{"title":"复杂分子电子碰撞电离截面研究","authors":"Zhan-Bin Chen","doi":"10.1002/qua.27422","DOIUrl":null,"url":null,"abstract":"<p>An accurate and computationally efficient determination of the cross sections for electron collision ionization of molecules has various applications, such as plasma physics and atmospheric science. In the case of large molecules, ab initio calculations are often difficult and time-consuming. Here, we develop a feed forward neural network to predict the electron impact ionization cross sections of complex molecules. The training (predicting) set in the method consists of a series of theoretical ionization cross sections for small (large) molecules obtained from the combined model, which integrates the Binary-Encounter-Bethe and Deutsch-Märk models. Several complex systems or targets involving electron collision ionization are evaluated, including molecules such as CH<span></span><math>\n <semantics>\n <mrow>\n <msub>\n <mrow></mrow>\n <mrow>\n <mn>4</mn>\n </mrow>\n </msub>\n </mrow>\n <annotation>$$ {}_4 $$</annotation>\n </semantics></math>, C<span></span><math>\n <semantics>\n <mrow>\n <msub>\n <mrow></mrow>\n <mrow>\n <mn>3</mn>\n </mrow>\n </msub>\n </mrow>\n <annotation>$$ {}_3 $$</annotation>\n </semantics></math>H<span></span><math>\n <semantics>\n <mrow>\n <msub>\n <mrow></mrow>\n <mrow>\n <mn>8</mn>\n </mrow>\n </msub>\n </mrow>\n <annotation>$$ {}_8 $$</annotation>\n </semantics></math>, C<span></span><math>\n <semantics>\n <mrow>\n <msub>\n <mrow></mrow>\n <mrow>\n <mn>5</mn>\n </mrow>\n </msub>\n </mrow>\n <annotation>$$ {}_5 $$</annotation>\n </semantics></math>H<span></span><math>\n <semantics>\n <mrow>\n <msub>\n <mrow></mrow>\n <mrow>\n <mn>8</mn>\n </mrow>\n </msub>\n </mrow>\n <annotation>$$ {}_8 $$</annotation>\n </semantics></math>, C<span></span><math>\n <semantics>\n <mrow>\n <msub>\n <mrow></mrow>\n <mrow>\n <mn>6</mn>\n </mrow>\n </msub>\n </mrow>\n <annotation>$$ {}_6 $$</annotation>\n </semantics></math>H<span></span><math>\n <semantics>\n <mrow>\n <msub>\n <mrow></mrow>\n <mrow>\n <mn>10</mn>\n </mrow>\n </msub>\n </mrow>\n <annotation>$$ {}_{10} $$</annotation>\n </semantics></math>, C<span></span><math>\n <semantics>\n <mrow>\n <msub>\n <mrow></mrow>\n <mrow>\n <mn>6</mn>\n </mrow>\n </msub>\n </mrow>\n <annotation>$$ {}_6 $$</annotation>\n </semantics></math>, C<span></span><math>\n <semantics>\n <mrow>\n <msub>\n <mrow></mrow>\n <mrow>\n <mn>2</mn>\n </mrow>\n </msub>\n </mrow>\n <annotation>$$ {}_2 $$</annotation>\n </semantics></math>H<span></span><math>\n <semantics>\n <mrow>\n <msub>\n <mrow></mrow>\n <mrow>\n <mn>6</mn>\n </mrow>\n </msub>\n </mrow>\n <annotation>$$ {}_6 $$</annotation>\n </semantics></math>O, and C<span></span><math>\n <semantics>\n <mrow>\n <msub>\n <mrow></mrow>\n <mrow>\n <mn>6</mn>\n </mrow>\n </msub>\n </mrow>\n <annotation>$$ {}_6 $$</annotation>\n </semantics></math>H<span></span><math>\n <semantics>\n <mrow>\n <msub>\n <mrow></mrow>\n <mrow>\n <mn>6</mn>\n </mrow>\n </msub>\n </mrow>\n <annotation>$$ {}_6 $$</annotation>\n </semantics></math>O. The root mean square errors of the trained and predicted cross sections by the <span></span><math>\n <semantics>\n <mrow>\n <mn>2</mn>\n <mo>×</mo>\n <mn>3</mn>\n <mo>×</mo>\n <mn>3</mn>\n <mo>×</mo>\n <mn>1</mn>\n </mrow>\n <annotation>$$ 2\\times 3\\times 3\\times 1 $$</annotation>\n </semantics></math> neural network (compared to the values from the combined model) are found to be approximately .0086 and .0930 (in 10<span></span><math>\n <semantics>\n <mrow>\n <msup>\n <mrow></mrow>\n <mrow>\n <mo>−</mo>\n <mn>20</mn>\n </mrow>\n </msup>\n </mrow>\n <annotation>$$ {}^{-20} $$</annotation>\n </semantics></math> cm<span></span><math>\n <semantics>\n <mrow>\n <msup>\n <mrow></mrow>\n <mrow>\n <mn>2</mn>\n </mrow>\n </msup>\n </mrow>\n <annotation>$$ {}^2 $$</annotation>\n </semantics></math>), respectively, (using the C<span></span><math>\n <semantics>\n <mrow>\n <msub>\n <mrow></mrow>\n <mrow>\n <mn>2</mn>\n </mrow>\n </msub>\n </mrow>\n <annotation>$$ {}_2 $$</annotation>\n </semantics></math>H<span></span><math>\n <semantics>\n <mrow>\n <msub>\n <mrow></mrow>\n <mrow>\n <mn>6</mn>\n </mrow>\n </msub>\n </mrow>\n <annotation>$$ {}_6 $$</annotation>\n </semantics></math>O molecule as an example), indicating our results are very high accuracy. The excellent agreement between the predicted values and the actual values indicates that the neural network is a practical and powerful tool for determining the electron collision ionization cross sections of complex molecules and can provide valuable insights into the dynamics process. Apart from its fundamental importance, this study has far-reaching implications for gas discharge, low-temperature plasmas, and fusion edge plasmas and so forth.