Open Problems - Graph Theory and Combinatorics
This site was intended as a resource for research in graph theory and
combinatorics but is now long neglected.
Also available is a Glossary of Terms.
A more thorough collection of
open problems and information about them appears at the
Open Problem Garden.
This directory mostly lists some well-known problems for which more detailed
pages may be created later.
The organization of topics roughly follows the four volumes of The Art of
Combinatorics under development by D.B. West. Thus the four main headings
are
Extremal Graph Theory,
Structure of Graphs,
Order and Optimization, and
Arrangements and Methods.
Note: Here is a discussion of the notation
for the number of vertices and the number of edges of a graph G.
Other directories of open problems pages can be found as follows:
Graph Theory,
Combinatorics,
Optimization.
(emphasizing graph theory, combinatorics, number theory, and discrete
geometry) is at the
Open Problem Garden at
Simon Fraser University.
Topics in this section include
distance,
matching and independence,
coloring,
perfect graphs,
classical extremal problems,
etc.
Diameter
Average distance
- Average distance in vertex-transitive graphs (Alan Kaplan): (in a
vertex-transitive graph, the average distance from a given vertex to the other
vertices exceeds half the diameter; PROVED in a more general context by Mark
Herman and Jonathan Pakianathan -- see arXiv article)
Matchings and Factors
Independent Sets
Domination
Vertex coloring
- Erdős-Faber-Lovász Conjecture (every union
of n pairwise edge-disjoint copies of Kn is
n-colorable)
- Alon-Saks-Seymour Conjecture (every
union of m pairwise edge-disjoint complete bipartite graphs is
(m+1)-colorable - FALSE)
- Lovász Doubly Critical Graph Conjecture (if χ(G-x-y)=χ(G)-2
for every edge xy in G, then G is a complete graph)
- Gyárfás-Sumner Conjecture (for every forest T, there exists a
function fT such that every graph G with no induced
subgraph isomorphic to T has chromatic number at most
f(ω(G)))
- Reed's upper bound (for every graph G, the chromatic number
is at most the ceiling of (1+Δ(G)+ω(G))/2
- Dense triangle-free graph conjecture (G
is 4-colorable if G is triangle-free and δ(G)≥n(G)/3)
- Brandt's regular supergraph conjecture (every
maximal triangle-free graph G with δ(G)≥n(G)/3 has a
regular supergraph obtained by vertex multiplications)
Edge-coloring
- Goldberg-Seymour Conjecture (every multigraph G has a proper
edge-coloring using at most max{μ(G),Δ(G)+1} colors, where
μ(G) is the maximum of
⌈|E(H)|/(½(|V(H)|-1))⌉ over
subgraphs H of G) with |V(H)| odd)
- Overfull Conjecture (Class 2 requires overfull
subgraph when Δ(G)>n/3)
- 1-Factorization Conjecture (if G is a
2m-vertex regular graph with degree at least 2⌈m/2⌉-1,
then G is 1-factorable --- implied by Overfull
Conjecture)
- Total Coloring Conjecture (the vertices and edges of a graph G can
be colored with Δ(G)+2 colors such that adjacent vertices have
different colors, incident edges have different colors, and incident edge
and vertices have different colors)
- Strong Chromatic Index (for edge-colorings
in which distinct edges of the same color cannot have adjacent vertices, what
are the best bounds on the number of colors needed in terms of the maximum
degree?)
List Coloring
- List Coloring Conjecture (edge-choosability equals edge-chromatic number)
- Square List Coloring Conjecture (choosability equals chromatic number) for the square of every graph
- 4-Choosability of 5-connected planar graphs (would imply 4-color Theorem;
all known planar graphs that are not 4-choosable are not 5-connected -
Kawarabayashi-Toft)
- List coloring of locally sparse graphs (for graphs with maximum degree
D and maximum codegree cD, choosability is bounded by
(c+o(1))D, where the codegree of vertices u and v
is their number of common neighbors)
Graph Homomorphism and Circular Coloring
A homomorphism from a graph G to a graph H is a
function f: V(G)→V(H) such that the images of adjacent vertices
are adjacent. A (k,d)-coloring of G is an assignment of
congruence classes modulo k to vertices of G such that
adjacent vertices have colors differing by at least d. Such an
assignment is a homomorphism from G into the complement of the
dth power of a k-cycle. The circular chromatic number
χc of G is the minimum ratio k/d such that
G has a (k,d)-coloring; the ceiling of this is the ordinary
chromatic number. A homomorphism from G into H is an
H-coloring of G.
- Jaeger's Conjecture, special case (every planar graph with girth at least
4k is C2k+1-colorable)
- Albertson-Chan-Haas Problem (is it true that
every n-vertex graph with odd girth at least 2k+1 and minimum
degree greater than 3n/(4k) is C2k+1-colorable?)
