MEMS Pressure Sensors- An Overview of Challenges in ... · PDF fileMicrosystem packaging...

INSTITUTE OF SMART STRUCTURES AND SYSTEMS (ISSS JOURNAL OF ISSS

J. ISSS Vol. 1 No. 1, pp. 1-10, Sept 2012. REGULAR PAPER

Journal of ISSS 39

MEMS Pressure Sensors-An Overview of Challengesin Technology andPackaging

K. N. Bhat and M. M. NayakCentre for Nano Science and Engineering,Indian Institute of Science,Bangalore-560012

AbstractPressure sensors are required in all walks of life, irrespective of civil-ian, defense, aerospace, biomedical, automobile, Oceanography ordomestic applications. Naturally, rapid progress has been made inmicromachined pressure sensors and the microsystems using thesesensors. Starting with metal strain gauges and moving forward withsilicon based pressure sensors with flat diaphragms, the search fordevices which can operate in harsh environments involving corro-sive fluids and high temperatures has spurred activities which leadto pressure sensors using harder materials such as Silicon Carbide(SiC) and Carbon Nano Tubes (CNT). The present article providesan overview of these pressure sensors including their design, engi-neering, technology and packaging challenges.

INSTITUTE OF SMART STRUCTURES AND SYSTEMS (ISSS) JOURNAL OF ISSS

J. ISSS Vol. 2 No. 1, pp. 39-71, March 2013. REGULAR PAPER

Available online at www.isssonline.in/journal/02paper05.pdf 39

Keywords:

Pressure sensors(Low pressure, High pressure),Microsystems, MEMS,Silicon Carbide,Piezo resistors (Si, PolySi, SiC, CNT),MOS integrated pressure sensors,Microsystem packaging technology.

1. Introduction

Pressure sensors in their primitive form existedas strain gauges for over several decades. Theminiaturization of pressure sensors and othermechanical sensors gained considerable attentionsoon after the invention of piezoresistivity in siliconand germanium [Smith, 1954]. This activity gainedfurther momentum with the recognition of theexcellent mechanical properties of silicon [Peterson,1982]. The advent of the micromachining of siliconto carve out mechanical microstructures in siliconand the already existing expertise in manufacturingmicroelectronic devices and integrated circuits insilicon, opened the doors of the highlyinterdisciplinary area of MEMS and microsystems.During the past two decades, several industries andacademic institutions all over the globe have beeninvolved in the development and commercializationof micro sensors for industrial, automobile, defense,space and biomedical applications. Among thevarious devices, pressure sensors using MEMStechnology have received great attention because

the pressure sensors find applications in everydaylife involving sensing, monitoring and controllingpressure, and they therefore constitute 60 to 70percent of the market amongst the various MEMSdevices.

As the requirements widen, new challengesemerge because the pressures range from a fewPascal (Pa) to several Mega Pascal (MPa)depending on the application and the environment,which vary from being very sensitive in biomedicalapplications to being very harsh in industrial andautomobile applications. Hence, they need to bebiocompatible in some applications while they needto be rugged and capable of performing reliably intemperatures well in excess of 80°C to 100°C.Theyalso need to survive in corrosive fluids like oceanwater in applications such as Oceanography. Inseveral applications, such as mapping the pressureon the aero foil of an aircraft, the package needs tobe flat and the device height needs to be restrictedto below a millimeter. Similarly, in biomedicalapplications such as an intra cranial pressure (ICP)

40INSTITUTE OF SMART STRUCTURES AND SYSTEMS

sensor, where the sensor is inserted into theventricle, the packaged size should not exceed adiameter of 1mm.

As a result of these several constraints, thepackaging of pressure sensors is not a universaltechnique, and has to be tailor-made depending uponthe application. These restrictions on the size ofthe finished device also impose tremendousconstraints on the pressure sensor chip design andfabrication. The demand for pressure sensors overa wide range of pressures varying from a few Pato several MPa, operation capability at temperaturesfrom -25 C to +125C for aerospace applicationsand the growing demand for pressuremeasurements in high-temperature environments(>500 °C) have spurred the development of robust,reliable MEMS-based pressure sensor technologiesinvolving silicon, Silicon on Insulator (SOI), Siliconon Sapphire (SOS), Silicon Carbide (SiC) andCarbon Nanotubes (CNT). As a result, theconstraints and the capability requirements on theruggedness of the microsystem packagingtechnology have been tremendous.

This paper aims at providing an overall scenarioof pressure sensor technology, beginning from basicprinciples. This is followed by design criteria fordifferent ranges of pressures and the relatedtechnology, based on silicon as the diaphragmmaterial and piezoresistor, SOI and polysiliconpiezoresistors for higher temperatures and highpressures followed by more exotic materials likeSiC and CNT for capacitive and piezoresistivepressure sensors. The paper elucidates the use ofbossed structures, which result in sculptureddiaphragms for low-pressure applications withgreater linearity. The paper also gives a glimpse ofsome of our work which has been supported by theNPMASS program at the Centre for Nano Scienceand Engineering (CeNSE ) at IISc and includes asection on the challenges of reliable pressure sensorpackaging technology for operation in harshcorrosive environments.

2. Pressure Sensor Types and Classification

Pressure sensors are categorized as absolute,gauge and differential pressure sensors based onthe reference pressure with respect to which themeasurement is carried out. Particular applicationsare as follows.

(a) Absolute Pressure Sensors measure thepressure relative to a reference vacuumencapsulated within the sensor as shown in Figure1. Such devices are used for atmospheric pressuremeasurement and as manifold absolute pressure(MAP) sensors for automobile ignition and airflowcontrol systems. Pressure sensors used for cabinpressure control, launch vehicles, and satellites alsobelong to this category.

Fig. 1 Schematic diagram of Absolute Pressure sensor

(b) Gauge Pressure sensors measure pressurerelative to atmospheric pressure. One side of thediaphragm is vented to atmospheric pressure asshown in Figure 2. Blood pressure (BP), intra-cranialpressure (ICP), gas cylinder pressure and most ofground-based pressure measurements are gaugepressure sensors. Vacuum sensors are gaugesensors designed to operate in the negative pressureregion.

Fig. 2 Schematic diagram of Gauge Pressure sensor

(c) Differential Pressure Sensors measureaccurately the difference ΔP between twopressures P1 and P2 across the diaphragm (withΔP << P1 or P2 ), and hence need two pressureports as shown in Figure 3. They find applicationsin airplanes used in warfare. They are also used inhigh pressure oxidation systems where it is required

MEMs Pressure Sensors-An Overview of Challenges in Techology and Packaging



Journal of ISSS 41

to maintain an oxygen pressure ranging from 1 to10 atmospheres inside a quartz tube during theoxidation of silicon. In this system, the outside ofthe quartz tube is maintained at a slightly highergas pressure of nitrogen, and the pressure differenceis monitored using a differential pressure sensorwhich ensures that the quartz tube does notexperience a differential pressure greater than itsrupture stress of 1 atmosphere (105 Pascal). Thedifferential pressure sensor is also used in someapplications where it is desirable to detect smalldifferential pressures superimposed on large staticpressures.

In almost all types of pressure sensors, thebasic sensing element is the diaphragm, whichdeflects in response to the pressure. As thedeflections in diaphragm-based sensors are smallthey cannot be directly measured. This mechanicaldeflection or the resulting strain in the diaphragm isconverted ultimately into electrical signals usingsuitable transduction mechanisms, namely,capacitive, piezoresistive or piezo-electrictechniques, which are usually employed asadjectives for the pressure sensors as describedbelow:

(i) Strain gauges and Piezoresistive pressuresensors: In traditional metal diaphragm-basedpressure sensors, the most common method hasbeen to locate metal strain gauges (foil type) on themetal diaphragm, in positions of maximum stress tomaximize the sensitivity. With the invention of piezo-resistivity in Silicon, and silicon micromachining fordiaphragm realization, boron-doped siliconpiezoresistors have replaced the metal strain gauges.In this approach, much higher sensitivities have beenachieved because the piezo-resistors are embedded

Fig. 3 Schematic diagram of Differential Pressuresensors

directly on the silicon diaphragm by implanting ordiffusing boron in the selected regions of maximumstress as shown in Figure 4(a). These resistors areconnected in the form of a Whetstone Bridge whichgives an output when the resistors are strained underthe action of the pressure sensed by the diaphragm.It will be seen in subsequent sections thatpiezoresistive pressure sensors enable linearoperation over a wide range of pressures. Theyare also simple to fabricate. As a result, they havecaptured the major market of pressure sensorsencompassing the automobile industry, defense,space as well as biomedical applications.

Other transduction techniques are capacitanceand piezoelectric approaches and they are identifiedas capacitive pressure sensor and piezoelectricpressure sensors respectively.

(ii) Capacitive Pressure Sensor: A schematicdiagram of a silicon micro machined sensor of thistype is shown in Figure 4(b).This approach usesthe diaphragm as one electrode of a parallel platecapacitor structure and diaphragm displacementcauses a change in capacitance with respect to afixed electrode. The merits of capacitive pressuresensors are their high sensitivity, which is practicallyinvariant with temperature. However, in this case,an electronic circuit is required to convert the

Fig. 4 Schematic representation of micrmachinedsilicon pressure sensors (a) Piezoresistive (b) capacitive(c) Piezoelectric

Bhat & Nayak


capacitance change into an electrical output. Anadditional disadvantage of this approach is thenonlinear relationship between the capacitance anddisplacement and hence a force-balancing andlinearizing electronic circuit is essential to capturea wide range of pressures.

(iii) Piezoelectric pressure sensors: Silicon doesnot show a piezoelectric effect. Therefore, apiezoelectric sensing element, such as LeadZirconate Titanate (PZT) or Zinc Oxide (ZnO) areplaced/deposited on to the silicon diaphragm asshown in Figure 4(c). The deflection of thediaphragm induces strain in the piezoelectricmaterial and hence a charge is generated. Thesesensors are only suitable for measuring dynamicpressures and are not suitable for static pressuresensing because piezoelectric materials onlyrespond to changing strains. The major advantageof this approach is that an external power supply isnot required.

In another approach, in which pressure sensorsare identified as resonant sensors, the vibrationfrequency of a mechanical beam or a membrane,which depends on the extent to which it is stretched,is used This is similar to the vibration frequency ofa violin string. The output signal from a resonantsensor is a frequency, which can easily betransferred into a digital signal and interfaced withcomputer systems, without having to use an Analogto Digital converter. In most of the cases, a quartzresonating beam is used because these sensors arenoted for their high stability and high resolution as afrequency signal is much more robust than anamplitude (e.g. a voltage). The stability isdetermined only by the mechanical properties ofthe resonator material, which is generally verystable. On the flip side, resonant silicon sensors arenot easy to fabricate and hence become expensive.The high costs may be compensated by innovativesimpler mechanical structures and by simplerelectronics.

In all the types of pressure sensors, thediaphragm is invariably, if not always, chosen asthe sensing element. The static and dynamicperformance characteristics of the diaphragm areof great importance and provide suitable designguidelines for designing pressure sensors. Theseaspects of micromachined diaphragms arediscussed in the following section.

3. Static and Dynamic Analysis ofMicromachined Si Diaphragms

3.1 Linear Range of Operation

The diaphragm design is the most crucial stepamong the various stages of pressure sensorrealization. The dimensions of the diaphragm needto be chosen to ensure linearity of operation overthe entire pressure range of operation of the sensor.In most situations, the diaphragms of the pressuresensors turn out to be square or rectangular in thelateral direction while being rigidly anchored at theedges, as shown in Figure 5. Consider a squarediaphragm of thickness h and side length = 2a,subjected to a uniform pressure P. From the theoryof plates [Meleshenko, 1997], the maximumdeflection at the center of the diaphragm is givenby the equation,

Where E is Young’s modulus and g1 and g

2

are constants related to Poisson’s ratio,ν , by therelation

Substituting v = 0.3 for silicon, g1 and g

2 turn

out to be 4.54 and 2.33 respectively. Thus themaximum deflection w

o is linearly related to pressure

P till wo <<h. The second term inside the bracket is

about 0.5 % of the first term whenw

o = 0.1h and the deflection w

oin the linear region

of operation can be expressed as follows for asquare diaphragm

Using (2) for a square diaphragm, of silicon(E=170GPa), with side length 2a = 500 μm andthickness h=10 μm, the maximum deflection isestimated to be 0.5 μm when P=105 Pascal = 1bar.

