<wpwrak> in the end, you'll have a joystick with which you direct the death ray that picks off those last molecules that resisted everything else you threw at them ;-)
<wpwrak> azonenberg: you guys lost me at KOH vs. HF. i live in HCl+H2O2 stone age. what do these critters do ? i suppose at least one of them is an etchant, but does this make the other ?
<azonenberg> wpwrak: HF is hydrofluoric acid, which eats
<azonenberg> SiO2 (glass) and related materials
<azonenberg> But not silicon or polymer materials
<wpwrak> sound like nasty stuff :)
<azonenberg> HF is quite nasty, thats why i work with 2% solution and am still paranoid with double gloves, a face shield, etc :P
<azonenberg> Overkill considering the stuff in this concentration is household rust remover
<wpwrak> i usually consider glass and ceramics as my final barrier ;-)
<azonenberg> But i've seen what the strong stuff can do
<azonenberg> My photoresist is a polymer that dissolves in a weak solution NaOH (household lye) when it's been exposed to light
<azonenberg> UV in particular
<azonenberg> So you shine UV through your mask onto the resist, then develop
<azonenberg> Then you can use HF to eat away the glassy layer under it
<azonenberg> but only where light hit the mask
<wpwrak> (rust remover) hmm, interesting.
<azonenberg> It's rare, most rust remover is phosphoric acid
<azonenberg> But Whink brand is HF based
<azonenberg> KOH is potassium hydroxide (caustic potash) which will eat silicon
<azonenberg> It also attacks photoresist so i need the Ta2O5 glass layer as a mask
<azonenberg> i.e. pattern hardmask with photoresist, then pattern silicon with hardmask
<azonenberg> using two etch steps
<azonenberg> The nice thing about KOH is that it's sensitive to the bond structure in the silicon crystal
<azonenberg> It eats nearly 100x faster along one axis than the other one
<azonenberg> So by choosing the angle that you cut the wafer vs the bond planes you can get almost no etching (surface on the <111> vector), 57 degree sloped sidewalls (surface on the <100> vector), or vertical sidewalls (surface on the <110> vector)
<wpwrak> brr. messy. i kinda liked whatshername's transistor-making process you once mentioned. obviously, the process control roles are reversed there, but it seems that isn't nothing too horrible to do
<azonenberg> Jeri used HF
<azonenberg> Same concentration as I am - 2% from Whink rust remover
<azonenberg> The main difference is she was doing patterning by hand with electrical-tape masks etc
<wpwrak> yeah ;-)
<azonenberg> I'm doing real lithography at several-micron scales
<azonenberg> And i'm not to the point i can do transistors at this scale by far
<azonenberg> The first step is MEMS
<azonenberg> Build a comb drive
<azonenberg> And before that, basic 2D engraving in Si
<azonenberg> Shallow etching, 10 microns or so
<azonenberg> just characterizing the process
<wpwrak> i don't know much about mems or their use. i have some ideas what a transistor can be good for, though :) one problem i'd envision is just connectivity. that's some more layers.
<azonenberg> Connectivity, well
<azonenberg> Etching metal is a solved problem, so is alignment
<azonenberg> The unsolved problem is how to get the metal there in the first place
<wpwrak> yup
<azonenberg> there are several deposition processes (sputtering and evaporation are the two big ones) but both need high vacuum
<wpwrak> which requires some alchimist's lab in paris
<azonenberg> Lol, or something i build here in new york
<azonenberg> The first step will be etching these patterns http://i.imgur.com/VDW36.png just a few microns deep
<wpwrak> and people are worred about cern's black holes eating geneva ;-)
<azonenberg> Lol
<azonenberg> Black is informational only (not on the actual mask)
<azonenberg> The outer disk around each pattern is the field of view of my exposure system (about 500 microns)
<azonenberg> Blue areas are masked off, white is to be etched
<azonenberg> Everything outside the disk is not etched
<wpwrak> damn small. i still kinda like the idea of a dual BR direct exposure system.
<azonenberg> I'm still thinking of direct write but not for exposure on the wafer
<azonenberg> it'll be used for making metal-on-glass masks
<azonenberg> that are then used in my projection system
<wpwrak> if you have it, why not use it for everything ?
