Density Functional Tight Binding Methods

Density Functional Tight Binding (DFTB) methods have been shown to be capable of producing reliable molecular structures and energetics at a significantly reduced computational cost. The development and characterization of such methods will permit calculations of properties and simulations of dynamic chemical processes for molecules with many thousands of atoms. This capability will facilitate the study of proteins, nanoparticles, and other interesting species in realistic chemical environments. Ultimately, these methods will aid in the interpretation of experimental data and in the optimization of molecules targeting specific properties.

Presently the study by quantum chemistry techniques of large (several thousands of atoms) molecular systems is limited by large computer time and space (memory, disk) requirements. The required time, memory, and disk space to solve the requisite equations typically increase exponentially with the number of atoms in the systems whereas the speed and storage capacity of computers is increasing only linearly. In order to meet the requirements of the current and next generation of research topics, reliable methods which scale much more favorably with the problem size must be implemented. Tight binding (TB) methods based on density functional theory (DFT) have been shown to produce results of good accuracy at a significantly reduced computational expense. The use of such methods allows computational investigation of large molecular systems on resources which are readily available to most researchers.

Many chemical problems of current interest involve large numbers of atoms (relative to the current capability). Examples of such problems include the simulation of proteins in aqueous solution, prediction of the structure and properties of core-shell nanostructures, prediction of the current-voltage behavior of nano-molecular electronics, and simulation of aerosols. The implementation of the DFTB method described above will permit study of each of these systems via quantum chemical techniques, something which is difficult or impossible with traditional quantum chemistry methods.

In FY2008, our goal was to take the existing DFTB base code and extend it in a number of ways. Code was developed to (a) optimize molecular geometries; (b) compute vibrational frequencies; (c) compute thermodynamic quantities; (d) achieve (approximate) linear scaling in computational cost; (e) permit parallel execution on a wide variety of parallel architectures; and (f) assist users in fitting application-specific parameters. Parameter sets for simulations of amino acids and aluminum clusters were also developed.

Publications

Hasmy, A., Rincon, L., Hernandez, R., Mujica,V., Marquez, M., Gonzalez, C., On the Formation of Suspended Noble-Metal Monatomic Chains, Phys., Rev. B, 2008, 78, pp. 115409

• computation of nuclear Hessian and vibrational frequencies
  
• computation of thermochemical quantities
  
• implementation of sparse matrix techniques
  
• parallelization of code
  
• development of parameter fitting techniques
  
• development of parameters for amino acid and metallic systems