Understand
Objectives
- Derive both graph color polynomials and normalizations.
- Calculate the C₄ and tree gaps without conflating their matroids.
- Mark computed equalities separately from general theorem inputs.
Start with something you can see
Draw C₄ and remove one edge to form a path. Before computing, predict which changes in rank, spanning-tree bases and color counts. A count for one graph is evidence about that graph; the general inequality needs a theorem.
Give the idea a precise name
The graphic matroid has the graph edges as its ground set. For a graph with c components, P_G(q)=qᶜχ_M(q). For a nonempty loopless matroid, χ̄_M(q)=χ_M(q)/(q−1). A rank r+1 matroid has reduced degree r and Chow top degree r.
Choose the first vertex in q ways and the second in q−1 ways. If the third equals the first, the fourth has q−1 choices; otherwise the third and fourth each have q−2 choices. Thus q(q−1)[(q−1)+(q−2)²]=q⁴−4q³+6q²−3q. Dividing by q gives χ=q³−4q²+6q−3=(q−1)(q²−3q+3). The rank is 3 and four edge deletions give four spanning-tree bases.
For a finite loopless matroid of rank r+1, the mixed-degree identity μᵏ=deg(αr−kβᵏ) and the Hodge–Riemann relations imply weak log-concavity of its absolute reduced characteristic coefficients. This general input is due to Adiprasito–Huh–Katz; the four-cycle arithmetic above is an illustration, not its proof.
Work one small world
At q=3, P=18. The absolute reduced coefficients are (1,3,3); the gap is 3²−1×3=6. Delete an edge: the tree has P=q(q−1)³, χ̄=(q−1)² and coefficients (1,2,1), gap 3. This tree’s free matroid U₃,₃ is the earlier computed Chow toy case, not C₄’s ring.
Counterexample to the proof shortcut: checking C₄ and its tree proves only their own gaps. It cannot establish log-concavity for arbitrary, possibly nonrepresentable matroids. The deep coefficient identity and Hodge–Riemann statement are invoked, not proved by two computations.
THE BRIDGEThe final capstone replaces a counting inequality by a rank-and-noise comparison.
WHERE THIS IDEA STOPSWeak log-concavity permits equality; graph calculations do not prove the theorem for general matroids.
Experiment
Use the graph lab’s cycle and path presets, then make the proof-map table before looking at the numerical anchors. Predict rank 3 for both, bases 4 versus 1, and the changed reduced coefficients.
What changes, and what remains invariant, when you remove exactly one edge?
Open the interactive experiment
Check & explain
How many spanning-tree bases does a simple six-cycle C₆ have?
For the seven-vertex path, the reduced coefficients are those of (q−1)⁵. What is the first interior log-concavity gap?
Teach back
In the teach-back field, submit a six-arrow proof map from C₄ to P, χ, χ̄, mixed degrees and the Hodge inequality. Label each calculation you performed and each theorem you invoked. Include the deleted-edge comparison.
- If an edge/vertex mix-up appears, identify colorings as vertex assignments and matroid elements as edges.
- If a normalization is missing, show qᶜ then q−1 and state the hypotheses.
- If a generality leap appears, mark Proposition 9.5 and Theorem 1.4 as inputs, not consequences of two graphs.
Read deeper
Matroids: A Geometric Introduction · contents
Gary Gordon & Jennifer McNulty · Publisher contents
Use the publisher contents to locate §9.6, pp.340–348 on chromatic polynomials; return to lesson 24 for the AHK paper and its numbered theorem inputs.
Open the source
Lecture 6: Graph Theory and Coloring
MIT OpenCourseWare · Course lecture · MIT OCW
Lecture 6 gives proper coloring as a prerequisite; it does not establish the matroid polynomial or Hodge arrows.
Open the sourceOpen the interactive lesson