Fixed point stability for an iteration (source code)

= Fixed point stability for an iteration

For a scalar iteration $x_{n+1}=g(x_n)$, a <fixed point> satisfies $g(x_*)=x_*$. A small deviation obeys $\delta x_{n+1}\simeq g'(x_*)\delta x_n$, so $|g'(x_*)|<1$ gives local asymptotic stability and $|g'(x_*)|>1$ gives local instability when the map is differentiable nearby.

The multiplier's sign also describes the approach. A positive multiplier below one preserves the sign of sufficiently small deviations, while a negative multiplier of magnitude below one makes successive deviations alternate around the <fixed point>. For example, $x_{n+1}=-x_n/2$ converges while changing sign at each step. At multiplier magnitude one, higher-order terms can decide the outcome, so the strict test is inconclusive. This differs from the sign-of-real-part condition for a continuous-time <autonomous differential equation>; an iteration advances by finite steps.