Lowering Production Cost of "Big MEMS (and Sensors)" Chip Technologies using Large Area Manufacturing Techniques

www.advanced-energy.com
1
REDUCING PRODUCTION COST OF “BIG MEMS”
CHIP TECHNOLOGIES USING LARGE AREA
MANUFACTURING TECHNIQUES
Dr. Robert G. Andosca
Director, Worldwide Applications Technology
Advanced Materials Processes (AMP)
Advanced Energy Industries Inc.
An Invited Keynote Address – given May 2nd 2019

ABOUT THE SPEAKER
2
Robert G. Andosca, Ph.D.
Director
Worldwide Applications Technology
Advanced Energy Inc.
• 25+ years semiconductor and MEMS / sensor industry experience
– C-level and operations management, business development
• Formed numerous strategic partnerships and strong business relationships
• Wrote $4.4M in awarded government proposals and corporate JDA
– Entrepreneur – Founder and former CEO/CTO, MicroGen Systems Inc.
• Piezo-MEMS based vibration energy harvester products to power
various automotive and industrial Internet of Things (IoT) sensor modules
• Raised $8M in strategic corporate and angel venture investment
– Scientist and engineer
• Ph.D. / M.S. Materials Science (EE, ME & Physics), The University of Vermont
• Design – Semiconductor IC, MEMS, sensors and photovoltaics
• Process specialty – PVD, PECVD and etch (e.g. DRIE) of various thin films
– 12 publications, 25 issued US and international patents (another 11 pending)
– Invited speaker worldwide (28X)
• IoT, energy harvesting and various thin film based technologies

INTERNET OF THINGS (IOT)
MEMS / sensors everywhere!

IOT – MEMS AND SENSORS EVERYWHERE IN EVERYTHING!
4

IOT – TRILLION SENSOR VISION
5
Dr. Janusz Bryzek, sometimes referred to as “The Father
of Sensors” and “The Trillion Sensor Man”, is the Chairman
and CEO of Trillion Sensors Summit.
Co-Founder of NovaSensor (acquired by GE), LV Sensors,
InvenSense (acquired by TDK), Jyve (acquired by Fairchild),
eXo (currently CEO) and several other MEMS companies.
60T / year
in 2035
https://www.eenewsanalog.com/news/janusz-bryzek-trillion-sensor-man-part-1/page/0/1
https://www.eenewsanalog.com/news/janusz-bryzek-trillion-sensor-man-part-2-0

MEMS / SENSORS
My vision for the future of MEMS manufacturing for low
production cost to penetrate markets quickly

WHAT IS MEMS?
7
✓ Micro Electro Mechanical Systems (MEMS)
• Micro-scale dimensions – typical features < 100 microns
• Electrical and Mechanical features
• Systems – internal and external features combine to form a function
✓ Typical fabrication techniques
• Originally only used IC fabrication techniques on Si substrates
• Batch processing is used to lower cost
• More MEMS specific processes (e.g. DRIE) and materials
(e.g. glass and flexible substrates) now in use
Majority of MEMS devices
are < 10 mm2 in size
1-axis MACRO-
accelerometer (early 1990’s)
3-axis MEMS accelerometer
(up to 12-axes today)
✓ MEMS is an enabling technology
• Able to reduce size | macro-scale → micro-scale
(e.g. accelerometers – see pictures)
• Can lower unit cost
• Can have more precise functionality

MEMS – SURFACE -VS- BULK MICROMACHINING
8
Surface micromachining (2D)
✓ Multilayered deposition, patterning, etch and release of
structures only on surface of substrate
✓ No etching of the substrate – serves as foundation only
Bulk micromachining (3D)
✓ Includes surface micromachining
✓ Utilizes substrate as a functional mechanism
(e.g. membrane or spring) via etching
✓ Substrate bonding may be used as well
Mirror (popped up) Gear train
Membrane pressure sensor
KOH etched Si forming
54.7° wall angle – (111) plane
Can be DRIE with 90° walls
for chip packing density
substrate sideview
surface structures
top down view
2nd substrate sideview
substrate sideview bonding
surface structures
backside port
cavity

