This Title All WIREs
How to cite this WIREs title:
WIREs Comput Mol Sci
Impact Factor: 16.778

TeraChem: A graphical processing unit‐accelerated electronic structure package for large‐scale ab initio molecular dynamics

Full article on Wiley Online Library:   HTML PDF

Can't access this content? Tell your librarian.

Abstract TeraChem was born in 2008 with the goal of providing fast on‐the‐fly electronic structure calculations to facilitate ab initio molecular dynamics studies of large biochemical systems such as photoswitchable proteins and multichromophoric antenna complexes. Originally developed for videogaming applications, graphics processing units (GPUs) offered a low‐cost parallel computer architecture that became more accessible for general‐purpose GPU computing with the release of CUDA in 2007. The evaluation of the electron repulsion integrals (ERIs) is a major bottleneck in electronic structure codes and provides an attractive target for acceleration on GPUs. Thus, highly efficient routines for evaluation of and contractions between the ERIs and density matrices were implemented in TeraChem. Electronic structure methods were developed and implemented to leverage these integral contraction routines, resulting in the first quantum chemistry package designed from the ground up for GPUs. This GPU acceleration makes TeraChem capable of performing large‐scale ground and excited state calculations in the gas and condensed phase. Today, TeraChem's speed forms the basis for a suite of quantum chemistry applications, including optimization and dynamics of proteins, automated and interactive chemical discovery tools, and large‐scale nonadiabatic dynamics simulations. This article is categorized under: Electronic Structure Theory > Ab Initio Electronic Structure Methods Software > Quantum Chemistry Structure and Mechanism > Computational Biochemistry and Biophysics
Performance of single point energy and gradient calculations for TrpCage (PDB ID: 2JOF, left), a BuckycatcherII complex (center), and bovine pancreatic trypsin inhibitor (PDB ID: 6PTI, right) over several generations of GeForce and Tesla GPUs. The same Hartree–Fock implementation is used on all GPUs and speedups are reported compared to the GeForce GTX 680 for each molecule. Each calculation used a single GPU and single core of an Intel Xeon CPU, either a 2.8 GHz E5‐2680 (K80), a 2.5 GHz E5‐2640 CPU (680/970), a 2.6 GHz E5‐2660v3 (P100), or a 3.4 GHz E5‐2643v4 (1080Ti/V100)
[ Normal View | Magnified View ]
TeraChem's capability to simulate ab initio molecular dynamics for proteins as GPU hardware and algorithms improve, from ground state dynamics and optimization in 2011135,136 to multireference nonadiabatic dynamics in 2019141
[ Normal View | Magnified View ]
Schematic overview of the ab initio nanoreactor framework. Reaction networks are iteratively built up through three phases: reaction discovery through accelerated molecular dynamics, rate determination through minimum energy pathway optimization, and running kinetic models to generate new concentration profiles for discovery runs and pinpoint rate‐limiting intermediates for further refinement
[ Normal View | Magnified View ]
Mechanochemical ring opening of benzocyclobutene using real‐time interactive molecular dynamics with GFN2‐xTB. Dynamic bond orders and molecular orbitals show (a) the sigma bond in the reactant, (b) the bond rearrangement at the transition state, and (c) the new π orbital in the product
[ Normal View | Magnified View ]
The ab initio exciton model with locally excited (LE) and charge transfer (CT) states applied to the 18 BChl‐a chromophore B850 assembly of the light harvesting system II (LH2). In recent work, Li et al.110 benchmarked against TDDFT and EOM‐CC2 calculations of the supersystem to show the robustness and accuracy of the exciton model
[ Normal View | Magnified View ]
Timings (in seconds) for hybrid QM/MM calculations of myoglobin (PDB ID: 3RGK) at the HF‐CAS‐(16,16)‐CI/6‐31G level of theory using a single V100 GPU and a 3.4 GHz Intel Xeon E5‐2643v4 CPU. The various QM regions were carved out by including entire residues within a given distance of the heme cofactor. Note that the CASCI portion of the calculation dominates for small QM sizes but remains fairly constant and is heavily outweighed by the SCF procedure for the entire protein
[ Normal View | Magnified View ]
Computational wall times of THC‐MP2 calculations, where bar colors represent different components, that is, constructing the LS‐THC‐MO tensors, MP2 Coulomb‐like energies, and MP2 exchange‐like energies. Calculations use cc‐pVDZ basis set, cc‐pVDZ‐RI auxiliary basis set for density fitting, and THC grids optimized for cc‐pVDZ. Timings were run using a single GTX 980 or 1080Ti GPU and a single thread of a 3.33 GHz Intel Xeon X5680 CPU
[ Normal View | Magnified View ]
Comparison of the double and dynamic precision schemes7 between Pascal‐generation GeForce and Tesla GPUs. The same Hartree–Fock implementation is used on all GPUs and speedups are reported compared to double precision on the GeForce GTX 1080Ti for each molecule. Each calculation used a single GPU and single core of an Intel Xeon CPU as described in Figure 1
[ Normal View | Magnified View ]

Browse by Topic

Structure and Mechanism > Computational Biochemistry and Biophysics
Software > Quantum Chemistry
Electronic Structure Theory > Ab Initio Electronic Structure Methods

Access to this WIREs title is by subscription only.

Recommend to Your
Librarian Now!

The latest WIREs articles in your inbox

Sign Up for Article Alerts