---
title: Berge Hamilton Cycle in Hypergraphs
url: https://www.emergentmind.com/topics/berge-hamilton-cycle
type: topic
---

# Berge Hamilton Cycle in Hypergraphs

A Berge Hamilton cycle is a Hamiltonian Berge cycle in a hypergraph: an alternating sequence of distinct vertices and distinct hyperedges
\[
(v_1,e_1,v_2,e_2,\dots,v_n,e_n)
\]
such that each \(e_i\) contains \(v_i\) and \(v_{i+1}\) with indices taken cyclically, and the vertex sequence uses every vertex of the hypergraph. In the modern literature this is the earliest and most classical hypergraph analogue of a graph cycle, and it remains central because it is flexible enough to interact with extremal, probabilistic, decomposition, and Ramsey-theoretic methods while still retaining a genuinely spanning notion of Hamiltonicity [1403.7932][1903.09057].

## 1. Definition, formalizations, and relation to other cycle notions

The Berge notion admits several equivalent formalisms. In the general Berge-\(G\) framework, a hypergraph \(\mathcal H\) is a Berge-\(G\) if there is a bijection \(f:E(G)\to E(\mathcal H)\) such that \(e\subset f(e)\) for every graph edge \(e\in E(G)\). Specializing to \(G=C_n\) yields a Berge cycle of length \(n\); when that cycle spans the full vertex set, one obtains a Berge Hamilton cycle [1505.08127].

A recurring distinction is between ordinary Berge cycles and weak Berge cycles. In a weak Hamiltonian Berge cycle the hyperedges in the alternating sequence are not required to be distinct, whereas in a Berge Hamilton cycle they are. This distinction is immaterial in ordinary graphs but structurally significant in hypergraphs, particularly in threshold and stopping-time results [1809.03596].

Berge Hamilton cycles should also be separated from \(\ell\)-cycles. Every \(\ell\)-cycle is a Berge cycle, and every tight Hamilton cycle is a Hamilton Berge cycle. By contrast, for \(\ell<r-1\), a Hamilton \(\ell\)-cycle is not a Hamilton Berge cycle in general because it has too few hyperedges [1903.09057].

An important representation is via the incidence bipartite graph. Berge cycles correspond exactly to cycles in that incidence graph, and Hamiltonian Berge cycles correspond to cycles covering all vertex-vertices on the hypergraph side. This translation is especially useful in connectivity and longest-cycle arguments, where bipartite graph techniques can be applied directly to hypergraphs [2111.06710][2004.08291].

## 2. Degree conditions, exact-type criteria, and structural forcing

A large part of the theory asks which local density conditions force a Berge Hamilton cycle. One sharp-type direction is Pósa-style degree-sequence forcing. For \(r\)-uniform hypergraphs with degree sequence \((d_1,\dots,d_n)\), \(n>2r\), and \(r\ge 3\), the degree conditions
\[
d_i>i \quad \text{for } 1\le i<r,
\]
and
\[
d_i>\binom{i}{r-1} \quad \text{for } r\le i\le \frac{n-1}{2},
\]
together with an additional even-\(n\) correction, are sufficient for Berge Hamiltonicity, and the paper gives explicit extremal constructions showing the conditions are sharp. The same work also proves a non-uniform analogue with thresholds \(d_i>2^i\) up to the middle of the degree sequence, again with an even-\(n\) correction [2111.06710].

A different exact-type criterion uses connectivity together with minimum degree. Every \(3\)-connected hypergraph \(H\) with
\[
\delta(H)\ge \max\{|V(H)|,\tfrac{|E(H)|+10}{4}\}
\]
has a hamiltonian Berge cycle. This result is sharp and is obtained by translating the problem to \(3\)-connected bipartite incidence graphs that contain a cycle through all vertices on one side [2004.08291].

Sparse random analogues of Dirac-type forcing are also available. For the binomial random \(r\)-uniform hypergraph \(H^{(r)}(n,p)\), asymptotically almost surely every spanning subgraph \(H\subseteq H^{(r)}(n,p)\) with
\[
\delta_1(H)\ge \left(\frac{1}{2^{r-1}+\gamma}\right)p\binom{n}{r-1}
\]
contains a Hamilton Berge cycle, provided
\[
p \ge \frac{\log^{17r} n}{n^{r-1}}.
\]
The minimum-degree constant is asymptotically tight, with the standard obstruction given by splitting the vertex set into two nearly equal parts and retaining only induced edges inside the parts [1903.09057].

These criteria collectively show that Berge Hamiltonicity is governed by several distinct thresholds: exact degree-sequence conditions in deterministic uniform and non-uniform settings, connectivity-sensitive incidence-graph thresholds, and resilience-type minimum-degree conditions in sparse random environments. This suggests that the Berge notion is flexible enough to absorb techniques from both graph Hamiltonicity and hypergraph expansion, but rigid enough to sustain sharp extremal barriers [2111.06710][1903.09057].

