---
title: 'Smith''s Conjecture: Multi-Domain Insights'
url: https://www.emergentmind.com/topics/smith-s-conjecture
type: topic
---

# Smith's Conjecture: Multi-Domain Insights

Searching arXiv for recent papers on “Smith’s Conjecture” across domains to ground the article.
“Smith’s Conjecture” is not a single invariant statement across the arXiv literature represented here. The name appears in at least three distinct technical settings: a graph-theoretic conjecture asserting that in any \(k\)-connected graph any two longest cycles intersect in at least \(k\) vertices [2310.03849]; the Gross–Smith generalization of Hardy–Littlewood prime pair conjectures to algebraic number fields, presented in one exposition as “Smith’s Conjecture” in number fields [2001.09513]; and Llewellyn Smith’s conjecture in quantum field theory, which ties tree-level unitarity bounds to perturbative renormalizability in local, Lorentz-invariant theories with a positive-definite Hilbert space [2210.13666]. This multiplicity of usage suggests that the term is domain-sensitive rather than canonical.

## 1. Principal usages of the name

The following table summarizes the three usages documented in the supplied arXiv material.

| Domain | Conjectural content | arXiv |
|---|---|---|
| Graph theory | Every pair of longest cycles in a \(k\)-connected graph intersect in at least \(k\) vertices | [2310.03849] |
| Algebraic number theory | Prime-element pairs in \(O_K\) at fixed difference \(\eta\) are governed by a singular series \(\mathfrak{S}(\eta)\) | [2001.09513] |
| Quantum field theory | Tree-level unitarity bounds are tied to perturbative renormalizability in positive-metric theories | [2210.13666] |

In graph theory, the paper “On two conjectures about the intersection of longest paths and cycles” treats Smith’s Conjecture as the diagonal case of a family \(C(k,r)\) of cycle-intersection statements:
$$
\text{Conjecture C(k,k):}\quad \text{In any }k\text{-connected graph }G,\text{ any two longest cycles intersect in at least }k\text{ vertices.}
$$
The same paper places this beside Hippchen’s path analogue \(P(k,k)\) and makes the cycle statement the central object [2310.03849].

In algebraic number theory, Gross and Smith formulate a prime-pair conjecture over algebraic number fields, and the supplied exposition explicitly identifies this framework as “Smith’s Conjecture” in number fields. The conjecture is a Hardy–Littlewood-type prediction for prime elements in \(O_K\), with local factors indexed by prime ideals \(\mathfrak p\) and a singular series \(\mathfrak S(\eta)\) depending on a fixed nonzero shift \(\eta\) [2001.09513].

In high-energy theory, the relevant name is more precise: Llewellyn Smith’s conjecture. It concerns the relation between tree-level partial-wave unitarity bounds and perturbative renormalizability. The supplied papers on quadratic gravity analyze its domain of validity and exhibit a counterexample once negative-norm states are present [2210.13666].

## 2. Smith’s Conjecture in graph theory

The graph-theoretic formulation is the most literal use of the bare name in the supplied material. Let \(G\) be a simple graph, let \(n=|V(G)|\), and let \(G\) be \(k\)-connected in the sense that every separator \(S\subseteq V(G)\) has \(|S|\ge k\). If \(C\) and \(D\) are longest cycles in \(G\), Smith’s Conjecture asserts
$$
|V(C)\cap V(D)|\ge k.
$$
The paper formalizes this as \(C(k,k)\) and treats it as the cycle member of a broader family \(C(k,r)\) [2310.03849].

A central structural relation is the reduction
$$
\forall k\ge 1,\ \forall r\le k,\quad C(k+1,r+1)\ \Rightarrow\ P(k,r),
$$
where \(P(k,r)\) is the analogous conjecture for longest paths. The reduction adds a universal vertex \(s\) to a \(k\)-connected graph \(G\), producing a \((k+1)\)-connected graph \(G'\) in which longest paths \(P,Q\) in \(G\) become longest cycles \(P\cup\{s\}\) and \(Q\cup\{s\}\). Removing \(s\) transfers any cycle-intersection lower bound back to a path-intersection lower bound [2310.03849].

The paper proves the following cycle theorem:
$$
\forall k\ge 2,\ \forall n,\quad \bigl|V(C)\cap V(D)\bigr|\ \ge\ \min\{\,k\,,\,8k-n-16\,\}
$$
for any two longest cycles \(C,D\) in a \(k\)-connected graph on \(n\) vertices. A direct corollary is that Smith’s Conjecture holds whenever
$$
k\ge \frac{n+16}{7}.
$$
This confirms the conjecture in a high-connectivity range relative to the order \(n\) [2310.03849].

