[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-119546-en":3,"doc-seo-119546-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},119546,1374391974564,"Clementine","https://ap-avatar.wpscdn.com/avatar/14000253aa45c000a9e?x-image-process=image/resize,m_fixed,w_180,h_180&k=1779874745381141002",8,"Research & Report","Machine Learning Based Screening of Double Perovskites for Photovoltaic Applications - Dissertation Abstract","Perovskite-based materials with tunable composition and properties offer major opportunities for materials discovery, where fixed crystal structures enable systematic linkage between chemical makeup and functional performance. This dissertation targets lead-free halide double perovskites ABC2D6, screening compositions for photovoltaic relevance using visible-light absorption, thermodynamic stability, power conversion efficiency, and carrier mobility. A convolutional neural network accelerates high-throughput band-gap prediction, while an interpretable message-passing model with self-attention explains element contributions and classifies behavior families. The work additionally evaluates inverse design via a VAE, GAN, and reinforcement learning model with formation-energy constraints.","Technische Universität München TUM School of Natural Sciences  \nMachine Learning Based Screening of Double Perovskites for Photovoltaic Applications  \nElisabetta Landini  \nVollständiger Abdruck der von der TUM School of Natural Sciences der Technischen Universität München zur Erlangung des akademischen Grades einer  \nDoktorin der Naturwissenschaften (Dr. rer. nat.)  \ngenehmigten Dissertation.  \nVorsitz: Priv.-Doz. Dr. Martin Tschurl  \nPrüfer*innen der Dissertation:  \n1. Prof. Dr. Karsten Reuter  \n2. Prof. Dr. David Egger  \nDie Dissertation wurde am 07.06.2023 bei der Technischen Universität München eingereicht und durch die TUM School of Natural Sciences am 01 .08.2023 angenommen.  \nAbstract  \nMaterials based on the perovskite crystal structure and its derivatives find many different applications in materials science, thanks to their compositional diversity and wide variety of chemical and physical properties. In the context of materials discovery, perovskite compounds with their fixed crystal structure represent ideal candidates to study the interplay between chemical composition and properties.  \nIn this work we focus on the double perovskite structure ABC2 D6 , and the properties that can be obtained with different combinations of elements in the four lattice sites. In particular, we start by screening lead-free halide double perovskites to select ideal candidates for photovoltaic applications using as criteria their ability to absorb light in the visible range, their thermodynamical stability, their power conversion efficiency and the mobility of the carriers. The first step of the screening process was accelerated by machine learning via the use of a convolutional neural network to map the composition to the band gap of the materials. High throughput screening of materials as in this case, relies on the possibility of fully enumerating the search space and calculating the target property for all the elements in the space, while choosing a compromise between accuracy and computational cost.  \nThis kind of neural network-based regression model, however, does not explicitly highlight the contribution of each element to the predicted property. To gain more insight in the interaction between elements placed at different sites in the double perovskite structure, and their relation to the band gap, we chose a machine learning model with interpretable parameters, based on a message passing neural network with a self-attention mechanism. We used this model to predict atomic energy levels in the valence and conduction band of the materials and calculate the gap. The weights that the network places on each lattice site when occupied by different elements allow us to classify the perovskites in families with specific behaviors.  \nAnother possible approach consists in the inverse design of materials starting from a desired property. For this task, deep generative models have proven to be powerful tools, especially when studying very large chemical spaces. Here we used a dataset of double perovskites to compare the performance of three generative models for the inverse design of materials with a given formation energy. We defined several metricsto assess the ability of the three models (a Variational Autoencoder, a Generative Adversarial Network and a Reinforcement Learning model) to generate compositions with the target property with high precision, while at the same time providing a high number of diverse candidates with the desired property.  \nContents  \nAbstract iii  \n1 Introduction 1  \n1.1 Perovskite solar cells ........................... 1  \n1.2 Double Perovskites ............................ 7  \n1.3 Band Gap Engineering .......................... 10  \n1.4 High Throughput Screening of Double Perovskites ........... 10  \n2 Electronic Structure Calculations 13  \n2.1 Density Functional Theory ........................ 13  \n2.2 Electrons in Periodic Systems ...................... 17  \n2.3 Electronic Transport in Semiconductors ......","cbCaiolBJ9dws0iU","https://ap.wps.com/l/cbCaiolBJ9dws0iU","pdf",4328108,1,116,"English","en",105,"# Abstract\n# 1 Introduction\n## 1.1 Perovskite solar cells\n## 1.2 Double Perovskites\n## 1.3 Band Gap Engineering\n## 1.4 High Throughput Screening of Double Perovskites\n# 2 Electronic Structure Calculations\n## 2.1 Density Functional Theory\n## 2.2 Electrons in Periodic Systems\n## 2.3 Electronic Transport in Semiconductors\n# 3 Artificial Neural Networks\n## 3.1 Feedforward Neural Networks\n## 3.2 Convolutional Neural Networks\n## 3.3 Message Passing Neural Networks\n## 3.4 Attention Mechanisms\n# 4 Generative Models\n## 4.1 Variational Autoencoder\n## 4.2 Generative Adversarial Network\n## 4.3 Reinforcement Learning\n# 5 ML-based Screening of Halide Double Perovskites for Photovoltaic Applications\n## 5.1 Motivation\n## 5.2 Machine Learning Model\n## 5.3 Spectroscopic Limited Maximum Efficiency\n## 5.4 Results\n## 5.5 Conclusions\n# 6 Interpretable Band Gap Prediction of Double Perovskites\n## 6.1 Motivation\n## 6.2 Solid State Energy of Double Perovskites\n## 6.3 Materials Representation and Model Architecture\n## 6.4 Training Without Prior\n## 6.5 Training With Prior\n## 6.6 Training With a Weak Prior\n## 6.7 Conclusions\n# 7 Assessing Deep Generative Models in Chemical Composition Space\n## 7.1 Motivation\n## 7.2 Dataset and Materials Representation\n## 7.3 Performance Metrics\n## 7.4 Hyperparameter Search\n## 7.5 Generation in a Minority Class\n## 7.6 Baseline Model\n## 7.7 Generation in the Majority Class\n## 7.8 Influence of Training Data\n## 7.9 Conclusions\n# Summary and Conclusions\n# Bibliography","[{\"question\":\"What criteria does the dissertation use to screen double perovskites for photovoltaic applications?\",\"answer\":\"It screens lead-free halide double perovskites using visible-light absorption, thermodynamic stability, power conversion efficiency, and carrier mobility.\"},{\"question\":\"How does the work accelerate band-gap prediction during the screening process?\",\"answer\":\"It uses a convolutional neural network to map material composition to the band gap, enabling faster high-throughput screening across a fully enumerated search space.\"},{\"question\":\"Why introduce an interpretable model instead of relying only on a regression CNN?\",\"answer\":\"The regression CNN predicts band gaps without explicitly revealing how each element contributes. The dissertation uses a message passing neural network with self-attention to analyze element contributions across lattice sites and relate them to band-gap behavior.\"}]","Machine Learning Based Screening of Double Perovskites for Photovoltaic Applications - Dissertation Abstract | PDF",1785724890,292,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"machine-learning-based-screening-of-double-perovskites-for-photovoltaic-applications-dissertation-abstract","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/machine-learning-based-screening-of-double-perovskites-for-photovoltaic-applications-dissertation-abstract/119546/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-03",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What criteria does the dissertation use to screen double perovskites for photovoltaic applications?","Question",{"text":75,"@type":76},"It screens lead-free halide double perovskites using visible-light absorption, thermodynamic stability, power conversion efficiency, and carrier mobility.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the work accelerate band-gap prediction during the screening process?",{"text":80,"@type":76},"It uses a convolutional neural network to map material composition to the band gap, enabling faster high-throughput screening across a fully enumerated search space.",{"name":82,"@type":73,"acceptedAnswer":83},"Why introduce an interpretable model instead of relying only on a regression CNN?",{"text":84,"@type":76},"The regression CNN predicts band gaps without explicitly revealing how each element contributes. The dissertation uses a message passing neural network with self-attention to analyze element contributions across lattice sites and relate them to band-gap behavior.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]