Accelerated Articles Anal. Chem. 1995,67, 4221 -4228
A Renewable Liquid Droplet as a Sampler and a Windowless Optical Cell. Automated Sensor for Gaseous Chlorine Hanghui Liu and Pumendu K. Dasgupta*
Department of Chemistry and Biochemistry, Texas Tech Univemify, Lubbock, Texas 79409-1061
A droplet of a reagent solution is formed at the tip of a tube centered in a cylindrical chamber through which a gas sample is aspirated. The solution is continuously pumped; the drop grows, falls, and another drop grows again. The droplet serves not only as a reproducible collector for the sample gas flowing around it but also as a reactor for a chromogenic reaction and as a windowless optical cell. The design and characteristics of this dynamically growing/hUhg droplet-basedgas sensor system are described; the performance parameters are attractive relative to a static drop. In particular, such systems can be intemally calibrated for any humidity effects: at a constant pumping rate, the drop periodhquency is a predictable function of sample relative humidity. The feasibility of the sensor is demonstrated by the fully automated and continuous determination of gaseous chlorine using tetramethylbenzidine solution as a chromogenic collection liquid. At levels relevant to industrial hygiene monitoring, an 18 p L droplet-based sensor equipped with a light-emitting diode photodiode-based detector shows a relative standard deviation of 1.2%(pC12 -900 ppbv, drop period 1.1min) while the corresponding blank standard deviation is equivalent to -1 ppbv. There appears to be a great potential for such drop-based collectors with in situ photometric or electrochemical signal transducers as automated sensor devices in biphasic measurements (trace gases, solvent extraction, etc.). Some 130 years ago, Tatel formulated a remarkable series of laws regarding the magnitude of a drop of liquid formed under different conditions. Even with measurement techniques that were basic by prevailing standards, the remarkable uniformity of repeated drops impressed him: “...notvarying in any case by more (1) Tate, T. Philos. Mug. 1864,27, 176-180. 0003-2700/95/0367-4221$9.00/0 0 1995 American Chemical Society
than the hundredth part of itself’. As a drop forms at the tip of an orifice, initially it is directly attached at the base (i.e., without a neck or stem, such a pendant drop is called sessile) and can be studied under this condition in a stationarymode by stopping the flow. As flow continues, a neck is formed at an accelerating rate until the drop eventually becomes scissile. Drops can also be studied under this dynamic condition of growth and scission. Fifteen years after Tate, Lord Rayleigh provided the first quantitative description of the dynamics of the process and remained interested in various aspects of drop formation at least through the next two decades.2 In more recent years, McMillan et al. carried out much fundamental and applied work using a fiber drop analyzer, a small head containing a liquid output orifice and two fibers, one to bring in the probe light beam and the other to conduct the transmitted light to a d e t e ~ t o r .Much ~ of the relevant earlier literature has been reviewed by McMillan et al. and will not be repeated here. The optical signal during the life of a single drop, termed fiber drop trace, contains much information. The effect of an electric field on the droplet and the resulting optical signal have been studied, and the optical signal resulting from a mechanical shock applied to a stationary sessile drop has been studied as well. viscosity, surface tension, absorbance, turbidity, etc., are some properties of a liquid that can be measured. Like Tate, McMillan et al. find that defining the limits of the reproducibility of the physical aspects of drop formation taxes extant instrumentation: (2) Rayleigh, J. W. S. Proc. R. SOC. 1879,29, 71-97; Philos. Mug. 1899,48, 321-337. (3) McMillan, N. D.; Finlayson, 0.;Fortune, F. M.; Fingleton, M.; Townsend, D.; McMillan, D. D. G.; Dalton, M. J. Meus. Sci. Techno!. 1992,3, 746764. McMillan, N. D.; Fortune, F. J. M.; Finlayson, 0.;McMillan, D. D. G.; Townsend, D. E.; Daly, D. M.; Fingleton, M.; Dalton, M. G.; Cryan, C. V. Reu. Sci. Instmm. 1992,63,3432-3454. McMillan, N. D.; O’Mongain, E.; Walsh, J. E.; Orr,D.; Ge, Z. C.; Lawlor, V. PYOC. SPIE 1993,2005,216227. McMillan, N. D.; O’Mongain, E.; Walsh, J.; Breen L.; McMillan, D. G. E.; Power, M. J.; O’Dea, J. P.; Kinsella, S. M.; Kelly, M. P.; Hammil, C.; Orr, D. Opt. Eng. 1994,33, 3871-3890.
