Characterizing k-th powers of Hamilton cycles in G(n,W) via K_{k+1}-fractional covers

Establish that for every fixed integer k ≥ 2 and graphon W, the inhomogeneous random graph G(n,W) is asymptotically almost surely Hamiltonian to the k-th power if and only if: (1) W is connected; (2) the small-degree tail is light, i.e., lim_{alpha→0} μ({x ∈ Ω : deg_W(x) ≤ α})/α = 0; and (3) the only K_{k+1}-fractional cover of W with L1 norm at most 1/(k+1) is the constant function equal to 1/(k+1), where a K_{k+1}-fractional cover is a measurable f: Ω → [0,1] such that for μ^{k+1}-almost every (x1,…,x_{k+1}) either Σ_{i=1}^{k+1} f(x_i) ≥ 1 or at least one edge of the (k+1)-clique is absent, i.e., ∏_{1≤i<j≤k+1} W(x_i,x_j) = 0.

Background

The main theorem gives a characterization for ordinary Hamilton cycles (k = 1) in G(n,W) using connectedness, light small-degree tail, and the absence of peninsulae (a structural obstruction tied to perfect fractional matchings). To extend this to higher powers of Hamilton cycles, the authors draw on the graphon tiling framework and duality via fractional covers for cliques.

They conjecture that the correct generalization replaces the peninsula condition by a condition on K_{k+1}-fractional covers: the only low-norm cover should be the uniform constant value 1/(k+1). Together with connectedness and a light degree tail, this would yield an exact if-and-only-if criterion for the appearance of the k-th power of a Hamilton cycle in G(n,W).

References

Hence, our conjecture, whose case k=1 corresponds to Theorem~\ref{thm:mainHC}, is as follows. Suppose that $W:\Omega2\to[0,1]$ is a graphon, and $k\ge 2$ is fixed. Then $G(n,W)$ contains a.a.s.\ the $k$-th power of a Hamilton cycle if and only if the following three conditions are fulfilled: (i) $W$ is a connected graphon, (ii) we have $\lim_{\alpha\searrow 0}\frac{\mu(D_\alpha(W))}\alpha=0$, and (iii) the only $K_{k+1}$-fractional cover of $W$ of $L1$-norm at most $\frac1{k+1}$ is the constant-$\frac{1}{k+1}$ function.

Hamiltonicity of inhomogeneous random graphs  (2604.00899 - Garbe et al., 1 Apr 2026) in Concluding remarks, Subsection “Other spanning structures in G(n,W)” (Conjecture)