{"id":1075809,"date":"2024-05-14T09:50:08","date_gmt":"2024-05-14T16:50:08","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-blog-post&#038;p=1075809"},"modified":"2025-07-17T08:47:13","modified_gmt":"2025-07-17T15:47:13","slug":"mattersim","status":"publish","type":"msr-blog-post","link":"https:\/\/www.microsoft.com\/en-us\/research\/articles\/mattersim\/","title":{"rendered":"MatterSim\uff1a\u4eba\u5de5\u667a\u80fd\u89e3\u9501\u6750\u6599\u8bbe\u8ba1\u7684\u65e0\u9650\u53ef\u80fd"},"content":{"rendered":"\n<p><em>\u4f5c\u8005\uff1a\u79d1\u5b66\u667a\u80fd\u4e2d\u5fc3<\/em><\/p>\n\n\n\n<p>\u7f16\u8005\u6309\uff1a\u968f\u7740\u79d1\u6280\u7684\u4e0d\u65ad\u53d1\u5c55\uff0c\u65b0\u6750\u6599\u5df2\u6210\u4e3a\u589e\u5f3a\u4ea7\u4e1a\u7ade\u4e89\u529b\u7684\u5173\u952e\u56e0\u7d20\u3002\u7136\u800c\uff0c\u65b0\u6750\u6599\u7684\u7269\u7406\u548c\u5316\u5b66\u7279\u6027\u590d\u6742\u591a\u53d8\uff0c\u51c6\u786e\u9884\u6d4b\u5176\u5c5e\u6027\uff0c\u7279\u522b\u662f\u5b9e\u9645\u5408\u6210\u548c\u4f7f\u7528\u6761\u4ef6\u4e0b\u7684\u5c5e\u6027\uff0c\u662f\u7269\u8d28\u79d1\u5b66\u9886\u57df\u4e2d\u957f\u671f\u5b58\u5728\u7684\u6311\u6218\uff0c\u4e5f\u662f\u6750\u6599\u5de5\u4e1a\u6570\u5b57\u5316\u8f6c\u578b\u7684\u6838\u5fc3\u6311\u6218\u4e4b\u4e00\u3002<\/p>\n\n\n\n<p>\u4e3a\u4e86\u7834\u89e3\u8fd9\u4e00\u96be\u9898\uff0c\u5fae\u8f6f\u7814\u7a76\u9662\u79d1\u5b66\u667a\u80fd\u4e2d\u5fc3\uff08Microsoft Research AI for Science\uff09\u5f00\u53d1\u4e86\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b MatterSim\uff0c\u80fd\u591f\u5728\u5e7f\u6cdb\u7684\u5143\u7d20\u3001\u6e29\u5ea6\u548c\u538b\u529b\u8303\u56f4\u5185\u5b9e\u73b0\u51c6\u786e\u9ad8\u6548\u7684\u6750\u6599\u6a21\u62df\u4e0e\u6027\u8d28\u9884\u6d4b\uff0c\u4e3a\u6750\u6599\u8bbe\u8ba1\u7684\u6570\u5b57\u5316\u8f6c\u578b\u63d0\u4f9b\u4e86\u5f3a\u6709\u529b\u7684\u652f\u6301\u3002<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"1400\" height=\"788\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/NEWMatterSim2024-BlogHeroFeature-1400x788-1.png\" alt=\"The image features a complex network of interconnected nodes with a molecular structure, illuminated in blue against a dark background.\" class=\"wp-image-1033260\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/NEWMatterSim2024-BlogHeroFeature-1400x788-1.png 1400w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/NEWMatterSim2024-BlogHeroFeature-1400x788-1-300x169.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/NEWMatterSim2024-BlogHeroFeature-1400x788-1-1024x576.png 1024w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/NEWMatterSim2024-BlogHeroFeature-1400x788-1-768x432.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/NEWMatterSim2024-BlogHeroFeature-1400x788-1-1066x600.png 1066w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/NEWMatterSim2024-BlogHeroFeature-1400x788-1-655x368.png 655w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/NEWMatterSim2024-BlogHeroFeature-1400x788-1-240x135.png 240w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/NEWMatterSim2024-BlogHeroFeature-1400x788-1-640x360.png 640w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/NEWMatterSim2024-BlogHeroFeature-1400x788-1-960x540.png 960w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/NEWMatterSim2024-BlogHeroFeature-1400x788-1-1280x720.png 