I'm working on a complicated fabrication project, and it just so happens that if I was able to wire bond through a thin dielectric of approximately 40nm, it would make this fab considerably less time consuming. It seems likely that this would work to me, but I haven't tried. I'm using a fairly modern Westgate wedge bonder.
It would be gold wire bonding to gold pads, roughly 180x120 um in size. The dielectric is ALD Al2O3 (40nm of it).
Side Note - if you're wondering why I can't just etch vias. Well, I can, but I have to dice up my samples into chips to do really fine e-beam lithography. We use a converted SEM to do our e-beam writing so I can't write on full wafers, they must be small chips. Writing large bond pad vias on an SEM takes forever. The dieelctric is deposited in between e-beam steps, so I can't etch vias before I dice. I would much rather just skip what turns out to be an entire days worth of writing if possible.
I suppose I could always just try this out, but if anybody has tried doing this before it would be good to know in advance! Thanks.
I am afraid that in order to get a good contact you have to etch Al2O3. If you do not, 40 nm of high quality ALD Al2O3 will be pretty resistive. Depending on applied voltage, contact resistance will be in the region of MOhms or more. This will kill performance of any semiconductor device.
I am not doing such things, but just a thought for your technically challenging problem.
Ti/Au and Cr/Au are known to form adherent contact layers on insulators.
I would think of the following. The location where you want to wire bond, deposit a circular dot area of (1 to 2mm dia) with either Ti or Cr, through a shadow mask/lift off technique. Then bond your Gold wire to this Ti, or Cr patch. Well this will definitely allow you to bond the gold wire conveniently because ti/Au and Cr/Au can form an alloy and given you an adherent wire-bond contact.
The questions arises if your wire bond has gone through the thin dielectric layer, as per your desire for such a thing for some specific reason.
You need to ensure if Cr or Ti can diffuse through the thin dielectric layer, or alternately with any other metal which can diffuse through your 40 nm dielectric layer. In this way you are selectively making the dielectric area turn -metallic-, and then bonding your gold wire to it.
K. Sreenivas
Al2O3 ist 'hard stuff'' in more than one way.
What I'd try - provided that your bonder has ultrasonic support - is to try to get the Al2O3 cracked by applying ultrasonics. Might require to fiddle with the frequency to achieve maximum effect and is a quite 'experimental' approach, but worth a try.
Otherwise - as already stated by others - I doubt you'll get a reasonable contact. If a contact at all.
Regards
An afterthought: as your bond pad geometry is not that minute: would you be able to apply photo resist and some sort of "shadow mask" (could even be some Al sheet metal with holes drilled or etched) - exposing the masked dice to a 'normal' light source?
I got such things done (by others) in some development project. Worked quite well with positioning precision better than +- 25 µs. And this is certainly not the best precision - it was just what did suffice our geometric needs.
The short answer is no - most likely either wire would not bond, or if it bonds mechanical strength of bond and electrical connection would be very poor.
But with such large bonding pads you do not need to use EBL - very crude optical lithography would do. Coat chips with photoresist, and expose pads either in optical microscope with halogen bulb illuminator or (if you have it and the parameters are suitable) in laser cutter.
For OM exposure insert red or yellow filter, navigate manually to the bondpad location, close field aperture to have just the bondpad illuminated, remove filter and wait. Area illuminated by white light of halogen illuminator in a matter of seconds and you can insert red/yeloow filter and navigate to the next bondpad.
In laser cutter - set very low power with blue laser light, locate "cutting box" over the bondpad, and expose photoresist by a few flashes.
You would need to tune time of OM illumination or power/time of laser flashes in the cutter to get adequate exposure of the resist over bondpads.
You can always try scratching through the oxide using a diamond scribe at the point where you need to bond - if that is an option. Works if you don't need to too many devices.
- Badges
- Science topic
- Similar topics
- Correction