PQS (software)

The origin of PQS program was developed by Meyer and Pulay in the late 1960s. They both were at the Max-Planck Institute for Physics and Astrophysics in Munich when they began to write a new ab initio program.[1] The main purpose was to establish new ab initio techniques. Pulay and Meyer had slightly different interests.[1] Pulay was interested in implementing gradient geometry optimization, analytical energy derivatives (force), and force constant calculations via the numerical differentiation of analytical forces, while Meyer enthused about the coupled-electron pair approximation (CEPA), spin density calculation, and extremely accurate correlation methods like pseudonatural orbital-configuration interaction (PNO-CI).[1] At that time, analytical gradients were limited to closed-shell Hartree-Fock wavefunctions. However, they were able to do it for unrestricted (UHF) and restricted open-shell (ROHF) methods in 1970. The first version of the code was completed in 1969 at Max-Planck institute and University of Stuttgart. Then, Meyer named it “MOLPRO” and used Gaussian lobe basis sets.[1] In the 1970s, the current version of MOLPRO added a number of advanced methods such as multiconfiguration self-consistent field (MC-SCF) and internally contracted multireference configuration interaction (MR-CI). Simultaneously, in the 1980s, MOLPRO was extended and mostly rewritten by Hans-Joachim Werner, Peter Knowles and Meyer's coworkers.[1]

Meanwhile, in 1976, Pulay was visiting Boggs at the University of Texas, Austin and Schaefer at the University of California. They wrote a new program called TEXAS based on the original MOLPRO and replaced Gaussian lobe functions with the standard Gaussian functions.[1] TEXAS emphasized large molecules, SCF convergence, geometry optimization techniques, and vibrational spectroscopy-related calculations. From 1982 onward, the program was further developed at the University of Arkansas.[1]

The primary significant expansion was the usage of a few new electron correlation methods by Saebo and a first- order MC-SCF program by Hamilton. A critical option was the implementation of the first practical gauge-invariant atomic orbital (GIAO) NMR program by Wolinski, who additionally included a highly efficient integral package.[1] Bofill executed an unhindered natural orbital-complete active space (UNO-CAS) program including analytical gradients; this is a minimal-cost alternative to MC-SCF and works just as well in most cases. TEXAS was initially parallelized in 1995–1996 on a cluster of 10 IBM RS6000 workstations.[1]

In 1996, Baker joined Pulay and, around the same time, Intel brought out the Pentium Pro, a processor for PC that was competitive with low-end workstations and less costly by around an order of magnitude. Understanding the capability of this improvement for computational chemistry, PQS was formed and a SBIR grant application was submitted in July 1997 for the commercial development of PC clusters for parallel ab initio calculations.[1] In the meantime, the Pulay group, financed by a National Science Foundation grant, set about constructing a Linux cluster using 300 MHz Pentium II processors. By a fortunate circumstance, a few capable and PC proficient graduate students were available, specifically Magyarfalvi and Shirel.[1] The PC cluster was a complete success, and significantly outperformed the IBM Workstation cluster that was the group's computational mainstay at a small amount of its expense.[1]

The PQS programming was demonstrated on the TEXAS code and parts of it, chiefly the NMR code, were authorized to PQS from the University of Arkansas.[1] Much of the code was widely changed to comply with the twin points of (a) having all major functionality fully parallel; and (b) having the capacity to routinely perform calculations on extensive systems.[1] They aimed primarily for a modest level of parallelism (from 8 to 32 CPUs), as this is the most widely recognized size for an individual or group resource. Indeed, even on very large clusters it is normal for any given user to be allocated only a percentage of the available processors.[1]

The basis capabilities in high-level correlated energies for PQS ab initio v. 4.0 include MP3, MP4, CID, CISD, CEPA-0, CEPA-2, QCISD, QCISD(T), CCD, CCSD and CCSD(T) wavefunctions; enforced geometry optimization (used, among other things, to simulate the results of Atomic Force Microscopy (AFM) experiments); full-accuracy, canonical UMP2 energies and analytical polarizabilities and hyperpolarizabilities for HF and DFT wavefunctions.[2]

• An efficient vectorized Gaussian integral package allowing high angular momentum basis functions and general contractions.
• Abelian point group symmetry throughout; utilizes full point group symmetry (up to Ih) for geometry optimization step and Hessian (2nd derivative) CPHF.
• Closed-shell (RHF) and open-shell (UHF) SCF energies and gradients, including several initial wavefunction guess options. Improved SCF convergence for open-shell systems.
• Closed-shell (RHF) and open-shell (UHF) density functional energies and gradients including all popular exchange-correlation functionals: VWN, B88, OPTX, LYP, P86, PW91, PBE, B97, HCTH, B3LYP, make up your own functional etc.
• Fast and accurate pure DFT energies and gradients for large basis sets using the Fourier Transform Coulomb (FTC) method.
• Productive, adaptable geometry optimization for every one of these methods including Eigenvector Following (EF) algorithm for minimization and saddle-point search, Pulay's GDIIS algorithm for minimization, use of Cartesian, Z-matrix and delocalized internal coordinates. Includes new coordinates for efficient optimization of molecular clusters and adsorption/reaction on model surfaces.[3]
• Full range of geometrical constraints including fixed distances, planar bends, torsions and out-of-plane bends between any atoms in the molecule and frozen (fixed) atoms. Atoms involved in constraints do not need to be formally bonded and - unlike with a Z matrix - desired constraints do not need to be satisfied in the starting geometry.
• Expository second subsidiaries for every one of these systems, including the calculation of vibrational frequencies, IR intensities and thermodynamic analysis.[3]
• Efficient NMR Chemical Shifts for closed-shell HF and DFT wavefunctions.
• A full range of effective core potentials (ECPs), both relativistic and non-relativistic, with energies, gradients, analytical second derivatives and NMR.
• Closed-shell MP2 energies and analytical gradients and dual-basis MP2 energies; numerical MP2 second derivatives.
• Potential scan, including scan + optimization of all other degrees of freedom.
• Reaction Path (IRC) following using either Z-matrix, Cartesian or mass-weighted Cartesian coordinates.
• Conductor-like screening solvation model (COSMO) including energies, analytical gradients, numerical second derivatives and NMR.
• Population analysis, including bond orders and atomic valencies (with free valencies for open-shell systems); CHELP and Cioslowski charges.
• Weinhold's Natural Bond Order (NBO) analysis, including natural population and steric analysis.
• Properties module with charge, spin-density and electric field gradient at the nucleus.
• Polarizabilities and dipole and polarizability derivatives; Raman intensities.
• Full Semiempirical package, both open (unrestricted) and closed-shell energies and gradients, including MINDO/3, MNDO, AM1 and PM3. For the latter, all main group elements through the fourth row (except the noble gases) as well as Zinc and Cadmium, have been parametrized.
• Molecular Mechanics using the Sybyl 5.2 and UFF Force Fields.
• QM/MM using the ONIOM method.
• Molecular dynamics using the simple Verlet algorithm.
• Pople-style input for quick input generation and compatibility with other programs.
• Graphical input generation and display
• All major ab initio functionality is fully parallel (except MP2 gradients which is serial only - parallel version under development).
• Calculate molecular structure and vibrational spectra for transition state, infrared (IR), Raman and Vibrational circular dichroism (VCD).[4]

• Parallel Quantum Solutions