NPN Liftoff Update – 7/7/16

The NSF STTR phase I project officially ended last week, and we are writing the final report and preparing to submit a Phase II proposal.  I will attempt here to summarize the program achievements and list some of the persisting technical issues we are facing.

Program Technical Objectives:

1. Optimize nanoporous silicon nitride (NPN) membrane properties and related LO processes to create nominally five by ten cm size sections that are free of pin-hole or other defects. Target a yield of two such membrane samples per ø150 mm wafer
2. Optimize dialyzer module design for prototype device by:
   • providing necessary membrane mechanical support and fluid sealing, maximizing open area and minimizing pressure drop and recirculation zones in blood flow path;
   • fabricating hemodialysis prototypes from biocompatible materials at sizes that can readily be scaled for commercial applications.
3. Demonstrate the efficacy of hemodialysis prototypes in bench test using bovine blood by measuring clearance rates of urea and β2-microglobulin with the following targets:
   • clearance rates with the silicon nanomembranes that are more than 10-fold greater than those achieved with commercial membrane materials with the same active area;
   • serum albumin retention of > 99%

Summary of Achievements:

Objective 1 — Optimized NPN Membranes and LO Procedures: We basically achieved this objective.  We focused on smaller area samples largely because the modeling results of the dialyzer module indicate that ~9 cm2 membrane area is all that is needed for a continuous treatment device.  Specific approaches used to strengthen the membrane include:

• Selective regions of nanopore transfer for added strength, using a hybrid micro-nano pore pattern to create a contiguous, non-porous matrix that supports nanoporous regions.
• 50% added thickness to nanomembranes over initial materials, using a more refractory nanoporous template that permits longer pore transfer etches into a thicker, underlying film. 75 nm (vs. initial 50 nm thick) NPN LO membranes were developed and are strong enough to be lifted off without the micro-nano pore hybrid pattern.
• Developed transfer processes using a static charging of plastic rod to roll up membrane and move to gasket supports. Also developed a combination static charge, vacuum chuck capture method for transferring released membranes from wafers onto gasket supports, to help maintain wrinkle-free membranes during transfer.
• As an alternative to both traditional on-chip NPN membranes (see below and the LO membranes developed in this project, we established initial feasibility for an alternative high active area membrane fabrication strategy using an in situ through-pore etch technique to create fluidic channels directly underlying NPN membranes. These “trench” channels offer advantageous dialysis and fabrication properties.

Objective 2 — Module design:  We accomplished this objective.  Modeling of mass transfer within NPN membrane-enabled dialyzers revealed that a dialyzer incorporating ~1.5 x 1.5” (or 9 cm²) of NPN membrane area is sufficient for continuous dialysis treatment.  This membrane area requirement assumes certain (yet practical) fluidic channel heights and blood/dialysate flow rates. The revealed design incorporates flow features for near elimination of regions of high shear or recirculation.  This membrane area target sets requirements for Phase II, as well as needs for successful future large mammal and initial human feasibility testing trials. Modeling results were published in a peer-reviewed journal (Burgin et al., 2015).

Objective 3 — Clearance Performance in Prototype Dialyzers: We are still working on this objective. Despite the improvements we made in NPN membrane strength, etching techniques to release NPN membranes from wafer substrates and initial success in capturing released membranes, we have been unable to assemble defect-free membranes into fluidic devices.  We hypothesize that membrane wrinkling may be causing defects that prevent our ability to seal LO membranes fluidically within prototype dialyzers. We implemented a wide range of mitigation strategies to avoid and/or repair wrinkles and defects, as well as different fluid channel designs and flow protocols but have yet to produce and then successfully test dialysis performance in a defect-free device incorporating LO membranes.  We will continue to document failure modes, attempt additional iterations and mitigation strategies and the alternative trench channel membrane approach during the intervening time until our Phase II submission.

Summary of Persisting Technical Issues:

Release Etch:  We are still experiencing variability in the efficacy of the through-pore, XeF2 release etch.  In some cases the membrane “clamps” to the underlying SiO2 layer once the sacrificial poly-Si is removed.  It is not completely clear what causes this to occur sometimes.  Once possible issue it that the pressure cycling in the Xactix tool (successive pulse of XeF2 and purging N2 w/ intermediate vacuum evacuation) cause the partially released membrane to flap back and forth.  We also suspect that various static charges can cause the membrane to be pushed down onto the substrate.

Transfer Process:  The vacuum chuck process seems to work fairly repeatably even without addition of static charging.  The main issue is that the membrane still have some wrinkles from the fairly open mesh used to hold the membrane corners down during the XeF2 etch.  This can likely be solved by moving to a finer mesh material.

Failures during device filling: To date, we have not succeeded in filling a device with the LO membrane.  Generally, a large cross-over leak develops when filling the channels.  Analysis of failed devices indicates that the failures occur at random areas – i.e., not just at the interface of the membrane and the pdms gasket.  These failures could be a result of damage that exists before filling (caused by the etch and/or transfer process), related to areas that are wrinkled, or simply because the membranes are too week to withstand the forces of the filling process.  We are using much wider channels (2 mm vs. 0.7 mm for the chip-supported membranes).  Going to very narrow channels is an issue because they can get clogged with Sylgard when doing the spot defect repairs during device fabrication.