Anal. Chem. W O O , 62, 2547-2552
proposed peak distortion method could be used to detect “self-interference” or nonlinearity in the calibration curve from inner-filter effects at high analyte concentrations. It is noted that quenching or inner-filter effects will generally be less significant with FIA relative to batch measurements in a conventional sample cell. Both effects are less significant because of the dilution of the sample due to dispersion. Additionally, inner-filter effects will become significant (ignoring dispersion effects) at about 10 times the interferent concentration in a conventional 1 cm path length cell because of the smaller path length of flow cells. Although the peak-distortion method was tested with a single-line FIA system with fluorescence detection for the case that the analyte does not react with reagents, it should also be applicable to other situations. For example, the method could be adapted to cases where a fluorescent product is formed by injection of the sample into a reagent stream in a single-line system or with sequential additional of reagents to the sample carrier stream in a multiple-line system. In these cases, the reference profile would be established by injecting an interference-free standard into whatever FIA manifold is used. The proposed method should also be extendable to FIA systems using detection by methods other than fluorescence. It is only necessary that the interference effect is dependent on the relative interferent concentration and independent of the analyte concentration. The proposed method does not
2547
detect additive interferences because the signal due to interferent varies as analyte signal such that the shape of the peak profile does not change. Dispersion normalization of FIA peak profiles also allows extraction of kinetic information. This application is disc& in ref 15.
LITERATURE CITED (1) Ingle, J. D., Jr.; Crouch, S. R. Specfrochemlcal Analyss; PrenticeHall: Englewood Cliffs, NJ, 1988; Chapter 15. (2) Adamson, K.; Sell, J. E.; Holland, J. F.; Tlmnlck, A. Am. Lab. 1984, 16 ( l l ) , 18-23. (3) Yappert, M. C.; Ingle, J. D., Jr. Appl. Specfrosc. 1989, 43, 759-767. (4) Street, K. W., Jr.; Tamer, M. Anawst 1985, 110, 1189-1172. (5) Hlettje, G. M.; Haugen, G. R. Anal. Chlm. Acta 1981, 123. 255-261. (6) Campi, G. L.; Ingle, J. D., Jr. Anal. Chlm. Acta 1989, 224, 225-234. (7) Ruzlcka, J. Anal. Chem. 1983, 5 5 , 1040A-1053A. ( 8 ) Olsen, S.; Ruzicka, J.; Hansen, E. H. Anal. Chim. Acta 1982, 136, 101-112. (9) BetterMge, D. Anal. Chem. 1978, 50. 832A-846A. (10) Ruzicka, J.; Hansen, E. H. Anal. Chlm. Acta 1978, 9 9 , 37-78. (11) Ruzlcka, J.; Hansen, E. H.; Mosbaek, H. Anal. Chim. Acta 1977. 9 2 , 235-249. (12) Ramsing, A. U.; Ruzicka, J.; Hansen, E. H. Anal. Chlm. Acta 1981, 129, 1-17. (13) Wilson, R. L.; Ingle, J. D., Jr. Anal. Chem. 1977. 4 9 , 1080-1065. (14) Ryan, M. A.; Ingle, J. D., Jr. Anal. Chem. 1980, 5 2 , 2177-2184. (15) Chung, H. K.; Ingle, S. D. Anal. Chem., following paper In this issue
RECEIVEDfor review June 4,1990. Accepted September 4, 1990.
Kinetic Data from Individual Peak Profiles Corrected for Dispersion in Flow Injection Analysis Applied to the Determination of Aluminum H. K. Chung and J. D. Ingle, Jr.* Department of Chemistry, Oregon State University, Gilbert Hall 153, Corvallis, Oregon 97331-4003
A new procedure for obtalnlng analytical klnetlc lnformatlon In singlavne flow InJectlonanaiysls (FIA) system b described. A microcomputer-controlled FIA system Is flrst used to acquire and store the peak profiles of a nonreactlng reference solutlon and a reacthg sample oolutlon. Next the sample FIA response Is normallzed In the tlme domaln by multlplylngthe a n a l response by the relathre sample dlqoerslon coefflclents calculated from the reference proflle. I n the absence of a chemlcal reaction, the m a l l z e d sample proflle b flat. When the slow chemlcal reactlon Is In progress durlng the tlme between InJectlon and detectlon, the signal varles across the normallzed proflle and analytlcal klnetlc lnformatlon can be extracted under certaln condltlons. Chemlcal and Instrumental varlables were optimized for a fluorometric klnetlc daermlnatbn of AP+ wlth the fixed-time method. A 13 ng/mL de4ectlon Wmtl was achieved. The proposed method Is shown to dlscrlmlnate against the Interference from rlboflavln, a nonreactlng fluorescent specles.
