Mayers–Yao Local Isometry

• Mayers–Yao local isometry is a framework that certifies quantum correlations by mapping unknown physical states to ideal entangled subsystems through local isometries.
• It employs a swap-isometry construction to extract a certified qubit or qudit pair from observed input–output statistics, ensuring matched operator relations.
• The method extends to qudit systems via Heisenberg–Weyl operators, with robust error bounds enabling device-independent self-testing across various dimensions.

Searching arXiv for the specified papers to ground the article and confirm citation details. Mayers–Yao local isometry denotes the local-isometric equivalence used in device-independent self-testing: from observed input–output statistics alone, one concludes that there exist local isometries which map an unknown physical realization to an ideal entangled reference subsystem tensored with an arbitrary normalized “junk” state, and which map the physical measurements to the corresponding target observables on that subsystem. In the qubit setting, this framework certifies a singlet and Pauli measurements from Mayers–Yao or CHSH-type data (McKague et al., 2012). In the qudit setting, it is extended to maximally entangled states in every finite dimension by reconstructing Heisenberg–Weyl structure and applying a qudit SWAP isometry (Meyer et al., 1 Aug 2025).

1. Formal notion of local-isometric certification

In self-testing, no assumption is made about the internal Hilbert spaces or implementations of the devices; only the observed correlations are trusted. The basic conclusion is not literal equality between the physical state and a fixed reference state, but equivalence up to local isometries. For qubits, the robust formulation uses local isometries

$$\Phi_A : \mathcal{H}_A \to \mathcal{H}_A' \otimes \mathbb{C}^2, \qquad \Phi_B : \mathcal{H}_B \to \mathcal{H}_B' \otimes \mathbb{C}^2,$$

such that

$$(\Phi_A \otimes \Phi_B)\lvert\psi'_{AB}\rangle = \lvert \mathrm{junk}_{AB}\rangle \otimes \lvert \phi_+\rangle,$$

and, for the relevant observables,

$$(\Phi_A \otimes \Phi_B)\big(M'_A N'_B \lvert\psi'_{AB}\rangle\big) = \lvert \mathrm{junk}_{AB}\rangle \otimes M_A N_B \lvert \phi_+\rangle.$$

Here $$\lvert \phi_+\rangle = \frac{1}{\sqrt{2}}(\lvert 00\rangle+\lvert 11\rangle)$$, and the junk system absorbs all degrees of freedom irrelevant to the certified qubit pair (McKague et al., 2012).

The same structure appears in the qudit generalization. For odd prime $d$, the main self-testing theorem asserts the existence of local unitaries

$$V_A : \mathcal{H}_A \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{A'}, \qquad V_B : \mathcal{H}_B \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{B'},$$

such that the unknown state is mapped close to a maximally entangled $d$-dimensional reference state tensored with an auxiliary state, and the measurement operators are mapped to canonical Heisenberg–Weyl observables tensored with identity on the auxiliary spaces (Meyer et al., 1 Aug 2025).

An isometry preserves inner products. Operationally, “add an ancilla and apply a unitary” is an isometry. The use of an isometry rather than a unitary is essential because the physical Hilbert space need not have the same dimension as the extracted reference subsystem. This is the mathematical form of the claim that a device may contain extra degrees of freedom, a different local basis, or an encoding of the target subsystem inside a larger space.

2. The qubit Mayers–Yao setting

In the qubit Mayers–Yao test, Alice has observables $X'_A, Z'_A$, while Bob has $X'_B, Z'_B, D'_B$, all with eigenvalues $\pm 1$. The ideal target configuration consists of the Bell state $$(\Phi_A \otimes \Phi_B)\lvert\psi'_{AB}\rangle = \lvert \mathrm{junk}_{AB}\rangle \otimes \lvert \phi_+\rangle,$$0, Pauli $$(\Phi_A \otimes \Phi_B)\lvert\psi'_{AB}\rangle = \lvert \mathrm{junk}_{AB}\rangle \otimes \lvert \phi_+\rangle,$$1 and $$(\Phi_A \otimes \Phi_B)\lvert\psi'_{AB}\rangle = \lvert \mathrm{junk}_{AB}\rangle \otimes \lvert \phi_+\rangle,$$2 on both sides, and the additional observable

$$(\Phi_A \otimes \Phi_B)\lvert\psi'_{AB}\rangle = \lvert \mathrm{junk}_{AB}\rangle \otimes \lvert \phi_+\rangle,$$3

on Bob’s side (McKague et al., 2012).

