, without prior knowledge of the system. However, the features generated by such models lack interpretability. Thus, the resulting model is a black-box that effectively reduces the complexity of the system but does not provide insights into the meaning of the latent features. To address this critical issue, we introduce a novel interpretable CAE approach for high-dimensional fluid flow data that maintains the reconstruction quality of conventional CAEs and allows for feature interpretation. Our method can be easily integrated into any existing CAE architecture with minor modifications of the training process. We compare our approach to Proper Orthogonal Decomposition (POD) and two existing methods for interpretable CAEs. We apply all methods to three different experimental turbulent Rayleigh–Bénard convection datasets with varying complexity. Our results show that the proposed method is lightweight, easy to train, and achieves relative reconstruction performance improvements of up to 6.4% over POD for 64 modes. The relative improvement increases to up to 229.8% as the number of modes decreases. In addition, our method delivers interpretable features similar to those of POD and is significantly less resource-intensive than existing CAE approaches, using less than 2% of the parameters. These approaches either trade interpretability for reconstruction performance or only provide interpretability to a limited extent. |
| Author | Sharifi Ghazijahani, Mohammad Mäder, Patrick Teutsch, Philipp Schumacher, Jörg Pfeffer, Philipp Cierpka, Christian |
| Author_xml | – sequence: 1 givenname: Philipp orcidid: 0000-0002-4247-7961 surname: Teutsch fullname: Teutsch, Philipp – sequence: 2 givenname: Philipp orcidid: 0000-0001-8503-774X surname: Pfeffer fullname: Pfeffer, Philipp – sequence: 3 givenname: Mohammad orcidid: 0000-0001-5885-4138 surname: Sharifi Ghazijahani fullname: Sharifi Ghazijahani, Mohammad – sequence: 4 givenname: Christian orcidid: 0000-0002-8464-5513 surname: Cierpka fullname: Cierpka, Christian – sequence: 5 givenname: Jörg orcidid: 0000-0002-1359-4536 surname: Schumacher fullname: Schumacher, Jörg – sequence: 6 givenname: Patrick orcidid: 0000-0001-6871-2707 surname: Mäder fullname: Mäder, Patrick |
| BookMark | eNpNkU1LxDAQhoMoqKsH_0GuHqqTj7bboyx-LAge1HOZJhOppM2SpOJe_O1WdxFPMzwvPAPznrLDMYzE2IWAKwGVui6vQGqtRXXATmRdQ9GAaA7_7cfsPKV3AJBNAzXACft69v3Ah8nnvkgGPXETxo_gp9yHET3HKQcaTbAUiw4TWR7JToZsEeLM-DAnPnEXIu_HTHETKWM3axxhniIlHhzHWTpsPH1yux1x6Oc7PG1TpuGMHTn0ic73c8Fe725fVg_F49P9enXzWBgpIRdaOCiXWrlKC2isrbWutQSnSwNWKYukKxIdVsuaVGWUUcrNr6hdaUmUUqgFW--8NuB7u4n9gHHbBuzbXxDiW4sx98ZTK1XlliSc7KjSAAKFbYywdaeW-INn1-XOZWJIKZL78wlof3poy3bfg_oGoGB9rw |
| Cites_doi | 10.1063/1.4979665 10.1006/jsvi.2001.4041 10.3847/1538-4357/ad1c55 10.1146/annurev-fluid-010719-060214 10.1140/epje/s10189-023-00356-w 10.1038/s42256-024-00810-0 10.1140/epje/i2012-12058-1 10.1063/5.0191403 10.1063/5.0213700 10.1146/annurev.fluid.39.050905.110308 10.1063/5.0087977 10.1126/science.aaw4741 10.1098/rspa.2017.0844 10.1002/nme.6681 10.1146/annurev.fluid.25.1.539 10.1017/jfm.2019.822 10.1063/5.0074310 10.1631/fitee.1700808 10.1145/1553374.1553380 10.5555/2627435.2670313 10.1515/teme-2022-0121 10.1103/physrevfluids.5.113506 10.1073/pnas.1900358116 10.1063/5.0065637 10.1016/j.cma.2021.114181 10.1073/pnas.2213638120 10.1103/physrevresearch.4.033176 10.1063/5.0039986 10.1007/s00348-023-03736-2 10.1063/5.0020721 |
| ContentType | Journal Article |
| DBID | AAYXX CITATION DOA |
| DOI | 10.1063/5.0244416 |
| DatabaseName | CrossRef Open Access: DOAJ - Directory of Open Access Journals |
| DatabaseTitle | CrossRef |
| DatabaseTitleList | CrossRef |
| Database_xml | – sequence: 1 dbid: DOA name: DOAJ Directory of Open Access Journals url: https://www.doaj.org/ sourceTypes: Open Website |
