Improving the MEMS Fab Supply Chain with a 3D Printer Virtual Warehouse

Early 2020 was fraught with supply chain problems for wafers fabs, electronic assembly sites and micro-nano businesses as COVID-19 shut-down international travel, suppliers and customers. As employees fell sick and quarantines tightened, work slowed in affected regions. Fabs in Malaysia and the EU reported temporarily reduced production and staffing during peak infection/lockdown periods earlier this year. While large fabs typically have significant spare parts on hand, smaller fabs, often MEMS fabs, rely on a Just-in-Time (JIT) inventory management approach. If a fab tool goes down and requires a part, it is ordered and shipped to the facility. Ideally the part arrives in a few days, but in some cases for older legacy tools this can take weeks. During 2020, due to travel bans, this lag could be months if the part had to come from another country.  Traditional JIT relying on international supply lines failed in the COVID-19 pandemic. In response, many have called for shorter, more local supply chains or a redundant suppliers and higher levels of physical inventory. 

An alternative method of improving the resilience of some segments of the supply chain can include the use of 3D printing as a virtual warehouse of many replacement parts for businesses and fabs, greatly reducing down times for key tools and facilities. 3D printing is also an ideal tool for speeding development and prototyping, capable of fabricating cavity packages for sensors and even structured MEMS wafers from a wide spectrum of materials. The DoD and NASA were some of the early adopters for using 3D printing in remote applications. Naval ships like carriers, forward bases in Afghanistan and even the international space station have all started using additive manufacturing (AM) to shorten their supply lines and speed repairs.

The semiconductor industry has already started using AM to make complex parts for legacy tools. Many of the older 100, 150 and 200mm diameter wafer tools no longer have OEM support. Some of these old parts were made by companies that no longer exist, or the volume of parts may be so low that casting or machining and assembly of the old parts is not financially feasible. Just like other markets for antique and legacy parts, 3D printing offers a faster and less expensive way to make replacement parts for semiconductor equipment. Materials for replacement parts commonly used in fabs like plastics, filled plastics, Teflon, Poly Vinyl Chloride (PVC), various stainless steels, titanium, copper, aluminum can all be 3D printed. Multilayer printed circuit boards with discretes, can even be 3D printed (1) for support of electrical systems.

When CAD files or paper prints of the part no longer can be found, the object can be scanned using a video camera or even smart phone to generate a rough CAD file that can be tweaked on a laptop. The .stl file can then be sent for 3D printing. AM can get around logistic nightmares we’ve experienced this year as travel bans were applied across the world. On-site AM can also avoid import taxes, duties, shipping fees and associated custom delays. In the future OEM tool and fitting suppliers will offer parts from a virtual warehouse in the cloud to certified 3D printers.

Quality: ISO & ASTM Certification

To ensure safety and enter main stream chemical and wafer processing applications, found in MEMS & semiconductor fabs, AM products and processes will have to pass the same quality certification processes that other metal & plastic forming methods and materials have in the past. The aerospace industry has pushed this technology into compliance over the last 10 years, with existing standards and is constantly creating new AM specific standards. Both ISO and ASTM have generated a variety of metal 3D printing standards. Additive manufacture is defined by ASTM: F2792 - 12a (2012) as the "process of joining materials to make objects from 3D model data, usually layer upon layer, as opposed to subtractive manufacturing methodologies, such as traditional machining”. ASTM F3301 - 18 is a post AM processing standard. Other examples include AM standards for materials like titanium and nickel alloys and fabrication techniques using the DMLS powder bed laser method. While various corrosion resistant stainless steels can be 3D printed with laser and e-beam systems, the same approach to the standards and certification process will be required for the adoption of AM manufactured repair parts made of other corrosion resistant alloys, plastics and PCBs for wafer fab facilities and tools. For spare part, the fab tool OEMs may require standards for materials like powder and filament as well as AM methods in addition to providing certified .stl files of the replacement parts.

Post AM Processing

Just like casting, injection molding and other fabrication processes, additive manufacturing does not stop after the 3D printing step. For applications demanding reliability and high quality, post processing will often be required. Post processing of 3D printed parts can start immediately after the printing step. A small machine shop, typical of most fab maintenance & facilities departments, is needed for this post-AM processing. For laser and e-beam fabricated metal the product is in a state very similar to a welded metal part. Significant stress is found in the metal structure. As in the case of welding, an anneal step can reduce some of this built in stress, which reduces the likelihood of warpage and cracking. The anneal, often in an inert atmosphere or vacuum, can be done before or after the part is removed from the build plate. Hot isostatic pressure is used in some aerospace parts to reduce voids and improve fatigue life of AM metal parts. If we want to fabricate vacuum pump turbine parts on-site, the AM processes & QC steps may be similar to those used to make 3D printed aerospace turbine blades. A band saw or Electrical Discharge Machining (EDM) is often used to remove the part from the printed build plate. Many printed objects require supports to maintain dimensional stability during the printing process. These fine support structures must be removed. Trapped metal powder may need to be shook out of cavities. Next the relatively, as-printed matte surface finish can be improved with machining, bead / sand blasting or tumbling. Additional machining and surface treatments are often common to insure dimensional specification and smooth outer surfaces. Plastic parts have use solvent sprays to smooth surfaces and remove supports. 