The Mayers–Yao correlations are encoded by the requirement

$$(\Phi_A \otimes \Phi_B)\lvert\psi'_{AB}\rangle = \lvert \mathrm{junk}_{AB}\rangle \otimes \lvert \phi_+\rangle,$$4

for all $$(\Phi_A \otimes \Phi_B)\lvert\psi'_{AB}\rangle = \lvert \mathrm{junk}_{AB}\rangle \otimes \lvert \phi_+\rangle,$$5 and $$(\Phi_A \otimes \Phi_B)\lvert\psi'_{AB}\rangle = \lvert \mathrm{junk}_{AB}\rangle \otimes \lvert \phi_+\rangle,$$6. In the ideal case, these reproduce the six expectation values of the target qubit model. In particular,

$$(\Phi_A \otimes \Phi_B)\lvert\psi'_{AB}\rangle = \lvert \mathrm{junk}_{AB}\rangle \otimes \lvert \phi_+\rangle,$$7

and

$$(\Phi_A \otimes \Phi_B)\lvert\psi'_{AB}\rangle = \lvert \mathrm{junk}_{AB}\rangle \otimes \lvert \phi_+\rangle,$$8

The logical role of these correlations is twofold. First, they force $$(\Phi_A \otimes \Phi_B)\lvert\psi'_{AB}\rangle = \lvert \mathrm{junk}_{AB}\rangle \otimes \lvert \phi_+\rangle,$$9 and $$(\Phi_A \otimes \Phi_B)\big(M'_A N'_B \lvert\psi'_{AB}\rangle\big) = \lvert \mathrm{junk}_{AB}\rangle \otimes M_A N_B \lvert \phi_+\rangle.$$0 to agree with $$(\Phi_A \otimes \Phi_B)\big(M'_A N'_B \lvert\psi'_{AB}\rangle\big) = \lvert \mathrm{junk}_{AB}\rangle \otimes M_A N_B \lvert \phi_+\rangle.$$1 and $$(\Phi_A \otimes \Phi_B)\big(M'_A N'_B \lvert\psi'_{AB}\rangle\big) = \lvert \mathrm{junk}_{AB}\rangle \otimes M_A N_B \lvert \phi_+\rangle.$$2 on the support of the state: $$(\Phi_A \otimes \Phi_B)\big(M'_A N'_B \lvert\psi'_{AB}\rangle\big) = \lvert \mathrm{junk}_{AB}\rangle \otimes M_A N_B \lvert \phi_+\rangle.$$3 Second, the observable $$(\Phi_A \otimes \Phi_B)\big(M'_A N'_B \lvert\psi'_{AB}\rangle\big) = \lvert \mathrm{junk}_{AB}\rangle \otimes M_A N_B \lvert \phi_+\rangle.$$4, which behaves like a $$(\Phi_A \otimes \Phi_B)\big(M'_A N'_B \lvert\psi'_{AB}\rangle\big) = \lvert \mathrm{junk}_{AB}\rangle \otimes M_A N_B \lvert \phi_+\rangle.$$5 measurement between $$(\Phi_A \otimes \Phi_B)\big(M'_A N'_B \lvert\psi'_{AB}\rangle\big) = \lvert \mathrm{junk}_{AB}\rangle \otimes M_A N_B \lvert \phi_+\rangle.$$6 and $$(\Phi_A \otimes \Phi_B)\big(M'_A N'_B \lvert\psi'_{AB}\rangle\big) = \lvert \mathrm{junk}_{AB}\rangle \otimes M_A N_B \lvert \phi_+\rangle.$$7, is used to force approximate anti-commutation: $$(\Phi_A \otimes \Phi_B)\big(M'_A N'_B \lvert\psi'_{AB}\rangle\big) = \lvert \mathrm{junk}_{AB}\rangle \otimes M_A N_B \lvert \phi_+\rangle.$$8 and similarly on Bob’s side. Once approximate anti-commutation and approximate equality are available, the local isometry can extract an effective qubit pair (McKague et al., 2012).

