arxiv: 2604.27015 · v2 · submitted 2026-04-29 · 🪐 quant-ph

Recognition: unknown

Congestion-free routing on quantum chips

Mithilesh Kumar , Yusuf Tahir , Varun Daiya , Sanjana Mattaparthi , Aarav Shaurya

Authors on Pith no claims yet

Pith reviewed 2026-05-07 13:31 UTC · model grok-4.3

classification 🪐 quant-ph

keywords qudit routingspectral busesswap-freequantum congestionnonlocal gatesquantum compilationHilbert space expansionBoolean fan-in

0 comments

The pith

Higher levels of qudits serve as spectral buses that carry control signals without moving computational states or creating routing congestion.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces a routing approach for quantum hardware with only nearest-neighbor links, where extra energy levels in each qudit act as independent channels for control information. Standard SWAP-based methods move the actual qubit states along paths, which both adds depth and blocks other operations when routes overlap. Here the computational state stays put while control travels on orthogonal subspaces encoded in the same devices, cutting the primitive count for a distance-L gate from 3L to 2L+1. The same mechanism supports multi-control operations by letting several incoming buses trigger a local Boolean function at the target. The approach is shown to be correct for CNOT and fan-in gates, with simulations confirming no crosstalk and reduced circuit depth on QFT, QAOA, and mirror circuits.

Core claim

We present a swap-free routing framework in which higher levels of a qudit act as orthogonal spectral buses that transport control information without moving the computational state. We show that exact congestion relief in nearest-neighbor architectures requires local Hilbert-space expansion. In this model, a nonlocal operation over a path of length L requires 2L+1 logical routing primitives, compared to the 3L baseline. Overlapping routes remain distinguishable through bus labels encoded in the same physical qudits. This routing algebra extends to Boolean fan-in at a common target, yielding multi-control operations of depth 2L + D_g + O(1).

What carries the argument

The spectral bus realized by higher qudit levels, which encodes and transports control labels in an orthogonal subspace while leaving the computational state stationary.

If this is right

A nonlocal gate along a path of length L costs 2L+1 routing primitives rather than 3L.
Distinct bus labels allow overlapping routes to execute simultaneously on the same physical qudits without interference.
Boolean fan-in of size K at a target produces multi-control gates of depth 2L + D_g + O(1).
Exact overlap routing requires qudit dimension d at least 2^{K+1} for K controls.
Compiler runs on QFT, QAOA, and mirror-interaction circuits exhibit the predicted drop in transport overhead.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Hardware priorities may shift toward devices that already provide controllable higher levels rather than adding more physical couplers.
The clean separation between control delivery on buses and local aggregation at targets could simplify compilation for other multi-qubit primitives beyond CNOT.
The dimension lower bound suggests a concrete test: implement fan-in K on a qudit whose size is just above or below 2^{K+1} and measure whether exact routing fails.

Load-bearing premise

Higher levels of the qudits can be controlled and maintained with sufficient coherence and speed to realize the spectral buses without introducing prohibitive noise or crosstalk, and that local Hilbert-space expansion is practically feasible on the target hardware.

What would settle it

An experiment or simulation measuring a distance-2 CNOT that records whether the total routing steps stay at or below 5 primitives while fidelity remains comparable to the SWAP baseline, or whether higher-level decoherence exceeds the saved depth.

Figures

Figures reproduced from arXiv: 2604.27015 by Aarav Shaurya, Mithilesh Kumar, Sanjana Mattaparthi, Varun Daiya, Yusuf Tahir.

**Figure 1.** Figure 1: FIG. 1. Routed Boolean fan-in in the spectral-routing model. Multiple source controls view at source ↗

**Figure 2.** Figure 2: FIG. 2. Logical routing cost decomposition for Boolean fan-in. At the level of routing primitives, the spectral-routing framework view at source ↗

**Figure 3.** Figure 3: FIG. 3. QuTiP length-sweep plots for routed fan-in. Left: ideal-limit run with view at source ↗

**Figure 4.** Figure 4: FIG. 4. Concrete Cirq depth scaling for the routed fan-in construction. The plotted circuit moments agree with the linear-in- view at source ↗

