Artificial intelligence for accelerating polymer design: recent advances and future perspectives

doi:10.11949/0438-1157.20221077

Abstract

Abstract:

Due to the enormous chemical and configurational space, the optimal design of candidates for the next-generation soft materials is still a challenging task. It is cumbersome to conduct trial-and-error research using high-throughput computations or experiments to evaluate the properties of a large number of materials and select the best candidates for future investigations. Using artificial intelligence approaches in combination with computer simulations and experiments, researchers are able to reliably predict properties of materials over a vast structural and property space, breaking the traditional model of “empirically guided experiments” and gradually overcoming various bottlenecks in the process of the polymer design. This review begins with a historical look at the difficulties in polymer engineering over the preceding decades. The concept of data-driven techniques is then given and examined in detail, along with how they are used in polymer design. The following section highlights some noteworthy developments in identifying novel polymers with specific characteristics using data-driven approaches. In conclusion, this review provides a synopsis of recent tendencies and outlines the opportunities for intelligent design in polymer engineering. Artificial intelligence, rapid computational simulation, and the availability of enormous amounts of open-source homogeneous data combined with experiments will revolutionize polymer research and accelerate the industrial application of designed polymeric materials. Finally, the current industry development trend is summarized, and the large-scale application prospects of intelligent design in the research of new polymers are prospected.

Key words: artificial intelligence, polymer design, data-driven, structure-property relationship, inverse design

摘要：

广阔的化学空间蕴藏着近乎无限的可能，高性能聚合物材料的设计至今仍是一项充满挑战的工作。利用实验或高通量计算广泛探索大量样本，选择其中性能较好的候选材料进行深入研究的传统试错方式，越发难以在庞大的组合序列中筛选出满足实际多项性能需求的新材料。由先进计算方法、自动化实验和大数据技术耦合形成的新型智能研发模式，突破了“经验指导实验”的传统思路，有望实现在广阔的结构与性质空间中进行聚合物材料的特性预测，逐渐成为克服各类瓶颈问题的得力助手，大大提高了高性能聚合物设计的效率。本文回顾了以往聚合物设计的困境，讨论了人工智能方法的一般思路及其在聚合物设计中的工作原理，列举了采用智能方法在高性能聚合物研发工作中取得突破性进展的典型案例，最后对当前行业发展趋势进行总结，展望了智能设计在新型聚合物研究中的规模化应用前景。

关键词: 人工智能, 聚合物设计, 数据驱动, 结构-性能关系, 反求设计

CLC Number:

TQ 31

Tianhang ZHOU, Xingying LAN, Chunming XU. Artificial intelligence for accelerating polymer design: recent advances and future perspectives[J]. CIESC Journal, 2023, 74(1): 14-28.

周天航, 蓝兴英, 徐春明. 人工智能加速聚合物设计的最新进展和未来前景[J]. 化工学报, 2023, 74(1): 14-28.

