Finding incomplete solutions of second order ODE system

390 Views Asked by At

For a smooth, 1-periodic, positive function $c:\mathbb{R}\rightarrow (0,\infty)$ consider the following second order ODE system for the curve $\gamma(t)= (x(t),y(t),z(t)):\mathbb{R}\rightarrow \mathbb{R}^3$ (note that I will denote the derivative of $c$ by a prime and the derivatives of the curve components $x,y,z$ by dots).

\begin{align} (i) \hspace{15mm} \ddot{x}(t)+ \frac{c'(z(t))}{c(z(t))}~ \dot{x}(t)\dot{z}(t)=0 \\ (ii) \hspace{14mm} \ddot{y}(t) +\frac{c'(z(t))}{c(z(t)))}~\dot{y}(t)\dot{z}(t)=0 \\ (iii) \hspace{16mm} \ddot{z}(t)-c'(z(t)) ~ \dot{x}(t)\dot{y}(t)=0 \end{align}

My question: I'm interested in finding a function $c$ as described above such that the system has "incomplete" solutions, i.e. solutions $\gamma:\mathbb{R}\rightarrow \mathbb{R}^3$ which are not defined for all times $t\in\mathbb{R}$. Or alternatively I want to show that for all $c$ as above there are no incomplete solutions.

Background: the background to this problem is a geometric one: I have a family of metrics $\{g_c\}$ given on $\mathbb{R}^3$ (where $c$ is as described above) and I'm asking myself if all those metrics are geodasically complete. Solving the geodesic equation for those metrics gives the above system of ODEs.

1

There are 1 best solutions below

4
On

The equation system has the following first integrals: $$c(z)^2\left(\dot x^2+\dot y^2\right)=I_1=const$$ $$c(z)\left(y\dot x-x\dot y\right)=I_2=const$$ If $c_{MIN}<c(z)<c_{MAX}$, then $\dot x$ and $\dot y$ are bounded: $$|\dot x|<\frac{\sqrt{I_1}}{c_{MIN}}$$ $$|\dot y|<\frac{\sqrt{I_1}}{c_{MIN}}$$ Assuming that $c'(z)$ is a bounded smooth function, $\ddot z(t)$ is also bounded. This means that there are no singular solutions (such that $x$, $y$ or $z$ reach infinity for finite $t$). Also, existence theorem guarantees existence of solution to the initial value problem in the neighborhood of every $t=t_0$, so there are no incomplete solutions.