⚛️ quantum physics

Constrained Quantum Imaginary Time in Krylov Space

This paper introduces Constrained Quantum Imaginary Time Evolution (CQITE), a method that leverages a Krylov-space-derived hierarchy of nested anti-commutators to generate compact, systematic unitary approximations for preparing ground states under prescribed expectation-value constraints, demonstrating high accuracy and low constraint deviation in numerical tests.

Original author: Daniele Cuomo, Luis Cort Barrada, Michał Studziński, Robert van Leeuwen

Published 2026-11-06

📖 4 min read🧠 Deep dive

CC BY 4.0

Original author: Daniele Cuomo, Luis Cort Barrada, Michał Studziński, Robert van Leeuwen

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the author. For technical accuracy, refer to the original paper. Read full disclaimer

Finding the lowest energy state of a complex quantum system is a fundamental challenge in physics and chemistry. It is the key to understanding how molecules bond, how new materials might conduct electricity, or how to solve difficult optimization problems. In the quantum world, a common strategy to find this lowest energy state, or "ground state," involves a process called imaginary time evolution. Imagine a system that naturally wants to settle down; if you could run time backward in a specific mathematical way, the system would rapidly lose its excited, high-energy components and collapse into its most stable configuration. However, this process is not something that can be run directly on a quantum computer. Real quantum computers operate with rules that require all changes to be reversible and preserve information, whereas this imaginary time process is inherently one-way and irreversible. To bridge this gap, scientists have developed methods to approximate this settling process using a series of reversible steps, but doing so while keeping the system under strict control has remained a difficult hurdle.

Researchers Daniele Cuomo, Luis Cort Barrada, Michał Studziński, and Robert van Leeuwen have introduced a new method called constrained quantum imaginary time evolution to solve this problem. Their work addresses a specific scenario where the ground state is not just the lowest energy state, but one that must also satisfy certain physical rules, such as having a specific number of particles or a particular magnetic alignment. In many real-world applications, simply finding the lowest energy is not enough; the solution must also fit within a set of constraints, much like finding the shortest path between two cities that must also pass through a specific landmark. The team's approach allows a quantum computer to evolve a system toward its ground state while strictly adhering to these extra conditions, ensuring the final result is both energetically optimal and physically valid.

The core difficulty in simulating these systems on a quantum computer is that the mathematical description of the system grows explosively large as more particles are added. A full description of the system's behavior would require a list of instructions so vast that it would be impossible to store or process. The researchers solved this by discovering a way to generate a compact, manageable list of instructions that captures the essential behavior of the system without needing the full, overwhelming list. They achieved this by looking at how the system's energy operator, which dictates how the system changes, interacts with the current state of the system. By repeatedly applying this interaction, they built a structured hierarchy of directions in which the system naturally wants to move. They found that they could stop this process after a few steps and still capture the most important movements needed to reach the ground state. This creates a "pool" of possible moves that is small enough to be handled by current technology but large enough to be highly accurate.

In their simulations, the team tested this method on several different models, including systems of interacting spins and models of electrons in a triangular arrangement. For the spin models, they compared their new method against the exact mathematical solution for the constrained problem. The results showed that the new method tracked the exact solution almost perfectly, keeping the energy low while maintaining the required constraints with very high precision. In one test involving a six-spin model, the method kept the constraints satisfied with an error as small as one part in a thousand. In a more complex test involving fermions, which are particles like electrons, the method produced energy values that were extremely close to the known theoretical limits. The researchers observed that the method successfully navigated the complex landscape of possible states, finding the correct minimum even when the constraints forced the system into a difficult configuration.

A key feature of this new approach is that the list of instructions, or the operator pool, does not need to be rebuilt every time the system changes slightly. Instead, the pool is constructed once based on the system's size and the complexity of the interactions, and it remains valid throughout the entire evolution. This makes the method efficient and scalable. The researchers demonstrated that for a fixed level of accuracy, the size of this list grows at a manageable rate as the system gets larger, rather than exploding into impossibility. This provides a reliable path for using quantum computers to solve constrained problems that were previously out of reach. The work suggests that by using this structured hierarchy of directions, scientists can prepare quantum states that are not only low in energy but also strictly obey the physical laws and constraints of the real world, opening the door to more accurate simulations of materials and chemical reactions.

Technical Summary: Constrained Quantum Imaginary Time in Krylov Space

Problem Statement
Preparing ground states of quantum many-body systems is a fundamental challenge in quantum simulation, chemistry, and optimization. Imaginary time evolution (ITE) offers a theoretical route to ground-state preparation by exponentially suppressing excited-state components. However, its direct application faces two primary obstacles: the exponential growth of the Hilbert space makes classical simulation intractable for large systems, and the intrinsically non-unitary nature of ITE prevents direct implementation on gate-based quantum computers.