Figures/Tables 11

References 78

1	Allcock H R. Rational design and synthesis of new polymeric material[J]. Science, 1992, 255(5048): 1106-1112.
2	Le T, Epa V C, Burden F R, et al. Quantitative structure-property relationship modeling of diverse materials properties[J]. Chemical Reviews, 2012, 112(5): 2889-2919.
3	Bereau T, Andrienko D, Kremer K. Research update: computational materials discovery in soft matter[J]. APL Materials, 2016, 4(5): 053101.
4	Patra T K. Data-driven methods for accelerating polymer design[J]. ACS Polymers Au, 2022, 2(1): 8-26.
5	Sattari K, Xie Y, Lin J. Data-driven algorithms for inverse design of polymers[J]. Soft Matter, 2021, 17(33): 7607-7622.
6	Zhu M X, Deng T, Dong L, et al. Review of machine learning-driven design of polymer-based dielectrics[J]. IET Nanodielectrics, 2022, 5(1): 24-38.
7	Kumar J, Li Q, Ye J. Challenges and opportunities of polymer design with machine learning and high throughput experimentation[J]. MRS Communications, 2019, 9(2): 1-8.
8	Zhou T, Wu Z, Chilukoti H K, et al. Sequence-engineering polyethylene-polypropylene copolymers with high thermal conductivity using a molecular-dynamics-based genetic algorithm[J]. Journal of Chemical Theory and Computation, 2021, 17(6): 3772-3782.
9	DeStefano A J, Segalman R A, Davidson E C. Where biology and traditional polymers meet: the potential of associating sequence-defined polymers for materials science[J]. JACS Au, 2021, 1(10):1556-1571.
10	Perry S L, Sing C E. 100th anniversary of macromolecular science viewpoint: opportunities in the physics of sequence-defined polymers[J]. ACS Macro Letters, 2020, 9(2): 216-225.
11	Khokhlov A R, Khalatur P G. Conformation-dependent sequence design (engineering) of AB copolymers[J]. Physical Review Letters, 1999, 82(17): 3456-3459.
12	Zhou T, Song Z, Sundmacher K. Big data creates new opportunities for materials research: a review on methods and applications of machine learning for materials design[J]. Engineering, 2019, 5(6): 1017-1026.
13	Webb M A, Jackson N E, Gil P S, et al. Targeted sequence design within the coarse-grained polymer genome[J]. Science Advances, 2020, 6(43): eabc6216.
14	Meenakshisundaram V, Hung J H, Patra T K, et al. Designing sequence-specific copolymer compatibilizers using a molecular-dynamics-simulation-based genetic algorithm[J]. Macromolecules, 2017, 50(3): 1155-1166.
15	Nguyen D, Tao L, Li Y. Integration of machine learning and coarse-grained molecular simulations for polymer materials: physical understandings and molecular design[J]. Frontiers in Chemistry, 2022, 9: 820417.
16	Bernazzani L, Duce C, Micheli A, et al. Predicting physical-chemical properties of compounds from molecular structures by recursive neural networks[J]. Journal of Chemical Information and Modeling, 2006, 46(5): 2030-2042.
17	Miccio L A, Schwartz G A. From chemical structure to quantitative polymer properties prediction through convolutional neural networks[J]. Polymer, 2020, 193: 122341.
18	Zhou T, Schneider J, Wu Z, et al. Compatibilization efficiency of additives in homopolymer blends: a dissipative particle dynamics study[J]. Macromolecules, 2021, 54(20): 9551-9564.
19	Eastwood E A, Dadmun M D. Multiblock copolymers in the compatibilization of polystyrene and poly(methyl methacrylate) blends: role of polymer architecture[J]. Macromolecules, 2002, 35(13): 5069-5077.
20	Chen G, Shen Z, Iyer A, et al. Machine-learning-assisted de novo design of organic molecules and polymers: opportunities and challenges[J]. Polymers, 2020, 12(1): 163.
21	Bowman A L, Mun S, Nouranian S, et al. Free volume and internal structural evolution during creep in model amorphous polyethylene by molecular dynamics simulations[J]. Polymer, 2019, 170: 85-100.
22	Choy C L, Wong Y W, Yang G W, et al. Elastic modulus and thermal conductivity of ultradrawn polyethylene[J]. Journal of Polymer Science Part B: Polymer Physics, 1999, 37(23): 3359-3367.
23	Blaber S, Mahmoudi P, Spencer R K W, et al. Effect of chain stiffness on the entropic segregation of chain ends to the surface of a polymer melt[J]. The Journal of Chemical Physics, 2019, 150(1): 014904.
24	Agrawal A, Choudhary A. Perspective: materials informatics and big data: realization of the “fourth paradigm” of science in materials science[J]. APL Materials, 2016, 4(5): 053208.
25	Behler J, Parrinello M. Generalized neural-network representation of high-dimensional potential-energy surfaces[J]. Physical Review Letters, 2007, 98(14): 146401.
