2058
Anal. Chem. 1982, 5 4 , 2056-2061
meaningless. However, retaining all data points will still have advantages in computing variances and confidence intervals. Therefore, in the characterization of a reaction mechanism, replicate measurements have two significant advantages. First, they allow an estimate of the inherent (or pure) error in an analytical measurement, which can be compared with the error due to lack of fit. From this comparison, an objective judgment can be made of the adequacy of a postulated reaction mechanism, allowing any necessary revisions to be made. Once a satisfactory mechanism has been determined, replicate measurements have a second advantage in decreasing the size of confidence intervals for estimated rate constants. In the absence of replications, alternative methods allow for comparisons between reaction models (14). As pointed out by one of the referees, the use of averages in regression procedures is often a case of double-filtering of the data which can lead to distortions, missing real features by oversmoothing, etc.
LITERATURE CITED (1) Guilbault. G. G. I n "Treatise on Analytlcal Chemistry", 2nd ed.; Kolthoff, I.M., Elvlng, P. J., Eds.: Wiley-Intersclence: New York, 1978; Pari I,Vol. 1, pp 663-710.
(2) Guilbault, G. G. "Handbook of Enzymatic Analysis"; Marcel Dekker: New York. 1977. (3) Mark, H. B.; Rechnitz, G. A. "Kinetics in Analytical Chemistry"; Wiley: New York. 1968. (4) Horacio, A. M.; Mark, H. 13.Anal. Chem. 1980, 52, 31R-40R. (5) Greinke, R. A.; Mark, H. E. Anal. Chem. 1978, 50, 70R-76R. (6) Snedecor, G. W.; Cochran, W. G. "Statistical Methods", 7th ed.; Iowa State Unlverslty Press: Ames, IA, 1980; pp 171-172. (7) Deming, S. N.; Morgan, S. L. Clin. Chem. (Winston-Salem, N . C . ) 1979, 25, 840-855. (8) Draper, N. R.; Smith, H. "Applled Regression Analysis", 2nd ed.; Wiley: New York, 1961; p 94. (9) Draper, N. R.; Smith, H. "Applied Regression Analysis", 2nd ed.; Why: New York, 1981; pp 33-42. (IO) Warrick, P., Westminster College, unpublished results. (11) Warrlck, P.; Auborn, J. J.; Eyring, E. M. J. Phys. Chem. 1972, 76, 1184-1 191. (12) Olsen, S. L.; Holmes, L. P.; Eyring, E. M. Rev. Sci. Instrum. 1974, 45,859. (13) Purdie, W.; Eyring, E. M.; Rodriguez, L. I n "Techniques of Chemistry"; Welssberger, A., Rossiter, E., Eds.; Wiiey: New York, 1980; Voi. 9. (14) Shiilington, E. R. Can. J. Stat. 1979, 7 , 137-146.
RECEIVED for review May 5, 1982. Accepted July 12, 1982. Financial support of this work by a contract from the Department of Energy (Office of Basic Energy Sciences) is gratefully acknowledged.
Optimization of a Flow Injection Analysis System for Multiple Solvent Extraction Thomas
M. Rossl, Dennis C.
Shelly, and Isiah
M. Warner"
Department of Chemistty, Texas A&M University, College Station, Texas 77843
The performance of a multlstage flow Injection analysls solvent extractlon system has been optlmlzed. The effect of solvent segmentatlon devices, extractlon colls, and phase separators on performance characterlstlcs is dlscussed. Theoretical conslderatlon Is glven to the effects and determlnatlon of dlspersion and the extractlon dynamlcs within both glass and Teflon extractlon coils. The optimized system has a sample recovery slmllar to an identical manual procedure and a 1.5 % relatlve standard devlatlon between InJectIons. Sample throughput time is under 5 mln. These characteristlcs represent signlflcant Improvements over the performance of the same system before optlmlzatlon.
into alternating segments. Modified A8 connectors (4-8) and various configurations of glass capillary tubes (9,10) have been used for phase segmentation. The extraction process requires that the segmented phases be allowed to remain in contact while the analyte approaches a state of thermodynamic equilibrium by partitioning between the two phases. This process is achieved while the segmented phases are pumped through an extraction coil. Finally, the phase separation process involves a partitioning of the segmented phases. During this process, the unwanted phase is sent to waste while the other phase is "resampled" or pumped through for detection. Phase separation is usually accomplished by the use of a modified A4 connector (4-7) or a membrane separator (8-10).
