MICROSTRUCTURE
SENSORS
H. Guckel, T. R. Christenson, K. J. Skrobis, J. J. Sniegowski. J. W. Kang, B. Choi,
E. G. Lovell
Wisconsin Center For Applied Microelecnonics
Department of Electrical and Computer Engineering
Universify of Wisconsin, Madison, WI 53706
ABSTRACT
Surface micromachined polysilicon ressure sensors offer an
amactive, cost effective, hi h performance technoPo y if material repeatability
issues can be solved. Tkis has been accomplised and fully packaged
y
o
t
o
w devices have been tested extensively and confum the exptahons
or ,this technology. The ressure ranges over which these devices are
desienable has been estabkhed and is set by the maximum allowable
diaphragm stress and the minimum acceptable ouput level.
The low uressure limitation can be eliminated bv imurovinrr the
transducer sensitihy. The substitution of the resonatihg fdrce s$sor,
another polysilicon component, for piezo,resistorscan accomplish this but
complicates processing. This a proach is ustified for hlgh performance
appl!cations.. The altemeve is Found in a differential transducer which I s
fumished with double-sided over- ressure stops One of these stops is
provided by the silicon substrate 8 e other by an electroplated air brid e
which is formed b x ra lithograihy and electroplating in the presence ofa
sacrificial layer.
result is a ngd, thick, accurate1 spaced stop which
results in devices which are rug ed and can measure difYerentials in the 1
water range. The read-out for tfis class of transducers can,be designed in
such a way that simultaneous capacitive and piezoresistive interrogation is
possible.
is sensed electronically andapmduces (he electronic pressure signal. The
discussion indicates a natur subdivision of ill box formation vacuum
sealing and electronic sensing for the device. ?t ,also implies thai pill box
behavior and electronic sensing conmbute to dev!ce performance together.
Thus very small deformahonsof a mechanicall sbff system are acceptable If
the sinsjn scheme is sufficientlySensitive. d e concept of an overpressure
sto$ whicf is either provided by the device or the package becomes apparent
an necessary if one considers that increasing pressures cause increasing
deflections and will eventually lead to ill box fdure.
In the surface miqwnachin3 version of the pressure transducer the
ConshuChOn techruque of Figure 1 applies [l].
INTRODUCTION
Sensors are devices which typically convert physical variables which
are to be measured to electronic signals which become part of a control
s stem They consist of two parts: The Sensor structure which will be
d&cusskd here and the package which protects the device from environments
which art?.often difficult to handle.
Size reductions in the sensor structure are near1 always beneficial.
They allow cost reductions via batch fabrication teclnigues which ,are
borrowed from microelectronics and lead to the term micromechanics.
Micromechanical sensors or microsensors can sometimes be combined with
ormance improvements and
cofabricated microelectronics which produces
can result in structures which are i g i f i e d ,as smart sensors.
Microminiaturization can expand sensor application areas. This is
exem lified b physic$ sensors for biological systems. Blqod pressure and
b l d ana$'.
SIS devices
. must be small m order to be effeChVe
g"The abncapon techniques which a y most d k t l y 'available for
microsensor fabncauon have theu onmn in microelectronics. The central
difficulty which one experiences is'based on the fact that sensors are
fundamentallv three-dimensional structrures and integrated circuit
construction isbased on the concept of planar processing wh>h is of course
two-dimensional. It is therefore not very difficult to understand that presently
nearly all microsensor construction techniques are adaptations of planar
integrated circuit processing with modest three-dimensional extensions.
Thus,,in wafer to wafer bondmg IC,processingis co,mbmed with silicon bulk
machming and wafer to wafer b o d n g to acheve micrOSensor produchon. In
surface micromachining a technolo y which is of interest here planar
processing and lateral etching are com%ined to achieve the neces& threedimensionalit$, This
,
situation,
,
and therefore the, shift towards ,threedimensional f ncahon, is slowly changmg and non-silicon technologies are
becoming more important.