The maximum stress σmax

, which occurs inthe middle at the edge of the square diaphragm (ie,

(1)

(2)




Journal of ISSS 43

at x = ± a in Figure 5), can be expressed by theanalytical expression for a square diaphragm as,

The stress distribution on the top surface of arectangular diaphragm of side lengths 2a and 2bshown in Figure 5 are estimated with FEMtechniques using the commercial simulator ANSYS.The results of a simulation study [Bhat et al., 2005]undertaken for a rectangular diaphragm, with theaspect ratio L/W ranging from 1 to 2, are presentedin Figure 6, which shows the longitudinal stresscomponent along the x-direction as x is varied from-a to + a along the dotted line shown in Figure 5. Inthis study the width w = 2a is kept fixed at 2a = 500μm and the diaphragm thickness h is held at10μm.The simulation was carried out with thediaphragm subjected to a uniform pressure P=3bardirected perpendicularly into the top surface of thediaphragm as shown in Figure 5. It is evident, fromthe results in Figure 6, that the x-component of thestress is tensile and is greatest at the middle of thediaphragm edges. This stress component decreasesas one moves towards the center of the diaphragm,becomes zero and then changes to compressive

pressure sensor, the locations of maximum tensileand compressive stress on the diaphragm surfaceare important to achieve the maximum sensitivityby placing the resistors in those locations. Figures8 and 9 show these longitudinal and transversestress components respectively at the diaphragmedge (x=±a, y=0) and at the center ( x=0, y=0) for

Fig. 5 Micromachined rectangular diaphragm of siliconshowing the cross section and the diaphragm plan view

stress at the center of the diaphragm where thecompressive stress takes a maximum magnitude.Similarly, Figure 7 presents [Bhat et al., 2005] thetransverse stress component in the y-direction, asx is varied from -a to +a . For a piezoresistive

Fig. 6 ANSYS simulation results on longitudinal Stressacross a diaphragm for different aspectratios [Bhat et al, 2005].

Fig. 7 ANSYS simulation results on transverse Stressacross a diaphragm for different aspect ratios [Bhat etal, 2005]

Fig. 8 Longitudinal and transverse stress in he middleat the diaphragm edge [Bhat et al, 2005 ].

(3)

Bhat & Nayak


different magnitudes of the diaphragm length, Lranging from 500 μmto1000μm [Bhat et al., 2005],for the case h=10μm and the width, W=2a=500μm,when the uniform applied pressure is P=3bar. Inboth the figures, the positive signs indicate that thestress is in the same direction as marked in the insetin these figures. The negative signs indicate thatthe stresses are opposite in direction to that markedin the inset figure.

Fig. 9 Longitudinal and transverse stress at the centerof the diaphragm [ Bhat et al, 2005].

The analysis above demonstrates that for agiven diaphragm thickness h, and pressure, thelongitudinal tensile stress component (indicated bythe positive sign) at the diaphragm edge as well asthe longitudinal compressive stress (indicated by thenegative sign) at the diaphragm center (x=0, y=0)increase with an increase in the aspect ratio (L/W)of the diaphragm, and that both of them almostsaturate when L/W is 2. These criteria need to beconsidered while designing the piezoresistor size andits location on the chip.

3.2 The Proof Pressure and Burst Pressure

The proof pressure is normally defined as 1.5times the nominal pressure of the sensor. The sensoris required to operate up to this pressure whilemaintaining the overall specifications. Burst pressureis another important design consideration for thediaphragm dimensions, because this limits theultimate stress to which the diaphragm can besubjected. This is the pressure at which themaximum stress σ

max on the diaphragm becomes

equal to the critical stress σc which is actually the

yield strength of the material. For the case of asingle crystal silicon, σ

c=7GPa. Thus, for a square

diaphragm having side length 2a and thickness, h,

the burst pressure PB

is determined bysubstituting σ

max=σ

c in equation (3) and can be

written as in (4).

As the magnitude of maximum stress is higherin rectangular diaphragms, the burst pressure of thediaphragms having (L/W) > 1 is lower than thecorresponding square diaphragm having (L/W) =1.This can, indeed, be seen from the ANSYSsimulation results [Bhat et al., 2005] presented inFigure 10 for rectangular diaphragms of threedifferent thicknesses with a diaphragm widthW=2a=500μm. On considering a diaphragm ofthickness h=10 microns, the burst pressure isreduced from 95 bar to 65 bar when the diaphragmlength, L, is varied from 500μm to 1000μm.

3.3 Dynamic Response of the Pressure Sensor

Apart from static response and sensitivity, thefrequency response of the pressure sensor is animportant parameter which becomes important insituations where it is required to monitor the changesin pressure over small intervals of time, as in thecase of blood pressure (BP) or Intracranial pressure(ICP) monitoring. The frequency response indicatesthe ability of the sensing system to precisely respondto dynamic changes in pressure. The frequencyresponse is governed by the sensing element whichacts like a spring –mass system. The naturalfrequency of silicon micromachined pressuresensors is generally high because of their small sizeand high Young’s modulus of silicon. Based on thetheory of plates [Timoshenko & Woinowski-Krieger1983], the resonance frequency of a clamped squarediaphragm is given by the relation involving thediaphragm thickness h, width 2a, the materialproperties, namely, Young’s modulus E and thedensity ρ, as follows:

For a square diaphragm with 2a= 0.5mm andthickness 10 μm the resonance frequency isestimated to be 2.15MHz. Thicker diaphragms inhigher pressure sensors give proportionately higherresonance frequencies. Figure 11 shows some of

(4)


(5)



Journal of ISSS 45

Fig. 10 Burst Pressure PB versus diaphragm aspect

ratio for fixed width=500mm [Bhat et al., 2005].

Fig. 11 Resonance frequency of rectangular diaphragmshaving different thickness lateral dimension of1mmx0.5mm

the results obtained using ANSYS simulations ofrigidly clamped rectangular diaphragms of differentthickness. It can be seen from this simulation thatthe resonance frequency of 400kHz can be achievedusing a 1mm x 0.5mm rectangular diaphragm with10μm thickness. This gives the micro machinedsilicon diaphragm excellent inherent dynamicresponse characteristics.

However, when isolation of the diaphragmfrom a pressurized environment is required, astainless steel barrier diaphragm is employedbetween the pressure sensor and the pressurizedmedia. The volume between the steel diaphragmand the silicon sensor is filled with hydraulic oil thattransmits the pressure to the sensor die. Thisadditional barrier between the pressurized mediaand the sensor has a dampening effect and hencelowers the resonance frequency of the sensor as awhole.

4. Piezoresitive Transducer Analyses andDesign

Among the various types of pressure sensors,piezoresistive pressure sensors and piezoelectricpressure sensors work on the principle of convertingthe strain developed on the sensor chip into changein resistance and voltage respectively. Therefore,the sensitivity of these two types of pressure sensorsis governed by the location of maximum stressregions on the chip when the diaphragm is subjectedto pressure. On the other hand, the performance ofcapacitive sensors is judged by the extent of changein the capacitance of a movable electrode withrespect to a fixed reference electrode. Therefore,the operation range and sensitivity of the capacitivesensor is determined (i) by the extent to which theelectrode can deflect, (ii) the ability to minimizestray capacitances and (iii) capacitance voltageconversion circuits.

Due to their simplicity of fabrication and thewide range of pressures, ranging from a few kPato several hundreds of kPa over which they canperform with excellent linearity and accuracy,piezoresistive pressure sensors have receivedmaximum attention and acceptance in automobiles,airplanes, missiles, and rockets as well as inbiomedical applications. Therefore, in this section,we consider the various design criteria andfabrication process steps involved in thedevelopment of piezoresitive pressure sensors.Following the diaphragm design and identificationof the regions of high stress components on thediaphragm, the next important design parameter isthe design of resistors and their dimensions.

In the following sections, we first focus onvarious parameters such as the gauge factor andpiezorestive coefficient and the factors influencingtheir magnitudes, and hence, the sensitivity of thepiezoresistive pressure sensors

4.1 The Gauge Factor and the PiezoresistiveEffect

The piezoresistive effect can be quantifiedusing the gauge factor which is defined as the ratioof the relative change in resistance (ΔR / R) whenthe resistor is subjected to a strain, ε, and isexpressed by the relation,

Bhat & Nayak


The resistance R of a rectangular resistor oflength L, width W, thickness H and the resistivity, r,is expressed by the relation,

When the resistor is subjected to strain, thefollowing relation gives the relative change inresistance (ΔR/R).

Here ΔL, ΔW, ΔH and Δρ are the changes inthe respective parameters due to the strain. If theresistors experience tensile stress along their length,the thickness and width of the resistors will decreasewhereas the length will increase. Using Poisson’sratio, ν, the change, ΔL, in length is correlated tothe change, ΔW in width, and the change, ΔH, inthickness of the piezoresistors by the followingequation

Combining equations (8) and (9), and usingthe definition given in (6), the gauge factor G isobtained as:

where is the strain. The first two terms in

equation (10) represent the change in resistancedue to dimensional changes and are dominant inmetal gauges, while the last term is due to a changein resistivity. Table -1 gives the gauge factors ofdifferent types of strain gauges.

Thus the gauge factors of different types ofstrain gauges can be vastly different. This is mainlydue to the difference in the extent of the resistivity,ρ, change under the influence of strain. For metals,ρ does not vary significantly with strain andPoisson’s ratio, ν, is typically in the range of 0.3 to0.5, leading to gauge factors of only about 2 to 5 inmetal strain gauges. In semiconductor strain

Type of Strain Gauge Gauge Factor

Metal Foil 1 to 5

Thin-Film Metal ≈ 2

Diffused Semiconductor 80 to 200

Polycrystalline Silicon ≈ 30

Polycrystalline SiC 3 to 5

Single crystal SiC 10 to 30

Carbon Nano Tube (CNT) 200 to 1000

Table-1. Gauge factors of different types of strain gauges[Kovacs 1998; Mosser et al., 1991]

gauges, the piezoresistive effect, which causes largechanges in ρ is dominant and hence the gaugefactor is considerably high. Gauge factors up to 200for P-type silicon and up to 140 for N-type siliconhave been reported.

It may be noted that in metal strain gauges,the gauge factor is small and in the range 1 to 5because in metals the change in the resistance dueto strain is mainly attributed to a change in physicaldimensions. On the other hand, gauge factors inthe range of 80 to 200 have been observed indiffused semiconductor resistors. This is attributedto the piezoresistive effect, which results in largechange in the resistivity (ρ) in semiconductors,whereas in metal foils and thin film metals thechange in resistivity is very small. From Table-1, itcan also be seen that Carbon Nano Tubes (CNT)show a very high gauge factor close to even 1000,and this has recently led to research that uses CNTas the piezoresistive element instead of silicon

4.2 Atomistic Model for the PiezoresistiveEffect in Semiconductors

Piezoresistivity in silicon is usually explainedbased on the deformation of energy bands as a resultof stress. This affects the effective mass, andconsequently the mobility of electrons and holes,and hence, modifyies the resistivity. However, aqualitative model based on an atomistic model of asemiconductor provides a more physical explanationto account for the piezoresistive effect in both P-type and N-type semiconductors. In this model, theeffect of scattering due to changes in the mean freepath is considered to account for the change in

(6)

(7)


(10)

(9)

(8)



Journal of ISSS 47

mobility. This model is presented below for bothN-type and P-type materials.

(i) N-type semiconductor: When an N-typesemiconductor resistor is stretched along itslongitudinal direction, the lattice spacing increasesby a small extent causing the lattice scattering meanfree path length to increase, thus reduces theprobability of the scattering of electrons by thelattice atoms, thus leading to an increase in electronmobility. This causes a reduction in resistivity, ρ,which varies inversely as the mobility μ

n as given

by the relation,

where q is the electron charge and n is theelectron concentration. Similarly when the N-typeresistor is subjected to longitudinal compressivestress the mean free path decreases, causing anincrease in the scattering probability, thus leadingto a reduction in μ

nand hence an increase in

resistivity, ρ.

