16

I am proving $(x^n)'=nx^{n-1}$ by the definition of the derivative:

\begin{align} (x^n)'&=\lim_{h \to 0} {(x+h)^n-x^n\over h}\\ &=\lim_{h \to 0} {x^n+nx^{n-1}h+{n(n-1)\over 2}x^{n-2}h^2+\cdots+h^n-x^n\over h} \\ &=\lim_{h \to 0} \left[ nx^{n-1}+{n(n-1)\over 2}x^{n-2}h+\cdots+h^{n-1} \right] \end{align}

Because polynomial is continuous for every $x$, we can conclude that $\lim_{x_0\to 0}(x_0)^n=0$. Therefore $$\lim_{h \to 0} \left[ nx^{n-1}+{n(n-1)\over 2}x^{n-2}h+\dots+h^{n-1} \right]= nx^{n-1}$$

Is this proof valid?

Cookie
  • 13,952
gbox
  • 13,645
  • 2
    this looks entirely correct – ncmathsadist Jun 22 '15 at 21:39
  • 4
    Well, unless $n$ is not an integer... but it seems you're set for the simple version. – MichaelChirico Jun 22 '15 at 21:40
  • 1
    The proof you give is correct only when $n$ is a positive integer. it is easy to extend the proof for negative integer $n$ as well. There are slight complications when proving the formula for rational $n$. The derivative formula amounts to the following standard limit $$\lim_{x \to a}\frac{x^{n} - a^{n}}{x - a} = na^{n - 1}$$ which I have proved in my post http://paramanands.blogspot.com/2013/11/teach-yourself-limits-in-8-hours-part-4.html (see "Proof of Standard Limits"). – Paramanand Singh Jun 23 '15 at 04:03

5 Answers5

13

It is correct, but too many ellipses tend to obscure a bit the proof. And ellipses are not very rigurous, if you ask me.

I'd rewrite it using this:

For $n\ge 2$ there exists some polynomial $P(x,h)$ such that $$(x+h)^n=x^n+nhx^{n-1}+h^2P(x,h)$$

ajotatxe
  • 66,849
11

The proof in OP is correct if $n$ is a positive integer. For a generic real exponent $a$ we can start from the derivative of the exponential function $y=e^x \rightarrow y'=e^x$.

From the inverse function differentiation rule we find $y=\log x \rightarrow y'=\dfrac{1}{x}$ and (using the chain rule):

$$ y=x^a =e^{a \log x} \rightarrow y'=e^{a \log x} (a\log x)'=e^{a \log x} \dfrac{a}{x}=ax^{a-1} $$

Emilio Novati
  • 64,377
  • I think you skipped the part when $n$ is rational. When $n$ is rational the derivative of $x^{n}$ can and should be calculated via algebraical means rather than using the theory of exponential and logarithmic functions. – Paramanand Singh Jun 23 '15 at 03:59
  • 2
    @Paramanand: This proof is valid for any $a \in \mathbb{R}$, so also rational. We can give a proof for rational exponents ( positive or negative) without use of the exponential function, but it is more laborious. Note that the derivative of the exponential function can be found very quickly by the definition of derivative using the fact that $e^{x+h}=e^xe^h$. – Emilio Novati Jun 23 '15 at 07:54
  • I can see this works for $x>0$, but polynomials are defined in the whole line, right? – anchova May 20 '25 at 05:30
2
Quote of the Question to Make it Easier to Reference in the Answer

I am proving $(x^n)'=nx^{n-1}$ by the definition of the derivative: $$ \begin{align} (x^n)'&=\lim_{h \to 0} {(x+h)^n-x^n\over h}\\ &=\lim_{h \to 0} {x^n+nx^{n-1}h+{n(n-1)\over 2}x^{n-2}h^2+\cdots+h^n-x^n\over h} \\ &=\lim_{h \to 0} \left[ nx^{n-1}+{n(n-1)\over 2}x^{n-2}h+\cdots+h^{n-1} \right] \end{align} $$ Because polynomial is continuous for every $x$, we can conclude that $\lim_{x_0\to 0}(x_0)^n=0$. Therefore $$\lim_{h \to 0} \left[ nx^{n-1}+{n(n-1)\over 2}x^{n-2}h+\dots+h^{n-1} \right]= nx^{n-1}$$ Is this proof valid?

Steps Towards the Answer

By application of the Binomial Theorem for example for $(x+y)^4$: (x+y)^4 The coefficient a in the term of $ax^by^c$ is known as the binomial coefficient ${\displaystyle {\tbinom {n}{b}}}$ or ${\displaystyle {\tbinom {n}{c}}}$ (the two have the same value). These coefficients for varying n and b can be arranged to form Pascal's triangle.

Pascal's Triangle

The derivative of a constant $C$ $\left(n=0\right)$ is zero since $\displaystyle \lim_{h \to 0}\frac{C-C}{h}=\displaystyle \lim_{h \to 0}\frac{0}{h}=0$. Now when $\displaystyle \lim_{h\to 0}$ then for the case $n=1$: $$x^{'}=\lim_{h \to 0}\frac{x+h-x}{h}=\lim_{h \to 0} \frac{h}{h}=1$$ Now consider $n \ge 2$ so that:

