Evatec LAYERS 4 2018-2019 - MEMS

EDITION 4

SEMICONDUCTOR

MEMS

Solving new challenges every day in Power Devices, MEMS and Wireless EXTRACTS FROM LAYERS 4

CHAPTER

SEMICONDUCTOR Semiconductor – Meet the Team MEMS Industry trends by Yole Développement Vaisala – Technology to keep our airports safe PragmatIC – Reinventing electronics for mass market applications Thin-film integrated passive devices (IPD) AlScN update Wireless Industry trends by Yole Développement Magnetron sputter epitaxy of AlScN Stress improvement for FBAR electrodes Power Devices Industry trends by Yole Développement Nexperia – A vision for 2020 Wide Band Gap technologies by Yole Développement

4

LAYERS 4 | SEMICONDUCTOR | MEET THE TEAM

SEMICONDUCTOR SEMICONDUCTOR BUSINESS UNIT SEMICONDUCTOR NEW CHALLENGES EVERY DAY! SEMICONDUCTOR Power Devices SEMICONDUCTOR Wireless Power Devices MEMS MEMS Memory Wireless Memory

Our Semiconductor chapter covers technology and applications across Power Devices, MEMS and Wireless – from the latest thin film integrated magnetic passive devices for RF applications to high performance sensor technologies that keep our airports operational 24/7.

Silvan

Silvan Wuethrich Head of BU Semiconductor Silvan qualified as an engineer for system technology in Switzerland, specialising in micro and nanotechnology. He joined Evatec as an application engineer on his graduation in 2006 before holding positions in both project and product management for the BAK Box Coater.

He was responsible for all Evatec’s Batch Systems business before becoming Head of BU Semiconductor.

Semiconductor

Maurus Tschirky Senior Product Marketing Manager Maurus is responsible for activities within the MEMS market segment with a particular interest in advanced functional materials such as piezoelectric and magnetic materials. Prior to joining Evatec in 2015, he held a number of positions in the PVD-equipment industry in application, system engineering, project and product management. He holds a Masters in Business Engineering / International Marketing.

Hans Auer Senior Product Marketing Manager With a degree in electrical engineering from Switzerland Hans took his first position in vacuum and thin film technology in the equipment engineering team at Unaxis in 1981. Now with 35 years experience in engineering, product and business management in the semiconductor industry he is responsible for Evatec’s Power Device business.

5

6

LAYERS 4 | SEMICONDUCTOR | MEMS | INDUSTRY TRENDS

INDUSTRY TRENDS: MEMS The future of MEMS1 and Sensors: beyond the human senses!

2017 has been quite good year for the MEMS markets and although the MEMS industry reached maturity, it is still expected to grow at a significant rate: 18% in value and 27% in units, over 2018-23. In 2023, the MEMS market should be a US$31 billion market with 88 billion units.

Moreover, with new mega trends such as robotic cars, autonomous vehicles, AI2, AR/VR3, 5G, and Industry 4.0 … the demand for sensors will grow as for MEMS. It is still a domain with a lot of innovation as new devices are in R&D (speakers, gas sensors, hyperspectral imagers …). This wave of innovation is also confirmed by the good 2017 business year realised by most of the MEMS foundries. This business is highly dynamic, as shown by the shuffle of the MEMS players ranking in 2017/2016 where RF4 MEMS players are moving to higher ranks.

Over the years, sensors have shifted from detectors to awareness sensors.

In the 1970s, sensors were first developed and used for physical sensing: shock, pressure, and then acceleration and rotation. As more effort was been put into R&D, the use of MEMS shifted from physical sensors to light management (e.g. micro mirrors), and then to uncooled infra-red sensing (e.g. microbolometers). It opened the way to the first sensor that can sense beyond human senses.

After physical/light sensing, MEMS development has been driven by sound, with microphones. Nowadays,

MEMS and sensor developments are aiming to go far beyond human capability with sensing in ultra-sonic, hyperspectral and radio-frequency.

We can imagine a next generation of sensors that can be used for emotion/ empathy sensing in the long term.

Over the years, my MEMS industry experience made me realised that the MEMS business has moved through three different eras: 1. The “detection era” in the very first years (simple sensors to detect a shock, a level) 2. The “measuring era” when sensors could not only sense and detect but also measure (e.g. a rotation) 3. The “global perception awareness era” when sensors are increasingly used for a mapping of the environment (e.g. 3D with LiDAR, air quality with environmental sensors, gesture recognition and biometry). This is possible thanks to sensor fusion of multiple parameters together with artificial intelligence.

This trend has been possible thanks to the implementation of numerous technological breakthroughs over the years: new sensor designs, new processes and materials, new integration approaches, new packaging, sensor fusion and new detection principles.

1.

2.

3.

4.

MEMS: Micro Electro Mechanical Systems AI: Artificial Intelligence AR/VR: Augmented Reality/Virtual Reality RF: Radio Frequency

With almost 20 years of experience in MEMS, Sensors and Photonics applications, markets, and technology analyses, Dr. Eric Mounier provides deep industry insight into current and future trends.

As a Fellow Analyst, Technology & Market, MEMS & Photonics, in the Photonics, Sensing & Display division, he is a daily contributor to the development of MEMS and Photonics activities at Yole Développement, with a large collection of market and technology reports as well as multiple custom consulting projects: business strategy, identification of investments or acquisition targets, due diligences (buy/sell side), market and technology analysis, cost modelling, technology scouting, etc.

Semiconductor

2017 – 2023 MEMS market forecast by segment Source: Status of the MEMS Industry report, Yole Développement, 2018

Year 2023 – US$ 31 billion

Automotive US$ 3.3B

Medical US$ 1.2B

Telecom US$ 600M

Industrial US$ 2.8B

Aeronautics Defence US$ 100M US$ 700M

Consumer US$ 22B

CAGR = 17.4% (Compound Annual Growth Rate)

Year 2017 – US$ 12 billion Telecom US$ 71M

Aeronautics US$ 67M

Industrial US$ 1.15B

Medical US$ 700M

Defence US$ 500M

Automotive US$ 2.6B

Consumer US$ 6.6B

How do we perceive the external world?

Source: Status of the MEMS Industry report, Yole Développement, 2018

Smell 4%

Touch 1%

Audio is the next innovation!

Taste 1%

Psychological study in 1994 by Hatwell (Hatwell, Y., 1994,. Traité de psychologie experimentale.

Paris, P.U.F.) showed that 83% of our external world perception is through our vision, followed by hearing which represents 11% of our perception.

Hearing 11% Vision 83%

Thus, a high quality image is greatly valued for the user. Today the smartphone bill of materials for camera modules is $10 per unit.

The next most-used sense is hearing. We believe the next innovation in MEMS and sensors will be audio for sound and voice control.

Gas sensors could quickly follow microphones as valued functions for consumer applications.

7

8

LAYERS 4 | SEMICONDUCTOR | MEMS | KEEPING OUR AIRPORTS SAFE

Semiconductor

OBSERVATIONS FOR A SAFER WORLD 38 million flights every year from 40,000 airports around the globe, carrying 2.5 million passengers every day and all relying on safe procedures to get them in and out of the air. Offering Manager Kari Luukkonen from Vaisala explains how Vaisala technology does just that at airports with some of the most challenging conditions in the world.

9

10

LAYERS 4 | SEMICONDUCTOR | MEMS | KEEPING OUR AIRPORTS SAFE

A world full of sensors

Sensor manufacturing challenges

Vaisala's history may have begun over 80 years with commercialisation of a single “radiosonde” product, designed to measure temperature, humidity, pressure, wind speed and direction in the upper atmosphere but today the company produces over 6,000 products. From high end humidity and carbon dioxide measurement for demanding industrial applications, to air quality measurement that protects our citizens, all these sensor technologies help us to better understand and influence our environment. There is no better example of where sensing takes a centre stage than at airports around the world. Over the last 40 years Vaisala has installed weather systems at airports in over 100 countries.

The accuracy of measurements is the key in our instruments and systems and the sensors at their heart are critical. Vaisala sensors are designed for harsh and extreme conditions and customers’ expectations are that performance of the product is stable in all conditions.

