Environ. Sci. Technol. 2008, 42, 2072–2078
Laboratory Tests on Mercury Emission Monitoring with Resonating Gold-coated Silicon Cantilevers J A R O S L A W D R E L I C H , * ,† CALVIN L. WHITE,† AND ZHENGHE XU‡ Department of Materials Science and Engineering, Michigan Technological University, Houghton, Michigan 49931, USA and Department of Chemical and Materials Engineering, University of Alberta, Edmonton, Canada
Received July 12, 2007. Revised manuscript received October 9, 2007. Accepted December 26, 2007.
To measure extremely low concentrations of mercury vapor in gases as encountered in flue gases of coal-fired power plants, accurate and reliable online and/or portable mercury detection systems are needed. As discussed in this communication, resonating silicon-based cantilevers coated with thin films of gold change their resonant frequency when exposed to mercury vapors and could serve as the basis for such sensing devices. Two different types of commercial AFM cantilevers, which differed by physical dimensions and surface finish, were coated with a 10 nm film of gold and were tested in streams of argon containing mercury. The argon flow rates ranged from 5.7 to 57.4 mL/min, carrying mercury vapors at concentrations between 37 and 700 µg/m3. The results show that smaller cantilevers (∼140 µm × 40 µm × 4 µm) with a resonant frequency of 270–275 kHz were sensitive to less than 10 picograms of mercury, whereas larger cantilevers (∼245 µm × 50 µm × 7 µm) with a resonant frequency of 155-165 kHz have a sensitivity about 10 times lower. The results indicate that the kinetics of mercury capture by the gold coating follows a simple power law-correlation with the mass change (∆m) being proportional to tn, where t is the capture time and n depends strongly on the concentration of mercury in the gas. It is also demonstrated that the mercury can be stripped off the gold coating by heating to 350 °C, which would allow the cantilevers to be regenerated and reused.
systems are needed. There are a few commercial mercury vapor monitoring systems on market that rely on either cold vapor atomic absorption spectroscopy, cold vapor atomic fluorescence spectroscopy, atomic emission microscopy, UV differential optical absorption spectroscopy, or other analytical techniques (4, 5). All of these techniques rely on a pretreatment unit to covert nonelemental mercury to elemental mercury prior to spectroscopic measurement. Similar pretreatment units might be useful in combination with the resonating cantilever based techniques described here. Commercially available sensors suitable for in situ monitoring of low mercury concentrations are often mercury capture cartridges filled with gold- or silver-coated beads of glass, silica, alumina, or other material. These cartridges preconcentrate mercury for subsequent ex situ analysis with conventional laboratory analytical instruments such as a cold vapor atomic absorption spectrometer (6). Preconcentration of the mercury is necessary because of the low concentration of mercury in flue gases, which is below the detection limit of atomic absorption spectrometers. Some of the more sensitive spectroscopic techniques mentioned above can detect mercury vapor at a concentration of 1-10 µg/m3 in inert gas, but their sensitivities generally drop 50–100 times in air. This communication describes a mercury detection technique based on the change in resonant frequency of gold-coated cantilevers resulting from mercury adsorption and discusses the prospects for its use as a mercury sensor. A general concept for sensing mercury vapor in gas streams is schematically illustrated in Figure 1. It relies on measuring the resonant frequency of a miniature beam (cantilever) made of silicon or other elastically stiff material. The beam is coated with a thin film of noble metal such as gold, silver, or copper. Noble metals have a high affinity for mercury and form amalgams with metallic mercury. The resonant frequency of the beam decreases with increasing mass during the mercury adsorption/amalgamation process. Quantitatively, the increase in the mass of the beam (∆m) is given by eq 1 (7), ∆m )
(
1 1 k - 2 (2π)2b fHg2 f
)
(1)
where k is the cantilever spring constant (also known as force constant), fHg and f are resonant frequencies of loaded and unloaded cantilevers, respectively, and b is a geometrical
Keywords: Mercury emission monitoring; mercury sensor; mercury adsorption; oscillating cantilever
