Enhancement of flow injection optosensing by sorbent extraction and

Jul 15, 1990 - Lacy, Gary D. Christian, and Jaromir. ... Miró , Matias Manera , José Manuel Estela , Vı́ctor Cerdà , M.Renata S Souto , António ...
0 downloads 0 Views 2MB Size
Anal. Chem. 1990, 62, 1482-1490

1482

Enhancement of Flow Injection Optosensing by Sorbent Extraction and Reaction Rate Measurement Nathan Lacy, Gary D. Christian,* and Jaromir Ruzicka* Department of Chemistry, BG-IO, University of Washington, Seattle, Washington 98195

The molybdenum blue reactlon for the determination of phosphate Is used as a model to Illustrate the extension of the use of hydrophoblc sorbents In flow Injection analyds for the preconcentratlon of an anlon and for on-cdumn detec#on, Le., optosenslng. Optosenslng provides for real-time monltorlng of the rate of slgnai change, dA /d T , so that the rate of color devebpment durlng the reductlon step of the analysis can be measured. The synerglstlc relationship between the rates of formation of the phosphate and sHlcate heteropdy complexes is examlned. A kinetic optosenslng method has been developed In which the difference In reduction rate for the heteropoly complexes allows simultaneous determlnatlon of phosphate, In the parts per bllllon range, and slllcate, in the parts per mllllon range. The system used an Inexpensive mlnlspec-20 spectrophotometer wtth a net path length through the sorbent of less than 2 mm. Partlal least-squares analyds was used to analyze the data, and predlctlon errors of approxlmately 10 % were obtained for both components.

INTRODUCTION Enhancement of the selectivity of instrumental methods by flow injection analysis (FIA) has been accomplished by various means, such as the use of solvent extraction ( 1 ) and gas diffusion techniques (2). These may be considered homogeneous techniques in that the chemistry occurs in the homogeneous media of the flowing stream itself. Additionally, sensitivity has been increased by on-line preconcentration using ion exchange ( 3 ) . The latter technique may be viewed as heterogeneous since isolation of the analyte on a surface has occurred, creating a region of increased concentration within the flowing stream and may be considered similar to solute focusing (4). Only very recently, attention is being given to the concept of sorbent extraction in FIA, which has been used so far only for the preconcentration of metal complexes (5) using hydrophobic interactions on C-18. Further increases in sensitivity can be achieved by flow injection optosensing (6),which is based on the measurement of an analyte retained on the solid sorbent surface. Use of cation exchangers for this purpose has been described by Yoshimura and Waki (7) while Valcarcel and co-workers (8) used an anion exchanger in a detailed study of the iron(II1) thiocyanate complex. This paper deals, for the first time, with the use of sorbent extraction for the optosensing of anionic complexes and exploits in a novel way the differential kinetics of competing species. The present study of differential kinetics allows for the simultaneous determination of two analytes or the determination of one analyte in the presence of an interferent. Thus, while to date the use of sorbents has been driven by the "all or nothing" philosophy that focuses on steady-state chemistry, it is shown in this work how the horizon of applications for sorbent use in FIA can be greatly broadened with the incorporation of kinetic techniques. Previous work in this laboratory has focused on the enhancement of instrumental response at room temperature for the determination of phosphate by the molydenum blue reaction: 0003-2700/90/0362-1482$02.50/0

-

H3P04+ ~ M O ~ ' , , ~

12-molybdophosphoric acid

+ 9H'

(1)

utilizing antimony to enhance the reaction rate (9,lO). For further specifics on the molybdenum blue reaction the reader is referred to the above reference and references therein. While significant enhancement of the homogeneous FIA detection was achieved, further enhancement was desired. Most FIA techniques for phosphate have centered around the formation of the colored complex using homogeneous chemistry (11-17). A few FIA methodologies for phosphate have used an anion exchange resin, but this requires elution of the anion with postelution chemistry being performed in order to generate a detectable species (18). Silicate interferes significantly with the molybdenum blue determination of phosphate and must be eliminated, or accounted for, in order to correctly determine the amount of phosphate present in the sample (19, 20). In the past this has been done by examining the differences in the rates of formation of the heteropoly complexes of phosphate and silicate, but the difference in the rate of reduction of the heteropoly complexes to the blue complex was not used due to the problems of longer reaction times, deviations from Beer's law, unstable reagents and products, and high blank absorbances (21). Our goal was to achieve not only preconcentration of the heteropoly complex but also the generation of a detectable species, the blue complex, on the sorbent material itself through on-column chemistry while monitoring this chemistry through optosensing. In this way, the removal of the silicate interference and/or the simultaneous determination of silicate along with phosphate would be achieved, thereby showing the utility of this methodology, its relevance to competitive chemical reactions, and the FIA advantage of eliminating problems associated with unstable reagents and products.

EXPERIMENTAL SECTION Two approaches were used for detection of the signal. These consisted of (1)reversed flow elution (RFE)of the analyte previously adsorbed onto a microcolumn wherein the absorbance (peak height and peak area) of the eluted analyte was measured, and (2) optosensing of the change in absorbance due to the adsorption of the analyte onto silica based C-18 placed into a specially designed flow cell. Apparatus for Reversed Flow Elution. Figure 1 illustrates the manifold used for the reversed flow elution work. All Tygon pump tubes were of identical 0.64-mm i.d. The reaction coil consisted of a knotted 2-m length of 0.5-mm Tygon tubing. Alitea-XV peristaltic pumps were used, which in conjunction with the software used (22) provided for versatile flow control. The pump speed was adjusted to ensure delivery rates of 1.5 mL/min for all channels. The microcolumn (Figure 2) consisted of a plexiglass block with a 1-mm-i.d. hole bored into it and having a length of 10 mm. The microcolumn was packed with the hydrophobic sorbent C-18 (silica based and having a particle size distribution of 15-100 pm) as supplied by Analytichem International in their BondElut cartridges. A Valco 10-port electrically actuated valve was used to switch the microcolumn from the reagent/sample stream to the eluent stream and, as can be seen from the manifold, eluent flow was the reverse of the reagent/ sample flow. Elution of the analyte was performed by using reversed flow to minimize band broadening and the associated 0 1990 American Chemical Society

ANALYTICAL CHEMISTRY. VOL. 62. NO. 14. JULY 15, 1990

1483

MIMSPEC-20

nmkvoL

n U

LOADING POSITION B Flgure 3. Flow injection manifold used lor optosensing.

