Fault-tolerant embedding of quantum circuits on hardware architectures via swap gates

arXiv:2406.17044 · quant-ph · Submitted 2024-06-24 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Quantum Radio. Generated commentary on the latest quantum physics and condensed matter papers.

Kai: We're looking at the paper: "Fault-tolerant embedding of quantum circuits on hardware architectures via swap gates".

Paper summary: Kai: So, looking at the paper "Fault-tolerant embedding of quantum circuits on hardware architectures via swap gates," the main point is this simple strategy for designing swap schemes to embed abstract circuits onto physical hardware with limited connectivity while keeping fault tolerance intact.

Mira: It claims that by restricting allowed swap moves and using something called an error-pattern-preserving property, you can ensure the embedded circuit’s errors still resemble those of the original circuit.

Lev: The results on the surface code show that this works, because they found a preservation of fault tolerance with a worsening of the threshold by a factor of three point six three in the heavy-hexagonal case.

Kai: So for someone just listening to this, it means you don't have to wait for perfect hardware with all those connections before you can start implementing circuits that need error correction.

Mira: It gives a concrete path forward for people who are working on abstract quantum circuits who want to map them onto real devices without having to fundamentally change the circuit itself.

Lev: It also motivates the need for better swap gates, which is a nice direction to look at for future hardware development.

Conclusion: Kai: So we’ve been talking about how they’re using swap gates to fit abstract quantum circuits onto real hardware that has limited connections, and now we're wrapping up with some thoughts on the big picture for this paper, "Fault-tolerant embedding of quantum circuits on hardware architectures via swap gates."

Mira: It really boils down to taking a circuit you design in a perfect world—say, something very complex—and figuring out how to physically run it on a chip where the qubits aren't all connected.

Lev: And they found this method using simple swap gates as the way to route information between those disconnected parts without totally ruining the error correction you built in.

Kai: So if I get it right, they’re not redesigning my abstract circuit to fit the hardware; they’re finding a smart way to map it onto what I already have, just with some noise introduced by these necessary routing swaps.

Mira: Exactly. The core idea is that even though the embedded circuit gets noisier because of those swaps, the error patterns that show up in the physical hardware still look like they come from your original, cleaner circuit.

Lev: That's what’s important for me. Because if the error patterns are preserved, then whatever fault tolerance you designed for your abstract code actually carries over to this noisy version on the actual hardware.

Kai: And they showed this works even on specific types of hardware like heavy-hexagonal and hexagonal lattices, showing that the noise doesn't explode too badly compared to native implementations.

Mira: They calculated that for the surface code example, you can actually get a worse threshold for your circuit by a factor of three point six three in the heavy-hexagonal case.

Lev: That’s a significant number because it shows us exactly how much noise we have to deal with when we try this approach. The main implication is that this gives us a concrete way to implement circuits with fault tolerance on current hardware without having to completely redesign the abstract circuit beforehand.

Kai: It’s a shortcut, basically, for getting fault tolerance onto real devices right now while we wait for better hardware architectures.

Mira: Right. It’s a practical tool that bridges the gap between theoretical circuit design and what can actually run on today's noisy quantum computers.

Lev: And it still points toward needing better native swap gates, because they showed that the error comes from how those swaps are done, which is where we need to focus next.

Shao-Hen Chiew, *Ezequiel Ignacio Rodríguez Chiacchio, Vishal Sharma, Jing Hao Chai, Hui Khoon Ng

Entropica Labs · Yale-NUS College Singapore · Centre for Quantum Technologies National University of Singapore

quant-ph

Submitted: 2024-06-24

Updated: 2024-06-24

Journal ref: Phys. Rev. Research 8, 033081 (2026)

DOI: 10.1103/crv2-99qd

License: http://creativecommons.org/licenses/by/4.0/

The gist: Fault-tolerant embedding of quantum circuits on hardware architectures via swap gates is presented as a simple strategy to design a swap scheme needed to embed an abstract circuit onto a physical…

Key concepts

Swap Gates
These are operations that permute qubits, moving quantum information between them. In this context, they are used not just for simple movement but as routing tools to move computation across physically disconnected parts of the hardware.
Error-Pattern-Preserving (EPP) Property
This property means that errors introduced by faulty swap gates in the physical layout can be simulated by equivalent or weaker errors occurring only within the computational operations. This ensures that the fault tolerance properties of the original circuit are inherited by the embedded, noisier circuit.
Surface Codes
Surface codes are a specific type of quantum error correction scheme used to protect quantum information. The authors tested their embedding technique on surface codes to demonstrate that even when mapped onto constrained hardware lattices, the logical error probability remains manageable.