Generalized Coloring
Stronger Coloring Requirements
- L(2,1)-labeling (every graph with maximum
degree k can be colored with colors 0,...,k2 such that
the colors on adjacent vertices differ by at least 2 and the colors on
vertices at distance 2 differ by at least 1)
- Equitable Coloring Conjecture (a connected graph with maximum degree
k has a proper k-coloring with color classes differing in
size by at most 1 if and only if it is not an odd cycle, a complete graph,
or the complete bipartite graph Kk,k with k
odd)
Hypergraphs
Graph Representations
Intersection Graphs
Forced Subgraphs
- Erdős-Sós Conjecture (every graph with average degree greater than
m-1 contains every tree with m edges)
Graph Decomposition
- Gallai's Conjecture (every n-vertex graph decomposes into
⌈n/2⌉ paths)
- Hajós' Conjecture (every even n-vertex graph decomposes into
⌊n/2⌋ cycles)
- Linear Arboricity Conjecture (every graph decomposes into
⌈(Δ(G)+1)/2⌉ unions of disjoint paths)
- Favaron-Genest-Kouider Conjecture [2010] (every (2k+1)-regular graph
having a perfect matching decomposes into paths with 2k+1 edges;
proved for k=1 by Kotzig, for k=2 in the triangle-free case
by Botler-Mota-Wakabayashi [2015])
- Nash-Williams Conjecture [1970] (for sufficiently large n, every
n-vertex graph with vertex degrees even, number of edges divisible by
3, and minimum degree at least 3n/4 decomposes into triangles
Graph Ramsey theory
anti-Ramsey problems
Topics in this section include
existence questions,
connectivity,
cycles,
planarity and topological graph theory,
graph minors,
integer flows,
algebraic graph theory,
etc.
Isomorphism
- Kelly-Ulam Reconstruction Conjecture (every graph with at least 3 vertices
is reconstructible from its deck of single vertex-deleted subgraphs)
Decomposition
- Ringel Tree Decomposition Conjecture (if T is a tree of
order n, then K2n-1 decomposes into copies of
T
- Tree Packing Conjecture (any trees of sizes 1,...,n-1 decompose
Kn)
- Alspach's Conjecture (if n is odd and
c1,...,cm are integers between 3 and n
that sum to (2n), then Kn
decomposes into cycles of lengths c1,...,cm)
- Hoffman-Singleton packing (does
K50 decompose into 7 copies of the Hoffman-Singleton
graph?)
Labelings
- Kotzig-Ringel Graceful Tree Conjecture (every tree has a vertex numbering
with 1,...,n such that n-1 distinct differences appear on the
edges)
Packing
Vertex Degrees
Digraph Partition Problem (Thomassen) - There exists a least number
f(s,t) such that the vertices of any simple digraph with minimum
outdegree at least f(s,t) can be partitioned into two sets inducing
subdigraphs with minimum outdegree at least s and at least t,
respectively. [It is known that f(1,1)=3.
Tournaments
Graph Products
- Jaeger's Dual-Hamiltonian Conjecture (every cyclically 4-connected cubic
graph G has a bond of size e(G)-n(G)+2; equivalently, every
cylically 4-connected cubic graph decomposes into two trees)
Digraphs
- Caccetta-Häggkvist Conjecture (every simple
n-vertex digraph with minimum outdegree at least r has a cycle of
length at most the ceiling of n/r)
- Bermond-Thomassen Conjecture [1981] on Disjoint Cycles (every simple
digraph with minimum outdegree at least 2k-1 has k disjoint
cycles)
Cycle lengths
Cages
A (k,g)-cage is a graph with minimum order among all
k-regular graphs with girth g
- Fu-Huang-Rodger Conjecture (every
(k,g)-cage is k-connected)
- Bipartite Cage Conjecture (every cage with even girth is bipartite;
Pisanski-Boben-Marusic-Orabnic-Graovac)
Longest paths
- Hitting all longest paths (does every chordal
graph have a vertex that appears in all longest paths?
What is the largest q such that in every connected graph, every set of
q longest paths have a common vertex?)
Longest cycles (and Hamiltonian cycles)
- Chvatal's Conjecture (some value of toughness is a sufficient to force
a spanning cycle)
- Barnette's Conjecture (every planar 3-connected 3-regular bipartite graph
is Hamiltonian)
- Seymour's kth-power Hamiltonian Cycle
Conjecture (every n-vertex
graph with minimum degree at least kn/(k+1) contains
Cnk)
- Zhang's Hamiltonian weight conjecture
(every 3-regular graph having a Hamiltonian weight arises from
K4 via Delta-Wye operations)
Hamiltonian decomposition
Definition: A Hamiltonian decomposition is a decomposition of a
regular graph into spanning cycles (if the degree is even) or into
spanning cycles and a single 1-factor (if the degree is odd).
- Nash-Williams' Conjecture (every 2k-regular graph with at most
4k+1vertices has a Hamiltonian decomposition)
- Bermond's Conjecture (the cartesian product of
two graphs having decompositions into Hamiltonian cycles also has such a
decomposition)
Subgraphs of Planar Graphs
- Induced forests in planar graphs (does every
planar subgraph have an induced forest with at least half of the vertices?
an induced linear forest with at least 4/9 of the vertices?)
