Unveiling the Importance of Longer Paths in Quantum Networks

Abstract

The advancement of quantum communication technologies is calling for a better understanding of quantum network (QN) design from first principles, approached through network science. Pioneering studies have established a classical percolation map** to model the task of entanglement transmission across QN. Yet, this map** does not capture the stronger, yet not fully understood connectivity observed in QNs, which facilitates more efficient entanglement transmission than predicted by classical percolation. In this work, we explore the critical phenomena of the potential statistical theory underlying this enhanced connectivity, known as concurrence percolation. Compared to classical percolation, the concurrence percolation map** employs a unique approach of “superposing” path connectivities, utilizing a different set of path connectivity rules, thereby boosting the overall network connectivity. Firstly, we analytically derive the percolation critical exponents for hierarchical, scale-free networks, particularly the ($U,V$) flower model, characterized by two distinct network length scales, $U\leq V$. Our analysis confirms that classical and concurrence percolations, albeit both satisfying the hyperscaling relation, fall into separate universality classes. Most importantly, this separation stems from their different treatment of non-shortest path contributions to overall connectivity. Notably, as the longer path scale $V$ increases, concurrence percolation retains unignorable dependence of both its critical threshold and critical exponents on $V$ and thus, comparing with its classical counterpart, shows a higher resilience to the weakening of non-shortest paths. This higher resilience is also observed in real-world network topology, e.g., the Internet. Our findings reveal a first principle for QN design: longer paths still contribute significantly to QN connectivity—as long as they are abundant.

The emerging prospect of quantum Internet [1] is driving the field of quantum communication to explore larger and more complex scales, entering the realms of statistical physics [2] and network science [1]. As a response, a quantum network (QN) model [3] has been introduced to explore the potentials of transmitting quantum resource—most notably, entanglement [4]—across complex network topologies. In the QN model, each link represents an identical, partially entangled state $\left|\psi(\theta)\right\rangle=\cos\theta\left|00\right\rangle+\sin\theta\left|11\right\rangle$, shared between two qubits that come from different nodes [Fig. 1a]. The parameter $\theta\in[0,\pi/4]$ serves as a “link weight” that quantifies the degree of entanglement. Specifically, $\theta=\pi/4$ signifies a perfectly entangled state, while $\theta=0$ indicates zero entanglement. This homogeneous, pure-state QN maps onto a classical percolation problem [3], where each link can employ certain quantum operations as a sort of “gambling” to enhance its level of entanglement, with allows a probability $p\equiv 2\sin^{2}\theta$ of becoming perfectly entangled and $1-p$ of losing all entanglement [3]. Within this percolation analogy, entanglement transmission between two distant nodes—say, Alice (A) and Bob (B)—translates to the probability of having a path comprising only maximally entangled links between A and B. The presence of such a path further ensures that infinite-distance entanglement can be established between A and B [5], marking a successful event of entanglement transmission.

In the thermodynamic limit (i.e., when the number of nodes $N\to\infty$), classical percolation exhibits a critical threshold $p_{\text{th}}$ [6, *galam1996universal, *cohen2002percolation]. It implies that if $\theta<\sin^{-1}{\sqrt{p_{\text{th}}/2}}$, no entanglement transmission can happen between A and B when their distance $L\to\infty$. This is, however, refuted by recent discovery of an alternate, more effective map** from QN to a percolation-like statistical theory of concurrence percolation [5].

The best way to understand the difference between classical and concurrence percolation is to express the percolation connectivity between A and B as a “superposition” of all paths’ connectivity [Fig. 1b]: In classical percolation, the overall probability of connection of A and B can be computed through each individual path, contingent upon $p$. The superposition can be simplified into connectivity rules, such as the series and parallel rules [Fig. 1a, left]. Similarly, in concurrence percolation, a distinct parameter $c$ called concurrence, a conventional measure of entanglement, can be defined via $\theta$ in the QN, $c\equiv\sin 2\theta$. The superposition of concurrence paths adheres to a distinct set of connectivity rules [Fig. 1a, right]. Concurrence percolation also reveals a critical threshold $c_{\text{th}}$, which, in terms of $\theta$, is always lower than $p_{\text{th}}$: $\left(\sin^{-1}{c_{\text{th}}}\right)/2\leq\sin^{-1}{\sqrt{p_{\text{th}}/2}}$ [9], indicating that QNs potentially have stronger connectivity than classical percolation predicts.