INTRODUCTION Kinetic information can be obtained from flow injection analysis (FIA) measurements if the time for a reaction to reach 0003-2700/90/0362-2547$02.50/0
completion is much longer than the time between sample injection and detection (1,2).In a single-line FIA manifold with injection of the sample into a reagent stream, the peak height or area is proportional to the amount of product formed in a fixed-time period ( I , 2). This single-point kinetic method does not provide the advantages of a true kinetic method because an absolute signal rather than a change in signal is measured. A nonreacting species in the sample that contributes to the detector response will cause error. To measure a change in the detector signal due to a chemical reaction with FIA, several methods have been suggested. These include a two-point kinetic assay based on the difference in signals from two flow detector cells (31, the stopped-flow technique in which the dispersed sample zone is stopped when it reaches the detection cell for a predetermined period during which the reaction-rate measurement is performed (4-6), a two-point kinetic method based on the simultaneous injection of the same sample with two serial injection valves into one reagent carrier stream (7),and a multiple-point method based on an array of detectors along an observation tube (8). Valcarcel and co-workers (9)demonstrated that kinetic information can be obtained in a single-line manifold by injecting a relatively large sample volume (e.g., 1000 pL) into the reagent stream. Two reaction zones and two peaks are produced because little or no reagent/sample mixing occurs 0 1990 American Chemical Society
2548
ANALYTICAL CHEMISTRY, VOL. 62, NO. 23, DECEMBER 1, 1990
in the middle of the sample zone. Kinetic determinations are based on the difference in peak heights or areas. A number of researchers have described the effect of chemical kinetics on the shape of profiles in FIA. An excellent paper by Hungerford and Christian (10) reviews much of the work in this area and includes a theoretical model that successfully predicts the shape of experimental FIA profiles by considering the effects of sample dispersion, reagent dispersion, and reaction kinetics. In this paper, a new method is described for obtaining rate information from the peak profile shape in a single-line FIA manifold. A reference profile is first obtained with a nonreacting sample and used to determine the relative physical dispersion coefficients. These dispersion coefficients are then used to normalize the FIA profile when a chemical reaction is occurring. Under suitable conditions, an increase in signal with time is observed over a portion of the normalized profile, and rate information is extracted with the fixed-time method. The technique is demonstrated by adapting a kinetic procedure developed by Campi and Ingle (11)for the determination of AI3+to the FIA system. I t is based on the formation of a fluorescent complex between A13+ and acid alizarine garnet R (AAGR). EXPERIMENTAL SECTION Instrumentation. All measurements were performed with a microcomputer-controlled single-line FIA manifold and a fluorescence detection system that have been described (12). Timing and data acquisition parameters were as previously noted (12) except the integration time per data point was 0.1 s. The excitation and emission monochromators were used with a 17-nm spectral bandpass and a 21-nm spectral bandpass, respectively. Solution Preparation. All chemicals were reagent grade with the exception of the AAGR (3-[(2,4-dihydroxyphenyl)azo1-4hydroxysulfonic acid) which is supplied by Berncolors-Poughkeepsie, Inc., as a sample from large batches of the dye. The major stock solutions were 100 pg/mL quinine sulfate (QS) in 0.05 M H,S04, 100 pg/mL AI3+ in 0.01 M HN03, 500 pg/mL AAGR, 100 pg/mL riboflavin, and an acetate buffer (2 mol of acetic acid adjusted to pH 4.75 with 10 M NaOH and diluted to 1 L for most studies). The buffer solution was further cleaned with an ionexchange resin (Chelex-100) to remove the metal contamination as previously described (11). The AAGR carrier stream solutions were prepared by adding the appropriate volume of AAGR and 50 mL of the acetate buffer and diluting to 100 mL with water at the time of analysis. All water used for solution preparation and rinsing glassware was double deionized water from a Millipore Milli-Q system. Procedures. QS solutions were injected into a 0.05 M H2S04 carrier stream to evaluate the physical dispersion of a sample zone. The excitation and emission wavelengths were 366 and 460 nm, respectively. Test solutions of AP+, the blank, and the reference solutions were injected into a carrier stream of AAGR of various concentrations. The blank was 0.01 M HN03 for AI3+test solutions and 150 pg/mL AAGR in buffer for the reference solution. To prepare the reference solution (an equilibrated solution of the AP-AAGR complex), 4 mL of a 0.05 pg/mL Ala+solution was mixed with 2 mL of a reagent mixture that contained 50 pg/mL AAGR in a 1.0 M acetate buffer solution of pH 4.75 and the mixture was allowed to reach equilibrium (over 30 min). The excitation and emission wavelengths were 366 and 435 nm, respectively. Unless otherwise specified, standard and sample solutions were injected in triplicate, and thus all rate values are the average rates of three runs. Software for Data Analysis. As previously described (12), for each injection of a test solution,a data file (fluorescence signal [ S ( t )vs ] time) is acquired and stored for later data analysis. The FIA response curve is plotted on the monitor, and the program performs baseline correction to obtain the net fluorescence signal at all times in the measurement period. Various peak parameters are calculated and reported including the signal and time for the peak maximum, the peak area, and the beginning and end times of the peak profile.
