A bounded set is a set whose elements stay within a finite distance of a chosen reference point. This is one of the first places where the language of [metric spaces](/page/Metric%20Space) turns geometric intuition into a reusable definition: an interval may have endpoints or fail to have endpoints, a subset may be open or closed, but boundedness asks a separate question about how far the set can spread. In real analysis, boundedness is the hypothesis that prevents sequences, functions, and domains from escaping to infinity while other limiting processes are studied. Boundedness is not a topological notion by itself. It depends on a metric or a norm, so it belongs naturally beside [normed vector spaces](/page/Normed%20Vector%20Space), compact sets, and the study of [sequences](/page/Sequence). In Euclidean space it is tied to compactness through the [Heine-Borel theorem](/theorems/309); in general metric spaces it separates from compactness and [total boundedness](/page/Total%20Boundedness), making it a useful test case for understanding which results are genuinely Euclidean. ## Definition The basic problem is to say that a set has finite extent without referring to coordinates or to an ambient line. A metric provides the only structure needed: it lets us measure how far a point of the set is from a reference point. For a point $x_0\in X$ and a radius $R>0$, the notation $B(x_0,R)$ means the open metric ball \begin{align*} B(x_0,R)=\{x\in X:d(x,x_0)<R\}. \end{align*} [definition: Bounded Set in a Metric Space] Let $(X,d)$ be a metric space and let $A \subset X$. The set $A$ is bounded in $(X,d)$ if $A=\varnothing$, or if there exist $x_0 \in X$ and $R > 0$ such that \begin{align*} A \subset B(x_0,R). \end{align*} [/definition] Most applications in elementary analysis take place in $\mathbb{R}^n$, where distances are computed from the Euclidean norm. A separate Euclidean formulation is needed to make the estimate usable in coordinates, in inequalities, and in comparisons with balls and boxes. The next definition also fixes the convention that bounded means bounded for the usual [Euclidean metric](/page/Euclidean%20Metric) unless another metric is named. [definition: Bounded Set in Euclidean Space] Let $A \subset \mathbb{R}^n$. The set $A$ is bounded if there exist $x_0 \in \mathbb{R}^n$ and $R>0$ such that \begin{align*} |x-x_0| < R \qquad \text{for every } x \in A. \end{align*} [/definition] The reference point is often taken to be the origin. That choice is convenient for computations, but it is not part of the concept. To compare different choices of centre and to state size estimates without naming any centre at all, it is useful to measure the largest distance between two points of the set. [definition: Diameter] Let $(X,d)$ be a metric space. Write $\mathcal{P}(X)$ for the power set of $X$, the collection of all subsets of $X$. The diameter functional is the map \begin{align*} \operatorname{diam}:\mathcal{P}(X)\to[0,\infty] \end{align*} defined by \begin{align*} \operatorname{diam}(A)=\sup\{d(x,y):x,y\in A\}, \end{align*} with the convention \begin{align*} \operatorname{diam}(\varnothing):=0. \end{align*} [/definition] With this convention, a subset has finite diameter exactly when this functional takes a finite value on it. The empty set is assigned diameter $0$ so that finite-size statements do not need to exclude it separately. The definitions above can be checked immediately on the sets that first motivate the idea. Bounded intervals fit inside large balls, while rays do not. [example: Interval and Ray in the Real Line] In $\mathbb{R}$ with its usual metric, the interval $(0,1)$ is bounded. If $x\in(0,1)$, then $0<x<1$, so \begin{align*} |x-0|=x<1<2. \end{align*} Thus every $x\in(0,1)$ lies in $B(0,2)$, and hence \begin{align*} (0,1)\subset B(0,2). \end{align*} By contrast, the ray $[0,\infty)$ is unbounded. To see this, let $x_0\in\mathbb{R}$ and $R>0$ be arbitrary, and set \begin{align*} y=|x_0|+R+1. \end{align*} Then $y\ge 0$, so $y\in[0,\infty)$. If $x_0\ge 0$, then $|x_0|=x_0$, and \begin{align*} |y-x_0|=|x_0+R+1-x_0|=R+1>R. \end{align*} If $x_0<0$, then $|x_0|=-x_0$, and \begin{align*} |y-x_0|=|-x_0+R+1-x_0|=R+1-2x_0>R. \end{align*} In either case, $y\notin B(x_0,R)$. Therefore