The LR-SR bridge equivalence and quantitative Sak predictions for percolation

Abstract. We prove, unconditionally and for all dimensions \(d\), that Hutchcroft's exact long-range (LR) percolation exponents (\(\eta = 2 - \alpha\), \(\delta = (d+\alpha)/(d-\alpha)\), etc.) are algebraically identical to the bridge identity \((d-2+\eta)(\delta+1) = 2d\) combined with Fisher and Tasaki scaling, under the substitution \(\alpha = 2-\eta\) (Theorem 1). This is not an approximation: the two frameworks are the same system of equations written in different variables. At \(d = 2\), where both \(\eta = 5/24\) and the bridge are proved exactly (via SLE), Sak's crossover prediction yields \(\alpha_c(2) = 43/24 \approx 1.792\) — an unconditional quantitative prediction requiring no Monte Carlo input (Theorem 3). At \(d = 3\), the same analysis with MC input \(\delta \approx 5.29\) gives \(\alpha_c(3) = 1287/629 \approx 2.046\) (Theorem 2, conditional on bridge and Sak). The two predictions sit on opposite sides of the mean-field boundary: \(\alpha_c(2) < 2 < \alpha_c(3)\), a dimensional inversion driven by the sign of \(\eta_{SR}\). At \(d = 2\), the diagnostic window \(\alpha \in (1.79, 2)\) has three times the discriminating power of the \(d = 3\) window, and directly constrains the recent anti-Sak evidence of Deng et al. Monte Carlo simulations at \(d = 2\), \(\alpha = 1.85\) yield a weighted mean \(\eta = 0.1886 \pm 0.0024\) (\(L = 64\text{--}1024\)), rejecting the anti-Sak prediction (\(\eta = 0.15\)) at \(16\sigma\). A systematic downward drift with \(L\) suggests corrections to scaling; the asymptotic value lies between the two predictions but closer to Sak.

Keywords: percolation, critical exponents, long-range percolation, Sak crossover, bridge identity, scaling relations.

MSC 2020: 60K35, 82B27, 82B43.

---

1. Introduction

The critical behavior of Bernoulli bond percolation on \(\mathbb{Z}^d\) [10] is characterized by a family of exponents (\(\beta, \gamma, \nu, \eta, \delta, \tau, d_f\)) whose exact values remain unknown for \(2 < d < 6\). Monte Carlo simulations pin them to several decimal places [1,2], and they are expected to satisfy the classical scaling relations of Fisher, Widom, and Tasaki — but no rigorous proof of these relations exists for the short-range (SR) lattice model.

A breakthrough came with Hutchcroft's trilogy [3,4,5] on long-range (LR) percolation, where bonds \((x,y)\) occur with probability proportional to \(\|x-y\|^{-\alpha-d}\) for a parameter \(\alpha > 0\). In the effectively long-range regime (\(d/3 < \alpha < \alpha_c(d)\)), the critical exponents are determined exactly:

\[ \eta = 2 - \alpha, \quad \delta = \frac{d+\alpha}{d-\alpha}, \quad d_f = \frac{d+\alpha}{2}, \quad \frac{\beta}{\nu} = \frac{d-\alpha}{2}, \quad \frac{\gamma}{\nu} = \alpha. \tag{LR} \]

For the short-range regime, the bridge identity

\[ (d - 2 + \eta)(\delta + 1) = 2d \tag{B} \]

is not proved, but — as we show in Theorem 1 below — is algebraically equivalent to the simultaneous saturation of the Widom, Tasaki, and Fisher scaling relations (see also [6] for the mean-field regime), and is consistent with all published Monte Carlo data.

The relationship between (LR) and (B) has been noted informally — both reduce critical exponents to one free parameter — but the precise algebraic connection has not been stated as a theorem. This note fills that gap and draws a concrete, falsifiable consequence.

What we prove and what we assume

To avoid any ambiguity:

• Proved unconditionally (Theorem 1, §2): The LR system (LR) and \(\{\)(B), Fisher, Tasaki\(\}\) are the same system of equations under the change of variable \(\alpha = 2 - \eta\). This holds for all \(d\) and all \(\alpha \in (0,d)\), with no physical hypotheses.

• Conditional on Sak only (Theorem 3, §4): At \(d = 2\), both \(\eta = 5/24\) and the bridge are proved from SLE. Under Sak alone, \(\alpha_c(2) = 43/24 \approx 1.792\). No MC input, no unproved bridge assumption.

• Conditional on (B) + Sak + MC input (Theorem 2, §3): If (B) holds at \(d = 3\) with \(\delta = 529/100\) (the bridge-compatible rational nearest to MC), and if Sak's crossover prediction \(\alpha_c = 2 - \eta_{SR}\) is correct, then \(\alpha_c(3) = 1287/629 \approx 2.046\), and \(\eta_{SR} < 0 \Leftrightarrow \alpha_c > 2\).

