Spark Spectroscopy Using Charge Transfer Devices: Analysis, Automated Systems, and Imaging
R O B E R T S. P O M E R O Y , R A F I D. J A L K I A N , and M . B. D E N T O N * Department of Chemistry, University of Arizona, Tucson, Arizona 85721
An atomic emission spectroscopic system utilizing a spark source for excitation has been developed. The instrument employs a custom echelle spectrometer and a charge injection device (CID) array detector system. This system simultaneously covers wavelengths from 200 to 450 nm with a resolution of 0.02 nm at 300 nm. Solids sample analyses of steels and aluminums were used to demonstrate this system's speed, sensitivity, and flexibility. Automated systems for rapid qualitative and semi-quan- titative screening of these materials will also be discussed. Another spectroscopic system based on a commercial imaging spectrograph and a charge-coupled device (CCD) array detector has been used to obtain temporally resolved spectral images of single sparks discharges.
Spectrochemical analysis is considered to be the most generally applicable method of instrumental analysis in a variety of fields such as agriculture, astronomy, biology, chemistry, forensic science, geology, medicine, metallur- gy, physics, etc. The wide applicability enjoyed by spec- trochemical analysis is due to the simple, rapid, and eas- ily au toma ted na ture of these methods . Although spectrochemical analysis may not meet all analytical needs in practice, it generally provides satisfactory ac- curacy, precision, and limits of detection for both metals and nonmetals at a low cost per analysis.
In the analysis of solid samples, there are two distinct advantages to performing direct analysis on the solid sample: (1) minimal sample preparation is required, and (2) potential contamination from the reagents used in the dissolution process is avoided. The two most common optical emission techniques for direct solids analysis are arc and spark emission spectroscopy. These techniques can be applied to over 70 elements; they demonstrate a high degree of specificity as well as adequate sensitivity for most applications, and, with proper matrix-matching, interelement interferences can be minimized. Although the initial investment for the instrument can be expen- sive, the high sample throughput, and hence low cost per analysis, combined with the above characteristics ac- counts for the widespread use of these techniques in industry.
The most important drawback associated with arc and spark analysis is in the acquisition and interpretation of the spectrum. Instruments have employed either pho- tographic film or banks of photomultiplier tubes (PMTs) for the detection of the emitted radiation.
Received 14 February 1991. * Author to whom correspondence should be sent.
A substantial length of time is required in order to develop and read photographic film, and it suffers from such problems as nonzero film fog, nonlinear response, limited linear dynamic range, and reciprocity limitations. Photomultiplier tube (PMT) detection is a more trac- table means of signal acquisition and overcomes the above-mentioned problems with photographic film. The drawback with P M T detection is the device's inherent single-channel nature. In order to monitor several emis- sion lines simultaneously, a P M T needs to be dedicated to each spectral feature. This setup entails a significant expense and has a practical limit as to the number of PMTs that can be placed at the focal plane of a spec- trometer. What is desirable is a detector system that would combine the continuous wavelength coverage af- forded by photographic film with the sensitive, electronic output of a PMT.
The development of a custom echelle spectrometer with charge injection device (CID) array detection car- ried out in these laboratories should be particularly well suited for arc and spark emission spectroscopy. CIDs exhibit many of the best characteristics of photographic film and P M T detection while providing the added ad- vantage of nondestructive readout and Random Access Integration (RAI). Further details on the performance characteristics, operation, and application of CIDs are available. 1-~
The echelle/CID spectrometer provides simultaneous acquisition of all of the spectral information over a wide wavelength region with adequate resolution for atomic spectroscopy. This allows the analyst to (1) use multiple lines for comparative analysis; (2) monitor the back- ground for changes in excitation, thereby indicating po- tential problems with the analysis; and (3) determine in the course of the analysis the optimum lines to be used for quantification. 7,8 Effective utilization of the large da- tabase of spectral information has led to the development of expert systems such as an automated qualitative and a semi-quantitative analysis routine.
