T2MAT (文本到材料)：基于文本生成目标性能材料结构的通用智能体

引用本文: 宋志龙, 陆帅华, 周跫桦, 王金兰. T2MAT (文本到材料)：基于文本生成目标性能材料结构的通用智能体[J]. 物理化学学报, 2026, 42(5): 100213. doi: 10.1016/j.actphy.2025.100213 shu

Citation: Zhilong Song, Shuaihua Lu, Qionghua Zhou, Jinlan Wang. T2MAT (text-to-material): A universal agent for generating material structures with goal properties from a single sentence[J]. Acta Physico-Chimica Sinica, 2026, 42(5): 100213. doi: 10.1016/j.actphy.2025.100213 shu

T2MAT (文本到材料)：基于文本生成目标性能材料结构的通用智能体

宋志龙 ^{1, 2,} ,
陆帅华 ^1, ,
周跫桦 ^{1, 2,
,} ,
王金兰 ^{1, 2,
,}

1.
东南大学物理学院量子材料与器件教育部重点实验室, 江苏南京 211189
2.
苏州实验室, 江苏苏州 215004

通讯作者: Email: qh.zhou@seu.edu.cn (周跫桦); jlwang@seu.edu.cn (王金兰)

摘要: 人工智能生成内容(AIGC)——由AI系统无需人工干预自主生成的内容——已显著提升多个领域的效率。然而，材料科学中的AIGC技术面临双重挑战：既要高效发现超越现有数据库范围的新型材料，又需确保晶体结构的对称性与稳定性。为应对该挑战，我们开发了T2MAT(文本到材料)，这是一种端到端智能体，能够通过全面探索化学空间并结合全自动第一性原理验证，将用户输入的文本转化为超越现有数据库范围、具有目标性能的新型材料结构的逆向设计。此外，我们提出CGTNet (晶体图变换器网络)，这是一种专门用于捕捉长程相互作用的图神经网络，可显著提高性质预测的准确性和数据效率，从而增强逆向设计的可靠性。通过这些创新，T2MAT降低了对人类专业知识的依赖，加速了高性能功能材料的发现，为真正全自动的材料设计铺平了道路。

关键词:

English

T2MAT (text-to-material): A universal agent for generating material structures with goal properties from a single sentence

1.
Key Laboratory of Quantum Materials and Devices of Ministry of Education, School of Physics, Southeast University, Nanjing 211189, Jiangsu Province, China
2.
Suzhou Laboratory, Suzhou 215004, Jiangsu Province, China

Corresponding author: Email: qh.zhou@seu.edu.cn (Qionghua Zhou); jlwang@seu.edu.cn (Jinlan Wang)

Received Date: 25 August 2025
Accepted Date: 27 October 2025
Revised Date: 27 October 2025
Available Online: 15 May 2026

Abstract: Artificial Intelligence-Generated Content (AIGC)—content autonomously produced by AI systems without human intervention—has significantly boosted efficiency across various fields. However, AIGC in material science faces challenges in efficiently discovering novel materials that surpass existing databases, while ensuring the invariance and stability of crystal structures. To address these challenges, we develop T2MAT (text-to-material), an end-to-end agent that transforms user-input text into the inverse design of novel material structures with target properties beyond existing database, enabled by comprehensive exploration of chemical space and fully automated first-principles validation. Furthermore, we propose CGTNet (Crystal Graph Transformer NETwork), a graph neural network specifically designed to capture long-range interactions, which dramatically improves the accuracy and data efficiency of property predictions and thereby strengthens the reliability of inverse design. Through these contributions, T2MAT reduces the reliance on human expertise and accelerates the discovery of high-performance functional materials, paving the way for truly autonomous material design.

Key words:

