This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Gorenstein and Cohen-Macaulay Matching Complexes

Ashkan Nikseresht
Department of Mathematics, College of Science, Shiraz University,
71457-13565, Shiraz, Iran
E-mail: [email protected]
Abstract

Let HH be a simple undirected graph. The family of all matchings of HH forms a simplicial complex called the matching complex of HH. Here , we give a classification of all graphs with a Gorenstein matching complex. Also we study when the matching complex of HH is Cohen-Macaulay and, in certain classes of graphs, we fully characterize those graphs which have a Cohen-Macaulay matching complex. In particular, we characterize when the matching complex of a graph with girth at least 5 or a complete graph is Cohen-Macaulay.

Keywords: Cohen-Macaulay ring; Gorenstein ring; Matching complex; Line graph; Edge ideal;
Mathematics Subject Classification (2020): 13F55; 05E40; 05E45.

1 Introduction

In this paper, KK denotes a field and S=K[x1,,xn]S=K[x_{1},\ldots,x_{n}]. Let GG be a simple graph on vertex set V(G)={v1,,vn}\mathrm{V}(G)=\{v_{1},\ldots,v_{n}\} and edge set E(G)\mathrm{E}(G). Then the edge ideal I(G)I(G) of GG is the ideal of SS generated by {xixj|vivjE(G)}\{x_{i}x_{j}|v_{i}v_{j}\in\mathrm{E}(G)\}. A graph GG is called Cohen-Macaulay (resp. Gorenstein) when S/I(G)S/I(G) is Cohen-Macaulay (resp. Gorenstein) for every field KK. Many researchers have tried to combinatorially characterize Cohen-Macaulay (CM, for short) or Gorenstein graphs in specific classes of graphs, see for example, [19, 8, 13, 15, 4, 6, 7]).

Note that the family of independent sets of GG is a simplicial complex which we denote by Δ(G¯)\Delta(\overline{G}), and is called the independence complex of GG. Suppose that IΔSI_{\Delta}\subseteq S is the Stanley-Reisner ideal of a simplicial complex Δ\Delta. Then Δ\Delta is called CM (resp. Gorenstein) when S/IΔS/I_{\Delta} is CM (resp. Gorenstein) for every field KK. Noting that IΔ(G¯)=I(G)I_{\Delta(\overline{G})}=I(G), we see that GG is CM or Gorenstein if and only if Δ(G¯)\Delta(\overline{G}) is so.

Assume that HH is a simple undirected graph and G=L(H)G=\mathrm{L}(H) is the line graph of HH, that is, edges of HH are vertices of GG and two vertices of GG are adjacent if they share a common endpoint in HH. Line graphs are well-known in graph theory and have many applications. In particular, there are some well-known characterizations of line graphs (see for example [17, Section 7.1]). Also a linear time algorithm is presented by Lehot in [10] that given a graph GG, checks whether GG is a line graph and if GG is a line graph, returns a graph HH such that G=L(H)G=\mathrm{L}(H). Note that faces of Δ(G¯)\Delta(\overline{G}) are exactly the matchings of HH, hence Δ(G¯)\Delta(\overline{G}) is known and studied as the matching complex of HH in the literature (see, for example, [16, 1, 9] and the references therein). In the survey article [16], one can find many known properties of matching complexes of complete and complete bipartite graphs. In particular, it is known that the matching complex of a complete bipartite graph with part sizes mm and nn with mnm\leq n is CM if and only if n2m1n\geq 2m-1.

Here, first in Section 2, we characterize graphs with a Gorenstein matching complex. Then in Section 3, we focus on the Cohen-Macaulay property and give classifications of graphs with a CM matching complex which are in one of the following classes: Cameron-Walker graphs, graphs with girth at least 5, complete graphs. Note that, in other words, we present characterizations of Gorenstein line graphs and CM line graphs in the aforementioned classes.

For definitions and basic properties of simplicial complexes and graphs one can see [6] and [17], respectively. In particular, all notations used in the sequel whose definitions are not stated, are as in these references.

