Enhancing GCMS analysis of trace compounds using a new dynamic baseline compensation algorithm to reduce background interference Abstract The advantages of mass spectrometry (MS) in combination with gas chromatography (GC) in terms of providing both qualitative and quantitative information - are well known. However, confident identification and accurate measurement of trace toxic and odorous species at the lowest levels of interest may be compromised by chromatographic anomalies such as; column bleed, extended solvent tails, air /water interference and unresolved sample matrix components. To address this issue, a new software algorithm has been developed for reprocessing stored GCMS data. It uses an innovative mathematical approach to distinguish and eliminate background interference from the mass ions contributing to real chromatographic peaks, even at the lowest levels. This should be of tangible benefit to any GCMS work involving detection and measurement of trace target compounds in complex uncharacterised samples environmental applications, homeland defence, forensic studies, food tests, materials emissions tests, flavour/fragrance profiling, odor monitoring, etc.. The performance and potential advantages of the new software to trace GCMS applications generally have been evaluated and are reported here. The examples presented include applications requiring trace enrichment carried out using thermal desorption (TD)-GCMS and high speed analyses carried out using fast- GC with TOFMS. Introduction Confident quantitative and qualitative analysis of trace compounds using GCMS can be compromised by chromatographic anomalies such as; column bleed, extended solvent tails, air /water interference and unresolved sample matrix components. A new GCMS data reprocessing algorithm has been developed which uses an innovative mathematical approach to distinguish and eliminate background interference from real chromatographic peaks, even at the lowest levels. It compensates dynamically for chromatographic background interference as it changes throughout a run without compromising any of the peak-related information. This dynamic baseline compensation (DBC) algorithm can be applied to total ion (TIC) or extracted ion (EIC) data and produces a separate DBC-reprocessed data file in addition to the original GCMS data file. An example of the software in operation is shown below (Figure 1.) Figure 1 shows the TD-GCMS analysis of a landfill gas standard carried out using automated thermal desorption with a thick-film, high-bleed capillary column
and quadrupole MS in full scan mode. The TIC trace is shown both before and after DBC-reprocessing. The dramatic impact on the chromatographic baseline is immediately apparent, but what the TIC data doesn t show is the equally dramatic impact on spectral purity and signal to noise. Figures 2 and 3 Figure 1: TD-GCMS analysis of a landfill-gas standard using a thick film capillary column with high bleed. Original TIC data file shown (black), DBC-reprocessed data file (blue). Apparatus: ULTRA-UNITY TD system (Markes International) with 6890-5973 GCMS (Agilent Technologies),) and ClearView GCMS reprocessing software Original GCMS data S/N ~ 3:1 Figure 2: Automatic analysis of the original GCMS data file does not correctly identify the trace component at 16.486 minutes because high level background ions interfere with the mass ion fragmentation (MIF) pattern ClearView reprocessed data S/N ~ 30:1 Figure 3: Automatic analysis of the reprocessed data file shows greatly enhanced spectral purity allowing automatic
On the face of it, this new automated approach to eliminating background interference could offer significant benefits to GCMS analyses both in terms of data quality and laboratory throughput. The purpose of this study was to evaluate the performance of this new algorithm with a range of challenging GCMS applications, in order to test its advantages and limitations under different conditions and assess how widely it could be applied. Examination of potential sensitivity enhancement As shown in the above example, selective removal of GCMS background ion contributions reduces the level of noise observed in full scan (TIC) data. The observed increase in signal to noise (S/N) ratio for the thiophene peak was from ~3:1 (RMS) in the original GCMS data file (figure 2) to ~30:1 in the reprocessed data file (figure 3). This represents a ten fold increase in sensitivity. Given the observed significant reduction in background interference, the new software should also facilitate calibration and quantification based upon a sum of characteristic ions rather than just relying on the intensity from a single extracted ion.. For reasons of selectivity, response factors for individual compounds are typically calculated based on a single target ion within the mass ion fragmentation (MIF) pattern of interest. The mass ion selected is typically only one of many generated during the ionisation process and may not even be the most significant ion in the fragmentation pattern (spectrum). This means that, in many cases, only a small proportion of the MIF is actually being used for quantitative analysis. If the risk of interference can be removed to negligible levels, the choice of mass ion selection for quantification could be expanded to encompass potentially the entire MIF profile, a range of ions, or a summation of specific mass ions. Examining the MIF pattern for the peak at 16.486 minutes in the original data file (figure 2), it is clear