ANALYTICAL CHEMISTRY, VOL 51, NO 12, OCTOBER 1979
-
330
our data are plotted along with Van Geet's equation. Our ethylene glycol data give a least-squares fit of
T(K) = 466.5 - 0.461 (hJ
TEllP I K
I
2051
(3)
which is in excellent agreement with Van Geet'c; equation ( 4 ) , scaled t o 220 MHz,
t
T(K) = 466.0- 0.462 I A v ~
(4)
: ' . . . T
2sc 230
Our new data thus demonstrate that Van Geet's 60-MHz calibration equations may be confidently scaled u p to 220 MHz, and the agreement suggests that scaling to even higher frequencies should not introduce appreciably larger errors.
300
350
400
450
500
LITERATURE CITED
SHIFT I N HZ 12201
Figure 1. Temperature dependence of the chemical shift separation of CH3 and OH proton lines of methanol at 220 MHz. The data points are taken from Table I; the line represents Van Geet's 60-MHz equation scaled to 220 MHz (Equation 2 in text)
where Av is in Hz a t 220 MHz. Van Geet's equation ( 3 ) ,scaled t o 220 MHz, is T(K) = 403.0 - 0.134 IAvl - 4.92 X (Av)'
(2)
The apparent differences turn out to be insignificant over the temperature range 24C-310 K, as indicated in Figure 1 where
(1) Variable Temperature Operation in HR-220 Spectrometer System, Technical Manual , Publication no 87 122-003, Section 2 4, Varian Associates, Palo Alto, Calif. (2) R . R. Shoup, quoted in R. R. Shoup, M. L. McNeel, and E. D. Becker, J . Phys. Chem., 76, 71 (1972). (3) A. L. Van Geet, Anal. Chem., 42, 679 (1970) (4) A. L. Van Geet, Anal. Chem., 40, 2227 (1968). (5) G. L. Knott, "MLAB: An On-Line Modeling Laboratory", 7th ed.,Juv 1977, Division of Computer Research and Technology, National Institutes of Health, Bethesda, Md.
RECEIVED for review April 4, 1979. Accepted July 10, 1979.
Determination of Cobalt by Lophine Chemiluminescence Dean F. Marino, Fred Wolff, and J. D. Ingle, Jr." Department of Chemistry, Oregon State University, Corvallis, Oregon 9733 7
Interest has recently been generated in the applications of solution chemiluminescence (CL) t o trace metal analysis because of the simplicity of CL instrumentation and the low detection limits available for some metals via this approach ( 1 4 ) . Systems based on luminol and lucigenin appear to be the most popular (I+?), although other CL reagents such as gallic acid and pyrogallol have been utilized for trace metal analysis (9, 10). T h e CL of lophine (2,4,5-triphenylimidazole)has been known since 1877 (11)and the mechanism of the CL reaction of lophine and some of its derivatives have been studied (12-14). A preliminary study (15) indicated that lophine CL might be useful for trace metal determinations although Co(I1) was not identified as an activator. Our initial studies indicated that, under proper conditions, ultratrace Co(I1) concentrations enhanced the CL of lophine in basic H,Oz solutions. Thus reagent concentrations were optimized for a low Co(I1) detection limit and the interference from other species was investigated.
EXPERIMENTAL All measurements were obtained with a discrete sampling CL photometer system reported earlier (16) and with the modifications and approximate experimental conditions previously described (7, 10). Lophine (Aldrich) was used without further purification. Lophine solution preparation and storage proved to be somewhat critical. Best results were obtained by degassing reagent grade methanol via boiling followed by dissolution of the lophine in the hot methanol. Lophine solutions were refrigerated when not in use. All other solutions (e.g., HzOz,Co(II), and other metals) were prepared as previously described (7, IO). The general analysis procedure consisted of addition with Eppendorf pipets of the following quantities of the equilibrated solutions into the reaction cell: 1.0 mL sample or blank, 0.5 mL 0003-2700/79/0351-2051$01.00/0
lophine solution, and 0.5 mL H202solution. The contents of the reaction cell were allowed to mix for 10 s prior to injection of 0.5 mL of KOH solution with an automatic dispensing syringe to initiate the reaction. The CL analytical signal is taken as the difference in the CL peak height between a blank and analyte run. The cell was then evacuated and rinsed twice with 0.1 M "OB followed by two Millipore water rinses. This wash solution proved to be critical in elimination of memory effect!, and ensuring reproducibility. Typical peak shapes for blank and Co runs are shown in Figure 1. Interference studies were conducted as previously described to determine detection limits for all species and intaference levels for some species with respect to the determiniition of Co(I1). The detection limit (DL) is defined as the concentration of analyte solution yielding an analytical signal equal to twice the standard deviation of the reagent blank CL signal. For this work, the standard deviation of the blank signal was determined by the irreproducibility of the blank signal (about 5 1 0 % relative standard deviation (RSD)) and not by noise in the dark current or background CL signal. The interference level is defined as the concentration of the species which causes the mean 1 ppb Co(I1) CL signal in the presence of the species to differ from the mean Co(I1) CL signal in the absence of the species by two standard deviations in the 1 ppb Co(I1) signal.
