This page is intended as a resource for entering current operation parameters for tools at the CNF, such as growth/etch rates, focus, exposure times etc. This will save future researchers from having to start at ground zero when encountering a new tool.

Tools used by our group at the CNF:

Photolithography: Pattern generator Spinners 5x g-line stepper Autostep 10x i-line stepper HTG Contact Aligner ABM Contact Aligner

Etchers: Oxford 80 1/2 Unaxis 770

Furnace Processing:

Deposition: CVC 4500 even/odd (odd works much faster and more reliably)

Misc. Critical Point Dryer

Always use the odd-hour tool if you can, it pumps down twice as fast as the even-hour one and also maintains a much more stable deposition rate. Unfortunately, its schedule also fills up much quicker because many people know about it, so make sure you plan ahead. When you schedule a time, be sure to note what metals you plan to deposit in the comments section or the reservation will be deleted. We have an iron source of our own in the group drawers, a crucible labeled as “Markus's Fe Source.” Follow standard instructions for the e-beam evaporator, but change the Tooling Factors (both of them) to five times what they normally read. If the preset is 80.0 (as it is on the odd-hour tool), set it to 400.0. Note that this is not entirely condoned by the CNF staff, but it doesn't damage anything as long as you remember to reset it to the standard when you finish. The tooling factor is an additional multiplicative factor on the crystal mass readout; so if you want to deposit 1 angstrom of material, and set your tooling factor at a multiple of five, then you would wait for the readout to say that you've deposited five angstroms of material. This is essentially a way of increasing your deposition rate precision. For aligned arrays we've been using catalyst layers of 1-3 angstroms. Depositing such a layer should take about an hour on the good evaporator, from start to finish. When you are done, be sure to reset the tooling factor. In the CAC log-out, the material deposited is “Other” on the first screen, then say “Fe (own source)” when it prompts for details.

For a 500um single-side-polished substrate: Vapor prime with the YES oven; the bake takes ~35 minutes. Apply 1813 photoresist from the McEuen Group bottle, spinning for 30s at 4000 rpm, with a 1000 rpm ramp. The 5x g-line stepper should be set to “transparent” (switch below the computer). We use a 0.7s exposure time, with an offset of +110. Use recipe 6 on the developer (300 MIF for 90s).

Cover NTs with 100nm of gold. Spin a layer of PVA (I use Emulsitone Laser Scribing Solution 1146, which is PVA in water and can be stored at room temperature) at 3000 RPM, 1000 RMP ramp, for 30s. Bake the chips for 30 minutes in a 70 C oven. Using tweezers, peel the NT/Au/PVA PVA film off of the donor substrate and place it on the receiving substrate. Use a gloved finger to smooth the film down and eliminate any bubbles. If you don't get decent adhesion to the surface - i.e. if the film would rather curl up or peel off than stick to the receiving substrate - then your PVA layer is probably too thick. Carefully place the chip in a warm DI water bath for one hour. Do not flow water over the chip or otherwise disturb it, as that can cause the film to release from the surface. Let the chip air-dry. You will notice that the NT/Au film will be mostly free of PVA, and will look as though it was evaporated onto your new surface rather than just being placed on top. Run the dry chip through an oxygen clean for 10-30 minutes (150 mtorr, 20 sccm, 150W). Place the clean chip in gold etchant (TFA) for two minutes. Rinse with water.

Cover NTs with 100nm of gold. Oxygen clean your receiving substrate. Wearing gloves, cover nanotubes with two layers of PVA tape from 3M, and smooth down. Peel NT/Au/PVA tape off of the donor substrate and place it on the receiving substrate in the desired orientation. Use a gloved finger to smooth the film down and eliminate any bubbles. Drip DI water onto the top of the tape and let it sit for a few minutes until the surface looks ripply, then carefully place the chip in a DI water bath for thirty minutes. Move the chip to a new bath; at this point the PVA tape should have come off of the surface, leaving the gold behind, and will be floating around in the water. You will notice that the NT/Au film will be mostly free of PVA, but may still be covered with a layer of goop. A few more water baths should clear that up, and you can also put it in a mild HCl bath if you're worried about cleanliness. Then dry the chip; it should look as though the gold was evaporated onto your new surface rather than just being placed on top. Run the dry chip through an oxygen clean for 10-30 minutes (150 mtorr, 20 sccm, 150W). Place the clean chip in gold etchant (TFA) for two minutes. Rinse with water.

