Graph minors and treewidth
Measure tree-like structure with bags, distinguish minors from subgraphs, and state the grid-minor obstruction with its proper quantifiers.
On this page
The question for today
What you’ll understand
- Check all three tree-decomposition conditions and calculate its width.
- Explain minor monotonicity without conflating contraction and deletion.
- Use the grid-minor theorem to distinguish planar-exclusion structure from mere sparsity.
Density measures how many edges exist, but it does not fully describe how complicated their interactions are. A tree can have arbitrarily many vertices while admitting a simple recursive structure. Treewidth asks how closely a graph can be assembled from small overlapping pieces arranged along a tree. We use finite undirected simple graphs; when an edge is contracted, resulting loops are discarded and parallel edges are merged to retain this convention.
Contraction changes the pattern being searched for
A graph H is a minor of G if one can obtain H by deleting vertices, deleting edges, and contracting edges. Equivalently, assign to each vertex of H a disjoint nonempty connected branch set in G, with an edge between branch sets whenever H requires one. Contracting each branch set then realizes H after deleting extras. Unlike a subgraph, a minor can shorten long connected routes. A cycle of length four has a triangle minor even though it contains no triangle subgraph.
A class is minor-closed if every minor of a graph in the class remains inside it. Planarity is an example: deletion and contraction preserve a planar drawing after suitable local adjustment. Neither a numerical density nor a fixed vertex labeling is the defining object. A minor theorem addresses an entire pattern after permitted contractions, whereas a subgraph extremal theorem addresses literal vertex-edge configurations. Switching these meanings can turn a true bound into a false one.
Bags must remain coherent along the indexing tree
A tree decomposition consists of a tree T and a vertex bag B_t at each tree node. Every graph vertex must appear in some bag, every graph edge must have both endpoints together in some bag, and the bags containing any fixed graph vertex must induce a connected subtree of T. This last running-intersection condition prevents incompatible copies of a vertex from appearing in separated pieces. Merely covering each edge with a small bag is not enough.
The width is the largest bag size minus one, and treewidth is the minimum width over all decompositions. A nontrivial tree has treewidth one; a cycle has treewidth two. Every graph with treewidth at most k has a vertex of degree at most k, and the property persists in subgraphs. Thus it is k-degenerate and has at most kn minus k(k+1)/2 edges when n is at least k+1. The converse fails: low edge count does not force a small-width decomposition.
Large grids obstruct uniformly small width
The grid-minor theorem states that for each positive integer r there is an integer f(r) such that every graph of treewidth at least f(r) contains the r-by-r grid as a minor. This qualitative theorem is stated here, not proved, and no sharp numerical bound for f is asserted. Large square grids have unbounded treewidth despite being planar with maximum degree four. Consequently sparsity and planarity alone do not make treewidth uniformly bounded.
For a fixed finite graph H, the class of H-minor-free graphs has bounded treewidth exactly when H is planar. The forward obstruction for nonplanar H is immediate from grids: all grids are planar and avoid H, yet their widths grow. The converse combines the grid theorem with the fact that every fixed planar graph is a minor of a sufficiently large grid. This is a uniform statement about a graph class, not a claim that excluding a minor makes every individual graph literally a tree.
Work it through
A width-two decomposition of a four-cycle
For the cycle with edges 12, 23, 34, and 41, verify the two bags {1,2,3} and {1,3,4} joined by one tree edge. Find its exact treewidth.
All four vertices appear. Edges 12 and 23 lie in the first bag, while 34 and 41 lie in the second, so every graph edge is covered.
Vertices 1 and 3 appear in both adjacent bags; vertices 2 and 4 each appear in only one. Each vertex's bag set is connected. The largest bag size is three, giving width two.
Contracting edge 12 produces a triangle minor. A triangle has treewidth two, so minor monotonicity supplies the lower bound two. Combined with the displayed decomposition, the original cycle has treewidth exactly two.
The decomposition is valid and optimal. The shared pair {1,3} need not itself be an edge of the original graph.
The living notebook
Explore the model
Interactive lab · modular
Compare a cycle with a contracted cycle
Pause, reset, then set modulus m=4 and additive step a=1. Predict the orbit 0,1,2,3,0 and interpret its arrows as an undirected four-cycle. For a fresh comparison, pause and reset, set m=3 and a=1, and predict a three-cycle. On paper, contract one edge of the four-cycle to obtain that triangle. The instrument draws residue orbits; it does not perform contractions or compute tree decompositions.
- Residue r0
- 0
- Orbit period · m/gcd(m,a)
- 12
- Distinct residues visited
- 1 of 12
Addition modulo m starts at zero: rₙ = na mod m. Playback advances two additions per second. The orbit has period m/gcd(m,a), including period one for a = 0. This is not a cryptographic security test or a primality algorithm. Parameter changes pause playback and recompute the model at the current elapsed time. Reset restores both parameters and returns time to zero. Playback starts paused. Adjust either parameter, then choose Play or Step.
Look forA triangle can be a minor of a triangle-free cycle, so minor containment differs from literal subgraph containment.
Pause & reason
Check your understanding
Your turn
Try it yourself
Why does excluding K₅ as a minor not imply a uniform bound on treewidth? Give a graph family and explain both required properties.
Take a moment to make a prediction or work it out on paper. Open a hint when you need one.
Give me a hint
Use square grids and remember that every minor of a planar graph is planar.
Compare with the solution
The r-by-r square grids are planar. Since K₅ is nonplanar and planarity is preserved by taking minors, every such grid is K₅-minor-free. Their treewidth is unbounded as r grows. Thus the whole K₅-minor-free class has no uniform treewidth bound. This does not contradict the fixed planar-excluded-minor result, because K₅ is not planar.
Keep the idea
Take it with you
- Minor containment permits connected branch-set contractions and is weaker than subgraph containment.
- Tree decompositions need edge coverage and a connected bag set for every vertex.
- Grid minors distinguish bounded treewidth from sparsity or planarity alone.
Read further
Original sources and deeper explanations to follow at your own pace.
Optional progress, saved in this browser when available. Revisit any lesson at any time.
Lesson not marked complete.