The [Bolzano-Weierstrass theorem](/theorems/628) is one of the most powerful results in elementary analysis: every bounded [sequence](/page/Sequence) in $\mathbb{R}^n$ has a convergent subsequence. This single statement underpins the Extreme Value Theorem, the existence of minimisers in optimisation, and the convergence of approximation schemes throughout classical analysis. The natural question is: in which spaces does this property persist? The answer depends entirely on the topology of the ambient space. In [metric spaces](/page/Metric%20Space), the ability to extract convergent subsequences from arbitrary sequences turns out to be equivalent to [compactness](/page/Compact%20Space) --- the open-cover property that, at first glance, appears to have nothing to do with sequences. This equivalence is the reason analysts in $\mathbb{R}^n$ can move freely between "every open cover has a finite subcover" and "every sequence has a convergent subsequence" without comment. But this equivalence is a theorem about metric spaces, not a tautology. In general topological spaces, the two properties diverge: a space can be compact without being sequentially compact, and sequentially compact without being compact. Understanding where and why this divergence occurs --- and what additional structure restores the equivalence --- is essential for working with the weak and weak-$*$ topologies that dominate modern functional analysis. [example: Where Sequences Fail to Detect Compactness] Consider the product space $X = \{0, 1\}^{[0,1]}$, where each factor $\{0, 1\}$ carries the discrete topology and the index set is the uncountable interval $[0, 1]$. Elements of $X$ are functions $f: [0, 1] \to \{0, 1\}$, and convergence in the [product topology](/page/Product%20Topology) means pointwise convergence: $f_n \to f$ if and only if $f_n(t) \to f(t)$ for every $t \in [0, 1]$. By [Tychonoff's theorem](/theorems/953), $X$ is [compact](/page/Compact%20Space) (it is a product of compact spaces). However, $X$ is *not* sequentially compact. For each $n \in \mathbb{N}$, write $n$ in binary as $n = \sum_{k=0}^\infty b_k(n) \, 2^k$ (where $b_k(n) \in \{0,1\}$ and only finitely many digits are nonzero), and define $f_n: [0, 1] \to \{0, 1\}$ by \begin{align*} f_n(t) = b_k(n) \quad \text{where } t \in [k \cdot 2^{-N}, (k+1) \cdot 2^{-N}) \end{align*} for an appropriate dyadic decomposition (more precisely, define $f_n$ to encode the binary digits of $n$ across the coordinates). A cleaner construction uses the following observation: since $|[0,1]| = \mathfrak{c} = |2^{\mathbb{N}}|$, there exists a bijection $\varphi: [0,1] \to 2^{\mathbb{N}}$, where $2^{\mathbb{N}}$ denotes the set of all subsets of $\mathbb{N}$. Define $f_n: [0,1] \to \{0,1\}$ by $f_n(t) = \mathbb{1}_{\{n \in \varphi(t)\}}$, so $f_n(t) = 1$ if $n$ belongs to the subset $\varphi(t)$, and $f_n(t) = 0$ otherwise. We claim no subsequence of $\{f_n\}$ converges pointwise. Suppose for contradiction that $\{f_{n_j}\}_{j=1}^\infty$ converges pointwise to some $f: [0, 1] \to \{0, 1\}$. Since each $f_{n_j}(t) \in \{0, 1\}$ and the limit $f(t)$ must also lie in $\{0, 1\}$, convergence at $t$ means $f_{n_j}(t)$ is eventually constant: either eventually $1$ or eventually $0$. Define the set $A = \{n_j : j \text{ is odd}\} \subset \mathbb{N}$. Since $\varphi$ is a bijection from $[0,1]$ onto $2^{\mathbb{N}}$, there exists $t_0 \in [0,1]$ with $\varphi(t_0) = A$. Then $f_{n_j}(t_0) = \mathbb{1}_{\{n_j \in A\}}$, which equals $1$ when $j$ is odd and $0$ when $j$ is even. The sequence $\{f_{n_j}(t_0)\}$ alternates between $0$ and $1$ and does not converge --- contradicting the assumption that $\{f_{n_j}\}$ converges pointwise at every $t$. This space is compact (by Tychonoff) but not sequentially compact. The open-cover definition detects compactness; sequences do not. [/example] This example reveals a fundamental limitation: sequences, being indexed by $\mathbb{N}$, are a *countable* probe into the topology. When the topology is determined by uncountably many coordinates (as in uncountable products) or when it lacks a countable local base (as in non-first-countable spaces), sequences may fail to capture the full topological structure. The correct generalisation of sequential compactness to arbitrary topological spaces uses *nets* (Moore-Smith sequences) or *filters*, which can be indexed by arbitrary directed sets. But in the settings that matter most for analysis --- metric