Investigation of dimethyl sulfide formation during GC analysis of permethrin API: study of the reaction kinetics and estimation of the activation energy of the reaction

Abstract

An unknown peak in the chromatogram was observed during gas chromatography (GC) analysis of the active pharmaceutical ingredient (API) permethrin (PMN). Further investigations revealed that the identity of the unknown peak is dimethyl sulfide (DMS) which was formed via a reaction between dimethyl sulfoxide (DMSO) and one of the impurities of PMN named 3-phenoxylbenzyl chloride. Kinetics of the reaction was studied and the activation energy of the reaction was determined through Arrhenius analysis.

---

Introduction

Permethrin is used as an API in human and veterinary drugs to treat scabies for human and also to control fleas and ticks for dogs.^1–5 As per regulatory requirements and International Conference on Harmonisation (ICH) guidelines,⁶ residual solvents in API batches should be controlled to ensure safety and quality of the finished product. Previously, we have reported the development and validation of a headspace GC method for determination of residual solvents in PMN.⁷ During the headspace GC method development, it was noticed that an unknown peak was formed when DMSO was used as a solvent to dissolve PMN. In this paper, we are reporting our study results for the root cause and formation mechanism of this peak; the kinetics of the reaction has also been studied.

Results and discussion

As shown in Fig. 1, an unknown peak was generated when PMN was heated in DMSO. This unknown peak was not observed when solvent was changed from DMSO to N,N-dimethylformamide (DMF). Neither did the peak appear when only neat PMN was heated and analyzed by GC. Based on GC-MS analysis and further confirmation by spiking authentic reference material into PMN sample, the peak was confirmed to be dimethyl sulfide (DMS). When PMN in DMSO was heated from low to high temperatures, the DMS peak also increased with the increase of temperature. Since the typical manufacturing process for DMSO is through oxidation of DMS, DMS is a common impurity in commercial DMSO solvents.⁸ As shown in Fig. 1, both DMSO and PMN in DMSO were heated at 130 °C for 30 minutes and then analyzed by GC. The DMS peak was generated when PMN was heated in DMSO and no increase of DMS peak in the DMSO blank. Therefore, the DMS peak was not due to the existing DMS in DMSO and must be from a different pathway. And all evidences indicated that a chemical reaction occurred when PMN sample was heated with DMSO.

Fig. 1 Overlaid GC chromatograms from bottom to top: DMSO, DMF, PMN in DMF, PMN in DMSO and neat PMN API (all samples were heated at 130 °C for 30 minutes in headpace GC vials prior to GC analysis). RS1: 2-methylpentane, RS2: 3-methylpentane, RS3: n-hexane, RS4: methylcyclopentane, RS5: cyclohexane.

DMSO, due to its excellent solubility for polar and nonpolar compounds, has been widely used as the most preferred solvent to prepare samples for analysis of residual solvents using GC.⁸ DMSO has a high boiling point of 189 °C and is also thermally stable⁹ at relatively high temperature and is one of the most popularly used solvents for gas chromatography (GC) for residual solvent testing in the pharmaceutical industry.^10,11 DMSO is also the solvent of choice for residual solvent testing prescribed in the compendia methods of the United States Pharmacopeia and the European Pharmacopoeia.^12,13

DMSO has also been used as a weak oxidizer in organic synthesis for oxidation of primary and secondary alcohols.^14–18 In oxidation reactions, DMSO is usually activated to form a sulfonium ylide intermediate^19,20 with the target molecules which further reacts with a base or an acid to form the oxidation product. In Pfitzner–Moffatt oxidation, DMSO is activated with dicyclohexyl carbodiimide (DCC) and then catalyzed by phosphoric acid to form the carbonyl oxidation product of the target alcohol. In the Swern oxidation, DMSO reacts with oxalyl chloride first and then forms sulfonium ylide intermediate which further reacts with triethylamine to form the alcohol oxidation products. DMSO has also been reported to be able to react with organic halide compounds and tosylates.^21–23

In the resonance structure of DMSO, the oxygen atom is slightly electron rich and therefore nucleophilic. The oxygen atom in DMSO could potentially attack the nuclear center of another molecule and therefore trigger an oxidative process. As shown in Fig. 2 of the structures of PMN and known impurities that are present in typical batches of PMN (Ia and Ib), 3-phenoxybenzyl alcohol (III), 3-phenoxylbenzaldehyde (IV) and 3-phenoxylbenzyl chloride (V) could potentially be oxidized by DMSO to form 3-phenoxylbenzaldehyde or 3-phenoxylbenzoic acid. All these three impurities are present at low levels in typical batches of PMN.

Fig. 2 Structures of permethrin and major known impurities.