Analytical Chemistry, Vol. 67, No. 23, December 1, 1995 4221
viscosity can be measured with a precision of 0.01%. Depending on the angle of the fibers, the probe beam can undergo total internal reflection many times-by using a multiwavelength detector and 1 ppb Rhodamine B as the analyte, the authors estimated that as little as 0.04 ppb of the dye can be detected if the ratio of the intensity at the absorbance maximum to that at the emission maximum is used as the index. We have been interested in droplets or films as reproducible collectors of gases. It is often necessary to discriminate between gases and particles by some physical means because the same analyte may exist in both phases. An interesting aspect of using a droplet as a collector for gases is that the flux of the vapor leaving the surface due to liquid evaporation inhibits the approach of particles. This characteristic, referred to in the aerosol science literature as diffusiophoresis caused by Stefan flow, may make it possible to eliminate interference from particles without special design. In this vein, we have reported the uptake of ammonia by an acidic drop (with subsequent retraction of the drop and introduction of the contents into a capillary format sequential injection analysis system),' the determination of gaseous nitrogen dioxide by monitoring the temporal development of absorbance in a single stationary drop formed on an inverted U-shaped wire guide as measured by two optical fibers in contact with the liquid? and the uptake of a variety of acid gases on a thin film formed on a wire loop located at the tip of a capillary with subsequent anaiysis of the multiple collected components by capillary elech-ophoresis.6 In the present paper, we report on the analytical use of a dynamic droplet-based gas collection and analysis system. This renewable sensor has many unique attributes not shared by any of its predecessors, perhaps the most important of which is its ability to intrinsically sense the relative humidity of the sample air. The design and characteristicsof a lightemitting diode (LED) photodiodebased system of this type are demonstrated using the colorimetric determination of chlorine gas via its uptake by a droplet of dilute colorless solution of tetramethylbenzidine m B ) and onestep rapid reaction of Clzwith TMB' at the droplet surface to form a yellow product. EXPERIMENTAL SECTION
Chlorine Gas Generation and Gas Sampling. The sampling manifold and the chlorine gas generation device are shown in Figure 1. The gas flow rate of each line is metered by a mass flow controller (Model FC-280, Q l a n General, Torrance, CA). Compressed house air is fed through a pressure regulator (G) and pur8ed by a column containing, in sequence, silica gel, activated carbon, and soda lime. The flow of the pure air is then divided into three streams. One stream, metered through M5. is fnlly humidified by passage through two sequential water-filled bubblers and is then merged with the dry air stream metered by M4. The relative humidity of the merged stream is thus controlled by choosing an appropriate flow rate ratio via M5 and M4. The dry air stream metered by M3 is thermally equilibrated by coil C and flows into a permeation chamber containing a chlorine permeation device P (VICI Metronics, Santa Clara, CA; gravimetrically calibrated to be emitting at 724 ng/min) at 40 cm3(STP)/min. The permeation chamber and the thermal equilibration coil are housed in a thermostated enclosure T maintained at (4) Liu. S.; Dasgupta, P. K Anal. Chem. 1995,67. 2042-2049. (5) Cardoso, A; Dasgupta, P. K Anal. Chem. 1995,67,2562-2566, (6)Kar. S.;Dasgupta, P. K Anal. Chem., in press. (7) %mat. F. B. Tolnnta 1994,42, 2091-2094.
4222 Analytical Chemistry, Vol. 67,No. 23,December 1, 1995
Detection Unit
3-way Valve
I
. .
veht
N3
Vent
N4
T
P
C
I
I
L
s"
'
Compressed
Bubblers
Filtered House Air
Figure 1. Gas flow schematic: G, pressure gauge and regulator: S. air purifying column containing, in sequence, silica gel, active carbon, and soda lime; N1-4, needle valves; M1-5, mass flow controllers; T, thermostated enclosure housing thermal equilibration coil C and permeation tube P.
