For a topology $\mathcal{T}$ on a set $S$, where $\mathcal{T}$ does not have a finite base, I define the grasp $g(\mathcal{T})$ to be the least infinite cardinal $\kappa$ such that $\mathcal{T}$ has a base $\mathcal{B}$ with $|\mathcal{B}|= w(\mathcal{T})$ , ($w$ is the usual weight function), satisfying $$\mathcal{T}=\{\bigcup V : V\in [\mathcal{B}]^{\leq \kappa}\}.$$

That is, every open set is the union of at most $\kappa$ members of $\mathcal{B}$, and $|\mathcal{B}|$ is minimum among all bases.

The following can be shown by elementary means:

  1. $g(\mathcal{T})\leq w(\mathcal{T})$. (Obvious).
  2. $|\mathcal{T}|\leq w(\mathcal{T})^{g(\mathcal{T})}$. (Obvious).
  3. $g(\mathcal{T})\leq g(D(w(\mathcal{T})))$ where $D(\lambda)$ is discrete space of cardinality $\lambda$.
  4. If $Y$ is a subspace of $X$ and $w(Y)=w(X)$ then $g(Y)\leq g(X)$.
  5. When $\kappa$ is an infinite cardinal with the discrete or the order topology then

    1. $\operatorname{cf}(\kappa)\leq g(\kappa)\leq \kappa$.
    2. If $\kappa$ is a singular strong limit then $g(\kappa)=\operatorname{cf}(\kappa)$.

    (This last one helps to distinguish $g$ from other topological cardinal functions.)

I have two questions.

Question 1: Referring to (2) above, $g(T)$ is not necessarily the least cardinal $\lambda$ such that $|\mathcal{T}|\leq w(\mathcal{T})^\lambda$ because if we assume $2^{\omega}=2^{\omega_1}$ and let $\mathcal{T}$ be the discrete topology on $\omega_1$ then $|\mathcal{T}|=\omega_1^{\omega}$ but by (5)(i) $g(\mathcal{T})=\omega_1$. So is there an example like this in $\mathsf{ZFC}$?

Question 2: By (5)(i) we have $\operatorname{cf}(2^{\omega})\leq g(D(2^{\omega}))\leq 2^{\omega}$. What values for $g(D(2^{\omega}))$ other than $2^{\omega}$ are consistent?

  • 3
    $\begingroup$ Originally on Mathematics: Topological cardinal function : The grasp. $\endgroup$
    – user642796
    Oct 17, 2015 at 6:11
  • $\begingroup$ I got a lot of comments there but no progress so I decided to try the professional site.As I said on Math Exchange, last year by e-mail with Prof. Franklin Tall,he said he had not heard of this function.It just occurred to me to also ask whether g(T) can ever be singular.I am not a professional. $\endgroup$ Oct 17, 2015 at 6:19

2 Answers 2


Proposition (with Z. Szentmiklóssy): It is consistent that $\omega_1=cf (2^{\omega})<g(D(2^{\omega}))=2^{\omega}$.

Proof: Assume GCH in the ground model. For ${\alpha}<{\omega}_1$ let $$ P({\alpha})=Fn({\omega}_{{\alpha}+1}\times {\omega}_{{\omega}_1+1},2;{\omega}_{{\alpha}+1}), $$ and $$ P=\prod_{{\alpha}<{\omega}_1}P({\alpha}),$$ where the product is the product with full support.

So $P$ is just an Easton-forcing, hence it preserves all the cardinals, and $2^{{\omega}_{{\alpha}+1}}={\omega}_{{\omega}_1+1}$ for all ${\alpha}<{\omega}_1$ in $V^P$.

Claim: For each ${\beta}<{\omega}_1$ there is an almost disjoint family $\mathcal F_{\beta}\subset [{\omega}_{{\beta}+1}]^{{\omega}_{{\beta}+1}}$ with $|\mathcal F_{\beta}|={\omega}_{{\omega}_1+1}$ in $V^P$.

Proof of the Claim: Write $P^{\beta}=\prod_{{\beta}\le {\alpha}<{\omega}_1}P({\alpha})$ and
$P_{\beta}=\prod_{{\alpha}<{\beta}}P({\alpha})$. Since $P^{\beta}$ is ${\omega}_{{\beta}+1}$-complete, and forcing with $P({\beta})$ introduces ${\omega}_{{\omega}_1+1}$ new subsets of ${\omega}_{{\beta}+1}$ , we have $$ V^{P^{\beta}}\models 2^{{\omega}_{\beta}}={\omega}_{{\beta}+1}\land 2^{{\omega}_{{\beta}+1}}\ge {\omega}_{{\omega}_1+1}, $$ hence in the model $V^{P^{\beta}}$ we have an almost disjoint family $\mathcal F_{\beta}\subset [{\omega}_{{\beta}+1}]^{{\omega}_{{\beta}+1}}$ with $|\mathcal F_{\beta}|={\omega}_{{\omega}_1+1}$.

Since $P=P^{{\beta}}*P_{\beta}$, we have such an $\mathcal F_{\beta}$ even in $V^P$. So we proved the Claim. Q.E.D

Let $Q=P*Fn ({\omega}_{{\omega}_1},2)$.