ELUTION POSITION m e 1. Flow h w n m a n W . usd IMreversed fbw elmn. stmwn in the loading position (A) and elution position (6).

A

F~UIO 2.

Minicolumn used IMreversed flow elution

decrease in precision and sensitivitythat occurs when the analyte passes through and interacts with the entire length of the microcolumn. A FIAstar 5023 spectrophotometer was used that utilized a flow-through cell having a volume of 18 fiL and a path length of 10 mm. Absorbance was measured at 660 nm. The spectrophotometer was interfaced with a FIAstar 5032 controller, the output of which was processed through an amplifier to ensure compatibility with an IBM DACA board that interfaced an ATcompatible PC with the pumps and spectrophotometer. Flow program control was achieved by using the above mentioned AT-compatible PC via a FIA program developed by Clark et al. at the University of Washington (22). Optosensing Apparatus. Figure 3 illustrates the optosensing apparatus used. The primary difference between this manifold and that used in the RFE work was that the solid sorbent was placed within the spectrophotometer's light path and that elution of the analyte occurred in the same direction as the reagent/ sample stream flow. The spectrophotometer was a Milton Roy Co. Minispectronic-20. The sample cell compartment was modified to accomcdate the packed flow through cell shown in Figure 4. Due to high absorbance and light scattering hy the solid sorbent, the original light source had to be replaced by an intense Intralux 4000 fiberoptic variable light source, which provided sufficientlight to pass on to the detector. Flow programming was

Figure 4. Microcolumn block (A) and microcolumn component ar-

rangement (6) used lor optosensing. controlled by the same apparatus as described above for the RFE work. Reagents. HPLC grade methanol was used as the column eluent. All other reagents used were ACS reagent grade, and all solutions were prepared by using deionized degassed water and stored in polyethylene containers to prevent silicate contamination. A 0.0025 M ammonium heptamolybdate solution was prepared that was also 0.2 M in "0,. L-Ascorhic acid solutions of 2%,

1484

ANALYTICAL CHEMISTRY, VOL. 62, NO. 14, JULY 15, 1990

0.3 PPM

0.3

IO0

I15

130

145

160

175

TIME (SEC) Figure 5. FIA raw data profile for reversed flow elution: 1 mL of

sample was preconcentrated so that concentrations, in parts per million, conespond to column loadings, in micrograms. Note the shifting time of the peak maxima. 0.5%, 0.4%, 0.2%, 0.1%, and 0.05% (w/v), which were also 100 ppm in antimony (as Sb(II1)) prepared from a 400 ppm stock potassium antimonyl tartrate solution, were prepared fresh daily to eliminate problems of air oxidation. Unlike our prior studies, the ascorbic acid solution did not contain glycerine because its use hindered adsorption of the analyte onto the C-18. A 100 ppm phosphate (as P) solution was prepared by using KH2P04,and a 1000 ppm silica (as Si) standard was prepared from sodium silicate. Standards for analysis were prepared by appropriate dilutions of the phosphate and silicate standards. On-ColumnRedox Duration Parameters. Column loading and the on-column reduction of the heteropoly complex were achieved by timed sequencing of both the 10-port valve and the peristaltic pumps. For the RFE work (Figure 1)the system was originally in the elution position with all pumps turned on. Upon initiation of the flow control program, the 10-port valve was switched to the load position and the ascorbic acid pump was stopped. This was maintained until the proper amount of sample, typically 1mL, had been passed through the sorbent bed. Upon completion of this loading phase, the heptamolybdate/sample pump was stopped and the ascorbic acid pump was turned on for a period of time just sufficient to fill the microcolumn with the ascorbic acid reducing solution. At this time the ascorbic acid pump was shut off to prevent migration of the blue complex, which is less strongly adsorbed than the heteropoly complex, down the column bed, which occurs if ascorbic acid flow is maintained. The ascorbic acid pump was then turned on for 1 s every 15 s to get fresh reductant into the column. Three different total reduction times were used, which we termed as short, medium, and long. The short reduction time was two 15-s cycles, the medium was four 15-scycles, and the long was six 15-9 cycles for total reduction times of 30,60,and 90 s, respectively. At the completion of these reduction cycles the 10-port valve was returned to the elution position and the analyte was eluted and passed through the spectrophotometer for measurement. For greater sample loading volumes, the heptamolybdate/sample pump was allowed to run for longer periods of time prior to initiating the reduction cycles. For our particular system, a sampling time of 40 s corresponded to a sample volume of 1 mL. Accordingly, sampling for 80 s provided a 2-mL sample, and so forth. For the optosensing measurements, detector loading and adsorbed analyte reduction were achieved as described above, with the exception that only one reduction sequence of 90 s was used. That is, the ascorbic acid was pumped into the packed detector, the pump was stopped, and the reduction was allowed to proceed for 90 s under continuous monitoring of the absorbance.