To ensure repeatable, accurate sensor performance in demanding conditions, our sensor manufacturing processes and our manufacturing equipment must also be reliable.

Keeping our airports operational We all expect our airports to remain operational whatever the weather and that demands a whole series of measurements. Fig 1. illustrates some typical measurements and the principles employed.

Demanding Heights and Low Temperatures Located well over 4 km above sea level, Daocheng Yading Airport in China is the highest civil airport in the world. The challenges posed by its altitude and low temperatures even on a runway over 4000m in length make it unprecedentedly difficult for aircraft to approach and land at the airport. Aircraft performance in extreme conditions is reduced and its essential to measure the airport’s outline weather parameters, such as wind, pressure, temperature and humidity, as well as specialised aviational parameters like runway visual range. Winds can be much stronger at high altitude making it crucial to measure their strength and impact on airport operations. Accurate measurement of clouds height is critical as they can be very close to the ground while reliable measurement of atmospheric pressure in thin air conditions is needed to understand any impact on engine performance.

To do their job properly, the sensors must be protected against windblown particles such as sand which could otherwise affect the measurement accuracy in the windy conditions and the sensor technologies themselves must also be suitable for operation in varying climates from the tropics to the arctic or where temperature shift between day and nighttime is large. In the case of Daocheng Yading we use high-power heaters that compensate for the effects of the cold climate as well as the accumulation of snow.

The goal for our sensor manufacturing equipment is to keep the overall system uptime high and maintenance low. All sensors have thin film technology at their heart and we use custom thin film processes and substrate handling in our evaporation and sputter systems. The challenges are to maximise process yields and in handling peaking demand across the many sensor types in our portfolio.

Vaisala – at an airport near you Next time you are at an airport taxing for take off in windy, low visibility or other harsh weather conditions I hope you will take a look out of the window. It may well be Vaisala technology that’s keeping you safe.

About Vaisala We deliver products and services for environmental and industrial measurement from our headquaters based in Helsinki, Finland. Our company employs approximately 1,600 people and exports 98% of its production to over 150 countries. Innovation and the desire to meet challenges are at Vaisala’s core.

To do this the company spends 12% of its annual net sales revenue in R&D. To find our more go to www.vaisala.com

Semiconductor

Measurement Barometric pressure

Relevance for aviation Aircraft altitude reading

Measurement principle PTB330 uses micromechanical BAROCAP sensor that uses dimensional changes in its silicon membrane to measure pressure.

As the surrounding pressure increases or decreases, the membrane bends, causing changes in the capacitance of the sensor. thereby increasing or decreasing the height of the vacuum gap inside the sensor. The opposite sides of the vacuum gap act as electrodes, and as the distance between the two electrodes changes, the sensor capacitance changes.

The capacitance is measured and converted into a pressure reading.

Temperature & humidity

Aircraft take-off/landing speed Aircraft maximum loading weight

Using HMP155, air temperature is measured by a platinum type (PT100) sensor and relative humidity is measured by a thin film type sensor HUMICAP®180R(C). Changes in humidity is detected by a change of capacitance in the polymer layer of the sensor.

Windspeed & direction

Aircraft take-off/landing ground speed (headwind) Landing safety (crosswind)

WMT700 series uses ultrasound to determine horizontal wind speed and direction. The measurement is based on transit time, the time it takes for the ultrasound to travel from one transducer to another, depending on the wind speed. The transit time is measured in both directions for a pair of transducer heads. Using two measurements for each of the three ultrasonic paths at 60° angles to each other, WMT700 computes the wind speed and direction.

Cloud height

Pilots ability to see airport (vertical) Approach method: Visual (VFR) or instrument (IFR)

The CL31 employs pulsed diode laser LIDAR technology, where short and powerful laser pulses are sent out in a vertical or near-vertical direction. The reflection of light, backscatter, caused by clouds, precipitation etc. is measured as the laser pulses traverse the sky. The resulting backscatter profile, i.e. signal versus height, is stored and processed and the cloud bases are detected.

Visibility & RVR

Pilots ability to see airport (horizontal) Approach method: Visual (VFR) or instrument (IFR)

FS11 transmits pulses of infrared light and detects the light scattered by airborne particles. The intensity of the received pulses is measured and converted to Meteorological Optical Range (MOR) using algorithms of FS11 sensor.

Lightning

Lightning damage, Microbursts / Wind Shear, severe turbulence Ground personnel safety, no refuelling

TSS928 detects optical, magnetic, and electrostatic pulses from lightning events to report cloud and cloud-to-ground lightning within 30 nautical miles (56 km).

Fig 1

“Thin film technology is at the heart of each and every sensor we make.”

11

12

LAYERS 4 | SEMICONDUCTOR | MEMS | A NEW APPROACH TO ELECTRONICS

ENABLING TRILLIONS OF SMART OBJECTS

PragmatIC is reinventing electronics for mass market applications. Its unique technology platform delivers ultralow cost flexible integrated circuits (FlexICs) thinner than a human hair that can easily be embedded into everyday objects.

Revolutionising Near Field Communications (NFC) using PragmatIC’s unique technology looks set to change the shopping experience for millions of consumers in the years to come. PragmatIC’s CTO Richard Price tells us how.

Semiconductor

In Store

At Home

End of Life

Product information

Product information

Product information

- Authentication - Ratings / reviews / testimonials - Product contents - Product story (source-to-shelf) - Multiple languages - Complementary products - Brand story

- Product tutorials - Extended product information - User manual - FAQs - Assembly instructions

- Add ratings / reviews - Product recommendations

Promotions - Discount coupons - Loyalty points

Product engagement - Product registration - Direct feedback to brand - Connect with other product users (social media)

The power of tagging and tapping Embedding NFC enabled tags into supermarket products offers brands the opportunity to transform everyday items into key marketing assets, influencing purchasing behaviour and building direct connections with customers in an ever more competitive landscape. Consumers can benefit from interactive personalised experiences, with access to product information before, during and after purchase, simply by tapping their NFC enabled smartphones.

Re-order - Online - Where to buy - Automatic stock check

Disposal - Recycling information - Automated waste sorting

The solution for mass market tagging is now here Around 80% of smartphones are already NFC enabled and current estimates suggest there will be a total installed base of around 3.5 billion capable handsets by 2019. That means there’s plenty of consumers ready to use the technology.

However, so far campaigns have been limited to premium products or special promotions. Although conventional tagging technology based on classic silicon integrated circuits (ICs) has indeed achieved cost reductions of over 50% in the last 10 years, at around 10¢ per tag the costs still remain a factor of 10 too high for brands in the Fast Moving Consumer Goods (FMCG) sector. Plus, from a practical perspective, the “hard” silicon-based tags can be susceptible to impact damage and in some cases require compromise on the packaging itself. That’s where PragmatIC’s FlexIC technology now steps in.

Tags based on FlexICs are ultra-thin and flexible.

Unlike silicon-based tags they can be integrated invisibly on any type of packaging and unlike other approaches like barcodes and QR codes they are also smaller, can be integrated behind labels and packaging without affecting brand identity and offer much greater functionality. Most importantly of all though, they offer the potential to achieve that all important 1¢ manufacturing cost target essential for the FMCG industry.

13

14

LAYERS 4 | SEMICONDUCTOR | MEMS | A NEW APPROACH TO ELECTRONICS

Bringing the technology to market PragmatIC has developed FlexLogIC™ - its own complete “fab-in-a-box” manufacturing line for its customers to produce FlexICs economically at high volumes. This is a fully automated system which PragmatIC can deliver, install and set up, even at manufacturing locations such as label and packaging facilities with little or no experience in the semiconductor industry. The detailed material recipes, end-to-end process flow, in-line quality monitoring and feedback control loops are implemented within the equipment and software design to ensure reliable production without operator intervention.

All the technologies required, like PVD and lithography are implemented in a self-contained clean environment. Capital investment is 100 to 1000

times less than for a conventional silicon IC fab and set up time is typically 6 months. That compares with a typical silicon IC fab calling for billions of dollars in upfront investment and a 2 year lead time.

