Skip to content

6. classifying sets of graphs

problem

For each of the following sets of Graphs\mathsf{Graphs}, determine whether it is meager, comeager, or neither, as well as whether it is null, conull, or neither with respect to the fair coin-flip measure.

  1. Bip{GGraphs:G is bipartite}\mathsf{Bip}≔\{G∈\mathsf{Graphs}:\text{\(G\) is bipartite}\}
  2. Conn{GGraphs:G is connected}\mathsf{Conn}≔\{G∈\mathsf{Graphs}:\text{\(G\) is connected}\}
  3. Mat{GGraphs:G is has a perfect matching}\mathsf{Mat}≔\{G∈\mathsf{Graphs}:\text{\(G\) is has a perfect matching}\}

  1. claim

    Bip\mathsf{Bip} is nowhere dense (and therefore meager).

    proof

    We wish to show that given any non-empty open set UGraphsU⊆\mathsf{Graphs} there exists a non-empty open VUV⊆U such that VBip=V∩\mathsf{Bip}=∅. It suffices to show this for just basic open sets, i.e. finitely-determined sets.

    Let UGraphsU⊆\mathsf{Graphs} be finitely-determined. That is to say, membership in UU depends only on whether a graph contains edges among some fixed finite set I{{u,v}:u,vN,  uv}I⊂\{\{u,v\}:u,v∈ℕ,\;u≠v\}. Let nNn∈ℕ be so large that it isn’t incident to any of the edges in II; i.e., membership in UU doesn’t depend on any edges involving nn.

    Then let V={GU:G contains the triangle {n,n+1,n+2}}V=\{G∈U:\text{\(G\) contains the triangle \(\{n,n+1,n+2\}\)}\}, which is non-empty. And notice that no graphs in VV are bipartite since they contain a triangle (odd cycle).

    claim

    Bip\mathsf{Bip} is null.

    proof

    For each nNn∈ℕ let

    Tn={GGraphs:G doesn’t contain the triangle {3n,3n+1,3n+2}}, T_n=\{G∈\mathsf{Graphs}:\text{\(G\) doesn’t contain the triangle \(\{3n,3n+1,3n+2\}\)}\},

    and let S=nNTnS=⋂_{n∈ℕ} T_n. Observe that BipS\mathsf{Bip}⊆S since every bipartite graph must be triangle-free. Also observe that the kk-th partial intersection has measure

    μ(nkTnSkk)(k)S=(78)k+1. μ\left(\smash{\underbrace{⋂_{n≤k} T_n}_{≕S_k}} \vphantom{⋂_k}\right) \vphantom{\underbrace{\left(⋂_k\right)}_S} = \left(\frac 7 8\right)^{k+1}.

    Then μ(Bip)μ(S)μ(Sk)=(7/8)k+1μ(\mathsf{Bip})≤μ(S)≤μ(S_k)=(7/8)^{k+1}. By sending kk→∞ we obtain μ(Bip)=0μ(\mathsf{Bip})=0.

  2. claim

    Conn\mathsf{Conn} is comeager.

    proof

    We will show that Conn\mathsf{Conn} is dense and GδG_δ.

    • Dense: Fix any finitely-determined UGraphsU⊆\mathsf{Graphs}. Let nNn∈ℕ be large enough so that no edges involving nn affect membership in UU. Let GUG∈U, and add edges {k,n}\{k,n\} for all knk≠n to obtain a new graph GG'. Since membership in UU is unaffected by edges involving nn, we have that GUG'∈U as well. And also GG' is connected because all vertices are adjacent to nn.
    • GδG_δ: For each pair i,jNi,j∈ℕ, define Sij={GGraphs:path between i and j in G}. S_{ij}=\{G∈\mathsf{Graphs}:∃\text{path between \(i\) and \(j\) in \(G\)}\}. By writing Sij=pN{GGraphs:p is a path between i and j in G} S_{ij}=⋃_{p∈ℕ^*} \{G∈\mathsf{Graphs}:\text{\(p\) is a path between \(i\) and \(j\) in \(G\)}\} observe that SijS_{ij} is open. And also observe that connectedness means exactly that all pairs of vertices are joined by some path, i.e. Conn=i,jNSij\mathsf{Conn}=⋂_{i,j∈ℕ}S_{ij}.
    claim

    Conn\mathsf{Conn} is conull.

    proof

    Fix any distinct i,jNi,j∈ℕ, and define

    Sij={GGraphs:there are no paths between i and j in G}. S_{ij}=\{G∈\mathsf{Graphs}:\text{there are no paths between \(i\) and \(j\) in \(G\)}\}.

    Next for every kN{i,j}k∈ℕ∖\{i,j\} define

    Tijk={GGraphs:ikj is not a path in G}. T_{ijk}=\{G∈\mathsf{Graphs}:\text{\(i∼k∼j\) is not a path in \(G\)}\}.

    Note that SijkN{i,j}TijkS_{ij}⊆⋂_{k∈ℕ∖\{i,j\}} T_{ijk}, so μ(Sij)(3/4)nμ(S_{ij})≤(3/4)^n for all nNn∈ℕ, and therefore μ(Sij)=0μ(S_{ij})=0.

    Meanwhile observe that GConnG∈\mathsf{Conn}^∁ iff there exist i,jNi,j∈ℕ not connected by a path; i.e.,

    Conn=ijNSij. \mathsf{Conn}^∁ = ⋃_{i≠j∈ℕ} S_{ij}.

    By countable (sub-)additivity μ(Conn)μ(Sij)=0μ(\mathsf{Conn}^∁)≤∑μ(S_{ij})=0.