$$\lim_{h \to 0}\frac{(x+h)^n-x^n}{h}= \lim_{h \to 0}\frac{ {\tbinom {n}{n}} \,x^n + h \, {\tbinom {n}{n-1}} \,x^{n-1}+ {\tbinom {n}{n-2}}\,x^{n-2}*h^2 + \, ...\text{Higher Order Terms of $h$} -x^n}{h} $$ Since the binomial coefficient ${\tbinom {n}{n}}=1$ and the binomial coefficient ${\tbinom {n}{n-1}}=n$ and $x^2-x^2=0$, then: $$ \lim_{h \to 0}\frac{(x+h)^n-x^n}{h}= \lim_{h \to 0}\frac{ n \, h \, x^{n-1}+ {\tbinom {n}{n-2}}\,x^{n-2}*h^2 + \, ...\text{Higher Order Terms of $h$} }{h} $$ Taking the limit $\displaystyle \lim_{h \to 0}$ where $\displaystyle \lim_{h \to 0} h=0 $ and Higher Order Terms of h $\displaystyle \lim_{h \to 0} \, \text{and (Higher Order Terms of h) } \to 0 \, \text{ so: }$ $$ \begin{align} \lim_{h \to 0}\frac{(x+h)^n-x^n}{h}&= \lim_{h \to 0}\frac{ n \, h \, x^{n-1}+ {\tbinom {n}{n-2}}\,x^{n-2}*h^2 + \, ...\text{Higher Order Terms of $h$} }{h} \\ &=\lim_{h \to 0} n \, x^{n-1}+ {\tbinom {n}{n-2}}\,x^{n-2}*h + \, ...\text{Higher Order Terms of $h$} \\ &=\lim_{h \to 0} n \, x^{n-1}=n \, x^{n-1} \end{align} $$ So, $$\boxed{ \left(x^n\right)^{'}=\lim_{h \to 0}\frac{(x+h)^n-x^n}{h}=n \, x^{n-1} }$$ And the proof is complete. Aside from a few formalities and additional details, these were basically the same steps as summarized with the question's proof. So the summary proof in the question is complete and correct.

  • This is very similar to what I've done in calculus 1 classes (roughly mid 1980s to mid 2000s). I also usually wrote down two versions of Pascal's triangle, one of which is written using the binomial symbols. (But it's a lot easier to do that on a blackboard than here!) For a similar-type calculation of the derivative of $x^r$ where $r$ is any rational number, see my answer to Differentiation using first principles with rational powers. (Also, pinging @gbox in case still interested in these sorts of calculations.) – Dave L. Renfro Jun 25 '24 at 11:15
  • @DaveL.Renfro. Thank you so much for your comment and link. Nice job! – Stephen Elliott Jun 25 '24 at 12:21
  • Most of the elementary calculus texts I have taught out of make an argument similar to this---the idea, I think, is that students are supposed to have seen the Binomial Theorem in a precalculus class, so there is no problem using it in a calculus class. Unfortunately, it seems that many of them either have never seen the result, or have completely forgotten it by the time they get to calculus. As such, I am always made uncomfortable by this argument---I think that it makes more sense to just give an inductive proof of the Power Rule... – Xander Henderson Jun 26 '24 at 14:00
  • Otherwise, I like to give an inductive proof of the Binomial Theorem (this is what I do in my own precalculus classes). In either case, I always feel like something is missing in the "usual" presentation. This is not a knock against your argument---again, it is pretty standard. It is just an argument which I have always found a little discomfiting. There is a lacuna in there, somewhere, which bothers me. – Xander Henderson Jun 26 '24 at 14:02
0

Another proof of

$$(x^n)'=nx^{n-1}, \forall n \in \mathbb{N}$$

can be implemented with induction on $n$ using the product rule

$$(f \cdot g)' = f'g+fg'$$

and the trivial identity derivative

$$(x)'=1$$

Specifically:

  1. For $n=0$, $(x^0)'=(1)'=0=0 \cdot x^{-1}$ holds
  2. For $n=1$, $(x^1)'=1=1 \cdot x^{0}$ holds
  3. Assume that $(x^k)'=kx^{k-1}$ for some $k \in \mathbb{N}$ holds
  4. For $k+1$ using the product rule we have:
    $$(x^{k+1})' = (x \cdot x^k)'$$ $$(x^{k+1})' = x \cdot kx^{k-1}+1 \cdot x^k$$ $$(x^{k+1})' = (k+1)x^{k}$$
  5. Thus $(x^n)'=nx^{n-1}, \forall n \in \mathbb{N}$ holds. QED

Furthermore:

  1. $(x^{-n})'=-nx^{-n-1}, \forall n \in \mathbb{N}$ can be proved with chain rule on $((\frac{1}{x})^n)'$
  2. $(x^{1/n})'=\frac{1}{n}x^{1/n-1}, \forall n \in \mathbb{N}$ can be proved with inverse rule on $f^{-1}(x)=x^n \implies f'=1/{f^{-1}}'$
  3. $(x^{n/m})'=\frac{n}{m}x^{n/m-1}, \forall n,m \in \mathbb{N}$ can be proved with chain rule on $((x^{1/m})^{n})'$
  4. Similarly for $(x^{-n/m})'=-\frac{n}{m}x^{-n/m-1}, \forall n,m \in \mathbb{N}$
  5. Thus, $(x^q)'=qx^{q-1}, \forall q=\frac{n}{m} \in \mathbb{Q}$. QED
Nikos M.
  • 2,326
0

This is correct if $n\neq 0$: $\lim_{x_0\to 0}(x_0)^n=0$. Otherwise, it is $1$. You should have said that $\lim_{h\to 0}h^k=0$ for $k\geq 1$.

And your prove, as it is written, is valid for $n\in\Bbb N$. Here is a cute implicit differentiation proof for all $n\in\Bbb R$: $y=x^n\implies \ln y=n\ln x\implies \frac{y'}y=\frac nx\implies y'=nx^{n-1},$ since we know that $(\ln x)'=\frac1x.$

Bob Dobbs
  • 15,712