RESULTS AND DISCUSSION Reversed Flow Elution. Figure 5 shows a series of overlying FIA profiles for 0.0, 0.1, 0.2,0.3, and 0.4 ppm phosphate standards. The blank profile shows a definite refractive index effect as can be seen from the negative deflection followed by the positive deflection from base line. This corresponds to measurement of the aqueous plug, present in the packed detector at the time when the valve switches to the elution

100

115

130

146

160

175

TIME (SEC) Figure 6. Blank subtracted FIA profiles for reversed flow elution. Note

that all peak maxima occur at the same time. position, on the background of the methanol eluent stream. As the amount of preconcentrated analyte increases, the time a t which the peak maximum occurs decreases. This is the result of the analyte peak being superimposed over the blank profile, with the analyte peak occurring between the negative and positive deflections of the aqueous blank. Since the sample volume was 1 mL, the parts per million concentration of the sample was equivalent to the total amount of phosphate loaded onto the detector in micrograms. Because sampling times could vary, absorbances were plotted against the detector loadings; a plot of peak height versus detector loading for the peaks shown in Figure 5 resulted in a nonlinear plot with a positive deviation. Because of the blank profile in the region where the analyte peak maxima occur, this upward trend was most likely the result of this underlying refractive index effect. With use of the program MATLAB,the blank FIA profile was subtracted from the FIA profiles of the samples. The resulting ”blank subtracted” FIA profiles are shown in Figure 6. All peak maximum times for these “blank subtracted” FIA profiles were identical, and a plot of peak height versus column loading was linear, having a correlation coefficient of 0.999. Since data processing with a program such as MATLAB can be time consuming, a simpler method of quantitation was sought. The refractive index effect is the same for each sample, so a fixed time evaluation of peak height (at the analyte peak time) was evaluated. A liner relationship having a correlation coefficient of 0.996 was obtained. However, if the fixed time chosen for the peak height evaluation does not correspond with the exact time of the peak maximum found when the blank profile has been subtracted, then a slight reduction in sensitivity will occur. The magnitude of this sensitivity loss will depend on the relative value of the peak height at the time chosen as compared to the value of the true peak maximum. If a slight loss in sensitivity is not critical, then selection of a time in the region of the peak maximum for a sample might be adequate. If, however, maximum sensitivity is required then blank subtraction with a program such as MATLAB should be performed to ascertain the exact time at which a fixed-time evaluation of peak height should be performed. An equally suitable method for quantitation is the measurement of peak area. A plot of peak area versus detector loading was also h e a r , having a correlation coefficient of 0.999. While both peak area and fixed-time measurements are easy to perform, relative precision data (Table I) showed a slight advantage for peak area over fixed-time peak height measurements a t low detector loadings. This advantage results from the summing of values over the entire peak as opposed to relying on only one data point on the FIA profile. However, peak area measurements have the inherent disadvantage of requiring a more elaborate data handling system than that

ANALYTICAL CHEMISTRY, VOL. 62, NO. 14, JULY 15, 1990 Table I. Percent Relative Standard Deviation (70rsd) Data for Column Loadings of 0.1-0.4 pg Phosphate (as P) col loading, pg P

peak area rsd, %

peak height rsd, %

0.1

1.63

0.2

1.11

0.3 0.4

1.33 1.09

3.68 1.24 1.29

‘I

1485

1

1.00

AL

I

n

L 0

10

a,

PHOSPHATE m

30

40

PPM SI

Flgure 8. Linear calibration curves for phosphate at varying silicate concentrations. The calibration curves are nearly parallel but do show slightly greater separation at the higher silicate concentrations. M

L S

BLANK PHOSPHATE

n 0

10

D

30

4l

PPM Si Flgure 7. Reversed flow elution curves for 0, 0.2, and 0.4 ppm phosphate samples with varying silicate concentrations. Note the increasing absorbance differences In going from short to long reduction times at 40 ppm silicate (as Si). required for peak height measurements. Silicate is a known interferent in the molybdenum blue reaction for phosphate since the silicate anion also forms a heteropoly complex with heptamolybdate (silicomolybdate) (I9,20). It is well documented that the rate of formation of the silicomolybdate is much slower than the rate of formation of phosphomolybdate, as is also the rate of its reduction to the blue complex (21). For the homogeneous chemistry, once the heteropoly complexes of silicate and phosphate have been formed, selective destruction of the phosphate complex can be achieved through the addition of organic acids such as tartaric and oxalic acids (19,23). This would allow separate measurement of, and correction for, the silicate. However, the length of reaction coils required and the analysis times needed to achieve these chemistries prior to adsorption onto the C-18 detracted from their desirability. Attempts at their use in the preconcentration system as designed (i.e., after adsorption of the analyte) yielded unsatisfactory results since once the heteropoly complexes were adsorbed onto the C-18, selective destruction could not be achieved through the use of these different organic acids. Because the silicomolybdate is coadsorbed onto the (2-18 and is also reduced on-column by ascorbic acid, it creates an interference for the determination of phosphate identical with that present in their homogeneous chemistries. Additionally, the rate of reduction for silicomolybdate is similarly slower than that of the phosphomolybdate, and Figure 7 illustrates data taken for a series of samples wherein the concentrations of both silicate and phosphate were varied. There is a nonlinear variation of the response curves at 0.2 and 0.4 ppm P for medium and long reduction times and high Si concentrations, but linear response curves are obtained when Si and P are measured individually. Accordingly, assuming that each species was acting independently, one would expect to see in Figure 7 a series of three parallel lines (corresponding to the short, medium, and long reduction times) at each phosphate concentration as the Si ranges from 0 to 40 ppm rather than the curves shown in Figure 7 . The color development reaction for the phosphate is complete in about 30 s, and as can be seen for the 0.2 and 0.4 ppm P standards at 0 ppm Si, there is actually some loss of signal

