Anomaly Detection Algorithm for Searching Selective Catalyst Differentiating Linear and Cyclic Alkanes in Oxidation

Anomaly Detection Algorithm for Searching Selective Catalyst Differentiating Linear and Cyclic Alkanes in Oxidation | Authorea try { document.documentElement.classList.add('js'); } catch (e) { } var _gaq = _gaq || []; _gaq.push(['_setAccount', 'G-8VDV14Y67G']); _gaq.push(['_trackPageview']); (function() { var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true; ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js'; var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s); })(); Skip to main content Preprints Collections Wiley Open Research IET Open Research Ecological Society of Japan All Collections About About Authorea FAQs Contact Us Quick Search anywhere Search for preprint articles, keywords, etc. Search Search ADVANCED SEARCH SCROLL Chinese Journal of Chemistry This is a preprint and has not been peer reviewed. Data may be preliminary. 9 January 2025 V1 Latest version Share on Anomaly Detection Algorithm for Searching Selective Catalyst Differentiating Linear and Cyclic Alkanes in Oxidation Authors : 稼兴 刘 0000-0001-8985-3664 , Pengkun Su , Bingling Dai , Da Zhou , and Cheng Wang 0000-0002-7906-8061 [email protected] Authors Info & Affiliations https://doi.org/10.22541/au.173641813.38940605/v1 342 views 220 downloads Contents Abstract Information & Authors Metrics & Citations View Options References Figures Tables Media Share Abstract Selective catalysis, particularly when differentiating substrates with similar reactivities in mixture, is a significant challenge. In this study, anomaly detection algorithms---tools traditionally used for identifying outliers in data cleaning---are applied to catalyst screening. We focus on developing catalytic methods to selectively oxidize cyclic alkanes over linear alkanes in mixtures such as naphtha. By inserting cyclohexane oxidation data one by one into a database of n-hexane oxidization, we used several anomaly detection algorithms to evaluate whether the inserted cyclohexane oxidation data could be considered anomalous. Conditions identified as anomalies imply that they are likely not suitable for n-hexane oxidization. However, these anomalies come from conditions for cyclohexane oxidation. As a result, they are promising conditions for selective oxidation of cyclohexane while leaving n-hexane unaltered. These anomalies were thus further investigated, leading to the discovery of a specific catalytic approach that selectively oxidizes cyclohexane. This application of anomaly detection offers a novel method to search for selective catalyst for chemical reactions involving mixed substrates. Anomaly Detection Algorithm for Searching Selective Catalyst Differentiating Linear and Cyclic Alkanes in Oxidation Jiaxing Liu [a,b] , Pengkun Su [b] , Bingling Dai [b] , Da Zhou [c]，* , Cheng Wang [b]，* [a] J. Liu Institute of Artificial Intelligence, Xiamen University, Xiamen, Fujian 361005, P. R. China [b] Dr. P. Su, B. Dai, Prof. C. Wang State Key Laboratory for Physical Chemistry of Solid Surfaces, iChEM, College of Chemistry and Chemical Engineering, Xiamen University, Xiamen, Fujian 361005, P. R. China E-mail: [email protected] [c] Prof. D. Zhou School of Mathematical Sciences/National Institute for Data Science in Health and Medicine, Xiamen University, Xiamen, Fujian 361005, P. R. China E-mail: [email protected] Abstract: Selective catalysis, particularly when differentiating substrates with similar reactivities in mixture, is a significant challenge. In this study, anomaly detection algorithms—tools traditionally used for identifying outliers in data cleaning—are applied to catalyst screening. We focus on developing catalytic methods to selectively oxidize cyclic alkanes over linear alkanes in mixtures such as naphtha. By