Jiaqing Jiang of the California Institute of Technology and Sandy Irani of the University of California, Irvine, have designed a new quantum Gibbs sampler that simplifies existing methods. Their work addresses a long-standing challenge in quantum computing: creating algorithms that accurately prepare Gibbs states, essential for simulating quantum systems and modeling thermal equilibrium.

The researchers revisited an earlier approach to mimicking the classical Metropolis algorithm, but significantly streamlined it by removing the Marriott-Watrous rewinding technique. This new method also circumvents the need for a potentially non-existent property of Quantum Phase Estimation, resulting in a conceptually simple and provably correct sampler with the Gibbs state as its approximate fixed point.

Quantum Metropolis Sampling via Weak Measurement

Circumventing a reliance on the Marriott-Watrous rewinding technique and significantly simplifying the algorithm previously used to mimic the classical Metropolis method for quantum Hamiltonians, the design of a new quantum Gibbs sampler achieves this advancement. Central to this advancement is a method that avoids the need for a shift-invariant version of Boosted Quantum Phase Estimation (QPE).

Earlier algorithms depended on this potentially non-existent property for analysis, creating a significant hurdle; the current work uses a Boosted QPE that takes the median of multiple runs, but does not require the problematic shift-invariant characteristic. This parameter set allows for a flexible approach to simulating quantum systems at varying levels of precision.

Unlike currently provably correct quantum Gibbs samplers which rely on simulating Davies generators, this new method aims for provable correctness while approximating the Gibbs state as its unique fixed point. The researchers define a Completely Positive and Trace Preserving (CPTP) linear map, essential for ensuring the algorithm’s validity, and use the trace norm to quantify how much the map can scale. The work further specifies that the algorithm’s performance is influenced by parameters r and g, which control the precision of the simulation.

Boosted Quantum Phase Estimation for Gibbs State Preparation

The removal of the Marriott-Watrous rewinding technique represents a significant simplification in the creation of quantum Gibbs samplers. This streamlining allows for a more direct implementation of the algorithm, potentially reducing computational overhead and improving efficiency. The set of energy estimations, S, relies on integer multiples of w: = $κ_{H} \cdot$ $2^{- r}$ , and the algorithm satisfies an approximate version of the fixed point property.

Classical Metropolis Algorithm Challenges in Quantum Systems

The manuscript details that the key difference lies in directly designing the Gibbs sampler from QPE, rather than through the more complex process of simulating a Davies generator. As stated in the manuscript, “the key difference is that our Gibbs sampler is directly designed from QPE instead of a Davies generator.” This approach, detailed in the paper, offers a conceptually simpler alternative for generating accurate quantum simulations.

Gibbs Sampling with Davies Generators and Lindbladians

Recent advances in quantum Gibbs sampling offer a conceptually simpler path toward accurate simulations by directly using quantum phase estimation (QPE), rather than relying on the simulation of Davies generators. Removing this step simplifies the algorithm considerably, reducing computational overhead without sacrificing provable correctness. This alternative method, they state, exactly fixes the Gibbs states and can be efficiently implemented on a quantum computer. The current work, however, presents a distinct strategy for achieving provable correctness in quantum Gibbs sampling, offering a new direction for exploration in the field.

Weak Measurement Simplifies Quantum Move Acceptance

The complexity of mimicking the classical Metropolis algorithm in a quantum context has been reduced through the incorporation of weak measurement, offering a conceptually simpler path to provably correct quantum Gibbs sampling. A key challenge in adapting the classical Metropolis algorithm to quantum systems lies in the difficulty of reverting a quantum measurement when a proposed move is rejected. The algorithm satisfies an approximate version of the fixed point property.

Approximate Fixed Point Property of the New Sampler

The newly detailed quantum Gibbs sampler achieves an approximate fixed point property, an essential characteristic for ensuring the algorithm’s correctness in preparing equilibrium states for quantum systems. The researchers detail that this streamlining significantly reduces computational overhead without sacrificing accuracy. The algorithm’s performance is quantified by Lemma 22, which establishes an upper bound on the difference between the transformed density matrix and the original.

Specifically, the analysis demonstrates that the trace norm distance, a measure of how dissimilar two quantum states are, is bounded by +, where ‘g’ and ‘n’ represent system parameters and ‘r’ is a parameter related to the Hamiltonian. This bound suggests manageable scaling with system size, a critical factor for practical implementation.

Eliminating Shift-Invariance from Boosted QPE Analysis

Removing a requirement for shift-invariance within Boosted Quantum Phase Estimation (QPE) represents a key simplification in the pursuit of provably correct quantum Gibbs samplers. This new approach circumvents that need, offering a potentially more viable path toward scalable quantum simulation. The work builds upon an earlier method employing QPE and the Marriott-Watrous rewinding technique, but diverges by eliminating the latter.

Technically, the analysis introduces a new method for bounding errors stemming from imperfect energy estimation within QPE. This is achieved by grouping error terms into finite classes.

As they state, “Our algorithm uses a Boosted QPE which takes the median value of multiple independent runs of the standard QPE. We do not require a version of QPE that satisfies shift-invariance.” The number “two” itself is not critical to the analysis, as long as a constant value is used. The team’s approach focuses on conceptual simplicity and a reduction in computational demands, potentially broadening the applicability of quantum Gibbs sampling techniques.

Alternate Accept Strategy Reduces Rewinding Complexity

Removing the Marriott-Watrous rewinding technique, a component of prior quantum Gibbs samplers, significantly reduces the complexity of this new method. The previous approach required a polynomial number of unitaries and measurements to successfully rewind the system, whereas this work achieves the same result with a single qubit measurement and unitary operation. This simplification stems from the observation that, after an unsuccessful rewinding attempt, the quantum state closely resembles the accepted state, allowing for immediate iteration completion.

The team’s approach introduces an “Alternate Accept” case, a critical feature enabling the elimination of rewinding. Without this, even with the implementation of weak measurement, the computationally intensive Marriott-Watrous rewinding would still be necessary. Weak measurement itself introduces a trade-off, increasing the number of iterations by a polynomial factor, but this is offset by the overall reduction in algorithmic demands.

The analysis details that this method uses a Boosted Quantum Phase Estimation (QPE), but it does not rely on the shift-invariant property. Defining the operators for the accept case is central to the method. The operator $W^{10}$ is expressed as a summation over energies in the system, incorporating a factor representing the probability of transitioning between energy levels.

The acceptance operator, $M_{C}^{a} (ρ)$ , is then defined as a partial trace over specific subsystems of the combined system and unitary transformation. As the researchers state, “we use only one single-qubit measurement and one unitary for rewinding.” This streamlined approach represents a departure from previous methods and offers a potentially more scalable path toward accurate quantum simulations.

👉 More information
🗞 Quantum Metropolis Sampling via Weak Measurement
✍️ Jiaqing Jiang and Sandy Irani
🧠 DOI: http://link.aps.org/doi/10.1103/xxq6-d82l

More like this