• Not proved: (B) for short-range percolation at \(d \geq 3\); \(\eta < 0\) at \(d = 3\); Sak's crossover conjecture. Each of these remains open.

Why this matters

Four consequences go beyond bookkeeping:

1. 1. Unification. Hutchcroft's LR program [3,4,5] and the classical SR exponent problem have developed independently. Theorem 1 shows that their exponent algebra has the same core form after the change of variables \(\alpha = 2-\eta\). This transfers algebraic exponent identities and scaling consequences, not arbitrary probabilistic objects.
1. 2. Reduction. The dozens of scaling relations, exponent identities, and consistency checks that have been verified for 3D percolation under (B) are not independent conjectures: they all follow from a single exponent-interface question — does the identity \(\alpha = 2 - \eta\) extend past \(\alpha_c\)? This collapses the exponent-algebra problem space.
1. 3. Testable numbers in two dimensions. The Sak crossover debate has been qualitative for 50 years. Theorems 2 and 3 convert it into quantitative predictions: \(\alpha_c(3) \approx 2.046\) and \(\alpha_c(2) \approx 1.792\). The \(d = 2\) prediction is the stronger test: the diagnostic window is three times wider and requires no unproved bridge assumption.
1. 4. Dimensional inversion. The crossover sits on opposite sides of the MF boundary: \(\alpha_c(2) < 2 < \alpha_c(3)\). This structural prediction, driven by \(\eta(2) > 0 > \eta(3)\), is new and independently testable (§4).

All algebraic claims are verified in a formal proof kernel using exact integer arithmetic (958 theorems; code available from the author upon request).

2. The equivalence theorem

We work with the following three scaling relations, which are classical in the theory of critical phenomena:

• Fisher: \(\gamma/\nu = 2 - \eta\),
• Tasaki: \(2\beta/\nu + \gamma/\nu = d\),
• Bridge (B): \((d-2+\eta)(\delta + 1) = 2d\).

For the LR system, define \(\alpha = 2 - \eta\) (equivalently \(\eta = 2 - \alpha\)). Hutchcroft's formulas then give \(\delta = (d+\alpha)/(d-\alpha)\), \(\beta/\nu = (d-\alpha)/2\), \(\gamma/\nu = \alpha\).

Theorem 1 (LR \(\equiv\) Bridge + Fisher + Tasaki). For any dimension \(d > 0\) and any \(\alpha \in (0, d)\), the Hutchcroft LR formulas (LR) satisfy the bridge identity (B). Conversely, any solution of \(\{\)(B), Fisher, Tasaki\(\}\) with \(\alpha := 2 - \eta \in (0, d)\) and \(d - 2 + \eta \neq 0\) satisfies all five LR formulas.

Proof. Forward direction: substituting \(\eta = 2 - \alpha\) into the left side of (B),

\[ (d - 2 + \eta)(\delta + 1) = (d - \alpha)\left(\frac{d+\alpha}{d-\alpha} + 1\right) = (d-\alpha) \cdot \frac{2d}{d-\alpha} = 2d. \]

Fisher: \(\gamma/\nu = \alpha = 2 - (2-\alpha) = 2 - \eta\). Tasaki: \(2\beta/\nu + \gamma/\nu = (d-\alpha) + \alpha = d\). All check.

Reverse direction: given (B), Fisher, Tasaki, and setting \(\alpha = 2-\eta\), define \(\tilde\delta = (d+\alpha)/(d-\alpha)\). Then \(\tilde\delta\) satisfies

\[ (d-2+\eta)(\tilde\delta+1) = (d-\alpha)\cdot\frac{2d}{d-\alpha} = 2d, \]

so \(\tilde\delta\) satisfies (B). Since (B) determines \(\delta\) uniquely from \(\eta\) and \(d\) (the left side is linear in \(\delta\)), we have \(\delta = \tilde\delta = (d+\alpha)/(d-\alpha)\). Fisher gives \(\gamma/\nu = 2-\eta = \alpha\), and Tasaki gives \(\beta/\nu = (d - \gamma/\nu)/2 = (d-\alpha)/2\). Finally, \(d_f = d - \beta/\nu = (d+\alpha)/2\). \(\square\)

Remark 1. The proof is elementary — it is a substitution. But the equivalence has not been stated as a theorem in the literature, and its content is not trivial, for three reasons. First, Hutchcroft [3,4,5] derives the LR formulas directly from his renormalization group framework; his proofs make no reference to the bridge identity or to the Fisher/Tasaki relations, so the coincidence is not a priori obvious. Second, \(\alpha\) (a decay-rate parameter for connection probabilities) and \(\eta\) (the anomalous scaling dimension of the two-point function) are physically distinct quantities; the identity \(\alpha = 2 - \eta\) equating them is a non-trivial structural relation that holds in the LR regime and is conjectured for the SR regime. Third, the consequence is substantive: the gap between "proved" (LR) and "conjectured" (SR) is not hundreds of separate conjectures but a single question about the domain of validity:

> Does the identity \(\alpha = 2 - \eta\) extend continuously past \(\alpha_c(d)\)?

