On almost hypohamiltonian graphs
Extracted source text ยท arXiv 1606.06577
%\documentclass{amcjoucc}
%\documentclass[review]{siamart}
\documentclass{siamart}
%\usepackage{graphicx}
%...
%\usepackage{amsmath, amsfonts, amsthm, amssymb, graphicx, dsfont}
\usepackage{amsmath, amsfonts, amssymb, graphicx, dsfont}
%\usepackage[numbers,square, comma, sort&compress]{natbib}
\usepackage[small,bf,hang]{caption} % Improved caption
\usepackage{subfig} % Figures which consist of multiple subfigures
\usepackage{hyperref}
\usepackage{algorithm, algorithmic}
\usepackage{multirow}
%\usepackage{epstopdf}
\usepackage[T1]{fontenc} %need this for v-accent over i in Russian name
\usepackage{color}
\newcommand{\todo}[1] {
{\tt \color{red}{[TODO: {#1}]}}
}
\newcommand{\jg}[1] {
{\color{blue}{{\tt [JGSAYS: }{#1}]}}
}
\newcommand{\cz}[1] {
{\color{magenta}{{\tt [CZSAYS: }{#1}]}}
}
\hyphenation{krijgs-laan hypo-cyclic hypo-hamilto-nian}
%\newtheorem{theorem}{Theorem}
%\newtheorem{definition}[thm]{Definition}
%\newtheorem{lemma}[theorem]{Lemma}
%\newtheorem{corollary}[theorem]{Corollary}
%\newtheorem{proposition}[theorem]{Proposition}
\graphicspath{{figures/}}
%\begin{frontmatter} %% Title, information about author, abstract, etc.
\title{New results on hypohamiltonian and almost hypohamiltonian graphs}
%\footnote{A preliminary version of this paper is submitted to the 42nd International Workshop on Graph-Theoretic Concepts in Computer Science (WG 2016).}
\author{Jan Goedgebeur\thanks{Department of Applied Mathematics, Computer Science \& Statistics, Ghent University, Krijgslaan 281-S9, 9000 Ghent, Belgium (\email{jan.goedgebeur@ugent.be}, \email{czamfirescu@gmail.com}).
The first author is supported by a Postdoctoral Fellowship of the Research Foundation Flanders (FWO). The second author is supported by a PhD fellowship at Ghent University on the Special Research Fund (BOF) scholarship~01DI1015.}
\and Carol T.\ Zamfirescu\footnotemark[1]}
\date{}
%\end{frontmatter} %% End of the front matter
%% Your article
%------------------------------------------------------------------------------------------------------------------------------
\begin{document}
\maketitle
\begin{abstract}
Consider a non-hamiltonian graph $G$. $G$ is \emph{hypohamiltonian} if for every vertex $v$ in $G$, the graph $G - v$ is hamiltonian. $G$ is \emph{almost hypohamiltonian} if there exists a vertex $w$ in $G$ such that $G - w$ is non-hamiltonian, and $G - v$ is hamiltonian for every vertex $v \ne w$. McKay asked in [\emph{J. Graph Theory}, doi: 10.1002/jgt.22043] whether infinitely many planar cubic hypohamiltonian graphs of girth~5 exist. We settle this question affirmatively.
The second author asked in [\emph{J. Graph Theory} \textbf{79} (2015) 63--81] to find all orders for which almost hypohamiltonian graphs exist---we solve this problem as well. To this end, we present a specialised algorithm which generates complete sets of almost hypohamiltonian graphs for various orders. Furthermore, we show that the smallest cubic almost hypohamiltonian graphs have order~26. We also provide a lower bound for the order of the smallest planar almost hypohamiltonian graph and improve the upper bound for the order of the smallest planar almost hypohamiltonian graph containing a cubic vertex. Finally, we determine the smallest planar almost hypohamiltonian graphs of girth~5, both in the general and cubic case.
\end{abstract}
%\noindent
\begin{keywords}
Hamiltonian, hypohamiltonian, almost hypohamiltonian, planar, cubic, exhaustive generation
\end{keywords}
% REQUIRED
\begin{AMS}
05C10, 05C35, 05C38, 05C45, 05C85
\end{AMS}
%------------------------------------------------------------------------------------------------------------------------------
\section{Introduction}
\label{section:intro}
Every graph in this paper will be undirected, finite, connected, and neither contain loops nor multiple edges, unless stated otherwise. A graph is \emph{hamiltonian} if it contains a cycle visiting every vertex of the graph---such a cycle is called \emph{hamiltonian}. A graph $G$ is \emph{hypohamiltonian} if $G$ itself is non-hamiltonian, but for every vertex $v$ in $G$, the graph $G - v$ is hamiltonian. The study of hypohamiltonicity goes back to a paper of Sousselier~\cite{So63} from the early sixties. The survey of Holton and Sheehan~\cite{HS93} provides a good overview. New material published after the appearance of this survey includes an article of Jooyandeh, McKay, \"{O}sterg{\aa}rd, Pettersson, and the second author~\cite{JMOPZ}, work of McKay~\cite{Mc16}, and the paper~\cite{GZ} by the two authors of this work.
We now introduce our terminology---see \cite{Di10} for undefined notions. We call a vertex \emph{cubic} if it has degree~3, and a graph \emph{cubic} if all of its vertices are cubic. Consider a graph~$G$. Two edges of $G$ are \emph{independent} if they have no common vertices. We use $\deg(v)$ to denote the degree of a vertex~$v$ and put $\Delta(G) = \max_{v \in V(G)} \deg(v)$. The \emph{girth} of a graph is the length of its shortest cycle. A cycle of length~$k$ will be called a \emph{$k$-cycle}. For $S \subset V(G)$, $G[S]$ shall denote the graph induced by $S$. A not necessarily connected subgraph $G' = (V',E') \subset G = (V,E)$ is \emph{spanning} if $V' = V$. For a set $X$, we denote by $|X|$ its cardinality.
Following Aldred, McKay, and Wormald~\cite{AMW97}, we say that a graph $G$ is \emph{hypocyclic} if for every $v \in V(G)$, the graph~$G - v$ is hamiltonian. Hamiltonian hypocyclic graphs are called ``1-hamiltonian'', see e.g.~\cite{CKL70}, so the family of all hypocyclic graphs is the disjoint union of the families of all 1-hamiltonian and hypohamiltonian graphs. A graph $G$ is \emph{almost hypocyclic} if there exists a $w \in V(G)$ such that $G - w$ is non-hamiltonian, but for every vertex $v \ne w$ in $G$, the graph~$G - v$ is hamiltonian. We call $w$ the \emph{exceptional vertex} of $G$. If $G$ is additionally non-hamiltonian, then $G$ is called \emph{almost hypohamiltonian} (with exceptional vertex $w$).
This article is structured as follows. In Section~\ref{section:McKay} we solve a problem of McKay: he asked~\cite{Mc16} whether infinitely many planar cubic hypohamiltonian graphs of girth~5 exist. We settle this question affirmatively.
In Section~\ref{section:almhypoham} we deal with almost hypohamiltonian graphs, extending a series of results (and settling certain open questions) of the second author~\cite{Za15,Za16}. In particular, in Section~\ref{subsect:generation_algo} we present a specialised algorithm for generating all almost hypohamiltonian graphs. We also use our implementation of this algorithm to generate complete sets of almost hypohamiltonian graphs for various orders and in Section~\ref{subsect:gen_case} we show that there exists an almost hypohamiltonian graph of order~$n$ iff $n \ge 17$, solving~\cite[Problem~2]{Za15}.
In Section~\ref{subsect:cubic_case} we show that the smallest cubic almost hypohamiltonian graphs have order~26, which settles the non-planar variant of~\cite[Problem~1]{Za15}---the planar version had been solved by McKay (private communication). In Section~\ref{subsect:planar_case} we give a lower bound for the order of the smallest planar almost hypohamiltonian graphs, and improve the upper bound for the order of the smallest planar almost hypohamiltonian graph whose exceptional vertex is cubic (from 47, see~\cite{Za15}, to 36), which are particularly important due to results from~\cite{Za15}. This answers \cite[Problem~6(a)]{Za16} and \cite[Problem~7]{Za16}. We also prove that such a graph of order $n$ exists for every $n \ge 73$, partially answering \cite[Problem~6(b)]{Za16}. Furthermore, we show that the smallest planar almost hypohamiltonian graph of girth~5 has order 44. (In the hypohamiltonian case, this number is~45, see~\cite{JMOPZ}.) Finally, in Section~\ref{subsect:planar_cubic_case} we determine that the smallest planar cubic almost hypohamiltonian graph has order at least~54, and that the smallest such graph which additionally has girth~5, is of order~68.
\section{McKay's Problem on planar cubic hypohamiltonian graphs}
\label{section:McKay}
In~\cite{Mc16}, McKay constructed three planar cubic hypohamiltonian graphs of girth~5 and raised the natural question whether infinitely many such graphs exist. In this section we prove that this question has a positive answer. Let $G$ be a graph containing vertices $v,w,x,y$. Following Chv\'atal~\cite{Ch73}, the pair of vertices $(v,w)$ is \emph{good} in $G$ if there exists a hamiltonian path in $G$ with end-vertices $v$ and $w$. Two pairs of vertices $((v,w),(x,y))$ are \emph{good} in $G$ if there exist two disjoint paths which together span $G$, and which have end-vertices $v$ and $w$, and $x$ and $y$, respectively.
Let $H$ and $H'$ be cubic graphs on at least 6 vertices. Consider $J = H - \{ ab, cd \}$, where $ab$ and $cd$ are independent edges in $H$, $J' = H' - \{ x,y \}$, where $x$ and $y$ are adjacent vertices in $H'$, and let $a',b'$ and $c',d'$ be the other neighbours of $x$ and $y$ in $H'$, respectively. Then the \emph{dot product} $H \cdot H'$ is defined as the graph $$(V(H) \cup V(J'), E(J) \cup E(J') \cup \{ aa', bb', cc', dd' \}).$$ (Note that under above conditions, the dot product may be disconnected. The dot product was defined by Isaacs in~\cite{Is75}. In fact, following Skupie\'n~\cite{Sk16}, it seems that this operation was invented earlier by Adel'son-Vel'ski\v{\i} and Titov~\cite{AT73}.)