1. [1]
  Y. Cao, S. Li, Y. Liu, Z. Yan, Y. Dai, P.S. Yu, L. Sun, arXiv: 2303.04226, http://arxiv.org/abs/2303.04226.
2. [2]
  Editorial, Nat. Mach. Intell. 4 (2022) 733, https://doi.org/10.1038/s42256-022-00539-8. doi: 10.1038/s42256-022-00539-8
3. [3]
  K.T. Butler, D.W. Davies, H. Cartwright, O. Isayev, A. Walsh, Nature 559 (2018) 547, https://doi.org/10.1038/s41586-018-0337-2. doi: 10.1038/s41586-018-0337-2
4. [4]
  M. Zhong, K. Tran, Y. Min, C. Wang, Z. Wang, C.T. Dinh, P. De Luna, Z. Yu, A.S. Rasouli, P. Brodersen, et al., Nature 581 (2020) 178, https://doi.org/10.1038/s41586-020-2242-8. doi: 10.1038/s41586-020-2242-8
5. [5]
  B. Weng, Z. Song, R. Zhu, Q. Yan, Q. Sun, C.G. Grice, Y. Yan, W.-J. Yin, Nat. Commun. 11 (2020) 3513, https://doi.org/10.1038/s41467-020-17263-9. doi: 10.1038/s41467-020-17263-9
6. [6]
  Z. Song, X. Wang, F. Liu, Q. Zhou, W.-J. Yin, H. Wu, W. Deng, J. Wang, Mater. Horizons 10 (2023) 1651, https://doi.org/10.1039/D3MH00157A. doi: 10.1039/D3MH00157A
7. [7]
  S. Lu, Q. Zhou, Y. Ouyang, Y. Guo, Q. Li, J. Wang, Nat. Commun. 9 (2018) 3405, https://doi.org/10.1038/s41467-018-05761-w. doi: 10.1038/s41467-018-05761-w
8. [8]
  S. Lu, Q. Zhou, Y. Guo, J. Wang, Chem 8 (2022) 769, https://doi.org/10.1016/j.chempr.2021.11.009. doi: 10.1016/j.chempr.2021.11.009
9. [9]
  B. Sanchez-Lengeling, A. Aspuru-Guzik, Science 361 (2018) 360, https://doi.org/10.1126/science.aat2663. doi: 10.1126/science.aat2663
10. [10]
  J. Noh, G.H. Gu, S. Kim, Y. Jung, Chem. Sci. 11 (2020) 4871, https://doi.org/10.1039/d0sc00594k. doi: 10.1039/d0sc00594k
11. [11]
  S. Lu, Q. Zhou, X. Chen, Z. Song, J. Wang, Natl. Sci. Rev. 9 (2022) 9, https://doi.org/10.1093/nsr/nwac111. doi: 10.1093/nsr/nwac111
12. [12]
  Y. Zhao, M. Al‐Fahdi, M. Hu, E.M.D. Siriwardane, Y. Song, A. Nasiri, J. Hu, Adv. Sci. 8 (2021) 14, https://doi.org/10.1002/advs.202100566. doi: 10.1002/advs.202100566
13. [13]
  Z. Yao, B. Sánchez-Lengeling, N.S. Bobbitt, B.J. Bucior, S.G.H. Kumar, S.P. Collins, T. Burns, T.K. Woo, O.K. Farha, R.Q. Snurr, et al., Nat. Mach. Intell. 3 (2021) 76, https://doi.org/10.1038/s42256-020-00271-1. doi: 10.1038/s42256-020-00271-1
14. [14]
  B. Kim, S. Lee, J. Kim, Sci. Adv. 6 (2020) eaax9324, https://doi.org/10.1126/sciadv.aax9324. doi: 10.1126/sciadv.aax9324
15. [15]
  T. Salimans, I. Goodfellow, W. Zaremba, V. Cheung, A. Radford, X. Chen, Adv. Neural Inf. Process. Syst. (2016) 2234, https://dl.acm.org/doi/10.5555/3157096.3157346 doi: 10.5555/3157096.3157346
16. [16]
  K.M. Jablonka, P. Schwaller, A. Ortega-Guerrero, B. Smit, Nat. Mach. Intell. 6 (2024) 161, https://doi.org/10.1038/s42256-023-00788-1. doi: 10.1038/s42256-023-00788-1
17. [17]
  L.M. Antunes, K.T. Butler, R. Grau-Crespo, Nat. Coummn. (2024) 10570, https://doi.org/10.1038/s41467-024-54639-7. doi: 10.1038/s41467-024-54639-7
18. [18]
  T. Xie, J.C. Grossman, Phys. Rev. Lett. 120 (2018) 145301, https://doi.org/10.1103/PhysRevLett.120.145301. doi: 10.1103/PhysRevLett.120.145301
19. [19]
  J. Gasteiger, J. Groß, S. Günnemann, arXiv: 2003.03123, https://arxiv.org/abs/2003.03123.
20. [20]
  K.T. Schütt, H.E. Sauceda, P.-J. Kindermans, A. Tkatchenko, K.-R. Müller, J. Chem. Phys. 148 (2018) 241722, https://doi.org/10.1063/1.5019779. doi: 10.1063/1.5019779
21. [21]
  M. Shuaibi, A. Kolluru, A. Das, A. Grover, A. Sriram, Z. Ulissi, C.L. Zitnick, arXiv: 2106.09575, http://arxiv.org/abs/2106.09575.
22. [22]
  J. Gasteiger, F. Becker, S. Günnemann, GemNet: Universal Directional Graph Neural Networks for Molecules, in Advances in Neural Information Processing Systems 34 (NeurIPS 2021), Curran Associates, Inc., 2021, pp. 6790–6802, https://proceedings.neurips.cc/paper/2021/hash/35cf8659cfcb13224cbd47863a34fc58-Abstract.html.
