# Bounded set (topological vector space)

In functional analysis and related areas of mathematics, a set in a topological vector space is called bounded or von Neumann bounded, if every neighborhood of the zero vector can be inflated to include the set. A set that is not bounded is called unbounded.

Bounded sets are a natural way to define locally convex polar topologies on the vector spaces in a dual pair, as the polar set of a bounded set is an absolutely convex and absorbing set. The concept was first introduced by John von Neumann and Andrey Kolmogorov in 1935.

## Definition

Suppose ${\displaystyle X}$ is a topological vector space (TVS) over a field ${\displaystyle \mathbb {K} .}$

A subset ${\displaystyle B}$ of ${\displaystyle X}$ is called von Neumann bounded or just bounded in ${\displaystyle X}$ if any of the following equivalent conditions are satisfied:

1. Definition: For every neighborhood ${\displaystyle V}$ of the origin there exists a real ${\displaystyle r>0}$ such that ${\displaystyle B\subseteq sV}$[note 1] for all scalars ${\displaystyle s}$ satisfying ${\displaystyle |s|\geq r.}$[1]
2. ${\displaystyle B}$ is absorbed by every neighborhood of the origin.[2]
3. For every neighborhood ${\displaystyle V}$ of the origin there exists a scalar ${\displaystyle s}$ such that ${\displaystyle B\subseteq sV.}$
4. For every neighborhood ${\displaystyle V}$ of the origin there exists a real ${\displaystyle r>0}$ such that ${\displaystyle sB\subseteq V}$ for all scalars ${\displaystyle s}$ satisfying ${\displaystyle |s|\leq r.}$[1]
5. For every neighborhood ${\displaystyle V}$ of the origin there exists a real ${\displaystyle r>0}$ such that ${\displaystyle tB\subseteq V}$ for all real ${\displaystyle 0[3]
6. Any one of statements (1) through (5) above but with the word "neighborhood" replaced by any of the following: "balanced neighborhood," "open balanced neighborhood," "closed balanced neighborhood," "open neighborhood," "closed neighborhood".
• e.g. Statement (2) may become: ${\displaystyle B}$ is bounded if and only if ${\displaystyle B}$ is absorbed by every balanced neighborhood of the origin.[1]
• If ${\displaystyle X}$ is locally convex then the adjective "convex" may be also be added to any of these 5 replacements.
7. For every sequence of scalars ${\displaystyle s_{1},s_{2},s_{3},\ldots }$ that converges to ${\displaystyle 0}$ and every sequence ${\displaystyle b_{1},b_{2},b_{3},\ldots }$ in ${\displaystyle B,}$ the sequence ${\displaystyle s_{1}b_{1},s_{2}b_{2},s_{3}b_{3},\ldots }$ converges to ${\displaystyle 0}$ in ${\displaystyle X.}$[1]
• This was the definition of "bounded" that Andrey Kolmogorov used in 1934, which is the same as the definition introduced by Stanisław Mazur and Władysław Orlicz in 1933 for metrizable TVS. Kolmogorov used this definition to prove that a TVS is seminormable if and only if it has a bounded convex neighborhood of the origin.[1]
8. For every sequence ${\displaystyle b_{1},b_{2},b_{3},\ldots }$ in ${\displaystyle B,}$ the sequence ${\textstyle \left({\tfrac {1}{i}}b_{i}\right)_{i=1}^{\infty }}$ converges to ${\displaystyle 0}$ in ${\displaystyle X.}$[4]
9. Every countable subset of ${\displaystyle B}$ is bounded (according to any defining condition other than this one).[1]

If ${\displaystyle {\mathcal {B}}}$ is a neighborhood basis for ${\displaystyle X}$ at the origin then this list may be extended to include:

1. Any one of statements (1) through (5) above but with the neighborhoods limited to those belonging to ${\displaystyle {\mathcal {B}}.}$
• e.g. Statement (3) may become: For every ${\displaystyle V\in {\mathcal {B}}}$ there exists a scalar ${\displaystyle s}$ such that ${\displaystyle B\subseteq sV.}$

