skip to main content

Growing Scope of 200mm Applications Drives Advances in MEMS Process Technologies

Mike Rosa PhD

As we all know, the 200mm wafer fabrication equipment (WFE) market is steadily shrinking and will be relegated to history in the not-too-distant future. After all, shipments of 300mm WFE have outpaced those of 200mm platforms since 2004–05 and now the industry is even contemplating 450mm platforms. In 2014, the 200mm WFE market was approximately $1.5B (down from $1.6B in 2013 with 4% year-over-year decrease[1]). Cost efficiencies offered by the 300mm toolsets for the production of advanced memory and microprocessor devices have made it increasingly difficult to justify continued 200mm operation—right?

Interestingly, evidence suggests that these might be unduly pessimistic conclusions, similar to the misrepresentation that prompted Mark Twain to comment, “The report of my death was an exaggeration.” Even with the transition to 300mm wafers, fabs running 200mm wafers have continued to fabricate specialty memories; image sensors; micro-controller units (MCUs); analog products; motion, audio, and radio-frequency (RF) communication devices embodying MEMS; and discrete passive components such as resistors, capacitors, antennas, filters and switches.

Production of such devices makes good economic sense in fully depreciated 200mm fabs, including those formerly used for making devices now produced on 300mm wafers;[2] furthermore, new market forces are re-energizing the 200mm equipment scene. Leveraging its core capabilities in precision materials engineering, Applied Materials is addressing the new challenges accompanying these rapidly growing market segments through sustained investment in R&D to expand the capabilities of its 200mm systems.

The drive to reduce chip costs was the leading factor in the transition to 300mm wafers, which can accommodate more than twice as many dies as 200mm wafers. By 2013, 300mm wafers represented approximately 57% of overall semiconductor capacity, with IC Insights forecasting a rise to 64% in 2018.[3] Production on 300mm wafers is predominantly in high-volume commodity chips, including DRAM, flash memory, and microprocessors. Foundries and IDMs favor them to best amortize their manufacturing costs per die. However, 300mm systems are also used for some production of image sensors, power management devices, and complex logic and micro-component ICs with large die sizes.

The automotive industry was among the first users of the devices fabricated on 200mm wafers and has been steadily expanding its use of sensors for improving vehicle efficiencies and safety. Similarly, industrial and medical applications have contributed to this market. However, the advent of wireless and motion-tracking technologies in consumer electronics has been the major stimulus for growth in applications produced on 200mm platforms. Sold in the hundreds of millions per year, first the iPhone and then the proliferation of other smartphones, tablet PCs, and related digital devices that followed spurred adoption of “More-than-Moore” devices. These comprise micro-scale, systems-inpackage or system-on-chip devices, integrating processing power with sensors, RF communication, and a range of other functionalities. They can be fabricated using die with feature sizes greater than 1μm and profitably produced on 200mm systems.

Although ~5% of the total WFE market in 2014,[1] 200mm applications are steadily multiplying (see figure 1). In 2014, fab investments by leading foundries and IDMs resulted in a 45% increase in spending for secondary 200mm equipment.[4] Foundries account for two-thirds of this increase, responding to expanding demand for MCUs for automotive devices, consumer electronics, power management applications, and emerging IoT-related MEMS and sensors.[5]

Figure 1. Production on 200mm tools is making small but steady gains.[6]

KEY 200mm MARKET SEGMENTS

The automotive industry and consumer electronics sectors have been responsible for most of the 200mm production since before the 2008–09 industry downturn (see figure 2). Since the early 2000s, cars produced in the U.S. have been equipped with devices like tire pressure sensors that send data to the vehicles’ central computers. The average new car now contains 60 microprocessors; electronics for such functions as advanced driver assistance, heating and air conditioning, and rear-view cameras account for 40% of a car’s cost.[6] According to Gartner, sensors will represent 31.3% of all IoT-enabled automotive semiconductors in 2020, with automotive safety applications the fastest growing segment for image sensors. Automotive applications will also be the biggest consumer of inertial/motion sensors (43%).

Figure 2. The automotive and consumer electronics industries are the most significant drivers of resurgent 200mm production.[8,9]

The automotive MEMS sensor market is expected to grow at a CAGR of 4.7% from 2014 to 2020 and become a $3.6B market. [7] The consumer electronics story is epitomized by the global explosion in the use of digital cameras, iPods, smartphones, and tablets over the past 5 years. The value of MEMS produced for cell phones and tablets alone grew by $500M in 2012.[10] “Wearables” are now adding to the growth of this sector. These miniature electronic devices—such as the Nike+ system for tracking time, distance, pace, and calories through a sensor in the shoe, or the Apple Watch that integrates timekeeping and calendar information with address book, communication, and Internet capabilities as well as several activity monitoring functions—are driving exponential proliferation in sensor applications.

