Doug Sparks, president, M2N Technologies
The semiconductor and microelectromechanical system (MEMS) supply chain crisis, which started in 2021, is still ongoing, with rising prices for foundry services and smaller volume customers being turned away by established complementary metal-oxide-semiconductor (CMOS) and MEMS wafer fabs. MEMS foundries known for accepting low-volume (c. 300-1,000) wafer orders are turning them away or charging significantly more for them. Moreover, new MEMS-based product development cycles have increased from a year to up to three years in 2022 due to lack of fab capacity.
MEMS-based products require not only MEMS chips, but also CMOS application-specific integrated circuit (ASIC) chips, packaging and customised test and calibration. Even if the company can create a new MEMS chip for their product, the same high-capacity utilisation problems exist in the CMOS wafer fabs where the ASIC chips are made. Fabless startups are perhaps the most adversely affected by the new supply chain pinch. If the new product requires a different material set than traditional silicon, it is almost impossible to move the startup’s wafer process out of a university fab into industrial manufacturing.
MEMS has been a part of the More than Moore (MtM) growth, which is moving beyond just shrinking the linewidths of silicon wafer circuits. MtM involves new processes, wafer stacking, 3D structures, moving parts on a chip and new wafer materials. Traditional wafer fabs will not allow these alternate substrates into their fabs because of the possibility of cross-contamination between these new wafer materials with the fab equipment or silicon wafers. Sodium or transition metals can cause high PN junction leakage currents, emitter-collector pipes and degradation of minority carrier lifetimes and gate oxide, plus transition metals are known to diffuse rapidly through silicon wafers.
Dedicated wafer fabs for these specialty wafers, such as silicon carbide (SiC), III-V compounds and even stainless steel, have been developed. Nagano Keiki, a Japan-based manufacturer of pressure gauges, pressure sensors and measurement control equipment, has its own internal sensor fab for making stainless steel high-pressure sensors such as the one shown in figure 1. Each corrosion-resistant, stainless-steel diaphragm is machined from rod stock into a thimble-shaped sensing element, the flat top surface is polished, and then these sensing pieces are processed in a fab in trays, just like a wafer. Both physical vapour deposition (PVD) and plasma enhanced chemical vapour deposition (PECVD) processes are used to form thin-film Wheatstone bridge circuits on the stainless-steel surfaces of the sensor diaphragms. The piezoresistive layer is a doped PECVD polysilicon film. This custom-designed, dedicated wafer fab has allowed Nagano Keiki to merge semiconductor processes with stainless steel for industrial and automotive pressure sensor applications.
Figure 1
A stainless steel sensor high-pressure produced by Nagano Keiki.
Other MEMS processes have been developed for high-temperature SiC and piezoelectric quartz wafers. The University of California, Santa Barbara (UC Santa Barbara) developed a titanium wafer MEMS process complete with titanium deep reactive ion etching (DRIE) and laser and wafer-to-wafer (W2W) bonding. This was commercialised by PiMEMS, a US-based manufacturer of titanium MEMS devices, using the university’s wafer fab. Sodium-containing BOROFLOAT has been employed for anodically bonding to silicon and as a microfluidic channel wafer. BOROFLOAT and quartz can be ultrasonically machined and even plasma etched to form cavities, channels and thin walls. A variety of polymers are used for microfluidic medical diagnostic and drug infusion applications. However, outgassing of the polymers often prevents most polymer substrates from being processed in vacuum-based semiconductor manufacturing equipment in a typical wafer fab.
3D printing has recently been adapted to rapidly produce new MEMS wafers1. A significant advantage of 3D printing wafers of complex MEMS and microfluidic structures is that less processing steps are required compared with traditional silicon MEMS fabrication methods. The most time-consuming of these methods are typically DRIE and W2W bonding, which only allow one or two wafers to be processed at a time and can each take more than an hour. Furthermore, DRIE is somewhat limited in etching direction, specifically straight down or at a slightly fixed angle into the silicon or glass substrate.
3D printing allows meandering channels and tubes to be formed as part of the wafer in a single step, as shown in figure 2. These wafers can be printed using a variety of corrosion resistant metals and polymers. The wafer shown in figure 2 is titanium while the one-piece printed optical pressure sensor package on top of the wafer is stainless steel. The printed microfluidic resonator tube and cantilever chips shown in figure 2 are titanium. Circuit layers can be formed on the top, flat surface of the printed wafer.
Figure 2
A 3D printed titanium MEMS wafer and stainless-steel optical pressure sensor package (main image); and 3D printed titanium microfluidic resonator tube and cantilever chips (inset image).
The 3D printing technology known as photopolymerisation is being used to create the micron and even submicron microfluidic channels and filaments of stents, drug delivery devices and sensors by companies and institutions including Boston Micro Fabrication (BMF), in the US, Microlight3D, in France, Multiphoton Optics, in Germany, and Northeastern University, in the US. On a modified platform, similar to conventional photolithography tools, they are using microstereolithography, precision optics and platform motion control to print complex polymer substrates and devices using photosensitive materials. Northeastern University has adopted this type of technology to 3D print micron-sized structures on silicon wafers.
