4
$\begingroup$

I asked the same question on MathStackExchange a month ago and received no answer. I feel that this would be more suitable for MathOverflow.

Compact open topology is one of the most common ways of turning the set of continuous maps between two topological spaces into a topological space. Suppose $X, Y$ are topological spaces, and let $K\subseteq X$ be a compact subset and $U\subseteq Y$ be an open subset. $$V_{K, U}=\{f: X\to Y \mid f(K)\subseteq U\}$$ Then the collection of all such $V_{K, U}$ is a subbase for the compact-open topology on $C(X, Y).$

A partial map $f: X\rightharpoonup Y$ is a function from some subset $X_f\subseteq X$ to $Y.$ To define continuity in this context, consider the totalization $$\widetilde{f}(x) = \begin{cases} f(x), & \text{if $x\in X_f$} \\ \star, & \text{if $x\notin X_f$} \end{cases}$$ where $\star\notin Y,$ and equip $\widetilde{Y}=Y\cup\{\star\}$ with the smallest topology extending that of $Y.$ i.e., $U\subseteq \widetilde{Y}$ is open iff $U=\widetilde{Y}$ or $U$ is open in $Y.$ Then claim $f: X\rightharpoonup Y$ is continuous iff $\widetilde{f}: X\to\widetilde{Y}$ is continuous. Observe that this definition doesn't alter the continuity of total maps.

Now, I want to understand how we can have a nice topology on the set of all continuous partial maps between $X$ and $Y.$ Since there is a bijection between $\{X\rightharpoonup Y\}$ and $\{X\to \widetilde{Y}\},$ I'm thinking that we can simply look at the compact-open topology on $C(X, \widetilde{Y}).$ But I'm not sure that this is the correct/best way to topologize the space of partial maps. I need the opinion of somebody who has some experience in this area.

  • Are there other ways of defining continuity for partial maps?
  • If so, how is this topology different from those?
  • Has anyone studied this problem before?
$\endgroup$
3
  • $\begingroup$ Another possibility, not sure how useful, would be to view partial maps as functional relations, those relations $R\colon X\to Y$ such that $R(a)$ has at most one element for each $a\in A$. Then we could apply some of the stuff that has been developed for relations between topological spaces, their continuity, and the spaces of maps between them. $\endgroup$
    – Emily
    Commented Dec 9, 2023 at 0:57
  • $\begingroup$ For example, the definitions of lower semicontinuous and upper semicontinuous relations here would lead to two other notions of continuity, which may or may not agree with the definition you gave (I haven't checked). We could also view relations as maps $\mathcal{P}(X)\to Y$, and use one of the topologies on $\mathrm{Hom}(\mathcal{P}(X),Y)$, choosing one for $\mathcal{P}(X)$ too (there are many choices) if needed. Then we use the subspace topology on the subspace of $\mathrm{Hom}(\mathcal{P}(X),Y)$ spanned by the partial maps/functional relations $\endgroup$
    – Emily
    Commented Dec 9, 2023 at 0:58
  • $\begingroup$ This should lead to a bunch of different notions and potential definitions/topologies, but I'm not sure how useful this point of view might be $\endgroup$
    – Emily
    Commented Dec 9, 2023 at 1:00

1 Answer 1

2
$\begingroup$

There is a way to promote function space topologies so that convergence of nets behaves the way that one would expect it to behave. There are probably other ways of getting a topology for spaces of partial functions.

Suppose that $X,Y$ are topological spaces. Let $\mathcal{C}\subseteq P(X)$. Suppose that $A_{R}\subseteq Y^X$ whenever $R\in\mathcal{C}$. Give each space $A_R$ a topology $\mathcal{T}_R$, and suppose that the restriction mapping $j_{R,S}:A_R\rightarrow A_S$ is continuous whenever $R,S\in\mathcal{C},S\subseteq R$. Let $\mathcal{T}$ be the topology on $\bigcup_{R\in\mathcal{C}}A_R$ where $U\in\mathcal{T}$ precisely when $U\cap A_R$ is open in $A_R$ for each $R\subseteq X$ and whenever $S\subseteq R$, we have $j_{R,S}^{-1}[U\cap A_S]\subseteq U\cap A_R$. The canonical basis for the topology on $\bigcup_{R\in\mathcal{C}}A_R$ consists of the sets of the form $\bigcup_{R\supseteq S}j_{R,S}^{-1}[U]$ where $U$ is open in $A_S$. Therefore, the topology $\mathcal{T}_R$ is the subspace topology inherited from $\mathcal{T}$.