Coloring of Graphs on Surfaces
Parameters of Planar Graphs
Crossing Number
The crossing number cr(G) of a graph G is the
minimum number of edge-crossings in a drawing of G in the plane.
In an optimal drawing, it may be assumed that edges cross at most once,
that no vertex is an internal point of an edge, that no three edges share
an internal point, and that no two edges are tangent.
- Zarankiewicz Conjecture (cr(Km,n) =
⌊m/2⌋⌊(m-1)/2⌋⌊n/2⌋⌊(n-1)/2⌋)
- cr(Kn) =
¼⌊n/2⌋⌊(n-1)/2⌋⌊(n-2)/2⌋⌊(n-3)/2⌋
- Crossing number of cartesian product of Cm and
Cn is m(n-2) for m≤n
- Are the pair-crossing number, the odd-crossing number, and the
independent odd-crossing number always equal to the crossing number?
(No, the odd-crossing number can be smaller)
Embeddings on Surfaces
An integer flow on a graph is a pair consisting of an orientation of
the graph and an assignment of integer weights to the edges such that for each
vertex the total weight on exiting edges equals the total weight on entering
edges. It is a k-flow if all weight have absolute value less than
k, and it is nowhere-zero if weight 0 is never used. A graph
having a nowhere-zero k-flow is k-flowable; this is a dual
notion to k-colorability. Note that every nowhere-zero k-flow
is a nowhere-zero k+1-flow.
Coverings and Embeddings
- Strong Embedding Conjecture
(every 2-connected
graph has an embedding on some orientable surface such that every face boundary
is a cycle)
- Cycle Double Cover Conjecture (every
2-edge-connected graph has an exact double covering of its edges by cycles)
- Fulkerson (or Berge-Fulkerson) Conjecture
(Archdeacon,
Mohar)
(every 2-connected 3-regular graph has an exact double covering of its edges
by six matchings)
Snarks
- Nedela-Skoviera Conjecture (every irreducible snark has girth at most 6)
- Jaeger-Swart Conjecture (every snark has cyclic edge-connectivity at most 6)
Graphs and Vector Spaces
Graph Eigenvalues
Distance-Regular Graphs
Topics in this section include
structure of posets,
linear extensions,
extremal problems on posets,
linear and integer programming,
matroids and related topics,
etc.
Antichains and Sperner Theory
Chain Decompositions
- Saturated Partitions & LYM orders (does every LYM order have a chain
partition that is k-saturated for every k?)
- L(m,n) (does the poset consisting of nondecreasing nonnegative integer
m-tuples bounded by n, ordered by inclusion, have a symmetric
chain decomposition?)
Lattices
Structure of Special Families
Dimension of Partial Orders
The dimension of a partially ordered set P is the minimum number
of linear extensions of P whose intersection is P. That is, it
is the minimum number of linear orders on the elements of P such that
x < y in P if and only if x < y in each of the linear
orders. Equivalently, it is the minimum t such that P is a
subposet of Rt under the componentwise order.
- Removable Pair Conjecture (every poset has two
points whose removal drops the dimension by at most 1)
- Kelly-Trotter Product Conjecture (dim(P x Q)≥dim(P)+dim(Q)-2)
Sorting and Searching
- 1/3-2/3 Conjecture (every non-chain poset has two elements x and
y such that the fraction of the linear extensions with x above
y is between 1/3 and 2/3)
Order Ideals
- Chvatal's Conjecture (in every ideal of sets, the largest intersecting
family consists of the sets containing one specified element)
Families of Finite Sets (hypergraphs)
General Partial Orders
Incidence Algebra and M?bius Functions
Packing and Covering
Network Flow
Polyhedral Methods
Matroids
Antimatroids and Greedoids
Oriented Matroids
Topics in this section include
enumeration,
probabilistic methods,
numbers and games,
designs and coding,
discrete geometry,
etc.
Permutations
- The Pancake Problems (what is the worst-case
number of prefix reversals needed to sort a permutation (or a signed
permutation) of [n]?)
Classical Enumeration
Asymptotic Enumeration
Theory of Generating Functions
Random Structures
Ramsey Numbers
- Diagonal Ramsey Problem (is there a limit for log(R(k,k))/k,
and if so, what is it?)
Combinatorial Number Theory
Special Combinatorial Configurations
Combinatorial Gray codes
Named for the classical Gray code listing binary vectors (cyclically) with
one bit change between successive vectors, A combinatorial Gray code is
a listing of the objects in a set using only specified changes between
successive objects. The last item should also be close to the first, so what
is sought is a Hamiltonian cycle in the graph defined by the permitted
adjacencies.
- Revolving Door (Middle Levels) Conjecture
(there is a cycle through the subsets of [2k+1] with sizes k
and k+1 by adding or deleting one element at each step) ---
UPDATE: This has been proved by Torsten Mütze
- Traversal by Prime Sum (for m≥2,
does the graph with vertex set [2m] and edges joining numbers whose sum
is prime always have a Hamiltonian cycle?)
Block Designs
Coding Theory
Configurations of Points
Hyperplane Arrangements
Geometric Graphs