The upfront design costs for FlexICs are also considerably lower, so new flexible electronic solutions can be developed, tested and rolled out in weeks rather than months, and the end products brought to market much more quickly. The production cycle is less than a day. This reduction in both cost and time lends itself well to a scalable distributed production model. Compared with silicon ICs, where a few huge foundry companies produce most of the world’s supply, the FlexLogIC model can support a large number of global manufacturers, across multiple locations, each one closer to the original product manufacturer.

Semiconductor

“A trillion objects by 2025” From pilot line to mass production

The future

The roadmap is clear. PragmatIC is already demonstrating its capabilities with customer trials through its own FlexLogIC “fab-in-a-box” production line at a state-of-the-art production facility in the UK. Evatec have been a technology partner from the early stages of our thin film process development. The Evatec CLUSTERLINE® with its inbuilt flexibility provides PragmatIC with a tool that not only supports the initial FlexIC products but also enables further evolution of PragmatIC’s technology platform.

NFC tags in consumer products are just the start.

PragmatIC sees lots of potential in areas such as authentication and even medical sensing.

Automotive technology is also a huge area where FlexIC technology could make its mark. RFID technology is already in widespread use, especially where the products have a relatively high value, but the much lower cost of FlexICs opens up many more possibilities for low value high volume products. The idea of trillions of smart objects could really become a reality.

The production line has capacity for 1 billion ICs per year and future lines will have even higher capacity.

Overall PragmatIC expects to see a trillion smart products in the market by 2025.

Just watch this space!

About PragmatIC PragmatIC is headquartered in Cambridge, UK, with a new billion-unit production facility in NETPark, Sedgefield.

For more information visit www.pragmatic.tech

15

16

LAYERS 4 | SEMICONDUCTOR | MEMS | THIN-FILM INTEGRATED PASSIVE DEVICES (IPD)

BREAKTHROUGH IN THIN FILM BASED INTEGRATED MAGNETIC PASSIVE DEVICES FOR RF APPLICATIONS The integration of on-chip passive devices (e.g. inductors and transformers) with magnetic materials into silicon technology has been for decades a major challenge in the move towards monolithic solutions for wireless communications, RF integrated circuits, power delivery and management, and EMI noise reduction. Senior scientist Dr. Claudiu Valentin Falub explains how the LLS EVO II allows engineering of superior soft magnetic multilayers that ultimately led to ultra-low profile integrated magnetic solenoid micro-inductors with record inductance density and quality factor.

Semiconductor

Thin-film integrated passive devices (IPD) Integrated Passive Devices (IPDs) are attracting an increasing interest due to constant needs for lighter, smaller, faster, “smarter”, and more economical and sophisticated mobile devices. IPDs are multiple passive components sharing a substrate and a package, which can be designed as flip-chip mountable or wire bondable components, and are generally fabricated on silicon, silicon-oninsulator (SOI), GaAs, sapphire, or glass substrates using standard wafer fabrication technologies, such as thin film and photolithography processing. A variety of functional blocks, e.g. impedance matching

circuits, harmonic filters, couplers and baluns, power combiner/divider, etc. can be realised by IPD technology (see Fig. 1).

Inductors and transformers are the passive electrical components that can store energy in the magnetic field created by the current passing through them, and form together with the resistors and capacitors the building blocks of the IPDs. Depending on the final application, there are planar (2D) and 3D inductor designs. The miniaturisation of these components has, however, been a major challenge for decades. Hence, integrated thin film magnetic cores with high magnetic permeability (µr) were proposed, since the inductance

(L) of magnetic-core inductors (and corresponding solenoids associated with magnetic-core microtransformers) is proportional to µr.

Subsequently, Intel demonstrated that planar inductors with magnetic cores can be integrated with 130 nm and 90 nm CMOS processes [1,2]. More recently, magnetic-core 3D inductors were fabricated using CMOS manufacturing equipment and process by Ferric and TSMC [3].

The figure-of-merit (i.e. quality of the inductor) is given by the quality factor (Q), which is a dimensionless number defined by Q = 2πf×L / (Rdc + Rac + Rd), where f is the frequency, Rdc is the DC winding resistance (winding loss) that scales with the wire cross-section

Fig. 1: Applications of the thin-film integrated passive devices (IPD). Left: IPD technologies for telecommunications, e.g. RF, digital and mixed signal devices, and ESD/EMI protection, which are usually realised on 8” or smaller substrates. Some typical IPD devices found in a mobile phone are depicted by yellow circles. Right: complex system-onchip (SOC) technologies, e.g. RF CMOS and on-chip power converters, which have to be realised on 12” wafers since they need to be compatible with Si CMOS technology.

Air-core spiral coils depicted by yellow circles occupy a substantial chip area, illustrating the need to shrink these components for the next generation mobile devices.

Fig. 2: Ferromagnetic core allows for a much higher inductance density, but the quality factor is reduced due to various losses associated with the magnetic material.

The inset in the low left corner shows the equivalent circuit of a “real” inductor, where L is the inductance, R is the series resistance associated with various losses, and C is the turn-to-turn and turn-to-core distributed capacitance. The inset in the top right corner indicates that inductor’s maximum quality factor can be increased by either lowering the series resistance (blue curve), or by increasing the resonance frequency (green curve).

17

18

LAYERS 4 | SEMICONDUCTOR | MEMS | THIN-FILM INTEGRATED PASSIVE DEVICES (IPD)

and total length, and the specific resistivity of wire material (e.g. copper, gold), Rac is the resistance associated with the core loss (eddy currents and hysteresis) and skin effect, and Rd is the resistance associated with the dielectric losses caused by the capacitance of the coil turns with the wire acting as dielectric.

Typically, Q-factor is plotted against frequency (f), as shown in the inset of Fig. 2. For a given inductor size and core permeability, the Q-factor curves converge on the low frequency of the curve, where the losses are primarily determined by the DC resistance of the wires, and the frequency dependence of Q-factor is almost linear. By increasing the frequency, the Q-factor curves start to diverge and reach a peak (Q = Qmax) at a frequency where the copper and magnetic-core losses are equal. For a chosen core material the frequency at which this peak occurs is inversely proportional with the core size. Thus, for larger cores Q-factor reaches a maximum at lower frequency; moreover, larger cores have higher peak Q-factor than smaller cores.

Beyond this region the magnetic-core losses prevail and Q-factor drops rapidly.

A high Q-factor is the inductor’s most desired feature, and hence in the design and manufacturing process the series resistance (R) and distributed capacitance (C) should be as low as possible (see Fig. 2). Since, depending on the inductor design used (2D or 3D) one can increase Q by making the windings larger and using thicker metal layers (i.e. longer and thicker wires) at the expense of size, and using lower resistivity metals (e.g. Cu or Au instead of Al). Another method of improving Q-factor is to increase the resonance frequency of the inductor (see Fig. 2), which can be realised by increasing the spacing between the turns of the inductor (i.e. lower C) also at the expense of size, lowering the dielectric constant of the material between inductor’s windings, using higher resistivity substrates and thick oxide layers between the substrate and metallic layers, and increasing the ferromagnetic resonance frequency of the magnetic-core material. The latter can be realised by selecting materials with high saturation magnetisation (Ms) and by increasing the anisotropy field (Hk) [4].

Fig. 3: Schematic diagrams of the LLS EVO II batch sputter system with 5 process modules that can operate, one or more at a time, in order to fabricate thin films based on single or multiple materials. To induce the inplane magnetic anisotropy in the sputtered thin films, aligning filed systems can be mounted in the cage housing (i.e. outside vacuum) in front of each magnetic target.

Fig. 4: Schematics of an integrated 3D inductor for on-chip RF applications, the magnetic core of which consists of a sputtered multilayer based on a low-loss soft magnetic material. The inset shows a cross-sectional transmission electron microscopy (TEM) analysis of a CoTaZr/Al2O3 soft magnetic multilayer sputtered on 8” Si/200nm-SiO2 wafer. To lower the hysteresis and eddy current losses the ~80 nm thick CoTaZr layers are laminated with 4 nm thick Al2O3 dielectric interlayers.

Semiconductor

FIG. 5: a) Q-factor vs. frequency for a 100 µm x 400 µm magneticcore solenoidal inductor depicted in the optical micrograph. b) Peak Q-factor vs. inductance density of integrated inductors on Si substrates from published on-chip inductor measurements (adapted from Reference [2]). The colors represent the frequency of the peak Q-factor.

