Noa Marom

Molecular crystals have applications in nonlinear optics, organic electronics, and particularly in pharmaceuticals because most drugs are marketed as crystals of the pharmaceutically active ingredient. Molecular crystals are held together by van der Waals (vdW) interactions (also known as dispersion interactions) between molecules. Unlike chemical bonds, van der Waals interactions do not involve overlap of electron densities. Rather, they arise from quantum fluctuations of the electron density that lead to the formation of dipoles and higher order multipoles. The electrostatic interaction between these generates a weak but long-ranged attractive force. Owing to the weak nature of van der Waals interactions, a given molecule may crystallize in more than one structure. This is known as polymorphism. Polymorphic forms of the same molecule may possess markedly different physical and chemical properties. Crystal structure may profoundly influence the bioavailability, toxicity, manufacturability, and stability of drugs. In the context of technological applications, crystal structure affects the electronic and optical properties. We use computer simulations to perform structure prediction and design of molecular crystals from first principles, based solely on the knowledge of their elemental composition and the laws of quantum mechanics. We develop genetic algorithms, which are guided to the most promising regions of the configuration space by the evolutionary principle of survival of the fittest. Offspring are generated by combining structural “genes” of the fittest structures in the population to propagate desirable features, while random mutations are employed to maintain diversity. We are particularly interested in optimizing crystal packing for high-performance organic electronics and solar cells.

Computational materials design offers tremendous potential for discovery and innovation. This powerful concept relies on computational exploration of the vast configuration space of materials structure and composition to identify promising candidates with desired properties for target applications. In fact, many applications do not rely on a single material but on the combination of several materials in a functional nano-structure. Examples for functional nano-structures include the dye-oxide interface, at which charge separation is achieved in dye-sensitized solar cells, and nanocatalysts based on clusters dispersed on a large surface area support. Therefore, we would like to design not just a material, but a functional nano-structure. This requires the combination of accurate electronic structure methods with efficient optimization algorithms.
The electronic properties and the resulting functionality of a nano-structure cannot be deduced directly from those of its isolated constituents. Rather, they emerge from a complex interplay of quantum mechanical interactions that depend on the local environment at the nano-scale. Describing these effects requires a fully quantum mechanical first principles approach. In the first part of the talk, many-body perturbation theory within the GW approximation, where G is the one-particle Green’s function and W is the screened Coulomb interaction, is used to elucidate the size effects in the energy level alignment at the interface between dye molecules and TiO2 clusters of increasing size.
In the second part of the talk, a new approach is presented for computational design of clusters using property-based genetic algorithms (GAs). These algorithms perform optimization by simulating an evolutionary process, whereby child structures are created by combining fragments (“mating”) of the fittest parent structures with respect to the target property. Property-based GAs tailored to search for low energy, high vertical electron affinity (VEA), and low vertical ionization potential (VIP) are applied to TiO2 clusters with up to 20 stoichiometric units. Analysis of the resulting structures reveals the structural features associated with a high VEA and a low VIP and explains the absence of the expected size trends.