mained essentially constant. To ensure accuracy, however, response factors were checked weekly and, where necessary, the sample weights for unknown samples were adjusted so that the area ratios for each compound were within those used for calibration. As was mentioned previously, the samples analyzed contained high molecular weight oligomeric polyester compounds and, consequently, exhibited various degrees of solubility in the reaction mixture. The samples normally dissolve completely upon heating. When cooled, the reaction mixture becomes cloudy, apparently because of the precipitation of the high molecular weight compounds. This does not, however, appear to affect theresults. In some instances the samples were highly polymerized and total solution was not possible. For such cases, longer heat-
ing times were required to extract the compounds of interest from the crystal lattice of the polymeric material The time necessary for this varied with the particular sample and was established by plotting the per cent compound cs. heating time (see Figure 2). Results obtained from samples of this type may or may not be accurate because of the uncertainty of effectively removing all the compounds from the insoluble materials. ACKNOWLEDGMENT
The authors thank B. D. Stabley and E. J. Corrigan for their helpful discussions, and C. H. Wilson and R. G. Cozart for their excellent technical assistance. RECEIVED August 19,1970. Accepted November 10,1970.
Gas-Liquid Chromatography of Some Irritants at Various Concentrations Samuel Sass, Timothy L. Fisher, Michael J. Jascot, and John Herban Chemical Research Laboratory, Research Laboratories, Edgewood Arsenal, Edgewood Arsenal, Md. 21010 THIS REPORT DESCRIBES the application of gas-liquid chromatography (GLC) to the detection and quantitative analysis of irritant compounds, some of which have been used by military and law enforcement agencies. The compounds used as examples in this work were a-bromobenzylnitrile (brombenzylcyanide, CA), o-chlorobenzalmalononitrile (CS), and a-chloroacetophenone (CN). A previous report published by these laboratories employed thin-layer chromatography for the detection of these compounds ( I ) . The GLC procedures discussed here further increase the possibilities for absolute identification of these as well as similar compounds while also allowing their quantitative analysis from the milligram through the nanogram range. Also included here are methods for the assay of individual irritant samples and for the detection of some characteristic impurities that could represent hydrolysis or other residues of these irritant compounds. The procedures are applicable to the estimation of irritants as concentrated and dilute solutions and when sampled directly as vapor or aerosol. EXPERIMENTAL
Equipment and Materials. The studies described here were performed on F & M Scientific Corp. (now a division of Hewlett-Packard) Model No. 810 gas chromatograph equipped with thermal conductivity (TC), flame ionization (FI), or electron capture (EC) detector; and a 1-mV MinneapolisHoneywell recorder with a disc integrator. Injections were performed using Hamilton syringes. The column coatings and supports were obtained from Applied Science Laboratories, State College, Pa. All solvents used were of CP grade except for the spectro-quality hexane used in determinations employing the electron capture detector. The 1,lOdibromodecane (used as the internal standard) was obtained through Eastman Organic Chemicals, Rochester, N. Y . Irritant Standards. CA was purified by fractional freezing followed by several recrystallizations from ethyl alcohol and (1) W. D. Ludemann, M. H. Stutz, and S. Sass, ANAL.CHEM., 41, 679 (1969). 462
ANALYTICAL CHEMISTRY, VOL. 43, NO. 3, MARCH 1971
washing with cold petroleum ether. The resultant CA produced near theoretical results by elemental analysis and showed no detectable impurity by thin-layer chromatography ( I ) and only the product peak by GLC. The purity was better than 99.5% and the melting point, 25 "C. CS, originally of 96.5 purity (mp 93-95 "C), was recrystallized from cyclohexane as a white solid melting at 96 "C, with a single GLC peak and no detectable impurity by TLC ( I ) , and with a determined purity of 99.5 %. CN, as recrystallized material, was of better than 98% purity based on melting point (54.5 "C), elemental analysis, ketone determination, GLC peak, and TLC detection. PROCEDURES
