1. Vertex colorings Definition 1.1. A (proper) vertex coloring of a graph is an assignment of colors to its vertices such that the colors of any two adjacent vertices are distinct. Without loss of generality, we can assume that our graphs are simple. Definition 1.2. A graph is called k-colorable if its vertices can be properly colored using ≤ k colors. A 2-colorable graph is called bipartite. Example 1.3. The cube is the only Platonic solid whose skeleton is bipartite. Example 1.4. Petersen’s graph is 3-colorable but not 2-colorable. Theorem 1.5 (K. Appel and W. Haken, 1976). Every planar graph is 4-colorable. This result, first conjectured in 1852, was the first important theorem whose proof relied, in an essential way, on the use of computers. The proof of the four color theorem requires the use of computers in an essential way. We will prove a much weaker result. Theorem 1.6 (The five-color theorem). Any planar graph G satisfies χ(G) ≤ 5. We will give two different proofs. Proof 1. We first show that any planar graph has a vertex of degree ≤ 5. If |V | < 3, this is trivally true, so assume |V | ≥ 3. Suppose there is no vertex of degree < 5. Then every vertex has degree ≥ 6. Therefore, 2|E| ≥ 6|V | so |E| ≥ 3|V |. On the other hand, since |V | ≥ 3, we have |E| ≤ 3|V | − 6, a contradiction. We now proceed by induction on |V |. If |V | < 5, then this is trivial. Suppose |V | > 5. Let v be a vertex of degree ≤ 5. If deg(v) ≤ 4, then by the induction hypothesis, G − v has a coloring with 5 colors and we can color v using a color not used by its neigbors. Therefore, assume deg(v) = 5. Let t, u, x, z, y denote the vertices in clockwise cyclic order incident to v. By the induction hypothesis, there is a 5-coloring c of G − v. If any of the 5 colors is not used by a neighbor of v, then we can color v using that color. Therefore, assume that all 5 neighbors have different colors. Consider x, y and let Vx,y := {u ∈ V (G − v) : c(u) ∈ {c(x), c(y)}} denote the vertices that have the same color as either x or y. Now, we have two cases: ′ (1) There is no path in G − v only using vertices in Vx,y from x to y. Let Vx,y denote the vertices that can be reached by a path from x only using vertics in Vx,y . Clearly, ′ ′ ′ x ∈ Vx,y and y ∈/ Vx,y . Interchange the colors of all vertices in Vx,y to get another coloring of G − v in which x and y have the same color. Since only 4 colors are now used by the neighbors of v, we can color v using the fifth color. 1 2 MATH 180 - GRAPH THEORY LECTURE 16 – COLORING
(2) There is a path P in G − v only using vertices in Vx,y from x to y in G. Consider z, t.
Since P ∪ {{v, x}, {v, y}} is a cycle with z and t on opposite sides, any path from z to t in G − v must intersect P , so there is no path in G − v only using vertices in Vz,t from z to t. Now we are in Case (1). □ MATH 180 - GRAPH THEORY LECTURE 17 – THE CHROMATIC POLYNOMIAL
1. Vertex colorings Definition 1.1. A (proper) vertex coloring of a graph is an assignment of colors to its vertices such that the colors of any two adjacent vertices are distinct. Without loss of generality, we can assume that our graphs are simple. Definition 1.2. A graph is called k-colorable if its vertices can be properly colored using ≤ k colors. A 2-colorable graph is called bipartite. Example 1.3. The cube is the only Platonic solid whose skeleton is bipartite. Example 1.4. Petersen’s graph is 3-colorable but not 2-colorable. Definition 1.5. The chromatic number χ(G) of a graph G is the smallest k for which G is k-colorable. Example 1.6. (1) If G is a bipartite graph with at least one edge, χ(G) = 2. (2) For a complete graph, we have χ(Kn ) = n. (3) If G is an n-cycle graph, n even, then χ(G) = 2. (4) If G is an n-cycle graph, n ≥ 3 odd, then χ(G) = 3. No efficient algorithm for computing χ(G) is known. Definition 1.7. Let G = (V, E) be a (simple) graph. A clique in G is a subset of V in which any two vertices are adjacent. The clique number ω(G) of G is the largest size of a clique in G. Proposition 1.8. χ(G) ≥ ω(G). Example 1.9. In general, χ(G) ̸= ω(G). For example, (1) for an odd cycle with ≥ 5 vertices, ω(G) = 2 but χ(G) = 3; (2) for the graph below, ω(G) = 3 but χ(G) = 4.
Definition 1.10. An independent set in a graph G = (V, E) is a subset of V in which no two
vertices are adjacent. The independence number α(G) is the largest size of an independent set in G. Proposition 1.11. |V | χ(G) ≥ . α(G) Proof. The set of vertices colored in a particular color is an independent set. Therefore |V | ≤ χ(G)α(G). □ 1 2 MATH 180 - GRAPH THEORY LECTURE 17 – THE CHROMATIC POLYNOMIAL
Example 1.12. For an odd cycle, the last proposition implies that χ(G) ≥ 3. Proposition 1.13. χ(G) ≤ max deg(v) + 1. v∈V
Proof. Color vertices one by one, any way we please. There will always be an allowable choice of color. □ Example 1.14. For an odd cycle, the last proposition gives χ(G) ≤ 3.