inserting cyclohexane oxidation data one by one into a database of n-hexane oxidization, we used several anomaly detection algorithms to evaluate whether the inserted cyclohexane oxidation data could be considered anomalous. Conditions identified as anomalies imply that they are likely not suitable for n-hexane oxidization. However, these anomalies come from conditions for cyclohexane oxidation. As a result, they are promising conditions for selective oxidation of cyclohexane while leaving n-hexane unaltered. These anomalies were thus further investigated, leading to the discovery of a specific catalytic approach that selectively oxidizes cyclohexane. This application of anomaly detection offers a novel method to search for selective catalyst for chemical reactions involving mixed substrates. Selectivity in catalytic conversion is crucial for optimizing yield, reducing waste, and improving the efficiency of chemical processes. Selective oxidation of alkanes, for instance, plays a vital role in transforming light alkanes into high-value chemical products [1] . Although high selectivity in single-substrate alkane transformations were achieved using various catalysts, [2] the selective oxidation of specific alkanes in the presence of multiple alkane substrates is less explored. [3] This challenge is particularly pronounced in industrial settings where separating complex alkane mixtures, such as naphtha, is economically and technically demanding. [4] A selective catalytic oxidation method to only convert cycloalkanes would aid in their upgrading and separation from the mixture, while the linear alkanes left can be cracked to form short olefins. There are multiple secondary sp 3 C‒H bonds both in cycloalkanes and linear alkanes, but the bond dissociation energies of them can be slightly different. In cycloalkanes, the bond dissociation energy (BDE) of C‒H bond in cyclohexane is 99.5-100.0 kcal/mol, [5] while in linear alkanes, the BDE of sp 3 C‒H in n-hexane are 101.1 kcal/mol for primary C‒H, 98.6 kcal/mol for secondary C‒H at 2-C and 96.5 kcal/mol for secondary C‒H at 3-C, [5a, 6] possibly due to difference in hyperconjugation. The different steric environments of the sp 3 C‒H bonds in these two types of alkanes also likely led to some reactivity difference. For example, a polyoxometalate-supported Uranyl catalyzes fluorination of cycloalkanes but not linear alkanes [7] . However, we failed to find reports of catalysts capable of selectively oxidizing cycloalkanes over linear alkanes to produce oxygenates. Even reports testing one catalyst on both cycloalkanes and linear alkanes for generating oxygenates are scarce. [8] We plan to use data science [9] to dig out information from the rich literature on alkane oxidations to find a catalyst that is simple, robust, and cost-effective for selectively oxidizing cycloalkanes. Ideally, a comprehensive database containing data on catalysts tested for converting representative mixtures like cyclohexane plus n-hexane would facilitate the search for highly selective catalysts. However, such a database is not publicly available. An alternative approach would be to utilize a database of catalysts tested separately on cyclohexane and n-hexane oxidation, though data are limited in this area as well. Most available data from literature searches focus either on cyclohexane or n-hexane but rarely on both. Given these constraints, we propose using an anomaly detection approach [10] : adding each catalytic condition known to oxidize cyclohexane to the n-hexane oxidation database. We then assess whether the cyclohexane oxidation data are identified as outliers in the n-hexane oxidation dataset. If so, the catalytic condition likely does not facilitate the oxidation of n-hexane, thus indicating selectivity for cyclohexane over n-hexane. Figure 1. Illustration of anomaly