Overview

Graph theory problems often rely on invariants such as degree sums and connectivity.

Key Ideas

  • Handshaking lemma: the sum of degrees equals 2E2E.
  • Trees with nn vertices have n−1n-1 edges.
  • Parity arguments are common in graph proofs.

Worked Example

Show that any graph has an even number of vertices of odd degree.

The sum of degrees is 2E2E, an even number. A sum of degrees is even only if the count of odd terms is even.

More Examples (Warm-up):

Example 1: Interesting Degrees Problem

Let GG be a simple graph on nn vertices with degrees d1,d2,…,dnd_1,d_2,\dots,d_n. Suppose that

∑i=1ndi2>n⌊n24⌋.\sum_{i=1}^{n} d_i^2 > n\left\lfloor \frac{n^2}{4}\right\rfloor.

Your job is to prove that GG contains a triangle.

First, we assume GG is triangle-free. Then, by Mantel's Theorem, we have that

m≤⌊n24⌋.m\le \left\lfloor \frac{n^2}{4}\right\rfloor.

Also,

∑vd(v)2=∑uv∈E(d(u)+d(v)).\sum_v d(v)^2=\sum_{uv\in E}(d(u)+d(v)).

For every edge uvuv, we have d(u)+d(v)≤nd(u)+d(v)\le n, otherwise uu and vv would have a common neighbor and form a triangle. Thus,

∑vd(v)2≤nm≤n⌊n24⌋,\sum_v d(v)^2\le nm\le n\left\lfloor \frac{n^2}{4}\right\rfloor,

which contradicts the claim, so therefore GG contains a triangle.

Example 2: Hall's Matching Theorem (classic!)

Let G=(X,Y)G=(X,Y) be a bipartite graph. Suppose that every vertex of XX has degree at least rr, and every vertex of YY has degree at most rr.

Your job is to prove that GG contains a matching saturating (at least one edge for every vertex) XX.

First of all, for any subset S⊆XS\subseteq X, we can count edges between SS and N(S)N(S):

r∣S∣≤e(S,N(S))≤r∣N(S)∣.r|S|\le e(S,N(S))\le r|N(S)|.

Thus, ∣N(S)∣≥∣S∣|N(S)| \ge |S|

Hall's condition holds, so Hall's Theorem guarantees a matching saturating all vertices of XX, and we are done!

Example 3: Tricky Paths Problem

Let GG be a graph on n≥4n\ge4 vertices. Suppose that every pair of nonadjacent vertices u,vu,v satisfies the following:

d(u)+d(v)≥n+1.d(u)+d(v)\ge n+1.

Your job is to prove that any two vertices are connected by at least three internally vertex-disjoint paths (make sure you read this carefully!).

First, assume that there is a vertex cut SS with ∣S∣≤2|S|\le2. Let AA and BB be distinct components of G−SG-S, and here, we will choose a∈Aa\in A, b∈Bb\in B.

Then

d(a)≤∣A∣−1+∣S∣,d(b)≤∣B∣−1+∣S∣.d(a)\le |A|-1+|S|, \qquad d(b)\le |B|-1+|S|.

From here, Adding and using ∣A∣+∣B∣≤n−∣S∣|A|+|B|\le n-|S| gives us:

d(a)+d(b)≤n+∣S∣−2≤n,d(a)+d(b)\le n+|S|-2\le n,

which contradicts the claim. Thus κ(G)≥3\kappa(G)\ge3, and Menger's Theorem (important!) means that there are at least three internally vertex-disjoint paths between any two vertices, and here, we are done!

Example 4: Wonderful Bipartite Matching Problem!

Let G=(X,Y)G=(X,Y) be a bipartite graph with

∣X∣=∣Y∣=n.|X|=|Y|=n.

Suppose every vertex has degree at least n/2n/2.

Your job is to prove that GG contains a perfect matching (every vertex has exactly 1 edge).

First, by Hall's Theorem, realize that it suffices to show ∣N(S)∣≥∣S∣|N(S)|\ge |S| for every S⊆XS\subseteq X.

Assume ∣N(S)∣<∣S∣|N(S)|<|S|. Now, notice that since every vertex of SS has degree at least n/2n/2,

e(S,N(S))≥n2∣S∣.e(S,N(S))\ge \frac n2|S|.

Because e(S,N(S))≤∣S∣∣N(S)∣e(S,N(S))\le |S||N(S)|, we get that:

∣N(S)∣≥n2.|N(S)|\ge \frac n2.

Thus ∣S∣>n2|S|>\frac n2. Let T=Y∖N(S)T=Y\setminus N(S). Then

∣T∣=n−∣N(S)∣>n−∣S∣.|T|=n-|N(S)|>n-|S|.

Every vertex of TT has degree at least n/2n/2 and no neighbors in SS, so we have:

∣X∖S∣≥n2.|X\setminus S|\ge \frac n2.

Hence ∣S∣≤n/2|S|\le n/2, which is a contradiction! Thus, Hall's condition holds, and GG has a perfect matching, and we are done!

