Project 1000 x 1000

A cotton-candy (Rotary jet spinning) machine for distributed manufacturing of N95 mask filter material

Created: March 19, 2020

Last Updated: April 21, 2020

Documentation License: Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0)

Short Title: Replicable/simple table-top setup to produce nano-fiber based material for N95-grade filters

Detailed Title: Distributed manufacturing approach for a simple, replicable semi-industrial setup using standard motors (i.e. CNC routers, dremel bits) to produce nano-fibers to locally manufacture N95-grade filters.

Contributors: Prakash Lab members

Edward Andrew Mazenc <mazenc@stanford.edu>, Yuri Lensky<ydlensky@gmail.com>, Daniel Ranard <danranard@gmail.com>, Abby Kate Grosskopf <abbykate@stanford.edu>

Project Leads:  Anton Molina <antonsf@stanford.edu>, Pranav Vyas <vpranav@stanford.edu>, Anesta Kothari <anestak@stanford.edu>, Shailabh Kumar <shailabh@stanford.edu>, Nikita Khlystov <nikitak@stanford.edu>  

Document Steward/Editor: Manu Prakash <manup@stanford.edu>

This document proposes a distributed manufacturing process (rotary jet spinning) as a scalable method to enable manufacturing of N95 grade filter material (a nano-fiber porous mesh) - a key need of the hour. Current manufacturing of N95 grade filter material is centralized and relies on a melt-blown process which can not scale beyond one or two million masks/day per factory. We provide an alternative approach of a medium scale manufacturing (filter material for 10,000 masks per day) and if replicated by 1000 small businesses and groups across the country, would produce enough material for one to ten million masks per day. That’s why we call this approach Project 1000x1000.

Here we provide the layout of a simple machine (a single upright mounted Dremel or a machining router, something very similar to a cotton candy machine) anyone can find in a hardware store, and the exact process that takes styrofoam, a common waste material, as input raw material and converts it into a nano-fiber based N95 grade filter material. We further provide estimates for optimizing this process for larger scale production. This is an ongoing work and we are sharing this document as a live open lab-notebook - we will keep updating this and provide final protocols very soon.

We propose that a small group of 2-3 to 10 people can set-up 10-20 rotary jet spinning setups, and produce enough N95 grade material for 10,000 to 20,000 masks per day; locally where they are needed. We provide instructions for how to do quality control and measurements for the final product. And a 1000 groups across the country doing this will exceed the 1 million mask a day output capacity easily. The N95 grade filter material can be integrated in a number of mask designs currently being pursued.

We know this is a radical approach - but the acute shortage of PPE for front line workers necessitates that we try to think differently about manufacturing.

Quick Links:

• Directory of other Prakash Lab COVID-19 response projects

Table of contents:

Context

Goal

Design Specifications

Proposed Test Setup

Basic Scaling and Production

Custom Dispersion Tools

Polymer Preparation

Polymer available from chemical supply stores

Production of nano-fiber from waste sources - utilizing styrofoam

Current Progress Summary

2020-03-21

2020-03-22

2020-03-23

2020-03-24

2020-03-25

2020-03-26

2020-03-27

2020-03-28

2020-03-29

2020-03-30

2020-03-31

2020-04-01

2020-04-04

2020-04-05

2020-04-06

Electrostatic Charging of Fibers

Compaction of Fibers

Processing Techniques Studied

Rheological Characterization

Planned Tests

Filter Test

Pressure Drop Test

Relevant patents

References

Request for comments and ways to help

Documentation

Forking and Working Independently

Context

According to Project N95, 81+ Million PPE are currently needed in the United States alone and this number grows by the hour. And this does not even count the global need, including LMICs. As hospitals run out of N95 masks (critical protective gear necessary to capture 95% of particles at 0.3um or greater) and the supply chain isn’t yet able to keep up with the escalating demand, we have been looking for alternative means to produce similarly-performing filter materials. The fabrication of micro- and nanofibers to produce N95-equivalent filters is currently done via a melt-blow based extrusion process which requires extensive tooling of the head (too complex to be scaled up in a short period of time). At a smaller scale though, other methods such as electro-spinning, and rotary jet spinning can be used to make nanofibers. Our own calculations and literature review, we find that rotary jet spinning (RJS) has the capacity to meet the demand of medium-scale production of N95 grade nano-fiber material and seems to be the one that would be most applicable in this scenario[1]. RJS can produce nanofibers about 50x faster than electrospinning and does not involve the danger of high voltage [2] and is very simple to set up.   

Goal

Project 1000x1000 refers to the idea of distributed medium scale manufacturing of N95 grade nano-fiber material across the country, in small factories with 2-10 people involved with a capacity to make enough filter material for 10,000 to 20,000 masks a day. This allows for an alternative approach - where if 1000 small businesses engaged across the country - roughly 1 to 10 million masks could be produced in a distributed manner.

We propose to use waste styrofoam - commonly available everywhere - as a raw material to convert it into N95 grade nano-fiber material which provides the key ingredient needed for making high grade PPE masks necessary for health care workers. We list the hardware requirements of a very simple manufacturing setup (just a simple Dremel attached to a cup), process for handling styrofoam (and other polymers) to produce the nano-fibers and subsequent process to convert this material into a fabric for mask production. We provide basic calculations for scale up and optimization of this process, implementation of quality control in the production process and list a plan for engaging broader community and small businesses  to test, optimize and scale this process for global production.

Figure 1: Schematic and actual production of nanofibers starting with styrofoam as raw material. The setup consists of a simple Dremel mounted vertically attached to a 3D printed cup. 1 minute of spinning at 10K rpm produces (2.4 g) micro-fibers depicted above. See SEM for morphology. See link for video here for  Nano-fiber production process.  

