LITERATURE CITED (1) P. D. Goulden and P. Brooksbank, Anal. Chem., 46, 1431 (1974). (2) F. D.Pierce, T. C. Larnoreaux, H. R. Brown, and R. S. Fraser, Appl. Specfrosc., 30, 38 (1976). (3) R. S.Brarnan and C. C. Foreback, Science, 182, 1247 (1973). (4) Y. Talmi and D. T. Bostick, Anal. Chem., 47, 2145 (1975). (5) J. S. Edrnonds and K. A. Francesconi, Anal. Chem., 48, 2019 (1976). (6) A. A. El-Awady, R. B. Miller, and M. J. Carter, Anal. Chem., 48, 110 (1976).
(7) P. D. Goulden, 8.K. Afghan, and P. Brooksbank, Anal. Chem., 44, 1845 (1972). (8) F. D. Pierce and H. R. Brown, Anal. Chem., 48, 693 (1976).
RECEIVED for review May 2, 1977. Accepted June 29, 1977. The use of the brand name in this report is for identification purposes Only and does not endorsement by the u.s* Geological Survey.
Determination of Ethylene Oxide in Gas Sterilants by Fourier Transform Infrared Spectrometry P. V. Allen and A. J. Vanderwielen" Control Analytical Research and Development, The Upjohn Company, Kalamazoo, Michigan 4900 1
A quantitative infrared determination of ethylene oxide in dichlorodifluoromethane has been described. The method was developed on a computerized single beam Fourier Transform Infrared (FTIR) system to overcome some of the problems associated with dispersive infrared spectrometers. The FTIR method gave accurate resuits between 5 and 25 %, calculated on a weight basis. A comparison was made between a gas chromatographic method, an infrared analysis on a dispersive infrared spectrophotometer, and the FTIR determination. A brief study on difference spectrometry has been made and showed promise as a quality control technlque.
Gas sterilization using ethylene oxide or propylene oxide as the active ingredient has become a preferred method for treating a wide variety of foodstuffs, drugs, and medical supplies and equipment. Basically, gas sterilization is used because certain materials cannot be readily sterilized with dry heat, steam, or chemical soaks without damage (1, 2). Present analytical methods utilize gas chromatography to determine the amount of active ingredient present in the sterilizing gas mixture. Gas chromatography using a flame ionization detector has been investigated (3-5) and satisfactory results were obtained by using a gas-sampling loop as verified by two independent methods ( 5 ) . However, a large fraction of some gas sterilants consists of halocarbons (e.g., CClzFz is a common inert propellant) and flame ionization detectors are not well suited for halocarbon analysis. To date, there have been no studies that have examined the possibility of utilizing infrared spectrometry as an analytical method to identify and quantitate the amount of ethylene oxide in gas sterilants, although infrared analyzers have been used to monitor the amount of ethylene oxide in sterilization units ( I ) . The objective of the present research was to examine the viability of using infrared spectrometry as an analytical method for determining the ethylene oxide content in dichlorodifluoromethane (Freon 12) in cylinders of sterilizing gas using Fourier Transform Infrared spectrometry. In this paper we describe a quantitative infrared method for determining the ethylene oxide concentration in cylinders of sterilizing gas containing ethylene oxide and Freon-12. Procedures for taking samples from the gas cylinders and analyzing these gas mixtures have been established. The results are compared to quantitative methods using a con1602
ANALYTICAL CHEMISTRY, VOL. 49, NO. 11, SEPTEMBER 1977
ventional double-beam grating spectrophotometer and an analytical gas chromatograph. We also report the results of a brief study to determine the usefulness of difference spectrometry as a quality control method (6).
