From OpenWetWare
Jump to navigationJump to search

Low-stress Nitride & Silicon Microfab (written by Emory Chan)

Use Sink 3 to etch Si along the (111) crystallographic planes. The reason it etches along the (111) planes is that the atoms in the 111 are hexagonally close packed (i.e. very dense), and if you were to cleave the Si along this plane, the surface atoms would only have 1 dangling bond. Thus, this surface has the lowest surface energy and is least reactive.

KOH etch rates:

for 1:2 vol/vol KOH:H2O:

80C: ~ 1 um/min

90C: ~ 2 um/min -- much more undercutting, but still useable

If your mask features are not aligned to the 110 direction (the 111 plane projected on to the surface of the wafer), then your features will be undercut and may not be at the right angle or of the right shape. If you need very good angular or spatial resolution (e.g. for thin features), you need to align your mask pattern to the 110 direction of the Si lattice. Since the wafer flats are not really accurate reflections of the crystal orientation (~3 degree error), the only way to find the actual orientation is to pre-etch alignment marks into the Si. While you could do this via lithography, that is time consuming. Here is the ghetto method I found on the web:

1. Using a diamond scribe, make a scratch in the nitride film that is somewhat diagonal to the crystal plane. Make the scribe on the right or left hand edge of the wafer, but centered vertically.

2. In the Sink 3 KOH bath, etch all the way through the wafer (use 90C to speed this up) until a well-defined rectangular window (or shape with straight edges) appears on the other side.

3. On quintel, use the backside alignment light pipes to illuminate the etched alignment marks so that the straight edges align to some straight feature of your mask. Quintel actually works better for this purpose than KSaligner due to ksaligner's bad contrast and poor backside illumination.

4. Make sure to strip your photoresist after etching the nitride in Lam2. Failure to do so will result in isotropic etching in KOH rather than anisotropic etching.

5. After etching thin membranes, be very careful when cleaning and drying wafers with these membranes. Let the wafers cool with their bottom edges touching a partially filled QDR tank for about 1 minute. Make sure the wafers are aligned parallel to the water jets. You can use the QDR cycles regularly after that. Blow dry at your own risk. definitely do not blow directly on the membranes, and do not use the full air velocity. After QDR, I normally let the wafer sit on clean wipes with the bottoms of the wafers touching the wipes so as to wick off the water. After 5-10 minutes, I use careful blow drying to dry the wafers partially, then complete drying by putting wafers in VWR-oven or in a metal cassette sitting on top of the 120C hot plate.

6. Wafers etched in KOH will have a lot of particles on them, especially in the etched regions. These particles are iron hydoxy oxide (FeOOH) that nucleate from iron contaminants in the KOH powder (and possibly from the bath container walls). The best way to get rid of these is to clean the wafers in piranha for 20 minutes. HCl will also work, but is slower.

KSBA6 -- Karl Suss Bond Aligner

Used for aligning two wafers for bonding. In general, it's a pain because you have to take apart ksalilgner every time you want to use it. Therefore, you should make reservations under ksaligner instead of ksba6.

The anodic bonding program assumes that the glass wafer will be on top and the silicon wafer will be on bottom. You align from the top, through the glass wafer.

There were times when I wanted to align the wafers with the glass wafer on the bottom and use the backside alignment mode. There's no obvious program to do this, but here is how I figured out how:

1. Use the 6" anodic bonding fixture (stored on ksbonder)

2. Use the 4" Si fusion chuck (not the vacuum one)

3. Use Program [2] (Si fusion, Load:slide, Unload:fixture)

  • Basically, ksba6 doesn't know the difference between the 2 fixtures, but we want the anodic bonding fixture so that we can clamp the aligned wafers rather than dumping them on the chuck again.

4. Load top (Si) wafer and capture an image of the back side.

5. Load bottom (glass) wafer and align to image of top wafer

6. When aligned, press "Clamp"

7. Remove the fixture and use a binder clip to clamp the wafers in place.

8. Remove the spacer flags carefully.

9. Undo the fixture clamps (but leave the binder clip in place)

10. You can flip the wafer stack over if you are doing anodic bonding (b/c it needs the opposite orientation).

11. If you are doing Si-Si bonding where you can't see the bond quality you need to use the IR microscope...

IR Microscope / Reichert

The IR microscope is located above Reichert (next to Nanospec, across from ASIQ). It is stored and used on the shelf above reichert, and it looks like a cheap closed-caption camera facing down over a ring stand. You use the Reichert computer to take images. The username and password for this PC are posted on the underside of the keyboard. I think they are login: riechert, password: reichert. Here is how you would use the IR microscope:

1. Turn on the white bucket lamp located underneath the camera. The switch is a rotary switch located on the power cord near the lamp end. The lamp is just a simple light bulb. If it burns out, you can ask the Microlab office for a new one, or, if it's the weekend and you're desperate, you can just go to ACE hardware. Flood lamps work better because they have a flatter intensity profile, but any lamp will do. You might want to check the wattage, though.

2. On the computer's Windows desktop, open XCAP for windows

3. After agreeing to some conditions (which sell your firstborn children to Mr. XCAP), you will see a window that has a big square for the video/snapshots, and a diagram of the back of a PCI video input card on the left side of the screen. In this EPIX® PIXCI® SV: Capture & Adjust window, click on the UPPER middle circle so that it is highlighted in red. This is the video input port for the IR Camera (the lower circle is for Reichert). Then select the button for Live. The window EPIX® PIXCI®: View #1 should now be showing live video of the IR camera.

