I want to find $\lim_{n \to \infty} \int_a^b g(x) \sin^n(x)dx$ where $a<b$ and $g(x)$ is continuous.
I've been able to show that if $a=0$ and $b=1$, then the limit is $0$. But I'm not sure if this holds in general.
Is there a general proof that the limit is $0$? If so, what does it look like?
UPDATE: I should have mentioned that I can't use measure theory.
This may be cheating, but it is essentially a much less elegant proof via the other answers. Fix $\epsilon>0$, and denote the points where $\vert \sin(x) \vert=1$ by $\{ y_l \}_{l=1}^k\subset [a,b]$. Then $g(x)\cdot \sin^n(x)\rightarrow 0$ converges uniformly to $0$ on $[a,b]\setminus \bigcup (y_l- \frac {1}{2} \epsilon, y_l+ \frac{1}{2} \epsilon)$, therefore if you denote $E:=[a,b]\setminus \bigcup_{l=1}^k (y_l- \frac {1}{2} \epsilon, y_l+ \frac{1}{2} \epsilon)$, you have:
$\vert\int_{a}^{b}g(x)\sin^n(x)dx\vert \leq \sum_{l=1}^k\int_{y_l-\epsilon}^{y_l+\epsilon}\vert g(x) \sin^n(x)\vert dx+\int_E \vert g(x)\sin^n(x)\vert dx \leq$
$\overset{\vert \sin^n(x)\vert \leq 1}{\leq} \sum_{l=1}^k\int_{y_l-\epsilon}^{y_l+\epsilon}\vert g(x)\vert dx+ \int_E \vert g(x)\sin^n(x)\vert dx \leq$
$\overset{\vert g(x)\vert \leq M}{\leq} M\cdot\sum_{l=1}^k\int_{y_l-\epsilon}^{y_l+\epsilon}1 dx+\leq M\cdot \epsilon k + \int_E \vert g(x)\sin^n(x)\vert dx \rightarrow Mk \epsilon+0$.
where $M$ is an upper bound on $G$ obtained from continuity on $[a,b]$. Since $M$ and $k$ are fixed, and $\epsilon$ is arbitrary, this must imply that:
$\underset{n\rightarrow \infty}{\lim} \vert\int_{a}^{b}g(x)\sin^n(x)dx\vert \leq 0$.
There are issues on what the intervals that should be excluded are since right now, we could be integrating outside $[a,b]$, but this is less important than the underlining argument.
Essentially, you take intervals of length $\epsilon$ around where $\sin(x)=1$, and use uniform convergence on the complement. The integrand on these intervals is bounded, and the length of the intervals can be a small you want.