14.2. The Hahn-Banach Theorem 1

14.2.
The Hahn-Banach Theorem 1
Section 14.2. The Hahn-Banach Theorem
Note. In this section we state and prove the Hahn-Banach Theorem. It involves
extending a certain type of linear functional from a subspace of a linear to the
whole space. It will ultimately give information about the dual space of the linear
space. We’ll see implications of the theorem in this section and throughout the
remainder of this chapter.
Definition. A functional p : X → [0, ∞) on a linear space x is said to be positively

homogeneous provided [(λx) = λp(x) for all x ∈ X, λ > 0. It is subadditive provided
p(x + y) ≤ p(x) + p(y) for all x, y ∈ X.
Note. In Fundamentals of Functional Analysis (MATH 5740) a positively homoge-

neous subadditive functional is called a “Minkowski function” (see the online notes
for Section 5.2: http://faculty.etsu.edu/gardnerr/Func/notes/5-2.pdf).
Definition. Let X be a linear space and let X0 be a linear subspace. If X0 has the
property that there is some x0 ∈ X, x0 6= 0, for which X = X0 + span{x0 }, then
X0 is a linear subspace of codimension 1 in X. A translate of a linear subspace of
codimension 1 is called a hyperplane.
Note. Any norm on linear space X is an example of a positively homogeneous

subadditive functional.
14.2. The Hahn-Banach Theorem 2
Note. We need a preliminary lemma before proving out big result, the Hahn-
Banach Theorem. The lemma involves an extension of a linear functional from a
subspace to a space “one dimension large.”
The Hahn-Banach Lemma. Let p be a positively homogeneous, subadditive

functional on the linear space X and Y a subspace of X on which there is defined
a linear functional ψ for which ψ ≤ p on Y . Let z belong to X \ Y . Then ψ can be
extended to a linear functional ψ on span[Y + z] for which ψ ≤ p on span[Y + z].
Note. The Hahn-Banach Theorem now allows us to extend ψ to all of X. The

proof is the same given by Reed and Simon’s Functional Analysis I (Academic
Press, Methods of Modern Mathematical Physics, 1980) and Promislow’s A First
Course in Functional Analysis (Wiley, 2008) (though these references incorporate
the Hahn-Banach Lemma into the proof of the theorem). It requires Zorn’s Lemma.
Hahn-Banach Theorem. Let p be a positively homogeneous, subadditive func-

tional on a linear space X and Y a subspace of X on which there is defined a linear
functional ψ for which ψ ≤ p on Y . Then ψ may be extended to a linear functional
ψ on all of X for which ψ ≤ p on all of X.
Note. The Hahn-Banach Theorem is named for Hans Hahn (1879–1934) and
Stefan Banach (1892–1945). Hahn published the result for normed linear spaces in
1927 (“Über lineare Gleichungssysteme in linearen Raümen, Journal für die Reine
und Angewandte Math. 157 1927), 241–229). Banach proved an analytic version in
1932 (“Sur les fonctionelles Linéaires (II),” Studia Mathematica 4 (1933), 223–239).
An earlier version was proved for the space C([a, b]) by Eduard Helly (1884–1943)
in 1912. We will explore other versions in versions of the Hahn-Banach Theorem
in this chapter. These notes are based on the Saint Andrews MacTutor History
of Mathematics biography of Eduard Helly, and John Saccoman’s “Evolution of
the Geometric Hahn-Banach Theorem,” Rivista di matematica della Università di
Parma, 17 (1991), 257–264.
Hans Hahn (1879–1934) Stefan Banach (1892–1945)
Theorem 14.7. Let X0 be a linear subspace of a normed linear space X. Then

each bounded linear functional ψ on X0 has an extension to a bounded linear
functional on all of X that has the same norm as ψ. In particular, for each x ∈ X
with x 6= 0 there is ψ ∈ X ∗ for which ψ(x) = kxk and kψk = 1.
Note. Theorem 14.7 shows that for every x ∈ X there is ψ ∈ X ∗, so as long as

X is a nontrivial normed linear space then X ∗ contains nonzero bounded linear
functionals (potentially, many).
Example. To illustrate how Theorem 14.7 can be used to establish the existence
of elements of X ∗, consider X = L∞([a, b]) and X0 = C([a, b]) (C([a.b]) may be
considered a subspace of L∞([a, b]) by Exercise 14.27). For any fixed x0 ∈ [a, b],
define ψ(f ) = f (x0 ) for all f ∈ C([a, b])). Then
ψ(αf + βg) = (αf + βg)(x0 ) = αf (x0 ) + βg(x0 ) = αψ(f ) + βψ(g),
so ψ is linear. Also, ψ(f ) = f (x0 ) ≤ kf k|x0 |, of ψ is bounded by |x0 |. That is,

