Non-Target Characterization and Simultaneous Quantitation for Reformulation of Complex Fragrance Samples with GCxGC–MS/FID: An Interview with Elizabeth Humston-Fulmer


The determination of the many chemical components that can make up a fragrance sample is important for multiple objectives, including reformulation, which typically involves determining the identification and relative amounts of components in a sample. As it is usually not practical to quantify and calibrate all the fragrance components using reference standards, workflows to achieve fragrance analysis use mass spectrometry (MS) for identification with a flame ionization detector (FID) for relative area percent quantification. At the 2024 Pittcon Conference, Elizabeth Humston-Fulmer of LECO Corporation presented an improved workflow which uses two-dimensional gas chromatography (GCxGC) with dual time of flight (TOF) MS/FID detection. Humston-Fulmer has found that this process yields reliable identification and area percent quantification using a single injection, rather than the traditional multiple injection requirements. Shortly after Pittcon, Humston-Fulmer discussed this analytical method with us.

Liz Humston-Fulmer is an Application Chemist at LECO Corporation in Saint Joseph, MI. She received her B.S. in Chemistry from the University of Pittsburgh and her Ph.D. in Analytical Chemistry from the University of Washington. She currently works with GC, GCxGC, TOFMS, and FID on a variety of food, flavor, and fragrance applications.

Liz Humston-Fulmer is an Application Chemist at LECO Corporation in Saint Joseph, MI. She received her B.S. in Chemistry from the University of Pittsburgh and her Ph.D. in Analytical Chemistry from the University of Washington. She currently works with GC, GCxGC, TOFMS, and FID on a variety of food, flavor, and fragrance applications.

From a separation standpoint, what are some of the challenges associated with fragrance analysis and formulation?

There are a few reasons why fragrance analyses and formulation tasks can be challenging. One of the reasons is that the samples tend to be pretty complex; there are often hundreds of analytes in a fragrance mixture, which can make it hard to determine all the individual components. Another reason these tasks are challenging is because they are usually non-targeted, and we don’t know which analytes will be important when we start. You are also trying to learn what analytes are present and their relative amounts, so it is a lot of information to try to understand about a complex sample.

In your presentation, you demonstrated an improved workflow using GCxGC–TOF-MS-FID detection to yield reliable identification and area percent quantification in a single injection. What made you decide to choose this analytical technique?

We really like the approach of using GCxGC with dual MS and FID detection for these kinds of projects because we get a lot of information from a single injection. GC is a common tool for fragrance analyses because the analytes that contribute to the aroma profile tend to have volatilities that are a good match with GC. When the samples are complex, adding a second dimension of separation with GCxGC can be helpful because the sample is separated with two complementary stationary phases in the same analysis. If analytes are coeluted on one phase, they often separate on the other, so we end up with more of the individual analytes chromatographically separated. This improved separation sometimes reveals analytes that were coeluted to the point that they were hard to find in the 1D data, so we can usually learn new information about the samples this way. It’s also helpful to have both MS and FID data because we use the MS information to identify the non-target analytes and the FID information for area percent determinations. Injecting once and splitting the effluent to both FID and MS saves us from having to make separate injections for each detector.

What were some of the specific hardware and data analysis methods used for this workflow to ensure reliable quantitative results?

The hardware and software for this system were both designed carefully to try to ensure that the peak areas would be reliable so they could be used for area percent calculations. The modulator and the splitter were two components of the instrument that received a lot of attention. The system uses a reverse fill/flush (RFF) flow modulator where the primary column effluent is collected in a sample loop, which is then flushed to the second column at the modulation period interval. This type of modulator can achieve total transfer, where everything that is on the first column is transferred to the second column, but there are a handful of parameters like flow, fill time, loop size, and so on that must be considered to make sure that happens. If the parameters are not set appropriately, the sample loop can be overfilled or underflushed, which can cause inconsistencies in the peak areas. This system has method checks and calculators built into the software to verify all these conditions and ensure that all the primary column effluent gets on to the second column and that everything will be reliable. In addition to that, the MS and FID splitter was designed with back-pressure regulation as a way to ensure that there would be a consistent split ratio and consistent flow to both detectors, even when the GC temperature conditions change through the separation. All of this makes sure that the FID peak areas are reliable so you can use those for area percent to better understand your sample.

How does your work differ from previous approaches by yourself or others for fragrance analysis?

Most of the work we have done with fragrances in the past was either a non-target qualitative analysis or a targeted quantitative analysis. We would use GC or GCxGC with MS to characterize the sample and try to learn what analytes were present, or we would analyze calibration standards for specific target analytes to get quantitative information. This new approach maintains the non-target qualitative information and adds some approximate quantitative information for these non-target compounds. FID area percent is more approximate than calibration standards, but it’s a common approach for describing relative amounts of the components within a sample, and we can reliably add that information to our non-target characterization.

A lot of existing workflows for reformulation will also use MS for identifications and FID for area percent calculations. Sometimes this involves multiple injections to the different detectors or multiple injections to different GC columns with different stationary phases to try to address coelutions and to get both the identification information and the FID peak areas for relative quantitative analysis. This approach aims to achieve this without multiple injections.

Please summarize your findings of what was learned from this study.

We used this system to explore a variety of different fragrance and perfume samples in a non-targeted way. We looked at standards, essential oils, and perfumes. In each case, we were able to get good characterization information about the sample. We searched the MS data against spectral libraries to get the analyte identifications. We also used the GC information, both first dimension retention index (RI) and second dimension eluted position, to support and add confidence to these identifications. The MS peaks were then linked to their associated FID peaks, so we could use that for area percent determinations. This gave us good reformulation information on each of the samples.

We found several analyte pairs in these samples that coelute and appear as a single chromatographic peak in the 1D data that were chromatographically separated in the second dimension with GCxGC. If we hadn’t used GCxGC for our analysis, we would have overcounted the amount of one of the analytes and missed the other in each of these pairs.

We were also able to make some interesting comparisons between different samples. One of the comparisons was a name brand perfume to drug store imitations of that brand. We could see how the aroma compounds and diluents differed. For example, all the perfumes had analytes with “musk” and “floral” aroma descriptions, but there were differences in the specific compounds with those aroma notes.

Can you please summarize the feedback that you have received from others regarding this work?

This approach can both save time and improve the characterization of the sample.

Do you imagine these techniques to be adaptable to other applications in other industries?

We do see applicability for these tools for other samples that have similar challenges and goals. This project was characterizing fragrances and perfumes, but we have other projects in our laboratory where the same tools are being used for other types of samples. For example, we are also looking at alternative aviation fuels where the quantitative aspect is important for meeting blending specifications and the non-target qualitative information can be used to screen for compounds that could cause fouling of engines or other performance problems.

What are the next steps for your research?

We are excited about the capabilities of this system and how we can get so much information from one injection. We have plans to explore additional fragrance samples and are also looking at analyzing other related samples.

Related Videos
Toby Astill | Image Credit: © Thermo Fisher Scientific
Robert Kennedy
John McLean | Image Credit: © Aaron Acevedo