#homecmos IRC log for Friday, 2011-10-21

Action: azonenberg starts designing mask for fab run tomorrow07:27
azonenberghi guys08:16
soul-dand added to startup  join :P08:16
soul-dwell that was my point azonenberg  i usaly don't understand stuff  untill one works backwards and make stuff yourself08:17
soul-dcool stuff so this is where  digital meets chemical ?08:21
azonenbergsoul-d: digital, and mechanical08:21
azonenberghttp://i.imgur.com/wVEEB.png full wafer08:21
azonenbergI am going to be doing the first two masks by contact lithography on a whole wafer (they dont need to be very accurate and will be the same for all dies)08:22
R0b0t1Which program is that?08:22
azonenbergfree but closed source08:22
azonenbergnone of the open tools were to my liking so far, i'm still searching08:22
azonenberghttp://i.imgur.com/rTEkR.png is a single die08:22
R0b0t1Are you at the point where you are actually going to etch a full wafer? What's on the chips?08:22
azonenbergnot yet fully designed08:22
azonenbergOk, here's the process overview08:23
azonenbergSpin coat resist, pattern "dicing" mask (blue rectangle in the background, full chip) by contact litho over a full 2 inch wafer08:23
azonenbergThis will leave lines (the space between the white and blue outlines) of photoresist with space between them wher ethe dies are08:23
azonenbergEvaporate 1000nm of Cr on both sides of the wafer08:24
azonenbergLift off in acetone08:24
azonenbergthis will result in the back of the wafer being solid Cr and the front being Cr over dies and bare silicon over the spaces between them08:25
azonenbergEtch in hot KOH for half an hour or so08:25
azonenbergactually more like 45 mins08:25
azonenbergEnough to go down about 100 microns08:25
azonenbergSo now i have Cr over unetched Si on the front and back of the dies and trenches 100um deep between them08:25
soul-dsounds like fun08:25
azonenbergNext step is to do a second contact mask08:25
azonenbergthe pink "backside" mask08:26
azonenbergalign to the edge of the dies08:26
azonenbergand etch the Cr in HCl : H2O208:26
azonenbergthen return to the KOH08:26
azonenbergnow i am etching both in the center of the die and between the dies08:27
R0b0t1So your photoresist is positive, turn that into a negative of Cr, then etch?08:27
azonenbergbut the trenches are 100um deeper08:27
soul-dsee lots of  fun chemical processes08:27
azonenbergR0b0t1: I havent figured out all of the polarities for the masks yet08:27
azonenbergin the CAD right now, filled areas are to not be etched08:27
R0b0t1Well, photo -> chrome -> etch?08:27
azonenbergi may invert them before fab08:27
azonenbergdepending on where i do liftoff08:27
azonenberganyway so i then KOH a second time08:27
azonenberguntil the dies separate08:27
azonenbergAt this point, i have 76 loose chips08:28
azonenberg500um deep08:28
azonenbergwith 400um deep rectangular holes in the backs08:28
azonenbergFrom now on, processing is done by projection litho one die at a time08:28
azonenbergas the process is a little more experimental (and features are smaller)08:28
azonenbergI then pattern the "fingers" mask08:29
azonenbergWhich i havent finished designing08:29
soul-dok so  you made  the "base"  first ?08:29
R0b0t1So is this etch of the whole wafer to get you separate chips?08:29
azonenbergR0b0t1: Yes, as well as to thin the dies in the center08:29
R0b0t1ah okay08:29
azonenbergleaving a full-thickness ring around the thin part for easier handling08:29
azonenbergThen i do the finger mask which is my actual MEMS geometry08:29
azonenberggarden-variety comb drive08:29
azonenberg20um fingers at 20um spacing08:29
azonenbergEtch this into the top surface08:30
azonenbergWhile leaving the 400um deep hole unmasked08:30
azonenbergSo i'm now etching down and up simultaneously08:30
azonenbergThey'll meet around 50um from the top of the die08:30
azonenbergleaving 20um wide and 50um tall fingers08:30
azonenbergI then deposit an insulator - either spin on glass, thermal oxidation, or something TBD08:30
azonenbergevaporate metal (aluminum most likely)08:31
azonenbergand pattern the metal1 mask to separate the stator from the rotor08:31
azonenbergas well as adding probe pads that i can test the device with08:31
azonenbergSonud like a reasonable process?08:32
soul-dsounds like fun :)08:32
R0b0t1It makes sense, but seems rather convoluted08:32
R0b0t1But right now you are only using one type of silicon?08:32
R0b0t1What about making transistors and having to use the P and N silicon?08:33
azonenbergR0b0t1: I am only doing MEMS atm so doping is unimportant08:33