spaces and the weak topology on separable Banach spaces --- sequences suffice, and understanding precisely when they suffice is the subject of this article. ## Definition [definition: Sequential Compactness] A topological space $X$ is **sequentially compact** if every sequence $\{x_k\}_{k=1}^\infty$ in $X$ has a convergent subsequence. That is, there exist a strictly increasing function $k: \mathbb{N} \to \mathbb{N}$ (written $k_1 < k_2 < k_3 < \cdots$) and a point $x \in X$ such that $x_{k_j} \to x$ as $j \to \infty$. [/definition] Several points of clarification are worth recording immediately. [remark: Convergence of Subsequences Versus Cluster Points] A point $x \in X$ is a **cluster point** (or accumulation point) of a sequence $\{x_k\}_{k=1}^\infty$ if every neighbourhood of $x$ contains $x_k$ for infinitely many values of $k$. In a metric space, $x$ is a cluster point of $\{x_k\}$ if and only if some subsequence of $\{x_k\}$ converges to $x$. In a general topological space, this equivalence can fail: a sequence may have cluster points without having any convergent subsequence (though this pathology does not arise in first-countable spaces). Sequential compactness requires the existence of a *convergent subsequence*, which is a stronger demand than the existence of a cluster point in general, but equivalent in first-countable spaces. [/remark] [remark: Sequential Compactness Versus Compactness] In full generality, neither sequential compactness nor compactness implies the other: - **Compact but not sequentially compact:** $\{0, 1\}^{\mathfrak{c}}$ (as shown above). - **Sequentially compact but not compact:** The ordinal space $\omega_1 = [0, \omega_1)$ with the order topology, where $\omega_1$ is the first uncountable ordinal. Every sequence in $\omega_1$ is bounded (since a countable union of countable sets is countable in ZFC, and each $\{x_k\}$ has a countable supremum), and hence contained in a closed interval $[0, \alpha]$ for some countable ordinal $\alpha$. This interval is compact (it is order-isomorphic to a well-ordered subset of $\mathbb{R}$), so every sequence has a convergent subsequence. But $\omega_1$ itself is not compact: the open cover $\{[0, \alpha) : \alpha < \omega_1\}$ has no finite subcover. The two notions coincide in metric spaces (and more generally in second-countable spaces), which is the content of the main equivalence theorem below. [/remark] ## The Equivalence in Metric Spaces The most important fact about sequential compactness is that, in the metric setting, it is simply another name for compactness. This equivalence is not an accident --- it reflects the fact that metric topologies are completely determined by convergent sequences. But the equivalence is non-trivial: it requires linking the "global" open-cover property to the "local" sequential property via an intermediate geometric condition. The bridge is **total boundedness** --- a quantitative refinement of boundedness that measures how efficiently the space can be approximated by finite sets. [definition: Total Boundedness] A [metric space](/page/Metric%20Space) $(M, d)$ is **totally bounded** if for every $\varepsilon > 0$, there exists a finite set $\{x_1, \ldots, x_N\} \subset M$ (an **$\varepsilon$-net**) such that \begin{align*} M = \bigcup_{i=1}^N B(x_i, \varepsilon). \end{align*} [/definition] Total boundedness is strictly stronger than boundedness in general. Every bounded subset of $\mathbb{R}^n$ is totally bounded (one can cover a bounded set by finitely many cubes of any prescribed side length), but in infinite-dimensional normed spaces, the closed unit ball is bounded yet *never* totally bounded --- the standard basis vectors in $\ell^2$ are mutually separated by $\sqrt{2}$, so no finite collection of balls of radius less than $\sqrt{2}/2$ can cover them all. Total boundedness is also strictly weaker than compactness on its own: the open interval $(0, 1)$ with the standard metric is totally bounded (it inherits total boundedness from the compact set $[0, 1]$) but not compact, because it is not complete. The missing ingredient is completeness, and the following theorem makes this precise. [quotetheorem:316] This three-way equivalence is the reason that compactness in metric spaces is such a robust notion. Each characterisation has a distinct practical role: **Characterisation (1)** (open covers) is needed for proofs that proceed by covering the space with sets having a local