<Laurenceb>
didnt know there was a channel for this :P
<Laurenceb>
i was thinking about how it might be possible to make a small chip fab for rapid prototyping
<Laurenceb>
has anyone looked at colloidal microjets?
<azonenberg>
Laurenceb: Not that I know of
<azonenberg>
My current focus is MEMS
<Laurenceb>
yeah
<azonenberg>
I have a 20 micron photolithography process working decently well and am in the process of working out some yield issues in a tantalum oxide hardmask
<azonenberg>
to be used for KOH wet etch
<Laurenceb>
wow
<Laurenceb>
i was interested in the possiblility of avoiding masks
<azonenberg>
Hoping to have the body of a comb drive (i.e. minus metal layers) by end of summer
<Laurenceb>
and directly printing
<azonenberg>
Direct write is interesting but wont get you good resolution
<Laurenceb>
colloidal microjets can get very small
<azonenberg>
On say a 600dpi laser printer your lambda (smallest addressible pixel size) is around 42 microns
<azonenberg>
and your design rule is normally 4-5 pixels
<azonenberg>
I do 10x or 40x optical reduction on that to get either a 20 or 5 micron design rule
<Laurenceb>
yeah, i was thinking build from scratch
<Laurenceb>
thats impressive stuff
<azonenberg>
Photolithography is the de facto standard method for a reason :P
<azonenberg>
If you want to avoid masks you're better off using maskless lithography of some sort
<azonenberg>
meaning e-beam or (more affordable) laser direct write
<azonenberg>
Most of these are unmagnified or through one of my light microscopes but i did get two sessions on a SEM on campus to help debug a yield problem
<azonenberg>
Lighter gray is tantalum chloride (Emulsitone Tantalumfilm, which would, later in processing, be baked to oxidize it into tantalum oxide to be used as a hardmask)
<azonenberg>
The white particles on the background are also under the tantalum layer, causing serious cracking which was destroying my yields for a while
<azonenberg>
The SEM imaging session was intended to diagnose the source of the cracks
<azonenberg>
I eventually traced it to silicon dust left over from cleaving the wafer into indididual dies
<azonenberg>
I had done an RCA clean on some of the dies, which didnt seem to make a difference, so i had ruled out chemical contamination
<azonenberg>
But large (micron or more) particulates wont be removed by an RCA clean
<azonenberg>
And i knew there were too many to be room dust
<azonenberg>
I finally identified the contaminant when EDS on a contaminated area showed nothing but silicon
<azonenberg>
I had initially thought it wasnt sensitive enough to pick up the contaminant
<azonenberg>
then i realized the contaminant might actually BE silicon
<Laurenceb>
ESD?
<Laurenceb>
i mean eds
<azonenberg>
Energy dispersive X-ray spectroscopy
<Laurenceb>
ah
<azonenberg>
You hit the sample with an electron beam like in imaging, but you read x-rays coming off the sample
<Laurenceb>
so those cracks are due to the dust?
<azonenberg>
The electrons in the atoms are excited and then drop down to a lower energy level, fluorescing in the process
<azonenberg>
by measuring the x-ray energy level you can identify elements (but not compounds)
<azonenberg>
And correct
<azonenberg>
The coating is made by depositing tantalum chloride dissolved in ethanol by spin coating
<azonenberg>
As it dries, it cracks alogn the edges of the particle
<azonenberg>
The guy who runs the electron microscopy lab at my school is very supportive of my work and is willing to give me a few hours of scope time here and there for free :)
<azonenberg>
And even if i need more, its only $45 an hour which i can afford (as compared to almost $200/hr for the nice Zeiss in the cleanroom, which i clearly cant afford :P )
<azonenberg>
The probes are for doing testing of finished devices
<azonenberg>
i dont have a wire bonder
<azonenberg>
I do, however, have a wentworth labs probing station and a couple of micropositioners
<azonenberg_work>
bart416: Go to a school with that kind of gear
<azonenberg_work>
if you can possibly find one
<bart416>
Belgian universities/colleges don't have that sort of resources available themselves generally
<bart416>
Usually research spin offs do though
<bart416>
Or they buy time on private ones as needed
<azonenberg_work>
i see
<azonenberg_work>
RPI is a bit better funded, i guess
<bart416>
How much do you pay per year for college/university?
