Chapter 2

Fundamentals of Graph Theory

[52/56%]
Show
LaTeX
Problem 2.1.1

Every graph has a bipartite subgraph containing at least half of the edges of the graph.

?
Problem 2.1.2

A path in a graph is a sequence v1e1v2e2…en−1vnv_{1} e_{1} v_{2} e_{2} \ldots e_{n-1} v_{n} in which, for every ii, the edge eie_{i} joins the vertices viv_{i} and vi+1v_{i+1}. The number n−1n-1 is called the length of the path. (The edges e1,e2,…,en−1e_{1}, e_{2}, \ldots , e_{n-1} need not be pairwise distinct.)

A cycle in a graph is a sequence v1e1v2e2…en−1vnenv_{1} e_{1} v_{2} e_{2} \ldots e_{n-1} v_{n} e_{n} in which, for every i<ni < n, the edge eie_{i} joins the vertices viv_{i} and vi+1v_{i+1}, and the edge ene_{n} joins the vertices vnv_{n} and v1v_{1}. Cycles are considered the same if they differ by a cyclic shift of the sequence. The number nn is called the length of the cycle.

A cycle is called non-self-intersecting if its vertices v1,v2,…,vnv_{1}, v_{2}, \ldots , v_{n} are pairwise distinct and its edges e1,e2,…,ene_{1}, e_{2}, \ldots , e_{n} are pairwise distinct. The standard term (less convenient for beginners) is a simple cycle.

?
(1)

Every cycle that does not pass along any edge twice contains a non-self-intersecting cycle.

(2)

Every cycle of odd length contains a non-self-intersecting cycle of odd length.

(3)

Does the analogous statement hold for cycles of even length that do not pass along any edge twice?

(4)

A graph has a non-self-intersecting cycle passing through the edges aa and bb, and also a non-self-intersecting cycle passing through the edges bb and cc. Then there is a non-self-intersecting cycle passing through the edges aa and cc.

Problem 2.1.3

If the degree of each of the nn vertices of a graph is greater than n2−1\frac{n}{2}-1, then the graph is connected.

?
Problem 2.1.4

Let a directed graph GG be given, with a weight f(u)f(u) written on each edge uu. (This weight can be understood as the work that must be expended in order to traverse the edge from its start to its end.) A function p:V(G)→Rp: V(G) \to \mathbb {R} (a potential'') such that $f(x,y) = p(x) - p(y)$ for every edge $u = (x,y)$ exists if and only if the sum of the weights of the edges of every cycle equals zero (when an edge is traversed along the cycle in the direction opposite to its orientation, its weight enters the sum with a minus'' sign).

?
Problem 2.2.1
?
(1)

Every tree has a leaf, i.e., a vertex of degree 1.

(2)

Every tree with nn vertices has n−1n-1 edges.

(3)

In every tree, between any two vertices there exists a unique non-self-intersecting path.

(4)

A sequence of nn positive integers is the degree sequence of some tree if and only if the sum of its terms equals 2n−22n-2.

Problem 2.2.2

Note that the graphs ({1,2,3},{{1,2}})(\left\{ 1,2,3\right\} , \left\{ \left\{ 1,2\right\} \right\} ) and ({1,2,3},{{1,3}})(\left\{ 1,2,3\right\} , \left\{ \left\{ 1,3\right\} \right\} ) are distinct. A graph means precisely a graph, not an isomorphism class of graphs. Or, informally speaking, the vertices of graphs are considered to be numbered. Therefore the term labeled graph'' is sometimes used instead of the word graph''.

Which graphs with nn given vertices are more numerous:

?
(1)

those having an isolated vertex or those not having one?

(2)

connected ones or disconnected ones?

Problem 2.2.3
?
(1)

Cayley's formula. The number of trees with nn given vertices equals nn−2n^{n-2}.

(2)

If the sum of positive integers d1,…,dnd_{1}, \ldots , d_{n} equals 2n−22n-2, then the number of trees with nn given vertices in which the ii-th vertex has degree did_{i} equals (n−2)!(d1−1)!⋅…⋅(dn−1)!\dfrac {(n-2)!}{(d_{1}-1)! \cdot \ldots \cdot (d_{n}-1)!}.