1280w\" sizes=\"auto, (max-width: 1400px) 100vw, 1400px\" \/><\/figure>\n\n\n\n<p>\u65b0\u6750\u6599\u63a2\u7d22\u5bf9\u7eb3\u7c73\u7535\u5b50\u5b66\u3001\u80fd\u91cf\u50a8\u5b58\u548c\u533b\u7597\u5065\u5eb7\u7b49\u591a\u4e2a\u9886\u57df\u7684\u6280\u672f\u8fdb\u6b65\u81f3\u5173\u91cd\u8981\u3002\u6750\u6599\u8bbe\u8ba1\u4e2d\u7684\u4e00\u4e2a\u6838\u5fc3\u96be\u70b9\u662f\u5982\u4f55\u5728\u4e0d\u8fdb\u884c\u5b9e\u9645\u5408\u6210\u548c\u6d4b\u8bd5\u7684\u60c5\u51b5\u4e0b\u9884\u6d4b\u6750\u6599\u5c5e\u6027\u3002\u7531\u4e8e\u65b0\u6750\u6599\u53ef\u80fd\u6d89\u53ca\u5143\u7d20\u5468\u671f\u8868\u4e2d118\u79cd\u5143\u7d20\u7684\u4efb\u610f\u7ec4\u5408\uff0c\u4e14\u5176\u5408\u6210\u548c\u5de5\u4f5c\u6e29\u5ea6\u3001\u538b\u529b\u8303\u56f4\u6781\u5e7f\uff0c\u8fd9\u4e9b\u56e0\u7d20\u6781\u5927\u5730\u5f71\u54cd\u4e86\u6750\u6599\u5185\u90e8\u539f\u5b50\u7684\u76f8\u4e92\u4f5c\u7528\uff0c\u4f7f\u5f97\u51c6\u786e\u9884\u6d4b\u6750\u6599\u5c5e\u6027\u548c\u884c\u4e3a\u6a21\u62df\u53d8\u5f97\u6781\u4e3a\u56f0\u96be\u3002<\/p>\n\n\n\n<p>\u4e3a\u6b64\uff0c\u5fae\u8f6f\u7814\u7a76\u9662\u79d1\u5b66\u667a\u80fd\u4e2d\u5fc3\uff08Microsoft Research AI for Science\uff09\u63a8\u51fa\u4e86 MatterSim \u6a21\u578b\u3002\u8be5\u6a21\u578b\u5c06\u6df1\u5ea6\u5b66\u4e60\u6280\u672f\u548c\u5927\u89c4\u6a21\u7b2c\u4e00\u6027\u539f\u7406\u8ba1\u7b97\u76f8\u7ed3\u5408\u6765\u5b66\u4e60\u539f\u5b50\u4e4b\u95f4\u7684\u76f8\u4e92\u4f5c\u7528\uff0c\u5b66\u4e60\u7684\u6750\u6599\u7a7a\u95f4\u4ece\u7edd\u5bf9\u96f6\u5ea6\u52305000\u5f00\u5c14\u6587\uff0c\u4ece\u6807\u51c6\u5927\u6c14\u538b\u5230\u4e00\u5343\u4e07\u500d\u5927\u6c14\u538b\uff0c\u80fd\u591f\u9ad8\u6548\u5904\u7406\u591a\u79cd\u6750\u6599\u7684\u6a21\u62df\uff0c\u5305\u62ec\u4f46\u4e0d\u9650\u4e8e\u91d1\u5c5e\u3001\u6c27\u5316\u7269\u3001\u786b\u5316\u7269\u3001\u5364\u5316\u7269\u53ca\u5176\u4e0d\u540c\u72b6\u6001\uff08\u5982\u6676\u4f53\u3001\u975e\u6676\u56fa\u4f53\u548c\u6db2\u4f53\uff09\u3002\u6b64\u5916\uff0cMatterSim \u8fd8\u63d0\u4f9b\u5b9a\u5236\u9009\u9879\uff0c\u53ef\u4ee5\u901a\u8fc7\u6574\u5408\u4e0b\u6e38\u573a\u666f\u7684\u6570\u636e\u6765\u6267\u884c\u66f4\u4e3a\u590d\u6742\u7684\u9884\u6d4b\u4efb\u52a1\u3002<\/p>\n\n\n\n<p><strong>MatterSim: A Deep Learning Atomistic Model Across Elements, Temperatures and Pressures<\/strong><\/p>\n\n\n\n<p><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/arxiv.org\/abs\/2405.04967\">https:\/\/arxiv.org\/abs\/2405.04967<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"1400\" height=\"450\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/MatterSim-figure1-1400x450-1.png\" alt=\"Figure 1: There are two subfigures. On the left-hand side, atomic structures of 12 materials belonging to metals, oxides, sulfides, halides, and organic molecules are shown. On the right-hand side, the temperature and pressure ranges of materials' application and synthesis are plotted.\" class=\"wp-image-1031901\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/MatterSim-figure1-1400x450-1.png 1400w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/MatterSim-figure1-1400x450-1-300x96.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/MatterSim-figure1-1400x450-1-1024x329.png 1024w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/MatterSim-figure1-1400x450-1-768x247.