| DeliveryMethod | fulltext_linktorsrc |
| Discipline | Computer Science |
| EISSN | 2770-9019 |
| EndPage | 016112-15 |
| ExternalDocumentID | oai_doaj_org_article_236f8e1f2be64001a1d9c1d7b38ae1f2 10_1063_5_0244416 |
| GroupedDBID | AAFWJ AAYXX ABDBF ABJGX AFPKN AGLKD AJDQP AKSGC ALMA_UNASSIGNED_HOLDINGS CITATION GROUPED_DOAJ M~E RQS |
| ID | FETCH-LOGICAL-c220t-41f05843f64109dd7447420f45c0d33dae46e1ba687e36c3c33f0247f5de15213 |
| IEDL.DBID | DOA |
| ISICitedReferencesCount | 1 |
| ISICitedReferencesURI | http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001493756700019&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D |
| ISSN | 2770-9019 |
| IngestDate | Mon Nov 10 04:31:02 EST 2025 Sat Nov 29 07:57:38 EST 2025 |
| IsDoiOpenAccess | true |
| IsOpenAccess | true |
| IsPeerReviewed | true |
| IsScholarly | true |
| Issue | 1 |
| Language | English |
| LinkModel | DirectLink |
| MergedId | FETCHMERGED-LOGICAL-c220t-41f05843f64109dd7447420f45c0d33dae46e1ba687e36c3c33f0247f5de15213 |
| ORCID | 0000-0001-5885-4138 0000-0001-8503-774X 0000-0002-1359-4536 0000-0001-6871-2707 0000-0002-8464-5513 0000-0002-4247-7961 |
| OpenAccessLink | https://doaj.org/article/236f8e1f2be64001a1d9c1d7b38ae1f2 |
| ParticipantIDs | doaj_primary_oai_doaj_org_article_236f8e1f2be64001a1d9c1d7b38ae1f2 crossref_primary_10_1063_5_0244416 |
| PublicationCentury | 2000 |
| PublicationDate | 2025-03-01 |
| PublicationDateYYYYMMDD | 2025-03-01 |
| PublicationDate_xml | – month: 03 year: 2025 text: 2025-03-01 day: 01 |
| PublicationDecade | 2020 |
| PublicationTitle | APL machine learning |
| PublicationYear | 2025 |
| Publisher | AIP Publishing LLC |
| Publisher_xml | – name: AIP Publishing LLC |
| References | 2025022614075408400_c18 (2025022614075408400_c42) 2014; 27 2025022614075408400_c16 (2025022614075408400_c41) 2014; 15 2025022614075408400_c36 2025022614075408400_c34 (2025022614075408400_c37) 2025; 3 (2025022614075408400_c13) 2021; 33 (2025022614075408400_c45) 2009 (2025022614075408400_c5) 2019; 116 (2025022614075408400_c46) 2022 (2025022614075408400_c4) 2023; 120 (2025022614075408400_c21) 2021; 33 (2025022614075408400_c39) 2013; 30 (2025022614075408400_c30) 2012; 35 (2025022614075408400_c9) 2023; 46 (2025022614075408400_c22) 2021; 11 (2025022614075408400_c17) 1993; 25 (2025022614075408400_c27) 2022; 388 2025022614075408400_c20 (2025022614075408400_c26) 2007; 39 (2025022614075408400_c14) 2021; 122 (2025022614075408400_c19) 2000; 78 (2025022614075408400_c43) 2016 (2025022614075408400_c25) 2019 (2025022614075408400_c3) 2020; 367 (2025022614075408400_c29) 2020; 32 (2025022614075408400_c38) 2002; 252 (2025022614075408400_c32) 2023; 90 2025022614075408400_c44 (2025022614075408400_c6) 2020; 5 (2025022614075408400_c35) 2019 (2025022614075408400_c11) 2024; 964 (2025022614075408400_c2) 2017; 27 (2025022614075408400_c1) 2020; 52 (2025022614075408400_c28) 2020; 882 (2025022614075408400_c12) 2018; 474 (2025022614075408400_c8) 2023; 64 (2025022614075408400_c10) 2024; 6 (2025022614075408400_c40) 2015 (2025022614075408400_c31) 2018 (2025022614075408400_c33) 2024; 36 (2025022614075408400_c7) 2022; 4 (2025022614075408400_c23) 2017 (2025022614075408400_c24) 2018; 19 (2025022614075408400_c15) 2022; 34 |
