Let $f\in C^{1}[0,1]$,is this true?

\begin{align*} \lim_{n\to\infty}\frac{\int_{\frac{i}{n}}^{\frac{i+1}{n}}f(x)dx-\frac{1}{2n}[f(\frac{i}{n})+f(\frac{i+1}{n})]}{\frac{1}{n^2}}=0 \end{align*} ($0\leq i\leq n-1,i\in \mathbf{N}$)


We have $$f(x) = f(i/n) + (x-i/n)f'(\xi_1) \,\,\,\,\,\,\,\,\, (\star)$$ $$f(x) = f((i+1)/n) + (x-(i+1)/n)f'(\xi_2) \,\,\,\,\,\,\,\,\, (\dagger)$$ where $\xi_1,\xi_2 \in [1/n,1/(n+1)]$. Hence, averaging $(\star)$ and $(\dagger)$, we get that $$f(x) = \dfrac{f(i/n) + f((i+1)/n)}2 + \dfrac12(x-i/n)f'(\xi_1) + \dfrac12(x-(i+1)/n)f'(\xi_2)$$ Hence, $$\int_{i/n}^{(i+1)/n} f(x) dx = \dfrac{f(i/n) + f((i+1)/n)}{2n} + \dfrac{f'(\xi_1)}{4n^2} - \dfrac{f'(\xi_2)}{4n^2}$$ Hence, we get that $$n^2 \left(\int_{i/n}^{(i+1)/n} f(x) dx - \dfrac{f(i/n) + f((i+1)/n)}{2n} \right) = \dfrac{f'(\xi_1)-f'(\xi_2)}{4}$$ Hence,$$\overbrace{\lim_{n \to \infty} n^2 \left(\int_{i/n}^{(i+1)/n} f(x) dx - \dfrac{f(i/n) + f((i+1)/n)}{2n} \right) = \lim_{\xi_1 \to \xi_2}\dfrac{f'(\xi_1)-f'(\xi_2)}{4}}^{\text{Because $\xi_1 , \xi_2 \in [i/n,(i+1)/n]$ and hence as $n \to \infty$, we get that $\xi_1 \to \xi_2$}}$$ But, $$\underbrace{\lim_{\xi_1 \to \xi_2}\dfrac{f'(\xi_1)-f'(\xi_2)}{4} = 0}_{\text{Because $f \in C^1$}}$$ Hence, we get what we want.

  • $\begingroup$ I found one can also prove it by L'Hospital's law. $\endgroup$
    – Luqing Ye
    Jan 13 '13 at 8:00
  • $\begingroup$ Yes, You could also prove using L'Hospital rule. In general, though I prefer not to use L'Hospital rule when I can use other more direct methods. $\endgroup$
    – user17762
    Jan 13 '13 at 8:05

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.