A topology is meant to record which subsets of a space are visible as neighbourhoods, but the axioms alone allow spaces whose open sets are far from metric intuition. The cofinite topology is one of the first useful tests of that intuition. It asks what happens if the only closed sets small enough to name are finite sets, while every set with finite complement is declared open. The surprising feature is that this topology is usually not discrete, not Hausdorff, and not metrizable, yet it still has strong compactness properties. It is therefore a compactness machine: it shows that compactness can come from having very few open sets, not only from boundedness or completeness. [example: Two Infinite Open Sets Must Meet] Let $X$ be an infinite set, and suppose $U,V\subset X$ have finite complements. For each $x\in X$, \begin{align*} x\in X\setminus (U\cap V) \Longleftrightarrow x\notin U\cap V \Longleftrightarrow (x\notin U)\text{ or }(x\notin V). \end{align*} The last condition is equivalent to $x\in (X\setminus U)\cup (X\setminus V)$, so \begin{align*} X\setminus (U\cap V)=(X\setminus U)\cup (X\setminus V). \end{align*} Since $X\setminus U$ and $X\setminus V$ are finite, their union is finite. Hence $X\setminus (U\cap V)$ is finite, so $U\cap V$ has finite complement. If $U\cap V=\varnothing$, then \begin{align*} X\setminus (U\cap V)=X\setminus \varnothing=X, \end{align*} which would make $X$ finite, contradicting the assumption that $X$ is infinite. Therefore $U\cap V\ne\varnothing$. This example shows the basic failure of Hausdorff intuition: in the cofinite topology on an infinite set, two nonempty open sets cannot be separated from each other. [/example] This single computation drives the whole chapter. Separation fails because open sets are too large; compactness holds because open covers cannot avoid covering all but finitely many points at once; convergence behaves strangely because to converge to a point, a sequence or net only needs eventually to enter every neighbourhood of that point, and each such neighbourhood contains a cofinite [open set](/page/Open%20Set). Here a net is indexed by a directed set $(D,\le)$: this means $\le$ is reflexive and transitive, and for any $d_1,d_2\in D$ there is some $e\in D$ with $d_1\le e$ and $d_2\le e$. This common-upper-bound condition is what makes phrases such as "eventually for all $d\ge d_0$" meaningful beyond ordinary sequences. ## Definition Before defining the topology, we should isolate the question it answers. Suppose we want a topology on a set $X$ in which finite sets are the only closed sets forced by point data. In a [metric space](/page/Metric%20Space), points are closed, but many infinite sets are closed as well. The cofinite topology keeps the condition that points are closed while refusing to distinguish most infinite subsets from the whole space. [definition: Cofinite Topology] Let $X$ be a set. The cofinite topology on $X$ is the collection \begin{align*} \tau_{\mathrm{cof}} = \{\varnothing\} \cup \{U \subset X : X \setminus U \text{ is finite}\}. \end{align*} The [topological space](/page/Topological%20Space) $(X,\tau_{\mathrm{cof}})$ is called a cofinite topological space. [/definition] The definition says that a nonempty open set is allowed to miss only finitely many points. On a finite set this is no restriction at all, because every complement is finite. On an infinite set it is a severe restriction: there are many subsets of $X$ which are neither open nor closed. The open-set definition is often easier to use for covers, while the closed-set form is better for separation and closure arguments. To use the topology effectively, we need to know exactly which subsets can be closed. The obstruction is that an infinite proper subset cannot be detected as closed: only finite exceptional sets, together with the whole space, survive as closed sets. [quotetheorem:8843] This theorem is the main translation device for the page. Whenever an argument mentions closure, density, or separation, it usually reduces to asking whether a finite set or the whole set is involved. The finite case is worth separating at the beginning because it prevents many false contrasts later. In finite spaces the cofinite topology does not create an exotic example; it gives the most detailed possible topology. [example: Finite Sets Become Discrete] Let $X=\{1,2,3\}$. To compute the cofinite topology, list the complements of all subsets of $X$: \begin{align*} X\setminus \varnothing=\{1,2,3\}. \end{align*} \begin{align*} X\setminus \{1\}=\{2,3\}, \qquad X\setminus \{2\}=\{1,3\}, \qquad X\setminus \{3\}=\{1,2\}. \end{align*} \begin{align*} X\setminus \{1,2\}=\{3\}, \qquad X\setminus \{1,3\}=\{2\}, \qquad X\setminus \{2,3\}=\{1\}. \end{align*} \begin{align*} X\setminus X=\varnothing. \end{align*} Each complement displayed is finite, because each is a subset of the finite set $X$. Therefore every subset of $X$ has finite complement, so every subset of $X$ is open in the cofinite topology: \begin{align*} \tau_{\mathrm{cof}}=\{\varnothing,\{1\},\{2\},\{3\},\{1,2\},\{1,3\},\{2,3\},X\}. \end{align*} Since the power set $\mathcal{P}(X)$ is exactly the collection of all subsets of $X$, we have \begin{align*} \tau_{\mathrm{cof}}=\mathcal{P}(X). \end{align*} Thus on this finite set the cofinite topology is the [discrete topology](/page/Discrete%20Topology), so the unusual behaviour of the cofinite topology depends on $X$ being infinite. [/example] The infinite case is the real model. It is the smallest topology among those in which every finite set is closed, and it is a canonical example of a topology determined by a size condition rather than by distances. ## Open Sets and Closed Sets ### Large Open Sets The cofinite topology is governed by a simple principle: a nonempty open set is large. This raises the basic consistency check for the definition: if two allowed open sets each miss only finitely many points, their intersection must still miss only finitely many points, while unions must not introduce any new obstruction. That finite-exception calculation is the reason the proposed collection is a topology. [quotetheorem:8844] The finite-intersection axiom is the only point where the word finite matters. Intersecting finitely many cofinite sets removes only finitely many points. Arbitrary intersections are not part of the topology axioms, and in fact arbitrary intersections of cofinite open sets need not be open. To speak efficiently about neighbourhoods, we need the local version of openness. In a metric space a neighbourhood can be very small around a point. In a cofinite space every neighbourhood of a point contains almost every point of the space, so local information is already global information. [definition: Neighbourhood] Let $(X,\tau)$ be a topological space and let $x \in X$. A subset $N \subset X$ is a neighbourhood of $x$ if there exists an open set $U \in \tau$ such that \begin{align*} x \in U \subset N. \end{align*} [/definition] In the cofinite topology, every neighbourhood of a point contains a cofinite open set. Thus a statement that holds eventually inside every neighbourhood of $x$ is a statement that survives after removing any prescribed finite set of exceptions. [example: Neighbourhoods on $\mathbb{N}$] Give $\mathbb{N}$ the cofinite topology, and set \begin{align*} U=\mathbb{N}\setminus\{2,5,9\}. \end{align*} Its complement is \begin{align*} \mathbb{N}\setminus U=\mathbb{N}\setminus(\mathbb{N}\setminus\{2,5,9\})=\{2,5,9\}, \end{align*} which is finite, so $U$ is open in the cofinite topology. Since $1\notin\{2,5,9\}$, we have $1\in U$, and therefore $U$ is an open neighbourhood of $1$. More generally, if $m\in U$, then the same open set satisfies \begin{align*} m\in U\subset U, \end{align*} so $U$ is a neighbourhood of every point $m\in U$. Now consider $A=\{1,3,7\}$. If $A$ were a neighbourhood of $1$, there would be an open set $W$ such that \begin{align*} 1\in W\subset A. \end{align*} Because $1\in W$, the set $W$ is nonempty, so in the cofinite topology its complement $\mathbb{N}\setminus W$ is finite. From $W\subset A$ we get \begin{align*} \mathbb{N}\setminus A\subset \mathbb{N}\setminus W. \end{align*} But \begin{align*} \mathbb{N}\setminus A=\mathbb{N}\setminus\{1,3,7\} \end{align*} is infinite, since it contains every integer $n\ge 8$. An infinite set cannot be contained in a finite set, contradicting that $\mathbb{N}\setminus W$ is finite. Hence $\{1,3,7\}$ is not a neighbourhood of $1$. This shows that in the cofinite topology, neighbourhoods of a point must still contain almost all natural numbers, not merely a few nearby or selected points. [/example] ### Closure and Density Closure is where the topology first becomes counterintuitive. In Euclidean spaces, infinite sets may still be sparse and have small closure. To make that contrast precise, we need the topological operation that enlarges a subset by adding all points that cannot be separated from it by closed containment. In the cofinite topology this operation will distinguish finite subsets from infinite ones. [definition: Closure] Let $(X,\tau)$ be a topological space and let $A \subset X$. The closure of $A$, denoted $\overline{A}$, is the intersection of all closed subsets of $X$ that contain $A$. [/definition] This definition is formal, but in a cofinite space it becomes a dichotomy. Finite sets remain closed; infinite sets touch every nonempty open set and therefore fill the whole space after closure. [quotetheorem:8845] The theorem explains why density is cheap in this topology. Any infinite subset is dense, even if it looks arithmetically thin or scattered. [example: The Even Numbers Are Dense] Give $\mathbb{N}$ the cofinite topology and let \begin{align*} E=\{2n:n\in\mathbb{N}\}. \end{align*} The set $E$ is infinite, since the map $\mathbb{N}\to E$ defined by $n\mapsto 2n$ is injective: if $2m=2n$, then subtracting $2n$ from both sides gives $2(m-n)=0$, hence $m-n=0$, so $m=n$. By the *[Closure Dichotomy in the Cofinite Topology](/theorems/8845)*, every infinite subset of an infinite cofinite space has closure equal to the whole space. Applying this to $E\subset\mathbb{N}$ gives \begin{align*} \overline{E}=\mathbb{N}. \end{align*} Equivalently, let $U\subset\mathbb{N}$ be a nonempty open set. Then $\mathbb{N}\setminus U$ is finite. If $U\cap E=\varnothing$, then every even number lies outside $U$, so \begin{align*} E\subset \mathbb{N}\setminus U. \end{align*} This would put the infinite set $E$ inside the finite set $\mathbb{N}\setminus U$, impossible. Therefore $U\cap E\ne\varnothing$. Thus the even numbers are dense because no nonempty cofinite open set can remove all of them. [/example] ## Separation and Metrizability ### Hausdorff Failure Metric spaces let distinct points be pulled apart by disjoint balls. The cofinite topology on an infinite set prevents this from the start: every two nonempty open sets meet. This makes it a compact example that violates the strongest separation properties usually attached to analysis. To measure exactly what fails, we recall the Hausdorff condition. It formalizes the idea that distinct points have independent local worlds. [definition: Hausdorff Space] A topological space $(X,\tau)$ is Hausdorff if for every pair of distinct points $x,y \in X$, there exist open sets $U,V \in \tau$ such that \begin{align*} x \in U, \qquad y \in V, \qquad U \cap V = \varnothing. \end{align*} [/definition] In an infinite cofinite space, the first example of the page says that no two nonempty open sets are disjoint. Since neighbourhoods of points must be nonempty, this leaves no room for Hausdorff separation. The next theorem records the exact obstruction so that later non-metrizability will follow from a standard implication. [quotetheorem:8846] This failure should not be confused with total lack of separation. Points are still closed, because singleton sets are finite. The cofinite topology is therefore a standard example of a space that satisfies the $T_1$ condition but need not be Hausdorff. ### The $T_1$ Boundary The $T_1$ condition isolates exactly the property that single points are topologically detectable. It is weaker than Hausdorffness but strong enough to make finite sets closed. [definition: $T_1$ Space] A topological space $(X,\tau)$ is a $T_1$ space if for every pair of distinct points $x,y \in X$, there exists an open set $U \in \tau$ such that \begin{align*} x \in U, \qquad y \notin U. \end{align*} [/definition] For cofinite spaces, separating one point from another is possible by removing the unwanted point. This is the positive separation property that survives after Hausdorffness fails, and it explains why finite sets behave like closed sets even in a non-[Hausdorff space](/page/Hausdorff%20Space). We need this theorem to locate the cofinite topology precisely between the indiscrete and Hausdorff worlds. [quotetheorem:8847] The contrast between $T_1$ and Hausdorff is one of the main reasons the cofinite topology belongs in every first topology course. It separates two conditions that often appear together in metric examples. [example: Separating One Way but Not Two Ways] Let $X=\mathbb{N}$ with the cofinite topology. To separate $1$ from $2$ in the $T_1$ sense, take \begin{align*} U=\mathbb{N}\setminus\{2\}. \end{align*} Since $1\ne 2$, we have $1\notin\{2\}$, so $1\in \mathbb{N}\setminus\{2\}=U$. Also $2\in\{2\}$, so $2\notin \mathbb{N}\setminus\{2\}=U$. Thus this single open set contains $1$ while excluding $2$. It is open because \begin{align*} \mathbb{N}\setminus U=\mathbb{N}\setminus(\mathbb{N}\setminus\{2\})=\{2\}, \end{align*} and $\{2\}$ is finite. Now let $V$ be any open set containing $2$. Since $2\in V$, the set $V$ is nonempty, so in the cofinite topology its complement $\mathbb{N}\setminus V$ is finite. The complement of $U\cap V$ is \begin{align*} \mathbb{N}\setminus(U\cap V)=(\mathbb{N}\setminus U)\cup(\mathbb{N}\setminus V)=\{2\}\cup(\mathbb{N}\setminus V). \end{align*} The set $\{2\}$ is finite and $\mathbb{N}\setminus V$ is finite, so $\mathbb{N}\setminus(U\cap V)$ is finite. If $U\cap V=\varnothing$, then \begin{align*} \mathbb{N}\setminus(U\cap V)=\mathbb{N}\setminus\varnothing=\mathbb{N}, \end{align*} which would make $\mathbb{N}$ finite, a contradiction. Therefore $U\cap V\ne\varnothing$. So the cofinite topology can separate $1$ from $2$ in the one-sided $T_1$ sense, but it cannot provide disjoint open neighbourhoods separating them in the Hausdorff sense. [/example] ### No Metric Model Metrizability now fails for a structural reason. Every metric space is Hausdorff, so an infinite cofinite space with at least two points cannot come from any metric. [definition: Metrizable Space] A topological space $(X,\tau)$ is metrizable if there exists a metric $d:X\times X\to [0,\infty)$ such that $\tau$ is the topology generated by the open balls \begin{align*} B(x_0,r)=\{x\in X:d(x,x_0)<r\}. \end{align*} [/definition] This definition connects topology back to analysis. If a cofinite topology on an infinite set came from a metric, then the metric would provide disjoint balls around distinct points. The next theorem turns the previous Hausdorff failure into the promised obstruction to any distance model. [quotetheorem:8848] The obstruction is not size alone. An infinite set can carry many metrizable topologies, including the discrete topology. The obstruction is the particular pattern of large open sets. ## Compactness The cofinite topology is most useful as a compactness example. An [open cover](/page/Open%20Cover) of an infinite cofinite space may contain many sets, but once it contains a single