Quantitation of Cannabinoids in Dried Ground Hemp by Mid-Infrared Spectroscopy

A novel, small, portable, general purpose quantitative mid-infrared (IR) spectrometer has been invented and applied to the analysis of dried, ground hemp. The unit was calibrated using cannabinoid concentrations determined by high performance liquid chromatography (HPLC) at a state licensed laboratory, and mid-IR spectra measured on the same samples. The analyzer was validated using the leave-one-out cross validation method. Mid-IR calibration models for delta-9-tetrahydrocannabinol (Δ⁹-THC), tetrahydrocannabinolic acid (THCA), total THC, cannabidiolic acid (CBDA), cannabidiol (CBD), total CBD, cannabigerolic acid (CBGA), and cannabichromene (CBC) were constructed. The accuracy for the determination of total THC, which is important in determining the legality of a hemp crop, is ±0.04 weight percent, more than sufficient for compliance testing. The analyzer requires little sample preparation, features push button operation, produces results in 2 min, and at a cost of $0/sample. Hemp is a naturally variable material, so obtaining representative data on a grow requires averaging results from many samples. The speed and ease-of-use of mid-IR spectroscopy makes this feasible, as opposed to chromatography where typically only one or a few samples from a grow are analyzed so representative data are not obtained. Applications of this analyzer for hemp farmers, hemp extractors, state regulators, and law enforcement are discussed.

Because of the 2018 Farm Bill, growing industrial hemp in the United States is legal if the sample contains not more than 0.3 dry weight percent (wt.%) total tetrahydrocannabinol (THC) (1,2). There is a need then to test hemp samples to insure they comply with the new law. In the past, cannabis and hemp have been analyzed for cannabinoid content via high performance liquid chromatography (HPLC) (3,4) or gas chromatography (GC) (5). However, chromatography suffers from several problems. Samples must be weighed, ground, extracted, vortexed, diluted, and filtered before injection (3–5). To perform these many manual sample preparation steps properly takes a skilled analyst several minutes and involves the use of expensive consumables. Also, chromatography runs can take at least 5 min or more (3–5), which when combined with the sample preparation time means it takes at least 10 min to analyze one sample. Between the consumables and labor, the cost per sample can be $20 or more. It is hard to imagine laypeople having the time or skill to perform these steps properly, which is why cannabis regulations require highly trained scientists to operate chromatographs (2).

Another concern with chromatography is the lack of representative sampling. Cuttings from adjacent plants in a grow, and even buds from the same plant, can vary by several weight percent in their cannabinoid content. This is illustrated in Table I, which shows the results for five different cannabis strains. The numbers represent weight percent tetrahydrocannabinolic acid (THCA) as measured by HPLC, and the low, medium, and high values represent samples take from different positions on the same plant (3).

Plants from five different strains in Table I are represented. Note that the potency variation across an individual plant can vary by almost 3 wt.%. This means cannabis is a heterogeneous, naturally varying material. The scientifically correct way then to sample a cannabis grow is to collect composite samples from many places in a grow, analyze them, and average the results (6,7). This may mean dozens or even hundreds of samples need to be analyzed, and these analyses must be done over time to insure a hemp grow does not go above the 0.3% total THC limit. Analyzing this many samples by chromatography would be prohibitively expensive. Because of the time, expense, and trouble of chromatographic analyses, typically only one sample from a grow is analyzed. This is by definition not representative (6,7). The danger here is that by using chromatography hemp farmers may be obtaining nonrepresentative data on their grow, misleading them as to its legality and economic value.

Another problem with chromatographic techniques is the lack of standard reference materials and methods for hemp analysis, leading to the problem of inter-laboratory variation—different laboratories obtaining markedly different cannabinoid values on the same samples (8–10). A large part of this problem is variation in chromatographic sample preparation techniques across cannabis laboratories resulting in varying extraction efficiencies and significant offsets between potency values on the same samples (8). An analysis technique that involves little sample preparation would go a long way towards solving this problem. There is then a need for a faster, easier, and less expensive test with minimal sample preparation to measure cannabinoids in hemp, and that will allow more representative data to be obtained on hemp grows.