What is the origin of this stronger connectivity, from a statistical physics perspective? It is conceivable that concurrence percolation might simply be a variant of classical percolation (albeit under a different set of variables), thereby belonging to the same universality class. Alternatively, concurrence percolation could represent a fundamentally distinct phenomenon from classical percolation, characterized by distinct critical exponents. Previous research, unfortunately, has been unable to identify the universality class of concurrence percolation due to computational constraints [10]. Gaining a deeper insight into the nature of this stronger concurrence connectivity could shed light on QN design from first principles, as explored through the lens of network science.

In this study, we investigate the critical phenomena of concurrence percolation in networks displaying hierarchical and scale-free structures, typical characteristics found in numerous real-worlds networks such as the Internet [11]. Firstly, we exactly determine the critical exponents for a family of hierarchical scale-free network models known as the ($U$,$V$) flowers [12], which are characterized by two distinct network length scales, $U\leq V$. Our analysis confirms that concurrence percolation falls into a distinct universality class, separated from that of classical percolation.

Most importantly, we highlight that this separation is largely reflected in how the two percolation theories respond to an increase in the longer length scale $V$, controlling the lengths of non-shortest paths. When $V\to\infty$, we find that the classical critical exponents become decoupled from $V$, depending only on the shorter length scale $U$. In contrast, the concurrence critical exponents depend on both $U$ and $V$. This distinction extends to the behavior of critical thresholds: while both $p_{\text{th}}$ and $c_{\text{th}}$ tend towards unity as $V\to\infty$, the concurrence threshold $c_{\text{th}}$ shows a slower rate of approach. This implies a higher resilience of concurrence connectivity against increase in $V$, a phenomenon we also observe in real-world network topology. Our findings underscore the role of longer paths in QN: despite the exponential decay of entanglement along longer paths, these paths cannot be ignored in concurrence percolation. In fact, longer paths still contribute significantly to QN connectivity—as long as they are abundant. In practice, this principle suggests that an effective design of QN should move beyond a focus solely on the shortest paths to include a strategic incorporation of longer paths as well. This view may open up opportunities to explore advanced quantum communication technologies, such as path routing [13] and network coding [14], in a synergistic manner.

(a)

(b)

Figure 1: Percolation map**s of a quantum network (QN). (a) In a QN, a node comprises several qubits, each entangled with those in other nodes, and a link symbolizes a partially entangled bipartite state $\left|\phi\right\rangle$ shared between two nodes. Entanglement transmission across QN can be understood from two distinct statistical physics map**s: classical percolation (blue) vs. concurrence “quantum” percolation (red). Each map** is based on its own set of path connectivity rules—series and parallel rules—which dictate the “superposition” of multiple path connectivities to determine the total percolation connectivity between two distant nodes, say, Alice (A) and Bob (B). (b) Different connectivity rules result in a varied emphasis on path length in classical vs. concurrence percolations. While in classical percolation, the shortest path between A and B dominates the overall network connectivity, in concurrence percolation, longer paths also contribute to the concurrence connectivity if they are abundant.

Figure 2: ($U$,$V$) flower as hierarchical scale-free network. The ($U$,$V$) flower is a self-similar network, constructed by iteratively replacing each link with a motif comprising two nodes connected by two parallel paths of varying lengths $U\leq V$. The $n$-th generation ($U$,$V$) flower can be fully decomposed into $2^{n}$ non-overlap** paths between A and B, with varying lengths $(U^{n},VU^{n-1},...,V^{n})$ and corresponding count of paths $(C^{0}_{n},C^{1}_{n},...,C^{n}_{n})$. The example shows $U=2$ and $V=3$.

Figure 3: Critical phenomena on ($U$,$V$) flower. (a-e) Critical exponent $\nu$ in classical (blue) and quantum (red) map**s. (f-j) Critical exponents $\beta$ and $d_{f}$.

I Critical Exponents and Hyperscaling

We begin by focusing on ($U$,$V$) flowers [12], a canonical example of hierarchical, scale-free networks. In a ($U$,$V$) flower, the $(n+1)$-th generation is built by replacing each existing link in the $n$-th generation by a basic motif, formed by a shorter path with $U$ links and a longer path with $V$ links between two nodes A and B (Fig. 2). The two scales $U$ and $V$ determine different network characteristics: while the shorter length scale $U$ governs the shortest path length $L=U^{n}$ between A and B, which is asymptotically the same as the diameter of the network [15], the longer length scale $V$ controls the lengths of non-shortest paths, thereby controlling the dimension of the network, $d=\ln N/\ln L$, where $N\sim\left(U+V\right)^{n}$ is the number of nodes in the network [15]. Note that all length scales are established in the chemical (“time”) space, thus differing from traditional Euclidean length scales $\xi$ by an additional dynamic exponent $z$, related by $L\sim\xi^{z}$ [12].