DR. JANUSZ BRYZEK MEMS MARKET VISION (NEEDS UPDATING)
9

MEMS MARKET EVOLUTION
10
Optical
waveguides / switches
✓ Typically < 10 mm2
size devices
✓ Now many MEMS chips
are becoming larger,
which have difficulty
being miniaturized
BioMEMS

WHAT IS “BIG MEMS”?
11
• Big MEMS – large devices that cannot miniaturize well, yet could benefit from alternative
high volume manufacturing techniques and the resulting lower production costs
✓ External connections (non-electrical) –
• Microfluidic bio-devices – e.g. connections to external tubing
• Optical switching – e.g. connections to fiber optics
✓ Surface area dependency –
• Electrical power generating devices
• Piezoelectric and thermal energy harvesting – e.g. powering IoT wireless sensors!
• Micro-fuel cells (MFC) – e.g. to enable mobile electronics
• Biometric sensors
• Piezoresistive (and piezoelectric) pressure sensors
• Microphones / micro speakers – e.g. mobile phones
Small chip size, but very
high production volume
Flexible Fingerprint Sensor
2D optical cross connect

BIG MEMS – E.G.
12
Solid Oxide micro-Fuel Cell (SOFC)
✓ Seven (7) 4 x 3 cm2 MEMS die stacked and glass-frit bonded
✓ Up to 8-photo mask levels per wafer (average 4 levels per wafer)
✓ Required nano- to micro-layer thickness control with high
uniformity and low stress
✓ Otherwise contains large printed feature dimensions
piezoMEMS Vibration Energy Harvester (pVEH)
✓ Three (3) 1-1.5 cm x 1-1.5 cm MEMS die stacked, including
glass wafer-level-packaging (glass-frit bonded)
✓ 6-photo mask levels excluding wafer-level packaging (WLP)
✓ Required nano- to micro-layer thickness control with high
uniformity and low stress
✓ Otherwise contains large printed feature dimensions

BIG MEMS – E.G.
BioMEMS microfluidic chip
• An automated FISH microfluidic chip,
which integrates a reagent multiplexer, a
cell chamber with a thin-film heater layer,
and a peristaltic pump.
13
Fluorescence in situ hybridization (FISH) is a molecular
cytogenetic technique that uses fluorescent probes that bind
to only those parts of a nucleic acid sequence with a high
degree of sequence complementarity.

BIG MEMS – E.G.
14
• Most optical switching companies went out of business after the Telecom Bubble crash in 2001
• LA manufacturing could revitalize such MEMS optical switch products for telecom today!

BIG MEMS – E.G. BIOMETRIC SENSORS
15
All Fingers Entire Hand
• Example backlit images of individual fingerprints and full hand
• Reflected light is detected by the dpiX digital a-Si TFT & photodiode array
• Images are grey scale (light intensity measured, not color)

IMMEDIATE HIGH VOLUME SMALL MEMS EXAMPLES FOR LA PRODUCTION
Pressure sensors (e.g. automotive TPMS) piezoMEMS microphones (e.g. mobile devices)
16
• 500M TPMS units per year to
be sold in 2025
• Used within Amazon Echo®
✓ Water proof microphone design

www.advanced-energy.com 17
MEMS
IC
• As IC substrate sizes increased MEMS
companies and foundries adopted the older
equipment technology at low cost
Diameter Chip size
+ 200 mm scribe
Chips / wafer Est. chip cost low volume
6-masks, 90% yield
Est. chip cost high volume
6-masks, 90% yield
150 mm 1 x 1 cm2 113 $28 chip only | $49 WLP $3 chip only | $5 WLP
200 mm 1 x 1 cm2 216 $21 chip only | $37 WLP $2 chip only | $4 WLP
ECONOMIES OF SCALE → MEMS MARKET ADOPTION
(adoption year?)
• Now, the IC industry is stalled at 300 mm
• Consequently, MEMS is mired at 200 mm
✓ MEMS market entry/adoption
is being blocked by high unit
cost in low volume