For identification of the potential reactant with DMSO, PMN API was heated in DMSO for different times at 130 °C and then analyzed by high performance liquid chromatography (HPLC). From the overlaid HPLC chromatograms as shown in Fig. 3, it clearly demonstrates that the magnitude of peak V decreased with time upon heating and at the same time the magnitude of peak IV increased.

Fig. 3 Overlaid HPLC chromatograms of permethrin in DMSO after heating at 130 °C for 0, 10, 30, 60, 120 and 240 minutes (HPLC detection wavelength: UV at 215 nm).

Peak area percentages at each time point for related compounds in PMN API are listed in Table 1. As shown in the table, after PMN API was heated in DMSO for 240 minutes, peak area percentage of peak V decreased from 0.14% to 0.03% and the peak area of peak IV increased from 0.11% to 0.22%. Peak area percentages of other impurities remained unchanged. The change of peak area percentage for peak V and peak IV during heating strongly indicated that impurity V was the most probable reactant with DMSO.

Table 1 Peak area percentage by HPLC after PMN heated in DMSO at 130 °C

Time (min)	% peak area of each compounds by HPLC
	Ia	Ib	IIa	IIb	III	IV	V	VI	Other
Initial	58.41	40.49	0.11	0.03	0.24	0.11	0.14	0.04	0.43
10	58.41	40.48	0.10	0.03	0.24	0.13	0.14	0.04	0.43
30	58.42	40.47	0.10	0.04	0.24	0.16	0.11	0.04	0.42
60	58.41	40.47	0.11	0.03	0.25	0.18	0.08	0.03	0.44
120	58.41	40.49	0.10	0.03	0.25	0.20	0.05	0.04	0.43
240	58.40	40.50	0.10	0.03	0.26	0.22	0.03	0.04	0.42

---

To confirm compound V was the real reactant with DMSO to form DMS, authentic reference material of V was spiked into DMSO to reach a concentration of 16 μg mL⁻¹ (equivalent to about 7 × 10⁻⁵ mol L⁻¹). The solution was heated at 130 °C for 15, 30, 60, 120 and 180 minutes and analyzed by headspace GC and HPLC. For each time point determination, an aliquot (5.0 mL) of the solution was transferred to a headspace GC vial, tightly capped and heated at 130 °C in the headspace oven of GC. When the equilibration time point was reached, 1.0 mL of the vapor was injected into the GC for detection of DMS. DMS formed in the solution and was identified and quantitated by the injections of DMS external reference standard solutions (see ESI† for detailed experimental information). After the GC analysis was completed, the vial was cooled down to room temperature and the remaining solution was diluted with acetonitrile/water and analyzed by HPLC for the detection of compounds IV and V. Compounds IV and V in the solution and were identified and quantitated by injecting standard solutions containing known concentrations of IV and V. From the results of these experiments, both DMS and IV were generated and no other product was detected after V was heated in DMSO. As shown in Fig. 4, the concentration of compound V started to decrease after heating and both IV and DMS were generated when V started to decrease. The magnitude of V concentration consumed was comparable to the concentrations increase of DMS and IV.

Fig. 4 Concentrations of 3-phenoxybenzaldehyde (IV), DMS and 3-phenoxybenzyl chloride (V) after V heated in DMSO at 130 °C for different times.

The results as shown in Fig. 4 confirmed that compound V was the reactant in PMN API that reacted with DMSO to form DMS. The proposed reaction mechanism is shown in Fig. 5. The oxygen in DMSO first attacks the benzyl carbon and chloride left to form a sulfonium ylide intermediate. This intermediate further rearranged through elimination to form three species namely DMS, compound IV and hydrogen chloride. With positive identification of the products of this reaction via HPLC and headspace GC experiments (shown in Fig. 4), the proposed reaction mechanism further confirms the proposition described by Nace and Monagle for DMSO oxidation of organic halides.²²

Fig. 5 Proposed reaction mechanism of DMS formation.

Organic halides oxidations by DMSO are usually conducted by converting halides into tosylates or catalyzed by a base.^21,22 Results from our study strongly demonstrated that the DMSO oxidation of certain chemical entities (compound V in our case) could proceed even without the addition of a base.

Further studies were conducted to determine the kinetics of the reaction. Compound V was dissolved in DMSO at about 16 μg mL⁻¹ (about 7 × 10⁻⁵ mol L⁻¹) and heated for 15, 30, 60, 120 and 180 minutes in separate headspace GC vials. At each time point, the reaction mixture was cooled to room temperature, diluted and analyzed by HPLC. Duplicate determinations at each time point were conducted to obtain higher confidence on the data. Concentrations of compound V from these experiments are shown in Table 2.