POLYSILICON PRESSURE TRANSDUCERS
Pressure transducers are the most used and therefore the best
understood sensors. They fall into two classes: Relative or differential
devices and absolute transducers. This sensor the absolute device has been
chosen as the test vehicle for microminiaturization via' surface
micromachinin
Absokte pressure tqnsducer;~are fupdamentally vacuum sealed pill
boxes which deform geometrically with applied pressure. Thls deformation
Fig. 1
Basic surface micromachined pill box.
In this s-tufe
a silicon wafer is fumhhed with a rof de th G
and sa square side dunension a. Thls recess IS covered with a polys8con
latk ofihickness h. Typical dimensions for G are 0.8 micromep or so.
Righ pressure T l s a u o n s would cause dm hragm deflechons which could
not e x d G. ence, with proper design 8gure 1 contains a pill box with
overpressurestop.
,
,
Fi ure also idenufies a &posited plysilicon f@n as
of the ill
box, The $ickness of this film, h, is Fmcted to &mensio% w&ch can
achieved in reasonable depositions hmes. Fi e 1 also insists that the
deposited film is locally defined which typicarrequires plasma etching.
Since @e etch time and pattem fidelity must fit into certain 1,imitsa second
restnchon of fiim thickness becomes evident. Practical expenence based on
the above$$ts
leads to the conclusion that typical film thicknesses range
from 1 to m i m e F r .
The dimension a depends on the intended pressure range and the
behavior of the polysilicon film. It becomes a designable quantity only if the
mechanical pro rties of the deposit,are constant with .messin This very
complex issue 6 s to the conclusio? that constant E m morp%o!ogy is,a
necessary condiuon Funhermore since isotropic rather that anisotro IC
behavior is advantagks, morpholdgies with orientated crystalliti?
$so
not ,acceptable. A lycrystallinefilm with randomized, small p n s is most
desuable. The qua& of the deposlt can be evaluated by measunng Young s
modulus for the film It should approach that of smgle crystal m a t e d [2
There is a'second consideration. Figure 1 can be realized.by
forming an oxide post via isoplanar r s s i n g @ the substrate. 'Ihe e n c l w
oxide must be m o v e d by lateral e& mg. TIUS mphes extended hydrofluonc
acid exposure of the polysilicon. Any oxide and rutride inclusions as well as
chemical modifications due to impunties must therefore be avoided because
this would cause modifications in mechanical film characteristics due to the
etching process. They are not acceptable and films that are modifiedin this
manner are not applicable to microsensor duction.
An opumized silicon film of $is t
will have three important
properties. It will have a Young's modulus, r w h i c h in the
1.65 X l0l2 dynes/cm2 and does not deviate by more than 1 0 1 c $ z . i s
&
26.1.1
CH2865-419010000-0613 $1.00 0 1990 IEEE
IEDM 90-613
r
The film will be able to su port a maximum strain of 1.5% before it
fractures. It will also have a buit-in strain field. This strain level must either
be controllableand thereforebecomes a part of the design process or the film
must have the propetty that processing techniques exist whch cause the builtm s t " to & s a y In the p""t case stnun zroin will be assumed even
though strain fie d c&trol is easiblefor Pplysicon
,
In order to understand the unp icauon of e microsensor vacuum
formation in Figure 1 is postulated. It is then assumed that the transducer is
square and that the maximum pressure range is defined b a diaphragm
deflection at the center of the diaphragm which is e q d t o the cavity
dimension G. This condition is given by
$.
(1)
G=
0.0152q a (1 - v 2 )
Eh3
where q is the applied pressure which causes touch-down and v is the
Poisson ratio for th~stype of polysilicon.
The above deflection induces diaphragm strain. The sash field
maximizes at the clamped edge midway between the comers. This stran
value is given by
(2)
Ex=
0.308 q a2 (1 - v')
of polysilicon which can be used to roduce excellent stable and
dielectrically isolated sensin structures. &is is accomplished by coverin
the structure in Figure 1 witf a silicon nitride la er,. a pattemed and dopd
polysilicon iayer and a protective "de
iayer. Me lssue becomes then one
of performance evaluauon for particular resistor doping levels and resistor
placements.