A common misconception is that the Mayers–Yao test is simply a reformulation of CHSH. The two are closely related in the robust framework, but they are not identical. CHSH uses near-maximal Bell inequality violation; Mayers–Yao uses a specified pattern of correlations involving the additional diagonal observable $$(\Phi_A \otimes \Phi_B)\big(M'_A N'_B \lvert\psi'_{AB}\rangle\big) = \lvert \mathrm{junk}_{AB}\rangle \otimes M_A N_B \lvert \phi_+\rangle.$$9.

3. Explicit SWAP construction in the qubit case

The Mayers–Yao local isometry is realized as a swap-type circuit. Each party appends a fresh ancilla qubit in state $$\lvert \phi_+\rangle = \frac{1}{\sqrt{2}}(\lvert 00\rangle+\lvert 11\rangle)$$0, applies a Hadamard $$\lvert \phi_+\rangle = \frac{1}{\sqrt{2}}(\lvert 00\rangle+\lvert 11\rangle)$$1 to the ancilla, applies controlled operations using the physical observables, applies another Hadamard, and then applies controlled operations with the other Pauli-type observable. Formally, for party $$\lvert \phi_+\rangle = \frac{1}{\sqrt{2}}(\lvert 00\rangle+\lvert 11\rangle)$$2,

$$\lvert \phi_+\rangle = \frac{1}{\sqrt{2}}(\lvert 00\rangle+\lvert 11\rangle)$$3

with ancilla as control (McKague et al., 2012).

The joint action on the physical state admits an explicit expansion: $$\lvert \phi_+\rangle = \frac{1}{\sqrt{2}}(\lvert 00\rangle+\lvert 11\rangle)$$4 When the operator relations hold exactly, the second and third terms vanish, and the fourth is matched to the first in such a way that

$$\lvert \phi_+\rangle = \frac{1}{\sqrt{2}}(\lvert 00\rangle+\lvert 11\rangle)$$5

Thus the ancilla qubits become the certified EPR pair, while the original physical degrees of freedom are relegated to junk.

The same calculation extends to observables. For $$\lvert \phi_+\rangle = \frac{1}{\sqrt{2}}(\lvert 00\rangle+\lvert 11\rangle)$$6, the image of $$\lvert \phi_+\rangle = \frac{1}{\sqrt{2}}(\lvert 00\rangle+\lvert 11\rangle)$$7 under the isometry is close to

$$\lvert \phi_+\rangle = \frac{1}{\sqrt{2}}(\lvert 00\rangle+\lvert 11\rangle)$$8

which is the formal sense in which the physical measurements become Pauli observables on the extracted qubits (McKague et al., 2012).

4. Robust self-testing and operator-level criteria

The qubit theory separates the local isometry from the derivation of the operator relations. The abstract robust theorem assumes Hermitian unitaries $$\lvert \phi_+\rangle = \frac{1}{\sqrt{2}}(\lvert 00\rangle+\lvert 11\rangle)$$9 satisfying

$d$0

$d$1

$d$2

Under these conditions there exists a local isometry $d$3 and a state $d$4 such that

$d$5

for all $d$6, with

$d$7

(McKague et al., 2012).

For the Mayers–Yao correlations, the same work derives explicit operator errors. In particular,

$d$8

and the resulting global robustness function is $d$9. For CHSH, the same framework applies after constructing effective $$V_A : \mathcal{H}_A \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{A'}, \qquad V_B : \mathcal{H}_B \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{B'},$$0 and $$V_A : \mathcal{H}_A \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{A'}, \qquad V_B : \mathcal{H}_B \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{B'},$$1 from linear combinations of Bob’s observables, so CHSH and Mayers–Yao become two routes to the same swap-isometry theorem.

This separation between operator rigidity and isometric extraction is one of the lasting features of the Mayers–Yao paradigm. The isometry is fixed and explicit; the nontrivial analytic work lies in deriving the required algebraic relations from observed statistics.