If ${\displaystyle X}$ is a locally convex space whose topology is defined by a family ${\displaystyle {\mathcal {P}}}$ of continuous seminorms, then this list may be extended to include:

1. ${\displaystyle p(B)}$ is bounded for all ${\displaystyle p\in {\mathcal {P}}.}$[1]
2. There exists a sequence of non-zero scalars ${\displaystyle s_{1},s_{2},s_{3},\ldots }$ such that for every sequence ${\displaystyle b_{1},b_{2},b_{3},\ldots }$ in ${\displaystyle B,}$ the sequence ${\displaystyle b_{1}s_{1},b_{2}s_{2},b_{3}s_{3},\ldots }$ is bounded in ${\displaystyle X}$ (according to any defining condition other than this one).[1]
3. For all ${\displaystyle p\in {\mathcal {P}},}$ ${\displaystyle B}$ is bounded (according to any defining condition other than this one) in the semi normed space ${\displaystyle (X,p).}$
4. B is weakly bounded, i.e. every continuous linear functional is bounded on B[5]

If ${\displaystyle X}$ is a normed space with norm ${\displaystyle \|\cdot \|}$ (or more generally, if it is a seminormed space and ${\displaystyle \|\cdot \|}$ is merely a seminorm),[note 2] then this list may be extended to include:

1. ${\displaystyle B}$ is a norm bounded subset of ${\displaystyle (X,\|\cdot \|).}$ By definition, this means that there exists a real number ${\displaystyle r>0}$ such that ${\displaystyle \|b\|\leq r}$ for all ${\displaystyle b\in B.}$[1]
2. ${\displaystyle \sup _{b\in B}\|b\|<\infty .}$
• Thus, if ${\displaystyle L:(X,\|\cdot \|)\to (Y,\|\cdot \|)}$ is a linear map between two normed (or seminormed) spaces and if ${\displaystyle B}$ is the closed (alternatively, open) unit ball in ${\displaystyle (X,\|\cdot \|)}$ centered at the origin, then ${\displaystyle L}$ is a bounded linear operator (which recall means that its operator norm ${\displaystyle \|L\|:=\sup _{b\in B}\|L(b)\|<\infty }$ is finite) if and only if the image ${\displaystyle L(B)}$ of this ball under ${\displaystyle L}$ is a norm bounded subset of ${\displaystyle (Y,\|\cdot \|).}$
3. ${\displaystyle B}$ is a subset of some (open or closed) ball.[note 3]
• This ball need not be centered at the origin, but its radius must (as usual) be positive and finite.

If ${\displaystyle B}$ is a vector subspace of the TVS ${\displaystyle X}$ then this list may be extended to include:

1. ${\displaystyle B}$ is contained in the closure of ${\displaystyle \{0\}.}$[1]
• In other words, a vector subspace of ${\displaystyle X}$ is bounded if and only if it is a subset of (the vector space) ${\displaystyle \operatorname {cl} _{X}\{0\}.}$
• Recall that ${\displaystyle X}$ is a Hausdorff space if and only if ${\displaystyle \{0\}}$ is closed in ${\displaystyle X.}$ So the only bounded vector subspace of a Hausdorff TVS is ${\displaystyle \{0\}.}$

A subset that is not bounded is called unbounded.

### Bornology and fundamental systems of bounded sets

The collection of all bounded sets on a topological vector space ${\displaystyle X}$ is called the von Neumann bornology or the (canonical) bornology of ${\displaystyle X.}$

A base or fundamental system of bounded sets of ${\displaystyle X}$ is a set ${\displaystyle {\mathcal {B}}}$ of bounded subsets of ${\displaystyle X}$ such that every bounded subset of ${\displaystyle X}$ is a subset of some ${\displaystyle B\in {\mathcal {B}}.}$[1] The set of all bounded subsets of ${\displaystyle X}$ trivially forms a fundamental system of bounded sets of ${\displaystyle X.}$

#### Examples

In any locally convex TVS, the set of closed and bounded disks are a base of bounded set.[1]

## Examples and sufficient conditions

Unless indicated otherwise, a topological vector space (TVS) need not be Hausdorff nor locally convex.