EXPERIMENTAL Infrared. Spectra were obtained on a Digilab Model FTS-1OM Fourier Transform Infrared Spectrophotometer (FTIR) and on a Perkin-Elmer 421 grating spectrophotometer. The P-E 421 was set to scan at approximately 9 cm-'/s using a fivefold gain and no suppression. These conditions give a nominal resolution of about 6 cm-'. Calibration curves were constructed using standard mixtures which were prepared in a vacuum line by combining pure Linde ethylene oxide and dichlorodifluoromethane. Spectra measured on the FTS-1OM were collected in a single beam mode and ratioed against a stored spectrum of the empty gas cell. The spectrophotometer was purged using dry nitrogen. From 50 to 500 scans were cc-added to obtain an acceptable signal to noise ratio at 4 cm-' nominal resolution (as given by the FTS-1OM instrument settings). All calculations (except one as will be indicated later) were carried out in single precision without apodization of the interferograms. Except for the difference spectra obtained using the absorbance subtraction technique, spectra were plotted on the transmittance scale and the intensities of the peaks were measured by the baseline method. This worked very well since the empty gas cell provides the ideal reference and flat baselines were obtained for the ratioed spectra. Gas Chromatography. The sterilizing gas mixture was also analyzed by gas chromatography (GC) using a Hewlett-Packard Model 402 instrument with a flame ionization detector. The GC conditions used were those adapted by P. B. Bowman and P. A. Hartman (7) for headspace analysis (8)and are as follows: (i) Column: 4 f t X 0.25 inch 0.d. glass, packed with Porapak R; (ii) Temperatures: Oven at 67 "C, injector at 64 OC, detector at 120 "C; Carrier Gas: Helium at 40 mL/min. A 100-pL gas tight syringe was used to withdraw samples from the cylinder (method discussed under Sampling Procedures) and inject approximately 30 pL of gas onto the GC column. Sampling Technique. We attached a two-stage regulator to the tanks of sterilizing gas and assembled the samplingapparatus shown in the diagram depicted in Figure 1. The regulator was adjusted to deliver 5-10 psi of gas to the 250-mL sampling bulb. Special care must be taken in sampling to prevent fractionation of the gas mixture. The sampling apparatus was purged six times in a stepwise fashion so that there was never a dynamic flow through the system which could cool the regulator enough to cause fractionation (whenever practical, a heat gun or heating tape should be used to keep the regulator at or above room temperature). The sample was then trapped in a small bulb on a vacuum system using liquid nitrogen, warmed to ambient
c
w "ll".
YllVl
snrnpl. bulb
o o O.rlnQ nniollon
Flguro 1. Sampling apparatus for collecting gas samples for infrared analysis. First stage pressure is approximately 50 psi at ambient temperature. Second stage pressure is set at 5-10 psi
65
, . . . . .1 . . Em-1 . . . I,
. . , .
,
..
-.I. 2650 2350
3110
..... .... .... . 2150
F R E l l U E N C Y Icm 11
Flgure 3. Gas phase spectrum of (A) ethylene oxlde (9% wlw) In CC12F, (cell pressure, 24 Torr, 500 scans), and (B) the scale expanded spectrum of the region of interest 3800
2000
660
FREOUENCY lcm-ll
Figure 2. Gas phase spectra collected by 50 scans on the FTS-1OM instrument: (A) is ethylene oxide at 36 Torr and (B) is CCI,Fp at 12 Torr
temperature, allowed to equilibrate for 5-10 min, and finally expanded into an evacuated 10-cm gas cell with NaCl windows. All of the quantitative analyses were preformed using sample pressure between 100 and 125 Torr. Mixtures for the experiments on difference spectrometry were prepared using the standard mixture and adding small amounts of propylene oxide obtained from Union Carbide, Inc. The spectra were obtained using the same 10-cm cell filled to a pressure between 20 and 50 Torr of gas mixture. For the gas chromatographic analyses, the method of sampling was altered; for this experiment a septum and two needle valves were attached to the regulator and the gas was sampled with a gas-tight syringe. The same precautions as noted above were taken during purging.