This statement can be reformulated as follows:

(x1+…+xn)n−2=∑Tx1deg⁡T(1)−1⋅…⋅xndeg⁡T(n)−1, (x_{1}+\ldots +x_{n})^{n-2} = \sum _{T} x_{1}^{\deg _{T}(1)-1} \cdot \ldots \cdot x_{n}^{\deg _{T}(n)-1},

where the sum is taken over all trees TT with vertices 1,2,…,n1,2,\ldots ,n, and deg⁡T(k)\deg_{T}(k) denotes the degree of the vertex kk of the tree TT.

(3)

Let T1,…,TrT_{1}, \ldots , T_{r} be trees whose vertex sets are pairwise disjoint. How many trees are there whose vertex set is the union of the vertex sets of these rr trees and which contain T1,…,TrT_{1}, \ldots , T_{r}?

Problem 2.2.4

The Prüfer code assigns to a tree with vertices 1,2,…,n1,2,\ldots ,n a sequence of numbers from 1 to nn by the following algorithm.

Initially the Prüfer code is the empty word. While the number of vertices is greater than two:

  1. a leaf (see Problem 2.2.1) vv with the smallest number is chosen;

  2. the number of the vertex adjacent to vv is appended to the Prüfer code;

  3. the vertex vv and the edge incident to it are removed from the tree. When two vertices remain, the algorithm terminates.

?
(1)

Find the Prüfer code of the tree with vertices 1,2,…,101,2,\ldots ,10 and edges (8,9),(8,4),(4,10),(10,3),(3,5),(10,6),(10,1),(1,7),(1,2)(8,9), (8,4), (4,10), (10,3), (3,5), (10,6), (10,1), (1,7), (1,2).

(2)

Reconstruct the tree from the Prüfer code 1,1,2,5,4,2,71,1,2,5,4,2,7.

(3)

The Prüfer code defines a bijection between the set of trees with nn given vertices and the set of words of length n−2n-2 over the numbers from 1 to nn.

(4)

In the Prüfer code, a vertex of degree dd occurs d−1d-1 times.

Problem 2.2.5

A graph is called unicyclic if it becomes a tree after the removal of some edge. (Or, equivalently, if it is connected and has exactly one non-self-intersecting cycle, up to cyclic shift and symmetry.)

?
(1)

Which are more numerous: trees with 100 given vertices or unicyclic graphs with 98 given vertices?

(2)

Express the number of unicyclic graphs with nn given vertices as a sum of at most nn terms.

Problem 2.2.6
?
(1)

A tree has no nonempty subgraphs in which the degree of every vertex is even and positive.

(2)

For a graph GG, denote by h1(G)h_{1}(G) the number of its subgraphs without isolated vertices in which the degree of every vertex is even. (The empty subgraph satisfies this condition.) Prove that h1(G)h_{1}(G) is a power of two. Express h1(G)h_{1}(G) in terms of the number nn of vertices, the number ee of edges, and the number kk of connected components of the graph.

(3)

The edges of a tree are marked with the signs ++ and −-. It is allowed to change the signs on all edges issuing from one vertex. Then any arrangement can be obtained from any other.

(4)

For a graph GG, denote by h1(G)h^{1}(G) the largest number of arrangements of the signs ++ and −- on its edges such that none of them can be obtained from another by the operations described above. Prove that h1(G)h^{1}(G) is a power of two. Express h1(G)h^{1}(G) in terms of nn, ee, and kk.

(5)

Prove that h1(G)h_{1}(G) and h1(G)h^{1}(G) do not change under the contraction of an edge, and deduce from this that h1(G)=h1(G)h_{1}(G) = h^{1}(G).

Remark.
?

Remark. Such subgraphs (from part (2)) are called cycles in the sense of homology theory (not to be confused with cycles in the sense of graph theory). How they arise is described, for example, in [S1, 6].

Remark. In cohomology theory such arrangements (from part (4)) are called cocycles, and the equivalence relation on cocycles given above is called cohomologousness. How cohomologousness of cocycles arises is described, for example, in [S4, Sec. 11.1, 11.2].