The same paper records prior progress: for \(2\le k\le 7\), \(C(k,k)\) holds. It also cites Chen et al.’s general lower bound
$$
C\bigl(k,\,c\,k^{3/5}\bigr)\text{ holds for all }k\ge 2,\qquad
c=\bigl(\sqrt[3]{256}+3\bigr)^{-3/5}\approx 0.2615.
$$
Thus the current status described in the supplied literature is twofold: exact intersection at level \(k\) is known for \(k\le 7\) and for \(k\ge (n+16)/7\), while intermediate regimes remain open [2310.03849].

## 3. Proof architecture for the graph-theoretic result

The proof of the bound
$$
|V(C)\cap V(D)|\ge \min\{k,8k-n-16\}
$$
is organized around fan constructions, circumference estimates, and inclusion–exclusion. Let \(C\) and \(D\) be longest cycles of common length \(L\), and write
$$
X:=V(C)\cap V(D).
$$
The argument selects a “short” path \(R\) of length at most \(2\) from a component of \(G-C\) or \(G-D\), and without loss takes \(R\subseteq G-C\), with endpoints \(p,r\). Applying the fan lemma to \(p\) and \(r\) with target \(V(C)\) yields two families \(A\) and \(B\) of internally disjoint paths from \(p\) and \(r\) to \(C\), each of size at least \(k-|R|\), with distinct endpoints on \(C\) [2310.03849].

From the endpoint set \(F\subseteq V(C)\), the proof constructs an auxiliary weighted cycle \(C^*\). The edge weights record arc lengths along \(C\), and the endpoints are colored according to whether they come from \(A\) or \(B\). The resulting estimates are:
$$
w(uv)\ge 2\quad\text{for }uv\in C''=E(C^*)\setminus C',
$$
and
$$
w(uv)\ge 2+|R|\quad\text{for }uv\in C'.
$$
These feed into the circumference lower bound
$$
L=w(C^*)\ge 4\,(k-|R|)+|R|\,|C'|-2\,|F'|.
$$
The bridge to intersection size is the inclusion–exclusion inequality
$$
|X|\ge |V(C)|+|V(D)|-n = 2L-n.
$$
When \(|R|=2\), the fan estimate gives \(L\ge 4(k-2)\), hence
$$
|X|\ge 2L-n\ge 8k-n-16.
$$
When \(|R|\le 1\), a refined analysis of bicolored vertices, covered edges, and components of \(D-C\) yields either \(|X|\ge k\) directly or a fallback to Dirac’s circumference bound. This explains the appearance of the constants \(8\) and \(16\): they come from combining \(L\ge 4(k-2)\) with \( |X|\ge 2L-n\) [2310.03849].

The path analogue follows from the cycle-to-path reduction. The paper deduces
$$
\forall k\ge 2,\ \forall n,\quad \bigl|V(P)\cap V(Q)\bigr|\ \ge\ \min\{\,k\,,\,8k-n-9\,\}
$$
for any two longest paths \(P,Q\) in a \(k\)-connected graph on \(n\) vertices, and therefore confirms Hippchen’s conjecture when \(k\le 6\) or \(k\ge (n+9)/7\) [2310.03849].

## 4. “Smith’s Conjecture” in algebraic number fields: the Gross–Smith framework

In the algebraic-number-theoretic usage, let \(K\) be an algebraic number field with ring of integers \(O_K\), degree \(n=[K:\mathbf Q]\), and Dedekind zeta residue
$$
\mathfrak R_K:=\operatorname{Res}_{s=1}\zeta_K(s).
$$
A prime element \(\omega\in O_K\) means that the principal ideal \((\omega)\) is prime. Fix an integral basis \(\{\alpha_1,\dots,\alpha_n\}\) of \(O_K\), define the embedding \(m:K\to \mathbf R^n\) by coordinates in that basis, and measure short intervals by sets of the form
$$
\{\alpha\in O_K:\ \|m(\alpha)-x\|\le H\}.
$$
The Gross–Smith prime pair conjecture then states that for fixed nonzero \(\eta\in O_K\),
$$
\sum_{\|m(\alpha)\|\le X} \mathbf{1}_{\mathcal P}(\alpha)\,\mathbf{1}_{\mathcal P}(\alpha+\eta)
\;\sim\;
\mathfrak S(\eta)\,\frac{1}{\mathfrak R_K^2}\,
\sum_{\substack{\|m(\alpha)\|\le X\\ |N(\alpha)|>1}}
\frac{1}{\log^2|N(\alpha)|},
$$
as \(X\to\infty\), where \(\mathbf 1_{\mathcal P}\) is the indicator of prime elements [2001.09513].

