The speedup provided by quantum algorithms with respect to their classical counterparts is at the origin of scientific interest in quantum computation. However, the fundamental reasons for such a speedup are not yet completely understood and deserve further attention. In this context, the classical simulation of quantum algorithms is a useful tool that can help us in gaining insight. Starting from the study of general conditions for classical simulation, we highlight several important differences between two nonequivalent classes of quantum algorithms. We investigate their performance under realistic conditions by quantitatively studying their resilience with respect to static noise. This latter refers to errors affecting the initial preparation of the register used to run an algorithm. We also compare the evolution of the entanglement involved in the different computational processes.
The speedup provided by quantum algorithms with respect to their classical counterparts is at the origin of scientific interest in quantum computation. However, the fundamental reasons for such a speedup are not yet completely understood and deserve further attention. In this context, the classical simulation of quantum algorithms is a useful tool that can help us in gaining insight. Starting from the study of general conditions for classical simulation, we highlight several important differences between two nonequivalent classes of quantum algorithms. We investigate their performance under realistic conditions by quantitatively studying their resilience with respect to static noise. This latter refers to errors affecting the initial preparation of the register used to run an algorithm. We also compare the evolution of the entanglement involved in the different computational processes.
Entanglement is generally accepted as the resource that allows quantum systems to be more efficient in performing classical computation, but the practical utility of entanglement in a variety of quantum information-processing processes is inadequately comprehended. The following paper examines the operational importance, constraints, and trade-offs of entanglement in quantum algorithms, communication protocols, error-correction frameworks and complexity-theoretic settings. Though entanglement (between two or more systems) is a requirement of uniquely quantum behaviours: teleportation, superdense coding, and non-classical correlations, we have found that a larger entanglement does not necessarily improve algorithmic performance. We find through analytical modelling, small-circuit entropy analysis, and comparative analysis of representative use-cases, an optimum entanglement zone, at which performance gains cease to improve with noise, gate infidelity, circuit-depth and hardware decoherence. It has also been found that moderate, noise-resilient entanglement is most useful in algorithmic tasks, structural connectivity is essential in communication tasks, and that quantum error-correction needs highly stable, topology-optimised entangled states rather than maximally entangled states. In addition, resource-theory views indicate conflicts between entanglement entropy, gate cost and classical stimulability thresholds and point to the importance of designing entanglement-efficiently. The paper identifies multiparty entanglement as the bottleneck to scalable quantum advantage, especially with the constraints in NISQ era. Comprehensively, this paper has provided a single analytical framework explaining when, where and how entanglement has real computational benefit. The analysis argues the development of algorithms and hardware in a so-called just-enough entanglement paradigm, to optimise fidelity, limit the effects of noise, and have scalable quantum architectures.
In principle, a quantum system could be used to simulate another quantum system. The purpose of such a simulation would be to obtain information about problems which are difficult to simulate on a classical computer due to the exponential increase of the Hilbert space with the size of the system and which cannot be readily measured or controlled in an experiment. The system will interact with the surrounding environment and with the other particles in the system, and be implemented using imperfect controls, making it subject to noise. It has been suggested that noise does not need to be controlled to the same extent as it must be for general quantum computing. However, the effects of noise in quantum simulations are not well understood and how best to treat them in most cases is not known. In this paper we study an existing quantum algorithm for simulating the one-dimensional Fano-Anderson model using a liquid-state NMR device. We examine models of noise in the evolution using different initial states in the original model. We also add interacting spins to simulate a realistic situation where an environment of spins is present. We find that states which are entangled with their environment, and sometimes correlated but not necessarily entangled, have an evolution which is described by maps which are not completely positive. We discuss the conditions for this to occur and also the implications.
Quantum computation is by far the most powerful computational model allowed by the laws of physics. By carefully manipulating microscopic systems governed by quantum mechanics, one can efficiently solve computational problems that may be classically intractable; on the contrary, such speed-ups are rarely possible without the help of classical computation, since most quantum algorithms heavily rely on subroutines that are purely classical. A better understanding of the relationship between classical and quantum computation is indispensable, in particular in an era where the first quantum device exceeding classical computational power is within reach. In the first part of the thesis, we study some differences between classical and quantum computation. We first show that quantum cryptographic hashing is maximally resilient against classical leakage, a property beyond reach for any classical hash function. Next, we consider the limitation of strong (amplitude-wise) simulation of quantum computation. We prove an unconditional and explicit complexity lower bound for a category of simulations called monotone strong simulation and further prove conditional complexity lower bounds for general strong simulation techniques. Both results indicate that strong simulation is fundamentally unscalable. In the second part of the thesis, we propose classical algorithms that facilitate quantum computing. We propose a new classical algorithm for the synthesis of a quantum algorithm paradigm called quantum signal processing. Empirically, our algorithm demonstrates numerical stability and acceleration of more than one magnitude compared to state-of-the-art algorithms. Finally, we propose a randomized algorithm for transversally switching between arbitrary stabilizer quantum error-correcting codes. It has the property of preserving the code distance and thus might prove useful for designing fault-tolerant code-switching schemes.