<azonenberg_work>
The cleanroom has a zeiss supra SEM as well as a FIB
<azonenberg_work>
And an AFM
<azonenberg_work>
The mat sci department has a TEM, a FESEM, another SEM, and i think maybe an STM
<azonenberg_work>
And hmm
<azonenberg_work>
Before scholarships etc i think the cost of attendance (tuition + room and board) is around $50k / yr
<bart416>
If they have a AFM they most likely have an STM as well
<bart416>
STMS are cheaper
<azonenberg_work>
I paid much less than that
<bart416>
By law it's set to about 550 euro / year here ;)
<azonenberg_work>
And now that i'm a grad student i get free tuition + stipend... yay teaching assistantships
<bart416>
There are a few private universities but those are for humanities
<bart416>
Not sciences
<azonenberg_work>
I see
<azonenberg_work>
RPI is the oldest engineering school in the USA... they've been around for a while :)
<Laurenceb_>
RPI?
<Laurenceb_>
thats some crazy tuition fees
<bart416>
Laurenceb, exactly
<bart416>
The tuition fees for 10 students of theirs are probably more than the budget the electronics department at our college gets for research and equipment >_>
<azonenberg_work>
Laurenceb_: The school president is also getting rich off our tuition :(
<azonenberg_work>
She makes like 1.5M USD a year
<Laurenceb_>
i graduated 2 years ago and owed UKP20K
<azonenberg_work>
Whats that in USD?
<bart416>
I'll graduate in 2 years assuming I stay here and I'll owe nobody anything
<azonenberg_work>
bart416: I dont actually owe much due to scholarships etc
<azonenberg_work>
But on paper the tuition is still insanely high
<bart416>
Due to me being lazy in highschool it's near impossible to get scholarships :(
<azonenberg_work>
i see
<bart416>
+ my father alone already earns too much for me to be able to apply to a government scholarship as well
<bart416>
And a NCO doesn't earn that much...
<azonenberg_work>
I see
<Laurenceb_>
about 14 or something
<azonenberg_work>
Well in any case, homebrew is always cheaper than buying this kind of stuff commercially
<azonenberg_work>
I'm just using their resources because they're available, and with the intention of it being temporary
<azonenberg_work>
test on their unit, then build a version of my own
<bart416>
heh
<azonenberg_work>
And then document how i did it so people without these opportunities can build straight off my lab notes
<Laurenceb_>
i dont see why FIB is impractical for this sort of stuff
<Laurenceb_>
but you cant FIB conductors
<azonenberg_work>
Laurenceb_: It's possible
<azonenberg_work>
Just very slow
<azonenberg_work>
You can, actually
<azonenberg_work>
Two ways
<azonenberg_work>
First is subtractive - sputter/evaporate the metal and then use the ion beam to cut traces out of the background
<azonenberg_work>
Second is additive
<azonenberg_work>
I dont fully understand the process but i recall it being something like targeted sputtering
<azonenberg_work>
the ion beam hits a chunk of (say) tungsten or platinum
<Laurenceb_>
dont you need really powerful beam for cutting?