**Figure 5.** Figure 5: FIG. 5. Compiled circuit depth for naive SWAP routing and the proposed qudit-based routing protocol as a function of path view at source ↗

**Figure 6.** Figure 6: FIG. 6. Validation of the congestion-round theorem on representative conflict-graph families. Left: line-overlap routes. view at source ↗

**Figure 7.** Figure 7: FIG. 7. State fidelity as a function of path length view at source ↗

**Figure 8.** Figure 8: FIG. 8. Win region for the routed protocol under decoher view at source ↗

read the original abstract

Limited connectivity makes nonlocal quantum gates expensive on near-neighbor hardware, where compilation typically relies on SWAP transport, inheriting both depth overhead and path congestion. We present a swap-free routing framework in which higher levels of a qudit act as orthogonal spectral buses that transport control information without moving the computational state. We show that exact congestion relief in nearest-neighbor architectures requires local Hilbert-space expansion. In this model, a nonlocal operation over a path of length $L$ requires $2L+1$ logical routing primitives, compared to the $3L$ baseline. Overlapping routes remain distinguishable through bus labels encoded in the same physical qudits. This routing algebra extends to Boolean fan-in at a common target: multiple controls arriving on distinct buses trigger a local unitary based on an arbitrary Boolean function of bus digits, yielding multi-control operations of depth $2L + D_g + O(1)$ for fan-in size $K$ and target-synthesis cost $D_g$. We prove decodability, reversibility, and correctness for CNOT and Boolean fan-in, along with a state-count lower bound $d \geq 2^{K+1}$ for exact overlap routing. Cirq simulations confirm single-control correctness and zero crosstalk. Compiler-level benchmarks on QFT, QAOA, and mirror-interaction circuits verify the predicted congestion law and transport reduction. Noisy QuTiP simulations show that the architectural advantage depends on higher-level coherence and speed. These results identify spectral qudit routing as a congestion-relief architecture that separates nonlocal control delivery from local target-side aggregation, providing a minimal mechanism for overcoming qubit routing limitations.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The qudit spectral bus idea claims a clean 2L+1 routing depth by moving control info in higher levels instead of SWAPs, but the practical edge still depends on unshown coherence in those levels.

read the letter

This paper's main contribution is a routing framework that encodes control signals in extra qudit levels as orthogonal spectral buses. The claim is that this avoids moving computational states along paths, delivering nonlocal operations in 2L+1 primitives rather than the usual 3L on nearest-neighbor hardware. Overlaps are handled by bus labels, and the same setup extends to multi-control Boolean fan-in at a target with depth 2L plus the local synthesis cost plus a small constant. They also derive a state-count lower bound d at least 2 to the K+1 for exact routing with K controls. Cirq runs confirm single-control correctness with no crosstalk, and compiler benchmarks on QFT, QAOA, and mirror circuits line up with the predicted savings. Noisy QuTiP runs are mentioned to show the benefit tracks higher-level coherence and speed. The algebra for reversibility and decodability looks internally consistent on paper. The separation between routing delivery and target aggregation is a useful way to think about the problem. The soft spot is the hardware premise. The advantage vanishes if higher levels introduce leakage, crosstalk, or slower gates, and the abstract only notes that the noisy simulations depend on those factors without giving concrete rates, thresholds, or a mapping to any real device. Full derivations are not visible from the abstract, so the soundness of the proofs cannot be checked in detail. This is for compilation researchers and people working on qudit or limited-connectivity hardware who want alternatives to standard SWAP networks. A reader interested in routing bounds or fan-in constructions would find the primitives and the dimension requirement worth looking at. It deserves a serious referee because the framework is distinct from existing techniques and has initial simulation support, even though the practical claims need tighter evidence on noise. I would send it for review and ask for the full proofs plus a quantitative noise analysis to see whether the depth reduction actually materializes before errors dominate.

