64 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 12, NO. 1, FEBRUARY 2003

Development of an End-Point Detector for ParyleneDeposition Process

Wongso Sutomo, Xuefeng Wang, David Bullen, Sarah K. Braden, and Chang Liu, Senior Member, IEEE

Abstract—Parylene is an emerging material for MEMS. It is anorganic material that is grown by using the chemical vapor deposi-tion method at room temperature. The deposition thickness is com-monly controlled by the amount of solid-phase dimer loaded in asublimation chamber. In a conventional deposition machine, theend point of the process is designated by the moment the dimeris exhausted. However, this end-of-process criterion does not offerprecise, repeatable control of film thickness. We present the resultsof the development of anin situ end-point detector for a Parylenechemical vapor deposition process. The detector is based on thethermal transfer principle and can be implemented on commer-cial Parylene deposition systems with minimal system modifica-tion. Such a sensor enables a user to stop the deposition when atargeted thickness is reached. The end point detector is very simpleto implement on existing Parylene deposition systems. A series ofsuch sensors with different target deposition thickness would allowextraction of the actual deposition rate within a deposition run.

[813]

I. BACKGROUND

T HE chemical vapor deposition process of Parylene (poly-para-xylylene) is well established [1]. Parylene is a di-

electric polymer material that may be grown by chemical vapordeposition methods at room temperature. It is an exceptionallyconformal coating material with very low pinhole density. Theintrinsic stress of Parylene is relatively low compared with thatof silicon oxide and silicon nitride. Due to its unique abilityto seal and encapsulate [2], the material is useful for telecom-munications, semiconductors, micromachining, and documentpreservation, etc. In recent years, Parylene has been used to fab-ricate micro electromechanical devices such as microfluidic cir-cuits [3], [4], micro-injectors [5], and valves/pumps [6], to namea few. Since the film is used to function as a mechanical struc-ture, the thickness becomes an important parameter that deter-mines the performance specifications of sensors and actuators.

At present, it is difficult to predict and control the depositionthickness. The method for controlling the thickness is to preloadthe deposition system with a controlled amount of dimer mate-rial. The end-of-process criterion is the complete sublimation

Manuscript received February 12, 2002; revised August 28, 2002. This workis supported by the National Science Foundation (NSF) CAREER award toC. Liu. Subject Editor H. Fujita.

W. Sutomo, X. Wang, and D. Bullen are with the Micro Actuators, Sensorsand Systems Group Micro and Nanotechnology Laboratory University of Illi-nois at Urbana-Champaign Champaign, IL 61801 USA.

S. K. Braden is with the Department of Electrical Engineering, Stanford Uni-versity, Palo Alto, CA 94305 USA.

C. Liu is with the Micro and Nanotechnology Laboratory, Urbana, IL 61801USA (e-mail: changliu@uiuc.edu).

Digital Object Identifier 10.1109/JMEMS.2002.807473

Fig. 1. Actual measured Parylene thickness as a function of amount of dimerused. Each data point represents one actual deposition run conducted at ourlaboratory.

of all dimer material. One major manufacturer of Parylene de-position systems predicts that final Parylene thickness can becontrolled to within 10% of desired thickness using thismethod. This degree of accuracy may not be satisfactory formany MEMS applications, however. Actual experimental datagathered at our laboratory demonstrates that the Parylene thick-ness exhibits even greater run-to-run variance than the manu-facturer suggests (see Fig. 1). Many factors, including changingbaffle geometry due to prior deposition and inexperience ofuser, inevitably contributes to the unpredictability of the thick-ness. The inaccuracy of predicting and controlling the depo-sition thickness is especially problematic when the depositionthickness is small (e.g., less than 5m). In the same token, it iseven more difficult to predict the duration of a process. In ourlaboratory, 50% run-to-run time variation for the same targetthickness has been found.

We have developed anin situ end-point detector for accu-rately monitoring the deposition thickness of Parylene. The de-position process is to be stopped when the desired thickness isreached; preferably before the preloaded dimer material is ex-hausted. The sensor should satisfy a number of criteria. For ex-ample, we believe it is important to be able to implement thesensor in commercial Parylene deposition systems with minimalmodification to existing machines. The sensor should be inex-pensive and use simple electronics for signal processing. Thesensor should be small to avoid interfering with the depositionprocess.

1057-7157/03$17.00 © 2003 IEEE

SUTOMOet al.: DEVELOPMENT OF AN END-POINT DETECTOR FOR PARYLENE DEPOSITION PROCESS 65

Fig. 2. Schematic diagram of the end-point sensor.

