What substrate material are you depositing on?

Whoops of course, SiO2; sorry for the omission.

Hi Darin,

I had similar problem when I fabricated OTFT with a short channel. Only in my case the gap between the Gold electrodes is much wider (~50nm) and I don’t produce it via electromigration.

Eventually I used Cr as adhesion layer, and after producing the gap I used a short step of Cr etch to remove the Cr from the gap. It couldn’t work with Ti since our Ti etcher contain HF so it attacks the SiO2

I don't know if it helps.  When I do ebeam evaporation, there are two options.  One is vertical evaporation, the other is rotational evaporation (2-axis substrate rotation). The resulting uniformity and step-coverage will be different. 

Thanks all!

Roy - good trick, unfortunately i need to leave my devices in vacuum after formation to preserve the gap.

CJ - I use a rotating chuck. I don't have any evidence that the Ti isn't uniform, but I'd heard that Ti doesn't uniformly coat below 4 nm, although I haven't been able to find a paper stating that.

In our experience with Ti on both Si and SiO2 we tend to aim for 3 to 5nm, (so about 4nm on average) and I'm not aware of us ever having problems with this thickness, although of course we are not doing your experiment. One thing I will say is how often do you check the deposition rate of your system and what thicknesses do you check if it all? They all tend to drift around a bit over time and even if you have a set recipe it will change due to system pressure variation, if the target has a bit of oxide on it, the energy and current of the e-beam drift and many other subtle variations, all of which might make the difference of aiming for 2nm and getting a bit less that might make for a continuous film or not. We tend to test ours at few difference target thicknesses such as 5nm 10nm 20nm about once every couple of months even if we thing it is working as it should. Basically don't trust the machine, it may be in error and you can't afford much of error with 2nm.

Hi Darin,

In my experience we use roughly 5 nm Ti as adhesion layer for Au. 

The concerns you've raised are valid; have you considered losing the Ti? perhaps using a substrate with better adhesion to Au, or some surface preparation process. For Au nanowire catalysts we do not use any adhesion layer; the substrate is different of course.

Another option I could think of is using only Au for the the device itself and then contacting it by a standard Ti / Au metallization ( a two step process).

Good luck!

Great responses, I have a much better sense of the mechanics of this system now. I'm still undecided, but at least I know what's at stake.

Would like to suggest something from my own experience, and maybe useful for your work

Ti-Au gold contacts layers need a sputtering system, and preferably insitu deposition and some substrate temperature would be good.

since Ti is being used as an adhesion layer for the Ti-Au contacts.

Titanium thickness should not be more than 200 anstroms.

Alternately you can also think of Cr-Au. again Cr thickness should be in the range 200 to 300 angstroms, it should look like a semi transparent greyish film. 

Etching of Au and then Chromium is relatively easy. Insitu deposition of Cr-Au in evaporation system, at a substrate temperature of 200 deg C, would give you very adherent contacts.

K. Sreenivas

Just to add to the wealth of answers:

I have found 2 nm of Ti sufficient (rotating chuck), but have also heard through the grape vine that the crystalline structure is altered at 3-4 nm so I usually aim for 5 nm.

A word of caution with Cr as an adhesion layer; if the Au is to be used as a sensing surface, be wary that annealing occurs at surprisingly low temperatures and Cr will readily diffuse into the Au layer changing the characteristics (see two attached papers). As I tend to use the Au surface for electrochemical biosensing applications I have since switched from Cr to Ti as this allows the sensors to undergo various heat treatments whilst retaining more stable baseline responses.

Yonaton - that's an excellent idea, can I ask what substrates you use? My understanding however was that the Ti or Cr adhesion layer is there less to interface with the oxide, and more to interface with organic contamination on the substrate surface. Being oxidizable, Ti and Cr bind to the organics more readily than Au, providing essentially a cleaner surface to deposit onto. To that end we've also considered using a directed plasma etch right before metal deposition in the cleanroom to minimize organic contamination and hence Au adhesion, but I suspect we'd still have adhesion issues even from adsorption of the PMMA solvent (anisole) degassing in the vacuum chamber.

Jules - excellent comment, thanks for the advice. We'll stick with Ti then.

With further experimentation, we've found that when the Ti layer is too thick it does indeed contribute to the overall current flow in a manner that interferes with our Au electromigration. To avoid that, we only deposit 3-5 angstroms of Ti. I suspect that does not form a continuous layer, which is good for us as that will minimize the role it plays, but seems to still sufficiently bind the Au to the surface. We are still conducting experimentation to see if this approach is effective.

We check the thickness via AFM to make sure it hasn't drifted.