Flow injection analysis (FIA) has become an important technique for laboratory automation. Many simple procedures involving FIA have already been investigated (1). For example, titration (2),extraction, chemiluminescent (3-11), and electrochemical (12) systems have been automated by FIA technology. In general, FIA consists of three primary steps. First, a sample is injected into a flowing stream. Second, some chemical or physical operation is performed on the sample. Third, and finally, the resultant product is detected. Karlberg and Thelander (4) first demonstrated the use of FIA for solvent/solvent extraction. In their study, they reported the automated extraction of caffeine from an aqueous medium to an organic medium. Since this preliminary investigation, many FIA automated solvent/solvent extraction systems have been reported (4-11). Three physical operations are essential to FIA automated solvent/solvent extraction systems. These are (1) phase segmentation, (2) extraction, and (3) phase separation. Phase segmentation involves dividing the aqueous and organic phases
The automated extraction systems described above involve simple, single-stage extractions. Complex multistage procedures have been automated by continuous flow analysis (CFA, air segmented streams), but it was considered impractical to automate such procedures using FIA. Complex FIA systems are constructed by linking together several simpler FIA operations. After interfacing, the complete system must be examined and concomitant loss of performance due to interfacing the individual operations must be minimized. It is important to recognize that merely interfacing the individual working operations does not necessarily produce a working complex FIA system. It is quite often found that interfacing individual components will produce problems such as back-flushing which are not present in the individual operations. Therefore, it is essential to optimize the performance of the entire system. Recently, Shelly et al. reported the automation of a complex solvent/solvent extraction procedure using FIA technology (14). The extraction was a three-stage manual procedure originally reported by Natusch and Tomkins (15). This ex-
0003-2700/82/0354-2056$01.25/0 1982 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 54, NO. 12, OCTOBER 1982
traction is useful for the isolation of polynuclear aromatics (PNAs) from a complex organic matrix. In the preliminary study the manual extraction was first dissected into its simplest components. An automated system for each extraction and back-extraction was then designed. Finally, the three automated extraction stages were interfaced, producing the final system. In this paper we explore the optimization of the three-stage extraction system dtesigned by Shelly et al. This study is designed to improve three basic performance characteristics. These characteristics are (1)increased sample recovery, (2) increased reproducibility of recovered sample, and (3) decreased analysis time.
THEORY General. Having identified the performance characteristics of the original system which need to lbe improved, it is necessary to establish haw the physical parameters of the system affect each of these characteristics. This can be done by sequentially conside ring each component involved in an automated system. First, the solvent ]pumping system must provide a stable, reproducible flow rate through each channel. Also, it is desirable to have a pump with the capability of independently varying the flow rate of each channel (12). The next componeint of importance is the injection valve. The requirements for a good injection system are well-known (13). In our multiple extraction system only the first extraction section usefi an injection valve. After sample injection is accomplished, the two immiscible phases necessary for the extraction aire brought together in the solvent segmentation device. There are two major variables to be examined here: (1) the reproducibility of segmentation and (2) segment size. While segment size may not affect extraction efficiency in a kinetically fast extraction, it could theoretically aYfect efficiency in a slower system. This will be discussed in rnore detail later. Fluctuation in solvent ratios caused by inconsistent segmentation has the adverse effect of forcing some of the phase intended for resampling out to waste a t the phase separator. Hence, uneven segmentation leads to EL decrease in sample recovery. The extraction coil may also influence both sample dispersion and extraction efficiency. There are two considerations to be made when constructing an extraction coil. First, one must decide what material to use. Ih the past both Teflon (3-12) and glass (141 extraction coils have been used. The choice of which coil material to use is based on whether the sample will be extracting from an aqueous to an organic phase or vice versa. The ,second choice to make concerning the extraction coil is how long the coil should be. The length of the coil will affect dispersion of the samples as well as extraction efficiency. The coil can be made long enough such that transfer kineticti is not the limiting factor in extraction efficiency. Any increase in coil length beyond this point will result in increased sample dispersion without a concurrent increase in sample recovery and is therefore undesirable. Finally, the extraction has been completed and the two immiscible phases must be separated. Thus, a phase separator is the necessary finad component. The separator must be designed to provide complete separation of the two phases with very little waste of the resampled phase. Also a small internal volume is desirable so as to limit the amount of sample dispersion occurring at this point. Dispersion. Dispersion of the sample as it travels through an FIA system produces a broadening of the output peak. This peak broadening results in a decreased sample throughput rate and hence must be minimized (15). Dispersion is nornially expressed atr a function of sample concentration. In the past the followiing equation has been
2057
Figure 1. Diagrammatic representation of two immiscible phases “a” and “b” as they flow through an extraction coil. The arrows indicate the general flow of solutes as they transfer from phase “b” to phase “a”.