Design Specifications

Although the filtration capacity for an N95 is stated to filter 95% of particles at or larger than 0.3µm, the fiber diameters are much larger than 0.3µm. Based on an SEM image analysis of a commercial N95 filter (Kimberely-Clark KC 200 N95), the average diameter of fibers were seen to be on the order of 3µm.

The thickness of the filter layer was approximately 500 µm. Some loose fibers are visible where the cut was made to image the cross-section, while tightly-packed fibers are visible in the lower layers (Figure 1, right). This indicates that while the size of the fibers can be larger than 0.3µm, many-layered compressive packing of the filter material can enable reliable production of N95 filters for respiratory masks.

Figure 2:  SEM image of the top surface of a N95 filter (left). Cross-section of a N95 mask filter (right).

Our goal is to be able to replicate the production of this material at medium-scale in locally produced factories.Combined with the capacity to test for particle capture quality using handheld particle counters (Lighthouse Apex Z), we can build local capacity to produce this material.

Proposed Test Setup

Figure 3: Schematic of a rotary jet spinning (RJS) machine aka cotton candy machine. Proposed multiplicity of 9 machines running in parallel.

Basic Scaling and Production

A rotary jet spinning machine ejects a viscous polymer solution from nozzles embedded in a spinning reservoir towards a collecting surface. Thin polymer fibers are formed as the jets, emitted from the nozzles (center), travel to the collecting drum. A schematic of our test machine is shown in Figure 3.

The goal of the project is to produce high-throughput sub-micron fibers through a distributed production. With multiple dremels/CNC routers available in many shops, there is an opportunity to produce the comparable amount of raw filter material that commercial/industrial manufacturers are able to output per day.

It has been reported that lab-scale versions of RJS can have a production rate of 60 g/hr per needle orifice (where the solution comes spinning out of the central axis). Furthermore, industrial scale RJS setups (such as the FibeRio CycloneTM L1000M laboratory machine) have been reported to produce around 12,000 g/hr continuously.[Reference: Rogalski et al. (2017). Rotary jet spinning review–a potential high yield future for polymer nanofibers. Nanocomposites, 3(4), 97-121.]. In another report, a lab-based setup was able to produce styrofoam fibers at approximately 0.26 g/min or 16 g/hr [Fauzi, Ahmad, et al. "Synthesis of Styrofoam fibers using Rotary Forcespinning technique." Materials Science Forum. Vol. 827. Trans Tech Publications Ltd, 2015.]. Based on these reports, we nominate our targeted production rate as 60g/hr, as obtained for a nozzle-based system previously. The fibre width can be controlled by experimental parameters [A simple model for nanofiber formation by rotary jet-spinning]; a scaling relation for the fibre width  is

where is the orifice area, the typical orifice exit speed, the kinematic viscosity (where is the extensional viscosity),  the radius of the device, and the angular speed of the spinneret.

Next, we dissected N95 masks ((Kimberely-Clark KC 200 N95 and 3M™ N95 mask (8200/07023 (AAD)) and measured the weight of the mask filter material to be 1.7 grams.

Figure 4: N95 mask filter weighed to calculate the mass per unit cross sectional area.

A single lab-scale RJS nozzle system should be able to make filter material worth 35 masks/hour. With a simple set-up of nine dremels (for example), we would be able to produce 315 masks/hr or 7,560 masks/day. The upper limit for production is almost 1000 times higher; but nominally we expect to generate material for 20,000+ masks per day with 10 machines running continuously.

Custom Dispersion Tools

We initially tested this set up using a hand-held dremel (Dremel 4000) as the rotating mechanism, which has a variable speed from 5,000 to 35,000 rpm. Using the standard screw-top attachment standard for most dremels when used with an abrasive sanding disc, we had a 3D-printed custom holder which attaches and is screwed onto the rotating end of the dremel and has 6 slots to receive six 1.5mL eppendorf tubes. Each of the eppendorf tubes was punctured at the tip, through which a pipette tip was slotted and secured to the tube using electrical tape. Each tube was then filled with the liquid polymer and recapped (see Polymer Preparation).The assembly was attached onto the dremel and spun at 10,000 rpm within a box enclosure, wrapped in aluminum foil.  

Figure 5: 3D-print tube holder that allows for multiple nozzles to be running simultaneously.

The next set up we are planning to test uses the same dremel, but replacing the eppendorf tubes with a custom 3D-printed cup that directly mounts onto the dremel. The cup would allow for a larger volume of liquid filament to allow for a larger volume output. There is an option to have the cup be “lifted” off from the dremel screw, through a twist-on assembly. The cap cover is optional to prevent spillage of filament through the process.

Figure 6: (top) basic 3D-print cup, (bottom) 3D-print cup with cap and base

The cup may pose a limit on the amount of volume we can output per cycle. Rather than a fixed volume, we iterated on the dispersion tool to allow for a dripper from the top as a continuous filament feeder. This design also allows for a more secured connection for the gauge needles by using luer locks as the intermediate interface. For this prototype, we simply fitted and hot glued the luer locks directly onto the 3D-printed part.

Figure 7: 3D-printed nozzle (trio) for dremel

The next scale-up set up would be to take advantage of a standard CNC router. This model assumes the ability for the router coupler to be mounted upside down, so that the collet is facing upward. For our purposes, we have used the standard shapeoko CNC router. As an initial prototype, we looked at using a similar cup set up as we did for the dremel but with a vertical shaft that connects and secures into an existing CNC router head. This would be best with an aluminum milled piece, as typical rotors on a CNC machine can spin up to 27,000 rpm. With this set up, we hope to have a significantly higher throughput and consistency.

Figure 8: 3D-milled cup mounted on existing CNC router

As we had done with the dremel model, we shifted this design to also allow for a continuous filament feeder from the top. This would also be milled out of aluminum since we are also assuming a high rpm of 20,000 or greater. The funnel would allow for tolerance from the continuous feeder. The nozzles can either be custom made (depending on your milling skills) or off-the-shelf paint nozzles that can be tapped into a simpler milled piece.

