1

So I got this really strange contradiction that I couldn't explain to myself and I'd really love some clarification on the pitfalls that I'm potentially facing.

Let $$f(x) = \sqrt[3]{(x-1)(x-2021)}|\sin(\pi x)|$$

The very strange thing that's happening is that when I try to evaluate $f'(1)$ with the direct definition as: $$f'(1) = \lim_{x\to 1}\frac{f(x) - f(1)}{x - 1} = \lim_{x\to 1}(x-1)^{-\frac{2}{3}}(x-2021)^\frac{1}{3}|\sin(\pi x)| = 0$$

But when I try to explicitly find $f'(x)$ as an expression for $x$: $$f'(x) = \frac{(4x-4044)|\sin(\pi x)|^2 + 3\pi(x-2021)(x-1)\sin(2\pi x)}{6\left((x-1)(x-2021)\right)^\frac{2}{3}|\sin(\pi x)|}$$ and then plug in $x = 1$, I clearly get a $0$ in the denominator, hence $f'(1)$ is undefined.

There's obviously something smelly, but I find this one to be a really interesting case that I couldn't explain to myself. Any advice is welcome and appreciated! Thanks in advance!

D. Petrov
  • 851
  • 4
  • 12
  • 2
    It should be $(x-1)^{-2/3}$ in $f'(1)$ in first case. – Koro Mar 23 '21 at 02:23
  • @Koro Correct, I've fixed it, but that's not the point - the limit still exists and is 0. :D – D. Petrov Mar 23 '21 at 02:27
  • 1
    $$ \lim_{x \to 1} (x-1)^\alpha\sin(\pi x) =0 ;;;;; \alpha\in (-1,\infty) $$, so both limits go to zero. – WW1 Mar 23 '21 at 03:01
  • As many have pointed out, $$\begin{align} \lim_{x \to 1} (x - 1)^{-\frac{2}{3}} &= \lim_{x \to 1} \frac{1}{(x - 1)^{\frac{2}{3}}} \ &= \frac{1}{\lim_{x \to 1} {(x - 1)^{\frac{2}{3}}}} \ &= \frac{1}{{\left(\lim_{x \to 1} (x - 1)^{\frac{1}{3}}\right)^2}} \ &= +\infty \end{align}$$ because both $\frac{1}{x}, x^2$ are continuous near $x = 1$ and the limit has the form $``\frac{1}{(\pm0)^2}\text{"}$. – 光復香港 時代革命 Free Hong Kong Mar 23 '21 at 04:22
  • Thanks for the comments, I know that the limit is 0, the whole problem is with the contradiction between why it is 0 when directly evaluating with limits and OTOH it's undefined when plugging $x=1$ into the expression for $f'(x)$. Or did I not get something you guys implied with those comments? – D. Petrov Mar 23 '21 at 14:20

2 Answers2

3

$0$ is the correct derivative, which you got in the first calculation. I assume however that you overlooked something in the calculation, as the limit is far from simple.

When evaluating the derivation limit directly, you correctly get that it is the limit of a product of 3 terms:

$$f'(1)=\lim_{x\to 1}(x-1)^{-2/3}(x-2021)^{1/3}|\sin(\pi x)|$$

Now the 'easy' solution to this would be if we could calculate the limit of each term directly, then multiply their results. But it turns out this doesn't work, as one term tends to infinity:

$$\lim_{x\to 1}(x-1)^{-2/3} = \infty,$$ $$\lim_{x\to 1}(x-2021)^{1/3} = \sqrt[3]{-2020},$$ $$\lim_{x\to 1}|\sin(\pi x)| = 0,$$

the last 2 limits calculated simply be realizing that the functions taken to the limit are continuous at the given point. So one limit is $\infty$, one is $0$ and one is a finite, non-zero number. That means the limit of their product could be anything: $0$, some other number, $\infty$ or "diverging" in the strict sense (not tending to a finite number or infinity). We need to evaluate the limit of the product

$$\lim_{x\to 1}(x-1)^{-2/3}|\sin(\pi x)|,$$

and whatever we find as a result (if there is one at all), it will just have to be multplied by $\sqrt[3]{-2020}$ in order to get the derivative.

