!1 Special Lecture Series Biosensors and Instrumentation Lecture 6: Micromechanical Sensors 1 This is the first part of the material on micromechanical sensors which deals with piezoresistive and piezoelectric systems. The second part will deal with sensors based on capacitance along with looking at some other issues for mechanical sensors and switched capacitor amplifiers. These are the three main classes of micromechanical sensors that we ll deal with but it should be understood that there are other techniques for actuation and sensing such as the use of magnetism. Piezoresistance Piezoresistance is probably the simplest way to transduce movement into an electrical signal. It is a change in the resistance of a piece of material that is subjected to a mechanical force and was discovered by Lord Kelvin around 1856. It is also referred to as the strain gauge effect and most electrically readable strain gauges will operate in this way. It lends itself to microfabrication as thin film deposited or diffused resistors make good strain gauges. There is also a greater effect in semiconducting materials than there is in metallic conductors. A typical strain gauge (such as the figure in the slides which is taken from http:// en.wikipedia.org/wiki/strain_gauge) is fabricated in a thin metal layer on a flexible support, probably an insulating polymer. Bending in the long direction causes the largest change in resistance and the markings are there to aid in alignment with the direction of the strain that is to be measured. Stretching the gauge in this direction, imagine the middle curling up out of the page will put the conductor in tension making it longer and thinner, which increases R. Similarly putting it into compression, imagine the ends curling up, the middle down, will squash the conductor making it shorter and wider so R decreases. The Wikipedia page above has some nice 3D diagrams of this. A gauge is characterised by its GAUGE FACTOR, this is the relative change in resistance ( R/R) divided by the applied strain ε. Typical maximum gauge factor for metal strain gauges with strain in the long direction is 2 though there are some metals with higher values. Strains of up to 10% can be measured with a good design but this requires good connection between the gauge and what s being measured. They are typically glued down and if there is any slippage due to the glue that can reduce the strain being measured or affect the time constant of the sensor. The biggest issue to control in strain gauges is the effect of temperature as all materials used in strain gauges will have a resistivity that varies with temperature, as we ve seen in the section of the course on temperature sensing. In addition there will be thermal expansion of the gauge and this can add to the mechanical strain being measured. It is possible to make the gauge out of an alloy like constantan which has very low thermal coefficients of resistance and expansion. However, there is another way to compensate for thermal effects. There are a number of different ways to arrange strain gauges in a Wheatstone bridge. The simplest is to replace R1 and R2 with strain gauges but only the R1 gauge should be fixed to the surface being measured. The other will be a dummy gauge that experiences the same thermal environment as the active gauge but does not change with applied stress. R3 and R4 should be chosen to have the same resistance of the unstrained gauges but may not be located
!2 near to the measurement. In an alternative setup we can use 4 gauges at the same time but arrange them so that two, opposite, gauges will be compressed, reducing R and two will be in tension, increasing R. This gives a much larger imbalance to the bridge for the same input so the resulting output voltage will be much higher. It also serves to linearise the output compared to using a single gauge. In the lecture we next explored a bit of the history of these circuits by looking at Murphy s Law. That won t be repeated here but you can read more about the story here: http://improbable.com/airchives/paperair/volume9/v9i5/murphy/murphy0.html As previously stated the piezoresistive effect is generally greater in semiconductors than in metals. The effect is not solely down to changes in the resistor geometry as it is stretched but there are additional effects due to strain changing the mobility of charge carriers in the semiconductor. This means that the gauge factor is significantly higher than with metals, up to a GF of 100 or more with single crystal diffused resistors. These effects on mobility are exploited in modern microelectronics. For example in advanced CMOS the lattice mismatch between Si and SiGe is used to create strain in the channel region of transistors to improve the mobility. The figure in slide 9 is a cross section through a diffused resistor created in a lightly doped n- type well. The function for the voltage drop when the resistor is strained is given on that slide and includes terms for the unstrained resistance, applied current along with coefficients representing the piezoresistive effect both along the length of the resistor and across it. Polysilicon resistors are also possible but will generally have a lower gauge factor. The important concept here is that you need to consider the transverse strain as well as axial/ longitudinal strain as the piezoresistive sensitivity will vary with orientation in single crystal silicon. Example Applications of Piezoresistive Sensing Silicon strain gauges are easily integrated with micromechanical sensor systems and we ll go on to look at a couple of examples of systems designed and fabricated at Edinburgh. The first of these is a pressure sensor consisting of two silicon substrates bonded together where one has a bulk micromachined cavity sealed with a flexible membrane. This could be fabricated from silicon dioxide or silicon nitride depending on the mechanical properties required. For example it s not usually a good idea to have significant compressive stress in a membrane like this as it can lead to buckling. Anyway, once we have this sealed cavity any pressure difference between the cavity and the atmosphere will lead to movement of the membrane so we just need to be able to measure it. The pressure sensor designed here used polysilicon strain gauges in a Wheatstone bridge arrangement. The poly piezoresistors are the short sections of track running in the vertical direction in the design shown on slide 12. These are arranged in serpentine meanders to increase the sensitivity and in this design there are 4 resistors in each arm of the bridge. The gauges are laid out so that two sets of resistors straddle the edge of the membrane (pink square) while the other two are parallel to the edge and shouldn t see significant stress. This is something like the push-pull bridge seen earlier but here only two resistors change. Slide 13 shows a close up microscope image of an actual sensor with 8 resistors in each arm. The cavities in the silicon were etched with an anisotropic wet process using TMAH which is selective to certain crystal planes. The devices required customised packaging which included a vacuum connection to allow the pressure to be controlled. The package was a standard dualin-line ceramic package which had a hole cut through it with a high speed drill. The chip was