Spatially and spectrally resolved images of single spark discharges have been previously obtained with the use of a charge-coupled device (CCD) detector and an ab- erration-corrected imaging spectrograph. Acquisition of temporally resolved spectral images is desirable in order to provide insight into the dynamics of the spark dis- charge. Temporal resolution can be accomplished by three technologies: (1) a rotating slot or mirror, (2) a pockel cell as an optical shutter, and (3) fast spectral framing with a CCD. The first method suffers from limited tem- poral bandwidth and/or mechanical complexity. The sec- ond approach is limited in the spectral region in which opto-electronic shutters can operate. Typically opto-
Standard Elements present Elements detected Elements suspected Not detected
SS 406/1 Fe, C, Si, Mn, P, S, Cr Fe, C, Si, Mn, Cr, Mo, Ni P--0.009% Mo, Ni, Co, Cu, V Cu, V Co--0.006 % S--No database
SS 407/1 Fe, C, Si, Mn, P, S, Cr Fe, C, Si, Mn, Cr, Mo, Ni P--0.030% S--No database Mo, Ni, Cu, V Cu, V
SS 408/1 Fe, C, Si, Mn, P, S, Cr Fe, C, Si, Mn, Cr, Mo, Ni P--0.037% S--No database Mo, Ni, Cu, V Cu, V
Co--0.014 To SS 409/1 Fe, C, Si, Mn, P, S, Cr Fe, C, Si, Mn, Cr, Mo, Ni P--0.025% S--No database
Mo, Ni, Co, Cu, V Cu, V SS 410/1 Fe, C, Si, Mn, P, S, Cr Fe, C, Si, Mn, Cr, Mo, Ni P--0.072 % S--No database
Mo, Ni, Cu, V Cu, V
electronic shutters have throughput only in the visible region of the spectrum and hence gather no information in the UV, where a number of interesting transitions occur. Preliminary results from the use of an imaging spectrograph and a thinned, backside-illuminated CCD detector (improved UV response) operated in the fast spectral framing mode will be presented.
EXPERIMENTAL
The echelle/CID spectroscopic system, with a few al- terations, has been described elsewhere. 7 The optics were stripped of an anti-reflection coating that severely at- tenuated the light throughput below 300 nm, and the prism was recoated with MgF2. The source utilized in these studies was a Baird KH-5 unidirectional spark source (Baird, Milford, MA). The source delivers a peak voltage of 950 V with a peak current of 275 A for 250 ms at a repetition rate of 120 Hz in the preburn mode and a peak voltage of 950 V with a peak current of 100 A for 130 ms at a repetition rate of 120 Hz in the expose mode. The sample serves as the cathode, and the anode is a pure silver electrode. The analytical gap was set at 5 mm, for all analyses, with a gap-setting tool provided by the manufacturer. The stand was flushed with argon before each use, and a constant back pressure supplied with an exhaust trap recommended by the manufacturer. It is necessary to protect the computer system from the radio- frequency interference (RFI) generated by the spark source. This was accomplished by mounting the com- puter into a rack with copper screening covering all ex- posed portions (a Faraday cage). All wiring to and from the computer was shielded and passed through filters. The samples were prepared with #60 A1203 paper or turned down on a lathe. The samples were subjected to a 25-s preburn prior to analysis. Analysis of the steel samples was performed at the full expose-mode power; the aluminum samples were analyzed at half expose- mode power. The algorithms for the quantitative anal-
ysis, automated qualitative analysis, and semi-quanti- tative analysis have been described. 9,1°
The experimental apparatus used to obtain spectrally resolved images has been previously describedY ,12 For the temporally resolved images, the following changes to the above system were made: The detector, an RCA 501-- a thinned backside-illuminated device--was operated in the fast spectral framing mode 13 by masking all but 4 rows of the CCD. The dispersive device used is an ISA CP-200 (Jobin Yvon/Instruments SA), and the spark source had a modification to the trigger circuit on the ignition card so it could operate in a single spark mode.
RESULTS AND DISCUSSION
Auto-Qua l i ta t ive A n a l y s i s . The flow chart and the analysis scheme are described elsewhere? ° Results of the auto-qualitative analysis of steels are shown in Table I. Note that there is a good correlation between the re- ported and found elements and no false positives. Sus- pected elements are simply present at low concentrations so that fewer than half of the emissions lines in the database exhibit an S/N _>5. Those elements not de- tected were present at a concentration lower than the limit of detection (LOD), or no database is currently present for that element. Similar results were obtained for the auto-qualitative analysis of aluminums and are reported in Table II. Again there is good agreement be- tween the known and the reported elements present.
