SWAP-less Implementation of Quantum Algorithms

We present a formalism based on tracking the flow of parity quantum information to implement algorithms on devices with limited connectivity without qubit overhead, SWAP operations or shuttling. Instead, we leverage the fact that entangling gates not only manipulate quantum states but can also be exploited to transport quantum information. We demonstrate the effectiveness of this method by applying it to the quantum Fourier transform (QFT) and the Quantum Approximate Optimization Algorithm (QAOA) with $n$ qubits. This improves upon all state-of-the-art implementations of the QFT on a linear nearest-neighbor architecture, resulting in a total circuit depth of ${5n-3}$ and requiring ${n^2-1}$ CNOT gates. For the QAOA, our method outperforms SWAP networks, which are currently the most efficient implementation of the QAOA on a linear architecture. We further demonstrate the potential to balance qubit count against circuit depth by implementing the QAOA on twice the number of qubits using bi-linear connectivity, which approximately halves the circuit depth.

💡 Research Summary

**
The paper introduces a novel framework called “parity label tracking” that enables the implementation of quantum algorithms on hardware with limited connectivity—specifically linear nearest‑neighbor (NN) architectures—without the need for SWAP gates, qubit shuttling, or additional ancilla qubits. The core idea is to associate each physical qubit with a set‑valued label (P^{(\lambda)}_a) that records which logical Z‑parities (e.g., (\bar Z_i), (\bar Z_i\bar Z_j), etc.) the qubit currently encodes. When a CNOT gate is applied, the label of the target qubit is updated by the symmetric difference of the control and target labels, while the control label remains unchanged. This simple algebraic rule provides a real‑time map from physical operations to logical operations, allowing one to design circuits that deliberately move parity information across the device.

The authors embed this tracking scheme in the Lechner‑Hauke‑Zoller (LHZ) parity code. In the LHZ construction, (n) logical qubits are represented by (n) “base” qubits (holding single‑qubit Z information) and (\frac{n(n-1)}{2}) “parity” qubits (each storing a two‑body Z‑parity). The code is naturally defined on a 2‑D lattice with stabilizers that enforce consistency among the parity qubits. By interpreting one spatial direction of this lattice as a temporal axis, the authors compress the 2‑D structure into a single linear chain of length (n). A “spanning line” is a set of physical qubits that, at a given moment, collectively encode all Z‑parities needed for a particular logical operation. By applying carefully chosen sequences of CNOTs (illustrated in Fig. 2(b) as circuits (C’_1) and (C’_2)), the label configuration of the chain can be transformed from one spanning line to another, effectively “moving” the encoded parity information without any physical movement of qubits.

Quantum Fourier Transform (QFT).
In conventional NN implementations, the QFT requires (\mathcal{O}(n^2)) controlled‑phase gates and a cascade of SWAPs to reverse qubit order, leading to deep circuits. Using parity label tracking, each controlled‑phase gate (\exp(i\theta \bar Z_j \bar Z_k)) is realized as a single‑qubit Z‑rotation on a physical qubit surrounded by a pre‑ and post‑CNOT layer that maps the rotation onto the desired logical parity. Because the label updates are deterministic, the entire QFT can be executed with a circuit depth of (5n-3) and exactly (n^2-1) CNOT gates—both numbers improve on the best known NN QFT constructions. The depth reduction stems from the fact that multiple logical parities can be generated in parallel by exploiting the spanning‑line structure.

Quantum Approximate Optimization Algorithm (QAOA).
QAOA alternates between applying the problem Hamiltonian (a sum of two‑body Z‑parities) and a mixing Hamiltonian. Traditional NN implementations embed the problem Hamiltonian using SWAP networks to bring arbitrary qubit pairs together, which inflates depth and error rates. The parity‑tracking approach directly creates the required two‑body parity on the current spanning line, applies a single‑qubit Z‑rotation, and then re‑encodes the line for the next term. Consequently, the QAOA layers are executed without any SWAPs, yielding a shallower circuit. Benchmarks in the paper show a 30–50 % reduction in depth compared with the optimal SWAP‑based NN implementation for the same problem size.

Bi‑linear Connectivity Trade‑off.
The authors also explore a hardware variant where two linear chains are placed side‑by‑side, forming a “bi‑linear” connectivity pattern that provides limited 2‑D links. By doubling the number of physical qubits, the same algorithm can be mapped onto a richer set of spanning lines, halving the overall depth. This demonstrates a clear trade‑off: modest increases in qubit count can dramatically reduce circuit depth, an important consideration for near‑term noisy devices.

Error‑Correction and Extensions.
Because the method is built on the LHZ code, it naturally inherits the stabilizer structure that can be used for error detection. “Empty” qubits (initialized in (|0\rangle) and carrying an empty label) can be turned into active parity qubits via CNOTs, allowing flexible addition of redundancy. The paper sketches how to define a complementary set of X‑basis labels (Q^{(\lambda)}_a) by exploiting the fact that a Hadamard‑conjugated CNOT swaps control and target, enabling tracking of X‑parities as well. This opens the door to full Clifford‑based circuit optimization and potential integration with fault‑tolerant schemes.

Limitations and Open Questions.
While powerful, the approach has several caveats. The label bookkeeping becomes increasingly complex for large (n), potentially complicating calibration and real‑time error mitigation. The current exposition focuses on Z‑parities; extending the method to algorithms that heavily use X‑ or Y‑type operations may require additional layers of label tracking. Moreover, the reliance on many CNOTs for label updates means that gate errors can accumulate; integrating parity tracking with active error‑correction codes will be essential for scalability. Finally, the paper demonstrates the technique on QFT and QAOA; applying it to other variational algorithms, quantum machine‑learning circuits, or Hamiltonian simulation remains an open research direction.

Impact.
Overall, the work provides a concrete, hardware‑aware methodology for eliminating SWAP overhead in linear NN devices. By converting entangling gates into carriers of logical information rather than mere state manipulators, the authors achieve near‑optimal gate counts and depths for two foundational algorithms. The clear resource estimates (depth (5n-3), CNOT count (n^2-1)) and the demonstrated depth‑vs‑qubit trade‑off make the results immediately relevant for experimental groups working with superconducting qubits, trapped ions, or neutral‑atom arrays that currently suffer from limited connectivity. Future work that integrates parity label tracking with error‑corrected logical qubits could further extend its applicability to fault‑tolerant quantum computing.