Mid-infrared (IR) spectroscopy has been used for decades for qualitative and quantitative analysis of samples (11–13). More recently it has been used to accurately measure cannabinoid profiles in marijuana plant material (14), to study the rate and mechanism of potency degradation in cannabis oils (15), to measure cannabinoid and terpene profiles in cannabis oils (16,17), and to measure potency in cannabis distillates (8). The latter study found that for measurements of THC in cannabis distillates mid-IR was four times more precise than HPLC. Mid-IR analyses feature little sample preparation, no expensive consumables, and a 2-min sample analysis time.

Given the success of mid-IR in measuring cannabinoids in marijuana, marijuana extracts, and distillates, the application of mid-IR to the analysis of dried, ground hemp was investigated. Such analyses would allow hemp farmers to monitor their grows and to harvest at the right time thereby lessening the risk of losing their crop (18). The molecules in hemp of economic value include cannabidiolic acid (CBDA) and cannabidiol (CBD). An analyzer that quantitates these molecules would allow hemp farmers to conduct rational experiments varying soil, water, light, fertilizer, and other inputs to maximize the amounts of CBDA and CBD generated by a grow while minimizing the cost of production. Such an analyzer could also be used for on-the-spot testing to set the price when hemp biomass is bought and sold.

State governments are currently tasked with determining the legality of hemp crops (19). Many states collect samples from a grow and then transport them to a laboratory for analysis (19). This is expensive and time consuming. Additionally, it can take state laboratories weeks to complete their analyses. An accurate, portable field hemp analyzer would allow state governments to perform hemp compliance testing on the spot saving taxpayer dollars and giving more representative data.

There has been controversy around the legality of shipping hemp across state lines in the United States. For example, in early 2019 hemp shipments in two states were seized by state law enforcement agencies (20), who argued that hemp that contained any THC violated state law. While the courts sort this out, truck drivers sit in jail and valuable hemp is sequestered in government warehouses (20). If law enforcement agencies had a field portable hemp analyzer, they could test hemp on the spot for compliance, perhaps avoiding false arrests and needlessly seizing legal plant material.

Experimental

Hemp samples were prepared by taking approximately 5 g of dried plant material and grinding it for 1 min in a standard coffee grinder. These samples were scanned using a BSS 3000 Hemp Analyzer (Big Sur Scientific). The BSS 3000 is about 6-in. x 5-in x 4-in., weighs about 3 lbs, and is portable. The BSS 3000 is a general purpose quantitative mid-IR spectrometer, and was equipped with a triple bounce zinc selenide (ZnSe) attenuated total reflectance (ATR) crystal (8–10). Ground plant material was placed on the BSS 3000’s sampling window, as seen in Figure 1.

The powdered sample is then secured with a clamp, which can be seen in the background in Figure 1. The clamp features a mechanism that slips at a given torque to insure reproducible pressure is applied to all samples. Mid-IR spectra were measured from 1250 to 952 cm^-1 at 12 cm^-1 instrumental resolution. This spectral region was chosen because THC and other cannabinoids absorb strongly here, enhancing sensitivity.

Big Sur Scientific Cannabis Analyzer software was used to control and scan the analyzer. The software runs on a standard personal computer equipped with the Windows 10 Pro operating system. Data is transferred from the analyzer to the computer via a USB cable. The software allows the user to choose different sample types, such as dried ground hemp, extracts, or distillates. A background spectrum is run, and then the user is prompted to scan three separate aliquots of each sample. The calibration models are applied to each spectrum in turn, the cannabinoid weight percents are predicted, then these are averaged to ensure representative data, and the results are presented to the user on the computer screen. The user then has the option to print the results, copy and paste the results into a different application, or generate a certificate of analysis in Adobe Acrobat PDF format. A video on how the BSS 3000 is used to analyze hemp is available (21). Mid-IR spectra of 12 standard hemp samples, spanning a range of cannabinoid concentrations, were measured in triplicate for a total of 36 reference spectra. The same samples were analyzed via HPLC to yield cannabinoid wt.% values. HPLC analyses were performed by ProVerde Labs, a state licensed, ISO certified, third party laboratory.

