【作者】 赵倩；

【导师】 陈立泉； 施思齐；

【作者基本信息】 上海大学 ， 材料物理与化学， 2021， 博士

【摘要】 材料基因组是一种通过融合高通量计算、高通量实验、材料数据库等三大要素高效地推动新材料研发的新理念、新方法。基于材料基因组的数据驱动模式代表了材料研发的新范式。在锂电池领域,无机固态电解质由于较高的安全性而受到了广泛的关注。然而,较低的离子电导率限制了它的发展和应用。理解离子传导机制并挖掘影响离子传导快慢的原因及因素,有助于合理有效地发现、设计高离子电导率的无机固态电解质材料。鉴于此,本论文基于材料基因组的思想,结合统计分析和机器学习的方法,对无机固态电解质材料离子传导问题进行了系统的定性和定量研究。主要围绕离子传导相关的描述符等数据的产生、“描述符-离子传导性能”之间的关系分析展开。本论文主要研究内容与结果如下:1.从动静结合的角度系统地分析了影响离子传导的可能性因素。通过整合无机固态电解质中离子传导有关的理论、经验知识和设计规则等,从动静结合的角度系统地分析了离子传导相关的因素。主要涉及三个维度（晶体结构、能量、传导动力学）,六个层面（离子的特性、离子的空间排列特点、缺陷和离子传导机制、离子势能变化、晶格动力学、旋转动力学）,并且以体积效应和明伦（paddle-wheel）效应分别作为静、动态因素分析的代表进行实例研究。通过α-Li₂SO₄基和α-Na₃PO₄基无机塑晶电解质以及Li₁₀MP₂S₁₂（M=Si,Ge,Se）电解质的相结构演变和离子电导率的实验以及计算结果,揭示了聚阴离子的旋转运动作用（即明伦效应）受到了抑制是较大尺寸的聚阴离子替换虽然扩展了骨架晶格却增大了激活能、降低了体系的离子电导率的原因。结果表明,为了研发高离子电导率的无机固态电解质材料,需要在增强体积效应的同时促进聚阴离子的明伦效应,证实了动静结合分析影响离子传导因素的重要性和必要性。2.提出了一种通用的分层编码晶体结构基（HECS）描述符。基于离子传导相关因素的分析,通过整合全局及局域离子传导环境对离子传导快慢的影响,构建了一套通用的分层编码晶体结构基（HECS）描述符,包含组成、结构、传导通道、离子分布、特殊离子五个方面。以立方相锂-硫银锗矿型电解质材料作为模型体系进行实例研究,建立了32个HECS描述符,并借助偏相关分析方法推断各描述符与激活能之间的因果关系。结果表明,小尺寸的阴离子替换有利于立方相锂-硫银锗矿中锂离子的传导,这是由受阴离子位置无序控制的晶格空间、瓶颈尺寸等相互竞争作用所致。藉此在已有立方相锂-硫银锗矿型电解质材料基础上,预测出潜在的高离子传导性能的无机固态电解质的优化成分范围:Li_6-xPS_5-xCl_1+x（0<x<1）,Li_6+xPS_5+xBr_1-x（0<x<1）,Li₆PS₅Cl_0.5+xBr_0.5-x（0<x<0.5）,Li_6+xPS_5+xBr_0.75I_0.25-x（0<x<0.25）以及Li₆PS₅Br_0.75+xI_0.25-x（0<x<0.25）。其中Li_5.5PS_4.5Cl_1.5(Li_6-xPS_5-xCl_1+x,x=0.5)已被实验证实为具有应用前景的无机固态电解质材料,其室温电导率达9.4m S/cm,比相同加工条件下制备的Li₆PS₅Cl高约4倍,且激活能仅0.29 e V;烧结导致更高的电导率（12±0.2 m S/cm）,可与Li₁₀Ge P₂S₁₂相媲美,但Li_5.5PS_4.5Cl_1.5更为便宜。3.建立了一个通用的基于HECS描述符的机器学习预测无机固态电解质材料激活能的框架。基于该框架,以立方相锂-硫银锗矿型电解质材料作为模型体系进行实例研究,采用偏最小二乘方法（PLS）建立了高性能的（训练集:R²,88.7%;RMSE,0.02e V;测试集:R²,82.0%;RMSE,0.02e V）预测激活能的模型。变量投影重要性（VIP）分析表明了全局及局域离子传导环境对激活能的联合作用,其中平均阴离子尺寸以及与阴离子位置无序密切相关的结构的改变对激活能值的贡献尤为突出,这一发现有助于进一步指导发现或设计新的无机固态电解质材料。同时,对该模型进行的知识提取表明,可以通过增大瓶颈尺寸、引发阴离子位置无序、激活离子协同迁移等优化和设计出具有高离子传导性能的新的无机固态电解质材料,如Li_6-xPS_5-xCl_1+x（E_a<0.322 e V）,Li_6+xPS_5+xBr_1-x（E_a<0.273 e V）,Li_6+xPS_5+xBr_0.25I_0.75-x（E_a<0.352 e V）,Li_6+（5-n）yP_1-yN_yS₅I（E_a<0.420 e V）,Li_6+（5-n）yAs_1-yN_yS₅I（E_a<0.371 e V）,Li_6+（5-n）yAs_1-yN_ySe₅I（E_a<0.450 e V）等。更多还原

