PsiQuantum's take on quantum chemistry using quantum computers

PsiQuantum takes a contrarian view on quantum computing: forget NISQ, focus on fault-tolerant systems with millions of qubits and error correction.

A few weeks ago PsiQuantum published a paper analyzing the resources required to carry out commercially-relevant simulations of battery chemistry: Fault-tolerant resource estimate for quantum chemical simulations: Case study on Li-ion battery electrolyte molecules. They also posted a summary on their website.

So far I only had a chance to skim the paper, but as someone focusing on NISQ algorithms I found the different approaches required for fault-tolerant algorithms quite interesting. For example, I had thought that state preparation might be a bottleneck in applying quantum phase estimation to classically-intractable quantum systems, but this doesn't appear to be a problem for high precision quantum chemistry applications:

Concerns over the viability of preparing such an ansatz efficiently are encapsulated by the “orthogonality catastrophe,” the observation that the overlap between the true ground state and some ansatz wave function decreases exponentially as the system size increases. However, it has been shown that there are sophisticated classical methods for preparing trial wavefunctions with sufficient overlap, for states of up to O(100) orbitals using simple-to-prepare states such as the single Slater determinant obtained from the Hartree-Fock method (alternatively, methods for multideterminant state preparation can be used) [arXiv:1809.05523]

Another part of the paper I found interesting was Appendix C, which presents a logarithmic-depth circuit for state preparation using Givens rotations. From what I understand, this construction is limited to the fault-tolerant regime, since it requires multi-qubit Pauli operations.