Fermionic Swap Networks
Reordering orbitals with fermionic swap gates so that all interaction terms become local, taming Jordan-Wigner string overhead on limited-connectivity hardware.
The connectivity problem
Molecular Hamiltonians couple every pair of orbitals, but real hardware connects only neighboring qubits. Under Jordan-Wigner, a term between distant orbitals also carries a long Z string. Both problems, hardware layout and encoding non-locality, can be handled together with fermionic swap networks.
The fermionic swap gate
An ordinary swap exchanges two qubits' states. A fermionic swap (fswap) exchanges two adjacent orbitals while correctly tracking the antisymmetry sign, effectively renumbering the Jordan-Wigner ordering. After an fswap, two orbitals that were far apart become adjacent, so their interaction term is now local and its Z string vanishes.
The network
A scheduled sequence of fswaps, arranged like a sorting network, brings every pair of orbitals adjacent at some point during the sweep. As each pair meets, the corresponding Hamiltonian term is applied as a local gate. Kivlichen et al. showed that a linear swap network lets a full electronic-structure Trotter step run on a line of qubits with depth O(N) rather than O(N^2).
- fswaps convert non-local terms into local gates by reordering orbitals.
- A sorting-network schedule ensures every pair becomes adjacent once.
- Trotter-step depth drops to O(N) on a linear qubit array.
- Only nearest-neighbor connectivity is required, matching many devices.
Why it matters
Without swap networks, dense chemistry Hamiltonians would demand long-range gates or long Jordan-Wigner strings, both expensive and error-prone. Swap networks make the geometry of the hardware and the ordering of the encoding work in concert, and they are a standard ingredient in low-depth chemistry compilation. They illustrate a general lesson in quantum simulation: the physical layout, the fermionic encoding, and the gate schedule must be co-designed to keep circuits shallow.