Fixed elements of pircon automorphisms

Since P is finite, \theta has finite order K. Observe that each automorphism \theta^i, i \in [K], transforms the SPM M into an SPM M_i, i.e., \theta^i \circ M=M_i \circ \theta^i. In particular, M_K=M.

For a given p \in P, define

Thus, C(p) consists of the elements that are connected to p by the SPMs M_i. By abuse of notation, we also let C(p) denote the subposet of P induced by the set C(p). For a \in C(p) given by a=M_{i_t} \circ M_{i_{t-1}} \circ \cdots \circ M_{i_1}(p), define b \in C(p) as b=M_{i_t}' \circ M_{i_{t-1}}' \circ \cdots \circ M_{i_1}'(p), where the M_{i_j}' are recursively given by

For a_j,b_j \in C(p), where a_j=M_{i_j} \circ \cdots \circ M_{i_1}(p) and b_j=M_{i_j}' \circ \cdots \circ M_{i_1}'(p), 1 \leq j \leq t, we have b_j \leq a_j. To see it, we use induction by assuming that b_{j-1} \leq b_{a-1}. We have the following three cases.

• (i) If a_j \geq a_{j-1}, there is nothing to prove.

• (ii) If a_j<a_{j-1} and M_{i_j}'=M_{i_j}, we have b_j= M_{i_j}(b_{j-1})<b_{j-1} and M_{i_j}(a_{j-1})=a_j<a_{j-1}. Hence, we may apply the lifting property to get b_j \leq a_j (with equality if and only if a_{j-1}=b_{j-1}).

• (iii) Finally, if a_j<a_{j-1} and M_{i_j}'=\mathrm{id}, we have M_{i_j}(b_{j-1}) \geq b_{j-1} and M_{i_j}(a_{j-1})=a_j<a_{j-1}. We can then apply the lifting property to get b_j=b_{j-1} \leq a_j.

Thus b_j \leq a_j, proving the claim. In particular, b \leq a. By construction, b \leq p, so if p and a are both minimal in C(p), we have a=b=p. Hence C(p) contains a unique minimal element. Analogously, if we reverse the inequality in (Proof), C(p) also contains a unique maximal element. (Observe that P need not have a minimum. However, Lemma 2.3 will still hold, with the same proof as in [A-H-H_pircons], even if P does not have a maximum, as long as M satisfies the other three conditions in Definition 2.2.)

If p is not minimal in C(p), then M_i(p) \vartriangleleft p for some i. To see this, note that if a<p in the above construction, then b \neq p and so we may choose i=i_j for the minimal j which satisfies M_{i_j}'=M_{i_j}. Reversing the inequality in (Proof), a similar reasoning shows that if p is not maximal in C(p), then there is at least one i such that p \vartriangleleft M_i(p).

If p \in P^\theta and p \leq M_i(p) for at least one i, then this will hold for all i and hence p=\min C(p). Similarly, if p \in P^\theta and M_i(p) \leq p, then p=\max C(p). Therefore, for any p \in P^\theta we have p=\min C(p) or p=\max C(p), both if C(p)=\{p\}.

Since \theta permutes the SPMs M_i, we have \theta(C(p))=C(\theta(p)) for all p \in P. Thus the claim below follows.

Moreover we have:

Let \vartriangleleft_\theta denote the covering relation in P^\theta.

Define the function \mathrm{M}^\theta \colon P^\theta\to P^\theta by

By Claim Proof, \mathrm{M}^\theta is well defined. We show that \mathrm{M}^\theta is an SPM on P^\theta by verifying the conditions in Definition 2.2.

• For every x \in P^\theta, (\mathrm{M}^\theta)^2(x)=x. That is, (\mathrm{M}^\theta)^2=\mathrm{id}.

• Since \theta is an automorphism of P, \theta(\hat{1})=\hat{1}, so \hat{1} \in P^\theta. Because M_i is an SPM, M_i(\hat{1}) \neq \hat{1}, so \hat{1} \neq \min C(\hat{1}). By Claim Proof, \mathrm{M}^\theta(\hat{1})=\min C(\hat{1}) \vartriangleleft_\theta \hat{1}.

• For any x \in P^\theta, Claim Proof shows that \mathrm{M}^\theta(x) \vartriangleleft_\theta x, x \vartriangleleft_\theta \mathrm{M}^\theta(x), or \mathrm{M}^\theta(x)=x.

• Let x \vartriangleleft_\theta y and \mathrm{M}^\theta(x) \neq y. We must show that \mathrm{M}^\theta(x)<\mathrm{M}^\theta(y). By Claim Proof, \min C(x) \leq \min C(y) and \max C(x) \leq \max C(y). Moreover, both inequalities are strict since \mathrm{M}^\theta(x) \neq y.