used to calculate the dispersion of a sample in a FIA system (16): where Co is the original concentration of the analyte,, C , is the maximum concentration of the analyte a t the point of detection, and D is a dimensionless dispersion number. The use of this equation is justified when the only cause for a reduction of, ,C below the value of C, would be an increase in the volume of solution which contains the analyte a t the point of detection relative to the injection volume. However, in a system where some analyte is lost during analysis,, ,C could be reduced below the value of C, without any change in the volume of solution containing the analyte. Extraction efficiencies of less than 100% are frequently encountered in organic FIA extraction systems. In these systems, then, D is not a reliable indication of the true dispersive properties of the system. The inadequacy of D to truely represent dispersion has been mentioned by Karlberg et al. ( 5 ) ,but a solution to this problem was not proposed. When attempting to optimize and understand the performance of an FIA system, it is important to view the dispersion of the system independently of any change in sample recovery. For this reason we propose the following equation as a more reliable means of expressing dispersion: where S~ is the standard deviation of the injected peak width in terms of volume, S, is the standard deviation of the output peak width in terms of volume, and D, is a dimensionless number expressing dispersion, henceforth to be known as the volumetric dispersion number. The injected “peak” width is determined by assuming that the sample is injected in a square wave profiie with a width proportional to the injection volume. Changes in analyte concentration do not influence D,. Hence, by monitoring D, we can observe the effect of changing the physical conditions of extraction on dispersion, independent of the effect on extraction efficiency. It is important to study dispersion because it will affect two aspects of the system’s output. The greater the dispersion, the more dilute the analyte will be at the output of the system causing an increase in the detection limit. Also, the more the sample is dispersed, the longer sample throughput time becomes. As a general rule for resolution between peaks of R = l(17)
Tt = 4St where Tt is the sample throughput time and S, is the standard deviation of the peak width in time. Note that sample throughput rate is directly proportional to the output peak width in time. Considering the foregoing discussion, it seems desirable to examine both D, and S, of an FIA system when optimizing the system. If this is done, it is possible to monitor the relative detection limits and sample throughput rates separately as functions of D, and St, respectively. Extraction Dynamics. The spacial orientation of the immiscible phases inside the coil is shown in Figure 1. In this diagram, phase “a” wets the coil wall while phase “b” forms
2058
ANALYTICAL ChEMISTRY, VOL. 54, NO. 12, OCTOBER 1982
evenly spaced droplets or plugs completely surrounded by phase “a”. According to the principle of selective wetting, if the extraction coil is constructed of Teflon, phase “a” is the organic phase. Conversely, if the extraction coil is constructed of glass, phase “a” will be the aqueous phase. According to Shelly et al. (14), in order to limit sample carryover it is desirable to choose the extraction coil material such that the initial concentration of analyte will be contained in phase “b”. Another criterion for choosing which extraction coil to use has not been previously mentioned and deserves mentioning here. Two factors will influence the kinetics of an extraction. One factor is the ratio of the area of the phase interface to the volume of the phase initially containing the analyte. The higher this ratio, the more favorable are the extraction conditions. The second factor is the ease with which the analyte reaches the phase boundary from anywhere in the original solution. In an FIA extraction coil, the ratio of phase contact area to the original