Mesoline, in the Netherlands, has developed a microfluidic polymer mask to 3D print across an entire wafer at one time with microdimensional resolution. The Mesoline approach uses microchannel particle deposition (MPD), which allows for the deposition of nanoparticles in a highly scalable and precise way. This 3D printing method is capable of patterning an entire 6–8 in. wafer in less than 15 minutes, outperforming incumbent technologies such as drop-casting and inkjet printing. MPD enables the fabrication of highly porous, high-aspect ratio structures, used as sensing elements of a metal-oxide gas sensor or getters.
Others are building upon the lithography, electroplating and moulding (LIGA for Lithographie, Galvanik und Abformung) technology developed in the 1990s to form 3D electroplated metal structures, formed in thick photoresist moulds and patterned on top of CMOS wafers. Exaddon, in Switzerland, has mounted a nozzle above a stepper motor platform to electro 3D print gold and copper antenna and microneedles on substrates.
All of these methods combining 3D printing with conventional wafer fab lithography circuit fabrication can form the same type of 3D microstructures often associated with MEMS devices, offering new paths in MEMS fabrication. The hundreds of MEMS wafer fab processing steps needed to form the complex cantilevers, channels, diaphragms and suspended tubes can be accomplished in a single print step, saving fab processing time and cost.
Another weak link in the MEMS and sensor supply chain is packaging. 3D printing has been used to fabricate cavity packages for MEMS sensors and microfluidic prototypes. By combining this printed cavity package with a printed circuit board (PCB) using off-the-shelf low impedance amplifiers and field programmable gate arrays (FPGAs), as shown in figure 3, the need for custom sensor ASICs can also be eliminated for startup prototyping.
Figure 3
A 3D printed cavity package with a printed circuit board (PCB) using off-the-shelf low impedance amplifiers and field programmable gate arrays (FPGAs).
In response to the logistics problems associated with COVID-19 travel restrictions and quarantines in 2020 and 2021, 3D printing has also been applied to wafer fab maintenance support. By having both polymer and metal 3D printers, replacement parts for processing equipment at fab facilities can be fabricated onsite. In addition, possessing a cloud-based directory of standard triangle language (STL) CAD files for replacement parts means that a large percentage of components can be printed on demand to support semiconductor processing tools and associated facilities. This can boost uptime by reducing replacement part shipping times and lower overheads by cutting down on spare part inventory levels.
InchFab, in the US, is another company addressing the production gap by providing fast cycle time fab services using its modular and scalable foundry solution, shown in figure 4. It can be assembled to provide the entire suite of MEMS foundry processes. This concept can save hundreds of millions of dollars in capital expenditure (CapEx) and years of construction time associated with a traditional wafer fab.
Figure 4
The Inchfab modular foundry solution.
The CHIPS Act in the US, one of many governmental funding schemes introduced around the globe to encourage semiconductor firms to build wafer fabs in their native countries, primarily supports large semiconductor firms in building large fabs similar to the ones they already have. InchFab’s modular foundry approach is geared towards fabless companies that do not have the capital and may not have benefited directly from the billions of dollars of government support to build a wafer fab. It enables them to run a local, flexible, possibly dedicated foundry for customers virtually anywhere in the world.
An InchFab modular foundry can be placed inside an existing industrial park building. The process chambers are compact and the gas requirements low, thus keeping costs low. Small gas cylinders are located under the main process chambers in the modules. Furthermore, the small, onsite gas requirements greatly reduce the time required for environmental approvals relating to the startup of a new MEMS fab. These improvements have been made possible by scaling down the processed substrate size to 2 in. This enables the reduction in the cost per chip, capital cost and physical footprint of the process chambers and fab by two to three orders of magnitude, while still retaining a similar level of performance. InchFab modular foundries can and have been placed in industrial park buildings, with no construction needed other than basic facilitation, e.g., power, exhaust and compressed air, needed.
This shrinking of the wafer foundry significantly reduces the material (wafers, gases, masks, etc.) and facilities cost of wafer fabrication. Additionally, the flexible design of the Inchfab allows users to develop fully customisable fabrication processes, accommodating not just silicon (Si) but non-traditional materials such as SiC, diamond, gallium nitride (GaN), gallium oxide (GaO), glass, metals and polymers and associated process flows needed for MtM products. An InchFab MEMS foundry can be placed virtually anywhere in the world to fill the foundry gap in a regional or even corporate MEMS supply chain.
In summary, while there are still several semiconductor and MEMS supply chain problems, new approaches are being taken to address the related difficulties startups as well as small and mid-sized sensor companies are facing. 3D printing can alleviate spare part and logistic difficulties for fab support as well as provide a new platform to quickly manufacture micromachined substrates and packages. Custom wafer or sensor fabs and flexible, modular MEMS foundries are being adopted to service low-volume customers and process new materials as well as to enable regional, national and even corporate MEMS supply chains. Together, these new fabrication methods are enabling the continued ramp up and commercialisation of micromanufactured products.
M2N Technologies
Reference1Sparks, D. (2019). The advantages of using additive micromanufacturing in the fabrication of MEMS wafers and sensors. CMM, 12(6), pp.20–26,Available at: https://bit.ly/2Vjv14A