#homecmos IRC log for Monday, 2011-06-27

R0b0t1So uh00:52
R0b0t1azonenberg: How do you draw onto the miscoscope slide to backwards-project it?00:53
R0b0t1(if that made any sense...00:54
azonenbergR0b0t1: I laser-print on transparency film00:55
R0b0t1I assume you can get more resolution that way than you could reasonably etch at this point?00:56
R0b0t1I think you said your feature limit was something like 250┬Ám, could the projection be finer than this with printing?00:58
azonenbergR0b0t1: My printing has a 200 um design rule00:59
R0b0t1Is that the printer or your ability to etch it?00:59
azonenbergThe printer00:59
azonenberg600 DPI times about four or five pixels minimum00:59
azonenbergThen I reduce that optically01:00
azonenbergAt 40x reduction, that gives me 5um line width01:00
azonenbergAnd i've fabricated 5um patterns in photoresist before01:00
azonenbergHowever, my illumination is uneven in the 5um process so i cannot get 5um across the entire FOV01:00
azonenbergAlso, none of my fab processes are able to (yet) transfer a PR pattern that small onto something else01:01
azonenbergMy liftoff is close to debugged but i havent tried anything smaller than 20um with it01:01
R0b0t11) What causes the illumination to be uneven, and how would you fix it? 2) What are the typical trace widths in modern ICs?01:01
azonenbergAnd i need to get thin low-stress films to prevent cracking during the KOH etch01:01
azonenbergThe illumination is uneven because i'm using a pocket flashlight as my exposure source :P01:01
R0b0t1.... XD01:02
azonenbergI plan to fix it by buying a fiber-optic microscope illuminator and using that instead01:02
azonenbergwhen i have $200ish to spare, thats a solved problem01:02
azonenbergTypical trace widths in modern ICs are very roughly double the technology generation size01:02
azonenbergdouble to triple*01:02
azonenbergSo for a 350nm chip a 1um trace size is typical on top metal01:02
R0b0t1When you say "350nm chip" what exactly does that mean?01:03
azonenbergTechnically its the half-pitch of a SRAM cell made on that technology generation01:04
azonenbergWhich means very little01:04
R0b0t1Oh is it some weird industrial standard?01:04
azonenbergPretty much01:04
R0b0t1Do traces not go below that or...?01:05
azonenbergTraces are usually several times that01:05
azonenbergThe width of a single transistor gate can be less01:05
R0b0t1Oh, I heard something about light... and one micron...01:05
azonenbergfor example in microchip's 350nm process i've measured traces on top metal to be around a micron wide01:05
azonenbergI prefer to use more concrete numbers, namely the minimum size design rules for line features, point features, and so on01:05
azonenbergWhich accurately describes what i can actually fabricate01:06
azonenbergSo I would describe a process as being a 5um design rule with a 1.058um lambda01:06
azonenbergmeaning that all dimensions are multiples of 1.058 um (1/600 inch, the printer dot size, divided by a 40x objective)01:07
azonenbergAnd that no single structure can be smaller than 5um01:07
R0b0t1Ok, thanks for the answers01:09
R0b0t1I heard some laser engravers can etch glass, so I wondered if they'd be able to etch channels in microscope slides01:10
R0b0t1and then you use capillary action to fill them with ink01:10
R0b0t1but w/e01:11
azonenbergIf anything, i'd plate the slide in metal01:11
azonenbergand use the engraver on that01:11
R0b0t1I realized it would be easier to just coat it01:12
R0b0t1then remove it01:12
R0b0t1Which kinda could be done with most things, I guess01:12
R0b0t1Or well idk01:12
azonenbergMetal would work better since its more opaque01:12
azonenbergIn any case mask making is not the current bottleneck01:13
R0b0t1Yeah that's what I realized01:13
R0b0t1so everything I was thinking of was a bit too preemptive :p01:13
azonenbergThe problem is little details of trying to get the hardmask to work properly01:14
azonenbergWhich i hope to solve fairly soon01:14
R0b0t1"work properly"?01:14
azonenbergHF wet etching through PR doesnt work when your photoresist is permeable to F- ions01:15
azonenbergAnd liftoff is a little tricky if your photoresist is slightly soluble in the hardmask's solvent01:15
azonenbergI'm experimenting and hope to have a workable process soon01:15
azonenbergI think if i hardbake the PR a little longer it will work well with liftoff01:15
azonenbergi just need to densify it to the point that it's just a little less soluble01:16
berndjazonenberg, is the hardmask solution [close to] saturated?17:35
azonenbergberndj: I'm actually not sure21:48
azonenbergit's 16%21:49
azonenbergi havent looked up solubiltiy figures21:49
azonenbergberndj: I am actually thinking of diluting it though21:59
azonenbergBecause my current coating method gives poor repeatability due to lack of precise spin speed control22:00
berndjmy lone of thinking was, maybe you could concentrate the solution further, hoping to reduce how quickly the PR dissolves?22:00
azonenbergNo, it will leave way too thick a layer22:00
azonenbergIf i spin really fast i get poor repeatability but low dissolution22:00
berndjbut that's by no means a guaranteed result.  and repeatability seems like a more important goal22:00
azonenbergSo my thinking is to dilute, quickly coat one drop22:00
azonenbergthen use color to determine if another drop is needed22:00
azonenbergAll i need is a continuous film22:01
azonenbergI'm using liftoff to pattern so i dont care about exact thickness since i'm not etching it22:01
berndjor if you add some other solute that doesn't influence viscocity so much?22:01
azonenbergNot sure if i could do that without affecting the generated film22:01
berndji'm still surprised that spin coating gives a roughly uniform-thickness coat22:01