The sensitivity of the auto-qualitative analysis routine is demonstrated by using steel standards with decreasing amounts of aluminum. The results are shown in Table III. For steel SRMs 1262a and 1263a, where the concen- trations are well above the detection limit, aluminum was reported as present. In SRM 1261a, where the alu- minum concentration is above the LOD on some of the emission lines but not enough to ensure its presence, aluminum is reported as suspected. For SRMs 1264a and 1265a, no aluminum was detected by the auto-qualitative
TABLE II. Auto-qual analysis of aluminums.
Standard Elements detected Elements suspected Elements missed Ni--0.003 %
SS-1100-AN AI, Mg, Si, Fe, Cu, Zn, Ti Cr, Mn Ga--No database SS-2011-X A1, Mg, Si, Fe, Cu, Cr, Zn Mn, Cr, Ni Ti--~.005% SS-2618-F A1, Mg, Si, Fe, Cu, Ni Mn, Zn, Ti Cr-~.001%
TABLE III. Sensitivity of auto-qual routine: Detectability of AI in SRMs.
SRM A1 concentration, % Detected (Y/N)
1261a 0.02 Suspect 1262a 0.09 Present 1263a 0.24 Present 1264a None Not found 1265a 0.0007 Not found
routine, in the case of 1264a, no aluminum was actually present, and no false positive was reported. In 1265a, the concentration of aluminum was below the S/N >_5 threshold limit on all A1 emission lines and was hence listed as not present.
The sensitive and flexible qualitative analysis of metals provides a rapid and sure method for screening or cursory evaluation of samples. The multi-line approach effec- tively eliminates the occurrence of false positives gen- erated from spectral interferences. This algorithm has been in use in our laboratories for over a year and has been successful in analyses ranging from investigation of geological samples to identification of unlabeled labo- ratory waste. Where this routine has failed is in identi- fying elements whose strong emission lines occur outside the wavelength region covered by the system (S, As, Na, Cs, K) or are in a region where the spectrometer through- put or detector quantum efficiency is low (Cd, P). New spectrometer design and UV spectral response enhance- ment with a fluorescent down-converter should address these problems. 14
Semi-Quant i tat ive Analys is . The next logical step in automated analysis would be to use the auto-qualitative algorithm to identify the elements present and then em- ploy a procedure to obtain semi-quantitative analysis. The semi-quantitative approach chosen was to use an internal standard. This method corrects for changes in excitation and has been extensively used in the past for the spark analysis of alloys. In the two materials tested, steels and aluminums, the major matrix components, iron and aluminum, are used as the internal standard. A wealth of information about internal standards and ap- propriate emission lines to be ratioed is available. 15,16
7000
.x
~ t • .
0 i i i i i i i 0 . 0 O . S
C o n c e n ~ P a t J o n o f Cr. X
FIG. 1. Working curve for Cr in a steel matrix with the use of Fe as the internal standard.
I I I I I I ] I i i } I I I I } i I I I I I I I i
g / 3 0 5 . 0 A I
m L : ~ t 0 4
t 03 I t I I I I I I I I I I t t I I I 1 t I I I I I I
i0-2 I0-2 iO o iO i
C o n c e n t r a t i o n o f t4g. X
FIG. 2. Working curve for Mg in an aluminum matrix with the use of A1 as the internal standard.
