1584
ANALYTICAL CHEMISTRY, VOL. 50, NO. 11, SEPTEMBER 1978
r
W +
5
400
J
W
>
9 z
P
300
400
500
600
EMISSION WAVELENGTH, nm
Figure 1. Fluorogram of 7-hydroxybenzo[alpyrene (0.2 ng/mL), adapted from ref. 2. Curves A and B are the conventional excitation and emission spectra, respectively. Curve C is the spectrum obtained by a synchronous scan with AA = 50 nrn. Curve C is a plot of the emission intensities vs. ,A, measured along the dashed diagonal line. The contours are 10% increments of the maximum emission intensity. Curve A is the profile section of the three-dimensional surface taken at ,A, = 440 nm. Curve B is a similar profile section taken at A, = 390 nm. Curve C is the profile section taken along the dashed line and projected onto the emission wavelength axis
would be generated by a synchronous scan where AA = ,A, - ,X , = (440 - 390) = 50 nm, the wavelength difference between the maxima of the conventional excitation and emission spectra. The spectrum labeled C is the synchronous spectrum obtained by plotting the variation of emission intensity along the diagonal dashed line vs. the emission wavelength. A synchronous measurement made according to the method of Vo-Dinh ( 1 ) will give spectrum C if AA is 50 nm.
Vo-Dinh’s experimental arrangement can produce the three-dimensional contour plot of Rho and Stuart by making repeated scans where AA is changed in fixed increments. Rho and Stuart‘s method can give Vo-Dinh’s results by the technique described above. Vo-Dinh’s experiment is simpler and is considerably faster if only a single scan is necessary, but it discards much information that is retained in the three-dimensional plot. The main difficulty with the synchronous scan method is that the best value for AA must be known beforehand for optimum results. Also, in some multicomponent systems, several different values of AA might be necessary for complete identification. The best values for AA can be determined easily from a three-dimensional plot, such as Figure 1. A threedimensional plot also shows whether it would be possible, in a synchronous scan, to suppress some components and observe only selected components of mixture. On the other hand, the three-dimensional plotting method, by itself, will often be more complicated than necessary for specific analytical purposes and it requires interpolation between contours for quantitative results. It appears that the simplest adaptation of these two methods would serve many analytical applications. This would be to use the synchronous scan method, but by first generating a three-dimensional plot to completely characterize the sample spectrum. We have found that a crude threedimensional plot can be generated in 2 or 3 h by hand plotting, if we wish only to locate the “mountain peak” positions without quantitative details of the contour shapes. In complicated systems, these crude plots have been completely adequate for determining the most useful values of AX for subsequently more rapid and quantitative analysis. LITERATURE CITED (1) T. Vo-Dinh, Anal. Chem., 5 0 , 396 (1978). ( 2 ) J. H. Rho and J. L. Stuart, Anal. Chem., 5 0 , 620 (1978)
Eugene R. Weiner Department of Chemistry University of Denver Denver, Colorado 80208 RECEIVED for review May 1, 1978. Accepted June 22, 1978.
Comments on Determination of Fluorine in Petroleum and Petroleum Process Catalysts with a Fluoride Electrode Sir: In 1973 we published a paper describing the determination of fluorine in petroleum and related materials which included analytical results for the fluorine contents of ten crude oils indicating levels, with one exception. substantially below 1 mg/kg ( 2 ) . Subsequently, data from a cooperative exercise presented by other workers ( 2 ) suggested that even the low levels we had measured might be erroneously high. The toxic effects of fluorine and its compounds can be very severe, and we considered it important to resolve this difference. In outline, our procedure consists of C-F bond cleavage under nonaqueous conditions with sodium biphenyl. followed by aqueous extraction of NaF and measurement of fluoride ion concentration with a solid state fluoride electrode. Because of the strong tendency of crude oils and residues to form stable emulsions during the extraction process, the size of sample which could be taken was limited to a maximum of 2 g. 0003-2700/78/0350-1584$01 .OO/O
However, we have found that addition of a cationic detergent such as methyl-tricaprylyl ammonium chloride. (Aliquat 336, General Mills Chemicals Inc., Des Plaines, Ill.), immediately before extraction prevents emulsion formation. This not only gives a cleaner separation but also permits the use of much larger samples (in practice 20-25 g.). The crude oils which we originally examined were Alaskan (North Slope), Arabian Light and Heavy, Iranian Light and Heavy, Iraq (Basra) and Iraq (Med), Kuwait, Libyan (Sarir), and Nigerian (Forcados). Re-examination of these materials by the modified procedure has established fluorine levels for these crude oils to be below 0.01 mg/kg, and probably below 0.005 mg/kg. These figures are in complete agreement with the results of the cooperative program (2). We have concluded that our original results were high because of interference with the electrode response by traces of entrained organic material. These new findings do not affect results reported in our 8 1978 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 50, NO. 11, SEPTEMBER 1978
paper for other materials, including petroleum products and petroleum process catalysts. LITERATURE CITED
1585
The British Petroleum Company Limited BP Research Centre Chertsey Road Sunbury-on-Thames, Middlesex TW16 7LN
(1) J. N. Wilson and C . Z. Marczewski, Anal. Chem., 4 5 , 2409 (1973). (2) Ernst Hohmann et al., €rdolKb;hle, 26, 647 (1973).