If you have any questions or need more detail my email is davis252 -at- gmail.
Method A:
- Clean wafer and blow dry it immediately before spinning to avoid comets.
- Spin Brewer XHRiC-16 Anti-reflective coating (ARC): Static dispense, spin 3000/5000/30 (spin rpm/ramp/time). Bake at 170 C for 60s on the orange hot plate.
- Spin OiR 620-7i photoresist: Static dispense, spin 4000/5000/30. Bake at 90 C for 60s.
- Expose on Autostep. Dose usually runs between 0.100 ms and 0.170 ms. Smaller doses will create shorter gaps, but too small will not get a gap at all. Choose an exposure time 0.020 ms above the minimum dose for gap formation. (Don't trust a dose test until after image reversal). The optical microscope in the spinner room has a ruler in the eyepiece that measures 1um/tick with objective 4 and 667nm/tick on obj 5. Usually you can tell how small the gaps are optically; if it's 2/3 of a tick on obj 5 then it's about 400 nm. As you get smaller diffraction will make the gap look like a dot in the middle of the two electrodes. It's still good, don't worry. If you can see the resist cleared all the way through the gap it also could still be there, but you would need SEM or AFM to tell for sure.
- Image reversal oven.
- Flood expose in the little JBA deep UV cure lamp in the middle of the contact aligner room for 60s. Let the machine warm up for a few minutes before you use it.
- Develop in MF321 for 60s. Auto develop is fine. Note that this ARC is insoluble in developer, acetone, and 1165; it needs to be ashed before evaporation.
- Ash the ARC in Oxford 80. I used the following modification to the oxygen clean recipe: 50sccm O2, 40 mT pressure, 125 W power, 1.0-1.1 minutes. The reason for the modification is to improve selectivity of ashing the ARC to the photoresist. Note also that the ARC migrates into the PR during the etch so if you lose too much resist on top the PR will be insoluble during liftoff. I usually etched for 1.0-1.1 minutes, which got rid of all the ARC, but less might work too if you can't afford to chew up as much resist. Also Vince has a recipe with 20 mT pressure and 100 W power, but to do that you will have to check the box telling the tool to strike the plasma at a higher pressure or it won't ignite. I did this all in Oxford 81.
- Evaporate
- Liftoff in 1165 or acetone or both.. You should see liftoff starting within 10 minutes of putting it in, but it will take a good day or longer to finish. Sonication is often necessary. If small features don't liftoff you can finish the liftoff in cold nanostrip for 60s or so. If no liftoff happens at all then there wasn't a layer of soluble resist under the pattern to liftoff. You can try the whole liftoff in nanostrip and that works OK, but if everything went right the liftoff should work without much trouble.
- Remove the rest of the ARC (any oxidizing method works - you can ash it or use nanostrip/piranha/RCA clean).

This worked to make gaps ~200 nm across an entire wafer.

Method B (most of the comments above apply to this as well):
- Clean and blow dry immediately before spinning.
- Spin LOR-5A in spinner bowl 2. Dispense at 100/10/10 then spin 5000/10000/30. Bake at 190 C for 3 min on the orange hot plate. I kept the same recipe for consistency but if I did it again I'd increase the dispense to 500/75/10 for a little better uniformity.
- Spin OiR 620-7i in spinner 1: Static dispense, 4000/5000/30. Bake at 90 C for 60s.
- Expose on Autostep.
- Image reversal oven.
- Flood expose.
- Develop in MF321 for 55-65 s. Auto develop might work, but sometimes it needs just a little more or less. In an optical microscope, you should be able to focus on the surface of the resist and measure it, then focus on the LOR layer and measure it to ensure there is an undercut.
- Descum in Oxford 80. I used the 40mT / 125 W recipe for 20s in Oxford 82, but regular O2 recipe is probably also OK.
- Evaporate
- Liftoff
- Brief descum, O2 plasma, cold nanostrip or piranha.

This will probably not get features less than 400 nm but it did a good job at getting 400 nm features. There will likely be a swing curve so the dose on one day will not necessarily be the same as on other days, but the liftoff is much faster (roughly an hour at most, if you agitate it every 15 minutes and apply heat).

—–Notes in general——
- Never use the Glen, even for descum. Good restaurants are always busy, and that's why the Oxford etchers are harder to book than the Glen. Also the Glen sometimes leaves spots shaped like the shelf trays and it has ruined several people's wafers (including my own). Avoid it.
- Another possibility (that I have not tried) to avoid problems with liftoff is to put the metal layer down first then use a directional dry etch afterward.
- Develop time is key to this entire process. Image reversal will not produce a reentrant resist profile without overdeveloping. LOR also needs time to undercut after the resist clears. However, if you use ARC, overdeveloping chews up the top layer of soluble resist, so don't overdevelop too long. A good rule of thumb is (resist clear time)*1.3.

It is usually a good idea to start from an old mask as your template, since they contain the GDS alignment keys already.

Our furnace is 23mm in diameter, so your final chips should not be larger than 20mm across.

If you plan to wire bond to your sample, contact pads should never be smaller than 50x50um.

Be sure to check the flash count on your completed file. The PG runs between 3,000 and 5,000 flashes per hour, so you can get an estimate of how long the mask will take to write.

To print a mask on the PG3600, you must convert your original L-Edit file to a GDSII file, then to a pattern generator file.

1) In L-Edit, go to Export in the File menu, and export as a GDSII file. Fracture polygons to 8 vertices, and export Cell0 and its hierarchy. Be sure to note what layer numbers correspond to what content on your mask.

2) Open the program WS_FTP. Set upload option to binary.

3) Log in with your CNF CAC account, and drag your GDSII file into the directory given.

4) Open the program Putty. Telnet to nnfvax.cnf.cornell.edu, and log in using your CNF account.

5) Type GDSPG. Press return to use the current directory. Give it the top-level cell (Cell0).

If you want to use image reversal, you will need to export the frame layer, and convert the layer to be image-reversed, using the frame layer simply to define its boundaries when asked (i.e. do not convert the frame layer as a layer of its own).