## 3. Pancyclicity, covering conditions, and local surplus phenomena

In some regimes the natural conclusion is stronger than Hamiltonicity. For \(3\)-uniform covering hypergraphs, where every vertex pair lies in some hyperedge, every covering \([3]\)-graph on \(n\ge 6\) vertices contains a Berge cycle \(C_s\) for every \(3\le s\le n\). In particular, every such hypergraph contains a Hamiltonian Berge cycle. The proof proceeds through a Hamiltonian Berge path and then a structural analysis of an \((n-1)\)-cycle with one external vertex [1901.06042].

A sharp Dirac-type pancyclicity theorem is known for larger uniformity. If \(n\ge 70\), \(3\le r\le \lfloor (n-1)/2\rfloor-2\), and an \(n\)-vertex \(r\)-uniform hypergraph \(\mathcal H\) satisfies
\[
\delta(\mathcal H)\ge \binom{\lfloor (n-1)/2\rfloor}{r-1}+1,
\]
then \(\mathcal H\) is pancyclic, meaning that it contains Berge cycles of every length from \(3\) to \(n\). The Hamiltonian Berge cycle is then simply the length-\(n\) case. The threshold is sharp: the paper gives constructions with minimum degree exactly \(\binom{\lfloor (n-1)/2\rfloor}{r-1}\) but no Berge Hamilton cycle [2410.21733].

Odd-order Hamiltonian Berge cycles exhibit an additional local robustness phenomenon. Let \(n=2r+1\), let \(C\) be a Hamiltonian Berge cycle, and let \(\mathcal G\) be a family of hyperedges, all of size at least \(r\), containing the hyperedges of \(C\). If \(D\subseteq\{2,\ldots,r\}\) and \(|\mathcal G|\ge n+|D|\), then the hyperedges can be reassigned to the adjacent pairs of the same cyclic order so that, for each \(d\in D\), a distinct unused hyperedge realizes cyclic distance \(d\). Consequently, the odd-order case of the one-extra-edge question of Bailey, Hollars, Li and Luo has an affirmative answer for all \(n=2r+1\ge 7\), in the convention including Berge cycles of length \(2\) [2606.12230].

These results clarify a common misconception. A Berge Hamilton cycle is not merely a terminal longest-cycle statement; in several dense or covering regimes it lies inside a substantially richer cycle structure, sometimes full pancyclicity and sometimes a reassignable Hamiltonian framework with prescribed leftover chords [1901.06042][2606.12230].

## 4. Random hypergraphs and hitting-time theory

The random theory of Berge Hamilton cycles is notably sharp. In the random \(r\)-uniform hypergraph process \(\mathcal H(t)\), where edges are added sequentially, the stopping-time theorem states that for fixed \(r\ge 3\),
\[
\mathcal H(T_1)
\]
almost surely has a weak Hamiltonian Berge cycle, while
\[
\mathcal H(T_2)
\]
almost surely has a Hamiltonian Berge cycle, where \(T_k\) is the first time the minimum degree reaches \(k\). Thus the natural obstruction for full Berge Hamiltonicity is exactly minimum degree \(2\) [1809.03596].

The same paper identifies the threshold in the Erdős–Rényi model \(G(n,r,p)\). If
\[
p=(r-1)!\,\frac{\log n+\log\log n+c_n}{n^{r-1}},
\]
then the probability of Berge Hamiltonicity has the limiting behavior
\[
0,\quad e^{-e^{-c}},\quad 1
\]
according as \(c_n\to -\infty\), \(c_n\to c\in\mathbb R\), or \(c_n\to\infty\). The \(2\)-out model is also sufficient for ordinary Berge Hamiltonicity when \(r\ge 4\), whereas the \(1\)-out model is not [1809.03596].

A dense-host extension of the hitting-time paradigm is now available. In a random process constrained to a dense \(r\)-uniform host hypergraph \(H\), the hitting time \(\tau_2(H)\) for minimum degree \(2\) coincides with the hitting time \(\tau_{\mathcal{BH}}(H)\) for the first Berge Hamilton cycle with high probability, provided the host satisfies either
\[
\delta_1(H)\ge \left(\frac{1}{2^{r-1}+\varepsilon}\right)\binom{n-1}{r-1}
\]
or
\[
\delta_2(H)\ge \varepsilon n^{r-2},
\]
and, more generally, a local codegree-overlap condition. This extends the complete-host result of Bal, Berkowitz, Devlin, and Schacht to dense host hypergraphs [2512.06675].

Random results also reveal the boundary between graph Hamiltonicity and Berge Hamiltonicity. The weak cycle threshold is tied to disappearance of isolated vertices, whereas ordinary Berge Hamiltonicity requires the distinct-edge constraint and therefore the stronger minimum-degree-\(2\) barrier. This sharp separation has no direct graph analogue because in graphs repeated-edge and distinct-edge formulations coincide [1809.03596].