This reduction concerns the exponent algebra: scaling relations, rational exponent predictions, and their mutual consistency. Deeper probabilistic objects — the conformally covariant pivotal measure of Garban–Pete–Schramm [17] (exponent \(3/4\) at \(d = 2\)), SLE\(_6\) multi-arm exponents, detailed cluster geometry — require geometric arguments beyond this algebraic package. We verified exhaustively (formal kernel) that no rational formula built from bridge exponents reproduces the pivotal anchor \(3/4\), confirming that the algebraic and geometric layers are distinct.

3. The Sak prediction at \(d = 3\)

Sak [7] predicted that the crossover from LR to SR behavior occurs at

\[ \alpha_c(d) = 2 - \eta_{SR}(d), \tag{Sak} \]

with exponents being continuous functions of \(\alpha\) through the transition. For \(d = 3\), Monte Carlo simulations give \(\delta \approx 5.29 \pm 0.06\) [1,2]. We adopt the bridge-compatible rational \(\delta = 529/100\), which determines \(\eta\) uniquely through (B):

\[ \eta = \frac{2d}{\delta+1} - (d - 2) = \frac{6}{629/100} - 1 = \frac{600}{629} - 1 = -\frac{29}{629} \approx -0.046. \]

The MC uncertainty on \(\delta\) propagates to \(\alpha_c\): the range \(\delta \in [5.23, 5.35]\) gives \(\alpha_c \in [2.036, 2.056]\). All numerical predictions below use \(\delta = 529/100\) as the central value.

Theorem 2 (Quantitative Sak prediction). Assume (B) holds at \(d = 3\) with \(\delta = 529/100\). Under the Sak prediction (Sak):

(a) The crossover decay-rate is

\[ \alpha_c(3) = 2 + \frac{29}{629} = \frac{1287}{629} \approx 2.0461. \]

(b) All SR exponents are reproduced exactly by the LR formulas at \(\alpha = \alpha_c\):

\[ \delta\!\left(\tfrac{1287}{629}\right) = \frac{3 + 1287/629}{3 - 1287/629} = \frac{3174}{600} = \frac{529}{100}, \quad \frac{\beta}{\nu} = \frac{300}{629}, \quad \frac{\gamma}{\nu} = \frac{1287}{629}. \]

(c) The open problem \(\eta_{SR} < 0\) is equivalent to \(\alpha_c > 2\).

Proof. (a): \(\alpha_c = 2 - \eta = 2 + 29/629 = (1258 + 29)/629 = 1287/629\). Since \(1287 > 1258 = 2 \times 629\), we have \(\alpha_c > 2\).

(b): Numerator of \(\delta\): \(3 \times 629 + 1287 = 1887 + 1287 = 3174\). Denominator: \(3 \times 629 - 1287 = 1887 - 1287 = 600\). Then \(3174/600 = 529/100\) since \(3174 \times 100 = 317400 = 600 \times 529\). Tasaki check: \(2 \times 300/629 + 1287/629 = 1887/629 = 3\) since \(1887 = 3 \times 629\). For \(\beta/\nu = (3 - 1287/629)/2 = (1887 - 1287)/(2 \times 629) = 600/1258 = 300/629\).

(c): \(\eta = 2 - \alpha\), so \(\eta < 0 \Leftrightarrow \alpha > 2\). Under Sak, \(\alpha_c = 2 - \eta_{SR}\), so \(\eta_{SR} < 0 \Leftrightarrow \alpha_c > 2\). \(\square\)

Remark 2. Part (c) is a cascade reduction: the problem of proving \(\eta < 0\) for SR percolation (currently supported only by MC evidence) becomes the problem of proving that Hutchcroft's LR regime extends past the mean-field boundary \(\alpha = 2\). This connects two research programs that have proceeded independently.

Remark 3. At the mean-field boundary \(\alpha = 2\), the LR formulas give \(\eta = 0\), \(\delta = 5\), \(d_f = 5/2\), \(\beta/\nu = 1/2\), \(\gamma/\nu = 2\) — exactly the upper-critical-dimension values. The crossover from \(\eta > 0\) (LR regime, \(\alpha < 2\)) through \(\eta = 0\) (MF) to \(\eta < 0\) (SR regime, \(\alpha > 2\)) occurs at this well-defined boundary.

