---
title: Boosted Type-I Fusion Gate
url: https://www.emergentmind.com/topics/boosted-type-i-fusion-gate
type: topic
---

# Boosted Type-I Fusion Gate

Searching arXiv for the core and related fusion-gate papers to ground the article in the current literature.
A boosted Type-I fusion gate is an ancilla-assisted fusion measurement for linear-optical quantum information processing that raises the heralded success probability of Type-I fusion above the ancilla-free \(1/2\) limit. In the photonic setting, Type-I and Type-II fusion gates are used to combine small entangled resource states into larger graph states, cluster states, or other multipartite resources. The contemporary literature distinguishes several closely related notions: a direct single-photon-ancilla implementation of boosted Type-I fusion with total success probability \(3/4\) [2603.27315], experimentally realized boosted fusion for graph-state construction based on an enhanced Bell-state measurement primitive with measured success probability \(71.0(7)\%\) [2412.18882], and high-dimensional pairwise fusion gates that play the role of a boosted Type-I fusion in qudit encodings [2407.10893]. Across these variants, the common objective is to convert more measurement outcomes from failure into useful heralded fusion events, thereby reducing the resource overhead of scalable photonic quantum computing.

## 1. Definition and conceptual scope

In its standard dual-rail photonic form, a Type-I fusion gate acts on two qubits encoded in modes \((1,2)\) and \((3,4)\), uses a \(50{:}50\) beam splitter and photon-number-resolving detection on measured modes \(1\) and \(4\), and succeeds when exactly one photon is detected in mode \(1\) or mode \(4\) [2603.27315]. The standard success probability is
\[
P^s=\frac{1}{2},
\]
which is the usual ancilla-free Type-I limit and matches the general no-go result for linear optics without auxiliary resources [2603.27315].

The qualifier “boosted” denotes a modification that increases this intrinsic heralding efficiency beyond \(1/2\). In the most direct recent formulation, the gate uses four ancillary single photons and passive linear optics, achieves a direct success probability of \(5/8\), and reaches a total success probability of \(3/4\) after a distillation step on a partially entangled branch [2603.27315]. Related literature uses the same basic idea in adjacent settings: ancilla-assisted Bell-state discrimination for graph-state fusion [2412.18882], qudit pairwise Bell projections with success \(1-d^{-1}\) or \(1-d^{-(k+1)}\) [2407.10893], and, in a broader analogical sense, ancilla-assisted fusion enhancements for W-state networks [1303.4008] and fusion-and-measurement protocols in SU\((2)_4\) anyon models [1504.02098].

A useful organizing distinction is between three layers of “boosting.” First, the fusion device itself can be modified so that more detector patterns correspond to successful projections. Second, partially useful output branches can be distilled into standard fusion outputs. Third, surrounding switch networks and muxing stages can increase the probability that the fusion device receives valid inputs at all [2109.13760]. These are conceptually distinct, although all are directed toward the same architectural bottleneck.

## 2. Baseline Type-I fusion and the source of the \(1/2\) limit

The baseline Type-I gate succeeds only on a subset of the possible photon-count outcomes after linear-optical interference. In the formulation given in "Single-photon-boosted type-I fusion gates" [2603.27315], the successful Kraus operators are associated with the single-photon outcomes on the measured modes, while the remaining branches correspond to vacuum or two-photon patterns and do not yield the desired fused logical output. When the gate succeeds, the output state is
\[
|\psi^{(\mathrm{out})}\rangle
=
\frac{1}{\sqrt{2}}
\bigl(
|A_0B_0\rangle|0\rangle_c
\pm
|A_1B_1\rangle|1\rangle_c
\bigr),
\]
and the success probability is \(1/2\) [2603.27315].

This limitation is structurally related to the restricted Bell-state discrimination available with passive linear optics. The graph-state experiment of 2024 states the same operational constraint in a different language: standard Type-I fusion has a maximum success probability of \(50\%\), which is insufficient for scalable graph-state growth under the relevant percolation constraints [2412.18882]. The qudit pairwise-fusion literature makes the same comparison explicitly, taking “conventional qubit Type-I fusion” as the \(1/2\) benchmark that high-dimensional schemes aim to surpass [2407.10893].

