Anal. Chem. 1980, 52, 349-351
TO PUMP
OBSERVATION ,,-WINDOW
-
- 9 cm DIAMETER OUARTZ TUBE M ON 0CH R O MATOR r AhD DETECTOR
{M-RING INJECTOR RESISTIVELY HEATEO TANTALUM BOAT
MICROWAVE CAVITY
[sample s o m i I
N 2 in
1
Ar in Figure 1. MTES apparatus with sample source adapted for liquid
samples
@ 0 B
s
9
F
0
349
Care must be taken to keep the system as leak tight as possible since N O y-system fluorescence is produced by the reaction of atmospheric oxygen and water vapor with active nitrogen. Should this become a problem, however, ancillary P b lines at 280.2 and 405.8 nm are suitable for lead detection by MTES. Recent data from a redesigned apparatus have extended our range in both directions. The new system differs mainly in having a 4-cm diameter quartz viewing tube (Figure 1) with the ring injector built into the wall. Operating a t a higher pressure of 4.2 Torr, this design concentrates the sample fluorescence, provides some light baffling, and does not mask or block the effusion of lead vapor from the boat. With this modification, our system response to lead content in the sample is linear from 1.0 to 20000 ng. A least-squares fit to our new data over this range gives a slope of 0.96 with a coefficient of determination of 1.0. Our minimum detection limit is 0.2 ng (20 FL of 10 ppb solution). The measurement precision is also improving. Ten successive determinations of a solution containing 10 ppm lead were found to have a standard deviation equal to 10% of the mean value. In order to test our system on a "real" sample, it was decided to look at the accumulation of lead emitted from automobiles. Five leaves of similar size and shape weighing 4.8 g were taken from a eucalyptus tree adjacent to the San Diego Freeway near the Aerospace Corporation. These leaves were washed with 5.0 mL of the same 0.3 N HN03 solution used for diluting the 1000 ppm standard. By the method of standard additions, the concentration of lead in the washing solution was determined to be 22 ppm using the 283.3-nm line. The 405.8-nm line was also used to verify the signal profile. Thus the average 1-g leaf carries on its surface the equivalent of 23 pg of lead. This is particularly interesting since there had been a rainstorm only three days prior to our collection of the leaves. LITERATURE CITED
Flgure 2. Response of the MTES system to nanograms of lead contained in solution. Note the characteristic flattening of the response curve as saturation is approached
of lead content in the sample. If one excludes the points obtained with 1, 2, and 20000 ng of lead, a least-squares fitting procedure on the remaining points yields both a slope and coefficient of determination of 0.99. In other words, the response is linear from 10 to 10 000 ng. Eventually a flattening of the response curve is observed ( 4 , 5 ) as saturation is approached. The response to samples containing less than 10 ng of lead deviates in the positive direction from linearity. We believe this is due to background interference from the 285.8-nm band of the N O y-system. This background is currently limiting our sensitivity although dark current fluctuations become significant in comparison to signal with samples containing less than 0.5 ng lead.
(1) U S . Patent No. 4 150951 (April 24, 1979). (2) G. A. Capelle and D. G. Sutton, Appl. Phys. Lett., 30, 407 (1977). (3) G. A. Capelle and D. G. Sutton, Rev. Sci. Insfrum., 49, 1124 (1978). (4) D. G. Sutton, J. E. Melzer, and G. A. Capelle, Anal. Chem., 50, 1247 (1978). (5) D.G. Sutton, K. R. Westberg, and J. E.Melzer, Anal. Chem., 51, 1399 (1979).
J. E. Melzer' J. L. Jordan D. G . Sutton Aerophysics Laboratory Laboratory Operations The Aerospace Corporation El Segundo, California 90245 RECEIVED for review August 23, 1970. Accepted November 5, 1979. This work was funded by the Aerospace Sponsored Research Program which the authors gratefully acknowledge.
Speciation and Elemental Analysis of Mixtures by High Performance Liquid Chromatography with Inductively Coupled Argon Plasma Emission Spectrometric Detection , Sir: Chromatography is a primary method for speciation and analysis of multicomponent mixtures. For complete resolution or characterization of species, detectors responsive solely to the element(s) of interest have been devised to sim0003-2700/80/0352-0349$01 .OO/O
plify both qualitative and quantitative requirements (1-5). If the detector has multielement capability, one can obtain a multielement chromatogram with the elemental composition of the resolved components. Microwave plasma has already 0 1980 American Chemical Society
350
ANALYTICAL CHEMISTRY, VOL. 52, NO. 2, FEBRUARY 1980
t
Vitamin Biz 36'
H90Co
i!
N14°14
Zn
.................................................... . . . .,,-............:... ................. . .
p
..............................
_. .
:...................
....
..... ...... co ._'
n
_ _ _I_ 10 m
c
i....
I! I1 ~
30
40
50 Min
Mn
............
:.-- . . . . .-
...................
,
. . .