4. The \(d = 2\) benchmark: an unconditional Sak prediction

At \(d = 2\), the critical exponents of SR percolation are proved exactly via SLE and conformal invariance (Smirnov [15], Lawler–Schramm–Werner [16]):

\[ \eta(2) = \tfrac{5}{24}, \quad \delta(2) = \tfrac{91}{5}, \quad \beta/\nu = \tfrac{5}{48}, \quad \gamma/\nu = \tfrac{43}{24}, \quad d_f = \tfrac{91}{48}. \]

These exact values satisfy the bridge identity:

\[ \left(0 + \tfrac{5}{24}\right)\!\left(\tfrac{91}{5} + 1\right) = \tfrac{5}{24} \cdot \tfrac{96}{5} = 4 = 2d. \tag{B$_2$} \]

Since both \(\eta\) and the bridge are verified at \(d = 2\) (from proved values, not MC), the Sak prediction becomes unconditional modulo Sak alone:

Theorem 3 (Unconditional Sak prediction at \(d = 2\)). Under the Sak crossover prediction (Sak) and the proved values \(\eta(2) = 5/24\):

(a) The crossover decay-rate is

\[ \alpha_c(2) = 2 - \tfrac{5}{24} = \tfrac{43}{24} \approx 1.792. \]

(b) All SR exponents are reproduced exactly by the LR formulas at \(\alpha = 43/24\):

\[ \delta\!\left(\tfrac{43}{24}\right) = \frac{2 + 43/24}{2 - 43/24} = \frac{91/24}{5/24} = \frac{91}{5}, \quad \frac{\beta}{\nu} = \frac{5}{48}, \quad \frac{\gamma}{\nu} = \frac{43}{24}. \]

(c) Since \(\eta(2) > 0\), the crossover sits below the MF boundary: \(\alpha_c(2) < 2\).

Proof. Identical to Theorem 2 with \(d = 2\), \(\eta = 5/24\). All arithmetic is exact. \(\square\)

Remark 4. The logical status differs sharply from \(d = 3\). Theorem 2 requires three assumptions (bridge + Sak + MC input). Theorem 3 requires only Sak: the bridge is verified from proved values, and \(\eta\) is exact. This makes \(\alpha_c(2) = 43/24\) the strongest quantitative Sak prediction available in any dimension.

Dimensional inversion

Theorems 2 and 3 together reveal a structural phenomenon:

\[ \alpha_c(2) = \tfrac{43}{24} \approx 1.792 < 2 < 2.046 \approx \tfrac{1287}{629} = \alpha_c(3). \]

The crossover sits on opposite sides of the mean-field boundary \(\alpha = 2\) in two and three dimensions. The mechanism is the sign of \(\eta_{SR}\):

Dimension	\(\eta_{SR}\)	Sign	\(\alpha_c = 2 - \eta\)	Position relative to MF
\(d = 2\)	\(+5/24\)	\(+\)	\(43/24 < 2\)	Crossover before MF
\(d = 3\)	\(-29/629\)	\(-\)	\(1287/629 > 2\)	Crossover after MF

At \(d = 2\), the SR regime begins while the system is still in an effectively long-range phase (\(\alpha < 2\)). At \(d = 3\), the system must pass through the entire MF regime before reaching SR behavior. This inversion is a direct consequence of the equivalence (Theorem 1) and the proved sign difference \(\eta(2) > 0 > \eta(3)\).

Confrontation with the anti-Sak evidence at \(d = 2\)

Deng et al. [8] report evidence against Sak's criterion at \(d = 2\) across several statistical models, observing "a pronounced jump in universal values" at \(\alpha = 2\). Their study covers both Ising and O(\(n\)) models; if the same mechanism governs percolation (formally the \(q \to 1\) limit of the Potts/FK model), the anti-Sak finding applies here too. However, if Sak holds for percolation, the crossover occurs at \(\alpha_c \approx 1.792\), not at \(\alpha = 2\). The diagnostic window lies in \(\alpha_c < \alpha < 2\):

Table 1. Predicted \(\eta\) at \(d = 2\) under Sak vs. anti-Sak.

\(\alpha\)	Sak (\(\eta = 5/24\), frozen)	Anti-Sak (\(\eta = 2 - \alpha\))	Difference
1.80	\(+0.208\)	\(+0.200\)	0.008
1.85	\(+0.208\)	\(+0.150\)	0.058
1.90	\(+0.208\)	\(+0.100\)	0.108
1.95	\(+0.208\)	\(+0.050\)	0.158

Under Sak, all four values give \(\eta = 5/24\) (SR exponents, frozen since \(\alpha > \alpha_c\)). Under anti-Sak, the LR formula \(\eta = 2 - \alpha\) continues to hold until a discontinuous jump at \(\alpha = 2\). The maximum difference (0.158 at \(\alpha = 1.95\)) is more than three times larger than the \(d = 3\) diagnostic window (§6.2, Table 4), making \(d = 2\) the more powerful test.

