arXiv:2609.37527
Digital quantum state preparation by nucleation
Jean Paul Sadia, Morten Hjorth-Jensen, Dean Lee, Ryan LaRose
Abstract
Quantum state preparation is crucial for quantum algorithms, and improved state preparation methods can reduce resource requirements by orders of magnitude. In this work, we introduce a state preparation method for lattice Hamiltonians on digital quantum computers inspired by nucleation. In our method, the exact ground state of a small lattice Hamiltonian is prepared in a quantum circuit, then a series of growth stages are performed via Trotterized adiabatic evolution to increase the size of the lattice to the desired model. Using the adiabatic theorem and Trotter error bounds, we provide sufficient conditions on the total evolution time and number of Trotter steps required for nucleation to prepare an initial state with desired fidelity. In addition, we introduce optimized nucleation which considers variational optimization on top of the nucleation structure. For a two-dimensional Ising model, we show that optimized nucleation can outperform a current leading method for state preparation on digital quantum computers.
AI Summary
The paper studies ground-state preparation for lattice Hamiltonians on digital quantum computers using incremental lattice growth. Its “nucleation” procedure exactly prepares a small seed, adds sites in product states, and turns on their couplings through Trotterized adiabatic evolution. A variational version retains this circuit structure, initializes parameters from the adiabatic schedule, and optimizes energy layer by layer. The guarantees combine an existing adiabatic theorem with freezing and first-order product-formula error bounds. For one growth stage, adiabatic infidelity at most $\epsilon$ is ensured by $T\ge\epsilon^{-1/2}\int_0^1(\|\ddot H\|/\Delta^2+7\|\dot H\|^2/\Delta^3)\,du$, assuming a smooth path with a unique, gapped ground state. Digitization operator error at most $\eta$ follows from $N\ge[ T\Lambda_1/4+T^2\Lambda_2/2]/\eta$, where $\Lambda_1=\max\|\dot H\|$ and $\Lambda_2=\max\|[A,B]\|$ for $H=A+B$. Classical simulations grow a one-dimensional transverse-field Ising chain from 2 to 12 sites. They reach ground-state overlap squared $0.997$, versus $0.022$ initially, using approximately $T=40$ and 400 Trotter steps per stage across five stages—far below the reported conservative estimates. For a $5\times5$ Ising lattice, optimized nucleation achieves lower energy error with fewer two-qubit gates than an MPS-to-circuit method, including its variationally optimized version. The two-dimensional benchmark is a single instance explicitly chosen to favor nucleation; energies use finite-bond-dimension MPS simulations and a DMRG reference. Total-cost asymptotic scaling, broader benchmarking, growth-strategy selection, hardware tests, and fault-tolerant resource estimates remain open.
Error found: The paper presents single-growth-stage error bounds as preparation guarantees without specifying how adiabatic and Trotter errors accumulate across all growth stages. The numerical resource estimates use an even-parity-sector gap although the stated theorem defines its gap using the full spectrum, without explicitly applying the theorem to the invariant sector. The reported adiabatic-time estimate is approximately twice the value obtained from its stated parameters and displayed formula, with no explanation of the extra factor.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 10 pages, 7 figures