1.1.1 A short history of MEMS
The abbreviation MEMS stands for Micro-Electro-Mechanical-Systems. This name has been in use since the mid-1980's and basically refers to systems where mechanical parts and some electronic interfacing are integrated on a single chip or in a single package. Since 1987 there is a yearly IEEE MEMS conference, which is the major conference in the field. Other names that are used instead of "MEMS" are "Micro Systems Technology" (MST) and "Micromachines".
Although the name MEMS started to appear in the 1980's, the first actual devices were already made much earlier. Already in 1962 the first pressure sensors were made consisting of thin silicon membranes with piezoresistive readout (Tufte et al., Honeywell, 1962), and in 1967 a "resonant gate transistor" was presented (Nathanson, 1967) where the gate electrode of a field effect transistor (FET) was replaced by a vibrating metal electrode as indicated in Figure 1.1.1.
Figure 1.1.1: Resonant gate transistor presented by Nathanson. The gate electrode in a field effect transistor is replaced by a suspended metal cantilever beam. An electrostatic force is applied to the tip of the beam by a voltage difference between the beam and the "input force plate" underneath the beam. By applying an AC voltage the beam can be made to vibrate up and down at its resonance frequency. This results in modulation of the current through the transistor at the same frequency.
The first silicon accelerometer, or acceleration sensor, was presented by Roylance et al. in 1979. Figure 1.1.2 shows a picture of this device. In the same year the first silicon inkjet printer nozzle was presented by Peterson et al..
Figure 1.1.2. First silicon accelerometer, consisting of a silicon proof mass attached to the tip of a silicon cantilever beam. An acceleration results in displacement of the proof mass and bending of the cantilever beam. The latter is detected by integrated piezoresistors.
The first person who thought of the great potential of MEMS devices and who gave a lecture on this in 1960 is Richard Feynman, who you already know from his lecture notes in the module Fields and Waves. If you are interested you can find a transcript of his talk on internet. Another important paper that you should read if you are interested in the origin of MEMS is Silicon as a mechanical material by Kurt Petersen, which was published in 1982. In this paper Peterson describes the attractive properties of silicon for mechanical devices: it is strong, comparable to steel, especially in crystalline form it acts as a perfect spring, without creep or hysteresis, and it allows batch fabrication, i.e. many devices can be fabricated simultaneously on a wafer. One property of silicon can be regarded as a disadvantage: it is brittle, i.e. when overloaded silicon will break whereas steel would first deform plastically (resulting in a permanent deformation). But when a device breaks, at least you notice that it is broken; a device made from steel might seem ok while it is not.
1.1.2 Key technologies
There are a number of key technologies that are used to fabricate MEMS devices. Here we will give a short overview.
Photolithography
In photolithography light impinges a transparent plate with opaque regions, called a mask. Below the mask a photosensitive material is placed, which can be “developed” after illumination. In this process, the illuminated regions (or the non-illuminated ones) are dissolved, and the underlying material is accessible to further processing. Figure 1.1.3 shows an example.
The basis of silicon micromachining is photolithography. Photolithography defines regions on a silicon wafer where the machining is done. The machining includes etching (in a liquid, plasma, gas or by high energy ions or particles), doping and deposition of thin films. It is clear then that micromachining allows the fabrication of three-dimensional structures with complex forms only in two dimensions, everything in the third dimension is in some way extended from two dimensions. Some extension to really three-dimensional structures can be accomplished by means of wafer bonding (see below).
Figure 1.1.3: An example of using photolithography to etch a cavity in silicon.
Thin layer deposition
It is impossible to make a sensor only from silicon. One has to be able to etch silicon, therefore materials are required which are inert against the etching agents used to machine silicon. Sensors have some kind of electric function, so resistors and conductors are needed. Finally one needs special materials for transduction. These materials typically are used in the form of thin films deposited on top of silicon wafers. Doping can be used to alter the (electrical) properties of silicon.