A common misconception is to treat the \(1/2\) limit as a property of all fusion-based architectures. The current literature instead treats it as the limit of ancilla-free linear-optical Type-I or Type-II fusion. Ancilla photons, ancillary entangled states, enlarged Hilbert spaces, or repeat-until-success measurement structures can all increase the effective success probability above \(1/2\) [2603.27315] [2407.10893] [1504.02098].

## 3. Single-photon-ancilla boosted Type-I fusion

The most direct instantiation of the topic is given in "Single-photon-boosted type-I fusion gates" [2603.27315]. The gate uses four ancillary single photons, only passive linear optics, and no ancillary Bell-pair resource states. Its central purpose is to exceed the ancilla-free \(1/2\) limit while avoiding the probabilistic and resource-intensive preparation of ancillary Bell states required in earlier \(3/4\)-efficient Type-I schemes [2603.27315].

The circuit has two parts. In part (I), the ancillary input
\[
|1_5 1_6 1_7 1_8\rangle
\]
is transformed into a product of two two-photon NOON-like states,
\[
|\mathrm{anc}\rangle
=
\frac{1}{2}
\bigl(|20\rangle-|02\rangle\bigr)\otimes
\bigl(|20\rangle-|02\rangle\bigr),
\]
so the four single photons are reshaped into two \((|20\rangle-|02\rangle)/\sqrt{2}\) states [2603.27315]. In part (II), the input qubits and the ancilla are interfered, modes \(1,4,5,7\) are measured, and the remaining modes carry the surviving output [2603.27315].

The protocol classifies events by the total detected photon number \(n_d\) in modes \(1,4,5,7\). The odd-photon branches \(n_d=1,3,5\) reproduce the standard Type-I fused output and contribute total probability
\[
P^s_{\mathrm{odd}}=\frac{1}{2}.
\]
A distinct class within the four-photon branch \(n_d=4\) yields an additional useful two-qubit entangled output with probability
\[
P^s_{n_d=4}=\frac{1}{8}.
\]
Consequently, the direct success probability is
\[
P^s_{\mathrm{direct}}=\frac{1}{2}+\frac{1}{8}=\frac{5}{8}.
\]
This is the direct boost over the standard \(1/2\) limit [2603.27315].

The \(n_d=2\) branch occurs with probability
\[
P_{n_d=2}=\frac{3}{16}
\]
and produces partially entangled states that are not yet standard fusion successes but can be distilled [2603.27315]. The distillation circuit uses balanced beam splitters and beam splitters with equal transmission coefficient \(t\). A successful round is heralded by vacuum detection in mode \(3\) and exactly one photon in either mode \(9\) or mode \(10\); two photons in mode \(3\), or in modes \(9\) or \(10\), constitute failure; and if no photons are detected in modes \(3,9,10\), the procedure may be repeated [2603.27315]. The asymptotic contribution of this branch is
\[
P^s_{n_d=2}=\frac{1}{8},
\]
so the total fusion success probability becomes
\[
P^s=\frac{5}{8}+\frac{1}{8}=\frac{3}{4}.
\]
The paper also gives the one-stage special case with balanced beam splitters \(t=1/2\):
\[
P^s_{n_d=2}=\frac{1}{16},
\qquad
P^s=\frac{11}{16}.
\]
Accordingly, the literature distinguishes four efficiency levels for this gate: baseline \(1/2\), direct boosted \(5/8\), one-stage distillation \(11/16\), and full asymptotic \(3/4\) [2603.27315].

This scheme is stated to be the first Type-I fusion protocol to exceed \(50\%\) success using only single-photon ancillary inputs [2603.27315]. A plausible implication is that the gate isolates the architectural gain of ancilla-assisted fusion from the separate challenge of preparing entangled ancillary Bell resources.