%more precisely, the hypohamiltonicity part. \jg{It is a bit ambiguous what is meant here: is the hypoham part too weak or is the hypoham part the result which he advertises?} (The fact that the dot product of two snarks is a snark was known since the early seventies.) \cz{I am very unsure what to do here. If we leave out the Cavicchioli et al.\ part, a referee can easily say that our Thm below was already shown by them---if we include it, we certainly do not make friends. I tend to include it, since it motivates giving the full proof here. Our proof has 3 advantages over Cavicchioli et al.: (1) it is more general, (2) it uses weaker hypotheses, (3) it does not make the $v \in \{ a,b,c,d \}$ mistake. Also note that the Thm below is the general version---I added a remark at the end of this subsection.} \jg{Ok then let's include it. What exactly do you want to include? The commented text below? Maybe we should be a bit careful with the wording, e.g. write the assumptions are "not entirely correct" rather than writing "wrong"?} %A discussion of this omission and a new statement and proof of the theorem can be found in Section~3 of~\cite{CMRS03}---note that although Fiorini's hypotheses are too weak, the hypotheses in \cite[Theorem~3.2]{CMRS03} are unrealisable (in the second part, i.e.\ their hypothesis (2)): using the notation from~\cite[Theorem~3.2]{CMRS03}, if $v \in \{ a,b \}$, then $(a,b)$ cannot be good in $G - \{ v, ab, cd \}$, but the authors demand both $(c,d)$ \emph{and} $(a,b)$ to be good (whereas Fiorini and we ask for at least one of $(a,b)$ and $(c,d)$ to be good).
\begin{theorem}\label{product}
Let $H$ and $H'$ be cubic graphs, $H$ non-hamiltonian and $H'$ hypohamiltonian. Consider independent edges $ab, cd \in E(H)$. If each of $$(a,c), \ (a,d), \ (b,c), \ (b,d), \ {\rm {\it and}} \ ((a,b),(c,d))$$ is good in $H - \{ ab, cd \}$, and for each vertex $t \in V(H)$, at least one of $(a,b)$ or $(c,d)$ is good in $H - \{ t, ab, cd \}$, then $H \cdot H'$ is a cubic hypohamiltonian graph as well. If $H$ and $H'$ are planar, and $ab$ and $cd$ lie on the same facial cycle, then the dot product can be applied such that $H \cdot H'$ is planar, as well. If $g$ ($g'$) is the girth of $H$ ($H'$), then the girth of $H \cdot H'$ is at least $\min \{ g, g' \}$.
\end{theorem}
\begin{proof}
Put $G = H \cdot H'$ and denote by $x,y$ the (adjacent) vertices which were deleted from $H'$ to form the dot product. $G$ is obviously cubic. Put $N(x) = \{ a', b', y \}$ and $N(y) = \{ c', d', x \}$ such that the unique neighbour of $a'$ ($b'$, $c'$, $d'$) in $G$ is $a$ ($b$, $c$, $d$). We treat $H - \{ ab, cd \} = J$ and $H' - \{ x,y \} = J'$ as subgraphs of $G$. W.l.o.g.\ we may assume that the aforementioned labeling of vertices satisfies that if $H$ and $H'$ are planar, then $G$ is also planar.
We first show that $G$ is non-hamiltonian. Assume there exists a hamiltonian cycle ${\mathfrak h}$ in $G$. If ${\mathfrak h} \cap J$ consists of two components $P$ and $P'$, then either ${\mathfrak h} \cap J + ab + cd$ is a hamiltonian cycle in $H$, a contradiction since $H$ is non-hamiltonian, or $P$ is a path with end-vertices $a$ and $b$, and $P'$ is a path with end-vertices $c$ and $d$. But then ${\mathfrak h} \cap J'$ consists of two disjoint paths which together with $a'xb'$ and $c'yd'$ form a hamiltonian cycle in $H'$, a contradiction since $H'$ is hypohamiltonian. So ${\mathfrak h} \cap J$ consists of one component. If ${\mathfrak h} \cap J$ has end-vertices $a,b$ or $c,d$, then we immediately obtain a contradiction (by adding the edge $ab$ or $cd$) to the non-hamiltonicity of $H$, so we may assume that ${\mathfrak h} \cap J'$ is a hamiltonian path $P$ with end-vertices $p \in \{ a',b' \}$ and $q \in \{ c',d' \}$. Then $P \cup pxyq$ is a hamiltonian cycle in $H'$, once again a contradiction.
It remains to show that $G - s$ is hypocyclic for every $s \in V(G)$. Let $s \in V(J')$. Since $H'$ is hypohamiltonian, there exists a hamiltonian cycle ${\mathfrak h}$ in $H' - s$. Put $Q = a'xb'$ and $Q' = c'yd'$. Note that if exactly one of $Q$ and $Q'$ lies in ${\mathfrak h}$, then necessarily $s \in \{ x,y \}$, which is impossible, as neither $x$ nor $y$ lie in $J'$.
\smallskip
Case 1: Neither $Q$ nor $Q'$ lie in ${\mathfrak h}$. In this situation, $xy \in E({\mathfrak h})$, and say $a'$ and $d'$ are the end-vertices of the path $P' = {\mathfrak h} - \{ x,y \}$, which is a hamiltonian path in $J'$. $(a,d)$ is good in $J$, i.e.\ there exists a hamiltonian path $P$ in $J$ with end-vertices $a,d$, so we obtain the hamiltonian cycle $P' \cup P + aa' + dd'$ in $G - s$ as desired. (If $P'$ has end-vertices other than $a',d'$, the argument is very similar. Likewise if $s \in \{ a',b',c',d' \}$.)
\smallskip
%Case 2: Either $Q$ or $Q'$ lie in ${\mathfrak h}$, w.l.o.g.\ $Q \subset {\mathfrak h}$. Then deleting $x$ from ${\mathfrak h}$ and the fact that $(a,b)$ is good in $H$ leads to a hamiltonian cycle in $G - s$.
Case 2: Both $Q$ and $Q'$ lie in ${\mathfrak h}$. $((a,b),(c,d))$ is good in $H$, so there exist disjoint paths $P_1,P_2$ in $H$ which together span $H$, with end-vertices $a,b$ and $c,d$, respectively. Denote by $P'_1,P'_2$ the components of ${\mathfrak h} - \{ x,y \}$. Then $P_1 \cup P_2 \cup P'_1 \cup P'_2 + aa' + bb' + cc' + dd'$ is a hamiltonian cycle in $G - s$.
\smallskip
Now consider $s \in V(H)$. We know that at least one of $(a,b)$ and $(c,d)$ is good in $H - \{ s, ab, cd \}$, say $(a,b)$. So there exists a hamiltonian path $P$ in $H - \{ s, ab, cd \}$ with end-vertices $a$ and $b$. Since $H'$ is hypohamiltonian, $H' - y$ is hamiltonian, which implies that there exists a hamiltonian path $P'$ in $J'$ with end-vertices $a'$ and $b'$. Now $P \cup P' + aa' + bb'$ is a hamiltonian cycle in $G - s$ (here it is crucial that $cd \notin E(P)$, as $cd \in E(H)$, but $cd \notin E(G)$). If $(c,d)$ is good in $H - \{ s, ab, cd \}$ and not $(a,b)$, we use the fact that $H' - x$ is hamiltonian.)
Finally, let $g$ be the girth of $H$ and $g'$ be the girth of $H'$. Assume the girth of $H \cdot H'$ is $k < \min \{ g, g' \}$, and let $C$ be a $k$-cycle in $H \cdot H'$. Clearly, $C$ contains (at least) two edges $e_1, e_2 \in \{ aa', bb', cc', dd' \}$ with $e_1 = aa'$ w.l.o.g. If $e_2 = bb'$, then %$(C \cap J) + ab$ is a cycle of length at most $k - 3$ in $H$ and
$(C \cap J') \cup a'xb'$ is a cycle of length at most $k - 2$ in $H'$. If $e_2 = cc'$, then $(C \cap J') \cup a'xyc'$ is a cycle of length at most $k$ in $H'$. If $k < g'$, in both cases we have a contradiction, so $k \ge g'$. But this is impossible, as $k < \min \{ g, g' \}$. We have shown that the girth of $H \cdot H'$ is at least $\min \{ g, g' \}$.
\end{proof}
The above result is inspired by a theorem of Fiorini~\cite{Fi83}, who showed that gluing two hypohamiltonian snarks using the dot product yields a new hypohamiltonian snark. Note that in Fiorini's statement the hypotheses he formulates are too weak to prove the result he advertises; Cavicchioli et al.~\cite{CMRS03} have identified this omission, but we believe their version is also not entirely correct. A detailed account will be given in the note~\cite{GSZ}.
In the following, we will call the pair of edges $ab,cd$ from the statement of Theorem~\ref{product} \emph{suitable}. Let $H$ be the graph from Figure~\ref{fig:hypo_76}. It is one of the three planar cubic hypohamiltonian graphs of girth~5 on 76~vertices found by McKay~\cite{Mc16}. Using a computer, we found several suitable pairs of edges $ab,cd$ in this graph. Since we will use double-digit numbers to label the vertices, we shall denote the edge $vw$ by $(v, w)$ for better readability.
\begin{figure}[h!t]
\centering
%\includegraphics[width=1.0\textwidth]{Hypohamiltonian_graphs_76v-g5-cyc4-graph1.png}
\includegraphics[trim = 10mm 0mm 10mm 0mm, clip,width=1.0\textwidth]{Hypohamiltonian_graphs_76v-g5-cyc4-graph1-bold.pdf}
\caption{The planar cubic hypohamiltonian graph $H$. It has girth~5 and order~76. $H$ is one of the three smallest planar cubic hypohamiltonian graphs of girth~5, and was found by McKay~\cite{Mc16}. The suitable edge pair $( 20,28 ), ( 52,62 )$ is marked in bold.}
\label{fig:hypo_76}
\end{figure}
\begin{lemma}\label{suitable}
Let $H$ be the graph from Figure~\ref{fig:hypo_76}. The edge pair $( 20,28 ), ( 52,62 )$ is suitable in $H$.