no ball in $\mathbb{R}$ contains all of $[0,\infty)$, so the ray is unbounded. This example separates finite extent from the presence of endpoints. [/example] Limits in analysis are usually expressed through sequences, so size control for a sequence must mean more than each individual term being a point of the space. The relevant question is whether all terms live in a single bounded region. This definition lets boundedness enter compactness criteria and convergence arguments for sequences. [definition: Bounded Sequence] Let $(X,d)$ be a metric space and let $(x_k)_{k=1}^{\infty}$ be a sequence in $X$. The sequence $(x_k)_{k=1}^{\infty}$ is bounded if the set \begin{align*} \{x_k:k\in \mathbb{N}\} \end{align*} is a bounded subset of $X$. [/definition] For real-valued functions, the object whose size matters is usually the image of the domain. A function may be defined on a complicated set, but boundedness asks whether all its output values fit into a finite interval. [definition: Bounded Function] Let $X$ be a set and let $f:X\to \mathbb{R}$ be a function. The function $f$ is bounded on $X$ if the set $f(X)\subset \mathbb{R}$ is bounded. [/definition] For functions into $\mathbb{R}^m$ or a normed [vector space](/page/Vector%20Space), the same definition is used with the metric induced by the norm on the codomain. A quick example shows why the definition looks at the image rather than at the domain. [example: Bounded and Unbounded Images on the Same Domain] Let $X=(0,1)\subset\mathbb{R}$. Define $f:X\to\mathbb{R}$ by $f(x)=x$, and define $g:X\to\mathbb{R}$ by $g(x)=1/x$. We first compute the image of $f$. If $x\in X$, then $0<x<1$ and $f(x)=x$, so $f(x)\in(0,1)$. Conversely, if $y\in(0,1)$, then $y\in X$ and $f(y)=y$, so $y\in f(X)$. Hence \begin{align*} f(X)=(0,1). \end{align*} For every $y\in f(X)$, we have $0<y<1$, and therefore \begin{align*} |y-0|=y<1<2. \end{align*} Thus $f(X)\subset B(0,2)$, so $f$ is bounded on $X$. Now consider $g$. To show that $g(X)$ is not bounded, let $x_0\in\mathbb{R}$ and $R>0$ be arbitrary. Set \begin{align*} t=|x_0|+R+1. \end{align*} Then $t>1$, so $x=1/t$ satisfies $0<x<1$ and hence $x\in X$. For this $x$, \begin{align*} g(x)=\frac{1}{1/t}=t=|x_0|+R+1. \end{align*} Since $x_0\le |x_0|$, we have \begin{align*} g(x)-x_0=|x_0|+R+1-x_0\ge R+1>R. \end{align*} In particular $g(x)-x_0>0$, so \begin{align*} |g(x)-x_0|=g(x)-x_0>R. \end{align*} Therefore $g(x)\notin B(x_0,R)$. Since no choice of $x_0$ and $R$ captures all of $g(X)$, the function $g$ is not bounded on $X$. Thus boundedness of a real-valued function is a condition on its output values, not merely on the size of its domain. [/example] ## Equivalent Characterisations The metric definition uses containment in a ball. The diameter formulation says the same thing using pairwise distances inside the set. This is often the more natural form when proving stability under unions, images, and estimates, because it avoids carrying around a preferred centre. [quotetheorem:8716] This characterisation is useful because it removes the arbitrary centre from the definition. To show a set is bounded, it is enough to control all distances between pairs of its own points; conversely, a finite-radius ball automatically gives a finite pairwise distance bound. The equivalence also explains why the empty set and singleton sets are harmless edge cases: their internal spread is already finite. Later estimates often use diameter because it behaves naturally under inclusions and lets one compare sizes without first choosing an origin. Componentwise control is often the first form in which a multivariable estimate is obtained. Pairwise control is still not always the easiest way to prove boundedness in $\mathbb{R}^n$, because computations usually produce separate inequalities for separate coordinates. If every coordinate projection has finite range, then the whole set should not be able to escape in any direction. The following result turns that intuition into an exact criterion for subsets of Euclidean space. [quotetheorem:8717] The coordinate criterion is the bridge between geometric boundedness and the inequalities that appear in calculations. It says that no hidden diagonal escape is possible in finite-dimensional Euclidean space: controlling each coordinate separately controls the Euclidean norm, and any Euclidean ball automatically controls every coordinate. This equivalence is special to having finitely many coordinates; it is one reason finite-dimensional Euclidean arguments do not transfer unchanged to infinite-dimensional function spaces. Coordinate bounds are often packaged geometrically as containment in a box. Boxes are useful in integration, compactness arguments, and estimates that separate variables. The next criterion gives a convenient rectangular version of Euclidean boundedness. [quotetheorem:8718] The box formulation is often the most practical boundedness test in multivariable work. A box gives explicit upper and lower coordinate bounds, so it is well suited to estimates on rectangular domains, integration regions, and compactness arguments based on closed bounded boxes. It is equivalent to ball containment for Euclidean boundedness, but it records more coordinate-level information than a single radius does. The limitation is also important: the box must have finite side lengths in every coordinate direction. Function estimates are usually written as inequalities rather than as statements about images of sets. To say that the image of $f$ is bounded should mean that all values of $f$ stay within a fixed distance of $0$, independent of the input. The possible obstruction is that the values may remain finite pointwise while still requiring larger and larger bounds. The useful test is therefore whether one constant controls $|f(x)|$ for every point of the domain, the form needed for sup norms and estimates for [continuous functions](/page/Continuous%20Function). [quotetheorem:8719] The least possible such $M$ is the supremum of $|f|$, when that supremum exists as a finite real number. This connects bounded functions to the [supremum](/page/Supremum) and to function-space norms. ## Standard Examples The simplest examples show that boundedness is independent of openness and closedness. A set can fail to contain its endpoint while still staying inside a finite interval. [example: Open Interval Is Bounded] In $\mathbb{R}$ with its usual metric, the interval $(0,1)$ is bounded. Choose the centre $x_0=0$ and radius $R=2$. If $x\in(0,1)$, then $0<x<1$, so $x>0$ and therefore \begin{align*} |x-0|=|x|=x<1<2. \end{align*} Thus every $x\in(0,1)$ satisfies $|x-0|<2$, which means $x\in B(0,2)$. Hence \begin{align*} (0,1)\subset B(0,2). \end{align*} Since the usual open ball around $0$ of radius $2$ is \begin{align*} B(0,2)=\{y\in\mathbb{R}:|y|<2\}=(-2,2), \end{align*} this is exactly a containment in a finite-radius ball. The interval is not closed in $\mathbb{R}$: the point $0$ is not in $(0,1)$, but every ball $B(0,r)$ with $r>0$ contains the point $\min\{r/2,1/2\}$, and this point lies in $(0,1)$. Thus boundedness does not require closedness. [/example] The next example separates boundedness from having endpoints included. It also shows why compactness needs more than boundedness in Euclidean space. [example: Bounded Set Need Not Be Compact] The set $A=(0,1)\subset\mathbb{R}$ is bounded. If $x\in A$, then $0<x<1$, so \begin{align*} |x-0|=|x|=x<1<2. \end{align*} Thus $A\subset B(0,2)$. The set is not compact. For each $n\in\mathbb{N}$, let \begin{align*} U_n=\left(\frac{1}{n},1\right). \end{align*} Each $U_n$ is open in $\mathbb{R}$. If $x\in(0,1)$, then $1/x>1$, so there exists $n\in\mathbb{N}$ with $n>1/x$. Hence $1/n<x$, and therefore $x\in U_n$. Thus $\{U_n:n\in\mathbb{N}\}$ is an [open cover](/page/Open%20Cover) of $A$. No finite subcollection covers $A$. Indeed, if $U_{n_1},\dots,U_{n_m}$ are chosen and $N=\max\{n_1,\dots,n_m\}$, then \begin{align*} U_{n_1}\cup\cdots\cup U_{n_m}\subset \left(\frac{1}{N},1\right). \end{align*} But $1/(N+1)\in(0,1)$ and \begin{align*} \frac{1}{N+1}<\frac{1}{N}, \end{align*} so $1/(N+1)$ is not contained in that finite union. Therefore $A$ is bounded but not compact. Equivalently, the sequence $x_k=1/k$ stays in $A$ and converges in $\mathbb{R}$ to $0\notin A$, showing exactly where compactness fails. [/example] Unboundedness does not mean that every coordinate or every local slice behaves badly. A set may look bounded in many finite windows while still escaping in one