Figure 9: 3D-milled nozzle with funnel for CNC router

The following are images of our prototypes and set-ups for both the dremel and CNC versions.

Figure 10: Experimental set-up: (left) 3D-printed cup mounted on a dremel (model: Dremel 4000), (center) 3D-printed eppendorf tube holder, (right) 3D-printed “basic” cup with attached gauge needles.

Figure 11: Experimental Setup: (left) initial setup using posts arrayed around the central cup with aluminum foil wrapped around the outside perimeter of the posts, (right) a larger setup to allow for a larger collection space and increased travel distance for the droplets.

Figure 12: Experimental Setup: (left) latest outcome using the dremel 3D-printed cup, (right) Shapeoko router mounted in a more stable enclosed chamber along with a milled aluminium cup to allow probing higher temperatures for the RJS process.

Figure 13: Experimental Setup: (left) latest outcome using the dremel 3D-printed trio nozzle, (right) current milled nozzle with funnel for the CNC.

Associated Print Files:

• Google Drive
• For Dremel:

• Dremel 1.5mL Tube - contains .stl and .stp files for the tube holder model
• Dremel Cup - contains .stl and .stp files for both versions of the cup (w/ and w/out base and cap)
• Dremel Nozzle Trio - contains .stl and .stp files for three-nozzle to dremel model
• Dremel Collector Stage - contains .ai, .dwg, and .dxf files for for collection stage

• For CNC router:

• CNC Cup - contains .stl and .dwg files for cup to fit on CNC router bit holder
• CNC Cup (simple) - contains .stp files of two simplified versions (a and b) of CNC router bit holder
• CNC Nozzle Funnel - contains .stl and .stp files for nozzle-applied funnel for CNC router bit holder

Polymer Preparation

Polymer available from chemical supply stores

We did the first series of experiments with PVA. Polymer solution: initially a 13.5 wt % (w/w) solution of PVA (MW 85,000-125,000 g/mol 99% hydrolyzed) was dissolved in water by magnetic stirring for 5 h just below boiling (80-90 C). This concentration was too high and resulted in the formation of large aggregates. Dilution to approximately 8%  produced a homogenous solution that could be used for RJS; in future experiments we will attempt to get more precise measurements of concentration.

The tip of an eppendorf microcentrifuge tube (polypropylene) was clipped off with scissors. Next a luer lock needle with 24 gauge was added and sealed with hot glue. The eppendorf tube was loaded with 1 mL of the PVA solution and loaded into the 3D printed RJS chuck. The dremel was initially set to 5k RPM, this was too long and no material was extruded. Increasing the rotation to 10k RPM led to the extrusion of material and successful fiber formation. Fiber was collected approximately 10cm away from the chuck onto aluminum foil.

Production of nano-fiber from waste sources - utilizing styrofoam

We started using styrofoam as our next testing material, as previously demonstrated in another paper. This is advantageous since waste styrofoam is available globally everywhere and if we can turn a waste product into a functional material, it can drive the costs down significantly. We successfully spun nano-fibers from Styrofoam and are continuing to refine this process for production of filters.

Current Progress Summary 

2020-03-21

We successfully spun a small amount of microfiber by creating an ad hoc RJS system using a dremel and eppendorf tubes. Although the fibers contain beads (from each initial droplet), we believe that we can get rid of them by increasing the concentration and/or viscosity of the polymer solution and increasing the rotational speed of our system. We believe that we can achieve a thread smaller diameter by increasing the rpm and increasing the capture distance. Also testing multiple materials remains to be explored for reducing fiber diameter and filtration efficiency, e.g. blends with cationic polymers.

Figure 14: Left: With a 4x lens, hierarchical branching networks of the polymer are evident. Right: Here is a big stem with a thickness of almost 20 microns. The stem is surrounded by several smaller threads.

Figure 15: Left: Here is an individual smaller thread. The smaller threads appear to consist of beads intermittently spaced by thin portions. This probably occurs as a result of some instability in the jet due to unoptimised process parameters. Right: Here is a zoomed in view of the beads. The thinnest sections of the thread varied from about 0.8 microns to 2.0 microns across the sample, while the bead size varied from 2.5-5 microns.

Figure 16: Left: A bead on one thread can break into smaller threads with smaller beads showing that there is hierarchy in bead formation as well. Right: Elongated beads are also visible in some regions of the sample.

Figure 17: Left: Smaller fibers may also entangle as a bundle. Right: Zoomed in view of the same.

SEM Analysis and comparison with N95 mask filter

Figure 18: Left: A bunch of fibers with beads visible. Right: A single fiber with diameter approximately 1 µm.

2020-03-22

We attempted to spin a larger volume of filament using the cup configuration attached onto a dremel. We were not able to generate a large volume of fibers, possibly due to either the current chamber temperature (not warm enough) or the filament solution (concentration of PVA). We were able to generate some fibers, which accumulated on the foil around where the posts are located behind it.

We plan to further optimize and investigate the PVA polymer mixture along with the chamber conditions for the next experiment. We are also planning to test with styrofoam as our next polymer.

2020-03-23

We have been able to produce a small amount of nanofiber using the cup and dremel setup with a styrofoam melt in acetone. The set-up was run for roughly 20 seconds at 10,000 rpm. Fibers accumulated not on the collector surface, but rather across the extended needle tips.

Figure 19: Production of fibers from styrofoam melt using acetone. Left: Styrofoam piece from waste packaging. Right: Initial test run on the Dremel 4000 setup.

Figure 20: Left: At 4x magnification, we see a rare mesh of fibres. Right: Zoomed in at 20x magnification.