A promising feature of ($U$,$V$) flowers is that they are also series-parallel networks [16]. Consequently, they allow for an analytical calculation of the classical sponge-crossing probability [6, 17], $P_{\text{sc}}$, which is the probability of connection between the two boundary nodes A and B (a.k.a. $\Pi_{\infty}(p;L)$ in boundary conformal field theories [18]). Specifically, $P_{\text{sc}}$ can be calculated through iteratively applying an exact renormalization-group (RG) function that is constructed from series and parallel rules [Fig. 1a]:

$$\mathcal{R}(P_{\text{sc}})=\text{para}(\text{seri}(\stackrel{{\scriptstyle U}}{{\overbrace{P_{\text{sc}},\cdots,P_{\text{sc}}}}}),\text{seri}(\stackrel{{\scriptstyle V}}{{\overbrace{P_{\text{sc}},\cdots,P_{\text{sc}}}}})).$$ (1)

Equation (1) allows us to derive the $n$-th generation $P_{\text{sc}}$ by nesting $\mathcal{R}$ $n$ times, given by $P_{\text{sc}}=\stackrel{{\scriptstyle n}}{{\overbrace{\mathcal{R}(\mathcal{R}(\mathcal{R}(\cdots\mathcal{R}}}}(p))))$. The critical thresholds $p_{\text{th}}$ can be determined by locating the nontrivial fixed point of $p$ that satisfies the RG fixed point equation, $p=\mathcal{R}(p)$. It also means that the sponge-crossing probability $P_{\text{sc}}$ at criticality is equal to $p_{\text{th}}$ when $n\to\infty$.

Similarly, the sponge-crossing concurrence $C_{\text{sc}}$ and the concurrence threshold $c_{\text{th}}$ (hereafter referred to as “quantum” results) can be calculated by replacing $P_{\text{sc}}$ to $C_{\text{sc}}$ in Eq. (1) and the corresponding classical series/parallel rules to their quantum counterparts [Fig. 1a]. For example, in a ($2$,$2$) flower, for concurrence percolation, we find $c_{\text{th}}=0.759\cdots$, or $\theta\approx 0.549\pi/4$, while for classical percolation, $p_{\text{th}}=(\sqrt{5}-1)/2\approx 0.618$ [15], corresponding to $\theta\approx 0.750\pi/4$ [Fig. 3c]. The results reinforce that the quantum threshold is lower than the classical one [5].

Near criticality, the thermal exponent $\nu$ characterizes the divergence of the correlation length as one approaches the critical point. We can obtain an exact expression for $\nu$ by evaluating [15]

$$\frac{\partial R(p)}{\partial p}\big{|}_{p=p_{\text{th}}}=\left[\left(UL\right)/L\right]^{1/\nu}\Rightarrow\nu=\frac{\ln{U}}{\ln\left({\partial R(p)}/{\partial p}\right)}\big{|}_{p=p_{\text{th}}},$$ (2)

where $L$ and $UL$ correspond to the diameters of the $n$-th and $\left(n+1\right)$-th generations of the network, respectively. Compared to the established classical value of $\nu=1.635\cdots$ [15], we find a distinct quantum value of $\nu=1.352\cdots$. These exact values are numerically verified by finite-size scaling analysis: assuming that the finite-size critical threshold $p_{\text{th}}(L)$ deviates from the true threshold $p_{\text{th}}$ by $|p_{\text{th}}(L)-p_{\text{th}}|\sim L^{-1/\nu}$, we find $\nu\approx 1.63528(1)$ for classical and $\nu\approx 1.3530(1)$ for quantum on both sides of the critical threshold of the ($2$,$2$) flower [Figs. 3a,b,d,e] (also see SI, Section IV).