While it seems possible that quantum computers may allow for algorithms offering a computational speed-up over classical algorithms for some problems, the issue is poorly understood. We explore this computational speed-up by investigating the ability to de-quantise quantum algorithms into classical simulations of the algorithms which are as efficient in both time and space as the original quantum algorithms. The process of de-quantisation helps formulate conditions to determine if a quantum algorithm provides a real speed-up over classical algorithms. These conditions can be used to develop new quantum algorithms more effectively (by avoiding features that could allow the algorithm to be efficiently classically simulated) and to create new classical algorithms (by using features which have proved valuable for quantum algorithms). Results on many different methods of de-quantisations are presented, as well as a general formal definition of de-quantisation. De-quantisations employing higher-dimensional classical bits, as well as those using matrix-simulations, put emphasis on entanglement in quantum algorithms; a key result is that any algorithm in which the entanglement is bounded is de-quantisable. These methods are contrasted with the stabiliser formalism de-quantisations due to the Gottesman-Knill Theorem, as well as those which take advantage of the topology of the circuit for a quantum algorithm. The benefits and limits of the different methods are discussed, and the importance of utilising a range of techniques is emphasised. We further discuss some features of quantum algorithms which current de-quantisation methods do not cover and highlight several important open questions in the area.
As quantum processors move into the noisy intermediate-scale quantum (NISQ) era, the central question is no longer whether a quantum algorithm is asymptotically faster than its classical counterpart, but whether it can retain a meaningful speedup under realistic noise, compilation, and measurement constraints. This work develops a quantitative framework for characterizing the noise resilience of quantum advantage. We introduce the advantage robustness function RA(ε), which measures how the asymptotic and concrete performance of an algorithm A degrades with physical error rate ε, and we compute it analytically or numerically for twelve representative algorithm families, including Grover search, amplitude estimation, phase estimation, HHL, VQE/QAOA, Hamiltonian simulation, and sampling algorithms. For each class, we derive critical noise thresholds and depth limits beyond which classical algorithms regain supremacy, explicitly incorporating realistic connectivity and compilation overheads. We show that algorithms based on coherent amplitude amplification are more fragile to coherent errors than depolarizing noise, while variational algorithms display a tradeoff between expressivity and robustness. By combining these algorithm-specific robustness curves with current and projected hardware error rates for superconducting, trapped-ion, neutralatom, and photonic platforms, we map out a phase diagram of practical quantum advantage that links problem size, noise rate, and algorithm type. Our analysis identifies promising near-term application windows in risk analytics, small-molecule chemistry, and structured optimization, and also delineates regions where no advantage is realistic, providing valuable guidance for both algorithm designers and hardware architects.
While it seems possible that quantum computers may allow for algorithms offering a computational speed-up over classical algorithms for some problems, the issue is poorly understood. We explore this computational speed-up by investigating the ability to de-quantise quantum algorithms into classical simulations of the algorithms which are as efficient in both time and space as the original quantum algorithms. The process of de-quantisation helps formulate conditions to determine if a quantum algorithm provides a real speed-up over classical algorithms. These conditions can be used to develop new quantum algorithms more effectively (by avoiding features that could allow the algorithm to be efficiently classically simulated), as well as providing the potential to create new classical algorithms (by using features which have proved valuable for quantum algorithms). Results on many different methods of de-quantisations are presented, as well as a general formal definition of de-quantisation. De-quantisations employing higher-dimensional classical bits, as well as those using matrix-simulations, put emphasis on entanglement in quantum algorithms; a key result is that any algorithm in which the entanglement is bounded is de-quantisable. These methods are contrasted with the stabiliser formalism de-quantisations due to the Gottesman-Knill Theorem, as well as those which take advantage of the topology of the circuit for a quantum algorithm. The benefits of the different methods are contrasted, and the importance of a range of techniques is emphasised. We further discuss some features of quantum algorithms which current de-quantisation methods do not cover.