|
imomath
Proof of L’Hopital’s theorem
We will first prove a discrete version of the L’Hopital’s theorem: the case when instead of functions we are dealing with sequences. We may see a sequence \((a_n)_{n=1}^{\infty}\) as a special case of a function. The finite difference \(\frac{a_{n+1}-a_n}{(n+1)-n}\) plays a role that derivative plays when studying functions.
Cesaro-Stolz’s theorem
Assume that \((x_n)_{n=1}^{\infty}\) and \((y_n)_{n=1}^{\infty}\) are two sequences of real numbers such that \(y_n\) is increasing and \(\lim_{n\to\infty}y_n=+\infty\).
(a) The following three inequalities hold: \[ \liminf \frac{x_{n+1}-x_n}{y_{n+1}-y_n}\leq \liminf\frac{x_n}{y_n}\leq \limsup\frac{x_n}{y_n}\leq \limsup \frac{x_{n+1}-x_n}{y_{n+1}-y_n}.\]
(b) Assume that \(\displaystyle\lim_{n\to\infty}\frac{x_{n+1}-x_n}{y_{n+1}-y_n}=L\). Then the limit \(\displaystyle \lim_{n\to\infty}\frac{x_n}{y_n}\) exists and is equal to \(L\).
Notice that (b) is a consequence of (a), so we will only prove part (a).
The middle inequality is obvious as \(\limsup\) is always bigger than \(\lim\inf\). The leftmost inequality is analogous to the rightmost, so we will only prove one of them: the right-most.
Let \(\gamma\) be any number greater than \(\displaystyle \frac{x_{n+1}-x_n}{y_{n+1}-y_n}\). Since \(\gamma\) is not \(\limsup\) of the sequence, there exists \(N\) such that for all \(n > N\) the following inequality holds \(\displaystyle \frac{x_{n+1}-x_n}{y_{n+1}-y_n} < \gamma\). Let \(m\in\mathbb N\) be any integer greater than \(N\). According to the previous inequality we have: \begin{eqnarray*} x_{N+1}-x_N&< &\gamma\left(y_{N+1}-y_N\right)\newline x_{N+2}-x_{N+1}&< &\gamma\left(y_{N+2}-y_{N+1}\right)\newline x_{N+3}-x_{N+2}&< &\gamma\left(y_{N+3}-y_{N+2}\right)\newline &\vdots&\newline x_{m}-x_{m-1}&< &\gamma\left(y_{m}-y_{m-1}\right). \end{eqnarray*} Adding all of the previous \(m-N\) inequalities yield: \[x_m-x_N < \gamma\left(y_m-y_N\right) \quad\quad\Rightarrow \quad\quad \frac{x_m}{y_m}-\frac{x_N}{y_m} < \gamma-\gamma\frac{y_N}{y_m}.\] Since \(y_m\to+\infty\) as \(n\to\infty\) we conclude that \[\limsup \frac{x_m}{y_m}\leq\gamma.\] This holds for every \(\displaystyle \gamma > \limsup \frac{x_{n+1}-x_n}{y_{n+1}-y_n}\) thus \[\limsup \frac{x_m}{y_m}\leq\limsup\frac{x_{n+1}-x_n}{y_{n+1}-y_n}.\]
L’Hopital’s theorem
Assume that \(f\) and \(g\) are differentiable functions on an open interval containing real number \(a\). Assume that \(g\neq 0\) and \(g^{\prime}\neq 0\) on that interval. Assume that \(\displaystyle \lim_{x\to a} \frac{f^{\prime}(x)}{g^{\prime}(x)}=L\) for some real number \(L\) and assume that one of the following two conditions is satisfied:
(a) \(\displaystyle \lim_{x\to a}g(x)=+\infty\).
(b) \(\displaystyle \lim_{x\to a}f(x)=\lim_{x\to a}g(x)=0\).
Then \(\displaystyle \lim_{x\to a}\frac{f(x)}{g(x)}=L\).
Proof
(a) Since we assumed that \(g^{\prime}\neq 0\), according to Darboux’s mean value theorem we may assume that \(g^{\prime}\) is either positive or negative. That means that \(g\) is monotone. It suffices to prove that for each monotone sequence \(x_n\) such that \(\displaystyle \lim_{n\to \infty}x_n=a\) we have \(\displaystyle\lim_{n\to \infty}\frac{f(x_n)}{g(x_n)}=L\).
The sequence \(g(x_n)\) is increasing and converges to \(+\infty\) hence according to Cesaro-Stolz theorem we have \[\liminf \frac{f(x_{n+1})-f(x_{n})}{g(x_{n+1})-g(x_n)}\leq \liminf \frac{f(x_n)}{g(x_n)}\leq \limsup \frac{f(x_n)}{g(x_n)}\leq \limsup \frac{f(x_{n+1})-f(x_n)}{g(x_{n+1})-g(x_n)}.\] According to Cauchy mean-value theorem we have that there exists a sequence of numbers \(\xi_n\in(x_n,x_{n+1})\) such that \[\frac{f(x_{n+1})-f(x_n)}{g(x_{n+1})-g(x_n)}=\frac{f^{\prime}(\xi_n)}{g^{\prime}(\xi_n)}.\] We must have \(\xi_n\to a\) since \(\xi_n\in(x_n,x_{n+1})\). According to our assumption that \(\displaystyle \lim_{x\to a}\frac{f^{\prime}(x)}{g^{\prime}(x)}=L\) we have: \[ L=\liminf \frac{f’(\xi_n)}{g’(\xi_n)}\leq \liminf \frac{f(x_n)}{g(x_n)}\leq \limsup \frac{f(x_n)}{g(x_n)}\leq \limsup \frac{f’(\xi_n)}{g’(\xi_n)}=L. \] This completes the proof in the case (a).
(b) Let \(\varepsilon > 0\). There exists a neighborhood \(U\) of \(a\) such that \(x\in U\) implies that \(\displaystyle \left|\frac{f^{\prime}(x)}{g^{\prime}(x)}-L\right| < \varepsilon \). Take any \(x\in U\) and let \(y_n\in U\) be a monotone sequence converging to \(a\). According to Cauchy’s mean value theorem we have that there exists a sequence \((\xi_n)_{n=1}^{\infty}\) such that \(\xi_n\in (x,y_n)\) and \[\frac{f(x)-f(y_n)}{g(x)-g(y_n)}=\frac{f^{\prime}(\xi_n)}{g^{\prime}(\xi_n)}\in (L-\varepsilon, L+\varepsilon).\] Therefore the sequence \(\displaystyle \left(\frac{f(x)-f(y_n)}{g(x)-g(y_n)}\right)_{n=1}^{\infty}\) has all its terms in \((L-\varepsilon, L+\varepsilon)\). It remains to notice that \(\displaystyle\lim_{n\to\infty} \frac{f(x)-f(y_n)}{g(x)-g(y_n)}=\frac{f(x)}{g(x)}\) hence \(\displaystyle \frac{f(x)}{g(x)}\in (L-\varepsilon, L+\varepsilon)\) which completes the proof of the part (b) of the theorem. |
|