Making MEMS: A short guide

Micro-Electro-Mechanical-Systems (MEMS) are three-dimensional structures made using silicon micromachining technologies. They made their first appearance in semiconductor fabs in the sixties and, among many applications, they can be used to sense acceleration, angular rate, pressure, and sound pressure.

Our daily life is full of micromachined physical sensors. In the car, all active and passive safety systems, like the vehicle dynamic control and air bags use acceleration and yaw rate sensors. Also, pressure sensors in engine manifolds and fuel lines can be used to keep petrol consumption to a minimum.

However, mems are experiencing a 'consumerisation wave'. Portable pcs now use a 3-axis accelerometer to protect data stored on the hard disk drive in case of an accidental fall. Meawhile, a number of mobile phones exploit the sensing ability of tiny accelerometers to simplify the interface between people and their equipment. Last but not least, gaming devices such as the Nintendo Wii or Sony PS3 enable us to really appreciate the experience thanks to the remote controller motion-sensing feature.

The penetration of motion sensors, like accelerometers and gyroscopes, will continue to increase in the automotive market, driven partly by regulations, but their penetration in the consumer market will happen at a higher rate. For example, aside from vehicle dynamic control systems, yaw rate sensors are being used to improve image stabilisation in camcorders and digital still cameras. Moreover, motion sensors and geo-magnetometers are expected to cluster together in Motion Measurement Units to enable personal navigation in portable devices, thus fostering the deployment of location-based services by telecommunication operators.

Tiny pressure sensors today are widely used in the automotive and the medical markets. Their penetration will rapidly increase in the automotive sector thanks to the tyre pressure monitoring application. However, recently developed thin, small, and inexpensive pressure sensors will also appeal to the consumer market by enabling new applications and bringing new sources of revenue to wireless operators. Capacitive silicon microphones are also competing with non-surface mountable electret-condenser microphones in mobile phones and laptops.

Clustering several sensors, like accelerometers, gyroscopes, and pressure sensors, into a single module will happen, driven by customer demands to suit specific applications. STMicroelectronics has two technology platforms, THELMA and VENSENS, designed for sensor integration. So far, ST has developed multiple axis gyroscopes, pressure sensors, and microphones and is open to partnering with customers to develop new sensor devices, such as magnetometers. Production and development activities are all run in an 8in MEMS fab to accommodate the fast time-to-market required by the consumer market.

Physical sensors are micromachined, sharing the same processing steps as basic integrated circuits. The end result though is typically a 3D mechanical structure, most often on a silicon substrate. Other materials that can be micro-machined or micro-formed include quartz, glass, plastic and ceramic. Quartz and ceramic are used for crystal resonators and for Coriolis-based gyroscopes. However, silicon is becoming increasingly popular thanks to its excellent electrical, mechanical, and thermal properties.

In addition to its physical properties, silicon's attractiveness is down to the fact that manufacturers can make thousands of micro-machined components – MEMS is benefiting from the same economies of scale that made microelectronics such a success. Moreover, since the components are made side-by-side on wafers and with an extremely well-controlled process, the devices can be made much more precisely and repeatably than similar products manufactured in different ways.

The primary difference between MEMS devices and their CMOS transistor cousins is that in the case of the former, it is not just the electrons that are moving. Once combined with ics, electrical signals generated by moving structures such as diaphragms and cantilevers provide perception and control capabilities to create sensors for a large variety of applications.

Many of the micromachining processing steps are similar to basic IC manufacturing: photolithography, material deposition, reactive ion, and chemical etching. However, while the CMOS roadmap aims to pack more and more devices in plane and thickness, micro-machined devices usually have dimensions in the tenths of a millimeter and thicknesses of several tens of microns. Wet etching, grown or electroplated thick films, stacks of two/three bonded wafers, through silicon vias/holes, and high aspect-ratio dry etches are common steps for micromachining technology. Meanwhile, MEMS devices use materials like gold or Glass Frit that are forbidden in a CMOS process.