In metals, as the atoms are more closelypacked, the change in the mean free path and hencethe change in mobility is not noticed particularlybecause the electron concentration in metals isseveral orders higher than in semiconductorresistors.

(ii) P-type semiconductor: Consider a P-typesemiconductor resistor, when it is stretched alongits longitudinal direction. In the atomistic model ofsemiconductors the hole transport in a P-typesemiconductor takes place by the jump movementof electrons from an occupied site to a nearbyvacant site on the lattice atom. The distancebetween the lattice atoms increases by a smallextent when the lattice is stretched; hence, the

electron jump to the nearby lattice becomes moredifficult. This, indeed, hampers the hole transportand appears to increase its effective mass andhence, lowers hole mobility. This leads to an increasein the resistivity of P-type resistors when subjectedto longitudinal tensile stress. By the same argument,it can be seen that the resistivity of the P-typesemiconductor resistor falls when subjected tolongitudinal compressive stress. The resistivity ofthe P-type semiconductor is given by equation (11)replacing electron concentration n with holeconcentration p, and electron mobility with holemobility .

4.3 Piezoresistive Coefficients

The resistance change can be calculated as afunction of stress using the concept of thepiezoresistive coefficient. Contributions to theresistance change come from longitudinal stress(σ

L) and transverse stress (σ

t) with respect to the

current flow. Assuming that mechanical stressesare constant over the resistors, the resistancechange ΔR with respect to the resistance R is givenby,

Silicon type Orientation of resistor πππππ1(10-11 / Pa) πππππt

(10-11 / Pa)

P-type In <100> Direction 0 0

P-type In <110> direction 72 -65

N-type In <100> Direction -102 53

N-type In <110> direction -32 0

Table-2 Piezoresistive coefficients for (100) Si and doping density less than 1018 / cm3 [Middlelhoek & Audet,1989]

Where πL

and πt are the longitudinal and

transverse piezoresistive coefficients, respectively.

Table-2 shows the piezoresistive coefficients[Middlelhoek & Audet, 1989] for N-type and P-type {100} silicon and doping levels below 1018

cm-3. The values decrease at higher dopingconcentrations. For {100} wafers, the piezoresistivecoefficients for P-type elements are maximum inthe <110> directions and are zero along the <100>directions. Therefore, P-type piezoresistors must

(12)

Bhat & Nayak

(11)


be oriented along the <110> directions to measurestress. Because of the higher piezoresitivecoefficients in P-type resistors, most piezoresistivepressure sensors employ P-type resistors on (100)wafers and align them in the <110.>directions.

5. Single Crystal Piezoresistive PressureSensor with Flat Diaphragms

A micromachined piezoresistive pressuresensor makes use of a square or rectangulardiaphragm of N-type silicon, which acts as thesensing element. Figures 12(a) and (b) respectively,show the cut away isometric view, and the top viewof the piezoresistive pressure sensor. The Four P-type piezoresistors equal magnitude, marked R

1, R

2,

R3 and R

4 for identification, are embedded on to

this diaphragm by ion implantation or by thediffusion of boron. They are then laid out as shownand are connected in a Wheatstone bridgeconfiguration as shown in Figure 12(c). PNjunctions formed by individual P-type resistors withthe N-type diaphragm provide the isolation requiredbetween the resistors. Though it would bepreferable to use circular diaphragms to preventunwanted stress concentration, silicon diaphragmsturn out to be square or rectangular when they arefabricated using anisotropic wet chemical etchingon silicon wafers of <100> orientation. It is alsoeasier to align the resistors parallel andperpendicular to the edges of the diaphragm whichare in the <110> direction, thus ensuring that thepiezoresistive coefficients π

l and π

t are maximum

along this direction.

5.1 Diaphragm Design

In miniaturized pressure sensors, the lateraldimension has an upper limit. For instance, inbiomedical applications such as intra-cranialpressure monitoring, the chip size needs to be keptbelow an upper limit of 1mmx1mm. This limits thediaphragm size to below 500μm. Once this lateraldimension is finalized, the thickness of thediaphragm can be chosen based on linearity andburst pressure conditions.

As discussed in Section-3, the maximumoperating pressure is limited by two parameters (a)the linearity limit and (b) the burst pressure limit.The linearity limit can be arrived at using equation(1). In this equation, the first term is due to the

stress distribution caused by pure bending whenthe central plane of the diaphragm is not stretchedor compressed. This is true if the deflection,w

o, of

the diaphragm is small compared to its thickness,h. If w

o is not small when compared to h, the central

plane of the diaphragm will be stretched like aballoon. The second term in this equation representsthe stress caused by the stretching of the centralplane and makes the deflection nonlinear withrespect to the pressure. Therefore, even though areduction in thickness would increase deflectionw

o, and hence, the sensitivity of the pressure sensor,

it is necessary to keep h sufficiently thick so thatw

o at the maximum operation pressure is small

when compared to h. The burst pressure limit fora square diaphragm is determined using equation(4). This limit would come into the picture at a muchhigher pressure compared to the linear limit on thethickness for a given size of the diaphragm. It mayalso be noted from Figure 10 that for a giventhickness of the diaphragm, the burst pressurebecomes lower when the diaphragm aspect ratiob/a is increased. In general, the diaphragmdimensions (thickness and aspect ratio) are sochosen that the proof pressure is below one fifthof the burst pressure (P

B/5), further guided by the

permitted nonlinearity.

5.2 Positioning the Piezoresistors for BestSensitivity

The proper positioning of the piezoresistorson the diaphragm needs to be chosen to achievethe best sensitivity with the diaphragm dimensionsdesigned according to linearity and burst pressureconsiderations. In the case of a single crystalpiezoresistive pressure sensor, the four diffused orimplanted P-type piezoresistive transducerelements are embedded near the edge of thediaphragm, as shown in Figure 12(b), where thestress is maximum as seen by the simulation resultsin Figure 6. Here, the two piezoresistors R

1 and R

3

are placed parallel to opposite edges of themembrane, and the other two resistors R

2 and R

4

are placed perpendicular to the other two edges.When a uniform pressure is applied on the backsurface as shown in Figure 12(a), the diaphragmdeflects upwards, causing compressive stress atthe edges of the membrane surface. This causesthe resistors R

1 and R

3 to experience transverse




Journal of ISSS 49

Fig. 12 Silicon micromachined pressure sensor

compressive stress and the longitudinal tensilestress given by the relations (the negative sign onis indicative of compressive stress).

Hence they show an increase in resistance ΔRgiven by the relation,

Similarly the resistors R2 and R

4 laid out

perpendicular to the edge of the diaphragmexperience longitudinal and transverse stressesgiven by,

and they show a decrease in resistance given bythe relation

In single crystal (100) silicon, πl and π

t are

equal to each other in magnitude. Therefore, themaximum sensitivity can be achieved when the fourresistors fabricated by diffusion on to a single crystalsilicon diaphragm are arranged in locations as shownin Figure 12. All the four resistors, (R

1, R

2. R

3 and

R4), are chosen to be equal in magnitude so that the

bridge is balanced when the v 0pressure is zero.When the diaphragm is subjected to uniformpressure, P, an imbalance is created in the bridgewhich gives rise to an electrical output v

0of polarity

shown in Figure 12(c) for the applied voltage Vin

and is given by the relation,

Where R is the zero-stress resistance of eachof the resistors.

Equation (17) is derived based on thesimplifying assumption that the magnitude oftransverse stress on the resistors marked R

1 and

R3 is the same as the magnitude of longitudinal

stress on the resistors marked R2 and R

4, so that

the ΔR is same in magnitude on all the four resistors.In practice, this is not true because, as can be seenfrom Figure 6, the average longitudinal stress onthe resistors R

2 and R

4 would decrease along its

length considerably, whereas the variation in thestress along the length of the resistors R

1 and R

3 is

rather small. The stress variation along the lengthof all the resistors is minimized by splitting each ofthe resistors into resistors of shorter length adjacentlylaid out and connected in series as shown in Figure12(b). It can also be seen that the resistors areterminated on heavily doped P-type layersdesignated P+ regions, to reduce the contactresistance between the resistor and the aluminummetal contact

Typically most commercial pressure sensorsare operated at supply voltagesV

in = 5 Volts.

Assuming the typical values of π1= 70 x 10-11 /Pa

and ν=0.3 for a single crystal silicon, with a pressuresensor having a square diaphragm of 2a = 1000μm,h=25μm, the pressure sensor output voltageestimated using (17) is approximately equal toV

o = 117mV, when P=1.2bar = 1.2x105 N/m2.(Note

that 1N/m2 =1 Pascal).

6. Performance Characteristics of PressureSensors

The main parameters and specifications of thepressure sensor are sensitivity in the linear rangeof operation, offset voltage, maximum V

0 in the

linear range of operation, non-linearity for aspecified maximum pressure of operation, thetemperature coefficient of resistivity (TCR) and thetemperature coefficient of sensitivity (TCS),hysteresis and creep. For the sake of clarity, theexperimentally measured output characteristics ofa pressure sensor having resistors fabricated on to

(13)

(14)

(15)

(16)

(17)

Bhat & Nayak


a polysilicon layer on oxide over a diaphragm isshown in Figure 13, to understand the variousparameters and specifications. These outputcharacteristics show a non zero output voltage, V

o,

when the pressure is zero, followed by a linear Vo

versus pressure region, and then a non-linear regionbeyond a certain pressure and are detailed below.

Fig. 13 Typical Output characteristics of apiezoresistive Pressure sensor, showing the Offsetvoltage and the nonlinear behavior beyond 2.0 bar

6.1 Offset Voltage

The output voltage of the pressure sensor atzero input pressure is designated as the offsetvoltage. This is due mainly to two reasons. Thefirst one is due to some residual stress on themembrane. The second main reason is thevariability in the four resistors. Even though theresistors are processed by diffusion simultaneously,there are some variations due to non-uniformity inthe starting polysilicon layer or due to non-uniformityin dopant diffusion in the polysilicon resistor. Evena 5 Ω difference in the 1kΩ resistance values cancause an imbalance in the Wheatstone bridge andresult in an output offset voltage of 50 mV for aninput voltage of 10 Volt. Therefore, piezoresistivepressure sensors invariably show offset voltage.In Figure 13, this is marked and is equal to 70mVfor an input voltage of 10V. One of the approachesused for offset voltage compensation has been toconnect the external resistors. This involves the useof an open bridge configuration and the use ofexternal precision resistors to complete the bridgeduring the packaging stage. Over the years, moreelegant and efficient techniques of compensatingfor the offset voltage using electronics haveevolved.

6.2 Sensitivity, S

The sensitivity of a pressure sensor is defined

as at a particular input voltage.

Considering the device characteristics shown inFigure13, the V

0 increases from 70mV to 210 mV

linearly as the pressure is increased from 0 to2.0bar. The sensitivity for this device is

mV/bar and this is equal to 70mV /bar

when the input voltage VIN

=10V The sensitivity israther low for this device because the resistors inthis case were fabricated using polycrystallinesilicon whose gauge factor is low compared to thatof single crystal diffused resistors. Highersensitivities and lower noise levels are typicallyachieved with diffused or ion implanted single crystalsilicon strain gauges.

