Proving piecewise function is k-differentiable

member 731016
Homework Statement
Please see below
Relevant Equations
Please see below
For this problem,
1716937553405.png

My solution is,

##F(x)=\left\{\begin{array}{ll} e^{-\frac{1}{x}} & \text { if } x>0 \\ 0 & \text { if } x \leq 0\end{array}\right.##

The we differentiate both sub-function of the piecewise function. Note I assume differentiable since we are proving a result that the function is differentiable, so I assume that ##F^{k}(0)## exists, that is the function is k-differentiable at zero assuming the limit exists. I will prove for first derivative k = 1 below.

##F'(x)=\left\{\begin{array}{ll} -\frac{1}{x}e^{-\frac{1}{x}} & \text { if } x>0 \\ 0 & \text { if } x \leq 0\end{array}\right.##

##F''(x)=\left\{\begin{array}{ll} \frac{1}{x^2}e^{-\frac{1}{x}} & \text { if } x>0 \\ 0 & \text { if } x \leq 0\end{array}\right.##

One can continue considering ##k## number of cases

Thus from our proof, the polynomial ##P_k## can be written explicitly as ##P_k = \frac{1}{x^{k - 1}(-1)^k}## for ##k > 0##

Thus we consider two cases for ##k##,

##
F^{(k)}(x)= \begin{cases}\frac{1}{x^{k - 1}(-1)^k}\left(x^{-1}\right) \exp \left(-x^{-1}\right) & \text { if } x>0 \\ 0 & \text { if } x \leq 0\end{cases}
##

for ##k > 0##

And for ##k = 0##, we have the original piece wise function,

##F(x)=\left\{\begin{array}{ll} \exp \left(-x^{-1}\right) & \text { if } x>0 \\ 0 & \text { if } x \leq 0\end{array}\right.##

QED

So as part of our proof, we also consider a sub-proof of the reason why the function is ##F^{k}## at ##x = 0##. We just consider the trival case for ##k = 1##,

In order to be differentiable at zero then ##lim_{x \to 0} F(x)## must exist. This is equivalent to both the Left hand and right hand limits existing.

##\lim_{x \to 0^-} F(x) = \lim_{x \to 0^-} 0 = 0##. Therefore, left hand limit exists.

Now for right hand limit, we have,

##\lim_{x \to 0^+} e^{-\frac{1}{x}}## DNE as I don't know the limit as ##\frac{1}{x}## goes to zero. However, for some reason, wolfram alpha says that ##\lim_{x \to 0^+} e^{-\frac{1}{x}} = 0##. Does someone please know why this is true?

I will assume that that is true, and I think the next part of the sub-proof is to generalize to higher order derivative

In order to find ##F^{k}(0)## then this limit ##\lim_{x \to 0} F^{k}## must exist.

That is, ##\lim_{x \to 0^-} F^{k}(x) = \lim_{x \to 0^+} F^{k}(x)##

##\lim_{x \to 0^-} F^{k}(x) = \lim_{x \to 0^-} 0 = 0##

##\lim_{x \to 0^+} F^{k}(x) = \lim_{x \to 0^+} \frac{1}{x^k(-1)^k}e^{-\frac{1}{x}} = \frac{1}{(-1)^k}\lim_{x \to 0^+} \frac{1}{x^k}e^{-\frac{1}{x}} = 0##

Does someone also please know whether this a proof by induction (or what sort of proof it is)? Or can we just use a informal generalization proof?

Thanks!
 
Physics news on Phys.org
Please remind yourself of the correct definition of the chain rule ##\frac{d}{dx}f(g(x))=f'(g(x))g'(x)## and re-check your work for $$\frac{d}{dx}e^{-1/x}.$$

And you haven't yet shown that ##P_k = \frac{1}{x^{k - 1}(-1)^k}## is true for ##k>2##. There are specific requirements for a valid proof by induction, and you should try to follow the steps (even if it's boring) until you understand how it works.

In order to be differentiable at zero, you need the function to be continuous at 0 and have $$\lim_{h\to 0^+}\frac{f(x+h)-f(x)}{h} =\lim_{h\to 0^-}\frac{f(x+h)-f(x)}{h}.$$
By using the hospital rule, you should be able to show that both are equal to 0 for any ##k## (this would be another proof by induction).
 
