F-Purity of Binomial Edge Ideals

The graphs appearing in section (1) and (2) of Theorem 5.5 (the “sporadic graphs" and “regular families", respectively) are not F-pure in any positive characteristic, since each of these graphs contain an asteroidal triple (Theorem 4.11, Corollary 4.13).

Theorems 9.7, 9.18, 9.20, and 9.36 establish that the graphs appearing in section (3) of Theorem 5.5 (the “co-regular families") are not F-pure in characteristic 2. Matsuda’s conjecture A now follows from Proposition 5.6.

By Fedder’s criterion (Theorem 2.2), we have that R/J_{G} is F-pure if and only if

Since the map \Bbbk \hookrightarrow \mathbb{K} is faithfully flat and S \cong R \otimes_{\Bbbk} \mathbb{K}, it follows that equation (Proof) is equivalent to

By applying Fedder’s criterion (Theorem 2.2), equation (Proof) is equivalent to S/J_{G}S being F-pure. The result now follows by observing that J_{G}S is precisely the binomial edge ideal of G defined over the ring S.

Since \Bbbk = \mathbb{Z}/p\mathbb{Z} is F-finite, the property of R_{G} or R_{H} being F-pure is equivalent to R_{G} or R_{H} being F-split. Thus, we may suppose that R_{G} is F-split, and it suffices to show that R_{H} is F-split. Since H is an induced subgraph of G, we have that R_{H} is a \Bbbk-algebra retract of R_{G}; in other words, we have that R_{H} \subseteq R_{G} as \Bbbk-algebras and that there is a map \varepsilon : R_{G} \rightarrow R_{H} which is the identity on R_{H}. Because R_{G} is F-split, there is a map \phi : F_{\ast} R_{G} \rightarrow R_{G} which splits the Frobenius map F : R_{G} \rightarrow F_{\ast} R_{G}. Thus we have the following commutative diagram.

We claim that the composition of maps

splits the map F : R_{H} \rightarrow F_{\ast} R_{H}. Indeed, we have that for r \in R_{H} that

We note that each of the maps F_{\ast}\iota, \phi, and \varepsilon are R_{H}-linear, and hence their composition is also R_{H}-linear.

First, we show that R_{G\smallsetminus v} is F-pure. Lemma 3.2 implies that G \smallsetminus v defines an F-pure binomial edge ideal since G \smallsetminus v is an induced subgraph of G.

Next, we show that R_{G_{v}} is F-pure. By Fedder’s criterion (Theorem 2.2), it suffices to show that

We will show that

Since R_{G} is F-pure, we have that

Hence, equation (?) will imply equation (Proof), which will complete the proof.

Let r \in J_{G}^{[p]}:J_{G}, and let \{i,j\} \in E(G_{v}). If \{i,j\} \in E(G), then we have that

Suppose that \{i,j\} \in E(G_{v}) \smallsetminus E(G). Then, i,j \in N_{G}(v). By Plücker’s relations (see e.g., [bruns2022determinants]), we have that x_{v} \cdot f_{ij} = \pm x_{i} \cdot f_{vj} \mp x_{j} \cdot f_{vi}. Because \{v,j\} and \{v,i\} are edges of G, it follows from the previous case that

Observe that

by [peskine1973dimension] and [bolognini2022cohen, Lemma 4.5]. Proposition 2.4 implies that x_{v} is a non-zerodivisor on R/J_{G_{v}}^{[p]}. Consequently, r \cdot f_{ij} \in J_{G_{v}}^{[p]}. This proves that r \in J_{G_{v}}^{[p]}:J_{G_{v}}, and thus establishes equation (?), which completes the proof.

Finally, we observe that the above two cases imply that G_{v} \smallsetminus v defines an F-pure binomial edge ideal.

The proof is an exercise in elementary number theory. A write-up can be found in [binomialIdentity].

The result can be verified by hand when p = 2. Hence, we may suppose that p > 2 is an odd positive prime number. By the Pigeon hole principle, it is clear that every non-zero term of (f_{a,b}f_{b,c})^{p-1} is divisible by x_b^{p}, y_b^{p}, or x_b^{p-1}y_b^{p-1}. The term of (f_{a,b}f_{b,c})^{p-1} divisible by x_b^{p-1}y_b^{p-1} is

which by Lemma 4.2 is equal to

By the binomial theorem, the polynomial f_{a,c}^{p-1} can be expanded as

As p is in particular an odd number, Lemma 4.2 implies that expression (Proof) is equal to

The result now follows by substituting f_{a,c}^{p-1} into expression (Proof).

By Lemma 4.3, we have that

Every term of (f_{1,3})^{2p-2} belongs to \mathfrak{m}^{[p]} except for the term (x_1y_3x_3y_1)^{p-1}. The coefficient on this term is

By Theorem 4.1, we have that

Let r \in \mathfrak{p}^{[p]}:(f_{i,j},f_{j,k}) in which case rf_{i,j} and rf_{j,k} belong to \mathfrak{p}^{[p]}. From the following relation

we see that

Since \sqrt{\mathfrak{p}^{[p]}} = \mathfrak{p} does not contain x_j, it follows that rf_{i,k} \in \mathfrak{p}^{[p]}. Thus, r \in \mathfrak{p}^{[p]}:(f_{i,k}) as claimed.

Since G[A] is connected by assumption, there is a path i = v_0,v_1,\ldots,v_r = j in G[A] with r \geq 1. Thus f_{v_{k-1},v_k} \in J_G for 1 \le k \le r. Therefore, P(S,A)^{[p]}:J_G \subseteq P(S,A)^{[p]}:(f_{v_0,v_1},f_{v_1,v_2},\ldots,f_{v_{r-1},v_r}).

First we show that P(S,A)^{[p]}:J_G \subseteq P(S,A)^{[p]}:(f_{i,j}). We observe that

where the first containment follows from Lemma 4.5. By repeating this argument it follows that P(S,A)^{[p]}:J_G \subseteq P(S,A)^{[p]}:(f_{i,j}).

