2D root finding is failing on realistic convex potentials

[keiwanjamaly]

When modelling a 2D piecewise function with a linear decent at the origin and a quadratic increase at $x=x'$ results at a false root found in the origin.

To be precise, the function chosen to model a typical RG flow behaviour was

$u(x) = m\cdot x \quad \text{if} \quad x < 0\quad \text{else} \quad a*x^2 + b$

with b chosen, such that u is continuous.
A false root was found, if $m<0$.