constructed. The weight percent of E T 0 was calculated according to:
where P and MW refer to the pressure in mm of Hg (torr) and molecular weight in g/mol of each gas. A plot of the ratio of the peak intensities (IC-HIICFJ vs. the 70ET0 (w/w) gave a slope of 0.052 and an intercept of 0.256 using the FTS-IOM FTIR and somewhat different values on the P-E 421 instrument (Slope, 0.070, intercept, 0.619). The correlation coefficient in both cases was found to be 0.998. At 10-15% ET0 the 95% confidence limits were found to be *0.8% when a linear regression analysis was performed. Plotting the ratio of absorbances vs. the ratio of weights in general would be expected to produce a linear calibration curve over a wide range of values. However, as is indicated by our data, the ratio of the peak intensities vs. the weight percent was h e a r over the region from 5 to 25% E T 0 by weight. This method of presentation was best suited for our needs in terms of data collection and ease of calculation by the laboratory analyst. Using absorbance values would be advised for general application to give a wider range. Plots of the absorbance ratio vs. % E T 0 by weight were linear, but a non-zero intercept (0.2) was still observed. This would indicate that Beer's law is not strictly obeyed or that there is a low intensity band of the CClzFz under the ET0 peaks of interest. The weight percentages of samples then calculated according to:
RESULTS Spectra were obtained of pure ethylene oxide (ETO) and CC12F2(Figure 2, A and B, respectively) and the peaks of interest were found to be a C-H stretching band between 3150 cm-l and 2850 cm-I and a CF2 overtone between 2350 cm-l and 2150 cm-' (Figure 3A and 3B). A calibration curve of the ratio of the peak intensities vs. % ET0 (by weight) was
where b = Intercept and m = l/slope. IC-H = Peak intensity at 3110 cm-' for FTS-1OM and at 3030 cm-' for the P-E 421. ICFz = Peak intensity at 2240 cm-' for the FTS-1OM and a shoulder on the 2240 cm-l peak for the PE-421. ANALYTICAL CHEMISTRY, VOL. 49, NO. 11, SEPTEMBER 1977
1803
Table I. Multiple Samples from One Tank Results on FTS-1OM Intensity Intensity of peak of peak at 3110 at 2240 Sample cm-l, cm cm-', cm Ratio A
B C D E F
14.43 13.61 13.08 12.55 12.88 12.65
15.27 13.64 13.39 13.26 13.74 12.78
C D E F
9.30 9.42 9.18 9.48 9.25 9.50
5.54 5.60 5.48 5.58 5.79 5.69
Sample A B C D E F
% ET0 by
weight
0.945 0.998 0.977 0.946 0.937 0.990
13.25 14.27 13.87 13.28 13.10 14.16 Av= 13.7% Std dev = 0.5
Results on P-E 421 Intensity Intensity of peak of shoulder at 3030 at 2240 Sample cm+, cm cm-', cm Ratio A B
Table 111. Gas Chromatography Resultsa
H
13.72 13.89
14.22 14.48
% ET0 by
weight
1.678 1.682 1.675 1.699 1.598 1.670
15.13 15.18 15.08 15.32 13.97 15.00 A v = 15.0% Std d e v = 0.5
0.965 0.959
13.63 13.53
13"
Multiple samples were taken from one tank of sterilizing gas and analyzed using the FTS-1OM. The same samples were also run on the P-E 421 for comparison. The results are given in Table I and typical spectra are shown in Figures 3-5. Duplicate samples of four other cylinders were run on the FTS-10M. The results listed in Table I1 show good agreement between the various duplicate experiments. The reproducibility of the results indicates that our sampling procedure is adequate for this gas mixture. Table I reveals that while the FTS-1OM and P-E 421 have nearly identical precision, the average obtained on the PE-421 is significantly higher (significant difference = 0.8% E T 0 at the 95% confidence level). For the chromatographic analysis of the sterilant mixtures, areas under the curve were obtained by using a ball and disc integrator. The percent ethylene oxide was calculated from the peak areas by normalization with response factors (5):
(3) where Al is the area of the E T 0 peak, A , is the area of the CC12F2peak, and R2 is the correction factor for CClzFzto E T 0 (R, = 5.8 was used). The results (Table 111)indicate that there 1604
CCI,F, area
1205 1076 1060 1274 2 577 1860
1295 1307 1266 1250 1797 1827
% ET0 by
weight
19.6 12.4 12.6 14.9 19.8 14.9 Mean 15.7 Std dev 3 . 3
a Samples taken from the same tank used during the infrared analysis listed in Table I.
