Experimental Considerations in Headspace Gas Chromatography

May 2, 2010
Pharmaceutical Technology, Pharmaceutical Technology-05-02-2010, Volume 34, Issue 5

In this case study of amines, the authors discuss several parameters to be considered in developing a headspace GC method.

Headspace gas chromatography (GC) has been used since the 1980s, but only recently has it become part of mainstream pharmaceutical analysis. The United States Pharmacopeia (USP) incorporated the technique in 2007 into its General Chapter <467> "Residual Solvents." USP defines the use of headspace GC for the analysis of organic volatile impurities (OVIs) or, more commonly, residual solvents (1). A primary benefit of using headspace GC is that one can analyze small amounts of residual solvents buried in a large amount of matrix (i.e., active pharmaceutical ingredient) without having to inject the matrix onto the chromatography column. This technique results in clean, easy sample preparation coupled with less wear on the chromatography columns and the GC instrument because there is less compound passing through the injector or injection liner and detector.

Definition of headspace

The process described in this study is more accurately known as static headspace. However, because no other headspace techniques will be discussed, only the term headspace will be used. The methodology for headspace is quite simple. A sample is diluted and sealed in a closed container that is large enough to allow some headspace in the vial where vapors can collect. The sample is thermostated for a period of time to allow the solvents in the matrix to volatilize, enter the gas phase, and reach equilibrium with the remaining solvent(s) in the vial. This gas phase is the headspace, which is sampled and injected onto the column using pressure as the injection mode. The injection precision of a headspace injector can consistently give < 10.0% relative standard deviation (RSD) and frequently gives ≤ 2.0% RSD, similar to the precision found with liquid-chromatography injection.

The authors studied small, basic organic amines, which are commonly used in pharmaceutical processes instead of more traditional bases such as sodium hydroxide or sodium ethoxide. Use of these amines presents several analytical challenges. The International Conference on Harmonization (ICH) guidelines do not include limits for many such amines; pyridine being the major exception (2). The target value for the specification for such amines can be assigned by the chemist or quality assurance department, which can lead to the same amine having very different set limits. Because specifications are often set at the lowest level the process will tolerate, having an efficient method that can detect low levels of these small organic amines is important. Additional analytical problems with these compounds include non-Gaussian peak shape due to interactions with the GC column, and low flame-ionization detection (FID) based on the size of the compound and low amounts of oxidizable (via burning) carbon groups.

Experimental considerations

This investigation focused on some headspace parameters that should be considered when developing a headspace method, with the goal of maximizing the signal and sensitivity while efficiently using time and materials. All these parameters are important in the early stages of drug development, however, a fast, efficient, sensitive method is also beneficial to the quality-control laboratories that release approved drugs. The authors evaluated experimental parameters including sample of volume in the headspace vial, incubation time in the headspace oven, and the composition of the diluent. The three amines chosen (triethylamine, n-butylamine, and allylamine) represent a cross-section of organic amines that have been analyzed by the authors during scale-up processes.

Sample volume. Headspace sample vials typically come in 10-mL and 20-mL sizes. Often, the analytical chemist needs to analyze a very small amount of a residual solvent in a matrix, which raises the issue of sensitivity. Traditionally, the analytical chemist would increase the concentration of a sample or inject more sample onto a column to get a better signal, thereby increasing the signal strength. With headspace, more sample volume does not always provide the expected increase in area counts because the greater the sample volume, the smaller the actual headspace volume becomes (see Figure 1). The article describes the effect that sample volume has on response for triethylamine, n-butylamine, and allylamine in different diluents.

Figure 1

Incubation time. Solvent-vapor equilibria play a role in headspace analysis. The more readily a solvent can be evaporated into the headspace, the more of that particular solvent will be injected onto the column. Incubation of the headspace vial is an important consideration in headspace-GC method development. If the sample is incubated for too short a period of time, less of the analyte will be in the headspace, which can affect overall area counts. After a certain point, however, the analyte and the solution from which it came will settle into equilibrium; more incubation will not result in any more sample entering the vapor phase and may result in sample degradation or cause secondary reactions. The authors varied the incubation time of triethylamine, n-butylamine, and allylamine samples and reported the change in solvent response over time.

Diluent composition. Salt raises the boiling point of water by affecting the equilibrium of a solvent with its vapor. Because headspace also depends on solvent-vapor equilibria, such techniques can also be used to increase vapor concentration in a headspace-GC sample. Salts have been used in headspace analysis (3–5); however, this technique works best for aqueous-based solvents, which are not the best solvents for GC analysis. The authors, therefore, used high-boiling solvents such as dimethylsulfoxide (DMSO) and dimethylformamide (DMF) as sample diluents. The authors varied the percent of water (aqueous) in the diluent to see whether mixing in a solvent with a higher vapor pressure would affect the concentration of volatile solvents in the headspace.

