Magic Square Game in Quantum Information

• Magic Square Game is a cooperative nonlocal game defined on a 3×3 grid that uses parity constraints to expose the limitations of classical strategies and demonstrate the power of quantum entanglement.
• It leverages entangled measurements with EPR pairs to achieve a perfect winning probability, embodying quantum pseudo-telepathy and robust device-independent self-testing.
• Experimental implementations include fault-tolerant protocols and noise mitigation techniques that bolster advances in quantum cryptography and large-scale entanglement certification.

The Magic Square Game (MSG) is a cooperative two-player nonlocal game that encapsulates quantum pseudo-telepathy and device-independent quantum self-testing. In its canonical formulation, a referee privately assigns a row index to Alice and a column index to Bob on a $3 \times 3$ grid. Each player, without communication, must fill their row/column with bits that satisfy parity constraints—row parities even for Alice, column parities odd for Bob—with the crucial requirement that their answers coincide at the unique intersection cell. While classical strategies are limited to an $8/9$ winning probability, quantum entanglement allows perfect success, exemplifying the separation between quantum and classical correlations. The MSG is central to modern quantum information, with foundational roles in nonlocal games, self-testing, quantum cryptography, and experimental entanglement verification.

1. Formal Definition and Parity Constraints

The original MSG involves a $3 \times 3$ array where Alice and Bob are assigned a row $r \in \{0,1,2\}$ or column $c \in \{0,1,2\}$, respectively. Each must output three bits (classically, $\pm 1$ or $0/1$ for each cell):

• Alice (row $r$): $(a_{r,0}, a_{r,1}, a_{r,2})$ with constraint $\prod_{j=0}^2 a_{r,j} = +1$ (even parity).
• Bob (column $c$): $(b_{0,c}, b_{1,c}, b_{2,c})$ with constraint $\prod_{i=0}^2 b_{i,c} = -1$ (odd parity).
• Consistency constraint: $a_{r,c} = b_{r,c}$ for the intersection cell.

There does not exist a classical assignment of bits fulfilling all row/column constraints simultaneously. The best classical protocol reaches a maximal success probability of $8/9$ by covering all but one input pair (Zhen et al., 2023, Bugu et al., 2020, Pawela et al., 2012). Quantumly, in the noise-free limit, unit probability is achievable.

2. Quantum Strategy and Algebraic Realization

The quantum winning strategy uses entanglement and appropriate local measurements. Canonically:

• Resource state: Alice and Bob share two EPR pairs, $\ket{\Psi^+} = (|00\rangle + |11\rangle)/\sqrt{2}$, or equivalently the four-qubit “magic square” state:

$$|\phi\rangle = \frac{1}{2} \Big(|0011\rangle + |1100\rangle - |0110\rangle - |1001\rangle\Big)$$

(Bugu et al., 2020, Pawela et al., 2012, Zhen et al., 2023).

• Measurement observables: For each possible row or column, the players implement commuting Pauli product observables. E.g., Alice’s row measurements for $x = 0$ are $\{Z \otimes I, I \otimes Z, Z \otimes Z\}$, with product $+I$; Bob’s column measurements analogously, but with product $-I$ (Zhen et al., 2023).
• Output bits: Measurement results are mapped to bits such that the parity requirements are automatically satisfied, and matching at the intersection is enforced by the entangled state (Arkhipov, 2012).

This algebraic structure ensures perfect winning (i.e., probability 1) for all $9$ input pairs, establishing quantum pseudo-telepathy.

3. Generalizations and Rigidity

MSG generalizes to the “Parity Telepathy Game” on arbitrary hypergraph arrangements, characterized by parity constraints over intersecting sets. The central result is that quantum pseudo-telepathy is present if and only if the intersection graph is nonplanar, i.e., contains $K_{3,3}$ or $K_5$ as a minor. This is a necessary and sufficient criterion for the “magic” property (Arkhipov, 2012). For all such games, three Bell pairs and measurements from the three-qubit Pauli group suffice to achieve certainty.

Rigidity phenomena are crucial: robust self-testing guarantees that any quantum strategy winning the MSG with probability $1-\varepsilon$ must be close to the ideal strategy up to local isometry and operator approximation. Parallel repetition preserves rigidity, certifying $2n$ EPR pairs and the full $n$-qubit Pauli group with error scaling polynomially in $n\varepsilon$ (Coudron et al., 2016). This underpins device-independent certification of large-scale entanglement.