Referee Report

3 major / 2 minor

Summary. The manuscript proposes a swap-free routing framework for nearest-neighbor quantum hardware in which higher levels of qudits serve as orthogonal spectral buses to transport control information without displacing the computational state. It claims that a nonlocal operation along a path of length L requires only 2L+1 logical routing primitives (versus the 3L SWAP baseline), that overlapping routes remain distinguishable via bus labels, and that the approach extends to Boolean fan-in multi-control gates with depth 2L + D_g + O(1). The paper states proofs of decodability, reversibility, and correctness for CNOT and fan-in operations, a state-count lower bound d ≥ 2^{K+1}, and supports the claims with Cirq simulations (zero crosstalk) plus compiler benchmarks on QFT/QAOA/mirror circuits; noisy QuTiP runs indicate the advantage depends on higher-level coherence and speed.

Significance. If the higher-level control and coherence assumptions hold, the framework offers a concrete architectural mechanism to decouple control delivery from state transport, potentially lowering routing depth and congestion in scalable quantum compilers beyond what SWAP networks or teleportation achieve. The explicit primitive count, bus-label distinguishability, and Boolean aggregation algebra provide falsifiable predictions that could be tested on qudit hardware.

major comments (3)

[Proofs of decodability, reversibility, and correctness] The abstract and introduction assert proofs of decodability, reversibility, and correctness for the CNOT and Boolean fan-in cases, yet the manuscript supplies neither the full derivations nor the key algebraic steps establishing that 2L+1 primitives suffice while preserving orthogonality; without these, the central depth-reduction claim cannot be verified as load-bearing.
[Noisy QuTiP simulations] Noisy QuTiP simulations are invoked to show that the 2L+1 advantage depends on higher-level coherence and speed, but the text provides no concrete error model, leakage rates, crosstalk bounds, or threshold calculation indicating the regime in which 2L+1 outperforms 3L before noise dominates; this assumption is load-bearing for the practical congestion-relief claim.
[State-count lower bound] The state-count lower bound d ≥ 2^{K+1} for exact overlap routing with fan-in K is stated as necessary, but the derivation or tightness argument is not supplied, leaving open whether the bound is tight or whether smaller d suffices under the bus-label encoding.

minor comments (2)

[Abstract] The abstract references Cirq simulations confirming single-control correctness and zero crosstalk, yet no corresponding figure, table, or data excerpt is cited in the summary or results section.
[Routing primitives] Notation for bus labels and the distinction between logical routing primitives and physical gates is introduced without an early definition or table, which may hinder readability for readers unfamiliar with the spectral-bus model.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the careful review and valuable suggestions. Below we respond point-by-point to the major comments, outlining the revisions we will implement.

read point-by-point responses

Referee: [Proofs of decodability, reversibility, and correctness] The abstract and introduction assert proofs of decodability, reversibility, and correctness for the CNOT and Boolean fan-in cases, yet the manuscript supplies neither the full derivations nor the key algebraic steps establishing that 2L+1 primitives suffice while preserving orthogonality; without these, the central depth-reduction claim cannot be verified as load-bearing.

Authors: We agree that the full derivations should be more prominently featured. The manuscript states the proofs and outlines the routing algebra in Section 3, with supporting steps in the appendix, but the explicit algebraic verification that 2L+1 primitives preserve orthogonality was not expanded in the main text. We will revise by adding the complete derivations for decodability, reversibility, and correctness, including the step-by-step mappings for both CNOT and fan-in cases. revision: yes
Referee: [Noisy QuTiP simulations] Noisy QuTiP simulations are invoked to show that the 2L+1 advantage depends on higher-level coherence and speed, but the text provides no concrete error model, leakage rates, crosstalk bounds, or threshold calculation indicating the regime in which 2L+1 outperforms 3L before noise dominates; this assumption is load-bearing for the practical congestion-relief claim.

Authors: We acknowledge that the noisy simulations use generic models without explicit parameters. In the revised manuscript we will add a concrete error model specifying leakage rates to higher levels, crosstalk bounds between spectral modes, and a threshold calculation based on representative qudit coherence times to delineate the regime where the 2L+1 routing outperforms the 3L baseline. revision: yes
Referee: [State-count lower bound] The state-count lower bound d ≥ 2^{K+1} for exact overlap routing with fan-in K is stated as necessary, but the derivation or tightness argument is not supplied, leaving open whether the bound is tight or whether smaller d suffices under the bus-label encoding.

