π-base: Difficulty 5

Difficulty: 5

For proofs that take a little bit of time. Took me some time thinking, probably had to look at a hint or glance at a proof.

69

CGWH     \implies k3k_3-space

Added:

Mar 23, 2026

Difficulty:

Nice. Let AA be a set such that for any KXK \subseteq X compact and T2T_2, AKA \cap K is open in KK. It suffices to show AA is open in XX. We wish to prove so using the k2k_2-space property, so let KK be any compact T2T_2 space and f:KXf : K \to X any continuous map. f(K)f(K) is immediately compact, so our objective is to prove it is T2T_2. Once that’s done, that means Af(K)A \cap f(K) is open, and so f1(A)=f1(Af(K))f^{-1}(A) = f^{-1}(A \cap f(K)) is open, proving AA is open.

XX is Weak Hausdorff, so f(K)f(K) is closed. Furthermore, for any xXx \in X, the inclusion map id:{x}X\text{id} : \{x\} \to X is continuous and {x}\{x\} is trivially compact and T2T_2, so f({x})={x}f(\{x\}) = \{x\} is closed, proving XX is T1T_1, and consequently, f(K)f(K) is as well. We now verify that f:Kf(K)f : K \to f(K) is a closed map: If AKA \subseteq K is closed in KK, since it must be also compact and T2T_2, f(A)f(A) is closed in XX, hence, in f(K)f(K) as well.

Now take aba \ne b in f(K)f(K). Being T1T_1, we know A=f1({a})A = f^{-1}(\{a\}) and B=f1({b})B = f^{-1}(\{b\}) are closed sets. They’re compact, and KK is T2T_2, so let AUA \subseteq U and BVB \subseteq V with UV=U \cap V = \emptyset. Then let U=f(K)f(UC)U' = f(K) \setminus f(U^C) and V=f(K)f(VC)V' = f(K) \setminus f(V^C). Being a closed map, these are open sets of f(K)f(K). Clearly aUa \in U' and bVb \in V'. zUVz \in U' \cap V' would imply z=f(w)z = f(w) with wUVw \in U \cap V, so UV=U' \cap V' = \emptyset.

Source: The category of CGWH spaces, N. P. Strickland

107

(Countably compact ∧ Meta-Lindelöf)     \implies Compact

Added:

Mar 13, 2026

Difficulty:

Let R={Rj}\mathcal{R} = \{R_j\} be a point-countable open refinement of some open cover {Ui}\{U_i\}. Contrapositively, suppose no countable subcover exists. Recursively, define x0x_0 as any point and R0\mathcal{R}_0 the countable collection of sets in R\mathcal{R} containing xx. Then XR0X \setminus \mathcal{R}_0 is nonempty and we can choose x1x_1 from that. Inductively, let RnR\mathcal{R}_n \subseteq \mathcal{R} of the sets containing xnx_n, so that j=0nRj\bigcup_{j=0}^n \mathcal{R}_j is countable and so xn+1X(j=0nRj)x_{n+1} \in X \setminus (\bigcup_{j=0}^n \mathcal{R}_j).

The sequence Y={xn}Y = \{x_n\} is infinite and has no cluster point: If xx were to be, take xRjx \in R_j and the minimum nn such that xnRjx_n \in R_j. By construction, it is the only element of YY in RjR_j. For any sequence with no cluster point, we can construct a countable cover with no finite subcover. Take U0U_0 nbd of x0x_0, let mm be the least element of which xmU0x_m \notin U_0 and U1U_1 a nbd of xmx_m. Then if mm is the least element with xmknUkx_m \notin \bigcup_{k \le n} U_k, let Un+1U_{n+1} a nbd of xmx_m. Finally, if V=XUnV = X \setminus \bigcup U_n, then {V}{Un}\{V\} \cup \{U_n\} is a countable cover of XX with no finite subcover.