Table I. Comparison of Q S Calibration Data" time interval, s
slope,
RSD,*
counts x 104/(gg/mL)
5%
8.8
2.72 2.71 2.71 2.71 2.71 2.71
0.4 0.6 0.7 0.6 0.6
12.4
2.70
0.6
0.2
2.2 3.6 4.6 7.4
0.7
RSD of slope, 70 0.25 0.30 0.35 0.36 0.34
0.38 0.44
a All values from three repetitive measurements: sample volume, 30 p L ; flow rate, 2 mL/min; integration time per data point, 0.2 s. In the normalized signal for 1 pg/mL QS.
A separate QuickBASIC program was written for further data processing of the stored files. The digitized reference profiles, which involve purely physical dispersion, are averaged and used to calculate the dispersion coefficient at time t throughout the sample zone. For most studies, the relative dispersion coefficient (D'(t) = S,,/S(t)) was calculated where S, is the peak maximum signal. The absolute dispersion coefficient ( D ( t ) )was calculated from the fluorescence signal of the undiluted reference solution. The dispersion-normalized signals ( S ' ( t ) )for the test solutions were calculated from Dit) x S(t). Before normalization for dispersion, the appropriate blank profiles were averaged and subtracted from the reference or test solution profiles for blank correction. To calculate the reaction rate, the fixed-time method is applied to the dispersion-normalized data files for the test solutions. The initial time and total computational time are specified by the operator. The sum of the data points over the first half of the computational period are subtracted from the sum of the data points over the second half of the computational period.
RESULTS AND DISCUSSION Initial Studies. To test the reproducibility of the FIA system, repetitive injections of 30 pL of 1 , 3 , 5 , and 10 wg/mL QS solutions were made into a 0.05 M H I S O carrier stream. The 10 pg/mL QS solution was selected as the reference solution from which the relative dispersion coefficients (D'(t)) were calculated. The relative standard deviation (RSD) of the signals or then relative dispersion coefficients was 1.1% or better for signal values as small as 1% of the peak maximum. The average dispersion-corrected signals for 1, 3, and 5 pg/mL QS solutions over different time intervals were calculated by summing n data points, starting at the peak maximum, and dividing by n. The value of n varied from 1 to 62 corresponding to time intervals from 0.2 to 12.4 s. Calibration data of the mean dispersion-corrected signal versus QS concentration for different time intervals were fit by a linear least-squares program. Table I summarizes the data. The calibration curves are linear with a correlation coefficient of greater than 0.99999 in all cases. The RSD in the slope of the calibration curve increases slightly as the total measurement time increases. The range of slopes is about 0.7% of the mean value. These results indicate the good long-term reproducibility of the peak profile and the accuracy of the normalization scheme. For 1 pg/mL QS, the RSD in the normalized mean signal for different integration times ranged from 0.4 to 0.770. Initially for rate measurements, reference profiles derived from injection of QS solutions into a 0.05 M H2S04carrier stream or the equilibrated AI3+-AAGR complex into a H20 or AAGR stream were tested. The nature of the reference solution or the carrier stream or the use or not of blank correction had less than a 0.5% effect on the calculated relative dispersion coefficients. For all further studies, the equilibrated solution of AI3+-AAGR mixture and reagent blank were injected into the AAGR reagent carrier stream to obtain
ANALYTICAL CHEMISTRY, VOL. 62,NO. 23,DECEMBER 1, 1990
2549
d +l
‘1
C
-8 3
b 0 . : 15
20
: 26
: 30
: 36
, .*
:
I
40
46
50
Time
Figure 1. Dependence of the shape of the peak on the injected sample volume for QS (dotted line) and AI3+ (solid line) injected into a 150 pg/mL AAGR carrier stream: injection volume, (a) 30 pL; (b) 100 wL; (c) 200 pL; (d) 500 pL; QS concentration, 1 pg/mL; AI3+ concentration, 1 pg/mL; flow rate, 2.0 mL/min. blank-corrected dispersion coefficients as a function of time. This procedure assures that viscosities of the reference, analyte, and carrier stream solutions are similar and that the relative dispersion coefficients of the reference profile apply to the analyte profiles. Effect of Sample Loop Volume. To determine the effect of sample loop volume, QS solutions were first injected into the single-line FIA manifold with a 0.05 M H2S04carrier stream to obtain nonkinetic peaks. The sample volume ranged from 30 to 500 pL. The shapes of the nonkinetic profiles, shown in Figure 1 (dashed line), depend purely on physical dispersion (or dilution). The peak height and area determined increase with increasing injected sample volume. Even though the peak area yields no information about the shape of the peak, it gives more direct information about the nature of the peak than the peak height. The linear relationship between the peak area and the injected sample volume observed indicates that there is not chemical reaction involved, as the area is proportional to the total number of moles of analyte injected. To test the effect of sample volume in the FIA system for the situation involving a chemical reaction, the same volumes of A13+ samples were injected in the reagent carrier (150 pg/mL AAGR in a 1M HACpH 4.75 buffer) without changing the FIA manifold. Figure 1 shows superimposed detector responses of nonkinetic (dashed line) and kinetic (solid line) cases. When a slow chemical reaction is still in progress a t the time of detection, the peak shape is affected both by physical dispersion and reaction kinetics and the sample loop volume affects the shape of peak much more dramatically. Consider a second-order reaction between the analyte and a reagent in a single-line FIA manifold for which the product is monitored. The amount of product formed, and hence the signal observed, at any point in the dispersed sample zone a t the detector depends on several competing factors. First, the concentration of analyte varies throughout the sample zone as given by C A o / D ( t )where , CAo is the initial analyte concentration and is greatest at the maximum of the profile for a nonkinetic sample. This factor by itself would result in more product being formed in the center of the profile than a t the edges of the profile. Second, the reagent concentration varies throughout the sample zone as described by cRo(D(t)- 1)/ D ( t ) ,where CRois the original reagent concentration in the carrier stream, and is least at the maximum of the profile for a nonreacting sample. This factor by itself would result in more product being formed away from the center of the profile. For small sample volumes yielding a large dispersion, ( D ( t ) - l ) / D ( t ) N 1 throughout the sample zone and effect of reagent dispersion is less significant. Third, the analyte a t the edges of the initial sample plug injected is in contact with the reagent longer than the analyte in the center of the plug.
Time (s)
Figure 2. Dependence of dispersion-normalized profiles on injection volume: (a) 30 pL; (b) 100 pL: (c) 200 pL; (d) 500 pL. The arrow marks the position of the maximum for the reference profile. The absolute dispersion coefficients were used for normalization (6.54,2.30, 1.37,and 1.00, respectively, at the maximum). Other conditions are the same as in Figure 1.