Introduction Coal-fired power plants are a major source of mercury emission with a total global mercury emission at 2–4 kt/ year worldwide (1). The concentration of mercury in coal ranges from 0.01 to 3.3 ppm (2), and the concentration of mercury in flue gas is typically from 1 to 10 µg/m3 (1, 3). To measure such low concentrations of mercury vapor in flue gases, accurate and reliable online mercury detection * Corresponding author e-mail:
[email protected]; phone: 906487-2932. † Michigan Technological University. ‡ University of Alberta. 2072
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 6, 2008
FIGURE 1. Illustration of the concept of the resonating cantilever sensor for monitoring mercury concentration in flue gas. 10.1021/es071724e CCC: $40.75
2008 American Chemical Society
Published on Web 02/19/2008
parameter. [b ) 0.24 for rectangular beams with an additional mass uniformly distributed over the surface of the beam, and n ) 1 for rectangular beams with a mass added to the end of the beam.] Micromachined, gold-coated cantilevers used in atomic force microscopy (AFM) have been explored as possible chemical sensors for detection of mercury vapor (8–10). Thundat et al. (8) showed that the commercial silicon nitride microcantilevers coated only at the end with gold decreased their resonant frequency linearly with exposure time when exposed to air containing approximately 30 µg/m3 of mercury (presumably at room temperature), in qualitative agreement with eq 1. The resonant frequencies of similarly exposed cantilevers that were coated along their entire length increased, rather than decreased, in a nonlinear manner with respect to exposure time. The increase in resonance frequency in the latter case was attributed to changes in the stress state and stiffness, associated with amalgamation of the mercury with gold. These fully-coated cantilevers also exhibited a static deflection that increased with exposure time, possibly explaining the build-up of stresses during mercury adsorption. The frequency responses of both types of cantilevers were sensitive to adsorption of as little as a picogram of mercury, and the response times were on the order of a few minutes. Thundat et al. also reported that mercury could be desorbed from the contaminated cantilevers by heating to 150–170 °C, allowing them to be reused. In a subsequent study directed primarily at the build-up of adsorption-induced stresses in cantilevers, Hu et al. (9) found a little difference in the response of cantilevers coated with gold films of different thickness, suggesting very limited mercury diffusion into the bulk of a gold film at room temperature. This result seems to be consistent with other research indicating that mercury adsorbed at room temperature remains at or very near the top layer of the gold surface (11, 12). At higher temperatures, the mercury can amalgamate with gold and can diffuse as much as several nanometers into the gold (11). Whereas Thundat et al. employed silicon nitride cantilevers having a triangular cross section, Rogers et al. (10) tested thin silicon cantilevers having a roughly rectangular cross section and manufactured with an integrated piezoelectric ZnO actuating element and a gold coating for sensing along most of the length of one side. Like Thundat et al., Rogers et al. also report increased (rather than decreased) resonant frequencies upon exposure to mercury vapor. In this paper we report on the kinetics of changes in the resonant frequency of gold-coated single-beam silicon AFM cantilevers exposed to argon-mercury vapor mixtures having a range of mercury concentratios. The effect of mercury vapor concentration on oscillation characteristics of such cantilevers does not appear to have been previously studied. Although the kinetic responses of cantilevers to mercury adsorption were nonlinear, analysis of the experimental data indicates that the weight gain (∆m) depends on tn (t ) time), where the values of n depend strongly on the concentration of mercury in the gas. Our tests involved two types of commercial cantilevers, which differed by physical dimensions and surface finish. In contrast to the previous research cited above, our results indicate that mercury captured by the gold coating on the cantilever has a negligible influence on the spring constant of the cantilever and can be ignored in analysis of the cantilever responses to mercury vapors. Finally, in agreement with the results reported by Thundat et al., we also show that, once captured by the cantilever, mercury can be stripped off the gold coating at elevated temperatures, allowing the cantilever to be reused.