Fe
Figure 1. Mukielement chromatogram of vitamin B12;75 Kg of vitamin B12was injected
been applied for the multielement detector in gas chromatography (6-8). High performance liquid chromatography (HPLC) has many advantages over gas chromatography. Nonvolatile, polar, thermally unstable or high molecular weight compounds can be separated. As a detector for HPLC, inductively coupled plasma atomic emission spectrometry (ICP-AES) is particularly suitable because it has simultaneous multielemental capability for major to ultra trace levels, and interelement interference effects are considerably less than those observed in atomic absorption spectrometry (9). An approach using dc argon plasma as an HPLC detector has recently been reported by Uden et al. (IO). The gel permeation chromatography is chosen here to demonstrate the suitability of the ICP-AES-HPLC system. The high sensitivity of ICPAES to metal atoms is particularly suitable in the analysis of metalloproteins. In this study the outlet of an HPLC (Toyo 802) was directly connected to the nebulizer of an argon plasma atomic emission spectrometer (Jarrell-Ash Atomcomp). The nebulizer was a conventional cross flow nebulizer. Gel permeation chromatography was performed under the following conditions: column, TSK GEL 3000 SW (600 mm X 2); elutant, 0.9 7'0 NaCl aqueous solution; flow rate, 1.0 mL/min; and temperature, room temperature. The recently developed gel columns, Toyo Soda TSK-Gel series, have shown good performance for the separation of proteins in a short analytical time although the column substrate is not reported. The plasma was operated at 1.25 kW radiofrequency power, with 1,0.5,and 18 L/min. of argon gas for the plasma, sample, and coolant gas flow, respectively. The elements (and the respective wavelength) programmed in the polychromator were Co (228.6 nm), Cu (324.9 nm), Fe (259.9 nm), Mn (257.6 nm), P (241.9 nm), and Zn (213.8 nm). Carbon (247.6 nm) was also determined simultaneously with the monochromator attachment. Emission intensities were measured for a burning time of 10 s. The intensities were converted to concentration using the calibration curve which was obtained by pumping known concentrations of metal solutions or phosphate solution with or without. glycylglycine (carbon standard) into the nebulizer at the same flow rate. Emission intensities for the elements showed wide dynamic range and were not interfered by carbon present in the media up to 4000 ppm. Spectral interference by other elements was negligibly small. A demonstration of the operation of the system is given for vitamin Biz. The multielement chromatogram gave an apparent peak in C, Co, and P chromatograms (Figure 1). From the peak area in the C, Co, and P chromatograms, the amount of these elements in vitamin B12was estimated. The relative
cu
, .
: ........................................._.. .....
._..............
C
Flgure 2. Multielement chromatogram of proteins; 20 pL of mixed protein solution (0.5% for each protein) was injected
emission intensity per gram-atom was 10690/528/1 (Co/P/C) and the relative peak area of vitamin B12was 166.5/7.65/1 (Co/P/C). Thus, the atom number ratio was calculated to be Cs4,2Po,93C01, which is close to the theoretical value of CS3PCo. Vitamin BI2was not recovered completely in this experiment probably owing to the adsorption on stainless tubing. Therefore, it was impossible to decide the absolute amount of each element. The protein kit for molecular weight calibration (Combiteck) was also applied to the system (Figure 2). Proteins are known to elute in the order of molecular weight and to provide quantitative recovery. The highest molecular weight protein, ferritin, and the second largest protein, catalase, were not resolved. Both proteins are iron-containing proteins which exhibit a large peak for the iron chromatogram. It is interesting that divalent metals and phosphorus are also observed at this position. These elements were contained in the ferritin we used in this experiment. The protein eluted last, cytochrome c from horse heart, was also checked quantitatively. Assuming the complete recovery of injected cytochrome c (100 hg), the carbon and iron content was calculated. The observed values of 0.47% and 55%, respectively, for iron and carbon were close to the theoretical values of 0.44% and 55%, respectively. The present method provides information on molecular weight, spectrometric characterization along with elemental composition. Thus it is expected to have wide application
Anal. Chem. 1980,
especially in the field of biochemistry because many enzymes have metal atoms in their functional sites. ACKNOWLEDGMENT The authors express their thanks to C. MacLeod and K. Okamoto for their encouraging discussions. LITERATURE CITED (1) Kolb, 6.; Kernrnner, G.; Schleser, F.; Wiedeking, E. Fresenius’ Z.Anal. Chem. 1988, 227, 166-75. (2) Segar, D. A. Anal. Lett. 1974, 7, 89-95. (3) Botre, C.; Cacace. F.; Cozzani, R. Anal. Lett. 1978, 9 , 825-30. (4) Fernandez. F. J. Chromafogr. News/. 1977, 5, 17-21. (5) Koizurni, H.; Hideishi, T.; McLaughiin, R. Anal. Chem. 1978, 50, 1970-71. (6) Bache, C. A.; Liske, D. J. Anal. Chem. 1987, 39, 786-89.
