[mems-talk] Metallization Peel on Anodic Bonded Glass Wafer

[Brubaker Chad]

John,

Out of curiosity, what is the exact makeup of the stack.  Based on your description, it sounds as if the effective arrangement of the stack is as follows:

Silicon
<<<Bond Interface>>>
Au
Pt
Ti
Glass

If this is the case, then the issue may be more of an issue with physics.

In general, anodic bonding can only occur between amorphous materials bearing a positive ionic species (for example, borosilicate glass containing Na+ ions) , and an oxidizable material (silicon in most cases, but also GaAs, kovar, etc...)

During an anodic bond, the heating of the substrates increases the mobility of the Na+ within the amorphous structure of the glass.  Application of a negative potential to the top of the glass wafer causes the Na+ to be drawn to the negative potential.  This leaves unbalanced negatively charged oxygen at the bond interface, which then chemically bonds to the silicon

In this case, you are expecting an adhesion between gold and silicon - this reaction cannot happen by an anodic bonding process, because first, the gold has no positive ions available for migration, and second, because silicon will not chemically bond to gold.

I'm guessing that the success of the 4" bonds was due to one of two other mechanisms - 

1) - the anodic bonding worked, but it was between the non-metallized glass and the silicon.  The electric field that developed was enough to cause the thin (~500µm) substrates to flex slightly until glass and silicon were in contact. This may not be working on the larger wafer size because the standard thickness for 6" is much greater (650 µm).  If the pitch of your patterns is similar, then the amount of flexing wil be much less.

2) - The bonding mechanism is not what you thought - instead, you accomplished a gold-silicon eutectic bond.  The eutectic point of gold-silicon is 363ºC, so if you are processing your bond at 400ºC, you're definitely above this point.  If you are applying enough bond force (or if your metallization method produces a smooth enough surface), the Si and Au inter-diffuse and liquefy.  This may not work at 6" because either the metal surface is rougher, or you are using the same bond force, which, considering that the area is 225% that of 4", is insufficient to press enough of the bonding area together to get sufficient flux from diffusion.

Best Regards,
Chad Brubaker

EV Group       invent * innovate * implement
Senior Process Engineer - Technology - Tel:  480.727.9635, Fax:  480.727.9700  e-mail: c.brubaker at EVGroup.com, www.EVGroup.com

-----Original Message-----
From: mems-talk-bounces at memsnet.org [mailto:mems-talk-bounces at memsnet.org] On Behalf Of John Dangtran
Sent: Tuesday, April 10, 2007 12:32 PM
To: mems-talk at memsnet.org
Subject: [mems-talk] Metallization Peel on Anodic Bonded Glass Wafer

Dear all, 

Could anyone shed us a light to the problem?  I use Ti-Pt-Au on glass, which
is anodically bonded to silicon substrate with bad results, the metal
adhesion failed.  In the past, we had good results on the 4" wafers until we
switched to the 6".  We sputter- etched 500A of glass before depositing
Ti-Pt-Au.