检索问题
机器学习辅助钙钛矿材料稳定性提升的研究进展
机器学习辅助钙钛矿材料稳定性提升的研究进展
钙钛矿太阳能电池(PSCs)因其高光电转换效率和低成本制造潜力,被认为是下一代光伏技术的有力竞争者 。然而,其长期稳定性问题一直是阻碍其商业化应用的关键瓶颈 。传统的试错法在PSC的筛选、开发和稳定性测试方面速度慢且劳动密集 。近年来,人工智能(AI)和机器学习(ML)技术在材料科学领域迅速发展,为解决这一难题提供了新的工具和方法,在提升钙钛矿材料的稳定性方面展现出独特的优势 。
1. 稳定性预测与筛选
机器学习在钙钛矿材料稳定性研究中的一个核心应用是预测和筛选具有高稳定性的新材料。通过建立机器学习模型,研究人员可以快速评估大量候选材料的稳定性,从而避免耗时耗力的实验尝试。
- 数据驱动的预测模型:机器学习模型能够基于现有数据进行快速预测,而不是通过直接实验或求解基本方程的计算/模拟 。这些模型通常通过将材料数值化(即“指纹化”或“描述符”)来工作,然后通过学习算法建立指纹与目标性质(如稳定性)之间的映射关系 。例如,有研究通过机器学习方法构建了统计模型,能够预测新型ABO3钙钛矿化合物及其晶体结构的可能性,其中考虑了A位和B位香农离子半径、容忍因子和八面体因子等几何描述符 。
- 多因素影响分析:钙钛矿材料的稳定性受多种环境和内部因素影响。机器学习方法能够处理这些复杂的多变量体系。一项研究引入了一种机器学习辅助方法,使用多头注意力机制同时处理包括外部和内部参数在内的多样化输入数据,并结合挤压-激励残差网络(SEResNet),实现了对PSC稳定性预测的高精度,决定系数(R2)达到0.972,皮尔逊相关系数(r)达到0.986 。此外,该研究还利用SHapley Additive exPlanations(SHAP)算法识别了影响稳定性的关键因素,并通过高通量预测探索了这些关键因素的相互作用,从而全面理解它们对器件稳定性的影响 。
- 关键因素识别:机器学习有助于揭示影响效率和稳定性的关键因素。例如,一项研究通过机器学习揭示了晶粒尺寸、缺陷密度、带隙、荧光寿命和表面粗糙度这五个因素对PSC效率和稳定性的影响。结果表明,带隙对效率影响最大,而表面粗糙度和晶粒尺寸对长期稳定性影响最大 。这项工作还提供了一个基于荧光寿命和带隙预测效率的数学模型,并指导实验获得了具有23.4%效率和优异长期稳定性的PSC,在环境条件下老化3288小时后仍保持97.6%的初始效率 。
- 多目标优化:除了预测稳定性,机器学习还可以用于多目标优化,同时考虑材料的其他重要性能。例如,有研究提出了一个多步多阶段筛选方案,结合机器学习和高通量计算,从大量候选材料中加速发现具有优异效率和热稳定性的无铅杂化有机-无机钙钛矿A2BB'X6 。通过一系列筛选,建立了A2BB'X6性质的结构-性能关系,并成功筛选出四个具有良好稳定性、高德拜温度和合适带隙的实验可行候选材料 。
2. 结构与组分工程
机器学习在钙钛矿材料的结构和组分设计方面也取得了显著进展,帮助研究人员发现新的、更稳定的化合物。
- 晶体结构预测:低维杂化有机-无机钙钛矿(HOIPs)具有巨大的设计空间,通过结合不同有机阳离子和无机骨架可以调控其电子和机械性能 。机器学习可以用于加速这些材料的晶体结构预测。一项工作提出了一种准确、高效且广泛适用的机器学习原子间势(MLIP),该MLIP在86种多样化的实验报道HOIP材料上进行训练,并在73种实验报道的钙钛矿组分上进行了测试,相对于密度泛函理论(DFT)实现了高精度 。结合新颖的随机结构搜索算法,该方法仅通过指定有机分子和无机阳离子/卤化物就能可靠地恢复14种已知二维钙钛矿的晶体结构 。这种方法预测了一种全新HOIP的结构,并通过后续合成和表征证实了预测的准确性,这将使成千上万种有机阳离子和无机层组合的高效准确筛选成为可能 。
- 无铅钙钛矿的发现:传统钙钛矿材料中铅的毒性是其商业化的一个主要障碍 。机器学习被用于加速无铅钙钛矿的发现。一项研究开发了一种目标驱动的方法,结合机器学习技术和密度泛函理论计算,从5158种未探索的HOIPs中快速筛选出六种具有适当带隙和室温热稳定性的正交无铅HOIPs,其中两种表现出可见光区的直接带隙和优异的环境稳定性 。另一项研究通过结合机器学习和高通量计算,从大量候选材料中筛选出了四个无铅钙钛矿候选材料,它们具有良好的稳定性、高德拜温度和合适的带隙,并通过DFT计算进行了验证,预测其效率高达27.6% 。
- 组分工程对稳定性的影响:阳离子工程对钙钛矿的稳定性有重要影响。研究发现,在多阳离子钙钛矿中,在高温老化测试(例如140°C)下,增加有机阳离子(如甲基铵)或减少无机阳离子(如铯)对光/热稳定性有不利影响;但在100°C以下,这种影响则相反 。通过机器学习和高通量实验,研究人员揭示了这种稳定性反转行为,并确定了通过掺入至少10 mol.% MA和最多5 mol.% Cs/Rb可以最大化器件在操作温度(<100°C)下的稳定性 。
- 缺陷控制与组分协同工程:相不稳定性是甲脒铅碘(FAPbI3)基太阳能电池商业化的严重挑战 。结合密度泛函理论和机器学习分子动力学模拟,研究人员发现普遍存在的碘空位和间隙可以通过在过渡态诱导强大的共价性来显著加速结构相变动力学 。研究还发现,A位工程主要控制热力学,而B位掺杂可以有效操纵FAPbI3相变的动力学,其中镧系离子被认为是很有前景的B位替代物 。A-B混合掺杂被认为是一种协同稳定α-FAPbI3的有效策略,通过实验证实了Cs-Eu掺杂的FAPbI3相比于Cs掺杂的FAPbI3具有显著更高的初始光电特性和增强的相稳定性 。
3. 界面工程与优化
埋藏界面工程对钙钛矿太阳能电池的性能至关重要,机器学习也为此提供了强大的工具。
