I converted the set of 134,000 molecules published in Scientific Data into a NoSQL database powered by MongoDB. This dataset includes properties such as structures, energies, charges and frequencies calculated with high-quality quantum mechanical methods (at the B3LYP or G4MP2 level). All the data were released as a set of text files, which are not straightforwardly searchable.

I wrote a php-based web interface to the database that allows one to search it using complex queries for molecule properties. The user can then plot the data to visualize values and trends, and to study correlations. Furthermore, a "details" page summarizes all the properties available for a given molecule and displays an interactive 3D structure. The database back-end is pleasantly fast: the entire dataset can be searched within seconds, and plots of up to 15000 data points are generated on the fly with minimal latency.

The GDB9 Database Browser can be found at: https://gdb.koitz.info/gdbrowse. Please give it a try -- I'll be happy to hear any feedback you may have.

Since the advent of Graphene, two-dimensional sheets and networks have come to the forefront of materials research. However, preparing and isolating tailored two-dimensional graphene analogs has remained a challenge. Recently, a new method to prepare 2D polymers was presented, employing suitable multifunctional monomers confined on a water surface and using metal ions from the liquid phase to link them via coordination bonds.

I study the molecular details of sheet formation from a prototype molecule (TTPB) on water. I use large-scale quantum mechanical simulations (molecular dynamics) to investigate the properties of the monomer on water, and the binding of Zn ions from the liquid. The image to the right shows the path of a Zn ion as it moves through liquid water to the binding site of TTPB.

Molecular sheets such as graphene and hexagonal boron nitride (h-BN) can be adsorbed on many metal surfaces, such as ruthenium, rhodium, copper, nickel, iron, and others. These systems are very promising for applications in nanoelectronics, chemical sensing and molecular assembly through templating. Depending on the interaction strength between h-BN and the metal, as well as on the lattice mismatch between the two materials, different structural phenomena such as corrugation and buckling can be observed.

As part of my doctoral project I studied h-BN adsorbed on a Cu(111) surface. Here the lattice mismatch is rather small and the interaction with the metal weak, so that the monolayer remains almost perfectly flat as it is adsorbed. However, experimental evidence with some theoretical input from our side has shown that the BN-layer is nevertheless corrugated electronically [Joshi, Ecija, RK, et al, Nano Letters, 2012]. The reason for this could be traced to the lateral variation of the Cu-BN distance, due to the slight incommensurability of the structures [RK et al, Nanoscale, 2013]. More recently I have been looking at h-BN on Ni and various other metals in the context of O₂ activation for catalysis and fuel cell applications.

Metal nanoclusters are an interesting and versatile class of materials and find applications in many fields of nanoscience, especially heterogenous catalysis. At cluster sizes larger than a threshold of about 100 atoms properties of clusters become scalable, i.e. proportional to the cluster size, and can thus be extrapolated to bulk quantities. The present thesis studied the scalability of structural, energetic, and magnetic cluster properties, using a set of ten exchange-correlation functionals, among them some very recently published ones. A set of six M_n clusters, where n = 13, 19, 38, 55, 79, 147 and M = Pd and Au, was studied, focusing on the average intermetallic distance d_av, the cohesive energy E_coh, and for Pd the Ionization Potential and Electron Affinity. These quantities scale with the cluster size. They can be fitted with linear functions of n^-1/3, a quantity proportional to the cluster radius.

While all of the functionals predict the scalable behavior of the clusters and the corresponding size-dependent trends, none of them can be considered "optimal" for the computation of all of the quantities. The results indicate a fundamental trade-off between accurate geometries and accurate energies, which seem to be mutually exclusive within the set functionals. The four novel GGA functionals of the VMT- and VT{84} families are also subject to this fact, and do not improve on the results by the older GGA methods. The use of a meta-GGA functional, M06-L, does not improve results, either; indeed, it yields unrealistically long bond lengths, and energies one order of mangitude worse than those found with the best GGA functionals. From the observations made in this study it is evident that the choice of XC approximation has a profound influence on the obtained results (geometries differing by 8 pm and energies by up to 160 kJ/mol), so that a judicious selection of functional according to the desired property is crucial.

Download the pdf here