52,351-352
351
(7) McLean, W. R.; Stanton, D. L.: Penketh, G.E. Analyst (London) 1973, 98,432-42. (8) Quimby, B. D.; Uden, P. C.; Barnes, 13. M. Anal. Chem. 1978, 50, 2112-18. (9) Fassel, V. A. Science 1978, 202, 183-90. (10) Uden, P. C.; Quimby, 6. D.; Barnes, R. M.; Elliot, W. G.Anal. Chlm. Acta 1978, 701, 99-109.
Masatoshi Morita* Takashi Uehiro Keiichiro Fuwa National Institute for Environmental Studies Yatabe, Tsukuba, Ibaragi, 300-21 Japan
RECEIVED for review June 19, 1979. Accepted October 19, 1979.
Comments on Critical Evaluation of Curve Fitting in Infrared Spectrometry Sir: Vandeginste and De Galan ( I ) have proposed objective criteria for the resolvability of overlapping infrared absorption bands, based on the separation between the origins of a pair of bands. As the band separation decreases the first critical separation, termed the “shoulder limit”, is reached, defined when the experimental curve y has a critical point at which D y = D2y = 0, and illustrated in Figure l a . As the bands are brought closer together the second critical separation, termed the “detection limit” is reached, defined by the existence of a critical point at which D2y = D3y = 0, and illustrated in Figure lb. In their submission, their infrared absorption curves could be “reliably” resolved only if the bands were further apart than the second critical separation. We have examined the effect of the signal-to-noise ratio, S I N , of the data on resolvability and have found that the minimum band separation at which acceptable resolutions can be obtained decreases as S I N increases. To illustrate this point, we have synthesized a number of realistic curves of known composition and have subjected them to curve resolution (2) in order to compare calculated parameter values with known values. The curves consisted of three overlapping Lorenzian bands of equal half-width, with intensities of 200, 100, and 200 arbitrary units. Adjacent bands were separated by the same distance s which is conveniently expressed in terms of the band half-width (fwhh). An experimentally derived spectrum of noise with a slightly nonrandom distribution, rms 2.25 units, was superimposed on the sum of Lorenzians. A second set of curves was synthesized in the same way but with one fifth of the noise level, rms = 0.45 unit. Thus, the S I N ratio, defined as maximum signallrms noise, was approximately 100 and 500 in the two sets of curves. A typical curve from the first set, its resolution, and the noise are shown in Figure 2. When the band parameters are known as in the present case or as when spectra are obtained from mixtures of known composition, the reliability of a resolution may be expressed in terms of the “discrepancy ratio”. This is the ratio of the discrepancy between the value of a parameter calculated by curve resolution and its value in the synthesized curve and the standard deviation of that parameter as calculated by the principle and method of least squares ( 3 ) . Some computed discrepancy ratios, which are of course scale-independent, are given in Tables I and 11. The correlation coefficients between height and width were greater than 0.9 for s less than 0.75 0003-2700/80/0352-035l$Ol .OO/O
half-widths, and so were included in the calculation of errors in band areas by error propagation ( 3 ) ,as the correlation term may contribute up to half the total relative variance in band area. The results obtained at SIN E 100 show that the estimated values of the band parameters position, height, width, and area are on average within ca. three standard deviations of the values used to synthesize the spectra from s = 1.15 to s = 0.45 half-widths, which corresponds to the “resolution limit”. With separations less than 0.45 half-width some discrepancy ratios such as those of the height, width, and area of the middle band become increasingly large. This means that not only are the computed parameter values in error but also, more significantly, the errors themselves are seriously underestimated. We conclude that for separations greater than the second critical value, the curves have been reliably resolved if we estimate the errors as three standard deviations. This is in agreement with the results of Vandeginste and De Galan obtained with infrared spectra. However, for s less than 0.45, the resolutions are unreliable in the sense that estimates of errors are unreliable. At S I N E 500 we conclude that the curves with s greater than 0.25 half-width have been reliably resolved if we estimate the errors in band parameters as four standard deviations. This margin is smaller in absolute terms than in the high noise case since the standard deviations are some 3-4 times larger in that case. Thus reliable resolutions have been achieved with separations much less than the “resolution limit”; s = 0.25 corresponds to the fifth critical separation as defined below. We suggest that one reason why standard deviations may be underestimates of error is that they reflect only random errors in the data, and that systematic errors must inevitably be present locally in a spectrum even if overall the errors follow an appropriate random distribution. Further, the errors are based on the approximation that the model is linear in the region of best fit ( 3 ) ;the degree of nonlinearity is uncertain, though probably small. Other systematic errors may derive from the model, if the base line must be parameterized (base-line height was optimized in our model calculations) or the shape-function of the bands is inaccurate as for example, when the bands are slightly asymmetric ( I ) . The effects of finite slit-widths can in principle be eliminated by deconvolution ( 4 ) . If these sources of systematic errors can be eliminated, our results show 1980 American Chemical Society