30 "C. The exit gas is diluted by air metered by M2. Either this calibrant gas or pure air is selected by a threeway valve and flows around the droplet inside the detection unit. The gas is aspirated by pump P at a flow rate determined by M1 (180 cm3(STP)/min, unless otherwise stated). All remaining flow streams are vented. Except as stated, 670 ppbv Clz was used in the experiments. Sampler/Detedor. The sampler/detector unit is schematically shown in Figure 2. The main chamber body is composed of an opaque plastic tube (high-density polyethylene, 9.1 mm id.; 31.1 mm o.d.), machined from rod stock. The gas stream is introduced into the cylindrical chamber at the top, from two oppositely located gas inlets, to achieve more symmetric flow profile inside the chamber. A stainless steel tube (20 gauge, 0.58 mm id., 0.90 mm o.d.), concentrically k e d in the chamber, is tipped with a 0.5 cm long FTFE tube (0.44 mm i.d., 1.02 mm 0.d.) and is used to deliver TMB solution. Without the FTFE tip, we iind that the wettability of a stainless steel tube changes over a period of time, resulting in altered drop characteristics. The TMB solution was pumped by a syringe pump (Model 33, Harvard Apparatus Inc., South Natick, MA) with a 1mL syringe. The flow rate used was 1 mWh, except as stated. This arrangement produces a -18 p L droplet. The light from a high brightness (1oM) mcd) blue (450 nm) LED (BF28OCWBlK-3.6vF-O5W, Ledtronics,Torrance, CA) passes through a 2.4 mm diameter hole (15 mm between the head of the LED and the chamber wall) and illuminates the droplet. On the diametrically opposite side, the transmitted light is conducted by a 1.5 mm 0.d. plastic optical fiber (inserted to a depth of 1.6 mm inside the chamber) to a photodiode equipped with an interference filter centered at 460 nm (RXTK-460, half-bandwidth 10 nm; Intor, Socorro, NM). The center line of the optical fiber is 1.95 mm below the tip of the dropforming capillary; this distance (denoted as h in Figure 2)
z
Reagent Delivery Tube
-a
Gas Inlet Tygon Tube
PTFE Tube,
1. L:-I
Reference Photodiode
\QL;rtz Window Rubber Ring
0.0
Detect, Photodiode
0.0
I
I
20.0
40.0
‘
60.0
1
t
Time, s Figure 3. Absorbance signal as a function of the drop head height h. Solid and dashed lines represent traces for pure air and sample air, respectively. The value of h is (a) 0.75. (b) 1.35,(c) 1.95,and (d) 2.35 mm. The drop period is indicated by t.
Figure 2. Schematicdiagram of the detection unit. his the vertical distance between the tip of the droplet forming capillaly and the center
line of the optical fiber. can be varied (vide infra). A reference photodiode (S2007. Electronic Goldmine, Phoenix, AZ) is placed at the back plane of the LED to provide a dual-beam, referenced detection arrange. ment. The detection system is controlled and the data are acquired by a lZbit data acquisitionboard (DAS16G1, Mehabyte, Taunton, MA) hosted by a personal computer equipped with a 80486 class processor. The overall electronic arrangement and software control is similar to that previously described elsewhere! Briefly, the LED is pulsed on and off by the computer at 20 Hz. The photocurrents for both the detector and the reference photodiode are converted to voltage signals and digitized. A total of 500 individual digitized results for each photodiode made over a period of 16.667 ms are averaged and stored as a single number. The absorbance is then computed and stored as a single datum. The “dark” photocurrent, which includes the contribution of ambient light lealdng to the photodiodes, is monitored in the LED off period and is thus compensated for. The combination of a pulsed bright source, a wavelength-selective signal photodiode, efficient coupling between the light transmitted through the drop and the collector fiber, and signal averaging makes possible highprecision absorbance measurement without extraordinary measures to shield the detector from ambient light. Absorbance is computed as -log[(Za - Id?/(& - &’)I KO where Id and I, are the respective photocurrents generated by the detector and the reference photodiode, I{ and I: are the corresponding dark currents, and KO is an offset constant that compensates for the inequality between the intensities of the reference beam and the probe beam directed to the drop. (In effect, & is the equivalent of a zero adjustment control). The detector photocurrent Id is affected not only by the absorption of light by the drop but also by reflection, scattering, and refraction.
+
(8)Dasgupta. P. II: Bell.”. H.S.: Uu. H.:Lopez,1. L:Loree. E. L Monis. K Petersen. IC: Mir, K A Tolnnfn 1993,40, 53-73.
