Resource Hub / ChromaBLOGraphy / Cannabis Residual Solvents Using MS Detection – I’m Not Hungry but I’ll Eat My Words Anyway

Cannabis Residual Solvents Using MS Detection – I’m Not Hungry but I’ll Eat My Words Anyway

12 Jul 2016

After coming back from a huge lunch at the Bellefonte Wok, a favorite Restek lunch spot, I’m completely stuffed, but I have to eat my words from a previous blog. In this blog, I made the case against using MS detection for headspace analysis of residual solvents in cannabis concentrates due to interference between the air peak from the headspace injection and propane. It turns out that’s absolutely not the case, as my colleague and Restek air chemist Jason Herrington argued when I first posted the blog.

During a trip to Trace Analytics in Spokane, WA, I had the opportunity to run cannabis residual solvents using the full evaporation technique with headspace GC/MS (FET-HS-GC/MS), and lo and behold, propane is resolved from the air peak as shown in Figure 1.

Figure 1: Propane is Resolved from Air Peak on the Rxi-624Sil MS using FET-HS-GC/MS (25ppm Standard)Figure 1

Propane is still well-resolved even at higher concentrations where band broadening may become an issue (Figure 2):

Figure 2: Propane is Resolved from Air Peak Even at Higher Concentrations (500ppm Standard)

Figure 2

Given the close proximity of propane to air, the two peaks may co-elute under sub-optimal injection conditions.

Although I dismissed the use of MS originally, there are benefits that I failed to mention. In addition to the ability to confirm peak identity using mass spectra, sensitivity is improved over GC-FID for the later eluting compounds like benzene, toluene, and the xylenes (BTX). This is because these compounds produce higher molecular weight fragments (78, 91, 106 m/z), and MS detectors aren’t great at reliably detecting low mass fragments from solvents like propane, methanol, and butane (29, 31, 43 m/z). Luckily, the solvents that produce the lower molecular-weight fragments have much higher regulatory cutoffs than butane, toluene, and xylenes, so detectability via MS isn’t an issue. Figure 3 shows an extracted ion chromatogram of our low standard (0.5ppm) and the signal-to-noise ratio is very good for BTX. Note that this chromatogram was collected in scan mode, so even more sensitivity can be gained from the development of a selected ion monitoring (SIM) method.

Figure 3: Good Sensitivity is Achieved for BTX using FET-HS-GC/MS (XIC of 78, 91, 106m/z, 0.5ppm Level)Figure 3

All this being said, there is one big caveat to the use of MS for analysis of residual solvents in cannabis concentrates: the limited linear dynamic range of MS detectors when compared to FIDs. Where a FID can produce a linear curve over several orders of magnitude, an MS detector has a dynamic range limited to about three orders of magnitude. This means that one curve covering the entire regulatory range (e.g. 0.5 – 5500ppm) cannot be run using MS. In fact, under the conditions listed in the protocol for this method, BTX ions begin to saturate the detector at 250ppm (Figure 4), and almost all major ions for all analytes are saturated at 500ppm (Figure 5). This results in quadratic calibration curves for our analytes over even this limited range of quantification (Figure 4 inlay). If we tried to extend the curve beyond 500ppm, even a quadratic fit would be inappropriate.

Figure 4: BTX are Saturated at 250ppm, Resulting in a Quadratic Calibration Curve (Inlay)

Figure 4


Figure 5: Almost All Compounds are Saturated at 500ppm

chart, histogram

While quadratic curves aren’t inherently bad, most guidance suggests the use of a linear calibration curve if feasible. Since even a quadratic curve wouldn’t cover the full regulatory range, in the cannabis residual solvents analysis protocol I suggest breaking up calibration into high and low segments. These segments will vary by state due to different regulations, but the approach is the same. For some states with very high cutoffs (I’m talking to you, Oregon), a higher split will most likely have to be used for the higher range curves in order to avoid overloading the column. Please note that this protocol has been partially validated, but still needs some work in terms of internal standard choice and extending the range of quantification above 500ppm.

While I learned a lot in the development of this method, I think the most valuable lesson I learned is to never argue with the air chemist when it comes to volatiles analyses.