\end{lemma}
\begin{proof}
It is clear from Figure~\ref{fig:hypo_76} that the edges $( 20,28 )$ and $( 52,62 )$ are on the same facial cycle. (This guarantees that the dot product is planar.)
$(20,52), (20,62), (28,52)$ and $(28,64)$ are good in $H - \{(20,28),(52,62)\}$ due to the following hamiltonian paths, respectively:
\smallskip
\begin{itemize}
\item 52, 44, 36, 26, 16, 8, 3, 7, 14, 22, 23, 15, 24, 25, 34, 41, 42, 35, 43, 51, 61, 71, 72, 62, 53, 45, 37, 29, 30, 38, 46, 55, 54, 64, 63, 73, 76, 70, 60, 69, 75, 74, 66, 65, 56, 47, 57, 67, 68, 59, 50, 58, 49, 48, 40, 33, 32, 39, 31, 21, 13, 6, 2, 1, 4, 9, 17, 27, 28, 19, 18, 10, 5, 11, 12, 20
\item 62, 53, 45, 37, 29, 30, 38, 46, 55, 54, 64, 63, 73, 72, 71, 61, 52, 44, 36, 26, 35, 43, 51, 60, 70, 76, 75, 69, 59, 68, 74, 66, 65, 56, 47, 57, 67, 58, 50, 42, 41, 49, 48, 39, 32, 23, 15, 24, 33, 40, 34, 25, 16, 8, 3, 7, 14, 22, 31, 21, 13, 6, 2, 1, 4, 9, 17, 27, 28, 19, 18, 10, 5, 11, 12, 20
\item 52, 44, 36, 26, 16, 25, 24, 15, 7, 14, 13, 21, 30, 38, 46, 55, 54, 64, 63, 73, 76, 70, 60, 69, 75, 74, 66, 65, 56, 47, 57, 67, 68, 59, 50, 58, 49, 48, 39, 31, 22, 23, 32, 33, 40, 34, 41, 42, 35, 43, 51, 61, 71, 72, 62, 53, 45, 37, 29, 20, 12, 6, 2, 1, 3, 8, 9, 4, 10, 5, 11, 19, 18, 17, 27, 28
\item 62, 53, 45, 37, 29, 20, 12, 6, 13, 14, 7, 15, 23, 22, 31, 21, 30, 38, 46, 55, 54, 64, 63, 73, 72, 71, 70, 76, 75, 69, 60, 51, 61, 52, 44, 43, 35, 42, 50, 59, 68, 74, 66, 65, 56, 47, 57, 67, 58, 49, 41, 34, 40, 48, 39, 32, 33, 24, 25, 16, 26, 36, 27, 17, 18, 10, 4, 9, 8, 3, 1, 2, 5, 11, 19, 28
\end{itemize}
\smallskip
$((20,28),(52,62))$ is good in $H - \{(20,28),(52,62)\}$ due to the following two disjoint paths with end-vertices $20$ and $28$, and $52$ and $62$, respectively, which together span $H$.
\smallskip
\begin{itemize}
\item 28, 19, 11, 12, 20
\item 62, 53, 45, 37, 29, 30, 21, 13, 6, 2, 5, 10, 18, 17, 27, 36, 26, 35, 42, 41, 34, 25, 16, 8, 9, 4, 1, 3, 7, 14, 22, 31, 39, 32, 23, 15, 24, 33, 40, 48, 49, 58, 50, 59, 69, 60, 70, 76, 75, 74, 68, 67, 57, 66, 65, 56, 47, 38, 46, 55, 54, 64, 63, 73, 72, 71, 61, 51, 43, 44, 52
\end{itemize}
\smallskip
The fact that at least one of $(20,28)$ or $(52,62)$ is good in $H - \{ t, (20,28), (52,62)\}$ for every $t \in V(H)$ was shown by computer. We do not include the 76~hamiltonian paths here due to space constraints---76~paths are necessary since $H$ has trivial automorphism group~\cite{Mc16}---, but in each case we verified that the path found by the computer is indeed a valid hamiltonian path in the graph.
\end{proof}
%\cz{Let $\Omega$ be the planar cubic hypohamiltonian graph of girth~5 with non-trivial automorphism group we constructed in~\cite[Fig.~5]{GZ} OR one of McKay's graph---it depends on where we find suitable edges. Insert here details concerning suitable edges}.
%\jg{This will have to be rephrased this a little. Also in the lemma above I called the graph $H$ instead of $\Omega$ -- feel free to change this.}
As in Lemma~\ref{suitable}, $H$ will denote in the following the 76-vertex graph from Figure~\ref{fig:hypo_76}. Consider a copy $H'$ of $H$. Using the pair of suitable edges lying on the same facial cycle in $H$ we have found in Lemma~\ref{suitable}, and any pair of adjacent vertices in $H'$, we form the graph $H \cdot H'$. By Theorem~\ref{product}, $H \cdot H'$ is a planar cubic hypohamiltonian graph of girth~5. (Theorem~\ref{product} states that the girth of $H \cdot H'$ is at least~5, and a planar 3-connected graph cannot have girth greater than~5.) Iterating this process (i.e., for a copy $H''$ of $H$, forming the dot product $H'' \cdot (H \cdot H')$, etc.) we obtain the following answer to McKay's question~\cite{Mc16}. The second part of the statement was obtained by using the 78-vertex planar cubic hypohamiltonian graph of girth~5 constructed in~\cite{GZ} as $H'$ in Theorem~\ref{product}.
%\begin{lemma}\label{lemma-inductio}
%Let $k \ge 1$. If $\Omega_k$ is hypohamiltonian, then so is $\Omega_{k+1}$.
%\end{lemma}
%\begin{proof}
%$\Omega_k$ is planar, cubic, and hypohamiltonian, and so is $\Omega$, as shown in~\cite{GZ}. We apply Theorem~\ref{product} with $H = \Omega$ and $H' = \Omega_k$. It remains to find a pair of suitable edges in $\Omega$. \cz{Can you do this? Double- and triple-check the argument!}
%\end{proof}
\begin{theorem}
There are infinitely many planar cubic hypohamiltonian graphs of girth~$5$. More specifically, there exist such graphs of order $n$ for every $n = 74k + a$, where $k \ge 2$ and $a \in \{ 2,4 \}$.
\end{theorem}
We remark that Theorem~\ref{product} may in fact be applied to show that there are smaller graphs which could be used in the above construction, since in its statement we only ask of $H$ to be non-hamiltonian, not hypohamiltonian. However, our focus here was to solve McKay's problem---future work may include a more careful analysis of the graphs which may take the place of $H$.
We shall return to planar 3-connected cubic graphs with interesting hypohamiltonian properties at the end of this paper.
%\newpage
\section{Almost hypohamiltonian graphs}
\label{section:almhypoham}
\subsection{The generation algorithm}
\label{subsect:generation_algo}
In~\cite{GZ} we presented an algorithm for the exhaustive generation of hypohamiltonian graphs. It is based on work of Aldred, McKay, and Wormald~\cite{AMW97}. We now describe how we modified this algorithm to generate almost hypohamiltonian graphs exhaustively. We point out that in the following arguments there is significant overlap with~\cite{GZ}---still, we choose to present these results for two reasons: firstly, small changes with respect to~\cite{GZ} are nonetheless necessary and important, and secondly, we would like to make this paper to a large extent self-contained.
Let $G$ be a possibly disconnected graph, and $p(G)$ the minimum number of disjoint paths needed to cover all vertices of $G$. Denote by $V_1(G)$ the vertices of degree~1 in $G$, and by $I(G)$ the set of all isolated vertices and all isolated edges of~$G$. Put
$$k(G) = \begin{cases}
0 & {\rm if } \ G \ {\rm is \ empty,}\\
\max \left\{ 1, \left\lceil{\frac{|V_1|}{2}}\right\rceil \right\} & {\rm if } \ I(G) = \emptyset \ {\rm but} \ G \ {\rm is \ not \ empty},\\
|I(G)| + k(G - I(G)) & {\rm else}.
\end{cases}$$
We now adapt two lemmas of Aldred, McKay, and Wormald~\cite{AMW97} from the hypocyclic to the almost hypocyclic case.
\begin{lemma}\label{lem:A+B}
Given an almost hypocyclic graph $G$ with exceptional vertex $w$, for any partition $(W,X)$ of the vertices of $G$ with $|W| > 1$ and $|X| > 1$, we have that $$p(G[W]) < |X| \quad {\rm {\it and}} \quad k(G[W]) < |X|.$$
\end{lemma}
Consider a graph $G$ containing a partition $(W,X)$ of its vertices with $|W| > 1$ and $|X| > 1$. If $p(G[W]) \ge |X|$, then we call $(W,X)$ a \emph{type A obstruction}, and if $k(G[W]) \ge |X|$, then we speak of a \emph{type B obstruction}. For efficiency reasons we only consider type A obstructions where $G[W]$ is a union of disjoint paths in the algorithm.
\begin{lemma}\label{lem:C}
Let $G$ be an almost hypocyclic graph with exceptional vertex $w$, and consider a partition $(W,X)$ of the vertices of $G$ with $|W| > 1$ and $|X| > 1$ such that $W$ is an independent set. Furthermore, for some vertex $v \in X$, define $n_1$ and $n_2$ to be the number of vertices of $X - v$ joined to one or more than one vertex of $W$, respectively. Then we have $2n_2 + n_1 \ge 2 |W|$ for every $v \in X$.
