2
$\begingroup$

From the paper Some further results on ideal convergence in topological spaces by Pratulananda Das (Topology and Its Applications, vol.159, pp.2621-2626):

Theorem 3: . Let $X$ be a space with $\operatorname{hcld}(X) = \omega$. Then for each $F_\sigma$ -set $A$ in $X$ there exists a sequence $x = (x_n)_{n∈\mathbb{N}}$ in $X$ such that $A = \mathcal{I}(L_x)$ provided $\mathcal{I}$ is an analytic P-ideal.

Proof:. Let $A =\bigcup_{i=1}^{\infty} A_i$ where each $A_i$ is a closed subset of $X$. By Lemma 1, for each $i$, we can find a sequence $(y_{i,j})_{j∈\mathbb{N}} \subset A_i$... such that $A_i = L((y_{i,j}))$.

Before we proceed, we first observe that if $K \notin \mathcal{I}$ then $\lim_{n\to\infty} \varphi (K\setminus [1,n]) = \beta$ (say) $\neq 0$. Then $\beta > 0$ (possibly $\beta = \infty$). From the lower semicontinuity of $\varphi$, there are finite pairwise disjoint sets $C_j$ , $j \in \mathbb{N}$ with $C_j \subset K$ and $\lim_{j\to\infty} \varphi(C_j) = \beta$. Let $\mathbb{N} =\bigcup_{i=1}^{\infty}D_i$ be a decomposition of $\mathbb N$ into pairwise disjoint subsets of $\mathbb N$. Put $K_1 = (K\setminus \bigcup_j C_j) \cup ( \bigcup_{j\in D_1} C_j )$ and for $i > 1$ $K_i=\bigcup_{j∈D_i}C_j$ . Then one can check that the sets $K_i$’s are pairwise disjoint subsets of $K$, $K =\bigcup_{i=1}^{\infty}K_i$ . Further it follows that $$\lim_{n\to\infty} \varphi(K_i\setminus[1,n])= \beta \quad \forall i \in \mathbb{N}.$$ Clearly then $K_i \notin \mathcal{I}$ $\forall i$.

We now come back to the main proof. First decompose $\mathbb{N}$ into pairwise disjoint sets $\mathbb N =\bigcup_{i∈\mathbb N} M_i$ where $M_i \notin \mathcal{I}$ for any $i$. Further for each $i \in \mathbb N$, decompose $M_i=\bigcup_{j∈\mathbb N} B_{i,j}$ where $$\lim_{n\to\infty} \varphi(B_{i,j}\setminus [1,n])=\lim_{n\to\infty} \varphi(M_i\setminus [1,n])=β_i \text{ (say) } \forall j \in \mathbb{N}$$ and $B_{i,j} \notin \mathcal{I}$ $\forall j \in \mathbb N$. Define a sequence $x = (x_n)_{n\in\mathbb N}$ as follows: $$x_n = y_{i,j} \quad\text{for each } n ∈ B_{i,j}.$$ We shall show that $A = \mathcal{I}(L_x)$. The following two cases arise.

And so do my problems again

I believe at this point he is proving the equality by inclusion-reverse inclusion method. (i), below, will show $L(x)\subset A$ by contrapositivity and (ii) will show $A\subset L(x)$ and thus arriving at the desired conclusion.

(i) First suppose that there is a subsequence $(x_{n_k})_{k\in \mathbb N}$ of $x$ converging to $y$ and $y\notin A$.Then for every $i$,$\bigcup_{j=1}^i M_j$ contains only finite number of terms $n_1,n_2,\ldots ,n_s$ but $\{n_1,n_2, \ldots ,n_s\} \in \mathcal I$ (since $\mathcal I$ is admissible). So $y \notin \mathcal{I}(L_x)$.

I cannot understand how we know that $\bigcup_{j=1}^i M_j$ is finite here. $y\notin A \Rightarrow y\notin A_i$ for any $i$. If possible let infinitely many $n_k$'s are in $\bigcup_{j=1}^i M_j$.Wh ere is the contradiction? $y$ is a limit point of that sequence.

(ii) Now take $y \in A$. Then $y \in A_i$ for some $i \in \mathbb N$. Then there is a subsequence $(x_{i,j_k})_{k\in \mathbb N}$ of $(x_{i,j})_{j\in\mathbb N}$ which converges to $y$.

Where did this sequence $(x_{i,j})_{j \in \mathbb N}$ come from? And why would we assume that $y\in A_i$ implies there is a subsequence of $x_{i,j}$ converging to $y$?

Choose an $\epsilon > 0$ such that $\epsilon < \beta_i$ . Following the proof in Theorem 1 we can find a sequence of positive integers $t_1 < t_2 < t_3 < \cdots$ such that $$\varphi (B_{i,j_k} \cap (t_k,t_{k+1}])\ge \beta_i−\epsilon.$$ We now define $B=\bigcup_{k\in \mathbb N}\{B_{i,j_k} \cap (t_k,t_{k+1}]\}$. Then it readily follows that $$\lim_{n\to \infty} \varphi (B\setminus [1,n])\ge \beta_i−\epsilon.$$

I do not know how.

Therefore $B\notin \mathcal{I}$ and it is easy to observe that $\lim_{n\to \infty,n\in B} x_n = y$.

I cannot understand this one either.

This shows that $y \in \mathcal{I}(L_x)$.

I managed the earlier parts thanks to Martin Sleziak. Please explain these fourthings to me. I wish I could speak to the author personally.

  • 0
    This question contains a lot of symbols that are not defined within it. It would be best if you included all pertinent definitions, results, etc. within the question itself, and not require potential answerers to read through the linked paper.2017-02-03
  • 0
    @arjafi: I know. Like I said I posted half asleep. I'll fix things soon.2017-02-03

0 Answers 0