Soft magnetic thin films at LLS EVO II The vertical batch sputter system LLS EVO II (see Fig. 3) is a very versatile economical tool for depositing micrometer thick soft magnetic on substrates up to 200 x 230 mm.

This system has several knobs for tuning the in-plane anisotropy of the sputtered soft magnetic layers.

Thus, the performance of the magnetic cores can be tailored by appropriate choice of the magnetic material (saturation magnetization, electrical resistivity) and dielectric interlayer (dielectric constant) [5].

Further tailoring of the soft magnetic multilayer properties can be done by tuning the process parameters (e.g. pressure, power, deposition temperature, angular distribution) [6]. Last but not least, since in the LLS EVO II system the substrate cage rotates continuously during deposition, so that the substrates face different targets alternatively, each ferromagnetic sublayer in the multilayer stack may consist of a fine structure comprising alternating nanolayers with very sharp interfaces.

Adjusting the thickness of these individual nanolayers by changing the cage rotation speed and the power applied to each cathode,

allows to engineer new, composite ferromagnetic materials [4,7,8].

Integrated passive devices with record quality factor Ultra-low profile integrated magnetic solenoid inductors and transformers were fabricated at CEA Leti on 200 mm high-resistivity silicon wafers with back-end-of-line (BEOL) process [9]. Bottom and top conductors were formed by electroplating with 10 µm thick copper, 5 µm line width, and 5 µm spacing between the lines. In order to have a good insulation and to reduce topology, the lines were embedded into a thick polymer.

Then, the magnetic film consisting of a multilayer with alternating 80 nm thick CoZrTa amorphous soft magnetic layers and 4 nm thick Al2O3 dielectric interlayers were deposited by dynamic sputtering under a linear magnetic field using a LLS EVO II system (see Fig. 4).

Finally, the wafers were grinded to reduce the silicon thickness down to 100 µm. By varying the core size, i.e. width (l) and length (L), and the winding pitch, a record inductance surface density of 3500 nH x mm2 in the 1 MHz to 3 GHz frequency range with a record peak Q-factor

of 23 (see Fig. 5). This breakthrough sets a new benchmark of quality for cost-effective manufacturing of soft magnetic multilayers on silicon, and can therefore help manufacturers meet exacting standards for next generation thin film based integrated passive devices.

REFERENCES [1] D.S. Gardner, G. Schrom, P. Hazucha, F. Paillet, T. Karnik, S. Borkar, Integrated On-Chip Inductors with Magnetic Films, IEEE Trans. Magn. 43, pp. 2615-2617 (2007).

[2] D.S. Gardner, G. Schrom, F. Paillet, B. Jamieson, T. Karnik, S. Borkar, Review of On-Chip Inductor Structures with Magnetic Films, IEEE Trans. Magn. 45, pp. 4760-4766 (2009).

[3] N. Sturcken, R. Davies, H. Wu, M. Lekas, K. Shepard, K.W. Cheng, C.C. Chen, Y.S. Su, C.Y. Tsai, K.D. Wu, J.Y. Wu, Y.C. Wang, K.C. Liu, C.C. Hsu, C.L. Chang, W.C. Hua, A. Kalnitsky, Magnetic thin-film inductors for monolithic integration with CMOS, Proc. of the IEEE Int. Electron Devices Meeting (IEDM), 11.4.1-4 (2015).

[4] C.V. Falub, Innovate the soft magnetics for tomorrowʼs RF passive devices, LAYERS, vol. 3, pp. 50-55 (2017).

[5] C.V. Falub, R. Hida, M. Meduňa, J. Zweck, J.P. Michel, H. Sibuet, D. Schneider, M. Bless, J.H. Richter, H. Rohrmann, Structural and ferromagnetic properties of sputtered FeCoB/AlN soft magnetic multilayers for GHz applications, IEEE Trans. Magn. 53, pp. 202906/1-6 (2017).

[6] C. V. Falub, H. Rohrmann, M. Bless, M. Meduňa, M. Marioni, D. Schneider, J. Richter, M. Padrun, Tailoring the soft magnetic properties of sputtered multilayers by microstructure engineering for high frequency applications, AIP Advances 7, pp. 056414/1-7 (2017).

[7] C.V. Falub, M. Bless, R. Hida, M. Meduňa, Innovative soft magnetic multilayers with enhanced in-plane anisotropy and ferromagnetic resonance frequency for integrated RF passive devices, AIP Advances 8, pp.

048002/1-14 (2018).

[8] R. Hida, C.V. Falub, S. Perraudeau, C. Morin, S. Favier, Y. Mazel, Z. Saghi, J.P. Michel, “Nanolaminated FeCoB/ FeCo and FeCoB/NiFe soft magnetic thin films with tailored magnetic properties deposited by magnetron sputtering”, J. Magn. Magn. Mater. 453, pp. 211-219 (2018).

[9] J.-P. Michel, H. Sibuet, N. Buffet, J.-C. Bastien, R. Hida, C. Billard, B. Viala, P. Poveda, A.-S. Berneux-Dugast, C.V. Falub, Ultra-low Profile Integrated Magnetic Inductors and Transformers for HF Applications with IEEE Trans Magn. (submitted).

19

LAYERS 4 | SEMICONDUCTOR | MEMS | AlScN UPDATE

20

CLUSTERLINE® 200 II IS READY

LAYERS UPDATE

MASS PRODUCTION OF HIGH SCANDIUM CONCENTRATION Al(1-x)Sc(x)N FILMS!

Semiconductor

Evatec’s Dr. Bernd Heinz explains the latest step forward in delivering an economical mass production solution for high performance piezoelectric films.

Why AlScN?

The challenge

Its strongly enhanced piezoelectric response [1] makes aluminum scandium nitride (AlScxN) a very promising candidate for use in next generation RF filter devices, microphones and speakers, energy harvesting devices, piezoelectric micro-machined ultrasonic transducers and many other sensors and actuators. The industrial use of AlScxN requires a reliable deposition technology to control the growth of films in the correct (002) textured wurtzite structure within tight specifications regarding uniformity and repeatability. Further, the process has to be mastered on multiple substrate and electrode materials due to the wide variety of possible applications.

In a previous edition of LAYERS, we reported AlScxN films grown using Evatec Multisource Technology – a unique solution for deposition of AlScxN films by co-sputter of metallic aluminum and scandium targets.

It enables the deposition of films with any desired Sc concentration independent of the availability of AlScx compound targets.

For single target deposition of high uniformity AlScx compound films, targets with the size of 300mm diameter are required to coat 200mm substrates uniformly.

Until now such targets have only been available with a limited Sc concentration below 10at%. Quite recently, target manufacturers succeeded in providing the first 300mm prototype targets to Evatec with a high scandium concentration.

This article will report on AlScxN films sputtered from 300mm single compound targets with a nominal Sc concentration of 30at%. The obvious advantages of using larger compound targets instead of co-sputtering technology is the increased productivity (by a factor of 5) and significantly better film uniformity – in particular thickness and stress.

Please note that AlScN technology is subject to a number of patents and/or patent applications. Evatec recommends that customers check the patent situation within their field of business carefully and secure licenses if necessary prior to the start of their commercial production.

21

22

LAYERS 4 | SEMICONDUCTOR | MEMS | AlScN UPDATE

Results Standard AlN deposition technology available on the Evatec CLUSTERLINE® 200 II single wafer production tool was used for the AlScxN trials. Films were deposited on 200mm substrates with a film growth rate of about 1nm/s at a temperature of 300°C. The Sc concentration in the films was verified by EDX analysis.

It turned out that the in-film Sc concentration is slightly higher than the nominal target composition.

Depending on the specifics of the available targets, Sc concentration values of 31at% to 34at% were measured in the films.

The piezoelectric response for an AlSc31N film sputtered from

a 300mm compound target is highlighted in Fig. 1 where it is compared with selected films made by different methods on the Evatec CLUSTERLINE® 200 II. All values were determined using a Piezotest PM300 tester. Each data point represents the average value of d33 measured at different positions on each substrate while the bars indicate the variation of the piezoelectric response within one substrate. The increase of the d33 values with increasing scandium content is in good agreement with the piezoelectric response reported

by Akiyama [1]. The value of 16.1 pC/N for the single target sputtered film with a composition of AlSc31N

confirms our ability to match the Akiyama results even for higher Sc concentrations with a film grown under production relevant conditions.

