深入研究低阶煤的热解行为和热解反应机理对于开发先进的低阶煤转化技术具有重要意义。然而煤热解为高温自由基过程，期间涉及数量众多且相互耦合的化学反应路径并伴有大量自由基中间产物的生成与消耗，反应机理极其复杂。由于缺乏对自由基演化的原位检测手段，使得通过实验从分子层面研究煤热解的微观机理面临诸多困难。基于键级的化学反应力场ReaxFF与分子动力学结合的反应分子动力学方法（ReaxFF MD），可连续描述化学键的断裂与生成，其准确度接近于广泛应用的密度泛函理论，并且可以大幅度降低计算的复杂度，是极具潜力的深入研究煤热解机理的方法。本论文致力于采用ReaxFF MD模拟对府谷次烟煤的热解规律进行深入研究，提出了一种大规模多组分煤模型的构建方法，构建了府谷次烟煤的大规模多组分模型，针对该模型开展了慢速升温与恒温热解过程的ReaxFF MD模拟，获得了利用实验手段和其它计算方法难以得到的煤热解反应机理的认识。本论文还开展了府谷次烟煤的热解-同步辐射真空紫外光电离飞行时间质谱实验，以便从实验角度更好地认识府谷次烟煤的热解过程并为ReaxFF MD模拟结果评价提供参考并与其互为补充。本论文的主要工作与结论如下：基于中国科学技术大学国家同步辐射实验室的固体热解-同步辐射真空紫外光电离飞行时间质谱实验装置，开展了府谷次烟煤的热解实验研究，确定了主要热解产物的质谱峰归属。通过比较不同温度（500、600、700、800℃）条件下主要热解产物的质谱信号获得了主要气体产物和代表性芳香族产物随热解温度的演变规律。实验结果表明：高温有利于低分子烯烃类热解产物的生成；酚羟基的存在和芳香环的增大都会使芳香环的烷基取代基的反应活性增强。提出了一种基于煤样的常规实验表征数据（工业分析与元素分析、13C核磁共振分析和溶剂抽提实验）的大规模多组分煤模型的手动构建方法。该方法兼顾煤模型的结构多样性和构建效率，对煤大分子组分采用平均化的构建策略以快速达到煤模型与煤样的元素组成及主要结构参数的吻合，并通过改变组成煤大分子的结构单元间的连接关系以增加煤模型的结构多样性。该方法同时考虑了分布于煤大分子组分间的游离相小分子化合物以更好地描述煤的两相结构。基于该模型构建方法并结合府谷次烟煤的实验表征数据，构建了其大规模多组分三维模型。该模型总共包含23,898原子，由75个平均化的大分子组分模型和29个游离相小分子化合物组成，其中大分子平均组分包含20种不同的分子结构以描述煤的结构多样性。该多组分模型的元素组成、芳香碳含率及芳香团簇的平均尺寸都与煤样的实验表征数据较为吻合。该模型应用于ReaxFF MD热解模拟可获得合理的产物组成及其演化趋势。应用基于GPU并行的反应分子动力学模拟程序GMD-Reax，对府谷次烟煤的多组分模型开展了慢速升温（300~2500 K, 1 K/ps）条件下的ReaxFF MD热解模拟研究。获得了府谷次烟煤在慢速升温条件下热解过程的三个阶段：煤结构活化阶段（Stage-I）、快速热分解阶段（Stage-II）和缩聚结焦阶段（Stage-III），快速热分解阶段可再分为一次热解阶段（Stage-IIA）和二次热解阶段（Stage-IIB）。重点考察了主要组成元素（碳、氢、氧、氮）在热解产物中的分布与迁移规律，结果表明：碳元素在挥发性产物中的相对含量顺序为重质焦油>轻质焦油>热解气，即碳元素在慢速升温热解过程中逐渐向重质焦油和非挥发性组分中迁移富集；氢元素在慢速升温热解过程中主要向热解气中迁移富集；氧元素在挥发性产物中的相对含量顺序为热解气>轻质焦油>重质焦油，即氧元素在慢速升温热解过程中主要迁移富集到热解气中；氮元素在慢速升温热解的前两个阶段（模拟温度达2000 K前）未表现出明显的迁移倾向，而在缩聚结焦阶段迅速向热解气中迁移富集。应用化学反应分析与可视化软件VARxMD分析慢速升温热解的ReaxFF MD模拟结果，揭示了实验手段或其它计算方法难以直接获得的主要轻质气体产物（CO2、H2O、H2和CH4）的详细生成反应路径。结果表明：府谷次烟煤热解初期CO2、H2O和H2的生成反应都与羧基密切相关，CH4的初始生成反应则与甲氧基紧密关联，因此可以认为羧基和甲氧基等含氧官能团在府谷次烟煤热解初期，尤其是初始活化阶段发挥着至关重要的作用。利用ReaxFF MD在不同温度（1200、1400、1600、1800、2000、2200 K）条件下针对府谷次烟煤多组分模型开展了2000 ps的恒温热解模拟，考察了模拟温度和模拟时间尺度对府谷次烟煤的热解行为和产物产出的影响，获得了府谷煤在不同温度条件下的热解产物演化规律。与文献报道的府谷煤流化床热解实验结果相比，ReaxFF MD恒温模拟显著高估了煤焦油的产出、略微低估了热解气的产出。通过对府谷次烟煤高温较短时间尺度和低温较长时间尺度模拟的热解产物演化趋势的比较发现，1200–2200 K的模拟温度范围内，将模拟时间从2000 ps缩短为250 ps，则模拟温度需要提升400 K来获得相似的热解产物演化趋势，但会导致煤焦油和热解气产量的高估以及非挥发性产物产量的低估。升温模拟策略和模拟时间尺度对热解产物演化影响的探讨可为今后选择ReaxFF MD模拟策略提供依据。本论文提出的多组分煤模型的快速构建方法是构建可用于ReaxFF MD模拟煤热解的大规模煤模型的可行方法。合理的大规模煤模型结合基于高性能计算的ReaxFF MD模拟可以获得府谷次烟煤热解产物的全景式演化规律，有望扩展至其它低阶煤的热解模拟研究中，从而为更好地利用低阶煤提供理论支持。;Coal pyrolysis refers to the initial reaction step in most coal conversion processes, which plays a significant role in efficient and clean utilization of coal, especially for low-rank coals. Coal pyrolysis is well accepted as a radical driven process that involves myriad coupled reaction pathways with vast free radical intermediates generated. The heterogeneous nature of coal and the complexity of the pyrolysis process have made it very difficult to access the comprehensive mechanisms, even with the state-of-the-art experimental approach. When combined with molecular dynamics (MD), the ReaxFF reactive force filed