that occurs at longer reduction times and results from the minor amount of leaching of the blue complex from the microcolumn associated with the intermittent pumping of the ascorbic acid. An interesting feature of Figure 7 is the common point of intersection at 20 ppm Si for the 0.2 and 0.4 ppm phosphate samples. This intersection may be viewed as a “break-even’’point between the loss of signal due to leaching and the increase in signal due to the Si contribution. However, color development for the Si is not complete; that is, the reaction has not reached steady-state conditions. Yet, a t 40 ppm Si, we are seeing an increase in the signal in going from short to long reduction times when the concentration of phosphate is increased. Since the Si concentration is being held constant at 40 ppm and all samples are experiencing the same reduction conditions, we would expect the absorbance change from short to medium to long reduction times, at 0.0, 0.2, and 0.4 ppm phosphate, to be the same even though we are not at steady-state conditions. The enhancement of the Si signal observed as the concentration of phosphate is increased indicates that the phosphate is influencing the rate at which the silicomolybdate is either forming or being reduced. Crouch et al. described a synergistic effect between Si and P with regard to the rate at which their heteropoly complexes form (21). Our present data support and are appropriately explained by these earlier observations. Inverse least squares (ILS) analysis was used in an attempt to determine the concentrations of both the Si and P, and prediction errors of 10% for phosphate and over 50% for Si were obtained. This large error for Si results from trying to predict a large value of Si from a small signal. The error for phosphate is tolerable. Examination of the short reduction time curves, shown separately in Figure 8, showed that the system responded in a close to linear fashion to variations in the concentrations of both Si and P for the range examined. The near linear Si influence is seen as the P/Si calibration curves are nearly parallel, evenly spaced, and linear with respect to both Si and P. At 40 ppm Si, a 15% error for the 0.4 ppm P standard and a 21% error for the 0.2 ppm P standard were observed due to the uncorrected contribution of the Si. The relative errors are not exactly in proportion to the P concentration since the small absolute Si contribution does vary. While these errors for phosphate were greater than those observed by using ILS, no data processing was required. Accordingly, unless the data can be processed by ILS or the Si influence is constant, the present system should be Si free for determining phosphate and is unsuitable for the determination of Si. The reversed flow elution data presented here have shown that the measurements obtained are an indication of the condition of the chemical system at the time it reaches the

1486

ANALYTICAL CHEMISTRY, VOL. 62, NO. 14, JULY 15, 1990 0.6

j

E iELUTE L = LOAD R iREDUCE

I

1 U MIN

03 0

2

4

6

8

10

11

SAMPLING VOLUME (ML) Figure 9. Logarithmic decrease in response with increasing sample size. The total amount of phosphate in all samples was held constant at 100 na“ .(as PI.

E

L

R

E

L

R

E

L

R

E

I

sorbent. As the concentration of phosphate increases, the amount of the silicomolybdate complex formed is increased, thereby resulting in increase(-.absorbances over what would normally be expected (a normal linear dependence). These differences in absorbance can be correlated to the varying concentration ratios of P and Si and a measure of the synergistic kinetics occurring within the system can be attained. However, quantification studies of these synergistic effects were beyond the scope of the present research. Prior measurements of these kind have all been done “on the fly” in FIA, but for a chemical system such as the Si/P system werein homogeneous chemistry might not provide an adequate signal at the low concentrations being used (for phosphate), sorbents can be used to determine interrelated kinetic effects via preconcentration. The effect of increased sample volumes on the observed signal is shown in Figure 9. A logarithmic decrease in signal with volume is noted (correlated coefficient of 0.977). This may result from incomplete retention of the phosphomolybdate complex by the C-18 column due to the effect of partitioning of the analyte between the aqueous phase and the sorbent. The retention of an analyte by a sorbent is dependent upon the elutropic strength of the sample stream/carrier, which is determined by tne natures of both the analyte and sorbent. For an ion-exchange sorbent and an ionic analyte, water would have the greater elutropic strength, that is, it would be the stronger chromatographic mobile phase. However, for a hydrophobic sorbent and a neutral analyte, methanol would have the greater elutropic strength. Since the phosphate complex, 12-molybdophosphoric acid, can exist in both the ionic and neutral forms, a competition between hydrophobic interactions of the neutral species with C-18 and solvation of the ionic form by water probably exists. As a result, the water sample stream may carry some of the phosphate complex through the column, resulting in a loss of signal. Additionally, even though the pH of the sample stream is approximately 1 and the C-18 material is known to cleave a t these pH values, the fact that the analyte is retained from the water carrier and eluted with methanol lends credence to a hydrophobic type interaction between the analyte and the C-18 column as opposed to retention on silanol sites resulting from C-18 cleavage. Experiments were conducted wherein attempts were made to preconcentrate the phosphomolybdate complex on a SI sorbent that has unbonded, activated silica as the functional group. No preconcentration of the analyte was achieved, again indicating that the hydrophobic interaction of the (2-18 with the 12-molybdophosphoric acid was the primary factor in achieving preconcentration. However, because C-18 cleavage can occur in an acidic environment, thereby deactivating the column, the column was exposed to the acidic sample envi-