Terminology

Summary

Fault-tolerant embedding of quantum circuits on hardware architectures via swap gates is presented as a simple strategy to design a swap scheme needed to embed an abstract circuit onto a physical hardware with constrained connectivity in a manner that preserves the fault-tolerant properties of the abstract circuit. p. 0

Where the paper says it

Here, we present a simple strategy to design the swap scheme needed to embed an abstract circuit onto a physical hardware with constrained connectivity, in a manner that preserves the fault-tolerant properties of the abstract circuit. The embedded circuit will, of course, be noisier, compared to a native implementation of the abstract circuit, but we show in the examples of embedding surface codes on heavy-hexagonal and hexagonal lattices that the deterioration is not severe.

The embedded circuit will, of course, be noisier, compared to a native implementation of the abstract circuit, but we show in the examples of embedding surface codes on heavy-hexagonal and hexagonal lattices that the deterioration is not severe. p. 0
Where the paper says it

Here, we present a simple strategy to design the swap scheme needed to embed an abstract circuit onto a physical hardware with constrained connectivity, in a manner that preserves the fault-tolerant properties of the abstract circuit. The embedded circuit will, of course, be noisier, compared to a native implementation of the abstract circuit, but we show in the examples of embedding surface codes on heavy-hexagonal and hexagonal lattices that the deterioration is not severe.

This then offers a straightforward solution to implementing circuits with fault-tolerance properties on current hardware. p. 0
Where the paper says it

The embedded circuit will, of course, be noisier, compared to a native implementation of the abstract circuit, but we show in the examples of embedding surface codes on heavy-hexagonal and hexagonal lattices that the deterioration is not severe. This then offers a straightforward solution to implementing circuits with fault-tolerance properties on current hardware.

Pruning redundant reasoning tokens

The core problem addressed is that in near-term quantum devices, connectivity between qubits remains limited by architectural constraints, necessitating routing procedures to move information between unconnected qubits. Ideal swap gates just permute the qubits; real swap gates, however, have the added possibilities of causing simultaneous errors on the qubits involved and spreading errors across the circuit. p. 0

Where the paper says it

The simplest routing strategy involves the use of swap gates to swap the information carried by two unconnected qubits to connected ones. Ideal swap gates just permute the qubits; real swap gates, however, have the added possibilities of causing simultaneous errors on the qubits involved and spreading errors across the circuit.

A general swap scheme thus changes the error-propagation properties of a circuit, including those necessary for fault-tolerant functioning of a circuit. p. 0
Where the paper says it

Ideal swap gates just permute the qubits; real swap gates, however, have the added possibilities of causing simultaneous errors on the qubits involved and spreading errors across the circuit. A general swap scheme thus changes the error-propagation properties of a circuit, including those necessary for fault-tolerant functioning of a circuit.

The authors show that by a simple restriction of the allowed swap-moves, they can look for swap-based routing schemes that embed arbitrary circuits onto physical hardware with constrained connectivity in a manner that preserves the fault-tolerant properties. p. 0
Where the paper says it

com † huikhoon. sg Nevertheless, we show here that, by a simple restriction of the allowed swap-moves, we can look for swap-based routing schemes that embed arbitrary circuits onto physical hardware with constrained connectivity in a manner that preserves the fault-tolerant properties.

Specifically, the error-patterns from faults in the embedded circuit resemble those of the original circuit; the former thus inherits any fault-tolerant properties—determined by the error-patterns that can be corrected—of the latter. p. 0
Where the paper says it

sg Nevertheless, we show here that, by a simple restriction of the allowed swap-moves, we can look for swap-based routing schemes that embed arbitrary circuits onto physical hardware with constrained connectivity in a manner that preserves the fault-tolerant properties. Specifically, the error-patterns from faults in the embedded circuit resemble those of the original circuit; the former thus inherits any fault-tolerant properties—determined by the error-patterns that can be corrected—of the latter.

