Modular Quantum Computing
Modular quantum computing combines processors through quantum links so that a circuit can use qubits in different modules. The research question is when the benefits of smaller, more controllable modules outweigh remote-entanglement cost, waiting and additional faults. Logical all-to-all connectivity does not imply that all remote gates can be executed quickly or simultaneously.
Required background. Distributed quantum computing develops gate teleportation and circuit distribution.
A remote gate is a service with a queue
Section titled “A remote gate is a service with a queue”A teleported gate consumes entanglement, local gates, measurements and classical messages. Its latency depends on available cached pairs and the dependency graph of the algorithm. With link supply and consumption pairs per remote operation, a sustained rate cannot exceed before purification and contention are charged.
Pre-generation can hide link latency, but stored pairs decay. A compiler that minimizes the number of remote gates can still increase critical-path waits. Module size, communication-qubit allocation and circuit partition must therefore be optimized together.
Evidence and competing architectures
Section titled “Evidence and competing architectures”Main and colleagues’ 2025 experiment distributed operations between optically connected trapped-ion modules using heralded entanglement and teleported gates. It establishes small-scale remote-computing operations under the reported fidelity and timing conditions. It does not establish fault-tolerant scaling or an application speedup. Primary report.
Active: link multiplexing, purification, fault-tolerant remote parity measurements, network-aware compilation and preservation of data qubits during repeated link attempts. Optical links favor reconfigurability and separation; local physical connections may offer lower latency but less modular isolation. Conjectural: a universally optimal module size without a workload and hardware model.
Open questions
Section titled “Open questions”When does purification improve accepted logical-gate throughput after consuming extra pairs? Can circuit partitioning exploit slow but high-fidelity links? How should network failures be contained so one module does not stall the entire computation? Which cross-module operations preserve the intended code distance?
Graduate project: partition a circuit under stochastic links
Section titled “Graduate project: partition a circuit under stochastic links”Choose a family of small Clifford-plus-T circuits and distribute them across two or four modules. Compare a cut-minimizing partition with a latency-aware schedule. Simulate heralded link generation, finite entanglement buffers and memory decay. A laptop discrete-event simulator suffices; quantum states need only be simulated for correctness checks at small sizes.
Deliver remote-gate counts, runtime distributions, buffer occupation and a conservative error budget. The baseline uses deterministic link latency equal to its mean. Failure tests use burst outages, unequal links and a circuit dominated by adaptive dependencies. Research begins with a compiler or buffering rule that improves multiple workloads at equal failure probability using measured link parameters.
Entry points and review scope
Section titled “Entry points and review scope”Pair the experiment with fault-tolerant architectures before extrapolating to encoded modules. Sources checked on 11 October 2026; selected experimental anchor: 2025. Update when a complete protected remote operation or a substantially different link/noise regime changes the scheduling assumptions.