Adaptive Entanglement Management in Quantum Multi-Core Architectures

As quantum computing transitions from monolithic chips to scalable multi-core architectures, inter-core communication becomes a critical bottleneck. Because qubits are fragile, they cannot be physically moved between cores; instead, they rely on entanglement-assisted quantum teleportation. This paper investigates how to optimize the generation of entangled EPR pairs to minimize the latency of these inter-core operations.

The authors compare three entanglement management paradigms using an extended SeQUeNCe simulator on mesh-based multi-core architectures:

• On-Demand Generation (ODG): A reactive approach that generates entanglement only when a communication request is received.
• Continuous Pre-Generation (CGP): A proactive approach that generates EPR pairs in the background using idle communication qubits.
• Adaptive Continuous Pre-Generation (ACGP): An intelligent, proactive approach that dynamically updates the probability of generating entanglement with specific neighbors based on historical communication traffic patterns.

The researchers evaluated these protocols using standard quantum benchmarks (QFT, Cuccaro Adder, Draper Adder, and MCTMV) across various mesh sizes, measuring both average teleportation latency and the resulting entanglement fidelity.

The results demonstrate that ACGP consistently outperforms both ODG and CGP in reducing average teleportation latency. By learning which cores communicate most frequently, ACGP ensures that entangled pairs are ready when needed, effectively acting as a cache for quantum resources. While pre-generation can lead to fidelity degradation due to the limited coherence time of stored qubits, the authors note that entanglement purification can restore fidelity with minimal impact on overall system latency. This suggests that adaptive management is a highly effective strategy for improving communication efficiency in scalable quantum systems.