RESULTS AND DISCUSSION Optimization Studies. Because lophine is sparingly soluble in water, a water miscible organic solverit must be used to make the lophine stock solution. The use of organic solvents in analytical CL measurements has not been explored to a great extent, even though there are potential advantages: (i) higher concentrations of the CL reagent in the reaction mixture may provide better detection limits, (ii) water insoluble reaction products may be kept in solution and increase 0 1979 American Chemical Society
2052
ANALYTICAL CHEMISTRY, VOL.
51, NO. 12, OCTOBER 1979 ' O l
'IO
0
2 0
10
LOG
con
3 0
40
P ~ B
Figure 3. Log-log Co(I1) calibration curve
TIME SEC
Figure 1. Typical peak shapes for lophine CL signals using optimized
reagent concentrations.
(A) 5.0 ppb Co(II), ( 6 ) 0.01 M HNO,
blank 500
I
400
tlM
I
-2 0
I -10
I
0
LCG KOH M
Figure 2. KOH optimization curves. [H202] = 5.9 X lo3 M, [lophine] = 4.0 X lo4 M. (A) 1.0 ppb Co(II), (6)blank, (C) DL, (D) 1.O ppb Co(I1) 500 ppb Mg(I1). The DL plotted is the extrapolated value from the 1.0 ppb Co(I1) signal
+
precision and prevent memory effects, (iii) reaction kinetics and the CL quantum efficiency may be changed to provide better detection limits and selectivity. For this study, lophine stock and test solutions were prepared in 95% ethanol, DMSO, and methanol. Methanol was found to be the optimum solvent because it provided a detection limit for Co(I1) over an order of magnitude better than with the other two solvents and because the stock lophine solutions exhibited reasonable time stability. Lophine is the least soluble in 95% ethanol ( - 2 X M). Also, in 95% ethanol, the lophine decomposed as evidenced by a decreasing CL signal even over one day. In DMSO, the CL signals were much smaller than in the other two solvents. Because Mg(I1) was identified as causing a series depressive interference, reagent concentrations were optimized for both a low Co(I1) DL and a minimal interference from Mg(I1). The results of the KOH solution optimization are presented in Figure 2. Maximum CL intensity and the best Co(I1) DL occur a t one KOH concentration, 0.2 M (1.0 M initial concentration). However, a final KOH cell concentration of 0.5
M (2.5 M initial concentration) was chosen as optimal because the Co(I1) (curve A) and COW) + Mg(I1) (curve D) CL signals most closely approach each other a t this concentration. The Hz02concentration optimization study revealed a broad maximum for the Co(I1) CL signal and a broad minimum for the Co(I1) DL, both centered a t a final concentration of 5.9 x M HzOz(3.0 X M initial concentration) which was chosen as the final optimum HzOzconcentration for this system. The blank CL signal was essentially independent of the H 2 0 2concentration. For lophine, the maximum CL signal and best DL occur M (4.0 X M initial a t a lophine concentration of 9 X concentration). However, the final cell concentration chosen for analysis was 4 X M lophine because the Mg(I1) interference was the least at this concentration. The CL signal starts to fall off at higher lophine concentrations, but this is not due to self-absorption, as lophine absorbs a t 367 nm, fluoresces a t 438 mn, and the CL signal is a t a maximum of 530 nm. In addition, no evidence has been found for selfquenching (13). Finally, Figure 3 presents a calibration curve for Co(I1) run under the aforementioned optimum conditions. The log-log plot is linear (slope = l.O), from a detection limit of 0.1 ppb (experimentally measured) up to 10 ppb Co(I1). The RSDs for 1 and 10 ppb Co(I1) CL signals are typically 2% and 8%, respectively. Interference Studies. The detection limits and interference levels for several metals, nonmetals, anions, and common complexing agents are summarized in Table I. The calibration sensitivity or slope is measured near the detection limit. A negative slope indicates that the species depresses rather than enhances