| References_xml | – volume: 27 start-page: 041102 year: 2017 ident: 2025022614075408400_c2 article-title: Reservoir observers: Model-free inference of unmeasured variables in chaotic systems publication-title: Chaos doi: 10.1063/1.4979665 – volume: 252 start-page: 527 year: 2002 ident: 2025022614075408400_c38 article-title: Proper orthogonal decomposition and its applications—Part I: Theory publication-title: J. Sound Vib. doi: 10.1006/jsvi.2001.4041 – volume-title: Physics of Buoyant Flows: From Instabilities to Turbulence year: 2018 ident: 2025022614075408400_c31 – volume: 964 start-page: 2 year: 2024 ident: 2025022614075408400_c11 article-title: Extending a physics-informed machine-learning network for superresolution studies of Rayleigh–Bénard convection publication-title: Astrophys. J. doi: 10.3847/1538-4357/ad1c55 – volume: 52 start-page: 477 year: 2020 ident: 2025022614075408400_c1 article-title: Machine learning for fluid mechanics publication-title: Annu. Rev. Fluid Mech. doi: 10.1146/annurev-fluid-010719-060214 – ident: 2025022614075408400_c44 – volume: 46 start-page: 102 year: 2023 ident: 2025022614075408400_c9 article-title: Topical issue on quantitative AI in complex fluids and complex flows: Challenges and benchmarks publication-title: Eur. Phys. J. E doi: 10.1140/epje/s10189-023-00356-w – volume: 6 start-page: 393 year: 2024 ident: 2025022614075408400_c10 article-title: Synthetic Lagrangian turbulence by generative diffusion models publication-title: Nat. Mach. Intell. doi: 10.1038/s42256-024-00810-0 – start-page: 113 year: 2019 ident: 2025022614075408400_c35 article-title: AutoAugment: Learning augmentation strategies from data – start-page: 448 year: 2015 ident: 2025022614075408400_c40 article-title: Batch normalization: Accelerating deep network training by reducing internal covariate shift publication-title: Proceedings of Machine Learning Research – volume: 35 start-page: 58 year: 2012 ident: 2025022614075408400_c30 article-title: New perspectives in turbulent Rayleigh-Bénard convection publication-title: Eur. Phys. J. E doi: 10.1140/epje/i2012-12058-1 – volume: 36 start-page: 035120 year: 2024 ident: 2025022614075408400_c33 article-title: Spatio-temporal dynamics of superstructures and vortices in turbulent Rayleigh–Bénard convection publication-title: Phys. Fluids doi: 10.1063/5.0191403 – volume: 3 start-page: 016104 year: 2025 ident: 2025022614075408400_c37 article-title: Large-scale-aware data augmentation for reduced-order models of high-dimensional flows publication-title: APL Mach. Learn. doi: 10.1063/5.0213700 – volume: 39 start-page: 447 year: 2007 ident: 2025022614075408400_c26 article-title: Turbulence transition in pipe flow publication-title: Annu. Rev. Fluid Mech. doi: 10.1146/annurev.fluid.39.050905.110308 – volume-title: Transactions on Machine Learning Research year: 2022 ident: 2025022614075408400_c46 article-title: Flipped classroom: Effective teaching for time series forecasting – volume: 34 start-page: 045106 year: 2022 ident: 2025022614075408400_c15 article-title: Direct data-driven forecast of local turbulent heat flux in Rayleigh–Bénard convection publication-title: Phys. Fluids doi: 10.1063/5.0087977 – volume: 367 start-page: 1026 year: 2020 ident: 2025022614075408400_c3 article-title: Hidden fluid mechanics: Learning velocity and pressure fields from flow visualizations publication-title: Science doi: 10.1126/science.aaw4741 – ident: 2025022614075408400_c36 – ident: 2025022614075408400_c34 – ident: 2025022614075408400_c16 – volume: 474 start-page: 20170844 year: 2018 ident: 2025022614075408400_c12 article-title: Data-driven forecasting of high-dimensional chaotic systems with long short-term memory networks publication-title: Proc. R. Soc. A doi: 10.1098/rspa.2017.0844 – volume: 122 start-page: 3780 year: 2021 ident: 2025022614075408400_c14 article-title: An autoencoder-based reduced-order model for eigenvalue problems with application to neutron diffusion publication-title: Int. J. Numer. Methods Eng. doi: 10.1002/nme.6681 – volume: 25 start-page: 539 year: 1993 ident: 2025022614075408400_c17 article-title: The proper orthogonal decomposition in the analysis of turbulent flows publication-title: Annu. Rev. Fluid Mech. doi: 10.1146/annurev.fluid.25.1.539 – volume: 882 start-page: A13 