<azonenberg> Because it can pattern but has little alignment capability
<azonenberg> unless you build it in a far more advanced way
<azonenberg> The idea was, stick mask blank into the system
<azonenberg> zero it *somewhere* on the blank
<azonenberg> expose, develop, etch
<wpwrak> hmm. i'd try to align with perhipheral features. edges, holes made for that purpose, whatever.
<azonenberg> now you have your mask, 100nm of (say) nickel plated on glass
<azonenberg> With 50 micron (for example) features
<azonenberg> Feed that into my current exposure system
<azonenberg> Reduce 10x
<azonenberg> You now have 5um features and a 2mm FOV, which is enough for a pretty good sized die
<azonenberg> It'd be enough to build a 4004 i think
<wpwrak> 2 mm is plenty, yes
<azonenberg> 500um is FOV with 40x objective
<azonenberg> 2mm is FOV with 10x
<azonenberg> But that also means my feature size grows 4x so the only way to prevent that is to make the mask 4x smaller
<azonenberg> i.e. 50um half-pitch vs 200um (the limit of my printer)
<azonenberg> I could even hit 5mm if i used the 4x obj but that'd be pushing it
<wpwrak> or just use direct writing ;)
<azonenberg> Because i'd need 20um features on the mask
<azonenberg> And those would be so small i wouldn't be able to see them at 40x magnification easily and get good alignment
<azonenberg> 100x is about the minimum to get micron-level alignment
<azonenberg> And like i said, i'd love direct write onto the wafer
<azonenberg> I just think it's beyond trhe scope of a rev 1 process
<wpwrak> whatever rev 1 may be :) for me, your inverted scope is an excellent hack. not only for the idea itself, but also because it breaks down one of those imaginary barriers.
<azonenberg> Exactly
<azonenberg> That was the single biggest unsolved problem for me, my group at RPI was puzzling over lithography and mask alignment for months
<azonenberg> thinknig about building an optical column from scratch
<azonenberg> Then i realized we already had one, problem solved
<wpwrak> but for a more dependable process, it seems to me that BR drives are an excellent opportunity. they have the resolution and, thanks to mass-market volumes, they're relatively cheap.
<azonenberg> Yeah, I fully agree with you there
<azonenberg> It's just something that should be saved until the rest of the process is somewhat more mature
<azonenberg> If you want, i can set you up with access to the project http://code.google.com/p/homecmos/ so you can hack on the wiki and start throwing thoughts in there
<wpwrak> (building an optical column) looks like one of those problems that seem ridiculously simple - once you've solved them ;-)
<azonenberg> Exactly, once you have all the numbers and plans anybody can go build it
<azonenberg> It's the engineering that's hard
<azonenberg> And, once you've built it, the calibration
<azonenberg> getting everything precisely in focus
<wpwrak> i have a strong belief in the law of large numbers ;-)
<azonenberg> Lol
<azonenberg> And this is a solved problem with the microscope - which is a pretty nice piece of glass for the price
<azonenberg> The only big problem is chromatic aberration at higher powers
<azonenberg> But when using monochromatic light from an LED? Non-issue
<wpwrak> the microscope looks more like a mid-way point. it solves the problem at hand, showing that it can be solved, but it suffers many limitations as well (FOV, for one)
<azonenberg> Correcty
<azonenberg> Hence why it's a rev 1 process
<wpwrak> besides, it's not really "cheap". while BR is, by and large.
<azonenberg> The big problem, like i said, with BR is alignment
<azonenberg> It's going to have to be done like they do with steppers
<azonenberg> you have purpose-built alignment marks around the edge of the die
<azonenberg> that it calibrates off
<wpwrak> i stil think you can just use opportunistic alignment. find any features, and align to them.
<azonenberg> Maybe for rev 3? But it'd be tricky
<azonenberg> This would be quite the hack if we pulled it off though
<azonenberg> Direct write laser lithography system with 5um or better features
<wpwrak> you only care about repeating what you did before. but there's no absolute origin or orientation.
<wpwrak> indeed. Intel@Home ;-)
<azonenberg> Good point, only alignment matters
<azonenberg> But what about if you have a nearly blank mask?
<azonenberg> Also, when you say "align to them"
<azonenberg> The detector is easy
<azonenberg> But how do you see it?
<azonenberg> Illuminate with the laser? Bear in mind every laser pulse punches a hole in the photoresist
<wpwrak> i'd align with the chip/fragment. or if that doesn't work, make a hole and align with that.