In the traditional study of classical percolation, not only does the length scale but the cluster size distribution shows critical behaviors as well. Notably, the emergence of a giant component when $p\geq p_{\text{th}}$ [19], where the strength $P_{\infty}=N_{g}/N$ represents the giant component’s relative size $N_{g}$ to the total network size $N$, leads to a scaling behavior as $P_{\infty}\sim(p-p_{\text{th}})^{\beta}$ for $p\to p_{\text{th}}^{+}$. This cluster-based concept of $P_{\infty}$ has not been translated into the quantum counterpart, which relies on path connectivity rules and lacks a notion of “clusters.” Nevertheless, we notice that $P_{\infty}$ can be alternatively considered as the probability of a randomly chosen node in the bulk to reach the boundaries of the network [15]. This consideration allows us to redefine $P_{\infty}$, and crucially $C_{\infty}$, in terms of path connectivity from a randomly chosen node (which must be averaged across all nodes) to the boundaries. For the ($U,V$) flowers, the boundaries are represented by A and B. The construction is detailed in SI, Section I.

Using the new definition of $P_{\infty}$, we find an exponent $\beta\approx 0.1688290(1)$ [Fig. 3g], closely matching the cluster-defined value, $\beta=0.165\cdots$ on ($2$,$2$) flowers. (The minor difference arises because computing $P_{\infty}$ requires higher-order path connectivity rules (SI, Section I) that extend beyond the basic series and parallel rules. However, these higher-order rules can be approximated using only the series/parallel rules through a technique known as star-mesh transformations [5], yielding results that are close, but not precise.) In the quantum domain, we observe $\beta\approx 0.0765379(1)$ [Fig. 3f], markedly distinct from its classical counterpart.

Another critical exponent, $d_{f}$, defines the fractal dimension of the giant cluster at the critical point, with $N_{g}\sim L^{d_{f}}$. Again, lacking a “cluster” notion, $N_{g}=NP_{\infty}$ must be redefined through $P_{\infty}$ at the critical point. We find $d_{f}\approx 1.89675835(5)$ [Fig. 3j] in the classical scenario, nearly identical to the cluster-defined value $d_{f}=1.899\cdots.$ In contrast, the quantum scenario shows $d_{f}\approx 1.9434305(5)$ [Fig. 3i].

Promisingly, the three path-defined critical exponents still adhere to the hyperscaling relation, $d_{f}=d-\beta/\nu$ [20, 21], in both classical and concurrence percolation, underscoring a consistent path-connectivity framework across different percolation theories.

II Asymptotic dependence on the longer length scale

To investigate the impact of the longer length scale $V$ on the two percolation theories, we derive the asymptotic behaviors of both critical thresholds $p_{\text{th}}$ and $c_{\text{th}}$ in the $V\to\infty$ limit (SI, Section II), finding that for classical percolation,

$$p_{\text{th}}\simeq 1-\left(\ln\frac{U}{U-1}\right)V^{-1}+O(V^{-2}),$$ (3)

and, in comparison, for concurrence percolation,

$$c_{\text{th}}\simeq 1-\left(\frac{1}{4}\ln V\right)V^{-1}+O(V^{-1}\ln\ln V).$$ (4)

The $V^{-1}$ terms originate from the fact that as $V\to\infty$, loops within the ($U$,$V$) flower extend to infinite lengths, and the network essentially becomes tree-like. As a result, both thresholds should approach $1$ when $V\to\infty$. For classical percolation, the prefactor for $V^{-1}$ depends solely on $U$, signaling the significant role the shorter paths play in classical percolation near the critical point. In contrast, concurrence percolation shows a unique behavior. The presence of a logarithmic prefactor $\left(\ln V\right)/4$ highlights the importance of longer paths in concurrence percolation. This term indicates that connectivity in concurrence percolation near criticality is enhanced compared to its classical counterpart, implying $1-c_{\text{th}}\gg 1-p_{\text{th}}$ as $V\to\infty$. Moreover, this term is fully decoupled from $U$, suggesting that the shorter length scale no longer plays a determining role in $c_{\text{th}}$.

The asymptotic behaviors of the critical exponents are also different. We analytically determine $\nu$ and $d_{f}$ directly via the solution of $P_{\text{sc}}$ and $P_{\infty}$ (cf. $C_{\text{sc}}$ and $C_{\infty}$), and $\beta$ from the hyperscaling relation $d_{f}=d-\beta/\nu$. The results are summarized in Table 1. We find that in classical percolation, all dominant terms governing $\nu$, $d-d_{f}$, and $\beta$ are fully determined by $U$, once again signaling the dominance of shorter paths. In contrast, in concurrence percolation, we notice the emergence of a very slow correction ($\sim\ln\ln V$) affecting the dominant terms in $\nu$ and $d_{f}$. Interestingly, the corrections counterbalance each other in the calculation of $\beta$, resulting in a constant $\beta=2$ that is independent on either $U$ or $V$. Since $\beta$ is intrinsically tied to the definition of an order parameter, we hypothesize that this unique, constant value of $\beta$ reveals an entirely distinct symmetry associated with the order parameter in concurrence percolation on ($U$,$V$) flowers.