that, due to the high background at this point in the TIC, quantification of thiophene would require single extracted ion analysis using parent mass ion 84. The number and range of interfering mass ions precludes use of a mass range for quantification in this case. In contrast to this, it is clear from the reprocessed data file (figure 3) that masses ranging from around 30 to 84 amu could be used for thiophene quantification after dynamic baseline compensation. In this actual example, the extracted ion profile for mass 84 has a peak area of 144112, whereas using mass range 30-84 for quantification gives an area of 395085. This represents nearly 3-fold increase in actual signal / peak area. Chromatographic needles in TIC haystacks The performance of the new GCMS reprocessing algorithm was evaluated for even lower level compounds using an example of trace contaminants in pure air. In this case 500 ml of clean outdoor air was sampled and analysed using on-line thermal desorption with GC-TOF MS. Original and DBC reprocessed data are
shown below (Figure 4) with a close-up of the area of the TIC around 12.6 minutes shown in Figure 5. Figure 4: 500 ml outdoor air sampled on-line. Original GCMS data (black). DBC data (blue) Original MS data MIF from original data Tiny peak at 12.619 mins Spectrum after ClearView Reprocessed data Styrene Figure 5: Close-up of original and reprocessed data around 12.6 minutes. Insets: High mass ion interference in the original data file meant no identification was possible. Much lower mass ion interference in the reprocessed data allowed automatic library search identification of styrene. The high level of background interference in the original GCMS data file made it impossible to identify the trace component at 12.6 minutes without careful manual background subtraction. However, the enhanced spectral purity observed
for this peak in the DBC reprocessed data file allowed this component to be automatically identified as styrene without any manual intervention. Discussion of potential improvements in quantitative accuracy and repeatability In order to evaluate the performance of the new software with respect to quantitative accuracy and repeatability, it was presented with a particularly challenging GCMS application High speed analysis of diesel fuel (5 minute total run time) containing trace levels (3 pg on column) of octafluoronaphthalene (OFN). A Time-of-Flight (TOF) mass spectrometer (BenchTOF-dx from ALMSCO International.) was used in this case. Figure 6 shows the original GCMS data file overlaid with the reprocessed data file. Once again the impact of DBCreprocessing in terms of flattening baselines (this time by removing the hump of unresolved sample components) can clearly be seen. The process of flattening the baseline alone facilitates more accurate and repeatable automatic integration of resolved components. Figure 6: TIC analysis of trace levels (3 pg on column) of octafluoronapthalene (OFN) in diesel using high speed chromatography and Time-of- Flight (TOF) MS. 5 minute run time. Original GCMS data (black) shown overlaid with reprocessed data (blue) In order to assess any potential negative impact of DBC on the mass ions contributing to chromatographically resolved peaks, both the data files for this analysis (original and reprocessed) were first examined in extracted ion mode. While the trace OFN signal is buried within the diesel matrix in the TIC trace (figure 6) and cannot be detected, its presence and retention time can be confirmed by extracting the key mass 272 ion (Figure 7). Detailed close comparison of the OFN extracted ion signal from the original data file and the DBC-reprocessed file shows that the peak area data is identical. This indicates
that no mass data contributing to the chromatographic peak is lost during the DBC process and that full quantitative consistency with the original data is been maintained. Extracted Ion 272 Identical OFN Data Original GCMS data ClearView reprocessed data Figure 7: Diesel/OFN data in extracted ion mode (The match between the OFN 272 signals (original and reprocessed data) is so perfect that are shown inverted for clarity.) Such quantitative consistency cannot be claimed for conventional static background subtraction methods - as used by many commercial GCMS software systems to reduce background contribution to a TIC profile. The limitations of static background subtraction for ultra-trace level measurement are illustrated below in Figures 8 and 9. Original GCMS data DBC reprocessed data Figure 8: Overlaid extracted ion (mass 273) C 13 chromatograms (original and DBC reprocessed) from the high speed GC- BenchTOF analysis of diesel spiked with 3 pg OFN. Data offset by 5% along the x axis for clarity. Figure 8 show the extracted mass ion 273 chromatograms from the original and DBCreprocessed data files for the C 13 isomer of OFN. OFN molecules containing one C 13 atom in place of a C 12 atom are only ~11% of the abundance OFN molecules containing all C 12 so levels shown in this data are in the order of 300 femtograms. Despite working at these super-trace levels, close examination of the original and reprocessed data once again shows an identical response i.e. DBC reprocessing has not compromised the mass ion data contributing to the OFN peak area at 2.084 minutes. However, later in the chromatogram (from about 3.6 minutes) the positive impact of DBC in terms of removing the contribution of background 273 mass ion present in the unresolved diesel matrix can be seen.
