A new benchmark tests quantum algorithms with realistic noise

Researchers have established a new benchmark for evaluating quantum algorithms by testing their ability to prepare ground states amidst realistic noise. The work, centered on quadratic fermionic Hamiltonians exhibiting two distinct phases separated by a quantum phase transition, allows for both analytical and numerical comparisons of algorithms, a rare combination for assessing performance. Results show multi-frequency cooling algorithms, as proposed in Phys.

7 (2025) 033162, can outperform adiabatic evolution in the topological phase, where gap-closing typically limits adiabatic protocols. The team further demonstrated enhanced robustness to parameter imperfections in the cooling protocol, suggesting a practical advantage for noisy quantum state preparation.

Quadratic Fermionic Hamiltonians & Depolarizing Noise Benchmark

An exactly solvable family of quadratic fermionic Hamiltonians is now a benchmark for evaluating quantum algorithms under realistic conditions, allowing for both analytical and numerical validation of performance. This model facilitates a direct comparison of ground-state preparation methods when subjected to depolarizing noise, a common source of error in quantum systems. The Hamiltonian’s structure, with translationally invariant coupling, enables decomposition into independent momentum blocks, simplifying calculations and large-scale numerical simulations.

This factorization has notable consequences for the three algorithms tested: multi-frequency cooling, adiabatic evolution and the Quantum Approximate Optimization Algorithm (QAOA). Researchers found that cooling protocols exhibited enhanced robustness to parameter imperfections, a key advantage for practical implementation. This resilience stems from the algorithm’s convergence to a fixed point; the steady state acts as an attractor, minimizing the impact of small variations in coupling parameters.

In contrast, QAOA, which produces a single time-evolved state, proved more sensitive to parameter values, lacking the stabilizing effect of a fixed-point structure. The unified framework developed by the team provides a standardized approach for assessing ground-state preparation algorithms in noisy environments.

Adiabatic Algorithm Performance Across Quantum Phases

The performance of quantum algorithms shifts markedly across distinct phases of matter, with cooling protocols demonstrating an advantage over traditionally favored adiabatic evolution in topological phases. 7 (2025) 033162, becomes competitive or superior when the system enters the topological phase. This reversal occurs because gap-closing, a common limitation in adiabatic protocols, does not impede the cooling algorithm’s effectiveness. This study extends beyond simply identifying superior algorithms, also quantifying the impact of noise on each approach.

Researchers combined analytical derivations, determining achievable energy as a function of noise rate, with numerical simulations. The analytical approach developed here, in conjunction with numerical validation, establishes an extendable approach to benchmarking ground-state preparation algorithms. The team’s work demonstrates that QAOA performs competitively with cooling in the trivial phase, but is typically outperformed in the topological regime. The chosen Hamiltonian, with its clear phase transition, provides a standardized benchmark for evaluating ground-state preparation methods under realistic conditions.

Multi-Frequency Cooling Algorithm in Topological Regimes

Analytical derivations establish that achievable energy scales with noise rates differently across phases within the tested quadratic fermionic Hamiltonians. Specifically, in the trivial, gapped phase, adiabatic algorithms achieve a lower relative energy than multi-frequency cooling algorithms due to a small non-adiabatic prefactor and the power law governing this regime. However, this advantage reverses in the topological phase, where both algorithms scale similarly, but cooling demonstrates a more consistent performance.

This shift occurs because the adiabatic path crosses a quantum critical point, a known limitation for this protocol, while cooling’s performance remains largely unaffected by this transition. Simulations reveal that cooling yields lower energy throughout most of the topological phase, even with realistic imperfections in the initial state.

Quantum Approximate Optimization Algorithm (QAOA) Limitations

Analytical derivations and numerical validation combined to establish a repeatable method for evaluating ground-state preparation algorithms, an important step for benchmarking quantum performance. This approach moves beyond idealized conditions to assess algorithms with realistic imperfections in parameter settings. The researchers found that while all three algorithms ultimately suffer from extended runtimes, the mechanisms governing optimal time balance between coherent evolution and noise vary significantly. This algorithm-specific behavior forms the central analysis of a related section of the work.

The study also clarifies that increasing total runtime degrades performance across all three algorithms tested. “Preparing ground states of many-body Hamiltonians is a central task in quantum simulation,” the authors note, highlighting the broad relevance of their findings. This work builds on previous research published in *Phys. The research team made all data and code publicly available, permitting any use, distribution and reproduction with appropriate attribution.