Proposition: A net $(f_d)_{d\in D}$ in the space $\bigcup_{R\subseteq X}A_R$ converges to a function $f$ if and only if there exists some $d\in D$ where $\text{Dom}(f)\subseteq\text{Dom}(f_e)$ whenever $e\geq d$ and where the net $(f_e|_{\text{Dom}(f)})_{e\in D,e\geq d}$ converges to $f$ in the topology $(A_\text{Dom}(f),\mathcal{T}_\text{Dom}(f))$.

Proof:

$\rightarrow$ Suppose that $(f_d)_{d\in D}\rightarrow f$ in $\bigcup_{R\in\mathcal{C}}A_R$. Let $U=\bigcup_{R\in\mathcal{C},R\supseteq\text{Dom}(f)}A_R$. Then there exists some $d\in D$ where if $e\geq d$, then $f_e\in U$. But this implies that if $e\geq d$, then $\text{Dom}(f_e)\supseteq\text{Dom}(f)$. Suppose now that $O\in\mathcal{T}_{\text{Dom}(f)}$ and $f\in O$. Then $\bigcup_{R\subseteq\text{Dom}(f)}j_{R,\text{Dom}(f)}^{-1}[O]\in\mathcal{T}$ and $f\in\bigcup_{R\subseteq\text{Dom}(f)}j_{R,\text{Dom}(f)}^{-1}[O]$. Therefore, since $(f_d)_{d\in D}\rightarrow f$, there exists some $d_1\in D$ with $d_1\geq d$ where $f_e\in\bigcup_{R\subseteq\text{Dom}(f)}j_{R,\text{Dom}(f)}^{-1}[O]$ whenever $e\geq d_1$. However, since $f_e\in\bigcup_{R\subseteq\text{Dom}(f)}j_{R,\text{Dom}(f)}^{-1}[O]$, we know that $f_e|_{\text{Dom}(f)}=j_{\text{Dom}(f_e),\text{Dom}(f)}(f_e)\in O$. From this fact, we know that $(f_e|_{\text{Dom}(f)})_{e\geq d_1}$ converges to $f$ in the topology $(A_\text{Dom}(f),\mathcal{T}_\text{Dom}(f))$.

$\leftarrow.$ Suppose that $O\in\mathcal{T}$ and $f\in O$. Let $U=O\cap X_{\text{Dom}(f)}$. Then $U\in \mathcal{T}_{\text{Dom}(f)}$. Let $V=\bigcup_{R\supseteq\text{Dom}(f)}j_{R,\text{Dom}(f)}^{-1}[U]$. Then $V\in\mathcal{T}$ and $V\subseteq O$. Since there is some $d\in D$ where $\text{Dom}(f)\subseteq\text{Dom}(f_e)$ for $e\geq d$ and where $(f_e|_{\text{Dom}(f)})_{e\geq d}$ converges to $f$ in the topology $\mathcal{T}_{\text{Dom}(f)}$, we know that there is some $d_1\geq d$ where if $e\geq d_1$, then $f_e|_\text{Dom}(f)\in U$. But this implies that $f_e\in V\subseteq O$ whenever $e\geq d_1$. We conclude that $(f_d)_{d\in D}$ converges to $f$ in the space $\bigcup_{R\in\mathcal{C}}A_R$.

Q.E.D.

The topology $\mathcal{T}$ is not $T_1$ whenever there are $R,S\in\mathcal{C}$ with $R\subseteq S,R\neq S$. Let $\leq$ be the specialization ordering with respect to the topology $\mathcal{T}$. Then we observe that $f\geq f|_R$ whenever $R\in\mathcal{C},R\subseteq\text{Dom}(f)$. Since the specialization ordering is not the equality relation, we conclude that the topology $\mathcal{T}$ is not $T_1$.

If you are looking for a different topology on spaces of partial functions, one can always associate the partial functions with their graphs which are closed subsets of $X\times Y$, but we can give the collection of closed subsets of $X\times Y$ various topologies such as the Vietoris topology.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .