---
title: 'ChordEdit: Bell-Fusion Pairing Geometry'
url: https://www.emergentmind.com/topics/chordedit
type: topic
---

# ChordEdit: Bell-Fusion Pairing Geometry

In the setting of this paper, “Bell‑fusion pairing geometry” is essentially about how, where, and in what pattern photons can be paired and fused via Bell measurements, given a hard constraint: the amount of auxiliary *Schmidt rank* you can embed into the same two photons you’re measuring. The paper identifies this Schmidt rank as a sharp, quantifiable resource that directly shapes which Bell‑fusion geometries are possible with passive linear optics and no extra photons.

Below, I’ll first set up the model, then state and interpret the main theorems, and finally connect them to fusion‑based architectures and geometric/graph‑like pictures of how Bell fusion can be arranged.

---

## 1. Setting: Photonic Bell measurements in linear optics

The paper considers two main contexts:

- **Quantum communication:** teleportation, entanglement swapping, quantum repeaters, dense coding.
- **Fusion-based photonic quantum computation:** graph/fusion operations where Bell measurements fuse resource states into larger cluster/graph states.

In all of these, a *Bell measurement* (BM) is operationally:

- A projective measurement in a Bell basis on two *system* qudits, each of dimension \(d\).
- Implemented in optics by:
  - Two photons entering a passive linear-optical interferometer (static beam splitters, phase shifters; no squeezing, no nonlinearities).
  - Followed by photon-number-resolving (PNR) detection and classical post-processing.

The key restriction:

- **Exactly two photons are populated.**  
  No additional populated ancilla photons; any extra optical modes start in vacuum.
- Operations are **passive linear optics only** (no squeezing, no feed-forward unitaries between detection events).
- The two photons may carry *additional degrees of freedom* (DoFs) besides the system qudits, and these auxiliary DoFs may be entangled.

This is the “same‑photon‑assisted, ancilla‑photon‑free” regime: all physical resources are *embedded in the two photons themselves* via multiple DoFs.

---

## 2. System qudits, auxiliary state, and auxiliary Schmidt rank

Each of the two photons carries:

- A **system qudit**:
  - \(S_A\) on photon \(A\), \(S_B\) on photon \(B\), each of dimension \(d\).
- Possibly several **auxiliary degrees of freedom**:
  - For photon \(A\): \(R_A\) (maybe a tensor product of polarization, time-bin, frequency, OAM, etc.).
  - For photon \(B\): \(R_B\) similarly.

The combined one-photon Hilbert spaces are
\[
\mathcal{H}_A = \mathcal{H}_{S_A}\otimes \mathcal{H}_{R_A},\quad
\mathcal{H}_B = \mathcal{H}_{S_B}\otimes \mathcal{H}_{R_B}.
\]

The **unknown input** is one of the \(d^2\) generalized Bell states on \(S_A\otimes S_B\), tensored with a fixed, known auxiliary state on \(R_A\otimes R_B\),
\[
\mathcal{S}^{(d)}_{\Phi} = \left\{
|\Psi^{(d)}_{pq}\rangle_{S_A S_B}\otimes|\Phi\rangle_{R_A R_B} : p,q\in\mathbb{Z}_d
\right\}.
\]

### Generalized Bell states

Let \(X\) and \(Z\) be the generalized Pauli (Weyl) operators:
\[
X|n\rangle = |n+1\rangle,\quad Z|n\rangle = \omega^n|n\rangle,\quad \omega = e^{2\pi i/d},
\]
with addition modulo \(d\). Then the Bell matrices are
\[
C_{pq} = d^{-1/2} X^p Z^q,\qquad p,q\in\mathbb{Z}_d,
\]
and correspondingly the Bell states
\[
|\Psi^{(d)}_{pq}\rangle
= \frac{1}{\sqrt{d}}\sum_{n=0}^{d-1}\omega^{qn}|n+p\rangle_A|n\rangle_B.
\]
Each \(|\Psi^{(d)}_{pq}\rangle\) is maximally entangled and has Schmidt rank \(d\).

