Skip to content
Quantum.world

The Signal

ParityQC Turns Rydberg Embedding Overhead Into Error Correction

A Rydberg atom-array embedding can provide local error syndromes that help correct errors without adding hardware. On QuEra’s Aquila QPU, a Cluster decoder recovered optimization solutions for a 107-atom layout when raw data and standard greedy processing found none.

ParityQC Turns Rydberg Embedding Overhead Into Error Correction

For a 107-atom layout on the QuEra Aquila QPU, a Cluster decoder recovered exact optimization solutions with an average success rate of at least around 8 percent, even when raw experimental data and standard GreedyMIS processing found none. The result comes from an approach that treats the extra atoms used to embed an optimization problem as a resource for error correction, rather than only as additional physical overhead.

Source parityqc.com

From embedding overhead to a code

Rydberg atom arrays naturally encode the maximum independent set problem on geometric graphs. Solving general combinatorial optimization problems on more complex topologies requires an embedding to work around the hardware’s limited-range native interactions. The gadget arrangements used for these embeddings introduce ancillary atoms, increasing the physical system size and its exposure to physical errors.

In the ParityQC Architecture-based Rydberg embedding, the atoms are grouped into superatoms. The Rydberg blockade restricts each group to at most one excitation, while valid logical states require exactly one. Counting excitations locally can therefore flag a superatom with zero or multiple excitations as an error. The check extracts a local error syndrome without additional hardware.

The researchers call the resulting classical error-correcting code the Rydberg-Parity code. Its checks rely on counting excitations, rather than parity, making it strictly nonlinear. The code’s properties—including its qubit overhead, encoding rate and code distance—depend on the topology of the original optimization problem.

A decoder that grows unresolved clusters

Because standard decoders for linear codes do not apply to these nonlinear syndromes, the researchers developed the Cluster decoder for the Rydberg gadget structure. It forms clusters around detected syndromes and uses predefined lookup tables for the gadgets to find a unique valid configuration for each cluster.

When the decoder cannot identify a unique correction, it expands the affected cluster to neighboring superatom sets and merges overlapping clusters. This process repeats until the clusters are resolved or an unresolved cluster spans the graph. Each cluster that is resolved is guaranteed to be a codeword of the embedded graph. Numerical tests found an error-correction threshold of 26 percent bit-flip probability across different layout geometries.

Results on experimental hardware

The hardware benchmark used experimental data from the QuEra Aquila QPU for QUBO problems on complete and bipartite graphs, with layouts reaching 107 atoms. Both the Cluster decoder and standard GreedyMIS processing improved success rates relative to raw experimental data. The Cluster decoder’s success rate fell more slowly as problem size increased. On the largest layout, it recovered target solutions at an average success rate of at least around 8 percent, while the raw data and GreedyMIS-processed data found none.

The approach is not limited to the ParityQC Architecture-based embedding. The researchers also applied the same superatom syndrome and Cluster decoder to other blockade-based Rydberg gadget layouts, including crossing lattices and 3D quantum wires, which showed similar threshold behavior. The decoder can also apply to layouts made from a finite set of bounded-size gadgets when their admissible configurations are known and propagate deterministically from one gadget to the next.

The work also applies the Cluster decoder to results from a globally driven, ParityQC Architecture-based neutral-atom array, where it improved the target-state probability. Together, the results describe an architecture in which the embedding, computation and error correction arise from the same physical structure.

Source

  1. ParityQC

Related stories

Explore

More articles