Abstract
We give an explicit formula for the Hilbert–Poincaré series of the parity binomial edge ideal of a complete graph \(K_{n}\) or equivalently for the ideal generated by all \(2\times 2\)-permanents of a \(2\times n\)-matrix. It follows that the depth and Castelnuovo–Mumford regularity of these ideals are independent of n.
1 Introduction
Let \(R = \Bbbk [x_{1},\dots ,x_{n},y_{1},\dots ,y_{n}]\) be a standard graded polynomial ring in 2n indeterminates. The parity binomial edge ideal of an undirected simple graph G on \([n]=\{1,\dots ,n\}\) is
where E(G) is the edge set of G. This ideal was defined and studied in [11] in formal similarity to the binomial edge ideals of [7, 13]. If \({{\mathrm{char}}}(\Bbbk )\ne 2\), then the linear coordinate change \(x_{i} \mapsto (x_{i}-y_{i})\) and \(y_{i} \mapsto (x_{i}+y_{i})\) turns this ideal into the permanental edge ideal
We aim to understand homological properties of these ideals and we view such understanding as helpful in the context of complexity theory and the dichotomy of permanents and determinants. In linear algebra it is known that determinants can be evaluated quickly with Gaussian elimination, but permanents are \(\#\)P-complete and thus NP-hard to evaluate. This complexity distinction is also visible for ideals generated by determinants and permanents, as the permanental versions are often much harder to analyze and have nice properties much more rarely. For details and history we recommend [12] which treats ideals of \(2\times 2\)-permanents of \(m\times n\)-matrices in detail.
\(2\times 2\)-permanental ideals also arise from the study of orthogonal embeddings of graphs in \({\mathbb {R}}^{2}\) as the Lovász–Saks–Schrijver ideals of [8]. That paper also contains information about radicality and Gröbner bases of parity binomial edge ideals. Badiane, Burke and Sköldberg proved in [2] that the universal Gröbner basis and the Graver basis coincide for parity binomial edge ideals of complete graphs. The case of bipartite graphs is also special, as then binomial edge ideals and parity binomial edge ideals agree up to a linear coordinate change. A coherent presentation of our knowledge about these binomial ideals can be found in [6], in particular Chapter 7.
In this paper we are concerned with permanental ideals of \(2\times n\)-matrices, but switch to the representation as parity binomial edge ideals of complete graphs, as this seems easier to analyze. For example, the permanental ideal contains monomials by [12, Lemma 2.1] and these make the combinatorics more opaque [10]. Due to the linear coordinate change, our computations of homological invariants are valid for both ideals unless \({{\mathrm{char}}}(\Bbbk ) = 2\), in which case the permanental ideal and the determinantal ideal agree.
The binomial edge ideal of a complete graph, also known as the standard determinantal ideal of a generic \(2\times n\)-matrix, is well understood. It has a linear minimal free resolution independent of n, constructed explicitly by Eagon and Northcott [4]. Parity binomial edge ideals of complete graphs do not have a linear resolution and their Betti numbers have no obvious explanation.
Example 1.1
The package BinomialEdgeIdeals in Macaulay2 [5] easily generates the following Betti table of \({\mathcal {I}}_{K_{7}}\). The Betti table agrees with the Betti table of a permanental ideal of a generic \(2\times 7\)-matrix.
From computations for the first few n one can observe that the Castelnuovo–Mumford regularity (the index of the last row of the Betti table, see Sect. 2 for definitions) of \(R/{\mathcal {I}}_{K_{n}}\) appears to be independent of \(n \ge 4\). That \({{\mathrm{reg}}}(R/{\mathcal {I}}_{K_{n}}) = 3\) was conjectured by the second author and Krüsemann [9, Remark 2.15] and is now our Theorem 3.6. Our main results are explicit formulas for the Hilbert–Poincaré series, the depth, the Castelnuovo–Mumford regularity, and some extremal Betti numbers in the case of a complete graph. The proof of our theorem relies on good knowledge of the primary decomposition of \({\mathcal {I}}_{K_{n}}\) from [11] and the resulting exact sequences. At the moment it is not clear if the techniques can be generalized to other graphs or maybe even yield the conjectured upper bound \({{\mathrm{reg}}}(R/{\mathcal {I}}_{G}) \le n\) from [9, Remark 2.15].
