Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I was trying to prove that a continuous function $f:[a,b]\to\Bbb{R}$ is integrable and thought that I came up with an easy solution so I checked the internet and here is a long proof: This makes me think that something is wrong with my argument:

We know that $f$ attains its maximum and minimum on $[a,b]$, call them $A$ and $B$ resp. Assume that $f$ is not constant so that $A\not=B$ (when $f$ is constant it is trivial to prove that $f$ is integrable). Let $\epsilon>0$. It suffices to find a partition such that $U(P)-L(P)<\epsilon$. This is true for a partition with mesh $<\epsilon/(A-B)$. Isn't it?

share|cite|improve this question
Your choice of partition gives $n\epsilon$ for the value of $U-L$ where $n$ is the number of subintervals determined by your partition, so not smaller than $\epsilon$. – Santiago Canez Jun 6 '14 at 14:32
Correction: not $n\epsilon$ for the exact value of $U-L$, but rather as a bound for $U-L$. – Santiago Canez Jun 6 '14 at 14:46
Don't take this personally, but do you really think that so many mathematicians could have missed such a proof, if it were correct? – Siminore Jun 6 '14 at 14:54
yeah this is why I think something must be wrong. Duh – vilma Jun 6 '14 at 16:12
The counter arguments to your candidate proof show what you can learn from this: If your proposed bound depends on characteristics of a given function (in this case the upper and lower bound), there's probably a function somewhere that can break your proof. As Rene Schipperus points out, we need a stronger guarantee, i.e. uniform continuity. Lucky for us, every continuous function on a compact set is uniformly continuous. – rwols Jun 6 '14 at 18:24
up vote 12 down vote accepted

No, this proof is not correct - there's not a nice connection between the size of the mesh and the difference in $y$-values. Let's look at a specific function, $y = \sqrt x$, on the interval $[0,1]$. Choose $\epsilon = .1$, so the proposed mesh size is $0.1$.

Looking at the interval partitioned as $[0, .1] \cup [.1, .2] \cup ... \cup [.9, 1]$, we have that the difference between the partitions is at least $\sqrt{.1} > .1$ - just look at the very first piece of the partition.

share|cite|improve this answer

Consider $f_c\colon[0,2\pi]\to\mathbb R$, $x\mapsto \sin cx$ where $c$ is some constant. For all $c\ge 1$m you will find $A=1$ and $B=-1$, so your proof suggests that $U(P)-L(P)<\epsilon$ as soon as the mesh is $\delta<\frac{\epsilon}{A-B}$ - no matter what $c$ is. But for suitable $c$ (namely $c>\frac{2\pi}\delta$), you will notice that $f$ assumes the values $A$ and $B$ also in each interval of length $\delta$, hence any such partition will sitll have $U(P)-L(P)\ge 4\pi$.

share|cite|improve this answer

Use the fact that a continuous function on a closed interval is uniformly continuous to find $\delta$ such that $$|x-y| \leq \delta \rightarrow |f(x)-f(y)| \leq \frac{\epsilon}{b-a}$$ then $\delta$ is the size of the mesh.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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