Polysi!icon Fsjstors are. quite different from diffused silicon
resistors. The plezoresistlveeffect m these devlces IS roughly a factor of five
smaller than that of a well desi ed single crystal counterpart. Longitudinal
gage factors m typically slig&y above 20 and transverse ga e factqrs are
near -8 The tem rature coefficient of reslstance TCR can positwe or
ne tive and can,~ c l tqozero.
~ The noise figure;or &se devices mvolves
onf thermal noise which is normally only found to be true for very ood
metal film resistors. Polysilicon resistors are dielecmc isolated which d o w s
for higher temperature applications because junction leakage currents are
absent
The Dlacement issue for these devices is again auite different than
single crystaf placement. Here the rule is simpli to locate them in the
This would imulv lonaitudinal
maximum stress regions on the diauh".
sectiov which en& the d i a p y m - a t t&e support midway kiw~<comers.
There is however a roblem
nphra m sizes accodng to Fi e 2 will
typic@ibe less
m i w e on& side. The resisto~wf&xefm
be quite small with typical linewidths of 4 micrometer. This unplies that
alignment tolerances as well as line width shifts during polysilicon etching
must be considered. These problems whch can become very signficant have
led to an implementation which follows the rule of one resistor per
diaphragm. Figure 3 shows the layout.
&
Eh2
F d cannot exceed,the maximum allowed strain. Figure 2 summarizes and
illustratesthe StIJahon.
Fig. 3
PLATE SIZE ( r m )
Fig. 2
Pill box behavior at +e touch-down pressure in psia,versus uare
plate width a in micrometer with diaphragm thickness3,in
micrometer as a parameter. The upper curves are maxunum st" at
touch-down curves which are measured in percent
The maximum pressure sensing ability of this technology is clearly
illustrated by this graph. .
The structure m Figure 1 can be converted to a pressure transducerif
the sensing mechanism is added. Piezoresistive and capacitive techniques
form the most direct approach. Piezoresistive sensing is the most directly
implemented technique and profits in this technology from the availability
26.1.2
614-IEDM 90
Resistor layout for alignment error tolerant manufactUring.
This layout is reasonably insensitive to alignment erfors but does not avoid
the linevndth tolerance w e
A full "Iucer
&s four dev$a in the bridge confi@on.
"he
procedure involves two resistors which are pressure sensitwe and two
s ~ their pill box oxide has not
resistors which are insensitive t o ~ s because
been removed. W i e +is con igurauon, the hdf-active full brid e, the
maximum output in millivolt per volt of bndge excitatm can be calcu%ted at
the touch-down pressure. The results are shown in Figure 4.
The difficulties, small ou ut oyer span, for,the low ressure ranges are
clearly evident in F i r
This data e d R y e 2 &ntify the expected
design ranges over w ich thls technolo y ISuse ul Withm this design range
experimentalperfomance is very goofand in par~cularalways results in a
very small device. One can take advantage thls and Pr0;)uce a device which
contains multiple sensors. This is indicated in Figure 5 where four pressure
ran es are supplied on a single die which eventually is mounted in a single
Pa&=.
d
20
40
60
60
IO0
120
140
I60
PLATE SIZE
Fig. 4
180
200
220
243
:63
282
(pm)
This figure is related to figure 2. It contains the maximum cxpectcd
out ut voltage per volt of bridge excitation for a half active full
bri8ge.
Fig. 6
Resonating force transducer. This device contains a 200 micrometer
long and 40 micrometer wide polysilicon beam which is 2
micrometer jhick. The beam is completely enclosed by a silicon
shell which is evacuated. It is dnven elecuostatlcally at the beam
center and monitored with iezoresistors at the beam ends.
Resonant frequencies are near
khz with Q-factors near 2 5 . m .
