Definition 3.7.1 First indeterminate forms
Let \(a \in \mathbb{R}\) and let \(f(x)\) and \(g(x)\) be functions. If
then the limit
is called a \(\frac{0}{0}\) indeterminate form.
Let us return to limits (Chapter 1) and see how we can use derivatives to simplify certain families of limits called indeterminate forms. We know, from Theorem 1.4.3 on the arithmetic of limits, that if
and \(G\ne 0\text{,}\) then
\begin{align*} \lim_{x\rightarrow a}\frac{f(x)}{g(x)} &= \frac{F}{G} \end{align*}The requirement that \(G\ne 0\) is critical — we explored this in Example 1.4.7. Please reread that example.
Of course 1 Now it is not so surprising, but perhaps back when we started limits, this was not so obvious. it is not surprising that if \(F\ne 0\) and \(G= 0\text{,}\) then
and if \(F=0\) but \(G\neq 0\) then
However when both \(F,G=0\) then, as we saw in Example 1.4.7, almost anything can happen
Indeed after exploring Example 1.4.12 and 1.4.14 we gave ourselves the rule of thumb that if we found \(0/0\text{,}\) then there must be something that cancels.
Because the limit that results from these \(0/0\) situations is not immediately obvious, but also leads to some interesting mathematics, we should give it a name.
Let \(a \in \mathbb{R}\) and let \(f(x)\) and \(g(x)\) be functions. If
then the limit
is called a \(\frac{0}{0}\) indeterminate form.
There are quite a number of mathematical tools for evaluating such indeterminate forms — Taylor series for example. A simpler method, which works in quite a few cases, is L'Hôpital's rule 2 Named for the 17th century mathematician, Guillaume de l'Hôpital, who published the first textbook on differential calculus. The eponymous rule appears in that text, but is believed to have been developed by Johann Bernoulli. The book was the source of some controversy since it contained many results by Bernoulli, which l'Hôpital acknowledged in the preface, but Bernoulli felt that l'Hôpital got undue credit.
Let \(a\in\mathbb{R}\) and assume that
Then
We only give the proof for part (a). The proof of part (b) is not very difficult, but uses the Generalised Mean–Value Theorem (Theorem 3.4.38), which is optional and most readers have not seen it.
To prove part (b) we must work around the possibility that \(f'(a)\) and \(g'(a)\) do not exist or that \(f'(x)\) and \(g'(x)\) are not continuous at \(x=a\text{.}\) To do this, we make use of the Generalised Mean-Value Theorem (Theorem 3.4.38) that was used to prove Equation 3.4.33. We recommend you review the GMVT before proceeding.
For simplicity we consider the limit
By assumption, we know that
For simplicity, we also assume that \(f(a)=g(a)=0\text{.}\) This allows us to write
which is the right form for an application of the GMVT.
By assumption \(f'(x)\) and \(g'(x)\) exist, with \(g'(x)\) nonzero, in some open interval around \(a\text{,}\) except possibly at \(a\) itself. So we know that they exist, with \(g'(x)\ne 0\text{,}\) in some interval \((a,b]\) with \(b \gt a\text{.}\) Then the GMVT (Theorem 3.4.38) tells us that for \(x\in (a,b]\)
where \(c \in (a,x)\text{.}\) As we take the limit as \(x\to a\text{,}\) we also have that \(c\to a\text{,}\) and so
as required.