23. [23]
  K.T. Schütt, O.T. Unke, M. Gastegger, Equivariant Message Passing for the Prediction of Tensorial Properties and Molecular Spectra, in Proceedings of the 38th International Conference on Machine Learning (ICML 2021), PMLR, 2021, pp. 9377–9388, https://proceedings.mlr.press/v139/schutt21a.html.
24. [24]
  Y.-L. Liao, B. Wood, A. Das, T. Smidt, EquiformerV2: Improved Equivariant Transformer for Scaling to Higher-Degree Representations, in International Conference on Learning Representations (ICLR), 2024, https://proceedings.iclr.cc/paper_files/paper/2024/file/ab12e8f3443c1a789f595b18d8c597b4-Paper-Conference.pdf.
25. [25]
  M. Fey, J.E. Lenssen, Fast Graph Representation Learning with PyTorch Geometric, in ICLR Workshop on Representation Learning on Graphs and Manifolds, 2019, https://rlgm.github.io/papers/2.pdf.
26. [26]
  Z. Song, S. Lu, M. Ju, Q. Zhou, J. Wang, Nat. Commun. 16 (2025) 6530, https://doi.org/10.1038/s41467-025-61778-y. doi: 10.1038/s41467-025-61778-y
27. [27]
  D. Guo, D. Yang, H. Zhang, J. Song, R. Zhang, R. Xu, Q. Zhu, S. Ma, P. Wang, X. Bi, et al., Nature 645 (2025) 633, https://doi.org/10.1038/s41586-025-09422-z. doi: 10.1038/s41586-025-09422-z
28. [28]
  H. Touvron, L. Martin, K. Stone, P. Albert, A. Almahairi, Y. Babaei, N. Bashlykov, S. Batra, P. Bhargava, S. Bhosale, et al., arXiv: 2307.09288, http://arxiv.org/abs/2307.09288
29. [29]
  Z. Du, Y. Qian, X. Liu, M. Ding, J. Qiu, Z. Yang, J. Tang, Proc. Annu. Meet. Assoc. Comput. Linguist. 1 (2022) 320, https://doi.org/10.18653/v1/2022.acl-long.26. doi: 10.18653/v1/2022.acl-long.26
30. [30]
  T.B. Brown, B. Mann, N. Ryder, M. Subbiah, J. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askell, et al., Language Models are Few-Shot Learners, in Advances in Neural Information Processing Systems 33 (NeurIPS 2020), Curran Associates, Inc., 2020, pp. 1877–1901, https://papers.nips.cc/paper/2020/hash/1457c0d6bfcb4967418bfb8ac142f64a-Abstract.html
31. [31]
  S.P. Ong, W.D. Richards, A. Jain, G. Hautier, M. Kocher, S. Cholia, D. Gunter, V.L. Chevrier, K.A. Persson, G. Ceder, Comput. Mater. Sci. 68 (2013) 314, https://doi.org/10.1016/j.commatsci.2012.10.028. doi: 10.1016/j.commatsci.2012.10.028
32. [32]
  S. Haastrup, M. Strange, M. Pandey, T. Deilmann, P.S. Schmidt, N.F. Hinsche, M.N. Gjerding, D. Torelli, P.M. Larsen, A.C. Riis-Jensen, et al., 2D Mater. 5 (2018) 042002, https://doi.org/10.1088/2053-1583/aacfc1. doi: 10.1088/2053-1583/aacfc1
33. [33]
  M.N. Gjerding, A. Taghizadeh, A. Rasmussen, S. Ali, F. Bertoldo, T. Deilmann, N.R. Knøsgaard, M. Kruse, A.H. Larsen, S. Manti, et al., 2D Mater. 8 (2021) 044002, https://doi.org/10.1088/2053-1583/ac1059. doi: 10.1088/2053-1583/ac1059
34. [34]
  A.S. Rosen, S.M. Iyer, D. Ray, Z. Yao, A. Aspuru-Guzik, L. Gagliardi, J.M. Notestein, R.Q. Snurr, Matter 4 (2021) 1578, https://doi.org/10.1016/j.matt.2021.02.015. doi: 10.1016/j.matt.2021.02.015
35. [35]
  A.S. Rosen, V. Fung, P. Huck, C.T. O'Donnell, M.K. Horton, D.G. Truhlar, K.A. Persson, J.M. Notestein, R.Q. Snurr, npj Comput. Mater. 8 (2022) 112, https://doi.org/10.1038/s41524-022-00796-6. doi: 10.1038/s41524-022-00796-6
36. [36]
  T. Xie, X. Fu, O.-E. Ganea, R. Barzilay, T. Jaakkola, arXiv: 2110.06197, https://arxiv.org/abs/2110.06197.
37. [37]
  R. Jiao, W. Huang, Y. Liu, D. Zhao, Y. Liu, arXiv: 2402.03992, https://arxiv.org/abs/2402.03992.
38. [38]
  B.K. Miller, R.T.Q. Chen, A. Sriram, B.M. Wood. FlowMM: Generating Materials with Riemannian Flow Matching, in Proceedings of the 41st International Conference on Machine Learning, PMLR, 2024, pp. 35664–35686, https://proceedings.mlr.press/v235/miller24a.html.