direction. [example: A Parabola Is Unbounded] Let \begin{align*} A=\{(x,x^2):x\in\mathbb{R}\}\subset\mathbb{R}^2. \end{align*} We show that $A$ is unbounded by proving that no Euclidean ball can contain it. Let $(a,b)\in\mathbb{R}^2$ and $R>0$ be arbitrary, and choose \begin{align*} x=|a|+R+1. \end{align*} Then $(x,x^2)\in A$. Its distance from $(a,b)$ is \begin{align*} |(x,x^2)-(a,b)|=\bigl((x-a)^2+(x^2-b)^2\bigr)^{1/2}. \end{align*} Since $(x^2-b)^2\ge 0$, we have \begin{align*} \bigl((x-a)^2+(x^2-b)^2\bigr)^{1/2}\ge \bigl((x-a)^2\bigr)^{1/2}=|x-a|. \end{align*} Also \begin{align*} x-a=|a|+R+1-a\ge R+1, \end{align*} because $|a|\ge a$. Hence \begin{align*} |(x,x^2)-(a,b)|\ge |x-a|\ge R+1>R. \end{align*} Thus $(x,x^2)\notin B((a,b),R)$. Since the centre $(a,b)$ and radius $R$ were arbitrary, no ball in $\mathbb{R}^2$ contains all of $A$, so $A$ is unbounded. At the same time, each finite vertical window is bounded. Fix $M>0$ and suppose \begin{align*} (u,v)\in A\cap([-M,M]\times\mathbb{R}). \end{align*} Then $(u,v)=(x,x^2)$ for some $x\in\mathbb{R}$, and $u=x\in[-M,M]$. Therefore \begin{align*} 0\le v=x^2\le M^2. \end{align*} So \begin{align*} A\cap([-M,M]\times\mathbb{R})\subset [-M,M]\times[0,M^2]. \end{align*} If $(u,v)\in[-M,M]\times[0,M^2]$, then $u^2\le M^2$ and $v^2\le M^4$, so \begin{align*} |(u,v)|=(u^2+v^2)^{1/2}\le (M^2+M^4)^{1/2}< (M^2+M^4)^{1/2}+1. \end{align*} Thus $[-M,M]\times[0,M^2]$ is contained in the ball centered at $0$ with radius $(M^2+M^4)^{1/2}+1$, and the windowed intersection is bounded. The parabola escapes every global ball, even though every finite vertical window of it fits inside a finite-radius ball. [/example] Boundedness behaves differently in infinite-dimensional spaces. The unit ball has finite radius by definition, but it need not be compact. This is a key warning for functional analysis. [example: Infinite-Dimensional Unit Ball] Let $\ell^2$ be the [Hilbert space](/page/Hilbert%20Space) of square-summable real sequences with norm \begin{align*} \|a\|_{\ell^2}=\left(\sum_{k=1}^{\infty}|a_k|^2\right)^{1/2}. \end{align*} The closed unit ball \begin{align*} \overline{B}(0,1)=\{a\in\ell^2:\|a\|_{\ell^2}\le 1\} \end{align*} is bounded because if $a\in\overline{B}(0,1)$, then \begin{align*} d(a,0)=\|a-0\|_{\ell^2}=\|a\|_{\ell^2}\le 1<2. \end{align*} Hence $\overline{B}(0,1)\subset B(0,2)$. For each $k\in\mathbb{N}$, let $e_k$ be the sequence with $1$ in the $k$th position and $0$ in every other position. Then \begin{align*} \|e_k\|_{\ell^2}=\left(1^2+\sum_{i\ne k}0^2\right)^{1/2}=1, \end{align*} so $e_k\in\overline{B}(0,1)$. If $j\ne k$, then $e_j-e_k$ has entry $1$ in position $j$, entry $-1$ in position $k$, and entry $0$ elsewhere. Therefore \begin{align*} \|e_j-e_k\|_{\ell^2}=\left(1^2+(-1)^2+\sum_{i\ne j,k}0^2\right)^{1/2}=\sqrt{2}. \end{align*} Now let $(e_{k_m})_{m=1}^{\infty}$ be any subsequence of $(e_k)_{k=1}^{\infty}$. Since the indices $k_m$ are distinct, for $m\ne n$ we have \begin{align*} \|e_{k_m}-e_{k_n}\|_{\ell^2}=\sqrt{2}>1. \end{align*} Thus this subsequence is not Cauchy. A norm-convergent sequence is Cauchy: if $u_m\to u$, then for $\varepsilon>0$ choose $N$ such that $\|u_m-u\|_{\ell^2}<\varepsilon/2$ for $m\ge N$, and for $m,n\ge N$ the triangle inequality gives \begin{align*} \|u_m-u_n\|_{\ell^2}\le \|u_m-u\|_{\ell^2}+\|u_n-u\|_{\ell^2}<\varepsilon. \end{align*} Therefore no subsequence of $(e_k)$ converges in norm. If $\overline{B}(0,1)$ were compact, the metric-space [sequential compactness](/page/Sequential%20Compactness) criterion for compactness would give a norm-convergent subsequence of $(e_k)$. Since no such subsequence exists, $\overline{B}(0,1)$ is not compact. Thus closed and bounded sets need not be compact outside finite-dimensional Euclidean space. [/example] ## Properties Boundedness is stable under the basic finite set operations used throughout analysis. The first stability principle is monotonicity: estimates on a large set automatically apply to every smaller set, so it is worth isolating this inheritance property. [quotetheorem:8720] This monotonicity is often the quiet step in an argument: once a complicated set has been trapped inside a ball, every subset inherits the same trap without any new estimate. Its limitation