Anal. Chem. 1997, 69, 2221-2223
Analysis of USP Organic Volatile Impurities and Thirteen Other Common Residual Solvents by Static Headspace Analysis Rodney B. George* and Preston D. Wright
Analytical Research, R&D Services, Mallinckrodt Inc., P.O. Box 5439, Saint Louis, Missouri 63147-0339
Static headspace gas chromatography was investigated for the analysis of residual solvents in size-limited samples. The advantages of improved limits of detection at low ppm levels and decreased sample size requirements were realized. This methodology allows the measurement of 18 common residual solvents in 6 min using 1 mg or less of sample. Greater recovery of analytes from pure dissolution solvents without the use of salts became possible when smaller preparations of samples were combined with the use of smaller vials. Rapid equilibration of the static headspace sample preparations was an additional benefit of this modification. Optimized chromatography was developed to take advantage of the shorter equilibration time and to increase instrument productivity. Static headspace analysis has become a versatile technique for the analysis of trace volatile components in many matrices. It is commonly used in research, quality verification, and stability studies. Static headspace is based on the principle of dissolving a sample matrix in a suitable solvent, heating this mixture in an enclosed vial to equilibrate the volatile components between the gas and liquid phases, and injecting a sample of the gas phase from above the liquefied sample into a gas chromatograph for analysis. Many factors can adversely affect static headspace analysis. Care must be taken to seal the vials in a consistent manner. Thermostating the prepared vials must be performed in a precise and reproducible manner. Atmospheric contamination in the laboratory must be minimized for accurate determinations. Fixed sample loop injection systems characteristically can have large interfering carryover peaks. To evaluate headspace injection systems, two analyzers were compared, one using a fixed loop sampling system, and one using a timed direct injection system. Speed and accuracy are also important parameters in any laboratory setting. Cost of analysis for expensive or limited size samples can be a major concern in a quality control or research laboratory. This study demonstrates a methodology that uses no inorganic salts to increase analyte detectability, consumes only 1 mg of sample, separates 18 commonly used solvents in 6 min with a limit of detection of less than 1 ng (1 ppm), and has a total analysis run time of 15 min. EXPERIMENTAL SECTION Instrumentation. The study was performed on a Varian Model 3500 capillary gas chromatograph (Varian Corp., Palo Alto, CA 94304) fitted with a flame ionization detector (FID). Attached to the GC was a Tekmar 7000 headspace autosampler (Tekmar S0003-2700(97)00040-1 CCC: $14.00
© 1997 American Chemical Society
Co., Cincinnati, OH 45242-9576). This headspace analyzer utilizes a fixed-sample loop technology and standard vial sizes of 9-, 12-, or 22-mL volume. The generic stainless steel sampling needle and tubing supplied with the instrument were replaced with fusedsilica coated components, modified for Tekmar by Restek (Restek Corp., Bellefonte, PA 16823-8812). A Perkin-Elmer HS-40 headspace analyzer (The Perkin-Elmer Corp., Wilton, CT 06897) attached to the same Varian 3500 GC was used for comparison purposes. This instrument differs from the Tekmar unit in that it uses a valveless timed injection technique instead of a fixed-loop sample valve. Special 3.35-mL vials with custom inserts obtained from Perkin-Elmer were used in their analyzer and in the Tekmar unit. Procedure. A 1-mL sample loop was employed. Samples were heated at 85 °C for 6 min, mixed for 30 s, equilibrated for 30 s, and pressurized to 10 psi prior to sampling. A Silcosteel union from Restek connected the heated (100 °C) headspace analyzer transfer line to the GC analytical columns. The series of fused-silica open tubular analytical columns consisted of a 3 m × 0.53 mm i.d. Rtx-Wax [bonded poly(ethylene glycol)] of 1-µm film thickness, a 30 m × 0.53 mm i.d. bonded Rtx-1301 (bonded 6% cyanopropyl phenyl methyl silicone) of 3-µm film thickness, and a 5-m uncoated 0.25 mm i.d., each connected by fused-silica butt connectors. The column oven was programmed with a 2-min initial hold at 40 °C, increased at 20 °C/min to 120 °C, held for 0.5 min, increased to 180 °C at 40 °C/min, and held for 2 min. Helium was used for both the carrier gas and the FID make-up, and each was set at 20 mL/min. The FID was