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/MatterSim-figure1-1400x450-1-240x77.png 240w\" sizes=\"auto, (max-width: 1400px) 100vw, 1400px\" \/><figcaption class=\"wp-element-caption\"><em>\u56fe1\uff1aMatterSim \u80fd\u591f\u5728\u5e7f\u6cdb\u7684\u6e29\u5ea6\u3001\u538b\u529b\u6761\u4ef6\u4e0b\uff0c\u6a21\u62df\u91d1\u5c5e\u3001\u6c27\u5316\u7269\u3001\u786b\u5316\u7269\u3001\u5364\u5316\u7269\u7b49\u5404\u79cd\u6750\u6599\u7684\u6027\u8d28\u548c\u884c\u4e3a\u3002<\/em><\/figcaption><\/figure>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"\u5168\u5468\u671f\u8868-\u8de8\u6e29\u5ea6-\u8de8\u538b\u529b\u7684\u771f\u5b9e\u6761\u4ef6\u4e0b\u7684\u6750\u6599\u6a21\u62df\">\u5168\u5468\u671f\u8868\u3001\u8de8\u6e29\u5ea6\u3001\u8de8\u538b\u529b\u7684\u771f\u5b9e\u6761\u4ef6\u4e0b\u7684\u6750\u6599\u6a21\u62df<\/h2>\n\n\n\n<p>MatterSim \u7684\u8bad\u7ec3\u8fc7\u7a0b\u4f7f\u7528\u4e86\u5927\u89c4\u6a21\u7684\u5408\u6210\u6570\u636e\u3002\u4e3a\u4e86\u83b7\u5f97\u8fd9\u4e9b\u8bad\u7ec3\u6570\u636e\uff0c\u7814\u7a76\u5458\u4eec\u7ed3\u5408\u4e86\u4e3b\u52a8\u5b66\u4e60\u3001\u5206\u5b50\u52a8\u529b\u5b66\u6a21\u62df\u548c\u751f\u6210\u6a21\u578b\u7b49\u6280\u672f\uff0c\u6784\u5efa\u4e86\u9ad8\u6548\u7684\u6570\u636e\u751f\u6210\u65b9\u6848\u3002\u8fd9\u79cd\u6570\u636e\u751f\u6210\u7b56\u7565\u786e\u4fdd\u4e86\u6a21\u578b\u5bf9\u6750\u6599\u7a7a\u95f4\u7684\u5e7f\u6cdb\u8986\u76d6\uff0c\u4f7f\u5176\u80fd\u591f\u4ee5\u4e0e\u7b2c\u4e00\u6027\u539f\u7406\u9884\u6d4b\u76f8\u5f53\u7684\u51c6\u786e\u5ea6\uff0c\u9884\u6d4b\u6750\u6599\u5728\u539f\u5b50\u5c42\u9762\u7684\u80fd\u91cf\u3001\u529b\u548c\u5e94\u529b\u3002<\/p>\n\n\n\n<p>\u4e0e\u5f53\u524d\u7684 SOTA \u6a21\u578b\u76f8\u6bd4\uff0cMatterSim \u5728\u6709\u9650\u7684\u6e29\u5ea6\u548c\u538b\u529b\u4e0b\u5bf9\u6750\u6599\u5c5e\u6027\u9884\u6d4b\u7684\u51c6\u786e\u5ea6\u63d0\u9ad8\u4e8610\u500d\u3002\u7814\u7a76\u8868\u660e\uff0cMatterSim \u80fd\u591f\u51c6\u786e\u6a21\u62df\u5305\u62ec\u70ed\u529b\u5b66\u3001\u529b\u5b66\u548c\u8fd0\u8f93\u7279\u6027\u5728\u5185\u7684\u5e7f\u6cdb\u6750\u6599\u5c5e\u6027\u3002\u503c\u5f97\u4e00\u63d0\u7684\u662f\uff0cMatterSim \u53ef\u4ee5\u4ece\u5934\u6784\u5efa\u6750\u6599\u76f8\u56fe\uff0c\u4e3a\u65b0\u6750\u6599\u7684\u53ef\u5408\u6210\u6027\u4e0e\u5408\u6210\u6761\u4ef6\u63d0\u4f9b\u6307\u5bfc\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"1400\" height=\"451\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/MatterSim-figure2-1400x450-1.png\" alt=\"Figure 2: There are three subfigures. The panel on the left shows a comparison of the highest phonon frequency predicted by MatterSim and by first-principles methods. The two values are for each material is very close, leading to a nearly straight line in the parity plot. The middle panel depicts the same relation of free energies of around 50 materials and comparison between MatterSim and first-principles results. The right panel shows the phase diagram of MgO predicted using MatterSim. The x-axis denotes the temperature and the y-axis denotes the pressure. The pressure ranges of where MgO\u2019s B1 phase is below 500 GPa and this range decreases with temperature increase. The blue lines show the prediction from MatterSim and fits well with the shaded region which is the result from experiment measurement.