## 4. Relation to boosted fusion for graph-state generation

The 2024 experiment "Boosted fusion gates above the percolation threshold for scalable graph-state generation" reports an experimental realization of a boosted fusion gate for graph-state construction, more precisely a boosted Type-II fusion gate protocol used as a Bell-state measurement primitive [2412.18882]. Although it is not a Type-I gate, it is central to the broader literature because it establishes the regime in which boosted fusion becomes directly relevant to scalable photonic graph-state growth.

The motivating requirement is percolation. For three-photon GHZ states as resource states, theoretical analysis gives a refined threshold
\[
p_c = 58.98\%
\]
for scalable graph-state generation, replacing an earlier estimate of about \(62.5\%\) [2412.18882]. The paper also notes a simulation threshold of \(0.672\) for building \(2\mathrm{D}\) cluster states when photon loss is ignored, but the experimentally relevant percolation threshold for the demonstrated fusion gate is \(58.98\%\) [2412.18882].

The experiment implements the boosted gate using two Bell states as resource states, four auxiliary photons arranged into two auxiliary two-photon states, an enhanced Bell-state measurement based on linear optics, and pseudo-photon-number-resolving detectors [2412.18882]. Eight single photons labeled \(1\)–\(8\) are used: photons \(1\)–\(4\) prepare two Bell pairs,
\[
|\Phi^+\rangle_{12},
\qquad
|\Phi^+\rangle_{34},
\]
and photons \(5\)–\(8\) prepare two auxiliary states,
\[
|\gamma\rangle_{56},
\qquad
|\gamma\rangle_{78},
\]
with the auxiliary state written as
\[
|\gamma\rangle_{56}
=
\frac{1}{\sqrt{2}}
\left(
|2_H0_V\rangle_{56}+|0_H2_V\rangle_{56}
\right),
\]
and similarly for \( |\gamma\rangle_{78} \) [2412.18882]. These auxiliary states are two-photon N00N-like states generated deterministically through Hong–Ou–Mandel interference [2412.18882].

The reported theoretical success probability is \(75\%\), and the measured success probability is \(71.0(7)\%\) [2412.18882]. The corrected probabilities for identifying the relevant Bell outcomes are \(0.248(5)\) for \( |\Psi^-\rangle_{23} \), \(0.270(6)\) for \( |\Psi^+\rangle_{23} \), \(0.095(5)\) for \( |\Phi^-\rangle_{23} \), and \(0.097(4)\) for \( |\Phi^+\rangle_{23} \), summing to the measured total [2412.18882]. The same work further demonstrates fusion of two Bell states with a fidelity of \(67(2)\%\), extracted from correlation measurements in the \(+-, RL, HV\) bases, which is interpreted as direct evidence that the fusion process created the desired entanglement between photons \(1\) and \(4\) [2412.18882].

For the history of boosted Type-I fusion, the significance of this experiment is indirect but substantial. It shows that ancilla-assisted linear-optical fusion can be pushed above both the conventional \(50\%\) limit and the graph-percolation threshold required for scalable resource growth [2412.18882]. This suggests that the success-probability targets pursued by boosted Type-I designs are not only of abstract interest; they are aligned with concrete scalability thresholds in photonic graph-state architectures.

## 5. High-dimensional and other generalized notions of boosted fusion

The qudit pairwise fusion gate of "Linear-optical fusion boosted by high-dimensional entanglement" is described as a high-dimensional, boosted version of the usual Type-I linear-optical fusion gate for qubits [2407.10893]. Rather than discriminating a full \(d\)-dimensional Bell basis, the pairwise fusion gate performs a Bell-state projection on a chosen two-qubit subspace spanned by two levels \(i\) and \(j\), with pairwise Bell states
\[
|\psi_{i,j,\pm}\rangle
=
\frac{1}{\sqrt 2}\bigl(|i\rangle|j\rangle \pm |j\rangle|i\rangle\bigr),
\]
\[
|\phi_{i,j,\pm}\rangle
=
\frac{1}{\sqrt 2}\bigl(|i\rangle|i\rangle \pm |j\rangle|j\rangle\bigr).
\]
In the ancilla-free circuit, the success probability is
\[
p_f = 1-d^{-1},
\]
and with ancilla photons it becomes
\[
p_f = 1-d^{-(k+1)},
\]
using \(2(2^k-1)\) ancilla photons [2407.10893]. For \(k=1\), the ancilla is a \(d\)-dimensional Bell state, and the construction reduces to the well-known qubit-boosting approach of Grice when \(d=2\) [2407.10893]. The importance of this work for the Type-I literature is conceptual: the larger Hilbert space is used not merely for encoding, but to make fusion itself much more likely to succeed [2407.10893].