300 370
460
Gen 1 / 2
ECONOMIES OF SCALE – FABRICATE MEMS USING LARGE AREA TECHNIQUES
18
GEN Chip size
+ 200 mm scribe
Chips / substrate Est. chip cost low volume
6-masks, 90% yield
Est. chip cost high volume
6-masks, 90% yield
2.0 1 x 1 cm2 1406 $3.80 chip only | $8.00 WLP $0.38 chip | $0.80 WLP
4.0 1 x 1 cm2 4945 $1.46 chip only | $3.00 WLP $0.15 chip | $0.30 WLP
A single Gen 2 substrate area equivalency –
✓ 6.5 wafers @ 200 mm diameter
A single Gen 4 substrate area equivalency –
✓ 22 wafers @ 200 mm diameter
✓ Market entry is much more
tractable using large area
manufacturing techniques … let
alone the high volume cost-points

SVC TECHCON 2019 – LARGE AREA MEMS & SENSORS MANUFACTURING
19
. . . just imagine using large area manufacturing techniques
to drive down production cost!
At normally 1 US$ per 1 mm2 for each Si die . . .
Optical
waveguide switches
BioMEMS
✓ Could benefit from
LA fabrication
techniques

E.G. – EMERGING IOT TECH THAT IS NEEDING COST REDUCTION (1)
20
"Energy is a challenge. To power trillions of
sensors requires energy and per unit (cost) will
have to be reduced from today's levels. It will
need to be derived from light, vibration,
thermal energy scavengers. Particularly we
need to reduce the energy to power radios by
a factor of 100 to allow them to be powered by
scavenging,“
-- Dr. Janusz Bryzek
https://www.eenewsanalog.com/news/janusz-bryzek-trillion-sensor-man-part-2-0/page/0/1

E.G. – EMERGING IOT TECH THAT NEEDING COST REDUCTION (2)
21
piezoMEMS
vibration energy
harvester for
powering IoT
wireless sensors
• $75-100 → < $10 each
in low → high standard
200 mm diameter
manufacturing volume
(without energy management
electronics)
✓ $3 - 6 each → << $1 each in low → high
Gen 2 - 4 manufacturing volume (without
energy management electronics)

LARGE AREA COATERS AND ETCH TOOLS
Why not use for MEMS/sensor production?!

LARGE AREA GLASS COATERS
23

FLAT PANEL DISPLAYS HAVE EVOLVED IN SIZE AND DENSITY
24

MOBILE DEVICES DRIVING HIGHER RESOLUTION
25
HD TV → 758 pixels/ inch2
8K TV → 12K pixels/ inch2
iPhone X’s → 213K pixels/inch2
Large Screen
LCD-TV
High Resolution
Smart Phone
Pixel Structure

SUBSTRATE SIZE INCREASING FOR MANUFACTURING ECONOMICS
26
3.4 meters
Large area manufacturing
uses similar techniques as
IC processing
✓ 1X and stepper
photolithography
• Spinless resist coating
• Down to 1.2 mm linewidths
✓ DC and RF magnetron
sputtering and PECVD
deposition
• High uniformity
✓ RF plasma etch
• High uniformity
Substrate generations
jumbo

CHALLENGES OF ADOPTING LARGE AREA
MANUFACTURING METHODS FOR MEMS
A side by side comparison

