Variational determination of arbitrarily many eigenpairs in one quantum circuit

Guanglei Xu, Yibin Guo, Xuan Li, Zong-Sheng Zhou, Haijun Liao, T. XiangarXiv 2022 · DOI · arXiv

This paper, published by researchers at the Chinese Academy of Sciences and other institutions, introduces a more efficient way to find multiple energy levels of a quantum system using today’s "noisy" quantum computers. By using a technique called "purification" with extra qubits, they can find many excited states simultaneously in a single quantum circuit, avoiding the complex, error-prone recursive steps required by previous methods.

TL;DR

The authors present a novel variational quantum algorithm that determines multiple low-energy eigenpairs (eigenvalues and eigenvectors) simultaneously within a single quantum circuit. By leveraging ancillary qubits to maintain state orthogonality, the method significantly reduces circuit depth and readout errors compared to traditional recursive excited-state solvers.

What Problem Does This Paper Address?

In quantum mechanics, finding the ground state (lowest energy) of a Hamiltonian is crucial, but finding "excited states" (higher energy levels) is just as important for understanding chemical reactions and material properties.

The standard tool for this is the Variational Quantum Eigensolver (VQE). While VQE is great for ground states, extending it to excited states usually requires a recursive approach: you find the ground state, then you find the first excited state while mathematically forcing it to be "orthogonal" (perpendicular) to the first. This process is computationally expensive, requires increasingly deep (and thus noisy) circuits, and accumulates errors at every step. The authors sought a way to find these states all at once without the recursive overhead.

Methods

The core innovation is the use of ancillary qubits and a purification technique.

  1. Purification & Entanglement: The researchers introduce $N_a$ ancillary qubits. They prepare an initial state where the physical qubits are "maximally entangled" with these ancillas.
  2. Orthogonality by Design: A variational unitary transformation $U(\theta)$ is applied only to the physical qubits. Because the ancillas remain untouched, the orthogonality of the basis states in the ancilla space "protects" the orthogonality of the trial states in the physical space. This ensures that $K$ different trial states remain orthogonal throughout the entire optimization process.
  3. The Loss Function: Instead of minimizing a single energy, the algorithm minimizes the sum of the energy expectation values of the first $K$ states, guided by the generalized Rayleigh-Ritz variational principle.
  4. Ancilla-Assisted Post-Processing: To find the actual eigenstates (not just the subspace), they need off-diagonal matrix elements. Traditionally, this is very difficult. Here, they show that by rotating only the ancillary qubits at the end of the circuit, they can extract these elements easily. This allows them to diagonalize a small $K \times K$ matrix classically to get the final results.

Datasets

The algorithm was validated using numerical simulations of the one-dimensional transverse Ising model (TIM), a standard benchmark in condensed matter physics.

  • System Size: 8 physical qubits ($N_p = 8$), resulting in a Hilbert space dimension of 256.
  • Comparison: Results were compared against Exact Diagonalization (ED) to measure accuracy.
  • Parameters: The model was tested at the critical point ($h_x = 0.5, J = 1$), where the system's behavior is most complex.

Key Insights and Ideas

  • Simultaneity is Key: By finding $K$ states at once, the algorithm avoids the "error "creep" inherent in recursive methods like Variational Quantum Deflation (VQD).
  • The "Ancilla Trick": Manipulating ancillas to perform basis transformations on the physical states is a powerful shortcut. It allows for complex post-processing (like calculating thermal averages or off-diagonal elements) without adding any depth to the physical quantum circuit.
  • Error Cancellation: Because all eigenvalues are derived from the same optimized subspace in the same circuit, systematic errors tend to affect all levels similarly. This means the energy gaps (the difference between levels) are often more accurate than the absolute energy values themselves.

Key Results

  • Rapid Convergence: For the 8-qubit Ising model, the four lowest eigenvalues converged to high precision within a few hundred optimization iterations.
  • Shallow Circuits: The absolute errors ($\delta E_i$) dropped exponentially with circuit depth, reaching a floor of roughly $10^{-13}$ (near machine precision in simulation) with only 6 to 8 layers of quantum gates.
  • Subspace Rank: The researchers found that to successfully find $K$ states, the initial trial states must have sufficient "overlap" with the true eigenstates. In one test (Fig 3), using 3 ancillas (8 states) only yielded 6 of the lowest 8 states because the chosen ansatz lacked the "rank" to capture the 6th and 7th excited states, skipping instead to the 8th and 10th.

Limitations and Future Work

  • Qubit Overhead: The method requires $\log_2 K$ additional ancillary qubits. While manageable for a few states, finding hundreds of states would require significant qubit overhead.
  • Ansatz Selection: As noted in their results, if the variational ansatz (the structure of the circuit) doesn't overlap well with certain excited states, the algorithm might "skip" them. Choosing a sufficiently expressive ansatz remains a challenge.
  • Hardware Noise: While the theory suggests this method is resilient to noise because of its shallow depth, the paper relies on ideal simulations. Testing this on actual NISQ hardware (like superconducting or ion-trap processors) is the necessary next step to prove its "noise-resilient" claims.

Summary updated

Abstract

The state-of-the-art quantum computing hardware has entered the noisy intermediate-scale quantum (NISQ) era. Having been constrained by the limited number of qubits and shallow circuit depth, NISQ devices have nevertheless demonstrated the potential of applications on various subjects. One example is the variational quantum eigensolver (VQE) that was first introduced for computing ground states. Although VQE has now been extended to the study of excited states, the algorithms previously proposed involve a recursive optimization scheme which requires many extra operations with significantly deeper quantum circuits to ensure the orthogonality of different trial states. Here we propose a new algorithm to determine many low energy eigenstates simultaneously. By introducing ancillary qubits to purify the trial states so that they keep orthogonal to each other throughout the whole optimization process, our algorithm allows these states to be efficiently computed in one quantum circuit. Our algorithm reduces significantly the complexity of circuits and the readout errors, and enables flexible post-processing on the eigen-subspace from which the eigenpairs can be accurately determined. We demonstrate this algorithm by applying it to the transverse Ising model. By comparing the results obtained using this variational algorithm with the exact ones, we find that the eigenvalues of the Hamiltonian converge quickly with the increase of the circuit depth. The accuracies of the converged eigenvalues are of the same order, which implies that the difference between any two eigenvalues can be more accurately determined than the eigenvalues themselves.
BibTeX
@misc{xu2022variational,
  author = {Guanglei Xu and Yibin Guo and Xuan Li and Zong-Sheng Zhou and Haijun Liao and T. Xiang},
  title = {{Variational determination of arbitrarily many eigenpairs in one quantum circuit}},
  year = {2022},
  eprint = {2206.11036},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2206.11036},
}
Figure