TABLE 1. Source and Technical Data on Cantilevers Used in This Studya technical data
type A cantilever
type B cantilever
cantilever type/ distributor
NCL Nanosensors, Switzerland
material/coating
silicon: no coating
dimensions length [µm] width [µm] thickness [µm] force constant [N/m] resonant frequency [kHz]
BS-Tap300Al Budget Sensors, Bulgaria silicon coating: 30 nm aluminum
225 ( 5 38 ( 5 7.0 ( 0.5 31–71
125 ( 10 30 ( 5 4(1 20–75
170–210
200–400
a Technical data are those provided by the distributors of cantilevers.
Experimental Section Two types of commercial AFM silicon cantilevers, designated as type A and type B, were used in this study. In total, nine cantilevers were tested, five of type A and four of type B. Specifications for the cantilevers are provided in Table 1. Aside from having different dimensions, and therefore different resonant frequencies, type B cantilevers were coated with a 30 nm aluminum film. The as-received cantilevers were sputter-coated with a 10 nm thick gold film using a Hummer 6.2 sputtering system (Anatech Ltd.). The average pressure of the vacuum was 0.075 Torr with a plasma discharge current of 0.015 A during deposition. Morphology and coverage of gold coatings for type A and B cantilevers were examined under a fieldemission scanning electron microscope (FE-SEM, Hitachi S-4700, Hitachi Inc.). An accelerating voltage of 5 kV and working distance of 2.4 mm were used to resolve nanometerscale features within the films. Surfaces of two cantilevers were also imaged with the Dimension 3000 atomic force microscope (AFM, Digital Instruments) using a silicon cantilever (300kHz, 40N/m) in intermittent contact mode. Scan sizes of 1 × 1 µm2 were used. The surface roughness was quantified using three parameters of roughness: rootmean-square roughness (rms), average roughness (Ra), and roughness parameter (r). These parameters are defined by eqs 2-4,
rms ) Ra )
1 LxLy
∑ (Z
i
- Zave)2
(2)
N
Ly Lx
∫ ∫ |f(x, y)|dx dy 0
(3)
0
r )
A Ao
(4)
where Zave is the average of the Z values (Z-axis representing topographical height features) within the given area, Zi is the current Z value, N is the number of points within the given area, f(x, y) is the surface relative to the center plane, Lx and Ly are the dimensions of the surface, A is the surface area of the scanned gold coating, and Ao is the projected surface area of this scanned area. The total surface area of gold coating (AT) was calculated from measured dimensions of the cantilever multiplied by the roughness parameter (r),
(
AT ) r[w Lt +
h ] 2
)
VOL. 42, NO. 6, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
(5)
9
2073
FIGURE 2. Schematic of laboratory set up used in mercury emission monitoring. where Lt is the length of the rectangular section of the cantilever, h is the length of the triangular section of the cantilever, and w is the average width of the cantilever. The gold-coated cantilevers were exposed to inert gas mixed with mercury vapor for a desired period of time at a controlled temperature and at 1 atm total pressure. The laboratory set up used is schematically illustrated in Figure 2. A small mercury drop (2–3 mm in a diameter) was placed in a glass U-tube and was exposed to an argon flow rate ranging from 5.7 to 57.4 mL/min. The concentration of mercury vapor in the gas stream varied with the argon flow rate, and ranged from around 37 to 700 µg/m3. The mercury-argon mixture was directed into a glass jar (250 mL) in which an AFM holder with a mounted cantilever was placed. The gas leaving the glass jar was passed into a mercury trap filled with glass beads coated with gold to capture remaining mercury vapors. The concentration of mercury in the mercury-argon mixture (cHg) was determined from the amount of mercury trapped by gold-coated glass beads during control experiments (without cantilever), at constant flow rate (Q), and over a controlled time (t). The trap was then heated to 350 °C in a 40 mL/min stream of argon to release the captured mercury. The amount of released mercury (mHg) at elevated temperature was determined using a precalibrated TEKRAN Mercury Detector. The cHg value was calculated using eq 6. [Note that this approach, which is standard in measurement of mercury vapor concentrations, assumes that all mercury in the gas stream is captured by the gold-coated glass beads.] cHg )
mHg Qt
(6)