KEY PROCESS STEPS FOR CONVERSION
Process type Sub-process Large Area (LA) LA MEMS comments
PHOTOLITHOGRAPHY Resist coat ✓ Yes ✓ Yes
Align / expose ✓ Yes ✓ Yes
Develop ✓ Yes ✓ Yes
Resist strip ✓ Yes ✓ Yes
Metal liftoff ✓ Yes ✓ Yes
DEPOSITION Evaporation ✓ Yes ✓ Yes
LPCVD conformal coatings
@ 600-1100’C |
No No Temperature issues w/ glass
substrates
Electroplating ✓ Yes ✓ Yes
DC & RF magnetron
sputtering
✓ Yes ✓ Yes Non-high aspect ratio a = etch
depth/width conformal coatings can
be achieved.
PECVD ✓ Yes ✓ Yes Non-high a conformal coatings can
be achieved.
ETCH Wet (e.g. BOE) ✓ Yes ✓ Yes
Dry (e.g. HF and XeF2) ✓ Yes ✓ Yes
Plasma (e.g. ICP) ✓ Yes Yes, except Deep RIE of
glass. Oxide etch rate is
much slower than Si rate.
Through-glass substrate etching will
need work around (e.g. wet chemical
etching, sand blasting, laser).
28
No clear show stoppers

LARGE AREA PHOTOLITHOGRAPHY
EV Group (EVG) – Schärding, Austria

✓ Gen 2-4

LARGE AREA THIN FILM DEPOSITION
Using LA glass coaters and FPD tools

DC & RF MAGNETRON SPUTTERING
35
Cathode / target containing raw
material that is sputtered off by
the positive ions impacts
Anode / substrate where
thin film is deposited
substrate
Balanced
Slightly
unbalanced
Highly
unbalanced
plasma plasma
plasma
plasma
cathode / target
substrate
target

JUMBO GLASS COATERS CAN ACHIEVE NANOMETER LEVEL
DEPOSITION UNIFORMITY → PERFECT FOR MEMS !
36
Source – ULVAC
Power synchronization and
balancing plus superior arc
management across 16
cathodes to increase
uniformity and yield
✓ Advanced Energy is the
world’s leading expert for
plasma power technology!
The aforementioned requires . . .

VARIOUS LARGE AREA COATERS CAN BE USED FOR MEMS
Horizontal coaters Vertical coaters
37

BULK MICROMACHINING – GLASS ETCHING
SPTS – an Orbotech company

Note:
SPTS does
not currently
build large
area
platforms –
up to 300 mm
diameter
round
substrates

SPTS – SILICON VERSUS GLASS ETCHING SUMMARY
• Silicon Deep Reactive Ion Etching (DRIE) etch rates are very fast compared
to glass, yet similar etch profiles can be achieved
– >10 mm/min versus 0.3-0.8 mm/min today → needs improvement, but not
a show stopper since large area substrate processing will compensate for
3-4X added cost to this etch step
– 73-90 degree etch profiles can be achieved in glass w/ various masks
• Pure quartz and fused silica etch like thermal oxide – lower power, smooth
→ 83-90 degree wall angles
• Pyrex contains impurities – requires higher process power with resulting
rougher surfaces
→ 73-83 degree wall angles
• Large area glass DRIE equipment has to be designed/constructed for MEMS
– Requires market pull it takes just 1 large area Gen X MEMS foundry to get market
traction and compete with 200 mm MEMS foundries (see slides # 40-43)
41

DPIX GEN 4.5 MEMS FOUNDRY
The Future is Now!

GEN 4.5 FPD FOUNDRY – NOW ONLY LARGE AREA MEMS FOUNDRY IN WORLD
• World class cleanroom facility
– Location: Colorado Springs, CO, USA
– Building: 260,000 ft2
– Cleanroom: 65,000 ft2
– Substrate size: single G4.5 plate = (39) 6” wafers
– Single lot: (20) G4.5 plates = (780) 6” wafers
• Volumes
– Prototyping
– Pilot production
– Mass production
• Customer Benefits
– Provide customers a secure IP environment for
technology and product development
– Extensive design engineering expertise
• Open for business → MEMS April 2019 !
43
X-ray photo detector arrays
for medical imaging on
Gen 4.5 glass
X-ray photo detector arrays
for medical imaging on
Gen 4.5 flexible substrate