J. N. Wilson* C. Z. M a r c z e w s k i
RECEIVED for review April 26,1978. Accepted May 30,1978.
Exchange of Comments: Analysis of Complex Volatile Mixtures by a Combined Gas Chromatography-Mass Spectrometry-Computer System Sir: We are writing to correct several errors in the paper of Gates et al. (I)concerning previously published results from our laboratory ( 2 , 3 ) . In a detailed discussion of our work in the above paper, the authors seek to compare our methods with their own. During the course of this discussion, and in other places in their otherwise excellent paper, erroneous statements are made which in our judgment reflect badly on both laboratories and will result in considerable confusion in the minds of readers of the paper. First, a very important comparison is never explicitly stated in their paper. Their system ( I ) is designed specifically to quantitate previously observed compounds and is applicable to mixtures where the only interest is quantitation of these constituents. Our efforts (2, 3) have been directed toward the more general problem of detecting and quantitating both previously observed and new compounds. Both methods have their advantages and disadvantages. These are confused, however, by the discussions of Gates et al. (I). Second, we wish to correct the following specific points: (1)None of our methods utilize reverse library search ( I , p 439 and p 440). The HISLIB (3) program utilizes “forward” library search and comparison in all phases of operation. (2) T h e CLEANUP (2) program applies its doublet resolver recursively to deconvolute multiplets (2, p 1372). Although one might criticize this approximation, the program is not now, nor has it ever been, limited to simple doublets ( I , p 437 and
p 440) and routinely handles complex multiplets with limitations similar to those discussed by Gates et al. for their own procedures. (3) Our work makes explicit use of internal standards (3, p 1623 (abstract) and throughout the remainder of paper), not external standards as alleged ( I , p 440). (4) Although we have not published data on the accuracy, sensitivity, and linear range of our methods, we have published data on its precision (3).
Sir: In a recent paper ( I ) we made erroneous statements about the techniques developed at Stanford University (2,3) for the analysis of complex volatile mixtures by a combined gas chromatography-mass spectrometry-computer system. As Rindfleisch, Smith, and Yeager have indicated above, the HISLIB program (3)uses “forward” library search algorithms rather than “reverse” library search and quantitation is based on internal standards ( 3 ) ,not external standards. Furthermore, Smith et al. (3) have reported percentage standard deviation of their procedure in the analysis of relative concentrations of reference hydrocarbons and of urinary organic acids obtained either by organic solvent extraction or by anion-exchange chromatography. We regret that these errors have appeared in our discussion of the Stanford system, which is an alternative approach to our MSSMET system for quantitative metabolic profiling. I t does have several important advantages that are not yet imp!emented in MSSMET, including detection and quantitation of previously unobserved (novel) compounds. We certainly did not intend to denigrate the excellent work (2, 3) by this group.
The objection by Rindfleisch, Smith, and Yeager about the capability of their CLEANUP program (2) to deconvolute multiplets needs to be clarified. Our opinion, after having read their paper (2),was that several points were not clearly presented and documentation was minimal. In retrospect, we should not have said (1) that “their system can resolve doublets but not multiple overlapped mixtures of substances” because it was explicitly stated (2, p 1372) that sequential use of the doublet resolver can be used for the multiplet case. They go on to point out, however, that the full procedure has not been implemented beyond the doublet case (2, p 1372), and that the procedure may not perform very well with complex multiplets, in terms of producing accurate amplitude information (2, p 1374). We interpreted these statements to mean t h a t they were not running the system routinely on complex multiplets and had not yet documented a capability to deal with such mixtures effectively. The analysis of multiplets by MSSMET depends 011 the choice of a single, unique designate ion, rather than a cleaned-up spectrum, for each substance in the multiplet and has, indeed, been shown
0003-2700/78/0350-1585$01 .OO/O
LITERATURE CITED (1) S. C. Gates, M. J. Smisko, C. L. Ashendei, N. D. Young, J. F. Holland, and C. C. Sweeley, Anal. Chem., 5 0 , 433 (1978). (2) R. G. Drorney, M. J. Stefik, T. C. Rindfleisch, and A. M. Duffield, Anal. Chem.. 48, 1368 (1976). (3) D. H. Smith, M. Achenbach, W. J. Yeager, P. J. Anderson, W. L. Fitch, and T. C. Rindfieisch, Anal. Chem.. 49, 1623 (1977).
D e n n i s H. S m i t h * William J. Yeager T h o m a s C. Rindfleisch Department of Genetics Stanford University School of Medicine Stanford, California 94305
RECEIVED for review April 3, 1978. Accepted May 26, 1978.
1978 American Chemical Society