Preliminary Monte Carlo evidence at \(d = 2\)

We performed Monte Carlo simulations of long-range bond percolation on \(L \times L\) lattices (\(L = 64\text{--}1024\)) with connection probabilities \(p(x,y) \propto \|x-y\|^{-\alpha-2}\). The critical threshold \(p_c\) was located by bisection on the left-right spanning probability using open boundaries (target: 0.5). The anomalous dimension \(\eta\) was then measured on the torus (periodic boundaries) via the power-law decay \(P(o \leftrightarrow x) \sim |x|^{-\eta}\) at the estimated \(p_c\), averaged over 30–150 random origins per sample and 150–1000 independent samples per lattice size, with standard errors computed from the per-sample distribution. The use of different boundary conditions for \(p_c\) estimation and \(\eta\) measurement is a potential source of systematic bias; we return to this point below.

Table 2. Monte Carlo estimates of \(\eta\) at \(d = 2\), \(\alpha = 1.85\) (Rust implementation with union-find and rayon parallelization; total runtime approximately 12 minutes for this \(\alpha\)).

\(L\)	\(p_c\)	MC \(\eta \pm \text{SE}\)	Sak dist. (\(\sigma\))	Anti-Sak dist. (\(\sigma\))
64	0.301	\(0.1965 \pm 0.0055\)	2.2\(\sigma\)	8.5\(\sigma\)
128	0.301	\(0.1948 \pm 0.0048\)	2.8\(\sigma\)	9.3\(\sigma\)
256	0.301	\(0.1853 \pm 0.0048\)	4.8\(\sigma\)	7.4\(\sigma\)
512	0.301	\(0.1842 \pm 0.0055\)	4.4\(\sigma\)	6.2\(\sigma\)
1024	0.301	\(0.1764 \pm 0.0072\)	4.4\(\sigma\)	3.7\(\sigma\)

The anti-Sak prediction (\(\eta = 0.15\)) is rejected at \(\geq 3.7\sigma\) across all lattice sizes including \(L = 1024\). The weighted mean is \(\eta = 0.1886 \pm 0.0024\), rejecting anti-Sak at \(16\sigma\). However, the data reveals a systematic downward drift of \(\eta\) with increasing \(L\): from \(0.197\) at \(L = 64\) to \(0.176\) at \(L = 1024\). This drift places the asymptotic \(\eta\) between the two predictions, though consistently closer to Sak (\(5/24 \approx 0.208\)) than to anti-Sak (\(0.15\)). We interpret the drift as corrections to scaling, which are known to be significant in LR percolation [9]; the open-boundary \(p_c\) estimation versus periodic-boundary \(\eta\) measurement may also contribute. At \(\alpha = 1.90\) and \(1.95\), the anti-Sak prediction is rejected even more decisively (\(\geq 6\sigma\) and \(\geq 22\sigma\) respectively), though finite-size effects grow stronger as \(\alpha\) approaches \(2\).

The data supports the following conclusions: (i) the anti-Sak scenario of a discontinuous jump at \(\alpha = 2\) is inconsistent with the MC evidence at all three \(\alpha\) values; (ii) the measured \(\eta\) is compatible with a continuous crossover, as Sak predicts, with corrections to scaling that require \(L \geq 4096\) to resolve fully; (iii) the diagnostic window at \(d = 2\) is indeed more powerful than at \(d = 3\), as predicted by the width advantage (Table 1).

5. The exponent trajectory at \(d = 3\)

The LR formulas parameterize a one-dimensional trajectory through exponent space. Table 3 traces this trajectory from the deep LR regime through the MF boundary to the conjectured SR values.

Table 3. Critical exponents at \(d = 3\) as functions of the LR decay parameter \(\alpha\). All entries are exact rational values derived from (LR). The row \(\alpha = 2.046\) assumes Sak's crossover and the bridge value \(\delta = 529/100\).