CORE TECHNOLOGY AND CAPABILITIES
• Core Technology
– Substrates: Gen 4.5 glass 700 mm* thick and flexible PI
– Thin Film Transistors (TFTs): a-Si and IGZO**
– Photodiodes: amorphous-Si and organic
• Testing
– Parametric (test structures)
– Full contact (optical sensor arrays)
• Process Capability
– Photolithography
• Resist coating: Extrusion
• Align / expose: Stepper (2.25 μm feature size)
• Develop: Puddle
– Deposition
• PVD: metals, ITO and IGZO
• PECVD: dielectrics and a-Si
– Etching
• Wet: various, including BOE
• Plasma etch: metals, dielectrics
44
M
et
al
a-Si Photodiode
Glass Substrate
TFT
Polyimide Flex Substrate
TFT
a-Si Photodiode
GATE Line
Diode Bias
Top down view
Pixel (FLEX)
PECVD – AMAT tool shown
(As shown using optional polyimide (PI) and moisture
barrier layers for FLEX substrate)
Cross Section View
* 700 mm is the foundry standard, 500 mm thick is optional ** IGZO = Indium Gallium Zinc Oxide

POTENTIAL MEMS / SENSORS APPLICATIONS TODAY!
Substrate = Flex
Metal Electrode 2
Electrolyte
Metal Electrode 1 Substrate = Glass or Flex
TFT Backplane
Lens Array and Hardcoat
Photodiode
Substrate = Flex
TFT Backplane
Photodiode
LED
Finger
Substrate = Glass or Flex
Functionalized TFT Backplane
ââââ Environmental Species ââââ
Substrate = Glass
TFT Backplane
Photodiode
Species w/ Fluorescent Marker
ââââ Excitation ââââ
✓ Solid state battery ✓ Chem-bio
optical sensor
✓ Biometric sensor
✓ Oximeter (patient
O2 monitoring)
✓ Environmental
sensor
✓ And more
• Pressure sensors
• Energy harvesters
• Optical switches
• Fuel cells

CONCLUSION AND Q&A

47
60T
1
10,000
~60 Trillion IoT sensors per year deployed in 2035
Only 1 large area MEMS foundry today, yet this
number will increase as dpiX obtains market traction
and competes with 200 mm MEMS foundries
Estimated number of coaters/etchers needed to
manufacturer 60T sensors per year (back of
envelope calculation; needs substantiation)
(Below NOT SHOWN, but STATED)

MY FINAL ASSERTION (BELOW NOT SHOWN, BUT STATED)
1. Start by making dpiX’s Gen 4.5 MEMS foundry successful by transferring high volume
products as soon as possible to make a competitive impact on 150/200mm MEMS foundries.
2. Next transfer emerging MEMS products from 150/200mm diameter MEMS foundries to Gen 2
MEMS production, which also serves as a learning platform
✓ Remember, there is still quite a bit of equipment and process engineering learning needed for
this transfer to be successful (it is not a slam dunk) and scaling to even larger area Gen X
substrate sizes
✓ It can be done, because engineers love to solve problems!
3. Establish volume production level products and create market tension with 200 mm foundries
4. As volumes increase and price-points require reduction for IoT applications to become
ubiquitous → create more Gen 4+ MEMS production
✓ Keep Gen 2 as pilot line/ low volume production for emerging technologies
5. Don’t be the “quiet company”, make waves by getting out there and doing product and
promotional marketing!
48

ROBERT G. ANDOSCA, PH.D.
Director, Worldwide Applications Technology
1625 Sharp Point Drive, Fort Collins, CO 80525
+1 (970) 407-6380 office | +1 (970) 829-6107 cell
robert.andosca@aei.com
Precision. Power. Performance.