set at 200 °C and at the most sensitive amplifier setting. Data was collected by a Digital Equipment Corp. (DEC) Vax 4000/400 computer running Perkin-Elmer Access*Chrom version 1.96 software. Analytical grade reagents (Mallinckrodt Inc.) were used throughout the procedure. Each sample preparation of 25 µL of purified water was spiked with 1 µL of one of eight dimethylacetamide solutions containing the solvent analytes at different concentrations. These solvents were as follows: methanol, diethyl ether, ethanol, acetone, 2-propanol, dichloromethane, hexane, 1-propanol, ethyl acetate, tetrahydrofuran, chloroform, benzene, isobutanol, trichloroethylene, dioxane, methyl isobutyl ketone, toluene, and pyridine. These solutions were prepared by diluting a stock made from 250 mg of hexane, 1000 mg of chloroform, and 500 mg (all (1%) of each of the other 16 analytes dissolved into a final volume of 100 mL of dimethylacetamide. Final concentrations were 0, 0.50, 9.69, 25.32, 49.78, 136.76, 483.93, and 1000.25 µg/mL. Chloroform was at twice and hexane was at half these concentrations. The following number of spiked injections were made: six of the 0.50-ng level spikes, two of the 9.69-ng Analytical Chemistry, Vol. 69, No. 11, June 1, 1997 2221
Table 1 analytea
average (%) recovery
resolution factor
precision (% RSD)
linearity (R2)
methanol diethyl ether ethanol acetone isopropanol dichloromethane hexaneb 1-propanol ethyl acetate tetrahydrofuran chloroformc benzene isobutanol trichloroethylene dioxane MIBK toluene pyridine
95.2 97.0 98.7 96.3 96.6 96.1 96.5 95.5 96.0 96.4 99.2 96.1 95.8 94.8 95.2 96.2 96.1 88.5
5.07 1.64 1.82 2.55 2.61 2.55 5.21 1.58 1.88 2.10 3.45 1.69 3.63 3.71 5.21 2.07 2.02
2.8 2.1 3.7 1.8 2.8 2.6 2.9 3.5 2.2 2.0 3.9 3.6 2.6 4.4 4.7 2.9 5.0 4.8
0.999 59 0.999 85 0.999 81 0.999 82 0.999 82 0.999 74 0.999 81 0.999 77 0.999 88 0.999 88 0.999 69 0.999 74 0.999 82 0.999 53 0.999 77 0.999 88 0.999 52 0.995 17
a
Sample size, 50 ng; n ) 3. b Sample size, 25 ng. c n ) 6.
level spikes, six of the 25.32-ng level spikes, three of the 49.78-ng level spikes, and one each at the other levels. A comparison of relative recoveries using vials of 22-, 9-, and 3.35-mL volume was made at the 483.93-ng spike level to assess the impact (if any) of vial size. RESULTS AND DISCUSSION Comparison of the HS-40 and the Tekmar 7000 indicated that the Perkin-Elmer HS-40 was superior in reducing analyte carryover; injection to injection carryover was minimal. At most a single blank injection between samples was needed to essentially eliminate carryover. Through the use of 10 or more water injections, the carryover of the Tekmar instrument could be reduced sufficiently to allow proper quantitation at low levels. Subsequently, the Tekmar instrument was used for data gathering for this study. Alcohols, esters, dioxane, and pyridine were found to be most susceptible to absorption in rotor seals and adsorption to other active sites in the instrument. The carryover and adsorption of pyridine in the headspace system caused its recovery at a 25-ng level to be unacceptably high and its linearity (R2) to fall below 0.999. The unusual configuration of analytical columns in the chromatograph is dictated in order to focus the sample plug injected from the headspace analyzer. The bonded poly(ethylene glycol) column also changes the selectivity of the bonded 6% cyanopropyl phenyl methyl silicone column and permits the 18 solvents to be separated rapidly. Dissolution solvents such as dimethylacetamide or dimethyl sulfoxide and different sample matrices can increase or decrease the limit of detection as they affect the partition ratio of the volatile components between the gas and liquid phases. The suitability of each sample matrix or its dissolution solvent must be determined experimentally for the residual solvent analyte in question. A complete listing of the following analytical criteria calculations are found in results Table 1. Accuracy. Recovery of the 50-ng spikes was based on the regression line of the other seven spiked level injections. Average recoveries ranged from 88.5 to 99.2% for the 18 analytes. 2222 Analytical Chemistry, Vol. 69, No. 11, June 1, 1997
Figure 1. The 0.5- and 9.69-ng spiked preparations using 3.35mL vials. Peaks: (1) air, (2) methanol, (3) diethyl ether, (4) ethanol, (5) acetone, (6) isopropanol, (7) dichloromethane, (8) hexane, (9) propanol, (10) ethyl acetate, (11) tetrahydrofuran, (12) chloroform, (13) benzene, (14) isobutanol, (15) trichloroethylene, (16) dioxane, (17) methyl isobutyl ketone, (18) toluene, (19) pyridine, and (20) dimethylacetamide.