Shown in Fig. I is a working curve for chromium in steel where the Fe concentration is constant to within two tenths of a percent. Since the iron concentration is so consistent, Fig. 1 plots the emission ratio Cr/Fe vs. chro- mium concentration. Similarly, Fig. 2 shows the working curve for magnesium in aluminum alloys where, once again, the aluminum concentration was nearly constant. As shown, the linearity is excellent and hence should provide acceptable semi-quantitative results. A test of the effectiveness of the semi-quantitative analysis rou- tine is shown in Fig. 3. First, the chromium content in a low-alloy steel was measured by ten replicate analyses; this value is represented by the solid line at 0.055% Cr. Over the next 18 days, the sample was run and the chro- mium content determined from the slope of the emission ratio working curve and displayed on the chart. No stan- dards were run, and only routine calibration of the sys- tem with a Hg pen lamp was performed. By this method, the chromium content of the steel sample was obtained with no more than a 20 % error. This level of precision is certainly acceptable for a quick screening process to determine the rough composition and identification of
O. O0 ,
0 . 0 7
O. 06
O. 0 5
O. 04
0 0
0
O. 04 i i I i t i i i i i i i i i i t
17
Time, D a y s
FIG. 3. Daily X-QC chart demonstrating the fluctuation in Cr analysis in steel. Quantitation was performed by using the slope of the emission ratio curve. Note: no standards were run at the time of analysis.
1122 Volume 45, Number 7, 1991
TABLE IVA. Quantitative analysis of s t e e l - SRM 1264a.
Amount Amount Element present, % found, % Line pair
229.7 C C 0.87 0.88 230.9 Fe 293.9 Mn Mn 0.25 0.26 292.6 Fe 251.6 Si Si 0.067 0.070 249.5 Fe 327.4 Cu Cu 0.25 0.25 328.7 Fe 341.5 Ni Ni 0.14 0.14 339.9 Fe 267.7 Cr Cr 0.06 0.057 265.9 Fe 311.0 V V 0.10 0.10 307.6 Fe 281.6 Mo Mo 0.49 0.50 282.9 Fe 285.0 Co Co 0.15 0.16 259.2 Fe 399.4 Nb Nb 0.15 0.15 304.8 Fe
the type of alloy. If more precise measurements are nec- essary, the information gained in this rapid semi-quan- titative mode can be used to pick the most appropriate standard for best analytical results and can also be used to decide which emission lines will be the most appro- priate for for quantitative analysis.
Quantitative Analysis. The algorithm for quantitative analysis has been previously described. 1° The quantita- tive analysis of a steel and an aluminum reference ma- terial is reported in Tables IVA and IVB, respectively. Also listed are the wavelengths used in the emission ratio used for quantification. Excellent agreement exists be- tween the certified values and those found in our labo- ratory with the use of other reference materials as stan- dards. 17'1s Limits of detection were extrapolated from the available standards. 19 The LOD values for the alloying components in steel and aluminum are shown in Tables VA and VB, respectively. These LOD values are equiv- alent to those reported by other workers. A more recent spectrometer design increases the throughput of light
TABLE IVB. Quantitative analysis of aluminum--Alcoa SS-2011-X.
Amount Amount Element present, % found, % Line pair
288.1 Si Si 0.29 0.30 265.3 A1
Fe 0.52 0.50 259.9 Fe 305.0 A1 282.4 Cu Cu 5.51 5.44 305.0 A1 259.4 Mn Mn 0.042 0.043 305.0 A1
Cr 0.039 0.041 267.7 Cr 265.3 A1 341.5 Ni Ni 0.046 0.044 305.0 A1 334.5 Zn Zn 0.12 0.12 305.0 A1 334.9 Ti Ti 0.005 0.004 305.0 A1
TABLE VA. LOD for selected elements in steel.
Internal stan- Element LOD, % Wavelength, nm dard l ine--Fe
A1 0.0007 396.1 396.3 C 0.0025 229.7 230.9 Co 0.0025 228.6 232.7 Cr 0.0004 267.7 268.9 Cu 0.0003 327.3 328.6 Mn 0.0021 293.3 292.3 Mo 0.0029 281.6 282.9 Nb 0.0045 319.4 304.7 Ni 0.0005 243.7 339.9 Si 0.0030 251.6 249.5 V 0.0011 310.2 307.5
through the spectrometer by a factor of 8. In addition, use of a fluorescence down-converting coating on the detector, Meta-chrome II (Photometrics Limited, Tuc- son, AZ), increases the detector response to radiation in the UV region. These increases in spectral collection ef- ficiency are not likely to improve the LOD significantly in the visible region, as source drift is not the dominant noise source. However, this increased throughput should represent an appreciable decrease in the exposure time, and hence the time per analysis in the visible region, and should improve the LOD values in the UV region.