If, alternatively, a conventional manual, static mass ion fragmentation (MIF) background subtraction is performed at 4.2 minutes and applied to the original GCMS data file then the broad hump maximising at this time can be reduced to something similar to that seen in the DBC profile (Figure 9). However, the C 13 OFN peak at 2.084 minutes completely disappears under these conditions highlighting the limitations of static subtraction. No OFN visible (red plot) after static subtraction of original data 3 pg OFN (C 13 ) post CV reprocessing Figure 9: Overlaid extracted ion (mass 273) C 13 chromatograms (DBC reprocessed (blue) and static background subtraction (red)) from the high-speed GC- TOF analysis of diesel spiked with 3 pg OFN. As above, data offset by 5% along the x axis for clarity. An investigation of the impact of DBC reprocessing on GCMS quantification is illustrated below (Figures 10 and 11). Figure 10 shows the DBC reprocessed data from three consecutive high speed GC-TOF analyses of OFN-spiked diesel injected manually. A region of interest is highlighted in the upper plot, and is shown expanded to full-scale below. Visual inspection of the overlaid data in close-up shows the excellent run-to-run reproducibility obtained using DBC. This degree of repeatability, makes it possible to reliably implement sophisticated, time-dependent integration events within a GCMS data analysis method for application to all the runs in a data set For example, note the integration event inserted within the TIC (see blue oval) for applying a free-hand baseline between 3.1 and 3.17 minutes such that peak areas within this time window can be summed. The area reproducibility obtained was 4% RSD remarkable for summed peaks and manual injection.
Figure 10: Three overlaid repeat high speed GC-BenchTOF MS analyses of OFN-spiked diesel, all reprocessed using DBC. Full run (top), expanded area (bottom). Enhanced repeatability and baseline assignment can be seen. Figure 11: Overlay of the original GCMS data from the series of 3 diesel/ofn analyses as shown in figure 10 but without DBC reprocessing. The decrease in performance with respect to repeatability and chromatographic resolution are obvious. If the same data analysis method (i.e. incorporating the integration event for assigning a baseline and summing peaks between 3.1 and 3.17 minutes) was applied to the original GCMS data obtained from this series of three runs (Figure 11) it had no effect In other words, no baseline was constructed on any of the 3 original GCMS data files. When a second attempt was made to implement a timed integration event using one of the original GCMS data files, a baseline was drawn for this one file. However, when this data analysis method was subsequently applied to the other two original GCMS data files in the set, y-axis displacement (differences) between the runs meant it could not be implemented and no baseline was drawn. There are cases, such as that shown for diesel above where the eliminated background ion contribution comprises an unresolved hump of material from the sample itself This is relatively common in GCMS for example during the analysis of hydrocarbon fuels (as shown) and in TD-GCMS analysis of chemical emissions from materials. In all such cases both the original and reprocessed data files would be required for a complete analysis:
The original data file would allow the total unresolved area to be determined e.g. for simulated distillation-type applications in the petroleum industry or TVOC measurements in materials emissions testing The reprocessed data file would allow enhanced quantitative and qualitative analysis of the resolved components. Impact on quantitative accuracy and repeatability when common ions are present? As an ultimate challenge to the quantitative performance of the new algorithm, the landfill gas standard data (Figure 1) was used to evaluate how well the software performed when trying to discriminate background ion contributions from chromatographic data when the peak and background contained one or more common ions. As an example, the peak at 18.02 minutes (butanethiol) was examined. The spectrum of butanethiol includes a small contribution from mass ion 55 which was also present in the background. Figure 12 shows a spectrum obtained from the reprocessed data at the apex of the peak (which was confirmed as butanethiol by library search) and a spectrum from the background immediately after the peak. Both show a contribution from mass ion 55. * * Figure 12: Original and reprocessed data files for the landfill gas standards, highlighting butanethiol at ~18.1 minutes, shown together with the peak apex spectrum from the reprocessed data and a spectrum of the background signal immediately after the peak in the original GCMS data To evaluate whether the new software had correctly compensated for the mass ion 55 contribution due to the background and eliminated it from the reprocessed
data without impacting the contribution of mass ion 55 to the peak, the extracted ion 55 data (original and reprocessed) were overlaid with appropriate offset and alignment. (Figure 13). As mass 55 is a minor component of the butanethiol spectrum, any loss during reprocessing would be immediately apparent. However, the degree of correlation was remarkable, indicating that the new DBC algorithm appeared to compensate accurately for the background, without compromising chromatographic data, even when common ions were present. Figure 13: Extracted ion data for the original and reprocessed data. The overlaid trace was obtained by offsetting and aligning downwards the extracted ion response in the original data to match that in the reprocessed data Increasing the scanning mass range and reducing air/water interference Another key issue for trace TD-GCMS work can be air and water interference - Monitoring key ultra-volatile atmospheric pollutants such as freons and C 2 hydrocarbons requires use of super-strong, water retentive sorbents such as carbon molecular sieves which can retain significant masses of water and traces of carbon dioxide during environmental air monitoring studies. The impact of these interferences can usually be reduced by judicious use of dry purging both before and during thermal desorption-gcms analysis. However, traces may remain and cause baseline anomalies, particularly in the first part of the total ion chromatogram. In extreme cases, these perturbations can become significant and may compromise quantitative analysis when scanning below mass 45. However, when studying light compounds it is not always possible / advisable to exclude mass ions below 45. Low mass ions may comprise a major part of the spectrum of the light compounds of interest and their exclusion may compromise the match quality of some key target analytes. The ability of the new software to selectively eliminate this more unusual background interference was tested with an extreme example of air monitoring using strong, non-hydrophobic sorbent tubes, with insufficient dry purging and scanning from mass 15 (figure 14). The beneficial impact of dynamic baseline compensation is immediately apparent.
Figure 14: TD-GCMS of humid air sample scanned from mass 15. Original GCMS data file (black). DBC reprocessed data (blue) Summary This paper has illustrated the potential of the new dynamic baseline compensation algorithm to enhance the quality of GCMS data for trace components which is a critical issue for many applications. Improvements in spectral purity have been shown to facilitate more automated / higher quality identifications of trace components while, at the same time, the flatter baselines and reduced interference of the reprocessed data have been demonstrated to offer more sensitivity and more repeatable quantitative analysis without any loss of peak area data. It has also been shown to facilitate scanning from low masses and to apply to both conventional and high speed GCMS data. Over the range of examples tested it seems that the new software offers significant potential for improving the quality of GCMS data in terms of repeatability, sensitivity, ease-of-integration and spectral purity. By reducing the need for time-consuming manual re-integration of complex total ion chromatograms by skilled technicians, it also holds out the possibility of significant reduction in mean time / cost-per-analysis for GCMS applications requiring the detection and measurement of trace target analytes in uncharacterized and complex real-world sample matrices.