<Laurenceb_>
the targeted spluttering sounds sane
<bart416>
Depends on the thickness of the material
<azonenberg_work>
then sends those atoms flying at high speed into your substrate, where they stick
<azonenberg_work>
I know its done
<azonenberg_work>
And our FIB is capable of both additive and subtractive machining
<azonenberg_work>
But its normally waaaay too slow to be of much use for anything big
<azonenberg_work>
Its main use is thinning of samples for TEM
<Laurenceb_>
it doesnt sound completely impractical to build
<azonenberg_work>
Or for failure analysis
<Laurenceb_>
or for prototyping
<azonenberg_work>
FIB for prototyping would be as slow as ebeam
<azonenberg_work>
if not more
<Laurenceb_>
depends what you are making i guess
<azonenberg_work>
Yeah
<azonenberg_work>
I still like mask lithography
<azonenberg_work>
But i want to use laser direct write for making the masks
<bart416>
All a FIB really is is a particle accelerator
<azonenberg_work>
right now i'm doing printing onto a transparency
<azonenberg_work>
bart416: Not just
<bart416>
the basic form of it is
<azonenberg_work>
A particle accelerator plus the ability to aim the beam precisely :P
<Laurenceb_>
i have a laser galvo off ebay
<Laurenceb_>
but its analogue
<bart416>
I wonde rif you could hijack the electron cannon of a TV for this sort of thing
<bart416>
use an ion source instead of an electron source
<azonenberg_work>
I was actually considering using an electron gun for my homebrew SEM
<azonenberg_work>
But thats a year or more out
<azonenberg_work>
i want to get most of the fab worked out by then
<Laurenceb_>
i guess you could have a two axis mechanical rig then an electrostatic or magnetic two or single axis ion beam
<Laurenceb_>
for fine detail
<azonenberg_work>
Move the stage for coarse stuff
<azonenberg_work>
then beam shift for fine
<Laurenceb_>
yes
<azonenberg_work>
Thats how SEMs do it too
<Laurenceb_>
my origional idea for maskless was wet process using colloidal microjet
<Laurenceb_>
you can get to crazy small droplet sizes
<bart416>
You could use a pulsed laser maybe?
<bart416>
With a kerr lens
<azonenberg_work>
bart416: I want to do laser but not for maskless litho
<azonenberg_work>
It would be for generating the mask
<azonenberg_work>
sputter a microscope slide in a few hundred nm of metal
<Laurenceb_>
but youd still need a way to metal coat
<azonenberg_work>
Laurenceb_: Evaporation
<Laurenceb_>
- with microjet
<Laurenceb_>
yeah but then you need vaccuum
<azonenberg_work>
You'd need that for a lot of processes
<bart416>
yeah, but you can use the laser through the glass of the vaccuum chamber if you make it well
<azonenberg_work>
And the problem with microjet is that you cant do anything that isnt soluble in a compatible liquid
<azonenberg_work>
bart416: You misunderstand
<azonenberg_work>
Coat the sample in metal, then remove from vacuum
<azonenberg_work>
spin coat in photoresist
<azonenberg_work>
and use a UV laser to expose the resist
<bart416>
You could also do that
<azonenberg_work>
Then develop and etch
<azonenberg_work>
and use the result as a contact mask
<Laurenceb_>
i saw some ideas about using 'nanoparticles'/dust in a carrier liquid
<Laurenceb_>
then sintering/diffusing
<Laurenceb_>
cant see it working well with Al
<Laurenceb_>
chemical plating with two liquids might work
<azonenberg_work>
Hmm
<azonenberg_work>
Well, Al is being phased out due to higher resistance
<azonenberg_work>
Modern ICs are moving toward Cu, which has its own set of problems
<azonenberg_work>
the most obvious being that Cu diffuses into Si and disrupts PN junctions
<azonenberg_work>
so you need a barrier metal
<azonenberg_work>
usually its not even a metal, i've heard TiN (a ceramic) is commonly used
<azonenberg_work>
just a few nm to prevent diffusion
<Laurenceb_>
eww
<Laurenceb_>
i dont think its worth trying to emulate the state of the art
<azonenberg_work>
I'm not recommending it, just saying Al isnt the only option
<bart416>
Can't you use SiO2 as barrier?
<azonenberg_work>
bart416: No
<azonenberg_work>
a) you need metal-to-Si contacts, the barrier between them has to be conductive
<bart416>
ah
<azonenberg_work>
b) metal ions diffuse through SiO2
<bart416>
You meant contacts themselves
<azonenberg_work>
And the barrier has to completely surround the wire
<azonenberg_work>
Normally you do a dual damascene process