Author's personal copy
Available online at www.sciencedirect.com
Sensors and Actuators B 129 (2008) 241–245
Al2O3-coated microcantilevers for detection of moisture at ppm level
Xiaolei Shi, Qi Chen, Ji Fang, Koday Varahramyan, Hai-Feng Ji ∗
Institute for Micromanufacturing, Louisiana Tech University, Ruston, LA 71272, USA
Received 6 May 2007; received in revised form 1 August 2007; accepted 2 August 2007
Available online 14 August 2007
Abstract
We have demonstrated that the Al2 O3 -modified microcantilevers (MCLs) can be used to detect low level of moisture. The detection limit for
moisture in nitrogen was 10 ppm. The MCLs’ response time to moisture was less than 3 min. The sensors were stable for months stored under
ambient conditions. The bending amplitudes were proportional to the moisture level and temperature, and the detection of moisture was not affected
by alcohols in the environment.
© 2007 Elsevier B.V. All rights reserved.
Keywords: Microcantilever; Sensors; Moisture; Natural gas; Aluminum oxide; Al2 O3
1. Introduction
Natural gas, a combustible mixture of hydrocarbon gases,
is a vital component of the world supply of energy. Raw natural gas contains water vapor, hydrogen sulfide, alcohol, carbon
dioxide, etc., which need to be removed before the natural gas
enters the pipeline. Water vapor is typically removed by a dehydrating agent such as glycol that absorbs water vapor from the
gas stream. After dehydrating process, industrial gas manufacturers monitor moisture content to meet industry specifications
for pure, dry gas. Excess amounts of water vapor will not only
lower the burning efficiency, but also corrode the pipeline. The
moisture was controlled down to the parts-per-million (ppm) to
billion (ppb) level and monitored by moisture meters.
Current techniques to measure water vapor content include
cooled (chilled) mirrors, electrolytic cells, oscillating crystals,
infrared absorption, metal oxide or polymer capacitive films,
etc. [1,2]. The prices range from a couple of thousand to tens
of thousands US dollars. Besides these devices, length-of-stain
tubes occupy the low end of technology. These detector tubes are
quick, cost-effective, and convenient to operate, but the trade-off
is its low accuracy.
Each device has their advantages and disadvantages. The
optical devices suffer from higher cost and interference from
alcohols although they are the only devices that can handle high
∗
Corresponding author. Tel.: +1 318 257 5125; fax: +1 318 257 5104.
E-mail address: hji@chem.latech.edu (H.-F. Ji).
0925-4005/$ – see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.snb.2007.08.019
corrosive gases since these devices do not directly contact with
gases. In general, other devices are less sensitive than optical
devices, but more cost-effective. With advances in material and
microelectronic technology moving rapidly, manufacturers are
finding ways to overcome disadvantages inherent in these sensors. Furthermore, research on new systems is also undergoing.
Advances in the field of micro-electro-mechanical systems
(MEMS) now offer unique opportunities to design sensitive and
cost-effective analytical methods. Recently, microcantilevers
(MCLs) have been proven to be an attractive platform for sensors with on-chip electronic circuitry and extreme sensitivity
[3,4]. Because the micromechanical aspects of the MCL can be
integrated with on-chip electronic circuitry, it provided an outstanding platform for chemical [5,6] and biological sensors [7,8].
Extremely sensitive chemical vapor sensors based on MCLs
have been demonstrated using selective coatings on the MCLs.
MCLs undergo bending due to molecular adsorption by confining the adsorption to one side of the cantilever. Adsorption
or intercalation of the analyte will markedly change the surface
characteristics of the MCL, and results in the bending of the
MCL. Using Stoney’s formula [9], the radius of curvature of
bending of the MCL due to adsorption can be written as
1
6(1 − v)
=
␦s
R
Et 2
(1)
where R is the radius of curvature for the MCL, ν and E are Poisson’s ratio and Young’s modulus for the substrate, respectively,
t the thickness of the MCL, and ␦s is the film stress.
Author's personal copy
242
X. Shi et al. / Sensors and Actuators B 129 (2008) 241–245
MCL-based moisture sensors have been developed using
SiO2 , Si3 N4 , and polymer coatings [10–13]. However, these sensors are not sensitive enough for ppm level moisture detection.