solvent volume is determined by the solvent segment size. The ease with which the solute reaches the phase interface will be determined by whether the solution containing the initial concentration of the analyte is solvent “a” or “b” as diagrammed in Figure 1. If the analyte is initially contained in solution “b”, there should be free movement of the solute molecules within the bubble toward the phase interface. In a dynamic process such as in the FIA extraction coil secondary flow patterns will develop within the bubble of solvent “b”. These flow patterns will tend t o provide a constant supply of solute flowing to the phase boundary from within the bubble of solvent “b”. Once the solute has crossed the phase boundary and is contained in solvent “a”, it must be drawn away from the boundary in order t o maintain a chemical potential favorable for continued extraction. This should not be a problem a t the leading and trailing edges of the bubble “b”. Here secondary flow patterns will remove solute molecules from the interface toward the bulk of solvent “a”. In the region where solvent b is separated from the extraction coil wall by a thin layer of solvent “a”,a slightly different situation exists. Solute molecules extracting into this region cannot be brought directly into the bulk of solvent a by secondary flow patterns. Fortunately the linear velocity of “a” is low in this region; hence, the trailing portion of the thin layer of “a” is constantly being mixed with the bulk of “a”. A fresh solution of a is also being introduced a t the leading edge of this thin layer. Figure 1 is a graphic representation of these ideas. I t shows the general pathway traveled by the extracting species. Now consider the situation where the extraction must occur from solvent “a” to solvent “b”. In this case, the analyte must be channeled into the thin layer of solvent a in order to reach the “a/b” interface along the side of the bubble of “b”. Once it is in this layer of “a”, it must be extracted into “b” before it leaves the thin layer a t the trailing edge. Thus, the flow patterns in this situation make it difficult for extraction to occur. It follows from the above discussion that for an extraction proceeding from the organic to the aqueous phase, a glass extraction coil should encourage a higher extraction efficiency than a Teflon coil of equal length. Conversely, for an extraction proceeding from the aqueous t o the organic phase a Teflon coil should have a higher extraction efficiency than would a glass coil of equal length. EXPERIMENTAL S E C T I O N Reagents. Perylene was obtained from Sigma Chemical Co., St. Louis, MO. Omnisolve glass distilled dimethyl sulfoxide was obtained from MCB Manufacturing Chemists Co., Cincinnati, OH. Glass distilled cyclohexane and acetonitrile were both obtained from Burdick and Jackson Laboratories, Muskegan, MI. Apparatus. A schematic of the optimized extraction apparatus is shown in Figure 2. Solvent pumping was performed by using a Gilson Minipuls 2 , eight-channel peristaltic pump (Gilson
DMSO
-1
a
L
WATE‘I
‘r-
L ....2
4
5
CYCLOHEXANE
-
~
4
DETECTION AND COLLECTION
Flgure 2. Schematic diagram of the optimized FIA extraction system: (VI) sample injection valve with variable volume injection loop; (a)A 8 phase segmentation device: (b) 100-cm glass extraction coil; (c) A4
phase separator: (d) cooling coil; (e) 200-cm Teflon extraction coil; (w) waste. Pump tubing for the indivldual channels were as follows: (1) 1.01 mm solvent flexible; (2)1.42 mm Viton: (3) 1.42 mm Viton; (4) 1.01 mm silicone; (5) 1.42 mm solvent flexible; (6) 1.14 mm solvent flexible. Medical Electronics, Middleton,WI). Silicone and solvent flexible pump tubing were provided by Fisher Scientific, Fairlawn NJ. Black Viton tubing (Gilson Medical Electronics) was used to pump Me2S0. Teflon tubing and polypropylene fittings were obtained from Altex Scientific,Berkely CA, as were the pneumatically activated valves and pneumatic interface. Teflon phase segmentation devices were milled by Brazos Technology, College Station TX. A4 