\end{lemma}
If all assumptions of Lemma~\ref{lem:C} are met and $2n_2 + n_1 < 2 |W|$ for some $v \in X$, we call $(W,X,v)$ a \emph{type C obstruction}. The following proof is a straightforward adaptation of the proof of Aldred, McKay, and Wormald~\cite{AMW97}.
\medskip
\noindent
\emph{Proof of Lemmas~\ref{lem:A+B} and \ref{lem:C}.} Consider $v \in X$ with $v \ne w$ (this choice is possible, since $|X| > 1$), and let ${\mathfrak h}$ be a hamiltonian cycle in $G - v$. The number of components of ${\mathfrak h}$ restricted to $W$ must be at least $p(G[W])$, so the number of components in ${\mathfrak h}$ restricted to $X - v$ is at least $p(G[W])$ (and thus we have $|X| - 1 \ge p(G[W])$). This handles type A and, since clearly $k(G) \le p(G)$, type B obstructions. For type C, consider the same cycle ${\mathfrak h}$. The number of edges of ${\mathfrak h}$ between $W$ and $X$ must be $2|W|$, but $X$ can supply at most $2n_2 + n_1$. \hfill $\Box$
\begin{lemma}
Let $G$ be an almost hypohamiltonian graph with exceptional vertex $w$. Then $G$ is $3$-connected. In particular, $G$ has minimum degree~$3$.
\end{lemma}
\begin{proof}
Since for every vertex $v \ne w$, the graph $G - v$ is hamiltonian, $G$ must be 2-connected. Now let $\{ x,y \}$ be a 2-cut in $G$. Both $G - x$ and $G - y$ are non-hamiltonian, which is impossible. So $G$ does not contain 2-cuts, thus $G$ is 3-connected.
\end{proof}
Note that if we allow more than one exceptional vertex, 3-connectedness is not guaranteed anymore. Or put in rigorous terms: consider a 2-connected graph $G$ of circumference $|V(G)| - 1$ and let $W \subset V(G)$ be the (possibly empty) inclusion-maximal set of vertices such that for all $w \in W$ the graph $G - w$ is non-hamiltonian. (And thus, for all $v \in V(G) \setminus W$, the graph $G - v$ is hamiltonian.) We say that $G$ is \emph{$|W|$-hypohamiltonian}. Although 0-hypohamiltonian (i.e.\ hypohamiltonian) and 1-hypohamiltonian (i.e.\ almost hypohamiltonian) graphs are 3-connected, it is easy to see that 2-hypohamiltonian graphs of connectivity~2 exist: consider a 4-cycle $v_1 v_2 v_3 v_4$, add an extra vertex $v_5$, and the edges $v_2v_5$ and $v_4v_5$. This emphasises the intrinsic qualitative difference between $\{ 0,1 \}$-hypohamiltonian graphs and $k$-hypohamiltonian graphs with $k \ge 2$.
\begin{lemma}
Let $M$ be a $3$-edge-cut in an almost hypohamiltonian graph $G$ with exceptional vertex $w$. Then $G - M$ contains exactly two components $A_1$ and $A_2$ with $A = K_1$ and $A_2 \ne K_1$. In particular, almost hypohamiltonian graphs are cyclically $4$-edge-connected.
\end{lemma}
\begin{proof}
Let $G - M$ contain components $A_1,...,A_k$, $k \ge 2$, and assume that $A_1 \ne K_1$ and $A_2 \ne K_1$. We put $M = \{ a_1b_1, a_2b_2, a_3b_3 \}$, where $a_i \in V(A_1), b_i \in V(A_2)$ for all $i$. Since $G$ is 3-connected, the elements of the set $\{ a_1, a_2, a_3, b_1, b_2, b_3 \}$ are pairwise distinct. Since $G$ is almost hypohamiltonian, there exists an $i$ such that $G - b_i$ and $G - a_i$ both are hamiltonian. Hence, there is a hamiltonian path in $A_1$ with end-vertices $a_j$ and $a_k$, $j \ne i \ne k$, and a hamiltonian path in $A_2$ with end-vertices $b_j$ and $b_k$, $j \ne i \ne k$. These paths together with $a_jb_j$ and $a_kb_k$ yield a hamiltonian cycle in $G$, a contradiction. Note that if at least one of the components of $G - M$ is isomorphic to $K_1$, then $k = 2$. This completes the proof.
\end{proof}
Collier and Schmeichel~\cite[p.~196]{CS78} observed that the vertices of a triangle in a hypohamiltonian graph have degree at least~4. We now show that this holds for almost hypohamiltonian graphs, as well.
\begin{lemma}\label{lemma:no_deg3_triangle}
Let $G$ be an almost hypohamiltonian graph containing a triangle $T$. Then every vertex of $T$ has degree at least~$4$.
\end{lemma}
\begin{proof}
Put $V(T) = \{ v_1, v_2, v_3 \}$, where $v_3$ shall be cubic. Since $G$ is almost hypohamiltonian, at least one of $G - v_1$ and $G - v_2$ must be hamiltonian, say $G - v_1$. Let ${\mathfrak h}$ be a hamiltonian cycle in $G - v_1$. As $v_3v_2 \in E({\mathfrak h})$, replace in ${\mathfrak h}$ the edge $v_3v_2$ with the path $v_3v_1v_2$. We obtain a hamiltonian cycle in $G$, a contradiction.
\end{proof}
We make use of these lemmas in our algorithm to generate all almost hypohamiltonian graphs---first we give an intuitive description, which is followed by the rigorous argument. Intuitively, by a \emph{good $Y$-edge} we mean an edge which works towards the removal of a type $Y$~obstruction, where $Y \in \{ A, B, C \}$. Let us now formally define these good $Y$-edges.
Given a hypohamiltonian graph $G'$, and let $G$ be a spanning subgraph of $G'$ which contains a type A obstruction $(W,X)$ (where $G[W]$ is a union of disjoint paths), since $G'$ is hypohamiltonian it cannot contain a type A obstruction. Thus, there is an edge $e \in E(G') \setminus E(G)$ the endpoints of which lie in different components of $G[W]$. Furthermore, at least one of the endpoints of $e$ has degree at most one in $G[W]$. We call such an edge a \textit{good $A$-edge} for $(W,X)$. In the same vein, a \textit{good $B$-edge} for a type B obstruction $(W,X)$ in $G$ is a non-edge of $G$ that joins two vertices of $W$, where at least one of those vertices has degree at most one in $G[W]$. Finally, a \textit{good $C$-edge} for a type C obstruction $(W,X,v)$ in $G$ is a non-edge $e$ of $G$ for which one of the two following conditions holds:
\begin{itemize}
\item[(i)] Both endpoints of $e$ are in $W$.
\item[(ii)] One endpoint of $e$ is in $W$ and the other endpoint is in $X-v$ and has at most one neighbour in $W$.
\end{itemize}
It is straightforward that this is the only way to destroy a type B/C obstruction.
The pseudocode of our enumeration algorithm is given in Algorithm~\ref{algo:init-algo} and Algorithm~\ref{algo:construct}.
It is analogous to the algorithm we presented in~\cite{GZ} to generate all hypohamiltonian graphs. Nonetheless, we choose to present it here as well for the sake of convenience.
In order to generate all almost hypohamiltonian graphs with $n$~vertices we start from a graph $G$ which consists of an $(n-1)$-cycle and an isolated vertex $h$ (disjoint from the cycle), so $G - h$ is hamiltonian. We then connect $h$ to $D$~vertices of the $(n-1)$-cycle in all possible ways and then perform Algorithm~\ref{algo:construct} on these graphs. Algorithm~\ref{algo:construct} will continue to recursively add edges in all possible ways such that at most one vertex $w$, the exceptional vertex, has degree larger than~$D$.
It is essential for the efficiency of the algorithm that as few as possible edges are added by Algorithm~\ref{algo:construct}, while still guaranteeing that all almost hypohamiltonian graphs are found by the algorithm.
We omit the proof of the following theorem as it is analogous to the proof of~\cite[Theorem~2.8]{GZ}.
\begin{theorem}
If Algorithm~\ref{algo:init-algo} terminates, the list of graphs $\mathcal H$ outputted by the algorithm is the list of all almost hypohamiltonian graphs with $n$ vertices.
\end{theorem}
\begin{algorithm}[ht!]
\caption{Generate all almost hypohamiltonian graphs with $n$ vertices}
\label{algo:init-algo}
\begin{algorithmic}[1]
\STATE let $\mathcal H$ be an empty list
\STATE let $G := C_{n-1} + h$
\FORALL{$3 \le D \le n-1$}
\STATE // Generate all almost hypohamiltonian graphs where at least one vertex has degree $D$ and at most one vertex has degree larger than $D$
\FOR{every way of connecting $h$ of $G$ with $D$ vertices of the $C_{n-1}$}
\STATE Call the resulting graph $G'$
\STATE Construct($G', D$) \ // i.e.\ perform Algorithm~\ref{algo:construct}
\ENDFOR
\ENDFOR
\STATE Output $\mathcal H$
\end{algorithmic}
\end{algorithm}
\begin{algorithm}[ht!]