How the training signal is built

The paper introduces two types of swap gates allowed in routing schedules: (1) Type-1 SWAP– a swap gate between a computational qubit and a routing qubit; (2) Type-2 SWAP– a swap gate between two computational qubits that participate in the same operation or circuit location. The main result is to show that an error-pattern in the embedded circuit arising from faulty SWAPs can be mimicked by having an equivalent- or lower-weight faultpath on the computational operations only, which is referred to as the error-pattern-preserving (EPP) property of a routing schedule. p. 1

Where the paper says it

We do not use swaps between two routing qubits—these add no useful moves to routing the computational information (though they can change the noise properties of the embedded circuit). Our main result is to show that an error-pattern in the embedded circuit arising from faulty SWAPs can be mimicked by having an equivalent- or lower-weight faultpath on the computational operations only.

Theorem 1 states that for every error-pattern that arises in the embedded circuit from a fault-path in the physical circuit, there exists an equal- or lower-weight fault-path in the noisy abstract circuit that can give the same errorpattern. p. 2
Where the paper says it

For every error-pattern that arises in the embedded circuit from a fault-path in the physical circuit, there exists an equal- or lower-weight fault-path in the noisy abstract circuit that can give the same errorpattern. Theorem 1 allows us to argue that our routing schedules preserve the fault-tolerance properties of an abstract circuit.

This allows them to argue that their routing schedules preserve the fault-tolerance properties of an abstract circuit. p. 2
Where the paper says it

For every error-pattern that arises in the embedded circuit from a fault-path in the physical circuit, there exists an equal- or lower-weight fault-path in the noisy abstract circuit that can give the same errorpattern. Theorem 1 allows us to argue that our routing schedules preserve the fault-tolerance properties of an abstract circuit.

Where the gains come from

The authors apply their results to embed the error correction circuit for the rotated planar surface code on hardware with heavy-hexagonal and hexagonal lattices. They show that for the heavy-hexagonal case, the logical error probability pL versus the effective error probability peff is rescaled to a best-fit value of peff = 3. p. 4

Where the paper says it

The results for the heavy-hexagonal case are given in Fig. 2, with pL plotted against the effective error probability peff, equal to p for the abstract circuit, but rescaled to the best-fit value of peff = 3.

63p for the embedded circuit. p. 4
Where the paper says it

2, with pL plotted against the effective error probability peff, equal to p for the abstract circuit, but rescaled to the best-fit value of peff = 3. 63p for the embedded circuit.

The gradient of the pL versus peff (or p) line tells us the number of faults the circuit is tolerant to, and a preservation of the fault-tolerant nature of a circuit demands precisely an unchanged gradient. p. 4
Where the paper says it

The horizontal rescaling is done to bring the pL curves for the the abstract and embedded circuits on top of each other for easier comparison. We see that, for each d, the gradients of the curves for the abstract and embedded circuits match very well, demonstrating the EPP nature of our routing schedule: The gradient of the pL versus peff (or p) line tells us the number of faults the circuit is tolerant to, and a preservation of the fault-tolerant nature of a circuit demands precisely an unchanged gradient.

This demonstrates that their EPP routing schedule preserves the fault-tolerant nature of the surface code error correction circuit, with a worsening of the threshold by a factor of 3. p. 4
Where the paper says it

Consequently, the lines for different d also intersect at the same point on the pL versus peff graph, for both abstract and embedded circuits. We thus conclude that our EPP routing schedule preserves the fault-tolerant nature of the surface code error correction circuit, with a worsening of the threshold by a factor of 3.

63.

The effective noise model for depolarizing noise on the swap gate is given by Eq(·) ≡ Xα pσα(·)σα (B1). For surface-code examples under depolarizing noise, they derive a simpler, albeit, approximate, description using a single effective noise parameter p′eff ≡ p + pswap 4/5 n1 + n2 (B3). p. 9

Where the paper says it

The exact effective noise model thus depends on the details of the full routing schedule, and can be very complicated in general. For our surface-code examples under depolarizing noise, however, we can derive a simpler, albeit, approximate, description.

For the heavy-hexagonal case, this yields peff = 3. p. 4
Where the paper says it

B 2) yields an approximate expression for peff, giving peff ≃ 3. 25p, respectively, for the heavy-hexagonal and hexagonal cases.