The analysis of mercury concentration in gas was repeated three times, and average values are reported. Reproducibility of the cHg values determined in triple analysis was found to be better than 5%. After a cantilever was exposed to mercury vapor, it was mounted in a commercial AFM instrument (Nanoscope E, Digital Instruments Inc.) and oscillated by a piezoelectric transducer. Its resonant frequency was then determined using the laser beam-photodiode detection system of the AFM. [Although we use a laser-beam photodiode system of the commercial AFM instrument, other detecting systems could be used in development of industrial sensors (10, 13, 14).] When exposed to the mercury contaminated gas, a shift in the cantilever resonant frequency was observed due to an increase in its mass as a result of mercury capture by gold. A simplified schematic in Figure 1 presents the principles of the sensing technique. Additionally, in selected tests, the spring constant of the cantilever was measured as a function of loading with mercury using a thermal tune method, available in the AFM software package. Details on the thermal tune technique can be found elsewhere (15).
Results and Discussion 1. Morphology and Roughness of Cantilever Gold Coatings. Scanning electron microscopy (SEM) images of type A and 2074
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 6, 2008
B cantilevers are shown in Figure 3. Cantilever A, shown in Figure 3a, has a length of approximately 246 µm and a width of approximately 53 µm, whereas cantilever B (Figure 3b) is about 140 µm long and 37–38 µm wide. The measured dimensions of the cantilevers are slightly outside the specifications provided by the manufacturers (Table 1). The thickness of cantilevers was not examined in this study. Both the resonant frequency and the spring constant of the cantilevers were measured after coating them with a gold film. These measurements were not performed on the asreceived cantilevers, and therefore direct comparison to the manufacturer’s specification is not available. Because cantilever dimensions vary from one batch to another and even among cantilevers from the same batch and because gold coating can vary as well, quantification of the surface area of the gold coating is important if quantification of the mercury adsorption process and comparison of adsorption results between different cantilevers are to be accomplished. SEM images shown in Figure 3, panels c-f, illustrate the morphology of gold coatings sputtered over the top surface of the cantilevers A and B after mercury adsorption. Figure 4 shows AFM images for the same coatings, together with line-scans illustrating differences in roughness of the coatings. The scanning electron micrographs in Figure 3 show that the gold did not form a uniform film on the cantilever surface. On both type of cantilevers, the gold formed islands separated by small strips of uncoated (dark) areas of the cantilever. Both islands of gold and strips of uncoated cantilever surface were larger on the type A cantilever than on the type B cantilever. The differences between the morphology of gold coating on cantilevers A and B may be due to the differences in the binary Au-Si phase equilibrium (eutectic) and the Au-Al equilibrium (multiple intermetallic compounds) (16). The AFM line-scans of these coatings in Figure 4 also indicate that the features on cantilever A are larger than on cantilever B. The values of rms, Ra, and r, and the calculated total surface area of gold coating (Table 2), are larger for cantilever A than for cantilever B. Dimensions of surface features on the AFM images and resulting roughness parameters are approximate values. Both broadening of the asperities/islands and deconvolution of the AFM tip during surface imaging can distort the real dimensions of surface features (17, 18). Such distortions are negligible when imaging gold islands that are only a few nanometers tall and have diameters that are only a few tenths of nanometers, such as those in Figure 4. Theoretical models of AFM tip broadening effects (17, 18) suggest that the error in determination of the surface roughness is less than 5% for the gold-coated cantilevers in this study. 2. Sensitivity of Gold-coated Cantilevers to Mercury Vapor. The changes in resonant frequency, due to adsorption of mercury, are shown in Figure 5 for three single-beam goldcoated cantilevers. Unlike the results reported by Thundat et al. (8), we did not observe any increase in resonant frequency associated with exposure of the cantilevers to mercury vapor. In our experiments, the resonant frequency of cantilevers exposed to mercury vapor always decreased. We also measured the stiffness constants for the cantilevers and found them to be nearly independent of mercury adsorption. The stiffness constant of one of the type B cantilevers was 20.46 ( 0.01 N/m prior to exposure and was 20.39 ( 0.01 N/m after capturing 800 pg of Hg. The increase in the cantilever resonant frequency observed by Thundat et al. was attributed to increases in the cantilever spring constant due to amalgamation of the gold coating with mercury. As reported above, we observed a very small decrease in the cantilever’s spring constant following exposure to mercury vapor. This decrease was less than 1% and would have a negligible effect on oscillation of cantilever.