Furthermore, sensors developed by these sensing materials were
affected by alcohols existed in the background.
Aluminum oxide (Al2 O3 ), on the other hands, has been
demonstrated highly selective for moisture measurement
[14–16] and an excellent material for measurement of moisture
in most industrial gases. It is anticipated that the adsorption of
water molecules on the Al2 O3 thin film will result in the tensile
force on the Al2 O3 film that will deflect a MCL modified by the
Al2 O3 film. In this paper, we report the sensitivity, temperature
effects, and selectivity of Al2 O3 -modified MCLs for low level
moisture detection.
through the cell at a constant 100 mL/min flow rate during each
experiment. When the stable baseline was reached the moisture
gas was switched in for testing. The flow rate is relatively high
and the gas is expected to flow through the flow cell within 1 s.
The bending of the MCL was measured by monitoring the
position of a laser beam reflected from the gold-coated side of
the MCL onto a four-quadrant atomic force microscope (AFM)
photodiode. We define bending toward the gold side as “upward
bending”; “downward bending” refers to bending toward the
Al2 O3 side. When the adsorption occurs on the Al2 O3 surface, in general, the upward bending is caused by repulsion or
expansion of molecules on the Al2 O3 surface, which is so called
compressive stress.
3. Results and discussions
2. Experimental
3.1. Detection limit
2.1. Materials
In our experiments, we used commercially available silicon
MCLs (Veeco Instruments, Santa Barbara, CA). The dimensions
of the V-shaped silicon MCLs were 180 ␮m in length, 25 ␮m in
leg width, and 1 ␮m in thickness. One side of these MCLs were
covered with a thin film of chromium (3 nm) and followed by
a 20 nm layer of gold, both deposited by e-beam evaporation.
Another side of MCL was deposited by a layer of 100 nm thick
aluminium (Al). The Al film was oxidized by oxygen in a high
vacuum chamber while oxygen gas flowing through at 100 ◦ C.
In comparison with conventionally direct deposition of Al2 O3 ,
oxygen oxidation provides a way to obtain compact thin oxide
layers. Ethanol was used in experiments where alcohols were
mentioned.
2.2. Gas system
Instead of natural gas, dry nitrogen was used as the carrier gas
for sensing validation. This is reasonable and convenient since
alkanes do not interact with Al2 O3 . Dry nitrogen was passed
through a gas bubbler containing distilled water used to generate
wet gas. Dual stage gas regulators for wet and dry gases controlled the gas flow into a gas mixing setup. The desired moisture
level was obtained by controlled mixing of the dry and wet gases.
The magnetic heater and thermometer as well as a water-bath
provide the temperature control of the vapor generation system.
The moisture level of the final mixture was measured using a
Meeco Waterboy moisture meter (Warrington, PA) with a range
of 1 ppm to 5000 ppm and an accuracy of ±5%. The flow rate
of the gas inside the cell was 100 mL/min. For experiments at
temperatures of 30–50 ◦ C, a heated water-bath was used to maintain the gas temperature. The volume of the sample glass cell
including the plumbing was 0.5 cm3 , thus ensuring fast exchange
of gases. Typically 10–20 min will be needed to stabilize the
cantilevers to reach a stable baseline prior to the measurement.
Fig. 1 compares the bending response of the modified MCL
to various moisture levels in nitrogen at a 100 mL/min flow
rate. The moisture gas was switched in at the marked time. The
MCLs underwent upward bending and the maximum deflection amplitude depended on the concentration of moisture.
After approximately 3 min, the dry nitrogen was switched back
through the fluid cell, and the MCL bent downward back to their
original positions. Fig. 1 insert shows the reproducible response
of the MCL to 30 ppm level of moisture.
The sensing was independent of the aluminum thickness.