5. Heisenberg–Weyl extension to qudits

The qudit generalization replaces Pauli anti-commutation by Heisenberg–Weyl commutation. For odd prime $$V_A : \mathcal{H}_A \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{A'}, \qquad V_B : \mathcal{H}_B \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{B'},$$2, the basic operators are

$$V_A : \mathcal{H}_A \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{A'}, \qquad V_B : \mathcal{H}_B \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{B'},$$3

with $$V_A : \mathcal{H}_A \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{A'}, \qquad V_B : \mathcal{H}_B \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{B'},$$4. The Bell scenario has $$V_A : \mathcal{H}_A \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{A'}, \qquad V_B : \mathcal{H}_B \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{B'},$$5 measurement settings and $$V_A : \mathcal{H}_A \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{A'}, \qquad V_B : \mathcal{H}_B \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{B'},$$6 outcomes per party, and the Bell operator $$V_A : \mathcal{H}_A \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{A'}, \qquad V_B : \mathcal{H}_B \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{B'},$$7 is constructed from the powers $$V_A : \mathcal{H}_A \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{A'}, \qquad V_B : \mathcal{H}_B \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{B'},$$8 and $$V_A : \mathcal{H}_A \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{A'}, \qquad V_B : \mathcal{H}_B \rightarrow \mathbb{C}^d \otimes \mathcal{H}_{B'},$$9 together with coefficients determined by a phase polynomial $d$0. On the target maximally entangled state, the Bell operator achieves

$d$1

(Meyer et al., 1 Aug 2025).

The crucial analytic step is an exact sum-of-positive-operators decomposition,

$d$2

with

$d$3

At maximal violation, each positive term vanishes on $d$4, and this yields twisted Heisenberg–Weyl commutation relations: $d$5 In the robust case, approximate versions of these relations are proved, with errors of order $d$6.

The local isometry is then defined by a qudit SWAP circuit. Each party appends a fresh ancilla qudit in $d$7, applies the qudit Fourier transform $d$8, and uses controlled Heisenberg–Weyl operations. The effective generators are chosen from the physical observables as

$d$9

$X'_A, Z'_A$0

Because these satisfy the Heisenberg–Weyl relations on the support of the state, the SWAP calculation shows that the isometry maps the unknown realization to a rotated Bell state and the measurements to canonical Heisenberg–Weyl observables: $X'_A, Z'_A$1

The robust theorem states that if

$X'_A, Z'_A$2

then there exist local unitaries $X'_A, Z'_A$3 such that

$X'_A, Z'_A$4

with

$X'_A, Z'_A$5

and the implemented measurements are correspondingly close to the target observables. The stated scaling is $X'_A, Z'_A$6 (Meyer et al., 1 Aug 2025).

6. Interpretation, tensorization, and scope

The local isometry is not required to be physically implemented. It is a mathematical “virtual” change of picture showing that any realization producing the certified correlations is equivalent, up to local isometries, to the reference state and observables. This is central to device independence: only input–output statistics are trusted, while Hilbert-space dimension, encoding, and internal structure remain unrestricted.

In the qubit case, the algebraic content is approximate Pauli anti-commutation and agreement across the bipartition. In the qudit case, it is the Heisenberg–Weyl algebra. The same pattern recurs in both settings: derive operator relations from observed statistics, then feed them into a fixed swap-type isometry. This suggests a general extraction template in which rigidity of an operator algebra yields a canonical embedding into an ideal reference model.

The qudit construction is not confined to prime dimension. The tensor-factor argument states that any finite-dimensional Hilbert space decomposes uniquely into a tensor product of prime-power subsystems, so a robust self-test for each prime dimension yields a robust self-test for every composite dimension $X'_A, Z'_A$7. If

$X'_A, Z'_A$8

the system can be regarded as

$X'_A, Z'_A$9

and the full local isometry is built as the tensor product of the prime-block isometries. Because robustness bounds are stable under tensor products, with errors accumulating at most linearly in the number of blocks, the full system inherits an $X'_B, Z'_B, D'_B$0 robustness (Meyer et al., 1 Aug 2025).

A further point of scope concerns the target observables. In the qudit construction, the certified operations are standard Heisenberg–Weyl operators together with diagonal non-Clifford phase gates

$X'_B, Z'_B, D'_B$1

which are diagonal in the computational basis. The cited work states that this makes the protocol directly applicable to high-dimensional photonic and atomic platforms (Meyer et al., 1 Aug 2025).

Across both qubit and qudit settings, the Mayers–Yao local isometry is therefore best understood not as a single circuit tied to dimension $X'_B, Z'_B, D'_B$2, but as a device-independent equivalence principle: once the appropriate operator algebra has been certified on the support of the state, a local SWAP isometry extracts the ideal entangled subsystem and relegates all remaining degrees of freedom to junk.