ψ ∈ X ∗. By Theorem 14.7, ψ can be extended to a bounded linear functional
defined on all of X = L∞([a, b]).
Example. Let X = `∞. Define p({xn }) = lim sup{xn } for all {xn } ∈ `∞. Then
p(λ{xn }) = p({λxn }) = lim sup{λxn } = λ lim sup{xn }
for λ > 0. Also
p({xn } + {yn}) = p({xn + yn}) = lim sup{xn + yn }
≤ lim sup{xn } + lim sup{yn } = p({xn }) + p({yn }).
So p is positively homogeneous and subadditive. Let c0 ⊂

ell∞ be the subspace of convergent sequences. Define L on c0 as L({xn }) = lim(xn )
for all {xn } ∈ c0 . Then L is linear on c0 and L ≤ p on c0 . So by Theorem 14.7, L
has an extension to `∞ and the extension is ≤ p on `∞. Any such extension (we do
not necessarily have uniqueness) is called a Banach limit.
Note. Recall that is Section 13.4 we defined the closed linear complement in a
normed linear space. For X a normed linear space and V a subspace of X, a
(topologically) closed subspace W of X is the closed linear complement of X if
X = V ⊕ W . The next result shows that a finite dimensional subspace of a normed
linear space has a closed linear complement (recall by Corollary 13.6 that any finite
dimensional subspace of a normed linear space is [topologically] closed).
Corollary 14.8. Let X be a normed linear space. If X0 is a finite dimensional

subspace of X, then there is a closed linear subspace X1 of X for which X = X0 ⊕X1 .
That is, X0 has a closed linear complement in X.
Corollary 14.9. Let X be a normed linear space. Then the natural embedding
J : X → X ∗∗ is an isometry.
Theorem 14.10. Let X0 be a subspace of the normed linear space X. Then a

point x ∈ X belongs to the closure of X0 if and only if whenever a functional
ψ ∈ X ∗ vanishes on X0, it also vanishes at x.
Corollary 14.11. Let S be a subset of the normed linear space X. Then the
linear span of S is dense in X if and only if whenever ψ ∈ X ∗ vanishes on S, then
ψ = 0.
Theorem 14.12. Let X be a normed linear space. Then every weakly convergent
sequence in X is bounded. moreover, if {xn } * x in X, then kxk ≤ lim inf kxn k.
Note. Royden and Fitzpatrick comment (page 281): “. . . often by making a clever
choice of the functional p, [the Hahn-Banach Theorem] allows us to create basic
analytical, geometric, and topological tools for functional analysis.” We have seen
that Theorem 14.7 (whose proof depends on the Hahn-Banach Theorem) allows us
to find nonzero elements of X ∗ . In Section 14.4 we use the Hahn-Banach Theorem
with p as the “gauge functional” associated with a convex set to separate disjoint
convex subsets of a linear space by a hyperplane (this result is sometimes called the
“geometric Hahn-Banach Theorem”). In Chapter 15 we use the natural embedding
J : X → X ∗∗ to prove that the closed unit ball of a Banach space X is weakly
sequentially compact if and only if X is reflexive (i.e., J(X) = X ∗∗).
Revised: 5/1/2017

14.2. The Hahn-Banach Theorem 1

Uploaded by

Copyright:

Available Formats

14.2. The Hahn-Banach Theorem 1

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

14.2. The Hahn-Banach Theorem 1

Uploaded by

Copyright:

Available Formats

14.2.

The Hahn-Banach Theorem 1

Section 14.2. The Hahn-Banach Theorem

Definition. A functional p : X → [0, ∞) on a linear space x is said to be positively

Note. In Fundamentals of Functional Analysis (MATH 5740) a positively homoge-

Note. Any norm on linear space X is an example of a positively homogeneous

The Hahn-Banach Lemma. Let p be a positively homogeneous, subadditive

Note. The Hahn-Banach Theorem now allows us to extend ψ to all of X. The

Hahn-Banach Theorem. Let p be a positively homogeneous, subadditive func-

Hans Hahn (1879–1934) Stefan Banach (1892–1945)

Theorem 14.7. Let X0 be a linear subspace of a normed linear space X. Then

Note. Theorem 14.7 shows that for every x ∈ X there is ψ ∈ X ∗, so as long as

ψ(αf + βg) = (αf + βg)(x0 ) = αf (x0 ) + βg(x0 ) = αψ(f ) + βψ(g),

so ψ is linear. Also, ψ(f ) = f (x0 ) ≤ kf k|x0 |, of ψ is bounded by |x0 |. That is,

p(λ{xn }) = p({λxn }) = lim sup{λxn } = λ lim sup{xn }

for λ > 0. Also

p({xn } + {yn}) = p({xn + yn}) = lim sup{xn + yn }

≤ lim sup{xn } + lim sup{yn } = p({xn }) + p({yn }).

So p is positively homogeneous and subadditive. Let c0 ⊂

Corollary 14.8. Let X be a normed linear space. If X0 is a finite dimensional

Theorem 14.10. Let X0 be a subspace of the normed linear space X. Then a

You might also like