Table 1: Critical exponents in the $V\to\infty$ limit.

	Classical [$\sim\!f(U)+O(V^{-1})$]	Quantum [$\sim\!g(U,V)$]
$\nu$	$\frac{\ln U}{\ln\left[1+\left(U-1\right)\ln\frac{U}{U-1}\right]}\!+\!O(\!V^{-1}\!)$	$\frac{2\ln U}{\ln\ln V}\!+\!O\left(\frac{\ln\ln\ln V}{\left(\ln\ln V\right)^{2}}\right)$
$d\!-\!d_{f}\!$	$\frac{\!\ln{U}\!-\!\ln{\left[\left(1\!-\!U\right)+2/\left(\ln\!{\frac{U}{U-1}}\right)\right]}}{\ln{U}}\!+\!O(\!V^{-1}\!)$	$\frac{\ln\ln V}{\ln U}\!+\!O\left(\frac{\ln\ln V}{\ln V}\right)$
$\beta$	$\frac{\!\ln{U}\!-\!\ln{\left[\left(1\!-\!U\right)+2/\left(\ln\!{\frac{U}{U-1}}\right)\right]}}{\ln\left[1+\left(U-1\right)\ln\frac{U}{U-1}\right]}\!+\!O(\!V^{-1}\!)$	$2\!+\!O\left(\frac{\ln\ln\ln V}{\ln\ln V}\right)$

III Superposition of paths in general

The above results have broader implications beyond the specific ($U$,$V$) flower topology. To see this, it is worth noting that the $n$-th generation of a ($U$,$V$) flower could be fully decomposed into several non-overlap** but intersecting paths between A and B. In the language of graph theory, these paths are edge-disjoint, but not vertex-disjoint. Specifically, each of these paths can have a length that is given by one of the options $\{U^{n},VU^{n-1},\cdots,V^{n}\}$. The number of paths corresponding to each length option is determined by the binomial coefficients $\{C^{0}_{n},C^{1}_{n},\cdots,C^{n}_{n}\}$, summing up to $2^{n}$ paths in total.

The exact decomposition goes as follows. Initially, in the $1$-st generation, there are only two non-overlap** paths between A and B, characterized by lengths $U$ and $V$. As progressing to the $2$-nd generation, each link from the previous generation is systematically replaced by the basic motif, necessitating a decision for each link within a previous path: to replace it by either a $U$-length path or a $V$-length path. Consequently, there are two non-overlap** options: replacing all links entirely by $U$-length paths, or entirely by $V$-length paths. This bifurcation results in each earlier generation’s path splitting into two distinct paths. Thus, as a shorthand, each path can be encoded as a string of $n+1$ characters (e.g., “UVVUUV….”), each designated as ‘U’ or ‘V’. This string reflects the sequence of choices made at each generation, with ‘U’ for choosing all $U$-length paths and ‘V’ for $V$-length paths. Each non-overlap** path can be uniquely identified by such a string. The corresponding path length is $U^{n-x}V^{x}$, with $x$ representing the count of ‘V’s in the string. For each possible length, there are $C^{x}_{n}$ such choices, which scale as $C^{x}_{n}\simeq\left(2\pi x\right)^{-1/2}x^{-x}n^{x}e^{x}$ for $x\ll n$, $x\to\infty$, suggesting that longer paths are also exponentially more abundant.

When $V$ adjusts to $tV$, these lengths would adjust to $(U^{n},tVU^{n-1},\cdots,t^{n}V^{n})$, i.e., changed by $(1,t,\cdots,t^{n})$ sequentially, highlighting different growth patterns between shorter and longer paths. On the other hand, the number of these paths remains constant. This allows us to investigate the response to adjustment of the longer length scale $V$ for hierarchical, scale-free networks of more general topologies. Specifically, given a real-world network, we select two hub nodes A and B and non-overlap** (but potentially intersecting) paths between them (SI, Section V). These paths are ranked by their lengths and sequentially placed into different groups, such that each successive group encompasses an exponentially larger number of longer paths. The lengths of these grouped paths are then rescaled by $(1,t,t^{2},\cdots)$, reflecting the adjustment of $V$.

