Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Statement

Notation (printed pp. 1--2). For a primitive nn-th root of unity ζ\zeta, the cyclotomic matroid of order nn, μn\mu_n, is the matroid on the nn vectors Zn={1,ζ,ζ2,…,ζn−1}Z_n=\{1,\zeta,\zeta^2,\ldots,\zeta^{n-1}\} of the Q\mathbb Q-vector space Q(ζ)\mathbb Q(\zeta); its rank is ϕ(n)\phi(n). For a dd-dimensional simplicial complex Δd\Delta^d with H~d−1(Δ,F)=0\widetilde H_{d-1}(\Delta,\mathbb F)=0 and exactly nn facets, the simplicial matroid S(Δd,F)\mathcal S(\Delta^d,\mathbb F) is the matroid on the nn facets represented over F\mathbb F by the columns of the boundary map ∂d:C~d(Δd,F)→C~d−1(Δd,F)\partial_d:\widetilde C_d(\Delta^d,\mathbb F)\to\widetilde C_{d-1}(\Delta^d,\mathbb F). The complex Δn1,…,nrr−1\Delta^{r-1}_{n_1,\ldots,n_r} is the simplicial join of 00-dimensional complexes Δn10,…,Δnr0\Delta^0_{n_1},\ldots,\Delta^0_{n_r}, where Δni0\Delta^0_{n_i} consists of nin_i disjoint vertices; a face takes at most one vertex from each Δni0\Delta^0_{n_i}. The dual M∗M^* of a matroid MM is the matroid whose bases are the complements of the bases of MM (p. 1).

Theorem 1 (printed p. 2). "Let n=p1m1⋯prmrn=p_1^{m_1}\cdots p_r^{m_r}, with p1,…,prp_1,\ldots,p_r distinct primes and m1,…,mrm_1,\ldots,m_r positive integers.

Then the following two matroids representable over Q\mathbb Q are dual:

  • The cyclotomic matroid μn\mu_n.
  • The direct sum of p1m1−1⋯prmr−1p_1^{m_1-1}\cdots p_r^{m_r-1} copies of S(Δp1,…,prr−1,Q)\mathcal S(\Delta^{r-1}_{p_1,\ldots,p_r},\mathbb Q)."

The two ground sets both have nn elements: Δp1,…,prr−1\Delta^{r-1}_{p_1,\ldots,p_r} has p1⋯prp_1\cdots p_r facets, one for each choice of a vertex from every Δpi0\Delta^0_{p_i}, and the direct sum has p1m1−1⋯prmr−1p_1^{m_1-1}\cdots p_r^{m_r-1} copies of it. The statement leaves the matching of the ground sets implicit; the proof (pp. 3--4) makes it by splitting ZnZ_n into p1m1−1⋯prmr−1p_1^{m_1-1}\cdots p_r^{m_r-1} blocks, each a copy of μp1⋯pr\mu_{p_1\cdots p_r}, and, for square-free nn, through the Chinese Remainder Theorem, which matches a facet, one residue modulo each pip_i, with the root ζj\zeta^j whose exponent has those residues. This is a filing observation, not a review verdict.

Consequences recorded in the paper. Remark 6 (p. 5): if nn is divisible by at most two primes, the complex Δp1,p2r−1\Delta^{r-1}_{p_1,p_2} is a graph, the complete bipartite graph Kp1,p2K_{p_1,p_2} (p. 3), and μn\mu_n is cographic; if nn is odd, μ2n\mu_{2n} is the parallel extension of μn\mu_n with one parallel copy of each ground-set element, so that for odd primes p,qp,q the matroid μ2pq\mu_{2pq} is the cographic matroid of the graph obtained from Kp,qK_{p,q} by doubling every edge. Remark 5 (p. 5) records, citing Johnsen, that the primitive nn-th roots of unity form a Q\mathbb Q-basis of Q(ζ)\mathbb Q(\zeta) if and only if nn is square-free, and identifies the corresponding basis of the simplicial matroid for square-free nn as a union of vertex stars, a contractible complex.

Source. Jeremy L. Martin and Victor Reiner, "Cyclotomic and simplicial matroids," arXiv:math/0402206v1 (2004), published in Israel J. Math. 150 (2005), 229--240; Theorem 1 on printed p. 2 of the arXiv preprint. Labels and pages here are the preprint's; the edition read is identified in the source digest.

Read depth. Claims checked: the definitions (pp. 1--2), Theorem 1, Lemma 3 (p. 3) and Remarks 5 and 6 (p. 5) were read clause by clause on the page images of the preprint. The proof (pp. 3--4) was read for structure only; nothing here is independently reviewed.

Proof pointer

§ 2, pp. 3--4. Lemma 3 (p. 3) says that a matroid whose ground set splits into parts whose restricted ranks add up to the full rank is the direct sum of the restrictions. With s=p1⋯prs=p_1\cdots p_r and t=n/st=n/s, the sets Ej={ζj,ζj+t,…,ζj+(s−1)t}E_j=\{\zeta^j,\zeta^{j+t},\ldots,\zeta^{j+(s-1)t}\}, 0≤j≤t−10\le j\le t-1, partition ZnZ_n, each restriction is isomorphic to μs\mu_s, and ϕ(n)=t ϕ(s)\phi(n)=t\,\phi(s), so μn\mu_n is the direct sum of tt copies of μs\mu_s. Since duality commutes with direct sums, it remains to treat square-free nn. There the tensor product over the primes p∣np\mid n of the two-term complexes Q→Qp\mathbb Q\to\mathbb Q^p is the augmented cochain complex of the join Δp1,…,pr\Delta_{p_1,\ldots,p_r}; by the Künneth formula it is exact except at the top, where the Chinese Remainder Theorem identifies the map onto the top cohomology with Qn→Q(ζ)\mathbb Q^n\to\mathbb Q(\zeta), ej↦ζje_j\mapsto\zeta^j. So the last coboundary spans the kernel of that map, and its transpose, whose columns represent the simplicial matroid, represents the dual of μn\mu_n.

Dependencies

Lemma 3 (p. 3), which the paper calls a well-known general fact; the Künneth formula over Q\mathbb Q and the Chinese Remainder Theorem. Remark 5 cites K. Johnsen, Lineare Abhängigkeiten von Einheitswurzeln, Elem. Math. 40 (1985), 57--59, which has no library card.

Bears on

  • Problem 774: indirectly. The circuits of μn\mu_n are the minimal rational linear relations among the nn-th roots of unity; as circuits of a matroid are the complements of the hyperplanes of its dual, the theorem identifies them with complements of hyperplanes of the direct sum of simplicial matroids, which for nn with two distinct prime factors are the minimal edge cuts of a copy of Kp1,p2K_{p_1,p_2} (Remark 6). A Q\mathbb Q-linearly independent set of roots of unity is dissociated, but dissociation forbids only relations with coefficients in {−1,0,1}\{-1,0,1\}, so this dictionary describes a stronger condition. The theorem concerns roots of unity, not subsets of the natural numbers, and proves nothing about dissociated or proportionately dissociated sets.