Practical quantum computation requires high-fidelity instruction executions on qubits. Among them, Clifford instructions are relatively easy to perform, while non-Clifford instructions require the use of magic states. This makes magic state distillation a central procedure in fault-tolerant quantum computing. A magic state distillation factory consumes many low-fidelity input magic states and produces fewer, higher-fidelity states. To reach high fidelities, multiple distillation factories are typically chained together into a multi-level pipeline, consuming significant quantum computational resources. Our work optimizes the resource usage of distillation pipelines by introducing a novel dynamic pipeline architecture. Observing that distillation pipelines consume magic states in a burst-then-steady pattern, we develop dynamic factory scheduling and resource allocation techniques that go beyond existing static pipeline organizations. Dynamic pipelines reduce the qubit cost by 16%-70% for large-scale quantum applications and achieve average reductions of 26%-37% in qubit-time volume on generated distillation benchmarks compared to state-of-the-art static architectures. By significantly reducing the resource overhead of this building block, our work accelerates progress towards the practical realization of fault-tolerant quantum computers.
Practical quantum computation requires high-fidelity instruction executions on qubits. Among them, Clifford instructions are relatively easy to perform, while non-Clifford instructions require the use of magic states. This makes magic state distillation a central procedure in fault-tolerant quantum computing. A magic state distillation factory consumes many low-fidelity input magic states and produces fewer, higher-fidelity states. To reach high fidelities, multiple distillation factories are typically chained together into a multi-level pipeline, consuming significant quantum computational resources. Our work optimizes the resource usage of distillation pipelines by introducing a novel dynamic pipeline architecture. Observing that distillation pipelines consume magic states in a burst-then-steady pattern, we develop dynamic factory scheduling and resource allocation techniques that go beyond existing static pipeline organizations. Dynamic pipelines reduce the qubit cost by 16%-70% for large-scale quantum applications and achieve average reductions of 26%-37% in qubit-time volume on generated distillation benchmarks compared to state-of-the-art static architectures. By significantly reducing the resource overhead of this building block, our work accelerates progress towards the practical realization of fault-tolerant quantum computers.
Magic State Distillation is considered to be one of the promising methods for supplying the non-Clifford resources required to achieve universal fault tolerance. Conventional MSD protocols implemented in surface codes often require multiple code blocks and lattice surgery rounds, resulting in substantial qubit overhead, especially at low target error rates.
In this work, we present practical magic state distillation factories on Bivariate Bicycle (BB) codes that execute Pauli-measurement-based Clifford circuits inside a single BB code block. We formulate distillation circuit design as a joint optimization of logical qubit mapping, gate scheduling, measurement nativization, and protocol compression via qubit recycling. Based on detailed resource analysis and simulations, our BB factories have space-time volume comparable to that of leading distillation factories while delivering lower target error at a smaller qubit footprint, and are particularly compelling as second-round distillers following magic state cultivations.
Quantum computation promises exponential or polynomial speedups over classical computation for a wide range of problems, including integer factoring, quantum chemistry, and linear-system solving.However, current quantum devices remain too noisy to execute complex algorithms reliably.Overcoming this noise and achieving quantum advantage require a fault-tolerant quantum computer (FTQC).The resource overhead associated with fault-tolerant operation is substantial, and reducing this overhead remains one of the central challenges.A major source of this overhead arises from the implementation of non-Clifford gates, which are essential for universal quantum computation.A non-Clifford gate such as the T gate is implemented using a special quantum state known as a magic state, which is prepared via a process called magic state distillation.Magic state distillation is resource-intensive and remains a major bottleneck in FTQC.Hence, reducing the distillation overhead is necessary to lower the overall FTQC overhead.The large overhead of magic state distillation has significantly influenced FTQC architectures.In particular, it has motivated compilation schemes optimized under the constraint of expensive magic state distillation.While reducing the distillation overhead is crucial, it is not sufficient; lowering the overall FTQC overhead also requires revisiting design choices guided by this constraint.Therefore, efficient methods are required not only for preparing magic states but also for consuming them.In this thesis, we reduce the overhead of fault-tolerant quantum computation by focusing on the preparation and consumption of magic states.In contrast to traditional logical-level distillation, physical-level distillation allows direct implementation with physical gates and has the potential to drastically reduce the distillation overhead.However, existing physical-level distillation protocols, such as zero-level distillation, cannot directly yield high-fidelity magic states.We propose two magic state distillation protocols that use physical-level distillation techniques to efficiently prepare high-fidelity magic states.The first protocol is a two-level distillation protocol using physical-level distillation as the first-level stage.The second protocol utilizes lattice surgery to reduce the spacetime overhead of a physical-level distillation protocol.The former targets a high-fidelity regime, whereas the latter targets an intermediate-fidelity regime-lower than the former, but higher than zero-level distillation.Next, we investigate FTQC architectures that can efficiently consume the magic states prepared by these protocols.Because magic state distillation is probabilistic, occasional failures can delay computation.Moreover, physical-level distillation protocols generally exhibit higher failure probabilities than logical-level ones.This delay is amplified by parallel execution, making it an i ii important consideration for FTQC using physical-level distillation protocols.We therefore propose a scheme to mitigate the negative performance impact caused by magic state distillation failures by introducing a pool of magic states for each distillation factory.The proposed distillation protocols allow the simultaneous preparation of many magic states.Leveraging this capability, we propose an FTQC compilation scheme optimized for highly parallel quantum algorithms.By performing magic state distillation locally and consuming magic states in parallel, the scheme enables highly parallel execution and consequently reduces the computational time overhead.Overall, this thesis presents a unified approach to reducing the overhead of fault-tolerant quantum computation.By improving both the preparation and consumption of magic states, it advances the design of resource-efficient FTQC architectures.These results demonstrate that efficient magic state preparation motivates efficient magic state consumption, and vice versa.Together, they bring us closer to the practical realization of fault-tolerant quantum computers.