\caption{Construct(Graph $G$, int $D$)}
\label{algo:construct}
\begin{algorithmic}[1]
\IF{$G$ is non-hamiltonian AND not generated before} \label{line:hamiltonian}
\IF{$G$ contains a type A obstruction $(W,X)$}
\FOR{every good A-edge $e \notin E(G)$ for $(W,X)$ for which at most one vertex has degree larger than $D$ in $G+e$}
\STATE Construct($G+e, D$) \label{line:destroy_typeA}
\ENDFOR
\ELSIF{$G$ contains a vertex $v$ of degree 2}
\FOR{every edge $e \notin E(G)$ for which at most one vertex has degree larger than $D$ in $G+e$}
\STATE Construct($G+e, D$) \label{line:destroy_deg2}
\ENDFOR
\ELSIF{$G$ contains a type C obstruction $(W,X,v)$} \label{line:typeC_obstr}
\FOR{every good C-edge $e \notin E(G)$ for $(W,X,v)$ for which at most one vertex has degree larger than $D$ in $G+e$}
\STATE Construct($G+e, D$)
\ENDFOR
\ELSIF{$G$ contains a vertex $v$ of degree 3 which is part of a triangle} \label{line:deg3_triangle}
\FOR{every edge $e \notin E(G)$ which contains $v$ as an endpoint for which at most one vertex has degree larger than $D$ in $G+e$}
\STATE Construct($G+e, D$)
\ENDFOR
\ELSIF{$G$ contains a type B obstruction $(W,X)$} \label{line:typeB_obstr}
\FOR{every good B-edge $e \notin E(G)$ for $(W,X)$ for which at most one vertex has degree larger than $D$ in $G+e$}
\STATE Construct($G+e, D$)
\ENDFOR
\ELSE
\IF{$G$ is almost hypohamiltonian} \label{line:hypohamiltonian}
\STATE add $G$ to the list $\mathcal H$
\ENDIF
\FOR{every edge $e \notin E(G)$ for which at most one vertex has degree larger than $D$ in $G+e$}
\STATE Construct($G+e, D$)
\ENDFOR
\ENDIF
\ENDIF
\end{algorithmic}
\end{algorithm}
Note that since our algorithm only adds edges and never removes any vertices or edges, all graphs obtained by the algorithm from a graph with a $g$-cycle will have a cycle of length at most~$g$. So in case we only want to generate almost hypohamiltonian graphs with a given lower bound~$k$ on the girth, we can prune the construction when a graph with a cycle with length less than~$k$ is constructed.
For more details on the algorithm we refer to~\cite{GZ}.
\subsection{The general case}
\label{subsect:gen_case}
By using our implementation of the algorithm described in Section~\ref{subsect:generation_algo} we generated all almost hypohamiltonian graphs with girth at least~$g$ of order~$n$ for various values of $g$ and $n$. The counts of the number of almost hypohamiltonian graphs can be found in Table~\ref{table:number_of_almhypoham_graphs}. Therein, we also mention the number of almost hypohamiltonian graphs whose exceptional vertex is cubic---these are of special importance due to the fact that two such graphs can be combined to form a hypohamiltonian graph, see \cite[Theorem~3]{Za15} for details. In each case we went as far as computationally possible.
Note that we could not determine the complete set of all almost hypohamiltonian graphs with 24 vertices and girth~5---however, we succeeded in obtaining a sample, which was essential to complete the proof of Theorem~\ref{thm:all_orders}. Combining our findings with the second author's~\cite[Theorem~2]{Za15}, we have the following, settling~\cite[Problem~2]{Za15}.
\begin{theorem} \label{thm:all_orders}
An almost hypohamiltonian graph of order $n$ exists iff $n \ge 17$. The smallest almost hypohamiltonian graph of girth~$4$ (girth~$5$) has order~$18$ (order~$17$).
\end{theorem}
Hypohamiltonian graphs of order $n$ exist iff $n \in \{ 10, 13, 15, 16 \}$ or $n \ge 18$, see~\cite{AMW97}. The smallest hypohamiltonian graph of girth~4 (girth~5) has order~18 (order~10); for details we refer to~\cite{GZ}. No almost hypohamiltonian graphs of girths other than 4 or 5 are known, but it seems reasonable to believe that adapting a technique of Thomassen~\cite{Th74-2} will provide examples of girth~3. We know of the existence of hypohamiltonian graphs of girth~$g$ for every $g \in \{ 3,...,7\}$, see~\cite{GZ}, while no hypohamiltonian graphs of girth greater than~7 are known. (The smallest hypohamiltonian graph of girth~7 is Coxeter's graph.) In Figure~\ref{fig:almhypo_17} the two smallest almost hypohamiltonian graphs (each of order~17) are depicted---both have girth~5. Figure~\ref{fig:almhypo_18_g4} shows the smallest almost hypohamiltonian graph of girth~4 (which has order~18).
All graphs from Table~\ref{table:number_of_almhypoham_graphs} can also be
downloaded from the \textit{House of Graphs}~\cite{hog} at
\url{http://hog.grinvin.org/AlmostHypohamiltonian} and also be inspected in the database of interesting graphs by searching for the keywords ``almost hypohamiltonian''.
\begin{table}
\centering
\small
\begin{tabular}{|c || c | c | c | c |}
\hline
Order & \# alm hypoham. & $g \geq 4$ & $g \geq 5$ & $g \geq 6$\\
\hline
$0-16$ & 0 & 0 & 0 & 0 \\
17 & 2 {\small(1)} & 2 (1) & 2 (1) & 0 \\
18 & 2 (1) & 2 (1) & 1 (1) & 0 \\
19 & ? & 27 (11) & 4 (4) & 0 \\
20 & ? & ? & 14 (6) & 0 \\
21 & ? & ? & 27 (18) & 0 \\
22 & ? & ? & 133 (86) & 0 \\
23 & ? & ? & 404 (161) & 0 \\
24 & ? & ? & $\ge 68$ & 0 \\
25 & ? & ? & ? & 0 \\
26 & ? & ? & ? & 0 \\
\hline
\end{tabular}
\caption{The number of almost hypohamiltonian graphs. The columns with a header of the form $g \geq k$ contain the number of almost hypohamiltonian graphs with girth at least~$k$. In parentheses the number of almost hypohamiltonian graphs which have a cubic exceptional vertex are given.}
\label{table:number_of_almhypoham_graphs}
\end{table}
\begin{figure}[h!t]
\centering
\includegraphics[width=0.4\textwidth]{small_alm_hypo.pdf}
\caption{The almost hypohamiltonian graph of order~17 from~\cite{Za15}. Its exceptional vertex is~$w$. By deleting the edge $vw$, we obtain the smallest almost hypohamiltonian graph both in terms of order and size.}
\label{fig:almhypo_17}
\end{figure}
\begin{figure}[h!t]
\centering
\includegraphics[trim = 0mm 0mm 22mm 16mm, clip,width=0.6\textwidth]{almost_hypoham-18v-1-girth4.pdf}
\caption{The smallest almost hypohamiltonian graph of girth~4. It has 18 vertices. Its exceptional vertex is~$w$.}
\label{fig:almhypo_18_g4}
\end{figure}
Our generator for almost hypohamiltonian graphs was obtained by extending the code of our generator for hypohamiltonian graphs from~\cite{GZ}. In that paper we describe how we extensively tested the correctness of our implementation. We also used multiple independent programs to test hamiltonicity and almost hypohamiltonicity---one of those programs was kindly provided to us by Gunnar Brinkmann---and in each case the results were in complete agreement. Furthermore, the source code of our new generator can be downloaded and inspected at~\cite{genhypo-site}.
\subsection{The cubic case}
\label{subsect:cubic_case}
The algorithm described in Section~\ref{subsect:generation_algo} can also be used to only generate cubic almost hypohamiltonian graphs, but it is much more efficient to use a generator for cubic graphs and test which graphs are almost hypohamiltonian as a filter.
We used the program \textit{snarkhunter}~\cite{brinkmann_11,BG15} to generate all cubic graphs with girth at least~4 up to 32~vertices and tested them for almost hypohamiltonicity. (Note that by Lemma~\ref{lemma:no_deg3_triangle} cubic almost hypohamiltonian graphs must have girth at least~4.) The results are presented in Table~\ref{table:counts_almhypo_cubic}. The smallest cubic almost hypohamiltonian graphs of girth~4 and~5 are shown in Figure~\ref{fig:cubic_almhypo}.
We also used the complete lists of snarks from~\cite{snark-paper} to determine all almost hypohamiltonian snarks up to 36~vertices. A \textit{snark} is a cyclically 4-edge-connected cubic graph with girth at least~5 which is not 3-edge-colourable. The number of almost hypohamiltonian snarks is listed in the last column of Table~\ref{table:counts_almhypo_cubic}. The sudden increase at 34~vertices is striking and also occurs for hypohamiltonian snarks as described in~\cite{snark-paper}.
\begin{figure}[h!t]
\centering
\subfloat[]{\label{fig:cubic_almhypo_26_g5}\includegraphics[trim = 0mm 0mm 22mm 16mm, clip, width=0.4\textwidth]{almost_hypoham-cubic-26v-1-girth5.pdf}} \qquad
\subfloat[]{\label{fig:cubic_almhypo_28_g4}\includegraphics[trim = 0mm 0mm 22mm 16mm, clip, width=0.4\textwidth]{almost_hypoham-cubic-28v-1-girth5.pdf}}
\caption{Figure~\ref{fig:cubic_almhypo_26_g5} depicts one of the ten smallest cubic almost hypohamiltonian graphs of girth~5. It has 26~vertices. Figure~\ref{fig:cubic_almhypo_28_g4} shows one of the four smallest cubic almost hypohamiltonian graphs of girth~4. It has 28~vertices. In both figures the exceptional vertex is labelled $w$.}
\label{fig:cubic_almhypo}
\end{figure}
All almost hypohamiltonian graphs from Table~\ref{table:counts_almhypo_cubic} can be downloaded from the \textit{House of Graphs}~\cite{hog} at
\url{http://hog.grinvin.org/AlmostHypohamiltonian} and also be inspected in the database of interesting graphs by searching for the keywords ``almost hypohamiltonian''.
Using \textit{snarkhunter} we also verified that there are no cubic almost hypohamiltonian graphs with girth 6 up to at least 36 vertices.