connectors, used for phase separation, were obtained from Technicon Instruments, Tarrytown, NY. The glass extraction coil was constructed with glass-lined stainless steel tubing, 7 mm i.d. obtained from Altech Associates, Arlington Heights, IL. Instrumentation. A Farrand Model 801 spectrofluorimeter equipped with a 10-fiL flow cell was used to monitor the fluorescence intensity of perylene extracts. Later studies used a Perkin-Elmer LS-5 spectrofluorimeter. Liquid chromatography of extracted PNA mixtures was performed on an Altex Model 312 MP high-performance liquid chromatograph using a 25-cm Ultrasphere C18column. Isocratic elution with a 70% MeOH/H20 mobile phase was used. Procedure. The original FIA system is diagrammed in the earlier paper by Shelly et al. (14). Optimization was performed by examining each section of the system separately. The first process was to rebuild the first extraction section and to optimize its performance. Next, the second extraction section was added and the performance of the two-stage extraction system was studied. The third major step was to disconnect the two sections and examine the second section alone. Finally, the first and second extraction sections were reconnected, and the third extraction section was considered. The third section was originally designed to remove any residual MezSO from the cyclohexane. We eliminated the third extraction section and passed the extract through glass wool t o remove Me2S0 and other polar residuals. This simplified the overall system considerably. It also allowed for both decreased system dead volumn and increased sample recovery. The new system was mounted on a vertical section of Plexiglass approximately 10 X 12 X in. This allowed each extraction section to be mounted very close to the pump head reducing significantly the dead volume due t o resampling. Also during this study it was recognized that a replacement solvent for pentane was needed since its high vapor pressure produced pumping difficulties. Cyclohexane was selected because of its low vapor pressure and high purity. Solvent pump ratios were varied by using various inner diameter pump tubing. Five different sample injection volumes were tested. The sample percent recovery was used as a basis of comparison for each injection volume. Injection volumes were varied by changing the volume of the sample loop used in the pneumatic injection valve. Injection volumes tested were 100, 200, 300,400, and 500 PL. Three types of phase segmentation devices were tested. The devices were evaluated with regard to their consistency of operation and their effect on sample recovery.
ANALYTICAL CHEMISTRY, VOL. 54,NO. 12, OCTOBER 1982
Table-I
17
extraction channel
solvenit/solvent system
1
Me,SO,I cyclohexane water/Pde,SO cyclohexane/ water in Me,SO
2 2
2059
15
pump solvent ratio in ratio manual proced.ure 1:3.1
3:l
1:1.3 1:1.#6
3: 1
> Y e
1:l
Both glass and Teflon extraction coils were tested in the two extraction sections. Each coil was tested at different pump speeds. The extraction effciermy and dispersive properties of the channels were plotted as a function of flow rate. This formed the basis for choosing which extraction coil type! was to used in a given section and at what pump speed the system should be run. Solvent separation was accomplished by using the modified A4 connector. A memlbrane separator (8)was tested but was found unsuitable for the solvent systems used in our apparatus. Quantitative evaluation of the system was performed by using 500-bL injections of M perylene iin cyclohexane. For the evaluation of the second channel alone, a M solution of perylene in MezSO wm used. The resampled phase was pumped through the flow cell and the amount of perylene contained in this phase was determined by computing the area under the output curve. In order to qualitatively evaluate the performance of the system, we injected a mixture of PNAs dissolved in cyclohexane into the extraction system. The extracted sample was separated by HPLC and the chromatographic profile was compared to the original mixture.