63p. In contrast, for the hexagonal case, they find peff = 1. p. 4
Where the paper says it

The results for the heavy-hexagonal case are given in Fig. 2, with pL plotted against the effective error probability peff, equal to p for the abstract circuit, but rescaled to the best-fit value of peff = 3.

25p. These values are close to the best-fit numerical peff values given in the main text. p. 9
Where the paper says it

From Table I, we see that the effective noise channel for each CNOT (or, more generally, a two-qubit gate) in the embedded circuit is an asymmetric Pauli channel, with coefficients for the different Pauli errors that depend on the details of the routing schedule; the asymmetry arises from type-1 SWAPs. Nevertheless, one might hope for an approximate summary of the behavior by a single effective noise parameter p ′ eff (where the prime on p ′ eff distinguishes it from the best-fit peff used in the main text and Apps B 5) using spacetime averages of the routing schedule parameters and ignoring the asymmetry.

The authors describe a search algorithm that yields constant-depth routing schedules for topological codes such as the surface code, with a computational cost that is independent of code distance. This algorithm avoids searching across the entire interaction- and device-graph lattices by embedding a unit cell of the interaction graph onto a unit cell of the device graph. p. 7

Where the paper says it

Since the interaction graph of the surface code SE circuits and the device graphs are both periodic, with invariant local structures as the code-distance increases, the core idea behind our approach is to embed a unit cell of the interaction graph onto a unit cell of the device graph, and subsequently search for the routing schedule for this unit. This avoids the need to search across the entire interaction- and device-graph lattices, which increase in size with code distance.

The resulting schedule can then be tiled in a repeated manner up to the desired code-distance. p. 7
Where the paper says it

This avoids the need to search across the entire interaction- and device-graph lattices, which increase in size with code distance. The resulting schedule can then be tiled in a repeated manner up to the desired code-distance, with tiles at the code boundary appropriately truncated to ensure that lower-weight stabilizers at the boundaries are measured correctly.

The authors conclude that their work relaxes hardware connectivity requirements for near-term fault-tolerant tasks. Their general applicability provides a concrete route to implementing arbitrary circuits on hardware in a fault-tolerant manner, without having to redesign the abstract circuit. p. 5

Where the paper says it

All in all, our work relaxes hardware connectivity requirements for near-term fault-tolerant tasks. Its general applicability provides a concrete route to implementing arbitrary circuits on hardware in a fault-tolerant manner, without having to redesign the abstract circuit.

The work also motivates the development of native, high-quality swap gates. p. 5
Where the paper says it

Its general applicability provides a concrete route to implementing arbitrary circuits on hardware in a fault-tolerant manner, without having to redesign the abstract circuit. Our work also motivates the development of native, high-quality swap gates, such that the information transport itself gives minimal increase in the overall noise.

p. 5
Where the paper says it

Its general applicability provides a concrete route to implementing arbitrary circuits on hardware in a fault-tolerant manner, without having to redesign the abstract circuit. Our work also motivates the development of native, high-quality swap gates, such that the information transport itself gives minimal increase in the overall noise.

The authors note that their current routing search algorithm constrains SWAPs and computational gates into separate circuit layers. However, they suggest that a more flexible circuit compilation program that allows for simultaneous circuit compression and EPP routing search can potentially generate more solutions, including ones with desirable features. p. 5

Where the paper says it

Interestingly, this 6-SWAP-layer solution introduces only temporally, but not spatially, correlated errors, a property that may be helpful in limiting the spread of errors. A more flexible circuit compilation program that allows for simultaneous circuit compression and EPP routing search can potentially generate more solutions, including ones with desirable features.

The work provides a concrete route to implementing arbitrary circuits on hardware in a fault-tolerant manner, without having to redesign the abstract circuit. p. 5
Where the paper says it

All in all, our work relaxes hardware connectivity requirements for near-term fault-tolerant tasks. Its general applicability provides a concrete route to implementing arbitrary circuits on hardware in a fault-tolerant manner, without having to redesign the abstract circuit.