39. [39]
  C. Zeni, R. Pinsler, D. Zügner, A. Fowler, M. Horton, X. Fu, Z. Wang, A. Shysheya, J. Crabbé, S. Ueda, et al., Nature 639 (2025) 624, https://doi.org/10.1038/s41586-025-08628-5. doi: 10.1038/s41586-025-08628-5
40. [40]
  Z. Song, C. Ling, Q. Li, Q. Zhou, J. Wang, arXiv: 2507.19307, http://arxiv.org/abs/2507.19307.
41. [41]
  A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A.N. Gomez, L. Kaiser, I. Polosukhin. Attention is All you Need, in Advances in Neural Information Processing Systems 30 (NIPS 2017), Curran Associates, Inc., 2017, https://papers.neurips.cc/paper/7181-attention-is-all-you-need.pdf.
42. [42]
  J. Devlin, M.-W. Chang, K. Lee, K. Toutanova, BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding, in Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Association for Computational Linguistics: Minneapolis, Minnesota, 2019, pp. 4171–4186, https://aclanthology.org/N19-1423/.
43. [43]
  H. Huo, M. Rupp, Mach. Learn. Sci. Technol. 3 (2022) 045017, https://doi.org/10.1088/2632-2153/aca005. doi: 10.1088/2632-2153/aca005
44. [44]
  L. Yu, A. Zunger, Phys. Rev. Lett. 108 (2012) 068701, https://doi.org/10.1103/PhysRevLett.108.068701. doi: 10.1103/PhysRevLett.108.068701
45. [45]
  S. Kong, F. Ricci, D. Guevarra, J. B. Neaton, C. P. Gomes, J. M. Gregoire, Nat. Commun. 13 (2022) 949, https://doi.org/10.1038/s41467-022-28543-x. doi: 10.1038/s41467-022-28543-x
46. [46]
  C. Loh, T. Christensen, R. Dangovski, S. Kim, M. Soljačić, Nat. Commun. 13 (2022) 4223, https://doi.org/10.1038/s41467-022-31915-y. doi: 10.1038/s41467-022-31915-y
47. [47]
  S. Kirklin, J.E. Saal, B. Meredig, A. Thompson, J.W. Doak, M. Aykol, S. Rühl, C. Wolverton, npj Comput. Mater. 1 (2015) 15010, https://doi.org/10.1038/npjcompumats.2015.10. doi: 10.1038/npjcompumats.2015.10
48. [48]
  K. Choudhary, K.F. Garrity, A.C.E. Reid, B. DeCost, A.J. Biacchi, A.R. Hight Walker, Z. Trautt, J. Hattrick-Simpers, A.G. Kusne, A. Centrone, et al., npj Comput. Mater. 6 (2020) 173, https://doi.org/10.1038/s41524-020-00440-1. doi: 10.1038/s41524-020-00440-1
49. [49]
  Z. Yang, D. Bourgeois, J. You, M. Zitnik, J. Leskovec, GNNExplainer: Generating Explanations for Graph Neural Networks, in Advances in Neural Information Processing Systems 32 (NuerIPS 2019), Curran Associates, Inc., 2019, https://proceedings.neurips.cc/paper_files/paper/2019/file/d80b7040b773199015de6d3b4293c8ff-Paper.pdf.
50. [50]
  W.-J. Yin, T. Shi, Y. Yan, Adv. Mater. 26 (2014) 4653, https://doi.org/10.1002/adma.201306281. doi: 10.1002/adma.201306281
51. [51]
  W.-J. J. Yin, T. Shi, Y. Yan, Appl. Phys. Lett. 104 (2014) 063903, https://doi.org/10.1063/1.4864778. doi: 10.1063/1.4864778
52. [52]
  S.M. Lundberg, S.I. Lee, A Unified Approach to Interpreting Model Predictions, in Advances in Neural Information Processing Systems 30 (NIPS 2017), Curran Associates, Inc., 2017, https://proceedings.neurips.cc/paper/7062-a-unified-approach-to-interpreting-model-predictions.pdf.
53. [53]
  K. Momma, F. Izumi, J. Appl. Crystallogr. 44 (2011) 1272, https://doi.org/10.1107/S0021889811038970. doi: 10.1107/S0021889811038970
54. [54]
  G. Kresse, J. Furthmüller, Phys. Rev. B 54 (1996) 11169, https://doi.org/10.1103/PhysRevB.54.11169. doi: 10.1103/PhysRevB.54.11169
55. [55]
  P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, G. L. Chiarotti, M. Cococcioni, I. Dabo, et al., J. Phys. Condens. Matter 21 (2009) 395502, https://doi.org/10.1088/0953-8984/21/39/395502. doi: 10.1088/0953-8984/21/39/395502