property and then extracting a finite subcover to obtain a *global* conclusion. The proof that continuous functions on compact sets are uniformly continuous is the prototype. **Characterisation (2)** (sequential compactness) is the workhorse of existence theory. The direct method in the [calculus of variations](/page/Calculus%20of%20Variations), Peano's existence theorem for ODEs, and most PDE existence proofs all rely on extracting convergent subsequences from bounded sequences of approximate solutions. **Characterisation (3)** (complete + totally bounded) is the most useful for *verifying* compactness. Completeness is typically straightforward to check (closed subsets of complete spaces are complete), and total boundedness can be verified by constructing explicit $\varepsilon$-nets. The [Arzel\`a-Ascoli theorem](/theorems/66) and the [Rellich-Kondrachov theorem](/theorems/64) are both, at their core, theorems about total boundedness: they establish that controlling derivatives or moduli of continuity prevents sequences from "spreading out," ensuring that the relevant function space is totally bounded in a weaker norm. The equivalence also clarifies the two distinct ways a metric space can fail to be compact. A metric space is non-compact if and only if it is either: - **Incomplete:** sequences can converge to a point "outside" the space (e.g., $(0, 1) \subset \mathbb{R}$), or - **Not totally bounded:** the space is "too spread out" to be covered by finitely many small balls (e.g., $\mathbb{R}$ itself, or the unit ball of an infinite-dimensional Banach space). Both failures are visible through sequences. Incompleteness produces Cauchy sequences with no limit; failure of total boundedness produces sequences with no Cauchy subsequence (constructed greedily, as in the proof of $(2) \Rightarrow (3)$). ## Countable Compactness Between compactness and sequential compactness sits a third notion that clarifies the logical relationships. [definition: Countable Compactness] A topological space $X$ is **countably compact** if every *countable* open cover of $X$ has a finite subcover. [/definition] Compactness requires finite subcovers for *arbitrary* open covers; countable compactness weakens this to countable covers. The relationship between the three notions is: \begin{align*} \text{compact} \implies \text{countably compact} \impliedby \text{sequentially compact}. \end{align*} The first implication is immediate (a countable cover is a special case of an arbitrary cover). The second holds because if $X$ is sequentially compact and $\{U_n\}_{n=1}^\infty$ is a countable open cover with no finite subcover, one can choose $x_n \in X \setminus \bigcup_{k=1}^n U_k$ for each $n$, producing a sequence with no convergent subsequence (any limit point would lie in some $U_N$, and the subsequence would eventually be trapped in $\bigcup_{k=1}^N U_k$, contradicting the construction). Neither implication reverses in general: - Countable compactness does not imply sequential compactness: the space $\{0, 1\}^{\mathfrak{c}}$ is compact (hence countably compact) but not sequentially compact. - Countable compactness does not imply compactness: the first uncountable ordinal $\omega_1$ is countably compact (in fact, sequentially compact) but not compact. However, in the presence of second countability (equivalently, in [metrizable](/page/Metrizable%20Space) [separable](/page/Separable%20Space) spaces), all three notions collapse. A second-countable space is Lindel\"of (every open cover has a *countable* subcover), so compactness and countable compactness coincide. Combined with the metric-space equivalence theorem, this yields: [quotetheorem:1042] The distinction between these notions becomes essential in functional analysis, where one routinely works with topologies that are *not* metrizable. The [weak topology](/page/Weak%20Topology) on an infinite-dimensional [Banach space](/page/Banach%20Space) is never metrizable (on the whole space), and the [weak-$*$ topology](/page/Weak*%20Topology) is metrizable on bounded subsets only when the predual is [separable](/page/Separable%20Space). In these settings, the gap between compactness and sequential compactness is a genuine analytic obstacle, and bridging it requires tools like the [Eberlein-Smulian theorem](/theorems/987). ## Sequential Compactness in Banach Spaces The failure of norm compactness for bounded subsets of infinite-dimensional [Banach spaces](/page/Banach%20Space) --- the