111

(Biconnected ∧ Cardinality 4\geq 4)     \implies T0T_0

Added:

Mar 24, 2026

Difficulty:

By contradiction, suppose xyx \ne y are topologically indistinguishable. Then Y=X{x,y}Y = X \setminus \{x, y\} must be disconnected, since {x,y}\{x, y\} is connected and YY has at least two points. Then there are open sets U,VU,V such that YUVY \subseteq U \cup V and UVY=U \cap V \cap Y = \emptyset. Note that UU can’t contain any of xx or yy, otherwise it must contain both and UU and VV are a separation of XX, which should be connected. Same goes for VV. So {x,y}=(UV)C\{x, y\} = (U \cup V)^C. But notice U{x}U \cup \{x\} is connected: If A,BA,B is a separation by open sets with xAx \in A, then yAy \in A and the pair AVA \cup V,BUB \cap U is a separation of XX. Through a similar reasoning, V{y}V \cup \{y\} is connected. But then XX is not biconnected.

192

(k2k_2-space ∧ k2k_2-Hausdorff)     \implies CGWH

Added:

Mar 26, 2026

Difficulty:

It suffices to prove Weak Hausdorff. Thus, let KK be compact T2T_2 and f:KXf : K \to X continuous. Using the k2k_2-space property, we wish to show that for any other KK' compact T2T_2 and g:KXg : K' \to X continuous, that g1(f(K))g^{-1}(f(K)) is closed. Let Δ\Delta be the diagonal of X2X^2. If we define the function f×g:K×KX2f \times g : K \times K' \to X^2 as (f×g)(x,y)=(f(x),g(x))(f \times g)(x, y) = (f(x), g(x)), then the set L=(f×g)1(Δ)L = (f \times g)^{-1}(\Delta) is closed, by the k2k_2-Hausdorff property (K×KK \times K' is compact T2T_2). If π2:K×KK\pi_2 : K \times K' \to K' is the projection map, since KK is compact, π2\pi_2 is a closed map. But π2(L)=g1(f(K))\pi_2(L) = g^{-1}(f(K)).

211

(Countably tight ∧ Radial)     \implies Fréchet Urysohn

Added:

Mar 13, 2026

Difficulty:

Given pAp \in \overline{A}, choose DAD \subseteq A countable with pDp \in \overline{D}. Suppose (xα)α<λ(x_\alpha)_{\alpha < \lambda} is some transfinite sequence in DD with xαpx_\alpha \to p. We now construct a sequence in DD which converges to pp and we’re done (thanks to PatrickR).

If some yDy \in D is cofinal in (xα)(x_\alpha), this means any neighborhood UU of xx must contain yy, so we can construct a constant sequence yn=yy_n = y with ynpy_n \to p. If no yDy \in D is cofinal, then each Iy={α<λ : xα=y}I_y = \{ \alpha < \lambda \ : \ x_\alpha = y \} is finite and λ=yDIy\lambda = \bigcup_{y \in D} I_y is a countable union of finite sets, so it is a countable ordinal. We can assume λ\lambda to be a regular cardinal, so it is already a sequence.

216

GO-space     \implies Radial

Added:

Mar 26, 2026

Difficulty:

Let VV be an order convex nbd of pAp \in \cl{A}. If A=(,p)A = ({\leftarrow}, p) and B=(p,)B = (p, {\rightarrow}), then we could assume either AVA \cap V has no maximum or BVB \cap V has no minimum. Otherwise, then we could take x=maxAx = \max A and y=minBy = \min B then (x,y)V={p}(x, y) \cap V = \{p\} would be a nbd of pp and so pAp \in A (it would then suffice to take the constant sequence xα=px_\alpha = p). Without loss of generality, say AVA \cap V is infinite.

Using the axiom of choice, specifically Zorn’s lemma, we can take the collection of all well-ordered subsets of AVA \cap V, and any maximal element WW must be cofinal in AVA \cap V. Since this set is well-ordered, there is an isomorphism f:ξWf : \xi \to W from some ordinal ξ\xi and xα=f(α)x_\alpha = f(\alpha) is a sequence with xαpx_\alpha \to p by cofinality.