Table 11. Duplicate Samples from Four Tanks Intensity Intensity of peak of peak at 3110 at 2240 %ET0 by Av, Sample cm-', cm cm-', cm Ratio weight % Tank No. 1 A 11.68 12.78 0.914 12.66 B 13.21 14.22 0.929 12.93 12*' Tank No. 2 C 13.72 0.953 13.41 14.40 D 13.21 14.22 0.929 12.93 13*2 Tank No. 3 E 14.33 13.74 1.043 15.13 15,2 F 14.58 13.89 1.050 15.27 Tank No. 4 G
ET0 area
ANALYTICAL CHEMISTRY, VOL. 49, NO. 11, SEPTEMBER 1977
are several problems associated with the gas chromatographic assay when a gas syringe is used; (i) the large variability indicates that it is very difficult to obtain representative samples using a gas syringe, and (ii) the results may be biased high due to gas fractionation (CC12F2has a higher vapor pressure than ETO). The results of the absorbance subtraction experiments (Figure 6) indicate that a 0.7% impurity is easily detectable. The spectral data for the difference spectrum depicted in Figure 6B was collected by 100 double precision scans (resolution = 4 cm-l) on the FTS-1OM FTIR. The Fourier Transform calculations were carried out in double precision and used a rectangular apodization function (this results in essentially no weighing of data points of the interferogram during the transform calculation).
DISCUSSION As is so often the case, sampling is the most critical aspect in the analysis of the sterilizing gas mixture. Sampling and subsequent use of the sterilizing "gas" is facilitated by withdrawing the liquid mixture (tank is under pressure and liquid is forced out through a tube that extends to the bottom) and vaporizing it. By using the apparatus (Figure 1)and the sampling procedure outlined in the Experimental section, liquid samples are permitted to vaporize completely (no fractionation), and the gas mixture, therefore, is representative of the liquid. When the ratio of peak intensities is to be used to determine the relative concentrations of two gas phase components by infrared spectrometry, it would be desirable to choose a peak for each component which is distinct, reproducible, and totally free of interference from peaks due to other components. The C-H stretching band at 3110 cm-l and the CF2 overtone at 2240 cm-l adequately fulfill these conditions. As Figures 2A and 2B show, the peak at 3110 cm-l is quite distinct, essentially independent of any peak from CC12F2and nearly free of interference from rotational spikes (unlike the slightly more intense peak at 3030 cm-'). The CF2 overtone a t 2240 cm-' is reproducible, very sharp, and despite the fact that it falls in the midst of some weak ET0 peaks, sufficiently independent of E T 0 to be useful. The band at 3110 cm-I and the peak at 2240 cm-' which were used when the analysis was done on the FTS-1OM could not be used on the PE-421 because of the lack of reproducibility for these sharp peaks. It was necessary to use the broader peak at 3030 cm-I and the shoulder on the 2240 cm-' peak during analyses with the PE-421. We chose to use a rectangular (box car) apodization function to calculate the Fourier Transform of the interferogram to obtain a signal to noise ratio of approximately 200 in a reasonable time (9) and to avoid the possible error than can occur in the absorbance measurement when a triangular apodization function is used (10). Anderson and Griffiths indicate that for a 4-cm-l wide band measured with a nominal resolution of 2 cm-l, the measured peak absorbance will be
Table IV. Pressure Effects Ratio, Pressure, ICH/ICF, mm Hg 150 0.768 125 0.766 100 0.752 75 0.719 50 0.708
Ethylene oxide,a % by weight 9.85 9.81b 9.55b 8.91 8.69
a Standard provided by Union Carbide was 10.1% (w/w) Average obtained under norethylene oxide in CCl,F,. mal experimental conditions was 9.68% E T 0 (w/w).