closed unit ball is *never* compact in the norm topology, by [Riesz's theorem](/theorems/878) --- forces a fundamental shift in strategy. Existence proofs that rely on extracting convergent subsequences from bounded sequences must either: 1. work in a weaker topology (where bounded sets *are* compact), or 2. impose additional structural conditions (equicontinuity, control of derivatives) that restore total boundedness in the norm topology. The first strategy leads to weak and weak-$*$ compactness; the second leads to the [Arzel\`a-Ascoli theorem](/theorems/66) and the [Rellich-Kondrachov theorem](/theorems/64). Both strategies rely on sequential compactness in an essential way --- the analyst needs *subsequences*, not abstract subnets. ### The Eberlein-Smulian Theorem The [Banach-Alaoglu theorem](/theorems/212) guarantees that the dual unit ball $B_{X^*}$ is compact in the weak-$*$ topology for any normed space $X$. But this is compactness in the sense of open covers (or nets), not sequential compactness. The weak-$*$ topology on $B_{X^*}$ is metrizable only when $X$ is separable (by the [Weak-$*$ Metrizability theorem](/theorems/547)), and without metrizability, one cannot extract convergent *subsequences* from arbitrary sequences. For the weak topology on $X$ itself, the situation is even more delicate. The closed unit ball $B_X$ is weakly compact if and only if $X$ is [reflexive](/page/Reflexive%20Space) ([Kakutani's theorem](/theorems/897)), but even in a reflexive space, the weak topology on $X$ is not metrizable (unless $X$ is finite-dimensional). So one faces the question: does weak compactness of bounded sets imply weak *sequential* compactness? The Eberlein-Smulian theorem answers this affirmatively, in full generality, for bounded subsets of any Banach space. [quotetheorem:987] The equivalence $(1) \Leftrightarrow (2)$ is remarkable for two reasons. First, it shows that for *bounded* subsets of Banach spaces, the pathology of the opening example ($\{0, 1\}^{\mathfrak{c}}$ being compact but not sequentially compact) does not arise in the weak topology. Bounded weakly compact sets are always weakly sequentially compact. This is a deep structural property of Banach spaces that has no purely topological explanation --- it relies on the linear structure and the duality between $X$ and $X^*$. Second, it applies to *all* Banach spaces, not just reflexive ones. The theorem does not assert that bounded sets in a Banach space are weakly compact --- that is the definition of reflexivity. Rather, it asserts that *whenever* a bounded set happens to be relatively weakly compact, sequential compactness comes for free. The direction $(2) \Rightarrow (1)$ is equally powerful: if every sequence in a bounded set has a weakly convergent subsequence, then the set is automatically relatively weakly compact in the full topological sense. ### Weak Sequential Compactness in Reflexive Spaces Combining the Eberlein-Smulian theorem with [Kakutani's characterisation of reflexivity](/theorems/897) yields the result that is most directly applied in PDE theory and the calculus of variations. [quotetheorem:214] This is the infinite-dimensional replacement for the Bolzano-Weierstrass theorem. In $\mathbb{R}^n$, every bounded sequence has a *norm*-convergent subsequence. In a reflexive Banach space, bounded sequences have *weakly* convergent subsequences --- the convergence is weaker, but it is often sufficient for passing to limits in variational problems. The hypotheses deserve careful scrutiny: **Reflexivity is necessary.** In $\ell^1(\mathbb{N})$, which is not reflexive, the standard basis sequence $\{e_n\}_{n=1}^\infty$ satisfies $\|e_n\|_{\ell^1} = 1$ for all $n$, but has no weakly convergent subsequence. This is a consequence of [Schur's theorem](/page/Weak%20Topology): in $\ell^1$, weak convergence and norm convergence coincide. Since $\|e_m - e_n\|_{\ell^1} = 2$ for $m \neq n$, no subsequence converges in norm, and therefore no subsequence converges weakly. The failure is not a technicality --- it reflects the fundamental structural deficiency of non-reflexive spaces for variational methods. **Boundedness is necessary.