### Auxiliary entangled state and its Schmidt rank

The auxiliary state on the extra DoFs is a fixed pure state
\[
|\Phi\rangle_{R_A R_B} = \sum_{a,b}\Phi_{ab}\,|a\rangle_{R_A}|b\rangle_{R_B}.
\]

The **total auxiliary Schmidt rank**
\[
r_\Phi = \operatorname{rank}\Phi
\]
is the Schmidt rank across the partition \(R_A|R_B\). If there are multiple auxiliary DoFs \(\kappa\), and
\(|\Phi\rangle = \bigotimes_\kappa |\Phi^{(\kappa)}\rangle\), then
\[
r_\Phi = \prod_\kappa r_{\Phi^{(\kappa)}},
\]
so Schmidt rank multiplies across auxiliary DoFs.

Crucially, the paper treats **\(r_\Phi\) as the resource**: it quantifies how much auxiliary entanglement is embedded into the same two photons.

---

## 3. Bare two-photon limitation

Before adding auxiliaries, consider the bare two-qudit case (no auxiliary entanglement, only system qudits \(S_A, S_B\)).

An arbitrary two-qudit state is
\[
|C\rangle = \sum_{m,n=0}^{d-1} C_{mn}\,a_m^\dagger b_n^\dagger|\mathrm{vac}\rangle,
\quad \mathrm{Tr}(C^\dagger C)=1,
\]
where \(a_m^\dagger\) creates mode \(m\) of qudit \(A\), and similarly for \(b_n^\dagger\).

A passive interferometer implements linear mode mixing; a fine-grained PNR outcome with photons in output modes \((\mu,\nu)\) has amplitude
\[
A_{\mu,\nu}(C) = \langle\mu,\nu|C\rangle
= \frac{\alpha_\mu^T C \beta_\nu + \alpha_\nu^T C\beta_\mu}{\sqrt{1+\delta_{\mu\nu}}},
\]
equivalently
\[
A_{\mu,\nu}(C)=\mathrm{Tr}\big(P_{\mu\nu}^T C\big),
\]
with
\[
P_{\mu\nu} = \frac{\alpha_\mu\beta_\nu^T + \alpha_\nu\beta_\mu^T}{\sqrt{1+\delta_{\mu\nu}}}.
\]

Here:

- \(\alpha_\mu, \beta_\mu\) are the pieces of output row \(\mu\) restricted to the system registers \(A,B\).
- \(P_{\mu\nu}\) is a sum of two outer products; hence
  \[
  \operatorname{rank} P_{\mu\nu} \le 2.
  \]

A *microscopic pattern* \((\mu,\nu)\) is conclusive for a Bell label \((p,q)\) only if its associated functional is proportional to the target Bell matrix:
\[
P_{\mu\nu} \propto C_{pq}.
\]
But every \(C_{pq}\) is rank \(d\), while \(P_{\mu\nu}\) is rank \(\le 2\). So for \(d>2\):

- No fine-grained detection pattern can be conclusive for a Bell label.
- By extension, no coarse-grained outcome can be conclusive either.

This is a very strong statement: **with two photons, passive linear optics, and no auxiliary entanglement, there is no conclusive high‑dimensional Bell measurement at all for \(d>2\)**, not even probabilistic.

Geometrically, in a “fusion graph” picture where edges represent Bell links, the bare two‑photon node has *zero ability to resolve any of the \(d^2\) edge labels* for \(d>2\). It can’t tell you which Bell edge you have.

---

## 4. Adding auxiliary DoFs: effective Bell-label functionals

Now include the auxiliary entanglement \(|\Phi\rangle\) between \(R_A\) and \(R_B\).