Next, we observe that for g \in P(S,A)^{[p]}:(f_{i,j})

The conclusion now follows from the fact that \{x_l,y_l \mid l \in S \cup A - \{i,j\}\} \cup \{f_{i,j}\} form a regular sequence.

First, note that the subring S = \Bbbk[f_1,\ldots,f_t] is isomorphic to a polynomial ring in t variables. In S, the intersection amounts to a statement about the intersection of monomial ideals, which is well-known; see e.g. [MRSW2018, Proposition 2.1.5]. The conclusion follows since the extension S \to R is flat.

By Fedder’s Criterion Theorem (2.2), it suffices to show that J_G^{[p]}:J_G \subseteq \mathfrak{m}^{[p]}. First, we observe that the hypothesis that N(v_{i}) is a minimal vertex isolator of G implies that

is a minimal prime of J_{G}. Lemma 4.7 implies that

for distinct i,j,k \in \{1,2,3\}. Write \eta = (x_l,y_l \mid l \in (V(G) \smallsetminus \{v_i,v_j,v_k\})). Then, we compute that

We have that: step (Proof) follows from J_G^{[p]} \subseteq \bigcap_{l = 1}^3 \mathfrak{p}_i; step (Proof) follows from the definition of Frobenius power; step (Proof) follows from the definition of colon ideals; step (Proof) follows from Lemma 4.7; step (Proof) follows from Proposition 4.8 since \{x_l,y_l \mid l \in (V(G) - \{v_i,v_j,v_k\})\} \cup \{f_{v_i,v_j},f_{v_i,v_k}\} forms a regular sequence; and step (Proof) also follows from Proposition 4.8 since \{x_l,y_l \mid l \in (V(G) - \{v_i,v_j,v_k\})\} \cup \{f_{v_i,v_j} \cdot f_{v_i,v_k},f_{v_j,v_k}\} is a regular sequence. The final step (Proof) follows from Lemma 4.4.

We prove the contrapositive. Let G be a graph containing an asteroidal triple. Theorem 4.11 implies that G contains one of the listed subgraph as an induced subgraph. It is readily verified that every graph appearing in the list in Theorem 4.11 satisfies the hypotheses of Theorem 4.9. Consequently, the binomial edge ideal of every graph listed in Theorem 4.11 is not F-pure in any positive characteristic. Lemma 3.2 implies that S/J_{G} is not F-pure in any positive characteristic.

Mastuda’s Theorem 5.3 implies that every weakly closed graph defines an F-pure ring in any characteristic, and hence, in particular, in characteristic 2. To prove the converse of Conjecture A, it suffices to show that if a graph G is not weakly closed, then R_{G} is not an F-pure ring in characteristic 2. By Theorems 5.4 and 5.5, we have that if G is not weakly closed, then G contains as an induced subgraph H where H belongs to the list of graphs in Theorem 5.5. By Lemma 3.2, if R_{H} is not an F-pure ring in characteristic 2, then R_{G} cannot be an F-pure ring in characteristic 2 either.

\mathit{(1) \Rightarrow (2)} is obvious. \mathit{(2) \Rightarrow (3)} follows from Corollary 4.13. \mathit{(3) \Rightarrow (4)} follows from Gallai’s Theorem 5.5 and Matsuda’s Theorem 5.4, since chordal graphs are AT-free if and only if they are co-comparability [isgciATFreeANDChordal], while \mathit{(4) \Rightarrow (1)} follows from Matsuda’s Theorem 5.3.

Since the element \alpha \cdot r is not divisible by x_{k}^{2} or by y_{k}^{2}, it follows that

This still holds after evaluating x_{k} and y_{k} at one. When \alpha \in \{x_{k},y_{k},x_{k}y_{k}\}, the lemma follows. When \alpha = f_{k,b}, we obtain that

Since x_{b} + y_{b} is a regular element on P(A\smallsetminus \{k\},B)^{[2]}, it follows that r \in P(A\smallsetminus \{k\},B)^{[2]}, which completes the proof.

We first prove the reverse inclusion. It suffices to prove for every edge e of G that f_{e} \, \omega_{A} \in P^{[2]}. Since e is of the form \{k,j\} for some vertex k \in A and j \in V(G) \smallsetminus \{k\}, it follows that f_{k,j} \omega_{A} \in (x_{k}^{2},y_{k}^{2}) \subseteq P^{[2]}.

We prove the forward inclusion by induction on \#A. If \#A = 0, then \omega_{A} = 1 and J_{G} = (0) by condition (4), in which case equation (6.3) holds. Hence, we may suppose that \#A > 0, and that equation (6.3) holds for all graphs G' satisfying conditions (1), (2), (3), and (4) and for all prime ideals P(A',B') as in the statement of the lemma whenever \# A' < \# A.

Pick r \in P^{[2]}:J_{G}. After subtracting from r those monomials belonging to the support of r and contained in the ideal ( \{x_{j}^{2},y_{j}^{2} \mid j \in A\}), we may suppose that every monomial belonging to the support of r is not contained in P^{[2]}. Next, pick k \in A. Since every monomial in the support of r is not divisible by x_{k}^{2} or y_{k}^{2}, we may express r as:

for some r_{1}, r_{2}, r_{3}, r_{4} \in \Bbbk[x_{i},y_{i} \mid i \in [n] \smallsetminus \{k\}]. The ideals J_{G} and P^{[2]} are both \mathbb{N}^{n}-graded ideals, where n = \# V(G) and \deg(x_{i}) = \deg(y_{i}) = e_{i} is the i-th standard basis element. Hence, the ideal P^{[2]}:J_{G} is also \mathbb{N}^{n}-graded. Thus, for s \in \{x_{k}y_{k} r_{1}, x_{k}r_{2} + y_{k} r_{3}, r_{4} \}, we have that

It now suffices to show that each element of \{x_{k}y_{k} r_{1}, x_{k}r_{2} + y_{k} r_{3}, r_{4} \} belongs to P^{[2]} + (\omega_{A}).

