Quantum algorithms for combinatorial optimisation
Lennart Binkowski · Leibniz Universität Hannover · 2026
Combinatorial optimisation constitutes a central class of computational problems with broad relevance across science and industry, combining foundational theoretical interest with significant real-world impact. Many problems of practical importance admit concise mathematical formulations, yet resist efficient classical solution methods due to their NP-hard structure. Quantum computing offers several algorithmic paradigms that promise structural speedups for such problems. Most quantum and hybrid frameworks fall into one of the following three categories: algorithms based on continuous adiabatic evolution, such as quantum annealing; circuit-based variational quantum algorithms; circuit-based approaches based on Grover's algorithm. These three algorithmic families are conceptualised for different kinds and generations of quantum architectures, often studied in isolation from each other and from preexisting sophisticated classical methods. This thesis develops a unified, predominantly circuit-based treatment of quantum algorithms for combinatorial optimisation, with a particular focus on principled design transfers between near-term parametric algorithms and far-term fault-tolerant approaches. Starting from the Boolean circuit model, reversible and quantum circuit formulations are derived that allow for precise size and depth analysis of classical control logic, reversible embeddings, and quantum realisations of Boolean and pseudo-Boolean functions. These constructions form a common backbone for optimisation algorithms based on adiabatic evolution, variational principles, and amplitude amplification. Adiabatic quantum computation is analysed within this circuit framework, leading to explicit convergence conditions and resource bounds for discretised adiabatic evolutions driven by pseudo-Boolean cost functions of bounded degree. Variational quantum algorithms are then revisited with an emphasis on feasibility-preserving operations and mixing families. In this context, a novel class of exhaustive variational quantum algorithms is introduced, characterised by finite parameterisations that provably reach optimal solutions. These constructions naturally induce state-preparation routines that guarantee to generate superpositions of all feasible solutions. Finally, Grover’s algorithm and its generalisations are studied as optimisation methods, highlighting their structural dependence on feasible-state preparation and their conceptual proximity to variational and adiabatic approaches. By establishing explicit design transfer principles between these paradigms, this thesis provides a unified perspective on quantum optimisation algorithms, accompanied by concrete circuit constructions and quantitative resource analyses.