- 界面材料设计:埋藏界面的自组装单层和缓冲层可以优化电荷转移并减少复合损失 。然而,复杂的机制和选择合适的官能团带来了巨大挑战。机器学习为筛选和识别有效的界面改性结构提供了强大工具 。一项研究通过机器学习驱动的方法,成功制备了两种有前景的有机分子PAPzO和PAPz,它们与SnO2和钙钛矿表现出协同作用。这些分子能够降低电荷陷阱密度,延长载流子寿命,并延缓钙钛矿结晶 。其中,PAPzO由于更强的结合能和更好的能级匹配,实现了26.04%的功率转换效率(PCE)和长期稳定性,在连续最大功率点跟踪1200小时后仍保持91.24%的原始PCE 。
- 后处理策略:机器学习还可以指导发现有效的后处理方法来提高稳定性。一项研究受到机器学习结果的启发,发现通过在CH3NH3PbI3层上涂覆一层薄薄的4,4′-二溴三苯胺(DBTPA)可以提高所得太阳能电池的稳定性 。钝化后的器件在85°C的N2气氛下未封装1000小时后仍保持96%的原始功率转换效率。通过近环境压X射线光电子能谱(XPS)研究表明,DBTPA处理抑制了碘化物和甲基胺在高达150°C和5毫巴湿度下的逸出,并通过路易斯酸碱和阳离子-π相互作用与钙钛矿配位 。
4. 结合第一性原理计算与高通量筛选
机器学习与第一性原理计算(如密度泛函理论,DFT)的结合是钙钛矿材料稳定性研究中的一个强大策略,它能够结合机器学习的高效率和DFT的高精度。
- 加速材料发现:传统的试错法在处理大量候选材料时效率低下,特别是对于具有复杂晶体结构和多目标属性的材料 。机器学习可以作为预筛选工具,显著缩小搜索空间。例如,一项研究结合机器学习和DFT计算,开发了一种目标驱动的方法,旨在从数万种双钙钛矿材料中筛选出具有合适光电性能和热稳定性的功能材料。通过训练优秀的机器学习模型、多条件组合筛选和DFT计算,最终从16,400个候选材料中筛选出10种优异的双钙钛矿材料,以指导后续的实验合成 。
- 弥补各自局限性:这种结合策略弥补了机器学习和DFT各自的缺点,提高了发现新材料的效率 。DFT计算虽然准确,但计算成本高昂,难以应用于大规模筛选。机器学习模型则可以快速生成预测,然后用DFT对预测结果进行验证和精炼,从而实现高通量、高精度的材料设计和发现 。
- 通用原子间势:在材料发现中,机器学习能量模型作为预过滤器应用于第一性原理计算数据,以高通量搜索稳定的无机晶体 。基准测试结果表明,通用原子间势(universal interatomic potentials)已经足够先进,可以有效且廉价地在未来高通量材料数据库的扩展中预筛选热力学稳定的假设材料 。MACE-OFF就是一个利用最先进机器学习技术和高水平量子力学理论计算的第一性原理参考数据创建的短程可转移力场系列,用于有机分子,能够准确预测分子系统的多种气相和凝聚相性质,包括分子晶体和液体的可靠描述,以及量子核效应 。
5. 挑战与展望
尽管机器学习在提升钙钛矿材料稳定性方面取得了显著进展,但仍面临一些挑战和发展趋势。
- 数据稀缺性与质量:机器学习模型的性能严重依赖于高质量、足够大的数据集。在材料科学领域,尤其是在钙钛矿材料稳定性方面,实验数据的收集仍然耗时且昂贵,导致数据稀缺 。未来的发展需要更高效的数据生成方法和更广泛的数据共享标准化 。
- 模型可解释性:尽管机器学习模型可以进行准确预测,但其“黑箱”特性有时会阻碍研究人员理解材料行为背后的物理机制 。提升模型的可解释性,例如通过特征重要性分析(如SHAP算法),将有助于揭示结构-性质-性能关系,并指导更深入的物理理解 。
- 多尺度协同建模:钙钛矿材料的稳定性是一个复杂的多尺度问题,涉及原子级相互作用、微观结构以及宏观器件性能。未来的研究需要开发能够整合不同尺度信息的机器学习模型,实现多尺度协同建模 。
- 与实验的闭环结合:将机器学习与实验形成一个闭环系统,即机器学习模型指导实验设计,实验结果反过来优化模型,是加速材料发现和优化的理想模式 。这种自主实验系统将进一步加速钙钛矿材料的开发和稳定性测试 .
- 描述符的选择与开发:机器学习描述符在连接材料的经验语言和机器学习算法的数值输入之间起到关键作用 。将领域知识转化为可计算形式的描述符可以显著增强模型的可解释性,并帮助研究人员发现控制PSC行为的潜在物理机制 。未来的研究需要进一步探索和优化钙钛矿材料晶体结构编码、薄膜微观结构量化以及材料和器件稳定性评估的描述符 。
总的来说,机器学习为解决钙钛矿太阳能电池的长期稳定性问题提供了强大的新工具。通过结合高通量实验、第一性原理计算和先进的机器学习算法,研究人员正在加速发现新的、更稳定的钙钛矿材料,优化器件结构,并深入理解影响稳定性的基本机制。随着数据积累、算法优化以及与实验的更紧密结合,机器学习有望在钙钛矿材料的商业化进程中发挥越来越重要的作用 。
References
1Machine Learning Roadmap for Perovskite PhotovoltaicsOpenAlex
Meghna Srivastava, John M. Howard, Tao Gong, et al.