Figure 21: At higher (60x*1.5x) magnification, single fibers can be seen ranging from 20 to 5 microns.

2020-03-24

We’ve successfully produced a sizable amount of nano-/micro-fibers using the cup and dremel setup with melted styrofoam in acetone. We have experimented with different solution concentrations. Preparing the styrofoam melt directly in the dremel cup attachment is the most convenient. Currently, we have left the max rpm at 10,000 to maintain balance on the attached cup.

2020-03-25

Preliminary viscosity measurements indicate that there are no significant differences in rheology for different wt% polymer melts and these values are stable for at least 24 hours. Practically this means the styrofoam does not have to be accurately weighed out to give reproducible results and these melts can be prepared in advance of processing. We measured the mass of the high yield fibers from (3.24.20) to be 2.4 g. However, these thick fibers dry out and become quite brittle, thinner fibers appear to retain their mechanical properties.

We have begun to explore alternative solvents, specifically dimethylformamide (DMF). Expanded polystyrene dissolved (EPS)  in DMF dissolves over several minutes to produce an opaque, homogeneous, viscous solution with a slight yellow color. Preliminary investigation suggests that this mixture is capable of producing thinner fibers than polystyrene in acetone. This will be investigated further.

Figure 22: Left: Fibers created from Polystyrene dissolved in DMF imaged at 4x*1.5x magnification. Right: Zoomed in view at 60x*1.5x magnification. Fiber diameters lie predominantly in the range 5-10 microns.

Figure 23: The smallest achievable fiber size was ~1.5 microns

2020-03-26

1. We investigated three different concentrations of EPS in DMF at two different speeds (5k and 10k RPM). Concentrations prepared were 5, 7.5, and 10 g in 25 mL DMF. All experiments were done using a 24 gauge tapered (red) plastic luer lock tip. The highest yield of fibers were produced at 7.5 g. These experiments were performed using a continuous drip based flow system with a bi or tri nozzle design that will be useful for improving throughput. Both the design and DMF based solvation strategy improved the throughput to a scale that allowed us to collect multiple layers of fibers.
2. We also attempted to fabricate polypropylene fibers using a solution based approach. Briefly, polypropylene (sourced from Eppendorf microcentrifuge tubes) was dissolved in warm xylene (100 C) and let to cool. The cooled mixture formed a gel that was too viscous to process.

Figure 24: 7.5 gm Polystyrene + 25 ml DMF spun out using a 24 gauge tapered tip at 5000 rpm. Left: 4x*1.5x magnification. Right: 20x*1.5x magnification. Fiber size is dominated by fibers in the range 1.5-3 microns, but excessive beading caused the fibers to stick together. We are getting smaller fibers and larger output!!

Figure 25: 7.5 gm Polystyrene + 25 ml DMF spun out using a 24 gauge tapered tip at 10000 rpm. Left: 4x*1.5x magnification. Right: 20x*1.5x magnification. Fiber size dominated by fibers in the range 3-7 microns. But the fiber morphology was a lot smoother and allowed us to wrap multiple layers to create testable filter like mesh.

2020-03-27

Since the fibers produced due to Polystyrene + DMF were too brittle, we decided to try out making fibers from polymer melts. In order to do so, we needed a chuck that could keep the polymer in a molten state at a temperature higher than the melting point of the polymer. Hence, we utilized the metal chucks designed earlier by the team and tried maintaining higher temperatures to melt polyethene bags using an electric heat gun. Unfortunately the temperature was not enough to keep the polymer in a molten state and it solidified quickly. Thus, we decided to explore other ways of creating polymer melts.

2020-03-28

[Filtration test results for the solvent based sample Polystyrene + DMF]

Polypropylene work continued because the fibers produced using polystyrene + DMF are too brittle after degassing, likely due to the polystyrene’s glass transition temperature being substantially higher than room temperature (100 C). Polypropylene was successfully melted without necessitating solvent. This required a powdered form, which was produced efficiently from HPLC vials (Agilent 5190-2243). Because the glass transition temperature of polypropylene is ~0 C, vials were first frozen in liquid nitrogen to render the plastic brittle enough to fragment between two metal plates repeatedly struck with a rubber mallet (Figure 2020-03-28-1). While still cold, these fragments were then processed into powder-like form using a coffee grinder to particles <2 mm in diameter (Figure 2020-03-28-2). Upon heating to 140 C (confirmed by FLIR camera imaging), the powder fused into a molten state with molasses-like viscosity and very high surface tension. Upon contacting and drawing using a metal spatula, this molten polymer was successfully drawn into continuous fibers that was highly ductile and not brittle after drying (Figure 2020-03-28-3).

                                 Figure 2020-03-28-1                                Figure 2020-03-28-2

Figure 2020-03-28-3

2020-03-29

In order to move towards a melt based approach form the solvent based approach we had been pursuing, we ordered a couple of cotton candy machines to utilize the idea of melt spinning. The following purchase link can be followed for the product - https://www.amazon.com/Paragon-Professional-Concessionaires-Commercial-Construction/dp/B01DO7IF52/ref=sr_1_5?dchild=1&keywords=paragon+cotton+candy+machine&qid=1585804286&sr=8-5

(Although and other machines can be used for initial exploration)

The sugar used to make cotton candy melts at about 186oC which is higher than the melting point for Polypropylene 160oC making the heating chucks suitable for generating polymer melts. The machine comes in without a speed control and the motor runs at a fixed running speed which is around 3500 rpm. Ideally, we would like to have a setup where we could vary the motor speed as an independent parameter to test its effects on the diameter, but this serves as a good start. We have plans to later on switch the motor with a brushless spindle motor used in CNC operations (https://www.amazon.com/Spindle-Cooled-Milling-Converter-Engraving/dp/B01LNBOCDA/ref=sr_1_1?dchild=1&keywords=spindle+motor&qid=1585805092&sr=8-1).