26	Jackson N E, Webb M A, de Pablo J J. Recent advances in machine learning towards multiscale soft materials design[J]. Current Opinion in Chemical Engineering, 2019, 23: 106-114.
27	Mannodi-Kanakkithodi A, Chandrasekaran A, Kim C, et al. Scoping the polymer genome: a roadmap for rational polymer dielectrics design and beyond[J]. Materials Today, 2018, 21(7): 785-796.
28	Kresse G, Furthmüller J. Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set[J]. Computational Materials Science, 1996, 6(1): 15-50.
29	Kresse G. Ab initio molecular dynamics for liquid metals[J]. Journal of Non-Crystalline Solids, 1995, 192/193: 222-229.
30	Giannozzi P, Baroni S, Bonini N, et al. QUANTUM ESPRESSO: a modular and open-source software project for quantum simulations of materials[J]. Journal of Physics. Condensed Matter: an Institute of Physics Journal, 2009, 21(39): 395502.
31	Giannozzi P, Baseggio O, Bonfà P, et al. QUANTUM ESPRESSO toward the exascale[J]. The Journal of Chemical Physics, 2020, 152(15): 154105.
32	Gaussian, Inc. Expanding the Limits of Computational Chemistry[EB/OL]. [2022-07-10]. .
33	Plimpton S. Fast parallel algorithms for short-range molecular dynamics[J]. Journal of Computational Physics, 1995, 117(1): 1-19.
34	Thompson A P, Aktulga H M, Berger R, et al. LAMMPS—a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales[J]. Computer Physics Communications, 2022, 271: 108171.
35	van der Spoel D, Lindahl E, Hess B, et al. GROMACS: fast, flexible, and free[J]. Journal of Computational Chemistry, 2005, 26(16): 1701-1718.
36	Hess B, Kutzner C, van der Spoel D, et al. GROMACS 4: algorithms for highly efficient, load-balanced, and scalable molecular simulation[J]. Journal of Chemical Theory and Computation, 2008, 4(3): 435-447.
37	Phillips J C, Hardy D J, Maia J D C, et al. Scalable molecular dynamics on CPU and GPU architectures with NAMD[J]. The Journal of Chemical Physics, 2020, 153(4): 044130.
38	Smith W, Yong C W, Rodger P M. DL_POLY: application to molecular simulation[J]. Molecular Simulation, 2002, 28(5): 385-471.
39	Karimi-Varzaneh H A, Qian H J, Chen X, et al. IBIsCO: a molecular dynamics simulation package for coarse-grained simulation[J]. Journal of Computational Chemistry, 2011, 32(7): 1475-1487.
40	Pilania G, Wang C, Jiang X, et al. Accelerating materials property predictions using machine learning[J]. Scientific Reports, 2013, 3: 2810.
41	SMILES Weininger D., a chemical language and information system (1): Introduction to methodology and encoding rules[J]. Journal of Chemical Information and Computer Sciences, 1988, 28(1): 31-36.
42	Heller S R, McNaught A, Pletnev I, et al. InChI, the IUPAC international chemical identifier[J]. Journal of Cheminformatics, 2015, 7: 23.
43	Krenn M, Häse F, Nigam A, et al. Self-referencing embedded strings (SELFIES): a 100% robust molecular string representation[J]. Machine Learning: Science and Technology, 2020, 1: 045024.
44	Daylight Chemical Information Systems, Inc. Reaction SMILES and SMIRKS[EB/OL]. [2022-07-11]. .
45	Grethe G, Goodman J M, Allen C H. International chemical identifier for reactions (RInChI)[J]. Journal of Cheminformatics, 2018, 10(1): 45.
46	Vu K, Snyder J C, Li L, et al. Understanding kernel ridge regression: common behaviors from simple functions to density functionals[J]. International Journal of Quantum Chemistry, 2015, 115(16): 1115-1128.
47	Balachandran P V, Xue D, Theiler J, et al. Adaptive strategies for materials design using uncertainties[J]. Scientific Reports, 2016, 6: 19660.
48	Vasudevan R, Pilania G, Balachandran P V. Machine learning for materials design and discovery[J]. Journal of Applied Physics, 2021, 129: 070401.
49	Pedregosa F, Varoquaux G, Gramfort A, et al. Scikit-learn: machine learning in Python[J]. The Journal of Machine Learning Research, 2011, 12: 2825-2830.
50	Learn Scikit. Scikit-learn 1.1.1 Documentation[EB/OL]. [2022-07-11]. .
51	Keras. The Python Deep Learning API[EB/OL]. [2022-07-11]. .
52	PyTorch[EB/OL]. [2022-07-11]. .
53	Koros W J, Zhang C. Materials for next-generation molecularly selective synthetic membranes[J]. Nature Materials, 2017, 16(3): 289-297.
54	Baker R W, Low B T. Gas separation membrane materials: a perspective[J]. Macromolecules, 2014, 47(20): 6999-7013.
55	Robeson L M. The upper bound revisited[J]. Journal of Membrane Science, 2008, 320(1/2): 390-400.