detection algorithms in identifying selective catalysts for the oxidation of cyclic alkanes (cyclohexane, CyH) versus linear alkanes ( n -hexane, Hex). The table on the left lists the catalytic conditions tested for both n -hexane and cyclohexane oxidations. A data point from CyH oxidation was added to Hex data points for anomaly test to identify promising for selective candidate. We have collected over 535 data points for cyclohexane oxidation and 180 for n -hexane, primarily from journal articles. As illustrated in Figure 2, the active metals in the catalysts predominantly include Fe, Mn, Cu, and Co, while hydrogen peroxide and oxygen are the most frequently used oxidants. Acetonitrile, water, and dichloromethane are the main solvents, chosen likely for their resistance to these oxidants. The reaction temperatures vary widely. In this highly overlapping chemical space, discerning which conditions selectively oxidize cyclohexane through human intuition alone is exceedingly challenging. Figure 2. Pie charts comparing the metals, oxidants, solvents, and reaction temperatures for cyclohexane and n-hexane oxidation data. For further analysis, data encoding was performed using a one-hot encoding strategy, where the presence of a specific active metal, oxidant, or solvent in the reactions was marked as ’1’, and its absence as ’0’ (Figure 3b). Reaction temperatures were also recorded, providing a comprehensive profile of each experimental setup consisting of 30 features. We attempted a classification model using Decision Tree (DT), Random Forest (RF), K-Nearest Neighbors (KNN), and Logistic Regression (LR) to try to determine if a reaction condition comes from reported n -hexane oxidation or cyclohexane oxidation. We must note that this task itself has a serious limitation, as many of the reported conditions can actually oxidize both n -hexane and cyclohexane but the corresponding paper only reports result on one of the substrates. The data is imbalanced due to the different number of collected n-hexane and cyclohexane oxidation data. As expected, we had poor precision, recall, and F1 score for all the models (Table S2). Traditional classification models cannot solve the challenge here. We then opt to the anomaly detection algorithm to identify potentially selective conditions. We added each data point from the cyclohexane oxidation experiments one by one into the n-hexane oxidation dataset for anomaly detection using the Z-score, Isolation Forest (IOF) [11] , and K-Nearest Neighbors (KNN), Local Outlier Factor (LOF) [12] methods. The entire process is illustrated in Figure 3. The Z-score algorithm measures the standard deviation of a data point with respect to the mean of the dataset. Data points with a Z-value greater than 3 or less than -3 are considered outliers, indicating that they are unusually far from the average values. Both IOF and KNN are distance-based methods. IOF detects outliers by calculating the average path length needed to isolate a point, with shorter paths indicating a higher probability of being an outlier. KNN evaluates the distance of points from their nearest neighbors, identifying those far from the center as more likely to be outliers. LOF compares the data density around a data point to the density around its neighbors to determine how isolated the point is with respect to the surrounding neighborhood to determine if it is an outlier. We analyzed the results of the anomaly detection, as shown in Figure 4a. The number of “Active Metal” anomalies indicates the number of data points identified as anomalies due to using metals other than Fe, Mn, Cu, Co, and Ti, which are the main active metals used in n-hexane oxidation. The number of “Oxi.” anomalies represent the data points that use oxidants uncommon or absent in the n-hexane oxidation data. The number of “Sol” anomalies