FIGURE 3. Field emission scanning electron microscopy micrographs of type A and B cantilevers used in mercury emission tests. The resonant frequency of the oscillating cantilever decreased monotonically with time (Figure 5). Two distinctly different regions of curves in Figure 5 are observed. Within the first minute or two, a very significant decrease in the resonant frequency occurred. The frequency continued to decrease, albeit at a lower rate, approaching a nearly linear relationship after about 30 min. No clear correlation between the cantilever resonant frequency change and either mercury concentration or gas flow rate is evident from the results in Figure 5. To make a better comparison among the results, two parameters were calculated: (i) mass flow rate of mercury as supplied to the cantilever per unit surface area of the gold coating (D) using eq 7, and (ii) mass added to the cantilever during mercury emission monitoring experiment (∆m) using eq 1 and assuming a rectangular shape of the cantilevers used (b ) 0.24). D )
cHgQ AT
(7)
Other parameters in eq 7 are the same as in previous equations. The results from Figure 5 are replotted in Figure 6a, showing ∆m as a function of t and, in Figure 6b, as log(∆m) versus log(t). The nonlinear response of the cantilever (Figure 6a) suggests that mercury adsorbs on the surface of gold and saturates the surface region very quickly at room temperature. The slower rate of frequency change after the first few minutes is suggestive of a thermally
activated reaction, involving much slower capture of mercury by the gold coating. The mercury capture levels corresponding to one monolayer of mercury (as calculated for a closed-packed surface layer) are shown for cantilevers A and B in Figure 6a. Saturation levels on both cantilevers are in excess of one monolayer, and for cantilever A it is in the range of 4–5 monolayers of mercury. Previous research indicated that mercury stays on the top layer or near the surface of gold surface when adsorbed at room temperatures (11, 12). At higher temperatures, amalgamation of gold with mercury takes place and mercury can diffuse up to several nanometers from the surface (11). In a subsequent study, Hu et al. (9) found little difference in response of cantilevers coated with thin gold films of different thicknesses, suggesting very little (if any) diffusion of mercury into the bulk of gold film at room temperatures. The saturation effects indicated in Figure 6a suggest that mercury either formed a multilayer on the gold surface or penetrated a very short distance into the gold film. Figure 6b shows that, except at very short adsorption times, the kinetics of mercury adsorption for a given cantilever and adsorption conditions can be well represented by a straight line. For type A cantilevers, the slope depends on the D value for the exposure conditions. Figure 7 shows five sets of n and D values obtained in experiments where the spring constant of the cantilever was measured, VOL. 42, NO. 6, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2075
FIGURE 4. AFM topographic images of the cantilevers’ gold coatings and their cross-sections.
