Let $\nu_t = b^t \nu_0$ be a geometric sequence where $\nu_0>0$, $0<b<1$, $t = 0,1,2,\dots$. Let $h_0>0$ and $0<a<1$. Define the sequence $h_{t+1} = a h_t+\nu_t$. Show that $h_n$ is linearly converging.
----My Attempt:
Let $c:=\max\{a+\epsilon,b\}$ for any $\epsilon>0$. We prove by induction that $h_t\leq \frac{c^t}{\epsilon}(h_0+\nu_0)$ for any $\epsilon>0$. It follows by the induction hypothesis that \begin{equation} \begin{aligned} h_{t+1}&= a h_t +\nu_t \leq \frac{a}{\epsilon} c^t (h_0+\nu_0)+\frac{\epsilon}{\epsilon} c^t \nu_0\\ &\leq \frac{a}{\epsilon} c^t (h_0+\nu_0)+\frac{c-a}{\epsilon} c^t (h_0+\nu_0) = c^{t+1}(h_0+\nu_0)/\epsilon, \end{aligned} \end{equation} where we used the fact that $c-a\geq \epsilon$ and $\nu_0 \leq h_0+\nu_0$.
----What I need:
I'd like an approach that does not introduce $\epsilon$. Any help is much appreciated.
If you calculate $h_n$ for $n=0,1,2,3,4$, it’s easy to conjecture that
$$h_n=a^nh_0+\nu_0\sum_{k=0}^{n-1}a^kb^{n-1-k}$$
and prove it by induction. If $a\ne b$ it can then be written more compactly as
$$h_n=a^nh_0+\frac{(a^n-b^n)\nu_0}{a-b}=(h_0+c)a^n-cb^n$$
where $c=\frac{\nu_0}{a-b}$. If $a=b$, $h_n=h_0a^n+n\nu_0b^n$.