2 Basics of (parity) binomial edge ideals
Throughout this paper, let G be a simple (i.e. finite, undirected, loopless and without multiple edges) graph on the vertex set \(V(G)=[n] := \{1,\ldots ,n\}\). Let E(G) denote the set of edges of G. Each graded R-module and in particular \(R/{\mathcal {I}}_{G}\) has a minimal graded free resolution
where \(R(-j)\) denotes the free R-module obtained by shifting the degrees of R by j. The number \(\beta _{i,j}(R/{\mathcal {I}}_G)\) is the (i, j)-th graded Betti number of \(R/{\mathcal {I}}_G\). Let \(H_{R/{\mathcal {I}}_G}\) be the Hilbert function of \(R/{\mathcal {I}}_G\). The Hilbert–Poincaré series of the R-module \(R/{\mathcal {I}}_G\) is
By [14, Theorem 16.2], this series has a rational expression
The numerator is the Hilbert–Poincaré polynomial of \(R/{\mathcal {I}}_G\) and has the form
It encodes different homological invariants of \(R/{\mathcal {I}}_{G}\) of which we are particulary interested in the Castelnuovo–Mumford regularity
and the projective dimension of \(R/{\mathcal {I}}_G\):
In terms of Betti tables, the regularity is the index of the last non-vanishing row, while the projective dimension is the index of the last non-vanishing column of the Betti table. Both are finite for any R-module as R is a regular ring.
The Auslander–Buchsbaum formula [6, Theorem 2.15] relates depth and projective dimension over R as \({{\mathrm{depth}}}(R/{\mathcal {I}}_G) = 2n-{{\mathrm{pdim}}}(R/{\mathcal {I}}_G)\). The Castelnuovo–Mumford regularity and depth could also be computed from vanishing of local cohomology. Using that definition allows to easily deduce some basic properties of the regularity and depth. For instance, the regularity and depth behave well in a short exact sequence. The following lemma appears as [14, Corollary 18.7].
Lemma 2.1
If \(0 \rightarrow A \rightarrow B \rightarrow C\rightarrow 0\) is a short exact sequence of finitely generated graded R-modules with homomorphisms of degree 0, then
-
(1)
\({{\mathrm{reg}}}(B)\le \max \{{{\mathrm{reg}}}(A), {{\mathrm{reg}}}(C)\}\),
-
(2)
\({{\mathrm{reg}}}(A)\le \max \{{{\mathrm{reg}}}(B), {{\mathrm{reg}}}(C)+1\}\),
-
(3)
\({{\mathrm{reg}}}(C)\le \max \{{{\mathrm{reg}}}(A)-1, {{\mathrm{reg}}}(B)\}\),
-
(4)
\({{\mathrm{depth}}}(B)\ge \min \{{{\mathrm{depth}}}(A),{{\mathrm{depth}}}(C)\}\),
-
(5)
\({{\mathrm{depth}}}(A)\ge \min \{{{\mathrm{depth}}}(B),{{\mathrm{depth}}}(C)+1\}\),
-
(6)
\({{\mathrm{depth}}}(C)\ge \min \{{{\mathrm{depth}}}(A)-1,{{\mathrm{depth}}}(B)\}\).
As with any binomial ideal, the saturation at the coordinate hyperplanes plays a central role. To this end, let \(g = \prod _{i\in [n]} x_iy_i\) and let
By [11, Proposition 2.7], the generators of the saturation \({\mathcal {J}}_G\) can be explained using walks in G. For our purposes it suffices to know the following generating set which can be derived from [11, Section 2].
Proposition 2.2
If G is a non-bipartite connected graph, then
3 Parity binomial edge ideals of complete graphs
We now consider the parity binomial edge ideal \({\mathcal {I}}_{K_{n}}\) of a complete graph \(K_{n}\) on \(n\ge 3\) vertices. For \(1\le i<j\le n\), let
The parity binomial edge ideal of the complete graph is \({\mathcal {I}}_{K_n} = \left( g_{ij}\mid 1\le i<j\le n\right)\).
We need some further notation. For any \(I\subseteq [n]\) we denote \({\mathfrak {m}}_I := (x_i,y_i\mid i\in I)\). Let \({\mathfrak {p}}^{+} := (x_i+y_i\mid i\in [n])\) and \({\mathfrak {p}}^{-} :=(x_i-y_i\mid i\in [n])\). Denote \(P_{ij} := (g_{ij})+ {\mathfrak {m}}_{[n]\backslash \{i,j\}}\). By [11, Theorem 5.9], there is a decomposition of \({\mathcal {I}}_{K_n}\) as follows.
Proposition 3.1
For \(n\ge 3\), we have
In particular, \(\dim (R/{\mathcal {I}}_{K_n}) = n\).
We analyze \({\mathcal {I}}_{K_{n}}\) by regular sequences arising from successively adding the polynomials \(f_{kn}\) or saturating with respect to them. Let \(I_0 := {\mathcal {I}}_{K_n}\) and, inductively for \(1\le k\le n-1\), \(I_k := I_{k-1} + (f_{kn})\).