Thus, this factor by itself would result in more product being formed at the edges of the profile due to a longer reaction time. The shapes of the kinetic profiles shown in Figure 1 illustrate how the relative importance of these competing factors varies with the sample injection volume. For lower injected sample volumes (30 or 100 pL), the time for the peak maximum for kinetic peaks shifts slightly to longer times relative to nonkinetic peaks but the profiles are similar in shape. The dominant factor that affects the profile shape is the dispersion of the analyte. For smaller sample volumes, the absolute dispersion coefficient at the maximum is 3 or greater such that the reagent concentration (CR) throughout the sample is relatively constant (i.e., 2CR0/3 < CR < CRO). However, for larger sample volumes (200 p L or larger), dual peaks are observed, which indicates that the shape of peak is affected significantly by the dispersion of the reagent and the reaction time. With the largest sample volume tested (500 pL), no mixing occurred between the sample and carrier stream in the middle of the sample zone (i.e., a steady-state signal due to undiluted analyte is observed). Near the region of no mixing, the limiting factor for the formation of product is the lower concentration of the reagent and the shorter reaction time. The data in Figure 1 also show that the shape of the profile in the leading region of the sample zone is fairly independent of the sample volume. This occurs because the dispersion coefficients (D’(t))in the leading region are independent of the sample volume. As shown in Figure 2, the shape of the dispersion-corrected profiles changes significantly with the injected sample volume. For the 30-pL sample volume, the corrected profile is similar in shape to a normal FIA profile without a kinetic contribution. In this case, the absolute dispersion is relatively large throughout the profile such that the reagent concentration and the reaction time are similar throughout the profile. The maximum is near the center of the profile because the A13+ concentration is the highest. When the sample volume is increased, the concentration of the reagent is significantly less near the center of the sample zone that a t the edges. Also the A13+originally in the center of the sample zone had had much less time to react with the reagent. For sample volumes of 200 and 500 pL, dual peaks in the uncorrected profile and U-shaped corrected profiles (Figure 2c,d) are observed. At or near the dip of the profile, there is little contribution from the chemical reaction because of the lower reagent concentration and shorter reaction time. For the 500-pL sample volume, the fluorescence signal is effectively zero because no
2550
ANALYTICAL CHEMISTRY, VOL. 62, NO. 23, DECEMBER 1, 1990 I
25
f I 0
100
200
300
0 0
500
400
251
1
2
3
4
Flow Rate (mL/min)
Injected Sample Volume (pL)
Figure 5. Dependence of the reaction rate on the injected sample volume. The starting times for the fixed-time calculation were 18.2, 21.1, 23.1, 27.1, 32.0, and 36.1 s for 30, 60, 100, 200, 350, and 500-pL injection volumes, respectively. Other conditions are the same as In Figure 2.
Flgwe 5. Dependence of the reaction rate on the flow rate. The rate Computational period was 4.0 s staring at peak maximum time for reference profile. Other condltlons are the same as in Figure 4.
a
Time (s) 0
20
40 Time
60
a0
(6)
Figure 6. Dependence of the FIA profiles for Ai3+ injected into a AAGR carrier stream on the concentration of AAGR: (a) equilibrated
Figure 4. Effect of flow rate on the FIA profiles for AI3+ injected into a 150 Mg/mL AAGR reagent carrier stream: (a) 1.0 mL/min, (b) 2.0 mL/min, (c) 3.0 mL/mln, (d) 4.0 mL/min; injected sample, 1 pg/mL AV+; injection volume, 200 p ~ .
AP+-AAGR solution, (b) 50 pg/mL, (c) 100 pg/mL, (d) 150 pg/mL, (e) 200 pg/mL, (f) 250 pg/mL; AI3+ concentration, 1 pg/mL; injection volume, 200 pL; flow rate, 2.5 mL/min. Rate computational period was from 19.5 to 22.1 s.