ABSTRACT
• Since the 1980’s microelectromechanical systems (“MEMS”) based devices have been manufactured
primarily on round silicon (“Si”) substrates. This has been accomplished by primarily riding the “coattails”
of the semiconductor (“SEMI”) integrated circuit chip industry, where Si substrate diameters have grown
from less than 50 mm to 300 mm. As new larger diameter fabrication equipment was needed the previous
generation tools (refurbished) were adopted by the MEMS industry at much lower price points.
• Today, the SEMI industry has stalled at 300 mm, likewise the MEMS industry is mired at 200 mm diameter.
The issue is that many MEMS chip dimensions can be large, greater than 10 x 10 mm2 in area and can
have expensive wafer-level packaging (“WLP”) utilized to protect its moving parts from inexpensive plastic
molded packaging. When considering the $1 per mm2 ‘rule of thumb’ for unyielded chip production cost,
these “Big MEMS” chips are very difficult to fabricate cost effectively for their accompanying product
market adoption.
• Meanwhile over the last two decades of flat panel display (FPD) technology requirements have continued
to increase in complexity and manufacturing capabilities. This includes increasing FPD resolution from
today’s 4K to 8K and glass substrate size up to 3.1 x 3.1 m2, a.k.a. ‘Gen 10 (G10)’ glass. To achieve these
challenging levels many manufacturing obstacles have had to be overcome, such as magnetron sputtering
over large areas, including deposition thickness uniformity and optical property uniformity, the reduction of
yield detractors, such as particles generated due to plasma arcing, and other process challenges.
• What if the MEMS industry wasn’t restricted in substrate size, such as by utilizing G8 (2.1 x 2.4 m2) or
older (smaller area) fabrication equipment? Then, the chip cost could dramatically decrease.
50

SPEAKER BIO
Dr. Robert Andosca (www.linkedin.com/in/randosca) is the Director, Worldwide
Applications Technology focusing on plasma-based deposition and etch of thin films
materials for Advanced Energy (www.advancedenergy.com) headquartered in Fort
Collins, CO. He has 25+ years’ experience in the semiconductor,
microelectromechanical systems (MEMS) and photovoltaic industries.
Dr. Andosca's professional experience ranges from C-level to operational to
engineering management and business development, and has been a scientist and
engineer focusing on many thin film based products. Dr. Andosca is the founder and
former CEO of MicroGen Systems Inc. (www.microgensystems.com), has held
senior level positions at the Smart System Technology & Commercialization Center,
Lilliputian Systems, Umicore, Corning IntelliSense, Clare, Lockheed Martin and
Irvine Sensors, and is an adjunct professor in the Rochester Institute of
Technology’s Mechanical Engineering (www.rit.edu/kgcoe/mechanical) Department.
Dr. Andosca completed his Ph.D. from The University of Vermont (www.uvm.edu) in
Materials Science (multi-disciplinary program between EE, ME and Physics). His
dissertation research was on theoretical and experimental studies of piezoelectric
MEMS-based vibration energy harvester devices and sensors. He also holds an
M.S. in Materials Science from UVM, and B.S. degrees in Mathematics and Physics
from Keene State College. He is an author on twelve (12) published scientific
papers, and is an inventor on twenty-five (25) issued US and international patents
and has another eleven (11) pending. Dr. Andosca has been an invited speaker
worldwide on Internet of Things, energy harvesting and various thin film based
product technologies.
51
Robert G. Andosca, Ph.D.
Director
Worldwide Applications Technology
Advanced Energy Inc.

Lowering Production Cost of "Big MEMS (and Sensors)" Chip Technologies using Large Area Manufacturing Techniques

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Similar to Lowering Production Cost of "Big MEMS (and Sensors)" Chip Technologies using Large Area Manufacturing Techniques

Similar to Lowering Production Cost of "Big MEMS (and Sensors)" Chip Technologies using Large Area Manufacturing Techniques (20)

Recently uploaded

Recently uploaded (20)

Lowering Production Cost of "Big MEMS (and Sensors)" Chip Technologies using Large Area Manufacturing Techniques