is equally important: boundedness need not pass upward to larger sets, so one cannot prove a union or enlargement is bounded merely by checking one bounded piece. That is why the next stability question asks when several bounded pieces can be combined without losing control. The next stability principle allows finitely many separately controlled pieces to be combined. This is the set-theoretic version of taking the maximum of finitely many bounds, and it is used whenever an argument splits a domain into finitely many regions. [quotetheorem:8721] The word finite matters. In $\mathbb{R}$, every singleton $\{k\}$ is bounded, but the union $\bigcup_{k=1}^{\infty}\{k\}=\mathbb{N}$ is unbounded. In vector spaces with a norm, the next question is whether algebraic operations preserve finite size control. [quotetheorem:8722] Translation and scaling invariance explain why boundedness is a geometric size condition rather than a condition tied to a particular origin or unit of measurement. Translating a set changes where it sits, while multiplying by a scalar changes its scale; neither operation introduces infinitely distant points from a set that was already controlled. These rules are used constantly when sets are centered, normalized, rescaled, or compared through affine changes of variables, but they also show the boundary of the idea: boundedness is preserved by fixed finite transformations, not by processes whose scale factors themselves become unbounded. Compactness is stronger than boundedness, and in Euclidean spaces the precise relationship is especially clean: boundedness supplies the size control, while closedness supplies the missing boundary points. The next result records this Euclidean compactness criterion and explains why closed bounded intervals behave so differently from open bounded intervals. Euclidean spaces have a special compactness criterion. A subset of $\mathbb{R}^n$ becomes compact exactly when boundedness is paired with the topological requirement of closedness. The theorem below explains why bounded intervals with endpoints included behave so differently from open intervals. [quotetheorem:271] Many boundedness results for functions are obtained indirectly through compactness. Continuity alone prevents jumps, but it does not stop values from growing without a compact domain to keep sequences from escaping or approaching a missing boundary point. This raises a function-level version of the compactness question: when does topological control of the domain force numerical control of all function values at once? The next result answers this by combining continuity with compactness of the domain, turning compact images in Euclidean space into bounded ranges. [quotetheorem:305] The compactness hypothesis cannot be replaced by boundedness of the domain. The map \begin{align*} f:(0,1)\to\mathbb{R} \end{align*} defined by \begin{align*} f(x)=\frac{1}{x} \end{align*} has bounded domain but unbounded image. ## Relationship to Other Concepts Boundedness is weaker than compactness, and the difference is one of the main organizing lessons in analysis. In $\mathbb{R}^n$, compact sets are exactly the closed bounded sets, but in a general metric space closed bounded sets may fail to be compact. Total boundedness is a stronger metric condition that controls a set at every scale. Ordinary boundedness uses one ball of possibly large radius; total boundedness asks for finitely many balls at every prescribed small radius, which is much closer to compactness. [definition: Totally Bounded Set] Let $(X,d)$ be a metric space and let $A\subset X$. The set $A$ is totally bounded if for every $\varepsilon>0$ there exists a finite subset $F\subset X$ such that \begin{align*} A\subset \bigcup_{x\in F} B(x,\varepsilon). \end{align*} [/definition] Total boundedness is designed as a compactness-scale strengthening of boundedness. The definition gives many small balls rather than one large ball, so it is not literally the same shape as ordinary boundedness. The useful comparison question is whether this stronger covering condition at every scale automatically gives ordinary boundedness at one scale. The following result records that implication, separating the metric idea of finite small-scale covers from the