II. PRINCIPLE OFPARYLENE END-POINT DETECTOR

The schematic diagram of the Parylene sensor is shown inFig. 2. The sensor consists of a heating element and a tempera-ture sensor. The heater and the temperature sensor are located atdistal ends of two diving-board-type cantilever beams. The dis-tance between the distal ends of the two cantilever beams, de-noted , is well defined in the mask layout. Using microlithog-raphy, the size of the gap can be accurately defined.

The heater, made of thin-film metal coil, generates ohmicheating when an electrical current passes through. This heat maybe transmitted to the sensor by two possible heat-transfer modes.Under the first transfer mode, the heat can be conducted throughthe gap between the distal ends of the two beams if a thermallyconducting medium (such as air or Parylene) is present. Underthe second transfer mode, the heat can travel the lengths of thetwo cantilever beams and the supporting silicon substrate. Ob-viously, the second heat transfer mode involves a much longerheat conduction path and greater thermal mass.

Parylene is deposited in a low-pressure environment, with thetypical deposition pressure ranging from 20 to 40 mtorr. When asensor with an open gap is placed in a vacuum, the thermal con-duction through the gap is negligible. The second heat transfermode dominates at this point.

As Parylene is deposited in a conformal fashion, the distancebetween the two distal ends of cantilevers is gradually reduced[see Fig. 3(a)]. When the Parylene thickness reaches, thetwo Parylene fronts will meet, thereby filling the gap and com-pleting a thermal conduction path [see Fig. 3(b)]. As the gap isfilled with Parylene, a thermally conducting medium, heat canbe transferred by both the first and the second transfer modes.Heat generated by the heater now has a “thermal short-cut” toreach the temperature sensor. This change of thermal transfercharacteristic is used to infer the process end point. A singlesensor with a gap can indicate when the Parylene thicknessreaches d .

Experimentally, there are two ways to identify the end point.The first method involves evaluating the time response of thetemperature sensor when heat pulses are applied. The thermaltime constants associated with the two transfer modes can be

Fig. 3. Side profile of sensor during the Parylene deposition process.

Fig. 4. Dimensions of a typical sensor that consists of a heater and a thermalresistor.

made vastly different by design. Specifically, the time constantassociated with the first mode could be much shorter than that ofthe second mode. The second method is to measure the steady-state temperature of the sensor under constant power input tothe heater. We conjecture that the spatial distributions of tem-perature before and after the gap closing will be significantlydifferent. The temperature of the sensor is expected to increaseafter the gap is closed.

It is noteworthy that microfabrication technology is essentialfor the successful implementation of such an end-point detector.Optical lithography and micromachining allow the cantileverbeams to be narrow and thin, thereby reducing the heat transferfrom the heater to the substrate. This increases the time con-stant associated with the second heat-transfer mode, allowingthe presence of the first transfer modes to be detected easily.It also reduces unnecessary heat loss and power consumption.Furthermore, optical lithography is critically important to pre-cisely define the gap distance.

III. SENSORDESIGN AND FABRICATION

The configuration of a typical sensor is shown in Fig. 4. Thethickness of the cantilever beams is 40m. The value of theeffective thermal resistances associated with the temperaturesensor and the heater are , assuming that thethermal conductivity of silicon is 149 .

Generally speaking, it is advantageous to use thin and narrowbeams to increase the thermal resistances. However, metal coils

66 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 12, NO. 1, FEBRUARY 2003

Fig. 5. An optical micrograph of an end-point detector with a heater and atemperature sensor.

(heater) and associated wire leads exhibit intrinsic stress, whichmay bend the supporting cantilever beams. The bending couldbecome significant compared to the gap spacingif the sil-icon beam is overly thin. This would alter the effective distancesbetween the heater and sensor. In our studies, space distortionhas been found in samples with the cantilever beam thicknessbeing less than 8m. On the other hand, if the beams are overlythick, the heat transfer associated with the heater and temper-ature sensor will be reduced. The device would require morepower to operate.

To avoid distortion of gap spacing due to intrinsic stress, weuse single crystal silicon for the cantilever beams. Single crystalsilicon material has very little intrinsic stress. The thickness ofthe beams is carefully controlled in process.

The sensor fabrication process is described as follows.Starting with a -oriented silicon wafer, we growthin thermal oxide as a mask for doping. We then performselective doping to form the silicon thermistor. Thin-film metal(200-nm-thick gold) is evaporated and patterned to form heatingresistors. Bulk etch is performed from the backside of the wafer,using either deep reactive ion etching (DRIE) or anisotropicsilicon etching. The wafer is etched to a predetermined depth.The thickness of the resultant single-crystal silicon membraneis controlled by timed etch. This is followed by reactive ionetching from the front side to define the cantilever beam. Theheater and the temperature sensors are located at distal ends oftwo cantilever beams.

