Investigating Arson Attacks Using Multidimensional Gas Chromatography (GC×GC)

Q. You recently developed a chemical fingerprinting method to analyse petrochemicals in arson investigations for forensic analysis. How did this project arise? (1)

OJ: Like many of the most enjoyable projects I have been involved with, this one came about from somebody asking: “I wonder if that would work?” I have been working with Jim Pearson (one of the coauthors of this paper (1) from the Victoria Police Forensic Services Department for several years now. We have completed a number of research projects together and Jim regularly gives a seminar on forensic science that goes down very well with our undergraduate students. One time he gave this talk, he mentioned the difficulties of identifying accelerants in arson cases because of the complexities of the sample.

Imagine you are investigating an arson attack and have identified that a particular accelerant was used. There is a suspect in custody and when arrested they had a container of petrol in their car and petrol was on their clothes. Is this enough to convict, or even to charge the suspect? Clearly not. Just because somebody has some petrol in their car or on their clothes does not mean much; on its own, it is just coincidence. However, if you could say the petrol in the car was the same as that on the clothes, and both those samples matched the specific brand and age of the accelerant used in the fire you are investigating, then you might be more confident that the suspect was potentially involved.

This sounds simple but as anyone who has ever analysed them can tell you, hydrocarbons like petrol are very complex. Aside from that, petrol samples fresh from the tank are one thing but in a real arson case the samples might have been out in the environment for some time which makes a difficult analysis even harder.

I had been using multidimensional gas chromatography for my metabolomics work (metabolomics is the comprehensive analysis of small biological metabolites) and I asked Jim if he would be interested in exploring if it would work for fuels. He said he would. We happened to be setting up our third‑year undergraduate science projects at RMIT at the time so had the analytical facilities and willing students ready to go relatively quickly.

We were able to see variation in different fuel types, and even different brands of fuels fairly early on. Different manufacturers put a diverse range of additives in their petrol on purpose, which makes differentiating them a little easier. Jim then asked if we could look at how the chemical fingerprints changed over time as well? By this he meant does brand “X” that has been sitting out in the open air for a week look like a fresh sample and can you still tell it from brand “Y”? The ongoing feedback and collaborations with Jim were really critical in getting this project going and making sure it was relevant.

Q. Why is chemical fingerprinting of hydrocarbons useful in this instance? What information did you hope to obtain?

JP: We used chemical fingerprinting of hydrocarbons in a forensic application, more specifically to profile readily available petrochemicals that could be used as a potential accelerant. Gas chromatography–mass spectrometry (GC–MS) is the current chemical profiling technique used during arson investigations and it clearly has major limitations in separating complex petrochemicals that contain hundreds of compounds, let alone distinguishing between the same petrol coming from different petrol stations or determining the time a fire started.

Our approach to chemical fingerprinting is useful in arson cases as it allows the separation of the components of complex fuels, such as diesel and kerosene, that were previously co-eluting using standard GC–MS methods. This gain in separation power gives forensic scientists/investigators an advantage and enables them to identify “markers” that could be associated to a specific petrol station or a house where the fuel could have been stored before being used in a crime scene. From our experiment, we hope to provide chemical fingerprints for easily available petrochemicals and investigate their weathering patterns.

Q. You used multidimensional GC with flame ionization detection (FID) and multivariate analysis. Why did you choose multidimensional GC? What is novel about this approach and what benefits does it offer the analyst?

OJ: Gas chromatography (GC) is the obvious analytical tool to use for volatile samples like petrol. We chose comprehensive multidimensional GC (or GC×GC) for the extra separation space and resolving power it provides. You can also infer some structural information from peak location in GC×GC.

Resolving power (in a reasonable time) is what it is all about in separation science Unfortunately, the chromatograms from petrol are very complex with lots of overlapping peaks and unresolved humps. Using standard GC just doesn’t provide enough information to separate out sample types.

With GC×GC (and its younger brother LC×LC) the increase in resolution power is based on linking two columns with different stationary phases, for example nonpolar and polar, via a modulator. Fractions exiting the first column are refocussed via the modulator and then injected into the second column where they are separated again based on a different chromatographic selectivity. In a sense, the addition of the second dimension can be thought of as being like standard GC with an extra chemically‑selective detector added on.

The total peak capacity of a two‑dimensional chromatography system is almost the product of peak capacities of the individual dimensions, so the resulting separation space far exceeds that of standard GC systems. This extra resolution and separation power is one reason the oil and gas industry were early adopters of the technology and it is now widely used in the analysis of all types of oils, including high value products like olive oils.

Q. Can you elaborate on the role of multivariate analysis in this research?

JP: Multivariate analysis was critical to this project. The GC×GC plots of the petrochemicals were very distinct and the difference in the chemical fingerprints of each unweathered sample was obvious, making it straightforward for identification by visual comparison alone. The challenge arose when studying the weathered samples. As a result of the high similarity of chemical components in the petrochemicals matrix and the minute differences in its content between the chosen weathered timepoints, a multivariate analysis was essential to reduce the dimensionality (ironically) and, in turn, maximise the visualisation of the variance. This can make patterns in the dataset clearer, which is just what we needed in this case, and allowed us to separate the weathered fuels at different timepoints.

Q. What were the main challenges you encountered when developing this method, and how did you overcome them?

OJ: There were two main challenges. The first was around which was the best column combination to use. Getting the best separation in GC×GC relies on using columns that are as statistically independent (different) as possible. We spent a fair bit of time running the samples through different column set ups to optimize the separation before we could run the majority of the samples.