• Finite sets are bounded.[1]
• Every totally bounded subset of a TVS is bounded.[1]
• Every relatively compact set in a topological vector space is bounded. If the space is equipped with the weak topology the converse is also true.
• The set of points of a Cauchy sequence is bounded, the set of points of a Cauchy net need not be bounded.
• The closure of the origin (referring to the closure of the set ${\displaystyle \{0\}}$) is always a bounded closed vector subspace. This set ${\displaystyle \operatorname {cl} _{X}\{0\}}$ is the unique largest (with respect to set inclusion ${\displaystyle \,\subseteq \,}$) bounded vector subspace of ${\displaystyle X.}$ In particular, if ${\displaystyle B\subseteq X}$ is a bounded subset of ${\displaystyle X}$ then so is ${\displaystyle B+\operatorname {cl} _{X}\{0\}.}$

Unbounded sets

A set that is not bounded is said to be unbounded.

Any vector subspace of a TVS that is not a contained in the closure of ${\displaystyle \{0\}}$ is unbounded

There exists a Fréchet space ${\displaystyle X}$ having a bounded subset ${\displaystyle B}$ and also a dense vector subspace ${\displaystyle M}$ such that ${\displaystyle B}$ is not contained in the closure (in ${\displaystyle X}$) of any bounded subset of ${\displaystyle M.}$[6]

### Stability properties

• In any TVS, finite unions, finite Minkowski sums, scalar multiples, translations, subsets, closures, interiors, and balanced hulls of bounded sets are again bounded.[1]
• In any locally convex TVS, the convex hull (also called the convex envelope) of a bounded set is again bounded.[7] However, this may be false if the space is not locally convex, as the (non-locally convex) Lp space ${\displaystyle L^{p}}$ spaces for ${\displaystyle 0 have no nontrivial open convex subsets.[7]
• The image of a bounded set under a continuous linear map is a bounded subset of the codomain.[1]
• A subset of an arbitrary (Cartesian) product of TVSs is bounded if and only if its image under every coordinate projections is bounded.
• If ${\displaystyle S\subseteq X\subseteq Y}$ and ${\displaystyle X}$ is a topological vector subspace of ${\displaystyle Y,}$ then ${\displaystyle S}$ is bounded in ${\displaystyle X}$ if and only if ${\displaystyle S}$ is bounded in ${\displaystyle Y.}$[1]
• In other words, a subset ${\displaystyle S\subseteq X}$ is bounded in ${\displaystyle X}$ if and only if it is bounded in every (or equivalently, in some) topological vector superspace of ${\displaystyle X.}$

## Properties

A locally convex topological vector space has a bounded neighborhood of zero if and only if its topology can be defined by a single seminorm.

The polar of a bounded set is an absolutely convex and absorbing set.

Mackey's countability condition[8] — If ${\displaystyle B_{1},B_{2},B_{3},\ldots }$ is a countable sequence of bounded subsets of a metrizable locally convex topological vector space ${\displaystyle X,}$ then there exists a bounded subset ${\displaystyle B}$ of ${\displaystyle X}$ and a sequence ${\displaystyle r_{1},r_{2},r_{3},\ldots }$ of positive real numbers such that ${\displaystyle B_{i}\subseteq r_{i}B}$ for all ${\displaystyle i\in \mathbb {N} }$ (or equivalently, such that ${\displaystyle {\tfrac {1}{r_{1}}}B_{1}\cup {\tfrac {1}{r_{2}}}B_{2}\cup {\tfrac {1}{r_{3}}}B_{3}\cup \cdots \subseteq B}$).

Using the definition of uniformly bounded sets given below, Mackey's countability condition can be restated as: If ${\displaystyle B_{1},B_{2},B_{3},\ldots }$ are bounded subsets of a metrizable locally convex space then there exists a sequence ${\displaystyle t_{1},t_{2},t_{3},\ldots }$ of positive real numbers such that ${\displaystyle t_{1}B_{1},\,t_{2}B_{2},\,t_{3}B_{3},\ldots }$ are uniformly bounded. In words, given any countable family of bounded sets in a metrizable locally convex space, it is possible to scale each set by its own positive real so that they become uniformly bounded.