nonempty open set, only finitely many points remain uncovered. Those leftover points can then be handled one at a time. We first recall the covering definition. Compactness is stated in terms of arbitrary open covers, not sequences, because sequence-based compactness can behave differently outside metric spaces. [definition: Compact Space] A topological space $(X,\tau)$ is compact if for every collection $\mathcal{U}\subset \tau$ such that \begin{align*} X = \bigcup_{U\in \mathcal{U}} U, \end{align*} there exists a finite subcollection $\mathcal{V}\subset \mathcal{U}$ such that \begin{align*} X = \bigcup_{V\in \mathcal{V}} V. \end{align*} [/definition] In Euclidean space, compactness often appears as closedness plus boundedness. The cofinite topology gives a different mechanism: every nonempty open set is almost the whole space. This suggests that an arbitrary open cover should become finite after choosing one nonempty member and then covering the few missed points, so the next theorem is the central compactness result for the chapter. [quotetheorem:8849] This theorem is a compactness proof template. Choose one nonempty member of the cover; its complement is finite; cover those finitely many missing points with finitely many additional members of the cover. [example: A Compact Non-Hausdorff Space] Let $X=\mathbb{N}$ with the cofinite topology. We first verify compactness from the open-cover definition. Let $\mathcal{U}$ be an open cover of $\mathbb{N}$, so \begin{align*} \mathbb{N}=\bigcup_{U\in\mathcal{U}}U. \end{align*} Since $1\in\mathbb{N}$, some $U_0\in\mathcal{U}$ satisfies $1\in U_0$. Thus $U_0\ne\varnothing$, and by the definition of the cofinite topology the complement $\mathbb{N}\setminus U_0$ is finite. Write \begin{align*} \mathbb{N}\setminus U_0=\{a_1,\ldots,a_m\} \end{align*} for some $m\ge 0$, where the case $m=0$ means the complement is empty. For each $a_i$, the cover property gives a set $U_i\in\mathcal{U}$ such that $a_i\in U_i$. Then every $n\in\mathbb{N}$ either lies in $U_0$ or lies in $\mathbb{N}\setminus U_0=\{a_1,\ldots,a_m\}$, in which case $n=a_i$ for some $i$ and hence $n\in U_i$. Therefore \begin{align*} \mathbb{N}=U_0\cup U_1\cup\cdots\cup U_m. \end{align*} This is a finite subcover, so $\mathbb{N}$ is compact. Now take the two distinct points $1,2\in\mathbb{N}$. Suppose there were open sets $U,V$ with \begin{align*} 1\in U,\qquad 2\in V,\qquad U\cap V=\varnothing. \end{align*} Because $1\in U$ and $2\in V$, both $U$ and $V$ are nonempty, so $\mathbb{N}\setminus U$ and $\mathbb{N}\setminus V$ are finite. For any $n\in\mathbb{N}$, \begin{align*} n\in \mathbb{N}\setminus(U\cap V)\Longleftrightarrow n\notin U\cap V\Longleftrightarrow n\notin U\text{ or }n\notin V. \end{align*} Hence \begin{align*} \mathbb{N}\setminus(U\cap V)=(\mathbb{N}\setminus U)\cup(\mathbb{N}\setminus V). \end{align*} The right-hand side is finite, so $\mathbb{N}\setminus(U\cap V)$ is finite. But $U\cap V=\varnothing$ would give \begin{align*} \mathbb{N}\setminus(U\cap V)=\mathbb{N}\setminus\varnothing=\mathbb{N}, \end{align*} which is infinite. This contradiction shows that no disjoint open neighbourhoods separate $1$ and $2$, so the space is not Hausdorff. Thus the cofinite topology on $\mathbb{N}$ gives a [compact space](/page/Compact%20Space) where metric-style separation fails. [/example] The inclusion relation with the discrete topology is simpler: every cofinite-open subset is still a subset of $X$, hence open in the discrete topology, so the cofinite topology is coarser than the discrete topology. Compactness explains the consequence of this coarsening. A discrete topology on $\mathbb{N}$ is not compact, because the cover by singletons has no finite subcover; the cofinite topology removes enough open sets to force compactness. Compact subsets of a cofinite space have a mixed character. The whole space is compact, all finite subsets are compact, and subspaces should not depend on the ambient set in a mysterious way. The next theorem verifies that restricting the topology to a subset simply recreates the same cofinite construction on that subset. [quotetheorem:8850] This result says that the cofinite topology is stable under passage to subsets. Unlike many topological constructions, it does not depend on a surrounding metric or ambient geometry. [example: The Odd Numbers as a Subspace] Let $X=\mathbb{N}$ have the cofinite topology, and let \begin{align*} Y=\{2n-1:n\in\mathbb{N}\}. \end{align*} We show that the [subspace topology](/page/Subspace%20Topology) on $Y$ is exactly the cofinite topology on $Y$. First let $O$ be open in the subspace topology on $Y$. Then $O=Y\cap U$ for some open set $U\subset X$. If $O=\varnothing$, then $O$ is open in the cofinite topology on $Y$ by definition. If $O\ne\varnothing$, then $U\ne\varnothing$, so $X\setminus U$ is finite. For each $y\in Y$, \begin{align*} y\in Y\setminus(Y\cap U)\Longleftrightarrow y\in Y\text{ and }y\notin U. \end{align*} Thus \begin{align*} Y\setminus(Y\cap U)=Y\cap(X\setminus U). \end{align*} Since $X\setminus U$ is finite, its subset $Y\cap(X\setminus U)$ is finite. Hence $O$ has finite complement in $Y$. Conversely, let $O\subset Y$ have finite complement in $Y$. Put \begin{align*} F=Y\setminus O. \end{align*} Then $F$ is finite, and since $F\subset Y\subset X$, the set \begin{align*} U=X\setminus F \end{align*} is open in the cofinite topology on $X$. Now \begin{align*} Y\cap U=Y\cap(X\setminus F)=Y\setminus F=Y\setminus(Y\setminus O)=O, \end{align*} where the last equality uses $O\subset Y$. Therefore $O$ is open in the subspace topology. The open subsets of $Y$ are exactly $\varnothing$ together with the subsets of $Y$ whose complements in $Y$ are finite, so the odd numbers inherit the cofinite topology. [/example] ## Convergence ### Sequential Behaviour Convergence in a topology is controlled by neighbourhoods. Since neighbourhoods in a cofinite space are large, convergence is correspondingly weak. A sequence may converge to many points, and a net may be the more natural object for expressing closure. We begin with the familiar sequential definition. It is enough to reveal the main pathology, even though general topological spaces require nets for the full closure theory. [definition: Sequence Convergence] Let $(X,\tau)$ be a topological space. A sequence in $X$ is a function \begin{align*} x:\mathbb{N}\to X. \end{align*} We write $x_n$ for the value of this function at $n$. The sequence $(x_n)_{n\in\mathbb{N}}$ converges to $x\in X$ if for every neighbourhood $N$ of $x$, there exists $n_0\in\mathbb{N}$ such that for all $n\ge n_0$, \begin{align*} x_n\in N. \end{align*} [/definition] In a Hausdorff space, convergent sequences have unique limits. The cofinite topology breaks that conclusion because the Hausdorff hypothesis is missing. The relevant sequences are those that eventually avoid each finite obstruction; every cofinite neighbourhood contains all terms after such an obstruction has been escaped. [quotetheorem:8851] This criterion is more precise than saying that convergent sequences must escape every finite set. A sequence may visit its proposed limit value infinitely often; what convergence forbids is infinitely many visits to any other single point. For applications, however, it is useful to have a condition that does not name a proposed limit in advance. If a sequence eventually avoids each finite subset of $X$, then no point can serve as an obstruction to convergence toward any other point. This gives a clean sufficient condition for universal convergence, and in a countably infinite set an enumeration without repetition has exactly this property. [quotetheorem:8852] [example: The Sequence $n$ Converges to Every Natural Number] Give $\mathbb{N}$ the cofinite topology and consider the sequence $x_n=n$. We show that this sequence converges to every $k\in\mathbb{N}$. Fix $k\in\mathbb{N}$, and let $M$ be a neighbourhood of $k$. By the definition of neighbourhood, there is an open set $U$ such that \begin{align*} k\in U\subset M. \end{align*} Since $k\in U$, the set $U$ is nonempty. In the cofinite topology, every nonempty open set has finite complement, so \begin{align*} F=\mathbb{N}\setminus U \end{align*} is finite. If $F=\varnothing$, then every $n\in\mathbb{N}$ lies in $U$. If $F\ne\varnothing$, let $m=\max F$, which exists because $F$ is a nonempty finite subset of $\mathbb{N}$. Put $n_0=m+1$. Then for every $n\ge n_0$ we have $n>m$, so $n\notin F$. Therefore \begin{align*} x_n=n\in \mathbb{N}\setminus F=U\subset M. \end{align*} Thus for every neighbourhood $M$ of $k$, all sufficiently large terms of the sequence lie in $M$, so $x_n\to k$. Since $k\in\mathbb{N}$ was arbitrary, the sequence $x_n=n$ converges to every natural number in the cofinite topology. [/example] This example is not a defect in the definition of convergence. It is evidence that [uniqueness of limits](/theorems/625) is a separation theorem, not a built-in feature of convergence. ### Nets and Finite Avoidance Sequences do not describe closure well in every topological space, so general topology uses directed index sets. The cofinite topology makes this need visible on arbitrary sets, where there may be no countable enumeration that visits an infinite subset in the right way. Nets provide the replacement object: an indexed family flexible enough to be ordered by finite obstructions. [definition: Net] Let $(D,\le)$ be a directed set and let $X$ be a set. A net in $X$ indexed by $D$ is a function \begin{align*} x:D&\to X. \end{align*} [/definition] A net is a generalized sequence whose index set can be adapted to the neighbourhood structure of the space. To use nets topologically, we need the analogue of eventually lying in every neighbourhood. This definition is what lets finite avoidance become a convergence statement in cofinite spaces. [definition: Net Convergence] Let $(X,\tau)$ be a topological space, let $(D,\le)$ be a directed set, and let $(x_d)_{d\in D}$ be a net in $X$. The net converges to $x\in X$ if for every neighbourhood $N$ of $x$, there exists $d_0\in D$ such that for all $d\ge d_0$, \begin{align*} x_d\in N. \end{align*} [/definition] The cofinite topology turns finite avoidance into convergence. For sequences this principle can be too small when $X$ is uncountable, because a sequence may not be able to range through enough of the space while avoiding every prescribed finite