Precision. Relative standard deviations (RSDs) of 5% or less were attained for each analyte from the 50-ng spiked preparations. Linearity. Regression of the blank and the seven spiked level injections gave coefficient of correlation (R2) values of 0.995170.99988. Selectivity. This system was able to achieve adequate resolution for the identification and quantitation of 18 common residual solvents. Resolution factors (R) greater than 1.5 were calculated from the injection made at the 1000-ng spike level. No single chromatography system can be expected to separate all possible analytes. Additional work outside this study has demonstrated suitable linearity, recovery, and precision for dimethyl sulfide, acetonitrile, methyl ethyl ketone, and dimethyl sulfite solvent analytes, but certain coelutions exist. Quantitation is possible if only one of a coeluting pair is present. The USP OVI analytes were resolved at a resolution factor (R) of >3 with RSD values of less than 5% at a level of 50 ppm for each using a theoretical sample weight of only 1 mg. Limit of Detection. The Varian chromatograph with its capillary-optimized detectors has superb sensitivity. The criterion for a detectable peak is to have a peak height that is 3 times the level of the detector noise. All 18 analytes were detected at levels less than 1 ng. Certain problems are manifested at this low level of detection. The dimethylacetamide produces artifactual peaks which coelute with methanol and dichloromethane and are adjacent to hexane, isobutanol, and dioxane. The pyridine peak is much larger than expected due to carryover in the fixed-loop analyzer from previous injections containing pyridine. Compare
Vial Size Effects. Vial size is a primary factor in analysis by static headspace of microsamples. The three chromatograms in Figure 2 are of the same 483.93-ng spike sample in 22-, 9-, and 3.35-mL volume vials. Each was prepared and injected with the same parameters. The improved response of trace volatile impurities when vials of smaller volume are used for the analysis of submilligram sample preparations is clearly visible. (See Figure 2.) Sample equilibration times were reduced from the prescribed 60 min1 to less than 10. CONCLUSIONS Static headspace GC analysis is capable of separating and measuring the USP OVI analytes and 13 other common residual solvents accurately using 1 mg or less of sample, while meeting the USP parameters of resolution and precision for the OVI analytes and reducing the prescribed runtime of about 15-45 min.1 Utilization of a smaller vial size can improve the sensitivity and the speed of trace residual solvent analyses. Extrapolation of this finding leads to the idea that a better solution would be an instrument that injects the entire vial headspace contents. To date no manufacturer has made this solution available to laboratory analysts.
Figure 2. The 483.93-ng spiked preparations using 22-, 9-, and 3.35-mL vials.
the relative peak heights of pyridine in the chromatogram of the 0.5-ng spiked sample in Figure 1 injected after more than 10 water blank injections to the relative height of pyridine in the chromatograms of the 483.93-ng spiked samples in Figure 2.
ACKNOWLEDGMENT The authors gratefully acknowledge Valerie Lopez-Naughton and Ed Lewis of Tekmar Co. for donating the fused-silica coated Tekmar 7000 components. We also thank Pat Pijanowski, Kaj Petersen, and Lee Marotta of the Perkin-Elmer Corp. for the extended use of the HS-40 headspace analyzer, the 3.35-mL vials, the custom vial inserts, and their invaluable technical assistance. Received for review January 14, 1997. Accepted March 26, 1997.X AC970040U
(1) United States Pharmacopeia 23 (USP23-NF18); United States Pharmocopeia Convention, Inc.: Rockville, MD, 1995; pp 1746-1747.
X
Abstract published in Advance ACS Abstracts, May 1, 1997.
Analytical Chemistry, Vol. 69, No. 11, June 1, 1997
2223