SEMIMICRO TO MACRODETERMINATION (MILLIGRAM QUANTITIES). Relative-thermal conductivity-response factor (RTCF) determinations were made for each irritant 1;s. the internal standard (1 ,lo-dibromodecane). Separate benzene solutions containing 50 mg/ml of the individual irritants were prepared along with a 50 mg/ml solution of 1,lO-dibromodecane. Under identical conditions, each solution was injected in 5-111 increments from 5- to 50-pl volumes. The respective slopes of the irritant concentration us. response curves were normalized with respect to that of the 1,lOdibromodecane. Similar solutions containing each irritant in combination with the internal standard were prepared and analyzed. The resultant normalized slopes were identical to those obtained with the separate solutions. Assigning the value of 1.000 to 1,lO-dibromodecane resulted in RTCF values of 1.030 for CA, 1.176 for CN, and 1.111 for CS. To determine purity, a weighed sample of irritant (CA, CN, or CS) and a similar weight of internal standard (1,lOdibromodecane) were placed in a 10-ml volumetric flask and made up to volume with benzene to yield an agent concentration of about 50 mg/ml. Aliquots of sample were injected directly onto a 5.5-ft x 1/4-in,0.d. borosilicate glass column packed with 10% QF-1 on 60/80 mesh Gas-Chrom Q. A helium carrier gas-flow of 90 ml/min was used with temperature-programming from 65-200 OC at 6 "C/min. MICRODETERMINATION (MICROGRAMS). To detect irritants (CA, CN, and CS) in microgram quantities, the chromato-
graph was fitted with its dual hydrogen-flame ionization detector and employed the identical column and conditions as described previously for the macrodetermination. The hydrogen and air-flow rates were 50 ml and 250 ml/min, respectively. A weighed sample of irritant of known purity (as determined via internal standard using TC detection) was dissolved in solvent (benzene, chloroform, or ether) and made up to volume prior to constructing a calibration curve. Using this curve, the irritant peak areas from the unknown samples were measured and extrapolated to the abscissa to determine the agent content. TRACEDETERMINATION (NANOGRAMS).The gas chromatograph was fitted with a tritium-source electron capture detector. The identical QF-l packed column (under isothermal temperatures and with a carrier gas mixture comprised of 90% argon and 10% methane) was used for the detection and estimation of nanogram quantities of the irritants. The carrier gas flow rate was maintained at 85 mljmin while the detector was purged with a flow of 10 ml/min of the identical gas mixture. Samples and standards were prepared in spectro grade hexane with calibration performed in the range of 5 to 50 ng of irritant. The isothermal column temperature was 150 "C for CS and 130 "C for CA and CN. RESULTS AND DISCUSSION
Semimicro to Macrodetermination. The elution data obtained for CA using the GLC system with thermal conductivity detection showed elution times of 15 rnin (153 "C) for CA, 12 min (137 "C) for benzylcyanide, and 20 rnin (185 "C) for 1,lO-dibromodecane. A standard deviation of 1.0% was calculated on the basis of 16 separate samples involving more than 50 individual sample injections. The maximum range found for a given sample was h2.2 %. As no other absolute quantitative method was available for comparison purposes, these values reflected only the precision of the GLC system. However, confirmation of accuracy was obtained through the analysis of synthetic CA samples containing weighed amounts of added known impurities of CA including benzylcyanide, a-bromobenzylimidobromide, cu,cu'-dibromobenzylcyanide and a combination of all of these. For determining C N purity, the system gave a standard deviation of 1.0% on the basis of repetitive analysis. Weighed mixtures of 1,IO-dibromodecane, CN, and acetophenone were dissolved in benzene or chloroform and aliquots of these were analyzed by GLC. A range of 50 to 99.5% CN purity was simulated in this test. Under conditions of the prescribed procedure, CN eluted at 14.5 rnin at 149 "C while acetophenone eluted in 9 min at 119 "C. The determination of CS purity using this method showed a standard deviation for accuracy and precision of 1 %, based on both the inherent system and on comparison with a macrovolumetric method (2). For this study, a batch of CS (96.5 %>,recrystallized product (99.9 % from cyclohexane), and mixtures with impurities were used. Since prepared mixtures of CS with o-chlorobenzaldehyde and malononitrile could, on solution, produce some additional CS, the effect of analysis on one or the other of these two intermediates was simulated. Tested mixtures represented CS in purities from 50 to 99.9%. With this procedure CS elutes in 22 rnin at 193 "C, malononitrile in 7 min at 101 "C, and ochlorobenzaldehyde in 7.5 min at 110 "C. (2) S. Sass et al., CWLR 2396, "Analytical Methods for CS. Part 11. Volumetric Determination of CS Purity" (1960), Edgewood
Arsenal, Md. internal publication.
