Programmable photonic state fusion via heralded storage of asynchronously generated resources

Technical Summary: Programmable Photonic State Fusion via Heralded Storage

Problem Statement
Probabilistic photonic sources generate elementary quantum states in discrete, asynchronous trials, whereas multiphoton protocols typically require these states to interfere within common temporal modes. While active switching and temporal multiplexing can collect events from different time bins, they often fail to place fused photons into identical temporal modes, leaving the resulting state with "time-bin labels" that encode generation history. For specific applications like clock interferometry, these labels must be removed to allow amplitudes associated with different generation histories to interfere coherently. Existing architectures, including loss-tolerant and fusion-based schemes, struggle to synchronize independently generated resources without retaining this temporal information or suffering from rapidly decreasing success probabilities.

Methodology
The authors propose a sequential fusion protocol that utilizes active storage loops to load independently heralded photonic states ("blocks"). The core mechanism involves:

1. Heralded Loading: Elementary blocks, defined as coherent superpositions of two logical alternatives (e.g., $|\psi_j\rangle = u_j X^\dagger_j + v_j Y^\dagger_j |0\rangle$), are generated probabilistically. Upon a successful herald, the block is routed to a storage loop.
2. Dump Mode Monitoring: The system monitors "dump modes" to post-select events where the incoming photons are successfully transferred into the same circulating modes as previously stored photons. A successful load removes the generation time label of the new block; a "dump" event (indicating a failed transfer) rejects the sequence.
3. Matched Transformations: The protocol requires that the two logical alternatives ($X$ and $Y$) of each block undergo identical complex loading transformations (spectral, temporal, and phase response) at every step. This ensures the loading amplitude is independent of the logical composition of the state already stored.
4. Polynomial Factorization: Under these conditions, the accumulated state is described by a product of linear factors. To synthesize a specific target superposition, the authors utilize polynomial factorization: the coefficients of the desired target state polynomial determine the required amplitudes ($u_j, v_j$) and phases of the elementary blocks.
5. Adaptive Coupling: To optimize efficiency, the coupling strength of the storage loops is adjusted as the number of stored photons increases. Specifically, the coupling parameter $x_j$ is set to $1/(1+j)$, where $j$ is the loading step index.

Key Contributions

• Time-Label Removal: The protocol successfully removes generation time labels by ensuring that newly loaded blocks occupy the same circulating modes as stored photons, enabling coherent interference of different generation histories.
• Programmable State Synthesis: The authors provide a direct prescription for constructing arbitrary target superpositions in a specific subspace by factorizing the target state's associated polynomial. This allows for the construction of complex states from simple, probabilistic blocks.
• Scaling Improvement: A critical theoretical contribution is the demonstration that adapting storage loop couplings changes the loading efficiency scaling. Fixed balanced couplers result in a "faster-than-exponential" penalty for success probability as the state grows. In contrast, the proposed adaptive coupling ($x_j = 1/(1+j)$) reduces this penalty to exponential scaling, significantly improving the viability of assembling large multiphoton states.
• Clock-State Realization: The paper applies this framework to a specific two-photon path–frequency state used in photonic clock interferometry. It details the generation of "clock blocks" via entanglement swapping on idler photons and the subsequent fusion into a NOON-type state where endpoints are distinguished by exchanged frequency assignments rather than spatial modes.

Results and Performance Estimates

• Success Probability: For the "roots of unity" setting (generating NOON-type states), the optimal loading probability is derived as $p_{opt, ROU} \propto 2^{1-N} \left( \frac{(N-1)!}{N^{N-1}} \right)^{n_b}$. Using Stirling's approximation, this probability decreases exponentially with $N$, which is a marked improvement over the faster-than-exponential decay of fixed couplers.
• Detected Rates: The authors estimate the detected event rate ($R_{out, N}$) for a sequential pulsed source, accounting for source waiting time, round-trip loss ($\lambda$), and detection efficiency ($\eta$).
  • For a reference round-trip loss of 1.2% and a source probability $p_h = 10^{-2}$, the estimated rate for $N=2$ is $1.2 \times 10^3$ s$^{-1}$, dropping to 11 s$^{-1}$ for $N=3$.
  • Achieving a rate of 1 s$^{-1}$ at $N=4$ requires reducing the round-trip loss to approximately $3.4 \times 10^{-3}$.
• Regime Analysis: The performance analysis indicates that for typical parameters, the system operates in a "storage-loss-limited" regime, where the survival probability is heavily influenced by the ratio of block supply probability to round-trip loss.

Significance and Claims
The paper claims that this architecture offers a route to assembling asynchronously generated entangled resources across different photonic encodings while preserving the delivered photons for subsequent sensing or communication. Key advantages highlighted include:

• Spectral Preservation: Unlike projective fusion methods that may require active frequency conversion or coherent rotations between frequency bins before detection, this protocol performs Bell projections only on the heralding idler photons. The signal photons remain spectrally untouched throughout the fusion sequence.
• Modularity: The scheme relies on mode-matched loop coupling, phase stability, and low-loss storage, ingredients that are compatible with existing integrated and microresonator platforms.
• Scalability: The architecture naturally extends to larger fusion sequences and multiplexed block sources, where parallel generation could further suppress stochastic waiting times.

The authors maintain a modest tone, presenting the work as a theoretical proposal and performance estimate rather than a completed experimental demonstration, noting that the specific loss values cited are reference points rather than demonstrated end-to-end operating points.