Authors: The bound follows from the need to distinguish 2^K control combinations plus the target state under overlapping routes. We will include the full derivation in the revision, showing that d < 2^{K+1} produces collisions in the bus-label encoding, together with a tightness argument via an explicit encoding that saturates the bound. revision: yes

Circularity Check

0 steps flagged

No circularity: routing counts derived from explicit bus model definitions

full rationale

The paper defines a new swap-free routing scheme via higher qudit levels acting as labeled spectral buses, then derives the 2L+1 primitive count, decodability, and Boolean fan-in depth directly from those definitions and local unitaries. No fitted parameters are renamed as predictions, no self-citations load-bear the central claims, and no uniqueness theorems or ansatzes are imported from prior author work. The algebra is internally consistent by construction of the model (bus labels, orthogonality, local aggregation) and is verified by explicit proofs plus Cirq/QuTiP simulations rather than reducing to inputs. This is a standard non-circular proposal of an architecture with its own derived costs.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 2 invented entities

The central claim rests on expanding the local Hilbert space to create distinguishable spectral channels and on the assumption that higher-level operations can be performed coherently; these are not standard in qubit routing but are introduced here as architectural primitives.

free parameters (1)

qudit dimension d
Must be chosen at least 2^{K+1} to support exact overlap routing for fan-in size K

axioms (1)

standard math Standard quantum mechanics on finite-dimensional qudit Hilbert spaces
Assumes unitary operations and orthogonality of levels as in conventional qudit models

invented entities (2)

spectral bus no independent evidence
purpose: Transport control information orthogonally without displacing the computational state
New conceptual channel encoded in higher qudit levels
bus label no independent evidence
purpose: Distinguish overlapping routes encoded in the same physical qudits
Mechanism for zero-crosstalk overlap routing

pith-pipeline@v0.9.0 · 5605 in / 1542 out tokens · 54963 ms · 2026-05-07T13:31:40.791608+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

31 extracted references · 5 canonical work pages

[1]

G. Li, Y. Ding, and Y. Xie, Tackling the qubit map- ping problem for nisq-era quantum devices (2019), arXiv:1809.02573 [cs.ET]

work page arXiv 2019
[2]

On the qubit routing problem.arXiv preprint arXiv:1902.08091, 2019

A. Cowtan, S. Dilkes, R. Duncan, W. Simmons, and S. Sivarajah, On the qubit routing problem, arXivpreprint arXiv:1902.08091 (2019)

work page arXiv 1902
[3]

Murali, J

P. Murali, J. M. Baker, A. J. Abhari, F. T. Chong, and M. Martonosi, Noise-adaptive compiler mappings for noisy intermediate-scale quantum computers, inProceedings of the Twenty-Fourth International Conference on Architec- tural Support for Programming Languages and Operating Systems(2019) pp. 367–380

2019
[4]

B. P. Lanyon, M. Barbieri, M. P. Almeida, T. Jennewein, M. S. Rashid, and A. G. White, Simplifying quantum logic using higher-dimensional hilbert spaces, Nature Physics 5, 134 (2009)

2009
[5]

Gokhale, J

P. Gokhale, J. M. Baker, C. Duckering, N. C. Brown, K. R. Brown, and F. T. Chong, Asymptotic algorithms for qu- dit quantum computing, arXiv preprint arXiv:1905.10481 (2019)

work page arXiv 1905
[6]

Y. Wang, Z. Hu, B. C. Sanders, and S. Kais, Qudits and high-dimensional quantum computing, Frontiers in PhysicsVolume 8 - 2020, 10.3389/fphy.2020.589504 (2020)

work page doi:10.3389/fphy.2020.589504 2020
[7]

Klaver, S

B. Klaver, S. M. A. Rombouts, M. Fellner, A. Messinger, K. Ender, K. Ludwig, and W. Lechner, Swap-less im- plementation of quantum algorithms, Phys. Rev. A113, 012443 (2026)

2026
[8]

J. Koch, T. M. Yu, J. Gambetta, A. A. Houck, D. I. Schuster, J. Majer, A. Blais, M. H. Devoret, S. M. Girvin, and R. J. Schoelkopf, Charge-insensitive qubit design derived from the cooper pair box, Physical Review A76, 042319 (2007)