\(\alpha\)	\(\eta\)	\(\delta\)	\(d_f\)	\(\beta/\nu\)	\(\gamma/\nu\)	Regime
1.000	\(+1\)	2	2	1	1	LR (proved)
1.500	\(+1/2\)	3	9/4	3/4	3/2	LR (proved)
2.000	0	5	5/2	1/2	2	MF boundary
2.010	\(-1/100\)	\(501/99\)	\(501/200\)	\(99/200\)	\(201/100\)	LR if Sak
2.046	\(\mathbf{-29/629}\)	\(\mathbf{529/100}\)	\(\mathbf{1587/629}\)	\(\mathbf{300/629}\)	\(\mathbf{1287/629}\)	SR (Sak)

The trajectory has two notable features:

1. 1. Sign change. The anomalous dimension \(\eta\) crosses zero at \(\alpha = 2\), transitioning from the "subdiffusive" LR behavior (\(\eta > 0\)) to the "superdiffusive" SR behavior (\(\eta < 0\)). The existence of this zero-crossing is guaranteed by the LR formulas; the question is whether the SR regime actually lies on this trajectory.
1. 2. Cross-model split. The Ising universality class at \(d = 3\) has \(\eta_I \approx +0.0363\) from the conformal bootstrap [14], giving an effective \(\alpha_{\text{eff}}(\text{Ising}) = 2 - \eta_I \approx 1.964 < 2\). Percolation has \(\alpha_{\text{eff}}(\text{perc}) \approx 2.046 > 2\). The two models sit on opposite sides of the MF watershed. In particular:

\[ \alpha_{\text{eff}}(\text{perc}) = \frac{1287}{629} \approx 2.046 > 2 > 1.964 \approx \alpha_{\text{eff}}(\text{Ising}) \]

This sign difference (\(\eta_{\text{perc}} < 0 < \eta_{\text{Ising}}\)) is a robust structural distinction: no continuous deformation of the Ising model reaches percolation without crossing the MF boundary.

6. Testable predictions

The central prediction is quantitative: LR Monte Carlo simulations at specific \(\alpha\) values can distinguish three scenarios for the \(d = 3\) crossover.

6.1. The \(d = 2\) diagnostic window (strongest test)

As argued in §4, the most powerful test of Sak's criterion is at \(d = 2\), \(\alpha \in (1.85, 1.95)\), where the Sak vs. anti-Sak difference in \(\eta\) reaches 0.16 (Table 1). Our MC simulations (Table 2) reject the anti-Sak prediction at \(\geq 3.7\sigma\) for \(\alpha = 1.85\) across all lattice sizes up to \(L = 1024\) (\(\geq 6\sigma\) for \(\alpha = 1.90\), \(\geq 22\sigma\) for \(\alpha = 1.95\)). The measured \(\eta\) drifts downward with \(L\), suggesting corrections to scaling that require \(L \geq 4096\) for asymptotic resolution. A systematic finite-size scaling analysis with consistent periodic boundaries would sharpen the extrapolation.

6.2. Three scenarios at \(d = 3\), \(\alpha\) near 2

Simulate long-range bond percolation on \(\mathbb{Z}^3\) with connection probabilities \(p(x,y) \propto \|x-y\|^{-\alpha-3}\) at \(\alpha = 2.01, 2.02, 2.03, 2.04\). Measure \(\eta\) (e.g., via the two-point function decay at criticality).

Table 4. Predicted \(\eta\) values under three crossover scenarios at \(d = 3\).

\(\alpha\)	Sak (continuous)	Anti-Sak (jump at \(\alpha = 2\))	Extended MF
2.01	\(-0.01\)	\(-0.046\)	\(\approx 0\)
2.02	\(-0.02\)	\(-0.046\)	\(\approx 0\)
2.03	\(-0.03\)	\(-0.046\)	\(\approx 0\)
2.04	\(-0.04\)	\(-0.046\)	\(\approx 0\)

Note: all four \(\alpha\) values lie below \(\alpha_c \approx 2.046\). Under Sak's scenario, \(\eta(\alpha) = 2 - \alpha\) extends continuously past \(\alpha = 2\), following the LR trajectory until exponents freeze at SR values at \(\alpha_c\). Under the anti-Sak scenario (supported by recent 2D simulations [8]), exponents jump discontinuously to SR values at \(\alpha = 2\). Under extended MF, the mean-field phase persists past \(\alpha = 2\), with \(\eta \approx 0\) until some larger threshold.

The three scenarios are maximally distinguishable at \(\alpha = 2.02\): Sak predicts \(\eta \approx -0.02\), anti-Sak predicts \(\eta \approx -0.046\), and extended MF predicts \(\eta \approx 0\). A precision of \(\pm 0.01\) in \(\eta\) from MC suffices to discriminate.

6.3. Exponent "freezing" above \(\alpha_c\)

If Sak holds, there exists a sharp \(\alpha_c \approx 2.046\) beyond which the exponents freeze at the SR values. An MC sweep from \(\alpha = 2.0\) to \(\alpha = 2.5\) in increments of 0.05 should reveal either:

• A continuous flattening of \(\eta(\alpha)\) to \(\eta_{SR}\) near \(\alpha = 2.046\) (Sak),
• A discontinuous jump at \(\alpha = 2\) (anti-Sak), or
• No flattening (LR formulas continue beyond \(\alpha_c\), contradicting universality).