The two‑photon detection still comes from a rank‑2 two‑click vector in the enlarged space \((S_A\otimes R_A)\otimes(S_B\otimes R_B)\). Let the coefficient matrix of a microscopic two‑click vector be \(P\). Contracting with the known auxiliary state yields an **effective system‑level functional**
\[
[\Gamma_\Phi(P)]_{ij} = \sum_{a,b} P_{(i,a),(j,b)}\,\Phi_{ab}^*,
\]
so that for any system state \(|C\rangle\),
\[
\langle P|C\otimes\Phi\rangle = \mathrm{Tr}\,[\Gamma_\Phi(P)^\dagger C].
\]

If \(P\) has Schmidt rank \(\le 2\) on the enlarged bipartition, it can be decomposed as
\[
P = u_1 v_1^T + u_2 v_2^T
\]
and, after reshaping, one finds
\[
\Gamma_\Phi(P) = \sum_{t=1}^2 U_t\,\Phi^*\,V_t^T,
\]
with \(U_t,V_t : \mathbb{C}^{r_\Phi}\to\mathbb{C}^d\).

Each term \(U_t\Phi^*V_t^T\) has rank \(\le r_\Phi\), so
\[
\operatorname{rank} \Gamma_\Phi(P) \le \min(d, 2 r_\Phi).
\]

Conversely, any \(d\times d\) matrix of rank \(\le \min(d,2r_\Phi)\) can be written in this form, so the system‑level contractions accessible from some two‑click pattern fill the variety
\[
\mathcal{K}^{(2)}_\Phi = \{ Q\in\mathbb{C}^{d\times d} : \operatorname{rank}Q\le \min(d,2r_\Phi)\}.
\]

In other words: **auxiliary entanglement “amplifies” the rank of the effective functional from \(\le 2\) up to \(\le 2r_\Phi\)**.

---

## 5. Main Theorem 1: Single conclusive Bell-label functional

A microscopic outcome is conclusive for label \((p,q)\) if its effective functional is orthogonal to all \(C_{p'q'}\) with \((p',q')\neq(p,q)\), which forces
\[
\Gamma_\Phi(P) \propto C_{pq}.
\]
Since \(C_{pq}\) has rank \(d\), such a functional exists in \(\mathcal{K}^{(2)}_\Phi\) iff
\[
\min(d,2r_\Phi) \ge d \quad\Longleftrightarrow\quad 2r_\Phi \ge d.
\]

So we obtain the **single‑outcome threshold**:
\[
\boxed{ r_\Phi \;\ge\; \big\lceil \tfrac{d}{2} \big\rceil \iff \text{a single conclusive Bell-label functional is possible.} }
\]

Interpretation:

- If you can embed auxiliary entanglement with Schmidt rank at least \(\lceil d/2\rceil\) into the same two photons, there exists at least one detection pattern that uniquely identifies some Bell label \((p,q)\).
- For example, for qutrits \(d=3\), an auxiliary qubit entangled state (\(r_\Phi=2\)) suffices to construct exactly one microscopic pattern that fires only on a chosen Bell state—though not enough to make the whole measurement deterministic.

In fusion geometry terms:

- With \(r_\Phi \ge \lceil d/2\rceil\), a given node can host *isolated fusion edges* whose presence can sometimes be identified unambiguously (single conclusive patterns), but the node cannot yet support a *complete deterministic Bell‑fusion interface* for all possible Bell labels.

---

## 6. Main Theorem 2: Deterministic discrimination of all \(d^2\) Bell states

Deterministic full-label BM imposes a much stronger condition:

- Every fine‑grained two‑photon Fock pattern (including “bunching” where both photons exit the same mode) must either:
  - Be impossible for all inputs (zero probability), or
  - Be compatible with exactly one Bell label \((p,q)\).

Since the Bell matrices form a complete orthonormal Hilbert–Schmidt basis, any nonzero contraction that vanishes on all non‑target labels must be proportional to the corresponding \(C_{pq}\), hence full rank \(d\).

The core of the second main theorem is a rank argument that uses the **same‑mode events** \((\mu,\mu)\). Summarizing:

1. Put \(|\Phi\rangle\) in Schmidt form
   \[
   |\Phi\rangle = \sum_{a=1}^r \lambda_a |a\rangle_{R_A}|a\rangle_{R_B},\quad r = r_\Phi,
   \]
   with invertible \(\Lambda = \operatorname{diag}(\lambda_1,\ldots,\lambda_r)\).

