-1

We use the chain rule for functions that can be seen as composed functions. However, for a simple function like $y=x^2$ we can also say that this is just a composition of 2 functions: $y=x$ and $y=(\text{something})^2$

Now, if we use the chain rule instead of the power rule, it still works and we still get $y'=2x$ But I'm wondering if the "power rule" is really just a shortcut of the chain rule for polynomials (polynomials meaning, no composite functions).

*NOTE: the answer was given to me in one of comments, which is basically that the power rule is applied when the composed functions are only the identity function (f(x)=x) and that of course applies only in polynomials.

  • That doesn't really answer my question. Is the Chain Rule ALWAYS an alternative to using the power rule? – Pineapple29 Sep 22 '16 at 21:58
  • 2
    Chain Rule can be applied in any situation, but if the derivative of the inside function is $1$, how does it contribute to the overall derivative? – Simply Beautiful Art Sep 22 '16 at 22:01
  • You're missing the bigger question. I just gave an example to a function that you would normally apply the Power Rule on, but I'm asking if the Power Rule can be substituted for the Chain Rule in ANY instance. Don't focus on the specific example I gave. – Pineapple29 Sep 22 '16 at 22:03
  • 2
    $y=x^2$ is the same as $y = (something)^2$ it's just you are writing it without coordinates, in both cases you're differentiating the squaring function. Furthermore, you can always write $f = f\circ id$ but doing so is rather pointless as it just adds work to make it into more of a clear composition. – Adam Hughes Sep 22 '16 at 22:06
  • 1
    "the "power rule" is really just a shortcut of the chain rule for polynomials": what the hell does this mean ? –  Sep 22 '16 at 22:07
  • It means, as explained in a comment below, that you can always view a function as composition of multiple functions, and that's why you'd use the chain rule. But when it's just a polynomial, then you'd use the Power Rule, not the chain rule, even though you can view the polynomial as a composition of multiple functions, such as y=x^2 – Pineapple29 Sep 22 '16 at 22:15
  • 1
    @Pineapple29 with the composition as you explain it in your questions, one of the functions in your composition is still $f(x) = x^n$. So even if you've "made the power rule problem into a chain rule problem", you're still using the power rule within your implementation of the chain rule. – Ben Grossmann Sep 22 '16 at 22:23
  • @Pineapple29: "you can view the polynomial as a composition of multiple functions". Ah, how ? Could you show it with the polynomial $x+x^2$ ? –  Sep 22 '16 at 22:24
  • Yes, of course, any function $f(x)$ can be thought of as the function $f(\text{something})$ combined with the function $x$, but when you expand this with the chain rule you just get $1\cdot f'(x)$, which doesn't help you. – Jack M Sep 22 '16 at 22:25

4 Answers4

3

I would say that you can always make a power rule problem into a chain rule problem in an unenlightening way. Yes, if $f(x) = x^n$, then we can define $g(x) = x$, so that $$ \frac{d}{dx}f(x) = \frac{d}{dx}f(g(x)) = f'(g(x))g'(x) = n [g(x)]^{n-1} \cdot 1 = nx^{n-1} $$ Note, however, that we would need to know the power rule in order to find the "outer" derivative $f'(x)$.

Slightly more interesting is the fact that if $f(x) = x^n$ and $g(x) = x^m$, then we can write $$ \frac{d}{dx}x^{mn} = \frac{d}{dx}(x^m)^n = \frac{d}{dx}f(g(x)) = n[g(x)]^{n-1} g'(x) = n[x^{m}]^{n-1} \cdot mx^{m-1} =\\ (mn)x^{m(n-1) + m-1} = (mn)x^{mn - 1} $$ Of course, it is easier to use the power rule outright.