In particular, the receivingfiber optic is of finite size and numerical aperture. The refracting properties of the drop changes with its growth, and this can dramaticallyaffect the measured absorbance. For these reasons, the absorbance, as computed above, is referred throughout the rest of this work as apparent absorbance. Reagents. TMB solution was prepared by dissolving 0.100 g of TMB (Aldrich) in 25 mL. of NSvrdimethylfonnamideand diluting with 50 mL of @phosphoricacid and 25 mL of distilled deionized water? A 4 mL aliquot was diluted to 100 mL with water as the working reagent. TMB has been shown to be a superior choice over other reagents such as syringaldazine or @tolidinefor the selective and sensitive determination of chlorine. RESULTS AND DISCUSSION Position of The Drop Head and Collector Fiber. In the optical train,the droplet acts like a ball lens of temporally variable shape and dimensions. As previously stated, the exact vertical position of the collector fiber relative to the drop head is given by the parameter h. Irrespective of the presence of any analyte, the choice of h determines (1) the precise time during the life of a droplet at which Id reaches a maximum, (2) the maximum value of l a , (3) the specific part of the droplet that is probed by the incident beam at a specilic time, and (4) the path length and orientation of a ray passing through the droplet. Obviously, these factors will affect the detection sensitivity and precision. Figure 3 shows the apparent absorbance recorded during the lifetime of a drop for different values of h. In each case, traces for pure air and the sample gas are shown, respectively represented by solid and dashed lines. The minima in these traces represent the respective maxima in the coupling efficiencybetween the source and the collector fiber during a drop period. The moment of scission of the last droplet is taken to be time zero; the drop period t i s -67 s, and at the end of this time a new drop begins and the clock is reset to time zero. Briefly at the beginning of each trace, the signal is invariant because the droplet has not yet grown enough to interpose itself between the source and the detector fiber; in the following, we refer to this period as the preinterposition period. Trace a is an exception; it shows no preinter-
Analyiical Chemistry, Vol. 67,No. 23, December 1,
1995
4223
2.0
-
c1.6
-
4
/
0.3 5
C 0)
5
E!
Q)
L
0.2
5 1.2 0 0
c
a 0.8 0.4
-
0.1
m
e 8 n a
e Photocurrent
o.4
Absorbance Signal
1
Photoc;rren;
,
+ Absorbance Signal
1
0.0
0.0 0.0
1.0
2.0
3.0
Fiber Protrusion Depth, m m Figure 4. (a) Photocurrent /d with pure air at absorbance minimum and net analytical signal (Anet)for 670 ppbv Clz for different values of h, fiber protrusion depth 1.6 mm; (b) same parameters as a function of the fiber protrusion depth, h = 1.95 mm.
position period. In this case, h is minimum within the tested range.) As may be expected, the preinterposition period increases with increasing h because the droplet has to grow larger to reach the optical path when the dropforming tip is raised to a higher position. After the drop grows large enough to interpose itself between the source and the collector fiber, less and less light reaches the detector and the apparent absorbance eventually attains a maximum value. Past this point, the droplet focuses increasingly more light on to the fiber, causing the absorbance to decrease which eventually reaches a broad minimum (except for case a). The position of the minimum shifts to longer times as h increases. These minima exhibit even lower apparent absorbance relative to that during the preinterposition period, clearly establishing the light-gathering and focusing behavior of the drop. Extraction of the analytical information from the data near the absorbance minima is advantageous because (a) high Id minimizes the effect of ambient light and enhances the signal-tonoise ratio, (b) the absorbance does not change rapidly near the minimum, and (c) the difference between the calibrant gas and the corresponding pure air trace is greatest near the absorbance minimum. Once the temporal location (t,iJ of the absorbance minimum and its absolute magnitude, &min, have been determined for the pure air trace, the analytical signal for a drop with sample gas flowing can simply be taken as the absorbance observed for this drop at tminfrom which the blank value, &in, is subtracted. In this work, we have taken a somewhat more complex approach that, however, is no more difticult to implement when this is practiced through software. This approach, described below, can compensate for any drift in the detector baseline, caused, for example, by dust that has deposited on the receptor fiber since the pure air run. In our method, the analytical signal is obtained in the following manner: (a) 1 s worth of data during the preinterposition period is averaged, (b) 6 s of absorbance data, centered at tmin,is averaged and the baseline value obtained in (a) is subtracted from it; (c) steps identical to (a) and (b) are carried out for pure air and this blank is then subtracted from the value in step b to obtain the blankcorrected absorbance signal. All results reported henceforth are discussed in terms of this absorbance signal, A,,,. 4224
