Fusion bonding of silicon or silicon dioxide requires that both surfaces be
highly polished and smooth.  If the oxide you are putting on the wafers is
deposited, the roughness will probably be too high.  Try thermal oxidation
of your wafers.  If you are using junk wafers then you may have to do a
denuding oxide and strip and reoxidize the wafers.

The second limiting factor is surface chemistry.  Hydrophobic (such as
cleaned in HF) are harder to bond than hydrophillic.  That's why you had
some success with bonding with water.  It is generally accepted that for
silicon direct bonding, hydrogen bonds dominate the strength or degree of
bonding.  Try cleaning your wafers in Piranha clean or some other oxidizing
bath immediately prior to bonding.

The final factor (which is really part of the first) is wafer flatness or
flexibility.  If you are using normal thickness (~ 400-500 micron) wafers,
the bow of the wafers must be very small as well as the roughness.  If the
wafers are warped, then you won't have enough surface area in close enough
proximity to bond.  Thinner wafers deform easier and can overcome roughness
that thick wafers will not bond to normally.

For more information on wafer bonding, look for papers by Q-Y. Tong and U.
Gosele in the Electrochemical Society journals and conference proceedings.
I've also got a paper in the fourth wafer bonding conference on the use of
very thin (~50 micron) Si wafers for bonding.

Good luck in your future attempts,

Mike


p.s..  having water at the interface is really not a good idea when doing
the anneal.  The water will tend to push the wafers apart during heating.
But you could try leaving them in for a day or so.  The water may then break
up chemically and the constituents either help with the bonding or diffuse
into the silicon.  The bubbles may also reduce due to flow of the oxide.