Common techniques for thin layer deposition are: 1) Oxidation to create a silicon dioxide (often called "oxide") layer, 2) Chemical Vapour Deposition (CVD) and Low Pressure Chemical Vapour Deposition (LPCVD), see Figure 1.1.4., to create e.g. layers of polycrystalline silicon ("polysilicon", or "poly"), silicon nitride or phosphor silicate glass (PSG), 3) Evaporation and sputtering to deposit metal films.
Figure 1.1.4: Thin layer deposition by LPCVD - Low Pressure Chemical Vapour Deposition.
Wet etching
Wet chemical etching is the oldest silicon micromachining technology. We distinguish between isotropic and anisotropic wet chemical etching.
In isotropic etching the etch rate is independent of the crystallographic direction. Typical etching solutions are mixtures of concentrated HF and HNO3. These are very corrosive liquids; one of the few materials not attacked by these mixtures is Teflon, from which etching vessels and holders are made. Also noble metals are resistant and fortunately silicon nitride, which can be used best as a mask material.
The etch rate of silicon in OH-containing solutions is anisotropic. There are deep etch rate minima in the crystallographic <111> directions, and - depending on the solution - shallow minima in <001> (KOH) and <110> (EDP, KOH with isopropyl alcohol). Mask materials are silicon dioxide thermally grown in a wet oxidation oven or silicon nitride grown in an LPCVD reactor.
Figure 1.1.5 shows how a wet etching process in performed in a wet bench. Figure 1.1.6 shows anisotropic and isotropic etched cavities.
Figure 1.1.5: Wet etching of silicon wafers in a wet bench.
Figure 1.1.6: Left: Anisotropic etching where the cavity a shaped by the silicon crystal. Right: Isotropic etching.
Wafer bonding
Wafer bonding refers to the fixation of wafers on top of each other. It is an indispensable technique for silicon micromachining. Wafer bonding not only allows us to fabricate constructions with rather complex structures in three dimensions, but it also plays a role in packaging. Because whole wafers can be bonded containing many single components, tedious and expensive assembly can often be circumvented by wafer bonding.
Silicon is very smooth. Wafers, when contacted, do not touch each other at three small spots, as normally is the case, but on extended regions. Wafers, when clean and flat (not bent too much), spontaneously adhere to each other. This tendency can be enhanced using external forces, such as a weight on top of the wafer package or – much more efficient – electric fields. Bonding without the help of external forces between silicon wafers with or without thin films on top is called “direct silicon bonding”, DSB, or “silicon fusion bonding”, SFB (although the process has nothing to do with fusion). Using electric forces requires additional materials (glass), and is referred to as “anodic bonding”.
Figures 1.1.7 and 1.1.8 show two devices that required wafer bonding to make them.
Figure 1.1.7: Angular rate sensor from Robert Bosch GmbH: a cap is bonded on top of the sensor chip and hermetically seals it as the device needs to operate in vacuum.
Figure 1.1.8. LIS3DH three-axis linear accelerometer from ST Microelectronics consisting of three chips on top of each other: the bottom chip contains the accelerometer structures, the middle chip is bonded on top for protection at wafer scale, the top chip is added later and contains the interface electronics.
Plasma etching
Wet chemical etching is confined either by the <111> planes of single crystalline silicon or by poorly controllable diffusion fields. Both techniques severely limit the possible design space for mechanical construction. Plasma etching offers much more. In plasma etching ions impinge on the substrate, and neutral particles reach the substrate by diffusion. The combined action - which is quite complex - leads to anisotropic etching. The anisotropy here is not the result of crystallographic properties but a result of the direction of the ion flux towards the substrate. In particular reactive ion etching (RIE) and related techniques have been developed for micromachining. Originally developed for IC processing (etching of thin films, pattern transfer) for comparatively shallow etching (etch depth not more than 1 - 3 μm), MEMS has a need of great structure height, so etching must be deep, with neat sidewalls, vertically straight through the wafer. Several processes have been developed to make this possible. Figure 1.1.9 shows an example of etched trenches. Figure 1.1.10 gives an impression of how a plasma etching process works.