Problem 2.3.1

Roughly speaking, graphs are isomorphic if they are the same (although their drawings in the plane may differ). Formally, graphs G1G_{1} and G2G_{2} are called isomorphic if there exists a one-to-one mapping f:V(G1)→V(G2)f: V(G_{1}) \to V(G_{2}) satisfying the condition: vertices A,B∈V(G1)A, B \in V(G_{1}) are joined by an edge if and only if the vertices f(A),f(B)∈V(G2)f(A), f(B) \in V(G_{2}) are joined by an edge.

Which of the graphs in Fig. 3 are isomorphic?

Fig. 3. Which of the graphs in the figure are isomorphic?Fig. 3. Which of the graphs in the figure are isomorphic?

?
Problem 2.3.2

For arbitrary k,l,m,n∈Nk,l,m,n \in \mathbb {N}, find the number of

?
(1)

cliques of size kk in the graph KnK_{n},

(2)

cliques of size kk in the graph Km,nK_{m,n},

(3)

independent sets of size kk in the graph KnK_{n},

(4)

independent sets of size kk in the graph Km,nK_{m,n},

(5)

subgraphs of KnK_{n} isomorphic to Kk,lK_{k,l},

(6)

subgraphs of Km,nK_{m,n} isomorphic to Kk,lK_{k,l}.

Be careful: these problems are simple, but almost all of them require a case analysis.

Problem 2.3.3

List all pairwise non-isomorphic

?
(1)

graphs with four vertices,

(2)

connected graphs with five vertices and five edges,

(3)

disconnected graphs with five vertices.

Problem 2.3.4

How many pairwise non-isomorphic graphs with 8 vertices and 25 edges are there?

?
Problem 2.3.5

The number of isomorphism classes of trees with nn vertices (i.e., the number of distinct trees with nn unlabeled vertices) is less than 4n4^{n}.

?
Problem 2.4.1

A plane graph with triangular faces and more than three vertices is given. A vertex was deleted together with the edges emanating from it.

?
(1)

Is it true that the resulting face is bounded by a non-self-intersecting cycle?

(2)

Is it true that if one more vertex is removed, then all faces will again be bounded by non-self-intersecting cycles?

(3)

Suppose that in the resulting graph every vertex has degree at least 3. Is it true that any vertex of the new graph can be deleted so as to obtain a graph all of whose faces are bounded by non-self-intersecting cycles?

Remark.
?

A plane graph is a drawing of a graph in the plane in which any two edges intersect only at their common vertices (in particular, if there are no such vertices, they do not intersect). A plane graph divides the plane into parts called the faces of the graph; one of these parts is ``infinite''.

Problem 2.4.2
?
(1)

Euler's formula. For any connected plane graph with ff faces, the equality n−e+f=2n - e + f = 2 holds.

(For a proof see, e.g., [R]. From now on this result may be used without proof. This formula is often written in the form V−E+F=2V - E + F = 2.)

(2)

Find an analogue of Euler's formula for a plane graph with kk connected components.

Problem 2.4.3
?
(1)

Neither of the graphs K5K_{5} and K3,3K_{3,3} can be drawn in the plane without self-intersections.

Fig. 5. Nonplanar graphsFig. 5. Nonplanar graphs

(2)

In the plane, nn points are marked. It is allowed to join some two of them by a polygonal line not passing through the other points. Two players take turns joining some two not yet joined points by a polygonal line. It is required that these polygonal lines do not self-intersect and do not intersect anywhere except at the marked points. The player who cannot make a move loses. For which nn does the player who moves first win with correct play?

(3)

List all connected plane graphs (up to isomorphism) in which the degrees of all vertices are equal and the ``degrees'' of all faces are equal (i.e., the boundary of each face consists of the same number of edges).

Remark. If parts (1), (2), or (3) do not work out, solve the following parts. For the definition of isomorphism see Sec. 2.3.