The singular series is
$$
\mathfrak S(\eta)
=
\prod_{\mathfrak p}
\frac{1-\nu_\eta(\mathfrak p)/N\mathfrak p}{(1-1/N\mathfrak p)^2},
\qquad
\nu_\eta(\mathfrak p)=
\begin{cases}
2 & \text{if }\eta\notin\mathfrak p,\\
1 & \text{if }\eta\in\mathfrak p.
\end{cases}
$$
This is the number-field analogue of the classical Hardy–Littlewood singular series, with prime ideals replacing rational primes and \(N\mathfrak p\) replacing \(p\). The paper also gives a Ramanujan expansion:
$$
\mathfrak S(\eta)
=
\sum_{\mathfrak q\ne 0}\frac{\mu(\mathfrak q)^2}{\phi(\mathfrak q)^2}\,c_{\mathfrak q}(\eta),
$$
where \(c_{\mathfrak q}(\eta)\) is a Ramanujan sum over ideals, \(\mu(\cdot)\) is the Möbius function on ideals, and
$$
\phi(\mathfrak q):=N\mathfrak q \prod_{\mathfrak p\mid \mathfrak q}\left(1-\frac{1}{N\mathfrak p}\right).
$$
The proof strategy for the main analytic theorem does not rely on multiplicativity of singular series outside PIDs; instead it uses Ramanujan expansions, Fourier analysis, lattice counting, and analytic properties of \(\zeta_K(s)\) [2001.09513].

A “nice” smoothing function is a compactly supported \(w:\mathbf R^n\to\mathbf R\) whose Fourier transform satisfies
$$
\widehat w(\xi)=O(|\xi|^{-(n+1)}).
$$
For such \(w\), the principal unconditional theorem is
$$
\sum_{\substack{\eta\in O_K\\ \eta\ne 0}}
\bigl(\mathfrak S(\eta)-1\bigr)\,
w\!\left(\frac{m(\eta)}{H}\right)
\sim
-\,w(0)\,\mathfrak R_K\,\log(H^n),
\qquad \text{error }O(1),
$$
as \(H\to\infty\). In the case \(K=\mathbf Q\) and \(n=1\), this recovers Montgomery’s classical identity with the triangular weight \((1-|x|)_+\) [2001.09513].

## 5. Short-interval prime statistics and the universal variance factor

The same number-field paper uses the Gross–Smith framework to formulate a variance conjecture for prime elements in short intervals. Define
$$
\pi_K(x;H):=\#\{\omega\in O_K\text{ prime}:\ \|m(\omega)-x\|\le H\},
$$
and average over \(x\) uniformly in \(B_X^{\|\cdot\|}\):
$$
E_K(X;H)
:=
\frac{1}{\operatorname{vol}(B_X^{\|\cdot\|})}
\int_{\|x\|\le X}\pi_K(x;H)\,d^n x.
$$
Using Mitsui’s generalized prime number theorem together with log-weighted lattice-sum asymptotics, the paper obtains
$$
E_K(X;H)\sim
\frac{\operatorname{vol}(B_H^{\|\cdot\|})}
{\mathfrak R_K\,\log\operatorname{vol}(B_X^{\|\cdot\|})},
\qquad
\log\operatorname{vol}(B_X^{\|\cdot\|})\sim \log(X^n).
$$
The centered count is
$$
\widetilde{\pi}_K(x;H)
=
\pi_K(x;H)
-\frac{1}{\mathfrak R_K}
\sum_{\substack{\|m(\alpha)-x\|\le H\\ |N(\alpha)|>1}}
\frac{1}{\log|N(\alpha)|},
$$
and the variance proxy is
$$
V_K(X;H)
:=
\frac{1}{\operatorname{vol}(B_X^{\|\cdot\|})}
\int_{\|x\|\le X}\widetilde{\pi}_K(x;H)^2\,d^n x.
$$
The conjectural prediction for \(H=X^\delta\), \(\delta\in(0,1)\), is
$$
V_K(X;H)\sim (1-\delta)\,E_K(X;H),
$$
equivalently
$$
\delta=
\frac{\log\operatorname{vol}(B_H^{\|\cdot\|})}
{\log\operatorname{vol}(B_X^{\|\cdot\|})},
\qquad
V_K(X;H)\sim
\left(1-\frac{\log\operatorname{vol}(B_H^{\|\cdot\|})}{\log\operatorname{vol}(B_X^{\|\cdot\|})}\right)E_K(X;H).
$$
This differs from a Cramér random model prediction,
$$
V_K(X;H)\sim E_K(X;H),
$$
by a universal factor \(1-\delta\), independent of \(K\), the embedding \(m\), and the norm \(\|\cdot\|\) except through the volume ratio entering \(\delta\) [2001.09513].