product has formed in the undiluted central sample zone. Even though the ratio of reagent to sample concentrations varies throughout the sample profile due to the inherent characteristics of a single-line manifold, the fixed-time method can be applied to the dispersion-normalizedprofiles to obtain reaction-rate information which compensates for the background signal due to nonreacting species. The total measurement time was fixed (4.2 s) and the initial time for the calculation was adjusted to obtain the greatest rate for all cases. Figure 3 shows the effect of injected sample volume on the reaction rate. The RSD in the rate did not vary significantly with the sample volume. The calculated rate increases about a factor of 3 as the sample volume is increased from 60 to 500 pL. The rate for the 30-pL sample size is greater than that for 60-pL sample size. Note from the normalized profiles in Figure 2 that as the sample volume increases, a longer computational time for the fixed-time method can be used because the dispersed sample zone is larger. For further studies, a 200-pL sample volume was chosen. This sample volume yields a rate within 30% of that obtained with the largest sample volume tested (500 pL). Use of a 200-pL rather than 500-pL sample size has several advantages including lower sample consumption, higher sample throughput, and the fact that the minimum value of the normalized profile is at the maximum of the profile of the reference solution. The last factor makes it simple to identify the starting time for the fixed-time rate calculation. With a 500-rL sample loop, the central portion of the profile cannot be used because the absolute dispersion coefficients are at or near 1. Effect of Flow Rate. Figure 4 shows that increasing the flow rate from 1to 4 mL/min decreases both the peak height
and the peak area when A13+ is injected into the reagent carrier stream. The peak height decreases with increasing flow rate in contrast to that observed from nonkinetic sample, primarily due to the smaller reaction time between the sample and reagent (the residence time between sample injection and detection). However, the general shape of the profiles appears to be fairly independent of flow rate even though the rate of change of the dispersion coefficient of the sample zone with respect to time is greater at higher flow rates. Figure 5 illustrates that the calculated reaction rate increases somewhat with increasing flow rate up to about 2.5 mL/min and then varies little with flow rate. This behavior is due to the decrease in the total reaction time with increasing flow rate. At flow rates of 2.5 mL/min or greater, a small enough fraction of A13+ reacts over the 4-9 computational period such that the kinetics are pseudo zero order. A flow rate of 2.5 mL/min was selected for further studies. The RSD of the rate did not significantly vary with flow rate. Effect of t h e Reagent Concentration. With a 200-pL sample size and a 2.5 mL/min flow rate, the absolute dispersion coefficient varies from 1.5 to -, which means that the reagent concentration varies from its initial value to about 33% of the initial value at the center of the sample zone. When A13+solution is injected into carrier streams containing five different AAGR concentrations (50-250 pg/mL), the peak profiles shown in Figure 6 are obtained. As the AAGR concentration increases, the two peak maxima shift toward the center of the sample zone and the peak area increases as also shown in Figure 7. Figure 7 illustrates that the calculated rate from the dispersion normalized profiles reached a maximum value at about 150 pg/mL AAGR. This concentration was chosen for further
ANALYTICAL CHEMISTRY, VOL. 62, NO. 23, DECEMBER 1, 1990
2551
0 0
50
100 150 200 250 Conc. of AAGR (pg/mL)
300
Time (s)
Flgure 7. Dependence of the reaction rate on the concentration of
AAGR. Conditions were the same as in Figure 6.
Table 11. Calibration Data Based on the Peak Height, Peak Area, Double-Peak, and Reaction-Rate Methods slope,
RSD,"
detection
method
counts/(pg/mL)
%
limit, ng/mL
peak height peak areab double-peak rate
431 405 X lo2 150 198 X lo2
0.33 1.1 0.96 0.98
7 9 24 13
aFor 1 pg/mL A13+. bArea from peak maximum to 1% of maximum.
studies because the dependence of the rate on AAGR concentration is slight. Campi and Ingle found that the rate measured in a conventional sample cell was fairly independent of AAGR concentration above a reaction mixture AAGR concentration of about 40 pg/mL (11). With a 150 pg/mL AAGR carrier stream, the AAGR concentration at the center of the profile is about 50 pg/mL. Thus, the results observed in this study are consistent with the data reported by Campi and Ingle. Characteristics of the Analytical Method. With the optimized experimental conditions described in the previous sections, the FIA peak profiles for the A13+standards of 0.025 to 10 pg/mL were acquired and stored. The rate was calculated with computational period of 6.4 s, which corresponds to the time between the maximum and 10% of the maximum for the reference profile (equilibrated A13+ and reagent) in the same FIA manifold. Over this time interval, the sample dispersion coefficient varies from 1.4 to 14. The bmeline-width time was 14.9 s for the reference and 20.6 s for the sample. The calibration curves based on the peak height (the second peak corresponding to a longer reaction time) and the area, which are most common descriptors of flow injection peaks, are linear. This is expected if the reagent is in great excess because the relative dispersion remains constant at every point of a given sample profile, but the amount of product formed a t any point is proportional to the A13+concentration. The double-peak method of Valcarcel and co-workers (9) (Le., the difference between the heights of the two maxima) also provides a linear plot. The calibration curve based on the fixed-time reaction-rate method from the normalized profile is linear up to 7 pg/mL A13+. This suggests that the reaction kinetics are near pseudo first order in A13+. Table I1 summarizes the calibration data for the peak height, peak area, double-peak, and reaction-rate methods. The blank signal for the peak height, peak area, or double-peak methods was obtained at the same times or over the same time interval used for sample injections. The precision at 1pg/mL A13+ is best for the peak height method. The peak height would be expected to be least affected by the variance in dispersion due to between-run variations in injection time or
Flgure 8. Effect of riboflavin on the peak profile of AI3+: (a) without riboflavin, (b) with riboflavin; AI3+ concentration, 1 pg/mL; riboflavin concentration, 5 pg/mL.
