In quantum many-body systems, the existence of a spectral gap above the ground state has far-reaching consequences. In this paper, we discuss "finite-size" criteria for having a spectral gap in frustration-free spin systems and their applications.We extend a criterion that was originally developed for periodic systems by Knabe and Gosset-Mozgunov to systems with a boundary. Our finite-size criterion says that if the spectral gaps at linear system size n exceed an explicit threshold of order n −3/2 , then the whole system is gapped. The criterion takes into account both "bulk gaps" and "edge gaps" of the finite system in a precise way. The n −3/2 scaling is robust: it holds in 1D and 2D systems, on arbitrary lattices and with arbitrary finite-range interactions. One application of our results is to give a rigorous foundation to the folklore that 2D frustration-free models cannot host chiral edge modes (whose finite-size spectral gap would scale like n −1 ).
arXiv:1801.08915v2 [quant-ph] 3 May 2019It is well-known that the existence of a spectral gap may depend on the imposed boundary conditions, and this fact is at the core of our work.In general, the question whether a quantum spin system is gapped or gapless is difficult: In 1D, the Haldane conjecture [15,16] ("antiferromagnetic, integer-spin Heisenberg chains are gapped") remains open after 30 years of investigation. In 2D, the "gapped versus gapless" dichotomy is in fact undecidable in general [10], even among the class of translation-invariant, nearest-neighbor Hamiltonians.This paper studies the spectral gaps of a comparatively simple class of models: 1D and 2D frustration-free (FF) quantum spin systems with a non-trivial boundary (i.e., open boundary conditions). (A famous FF spin system is the AKLT chain [1]. One general reason why FF systems arise is that any quantum state which is only locally correlated can be realized as the ground state of an appropriate FF "parent Hamiltonian" [12,32,35].)Specifically, we are interested in finite-size criteria for having a spectral gap in such systems. Let us explain what we mean by this.Let γ m denote the spectral gap of the Hamiltonian of interest, when it acts on systems of linear size m. Letγ n be the "local gap", i.e., the spectral gap of a subsystem of linear size up to n. (We will be more precise later.) The finite-size criterion is a bound of the form γ m ≥ c n (γ n − t n ), (1.1) for all m sufficiently large compared to n (say m ≥ 2n).Here c n > 0 is an unimportant constant, but the value of t n (in particular its ndependence) is critical. Indeed, if for some fixed n 0 , we know from somewhere that γ n 0 > t n 0 , then (1.1) gives a uniform lower bound on the spectral gap γ m for all sufficiently large m. Accordingly, we call t n the "local gap threshold".The general idea to prove a finite-size criterion like (1.1) is that the Hamiltonian on systems of linear size m can be constructed out of smaller Hamiltonians acting on subsystems of linear size up to n, and these can be controlled in terms ofγ n . We emph...