4. Fault-Tolerant Realization and Experimental Implementations

Physical imperfections—noise, gate errors, and decoherence—inevitably degrade the unit probability. Variational quantum algorithms have been crafted to optimize parameterized quantum circuits that encode the MSG parity and consistency constraints into a value Hamiltonian minimized via hybrid solvers; e.g.,

$$H_{\mathrm{value}} = \sum_{r=0}^2 \frac{1}{2}(I - A_r) + \sum_{c=0}^2 \frac{1}{2}(I + B_c) + \sum_{r=0}^2\sum_{c=0}^2 \frac{1}{2}(I - A_r \otimes B_c)$$

with optimal strategies reconstructing the canonical measurement assignments (Chehade et al., 19 May 2025). Physical proposals using distant quantum dots in optical cavities exploit spin-photon CZ gates realized by single-photon ancillae, achieving quantum advantage under realistic experimental parameters ($g \gg \sqrt{\kappa\gamma}$, $\theta \approx 0$) (Bugu et al., 2020).

Fault-tolerant MSG protocols encode qubits into concatenated Steane codes and apply a logical entanglement purification protocol, with significant Bell-pair savings and enhanced error thresholds ($p_{th}=4.70\times 10^{-4}$, $F_{max}=18.3\%$) (Liu et al., 17 Mar 2025). These methods delineate necessary conditions on gate and resource fidelity for achieving near-perfect nonlocality in real devices.

5. Noise Robustness and Recovery Techniques

Noise models (depolarizing, amplitude/phase damping, bit/phase/bit-phase flip) impact the MSG quantum-classical gap. Without recovery, quantum advantage vanishes at modest noise levels (e.g., $\alpha \sim 0.05$ for depolarizing). Semidefinite programming (SDP) enables construction of local reversing channels that substantially extend the region of pseudo-telepathy, restoring quantum advantage even at high noise $\alpha$ for flip channels and up to $\alpha \sim 0.30$ for depolarizing (Pawela et al., 2012). This technique is robust for uncorrelated noise and is limited primarily by channel separability and knowledge of noise parameters.

6. MSG in Quantum Cryptography and Verification Protocols

MSG has been embedded effectively into device-independent quantum key distribution (DIQKD) schemes. The protocol leverages the game’s nonlocality to certify shared secret key bits, with the quantum-classical gap yielding increased key rates relative to those based on CHSH, provided state visibility and detector efficiency thresholds ($\nu > 0.959$, $\eta > 0.969$) are met (Zhen et al., 2023). This construction exploits the full self-testing property of the MSG, with achievable key rates $R \to \gamma$ per round as opposed to $R_{CHSH} \leq \gamma/2$ for conventional protocols.

Applications extend to randomness expansion, multi-prover interactive proofs, and robust benchmarks for modular quantum architectures and networks (Coudron et al., 2016, Liu et al., 17 Mar 2025).

7. Glued Magic Square Game and Convex Self-Testing

Composite games formed by “gluing” two MSGs along parity constraints, termed Glued Magic Square games (GMS), admit multiple inequivalent perfect quantum strategies. This necessitates a convex self-testing framework: every perfect GMS strategy decomposes into a convex combination of two substrategies, each a locally dilated canonical MSG strategy. Robust convex self-testing extends, with any strategy succeeding with probability $1-\varepsilon$ lying $O(\sqrt{\varepsilon})$-close, in fidelity, to a convex combination of reference strategies (Mančinska et al., 2021). These ideas generalize to glued pentagram and mixed games, with self-tested resource states reflecting the intersection of base game Schmidt ranks.

Table: Classical vs. Quantum Bound and Key Features

Feature	Classical MSG	Quantum MSG
Winning probability	$8/9$ (max)	$1$ (perfect)
Shared resource	Shared randomness	Two EPR pairs (qubits)
Parity constraints	At most 8/9 satisfied	All 9 satisfied
Robust to noise	N/A	With error mitigation SDP/logical EPP
Device-independent test	No	Yes

Classical bounds for MSG are strictly limited by parity constraints, with only $8/9$ maximal achievable success. Quantum strategies leverage entanglement to fulfill all parity and consistency conditions, rendering the game an archetypal nonlocality witness.

---

The MSG and its generalizations serve as testbeds for quantum nonlocality, entanglement certification, fault-tolerant protocol design, cryptographic primitives, and experimental demonstrations on scalable quantum platforms. Key technical advances include algebraic realization, self-testing rigidity, noise-robust recovery, convex self-testing in glued games, and physically-motivated variational protocols, consolidating the MSG’s role as a foundational object in quantum information science.