8

Let $f: \mathbb{R} \to \mathbb{R}$ be a function such that $f(x) = f^{(4)}(x)$ with $f(0) = f’(0) = f’’(0) = f’’’(0) = 0.$ Prove $f(x) = 0$ for all $x \in \mathbb{R}$

My Attempts:

Suppose $x \in \mathbb{R}$. Note that $\displaystyle f'(0) = \lim_{x\to 0} \frac{f(x)-f(0)}{x-0} = \lim_{x\to 0} \frac{f(x)}{x} = \lim_{x\to 0} \frac{f'(x)}{1}=0$. (L'Hôpital's Rule was used in the second to last limit due to the form $\frac{0}{0}$). With this approach, I am not necessarily finding if $x = 0$ on the whole real line. This led me to a different approach:

Suppose $x \in [0,b]$. By Mean Value Theorem, there exists $c \in (0,b)$ so that $\displaystyle \frac{f(b)-f(0)}{b-0} = f'(c)$. This approach doesn't bring me anywhere either, even if I repeatedly use Mean Value Theorem. Any suggestions on how to proceed and conclude? (I am currently reading/finishing the chapter on Differentiation in baby Rudin.)

operatorerror
  • 29,103
  • Have you tried proving this by contradiction? – Yunus Syed Feb 02 '18 at 01:44
  • 1
    Try $f(x)=\begin{cases}e^{-\frac{1}{x^2}}&\text{when }x\neq 0\0&\text{when }x=0\end{cases}$.

    This gives a Taylor polynomial of 0, so it can't help much.

    – ElfHog Feb 02 '18 at 01:47
  • @астонвіллаолофмэллбэрг not necessarily real analytic. – AHusain Feb 02 '18 at 01:47
  • By repeatedly using the mean value theorem you get a sequence of $x_i$ such that $f (x_i ) \neq 0$. Can you get a contradiction from that? – AHusain Feb 02 '18 at 01:54
  • what does $f^4(x)$ means? – Bad English Feb 02 '18 at 01:58
  • 1
    @BadEnglish $f^{(4)}(x)$ means the 4-th derivative of $f$ at $x$. – Masacroso Feb 02 '18 at 02:03
  • Well. It’s a fourth order linear ode with zero initial conditions. So... – Jacky Chong Feb 02 '18 at 02:05
  • @AHusain Could you give another hint following what you mentioned about those sequence of nonzero points? If I’m trying to show the function is $0$ for all $x$, how would nonzero terms prove the result? (Are you considering the contrapositive?) – Dominated Convergence Theorem Feb 02 '18 at 02:20
  • $f = c_1 e^{x} + c_2 e^{-x} + c_3\cos x + c_4\sin x$ But with those initial conditions your are not getting started. – Doug M Feb 02 '18 at 02:21
  • $f^{(4n+k)}(0) = f^{(k)}(0) = 0$ which means that the Taylor polynomial at $0$ is the $0$ polynomial. – Doug M Feb 02 '18 at 02:23
  • Nice question. +1 I have used Taylor's theorem in my answer, but you can modify it to use just the mean value theorem and you will have the inequality $|f(x) |\leq x^4|f(c)|$ and this also works in similar manner. – Paramanand Singh Feb 02 '18 at 05:05
  • @ParamanandSingh I was actually starting to consider using Taylor's Theorem but I couldn't see how to apply it. It's great that you were able to, I really liked the proof you gave. I feel that even though I keep practicing these problems, I am not effectively studying and learning these concepts. Do you have any suggestions, or mind explaining how you came up with your proof? (And how to generally approach problems like these and study them effectively in Analysis? I'm trying to improve.) – Dominated Convergence Theorem Feb 02 '18 at 05:14
  • To be frank, you need some experience doing a variety of problems and there is no way other than putting more effort. However your job is easy if you get the right book to have good conceptual foundation. I studied calculus on my own via Hardy's A Course of Pure Mathematics. – Paramanand Singh Feb 02 '18 at 05:17
  • When I wrote the Taylor's formula for degree $4$ I could see the recursive nature of the equation and one just needs to convert this into an inequality to complete the recurrence and finish the proof. But perhaps this might not be obvious to everyone (especially beginners). – Paramanand Singh Feb 02 '18 at 05:19
  • @ParamanandSingh I still am a beginner, so I can see what you are saying. I haven't heard about that book, I will look into it. If you're looking for another analysis question, I just posted one up regarding how to rigorously find the value of a derivative at a point. (I think it requires induction near the end and Mean Value Theorem, but I am unsure.) – Dominated Convergence Theorem Feb 02 '18 at 05:30

4 Answers4

3

The general solution to the ode is the sum $$ c_1e^x+c_2e^{-x}+c_3\sin x+c_4\cos x=f(x) $$ using the usual methods.