First, we show that x_{k}y_{k} r_{1} \in P^{[2]} + (\omega_{A}). By applying Lemma 6.2 to x_{k}y_{k}r_{1}\cdot f_{i,j} for every edge \{i,j\} \in E(G \smallsetminus \{k\}), it follows that

Since G \smallsetminus k, A' := A\smallsetminus k, and B' := B satisfy the hypotheses of Lemma 6.3, it follows by induction hypothesis that

Thus, x_{k}y_{k}r_{1} \in P^{[2]} + (\omega_{A}).

Next, we show that r_{4} \in P^{[2]} + (\omega_{A}). We observe that

By Lemma 6.2, it follows that r_{4} \in P^{[2]}.

Finally, we show that x_{k}r_{2} + y_{k} r_{3} \in P^{[2]} + (\omega_{A}). We consider the element

After expanding this product and subtracting terms divisible by x_{k}^{2} or by y_{k}^{2}, we obtain that

Lemma 6.2 implies that

Write

where every monomial in the support of r_{2}'' (respectively r_{3}'') is not divisible by y_{b} (respectively x_{b}). Hence, we have that

In (?), setting x_{b} = 0 implies that

Since y_{b} is a regular element on P(A\smallsetminus \{k\},B)^{[2]}, it follows that

Likewise, setting y_{b} = 0 in (?) and using that x_{b} is a regular element on P(A\smallsetminus \{k\},B)^{[2]}, it follows that

Combining (?), (?), and (?) together with the fact that x_{b}y_{b} is a regular element on P(A\smallsetminus \{k\},B)^{[2]} implies

From (?), (?), (?), and (?) we have that after subtracting terms belonging to P(A\smallsetminus \{k\},B)^{[2]} that we may assume that

Without loss of generality, we may assume by condition (3) that \{a,k\} \in E(G). It follows that

Since f_{a,b} is a regular element on P^{[2]}, it follows that

Lemma 6.2 implies that r_{2}' \in P(A \smallsetminus \{k\}, B)^{[2]} \subseteq P^{[2]}. It thus follows that r_{2} and r_{3} belong to P^{[2]}, which completes the proof.

We prove the forward inclusion, as the reverse inclusion is clear. Let r \in P^{[2]} + (\omega_{A}). Then we may express r as r = r_{1} + p \, \omega_{A}, with p \in \Bbbk[x_{j},y_{j} \mid j \in [n] \smallsetminus A] and r_{1} \in P^{[2]}. Since r \in P^{[2]} + Q, it follows that p \, \omega_{A} \in P^{[2]} + Q. For degree reasons, it follows that

Since \omega_{A} is a regular element on P(\varnothing,B)^{[2]} + Q, it follows that p \in P^{[2]} + Q, which completes the proof.

We prove the forward inclusion, as the reverse inclusion is clear. Let r \in P^{[2]}:J_{G}. Subtract from r all monomials belonging to P(A,\varnothing)^{[2]}. Since V(G) \subseteq B, it follows for degree reasons that for all \{i,j\} \in E(G) that r\cdot f_{i,j} \in P(\varnothing,B)^{[2]}, which completes the proof.

The proof of equation (6.6) is clear. The proof of equation (6.6) can be verified via the computer program Macaulay2.

We will prove the case that B = \{a,c,d\} since the proof of the case where B = \{a,c\} is similar. Consider the following subgraphs of G: the graph H_{1} obtained as the induced subgraph of G on the vertices \{a,c,d\}; and H_{2}, the graph obtained from G after deleting the edges E(H_{1}). Since the edges of H_{1} and H_{2} cover the edges of G, it follows from elementary properties of colon ideals that

Lemmas 6.5 and 6.6 imply that

Lemma 6.3 implies that

The claim now follows from Lemma 6.4.

Since r \in P(A,B)^{[2]} + (f_{a,c} \, \omega_{A}), we may express r as:

for some \alpha_{j},\beta_{j},\gamma,\delta \in R and all j \in A. The monomial w := \frac{x_{a}}{y_{a}} \cdot \frac{p}{q} \cdot m is not divisible by x_{j}^{2} or by y_{j}^{2} for all j \in A \cup \{c\}, since p is not divisible by these variables. It follows that

Since w is not divisible by y_{a}, it follows that there exists \delta' \in \operatorname{Supp}(\delta) such that

Consequently, \delta' = \frac{1}{y_{a}y_{c}} \cdot \frac{p}{q} \cdot \frac{m}{\omega_{A}}. The assumption that q is not divisible by x_{j} or by y_{j} for all j \in A \cup \{a,c\} implies that \delta' is a bona fide monomial. We compute that

The monomial w' := \frac{x_{c}}{y_{c}} \cdot \frac{p}{q} \cdot m appearing in equation (Proof) is not divisible by x_{j}^{2} or by y_{j}^{2} for j \in A \cup \{a\}, since p is not divisible by these variables. Consequently, the computation in Equation (Proof) is the unique way to obtain w' as a term appearing in the expansion in Equation (Proof). Thus, the coefficient of w in r is equal to the coefficient of w' in r. Thus, w \in \operatorname{Supp}(r) if and only if w' \in \operatorname{Supp}(r).

Since r \in P^{[2]} + L_{P}, we have that

where \gamma \in P(A,\varnothing)^{[2]} + (f_{a,d}^{2},f_{c,d}^{2}) and \mu, \mu', \mu'', and \mu''' are polynomials in R. We observe that m \in \operatorname{Supp}(r) implies

After expanding out the binomials in equation (Proof) and comparing terms, we can determine the monomials belonging to \mu', \mu'', and \mu''' for which equation (Proof) is satisfied. We find that for \tilde{\mu} \in \operatorname{Supp}(\mu')

if and only if \tilde{\mu} \in \{\mu_{2},\mu_{3}\}, where

Similarly, for \tilde{\mu} \in \operatorname{Supp}(\mu''), we have that

if and only if \tilde{\mu} \in \{\mu_{4},\mu_{5}\} where

Finally, for \tilde{\mu} \in \operatorname{Supp}(\mu'''), we have that

if and only if \tilde{\mu} \in \{\mu_{6},\mu_{7}\}, where

Next, we compute that

and that

The polynomials \mu_{i} \cdot f_{a,c}f_{c,d} \, \omega_{A} and \mu_{j} \cdot f_{a,d}f_{c,d} \, \omega_{A} can be computed for i = 4,5 and j = 6,7 by an appropriate change of variables applied to equations (Proof) and (Proof).

