Instability of an equilibrium in a potential field (Q1823036)

We study the instability of an equilibrium position (located at the origin in \({\mathbb{R}}^ n)\) of the mechanical system (1) \((d/dt)L_{x'}(x,x')=L_ x(x,x')\). The Lagrangian L: \({\mathbb{R}}^ n\times {\mathbb{R}}^ n\to {\mathbb{R}}\) is given by (2) \(L(x,x')=T(x,x')-U(x)\) where, for each \(x\in {\mathbb{R}}^ n\), the kinetic energy, T(x,x'), is a positive-definite quadratic form in x'. Since adding a constant to the potential energy, U, does not change (1), we can assume \(U(0)=0.\) If L is continuously differentiable and U assumes a strict local minimum at \(x=0\), then the Lagrange-Dirichlet theorem states that the origin for (1) is stable (that is, for each \(\epsilon >0\), there exists \(\delta >0\) such that if x(t) is a solution of (1) with \(\| x(0)\| +\| x'(0)\| <\delta\) then \(\| x(t)\| +\| x'(t)\| <\epsilon\) for all \(t>0).\) Several authors have incorrectly claimed that the converse of the Lagrange-Dirichlet theorem is true. However it was shown elsewhere that if \(n=2\), \(T(x,x')=x'\cdot x'\), and U: \({\mathbb{R}}^ 2\to {\mathbb{R}}\) is the unique infinitely differentiable function satisfying \[ U(x_ 1,x_ 2)=e^{-x_ 1^{-2}}\cos (1/x_ 1)-e^{-x_ 2^{-2}}(x^ 2_ 2+\cos (1/x_ 2)),\quad x_ 1x_ 2\neq 0, \] which is negative and decreasing on the line \(x_ 2=x_ 1\), \(x_ 1>0\), then the origin for (1) is stable. In this paper we assume that T: \({\mathbb{R}}^ n\times {\mathbb{R}}^ n\to {\mathbb{R}}\) and U: \({\mathbb{R}}^ n-\{0\}\to {\mathbb{R}}\) are twice continuously differentiable and, for \(i=0\), 1, 2, \(U^{(i)}(x)=U_ p^{(i)}(x)+O(\| x\|^{p+\epsilon -i})\) as \(x\to 0\) where \(p>0\) and \(0<\epsilon \leq 1\) are real numbers, and \(U_ p:\) \({\mathbb{R}}^ n- \{0\}\to {\mathbb{R}}\) is positively homogeneous of degree p and three times continuously differentiable. Thus \(\lim_{x\to 0}U(x)=0\). However the origin is an equilibrium position for (1) (that is \(\lim_{x\to 0}U'(x)=0)\) if and only if \(p>1\). Since (1) is autonomous, the t- translation of any solution of (1) is also a solution of (1); that is if \(t_ 0\in {\mathbb{R}}\) and x(t) is a solution of (1), then \(\hat x(t)=x(t+t_ 0)\) is also a solution of (1). If, for some \(x\in {\mathbb{R}}^ n-\{0\}\), \(U_ p(x)<0\) then we will prove that there is a solution, x(t), of (1) such that \(\| x(t)\| +\| x'(t)\| \to 0\) as \(t\to 0^+\), when \(0<p<2\); and as \(t\to -\infty\), when \(p\geq 2\). In particular, since the t-translation of any solution of (1) is also a solution of (1), the origin for (1) is not stable.