4. If you don't have anything between the lamp and the camera, you'll probably just see a white screen because the intensity is too high (the camera is sensitive to visible light, too). Place your sample on the ring stand and see if you can see your features. You shouldn't need to adjust the focus, as the ring stand has been adjusted to the focal length of the camera.

5. If the intensity is still too high (white screen), there are usually one or two plain silicon wafers sitting next to the lamp. Put these directly on top of the lamp's top frame. These will attenuate the visible and IR light so that you can get better contrast/less saturation with the camera. If it is too dark, move the wafers so that they only partically cover the top of the lamp.

6. Sometimes, you'll see some weird white cloudy spots in your picture which don't really matter but look bad if you're trying to use a picture for a talk/publication. These white spots are glare from metal nearby (like the post of the adjustable camera mount). Usually a strategically placed technicloth can get rid of this glare.

7. When you are ready to take a picture, click the Unlive radio button to freeze the image. If you want to refresh the image, press Snap in Unlive mode.

8. Save the image with File -- Image Save. Create a directory for yourself and delete pictures when you have transferred them to external disk or server.

9. More info can be found in the Reichert chaper of the Microlab manuals.

You can confirm that it is live by moving the stage slightly.

Tensiometer (& Kruss contact angle measurements)

Tensiometer measures surface tension and dynamic contact angle. Dynamic contact angle gives more reproducible results than static contact angle (as measured by Kruss).

For surface tension, you should use the Du Nouy ring pull method. It's very very very important to clean the platinum ring before every measurement. I used one of those flame lighters that you normally use to light campfires, candles, etc. Clean the right (ring?) in ethanol and flame all parts until they are red hot. The force curves will tend to change over time, so you have to be consistent about which ones you decide to take (first 10, etc).

Due to a screwup in the enabling system, you enable tensiometer under "kruss." That means that getting qualified for tensiometer doesn't do you any good if you're not qualified for kruss. In fact, there are no superusers for tensiometer, and very few actual users to train you. So get qualified on kruss and teach yourself. The manual is pretty handy, as are some people in the Majda group.

In reality, the ring pull method is not great for low surface tension measurements. A better method is to use the hanging drop method, which supposedly Kruss can do.

Lam2 tips

1. Use programs "o" or "p". P doesn't have the over etch.

2. To prevent photoresist burning:


30 min in vwroven, or use uvbake (for Si only).

  • use 700 W instead of 850 W power. etch rate of oxide ~ 500 nm/min. nitride is a little less, but close.
  • don't let the temperature get over 14C. don't etch for much over 1 minute at a time. Wait for 1 minute in between etch steps.
  • use g-line resist + g-line uvbake program for best results.

STS / SPR220 tips

  • HF programs:

VEE6 (deep etch, preferred, ~3 um/min etch rate)

AARON6A (deep etch, OK)

HEXA250 (lots of undercutting, not so great)

LIPPMAN1B (smooth edges, slow etch)

  • Low frequency programs:

Use to prevent "footing" (widening of channels when you hit the bottom oxide or buried SOI layer). Basically, the platen power will cycle off and on, allowing the charged ions that build up near the insulating layer to dissipate and preventing distortion of electric fields that screw up anisotropic etching.

You CANNOT use HF programs for LF or vice versa. They use different power supplies and STS will yell at you.

  • Use 20 ms period, 25% duty cycle. The VEE9 LF program works very well, with ~ 2 um/min etch rate for big channels, ~1.67 um/min for 40 um channels. Other programs:

ROLF1B (not so great)

Lipman1C (~1.4 um/min)

Anita1B (lots of grass?)

  • For best results, run the COND3 recipe on a plain Si wafer before etching (35 minutes)
  • Use a 1 um oxide (tystar12 LTO) hardmask for deep etches and nice lines.
  • Etch for 10 s in Lam 2 (program P) to remove native oxide before STS. This increases uniformity and reduces grass.
  • For the best through etching, I use SPR-220 photoresist (10 um) and Cool Grease for handle wafer bonding. Here are my parameters.

o SPR220-7

1. Spin SPR220 on Spinner1 (much better than SVG coat)

10s ramp -- 500 RPM -- hold 10s -- 10s ramp -- 1.8 KRPM -- hold 30s

2. Soft bake 5 min, 115C hot plate, NO FOIL!

3. Expose 280 mJ/cm on ksaligner, soft contact mode

4. NO post-expose bake! It just induces cracking,etc.

5. Wait 30 min before developing for 1.5 min in LDD26W.

6. Hard Bake 15 min on 95C hot plate

7. Etch 1 um oxide hard mask on Lam2 (2 min, 700W)

o Cool Grease handle wafer bonding

1. Put a piece of foil on a 65C hot plate.

2. Place handle wafer (1 um oxide) on hot plate.

3. Spread AI Technologies Cool Grease on handle wafer with a glass slide. (cover completely)

4. Smooch your photoresist-covered wafer on dummy wafer while on hotplate. Make sure to align the flats. Use the big aluminum cooling block to smooch them evenly (use foil on top to protect wafers).

5. Use poly gloves for everything because the cool grease gets everywhere. 6. Leave block on wafer stack on hotplate for 5 min.

7. Clean off cool grease from edges/bottom of wafer stack, since it could screw up Lam and STS.

8. Bake in 60C OVEN for 15 min.


or instead, discuss this protocol.