ronment only during the preconcentration step to minimize any degradation of the column that might occur. As a result, no changes in the characteristics of the column were noted. A recent article by Sentell and Dorsey examined the retention mechanism present in reversed phase liquid chromatography (RPLC) columns (24) or, as we call them, sorbents. Their discussion focused on the two main theories of retention in RPLC, partitioning and adsorption. They stated that while in partitioning the solute is almost fully embedded within the sorbent, adsorption primarily involves surface contact of the solute with the sorbent and the solute is therefore not fully embedded. Retention, they stated, can result from either of these processes or a combination of both, and their work showed that partitioning was the primary mechanism involved. We are presently conducting studies to determine if this partitioning phenomenon is the cause of the logarithmic decrease in detector signal as sample volume increases. However, regardless of the source of the effect, it is reproducible and well defined so that for extended preconcentration times calibration curves for those specific analytical conditions may be generated to account for the reduced signal that is observed. Optosensing. Figure 10 shows a typical optosensing FIA profiie for the reduction of the phosphomolybdate. When the system is in the elute (E) position, a base line is established that drops by approximately 0.08 absorbance units when the valve is switched to the load (L) position. The exact reason for this baseline shift is unknown but is believed to be the result of differences in the amount of light scattering by the C-l8/methanol and C-l8/aqueous systems. After the analyte has been preconcentrated onto the C-18, ascorbic acid is then pumped across the sorbent bed and the flow is stopped. Reduction of the adsorbed phosphomolybdate occurs and a signal is generated, AA, whose value depends on the quantity of phosphate present in the sample. After the reduction step is complete, the valve is returned to the elution position, the analyte is eluted to waste, and the packed detector is ready for the next sample. The time required for the reduction step to reach completion for these detector loadings, in the nanogram range, is on the order of 20 s. Previous work showed that in homogeneous FIA the time required is approximately 5 min with antimony enhancement. Accordingly, for the phosphomolybdate complex, utilization of a sorbent accomplishes not only preconcentration but an enhancement of the reaction rate as well. It should be noted at this point that some base-line drift toward higher absorbance values was observed and can be seen in Figure 10. This was due to the collection of a precipitate at the front of the packed detector and was attributal to the antimony present in the ascorbic acid reducing reagent.

ANALYTICAL CHEMISTRY, VOL. 62, NO. 14, JULY 15, 1990 r-1

0.05

I

1487 1

W

!w l u

8

O.*: 0.03

D

z

0.01-

o.w+ 0

'

'

2

'

4

'

6

8

'

8

'

' 1 0 1 2 1 4 '

'

'

COLUMN LOADING (NG P)

Flgure 11. Column loading necessary to obtain an analytical signal. The consistent response at 0,2, and 4 ppb is due to adsorbed anaiyte on the head of the microcolumn, which is not within the light path of the spectrophotometer.

PRECONCENIRATION WrlIrREwFtmD FLOW ELUTION

(10MM PA'IELENG'I'H)

a HOMOGENEOUS

CHEMLSTRY (10MM PA'IELENGTW 0

100

200

300

400

500

600

PPB/NG PHOSPHATE! (AS P)

Figure 12. Comparison of the homogeneous, reversed flow elution, and optosensing methodologies. Optosensing provided a net enhancement of over 200-fold.

Because we used the change in absorbance, AA, for quantitation, this base-line shift presented no problems in that, for a given column loading, a consistent AA was obtained regardless of the base-line value. Complementary preconcentration work using sorption of pH indicators showed that reproducible base lines could be established in a system in which no residue/precipitate is left in the detector. Due to the design of the optosensing cell, a small amount of the phosphomolybdate collects at the front of the packed detector but does not enter the light path. This results in the recorded signal being the same for concentrations ranging from 0 to 4 ppb when 1 mL is preconcentrated (i.e., 0-4-ng column loading), Figure 11. The value of this smallest measurable concentration may be lowered by simply employing longer loading times. For the range 6-200-ng column loading, Figure 12, a linear relationship is observed that has a correlation coefficient of 0.999. Above 200 ng the c w e becomes nonlinear but is reproducible and well defined. Figure 12 compares the signals obtained by using optosensing with those obtained with the RFE and homogeneous chemistry techniques. It shows not only that significant signal enhancement can be gained through the use of sorbents in FIA but also that even greater signal enhancement is attained when sorbents are coupled with optosensing. For the molybdenum blue reaction, used as a model in this research, we observed an approximately 40-fold increase in signal with a path length of less than one-fifth of that used for the homogeneous FIA chemistry; i.e., a net enhancement of over 200-fold. Figure 13 shows a plot of absorbance versus time for various concentrations of ascorbic acid (25 ng of phosphate loaded onto the column in all cases) and provides a real-time plot of the kinetics of the reduction reaction as it proceeds. An

0

6

10

16

a0

TIME (SEC) Flgure 13. Reaction profiles for reductant solutions of ascorbic acid at 0.4% (A), 0.2% (B), 0.1 % (C), and 0.05% (D). The ascorbic acid solutions were also 100 ppm in Sb(II1). Column loadings of 25 ng of phosphate (as P) were used in all cases.

increase in dA JdT with increasing ascorbic acid concentration is observed, and the plot is seen to be "S" shaped. If the slopes in the linear portions of the curves for 0.1, 0.2, and 0.4% ascorbic acid are compared, a first-order relationship is seen for the ascorbic acid's effect on the reaction rate. That is, a plot of the normalized slope versus normalized ascorbic acid concentration yields a linear plot having a correlation coefficient of 0.994 and a slope of 1.02. The slope of the 0.05% ascorbic acid reagent deviates from the relationship described above, being lower than expected, and is believed to be due to the depletion of the ascorbic acid within the packed detector because as the reduction reaction proceeds all of the ascorbic acid is used up prior to reducing the total amount of phosphomolybdate that has been adsorbed.