Figure 1.1.9. An example of trenches etched by deep reactive ion etching.
Figure 1.1.10: A typical plasma etching process. During the process an inert layer is deposited everywhere in the reactor, in particular on the sidewalls of trenches and protect the sidewalls against chemical etching. The material is also deposited on the bottom of the trench but there it is removed by ions.
Surface micromachining
Surface micromachining is a process to fabricate free standing and freely moving microstructures in a large two dimensional design space. The process idea is shown in Figure 1.1.11. The trick is to deposit and pattern two thin films of materials that can be etched away selectively with respect to each other. The substrate now only plays a role as a mechanical carrier; the structure is made from a thin film material.
Although many combination of materials are possible for this process, the technology has been developed to a high standard for the combination of silicon dioxide or PSG as the sacrificial material and silicon as the structural material. In Figure 1.1.11, first a silicon dioxide layer is deposited and patterned. Next, a poly silicon layer is deposited and patterned, which will later form the mechanical structure. The oxide layer is then selectively etched in a HF solution. Next, care has to be taken that the released structures will not stick to the substrate. When etching the sacrificial layer the free-standing structures have the tendency to touch the substrate and stay there. The cause is thought to be the surface tension of the liquid during evaporation.
Figure 1.1.11: A simple surface micromachining process.
Instead of depositing the silicon dioxide and polysilicon layers, one can also start with a so-called Silicon-On-Insulator (SOI) wafer. An SOI wafer consists of a thick silicon "handle" layer that acts like the substrate, a thin burried silicon dioxide layer (the "BOX" layer), and a thin silicon "device" layer that will contain the mechanical structure. Figure 1.1.12 shows a fabrication outline of an SOI wafer based surface micromachining process. This is the process that is used in the MEMS Design project.
Figure 1.1.12: Surface micromachining using an SOI wafer. First, the wafer is oxidized to create a silicon dioxide layer at both sides of the wafer. Next, a pattern is etched in the device layer, and trenches are etched around the individual chips in the handle layer. Next, in step 4 the oxide is selectively etched in a HF solution. The etching is stopped before the structures are completely released to prevent stiction. In step 5, the selective etching of oxide is continued in HF vapour to prevent capillary forces from pulling the structures down to the substrate, until the structures in the device layer are completely released. At the same time, the chips are separated from the wafer so that no further separation in chips is needed.
1.1.3 Examples of devices
This section provides a large number of examples of MEMS devices without explaining the exact operating principles.
Force and pressure sensors
MEMS force and pressure sensors exist in a large variety of shapes and ranges, as illustrated by Figure 1.1.13. A force or pressure sensor typically consists of a silicon suspension or a membrane that deforms under an applied load. The amount of deformation is measured.
Figure 1.1.13: A large variety of MEMS force and pressure sensors exists. This figure shows some examples: a highly sensitive surface micromachined device that can measure the forces exerted by a fruit fly, a bulk micromachined chip that can measure up to 1000kg load on a 1cm2 area of silicon, a 6 Degree-of-Freedom sensor that measures 3 forces and 3 moments, two devices made by ISSYS, a very small and a very sensitive one, and a pressure sensor from Robert Bosch that includes interface electronics on the same chip.
Microphones
A MEMS microphone typically consists of a thin circular membrane that vibrates due to the sound pressure. Figure 1.1.14 should photographs of two commercial devices. Measurement of the membrane movement is typically done capacitively: the vibrating membrane is one electrode of a capacitor and the other electrode is formed by a second parallel membrane which is perforated. The second membrane will not vibrate as there is no pressure difference across this membrane (due to the perforation).
Figure 1.1.14: Two silicon microphone chips, from Infineon and Knowles.