A new phase in interconnectivity will be fueled by the emerging IoT, which is predicted to link wearables, consumer devices, home automation, appliances, healthcare/medical, retail/marketing, transportation, and everything in between over the next decade. Analysts predict that five years from now, IoT-enabled sensors will be a $10 billion industry.[11] This new development is reinforcing demands for economical production of cheaper, smaller, more capable, and more power-efficient devices—for which legacy or enhanced 200mm toolsets offer optimal solutions.

DELIVERING MORE THAN MOORE

MEMS are among the More-than-Moore devices rejuvenating 200mm production. However, manufacturing them poses challenges from material- or process requirements not encountered in traditional semiconductor fabrication. Applied Materials has focused expertise and investment in its Xi’an, China, 200mm development facilities to support customers’ success in these technically complex and price-sensitive segments. There, hardware- and process development work is underway to enable new applications specialized for MEMS, power devices, thin-film batteries, and camera image sensors. In a previous issue, we focused on key advances for power device fabrication;[12] here we highlight progress in MEMS-related processes.

Monolithic Integration

As die sizes continue to shrink, manufacturers strive to more efficiently package MEMS and CMOS devices. While several hybrid system-in-package devices have been developed, the industry trend is toward monolithic integration of CMOS and MEMS devices. This approach raises a number of new issues.

LOW-TEMPERATURE SILICON GERMANIUM (SiGe): Polysilicon up to 60μm thick is currently used to fabricate most of the MEMS mechanical structures. However, its deposition rate at the low process temperatures (<400 °C) necessary to avoid damaging the CMOS during MEMS integration is too slow to be production-worthy. SiGe offers a viable alternative to polysilicon as its mechanical and electrical attributes are equivalent or superior. In addition, its deposition rate is far higher at the lower temperatures required for full monolithic integration with CMOS. Nevertheless, depositing it in the thick layers required for MEMS is a challenging balance of stress and in-film particle control.

Applied Materials has successfully developed the only commercially available production-worthy, single-step/single-pass process for depositing thick SiGe. Using plasma-enhanced CVD (PECVD) on Centura DxZ or Producer mainframes, the process deposits a series of 2μm layers for a total thickness of 10μm. It is undergoing further refinement to enable deposition of layers more than 5μm thick with even fewer particles per pass. At the targeted total thickness of 40μm, it will meet customer roadmaps for comparability with today’s polysilicon layers (see figure 3).

Figure 3. Applied’s state-of-the-art, low-temperature SiGe deposition process is targeting >5μm/pass; the 4μm-thick layer in the SEM image shows the desired crystalline characteristics.

LOW-TEMPERATURE CONFORMAL NITRIDE:
Monolithic integration poses even more complexities in the case of MEMS-based microphones. In these hybrid devices, two chips are packaged as a unit; one contains the MEMS and the second contains the CMOS. The MEMS structure creates significant topography and is also subjected to a somewhat lengthy release process step in hydrofluoric acid (HF). The silicon nitride (SiN) used as a barrier layer over the two components must be deposited at low temperatures (<400 °C) for CMOS compatibility. Further, it must be deposited slowly to ensure that the film forms a conformal layer and that its crystalline structure results in a high wet etch rate ratio (WERR) with the sacrificial silicon dioxide removed by the HF. Monolithic integration with its CMOS drivers not only demands a higher deposition rate for production-worthiness, it means that the SiN layer may also be used as an electrically insulating dielectric, which imposes additional essential requirements for breakdown voltage and current leakage. Using the Producer PECVD platform, Applied has developed a low-temperature process that achieves not only a higher deposition rate without sacrificing WERR, but also conforms to the requisite electrical properties (see figure 4).

Figure 4. Applied’s low-temperature SiN film forms a highly conformal, void-free layer with no delamination.

Ultra-Thick SiO2 Deposition

Pressure sensor applications require thick films (> 20μm) that are suited to high-volume manufacturing (high deposition rates, better stress control, low particle counts). Deposition processes for both doped and undoped thick oxide have been developed on the Applied Producer PECVD system. These neutral-stress films can be deposited at increments of 5μm per pass with excellent particle performance (see figure 5). Continuing development work focuses on depositing greater thicknesses per pass while maintaining proportionally low particle counts.

