Quantum Simulations Become Faster with New Hybrid Algorithm Design

A new hybrid classical-quantum algorithm improves the simulation of Hamiltonians, offering a key addition to the set of tools for modelling physical systems. Nhat A. Nghiem and Tzu-Chieh Wei at State University of New York detail a procedure involving classical diagonalisation of Hamiltonian components, followed by quantum processing to achieve block-encoding and ultimately simulate time evolution. The research introduces three variants of this method and demonstrates its flexibility to time-dependent coefficients when the Hamiltonian components commute. Furthermore, the work uses a recent technique for quantum state preparation, potentially broadening its impact beyond Hamiltonian simulation itself.

Reduced quantum circuit complexity enables efficient Hamiltonian simulation

A threefold improvement in simulating Hamiltonians has been achieved, reducing the quantum circuit complexity to O(K²d|R|M log(d)t log 1/δ) using a novel approach, in contrast to existing methods reliant on quantum oracle access. This reduction in complexity is significant because it directly impacts the resources, specifically the number of quantum gates, required to perform the simulation. Traditional methods often require a quantum oracle to access information about the Hamiltonian, which can be computationally expensive. The ability to efficiently represent and evolve these systems is crucial for modelling materials and discovering new drugs, unlocking the simulation of larger, more complex systems previously intractable due to exponential scaling of computational resources. For instance, simulating the behaviour of molecules with more than a few atoms quickly becomes impossible for classical computers, but quantum simulation offers a potential pathway to overcome this limitation. The new hybrid classical-quantum algorithm classically diagonalises Hamiltonian components, utilising known matrix entries to efficiently block-encode the Hamiltonian and obtain the time evolution operator. Block-encoding is a technique that allows a Hamiltonian to be represented as a unitary operator, which can then be implemented on a quantum computer.

Success hinges on the feasibility of obtaining complete classical information, a task that may prove insurmountable for many real-world scenarios, yet this work refines techniques like block encoding and randomised truncation, broadening the toolkit available to quantum scientists. The algorithm’s efficiency is predicated on the assumption that the entries of all Hamiltonian components { H_i1, H_i2 , ⋯ , H_iM} (for all i) are classically known. While this is a strong requirement, it is not unrealistic for certain classes of problems, such as those arising in quantum chemistry where the atomic interactions are well-defined. Randomised truncation, a technique used to reduce the size of the quantum state space, further enhances the algorithm’s practicality. This new hybrid approach, blending classical and quantum techniques, expands the options available to researchers seeking efficient simulations. The algorithm classically diagonalises components of the Hamiltonian, breaking down the complex system into simpler, solvable parts with conventional computing power. This pre-processing allows for efficient block-encoding of the Hamiltonian and ultimately obtaining the time evolution operator. The Hamiltonian is expressed as the sum of K terms: H= ∑_i=1^K H_i = ∑_i=1^K H_i1 ⊗ H_i2 ⊗ ⋯ ⊗ H_iM. The tensor product notation (⊗) indicates that the Hamiltonian consists of multiple interacting subsystems. It offers a complementary approach to existing methods, particularly for systems with interactions limited to specific locations. By focusing on systems where the Hamiltonian can be decomposed into a sum of simpler terms, the algorithm avoids the need for complex quantum circuits that would otherwise be required to simulate the full Hamiltonian directly.

Hamiltonian knowledge limitations and the potential for hybrid quantum-classical simulations

Quantum simulation is steadily progressing towards practical application, offering the tantalising prospect of modelling complex systems beyond the reach of classical computers. The promise of quantum simulation lies in its ability to exploit the principles of quantum mechanics, superposition and entanglement, to solve problems that are intractable for classical computers. While complete classical information about a system’s Hamiltonian may be impossible to obtain in practice, this does not invalidate the current work; instead, it highlights the need for further exploration of alternative strategies. In many realistic scenarios, the Hamiltonian is only known approximately, or is derived from experimental data. Therefore, developing methods that can handle incomplete or noisy Hamiltonian information is a crucial area of research. This research presents a hybrid computational method, merging classical and quantum techniques for simulating Hamiltonians, which are mathematical descriptions of a system’s energy. The Hamiltonian operator describes the total energy of the system and governs its time evolution.

Classically determining the properties of individual Hamiltonian components first created a more efficient pathway for quantum processing. The classical diagonalisation step allows for a significant reduction in the complexity of the subsequent quantum computation. Diagonalising a matrix involves finding its eigenvalues and eigenvectors, which can be used to simplify the Hamiltonian and make it easier to simulate. Building upon recent advances in quantum state preparation, specifically randomised truncation, the work demonstrates its broader applicability beyond its initial purpose. Randomised truncation is a technique used to reduce the dimensionality of the quantum state space, which is essential for simulating large systems. By selectively discarding less important quantum states, the algorithm can significantly reduce the computational resources required. The resulting algorithm expands the options for modelling physical systems, offering a valuable addition to the quantum computational toolkit. This toolkit is constantly evolving, and new algorithms are needed to address the challenges of simulating increasingly complex systems. Further investigation into the limitations of classical pre-processing and the development of methods to mitigate these constraints will be key to realising the full potential of this hybrid approach. For example, exploring techniques for approximating the Hamiltonian components or developing error-correction schemes to mitigate the effects of noise could further enhance the algorithm’s performance and applicability. The three variants of the procedure presented offer flexibility in adapting to different system characteristics and computational resources, increasing the algorithm’s versatility.

The researchers developed a new hybrid algorithm combining classical and quantum computation to simulate Hamiltonians, which describe the total energy of a physical system. This method improves efficiency by first classically determining the properties of the Hamiltonian’s components before processing them with quantum techniques. The algorithm builds upon recent advances in quantum state preparation, demonstrating its wider use in quantum simulation. The authors suggest further work will focus on understanding the limits of classical pre-processing to optimise the algorithm’s performance.