Methods for Estimating Eigenvalues
A comparison of quantum approaches to extracting Hamiltonian eigenvalues, from variational bounds to phase estimation and spectral filtering.
Why eigenvalues
Ground- and excited-state energies determine chemistry, materials properties, and reaction rates. Estimating eigenvalues of a Hamiltonian is therefore a headline goal of quantum simulation. Several families of methods exist, differing in precision scaling, hardware demands, and whether they give guarantees.
Variational methods
- VQE returns an upper bound on the ground-state energy via the variational principle.
- Precision scales as one over the square root of shots (standard quantum limit).
- Shallow circuits, near-term friendly, no rigorous accuracy guarantee.
- Excited states reachable via constraints or subspace expansion.
Phase-estimation methods
- QPE and its iterative or Bayesian variants read eigenphases of e^(-iHt).
- Heisenberg-limited precision: error scales as one over total evolution time.
- Deep coherent circuits; fault-tolerant hardware required.
- Return a true eigenvalue sampled by the input state's overlaps.
Spectral-filter and signal methods
Newer approaches estimate eigenvalues from measured time-series of
Choosing a method
The choice depends on hardware maturity and the precision required. Near-term devices favor VQE and short-time signal methods that tolerate noise. Fault-tolerant machines favor qubitization-based phase estimation, which gives the best asymptotic scaling and rigorous eigenvalue outputs. Across all methods, the recurring prerequisite is good overlap between the prepared state and the target eigenstate, making state preparation an unavoidable companion problem.