!3 firmly glued in place over the hole and wire bonded to make electrical connection to the resistors. A plastic push fit vacuum connection completed the device. The Wheatstone bridge wiring includes a break in one arm which allows the addition of a potentiometer to balance the bridge at some set pressure. It could then be connected up to a differential amplifier or instrumentation amplifier to boost the signal. The second example of a MEMS sensor using piezoresistors is an accelerometer developed and designed at Heriot-Watt University in Edinburgh and fabricated in the SMC. The concept is of a robust, micromachined 3-axis accelerometer in hermetically sealed, biocompatible packaging. This would be stitched to the outside of the heart during a bypass operation and would monitor the motion of the heart to spot any problems after the operation. The picture in the slides of the packaged device stitched to a animal heart shows a commercially available accelerometer which is too large for safe use in humans. The Heriot-Watt device is significantly smaller and shown on top of a two pence coin (similar to a 10 coin). The MEMS accelerometer used in this case consists of 4 moving proof masses made from bulk etched silicon. They are suspended from thin cantilevers which have diffused resistors in the surface to monitor the movement of the masses. There are 2 cantilevers for each mass and two resistors on each cantilever. This means there are 16 piezoresistors in total, arranged in 4 Wheatstone bridges. This sensor takes advantage of the fact that the coefficients of piezoresistance are anisotropic in crystalline silicon. Arranging the resistors in different directions relative to the crystal planes allows equal and opposite effects in the longitudinal and transverse directions. More simply, if you have resistors at right angles to each other, you can cause one to increase by R while the other decreases by R by bending in a particular direction. In plane acceleration will cause differential movement of opposite masses. So, in the case when there is movement in the X direction, mass 1 moves up while mass 2 moves down. Motion in the opposite direction would obviously cause the opposite effect while acceleration in Y would involve movement of masses 3 and 4. In each case two resistors will be in compression and two in tension which is ideal for use in a Wheatstone bridge. Out of plane acceleration is a little more difficult to measure, all 4 masses will move up or down together and so two separate Wheatstone bridges are used. However, again it is possible to combine pairs of transverse and longitudinal resistors to obtain the maximum output sensitivity from each bridge. Piezoelectric Sensors The second way to directly transduce mechanical forces into electrical signals is the piezoelectric effect. While piezoresistors are passive sensors requiring external power the piezoelectric effect can be used to make active sensors. It is the ability of some materials to generate an electrical potential when mechanically stressed. This is a two way effect and electrical inputs can cause a piezoelectric material to change shape. One of the most common examples in medicine is ultrasound where piezoelectric transducers both produce the ultrasound and sense the returning echos. Piezo materials are usually crystalline and will contain fixed dipole charges which have some form of polarisation, i.e. the dipole charges are lined up in the same directions. Applying mechanical forces can move these charges around generating the measured potentials. The important point is that useful signals are only produced by changing the applied force, dc mechanical signals do not produce any output. Piezo materials range from natural substances
!4 like quartz crystal and bone, to synthetics like lead zirconate titanate (PZT) and lithium niobate. Aluminium nitride is interesting as it should be possible to deposit thin films using microfabrication techniques. Quartz Crystal Microbalance Quartz was the first piezoelectric material to be widely used and originally it was naturally occurring material which was carefully cut to give the desired quality. The first and most widely used sensor is the quartz crystal microbalance (QCM) which is a very thin slice of quartz with thin film electrodes on either side. These are driven with an AC signal which sets up a resonant standing wave in the crystal. This resonance has a very high quality factor, meaning a very narrow bandwidth around the resonant frequency. The resonance can be detected with the same electrodes. The graph in slide 26 shows a common way of characterising resonant devices like the QCM by measuring the conductance (really admittance for AC) against frequency. This gives you the characteristic peak at the resonant frequency with a bandwidth w. The quality factor (Q) is the resonant frequency divided by the bandwidth and gives a measure of the dissipation of the resonant energy. The resonant frequency is sensitive to the mass of the QCM and so they are commonly used as thickness monitors in deposition processes. If the material deposited is very thin or has similar mechanical properties to the QCM the frequency will change in proportion to the deposited mass. If the material deposited is soft or viscoelastic then the Q factor will change as well as the energy dissipation is increased. With very thin layers the mechanical properties of what you re putting down don t really matter and it s possible to sense mass changes using frequency