Acceleration sensors
Acceleration sensors or accelerometers basically consist of a suspended mass, the proof mass. When the chip experiences an acceleration, a force equal to the mass times the acceleration is needed to accelerate the mass. This force has to be supplied by the suspension springs. This results in a deformation of the springs and a displacement of the proof mass proportional to the applied acceleration. Hence, by measuring the proof mass displacement or the strain in the spring suspension a measure is obtained for the acceleration. Figure 1.1.15 gives some examples of accelerometer devices.
Figure 1.1.15: Some examples of MEMS accelerometers: a 3D accelerometer from ST Microelectronics with the interface electronics on a separate chip, a 2D accelerometer with integrated electronics by Analog Devices, and an ultra high resolution device developed at MESA+ in cooperation with the NIKHEF institute.
Angular rate sensors
Angular rate sensors or gyroscopes measure angular velocity. The operation is based on the fact that a moving mass will move in a straight line unless a force is applied. A commonly used structure consists of two masses that are continuously vibrating towards and away from each other. Without rotation, the masses will move along a straight line. If the chip rotates, the movement will deviate from a straight line - a force is needed to make the masses follow the rotation of the chip. The deviation can be sensed and is proportional to the angular velocity. A large variety of implementations exist, as illustrated in Figure 1.1.16. Typically one can distinguish a "drive" mode, which is a vibration that is generated on the chip, and a "sense" mode, which is a vibration mode induced by the rotation. The amplitude of the sense mode is proportional to the amplitude of the drive mode and to the angular velocity.
Figure 1.1.16: Some examples of angular rate sensors (also called gyroscopes): a device from Analog Devices containing two vibrating masses and with integrated electronics, the angular rate sensor used in the iPhone 4, which seems to be manufactured by ST Microelectronics, an early device developed at the University of California in Berkeley, and an early vibrating ring device developed by Delco.
Microfluidic devices
Figure 1.1.17 shows some examples of microfluidic devices: a micro Coriolis mass flow sensor chip with an integrated control valve, a Wobbe Index sensor to measure the energy content of natural gas, and a "Microflown", which measures the particle velocity associated with a sound wave.
Figure 1.1.17: Some microfluidic devices: a micro Coriolis mass flow sensor based on a vibrating tube, a Wobbe Index sensor which measures the energy content of natural gas by burning gas on-chip and measuring how much heat is generated, and a so-called Microflown, a thermal flow sensor fast enough to measure the air flow associated with a sound wave.
Ink jet heads
Ink jet printer heads are one of the most successful microsystems. One of the ways to dose drops of ink on a piece of paper is to squeeze the liquid ink through an opening – the nozzle – in such a way that a single drop with a reproducible volume and speed will leave the nozzle. This can be accomplished by a piezoelectric element or by a heater, using thermal expansion of ink vapour. Figure 1.1.18 shows two examples.
Figure 1.1.18: Two examples of MEMS inkjet printer heads.
Projection systems
Another very successful application area of microsystems is in projection systems. Such systems typically contain chips with arrays of moveable micromirrors, each mirror corresponding to a pixel of the projected image, or a single mirror that is used to scan a laser beam. Figure 1.1.19 shows two examples.
Figure 1.1.19: Some examples of chips with moveable micromirrors.
1.1.4 Master courses
Within the module Device Physics we can only offer a glimse of the broad field of MEMS and microtechnology. If you are interested to learn more, there are two master courses that you can choose after completing your bachelor:
Measurement Systems for Mechtronics - This course explains all about the operating principles of MEMS sensors.
Micro-Electro-Mechanical-System (MEMS) Design - This course forms an introduction in how to design MEMS devices. Part of the course is a project where you design your own device which is then fabricated in the MESA+ cleanroom.
If you can't wait until your master you can have a look at the book "Mechanical Microsensors", which can be downloaded from the Canvas site.