the blank CL signal. It can immediately be seen that a large degree of correlation within a family of elements is certainly lacking, and that the generalization made for the luminol-Co(I1) system, that the lower oxidation state of a metal is most active (17), does not apply to the lophine system [note the difference between Cr(II1) vs. Cr(VI)]. Some correlations, however, do exist. For example, all of the depressants listed are also depressants for the lucigenin-Co(I1) system, and have similar detection limits (7). Although most species influence the CL signal a t some level, only Mg, Cr, Fe, Os, Co, Ir, Cu, C10-, Fe(CN)63-,and Mn04have detection limits below 1 ppm. The ratio of the Co(I1) detection limit to the detection limits of species likely to be found in natural systems is over lo4 for most species except for Mg, Cr, Fe, and Cu for which the ratio is over lo3. The interference levels of most of the species tested (mainly depressors) are fairly close t o their detection limits, with only a few notable exceptions. Humic acid, C1-, S042-,and EDTA probably form complexes with Co(II), making their interference level for Co(I1) determinations lower than their de-
ANALYTICAL CHEMISTRY, VOL. 51, NO. 12, OCTOBER 1979
Table I. Detection Limits and Interference Levels calibration detection interference sensitivity, species limit, ppm level, ppm (nA/ppm) Na(I) W(II) Ca( 11) Sr(I1) La(II1) TI(1V) Zr(1V) Hf(1V) V(IV) Cr(II1) Cr(V1) Mo(V1) Mn(I1) Fe(II1) Ru(II1) Os(1V)
Co(I1) Rh(1V) Ir(1V) Ni(I1) CU(I1) A d 1) Zn(I1) Cd(I1) AI(111) Pb(I1) A4V)
Sn(11) Ce(II1) FCI c10Br ~
co,*NO 1
so,>
PO-*
NH,' CNBO,+
si0~3
Fe( CN), 3MnO; citrate EDTA humic acid
> 1 0 000. 0.10
3. 438. 4. 29. 38. 52. 5.
0.25 0.005 33. 15.
0.20 6. 0.003 0.0001
28. 0.56 >lo. 0.17 10.
> 100. 100. 34. 5. 10.
25. 5. 1.0 70. 10000. 0.15
35. 8.
10000. 92. 5. 2. 21. >10000. 50.
0.24 0.48 50. 8. 37.
0.17
2 -
-
-
-
-
-
-.
-
-
100.
50. -
4.
-
200. -
>5000. 1. -
38. 50,
-
-
-
1.
0.5
-0.7
0.225 0.00016 0.017 0.0024 0 .OO 39 0.0029 0.028 6.2 29. 0.0034 0.0049 0.484 0.01 3 33. 660. 0 .OO 32 0.125
-
0.4 0.0021 -
-0.004 - 0.002 3 0.012 0.0092 0.0022 -0.004 0.0712 0.00104 7.8 x 10-5
-
0.002 0.0055 -
0.00048
0.016 0.026 0.0014
-
0.00105 0.175 1.
0.002 -0.007 - 0.0012
tection limit. P043-,however, has an interference level that is greater by an order of magnitude than its detection limit. Phosphate enhances the blank reaction a t any concentration over 5 ppm, and yet does not affect the Co(I1) CL signal until it reaches the 50-ppm level, a t which point the blank and Co(I1) signals are nearly equal. The reason for this behavior is uncertain. In natural water and biological samples, Ca, Mg, Fe, and Cu are the most likely metal interferences because of the high concentrations of these elements relative to Co. Standard techniques such as flame atomic absorption can be applied to the samples to evaluate if the concentrations of these metals are at their interference levels. All of the above metals except Mg cause an enhancement effect so that it is possible to correct the measured CL signal for the effects of these metals. The Mg interference is most critical because it depresses the blank and Co CL signal such that the Co detection limit is dramatically worsened in the presence of ppm levels of Mg. Since
2053
the concentration of Mg(I1) in natural water is typically in the 1-10 ppm range, a means of minimizing Mg(I1) in the Co(I1) reaction becomes highly desirable. Under the optimum reagent concentrations chosen, general masking agents for Mg(I1) such as P043-,P20j2-,and F- (18) were ineffective in masking the Mg(I1) depressive effect. As previously indicated, reagent concentrations were optimized to reduce the interference of Mg(I1) with the net result that the Mg(I1) interference can be only partially alleviated and the optimum reagent concentrations chosen are those which minimize the Mg interference to the largest extent.