(4)

There are exactly 5 convex regular polyhedra (all faces are regular polygons with the same number of sides, and the degrees of all vertices are equal) (up to isomorphism of their graphs). A construction of the corresponding polyhedra must be given; it is not assumed to be known.

(5)

For any connected plane graph without loops and multiple edges having more than two vertices, 2e≥3f2e \geq 3f and e≤3n−6e \leq 3n - 6.

(6)

Every plane graph has a vertex of degree at most 5.

(7)

If every vertex of a connected plane graph has degree dd and the boundary of every face consists of exactly k≥3k \geq 3 edges, then

1d+1k=12+1e. \frac{1}{d} + \frac{1}{k} = \frac{1}{2} + \frac{1}{e}.
Problem 2.4.4

A map in the plane is a partition of the plane into a finite number of polygons (possibly ``infinite'' ones). A coloring of a map is called proper if distinct polygons having a common boundary segment have distinct colors. Prove that every map can be properly colored with

?
(1)

6 colors;

(2)

5 colors;

(3)

[You won't be able to prove it anyway.]

Remark.
?

Remark. Using the construction of the dual graph, one can prove that proper colorability of every map in the plane with dd colors is equivalent to proper colorability of every plane graph with dd colors (Sec. 3.1).

Problem 2.4.5

Devise an algorithm

?
(1)

for recognizing whether a graph is planar (here you may use Kuratowski's theorem without proof);

(2)

for drawing a graph that is known to be planar in the plane without self-intersections.

Find the asymptotics of the complexity of your algorithm as a function of the number nn of edges of the graph, i.e., the asymptotics of the maximum, over graphs with nn edges, of the number of steps of the algorithm applied to the given graph. See the ``definition'' of finding asymptotics in Sec. 6.1.

Hopcroft--Tarjan theorem. There exists an algorithm for recognizing planarity of a graph that is linear in the number of edges.

Remark.
?

The operation of subdividing an edge of a graph: the edge (A,B)(A,B) is removed, and in its place a new vertex DD and two new edges (A,D)(A,D) and (D,B)(D,B) are added.

Two graphs are called homeomorphic if one can pass from one to the other by means of edge-subdivision operations and their inverses; or, equivalently, if there exists a graph obtained from each of the given graphs by edge-subdivision operations.

Clearly, homeomorphic graphs are either both planar or both non-planar.

Kuratowski's theorem. A graph is planar if and only if it contains no subgraph homeomorphic to the graph K5K_{5} or K3,3K_{3,3}. (For a proof of this theorem see, e.g., [S5].)

Problem 2.4.6
?
(1)

Fáry's theorem. A plane graph can be drawn in the plane without self-intersections so that all edges are line segments.

(2)

A non-convex polygon with opaque sides is given. Call a vertex AA visible from a point XX if the interior of the segment AXAX does not intersect the boundary of the polygon. Call the kernel of the polygon the set of its interior points from which all vertices of the polygon are visible. Prove that if a point of the kernel is joined by segments to several arbitrarily chosen vertices of the original polygon (at least two, and not necessarily all of them), then the polygon splits into polygons with nonempty kernels.

Problem 2.4.7

Draw on the torus without self-intersections the graph

?
(1)

K5K_{5};

(2)

K3,3K_{3,3};

(3)

K6K_{6};

(4)

K3,4K_{3,4};

(5)

K7K_{7};

(6)

K4,4K_{4,4}.

Remark.
?

The torus and the Möbius band are assumed to be transparent, i.e., a point (or a subset) lying on one side of the surface'' also lies on the other side''. This is analogous to the fact that in studying geometry we speak, for example, of a triangle in the plane, and not of a triangle on the upper (or lower) side of the plane.

Problem 2.4.8

Draw on the Möbius band without self-intersections the graph

?
(1)

K5K_{5};

(2)

K3,3K_{3,3};

(3)

K6K_{6};

(4)

K3,4K_{3,4}.

Problem 2.4.9

A map on the torus is a partition of the torus into finitely many (curvilinear and bent) polygons. A coloring of a map on the torus is called proper if distinct polygons having a common boundary curve have distinct colors. Can every map on the torus be properly colored

?
(1)

with 5 colors;

(2)

with 6 colors;

(3)

with 7 colors?