2. Chromatic polynomial Definition 2.1. The chromatic polynomial of a graph G is a function pG : Z≥0 → Z defined by pG (k) = number of proper vertex colorings of G in k colors. This terminology will be justified later on. The chromatic polynomial determines the chromatic number: χ(G) = (smallest k such that pG (k) ̸= 0).
Example 2.2. Let G = Kn . Then pG (k) = k(k − 1) · · · (k − n + 1).
Proposition 2.3. If G is a tree on n vertices, then pG (k) = k(k − 1)n−1 . Proof. Induction on n. For n = 1, we have pG (k) = k. For n ≥ 2, color all vertices except for a leaf v; we then have k − 1 options left for v. □ Corollary 2.4. If G is a forest on n vertices with c connected components, then pG (k) = k c (k − 1)n−c . Example 2.5. Let G be a 4-cycle. Pick two opposite vertices u and v. In each coloring, u and v are either of the same color or not. Accordingly, pG (k) = k(k − 1)2 + k(k − 1)(k − 2)2 = k(k − 1)(k 2 − 3k + 3).
3. The deletion-contraction formula
Definition 3.1. Let G = (V, E) be a simple graph, and e ∈ E an edge. We denote G − e = graph obtained from G by the deletion of e; G/e = graph obtained from G by the contraction of e. After contracting e, we get rid of multiple edges, if any. Lemma 3.2 (The deletion-contraction formula). pG (k) = pG−e (k) − pG/e (k). Proof. In each proper coloring of G − e in k colors, the (former) endpoints of e are either colored the same color or not. This yields pG−e (k) = pG/e (k) + pG (k). □ MATH 180 - GRAPH THEORY LECTURE 17 – THE CHROMATIC POLYNOMIAL 3
Example 3.3. Let G be a 4-cycle. Let e be an edge in it. Then
G − e is a tree ⇒ pG−e (k) = k(k − 1)3 G/e ∼ = K3 ⇒ pG/e (k) = k(k − 1)(k − 2) pG (k) = pG−e (k) − pG/e (k) = k(k − 1)(k 2 − 3k + 3) Example 3.4. Let G be a 5-cycle. Let e be an edge in it. Then G − e is a tree ⇒ pG−e (k) = k(k − 1)4 G/e is a 4-cycle ⇒ pG/e (k) = k(k − 1)(k 2 − 3k + 3) pG (k) = pG−e (k) − pG/e (k) = k(k − 1)(k − 2)(k 2 − 2k + 2). Note that pG (2) = 0, in agreement with χ(G) = 3. Lemma 3.5. If G is an edgeless graph on n vertices, then pG (k) = k n . The deletion-contraction formula pG (k) = pG−e (k) − pG/e (k) together with the above observation, provides a recursive algorithm for computing pG (k). This algorithm has exponential complexity. There are many versions of this algorithm, de- pending on the order in which we delete/contract the edges. Note that if we did not know the meaning the of pG (k), then it would not be clear at all why the output of the algorithm does not depend on the choices of edges to be deleted/contracted. Theorem 3.6. pG (k) is a polynomial in k. Proof. Induction on the number of edges in G. Base: if G has no edges, then pG (k) = k n . Induction step: use the deletion-contraction formula. □ Proposition 3.7. The chromatic polynomial pG (k) is monic, of degree n = |V |. Proof. We need to show that pG (k) ∼ k n as k → ∞. Note that pGk(k) n is the probability that a random assignment of colors to vertices yields a proper coloring. For k large, almost all colorings are proper. Alternatively, we can use induction on the number of edges. Suppose |E| = 0. Then, G consists of n isolated vertices, so pG (k) = k n , which is monic of degree n. Suppose the statement is true for all graphs with ≤ m − 1 edges, and suppose G is a graph with m edges. Let e be an edge. Both G − e and G/e have ≤ m − 1 edges, so pG−e (k) = k n + (lower degree terms) and pG/e (k) = k n−1 + (lower degree terms). Using deletion-contraction, pG (k) = k n + (lower degree terms). □ MATH 180 - GRAPH THEORY LECTURE 18 – DELETION-CONTRACTION
1. The deletion-contraction formula
Definition 1.1. Let G = (V, E) be a simple graph, and e ∈ E an edge. We denote G − e = graph obtained from G by the deletion of e; G/e = graph obtained from G by the contraction of e. After contracting e, we get rid of multiple edges, if any. Lemma 1.2 (The deletion-contraction formula). pG (k) = pG−e (k) − pG/e (k). Proof. In each proper coloring of G − e in k colors, the (former) endpoints of e are either colored the same color or not. This yields pG−e (k) = pG/e (k) + pG (k). □ Example 1.3. Let G be a 4-cycle. Let e be an edge in it. Then G − e is a tree ⇒ pG−e (k) = k(k − 1)3 G/e ∼ = K3 ⇒ pG/e (k) = k(k − 1)(k − 2) pG (k) = pG−e (k) − pG/e (k) = k(k − 1)(k 2 − 3k + 3) Example 1.4. Let G be a 5-cycle. Let e be an edge in it. Then G − e is a tree ⇒ pG−e (k) = k(k − 1)4 G/e is a 4-cycle ⇒ pG/e (k) = k(k − 1)(k 2 − 3k + 3) pG (k) = pG−e (k) − pG/e (k) = k(k − 1)(k − 2)(k 2 − 2k + 2). Note that pG (2) = 0, in agreement with χ(G) = 3. Lemma 1.5. If G is an edgeless graph on n vertices, then pG (k) = k n . The deletion-contraction formula pG (k) = pG−e (k) − pG/e (k) together with the above observation, provides a recursive algorithm for computing pG (k). This algorithm has exponential complexity. There are many versions of this algorithm, de- pending on the order in which we delete/contract the edges. Note that if we did not know the meaning the of pG (k), then it would not be clear at all why the output of the algorithm does not depend on the choices of edges to be deleted/contracted. Theorem 1.6. pG (k) is a polynomial in k. 1 2 MATH 180 - GRAPH THEORY LECTURE 18 – DELETION-CONTRACTION
Proof. Induction on the number of edges in G. Base: if G has no edges, then pG (k) = k n . Induction step: use the deletion-contraction formula. □ Proposition 1.7. The chromatic polynomial pG (k) is monic, of degree n = |V |.
