Derivative of inverse function using $dx/dy$ - what is going on here?

1.3k Views Asked by At

I'm about to learn about differentiating inverse functions at school, and the formula we're being told we'll be using is [assuming $g(x)=f^{-1}(x)$]:

$$f'(x)=\frac{1}{g'(f(x))}$$

In other places online, however, I have seen a much simpler formula, $dx/dy * dy/dx = 1$. (I am quite interested in Leibniz notation, as the old intuitive ideas often stun me in their elegance.)

However, here I cannot for the life of me figure out the connection between the Leibniz notation above and the more complicated "prime" notation above. I have searched far and wide across the Internet, but am still confused. Can anyone explain? How would I calculate the derivative of an inverse function using Leibniz notation, and how does that connect to the process of computing an inverse function's derivative using "prime" notation?

4

There are 4 best solutions below

10
On BEST ANSWER

I think that your teacher needs to clarify the notation with examples.

Definition: Let $g(x)$ be injective. Then $f(x)$ is an inverse function of $g(x)$ provided

$$g(f(x)) = f(g(x)) = x$$

Example: Suppose $g(x)=2x-1$, which is injective. Then, $g^{-1}(x)$ can be found by switching $x$ and $y$ and then solving for $y$

$$x=2y-1\implies y=\frac{x+1}{2} \implies g^{-1}(x)=\frac{x+1}{2}$$

Alternatively, if we define $f(x)=\dfrac{x+1}{2}$ then

$$g(f(x))=g\left(\frac{x+1}{2}\right)=2\left(\frac{x+1}{2}\right)-1=x$$

and

$$f(g(x))=f\left(\frac{(2x-1)+1}{2}\right)=x$$

so we see that the definition holds in this example. This leads to the result provided by your course notes.

Theorem: Let $g(x)$ be a function that is differentiable on an interval $I$. If $g(x)$ has an inverse function $f(x)$, then $f(x)$ is differentiable at any $x$ for which $g'(f(x))\neq 0$. Moreover,

$$f'(x)=\frac{1}{g'(f(x))},~~g'(f(x))\neq 0$$

To prove this theorem, we start with the definition

$$g(f(x)) = x$$

and then differentiate implicitly

$$\frac{d}{dx}\Big[g(f(x))\Big]=\frac{d}{dx}(x)$$

where we set $y=g(u)$ and $u=f(x)$ and then use the chain rule

$$\frac{dy}{dx}=\frac{dy}{du}\frac{du}{dx}=g'(u)f'(x)=g'(f(x))f'(x)$$

and since $\frac{d}{dx}(x)=1$, we know that

$$g'(f(x))f'(x)=1$$

then since $g'(f(x))\neq 0$, we can divide by it to form

$$f'(x)=\frac{1}{g'(f(x))}$$

Example: Suppose $x>0$ and

$$g(x)=x^2$$

then $f(x)=\sqrt{x}$ is its inverse. We have

$$f'(x)=\frac{1}{2\sqrt{x}}$$

and

$$g'(x)=2x$$

therefore

$$g'(f(x))=2(f(x))=2\sqrt{x} \implies \frac{1}{g'(f(x))}=\frac{1}{2\sqrt{x}}$$

so we see that

$$f'(x)=\frac{1}{g'(f(x))}$$


In terms of Leibniz notation, we can adjust the proof so that all of the primes are replaced by differentials. In this notation $$f'(x)=\frac{df(x)}{dx}=\frac{df}{dx}$$

and letting $y=f(x)$ and $x=g(y)$ forms

$$\frac{dy}{dx}=\frac{df}{dx},~~g'(f(x))=\frac{dg(y)}{dy}=\frac{dx}{dy}$$

so that

$$\frac{dy}{dx}\frac{dx}{dy}=1 \implies \frac{dy}{dx}=\frac{1}{\frac{dx}{dy}}$$

or

$$\frac{df}{dx}=\frac{1}{\frac{dg(y)}{dy}}$$

and we once again arrive at

$$f'(x)=\frac{1}{g'(f(x))}$$

as desired.

Example: Suppose that $y=f(x)=e^x$. Then, we want to show that

$$f'(x)=\frac{1}{g'(f(x))}$$

We are given

$$\frac{df}{dx}=\frac{dy}{dx}=e^x$$

Next, we can deduce that $x=g(y)=\ln(y)$ so that $g$ is the natural log function (the inverse of $e^x$ is the natural log function). Then

$$g'(f(x))=\frac{dx}{dy}=\frac{dg(y)}{dy}=\frac{d}{dy}\ln(y)=\frac{1}{y}=\frac{1}{e^x}$$

Therefore,

$$\frac{1}{g'(f(x))}=e^x$$

In terms of prime notation versus Leibniz notation, it is largely a matter of personal preference. Many people prefer using the Leibniz notation because the chain rule can be easily identified as

$$\frac{dy}{dx} = \frac{dy}{du} \frac{du}{dx}$$

instead of

$$\Big[f(g(x))\Big]' = f'(g(x))g'(x)$$

There is also the additional advantage that writing

$$\frac{dy}{dx}$$

makes it abundantly clear that $x$ is the independent variable and $y$ is the dependent variable. This can sometimes be confusing when you have multiple variables with primes.

I often switch between the two notations as needed. It might be best to work with both notations if the primes confuse you.

0
On

Note that $y=f(x)$, $x=g(y)$ $$\begin{align} \frac{dy}{dx}\cdot\frac{dx}{dy}&=1\\ \frac{dy}{dx}&=\frac{1}{\frac{dx}{dy}} \\ \frac{df(x)}{dx}=\frac{1}{\frac{dg(y)}{dy}} \\ f'(x)=\frac{1}{g'(f(x))} \end{align}$$

0
On

$g(x)=f^{-1}(x)$ means that $g(f(x))=x$. Take the derivative of both sides and we get $f'(x)g'(f(x))=1$ by the chain rule. But that's it as $y=f(x)$ and $x=g(y)$ implies $dx/dy = g'(y) = g'(f(x))$ and we're done.

0
On

I think this is best explained with physical quantities by interpreting the derivatives in terms of slope of a function. For example the kinetic energy depends on the speed of an object by the formula $E=\frac{1}{2} m v^2$. We can write this as $E = f(v)$. Suppose the object has mass $3\ kg$ and speed $8\ m s^{-1}$. Its kinetic energy is $\frac{1}{2} 3 \times 8^2 = 96\ J$. Now the slope of $f$ at that point is $24$ because of the relation $d E = m v d v$. It means that an increase of the speed by $1\ $ metre per second results in a increase of energy of $24$ joules. The meaning of the formula is that the slope of the inverse function is the inverse of the slope of $f$, that is to say $\frac{1}{24}$. Thus an increase of $1 J$ of kinetic energy results in an increase of $\frac{1}{24} m s^{-1}$ in speed. To sum things up, if $v=g(E) = f^{-1}(E)$, we have \begin{equation} f(8)=96\qquad g(96)=8\qquad f'(8) = 24\qquad g'(96) = \frac{1}{24} = \frac{1}{f'(8)} = \frac{1}{f'(g(96))} \end{equation} The key point to understand is that we take $g$ and $g'$ at the point $96 = f(8)$, but we take $f$ and $f'$ at the point $8 = g(96)$.