Some counterexamples to separation of variables (Q1898792)

An algebraic PDE system admits separation of variables if it has at least one differentially algebraic (DA) solution. Here a function is DA if there is a field of finite transcendence degree over \(\mathbb{C}\) that contains this function and all its partial derivatives. If \(u(x, y)\) is a DA solution of an algebraic PDE system, then \(u(x, y)\) can be computed in the following way: Starting, say in \((0, 0)\), we first compute \(u(x, 0)\) by solving an algebraic ODE along the \(x\) axis, then we solve an algebraic ODE along the vertical line \(x=\text{const}\) to find \(u(x, y)\). The main result of this paper is that the algebraic PDE \(u_y= (1- {x/y})\cdot u+ 1\) has no DA solutions. The fact that the indefinite integral with respect to \(y\) of a DA function \(f(x, y)\) need not be DA is at the heart of this counterexample. Concluding some open problems are presented.