<wpwrak> (not sure about the cleaning, though)
<azonenberg> Hole meaning drilled?
<azonenberg> Waaay too imprecise
<wpwrak> yes
<azonenberg> Do you have any idea how small five microns is?
<azonenberg> Hint, typical human hair is 20um across
<azonenberg> Let's say you drill a hole 1mm across
<wpwrak> you only care about finding the same edge several times again
<azonenberg> You need to be able to localize to within 1/50 of the hole diameter
<azonenberg> And you'd need >1 hole to get rotation locked down as well as position
<azonenberg> It'd just be too imprecise
<wpwrak> sure. make as many holes as you want. the first one is expensive, the rest is free ;-)
<azonenberg> lol
<azonenberg> the other thing is, in MEMS, puncturing the die can be problematic
<azonenberg> You might need a big piece for a heatsink or whatever
<azonenberg> It just strikes me as a really bad idea
<azonenberg> Whereas if the first exposure etched alignment marks into th esurface
<azonenberg> At 5um pitch
<azonenberg> We could then see those markings and align future stages to them
<wpwrak> you don't need the exact position relative to the hole center or such. all you need is an edge you can detect accurately (with significantly less than infinite tries ;). once you have that, you can always reuse that edge.
<azonenberg> The other thing is, silicon fractures along cleavage planes
<azonenberg> A hole is a potential fracture point
<wpwrak> (mems) yeah, dunno about mems.
<azonenberg> And, on top of that, each edge of the hole is going to be (at the microscale) broken along one of the cleaveage planes
<wpwrak> could you try to cut/mill a straight line ?
<wpwrak> you'd of course end up with a jagged edge, because you won't match the substrate's orientation
<azonenberg> Well, the substrate will have at least one edge (or a flat for a wafer) parallel to a cleavage plane
<azonenberg> you'd align theta to that
<azonenberg> So you're aligned to the crystal structure
<wpwrak> one degree of freedom solved :)
<azonenberg> (for the first level) and then etch your main pattern plus alignment markings
<azonenberg> crosses, vernier scale, etc
<azonenberg> Around the edges
<azonenberg> Then the next level just aligns to them
<wpwrak> doesn't sound too bad. you still need at least two edges to align to, though for each step.
<azonenberg> No, you need one edge for the first step since you know which plane it's on a priori
<azonenberg> From there on, you have a cross at each side of the die
<azonenberg> Align x/y to the center of them
<azonenberg> and theta so that the lines are parallel to your axes
<azonenberg> google "byu contact aligner" for Bringham Young Universtity's doc on their manual mask aligner
<azonenberg> We just need to design an automated system that can read those marks
<wpwrak> that's your microscope process. but what about the BR process ? there's not much a priori there. the machine has to discover everything on its own, each time the die is inserted.
<azonenberg> no, you misunderstand
<azonenberg> You stick the die into it, it scans the edge
<azonenberg> Determines the crystal orientation (if the first run)
<wpwrak> yes
<azonenberg> i.e. you tell it "this is a <100> wafer, align X axis to a <111> plane
<azonenberg> Then it'll center your pattern on the fragment approximately
<azonenberg> Future exposures will scan the edge, determine the approximate center (to within 100um or so, really coarse) and then locate the alignment crosses
<azonenberg> which it will use for the real fine alignment
<azonenberg> The problem is that we need to figure out how to see those marks
<azonenberg> and get data that a machine vision system can use to determine the actual position of the die and adjust the stage accordingly
<wpwrak> ah, you're trying to identify features on the die
<azonenberg> Correct
<azonenberg> Its the only way to get precise alignment
<azonenberg> below a hundred microns or so
<wpwrak> i was thinking of just (re-)identifying the die's shape
<azonenberg> Not good enough
<azonenberg> Some of my substrates are precise rectangles 1cm square
<azonenberg> you cant tell orientation from that
<azonenberg> Scroll down to the bottom
<wpwrak> you don't need a simple euclidian shape. i think it should be sufficient to identify the chaotic edges.
<azonenberg> Take it from someone who's done a good amount of machine vision
<azonenberg> Simple euclidean shapes are waaaay easier to work with
<azonenberg> And you can get better precision with them
<wpwrak> oh, fiducials are great - if you have them ;-)
<azonenberg> What do you mean?
