Detecting Volatile Compounds with Colloidal Nanoparticles and Portable Raman Spectroscopy

Detecting volatile organic compounds (VOCs) with surface-enhanced Raman spectroscopy (SERS) usually means working with specialized substrates that can be costly and hard to reproduce.

Amy Colleran, who is a Postdoctoral Researcher in the Centre of Metabolomics Research at the University of Liverpool, took a simpler route. During her PhD as part of a collaboration with Unilever to detect the volatile thiols behind underarm odor, she and her colleagues developed a method in which a gaseous analyte is bubbled directly into colloidal silver or gold nanoparticles while a portable 785-nm Raman spectrometer records the spectra in real time.^1,2 The work uses 3-mercaptohexanol (3MH) as a model compound.² 3MH aggregates hydroxylamine-stabilized silver nanoparticles without an aggregating agent, and the team found that a flow rate of 50 mL/min gave the most reproducible results.

Colleran presented her work and team’s findings at the 2026 SciX Conference in Sparks, Nevada. In this interview, Colleran sat down with Spectroscopy to explains why she chose colloidal nanoparticles, the practical challenges of getting a gas to interact reliably with a colloid, and why indole needed an added sodium chloride aggregating agent.

What pushed you toward colloidal nanoparticles, and how do cost, ease of synthesis, and reproducibility compare in practice?

At the time during my PhD, I was looking at trying to develop a SERS method for detecting volatile thiol compounds associated with underarm odor. We were looking for methods that were easy to replicate, inexpensive, and demonstrated good SERS enhancement for the gaseous thiols. In our previous research, we saw that one of the thiol compounds of interest (3-mercaptohexanol; 3MH) appeared to show strong enhancement when measuring the compound in solution using hydroxylamine-stabilized silver nanoparticles and no aggregating agent.¹ As a result, we wondered if we would see the same effect if we bubbled gaseous 3MH into silver or gold colloidal nanoparticles.² I needed a solution that was cost effective, usable for SERS nonexperts, and one that uses established methods and showed good reproducibility and enhancement. As there are a wide range of established methods for synthesizing colloidal nanoparticles which have been used across many different laboratories and are generally cheaper than other SERS substrates, they made for an ideal solution if they worked. Additionally, with optimization for consistent nanoparticle aggregation, we can achieve highly reproducible signal enhancement. This new method we developed utilizes the many benefits of using colloidal nanoparticles, making it a highly attractive choice for the detection of volatile compounds and motivating us to further investigate the capabilities of the system.

Can you walk us through the experimental setup? What practical challenges came up in getting the analyte to interact reliably with the colloid?

The actual setup is pretty straightforward and for those interested please read our latest paper on this.² First, 20 mL of air is pumped from a syringe using a syringe pump at a consistent flow rate through a piece of polytetrafluoroethylene (PTFE) tubing and into a conical flask containing 5 mL of a sample with our analyte of interest. This process helped to generate the gaseous analyte and cause the gas to flow through a second piece of PTFE tubing and bubble into 2 mL of colloidal nanoparticles. Before, during, and after the bubbling process, spectra were continually collected using a portable CBeX SnowyRange 785-nm Raman spectrometer in point-and-shoot mode with a 1 s acquisition time to track how the spectra changed during the bubbling process and while the nanoparticles were aggregating.²

The main practical challenges we initially faced when developing this method was making sure that we were primarily measuring gaseous particles and not aerosols, choosing a flow rate for pumping the air to allow for consistent interactions between the analyte and nanoparticles, and having tube lengths that were consistent between measurements for reproducible results. We also had to try and reduce the risk of any gas escaping from the system and that the tubing in the nanoparticles was at a suitable position to allow full interaction between the nanoparticles and the analyte. To prevent any aerosols entering the solution, we placed a frit with a 10-μm pore size on the end of the tubing facing the analyte and going from the analyte to the nanoparticles. We also had consistency for the two pieces of tubing, making sure this was measured every time and used the bottom of the point-and-shoot adapter head as the marker for how far to put the piece of tubing into the nanoparticle solution to keep it consistent between experiments. Finally, to try and reduce the amount of volatile loss, the tubing was held tightly in place in the flask with the analyte using a two-holed stopper and Safety Chemical Analysis Technology (SCAT^TM) fittings. The other piece of tubing in the nanoparticles was held tightly in place with Parafilm wrapped around the tubing and opening of the vial.

Why did you start with 3-mercaptohexanol and hydroxylamine-stabilized silver nanoparticles? What makes 3MH a good model analyte, and how important was the thiol group's affinity for silver, reflected in the ν(C–S) band at 634 cm⁻¹, to that first successful measurement?