A straightforward computation shows that if

for \tilde{\mu} a monomial and i,j,k \in \{a,c,d\} distinct, then \tilde{\mu} \in \{\mu_{3},\mu_{4},\mu_{6}\}. Likewise, if

for \tilde{\mu} a monomial and i,j,k \in \{a,c,d\} distinct, then \tilde{\mu} \in \{\mu_{2},\mu_{5},\mu_{7}\}.

For \nu = \alpha or \nu = \beta, we have that \nu \in \operatorname{Supp}(r) if and only if

We have characterized in the preceding paragraph necessary and sufficient conditions on \mu', \mu'', and \mu''' for which

A similar computation shows that

if and only if

Define the sets

It follows from the above observations that the coefficient on m, \alpha, and \beta in r is \# U_{m}, \# U_{\alpha}, and \# U_{\beta}, respectively. Since U_{m} = (U_{\alpha} \smallsetminus \{1\}) \sqcup (U_{\beta} \smallsetminus \{1\}) and \Bbbk = \mathbb{Z}/2\mathbb{Z}, it follows that \#U_{m} + \#U_{\alpha} + \# U_{\beta} \equiv 0 \mod{2}, which completes the proof.

For i \in \{a,c,d\}, we denote by \alpha_{i} the monomial

After relabeling the variables as necessary, we may suppose without loss of generality that

It follows that

where \alpha \in P^{[2]} and \mu, \mu', and \mu'' are polynomials in R. The constraints on p imply that

By expanding the binomials f_{a,c}f_{a,d} and comparing terms, we see that for \tilde{\mu} \in \operatorname{Supp}(\mu) that

if and only if

Similarly, for \tilde{\mu} \in \operatorname{Supp}(\mu') we have that

if and only if

Finally, for \tilde{\mu} \in \operatorname{Supp}(\mu'') we have that

if and only if

We compute that

The change of variables induced by swapping a and c applied to equation (Proof) computes \mu_{2} \cdot f_{a,c}f_{c,d}\, \omega_{A}. Likewise, the change of variables induced by swapping a and d applied to equation (Proof) computes \mu_{3} \cdot f_{a,d}f_{c,d}\, \omega_{A}. It now follows that \alpha_{i} belongs to the support of \mu_{2} \cdot f_{a,c}f_{a,d}\, \omega_{A} and \mu_{3} \cdot f_{a,d}f_{c,d}\, \omega_{A} for all i \in \{a,c,d\}.

A similar argument shows that for i \in \{c,d\} that

It follows that the coefficient on \alpha_{i} in r is the same for all i \in \{a,c,d\}, which completes the proof.

We denote by r' the monomial

Since r \in P^{[2]} + L_{P}, we have that

where \alpha \in P^{[2]} and \beta \in L_{P}. The constraints on p imply that r' \notin \operatorname{Supp}(\alpha).

Next, we observe that every monomial of f_{a,c} is divisible by y_{a} or by y_{c}. The constraints on p imply that r' is not divisible by y_{a} nor by y_{c}. Hence, it follows that r' \notin \operatorname{Supp}(f_{a,c}f_{a,d}) and that r' \notin \operatorname{Supp}(f_{a,c}f_{c,d}).

Lastly, we observe that every monomial of f_{a,d}f_{c,d} is divisible by y_{a}, by y_{c}, or by x_{a}x_{c}y_{d}^{2}. Since r' is not divisible by y_{a} nor by y_{c} nor by y_{d}^{2}, it follows that r' \notin \operatorname{Supp}(f_{a,d}f_{c,d}).

This proves that r' \notin \operatorname{Supp}(\beta). Hence, in particular, r' \notin \operatorname{Supp}(r).

We will establish the statement for T_{1} as the statement for T_{2} is proven analogously. First, we observe that for all r' \in T_{1} that r' is not divisible by x_{j}^{2} or y_{j}^{2} for all j \in A, and r' is not divisible by x_{i}^{2}y_{j}^{2} for all i,j \in \{a,c,d\} distinct. Consequently,

Second, we observe that for r' \in T_{1}, r' \notin \operatorname{Supp}(\mu \cdot f_{a,c}f_{a,d} \, \omega_{A}), because every monomial of f_{a,c}f_{a,d} is divisible by x_{a}^{2}, x_{a}y_{a}, or y_{a}^{2}, whereas r' is not divisible by these monomials. Consequently,

Next, we observe that if

then

for \tilde{\mu} \in \operatorname{Supp}(\mu') if and only if

Equivalently,

It can be verified that for every r' \in T_{1},

for \tilde{\mu} \in \operatorname{Supp}(\mu') if and only if \tilde{\mu} = \mu_{1}. Similarly, it can be verified that

for \tilde{\mu} \in \operatorname{Supp}(\mu'') if and only if

We compute that

Finally, we compute that

Since the coefficient field is \mathbb{Z}/2\mathbb{Z}, it follows for parity reasons, if r' \in \operatorname{Supp}(r) for some r' \in T_{1}, then either \mu_{1} \in \operatorname{Supp}(\mu') or \mu_{2} \in \operatorname{Supp}(\mu''). In which case, equations (Proof) and (Proof) imply the claim.

Expanding r in terms of the generators of P_{a,c}^{[2]} + (f_{a,c}\, \omega_{A}), we have that

for polynomials \alpha_{j},\beta_{j},\gamma,\mu. The constraints on r' imply that r' \in \operatorname{Supp}(\mu \cdot f_{a,c} \, \omega_{A}).

It follows that

We observe that every monomial belonging to the support of f_{a,c}f_{a,d} is divisible by x_{k} or y_{k} for each k \in B. Similarly, this holds for f_{a,c}f_{c,d} and f_{a,d}f_{c,d}. This proves the first claim. The second claim follows immediately from equation (Proof).

For the third claim, let us suppose that y_{d}^{2} \nmid r'. The case in which x_{d}^{2} \nmid r' is proved similarly. We have that equation (Proof) holds. We observe that every monomial belonging to the support of f_{a,c} is divisible by y_{a} or y_{c}, and that every monomial belonging to the support of f_{a,d}f_{c,d} is divisible by y_{a}, y_{c}, or y_{d}^{2}. This proves the third claim.