6.3 Nonlinearity and the Ballooning Effect

Nonlinearity is a key parameter of pressuresensors and it is specified as a percentage of thefull-scale output voltage of the sensor. Referring tothe experimental characteristics shown inFigure 13, the full scale output voltage V

om = 435mV

at Pm = 6 bar and Voff

= 70mV when P=0. Thenonlinearity NL

i of this experimental pressure

sensor at any intermediate pressure Pi is defined

with reference to the end point straight line joiningthe point V

o = V

om at P

m = 6 bar and at P= 0 by the

following equation,

In this expression, Voi is the sensor output

voltage at pressure Pi and V

oL is the value of V

o that

would have occurred if the Output characteristicscoincided with the end point straight line. It is givenby the relation,

It is seen from Figure 13 that at, Pi = 3bar,

Voi = 275 mV and V

oL = 230mV. The nonlinearity is

. However, when the Pm

recommended for the same pressure sensor is

(19)


(18)



Journal of ISSS 51

changed to 3 bar, the corresponding Vm = 275mV

and the end point line is as shown in Figure 14. Thecorresponding values of V

oi and V

oL are respectively

215mV and 205mV at a pressure Pi=2bar. Hence,

the nonlinearity at Pi=2bar with respect to the

corresponding to V0

is considerably lower and isgiven by,

Fig. 14 Typical Output characteristics of apiezoresistive Pressure sensor, showing that thenonlinearity is lower if is chosen to be 3bar

Thus if the range of operation is restricted tothe linear range the maximum nonlinearity withinthe operating range remains at an acceptable level.

Occasionally, one may encounter negativeoffset voltage. In such cases the above definitionholds good and the sign should be appropriately takeninto account. The nonlinearity in the piezoresitivepressure sensors is caused mainly by the followingfactors:

(i) The nonlinear relationship between the stressand the pressure applied. As discussed in section 3,if the deflection of the diaphragm is large comparedto its thickness, the central plane of the diaphragmstretches like a balloon. Due to this balloon effect,the diaphragm is subjected to a stretching stresscomponent, σ

s, in addition to the stress, σ

b caused

by the bending of the diaphragm. The stress,σb,

caused by bending, is reduced in magnitude as thestretch of the diaphragm takes part of the pressureload and this results in nonlinearity. The nonlinearitycaused by the balloon effect (ie the stretch) issmaller when the sensor is subjected to pressurefrom the front side where the resistors are located.

This is because σs is always positive irrespective

of its position in the diaphragm and the direction ofthe applied pressure, whereas the polarity of σ

b

can be either positive or negative depending on itsposition in the diaphragm and the sign of the appliedpressure. Thus, both σ

s and σ

b are positive at the

diaphragm edge when the pressure is applied fromthe front whereas σ

b is negative and σ

s is positive

when the pressure is applied from the backside.Hence when the pressure is applied from the frontside the stresses add up and the total stress tendsto be closer to the linear theory which assumes thatthe stress distribution is a result of pure bending.

(ii) The piezoresistive coefficient of silicon isgenerally considered to be independent of stress.In practice, this is not really true when examinedwith high accuracy. The nonlinear relationshipbetween the piezoresistive coefficient and the stressis thus another source of nonlinearity in piezoresitivepressure sensors.

(iii) The third cause of the nonlinear output voltageis due to the difference in piezoresistive sensitivitybetween the resistors of the Wheatstone bridge.

For low- pressure sensors, linearity becomesan issue. Bossed diaphragms or sculptureddiaphragms, as will be discussed in subsequentsections, are used to overcome this problem.

6.4 Hysteresis

Hysteresis is yet another parameter ofpressure sensors and this is also specified as apercentage of the full-scale output voltage of thesensor. This parameter is a matter of concern inpressure sensors, which employ metaldiaphragms, because of the non-elasticcharacteristics of ductile metals, even with verygood spring materials. However, single crystalsilicon is an excellent spring material. Attemperatures below 600°C, the silicon stress versusstrain curve has no plastic zone and the materialhas essentially no creep. Pressure sensorsemploying a silicon diaphragm as a sensing elementhave hysteresis below 0.1%, which is very low.

6.5 Temperature Coefficients of Resistivity andSensitivity (TCR and TCS)

Piezoresistivity is a consequence of the changein resistance due to a change in the mobility of

Bhat & Nayak


carriers due to pressure-induced strain. It is wellknown that resistivity is also a function oftemperature due to a reduction in mobility at highertemperatures. The TCR is always positive indiffused single crystal resistors. The TCR ofpiezoresistors affects the sensitivity of the pressuresensor and leads to TCS which we show below tobe always negative.

Consider the Wheatstone bridge circuit shownin Figure 12, which is balanced at room temperatureand at zero pressure difference across thediaphragm. Due to the positive TCR, all theresistance values increase by ΔR

T when the

temperature goes up by ΔT . This, alone, would notaffect the output voltage because all the resistorvalues go up equally by ΔR

T. However, due to a

change in the pressure, the P-type resistances R1

and R3 in Figure 12 increase by ΔR, and the

resistance values R2 and R

4 decrease by ΔR

assuming them for simplicity to be the same inmagnitude when pressurized from the backside ofthe diaphragm. The combined effect of pressure Pand the temperature rise ΔT on the V

0 can be

expressed as given below.

The temperature dependent term ΔRt=Rα ΔΤwhere α is the TCR of the resistors R. Substitutingthis in equation (20),we have,

In this equation from (16)

and hence

and the sensitivity can be expressed as

where So is the sensitivity when the temperature

is constant. Generally, as αΔT <<1, (23) can bewritten as,

Equation (24) shows clearly that TCS inpiezoresistive pressure sensors is reduced and thatthe extent of reduction in sensitivity depends uponTCR and the temperature change. In diffused singlecrystal resistors, TCR (=α) is always positive[Mosser et al., 1991] with its magnitude varyingfrom 1x10-3 / °C to 2x10- 3 /°C for dopingconcentrations in the range of 1018 / cm3 and 1020 /cm3. Assuming the value of α = 0.001/°C, whenthe temperature of the sensor goes up by 100°C,αΔΤ = 0.1 then using this in (24), S= 0.9 S

0

indicating a 10% fall in sensitivity.

The TCS can cause severe errors in thepressure being monitored. The magnitude of theerror will be severe when ΔR

T is comparable to the

change in the resistivity ΔR caused by the pressure.Commercial ASIC chips are used to compensatefor the TCS and the offset voltage. Anotherapproach to minimize this effect is to usepolycrystalline silicon (poly-Si) piezoresistors withthe TCR tailored suitably by altering the dopingconcentration [Mosser et al., 1991].

7. Polycrystalline Piezoresistive PressureSensors

In conventional single crystal siliconpiezoresistive pressure sensors, the four P-typepiezoresistors constituting the Wheatstone bridgeare embedded on an N-type silicon diaphragm eitherby ion implantation or a diffusion of boron process.Thus the resistors are isolated from each other onthe diaphragm by PN junctions which can beeffective only till about 100 °C. Yet another issuewith this approach is that the resistors need to beisolated from the header on which the device ismounted either by anodic bonding glass to siliconor by Fusion bonding an oxidized silicon wafer. Theabove problems can be sorted out by realizing theresistors on an oxide layer grown on a Silicon OnInsulator (SOI) wafer.

7.1 Polycrystalline Piezoressitive PressureSensor On SOI Wafer

In this approach the SOI layer itself serves asthe diaphragm, with the handle wafer etched usinganisotropic etching from the backside. P-typepolycrystalline silicon resistors are realized on theoxide grown on the SOI layer as shown in Figure

(22)

(23)

(24)




Journal of ISSS 53

15 (a), by depositing, doping and patterning thepolysilicon. The deposition of polysilicon is carriedout by a low pressure chemical vapor deposition(LPCVD) process, and the P-type doping isachieved by the ion implantation of boron followedby annealing. With this approach, P-type polysiliconresistors are isolated from each other by the oxidelayer. Therefore, the isolation between the resistorsis maintained even at temperatures in excess of300 °C. The pressure sensor fabrication iscompleted at the wafer level by depositing andpatterning chrome /gold or Al as shown in the topview in Figure 15(b) to connect them in the form ofthe Wheatstone bridge shown in Figure 15(c).

7.1.1 Resistivity and TCR

The resistance of grains and that of grainboundaries determine the effective resistance ofpolysilicon resistors, the latter being the mostimportant aspect. Within the grains, the currenttransport is by carrier drift, and the resistivity ofthe grain region behaves essentially like that of thesingle-crystal silicon resistor. Therefore, it increaseswith temperature due to a reduction in mobility. Onthe other hand, the grain boundaries offer potentialbarriers for carrier transport across them. At highertemperatures, more carriers gain energy sufficientto cross over these barriers, and the grain boundaryresistance decreases with temperature rise. Thebarrier height at the grain boundary is a function ofdoping concentration. Therefore, the temperaturecoefficient of resistivity (TCR) of polysilicon canbe tailored and made almost zero by adjusting theboron doping concentration [Mosser et al., 1991]

so that the positive TCR of the grains can bebalanced by the negative TCR of the grain boundaryregions. The experimental results reported recently[Raman et al., 2006; Manjula, 2005] onboron-implanted LPCVD polysilicon films ofthickness 0.3μm are presented in Figure 16 to showthat the TCR of LPCVD polysilicon resistors canbe negative, approach zero, or be positive dependingon the doping density. It can be seen that theTCR can be reduced to a low value equal to1.75x10-5/°K with the room temperature resistivity,ρ=1.05 x 10-2 Ω - cm for doping concentration N

A

= 2.66 x 1019 / cm3.

7.1.2 Alternate Layout on a RectangularDiaphragm

The multiple orientation of the crystallinesilicon in the grain makes an average contributionin all possible directions to the piezoresistance ofthe polysilicon film. Therefore the piezoresistiveeffect in polysilicon film is isotropic [Xiaowei et al.,1998] and is less than in single crystal silicon. Themagnitude of the longitudinal gauge factor (G

L=30)

of boron-implanted polysilicon is larger than thetransverse gauge factor (G

T = -10) [Mosser et al.,

1991] by a factor of about 3. Polysiliconpiezoresistors therefore are arranged n thediaphragm in such a way that they experiencemaximum longitudinal stress in order to achievebetter sensitivity. The results of a study presentedin Figure 6 in section 3 show that the x-componentof the stress is a compressive maximum at x=±a,y=0and it is a tensile maximum at x=0, y=0 when thepressure is applied as shown in Fig.15. Thesimulation results presented in Figure 8 and 9 alsodemonstrate that the magnitude of these stressesare maximum when the diaphragm is rectangularwith an aspect ratio b/a = 2. The rectangulardiaphragm is deliberately chosen and the resistorsare placed as shown in Figure 15(b) along thenarrower direction and connected in the Wheatstonebridge as in Figure 15(c) to achieve a maximumpossible sensitivity.

Several possible arrangement patterns ofpolysilicon resistors have been studied for differentgeometrical sizes of the membrane and the results[Xiaowei et al., 1998] have shown the following:(1) The arrangement Pattern shown in Figure 15(b)gives the best results in terms of maximizing the

Fig. 15 Polysilicon pizoresistive pressure sensor onoxidized SOI diaphragm

Bhat & Nayak


output voltage of the sensor because, thepiezoresistors take advantage of the longitudinalpiezoresistive effect of polysilicon and placed alongthe narrower direction. (2) In the sensor design,the length to width ratio of the rectangular membraneshould be increased to some degree to enhance theoutput.

7.1.3 Experimental Results

Based on these studies, boron implanted andannealed LPCVD polysilicon resistors are arrangedon the oxide grown on a silicon diaphragm as shownin Figure 15(b) and connected in a Wheatstonebridge configuration as shown in Figure 15(c) onan oxidized rectangular diaphragm of thickness15μm and a lateral dimension equal to500μm x 875μm. The individual devices were diced and mountedon the TO-5 header, wire bonding was carried outand a cap with a pressure port was welded. Figure17 shows the photograph of the packaged devicewith the pressure port. The figure also shows theclose-up photograph of the chip, four wires bondedon to the pads and the photograph of the chipmounted inside the package on the header [Bhat etal., 2006]. The output voltage Vo as a function ofpressure in the range P=0 to 15bar of thesepackaged devices is obtained for two differentapplied input voltages V

in=1V and V

in =2V. The

typical results obtained are shown in Figure 18. Itcan be seen that the characteristics are linear overthe entire range of pressure upto 15 bar. As thegauge factor of polysilicon is rather low, and alsobecause the diaphragm was designed with athickness of 15 μm for operation up to 15barpressure with burst pressure in excess of 100bar,the sensitivity in these devices were found to be inthe range of 1.8mV per bar forV

in = 1.0 Volt. The

temperature sensitivity of these devices was testedin the temperature range 30°C to 70°C and theresults [Bhat et al., 2006] are shown in Figure 19.The variation in sensitivity was found to be withinabout 5% of the room temperature value. Similarresults are seen in Figure 19 on another sensor witha diaphragm thickness of 10μm. The sensitivity ofthis sensor is much higher (6mV/bar) as expectedbecause of the lower thickness. As a result, thetemperature sensitivity of these devices was foundto be slightly lower than 3% because the TCR termΔRT is invariant with the diaphragm thickness

Fig. 16 Experimental results on the effect of borondoping concentration on the TCR of polysiliconresistors [Manjula, 2005].

