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

At the moment, I'm studying Ergodic Theory and I find myself a little stuck.

The Weyl Equidistribution Theorem states that the following are equivalent: 1. For any $f \in L^{1}([0,1])$ and sequence $\{ u_{k} \}_{k \geq 1}$ on $[0,1]$, \begin{eqnarray} \lim_{n \to \infty} \frac{1}{n} \sum_{k = 1}^{n} f( u_{k} ) = \int_{0}^{1} f(x) \, dx. \end{eqnarray} 2. The sequence $\{ u_{k} \}_{k \geq 1}$ is uniformly distributed modulo 1, i.e., equidistributed on $[0,1]$.

For fixed real $b, t > 0$, how does one compute the following limit? \begin{eqnarray} \lim_{a \to \infty} \tfrac{1}{a t} \sum_{i = 1}^{at} \lbrace b ( t - \tfrac{i}{a} ) \rbrace \end{eqnarray} where $\lbrace \cdot \rbrace$ denotes the fractional part function. It is known that the sequence of fractional parts of the integral multiples of the irrational $\alpha$, i.e., $\{ \{ k \alpha \} \}_{k \geq 1}$, is equidistributed on $[0,1]$. I'm not sure how to work with the fractional part in the sum above.

Update: For fixed real $a, b > 0$, how does one compute the following limit? \begin{eqnarray} \lim_{t \to \infty} \tfrac{1}{a t} \sum_{i = 1}^{at} \lbrace b ( t - \tfrac{i}{a} ) \rbrace \end{eqnarray} Unfortunately, the Riemann summation trick doesn't work in this case. (I believe the answer is $\frac{1}{2}$ in the general case, and I think it follows from the above theorem and the average value of $\{ \cdot \}$ on $[0,1]$. However, if $a$ divides $b$, then the limit is $0$.)


share|cite|improve this question
For your second sum, try making the substitution $n = a t$ and try to get a sum where the summand is of the form $\{ \alpha i \}$ – yjj Feb 22 '11 at 14:58
up vote 3 down vote accepted

First make the substitution $n = a t$, then your sum becomes $$\frac{1}{n} \sum_{i=1}^n \{(b t) (1 - i/n) \}.$$ Now this is really just a Riemann sum which converges (as $n \rightarrow \infty$) to $$ \int_0^1 \{(bt) (1-i) \} di$$ Putting $u = (bt)(1-i)$ the integral becomes $$ \frac{1}{b t} \int_0^{bt} \{u\} du $$ and this is equal to $$\frac{1}{2 bt} ( \lfloor{b t\rfloor} + \{ b t \}^2 ).$$ (Because there are $\lfloor{b t\rfloor}$ triangles of area $1/2$ and one triangle of area $\{ b t \}^2 / 2$.)

share|cite|improve this answer
Thanks for the post. Does it follow that Weyl's Theorem doesn't necessarily play a role here? Or is it just lurking behind the Riemann summation? – user02138 Feb 21 '11 at 8:29
I'm not sure. I have always thought that what you called Weyl's theorem to be just a simple characterization of equidistributed sequences. I think Weyl's theorem refers to's_criterion_for_equidistribution . Anyhow, to use the sum-integral equivalence, you need to know your sequence is equidstributed which I think is doable with what I call Weyl's theorem. – yjj Feb 21 '11 at 9:10
The "reason" that your problem has nothing to do with Weyl's theorem is that the sequence $u_k$ does not depend on $n$, whereas your summands do depend on $a = n/t$. – Greg Graviton Feb 22 '11 at 15:24
Thanks for the comment. Yes, I finally realized this point after I worked through yjj's answer. – user02138 Feb 22 '11 at 23:09

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.