New definitions of exponential, hyperbolic and trigonometric functions on time scales (Q662045)

This interesting paper compares the definitions of exponential, hyperbolic and trigonometric functions on time scales introduced by Stefan Hilger on one side, and by Martin Bohner and Allan Peterson on the other side. The author proposes a new definition of the exponential function inspired by the Cayley transform. Given an rd-continuous function \(\alpha:\mathbb T\to\mathbb R\) satisfying \(\mu(t)\alpha(t)\neq\pm 2\), where \(\mu\) denotes the graininess of the time scale \(\mathbb T\), the new exponential function is defined as \[ E_\alpha(t,t_0)=\exp\left(\int_{t_0}^t\xi_{\mu(s)}(\alpha(s))\Delta s\right), \] where \[ \xi_h(z)=\frac{1}{h}\log\frac{1+\frac{1}{2}zh}{1-\frac{1}{2}zh} \] for \(h>0\) and \(\xi_0(z)=z\). Equivalently, the function \(t\mapsto E_\alpha(t,t_0)\) can be characterized as the unique solution of the initial-value problem \(x^\Delta(t)=\alpha(t)\langle x(t)\rangle\), \(x(t_0)=1\), where \[ \langle x(t)\rangle=\frac{x(t)+x(\sigma(t))}{2}. \] The relation between the new exponential function \(E\) and the usual exponential function \(e\) on time scales introduced by Stefan Hilger is given by the identity \(E_\alpha(t,t_0)=e_\beta(t,t_0)\), where \[ \beta(t)=\frac{\alpha(t)}{1-\frac{1}{2}\mu(t)\alpha(t)}. \] After establishing the properties of the function \(E\), the author defines \(E_\alpha(t)=E_\alpha(t,0)\) and introduces hyperbolic and trigonometric functions by the formulas \[ \mathrm{Cosh}_\alpha(t)=\frac{E_\alpha(t)+E_{-\alpha}(t)}{2},\;\;\;\mathrm{Sinh}_\alpha(t)=\frac{E_\alpha(t)-E_{-\alpha}(t)}{2}, \] \[ \mathrm{Cos}_\alpha(t)=\frac{E_{i\alpha}(t)+E_{-i\alpha}(t)}{2},\;\;\;\mathrm{Sin}_\alpha(t)=\frac{E_{i\alpha}(t)-E_{-i\alpha}(t)}{2i}. \] However, since 0 does not need to belong to the time scale, it would be better to fix a point \(t_0\in\mathbb T\) and let \(E(t)=E(t,t_0)\). An alternative is to define the hyperbolic and trigonometric functions as functions of two variables, \(t\) and \(t_0\) (this is the usual approach in the time scale literature). The new hyperbolic functions satisfy \[ \mathrm{Cosh}_\alpha^2(t)-\mathrm{Sinh}_\alpha^2(t)=1, \] \[ \mathrm{Cosh}_\alpha^\Delta(t)=\alpha(t)\langle\mathrm{Sinh}_\alpha(t)\rangle,\;\;\;\mathrm{Sinh}_\alpha^\Delta(t)=\alpha(t)\langle\mathrm{Cosh}_\alpha(t)\rangle. \] Similarly, the new trigonometric functions satisfy \[ \mathrm{Cos}_\alpha^2(t)+\mathrm{Sin}_\alpha^2(t)=1, \] \[ \mathrm{Cos}_\alpha^\Delta(t)=-\alpha(t)\langle\mathrm{Sin}_\alpha(t)\rangle,\;\;\;\mathrm{Sin}_\alpha^\Delta(t)=\alpha(t)\langle\mathrm{Cos}_\alpha(t)\rangle. \] Finally, the author turns his attention to the case when \(\alpha\) is a constant function and discusses the possibility of introducing elementary functions on time scales by restricting the domain of the usual elementary functions defined on the whole real line. As he demonstrates, the disadvantage of this approach is that it leads to dynamic equations which are rather awkward.