2 Gorenstein Matching Complexes

In what follows, HH is a simple undirected graph with at least one edge, G=L(H)G=\mathrm{L}(H) is the line graph of HH and Δ=Δ(G¯)\Delta=\Delta(\overline{G}) is the matching complex of HH. First we characterize when this complex is Gorenstein. It should be mentioned that this characterization also follows from Theorem 4.3 of the recent paper [1] and Theorem 5.1 in chapter II of [14], though, the proof presented here is quite different and does not need homological concepts. Here if FV(G)=E(H)F\subseteq\mathrm{V}(G)=\mathrm{E}(H), then by N[F]\mathrm{N}[F] (or NG[F]N_{G}[F]) we mean F{vV(G)|uvE(G) for some uF}F\cup\{v\in\mathrm{V}(G)|uv\in\mathrm{E}(G)\text{ for some }u\in F\}, and we set GF=GN[F]G_{F}=G\setminus\mathrm{N}[F] and HF=HV(F)H_{F}=H\setminus\mathrm{V}(F), where V(F)\mathrm{V}(F) is the set of all vertices of HH incident to an edge in FF. It is routine to check that GF=L(HF)G_{F}=\mathrm{L}(H_{F}).

Theorem 2.1.

Suppose that HH is a graph with at least one edge, G=L(H)G=\mathrm{L}(H) and KK be an arbitrary field. Then GG (or equivalently, the matching complex of HH) is Gorenstein over KK if and only if each connected component of HH is either a 5-cycle or a path with length 2\leq 2 or the graph in Figure 1.

Refer to caption
Figure 1: A graph whose line graph has a Gorenstein independence complex.
Proof.

Note that GG is Gorenstein if and only if each connected component of GG is so. Since each connected component of GG is the line graph of a connected component of HH, we assume that HH is connected. The line graph of the stated graphs are either K2K_{2} or K1K_{1} or the complement of cycles of length 5 or 6 and hence are Gorenstein by [13, Corollary 2.4]).

Conversely assume that GG is Gorenstein. If dimΔ(G¯)=0\dim\Delta(\overline{G})=0, then GG is a Gorenstein complete graph and thus either GG is a vertex and hence HH is K2K_{2} or GG is K2K_{2} and hence HH is a path of length 2. Suppose dimΔ(G¯)1\dim\Delta(\overline{G})\geq 1. Then by [13, Theorem 2.3], for each maximum size matching MM of HH, if F=M{a,b}F=M\setminus\{a,b\} for some a,bMa,b\in M, then GFG_{F} is the complement of a cycle of length at least 4. In particular, as abE(GF¯)ab\in\mathrm{E}(\overline{G_{F}}) there are edges c,dE(HF)=V(GF)c,d\in\mathrm{E}(H_{F})=\mathrm{V}(G_{F}) such that ac,bdE(G)ac,bd\in\mathrm{E}(G) but bc,adE(G)bc,ad\notin\mathrm{E}(G). Two cases are possible; either cc and dd have a common endpoint in HH or cc and dd do not have a common endpoint in HH. In the latter case, cc and dd are adjacent in GF¯\overline{G_{F}} and hence GF¯\overline{G_{F}} is cycle of length 4 and HFH_{F} is the disjoint union of two paths of length 2. In the former case, any other edge of HFH_{F} is adjacent to both aa and bb in GG (since in the cycle GF¯\overline{G_{F}}, degree of each vertex is 2, and aa and bb are already adjacent to two vertices among a,b,c,da,b,c,d), therefore in HFH_{F} any edge except a,b,c,da,b,c,d must be incident to an endpoint of aa and also to an endpoint of bb. There are 4 possible such edges and hence only 16 possibilities exist for HFH_{F}. Because GF¯\overline{G_{F}} is also the complement of a cycle, it follows that the only possible graphs for HFH_{F} is the 5-cycle or the graph in Figure 1.