While Quantum Imaginary Time Evolution (QITE) addresses the unitarity issue by approximating ITE steps with locally optimal unitary transformations, many practical applications require the ground state to satisfy additional physical constraints. These include prescribed expectation values of conserved observables, symmetry conditions, or feasibility requirements in constrained optimization. Extending QITE to these "constrained-search" dynamics (CQITE) is non-trivial because the resulting dynamics remain non-unitary, and constructing a compact unitary representation that maintains accuracy while respecting constraints is a significant hurdle. Existing adaptive approaches often lack a direct characterization of the operator directions generated by the evolution.

Methodology
The paper introduces the Constrained Quantum Imaginary Time Evolution (CQITE) algorithm, which extends standard QITE to handle constrained-search dynamics. The methodology is built upon three core components:

Constrained-Search Formalism: The algorithm solves a constrained optimization problem by introducing time-dependent Lagrange multipliers ( $v(\tau)$ ) that enforce constraints $C$ throughout the evolution. The dynamics are governed by a modified generator $G(\tau) = H + v(\tau)^T C$ , where the multipliers are determined by a linear system involving the symmetrized covariance of the constraints and the Hamiltonian.
Unitary Approximation: To implement this on quantum hardware, the non-unitary step is approximated by a unitary operator $e^{-i\delta A}$ , where $A$ is a Hermitian operator. The coefficients of $A$ are determined variationally by minimizing the distance between the exact imaginary-time tangent vector and the variational ansatz.
Static Polynomial Krylov Operator Pool: A central contribution is the derivation of a systematic, static operator pool to represent the generator $A$ $A$ . Instead of heuristic selection or adaptive growth, the authors derive an exact characterization of the imaginary-time evolution using a Krylov space generated by repeated anti-commutator actions of the Hamiltonian on the initial state.
- The exact generator directions are shown to lie within the Krylov subspace spanned by $\{ \text{ac}_H^j A(0) \}_j$ , where $\text{ac}_H$ is the anti-commutator superoperator.
- By truncating this hierarchy at a fixed order $c$ , the authors construct a static pool of Pauli operators. This pool is derived directly from powers of the Hamiltonian ( $H^j$ ), avoiding the need for an explicit Pauli expansion of the initial density matrix.
- The size of this pool scales polynomially with the system size ( $O(n^c)$ ) for a fixed truncation order, providing a principled route to compact unitary approximations.

Key Contributions

Algorithmic Extension: The paper formalizes CQITE, extending QITE to constrained-search dynamics, enabling ground-state preparation with prescribed expectation values.
Exact Krylov Characterization: It provides an exact characterization of the operator directions generated by imaginary-time evolution via a hierarchy of nested anti-commutators. This contrasts with conventional quantum Krylov methods that rely on Hamiltonian powers or real-time evolution.
Static Operator Pool: The derivation yields a static operator pool that is independent of the instantaneous variational coefficients and the imaginary time parameter $\tau$ . This pool is constructed solely from the Hamiltonian and the truncation order, offering a systematic sequence of approximations with increasing expressive power.
Theoretical Bounds: The authors derive uniform error bounds for the truncation of the Krylov hierarchy, showing that the residual error decays factorially with the truncation order $c$ in the fast-convergence regime.

Numerical Results
The paper validates the CQITE algorithm through numerical simulations on spin and fermionic models:

Three-Spin Ising Model: The algorithm successfully minimizes energy under transverse magnetization constraints. The CQITE trajectory closely matches the exact constrained imaginary-time evolution (CITE), and the constraints are accurately preserved throughout the optimization.
Six-Spin Triangular-Ladder Model: In a larger system with local magnetization constraints, CQITE again demonstrates near-indistinguishable energy evolution compared to exact CITE. Constraint violations remain small (order $\sim 10^{-3}$ ), though minor residual deviations are observed.
Two-Particle Fermionic Triangle: Applied to a fermionic system (Penz–van Leeuwen model), the algorithm produces energies close to the analytic constrained-search reference ( $F(\rho)$ ). The maximum absolute energy difference across the density domain is approximately $1.4 \times 10^{-4}$ . The results confirm that the algorithm navigates the constrained energy landscape effectively, even in regions where the constraint potential diverges.

Significance and Claims
The paper claims that CQITE provides a practical algorithm for constrained ground-state preparation on gate-based quantum devices. Its significance lies in two complementary advances:

It extends quantum imaginary-time evolution to constrained-search dynamics, addressing a gap in existing QITE implementations.
It establishes an exact Krylov characterization of the operator directions, offering a systematic, polynomially-scaling alternative to variational operator pools.

The authors note that while the method shows close agreement with exact dynamics in numerical tests, the approach relies on a static pool constructed from the Hamiltonian. They modestly suggest that future work could investigate sharper truncation bounds, apply the Krylov construction to unconstrained Hamiltonians, and benchmark the algorithm on actual quantum hardware to assess the impact of noise and finite sampling. The work does not claim to solve all challenges of quantum simulation but offers a structured framework for constrained state preparation.

Technical Summary: Constrained Quantum Imaginary Time in Krylov Space

More like this