Rate Measurement for Mixtures Using Partial Least Squares Analysis. Optosensing provides us with the ability to monitor a reaction as it proceeds and obtain kinetic information about a chemical system. For a two-component system wherein similar chemistries proceed at different rates, such as in the molybdenum blue reactions for phosphate and silicate, differential kinetics can be used to establish the concentrations of both components. As was mentioned in the Introduction, the approach of differential kinetics was used for the homogeneous chemistry simultaneous determination of phosphate and silicate with respect to their rates of formation of the yellow heteropoly complexes. However, the differences in the rate of reduction of the heteropoly complex were not used due to a number of factors that included unstable reagents and products (21). These factors could be overcome through the highly reproducible methodologies of FIA and are shown here for the simultaneous determination of phosphate and silicate as based on the differences in reduction rates of the heteropoly complexes, monitored by optosensing. Figure 14, parts A and B, shows reduction profiles for phosphate and silicate standards, respectively, which range from 0 to 40 ppb phosphate (as P) and 0 to 40 ppm silicate (as Si), respectively. The phosphate reduction reaction is approximately 95% complete after only 6 s, while the silicate reduction reaction is still proceeding at 16 s, illustrating the difference that exists in the reduction rates of the two complexes. Figure 15 shows a series of reduction profiles for phosphate/silicate mixtures ranging from 0 to 40 ppb phosphate (as P) and 0 to 40 ppm silicate (as Si) in increments of 10 ppb and 10 ppm, respectively. Table I1 provides a listing of the samples and their phosphate and silicate concentrations. Significant differences exist between the different reduction profiles, providing an information-rich system that should be amenable to analysis using partial least-squares (PLS) methods. The reduction profiles of the 16 standards shown in Figure 15 were analyzed along with 8 additional sample profiles by

1488

a

ANALYTICAL CHEMISTRY, VOL. 62, NO. 14, JULY 15, 1990

c -I;

i

2

6

1

IO

8

14

I:!

REDUCTION TIME k

.-

s

5

9

i

I1

13

13

I

REDUCTION REACTION TIME (SEC)

Figure 16. Relative intensity of the contribution of each time slice to the first source of variance (A), second source of variance (B), and third source of variance (C).

2

4

8

6

10

12

14

REDUCTION TIME Figure 14. A: reduction profiles for 0 (A), 10 (B), 20 (C), and 30 ng of P (D). 9: reduction profiles for 0 (A), 10 (B), 20 (C), and 30 Mg of

Si (D).

--,

,

0

P

I

6

8

I,)

12

I4

REDUCTION REACTION TIME (SEC)

7

Figure 17. Reduction profiles for a 30 ppb phosphate (as P) standard in the presence of 0 (A), 10 (B),20 (C), and 30 ppm Si (D). Note the suppression of signal in the reduction reaction time range 1-3 s.

I-

0

+ 2

6

4

8

10

12

14

REDUCTION TIME Figure 15. Reduction profiles for the series of samples listed in Table 11.

Table 11. Concentrations of Mixtures of Phosphate (as P) and Silicate (as Si) by Sample Number, Measured in Figure 15 sample no.

pph P

ppm Si

1 2 3

0 0 0 0

0 10 20 30 0 10 20 30

4

5 6

8

8

10

10 10 10

sample no.

ppb P

ppm Si

9

20 20 20 20 30 30 30 30

0 10 20 30 0 10 20 30

10

11 12 13 14 15 16

cross validation using PLS. Prediction errors of 9.3% and 11.5% were obtained for phosphate and silicate, respectively, and are excellent considering the 1000-fold difference in the concentrations of the two components. Three latent variables were used in modeling the phosphate, while five were used in modeling the silicate. For a two-component system that is linear, only two latent variables would need to be used. However, as is discussed below, chemical interactions between the two components produce a system that is nonlinear and

requires the inclusion of more latent variables to model the subtle interactions that occur. Figure 16 shows plots of the three major sources of variance, as found from a principal-component decomposition (25),in the phosphate/silicate system, which account for 99.99% of the total variance. These figures can best be discussed by comparing them to Figures 14,15, and 17. Figure 17 is a series of reduction curves wherein the phosphate concentration was held constant while the silicate concentrtion was varied. Curve A in Figure 16 ilustrates the first source of variance (99.75% of the total) and, as might be expected, highly resembles the common reduction profile of increasing absorbance with time as the reduction reaction proceeds. Accordingly, almost all information obtained for this system simply results for the straightforward reduction of the heteropoly complexes. Curve B shows the second source of variance (0.19%),and illustrates an interaction effect occurring a t approximately 3 s reaction time. Comparing this to the reduction profiles in Figure 17 we can see that this results from a suppression of the phosphate reaction by silicate (lowered absorbance) at the early reduction times. That is, rather than the series of reduction profiles following a gradual upward trend as the silicate concentration ranges from 0 to 40 ppm, an initial decrease in absorbance occurs. This can be seen most clearly in the way the profile of the 30 ppm phosphate/lO ppm silicate standard (curve B) dips below the 30 ppm phosphate/O ppm silicate standard (curve A) from about 2.5 to 3.5 s. This suppression effect was seen in all phosphate standards when silicate was added, and then as the silicate concentration increased, the reduction profile showed the upward trend expected. Curve C of Figure 16 illustrates the third source of variance (0.05%). This can be interpreted as describing the differences in reduction rates for the two heteropoly complexes. The narrow peak at approximately 3 s describes

ANALYTICAL CHEMISTRY, VOL. 62, NO. 14, JULY 15, 1990

2

4

6

8

10

12

14

,e

SAMPLE NUMBER

Figure 18. Relative coordinates of the 16 standard samples along the first source of variance.

the faster phosphomolybdate reaction, whereas the broader trough centered at 7 s describes the slower silicomolybdate reduction. While the contribution of the second and third sources of variance are small, their contributions to the PLS model are very important since they describe the subtle chemistries necessary for acceptable predictions. A plot of the scores for each sample from Table I1 is provided in Figure 18 and shows how the samples are grouped along the first source of variance. Sample groups [ l , 2,3, 41, [5, 6, 7, 81, [9, 10, 11, 121, and [13, 14, 15, 161 correspond to increasing silicate concentration at constant phosphate, while sample groups [l,5, 9, 131, [2, 6, 10, 141, [3, 7, 11, 151, and [4,8,12,161 correspond to increasing phosphate at constant silicate. If one were to “connect the dots” for these two separate groupings, a series of four vertical and four horizontal lines would result. The amount of vertical separation between points on these lines is indicative of increasing silicate concentration (vertical line) or increasing phosphate concentration (horizontal line). The greatest vertical separation (largest difference in score) between points on any one line is seen for the series of horizontal lines. These correspond to increasing phosphate concentration, the second grouping of samples listed above. This is because a greater change in absorbance values is obtained for the phosphate standards as compared to silicate standards (shown in Figures 14 and 15) and hence a greater variance.

