Computer Spectroscopy on Classical and Quantum Computers

Event Date:
2019-03-04T13:30:00
2019-03-04T14:30:00
Event Location:
Hennings 318
Speaker:
Dr. Nike Dattani, University of Oxford, Chemistry, Research Scientist
Related Upcoming Events:
AMO Seminars
Intended Audience:
Graduate
Local Contact:

Kirk Madison

Event Information:

Ideally, the cataloging of spectroscopic linelists would not demand laborious and expensive experiments. If it were possible to obtain the exact same information by running a calculation on a computer, then when this information is needed for a new molecule, new isotopologue, new charge, new electronic state, or for new vibrational levels, we would not have to set up a new experiment; we could instead change some lines in our computer program's input file. 

 

QED (quantum electrodynamics) offers a description of electromagnetic interactions, QFD (quantum flavordynamics) also succeeds in describing weak-nuclear interactions, and QCD (quantum chromodynamics) succeeds in describing strong-nuclear interactions. Therefore a numerically converged QCD calculation (assuming available computing power) provides the fine, hyperfine, and finer than hyperfine splittings in atomic, molecular, and condensed matter spectra with only two physical approximations: neglect of gravity and neglect of "fifth" forces (a term used to describe any other missing piece which might make our calculation disagree with experiment). 

 

We have millions of high-precision experimental spectral lines to which we can compare our QCD-level calculations. In the cases where the calculations are converged to sub-cm-1 precision, they agree with experiments every time. We are therefore be confident that a spectral database numerically generated by a computer, can fill in gaps in the NIST database, despite our ignorance of how to deal with gravity and "fifth"

forces.

 

I will present results on the small systems where QCD-level calculations have been compared to experiments. The majority of the energy comes from the non-relativistic Schroedinger equation. A Monte Carlo implementation of FCI called FCIQMC allows for a numerically exact treatment of many electrons, and all other QED, QFD, and QCD interactions are treated with perturbation theory using the FCIQMC wavefunction. I will present FCIQMC results on up to 54 electrons, and perturbation theory corrections up to 7th order in the QED expansion for smaller active spaces. 

 

Going beyond 54 electrons is extremely difficult on a classical computer, where simulation of quantum mechanics is un-natural and scales exponentially with the number of electrons. It would be more natural if our computer treated the quantum effects at the hardware level. A quantum computer can do the calculations with cost scaling only polynomially with respect to the number of electrons, and I will present preliminary results for small calculations that are run on D-Wave's 2048 qubit annealer and IBM's 5-qubit machine (Yorktown), their 14-qubit machine (Melbourne), and their 20-qubit machine (Austin).

Add to Calendar 2019-03-04T13:30:00 2019-03-04T14:30:00 Computer Spectroscopy on Classical and Quantum Computers Event Information: Ideally, the cataloging of spectroscopic linelists would not demand laborious and expensive experiments. If it were possible to obtain the exact same information by running a calculation on a computer, then when this information is needed for a new molecule, new isotopologue, new charge, new electronic state, or for new vibrational levels, we would not have to set up a new experiment; we could instead change some lines in our computer program's input file.    QED (quantum electrodynamics) offers a description of electromagnetic interactions, QFD (quantum flavordynamics) also succeeds in describing weak-nuclear interactions, and QCD (quantum chromodynamics) succeeds in describing strong-nuclear interactions. Therefore a numerically converged QCD calculation (assuming available computing power) provides the fine, hyperfine, and finer than hyperfine splittings in atomic, molecular, and condensed matter spectra with only two physical approximations: neglect of gravity and neglect of "fifth" forces (a term used to describe any other missing piece which might make our calculation disagree with experiment).    We have millions of high-precision experimental spectral lines to which we can compare our QCD-level calculations. In the cases where the calculations are converged to sub-cm-1 precision, they agree with experiments every time. We are therefore be confident that a spectral database numerically generated by a computer, can fill in gaps in the NIST database, despite our ignorance of how to deal with gravity and "fifth" forces.   I will present results on the small systems where QCD-level calculations have been compared to experiments. The majority of the energy comes from the non-relativistic Schroedinger equation. A Monte Carlo implementation of FCI called FCIQMC allows for a numerically exact treatment of many electrons, and all other QED, QFD, and QCD interactions are treated with perturbation theory using the FCIQMC wavefunction. I will present FCIQMC results on up to 54 electrons, and perturbation theory corrections up to 7th order in the QED expansion for smaller active spaces.    Going beyond 54 electrons is extremely difficult on a classical computer, where simulation of quantum mechanics is un-natural and scales exponentially with the number of electrons. It would be more natural if our computer treated the quantum effects at the hardware level. A quantum computer can do the calculations with cost scaling only polynomially with respect to the number of electrons, and I will present preliminary results for small calculations that are run on D-Wave's 2048 qubit annealer and IBM's 5-qubit machine (Yorktown), their 14-qubit machine (Melbourne), and their 20-qubit machine (Austin). Event Location: Hennings 318