Cooling Protocol Robustness to Parameter Imperfections

This work specifically details how multi-frequency cooling protocols demonstrate resilience to parameter imperfections, a key consideration for real-world quantum devices susceptible to noise. Comparisons focused on quadratic fermionic Hamiltonians, a model chosen for its ability to support both analytical and numerical benchmarking, allowing researchers to thoroughly map algorithm performance across the Hamiltonian’s phase diagram. This robustness is not merely a matter of achieving lower energy states. It signifies a more stable and predictable algorithm in the face of realistic operating conditions. The cooling protocol’s enhanced stability stems from its inherent characteristics.

Analytical Derivation of Energy Scaling with Noise

Analytical derivations now detail how achievable relative energy scales with noise rate, a critical step for comparing quantum algorithms beyond idealized conditions. The work establishes a method for predicting performance degradation as noise increases, focusing on the relationship between noise strength and energy scaling, specifically, how quickly relative energy rises with increasing noise. Researchers determined that a superior scaling law indicates a slower increase in relative energy with a given parameter, providing a quantifiable metric for algorithm comparison.

In the trivial, or gapped, phase of the quadratic fermionic Hamiltonian model, adiabatic algorithms achieved a scaling of, yielding lower relative energies than multi-frequency cooling within the studied parameter range. Detailed analysis, supported by exact numerical optimization and approximations, reveals the interplay between algorithm choice and the Hamiltonian’s phase in determining optimal scaling.

The analytical predictions align with observed scaling behavior and the position of minima in numerical simulations, validating the derived laws. For depolarizing noise, the noise channel’s commutation with unitary evolution simplifies the analysis, allowing researchers to relate energy with and without noise through a defined factor.

This precise modeling of noise effects is important for assessing the viability of quantum algorithms in realistic hardware environments, where imperfections are unavoidable. The team’s work provides a framework for evaluating the resilience of different approaches to noise, moving beyond simple performance benchmarks to a deeper understanding of algorithmic stability.

Unlike adiabatic and QAOA methods which rely on a specific initial state and unitary evolution, cooling protocols converge on a stationary state largely independent of the starting conditions.

Optimized QAOA vs. Cooling: Trivial & Topological Phases

Analytical and numerical results reveal a strong link between algorithmic performance and the phase of the quantum system being studied; adiabatic evolution excels in the trivial phase, while cooling proves more effective in the topological phase. This phase-dependence extends to parameter optimization, with the study demonstrating that no single algorithm consistently outperforms the others across the entire phase diagram. Further analysis focused on the robustness of each approach when subjected to parameter imperfections, a critical consideration for practical quantum computation.

Cooling demonstrated greater resilience to these perturbations, reinforcing its potential advantage in real-world scenarios where precise parameter tuning is unlikely. The team compared relative energies achieved by both cooling and QAOA using optimally tuned parameters, then assessed how each algorithm fared when those parameters were slightly altered. Cooling achieved lower energy levels deep within the topological phase, exhibiting smaller prefactors for targeted values than QAOA. “No single algorithm dominates across the full phase diagram,” the authors note, emphasizing the importance of selecting an algorithm tailored to the specific Hamiltonian phase.

Benchmarking Approach for Ground-State Preparation Algorithms

An analytical framework extends benchmarking beyond specific models and noise channels. This approach moves beyond simply assessing performance in ideal conditions, addressing a critical need for realistic evaluation of quantum algorithms. The study’s methodology uses an exactly solvable Hamiltonian family subject to depolarizing noise, enabling both analytical derivation and numerical confirmation of scaling behavior. The team’s analysis also considered QAOA as a third approach to ground-state preparation, alongside cooling and adiabatic evolution. This analytical approach, they believe, provides a template for benchmarking ground-state preparation algorithms beyond the specific model and noise channel investigated.

More like this

Quantum PhysicsQuantum simulators may become advantageous in the high-energy regime.

Quantum PhysicsPauli Z dressing unlocks universal quantum gates in limited spaces

Quantum Research NewsIron-sulfur clusters within reach of early quantum computers

Quantum Research NewsKagome antiferromagnets studied with 36-spin quantum simulation