The other major challenge was the data analysis. It will come as no surprise to learn that using two dimensions of separations, leads to far more complex data plots. GC×GC files are much bigger than their one-dimensional counterparts, with final file sizes of 500 Mb or more.

Dealing with two dimensions of separation requires a large amount of processing power to enable visualisation and analysis (both quantitative and quantitative). This introduces two problems to be solved. One, how to collect and save the data files, and two, how to easily handle and process the acquired data. You can export chromatograms as text files, but they don’t go into Excel very easily (although you can do multivariate analysis this way). To handle the large data sets we used the R programming language which is very powerful and open-source. It allowed us to process and analyse the large datasets very quickly. We both also learned some new skills in coding along the way.

Another (minor) challenge was getting the samples. We needed a lot of different samples for the project to work so we had to do a bit of travel round Melbourne getting different brands of petrol for analysis and got a few funny looks at the petrol stations we visited as we only needed a small amount for testing rather than a full tank of fuel. The head of the RMIT Chemistry store very kindly helped us with this but did ask that I please let her know in advance if I wanted to do that sort of project in future.

Q. What were your main findings?

JP: We successfully separated numerous components in petrochemicals that were previously co-eluting using GC–FID. This paper is the first report of the application of GC×GC–FID to generate and compare detailed and sensitive chemical fingerprints of easily available ignitable liquids (1). We were able to show that we could distinguish between various petroleum products available on the market, but that multivariate analysis is needed to differentiate between various ignitable liquids and, moreover, to make a distinction between ignitable liquids that have been weathered.

It just goes to show that an “all-singing, all-dancing” mass spectrometer is not always needed to get good results in analytical science.

Q. Have you used GC×GC in any other areas of this research? Do you see it becoming more commonly used in routine analysis?

OJ: Although is it not that widely used compared to standard GC at present, GC×GC is actually a fairly mature technology in many areas. A simple search will illustrate the wide range of potential uses. GC×GC does have what might be termed a bit of an image problem in that it is often seen as overly complex and too difficult to use and operate for routine analysis. The instruments themselves are also more expensive than a standard GC; so the method is not without barriers to newcomers.

That said, I certainly see interest and use in this method growing in future. To see why, just step into any modern laboratory and look at the type of samples currently of interest to the analytical chemist. Such samples are amazingly complex, be they protein extractions, metabolites, water/soil samples or food products. There is also an increasing need to measure more compounds than ever, even in targeted analysis. For example, not long ago an analyst looking at pesticides may only need to test for 20–30 compounds, now they may need to look for 6 000 or more. Although not yet widespread, the increased separation power of multidimensional chromatography has great potential to meet this need in a large variety of areas, particularly for the analysis of compounds that are too sensitive for mass spectrometry (MS). Such areas include, but are not limited to, animal health, biomedicine, dairy science, environmental toxicology, food science, functional genomics, pharmacology, plant biology toxicology, and the -omic sciences.

The latest version of the Human Metabolome Database (version 4.0) (2), for example, lists 114 100 individual entries. This is nearly a three-fold increase from version 3.0 and even this large database does not include all possible compounds/isomers. This means that the actual number of human metabolites (both endogenous and exogenous) could potentially be far higher. Typically, however, metabolomics studies identify only around 100 metabolites and there may be many compounds listed as “unidentified”. This means that potentially useful information is being lost from publicly-funded studies each year. GC×GC was used in the early days of metabolomics, particularly for plant samples but this seems to have died back a bit recently. I think GC×GC still has a lot to offer in this area. I see the technique being of particular value in environmental toxicology‑based metabolomics and exposomics because you could perhaps measure the concentrations of biological compounds and pollutants in a sample at the same time. Then you could start to correlate pollutants with specific biochemical changes.

Q. Any final comments on mutidimensional chromatography?

OJ: Multidimensional chromatography (both gas and liquid) has a reputation of being too complex and difficult for everyday laboratory use and that you have to be an expert to use it. This understandably puts people off trying – even when it might be useful. GC×GC (and to a growing extent LC×LC) is actually quite a mature science with robust, established methods and software, and a growing number of applications. I encourage everyone to give it a try if you get the chance.

References

1. J. Pandohee, J.G. Hughes, J.R. Pearson, and O.A.H. Jones, Sci. & Justice: J. Foren. Sci. Soc. 60(4), 381–387 (2020). doi: 10.1016/j.scijus.2020.04.004
2. D.S. Wishart, Y.D. Feunang, A. Marcu, A.C. Guo, K. Liang, R. Vázquez-Fresno, T. Sajed, D. Johnson, C. Li, N. Karu, Z. Sayeeda, E. Lo, N. Assempour, M. Berjanskii, S. Singhal, D. Arndt, Y. Liang, H. Badran, J. Grant, A. Serra-Cayuela, Y. Liu, R. Mandal, V. Neveu, A. Pon, C. Knox, M. Wilson, C. Manach, and A. Scalbert, 2017. HMDB 4.0: the human metabolome database for 2018. Nucleic Acids Research 46(D1), D608-D617.

Oliver Jones is a Professor of Analytical Chemistry and Associate Dean (Biosciences and Food Technology) at RMIT University in Melbourne, Australia.

Jessica Pandohee was a PhD student at RMIT but is now a Research Fellow at the Centre for Crops and Disease Management at Curtin University in Perth, Australia.

E-mail: oliver.jones@rmit.edu.au
Website: www.rmit.edu.au/contact/staff-contacts/academic-staff/j/jones-professor-oliver