To enable production with high yield the control of the film stress uniformity across the 200mm substrate is key. Any larger variation in the stress value will directly affect device properties as the electromechanical coupling coefficient kt2 of AlScxN films depend directly on the film stress. With the newly developed AlN process hardware film stress uniformity results of better ± 100MPa across a 200mm wafer can be achieved, as demonstrated in Fig.2.

Fig. 1: Measured d33 values for AlScxN with a Sc concentration between 0 at% and 39 at%. The result achieved by sputtering the 300mm AlSc30 compound target is highlighted. For comparison, the data from Akiyama [1] is shown.

Fig. 2: Stress uniformity for AlSc31N across a 200mm wafer with an edge exclusion of 7mm.

Results for 3 different process settings are displayed.

Semiconductor

Substrate / Electrode

RC (FWHM)

Pt (111)

1.1º

Si (100)

1.2º

Mo (110)

1.3º

Table 1. XRD rocking curve values (FWHM) measured around the (002) diffraction peak for 1000nm thick AlSc33N films grown on different surfaces.

The XRD rocking curve values (FWHM) measured around the (002) diffraction peak are the commonly used quantity to rate the crystalline quality of AlScxN films. Table 1 summarises the rocking curve values for 1000nm thick AlSc33N films grown on bare and metallised Si substrates. Excellent values below 1.5° were achieved on well textured Pt(111) and Mo(100) electrodes as well as on bare silicon wafers.

The AlSc33N films deposited either on Pt or Mo electrodes or directly on silicon show high crystalline quality in the XRD measurement. However, AFM images (Fig.3) reveal significant differences in surface morphology correlating with the different types of substrate. The appearance of elevated, cone-like grains, presumably formed by miss-oriented AlScxN crystallites which are embedded in a matrix of the preferred (002) oriented crystallites can be observed. AlSc33N films grown on Si and on Mo are affected more severely compared to AlSc33N films grown on Pt. The number of unwanted crystallites is significantly suppressed in the films grown on Pt electrodes. The appearance of these crystallites is a well-known phenomenon, associated with increasing Sc concentration.

The appropriate type and condition of the substrate surface is key to mitigate the appearance of these grains. Deposition at reduced sputter pressure or the introduction of an AlN seed layers are known countermeasures to minimise their number but the deposition of Al(1-x) ScxN films with Sc concentration higher than 30% on other than Pt electrodes remains an ongoing challenge.

Fig. 3: AFM image of 1000nm thick AlSc33N films deposited on on (a) Pt (111) electrode (b) bare silicon and (c) on Mo(110) by sputtering a 300mm compound target. (a, b, c, still missing in the picture) [1] M. Akiyama, K. Kano, and A. Teshigahara, Appl. Phys. Lett. 95, 162107 (2009).

The road ahead Evatec is currently witnessing an increasing demand for new piezoelectric materials. AlScxN is still in pole position amongst the possible candidates and preparation for 5G mobile communication is recognised as one of the major driving forces.

Even though films with scandium concentrations well below 30% will be seen in mass-production initially, Evatec’s capability in mastering the challenges of AlScxN film deposition with higher Sc concentration means we will be well placed for future demands in this industry.

23

24

LAYERS 4 | SEMICONDUCTOR | WIRELESS | INDUSTRY TRENDS

INDUSTRY TRENDS: WIRELESS A bright future of RF technologies RF1 technologies enable wireless connectivity and sensing, which are key functions in any market segment from consumer to automotive.

In the mobile handset market, LTE evolution (LTE-Advanced, LTE-Pro) but also upcoming 5G which was just specified at the end of June 2018 through the 3GPP release 15 bring innovative RF technologies to the market such as carrier aggregation, MIMO, beam forming and dual connectivity in the sub 6 GHz or even millimeter-wave radio link.

The mobile RF Front End market is expected to enjoy a sustainable growth with a CAGR2 of 14% reaching US$35.2 billion in 2023.

This market opportunity translate into fierce competition between the current leader which are Broadcom, Skyworks, Qorvo and Murata and attract giant companies willing to expand from their core activities such as Qualcomm, Intel or HiSilicon.

The RF front-end industry not only involve front-end module companies but also impact surrounding business for foundries, epi house, substrate providers, OSAT3 for packaging, assembly and test and of course equipment providers.

In the automotive market, wireless radar sensing for anti-collision systems is achieving a good market penetration along with other sensors such as imaging, ultrasonic or even LiDAR. This is driven by government policy for road safety improvement which incentivises traditional car makers to embed sensors for car environment monitoring.

Another trend which favours radar implementation in the car is automated drive which will require 360° surveillance in real time. Radar is well suited for object detection in a cost effective way. It is operable in all weather conditions and has a promising technology roadmap to enable object classification and imaging. The radar market is expected to reach US$7.5 billion in 2022 based on a 25% CAGR between 2016 and 2022, at the module level.

A strong ecosystem serves this market with leaders such as Bosch and Continental at the module level, supported by Infineon and NXP at the chip level. Again, witnessing a strong market dynamic, fierce competition is occurring at both module and chip level. e.g. Texas Instruments disrupting current technology with an all integrated chip solution or even a strong Chinese ecosystem building up.

Both examples illustrate the bright future for the RF industry and many more cases including telecommunication infrastructure, AR/VR, connected vehicles, Internet of Things, remote surgery are all expected to contribute greatly to the picture.

1. RF: Radio Frequency 2. CAGR : Compound Annual Growth Rate 3. OSAT : Outsourced Semiconductor Assembly and Test

As a Technology & Market Analyst, specialised in RF devices & technologies within the Power & Wireless division at Yole Développement (Yole), Cédric Malaquin is involved in the development of technology & market reports as well as the production of custom consulting projects. Prior his mission at Yole, Cédric first served Soitec as a process integration engineer during 9 years, then as an electrical characterisation engineer during 6 years. He deeply contributed to FDSOI and RFSOI products characterization.

He has also authored or coauthored three patents and five international publications in the semiconductor field.

Semiconductor

2017 – 2023 RF front-end modules market outlook Source: 5G Impact on RF Front-End and Connectivity for Cell Phones report, Yole Développement, 2018

2023

US$ 35B

2017

US$ 1B

US$ 15B

CAGR +15%

US$ 22.5B CAGR +19%

$ 463M

US$ 7B CAGR +7%

US$ 8B

CAGR +14%

US$ 5B

mm-Wave

$ 423M

$ 246M

US$ 1B

US$ 602M

US$ 3B

CAGR +15%

Total RF components & FEM/PAMiD module manufacturers Filters Antenna tuners Switches PAs LNAs mmW FEM

CAGR +16%

25

26

LAYERS 4 | SEMICONDUCTOR | WIRELESS | SPUTTER EPITAXY

MAGNETRON SPUTTER EPITAXY OF ALUMINUM SCANDIUM NITRIDE (AlScN) THIN FILMS Dr. Agnė Žukauskaitė, group manager in the Epitaxy department at Fraunhofer Institute for Applied Solid State Physics IAF in Freiburg, Germany, is responsible for development of piezoelectric materials and shares the latest progress in sputtered AlScN thin films.

Ever since the discovery of enhanced piezoelectric properties of aluminum scandium nitride (Al1-xScxN, later denoted as AlScN) in 2009, the interest in it as a next generation material for broadband RF-filters in 5G communications is still growing. In addition, it is also very attractive for other applications, where piezoelectric transducers and actuators are required, such as bio-sensing, energy harvesting, or acoustics. Here at Fraunhofer IAF, the focus is on bridging the gap between the material science and device design in order to understand how to best implement, or even to open new horizons for this exciting material. Typically, for piezoelectric resonator applications, e.g. RF filters, highly c-axis oriented AlScN layers are preferred. However, growth of AlScN on silicon – the most common substrate in RF-MEMS applications – leads to textured films, i.e., the grains are c-axis oriented out-of-plane, but randomly oriented in-plane. This gives rise to additional acoustic losses in the fabricated devices, decreasing the quality factor Q. If one can go from textured growth to epitaxial growth (clearly defined in-plane epitaxial relationship between the substrate and the film), the device performance can be further improved through superior material quality.