3MH is definitely a different model compound to test when evaluating a new SERS technique. However, the reason this was chosen to test first was because of my first paper where I was looking at multiplex detection of four thiols associated with underarm odor in solution using SERS. When I was optimizing the experiments, I found that 3MH aggregated hydroxylamine-stabilized silver nanoparticles without an aggregating agent at concentrations of 444 ppm in solution. We wondered if the same could be seen if we tested the compound as a gas.

Luckily, we found that it worked in a similar way and caused nanoparticle aggregation. If 3MH did not show an affinity to the silver and did not cause aggregation, it would have been much harder to understand which analytes could initially work for this method. As 3MH only has one thiol group and only one functional group which will bind to the surface, the compound will orientate itself to the surface from the metal–sulfur bond and enhancement is primarily seen for the peaks corresponding to the bonds and vibrational modes closest to the thiol group. Additionally, it is a compound that is relevant to different applications, including food and beverage analysis as well as underarm odor, making it a suitable model for different areas of research. Finally, in comparison to other thiol compounds like thiophenol, it is a lot less toxic; therefore, it can be used in higher concentrations for detection while trying to optimize the method.

Why do you think 50 mL/min gave the most reproducible results, and is there a trade-off with higher flow rates?

When we tested eight different flow rates between 2–50 mL/min, we noticed an interesting trend in the change in peak height in the linear region of the kinetic curve over time before the peak height plateaued. It appeared to increase up to 8–10 mL/min and then decreased again. At the lower flow rates, it is understandable why this is the case as 20 mL of 3MH gas is being flowed into the nanoparticles over 10 minutes meaning less 3MH gas is interacting with silver nanoparticles over time causing slower aggregation. However, at 50 mL/min, 3MH is bubbled in over 24 s, meaning there is fast perturbation to the mixture of 3MH and nanoparticles, which means that after the bubbling process, any further interactions between 3MH and the nanoparticles relies on just the movement of both the nanoparticles and 3MH. Therefore, this practice likely slows down the aggregation process and may also contribute to the reproducibility as the flow rate is fast enough to cause immediate aggregation of the nanoparticles, and it is likely to be pretty consistent for each measurement. If the flow rate was higher though, there could be an issue with the aggregation of the nanoparticles in which the analyte is added too quickly for the bubbling to have much effect on the mixing process. Therefore, the SERS signal of 3MH would rely more on the aggregation because of the Brownian motion of the nanoparticles and 3MH to make the bond and cause aggregation. As a result, it is likely that the aggregation process would not be as reproducible between replicates.

Indole needed NaCl added as an aggregating agent. What does that tell us about which analytes suit this approach without modification, and how might users adapt the method for molecules that don't bind strongly to silver?

Indole needing sodium chloride added as an aggregating agent tells us that the analyte needs to bind to the nanoparticle used and also needs to cause some form of aggregation of the nanoparticles. Thiols are able to do this because they are weakly acidic, which allows for some disruption to the surface charge of the nanoparticles and for aggregation to occur, especially at higher concentrations. However, indole is not very basic and much less so in comparison to other nitrogen-containing compounds like ammonia. Therefore, it struggles to promote nanoparticle aggregation. If the molecules don’t cause this, then apart from adding an aggregating agent to the system, other methods around this could be to add a SERS tag to the nanoparticles to ensure that the analyte is captured and close to the surface to measure the signal. Doing so makes the method much more analyte specific, and different tags could be used to target different VOCs.

Given that colloidal nanoparticles pair with portable Raman spectrometers, which real-world applications do you have in mind, such as food and beverage aroma analysis, environmental monitoring, or breath analysis? What still needs to happen on quantitation, detection limits, or selectivity in complex gas mixtures before it's ready for field use?

Overall, we think that the method could be applied to numerous fields including aroma analysis, environmental monitoring, or detecting VOCs related to breath or other media that can provide further information on human health. There are still many steps that need to be taken before it is ready to use in real-life applications, including quantification, selectivity, and the ability to perform multiplex analysis. Additionally, there is still plenty of optimization, which should be performed on the system; in particular, the system needs to work with an aggregating agent to get consistent results with high SERS enhancement. Having greater containment of the system to reduce any gas loss and to have some control over environmental factors influencing the system would also make a difference to improve results and make sure there is no analyte loss. A more automated control with starting the syringe pump would also make the system easier to use and would significantly improve the overall system for real-world applications.

References

1. Colleran, A.; Lima, C.; Xu, Y.; Millichope, A.; Murray, S.; Goodacre, R. Using Surface-enhanced Raman Scattering for Simultaneous Multiplex Detection and Quantification of Thiols Associated to Axillary Malodour. Analyst 2024, 149 (15), 3989–4001. DOI: 10.1039/d4an00762j
2. Colleran, A.; Gotts, N.; Murray, S.; Goodacre, R. Analyzing Volatile Organic Compounds (VOCs) with SERS by Bubbling into Colloidal Nanoparticles. Analyst 2026, 151 (20), 5900-5910. DOI: 10.1039/d6an00673f