3400
FREOUENCY (cm 11
2000
Figure 4. Scale expanded spectrum of ethylene oxide (13.6% by weight in CCI,F,) obtained on a Perkin-Elmer 421 at 120-Torr pressure
at least 16% low if triangular apodization is used. They also suggest that apodization somewhere between triangular and rectangular might be an excellent compromise for analytical work. However, we feel that rectangular apodization gave us accurate results in this case with no interference from the secondary lobes due to neighboring peaks. The widths of absorption bands in gas phase spectra can be affected by the total sample pressure. A study was performed on the FTIR instrument to examine these pressure effects on the ratio of the band intensities. A standard (10.1% w/w ethylene oxide) was prepared for us by the Linde Division of Union Carbide. The analysis was performed at pressures ranging from 50 to 150 Torr and the % E T 0 was calculated using the calibration curves obtained under the normal experimental conditions (100-125 Torr). The results (Table IV) indicate that there is a pressure effect; however, only the results at 50 and 75 Torr differ from the actual value by more than 0.8% ( E T 0 95% confidence limit). It is of interest to compare the results obtained on the FTS-1OM to those obtained on the PE-421. As mentioned previously, the latter gives ethylene oxide concentrations that are higher (Table I) than those obtained on the FTIR instrument. The accuracy of the FTIR result was verified during the study on pressure effects (9.7% E T 0 measured vs. 10.1% E T 0 actual). There are several possible reasons for this anomaly. The bias could be caused by the uncertain baseline (see Figure 4) or by cell effects (e.g., NaCl window fogged). The problem is corrected on the FTIR instrument since the empty cell is used as the reference. On the P-E 421, excessive scanning time was necessary to minimize anomalies due to pen response and even then it was necessary to use the shoulder of the CF2 overtone a t 2240 cm-l to obtain reproducible results. The instrument line shape of a dispersion spectrophotometer is approximately represented by a triangle ( 1 1 ) . Therefore, it is quite possible that the resolution and absorbance errors observed with triangular apodized FTIR spectra (9) are likely to occur in spectra obtained on conventional dispersion spectrometers. The C-F2 band is con-
-0 024 3800
,
, , , ,
,
. , . . . 2000 iREQUENCY icm.?
I
800
Figure 5. Results of the absorbance subtraction of (A) two gas mixtures (both consisting of 89 % (w/w) CClpFp-t 11% ethylene oxide) and (B) mixture minus (mixture 0.7% (w/w) propylene oxide). Cell pressure of Sample A approximately 22 Torr and Sample B approximately 50 Torr
+
siderably narrower than the C-H band. The decrease in absorbance due to triangular apodization should, therefore, be expected to have a greater effect on I c - F ~than on Ic-H, causing a net increase in the apparent concentration of E T 0 as calculated from IC-H/Ic-F2. However, this effect should be accounted for in the calibration curve. It is more likely that the higher values for E T 0 concentration obtained on the P-E 421 are due to uncertainties in the baseline, a problem not encountered with the FTS-1OM. Infrared analysis of ethylene oxide concentration is only one of many techniques available today (12,13),but we feel that this method offers several advantages (speed, accuracy, and easy sample handling) and will be useful for many applications. It has been suggested that difference spectrometry on Fourier Transform instruments could be used as a quality control method (6). When the spectrum of an acceptable product or material is recorded and stored, it can be compared with any subsequent spectra. The difference spectrum in Figure 5A shows that when the absorbance spectra of two acceptable samples of sterilizing gas (11% E T 0 in each case) are subtracted from each other, the difference is less than 0.005 absorbance unit (AU) in the region 3800 cm-' to 1200 cm-'. The peaks observed a t lower wavenumbers are due to the difficulty in subtracting peaks due to strongly or totally absorbing vibrational modes. Figure 5B is the difference between two samples of gas sterilant (9% ETO), one of which contains 0.7% (w/w) of propylene oxide as an impurity. The difference in this case is 0.05 AU in the region 3800 cm-l to 1200 cm-l. A comparison of the resulting spectrum to that of pure propylene oxide (Figure 6) shows that there would be little difficulty in identifying the impurity. These results indicate that it is certainly plausible to utilize the capabilities of FTIR instruments for quality control decisions. In our example case, we might require that the difference in the region 1500 cm-I to 1300 cm-' be no greater than 0.01 AU (this corresponds to approximately 0.5% propylene oxide) and a different specification for the region 3200 cm-' to 2800 cm-' ANALYTICAL CHEMISTRY, VOL. 49, NO. 11, SEPTEMBER 1977
1605
ACKNOWLEDGMENT We thank M. F. Fuller for her help in obtaining the gas chromatographic data and R. W. Brink for the experimental work for the studies on pressure effects.
LITERATURE CITED
10 , 3800
, ,
, ,
1
(1) John J. Perkins, “Principles and Methods for Steriiizatlon in Health Sclences”, Charles C Thomas, Springfield, Ill., pp 501-530. (2) Alian Ciaghorn, Inhalatlon Therapy, 11, 77-84 (1966). (3) B. Warren, J n Pharm. Pharmacol., 23, 1705-1755 (1971). (4) Ben Berck, J. Chromatogr. Scl., 13, 256-267 (1975). (5) R. G. Wliilams, UpJohn Flne Chemicals Dlvlsbn, unpublished work, February ,
,
,
1976.