year: 2020 ident: 2025022614075408400_c28 article-title: Nonlinear mode decomposition with convolutional neural networks for fluid dynamics publication-title: J. Fluid Mech. doi: 10.1017/jfm.2019.822 – volume: 33 start-page: 121702 year: 2021 ident: 2025022614075408400_c21 article-title: Nonlinear proper orthogonal decomposition for convection-dominated flows publication-title: Phys. Fluids doi: 10.1063/5.0074310 – volume: 78 start-page: 808 year: 2000 ident: 2025022614075408400_c19 article-title: An introduction to the proper orthogonal decomposition publication-title: Curr. Sci. – volume: 19 start-page: 27 year: 2018 ident: 2025022614075408400_c24 article-title: Visual interpretability for deep learning: A survey publication-title: Front. Inf. Technol. Electron. Eng. doi: 10.1631/fitee.1700808 – ident: 2025022614075408400_c18 article-title: A tutorial on the proper orthogonal decomposition – start-page: 41 volume-title: ICML year: 2009 ident: 2025022614075408400_c45 article-title: Curriculum learning doi: 10.1145/1553374.1553380 – ident: 2025022614075408400_c20 – volume: 15 start-page: 1929 year: 2014 ident: 2025022614075408400_c41 article-title: Dropout: A simple way to prevent neural networks from overfitting publication-title: J. Mach. Learn. Res. doi: 10.5555/2627435.2670313 – volume: 90 start-page: 296 year: 2023 ident: 2025022614075408400_c32 article-title: The SCALEX facility – an apparatus for scaled fluid dynamical experiments publication-title: TM Tech. Mess. doi: 10.1515/teme-2022-0121 – start-page: 72 volume-title: Artificial Neural Networks and Machine Learning – ICANN 2016 year: 2016 ident: 2025022614075408400_c43 article-title: Analysis of dropout learning regarded as ensemble learning – volume: 27 start-page: 3365 year: 2014 ident: 2025022614075408400_c42 article-title: Learning with pseudo-ensembles publication-title: Adv. Neural Inf. Process. Syst. – volume: 5 start-page: 113506 year: 2020 ident: 2025022614075408400_c6 article-title: Reservoir computing model of two-dimensional turbulent convection publication-title: Phys. Rev. Fluids doi: 10.1103/physrevfluids.5.113506 – volume: 116 start-page: 8667 year: 2019 ident: 2025022614075408400_c5 article-title: Deep learning in turbulent convection networks publication-title: Proc. Natl. Acad. Sci. U. S. A. doi: 10.1073/pnas.1900358116 – volume: 11 start-page: 105211 year: 2021 ident: 2025022614075408400_c22 article-title: Feature extraction of fields of fluid dynamics data using sparse convolutional autoencoder publication-title: AIP Adv. doi: 10.1063/5.0065637 – start-page: 454 year: 2019 ident: 2025022614075408400_c25 article-title: Alime: Autoencoder based approach for local interpretability – volume: 388 start-page: 114181 year: 2022 ident: 2025022614075408400_c27 article-title: POD-DL-ROM: Enhancing deep learning-based reduced order models for nonlinear parametrized PDEs by proper orthogonal decomposition publication-title: Comput. Methods Appl. Mech. Eng. doi: 10.1016/j.cma.2021.114181 – volume: 30 start-page: 3 year: 2013 ident: 2025022614075408400_c39 article-title: Rectifier nonlinearities improve neural network acoustic models publication-title: Proc. ICML – volume: 120 start-page: e2213638120 year: 2023 ident: 2025022614075408400_c4 article-title: Lagrangian large eddy simulations via physics-informed machine learning publication-title: Proc. Natl. Acad. Sci. U. S. A. doi: 10.1073/pnas.2213638120 – volume: 4 start-page: 033176 year: 2022 ident: 2025022614075408400_c7 article-title: Hybrid quantum-classical reservoir computing of thermal convection flow publication-title: Phys. Rev. Res. doi: 10.1103/physrevresearch.4.033176 – volume: 33 start-page: 037106 year: 2021 ident: 2025022614075408400_c13 article-title: Reduced-order modeling