Lemma 3.2
For \(1\le k\le n-1\), we have
Proof
By Proposition 2.2, \(f_{1n}, \ldots , f_{(k-1)n}\in {\mathcal {J}}_{K_n}\). Moreover, for all \((\ell ,n)\ne (i,j)\) we have \(f_{\ell n}\in P_{ij}\). Thus
Together with Proposition 3.1 the lemma is proven. \(\square\)
Lemma 3.3
For \(1\le k\le n-1\), we have
In particular, \({{\mathrm{depth}}}(R/(I_{k-1}: f_{kn}))= 3\), \({{\mathrm{reg}}}(R/(I_{k-1}: f_{kn})) =1\) and \(P_{R/(I_{k-1}: f_{kn})}(t) = (1-t)^{2n-3}(1+t)\).
Proof
One can check that \(I_{k-1}: f_{kn} \supseteq P_{kn}\) (in fact \({\mathcal {I}}_{K_{n}} : f_{kn} \supseteq P_{kn}\)) by simple calculations like \(x_{1}f_{kn} \equiv -y_{k}g_{1n} \mod {\mathcal {I}}_{K_{n}}\). Now, for all \((k,n)\ne (i,j)\), one can see that \(f_{kn}\) is contained in both \(P_{ij}\) and \({\mathcal {J}}_{K_n}\). By [1, Lemma 4.4], \(P_{ij}: f_{kn} = {\mathcal {J}}_{K_n}: f_{kn} = R\) and \(P_{kn}: f_{kn} = P_{kn}\) because \(P_{kn}\) is a prime that does not contain \(f_{kn}\). Hence by Lemma 3.2, we have \(I_{k-1}: f_{kn} \subseteq P_{kn}\) and thus \(I_{k-1}: f_{kn} = P_{kn}\).
Using this result, the invariants can be computed for the prime \(P_{kn}\) as follows: \({{\mathrm{depth}}}(R/(I_{k-1}:f_{kn})) = {{\mathrm{depth}}}(R/P_{kn}) = 3\), \({{\mathrm{reg}}}(R/(I_{k-1}: f_{kn})) = {{\mathrm{reg}}}(R/P_{kn}) =1\), and \(P_{R/(I_{k-1}: f_{kn})}(t) = P_{R/P_{kn}}(t) = (1-t)^{2n-3}(1+t)\). \(\square\)
Lemma 3.4
In particular, \({{\mathrm{depth}}}(R/(I_{n-2}: (x_n+y_n)))\ge 3\), \({{\mathrm{reg}}}(R/(I_{n-2}: (x_{n}+y_n)))\le 1\) and \(P_{R/(I_{n-2}: (x_{n}+y_n))}(t) = (1-t)^n +2t(1-t)^{2n-3}\).
Proof
For the lexicographic ordering on \(\Bbbk [x_1,\ldots , x_n, y_1,\ldots , y_n,t]\) induced by \(x_1>\ldots>x_n>y_1>\ldots>y_n>t\), the Gröbner basis for \(J=t{\mathfrak {p}}^{-} + (1-t)P_{n-1,n}\) is
Thus,
This implies the containment \({\mathfrak {p}}^{-}\cap P_{n-1,n} \subseteq I_{n-2}: (x_{n}+y_n)\). Conversely, by Lemma 3.2
For all \(1\le i<j\le n-1\), it is clear that \(x_n+y_n\in P_{ij}\) and so \(P_{ij}:(x_n+y_n)=R\). By [1, Lemma 4.4], \(P_{n-1,n}: (x_n+y_n)=P_{n-1,n}\). Moreover, by Proposition 2.2, we obtain that \({\mathcal {J}}_{K_n}: (x_n+y_n) = {\mathfrak {p}}^{-}\). This implies that \(I_{n-2}: (x_n+y_n) \subseteq {\mathfrak {p}}^{-} \cap P_{n-1,n}\) and thus the conclusion \(I_{n-2}: (x_{n}+y_n) = {\mathfrak {p}}^{-}\cap P_{n-1,n}\).