For details see [S1, 2].

Remark.
?

Remark. Any connected graph with gg edges can be drawn on a sphere with $g$ handles'' (i.e., inside a regular $2g$-gon whose diametrically opposite sides are glued''), in such a way that some edges are line segments, and the remaining edges are unions of two disjoint segments, each of which has one endpoint at a vertex of the graph and the other endpoint on a side of the 2g2g-gon.

Problem 2.5.1

How many multigraphs are there on nn given vertices that are

?
(1)

directed, without multiple edges but possibly with loops?

(2)

undirected, without loops but possibly with multiple edges?

Remark.
?

A multigraph (or a graph with loops and multiple edges) is a square table of nonnegative integers that is symmetric with respect to the main diagonal. The number at the intersection of the ii-th row and the jj-th column is interpreted as the number of edges (or the multiplicity of the edge) between the vertices numbered ii and jj when i≠ji \neq j, and as the number of loops at the vertex numbered ii when i=ji = j. An edge is called multiple if its multiplicity is greater than one.

A directed multigraph (or a directed graph with loops and multiple edges) is a square table of nonnegative integers. If some cell (no matter whether diagonal or not) contains a number greater than 1, then the directed multigraph is said to have multiple edges.

Problem 2.5.2

How many multigraphs are there on nn given vertices that have kk edges and are

?
(1)

undirected, without loops and multiple edges?

(2)

undirected, with multiple edges and loops allowed?

Problem 2.5.3

An Eulerian cycle (path) in a multigraph is a cycle (path) that passes along every edge of the multigraph exactly once.

?
(1)

A connected multigraph has an Eulerian cycle if and only if the degree of each of its vertices is even.

(2)

A connected multigraph has an Eulerian cycle if and only if its edge set decomposes into non-self-intersecting cycles.

(3)

Under what condition does a multigraph have an Eulerian path?

(4)

Under what condition does a directed multigraph have a directed Eulerian cycle?

(5)

For which nn does the graph KnK_{n} have an Eulerian cycle?

(6)

The same for the graph Km,nK_{m,n}.

Problem 2.5.4
?
(1)

If the number of vertices of odd degree in a connected graph equals 2k2k, then its edge set can be represented as the union of kk paths, none of which passes along any edge twice and no two of which have common edges.

(2)

The edges of a graph in which the degree of each vertex is even can be supplied with arrows so that at each vertex the in-degree equals the out-degree.

(3)

All edges of a connected graph are colored in two colors. Each vertex has equally many edges of both colors leaving it. Then from any vertex one can reach any other one, changing the color of the edge at every step.

(4)

A connected graph drawn in the plane without self-intersections has an Eulerian cycle if and only if its faces can be properly colored in 2 colors, i.e., so that the color changes when crossing each edge.

Remark.
?

The in-degree of a vertex of a directed multigraph is the number of edges entering it (counted with multiplicity). The out-degree is defined similarly. Here a loop of multiplicity kk ``contributes'' kk both to the in-degree and to the out-degree.

Problem 2.5.5

A mathematician has forgotten the three-digit code of his lock. The lock opens if the three digits of the code are entered consecutively (even if other digits were entered before that). The mathematician enters one digit per second; each entered digit is appended at the end. Prove that the mathematician can open the lock within

?
(1)

29 seconds if only the digits 1, 3 and 7 can be used in the code;

(2)

1002 seconds if ten digits can be used in the code.

(3)

Formulate and prove the ``0<1<2<…<8<90 < 1 < 2 < \ldots < 8 < 9'' rule for opening the lock within 1002 seconds.

Remark.
?

A de Bruijn sequence (d.B.s.) with parameters nn and kk is a sequence whose elements belong to a given set of kk elements (usually {0,1,…,k−1}\{ 0,1,\ldots ,k-1\}), such that all of its subsequences of length nn are distinct and among these subsequences all knk^{n} possible sequences occur. (Thus the length of a d.B.s. equals kn+n−1k^{n}+n-1.)