Magnetron sputter epitaxy Magnetron sputter epitaxy (MSE) is a special type of sputtering process where – under particular conditions and provided that the substrate has a reasonably good lattice-match to the film material – it is possible to achieve epitaxial growth in a similar manner as in molecular beam epitaxy (MBE). In the case of group-III nitrides such as AlN or AlScN one of the most suitable substrates is sapphire, where a 30° rotation between the layer and substrate crystal lattices leads to the lowest possible latticemismatch. Recently, MSE process was successfully employed at Fraunhofer IAF on an Evatec CLUSTERLINE® RAD sputter tool

system using reactive co-sputtering from Al and Sc targets to produce 1 µm-thick, epitaxial, piezoelectric, single-phase Al1-xScxN/Al2O3 with up to x=0.4 (40%) scandium incorporation (Figure 1). The sputtered layer still has columnar microstructure typical for sputtered thin films, but at the same time each of these columnar grains has a defined crystallographic relationship with the substrate. A comparison of XRD pole figures is shown in Figure 2, where, for textured Al1-xScxN/Si films, a continuous ring is observed, while for epitaxial Al1-xScxN/Al2O3 six distinct spots appear instead. Samples showed 0002 reflection rocking curve FWHM values in the range of 0.9° in low-Sc films and up to 1.6° in high-Sc films, Figure 1. X-ray diffraction patterns for phase pure, c-axis oriented Al1-xScxN/Al2O3 thin films with different Sc concentrations denoted as x=0.06, 0.14, 0.17, 0.23, 0.32, and 0.40.

Semiconductor

27

Figure 3.

First surface acoustic wave (SAW) resonators fabricated based on epitaxial Al1-xScxN/Al2O3 layers at Fraunhofer IAF.

[1]: Y. Lu, M. Reusch, N. Kurz, A. Ding, T. Christoph, L. Kirste, V. Lebedev, A. Žukauskaitė, Phys. Status Solidi (2017) 1700559.

[2]: Y. Lu, M. Reusch, N. Kurz, A. Ding, T. Christoph, M. Prescher, L. Kirste, O. Ambacher, and A. Žukauskaitė, APL Materials 6(7), 076105 (2018).

Fraunhofer Institute for Applied Solid State Physics IAF, Tullastraße 72, D-79108 Freiburg Dr. Agnė Žukauskaitė agne.zukauskaite@iaf.fraunhofer.de www.iaf.fraunhofer.de

indicating that while highly c-axis oriented material was achieved, the material becomes more and more distorted, when enough Sc is replacing Al in the wurtzite crystal lattice.

As a first step in AlScN growth process optimization, the misoriented grains in co-sputtered AlScN/Si layers were identified and their density reduced by adjusting the target-tosubstrate distance and Ar:N2 ratio in the process gas, as is described in more detail in [1]. The same process window was used as a starting point for MSE of Al1-xScxN/Al2O3. However, in comparison to conventional reactive sputtering, epitaxial growth of AlScN brings additional challenges, such as higher residual stress in the films that can then lead to cracks.

At Fraunhofer IAF this problem is addressed in two ways. First, by lowering the growth temperature the residual stress generated due to the thermal-expansion mismatch was partially compensated. Second, to make up for the lower temperature

the total process pressure was reduced to increase the mean free path of sputtered species so that upon reaching the substrate surface they have higher kinetic energy and promote the growth of highlycrystalline material [2]. After the successful material optimization, first device performance evaluation was carried out as well. Surface acoustic wave (SAW) resonators were fabricated using AlScN with up to 32% Sc (Figure 3) and immediately showed improved electromechanical coupling as well as better overall device performance in comparison to conventional non-epitaxial layers with the same Sc concentration.

Conclusion To conclude, while it is more challenging to find the optimum process window to achieve the MSE mode, it is already clear that epitaxially-deposited piezoelectric AlScN layers offer a huge advantage over the textured films and should be investigated further.

Figure 2. Comparison of x-ray diffraction pole figures for textured Al0.86Sc0.14N/Si (top) & epitaxial Al0.86Sc0.14N/Al2O3 (bottom) layers.

28

LAYERS 4 | SEMICONDUCTOR | WIRELESS | FBAR ELECTRODES

STRESS IMPROVEMENT FOR FBAR ELECTRODES ON CLUSTERLINE® 200 II Evatec Scientist, Dr. Andrea Mazzalai, explains how know how in deposition of Molybdenum and Ruthenium electrodes with controlled stress now compliments processes for AlScN deposition to bring solutions for full thin film stack production for high performance FBARS on CLUSTERLINE® 200 II.

Within the the development of 5th generation wireless systems (5G) the quest for high performance duplexers is driving the development of the latest FBAR devices with resonant frequencies of several GHz. At this range, relatively small in-wafer deviations of the membrane bow can lead to substantial frequency shifts as well as significant variations of the coupling coefficient. For this reason, the strict requirements in terms of stress uniformity are no longer confined to the piezoelectric layer, but are becoming more and more important also for the electrodes.

The FBAR electrode material has to show a good balance between low specific resistivity and high acoustic impedance in order to minimise the resistive losses and to maximise the fraction of mechanical energy confined in the piezoelectric layer. The large majority of designs therefore employ Molybdenum (Mo); but recently Ruthenium (Ru) is also gaining more and more popularity.

We have therefore concentrated our efforts on bringing our Mo and Ru process solutions towards the same outstanding stress control and uniformity levels as we achieve for Al1-xScxN. Our accumulated know-how and experience from developing the piezo-layers themselves represented a valuable base on which we could further design specific process kits for the deposition of Mo and Ru with enhanced stress uniformities on our CLUSTERLINE® 200 II. Figures 1 and 2 illustrate the significant improvements achieved especially towards the edge of the wafers. With the our latest technology we are now able to offer production solutions for Mo with ±100MPa and Ru with ±150MPa stress range down to 7mm of edge exclusion. This now comes along with thickness uniformities better than 1.5% (1σ) for Mo and 1% (1σ) for Ru.

Semiconductor

Excellent stress and thickness uniformity are not the only features needed in the manufacturing of high performing FBARs: the bottom electrode also serves indeed as a template for the nucleation and growth of the fiber-textured piezoelectric layer on top. In order to achieve highest coupling coefficients and maximal yield, its grains have to grow with the narrowest alignment and the surface must be defect-free. A narrow FWHM of the bottom electrode rocking curve and a low surface roughness in as-deposited electrodes are therefore prerequisite for a good piezoelectric performance.

Figure 1. Mo stress uniformity

With the dedicated process kit we can now deposit Mo electrodes that combine the aforementioned stress and thickness uniformities with rocking curve peaks as narrow as 1.6Âş for a thickness of 300nm and an average roughness Ra of 0.4nm when deposited onto an AlN seed layer. This is the key to excellent crystallinity of the subsequent Al1-xScxN piezoelectric layer!

In combination with the excellent performance of the Al1-xScxN thin film deposition, we can now offer a complete production solution for the full film stack of high performing FBARs. I hope that this short example of our efforts in understanding and mastering the entire chain from vacuum systems to the material science represents an example of Evatec’s focus on PVD of advanced functional materials.

Figure 2. Ru stress uniformity

Figure 3. AFM scan of 300nm thick Mo bottom electrode grown on AlN seed layer.

A closer look at the edge exclusion The market is flooded by numerous values of stress uniformities these days. These indications can only be understood when quoted in conjunction with an edge exclusion value in mm. Many companies for example claim superior stress uniformities but they are often measured out to an exclusion of about 20mm. A short lesson in geometry reveals the importance of precise measurement as far out to the wafer edge as possible: on a 200mm wafer the addition to the useful surface of the annulus defined by 80mm and 93mm radii (20mm and 7mm respectively) represents an increase of the yield of 33%! Equally, the measurement method used should also be declared. This is due to the fact, that the values towards the edges are often extrapolated following fitting procedures which might differ significantly. We therefore recommend that you always check for the measurement conditions, or better compare two wafers from different vendors on the very same instrument.