azonenbergi think all of the wafers i have now are P type boron doped08:33
azonenbergWhen i get to transistors, i will be using spin coated dopants and diffusion08:34
azonenbergspin coat a layer of doped SiO2, photoresist over that, pattern08:34
azonenbergetch oxide in 3% HF08:34
azonenbergbackfill with undoped oxide08:34
azonenbergCook at ~1100C to diffuse dopants into the silicon08:34
azonenbergstrip oxide08:35
soul-dill leave the HF  to you though08:35
azonenbergdeposit field oxide08:35
azonenbergsoul-d: 3% is actually not THAT bad, though you still want to be careful08:35
azonenbergi buy it at the local grocery store08:35
azonenbergits an off-the-shelf rust remover08:35
soul-dmmm ?08:35
azonenbergwhink brand rust stain remover08:35
azonenbergbrown plastic bottle08:35
azonenberg1.5-3% HF08:35
soul-dok lol cool08:36
soul-dyeah ,y shed is chemical hazard anyhow08:36
azonenbergAt the same time, you want to wear gloves and preferably a lab coat08:36
azonenbergas with any acid08:36
soul-dgot them  and  full face shield08:36
soul-dabec1 filters08:36
azonenbergYep, me too08:36
azonenbergWhat i am actually working on setting up is local exhaust for solvents08:36
azonenbergi have a 6" duct fan but have to build the rest of the hood08:37
azonenbergIn any case its 4:30 in the morning and i have homework due at 170008:37
azonenbergand class at 140008:38
azonenbergI'd like to be somewhat awake for them08:38
azonenbergso i'm off to get some sleep08:38
Action: azonenberg AFKs: 503 Geek Temporarily Unavailable\r\nX-Server: azonenberg\r\nRetry-After: 21OCT2011 1030EDT\r\n\r\nThe requested geek is asleep. Your request should be repeated in six to eight hours.08:39
azonenbergAnd who said anything about night? It's the morning here :p08:40
soul-dheh true  nasty habbit i have to  night time seems to work better08:42
reportingsjrazonenberg: I would say it's /almost/ morning here, haha08:49
soul-dmultilayer ion paper :D12:11
soul-dwould be easy to make12:11
soul-dcould do fast prototypes if you have reliable draw surfaces   probably more reliable then breadboard :)12:13
CIA-67homecmos r147 | trunk/mems-tests/ (22 files in 12 dirs) | First draft of new comb drive mask set20:44
azonenberglekernel: http://colossus.cs.rpi.edu/~azonenberg/downloads/combdrive-wafer.gds22:00
azonenbergdoing the litho for the first mask today, then the evaporation and liftoff monday or tuesday (was gonna be today but was delayed by homework)22:01
azonenbergtwo contact masks on the whole wafer, then dicing and two projection masks on each die22:02
soul-dyou still where able to do homework  ? :D22:04
azonenbergsoul-d: Got everything done on time, yes22:04
azonenbergSkipped my one class of the day but its boring22:04
azonenbergits the EE version of stuff i already learned in CS22:05
azonenbergBut i need credits in my schedule22:05
soul-dcoulnd you jsut pretend to be there and do saomthing else22:05
azonenbergi took so many CS classes as an undergrad that i'm having problems taking enough as a grad22:05
azonenbergall of the interesting ones i already took22:05
azonenbergso i have to find another dozen classes to take over the next few years22:05
azonenbergi dont think there are 12 classes offered by the department i havent taken :p22:05
azonenbergand especially if you count ones like "numerical methods for differential equations" that are super math heavy?22:06
azonenbergSo i'm starting to branch out into other departments out of necessity22:06
azonenbergunfortunately, this often results in overlap22:07
azonenbergand by the time you figure out you're going to learn nothing in the class, and it's too late to add a new one and drop this one, you're stuck with it22:07
soul-dheh hard to get motivated22:07
azonenbergfunny thing is, this is the only EE class i've taken to date22:08
azonenbergbut its all stuff i learned in CS22:08
azonenbergAnyway so as soon as it gets dark enough that i can do litho without sunlight frying my wafer, i'm going to do a photoresist coating followed by my first attempt at full-wafer contact litho22:10
azonenbergGood news, the feature size is huge lol22:11
azonenbergsmallest geometry on this mask is 200 um22:11
Action: azonenberg should probably look into buying professionally made masks just once to see how good they are22:12
azonenberg12x18 inch mask for $5222:12
azonenbergi could fit an entire 4+ mask set for a 2-inch wafer on that sheet22:13
azonenbergat 8000 DPI22:13
azonenbergthat's around 3.175 micron lambda with a four pixel min-feature design rule22:13
soul-dlol wait  let me scroll back din't look for a sec22:13
