4

It is possible to find non trivial solutions $f(t) \in C_c^\infty$ to $\dot{f}(t)=2f(2t+1)-2f(2t-1),\,f(0)=1$ for the whole $\mathbb{R}$ domain?

I am trying to find examples of solutions of finite duration to differential equations (meaning here, they achieve zero in finite time and stays there forever after), and some functions that shows a similar behavior are smooth bump functions $\in C_c^\infty$ (I hope that from an initial condition and their differential equation I could made a finite duration solution), and in MathStackExchange I found this question with a lot of examples of smooth bump functions, but so far, the only differential equation I found describing them without definition issues is the one given in this answer, which is the equation I am asking for.

There, the equation is taken from this paper by Juan Arias de Reyna (2017), where the author found the Fourier Transform of $f(t)$ (highly nonlinear), but the explicit/exact formula for $f(t)$ is not given.

I tried to research about this differential equation since I never saw before a differential equation with arguments different from a simple variable $t$, so I am not sure if this equation belongs to a Functional Equation or instead to a Delayed Differential Equation, or other as which are studied in the book Generalized Solutions Of Functional Differential Equations by Joseph Wiener (I tried to read it "diagonally" looking for something similar unsuccessfully - it works with things with explicit discontinuities).

I have no more related experience than reading the related Wikipedia pages, but so far, it looks like that the differential equation stand the zero trivial solution $f(t) = 0$, but also it will solved by every constant $f(t)= c\equiv f(0)$ (since $\dot{c}=0=c-c$), so it looks like Uniqueness of solutions is not hold (actually Wolfram-Alpha gives this solution on its numerical plots).

I don't even sure if the initial condition given is enough information to solve it, so if required, as is done I think in the paper, include the following additional requirement: $$f(t) = 0,\,|t|\geq 1$$

I have also played in Wolfram-Alpha trying to fit things like the other examples given in the mentioned question, or trying things like $y(x) = e^{-\log(1-x^2)^2}(1-x^4+|1-x^4|)$ which looks like the draw of the paper, but it didn't work so far (is really hard to make the two displaced and contracted versions to match their derivative).

So I hope you can answer this:

  1. What kind of differential equation is this? (if it match any - this to look for references)
  2. It is possible to find the exact solution $f(t) \in C_c^\infty$? (maybe someone already prove is not possible)
  3. Obviously, which is the exact solution $f(t) \in C_c^\infty$?

Added later

My best attempt after trying "too many things" was this: $$y(x) = e^{-\log(\sqrt{1-x^4})^2}\left(\frac{1-x^4+|1-x^4|}{2}\right)^{14}$$ From what I am seeing, the derivative need to be more flat near $\pm 1/2$. The exponential function is so flat top that actually is the polynomial who is giving the form of the function, only been the exponential who made the function goes smoothly to zero at the edges $[-1,\,1]$. Maybe someone more experienced could made a polynomial fit.

Improved attempt: $$y(x) = e^{-\tanh^{-1}(x^4)^2}\left(\frac{1-|x|^\frac{5}{2}+|1-|x|^\frac{5}{2}|}{2}\right)^{4}$$ and as is seen in Desmos.


Added later 2

It looks the solution to the main differential equation behaves like a translated and extended version (with its own reflection to form a symmetric/odd function) of the also known Fabius function, on Wikipedia is shown some of their properties.

Joako
  • 1,957
  • 1
  • This is a functional differential equation. 2. Asking for a solution to be infinitely differentiable may be too strong here. This will need to be clarified. 3. In general, those equations do not admit closed-form solutions.
  • – KBS Apr 08 '22 at 17:09
  • 1
    @KBS Thanks for commenting. About point 2, I am following what it is said on the paper (or at least, what I am understanding from it, since I am not someone with expertise on the topic). – Joako Apr 08 '22 at 21:03
  • 1
    I have checked the paper and the author proved the existence and the uniqueness of such a solution. The author also provides 4 different expressions for the function. It seems unlikely that there is an explicit solution $f(t)$ in terms of elementary functions as you seem to be looking for. In fact, most equations have no solutions in terms of elementary functions. We usually do not use explicit expressions, but only need to know the properties of that function. – KBS Apr 08 '22 at 22:11
  • 1
    Thanks you very much for taking the time to see the paper. When you say that "have no solutions in terms of elementary functions", it means: (a) is been proved that the answer exists and also is proven that it cannot been represented through known functions, or (b) is been proved that the answer exists but nobody knows so far how to express it through known functions (so is unlikely I will find it by myself by trial and error)?? – Joako Apr 08 '22 at 22:24
  • 1
    I have not said "have no solutions in terms of elementary functions". I have said "It seems unlikely that there is an explicit solution in terms of elementary functions". We know that it exists, but we do not know whether it can be expressed as elementary functions but this is unlikely that it can be. I mean, I do not know why you are insisting so much in trying to solve differential equations in all your posts. Almost all of them are not solvable in terms of elementary functions. You do not need to find explicit expressions for the solutions to understand how they behave. – KBS Apr 08 '22 at 22:30
  • 1
    I believe that through examples is way more easy to understand something and using it as a tool (like a "toy model"), it works at least for me, this is why I am looking for them... been true there are alternative ways to figure out the behavior, if it keeps in the mathematician/physicist abstract world it doesn't become a tool for least trained people (like me)... so far, through searching for examples, now I learn that there exists solutions of finite duration (papers didn't have them explicitly), and also that no wide-known model could stand them since are traditionally Lipschitz ODEs. – Joako Apr 09 '22 at 00:03
  • Then, if you can learn in this way it is fine. Do you have any goal? What is the purpose of all your questions? – KBS Apr 09 '22 at 22:42
  • I am trying to figure out if using traditional linear (or non lipshitz) models could be introducing unwanted phenomena since classic phenomena have ending times and they cannot stand finite duration solutions... is a long shot, but many issues in nowadays physics are related with the arrow of time, and I am trying to understand if are because of using traditional approximations... in the short term, I am writing a question abou the use of this final duration solutions to model the bounded spiral the Euler's disk toy made, since I believe this ending time cause the blow up in the wobbling rate. – Joako Apr 09 '22 at 22:55
  • As example, I am thinking about the issues related with infinite response filters compared with finite ones. Also, I have started really lost about them since at first I didn't find any examples on the web (now there is one on Wikipedia, but because I add it). And now, with your help and the help of other users, I believe there is now a whole family of closed form examples as I proposed here, which I believe, at least for small angle aproximations will be usefull (since a quadratic form is included). So thanks for your help and patience. – Joako Apr 09 '22 at 23:05
  • 1
    Ok in that case, you should view your differential equations as dynamical systems which are integrated forward in time. Not all solutions of a differential equations will make sense for a dynamical system. For instance, if a dynamical system reaches an equilibrium point, it will never leave from that place unless it is perturbed. – KBS Apr 09 '22 at 23:08
  • @KBS thanks for your comment. Actually that is an important property I want to use on describing the model. In this question I present step by step how it can be used to properly model the time variable on a scalar autonomous ODEs (since it should be Non-Lipshitz if I want make the system to have a finite ending time). In few words, maybe current physics have miss this, or at least, accuracy could be improved (don't know if the same apply to PDEs) – Joako Apr 14 '22 at 17:54
  • @KBS Hi. I have made this another question where I explained what I am aimming to understand with all these questions. I hope you can see it too. Beforehand, thanks you very much. – Joako May 23 '22 at 23:05