Parity binomial edge ideals (Q737975)

Let \(G\) be a simple undirected graph on \(V(G)\) with edge set \(E(G)\). Let \(K\) be any field and \(K[X,Y]=K[x_i,y_i:i\in V(G)]\) the polynomial ring in \(2|V(G)|\) indeterminates. The parity binomial edge ideal of \(G\) is \(I_G=\langle x_ix_j-y_iy_j:\{i,j\} \in E(G)\rangle \subseteq K[X,Y]\). The main task of the paper under review is to decide whether this ideal is radical. This parity binomial edge ideal resembles the standard binomial edge ideal, and therefore shares a number of common properties. However, the combinatorics involved is more subtler in this case. A key observation here is that when the graph \(G\) is bipartite, then under a natural ring automorphism, the parity binomial edge ideal can be mapped to the standard binomial edge ideal, and therefore is indeed radical. This also explains the focus of this paper on the non-bipartite components of the graph \(G\) and its restrictions, when \(G\) is not bipartite. For the general cases, this paper first studies the minimal Markov basis of the saturation ideal \(J_G=I_G: (\prod_{i\in V(G)}x_iy_i)^\infty\) of \(I_G\) at the coordinate hyperplanes. It turns out that the generators of \(J_G\) can be described by the odd and even walks in \(G\). Next, the authors give an explicit lexicographic Gröbner basis of the parity binomial edge ideal \(I_G\). The generators are reduced binomials which also come as walks in \(G\). This result provides another way to see that in the bipartite case, \(I_G\) is radical. Notice that in general it does not have a square-free initial ideal, unlike the standard binomial edge ideal. Then, the authors characterize the minimal primes of \(I_G\). These primes all have the form \(\mathfrak{q}=\mathfrak{m}_S+\mathfrak{p}\). Here, \(\mathfrak{m}_S=\langle x_s,y_s: s\in S\rangle\) is the monomial part of \(\mathfrak{q}\), which is associated to a disconnector set of \(G\). And \(\mathfrak{p}\) is the binomial part of \(\mathfrak{q}\), which is a sign-split minimal prime of the saturation ideal \(J_{G_S}\) where \(G_S\) is the restriction of \(G\) to \(V(G)\setminus S\). To be relatively more precise, the prime ideal \(\mathfrak{p}\) corresponds to a special selection of parities for the non-bipartite components of \(G_S\), so that the resulting primes are minimal for \(I_G\). As the final step, the authors decomposed the parity binomial edge ideal as an intersection of relatively simpler unital binomial ideals. When \(\text{char}(K)\neq 2\), an induction by applying this decomposition shows that \(I_G\) is radical. When \(\text{char}(K)=2\), a similar decomposition turns out to be a primary decomposition of \(I_G\). Independently, \textit{J. Herzog} et al. [Adv. Appl. Math. 71, 146--173 (2015; Zbl 1322.05098)] considered similar questions by applying a different analysis.