Perovskite solar cells (PSC) are a favorable candidate for next-generation solar systems with efficiencies comparable to Si photovoltaics, but their long-term stability must be proven prior to commercialization. However, traditional trial-and-error approaches to PSC screening, development, and stability testing are slow and labor-intensive. In this Perspective, we present a survey of how machine learning (ML) and autonomous experimentation provide additional toolkits to gain physical understanding while accelerating practical device advancement. We propose a roadmap for applying ML to PSC research at all stages of design (compositional selection, perovskite material synthesis and testing, and full device evaluation). We also provide an overview of relevant concepts and baseline models that apply ML to diverse materials problems, demonstrating its broad relevance while highlighting promising research directions and associated challenges. Finally, we discuss our outlook for an integrated pipeline that encompasses all design stages and presents a path to commercialization.
2MACE-OFF: Short-Range Transferable Machine Learning Force Fields for Organic MoleculesOpenAlex
Dávid Péter Kovács, J. Harry Moore, Nicholas J. Browning, et al.
Classical empirical force fields have dominated biomolecular simulations for over 50 years. Although widely used in drug discovery, crystal structure prediction, and biomolecular dynamics, they generally lack the accuracy and transferability required for first-principles predictive modeling. In this paper, we introduce MACE-OFF, a series of short-range transferable force fields for organic molecules created using state-of-the-art machine learning technology and first-principles reference data computed with a high level of quantum mechanical theory. MACE-OFF demonstrates the remarkable capabilities of short-range models by accurately predicting a wide variety of gas- and condensed-phase properties of molecular systems. It produces accurate, easy-to-converge dihedral torsion scans of unseen molecules as well as reliable descriptions of molecular crystals and liquids, including quantum nuclear effects. We further demonstrate the capabilities of MACE-OFF by determining free energy surfaces in explicit solvent as well as the folding dynamics of peptides and nanosecond simulations of a fully solvated protein. These developments enable first-principles simulations of molecular systems for the broader chemistry community at high accuracy and relatively low computational cost.
3Accurate Crystal Structure Prediction of New 2D Hybrid Organic–Inorganic PerovskitesOpenAlex
Nima Karimitari, William J. Baldwin, Evan W. Muller, et al.