Note that changing the default weight distributions in a high speed rotating device is risky and must be done with utmost care to ensure balance.

We set one of the machines in our machine shop to test dropping in crushed polypropylene collected earlier directly into appropriately heated chuck (based on the instructions to remain in the right heating zone for sugar). The temperature of the rotating chuck was measured to be approx 150 C (FLIR camera https://www.flir.com/products/flir-one-pro/). This temperature may be too high for polypropylene (PP) used since we noticed a burning odor. This might also be due to a large size distribution in the granulated PP. When the PP was introduced into the rotating chamber, smaller particles were cast into the air. After the experiment, we noticed that a large portion of the granular material had not been extruded into fibers but had remained within the rotating chamber. It is likely that simply more time is needed to run the device to convert these plastic granules into a molten state. In the end we have successfully fabricated polypropylene fibers at high yield (7.5 grams in < 1 min), though not in the current diameter range and morphology.

Going forward we will optimize the pre-processing of the PP material. Importantly, this will involve filtering granules to remove small dusts since this poses an inhalation hazard and is likely responsible for the burning of some material. Furthermore, we will now try to refine the process to produce fibers that are of the correct diameter for use in an air filtration application (~ 1 micron).

Figure 2020-03-29-1. (Left) Spin Magic 5 (Paragon, USA) cotton candy machine sourced from amazon. (Middle) side view of heating coil of disassembled rotational chamber in cotton candy machine. (Right) top down view of disassembled rotational chamber - granulated material is poured onto the red platform. TODO: add image showing mesh network to compare to our planned Al foil hack.

The designs of the setup can be accessed through the online manual for the product - https://images.homedepot-static.com/catalog/pdfImages/17/1703fb4a-2f58-457e-a0bc-69f0bccd68dc.pdf

The rotary chuck, which is the critical part that heats and rotates to extrude the molten polymer, is composed of a helical heating element encased touching the curved walls of the cylindrical stage. Also the walls were made porous using a wire mesh to allow molten material to ooze out as the chuck rotates. The general mode of operation would involve any material to fly outward from the center towards the heated wall and remain stuck to it until it melts. Once molten, the material would be forced to ooze out from the porous walls. The chuck is approximately ~15 cm in diameter and 8 cm in height. Another design feature was a pair of leather straps that when attached on the top cap, can generate air drafts allowing the rotor to create its own convection, that would allow fibers to dry up and collect in specific ways. Kindly comment or write to us if you need additional critical dimensions.

Figure 2020-03-29-2: Borrowed from the product manual

Figure 2020-03-29 Res: First fibers from the cotton candy machine imaged at 4x*1.5x, 10x*1.5x and 20x*1.5x magnifications. The fiber morphology was mostly rough with an average diameter of ~40 microns. There were smaller threads appearing from bigger stems that were about ~5 microns in size.

2020-03-30

To obtain thin fibers we pursued the idea of covering the wire mesh screen curved wall with aluminium foil and then making small holes using thin pins to force polymer melt to extrude out through smaller diameters. We sealed the gaps between the rim of the cylindrical ring and the top cap with kapton tape. Leather strap floaters attached to top cap of the chuck that are designed to control turbulent air flow around the chuck as it rotates, provided a solid substrate for the fibers to settle and we accidentally obtained a thick patch of ~ 2 micron fiber diameter size dense patch, which looked pretty similar to the N95 filter in its morphology. This supported that idea that producing the filter material is indeed possible with this machine and maybe we need to explore hle diameter and substrate collection strategies.

Figure 2020-03-30-1: Last image shows comparison between the N95 mask filter material (left) and material obtained from leather floater (right).

Figure 2020-03-30-res-1: Micrographs of fibers formed from over the leather floater at 4x*1.5x, 20x*1.5x and 60x*1.5x magnifications.Most fiber diameters were in the range 2-5 microns.

Figure 2020-03-30-res-2: First trial on treatment of polystyrene fibers with Xylene. The fibers merge at specific locations where Xylene droplets might have settled hinting towards use of solvents in crosslinking these networks.

2020-03-31

To further our efforts with polypropylene melt we ordered the four commercially available grades of polypropylene pellets out of which 3 arrived today.  Unfortunately, polymer 4 (MW ~340,000 g/mol) will not be received until June.

1. 428116  Polypropylene, Isotactic, average Mw ~12,000, average Mn ~5,000,
2. 428175 Polypropylene, Amorphous,
3. 427888 Polypropylene, Isotactic, average Mw ~250,000, average Mn ~67,000,
4. 427861 Polypropylene, Isotactic, average Mw ~340,000, average Mn ~97,000

To standardize the positive results obtained from the aluminium foil experiments we designed a solid aluminium cylinder to replace the entire wire mesh which had 8 orifices of 1mm (??check??) in size. We processed each of the three polymers under identical conditions and compared the resulting fibers. Unsurprisingly, the different polymers yield fibers with different properties. What is surprising is that the differences in fibers translate to the fibers being deposited in very different ways in our collection chamber.  For example, the amorphous polymer deposits as a fairly dense mat. The resulting material is sticky to the touch. This stickiness can be removed by briefly soaking in a solvent (xylenes or DMF) followed by desiccation, suggesting that the amorphous polymer might contain very small polymer chains that dissolved and were removed from the solvent. The isotactic polymers deposit along the walls of the collection chamber, more closely resembling what one would expect when processing sugar in a cotton candy machine. The high MW polymer deposits in discrete clumps. One explanation is that fibers cling to the spinneret until they reach some critical size and detach together before being deposited on the collection chamber. In contrast, the low MW deposits continuously at the chamber walls. Both isotactic polymers are collected as fluff balls for further processing. These observations suggest that there is significant space for optimization with respect to the exact type of polypropylene resin added. We are currently looking into acquiring polymer resins optimized for fiber formation from ExxonMobile and formosa plastics. A previous paper (Raghavan, B. et al. Journal of Engineered Fibers and Fabrics  2013) on forcespinning polypropylene suggests that resins with a melt mass flow rate of ~ 30 g/10 min at 230 C are good candidates for producing fibers with similar dimensions as those observed in commercial N95 masks.