Now we have

$$(x-1)^{-2/3}|\sin(\pi x)| = \frac{|\sin(\pi x)|}{(x-1)^{2/3}} = \frac{|\sin(\pi x - \pi)|}{(x-1)^{2/3}} = \frac{|\sin(\pi(x - 1))|}{(x-1)^{2/3}} = (x-1)^{1/3}\frac{|\sin(\pi(x - 1))|}{(x-1)} = \pi(x-1)^{1/3}\frac{|\sin(\pi(x - 1))|}{\pi(x-1)}.$$

These are mostly simple equivalent transformations, where the fact that $|\sin(x)|$ has period $\pi$ has ben used for the second equality sign.

Now if you've seen a bit of trigonometric limits, the last term contains a part that is very near to what you may have seen before, if you set $y=\pi(x-1)$

$$\lim_{y\to 0}\frac{\sin(y)}y = 1$$

The only difference is that our term contains an absolute value. But that absolute value doesn't change that

$$\frac{|\sin(\pi(x - 1))|}{\pi(x-1)}$$

is bounded around $x=1$. So we see that

$$\pi(x-1)^{1/3}\frac{|\sin(\pi(x - 1))|}{\pi(x-1)}$$

is the product of 2 factors, the first of which ($\pi(x-1)^{1/3}$) tends to $0$ when $x\to 1$ and the second ($\frac{|\sin(\pi(x - 1))|}{\pi(x-1)}$) stays bounded when $x \to 1$. That means the product of those 2 functions tends to $0$ as well:

$$\lim_{x \to 0}\pi(x-1)^{1/3}\frac{|\sin(\pi(x - 1))|}{\pi(x-1)} = 0.$$

That finally means that

$$f'(1)=\lim_{x\to 1}(x-1)^{-2/3}(x-2021)^{1/3}|\sin(\pi x)| = \lim_{x \to 0}\pi(x-1)^{1/3}\frac{|\sin(\pi(x - 1))|}{\pi(x-1)} \times \lim_{x\to 1}(x-2021)^{1/3} = 0 \times \sqrt[3]{-2020}=0.$$

Now what happened when using the second method to calculate the derivative explicitly?

The function $f$ is given as the product of 2 functions, and it turns out that neither factor is differentiable at $x=1$. You can do the calculations yourself (you did already), but the Wolfram alpha picutures show that the root part has an "infinite" derivative:

picture of root factor of function f

The $\sin$ part OTOH has different left and right derivates (due to the absolute value function "flipping" the sin function at $x=1$):

enter image description here

So you simply can't use the product rule for the function at the point $x=1$, because that requires that both functions have a derivative at that point, which isn't true. One would guess that the product of 2 such functions would then also have no derivative for $x=1$, but the above calculation shows that is not true.

Finally, here comes the picture of how the function $f$ looks like around $x=1$:

enter image description here

It looks continuous, though one can't really see what the derivative at $x=1$ is, but that cleared up by the above calculation.

Ingix
  • 14,494
  • I see that the way I've simply put the result of the limit at the end has confused a lot of people. I knew it wasn't a simple limit, but the main point was that it existed and was 0, whereas the explicit function expression for $f'(x)$ was undefined for $x = 1$. So I guess you have pointed out the problem - in order to use the product rule, one has to assume both functions have a derivative at that point, right? But it still troubles me... how is $f'(x)$ the expression for the derivative of the product, when its undefined at $x=1$, but the derivative should be defined there and $f'(1) = 0$... – D. Petrov Mar 23 '21 at 14:18
  • 2
    Finding the derivation of the expression "the normal way" is correct when $x$ is not an integer. Then the $|\sin(\pi x)|$ part is not zero, so can be derived, and since the zeros of the root part are also integers, they don't give additioanl problem points.

    There is no "rule" that says the derivate of a function given by some formula must always exist and also be representable by a formula. Depending on what you use in the formula, the derivative may not exist at some points, and at other points the usual rules to combine the derivatives of the parts may not work, as explained here.

    – Ingix Mar 23 '21 at 14:30
  • 1
    See the answer by Lei Li to this question here https://math.stackexchange.com/questions/525860/can-a-continuous-function-have-a-non-continuous-derivative to see another often cited example where the usual way to calculate the derivate of a function breaks down at $x=0$, but the derivative still exists. – Ingix Mar 23 '21 at 14:33
0

$\lim_{x \to 1} (x-1)^{-2/3} =\lim_{x \to 1} \dfrac1{(x-1)^{2/3}} \to \infty$ since $\lim_{x \to 1}(x-1)^{2/3} =0 $.

marty cohen
  • 107,799