Table I. Accuracy of Assay Method for CA and CN Using Weighed Mixtures of Components
Samvle Nd. 1 2
3 4
Average irritant DifWeight. found, ference,
z -
Component
I
87.8 10.4 1.8 79.7 a-Bromobenzylimidobromideb 19.0 Others" 1.3 87.8 CA 10.7 a,a-Dibromobenzylcyanide 1.5 Others= 71.2 CA 8.0 Benzylcyanide a-Bromobenzylimidobromideb 11.3 8.5 a,a-Dibromobenzylcyanide 1.0 Otherso
CA Benzylcyanide Othersa CA
z
x
87.2
-0.6
79.7
0.0
86.6
-1.2
70.6
-0.6
-0.7 45.6 44.4 CN 54.4 Acetophenone $1.1 59.4 60.5 CN 6 40.6 Acetophenone +0.4 74.9 75.3 CN 7 25.1 Acetophenone $0.9 86.2 87.1 8 CN 13.8 Acetophenone -0.7 94.2 93.5 CN 9 5.8 Acetophenone -0.4 99.5 99.1 10 CN a "Others" refers to unidentified contaminants found in the samples of CA and components used. b Bromobenzylimidobromide did not dissolve in significant quantity. 5
Chemical assay methods previously established for the quality control of CA and C N had entailed, essentially, elemental analysis and physical methods such as melting point determination. The application of GLC to the assay of these irritants and also to CS increased the specificity of analysis while allowing the identification and estimation of elutable impurities. Data obtained on prepared synthetic mixtures of the separate irritants with characteristic impurities are summarized in Tables I and 11. Microdetermination. The results obtained via flame ionization detection proved this system was most useful for analyzing irritants in solution concentrations in excess of 50 pg/ml. The minimum quantity of irritant detectable was 2 pg while the overall accuracy was +2.0% at the 10- to 200-pg level for all the irritants. The elution data were identical to those reported with the semimicro to macrodetermination. Trace Determination. Results obtained uia the electron capture detector indicated that solutions containing CA in quantities as low as 0.2 pg/ml could be analyzed. A measure of the precision of this system was calculated on the basis of replicate 12.5-p1 injections at the 7 pg/ml concentration level. The standard deviation thus derived was 2.0%. As with most methods, the accuracy of the method declined as the lower limit of detection was approached. For example, replicate CA vapor concentration determinations in the 3035 mg/m3 range showed a deviation of approximately 1.4 mg/m3; while in the 5-10 mg/m3 range, the deviation was of the order of 1 mg/m3. The minimum amount of CA detectable using this method was about 4 ng (or 4 X lou98). The retention time of CA was 6.5 rnin with a column temperature of 130 "C. ANALYTICAL CHEMISTRY, VOL. 43, NO. 3, MARCH 1971
463
Table 11. Accuracy of Assay Method for CS Using Known Mixtures of Components and Comparison with Macro-Volumetric Method
Sample CS, recrystallized CS, production lot 1
Component mixture
cs cs
cs
Malononitrile 0-Chlorobenzaldehyde
2
cs
Malononitrile 3
cs
Malononitrile 4
cs cs
o-Chlorobenzaldehyde 5
o-Chlorobenzaldehyde a
Component weight, % 99.9 96.5 54.9 23.7 21.4 88.5 11.5 63.7 36.3 76.0 24.0 49.9 50.1
GLC, %
CS found Macro-Volumetric Method, %a
Difference of GLC from Weighed Mixture, %b
99.5 95.9 85. 3c
99.9 96.5
-0.4 -0.6
89.4
89.4
+0.9
64.0
62.8
$0.3
76.8
75.5
f0.8
48.9
50.5
-1.0
...
...
The standard deviation of the macro-volumetic method (2) is 0.7%. For CS with no additive, the purity value is assumed that of the macro-volumetricmethod. The total calculated quantity of CS present, if other components reacted 100% to form additional CS, is 86.3%:.