The authors show that certain routing qubits in this schedule are involved in type-1 SWAPs with computational qubits that are not connected in the interaction graph, giving rise to the possibility of spatio-temporally correlated errors as mentioned in the main text. These correlated errors can be directly included as hyper-edges (i. p. 2

Where the paper says it

To show the EPP property, we first consider routing schedules comprising only type-1 SWAPs. Faulty type-1 SWAPs can introduce spatio-temporally correlated errors in the embedded circuit absent in the abstract circuit.

e., edges joining more than 2 nodes) in the matching graph, instead of decomposing them in terms of independent error mechanisms. p. 7
Where the paper says it

, UL, respectively. The interaction graph is denoted as GI ≡ (VI , EI ), where VI (EI ) is the set of nodes(edges) of the graph; similarly, the device graph is GD ≡ (VD, ED).

This improves the performance of decoding and thus lowers logical error probabilities and threshold requirements. p. 2
Where the paper says it

These > t-fault situations give the failure modes of the abstract or embedded circuit, and determine the resulting logical error rate. The noisier embedded circuit will thus have a larger logical error rate, which translates into a more stringent fault-tolerance noise threshold below which errors can be

The work relaxes hardware connectivity requirements for near-term fault-tolerant tasks. p. 5
Where the paper says it

A more flexible circuit compilation program that allows for simultaneous circuit compression and EPP routing search can potentially generate more solutions, including ones with desirable features. All in all, our work relaxes hardware connectivity requirements for near-term fault-tolerant tasks.

The general applicability provides a concrete route to implementing arbitrary circuits on hardware in a fault-tolerant manner, without having to redesign the abstract circuit. p. 5
Where the paper says it

All in all, our work relaxes hardware connectivity requirements for near-term fault-tolerant tasks. Its general applicability provides a concrete route to implementing arbitrary circuits on hardware in a fault-tolerant manner, without having to redesign the abstract circuit.

The work also motivates the development of native, high-quality swap gates.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Accurate circuit embedding for near-term quantum hardware can be achieved by designing swap schemes that preserve fault-tolerant properties of an abstract circuit, as shown by Theorem 1 (EPP routing schedules). This allows implementation of circuits with fault-tolerance properties on current hardware, even though the embedded circuit will be noisier.

  2. Circuit compilation and routing algorithms can be adapted to incorporate error-pattern preserving constraints by limiting allowed swap moves to types-1 and 2 SWAPs. This adaptation makes existing search algorithms easily applicable for embedding abstract circuits onto constrained hardware topologies.

  3. The system can generate effective noise models for embedded circuits using a single parameter p'eff derived from spacetime averages of routing schedule parameters, such as p'eff = p + pswap 4/5 n1 + n2. This provides a good summary of noise deterioration introduced by the routing schedule, even when the exact effective noise channel is asymmetric.

  4. Syndrome decoding for embedded circuits can be enhanced by incorporating error-correlation information into the matching graph, allowing for order-2s hyper-edges to represent correlated errors. This can lower logical error probabilities and improve threshold requirements during syndrome decoding via minimum-weight perfect matching algorithms.

  5. The system can utilize symmetry in topological codes to develop constant-depth routing schedules independent of code distance, avoiding the need to search across the entire lattices. This involves embedding a unit cell of the interaction graph onto a unit cell of the device graph and tiling the resulting schedule for desired code distances.

Abstract

In near-term quantum computing devices, connectivity between qubits remain limited by architectural constraints. A computational circuit with given connectivity requirements necessary for multi-qubit gates have to be embedded within physical hardware with fixed connectivity. Long-distance gates have to be done by first routing the relevant qubits together. The simplest routing strategy involves the use of swap gates to swap the information carried by two unconnected qubits to connected ones. Ideal swap gates just permute the qubits; real swap gates, however, have the added possibilities of causing simultaneous errors on the qubits involved and spreading errors across the circuit. A general swap scheme thus changes the error-propagation properties of a circuit, including those necessary for fault-tolerant functioning of a circuit. Here, we present a simple strategy to design the swap scheme needed to embed an abstract circuit onto a physical hardware with constrained connectivity, in a manner that preserves the fault-tolerant properties of the abstract circuit. The embedded circuit will, of course, be noisier, compared to a native implementation of the abstract circuit, but we show in the examples of embedding surface codes on heavy-hexagonal and hexagonal lattices that the deterioration is not severe. This then offers a straightforward solution to implementing circuits with fault-tolerance properties on current hardware.

Sources

Related papers