56	Robeson L M, Liu Q, Freeman B D, et al. Comparison of transport properties of rubbery and glassy polymers and the relevance to the upper bound relationship[J]. Journal of Membrane Science, 2015, 476: 421-431.
57	Park H B, Kamcev J, Robeson L M, et al. Maximizing the right stuff: the trade-off between membrane permeability and selectivity[J]. Science, 2017, 356(6343): eaab0530.
58	Barnett J W, Bilchak C R, Wang Y W, et al. Designing exceptional gas-separation polymer membranes using machine learning[J]. Science Advances, 2020, 6(20): eaaz4301.
59	Ho J S, Greenbaum S G. Polymer capacitor dielectrics for high temperature applications[J]. ACS Applied Materials & Interfaces, 2018, 10(35): 29189-29218.
60	Tan D, Zhang L L, Chen Q, et al. High-temperature capacitor polymer films[J]. Journal of Electronic Materials, 2014, 43(12): 4569-4575.
61	Zhou Y, Li Q, Dang B, et al. A scalable, high-throughput, and environmentally benign approach to polymer dielectrics exhibiting significantly improved capacitive performance at high temperatures[J]. Advanced Materials, 2018, 30(49): 1805672.
62	Wu C, Deshmukh A A, Li Z, et al. Flexible temperature-invariant polymer dielectrics with large bandgap[J]. Advanced Materials, 2020, 32(21): 2000499.
63	Kim C, Pilania G, Ramprasad R. From organized high-throughput data to phenomenological theory using machine learning: the example of dielectric breakdown[J]. Chemistry of Materials, 2016, 28(5): 1304-1311.
64	Batra R, Dai H, Huan T D, et al. Polymers for extreme conditions designed using syntax-directed variational autoencoders[J]. Chemistry of Materials, 2020, 32(24): 10489-10500.
65	Majewski P W, Gopinadhan M, Jang W S, et al. Anisotropic ionic conductivity in block copolymer membranes by magnetic field alignment[J]. Journal of the American Chemical Society, 2010, 132(49): 17516-17522.
66	Weber R L, Ye Y, Schmitt A L, et al. Effect of nanoscale morphology on the conductivity of polymerized ionic liquid block copolymers[J]. Macromolecules, 2011, 44(14): 5727-5735.
67	Zhang S, Lee K H, Frisbie C, et al. Ionic conductivity, capacitance, and viscoelastic properties of block copolymer-based ion gels[J]. Macromolecules, 2011, 44: 940-949.
68	Guan T, Qian S, Guo Y, et al. Star brush block copolymer electrolytes with high ambient-temperature ionic conductivity for quasi-solid-state lithium batteries[J]. ACS Materials Letters, 2019, 1(6): 606-612.
69	Han Z, Fina A. Thermal conductivity of carbon nanotubes and their polymer nanocomposites: a review[J]. Progress in Polymer Science, 2011, 36(7): 914-944.
70	Yan X R, Gu J W, Zheng G Q, et al. Lowly loaded carbon nanotubes induced high electrical conductivity and giant magnetoresistance in ethylene/1-octene copolymers[J]. Polymer, 2016, 103: 315-327.
71	Ghahramani N, Esfahani S A S, Mehranpour M, et al. The effect of filler localization on morphology and thermal conductivity of the polyamide/cyclic olefin copolymer blends filled with boron nitride[J]. Journal of Materials Science, 2018, 53(23): 16146-16159.
72	Zhang W F, Li H X, Jiang H Y, et al. Influence of surface defects on the thermal conductivity of hexagonal boron nitride/poly(dimethylsiloxane) nanocomposites: a molecular dynamics simulation[J]. Langmuir, 2021, 37(41): 12038-12048.
73	Gao Y Y, Müller-Plathe F. Molecular dynamics study on the thermal conductivity of the end-grafted carbon nanotubes filled polyamide-6.6 nanocomposites[J]. The Journal of Physical Chemistry C, 2018, 122(2): 1412-1421.
74	Gao Y Y, Müller-Plathe F. Effect of grafted chains on the heat transfer between carbon nanotubes in a polyamide-6.6 matrix: a molecular dynamics study[J]. Polymer, 2017, 129: 228-234.
75	Wei X F, Luo T F. The effect of the block ratio on the thermal conductivity of amorphous polyethylene-polypropylene(PE-PP) diblock copolymers[J]. Physical Chemistry Chemical Physics: PCCP, 2018, 20(31): 20534-20539.
76	Li H, Collins C R, Ribelli T G, et al. Tuning the molecular weight distribution from atom transfer radical polymerization using deep reinforcement learning[J]. Molecular Systems Design & Engineering, 2018, 3(3): 496-508.
77	Liu H, Xue Y H, Zhu Y L, et al. Inverse design of molecular weight distribution in controlled polymerization via a one-pot reaction strategy[J]. Macromolecules, 2020, 53(15): 6409-6419.
78	Li J L, Lim K, Yang H T, et al. AI applications through the whole life cycle of material discovery[J]. Matter, 2020, 3(2): 393-432.