To build mid-IR calibration models, HPLC cannabinoid weight percents and mid-IR spectra were input into the Big Sur Scientific Model Builder software. Spectra were preprocessed by calculating their first derivative using the Savitsky-Golay algorithm (13) using a second order polynomial and filter width of three points. The purpose of the derivative is to remove any baseline offsets that may be present in the spectral data.

Both THCA and THC exhibit a strong absorbance around 1160 cm^-1. The size of this absorbance is thus proportional to the total THC concentration in a hemp sample. This is illustrated in Figure 2.

First derivative spectra are shown as these are the spectra that were used in building the calibration model. The size of derivative features follow Beer’s Law and are proportional to concentration (13). The top trace in Figure 2 is from a hemp sample containing 0.77 wt.% total THC, the bottom trace is from a hemp sample that contains 0.27% total THC. The features are clearly of different sizes, and it is the variation in the size of these features that gives rise to the correlation between total THC values measured by HPLC and mid-IR as discussed below.

Pure cannabinoid standards were not used to calibrate the spectrometer because it would have been inappropriate (13). In use, the spectrometer analyzes dried, ground hemp not solutions of pure cannabinoids. For the spectroscopic models to be applicable to actual hemp samples, spectra and cannabinoid weight percents for dried, ground hemp samples must be used to build calibrations. A partial least squares algorithm (PLS1) was used to build the calibration models (13). The advantage of this algorithm compared to traditional single peak Beer’s Law analyses is that it works well even if spectral peaks contain overlaps from multiple analytes (13). Thus, there is no need for each analyte to have a spectrally resolved peak for quantitation to be achieved (13). Separate mid-IR calibration models for delta-9-tetrahydrocannabinol (Δ⁹-THC), tetrahydrocannabinolic acid (THCA), total THC, CBDA, CBD, total CBD, cannabigerolic acid (CBGA), and cannabichromene (CBC) were developed.

Models were validated using the leave-one-out cross validation method (13). In this technique, one standard sample’s spectra and reference concentrations are removed from the data set, and a calibration is generated using the remaining standard spectra and cannabinoid weight percents. This model is then applied to the spectra of the standard sample left out to give mid-IR predicted concentrations. This process is completed in turn for each standard sample until each one’s data has been left out, yielding a set of predicted concentrations as measured by mid-IR, which are compared to the HPLC reference cannabinoid values for the sample set to calculate accuracy. This is a validation because the sample’s data whose concentrations are being predicted are not included in the calibration model applied.

Accuracies were calculated as average cross validation standard errors of prediction (ACVSEP). The need for this metric arises because for each calibration sample there was one HPLC cannabinoid reference value and three mid-IR predicted values. This arose because spectra of each sample were measured in triplicate, and thus three mid-IR predicted values for each standard were generated.

The first step in determining the ACVSEP is to calculate the average predicted mid-IR value for a standard sample and given analyte using equation 1.

C_apc = [Σ_n(C_p)]/ⁿ                  [1]

where C_apc is the average predicted concentration by mid-IR for a particular sample and analyte; C_p is a predicted concentration for that sample and analyte; and n is the number of predicted concentrations for a standard. In the present case n = 3.

In the present case, there were 12 HPLC and 36 mid-IR concentration values for each analyte. Equation 1 is used to calculate the average predicted mid-IR value for each standard. We now have 12 HPLC and 12 average mid-IR predicted values for comparison.

The ACVSEP for a particular analyte is calculated using this data set and equation 2.

ACVSEP = [Σ_N(C_ref – C_apc)²/N-1]^1/2     [2]

where C_ref is the reference sample weight percent value by HPLC; C_apc is the average predicted value as calculated using equation 1; and N is the number of reference samples.

The ACVSEP is the standard deviation of the HPLC reference values and the C_apc values as determined by mid-IR. The ACVSEP is an excellent measure of accuracy since it is determined by how well the model predicts concentrations for samples not included in the calibration (13), which is precisely how calibrations are used in real life. Correlation coefficients were determined for the mid-IR calibration models by plotting predicted wt.%  values for the standards as determined by mid-IR versus the values for the same samples as measured by HPLC.