A second related but distinct strand is W-state fusion. "Enhancing the W State Quantum Network Fusion Process with A Single Fredkin Gate" augments the Ozdemir et al. fusion gate with a single Fredkin gate and an ancillary \(H\)-polarized photon, converting the only failure case \((V,V)\) into a success \((V,H)\) at the original fusion-gate input [1303.4008]. The success probability rises from
\[
P_s^{FG}=\frac{n+m-2}{nm}
\]
to
\[
P_s^{FGF}=\frac{n+m-1}{nm},
\]
and the successful output becomes
\[
|W_{n+m-1}\rangle
\]
rather than \( |W_{n+m-2}\rangle \) because the ancilla is not consumed [1303.4008]. In the special Bell-state case, the protocol can create a \(W_3\) state with success probability \(3/4\) [1303.4008]. The paper does not use the phrase “boosted Type-I fusion gate” explicitly, but it is closely analogous in the sense that ancilla-assisted circuitry converts a previous failure class into success [1303.4008].

A broader analogical use appears in topological quantum computation. "Universal Gates via Fusion and Measurement Operations on SU\((2)_4\) Anyons" uses topological qubit fusion, forced measurement, and semi-forced measurement to convert ancillary anyon states into gates, including an exact irrational phase gate
\[
K = R_{\alpha/2}
=
\begin{pmatrix}
1 & 0\\
0 & e^{i\alpha}
\end{pmatrix},
\]
with
\[
e^{i\alpha}=\frac{1+i\,4\sqrt{3}}{7}
\]
as given in the source material [1504.02098]. Here the “boosted” aspect comes from repeated measurement structures and ancilla preparation rather than from photonic Bell-state discrimination [1504.02098]. This is not a Type-I photonic fusion gate, but it shows that “boosted fusion” has become a broader methodological idea: ancillary resources and repeat-until-success logic can upgrade fusion from a fragile projection into a reliable gate-conversion primitive.

## 6. Resource overhead, switch networks, and architectural significance

The main architectural justification for boosted Type-I fusion is resource reduction in repeated-fusion constructions. In the 2026 Type-I analysis, the average number of input single photons per successful target state is estimated under “perfectly resource-efficient multiplexing,” where a subroutine using \(n_0\) photons per attempt and succeeding with probability \(p_0\) has average cost
\[
\frac{n_0}{p_0}
\]
in the large-parallelism limit [2603.27315]. Using seed-state generation costs from the literature—Bell state from four single photons with success probability \(1/8\), and three-qubit GHZ state from six single photons with success probability \(1/32\)—the paper evaluates 4-qubit GHZ and 6-ring graph-state constructions [2603.27315].

For 4-qubit GHZ generation and 6-ring graph states, the reported average photon costs are as follows [2603.27315].

| Scheme | Standard / direct / one-stage / full |
|---|---:|
| \( |1\rangle^{\otimes10} \rightarrow |\mathrm{GHZ}_2\rangle + |\mathrm{GHZ}_3\rangle \rightarrow |\mathrm{GHZ}_4\rangle \) | 448 / 365 / 332 / 304 |
| \( |1\rangle^{\otimes12} \rightarrow |\mathrm{GHZ}_2\rangle^{\otimes3} \rightarrow |\mathrm{GHZ}_4\rangle \) | 320 / 232 or 197 / 197 or 172 / 169 or 152 |
| \( |1\rangle^{\otimes18} \rightarrow |\mathrm{GHZ}_3\rangle^{\otimes3} \rightarrow |G_6\rangle \) | 3840 / 2097 / 1615 / 1273 |
| \( |1\rangle^{\otimes24} \rightarrow |\mathrm{GHZ}_2\rangle^{\otimes6} \rightarrow |G_6\rangle \) | 2560 / 1203 / 845 / 613 |

These numbers correspond, respectively, to standard \(P^s=1/2\), direct boost \(P^s=5/8\), one-stage distillation \(P^s=11/16\), and full distillation \(P^s=3/4\) [2603.27315]. The paper’s explicit conclusion is that improving a single fusion primitive from \(1/2\) to \(3/4\) can have a large compounding effect in multiplexed, fusion-based architectures [2603.27315].

