Choosing the next link can help quantum networks connect sooner

A quantum network can have many links without yet offering a connected route between distant nodes. How should it choose which imperfect link to try to strengthen next? Yanxuan Shao and Adilson E. Motter model a way to make that choice: compare two candidate links and favor the one that would join larger connected groups. Their simulations indicate that this “power-of-choice” strategy can create a large connected cluster earlier than choosing links at random.

The setting is a network whose links each hold a partially entangled pair of qubits. The paper assumes these pairs are pure, identical, and unaffected by decoherence during the process. A local operation together with classical communication can attempt to convert a pair into a maximally entangled Bell pair, suitable as a stronger network link. In the model, each attempt succeeds with probability p; failure loses the entanglement in that link. Thus, the network grows through successful conversions, but unsuccessful attempts still consume opportunities and are counted in the resource measure.

The baseline is simple: pick an unconverted link uniformly at random and attempt conversion. In the power-of-choice (PoC) protocol, pick two distinct unconverted links at each step. If a candidate already connects nodes within the same Bell-connected cluster, treat it as redundant and choose the other; if both are redundant, choose randomly. Otherwise, compare the total sizes of the two clusters each link would connect, and attempt the link with the larger combined size. Ties are broken at random. The paper also studies selecting among more than two candidates; more choices move the transition earlier, with diminishing additional benefit.

The authors simulate networks and analyze how clusters evolve using rate equations. In the Erdős–Rényi (ER) examples, links are initially placed randomly, and each node has an average of eight possible connections in a representative case. For N = 1,000,000 nodes and conversion probability p = 0.75, the reported curves are averages over ten simulated network realizations. Compared with random selection, PoC shifts the onset of a giant connected component to an earlier point in the sequence of conversion attempts and reaches the same later cluster sizes with fewer attempts. The advantage is about the selection of attempts—not a higher probability that any individual conversion succeeds.

The transition remains continuous in the authors’ analysis: the giant cluster grows from zero rather than appearing at a finite size abruptly. At the transition, their calculations and simulations support a cluster-size distribution with a power-law exponent of 2.5. They also test a local power-of-choice (LPoC) variant. Instead of comparing whole-cluster sizes, it compares the sums of the endpoint nodes’ degrees in the already converted Bell-link network, favoring the candidate with the larger sum. In the tested networks, this local rule produces a percolation curve very close to the global PoC rule. This is useful because tracking degree is more local than maintaining cluster-size information across the network.

The pattern is not confined to ER networks in the reported simulations: the authors also examine regular networks, scale-free networks and core–periphery structures. Across these models, PoC and LPoC tend to accelerate growth of the giant component relative to random selection. A parallel PoC variant reaches the transition in only a few time steps because as many as N conversions can be attempted in one step; the paper cautions that the total number of conversion operations is generally not smaller than for sequential PoC and can be larger under the same parameters. In ER networks, the minimum average degree needed for percolation is 1/p for all three strategies; choice changes how quickly the transition is reached, not that minimum condition.

These are theoretical and numerical results, not an experimental demonstration. The idealized starting links omit decoherence and assume identical pure states; practical networks may have noisy, unequal links and communication or memory constraints. Even the local version needs coordination, record-keeping and, after clusters form, some information propagation. The simulations are finite, and the sharp transition is a large-network concept estimated from finite systems. The work therefore suggests a promising link-selection rule, not a ready-made guarantee of lower resource use in deployed quantum networks.