Temporally Resolved Spectral Images of Single Spark Discharges. Temporal resolution of single-spark dis- charges has been obtained with an experimental setup similar to that described for spatially resolved images of single sparks. 11,12 By masking all but four rows of the CCD and orienting the wavelength dispersion axis of the CP-200 onto the exposed area of the CCD, one can use a technique known as fast spectral framing to simulta- neously obtain a time-resolved spectrum. A more de- tailed discussion of fast spectral framing can be obtained from Aikens et al. 13 A wavelength coverage of 300 nm is obtained with a 200 groove/mm grating at a resolution of 1.0 nm/detector element. The time resolution of 10 ~s was measured independently by using the internal clock of the computer to measure the time required to move the acquired spectra from the imaging area to the region under the mask. The resulting temporally resolved spec- tral images for a single-spark discharge of aluminum are displayed in Figs. 4A and 4B. Figure 4A represents spec- tra obtained before the peak maximum. The first spec- trum obtained is at the bottom, and subsequent spectra are placed above. Figure 4B shows the spectra obtained after the peak emission, with the maximum emission at the top and subsequent spectra placed below. The emis- sion line at 396 nm is an atomic line of aluminum, and the pair of lines at 359 and 361 nm are a neutral silver doublet. The peak intensity occurred at ~ 50 ~s after the
TABLE VB. LOD for selected elements in aluminum.
Internal stan- Element LOD, % Wavelength, nm dard line--A1
Cr 0.0005 359.4 347.9 Cu 0.0001 324.7 305.0 Fe 0.0010 259.9 305.0 Mg 0.00005 279.5 305.0 Mn 0.0011 280.1 305.0 Ni 0.001 341.5 347.9 Si 0.001 288.1 265.3
APPLIED SPECTROSCOPY 1123
t 6 0 0
i I i t
340 350 360 370 380 390
Wave,length (n~
400 410
FIG. 4A. Series of spectra from a single-spark discharge. Moving from the bottom to the top marks increases in time; the top spectrum rep- resents the emission maximum.
initiation of the spark. This finding is consistent with the findings obtained by Zoellner and Scheeline. 19 The silver lines result from sampling of the silver counter electrode. Although the power circuit of the spark source is rectified, the ignition circuit is oscillatory; hence, in the trigger pulse to break down the analytical gap, the silver counter pin is sampled. The point to be emphasized here is the acquisition of 128 separate spectra of a single- spark discharge. The spectral resolution obtained is too low for many fundamental studies of spark excitation dynamics; however higher-resolution gratings are avail- able, the compromise being reduced spectral range. The distinct advantage in employing this approach is the in- strumental simplicity, as compared to other methods of obtaining time-resolved spectral information29'2°
CONCLUSIONS
The analysis of metal samples by spark emission spec- troscopy was successfully carried out with the use of the echelle/CID spectrometer. This system represents an at- tractive alternative to spectroscopic systems employing photographic film or banks of photomultiplier tubes. Since the data are acquired under computer control and collect wavelength information over a continuous 300- nm region, automated routines for qualitative, semi- quantitative, and quantitative analysis can be employed. The use of these computer algorithms greatly aids in the reduction of the analytical data and helps to provide the user with rapid and reliable results. Feedback about the excitation conditions and/or spectral interferences gives the analyst better insight into the interpretation of the data. Because of the system's compact size, modest power requirements, ease of calibration, and mechanical sta- bility, it can be potentially employed as a portable spec- t rometer used to sort metals or verify their composition in the field.