\" class=\"wp-image-1031904\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/MatterSim-figure2-1400x450-1.png 1400w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/MatterSim-figure2-1400x450-1-300x97.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/MatterSim-figure2-1400x450-1-1024x330.png 1024w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/MatterSim-figure2-1400x450-1-768x247.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/MatterSim-figure2-1400x450-1-240x77.png 240w\" sizes=\"auto, (max-width: 1400px) 100vw, 1400px\" \/><figcaption class=\"wp-element-caption\"><em>\u56fe2\uff1aMatterSim \u9884\u6d4b\u5c5e\u6027\u4e0e\u5b9e\u9a8c\u6d4b\u91cf\u7ed3\u679c\u7684\u6bd4\u8f83\u3002MatterSim \u5728\u9884\u6d4b\u6750\u6599\u7684\u632f\u52a8\u6027\u8d28\u3001\u673a\u68b0\u6027\u8d28\u548c\u76f8\u56fe\u65b9\u9762\u5747\u8fbe\u5230\u4e86\u4e0e\u7b2c\u4e00\u6027\u539f\u7406\u8ba1\u7b97\u548c\u5b9e\u9a8c\u6d4b\u91cf\u76f8\u5f53\u7684\u7cbe\u5ea6\u3002<\/em><\/figcaption><\/figure>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"\u5b9a\u5236\u5316\u7cbe\u8c03-\u7528\u66f4\u5c11\u7684\u8bad\u7ec3\u6570\u636e\u8fbe\u5230\u5b9e\u9a8c\u7cbe\u5ea6\u7684\u6a21\u62df\">\u5b9a\u5236\u5316\u7cbe\u8c03\uff1a\u7528\u66f4\u5c11\u7684\u8bad\u7ec3\u6570\u636e\u8fbe\u5230\u5b9e\u9a8c\u7cbe\u5ea6\u7684\u6a21\u62df<\/h2>\n\n\n\n<p>\u5c3d\u7ba1\u57fa\u4e8e\u5927\u91cf\u7684\u5408\u6210\u6570\u636e\u96c6\u8fdb\u884c\u8bad\u7ec3\uff0cMatterSim \u53ef\u4ee5\u901a\u8fc7\u6574\u5408\u989d\u5916\u7684\u6570\u636e\u6765\u6ee1\u8db3\u7279\u5b9a\u7684\u8bbe\u8ba1\u8981\u6c42\u3002MatterSim \u80fd\u591f\u5229\u7528\u4e3b\u52a8\u5b66\u4e60\u548c\u5fae\u8c03\uff0c\u5b9a\u5236\u5316\u5b8c\u6210\u7279\u5b9a\u573a\u5408\u4e0b\u9ad8\u5ea6\u590d\u6742\u7684\u6750\u6599\u6a21\u62df\u548c\u8bbe\u8ba1\u4efb\u52a1\u3002<\/p>\n\n\n\n<p>\u4ee5\u7269\u8d28\u8ba1\u7b97\u9886\u57df\u7684\u7ecf\u5178\u4efb\u52a1\u2014\u2014\u6db2\u6001\u6c34\u6027\u8d28\u7684\u6a21\u62df\u4e3a\u4f8b\uff0c\u8fd9\u4e2a\u4efb\u52a1\u770b\u4f3c\u7b80\u5355\uff0c\u5b9e\u9645\u4e0a\u5374\u9700\u8981\u5927\u91cf\u7684\u8ba1\u7b97\u3002\u901a\u8fc7 MatterSim \u7684\u5b9a\u5236\u5316\u529f\u80fd\u5bf9\u8be5\u4efb\u52a1\u8fdb\u884c\u4f18\u5316\uff0cMatterSim \u53ea\u9700\u89813%\u7684\u539f\u59cb\u6570\u636e\uff0c\u5c31\u80fd\u8fbe\u5230\u9884\u671f\u7684\u5b9e\u9a8c\u7cbe\u5ea6\u6a21\u62df\u3002\u76f8\u6bd4\u4e4b\u4e0b\uff0c\u7528\u4f20\u7edf\u65b9\u6cd5\u8bad\u7ec3\u7684\u4e13\u6709\u673a\u5668\u5b66\u4e60\u6a21\u578b\uff0c\u5219\u9700\u8981\u63d0\u4f9b30\u500d\u7684\u8d44\u6e90\uff0c\u800c\u7b2c\u4e00\u6027\u539f\u7406\u7684\u65b9\u6cd5\u5219\u9700\u8981\u589e\u52a0\u6307\u6570\u7ea7\u7684\u8d44\u6e90\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"1400\" height=\"451\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/MatterSim-figure3-1400x450-1.png\" alt=\"Figure 3: There are two panels in this figure. The right panel shows the structure of Li2B12H12, a complex material system used for solid-state batteries. This system is used in the benchmark of the performance of MatterSim. The left panel panels show the comparison between number of data point needed to train a model from scratch and customize from MatterSim to achieve the same accuracy. MatterSim requires 3% and 10% of the data for the two tasks compared with training from scratch.