Zero-level CCZ Distillation
Magic state distillation is a key component of fault-tolerant quantum computation, as it enables the implementation of non-Clifford gates such as the T gate and the CCZ gate via gate teleportation. However, conventional distillation protocols require a large number of logical qubits and introduce substantial spatial and temporal overhead, posing a significant bottleneck for scalable fault-tolerant quantum computation. In this work, we propose a zero-level distillation protocol that efficiently generates a high-fidelity logical CCZ magic state using only physical qubits on a two-dimensional square lattice with nearest-neighbor interactions. Our method leverages the transversal T/T† operation of the [[8,3,2]] code to fault-tolerantly encode the state CCZ∣+++⟩, which is subsequently teleported to three surface-code logical qubits via lattice surgery. To enable teleportation between codes with different distances, we introduce adaptively initialized teleportation (AIT), a tailored initialization procedure for the surface code. Numerical simulations demonstrate that the logical error rate scales as pL≃300×p2 with respect to the physical error rate p. For example, the proposed method improves the logical error rate by approximately one and two orders of magnitude at p=10−3 and p=10−4, respectively, compared to conventional seven-T-gate approaches. The distillation circuit requires only 22 physical qubits, 3 logical qubits, and a circuit depth of 24, reducing the space-time overhead by a factor of approximately 5-10 compared to previous methods. This result highlights the practicality of CCZ-state distillation in early fault-tolerant quantum computation and offers a new direction toward resource-efficient physical-level magic state distillation beyond conventional T-state generation.
We show how looped pipeline architectures-which use short-range shuttling of physical qubits to achieve bounded non-local connectivity-can efficiently implement the fault-tolerant non-Clifford gate between 2D surface codes described in (Sci. Adv. 6, eaay4929 (2020)). The shuttling schedule required is only marginally more complex than is required for implementing the standard 2D surface code in this architecture. We compare the resource cost with the cost of magic state distillation and find that, at present, this comparison is heavily in favour of distillation. The high cost of the non-Clifford gate is largely due to the relatively low performance of the just-in-time decoder used in the procedure, which necessitates very large code distances in order to achieve suitably low logical error rates. We argue that, as little attention has been given to the study and optimisation of these decoders, there are potentially significant improvements to be made in this area.
Despite significant overhead reductions since its first proposal, magic state distillation is often considered to be a very costly procedure that dominates the resource cost of fault-tolerant quantum computers. The goal of this work is to demonstrate that this is not true. By writing distillation circuits in a form that separates qubits that are capable of error detection from those that are not, most logical qubits used for distillation can be encoded at a very low code distance. This significantly reduces the space-time cost of distillation, as well as the number of qubits. In extreme cases, it can cost less to distill a magic state than to perform a logical Clifford gate on full-distance logical qubits.
A promising approach to achieving scalable fault-tolerant quantum computation is the use of quantum error correction (QEC) codes augmented with magic states i.e. resource states produced via distillation, cultivation, or Rz synthesis and teleported into the circuit as needed. Because magic-state production dominates the space-time volume of fault-tolerant programs, system architects must decide how many production units to allocate. Current approaches rely on deterministic analysis that either provisions for worst-case peak demand (wasting valuable qubit resources on factories that are never simultaneously utilized) or assumes average demand, which increases execution time. In this work, we build a simulation framework that couples circuit scheduling with different stochastic magic state production models, and use it to quantify the impact of non-determinism on circuit execution. We show that non-determinism has a dual effect that deterministic models cannot capture: it inflates total execution time (the price), while deflating peak per-cycle resource demand (the payoff). For distillation-based architectures, this demand smoothing shifts the space-time-optimal provisioning point: fewer factories are needed to minimize space-time volume than deterministic analysis predicts. Across benchmarks, stochastic-aware provisioning reduces space-time volume by up to 27% compared to the deterministic optimum for distillation, while requiring up to 30% fewer factories. We characterize these effects across each preparation mechanism, map the resulting design-space tradeoffs, and demonstrate that static resource estimation systematically mis-characterizes the cost of fault-tolerant execution. Our results establish that stochastic-aware analysis is necessary for right-sizing the factory allocations and should replace deterministic heuristics as the standard methodology for FTQC resource planning.