Although each MEMS manufacturer uses a specific micro-machining process, all of the processes can be classified into two broad classes:

A. Bulk Micromachining: this is a subtractive process because a large portion of the substrate is removed to form whatever structure is desired. This technique requires less precision than surface micromachining. Thicker structures are easier to fabricate because the substrate thickness can be chosen quite freely, but the crystal planes of the silicon substrate limit the shape of the micro-machined structure. This technology is quite old and is reaching the end of life.

B. Surface Micromachining: it is an additive process requiring the building of various layers of materials that are selectively left behind or removed by subsequent processing. The bulk of the substrate remains essentially untouched. This technique was initially limited to thin devices (under 2um), since only thin films could be deposited or grown on the substrate. However, the use of thicker films, as well as new wafer bonding techniques is helping to create thicker devices. By exploiting all the tricks offered by photolithography, the manufacturing of very complex and innovative mechanical structures is fairly simple. This class of processes has a longer roadmap, and it's the most popular for motion sensors.

STMicroelectronics has two different micromachining processes for manufacturing MEMS;

THELMA: THick Epitaxial Layer for Microgyroscopes and Accelerometers;
VENSENS: : VENice process for SENSor.

The first process is for manufacturing high-performance and low-cost motion sensors, like accelerometers and gyroscopes, and microphones, while the latter enables extremely small pressure sensors to be made. Both processes are a proprietary combination of the manufacturing steps of Bulk and Surface Micromachining technologies.

The THELMA process begins with a standard silicon wafer onto which a layer of oxide (under 2um) is first grown for electrical isolation. A thin poly-silicon layer used for interconnections and a second layer of sacrificial oxide (under u¼m) are then deposited. Into this layer, holes are etched at the points corresponding to the supports for fixed elements and anchors for moving elements. A thicker poly-silicon epitaxial layer (under 15 um) is grown on top of this, and into this third layer the structures for the moving and fixed elements of the device are etched with a single mask. Finally, the sacrificial oxide layer beneath the structures is removed by an isotropic etching operation to free the moving parts.

The open space around the structures is filled with a gas, usually dry nitrogen, to reduce or eliminate effects caused by humidity or variations in gas density, which would affect the resonant frequencies of the device. A second wafer is then bonded to the first one to protect the tiny structures during an injection molding process during which high pressures are applied.

The VENSENS process begins with a standard silicon wafer. A proprietary combination of wet and dry silicon etching steps enables the formation of a sacrificial layer on top of which a monocrystal silicon layer is grown. The thickness of the sacrificial layer is less than 3 um and the thickness of the structural layer can reach 20 um. The end result is very similar to what is possible to get with bulk micromachining wafer-to-wafer bonding. But there is one important advantage: the resulting thinner, smaller, and mechanically more robust chips . Moreover, the sealing of the cavity doesn't require wafer-to-wafer bonding, and thus the reliability of the sealing joint is higher.

Because of the electrical properties of the monocrystal silicon, quality, stable resistors can be integrated in the structural layer through implantation or diffusion process. Then these resistors are connected with an aluminum metal layer to realize the four branches of a Wheatstone bridge. The bridge is sensitive to pressure changes due to the excellent piezoresitive properties of the monocrystal silicon layer. Finally, the metal layer is then covered with a standard dielectric, like silicon-oxy-nitride, to provide the required protection against the external corrosive agents.


Although the micron-scale of current MEMS devices mean that they can be made in older 6in wafer fabs, many companies are making the switch to 8in lines to meet the fast growth in demand and price pressure of the consumer market. STMicroelectronics has already made this transition. On the technical side, the scope and use of MEMS is primarily due to their extremely small size, terrific reliability, and low power consumption, which, in many instances, allows MEMS to be capable of faster and more precise operations than their macroscopic equivalents. But the cost advantage for the customer, especially in the consumer market, cannot be ignored.

0 comments on “Making MEMS: A short guide

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.