Initial NMR analysis of the major peaks in the collected spectra show that the major product is consistent with published chemical shifts of Δ-8 THC [4] (Figure 2). With the integral for the proton signal on carbon 5′ calibrated to 1.0 proton, few of the other integrals match up to the expected values. As a result, the overall proton count by integration is 31 for Figure 2. A proton integration of 27 is expected for Δ-8 THC. While the deviations are slight, typically less than 10% deviation is expected, examination of the minor peaks show the presence of many impurities.

Upon closer inspection of the region surrounding the peak from the proton on carbon 5′ in sample 3 it becomes very apparent that there are multiple products in this sample (Figure 3). This sample was certified as 93.43% pure Δ-8 THC with no other cannabinoids detected.

The presence of four extraneous doublet peaks, all with J values of 1.6–1.7 Hz suggests that these peaks may well be equivalent protons on isomers of Δ-8 THC or some other form of cannabinoid. All other samples show similar impurities with the expanded spectral excerpts available in the Supplementary Materials (Sample S3). Peaks “A” and “D” in Figure 3 may potentially belong to compounds 1 and 2 described by Radwan [5]. Though the relatively congested peak area for carbon 3′ shows no clear peak of similar integral value at 6.27 ppm and 6.24 ppm, respectively (Figure 4).

When the integral of the major product peak “E” (Figure 3) is calibrated to 100 protons it becomes quite clear that over 15% of this product is not consistent with Δ-8 THC, even when ignoring the extraneous peaks in the lower frequency portions of the spectrum. Table 1 details the relative impurity totals for all samples examined.

With the Δ-8 THC 5′ peak (peak E) integral value adjusted to 100, a simple percentage value of structurally similar peaks in the same region of the spectrum can be calculated. Due to having similar, but slightly different, resonant frequencies peaks A, B, C, D, and F can be presumed to be compounds with a substituted arene moiety similar to that of THC. However, as these peaks have differing chemical shifts than the majority Δ-8 THC product, they are clearly not the same compound. When the peak areas are added together, it is clear that the COA purity values supplied with these products are not consistent with the measured values for each sample.

Due to the general unavailability of 800 MHz class NMR instruments, sample 3 was also analyzed on a 400 MHz NMR for comparison. While integration of the discrete peaks in a 1d proton spectrum is more challenging due to the lower resolution it appears to be a viable option for analysis of these products (Figure 5).

Samples 6 through 10 (Table 1) contain added terpenes with resonances in the region of interest. Therefore, care must be taken when integrating impurities in this region to avoid including terpene signals as impurities. Samples 8, 9, and 10 were supplied with identical COA’s suggesting that they were from the same lot of base product but considering the difference in impurities between sample 8 and the other two this appears to not be the case. Overall, none of the samples analyzed, save for sample 4, are close to having accurate COA purity values even when only investigating a single peak and ignoring the rest of the spectrum. Due to the chemical shift and J-value similarities of this peak and that found on Δ-8 THC, the compounds could reasonably be assumed to contain an arene and thus these products would very likely be UV active and detectable on an HPLC equipped with a UV detector or DAD.

Comparing the HPLC chromatogram supplied with sample 4 (Figure 6) to an HPLC-UV performed during this study (Figure 7), it is clear that the HPLC elution method used by the certifying laboratory is inappropriate for detecting impurities in this type of product.

Mass spectrometry and MS² analysis of peaks in sample 4 revealed several impurity peaks. Figure 8 is the total ion chromatogram for sample 4 over the mass range of 317–750 Da. Table 2 is a summary of the LC/MS experiment. The peak number in the first table column correspond to the numbers shown in Figure 8. The second column indicates the retention time of the peak. The “major peak” column indicates the mass of the base peak observed for each chromatographic peak. The last column shows the masses of other peaks observed in the mass spectrometry scan.

Table 2 shows the results of several MS/MS experiments. Each of the labeled peaks in Figure 8 were subjected to MS² analysis. The parent ion column in Table 3 is the same as the major peak listed in Table 2. The parent ion was isolated and then fragmented in the MS/MS experiment. The “Base Peak” in Table 3 indicates the mass of the largest ion intensity in the MS/MS spectrum. The last column identifies other fragment ions in the MS/MS spectra.