For CN, a standard deviation of 2 x at the 90-ng level was derived, based on injections of 7 p1 of a 12.5 pg/ml CN solution. Replicate C N vapor determinations in the 5 to 10-mg/m3range resulted in a deviation of 0.5 mg/m3. The minimum amount of C N detectable using this method was about 0.2 ng (or 2 X g) and the retention time was 6.0 min at a column temperature of 130 O C . For CS, the system also indicated a standard deviation of 2 z at the 80-ng level from injections of 8 pl of 10 pg/ml CS solution. Replicated CS vapor determinations in the 5 to 10-mg/m3 range resulted in a deviation of 0.4 mg/m3. The minimum amount detectable, using this method, was about 0.1 ng. The retention time for CS (column temperature = 150 “C) was 12 min. Minimum vapor concentrations detectable for each of the agents, using this procedure, are: CA, 0.4 mg/m3; CN, 0.02 mg/m3; and CS, 0.01 mg/m3. These values were determined with 10-ml vapor samples. It was demonstrated that cloth, soil, water, or other media could be “sniffed” for these compounds uia syringe or other gas sampling apparatus, providing the vapor concentration was above these detection thresholds. Most of the problem in vapor sampling appeared to be due to heterogeneity of sample resulting from irritant condensation on the walls of the gastight syringe. This could be precluded by using a heated syringe. Measurements of CA and C N vapor, using GLC, were actually used in determining the volatilities of these compounds at various temperatures (3). (In an associated study, it was found possible to analyze both solution and vapor samples of ethyl bromoacetate using the basic EC procedure given for the other irritants. As little as 0.1 ng of the compound could be detected as a small peak eluting at 4.8 min, using an isothermal column temperature of 60 “C.) The systems for “macro” to trace analysis of CA, CN, and CS were designed to fit the most difficult of the three compounds, namely CA. The compound was known to pyrolize readily in the presence of metals to form stilbene types and a variety of other products. To preclude metal-catalyzed
decomposition, all-glass systems were used wherever possible. Diluted solutions of the compound showed less of this effect. Samples injected via syringe (containing a metal needle) were retained in the syringe for only a minimum of time. With electron-capture, reducing the detector temperature from 220 to 160 “C increased the maximum sensitivity time from a week to greater than two months. When sluggishness was noted, periodic flushing of the column and/or detector with methanol restored the sharpness of peaks and also detector sensitivity. CN and CS showed far less thermolability in the presence of metals. A number of column coatings were tested for satisfactory elution of CA. These included SE-30 (methylsilicone gum rubber), DC-200 silicone oil, and QF-1 (trifluoropropylmethyl silicone fluid). None of these coatings eluted dibromobenzylcyanide or the bromobenzylimidobromide present in CA samples. Benzylcyanide, however, was eluted in all cases just prior to the CA peak. QF-1 was selected as the coating giving the cleanest separation and peaks. Direct calibration with CA was discarded as CA of near 100% purity would be difficult to keep on hand for long periods of time. The internal standard procedure of calibration required high purity CA only during the initial relative-thermal conductivity-response factor (RTCF) determinations. 1,lO-Dibromodecane was chosen as the internal standard for several reasons: it eluted close to CA but in an area free of peaks characteristic of CA impurities; it was a stable compound available in high purity or readily purified; and it also was a halogenated compound and did not react with constituents of the CA samples. Unlike CA, CN and CS were compatible with metal columns, injection port vaporization, and with Apiezon-N and DC-LSX-3-0295 (trifluoropropyl-vinyl-methyl silicone gum polymer) coatings. However, since a more universally applicable procedure was desired, the method developed for CA, including the internal standard, was applied to the other irritants.
(3) J. J. Martin, S. Sass, et al., EATR 4086, “Soil Stability of CA and CN. Analytical Methods for Trace Quantities of the Agents” (1967), Edgewood Arsenal, Md., internal publication.
RECEIVED for review August 26, 1970. Accepted November 10, 1970.
464
ANALYTICAL CHEMISTRY, VOL. 43, NO. 3, MARCH 1971