simpler consequence that the set fits inside a single ball. [quotetheorem:8723] The converse fails in general metric spaces, and the failure is visible even for a bounded set with very simple distances. [example: Bounded Set That Is Not Totally Bounded] Let $X=\mathbb{N}$ and define the [discrete metric](/page/Discrete%20Metric) $d:X\times X\to\mathbb{R}$ by \begin{align*} d(m,n)=0 \text{ if } m=n,\qquad d(m,n)=1 \text{ if } m\ne n. \end{align*} The set $X$ is bounded. Indeed, if $k\in\mathbb{N}$, then either $k=1$, in which case \begin{align*} d(k,1)=d(1,1)=0<2, \end{align*} or $k\ne 1$, in which case \begin{align*} d(k,1)=1<2. \end{align*} Thus every $k\in X$ lies in $B(1,2)$, so \begin{align*} X\subset B(1,2). \end{align*} We now show that $X$ is not totally bounded. Take $\varepsilon=1/2$. For any $n\in\mathbb{N}$, the ball $B(n,1/2)$ consists exactly of the single point $n$. If $m=n$, then \begin{align*} d(m,n)=d(n,n)=0<1/2, \end{align*} so $n\in B(n,1/2)$. If $m\ne n$, then \begin{align*} d(m,n)=1>1/2, \end{align*} so $m\notin B(n,1/2)$. Hence \begin{align*} B(n,1/2)=\{n\}. \end{align*} Now let $F\subset\mathbb{N}$ be finite. If $F=\varnothing$, then \begin{align*} \bigcup_{n\in F}B(n,1/2)=\varnothing\ne\mathbb{N}. \end{align*} If $F\ne\varnothing$, let $N=\max F$ and set $q=N+1$. Then $q\in\mathbb{N}$ and $q\notin F$. Since $B(n,1/2)=\{n\}$ for each $n\in F$, we have \begin{align*} \bigcup_{n\in F}B(n,1/2)=F, \end{align*} and therefore $q$ is not contained in this finite union. Thus no finite collection of balls of radius $1/2$ covers $X$. The space $X$ is bounded, but it is not totally bounded. [/example] The infinite-dimensional unit ball example above gives another bounded set that is not totally bounded in the norm metric. Boundedness for functions leads to a standard function space, where finite output size becomes a normed linear structure. [definition: Space of Bounded Real Functions] Let $X$ be a set. The space of bounded real functions on $X$ is \begin{align*} B(X):=\{f:X\to\mathbb{R}: f \text{ is bounded on } X\}. \end{align*} [/definition] A space of bounded functions becomes analytically useful only after it is equipped with a norm. The relevant norm should record the smallest uniform bound controlling the absolute value of the function over the entire domain. The definition below turns boundedness into a quantitative size measurement. [definition: Sup Norm on Bounded Functions] Let $X$ be a set. The sup norm on bounded real functions is the map \begin{align*} \|\cdot\|_\infty:B(X)\to[0,\infty) \end{align*} defined by \begin{align*} \|f\|_\infty=\sup_{x\in X}|f(x)|. \end{align*} When $X=\varnothing$, the supremum is taken to be $0$. [/definition] This construction links bounded sets to function spaces. Many analytic estimates amount to proving that a family lies in a bounded subset of some normed space. In topology, boundedness should be handled with care. A homeomorphism need not preserve boundedness, because it need not preserve distances. For instance, define \begin{align*} f:(0,1)\to\mathbb{R} \end{align*} by \begin{align*} f(x)=\frac{x}{1-x}. \end{align*} This map is a homeomorphism from a bounded interval onto an unbounded subset of $\mathbb{R}$. Thus boundedness is metric information, not purely topological information. [remark: Metric Dependence] The same underlying set can be bounded for one metric and unbounded for another. Boundedness is therefore a property of a subset together with the metric structure in which it is being measured. [/remark] A final useful perspective is that boundedness is a zeroth-order estimate. It controls size but not oscillation, smoothness, convergence, or compactness by itself. Stronger theories add further structure: equicontinuity in Arzela-Ascoli type results, total boundedness in metric compactness, weak compactness in functional analysis, and coercivity estimates in variational problems. ## References - [Metric Space](/page/Metric%20Space). - [Compact Space](/page/Compact%20Space). - [Normed Vector Space](/page/Normed%20Vector%20Space). - Walter Rudin, *Principles of Mathematical Analysis* (1976). - James R. Munkres, *Topology* (2000). - Haim Brezis, *Functional Analysis, Sobolev Spaces and Partial Differential Equations* (2011).