azonenbergI think the best bet is just to not leave it on the film long22:01
azonenbergon the resist*22:01
azonenbergI'll be testing that tonight22:02
azonenbergthen later in the week taking it up to rpi's materials research center22:02
azonenberga guy there is willing to give me a few hours on a SEM22:02
azonenbergSo i can do low-angle cross sections etc of my dies to date22:02
azonenbergand measure sharpness of my probe tips22:02
azonenbergMy work between now and then will focus on getting a wide variety of samples to image22:02
azonenbergfor example, resist on silicon that's been developed but not etched or anything22:03
azonenbergso i can study the adhesion at the interface22:03
berndjheh, reminds me of that old joke. stop me if you know it; i'll adapt to your probe tip22:03
azonenberghardmask that's been coated but not exposed to acetone22:03
azonenberghardmask that's been exposed to acetone before baking22:03
azonenberghardmask that's been lifted off both before and after baking22:04
azonenbergBasically i want dies at every stage of processing that might be causing the problem22:05
berndjworld's best inventor invents the sharpest needle the world has ever seen, shows it to the world's tech companies. the swiss look and say, 'no need for this'. germans say, 'impressive, ja? but we can't think of anything to do with it.' japanese say, 'hmm, interesting, we don't know what to do with it, but we drilled a hole down the middle for you'22:05
berndjdo you nead to cleave the dies to get a cross section, or do you do that with SEM magic?22:05
azonenbergAnd actually, if i cleave them the film will likely chip off the edge (though a micron or so away it'll be fine22:09
azonenbergMy plan is to place them at a steep angle (maybe 70 degrees)22:09
azonenbergand then image from that angle22:10
azonenbergI can also do a true xc with a 90 deg shot22:10
azonenbergBut i think high angle is more likely to be suitable than full perpendicular22:10
azonenbergIn any case i'll fool around once i hop on the scope22:12
azonenbergI'll need a few minutes to get familiar with it anyway as i havent used this instrument before22:12
azonenbergSEMs have more controls than optical scopes so learning which one does what can be a little tricky the first time you use one you havent used before22:12
azonenbergThe cleanroom on campus has a zeiss supra 55 which i've used before, but its expensive to use22:13
berndjwhy "expensive" - actual running costs or rather opportunity costs?22:13
azonenbergNot sure, all i know is that's what they charge people22:13
azonenbergThey charge a definite profit if you arent under an RPI grant22:14
azonenbergif you're a local researcher they charge your grant at a reduced rate22:14
azonenbergThis one is much cheaper to use (about $60 vs $190 an hour) than the one in the cleanroom (because it isnt in a cleanroom) and the guy is even willing to let me use it for a couple of hours at no charge22:14
azonenbergThough if more than a couple of hours is required i'll need to shell out some $$22:15
berndjprobably under "if we need it we'll kick you off" terms?22:15
azonenbergMost likely i'd be bottom of the priority queue, but thats to be expected22:15
berndji think i only ever saw one SEM22:15
R0b0t1They charge you to use the equipment...?22:15
berndjhehe, almost like an observatory's director's discretionary time22:16
azonenbergberndj: Pretty much, the guy i'm talking to is the director of the microscopy lab for the mat sci department22:16
azonenbergR0b0t1: Most resources are available at no charge but a few things like the SEMs are expensive to run (spare parts and such)22:16
azonenbergEspecially if they're in a high-maintenance lab like a cleanroom22:16
berndjwould you have to coat your chip?  or wouldn't there be any PR / hard mask solvent left to screw up the vacuum22:17
azonenbergberndj: Everything will be baked to drive off solvents etc before i bring it up there22:17
azonenbergI'll be coating them but not for that reason22:17
azonenbergIt's difficult to image non-conductive materials (like the hardmask) with a SEM because the electron beam builds up on the surface and gives it a negative charge22:18
azonenbergWhich repels the inbound electrons and eventually gives you a blank image22:18
berndjwhen i was little i used to think the bugs etc. in SEM pics were still alive22:18
azonenbergThe solution is to sputter it in a few nm of gold or platinum or similar22:18
azonenbergand ground it22:18
R0b0t1You mean they didn't just snap a still? D:22:18
azonenbergR0b0t1: Electron microscopes only work in vacuum22:18
azonenbergberndj: Generally if you're doing school-sponsored research they charge your research grant for the lab time22:19
berndji imagine many insects could survive a vacuum though22:19
azonenbergtypically at-cost and sometimes even a bit lower22:19
berndjyeah. well, *somebody* has to pay22:19
azonenbergBut if you're an outside user they charge the cost of time plus profit22:19
azonenbergSo i'm quite lucky to have found someone willing to give me even a couple of hours on one22:19
azonenbergAnd at a reasonable rate should i decide i need more time22:19
berndji suppose in a case like yours, sometimes you can pay in the form of having something really interesting to tell whoever owns the gear22:20
azonenbergberndj: Thats basically why he's letting me on lol22:20
azonenbergbut it's not enough to give me 50 hours of free sem time :P22:20
azonenbergMaybe 522:20
azonenbergberndj: Also, some field emission SEMs can run at decently low vacuum (lower resolution in that mode though)22:20
azonenbergAnd people have successfully imaged live insects on them22:21
azonenbergI dont remember the exact details but it's been done22:21
--- Tue Jun 28 201100:00

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