I will refer to Simpson's "Higgs bundles and local systems".
Proposition 1.4:
When $X$ is a smooth projective variety, one can build up the moduli space $\mathcal{M}(X,r)$ of rank $r$ Higgs bundles (of a certain type) over $X$ as a quasiprojective variety. Furthermore, there is a proper map from $\mathcal{M}(X,r)$ to a [finite dimensional] vector space.
The proper map this proposition talks about is the Hitchin fibration
$$ \chi \, \colon \, \mathcal{M}(X,r) \to \bigoplus_{k=1}^r H^0(X,\mathrm{Sym}^k\,\Omega^1(X)) $$
sending each isomorphism class $[(E,\phi)]$ to the coefficients of the characteristic polynomial of the Higgs field $\phi$. These are holomorphic sections of the symmetric powers of $\,T^*_X \otimes \mathbb{C}$.
There is a continuous $\mathbb{C}^*$-action on $\mathcal{M}(X,r)$ defined by rescaling the Higgs field by $z \in \mathbb{C}^*$:
$$ z \cdot [(E,\phi)] = [(E,z \phi)] $$
In the proof of Theorem 3:
Since $$ \lim_{z \to 0} \, \chi (z \cdot[(E,\phi)]) = 0 $$ and $\chi$ is a proper map, we have that the limit $$ \lim_{z \, \in \, \mathbb{C}^* , \,z \to 0} z \cdot[(E,\phi)] $$ exists and is a fixed point of the $\mathbb{C}^*$-action.
Clearly, if the limit exists, being unique, it provides a fixed point. But I do not follow Simpson's argument leading to the existence of $\lim_{z \to 0} z \cdot[(E,\phi)]$. Maybe, I'm leaving out some considerations.
Question: Why the limit $\lim_{z \to 0} z \cdot[(E,\phi)]$ does exist?
