2350
Anal. Chem. 1988,58,2350-2352
I
0.032 0.0.
C
I
c+ 2 mn.
F@m 4. Typlcei peak shapes of the standard sample with the valve: (A) sample lnjectlm in reegent Injection position; (B) reagent in)ecuon in reagent WcUm pos#kn; (C) sample and reagent hjectkn in reagent injection position; sample volume 100 pL, reagent volume 100 pL.
water was used as a solution of the flow stream. Deionized water was used throughout the testing and acetone was analytical reagent grade and used without further purification.
RESULTS AND DISCUSSION A typical peak shape when only a sample solution was injected into the flow in reagent injection position (the reagent loop was filled with the deionised water) is shown (A) in Figure 4. Another typical peak shape when only a reagent solution was injected into the flow (the sample loop was filled with deionized water) is shown (B)in Figure 4. Peak C in Figure 4 is a peak shape when the both the reagent and the sample loops were filled with the solution and the solutions were
injected into the flow. Though the loop volume and the diameter of the both loops were the same in case (A) and (B), the peak shapes of the two were not in agreement. It could be due to the differences in the shape of grooves and connectors at the rotor or the stator. We are currently investigating the problem. When flow rate was 0.5 or 1mL/min, the reproducibilities of the sample injection and the reagent injection were investigated and the results are shown in Table I. The reproducibility of a sample injection only was slightly worse than that with six-port injection valves. In the case of the sample and the reagent injection, the reproducibility was almost twice as bad as that in the case of an only sample injection. The reproducibility may depend upon the number of loops. As the timing of the slide of the rotor was slightly long (about 1s), we investigated the reproducibility of the injections at a flow rate of 0.5-1 mL/min. But under these conditions the results were almost the same. So in conclusion, under low pressure (about 1-3 kg/cm2), which usual FIA is performed, the timing would not affect the injection.
CONCLUSION We designed and assembled a multifunction valve for flow injection analysis and investigated the properties. On the whole it would be one of the suitable injection devices for flow injection analysis. LITERATURE CITED (1) Ruzicka, J.; Hansen, H. Now Injectbn Analysis; Wiley: New York, 1981. (2) Tyson, J. F. Ana&st(London) 1985. 170. 419. (3) Rocks, 8. F.; Sherwood, R. A.; Riley, C. Ciln. Chem. ( Whston-Salem, N . C . ) 1982, 28, 440. (4) Attiyat, A. S.; Christian, 0.D. Anal. Chsm. 1984, 56, 439. (5) Ruzicka, J.; Hansen, H. Anal. Chlm. Acta 1983, 745, 1. (6) Mindegeard, J. Anal. Chim. Acta 1979, 104, 185. (7) Bergamin, H. F.; Zagatto, G.; Krug, F.; Reis, F. An8l. Chim. Acta 1978, 101, 17. (8) Toei, J.; Baba, N. Bun&/ Kag8ku 1988, 35, 41 1. (9) Toei, J.; Baba, N., Japan patent pending, 1986.
RECEIVED for review March 18,1986. Accepted May 12,1986.
RapW Headspace Analysis in Sealed Drug Vials by Multichannel Raman Spectrometry Laura Porter Powell*' American Cyanamid Company, Chemical Research Division, 1937 West Main Street, Stamford, Connecticut 06904
Alan Campion Department of Chemistry, University of Texas, Austin, Texas 78712 Many pharmaceuticals slowly decompose when stored in the presence of oxygen and are therefore packaged in a nitrogen atmosphere. Gas chromatography is frequently used to verify the nitrogen gas purge and check for oxygen leaks. Raman spectrometry can potentially perform this determination more rapidly, reliably, and nondestructively. Previously reported Raman spectrometric headspace analyses have employed conventional scanning spectrometer systems (1,2).Residual oxygen content in sealed vials was determined nondestructively by using the nitrogen purge gas as an internal standard. The laser beam was focused inside the sealed vial and the internal Raman scattering was imaged Current address: American Cyanamid Co., Medical Research Division, Lederle Laboratories, Pearl River, NY 10965. 0003-2700/86/0358-2350$01.50/0
and measured. Both the pure rotational and the pure vibrational Raman peaks have been used; the rotational cross sections are approximately 10-fold larger than the vibrational cross sections (1,2).The time required for one scan was a t least 10 min, plus sample alignment and data analysis time. These techniques are not rapid enough for routine pharmaceutical quality control use. The well-known analysis speed of multichannel optical detection systems has been demonstrated for Raman spectrometry applications (3). This speed advantage was realized by Hug and Surbeck for gas-phase rotational Raman spectra of oxygen and nitrogen, each at atmospheric pressure. The authors suggest that pharmaceuticalheadspace analyses might be performed if specular reflections from the walls of the sample vials could be reduced by an iodine filter cell (4). 0 1988 American Chemlcai Society
ANALYTICAL CHEMISTRY. VOL. 58. NO. 11. SEPTEMBER 1986
/
Folding M i r r o r m
I TO
2351
Spectrometer
Figure 2. Common optical sampling and imaging geometries for Raman spectrometric pharmaceutical headspace measurements (side view): (a) vertical configuration. (b) horizontal configuration, and (c) mnfburatimn fw hozen iiauid samoles. The solid arrows reoresent me laser-beam; the dotted rectangles represent an image oi the spectrometer entrance slit.