2007
[9]

N. Goss, S. Ferracin, A. Hashim, A. Carignan-Dugas, J. M. Kreikebaum, R. K. Naik, D. I. Santiago, and I. Siddiqi, Extending the computational reach of a superconduct- ing qutrit processor, npj Quantum Information10, 101 (2024)

2024
[10]

Y. Chi, J. Huang, Z. Zhang, J. Mao, Z. Zhou, X. Chen, C. Zhai, J. Bao, T. Dai, H. Yuan,et al., A programmable qudit-based quantum processor, Nature Communications 13, 1166 (2022)

2022
[11]

R. W. Heeres, P. Reinhold, N. Ofek, L. Frunzio, L. Jiang, M. H. Devoret, and R. J. Schoelkopf, Implementing a uni- versal gate set on a logical qubit encoded in an oscillator, Nature Communications8, 1 (2017)

2017
[12]

K. N. Smith, G. S. Ravi, T. Alexander, N. T. Bronn, A. R. Carvalho, A. Cervera-Lierta, F. T. Chong, J. M. Chow, M. Cubeddu, A. Hashim,et al., Programming physical quantum systems with pulse-level control, Frontiers in Physics10, 900099 (2022)

2022
[13]

Ringbauer, M

M. Ringbauer, M. Meth, L. Postler, R. Stricker, R. Blatt, P. Schindler, and T. Monz, A universal qudit quantum processor with trapped ions, Nature Physics18, 1053 (2022)

2022
[14]

X. Shi, J. Sinanan-Singh, T. Burke, J. Chiaverini, and I. Chuang, Efficient implementation of a quantum algorithm with a trapped ion qudit, arXiv preprint arXiv:2506.09371 (2025)

work page arXiv 2025
[15]

J. Wang, S. Paesani, Y. Ding, R. Santagati, P. Skrzypczyk, A. Salavrakos, J. Tura, R. Augusiak, L. Mančinska, D. Bacco,et al., Multidimensional quantum entangle- ment with large-scale integrated optics, Science360, 285 (2018)

2018
[16]

E. H. Lieb and D. W. Robinson, The finite group velocity of quantum spin systems, Communications in Mathemat- 17 ical Physics28, 251 (1972)

1972
[17]

Kumar, Y

M. Kumar, Y. Tahir, V. Daiya, S. Mattaparthi, and A. Shaurya, Congestion-free routing on quantum chips (2026). Appendix A: Formal Guarantees for Routed Control Delivery This appendix collects the formal correctness and re- source statements for the routed single-control protocol and the basic parallel-routing abstraction

2026
[18]

, vL = v)of length L

Preliminaries Let G = (V, E)be a connected graph representing the hardware architecture, and letu, v∈V be two nodes con- nected by a pathP = (v0 = u, v1, . . . , vL = v)of length L. Let K denote the number of simultaneously addressable routing buses, with offsets∆ k = 2 k for k = 1 , . . . , K. The largest routing value produced by activating every bus wh...
[19]

Using the routing protocol defined in Sec

Correctness of the Routed Nonlocal CNOT Theorem 26(Swap-Free Non-Local CNOT).Let u, v∈ V be two nodes connected by a path of lengthL. Using the routing protocol defined in Sec. III, a logicalCX(u,v) operation can be implemented without SWAP gates, while restoring all intermediate nodes to their original states. Proof. Let the initial computational values ...
[20]

Complexity Bounds Proposition27(GateComplexity).The routing protocol implements a nonlocal CNOT over a path of lengthL using 2L+ 1primitive operations. Proof. The protocol consists of L forward propagation steps, one target interaction, andL reverse cleanup steps, yielding a total of2L+ 1operations. Proposition 28(Depth).The circuit depth of the pro- toco...
[21]

, K, and suppose that 18 d≥ 2K+1

Parallel Routing Capacity Theorem 29(Bus Parallelism).Let the available routing buses be∆ k = 2 k for k = 1 , . . . , K, and suppose that 18 d≥ 2K+1. Then up to K independent nonlocal routing operations can share nodes while remaining algebraically separable by their bus labels. In particular, overlapping routes assigned distinct buses do not suffer logic...
[22]

