{"id":990009,"date":"2023-12-07T12:32:32","date_gmt":"2023-12-07T20:32:32","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?p=990009"},"modified":"2024-06-10T10:12:11","modified_gmt":"2024-06-10T17:12:11","slug":"mattergen-property-guided-materials-design","status":"publish","type":"post","link":"https:\/\/www.microsoft.com\/en-us\/research\/blog\/mattergen-property-guided-materials-design\/","title":{"rendered":"MatterGen: Property-guided materials design"},"content":{"rendered":"\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"1400\" height=\"788\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1.jpg\" alt=\"MatterGen \" class=\"wp-image-990387\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1.jpg 1400w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-300x169.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-1024x576.jpg 1024w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-768x432.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-1066x600.jpg 1066w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-655x368.jpg 655w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-343x193.jpg 343w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-240x135.jpg 240w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-640x360.jpg 640w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-960x540.jpg 960w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-1280x720.jpg 1280w\" sizes=\"(max-width: 1400px) 100vw, 1400px\" \/><\/figure>\n\n\n\n<p><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-admin\/edit.php?post_type=post\"><\/a>Generative AI has revolutionized how we create text and images. How about designing novel materials? We at <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/lab\/microsoft-research-ai4science\/\">Microsoft Research AI4Science<\/a> are thrilled to announce <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/mattergen-a-generative-model-for-inorganic-materials-design\/\">MatterGen<\/a>, our generative model that enables broad property-guided materials design.<\/p>\n\n\n\n<div class=\"wp-block-buttons is-content-justification-center is-content-justification-center is-layout-flex wp-container-core-buttons-is-layout-1 wp-block-buttons-is-layout-flex\">\n<div class=\"wp-block-button\"><a data-bi-type=\"button\" class=\"wp-block-button__link wp-element-button\" href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/mattergen-a-generative-model-for-inorganic-materials-design\/\">Read the paper<\/a><\/div>\n<\/div>\n\n\n\n<div class=\"wp-block-columns aligncenter is-layout-flex wp-container-core-columns-is-layout-1 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-vertically-aligned-center aligncenter is-layout-constrained wp-block-column-is-layout-constrained\" style=\"flex-basis:45%\">\n<figure class=\"wp-block-video aligncenter\"><video autoplay controls loop src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/paper_final_video_firstframe.mp4\" playsinline><\/video><\/figure>\n<\/div>\n<\/div>\n\n\n\n<p>The central challenge in materials science is to discover materials with desired properties, e.g., high Li-ion conductivity for battery materials. Traditionally, this has been done by first finding novel materials and then filtering down based on the application. This is like trying to create the image of a cat by first generating a million different images and then searching for the one with a cat. In MatterGen, we directly generate novel materials with desired properties, similar to how DALL\u00b7E 3 tackles image generation.&nbsp;&nbsp;<\/p>\n\n\n\n<p>MatterGen is a diffusion model specifically designed for generating novel, stable materials. MatterGen also has adapter modules that can be fine-tuned to generate materials given a broad range of constraints, including chemistry, symmetry, and properties.&nbsp;MatterGen generates 2.9 times more stable (\u2264\u202f0.1 eV\/atom of our training + test data convex hull), novel, unique structures than a SOTA model (CDVAE). It also generates structures 17.5 times closer to energy local minimum.&nbsp;MatterGen can directly generate materials satisfying desired magnetic, electronic, mechanical properties via classifier-free guidance. We verify generated materials with DFT-based workflows.&nbsp;<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full is-resized\"><img loading=\"lazy\" decoding=\"async\" width=\"1600\" height=\"2133\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/blog_figure_1.jpg\" alt=\"Figure 1 (alt text) \n\nThis figure displays six pairs of crystalline structures, two for each property constrain. The property constraints are, top to bottom and left to right, high space group symmetry, high bulk modulus, target chemical system, target band gap, high magnetic density, combined high magnetic density and low HHI index.\" class=\"wp-image-990054\" style=\"width:698px;height:auto\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/blog_figure_1.jpg 1600w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/blog_figure_1-225x300.jpg 225w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/blog_figure_1-768x1024.