Lift-off of SiO2 continued

I posted previously about an issue we have with needing to strip the native oxide from the sacrificial silicon layer prior to performing the through-pore etch with XeF2 that enables separation of the oxide membrane from the wafer substrate.  The issue is that the membrane is generally attacked by whatever etch method is used to strip the native oxide.  Chris had suggested a vapor HF etch, and I machined a simple fixture to test this with two weeks ago.  I started with a 6″ x 6″ x 2″ block of polypropylene and machined a 4″ x 3″ x 1.5″ pocket to hold a shallow puddle of dilute HF acid.  I also made a lid and machined a slot and ledge that allows one to slide a plastic screen holding wafer samples into the closed chamber so that it is exposed to the nominal equilibrium vapor pressure of HF.

Figure 1:  Sketch of vapor HF fixture.  It’s machined from a 6″ x 6″ x 2″ block of polypropylene.  The lid is 1/2″ thick and has a lip to sit inside the ledge of the block.

I found a reference for HF vapor pressure (J. C. Brosheer, F. A. Lenfesty, and Kelly L. Elmore, “Vapor pressure of Hydrofluoric Acid Solutions,” Ind. Eng. Tech., 39 (1947) 423-427).  The summary is below.  Even at 2% concentration, the vapor pressure is on the order of 50 mTorr.  For reference, at 1 x 10^-6 Torr a surface experiences one monolayer for exposure every second.  So, 50 mTorr results in 50,000 monolayers of exposure every second.  Given that HF reacts so aggressively with SiO2, it can be expected that the strip will proceed quickly.  Allowing one to slide samples in through the narrow slot without removing the lid, helps maintain a more constant HF vapor pressure in the etch chamber.

Figure 2:  Table of equilibrium vapor pressures of HF in water.

I started with some concentrated (49% HF/water) solution and tested a couple monitor samples of wafers that have 100 and 300 nm layers of SiO2 deposited with TEOS in PE-CVD.  I placed them in the etch chamber for 15 minutes by placing them face-down on a polyproplene screen that can slide in from the side through the slot while leaving the lid in place.  After 15 minutes, the oxide layers were completely stripped, so I clearly need a more dilute acid solution.  I diluted the HF acid to 50:1 and repeated the etch test with new samples leaving them in the fixture for 10 minutes.  I measured the before and after oxide thickness using the Nanospec spectrophotometer and found that about 29 nm of the oxide was removed for an etch rate for annealed TEOS xide of 2.9 nm/min.

I then tested the efficacy of stripping the native oxide through micropores with a sample of 100 nm SiO2 on 1 µm poly-Si layer on 100 nm thermal oxide etch stop.  I placed the sample in the vapor HF fixture face down for for 60 s followed by rinse in DI water.  I found that it etched very efficiently in the Xactix XeF2 etcher, exposing the deep blue thermal oxide etch stop layer after just a few pulse of XeF2.