Finally, metrology is important to ensure the part meets the specifications and is free of voids and cracks. Quality tests for aerospace and medical devices are preferably nondestructive and can include dye-penetrating tests, ultrasonic scanning, X-ray and computed tomography (CT) scanning. For some applications witness coupons, powder chemistry and microstructure analysis are required. Internal cavities, channels and weight saving internal lattice structures can be a complicating factor in final inspection. This type of QC inspection may be required before a part can be used in a hazardous gas or liquid application in a wafer fab.

Plastic Cavity Packages

Plastic 3D printing generally requires feed stock made of filaments, powder, resin or rods. The types of plastic being printed include ABS (Acrylonitrile Butadiene Styrene), PLA (Polylactic Acid), PVA (Polyvinyl Alcohol), Nylon, PC (Polycarbonate) and HDPE (High Density Poly Ethylene). PC+PBT (Polycarbonate + Polybutylene Terephthalate), PVC and Teflon. These plastics are ideal for forming nozzles, unique structures for flow and pressure sensors as well as cavity packages and microfluidic channels. For early development, 3D printing offers considerable cost and time savings. For example, an injection mold can cost around $10,000 to $30,000 and required CAD design and months for mold and part fabrication time to produce the package. Using 3D printing for plastic parts avoids the mold costs and time waiting for mold fabrication.

3D printing was first used in the sensor community to make plastic models in the 1990’s. Figure 4 shows some examples of plastic packages that can be made using AM. The cavity package in the center of Figure 4 was 3D printed for a medical drug infusion device made back in 2004 (2). This AM cavity package had internal microfluidic channels and Luer fittings for interfacing with an IV infusion set. The microfluidic silicon/glass MEMS chip was glued to the 3D printed plastic package as was the FR-4 PCB. The MEMS chip was wire bonded to the PCB to make this prototype.

Another hurdle to overcome in AM plastic sensor package manufacturing is the joining of metal leads, bolt and screw inserts into the plastic housing. These are used in most electrical device and sensor package. One approach that has been taken for some plastics is to hot press metal fittings into the plastic after the printing step (3). This allows for abrasion resistant bolt and screw fittings insertion, used for mounting, to extend the life of the part. Metal inserts are shown on the right of Figure 4. Larger hand solderable electrical feedthroughs could use the same method already developed for the bolt and screw fittings.

Thermally insulating layers and well as insulators have also been printed on packages and PCBs. Package coatings have also been printed using layers of materials with varying atomic masses, also called graded Z materials, to improve radiation resistance for space applications (4). The process is similar to traditional 2D silicone gel coating of PCBs. These AM enhanced packages are typically used for memory chips, but can be applied to protect sensor ASICs, especially those with on-chip EPROMs.

3D Printing Structured MEMS Wafers

AM has been used to fabricate structured (3D printed micromachining) substrates. This could also be done on-site at MEMS wafer fabs to tighten the supply chain. Figure 5 shows an example of a 3D printed metal MEMS 100mm diameter wafer (5,6). This titanium wafer was printed using the direct metal laser sintering method.

The same types of cavities and bond pad openings made in traditional silicon wafers as a cap over MEMS devices using wafer bonding can be duplicated with various 3D printed metals and optically clear plastics. AM capping wafers or device wafers using a chemically active metal like titanium enables the fabrication of a gas gettering surface for vacuum packaging without the need for thin film getter (7) deposition and patterning steps. The rough as-printed titanium surface is well suited for a higher surface area impurity absorbing chip scale hermetic packaging. The printed wafer material itself can act as this absorbent getter material. Printing a clear plastic or glass also can be used in optical & RF sensor applications.

As the Figure 6 illustrates, wafer to wafer bonding is used in conventional silicon WLP to partially or fully enclose the fragile MEMS cantilevers and resonators device elements. 3D printing can combine the capping cavity and moving device element into one wafer. One big advantage to AM fabrication of complex MEMS and microfluidic structures is that it can reduce the number of wafer fab processing steps that would have been required with conventional silicon MEMS processing. The longest and hence most expensive MEMS fab processing steps are typically Deep Reactive Ion Etching (DRIE) and wafer to wafer bonding. In these steps only 1 or 2 wafers can be processes at a time and they both can take more than an hour to complete. DRIE is also somewhat limited in etching direction, straight down or at a slight, fixed angle, into the silicon or glass substrate. Wafer bonding to form channels can have problems with hermeticity at the bond interface (8) and be prone to burst failure. In general silicon has low fracture toughness and can rupture under pressure or shock, compared to metals like titanium and stainless steel. AM can simultaneously form vertical, horizontal and multi-directional channels and other structures, without a bonding interface and using fracture and corrosion resistant materials as illustrated in Figure 5 and 6. 