29

30

LAYERS 4 | SEMICONDUCTOR | POWER DEVICES | INDUSTRY TRENDS

INDUSTRY TRENDS: POWER DEVICES The power electronics market is definitely in excellent health The power electronics industry is enabling mega trends and will reach almost US$41.5 billion in 2023. This market showed impressive general growth in 2018 to achieve US$34.4 billion revenue at the end of the year. To be more precise, the discrete device market and the power IC1 market will grow with a CAGR2 20172023 of 2.7% and 4.6% respectively, whereas the power module market will have a CAGR2017-2023 of almost 8%3.

There are several reasons for this growth, but as confirmed by the 18% increase in 2017 in year-on-year global IGBT4 module sales, the major drive comes from dynamic EV/HEV5 power market6. Currently EV/HEV represents 29% of IGBT modules consumption, while by 2023 Yole Développement estimates it will represent over 43%. A similar situation is found in MOSFETs7 for EV/HEV applications, with a 5.9% market increase in 2017 over 2016. MOSFETs are widely used in various EV/HEV converters, in battery chargers handling roughly 3 to 6 kW, in 48V DC/DC8 converters and in micro-inverters for the stop-start function9. Power ICs for automotive is forecast to reach US$2 billion by 2023 with a CAGR of 6.2% from 2018 to 2023. This is possible not only because of the expected increase in power trains to cater for the increase of EV/HEV sales, but also due to the addition of ADAS10 which is the predominant safety product offered both by luxury automobile makers such as Audi, BMW, Mercedes-Benz, Tesla, etc. and for mass market

passenger vehicles such as those offered by VW, Toyota, Honda, Ford, etc.

For a long time, the semiconductor industry also profited from the growth of the PV11 segment, in part due to an accelerated installation boom in China, but today the status of this industry is different and future growth will be partially dependent on Chinese government subsidies.

However, the growth of the PV market in other geographical regions will also lead to an increasing market for the forecast period. Motor drives is another big segment pushing the growth of the IGBT module market due to aggressive regulation targets.

In fact, Yole Développement forecasts a CAGR of 4.6% for motor drives from 2017 to 2023.

The computing and storage market, including laptops and data servers is the second biggest market for MOSFETs, which is expected to achieve $1.7B by 2023. The network and telecommunications market will get a boost thanks to the arrival of new communications technology such as 5G, with a CAGR20172023 of 7.8%. Yole Développement forecasts Power ICs will benefit from multiple key end markets to deliver a 4.6% CAGR2017-2023, in line with the general trend of the overall semiconductor industry.

New directions in power electronics driven by the new requirements of EV/HEV are just a few examples of how technology is evolving. There are still several issues with cost, product

shortage, integration and reliability, but step by step the supply chain is stabilizing, passive components and drivers are being developed, and automotive qualification is starting.

It is still too early today to say how mainstream module technology will look like in an electric car in 10 years.

However innovations are accelerating the evolution of power electronics, and other industries will no doubt be able to take advantage of these costeffective emerging technologies.

The power electronics industry, which represents a large ecosystem from semiconductor to packaging material suppliers and from passive components to converter system designers, is definitely in excellent health.

1. IC: Integrated Circuit 2. CAGR: Compound Annual Growth Rate 3. Source: Status of the Power Electronics Industry report, Yole Développement, 2018 and Introduction to the power IC market 2018 report, Yole Développement, 2018 4. IGBT: Insulated-Gate Bipolar Transistor 5. EV/HEV: Electric and Hybrid Vehicles 6. Source: IGBT Market and Technology Trends 2017 report, Yole Développement, 2017 7. MOSFET: Metal Oxide Semiconductor FieldEffect Transistor 8. DC: Direct Current 9. Source: Power MOSFET 2017: Market and Technology Trends report, Yole Développement, 2018 10. ADAS: Advanced Driver Assistance Systems 11. PV: Photovoltaic

Semiconductor

2017 - 2023 power electronics driving application evolution Discrete and module IGBTs & MOSFETs markets Source: Status of Power Electronics Industry report, Yole DĂŠveloppement, 2018

2023 US$ 13.2B

2017 US$ 10.4B US$ 1.35B US$ 1.4B

US$ 2.0B

US$ 2.3B

$US 2.68B

CAGR: 4%

US$ 0.7B

US$ 1.51B US$ 1.52B

US$ 2.5B

US$ 3.7B

US$ 3.16B

Computing & storage

Industrial

Home appliance

Others

Consumer [1]

EV / HEV

[1] Consumer segment includes Portable & wireless and Audio & Image applications

Dr. Ana Villamor serves as a Technology & Market Analyst, Power Electronics at Yole DĂŠveloppement. She is involved in many custom studies and reports focused on emerging power electronics technologies, including device technology and reliability analysis (MOSFET, IGBT, HEMT, etc.). In addition, Previously Ana was involved in a high-added value collaboration on SJ Power MOSFETs, within the CNM research center for the leading power electronic company ON Semiconductor. During this partnership, and after two years as Silicon Development Engineer, she acquired extensive relevant technical expertise and a deep knowledge of the power electronics industry. Dr. Villamor is author and co-author of several papers as well as a patent.

US$ 0.8B

31

32

LAYERS 4 | SEMICONDUCTOR | POWER DEVICES | NEXPERIA - A VISION FOR 2020

$2BIL THE ROAD TO

Michael Müller Director Metallisation and Stefan Schwantes, General Manager Waferfab, Nexperia HH talk about the company’s vision for 2020 and the importance of working closely together with key partners like Evatec for successful realisation of the company ramp up plans.

Semiconductor

33

LION Our markets are growing

The combined market for Discretes, Logic and MOSFET products is expected to reach $10 Billion by 2020. Working in a market that’s growing is always great as it brings opportunity for us to grow too, but we know that our competitors will be just as keen to grab that extra business as we are, so having the right strategy is essential to make sure we achieve our goal of being a $2 Billion company by 2020.

34

LAYERS 4 | SEMICONDUCTOR | POWER DEVICES | NEXPERIA - A VISION FOR 2020

Building on Nexperia’s existing values Our focus remains on efficiency, producing consistently reliable semiconductor components at high volume to meet the stringent standards set by the Automotive industry. That means delivering small packages, produced in-house, combining power and thermal efficiency with best-in-class quality levels. Continuing to be adaptable to global mega trends in the Automotive Industry like electric vehicles will be essential to our success.

Alongside efficiency and quality, our customers value reliability and a consistent supply they can trust. We offer the highest capacity in the industry for various packages, and continuously invest in new capacities. We work at every step to safeguard the long-term availability of our manufacturing processes and products, to ensure secure supply for our customers. Delivering complete customer satisfaction will also be key.

Investing in manufacturing The majority of our capital investment over the coming years will go into expanding our production capacity.

In terms of thin film technology that means increasing capacity for front and backside metallisation processes on Evatec’s CLUSTERLINE® primarily at our Hamburg facility and a continued move from 6” to 8” processing. It’s a given that these processes will need to run at increasingly tough repeatability specifications and yield on tools running at ever higher uptimes.

Continuous improvement is a way of life Working together with key suppliers in partnership is the only way to achieve long term success. We need suppliers who will be there in the long term, offering not only hardware but also process know-how, global support wherever our manufacturing is based, and the capability to offer customised solutions just for Nexperia.

The collaboration between today’s Nexperia and Evatec companies goes back much further than the origins of these company names of course. The first equipment to come out of the then Balzers’ factory was installed at Philips in Hamburg 40 years ago and today more than 15 Evatec systems are in 24/7 production in Hamburg.

This long term collaboration led to several customised solutions, for example a worldwide unique process for sputtering and alloying a gold layer on ultra-thin silicon substrates in one process step.

The “OEE” improvement programmes we run with major suppliers continuously monitor existing tool performance.

Though today’s OEE programme together with Evatec we improve operating procedures on our side, implement hardware upgrades to improve existing process performance, introduce new processes or address obsolescence and ensure the best tool uptime across our complete installed base. Evatec has consistently achieved a supplier rating on the highest level over the last years.

Recognising good performance At the Nexperia Supplier Day in April 2018 we were pleased to present Evatec with the award in the “Front End Equipment” category, recognizing the strategic relationship, outstanding performance and Evatec’s commitment to our ambitious growth plans and priorities. Reaching good understanding needs regular communication and the executive management of our two companies meet twice a year to ensure our expectations are clear.