Camera Lens
Figure 1. Novel folded beam optical geometry for Raman spactromebic headspace measurements in sealed drug vials (sMe view: not drawn to scale). The dotted rectangle represents an image of the vertical spectrometer entrance slit Inside the vial.
However, sensitive and rapid determinations have not been demonstrated. In this work the pharmaceutical headspace determination is improved through the use of multichannel Raman detection together with a novel sample imaging technique. EXPERIMENTAL SECTION Instrumentation. The Coherent CR-6 argon ion laser was operated at 488.0 nm with 1 W power. The spectrometer was a Spex Triplemate 18IIA. In its spectrograph stage a 300 groove/mm grating was used with an entrance slit width of 300 pm, the spectral bandwidth was 60 em-'. The spectrometer wavelength was 540.0 nm (so that the central channel of the detector corresponded to 1973 cn-'). The multichannel detector wed was a EG&G PARC 1420 reticon diode m a y with an OMA2 controller. For each Spectrum the total data acquisition time was 100 8 , and a "dark" background spectrum was subtracted. Standards. Standards were prepared by evacuating vials to 0.1 torr, and then filling them to 740 torr (at 24 "C) with premixed ganes. The gases were purchased from Scott Specialty Gases,Inc., and their oxygen concentrations (in nitrogen) were certified to within 2-5%. Optical Configuration. A side view of the optical sampling and imaging geometry is depicted in Figure 1. The laser beam was parallel to and directly below the monochromatoroptic axis, polarized perpendicular to the entrance slit, and focused in the headspace. Some vials had a dimple in the center of the bottom, and were laterally translated to image the Raman scattering through a flat surface. A standard Canon 50-mm fl1.4 camera lens collected and collimated the Raman scattered light. The collimated beam was folded hy 9oowith a 2-in. mirror (larger than the camera lens diameter). It was then focused onto the vertical spectrometer slit with a 2-in. diameter achromatic doublet of 160-mm focal length (not shown). In the figure the dotted rectangle depicts the optical matching of the vertical spectrometer slit wth the horizontal sample image,
RESULTS AND DISCUSSION In this work the pure vibrational oxygen and nitrogen Raman peaks were used instead of the pure rotational peaks, for the following reasons: 'Real-world" pharmaceutical vials scatter the laser light (and can emit fluorescence), and the multichannel spectrometer separates the higher-frequency vibrational peaks from this Rayleigh light far more efficiently than at low frequencies. Also, the vibrational peaks can be measured by using wide spectrometer slit widths, providing a rugged method that requires less sample alignment time. An improved optical geometry was needed for this work because the configurations described in the literature are awkward and time-consuming to use and do not provide the optimal sensitivity. T o optimize the light throughput, it is desirable to match optically the entrance slit with a line image of the laser passing through the sample headspace. The dotted rectangles shown in Figure 2 depict how the vertical entrance slit in conventional spectrometers is optically matched within the sample (for three configurations). The vertical configuration normally used (I, 5),shown in Figure 2a, has two major drawbacks: (1)Both the incoming laser beam and the exiting Raman scattering interact with curved glass surfaces which bend the light. Therefore, the image is rotated and incompletely fhthe entrance slit and sensitivity is lost. In addition, according to Bailey and Moore, "variability of the ampul wall defocused the laser beam, ...reducing the count rates" (I). (2) The laser is sent into the sample and can cause thermal decomposition, espeeidy at the l-Wlaser power commonly used (I, 5). Also, for opaque samples there are higher stray light background levels, and fluorescing samples interfere more. This second difficulty is overcome in the horizontal sampling and imaging configuration shown in Figure 2h. However, it provides even less sensitivity than the vertical configuration because the sample image is perpendicular to the entrance slit. An improved geometry for liquid samples is illustrated in Figure 2c. The samples are frozen while standing upright and are then held horizontally with the laser beam directed parallel to the frozen surface (5). Freezing the samples adds minutes to hours to the analysis time, and the light must still pass through curved glass surfaces. The folded beam optical sampling and imaging geometry depicted in Figure 1provides many advantages. The entrance slit image matches the line image in the sample headspace, so the sensitivity is optimized. With the samples inverted for analysis, the Raman scattered light is collected through the vial bottom, which is the only (relatively) flat glass surface. Small headspace volumes can be analyzed in the inverted vials; the stopper would hlock the light in the conventional configurations. The laser is directed parallel to the drug surface, so thermal decomposition of the drug is not possible. This optical configuration better discriminates against additional Raman peaks from the drug or the glass; the drug does not
2352
ANALYTICAL CHEMISTRY, VOL. 58, NO. 11, SEPTEMBER 1986
N2 2331
02
1556
+wmumber (an-’)
Figwe 3. Raman spectnrm of the headspace for a standard contalnlng 2.0% O2 in N2 (acquhed in 100 s, 488.0-nm laser at 1 W). The O2 peak is scale expanded by a factor of 10.