** Even in a reflexive space, unbounded sequences need not have weakly convergent subsequences. This follows from the [Uniform Boundedness Principle](/page/Uniform%20Boundedness%20Principle): if $x_n \rightharpoonup x$ weakly, then $\sup_n \|x_n\|_X < \infty$. Hence any unbounded sequence --- regardless of the space --- has no weakly convergent subsequence. [example: The Direct Method Using Weak Sequential Compactness] We illustrate the standard application of weak sequential compactness: the *direct method* in the calculus of variations. Let $U \subset \mathbb{R}^n$ be a bounded open set with $C^1$ boundary, let $1 < p < \infty$, and consider the energy functional \begin{align*} E: W^{1,p}_0(U) &\to \mathbb{R} \\ u &\mapsto \int_U |\nabla u|^p \, d\mathcal{L}^n - \int_U f u \, d\mathcal{L}^n, \end{align*} where $f \in L^{p'}(U)$ with $p' = p/(p-1)$. We show $E$ attains its infimum. **Step 1: Coercivity.** By the [Poincar\'e inequality](/theorems/76), $\|u\|_{L^p(U)} \le C \|\nabla u\|_{L^p(U)}$ for $u \in W^{1,p}_0(U)$. By H\"older's inequality: \begin{align*} \left|\int_U f u \, d\mathcal{L}^n\right| \le \|f\|_{L^{p'}(U)} \|u\|_{L^p(U)} \le C \|f\|_{L^{p'}(U)} \|\nabla u\|_{L^p(U)}. \end{align*} Therefore $E(u) \ge \|\nabla u\|_{L^p}^p - C\|f\|_{L^{p'}} \|\nabla u\|_{L^p} \to +\infty$ as $\|\nabla u\|_{L^p} \to \infty$. The functional $E$ is coercive, so any minimising sequence $\{u_k\}$ (with $E(u_k) \to \inf E > -\infty$) is bounded in $W^{1,p}_0(U)$. **Step 2: Extraction of a weakly convergent subsequence.** The space $W^{1,p}_0(U)$ is reflexive (for $1 < p < \infty$). By the Weak Sequential Compactness theorem, the bounded sequence $\{u_k\}$ has a subsequence $\{u_{k_j}\}$ with $u_{k_j} \rightharpoonup u$ weakly in $W^{1,p}_0(U)$ for some $u \in W^{1,p}_0(U)$. **Step 3: Weak lower semicontinuity.** The norm is weakly lower semicontinuous: $\|\nabla u\|_{L^p} \le \liminf_{j \to \infty} \|\nabla u_{k_j}\|_{L^p}$. Since $t \mapsto t^p$ is convex and increasing for $t \ge 0$, this gives $\int_U |\nabla u|^p \le \liminf_{j} \int_U |\nabla u_{k_j}|^p$. Meanwhile, the linear term $\int_U f u_{k_j}$ converges to $\int_U f u$ by definition of weak convergence (since $v \mapsto \int_U fv$ is a bounded linear functional on $W^{1,p}_0$). Combining: \begin{align*} E(u) \le \liminf_{j \to \infty} E(u_{k_j}) = \inf E. \end{align*} Since $u \in W^{1,p}_0(U)$, we have $E(u) \ge \inf E$, so $E(u) = \inf E$. The minimiser exists. The entire argument hinges on Step 2 --- without weak sequential compactness in the reflexive space $W^{1,p}_0(U)$, we could not extract a convergent subsequence from the minimising sequence. The method fails for $p = 1$ because $W^{1,1}_0(U)$ is not reflexive (its dual is not separable), and indeed the minimiser may not exist in $W^{1,1}_0$. [/example] ## Divergence of Sequential and Open-Cover Compactness We have seen that in metric spaces, sequential compactness and compactness coincide, and in Banach spaces, the Eberlein-Smulian theorem restores the equivalence for bounded weakly compact sets. But in general topological spaces, the two notions are genuinely independent. Understanding the failure modes is important both for recognising when sequential arguments are valid and for appreciating why more sophisticated tools (nets, filters) are sometimes necessary. ### Compact but Not Sequentially Compact The prototype is the product space $\{0, 1\}^I$ with $|I| > \aleph_0$. By [Tychonoff's theorem](/theorems/953), this space is compact. The product topology requires convergence at *every* coordinate $\alpha \in I$, and when $I$ is uncountable, a sequence (indexed by $\mathbb{N}$) cannot "visit" enough coordinates to force convergence everywhere. The key obstruction is the failure of first countability. A topological space is **first countable** if every point has a countable neighbourhood base. In a first-countable compact space, sequential compactness follows: given a sequence, the first countability allows one to construct a convergent subsequence by a diagonal argument. The product $\{0, 1\}^I$ for uncountable $I$ is not first countable --- the "natural" neighbourhood base at a point involves prescribing finitely many coordinates, and there are uncountably many finite subsets of $I$ to consider, but the resulting neighbourhood system is still uncountable. [remark: First Countability and Sequences] A topological space $X$ is first countable if and only if sequences determine the closure operator: a point $x$ lies in the closure of a set $A$ if and only if some sequence in $A$ converges to $x$. In spaces that are not first countable, the closure of a set may contain points that are *not* limits of sequences from the set. This is why nets (indexed by arbitrary directed sets, not just $\mathbb{N}$) are needed to characterise compactness in general topological spaces. [/remark] ### Sequentially Compact but Not Compact The ordinal space $\omega_1 = [0, \omega_1)$ provides the standard example. Its sequential compactness follows from the fact that every countable subset of $\omega_1$ is bounded: if $\{x_k\}_{k=1}^\infty \subset \omega_1$, then $\alpha = \sup_k x_k$ is a countable ordinal (a countable supremum of countable ordinals is countable), and the closed interval $[0, \alpha + 1]$ is compact (it is order-isomorphic to a successor ordinal, which is compact in the order topology). Hence $\{x_k\}$ has a convergent subsequence. But $\omega_1$ is not compact: the open cover $\{[0, \alpha) : \alpha < \omega_1\}$ has no finite subcover. The space is "too long" to be covered by finitely many initial segments, even though every sequence is trapped in a compact initial segment. The obstruction is the failure of the Lindel\"of property. A Lindel\"of space (every open cover has a *countable* subcover) that is countably compact is compact, and a sequentially compact space is always countably compact. So the chain breaks at Lindel\"of: $\omega_1$ is not Lindel\"of. This analysis yields a useful sufficient condition: [quotetheorem:1043] In particular, since every [metrizable](/page/Metrizable%20Space) space that is separable is Lindel\"of (by the equivalence of separability, second countability, and the Lindel\"of property in metrizable spaces), this recovers the equivalence of compactness and sequential compactness for separable metric spaces --- and hence for all metric spaces, since a sequentially compact metric space is automatically separable (it is totally bounded, hence has a countable dense subset formed by taking $1/n$-nets for each $n$). ## Sequential Compactness in Non-Metrizable Function Spaces In applications to partial differential equations and the calculus of variations, one frequently encounters situations where the topology is not metrizable on the full space, but is metrizable (or at least "sequentially well-behaved") on bounded subsets. The interplay between metrizability, separability, and sequential compactness in these settings is a recurring source of subtlety. ### The Weak-$*$ Topology on Dual Spaces The [Banach-Alaoglu theorem](/theorems/212) guarantees that the closed unit ball $B_{X^*}$ is weak-$*$ compact for any normed space $X$. When $X$ is [separable](/page/Separable%20Space), the weak-$*$ topology on $B_{X^*}$ is [metrizable](/page/Metrizable%20Space) (by the [Weak-$*$ Metrizability theorem](/theorems/547)), and so compactness and sequential compactness coincide on $B_{X^*}$. This yields the [Sequential Banach-Alaoglu theorem](/theorems/496): every bounded sequence in $X^*$ has a weak-$*$ convergent subsequence. When $X$ is *not* separable, the weak-$*$ topology on $B_{X^*}$ is not metrizable, and the passage from compactness to sequential compactness requires the Eberlein-Smulian machinery (or its weak-$*$ analogue). This situation arises, for instance, when working with the dual of $L^\infty$, which contains finitely additive measures and is highly non-separable. [example: Sequential Weak-$*$ Compactness in $L^p$ Duals] Consider a bounded open set $U \subset \mathbb{R}^n$ and the space $X = L^p(U)$ for $1 \le p < \infty$. The space $X$ is separable (finite rational linear combinations of indicators of rational-endpoint boxes form a countable dense subset). The dual is $X^* = L^{p'}(U)$ where $p' = p/(p-1)$ (with $p' = \infty$ when $p = 1$). By Sequential Banach-Alaoglu, every bounded sequence $\{f_k\}_{k=1}^\infty$ in $L^{p'}(U)$ has a subsequence $\{f_{k_j}\}$ converging in the weak-$*$ topology on $L^{p'}$. Concretely, this means there exists $f \in L^{p'}(U)$ such that \begin{align*} \int_U f_{k_j} \, g \, d\mathcal{L}^n \to \int_U f \, g \, d\mathcal{L}^n \quad \text{for every } g \in L^p(U). \end{align*} For $1 < p < \infty$, the space $L^p(U)$ is reflexive, so the weak-$*$ topology on $L^{p'} = (L^p)^*$ coincides with the weak topology. In this case, weak-$*$ convergence in $L^{p'}$ is the same as weak convergence. For $p = 1$, the space $L^1(U)$ is *not* reflexive, and the dual $(L^1)^* = L^\infty(U)$. Sequential Banach-Alaoglu gives: every bounded sequence in $L^\infty(U)$ has a subsequence converging weak-$*$ in $L^\infty$, meaning $\int_U f_{k_j} g \to \int_U f g$ for every $g \in