Figure 2020-03-31-res-1: Top: Fibers produced from amorphous polypropylene at 4x*1.5x, 20x*1.5x magnifications. The fibers ranged from 10-15 microns in diameter. Bottom: Fibers post treatment with Xylene for about half and hour and images right after visible drying. Fibers swell to almost double the size. Overnight drying in vacuum makes the fibers brittle.

Figure 2020-03-31-res-2: Fibers produced from high molecular weight polypropylene at 4x*1.5x, 20x*1.5x magnifications. The fiber diameters were is the range 2-5 microns suitable for filter applications..

Figure 2020-03-31-res-3: Fibers produced from low molecular weight polypropylene at 4x*1.5x, 20x*1.5x magnifications. The fiber diameters were in the 10-20 micron diameter size range, the largest of all three polymer types.

2020-04-01

We took receipt of different aluminum shells for our chuck to test out the ideas of different orifice diameters (0.4 mm, 0.6 mm, 0.8 mm). However, we ran into electrical issues with the Paragon cotton candy machine, mostly likely a short-circuit associated with using the machine without any material in it - an attempt at a cleaning protocol. The machine appears to trip the MCB switches in the house circuit repeatedly. Running without material leads to a dramatic build up of heat since normally the polymer would absorb the energy during melting and keep internal temperatures lower. We are currently in the process of replacing the heating coils in the machine and testing out other cotton candy machine designs that might be drawing less current.

2020-04-04

A filtration testing capsule was produced using PVC pipes and laser-cut Plexiglas to appropriately accommodate RJS meltspun samples. The finalized setup involves two threaded PVC pipes between which two gridded Plexiglas washers are used to compress and air-seal RJS material for filtration testing. O-rings are used to ensure air-tightness. Preliminary filtration testing showed that RJS material was filtering some smoke from lit incense. Smoke was drawn through the filter capsule using a battery-powered camping mattress pump.

2020-04-05

Charge tests:

We used a surface voltmeter (Surface DC Voltmeter Model SVM2, Alphalabs Inc.) in order to measure the surface voltage and thereby calculate the surface charge of previously prepared polypropylene fiber mesh.
The voltmeter sensor was first set to zero using a reference (ground). The polypropylene fiber mesh was then placed 1 inch away from the voltmeter sensor (ensuring that no other charged surface was near the sensor) and the surface voltage was measured. Using the surface voltage, surface area of the fiber mesh and distance of the mesh away from the sensor - the surface charge density can be calculated as:

,

with f =

Where Q = surface charge,

A = area of fiber mesh,

V = measured voltage,

D = diameter of the fiber mesh,

L = distance of the mesh surface away from voltmeter sensor.

We measured the surface voltage of a Kimberely Clark N95 mask filter as well as lab-spun polypropylene (amorphous). Using our measurement we estimated the surface charge density of the N95 mask filter to be around -3500 nC/m2. Our estimate is in the same order of magnitude as charge density estimated for a polypropylene mask filter in literature (Chen et al., 1998). Exposure to isopropyl alcohol resulted in reduction in charge measured on the filter material, which is consistent with previous reports.

We also realized that the as-prepared amorphous polypropylene fiber mesh (roughly the same size and density as the tested N95 mask filter) had an estimated charge density of approximately 2000 nC/m2. This is the same order of magnitude as an N95 mask filter, however with charges reversed. Exposure to air plasma was able to neutralize positive static charge on the amorphous polypropylene fiber mesh.

Compaction: was tested for the first time on meltspun polypropylene using the cotton candy machine method. Samples were compacted using two heated metal plates on a hotplate set at 130 C. Temperature of metal plates was ~70 C by FLIR camera measurement. Compaction was performed for 30 seconds using manual force (a very rough estimate of 50 lbs over a 100 cm^2 area). Samples were then cut to fit the filtration capsule apparatus produced on 2020-04-04.

Meltspun polypropylene fibers (amorphous; isotactic low molecular weight; isotactic high molecular weight; all purchased from Sigma-Aldrich) were compacted using two heated metal plates on a hotplate set at 130 C. Temperature of metal plates was ~70 C by FLIR camera measurement. Compaction was performed for 30 seconds using manual force (a very rough estimate of 50 lbs over a 100 cm^2 area). Samples were then cut precisely to 17.25 mm diameter (233.71 cm^2) to fit the filtration testing apparatus. Mass and thickness of each sample are recorded below; A, B, C correspond to replicate compaction samples. Samples were excised from the center of material with maximal homogeneity (by visual inspection) using a scalpel. Thickness was measured using calipers.

2020-04-06

Evaluation of filtration efficiencies of disc cut samples using the new testing rig.

[Conclude with a discussion on the values obtained, possible directions of improvement that the results suggest]

Electrostatic Charging of Fibers

The electrostatic charge of N95 masks is a major contributor to their filtration efficiency, improving it at least 10-fold over uncharged fabric (Tsai et al., Journal of Electrostatics 2002; Peter Tsai, personal communication). Three main mechanisms of filtration occur when aerosol and solid particles enter a non-woven fibrous mask: 1) steric hindrance owing to inertial impaction and interception by the fibers, primarily in the case of large particles, removing them from air streamlines; 2) sequestration of smaller particles by the fibers out of air streamlines due to diffusive effects; and 3) electrostatic attraction between particles and fibers, which is size-agnostic.