By combining the results from Table~\ref{table:counts_almhypo_cubic} with the results on non-hamiltonian cubic graphs from~\cite{GZ} we obtain the following bounds for cubic almost hypohamiltonian graphs.
\begin{theorem}
Let $c_g$ denote the order of the smallest cubic almost hypohamiltonian graph with girth~$g$. We have $$c_4 = 28, \quad c_5 = 26, \quad c_6 \ge 38, \quad c_7 \ge 44, \quad c_8 \ge 50, \quad {\rm {\it and}} \quad c_9 \ge 66.$$
\end{theorem}
In comparison, if we denote by $c'_g$ the smallest cubic hypohamiltonian graph with girth~$g$, we have the following. (See \cite{GZ} for more details.)
$$c'_4 = 24, \quad c'_5 = 10, \quad c'_6 = c'_7 = 28, \quad c'_8 \ge 50, \quad {\rm and} \quad c'_9 \ge 66.$$
%\begin{itemize}
%\item The smallest cubic almost hypohamiltonian graph with girth 7 has at least 44 vertices.
%\item The smallest cubic almost hypohamiltonian graph with girth 8 has at least 50 vertices.
%\item The smallest cubic almost hypohamiltonian graph with girth 9 has at least 66 vertices.
%\end{itemize}
\begin{table}
\centering
%\small
%\footnotesize
\scriptsize
\begin{tabular}{|c || r | r | r | r | r || r |}
\hline
\multirow{2}{*}{Order} & \multirow{2}{*}{$g \ge 4$} & Non-ham. & Alm. & Alm. hypo. & Alm. hypo. & Alm. hypo.\\
& & and $g \ge 4$ & hypo. & and $g \ge 5$ & and $g \ge 6$ & snark\\
% \# hypoham.\ $g \ge 5$ & Time (s) & Increase & Max edges added\\
\hline
6 & 1 & 0 & 0 & 0 & 0 & 0 \\
8 & 2 & 0 & 0 & 0 & 0 & 0 \\
10 & 6 & 1 & 0 & 0 & 0 & 0 \\
12 & 22 & 0 & 0 & 0 & 0 & 0\\
14 & 110 & 2 & 0 & 0 & 0 & 0\\
16 & 792 & 8 & 0 & 0 & 0 & 0\\
18 & 7 805 & 59 & 0 & 0 & 0 & 0\\
20 & 97 546 & 425 & 0 & 0 & 0 & 0\\
22 & 1 435 720 & 3 862 & 0 & 0 & 0 & 0 \\
24 & 23 780 814 & 41 293 & 0 & 0 & 0 & 0 \\
26 & 432 757 568 & 518 159 & 10 & 10 & 0 & 10 \\
28 & 8 542 471 494 & 7 398 734 & 6 & 2 & 0 & 2 \\
30 & 181 492 137 812 & 117 963 348 & 25 & 12 & 0 & 11 \\
32 & 4 127 077 143 862 & 2 069 516 990 & 74 & 4 & 0 & 0 \\
34 & ? & ? & ? & ? & ? & 6 253 \\
36 & ? & ? & ? & ? & ? & 2 243 \\
\hline
\end{tabular}
\caption{Counts of almost hypohamiltonian graphs among cubic graphs. $g$ stands for girth. Note that due to Lemma~\ref{lemma:no_deg3_triangle}, the girth of a cubic almost hypohamiltonian graph is at least~4. The last column denotes the number of almost hypohamiltonian snarks.}
\label{table:counts_almhypo_cubic}
\end{table}
\subsection{The planar case}
\label{subsect:planar_case}
Until recently, the smallest known planar almost hypohamiltonian graph had order~39, see~\cite{Za15}. \cite[Problem~3]{Za15} and \cite[Problem~6]{Za16} ask for the smallest order of a planar almost hypohamiltonian graph, and the smallest number $n_0$ such that there exists a planar almost hypohamiltonian graph of order~$n$ for every $n \ge n_0$. Although we are not able to fully settle the questions, we can report progress on both the lower and upper bounds of the order of the smallest planar almost hypohamiltonian graph, as well as an improvement of $n_0$.
We shall strengthen the second author's~\cite[Theorem 1]{Za15} and \cite[Theorem 4]{Za15}. Although Wiener recently has shown~\cite{Wi} the existence of a planar almost hypohamiltonian graph of order~31 (which constitutes the currently best upper bound for the order of the smallest planar almost hypohamiltonian graph), the planar almost hypohamiltonian graphs we will construct in Theorem~\ref{plalmhypo}---the smallest among them has 36 vertices---have a cubic exceptional vertex, while the exceptional vertex in Wiener's 31-vertex graph has degree~4. Cubic exceptional vertices are of particular interest due to the fact that we can combine two almost hypohamiltonian graphs, each having a cubic exceptional vertex, into a hypohamiltonian graph, see~\cite[Theorem 3]{Za15}---no method is known to do this if the exceptional vertex is non-cubic.
Let us first focus on the lower bound for the order of the smallest planar almost hypohamiltonian graph. Since the algorithm for generating all almost hypohamiltonian graphs presented in Section~\ref{subsect:generation_algo} only adds edges and never removes any vertices or edges, all graphs obtained by the algorithm from a non-planar graph will remain non-planar. So in case we only want to generate planar almost hypohamiltonian graphs, we can prune the construction when a non-planar graph is constructed. We used Boyer and Myrvold's algorithm~\cite{boyer2004cutting} to test if a graph is planar.
Following this approach, we showed that the smallest planar almost hypohamiltonian graph has at least 22 vertices and that the smallest planar almost hypohamiltonian graph of girth~4 must has at least 26 vertices. For girth~5 it turned out to be more efficient to use the program \textit{plantri}~\cite{brinkmann_07} instead. Using \textit{plantri} we generated all planar 3-connected graphs with girth~5 up to 48~vertices and tested them for almost hypohamiltonicity. This yielded the following results.
\begin{theorem}\label{thm:planar}
The smallest planar almost hypohamiltonian graph with girth~$5$ has $44$~vertices. There is exactly one such graph of that order and it is shown in Figure~\ref{fig:planar_almhypo_44}. There are exactly $42$~planar almost hypohamiltonian graphs with girth~$5$ on $47$~vertices. These are the only planar almost hypohamiltonian graphs with girth~$5$ up to at least $48$~vertices.
\end{theorem}
The graphs from Theorem~\ref{thm:planar} can be downloaded from the database of interesting graphs at the \textit{House of Graphs}~\cite{hog} by searching for the keywords ``planar almost hypohamiltonian''.
\begin{figure}[h!t]
\centering
\includegraphics[width=0.5\textwidth]{alm_hypoham_44v-g5-40except.pdf}
\caption{The smallest planar almost hypohamiltonian graph of girth~5. It has 44 vertices. Its exceptional vertex is~$w$.}
\label{fig:planar_almhypo_44}
\end{figure}
In summary, we have the following lower bounds.
\begin{corollary}
Let $\bar{h}$ ($\bar{h}_g$) denote the order of the smallest planar almost hypohamiltonian graph (of girth~$g$). We have $$\bar{h} \ge 22, \quad \bar{h}_4 \ge 26, \quad {\rm {\it and}} \quad \bar{h}_5 = 44.$$
\end{corollary}
Wiener~\cite{Wi} showed that there exists a planar almost hypohamiltonian graph of order~31 and girth~4, so $\bar{h} \le \bar{h}_4 \le 31$. (No planar almost hypohamiltonian graph of girth~3 is known. Adapting a technique of Thomassen~\cite{Th74-2} could provide examples of girth~3, as mentioned earlier.) In comparison, if we denote by $\bar{h}'$ ($\bar{h}'_g$) the order of the smallest planar hypohamiltonian graph (of girth~$g$), we have the following---more details can be found in~\cite{GZ}. $$23 \le \bar{h}' \le 40, \quad 23 \le \bar{h}'_3 \le 240, \quad 25 \le \bar{h}'_4 \le 40, \quad {\rm and} \quad \bar{h}'_5 = 45.$$
We now prepare for showing that there exists a planar almost hypohamiltonian graph of order~36 whose exceptional vertex is cubic, as well as improving $n_0$ (defined in the first paragraph of this subsection) from 76 (see~\cite{Za15}) to 73. Let $G$ be a graph containing a $4$-cycle $v_1 v_2 v_3 v_4 = C$. We denote by ${\rm Th}(G_C)$ the graph obtained from $G$ by deleting the edges $v_1 v_2$ and $v_3 v_4$ and adding a 4-cycle $v'_1 v'_2 v'_3 v'_4$ disjoint from $G$, and the edges $v_i v'_i$, $1 \le i \le 4$, to $G$. (Note that the operation Th preserves planarity. It was introduced by Thomassen~\cite{Th81} to show that there exist infinitely many planar cubic hypohamiltonian graphs.) The statement of Lemma~1 is a slight modification of a claim of Thomassen, see~\cite{Th81}; we skip its proof.
\begin{lemma}[Thomassen~\cite{Th81}]\label{Thom-lemma}
Let $G$ be a planar non-hamiltonian graph containing a quadrilateral face bounded by the cycle $C$. Then ${\rm Th}(G_C)$ is planar and non-hamiltonian.
\end{lemma}
In a planar graph, we call a face \emph{cubic} if all of its vertices are cubic. In~\cite{Th81}, Thomassen also showed that given a planar hypohamiltonian graph $G$ which contains a cubic quadrilateral face bounded by the cycle $C$, we have that ${\rm Th}(G_C)$ is a planar hypohamiltonian graph, as well. We require a modified version of Thomassen's result.
\begin{lemma}[Zamfirescu~\cite{Za15}]\label{almhypo-quad}
Let $G$ be a planar almost hypohamiltonian graph with exceptional vertex $w$, and let $G$ contain a cubic quadrilateral face bounded by the cycle $C$ such that $w \notin V(C)$. Then ${\rm Th}(G_C)$ is a planar almost hypohamiltonian graph, as well.