TABLE 2. Roughness Characteristic of Gold Films Sputtered on the Cantilevers roughness characteristic
type A cantilever
type B cantilever
rms [nm] Ra [nm] r [-] total surface area [mm2]
3.6 ( 0.2 2.6 ( 0.2 1.06 ( 0.01 0.013 ( 0.001
1.3 ( 0.1 1.0 ( 0.1 1.02 ( 0.01 0.0049 ( 0.0003
allowing ∆m values to be calculated. The value of the constant n increased approximately linearly with increasing D value. These results clearly indicate that, except at very short exposure times, the kinetics of mercury uptake by the cantilever can be described by the simple power function (eq 8), n
∆m ) Bt
(8)
where B and n are constants. The lowest concentration of mercury used in this study was 37 µg/m3, which is still 4–10 times larger than concentration expected for metallic mercury in the flue gas. We were unable to reliably produce and characterize gas streams with lower levels of mercury in the laboratory setup used for this study. Eventually, a redesigned system will address this problem. The deviation of experimental observations from eq 8 at very short exposure time could arise from several sources. These deviations could result, at least in part, from the limited control that we had over exposure time during the shortinterval tests. The cantilevers were manually inserted and removed from the glass jar, and for short exposure times the measurement of time was subject to large relative errors. This same issue was insignificant for the experiments with longer exposure times or conducted with low concentrations 2076
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 6, 2008
FIGURE 5. Results of resonant frequency decrease for three different AFM cantilevers coated with ∼10 nm film of gold; f is the resonant frequency of the cantilever before its exposure to mercury vapour, cHg is the concentration of mercury in the gas, Q is the gas flow rate used in experiments, and D is the mass flow rate of mercury calculated per unit surface area of gold coating. The tests were conducted at a room temperature (∼21 °C). of mercury vapors. Therefore, the results for the experiments that took 5–10 min or longer and experiments with mercury concentration less than 40–50 µg/m3 were much more reproducible. It is also noteworthy that the transition from more rapid adsorption kinetics to kinetics that follow eq 8 generally occur at mercury capture levels that correspond roughly to one monolayer of mercury coverage. Ignoring any effects of the argon carrier gas, the rate of adsorption from a single component vapor phase (kads) can be analyzed by considering the rate at which atoms or molecules of the vapor strike the
FIGURE 8. The decrease in resonant frequency of the silicon cantilever coated with a 10 nm film of gold exposed to an argon stream of 34.4 mL/min with a concentration of mercury of 233 µg/m3 for 46 min and then after cantilever regeneration at 350 °C for 20 min to the argon stream of 5.7 mL/min having 692 µg/m3 mercury for 46 min. The mercury capture tests were carried out at 21 °C. atoms from the vapor phase strike a surface is given by eq 10 (19), ν ) 2.635 × 1019
FIGURE 6. Results of Figure 5 presented as amount of mercury captured by the gold-coated cantilever vs time of cantilever exposure to the mercury vapour, in both normal-normal and log–log graphs. The theoretical amount of mercury that form a monolayer on each type of cantilevers is marked with dashed line and was calculated for close-packed arranged Hg atoms.
FIGURE 7. The effect of the mass flow rate of mercury per unit surface area of gold coating (D) on the n value of eq 8. surface (ν), and the fraction (S) of those incident atoms that remain on the surface. S )
kads ν
(9)
The parameter S is called the sticking probability and depends on the adsorbate (Hg in this case), its partial pressure (PHg), the substrate (Au), and the temperature (T ). The rate at which
PHg
[
1 √MT cm2 sec
]
(10)
where PHg is in microbars, M is the atomic weight (200.6 g/mol for Hg), and T is the temperature in Kelvin. For the type A cantilever in Figure 6, the mercury concentration is 233 µg/m3, yielding PHg ≈ 0.028 µb, which yields ≈ 3 × 1015 cm-2s-1. From the results in Figure 6 for cantilever A (D ) 104 µg/(mm2 s)), the rate of adsorption at the very early stages of adsorption (0 < t < 1 min) is approximately 4 × 1017 atoms/ (m2s). This implies a sticking coefficient based on eq 9 of approximately 0.01. At somewhat later stages (5 < t < 20 min), the implied sticking coefficient drops to around 5 × 10-4. For the type B cantilever, the sticking coefficient for small coverages (