<wpwrak> i'd add them just for fun. and check them if anything goes wrong ;-)
<azonenberg> The first mask level is positioned approximately
<azonenberg> Adds marks for all subsequent alignment steps
<azonenberg> All of the others lock onto those marks
<wpwrak> well, our problem is that it's easy to write with high precision and within a large area, but it's hard to see things with high precision within a large area
<azonenberg> Unless you're built into a microscope
<wpwrak> where "see things" means a generalized kind of vision
<azonenberg> You need a camera of some sort, which means an optical column
<azonenberg> I still think the best option is direct write with little to no alignment onto a mask blank
<azonenberg> followed by manual alignment using the microscope
<wpwrak> now, there are some things you can see easily, e.g., whether an edge obscures a narrow beam or not
<azonenberg> But there's another problem
<azonenberg> Drift
<wpwrak> drift of what ?
<azonenberg> As you pan from one side of the die to the other
<azonenberg> Alignment will gradually be lost
<azonenberg> due to imprecisions in the motors etc
<azonenberg> I suspect you'll need to realign periodically during a long run
<azonenberg> Especially if stepping multiple patterns on multiple dies on one wafer
<azonenberg> IOW, if i move 1000um right and 1000um left
<azonenberg> I won't be exactly where i started
<wpwrak> if you lose alignment within the die, you're probably screwed. well, unless you can model the loss accurately :)
<wpwrak> yeah, big wafers would be an issue. i can see that.
<azonenberg> Well, i think we can all agree that the first draft of the bluray system will have little to no visio ncapability
<wpwrak> but perhaps that's a version 2 problem ? :)
<azonenberg> and just be direct patterning on mask blanks
<azonenberg> Vision is a v3 problem
<azonenberg> v2 is where we introduce laser direct write
<azonenberg> v1 is transparencies on microscope
<wpwrak> v3 then :)
<wpwrak> i think you still need simple vision for lowering the requirements for v2. 2 x BR player plus a photo diode is still a lot more accessible than a camera-ready microscope
<kristianpaul> hacking  a camera you mean?
<wpwrak> (fiducials) looking at the patterns, i wonder if they didn't exaggerate perhaps just a little (towards the end) ;-)
<kristianpaul> i wonder what can you do inverting some lenses on a cheap canon camera, if already on a cheap webcam is not that bad for DIY scopes~
<azonenberg> Exaggerate? What do you mean
<azonenberg> When you're trying to get alignment down to a few nm it's kinda important
<wpwrak> kristianpaul: ah, i think we covered the "mass market camera" side already :) the somethat disappointing result is that they don't go far enough (sample size = 1, so take this for whatever it means :)
<azonenberg> +/- 150nm for example
<wpwrak> azonenberg: (exaggerated) http://www.cnf.cornell.edu/image/stepper alignment global with microscope overlay.jpg
<azonenberg> what about it?
<azonenberg> You need to be able to start from any angle and determine the correct position
<wpwrak> what you really need is an edge. two for 2D alignment. i even get 45 deg crosses, but stepped bars, extra-narrow edges of rectangles with rounded corners ? :)
<azonenberg> The rounded corners are an artifact of the etch probably, and the long bars are for theta alignment
<azonenberg> you need a baseline that's good enough to find the other alignnment mark across the wafer
<azonenberg> then the small stepped bars are probably a vernier scale
<wpwrak> okay, but that's more process optimizations then
<azonenberg> for alignment to below the smallest feature scale you can etch
<azonenberg> i know that the stepper the guys at work are using (not the same model as the one on this page) has a vernier
<wpwrak> hmm. still looks suspiciously like a process designed for human vision to me.
<wpwrak> perhaps for some truly advances computer vision as well. but just to find a few edges ?