,
2000 FREQUENCY Icrn.11
BOO
Figure 6. Gas phase spectrum of propylene oxlde (50 scans, cell pressure is 30 Torr)
to allow for acceptable variations in ethylene oxide concentration. The results of our study on difference spectrometry indicate that the method can be developed as a quality control technique. However, computer programs will have to be written to automate the data manipulation before it can become a viable technique.
(6) J. L. Koenlg, Appl. Specfrosc., 20, 293-308, (1975). B. Bowman and P. A. Hartman, UpJohn, Product Control Division, Determination of Residues of Geseous Sterilants”, to be published. (8) S. J. Romano, J. A. Renner, and P. M. Leitner, Anal. Chem., 45,
(7)
E.
2327-2330 (1973). (9) Peter R. Griffiths, ”Chemical Infrared Fourier Transform Spectroscopy”, John Wiiey and Sons, New York, N.Y., 1975, Chapter 8. (10) R. J. Anderson and P. R. Grifflths, Anal. Chem., 47, 2339-2347 (1975). (11) K. S. Seshadri and R. N. Jones, Spectrochlm. Acta., I O , 1013 (1963). (12) S. J. Romano and J. A. Renner, J. Pharm. Sci., 64, 1412-1417 (1975). (13) N. A. Kolchina, Zh. Anal. Chlm., 20, 380-387 (1965).
RECEIVEDfor review January 12, 1977. Accepted June 20, 1977.
Enhancement of the Fluorescence Intensity of Derivatives of Amino Acids in Mixed Solvent Systems Peter M. Froehlich” and Larry D. Murphy Department of Chemistty, North Texas State University, Denton, Texas
A number of reagents are known to react with amino acids to form fluorescent derivatives and it has been shown that the fluorescence intensity of the derivatives can be increased by consideration of the solvent system used for its formation. The use of mixed solvent systems such as DMSOIwater in place of water for the formation and detection of o-phthaidehyde (opt) derivatives of amino acids was found to raise the fluorescence intensity slgnifkantiy (e.g., 64 % for tryptophan-opt in 17% DMSO (v/v) compared to the case where DMSO was absent). Similar observations were found wlth dansyl amino acids and fiuorescamine amino acids.
In an analytical procedure where the detection is performed by fluorometry, the ultimate sensitivity is directly related to the fluorescence quantum yield of the analyte (or of a derivative directly related to the analyte). The fluorescence quantum yield of many compounds is very sensitive to the environment of the excited state (1). It has been determined, for example, that water is capable of interacting with the excited state of indole (Z),tryptophan and its metabolites (3), and tryrosine (4) to form an excited state complex (exciplex). The formation of an exciplex is competitive with fluorescence; hence the fluorescence quantum yield is reduced, thus raising the minimum detection limit of an analytical procedure. A reasonable approach to the problem of exciplex formation in fluorometric analyses might be to perform the analysis under 1606
ANALYTICAL CHEMISTRY, VOL. 49, NO. 11, SEPTEMBER 1977
76203
-
Table I. Relative Fluorescence Intensity of o-Phthaldehyde Derivatives of Amino Acids in Mixed DMSO” % DMSO
Aminoacid
0
7
10
13.5
17
20
27
Alanine 1.00 1.00 1.34 1.72 1.91 1.57 1.66 Phenylalanine 1.00 1.11 1.15 1.20 1.23 1 . 3 4 1.40 Histidine 1.00 1.50 1.70 2.15 2.20 2.45 Tryptophan 1.00 1.40 1.69 1 . 6 3 1.64 1.58 1 . 7 ” All data relative t o the fluorescence of the derivative in 0% DMSO. nonaqueous conditions. In theory, this would serve the analyst quite nicely; however, the removal of water may be time consuming. In those cases, the removal of water is not a viable procedure for the analysis. We have recently shown that the fluorescence of various indoles, tryptophan metabolites, and tyrosine metabolites (3) is considerably higher in mixed DMSO/water or ethanol/ water than in neat water; this indicates that it may be very useful to consider the overall solvent system used in the analysis of these compounds. A common tactic in analytical chemistry is the formation of a derivative of the compound of interest. The final solvent system in such an analysis is a composite of the original solvent of the sample, solvents used to isolate the compound of interest from the matrix, and solvents used in the formation of