A thin metal barrier would be ideal but of course if you cannot alter the sample structure then perhaps a multiangle approach would at least allow you to vary the penetration depth and thereby understand how much light is getting to your layer of interest. Good Luck!