Analytical Chemistry, Vol. 67, No. 23, December 1, 1995
Figure 4a shows I d at tmin for pure air as the sample and A,,, for 670 ppbv Clz as a function of h. Within the range studied, 1.95 mm is the optimum value for h with respect to either I d or A,,,. This value of h was hence chosen for all further work. Figure 4b shows the effect of the fiber protrusion depth (the length of the collector fiber protruding beyond the chamber wall) on I d at tminfor pure air and A,,, for 670 ppbv Clz. A,,, increases monotonically and slowly as the fiber approaches the droplet. In contrast, I d goes through a distinct maximum at a fiber insertion depth of 1.6 mm; this was therefore chosen for further experiments. Effect of Droplet Si.Different size droplets were obtained by varying the diameter of the drop head (0.d. 0.61, 1.02, 1.32, and 1.88 mm); for reasonably thin-walled tubes, it is only the outer diameter that is of importance.* Figure 5a shows the drop volumes are linearly related to the tube diameter as suggested by Tate’s law? For each droplet size, we determined the optimum value of h based on the criteria outlined in the foregoing section and experiments were then conducted at this h. The reagent delivery rate was adjusted to obtain the same drop period regardless of the droplet size. The same parameters as in Figure 4 are plotted in Figure 5b as a function of the quasi-diameter of the droplet (calculated from the drop volume, assuming a spherical geometry). Since a larger lens gathers more light, the observed increase in photocurrent with increasing droplet size is expected. On the other hand, it is observed that A,,, increases with decreasing droplet diameter. We believe this is because the beam transmitted through the droplet preferentially explores the surface layer^.^ Consider that the analyte mass collected by a spherical droplet is proportional to the surface area (= rz). For a beam piercing straight through the droplet, A,,, is expected to be proportional to the product of the concentration (= mass/volume; hence rl)and the path length (= r). Thence, there should be no dependence of the signal on the droplet size at all. Due to the slow diffusion rate in the liquid phase and the lack of enough convection, it is likely the light-absorbing product from the chromogenic reaction between TMB and chlorine is preferentially present in the surface layers. 0~
(9)Adamson, A W. Physical Chemistry of Surfaces, 2nd ed.; Wiley: New York, 1967; pp 21-24.
3
47
r 0.16
3-
- 0.12
0)
U”
iij
C
Q
Q)
50
s0 E
j
2-
- 0.08
g (D
2 1-
- 0.04
8 9
u0
Photocurrent
Absorbance Slgnal
0 0.0 1.0 2.0 Diameter of Drop Head, mm
0.00
2.8
3.2 3.6 4.0 Drop Diameter, mm
Figure 5. (a) Drop size as a function of drop head outer diameter; (b) same parameters as Figure 4 plotted as a .function of drop diameter.
Nevertheless, if the light beam passes straight through the droplet in a radial fashion, exploring different regions of the droplet uniformly, the signal would be independent of the precise distribution of the absorbing product in the radial dimension because the number of absorbing molecules in the beam path would not be affected by the distribution. On the other hand, if the absorbing material is primarily on the outer layers and the incident beam is refracted preferentially through the outer layers, the observed results can be explained. We believe that the preferred guidance through the outer layer occurs because the outer shell of the spherical lens represents a more concentrated solution than the core (both because of evaporation and the uptake of the analyte) and it is also cooled by evaporative cooling. The greater solute content results in a greater refractive index and thus results in the optical behavior that we observe. The increase in A,,, with decreasing drop size is fortuitous; as practitioners of capillary-based analysis techniques well know, smaller is rarely better in so far as the nominal path length for optical absorbance detection is concerned. In this light, it is instructive to reexamine the data in Figure 3 and Figure 4a. Figure 4a shows that A,,, is essentially independent of h in the range of h = 1.35-2.35 mm. However, A,,, is derived from the the absorbance at th,. Figure 3 shows that th, shifts withii this range of h from about 40 to 62 s, by 55%. This increase in sampling time has negligible effect on Anetbecause, from the point of view of detection, the increase in h is tantamount to the use of a larger