A nonlinear Mercerian theorem (Q1819002): Difference between revisions

Let \(x: [0,\infty)\to\mathbb R\) be a continuous function and let \(\phi:[0,\infty)\to [0,\infty)\) be continuous and strictly increasing to infinity. The Cesàro averages of \(x\) with respect to \(\phi\) are defined by \(y(t) = (1/\phi(t))\int_0^tx(u)d\phi(u), t>0.\) The classical ``Mercerian'' theorem states that if, for some \(0 <\lambda < 1,\) the linear combination \(\lambda x(t) + (1 - \lambda)y(t)\to c,\) then \(x(t) \to c\) without any additional condition. Denote by \(D\) the diagonal line \(\{{\mathbf x}=(x,x):x\in\mathbb R\}.\) A continuously differentiable function \(F:\mathbb R^2\to\mathbb R\) is called good at \({\mathbf x}_0,\) if \(\partial F({\mathbf x}_0)/\partial x\neq 0.\) Using a dynamical systems method the author proves the nonlinear analogue of the Mercerian theorem. Let \(F:\mathbb R^2\to\mathbb R\) be continuous, \(c\) a real number such that \(F^{-1}(c)\cap D\) is finite with \(F(x,x)\) bounded away from \(c\) as \(x\to\pm\infty,\) and assume that \(F(x,x)\) is good at all \({\mathbf x}_0\in F^{-1}(c)\cap D.\) Let \(x:[0,\infty)\to\mathbb R\) be continuous and let \(y\) be the Cesàro averages of \(x.\) If \(F(x(t), y(t))\to c\) as \(t\to\infty\) then either (a) for all sufficiently large \(t\) we have either \(x(t)\geq y(t)\to\infty\) or \(x(t)\leq y(t)\to-\infty,\) or (b) \(x(t)\to x_0\) as \(t\to\infty\) for some \(x_0\in\mathbb R\) satisfying \(F(x_0,x_0)=c.\) For example, this theorem shows that if \(x^3(t) - y^2(t)\to 4\) then \(x(t) \to 2.\) A specific instance of the theorem has applications to fractal dimensions and Minkowski measurability.