## 5. Decompositions and Ramsey-type forcing

A classical global question asks when all edges of a large hypergraph can be partitioned into Hamilton Berge cycles. For the complete \(k\)-uniform hypergraph \(K_n^{(k)}\), the obvious necessary condition is
\[
n \mid \binom{n}{k},
\]
since each Hamilton Berge cycle uses exactly \(n\) hyperedges. Kühn and Osthus proved that this condition is sufficient whenever \(4\le k<n\), \(n>30\), and the divisibility condition holds. Their argument converts the decomposition problem into a perfect-matching problem in an auxiliary bipartite graph and uses the Kruskal–Katona theorem together with Tillson’s theorem on Hamilton decompositions of complete digraphs [1403.7932].

Ramsey-type forcing concerns monochromatic Berge Hamilton cycles in edge-colored complete uniform hypergraphs. One theorem shows that if \(r>t\ge 2\) and
\[
n\ge (r-1)\left\lfloor \frac{r-2}{t-1}\right\rfloor+2,
\]
then every
\[
\left\lfloor \frac{r-2}{t-1}\right\rfloor\text{-edge coloring of }K_n^r
\]
contains a monochromatic Hamiltonian \(t\)-tight Berge-cycle. For \(t=2\), this yields that every \((r-2)\)-edge coloring of \(K_n^r\) with \(n\ge r^2-3r+4\) contains a monochromatic Hamiltonian Berge-cycle; the same paper also proves that any \(3\)-edge coloring of \(K_n^4\) with \(n\ge 85\) contains such a cycle [1403.2894].

The original conjecture of Gyárfás, Lehel, Sárközy, and Schelp asked for \((r-1)\)-colorings rather than \((r-2)\)-colorings. This was proved for sufficiently large \(n\): if \(r\ge 4\) and
\[
n > 6r\binom{4r}{r-1},
\]
then every \((r-1)\)-edge-coloring of \(K_n^r\) contains a monochromatic Hamiltonian Berge-cycle. The proof uses the shadow graph, a notion of good colors on graph edges, and a Hamiltonicity argument in an auxiliary graph satisfying Chvátal-type degree conditions [1404.3385].

These decomposition and Ramsey results position Berge Hamilton cycles as a canonical spanning object in complete uniform hypergraphs. In decomposition problems they play the role of Walecki-type factors; in coloring problems they play the role of unavoidable monochromatic spanning structures [1403.7932][1404.3385].

## 6. Proof methods, representations, and scope

Several proof paradigms recur across the subject. Shadow graphs are fundamental: if a graph edge in the \(2\)-shadow can be matched to a distinct containing hyperedge, then graph cycles lift to Berge cycles. This idea drives both pancyclicity proofs and monochromatic forcing arguments [2410.21733][1403.2894].

The incidence-graph viewpoint is equally central. By encoding vertices and hyperedges as the two parts of a bipartite graph, Berge Hamiltonicity becomes a long-cycle problem in bipartite graphs. This is the natural language for \(3\)-connectedness results and for translating hypergraph questions into graph Hamiltonicity techniques involving fans, longest cycles, and structural obstructions [2004.08291][2111.06710].

Absorbing methods, expander methods, and rotation-extension ideas also transfer effectively. The random-hypergraph Dirac theorem uses an absorbing-reservoir decomposition with vertex sets \(Y\), \(Z\), and \(W\), together with pseudorandomness, matching lemmas, and connection lemmas for edge-disjoint Berge paths [1903.09057]. The random-process results use sparse expanders, Pósa rotation for Berge paths, and booster or booster-pair arguments to iteratively force Hamiltonicity [1809.03596][2512.06675].

At the combinatorial end, Hall-type matching and exchange arguments remain ubiquitous. They appear in decomposition theorems through auxiliary bipartite graphs [1403.7932], in colored settings through shadow-graph lifting [1403.2894], and in odd-order leftover-chord theorems through alternating matching exchange combined with an additive lemma in \(\mathbb Z_{2r+1}\) [2606.12230].

The scope of the concept is broad but not unlimited. A Berge Hamilton cycle is a spanning hypergraph cycle notion, but it is weaker than tight Hamiltonicity, stronger than weak Berge Hamiltonicity, and conceptually different from ordinary graph Hamilton cycles. Results about graph Hamiltonicity, Grinberg conditions, or cycle spaces may provide analogies, yet they do not automatically transfer to the Berge setting. The modern literature therefore treats Berge Hamiltonicity as a distinct branch of hypergraph Hamilton cycle theory, with its own thresholds, extremal constructions, random hitting times, and Ramsey phenomena [1903.09057][1809.03596].

Source: https://www.emergentmind.com/topics/berge-hamilton-cycle