(An infinite periodic sequence with period knk^{n}, each subsequence of length kn+n−1k^{n}+n-1 of which is a d.B.s. with parameters nn and kk, is also called a d.B.s.)

Problem 2.5.6

Construct a de Bruijn sequence with parameters k=2k=2 (``binary'') and

?
(1)

n=3n=3, starting with 111;

(2)

n=4n=4, starting with 1011;

(3)

n=4n=4, ending with 1010.

Problem 2.5.7

Consider the sequence of zeros and ones constructed by the following rules. It starts with kk ones. After that, we write 1 only if, when writing 0, not all subsequences of length kk of the new sequence are distinct. If even when writing 1 not all subsequences of length kk of the new sequence are distinct, then we stop writing the sequence. Prove that in this way a de Bruijn sequence is obtained.

?
Problem 2.5.8

A connected directed multigraph with nn vertices is given. The in-degree dkd_{k} of each vertex kk equals its out-degree.

?
(1)

There exists a tree containing all vertices of this multigraph, all of whose edges are directed towards vertex 1.

(2)

Fix a tree TT from part (1). We traverse this graph (along the arrows), passing along each edge at most once. First we leave vertex 1 in an arbitrary direction. Next, suppose we have arrived at some vertex vv. We leave it along any edge not belonging to TT, if this is possible. If it is impossible, we leave it along the edge belonging to TT (such an edge is unique). Prove that the walk ends at vertex 1 and that the result is a directed Eulerian cycle.

(3)

The number of directed Eulerian cycles in this multigraph is a multiple of the number (d1−1)!⋅…⋅(dn−1)!(d_{1}-1)! \cdot \ldots \cdot (d_{n}-1)!.

Problem 2.6.1

The faces of a Hamiltonian plane graph can be properly colored in 4 colors.

?
Remark.
?

A Hamiltonian path (cycle) in a graph is a path (cycle) passing through each vertex exactly once.

Recall (Sec. 2.1) that the length of a path is the number of its edges (not vertices).

Problem 2.6.2
?
(1)

If a graph is connected and 2e≥n2−3n+62e \geq n^{2} - 3n + 6, then it has a Hamiltonian cycle.

(2)

The Dirac--Ore theorem. A graph in which the sum of the degrees of any two non-adjacent vertices is at least nn has a Hamiltonian cycle.

(3)

Dirac's lemma. If a0…asa_{0} \ldots a_{s} is a maximal one among the paths in the graph that pass through each of their vertices only once, s≥3s \geq 3 and deg⁡a0+deg⁡as>s\deg a_{0} + \deg a_{s} > s, then this graph has a non-self-intersecting cycle of length ss.

(4)

If a connected graph has a non-self-intersecting cycle of length s<ns < n, then this graph has a path of length ss passing through each of its vertices only once.

(5)

A graph in which the sum of the degrees of any two non-adjacent vertices is at least n−1n-1 has a Hamiltonian path.

Problem 2.6.3

Suppose that for some graph and some integer k≥2k \geq 2, among any k+1k+1 vertices of the graph there is an edge, and after deleting any set of k−1k-1 vertices the graph remains connected. Then this graph has a Hamiltonian cycle.

?
Problem 2.6.4

Suppose that among any k+1k+1 vertices of a graph there is an edge, and after deleting any set of k−1k-1 vertices the graph remains connected.

?
(1)

This graph has at least one non-self-intersecting cycle.

(2)

Denote by v1,…,vsv_{1}, \ldots , v_{s} a maximal non-self-intersecting cycle in this graph. Denote by WW any connected component of the graph obtained from the original graph by deleting the vertices of this non-self-intersecting cycle. Denote by XX the set of vertices of the non-self-intersecting cycle that are adjacent to WW.

Then ∣X∣≥k\left|X\right| \geq k.

(3)

The vertices vi,vi+1v_{i}, v_{i+1} do not both lie in XX.

(4)

If vi,vj∈Xv_{i}, v_{j} \in X, then the graph has no edge vi+1vj+1v_{i+1} v_{j+1}.

Problem 2.6.5
?
(1)

Is there a Hamiltonian path in the graph in Fig. 8?