<azonenberg> It's advanced vision
<azonenberg> Those things cost $40M +
<azonenberg> Lol
<azonenberg> For my process i envision using a much simpler alignment strategy
<azonenberg> Probably just a couple of crosses at 0/90/180/270 degrees
<wpwrak> i guess that's cheap as far as such things are considered ;-)
<wpwrak> i'd just let the damn machine scan the ragged edge, that's tens of thousands of data points. if you can't mine a decent position from that, what else can you do ? ;-)
<azonenberg> The thing is that not all of the dies have ragged edges
<azonenberg> 1cm square
<wpwrak> hey, if it's perfect, even better ;-) find the edge on X, find it one Y, and you're done for the day ;-)
<azonenberg> Thats the thing, the edges are accurate to maybe 10um or 25um
<wpwrak> (well, twice each - unless you eliminated rotation as well :)
<azonenberg> i need alignment to 5um
<azonenberg> and you still need optical to determine which way is up
<wpwrak> ah, so there's the great return of the ragged edge then :)
<azonenberg> This is the price to beat, you can get commercial photoplots for $34 for a 12x18 inch film sheet
<azonenberg> at 4000 DPI
<azonenberg> Feature sizes on the masks are maybe 25 or 50 um
<azonenberg> if not better
<azonenberg> And 12x18 inch film is enough for a complete mask set
<azonenberg> several times over
<azonenberg> considering that my exposure system can handle up to a 2cm square mask
<wpwrak> comparing price is tricky. since we're talking about extremely low duty cycles, almost all of the cost is in up-front investment
<azonenberg> Correct
<azonenberg> I'm just saying, unless you are making a lot it may be the case that masks are not viable to homebrew
<azonenberg> On the other hand, the cost of wafer fab is insanely high and homebrew is the only affordable way to experiment
<wpwrak> i think direct write is a lot more exciting. also because it reduces the number of steps. each step increases the probability of failure.
<azonenberg> True
<azonenberg> I think I am going to place a single order from these guys over the summer
<azonenberg> Include a bunch of test patterns etc
<azonenberg> And see how small i can actually reach
<azonenberg> IOW my current lambda is limited by printer resolution
<azonenberg> So if i have a super high res mask i can push the rest of the process to its limit
<wpwrak> sounds like a good plan. if there's something you can already get industrially made, even better
<azonenberg> I'll do the CAD myself since i want to know all of the details of the mask layout
<azonenberg> $35 plus shipping for 12x18 inches is like 100 test patterns
<azonenberg> heck, for scale
<azonenberg> $35 is the cost of a 4-inch wafer
<wpwrak> the price is certainly affordable enough. well, if you're in the US. seems they don't even do international.
<azonenberg> They do canada and mexico
<azonenberg> not sure about overseas
<wpwrak> does the panama channel make Argentina overseas ? ;)
<azonenberg> No idea, they just mentioned canada and mexico as being extra fees for shipping etc
<azonenberg> The very-high-res is $52
<azonenberg> Line widths of 12.7um for axially aligned and 25um for diagonal
<azonenberg> The normal is limited ot 38um for axially aligned and 75 for diagonal
<wpwrak> line width = resolution * 2 ?
<wpwrak> no, wait
<wpwrak> units 1mil/4 micron -> 6.35
<wpwrak> (for 4000 dpi)
<wpwrak> your factor would be dpi*6
<azonenberg> Yeah, looks like it
<azonenberg> so 6pix is the smallest line they can reliably resolve
<azonenberg> And my printer is 1200 DPI = 21um per pixel
<azonenberg> times 6 pix = 127 microns
<azonenberg> 150 is the smallest i've actually tried doing, but 200 is the smallest that gives goodr results last time i tried
<azonenberg> Actually no, that explains it
<azonenberg> The printer is 600dpi, not 1200
<wpwrak> 5 pixels/line then. still a lot. but okay, considering rounding errors, blur, and all that
<lekernel> what's a BR drive?
<lekernel> btw this paper http://www.physics.rutgers.edu/ugrad/387/388s06/film_deposition/Musgraves05.pdf uses direct write with a DLP, which solves the alignment problem nicely
<lekernel> "The patterning of this actuator was done by synchrotron x-ray lithography" lol why?
<lekernel> it's 30 micron
<[florian]> lekernel: blu-ray ?
<lekernel> ah, yeah probably
<[florian]> morning btw :)
<wpwrak> (BR drive) yes. the idea is to make a direct high-resolution writer from the head assembly of two Blu-Ray readers: one that provides laser and movement in the X direction, and other to move the target in the Y direction.
<lekernel> roh: how are the rc3 cases going?
<kristianpaul> lekernel: had you made some tests with OpenCV and mm1?
<lekernel> no
<kristianpaul> or and idea where to start with, i think i can econrage a univesiry student to take a look..
<kristianpaul> ok
<kristianpaul> or you have own plans for "Open Vision" on mm1 may be?
<lekernel> nope
<kristianpaul> :D ok