Swap Networks for Limited Connectivity

Swap Networks for Limited Connectivity

Real quantum hardware rarely lets every qubit interact directly with every other; qubits are laid out with limited connectivity, so two-qubit gates between distant qubits must be routed by physically swapping qubit states until the operands are adjacent. This Cirq tutorial demonstrates swap networks, a structured and highly efficient way to perform such routing, using the acquaintance submodule. As a concrete application it implements a single round of QAOA for Maximum Cut on a linearly connected device, even when the underlying problem graph is arbitrary and dense. The key idea is that a carefully designed sequence of swaps brings every pair of qubits into adjacency exactly once, so all the required two-qubit interaction terms of the cost Hamiltonian can be applied along the way without any wasted movement. The example shows how to express the desired circuit (alternating mixing and phase-separation layers) abstractly in terms of logical qubits, then compile it onto a line of physical qubits using acquaintance strategies that interleave the interaction gates with the swaps. It is a practical, advanced look at the routing and compilation problem that every near-term algorithm must solve to run on connectivity-limited hardware.

Run it

pip install -r requirements.txt
python swap_networks.py

Source and license

Imported from examples/swap_networks.py in quantumlib/Cirq at v1.6.1, under the Apache License 2.0. Original authors: The Cirq Developers. The upstream LICENSE is included alongside this example.