The surrounding hardware layer is treated in "Switch networks for photonic fusion-based quantum computing" [2109.13760]. That work emphasizes that switch networks are required for muxing stages used to boost the probabilities for allocating quantum states to fusion gates and other operations [2109.13760]. The core muxing probability is
\[
p_{\rm mux}=1-(1-p)^N,
\]
with the approximation
\[
p_{\rm mux}\simeq 1-e^{-Np}
\]
for large \(N\) and small \(p\) [2109.13760]. It also reports that rail permutation can increase the success probability for generating a dual-rail encoded Bell state from \(1/8\) to \(3/16\), decreasing the amount of muxing needed to reach a target output probability by a factor of \(\sim 1.55\) [2109.13760]. More generally, one- or two-layer MZI networks can rearrange entangled qubits in random modes to pre-assigned mode bundles with \(>70\%\) or \(100\%\) efficiency, and an additional layer of MZIs can yield a \(666/64 \approx 10\times\) improvement for a six-photon task when \(Np\ll 6\) [2109.13760].

The architectural lesson is that boosted Type-I fusion should not be viewed solely as a better interferometric subroutine. In practical FBQC, the boost can be amplified or negated by muxing, blocking, pairwise couplers, optical loss, and feedforward latency [2109.13760]. This suggests that the effective performance of a boosted gate is a systems-level property: the intrinsic fusion success probability, the quality of ancillary-state preparation, and the efficiency of switch-network allocation must all be considered jointly.

## 7. Status, terminology, and common points of confusion

The phrase “boosted Type-I fusion gate” is used most directly for the 2026 passive-linear-optics protocol with four ancillary single photons [2603.27315]. A frequent source of confusion is the proximity of this terminology to the 2024 experiment on boosted fusion above the percolation threshold, which in fact reports a boosted Type-II fusion gate protocol used as a Bell-state measurement primitive [2412.18882]. The underlying objective is the same—raising fusion efficiency above the conventional linear-optical \(50\%\) limit—but the operational primitive differs.

Another recurring ambiguity concerns whether “boosting” means ancillary entanglement, ancillary single photons, high-dimensional encoding, or repeated measurement. The literature supports all four usages, but not interchangeably. The 2026 photonic Type-I gate uses four ancillary single photons and passive linear optics [2603.27315]. The 2024 graph-state experiment uses deterministically generated auxiliary two-photon states based on Hong–Ou–Mandel interference [2412.18882]. The qudit pairwise fusion gate exploits the multiplicity of two-dimensional subspaces in a \(d\)-level system [2407.10893]. The anyonic construction uses forced and semi-forced measurement together with ancilla preparation [1504.02098]. The unifying pattern is not a single circuit identity, but the enlargement of the set of heralded useful outcomes.

A final misconception is that a success probability such as \(3/4\) alone determines scalability. The graph-state literature makes clear that scalability depends on the relevant architecture and threshold. For three-photon GHZ resource states in percolated photonic graph-state generation, the threshold highlighted in the experiment is \(58.98\%\) [2412.18882]. Thus a \(3/4\)-efficient Type-I gate is significant not merely because it is greater than \(1/2\), but because it lies well above the success-probability regime associated with scalable graph growth in related photonic architectures. A plausible implication is that future comparisons among boosted fusion gates will increasingly focus on end-to-end overhead, loss tolerance, and compatibility with specific resource-state factories rather than on isolated success probabilities alone.

Source: https://www.emergentmind.com/topics/boosted-type-i-fusion-gate