RESULTS AND DISCVJSSIONS Solvent Delivery Ratios. Table I is a comparison of the solvent delivery ratios used in this system with the solvent ratios used in the manual extraction. It is evident from Table I that we could not achieve the same solvent/solvent ratios in the automated syaitem as we did in the manual procedure. This was a limitation imposed on us bly a small range of flow rates available when using one peristaltic pump. Also we were occasionally prohibited from using hi.gh flow rates in an individual channel by adverse back-pressure effects. Phase Segmenta.tion. Each of the phase segmentation devices used in this study provided solme advantages as well as disadvantages. The Teflon tee fittings had the major advantage of being inexpensive and easily available. They offered consistent segmentation but the segments were of fixed size and this size was large compared to the smallest segments possible with the A8 connector. Another problem with these segmentors was that ithey were not very good at flow direction. That is, too much ba'ck-pressureat the outlet of the tee could result in the direction of flow being reversed in one of the solvent delivery lines. The modified A8 connector had the advantages of having variable segment sizes and of being very good at maintaining direction of flow. The A8 connectors could tolerate the most back-pressure. The problem with thecie segmentors was that the separation between segments wat3 not consistent. Finally, the 45' "W" fittings were very consistent in segmentation. They did not however have variable segment sizes. They were almost asi good for flow direction as the A8 connectors. The 45' " W Teflon phase segmentation device was selected for use in the first extraction channel. In the second channel the A8 connector was used because of its superior flow directing capability. Extraction Coils. The performance of a 100-cm glass extraction coil with i3 100-cm Teflon coil was compared for the first extraction chlannel. Figure 3 slhows the results of this comparison. It is observed that the glmis extraction coil allows significantly higher sample recovery than does the Teflon coil a t the same residence time. This is in1 good agreement with
5
T
b
T
'L 0
200
-
,
30C
boo
boo
boa
PUMP SPEED P O T E N T I O M E T E R S E T T I N G
137
I80
2.41
, I
2 99
3 61
FLOW RATE m l l m l n
Figure 3. Graphic representation of the performance of glass (0)and Teflon (A)extraction coils in the first extraction channel: (a) % reM perylene vs. pump speed; (b) volumetric dispersion covery of (D,) vs. pump speed; (c) S , vs. pump speed. Error bars represent standard deviations.
the theoretical predictions made in the extraction dynamics section of this paper. Interestingly, an examination of Figure 3b reveals that the Teflon coil disperses the sample less than the glass coil does. This is contrary to the beliefs that first led us to try the glass coil (13)but not contrary to the theory developed in this study. Finally, a comparison of Figure 3b,c also serves to illustrate a point mentioned in the theoretical section of this paper. While the volumetric dispersion (D,) increases steadily with pump speed, the peak width in time decreases. Hence, while the sample is being dispersed to increasingly higher degrees, the sample throughput time is actually decreasing. One should, therefore, be careful not to equate volumetric sample dispersion with throughput time. On the basis of the results shown in Figure 3, we decided to use the glass extraction coil at a flow rate of 1.39 mL/min. Under these conditions, an acceptably low sample dispersion and the maximum possible sample recovery were maintained. Figure 3a shows that a further increase in sample residence time a t flow rates less than 1.39 mL/min will not yield any significant increase in sample recovery. This may be due to
2060
ANALYTICAL CHEMISTRY, VOL. 54, NO. 12, OCTOBER 1982
a
70i L
701
60
50
c
e
40
0
30
8
1
20C
4
----,-5 6
--7-------”
7
8
9
1
i l E l A T I V E S A h F L E RESIDFNCE T I M E
b P
7T
C
.2
+
c - t - ~ F - - . - t - - + 200
300
400
500
600
730
800
900
1000
357
415
h58
PUMP SPEED P O T E N T I O M E T E R S E T T I N G
I
090
I39
I88
222
267
321
FLOW RATE rnl/mln
Flgure 4. Graphic representation of the performance of glass (0) and Teflon (A)extraction coils in the second extraction channel: (a) % recovery of lo4 M perylene vs. pump speed; (b) D, vs. pump speed; (c) S , vs. pump speed. Error bars represent standard deviations.