In order to prove the second part, note that
Therefore one reads off \({{\mathrm{depth}}}(R/({\mathfrak {p}}^{-}+P_{n-1,n})) = 2\) and \({{\mathrm{reg}}}(R/({\mathfrak {p}}^{-}+P_{n-1,n}))=0\). It is clear that \({{\mathrm{depth}}}(R/{\mathfrak {p}}^{-}) =n\) and \({{\mathrm{reg}}}(R/{\mathfrak {p}}^{-})=0\). From the exact sequence
we obtain, using Lemma 2.1, that
and furthermore,
\(\square\)
Lemma 3.5
Let \(J :=(x_n+y_n, I_{n-2})\). Then
and \(P_{R/J}(t) = t(1-t)^n + (1-t)^2 P_{S/{\mathcal {I}}_{K_{n-1}}}(t),\) where \(S = \Bbbk [x_i,y_i\mid 1\le i \le n-1]\).
Proof
In order to prove the lemma, we first check two following claims:
Claim 1 \((J,x_n) = (x_n,y_n, {\mathcal {I}}_{K_{n-1}}).\)
Since \(y_n = (x_n+y_n) - x_n\in (x_n,J)\) and \({\mathcal {I}}_{K_{n-1}}\subseteq I_{n-2}\), we have \((x_n,y_n, {\mathcal {I}}_{K_{n-1}}) \subseteq (J,x_n)\). Conversely, \(x_n+y_n, g_{in}, f_{in} \in (x_n,y_n)\) for \(1\le i\le n-1\) and thus \((J,x_n)\subseteq (x_n,y_n, {\mathcal {I}}_{K_{n-1}})\).
Claim 2 \(J: x_n = {\mathfrak {p}}^+.\)
One can compute \(x_n(x_i+y_i) = (x_ix_n-y_iy_n) + y_i(x_n+y_n)\in J\) for \(1\le i\le n\), so that \(x_n{\mathfrak {p}}^{+} \subseteq J\) which implies that \({\mathfrak {p}}^{+} \subseteq J: x_n\). Conversely, for \(1\le i<j\le n\), we have
Thus, by Proposition 2.2, \({\mathcal {J}}_{K_n} \subseteq {\mathfrak {p}}^{+} \cap (x_1-y_1,\ldots ,x_{n-1}-y_{n-1}, x_n,y_n)\) and \(f_{kn}\in {\mathfrak {p}}^{+} \cap (x_1-y_1,\ldots ,x_{n-1}-y_{n-1}, x_n,y_n)\) for all \(1\le k\le n-2\). Together with Proposition 3.1,
By [1, Lemma 4.4], \(J: x_n \subseteq {\mathfrak {p}}^+\) and thus the claim holds.
Now, we turn to the proof of the lemma. By Claim 1,
Moreover, by Claim 2, we have
From the exact sequence
we obtain
Moreover,
as required. \(\square\)
Theorem 3.6
The Hilbert–Poincaré polynomial of \(R/{\mathcal {I}}_{K_{n}}\) is
In particular, \({{\mathrm{depth}}}(R/{\mathcal {I}}_{K_n}) \ge 3\) and \({{\mathrm{reg}}}(R/{\mathcal {I}}_{K_n})\le 3\).
Proof
The proof is by induction on n. If \(n=3\), then a simple calculation (e.g. in Macaulay2) gives the result. Now assume \(n\ge 4\). For any \(1\le k\le n-1\) there is an exact sequence
By Lemmas 2.1 and 3.3, \({{\mathrm{depth}}}(R/I_{k-1}) \ge \min \{3, {{\mathrm{depth}}}(R/I_{k})\}\), \({{\mathrm{reg}}}(R/I_{k-1}) \le \max \{3, {{\mathrm{reg}}}(R/I_{k})\}\) and \(P_{R/I_{k-1}}(t) = t^2(1-t)^{2n-3}(1+t)+ P_{R/I_{k}}(t)\). This implies that \({{\mathrm{depth}}}(R/I_0) \ge \min \{3,{{\mathrm{depth}}}(R/I_{n-2})\}\), \({{\mathrm{reg}}}(R/I_{0}) \le \max \{3,{{\mathrm{reg}}}(R/I_{n-2})\}\) and
Now consider the following exact sequence
Let \(S := \Bbbk [x_i,y_i\mid 1\le i\le n-1]\). By Lemmas 3.4 and 3.5, \({{\mathrm{depth}}}(R/I_{n-2}) \ge \min \{3,{{\mathrm{depth}}}(S/{\mathcal {I}}_{K_{n-1}})\}\), \({{\mathrm{reg}}}(R/I_{n-2}) \le \max \{1, {{\mathrm{reg}}}(S/{\mathcal {I}}_{K_{n-1}})\}\) and
The induction hypothesis yields \({{\mathrm{depth}}}(S/{\mathcal {I}}_{K_{n-1}})\ge 3\) and \({{\mathrm{reg}}}(S/{\mathcal {I}}_{K_{n-1}}) \le 3\). Therefore \({{\mathrm{depth}}}(R/I_{n-2}) \ge 3\) and \({{\mathrm{reg}}}(R/I_{n-2})\le 3\). This is enough to conclude that \({{\mathrm{depth}}}(R/{\mathcal {I}}_{K_n}) \ge 3\) and \({{\mathrm{reg}}}(R/{\mathcal {I}}_{K_n})\le 3\). Moreover,
as required. \(\square\)
If an ideal has a square-free initial ideal, its extremal Betti numbers agree with that of the initial ideal by [3]. Although the parity binomial edge ideal of a complete graph cannot have a square-free initial ideal (see [11, Remark 3.12]), the bottom right Betti number agrees with that of the initial ideal for any term order.