Low-dimensional hybrid organic–inorganic perovskites (HOIPs) are promising electronically active materials for light absorption and emission. The design space of HOIPs is extremely large, as a variety of organic cations can be combined with different inorganic frameworks. This not only allows for tunable electronic and mechanical properties but also necessitates the development of new tools for in silico high throughput analysis of candidate materials. In this work, we present an accurate, efficient, and widely applicable machine learning interatomic potential (MLIP) trained on 86 diverse experimentally reported HOIP materials. This MLIP was tested on 73 experimentally reported perovskite compositions and achieves a high accuracy, relative to density functional theory (DFT). We also introduce a novel random structure search algorithm designed for the crystal structure prediction of 2D HOIPs. The combination of MLIP and the structure search algorithm reliably recovers the crystal structure of 14 known 2D perovskites by specifying only the organic molecule and inorganic cation/halide. Performing this crystal structure search with ab initio methods would be computationally prohibitive but is relatively inexpensive with the MLIP. Finally, the developed procedure is used to predict the structure of a totally new HOIP with cation (cis-1,3-cyclohexanediamine). Subsequently, the new compound was synthesized and characterized, which matches the predicted structure, confirming the accuracy of our method. This capability will enable the efficient and accurate screening of thousands of combinations of organic cations and inorganic layers for further investigation.
4Machine learning in materials informatics: recent applications and prospectsOpenAlex
Rampi Ramprasad, Rohit Batra, Ghanshyam Pilania, et al.
Abstract Propelled partly by the Materials Genome Initiative, and partly by the algorithmic developments and the resounding successes of data-driven efforts in other domains, informatics strategies are beginning to take shape within materials science. These approaches lead to surrogate machine learning models that enable rapid predictions based purely on past data rather than by direct experimentation or by computations/simulations in which fundamental equations are explicitly solved. Data-centric informatics methods are becoming useful to determine material properties that are hard to measure or compute using traditional methods—due to the cost, time or effort involved—but for which reliable data either already exists or can be generated for at least a subset of the critical cases. Predictions are typically interpolative, involving fingerprinting a material numerically first, and then following a mapping (established via a learning algorithm) between the fingerprint and the property of interest. Fingerprints, also referred to as “descriptors”, may be of many types and scales, as dictated by the application domain and needs. Predictions may also be extrapolative—extending into new materials spaces—provided prediction uncertainties are properly taken into account. This article attempts to provide an overview of some of the recent successful data-driven “materials informatics” strategies undertaken in the last decade, with particular emphasis on the fingerprint or descriptor choices. The review also identifies some challenges the community is facing and those that should be overcome in the near future.
5AI-aided structural design and stability of hybrid semiconductor photovoltaicsOpenAlex
Shaoqiang Wang, Wei Liu, Yifan Li, et al.
有机-无机杂化钙钛矿太阳能电池(perovskite solar cells, PSCs)因其高光电转换效率与低成本制造潜力而成为新一代光伏技术的重要候选。然而,其长期稳定性问题仍是制约其商业化应用的关键瓶颈。近年来,人工智能(artificial intelligence, AI)技术在材料科学领域快速发展,逐渐展现出在稳定性调控中的独特优势。本文围绕AI在PSCs稳定性优化中的应用进展,系统梳理了AI在材料组分筛选、界面工程设计和缺陷调控方面的典型应用路径与代表性成果。通过引入机器学习、贝叶斯优化等前沿算法以及AI与密度泛函理论和分子动力学结合在复杂多变量体系中的建模能力与应用局限,展望了未来AI与实验闭环结合、多尺度协同建模、数据共享标准化等发展趋势。本文旨在为PSCs结构稳定性设计提供系统性参考,并推动AI技术在新型光伏器件可靠性提升中的更深层次融合。
6Predictions of new <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:mrow><mml:mi mathvariant="italic">AB</mml:mi><mml:msub><mml:mi mathvariant="normal">O</mml:mi><mml:mn>3</mml:mn></mml:msub></mml:mrow></mml:math> perovskite compounds by combining machine learning and density functional theoryOpenAlex
Prasanna V. Balachandran, Antoine Emery, J. E. Gubernatis, et al.