based on bond order can describe the dynamic evolution of bond breaking and forming smoothly with accuracy close to the widely used Density Functional Theory (DFT) method with very much reduced computational costs.The overall pyrolysis process and the underlying chemical reactions of Fugu subbituminous coal are studied in the thesis by combining large-scale ReaxFF MD simulations and solid pyrolysis experiments with in situ synchrotron vacuum ultraviolet photoionization time-of-flight mass spectrometry (SVUV-PI-TOF-MS). The main results are summarized as the following.The pyrolysis experiments of Fugu subbituminous coal were performed using the solid pyrolysis apparatus with the volatile species detected by SVUV-PI-TOF-MS at the National Synchrotron Radiation Laboratory (NSRL) in Hefei, China. The dynamic profiles of major light gases and aromatic-based tar products with temperature were obtained by comparison of major pyrolyzates at varied temperature conditions (500, 600, 700, and 800 oC). The experimental results reveal that the existence of phenolic hydroxyl groups and enlargement of aromatic nucleus can both promote the reactivity of alkyl substituents in aromatic rings.A strategy was proposed for constructing large and reasonable coal molecular models with multi-components manually based on the limited conventional characterization data of coal samples (proximate and ultimate analysis, 13C NMR, and solvent extraction experiments). Both the structure diversity of macro coal molecules and model construction efficiency were considered in the proposed strategy. Local structure variation and average strategy for elemental composition are combined in the construction of macro coal molecules, which allows for structure diversity and easy match in a multi-component coal model with experimental characterization data of coal samples. Following the proposed strategy, a multi-component molecular model was constructed for Fugu subbituminous coal. The constructed Fugu coal model is a large model containing 23,898 atoms. It consist of 75 macromolecules with 20 varied average structures for structural diversity and 29 varied small compounds to capture the mobile phase. The element composition, aromaticity and averaged aromatic nucleus size of the Fugu coal model constructed are in good agreement with the experimental characterization of Fugu subbituminous coal.Slow heat-up (1 K/ps) and long-time (2000 ps) isothermal simulations were performed using the constructed large Fugu coal model with the GPU-enabled ReaxFF MD code of GMD-Reax to investigate the effects of heating rate and temperature on the coal pyrolysis. The reaction analysis code of VARxMD was used for uncovering the associated reaction mechanism in coal pyrolysis. Analysis of the heat-up ReaxFF MD simulation trajectories shows that Fugu coal pyrolysis process can be divided into three stages: the activation stage of coal structure (Stage-I), the primary pyrolysis (Stage-IIA) and the secondary pyrolysis stage (Stage-IIB), and the recombination dominated stage (Stage-III). The element distribution and migration among pyrolyzates in coal pyrolysis were analyzed. The content of C element in pyrolyzates are ranked as heavy tar > light tar > gas, which indicate that C element migrates to heavy tar and non-volatile molecules. While H element migrate to gas products. The content of O element in pyrolyzates are ranked as gas > light tar > heavy tar. The N element tends to stay in coal molecules that will migrate to gas products at stage-III of high temperature.By taking advantage of VARxMD for reaction analysis for ReaxFF MD simulations, detailed reaction pathways for the early gas product generation of CO2, H2O, H2 and CH4 in the activation and primary pyrolysis stages of Fugu subbituminous coal are revealed on the basis of heat-up simulations, which is hardly accessible experimentally or by other computational approach. It can be concluded that the gas product generation of CO2, H2O, H2 and CH4 in the activation and primary pyrolysis stages are closely associated with carboxyl and methoxyl groups, indicating the critical role of oxygen-containing groups in the initialization of subbituminous coal pyrolysis.Furthermore, the effects of elevated simulation temperature and simulation time on pyrolyzate profiles and yields were first investigated by the long time (2000 ps) isothermal ReaxFF MD simulations. Compared with reported experiments of fluidized bed pyrolysis, the coal tar yield is over-predicted, while the gas yield is slightly under-estimated. By comparing the weight profiles between the short and long duration time at high and low temperature conditions, it is found that to shorten the simulation time from 2000 ps to 250 ps, an increase of 400 K in average for simulation temperature is needed in range of 1200–2200 K to get similar pyrolyzate profiles that will result in the over-prediction for the yields of gas and tar, as well as under-prediction for the non-volatile yield. What obtained can be used in refining simulation strategy for coal pyrolysis study.The proposed strategy for fast constructing large and reasonable coal models manually with varied chemical structures is practical for construction of large multicomponent molecular models for coal pyrolysis study with ReaxFF MD. Combining with the reasonable and large coal molecular model, the ReaxFF MD simulation approach is useful for getting an overall scenario of coal pyrolysis and deeper insight into the complex pyrolytic reaction of Fugu subbituminous coal. The method can be applied in studying pyrolysis reaction mechanism of other low rank coals for their better utilization.Coal pyrolysis refers to the initial reaction step in most coal conversion processes, which plays a significant role in efficient and clean coal utilization, especially for low-rank coals.