This question caught my eye from "How to integrate it" by Sean M. Stewart. I have attempted it for fun and am now stuck.


Let $f$ and $g$ be continuous bounded functions on some interval $[a,b]$ such that $\int_a^b|f(x)-g(x)|dx=0$. Show that $\int_a^b|f(x)-g(x)|^2dx=0$.


Note: $0\le |\int_a^bf(x)-g(x)dx |\le\int_a^b|f(x)-g(x)|dx=0$

We conclude $|\int_a^bf(x)-g(x)dx |=0$

Multiply both sides by $|\int_a^bf(x)-g(x)dx |$ giving $|\int_a^bf(x)-g(x)dx |^2=0$

This is where I'm stuck. Specifically how do I get both the square and the modulus back past the pesky $\int dx$

Possibly $|\left(\int_a^bf(x)-g(x)dx\right)^2|=0$ leading to $0=|\left(\int_a^bf(x)-g(x)dx\right)^2|\le?$

I feel I'm missing something with obvious with modulus but it is escaping me. Of course I could be on the wrong track altogether.

Side question: the afore mention book only has solutions to selected problems so if anyone knew of a full solution set somewhere that'd greatly help.

Thanks in advance for your help.


I now realise I had discarded $f(x)=g(x)$ because I thinking was thinking (incorrectly) of $\int_a^b f(x)-g(x)dx=0$ whereby I imagined a situation such as $f(x)=-x$ and $g(x)=x$ over an interval such as $[-1,1]$ I have accepted the answer that includes the case for discontinuous functions but appreciate all other efforts.

  • 1
    $\begingroup$ If the integral is zero, then doesn't that mean $f(x)=g(x)$? $\endgroup$ – King Tut Apr 6 '18 at 11:30
  • $\begingroup$ Thank you see update to question for my error. $\endgroup$ – Karl Apr 6 '18 at 11:51

Are you sure that the condition is for $f$ and $g$ to be continuous? In that case, if $f$ and $g$ differ at one point $x_0$, then $f-g$ has the same sign and $|f-g| > \epsilon > 0 $ in some small interval around $x_0$, so the integral can't be zero. In other words, under that condition we get the much better result that $f=g$ in $[a,b]$.

Not assuming contiuity, we get have that $|f| < M$ and $|g| < N$ for some constants $M,N$ (in $[a,b]$), so we have $|f-g| < M+N$ in that interval. This leads to $\int_a^b |f(x)-g(x)|^2dx = \int_a^b |f(x)-g(x)| \times |f(x)-g(x)|dx \le \int_a^b |f(x)-g(x)|(M+N)dx = (M+N)\int_a^b |f(x)-g(x)|dx = 0.$

| cite | improve this answer | |
  • $\begingroup$ Thank you see update to question for my error. $\endgroup$ – Karl Apr 6 '18 at 11:51

Replace $f- g$ with $h$. It suffices to show that $$\int_a^b |h|=0 \implies \int_a^b |h|^2=0 \quad \forall\, h \in C([a,b])$$ As @KingTut points out, this follows trivially from $$\forall\, h \in C([a,b]), \int_a^b |h|=0 \implies h \equiv 0.$$ Its contrapositive form has a straightforward proof: let $h(x_0) \ne 0$ at some point $x_0 \in [a,b]$. There exists $\delta > 0$ such that $|h|$ is positive in the $\delta$-neighbourhood of $x_0$. $$\text{i.e.} \forall\, x \in [a,b] \cap (x_0-\delta, x_0+\delta), |h(x)| > 0$$ $$\therefore \int_a^b h \ge \int_{[a,b] \cap (x_0-\delta, x_0+\delta)} |h| > 0$$

In fact, you can replace "continuous" with "measurable". The given condition shows that $f = g$ a.e. in $[a,b]$, so $|f-g|^2 = 0$ a.e. in $[a,b]$.

(Addded in response to OP's edit)

In your example $\int_{-1}^1 (f-g) = \int_{-1}^1 (-2x) dx = 0$, but $\int_{-1}^1 (f-g) = \int_{-1}^1 |-2x| dx = 2 \int_0^1 |x| dx = 1 > 0$, so the condition $\int_a^b|f(x)-g(x)|dx=0$ is not satisfied. As a result, $\int_a^b|f(x)-g(x)|^2dx=\int_{-1}^1 |-2x|^2 dx=2\cdot\frac43 > 0$.

| cite | improve this answer | |
  • $\begingroup$ Thank you see update to question for my error. $\endgroup$ – Karl Apr 6 '18 at 11:51

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.