232

(Totally path disconnected ∧ Embeddable in R\mathbb R)     \implies Zero dimensional

Added:

Mar 28, 2026

Difficulty:

If XYRX \simeq Y \subseteq \R, for each xXx \in X, we shall assume xx is not minimum or maximum element of LL, so that it has a local basis of sets of the form (a,b)X(a, b) \cap X for a,bLa,b \in L (otherwise the proof is similar but the intervals are either [x,b)[x, b) or (a,x](a, x]). Then there must be some c(a,x)c \in (a, x) with cXc \notin X, otherwise (a,x)X(a, x) \subseteq X would be a path connected subset1. Similarly, there must be some d(x,b)Xd \in (x, b) \setminus X so that x(c,d)X=[c,d]Xx \in (c, d) \cap X = [c, d] \cap X is a clopen nbd of xx in XX. So the clopen nbds of XX form a basis.

(1): This is the one time in the proof where we use the fact that we’re in R\R, because connected     \iff path connected in R\R. Otherwise, we could construct a very similar proof for (Totally disconnected ∧ GO-space)     \implies Zero dimensional

235

(Sequential ∧ US)     \implies Strongly KC

Added:

Mar 27, 2026

Difficulty:

Suppose AA is countably compact yet not closed. XX is sequential, so there is a sequence (xn)(x_n) in AA and pAAp \in \cl{A} \setminus A with xnpx_n \to p. Then (xn)(x_n) has some accumulation point yAy \in A. There cannot be infinitely many nn for which xn=yx_n = y, otherwise we could extract a subsequence with xnkyx_{n_k} \to y, and by US, that would imply y=py = p. Therefore, by taking a subsequence if necessary, we can assume yxny \ne x_n for all nn. But then yB={xn : n<ω}{p}y \notin B = \{x_n \ : \ n < \omega \} \cup \{p\} yet yBy \in \cl{B}. So BB isn’t closed, which would imply there is some sequence (yn)(y_n) in BB with ynqy_n \to q and qBBq \in \cl{B} \setminus B. Just as before, there can only be finitely many nn with yn=xmy_n = x_m or yn=py_n = p for any mm. So we could extract a subsequence (ynk)(y_{n_k}) such that ynk=xmky_{n_k} = x_{m_k} is also a subsequence of (xn)(x_n) and so ynkp=qy_{n_k} \to p = q, a contradiction.

257

(Normal ∧ GδG_\delta space)     \implies Perfectly normal

Added:

Mar 28, 2026

Difficulty:

We prove any closed set AA is a zero set. Let A=n=1VnA = \bigcap_{n=1}^\infty V_n be a GδG_\delta set. Without loss of generality, assume {Vn}\{V_n\} is decreasing. Define An=VnCA_n = V_n^C, and by normality, let fn:X[0,1]f_n : X \to [0, 1] a function where fn(A)={0}f_n(A) = \{0\} and fn(An)={1}f_n(A_n) = \{1\}. We define f:X[0,1]f : X \to [0, 1] as

f(x)=n=1fn(x)2nf(x) = \sum_{n=1}^\infty \frac{f_n(x)}{2^n}

It already has the property f1({0})=Af^{-1}(\{0\}) = A as we wished, since f(x)=0f(x) = 0 iff fn(x)f_n(x) for all nn and xAx \notin A implies xAnx \in A_n for some nn and fn(x)0f_n(x) \ne 0. To prove continuity, we use the Weierstrass M-test. Each gn(x)=fn(x)/2n2ng_n(x) = f_n(x)/2^n \le 2^{-n} and 2n=1\sum 2^{-n} = 1 converges, so gn(x)\sum g_n(x) converges absolutely and uniformly. As each finite sum n=1Ngn(x)\sum_{n=1}^N g_n(x) is continuous, their uniform limit ff is continuous.

258

(Regular ∧ Hereditarily Lindelöf)     \implies Perfectly normal

Added:

Mar 29, 2026

Difficulty:

We shall show XX is normal and GδG_\delta (T257). The former is a result of (Regular ∧ Lindelöf)     \implies Normal, and the proof of this fact can be seen in the second-half of (T26). There, we merely use the fact that the space is regular and closed sets are Lindelöf, which is clearly true in a Lindelöf space.

Now let AA be closed. By regularity, for each xAx \notin A there is a pair xUx \in U and AVA \subseteq V of open sets with UV=U \cap V = \emptyset. By joining all such pairs (U,V)(U, V) into a collection W\mathcal{W}, we see that π1(W)\pi_1(\mathcal{W}) is a cover of ACA^C. By hereditarily Lindelöf, there is a countable subcollection {(Un,Vn)}\{(U_n, V_n)\} in W\mathcal{W} which, for any xAx \notin A, there is some nn for which nVnn \notin V_n. Conversely, xAx \in A implies xVnx \notin V_n for all nn. So A=n<ωVnA = \bigcap_{n < \omega} V_n is a GδG_\delta set.