Example 5: Turán's Theorem

Let GG be a graph on nn vertices with more than

(1−1r)n22\left(1-\frac1r\right)\frac{n^2}{2}

edges.

Your goal is to prove that GG contains a clique (you should know this term!) of size r+1r+1.

First, we assume that GG contains no Kr+1K_{r+1}. So, by Turán's Theorem, we have:

e(G)≤e(Tr(n))≤(1−1r)n22.e(G)\le e(T_r(n)) \le \left(1-\frac1r\right)\frac{n^2}{2}.

However, this contradicts the claim! So, GG must contain a clique of size r+1r+1, and we are done here!

Example 6: A Truly Dazzling Triangles Problem :)

Let GG be a graph on nn vertices with

m>n24.m>\frac{n^2}{4}.

Your goal is to prove that there exists an edge contained in at least

2mn−n2\frac{2m}{n}-\frac n2

triangles.

First, for an edge uvuv, let

c(uv)=∣N(u)∩N(v)∣.c(uv)=|N(u)\cap N(v)|.

Since

c(uv)≥d(u)+d(v)−n,c(uv)\ge d(u)+d(v)-n,

we can sum over all edges to give us

∑uv∈Ec(uv)≥∑vd(v)2−mn.\sum_{uv\in E} c(uv) \ge \sum_v d(v)^2-mn.

Then, from here, we can use Cauchy–Schwarz, to get:

∑vd(v)2≥(2m)2n.\sum_v d(v)^2 \ge \frac{(2m)^2}{n}.

Hence

∑uv∈Ec(uv)≥4m2n−mn.\sum_{uv\in E} c(uv) \ge \frac{4m^2}{n}-mn.

Averaging over all mm edges,there must be some edge that satisfies

c(uv)≥4mn−n.c(uv)\ge \frac{4m}{n}-n.

Since

4mn−n≥2mn−n2,\frac{4m}{n}-n \ge \frac{2m}{n}-\frac n2,

that edge lies in at least

2mn−n2\frac{2m}{n}-\frac n2

triangles, and we are done!

Practice Problems

StatusSourceProblem NameDifficultyTags
USAMOVery Hard
Show TagsGraph Theory, Invariants
USAMOInsane
Show TagsGraph Theory
Berkeley Math Circle Monthly Contest 2Hard
Show TagsColoring schemes, extremal arguments, Graph Theory
Berkeley Math Circle Monthly Contest 2Hard
Show TagsGraph Theory, Pigeonhole principle
Berkeley Math CircleHard
Show TagsCounting two ways, Graph Theory, Invariants / monovariants
Berkeley Math Circle: Monthly Contest 7Hard
Show TagsColoring schemes, extremal arguments, Graph Theory
MathNetHard
Show TagsCounting two ways, Graph Theory, Invariants / monovariants
Berkeley Math Circle Monthly Contest 2Hard
Show TagsGraph Theory, Linear and quadratic inequalities
Berkeley Math Circle Monthly Contest 4Hard
Show TagsColoring schemes, extremal arguments, Graph Theory
BAMOHard
Show TagsColoring schemes, extremal arguments, Graph Theory, Invariants / monovariants
BAMOHard
Show TagsGraph Theory, Jensen / smoothing, Linear and quadratic inequalities
Berkeley Math CircleHard
Show TagsColoring schemes, extremal arguments, Graph Theory, Invariants / monovariants
Harvard-MIT Mathematics TournamentHard
Show TagsColoring schemes, extremal arguments, Graph Theory, Linear and quadratic inequalities
Harvard-MIT Mathematics TournamentHard
Show TagsColoring schemes, extremal arguments, Graph Theory
Harvard-MIT Mathematics TournamentHard
Show TagsColoring schemes, extremal arguments, Enumeration with symmetry, Graph Theory
Harvard-MIT Mathematics TournamentHard
Show TagsGraph Theory
Bay Area Mathematical OlympiadHard
Show TagsColoring schemes, extremal arguments, Graph Theory
Harvard-MIT Mathematics TournamentHard
Show TagsColoring schemes, extremal arguments, Counting two ways, Graph Theory
Harvard-MIT Mathematics TournamentHard
Show TagsEuler characteristic: V-E+F, Graph Theory, Invariants / monovariants
USAMOHard
Show TagsColoring schemes, extremal arguments, Graph Theory, Induction / smoothing
10th Annual Harvard-MIT Mathematics TournamentHard
Show TagsCounting two ways, Enumeration with symmetry, Graph Theory
Harvard-MIT Math TournamentHard
Show TagsColoring schemes, extremal arguments, Graph Theory, Induction / smoothing
Harvard-MIT Math TournamentHard
Show TagsColoring schemes, extremal arguments, Graph Theory, Prime numbers
Harvard-MIT Math TournamentHard
Show TagsColoring schemes, extremal arguments, Graph Theory
Harvard-MIT Math TournamentHard
Show TagsColoring schemes, extremal arguments, Graph Theory, Induction / smoothing, Pigeonhole principle

Module Progress:

Join the Discord Community!

Stuck on a problem, or don't understand a module? Join the Discord and get help with your doubts while making more math friends.