Researchers at Tsinghua University and Hefei National Laboratory have entangled 13 photonic qubits in a single microwave graph state, using a superconducting circuit that joins small entangled photon groups into larger ones. The team published its results in Nature Physics in October 2026. The demonstration matters because the fusion operation behind it is deterministic, programmable and nondestructive — three properties that conventional optical fusion methods lack.

The work targets a core scaling problem in photonic quantum computing. Large graph states — entangled qubit networks whose connections follow a mathematical graph — are the fuel for measurement-based quantum computing, quantum networks and quantum error correction. Building them in one uninterrupted step becomes impractical as the qubit count grows.

Why fusion has been the bottleneck

The standard workaround is called fusion: instead of creating one very large entangled state at once, researchers stitch together smaller resource states. Hongyi Zhang, co-senior author of the paper, explained the catch to Phys.org.

"A promising way to build them is 'fusion': rather than making one very large, entangled state in a single step, one connects smaller resource states together," Zhang said. "The difficulty is that conventional fusion methods are probabilistic, so many attempts fail and the resource overhead grows rapidly."

Conventional approaches rely on optical components such as mirrors and beam splitters. These do not always complete the fusion, so experiments need repeated attempts or extra hardware. Each failure wastes photons and the cost compounds as the target state grows.

The Chinese team asked a different question: whether a quantum non-demolition measurement could make fusion deterministic. "Our goal was to demonstrate a deterministic, programmable fusion operation and show that it can provide a practical route to assembling larger, reconfigurable photonic graph states," Zhang said.

How the superconducting circuit works

The mechanism runs in three stages. Superconducting circuits first generate microwave photons in small, entangled states — the building blocks. A quantum non-demolition detector then performs a parity measurement on selected photon pairs.

"A quantum non-demolition detector then performs a parity measurement on selected photon pairs, entangling them without destroying them and thereby connecting the two smaller graph states into a larger one," Zhang said. "Frequency tuning lets us choose which photons to fuse."

That last point is the programmable part. By tuning frequencies, the researchers select which photons to join, so the connectivity of the resulting graph — which qubit links to which — can be adjusted after the fact. Non-demolition measurement is the key enabler, because the detector reads out parity without absorbing the photons it entangles.

In plain terms: the circuit glues small entangled clusters together on demand, and the operator chooses where the glue goes.

What the 13-qubit demonstration showed

Using this scheme, the team demonstrated genuine multipartite entanglement across 13 photonic qubits. The term "genuine" carries weight here: the entanglement spanned the entire group of thirteen, not separate smaller clusters that merely sit side by side.

The researchers verified that the graph structure was reconfigurable, meaning the same hardware could produce different connection patterns between qubits. Size and connectivity both matter for the downstream applications, as Zhang noted.

"Our main contribution is a fusion operation that is deterministic, programmable and nondestructive," Zhang said. "More broadly, the work offers an architecture for scaling photonic graph states by connecting smaller, on-demand resource states."

He tied the result to three application areas: measurement-based quantum computing, quantum networking and potentially quantum error-correction schemes, where both the size and the connectivity of graph states determine what computations become possible.

The takeaway: a fusion step that succeeds on demand removes the main source of waste in photonic graph-state construction.

What comes next

The demonstration is a laboratory result, not a product roadmap, and the authors are explicit about the remaining engineering gaps. "Next, we will improve the device fidelity, photon-generation efficiency and detector performance," Zhang said. "We also plan to develop multidetector architectures that can perform more fusion operations and generate larger, higher-dimensional graph states."

Each of those three metrics — fidelity, generation efficiency, detector performance — sets a ceiling on how large a graph state the architecture can reach before errors overwhelm the result. Multidetector architectures would run several fusion operations in parallel, the practical route from thirteen qubits toward states large enough for fault-tolerant, measurement-based computation.

The study is published as Yan Li et al, "Deterministic and programmable fusion for the scalable generation of photonic graph states," Nature Physics (2026), DOI: 10.1038/s41567-026-03471-5.

If fidelity and detector performance improve along the planned path, deterministic fusion could become the standard assembly method for photonic graph states in future quantum processors and long-distance quantum communication networks.