Fig. 8. Does this graph have a Hamiltonian path?Fig. 8. Does this graph have a Hamiltonian path?

(2)

Is there a Hamiltonian cycle in the graph in Fig. 9?

Fig. 9. The graph of the Grinberg polyhedron. Does it have a Hamiltonian path?Fig. 9. The graph of the Grinberg polyhedron. Does it have a Hamiltonian path?

(3)

For which nn is there a Hamiltonian cycle in the graph whose vertices are the 3-element subsets of an nn-element set, two subsets being joined by an edge if they intersect in exactly one element?

Problem 2.6.6

The maximum number of pairwise edge-disjoint Hamiltonian cycles in the graph KnK_{n} equals [n−12]\left[\dfrac {n-1}{2}\right].

?
Problem 2.6.7
?
(1)

Every tournament has a directed Hamiltonian path.

(2)

For every nn there exists a tournament with nn vertices that has at least n!/2nn! / 2^{n} directed Hamiltonian paths.

Remark.
?

A tournament is a directed graph in which any two vertices are joined by an edge. (That is, for any two vertices v,wv, w of a tournament, its edges include (v,w)(v,w) or (w,v)(w,v), but not both.)

Problem 2.6.8

The line graph of a graph GG is the graph whose vertices are the edges of GG; two vertices of the line graph are joined by an edge if the corresponding edges of GG have a common vertex. Find, in terms of the graph GG, a necessary and sufficient condition for its line graph to have a Hamiltonian cycle.

See also [Ve].

?
Problem 2.7.1

The parts of this problem, except (2), are various versions and special cases of Turán's theorem.

A triangle in a graph is a cycle of length 3.

?
(1)

If a graph contains no triangles, then e≤n2/4e \leq n^{2}/4.

(2)

If e=[n2/4]+1e = [n^{2}/4] + 1, then the graph has at least [n/2][n/2] triangles.

(3)

If n=kmn = km and the graph contains no (k+1)(k+1)-clique, then 2e≤k(k−1)×m22e \leq k(k-1) \times m^{2}. (Passing to the complementary graph, we obtain that if n=kmn = km and the graph contains no (k+1)(k+1)-anticlique, then 2e≥km(m−1)2e \geq km(m-1).)

(4)

If the graph contains no (k+1)(k+1)-anticlique, then 2e≥km(m−1)+2mr2e \geq km(m-1) + 2mr, where m:=[n/k]m := [n/k] and r:=k{n/k}r := k\{ n/k\}.

Problem 2.7.2
?
(1)

If a graph contains no non-self-intersecting cycle of length 4, then e<n3/2e < n^{3/2}.

(2)

If a graph contains no subgraph K3,2K_{3,2}, then e<2n3/2e < 2n^{3/2}.

(3)

If a graph contains no subgraph K3,3K_{3,3}, then e<2n5/3e < 2n^{5/3}.

(4)

For any integers s,ts, t, 2≤s≤t2 \leq s \leq t, if a graph contains no subgraph Ks,tK_{s,t}, then e<tn2−1/se < t n^{2-1/s}.

Problem 2.7.3

For any nn points in the plane there are at most nn diameters, i.e., (unordered) pairs of points whose distance equals the maximum of all possible distances between pairs of these nn points.

?
Problem 2.7.4

For any nn points A1,…,AnA_{1}, \ldots , A_{n} in Rd\mathbb {R}^{d}, denote by D(A1,…,An)D(A_{1}, \ldots , A_{n}) the number of (unordered) pairs of points whose distance equals 1. Let

En(d)=max⁡{D(A1,…,An):A1,…,An∈Rd}. E_{n}(d) = \max \{ D(A_{1},\ldots ,A_{n}) : A_{1},\ldots ,A_{n} \in \mathbb {R}^{d}\} .

Then:

?
(1)

En(2)>n[log⁡2n]/4E_{n}(2) > n[\log_{2} n]/4;

(2)

En(2)≤2n3/2E_{n}(2) \leq 2n^{3/2};

(3)

En(3)≤2n5/3E_{n}(3) \leq 2n^{5/3};

(4)

(n−1)24≤En(4)≤2(n+4)25\dfrac {(n-1)^{2}}{4} \leq E_{n}(4) \leq \dfrac {2(n+4)^{2}}{5}.