We apply machine learning (ML) methods to a database of 390 experimentally reported $AB{\mathrm{O}}_{3}$ compounds to construct two statistical models that predict possible new perovskite materials and possible new cubic perovskites. The first ML model classified the 390 compounds into 254 perovskites and 136 that are not perovskites with a 90% average cross-validation (CV) accuracy; the second ML model further classified the perovskites into 22 known cubic perovskites and 232 known noncubic perovskites with a 94% average CV accuracy. We find that the most effective chemical descriptors affecting our classification include largely geometric constructs such as the $A$ and $B$ Shannon ionic radii, the tolerance and octahedral factors, the $A$-O and $B$-O bond length, and the $A$ and $B$ Villars' Mendeleev numbers. We then construct an additional list of $625AB{\mathrm{O}}_{3}$ compounds assembled from charge conserving combinations of $A$ and $B$ atoms absent from our list of known compounds. Then, using the two ML models constructed on the known compounds, we predict that 235 of the 625 exist in a perovskite structure with a confidence greater than 50% and among them that 20 exist in the cubic structure (albeit, the latter with only $\ensuremath{\sim}50%$ confidence). We find that the new perovskites are most likely to occur when the $A$ and $B$ atoms are a lanthanide or actinide, when the $A$ atom is an alkali, alkali earth, or late transition metal atom, or when the $B$ atom is a $p$-block atom. We also compare the ML findings with the density functional theory calculations and convex hull analyses in the Open Quantum Materials Database (OQMD), which predicts the $T=0$ K ground-state stability of all the $AB{\mathrm{O}}_{3}$ compounds. We find that OQMD predicts 186 of 254 of the perovskites in the experimental database to be thermodynamically stable within 100 meV/atom of the convex hull and predicts 87 of the 235 ML-predicted perovskite compounds to be thermodynamically stable within 100 meV/atom of the convex hull, including 6 of these to be in cubic structures. We suggest these 87 as the most promising candidates for future experimental synthesis of novel perovskites.
7A framework to evaluate machine learning crystal stability predictionsOpenAlex
Janosh Riebesell, Rhys E. A. Goodall, Philipp Benner, et al.
Abstract The rapid adoption of machine learning in various scientific domains calls for the development of best practices and community agreed-upon benchmarking tasks and metrics. We present Matbench Discovery as an example evaluation framework for machine learning energy models, here applied as pre-filters to first-principles computed data in a high-throughput search for stable inorganic crystals. We address the disconnect between (1) thermodynamic stability and formation energy and (2) retrospective and prospective benchmarking for materials discovery. Alongside this paper, we publish a Python package to aid with future model submissions and a growing online leaderboard with adaptive user-defined weighting of various performance metrics allowing researchers to prioritize the metrics they value most. To answer the question of which machine learning methodology performs best at materials discovery, our initial release includes random forests, graph neural networks, one-shot predictors, iterative Bayesian optimizers and universal interatomic potentials. We highlight a misalignment between commonly used regression metrics and more task-relevant classification metrics for materials discovery. Accurate regressors are susceptible to unexpectedly high false-positive rates if those accurate predictions lie close to the decision boundary at 0 eV per atom above the convex hull. The benchmark results demonstrate that universal interatomic potentials have advanced sufficiently to effectively and cheaply pre-screen thermodynamic stable hypothetical materials in future expansions of high-throughput materials databases.
8Accelerated discovery of stable lead-free hybrid organic-inorganic perovskites via machine learningOpenAlex
Shuaihua Lu, Qionghua Zhou, Yixin Ouyang, et al.
Rapidly discovering functional materials remains an open challenge because the traditional trial-and-error methods are usually inefficient especially when thousands of candidates are treated. Here, we develop a target-driven method to predict undiscovered hybrid organic-inorganic perovskites (HOIPs) for photovoltaics. This strategy, combining machine learning techniques and density functional theory calculations, aims to quickly screen the HOIPs based on bandgap and solve the problems of toxicity and poor environmental stability in HOIPs. Successfully, six orthorhombic lead-free HOIPs with proper bandgap for solar cells and room temperature thermal stability are screened out from 5158 unexplored HOIPs and two of them stand out with direct bandgaps in the visible region and excellent environmental stability. Essentially, a close structure-property relationship mapping the HOIPs bandgap is established. Our method can achieve high accuracy in a flash and be applicable to a broad class of functional material design.
9Machine Learning-Assisted Analysis of Perovskite Solar Cell Long-Term Stability under Multiple Environmental FactorsOpenAlex
Shanshan Zhao, Sijia Zhou, Zhongli Guo, et al.
Perovskite solar cells (PSCs) have made significant strides in the past decade. However, poor long-term stability is a major challenge for PSCs, which hinders large-scale commercialization. Traditional trial-and-error methods are limited by the complexity of the environmental conditions and device structures. This study introduces a machine learning (ML)-assisted approach to analyze factors affecting the PSC stability. A multihead attention mechanism is used to simultaneously process diverse input data, including external and internal parameters. Combined with a squeeze-and-excitation residual network (SEResNet), this approach achieves a coefficient of determination (R2) of 0.972 and a Pearson correlation coefficient (r) of 0.986. Furthermore, the SHapley Additive exPlanations (SHAP) algorithm identifies key factors influencing stability. Through high-throughput prediction of approximately 2000 PSCs, we explore the interactive effects of key factors, offering a comprehensive understanding of their influences on device stability. In addition, we also present the predicted optimal PSC system structure for stability at 85 °C and 85% relative humidity (RH). Subsequently, we conduct device lifetime experiments to present the consistency between experiment and predication results. Hence, this work demonstrates the potential of ML in terms of predicting the stability of PSCs and obtaining critical parameters, facilitating the translation of laboratory research findings into practical applications.