L^1(U)$. This is strictly weaker than weak convergence in $L^\infty$ (which would require convergence against every element of $(L^\infty)^*$, a much larger space than $L^1$). [/example] ### The Weak Topology on Infinite-Dimensional Spaces The [weak topology](/page/Weak%20Topology) $\sigma(X, X^*)$ on an infinite-dimensional Banach space $X$ is never metrizable on the whole space --- the weak topology has too few open sets to be generated by a single metric. However, when $X^*$ is separable, the weak topology *is* metrizable on bounded subsets of $X$ (since the weak topology on a bounded set is induced by countably many functionals from a dense subset of $X^*$). This leads to a useful criterion: if $X$ is a reflexive Banach space with separable dual, then for bounded sequences, weak convergence behaves exactly like metric convergence --- the extraction of convergent subsequences is guaranteed by compactness (Kakutani) and metrizability (on bounded sets), without needing the Eberlein-Smulian theorem. The Eberlein-Smulian theorem's power lies in handling the non-separable case, where metrizability is unavailable but sequential compactness still holds. ## Standard Techniques for Extracting Convergent Subsequences The theoretical equivalences above are only useful insofar as one can *apply* them in practice. This section describes the standard methods for leveraging sequential compactness in analysis. ### The Bolzano-Weierstrass Strategy The most basic pattern: construct a bounded sequence in a compact (or weakly compact) set and extract a convergent subsequence. **When to use:** Any time an existence proof begins with a sequence of approximate solutions --- minimising sequences in variational problems, Galerkin approximations in PDE theory, iterates of a nonlinear map. **The three-step template:** 1. **Bound the sequence.** Establish a uniform estimate $\|u_k\|_X \le C$ using energy estimates, a priori bounds, or coercivity of the functional. 2. **Extract a subsequence.** Invoke the appropriate compactness theorem: Bolzano-Weierstrass in $\mathbb{R}^n$, Weak Sequential Compactness in a reflexive Banach space, Sequential Banach-Alaoglu in the dual of a separable space, or Rellich-Kondrachov for norm convergence in a weaker space. 3. **Identify the limit.** Show that the limit satisfies the desired equation or optimality condition, using continuity, weak lower semicontinuity, or passage to the limit in the weak formulation. ### Diagonal Extraction When convergent subsequences must be extracted with respect to infinitely many criteria simultaneously, the diagonal argument combines countably many extraction steps into one. [example: Extracting a Subsequence Converging in Every $L^q$] Let $U \subset \mathbb{R}^n$ be a bounded open set with $C^1$ boundary, and let $\{u_k\}_{k=1}^\infty$ be a bounded sequence in $W^{1,p}(U)$ with $1 \le p < n$. By the [Rellich-Kondrachov theorem](/theorems/64), for each integer $q$ with $1 \le q < p^* = np/(n-p)$, the embedding $W^{1,p}(U) \hookrightarrow L^q(U)$ is compact, so there exists a subsequence converging in $L^q(U)$. But the subsequence depends on $q$ --- how do we obtain a *single* subsequence converging in *every* $L^q$ for $q < p^*$? **Step 1.** Choose a sequence of exponents $q_1 < q_2 < q_3 < \cdots$ with $q_j \nearrow p^*$ (for instance, $q_j = p^* - 1/j$ for large enough $j$). By Rellich-Kondrachov, extract a subsequence $\{u_{k,1}\}$ of $\{u_k\}$ converging in $L^{q_1}(U)$. **Step 2.** Extract a further subsequence $\{u_{k,2}\}$ of $\{u_{k,1}\}$ converging in $L^{q_2}(U)$. This subsequence still converges in $L^{q_1}$ (a subsequence of a convergent sequence converges to the same limit). **Step $j$.** Extract $\{u_{k,j}\}$ as a subsequence of $\{u_{k,j-1}\}$ converging in $L^{q_j}(U)$. **Diagonal.