of advection-dominated systems with recurrent neural networks and convolutional autoencoders publication-title: Phys. Fluids doi: 10.1063/5.0039986 – start-page: 1 year: 2017 ident: 2025022614075408400_c23 article-title: Interpretability of deep learning models: A survey of results – volume: 64 start-page: 191 year: 2023 ident: 2025022614075408400_c8 article-title: Data-driven estimation of scalar quantities from planar velocity measurements by deep learning applied to temperature in thermal convection publication-title: Exp. Fluids doi: 10.1007/s00348-023-03736-2 – volume: 32 start-page: 095110 year: 2020 ident: 2025022614075408400_c29 article-title: Convolutional neural network based hierarchical autoencoder for nonlinear mode decomposition of fluid field data publication-title: Phys. Fluids doi: 10.1063/5.0020721 |
| SSID | ssj0002990700 |
| Score | 2.291257 |
| Snippet | In recent years, data-driven deep learning models have gained significant importance in the analysis of turbulent dynamical systems. Within the context of... |
| SourceID | doaj crossref |
| SourceType | Open Website Index Database |
| StartPage | 016112 |
| Title | Slim multi-scale convolutional autoencoder-based reduced-order models for interpretable features of a complex dynamical system |
| URI | https://doaj.org/article/236f8e1f2be64001a1d9c1d7b38ae1f2 |
| Volume | 3 |
| WOSCitedRecordID | wos001493756700019&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D |
| hasFullText | 1 |
| inHoldings | 1 |
| isFullTextHit | |
| isPrint | |
| journalDatabaseRights | – providerCode: PRVAON databaseName: DOAJ Directory of Open Access Journals customDbUrl: eissn: 2770-9019 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0002990700 issn: 2770-9019 databaseCode: DOA dateStart: 20230101 isFulltext: true titleUrlDefault: https://www.doaj.org/ providerName: Directory of Open Access Journals – providerCode: PRVHPJ databaseName: ROAD: Directory of Open Access Scholarly Resources customDbUrl: eissn: 2770-9019 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0002990700 issn: 2770-9019 databaseCode: M~E dateStart: 20230101 isFulltext: true titleUrlDefault: https://road.issn.org providerName: ISSN International Centre |
| link | http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwrV1NS8QwEA2yePDit7h-EcRr3LTTpulRxcWDLoIKeytpJoEF3ZW2K572t5uk3WU9efHSQyhDeZnOTOjre4RcxTIVyA0yjYKzRHHFZGqBgbRCguZGQPhR-DEbjeR4nD-vWX15TlgrD9wCN4hBWGkiG5dGuHyLVIS5jjArQSq_7Ksvz_K1w5Svwb7IZpwvpYQEDNJr141c7xe_GtCaTn9oKMNdst1NgvSmfYI9smGm-2Rn6bJAu5fugCxe3icfNBD_WO0QNdQzxbuMcQHUvJl5NUo0FfM9CWnl5VgdMEFWkwavm5q64ZROVgzD0oWxJoh61nRmqaKBW26-KbYW9S5wK_J8SN6G9693D6xzTWA6jnnDkshyN1WAFUnEc8QsSdzxl9sk1RwBUJlEmKhUQmYGhAYNYB00mU3R-GYOR6Q3nU3NMaGYypJrP2FxSNDb92WyjHOFApVALfrkcgll8dmKYxTho7aAIi06vPvk1oO8usHrWYcFt8tFt8vFX7t88h9BTslW7N17A4PsjPSaam7Oyab-aiZ1dRESyF2fFvc_KTLOmw |
| linkProvider | Directory of Open Access Journals |
| openUrl | ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=Slim+multi-scale+convolutional+autoencoder-based+reduced-order+models+for+interpretable+features+of+a+complex+dynamical+system&rft.jtitle=APL+machine+learning&rft.au=Philipp+Teutsch&rft.au=Philipp+Pfeffer&rft.au=Mohammad+Sharifi+Ghazijahani&rft.au=Christian+Cierpka&rft.date=2025-03-01&rft.pub=AIP+Publishing+LLC&rft.eissn=2770-9019&rft.volume=3&rft.issue=1&rft.spage=016112&rft.epage=016112-15&rft_id=info:doi/10.1063%2F5.0244416&rft.externalDBID=DOA&rft.externalDocID=oai_doaj_org_article_236f8e1f2be64001a1d9c1d7b38ae1f2 |
| thumbnail_l | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=2770-9019&client=summon |
| thumbnail_m | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=2770-9019&client=summon |
| thumbnail_s | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=2770-9019&client=summon |