Figure 5. Applied’s PECVD technology meets requirements for ultra-thick SiO2 deposition with negligible particle counts.

All-In-One Etch

Figure 6. Applied’s all-in-one etch and resist removal process boosts fab productivity.

MEMS manufacturers were previously unable to perform the oxide hard mask open etch and deep silicon main etch steps in the same chamber. This situation increased wafer handling, added queue time complexities, and reduced overall throughput.

Based on the Applied DPS DTM 200mm deep reactive ion etch chamber, we have developed a two-step etch process (hard mask open, deep silicon etch) followed by a photoresist strip all in one chamber. Figure 6 shows representative results. Besides lowering wafer handling time and the overall cost per wafer, it improves the productivity of each chamber on the tool.

MEMS/CMOS Packaging

Bonding of MEMS wafers to CMOS wafers has become an increasingly popular method of integrating these devices in instances where the dual-layer bond interface uses aluminum and germanium (Ge) to form a hermetic seal between the two wafers and to simultaneously create selective electrical connections to bond pads between the two wafers. Volume production using this method depends on reliable delivery of a Ge layer at a high deposition rate and low particle count.

In this case, meticulous process tuning now enables Applied Endura PVD technology to deliver a pulsed DC process that deposits Ge at a rate exceeding 2600Å/min. with fewer than 200 particles at 200nm diameter (see figure 7).

Figure 7. Enhanced Endura PVD technology delivers thick Ge film needed for MEMS/CMOS packaging.

FUTURE DEVELOPMENTS

The growing variety of MEMS devices and functionalities is leading to the adoption of new metals, dielectrics, and ceramic materials in the manufacturing flow. Many require specialized RF PVD processes to meet deposition rate, target erosion uniformity, or other specifications. Recent forecasts predict growth in a number of devices that use piezoelectric ceramic materials such as lead zirconate titanate and various lead-free alternatives for piezo-actuators or sensors, aluminum oxide for optical coatings or high-k applications, and vanadium oxide for next-generation microbolometer applications, among others.

As these emerging applications take hold, reliable manufacturing solutions will be needed that extend beyond traditional pulsed DC PVD to enable manufacturability of these films and solve key issues such as deposition rate, uniformity, kit life, target erosion, pasting efficiency, and so forth. To address these anticipated customer needs, Applied is developing a new 200mm RF PVD chamber capable of both RF and pulsed DC deposition. Besides MEMS, the chamber will serve applications related to power devices and LEDs.

SUMMARY

A small but steady resurgence in 200mm production has marked recent years, driven in large part by the consumer electronics and automotive segments for devices fabricated with non-leading edge die and profitably produced by these systems. MEMS account for a significant portion of this production and have posed challenges that Applied is solving by enhancing deposition and etch processes, and designing innovative hardware. Through our 200mm development facilities in Xi’an and collaborations with our customers, we are pursuing continuous tool and technology improvement to support today’s broadly expanding cost-sensitive More-than-Moore device classes.

Thanks to Jeannette Hoffman for her contributions to this article.

For additional information, contact mike_rosa@amat.com

[1] Semi Secondary Fab Equipment Report, January 2015
[2] http://www.icinsights.com/news/bulletins/Installed-Capacity-For-300mm-Wa...
[3] http://www.icinsights.com/news/bulletins/300mm-Capacity-Dominates-But-Li...
[4] SEMI Secondary Fab Equipment Report, January 2015.
[5] Christian G. Dieseldorff, SEMI World Fab Forecast, SEMICON West TechXPOT, July 9, 2014.
[6] MIT Technology Review, "The Internet of Things," July/August 2014.
[7] Markets and Markets, "Automotive MEMS Sensor Market by Type (Inertial Sensor, Microphone, & Pressure Sensor), Application (ADAS, ECU, ESC, HVAC, Safety & Security, In-Car Navigation, OIS Camera, Microphone in Cabin, & TPMS), & Geography—Analysis & Forecast (2014–2020)," March 2015.
[8] Glenn G. Daves, Freescale: "Automotive Packaging Trends," SEMICON West TEchXPOT, June 9, 2014.
[9] MEMS Trends, Issue 15, July 2013.
[10] Yole Developpement, "Status of the MEMS Industry, 2012."
[11] Stephan Ohr, "Sensors on the Internet of Things—A Tale of Two Markets," Gartner, October 23, 2014.
[12] Nanochip Fab Solutions, Vol. 9, Issue 1, pp. 7-14, 2014.