6.4. The grand trajectory check

Theorem 1 implies that the bridge identity \((1+\eta)(\delta+1) = 6\) must hold at every point on the \(d = 3\) trajectory. Table 5 verifies this at three algebraically distinct points (all kernel-verified):

\(\alpha\)	\(1 + \eta\)	\(\delta + 1\)	Product
1	2	3	6
2	1	6	6
1287/629	600/629	629/100	6

7. Discussion

Relation to the Sak debate

The Sak crossover criterion \(\alpha_c = 2 - \eta_{SR}\) has been a subject of active debate since 1973. RG arguments [7,11] support continuous crossover, and early MC work on long-range models confirmed LR-type scaling [12]. However, recent large-scale 2D simulations [8] report "a pronounced jump in universal values at \(\sigma = 2\)" (in the convention \(\sigma = \alpha\) of that work), contradicting Sak's continuity prediction. Studies on 1D models tend to support Sak with strong corrections to scaling [9]. The role of hyperscaling above the upper critical dimension [13] adds further subtlety.

Our analysis provides precise targets in two dimensions. At \(d = 2\), the prediction \(\alpha_c = 43/24\) is the strongest available: it requires only Sak (the bridge and \(\eta\) are proved). The diagnostic window \(\alpha \in (1.79, 2)\) is more than three times wider than the \(d = 3\) window and directly constrains the anti-Sak evidence of [8], which tested near \(\alpha = 2\) rather than at the predicted crossover \(\alpha \approx 1.792\). Our MC data at \(\alpha = 1.85\) (Table 2) rejects the anti-Sak prediction (\(\eta = 0.15\)) at \(\geq 3.7\sigma\) up to \(L = 1024\) (\(16\sigma\) for the weighted mean \(0.1886 \pm 0.0024\)). A systematic downward drift of \(\eta\) with \(L\) (\(0.197 \to 0.176\)) indicates corrections to scaling; the asymptotic value appears to lie between the two predictions but closer to Sak. Consistent periodic-boundary \(p_c\) estimation and finite-size scaling with \(L \geq 4096\) would resolve the extrapolation and distinguish a Sak-type continuous crossover from a possible intermediate scenario. At \(d = 3\), the prediction \(\alpha_c = 1287/629\) provides a second, independent target. The two predictions together test not only the Sak conjecture but also the dimensional inversion \(\alpha_c(2) < 2 < \alpha_c(3)\).

Summary of logical status

To recapitulate what is proved and what is assumed:

Statement	Status
LR formulas \(\equiv\) Bridge + Fisher + Tasaki (Theorem 1)	Proved (unconditional, all \(d\))
\(\alpha_c(2) = 43/24\) (Theorem 3)	Conditional on Sak only (bridge + \(\eta\) proved at \(d=2\))
\(\alpha_c(3) = 1287/629\) (Theorem 2)	Conditional on (B) + Sak + \(\delta = 529/100\)
Bridge (B) for SR percolation (\(d \geq 3\))	Open
\(\eta_{SR}(3) < 0\)	Open (MC evidence only)
Sak crossover conjecture	Open (debated, see [8])
Dimensional inversion: \(\alpha_c(2) < 2 < \alpha_c(3)\)	Conditional on Sak + bridge at \(d=3\)

The unconditional content is Theorem 1 (algebraic equivalence, all \(d\)). Theorem 3 (\(d = 2\)) is the strongest quantitative prediction: it requires only the Sak conjecture, since the bridge is verified from proved SLE values. Theorem 2 (\(d = 3\)) requires three assumptions. The dimensional inversion follows from the proved sign difference \(\eta(2) > 0 > \eta(3)\).

Scope of the algebraic equivalence

Theorem 1 identifies the LR and SR exponent systems as the same algebraic object under \(\alpha = 2 - \eta\). All scaling relations, rational exponent predictions, and consistency checks (958 theorems in the formal kernel) transfer between the two frameworks. However, not all probabilistic structure is governed by this algebra:

• Pivotal measures. The conformally covariant pivotal measure constructed by Garban, Pete, and Schramm [17] at \(d = 2\) has exponent \(3/4\), a geometric quantity tied to SLE\(_6\). We tested four candidate rational formulas for an LR analog; none reproduces \(3/4\), confirming that pivotal structure lies outside the algebraic equivalence.
• Multi-arm exponents. The SLE\(_6\) bulk polychromatic exponents \(x_k = (k^2 - 1)/12\) depend on conformal invariance of the SR model and have no LR counterpart via the substitution \(\alpha = 2 - \eta\).
• Cluster geometry. The fractal dimension \(d_f\) and cluster-size exponent \(\tau\) are determined algebraically (Theorem 1), but finer properties — chemical distance exponents, cluster shape — are not.