a decrease in the rigor of extraction conditions a t low flow rates. The second extraction channel was then added to the system. An 80-cm glass extraction coil and a 0.8 mm internal diameter Teflon coil of equivalent volume were tested in this channel. The results of these test are shown graphically in Figure 4. Two facts are revealed by the displayed data. First, dispersion is relatively constant a t the pump speeds tested and the Teflon coil disperses the sample slightly less than does the glass coil. Second, the Teflon coil is somewhat more efficient than the glass coil. This is in agreement with the theory developed earlier in this paper. By examination of Figure 4a one might assume that a further increase in residence time in the Teflon coil would not yield any increase in sample recovery. That is, there is a plateau of extraction efficiency vs. pump speed as one decreases flow rates to less than 2.41 mL/min on the Teflon coil efficiency profile. This was surprising since it implies that about 16% extraction efficiency is optimum. However, it must be noted that Figure 4a represents the extraction efficiency profiles of the combined first and second channels.
Flgure 5. Graphic comparison of 100-cm (A)and 200-cm (0)Teflon extraction coils in the second extraction channel isolated from the first channel. Error bars represent standard deviations.
The above results prompted an investigation of the efficiency of the second extraction channel separated from the first channel. Figure 5 shows the results of testing both a 100-cm and a 200-cm Teflon extraction coil. It is apparent that the 100-cm coil cannot provide a sample residence time sufficient to allow maximum sample recovery. A plateau in sample recovery is reached with the 200-cm coil a t a sample residence time corresponding to a flow rate of 1.80 mL/min. Note that extraction efficiency is plotted as a function of sample residence time rather than directly as a function of pump speed. This is necessary since the extraction coils being compared are of different lengths; hence, at one pump speed the sample residence time is different for each coil. The 200-cm extraction coil with a flow rate of 1.80 mL/min was used in the final system. The information produced in this part of our study must also be qualitatively evaluated. The user of an extraction system must examine this information with respect to need. If a high sample throughput is of primary importance then one should make different selections for operating conditions than in the case where high sample recovery is of primary importance. In this study, maximum sample recovery was the important concern; hence, small sacrifices in sample throughput were acceptable. Some general features of the efficiency profiles of this section merit attention. First, it should be noted from Figures 3a and 4a that as pump speed increases there is a crossing of profiles for the two coils shown in each diagram. In general, when the extraction coils have reached their lowest sample recovery as a function of sample residence time, a further increase in flow rate results in an increased sample recovery. This is the result of the interaction of flow conditions with extraction kinetics. In the portion of the efficiency profile that decreases with decreased sample residence time, the decrease in the efficiency is well-known to be a result of the decreased residence time not allowing the extraction to reach thermodynamic equilibrium. However, as the residence time shortens, increased turbulence within the coil at higher flow rates will create more vigorous extraction conditions. The effect of these more vigorous conditions begins to dominate the efficiency profile when residence time is short enough such that further decreases in residence time are no longer the main factor influencing extraction efficiency. Another example of the effect of flow conditions on extraction efficiency can be seen in Figure 5 . In the region of the diagram where both coils are being tested at the same residence time, both have similar profiles with the extraction efficiency in the 200-cm coil being consistently greater than that in the shorter coil. This displacement of the two curves occurs because the flow rate in the 200-cm coil is twice that of the flow rate in the 100-cm coil for the same sample residence time. Thus, the extraction conditions in the longer coil
ANALYTICAL CHEMISTRY, VOL. 54, NO. 12, OCTOBER 1982
Table 11. Performanoe Characteristics of the Extraction System with Various Injection Volumes inj vol, pL
% recov
100 200 300 400 500
23.8 19.7 16.8 19.0 21.3
ered
% re1 S, peak std dev width, min T,, min
i.1.5 1.9 5.9 4.6 3.1
0.50
0.52 0.60 0.60 0.62
2.00 2.08 2.10 2.10 2.52
b
2061
Hence, for quantitative work a standard solution of the anal@ of interest should be extracted in order to determine its extraction efficiency before analysis of an unknown is performed. Also note that in Figure 6b the effect of passing the sample through a small plug of glass wool is demonstrated. After the sample was filtered with glass wool, all of the highly polar (i.e., early eluting) impurities have been removed. These impurities are believed to be introduced by the Viton tubing used to pump Me2S0. The contaminant is nonaromatic and hence should not interfere with LC or fluorometric analysis. Several alternative pumping arrangements are now being considered. The optimized system has a sample recovery for perylene of 23.8 f 1.5%. This is comparable to the manual extraction when cyclohexane is used as the organic phase. Sample throughput time as calculated by eq 3 is 2.0 min. Peak tailing results in an increase of Tt to -4.5 min. In addition to optimizing this extraction system, this study has aided in evaluating a theory of extraction dynamics within the coil. Also, this study proves conclusively that complex FIA extraction systems are workable and desirable. The logistics of the approach used in this study can be used in the design, construction, and optimization of many complex FIA systems. LITERATURE CITED
Flgure 6. Liquid chromatographyof PNA mixtures (70 % MeOH, 30 % H,O mobil phase, 1.5 mL/min: (a) unextracited mixture of PNAs at (- - -) 1.28 absorbance units full scale deflection and (-) 0.64 absorbance unit full scale deflection; (b) the same mixture after extraction, (- - -) 0.32 absorbance unlt full scale deflection, (-) 0.08 absorbance unit full scale deflection, and (- -) before filtering the
extract through glass wool.