The potential use of charge transfer device technology for fundamental spark excitation studies has also been shown. The advantage here is the simple design, which once again involves no moving parts. At present, the wavelength resolution is not high enough for atomic spec-
~600
0 340 350 360 370 380 390 400 4 t 0
Wavelength (me)
FIG. 4B. Continuation of the spectra series. Moving from the top to the bottom marks increases in time; the top spectrum represents the emission maximum.
troscopic applications involving samples with any spec- tral complexity; however, the feasibility of this approach has been adequately demonstrated for spatially 12 and temporally resolved spectral images. A temporal depen- dence of the S/N ratio of the analyte emission can be used to selectively integrate when the S/N is optimal, which should result in improved precision and limits of detection. The most promising device is a Thompson CSF TH7864 anti-blooming CCD. 21 The anti-blooming diode structure can be operated in such a way as to provide the functional equivalent of an optical iris that can be opened in a fraction of a microsecond. The device can be operated in the nonintegrating mode during the initial stage of the spark process when the S/N is poor, and then in the integrating mode for the remainder of the event. Multiple spectra can be acquired and then binned on chip. The result should be an increase in an- alytical performance, as suggested by Cousins e t a l . 22
The application of solid-state charge transfer device array detectors opens the door to new possibilities in experimental and instrumental design for HPLC anal- ysis, spectroscopy, and imaging. Demonstrated here are just a few of the examples of how the flexibility in using the CTDs as detectors can be implemented and the ad- vantages they offer the analyst.
ACKNOWLEDGMENTS The authors would like to acknowledge the financial and/or the ma-
terial support of Thermo Jarrel Ash, the Baird Corporation, the An- alytical Technologies Division of Eastman Kodak, and Alcoa.
1. H. A. Lewis and M. B. Denton, J. Auto. Chem. 3, 9 (1981). 2. R. B. Bilhorn, P. M. Epperson, J. V. Sweedler, and M. B. Denton,
Appl. Spectrosc. 41, 1125 (1987). 3. R. B. Bilhorn, J. V. Sweedler, P. M. Epperson, and M. B. Denton,
Appl. Spectrosc. 41, 1114 (1987). 4. J. V. Sweedler, R. B. Bilhorn, P. M. Epperson, G. R. Sims and M.
B. Denton, Anal. Chem. 60, 282A (1987). 5. P. M. Epperson, R. B. Bilhorn, J. V. Sweedler, G. R. Sims, and M.
B. Denton, Anal. Chem. 60, 327A (1987). 6. J. V. Sweedler, R. D. Jalkian, R. S. Pomeroy, and M. B. Denton,
Spectrochim. Acta 44B, 7, 683 (1989). 7. R. B. Bilhorn and M. B. Denton, Appl. Spectrosc. 43, 1 (1989).
1124 Volume 45, Number 7, 1991
8. R. S. Pomeroy, J. V. Sweedler, and M. B. Denton, Talanta 37, 15 (1990).
9. R. B. Bilhorn, Ph.D. dissertation, University of Arizona, Tucson (1987).
10. R. S. Pomeroy, J. D. Kolczynski, and M. B. Denton, Appl. Spec- trosc. 45, 1111 (1991).
11. J. D. Kolczynski, R. S. Pomeroy, R. D. Jalkian, and M. B. Denton, Appl. Spectrosc. 43, 5 (1989).
12. R. D. Jalkian, R. S. Pomeroy, J. D. Kolczynski, M. B. Denton, J. M. Lerner, and R. Grayzel, Am. Lab. 80 81, (1989).
13. R. S. Aikens, P. M. Epperson, and M. B. Denton, SPIE, 501 (1984). 14. M. J. Pilon, M. B. Denton, R. G. Schleicher, P. M. Moran, and S.
B. Smith, submitted.
15. T. Torok, J. Mika, and E Gegus, Emission Spectrochemical Anal- ysis (Crane Russak and Co., New York, 1978).
16. Analytical Emission Spectroscopy, E. L. Grove, Ed. (Marcel Deck- er, New York, 1972), Vol. 1.
17. NIST, Gaithersburg, Maryland, Nos. 1261a-1265a. 18. Aluminum Company of America, Alcoa Center, Pennsylvania. 19. M. J. Zoellner and A. Scheeline, Appl. Spectrosc. 41, 943 (1987). 20. J. P. Walters and S. A. Goldstein, Spectrochim. Acta 3911, 5 (1984). 21. Thompson CSF, New Jersey. 22. J. C. Cousins, A. Scheeline, and D. M. Colemen, Appl. Spectrosc.