An optical micrograph of a sensor before the gap is closedis shown in Fig. 5. The scanning electron micrograph of thegap between the heater and the sensor for a typical device isshown in Fig. 6(a). An SEM micrograph of an array of suchsensors with varying gap sizes is shown in Fig. 6(b). Usingan array like the one shown in Fig. 6(b), we conjecture itwould be possible to extract the continuous etch rate of theParylene process by timing prescribed stops. A comprehensivecharacterization of this operation mode was not carried out inour current studies.

(a)

(b)

Fig. 6. SEM of (a) the gap between the temperature sensor and the heater and(b) four pairs of heater/temperature sensor with varying gap sizes.

IV. SENSORCHARACTERIZATION

A. Temperature Sensitivity

We first calibrate the temperature sensitivity of the doped sil-icon resistor. We measure the change of resistance as a functionof the substrate temperature by placing the chip on top of a tem-perature-controlled stage. The resistance changes with respectto the stage temperature in a linear fashion. The temperaturecoefficient of resistance (TCR) is found to be 0.14%/C.

B. Substrate Heating

We have experimentally proven that, if power is applied to theheater for an extended period of time, the substrate of the sensorwill be heated to a higher temperature. We recorded the changein resistance value of a thermistor as square-wave power pulsesare applied (see Fig. 7). The magnitude of the input voltage is5 V. The input power during the on state is 0.1 W. It is shownthat the time constants associated with the heating and coolingprocesses are 336.8 s and 145 s, respectively. The time constantsfor two other sensors with identical cantilever design are within

SUTOMOet al.: DEVELOPMENT OF AN END-POINT DETECTOR FOR PARYLENE DEPOSITION PROCESS 67

Fig. 7. Variation of resistance of a representative temperature sensor as thestep input power is applied to the heater. The x-axis represents the time. Thelength of the entire plot is 10 min. During the first half of the test period, thepower is turned off. The power is turned on suddenly.

Fig. 8. Schematic diagram of the bending of opposing cantilevers.

1% of the above-mentioned values. The temperature differencebetween the unheated and the heated case is 16.9C, based onthe known TCR value.

Incidentally, we discovered that the thickness of the depositedParylene is much lower if the chip substrate is at a constant el-evated temperature. For example, if power is continuously ap-plied to the chip during a 4-h deposition, the thickness of Pary-lene on the two cantilever beams is less than 0.1m while thethickness elsewhere in the chamber is 12m. Detailed inves-tigation of this phenomenon is currently being conducted. Adetailed study of this effect, though interesting, is outside thescope of this sensor development work. Nonetheless, we con-cluded that the power to the heater must be provided in shortpulses to interrogate the sensor. This would avoid significantlyraising the temperature of the substrate.

When an electric current is temporarily applied to theheater, the temperature rise would cause the cantilever to bendslightly due to thermal bimetallic actuation. Likewise, thecurrent used to interrogate the resistance of the thermal resistor,being on the order of several mA, also may cause deflections.The horizontal displacement components of the ends of twoopposing cantilevers would vary the gap spacing (see Fig. 8).In the following, we provide analysis to shown that this effectis negligible.

For a bimetallic cantilever, the radius of curvature is denotedand the length is denoted. The amount of change of hori-

zontal spacing is. The amount of vertical displacement at theend of cantilevers is . We estimate the value ofby assuminga moderate temperature rise (10C) on both heater and sensorcantilevers (with dimensions specified in Fig. 4). The horizontal

Fig. 9. Schematic diagram illustrating the sensor location in the Parylenedeposition system.

displacement of the end of each cantilever is approximately byusing Taylor series expansion, withbeing much greater

than . The cantilever for the heater consists of two segmentswith different width (Fig. 4). For order-of-magnitude estima-tion, we assume the width of the entire cantilever is 120m.

Based on existing formula [7], the values offor the sensorcantilever and the heater cantilever are roughly 38 m and 39 m,respectively. Here, we assume that the thickness, thermal expan-sion coefficient, and Young’s modulus of the metal layer (Au)is 0.2 m, C, and 57 KPa, respective. For the sil-icon layer, the thickness, thermal expansion coefficient, and theYoung’s modulus are estimated as 40m, C, and190 KPa. The value of contributed by both cantilevers, is ap-proximately . This is much less than the initial gapsize. Therefore, the distortion of the gap due to heating can beignored.