2. For each output mode \(\mu\), define \(X_\mu,Y_\mu \in \mathbb{C}^{d\times r}\) as the system–auxiliary blocks of the interferometer row for registers \(A\) and \(B\). The contraction for pattern \((\mu,\nu)\) is
   \[
   Q_{\mu\nu} = \frac{X_\mu\Lambda Y_\nu^T + X_\nu\Lambda Y_\mu^T}{\sqrt{1+\delta_{\mu\nu}}}.
   \]

3. Same‑mode events: \(Q_{\mu\mu} = \sqrt{2}\,X_\mu\Lambda Y_\mu^T\) has rank \(\le r\). If \(r<d\), such a matrix **cannot** be proportional to any Bell matrix. Determinism then forces
   \[
   X_\mu\Lambda Y_\mu^T = 0 \quad\forall\mu.
   \]
   Let \(a_\mu = \operatorname{rank}X_\mu\), \(b_\mu = \operatorname{rank}Y_\mu\). Sylvester’s inequality gives
   \[
   0 = \operatorname{rank}(X_\mu\Lambda Y_\mu^T) \ge a_\mu + b_\mu - r
   \quad\Rightarrow\quad a_\mu + b_\mu \le r.
   \]

4. Off‑diagonal events: for \(\mu\neq\nu\),
   \[
   \operatorname{rank}Q_{\mu\nu}
   \le \min(a_\mu,b_\nu)+\min(a_\nu,b_\mu)
   \le \frac{a_\mu+b_\mu + a_\nu + b_\nu}{2}
   \le r < d.
   \]

   So no off‑diagonal microscopic pattern can have rank \(d\) either. No pattern can implement \(C_{pq}\).

Thus:

\[
\boxed{
r_\Phi < d \;\Rightarrow\; \text{deterministic full-label BM on }\mathcal{S}^{(d)}_\Phi \text{ is impossible in this model.}
}
\]

Conversely:

- The paper constructs an explicit saturating example showing **existence** of a deterministic analyzer when \(r_\Phi = d\).
- Therefore, the threshold is **exact** in the same‑photon, passive‑optics, two‑photon resource class:

\[
\boxed{
r_\Phi \ge d \;\Longleftrightarrow\; \text{deterministic discrimination of all \(d^2\) Bell states is possible (existentially).}
}
\]

From the viewpoint of “Bell‑fusion pairing geometry”:

- Nodes (two-photon fusion sites) with \(r_\Phi<d\) cannot serve as deterministic Bell‑fusion vertices: some Bell edges cannot be resolved, and the whole incoming Bell label space cannot be partitioned unambiguously across detection outcomes.
- Nodes with \(r_\Phi=d\) *can* be designed (in ideal mode control) to deterministically read out all Bell labels. Only such nodes can behave as **deterministic Bell‑fusion interfaces** in a graph-like architecture.

---

## 7. Achieving the bound: local Bell-basis sorting with a rank‑\(d\) auxiliary

The sufficiency side constructs an explicit deterministic analyzer for \(r_\Phi=d\).

Take a **maximally entangled auxiliary state of rank \(d\)**:
\[
|\Phi_d\rangle_{R_A R_B} = \frac{1}{\sqrt{d}} \sum_{a=0}^{d-1} |a\rangle_{R_A}|a\rangle_{R_B}.
\]

For each photon, define a *local* (one‑photon) Bell basis between its system and auxiliary:
\[
|\chi_{mn}\rangle_{S R} = \frac{1}{\sqrt{d}} \sum_{t=0}^{d-1} \omega^{nt} |t\rangle_S |t+m\rangle_R,
\quad m,n \in \mathbb{Z}_d.
\]