Results and Discussion

The accuracies, range, and correlation coefficients (R²) for the determination of cannabinoids in dried, ground hemp by mid-IR are seen in Table II.

The total THC calibration has an accuracy of ±0.04 wt.%, more than accurate enough to determine whether a hemp crop is above or below the 0.3 wt.% limit as stated in federal law (1,2). The accuracies for the other cannabinoids are surprisingly good for a 2-min, no sample preparation required analysis of a naturally variable material. Table II indicates mid-IR can accurately quantitate total THC, THC, THCA, total CBD, CBD, CBDA, CBGA, and CBC in dried, ground hemp plant material. Given the half-dozen cannabinoids quantitated, this method gives not only potency measurements but also a cannabinoid profile.

To establish the correlation between two analytical methods, a plot of the results on the same samples obtained using both techniques can be made and is called a correlation chart (13). In the present case, correlation charts are constructed by plotting the reference cannabinoid wt.% value for standard samples as measured by HPLC versus the C_apc value for the same samples as measured by mid-IR. The correlation chart for total THC is seen in Figure 3.

A measure of the quality of the correlation between two methods can be derived from this chart by calculating the “correlation coefficient” or R², value (13). An R2 of 1.0 means there is perfect agreement between the two methods, which doesn’t happen because of measurement error (13). A correlation coefficient of 0 means there is no correlation between the two methods. The correlation coefficients for all the cannabinoids determined in hemp by mid-IR are seen in Table II. The correlation coefficient for total THC in hemp when comparing HPLC and mid-IR values is 0.94. Most of the R² values in Table II are about 0.94 or better. This is very good given hemp’s natural variability and complexity (3). Table II also lists the concentration range of the standards used for each calibration, which is the range over which each model can be expected to be quantitatively accurate.

The mid-IR spectrometer used in this study requires little sample preparation for hemp apart from drying and grinding. There is no weighing, dissolving, or filtering to prepare samples as in chromatography (3–5). Mid-IR also requires no consumables, so the cost per analysis is $0, unlike chromatography where the consumables and solvent used cost several dollars per sample, not to mention the at least 10 min of labor required to run one sample (3–5). The mid-IR analyzer used in this study comes pre-calibrated and features push button operation so anyone can use it. Lastly, analyzing a sample takes about 2 min, significantly faster than chromatography (3–5).

The ability of mid-IR spectroscopy to determine cannabinoids in dried, ground hemp opens up some unique applications. Hemp farmers can take numerous cuttings of a plot, dry them, grind them, and then easily analyze them by mid-IR spectroscopy. By measuring more samples than is typical with chromatography, a representative picture of a grow is obtained, which is more valuable to the farmer. In essence, HPLC provides a snapshot of a grow, mid-IR provides a motion picture. Since the latter provides more information, it should be preferred. The ability of mid-IR spectroscopy to measure CBDA, CBD, and total CBD-molecules of value in hemp-means hemp farmers can conduct rational growing experiments to maximize the CBD and minimize the THC content of their crops.

State regulators currently only sample a fraction of the plants from a particular grow for testing (19). This is because samples are collected and taken to an offsite location for slow and expensive chromatographic testing. Whether the handful of samples measured is representative of a grow needs to be questioned (6,7). Instead, a portable mid-IR unit would allow state regulators to analyze many samples on the spot, giving a more representative picture of a grow, saving time and money.

Hemp extractors can also make use of a portable, accurate, mid-IR hemp analyzer to assess the potency of hemp biomass prior to purchase. For example, when purchasing plant material a hemp extractor can use mid-IR to analyze the material on the spot to ascertain what they are truly buying and set the proper purchase price. The spectrometer described above can also analyze cannabinoids in extracts and distillates (8,15,16).

A portable hemp analyzer could also be of utility to law enforcement. There has been great controversy since the passage of the Farm Bill about the legality of transporting hemp across state lines (18,20). A portable, easy-to-use, and accurate mid-IR spectrometer could be used by law enforcement to determine the THC level of hemp shipments roadside. Shipments that are legal could be sent on their way, reducing false arrests, lessening the sample load for forensic laboratories, and thus saving time and taxpayer money. Samples of shipments that test above the legal limit could be forwarded for confirmatory testing. Of course, analytical instrumentation produces data, is it up to district attorneys and law enforcement officials to determine who to arrest, how to prosecute, and what evidence to introduce in court. A quantitative, field portable hemp analyzer will give them more evidence to work with when making these decisions.