Note that in both cases exactly one endpoint of each of aa and bb is adjacent to a vertex of HH not in V(M)\mathrm{V}(M). It follows that if the set of all vertices of HH not covered by MM is {x1,,xt}\{x_{1},\ldots,x_{t}\} and M={u1v1,,urvr}M=\{u_{1}v_{1},\ldots,u_{r}v_{r}\} for ui,viV(H)u_{i},v_{i}\in\mathrm{V}(H), then for each 1ir1\leq i\leq r, exactly one of uiu_{i} or viv_{i} is adjacent to exactly one of the xjx_{j}’s. We can assume that u1,,ur1u_{1},\ldots,u_{r_{1}} are the vertices adjacent to x1x_{1}, ur1+1,,ur1+r2u_{r_{1}+1},\ldots,u_{r_{1}+r_{2}} are adjacent to x2x_{2} and …. Suppose t>1t>1. Note that HH is connected and there is no edge between xix_{i}’s, because MM is a maximum matching. Thus there should be an edge with one end on some uiviu_{i}v_{i} for i>r1i>r_{1} and one end on ujvju_{j}v_{j} for jr1j\leq r_{1}. But then setting a=uivia=u_{i}v_{i} and b=ujvjb=u_{j}v_{j}, we see that GF¯\overline{G_{F}} is not a cycle for F=M{a,b}F=M\setminus\{a,b\}. Therefore, t=1t=1.

Suppose that dimΔ(G¯)>1\dim\Delta(\overline{G})>1, MM is a maximum matching of HH and a,b,cMa,b,c\in M. Let F=M{a,b,c}F=M\setminus\{a,b,c\}. Then for each l{a,b,c}l\in\{a,b,c\}, HF{l}H_{F\cup\{l\}} is either a 5-cycle or the graph in Figure 1. It follows that, HFH_{F} is one of the graphs in Figure 2. If HH is any of the first three graphs from the left in Figure 2 and ee is the edge specified in the figure, then HF{e}H_{F\cup\{e\}} is neither a 5 cycle nor the graph in Figure 1, although, F{e}F\cup\{e\} is a matching with size dimΔ(G¯)1\dim\Delta(\overline{G})-1.

Refer to caption
Figure 2: Illustrations for the proof of Theorem 2.1.

Consequently, HFH_{F} must be the rightmost graph depicted in Figure 2. Suppose that I(G,x)I(G,x) is the independence polynomial of GG, that is, the coefficient of xnx^{n} is the number of independent sets of GG. Then one can compute that I(GF,x)=1+12x+36x2+24x3I(G_{F},x)=1+12x+36x^{2}+24x^{3}. It follows that I(GF,1)=11=(1)dimΔ(GF¯)+1I(G_{F},-1)=1\neq-1=(-1)^{\dim\Delta(\overline{G_{F}})+1} and hence GFG_{F} is not Gorenstein by [13, Theorem 2.3]. But according to [12, Corollary 2.4], if GG is Gorenstein, then GFG_{F} is Gorenstein for every independent set FF of GG, a contradiction. From this contradiction we deduce that dimΔ(G¯)=1\dim\Delta(\overline{G})=1 and as argued above, HH is either a 5-cycle or the graph in Figure 1. ∎

3 Cohen-Macaulay Matching Complexes

In this section we classify graphs which have a CM matching complex. We do this only when the graph is in one of the following classes of graphs: Cameron-Walker graphs and graphs with girth at least 5 (Subsection 3.1) and complete graphs (Subsection 3.2). Recall that here HH is a simple undirected graph with at least one edge, G=L(H)G=\mathrm{L}(H) is the line graph of HH and Δ=Δ(G¯)\Delta=\Delta(\overline{G}) is the matching complex of HH.

3.1 Cameron-Walker graphs and graphs with large girth

Recall that an induced matching of HH is a matching MM of HH such that for every pair of edges eeMe\neq e^{\prime}\in M, there is no edge ff of HH such that ff is incident to an endpoint of ee and an endpoint of ee^{\prime}. For studying Cohen-Macaulayness of matching complexes, we first restrict ourselves to the class of graphs, such as HH, in which the maximum size of induced matchings (denoted by im(H)\mathrm{im}(H)) and the maximum size of matchings (denoted by m(H)\mathrm{m}(H)) are equal. A classification of such graphs was first stated in [3]. Then in [7] a minor correction to this classification was presented.