Methods and materials

Instrumentation and sample materials. A gas chromatograph (Agilent 6890, Santa Clara, CA) with a G1888 (Agilent) headspace autosampler and 20-mL sample vials were used for the experiments. The column used for these amine analyses was a Restek RTX-5 AMINE (Bellefonte, PA), 30 m x 530 µm x 5 µm film thickness. Triethylamine (TEA) was purchased from FisherScientific (Division of Thermo Fisher Scientific, Waltham, MA); DMSO, n-butylamine, and allylamine were purchased from Sigma-Aldrich (St. Louis, MO); Water (high-performance liquid chromatography grade) was purchased from J.T. Baker (Phillipsburg, NJ); and dimethylformamide (DMF) was purchased from EMD Science (Darmstadt, Germany).

Stock TEA samples were prepared by diluting 138 µL into 100 mL of diluent, where the diluent was different ratios of either DMSO:water or DMSO:0.01M sodium hydroxide (NaOH). Analysis samples were further diluted 1 mL into 100 mL with the appropriate diluent for the experiment. This dilution resulted in a final TEA concentration of 0.0100 mg/mL. Stock n-butylamine samples were prepared by diluting 213 µL into 250 mL of diluent, where the diluent was either water, 0.1M NaOH, or DMSO. Analysis samples were further diluted 1 mL into 25 mL with the appropriate diluent for the experiment (different ratios of DMSO:0.1M NaOH or water). This dilution resulted in a final n-butylamine concentration of 0.0252 mg/mL. Stock allylamine samples were prepared by diluting 300 µL into 100 mL of diluent, where the diluent was either 0.001M NaOH, 0.01M NaOH, DMF (or DMSO), or ratios of DMF (or DMSO):base. Analysis samples were further diluted 1 mL into 100 mL with the appropriate diluent for the experiment. This dilution resulted in a final allylamine concentration of 0.0228 mg/mL.

Table I: Headspace parameters

GC method for TEA and n-butylamine: Three minutes isothermal at 50 °C, followed by a programmed temperature ramp at 20 °C/min to 130 °C, a hold period for 2 min, followed by a programmed ramp at 20 °C/m into 170 °C, a hold period for 3 min, followed by a ramp at 30 °C/min to 240 °C, with a four-minute hold period at the end of the temperature ramp. The injector temperature was 200 °C, and the detector (i.e., FID) temperature was 300 °C with a constant column flow of 3.8 mL/min of helium.

Figure 2

GC method for allylamine: Eight minutes isothermal at 45°C, followed by a programmed temperature ramp at 40 °C/min to 240 °C, with a four-minute hold period at the end of the temperature ramp. The injector temperature was 180 °C, and the detector (i.e., FID) temperature was 300 °C with a constant column flow of 3.8 mL/min of helium. Headspace parameters are shown in Table I.

Figure 3

Results and discussion

Sample volume. The sample volume for each of the amines was varied by running replicates of 100 µl, 500 µl, 1 mL, 2 mL, and 5 mL of solution in the headspace vials. Figures 2–4 show that sample volumes higher than 1 mL do not yield the expected increase in signal. The experiments with allylamine were originally run using DMF, which resulted in the opposite response to that shown by using DMSO as the diluent. The response of allylamine decreased with increasing amounts of sample in the headspace vial, thereby demonstrating that the choice of diluent is important.

Figure 4

Vial equilibration time. The sample equilibration time in the headspace oven was varied by running replicates at 5, 10, 15, 30, and 60 min. Figures 5–7 demonstrate that equilibration times longer than 5 or 10 min do not yield a significant increase in signal. For TEA, the 15-minute time point shows more scatter in the data than the other time points, which could be indicative of a secondary reaction or interexperimental differences because of different columns and analysts. It is of interest to note that USP recommends incubating the samples for 60 min. These data show 60 min to be excessive and indicate that the same results can be obtained in one-quarter of the time, with less risk of secondary reactions occurring.