Fig. 17 Photographs of polysilicon Pressure sensorinside the TO-5 Package and a close up view of thewire boned chip.

Fig. 18 Output voltage versus the input pressuremeasured on the pressure sensor shown in Fig 17.

whereas the pressure dependent term ΔR is higherwith the thinner diaphragm.

These results have indeed demonstrated thatthe TCS of piezoresistive polysilicon pressuresensors can be minimized by suitably doping




Journal of ISSS 55

LPCVD polysilicon resistors such that the TCR isbelow 10-4/C. The SOI approach for realizing thediaphragm, not only provides the high isolationrequired between the sensor and the body of thepackage, it also makes it easy to achieve adiaphragm of uniform thickness over the entirewafer, thus enhancing the reliability andreproducibility of the pressure sensors.

7.2 SOI Approach for Monolithic Integration

The ultimate success of silicon micromachiningin smart systems depends upon the ability tointegrate mechanical components with electronicson the same chip. The possibility of the monolithicintegration of piezoresistive pressure sensors withelectronics has been demonstrated at the IndianInstitute of Technology Madras [Vinoth Kumar etal., 2006; Vinoth Kumar, 2006]. Figure 20 (a) showsthe schematic cross section of the MOS Integratedpiezoresitive pressure sensor designed andfabricated on a single chip. In this approach, thestarting wafer was an SOI wafer and polycrystallinesilicon resistors deposited on oxide were used forachieving better electrical isolation between themas well as between the electronics circuit and thesensor. The sensor output is connected to the inputof a common source differential amplifier circuit

Fig. 19 Temperature sensitivity of two different PolySilicon piezoresistive pressure sensors having thediaphragm dimensions indicated on the plots [Bhat etal., 2006]

Fig. 20 (a) Schematic Cross section (b) Sensor and the amplifier circuit and (c) Composite mask layout of theMOS Integrated pressure sensor. Chip mounted on header and wire bonded (e) Packaged sensor showing thepressure port.

as shown in Figure 20(b). The resistors areembedded over the diaphragm, which is realizedby anisotropic etching, and the electronics circuit islaid out in the rest of the portion on the same chipas shown in Figure 20(a). Integration with fewerprocess steps has indeed been possible by mergingthe process steps of polysilicon piezoresistors withthe resistors of a common source differentialamplifier circuit. The self-aligned polysilicon gateMOSFET technology was used for the electronicscircuit in this chip. The MOSFET integrated

Bhat & Nayak


At this stage, it is to be noted that the aboveexercise of packaging was meant for testingthe integrated sensor at the laboratory. Howeveras will be seen in the subsequent section onPackaging, several requirements need to be metby the packages operation in harsh environmentsand they need to be tailor made to suit specificapplications.

8. Low -Pressure Sensors with SculpturedDiaphragm

As the maximum stress in a flat diaphragmwith lateral dimensions of 2a and thickness isproportional to P(a/h)2, the sensitivity of apiezoresistive pressure sensor fabricated on a flatdiaphragm can be increased by making the (a/h)ratio larger. However, it can be seen from equation(2) that the deflection w

o to thicknessb h ratio is

proportional to P (a/h)4. Therefore, in a low

piezoresistive pressure sensor is fabricated using aseven-mask process. A composite layout of theseven masks is shown in Figure 20(c). In this layout,provision is made for accessing the sensor outputas shown by the pads marked +V

SO and –V

SO

which are connected to the MOSFET gateelectrodes. The supply voltage terminals aremarked +V

DD and –V

DD and the output of the

amplifier is labelled as VO.

The photographs of the integrated chip, thechip die mounted on the TO-39 header and wirebonded, and the final package are shown in Figures21 (a), (b) and (c) respectively. The sensor outputvoltage V

SO and the amplified output voltage V

O

measured in the pressure range of 0 to 7 bar arepresented in Figure 21(d). The on-chip amplifiergain was designed to be about 5. The overallsensitivity of 270 mV /bar with 10V supply voltagehas been achieved with this MOS integratedpiezoresistive pressure sensor.


Fig. 21 Photographs of (a) MOS integrated Pressure sensor chip (b) die mounted on the TO-39 header and wirebonded (c) Packaged sensor showing the pressure port. (d) Output Characteristics versus pressure.



Journal of ISSS 57

pressure range (e.g. a full-scale pressure of about500Pa), flat-diaphragm pressure sensors are notsuitable because the sensitivity will have to beincreased considerably by making the (a/h) ratioextremely large and this would lead to largedeflections resulting in a high degree of nonlinearity.As discussed before, non-linearity is the result ofthe stretching of the middle plane, which becomessignificant when the deflection becomescomparable to the thickness of the diaphragm. Inorder to improve sensitivity and linearitysimultaneously, specialized geometries, such asdiaphragms with a rigid center or boss [Mallon etal., 1990] have been introduced for increasing thestiffness to limit the maximum deflection of thediaphragm, and for enhancing linearity. Such astructure is also known as the sculptured diaphragmor the bossed diaphragm and is schematically shownin Figure 22 [Mallon et al., 1990]. In this approach,the structure is locally stiffened to limit the overalldeflection, while maintaining a relatively thin sectionwhere the piezoresistors are placed. Thus the totalnonlinear deflection due to membrane stress isreduced. Since the deflection and resulting stressoccur at a localized area, and also because the stressis concentrated in relatively localized thin areas ofthe diaphragm, this technique is sometimes referredto as the stress concentration technique. This termis rather misleading because the stress at the edgeof a flat diaphragm of thickness h is found to behigher than the stress on the same diaphragm witha central double boss structure. However, thelinearity of deflection with pressure improvesconsiderably with the bossed structure. Thus, insummary, the sculptured diaphragm structure

Bhat & Nayak

Fig. 22 (a) Pressure sensor with sculptured diaphragm(b) Wheatstone bridge connection of the resistors (notethat R1 and R4 are on opposite arms and R2 and R3 areon the other opposite arms of the bridge

provides lower deflection and better linearity at agiven stress level but does not enhance the stressby concentrating it in a small area.

As can be seen from Figure 23 (a), in a doublebossed diaphragm, all the four resistors aresubjected to transverse stress when the structureis subjected to pressure. When pressure P isapplied to it from the top surface as shown,resistors R

1 and R

4 which are laid on the thin regions

(i) and (iii) of thickness, h , experience identicaltransverse tensile stress and are connected to theopposite arms of a Wheatstone bridge as shownin Figure 23(b). On the other hand, resistors R

2

and R3

which are laid on the thin region (ii) ofthickness, h, experience transverse compressivestress and are connected to the other opposite armsof the Wheatstone bridge. A detailed analysis[Bao, 2000] of this structure has shown that theaverage transverse tensile stress on resistors R

1

and R4 placed in the regions-(i) and (iii),

is=1.35 P(a/h)2 and the average transverse stresson resistors R

2 and R

3 placed in the region-(ii)

is=-1.35 P(a/h)2, the negative sign indicating thatthe stress in this case is compressive. It has alsobeen shown in [Bao, 2000] that the output voltagefrom the pressure sensor and the maximumdeflection w

o can be approximated in terms of the

thickness h of the thin region where the resistorsare placed on the twin –bossed diaphragm and thelateral dimension, a, shown in the Figure 23, bythe following expressions

Comparing vo in eqn (25) with the expression

voin eqn (17) for the flat diaphragm, it can be seen

that the sensitivity of the twin island structure sensoris lower by a factor of 0.75 compared to thesensitivity of a flat diaphragm of uniform thicknessh and the same side length equal to 2a. Similarly,comparing equation (26) with equation (2), it isevident that the w

o/h ratio in this twin island

structure is lower by a factor of 2 than in the flat

(26)

(25)


membrane situation, thus illustrating that linearitycan be maintained at higher pressures for a giventhickness. Alternatively a thinner diaphragm canbe used for better sensitivity and for maintainingbetter linearity compared to a sensor with a flatdiaphragm.

In summary, piezoresistive pressure sensorsusing the sculptured structure give better linearityand, hence, are used as a widely accepted techniquefor sorting out the difficulties faced when used withflat diaphragm sensors for applications in sensingvery low pressures.

An alternative solution for very low pressuresis the capacitive sensing approach. However,there are several problems associated withcapacitive sensing: (1) The stray capacitance needsto be minimized and compensated for (2)Capacitance to voltage conversion is a must (3)Fabrication of capacitive pressure sensors is moreinvolved than the piezoresistive counterpart. (4)Capacitive sensors are more susceptible to ElectroMagnetic Interference (EMI) effect. Consequently,many commercial pressure sensors arepiezoresistive.

Recently, resonant sensors have attractedconsiderable attention because of their ability toprovide high resolution, high stability and an easyinterface with digital circuits. Therefore, during thepast two decades, efforts have been directedtowards realizing resonant pressure sensors. In thisapproach, the change in the resonance frequencyof a resonating structure caused by mechanicalstress is monitored. Even though a few of themhave been commercialized, their use in commercialproducts has been slow because of the complexityin processing and the need for packaging them in avacuum to maintain high quality factor Q. A DigiQuartz resonating beam pressure sensor is beingused as a bottom pressure recorder (BPR) fortsunami sensing applications.

9. Silicon Carbide Pressure Sensors forHarsh Environments

9.1 Silicon Carbide Piezoresistive PressureSensors

The applications of conventional piezoresistivesilicon pressure sensors are limited to below 150°Cdue to the degradation of the PN junction isolation

between the resistors on the diaphragm. Silicon OnInsulator, along with oxide isolated piezoresistiveelements, has enabled their use up to about 300°C.However, the need for pressure measurements inhigh-temperature (>600°C) and harsh environments(automotive and aerospace applications: combustionprocesses or gas turbine control; oil industry;industrial process control; nuclear power) hasspurred the development of robust, reliable pressuresensors. This is mainly because of the poor thermo-mechanical properties of silicon, viz the degradationof the elastic modulus above 600°C, and its inabilityto withstand a corrosive environment. As a result,during the past decade, considerable effort has beendirected towards taking advantage of the superiorthermo mechanical properties of silicon carbide(SiC) to develop micro-pressure sensors that wouldextend the sensing capability to 600°C and beyond.The wide band gap of 3.3eV 6H-SiC enables hightemperature operation of the junctions. The fairlyinert nature of the SiC makes it suitable for use incorrosive environments, and the electronic circuitson SiC are radiation-insensitive. As alreadypresented in Table-1, the gauge factor of singlecrystal SiC is in the range 10 to 30 and inpolycrystalline SiC, it is in the range of 3 to 5. Theproperties of Single crystal SiC are presented inTable-3, and are compared with Silicon and diamondto show that the yield strength , Young’s modulusand hardness of SiC are considerably higher thanthose of Silicon and are closer to those of diamond.

In spite of its several merits, the developmentof SiC pressure sensors has been slow because ofthe difficulties involved in its micromachining.Initially, SiC piezoresistors were fabricated on oxidegrown on a silicon micromachined diaphragm. Inthis approach, silicon becomes the limiting factor.The benefits of SiC material can be fully utilized byfabricating both the piezoresistive sensing elementsand the sensing diaphragm on SiC. This has, indeed,been made possible, as reported in [Ned et al.,2004], by micromachining N-type SiC in the DRIEsystem in a gas plasma mixture containing SF

6 and

Oxygen to realize the SiC diaphragm. SiCpiezoresistors were realized on the SiC diaphragmby chemical vapor deposition of SiC using a gasmixture of SiH

4 and CH

4 and patterning it by RIE.