We analyzed a segment of the real-world Internet at the autonomous-system level [11]. By rearranging Eq. (3) and expanding the equation near $\theta_{\text{th}}(V\to\infty)=\pi/4$, we derive $V\left[\theta_{\text{th}}(\infty)-\theta_{\text{th}}^{(p)}(V)\right]\simeq V\left(1-p_{\text{th}}\right)/2\simeq\ln{\left[U/\left(U-1\right)\right]}/2$ as $V\to\infty$, where $\theta_{\text{th}}^{(p)}\equiv\sin^{-1}{\sqrt{p_{\text{th}}/2}}$. Similarly, from Eq. (4) we derive $V\left[\theta_{\text{th}}(\infty)-\theta_{\text{th}}^{(c)}(V)\right]\simeq V\sqrt{\left(1-c_{\text{th}}\right)/2}\simeq\sqrt{V\ln V/8}$ as $V\to\infty$, where $\theta_{\text{th}}^{(c)}\equiv\left(\sin^{-1}{c_{\text{th}}}\right)/2$. The distinct $\theta_{\text{th}}$ dependencies on $V$ (or $t$, assuming $V\sim t$) are illustrated for both the ($U$,$V$) flower [Fig. 4b] and the Internet [Fig. 4c]. The resemblance of the dependencies reveals a striking similarity between the two network topologies, highlighting a shared quantum characteristic: the slower approach of $\theta_{\text{th}}^{(c)}(V)$ towards to $\theta_{\text{th}}^{(c)}(V\to\infty)$, which indicates a higher resilience of concurrence connectivity. This observation underscores the presence of such resilience in general network topologies, characterized by non-overlap** but intersecting paths.

Given the very slow corrections $\sim\ln\ln V$ in the quantum critical exponents (Table 1), this extreme effect becomes eminent only at a very large length scale, far exceeding the Internet diameter. However, if we consider “fundamental” networks, such as spin networks in loop quantum gravity [22], where the length unit is extremely small (the Planck scale $10^{-35}$ m), the corrections could potentially become detectable.

(a)

(b)

(c)

Figure 4: Real-world path connectivity. (a) Paths between two hub nodes A and B in real-world networks (e.g., Zachary’s Karate club). Identifying non-overlap** paths between the hubs enables investigation of how the percolation critical thresholds ($p_{\text{th}}$ vs. $c_{\text{th}}$) depend on non-shortest path lengths (governed by the longer length scale, $V$) in general. (b) $p_{\text{th}}$ and $c_{\text{th}}$ show different scaling behaviors when translating to $\theta_{\text{th}}$ in ($U$,$V$) flowers. As $V$ increases, $p_{\text{th}}$ (blue) approaches unity at a constant rate $\sim V^{-1}$. In contrast, $c_{\text{th}}$ (red) shows a slower rate of approach ($\sim V^{-1}\ln V$). (c) Corresponding patterns are also observed in the Internet network topology [11].

IV Discussion

We have found analytical expressions for the critical thresholds and exponents of both classical percolation and concurrence percolation using the ($U,V$) flowers model, a hierarchical scale-free network structure. Our findings, which place concurrence percolation in a distinct universality class, rule out the possibility that differences between classical and concurrence percolation theories are confined merely to short-range (i.e., ultraviolet) details. Instead, variations also stem from long-range (i.e., infrared) behaviors. This distinction underscores the fact that concurrence percolation represents a fundamentally different statistical framework compared to its classical counterpart. The question of whether concurrence percolation can be described by a field theory akin to the $\phi^{3}$ model [23] for classical percolation remains open. A crucial step in this direction involves defining a cluster-based order parameter (field) specific to concurrence percolation. Our alternate definition of the giant “component” $C_{\infty}$ on ($U,V$) flowers might offer valuable insights for this.

From the perspective of quantum communication, our research indicates that having a path of maximally entangled states in QN is a sufficient, but not necessary criterion. It is possible to achieve entanglement transmission via longer, exponentially weaker paths composed of partially entangled links, provided there is a compensatory increase in the number of such links. This principle finds a parallel in classical branching processes, where infinite connectivity is achievable if the growing number of trees with given size exactly compensates for the exponential decline in the probability of such a tree existing at the critical point [19]. Further exploration of this concept through other models, such as the ($U,V,W$) flowers (with three different length scales), could provide deeper insights on both theoretical and practical aspects of entanglement distribution efficiency in QNs.

V Acknowledgements

S.H. thanks the EU H2020 DIT4Tram (Grant number 953783) and the Horizon Europe grant OMINO (grant number 101086321).