# A theorem about oscillation in Arnold's mathematical methods of classical mechanics

There is a theorem in page 100 of Arnold's Mathematical Methods of Classical Mechanics, which says that:

If $\cfrac{dx}{dt} = f(x) = Ax + R_2(x)$, where $A = \cfrac{\partial f}{\partial x}|_{x = 0}$, $R_2(x) = O(x^2)$, and $\cfrac{dy}{dt} = Ay$, $y(0) = x(0)$, then for any $\vphantom{\cfrac12} T>0$ and for any $\xi > 0$ there exists $\delta > 0$ such that if $|x(0)| < \delta$, then $\vphantom{\cfrac12}|x(t) - y(t)| < \xi \delta$ for all $t$ in the interval $0 < t < T$.

How can I prove this theorem rigorously?

Cross-posted at physics.se here.

• This follows from continuity of solution of initial value problems with respect to parameters. For example, for the equation with parameters $x'=Ax+\lambda R_2(x)$, with parameter $\lambda\in\mathbb R$. When $\lambda=0$ you get the linear one, with $\lambda=1$ the original non-linear one (and somework, depending on what statement of this theorem you have :-) ) – Mariano Suárez-Álvarez Dec 8 '13 at 2:53
• Thanks! Can you explain it further? – Eden Harder Dec 9 '13 at 14:31

Lets assume that $R$ is smooth. Therefore if $|f|,|g| < M \le 1$, then $$|R(f) - R(g)| \le C_1 M |f-g| .$$

Also define $$C_2 = \sup_{0 \le t \le T} \|e^{tA}\| .$$

Define $$B(\epsilon) = \{f(t) \in C([0,T]) : \|f\|_\infty \le \epsilon\}$$ $$G_{x_0}:B(\epsilon) \to C([0,T])$$ $$G_{x_0}f(t) = e^{tA} x_0 + \int_0^t e^{(t-s)A} R(f(s)) \, ds$$ Then $$\|G_{x_0}(f-g)\|_\infty \le T C_2 C_1\epsilon \|f-g\|_\infty$$ Hence there exists $\delta_1>0$ such that if $\epsilon<\delta_1$, then $G_{x_0}$ is a contraction mapping. Also $$\|G_{x_0}(f)\|_\infty \le C_2 |x_0| + T C_2 C_1\epsilon^2$$ Hence there exists a $\delta_2$ such that if $\epsilon, |x_0| < \delta_2$, then the range of $G_{x_0}$ is contained in $B(\epsilon)$.

Therefore if $\epsilon, |x_0| < \min\{\delta_1,\delta_2\}$, then $G$ has a fixed point. That will be the function $x(t)$.

Let $z = x-y$. Then $$\frac{dz}{dt} = Az + R(x) .$$ $$z(t) = \int_0^t e^{(t-s)A} R(x(s)) \, ds$$ Therefore $$\|z\|_\infty \le C_2 C_1 \epsilon^2 .$$ Choose $\delta = \min\{\delta_1,\delta_2,\xi/(C_1 C_2)\}$.

• Thanks very much! How can you make sure that $x(t) \in A(\delta)$? If can not, then $x(t)$ is not the fixed point of $T$. – Eden Harder Dec 8 '13 at 3:56
• I didn't explain that argument very well. But I edited the document to say that a similar argument will tell you this. – Stephen Montgomery-Smith Dec 8 '13 at 4:13
• Thanks! Can you explain $Tsup_{0≤t≤T}∥e^{tA}∥(∥R(f)−R(g)∥_∞)≤Kδ∥f−g∥_∞$ further more? I can not figure out where δ comes from. – Eden Harder Dec 8 '13 at 5:53
• If it were the case that $R(x) = x^2$, then we would have $R(f) - R(g) = (f+g)(f-g)$. I was saying that if $R$ is smooth, then we can do something similar, were the $f+g$ is replaced by something from a mean value theorem. – Stephen Montgomery-Smith Dec 8 '13 at 15:48
• Ok~ your proof assumes that $||x||_\infty \leq \delta$ at the last step while the theorem only under the condition that $|x(0)| \leq \delta$. – Eden Harder Dec 9 '13 at 0:10