Quantum Speed-Up Claims Challenged by New Classical Simulation Technique

A new classical simulation of ‘peaked’ quantum circuits challenges claims of quantum advantage. David Kremer and Nicolas Dupuis at IBM T.J. Watson Research Centre have demonstrated an efficient classical simulation of these circuits, previously thought to be intractable for conventional computers. Their method uses the unique mirrored structure of the circuits, reducing complex calculations to a manageable Matrix Product Operator via a technique called ‘unswapping’. This enables near-exact sampling and peak extraction in approximately one hour on a single GPU, sharply faster than the years estimated for classical simulation and even quicker than the original quantum hardware runtime.

Mirror symmetry reduction to Matrix Product Operators enables efficient simulation

Tensor network contraction, a technique for simplifying complex calculations by breaking them down into smaller, manageable pieces, underpinned this new simulation capability. Quantum circuits, at their core, describe the evolution of quantum states, and their complexity grows exponentially with the number of qubits involved. Representing and simulating these states classically requires immense computational resources. Tensor networks provide a way to represent these high-dimensional quantum states in a more compact form, allowing for efficient calculations. The process began by recognising the mirrored structure inherent in these peaked circuits, where identical operations occurred on either side of a central point. This symmetry is not a general property of all quantum circuits, but a specific design choice made in the construction of these peaked circuits. Iterative cancellation of corresponding layers from each half of the circuit then converted the entire structure into a Matrix Product Operator (MPO), a compact way to represent the quantum circuit, similar to how shorthand notation simplifies a long sentence. MPOs are particularly well-suited for representing circuits with limited entanglement, which is a key characteristic of these peaked circuits.

A variationally trained shallow structure, combined with an obfuscated permutation to increase circuit depth and hinder classical simulation, characterised the peaked quantum circuits constructed on Quantinuum’s 56-qubit H2 processor. The intention behind the obfuscation was to make the circuit’s structure less transparent to classical algorithms, thereby increasing the difficulty of simulation. However, the researchers found that the underlying mirrored structure still allowed for significant simplification. Classical simulation proved challenging due to the potential for exponential growth in computational cost with increasing circuit complexity. The number of parameters required to describe a quantum state grows exponentially with the number of qubits, making it difficult to store and manipulate these states on classical computers. However, the focus remained on exploiting the mirrored structure of the circuits to overcome these limitations. This approach highlights the importance of circuit structure in determining the feasibility of classical simulation.

Efficient tensor network contraction simulates a 56-qubit quantum circuit on a single GPU

Completion of the classical simulation of the largest peaked quantum circuit, a 56-qubit circuit with 1,917 two-qubit gates, took approximately one hour on a single GPU. This represents a sharp reduction from the original quantum hardware runtime and years previously estimated for classical methods. Prior to this work, simulating circuits of this size was considered beyond the reach of conventional computers. A new tensor network contraction method efficiently manages the circuit’s complex structure, overcoming limitations that previously restricted classical simulation to circuits of around 700 two-qubit gates. The ‘unswapping’ technique, central to this method, involves rearranging the order of operations in the circuit without changing the overall result, but in a way that simplifies the tensor network structure. Demonstrated was complete classical simulation of a 56-qubit peaked quantum circuit containing 1,917 two-qubit gates in approximately one hour using a single GPU, surpassing the original quantum hardware runtime. This sharply reduces the time previously projected for classical approaches. The use of a single GPU is particularly noteworthy, as it demonstrates the efficiency of the algorithm and its potential for wider accessibility.

The method accurately recovers the peak bitstring, the single output with the highest probability, confirming the circuit’s intended behaviour. This is a crucial validation step, as it ensures that the simulation is producing correct results. Numerical stability was ensured through the employment of a singular value cutoff, allowing near-exact sampling from the ideal circuit, and the circuit was processed almost twice as fast as its execution on Quantinuum’s H2 processor. The singular value cutoff is a technique used to truncate the tensor network, reducing the computational cost while maintaining a high degree of accuracy. Successfully simulating these circuits, however, does not resolve the fundamental question of designing quantum algorithms that demonstrably surpass classical capabilities. While this work demonstrates that these specific peaked circuits are not beyond the reach of classical simulation, it does not invalidate the potential of quantum computing as a whole.

The method is tailored to the specific construction of peaked circuits used by Gharibyan and colleagues, relying on the mirrored structure inherent within them. This raises a key tension: can this classical approach be generalised to more complex or differently obfuscated peaked circuits, or will increasingly intricate designs eventually prove resistant to such simplification. The effectiveness of this method is contingent on the presence of the mirrored symmetry. If the symmetry is broken or obscured, the simplification may no longer be possible. Further research is needed to determine the limits of this approach and to explore whether it can be extended to other types of quantum circuits. Despite demonstrating classical simulation of these specific peaked circuits, this work retains strong value as a methodological advance.

Techniques like iterative cancellation and ‘unswapping’ to reduce computational load offer a powerful new tool for quantum algorithm analysis, enabling efficient tackling of complex tensor networks. Classical simulation of recently proposed ‘peaked’ quantum circuits has been demonstrated, challenging claims of quantum advantage. Efficient classical simulation of these specifically constructed quantum circuits fundamentally alters approaches to verifying quantum computation and challenges interpretations of recent quantum advantage claims. By exploiting a mirrored structure within the circuits and employing a technique called ‘unswapping’, complex calculations were successfully converted into a manageable Matrix Product Operator, a condensed representation of the quantum circuit. This achievement demonstrates that circuits designed to concentrate results on a single outcome are not necessarily immune to classical simulation, prompting investigation into how to construct circuits that genuinely surpass classical capabilities. The development of more robust and versatile quantum algorithms remains a key goal in the field of quantum computing, and this work provides valuable insights into the challenges and opportunities that lie ahead.

Researchers demonstrated that recently proposed peaked quantum circuits, executed on a 56-qubit processor, can be efficiently simulated using classical methods. This finding challenges previous claims of quantum advantage for these circuits, which were designed to produce a sharply concentrated output on a single bitstring. The team achieved this by exploiting a mirrored structure within the circuits and converting complex calculations into a more manageable Matrix Product Operator, completing the simulation in approximately one hour on a single GPU. This work highlights the importance of rigorous classical verification when assessing the performance of quantum algorithms and informs future circuit design.