refers to data points that use mixed solvents or solvents not present in the n -hexane oxidation data. The number of “Temp” anomalies shows the data points where the reaction temperature is below 5°C or above 80°C. Out of 535 cyclohexane oxidation data points, 217 conditions were identified as anomalies by at least one of the anomaly detection algorithms. Among these, 10 conditions were selected by three of the four algorithms, and another 52 conditions were selected by two of the four algorithms. The Z-score method identified the fewest anomalies, with a total of 17 conditions. Of these 17 conditions, 8 used metal centers uncommon in n-hexane oxidation data, such as V, Ru, Au, and Os, and the other 9 conditions included bimetallic centers, mixed solvent systems, higher or lower reaction temperatures, and longer reaction times. In contrast, the IOF method identified the highest number of anomalies, including 93 conditions. Among these, 40 conditions used metal centers such as Zn, V, W, Mo, Ru, and Os, which are uncommon in n -hexane oxidation data. Seven conditions included bimetallic centers, and 54 conditions involved mixed solvents and reaction temperatures uncommon in n-hexane oxidation data. As for the conditions identified by three algorithms, seven of them have metal centers that are uncommon in the n-hexane oxidation data, such as Os, Au, Ru, and V, or use of more than two metals, and six of them involve reactions conducted in uncommon solvents or mixtures of multiple solvents. Figure 4b lists some of the anomaly detection results (see the Figure S1 for the Ligand information). Figure 3. (a) The oxidation data for cyclohexane and n -hexane from published literature or patents. (b) The data was characterized using one-hot encoding. (c) Performing data anomaly detection using four different algorithms. We need to note that the anomaly detection algorithms also identified 26 n-hexane oxidation conditions to be anomaly among other n-hexane oxidation conditions, with three of them detected by two of the four algorithms and the remaining ones detected by one algorithm (Table S3). These conditions mainly contain uncommon metal centers such as Cr, Ru, or special solvent / reaction temperature. Figure 4. (a) Statistical results of anomaly detection on four algorithms. (b) Anomaly detection scores. We then plan to experimentally test the candidate catalyst / operational condition obtained by the anomaly detection method. Before that, we need to check whether the identified system can meet the requirement on chemical robustness, simplicity, and cost-effectiveness. As we have a lot of candidates for the test, we used a stringent standard for selection. Firstly, we exclude entries containing expensive elements such as Os, Au, and Ru. Second, we removed catalysts of coordination complexes containing costly organic ligands. Additionally, entries with reaction temperatures significantly lower than room temperature are excluded due to their complexity of large-scale operations. After applying these criteria, we focus on eight catalysts, mainly metal oxides or metal salts, with simple reaction conditions (Figure 4b and Figure S2). However, under three of the eight conditions, only trace amounts of oxidation products were produced, and another three of these conditions was not selective for cyclohexane oxidation over n-hexane oxidation. Nevertheless, we found that using simple WO3 as the catalyst and H 2 O 2 as the oxidant at 100°C resulted in quite high selectivity. In a mixed substrate test, the catalyst oxidized cyclohexane with 100% selectivity without oxidizing n-hexane. We summarized the reaction results of the eight conditions in Table 1 and presented the gas chromatography (GC) trace and proton nuclear magnetic resonance ( 1 H-NMR) spectra for these eight conditions in Figure 5. Co(Ac) 2 