Figure 5

Diluent composition. Often, analyses require detecting small amounts of an analyte in the matrix. This function can highlight the power of headspace because one can use large amounts of sample in the headspace vial to get at the sensitivity required for detection without having to inject large amounts of the matrix onto the column. There are cases, however, when increasing sample size is insufficient, and other techniques are required to increase the signal. One option studied was to vary the percentage of water in the diluent to determine if the concentration of volatile solvents in the headspace could be affected by mixing the diluent with a solvent that has a higher vapor pressure. The following set of experiments was conducted testing the amines, using diulent systems containing some aqueous (as water, 0.001M NaOH, 0.01M NaOH, or 0.1M NaOH) and an organic solvent. Base was added to ensure that the amine was a volatile-free base and not present as a salt. Figure 8 presents the response of TEA with increasing percentages of water in the diluent. At 50% water, the peak splits into two peaks, likely due to water interactions with the amine and the liquid phase of the column. For these samples, the two peak areas are combined (denoted by –S). Using 0.01M NaOH does not appear to be necessary, because the responses with and without added base were similar. For TEA, the ratio of 80:20 DMSO:water appears to give the best response, thus affording an increase of about 60% over straight DMSO. Figure 9 presents the response of n-butylamine with increasing percentages of water in the diluent. Addition of the base 0.1M NaOH does appear to have an effect, as using only water results in lower responses. For n-butylamine, a ratio of 90:10 DMSO:0.1M NaOH appeared to give the highest response; however, the increase was only about 6% more than pure DMSO. The solvent selection for allylamine was more difficult, because the original solvent system of DMF showed low-area responses for the standard 1 mL sample volume as observed in the vial volume experiment previously outlined. The mixture of 90:10 DMF:0.001M NaOH does appear to yield a higher signal than other DMF solvent mixtures, but it is still much lower than similar mixtures using DMSO and a slightly stronger base. Other combinations were tried (e.g., DMF:water; DMF:0.0001M NaOH); however, nothing appeared to work better than neat DMSO (data not shown).

Figure 6


Three aspects of headspace method development were examined for a set of small basic amines. This compound class was chosen because these compounds are being used more frequently in the pharmaceutical industry in place of more traditional bases and because they often present problems in GC analyses because of their low sensitivity. It was shown that increasing the sample volume in the headspace vial does not always yield the desired increase in signal and that smaller amounts in the sammple vial can give the desired sensitivity. Specifically, the USP calls for solution volumes of 5 mL to be used, whereas these data show that 1 or 2 mL volumes can give equivalent signals using less material.

Figure 7

Equilibration time is another aspect of headspace method development that was examined. One must ensure that the sample has incubated long enough for all the solvents in the vial to achieve equilibrium with solvent in the liquid in the vial. If the sample does not reach equilibrium, the injection precision will vary from sample to sample. If the solution is in equilibrium with the gas phase, the injections will be consistent and precise. If the sample incubates too long, there is a risk of matrix degradation, which could add complexity to the residual-solvent analysis. The data showed that 15 min and often only 5 min were required for the residual amines to reach equilibrium. This finding is supported by the fact that upon longer incubation, there was no increase in signal. These data are in contrast to USP <467>, which sets a vial equilibrium time of 60 min. The data show that such an extended incubation is unnecessary; the time of analyses could be increased as much as 4–30 times and demonstrate the same sensitivity.

Figure 8

The final aspect of method development studied was multicomponent diluents. In some cases, it was possible to increase sensitivity by adding a small percentage of water or aqueous base to the sample diluents, as the water or aqueous base can aid in partitioning the solvents into the gas phase. In the case of triethylamine, the signal increased by 60% by using a diluent 80:20 ratio of DMSO:water instead of just DMSO. For allylamine, however, no combination of solvents yielded a signal higher than that with just using DMSO.

Figure 9

Many laboratories have generic residual-solvent methods that use a particular high-boiling solvent (i.e., DMSO, DMF, dimethlyacetamine, or n-methylpyrolidone) and a standard set of headspace conditions. These may work for many solvents/matrices; however, there are solvents that are more difficult to analyze, for example, the small organic amines that were studied in this work. This work has shown that by adjusting two of the GC headspace parameters (i.e., vial volume and equilibriation time), the consumption of sample and analysis time can be reduced. Finally, by trying different diluents (i.e., DMF versus DMSO for allylamine) and multicomponent diluents, containing a small amount of aqueous, the analysis of difficult analytes may be improved by enhanced sensitivity.

Laila Kott* is a senior scientist in analytical development, small molecules at Millennium Pharmaceuticals, 40 Landsdowne St., Cambridge, MA 02139, lkott@lanet.lv or laila.kott@mpi.com, and Hong Ming Chen is a quality control chemist at Ash Stevens Inc., 18655 Krause St., Riverview, MI 48193.

*To whom all correspondence should be addressed.

Submitted: May 19, 2009; Accepted Sept. 18, 2009


1. USP 32–NF 27 General Chapter <467> "Residual Solvents" (2007).

2. ICH, Q3(R4) Impurities: Guideline for Residual Solvents (July 2007).

3. H. Takamatsu and S. Ohe, J. Chem. Eng. Data 48 (2), 277–279 (2003).

4. F. Banat et al., Chem. Eng. Process. 42, 917–923 (2003).

5. A. Naddaf and J. Balla, Chrom. Suppl. 51, S283–S287 (2000).