Table 2 Concentrations of 3-phenoxybenzyl chloride (V) in DMSO after heating at different temperatures

Temperature (K)	Concentration of V (10^−5 mol L^−1) at each time point
	15 min	30 min	60 min	120 min	180 min
343	6.262	6.129	5.924	5.666	5.376
	6.263	6.125	5.938	5.673	5.388
353	6.115	5.925	5.609	5.023	4.863
	6.116	5.910	5.593	5.029	4.852
363	5.925	5.645	5.228	4.724	4.426
	5.921	5.622	5.210	4.719	4.438
373	5.550	5.113	4.892	4.267	3.906
	5.504	5.189	4.892	4.273	3.867
383	5.206	4.683	4.292	3.537	3.245
	5.198	4.676	4.313	3.544	3.230
393	4.917	4.332	3.674	2.988	2.601
	4.925	4.336	3.668	2.968	2.595
403	4.384	3.754	2.885	1.982	1.477
	4.482	3.748	2.890	1.979	1.478

---

Linear fit was obtained for the semi-logarithmic plot of concentrations of compound V vs. time via linear regression analysis. The results of the data analysis indicated a first order reaction kinetic for the oxidation reaction of compound V in DMSO.

The first order kinetic of this reaction obtained from experimental data is also consistent with theoretical expectation of the order of this reaction. During the reaction, large excess of DMSO was present in the reaction media and the concentration of DMSO remained relatively constant compared to the concentration of compound V. Therefore, the reaction rate was only dependent on the concentration of compound V. Further kinetic studies at different temperatures were also conducted and plots are shown in Fig. 6 which demonstrated good linear correlations with correlation coefficients (R) ranging from 0.979 to 0.996 (Table 3) at all seven temperatures.

Fig. 6 Semi-logarithmic plot of 3-phenoxybenzyl chloride (V) concentrations vs. time.

Table 3 Kinetics of 3-phenoxybenzyl chloride (V) oxidation by DMSO

Temperature (K)	k (s^−1)	Correlation coefficient (R)	Half-life (t^1/2) (hour)
343	1.480 ± 0.0460 × 10^−5	0.996	13.0
353	2.397 ± 0.178 × 10^−5	0.979	8.0
363	2.893 ± 0.193 × 10^−5	0.982	6.7
373	3.452 ± 0.260 × 10^−5	0.982	5.6
383	4.680 ± 0.331 × 10^−5	0.981	4.1
393	6.250 ± 0.466 × 10^−5	0.979	3.1
403	1.092 ± 0.051 × 10^−4	0.991	1.8

---

Based on the linear regression analysis of semi-logarithmic of compound V concentration vs. time, rate constant (k) at each temperature which are the slopes of each linear plot were calculated. Rate constants were obtained at confidence interval of 0.95. Half-lives of compound V in DMSO at different temperatures were also calculated by using the formula t^1/2 = ln(2)/k. The results of all the calculations are summarized in Table 3. As shown in Table 3, the k of compound V consumption increased from 1.480 × 10⁻⁵ s⁻¹ to 1.092 × 10⁻⁴ s⁻¹ when the temperature was increased from 343 K (70 °C) to 403 K (130 °C), and the half-life of the reaction significantly decreased from about 13 hours to less than 2 hours.

Arrhenius analysis was performed using the reaction kinetics data obtained from various temperatures. As shown in Fig. 7 of the Arrhenius plot, the slope obtained from the plot was −4127.5 K which was – (E_a/R). Since R equals to 8.314 J mol⁻¹ K⁻¹, therefore the activation energy of the reaction was calculated to be 34.3 kJ mol⁻¹ which was 8.2 kcal mol⁻¹. This activation energy is relatively small since the typical activation energy range are 10–30 kcal mol⁻¹ for common reactions.²⁴ The equation obtained from Arrhenius analysis is ln k = ln A + (−E_a/RT). Therefore, the rate of oxidation of compound V in DMSO at room temperature (298 K) could be predicted from the equation. The calculated value of k at 298 K was found to be 2.418 × 10⁻⁶ s⁻¹, with a half-life of about 80 hours.

Fig. 7 Arrhenius plot of 3-phenoxybenzyl chloride (V) in DMSO.

Conclusions

In conclusion, the DMS formation after PMN API was dissolved in DMSO was due to the reaction between DMSO and 3-phenoxybenzyl chloride (compound V). This reaction could proceed even without the addition of a catalyst. A thorough literature search reveals no report of the reaction kinetics study of the 3-phenoxybenzyl chloride oxidation in DMSO. Kinetics of the reaction was studied for the first time in this report and the activation energy of the oxidation reaction was estimated through Arrhenius equation and found to be about 8.2 kcal mol⁻¹. The small activation energy of this reaction may be due to the good leaving group of chloride in benzyl group, and also the phenoxy group on benzene facilitated this reaction via resonance stabilization of the sulfonium ylide intermediate. Our study results also suggested that careful evaluation needs to be conducted before selecting DMSO as a solvent for GC analysis because oxidation reactions can occur with DMSO depending on the physicochemical properties of the compounds in the sample.