Not only can AM reduce the MEMS processing step count by hundreds of fab operations, 3D printing can form substrate features that just cannot be fabricated using traditional silicon-based micromachining. The illustration on the right of Figure 6 shows how through wafer vias, cantilevers, suspended microtubes, curved horizontal and vertical surfaces can all be simultaneously printed in one step, which is not possible using a silicon wafer and standard processing. Through wafer vias (TWVs), including fluidic vias with female tubing inserts can printed along with other WLP elements. Figure 5 shows a printed TWV that has an aspect ratio of 10:1, comparable to silicon DRIE. Via formation through 5-millimeter-thick wafers, roughly 6 to 10 times thicker than a typical 100-200mm diameter silicon wafer, has been demonstrated. 

Convergence of Traditional IC/MEMS and AM Wafers

For sensors there is often a need for an electrical interface, generally on the device wafer. Screen, ink jet and laser 3D printing are generally limited to minimum feature dimensions for metal traces of 50 to 150+ microns across. MEMS silicon wafers integrate the electrical interface for sensor using wafer fabrication processes, leveraging Integrated Circuit (IC) fab technology. This can also be accomplished with AM+MEMS? wafers. By using wafer fab photolithography tools, the minimum feature dimensions for metal traces can go down to 5 to 2 microns for proximity tools and from 2 microns to less than 20 nanometers for stepper lithography tools. To use fab lithography tools requires a planar surface and a round wafer-shaped printed substrate.

CMOS wafer foundries will not allow transition metals and alloys like those used in many AM+MEMS? wafers. The cross contamination of silicon CMOS/BICMOS wafers with transition metals can cause high PN junction leakage currents, emitter-collector pipes, degraded minority carrier lifetimes and the gate oxide. Many traditional, pure MEMS foundries, which do not process CMOS circuit wafers in the frontend, have allowed the processing of sodium containing borofloat / Pyrex glass and other substrates like titanium wafers (9). This substrate material flexibility opens up the possibility not only prototyping of AM+MEMS? wafers but high-volume manufacturing with this technology. Metal and glass wafers of 100mm, 150mm and 200mm diameter are already being processed in some MEMS fabs and these facilities will be a natural manufacturing site for the speedy commercialization of this new AM+MEMS? technology. As AM substrate manufacturing grows, SEMI may need to generate standards for AM wafer materials and printing methods. We will need to rethink MEMS wafers and wafer fabs in light of the new capabilities in both materials and structures that 3D printing offers to substrate manufacturing.

For more information on 3D printing related to fab support, sensors, wafers and MEMS check out the references and come to:

COMS 2020 in Washington DC October 19-21 https://www.mancef.org/coms2020/ and the preconference session- Symposium #4, at Sensors Expo, November 16-18, 2020 in San Jose. https://www.sensorsexpo.com/

For more information on 3D printed MEMS wafers visit: https://3dprintwafer.com/

References

1. Sertogu, Hensoldt and Nano Dimension announce major breakthrough in high-performance electronics printing, 3D Printing Industry, May (2020). https://3dprintingindustry.com/news/

2. Sparks, et al., Preventing medication infusion errors and venous air embolisms using a micromachined specific gravity sensor, Drug Delivery Technology, Vol.4, No.4, p.82-86, May, (2004).

3. Vasquez, Threading 3D printed parts: How to use heat-set inserts, Hackaday, February 28, (2019).

4. Wrobel, et al, Versatile structural radiation shielding and thermal insulation through additive manufacturing, 27th AIAA/USU Conf. Small Satellites, SSC13-III-3, (2013).

5. Sparks, The advantages of using additive micromanufacturing in the fabrication of MEMS wafers and sensors,” Commercial Micro Manufacturing, Vol. 12, No.6, pp. 20-26, Dec. (2019).

6. Sparks, 3D-printed substrates for wafer-level and chip-scale fluidic packaging, Chip-Scale Review, Vol. 23, No. 6, pp.42-44, Nov-Dec. (2019).

7. Sparks, Thin film getters: from vacuum tubes to wafer scale MEMS packaging, Wafer & Device Packaging and Interconnect, Vol.1, p.19-22, June (2010).

8. Sparks, Advances in high-reliability, hermetic MEMS CSP, Chip-Scale Review, Vol. 20, No. 6, p.36-39, Sept-Oct, (2016).

9. Aimi, et al., High-aspect-ratio bulk micromachining of titanium, Nature Materials, Vol. 3(2), pp. 103-105, (2004).