Efficient global structure optimization in confined spaces: development and applications for design of advanced materials
Theoretical Chemistry: Electronic Structure, Dynamics, Simulation
Final Report Abstract
Full realization of the application potential of nanoparticular systems (NS) is faced with a number of challenges including fabrication of NS with a narrow size distribution allowing to control and finetune their properties, and assembly of NS into large (even macroscopic) and well-ordered superstructures and superlattices. A common avenue to address these challenges is encapsulation of NS in a host matrix. However, inclusion of embedding environment makes accurate theoretical description using first principle methods very demanding, even for such an essential step in the NS characterization, as is the detailed characterization of their atomic structure. Therefore within this project a tool for global structure optimization for clusters in confined spaces, GOCONS, has been developed, allowing location of the most stable structures of NS in host matrices in a computationally efficient way and its potential demonstrated for a system of current interest - PbS clusters confined in the zeolite host. The GOCONS tool employs the global structure optimization method based on well-established and efficient heuristics - the genetic algorithm (GA). Since the stability of confined clusters is no longer a function of their structure only but also of their location within the environment, this tool introduces new crossover operators for GA that reflect this fact and act differently depending on the strength of the cluster interaction with the environment. The same flexibility and capability to discriminate between otherwise similar cluster structures, which, however, might be found in different orientation within the environment, is also built-in to a very fast similarity recognition scheme, which ensures the diversity of the GA population. An obvious restriction imposed on the clusters to be located within the boundaries of the confinement is realized by dissecting the space around the host on-the-fly into sets of tetrahedra representing either the inner or outer voids using an algorithm combing weighted Voronoi decomposition of the environment with the alpha shape theory. To improve the computational efficiency of the GOCONS tool we implemented the lowmemory iterative density fitting (LMIDF) method that is applied in density functional theory (DFT)- based local geometry optimization carried out for each member of the cluster/host population. LMIDF reduces memory requirements of DFT calculations at least by half and thus all of the hundreds of DFT calculations that need to be carried for each GA run can be run over fewer cores achieving improved scalability of GA with the number of available CPU cores. The performance of the GOCONS tool was tested for PbS clusters encapsulated in the zeolite host – a system reported to exhibit extremely high nonlinear optical properties. Besides determining the structure of the embedded clusters we have also investigated their electronic, optical properties and determined the character of the interaction with the confining environment. We have observed sizable changes in cluster structures and even isomer stability ordering either with respect to the gas phase or as a result of the change of the zeolite composition (extraframework cations and Si:Al ratio). In some cases, even different types of isomers (wrt. to gas phase) have been identified as the most stable once the clusters become encapsulated in the host. For H-exchanged zeolite unexpected [PbS)nH]+ moieties were found to be formed upon cluster embedding. Presence of very small clusters in the zeolite cavities such as monomers or dimers is predicted since the energy gains due to embedding are significant being comparable in size with the gains due the cluster-enlargement. The stabilization is mostly due to dispersion and to lesser extent electrostatic interaction with the host while covalent bonding with the host can be largely disregarded. This is also confirmed by the calculated optical excitations of the PbS-zeolite composite, which are confined mostly to the PbS cluster, i.e., the zeolite can be regarded as a confining dielectric matrix that only modulates the optically active states of the PbS QDs. Importantly, the predicted spectra compare favorably to the experimental ones supporting the observations made regarding the structure of the PbS-zeolite system, preferred size of the embedded PbS cluster or the nature of the PbS-zeolite interaction. However, sizable computational effort is needed as inclusion of relativistic effects (spin-orbit coupling) at the hybrid DFT level is essential for proper assignment of the adsorption bands.
Publications
- Low-memory iterative density fitting. Journal of Computational Chemistry 2015, 36, 1521–1535
Grajciar, L.
(See online at https://doi.org/10.1002/jcc.23961) - Density functional theory for molecular and periodic systems using density fitting and continuous fast multipole method: Analytical gradients. Journal of Computational Chemistry 2016, 37, 2518–2526
Lazarski, R.; Burow, A. M.; Grajciar, L.; Sierka, M.
(See online at https://doi.org/10.1002/jcc.24477) - PbS Clusters Embedded in Sodalite Zeolite Cavities of Different Compositions: Unraveling the Structural Evolution and Optical Properties Using ab Initio Calculations. The Journal of Physical Chemistry C 2016, 120, 27050–27065
Grajciar, L.
(See online at https://doi.org/10.1021/acs.jpcc.6b09423)