The containment \supseteq always holds.

We prove that the relation \subseteq holds in this case. Pick r \in (IS + \operatorname{\mathfrak{n}}^{[2]}). Write r = a + b, where a \in IS and b \in \operatorname{\mathfrak{n}}^{[2]}. Express a as

where p_{i} \in S and g_{i} \in I for 1 \leq i \leq k. Since I \subseteq R, no monomial belonging to \operatorname{Supp}(g_{i}) belongs to \operatorname{\mathfrak{n}}^{[2]} for 1 \leq i \leq k. For c \in \operatorname{Supp}(p_{i}) and c \in \operatorname{\mathfrak{n}}^{[2]}, replace a by a - cg_{i} and b by b + cg_{i}. After repeating this process we may assume that no monomial belonging to \operatorname{Supp}(a) belongs to \operatorname{\mathfrak{n}}^{[2]}. Since r \in KS + \operatorname{\mathfrak{n}}^{[2]} and b \in \operatorname{\mathfrak{n}}^{[2]}, it follows that a = r - b \in KS+\operatorname{\mathfrak{n}}^{[2]}. Hence,

where q_{i} \in S, h_{i} \in K for 1 \leq i \leq \ell, and v \in \operatorname{\mathfrak{n}}^{[2]}. Since no non-zero term of a nor of h_{i} belongs to \operatorname{\mathfrak{n}}^{[2]} for any 1 \leq i \leq \ell, it follows that a non-zero term of q_{i} belonging to \operatorname{\mathfrak{n}}^{[2]} will cancel with a non-zero term of v. After performing these cancellations, it must be the case that v = 0 since the support of a is disjoint from \operatorname{\mathfrak{n}}^{[2]}. Thus, a \in KS, and the claim follows.

Let r \in Q^{[2]}:_{S} J_{G}, and f \in J_{H} a minimal generator. The relation r \cdot f \in Q^{[2]} holds mod \operatorname{\mathfrak{n}}^{[2]}, and we obtain that

Hence, we have that

Since the variables x_{m+1},\ldots,x_{n},y_{m+1},\ldots,y_{n} do not appear in a minimal generating set of J_{H} nor of Q', it follows that

Lastly, since R \rightarrow S is a flat map of rings, it follows that

which completes the proof.

We have that

Fedder’s Criterion (Theorem 2.2) implies that J_{G} is not F-pure.

Follows from Proposition 2.4 since the set [n] \smallsetminus \{i,i+1,i+2\} is a cut set of G for every 1 \leq i \leq n.

This follows from Theorem 6.8, where we take a = i, b = i+1, c = i+2, and d = n+1 belonging to [n+1], S := [n+1] \smallsetminus \{a,b,c,d\}, and T := \{a,c\}.

We prove (1). Suppose that

Lemma 7.2 applied with P = P_{1,3}, a = 1, c = 3, p = 1, and q = 1 implies that equation (Proof) is equivalent to

It follows from Lemmas 7.2 and 9.2 that we may continue in this manner, and thus obtain that (Proof) is equivalent to

This proves (1). The proof of (2) is analogous.

We prove (3). Suppose that

Lemma 7.2 applied with P = P_{1,3}, a = 1, c = 3, p = y_{2}, and q = x_{2} implies that equation (?) is equivalent to

Now Lemma 7.2 applied with P = P_{2,4}, a = 2, c = 4, p = x_{3}, and q = y_{3} implies that equation (?) is equivalent to

It follows from Lemmas 7.2 and 9.2 that we may continue in this manner, and thus obtain that equation (?) is equivalent to

This proves (3).

Since r \in P_{i,i+2}^{[2]}:J_{G} for i \in \{1,n\}, Lemma 7.7 implies that r' is divisible by \omega_{S} for S = [n] \smallsetminus \{1,2\}.

Case 1. Suppose that r' is not divisible by x_{i}^{2} or by y_{i}^{2} for i = 1,2. Then r \in P_{3,5}^{[2]}:J_{G}, and Lemma 7.7 implies that m \mid r', in which case, the constraints on r' arising from r' \notin L imply that r' = h_{1,1}.

Case 2. Suppose that x_{1}^{2} \mid r'. Since r' \notin \mathfrak{a}_{2} and r' \notin \mathfrak{a}_{3}, it follows that y_{1} and x_{2}^{2} do not divide r'. Because r \in P_{n-1,1}^{[2]}:J_{G}, Lemma 7.7 implies that y_{2}^{2} \mid r' or that x_{2}y_{2} \mid r'. When y_{2}^{2} \mid r' it follows that x_{2} \nmid r' (otherwise, r' \in \mathfrak{a}_{2}), and hence r' = h_{4,1}. When x_{2}y_{2} \mid r', it follows that r' = h_{2,1}.

Case 3. The remaining cases where x_{2}^{2}, y_{1}^{2}, or y_{2}^{2} divides r' are considered analogously to case (2).

Let r \in J. Let S := \operatorname{Supp}(r) \smallsetminus (L + \mathfrak{a}_{0}). Define the sets

Lemmas 9.4 and 9.5 imply that there exists 1 \leq t \leq 4 and 1 \leq i_{1} < \cdots < i_{t} \leq 4 such that

It follows that

which completes the proof.

When n = 5, we verify with Macaulay2 that J_{G} is not F-pure. Hence, we may suppose that n \geq 7 is an odd integer. We utilize the notation for h_{i,j} and L as they are defined in 9.5. We will use Proposition 8.4 to prove that J_{G} is not F-pure. In the notation coming from Proposition 8.4, we have that A = \varnothing, B = [n], Q = P(A,B), m = 5, B' = [5], Q' = P(\varnothing,B'), and H is the graph on [m] having edge set

We define the ideal

We define \mathcal{ C } := \{P_{i,i+2} \mid 1 \leq i \leq 2n+1\}. Proposition 9.6 implies that the containment (8.4) of Proposition 8.4 is satisfied. We verify using Macaulay2 that equation (8.4) of Proposition 8.4 is satisfied. Thus, Proposition 8.4 implies that J_{G} is not F-pure.