The adsorption follows a Langmuir model (Fig. 2). The rate
of formation of a fraction of a monolayer, θ, is proportional to
the concentration of water molecules and to the fraction of the
surface remaining free of adsorbate, 1 − θ. Thus, the cantilever
bending versus time follows the relationship [3]
3(1 − υ)L2
Z =
␦s ∝ 1 − exp(−kt)
(2)
ET 2
2.3. Deflection measurement
A MCL is placed in a flow-through glass cell (Veeco Instruments, Santa Barbara, CA) and dry nitrogen gas was passed
Fig. 1. Deflection of Al2 O3 -modified MCLs vs. time at various moisture levels
in nitrogen at 30 ◦ C. The gas flow rate was 100 mL/min. Inset: deflection of the
MCLs vs. time after repetitive exposure to 30 ppm moisture in nitrogen.
Author's personal copy
X. Shi et al. / Sensors and Actuators B 129 (2008) 241–245
Fig. 2. Deflection and simulation of Al2 O3 -modified MCLs vs. time at a
100 ppm moisture level in nitrogen at 30 ◦ C. The simulation was done using
Eq. (2).
where k is the reaction rate, and t is the time. The k was calculated
to be 0.02 s−1 using a non-linear curve-fitting method to fit the
observed experimental data.
These results supported the recent study that the hydrated
alumina surface is terminated by a monolayer of OH under ambient conditions and water can molecularly adsorb on top of the
OH-terminated surface [17].
Fig. 3 shows the MCL maximum deflection amplitude versus
the vapor concentration. The maximum deflection amplitudes of
the MCLs were proportional to the concentrations of moisture.
The MCL deflection increased as the concentration of moisture
increased. The lowest detectable concentration was obtained at
10 ppm, which was significantly improved over the SiO2 - and
Si3 N4 -modified MCLs with a detection limit at approximately
200 ppm [10,11]. However, the sensitivity cannot compete with
243
Fig. 4. Deflection of Al2 O3 -modified MCLs vs. time upon exposure to 200 ppm
of moisture in nitrogen at different temperatures. The flow rate was 100 mL/min.
other mature techniques yet. The detection limit of moisture
is of the ppb level with electrolytic technique [18] or optical
techniques such as FTIR spectroscopy [19] or semiconductors
[20].
The larger MCL deflection suggests larger surface stress
change on the MCL surface at higher moisture concentrations.
For a 32 nm maximum deflection corresponding to 200 ppm of
moisture, the surface stress change was 0.083 N/m according
to Eq. (1). A control experiment performed with an Al/Si/Au
MCL to moisture showed no deflection. The lifetime tests were
conducted on MCLs with 3 months storage under ambient condition. The deflection of these MCLs showed a similar profile
and bending amplitude to those in Fig. 1 (data is not shown).
3.2. Temperature effect
Temperature compensation in the final device is needed in
many moisture meters. MCL deflection versus time under different temperature (Fig. 4) were investigated to evaluate the
performance of the proposed sensors in a temperature range
between 30 ◦ C and 50 ◦ C. This is important for sensor accuracy; especially, the Al2 O3 -modified cantilevers deflect upon
temperature change [21,22].
The results showed that the MCL deflection amplitude
increased at high temperatures, which is a typical phenomenon
for metal oxide based sensors. Except for the bending amplitude, the MCL response and recover profiles were similar under
different temperatures. These results suggest that temperature
calibration is needed for accurate moisture measurement under
different temperatures.
3.3. Selectivity over alcohol
Fig. 3. Deflection amplitude of Al2 O3 -modified MCLs vs. the concentration of
moisture in nitrogen.
The chemicals in the natural gas include ppm level of mercaptans, CO2 , and alcohols. Mercaptans and CO2 do not interact
with Al2 O3 , and our experiments showed that the mercaptans
Author's personal copy
244
X. Shi et al. / Sensors and Actuators B 129 (2008) 241–245
controlled by the thermodynamics. It took 100 s to reach equilibrium. Comparing to the current moisture detection systems,
the microcantilever sensor has a relatively fast response. One
way to shorten the response time might be the collection of
the data at the fixed time before reaching equilibrium. Other
characteristics including the long-term stability, and especially
non-interference from alcohol make the cantilever approach very
competitive. The detection limit may be further improved by fine
tune of the coatings.