-
at similar sample residence times will be more vigorous. Injection Volumes,. Table I1 showci the extraction efficiency of the system at five different sample injection volumes. Changing injection volumes had little effect on extraction efficiency. The 1 0 0 - ~ Linjection volume produced the highest sample recovery and the highest reproducibility. In addition, the sample throughput time could be decreased with a 1 0 0 - ~ L injection. Qualitative a n d Quantitative Evaluation. Figure 6 is a comparison of the liquid chromatograms of the extracted and unextracted mixture of PNAs. All components of the original mixture have been retained in tlhe extracted sample. The relative concentrations of these coniponents are not the same in the extracted and unextracted mixtures. This is due to varying extraction efficiencies of the individual components.
Ranger, C. Anal. Chem. 1981,53, 20A-32A. Ruzicka, J.; Hansen, E. H. Anal. Chlm. Acta 1979, 706, 207. Rule, A.; Seitz, R. Clin. Chem. (Winston-Salem, N.C.) 1979, 25, 1635. Karlberg, B.; Thelander, S.Anal. Chim. Acta 1978,9 8 , 1-7. Karlberg, E.; Johansson, P.; Thelander, S. Anal. Chim. Acta 1979, 104, 21-28. Karlberg, 6.; Thelander, S. Anal. Chlm. Acta 1980, 114, 129-136. Johansson, P.; Karlberg, 8.; Thelander, S. Anal. Chim. Acta 1980, 1 1 4 , 215-226. Nord, L.; Karlberg, B. Anal. Chlm. Acta 1980, 118, 285-292. Kawase, J.; Nakae, A.; Uamanaka, M. Anal. Chem. 1979, 5 1 , 1640-1 643. Kawase, J. Anal. Chem. 1980,52,2124-2127. Bergamin, H.; Medeiros, J. X.; Reis, B. F.; Zagatto, E. A. G. Anal. Chlm. Acta 1978, 707, 9-16. Ramzing, A. U.: Janata. J.: Ruzicka, J.: L e w , M. Anal. Chim. Acta 1980, i78, 45-52. Ruzicka, J.; Hansen, E. H. Anal. Chim. Acta 1980, 714, 19-44. Shelly, D. C.;Rossl, T. M.; Warner, I. M. Anal. Chem. 1982, 5 4 ,
87-91. Natusch, D. F. S.; Tomkins, B. A. Anal. Chem. 1978,5 0 , 1429-1434. Ruzlcka, J.; Hansen, E. H. Anal. Chim. Acta 1980, 114, 19-44. Snyder, L. R. Anal. Chlm. Acta 1980, 114, 3-18.
RECEIVED for review April 13,1982. Accepted June 21,1982. This work was supported in part by grants from the Department of Energy (DE-AS05-80EV10404)and the Office of Naval Research. T.M.R. is also grateful for support by an American Chemical Society Analytical Division Fellowship sponsored by the Society for Analytical Chemists of Pittsburgh.