Ohmic contacts on the SiC resistors were obtainedwith a Ti/Al bi-layer having 125nm titanium thicknessand 1200nm of Aluminum. Both low and high




Journal of ISSS 59

pressure SiC piezoresistive pressure sensorsoperational up to 600°C have been demonstrated[Ned et al., 2004], using relatively thick (60 μm)diaphragms for the 1000 psi sensors, while thediaphragm for the 25 psi pressure sensor wasmicromachined to be significantly thinner withoptimized sculptured sensing diaphragms of the typediscussed in section-8. Although the performancecharacteristics such as sensitivity, linearity andhysteresis were reasonably good, a couple of issuesstill needed to be sorted out (1) A large thermal set(10mv) was observed in all SiC sensors afterexposure to 600°C. (2) The bridge resistance versustemperature showed a non-monotonic resistancechange with temperature, giving rise to a monotonicreduction in resistance up to 350 °C followed by anincrease in resistance. This continues to pose a bigchallenge for the temperature compensation of suchSiC sensors.

An exploratory effort [Vandelli, 2008] hasshown that capacitive SiC MEMS devicesfabricated and characterized showed a repeatablehysteresis less than 2.5% at 300°C. It was alsodemonstrated that the combined effect oftemperature and pressure on non-linearity andhysteresis was never much higher than 1% FSOover an operating temperature range of 0 to 200°C.

According to a SiC road map [Fraga et al.,2010], the development of SiC sensors is based onprogress in (1) the improved electrical andmechanical properties of SiC films produced by theoptimization of the SiC deposition process, (2) SiCfilm processing by the optimization of the etchingprocess and metallization appropriate for hightemperature applications, (3) micro-fabricationtechnology to fabricate miniaturized sensors and (4)sensor packaging for harsh environments. Theimmediate goal is to improve the performance ofSiC pressure sensors and strain gauges by optimizingall the four technologies above.

Property Yield strength Hardness Young’s modulus Melting point Gauge(GPa) (Kg/mm2) (GPa) (°C) factor

Si 7 850 190 1410 100-200

6H-SiC 21 2480 700 2830 (sublimes) 10-30

Diamond 53 7000 1035 4000 (Phase change) —-

Table-3 Some important properties of single crystal Si, SiC and diamond [Peterson, 1982]

Bhat & Nayak

9.2 SiC -Capacitive Pressure Sensors for HighTemperatures > 500°C

Capacitive pressure sensors are attractive forhigh-temperature applications because theirperformance is almost independent of temperature,and high sensitivity can be achieved with a low turn-on temperature drift. These devices are also lesssensitive to side stress and environment variations.Capacitive sensors are best suited for low pressureand high temperature sensing applications rangingfrom low temperatures up to 500 °C and beyondwith SiC diaphragms. A schematic structure of onesuch device [Du et al., 2003] fabricated with a 0.5ìm single crystal 3C-SiC diaphragm on a singlecrystal <100> silicon is shown in Figure 23. A thinlayer of phosphorous silicate glass (PSG) on thesurface of the DRIE-etched silicon substrateprovides electrical isolation between the SiCdiaphragm and the DRIE-etched silicon substrate.When the diaphragm is subjected to pressure, P asshown in Figure 23(a), the capacitance increasesnonlinearly due to its downward deflection till ittouches the substrate at a designed touch pressurepoint (TPP). At this pressure, the diaphragmtouches only at its middle. Beyond the TPP, thetouching area increases and the capacitanceincreases linearly as shown in Figure 23(b). Withthe capacitance structure having a circular SiCdiaphragm of 400μm, thickness 0.5 μm suspendedover a 2 μm cavity in the silicon substrate, highsensing capability up to 400 °C has been reportedin the literature [Du et al., 2003] with excellent linearcharacteristics in the pressure range between 1100Torr and 1760 Torr with a sensitivity of 7.7fF/Torr.

These results have indeed demonstrated thebenefits of using SiC diaphragms for hightemperature operations and a capacitive sensingapproach for low-pressure operations and this hasopened up new avenues for future work for sensorsusing SiC. The development of high temperature


SiC electronics with SiC pressure sensors and SiCtemperature sensors for intelligent engine systemsis being pursued in academic institutions andindustries. Semiconductor integrated circuit chipscapable of giving over 100-fold improvement at 500°C operational durability have been reported usingSilicon Carbide transistors and logic circuits[Mohamed et al., 2010]

bulk MWNT as the piezoresistive sensing elements.The development of the pressure sensor includesthe fabrication of 300 thick Polymethylmethacrylate(PMMA) diaphragms using an SU8 molding/hot-embossing technique and AC electrophoreticmanipulation of MWNT bundles on the diaphragms.The advantage of MWNT bundles is the ease withwhich they can be formed across electrodes ontop of pressure diaphragms to serve as sensingelements. A schematic structure of the pressuresensor reported in [Wong et al., 2000] is shown inFigure 24. It has been shown that the diaphragmdeflects due to an applied pressure and the MWNTpiezoresistor gets strained, as a result, causing achange of resistance of 4 kW at a pressure of 70kPafrom an initial value of 153kW at zero pressure.The CNT, as a sensing element on diaphragms usingvarious materials is of great interest, andconsiderable effort has been expended over thepast decade with varied degrees of success.


Fig. 23 SiC capacitive Pressure Sensor (a) theschematic structure (b) typical Capacitance, C, versusPressure, P, characteristics showing the Touch PointPressure (TPP)

Fig. 24 Schematic diagram of MWCNT based pressuresensor reported in [Mohamed et al., 2010]

10. Carbon Nano Tube (CNT) Based Sensors

In recent years CNTs-based pressure sensorshave drawn considerable attention due to their highsensitivity, small size, low power consumption, andstrong mechanical stability, in addition to thermalstability. The CNT-based pressure sensor can beoperated up to 250 °C as against the 125°C ofconventional PN junction-isolated silicon pressuresensors. The high gauge factors of about 1000 andthe small size(diameter 1 to 100nm) of CNTs makethem ideal for miniaturized pressure sensors forbiomedical applications as well. The piezoresistiveeffect in the aligned multi-walled carbon nanotube(CNT) array arises due to the buckling of individualcarbon nanotubes when compressed, leading to adecrease in the electrical resistance of the CNTarray. Recently, it has been experimentally provedby mechanical loading of the array that the decreasein resistance is almost fully recoverable once theload is removed [Mohamed et al., 2010]. In thesame paper, it has been shown that the change inthe resistance increases linearly when 100 g and500 g load are applied at a temperature of 20°C to180°C, thus demonstrating that the multi-walledcarbon nanotube (MWCNT) array is an excellentpressure and strain sensing element capable ofoperation at elevated temperatures.

A novel polymer-based MEMS pressuresensor has been reported [Fung et al., 2005] using

11. High Pressure Piezoresitve SiliconPressure Sensors

We present the details of achievements on ahigh pressure sensor project taken up recently inthe Centre for Nano Science and Engineering atthe Indian Institute of Science Bangalore, as partof the National Program on Micro And SmartSystems (NPMASS) directed by Dr V.K. Aatre,former Scientific Advisor to the Defense Minister.

11.1 Design Considerations

In this section, we present the designconsiderations and wafer level processing approachrequired for high pressure piezoresistive pressuresensors capable of operation in the pressure ranges100 bar to 400 bar (40MPa). They are presentedalong with packaging challenges and techniques for



Journal of ISSS 61

meeting the specifications after integrating with anASIC chip in the same package for offset andtemperature compensation.

As discussed in section 5, the pressure sensorchip design involves the mask layout based ondetailed simulation/analytical studies to determineboth the diaphragm dimensions and the resistorlocations and dimensions. The lateral dimensionsof the sensor are chosen to provide enough widthfor placing the resistors. Additional requirementssuch as the chip size of the pressure sensor enforcefurther constraints on the maximum size of thediaphragm. In most situations, the diaphragms aresquare with a side length L (=2a) = 500 to 1000μm.Thus, when the diaphragm thickness h is in the range10 to 50μm and is anchored at the bottom as shownin Figure 12(a), the theory of plates holds good.Hence, when the diaphragm is subjected to pressure,the maximum stress occurs in the middle of thediaphragm edges. Threfore, the resistors areembedded at the diaphragm edges as shown inFigure 12(b). However in sculptured diaphragmshaving bossed structures, and in high pressuresensors having very thick diaphragms, thediaphragm width L may turn out to be comparableto the thickness, h. In such situations, the simpletheory of plates cannot be applied to determine thelocation of maximum stress.

Recent simulation studies [Thyagarajan &Bhat, 2013] conducted using the FEM toolCOMSOL on micromachined diaphragms with thediaphragm anchored as shown in Figure 12(a) andFigure 15 have revealed that the location of thepeak longitudinal stress on the top surface of thechip depends upon the L/h ratio of the squarediaphragm side length L (=2a) and thickness h. Theexact location of the peak stress component is alsofound to depend on whether the diaphragm is carvedout by the wet chemical or dry etching process.Wet chemical anisotropic etching such as KOHetching results in the cavity having side wallsslanting at an angle of 54.74° to the diaphragm, asshown in Figure 12(a). On the other hand thesidewalls of a cavity etched by deep reactive ionetching (DRIE) are perpendicular to the diaphragm,as shown in Figure 15, making the transition fromthe diaphragm abrupt. As a result the stressdistributions in the two cases are different. Figure25 shows the simulation results obtained on the

location Xp of the peak of the stress component

along the width of the diaphragm on the top surface,for different L/h ratios, considering a few typicalthicknesses. It is clearly seen from this result thatthe location of the peak stress lies outside thediaphragm portion when the L/h ratio is below acertain minimum value which is determined by thecavity geometry (slanted or vertical) and thediaphragm thickness. The simulation results on thestress distribution [Thyagarajan & Bhat, 2013]across the diaphragm surface are shown in Figure26 for a specific thickness of the diaphragm whenthe L/h ratio is 3.75, for different applied pressuresranging from 10bar to 100bar. Here the location ofthe peak stress lies outside the diaphragm regionand that the location is independent of the magnitudeof the pressure. The peak value of this stress varieslinearly with pressure, suggesting that P-type piezo-resistors can be embedded around this regionorienting them either perpendicular or parallel tothe diaphragm edge which is in the <110> direction.

11.2 Fabrication, Characterization and Results

The thickness, h, and lateral dimensions, L ofthe diaphragm are chosen based on the maximumpressure of operation, while ensuring that the burstpressure is at least five times that of the maximumpressure of operation. The resistors are embeddedinside, outside or partly inside the diaphragm,depending upon its L/h ratio. An image of the 400barpressure piezoresistive pressure sensor chipfabricated in the National Nano Fab of the Centrefor Nano Science and Engineering (CeNSE) at theIndian Institute of Science, using the above designcriteria is shown in Figure 27(a). Here, each resistoris split into two portions to accommodate themwithin the high stress region, and are connected inseries using a P+ diffused region. The resistorswere fabricated by the boron diffusion process,which was standardized to achieve a sheetresistivity of 150-200 Ohms/square in the initial trialruns. Ion implantation was carried out at BharatElectronics (BEL), Bangalore to achieve precisesheet resistance in the final process runs. The chipsize is 2mm x 2mm, and the diaphragm is realizedusing DRIE. Aluminum metal was sputtered andpatterned to realize the Wheatstone bridge of thetransducer. Initially, the devices were tested at thewafer level by probing them to ascertain thefunctionality of the sensor and to determine the

Bhat & Nayak


offset voltage. The devices diced from the waferwere die mounted on headers, wire bonded andpackaged with oil filling along with a stainless steelsecond diaphragm and tested up to 400 bar. Aphotograph of the chip mounted on the header andwire bonded on to the pads is shown in Figure 27(b).


Fig. 27 Photographs of a 400bar piezoresistive pressure sensor fabricated at CeNSE, showing (a) the chip (b) chipmounted on header and wire bonded (c) wire connected to the electrical port of the package showing the chip inposition (d) completed package with pressure port and electrical connection.

Fig. 25 Location Xp of peak stress with respect to thediaphragm edge for different diaphragm for KOH andDRIE cases as a function of the aspect ratio L/h fortypical values of diaphragm thicknesses.

