# Techniques: Quantum Error Suppression with Subgroup Stabilisation

## What the paper contributes (one paragraph)
The paper supplies state purification gadgets that suppress errors in quantum states (not just expectation values) before full fault tolerance is available. Given M noisy copies of an unknown state, the gadgets project onto a symmetric subspace defined by projectors forming a subgroup of order M, rather than the full symmetric group of order M factorial used by Barenco et al. Two implementations are given: the Cyclic Group Gadget (CGG), using the cyclic group, and the Generalised SWAP Gadget (GSG), using the group of order M built from log(M) copies of Z/2Z. Both cut implementation cost to O(log(M)) ancilla measurements and O(M log(M)) controlled-SWAP gates while matching the 1/M suppression of coherent and stochastic errors per short evolution step. For depolarised inputs they reach purification rate O(p squared) at the asymptotically optimal sampling cost O(1/p). The authors also discuss applying the gadgets inside verification protocols.

## Techniques offered
- **state purification by subgroup stabilisation (Cyclic Group Gadget, CGG)** - projects M noisy copies onto a rotation-invariant subspace using a linear combination of cyclic-group permutation projectors plus post-selection, returning higher-purity states at reduced circuit cost.
  - guarantee: statistical bounds; suppresses coherent and stochastic errors by a factor of 1/M per short step and purifies depolarised inputs to O(p squared) at asymptotically optimal O(1/p) sampling cost.
  - quote: "In this work, we propose a lighter implementation for suppressing errors by projecting $M$ redundant states to their rotation invariant subspace using a linear combination of projectors forming the cyclic group $C_{M}$, which we name ``Cyclic Group Gadget (CGG)"."
- **state purification by subgroup stabilisation (Generalised SWAP Gadget, GSG)** - projects M noisy copies onto a symmetric subspace associated with the group (Z/2Z) raised to log(M), via generalised SWAP-test circuits and post-selection, with the same cost and suppression as CGG.
  - guarantee: statistical bounds; same error-suppression rate and sampling cost as CGG, and optimally amplifies the state overlap in the post-selection probability.
  - quote: "We also provide another implementation projecting $M$ inputs to a symmetric subspace associated with the group $\left(\mathbb{Z}/2\mathbb{Z}\right)^{\log\left(M\right)}$ inspired by the quantum circuit in Chabaud et al.~\cite{chabaud2018optimal}, which we name ``Generalised SWAP Gadget (GSG)"."

## Where it could apply
- Target primitive(s): sampling problems; any near- or middle-term circuit run repeatedly over M redundant copies; complements expectation-value error mitigation for BQP tasks.
- Target application group(s): error_correction (error suppression and partial QEC), platforms_tooling.
- Code family / hardware assumptions: not a stabiliser code; pre-fault-tolerant gadget needing no logical encoding; suited to photonic devices using Hadamard interferometers and to trapped-ion-plus-photon architectures able to perform nonlocal controlled-SWAP and permutation operations.

## Caveats
The paper is a methods contribution: the gadgets are unconditional purification primitives rather than a certificate supplied to an external result, and the analysis assumes the gadget itself runs perfectly. The 1/M suppression holds under first-order approximation in evolution time for coherent and stochastic errors; the O(p squared) rate and O(1/p) sampling cost are for depolarised inputs with small p, and there is an optimal M beyond which purity degrades. Performance under general noise, and integration with quantum verification protocols, are flagged as future work rather than established results.