We recall the corrected classification ([7, pp. 258, 259]). A star triangle is a graph obtained by joining several triangles at a common vertex and a pendant triangle is a triangle with two degree 2 vertices and a vertex with degree more than 2. Also a leaf edge is an edge incident to a leaf. Now for a connected graph HH, im(H)=m(H)\mathrm{im}(H)=\mathrm{m}(H) if and only if either HH is a star or a star triangle or is obtained from a connected bipartite graph with vertex partitions XX and YY, by adding some and at least one leaf edge to each vertex of XX and adding some and maybe zero pendant triangles to each vertex of YY. As in [7], we call a connected graph HH with im(H)=m(h)\mathrm{im}(H)=\mathrm{m}(h) and which is not a star or a star triangle, a Cameron-Walker graph. By a bipartition (X,Y)(X,Y) of a Cameron-Walker graph, we mean a bipartition of the underlying bipartite graph where XX and YY satisfy the aforementioned conditions. Figure 3 shows an example of a Cameron-Walker graph.

Refer to caption
Figure 3: An example of a Cameron-Walker graph

It is well-known that a CM simplicial complex is pure and the following shows that for graphs with im(H)=m(h)\mathrm{im}(H)=\mathrm{m}(h) the matching complex is pure.

Lemma 3.1.

If HH is a star or a star triangle or a Cameron-Walker graph, then Δ\Delta is pure.

Proof.

It is easy to see that each maximal matching of a star and a star triangle have size 1 and (n1)/2(n-1)/2, respectively, where nn denotes the number of vertices of the graph. Assume that HH is a Cameron-Walker graph with bipartition (X,Y)(X,Y). Let MM be a maximal matching of HH and M1M_{1} be those edges of MM which are in the underlying bipartite graph and M2=MM1M_{2}=M\setminus M_{1}. Suppose that X1X_{1} and Y1Y_{1} are vertices in XX and YY, respectively, incident to some edge in M1M_{1} and X2=XX1X_{2}=X\setminus X_{1}, Y2=YY1Y_{2}=Y\setminus Y_{1}. Then for each vertex xX2x\in X_{2} there should be exactly one leaf edge in M2M_{2} incident to xx and for xX1x\in X_{1}, none of the leaf edges at xx are in M2M_{2}. Also for each vertex yYy\in Y the pendant triangles at yy form a star triangle. If yY2y\in Y_{2}, then a maximal matching of the star triangle at yy must be in M2M_{2} and its size as mentioned before equals the number tyt_{y} of triangles in the star triangle at yy. If yY1y\in Y_{1}, then as yy is covered by M1M_{1}, only those edges of the star triangle can be in M2M_{2}, that do not meet yy and since MM is maximal, all of them must be in M2M_{2}. Therefore again the star triangle at yy contributes exactly tyt_{y} edges to M2M_{2}. Consequently, |M2|=|X2|+t|M_{2}|=|X_{2}|+t, where tt is the total number of pendant triangles in HH. Because |M1|=|X1||M_{1}|=|X_{1}|, it follows that |M|=|X|+t|M|=|X|+t is independent of the choice of MM. ∎

As a partial converse we have:

Lemma 3.2.

If HH is connected, Δ\Delta is pure and girth(H)5\mathrm{girth}(H)\geq 5, then either HH is a 5-cycle or a 7-cycle or a star or a Cameron-Walker graph.

Proof.

By [5, Theorem 1], HH is either a 5-cycle, a 7-cycle, an edge or a bipartite graph with bipartite sets V1V_{1} and V2V_{2} such that each vertex in V1V_{1} is adjacent to a leaf and each vertex in V2V_{2} is not adjacent to any leaf. Suppose that HH is not a star, a 5-cycle or a 7-cycle. If LL is the set of leaves of HH, then by setting X=V1X=V_{1} and Y=V2LY=V_{2}\setminus L, then HH is a Cameron-Walker graph with bipartition (X,Y)(X,Y). Note that the edges between XX and LL are the leaf edges at vertices of XX. ∎