CONCLUSIONS The use of sorbents in FIA is gaining wider acceptance as their range and usefulness increases. These methods, however, need not be limited to the study of cations and metal-ligand complexes at steady-state conditions only, as has been described in the literature thus far. A sorbent enables a sample to be isolated and preconcentrated while retaining its position relative to a flowing carrier/reagent stream. On this retained analyte, further chemical reactions, such as reduction or possibly derivatization, can be carried out while performing a spectrophotometric measurement, as demonstrated by optosensing performed in real time. This work has shown that a hydrophobic sorbent can be used in a novel way for such preconcentration and quantitation of an anion while exploiting the inherent kinetic aspects of FIA. In addition, for the phosphomolybdate and silicomolybdate complexes, the use of the sorbent C-18 reduces the analysis time by enhancing the rate at which the reduction step occurs possibly due to a surface catalytic effect. The overall result is 200-fold increase in the sensitivity of the determination as compared to conventional homogeneous spectrophotometry exploiting the same chemistry. For reversed flow elution work, problems associated with refractive index differences between the eluent and aqueous streams may be overcome by the use of peak area instead of peak height. However, a fixed time measurement of peak height may be used if data processing needs to be

1480

avoided and a slight reduction in sensitivity is not critical, or if subtraction of the blank is not possible to determine the true time at which the peak maximum occurs. The influence of partitioning in hydrophobic materials and its effect of reducing the intensity of the analytical signal warrants further study and is presently being investigated in this laboratory. Optosensing is a valuable technique for signal enhancement and has been shown to provide a new and novel way of monitoring reaction kinetics. Evaluations of dA/dT while varying the reagent stream composition provides kinetic information simply and effectively without the complex reagent/analyte ratio profile that is present in, and inherent to, homogeneous FIA chemistry. Additionally, kinetic information is obtained from the outset of the reaction, thereby avoiding the delay incurred, and the associated partial reaction completion that occurs in homogeneous FIA. It was shown that phosphate and silicate could be determined in the parts per billion range and parts per million range, respectively, with prediction errors of 9.3% and 11.5%, respectively, as based on the use of partial least-squares analysis with cross validation. Lower concentrations of silicate may be analyzed by increasing the time allowed for the silicomolybdate complex to form. This can be accomplished by lowering the flow rates or increasing the reaction coil lengths. With an increase in the reaction time, more silicomolybdate is formed and therefore adsorbed onto the column. Since the method actually measures the loading of the analyte on the detector, the net effect is to lower the concentrations of silicate that can be determined. Also, greater kinetic discrimination might be obtained by varying the reagent concentrations. We were able to decrease the prediction errors by reducing the ascorbic acid concentration, thereby providing greater kinetic discrimination between the phosphomolybdate and silicomolybdate reduction reactions. To minimize the prediction errors, an optimization study would need to be conducted that should be specific for the conditions under which the determination is to be performed. Performing reactions on solid surfaces holds great potential for expanding the applications of sorbents in FIA. When combined with optosensing, the range of applications is increased tremendously to include kinetic studies of systems that were previously unsuitable because of problems associated with unstable reagents and/or products. Also, since kinetic informaiton is obtained form the outset of the reaction, fast reactions that were incompatible with homogeneous FIA systems might now be compatible with optosensing FIA kinetic studies. A possible application for this technique that should be investigated further is in the biochemical/clinical area, wherein reaction rates are extensively used and the samples are typically expensive and small.

ACKNOWLEDGMENT We express our appreciation to Mary Beth Seasholtz of the Laboratory for Chemometrics at the University of Washington for her help with the principal-component decomposition of the data. Registry No. Phosphate, 14265-44-2; silicate, 12627-13-3; 12-molybdophosphoricacid, 12026-57-2.

LITERATURE CITED (1) Kariberg, B.; Thelander, S. Anal. Chim. Acta 1978, 98, 1-7. (2) Baadenhuijsen, H.; SeurenJacobs, H. E. H. Clin. Chem. 1878, 25. 443-452. (3) Bergamin, H.; Reis, B. F.; Jacintho, A. 0.; Zagatto. E. A. 0.Anal. Chim. Acta 1880, 717, 81-89. (4) Johnson, 8 . F.; Dorsey, J. G. SubmHted for publication In Anal. Chem. (5) Ruzicka, J.; Arndal, A. Anal. Chlm. Acta 1988, 276, 243-255. (6) Ruzicka, J.; Hansen, E. H. Anal. Chim. Acta 1985, 773, 3-21. (7) Yoshimura, K.; Waki, H. Talanta 1983, 3 2 , 345-352. (8) Lazaro, F.; Luque de Castro, M. D.; Valcarcei, M. Anal. Chim. Acta 1988, 219, 231-238. (9) Lacy, N.; Christian, G. D.; Ruzicka, J. Anal. Chim. Acta 1888, 224, 373-381.

1490

Anal. Chem. 1990, 62, 1490-1494

(10) Lacy, N.; Christlen, 0.D.; Ruzicka, J. Ouim. Anal. 1989,8 , 201-209. (11) Ruzicka, J.; Hansen, E. H. Flow lnjectlon Analysis; 2nd ed.; Wiley-Interscience: New York, 1988;pp 303-309. (12) Hlrai. Y.; Yoza, N.; Osashi, S. Anal. Ch/m. Acta 1980, 715, 269-277. (13) Motomlzu, S.; Wakimoto, T.; Toel, K. Talanta 1983. 30, 333-338. (14)Hansen, E. H.; Ruzicka, J. Anal. Chim. Acta 1978,8 7 , 353-363. (15) Linares. P.; Luque de Castro, M. D.;Valcarcel, M. Anal. Chem. 1986, 58, 120-124. (16)Johnson, K. S.;Petty, R. L. Anal. Chem. 1982,5 4 , 1185-1187. (17) Llnares, P.: Luque de Castro, M. D.; Valcarcel, M. Talanta 1986. 33, 889-893. (18) Hart, E. A. personal communication, 1989. (19)Kolthoff, I. M.:Eking, P. J. Treatise on Ana/yt/cal chemistry; Interscience Publishers: New York, 1961;pp 317-394.

