Problem 8b in Chapter 9 of Spivak's Calculus reads as follows:

Prove that if $g(x)=f(cx)$, then $g'(c)=c \cdot f'(cx)$.

Spivak's first part of the solution is written as:

$$\begin{align}g'(x)&=\displaystyle \lim_{h\to 0}\frac{g(x+h)-g(x)}{h}=\displaystyle \lim_{h\to 0}\frac{f(cx+ch)-f(cx)}{h} \\ &= \displaystyle \lim_{h\to 0}\frac{c[f(cx+ch)-f(cx)]}{ch}\color{red}{=} \displaystyle \lim_{k\to 0}\frac{c[f(cx+k)-f(cx)]}{k}\end{align}$$

My question is about the variable substitution that takes place with $k=ch$.

I previously wrote a post (found here: Question about my proof of: $\displaystyle \lim_{h \to 0}f(ch)=\displaystyle \lim_{ch \to 0}f(ch)$ for $c\neq 0$) that I hoped would clear up some of my confusion. The answer for this post alluded to the implicit usage of the following Theorem:

If we have $$\lim\limits_{x \to x_0} f(x) = \ell$$ for some $\ell \in \mathbb R$ (i.e. exists), and if $\varphi$ is a function such that $$\lim\limits_{t \to t_0} \varphi(t) = x_0$$ for some $t_0 \in \mathbb{R} \cup \{\pm \infty\}$, and $\varphi(t) \ne x_0 $ when $t$ is in some deleted nbhd of $t_0$, then $$\lim\limits_{t\to t_0} f(\varphi(t)) = \ell$$ that is $$\lim\limits_{x \to x_0} f(x) = \lim\limits_{t\to t_0} f(\varphi(t))$$

Although learning this theorem was valuable in and of itself, I am not sure I grasp how Spivak is employing it.

Firstly, I am unsure if it is correct to rewrite $\displaystyle \lim_{h\to 0}\frac{c[f(cx+ch)-f(cx)]}{ch}$ as $\displaystyle \lim_{h\to 0}F(c \cdot h)$. After playing around with several examples, I do not think such a conversion is acceptable (at least not in the context of applying the Theorem). If we were to go forward with this conversion, then that means if I could find an acceptable function of $k$...$\varphi (k)$...I should be able to claim:

$$\lim\limits_{h \to 0} f(c \cdot h) = \lim\limits_{k\to k_0} f(c \cdot\varphi(k))$$, but this does not feel right.

If anyone could budge me in the right direction, I would appreciate it.


Let $F(k) := \frac{c[f(cx + k) - f(cx)]}{k}$. Let $\varphi(h) = c h$. Then $F(\varphi(h)) = \frac{c[f(cx + ch) - f(cx)]}{ch}$. Your theorem implies $$\lim_{k \to 0} F(k) = \lim_{h \to 0} F(\varphi(h))$$ which is exactly the step that Spivak takes.


It's worthwhile using a general approach as you have done here. However, as an alternative, we can use a specific result for this problem. What follows is closer to Spivak's method, based partly on problem 5-14.

Lemma: $$\text{If }\lim_{x\to0} \frac{f(x)}{x} = \ell \text{ and }b \neq 0\text{, then }$$ $$\lim_{x\to 0} \frac{f(bx)}{x} = b\ell.$$

Lemma proof: Suppose $$\lim_{x\to0} \frac{f(x)}{x} = \ell.$$ This tells us for any $\varepsilon > 0$, there exists some $\delta > 0$ such that for all $x$ if $$0 < |x| < \delta \text{, then } \left|\frac{f(x)}{x} - \ell\right| < \varepsilon.$$

For this $\delta$, what if we have $$0 < |bx| < \delta, $$ or equivalently $$0 < |x| < \frac{\delta}{|b|}.$$

If $bx$ satisfies the $\delta$-requirement, we would have for all such $x$, $$\left|\frac{f(bx)}{bx} - \ell\right| < \varepsilon.$$

We set $\frac{\delta}{|b|} =\delta'$.

We see for all $x$ if $$0 < |x| < \delta' \text{, then } \left|\frac{f(bx)}{bx} - \ell\right| < \varepsilon,$$ or, $$\lim_{x\to 0} \frac{f(bx)}{bx} = \ell.$$ From this, we get $$\lim_{x\to 0} \frac{f(bx)}{x} = \lim_{x\to 0} b\cdot\frac{f(bx)}{bx}=b\ell.$$ $$\blacksquare$$ Now, returning to the problem:

9-8(b) Prove that if $g(x) = f(cx)$, then $g'(x) = c\cdot f'(cx)$.

Case 1: $c = 0$.

Suppose $c = 0$. In this case we have $$g(x) = f(0).$$ $$g'(x) = 0.$$ If $f'(0)$ exists, we indeed have $$ g'(x) = 0 = 0 \cdot f'(0)=c\cdot f'(cx).$$ However, if $f$ is not differentiable at $0$, the hypothesis will not be true. $g$ will still be constant, with $g'(x) = 0$, but we cannot write $g'(x) = c \cdot f'(0)$.

Case 2: $c \neq 0$. Suppose $c \neq 0$. From the definition of the derivative, we have \begin{align} g'(x) &= \lim_{h\to 0} \frac{g(x+h) - g(x)}{h}, \\ &= \lim_{h\to 0} \frac{f(c[x+h]) - f(cx)}{h}, \\ &= \lim_{h\to 0} \frac{f(cx+ch) - f(cx)}{h}. \end{align}

Similarly, if $f$ is differentiable at $cx$, we have $$f'(cx) = \lim_{h\to 0} \frac{f(cx + h) - f(cx)}{h}.$$ Let's take the numerator here and create a new function $F(h)$, with $$F(h) = f(cx + h) - f(cx).$$ By definition, $$\lim_{h\to 0} \frac{F(h)}{h} = f'(cx).$$ From our lemma, if $c \neq 0$ we have $$\lim_{h\to 0} \frac{F(ch)}{h} = cf'(cx),$$ but also \begin{align} \lim_{h\to 0} \frac{F(ch)}{h} &= \frac{f(cx+ch) - f(cx)}{h}, \\ &= g'(x). \end{align}

In other words, If $g'(x)$ exists, then $$g'(x) = \lim_{h\to 0} \frac{f(cx+ch) - f(cx)}{h} = c \cdot \lim_{h\to 0} \frac{f(cx+ h) - f(cx)}{h} = c \cdot f'(cx).$$ $$\blacksquare$$

  • 1
    $\begingroup$ Cheers for the alternative approach! $\endgroup$
    – S.Cramer
    Jul 28 at 8:15

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.