1. The proof that P_{i,i+2} is a minimal prime of J_{G} is analogous to the proof in Lemma 9.2.

2. We show that Q_{1} is a minimal prime of J_{G}. First, we observe that if S := \{j \mid 2 \leq j \leq 2n+3\} \cup \{2n+7\}, then G \smallsetminus S consists of two connected components G_{1} and G_{2} where V(G_{1}) = \{1,2n+4,2n+6\} and E(G_{1}) = \{ \{1,2n+6\}, \{2n+4,2n+6\} \}, and G_{2} is the graph on the singleton vertex \{2n+5\}. It is clear that S is a cut set of G, and hence Q_{1} is a minimal prime of J_{G} by Proposition 2.4. Similarly, it is shown that Q_{2} is a minimal prime of J_{G}.

3. We show that Q_{3} is a minimal prime of J_{G}. If S = \{2n+6,2n+7\}, then G \smallsetminus S consists of two connected components. The set G_{1} on the vertex set 1, and G_{2} on the vertices \{j \mid 2 \leq j \leq 2n+5\}. Adding 2n+6 or 2n+7 reconnects these connected components, and it follows that S is a cut set of G.

Follows from Theorem 6.8. When applying Theorem 6.8 to Q_{1}, we let a = 2n+4, b = 2n+5, c = 2n+6, and d = 1. When applying Theorem 6.8 to Q_{2}, we let a = 3, b = 2, c = 2n+7, and d = 1.

First we observe that Lemma 7.2 applied to P_{3,5} where p = x_{1} and q = y_{1} implies that

Repeatedly applying Lemma 7.2 we get the following

The statement now follows from Lemma 7.3 applied to Q_{2} as we have that

Lemma 7.4 applied to Q_{1} and Q_{2} implies that

respectively. Lemma 7.2 implies that

if and only if

respectively.

By Lemma 7.5 it follows for r \in \bigcap_{i=1,2} (Q_{i}^{[2]}:J_{G}) and r' \in \operatorname{Supp}(r) that

The claim now follows by applying Lemma 7.2.

Lemma 7.2 implies that

Lemma 7.6 applied to Q_{1} with p = x_{2n+5}, q = y_{2n+5}, a = 1, c = 2n+4, and d = 2n+6 implies that equation (Proof) is equivalent to

Lemma 7.2 implies that equation (Proof) is equivalent to

Now, Lemma 7.5 applied via Q_{2} with a = 3, c =1, d = 2n+7, p = y_{2}, and q = x_{2} implies equation (Proof). A similar argument shows that

Lemma 7.6 applied via Q_{1} shows that

Lemma 7.2 implies that

Equation (?) has been previously established in this proof. A similar argument proves that

We have that

Since r' \notin L + \mathfrak{a}_{0}, Lemma 7.7 implies that \omega_{S} \mid r' for S = [2n+7] \smallsetminus \{1,4,5\}.

Case 1. Suppose that r' is not divisible by x_{i}^{2} nor by y_{i}^{2} for i = 4,5. Then, Lemma 7.8 applied to Q_{1} implies that \omega_{T} \mid r' where T = [2n+7] \smallsetminus \{1\}, and that x_{1} \mid r' or that y_{1} \mid r'. Thus, we have that

for some d' \in \{x_{1},y_{1},x_{1}y_{1},x_{1}^{2},y_{1}^{2}\}. Corollary 9.15 implies that

1. d' = x_{1}y_{1}, in which case r' = h_{1,3},

2. d' = x_{1}^{2}, in which case r' = h_{2,1}.

3. d' = y_{1}^{2}, in which case r' = h_{3,1}.

Case 2. Suppose that r' is divisible by x_{4}^{2}. Then r' is not divisible by y_{i}^{2} for i=1,4,5, x_{1}^{2}, or x_{5}^{2} (otherwise, r' \in \mathfrak{a}_{2}). The requirement that

implies via Lemma 7.7 that x_{5}y_{5} \mid r'. Hence,

for some monomial d' with \operatorname{Supp}(d') \in \{1,x_{1},y_{1},x_{1}y_{1}\}. (We notice that y_{4} \nmid d'; otherwise, r' \in \mathfrak{a}_{3}.) Corollaries 9.14 and 9.15 imply that d' = x_{1}y_{1}, and thus that r' = h_{2,2}.

Case 3. Suppose that r' is divisible by x_{5}^{2}. This case is similar to case 2, and we find that r' = h_{2,3}.

Case 4. Suppose that r' is divisible by y_{4}^{2}. As in case 2 we have that

for some monomial d' with d' \in \{1,x_{1},y_{1},x_{1}y_{1},x_{1}^{2} \}. Corollary 9.14 implies that d' = x_{1}y_{1} (in which case r' = h_{3,2}) or that d' = x_{1}^{2} (in which case r' = h_{1,1}).

Case 5. Suppose that r' is divisible by y_{5}^{2}. This case is similar to case 4, and we find that r' = h_{1,2} or that r' = h_{3,3}.

Let r \in J. Let S := \operatorname{Supp}(r) \smallsetminus (L + \mathfrak{a}_{0}). Define the sets

Corollaries 9.12 and 9.13 and Lemma 9.16 imply that there exists 1 \leq t \leq 5 and 1 \leq i_{1} < \cdots < i_{t} \leq 5 satisfying \# \left( \{i_{1},\ldots,i_{t}\} \cap \{1,2,3\} \right) \leq 1 such that

It follows that

For 1 \leq i \leq 5 we have that

which completes the proof.