Since $P$ is $\sigma$-complete, we have $$ V^P\models 2^{\omega}={\omega}_1\land ({\omega}_{{\omega}_1})^{\omega}={\omega}_{{\omega}_1}, $$ and so $$ |2^{\omega}|^{V^Q}\le ((|Q|^{\omega})^{\omega})^{V^P}={\omega}_{{\omega}_1}. $$ So $2^{\omega}={\omega}_{{\omega}_1}$ in $V^Q$.

Assume on the contrary that $g(D({\omega}_{{\omega}_1}))={\omega}_{\alpha}<{\omega}_{{\omega}_1}$ in $V^Q$.

Let $\mathcal B$ be a base witnessing $g(D({\omega}_{{\omega}_1}))={\omega}_{\alpha}$.

Since every $F\in \mathcal F_{\alpha}\subset [{\omega}_{{\alpha}+1}]^{{\omega}_{{\alpha}+1}}$ is the union of at most ${\omega}_{\alpha}$-many elements of $\mathcal B$, so $F$ contains at least one element $B_F\in \mathcal B$ of cardinality ${\omega}_{{\alpha}+1}$. But the elements of $\mathcal F_{\alpha}$ are almost disjoint (i.e. the cardinalities of the intersections $\le {\omega}_{\alpha}$), so $B_F\ne B_{F'}$ for $F\ne F'$. Thus $|\mathcal B|\ge |\mathcal F_{\alpha}|>{\omega}_{{\omega}_1}$. Contradiction.

  • $\begingroup$ I am accepting this for now without without studying as it will take me some time and I will have to review some topics in my copy of Kunen's Set Theory.Which gives me something to do for the next week. Thank you. $\endgroup$ Oct 29, 2015 at 18:01

Proposition: It is consistent that $cf (2^{\omega})=g(D(2^{\omega}))<2^{\omega}$.

Proof: Assume GCH in the ground model. Let $P=Fn(\omega_{\omega_1},2)$. Since $2^\omega=\omega_{\omega_1}$ in $V^P$ it is enough to show that $g(D(2^{\omega})=\omega_1$ in the generic extension.

Let $G$ be a generic filter. In $V[G]$ for ${\alpha}<\omega_1$ let $$ \mathcal B_{\alpha}=\mathcal P(\omega_{\alpha})\cap V[G\cap Fn(\omega_{\alpha},2)]. $$ Since $$ |\mathcal B_{\alpha}|= (2^{{\omega}_{\alpha}})^{V[G\cap Fn(\omega_{\alpha},2)]}= {\omega}_{{\alpha}+1}, $$ the family $$ \mathcal B=\bigcup\{\mathcal B_{\alpha}:{\alpha}<\omega_1\} $$ has cardinality ${\omega}_{{\omega}_1}$. Since $\mathcal B$ clearly contains all the singletons of $D({\omega}_{{\omega}+1})$, it is enough to show that every $X\subset {\omega}_{{\omega}+1}$ is the union of $\omega_1$ many elements of $\mathcal B$.

Assume that $$ 1\Vdash \dot X\subset {\omega}_{{\omega}_1}. $$ For each ${\xi}\in {\omega}_{\omega_1}$ pick a maximal antichain $D_{\xi}$ in $P$ such that the element of $D_{\xi}$ decides if ${\xi}\in \dot X$ or not.

Define a function $f:{\omega}_{{\omega}_1}\to {\omega}_{{\omega}_1}$ in $V$ as follows: $$ f({\xi})=\min\{{\alpha}<\omega_1: D_{\xi}\subset Fn({\omega}_{\alpha},2)\}. $$ Let $$ Y_{\alpha}=\{{\xi}<{\omega}_{{\alpha}}:f({\xi})\le {\alpha}\}, $$ and let $\dot X_{\alpha}$ be an $Fn({\omega}_{\alpha},2)$-name of a subset of $Y_\alpha\subset {\omega}_{\alpha}$ such that for all ${\xi}\in Y_{\alpha}$ and for all $d\in D_{\xi}$ we have $$ d\Vdash {\xi}\in \dot X \text{ iff }d\Vdash {\xi}\in \dot X_{\alpha}. $$ and $$ d\Vdash {\xi}\notin \dot X \text{ iff }d\Vdash {\xi}\notin \dot X_{\alpha}. $$ Then in $V[G]$ we have $X\cap Y_{\alpha}=X_{\alpha}$ and $X_{\alpha}\in \mathcal B_{\alpha}$. Since $\bigcup\nolimits_{{\alpha}<{\omega}_1} Y_{\alpha}={\omega}_{\omega_1}$ we have

$$ V[G]\vDash X=\bigcup\nolimits_{{\alpha}<{\omega}_1} X_{\alpha}. $$

  • $\begingroup$ Very nice! In particular if a space $X$ ,e.g. $\beta N$, has a discrete subspace $Y$ with $|Y|=w(X)=2^{\omega}$ then $g(X)=g(D(2^{\omega}))$. It has also occurred to me to ask whether $g(X)$ can take singular values.Should I add this to my Q? $\endgroup$ Oct 23, 2015 at 17:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.