Recall that for a face FF of a simplicial complex Γ\Gamma, we define linkΓ(F)={GF|FGΓ}\mathrm{link}_{\Gamma}(F)=\{G\setminus F|F\subseteq G\in\Gamma\}. Also for a vertex vv of Γ\Gamma, Γv\Gamma-v is the simplicial complex with faces {FΓ|vF}\{F\in\Gamma|v\notin F\}. A vertex vv of a nonempty simplicial complex Γ\Gamma is called a shedding vertex, when no face of linkΓ(v)\mathrm{link}_{\Gamma}(v) is a facet of Γv\Gamma-v. A nonempty simplicial complex Γ\Gamma is called vertex decomposable, when either it is a simplex or there is a shedding vertex vv such that both linkΓ(v)\mathrm{link}_{\Gamma}(v) and Γv\Gamma-v are vertex decomposable. One should note that in some papers vv is called a shedding vertex when it is shedding in the sense defined here and also linkΓ(v)\mathrm{link}_{\Gamma}(v) and Γv\Gamma-v are vertex decomposable. The (1)(-1)-dimensional simplicial complex {}\{\emptyset\} is considered a simplex and hence is vertex decomposable.

Theorem 3.3.

Assume that im(H)=m(H)\mathrm{im}(H)=\mathrm{m}(H) and G=L(H)G=\mathrm{L}(H). Then Δ(G¯)\Delta(\overline{G}) is vertex decomposable.

Proof.

By [7], HH is either a star or a star triangle or a Cameron-Walker graph. If HH is a star then GG is a complete graph and by [19, Lemma 4], Δ=Δ(G¯)\Delta=\Delta(\overline{G}) is vertex decomposable.

For the case that HH is a star triangle or a Cameron-Walker graph, we use induction on the number of edges of HH. Suppose that HH is a star triangle in which the common vertex of the triangles is vv. Let ee be an edge incident to vv. Then linkΔ(e)=Δ(L(HNG[e])¯)\mathrm{link}_{\Delta}(e)=\Delta(\overline{\mathrm{L}(H-\mathrm{N}_{G}[e])}) and HNG[e]H-\mathrm{N}_{G}[e] is just a matching. Therefore linkΔ(e)\mathrm{link}_{\Delta}(e) is a simplex. Also Δe=Δ(L(He)¯)\Delta-e=\Delta(\overline{\mathrm{L}(H-e)}) and HeH-e is either a star or a Cameron-Walker graph with less edges that HH and hence Δe\Delta-e is vertex decomposable by the induction hypothesis. Also if MM is a maximal matching of HH containing ee, then (M{e}){e}(M\setminus\{e\})\cup\{e^{\prime}\} is a matching of HeH-e, where ee^{\prime} is the edge in the same triangle as ee but not adjacent to vv. This shows that ee is a shedding vertex of Δ\Delta and thus Δ\Delta is vertex decomposable.

Now assume that HH is a Cameron-Walker graph with bipartition (X,Y)(X,Y). Let ee be an edge of the underlying bipartite graph with an endpoint xXx\in X. By the definition of Cameron-Walker graphs, xx is incident to a leaf edge ee^{\prime}. Now every connected component of HeH-e and HNG[e]H-N_{G}[e] is either a star or a star triangle or a Cameron-Walker graph with less edges than HH. Consequently, both Δe\Delta-e and linkΔ(e)\mathrm{link}_{\Delta}(e) are vertex decomposable by induction hypothesis. Also if MM is a maximal matching of HH containing ee, then (M{e}){e}(M\setminus\{e\})\cup\{e^{\prime}\} is a matching of HeH-e, that is, ee is a shedding vertex of Δ\Delta. Thus Δ\Delta is vertex decomposable. ∎

A combinatorial property related to being CM is shellability. If there is an ordering F1,,FtF_{1},\ldots,F_{t} of all facets of Γ\Gamma such that for all ii we have F1,,FiFi+1\langle F_{1},\ldots,F_{i}\rangle\cap F_{i+1} is a pure simplicial complex of dimension =dimFi+11=\dim F_{i+1}-1, Γ\Gamma is called shellable. It is well-known that a vertex decomposable complex is shellable and a pure shellable complex is CM.

Corollary 3.4.

Suppose that HH is a connected graph with girth(H)5\mathrm{girth}(H)\geq 5, G=L(H)G=\mathrm{L}(H) and Δ=Δ(G¯)\Delta=\Delta(\overline{G}) be the matching complex of HH. Then the following are equivalent.