\" class=\"wp-image-1031907\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/MatterSim-figure3-1400x450-1.png 1400w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/MatterSim-figure3-1400x450-1-300x97.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/MatterSim-figure3-1400x450-1-1024x330.png 1024w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/MatterSim-figure3-1400x450-1-768x247.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/05\/MatterSim-figure3-1400x450-1-240x77.png 240w\" sizes=\"auto, (max-width: 1400px) 100vw, 1400px\" \/><figcaption class=\"wp-element-caption\"><em>\u56fe3\uff1aMatterSim \u5728\u8fdb\u884c\u7cbe\u7ec6\u6750\u6599\u6a21\u62df\u3001\u6027\u80fd\u9884\u6d4b\u65f6\uff0c\u80fd\u591f\u964d\u4f4e90%-97%\u7684\u6570\u636e\u9700\u6c42\u91cf\u3002<\/em><\/figcaption><\/figure>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"\u5f25\u5408\u6750\u6599\u5fae\u89c2\u6a21\u578b\u4e0e\u5b8f\u89c2\u5c5e\u6027\u95f4\u7684\u9e3f\u6c9f\">\u5f25\u5408\u6750\u6599\u5fae\u89c2\u6a21\u578b\u4e0e\u5b8f\u89c2\u5c5e\u6027\u95f4\u7684\u9e3f\u6c9f<\/h2>\n\n\n\n<p>\u4ece\u539f\u5b50\u7ed3\u6784\u9884\u6d4b\u6750\u6599\u5b8f\u89c2\u5c5e\u6027\u662f\u4e00\u9879\u590d\u6742\u7684\u4efb\u52a1\uff0c\u5bf9\u4e8e\u76ee\u524d\u57fa\u4e8e\u7edf\u8ba1\u529b\u5b66\u7684\u65b9\u6cd5\uff0c\u5982\u5206\u5b50\u52a8\u529b\u5b66\uff0c\u592a\u8fc7\u590d\u6742\u3002MatterSim \u901a\u8fc7\u6df1\u5ea6\u5b66\u4e60\u6280\u672f\u76f4\u63a5\u6620\u5c04\u8fd9\u4e9b\u5173\u7cfb\u6765\u89e3\u51b3\u8fd9\u4e2a\u95ee\u9898\u3002\u7814\u7a76\u5458\u4eec\u4e3a MatterSim \u4e13\u95e8\u8bbe\u8ba1\u4e86\u53ef\u4ee5\u5fae\u8c03\u7684\u9002\u914d\u5668\u6a21\u5757\uff0c\u53ef\u4ee5\u76f4\u63a5\u4ece\u7ed3\u6784\u6570\u636e\u9884\u6d4b\u6750\u6599\u5c5e\u6027\uff0c\u907f\u514d\u4e86\u590d\u6742\u6a21\u62df\u7684\u9700\u6c42\u3002\u5728\u8457\u540d\u7684\u6750\u6599\u5c5e\u6027\u9884\u6d4b\u57fa\u51c6\u6d4b\u8bd5\u96c6 MatBench \u4e0a\u7684\u57fa\u51c6\u6d4b\u8bd5\u4e2d\uff0cMatterSim \u7684\u7cbe\u786e\u5ea6\u6709\u4e86\u663e\u8457\u63d0\u5347\uff0c\u5e76\u4f18\u4e8e\u6240\u6709\u9488\u5bf9\u7279\u5b9a\u5c5e\u6027\u7684\u4e13\u6709\u6a21\u578b\uff0c\u5c55\u73b0\u4e86\u5176\u76f4\u63a5\u4ece\u9886\u57df\u7279\u5b9a\u6570\u636e\u9884\u6d4b\u6750\u6599\u5c5e\u6027\u7684\u5f3a\u5927\u80fd\u529b\u3002<\/p>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"\u5f00\u542f\u4eba\u5de5\u667a\u80fd\u8f85\u52a9\u6750\u6599\u8bbe\u8ba1\u65b0\u7bc7\u7ae0\">\u5f00\u542f\u4eba\u5de5\u667a\u80fd\u8f85\u52a9\u6750\u6599\u8bbe\u8ba1\u65b0\u7bc7\u7ae0<\/h2>\n\n\n\n<p>MatterSim \u4e3a\u9c81\u68d2\u4e14\u9ad8\u6548\u7684\u6750\u6599\u6a21\u62df\u548c\u6027\u8d28\u9884\u6d4b\u63d0\u4f9b\u4e86\u65b0\u7684\u53ef\u80fd\u6027\uff0c\u5176\u4e0e Distributional Graphormer \u7b49\u751f\u6210\u5f0f\u4eba\u5de5\u667a\u80fd\u6280\u672f\u548c\u5f3a\u5316\u5b66\u4e60\u6280\u672f\u7684\u7ed3\u5408\uff0c\u5c06\u6709\u671b\u5f7b\u5e95\u6539\u53d8\u6750\u6599\u79d1\u5b66\u7814\u7a76\u7684\u7406\u5ff5\u4e0e\u6a21\u5f0f\u3002\u4ee5 MatterSim\u3001MatterGen \u4e3a\u4ee3\u8868\u7684\u6750\u6599\u79d1\u5b66\u4e0e\u4eba\u5de5\u667a\u80fd\u6a21\u578b\uff0c\u4e5f\u5c06\u4f7f\u5b9a\u5236\u5316\u6750\u6599\u7684\u5f00\u53d1\u53d8\u5f97\u66f4\u52a0\u9ad8\u6548\uff0c\u5e94\u7528\u8303\u56f4\u66f4\u5e7f\uff0c\u4ece\u534a\u5bfc\u4f53\u6280\u672f\u5230\u751f\u7269\u533b\u836f\u5de5\u7a0b\u7b49\u8bf8\u591a\u9886\u57df\u90fd\u5c06\u4ece\u4e2d\u53d7\u76ca\u3002<\/p>\n\n\n\n<p>\u672a\u6765\uff0c\u7814\u7a76\u5458\u4eec\u5c06\u7ee7\u7eed\u63a8\u8fdb