drop. Although An,, increases with a decreasing droplet size, 1, decreases. The droplet size cannot therefore be indefinitely decreased; lower photocurrents result in increased noise levels. Additionally, a smaller droplet requires more stringent optical alignment. In this work, we have chosen to carry out most of the work with a droplet of -18 p L in volume as a compromise between sensitivity,precision, and facility of implementation.With a large-area optical detector placed directly in the system and means of focusing the incident light on the droplet (or with a tightly focused coherent source), it would doubtless be possible to work with smaller droplets. Effect of Reagent Delivery Rate. A decrease in the reagent delivery rate increases the drop period, effectively increasing the gas collection time, thus resulting in a higher absorbance signal. As may be expected, the drop period is proportional to the reciprocal of reagent delivery rate (Figure 6). While the actual
A
Drop Period, 8
0
Signal
d 0
I I I I I 0.0 4.0 8.0 12.0 Raclprocal of Reagent Delivery Rate, slpL I
Figure 6. Effect of reagent delivery rate on experimental param-
eters. volume of the drop at which scission occurs remains essentially independent of reagent flow rate (neglecting increased content of nonvolatile material due to evaporation and the resulting effects on surface tension, density, etc.; also, within the range of reagent delivery rate studied, hydrodynamic effects of drop volume at scission appear to be negligible), the volume of the reagent solution delivered per droplet, as calculated by multiplying the drop period with the reagent delivery rate, increases with decreasing reagent flow rate because of increased loss due to evaporation. This is also shown in Figure 6. The effect of the reagent delivery rate on the signal can obviously be used to choose the sensitivity and applicable dynamic range of a given detection method; higher flow rates also lead to greater convective mixing. The droplet-sensingtechnique is a unique measurement method where the effective integration period is determined phenomenologically, not by electronics or software. In many regulatory applications, the integration time is specified and the reagent delivery rate can be easily adjusted to match this parameter. Effect of the Sampling Flow Rate on the Analylical Signal. Traditional d ~ s i o n - b a s e dcollection/measurement systems tend to show a general trend that the signal initially increases steeply with increasing sample flow rate and then the slope decreases Analytical Chemistry, Vol. 67,No. 23, December 1, 1995
4225
t 3
140
-I/
f 0.04
1
Figure 7. Effect of
E Y
+ PeakHelght + DropPerlod
I
t
0.00 - -
0.00
60
0.10 0.20 0.30 Sampling Flow Rate, slpm
0
0.40
dVP = Q d t
(3)
sampling flow rate. pCl2 is 440 ppbv.
such that a virtual plateau is observed at higher flow rates.1° The extant theory has no provisions for rationalizing an actual decrease in the signal at high sampling rates. When we previously observed a decrease in the signal at high sampling rates for the determination of NO2 with a static/film drop sampler, we had believed that this occurs due to enhanced evaporation of the reagent components. Figure 7 shows the dependence observed in the present system; the signal is almost invariant with gas sampling rates above 100 cm3(STP)/min, but a minor decrease in the signal with sampling rates above 150 cm3(STP)/min will appear to be real, especially when one considers that the drop period is actually increasing in a minor fashion at higher sampling rates, thus increasing collection time. It is not likely that any changes in the drop composition due to evaporation play any roles in this. At present, we are uncertain as to the exact cause of this behavior (the small magnitude of this makes it difficult to study) but believe that the fluid dynamics in the boundary layer is involved. Effect of Humidity. Drop Period and Relative Humidity. Other factors remaining the same, the drop period is a function of the surface tension of a liquid, a fact that has been exploited as a detection principle in liquid chromatography.ll In the present situation, liquid composition and hence surface tension are essentially invariant but the amount of liquid lost from a droplet due to evaporation is related to the sample gas relative humidity (RH) at constant sample temperature and flow rate. For a static spherical water droplet in a stream of flowing air, Frossling12 previously derived the temporal dependence of the droplet radius r:
where Kl is a constant and Re is the Reynolds number. Equation 1 is readily put in terms of the change of volume due to evaporation, dV,: (10) Dasgupta, P. K. ACS Adv. Chem. Ser. 1993, No. 232, 41-90. (11) Lima, L. R. 111; Dunphy, D.R: Synovec, R E.Anal. Chem. 1994,66,12091216. (12) Frossling, N. Gerlands Beitr. Geophys. 1938, 52, 170.