300

Cardinality <c\lt\mathfrak c     \implies Strongly zero-dimensional

Added:

Mar 30, 2026

Difficulty:

Let A=f1(0)A = f^{-1}(0) and B=g1(0)B = g^{-1}(0) be disjoint cozero sets, where f,g:XRf,g : X \to \R are continuous. Then the function

h(x)=f(x)f(x)+g(x)h(x) = \frac{\abs{f(x)}}{\abs{f(x) + g(x)}}

is continuous, since the denominator would only be zero at xx if xABx \in A \cap B. hh maps XX to [0,1][0, 1]. Furthermore, A=h1(0)A = h^{-1}(0) and B=h1(1)B = h^{-1}(1). Note that {h1(t)}0t1\{h^{-1}(t)\}_{0 \le t \le 1} is an uncountable family, yet XX is countable, so h1(r)=h^{-1}(r) = \emptyset for some rr. Then h1([0,r))h^{-1}([0, r)) and h1((r,1])h^{-1}((r, 1]) are clopen disjoint sets of XX which separate AA and BB.

427

(Exhaustible by compacts ∧ KC)     \implies Paracompact

Added:

Apr 4, 2026

Difficulty:

Let {Kn}n=1\{K_n\}_{n=1}^\infty be a cover of compact nbds with Knint(Kn+1)K_n \subseteq \text{int}(K_{n+1}). Let U\mathcal{U} be any open cover. For each nn, there is a finite FnU\mathcal{F}_n \subseteq \mathcal{U} which covers KnK_n. Now we use the fact that each KnK_n is closed by KC, and so Un=int(Kn)Kn1U_n = \text{int}(K_n) \setminus K_{n-1} are all open (to define it at n=1n=1, denote K0=K_0 = \emptyset). Thus, each UnU_n is covered by the finite sets Fn\mathcal{F}_n, and if we define Rn={UnF : FFn}\mathcal{R}_n = \{ U_n \cap F \ : \ F \in \mathcal{F}_n \}, we get that R=n=1Rn\mathcal{R} = \bigcup_{n=1}^\infty \mathcal{R}_n is an open refinement, where for each xx, take the smallest nn for which xint(Kn)x \in \text{int}(K_n) so that xUnx \in U_n and so xx intersects at least one and at most all (finitely many) elements of Rn\mathcal{R}_n.

439

(Compact ∧ Countable)     \implies Sequentially compact

Added:

Apr 4, 2026

Difficulty:

Initially, for each xXx \in X, if a sequence (yn)(y_n) does not converge to xx, then there exists a nbd VV of xx such that for all n0ωn_0 \in \omega, there exists some n>n0n > n_0 such that ynVy_n \notin V. This means there are infinitely many nn for which ynVy_n \notin V, and so we can construct a subsequence (ynk)(y_{n_k}) such that (1) it doesn’t have a converging subsequence either and (2) ynkVy_{n_k} \notin V for all kk.

Let X={xn}X = \{x_n\}. Contrapositively, let y=(yn)\mathbf{y} = (y_n) be a sequence with no converging subsequence. Let V0V_0 be a nbd of x0x_0 and construct a subsequence y0\mathbf{y}_0 of y\mathbf{y} such that every term in y0\mathbf{y}_0 lies outside V0V_0. Inductively, given yn\mathbf{y}_n, choose Vn+1V_{n+1} a nbd of xn+1x_{n+1}, and let yn+1\mathbf{y}_{n+1} be a subsequence of yn\mathbf{y}_n which lies entirely outside of Vn+1V_{n+1}. Note that by construction, each yn\mathbf{y}_n is a subsequence of yk\mathbf{y}_k for any k<nk < n, so all of its terms lie outside of knVk\bigcup_{k \le n} V_k. This way, we’ve constructed an open cover {Vn}\{V_n\} of XX with no finite subcover, and so XX is not compact.