RSS icon
Twitter icon
Facebook icon
Vimeo icon
YouTube icon

The importance of the spectral gap in estimating ground-state energies

October 2, 2020 - 12:00pm
Speaker: 
Abhinav Deshpande

The field of quantum Hamiltonian complexity lies at the intersection of quantum many-body physics and computational complexity theory, with deep implications to both fields. The main object of study is the LocalHamiltonian problem, which is concerned with estimating the ground-state energy of a local Hamiltonian. A major challenge in the field is to understand the complexity of the LocalHamiltonian problem in more physically natural parameter regimes. One crucial parameter in understanding the ground space of any Hamiltonian in many-body physics is the spectral gap, which is the difference between the smallest two eigenvalues. Despite its importance in quantum many-body physics, the role played by the spectral gap in the complexity of the LocalHamiltonian is less well-understood. In this work, we make progress on this question by considering the "precise" regime, in which one estimates the ground-state energy to within inverse exponential precision.

In this setting of inverse-exponential precision, there is a surprising result that the complexity of LocalHamiltonian is magnified from a class called QMA, the quantum generalization of NP to a class called PSPACE, the class of problems solvable in polynomial space (but possibly exponential time). We clarify the reason behind this boost in complexity. Specifically, we show that the full complexity of the high precision case only comes about when the spectral gap is exponentially small. As a consequence of the proof techniques developed to show our results, we uncover important implications for the representability and circuit complexity of ground states of local Hamiltonians, the theory of uniqueness of quantum witnesses, and techniques for the amplification of quantum witnesses in the presence of postselection.

Seminar will be held via Zoom: https://umd.zoom.us/s/97099328991 and Meeting ID : 970 9932 8991

Zoom