## Generalizations

### Uniformly bounded sets

A family of sets ${\displaystyle {\mathcal {B}}}$ of subsets of a topological vector space ${\displaystyle Y}$ is said to be uniformly bounded in ${\displaystyle Y,}$ if there exists some bounded subset ${\displaystyle D}$ of ${\displaystyle Y}$ such that

${\displaystyle B\subseteq D\quad {\text{ for every }}B\in {\mathcal {B}},}$
which happens if and only if its union
${\displaystyle \cup {\mathcal {B}}~:=~\bigcup _{B\in {\mathcal {B}}}B}$
is a bounded subset of ${\displaystyle Y.}$ In the case of a normed (or seminormed) space, a family ${\displaystyle {\mathcal {B}}}$ is uniformly bounded if and only if its union ${\displaystyle \cup {\mathcal {B}}}$ is norm bounded, meaning that there exists some real ${\displaystyle M\geq 0}$ such that ${\textstyle \|b\|\leq M}$ for every ${\displaystyle b\in \cup {\mathcal {B}},}$ or equivalently, if and only if ${\textstyle \sup _{\stackrel {b\in B}{B\in {\mathcal {B}}}}\|b\|<\infty .}$

A set ${\displaystyle H}$ of maps from ${\displaystyle X}$ to ${\displaystyle Y}$ is said to be uniformly bounded on a given set ${\displaystyle C\subseteq X}$ if the family ${\displaystyle H(C):=\{h(C):h\in H\}}$ is uniformly bounded in ${\displaystyle Y,}$ which by definition means that there exists some bounded subset ${\displaystyle D}$ of ${\displaystyle Y}$ such that ${\displaystyle h(C)\subseteq D{\text{ for all }}h\in H,}$ or equivalently, if and only if ${\textstyle \cup H(C):=\bigcup _{h\in H}h(C)}$ is a bounded subset of ${\displaystyle Y.}$ A set ${\displaystyle H}$ of linear maps between two normed (or seminormed) spaces ${\displaystyle X}$ and ${\displaystyle Y}$ is uniformly bounded on some (or equivalently, every) open ball (and/or non-degenerate closed ball) in ${\displaystyle X}$ if and only if their operator norms are uniformly bounded; that is, if and only if ${\textstyle \sup _{h\in H}\|h\|<\infty .}$

Proposition[9] — Let ${\displaystyle H\subseteq L(X,Y)}$ be a set of continuous linear operators between two topological vector spaces ${\displaystyle X}$ and ${\displaystyle Y}$ and let ${\displaystyle C\subseteq X}$ be any bounded subset of ${\displaystyle X.}$ Then ${\displaystyle H}$ is uniformly bounded on ${\displaystyle C}$ (that is, the family ${\displaystyle \{h(C):h\in H\}}$ is uniformly bounded in ${\displaystyle Y}$) if any of the following conditions are satisfied:

1. ${\displaystyle H}$ is equicontinuous.
2. ${\displaystyle C}$ is a convex compact Hausdorff subspace of ${\displaystyle X}$ and for every ${\displaystyle c\in C,}$ the orbit ${\displaystyle H(c):=\{h(c):h\in H\}}$ is a bounded subset of ${\displaystyle Y.}$
Proof of part (1)[9]

Assume ${\displaystyle H}$ is equicontinuous and let ${\displaystyle W}$ be a neighborhood of the origin in ${\displaystyle Y.}$ Since ${\displaystyle H}$ is equicontinuous, there exists a neighborhood ${\displaystyle U}$ of the origin in ${\displaystyle X}$ such that ${\displaystyle h(U)\subseteq W}$ for every ${\displaystyle h\in H.}$ Because ${\displaystyle C}$ is bounded in ${\displaystyle X,}$ there exists some real ${\displaystyle r>0}$ such that if ${\displaystyle t\geq r}$ then ${\displaystyle C\subseteq tU.}$ So for every ${\displaystyle h\in H}$ and every ${\displaystyle t\geq r,}$ ${\displaystyle h(C)\subseteq h(tU)=th(U)\subseteq tW,}$ which implies that ${\textstyle \bigcup _{h\in H}h(C)\subseteq tW.}$ Thus ${\textstyle \bigcup _{h\in H}h(C)}$ is bounded in ${\displaystyle Y.}$ Q.E.D.

