Quantum Code Packs 130 Qubits into 192 Physical Bits

Quantum computing faces a scaling : current error-correcting codes require hundreds of physical qubits to encode just one logical qubit, making large-scale systems impractical. A new study introduces a quantum code that dramatically increases the density of logical qubits, encoding 130 logical qubits in only 192 physical qubits—a rate of 67.7%. This breakthrough uses the face-centered cubic (FCC) lattice, a structure known for its dense packing of spheres, to create a Calderbank-Shor-Steane (CSS) stabilizer code with uniform weight-12 checks. The code, verified computationally, offers a 24-fold higher encoding rate than the cubic 3D toric code, though at a lower minimum distance of 3, highlighting a trade-off between rate and error suppression.

The researchers constructed the code by placing physical qubits on the edges of an FCC lattice wrapped on a 3-torus with periodic boundaries. They defined Z-stabilizers acting on vertices and X-stabilizers acting on octahedral voids, each with a weight of 12, meaning every stabilizer interacts with exactly 12 qubits. Computational verification using Python scripts confirmed the CSS validity, with the product of check matrices yielding zero, and determined the code parameters: at lattice size L=4, there are 192 physical qubits, 130 logical qubits, and a minimum distance of 3. The rate approaches 2/3 asymptotically as the lattice size increases, with similar at L=6, where 434 logical qubits are encoded in 648 physical qubits.

Ology involved building sparse binary matrices for the stabilizers and performing Gaussian elimination over GF(2) to compute the number of logical qubits. The minimum distance was proven exactly by exhaustively checking all weight-1 and weight-2 vectors—18,336 combinations at L=4—and finding none in the kernel of the check matrices, while also constructing weight-3 non-stabilizer codewords. A minimum-weight perfect matching decoder adapted to the FCC geometry was used to simulate error correction, with Monte Carlo trials showing a 10-fold coding gain at a physical error rate of 0.001 and a 63-fold gain at 0.0005, achieving decode success rates up to 99.9%.

The high encoding rate originates from a structural surplus in the FCC lattice: it has 3L^3 edges but only L^3 independent stabilizer constraints, leaving 2L^3 + 2 logical degrees of freedom. This contrasts with the cubic toric code, which has a similar number of edges but nearly all consumed by constraints, resulting in a much lower rate. The FCC code's weight-12 checks may support single-shot error correction, though this requires further proof, and the logical qubits are conjectured to be associated with tetrahedral voids in the lattice. Physical implementation could leverage neutral-atom arrays, photonic networks, or superconducting qubits with multi-layer packaging.

Limitations of the code include its fixed minimum distance of 3, which prevents scalable error suppression as the system grows, unlike surface codes that achieve arbitrarily low logical error rates with increasing distance. The block logical error rate scales with the number of logical qubits, offering only constant suppression factors, such as a 10-fold gain at p=0.001. Open questions remain about increasing the distance while retaining high rate, explicitly constructing all 130 logical operators, and confirming single-shot correction capabilities. Despite these constraints, the FCC code provides a valuable complement for applications where many noisy logical qubits are preferable to a few clean ones, such as in variational algorithms or quantum simulation.