【Abstract】 Materials genome is a kind of new ideas and new methods to efficiently promote the research and development of new materials by integrating three elements of high-throughput computing,high-throughput experiment and material database.The data-driven mode based on the materials genome represents a new paradigm for material research and development.In the field of lithium-ion battery,inorganic solid electrolyte has been widely concerned due to its high safety,while the lower ionic conductivity limits its development and application.Understanding the ion conduction mechanism and extracting the cause and rule of ion conduction speed is helpful to reasonably and effectively find and design inorganic solid electrolyte materials with high ionic conductivity.Inspired by this,this thesis,based on the idea of the material genome and adopting the methods of statistical analysis and machine learning,mainly conducts a systematic qualitative and quantitative research on the basic problem of ion conduction.It mainly focuses on the generation of data such as descriptors for ion condction and the analysis of the relationship between descriptors and ion conduction performance.The main contents and results of this thesis are as follows:1.The factors influencing the ionic conductivity are systematically analyzed by combining static and dynamic lattice.By integrating the theory,experience knowledge and desigh rules of ion conduction in inorganic solid electrolyte,the factors related to ionic conduction are systemcatically analyzed from the point of the static and dynamic.It mainly involves three dimensions（crystal structure,energy,and conduction dynamics）and six aspects（ionic properties,ionic spatial arrangement characteristics,defects and ionic conduction mechanisms,the ionic potential energy changes,lattice dynamics and rotational dynamics）.The case studies are carried out by taking volume effect and paddle-wheel effect as the representations of static and dynamic factors respectively.From the experimental and calculated results on the phase structure evolution and ionic conductivities ofα-Li₂SO₄ andα-Na₃PO₄based inorganic plastic crystal electrolytes and Li₁₀MP₂S₁₂（M=Si,Ge,Se）electrolytes,it reveals the the suppression of rotational motion effect of polyanion,i.e.paddle-wheel is the reason of that the substitution of the bigger-size polyanion expands the skeleton lattice but ultimately increases activation energy and lowers the ionic conductivity.These results show that,in order to develop inorganic solid electrolyte materials with high ionic conductivity,it is necessary to enhance the volume effect and impove the paddle-wheel effect of polyanion at the same time,which verifies the importance and necessity of analyzing factors associated with ionic conduction by combing static and dynamic effect.2.A generic hierarchically encoding crystal structure-based descriptor is proposed.Based on the anlaysis on factor assiated with ion conduction,by integrating the global and local effects of Li⁺conduction environment,we develop the generic hierarchically encoding crystal structure-based（HECS）descriptors encompassing composition,structure,conduction pathway,ion distribution,and special ions derived from the unit cell information.The case study is carried out by taking cubic Li-argyrodite as the model system.32 HECS-descriptors are constructed and their causalities with Li⁺conduction are inferred via partial correlation analysis.It’s found that the smaller anion size plays a significant role in achieving lower activation energy,which results from the competing effects between the lattice space and bottleneck size controlled by framework site disorder.Based on this rule,the promising candidates are suggested,such as Li_6-xPS_5-xCl_1+x（0<x<1）,Li_6+xPS_5+xBr_1-x（0<x<1）,Li₆PS₅Cl_0.5+xBr_0.5-x（0<x<0.5）,Li_6+xPS_5+xBr_0.75I_0.25-x（0<x<0.25）and Li₆PS₅Br_0.75+xI_0.25-x（0<x<0.25）,in which Li_5.5PS_4.5Cl_1.5(Li_6-xPS_5-xCl_1+x,x=0.5)have been experimentally evaluated as excellent candidates for practical SSEs.Its room temperature ionicconductivity can reach 9.4 m S/cm,which is about 4 times higher than that of Li₆PS₅Cl prepared under the same processing conditions.And its activation energy is only 0.29 e V.Besides,sintering leads to a higher conductivity（12±0.2 m S/cm）of Li_5.5PS_4.5Cl_1.5,which is comparable to Li₁₀Ge P₂S₁₂,However,the raw materials of Li_5.5PS_4.5Cl_1.5 are cheaper than that of Li₁₀Ge P₂S₁₂.3.A generic framework of ML prediction for E_a in SSEs with hierarchically encoding crystal structure-based（HECS）descriptors is constructed.Under this framework,the case study is carried out by taking cubic Li-argyrodite as the model system.The E_a prediction model is developed to the coefficient of determination（R²）and root-mean-square error（RMSE）values of 0.887 and 0.02 e V for training dataset,and 0.820 and 0.02 e V for test dataset,respectively by partial least-squares regression analysis.The variable importance in projection（VIP）scores demonstrate the combined effects from global and local Li⁺conduction environments,especially the anion size and the resultant structural changes associated with anion site disorder make the significant contributions to the variation of E_a values.This analysis helps to further guide the discovery and design of new potential SSEs.Meantime,knowledge extracted from the ML model leads us to optimize and design new SSEs with low E_a,such as Li_6-xPS_5-xCl_1+x（E_a<0.322 e V）,Li_6+xPS_5+xBr_1-x（E_a<0.273 e V）,Li_6+xPS_5+xBr_0.25I_0.75-x（E_a<0.352 e V）;Li_6+（5-n）yP_1-yN_yS₅I（E_a<0.420 e V）,Li_6+（5-n）yAs_1-yN_yS₅I（E_a<0.371 e V）,Li_6+（5-n）yAs_1-yN_ySe₅I（E_a<0.450 e V）by broadening the bottleneck size,invoking anion site disorder and activating concerted conduction.更多还原