The central identity is a **same‑photon entanglement swapping decomposition**:
\[
|\Psi^{(d)}_{pq}\rangle_{S_A S_B} |\Phi_d\rangle_{R_A R_B}
= \frac{1}{d}\sum_{m,n\in\mathbb{Z}_d} \omega^{-np}\,
|\chi_{mn}\rangle_{S_A R_A} |\chi_{m+p, q-n}\rangle_{S_B R_B}.
\]
Thus, if you measure each photon in its local \(\{|\chi_{mn}\rangle\}\) basis, the outcomes \((m,n)\) on photon \(A\) and \((m',n')\) on photon \(B\) determine the original system Bell label via
\[
p = m' - m,\qquad q = n + n' \pmod d.
\]

Since \(\{|\chi_{mn}\rangle\}_{m,n}\) is an orthonormal single-photon basis, **there exists a passive single-photon unitary** (a mode sorter)
\[
U_{\text{loc}} : |s,r\rangle \mapsto |n,m\rangle,
\]
such that
\[
U_{\text{loc}}|\chi_{mn}\rangle = |n,m\rangle
\]
maps each Bell basis state to a distinct output mode pair \((n,m)\). Implement this \(U_{\text{loc}}\) separately on the modes of photon \(A\) and photon \(B\). Then:

- Photon \(A\): output mode indexed by \((n_A,m_A)\).
- Photon \(B\): output mode indexed by \((n_B,m_B)\).

PNR detection then returns \((n_A,m_A;n_B,m_B)\), and classical processing reconstructs \((p,q)\).

Thus:

- With a **maximally entangled rank‑\(d\) auxiliary** and
- Ideal local control of each photon’s \(S\otimes R\) modes,

one can implement a **deterministic, ancilla‑photon‑free Bell measurement**, using only passive linear optics and PNR detection.

This construction saturates the \(r_\Phi\ge d\) bound.

---

## 8. Implications for Bell‑fusion pairing geometry

Now we connect the theorems to geometric/graph-like pictures of fusion‑based architectures—how Bell pairs are fused, paired, routed, and read out.

Think of a **fusion node** as a two‑photon measurement site in a larger photonic graph (cluster/fusion network). Its local structure involves:

- Two incoming **system edges** (Bell pairs or logical links) attached to photons \(A\) and \(B\) in DoF \(S\).
- One or more **auxiliary entangled edges** embedded in other DoFs \(R\) on the same photons.
- A linear‑optical circuit that mixes all those modes and routes to detectors.

### 8.1. Auxiliary Schmidt rank as a node “capacity”

Given the two main thresholds:

- **Below** \(\lceil d/2\rceil\): no conclusive Bell label at all; the node cannot even occasionally identify a single Bell edge.
- **Between** \(\lceil d/2\rceil\) and \(d\): there exist *isolated* conclusive Bell‑label functionals, but **no deterministic full coverage** is possible.
- **At or above** \(d\): there exists at least one design achieving *deterministic* discrimination of all \(d^2\) Bell labels.

Interpret this as a **capacity constraint** at each fusion node:

- A node with total auxiliary Schmidt rank \(r_\Phi\) has an effective capacity to realize full‑rank functionals of rank at most \(\min(d,2r_\Phi)\).
- Deterministic Bell-fusion—where the node must correctly resolve any of the \(d^2\) possible incoming Bell labels—requires this capacity to be \(d\), forcing \(r_\Phi\ge d\).

This shapes geometry in several ways:

1. **Which Bell labels can be reliably identified.**  
   - With \(r_\Phi<d\), no microscopic detection pattern can be proportional to any Bell matrix when determinism is required; some labels (indeed all labels, under the theorem) cannot be deterministically resolved in that resource class.
   - With \(r_\Phi\ge d\), all labels can be resolved in principle, and the node can serve as a fully labeled fusion vertex.

2. **Deterministic vs probabilistic fusion attempts.**  
   - For deterministic fusion (no post-selection on measurement outcome), the node must have \(r_\Phi\ge d\).
   - If one is willing to accept *probabilistic* fusion or *grouped* outcomes (e.g., only parity information, or only a subset of labels), then nodes with smaller \(r_\Phi\) may still be useful, but the measurement is no longer a complete physical BM and the network geometry must accommodate probabilistic edges or coarse‑grained labels.