Analysis of the MS² spectra yielded interesting results showing a variety of what appear to be both impurities from low quality CBD feedstock and known side reaction products from the cyclization reaction used to convert CBD into Δ-8 THC.

Peak 1 exhibits a molecular mass of 328 Da, and as such, this impurity is certainly not CBD or Δ-8 THC. The MS² spectra compares well with published data describing cannabidihexol (CBDH) or tetrahydrocannabihexol. [5] (See supplemental spectrum Peak 1).

Peak 2 also exhibits a molecular mass of 328 Da and possesses a similar MS² spectrum to peak 1. Due to the longer retention time would suggest that this impurity may be an isomer of the impurity in peak 1.

Peak 3 with an odd mass of 253 Da appears to contain an amine. The use of amine containing reagents do not appear in any published literature describing the conversion of CBD to Δ-8 THC and as such the origin of this impurity is unknown. A mass transition of [M-85]⁺ may indicate the presence of a piperidine. However, further work would be required for definitive identification.

Peak 4 exhibits [M-18]⁺, [M-42]⁺, and [M-56]⁺ mass transitions suggesting that it is also a cannabinoid analogue. However, the mass of 344 Da suggests that this compound may be an O-alkylated cannabinoid analogue, which is supported by the [M-46]⁺ mass transition showing the loss of an ethyl ester.

MS² spectra for peak 5 appears to show transitions expected from 5′′-hydroxy-CBD or 5′′-hydroxy-THC. With the characteristic loss of water at [M-18]⁺, an [M-60]⁺ ion suggesting the loss of C₃H₇OH, an [M-72]⁺ transition for loss of C₄H₈OH, and an [M-74]⁺ transition for a loss of C₄H₁₀OH. Remaining [M + H]⁺ ions of 193 Da and 107 Da are suggestive that this is likely a 5′′-hydroxy CBD or THC analogue. This impurity likely arises from side reactions in the CBD to Δ-8 THC conversion reaction [6].

Peak 6 appears consistent with published MS2 spectra of cannabidivarin (CBDV) [7]. Due to being a metabolite found in cannabis, this is likely an impurity carried over from low purity CBD feedstock.

Peak 7 is quite unusual compared to the other impurities investigated in this study. In MS² spectra the base peak has a mass transition of [M-180]⁺ from the molecular ion suggesting the presence of a hexose moiety in this contaminant. The source for this impurity is unknown and further investigation would be required for full structural elucidation.

Peak 8 with an [M + H]⁺ molecular ion of 315 Da is consistent with MS² spectra of THC or CBD. However, due to the longer retention time of 9 min compared to a retention time of 6.46 min for CBD and the slightly longer retention time for Δ-8 THC of 10.5 min, it would appear to be a CBD or THC isomer.

Peak 9 exhibits MS² spectra similar to peak 8 with a slightly longer retention time of 9.5 min. This is also likely another CBD or THC isomer.

Peak 10 MS² transitions are similar to published spectra of cannabidihexol [6]. However, the [M-42]⁺ base peak, is significantly more intense in spectra collected in this study. This impurity is tentatively identified as cannabidihexol or an isomer thereof.

With several of these impurities now having tentative identifications it appears that the problem with these products is threefold: impure CBD feedstock, poor post-reaction purification, and poor analytical practices during certification of purity. With NMR, HPLC-UV, and HPLC-MS data, it is clear that the current analysis methods need to be improved. Inadequate purification and testing protocols gives rise to a situation where consumers make use of products with far higher levels of impurities than they were led to believe, a situation that could potentially give rise to catastrophic consequences. A less extreme and possibly even more concerning problem arises from considering that failures such as this could lead to a loss of public confidence in laboratory testing of consumer goods altogether.

Even a simple HPLC-UV analysis of these products shows that the certifying labs are failing their customers and consumers by using inappropriate HPLC conditions. While NMR and HPLC-MS could be an arguably more precise methods for detecting and analyzing these impurities the low cost of HPLC is undeniably appealing. HPLC is certainly still a viable method for analysis of these products but certifying laboratories must be vigilant regarding the effectiveness of the methods used in these analyses as demonstrated by the clear failings described above.