Experimental

DMSO (GC-headspace tested, ≥99.9%), DMF (GC-headspace tested, ≥99.9%), 3-phenoxylbenzyl chloride (590932, ≥97%) and 3-phenoxybenzaldehyde (≥98%) were purchased from Sigma-Aldrich. Acetonitrile (HPLC grade) was purchased from Burdick&Jackson. Water for HPLC analysis was in house generated MilliQ water. DMS (≥99%) was purchased from Alfa Aesar and phosphoric acid (85%) was purchased from EMD. Permethrin API was provided by Merial Inc. GC columns were purchased through VWR and HPLC columns were purchased from Waters.

The GC system was an Agilent GC 7890 equipped with a FID detector and an Agilent 1888 Headspace auto sampler. HPLC system was an Agilent 1200 equipped with diode array detector. Data acquisition and processing were accomplished using Empower® 3 software.

Details of sample preparation, GC conditions and HPLC conditions for PMN analysis and kinetic study are described in the ESI.†

Acknowledgements

The authors would like to thank colleagues from Merial Analytical R&D for their support and helpful suggestions during this study.

References

1. M. Elliott and N. F. Janes, Chem. Soc. Rev., 1978, 7, 473.
2. D. Taplin, T. L. Meinking, S. L. Porcelain, P. M. Castillero and J. A. Chen, J. Am. Acad. Dermatol., 1986, 15, 995.
3. V. Andriantsoanirina, A. Izri, F. Botterel, F. Foulet, D. Chosidow and R. Durand, Clin. Microbiol. Infect., 2014, 20, O139.
4. S. Oberoi, R. S. Ahmed, S. G. Suke, S. N. Bhattacharya, A. Chakraborti and B. D. Banerjee, Clin. Biochem., 2007, 40, 1321.
5. A. Anadón, M. R. Martínez-Larrañaga and M. A. Martínez, Vet. J., 2009, 182, 7.
6. Impurities Q3C(R5): Guideline for Residual Solvents, ICH Harmonised Tripartite Guideline, 2011.
7. J. Tian and A. Rustum, J. Pharm. Biomed. Anal., 2016, 128, 408.
8. G. S. Garusinghe, S. M. Bessey, C. Boyd, M. Aghamoosa, B. G. Frederick, M. R. Bruce and A. E. Bruce, RSC Adv., 2015, 5, 40603.
9. D. L. Head and C. G. McCarty, Tetrahedron Lett., 1973, 16, 1405.
10. C. Cheng, S. Liu, B. J. Mueller and Z. Yan, J. Chromatogr. A, 2010, 1217, 6413.
11. K. Mistry, N. P. Chetwyn, L. Dai, D. T. Yazzie, A. C. Quiroga, K. Zhang, H. B. Runes and M. W. Dong, LCGC North Am., 2010, 28, 56.
12. Identification and Control of Residual Solvents (2.4.24). European Pharmacopoeia, European Directorate for the Quality of Medicines & HealthCare, Strasbourg, France, 8th edn, 2013.
13. Residual Solvents 〈467〉. The United States Pharmacopoeia, USP38/NF33, The United States Pharmacopeial Convention, Rockville, MD, USA, 2012.
14. W. W. Epstein and F. W. Sweat, Chem. Rev., 1967, 67, 247.
15. D. H. R. Barton, B. J. Gardener and R. H. Wrightman, J. Chem. Soc., 1964, 1855.
16. K. E. Pfitzner and J. G. Moffatt, J. Am. Chem. Soc., 1963, 85, 3027.
17. K. E. Pfitzner and J. G. Moffatt, J. Am. Chem. Soc., 1965, 87, 5661.
18. K. Omura and D. Swern, Tetrahedron, 1978, 34, 1651.
19. S. G. Smith and S. Winstein, Tetrahedron, 1958, 3, 317.
20. K. Torssell, Tetrahedron, 1966, 37, 4445.
21. N. Kornblum, J. W. Powers, G. J. Anderson, W. J. Jones, H. O. Larson, O. Levand and W. M. Weaver, J. Am. Chem. Soc., 1957, 79, 6562.
22. H. R. Nace and J. J. Monagle, J. Org. Chem., 1959, 24, 1792.
23. N. Kornblum, W. J. Jones and G. J. Anderson, J. Am. Chem. Soc., 1959, 81, 4113.
24. J. E. Leffler and E. Grunwald, Rate and Equilibria of Organic reactions, Dover Publications, Inc., New York, 1989, p. 60.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra19204a

---