flow rate. The peak area, doublepeak, and rate methods yield comparable reproducibility. The detection limits for A13+with the four methods, also shown in Table 11, are within a factor of 3 of each other. The detection limit is defined as twice the standard deviation of the blank signal divided by the slope of the calibration curve. The standard deviation of the blank signal was obtained from ten repetitive injections of the reagent blank solution into the reagent carried stream. The detection limit with the rate method or the other methods in this study is considerably worse than that reported by Campi (0.1 ng/mL). The difference is due in part to the use of a shorter effective measurement time (6.4 s vs 16 s), the use of a l mm path length flow cell in place of a 10 mm path length cuvette, and the greater average dilution of the A13+. According to these results, the proposed reaction-rate method in a single-line manifold can be used as a true kinetics-based technique. Determination of A13+with Riboflavin. To evaluate the accuracy and applicability of the proposed reaction-rate method for samples containing a fluorescent interferent, mixtures of A13+and riboflavin were tested. Riboflavin was chosen as a test interferent because it does not react with AAGR reagent mixture but fluoresces significantly with the excitation and emission conditions used for the A13+ determination. The presence of riboflavin in a sample containing A13+causes the signal at all points in the profile to be too high as determined by the dispersion of the riboflavin and its concentration as shown in Figure 8. FIA profiles were obtained in triplicate for synthetic samples containing 1or 2 pg/mL A13+ and different concentrations of riboflavin. From the profiles, the peak height, peak area, rate with the double-peak method, and rate with the proposed method were obtained and are summarized in Table 111. The proposed reaction method does discriminate against the interference from a nonreacting fluorescent species. The maximum error is 4%. The peak height, area, and doublepeak methods suffer significant interference due to riboflavin. The increase in peak area and peak height for a given concentration of A13+ is proportional to the concentration of riboflavin. The double-peak method is not a true kinetic method because it does not in general discriminate against nonreacting fluorescent species. Moreover, the presence of an interfering species causes the times for the peak maxima to shift. The double-peak method would discriminate against a nonreacting species if the peak values were normalized for dispersion before the difference was taken. Figure 9 shows the FIA normalized profiles with and without riboflavin present. Note that the shape of the normalized profile and the slope are not affected by the presence of riboflavin. The riboflavin just contributes a dc level to the profile.
2552
ANALYTICAL CHEMISTRY, VOL. 62, NO. 23, DECEMBER 1, 1990
Table 111. Determination of A13+ in Synthetic Samples Containing Riboflavin riboflavin A13+ conc,
conc, pg/mL
pg/mL
0 1 2 3
1
1 1 1 1
5 0
2 2 2 2 2
1 2
3 5
riboflavin
peak
peak area: error, height,” error, counts % counts x lo2 % 179 (0.81) 199 (1.0) 221 (1.6) 244 (1.3) 288 (5.1)
11 23 36 61
245 (4.4) 277 (2.9) 302 (3.8) 329 (1.6) 392 (3.9)
13 23 34 60
350 (3.9) 371 (4.0) 392 (1.3) 411 (1.6) 457 (2.9)
6.0 12 17 30
464 (7.6) 487 (6.8) 509 (3.3) 537 (5.2) 590 (3.8)
4.9 9.7 16 27
double peak,n error, reaction rate: error, (counts x IO2) % counts %
AI3+ conc,
conc,
pg/mL
pg/mL
1 1 1 1 1
0 1 2 3 5
72 (4.4) 77 (4.3) 6.9 81 (4.7) 13 87 (6.0) 21 94 (2.0) 31
93.0 (2.8) 90.8 (2.3) 95.1 (2.3) 93.6 (4.5) 92.3 (3.7)
2.4 2.4 0.68 0.77
2 2 2 2 2
0 1 2 3 5
136 (1.4) 142 (1.6) 4.4 146 (3.6) 7.4 149 (3.8) 9.6 165 (2.9) 21
193 (0.93) 197 (3.0) 193 (1.8) 192 (2.7) 187 (2.4)
2.1 0.27 0.86 3.5
% RSD in
g
parentheses.
o
15
20
25
30
Time (s)
Flgure 9. Normalized profiles from the data in Figure 8: (a) without riboflavin: (b) with riboflavin.