azonenbergAnd if i do optical reduction of say 4x on that mask22:14
azonenbergi could do some pretty fun stuff22:15
azonenbergbut even without reduction, full contact litho would be cool22:15
azonenberghi cheater22:17
azonenbergI think once i get the comb drive mostly working on this process22:18
azonenbergi am going to try ordering a film mask set from laserlab22:18
azonenbergthe comb drive masks plus some test patterns22:18
azonenbergand see how it turns out22:18
soul-dyou should make  youtubes :P soem visual  stuff with your explaination might make more sense for me ;)22:19
lekernelwhat are the missing steps for the comb drive?22:19
azonenbergIn theory, i know how to do everything22:19
azonenbergthere are two big unknowns atm22:19
azonenbergfirst, how i am going to grow dielectric over the chip before metal 122:20
azonenbergProbably thermal oxide but i need to buy / borrow time in a furnace22:20
azonenbergsecond, how to deposit metal 122:20
azonenbergi did evaporation for everything so far22:20
azonenbergbut evaporated metal is tricky because it has bad step coverage22:20
azonenbergand i'm going to want to plate the sidewalls of the fingers22:20
azonenbergSputtering is an option but more complex22:21
azonenbergIn either case i am hopeful that by next weekend i will have at least one die with the full mechanical structure of a comb drive22:21
azonenbergat which point i can run the same process on another ~ten dies22:21
azonenbergand start screwing with metal and oxidation22:22
azonenbergI feel like i'm very close, the Cr hardmask for deep etching is working well so far22:22
azonenbergthat was my last big problem22:22
cheaterwhat is step coverage?23:23
azonenbergcheater: How conformal the coating is23:24
azonenbergGood step coverage means that you'll cover sidewalls of vertical holes, etc23:25
azonenbergpoor step coverage will cover the top and the floor but not the walls23:25
azonenbergThis is, however, good for lift-off processes23:25
azonenbergwhere you need exposed sidewalls of the resist to get your solvent into23:25
azonenbergWith liftoff, you put photoresist on a bare wafer23:25
azonenbergevaporate metal23:25
azonenbergthen dip in a solvent23:25
azonenbergwherever you had resist, the metal is on top of it23:25
azonenbergso it floats off when you put it in the solvent23:26
azonenbergGenerally speaking, spin coating and sputtering are conformal (good step coverage, bad liftoff profile)23:26
azonenbergso is CVD23:26
azonenbergwhile evaporation is a good liftoff profile23:26
cheatercan you mix these techniques at all?23:27
azonenbergI mostly use spin coating at home and evaporation on campus23:27
azonenbergi plan to build evaporation and sputtering rigs at home eventually23:27
azonenbergCVD is more complex and dangerous as it requires compressed cylinders of some distinctly unpleasant gases23:27
azonenbergi think i cna avoid it entirely23:27
cheaterno reason to die quickly23:32
azonenbergWell, worse than CVD would be RIE23:33
azonenbergOne of the nicer chemicals they use for RIE is chlorine gas23:33
azonenbergBut they've used stuff as nasty as chlorine trifluoride23:33
azonenbergthe two biggest baddest halogens all in one molecule23:33
azonenbergthe stuff ignites on contact with ANYTHING23:34
azonenbergincluding glass and asbestos23:34
cheaterwouldn't want to breathe that in23:34
azonenbergAnd produces hydrofluoric acid in the smoke :p23:34
cheaterdoesn't chlorine trifluoride get used as a rocket fuel too23:34
azonenbergcheater: they tried a few times23:35
azonenbergit's extremely effective23:35
azonenbergif you can manage to handle it23:35
cheaterwasn't there that funny blog post about it23:35
cheaterheh, is it difficult to fabricate?23:36
azonenbergFrom what i understand, get a cylinder of chlorine and fluorine23:36
azonenbergand mix them up lol23:37
azonenbergIf anything within a quarter mile of you survives the resulting reaction, you now have a tank of ClF323:37
cheateri wonder if you need to do this under specific pressure/temperature23:38
azonenbergI have no intention of attempting the reaction lol23:39
azonenbergOr even going anywhere near the precursors23:39
azonenbergthe strongest thing i use in my fab work is 1-3% HF23:39
cheaterin what23:42
cheaterthat sounds like mouth wash23:42
azonenberglol, no23:43
azonenbergyou're thinking NaF and its around 0.01%23:43
azonenbergmaybe less23:43
cheateryeah, one digit less or more.. who cares23:43
cheateras long as it keeps your breath fresh23:43
azonenbergmost definitely not mouthwash :p23:44
--- Sat Oct 22 201100:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!