10Predicting Perovskite Performance with Multiple Machine-Learning AlgorithmsOpenAlex
Ruoyu Li, Qin Deng, Dong Tian, et al.
Perovskites have attracted increasing attention because of their excellent physical and chemical properties in various fields, exhibiting a universal formula of ABO3 with matching compatible sizes of A-site and B-site cations. In this work, four different prediction models of machine learning algorithms, including support vector regression based on radial basis kernel function (SVM-RBF), ridge regression (RR), random forest (RF), and back propagation neural network (BPNN), are established to predict the formation energy, thermodynamic stability, crystal volume, and oxygen vacancy formation energy of perovskite materials. Combined with the fitting diagrams of the predicted values and DFT calculated values, the results show that SVM-RBF has a smaller bias in predicting the crystal volume. RR has a smaller bias in predicting the thermodynamic stability. RF has a smaller bias in predicting the formation energy, crystal volume, and thermodynamic stability. BPNN has a smaller bias in predicting the formation energy, thermodynamic stability, crystal volume, and oxygen vacancy formation energy. Obviously, different machine learning algorithms exhibit different sensitivity to data sample distribution, indicating that we should select different algorithms to predict different performance parameters of perovskite materials.
11Machine Learning Descriptors for Mapping Structure‐Property‐Performance Relationships of Perovskite Solar CellsOpenAlex
Yu Su, Kongxiang Wang, Xiang Guan, et al.
ABSTRACT In recent years, machine learning (ML) has emerged as a versatile tool for accelerating the development of perovskite solar cells (PSCs). A key challenge, however, lies in the scarcity of researchers possessing deep expertise in both material science and artificial intelligence. Pivotal to bridging this gap is ML descriptors, mediating between the empirical language of materials and the numerical inputs for ML algorithms. By translating domain knowledge into computationally tractable forms, the descriptors significantly enhance the model interpretability and empower researchers to uncover the underlying physical mechanisms governing behavior of PSCs. Therefore, it is crucial to overview the efforts translating the structure, property of perovskite materials and performance of PSCs into numerical descriptors compatible with ML models. This review summarized (1) the encoding of crystal structure in perovskites; (2) the quantification of microstructures in perovskite films; (3) the stability assessment of perovskite materials and devices. By synthesizing progress in these aspects, this work lays a solid foundation for constructing a universal model to elucidate the structure‐property‐performance relationships in PSCs, especially in forward prediction and backward inference.
12Discovery of temperature-induced stability reversal in perovskites using high-throughput robotic learningOpenAlex
Yicheng Zhao, Jiyun Zhang, Zhengwei Xu, et al.
Stability of perovskite-based photovoltaics remains a topic requiring further attention. Cation engineering influences perovskite stability, with the present-day understanding of the impact of cations based on accelerated ageing tests at higher-than-operating temperatures (e.g. 140°C). By coupling high-throughput experimentation with machine learning, we discover a weak correlation between high/low-temperature stability with a stability-reversal behavior. At high ageing temperatures, increasing organic cation (e.g. methylammonium) or decreasing inorganic cation (e.g. cesium) in multi-cation perovskites has detrimental impact on photo/thermal-stability; but below 100°C, the impact is reversed. The underlying mechanism is revealed by calculating the kinetic activation energy in perovskite decomposition. We further identify that incorporating at least 10 mol.% MA and up to 5 mol.% Cs/Rb to maximize the device stability at device-operating temperature (<100°C). We close by demonstrating the methylammonium-containing perovskite solar cells showing negligible efficiency loss compared to its initial efficiency after 1800 hours of working under illumination at 30°C.
13Machine‐Learning Modeling for Ultra‐Stable High‐Efficiency Perovskite Solar CellsOpenAlex
Yingjie Hu, Xiaobing Hu, Lu Zhang, et al.
Abstract Understanding the key factor driving the efficiency and stability of semiconductor devices is vital. To date, the key factor influencing the long‐term stability of perovskite solar cells (PSCs) remains unknown because of the many influencing factors. In this work, through machine learning, the influences of five factors, including grain size, defect density, bandgap, fluorescence lifetime, and surface roughness, on the efficiency and stability of PSCs have been revealed. It is found that the bandgap has the greatest influence on the efficiency, and the surface roughness and grain size are most influential to the long‐term stability. A mathematical model is given to predict efficiency based on fluorescence lifetime and bandgap. Guided by the model, four groups of experiments are conducted to confirm the machine‐learning predictions and a PSC with 23.4% efficiency and excellent long‐term stability is obtained, as manifested by retention of 97.6% of the initial efficiency after 3288 h aging in the ambient environment, the best stability under these conditions. This work shows that machine learning is an effective means to enrich semiconductor physical models.
