Question: Fix an interval $[a, b]$. Let $C[a,b]$ be the set of continuous functions from $[a,b]$ to $\mathbb{R}$. For $f,g \in C[a,b]$, define a dot product and norm by $f \cdot g = \int^b_a f(x)g(x) dx$ and $||f||_2 = \sqrt{f \cdot f}= \sqrt{\int^b_a |f(x)|^2 dx}$ (note that the absolute value is not actually necessary). The dot product is clearly bilinear and symmetric (you do not need to show this or that $\cdot$ defines a dot product). Show that $||.||_2$ is a norm on $C[a,b]$.


I know that I need to show three things:

  • $||f||_2 \ge 0$, with $||f||_2=0$ if and only if $f=0$ [Property A]
  • $||cf||_2 = |c|||f||_2$ for all $c \in \mathbb{R}$ and $f \in C[a,b]$ [Property B]
  • Triangle inequality: $||f+g||_2 \le ||f||_2+ ||g||_2$ for all $f,g \in C[a,b]$ [Property C]

For Property A: I am able to show that $\int^b_a |f(x)|^2 dx$ is nonnegative, but then get stuck. Since when you take the square root, $\sqrt{\int^b_a |f(x)|^2 dx}$, won't there be both a positive and negative solution? Also, to show $||f||_2=0$ if and only if $f=0$, I need to show both directions ($a$ iff $b$ is equivalent to $a$ implies $b$ AND $b$ implies $a$). I have no issue showing $f(x)=0$ implies $||f||_2 = 0$ but get stuck showing the other way around. I get to $0 = \int^b_a f(x)^2 dx$ but am not sure where to go from here (I'm guessing I must implement the fundamental theorem of calculus?).

I had no issue showing Property B holds as it is a simple matter of taking a constant out of an integral and then taking the square root.

For Property C I get really stuck. I need to show $\sqrt{\int^b_a (f+g)^2 dx} \le \sqrt{\int^b_a f^2 dx} + \sqrt{\int^b_a g^2 dx}$. My lecturer mentioned using the Cauchy–Schwarz inequality but I don't see it being of any value since it gives $|x \cdot y| \le ||x|| ||y||=\sqrt{x\cdot x} \sqrt{y \cdot y}$ which is a product rather than a sum.

Does anyone have any pointers?


1 Answer 1


For part (a), first let me clarify that if $a$ is a non-negative real, then $\sqrt{a}$ denotes the unique non-negative real whose square is $a$.

Now if $f:[a, b]\to \mathbb R$ is a continuous function and $\int_a^bf^2\ dx = 0$, then $f$ must be identically zero. This is becasue if $f(x_0)=u\neq 0$ for some $x_0\in [a, b]$, then in a small enough neighborhood $(x_0-\delta, x_0+\delta)\cap[a, b]$ around $x$, we would have $f$ takes values in $[u/2, \infty)$. This is becasue of the continuity of $f$. Thus $\int_a^b f^2\ dx \geq \delta u^2/4>0$ and we would have a contradiction.

For the Cauchy-Schwarz, you'd have to first construct a relevant vector space with an inner product. Here is how to do it.

Write $V=C[a, b]$ and note that $V$ is naturally a vector spaceunder pointwise addition and scaling by reals.

Now define an inner product on $V$ by writing $\langle f, g\rangle = \int_{a}^b fg\ dx$. It is easy to check that this is indeed an inner product.

Cauchy-Schwarz now becomes

$$ \int_a^b fg\ dx \leq \sqrt{\int_a^b f^2\ dx}\sqrt{\int_a^b g^2\ dx} $$

If you sqaure both the sides of

$$ \sqrt{\int^b_a (f+g)^2 dx} \le \sqrt{\int^b_a f^2 dx} + \sqrt{\int^b_a g^2 dx} $$ you will see that you are just left with the avatar of Cauchy-Schwarz.

You might want to look at $L^2$-spaces for more context.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.