Fig. 26 Stress distribution on the top surface of thediaphragm along the width of the diaphragm fordifferent magnitudes of pressure P (in bars) = 10, 25,50, 75, and 100 applied on a thick DRIE diaphragmhaving L/h <7

External wire connections were made to the deviceand brought out of the header as shown in thephotograph 27(c). The packaging was completedby appropriately welding the pressure port. Aphotograph of the completed device is shown inFigure 27(d).



Journal of ISSS 63

Another set of devices was packaged at BELwith oil filling and sealing as required by acommercial application. They were tested atCeNSE. The typical output characteristicsmeasured on the packaged 400bar sensor arepresented in Figure 28. A full scale output of 116mVat 400bar is achieved on the packaged sensor witha sensitivity of 29mV/100bar. Excellent sensitivityand linearity within 0.2% with respect to the fullscale output (FSO) of 400bar pressure wasachieved. As can be seen, the hysteresis isinsignificant.

Packaging these high-pressure sensors isquite different and more complicated than the TO-39 packages presented in section 7. The details ofthe technology and the challenges involved inpackaging these high pressure sensors for operationin harsh environments, and for absolute pressure ,gauge pressure and differential pressure sensing,are presented in the following section.

insensitive to the effects of signals which are outsidethe domain of the system. Thus for instance, if amicrosystem is realized for sensing and monitoringpressure, it should be able to track the effects oftemperature and correct the output to compensatefor the effect of temperature. The micromachinedcomponents of these microsystems are fragilestructures and they are invariably exposed to harshaggressive environments such as the corrosivewater of the ocean which comes into contact withthe pressure sensing microsystem used inoceanography. As a result, the packaging technologyfor microsystems plays a very important role inprotecting these micro devices from the shock, thechemical environment and temperature extremities.In some biomedical applications such as intracranialpressure (ICP) and blood pressure monitoringsensors, the package should invariably bebiocompatible. In aerospace applications which mapthe pressure in the aero foil, the package needs tobe flat and should not exceed 1mm in height. Thepackaging technology for microsystems is thus moreinvolved and varied than the packaging technologyfor integrated circuit chips. Here, we discuss thetechniques used for packaging microsystemsconsidering specific case of micromachined siliconpiezoresistive pressure sensors.

12.2 The Importance and Challenges

MEMS pressure sensor packaging includesthree major tasks, namely, (i) Assembly, (ii)Packaging and (iii) Testing, abbreviated as AP&T.The AP&T of microsystems contributes to asignificant portion of the overall cost of productionand could vary from 20% of the overall productioncost with plastic passivation designed for a friendlyenvironment in mass production, to as high as 95%of total cost for special pressure sensors for hightemperature application with toxic-pressurizingmedia. Thus the AP&T cost for microsystems andespecially for MEMS pressure sensors/ transducersvaries from one product to another. Currently, thiscost represents, on an average, 80% of the totalproduction cost. The source of the failure ofmicrosystems can be traced, in most cases, toinadequate packaging. As a result, microsystempackaging technology has become a key factor inMEMS product design and development, and it hasattracted great interest and the attention of themicrosystem community all over the world.

Bhat & Nayak

12. Packaging Technology

The requirements and constraints imposed onMEMS pressure sensor/transducer packages arebriefly presented here. The associated challengesand techniques used for realizing microsystempackages are presented with examples drawn fromvarious types (Absolute, Gauge or Relative andDifferential) of pressure sensors.

12.1 Introduction

Micromachined (MEMS) pressure sensorsplay an important role in present day microsystems.These microsystems are also required toaccommodate electronics to harness raw signalsinto acceptable levels and also to make the signal

Fig. 28 output characteristics of the 400bar pressuresensor


Packaging the micromachined component ofthe MEMS pressure transducer should be designedto meet the demands and requirements of the enduser, whether those requirements be generic orspecific. It consists of a complex matrix of solutionsand is a multi disciplinary field, irrespective ofwhether it is an application driven or performancedriven requirement. The reliable packaging of thesedevices and systems is a major challenge to theindustry because microsystem of packaging has notreached the maturity level of its counterpart, themicroelectronics packaging, has reached.

The packaging of the MEMS pressuretransducer is required to protect the sensing oractuating elements when the sensing device is incontact with the medium, which, in itself, is thesource of action, as in a gas sensor. Many mediaare hostile to these elements. The pressure sensor/transducer packaging is highly challenging becausethe core devices of microsystems, such as micro-sensors usually involve delicate, complex three-dimensional geometry made of layers of dissimilarmaterials [Gardener, 1994; Malshe et al., 1999].

12.3 Functional Requirements

The mechanical package and its associatedmanufacturing technologies must fulfill functions likemechanical protection for day-to-day use in theservice intended, and media protection whenoperated in harsh environments such as humidity,salt water, body fluids, fuels and gases etc. It shouldincorporate a convenient low cost means tointerface and to isolate a very sensitive silicon diefrom undesirable mechanical stresses. The sensorpackage must be carefully designed to isolate thestress and reject extraneous mechanical straininginputs from any other variables such as acceleration,vibration etc. Simultaneously the package shouldallow the efficient transfer of the mechanicalvariable of desired inputs i.e., pressure in this case.Silicon, which is used for fabricating micro sensorsand actuators, is a relatively low-expansion materialwhile most packaging materials such as metals andceramic exhibit considerably higher expansion.Hence, the package must allow for an appropriatemeans of reducing the undesirable effects ofthermally-created stresses due to expansion orcontraction.

An electrical interface is to be achieved bydepositing and patterning aluminum or a noble metalthin film layer on the die. Modern silicon sensorsoften incorporate calibration and compensationeither by using them on the silicon die itself or byemploying hybrid components / signal conditioningICs / resistors. The package must provide anappropriate means of incorporating and protectingthese additional components. Hence the packageprovided must be suitable for housing and protectingthe associated signal conditioning electronics.

In addition to the functional requirements thatthe package must accomplish, a suitable packagingscheme must have low cost, reliability, compatibilitywith subsequent assembly techniques and highvolume production capability.

12.4 Microsystem Packaging Techniques

The packaging techniques used in siliconsensors originate from two distinct roots. One rootis its adaptation from the well-established ICindustry. The second major root of sensorpackaging technology is derived from theconventional mechanical technology of aerospaceand other process control industries. This excellentpackaging technology involves the more traditional,reliable and well-established mechanical arts likemachining (Precision CNC turning, milling, grinding,drilling, honing, polishing, Lapping, electro dischargemachining etc.,), welding (Electron Beam Welding,Laser welding, precision TIG/MIG welding, Electricwelding), Brazing (Vacuum Brazing and microbrazing), Engraving, metal etching, electro plating,anodizing, Glass to metal sealing, casting of alloysetc. The modern silicon sensor package is a blendor amalgamation of these two arts. Hence, toproduce a package which is rugged and reliable,the silicon sensor-packaging engineer seeks toemploy the best from each art.

Microsystems packaging is categorized intothree levels, namely (i) die level, (ii) device leveland (iii) system level as shown schematically [Hsu,2002] in Figure 29. The primary objectives of dielevel or wafer level packaging are to protectthe die or other core elements from plasticdeformation or cracking; to protect the activecircuitry for signal transduction of the system; toprovide the necessary electrical and mechanicalisolation of these elements and to ensure that the




Journal of ISSS 65

system functions at both normal operating andoverload conditions.

The major challenges in device levelpackaging are associated with interfacerequirements like the interface of delicate dice andcore elements with other parts of the packagedproduct at radically different sizes. The interfacesof these delicate elements with the environment,particularly with regard to such factors astemperature, pressure and toxicity of the workingand the contacting media are some of the criticalissues to be addressed.

System level packaging involves thepackaging of primary signal circuitry with the dieor core element unit. System packaging requiresproper mechanical and thermal isolation as well aselectromagnetic shielding of the circuitry. Metal

housing usually gives excellent protection frommechanical and electromagnetic influences.Assembly tolerance is a more serious problem atthis level of packaging than at the device level [Hsu,2002; Judy, 2001; Li & Tseng, 2001].

Figures 30(a) and (b) illustrate various headersrequired for absolute and gauge/relative MEMSpressure sensor packaging respectively. Figures31(a) and 31(b) illustrate respectively the absolutepressure sensor header assembly and the gauge/relative pressure sensor header assembly withprovision for oil filling to isolate the silicon die fromcorrosive pressure media through a stainless steeldiaphragm. The header assembly for a differentialpressure sensor is illustrated in Figure 32. The dieand wire bonded package suitable for dry gas media(No oil filling and no isolated stainless steeldiaphragm) is shown in Figure 33. In this case, thepressure media is directly in contact with the silicondie. The packaging scheme required for variouscorrosive media, isolated by a stainless steeldiaphragm with silicone oil in between, is shown inFigure 34

12.5. Microsystem Development at CeNSE inIISc Bangalore

Recently, we [Bhat & Nayak, 2012] havetaken up activities on MEMS systems developmentat CeNSE in IISc, with the goal of achievingtechnological readiness for future Indian missions.Focused attention and efforts are needed to deliver

Fig. 30 Header assembly for (a) absolute pressure sensor (b) Gauge/relative pressure sensor

Bhat & Nayak

Fig. 29 Schematic representation of system packagingpressure transducer


Fig. 31 Header Assembly with provision for oil filling for (a) Absolute Pressure sensor (b) Gauge/RelativePressure Sensor

Fig. 32 Header Assembly for Differential Pressure Sensor (a) with provision for oil filling and extension forEvacuation /pressurization (b) with extension for easy evacuation/ pressurization.

Fig. 33 The package suitable for dry gas media showing the die bonded and wire bonded pressure sensor




Journal of ISSS 67

MEMS-based systems. One such device viz., anaerospace quality MEMS pressure transducer,which is in an advanced stage of development, isillustrated in Figure 35. The set-up used forcalibrating such sensors is shown in figure 36.

13 The Future and the Vision

The future is tending towards MEMS to Nanotechnology. Nano technology provides the ability towork at the molecular level, atom by atom, to createlarge structures with a fundamentally new molecular

Fig. 35 Aerospace quality MEMS pressure transducer

Fig. 34 Oil filled packaging scheme with stainless steel diaphragm. The wire bonded die mounted on the header isprotected against corrosive media by the silicone oil between silicon die and the steel diaphragm .

organization. The emerging approach for bothchemical and biological MEMS sensors is basedon carbon nanotube (CNT) and graphene-basedtechnology. It is essentially concerned withmaterials, devices and systems to achieve structuresand components which exhibit novel andsignificantly improved physical, chemical andbiological properties, phenomena and processes dueto their nano-scale size. These sensors have bettersensitivity, selectivity and stability than commerciallyavailable sensors.

Bhat & Nayak


Fig. 36 Pressure transducer calibration setup

Focused R & D, with more attention and effortin design and packaging is needed in India to deliverMEMS-based devices to commercial as well asstrategic programs. Effective collaborationsbetween successful organizations may dramaticallyshorten the time frame between the concept to thesystems. The following areas require carefulattention and efforts:

� Polymer-clay, Nano-micro composite materialfor load bearing structures.

� Composite materials with embedded MEMSsensors.

� Nano and micro sensors and devices for spaceand terrestrial applications.

� MEMS based vibration sensing and analysissystem for structural health monitoring

� Sensors for spacecraft structural healthmonitoring.

� Silicon on sapphire (SOS) based miniaturizedsensors for high temperature operation in upto30 °C for operation in harsh environment

� Silicon carbide (SiC) sensors for operation attemperatures in excess of 500°C and foroperation in automotive applications as wellas for corrosive fluids.

Acknowledgements

The authors are grateful to Dr V.K. Aatre,former Director General, Defence R&D, formerSA to Raksha Mantri and the present chairman,BSMART of the National Program on Micro AndSmart Systems (NPMASS) for his continuedsupport and encouragement for many of theactivities presented in this article which we carriedout in the first phase (NPSM) and the second phase(NPMASS) of this National Program. The authorsalso gratefully acknowledge the support receivedfrom the Centre for Nano science and Engineering(CeNSE) at the Indian Institute of Science forcarrying out the design, fabrication, packaging andcharacterization of these sensors. The authors alsoplace on record the unconditional support receivedfrom Bharat Electronics Ltd Bangalore for severalprocess steps such as the ion implantation of boron,dicing the devices from wafers and packaging thedevices.