Proof. We need to show that pG (k) ∼ k n as k → ∞. Note that pGk(k) n is the probability that a random assignment of colors to vertices yields a proper coloring. For k large, almost all colorings are proper. Alternatively, we can use induction on the number of edges. Suppose |E| = 0. Then, G consists of n isolated vertices, so pG (k) = k n , which is monic of degree n. Suppose the statement is true for all graphs with ≤ m − 1 edges, and suppose G is a graph with m edges. Let e be an edge. Both G − e and G/e have ≤ m − 1 edges, so pG−e (k) = k n + (lower degree terms) and pG/e (k) = k n−1 + (lower degree terms). Using deletion-contraction, pG (k) = k n + (lower degree terms). □
2. Acyclic orientations Definition 2.1. An acyclic orientation of a graph G is a directed graph obtained by orienting the edges of G without creating an oriented cycle. Let ao(G) denote the number of such orientations. Example 2.2. All edge orientations of a tree G = (V, E) are acyclic, so ao(G) = 2|E| . Example 2.3. Let G be an n-cycle. Then all edge orientations except for two are acyclic. Therefore ao(G) = 2n − 2. Example 2.4. The acyclic orientations of the complete graph Kn are in bijection with the permutations in Sn : (1) If w1 · · · wn ∈ Sn , orient the edge between wi and wj from wi to wj if i < j to get an acyclic orientation. (2) Given an acyclic orientation, we can recover the permutation as follows. There is a unique vertex v such that all edges are oriented into v: if not, then every vertex has an outgoing edge. Start at a vertex and keep walking along the outgoing edges to get a cycle. Let wn := v, remove wn and proceed recursively. Therefore, ao(G) = n!. Theorem 2.5 (R. P. Stanley, 1973). ao(G) = (−1)|V | · pG (−1). Although it does not make sense to set k = −1 in the original definition of pG (k), we can plug k = −1 into the polynomial pG (k) and miraculously we still get something combinato- rially meaningful. Such a result is called a combinatorial reciprocity theorem. Lemma 2.6 (Deletion-contraction formula for acyclic orientations). For an edge e ∈ E, we have ao(G) = ao(G − e) + ao(G/e). Proof. Let u and v be the endpoints of e. Any acyclic orientation of G − e extends to either one or two acyclic orientations of G, depending on whether u and v are connected by an oriented path (going in either direction) or not. There cannot be oriented paths going in MATH 180 - GRAPH THEORY LECTURE 18 – DELETION-CONTRACTION 3
both directions because this would have yielded an oriented cycle in G−e. Therefore acyclic orientations of G − e n o ao(G) = ao(G − e) + # with no paths u ⇝ v or v ⇝ u = ao(G − e) + ao(G/e). □ Example 2.7. Let G be a 4-cycle. Then pG (k) = k(k − 1)(k 2 − 3k + 3), so (−1)4 pG (−1) = (−1)(−2) · 7 = 14, matching ao(G) = 24 − 2 = 14. Exercise 2.8. Verify Stanley’s theorem when G is a tree and a complete graph Kn . Proof of Stanley’s theorem. The deletion-contraction formula pG (k) = pG−e (k) − pG/e (k) implies (−1)n pG (k) = (−1)n pG−e (k) + (−1)n−1 pG/e (k). Thus the polynomials (−1)n pG (k), and consequently the numbers (−1)n pG (−1), satisfy the same recurrence as the numbers ao(G). It remains to check the base case of the induction: an edgeless graph G on n vertices. In this case, pG (k) = k n and ao(G) = 1, so (−1)n pG (−1) = 1 = ao(G), as desired. □
