- The paper proves that the appearance of Hamilton ℓ-cycles is triggered exactly when their expected count diverges in r-uniform hypergraphs.
- Using meticulous second moment calculations and small subgraph conditioning, it confirms the conjectured exact threshold by Narayanan and Schacht.
- For Hamilton 2-cycles, the study demonstrates that the normalized cycle count follows a lognormal distribution driven by short subgraph fluctuations.
Exact Thresholds and Lognormal Limits for Non-Linear Hamilton Cycles in Random Hypergraphs
Background and Motivation
The appearance of spanning structures in random graphs and hypergraphs is a central topic in probabilistic combinatorics. The classical threshold results for subgraphs in random graphs (e.g., Bollobás, Ajtai-Komlós-Szemerédi, Komlós-Szemerédi) have been largely extended, yet determining sharp thresholds and limiting distributions for more complex spanning structures in hypergraphs remains challenging. In $r$-uniform hypergraphs, Hamilton $\ell$-cycles generalize the notion of Hamilton cycles: each edge consists of $r$ vertices, and consecutive edges intersect in exactly $\ell$ vertices, with $\ell$ varying from $1$ (loose/linear cycles) to $r-1$ (tight cycles).
Previous work, notably by Narayanan and Schacht [NS:20], established that the first moment threshold is sharp for the existence of non-linear Hamilton cycles but left open whether the threshold is exact in the sense that a Hamilton cycle appears as soon as the expected number just tends to infinity. This paper resolves that conjecture, provides sharp second moment calculations, and, for the case $\ell=2$, characterizes the limiting distribution as lognormal for the normalized count of Hamilton 2-cycles.
Main Results and Theoretical Advances
Sharp Threshold Characterization
For $r > \ell \geq 2$, the paper proves that the threshold for the appearance of a Hamilton $\ell$-cycle in the random $r$-uniform hypergraph $G_r(n, p)$ is precisely when $\mathbb{E}[Z((r))] \to \infty$, where $Z((r))$ is the random variable counting copies of the Hamilton $\ell$-cycle. This confirms the conjecture of Narayanan and Schacht and establishes an exact threshold: as soon as the expectation diverges, the probability of containing such a cycle tends to one.
This is achieved via careful second moment estimates, showing that for $\ell \geq 3$,
$\mathbb{E}[Z((r))^2] = (1+o(1)) \mathbb{E}[Z((r))]^2,$
so standard concentration inequalities (e.g. Chebyshev) are sufficient to deduce whp appearance when expectation diverges.
Lognormal Limiting Distribution for $\ell = 2$
For the special case $\ell = 2$, the normalized count $\frac{Z((r))}{\mathbb{E}[Z((r))]}$ converges in distribution to a lognormal random variable, as the main contribution to the second moment arises from small subgraph fluctuations. This is an explicit characterization:
$\frac{Z((r))}{\mathbb{E}[Z((r))]} \to \mathrm{Lognormal}\left(\mu, \sigma^2\right),$
with the mean and variance specified in terms of constants $A_k$ and explicit series dependent on $c, r, \ell$.
The proof utilizes the small subgraph conditioning method, previously applied in random graphs and statistical physics models [RW:94, J:95, MNS:15], but here adapted for dense hypergraph regimes. The approach involves controlling the effect of short path segments (small subgraphs), which multiplicatively influence the count of Hamilton 2-cycles, and mathematically removing their contribution via exponential correction, yielding sharp concentration and the lognormal limit.
Proof Structure and Technical Highlights
- Overlap Analysis: Detailed decomposition of the overlap structure between pairs of Hamilton cycles, analyzing connected and disconnected intersection subgraphs at various scales (from small path segments to global overlaps).
- Sharp Constant Evaluation: For $\ell \geq 3$ the overlap contributions are negligible except for trivial overlaps, so the second moment matches the square of the first.
- Small Subgraph Conditioning: For $\ell = 2$, the second moment deviation is explicitly attributed to fluctuations in counts of short paths ($\ell$-paths), which are handled by introducing an exponential correction variable $Y$ composed of normalized counts of short segments. This correction is shown to converge to a Gaussian via functional CLT [J:94], leading to the lognormal law for the normalized count.
- Central Limit Theorems and Independence: Independence properties for normalized subgraph counts are exploited, as are the mean and variance calculations under planted and double-planted distributions.
Numerical and Rigorous Outcomes
- Threshold Sharpness: The threshold for Hamilton $\ell$-cycles is shown to be exact: whp appearance coincides with expectation diverging, with rigorous second moment bounds.
- Distributional Limit for $\ell=2$: The count of Hamilton 2-cycles, normalized by expectation, converges to a lognormal distribution with explicit parameters, confirming that the variance is dominated by small subgraph fluctuations, not the full cycle structure.
- Confirmation of Narayanan-Schacht conjecture: The paper's results directly corroborate the conjectured sharp threshold for all $r > \ell > 1$.
Implications and Future Directions
Combinatorial and Probabilistic Theory
The exact threshold and lognormal limiting results deepen the understanding of spanning structures in random hypergraphs, especially distinguishing the behavior for $\ell \geq 3$ and $\ell = 2$. The techniques reinforce the centrality of second moment methods and small subgraph conditioning in threshold analysis, and the explicit lognormal law for cycle counts may inform further distributional studies in random combinatorial structures.
Practical Applications
In algorithmic or applied contexts involving hypergraph sampling or random combinatorial design (e.g. network reliability, random coding theory), knowing precise thresholds for Hamiltonicity directly informs feasibility regimes and expected sampling behavior.
Directions for Further Research
- Extension of these methods to other spanning structures (e.g. hypergraph factors, trees).
- Investigation of dynamical and local thresholds—e.g., hitting times for Hamiltonicity in random hypergraph processes.
- Distributional analysis for other parameter regimes or for cycles with $\ell=1$ (loose cycles), where different combinatorial phenomena arise.
- Asymptotic enumeration and probabilistic estimation in random CSPs (constraint satisfaction problems) with hypergraph structure, leveraging small subgraph conditioning insights.
Conclusion
The paper establishes the exact threshold for the appearance of non-linear Hamilton cycles in random $r$-uniform hypergraphs, with high probability coinciding precisely with the divergence of the expected count. It further characterizes, for the case $\ell=2$, the limiting distribution of the normalized count as lognormal, resolving the previously open question of threshold sharpness and confirming a prominent conjecture. The technical framework and novel adaptation of small subgraph conditioning have broader implications for combinatorial probability and related fields.
Reference: "Exact threshold and limiting distribution for non-linear Hamilton cycles" (2411.13452)