When n = 0, it can be verified via Macaulay2 that J_{G} is not F-pure. Hence, it suffices to consider the case n \geq 1. We utilize the notation for h_{i,j} and L as they are defined in 9.16. We will use Proposition 8.4 to prove that J_{G} is not F-pure. Adopting the notation from Proposition 8.4 we have that A = \{2n+6,2n+7\}, B = [2n+7] \smallsetminus (A\cup \{1\}), Q = P(A,B), m = 5, B' = [5] \smallsetminus \{1\}, Q' = P(\varnothing,B'). We define the ideal

We define \mathcal{ C } := \{P_{i,i+2} \mid 2 \leq i \leq 2n+3\}\cup \{Q_{i} \mid i \in \{1,2\}\}. Proposition 9.17 implies that the containment (8.4) of Proposition 8.4 is satisfied. We verify using Macaulay2 that the containment (8.4) of Proposition 8.4 is satisfied. Thus, Proposition 8.4 implies that J_{G} is not F-pure.

After completing G at the vertex 1 we have that vertex 2 is connected to the vertices 2n+7 and 2n+8. Now, the induced subgraph of G on the vertices [2n+8] \smallsetminus \{1\} is isomorphic to the graph \mathrm{co\!-\!XF}_1^{2n+3}. As \mathrm{co\!-\!XF}_1^{2n+3} is not F-pure for n\geq 0 (Theorem 9.18), it follows that \mathrm{co\!-\!XF}_5^{2n+3} is not F-pure by Lemma 3.2.

The proof is similar to the proof of Lemma 9.10

Follows from Theorem 6.8. When applying Theorem 6.8 to Q_{1}, we let a = 4, b = 3, c = 2n+7, and d = 2. When applying Theorem 6.8 to Q_{2}, we let a = 2n+4, b = 2n+5, c = 2n+6, and d = 1.

We have that

Lemma 7.5 applied to Q_{2} with a = 2n+4, c = 1, and q = x_{2} implies that

Similarly, it is proven that

Lemma 7.5 applied via Q_{1} implies that y_{2} does not divide q. Lemma 7.2 implies that

Lemma 7.5 applied via Q_{2} implies that y_{1} does not divide q.

We define the elements

Suppose that r' \in \operatorname{Supp}(r). Then, Lemma 7.2 implies that r'' and r''' also belong to \operatorname{Supp}(r). Applying Lemma 7.8 to r'' via Q_{1} implies that x_{2} or y_{2} divides r''. Hence, q is not divisible by x_{2} or by y_{2}. Similarly, applying Lemma 7.8 to r''' via Q_{2} implies that x_{1} or y_{1} divides r'''. Therefore, q is not divisible by x_{1} or by y_{1}. This leads to the conclusion that \# \operatorname{Supp}(q) \leq 2.

When p = 1, we have that

Lemma 7.5 applied to Q_{2} with a= 2n+4, c = 1, d = 2n+4, p = x_{2n+5}, and q = y_{2}y_{2n+5} implies that

When p = y_{1}, we have that

Lemma 7.5 applied to Q_{1} with a= 4, c = 2, d = 2n+7, p = x_{3}, and q = y_{3} implies that

When p = y_{2}, it is shown that

similarly as in the case where p = y_{1}.

Since r \in P_{i,i+2}^{[2]}:J_{G} for i = 3 and i = 4 and r' \notin L + \mathfrak{a}_{0}, Lemma 7.7 implies that \omega_{S} \mid r' for S = [2n+7] \smallsetminus \{1,2,4,5\}.

Case 1. Suppose that r{'} is not divisible by x_{4}^{2} nor by x_{5}^{2}. Since n \geq 4, r \in P_{2n+3,2n+5}^{[2]}:J_{G}, and r' \notin L + \mathfrak{a}_{0}, Lemma 7.7 implies that x_{4}y_{4}x_{5}y_{5} \mid r'. Consequently, we have that

for some monomial d' with \operatorname{Supp}(d') \subseteq \{x_{i},y_{i}\}_{i=1,2}. Lemma 7.8 applied to r' via Q_{1} and Q_{2} implies that d' is divisible by x_{i} or by y_{i} for both i = 1 and i = 2. After accounting for the constraint that r' does not belong to \mathfrak{a}_{1} nor to \mathfrak{a}_{2}, we see that

Expanding the products in (?) and removing those terms which would force r' \in L + \mathfrak{a}_{0}, we find that

Lemma 9.26 implies that d' \neq x_{1}x_{2} and that d' \neq y_{1}y_{2}. For the remaining values of d' we compute that

Case 2. Suppose that r' is divisible by x_{5}^{2}. Since r' \notin \mathfrak{a}_{3}, we conclude that r' is not divisible by x_{4}^{2}. Lemma 7.7 applied to r' via P_{5,7} implies that x_{4}y_{4} \mid r'. Hence we have that

for some monomial d' where \operatorname{Supp}(d') \subseteq \{x_{i},y_{i}\}_{i=1,2}. The fact that y_{5} \notin \operatorname{Supp}(d') follows from the condition that r' \notin \mathfrak{a}_{2}. Suitable applications of Lemma 7.2 together with Lemma 7.8 applied to r' via Q_{1} and Q_{2} implies that

Since x_{1}x_{2}x_{5}^{2} \in \mathfrak{a}_{3}, we may suppose that x_{1} x_{2} \nmid d'. If y_{1}y_{2} \mid d' and \operatorname{Supp}(d') \subseteq \{y_{1},y_{2}\}, then

for p \in \{1,y_{1},y_{2}\}. However, Lemma 9.29 implies that equation (?) is impossible. Thus, we are reduced to considering

For these values of d' we compute that

Case 3. Suppose that r' is divisible by x_{4}^{2} and that r' is not divisible by y_{2}^{2}. Since r' \notin \mathfrak{a}_{3}, it follows that x_{5}^{2} \nmid r'. Since y_{2}^{2} \nmid r', Lemma 7.8 applied to r' via Q_{1} implies that x_{5}y_{5} \mid r'. Thus,

for some monomial d' with \operatorname{Supp}(d') \subseteq \{ x_{i},y_{i} \}_{i=1,2}. The fact that y_{4} \notin \operatorname{Supp}(d') follows from the condition that x_{4}^{2}y_{4} \in \mathfrak{a}_{2}. Since r' is not divisible by y_{4}, Corollary 9.27 implies that y_{1}y_{2} \mid d'. Thus, we have that

