Device Fabrication File, 2000 - Author unknown

1) Cleaving: Wafer is placed on a few napkins over a glass slide, to provide a flat surface with a bit of give. The back of the wafer is scored lightly with a diamond scribe guided with a razor-blade, then the wafer is turned over and the razor blade is used to break the wafer along the score. Care is needed not to press too hard and crack the wafer.2) Cleaning: This must be done between almost every step in the processing, followed by visual inspection under a microscope to ensure that everything is as it should be. Ethan suggests “At the first stages of processing it is also wise to test for organic residues. If organic residue has not been removed (see the following cleaning instructions) metal evaporation will not stick. The test is easy, water will wet a clean oxide layer but will form droplets on the surface if organic residue exists. Plasma cleaning usually ensures that all organic residue is removed.”

-Before every exposure cycle, rinse wafer with acetone, then follow immediately with 2-propanol so that the acetone residue does not form on the wafer, and then blow dry. If there is hardened resist on the wafer, several minutes of soaking in acetone may be required. Note that care must be taken during the blow-drying that the wafer is not damaged; drying the wafer while it sits in a small Petri dish is thus a very good idea. Finally, resist residues must be ashed in a plasma asher (eg barrel plasma reactor), typically at an O pressure of 470 mTorr and 50W power for 2-3 minutes.

-After developing any resist layer, rinse with water before blow-drying, not acetone or 2-propanol, and then ash the resist residues in the oxygen plasma for 30-60s. The etch rate for photoresist under these conditions should be about 100A per minute, but I haven’t measured it yet. For very thin resist layers, 30s is usually sufficient, but where spatial resolution is less important than clean surfaces (eg, contact pads), 60s is a good idea.

3) Spinning Resist: The different resists and spinning conditions used are described below. - Shipley 1400-31: photoresist used in optical steps (mesa etch, contacts, optical gates). For etching large-scale structures ( 10 m), I spin at 6000 rpm for 30 s, followed by 10 minutes of baking in the 90ºC oven for 10 minutes. For smaller-scale structures, I have had better success with a speed of 7500 rpm. For the contacts, the baking is followed by a 5-minute immersion in chlorobenzene and then 5 more minutes of baking at 90ºC, in order to harden the resist and facilitate liftoff.

- 495K PMMA, 2% in chlorobenzene is what I use for e-beam lithography when I want to etch the wafer. This is spun at 6000 rpm for 30 s, then baked at 150ºC for 60 minutes.

- MMA, 8% in ??? and 950K PMMA: I use these in bilayer resists for gate deposition. The MMA is spun on first, at 000 rpm for 0 s, the resist is baked for 60 min. at 120ºC, the PMMA is spun on at 000 rpm for 0 s, and the whole thing is baked again for another 60 minutes.

Note that resist becomes thicker as it ages and so doesn’t spin as uniformly. Also, for very small wafers, the resist thickness tends to be quite uneven, being thin at the center and very thick at the edges. Slivers of wafer that hold 4-5 devices seem to be optimal in this respect. Also, because the resist hardens if left on the wafer, it is best to spin the resist for lift-off steps on the same day as the evaporation takes place.

4) Optical Lithography: I use the Quintel contact printer, which holds the wafer in vacuum contact against the optical mask. It is supposedly capable of writing features on the order of a micron, but in practice it is difficult to get 2 m features to develop properly. For best focus, the vacuum contact mode should be used, which is what I do for my smallest features. For larger features, the pressure contact mode (“cont. calib.”) mode is safer. Always ensure that the separation gauge is at maximum before aligning the mask, as the photoresist may smear otherwise! Exposures of 2.5 sec seem to work the best for the smallest features (where there is competition between exposing all the way through the resist and blurring the small features), while 3.5 - 4.0 sec works well for larger features. I have tried using 6.0 seconds for contact pads, to ensure that all the resist was exposed and prevent problems with the contacts sticking. It is not clear whether this is necessary, and it certainly degrades the resolution of the pattern.For best results, do not put the wafer into contact with the mask until everything is ready for exposure. This minimizes the risk of scratching the resist layer. Also, before placing the mask in the quintel, make sure to blow it off with compressed air, to remove any dust particles.

5) Etching: For GaAs/AlGaAs wafer, a mixture of 40 parts water to1 part H2O2 and 1 part H3SO4 etches at about 4 nm/sec when newly prepared. The etch rate is faster for an older solution, and also depends somewhat on the wafer. It is best to test the rate with a dummy piece of wafer; monitoring the rate under progress using the profilometer is not optimal, as the needle of the profilometer can actually scratch the resist layer. Getting good etches is quite hard, as usually the surfaces do not etch smoothly. It is important always to ash the resist residues after developing, otherwise a poor etch is virtually guaranteed, with pitted surfaces and sidewalls.

6) Contacts: These have given me problems, first with adhesion and then with proper alloying. What I have done to try and make the results consistent is expose longer and plasma ash the resist residues longer, to make sure the surface for deposition is clean. This improved results somewhat, though not fully.

Right before evaporation, the wafer should be etched with 5% HF for 5s, to createn a clean surface for deposition. Place the wafer on the holding stage with a very small dab of vacuum grease, making sure not to get any on the surface of the wafer!! (Note that the metal tweezers are easier for manipulating the smallest wafers than the Teflon tweezers.)

Pumping down overnight has given the best results so far, although a 3-hr pump-down has worked, too. The evaporation profile that I inherited from Geoff and Keith is: 30 nm Ni, 30 nm Ge, 130 nm gold. Note that Ge and Ni require much less current to evaporate than does gold, but they have a much greater tendency to explode. They need to be heated up much more slowly!!

After liftoff, annealing is done in the rapid thermal annealer, using a flow of 30mm H2/N2, and a temperature of 425C, for 60 s. The length of time seems not to make a big difference in either appearance or the resistance of the contacts—certainly up to 120 sec is still fine. Make sure to flush the annealing chamber thoroughly with the forming gas before annealing, however, to ensure that there is no oxygen, as GaAs will oxidize above 300 C. Also check all contacts after annealing at the I-V probe station, to see that they all work properly; if not, then subsequent lithography steps may need to be altered to accomodate the defective contacts.

7) E-beam Lithography: (by Ethan) There is a great chapter in “Handbook of …” describing the importance of different parameters when exposing and developing e-beam resist. It’s very helpful to understand small angle scattering, the relative importance of primary vs secondary electrons, the length scale of the proximity effect and how the developing process works. The SEM image showing a cross section of bilayer resist explains why bilayers are used. The electron scattering simulations also give good insight.

Patterning with Single Layer Resist

-rinse chip with acetone, then IPA the blow dry on technicloth. -spin PMMA (from my newest dropper bottle in microlab - the resist was fresh from the stock bottle). 30sec spin. 3500rpm. -small and large features were exposed at 330uA/cm^2 -developed for 60 sec in 1:3 developer (important that developing is long enough for the undercut to be made). Rinse with IPA and blow dry on technicloth. - Deposit Cr 5nm, Au 55nm. High Vacuum gives best results. -chip can sit in gel box overnight. -soak in acetone for 2 hours while riding bike in the hills. -squirt with hyperdermic syringe of acetone. -see gold release. cheak the small features using the probe station microscope. The chip should be STILL SUBMERGED in acetone when release is checked. (Adam recommends keeping the chip separate from the lifted off gold).