References

Bao M.-H., 2000, Micro Mechanical Transducers:Pressure Sensors Accelerometers and Gyroscopes,New York:Elsevier.

Bhat, K.N, P.R.S.Rao, Bhattachara,E., Dasgupta, A.,




Journal of ISSS 69

Dasgupta, N., Vinoth Kumar, V., Sivakumar, K.,Manjula, S.R., Madhavi, SP., Sushma,Y., & Daniel,R.J., 2005, “Micromachined polysiliconpiezoresistive pressure sensor & accelerometer withSOI approach,” Electronics Today (India), pp.65-67.

Bhat, K.N.,Vinoth Kumar, V.,Sivakumar, K., Madhavi, S.P.,DasGupta, A., Rao, P.R.S., ,Bhattacharya, E.,DasGupta, N. Manjula, S.R., Daniel, R.J., &Natarajan, K., 2006 “Silicon Micromachining forMOS Integrated Piezoresistive pressure Sensorswith SOI approach,” Indo-Japan Joint Seminar on‘Micro-Nano Manufacturing Science. Tokyo, pp.19 - 26.

Bhat, K.N. and Nayak, M.M. (Project coordinators) 2012,“Design, Fabrication, Packaging, QualificationTesting of Pressure sensors for Aerospaceapplications” This is a NPMASS sponsored projectjointly taken up by CeNSE at IISc and BEL Bangalorewith Rudra Pratap as PI, Navakanta Bhat, K.N.Bhatand M.M Nayak as Co-PIs from CeNSE; Y.PPrabhakara Rao, Kulkarni, Bharatidasan and Pramod,as Co-PIs from BEL. (June 2012-May 2014)

Du, J., Darrin J. Young, Christian A. Zorman, and WenH. Ko, 2003, “Single Crystal SiC Capacitive PressureSensor at 400C “, IEEE Electron Device meeting(IEDM).

Fraga, M.A., Rodrigo Sávio Pessoa, Homero SantiagoMaciel, Marcos Massi and Ivo de Castro Oliveira,2010, “Technology roadmap for development of SiCsensors at plasma processes laboratory”, J.Aerosp.Technol. Manag., São José dos Campos,Vol.2, No.2, pp. 219-224.

Fung, C. K. M. ,Zhang M Q. H. Zhang , Dong Z., and Li,W. J., 2005, “Fabrication of CNT-Based MEMSPiezoresistive Pressure Sensors Using DEP Nanoassembly” IEEE MEMS, pp.251-254.

Gardener, J.W., 1994, Microsensors –Principles andApplications, London: John Wiley & Sons .

Hsu T.-R., 2002 MEMS and Microsystems design andManufacture. Tata McGraw- Hill publishing Co. Ltd.

Judy, J. M. (2001) “Microelectromechanical systems(MEMS): Fabrication,Design and applications “,Institute of Physics Publishing. Vol. 10, pp 1115 –1134.

Kovacs, G.T.A., 1998, Micromachined TransducersSource Book,New York, WCB McGraw-Hill

Li, G., and Tseng, A. A., 2001, “Low stress packaging of

a Micromachined accelerometer,” IEEE Transactionson Electronics Packaging Manufacturing, Vol.24,pp. 18-25, January.

Mallon J.R., Pourahmadi F.,Petersen K., BarthT.Vermeulen, and Brezek J., 1990, “Low PressureSensors employing bossed diaphragms andprecision etch stopping,” Sensors and Actuators,Vol. A 21-A23, pp. 89-95.

Malshe, A, etal., 1999, “Challenges in the packaging ofMEMS “ Int. Journal of Microcircuits andElectronic packaging. Vol 22, no:3,.

Manjula, S.R., 2005, “Optimization of PolysiliconPiezoresistors for MEMS Pressure Sensor,” M.S.Thesis. Electrical Engineering Department, IndianInstitute of Technology, Madras, India.

Meleshenko, V.V., 1997, “Bending of an elasticrectangular clamped plate: exact versus Engineeringsolutions,” J. Elasticity, Vol. 48, pp.1-50.

Middlelhoek, S. & Audet, S.A., 1989, Silicon Sensors ,San Diego, CA, Academic Press.

Mohamed, N. M., Lai M. K. and Begam K. M., 2010,“Development of Aligned Carbon Nanotubes(CNTs) for Pressure Sensing Application”, NSTI-Nanotech, Vol.2, pp.130-133

Mosser,V., Susuki,J.,Goss,J, & Obermeier, E., 1991,“Piezoresistive Pressure Sensors Based onPolycrystalline Silicon,” Sensors and ActuatorsA,Vol.28, pp 113-132.

Ned, A. A., Kurtz, A.D., Beheim, G., Masheeb, F., andStefanescu, S., 2004) “Improved SiC LeadlessPressure Sensors For High Temperature, Low andHigh Pressure Applications”, Kulite SemiconductorProducts, Inc. Twenty-First Transducer WorkshopLexington, Maryland, June 22-23, 2004

Peterson, K.E., 1982, “Silicon as a mechanical material,”Proceedings of IEEE, Vol. 70, 420-457

Raman, M. S., Kifle, T., Bhattacharya, E., & Bhat, K. N.,2006, “Physical Model for the Resistivity andTemperature Coefficient of Resistivity in HeavilyDoped Polysilicon,” IEEE Transactions on ElectronDevices , Vol53, No.8, pp.1885-1892.

Smith, C., 1954, “Piezoresistance Effect in Germaniumand Silicon,” Physical Review, Vol.94, pp 42–49

Thyagarajan, V. R. and Bhat K.N., 2013, “Optimumlocation for Piezoresistors with Square Diaphragm

Bhat & Nayak


MEMS Pressure Sensors” (to be published).

Timoshenko, S. P., and S. Woinowski-Krieger, 1983,Theory of Plates and Shells, New York:McGraw-Hill.

Vandelli, N., 2008 “SiC MEMS Pressure Sensors ForHarsh Environment Applications”, MicroNano News, pp10-12 (April).

Vinoth Kumar, V., 2006 Design and Process Optimizationfor Monolithic Integration of Piezoresistive PressureSensor and MOSFET Amplifier with SOI Approach,MS Thesis, Electrical Engineering Department, IndianInstitute of Technology, Madras, India.

Vinoth Kumar, V., A. DasGupta , K.N.Bhat and K.Natarajan, 2006, “Process Optimization formonolithic integration of piezoresistive pressuresensor and MOSFET Amplifier with SOI Approach,”Journal of Physics Conference Series, Vol.34., pp.210-215

Wong V.T.S. et al., 2000, “Bulk Carbon Nanotubes asSensing Element for Temperature and AnemometryMicro Sensing”, Proc. IEEE MEMS 2003, pp. 41-44,

Xiaowei,L., Xin, L., Wei, W., Xilian, W., Wei C., &Zhenmao, L., 1998,”Computer simulation ofpolysilicon piezoresistive pressure sensors,” IEEconference proceedings, pp 891- 894.

Prof K.N.Bhat obtained his B.E degree inElectrical Engineering from theIndian Institute of ScienceBangalore, Masters degreefrom the RensselaerPolytechnic Institute (RPI)Troy NY and the PhD degreefrom the Indian Institute ofTechnology Madras.Subsequently, he did his post-

doctoral research work at the RPI where he carriedout extensive research work on the GaAspolycrystalline thin film solar cells and investigatedthe effects of grain boundary passivation on theGaAs polycrystalline solar cells. He has served asfaculty member of the EE Department at IITMadras at Chennai, India for nearly four decadesand taught several courses both at the UG and PGlevels on VLSI Technology, Semiconductor PowerDevices, Compound Semiconductor Devices,

Fundamentals of Semiconductor devices, and BasicElectronics circuits. He was evaluated by thestudents of IIT Madras as one of the top four bestteachers across all the departments of the Institute.. His lecture series on High Speed Devices are nowavailable in the youtube. He has also taught in theEE Department of the University of Washington atSeattle (USA) where he had spent a year onSabbatical Leave and carried out exploratoryresearch work along with Prof Robert BruceDarling and realized Micro machined Faraday CupArray Using Deep Reactive Ion Etching. He hasguided several Ph.D and MS research scholars andpublished extensively in reputed InternationalJournals and International Conference Proceedings.He has carried out pioneering work in the MEMSarea in India and has coauthored the book entitled“Micro and Smart Systems–Technology andModeling” published (2012) by John Wiley and Sons(USA). His book “Fundamentals of SemiconductorDevices” co-authored with Prof M.K. Achuthan,published (2007) by Tata McGgraw-Hill has beenvery popular all over the country.

Since 2006, he is a Visiting Professor at theCentre for Nano Science and Engineering, IndianInstitute of Science Bangalore, after hissuperannuation and retirement from IITM. ProfK.N. Bhat is a Fellow of the Indian NationalAcademy of Engineering

Dr. Manjunatha Nayak received D.IISc andPh.D. (Engg) degrees inElectronic Design Technologyand Instrumentation respectivelyfrom IISc Bangalore.Subsequently he carried out hisPost doctoral research in Micro

Electro Mechanical Systems (MEMS) at Universityof Delft and Twente – Netherlands under INSAInvitation fellowship programme, and at ToyohashiUniversity, Japan under JSPS fellowship.

He served in ISRO, Dept. of Space for nearly40 years since 1971 in various capacities. He hasgrown professionally to Scientist/Engineer ‘H’ andheld positions as Deputy Director, Semi-ConductorLaboratory, Chandigarh and Director, LaunchVehicle Programme Office, ISRO HQ Bangalore.During his 39 years of service at Liquid Propulsion




Journal of ISSS 71

Systems Centre (LPSC) he has developed andproductionised space qualified PressureTransducers for PSLV, GSLV, IRS, INSAT, andChandrayaan missions. The Ultrasonic liquid levelsensors, depletion sensors and Temperature Sensorsfor the Launch vehicle programmes of ISRO,Digital Pressure Sensors for Automatic weatherstations & Tsunami warning System were developedunder his leadership and guidance. Various MEMSSensors were designed, developed and fabricatedwith his guidance at SCL for Automotive, RadioSonde, wind tunnel and Shock tube applications

In recognition of his outstanding contributionsto ISRO’s programmatic activities in variouscapacities, he was awarded ISRO MERIT Awardin 2008. He has also received several other awardsfor his academic excellence. The best paper awardin NSI-1989, SEP France Vikas AgreementCompletion award in 1989, NRDC award in 1985,NPE- 1984 special award and Kirloskar Electriccompany best student award are only a fewexamples of such recognitions. He has more than35 publications in refereed international journals, 4patents and guided 7 Research students.

Superannuated from ISRO in June 2011,

Bhat & Nayak

currently he is a Visiting Professor at Centre forNano Science and Engineering, IISc, Bangalore,further pursuing his research activities in MEMSand Packaging. His research interests includeMEMS and NEMS based sensors and actuators,and advanced MEMS packaging.

In recognition of his outstanding contributionsto ISRO’s programmatic activities in variouscapacities, he was awarded ISRO MERIT Awardin 2008. He has also received several other awardsfor his academic excellence. The best paper awardin NSI-1989, SEP France Vikas AgreementCompletion award in 1989, NRDC award in 1985,NPE- 1984 special award and Kirloskar Electriccompany best student award are only a fewexamples of such recognitions. He has more than35 publications in refereed international journals, 4patents and guided 7 Research students.

Superannuated from ISRO in June 2011,currently he is a Visiting Professor at Centre forNano Science and Engineering, IISc, Bangalore,further pursuing his research activities in MEMSand Packaging. His research interests includeMEMS and NEMS based sensors and actuators,and advaned MEMS packaging.

MEMS Pressure Sensors- An Overview of Challenges in ... · PDF fileMicrosystem packaging...

Documents

Transcript of MEMS Pressure Sensors- An Overview of Challenges in ... · PDF fileMicrosystem packaging...