4226
where KZ= - 2 ~ K l R e ~ Frossling's .~. treatment of a static droplet is well established. Many exotic applications have been described: since the size of a sphere can be measured with extraordinary accuracy by Mie scattering (reportedly to 0.005%), the rate of evaporation of a drop (as determined from the change in size of a levitated droplet) can be used to estimate the vapor pressure of nonvolatile liquids.13 However, the dependence of the drop period of a dynamically growing and falling drop upon the vapor pressure of the liquid in the surrounding gas (in this case, RH), to our knowledge, has never been examined. The volume increment dV, due to pump delivery is given by
Analytical Chemistry, Vol. 67, No. 23, December 7, 7995
where Q is reagent delivery rate. It is obvious that a droplet can be held at the same size when the evaporative loss rate is exactly equal to the pumping rate whence at steady state the fractional relative humidity is given by
RH = 1 + Q/WzrsJ
(4)
It is beyond the scope of the present work, but it is obvious that a high-precision humidity sensor can be constructed by choosing a constant value of Q in the suitable range and measuring the steady state droplet radius ySs. In our situation, however, the pumping rate exceeds the evaporation rate sufficiently so that the droplet always grows to a point of scission. The overall volume change dVis given by the sum of eqs 2 and 3 and is also given by 4nZ dr; whence dt = 4 x 2 dr/(Q
+ K,r(l - RH))
(5)
Integration of eq 5 yields the expression of drop period tin terms of the droplet radius 7 and sample relative humidity:
+ 12/(2&(1 - RH)} + Q2/{K2(1 - RH)3)ln(l + (K,r(l - RH)/Q)11
t = 4x[-rQ/{Kl(l
- RH)'}
(6)
In eq 6, KZ is essentially a pseudodiffusion coefficient (with units of cm2/s), an index of the evaporation rate that depends on the hydrodynamics of the system, viz.,as governed by the inner diameter of the sensor chamber and the air flow rate. In Figure 8, four separate computed curves for drop period vs relative humidity are shown for four assumed values of KZfor a -18 pL volume (equivalent radius 1.62 mm) droplet, as used in the actual experiments. In computational trials, KZ was deliberately varied to see if a good match can be obtained with the experimentally observed results. It can be observed that the experimental data obtained under our test conditions are in excellent agreement with the predicted curve for KZ= cm2/s. A larger range of the drop period can be spanned as a function of relative humidity by using a lower Q if it is desirable to measure relative humidity via the drop period. The precision of the drop period measurement (13) Tang, I. N.: Munkelwitz, H. R. . IColloid . Inte&ce Sci. 1 9 9 1 , 141, 109-118.
Figure 9 shows the predicted vs observed response values over a large range of relative humidity utilizing this approach; clearly accurate compensation is possible. Reproducibility. Without a droplet in the optical train, as in the preinterposition period, the peak-to-peak absorbance noise is 0.2 mAU. The coefficient of variation for the measurement of 890 ppbv chlorine is 1.2% (n = lo). The standard deviation of the blank (pure air) measurements corresponds to a concentration of 1.1ppbv. This work was undertaken to determine the feasibility
K, ,1OJcmYa
_ - _ - - .-0.03
- -1.0 -3.0
a 80
'
- - -
\
-6.0
\
I
\ \
P
70
E
Q
p:
in the present arrangement is typically 150 ms or better; a relative precision of 0.2%should therefore be readily attainable. Calibration Behavior and Compensation for Relative Humidity Weds. The obvious effect of relative humidity is that the drop period and thus the effective sampling time is altered. There may be other effects of relative humidity; many analytes associate with water in the gas phase and the effective diffusion coefficient is thus decreased; we have observed this, for example, for NH3.5 Experimentally, we observe that at any given relative humidity, Clz concentration has no effect on the drop period. At any given relative humidity, Anetvaries linearly with chlorine concentration with an intercept that is statisticallyindistinguishable from zero; the regression equations (five individual concentrations between 0 and 1100 ppbv, n = 4 at each concentration)are listed in Table 1. It is interesting to note that, under our test conditions, the drop period changes from 65.4 s at 80%RJ3 to -67.7 s at 0% RH, a change of less than 4%. The calibration slope within this relative humidity range changes by 32%;clearly, effects other than an increase in sampling time are involved. The present system provides data not only for the observed signal but also as to the drop period, related to the sample relative humidity. If calibration data are available for different relative humidity values as a three-dimensional matrix (sample concentration, response, drop period), it is obvious that a variety of chemometric approaches will allow accurate evaluation of the concentration sampled given the response signal and the drop period. The linear behavior and essentially zero intercept of the present system allows a very simple approach, however. The calibration slopes are related linearly to the drop period within the relative humidity range studied (040%) as slope (mAu/ppbv ClJ = (2.0002 f 0.0726) x 10-'t - 1.1622 It 0.0482,
lZ = 0.9974 (7)
where t is the drop period. A single equation then allows the computation of the sample concentration, regardless of relative humidity:
C1, (ppbv) =Anet (mAU)/(2.0 x lo-?,
s - 1.162)
(8)
E
50
i/ 0 0
100 150 200 Obsewed Response, mAU
50
0 250
Figure 9. Experimentally observed response at different humidities and CI1concentrations vs values predicted from the transposed form of eq 8. The chlorine concentrationstested are 0, 440, 670,880, and 1110 ppbv. The first four points after the origin, centered at -75 mAU, represent the 440 ppbv measurements.The.line drawn is not the bestfit line but the 1:l corresponsence line of unit slope. The best-fit line has a slope of 0.9828.
