You are on page 1of 2

Proposition.

Given a vector space V equipped with norm kk, if the norm satisfies the following three
properties,
(1) v V, kvk 0, equality only holds when v = 0;
(2) u, v V, ku + vk kuk + kvk;
(3) v V, c R, kcvk = |c| kvk;
(4) Parallelogram law holds, i.e., u, v V, ku + vk2 + ku vk2 = 2 kuk2 + 2 kvk2 .
Then hu, vi := 14 (ku + vk2 ku vk2 ) defines an inner product on V.
Proof. The key to the proof is trying to get u, v, w V, hu, v + wi = hu, vi + hu, wi.
From the definition of the h, i and the parallelogram law, we have the following identity,
1
hu, v + wi = (ku + v + wk2 ku v wk2 )
4
1
= (2 k u + v k2 + 2 k w k2 k u + v w k2 + k u v + w k2 2 k u v k2 2 k w k2 )
4
1
1
= (ku v + wk2 ku + v wk2 ) + (ku + vk2 ku vk2 )
4
2
= hu, w vi + 2hu, vi.
By the symmetry of v, w, we also have

hu, v + wi = hu, v wi + 2hu, wi.


Combining the two identities will lead to

hu, w vi hu, v wi = 2hu, wi 2hu, vi,


and then

hu, w vi = hu, wi hu, vi


with the observation hu, w vi = hu, v wi.
Now weve proven the key fact stated at the beginning of the proof, the rest is a standard
exercise with what is called Cauchy equation.
By induction, it can be proven hu, nvi = nhu, vi, n Z. Then p, q Z, q 6= 0, hu, pvi =
p
p
p
qhu, q vi leads to hu, q vi = q hu, vi.
The triangular inequality tells us kk is continuous, therefore the h, i we defined is also continuous, by using a sequence of rational numbers to approximate any real number, we have

c R, hu, cvi = chu, vi,


hence the linearity on the second coordinate.
From here on, it is not a hard exercise to finish proving h, i being a well defined inner product.

Proposition. Suppose H is a Hilbert space and L is a continuous linear operator from H to H, then L is
bounded.
Proof #1,contrapositive argument. Assuming for contradiction that L in unbounded. Then we can
find a sequence of elements { xi }i=1 H with { Mi }i=1 R such that Mi and k L( xi )k >
Mi k x i k .
Consider yi := k L(xxi )k , then kyi k M1 i 0.
i
However with k L(yi )k = 1, this statement contradicts with the fact that L is continuous at
origin.

1

Proof #2, forward argument. Since L is continuous at origin, > 0 such that k x k implies
k L( x )k 1.
Now we take M = 2 , for every v 6= 0,



v 2 kvk

)
k L(v)k = L(
2 kvk


2 kvk
v

L(
)
=

2 kvk



v

= M kvk L(
) M kvk .
2 kvk
The inequality holds trivially for the origin.


You might also like