Corollary 3.7
In particular,
Proof
From Theorem 3.6 we obtain \(\beta _{p,p+r}(R/{\mathcal {I}}_{K_n}) =\frac{n^2-3n+2}{2}\ne 0\), where \(p={{\mathrm{pdim}}}(R/{\mathcal {I}}_{K_n})\) and \(r={{\mathrm{reg}}}(R/{\mathcal {I}}_{K_n})\). Thus, \(p+r=2n\). Since \(P_{R/{\mathcal {I}}_{K_n}}(t) = P_{R/{{\mathrm{in}}}_<({\mathcal {I}}_{K_n})}(t)\), we get
\(\beta _{p,p+r}(R/{\mathcal {I}}_{K_n}) =\beta _{p,p+r}(R/{{\mathrm{in}}}_<({\mathcal {I}}_{K_n}))\). On the other hand, \(r \le 3\) and \(p\le 2n-3\) by the Auslander–Buchsbaum formula. Thus, \(r = 3\) and \(p = 2n-3\). \(\square\)
References
Atiyah, M.F., Macdonald, I.G.: Introduction to Commutative Algebra. Addison-Wesley, Boston (1969)
Badiane, M., Burke, I., Sköldberg, E.: The Universal Gröbner Basis of a Binomial Edge Ideal. Electron. J. Comb. 24(4), 12 (2017)
Conca, A., Varbaro, M.: Square-free Gröbner degenerations. Invent. math. (2020). https://doi.org/10.1007/s00222-020-00958-7
Eagon, J.A., Northcott, D.G.: Ideals defined by matrices and a certain complex associated with them. Proc. R. Soc. Lond. Ser. A 269(1337), 188–204 (1962)
Grayson, D.R., Stillman, M.E.: Macaulay2, a software system for research in algebraic geometry. http://www.math.uiuc.edu/Macaulay2/. Version 1.15. Accessed Jan 2020
Herzog, J., Hibi, T., Ohsugi, H.: Binomial Ideals, Graduate Texts in Mathematics, 279. Springer, Cham (2018)
Herzog, J., Hibi, T., Hreinsdóttir, F., Kahle, T., Rauh, J.: Binomial edge ideals and conditional independence statements. Adv. Appl. Math. 45(3), 317–333 (2010)
Herzog, J., Macchia, A., Madani, S.S., Welker, V.: On the ideal of orthogonal representations of a graph in \({\mathbb{R}}^{2}\). Adv. Appl. Math. 71, 146–173 (2015)
Kahle, T., Krüsemann, J.: Binomial edge ideals of cographs. Preprint, arXiv:1906.05510
Kahle, T., Miller, E.: Decompositions of commutative monoid congruences and binomial ideals. Algebra Number Theory 8(6), 1297–1364 (2014)
Kahle, T., Sarmiento, C., Windisch, T.: Parity binomial edge ideals. J. Algebraic Comb. 44(1), 99–117 (2016)
Laubenbacher, R.C., Swanson, I.: Permanental ideals. J. Symb. Comput. 30, 195–205 (2000)
Ohtani, M.: Graphs and Ideals generated by some 2-minors. Commun. Algebra 39(3), 905–917 (2011)
Peeva, I.: Graded Syzygies. Algebra and Applications, 14. Springer, London (2011)
Acknowledgements
Open Access funding provided by Projekt DEAL. This paper was done when Do Trong Hoang visited Department of Mathematics, Otto-von-Guericke Universität Magdeburg with the support of Deutscher Akademischer Austauschdienst (DAAD). Thomas Kahle acknowledges support from the DFG (314838170, GRK 2297 MathCoRe).
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Do, T.H., Kahle, T. Hilbert–Poincaré series of parity binomial edge ideals and permanental ideals of complete graphs. Collect. Math. 72, 471–479 (2021). https://doi.org/10.1007/s13348-020-00294-2
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s13348-020-00294-2