扫一扫看文章

计量

PDF下载量: 1
文章访问数: 592
HTML全文浏览量: 44

通讯作者: 陈斌, bchen63@163.com

1.
沈阳化工大学材料科学与工程学院沈阳 110142

T2MAT (文本到材料)：基于文本生成目标性能材料结构的通用智能体

通讯作者: Email: qh.zhou@seu.edu.cn (周跫桦); jlwang@seu.edu.cn (王金兰)

English

T2MAT (text-to-material): A universal agent for generating material structures with goal properties from a single sentence

Corresponding author: Email: qh.zhou@seu.edu.cn (Qionghua Zhou); jlwang@seu.edu.cn (Jinlan Wang)

CCS Chemistry：嵌段共聚物自组装成 “三明治”二维有机材料，实现纳米级压力传感功能

CCS Chemistry：颜徐州、鲍哲南：你的皮肤可能是一片SPMs

CCS Chemistry：工作高效不疲劳，只须让催化剂动起来

CCS Chemistry：静电组装导向聚合- 聚电解质纳米凝胶量化制备新策略

CCS Chemistry：金黄色葡萄球菌醛基脱氢酶是如何识别特异性底物的？

计量

通讯作者: 陈斌, bchen63@163.com

中国化学会秘书处

T2MAT (文本到材料)：基于文本生成目标性能材料结构的通用智能体

通讯作者: Email: qh.zhou@seu.edu.cn (周跫桦); jlwang@seu.edu.cn (王金兰)

English

T2MAT (text-to-material): A universal agent for generating material structures with goal properties from a single sentence

Corresponding author: Email: qh.zhou@seu.edu.cn (Qionghua Zhou); jlwang@seu.edu.cn (Jinlan Wang)

CCS Chemistry：嵌段共聚物自组装成 “三明治”二维有机材料，实现纳米级压力传感功能

CCS Chemistry：颜徐州、鲍哲南： 你的皮肤可能是一片SPMs

CCS Chemistry：工作高效不疲劳，只须让催化剂动起来

CCS Chemistry：静电组装导向聚合- 聚电解质纳米凝胶量化制备新策略

CCS Chemistry：金黄色葡萄球菌醛基脱氢酶是如何识别特异性底物的？

计量

文章相关

通讯作者: 陈斌, bchen63@163.com

中国化学会秘书处

CCS Chemistry：颜徐州、鲍哲南：你的皮肤可能是一片SPMs