Anodic Bonding Vincent’s Chip Stack

ByGreg Madejski

Previous posts: Anodic Bonding Theory, Anodic Bonding Setup

 

I had originally hoped there would be enough residual impurity in the oxide layer to create a direct bond with Vincent’s chips. This was a foolish hope. Pyrex is at least 4-5% doping and standard p-doped silicon wafers are 5-6 orders of magnitude less than that, with sodium impurities being 5-6 orders of magnitude below that. Fortunately, it has been shown in literature that bonding can be performed from evaporated glass layers. This is trickier than it sounds. There must be enough sodium/dopant present in the evaporated glass to produce defect states that can be torn off later, but still maintain a solid crystal. This means that the glass must simultaneously maintain a coefficient of thermal expansion close to silicon while being doped in situ.

Anodic Bonding propagation
Anodic Bonding Silicon Nitride to Coverglass, 5-40 minutes 700V, 500C. The bonding front moves as more glass is sealed to create a bubble free sealing layer.

 

 

Anodic Bonding is strong enough to rip out pieces of silicon from a wafer
Anodic Bonding is strong enough to rip out pieces of silicon from a wafer

 

Coverglass coefficient of thermal expansion is not well matched to the nitride chip, causing it to crack and splinter as it cools.
Coverglass coefficient of thermal expansion is not well matched to the nitride chip, causing it to crack and splinter as it cools.

 

So I tried the easiest thing: crush up some different glasses that should allow anodic bonding (assuming it makes it to the chip), evaporate them over top the oxide spacer chips, and then run the anodic bonding process.

 

Pyrex

Corning 7740 is commonly sold under the trademark of Pyrex. So I found an old pyrex funnel that was broken and ground it up.

Mortar and Pestle used to break down chunks of glass
Mortar and Pestle used to break down chunks of glass

 

Pyrex itself doesn’t melt very easily, I used 12-13 mA to get an appreciable deposition rate. I deposited at 250 C, targeting 100 nm and 500 nm thickness (verified with profilometer +/- 10%).

Tried bonding a broken glass piece to silicon nitride. It failed, probably due to lack of contact.
Tried bonding a broken pyrex glass piece to silicon nitride. It failed, probably due to lack of contact.
500 nm of pyrex on an oxide spacer, and a pyrex spacer being bonded to a nitride flowchip. The channel height (~100 nm) is maintained from the evaporation.
500 nm of pyrex on an oxide spacer, and a pyrex spacer being bonded to a nitride flowchip. The channel height (~100 nm) is maintained from the evaporation.

No bonds were successful at 700 V for 20 minutes. Nothing I did with pyrex seemed to bond, even though there is literature associated with this material. Do I have a bad piece of pyrex? Is it not being evaporated well?

 

VWR coverglass

It has been very easy to get coverglass bonding to various chips. So I ground up some glass and evaporated it.

 

Ground up coverglass
Ground up coverglass

This deposition was poorly controlled, with wild fluctuations in the deposition rate. In the end, I only deposited about 50 nm of material.

 

The Coverglass evaporation (Top right, purple) chips, and the 500nm pyrex chips (left green).
The Coverglass evaporation (Top right, purple) chips, and the 500nm pyrex chips (left green).

Remarks

None of these combinations worked. Short circuit currents were apparent in the  bonding runs ( > 1 mA), meaning that no appreciable voltage would be observed across the oxide, meaning no electrostatic bonding. I am not getting enough of an insulating effect from the oxide, which is probably attributable to the low thickness of the oxide. I never tried annealing the glass at higher temperatures, though that may be a way forward to try and make a more homogenous crystal, and lessen the defect states that allow short circuits. A few membranes broke and stuck to the surface of the oxide spacer chips; I am not sure if they were anodically bonded or simply charged.

 

Future plans

Now that I have tried the easily available material, I will attempt products designed for this kind of bonding. Lithoglas, Schott 8329, Ground up Corning 7740 wafers. We have all sorts of ways to pattern the glass with lift off/spin on glass/evaporation, but we still have yet to get a successful bond with only a thin film of glass. Repeatable bonding has been demonstrated in literature at thicknesses of 1 micron, with 500nm glass layers being also reported. 100 nm thick bonds have been elusive.

Similar Posts