MatterSim \u7684\u5b9e\u9a8c\u9a8c\u8bc1\uff0c\u5e0c\u671b\u76f8\u5173\u7814\u7a76\u53ef\u4ee5\u5728\u9762\u5411\u53ef\u6301\u7eed\u53d1\u5c55\u7684\u50ac\u5316\u5242\u8bbe\u8ba1\u3001\u80fd\u6e90\u5b58\u50a8\u3001\u7eb3\u7c73\u6280\u672f\u7b49\u9886\u57df\u53d1\u6325\u4f5c\u7528\uff0c\u9020\u798f\u5168\u7403\u793e\u4f1a\u3002<\/p>\n\n\n\n<p>\u76f8\u5173\u94fe\u63a5\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.msra.cn\/zh-cn\/news\/features\/distributional-graphormer\" target=\"_blank\" rel=\"noopener noreferrer\">Distributional Graphormer\uff1a\u4ece\u5206\u5b50\u7ed3\u6784\u9884\u6d4b\u5230\u5e73\u8861\u5206\u5e03\u9884\u6d4b<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/blog\/mattergen-property-guided-materials-design\/\">MatterGen: Property-guided materials design<\/a><\/li>\n\n\n\n<li><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/matbench.materialsproject.org\/\" target=\"_blank\" rel=\"noopener noreferrer\">MatBench<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>\u4f5c\u8005\uff1a\u79d1\u5b66\u667a\u80fd\u4e2d\u5fc3 \u7f16\u8005\u6309\uff1a\u968f\u7740\u79d1\u6280\u7684\u4e0d\u65ad\u53d1\u5c55\uff0c\u65b0\u6750\u6599\u5df2\u6210\u4e3a\u589e\u5f3a\u4ea7\u4e1a\u7ade\u4e89\u529b\u7684\u5173\u952e\u56e0\u7d20\u3002\u7136\u800c\uff0c\u65b0\u6750\u6599\u7684\u7269\u7406\u548c\u5316\u5b66\u7279\u6027\u590d\u6742\u591a\u53d8\uff0c\u51c6\u786e\u9884\u6d4b\u5176\u5c5e\u6027\uff0c\u7279\u522b\u662f\u5b9e\u9645\u5408\u6210\u548c\u4f7f\u7528\u6761\u4ef6\u4e0b\u7684\u5c5e\u6027\uff0c\u662f\u7269\u8d28\u79d1\u5b66\u9886\u57df\u4e2d\u957f\u671f\u5b58\u5728\u7684\u6311\u6218\uff0c\u4e5f\u662f\u6750\u6599\u5de5\u4e1a\u6570\u5b57\u5316\u8f6c\u578b\u7684\u6838\u5fc3\u6311\u6218\u4e4b\u4e00\u3002 \u4e3a\u4e86\u7834\u89e3\u8fd9\u4e00\u96be\u9898\uff0c\u5fae\u8f6f\u7814\u7a76\u9662\u79d1\u5b66\u667a\u80fd\u4e2d\u5fc3\uff08Microsoft Research AI for Science\uff09\u5f00\u53d1\u4e86\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b MatterSim\uff0c\u80fd\u591f\u5728\u5e7f\u6cdb\u7684\u5143\u7d20\u3001\u6e29\u5ea6\u548c\u538b\u529b\u8303\u56f4\u5185\u5b9e\u73b0\u51c6\u786e\u9ad8\u6548\u7684\u6750\u6599\u6a21\u62df\u4e0e\u6027\u8d28\u9884\u6d4b\uff0c\u4e3a\u6750\u6599\u8bbe\u8ba1\u7684\u6570\u5b57\u5316\u8f6c\u578b\u63d0\u4f9b\u4e86\u5f3a\u6709\u529b\u7684\u652f\u6301\u3002 \u65b0\u6750\u6599\u63a2\u7d22\u5bf9\u7eb3\u7c73\u7535\u5b50\u5b66\u3001\u80fd\u91cf\u50a8\u5b58\u548c\u533b\u7597\u5065\u5eb7\u7b49\u591a\u4e2a\u9886\u57df\u7684\u6280\u672f\u8fdb\u6b65\u81f3\u5173\u91cd\u8981\u3002\u6750\u6599\u8bbe\u8ba1\u4e2d\u7684\u4e00\u4e2a\u6838\u5fc3\u96be\u70b9\u662f\u5982\u4f55\u5728\u4e0d\u8fdb\u884c\u5b9e\u9645\u5408\u6210\u548c\u6d4b\u8bd5\u7684\u60c5\u51b5\u4e0b\u9884\u6d4b\u6750\u6599\u5c5e\u6027\u3002\u7531\u4e8e\u65b0\u6750\u6599\u53ef\u80fd\u6d89\u53ca\u5143\u7d20\u5468\u671f\u8868\u4e2d118\u79cd\u5143\u7d20\u7684\u4efb\u610f\u7ec4\u5408\uff0c\u4e14\u5176\u5408\u6210\u548c\u5de5\u4f5c\u6e29\u5ea6\u3001\u538b\u529b\u8303\u56f4\u6781\u5e7f\uff0c\u8fd9\u4e9b\u56e0\u7d20\u6781\u5927\u5730\u5f71\u54cd\u4e86\u6750\u6599\u5185\u90e8\u539f\u5b50\u7684\u76f8\u4e92\u4f5c\u7528\uff0c\u4f7f\u5f97\u51c6\u786e\u9884\u6d4b\u6750\u6599\u5c5e\u6027\u548c\u884c\u4e3a\u6a21\u62df\u53d8\u5f97\u6781\u4e3a\u56f0\u96be\u3002 \u4e3a\u6b64\uff0c\u5fae\u8f6f\u7814\u7a76\u9662\u79d1\u5b66\u667a\u80fd\u4e2d\u5fc3\uff08Microsoft Research AI for Science\uff09\u63a8\u51fa\u4e86 MatterSim \u6a21\u578b\u3002\u8be5\u6a21\u578b\u5c06\u6df1\u5ea6\u5b66\u4e60\u6280\u672f\u548c\u5927\u89c4\u6a21\u7b2c\u4e00\u6027\u539f\u7406\u8ba1\u7b97\u76f8\u7ed3\u5408\u6765\u5b66\u4e60\u539f\u5b50\u4e4b\u95f4\u7684\u76f8\u4e92\u4f5c\u7528\uff0c\u5b66\u4e60\u7684\u6750\u6599\u7a7a\u95f4\u4ece\u7edd\u5bf9\u96f6\u5ea6\u52305000\u5f00\u5c14\u6587\uff0c\u4ece\u6807\u51c6\u5927\u6c14\u538b\u5230\u4e00\u5343\u4e07\u500d\u5927\u6c14\u538b\uff0c\u80fd\u591f\u9ad8\u6548\u5904\u7406\u591a\u79cd\u6750\u6599\u7684\u6a21\u62df\uff0c\u5305\u62ec\u4f46\u4e0d\u9650\u4e8e\u91d1\u5c5e\u3001\u6c27\u5316\u7269\u3001\u786b\u5316\u7269\u3001\u5364\u5316\u7269\u53ca\u5176\u4e0d\u540c\u72b6\u6001\uff08\u5982\u6676\u4f53\u3001\u975e\u6676\u56fa\u4f53\u548c\u6db2\u4f53\uff09\u3002\u6b64\u5916\uff0cMatterSim \u8fd8\u63d0\u4f9b\u5b9a\u5236\u9009\u9879\uff0c\u53ef\u4ee5\u901a\u8fc7\u6574\u5408\u4e0b\u6e38\u573a\u666f\u7684\u6570\u636e\u6765\u6267\u884c\u66f4\u4e3a\u590d\u6742\u7684\u9884\u6d4b\u4efb\u52a1\u3002 MatterSim: A Deep Learning Atomistic Model Across Elements, Temperatures and Pressures https:\/\/arxiv.org\/abs\/2405.04967 (opens in new tab) MatterSim \u7684\u8bad\u7ec3\u8fc7\u7a0b\u4f7f\u7528\u4e86\u5927\u89c4\u6a21\u7684\u5408\u6210\u6570\u636e\u3002\u4e3a\u4e86\u83b7\u5f97\u8fd9\u4e9b\u8bad\u7ec3\u6570\u636e\uff0c\u7814\u7a76\u5458\u4eec\u7ed3\u5408\u4e86\u4e3b\u52a8\u5b66\u4e60\u3001\u5206\u5b50\u52a8\u529b\u5b66\u6a21\u62df\u548c\u751f\u6210\u6a21\u578b\u7b49\u6280\u672f\uff0c\u6784\u5efa\u4e86\u9ad8\u6548\u7684\u6570\u636e\u751f\u6210\u65b9\u6848\u3002\u8fd9\u79cd\u6570\u636e\u751f\u6210\u7b56\u7565\u786e\u4fdd\u4e86\u6a21\u578b\u5bf9\u6750\u6599\u7a7a\u95f4\u7684\u5e7f\u6cdb\u8986\u76d6\uff0c\u4f7f\u5176\u80fd\u591f\u4ee5\u4e0e\u7b2c\u4e00\u6027\u539f\u7406\u9884\u6d4b\u76f8\u5f53\u7684\u51c6\u786e\u5ea6\uff0c\u9884\u6d4b\u6750\u6599\u5728\u539f\u5b50\u5c42\u9762\u7684\u80fd\u91cf\u3001\u529b\u548c\u5e94\u529b\u3002 \u4e0e\u5f53\u524d\u7684 SOTA \u6a21\u578b\u76f8\u6bd4\uff0cMatterSim \u5728\u6709\u9650\u7684\u6e29\u5ea6\u548c\u538b\u529b\u4e0b\u5bf9\u6750\u6599\u5c5e\u6027\u9884\u6d4b\u7684\u51c6\u786e\u5ea6\u63d0\u9ad8\u4e8610\u500d\u3002\u7814\u7a76\u8868\u660e\uff0cMatterSim \u80fd\u591f\u51c6\u786e\u6a21\u62df\u5305\u62ec\u70ed\u529b\u5b66\u3001\u529b\u5b66\u548c\u8fd0\u8f93\u7279\u6027\u5728\u5185\u7684\u5e7f\u6cdb\u6750\u6599\u5c5e\u6027\u3002\u503c\u5f97\u4e00\u63d0\u7684\u662f\uff0cMatterSim \u53ef\u4ee5\u4ece\u5934\u6784\u5efa\u6750\u6599\u76f8\u56fe\uff0c\u4e3a\u65b0\u6750\u6599\u7684\u53ef\u5408\u6210\u6027\u4e0e\u5408\u6210\u6761\u4ef6\u63d0\u4f9b\u6307\u5bfc\u3002 \u5c3d\u7ba1\u57fa\u4e8e\u5927\u91cf\u7684\u5408\u6210\u6570\u636e\u96c6\u8fdb\u884c\u8bad\u7ec3\uff0cMatterSim \u53ef\u4ee5\u901a\u8fc7\u6574\u5408\u989d\u5916\u7684\u6570\u636e\u6765\u6ee1\u8db3\u7279\u5b9a\u7684\u8bbe\u8ba1\u8981\u6c42\u3002MatterSim \u80fd\u591f\u5229\u7528\u4e3b\u52a8\u5b66\u4e60\u548c\u5fae\u8c03\uff0c\u5b9a\u5236\u5316\u5b8c\u6210\u7279\u5b9a\u573a\u5408\u4e0b\u9ad8\u5ea6\u590d\u6742\u7684\u6750\u6599\u6a21\u62df\u548c\u8bbe\u8ba1\u4efb\u52a1\u3002 \u4ee5\u7269\u8d28\u8ba1\u7b97\u9886\u57df\u7684\u7ecf\u5178\u4efb\u52a1\u2014\u2014\u6db2\u6001\u6c34\u6027\u8d28\u7684\u6a21\u62df\u4e3a\u4f8b\uff0c\u8fd9\u4e2a\u4efb\u52a1\u770b\u4f3c\u7b80\u5355\uff0c\u5b9e\u9645\u4e0a\u5374\u9700\u8981\u5927\u91cf\u7684\u8ba1\u7b97\u3002\u901a\u8fc7 MatterSim \u7684\u5b9a\u5236\u5316\u529f\u80fd\u5bf9\u8be5\u4efb\u52a1\u8fdb\u884c\u4f18\u5316\uff0cMatterSim \u53ea\u9700\u89813%\u7684\u539f\u59cb\u6570\u636e\uff0c\u5c31\u80fd\u8fbe\u5230\u9884\u671f\u7684\u5b9e\u9a8c\u7cbe\u5ea6\u6a21\u62df\u3002\u76f8\u6bd4\u4e4b\u4e0b\uff0c\u7528\u4f20\u7edf\u65b9\u6cd5\u8bad\u7ec3\u7684\u4e13\u6709\u673a\u5668\u5b66\u4e60\u6a21\u578b\uff0c\u5219\u9700\u8981\u63d0\u4f9b30\u500d\u7684\u8d44\u6e90\uff0c\u800c\u7b2c\u4e00\u6027\u539f\u7406\u7684\u65b9\u6cd5\u5219\u9700\u8981\u589e\u52a0\u6307\u6570\u7ea7\u7684\u8d44\u6e90\u3002 \u4ece\u539f\u5b50\u7ed3\u6784\u9884\u6d4b\u6750\u6599\u5b8f\u89c2\u5c5e\u6027\u662f\u4e00\u9879\u590d\u6742\u7684\u4efb\u52a1\uff0c\u5bf9\u4e8e\u76ee\u524d\u57fa\u4e8e\u7edf\u8ba1\u529b\u5b66\u7684\u65b9\u6cd5\uff0c\u5982\u5206\u5b50\u52a8\u529b\u5b66\uff0c\u592a\u8fc7\u590d\u6742\u3002MatterSim \u901a\u8fc7\u6df1\u5ea6\u5b66\u4e60\u6280\u672f\u76f4\u63a5\u6620\u5c04\u8fd9\u4e9b\u5173\u7cfb\u6765\u89e3\u51b3\u8fd9\u4e2a\u95ee\u9898\u3002\u7814\u7a76\u5458\u4eec\u4e3a MatterSim \u4e13\u95e8\u8bbe\u8ba1\u4e86\u53ef\u4ee5\u5fae\u8c03\u7684\u9002\u914d\u5668\u6a21\u5757\uff0c\u53ef\u4ee5\u76f4\u63a5\u4ece\u7ed3\u6784\u6570\u636e\u9884\u6d4b\u6750\u6599\u5c5e\u6027\uff0c\u907f\u514d\u4e86\u590d\u6742\u6a21\u62df\u7684\u9700\u6c42\u3002\u5728\u8457\u540d\u7684\u6750\u6599\u5c5e\u6027\u9884\u6d4b\u57fa\u51c6\u6d4b\u8bd5\u96c6 MatBench \u4e0a\u7684\u57fa\u51c6\u6d4b\u8bd5\u4e2d\uff0cMatterSim \u7684\u7cbe\u786e\u5ea6\u6709\u4e86\u663e\u8457\u63d0\u5347\uff0c\u5e76\u4f18\u4e8e\u6240\u6709\u9488\u5bf9\u7279\u5b9a\u5c5e\u6027\u7684\u4e13\u6709\u6a21\u578b\uff0c\u5c55\u73b0\u4e86\u5176\u76f4\u63a5\u4ece\u9886\u57df\u7279\u5b9a\u6570\u636e\u9884\u6d4b\u6750\u6599\u5c5e\u6027\u7684\u5f3a\u5927\u80fd\u529b\u3002 [&hellip;]<\/p>\n","protected":false},"author":42735,"featured_media":1076733,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr-content-parent":1012650,"msr_hide_image_in_river":null,"footnotes":""},"research-area":[13556],"msr-locale":[268881],"msr-post-option":[],"class_list":["post-1075809","msr-blog-post","type-msr-blog-post","status-publish","has-post-thumbnail","hentry","msr-research-area-artificial-intelligence","msr-locale-zh_cn"],"msr_assoc_parent":{"id":1012650,"type":"lab"},"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post\/1075809","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-blog-post"}],"author":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/users\/42735"}],"version-history":[{"count":3,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post\/1075809\/revisions"}],"predecessor-version":[{"id":1076736,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post\/1075809\/revisions\/1076736"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media\/1076733"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=1075809"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=1075809"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=1075809"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=1075809"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}