CONCLUSIONS Although the Co(I1)-lophine system provides a detection limit of 0.1 ppb and linearity of two orders of magnitude, its application is limited by a few interferences and in particular the Mg(1I) depressive interference. As a consequence, in most samples where Co concentrations are not high enough to allow sample dilution to bring metal concentrations below their interference levels, the Co(I1) must be separated from the interfering species. A method has been described for the extraction of Co(I1) for subsequent lucigenin CL analysis ( 8 ) although the reproducibility of the method is only fair and the method is tedious. The lophine CL method provides a lower detection limit for Co(I1) than that reported for any other instrumental technique except CL ( 8 ) . The detection limit for Co(I1) is a factor of five better than achieved with the gallic acid and pyrogallol CL systems ( 8 , 9 ) and a factor of five and ten worse than provided by the lucigenin and luminol CL systems, respectively (8). In lophine and all of the above CL systems, the detection limit of Co(I1) is the lowest of all species tested. It is interesting to note that, besides Co(II), a low detection limit is achieved for Cr(V1) ( 5 ppb). The luminol system cannot even detect Cr(V1) ( I ) and the lucigenin system can detect it only in the ppm range (7). Reoptimization for Cr(V1) analysis may, therefore, lead to a CL system that can be used for determination of Cr(V1) in the presence of Cr(II1).
LITERATURE CITED (1) Seitz, W. R.; Hercules, D. M. "Chemiluminescence and Bioluminescence"; Cormier, M. J., Hercules, D. M., Lee, J., Eds.; Plenum Press: New York, 1973; pp 427-449. (2) Isaacsson, V.; Wettermark, G. Anal. Chim. Acta, 1974, 68, 339. (3) Neary, M. P.; Seitz, W. R. In "Anaiytical and Clinical Chemistry: A Series of Current Topics"; Hercules, D. M., Cram, S., Hieftje, G., Melville, R., Eds.; Plenum Press: New York, 1976; Vol. 1. (4) Seitz, W. R. "Modern Fluorescence Spectroscopy"; Wehry, E. L., Ed.; Plenum Press: New York, 1976; Vol. 1. (5) Paul, D. B. Talanta 1978, 25, 377-382. (6) "Nasa Special Publication, Vol. 388: Analytical Applications of Bioluminescence and Chemiluminescence", Chappelle, E. W., Piccolo, G. L., Eds.; NASA, Washington, D.C. 1975. (7) Montano, L. A.; Ingle, J. D., Jr. Anal. Chem. 1979, 5 7 , 919-926. (8) Montano, L. A,; Ingle, J. D., Jr. Anal. Chem. 1979, 57, 926-930. (9) Stieg, S.;Nieman. T. A. Anal. Chem. 1977, 49, 1322-25. (10) Miller, R. J.; Ingle, J. D., Jr. "Determination of Cobalt by Pyrogallol Chemiluminescence", unpublished work, Oregon State University, Corvallii, Ore., 1979. (11) Pidziszewski, Chem. Ber. 1877, 70, 70. (12) White, E.; Harding, M. Anal. Chem. 1964, 3 6 , 5686. (13) Philbrook, G.; Maxwell, M.; Taylor, R.; Totter, J. Photochem. Photobiol. 1965, 4 , 1175-83. (14) White, E.; Harding, M. Photochem. fhotobiol. 1985, 4 , 1129-55. (15) Chan, K., B.S. Thesis, University of Illinois, Urbana, Ill., 1977. (16) Hoyt, S.; Ingle, J. D., Jr. Anal. Chim. Acta 1978, 87, 163. (17) Burdo, J.; Seitz, W. R. Anal. Chem. 1975, 4 7 , 1639. (18) Perrin, D. "Masking and Demasking of Chemical Reactions", WileyInterscience: New York, 1979; pp 158-177.
RECEIVED for review February 12, 1979. Accepted May 23, 1979. Acknowledgement is made to the NSF (grant rCHE-76-167-11) for partial support of this research.