</p>","PeriodicalId":182,"journal":{"name":"International Journal of Quantum Chemistry","volume":null,"pages":null},"PeriodicalIF":2.3000,"publicationDate":"2024-05-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Investigation of the cross sections for electron collision ionization of complex molecules\",\"authors\":\"Zhan-Bin Chen\",\"doi\":\"10.1002/qua.27422\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p>An accurate and computationally efficient determination of the cross sections for electron collision ionization of molecules has various applications, such as plasma physics and atmospheric science. In the case of large molecules, ab initio calculations are often difficult and time-consuming. Here, we develop a feed forward neural network to predict the electron impact ionization cross sections of complex molecules. The training (predicting) set in the method consists of a series of theoretical ionization cross sections for small (large) molecules obtained from the combined model, which integrates the Binary-Encounter-Bethe and Deutsch-Märk models. Several complex systems or targets involving electron collision ionization are evaluated, including molecules such as CH<span></span><math>\\n <semantics>\\n <mrow>\\n <msub>\\n <mrow></mrow>\\n <mrow>\\n <mn>4</mn>\\n </mrow>\\n </msub>\\n </mrow>\\n <annotation>$$ {}_4 $$</annotation>\\n </semantics></math>, C<span></span><math>\\n <semantics>\\n <mrow>\\n <msub>\\n <mrow></mrow>\\n <mrow>\\n <mn>3</mn>\\n </mrow>\\n </msub>\\n </mrow>\\n <annotation>$$ {}_3 $$</annotation>\\n </semantics></math>H<span></span><math>\\n <semantics>\\n <mrow>\\n <msub>\\n <mrow></mrow>\\n <mrow>\\n <mn>8</mn>\\n </mrow>\\n </msub>\\n </mrow>\\n <annotation>$$ {}_8 $$</annotation>\\n </semantics></math>, C<span></span><math>\\n <semantics>\\n <mrow>\\n <msub>\\n <mrow></mrow>\\n <mrow>\\n <mn>5</mn>\\n </mrow>\\n </msub>\\n </mrow>\\n <annotation>$$ {}_5 $$</annotation>\\n </semantics></math>H<span></span><math>\\n <semantics>\\n <mrow>\\n <msub>\\n <mrow></mrow>\\n <mrow>\\n <mn>8</mn>\\n </mrow>\\n </msub>\\n </mrow>\\n <annotation>$$ {}_8 $$</annotation>\\n </semantics></math>, C<span></span><math>\\n <semantics>\\n <mrow>\\n <msub>\\n <mrow></mrow>\\n <mrow>\\n <mn>6</mn>\\n </mrow>\\n </msub>\\n </mrow>\\n <annotation>$$ {}_6 $$</annotation>\\n </semantics></math>H<span></span><math>\\n <semantics>\\n <mrow>\\n <msub>\\n <mrow></mrow>\\n <mrow>\\n <mn>10</mn>\\n </mrow>\\n </msub>\\n </mrow>\\n <annotation>$$ {}_{10} $$</annotation>\\n </semantics></math>, C<span></span><math>\\n <semantics>\\n <mrow>\\n <msub>\\n <mrow></mrow>\\n <mrow>\\n <mn>6</mn>\\n </mrow>\\n </msub>\\n </mrow>\\n <annotation>$$ {}_6 $$</annotation>\\n </semantics></math>, C<span></span><math>\\n <semantics>\\n <mrow>\\n <msub>\\n <mrow></mrow>\\n <mrow>\\n <mn>2</mn>\\n </mrow>\\n </msub>\\n </mrow>\\n <annotation>$$ {}_2 $$</annotation>\\n </semantics></math>H<span></span><math>\\n <semantics>\\n <mrow>\\n <msub>\\n <mrow></mrow>\\n <mrow>\\n <mn>6</mn>\\n </mrow>\\n </msub>\\n </mrow>\\n <annotation>$$ {}_6 $$</annotation>\\n </semantics></math>O, and C<span></span><math>\\n <semantics>\\n <mrow>\\n <msub>\\n <mrow></mrow>\\n <mrow>\\n <mn>6</mn>\\n </mrow>\\n </msub>\\n </mrow>\\n <annotation>$$ {}_6 $$</annotation>\\n </semantics></math>H<span></span><math>\\n <semantics>\\n <mrow>\\n <msub>\\n <mrow></mrow>\\n <mrow>\\n <mn>6</mn>\\n </mrow>\\n </msub>\\n </mrow>\\n <annotation>$$ {}_6 $$</annotation>\\n </semantics></math>O. The root mean square errors of the trained and predicted cross sections by the <span></span><math>\\n <semantics>\\n <mrow>\\n <mn>2</mn>\\n <mo>×</mo>\\n <mn>3</mn>\\n <mo>×</mo>\\n <mn>3</mn>\\n <mo>×</mo>\\n <mn>1</mn>\\n </mrow>\\n <annotation>$$ 2\\\\times 3\\\\times 3\\\\times 1 $$</annotation>\\n </semantics></math> neural network (compared to the values from the combined model) are