  1. (i)

    Δ\Delta is pure vertex decomposable.

  2. (ii)

    Δ\Delta is pure shellable.

  3. (iii)

    Δ\Delta or equivalently GG is CM (over some field).

  4. (iv)

    HH is either a 5-cycle or a star or a Cameron-Walker graph.

Proof.

It is well-known that (i) \Rightarrow (ii) \Rightarrow (iii). Suppose that Δ\Delta is CM over a field KK. Then Δ\Delta is pure and hence according to (3.2), HH is either a 5-cycle or a 7-cycle or a star or a Cameron-Walker graph. But by [19, Theorem 10] a 7-cycle is not CM over any field, so (iv) follows. Now assume that HH is a star or a Cameron-Walker graph. Then by (3.1) and (3.3), Δ\Delta is pure vertex decomposable. For the 5-cycle, the result follows from [19, Theorem 10]. ∎

3.2 Complete graphs

At the end of this paper, we consider the matching complex of the complete graph KnK_{n} on nn vertices. Recall that a CM simplicial complex Γ\Gamma is strongly connected (or connected in codimension 1), that is, for each two facets A,BA,B of Γ\Gamma, there is a sequence of facets A=F1,F2,,Ft=BA=F_{1},F_{2},\ldots,F_{t}=B, such that |FiFi+1|=|Fi|1|F_{i}\cap F_{i+1}|=|F_{i}|-1. We call such a sequence of facets a strong path from AA to BB. First we find complete graphs with a strongly connected matching complex.

Proposition 3.5.

Suppose that G=L(Kn)G=\mathrm{L}(K_{n}) and Δ=Δ(G¯)\Delta=\Delta(\overline{G}). Then Δ\Delta is strongly connected, if and only if nn is odd or n=2n=2.

Proof.

Suppose that nn is even. Then each maximal matching of KnK_{n} has size n/2n/2. If MM and MM^{\prime} are maximal matchings of KnK_{n} and |MM|=|M|1|M\cap M^{\prime}|=|M|-1, then MMM\cap M^{\prime} saturates n2n-2 vertices of KnK_{n}. If uu and vv are the only unsaturated vertices, then the only maximal matching containing MMM\cap M^{\prime} is (MM){uv}(M\cap M^{\prime})\cup\{uv\}, so M=MM=M^{\prime}. Consequently, if KnK_{n} has more than one maximal matchings, which holds if n>2n>2, then Δ\Delta is not strongly connected.

Now assume that nn is odd. We use induction to show that there is a strong path between every two maximal matchings of KnK_{n}. Let M1M_{1} and M2M_{2} be maximal matchings of KnK_{n}. If uvM1M2uv\in M_{1}\cap M_{2} then there is a strong path between M1{uv}M_{1}\setminus\{uv\} and M2{uv}M_{2}\setminus\{uv\} in KnuvKn2K_{n}-u-v\cong K_{n-2}, by the induction hypothesis. Hence there is a strong path between M1M_{1} and M2M_{2}. Suppose that M1M2=M_{1}\cap M_{2}=\emptyset and aa is a vertex saturated by both M1M_{1} and M2M_{2}, say abM1ab\in M_{1} and acM2ac\in M_{2}. If cc is not saturated by M1M_{1}, then by the previous part there is a strong path between (M1{ab}){ac}(M_{1}\setminus\{ab\})\cup\{ac\} and M2M_{2}. Thus it follows that there is such a path between M1M_{1} and M2M_{2}. If cc is saturated by M1M_{1}, say cdM1cd\in M_{1}, then there is a strong path PP from (M1{ab,cd}){ac,bd}(M_{1}\setminus\{ab,cd\})\cup\{ac,bd\} to M2M_{2}. Note that since nn is odd, there is an vertex ee unsaturated in M1M_{1}. If we attach the following strong path at the start of PP we get a strong path from M1M_{1} to M2M_{2}: M1M_{1}, (M1{ab}){ae}(M_{1}\setminus\{ab\})\cup\{ae\}, (M1{ab,cd}){ae,bd}(M_{1}\setminus\{ab,cd\})\cup\{ae,bd\}. ∎

Recall that a perfect matching of HH is a matching that saturates all vertices of HH.