C. Measurement Setup

The schematic diagram of a measurement system is Fig. 9. Weapplied the sensor to a commercially available PDS 2010 Lab-coter 2 deposition system. Because the system is not equippedwith vacuum-sealed electrical access wires, we have modifiedthe observation window to provide electrical access. The ob-servation window, made of glass or plexiglass, can be removedeasily from the vacuum dome. Holes are drilled in a replacementwindow to allow wire leads to pass through. The holes are thensealed using an epoxy material. A sensor chip is located insidethe vacuum dome using a machined rack. The power supply andohmmeter are located outside the vacuum dome. A computerprovides long-term direct data acquisition (at 0.5 s intervals)from the ohmmeter through an RS-232 link. A photograph ofthe setup is shown in Fig. 10.

The accuracy of the end-point detection sensor is dependanton the length of intervals between interrogating heating pulsesas a fraction of the total deposition time. If the gap-closing eventoccurs between two interrogation pulses, the sensor would notknow it until the next pulse arrives. Suppose that the interval is

and the total deposition time is, the percentage margin of

68 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 12, NO. 1, FEBRUARY 2003

Fig. 10. An optical micrograph showing the vacuum dome and the packagedParylene end-point detection sensor.

Fig. 11. Response obtained in real time during a run that lasted 1 hour17 min. The signal strength increases visibly after roughly 57 min. The sensoris interrogated every 5 min after 25 min into the process.

error is given by . In order to increase the ac-curacy, it is advantageous to reduce. However, the value ofcannot be reduced arbitrarily as frequent heating may cause thelocal Parylene deposition thickness to change. We are currentlyconducting a systematic investigation to determine the minimalvalue of without compromising the deposition profile. How-ever, there is theoretically no limit to the accuracy. For example,it may be possible to use redundant sensors in one depositionchamber and interrogate multiple sensors sequentially to reducethe effective time interval,.

D. End-Point Sensing

We monitor the resistance of the temperature sensorthroughout courses of deposition runs by applying square-wavepulses with a constant magnitude (5 V) and pulse width (5 s).The average interval between pulses is 5 min. We believe thatthis brief heating does not cause sufficient substrate heatingto disrupt the thickness of the local Parylene deposition. Theresistance output of the sensor during a typical run is shownin Fig. 11. The thickness of the deposited Parylene material isindependently verified by measuring the Parylene thicknesson companion wafers using surface profilometry. A sectionof Parylene film is manually peeled off. A surface scan isconducted across the edge of the tear.

Fig. 12. Rise and fall of resistance under a step power input to the heater. Thisis obtained before the gap is sealed.

Fig. 13. Rise and fall of resistance under a step power input to the heater. Thisis obtained after the gap is sealed.

During a typical run that lasted 1 h, 15 peaks are registered,corresponding the application of 15 power pulses. The generalappearance of the first ten peaks is different from that of last fivepeaks. We conjectured that the appearance indicates whether thegap has been bridged. A typical waveform selected out of thefirst ten peaks is shown in Fig. 12. As the power is suddenly in-creased, the resistance (and therefore temperature) of the sensorincreases exponentially with a measured time constant on theorder of hundreds of seconds. After the power is cut off, the re-sistance value gradually returns to the original level.

A representative plot of the last five peaks is shown inFig. 13. It is obvious that the resistance of the temperaturesensor changes rapidly upon application of the power. Thisrapid change is caused solely by the fact that the heat pulsetravels directly across the gap, now bridged by Parylene, to thetemperature sensor. The rate of resistance change then slows,indicating that the substrate heating effect has taken over. Afterthe power is turned off, it is again seen that the resistancedecreases rapidly before the substrate heating effect catches up.

We performed 12 Parylene deposition runs using 12 sensorswith different gap spacing ( , 1, 2, 2.5, and 5 m).The actual measured thickness is plotted against the target thick-ness (see Fig. 14). The inaccuracy of data can derive from manysources, including inexperience of using the sensor at this point,as well as random and systematic sensor errors (e.g., gap sizevariation). We believe the accuracy will further improve withgreater number of runs and more expertise in using the sensor.Nonetheless, the effectiveness of the end point sensor is clearlyillustrated.

SUTOMOet al.: DEVELOPMENT OF AN END-POINT DETECTOR FOR PARYLENE DEPOSITION PROCESS 69

Fig. 14. Results of 12 deposition runs using sensors with five target thickness(d=2 = 0:5, 1, 2, 2.5, and 5�m). The vertical axis is the actual measuredthickness, obtained using surface profilometry on companion referencesubstrates. The horizontal axis is the target Parylene thickness of each sensor.Each data point is gathered with a distinct Parylene run.