&
DIFFERENTIAL MICROSENSORS
The absolute pressure transducer as discussed here can be converted
to a differential device if pressures are a plied to both sides of the
dia hm m. This requires one or more via hogs through the substrate which
cnrfin #ow channels which connect the backside pressure to the diaphragm.
The idea is illustrated in Figure 7.
i
Fig. 5
Prototype pressure transducer with four pressure ranges. Touchdowns occur at 20 sia 150psu. 300 psia and 650 psia. The dic
size is roughly O.O&*'xb.080 .
RESONATING SENSOR
The difficulties to which Figure 4 alludes are the result of
piezoresistive sensing. They can be removed by changing the sensing
technique or by converting the device from an absolute pressure sensor to a
differential transducer. Both appaches are receiving derailed attention. In
the first case the piemresistor IS laced by a new type of force sensor: The
vacuum sealed resonating beam%s
structure which is shown in Figure 6
evolves from the polysilicon technology which has been &scussed here [41.
It is essentially a pill box which contains a clampedclamped free standing
beam which can be excited elecrostatically. The transduchon mechanism is
s h ly axial ap lied force to frequency which becomes reasonable if one
th& of pitch atjjustments for a violin string. What is not intuitively obvious
is the very high sensitivity which allows for simpler and more precise
measurements. The draw-back is found in the increased complexity of the
necessary construction technique. The device becomes expensive and most
likely is useful only for the high end market where low pressure precision
mmurements are reqmred.
pb
Fig. 7
pb
Differential transducer with single over-pressure stop.
The construction technique preserves the over-pressure stop in the silicon
direction. The second stop which is needed for the condition P p > P f is
missing. It cwld be provided by p c k a 'n This will be somewhat difficult
because the gap dimension G is typic8y
than a micrometer. Wafer to
wafer bonding could be used but presents severe processing complications
because high temperature heat cycles are required. A less complicated
technolo IS requued. A viable and very interesting candidate for this is a
modifizersion of the LIGA process.
f&
26.1.3
IEDM 90-615
L E A processing was erst reported by W. Ehrfeld [5].,His concept
is based on a substrate which is covered with a suitable plaung base for
subsequent electroplatin The plating base is covered with a thick layer of
photoresist, say up to
micrometer. This matenal is exposed through a
suitable mask td high intensity x-ra fluxes with average wavelen ths in the
few An strom range. Diffraction eYfects will therefore be minimapand if thc
photon flux is normal to the substrate pattern run-out in the photoresist layer
due to exposure conditions will be essentially absent Optimized developin
will therefore produce photoresist free regions w5ch'are bounded by verticafi
walls. These regions can be filled with metals vla electroulauna. Figure 8
illustrales resurts which were obtained at Wisconsin'by this type of
processing [61.
3b
'
This has in fact been done. The processing uses the absolute transducer
constructiontechnique and augments it with bulk micromachined through-thewafer pressure tabs. It then uses a spun-on film of specjal polyimide to
produce the sacrificial layer at 0.8 micrometer thickness. Tlus layer is locally
dcfined via an o tical mask and photoresist processing. All excessfolylmidc
is removed w i g the photoresist develo er The remaining po yimide is
hardened by bakjng at 27p"C. Thls hariens the film sufficiently to allow
~
platin base applicauon vla sputtenng of 1 5 0 of~ utanium and 1 5 0 of
nickef However the pol imide can still be removed in a variet o basic
solutions. The blating xase process is followed by thick pxotoresist
application, x-ray exposure with an x-ra mask which must,be aligned to the
wafer, develo ing and electroplating. F!emoval of the laung base exposes
the s,acrificial&yer which can now be removed b latedetchmg. In essence
an a r bndge with ve tight dimensional controz and very significant metal
thickness results and%ms !he F o n d over-pressure sto
The resulting device is uite interesung.
goal, a 1" water
wqsducer, is easil achievable wi?h a 715x715 micrometer diaphragm a t 2
micrometer lysi8con thickness. The device can surave overpressures in
excess of 1fG psi without damage. ~ t piezoresistive
s
out ut is small.