found to be approximately .0086 and .0930 (in 10<span></span><math>\\n <semantics>\\n <mrow>\\n <msup>\\n <mrow></mrow>\\n <mrow>\\n <mo>−</mo>\\n <mn>20</mn>\\n </mrow>\\n </msup>\\n </mrow>\\n <annotation>$$ {}^{-20} $$</annotation>\\n </semantics></math> cm<span></span><math>\\n <semantics>\\n <mrow>\\n <msup>\\n <mrow></mrow>\\n <mrow>\\n <mn>2</mn>\\n </mrow>\\n </msup>\\n </mrow>\\n <annotation>$$ {}^2 $$</annotation>\\n </semantics></math>), respectively, (using the C<span></span><math>\\n <semantics>\\n <mrow>\\n <msub>\\n <mrow></mrow>\\n <mrow>\\n <mn>2</mn>\\n </mrow>\\n </msub>\\n </mrow>\\n <annotation>$$ {}_2 $$</annotation>\\n </semantics></math>H<span></span><math>\\n <semantics>\\n <mrow>\\n <msub>\\n <mrow></mrow>\\n <mrow>\\n <mn>6</mn>\\n </mrow>\\n </msub>\\n </mrow>\\n <annotation>$$ {}_6 $$</annotation>\\n </semantics></math>O molecule as an example), indicating our results are very high accuracy. The excellent agreement between the predicted values and the actual values indicates that the neural network is a practical and powerful tool for determining the electron collision ionization cross sections of complex molecules and can provide valuable insights into the dynamics process. Apart from its fundamental importance, this study has far-reaching implications for gas discharge, low-temperature plasmas, and fusion edge plasmas and so forth.</p>\",\"PeriodicalId\":182,\"journal\":{\"name\":\"International Journal of Quantum Chemistry\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":2.3000,\"publicationDate\":\"2024-05-21\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"International Journal of Quantum Chemistry\",\"FirstCategoryId\":\"92\",\"ListUrlMain\":\"https://onlinelibrary.wiley.com/doi/10.1002/qua.27422\",\"RegionNum\":3,\"RegionCategory\":\"化学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"CHEMISTRY, PHYSICAL\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"International Journal of Quantum Chemistry","FirstCategoryId":"92","ListUrlMain":"https://onlinelibrary.wiley.com/doi/10.1002/qua.27422","RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"CHEMISTRY, PHYSICAL","Score":null,"Total":0}
Investigation of the cross sections for electron collision ionization of complex molecules
An accurate and computationally efficient determination of the cross sections for electron collision ionization of molecules has various applications, such as plasma physics and atmospheric science. In the case of large molecules, ab initio calculations are often difficult and time-consuming. Here, we develop a feed forward neural network to predict the electron impact ionization cross sections of complex molecules. The training (predicting) set in the method consists of a series of theoretical ionization cross sections for small (large) molecules obtained from the combined model, which integrates the Binary-Encounter-Bethe and Deutsch-Märk models. Several complex systems or targets involving electron collision ionization are evaluated, including molecules such as CH, CH, CH, CH, C, CHO, and CHO. The root mean square errors of the trained and predicted cross sections by the neural network (compared to the values from the combined model) are found to be approximately .0086 and .0930 (in 10 cm), respectively, (using the CHO molecule as an example), indicating our results are very high accuracy. The excellent agreement between the predicted values and the actual values indicates that the neural network is a practical and powerful tool for determining the electron collision ionization cross sections of complex molecules and can provide valuable insights into the dynamics process. Apart from its fundamental importance, this study has far-reaching implications for gas discharge, low-temperature plasmas, and fusion edge plasmas and so forth.
期刊介绍:
Since its first formulation quantum chemistry has provided the conceptual and terminological framework necessary to understand atoms, molecules and the condensed matter. Over the past decades synergistic advances in the methodological developments, software and hardware have transformed quantum chemistry in a truly interdisciplinary science that has expanded beyond its traditional core of molecular sciences to fields as diverse as chemistry and catalysis, biophysics, nanotechnology and material science.