as a catalyst also showed high selectivity and activity, but acetaldehyde was required as a consumable cocatalyst, decreasing its economic feasibility in operation. Further extending the test of the WO3 catalyst to mixtures of C5-C8 alkanes and cycloalkanes, all showed conversion of cycloalkanes without conversion of n-alkanes, as summarized in Table S4. Cyclopentane had the highest activity of 156 μmol/g cat /h. Table 1. Summary of reaction results Reaction Conditions Substrate Selectivity (%) Activity (μmol/g cat /h) n -alkane Cyclo-alkane 1 200mg WO 3 + 0.2mL CyH + 0.2mL n-Hex + 3mL MeCN + 0. 5mL H 2 O 2 + 100°C + 12h / > 99 2.4 2 100mg Co(Ac) 2 + 1.5mL CyH + 1.5mL n-Hex + 1.5mL AcOH + 0.15mL Acetaldehyde + 100°C + 2h / > 99 2.8 3 5mg Mn(Ac) 2 + 0.2mL CyH + 0.2mL n-Hex + 1mL H 2 O + 1mL MeCN + 10mg C 6 H 5 IO + r.t. + 2h >99 / 260 4 5mg Mn(Ac) 2 + 0.2mL CyH + 0.2mL n-Hex + 1mL DCM + 1mL MeCN + 10mg C 6 H 5 IO + r.t. + 2h > 99 / 278 5 100 mg ZSM-5 + 0.5mL CyH + 0.5mL n-Hex + 3mL Acetone + 1mL H 2 O 2 + 100°C + 2h 86 14 148 6 Fe 2 (SO 4 ) 3 + HNO 3 + MeCN + O 2 + 25°C + 2h / / / 7 5mg TPP-Mn + 0.2mL CyH + 0.2mL n-Hex + 1mL MeCN + 1mL DCM + 10mg C 6 H 5 IO + r.t. + 2h / / / 8 5mg TPP-Fe + 0.2mL CyH + 0.2mL n-Hex + 1mL MeCN + 1mL DCM + 10mg C 6 H 5 IO + r.t. + 2h / / / H 2 O 2 was widely used as an oxidant for oxidation reactions. One common pathway of H 2 O 2 -involved oxidation is the Fenton or Fenton-like process, [13] where single electron transfer between H2O2 and a redox active center such as Fe 3+ /Fe 2+ generates •OH and •OOH radicals. On the other hand, non-redox d0 transition-metal centers in transition metal-doped molecular sieves (including W-doped ones) [14] , transition metal-doped polyoxometalates [15] , metal oxides (e.g. ZrO 2 , Nb 2 O 5 , Ta 2 O 5 , WO 3 ), could also activate H 2 O 2 for oxidation reaction. The key step may involve an M-OOH (hydroperoxide) intermediate [16] or M-(ƞ 2 -O 2 ) 2- (peroxide). [14c, 16c] Generation of M-(O 2 ) - (superperoxide) [17] was reported, likely due to single electron transfer between M-OOH and H 2 O 2 . The fact that WO3 is selective for cycloalkane vs. linear alkanes suggest that the oxygen species on non-redox Lewis acid center may be selective in such a reaction. Following this rationale, we further tested a bunch of other oxides as Lewis acid catalysts including MgO, CaO, B 2 O 3 , Al 2 O 3 , Sc 2 O 3 , SnO 2 , Ga 2 O 3 , MoO 3 , Gd 2 O 3 , ZrO 2 , among which Al 2 O 3 , Sc 2 O 3 , Ga 2 O 3 , MoO 3 also showed some conversion and decent selectivity for cyclohexane over n -hexane (Table S5 and Figure S3). In this study, we used anomaly detection methods to explore reaction conditions for selective oxidation of cycloalkanes but not linear alkanes. Despite the lack of experimental data in selectivity test from the reported data and a high degree of overlap in reaction conditions suitable for oxidizing the different substrates, a selective reaction condition was identified using the anomaly detection algorithms and then experimentally verified. This demonstrates the potential of anomaly detection methods in exploring chemical reactions and provides a different perspective for algorithms in chemical reaction exploration. Acknowledgements The authors acknowledge funding support from the National Key R&D Program of China (2021YFA1502500), the National Natural Science Foundation of China (22125502, 22071207, 22121001) and the Fundamental Research Funds for the Central Universities (20720220011, 20720240151). Large language models were used for language polishing. All data and codes are stored in github repository at https://github.com/Wang-Group/Anomaly-Detection-for-Selective-Catalysis. Keywords: Anomaly detection •Selectivity • Catalysis • Alkane Oxidation References: [1] a)T. Dalton, T. Faber, F. Glorius, ACS Central Science 2021 , 7 , 245-261; b)R. G. Bergman, Nature 2007 , 446 , 391-393; c)D. L. Golden, S. E. Suh, S. S. Stahl, Nature Reviews Chemistry 2022 , 6 , 405-427; d)N. F. Dummer, D. J. Willock, Q. He, M. J. Howard, R. J. Lewis, G. D. Qi, S. H. Taylor, J. Xu, D. Bethell, C. J. Kiely, G. J. Hutchings, Chemical Reviews 2023 , 123 , 6359-6411.