However since the polysilicon diaphragm can be covered w i 2 a thin metal
film withbut problems a pressure sensitive read-out via capacitive sensin
becomes feasible. The capacitor is formed between the metal covere8
diaphragm and the over-pressure stop. The fact that piezoresistive and
capacitive signals are available simultaneously is quite attractive because
electronic data verification becomes possible and is very attractive for high
reliability applicaeons. >e possibi ity of using,the, device as a uressurc
sensiuve switch exists and is currently under investigahon.
de
CONCLUSIONS
Construction techniques for surface micromachined absolute
pressure transducers are sufficiently dvanced toproduce m o t
pressure
uansducer chips which have been tested extensively. d e res% are very
encouragin and lead to the conclusion that this type of sensor is real and
economdfy viable.
The merging of polysilicon technolo ies with electroplated metal
structures is ust now swung. The expectd performance results in the
conclusion d a t man new structures will become feasible and that, in
particular, a reasonah 1 water transducer is in the near future. The
applicability of the double sided o v e y u r e stop to other devices such as
accelerometers is already bein studi and promises to have a major impact
on this, area. This is particu!arly true if these devices also include the
resonatlng beam force transducer.
ACKNOWLEDGEMENTS
Part of this work was supported by the National Science Foundation
under grant EET-8815285.
We are indebted to
T. D. F@m of,Brewer Science Inc.,, Rolla,
MO 65401 for the supply of PIRL matenals which were used to fabncate the
sacrificial layer for theinetal over-pressure stop.
e.
Fig. 8
Nickel gears which were produced by deep x-ray lithography at
Wisconsin.
I f the,baslc LIGA process is modified by the addtion of a sacnlicial
layer which is locally defined the over-pressure stop which has been
dscussed earlier can be fabricated. Figure 9 illustrates the concept
REFERENCES
1. H. Guckel D W Burns C. R. Rutigliano D. K. Showers J. Uglow,
Fine Gr&nd Pol silicdn and its Applicatibn to Pressure Trhsducers,"
Proceedings of t i e 4th Intematlonal Conference on Sensors and
Actuators,
277-282, Tok 0,Japan, June 1987.
2. H. Guckel
W. Bums
A. C. Tilmans D. W. DeRoo C. R.
Rutiglian;, "Mechanical hoperties of Fine G h n @ Polysilicoh: The
Repeatability Issue," IEEE Techn~calDi est for Solid Slate Sensors and
Actuators, p 96 99 Hilton Head SC fun, 1988.
3. H,Guckel
W.
C. C: G. \iissdr, H. A. C. Tilmans, D.,!kRoo,
Fine G h n e d Polys!licon Films with Built-in Tensile S m n , IEEE
Trans. on Elect. Devices. Vol. 35. No. 6, pp. 800-80J June 1988.
4 . H. Guckel, J. J. Sniego,wski, T. R. C,hpstenson,
and
Performance Charactenstlcs of Polysihcon Resonaung Beam Force
Transducers." Proc. of the Thud Toyota Conference, pp. 23-1,23-10,
Nisin Aichi Ja an 1989
5. W.Ehrteld. F: &U;, D. Miinchmeyer,. W. Schelb and D., Schmidt,
"LIGA hocess: Sensor Construction Techniques via X-Ray
Lithomhv," Technical Digest, IEEE Solid-State Sensor and Actuator
WorGhdp,-p
6. H. Guckel T. k%i:L\znK. J. Skrobis D. D.Denton B. Choi E.
G. Loveil J W. Lee S S. B a j h and T. W; Cha man "Deep X-Ra
and UV iithographiks for Micromechanics TecRnicai Di est, IEEZ
Solid-state Sensor and Actuator Workshop, ip. 118-122, I&.
%:
8
d
BGs,
Construction
E.
Fig. 9
Differential pressure transducer with two overpressure stops.
26.1.4
616-IEDM YO