\end{lemma}
Consider graphs $G$ and $H$, and the cubic vertices $x \in V(G)$ and $y \in V(H)$. Denote by $G_x H_y$ one of the graphs obtained from
$G - x$ and $H - y$ by identifying the vertices in $N(x)$ with those in $N(y)$ using a bijection (so $G_x H_y$ has $|V(G)| + |V(H)| - 5$ vertices). Thomassen~\cite{Th74-1} showed that if $G$ and $H$ are hypohamiltonian, then $G_x H_y$ is hypohamiltonian, as well. Note that if $G$ is hypohamiltonian, then $G$ contains no triangle with a cubic vertex (as shown by Collier and Schmeichel~\cite[p.~196]{CS78}). We will need the following lemma.
\begin{lemma}[Zamfirescu~\cite{Za15}]\label{hypo+almhypo}
Let $G$ be an almost hypohamiltonian graph which contains a cubic vertex $x$ different from the exceptional vertex $w$ of $G$, and $H$ a hypohamiltonian graph which contains a cubic vertex $y$. Then $G_x H_y$ is an almost hypohamiltonian graph with exceptional vertex $w$. If $G$ and $H$ are planar, then so is $G_x H_y$.
\end{lemma}
A further ingredient is the following useful result of Grinberg.
\begin{theorem}[Grinberg's Criterion~\cite{Gr68}]\label{Grinberg}
Given a plane graph with a hamiltonian cycle ${\mathfrak h}$ and exactly $f_i$ ($f'_i$) $i$-gons inside (outside) of ${\mathfrak h}$, we have $$\sum\limits_{i \ge 3} (i-2)(f_i - f'_i) = 0.\qquad (\dagger)$$
\end{theorem}
\begin{theorem}\label{plalmhypo}
There exists a planar almost hypohamiltonian graph of order $n$ and with cubic exceptional vertex for (i) $n = 36 + 4k$, $k \ge 0$, and (ii) every $n \ge 73$.
\end{theorem}
\begin{proof}
(i) We first show that the graph $G$ from Figure~\ref{fig:planar_alm_G} is almost hypohamiltonian. By Grinberg's Criterion, $G$ is non-hamiltonian, since all faces of $G$ are pentagons with exactly one exception, which is a quadrilateral---now the left-hand side of ($\dagger$) cannot vanish. $G - w$ is non-hamiltonian, as well: denote by $Q = a'b'c'd'$ the quadrilateral in $G$, by $a'$ the cubic vertex of $Q$, and be $w$ the exceptional vertex of $G$, as shown in Figure~\ref{fig:planar_alm_G}. Assume that there exists a hamiltonian cycle ${\mathfrak h}$ in $G - w$. By Grinberg's Criterion~\cite{Gr68}, in $G - w$, $Q$ and the nonagon $N$ must lie on the same side of ${\mathfrak h}$. But this is impossible, as $a' \in V({\mathfrak h})$ and thus, $Q$ and $N$ lie on different sides of ${\mathfrak h}$. Hence, $G - w$ is non-hamiltonian.
\begin{figure}[h!t]
\centering
\subfloat[]{\label{fig:planar_alm_G}\includegraphics[width=0.4\textwidth]{almhypo36.pdf}} \qquad
\subfloat[]{\label{fig:planar_alm_G_prime}\includegraphics[width=0.4\textwidth]{almhypo36+Q.pdf}}
\caption{$G$ and $G'$ (i.e.\ Figure~\ref{fig:planar_alm_G} and~\ref{fig:planar_alm_G_prime}, respectively), two planar almost hypohamiltonian graphs used in the proof of Theorem~\ref{plalmhypo}. Their respective cubic exceptional vertices are labelled $w$.}
\end{figure}
We skip the straightforward verification that for every vertex $v \ne w$, the graph $G - v$ is hamiltonian. Thus, we have shown that there exists a planar almost hypohamiltonian graph of order~36 whose exceptional vertex is cubic. Note that we cannot apply Lemma~\ref{almhypo-quad} to $G$ since $a'b'c'd'$ is not a cubic quadrilateral, thus we require the following claim.
\smallskip
\noindent \textbf{Claim}. \emph{The graph $G'$ from Figure~\ref{fig:planar_alm_G_prime} is planar and almost hypohamiltonian with exceptional vertex $w$.}
%\begin{figure}[h!t]
%\centering
%\includegraphics[width=0.45\textwidth]{almhypo39.eps}
%\caption{The graph $G'$ used in the proof of Theorem~\ref{plalmhypo}.}
%\label{fig:G_prime}
%\end{figure}
\smallskip
\emph{Proof of the Claim.} Since $G' = {\rm Th}(G_Q)$, by Lemma~\ref{Thom-lemma} $G'$ is planar and non-hamiltonian. Let ${\mathfrak h}$ be a hamiltonian cycle in $G' - w$. Certainly, $aa'd' \subset {\mathfrak h}$. If $dd' \in E({\mathfrak h})$, then $b'baa'd'dcc' \subset {\mathfrak h}$, which replaced with $b'a'd'c'$ gives a hamiltonian cycle in $G - w$, a contradiction, so $dd' \notin E({\mathfrak h})$. Hence $a'adcbb' \subset {\mathfrak h}$, which replaced with $a'b'$ gives a hamiltonian cycle in $G - w$, once more a contradiction.
It is quickly verified that there exists a hamiltonian cycle ${\mathfrak h}$ in $G - a'$ containing the path $b'c'd'$. Deleting from ${\mathfrak h}$ the edge $c'd'$ we obtain the path ${\mathfrak p}$ in $G'$. Now ${\mathfrak p} \cup d'dabcc'$ is a hamiltonian cycle in $G' - a'$. Finding hamiltonian cycles in $G' - b'$, $G' - c'$, and $G' - d'$ works in a similar (and in fact even simpler) fashion. ${\mathfrak p} \cup d'a'adcc'b'$ is a hamiltonian cycle in $G' - b$, and ${\mathfrak p} \cup d'a'abcc'b'$ is a hamiltonian cycle in $G' - d$. It is easy to see that there exists a hamiltonian cycle ${\mathfrak h}'$ in $G - d'$ containing the edge $b'a'$. Deleting this edge from ${\mathfrak h}'$ we obtain the path ${\mathfrak p}$ in $G'$. Now ${\mathfrak p} \cup b'bcdd'a'$ is a hamiltonian cycle in $G' - a$, and ${\mathfrak p} \cup b'badd'a'$ is a hamiltonian cycle in $G' - c$.
It remains to show that $G' - v'$ is hamiltonian for every $v' \in V(G') \setminus \{ w, a, a', b, b',$ $c, c', d, d' \}$. Let $v$ be the vertex in $G$ corresponding to $v'$ in $G'$. Let ${\mathfrak h}''$ be a hamiltonian cycle in $G - v$. If $a'b' \in E({\mathfrak h}'')$, replace $a'b'$ with $a'adcbb'$. If $a'b' \notin E({\mathfrak h}'')$, then necessarily $a'd' \in E({\mathfrak h}'')$, which we replace with $a'abcdd'$. In both cases we have obtained a hamiltonian cycle in $G' - v'$, which finishes the proof of the Claim.
\smallskip
Using the Claim and iterating Lemma~\ref{almhypo-quad} ad infinitum, the proof of (i) is complete.
\smallskip
(ii) In \cite{JMOPZ}, it was shown that for every $n \ge 42$ there exists a planar hypohamiltonian graph of order $n$, which we will call $H^n$. For arbitrary but fixed $n \ge 42$, let $x \in V(H^n)$ be cubic (Thomassen showed that every planar hypohamiltonian graph contains a cubic vertex~\cite{Th78}), and let $y \in V(G)$ be cubic, as well. (Where $G$ is the graph shown in Figure~\ref{fig:planar_alm_G}.) Applying Lemma~\ref{hypo+almhypo}, we obtain that $H^n_x G_y$ is an almost hypohamiltonian graph with a cubic exceptional vertex.
\end{proof}
%The corollary below follows from theorem~\ref{plalmhypo} and \cite[Theorem 3]{Za15}.
%\begin{corollary}
%There exist planar hypohamiltonian graphs of order $n$ for $n = 70 + 4k$, $k \ge 0$, and every $n \ge 150$.
%\end{corollary}
By applying~\cite[Theorem~3]{Za15}, Theorem~\ref{plalmhypo}~(i) yields an alternative proof of the fact that infinitely many planar hypohamiltonian graphs exist. (This was first shown by Thomassen~\cite{Th76}. Chv\'atal~\cite{Ch73} had asked whether the statement is true, and Gr\"unbaum~\cite[p.~37]{Gr74} had conjectured that it is not.)
\subsection{The planar cubic case}
\label{subsect:planar_cubic_case}
Using the program \textit{plantri}~\cite{brinkmann_07} we generated all planar cyclically 4-edge-connected cubic graphs with girth at least~4 up to 52~vertices and tested them for almost hypohamiltonicity. No such graphs were found, so we have:
\begin{theorem}
The smallest planar cubic almost hypohamiltonian graph has at least $54$~vertices.
\end{theorem}
%\jg{Can we also give an upper bound (better than 68)? }
With \textit{plantri} we also generated all planar cyclically 4-edge-connected cubic graphs with girth~5 up to 78~vertices. This yielded the following results. These results were independently obtained by McKay (private communication) but were not published.
\begin{theorem}\label{thm:planar_cubic}
The smallest planar cubic almost hypohamiltonian graph of girth~$5$ has $68$~vertices. There are exactly three such graphs of that order and they are shown in Figure~\ref{fig:planar_cubic_g5}. There are exactly $81$~planar cubic almost hypohamiltonian graphs with girth~$5$ on $74$~vertices. These are the only planar cubic almost hypohamiltonian graphs with girth~$5$ up to at least $78$~vertices.
