2156
ANALYTICAL CHEMISTRY, VOL. 50,
NO. 14, DECEMBER 1978
generalized for peaks with line width parameters that vary with energy. For instance, Van Espen et al. ( 4 ) have suggested t h e form u* = k l + k2E for empirically determined constants k l and k2. Equations 8 and 9 can be easily modified with the relationships
and
n
$(ti) =
-
sk)
where $(ti) is a normalized standard profile a t ti. However, for these more complicated functions, the required computation time should be expected to increase dramatically because the a,* and (3j* convolutions will involve n times the number of operations while the number of overlap correction calculations will increase as n2. The spectrum deconvolution method outlined in this paper is an extended and more nearly exact treatment of the technique originally developed by Brouwer and Jansen ( I ) . Peak locations and integrated intensities can be determined to an arbitrary degree of precision with the rapidly converging iteration procedure described. Although the simple theory presented seems to be directly applicable only to rather special cases, some practical suggestions have been made toward generalizing the prescription for more complicated spectra.
(13) where again At = t , - t,. Although typical high-intensity line profiles are not well described by the simple Gaussian shape assumed above, the technique given here could nevertheless be applied in a straightforeward but somewhat more complex way if the spectral lines can be represented as a linear combination of Gaussian functions
C ak+oh(ti k = l
LITERATURE CITED (1) (2) (3) (4)
G. Brouwer and J. A. J. Jansen, Anal. Chem., 45, 2239 (1973). P. J. Stratham, X-Ray Spectrum, 5 , 16 (1976). A. Savitzky and M. J. E. Golay, Anal. Chem., 3 6 , 1627 (1964). P. Van Espen, H. Nullens, and F. Adams, Nucl. Instrum. Methods, 145, 579 (1977).
RECEIVED for review May 8, 1978. Accepted September 15, 1978.
Digitizer for Generating Computer Readable Data Michael F. Delaney and Peter C. Uden” Department of Chemistry, GRC Tower I, University of Massachusetts, Amherst. Massachusetts 0 1003
One of the major obstacles deterring chemists from utilizing computerized data analysis techniques is the transfer of the experimental information from the measuring instrument to the computer. Furthermore, the evaluation of the increasingly sophisticated numerical techniques proposed for chemical studies can be performed only with a valid, comprehensive set of computer-readable test data. Although the increased availability and competitiveness of computer compatible instrumentation will greatly facilitate the interconnection of instruments with computers, the typical situation is too often a t present analog t o digital conversion by ruler and data transfer tediously, by hand and keyboard. This note describes a semiautomated data digitizer configured with a X-Y servorecorder and a laboratory minicomputer. EXPERIMENTAL A schematic representation of the digitizer is shown in Figure 1. The only components of the X-Y servorecorder (Plotamatic 690, Data Equipment Corp.) utilized were the pen movement assembly (wires, pulleys, gears) and the ten-turn precision potentiometers connected to each axis. The latter were used in separate voltage divider circuits, powered by a PS501 power supply module in a Tektronix TM 501 mainframe. The voltage output of each divider circuit, proportional to the pen position, was connected to an analog input channel in the “Laboratory Peripheral System” (LPS) ( I ) of a PDP-ll/lOe minicomputer (Digital Equipment Corp.). Each input passed through an analog multiplexer to a programmable gain sample and hold amplifier to a 12-bit analog to digital (ADC) converter capable of a 2 0 - p s conversion time. Each axis encompasses a 5-V range in the voltage divider corresponding to 2048 ADC counts. Accounting for the length 0003-2700/76/0350-2156$01 .OO/O
Table I. Digitization of 2-Heptanol (IR.053)’
peak position, cm-‘
peak position, cm-’
manual computer manual computer 3658 3652 1245 1244 2965 2960 1136 1133 2930 2931 1090 1091 2869 1055 1052 287 5 1462 1460 950 947 1382 1378 Manual vs. computer digitization of a typical vapor phase infrared spectrum. of each axis, the digitization resolution is 80 points per cm in the horizontal axis and 110 points per cm in the vertical. A given set of data was digitized by placing a recorded spectrum, chromatogram or other analog trace, on the plotter bed and positioning the pen at significant points. Although a complete curve could be traced out with care using appropriate software, it was found more useful to manually position the pen accurately over selected data points and signal readiness to the computer by a push-button connected to the Schmitt trigger in the LPS. This signal caused the software to initiate an analog to digital conversion on the two analog channels and to store the data in memory. For convenience, the acquisition routine was written in assembly language but the data array processing was programmed in FORTRAN. RESULTS A N D DISCUSSION T o demonstrate the utility of this digitizer, a library of 500 vapor phase infrared spectra (VPIR), produced by Sadtler Research Laboratories, Inc. (2) were digitized in terms of peak position and intensity over the 2.5-15.0 pm range. T h e spectra
e 1978 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 50, NO. 14. DECEMBER 1978 -
-1
c Pd
PDP-11
I
i
L-
SPECTRU'd E G TI;EFi Figure 1. Schematic representation of the data digitizer The area wrthin the dotted line IS contained wrthin the Laboratory Peripheral System (LPS) of the minicomputer averaged 16.8 peaks each. Six calibration points were included with each spectrum (2.5, 5.0, and 15.0 pm, each a t 0 and 100% transmission) so that the peaks obtained in ADC counts could be converted to wavelength (or wave number) and percent transmission. These spectra were digitized a t the rate of about one spectrum per minute. An example of the effectiveness of this digitizer is shown in Table I which compares the results of computer vs. manual digitization for a typical spectrum. The correspondence is seen to be within six reciprocal centimeters (averaging 3 cm-') which compares well with the stated instrumental resolution of 4 cm-'. T e n replicate digitizations of four different peaks yielded spectral accuracies averaging within 1 cm-' and intensity
2157
accuracy averaging within 1 % transmission unit. For these measurements, the average standard deviation for the wavelengths was 2.6 cn1-l and 0.4% transmission unit for intensities. This corresponds to a physical displacement of about 0.2 mm. Once the computerized data set has been obtained, it can be tested for correctness and restructured or abstracted to fit the needs of the desired investigation. T h e VPIR library described herein is presently being used in several ways to study the effectiveness of vapor phase spectra as a n identification tool for gas chromatography, namely, by pattern recognition ( 3 ) ,by library searching (4), and by selection of optimal wavelengths for selective single wavelength monitoring.