Problem 2.7.5
?
(1)

Let VV be an 11q11^{q}-element subset of the space Rq\mathbb {R}^{q} (for the definition of the space Rq\mathbb {R}^{q}, see Ch. 7) every 10q10^{q}-element subset of which contains two points x,yx,y at distance 1: ∣x−y∣=1\left|x-y\right|=1. Prove that for sufficiently large qq the number of unit distances between points of the set VV is greater than 12q/212^{q}/2:

12∣{(x,y)∈V×V:∣x−y∣=1}∣>12q2. \frac{1}{2} \left|\{ (x,y) \in V \times V : \left|x-y\right|=1\} \right| > \frac{12^{q}}{2}.
(2)

Prove that under the conditions of the previous part the number 12q/212^{q}/2 can be replaced by 12,1q12{,}1^{q}.

Problem 2.7.6

One can consider a generalization of Turán's problem (see Problem 2.7.1) in which, instead of cliques of a given size, other subgraphs are forbidden. Denote by ex⁡H(n)\operatorname {ex}_{H}(n) the maximum number of edges in a graph with nn vertices containing no subgraphs isomorphic to HH. For example, ex⁡Kk+1(n)\operatorname {ex}_{K_{k+1}}(n) is the maximum number of edges in a graph with nn vertices containing no (k+1)(k+1)-clique.

Prove that if H1H_{1} is a subgraph of the graph H2H_{2}, then ex⁡H1(n)≤ex⁡H2(n)\operatorname {ex}_{H_{1}}(n) \leq \operatorname {ex}_{H_{2}}(n).

See also Problems 6.1.2 and 6.1.3.

?
Problem 2.8.1

From every connected multigraph one can remove a vertex (together with all edges incident to it) so that it remains connected.

?
Remark.
?

A graph or multigraph is called 2-connected if it differs from K2K_{2} and remains connected after the removal of any vertex.

Problem 2.8.2
?
(1)

A special case of the vertex version of Menger's theorem. Any two distinct vertices of a 2-connected multigraph that are not joined by an edge lie on some non-self-intersecting cycle.

(2)

Is it true that for any path PP in a multigraph having at least three vertices there exists another path in the same multigraph with the same endpoints that does not intersect PP anywhere except at the endpoints?

Problem 2.8.3

If a multigraph has at least one edge and, after the removal of any edge, there is still a path between the vertices aa and bb, then the multigraph contains two paths between the vertices aa and bb having no edges in common.

?
Problem 2.8.4
?
(1)

For any edge uu of a 2-connected graph GG different from K3K_{3}, at least one of the graphs G−uG-u and G/uG/u is 2-connected.

(2)

For any two vertices of a convex polyhedron there exist three paths along its edges from one vertex to the other that are disjoint (everywhere except at these vertices). (Such graphs are called 3-connected.)

(3)

For a 3-connected graph GG with an edge xyxy (whose vertices are x,yx,y), the graph G/xyG/xy is 3-connected if and only if the graph G−x−yG-x-y is 2-connected.

Problem 2.8.5
?
(1)

Whitney's theorem (vertex version). A multigraph remains connected after the removal of any k−1k-1 vertices if and only if any two of its vertices can be joined by kk paths intersecting only at these two vertices.

(2)

Menger's theorem (vertex version). If vertices aa and bb of a multigraph GG that are not joined by an edge remain in the same connected component after the removal of any k−1k-1 other vertices, then aa and bb can be joined by kk paths intersecting only at these two vertices.

Problem 2.8.6

Call vertices AA and BB of a graph equivalent if there exists a sequence of vertices A=A0,A1,…,An=BA = A_{0}, A_{1}, \ldots , A_{n} = B such that any two consecutive vertices AiA_{i} and Ai+1A_{i+1} can be joined by kk paths having no common intermediate vertices. Then any two equivalent vertices can be joined by kk paths having no common edges.

?