Acknowledgements
This work was supported by NSF under SGER ECCS0643193 and Board of Regent Industrial Ties and Research
subprogram under contract number LEQSF(2005-04)-RD-B19.
References
Fig. 5. Deflection of Al2 O3 -modified MCLs vs. time upon exposure to 100 ppm
of moisture, 100 ppm of alcohol, 100 ppm of moisture, and 100 ppm of moisture + 100 ppm of alcohol in nitrogen at 40 ◦ C, respectively. The flow rate was
100 mL/min.
and CO2 , with a concentration as high as 1000 ppm, do not affect
the bending of the cantilevers (data is not shown). The effect of
alcohols is the major concern of many moisture meters for low
level moisture detection since the alcohols generally interfere
with moisture detection and cause error. Alcohols cause significant error to IR devices and relatively less effects on other devise
[1,2]. Our initial test showed that the SiO2 - and Si3 N4 -modified
MCLs were largely affected by alcohols (data is not shown),
which disqualify them for accurate moisture detection without
calibration when alcohols exist in the environment.
The potential interference of alcohols on the Al2 O3 -modified
MCLs was evaluated in this study, as shown in Fig. 5. In
these experiments, the MCL deflected 23.5 nm upon exposure
to 100 ppm moisture; no response to 100 ppm alcohol; deflected
again 24 nm upon exposure to 100 ppm moisture; and deflected
25 nm upon exposure to 100 ppm moisture + 100 ppm alcohol.
These results showed that (a) the MCL did not deflect upon
exposure to alcohol; (b) the MCL response to moisture was not
interfered by alcohol after exposure to alcohol; (c) the MCL
responses to moisture were the same with and without alcohol
in the environment.
4. Conclusion
The feasibility of a technology for low level moisture
detection depends on many factors: sensitivity, response time,
accuracy, long-term stability, temperature coefficient, and susceptibility to contaminants, such as alcohol, and cost. Our results
showed that Al2 O3 -coated MCLs are excellent sensors for low
level moisture detection and may be used for moisture monitoring in low level moisture environment, such as in the pipeline,
chambers to store moisture sensitive products, and high voltage
engineering and accelerator systems. The adsorption of moisture on the Al2 O3 surface followed a Langmuir model and is
[1] M.J. McDonough Jr., Device for water vapor and hydrocarbon dew point
determination, Proc. Int. School Hydrocarbon Meas. 75 (2000) 512–515.
[2] (a) B.J. Mychajliw, Determination of hydrocarbon dew point in natural gas,
Proc. Int. School Hydrocarbon Meas. 79 (2004) 111–113;
(b) B.J. Mychajliw, Devices for water vapor and hydrocarbon dew point
determination in natural gas, Proc. Int. School Hydrocarbon Meas. 78
(2003) 501–503.
[3] H.P. Lang, M.K. Baller, R. Berger, Ch. Gerber, J.K. Gimzewski, F.M. Battiston, P. Fornaro, J.P. Ramseyer, E. Meyer, H.J. Guntherodt, An artificial
nose based on a micromechanical cantilever array, Anal. Chim. Acta 393
(1999) 59–65.
[4] T. Thundat, E.A. Wachter, S.L. Sharp, R.J. Warmack, Detection of mercury vapor using resonating cantilevers, Appl. Phys. Lett. 66 (1995) 1695–
1697.
[5] H.-F. Ji, R. Dabestani, E. Finot, T. Thundat, G.M. Brown, P.F. Britt, A novel
self-assembled monolayer coated microcantilever for low level cesium
detection, Chem. Commun. (2000) 457–458.
[6] Z.G. Davis, G. Abadal, O. Kuhn, O. Hansen, F. Grey, A. Biosen, Fabrication
and characterization of nano-resonating devices for mass detection, J. Vac.
Sci. Technol. B 18 (2000) 612–616.
[7] M. Sepaniak, P. Datskos, N. Lavrik, C. Tipple, Microcantilever transducers: a new approach in sensor technology, Anal. Chem. 74 (2002) 568A–
575A.
[8] R. Raiteri, H.-J. Butt, Measuring electrochemically induced surface stress
with an atomic force microscope, J. Phys. Chem. 99 (1995) 15728–15732.