[2] a)S. Cheng, Q. Li, X. Cheng, Y.-M. Lin, L. Gong, Chinese Journal of Chemistry 2022 , 40 , 2825-2837; b)X. Tang, X. Jia, Z. Huang, Chemical Science 2018 , 9 , 288-299; c)A. Pulcinella, P. C. Tiwari, A. Luridiana, K. Yamazaki, D. Mazzarella, A. K. Sadhoe, A. I. Alfano, E. H. Tiekink, T. A. Hamlin, T. Noel, Angewandte Chemie (International ed. in English) 2024 , e202413846; d)Y. X. Xu, C. Wang, X. Y. Li, L. Q. Xiong, T. Y. Zhang, L. Q. Zhang, Q. H. Zhang, L. Gu, Y. Lan, J. W. Tang, Nature Sustainability 2024 ; e)W. Begum, M. Chauhan, R. Kalita, P. Gupta, N. Akhtar, N. Antil, R. Newar, K. Manna, Acs Catalysis 2024 , 14 , 10427-10436.[3] a)A. Corma, M. A. Camblor, P. Esteve, A. Martinez, J. Perezpariente, Journal of Catalysis 1994 , 145 , 151-158; b)D. P. Solowey, M. V. Mane, T. Kurogi, P. J. Carroll, B. C. Manor, M.-H. Baik, D. J. Mindiola, Nature Chemistry 2017 , 9 , 1126-1132.[4] F. Xie, L. Yu, H. Wang, J. Li, Angewandte Chemie-International Edition 2023 , 62 .[5] a)Z. Tian, A. Fattahi, L. Lis, S. R. Kass, Journal of the American Chemical Society 2006 , 128 , 17087-17092; b)S. Rayne, K. Forest, Nature Precedings 2010 .[6] L. Capaldo, D. Ravelli, M. Fagnoni, Chemical Reviews 2022 , 122 , 1875-1924.[7] C.-L. Chen, H.-Y. Wang, Z.-Z. Weng, L.-S. Long, L.-S. Zheng, X.-J. Kong, Inorganic Chemistry 2023 , 62 , 17041-17045.[8] a)I. Garcia-Bosch, M. A. Siegler, Angewandte Chemie International Edition 2016 , 55 , 12873-12876; b)G. M. Ucoski, V. H. A. Pinto, G. DeFreitas-Silva, J. S. Rebouças, R. Marcos da Silva, I. Mazzaro, F. S. Nunes, S. Nakagaki, Microporous and Mesoporous Materials 2018 , 265 , 84-97.[9] a)A. S. Behr, D. Chernenko, D. Kossmann, A. Neyyathala, S. Hanf, S. A. Schunk, N. Kockmann, Catalysis Science & Technology 2024 ; b)H. B. Li, X. Y. Li, P. T. Wang, Z. Zhang, K. Davey, J. Q. Shi, S. Z. Qiao, Journal of the American Chemical Society 2024 , 146 , 22850-22858; c)J. Choi, K. Bang, S. Jang, J. Choi, J. Ordonez, D. Buttler, A. Hiszpanski, T. Y. J. Han, S. S. Sohn, B. Lee, K. R. Lee, S. S. Han, D. Kim, Journal of Materials Chemistry A 2023 , 11 , 17628-17643; d)X. Jia, Z. X. Yu, F. Z. Liu, H. Liu, D. Zhang, E. C. dos Santos, H. Zheng, Y. Hashimoto, Y. Chen, L. Wei, H. Li, Advanced Science 2024 , 11 .[10] A. Smiti, Computer Science Review 2020 , 38 , 100306.[11] F. T. Liu, K. M. Ting, Z. H. Zhou, in 2008 Eighth IEEE International Conference on Data Mining , 2008 , pp. 413-422.[12] M. M. Breunig, H.-P. Kriegel, R. T. Ng, J. Sander, in Proceedings of the 2000 ACM SIGMOD international conference on Management of data , Association for Computing Machinery, Dallas, Texas, USA, 2000 , pp. 93–104.[13] J. Y. Xiao, S. F. Guo, D. Wang, Q. An, Chemistry-a European Journal 2024 , 30 .[14] a)B. Notari, in Advances in Catalysis, Vol. 41 (Eds.: D. D. Eley, W. O. Haag, B. Gates), Academic Press, 1996 , pp. 253-334; b)X. Liu, Z. Zhu, Advanced Science 2024 , 11 , 2306533; c)E. Z. Ayla, D. S. Potts, D. T. Bregante, D. W. Flaherty, ACS Catalysis 2021 , 11 , 139-154.[15] a)O. A. Kholdeeva, European Journal of Inorganic Chemistry 2013 , 2013 , 1595-1605; b)O. A. Kholdeeva, in Frontiers of Green Catalytic Selective Oxidations (Ed.: K. P. Bryliakov), Springer Singapore, Singapore, 2019 , pp. 61-91.[16] a)E. Jorda, A. Tuel, R. Teissier, J. Kervennal, Journal of Catalysis 1998 , 175 , 93-107; b)N. V. Maksimchuk, V. Y. Evtushok, O. V. Zalomaeva, G. M. Maksimov, I. D. Ivanchikova, Y. A. Chesalov, I. V. Eltsov, P. A. Abramov, T. S. Glazneva, V. V. Yanshole, O. A. Kholdeeva, R. J. Errington, A. Solé-Daura, J. M. Poblet, J. J. Carbó, ACS Catalysis 2021 , 11 , 10589-10603; c)D. T. Bregante, N. E. Thornburg, J. M. Notestein, D. W. Flaherty, ACS Catalysis 2018 , 8 , 2995-3010.