We could derive the power rule (for real numbers $n$) using the chain rule together with the rule $\frac{d}{dx} \ln(x)= \frac 1x$. In particular, if we define $y = x^n$, we can use "logarithmic differentiation". $$ y = x^n \implies\\ \ln(y) = \ln(x^n) \implies\\ \ln(y) = n\ln(x) \implies\\ \frac{d}{dx}\ln(y) = \frac{d}{dx}n \ln(x) \implies\\ \frac{1}y \frac{dy}{dx} = \frac{n}{x} \implies\\ \frac{dy}{dx} = \frac{ny}{x} $$ If we substitute $y$ back in, we find $$ \frac{dy}{dx} = \frac{n(x^n)}{x} = nx^{n-1} $$


We could also derive the power rule (for integers $n$) using just the multivariate chain rule. In particular, define $$ f(x_1,\dots,x_n) = \prod_{k=1}^n x_n, \quad g(t) = (t,\dots,t) $$ We then find that $$ \frac{\partial f}{\partial x_k} = \frac{1}{x_k} \prod_{k=1}^n x_n, \qquad g'(t) = (1,\dots,1) $$ It therefore follows that $$ \frac{d}{dt}f(g(t)) = \sum_{k=1}^n \frac{\partial f}{\partial x_k}(g(t)) \cdot g'(t) = \cdots = nt^{n-1} $$ so that's kind of neat.

Ben Grossmann
  • 225,327
  • Thanks, you're getting closer, but I'd like a more definitive answer than "I would say that you can..." :) – Pineapple29 Sep 22 '16 at 22:07
  • My uncertainty is not over whether my statement is true. My uncertainty is over whether my statement answers your question, since it isn't clear what exactly you're asking for. – Ben Grossmann Sep 22 '16 at 22:08
  • It's taught that to use the chain rule we need to write the function as a composition of multiple functions. But it's always ignored that even y=x^2 can be separated into a composition of 2 functions. That's why it's unclear to me where the distinction would be to using the chain rule or the power rule, because the distinction can't be just "viewed as a composition of multiple functions" as I've just explained – Pineapple29 Sep 22 '16 at 22:13
  • @Pineapple29 how about this for a distinction: if one of the functions in your composition is $g(x) = x$, then there's no reason to be using the chain rule. – Ben Grossmann Sep 22 '16 at 22:15
  • Thank you, that's actually an important distinction that isn't mentioned anywhere. That's why I described it as the special case for "polynomials", because polynomials are basically the identity function (f(x)=x) composed multiple times, and that's why we don't need to use the Chain Rule, because the Power Rule is sufficient in that case. – Pineapple29 Sep 22 '16 at 22:22
1

$x^n := \exp (\log x \cdot n)$ hence the derivative is

$\frac nx \cdot\exp (\log x \cdot n) = n \frac{\exp (\log x \cdot n)}{\exp \log x} = n\exp(\log x \cdot n - \log x) = n \exp(\log x(n-1)) = nx^{n-1} $

So yes, with this definition, the power rule is just a special case of the chain rule. This definition of $x^n$ is nice when $n \in \mathbb{R} \setminus \mathbb{Q}$.

  • This answer is my personal favorite becuase it shows that we can actually calculate the derivate WITHOUT the power rule and WITH the chain rule, just as required. – Peter Sep 22 '16 at 22:28
  • If I could accept this question , I would do it... – Peter Sep 22 '16 at 22:29
  • You've proven that the power rule is a special case of the chain rule, when the correct functions are applied. But that's why we have power rule, as a shortcut. – Kaynex Sep 22 '16 at 22:56
0

Your question is a nonsense, the chain rule is no substitute for the power rule.

Indeed, by the chain rule where you see the function as the composition of the identity ($f(x)=x$) and a power we have

$$(f^r(x))'=f'(x)\frac{df^r(x)}{df}=1\cdot rf(x)^{r-1}=rx^{r-1}.$$

and in this development we needed the power rule anyway.

0

The chain rule is best defined as: $$\frac{dy}{dx} = \frac{dy}{du}\frac{du}{dx}$$ This shows off what the chain rule does mechanically; it allows you to write any part of the function as $u$ and take the derivative in terms of that. If you are simply replacing $x$ with $u$ then the chain rule is applied as: $$\frac{dy}{dx} = \frac{dy}{dx}\frac{dx}{dx}$$ $$\frac{dy}{dx} = \frac{dy}{dx}$$ Which, as you can see above, becomes redundant. As I said above, the chain rule allows you to replace a function with another, and replacing a function ($x$) with the same function ($u$) means you're not actually chain ruling anything.

Kaynex
  • 2,448