Comparison with SWAP-Based and Alternative Routing Proposition 31(Reduction in Leading Routing Con- stant).For a path of length L, the proposed method re- alizes the routed operation using2L + 1logical routing primitives rather than the transport-only SWAP baseline of depth3 L, while requiring no additional physical qudits. Proof. Naive SWAP routing requi...
[23]

Consider two nodes u, v∈V separated by graph distanceL

Lower Bound on Information Propagation Theorem 32(Light-Cone Lower Bound).Let G = (V, E)be a graph representing a quantum architecture with nearest-neighbor interactions. Consider two nodes u, v∈V separated by graph distanceL. Any protocol that implements a non-localCX(u,v) gate using only local interactions requires depth at leastL. Proof.We argue using ...
[24]

III achieves depthO(L)for imple- menting a non-localCX(u,v) gate

Optimality of the Proposed Protocol Theorem 33(Asymptotic Optimality).The routing pro- tocol defined in Sec. III achieves depthO(L)for imple- menting a non-localCX(u,v) gate. Therefore, it is asymp- totically optimal with respect to circuit depth. Proof. From Sec. III, the protocol consists ofL forward propagation steps and L reverse cleanup steps. These ...
[25]

the Lower Bound While the depth lower bound applies to a single non- local operation, the proposed protocol allows multiple operations to proceed simultaneously

Parallelism vs. the Lower Bound While the depth lower bound applies to a single non- local operation, the proposed protocol allows multiple operations to proceed simultaneously. Proposition 34(Parallel Multiplexing Advantage).Us- ing k distinct routing buses, the protocol enables up tok independent nonlocal operations to be multiplexed within a common log...
[26]

, Pm, define theroute conflict graphΓto be the graph with vertex set {1,

Conflict-Graph Separation from Qubit-Only Routing Definition 7(Route Conflict Graph).Given a requested family of routed operations with pathsP1, . . . , Pm, define theroute conflict graphΓto be the graph with vertex set {1, . . . , m}in which j and ℓ are adjacent iffPj ∩P ℓ ̸= ∅. Thus an edge ofΓrecords exactly the pairs of routed operations that cannot s...
[27]

Comparison with the SWAP Baseline Corollary 42.Standard SWAP-based routing is not op- timal at the logical-routing level considered here, because it incurs a larger leading transport-depth constant than the proposed protocol. Proof. Naive SWAP-based routing requiresL swaps to move a qubit across distanceL, with each swap decom- posing into 3 CNOT gates. T...
[28]

Detailed Fan-In Sanity Checks To sanity-check the algebraic extension, we imple- mented small instances of the routed Boolean fan-in gate in both an ideal state-vector model (Cirq) and a multi- level open-system model (QuTiP). These computations are not intended as a broad hardware-performance claim; rather, they test whether explicit routed circuits beha...
[29]

Detailed Ideal Cirq Benchmarks To validate the ideal guarantees of the routing pro- tocol on small instances, we simulated the architecture in Google Cirq using exact state-vector evolution. The verification suite implements the generalizedCBL, BCP , and routing-controlled target primitives as explicit qudit unitaries and records the resulting benchmarkin...
[30]

First, an exact combi- natorial validator exhaustively checked the congestion theorem on every labeled conflict graph up to five ver- tices

Conflict-Graph and Compiler-Demand Validation We also supplemented the device-level simulations with two independent validation scripts aimed at the manuscript’s congestion claims. First, an exact combi- natorial validator exhaustively checked the congestion theorem on every labeled conflict graph up to five ver- tices. Across all1099graphs and for K = 1 ...
[31]

IVI, we simulated open- system dynamics in QuTiP using the Lindblad master equation

Detailed Noisy QuTiP Benchmarks To probe the physical limits of the routing idea beyond the ideal Cirq checks of Sec. IVI, we simulated open- system dynamics in QuTiP using the Lindblad master equation. The purpose of these noisy simulations is not to demonstrate immediate superiority over SWAP-based routing, but to identify the hardware regime in which t...