of a sensor for industrial hygiene applications that can provide an alarm within 2 min when the concentration of Clz reaches 1 ppmv. Because the performance obviously far exceeds this requirement, we made no explicit effort to determine the limit of detection. Based on the observed standard deviation of the blank, low ppbv levels should be detectable with the present experimental arrangement. Sensitivity could obviously be further improved by changing one or more operational parameters. CONCLUSIONS AND POTENTIAL APPLICATIONS
The feasibility of a dynamic droplet-based gas sensor using a windowless optical cell has been demonstrated. The system is simple, easily automated, and low cost (which could be lowered further by using a gravity-based or pressurized pumping system, presumably with some penalty to precision). Highly precise results are obtained with a indefinitelyrenewable collection surface with real-time detection and intrinsically adjustable integration time. A windowless cell, as represented by a drop, may solve some vexing problems in liquid phase optical detection, e.g., entrapment of gas bubbles in the illuminated volume or fouling of windows by adherent deposits. Both of these may be of particular importance in process analysis-turbidimetric applications can particularly benefit. In recent work conducted in this laboratory, the detection technique used with this sensing a p proach has been extended to luminometry and electrochemical Analytical Chemistry, Vol. 67, No. 23, December 1, 7995
4227
Table 1. Best Fit Calibration Equation at Various RHa
+
RH (%)
drop period, s
equation
0 26.7 53.3 80.0
67.72 f 0.12 66.78 i 0.16 66.08 0.16 65.36 f 0.14
absorbance, mAU = (0.1940 f 0.0029)pCl2,ppbv + (0.5 f 2.1) absorbance, mAU = (0.1747 f 0.0027)pC11,ppbv (2.0 & 2.0) absorbance, mAU = (0.1593 f O.O02l)pC12,ppbv - (0.2 i 1.5) absorbance, mAU = (0.1473 f O.O031)pC12,ppbv - (0.8 f 2.3)
*
+
0.99934 0.99928 0.99950 0.99864
Test concentration range 0-1100 ppbv.
detection. The merit of microdroplets for fluorescence detection is already well-kno~n.~* The droplet-sensing technique can obviously extend beyond gas measurement: solvent extraction with a drop of an immiscible solvent is an obvious possibility. Using an annular drop head, it will be possible to form one liquid film on top of an immiscible liquid drop, a novel path to a renewable liquid membrane-based sensor. Finally, a renewable exposed liquid surface as a collection/measurement interface may benefit from geometries other than that of a scissile or static drop; a thin liquid film offers higher (14) Barnes, M. D.; Whitten, W. B.; Ramsey, J. M.Anal. Chen,1995,67,418A427A.
4228 Analytical Chemistry, Vol. 67, No. 23, December 1, 1995
sensitivity because of increased surface-to-volume ratio.6 A liquid jet may act as an optical wave guide as well as a collection interface and permit extraordinarily fast response times. At this time, we believe that the potential of this type of systems is limited by imagination rather than practical obstacles. Received for review July 21, 1995. Accepted September 25, 1995.@ AC9507350 @Abstractpublished in Advance ACS Abstructs, November 1, 1995.