3. **Routing and allocation of auxiliary DoFs.**  
   Because the total Schmidt rank multiplies across auxiliary DoFs, you can distribute the entanglement across, say, polarization, time bins, and frequency modes:
   \[
   r_\Phi = r_{\text{pol}} \times r_{\text{time}} \times r_{\text{freq}} \times \cdots.
   \]
   Geometrically, this means you have flexibility in how you “embed” the necessary rank‑\(d\) entanglement into the photons, but you **cannot reduce the total Schmidt rank below \(d\)** and still keep deterministic Bell‑fusion capability.

4. **Same-mode events and local rank balance.**  
   The deterministic no‑go proof hinges on *same‑mode events* \((\mu,\mu)\) and the induced rank split
   \[
   a_\mu + b_\mu \le r_\Phi.
   \]
   This constraint is structural: any optical geometry (even large interferometers with vacuum ancillas) must allocate its rank budget across modes in such a way that if \(r_\Phi<d\), then ranks of off‑diagonal contractions are necessarily \(<d\).  
   Intuitively: when designing a mode‑mixing network for fusion, you cannot “hide” the rank deficiency in some tricky routing; same‑mode events force a local rank budget that propagates to all patterns.

### 8.2. Edge/graph picture

Picture the fusion‑based architecture as a graph:

- **Nodes:** photons or fusion sites.
- **Edges:** entangled links (Bell pairs) between photonic modes or logical qudits.

At a fusion site where two edges meet (two qudits \(S_A,S_B\)):

- Auxiliary entanglement \(|\Phi\rangle\) in other DoFs corresponds to *additional edges between the same two nodes* (A and B) in parallel layers (polarization layer, time‑bin layer, etc.).
- The **auxiliary Schmidt rank** counts how many independent entangled directions span these extra layers.

The theorems say:

- **Completeness of Bell fusion** (being able to fuse arbitrary Bell edges deterministically) demands that these auxiliary layers between \(A\) and \(B\) span a \(d\)-dimensional entangled subspace, i.e., an edge of Schmidt rank \(d\) between \(R_A\) and \(R_B\).
- With fewer than \(d\) auxiliary dimensions entangled, the fusion edge is *too narrow* to fully resolve and re‑route the \(d^2\) Bell labels between the graph’s system edges.

So the “pairing geometry” of Bell fusion is constrained at each node by:

- How many auxiliary entangled layers exist between the same two photons.
- How those layers’ Schmidt ranks multiply to reach or fall short of \(d\).

### 8.3. Trade-offs: auxiliary photons vs auxiliary DoFs vs Schmidt rank

There are three conceptually different knobs:

1. **Auxiliary photons** (ancilla-photon schemes):  
   Add more photons to the Fock sector. This is *not allowed* in the present model, but many BM schemes in the literature use it to circumvent rank limitations.

2. **Auxiliary DoFs on the same photons:**  
   Keep the photon count fixed, but extend each photon’s mode space; embed entanglement across those DoFs. This is precisely the model of the paper.

3. **Total auxiliary Schmidt rank \(r_\Phi\):**  
   The product over all auxiliary DoFs.

Trade-off implications:

- In an ancilla-photon-free geometry, you *must* pay for deterministic fusion in the currency of **auxiliary Schmidt rank** \(r_\Phi\ge d\).
- You can distribute this rank over many physical DoFs (spatial paths, time bins, polarization, frequency, OAM), but the *total* rank must reach \(d\).
- If you want to reduce the auxiliary Schmidt rank per DoF (e.g., for experimental simplicity in each mode), you must compensate by adding more auxiliary DoFs or, alternatively, by moving to a different resource class (adding ancilla photons or active operations).

