About constants in general solution of an ODE

75 Views Asked by At

Given the general form of an ODE of order $n$, namely: \begin{equation}\label{1} F\left(t,y(t),y'(t),\ldots,y^{(n-1)}(t),y^{(n)}(t)\right)=0 \end{equation} with: $u:t\in\mathbb{R}\longmapsto y(t)\in\mathbb{R}$ and $ F:\Omega\subseteq \underset{t}{\mathbb{R}}\times \underset{y(t)}{\mathbb{R}}\times \ldots\times \underset{y^{(n-1)}(t)}{\mathbb{R}}\times \underset{y^{(n)}(t)}{\mathbb{R}}\longrightarrow \mathbb{R} $ can we state that the general solution (supposing it exists) is a function of at least $n$ arbitrary constants? Without thinking about theorems of existence and uniqueness, this is a priori argument. First consider the case such that the ODE can be written in normal form, setting $(t,y(t),\ldots,y^{(n-1)}(t)\bigl)=z$: \begin{split} y^{(n)}(t)&=F_0\bigl(t,y(t),\ldots,y^{(n-1)}(t)\bigl)\\ \int y^{(n)}(t) dt&=\int F_0\bigl(t,y(t),\ldots,y^{(n-1)}(t)\bigl) dt\\ y^{(n-1)}(t)&=G_0(z)-c_1=F_1(z,c_1)\\ \int y^{(n-1)}(t) dt&=\int F_1(z,c_1) dt\\ y^{(n-2)}(t)&=G_1(z,c_1)-c_2=F_2(z,c_1,c_2)\\ &\vdots\\ y'(t)&=G_{n-2}(z,c_1,\ldots,c_{n-2})-c_{n-1}=F_{n-1}(z,c_1,\ldots,c_{n-1})\\ \int y'(t)dt&=\int F_{n-1}(z,c_1,\ldots,c_{n-1}) dt\\ y(t)&=G_{n-1}(z,x_1,\ldots,c_{n-1})-c_n=F_n(z,c_1,\ldots,c_n) \end{split} so [it seems that] the solution depends at least on the $n$ integrating constants: \begin{equation} y(t)=F_n(z,c_1,\ldots,c_n) \end{equation} Questions:

  1. Is it a reasonable proof for the normal case?
  2. If not, what can we say about it?
  3. Can we say that the case of an ODE in normal form in presence of Lipschitz regularity, is the case in witch we have exactly $n$ constants?
  4. In other words, can we state the following Theorem: given an ODE of order $n$ if the solution is not unique, than the solution depends on at least $n$ constants. Morover if the ODE can be written in normal form (and there is Lipschitz regularity) the constants are exactly $n$. , the solution Thanks for collaboration! :D

PS: a necessary hypothesis is that the ODE admists more than one solution a priori, e.g. don't consider:

\begin{equation} y'(t)=\sqrt{-|y(t)|} \end{equation}