“Working together for 40 years”

Semiconductor

About Nexperia Delivering benchmark solutions for today’s market requirements in semiconductors Nexperia leads in ultra-small, thermally enhanced packages

Nexperia offers the best solutions for Automotive Discretes, Logic, MOSFETs

» Everything is getting smaller

» Safety & multimedia electronic content

» Convergence of technologies

» Connected car (CtoC, Ctx)

» Increased electronic content

» Electrification of the car (braking, steering etc.) » Replacement of e/m relays by higher reliability MOSFETs

AECQ-100 / 101

Miniaturisation

Power efficiancy

ESD / EMI Protection

Nexperia delivers the benchmark in low power consumption devices

Nexperia offers the best protection solutions for every interface

» Extended battery lifetimes

» High-speed data rates

» Less heat in integrated applications

» Connected devices and media

» Environmental regulations

We are a dedicated supplier of Discretes, Logic and MOSFETs devices. Originally part of Philips, we became a business unit of NXP before becoming an independent company at the beginning of 2017.

Our company has 11,000 employees across Europe, USA and Asia, operates out of 5 manufacturing facilities (two front-end, three back-end sites) in Hamburg, UK, China, Malaysia and the Philipines to produce around 85 billion components annually from a portfolio of 10,000 active products to serve our global customer base.

More than 15 CLUSTERLINE® 200 in production in Hamburg.

applications everywhere

35

36

LAYERS 4 | SEMICONDUCTOR | POWER DEVICES | WIDE BAND GAP TECHNOLOIES

WIDE BAND GAP (WBG) TECHNOLOGIES A STATUS REPORT: SiC & GaN The strong dynamics of the power electronics industry is leading to ramp up in use of WBG1 based materials, in particular SiC2 and GaN3 as Si4 approaches its limits. Devices based on these materials are leading a next generation of energy efficiency and performance due to their intrinsic properties. In particular, GaN on Si power devices are more suitable for high frequency applications while SiC is better for high power density and high temperature inverters.

A market that is booming Although still relatively small compared to the Si power device market, the SiC market has already reached a relatively significant size compared to GaN due to its more mature technology.

In 2017, the SiC power device market was estimated at more than US$300 million5, roughly ten times that of GaN power devices. In fact, nowadays we can affirm that end-users are beginning to adopt SiC as the final solution whereas several years ago the market was still very small.

Today, 82% of the SiC market is driven by diodes used in PFC6 for power supplies, and in hybrid modules for applications such as PV7.

Yole Développement (Yole) expects that the transistor market will still grow with a CAGR2017-20238 of 56% with the introduction of these devices into applications such as EV/HEV9, including charging infrastructure, partly due to the implementation of full SiC modules. Indeed, this is a hot topic in the overall industry, where we see all the car manufacturers and their Tier One suppliers developing SiC solutions. We can already find

SiC devices in EV/HEVs in the main inverter of Tesla Model 310, and in the OBC11 from BYD. From industry feedback, it seems clear that the automotive segment will increasingly adopt SiC over the next 5-10 years.

By contrast, the GaN market is still some way behind, with a 2017 market estimated at at lower than US$20 million12. This is due to the lack of maturity of the devices. The main applications for GaN in the near future are fast charging adapters, as well as other high-end applications where the high performance of GaN is required such as LiDAR or wireless power. Last year we began to see some movement in the industry showing the potential market for these applications.

Technology developments for WBG devices The end user is interested in buying a solution that is cost effective and reliable, without considering the underlying technology: Silicon, GaN or SiC devices. For the cost conscious, device manufacturers claim that the total cost of the system will be about the same or lower than Si solutions.

Where reliability is concerned, no standards have currently been specified for GaN and SiC. As from the end of 2017, however, a JEDEC committee (JC-70) was created to set these standards. It is expected that once the JEDEC standards are specified, market competition will increase as end-users will be more confident that the technology is reliable.

In addition to the cost and reliability aspects, both GaN and SiC technologies still require some

additional development. In packaging, for example, some changes in terms of substrate or encapsulation materials need to be made to SiC modules compared to the standard IGBT module in order to sustain the highpower density.

In terms of device processing, it is not straightforward to change from Si to SiC or GaN for power electronics.

There are different requirements for clean rooms for both these materials.

For GaN (GaN-on-Si), different equipment is needed for the epitaxial growth (MOCVD13 manufacturing process), as well as every step where the surface of GaN is exposed, e.g., for contact etching.

About the authors Dr. Hong Lin has worked at Yole Développement, as a Senior Technology and Market Analyst, Compound Semiconductors & Emerging Materials within the Power & Wireless division since 2013. She specialises in compound semiconductors and provides technical, strategic and economic analysis, and is the author or co-author of various SiC and GaN market reports.

Dr. Ana Villamor serves as a Technology & Market Analyst, Power Electronics at Yole Développement.

She is involved in many custom studies and reports focused on emerging power electronics technologies, including device technology and reliability analysis (MOSFET, IGBT, HEMT, etc.).

Semiconductor GaN power device - Target application overview Source: Power SiC: Materials, Devices and Applications report, Yole Développement, 2018

Industrial GaN will compete directly with SiC. there is the need for high reliability and the cycle time to develop those technologies is about 3-5 years.

EV/ HEV

UPS UPS Server Data Centre

PV

GaN Envelope Tracking

LiDAR

Consumer In the coming years volume production of GaN for AC adapters for consumer applications will begin

Power Supply

High End Applications where GaN has a high performance benefit. It is a market that has started smoothly already.

Wireless Power Other App

Power vs frequency on electronics: power device technology positioning in 2018

Switching power (kW)

Source: Power SiC: Materials, Devices and Applications report, Yole Développement, 2018

SiC

103 102 101 100

Co mp etit i Ga zone on N/S iC/ Si

Thyris tor Si Bipolar

IGBT / IPM

1.

2.

3.

4.

5.

WBG : Wide Band Gap SiC : Silicon Carbide GaN : Gallium Nitride Si : Silicon Source: Power SiC report, Yole Développement, 2018 6. PFC : Power Factor Correction 7. PV : Photovoltaic 8. CAGR : Compound Annual Growth Rate 9. EV/HEV : Electrical Vehicles and Hybrid Electrical Vehicles 10. Source : Tesla Model 3 Inverter with SiC Power Module from STMicroelectronics report, System Plus Consulting, 2018 11. OBC : On Board Charger 12. Source: Power GaN report, Yole Développement, 2017 13. MOCVD : Metalorganic Chemical Vapor Deposition

GaN

MOSFET

103

104

105

106

Operating frequency (Hz)

Supply chain for GaN and SiC devices As well as developments in the both the device and market, an industrial supply chain for both SiC and GaN power devices has also had to be established, from wafer to epitaxy, bare die manufacturing, discrete/ module packaging and system end users.

The SiC power devices chain comprises companies with different business models: Vertically integrated companies from substrate to module, such as Wolfspeed and Rohm; Vertically integrated companies from bare die to end system such as Mitsubishi and Fuji Electric; Numerous players also occupy

a fragment of the supply chain, such as substrate suppliers, epi suppliers, device manufacturing and packaging.

A similar ecosystem for GaN power devices can be defined with different co-existing business models. We see established Si power players such as Infineon, On Semiconductor or Panasonic on one hand and start-ups using foundry models on the other.

Indeed, a foundry model is clearly developing which is facilitating both SiC and GaN fab-less and fab-lite companies in launching SiC and GaN products, thereby making the technology more accessible to the industry. For SiC power devices, the foundry model is currently driven by X-Fab supported by Power America.

But other foundries are also entering the market. On the GaN side, semiconductor giant TSMC is leading the business, partnering with different GaN start-ups such as Navitas and GaN systems.

SiC and GaN power device markets are still young compared to the well establish Si power device market.

The fast-evolving markets are seeing plenty of activity and changes from participants. We see players moving up and down the supply chain. Ever increasing industry development is coming, according to Yole. In terms of market ranking, the competition becomes more and more fierce day by day. The future will tell who has the last laugh.

37

www.evatecnet.com