4
\$\begingroup\$

I’m headed back to the CNF (http://www.cnf.cornell.edu/) shortly.
It’s a user clean room facility for semiconductor processing.
My goal is to make some very simple samples, for use in student teaching labs.
I’ll take Ge and Si wafers and sputter or evaporate metal on them. The first task is to make ohmic contacts for transport measurements. (Hall effect and conductivity.)
I then had the brilliant (or silly) idea of also making non-ohmic (Schottky) contacts to the sample. One can then do CV measurements and get a third measure of the doping density. As well as having a built in diode.. it could be used as a temperature sensor.

My first trip there was wildly successful.
I sputtered two dots of titanium followed by aluminum on the samples and then evaporated two dots of Al alone.

Table of results:
Sample           Ti/Al             Al only
Si-p           diode w/anneal      ohmic contact
Si-n            ohmic              diode w/anneal
Si-int.         ohmic             (unsure.. I broke it.)
Ge-n            ohmic              ohmic
Ge-int          ohmic              some diode behavior.
                                   Poor adhesion. 

Sample I-V was done back in Buf. And then samples were annealed up to 300C on a hot plate. Except for the intrinsic Ge the metal adhesion was good.
(They all pasted the scotch tape test.)
I must say that titanium looks to be a nice metal for contacts.

OK after that intro some questions.

I feel I’m reinventing the wheel, I’ve done a lot of on-line and library reading but there is not very much practical advice for making contacts to semiconductors.
(There is a lot of theory.) I keep wondering if there is some old Bell labs technical document that could help me. If anyone knows of anything like that I hope you’ll share.

I still would like to make a diode with Ge. Any ideas of a metal to try?
Probing the Ge samples with Gold pogo pins it did look like the gold made a rectifying contact. But I’m guessing gold won’t stick very well. I thought I’d try Indium.

The Ti/Al metal on Si developed a black layer around the edges.
I assume this is oxidation of the Al.
And I’m going to try Gold for the top layer.

Finally, are there any other things/ experiments I could make/try by putting down metal on semiconductors?
I should add that the diodes displayed a nice photo-response.

Thanks for reading this rather long question.

Edit: adding Link to I-V curves (I is vertical (y) axis, chan. 2, with gain shown.) https://www.dropbox.com/sh/v7roajd0dumt1o4/AABU1luSZvJFH2IND6SkRnnWa?dl=0 Excuse the messy lab pictures, I've now got pogo sticks on my "probe station."

\$\endgroup\$
  • \$\begingroup\$ Given that Si-M has such a low bandgap, would one really form with Ge-M? \$\endgroup\$ – Ignacio Vazquez-Abrams Oct 23 '14 at 16:49
  • \$\begingroup\$ @IgnacioVazquez-Abrams, I'm sorry I don't understand your question. "In theory" I can form a diode in Ge. (given the doping type and metal work function.) From my rather limited reading Ge has a lot of surface states that give rise to something called Fermi level pinning, which is not part of the standard work function/ fermi level theory. I also want a metal that has good adhesion. \$\endgroup\$ – George Herold Oct 23 '14 at 16:56
  • \$\begingroup\$ Try to not mind me too much; I'm not quite as learned regarding the nuances of the atomic-level physical characteristics of semiconductors. \$\endgroup\$ – Ignacio Vazquez-Abrams Oct 23 '14 at 17:21
  • \$\begingroup\$ @IgnacioVazquez-Abrams, no that's fine. In a schottky I usually think about the metal work function as either being above the conduction band with n-type, or below the valance band with p-type. If the metal work function is in the gap. (as I think you were asking about) well I don't know. This is a nice picture for n-type en.wikipedia.org/wiki/Schottky_barrier There Theta_sub_B is the barrier height, that's related to the work function. I really don't know... hey this a nice article ... thanks! en.wikipedia.org/wiki/Metal%E2%80%93semiconductor_junction \$\endgroup\$ – George Herold Oct 23 '14 at 23:34
2
\$\begingroup\$

My got to these questions is always Sze - "physics of semiconductor devices" Fig 13 in my version shows the doping capabilities of various elements against Si, Ge and GaAs. Ti in Si is 0.21 eV away from the conduction edge for example.

Fermi level pinning refers to the surface states having having trapping centers that respond slowing and thus dominate the fermi level in the bulk.

What is interesting is that the P-Type contact (or I am have it backwards - i.e. N-Type) under classical analysis cannot form a ohmic contact. It took a while to realize that the contact was dominated by QM effects.

A lot of contacts are actually silicides, in 0.5 um processes (if memory serves me) it is NiSi, CoSi is common in the 180 - > 90 nm era.

Gold in Si has two states, a Donor that is 0.26 eV above the valence band and an acceptor that is 0.54 eV below the conduction band.

Interestingly, in Ge most metals have multiple ionization levels but chromium has two closely spaced Donor levels that might work as a diode.

I am surprised that you got any annealing at 300 C it usually takes until 900 + C before Si can re-anneal from implant damage. That is what RTP does.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ Thanks for the answer, Yeah I've been all over Sze. (My copy is at work, but I think I know the figure.) That's an "in theory" graph. Sze also talks about gold doped with a bit of Ga or As as being used for contacts, but I don't have access to exotic type mixtures. (The wiki metal-semi-junc. link above is pretty good, there are some references I'll have to chase down.) Re: annealing this mostly made the diodes better, though it's a little hard to tell. I have two of the same contacts on each sample so imagine the I-V of two head to toe diodes. What's RTP? \$\endgroup\$ – George Herold Oct 24 '14 at 12:23
  • 1
    \$\begingroup\$ RTP = Rapid thermal processing, top sfc flash melt and subsequent regrowth. The Sze chart I reference is measured data.Andy Groves book, available as a reprint is also v. v. good.It is also a go to source "physics and technology of semiconductor devices" The other books that I use a lot are Wolf's "Silicon Processing" a 4 part series, but that is north of $1000 but does have details. If you want I can dig into my copy. \$\endgroup\$ – placeholder Oct 24 '14 at 12:50
  • \$\begingroup\$ OK I added a link to I-V data on my question. Re: Sze, I've got the 2nd ed. I remember a figure/table with metal work functions listed with/for Si, Ge and GaAs. I've ordered a copy of Groves, thanks. Re Wolf, thanks for the offer, but don't wast your time on my account. I'm trying to make these with a very simple process. No photolith or doping, just a few metal masks and bare wafers. I think Si is "in the bag" as long as I can put gold over Ti and Al. Ge is actually more interesting from a teaching stand point. (intrinsic at a lower temperature.) \$\endgroup\$ – George Herold Oct 24 '14 at 13:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.