  • Like
  • Love
Likes SammyS and member 731016
ChiralSuperfields said:
##\lim_{x \to 0^+} e^{-\frac{1}{x}}## DNE as I don't know the limit as ##\frac{1}{x}## goes to zero. However, for some reason, wolfram alpha says that ##\lim_{x \to 0^+} e^{-\frac{1}{x}} = 0##. Does someone please know why this is true?
$$\lim_{x\to 0^+}e^{-\frac{1}{x}}=\lim_{y\to \infty}e^{-y}=0.$$
 
  • Love
Likes member 731016
Thank you for your reply @docnet! I have some question about this solution,

1717043390531.png


Do you please know how they go from the ##f^{n}(x)## to ##f^{n + 1}(x)## step (I assume they differentiate ##f^{n}(x)##)? I think they made a typo there is something because they did not seem to differentiate the ##\frac{1}{x}## since it is three functions multiplied together so you have to use the product rule on ##p_nu## where ##u = \frac{1}{x}e^{-\frac{1}{x}}##.

I also don't understand how they got ##p_1(x) = x^2##. Should it not be ##p_1(x) = \frac{1}{x}##?

Thanks!
 
ChiralSuperfields said:
Thank you for your reply @docnet! I have some question about this solution,

View attachment 346210

Do you please know how they go from the ##f^{n}(x)## to ##f^{n + 1}(x)## step (I assume they differentiate ##f^{n}(x)##)? I think they made a typo there is something because they did not seem to differentiate the ##\frac{1}{x}## since it is three functions multiplied together so you have to use the product rule on ##p_nu## where ##u = \frac{1}{x}e^{-\frac{1}{x}}##.

I also don't understand how they got ##p_1(x) = x^2##. Should it not be ##p_1(x) = \frac{1}{x}##?

Thanks!
Let's take your last question first.

There is no way that ##\displaystyle \ p_1(x) = \frac{1}{x} \ ##. Fir one thing, that's not a polynomial at all.

If as they say, ##\displaystyle \ p_1(x) = x^2 \ , \ ## that then says that ##\displaystyle \ p_1\left( \frac 1 x \right) = \left( \frac 1 x \right)^2 \ , \ ##

Perhaps you're still having difficulties with ##\displaystyle \ F\,'(x) \ .\ ## As @docnet already told you:
docnet said:
Please remind yourself of the correct definition of the chain rule ##\frac{d}{dx}f(g(x))=f'(g(x))g'(x)## and re-check your work for $$\frac{d}{dx}e^{-1/x}.$$
So, no, ##\displaystyle \ p_1\left( \dfrac 1 x \right) \ne \frac{1}{x} \ ## because ##\displaystyle \ \frac{d}{dx}e^{-1/x}\ne
\frac{1}{x}e^{-1/x} \ .##

As for differentiating ##\displaystyle F^{(n)}(x)\,, \ ## it is the product of only two functions, not three. Note that the chain rule is involved in differentiating each of those two component functions.
 
Last edited:
  • Like
  • Love
Likes member 731016 and docnet
ChiralSuperfields said:
Thank you for your reply @docnet! I have some question about this solution,

View attachment 346210

Do you please know how they go from the ##f^{n}(x)## to ##f^{n + 1}(x)## step (I assume they differentiate ##f^{n}(x)##)?
Yes, differentiating ##f^n(x)## is a crucial step of the induction procedure.
ChiralSuperfields said:
I think they made a typo there is something because they did not seem to differentiate the ##\frac{1}{x}## since it is three functions multiplied together so you have to use the product rule on ##p_nu## where ##u = \frac{1}{x}e^{-\frac{1}{x}}##.

I also don't understand how they got ##p_1(x) = x^2##. Should it not be ##p_1(x) = \frac{1}{x}##?

Thanks!
##p_n\left(\frac{1}{x}\right)## indicates ##p_n## is a function of ##\frac{1}{x}##, like how ##f(x)## indicates ##f## is a function of ##x##.

ChiralSuperfields said:
I also don't understand how they got ##p_1(x) = x^2##. Should it not be ##p_1(x) = \frac{1}{x}##?
You should have ##p_1=\frac{1}{x^2}## because $$\frac{d}{dx}e^{-x}=\frac{1}{x^2}e^{-x}.$$

##\frac{1}{x^2}## isn't a polynomial but ##x^2## is one. It's true that
$$p_1\left(\frac{1}{x}\right)=\frac{1}{x^2}\Longleftrightarrow p_1(x)=x^2.$$
 
Last edited:
  • Like
  • Love
Likes SammyS and member 731016
There are two things I don't understand about this problem. First, when finding the nth root of a number, there should in theory be n solutions. However, the formula produces n+1 roots. Here is how. The first root is simply ##\left(r\right)^{\left(\frac{1}{n}\right)}##. Then you multiply this first root by n additional expressions given by the formula, as you go through k=0,1,...n-1. So you end up with n+1 roots, which cannot be correct. Let me illustrate what I mean. For this...
Back
Top