  • (i) If x \neq \min C(x), then \mathrm{M}^\theta(x)=\min C(x)<\min C(y) \leq \mathrm{M}^\theta(y).

  • (ii) If y \neq \max C(y), we have \mathrm{M}^\theta(x) \leq \max C(x)<\max C(y)=\mathrm{M}^\theta(y).

  • (iii) If x=\min C(x) and y=\max C(y), we have x=\min C(x)<\min C(y) \leq y and x \leq \max C(x)<\max C(y)=y. Because x \vartriangleleft_\theta y, we conclude that \min C(x)=\max C(x) and \min C(y)=\max C(y). Hence \mathrm{M}^\theta(x)=x \vartriangleleft_\theta y=\mathrm{M}^\theta(y).

For any non-minimal element p \in P, the order ideal P_{\leq p} is finite and has an SPM. Let p \in P^\theta be non-minimal. The elements in P^\theta_{\leq p} are the fixed elements of the restriction of \theta to P_{\leq p}. Therefore, by Theorem 3.1, P^\theta_{\leq p} has an SPM.

Let \sigma<\tau in F_n^B, let C_D=(\sigma \vartriangleleft \tau_k \vartriangleleft \cdots \vartriangleleft \tau_1 \vartriangleleft \tau) be the decreasing \sigma-\tau-chain in I_n^B, where k \geq 1, and let (i_\tau,j_\tau) be the label on \tau_1 \vartriangleleft \tau. Observe that if we can prove that \tau_1 \in F_n^B, it follows that \tau_1,\ldots,\tau_k \in F_n^B, because the decreasing \sigma-\tau_1-chain in I_n^B is \sigma \vartriangleleft \tau_k \vartriangleleft \cdots \vartriangleleft \tau_2 \vartriangleleft \tau_1.

Let h=\operatorname{di}(\sigma,\tau). Since \sigma has no fixed points, h is an exceedance of \sigma (i.e., \sigma(h)>h). Indeed, if \sigma(i)<i, then \sigma(i) is an exceedance of \sigma, and \sigma(\sigma(i))=\tau(\sigma(i)) \Leftrightarrow \sigma(i)=\tau(i). Consider any cover \pi \vartriangleleft \pi' in [\sigma,\tau]. Two applications of Definition 2.5 yield \operatorname{di}(\pi,\pi') \geq \operatorname{di}(\pi,\tau) \geq h. Since the label (i,j) on \pi \vartriangleleft \pi' satisfies i=\operatorname{di}(\pi,\pi'), it follows that h is in some label on C_D, and because C_D is decreasing, i_\tau=h.

Suppose \tau_1 \notin F_n^B so that \tau_1 \vartriangleleft \tau is a case where the number of fixed points decreases. The only such cases are E1.1, N1.1, N4.2, and H2.1, and in all of them, h is a fixed point of \tau_1. Observe that \tau_1 has fixed points also in E2.6a, N1.2a, N1.2b, N1.3a, N1.3b, N2.2, H1.2, H2.2a, H2.2b, and H2.3, and h is not always one of them, but the number of fixed points does not decrease in those cases. Therefore, \sigma[h,h+1]>\tau_1[h,h+1]. Since \sigma \leq \tau_1, it follows from Definition 2.5 that \tau_1 \in F_n^B.

We have proved that the decreasing \sigma-\tau-chain in I_n^B is contained in F_n^B. The next step is to show that this chain is lex-maximal. This is done in the same way as in the proof of [Hansson_graded_A, Theorem 1.4], where Incitti’s type A version of Lemma 5.3 is used. We include the argument here in order to make the proof complete.

To obtain a contradiction, consider the lex-maximal \sigma-\tau-chain in I_n^B, and suppose it is not decreasing; call it C=(\pi_1 \vartriangleleft \cdots \vartriangleleft \pi_{k+2}) and say that \lambda(\pi_1,\pi_2) \leq \lambda(\pi_2,\pi_3). By Lemma 5.3, \pi_1 \vartriangleleft \pi_2 \vartriangleleft \pi_3 is lex-minimal among the \pi_1-\pi_3-chains in I_n^B. Thus, \pi_1 \vartriangleleft \pi_2' \vartriangleleft \pi_3 \vartriangleleft \cdots \vartriangleleft \pi_{k+2}, where \pi_1 \vartriangleleft \pi_2' \vartriangleleft \pi_3 is the decreasing \pi_1-\pi_3-chain, is lex-larger than C, a contradiction.

So, the decreasing \sigma-\tau-chain in I_n^B is contained in F_n^B and it is lex-maximal. Now, by reversing the lexicographic order on the set \{(i,j) \in [\pm n]^2 \mid i<j\}, we obtain an edge-labelling of F_n^B such that in each interval [\sigma,\tau], there is an increasing \sigma-\tau-chain which is lex-minimal. By Lemma 5.4 and the remark following it, this is an EL-labelling of F_n^B.