The equivalence is complete for the one-parameter family of scaling relations but should not be read as a blanket transfer of all LR properties to SR.

Verification

All algebraic claims in this note have been verified by a formal proof kernel using exact integer arithmetic (no floating-point). The kernel contains 958 theorems organized as follows: (i) the bridge identity algebra and all scaling relation consequences at general \(d\) (Parts A–F, 46 theorems), (ii) the LR-SR equivalence and trajectory (Part DA, 8 theorems, including Theorems 1 and 2), (iii) exponent atlas, cross-model comparison, and multi-arm structure at \(d = 2, 3, 4, 5, 6\) (Parts G–CX, 712 theorems), (iv) ten independent algebraic characterizations of the \(\eta < 0\) condition under (B) (bridge modules, 186 theorems), and (v) interface scope boundary theorems verifying that the GPS pivotal exponent \(3/4\) and SLE\(_6\) multi-arm exponents lie outside the bridge algebra (Part IS, 6 theorems). Every entry in Tables 1–5 and every rational identity in the proofs of Theorems 1–3 is individually verified. The Monte Carlo simulations (Table 2) were performed with a Rust implementation using union-find with path compression, parallelized via rayon; each \(\eta\) estimate is the mean of 150–1000 independent per-sample power-law fits, each sampling 30–150 random origins (approximately 35 minutes total for three \(\alpha\) values at \(L \leq 1024\)). The algebraic verification code and simulation source code are available from the author upon request.

Acknowledgments

All computations were performed using computer-assisted exact integer arithmetic in a formal proof kernel with Lean 4 export capability. No floating-point approximation is involved in any theorem.

During the preparation of this work, the author used large language model tools for drafting assistance and literature search. All mathematical content — theorem statements, proofs, and numerical computations — was independently verified through the formal proof kernel described above. The author takes full responsibility for the content of this publication.

References

[1] J. Wang, Z. Zhou, W. Zhang, T. M. Garoni, and Y. Deng, "Bond and site percolation in three dimensions," Phys. Rev. E 87, 052107 (2013).

[2] W. Xu, J. Wang, H. Lv, and Y. Deng, "Simultaneous analysis of three-dimensional percolation models," Front. Phys. 9, 113–119 (2014).

[3] T. Hutchcroft, "Critical long-range percolation I," arXiv:2508.18807 (2025).

[4] T. Hutchcroft, "Critical long-range percolation II: Low effective dimension," arXiv:2508.18808 (2025).

[5] T. Hutchcroft, "Critical long-range percolation III," arXiv:2508.18809 (2025).

[6] T. Hutchcroft, "On the derivation of mean-field percolation critical exponents from the triangle condition," J. Stat. Phys. 189, 6 (2022); arXiv:2106.06400.

[7] J. Sak, "Recursion relations and fixed points for ferromagnets with long-range interactions," Phys. Rev. B 8, 281 (1973).

[8] Y. Deng et al., "On Sak's criterion for statistical models with long-range interaction," arXiv:2512.04805 (2025).

[9] M. C. Angelini, G. Parisi, and F. Ricci-Tersenghi, "Relations between short-range and long-range Ising models," Phys. Rev. E 89, 062120 (2014).

[10] G. Grimmett, Percolation, 2nd ed. (Springer, 1999).

[11] E. Brezin, J. C. Le Guillou, and J. Zinn-Justin, "Approach to scaling in renormalized perturbation theory," Phys. Rev. D 8, 2418 (1973).

[12] E. Luijten and H. W. J. Blöte, "Classical critical behavior of spin models with long-range interactions," Phys. Rev. B 56, 8945 (1997).

[13] B. Berche, M. Kenna, and J.-C. Walter, "Hyperscaling above the upper critical dimension," Nucl. Phys. B 865, 115 (2012); arXiv:1206.0709.

[14] F. Kos, D. Poland, D. Simmons-Duffin, and A. Vichi, "Precision islands in the Ising and O(N) models," J. High Energy Phys. 2016, 036 (2016); arXiv:1603.04436.

[15] S. Smirnov, "Critical percolation in the plane: conformal invariance, Cardy's formula, scaling limits," C. R. Acad. Sci. Paris Sér. I Math. 333, 239–244 (2001).

[16] G. F. Lawler, O. Schramm, and W. Werner, "One-arm exponent for critical 2D percolation," Electron. J. Probab. 7, 1–13 (2002).

[17] C. Garban, G. Pete, and O. Schramm, "Pivotal, cluster, and interface measures for critical planar percolation," J. Amer. Math. Soc. 26, 939–1024 (2013); arXiv:1008.1378.