LITERATURE CITED P. C. Uden, D. E. Henderson, and R. J. Lloyd, J . Cbromatogr., 126, 225 (1976). Specba 1-500 in "Atmospheric Pressure Vapor phase Infra-red Collection", Sadtler Research Laboratories, Inc., Philadelphia, Pa. M. F. Debney, P. C. Men, and D. E. Henderson, 29th PittsburghConference on Anavical Chemistry and Applied Spectroscopy, Cleveland, Ohio, 1978. Paper 478. M. F. Delaney, P. C. Denzer, and P. C. Uden, to be presented at the Fifth Annual Meeting of the Federation of Analytical Chemistry and Spectroscopy Societies, Boston, Mass., 1978.
RECEIVED for review July 17, 1978. .4ccepted September 5, 1978. The award of an American Chemical Society, Division of Analytical Chemistry summer fellowship, sponsored by the Olin Corporation is gratefully acknowledged (M.F.D.). The computer was obtained through National Science Foundation Grant G P 42542 to the University of Massachusetts.
Etching and Deactivating Glass Capillary Columns for Gas Chromatography Robert A. Heckman,' Charles R. Green, and Freddie W. Best Research Department, R. J. Reynolds Tobacco Company, Winston-Salem. North Carolina 27 102
T h e first step in preparing an efficient, wall-coated glass capillary column generally involves modification (usually etching) of the interior surface. This step is frequently required in column preparation for the promotion of film spreading and stability. T h e use of gaseous or aqueous H F for etching untreated glass capillary columns is not widespread. Aqueous HF does not impart a matte etch to borosilicate glass, and treatment of flint glass with either reagent is extremely difficult to control. One technique ( I ) applied to wide-bore columns relies on silica whisker formation resulting from attack of gaseous hydrogen fluoride which is generated at 350 "C from a deposit of ammonium hydrogen difluoride. Since 1974 we have used a similar reagent, potassium hydrogen difluoride, but have employed it under entirely different conditions and with significantly different results. The present etching technique employs aqueous K H F 2 solutions and has been used successfully in etching both borosilicate and flint glass capillary columns. Aqueous etching procedures have been shunned by some research groups (2) because of problems associated with deactivation of the etched layers. However, when properly eroded, the glass surface obtained by use of K H F z is already extensively deactivated relative to untreated glass. T h e procedure yields columns that are readily amenable to further deactivation by Cronin's adaptation (3) of Aue's method involving Carbowax 20M and silanization techniques similar 0003-2700/78/0350-2157$01.00/0
to the hexamethyldisilazane procedure of Welsch and coworkers ( 4 ) . T h e resulting columns have been statically and dynamically coated with a variety of polar and nonpolar stationary phases with excellent results; they have also been used extensively in several GC-MS systems. T h e etching of a flint glass column requires the passage of one column volume of 10% aqueous KHF,. This treatment is performed with a n 80-mL plastic-lined reservoir similar in construction to that described by Nikelly ( 5 ) . T h e development of a matte etch from the beginning to the end of the column can be followed visually and by a rapid change in the p H of the effluent from 6.5to 2.5. This is followed by passage of 1.5-2 mL of water per meter of column length. The water wash erodes the etched surface. I t can be followed visually and results in a uniform etch throughout the column. Both of these steps are conducted under 50-100 psig nitrogen pressure, depending on column length and diameter. Etching of columns made from borosilicate glasses requires a more concentrated K H F 2 solution (20%) and may require several treatments for development of a matte etch. Deactivation of the etched columns can generally be accomplished by the passage of 6 N hydrochloric acid. followed by sealing the ends of the nearly filled column with 6 N hydrochloric acid and heating overnight a t 100-150 "C. I t is sometimes advantageous to repeat this step. T h e columns are then rinsed with water and acetone and dried under a flow Q 1978 American Chemical Society
