Surface modification is highly required in such cases. For glass, clean it thoroughly in piranha solution  @ 20 min. and clean in DI water, thereafter Heat it at 150 deg. Cent. in oven for 30 min and treat in oxygen plasma for 30 sec and immediately coat by dip method or spin coating. For Silicon an oxide layer is required for better adhesion.

thanks......

I agree with @Pradip Kumar Dey. Basically you need a completely clean surface. And the method mentioned by him is good. i would just like to add that the cleaning step with DI water should be done properly and should involve 10 min sonication.

Another aspect you should pay attention to is that while cleaning the substrates should not get scratched and at no point in the cleaning process you should touch the top of the surface (even with forceps).

If you use the above method for cleaning it will even work for Si substrates, but of course it would be better if you can oxidize the substrates (depends on the applications you are looking for in the final stage). 

Surface modification also involves surface roughness.  The higher the surface roughness the better chance of surface adherence particularly in dip/spin coating. Try it and you shall get better adherence. Good luck.

Pre-treating the silicon surface or glass subtrate (with KOH for example) can help to attached something on the surface. I agree also with the others, surface modification in the key to for attaching something on a surface but it depends also on the functional groups of your coating material (OH, H, COOH etc...). 

the cleaning procedures described above are okay.......and I think everyone cleans the substrates before deposition by one method or another, but main still some basic growth mechanisms are responsible for not being deposited on these substrates. Crucial ones are large lattice mismatch between substrate and deposited material, the type of bonds in the deposited material on which interface dynamics depends very much while depositing, last one is the temperature of deposition which you may have tried at many but it is very important in the growth dynamics of the atoms and helps atoms to sit in the preferred low energy sites(ordered occupation in the crystallographic sites ). 

Try pre annealing of substrates as well to get rid of carbon and other oxide related contamination.

You can also use an anchoring layer first. When I was doing coatings of cellulose nanocrystals on glass by spin coating I used first an anchoring layer of polyallilamine hydrochloride and it worked. This polymer may not be (or may be) suitable for you, but you can take a look.

PS: I was always cleaning the glass with the piranha solution first.

Surface roughness is too small and difference is high but if seeding layer is applied then it possible 

Possibly Oxygen plasma along with argon plasma could help you more than only oxygen plasma

For glass: The surface should be cleaned by sonicating using a solution of acetone (10 min), solution mixtures of NH4OH : H2O2 : H2O (1 : 1 : 5, 30 min at 80 C), and  HCl : H2O2 : H2O (1 : 1 : 6, 30 min at 80 C), followed by rinsing and drying.

Contamination of the substrate is indeed a factor. Nevertheless, any thick sol gel film could not adhere properly. There is a thresshold of film thickness of 0.5μm in order to achieve good adhesion with the substrate. So, you should try either a lower viscosity of the solution or a more rapid dip and/or spin. Anyway, on the Si substrate its still possible you wont get what you need. I dont have an explanation to this, just from experience.

Either the surface is contaminated or contain any foreign materials. To overcome this issue clean either glass of silicon in Piranha solution (H2O2 and H2SO4) @ 20 min and dry in oven for 30 min @ 150 Deg. Cent. For glass substrate treat in oxygen plasma for 30 sec before deposition to promote adhesion and for silicon grow a thin oxide layer and then deposit any material.