Long-ranged gates in quantum computation architectures with limited connectivity

We propose a quantum computation architecture based on geometries with nearest-neighbor interactions, including e.g. planar structures. We show how to efficiently split the role of qubits into data and entanglement-generation qubits. Multipartite entangled states, e.g. 2D cluster states, are generated among the latter, and flexibly transformed via mid-circuit measurements to multiple, long-ranged Bell states, which are used to perform several two-qubit gates in parallel on data qubits. We introduce planar architectures with $n$ data and $n$ auxiliary qubits that allow one to perform $O(\sqrt n)$ long-ranged two-qubit gates simultaneously, with only one round of nearest neighbor gates and one round of mid-circuit measurements. We also show that our approach is applicable in existing superconducting quantum computation architectures, with only a constant overhead.