How to Distinguish Hemp from Marijuana with Mid-Infrared Spectroscopy

Published on: 
Cannabis Science and Technology, March 2021, Volume 4, Issue 2
Pages: 39-40

Columns | <b>Cannabis Voices</b>

In this interview, Brian Smith, PhD, founder, CEO, and CTO at Big Sur Scientific, discusses his work with mid-IR spectroscopy to test for total THC & more.

Mid-infrared (IR) spectroscopy has been around for a long time, but spectroscopists continue to apply this technology to new markets, such as hemp testing. We recently spoke to Brian C. Smith, PhD, founder, CEO, and Chief Technical Officer at Big Sur Scientific, about his work with mid-IR spectroscopy to test for total tetrahydrocannabinol (THC) and other cannabinoid levels in hemp and marijuana plant material, his plans for future developments, and more.

Can you tell us in detail how mid-infrared spectroscopy distinguishes the chemical differences between hemp and cannabis?

Brian Smith: Mid-infrared (IR) spectroscopy used infrared light, or heat, to examine samples. The wavelengths absorbed are unique to each individual molecule, which is how we determine what cannabinoids are present. The amount of light absorbed tells us the concentration of each cannabinoid. Current Federal law states that legal hemp contains 0.3% or less total tetrahydrocannabinol (THC), while illegal marijuana contains more than 0.3% total THC. This means that to legally distinguish hemp from marijuana, a total THC value must be measured. The hierarchical total THC algorithm we use, discussed in our peer-reviewed paper recently published in Cannabis Science and Technology(1), classifies cannabis plant material as hemp or marijuana with a high success rate.

Can you tell us how you were able to achieve 99+% discrimination (qualitative) accuracy for this method?

Smith: The key to our success is use of mid-IR spectroscopy combined with a Hierarchical Total THC Classification algorithm. The challenge with classifying cannabis plant material as hemp or marijuana is the broad total THC concentration range present in these materials, anywhere from 0.1–30%. Our method first uses a global total THC calibration to classify cannabis plant material samples as high or low total THC. High total THC samples are classified as marijuana and the analysis ends. Low total THC plant material may be legal hemp. In this case, a special high accuracy low total THC calibration is applied. Samples measuring 0.3% total THC or below are classified as hemp and samples containing more than 0.3% total THC are classified as marijuana.

What algorithm do your analyzers use to determine potency in samples?

Smith: The Global Total THC model and the low Total THC model were created using the partial least squares 1 (PLS-1) algorithm. These were incorporated into the Hierarchical Total THC algorithm. The first decision point in the Hierarchical algorithm determines whether a cannabis plant sample contains 5% or more total THC. Samples below this value are passed on to the next decision point, where it is determined whether the sample is above or below the 0.3% total THC legal limit.

How is accuracy measured on these machines to ensure the data is correct?


There are two figures of merit required for determining the accuracy of hemp and marijuana classification. The accuracy of the marijuana/hemp classification was calculated as a percent success rate. The algorithm was then applied to a set of validation samples known to be hemp or marijuana and the number correctly classified was divided by the total number of samples to determine the percent successfully classified.


The accuracy of the spectroscopic total THC calibration models used was determined as a standard error of prediction (SEP). In this calculation, a model is applied to a set of validation samples that were not included in the calibration but whose total THC values are known via high performance liquid chromatography (HPLC). This gives a set of total THC values as predicted by mid-infrared spectroscopy, and set of known values. The standard deviation between the known and predicted total THC values gives the SEP.

How does this method compare to other testing methods currently in use?

Smith: As pointed out in our article (1), for an analytical technique to be used to distinguish between hemp and marijuana per Federal law, it must produce a weight percent total THC value. Methods that have been touted in this application, such as color tests, genetic testing, and Raman spectroscopy, do not produce a total THC value and hence, should not be used in this application.

Gas and liquid chromatography can be used to accurately measure total THC levels in cannabis plant material. However, these techniques involve significant sample preparation, expensive solvents and consumables, and require a degree and significant training to be used properly, take 20+ minutes to produce a result, and are not portable. Therefore, chromatography is not practical for field analysis of cannabis plant material.

Our mid-IR spectroscopy analyzer is portable, pre-calibrated, features push button operation so anyone can use it, requires little sample preparation, and also gives accurate results in two minutes. It is ideally suited for use by law enforcement for on-the-spot classification of cannabis plant material, and by forensic labs for screening samples

Will mid-IR spectrometers ease the confusion for law enforcement so that fewer legal hemp shipments are confiscated and fewer individuals are arrested?

Smith: Yes, that is why we developed our analyzer. Mid-IR spectroscopy can be used roadside by police officers to determine whether suspect cannabis plant material is legal hemp or illegal marijuana. If a sample tests as legal, no action is taken. If a sample tests as illegal, it should be forwarded to a forensic laboratory for confirmatory analysis.

How will mid-IR spectroscopy benefit state departments of agriculture and forensic laboratories that are tasked with determining the legality of cannabis plant material?

Smith: Given the speed and ease of use of mid-IR, state department of agriculture agents can analyze plant material samples themselves in the field to determine legality. Samples that test illegal should be forwarded to a certified laboratory for confirmatory analysis. Field testing of samples by mid-IR will reduce the load of legal samples being sent to forensic laboratories, saving time and money.

Similarly, forensic labs can use mid-IR to screen incoming plant material samples and weed out the legal ones, saving the time and expense of analyzing samples that didn’t need to be analyzed.

How do you anticipate this instrument will help hemp farmers during the growing and harvesting seasons?

Smith: Hemp farmers have to tread a fine line between minimizing total THC levels to stay legal and maximizing total cannabidiol (CBD) levels to stay profitable. Our analyzer accurately measures total THC and total CBD in dried ground hemp (see our paper in the December 2019 edition of Cannabis Science and Technology [2]). Our analyzer allows hemp farmers to optimize their grow and know when to harvest to stay both legal and profitable.

What type of representative sampling methods are required for accurate analysis of cannabis grows?

Smith: Cannabis is a naturally variable, inhomogeneous material. Research has shown that adjacent plants, and cuttings from the same plant, vary significantly in cannabinoid concentration. This means taking one or a few cuttings from a grow and analyzing them will not give representative results, known as sampling error. This error is dangerous because you can be misled into thinking your grow is legal when it isn’t. The way to minimize sampling error is to test as many samples as is practical from a cannabis grow. In a recent column in Cannabis Science and Technology (3), I discussed my representative sampling plant for cannabis grows. To achieve the best compromise between sampling error and sample load, I advocate that we as an industry commit to analyzing 1 plant per 1000. Mid-IR can analyze a cannabis plant sample in two minutes and is able to handle hundreds of analyses per day. This technology is up to the task of handling the sample load required for true representative sampling.

What plans do you have on expanding your hemp and cannabis products?

Smith: The Big Sur Scientific family of cannabis analyzers accurately quantitates cannabinoids and terpenes in hemp, marijuana, extracts, distillates, vape pen formulations, tinctures, and final products. We are working on extending the analyzer to cannabis isolates and edibles. Lastly, we are also researching the use of our quantitative, general purpose mid-IR spectrometer to analyze hops, milk, olive oil, and ground meat.


  1. B.C. Smith and C.A. Fucetola, Cannabis Science and Technology 3(6), 24-38 (2020).
  2. B.C. Smith, Cannabis Science and Technology 2(6), 28-383(2019).
  3. B.C. Smith, Cannabis Science and Technology 3(6), 10-13 (2020).

How to Cite this Article

M. Colli, Cannabis Science and Technology 4(2), 39-40 (2021).