14Polyelemental, Multicomponent Perovskite Semiconductor Libraries through Combinatorial ScreeningOpenAlex
Michael Saliba
Abstract Recently, perovskites with multiple cations, metals, and anions have shown very high efficiencies and stabilities for perovskite solar cells. The novel materials frequently exhibit unexpected and beneficial properties, outperforming simpler counterparts. The trend of increasing material complexity requires a systematic strategy to explore polyelemental “multicomponent engineering.” Here, a combinatorial approach is introduced to generate all possible, unique combinations within a set of available components. Thus, with each new component, the combinatorial framework can generate the full theoretical parameter space. Based on reported components, the experimental parameter space can then be identified. The exceptional material versatility of perovskites is suited for high‐throughput screening, machine‐learning, or data mining, laying the foundation for a “perovskite genome project” that thoroughly catalogues the entire material family for desired properties. This can provide the framework for theoretical simulations toward understanding the fundamental working principles of perovskite materials enabling the “next big thing” after perovskites. Finally, informed by literature, a promising candidate list for future material exploration is presented including novel organic‐free, Pb‐free, and all‐inorganic perovskites. These compounds are primary contenders toward stable, high efficiency, and reproducible materials for rapid industrialization of perovskite solar cells, lasers, light‐emitting diodes, photo detectors, or particle detectors.
15Toward stabilization of formamidinium lead iodide perovskites by defect control and composition engineeringOpenAlex
Yuhang Liang, Feng Li, Xiangyuan Cui, et al.
Abstract Phase instability poses a serious challenge to the commercialization of formamidinium lead iodide (FAPbI 3 )-based solar cells and optoelectronic devices. Here, we combine density functional theory and machine learning molecular dynamics simulations, to investigate the mechanism driving the undesired α-δ phase transition of FAPbI 3 . Prevalent iodine vacancies and interstitials can significantly expedite the structural transition kinetics by inducing robust covalency during transition states. Extrinsically, the detrimental roles of atmospheric moisture and oxygen in degrading the FAPbI 3 perovskite phase are also rationalized. Significantly, we discover the compositional design principles by categorizing that A-site engineering primarily governs thermodynamics, whereas B-site doping can effectively manipulate the kinetics of the phase transition in FAPbI 3 , highlighting lanthanide ions as promising B-site substitutes. A-B mixed doping emerges as an efficient strategy to synergistically stabilize α-FAPbI 3 , as experimentally demonstrated by substantially higher initial optoelectronic characteristics and significantly enhanced phase stability in Cs-Eu doped FAPbI 3 as compared to its Cs-doped counterpart. This study provides scientific guidance for the design and optimization of long-term stable FAPbI 3 -based solar cells and other optoelectronic devices through defect control and synergetic composition engineering.
16Discovery of Lead‐Free Perovskites for High‐Performance Solar Cells via Machine Learning: Ultrabroadband Absorption, Low Radiative Combination, and Enhanced Thermal ConductivitiesOpenAlex
Xia Cai, Yiming Zhang, Zejiao Shi, et al.
Abstract Exploring lead‐free candidates and improving efficiency and stability remain the obstacle of hybrid organic‐inorganic perovskite‐based devices commercialization. Traditional trial‐and‐error methods seriously restrict the discovery especially for large search space, complex crystal structure and multi‐objective properties. Here, the authors propose a multi‐step and multi‐stage screening scheme to accelerate the discovery of hybrid organic‐inorganic perovskites A 2 BB′X 6 from a large number of candidates through combining machine learning with high‐throughput calculations for pursuing excellent efficiency and thermal stability in solar cells. Followed by a series of screenings, the structure‐property relationships mapping A 2 BB′X 6 properties are built and the predictions are close to reported experimental results. Successfully, four experimental‐feasibly candidates with good stability, high Debye temperature and suitable band gap are screened out and further verified by density‐functional theory calculations, in which the predicted efficiency for three lead‐free candidates ((CH 3 NH 3 ) 2 AgGaBr 6 , (CH 3 NH 3 ) 2 AgInBr 6 and (C 2 NH 6 ) 2 AgInBr 6 ) achieves 20.6%, 19.9% and 27.6% due to ultrabroadband absorption region ranging from UVC to IRC with excitonic radiative combination rates as low as 10 ps, large or intermediate polarons form with properties similar to CH 3 NH 3 PbI 3 and the calculated thermal conductivities are 5.04, 4.39 and 5.16 Wm −1 K −1 , respectively, with Debye temperatures larger than 500 K, beneficial for suppression of both nonradiative combination and heat‐induced degradation.
