Molecular materials are of great fundamental and applied importance in science and industry and provide fertile ground for the continuing development of theoretical methods. The applications of molecular materials are wide, ranging from pharmaceuticals to molecular sensors, hybrid organic-inorganic interfaces, as well as energy capture using dye-sensitised solar cells. With all of these potential applications, the design and control of molecular materials is one of the grand challenges of physical and chemical research. Computational modelling has much to offer in this regard. Designing, synthesising, and experimentally isolating and characterising molecular materials is very costly and time consuming. Computational approaches to molecular-materials design can enable scientists to explore the properties of materials in silico, identifying key properties of interest before the costly process of obtaining the material in a laboratory. For example, the activity of a drug molecule in binding to a key enzyme or protein is key to its application, but there can be many other properties, such as toxicity, solubility and dissolution rate, which can have a huge impact on its suitability for end use. Understanding how atomic-level interactions, structure, and dynamics in molecular materials dictate such macroscopic properties is essential for rational material design. Computational approaches that can successfully achieve this, leading to predictions of macroscopic properties of interest to industry and consumers, are highly desirable but remain largely elusive.
First-principles methods offer an attractive way to understand and rationalise the design of new materials, as they are not biased towards specific systems or properties. However, there are a number of challenges in applying such methods to molecular systems. The computational cost of first-principles electronic structure methods is typically far higher than that of those employing empirical potentials, and can scale rapidly with the size of the system, limiting the time and length scales accessible to these methods, ultimately requiring first-principles derived empirical potentials and multiscale approaches. There are also a plethora of first-principles approaches, many of which are suited only to specific properties. While DFT can provide reasonable molecular geometries it suffers from severe problems in predicting electronic properties, which calls for much more expensive many-body approaches. In addition, exploring the vast conformational and structural flexibility of molecular materials is challenging even for empirical potentials: predicting the stable polymorphic forms of a molecular crystal might require considering thousands of different structures. Even more challenging is the study of the kinetics of crystal growth and dissolution, which depend heavily on the chemical environment and solvent. This task requires the simulation of interfaces and of a large number of solvent molecules, which will remain a formidable computational challenge for first-principles methods for the foreseeable future.
In recent years there has been considerable progress in many areas related to molecular materials. The field of crystal-structure prediction is maturing, with recent blind trials showing many successes . DFT in particular is now capable of correctly ranking many challenging polymorphic systems . These improvements have gone hand-in-hand with a deeper understanding of the role of non-covalent interactions, particularly van der Waals interactions [3-5], in governing cohesive properties of molecular crystals and supramolecular assemblies. Synergistic, combined theoretical and experimental studies of structure and dynamics in molecular systems have also flourished in the past decade . Many groups are also working on developing and applying DFT and many-body methods for predicting electronic properties of solids and charge-transfer systems [7, 8]. In the field of simulation methodology there has been significant developments in multiscale techniques for modelling crystal growth and habit prediction [9, 10]. New techniques have been introduced for including nuclear quantum effects in molecular simulations at a fraction of the cost of path-integral simulations , paving the way for a better understanding of hydrogen bonding and proton transfer in molecular systems. Advances in linear-scaling codes  and explicitly correlated wavefunction methods  are also pushing the boundaries of first-principles methods.
These advances clearly point to the maturity of the field. However, many of these developments have taken place within self-contained communities, which focus on specific areas, such as structure, electronic properties, code development, simulation techniques. These boundaries are a hindrance to the development of unified frameworks for modelling molecular materials, limiting the scope of new developments and in some cases leading to computational methods that are mutually exclusive for different applications. Therefore, there is an urgent need for dialogue between the different communities. Achieving the goal of predictively evaluating the macroscopic properties of molecular materials will require a concerted effort from the different communities.