Remark 3.6.

The first part of proof of (3.5), indeed shows that if HH is a graph with a perfect matching, then its matching complex cannot be strongly connected and hence is not CM.

To find out when Δ(L(Kn)¯)\Delta(\overline{\mathrm{L}(K_{n})}) is CM, we need an algebraic tool. Let Γ\Gamma be a simplicial complex and denote by C~d(Γ)=C~d(Γ;K)\widetilde{C}_{d}(\Gamma)=\widetilde{C}_{d}(\Gamma;K) the free KK-module whose basis is the set of all dd-dimensional faces of Γ\Gamma. Consider the KK-map d:C~d(Γ)C~d1(Γ)\partial_{d}:\widetilde{C}_{d}(\Gamma)\to\widetilde{C}_{d-1}(\Gamma) defined by

d({v0,,vd})=i=0d(1)i{v0,,vi1,vi+1,,vd},\partial_{d}(\{v_{0},\ldots,v_{d}\})=\sum_{i=0}^{d}(-1)^{i}\{v_{0},\ldots,v_{i-1},v_{i+1},\ldots,v_{d}\},

where v0<<vdv_{0}<\cdots<v_{d} is a linear order. Then (C~,)(\widetilde{C}_{\bullet},\partial_{\bullet}) is a complex of free KK-modules and KK-homomorphisms called the augmented oriented chain complex of Γ\Gamma over KK. We denote the ii-th homology of this complex by H~i(Γ;K)\widetilde{H}_{i}(\Gamma;K). The Reisner theorem states that Γ\Gamma is CM over KK, if and only if for all faces FF of Γ\Gamma including the empty face and for all i<dimlinkΓ(F)i<\dim\mathrm{link}_{\Gamma}(F), one has H~i(linkΓ(F);K)=0\widetilde{H}_{i}(\mathrm{link}_{\Gamma}(F);K)=0 (see [6, Theorem 8.1.6]). Equivalently, this theorem states that Γ\Gamma is CM over KK if and only if for all vertices vv of Γ\Gamma, linkΓ(v)\mathrm{link}_{\Gamma}(v) is CM and H~i(Γ;K)=0\widetilde{H}_{i}(\Gamma;K)=0 for all i<dimΓi<\dim\Gamma.

Theorem 3.7.

Suppose that G=L(Kn)G=\mathrm{L}(K_{n}) and Δ=Δ(G¯)\Delta=\Delta(\overline{G}) is the matching complex of KnK_{n}. Then Δ\Delta is CM over a field KK, if and only if either

  1. (i)

    n3n\leq 3 or n=5n=5;

  2. (ii)

    or n=7n=7 and charK3\mathrm{char}K\neq 3.

Proof.

If n3n\leq 3, then dimΔ0\dim\Delta\leq 0 and Δ\Delta is CM. If n>2n>2 is even, then by (3.5) Δ\Delta is not strongly connected and hence not CM. If n=5n=5, then dimΔ=1\dim\Delta=1 and Δ\Delta is CM if and only if it is connected or equivalently strongly connected (see [2, Excercise 5.1.26]). Hence again the result follows from (3.5). Suppose that n7n\geq 7 is odd. Assume that ee is a vertex of Δ\Delta, that is, e=uve=uv, for vertices uu and vv of KnK_{n}. Then linkΔ(e)\mathrm{link}_{\Delta}(e) is the matching complex of KnuvKn2K_{n}-u-v\cong K_{n-2}. Thus if we show that for n=9n=9, Δ\Delta is not CM over any field and for n=7n=7, Δ\Delta is CM exactly when KK has characteristic 3\neq 3, then the proof is concluded. But letting n=7n=7, for all vertices ee of Δ\Delta, linkΔ(e)\mathrm{link}_{\Delta}(e) is CM. Hence Δ\Delta is CM if and only if H~i(Δ;K)=0\widetilde{H}_{i}(\Delta;K)=0 for all i<dimΔ=2i<\dim\Delta=2. By Table 6.1 of [16], we have H~1(Δ;)=3\widetilde{H}_{1}(\Delta;\mathbb{Z})=\mathbb{Z}_{3} and H~0(Δ;)=0\widetilde{H}_{0}(\Delta;\mathbb{Z})=0 and hence by the universal coefficient theorem, H~1(Δ;K)=3K\widetilde{H}_{1}(\Delta;K)=\mathbb{Z}_{3}\otimes K is zero if and only if charK3\mathrm{char}K\neq 3. Similar argument using Table 6.1 of [16] shows that if n=9n=9, then H~2(Δ,K)0\widetilde{H}_{2}(\Delta,K)\neq 0 for every field KK and Δ\Delta is not CM over any field. ∎