17Machine Learning Accelerates the Discovery of Light-Absorbing Materials for Double Perovskite Solar CellsOpenAlex
Zhenqing Yang, Yanbing Liu, Yun‐Qian Zhang, et al.
Double perovskite materials have excellent electronic and optical properties, which are the star material in the photovoltaic field. However, the large number of family members has brought difficulties to traditional material screening methods. The emergence of machine learning provides new ideas for this problem, and how to apply machine learning to search new functional materials has become a hot research topic. This paper combines machine learning and density functional theory (DFT) calculations to develop a goal-driven method to search for functional materials, aiming to select one that has suitable photoelectric properties and thermal stability from tens of thousands of double perovskite materials. Through a series of processes such as training excellent machine learning models, multicondition combination screening, and DFT calculations, 10 excellent double perovskite materials were finally selected from 16,400 candidate materials to guide subsequent experimental synthesis. Combining the high efficiency of machine learning and the accuracy of DFT calculations compensates for their respective shortcomings and improves the efficiency of finding new materials. In addition, this method also has a certain general applicability and can be extended to the screening of other materials.
18Machine learning-guided realization of full-color high-quantum-yield carbon quantum dotsOpenAlex
Huazhang Guo, Yuhao Lu, Zhendong Lei, et al.
Carbon quantum dots (CQDs) have versatile applications in luminescence, whereas identifying optimal synthesis conditions has been challenging due to numerous synthesis parameters and multiple desired outcomes, creating an enormous search space. In this study, we present a novel multi-objective optimization strategy utilizing a machine learning (ML) algorithm to intelligently guide the hydrothermal synthesis of CQDs. Our closed-loop approach learns from limited and sparse data, greatly reducing the research cycle and surpassing traditional trial-and-error methods. Moreover, it also reveals the intricate links between synthesis parameters and target properties and unifies the objective function to optimize multiple desired properties like full-color photoluminescence (PL) wavelength and high PL quantum yields (PLQY). With only 63 experiments, we achieve the synthesis of full-color fluorescent CQDs with high PLQY exceeding 60% across all colors. Our study represents a significant advancement in ML-guided CQDs synthesis, setting the stage for developing new materials with multiple desired properties.
19Ambient Pressure X-ray Photoelectron Spectroscopy Investigation of Thermally Stable Halide Perovskite Solar Cells via Post-TreatmentOpenAlex
Shougui Ning, Songwei Zhang, Jiaonan Sun, et al.
Long-term thermal stability is one limiting factor that impedes the commercialization of the perovskite solar cell. Inspired by our prior results from machine learning, we discover that coating a thin layer of 4,4′-dibromotriphenylamine (DBTPA) on top of a CH3NH3PbI3 layer can improve the stability of resultant solar cells. The passivated devices kept 96% of the original power conversion efficiency for 1000 h at 85 °C in a N2 atmosphere without encapsulation. Near-ambient pressure X-ray photoelectron spectroscopy (XPS) was employed to investigate the evolution of the composition and evaluate thermal and moisture stability by in situ studies. A comparison between pristine MAPbI3 films and DBTPA-treated films shows that the DBTPA treatment suppresses the escape of iodide and methylamine up to 150 °C under 5 mbar humidity. Furthermore, we have used attenuated total reflection Fourier transform infrared and XPS to probe the interactions between DBTPA and MAPbI3 surfaces. The results prove that DBTPA coordinates with the perovskite by Lewis acid–base and cation−π interaction. Compared with the 19.9% efficiency of the pristine sample, the champion efficiency of the passivated sample reaches 20.6%. Our results reveal DBTPA as a new post-treating molecule that leads not only to the improvement of the photovoltaic efficiency but also thermal and moisture stability.
20Machine-Learning-Assisted Design of Buried-Interface Engineering Materials for High-Efficiency and Stable Perovskite Solar CellsOpenAlex
Qi Zhang, Han Wang, Qiangqiang Zhao, et al.
Buried-interface engineering is crucial to the performance of perovskite solar cells. Self-assembled monolayers and buffer layers at the buried interface can optimize charge transfer and reduce recombination losses. However, the complex mechanisms and the difficulty in selecting suitable functional groups pose great challenges. Machine learning (ML) offers a powerful tool for screening and identifying effective structures for interface modification. Our ML-driven approach led to the preparation of two promising organic molecules, PAPzO and PAPz, which exhibit synergistic interactions with SnO2 and perovskites. These molecules decrease charge trap densities, elongate carrier lifetimes, and retard perovskite crystallization. PAPzO, with a stronger binding energy and better aligned energy levels, enables a power conversion efficiency (PCE) of 26.04% and long-term stability, maintaining 91.24% of its original PCE after 1,200 h of continuous maximum power point tracking. This ML-integrated approach marks a significant advancement in the development of efficient and stable perovskite photovoltaics.