This gives us the system of equations $$ f(0)=c_1+c_2+c_4=0\\ f'(0)=c_1-c_2+c_3=0\\ f''(0)=c_1+c_2-c_4=0\\ f'''(0)=c_1-c_2-c_3 $$ which has unique solution $c_1=c_2=c_3=c_4=0$.

operatorerror
  • 29,103
  • In Analysis aren’t we supposed to not consider the sine and cosine functions? I’m not sure if that is rigorous enough (it is clear on how to get this solution, but I was looking for more of a Mean Value Theorem, sequence, epsilon-delta, etc type proof). – Dominated Convergence Theorem Feb 02 '18 at 04:21
  • why can't you consider sine and cosine in analysis? – operatorerror Feb 02 '18 at 04:23
  • @Primes: there is nothing wrong in using any function and in particular circular functions. If you want to avoid the theorem from ode then you can directly solve the equation step by step and show that the function must of the form given in this answer. – Paramanand Singh Feb 02 '18 at 04:30
2

Here is one proof which avoids differential equations. Let us assume that $|x|\leq 2$. By Taylor's theorem we have $$f(x) =\frac{x^4}{4!}f^{(4)}(c_1)=\frac{x^4}{4!}f(c_1), x\neq 0$$ where $0<|c_1|<|x|$ and applying this repeatedly and noting that at each step we have $0<|c_n|<|x|$ we get $$|f(x)| \leq \left(\frac{x^4}{4!}\right)^{n}|f(c_n) |\leq \left(\frac{x^4}{24}\right)^{n}M$$ where $M$ is the maximum value of $|f|$ between $0$ and $x$. Since $|x|\leq 2$ by Squeeze Theorem we can see that $f(x) =0$ for all $x\in[-2,2]$. The proof is extended to other values of $x$ by using functions $g_1(x) =f(x-1),h_1(x)=f(x+1)$ and applying the above argument repeatedly to $g_1, h_1$ and their descendents $g_n, h_n$ given recursively as $$g_n(x) =g_{n-1}(x-1),h_n(x)=h_{n-1}(x+1)$$ One may also use the function $g(x) =f(kx) $ to extend the proof to all values of $x\in[-2k,2k]$ with much less effort.

Incidentally the same technique gives another proof of the following important theorem:

Theorem: If $f:\mathbb{R} \to\mathbb {R} $ is such that $f'(x) =f(x), f(0)=0$ then $f(x) =0$ for all $x \in\mathbb {R} $.


Update: As mentioned in my comment to the question, one can also use mean value theorem four times instead of Taylor's theorem to get the simpler inequality $|f(x) |\leq x^4|f(c)|$ for some $c$ between $0$ and $x$ and the proof can be continued with this inequality in a similar manner.

0

Notice that $f'(x)=f^{(5)}(x),\,f''(x)=f^{(6)}(x),\,f'''(x)=f^{(7)}(x),\,f^{(4)}(x)=f^{(8)}(x)=f(x)$ (assuming f posseses first, second, third derivatives for all x, that is true because we have $f(x)=f^{(4)}(x)$). So that if $i\in \{5, 6, 7,...\}\,$$$f^{(i)}=\begin{cases} f(x), & \text{if $i = 4,8,12,...$} \\ f'(x), & \text{if $i=5,9,13,...$} \\ f''(x), & \text{if $i=6,10,14,...$} \\ f'''(x), & \text{if $i=7,11,15,...$} \end{cases}$$ Then we claim that f is $C^\infty$, and since it's defined for x = 0, we have the McLaurin's power series of f given by $$f(x)=\sum_{n=0}^{inf}\frac{f^{(n)}(0)x^n}{n!}=\sum_{n=0}^{inf}\frac{(0)x^n}{n!}=0$$

  • But there are flat functions that they have 0 McLaurin's power series but still the function is non-zero. cf https://en.wikipedia.org/wiki/Flat_function – ElfHog Feb 02 '18 at 02:59
  • But it doesn't converge for all $x \in R$ – tobiasbriones Feb 02 '18 at 03:04
  • Your approach assumes that the Taylor series converges, but this is not guaranteed. You must somehow show that the series converges otherwise the proof is incomplete. – Paramanand Singh Feb 02 '18 at 04:32
0

So $f^{(4)}(x)-f(x)=0$, solving $y^{(4)}-y=0$, or the characteristic equation with $m^{4}-1=0$, then $(m+1)(m-1)\left(m+i\right)\left(m-i\right)=0$, so $f(x)=C_{1}e^{-x}+C_{2}e^{x}+C_{3}\cos x+C_{4}\sin x$.

user284331
  • 55,591