Acknowledgement

The author would like to thank M. Bayer and B. Goeckner for mentioning some interesting references that the author was not aware of and contained answers to some questions posed in a previous version of this paper.

References

  • [1] M. Bayer, B. Goeckner and M. J. MilutinoviĆ, Manifold matching complexes, Mathematika 66 (2020), 973–1002.
  • [2] W. Bruns and Jürgen Herzog, Cohen-Macaulay Rings, Cambridge University Press, Cambridge, 1993.
  • [3] K. Cameron and T. Walker, The graphs with maximum induced matching and maximum matching the same size, Discrete Math. 299 (2005), 49–55.
  • [4] M. Crupi, G. Rinaldo and N. Terai, Cohen-Macaulay edge ideals whose height is half of the number of vertices, Nagoya Math. J. 201 (2011), 116–130.
  • [5] A. Frendrup, B. Hartnell and P. D. Vestergaard, A note on equimatchable graphs Australasian J. Combin., 46 (2010), 185–190.
  • [6] J. Herzog and T. Hibi, Monomial Ideals, Springer-Verlag, London , 2011.
  • [7] T. Hibi, A. Higashitani, K. Kimura and A. B. O’Keefe, Algebraic study on Cameron-Walker graphs, J. Algebra, 422 (2015), 257–269.
  • [8] D. T. Hoang and T. N. Trung, A characterization of triangle-free Gorenstein graphs and Cohen-Macaulayness of second powers of edge ideals, J. Algebr. Combin. 43 (2016), 325–338.
  • [9] J. Jonsson, Exact sequences for the homology of the matching complex, J. Combin. Theory, Ser. A 115 (2008), 1504–1526.
  • [10] P. G. H. Lehot, An optimal algorithm to detect a line graph and output its root graph, J. ACM, 21 (1974), 569–-575.
  • [11] A. Nikseresht, Chordality of clutters with vertex decomposable dual and ascent of clutters, J. Combin. Theory —Ser. A, 168 (2019), 318–337, arXiv: 1708.07372.
  • [12] A. Nikseresht and M. R. Oboudi, Trung’s construction and the Charney-Davis conjecture, Bull. Malays. Math. Sci. Soc., 44 (2021), 9–16, arXiv: 1906.11482.
  • [13] M. R. Oboudi and A. Nikseresht, Some combinatorial characterizations of Gorenstein graphs with independence number less that four, Iran. J. Sci. Technol. Trans. Sci., 44 (2020), 1667–1671, arXiv: 1911.11406.
  • [14] R. P. Stanley, Combinatorics and Commutative Algebra, second ed., Birkhäuser, Boston, 1996.
  • [15] T. N. Trung, A characterization of Gorenstein planar graphs, in T. Hibi, ed., Adv. Stud. Pure Math., Vol. 77 (2018): The 50th Anniversary of Gröbner Bases, 399–409, arXiv:1603.00326v2.
  • [16] M. L. Wachs, Topology of matching, chessboard, and general bounded degree graph complexes, Algebra Univers., 49 (2003), 345–385.
  • [17] D. B. West, Introduction to Graph Theory, 2nd ed., Prentice Hall, 2001.
  • [18] R. Woodroofe, Chordal and sequentially Cohen-Macaulay clutters, Elec. J. Combin., 18 (2011), research paper p208.
  • [19] R. Woodroofe, Vertex decomposable graphs and obstructions to shellability, Proc. Amer. Math. Soc., 137 (2009), 3235–3246.