alone. As the thickness increases the viscoelastic properties become more important. Biological materials and use in liquid will increase energy dissipation and change the Q-factor as stated before. CQM-D is a patented technique from a company called Q-Sense, which looks at both resonant frequency and the dissipation factor D (1/Q). This makes the QCM sensitive to the mechanical properties of the deposited film as well as just the mass and is targeted at biological sensing. This is important when investigating chemical attachments in a liquid environment where there is significant damping. The equivalent circuit of a QCM, at a frequency close to resonance is shown in slide 28. It has two parallel arms where the left side has a capacitance CP representing the physical parallel plate capacitor formed by the QCM, plus any parasitics from the connections. The right hand side represents the resonant properties of the piezoelectric crystal. The inductance L and capacitance CS cancel out at the resonant frequency to leave the resistor R which represents the dissipative energy losses. If the resistance is low that is equivalent to a high quality factor. Loading the QCM by attaching something to the surface can change all of the resonant circuit elements. The initial paper on the QCM-D technique describes the instrumentation used to measure dissipation in a quartz crystal microbalance driven at its natural resonant frequency [1]. Prior to this the dissipation was often measured by monitoring the amplitude of the oscillation signal in the QCM, which is inversely proportional to D. However, this places limitations on how the microbalance is driven, for example there can be no automatic gain control. The solution detailed by Rodahl et al. used a setup similar to that shown in slide 29. Here the QCM is driven into resonance and once it is oscillating stably a computer controlled relay disconnects the input signal. Simultaneously, an oscilloscope is triggered to capture the output signal from the QCM as the oscillations decay. The rate of decay of the signal
!5 provides a measure of the dissipation factor of the microbalance setup. The inductor after the buffer in the sensing part of the circuit will act as a high pass filter, removing any DC offset introduced when the relay disconnects the drive oscillator. Surface Acoustic Wave Sensors The QCMs we ve been looking at are bulk acoustic wave devices but there is another class of piezoelectric devices that use acoustic waves confined to the surface of a material. The Surface Acoustic Wave (SAW) shown in slide 30 is a Rayleigh wave. This is the type of vibration that causes damage far from the epicentre of an earthquake but can also work with much smaller faster vibrations in thin film materials. The movement here is purely at the surface, the vibrations only reach about one wavelength into the material and the amplitude of Rayleigh waves is normal to the surface. SAW waves can be excited in a piezoelectric substrate using an interdigitated transducer electrode (IDT) fabricated in a thin metal film on the surface. The device shown in slide 31 is known as a SAW delay line where the waves are transmitted along the surface from the transmission IDT to a receiver IDT connected to a load. Another type might be a single transducer sending waves towards metal structures that reflect them back. They are often used to create high Q-factor filters and other RF components for use in things like mobile phones. Mobile devices with wireless connectivity rely heavily on SAW components. The resonant frequency (f) of the SAW device will depend on the acoustic wave velocity (v) of the piezoelectric material and the pitch (d) of the interdigitated transducer such that f = v/d. Typical SAW characteristic frequencies are in the range of 30-500 MHz. Changing the acoustic properties of the substrate are what turns a simple delay line into a sensor. The primary interaction mechanisms are those that affect the frequency or the delay between the transducers by changing the wave velocity, the IDT distance, or both. Temperature, strain, pressure, force, and properties of added surface materials are examples of quantities that can be measured. In particular, the accumulated surface mass produces a decrease in frequency so they can be used as deposition monitors in a similar way to the QCM. Biosensors can require a different type of SAW as Rayleigh waves are dissipated in liquid contact. Shear-Horizontal SAW waves require the cut of the piezo crystal to be orientated such that the induced waves are parallel rather than normal to the surface. Another alternative is the Love-wave mode of SAW. Here a top coating of a material with low acoustic wave velocity is applied which acts as a wave guide and prevents attenuation into the substrate or the surrounding liquid. Most SAW biosensors are going to operate by detecting mass loading as molecules attach to the prepared surface. By running two transducers in parallel where only one has the chemically/biologically selective surface you can compensate for cross-correlations with temperature or strain. As the devices have a characteristic resonant frequency they can be driven into oscillation by an amplifier connected between input and output. The sensor output is the difference in frequency between the reference and sensor device. The reference device may have an additional blocking layer on top to prevent non-specific attachment. SAW devices are also widely used in passive RFID tags. With an antenna attached to the IDT they can pick up pulses of EM radiation from an RF transmitter. This excites a pulse of SAW energy that is reflected back by reflecting elements spaced like a barcode. The antenna retransmits the reflected pulses and by measuring the timing the barcode ID can be worked out. A wireless SAW sensor would probably have a single reflector and the delay of the pulse would carry information about the state of the sensor. You can even have a differential
!6 measurement by having a double ended SAW sensor with the IDT in the centre. Only one SAW transmission area would be sensitive to whatever we re interested in so there would be a reference pulse to make the timing easier to interpret. Reference: [1] M. Rodahl et al., Quartz crystal microbalance setup for frequency and Q-factor measurements in gaseous and liquid environments, Review of Scientific Instruments, vol. 66, no. 7, p. 3924, 1995.