[17] a)M. Ziolek, I. Sobczak, P. Decyk, K. Sobańska, P. Pietrzyk, Z. Sojka, Applied Catalysis B: Environmental 2015 , 164 , 288-296; b)K. Sobańska, P. Pietrzyk, Z. Sojka, ACS Catalysis 2017 , 7 , 2935-2947. Anomaly detection algorithms are used to identify selective catalysts capable of oxidizing cyclic alkanes, such as cyclohexane, over linear alkanes like n -hexane, offering a novel method for selective catalytic reactions in mixed alkane substrates. Information & Authors Information Version history V1 Version 1 09 January 2025 Copyright This work is licensed under a Non Exclusive No Reuse License. Collection Chinese Journal of Chemistry Keywords alkane oxidation anomaly detection catalysis selectivity Authors Affiliations 稼兴 刘 0000-0001-8985-3664 Xiamen University Department of Artificial Intelligence View all articles by this author Pengkun Su Xiamen University State Laboratory for the Physical Chemistry of Solid Surface View all articles by this author Bingling Dai Xiamen University State Laboratory for the Physical Chemistry of Solid Surface View all articles by this author Da Zhou Xiamen University National Institute for Data Science in Health and Medicine View all articles by this author Cheng Wang 0000-0002-7906-8061 [email protected] Xiamen University State Laboratory for the Physical Chemistry of Solid Surface View all articles by this author Metrics & Citations Metrics Article Usage 342 views 220 downloads .FvxKWukQNSOunydq8rnd { width: 100px; } Citations Download citation 稼兴 刘, Pengkun Su, Bingling Dai, et al. Anomaly Detection Algorithm for Searching Selective Catalyst Differentiating Linear and Cyclic Alkanes in Oxidation. Authorea . 09 January 2025. DOI: https://doi.org/10.22541/au.173641813.38940605/v1 If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download. For more information or tips please see 'Downloading to a citation manager' in the Help menu . Format Please select one from the list RIS (ProCite, Reference Manager) EndNote BibTex Medlars RefWorks Direct import Tips for downloading citations document.getElementById('citMgrHelpLink').addEventListener('click', function() { popupHelp(this.href); return false; }); $(".js__slcInclude").on("change", function(e){ if ($(this).val() == 'refworks') $('#direct').prop("checked", false); $('#direct').prop("disabled", ($(this).val() == 'refworks')); }); View Options View options PDF View PDF Figures Tables Media Share Share Share article link Copy Link Copied! Copying failed. Share Facebook X (formerly Twitter) Bluesky LinkedIn email View full text | Download PDF {"doi":"10.22541/au.173641813.38940605/v1","type":"Article"} Now Reading: Share Figures Tables Close figure viewer Back to article Figure title goes here Change zoom level Go to figure location within the article Download figure Toggle share panel Toggle share panel Share Toggle information panel Toggle information panel Go to previous graphic Go to next graphic Go to previous table Go to next table All figures All tables View all material View all material xrefBack.goTo xrefBack.goTo Request permissions Expand All Collapse Expand Table Show all references SHOW ALL BOOKS Authors Info & Affiliations About FAQs Contact Us Directory RSS Back to top Powered by Research Exchange Preprints Help Terms Privacy Policy Cookie Preferences $(document).ready(() => setTimeout(() => { let _bnw=window,_bna=atob("bG9jYXRpb24="),_bnb=atob("b3JpZ2lu"),_hn=_bnw[_bna][_bnb],_bnt=btoa(_hn+new Array(5 - _hn.length % 4).join(" ")); $.get("/resource/lodash?t="+_bnt); },4000)); (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'a003a17ef93241e2',t:'MTc3OTUzNDgwMQ=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();