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/blog_figure_1-1152x1536.jpg 1152w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/blog_figure_1-1536x2048.jpg 1536w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/blog_figure_1-135x180.jpg 135w\" sizes=\"(max-width: 1600px) 100vw, 1600px\" \/><figcaption class=\"wp-element-caption\"><em>Figure 1: Stable and new materials generated by MatterGen while constrained on properties.&nbsp;<\/em><\/figcaption><\/figure>\n\n\n\n<p>Additionally, MatterGen can keep generating novel materials that satisfy a target property like high bulk modulus while screening methods instead saturate due to the exhaustion of materials in the database.<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full is-resized\"><img loading=\"lazy\" decoding=\"async\" width=\"1637\" height=\"955\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/bulk_modulus_line-1.jpg\" alt=\"This is a line plot. The x axis indicates the number of DFT property calculations calls; the y axis reports the number of structures found. The title of the plot says \"Bulk modulus greater than 400 gigapascal\". \nThe plot has three lines, one for MatterGen, one for Screening, and one for Labeled Data. The line for MatterGen increases linearly and reaches a y value of 260 at the x value of 500. The line for screening increases until an x value of 300 and then remains almost flat, plateauing at a level of y equaling 140 for x equaling 500. \nThe line for Labeled Data remains flat at the y value of 2 for every value of x.\" class=\"wp-image-990024\" style=\"width:672px;height:auto\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/bulk_modulus_line-1.jpg 1637w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/bulk_modulus_line-1-300x175.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/bulk_modulus_line-1-1024x597.jpg 1024w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/bulk_modulus_line-1-768x448.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/bulk_modulus_line-1-1536x896.jpg 1536w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/bulk_modulus_line-1-480x280.jpg 480w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/bulk_modulus_line-1-240x140.jpg 240w\" sizes=\"(max-width: 1637px) 100vw, 1637px\" \/><figcaption class=\"wp-element-caption\"><em>Figure 2: MatterGen discovers more novel stable high bulk modulus materials than the screening baseline, and does not plateau for increasing computational resources. MatterGen can find more than 250 materials with a bulk modulus > 400 GPa, while only 2 such materials are found in the reference dataset.<\/em><\/figcaption><\/figure>\n\n\n\n<p>MatterGen can also generate materials given target chemical systems. It outperforms substitution and random structure search baselines equipped with MLFF filtering, especially in challenging 5-element systems. MatterGen also generates structures given target space groups. Finally, we tackle the multi-property materials design problem of finding low-supply-chain risk magnets. MatterGen proposes structures that have both high magnetic density and a low supply-chain risk chemical composition.\u00a0<\/p>\n\n\n\n<p>We believe MatterGen is an important step forward in AI for materials design. Our results are currently verified via DFT, which has many known limitations. Experimental verification remains the ultimate test for real-word impact, and we hope to follow up with more results.&nbsp;<\/p>\n\n\n\n<p>None of this would be possible without the highly collaborative work between <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/fowlerandrew\/\">Andrew Fowler<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/claudiozeni\/\">Claudio Zeni<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/dzuegner\/\">Daniel Z\u00fcgner<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/mahorton\/\">Matthew Horton<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/rpinsler\/\">Robert Pinsler<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/ryoto\/\">Ryota Tomioka<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/tianxie\/\">Tian Xie<\/a> and our amazing interns Xiang Fu, Sasha Shysheya, and Jonathan Crabb\u00e9, as well as <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/jakesmith\/\">Jake Smith<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/lixinsun\/\">Lixin Sun<\/a> and the entire AI4Science Materials Design team.&nbsp;&nbsp;<\/p>\n\n\n\n<p>We are also grateful for all the help from Microsoft Research, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/lab\/microsoft-research-ai4science\/\">AI4Science<\/a>, and Azure Quantum.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>The central problem in materials science is to discover materials with desired properties.  MatterGen enables broad property-guided materials design.