We have that d' \neq x_{1}y_{1}x_{2}y_{2} (since x_{1}x_{2} y_{4}^{2} \in \mathfrak{a}_{3}), and we have that d' \neq y_{1}^{2}y_{2}^{2} (since y_{1}^{2}y_{2}^{2} \in \mathfrak{a}_{4}). For the remaining values of d', we compute that

Case 4. Suppose that r' is divisible by x_{4}^{2}y_{2}^{2}. Since r' \notin \mathfrak{a}_{2} it follows that y_{4} \nmid r' and that x_{2} \nmid r'. Hence

for some monomial d' with \operatorname{Supp}(d') \subseteq \{ x_{1},x_{5},y_{1}, y_{5} \}. Applying Lemma 7.2 to r' via P_{4,6} implies that

Since x_{5}^{2} \nmid d' (otherwise, r' \in \mathfrak{a}_{3}) and y_{5}^{2} \nmid d' (otherwise, r' \in \mathfrak{a}_{1}), Lemma 7.7 applied to r'' via P_{6,8} implies that x_{5}y_{5} \mid r''. Thus, we have that

for some monomial d' with \operatorname{Supp}(d') \subseteq \{ x_{1},y_{1} \}. Lemma 7.2 implies that

We observe that y_{1}^{2} \nmid r''' (otherwise, r' \in \mathfrak{a}_{4}), y_{2n+4}\nmid r''', and y_{2n+6}^{2} \nmid r'''. Consequently, Lemma 7.8 applied to r''' via Q_{2} implies that y_{1} divides r'''. Thus, y_{1} \mid d', and we have that

for some d' \in \{1,x_{1}\}. For d' \in \{1,x_{1}\}, we have that

When p \in \{1,y_{1}\}, we have that

Specializing to the case that p = 1, we have that

by applying Lemma 7.6 to Q_{2} with p' = x_{2n+5}, q' = y_{2}y_{2n+5}, a = 1, and c = 2n+4. Finally, Lemma 7.2 applied to equation (?) implies that

Specializing to the case that p = y_{1}, we have that

Case p = y_{2}: The proof is analogous to the case where p = y_{1}.

Case p = y_{1}y_{2}: We have that

(1) First, we observe that

by applying Lemma 7.6 to h_{1,2} with Q_{1} and p = 1, q = x_{1}, a = 2, c = 4, and d = 2n+7. Next, we observe that

The proof of statements (2) and (4) are similar to the proof of (1), and we do not repeat these proofs.

The proof of statement (3) is similar to the proof of (5). Thus, we give the proof of (5) below. We have that

Lemma 7.3 applied to Q_{2} implies that

Lemma 7.2 implies that

We observe that

which completes the proof.

Let r \in J. Let S := \operatorname{Supp}(r) \smallsetminus (L+\mathfrak{a}_{0}). Define the sets

Then, Lemmas 9.31, 9.32, 9.33, and 9.34 imply that there exists 1 \leq t \leq 12 and 1 \leq i_{1} < \cdots < i_{t} \leq 12 satisfying \# \left( \{i_{1},\ldots,i_{t}\} \cap \{6,7,8\} \right) \leq 1 such that

We observe that for the elements of S_{2} that we have the relation

A similar relation holds for the elements of S_{i} for 3 \leq i \leq 7. The elements of S_{8} satisfy the relation

It follows that

Now, equations (Proof) and (Proof) imply that

which completes the proof.

When n = 0 and n = 1, it can be verified via Macaulay2 that J_{G} is not F-pure. Hence, it suffices to consider the cases n \geq 3. We utilize the notation for h_{i,j} and L as they are defined in Lemma 9.31. We will use Proposition 8.4 to prove that J_{G} is not F-pure. Adopting the notation from Proposition 8.4 we have that A = \{2n+6,2n+7\}, B = [2n+7] \smallsetminus (A\cup \{1,2\}), Q = P(A,B), m = 7, B' = [7] \smallsetminus \{1,2\}, Q' = P(\varnothing,B'). We define the ideal

We define \mathcal{ C } := \{P_{i,i+2} \mid 3 \leq i \leq 2n+3\}\cup \{Q_{i} \mid i \in \{1,2\}\}. Proposition 9.35 implies that the containment (8.4) of Proposition 8.4 is satisfied. We verify using Macaulay2 that the containment (8.4) of Proposition 8.4 is satisfied. Thus, Proposition 8.4 implies that J_{G} is not F-pure.

Let P_{1},\ldots,P_{t} be as in Theorem 10.2. Let \alpha_{1},\ldots,\alpha_{g} denote the binomials corresponding to the edges of the P_{i}, where g = \operatorname{ht} J_{G}. By [pandey2024linkage, Corollary 3.3], if R/\mathfrak{a} is F-injective, where \mathfrak{a} = (\alpha_{1},\ldots,\alpha_{g}), then R/J_{G} is F-pure as \{\alpha_{i}\}_{i=1}^{g} form a regular sequence in J_G of length \operatorname{ht} J_{G}. The ideal \mathfrak{a} is even F-pure. Since the \alpha_{i} form a regular sequence, it follows from Proposition 4.8 that

The lead term of \prod_{i=1}^{g} \alpha_{i}^{p-1} with respect to the lexicographic term order induced by x_{1} > x_{2} > \cdots > x_{n} > y_{1} > y_{2} > \cdots > y_{n} does not belong to \mathfrak{m}^{[p]}. Thus,

Hence, Fedder’s criterion (Theorem 2.2) implies that R/\mathfrak{a} is F-pure.

Note that J_{G} being unmixed (resp., of König type) are combinatorial statements independent of the base field; see, for instance, Proposition 2.4 (resp., [laclair2023invariants]). Hence, we may assume that the base field has characteristic 2. Proposition 10.3 implies that R_{G} is F-pure, and Theorem A implies that G is weakly closed.

Since G is a weakly closed graph, R_{G} is F-pure whenever the base field \Bbbk has characteristic two (Theorem 5.3). Lemma 3.3 implies that G \smallsetminus v, G_{v}, and G_{v} \smallsetminus v define F-pure binomial edge ideals. Since the base field has characteristic two, Theorem A implies that the graphs G \smallsetminus v, G_{v}, and G_{v} \smallsetminus v are all weakly closed.