The main filtration element in N95 masks is the middle layer consisting of meltblown polypropylene nonwoven electret (Peter Tsai, personal communication). Electret formation results in a material with electric dipole polarization that is remarkably stable to charge decay for years. This quasi-permanent electrostatic charge can be imparted on a non-woven melt-blown fiber mask fabric by one of three methods: 1) corona charging; 2) tribocharging; and 3) electrostatic fiber spinning. It is industrially accomplished typically by corona charge (Tsai and Wadsworth, Particulate Science and Technology, 1994; Tsai et al., INJ 2000; US Patents 4588537 & 4375718) but also by non-discharging electric fields (US Patent 5401446).

Corona charging involves producing a voltage potential in the vicinity strong enough to exceed the ionization of air, resulting in electrical discharging. We are attempting to achieve this using Tesla coils. Alternatively, we are looking into ion generator modules that are often found in commercial air purifiers.

Figure 26. Corona charging setup. The irony of using coronas to fight coronavirus leads this particular author to question whether reality is a simulation.

Tribocharging is the most rudimentary form of introducing electrostatic charge to fibrous materials and is rarely employed at scale. Tribocharging involves the physical rubbing of materials with opposite charge on the triboelectric series. For the case of polymeric fabrics, glass serves as an example material to produce electrostatic charge in the fabric (Tsai et al., Journal of Electrostatics 2002). In dry conditions, this is the phenomenon that results in sparking when fabrics are rubbed together, and is capable of generating potentials that measure in the thousands of volts.  

Figure 27. An unfortunate consequence of tribocharging for this member of the feline species. [Wikipedia]

Electrostatic fiber spinning involves the application of high voltage potential (5 kilovolt and higher) between a syringe nozzle containing molten polymer or polymer melt and a ground plate. Fiber formation is achieved at a critical voltage where electrostatic repulsion counteracts surface tension and elongates an initial droplet at the tip into a continuous fiber. This is due to the net charge buildup within the liquid, which then is displaced to the surface of the fiber as the polymer melt solidifies or the polymer solution dries in flight. Unfortunately, electrostatic fiber spinning requires a strong voltage potential to be applied to the syringe equipment holding the polymer, making it impractical if not hazardous to deploy at unspecialized locations in a distributed manner. However, the application of an electric field during polymer melt cooling is a generalizable method of imparting charge to non-woven polymeric fabrics, and can be applied during the compaction of spun fibers (below), for example as in US Patent 2740184.

Electrostatic charge is typically measured using specialized voltmeters with high impedance to minimize electric disturbance to the substrate being measured. Plopeanu et al, 2011 describes one such approach using the following equipment: a TREK model 3450 electrostatic probe; a TREK model 341B electrostatic voltmeter; and a Keithley model 6514 electrometer. Liu et al, 2020 describes another setup using TREK-542A-2-CE and lists the surface potential of several conventional electret fiber mats (the order of magnitude is a few kV). In general, most non contacting electrostatic voltmeters with range >10kV should be able to provide an estimate for the surface charge.

Compaction of Fibers

We are currently exploring scalable options for compression of bundles of fibers into a tight-fitting mat/pad. SEM analysis of an N95 mask filter (Kimberely-Clark KC 200 N95) indicates that the filter thickness is approximately 500 µm.

Matting - similar to the process of manufacturing natural felt sheets, where wool fibers tend to tangle and aggregate when agitated. Fibers are first aligned through a rotating comb and later merged into thin fiber sheets which then are steam compressed and agitated between plates. We can attempt a similar process by placing a loosely bound sheet of fibers between two vibrating plates.

Compaction - We plan to perform baseline testing for material compression into a filter simply by tightly compacting the bundle of fibers between two heavy, parallel plates - leading to formation of a filter sheet.

Among other standard methods for processing of fibers into masks, we are exploring heat [You, Young, et al. "Thermal interfiber bonding of electrospun poly (l-lactic acid) nanofibers." Materials Letters 60.11 (2006): 1331-1333.

], solvent drying and chemical bonding as leading options.

Processing Techniques Studied

Rheological Characterization

To characterize the polymer solution of styrofoam in DMF before RJS, shear rheology was performed. The solution’s viscosity is on the order or 0.1 Pa*s or 100 cP. The solution viscosity decreases with temperature and exhibits slight shear thinning behavior.

Figure 16. Shear rheology of styrofoam in DMF. Data collected using a 60 mm cone and plate on a stress-controlled rheometer at 25 C.

Planned Tests

Filter Efficiency

Using an aerosol particle size spectrometer, we have built a test setup within a fume hood, consisting of two pvc tubes, a funnel, and a particle counter. One of the pvc tubes is attached to a funnel with its wider end covered by the filter material and the PVC tube is simply held within the center of the fume hood. Both ends of the PVC tubes would extend and exit out of the fume hood for measurement. We then push submicron-sized particles through an aerolizer (or similar methods) within the fume hood and sequentially measure the distribution of particles from the two PVC tubes to assess the difference in the particle distribution within the hood and through the filter material. We will be using this set up to test the outcome of fiber mats from the RJS experiments.

We have just finished setting up a particle counter based test for testing the efficiency of these materials.

We set up a simple experimental test rig and method for testing the particle filtration efficiency of various materials including N95-grade masks. The setup pictured in the above image includes a LightHouse handheld particle counter (Model 3016 IAQ), Intex QuickFil 6C Battery Pump, a rubber stopper with 2 holes covered by 2 kim wipes to mitigate the airflow, Incense: Satya Sai Baba Nag Champa 100 Gram, connectors (universal cuff adaptor, teleflex multi-adaptor). The pump with the rubber stopper, covered by 2 kim wipes, in it, provides an airflow within a range of 0.2-0.4 cfm to mimic that of breathing. The incense produces particles of various sizes, including those in the range picked up by the detector (0.3 µm - 10 µm).  With the pump on, we measure the number of particles produced by the incense. Then we place the filter on the setup and run the particle counter to measure the number of unfiltered particles. To calculate the filtration efficiency, we calculate the ratio of unfiltered particles to the number of particles produced by the incense, and then subtract from one.