CONCLUSIONS The signal in dispersion-normalized FIA profiles varies across the profile if the reaction between the analyte and reagents is relatively slow. It is demonstrated that analytical kinetic information can be extracted from the normalized profile. The kinetic method discriminates against nonreacting interferents that yield a detedor signal because the normalized signal due to these interferences does not vary with time. As the stopped-flow method, the proposed kinetic measurement technique can be used with a simple single-line FIA systems without additional flow components or detectors. Thus, it is simpler than FIA kinetics methods based on dual-injection values or multiple detectors. The sample throughput rate is greater than that for the stopped-flow technique. However, the stopped-flow method does provide some advantages compared to the proposed technique or dual-point FIA kinetic methods. These include no reagent
consumption during the stopped-flow period, direct observation of the reaction rate for a given analytelreagent concentration ratio, and rate measurements that are independent of the flow rate and sample volume. The dispersion-normalization kinetic technique does require microcomputer control and data acquisition and precise flow rates, which is becoming more commonplace. Although the technique was demonstrated with fluorometric detection and monitoring of a reaction product, it is applicable to other detection techniques and monitoring the disappearance of reactants. In the latter case, the normalized profile would be inverted relative to those presented (i.e., the normalized signal would decrease from the center to the edge of the profile). To use the new kinetic measurement technique, several conditions must be met. First, the reaction kinetics and the reagent concentrations must be adjusted such that it takes about a minute to tens of minutes for the reaction to reach completion in a normal sample cell. Second, the amount of product formed a t any point in the sample profile across the time interval used to extract kinetic information must be proportional to the analyte concentration. This condition requires that the reaction be first order in the analyte concentration and that the reagent concentrations be in sufficient excess relative to the maximum analyte concentration determined such that an insignificant amount of the reagents is consumed (effectively pseudo first order in the analyte at any point in the profile). Note that the reagent concentration necessarily does vary across the profile such that the normalized signal is not linear with time unless the reaction is zero order with respect to the reagent concentrations over the computational period. Third, the carrier stream flow rate, the sample injection volume, and reaction coil length must be adjusted so that the criteria for condition two above are met and that a detectable change in product concentration occurs over the rate computational period (typically 5-15 s). In general, relatively large sample loop volumes (200 p L or greater) must be used so that the dispersion at peak maximum of the reference solution is low (D, = 1-2) and the peak width is large enough (20-40 s) to obtain a sufficient reaction time. Under these conditions, the reaction time (the time the analyte and reagents are in contact before detection) increases from the center of the profile to the trailing edge as a first approximation, and the minimum signal in the normalized profile is a convenient point for the beginning of the rate computational period. LITERATURE CITED Ruzicka, J.; Hansen, E. H. Flow Injection Analysis, 2nd ed.; Wiley: New York, 1988. Valcarcel, M.; Luque de Castro, M. D. Flow Iniection Analysis: Principles and Applications; Ellis Horwood: Chichester, U.K., 1987. Hansen, E. H.; Ruzicka, J.; Rietz, E. Anal. Chlm. Acta 1977, 89, 241-254. Ruzicka, J.; Hansen, E. H. Anal. Chim. Acta 1979, 706, 207-224. Ruzicka, J.: Hansen, E. H. Anal. Chim. Acta 1980, 774, 19-44. Kagenow, H.; Jensen, A. Anal. Chim. Acta 1983, 745, 125-133. Fernandez, A.; Luque de Castro, M. D.; Valcarcel, M. Analyst 1987, 172, 803-807. Hooley, D. J.; Dessy, R. E. Anal. Chem. 1983, 55, 313-320. Fernandez, A.; Luque de Castro, M. D.; Valcarcel, M. Anal. Chim. Acta 1987. 793, 107-1 18. Hungerford, J. M.; Christian, G. D. Anal. C h h . Acta 1987, 200, 1-19. Campi, G.L.; Ingle, J. D.. Jr. Anal. Chim. Acta 1989, 224, 363-372. Chung, H. K.; Ingle, J. D., Jr. Detection and Correction of Multipllcatlve Interferences in Singleline Flow Injection Analysis whh Fluorescence Detection Anal. Chem ., preceding paper in this issue.
RECEIVEDfor review June 4, 1990. Accepted September 4, 1990.