(20)Snell F. D.; Snell, C. T. ColormetricMS#KXIS of Analysk, 3rd ed.;Van Nostrand: New Yo&, 1949;Vol. 2,pp 660-671. (21) Kircher, C. C.; Crouch, S. R. Anal. Chem. 1983,55, 248-253. (22) Clark, G. D.;Christian, G. D.; Ruzicka, J.; Anderson, G. F.; VanZee, J. A. Anal. Instrum. 1989, 78, 1-21. (23)Sandell, E. B.; Onishi, H. PhotomeMc Determination of Traces of N e ments; 4th ed.; Wiley Interscience: New York, 1978;Chemical Analysis Series, Vol. 3,pp 249-256. (24)Sentell, K. B.; Dorsey, J. G. Anal. Chem. 1989, 67, 930-934. (25) Stone, M. J . R . Statist. SOC.6 . 1974,36,111-133. RECEIVED

for review January 17, 1990. Accepted April 17,

1990.

Determination of Chloroaniline Traces in Environmental Waters by Selective Extraction with Two Traps in Tandem and Liquid Chromatography Antonio Di Corcia* and Roberto Samperi Dipartimento di Chimica, Universitci L a Sapienza di Roma, Piazzale Aldo Mor0 5, 00185 Roma, Italy

Selectlve llquld-solld extraction from environmental waters of 14 chloroanlllnes was achieved by uslng a two-trap tandem system, one contalnhg a nonspeclc adsorMng material, such as graphttlzed carbon black (Carbopack B), and the other one fllled with a resin-based strong catlon exchanger. After percolatlon through the Carbopack column (extractlon cartrldge) of water samples, the two traps were connected In serles, acetadtrHe ack#fted wlth HCI, 10 mmaUL, was allowed to flow along them, and chloroanlllnes displaced from the extractlon cartrldge were selectively readsorbed via hydrogen bondlng on the strong acld exchanger column (Isolation cartridge). After thls column was washed, the analytes were eluted from the lsolatlon cartrldge wlth 1 mL of aqueous acetonitrile basified with KOH, 0.1 mol/L. After sultable neutralzatlon, 0.2 mL of thls solution was dlrectty InJected Into the llquld chromatographlc apparatus, whlch was operated lsocratkally In the reverse-phase mode with UV detectlon at 240 nm. The analytlcal recoveries of the 14 chloroanlllnes were hlgher than 88%. The llmlts of detectability of the analytes consldered were well below 0.1 pg/L. The effectlveness In terms of recovery and selectlvity of the two traps In serles was compared wlth three other cartridges, which contalned a chemlcally bonded siliceous material (C,& Carbopack, and a catlon exchanger. These latter two cartrldges were the very same we used In tandem, but in thls case they were operated lndlvldually.

Aromatic amines may be present in the aquatic environment as a result of industrial discharges from factories using anilines as intermediates for the synthesis of special chemicals or as result of the degradation of some commonly used herbicides, such as phenylureas. Toxicological data are known for several aromatic amines and some are suspected to induce cancer (1,2). In this vein, the European Economic Community (EEC) has included many anilines in the list of priority pollutants which should be monitored in environmental *To whom correspondence should be addressed.

waters. Although several analytical schemes for the extraction, concentration, and detection of various basic compounds in water have appeared in the literature (3-6), several aniline derivatives, such as monochloro-, dichloro-, and methylchloroanilines are still included in that particular class of pollutants for which a reliable analytical procedure able to detect them a t 0.1 pg/L in water is unavailable. Various, additional reasons contribute to make conventional analytical schemes inadequate to the trace determination of aromatic bases. Firstly, the hydrophilic nature of these compounds results in partial extraction from relatively large water volumes by using either solvent or solid-phase extraction (7). Secondly, severe losses of the more volatile anilines may take place during the solvent removal step performed to increase the enrichment factor. Thirdly, the adoption of nonselective extraction procedures is ultimately reflected in great complexity of the final chromatogram when the simultaneous analysis of variously substituted anilines is performed. To enhance selectivity, extraction procedures of aromatic bases from water by cation exchangers have been proposed (8-11). However, these methods generally have detection limits higher than 1pg/L. Nielen et al. (12)reported a high-performance liquid chromatography (LC) procedure for determining traces of aromatic bases that involves on-line preconcentration on a short precolumn filled with a strong acid exchanger. This method, however, requires a sample pretreatment to eliminate calcium ions and it seems unsuitable for preconcentrating very weak bases, such as halogen ortho substituted anilines. The combination of on-line preconcentration by means of a nonselective adsorbent and LC subfractionation with electrochemical detection was found effective for determining four aromatic amines spiked in seawater samples (13). Because of a severe baseline deflection resulting from the sample back flushing onto the analytical column, this method appears to be uneffective for assaying aromatic amines at the level of 0.1 E/L. Recently, the isolation of triazine herbicides, which are weak bases, from an acetone extract of soil has been accomplished by exploiting the hydrogen bond formation between triazines and the acidic sites of a strong cation exchanger (14). More recently, a double trap tandem system, consisting of a cartridge

0003-2700/90/0362-1490$02.50/00 1990 American Chemical Society