Select Observations and Thought-Provoking Methods
If you have been active in the world of spectroscopy over the last decade, you are aware that Raman has become a widely accepted and streamlined technology for analyte identification. Once considered too advanced in both theory and instrumentation for most users, Raman spectroscopy today encompasses both simple handheld devices from a range of suppliers, including Ocean Optics, and advanced chemometrics and machine learning, which have hidden the difficult analyses behind the curtain.
Ocean Optics has offered modular, embedded and handheld Raman systems for many years, and to further enhance our Raman capabilities and portfolio we are also an industry-leading supplier of Surface Enhanced Raman (SERS) products. These chemistries greatly amplify the analytical Raman signal and therefore allow much lower limits of detection, which is key to applications such as security and food safety. Ocean Optics has sold many thousands of SERS substrates to users testing for pesticides, explosives, illicit drugs and pharmaceuticals, pathogens, fuel markers, and a wide range of samples beyond those.
Raman to Qualify … but to Quantify?
Raman is treated almost entirely as a qualification method rather than a quantification method. Raman peak signatures offer highly repeatable fingerprints for each analyte that can be deconvoluted via any number of numerical methods. This allows detection of complex mixtures and real-world samples that are quite “messy” versus an idealized laboratory sample. But what about concentration? Can Raman tell us how much of our analyte is present? The answer is a cautious “yes,” as we look at some approaches to quantification here.
The most basic approach to quantification via Raman is simple peak intensity. In Figure 1, we look at a sample of the fuel marker 1,2-di(2-pyridyl)ethylene, nicknamed BPE, in various concentrations. We see the BPE dilutions line up in perfect order and with a very strong linear fit on the 1605 cm-1 peak intensities.
This looks great, but it assumes a very steady optical setup and essentially, all experimental parameters as replicable as possible. What if we don’t have a stable setup along with six convenient known dilutions to reference?
When working in more typical spectroscopy modes such as absorbance or reflectance, it is important to account for any optical shifts by performing some type of baseline correction. Peak values, or any values in general, may be skewed by some environmental influence but can be corrected based on certain region(s) known to be independent of changing parameters. The same can be done for Raman spectra, and is perhaps the most critical spectral genre to correct if one wishes to extract quantification information.
Figure 2 shows our BPE analyte with no baseline correction. We can see the spectra hover above the x-axis and have about a 250-count range between spectra.
Around 580 cm-1 there is a nice, Raman-inactive region that is also a relative low-point for all spectra. If we average near those pixels and subtract across the spectrum, we will get a much cleaner arrangement of our concentration trends (Figure 3).
Validating Peak Ratios
Now that our peak values are more meaningful relative to the overall shape of the spectra, let’s look at some ratio trends of the major peaks for our various concentration samples. We processed a series of peak ratios, and will focus on two select scenarios within these ratios.
If we look at the 1197 cm-1 peak against the 1634 cm-1 peak across five BPE dilutions for five replicate runs, there is a highly repeatable ratio for all instances other than the lowest concentration (Figure 4).
Why is this valuable? What this provides is a potential check or additional reference on the measurement. If we were working with a government-subsidized fuel using BPE as the active Raman marker at, let’s say, 10-6 M, we could use this 1197 cm-1/1634 cm-1 ratio as another layer of confirmation that we have what we’re looking for. If that ratio suddenly came out to a wildly different value it would be a flag that something was not right, which may be missed if everything else appeared normal (i.e., someone using a near-identical BPE surrogate to pass the counterfeit check).
Now let’s look at another set of peaks that may give us a slightly different trend behavior. If we still look at the 1634 cm-1 peak but now ratio that against the 1337 cm-1 peak, there seems to be a rough yet consistent trend with BPE concentration (Figure 5).
This fit isn’t as tight as the first ratio we looked at, but there is enough here to give us some idea of where an unknown sample may fall. If we had taken these known samples as our references and then tested an unknown that yielded a 1634 cm-1/1337 cm-1 ratio less than 2, we may not know the exact concentration but could make a well-supported guess that it is below 10-6 M. This may be enough information for an alarm or basic threshold measurement.
While many will take Raman spectra at face value, we have shown here how very basic spectral and peak processing can clean up your combined plots and give you some potentially useful information. This gives us a deeper understanding of the spectral outputs and allows us to blend in additional layers of sample validation and rough concentration determination that were previously unavailable.
For larger data sets of known conditions, these outputs can be used as training data for PCA correlations and more intricate statistical analysis. Ocean Optics has strong experience processing and understanding spectral responses at the most basic and complex levels.