Proof of part (2)[10]

Let ${\displaystyle W}$ be a balanced neighborhood of the origin in ${\displaystyle Y}$ and let ${\displaystyle V}$ be a closed balanced neighborhood of the origin in ${\displaystyle Y}$ such that ${\displaystyle V+V\subseteq W.}$ Define

${\displaystyle E~:=~\bigcap _{h\in H}h^{-1}(V),}$
which is a closed subset of ${\displaystyle X}$ (since ${\displaystyle V}$ is closed while every ${\displaystyle h:X\to Y}$ is continuous) that satisfies ${\displaystyle h(E)\subseteq V}$ for every ${\displaystyle h\in H.}$ Note that for every non-zero scalar ${\displaystyle n\neq 0,}$ the set ${\displaystyle nE}$ is closed in ${\displaystyle X}$ (since scalar multiplication by ${\displaystyle n\neq 0}$ is a homeomorphism) and so every ${\displaystyle C\cap nE}$ is closed in ${\displaystyle C.}$

It will now be shown that ${\displaystyle C\subseteq \bigcup _{n\in \mathbb {N} }nE,}$ from which ${\displaystyle C=\bigcup _{n\in \mathbb {N} }(C\cap nE)}$ follows. If ${\displaystyle c\in C}$ then ${\displaystyle H(c)}$ being bounded guarantees the existence of some positive integer ${\displaystyle n=n_{c}\in \mathbb {N} }$ such that ${\displaystyle H(c)\subseteq n_{c}V,}$ where the linearity of every ${\displaystyle h\in H}$ now implies ${\displaystyle {\tfrac {1}{n_{c}}}c\in h^{-1}(V);}$ thus ${\displaystyle {\tfrac {1}{n_{c}}}c\in \bigcap _{h\in H}h^{-1}(V)=E}$ and hence ${\displaystyle C\subseteq \bigcup _{n\in \mathbb {N} }nE,}$ as desired.

Thus ${\textstyle C=(C\cap 1E)\cup (C\cap 2E)\cup (C\cap 3E)\cup \cdots }$ expresses ${\displaystyle C}$ as a countable union of closed (in ${\displaystyle C}$) sets. Since ${\displaystyle C}$ is a nonmeager subset of itself (as it is a Baire space by the Baire category theorem), this is only possible if there is some integer ${\displaystyle n\in \mathbb {N} }$ such that ${\displaystyle C\cap nE}$ has non-empty interior in ${\displaystyle C.}$ Let ${\displaystyle k\in \operatorname {Int} _{C}(C\cap nE)}$ be any point belonging to this open subset of ${\displaystyle C.}$ Let ${\displaystyle U}$ be any balanced open neighborhood of the origin in ${\displaystyle X}$ such that

${\displaystyle C\cap (k+U)~\subseteq ~\operatorname {Int} _{C}(C\cap nE).}$

The sets ${\displaystyle \{k+pU:p>1\}}$ form an increasing (meaning ${\displaystyle p\leq q}$ implies ${\displaystyle k+pU\subseteq k+qU}$) cover of the compact space ${\displaystyle C,}$ so there exists some ${\displaystyle p>1}$ such that ${\displaystyle C\subseteq k+pU}$ (and thus ${\displaystyle {\tfrac {1}{p}}(C-k)\subseteq U}$). It will be shown that ${\displaystyle h(C)\subseteq pnW}$ for every ${\displaystyle h\in H,}$ thus demonstrating that ${\displaystyle \{h(C):h\in H\}}$ is uniformly bounded in ${\displaystyle Y}$ and completing the proof. So fix ${\displaystyle h\in H}$ and ${\displaystyle c\in C.}$ Let