So for designing Bell-fusion networks:

- A **single physical photon pair** used as a fusion bond can be either:
  - “Thin”: few or no auxiliary entangled DoFs (bare or small \(r_\Phi\)), leading to probabilistic or grouped fusion; or
  - “Thick”: endowed with a rank‑\(d\) auxiliary entangled structure across its other DoFs, enabling deterministic, fully labeled fusion.

Different parts of a fusion-based architecture could mix these: deterministic high‑rank nodes where resource investment is justified, and cheaper low‑rank nodes where probabilistic fusion suffices.

---

## 9. “Ancilla‑photon‑free, embedded” BM and structural design

“Ancilla‑photon‑free, embedded” BM means:

- No extra photons beyond the two system photons;
- The auxiliary entanglement used to improve the BM is *embedded* in extra DoFs of the same photons.

Structurally, for a fusion‑based computation scheme, this implies:

1. **Local mode design problem:**  
   Each fusion site must realize an appropriate single-photon unitary \(U_{\text{loc}}\) acting on the composite mode space \(\mathcal{H}_{S}\otimes\mathcal{H}_{R}\). In the maximal case, this is a Bell‑basis sorter between \(S\) and \(R\).

2. **Graph-level constraints:**  
   The physical layout must provision auxiliary entangled states \(|\Phi\rangle\) with \(r_\Phi \ge d\) between the same two photons that are to be fused deterministically. This translates to specific geometric connections in auxiliary mode layers (e.g., time-bin entangled link + polarization entangled link, etc.) between those two photons.

3. **Separation from logical fusion schemes:**  
   Logical BM and fusion protocols that operate on encoded DOFs or accept grouped outcomes do not necessarily require full physical Bell-label decoding, and hence may not require \(r_\Phi\ge d\). They implement different tasks and obey different resource–success tradeoffs.

In a graph-like picture, “embedded” BM means that the auxiliary edges for Bell fusion live on the *same vertices* as the system edges, but possibly in different layers. The theorem then says: if you want a vertex to be a fully deterministic Bell‑fusion junction for physical qudits of dimension \(d\), the graph must contain, in those auxiliary layers, an entangled edge of Schmidt rank at least \(d\) between the same two vertices.

---

## 10. Summary: how the paper governs Bell‑fusion pairing geometry

- A two‑photon Bell measurement with passive linear optics is fundamentally constrained by the available **auxiliary Schmidt rank** \(r_\Phi\) embedded into the same photons.
- For dimension \(d>2\):
  - Without auxiliaries (\(r_\Phi=0\)), no conclusive generalized Bell‑label outcome exists at all.
  - A single conclusive Bell functional (one identifiable Bell label) is possible iff \(r_\Phi\ge\lceil d/2\rceil\).
  - Deterministic discrimination of all \(d^2\) Bell labels is possible (and only possible) when \(r_\Phi\ge d\); a maximally entangled rank‑\(d\) auxiliary state achieves this via local Bell‑basis sorting between each photon’s system and auxiliary DoFs.
- In **fusion‑based architectures**, these results translate into:

  - A local “capacity” constraint: each Bell‑fusion node must have total **auxiliary Schmidt rank \(r_\Phi\ge d\)** to serve as a deterministic physical Bell‑measurement vertex for system dimension \(d\).
  - The geometry of how Bell pairs are fused, paired, and routed is constrained by how auxiliary entanglement is distributed across physical DoFs and nodes: only nodes with rank‑\(d\) auxiliary entanglement can provide full Bell‑label resolving capability.
  - Trade-offs between using auxiliary photons vs auxiliary DoFs are now quantifiable: in the ancilla‑photon‑free model, you cannot lower the total Schmidt rank below \(d\) and still keep deterministic Bell fusion.

In this sense, the paper turns “Bell‑fusion pairing geometry” into a precise resource theory: the *auxiliary Schmidt rank* of same‑photon assistance is the minimal structural ingredient that dictates which Bell‑fusion geometries are physically realizable and deterministic in passive, two‑photon photonic systems.

Source: https://www.emergentmind.com/topics/chordedit