The heuristic derivation expands the variance, invokes Gross–Smith pair correlations for prime elements, and isolates a main contribution proportional to a smoothed sum of \(\mathfrak S(\eta)-1\). The theorem on smoothed singular-series sums then supplies the term \(-w(0)\mathfrak R_K\log(H^n)\), and the identity \(w(0)H^n=\operatorname{vol}(B_H^{\|\cdot\|})\) yields the correction
$$
-\frac{\log\operatorname{vol}(B_H^{\|\cdot\|})}{\log\operatorname{vol}(B_X^{\|\cdot\|})}E_K(X;H),
$$
hence \(V_K(X;H)\sim(1-\delta)E_K(X;H)\). The cancellation of \(\zeta_K\)-residue and volume constants explains the universality of the factor [2001.09513].

The numerical evidence reported in the paper comes from quadratic fields
\[
\mathbf Q(i),\ \mathbf Q(\sqrt{-3}),\ \mathbf Q(\sqrt{-5}),\ \mathbf Q(\sqrt{-7}),\ \mathbf Q(\sqrt2),\ \mathbf Q(\sqrt3),\ \mathbf Q(\sqrt{10}),
\]
using sup norms on coordinates with respect to the chosen integral basis. For \(X=1000\) and \(\delta\in\{0.1,0.2,\dots,0.9\}\), the computed ratios \(V_K(X;X^\delta)/E_K(X;X^\delta)\) align close to the line \(1-\delta\) in all tested quadratic fields. The paper also records that convergence is slow and that identifying secondary terms remains an open direction [2001.09513].

## 6. Llewellyn Smith’s conjecture in quantum field theory

In quantum field theory, the relevant conjecture is due to Llewellyn Smith and was developed in work of Bell and Cornwall–Levin–Tiktopoulos. In the positive-metric setting, \(S\)-matrix unitarity implies partial-wave bounds
$$
|a_\ell(s)|\le 1,
\qquad\text{equivalently}\qquad
|\operatorname{Re}a_\ell(s)|\le \tfrac12,
$$
for the partial-wave amplitudes
$$
a_\ell(s)=\frac{1}{32\pi}\int_{-1}^1 d\cos\theta\, P_\ell(\cos\theta)\,\mathcal M(s,t(\theta)).
$$
The traditional “tree unitarity” program constrains the high-energy behavior of tree-level \(2\to2\) amplitudes and connects those constraints to perturbative renormalizability in local, Lorentz-invariant theories with a positive-definite Hilbert space [2210.13666].

The quadratic-curvature theory studied in the supplied papers has action
$$
S_g=\int d^4x\,\sqrt{-g}\left(\Lambda+\frac{1}{\kappa^2}R+\alpha R^2+\beta R_{\mu\nu}R^{\mu\nu}\right),
$$
together with a scalar-matter sector
$$
S_m=\int d^4x\,\sqrt{-g}\left(-\frac12 g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi-\frac12 m^2\phi^2-\frac{\lambda}{4!}\phi^4+\xi\phi^2R\right).
$$
On flat background with \(\Lambda=0\), the graviton fluctuation decomposes as
$$
h_{\mu\nu}=H_{\mu\nu}+I_{\mu\nu},
$$
where \(H_{\mu\nu}\) is a massless spin-2 field of positive norm, while \(I_{\mu\nu}\) contains a massive spin-2 field of negative norm and a massive scalar of positive norm. The canonical commutators include
$$
[a_T^{(\sigma)}(\mathbf k),a_T^{(\tau)\dagger}(\mathbf p)]
=
-\delta_{\sigma\tau}\delta^3(\mathbf k-\mathbf p),
$$
which displays the negative norm of the massive spin-2 sector [2210.13666].

The graviton propagator in harmonic gauge is
$$
G_{\alpha\beta,\mu\nu}(p)=
\frac{2i}{\beta\,p^2\,(p^2+m_I^2)}\,P^{(2)}_{\alpha\beta,\mu\nu}
+
\frac{i}{2(3\alpha+\beta)\,p^2\,(p^2+m_S^2)}\,P^{(0)}_{\alpha\beta,\mu\nu},
$$
with
$$
m_I^2=-\frac{1}{\beta\kappa^2},
\qquad
m_S^2=\frac{1}{2\kappa^2(3\alpha+\beta)}.
$$
For graviton–scalar scattering at fixed angle, representative ultraviolet behavior is
$$
h^{(2,e)}\phi\to h^{(2,e)}\phi:
\qquad
\mathcal M(s,t)\sim \kappa^2 k^2\frac{1+\cos\theta}{1-\cos\theta}+\mathcal O(k^0),
$$
with the same leading growth for the odd helicity-2 combination. Since \(s\simeq 4k^2\), this is the familiar \(\kappa^2 s/t\) behavior, and the paper’s UV tables record in particular
\[
h^{(2)}\phi\to h^{(2)}\phi:\ \alpha=2,\qquad
h^{(2)}\phi\to h^{(1)}\phi:\ \alpha=1,\qquad
h^{(1)}\phi\to h^{(1)}\phi:\ \alpha=0.
\]
These amplitudes violate the traditional tree-level unitarity bound in the positive-definite sense [2210.13666].