** Define $v_k = u_{k,k}$. For each fixed $j$, the sequence $\{v_k\}_{k \ge j}$ is a subsequence of $\{u_{k,j}\}$, hence converges in $L^{q_j}(U)$. Therefore $\{v_k\}$ converges in $L^{q_j}(U)$ for every $j$. For any $q < p^*$, choose $j$ with $q \le q_j$. Since $U$ is bounded, $L^{q_j}(U) \hookrightarrow L^q(U)$ (by H\"older's inequality), so convergence in $L^{q_j}$ implies convergence in $L^q$. The single diagonal subsequence $\{v_k\}$ converges in $L^q(U)$ for *every* $1 \le q < p^*$. [/example] ### Compactness-and-Contradiction Sequential compactness enables a powerful indirect argument: assume a uniform estimate fails, construct a sequence of counterexamples, extract a convergent subsequence, and derive a contradiction from properties of the limit. This pattern is ubiquitous in PDE theory and geometric analysis. [example: A Nonlinear Poincar\'e Inequality via Compactness-and-Contradiction] We prove the following: let $U \subset \mathbb{R}^n$ be a bounded, connected open set with $C^1$ boundary, and let $1 \le p < \infty$. Then there exists a constant $C = C(U, p) > 0$ such that \begin{align*} \|u - \bar{u}\|_{L^p(U)} \le C \|\nabla u\|_{L^p(U)} \quad \text{for all } u \in W^{1,p}(U), \end{align*} where $\bar{u} = \frac{1}{\mathcal{L}^n(U)} \int_U u \, d\mathcal{L}^n$ is the average of $u$ over $U$. Suppose the inequality fails. Then for each $k \in \mathbb{N}$, there exists $u_k \in W^{1,p}(U)$ with $\|u_k - \bar{u}_k\|_{L^p(U)} > k \|\nabla u_k\|_{L^p(U)}$. Normalise by setting $v_k = (u_k - \bar{u}_k) / \|u_k - \bar{u}_k\|_{L^p}$. Then $\|v_k\|_{L^p} = 1$, $\bar{v}_k = 0$, and $\|\nabla v_k\|_{L^p} < 1/k$. **Bounding the $W^{1,p}$ norm.** By the Poincar\'e inequality for functions with zero average (which we are proving, but note that the $W^{1,p}$ bound follows from $\|v_k\|_{L^p} = 1$ and $\|\nabla v_k\|_{L^p} \le 1$, so $\|v_k\|_{W^{1,p}} \le 2$), the sequence $\{v_k\}$ is bounded in $W^{1,p}(U)$. **Extracting a convergent subsequence.** Since $U$ is bounded with $C^1$ boundary, the [Rellich-Kondrachov theorem](/theorems/64) (for $p < n$; the argument for $p \ge n$ is similar using the appropriate compact embedding) gives a subsequence $v_{k_j} \to v$ strongly in $L^p(U)$. **Identifying the limit.** Since $\|v_{k_j}\|_{L^p} = 1$, the limit satisfies $\|v\|_{L^p} = 1$, so $v \neq 0$. Since $\bar{v}_{k_j} = 0$ and $v_{k_j} \to v$ in $L^p$, we have $\bar{v} = 0$. Since $\|\nabla v_{k_j}\|_{L^p} < 1/k_j \to 0$, and weak derivatives are closed under $L^p$ convergence, we have $\nabla v = 0$ a.e. in $U$. But a function in $W^{1,p}(U)$ with $\nabla v = 0$ on a *connected* open set $U$ is constant a.e. (this follows from the characterisation of Sobolev functions via absolutely continuous representatives on lines). Since $\bar{v} = 0$, we have $v = 0$ a.e. --- contradicting $\|v\|_{L^p} = 1$. The compactness-and-contradiction pattern produced both the existence of the constant $C$ and the qualitative structure of the inequality. The argument is nonconstructive: it does not compute $C$, only proves its existence. This is typical of compactness arguments. [/example] ### Metrizability Reduction In functional analysis, one frequently encounters situations where the topology is not metrizable on the full space, but *is* metrizable on the relevant bounded subset. The standard technique is: 1. **Reduce to a bounded set.** Establish a uniform bound $\|x_k\| \le M$ on the sequence. 2. **Verify metrizability on bounded sets.** If $X$ is separable, the weak-$*$ topology on $B_M(X^*)$ is metrizable; if $X^*$ is separable, the weak topology on $B_M(X)$ is metrizable. 3. **Apply metric-space sequential compactness.** On the metrizable bounded set, compactness and sequential compactness coincide, so extract a convergent subsequence. This technique avoids the Eberlein-Smulian theorem entirely when metrizability on bounded sets is available. It is the standard approach in PDE theory, where the Banach spaces involved (Sobolev spaces, $L^p$ spaces for $1 < p < \infty$) are reflexive and separable, so both the space and its dual are separable. ## References - Munkres, J.R., *Topology* (2nd edition, 2000). Chapters 3 and 5 for general compactness, countable compactness, and sequential compactness in topological spaces. - Kelley, J.L., *General Topology* (1955). The definitive classical reference for nets, filters, and the general theory of compactness. - Brezis, H., *Functional Analysis, Sobolev Spaces and Partial Differential Equations* (2011). Chapters 3 and 4 for weak compactness, the Eberlein-Smulian theorem, and applications to variational problems. - Evans, L.C., *Partial Differential Equations* (2nd edition, 2010). Appendix D for weak convergence and compactness in Sobolev spaces; Chapter 8 for the direct method. - Dunford, N. and Schwartz, J.T., *Linear Operators, Part I: General Theory* (1958). The original treatment of the Eberlein-Smulian theorem and its applications. - Engelking, R., *General Topology* (2nd edition, 1989). Comprehensive treatment of compactness, sequential compactness, and countable compactness with examples and counterexamples.