interact with the laser, and the glass image is much farther from the laser focal spot. Therefore, wide spectrometer slits are used without any peak overlap or interferences. Because of the wide spectrometer slits and the simple technique for imaging the vials through a flat glass surface, sample alignment is rapid. A Raman spectrum of the headspace for a 2.0% O2in Nz standard is illustrated in Figure 3. Longer data acquisition times would yield improved signal-to-noise ratios, but for our quality control applications such 100-s scans are sufficient. Both the limit of detection (roughly 0.6% 0,) and the count rate (200 counts/s at the N2 peak) were on the order of the results reported by Bailey and Moore (1). The optical throughput of the configuration in this work was high because the focused laser beam was properly imaged and the spectrometer entrance slit was properly filled. The high optical throughput probably compensated for the decreased intensity caused by using the vibrational instead of the rotational Raman peaks. The nitrogen Raman peak was used as an internal standard for quantitation, similar to the method of Bailey and Moore (I). The oxygen peak area (or height) was ratioed to the nitrogen peak area (or height). The quantitative results were similar to the literature work (I). Changes in the peak shapes were caused by sample alignment differences, but they canceled out in the ratio. This work is the first demonstration of the advantages of multichannel detection for Raman spectrometric headspace analyses. The data acquisition speed was about 10 times faster
than in the literature work ( I ) ,even though the spectral range required for the vibrational peak measurement is 100 times larger than that for the rotational peak measurement. Another important advantage is that any changes in experimental conditions such as laser power, sample alignment, detector drift, etc., canceled out because they occurred for both peaks simultaneously. In both this measurement method and the literature method, the largest source of error is the noise associated with the background (1). It is difficult to analyze samples in vials that fluoresce or are of poor quality glass. We have successfully measured 2% O2Raman peaks against backgrounds 2 orders of magnitude larger. This work probably better discriminates against stray reflected background light (from poor quality glass or from the drug), because the vibrational peak frequencies are far from the Rayleigh line. This measurement method meets the goals of rapid, noninvasive headspace analysis, with no possibility for sample damage from the laser beam, in “real-world” drug vials. Further work is needed for routine quality control applications. If an interferometric technique for Raman headspace measurements could be developed, the measurement method would be far more rugged and less expensive. An automated (robotic) technique for changing the samples and optimizing their positions would also be highly desirable. The optical imaging technique described in this work could be extremely useful for general (industrial) Raman spectrometric measurements of solid samples. The samples could be placed on a horizontal surface and translated under the laser beam as on a microscope slide, if the laser entered at an angle instead of parallel to the optic axis. There would be no need for accessories to handle different sample sizes nor for containers to hold powdered samples.
ACKNOWLEDGMENT The authors wish to thank J. Baldoni and W. Wezel of the Medical Research Division, American Cyanamid Co., for supporting this work. We appreciate the cooperation of D. Hatzenbuhler and R. Chao, The Upjohn Co., who shared information about Raman spectrometric headspace analysis techniques. We also thank W. Haseltine, Chemical Research Division, American Cyanamid Co., for preparing the standard samples. Registry No. 02, 7782-44-7; N2, 7721-37-9.
LITERATURE CITED (1) Bailey, Glen F.; Moore, Herbert A,, Jr. J . Parenter. ONg Assoc. 1980, 3 4 , 127-133. (2) DiGregorio, S.; Casper, L. S E X Speaker 1977, 72, 8. (3) Freeman, J. J.; Heaviside, J.; Hendra, P. J.; Prior, J.; Reid, E. S. Appl. Specfrosc. 1901, 35, 196-202. (4) Hug, W.; Surbeck, H. J . Raman Specfrosc. 1982, 73,38-43.
(5) Hatzenbuhler, Douglas; Chao, Robert, The Upjohn Co., personal communications. Oct 1982 and March 1985.
RECEIVED for review February 3,1986. Accepted May 2,1986.
