Graphene based nano-porous materials (GNM) are potentially useful for all those applications needing a large specific surface area (SSA), typical of the bidimensional graphene, yet realized in the bulk dimensionality. Such applications include for instance gas storage and sorting, catalysis and electrochemical energy storage. While a reasonable control of the structure is achieved in micro-porous materials by using nano-micro particles as templates, the controlled production or even characterization of GNMs with porosity strictly at the nano-scale still raises issues. These are usually produced using dispersion of nano-flakes as precursors resulting in little control on the final structure, which in turn reflects in problems in the structural model building for computer simulations. In this work, we describe a strategy to build models for these materials with predetermined structural properties (SSA, density, porosity), which exploits molecular dynamics simulations, Monte Carlo methods and machine learning algorithms. Our strategy is inspired by the real synthesis process: starting from randomly distributed flakes, we include defects, perforation, structure deformation and edge saturation on the fly, and, after structural refinement, we obtain realistic models, with given structural features. We find relationships between the structural characteristics and size distributions of the starting flake suspension and the final structure, which can give indications for more efficient synthesis routes. We subsequently give a full characterization of the models vs H 2 adsorption, from which we extract quantitative relationship between the structural parameters and the gravimetric density. Our results quantitatively clarify the role of surfaces and edges relative amount in determining the H 2 adsorption, and suggest strategies to overcome the inherent physical limitations of these materials as adsorbers. We implemented the model building and analysis procedures in software tools, freely available upon request.

中文翻译：

---

纳米多孔石墨烯支架的计算机设计、构建和气体吸附

基于石墨烯的纳米多孔材料 (GNM) 可能适用于所有需要大比表面积 (SSA) 的应用，这是典型的二维石墨烯，但在体积维度中实现。此类应用包括例如气体储存和分拣、催化和电化学能量储存。虽然通过使用纳米微粒作为模板在微孔材料中实现了对结构的合理控制，但严格在纳米尺度上具有孔隙率的 GNM 的受控生产甚至表征仍然存在问题。这些通常使用纳米薄片的分散体作为前体来生产，导致对最终结构的控制很少，这反过来又反映了计算机模拟结构模型构建中的问题。在这项工作中，我们描述了一种为这些具有预定结构特性（SSA、密度、孔隙率）的材料建立模型的策略，该策略利用分子动力学模拟、蒙特卡罗方法和机器学习算法。我们的策略受到真实合成过程的启发：从随机分布的薄片开始，我们在运行中包括缺陷、穿孔、结构变形和边缘饱和，并且在结构细化后，我们获得具有给定结构特征的真实模型。我们发现了起始薄片悬浮液的结构特征和尺寸分布与最终结构之间的关系，这可以为更有效的合成路线提供指示。我们随后给出了模型与 H 2 吸附的完整表征，我们从中提取结构参数和重量密度之间的定量关系。我们的结果定量地阐明了表面和边缘相对量在确定 H 2 吸附中的作用，并提出了克服这些材料作为吸附剂的固有物理限制的策略。我们在软件工具中实施了模型构建和分析程序，可根据要求免费提供。