On integrable mappings of standard type (Q581898): Difference between revisions

For the difference equation \(x_{n+1}-2x_ n+x_{n-1}=\epsilon F(x,\epsilon)\) where \(F(x,\epsilon)=\sum^{\infty}_{k=0}\epsilon^ kf_ k(x)\) is holomorphic with respect to x in a band of finite width (independent of \(\epsilon)\) about the real axis, and \(\epsilon\) is a parameter close to 0, it is decided when there is a symmetric integral, i.e. a function \(\phi (x,y,\epsilon)=\phi_ 0(x,y)+\epsilon \phi_ 1(x,y)\) holomorphic in a domain \(| x-y| <\delta\) which satisfies \(\phi (x,y,\epsilon)=\phi (y,x,\epsilon)\), \(\phi (x_{m+1},x_ n,\epsilon)=\phi (x_ n,x_{n-1},\epsilon)\) and is not constant. This implies the existence of integrable difference schemes for certain differential equations ẍ\(=f(x)\), e.g. for \(f(x)=A+Bx+Cx^ 2+Dx^ 3\).