A method for solution of nonlinear ordinary differential equations (Q4940544)

The author proposes a scheme for solving the problem \[ u''+2u'/x+\beta ^2u^\nu =0,\;u(0)=0, \quad u'(0)=0. \] After rather cumbersome manipulations he obtains the integral equation \[ u(x)=1+\lambda ^{-1}\int_{0}^{x}\Biggl[(c\lambda +\xi)\psi [a](\xi)-\varepsilon ^{-\mu }(c\xi)^{-1}\int_{0}^{\xi }\zeta ^2a(\zeta) d\zeta \Biggr]d\xi \] with \(a=-\varepsilon ^{\mu }\beta ^2u^{\nu-1 },\;\mu =1/(2-\nu),\;c,\lambda \) are parameters, and \(\psi[a](\xi)\) is a linear Volterra operator with complicated kernel. The author finds a ``natural'' Volterra equation \(u(x)=1-\int_{0}^{x}s^{-2}\int_{0}^{s}\beta ^2\xi ^2u^\nu (\xi) d\xi ds\) to be inefficient for applications since its kernel contains the singularity \(s^{-2}\). But the same singularity appears in \(\psi\) for the presence of the term \((\varepsilon ^\mu cx^2)^{-1}\int_{0}^{x}\xi ^2a(\xi) d\xi \). So, the advantages of the author's approach are not obvious.