Fig. 15. Optical micrograph of a heater and a temperature sensor with the gapbridged by Parylene.

The micrograph of a sensor with fully deposited Parylene isshown in Fig. 15.

V. CONCLUSION

An end-of-process monitor for Parylene deposition has beendesigned, fabricated, and tested. The sensor, based on thermaltransfer principles, consists of a heater and a temperature sensorseparated by a well-defined gap,. When the Parylene deposi-tion reaches , the heat can conduct through the gap, in ad-dition to conduction through the substrate. The thermal signa-ture is used to determine the gap-closing event. We have foundthat the end point can be identified by the time constant of theheating and cooling process. Preliminary results of thicknesscontrol have been obtained.

REFERENCES

[1] Properties of Parylene, Parylene Coating Services. [Online]. Available:http://www.paryleneinc.com

[2] T. A. Harder, T. Yao, Q. He, C. Shih, and Y. C. Tai, “Residual stress inthin film Parylene-C,” inProc. 15th Int. Conf. On Micro Electro Me-chanical Systems, MEMS’02, Las Vegas, NE, 2002, pp. 435–438.

[3] P. F. Man, D. K. Jones, and C. H. Mastrangelo, “Microfluidic plasticcapillaries on silicon substrates: a new inexpensive technology for bio-analysis chips,” inProc. 10th Int. Workshop Micor Electro MechanicalSystems, MEMS’97, Nagoya, Japan, 1997, pp. 311–316.

[4] H.-S. Noh, P. J. Hesketh, and G. C. Frye-Mason, “Heating elementembedded Parylene microcolumn for miniature gas chromatograph,” inProc., 15th Int. Conf. Micro Electro Mechanical Systems, MEMS’02,Las Vegas, NE, 2002, pp. 73–76.

[5] X. Q. Wang, A. Desai, Y. C. Tai, L. Licklider, and T. D. Lee, “Polymer-based electrospray chips for mass spectrometry,” inProc. 12th Int. Conf.Micro Electro Mechanical Systems, Orlando, FL, 1999, pp. 523–528.

[6] X. Q. Wang, Q. Lin, and Y. C. Tai, “A Parylene micro check valve,” inProc. 12th Int. Conf. Micro Electro Mechanical Systems, Orlando, FL,1999, pp. 177–182.

[7] W. H. Chu, M. Mehregany, and R. L. Mullen, “Analysis of tip deflec-tion and force of a bimetallic cantilever microactuator,”J. Micromech.Microeng., vol. 3, pp. 4–7, 1993.

Wongso Sutomoreceived the B.S. degree in mechan-ical engineering from University of Texas at Austin inAugust 2000 and the M.S. degree in mechanical en-gineering from the University of Illinois at Urbana-Champaign in August 2002. His research interest isin the area of MEMS ith focus on parylene depositionsensors.

Xuefeng Wang, photograph and biography not available at the time of publica-tion.

David Bullen received the B.S. degree in mechanical engineering (magna cumlaude) from Colorado State University, Fort Collins, in 2000.

Since then, he has worked as a graduate research engineer in the Micro Actu-ators, Sensors, and Systems Group at the University of Illinois, Urbana-Cham-paign. His expertise is in the areas of heat transfer, mechanics, and systems inte-gration and design. His primary research responsibility is in the area of scanningprobe nonlithography.

Sarah K. Braden, photograph and biography not available at the time of pub-lication.

Chang Liu (S’92–A’95–M’00–SM’01) received theM.S. and Ph.D. degree from the California Instituteof Technology (Caltech), Pasadena, in 1991 and1995, respectively. His Ph.D. dissertation was titled“Micromachined sensors and actuators for fluidmechanics applications.”

In January 1996, he joined the MicroelectronicsLaboratory of the University of Illinois, Urbana, as aPostdoctoral Researcher. In January 1997, he becamean Assistant Professor with major appointment in theElectrical and Computer Engineering department and

minor appointment in the Mechanical and Industrial Engineering Department.His research interests cover microsensors, microfluidic lab-on-a-chip systems,and applications of MEMS for nanotechnology. He teaches undergraduate andgraduate courses covering the areas of MEMS, solid state electronics, and heattransfer. The website of his research group is http://mass.micro.uiuc.edu

Prof. Liu received the NSF CAREER award in 1998. He is currently an As-sociate Editor of the IEEE SENSORSJOURNAL. .