计算方法	可用软件	可计算特性
基于密度泛函理论（DFT）的第一性原理计算	VASP^[28-29]，QUANTUM ESPRESSO^[30-31]和GAUSSIAN^[32]	带隙、介电常数、折射率、吸附能、反应活化能等
经典分子动力学模拟（MD）和粗粒化（CG）模拟	LAMMPS^[33-34]，GROMACS^[35-36]，NAMD^[37]，DL_POLY^[38]和IBISCO^[39]	结构因子、回旋半径、相平衡、黏度、热导率、表面张力、离子扩散、气体渗透等

计算方法	可用软件	可计算特性
基于密度泛函理论（DFT）的第一性原理计算	VASP^[28-29]，QUANTUM ESPRESSO^[30-31]和GAUSSIAN^[32]	带隙、介电常数、折射率、吸附能、反应活化能等
经典分子动力学模拟（MD）和粗粒化（CG）模拟	LAMMPS^[33-34]，GROMACS^[35-36]，NAMD^[37]，DL_POLY^[38]和IBISCO^[39]	结构因子、回旋半径、相平衡、黏度、热导率、表面张力、离子扩散、气体渗透等

聚合物信息	方法	原理
分子组成和结构	简化分子输入线性输入系统（SMILES）^[41]、IUPAC国际化学标识符（InChIs）^[42]、自引用嵌入字符串（SELFIES）^[43]	分子信息被转换成特征向量或张量
化学反应	反应SMILESs^[44]、反应InChI （RInChI）^[45]、反应SMIRKS^[44]	基于描述分子组成的方法，从反应物的分子描述符（如指纹）开始，对其进行求和以及加减等操作
序列的特征化	One-hot encoding（OHE）^[17]	将聚合物的序列转换为数字表示，如将有两个单体的共聚物映射成一个“0”和“1”的二进制数字

聚合物信息	方法	原理
分子组成和结构	简化分子输入线性输入系统（SMILES）^[41]、IUPAC国际化学标识符（InChIs）^[42]、自引用嵌入字符串（SELFIES）^[43]	分子信息被转换成特征向量或张量
化学反应	反应SMILESs^[44]、反应InChI （RInChI）^[45]、反应SMIRKS^[44]	基于描述分子组成的方法，从反应物的分子描述符（如指纹）开始，对其进行求和以及加减等操作
序列的特征化	One-hot encoding（OHE）^[17]	将聚合物的序列转换为数字表示，如将有两个单体的共聚物映射成一个“0”和“1”的二进制数字

算法	优势	劣势
核脊回归（KRR）、支持向量机（SVM）	计算成本低	通常不适用于大数据集
高斯过程回归（GPR）	可以很好地预测目标值的不确定性	不具备在单个模型中处理多参数问题的能力
K-邻近算法（K-nearest neighbors）	直观和简单	所分类别的数量必须人为定义且对性质不均一的数据库效果不佳
朴素贝叶斯（naive Bayes, NB）	直观和简单，适用于大型数据集	假设所有特征是独立的，因此只适用于较简单的分类问题
随机森林（RF）、梯度提升（gradient boosting, GB）	适用于大型数据集，并且可提高每个描述符的重要性，有较强的可解释性	在相关的超参优化中，可能会创建过于复杂的决策树结构而导致过度拟合
人工神经网络（ANN）	从大规模数据集中捕捉非线性复杂关系的能力很强	需要较多的训练数据且缺乏可解释性