Next Generation MEMS Manufacturing The ConFab 2017 Alissa M. Fitzgerald, Ph.D., Founder & Managing Member
Overview About AMFitzgerald Retrospective: MEMS technology history and markets Next generation MEMS processes and devices A dilemma for foundries and MEMS startups Page 2
AMFitzgerald: Your Partner in Specialty Product Development MEMS Innovation MEMS Solutions Technology Strategy AMFitzgerald develops innovative MEMS and sensor solutions for specialty applications. We collaborate with our customers to create high value products enabled by customized micro-technology. With integrity, expertise, and attention to detail, we deliver what has never been done before. Page 3 The ConFab 2017, San Diego, CA AMFitzgerald 2017
Full development services from concept to production AMFitzgerald in-house Strategic partners Custom MEMS development for commercial products Rapid prototyping on state-of-the-art tools Sensor supply chain creation and management Focus on high-performance, specialty sensor technology Headquarters in Burlingame, CA (near SFO) Fab operations at 1500m 2 UCB Marvell Nanolab Page 4 The ConFab 2017, San Diego, CA AMFitzgerald 2017
Our work is at the leading edge in many markets Typical revenue breakdown, by market Aircraft, spacecraft sensors Cardiology guidewires, pacemakers, pumps; diagnostic chips Microphones, pollution detectors Atomic sensors, commercial print heads, quantum computers Fiber optic networking, laser system components, infrared detectors Page 5 The ConFab 2017, San Diego, CA AMFitzgerald 2017
MEMS Retrospective, 1980-2010
1980s MEMS Technology Key process innovation: Alkali etchants selective to silicon crystal planes Anodic- or fritbonded glass Piezoresistive pressure sensors Source: Merit Sensor Source: SMI Inkjet nozzles MEMS nozzles Source: mems-exchange.org Page 7
1990s MEMS Technology Key process innovation: -like MEMS, using sacrificial etch of layers TI Digital Light Projection (DLP) Sacrificial etch frees thin film mechanical structures Pixel array Accelerometer MEMS Bi readout circuitry Analog Devices ADXL50 Page 8
1990s-2000s MEMS Technology Key process innovation: Deep reactive ion etch (DRIE), or the Bosch process DRIE >20:1 aspect ratio Comb drive shapes used in many MEMS accelerometers, gyroscopes, actuators AMFitzgerald ST C5L24A -friendly piezoelectric: AlN Avago Film Bulk Acoustic Resonator (FBAR) Page 9
MEMS industry dynamics 1000 Annual sales, USD M$ 500 200 100 Major Players Dominate the two high volume markets: consumer and automotive The Long Tail The other 400+ MEMS companies with emerging technologies #1 #30 Company ranking by annual sales Page 10
Flashback: the MEMS industry in 2006 80s- 90s technology 90s- 00s technology Major players in 2006 dominated with mature process technology in wellestablished markets New process technologies and devices lurking in the Long Tail InvenSense had just started in 2003 Page 11
Flashback: the MEMS industry in 2006 Vertically-integrated IDM companies having captive fabs and lucrative non-mems business units prevalent 14 out of 30 using fabs Page 12
2006-2007: Two consumer products that changed the MEMS industry Nintendo Wii: proved value of MEMS motion sensors in consumer devices Apple iphone: dramatically expanded market for MEMS motion sensors and microphones Analog Devices ADXL330 3-axis accel STMicro LIS302DL 3-axis accel InvenSense IDG-600 2-axis gyro Page 13
2000s-2010s MEMS Technology Process innovation: aligned, eutectic wafer bonding with vacuum seal Au-In or Al-Ge Mechanical seal and electrical connection Gyroscopes, resonators, inertial measurement units, combo units Silicon cap to create vacuum cavity MEMS Source: InvenSense MEMS and wafers fabbed separately, bonded together: smaller chip size Page 14
Seven years later: MEMS market transformed 80s-90s tech 90s-00s tech 00s-10s tech, from the Long Tail Winners: companies having MEMS for smartphones Lexmark off the chart Epson last place Page 15
Early adopters of next gen MEMS manufacturing became the big winners Big risks, big payoffs STMicroelectronics builds the first dedicated 200mm MEMS fab in 2005 Bosch opens its 200mm MEMS fab in 2010, not too late Dedicated MEMS fabs helped give ST and Bosch market power Freedom to innovate without material restrictions Quickly expanded product lines Page 16
Current market 16 out of 30 using fabs Two foundries now on the list, each having a portfolio of fabless MEMS customers Page 17
MEMS retrospective insights Process innovation drives MEMS device innovation fabs have played a key role in MEMS since the 1990s Semiconductor industry infrastructure needed to meet end-user demands for cost, volume, and quality Today: MEMS + high volume 200mm and -compatibility Unless you have your own captive MEMS fab New entrants come from the Long Tail, with new technology Page 18