obstruction. Nets fix that mismatch by allowing the index set to be tailored to the finite subsets being avoided, so the same idea becomes fully topological. [quotetheorem:8853] This theorem is the net version of the sequence result. It is also the right way to understand why every infinite subset is dense: a net can move through the subset while avoiding any prescribed finite obstruction. ## Continuous Maps ### Maps Into Cofinite Spaces The cofinite topology is also useful for testing continuity. A map into a cofinite space is continuous exactly when preimages of finite closed sets are closed. A map out of a cofinite space often has to be nearly constant when the target is Hausdorff. We start with continuity in its topological form. The definition uses preimages of open sets because this formulation behaves well without metrics. [definition: Continuous Map] Let $(X,\tau_X)$ and $(Y,\tau_Y)$ be topological spaces. A function $f:X\to Y$ is continuous if for every open set $V\in\tau_Y$, the preimage \begin{align*} f^{-1}(V)=\{x\in X:f(x)\in V\} \end{align*} is open in $X$. [/definition] When the target has the cofinite topology, continuity can be checked against finite complements. Since closed sets in the target are finite or all of $Y$, continuity should amount to asking whether finite fibers and finite unions of fibers are closed in the domain. The next theorem gives this practical test. [quotetheorem:8854] The theorem is often most useful when $F$ is a singleton. If the domain is already $T_1$ and the fibers of $f$ are finite or closed, the finite-set condition may be easy to verify. [example: Injective Maps from $\mathbb{R}$ to a Cofinite Set] Let $Y$ have the cofinite topology, let $\mathbb{R}$ have its usual topology, and let $f:\mathbb{R}\to Y$ be injective. We show that the preimage of every open subset of $Y$ is open in $\mathbb{R}$. Let $V\subset Y$ be open. If $V=\varnothing$, then \begin{align*} f^{-1}(V)=f^{-1}(\varnothing)=\varnothing, \end{align*} which is open in $\mathbb{R}$. If $V\ne\varnothing$, then the cofinite topology gives that \begin{align*} F=Y\setminus V \end{align*} is finite. Write $F=\{y_1,\ldots,y_m\}$ for some $m\ge 0$. Since $f$ is injective, each fiber $f^{-1}(\{y_i\})$ contains at most one real number: if $a,b\in f^{-1}(\{y_i\})$, then $f(a)=y_i=f(b)$, so injectivity gives $a=b$. Therefore \begin{align*} f^{-1}(F)=f^{-1}(\{y_1,\ldots,y_m\})=f^{-1}(\{y_1\})\cup\cdots\cup f^{-1}(\{y_m\}) \end{align*} is finite, because it is a finite union of sets with at most one element. Now \begin{align*} f^{-1}(V)=f^{-1}(Y\setminus F)=\mathbb{R}\setminus f^{-1}(F). \end{align*} The set $f^{-1}(F)$ is finite, and every finite subset of $\mathbb{R}$ is closed in the usual topology: each singleton $\{a\}$ is closed because $\mathbb{R}\setminus\{a\}=(-\infty,a)\cup(a,\infty)$ is open, and a finite union of closed sets is closed. Hence $\mathbb{R}\setminus f^{-1}(F)$ is open, so $f^{-1}(V)$ is open. Thus every injective map from $\mathbb{R}$ with its usual topology into a cofinite space is continuous. [/example] ### Maps Out of Cofinite Spaces Maps in the opposite direction are rigid only after the target supplies extra information about its compact subsets. The real mechanism is compactness: a continuous image of a compact cofinite space is compact. Hausdorffness by itself does not rule out an infinite image; it only places that image inside the ordinary theory of compact subspaces of the target. The next theorem isolates the useful consequence for targets whose compact subsets are small. [quotetheorem:305] A sharper statement holds for maps into spaces where every value is isolated. To formulate that target condition, we need the topology in which all subsets are open. This is the topology that turns a function into a collection of open fibers, one fiber for each value. [definition: Discrete Topology] Let $X$ be a set. The discrete topology on $X$ is the topology \begin{align*} \mathcal{P}(X)=\{U:U\subset X\}. \end{align*} [/definition] The natural question is whether an infinite cofinite space can carry a nonconstant continuous label with isolated values. Such a label would split the domain into open fibers over distinct target points, so the theorem tests the strongest possible obstruction: discrete targets force every attempted separation to be visible as an open subset. [quotetheorem:8855] This theorem is a practical warning: cofinite domains are connected in a strong qualitative sense. They cannot continuously label infinitely many points by isolated values. [example: No Nonconstant Integer-Valued Continuous Functions] Let $X$ be an infinite set with the cofinite topology, and give $\mathbb{Z}$ the discrete topology. We show that every [continuous function](/page/Continuous%20Function) $f:X\to\mathbb{Z}$ is constant. Suppose, toward a contradiction, that $f$ is not constant. Then there exist $x_1,x_2\in X$ such that $f(x_1)\ne f(x_2)$. Set \begin{align*} a=f(x_1),\qquad b=f(x_2). \end{align*} Since $\mathbb{Z}$ has the discrete topology, the singleton sets $\{a\}$ and $\{b\}$ are open in $\mathbb{Z}$. By continuity of $f$, their preimages \begin{align*} A=f^{-1}(\{a\}),\qquad B=f^{-1}(\{b\}) \end{align*} are