<\/p>\n","protected":false},"author":42183,"featured_media":990138,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"footnotes":""},"categories":[1],"tags":[],"research-area":[13556],"msr-region":[],"msr-event-type":[],"msr-locale":[268875],"msr-post-option":[243984],"msr-impact-theme":[],"msr-promo-type":[],"msr-podcast-series":[],"class_list":["post-990009","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-research-blog","msr-research-area-artificial-intelligence","msr-locale-en_us","msr-post-option-blog-homepage-featured"],"msr_event_details":{"start":"","end":"","location":""},"podcast_url":"","podcast_episode":"","msr_research_lab":[],"msr_impact_theme":[],"related-publications":[],"related-downloads":[],"related-videos":[],"related-academic-programs":[],"related-groups":[],"related-projects":[],"related-events":[968280],"related-researchers":[{"type":"user_nicename","value":"Andrew Fowler","user_id":41850,"display_name":"Andrew Fowler","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/fowlerandrew\/\" aria-label=\"Visit the profile page for Andrew Fowler\">Andrew Fowler<\/a>","is_active":false,"last_first":"Fowler, Andrew","people_section":0,"alias":"fowlerandrew"},{"type":"user_nicename","value":"Matthew Horton","user_id":43053,"display_name":"Matthew Horton","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/mahorton\/\" aria-label=\"Visit the profile page for Matthew Horton\">Matthew Horton<\/a>","is_active":false,"last_first":"Horton, Matthew","people_section":0,"alias":"mahorton"},{"type":"user_nicename","value":"Ryota Tomioka","user_id":33483,"display_name":"Ryota Tomioka","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/ryoto\/\" aria-label=\"Visit the profile page for Ryota Tomioka\">Ryota Tomioka<\/a>","is_active":false,"last_first":"Tomioka, Ryota","people_section":0,"alias":"ryoto"},{"type":"user_nicename","value":"Robert Pinsler","user_id":41725,"display_name":"Robert Pinsler","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/rpinsler\/\" aria-label=\"Visit the profile page for Robert Pinsler\">Robert Pinsler<\/a>","is_active":false,"last_first":"Pinsler, Robert","people_section":0,"alias":"rpinsler"},{"type":"user_nicename","value":"Tian Xie","user_id":41413,"display_name":"Tian Xie","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/tianxie\/\" aria-label=\"Visit the profile page for Tian Xie\">Tian Xie<\/a>","is_active":false,"last_first":"Xie, Tian","people_section":0,"alias":"tianxie"},{"type":"user_nicename","value":"Claudio Zeni","user_id":41817,"display_name":"Claudio Zeni","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/claudiozeni\/\" aria-label=\"Visit the profile page for Claudio Zeni\">Claudio Zeni<\/a>","is_active":false,"last_first":"Zeni, Claudio","people_section":0,"alias":"claudiozeni"},{"type":"user_nicename","value":"Daniel Z\u00fcgner","user_id":42990,"display_name":"Daniel Z\u00fcgner","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/dzuegner\/\" aria-label=\"Visit the profile page for Daniel Z\u00fcgner\">Daniel Z\u00fcgner<\/a>","is_active":false,"last_first":"Z\u00fcgner, Daniel","people_section":0,"alias":"dzuegner"}],"msr_type":"Post","featured_image_thumbnail":"<img width=\"960\" height=\"540\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-960x540.jpg\" class=\"img-object-cover\" alt=\"four crystalline structures\" decoding=\"async\" loading=\"lazy\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-960x540.jpg 960w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-300x169.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-1024x576.jpg 1024w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-768x432.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-1066x600.jpg 1066w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-655x368.jpg 655w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-343x193.jpg 343w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-240x135.jpg 240w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-640x360.jpg 640w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1-1280x720.jpg 1280w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/12\/MatterGen-BlogHeroFeature-1400x788-1.jpg 1400w\" sizes=\"(max-width: 960px) 100vw, 960px\" \/>","byline":"","formattedDate":"December 7, 2023","formattedExcerpt":"The central problem in materials science is to discover materials with desired properties. MatterGen enables broad property-guided materials design.","locale":{"slug":"en_us","name":"English","native":"","english":"English"},"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts\/990009"}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/users\/42183"}],"replies":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/comments?post=990009"}],"version-history":[{"count":78,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts\/990009\/revisions"}],"predecessor-version":[{"id":990549,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts\/990009\/revisions\/990549"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media\/990138"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=990009"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/categories?post=990009"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/tags?post=990009"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=990009"},{"taxonomy":"msr-region","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-region?post=990009"},{"taxonomy":"msr-event-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event-type?post=990009"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=990009"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=990009"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=990009"},{"taxonomy":"msr-promo-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-promo-type?post=990009"},{"taxonomy":"msr-podcast-series","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-podcast-series?post=990009"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}