Next Generation MEMS, 2010- MEMS Millennials
Next generation MEMS are coming from academia, not industrial R&D Most new MEMS devices are being developed in research facilities, where process freedom exists New designs leverage new materials and processes Many innovations were not developed with -compatibility in mind Technology commercialization path: startup company or spin-out Meanwhile, most IDMs have been expanding their product lines through M&A, not research Page 20
Next generation: Piezoelectric (PZT) MEMS Wide range of sensors and actuators possible using piezoelectrics AlN resonators and microphones established, due to -compatibility Process innovation: significant recent advances in thin film PZT deposition tools Source: AMFitzgerald Major interest in PZT for superior d 33 compared to AlN PZT is not -compatible Page 21
Next generation: Glass (mid-process) MEMS Glass offers optical, electrical, and thermal advantages Market pull for optical telecom, RF devices MEMS switch: 5G mobile infrastructure Process innovation: Through Glass Vias (TGV) Deep glass etch process Borosilicate glass preferred CTE-matched to silicon to facilitate bonding But contains Na+, a contaminant Hyperspectral imaging Page 22
Next, next generation: Gallium nitride (GaN) MEMS GaN epitaxy on silicon wafer -compatible, but not widely available Potential applications: Low-loss resonators suitable for high power applications, filters Timing, frequency reference Harsh environment sensors: accelerometers, pressure sensors Source: U Mich, Rais-Zadeh lab Source: MIT, Weinstein lab Page 23
Next, next generation: Graphene MEMS Graphene formed in CVD process with nickel or copper catalysts Process still in development Potential applications: Mass or gas sensors Selective gas sensing using a single FET NH 3, NO 2, H 2 O, CH 3 OH Microwatt power Supercapacitors Source: UC Berkeley, Lin lab Page 24
The next generation of MEMS, almost ready for market Application Sensor/Device Type(s) Special Process Needs Example Company Audio Microspeakers Piezoelectrics (PZT) USound Inspection and Security IR and Hyperspectral Imagers Glass Unispectral Communication Switches, varactors Glass Menlo Micro Cameras Auto Focus Piezoelectrics (PZT), Glass polight Gesture Recognition Ultrasound Piezoelectrics (AlN) Chirp Microsystems Micro Power Energy harvester Piezoelectrics (AlN) microgen Page 25
The smartphone (and consumer electronics) is still the killer app Next generation MEMS devices are aiming for the smartphone Still the obvious high volume market Next generation MEMS startup companies are fabless Investors don t pay to build fabs! Access to high volume 200mm foundries will be key to scaling the business Apple, Samsung, etc. won t tolerate smaller suppliers Page 26
The Dilemma Irresistible force : Exciting new MEMS that cannot be made compatible without impairing device function Immovable object : High volume 200mm foundries cannot jeopardize their main business Page 27
Hard choices ahead for next generation MEMS startups Try to become -compatible ASAP Compromise on performance, function, and/or cost Use smaller MEMS-specific fabs first, then find some way to transfer the fabs Multiple fab transfers Risks of re-engineering for compatibility Slower time to market Could these MEMS be dead on arrival due to current lack of viable high volume manufacturing path? Page 28
Hard choices ahead for fabs Stick to -compatible Miss out on the next big opportunities in MEMS Existing MEMS business will become increasingly commoditized as technology ages Build MEMS-specific fabs (Or repurpose existing 200mm semiconductor fabs) Find ways to manage risks of handling non- materials mid-process Can equipment makers help find a solution? Page 29
200mm MEMS-specific foundries are responding STMicroelectronics Foundry Thin film PZT process already in production at Agrate fab Allows glass USound, polight publicly announced collaborations Silex Microsystems Adding thin film PZT Already allows glass STMicro Petra process Oxide Top Electrode PZT Top Electrode PZT Bottom Electrode Sloped sidewalls for good passivation step coverage Bottom Electrode PZT and the metallic electrodes patterned by dry-etch Page 30
Call to Action and Summary Next generation, fabless MEMS companies are seeking development on 200mm to serve high volume markets High volume foundries must consider adopting new processes or miss out on emerging MEMS opportunities The field is starting to move already. How will you find a path forward? Contact: amf@amfitzgerald.com Page 31