• Algorithms
  • Also: VQE

Variational Quantum Eigensolver

A hybrid quantum-classical algorithm that uses a quantum computer to estimate the ground state energy of a molecule, the leading near-term application for quantum chemistry.

The Variational Quantum Eigensolver (VQE) is the leading algorithm proposed for near-term quantum advantage in quantum chemistry. It estimates the ground state energy of a molecular Hamiltonian by iteratively optimizing a parameterized quantum circuit on a quantum processor, guided by a classical optimizer. The goal is to find the lowest possible energy state of a molecule: a number that determines molecular geometry, reaction rates, and binding affinities.

VQE was introduced in 2014 specifically for NISQ hardware. Unlike Shor’s algorithm or fully fault-tolerant quantum chemistry methods, VQE is designed to run on noisy qubits with short coherence times.

The details

The variational principle from quantum mechanics guarantees that for any trial state ∣ψ⟩|\psi\rangle:

Etrial=⟨ψ∣H∣ψ⟩≥EgroundE_{\text{trial}} = \langle \psi | H | \psi \rangle \geq E_{\text{ground}}

The trial energy is always an upper bound on the true ground state energy. By minimizing over all possible trial states, you approach the true ground state energy from above.

VQE implements this by parameterizing the trial state as a quantum circuit (the ansatz):

∣ψ(θ)⟩=U(θ)∣0⟩|\psi(\theta)\rangle = U(\theta)|0\rangle

where θ\theta is a vector of gate rotation angles. The algorithm runs in a loop:

  1. Prepare ∣ψ(θ)⟩|\psi(\theta)\rangle on the quantum processor by running the ansatz circuit
  2. Measure the expectation value E(θ)=⟨ψ(θ)∣H∣ψ(θ)⟩E(\theta) = \langle\psi(\theta)|H|\psi(\theta)\rangle by measuring Pauli terms of HH and summing
  3. Send E(θ)E(\theta) to a classical optimizer (COBYLA, SPSA, or gradient-based methods)
  4. Receive updated parameters θ′\theta' from the optimizer
  5. Repeat until convergence

The Hamiltonian HH for a molecule is expressed as a sum of Pauli operators using the Jordan-Wigner or Bravyi-Kitaev transformation:

H=∑iciPiH = \sum_i c_i P_i

where PiP_i are tensor products of Pauli matrices. Each term is measured separately, requiring many circuit executions per optimization step.

from qiskit.circuit.library import TwoLocal
from qiskit_algorithms import VQE
from qiskit_algorithms.optimizers import COBYLA
from qiskit.primitives import Estimator

# Ansatz: hardware-efficient parameterized circuit
ansatz = TwoLocal(num_qubits=2, rotation_blocks='ry', entanglement_blocks='cx', reps=2)

optimizer = COBYLA(maxiter=200)
estimator = Estimator()

vqe = VQE(estimator, ansatz, optimizer)
# result = vqe.compute_minimum_eigenvalue(hamiltonian)

Ansatz choices matter greatly:

  • Hardware-efficient ansatz: Uses native gates on the target hardware (low depth, can run on NISQ). No guarantee of reaching the ground state if it is chemically complex.
  • UCCSD (Unitary Coupled Cluster Singles and Doubles): Chemically motivated; good approximation for many molecules, but circuit depth is often too large for NISQ hardware without approximations.

Why it matters for learners

VQE represents the best current argument for near-term quantum utility. Classical computational chemistry scales exponentially with system size for exact methods. The ground state energy of the FeMo cofactor in nitrogenase (relevant to nitrogen fixation and fertilizer production) is beyond the reach of classical exact methods, yet its exact energy would have enormous practical value for catalysis research.

VQE illustrates several important concepts:

The hybrid quantum-classical loop: Not all computation happens on the quantum processor. The classical optimizer controls the overall search. Only the quantum parts (state preparation and expectation value estimation) run on quantum hardware. This hybrid structure is the design pattern for most near-term quantum algorithms.

Noise sensitivity: Each expectation value estimate on a NISQ device includes shot noise (statistical sampling error) and hardware noise (gate errors, readout errors). Both add uncertainty to E(θ)E(\theta), making the classical optimizer’s job harder and requiring more measurements.

Barren plateaus: As the ansatz gets deeper or broader, the gradient of E(θ)E(\theta) with respect to parameters often decreases exponentially to zero. Gradient-based optimizers fail in this regime. This is one of the most serious theoretical obstacles to VQE scaling to larger molecules.

Common misconceptions

Misconception 1: VQE is guaranteed to find the ground state energy. VQE is an upper bound method: the variational principle guarantees the energy found is never lower than the true ground state. But if the ansatz cannot represent the true ground state (which happens for complex molecules with strong correlation), VQE converges to a wrong value without knowing it is wrong. The quality of the result depends entirely on the expressiveness of the ansatz.

Misconception 2: VQE has already demonstrated quantum advantage. As of 2026, no VQE calculation has convincingly outperformed state-of-the-art classical methods on a molecule of practical interest. Classical methods like DMRG, FCI, and even empirically tuned classical neural network methods are competitive. Quantum advantage in chemistry remains an open challenge, not an accomplished fact.

Misconception 3: The quantum part does all the work. The classical optimizer runs many iterations and takes significant classical compute time. The quantum processor evaluates the energy at each set of parameters, but the optimization search is entirely classical. For small molecules, the classical simulation of the entire VQE loop (including the quantum part) can sometimes outperform running it on noisy real hardware.

See also