[9] P.R. Muller, R. Kern, About the measurement of absolute isotropic surface
stress of crystals, Surf. Sci. 301 (1994) 386–398.
[10] H.-J. Butt, A sensitive method to measure changes in the surface stress of
solids, J. Colloid Interface Sci. 180 (1996) 251–260.
[11] A.C. Stephan, E.L. Finot, H.-F. Ji, T.G. Thundat, L.F. Miller, R.J. Warmack,
Micromechanical measurement of active sites on silicon nitride using free
surface energy, Ultramicroscopy 91 (2002) 1–8.
[12] B.S. Berry, W.C. Pritchet, Bending-cantilever method for the study of
moisture swelling in polymers, IBM J. Res. Dev. 28 (1984) 662–667.
[13] C.Y. Lee, G.B. Lee, Micromachine-based humidity sensors with integrated
temperature sensors for signal drift compensation, J. Micromech. Microeng. 13 (2003) 620–627.
[14] L. Westcott, G. Rogers, Humidity sensitive MIS structure, J. Phys. E: Sci.
Instrum. 18 (1985) 577–586.
[15] R.S. Jachowitz, Electronics to microelectronics, in: Proceedings of the 4th
European Conference on Electrotechnics—EUROCON’80, 1980.
[16] M.G. Kovac, D. Chleck, P. Goodman, A new moisture sensor for in
situ monitoring of sealed packages, Solid State Technol. 21 (1978) 35–
39.
[17] N. Akmal, Y.N. Zhang, Trace-moisture sensor based on the electrolytic
technique, ACS Symp. Ser. 690 (1998) 174–187.
Author's personal copy
X. Shi et al. / Sensors and Actuators B 129 (2008) 241–245
[18] P.J. Eng, T.P. Traintor, G.E. Brown Jr., G.A. Waychunas, M. Newville, S.R.
Sutton, M.L. Rivers, Structure of the hydrated alpha-Al2 O3 (0001) surface,
Science 288 (2000) 1029–1033.
[19] B.R. Astallard, L.H. Espinoza, R.K. Rowe, Trace water-vapor detection in
nitrogen and corrosive-gases by FTIR spectroscopy, J. Electrochem. Soc.
142 (1995) 2777–2782.
[20] K.F.S. Boumsellek, Instrument design for sub-ppb oxygenated contaminants detection semiconductor processing, J. Inst. Environ. Sci. 40 (1997)
17–21.
[21] F.L. Li, Z.W. Jiang, Design and analysis of a biosensor transducer with
multifunctions, J. Intel. Mater. Syst. Struct. 17 (2006) 823–830.
[22] D. Son, J.H. Jeong, D. Kwon, Film-thickness considerations in microcantilever beam test in measuring mechanical properties of metal thin film,
Thin Solid Films 437 (2003) 182–187.
Biographies
Xiaolei Shi was a graduate student at Institute for Micromanufacturing,
Louisiana Tech University. She is currently at West Virginia University for her
PhD degree.
245
Qi Chen received his MS degree from Nanchang University, China, in 2001. He
is currently a PhD candidate at Institute for Micromanufacturing, Louisiana Tech
University. His research interests focus on microfabrication, MEMS devices, and
microsensor characterization.
Ji Fang received his BS degree in electrical engineering from Tianjing University, China in 1965. Since 2002, he is a senior research engineer at Institute for
micromanufacturing in Louisiana Tech University. His research interests include
microfluidic devices and system, sensor, microreactor, and total analysis system
on chip, optical lens system and micro/nano-fabrication technologies.
Kody Varahramyan received his PhD in electrical engineering, Rensselaer
Polytechnic Institute, in 1983. He is currently the director of the Institute for
micromanufacturing, Louisiana Tech University and the entergy professor of
electrical engineering. His research is focused on micro/nano-scale processes,
materials, devices and systems.
Hai-Feng Ji received his PhD degree of chemistry from Chinese Academy of
Science, China, in 1996. He is currently an associate professor at Institute for
Micromanufacturing, Louisiana Tech University. His research interests focus on
MEMS devices, surface modification, and nanoassembly.