What do I get when measuring on top of metal/insulator/metal by kelvin probe? I am using a macroscopic Kelvin Probe for work function measurement. There's one confusing condition. If I measure a trilayer thin film stack with a structure of ITO(bottom)/PMMA(100-200nm)/Al(top), what can I get?

Am I obtaining work function of top layer Al (after adding work function of the probe)? Or is it meaningless after all?
I thought since electrons of Al may not be able to transfer through the poorly conductive PMMA layer to align its fermi level to those of ITO and the probe, the result should be meaningless. But what I really got, after an experiment of it, I found the calculated work functions are as following:
ITO/PMMA/Al: 3.19eV
bare Al: 3.35-3.37eV
(note: Al oxidized at top yields a work function lower than intrinsic Al(4.2eV))
The about 0.1eV difference is big and confusing. Anyone knows why?
 A: Even if the floating Al film is (nearly) disconnected in DC, it is still capacitively coupled to the substrate. Since Kelvin probe techniques use capacitance modulation, the behaviour of the experiment will be more or less identical to the case of a connected film, since tip-film capacitance is so tiny compared to the film-substrate capacitance.
It's possible that the chemical potential of the Al film doesn't equilibrate exactly to the same value as the substrate, if your sample is freshly made or if it has recently undergone some process. Electrochemical forces present in the device's relaxation can pump charge, and if the leakage conductance is very low then you might get a voltage shift.
Indeed though, in this case what you measure is probably just the work function of the Al film, but I wouldn't count on it being that way every time.
By the way, is it really possible to be so precise about the work function of oxidized Al? I would have guessed that it would depend on many details of deposition, surface roughness, age, etc...