The central point is that quadratic gravity is nevertheless perturbatively renormalizable, as shown by Stelle. This makes it a counterexample to Llewellyn Smith’s conjecture in its traditional positive-metric form. The resolution proposed in the supplied papers is to replace the standard tree bound by the modified optical theorem appropriate to indefinite-metric theories:
$$
2\,\mathrm{Im}\,(\Psi\to\Psi)
=
\sum_\Phi \eta_\Phi\,\delta^4(p_\Psi-p_\Phi)\,|(\Psi\to\Phi)|^2,
$$
where \(\eta_\Phi=\pm1\) records the norm signature of the intermediate state. At tree level, the inequality
$$
|(\Psi\to\Psi)|
\ge
\mathrm{Im}\,(\Psi\to\Psi)
=
\sum_\Phi \eta_\Phi\,\delta^4(p_\Psi-p_\Phi)\,|(\Psi\to\Phi)|^2
$$
is taken as the analog of the unitarity bound. For forward scattering of a massless helicity-2 graviton and a scalar, the leading \(\mathcal O(E^4)\) terms on the right-hand side cancel between positive-norm massless graviton channels and negative-norm massive spin-2 ghost channels, leaving behavior of order \(E^2\) or softer. The papers therefore conclude that \(S^\dagger S=1\) is respected at tree level in the pseudo-unitary sense even though the traditional tree-level bound fails [2210.13666; 2012.01717].

The authors then propose a new conjecture: in local, Lorentz-invariant quantum field theories with polynomial interactions and an indefinite-metric Hilbert space, the analog of the tree-level unitarity bound appropriate to \(S\)-matrix unitarity gives conditions equivalent to those for perturbative renormalizability. Quadratic gravity is presented as a nontrivial example consistent with that reformulation [2012.01717].

## 7. Comparative status and open directions

Across these three usages, “Smith’s Conjecture” designates structurally different conjectures, but in each case current work combines a precise asymptotic or structural theorem with a remaining conjectural core.

For longest cycles in \(k\)-connected graphs, the open part lies between the verified regimes \(2\le k\le 7\) and \(k\ge (n+16)/7\). The paper explicitly identifies the gap at smaller \(k\) relative to \(n\) and points to tightening the constants \(8\) and \(16\), or finding extremal constructions, as natural next steps [2310.03849].

For the Gross–Smith conjecture over algebraic number fields, the theorem on smoothed sums of singular series is unconditional, while the variance prediction
$$
V_K(X;H)\sim (1-\delta)E_K(X;H)
$$
is heuristic and depends on the prime-pair conjecture together with smoothing approximations. The supplied exposition also identifies several open problems: sharpening error terms and secondary constants, extending smoothing to more general convex regions, obtaining the uniformity in \(\eta\) needed for a rigorous variance derivation, and moving from pair correlations to higher correlations and other prime patterns in \(O_K\) [2001.09513].

For Llewellyn Smith’s conjecture, quadratic gravity shows that the traditional tree-level partial-wave criterion is not equivalent to renormalizability once negative-norm states are allowed. The open direction is therefore not the original conjecture itself, which the papers present as refuted in that broader setting, but the status of the replacement principle asserting equivalence between renormalizability and the pseudo-unitary high-energy constraints derived from the modified optical theorem. The supplied papers verify that principle at tree level for matter–graviton scattering in \(R_{\mu\nu}^2\) gravity, but do not claim a complete proof beyond that scope [2210.13666; 2012.01717].

Taken together, these usages show that “Smith’s Conjecture” functions as a homonymous label rather than a uniquely determined theorem. In graph theory it is a longest-cycle intersection problem; in algebraic number theory it names a Hardy–Littlewood-type prime-pair framework over number fields; and in quantum field theory the relevant statement is Llewellyn Smith’s conjecture on tree unitarity and renormalizability.

Source: https://www.emergentmind.com/topics/smith-s-conjecture