${\displaystyle z~:=~{\tfrac {p-1}{p}}k+{\tfrac {1}{p}}c.}$

The convexity of ${\displaystyle C}$ guarantees ${\displaystyle z\in C}$ and moreover, ${\displaystyle z\in k+U}$ since

${\displaystyle z-k={\tfrac {-1}{p}}k+{\tfrac {1}{p}}c={\tfrac {1}{p}}(c-k)\in {\tfrac {1}{p}}(C-k)\subseteq U.}$
Thus ${\displaystyle z\in C\cap (k+U),}$ which is a subset of ${\displaystyle \operatorname {Int} _{C}(C\cap nE).}$ Since ${\displaystyle nV}$ is balanced and ${\displaystyle |1-p|=p-1 we have ${\displaystyle (1-p)nV\subseteq pnV,}$ which combined with ${\displaystyle h(E)\subseteq V}$ gives
${\displaystyle pnh(E)+(1-p)nh(E)~\subseteq ~pnV+(1-p)nV~\subseteq ~pnV+pnV~\subseteq ~pn(V+V)~\subseteq ~pnW.}$
Finally, ${\displaystyle c=pz+(1-p)k}$ and ${\displaystyle k,z\in nE}$ imply
${\displaystyle h(c)~=~ph(z)+(1-p)h(k)~\in ~pnh(E)+(1-p)nh(E)~\subseteq ~pnW,}$
as desired. Q.E.D.

Since every singleton subset of ${\displaystyle X}$ is also a bounded subset, it follows that if ${\displaystyle H\subseteq L(X,Y)}$ is an equicontinuous set of continuous linear operators between two topological vector spaces ${\displaystyle X}$ and ${\displaystyle Y}$ (not necessarily Hausdorff or locally convex), then the orbit ${\textstyle H(x):=\{h(x):h\in H\}}$ of every ${\displaystyle x\in X}$ is a bounded subset of ${\displaystyle Y.}$

### Bounded subsets of topological modules

The definition of bounded sets can be generalized to topological modules. A subset ${\displaystyle A}$ of a topological module ${\displaystyle M}$ over a topological ring ${\displaystyle R}$ is bounded if for any neighborhood ${\displaystyle N}$ of ${\displaystyle 0_{M}}$ there exists a neighborhood ${\displaystyle w}$ of ${\displaystyle 0_{R}}$ such that ${\displaystyle wA\subseteq B.}$

## References

1. Narici & Beckenstein 2011, pp. 156–175.
2. ^ Schaefer 1970, p. 25.
3. ^ Rudin 1991, p. 8.
4. ^ Wilansky 2013, p. 47.
5. ^ Narici Beckenstein (2011). Topological Vector Spaces (2nd ed.). pp. 253, Theorem 8.8.7. ISBN 978-1-58488-866-6.
6. ^ Wilansky 2013, p. 57.
7. ^ a b Narici & Beckenstein 2011, p. 162.
8. ^ Narici & Beckenstein 2011, p. 174.
9. ^ a b Rudin 1991, pp. 42−47.
10. ^ Rudin 1991, pp. 46−47.

Notes

1. ^ For any set ${\displaystyle A}$ and scalar ${\displaystyle s,}$ the notation ${\displaystyle sA}$ denotes the set ${\displaystyle sA:=\{sa:a\in A\}.}$
2. ^ This means that the topology on ${\displaystyle X}$ is equal to the topology induced on it by ${\displaystyle \|\cdot \|.}$ Note that every normed space is a seminormed space and every norm is a seminorm. The definition of the topology induced by a seminorm is identical to the definition of the topology induced by a norm.
3. ^ If ${\displaystyle (X,\|\cdot \|)}$ is a normed space or a seminormed space, then the open and closed balls of radius ${\displaystyle r>0}$ (where ${\displaystyle r\neq \infty }$ is a real number) centered at a point ${\displaystyle x\in X}$ are, respectively, the sets ${\textstyle B_{ and ${\textstyle B_{\leq r}(x):=\{z\in X:\|z-x\|\leq r\}.}$ Any such set is called a (non-degenerate) ball.