open in $X$. They are nonempty because $x_1\in A$ and $x_2\in B$. The sets $A$ and $B$ are disjoint. Indeed, if $x\in A\cap B$, then $f(x)=a$ and $f(x)=b$, so $a=b$, contradicting $a\ne b$. Thus \begin{align*} A\cap B=\varnothing. \end{align*} Since $A$ and $B$ are nonempty open sets in the cofinite topology, both complements $X\setminus A$ and $X\setminus B$ are finite. For each $x\in X$, \begin{align*} x\in X\setminus(A\cap B)\Longleftrightarrow x\notin A\cap B\Longleftrightarrow x\notin A\text{ or }x\notin B. \end{align*} Hence \begin{align*} X\setminus(A\cap B)=(X\setminus A)\cup(X\setminus B). \end{align*} The right-hand side is finite, so $X\setminus(A\cap B)$ is finite. But $A\cap B=\varnothing$, so \begin{align*} X\setminus(A\cap B)=X\setminus\varnothing=X. \end{align*} This would make $X$ finite, contradicting the assumption that $X$ is infinite. Therefore no two points of $X$ can have different integer values under $f$, so $f$ is constant. Integer-valued continuous functions on an infinite cofinite space cannot distinguish points. [/example] ## Connectedness and Irreducibility A space is connected when it cannot be split into two disjoint nonempty open pieces. Since nonempty open sets in an infinite cofinite space always meet, connectedness comes for free. This makes cofinite spaces useful as examples where compactness and connectedness coexist without Hausdorffness. We recall the definition in the form most suited to the cofinite topology: a disconnection is a partition into two open pieces. [definition: Connected Space] A topological space $(X,\tau)$ is connected if there do not exist nonempty open sets $U,V\in\tau$ such that \begin{align*} X=U\cup V, \qquad U\cap V=\varnothing. \end{align*} [/definition] The obstruction to disconnection is exactly the same computation that obstructs Hausdorffness: two nonempty cofinite open sets intersect. Since a disconnection would require two such disjoint sets whose union is the whole space, the cofinite topology gives connectedness without any path structure or metric intervals. [quotetheorem:8856] Connectedness only forbids a two-piece open separation. A stronger condition, useful in algebraic geometry and point-set topology, requires every two nonempty open sets to meet. [definition: Hyperconnected Space] A topological space $(X,\tau)$ is hyperconnected if every pair of nonempty open sets $U,V\in\tau$ satisfies \begin{align*} U\cap V\ne\varnothing. \end{align*} [/definition] The cofinite topology on an infinite set is the basic model of this phenomenon. It is not merely connected; its open sets overlap so much that no two nonempty open observations can be independent. The next theorem names this stronger conclusion so it can be used to explain the collapse of maps into discrete targets. [quotetheorem:8857] Hyperconnectedness helps explain why continuous maps from cofinite spaces into discrete spaces collapse. The preimages of two distinct discrete values would be disjoint open sets, which the topology does not allow. [example: A Connected Space with Many Closed Points] Let $X=\mathbb{N}$ with the cofinite topology. For each $n\in\mathbb{N}$, the complement of the singleton $\{n\}$ is \begin{align*} \mathbb{N}\setminus\{n\}. \end{align*} Since $\{n\}$ is finite, the set $\mathbb{N}\setminus(\mathbb{N}\setminus\{n\})=\{n\}$ is finite, so $\mathbb{N}\setminus\{n\}$ is open in the cofinite topology. Therefore $\{n\}$ is closed for every $n\in\mathbb{N}$. We now show that $X$ is connected. Let $U,V\subset\mathbb{N}$ be nonempty open sets. Since $U$ and $V$ are nonempty and open in the cofinite topology, both complements $\mathbb{N}\setminus U$ and $\mathbb{N}\setminus V$ are finite. For any $m\in\mathbb{N}$, \begin{align*} m\in \mathbb{N}\setminus(U\cap V)\Longleftrightarrow m\notin U\cap V\Longleftrightarrow m\notin U\text{ or }m\notin V. \end{align*} Thus \begin{align*} \mathbb{N}\setminus(U\cap V)=(\mathbb{N}\setminus U)\cup(\mathbb{N}\setminus V). \end{align*} The right-hand side is a union of two finite sets, hence finite. If $U\cap V=\varnothing$, then \begin{align*} \mathbb{N}\setminus(U\cap V)=\mathbb{N}\setminus\varnothing=\mathbb{N}, \end{align*} which is infinite, contradicting the finiteness just proved. Hence every two nonempty open sets meet, so no separation of $\mathbb{N}$ into two disjoint nonempty open sets exists. The space is connected even though every point is closed, so connectedness does not mean that individual points fail to be closed. [/example] ## Comparison with Related Topologies The cofinite topology belongs to a family of topologies defined by declaring complements small. Comparing it with the discrete, indiscrete, and cocountable topologies shows which properties come from finiteness and which come from largeness more generally. ### Extremal Topologies The coarsest possible topology is useful as a boundary case. It makes every function into the space hard to be continuous, and every function out of the space easy to be continuous only under special target conditions. [definition: Indiscrete Topology] Let $X$ be a set. The indiscrete topology on $X$ is \begin{align*} \{\varnothing,X\}. \end{align*} [/definition] The indiscrete topology is usually