Pressure Drop

We set up a simple experimental method for testing the pressure drop across various materials, including N95 masks, during inhalation and exhalation. The setup in Figures 1 and 2 includes an Intex QuickFil 6C Battery Pump, a Sensiron SNP37 Pressure Sensor, a Honeywell AWM700 Airflow sensor, a rubber stopper with 2 holes covered by 2 kim wipes to mitigate the airflow, connectors (universal cuff adaptor, teleflex multi-adaptor), and a sample N95 filter.  The pump with the rubber stopper, covered by 2 kim wipes, provides an inhalation or exhalation airflow within a range of 0.2-0.4 cfm to mimic that of breathing. With the pump on, we measure the airflow applied to the mask, and the differential pressure drop across the mask.

Quality Factor Calculation

Once the pressure drop and the filtration efficiency of a filter material are measured, the quality factor can be calculated using the following equation,

where P is the fraction of aerosol penetration through the filter material, and p is the pressure drop across the filter material.

Relevant patents

1. Melt spinning

1. US9610588B2 Electret nanofibrous web as air filtration media
2. US7857608B2 Fiber and nanofiber spinning apparatus
3. KR20170088910A Melt spun filtration media for respiratory device and face masks

2. Compaction

1. US5707468A Compaction-free method of increasing the integrity of a nonwoven web
2. US4003783 Method for compacting a nonwoven fabric impregnated with a thermoplastic binder

3. RJS Rig design and collection

1. EP2794972B1 Process for laying fibrous webs from a centrifugal spinning process
2. US9527257B2 Devices and methods for the production of microfibers and nanofibers having one or more additives
3. US9181635B2 Methods for the production of microfibers and nanofibers using a multiple chamber fiber producing device
4. US8858845B2 Systems and methods for the production of microfibers and nanofibers using a fluid level sensor
5. US9731466B2 Systems and methods of supplying materials to a rotating fiber producing device
6. US20160138194A1 Systems and methods for controlled laydown of materials in a fiber production system
7. US9889620B2 Devices and methods for the production of microfibers and nanofibers

4. Charging

1. US5401446A Method and apparatus for the electrostatic charging of a web or film
2. EP0702994B1 Nonwoven filter media for gas
3. US4588537A Method for manufacturing an electret filter medium
4. US9802187B2 Non-woven electret fibrous webs and methods of making same
5. CA2245861C Electrostatic fibrous filter web

5. N95 mask design

1. US8580182B2 Process of making a molded respirator
2. US6827764B2 Molded filter element that contains thermally bonded staple fibers and electrically-charged microfibers
3. US7858163B2 Molded monocomponent monolayer respirator with bimodal monolayer monocomponent media
4. KR20190049842A Exhalation valve and respirator including it
5. US8342180B2 Filtering face-piece respirator that has expandable mask body

References

1. Centrifugal melt spinning

2. Triboelectric charging

Request for comments and ways to help

Here is a list of the resources and help we need, for folks who would be serious about supporting and helping with our efforts:

• Feedback: If you are a material scientist, have worked in the filter industry, or have experience with rotary jet spinning process, please provide your feedback/suggestions to Anton Molina <antonsf@stanford.edu>  
• Replication: If you have access to either a dremel or CNC router and have the capacity to replicate and test the methods mentioned above, please do and share your findings with us. If you need access to a particle size spectrometer, please let us know and we can connect you to someone who does and can conduct the test for you. Contact Anton Molina <antonsf@stanford.edu
• Particle Size Testing: If you have access to an Aerosol Particle Size Spectrometer, please let us know and we can connect you with either the materials we’ve been able to produce or to others who are able to replicate/improve on the method of fabricating the fibers, but need access to a particle tester. Please send your interest for collaboration with Adam Larson <aglarson@stanford.edu>
• Pressure Drop Testing: If you have access to a UEi Test Instruments EM152 Dual Differential Digital Manometer and are willing to produce a small testing chamber, please let us know and we can connect you with either the materials we’ve been able to produce or to others who are able to replicate/improve on the method of fabricating the fibers, but need access to a pressure drop measuring device. Please send your interest for collaboration with Adam Larson <aglarson@stanford.edu>
• Suggest CAD of essential adapters: You can also participate by building better 3D models of the cups we are connecting to the CNC router for other CNC models available. We need to provide the best secured connection that can withstand the high rotational speeds proposed, while keeping the liquid filament intact within the cup (without spilling) and possibly have an opening through which we can pump in more liquid filament throughout the process. Contact Manu Prakash <manup@stanford.edu>
• 3D Printing and Milling: We would like people to print custom cup attachments as proposed by us and others to allow each proposed 3D model to be fabricated and tested in the set up. Please let us know if you’d have access and are willing to either print or mill parts for us/others collaborators.Contact Anesta Kothari <anestak@stanford.edu>.

Documentation

We are openly sharing all working design and draft documents related to COVID-19 projects in the lab before they become finalized or ready as products. So consider this document as our idea sheet, lab notebook, and more. These are working (living) documents. But it is important to note that this is a working draft design document and is not a final solution or any medically approved device. All regulatory hurdles still need to be crossed before any patients and health care workers can be served with these tools and solutions.

Every week on Friday we will release an update to this document, with new partners and progress. Note that released updates will have their own revision numbers (listed in the page headers), and that these updates will be listed as named versions in the Google Docs version history for this document.

Forking and Working Independently

We will continue to release our documents and designs here so that others can help us, replicate our results, work on similar ideas independently, and/or come up with better solutions to problems identified here. We hope they will also share their solutions for open access to others globally.