Conclusions

It has been shown that mid-IR spectroscopy can be used to accurately measure Δ⁹-THC, THCA, total THC, total CBD, CBDA, CBD, CBGA, and CBC in dried, ground hemp samples with good correlation to HPLC values measured by a state licensed, ISO certified laboratory. Unlike chromatographic methods, mid-IR analyses are fast, easy, inexpensive, and there is little sample preparation. This allows multiple samples to be scanned and averaged, giving hemp farmers a more accurate and representative picture of a grow, proving of greater value to hemp farmers. State regulators could use mid-IR for field compliance testing, hemp extractors are using mid-IR to assess biomass potency at the point of sale, and law enforcement could use such a device for roadside testing of questioned material.

Acknowledgement

I would like to thank Onchiotta Adornetto and the folks at Hudson Hemp for providing hemp samples.

References:

1. 115th United States Congress, Senate Bill S.2667, ”Hemp Farming Act of 2018.”
2. https://www.federalregister.gov/documents/2019/10/31/2019-23749/establishment-of-a-domestic-hemp-production-program.
3. M.W. Giese, M.A. Lewis, L. Giese, and K.M. Smith, J. AOAC Int. 98(6), 1503 (2015).
4. C. Giroud, CHIMIA Intl. Journal of Chemistry 56, 80 (2002).
5. T. Ruppel and M. Kuffel, "Cannabis Analysis: Potency Testing Identification and Quantification of THC and CBD by GC/FID and GC/MS," PerkinElmer Application Note (2013).
6. B.C. Smith, Cannabis Science and Technology 2(1), 14–19 (2019).
7. P. Atikins, Cannabis Science and Technology 2(2), 26–34 (2019).
8. B.C. Smith, P. Lessard, and R. Pearson, Cannabis Science and Technology 2(1), 48–53 (2019).
9. B.C. Smith, Cannabis Science and Technology 2(2), 12–17 (2019).
10. B.C. Smith, Cannabis Science and Technology 2(3), 10–14 (2019).
11. B.C. Smith, Fundamentals of Fourier Transform Infrared Spectroscopy, 2nd Edition (CRC Press, Boca Raton, Florida, 2011).
12. B.C. Smith, Infrared Spectral Interpretation: A Systematic Approach (CRC Press, Boca Raton, Florida, 1999).
13. B.C. Smith, Quantitative Spectroscopy: Theory and Practice (Elsevier, Boston, Massachusetts, 2002).
14. B.C. Smith, M. Lewis, and J. Mendez, "Optimization of Cannabis Grows Using Fourier Transform Mid-Infrared Spectroscopy," PerkinElmer Application Note (2016). https://www.perkinelmer.com/lab-solutions/resources/docs/APP_Determination_of_THC_and_CBD_CannabisFlower.pdf.
15. B.C. Smith, Terpenes and Testing Nov.-Dec., 48 (2017).
16. B.C. Smith, Terpenes and Testing Jan.-Feb., 32 (2018).
17. B.C. Smith, Manuscript in preparation.
18. https://www.mprnews.org/story/2019/06/06/hemp-production-minnesota-farmer-sues.
19. https://docs.google.com/spreadsheets/d/1x8doatlR6w1W3W6hA0hlu67qwe9uOSfYoHe3vjcYs6Y/edit#gid=855723386, courtesy Toni Anthony, Key to Life Inc.
20. https://cannabisindustryjournal.com/tag/big-sky-scientific/.
21. https://www.youtube.com/watch?v=DhQlp-cwQh0&t=3s.

About the Author

Brian C. Smith, PhD, is Founder, CEO, and Chief Technical Officer of Big Sur Scientific in Capitola, California. Direct correspondence to: brian@bigsurscientific.com

How to Cite this Article

B.C. Smith, Cannabis Science and Technology 2(6), 28-33 (2019).