not $T_1$ when $X$ has more than one point, while the cofinite topology is always $T_1$. This raises a universal question: once points are required to be closed, how many open sets are forced? The answer is exactly the cofinite topology. [quotetheorem:8858] This result gives the cofinite topology its universal role. It is not an arbitrary example; it is the least topology compatible with the requirement that points be closed. ### Countable Complements Another useful comparison replaces finite complements by countable complements. This asks what changes when the topology is still built from large open sets, but countable exceptional sets are now allowed. The resulting topology behaves similarly on uncountable sets while changing countability and convergence phenomena. [definition: Cocountable Topology] Let $X$ be a set. The cocountable topology on $X$ is \begin{align*} \tau_{\mathrm{coc}}=\{\varnothing\}\cup\{U\subset X:X\setminus U\text{ is countable}\}. \end{align*} [/definition] The cocountable topology is finer than the cofinite topology because every finite set is countable. On an uncountable set it is still far from discrete, but it makes countable subsets closed and changes the behaviour of sequences. [example: Cofinite versus Cocountable on $\mathbb{R}$] Let $U\subset\mathbb{R}$ be open in the cofinite topology. If $U=\varnothing$, then $U$ is also open in the cocountable topology by definition. If $U\ne\varnothing$, then $\mathbb{R}\setminus U$ is finite. Every finite set is countable, so $\mathbb{R}\setminus U$ is countable, and therefore $U$ is open in the cocountable topology. Hence every cofinite open subset of $\mathbb{R}$ is cocountable open. Now consider \begin{align*} A=\mathbb{R}\setminus\mathbb{Q}. \end{align*} Its complement in $\mathbb{R}$ is \begin{align*} \mathbb{R}\setminus A=\mathbb{R}\setminus(\mathbb{R}\setminus\mathbb{Q})=\mathbb{Q}. \end{align*} The set $\mathbb{Q}$ is countable: every rational number has the form $p/q$ with $p\in\mathbb{Z}$ and $q\in\mathbb{N}$, $q\ne 0$, so $\mathbb{Q}$ is the image of the [countable set](/page/Countable%20Set) $\mathbb{Z}\times(\mathbb{N}\setminus\{0\})$ under $(p,q)\mapsto p/q$. Therefore $A$ is open in the cocountable topology. But $\mathbb{Q}$ is infinite, since the map $\mathbb{N}\to\mathbb{Q}$ given by $n\mapsto n$ is injective. Thus $\mathbb{R}\setminus A=\mathbb{Q}$ is not finite, so $A$ is not open in the cofinite topology. Therefore the cocountable topology on $\mathbb{R}$ is strictly finer than the cofinite topology. [/example] The comparison with the discrete topology shows how compactness can be created by removing open sets. The comparison with the indiscrete topology shows how $T_1$ separation can be added without reaching Hausdorffness. The comparison with the cocountable topology shows how the size notion in the complement controls convergence and closure. ## Beyond and Connected Topics The cofinite topology is a gateway into general topology because it demonstrates that metric intuition is only a special case. The natural next topic is the theory of separation axioms, especially the distinction between $T_1$, Hausdorff, regular, and normal spaces. Cofinite spaces show why these conditions must be stated separately. Compactness is another continuation. In metric spaces, compactness is often taught through sequences, but the cofinite topology pushes the reader toward open covers, subspaces, and continuous images. This connects directly with the compactness material in [Cambridge IB Analysis and Topology](/page/Cambridge%20IB%20Analysis%20and%20Topology). Algebraic topology often assumes Hausdorff or locally nice hypotheses only when they are needed. Cofinite spaces are useful warning examples when studying quotient spaces and weak separation assumptions, especially before moving into [Cambridge II Algebraic Topology](/page/Cambridge%20II%20Algebraic%20Topology) and [Cambridge III Algebraic Topology](/page/Cambridge%20III%20Algebraic%20Topology). The cofinite topology also points toward the Zariski topology in algebraic geometry. In the affine line over an [algebraically closed field](/page/Algebraically%20Closed%20Field), the closed proper algebraic subsets are finite, so the Zariski topology on the line is cofinite. This connection explains why non-Hausdorff topologies can encode serious geometric information rather than merely serving as pathological examples. Finally, the page connects back to foundational analysis through [Cambridge IA Analysis Notes](/page/Cambridge%20IA%20Analysis%20Notes). The contrast between usual convergence on $\mathbb{R}$ and cofinite convergence clarifies which theorems depend on metrics, first countability, and Hausdorffness. ## References Androma, [Cambridge IB Analysis and Topology](/page/Cambridge%20IB%20Analysis%20and%20Topology). Androma, [Cambridge II Algebraic Topology](/page/Cambridge%20II%20Algebraic%20Topology). Androma, [Cambridge III Algebraic Topology](/page/Cambridge%20III%20Algebraic%20Topology). Androma, [Cambridge IA Analysis Notes](/page/Cambridge%20IA%20Analysis%20Notes). James R. Munkres, *Topology* (2000). John L. Kelley, *General Topology* (1955). Stephen Willard, *General Topology* (1970).