\end{theorem}
The graphs from Theorem~\ref{thm:planar_cubic} can be downloaded from the database of interesting graphs at the \textit{House of Graphs}~\cite{hog} by searching for the keywords ``planar cubic almost hypohamiltonian''.
%\jg{Add discussion: smallest planar cubic hypoham has 76 vertices (and also an example on 78 vertices, so no gap).}
\begin{figure}[h!t]
\centering
\subfloat[]{\includegraphics[width=0.28\textwidth]{Almost_hypohamiltonian_graphs_68v-g5-cyc4-1.pdf}} \quad
\subfloat[]{\includegraphics[width=0.28\textwidth]{Almost_hypohamiltonian_graphs_68v-g5-cyc4-2.pdf}} \quad
\subfloat[]{\includegraphics[width=0.28\textwidth]{Almost_hypohamiltonian_graphs_68v-g5-cyc4-3.pdf}}
\caption{All three smallest planar cubic almost hypohamiltonian graphs of girth~$5$. Each has 68~vertices. In each figure the exceptional vertex is labelled~$w$.}
\label{fig:planar_cubic_g5}
\end{figure}
The smallest planar cubic hypohamiltonian graph of girth~5 has order~76, as proven by McKay~\cite{Mc16}. In~\cite{GZ}, we showed that there exists exactly one such graph of order~78. Interestingly, McKay's three extremal graphs of order~76 have trivial automorphism group, whereas our 78-vertex example has $D_{3h}$ symmetry (as an abstract group, this is the dihedral group of order~12).
\section*{Acknowledgements}
Most computations for this work were carried out using the Stevin Supercomputer Infrastructure at Ghent University.
We also would like to thank Gunnar Brinkmann for providing us with an independent program for testing almost hypohamiltonicity and Brendan McKay for his advice on \textit{plantri}.
%% Bibliography
%\bibliographystyle{plain}
%\bibliographystyle{amcjoucc}
\bibliographystyle{siamplain}
\bibliography{references}
\end{document}
\begin{thebibliography}{10}
\bibitem{AT73}
{\sc G.~M. Adel'son-Vel'ski\v{\i} and V.~K. Titov}, {\em On edge 4-chromatic
cubic graphs (in {R}ussian)}, in Proc. Seminar of 1971 at Moscow Univ.,
Voprosy Kibernetiki [vol. not numbered], 1973, pp.~5--14.
\bibitem{AMW97}
{\sc R.~E.~L. Aldred, B.~D. McKay, and N.~C. Wormald}, {\em Small
hypohamiltonian graphs}, Journal of Combinatorial Mathematics and
Combinatorial Computing, 23 (1997), pp.~143--152.
\bibitem{boyer2004cutting}
{\sc J.~M. Boyer and W.~J. Myrvold}, {\em {On the Cutting Edge: Simplified
$O(n)$ Planarity by Edge Addition}}, Journal of Graph Algorithms and
Applications, 8 (2004), pp.~241--273.
\bibitem{hog}
{\sc G.~Brinkmann, K.~Coolsaet, J.~Goedgebeur, and H.~M{\'e}lot}, {\em {House
of Graphs: a database of interesting graphs}}, Discrete Applied Mathematics,
161 (2013), pp.~311--314.
\newblock Available at \url{http://hog.grinvin.org/}.
\bibitem{BG15}
{\sc G.~Brinkmann and J.~Goedgebeur}, {\em Generation of cubic graphs and
snarks with large girth}, arXiv preprint arXiv:1512.05944, (2015).
\bibitem{snark-paper}
{\sc G.~Brinkmann, J.~Goedgebeur, J.~H{\"a}gglund, and K.~Markstr{\"o}m}, {\em
Generation and properties of snarks}, Journal of Combinatorial Theory, Series
B, 103 (2013), pp.~468--488.
\bibitem{brinkmann_11}
{\sc G.~Brinkmann, J.~Goedgebeur, and B.~D. McKay}, {\em Generation of cubic
graphs}, Discrete Mathematics and Theoretical Computer Science, 13 (2011),
pp.~69--80.
\bibitem{brinkmann_07}
{\sc G.~Brinkmann and B.~D. McKay}, {\em Fast generation of planar graphs},
MATCH Commun. Math. Comput. Chem., 58 (2007), pp.~323--357.
\bibitem{CMRS03}
{\sc A.~Cavicchioli, T.~E. Murgolo, B.~Ruini, and F.~Spaggiari}, {\em Special
classes of snarks}, Acta Applicandae Mathematica, 76 (2003), pp.~57--88.
\bibitem{CKL70}
{\sc G.~Chartrand, S.~F. Kapoor, and D.~R. Lick}, {\em $n$-hamiltonian graphs},
Journal of Combinatorial Theory, 9 (1970), pp.~308--312.
\bibitem{Ch73}
{\sc V.~Chv\'atal}, {\em Flip-flops in hypohamiltonian graphs}, Canadian
Mathematical Bulletin, 16 (1973), pp.~33--41.
\bibitem{CS78}
{\sc J.~B. Collier and E.~F. Schmeichel}, {\em Systematic searches for
hypohamiltonian graphs}, Networks, 8 (1978), pp.~193--200.
\bibitem{Di10}
{\sc R.~Diestel}, {\em Graph {T}heory}, Graduate Texts in Mathematics, 173
(2010).
\bibitem{Fi83}
{\sc S.~Fiorini}, {\em Hypohamiltonian snarks}, in Graphs and Other
Combinatorial Topics (ed.: M. Fiedler), vol.~59, Teubner, 1983, pp.~70--75.
\bibitem{GSZ}
{\sc J.~Goedgebeur, E.~Steffen, and C.~T. Zamfirescu}, {\em On hypohamiltonian
snarks and fiorini's theorem}, In preparation.
\bibitem{genhypo-site}
{\sc J.~Goedgebeur and C.~T. Zamfirescu}, {\em Homepage of genhypohamiltonian:
\url{http://caagt.ugent.be/hypoham/}}.
\bibitem{GZ}
{\sc J.~Goedgebeur and C.~T. Zamfirescu}, {\em Improved bounds for
hypohamiltonian graphs}, arXiv preprint arXiv:1602.07171, (2016).
\bibitem{Gr68}
{\sc E.~J. Grinberg}, {\em Plane homogeneous graphs of degree three without
hamiltonian circuits (in {R}ussian)}, Latvian Math. Yearbook, 4 (1968),
pp.~51--58.
\bibitem{Gr74}
{\sc B.~Gr{\"u}nbaum}, {\em Vertices missed by longest paths or circuits},
Journal of Combinatorial Theory, Series A, 17 (1974), pp.~31--38.
\bibitem{HS93}
{\sc D.~A. Holton and J.~Sheehan}, {\em The {P}etersen graph, chapter 7:
Hypohamiltonian graphs}, 1993.
\bibitem{Is75}
{\sc R.~Isaacs}, {\em {Infinite families of nontrivial trivalent graphs which
are not Tait colorable}}, American Mathematical Monthly, 82 (1975),
pp.~221--239.
\bibitem{JMOPZ}
{\sc M.~Jooyandeh, B.~D. McKay, P.~R.~J. {\"O}sterg{\aa}rd, V.~H. Pettersson,
and C.~T. Zamfirescu}, {\em Planar hypohamiltonian graphs on $40$ vertices},
To appear in: Journal of Graph Theory, (2016).
\bibitem{Mc16}
{\sc B.~D. McKay}, {\em Hypohamiltonian planar cubic graphs with girth $5$}, To
appear in: Journal of Graph Theory, (2016).
\bibitem{Sk16}
{\sc Z.~Skupie\'n}, {\em Multi-compositions in exponential counting of
hypohamiltonian snarks}, in Convexity and Discrete Geometry Including Graph
Theory, Springer Proc. in Math. \& Stat. (eds.: K. Adiprasito, I. B\'ar\'any,
C. V\^{i}lcu), vol.~148, Springer, 2016, pp.~47--58.
\bibitem{So63}
{\sc R.~Sousselier}, {\em Probl\`{e}me no. 29: Le cercle des irascibles (in
{F}rench)}, Revue Fran{\c{c}}aise de Recherches Op{\'e}rationelle, 7 (1963),
pp.~405--406.
\bibitem{Th74-1}
{\sc C.~Thomassen}, {\em Hypohamiltonian and hypotraceable graphs}, Discrete
Mathematics, 9 (1974), pp.~91--96.
\bibitem{Th74-2}
{\sc C.~Thomassen}, {\em On hypohamiltonian graphs}, Discrete Mathematics, 10
(1974), pp.~383--390.
\bibitem{Th76}
{\sc C.~Thomassen}, {\em Planar and infinite hypohamiltonian and hypotraceable
graphs}, Discrete Mathematics, 14 (1976), pp.~377--389.
\bibitem{Th78}
{\sc C.~Thomassen}, {\em Hypohamiltonian graphs and digraphs}, in Theory and
Applications of Graphs, Springer, 1978, pp.~557--571.
\bibitem{Th81}
{\sc C.~Thomassen}, {\em Planar cubic hypohamiltonian and hypotraceable
graphs}, Journal of Combinatorial Theory, Series B, 30 (1981), pp.~36--44.
\bibitem{Wi}
{\sc G.~Wiener}, {\em New constructions of hypohamiltonian and hypotraceable
graphs}, In preparation.
\bibitem{Za15}
{\sc C.~T. Zamfirescu}, {\em On hypohamiltonian and almost hypohamiltonian
graphs}, Journal of Graph Theory, 79 (2015), pp.~63--81.
\bibitem{Za16}
{\sc C.~T. Zamfirescu}, {\em Seven problems on hypohamiltonian and almost
hypohamiltonian graphs}, in Convexity and Discrete Geometry Including Graph
Theory, Springer Proc. in Math. \& Stat. (eds.: K. Adiprasito, I. B\'ar\'any,
C. V\^{i}lcu), vol.~148, Springer, 2016, pp.~253--255.
\end{thebibliography}