Introduction

Salts with melting points lower than 100 °C are referred to as ionic liquids (ILs) [1]. Since their rediscovery at the end of the last century, these innovative fluids found applications in different branches of chemistry such as liquid-liquid extraction [2], microextractions [3, 4], solvents for extraction of DNA [5, 6], mass spectrometry [7,8,9,10], electrochemistry [11, 12], spectroscopy [4, 13], high-performance liquid chr``omatography [14, 15], and capillary electrophoresis [16, 17]. Thanks to their extremely low vapor pressure, non-flammability, wide liquid temperature range, and high thermal stability, ILs have been successfully applied in high temperature environments (such as liquid thermal storage media [18], high-temperature lubricants [19, 20], lubricant additives [21, 22], solvents for high-temperature reactions [23,24,25], gas chromatography (GC) stationary phases [26,27,28], solvents for headspace GC [29,30,31,32], high-temperature non-flammable electrolytes for batteries [33]) and high vacuum environments (high-performance lubricating greases for space [34] and high vacuum lubricants [35, 36]). In fact, the IL thermal stability can vary considerably depending on the nature of the IL anion and cation. Bis- or dicationic, imidazolium-based or phosphonium-based ILs often show decomposition temperatures higher than traditional monocationic ILs [37, 38]. The anion also can play a significant role in the IL thermal stability, e.g., nucleophilic halogens can attack the cation moiety [37, 39].

The most striking features of ILs are their low volatilities and high thermal stabilities, compared to common organic solvents. There are several methods reported for determining thermal stabilities of ILs. The short-term IL thermal stability is determined by thermogravimetric analysis (TGA) using temperature ramps of 5–10 °C/min and measuring the temperatures of 5% weight loss (Td) [38,39,40]. The Td temperatures are useful to compare the thermal stabilities of different ILs [38]. Since thermal stability is a time-dependent process, different results may be obtained when ILs are exposed to thermal stress for longer periods [41]. The long-term thermal stability, determined by monitoring weight loss by isothermal TGA experiments, provides detailed information on the stabilities/volatilities and/or rates of decomposition of ILs [39]. In kinetic studies, determination of frequency factor and activation energies, using the Arrhenius equation, provides information about the frequency of molecular collisions (collisions between cations and anions) and the proportion of collisions resulting in decomposition reactions respectively [42]. The long-term thermal stability is also determined by inverse gas chromatography, in which the column bleed is monitored by highly sensitive detectors as a function of temperature [27, 37, 38]. The observed mass loss upon heating can be due to volatilization (formation of neutral ion pairs (NIPs)) or to real thermal decomposition (TD) [39, 43]. Along with mass loss, ILs also can show color changes after heating at high temperatures, long storage times, and wear/friction in tribological applications which can be related to decomposition and formation of side products [23, 42, 44].

Most IL literature focuses on the determination of thermal stabilities, methods to determine thermal stabilities, rates of decomposition, etc. Little attention has been paid to the analysis of decomposition products, the reasons for weight loss, and color changes in ILs at higher temperatures. Recent reports have shown that the thermal stabilities of ILs can be altered by structural modifications [28, 38, 40, 45]. These reports compile trends of the “structure-thermal stability” relationships of ILs but do not provide any detailed explanation about the effects of various structural moieties on the IL decomposition. There are only a few reports on the thermal degradation products of some imidazolium and pyrrolidinium monocationic ILs [42, 46,47,48] and no reports on the analysis of decomposition products of dicationic ILs. The degradation studies of imidazolium-based monocationic ILs were done by pyrolysis GC-MS and established that the decomposition started by the removal of alkyl substituents on the imidazole ring [46, 47], while the monocationic pyrrolidinium-based ILs with cyano anions were observed to polymerize at higher temperatures [48].

Considering the popularity of dicationic ILs in high temperature applications, the results of a degradation study can be helpful for their optimal use, determining their changes at high temperatures and for providing information on possible structural modifications to further improve stability. The present work summarizes the results of a systematic and comprehensive thermal decomposition study performed on 15 bis(trifluoromethylsulfonyl)imide (NTf2) dicationic imidazolium, pyrrolidinium, and phosphonium-based ILs at 400–440 °C. The effect of different linkages such as straight alkane chains, polyethylene glycol chains, and branched alkane chains on the thermal behavior of ILs also was evaluated via these degradation studies. The soft ionization technique, electrospray ionization mass spectrometer (ESI-MS), was used to avoid the further fragmentation of decomposition fragments and to identify them, in contrast to what can be accomplished with pyrolysis GC-MS.

Materials and methods

Instruments

The analyses were performed on a mass spectrometer Finnigan LXQ ESI-MS (Thermo Fisher Scientific, San Jose, CA). The ESI analysis conditions were as follows: spray voltage 4.5 kV, sheath gas flow rate 35, aux gas flow rate 6, capillary temperature 275 °C, capillary voltage 16 V, tube lens 110 V, scan range 50–1000 m/z, positive mode.

Materials

1-Benzyl-2-methylimidazole (90%), 1,2-dimethylimidazole (98%), N-methylpyrrolidine (97%), tripropylphosphine (97%), 1-(2-hydroxyethyl)imidazole (97%), 1,9-dibromononane (97%), 1,5-dibromopentane (97%), 3-bromo-2-bromomethyl-1-propene (97%), 1-bromo-3-(2-bromoethyl)-4,4-diethylpentane, 1,3-dibromobutane (97%), tetraethylene glycol (99%), phosphorus tribromide (99%), and bis(trifluoromethanesulfonyl)imide lithium salt (99.95%) were purchased from MilliporeSigma (ex-Sigma-Aldrich, St. Louis, MO). 1,5-Dibromo-3-methylpentane (98+%) was purchased from Alfa Aesar (Haverhill, MA). 1,5-dibromo-3,3-dimethylpentane (95%) was obtained from Alfa Chemistry (Holtsville, NY). 1,3-Dibromo-2-methylpropane (95+%) was purchased from Oakwood Chemicals (Estill, SC). The detailed procedures for synthesis of ILs are reported in the literature [38, 49, 50].

Experimental design

A weighed amount of each IL was introduced in a closed vessel that could be heated up to 400 °C. Argon was circulated through the vessel to carry any volatiles generated by the IL upon heating. These volatiles were trapped in a flask cooled to − 78 °C by an acetone-dry ice bath. After 30 min at 400 °C, the vessel was cooled, and the collection flask was disconnected. The heated IL was visually inspected, and 10-mL methanol was introduced into the container. The methanolic solution was collected and filtered for MS analysis for identification of non-volatile (residual) degradation products in the heated IL. Similarly, 10-mL methanol was added to the volatile-collection flask for analysis by direct infusion MS and identification of volatile degradation products (see Electronic Supplementary Material (ESM) section 1.1 for full details).

Results and discussion

Fifteen different dicationic ILs (IL1 to IL15), either di-imidazolium, di-pyrrolidinium, or di-phosphonium paired with bis(trifluoromethylsulfonyl)imide (NTf2) anions were synthesized. Five dicationic ILs were made with a nonane (C9) alkyl spacer, three with a polyethylene glycol (PEG, three units) spacer, three with branched C3 spacers, one with a C5 linkage spacer, and three with branched C5 spacers [38, 49]. Their thermal behavior was tested as per the procedure outlined in the “Experimental design” section.

Analysis of straight chain linkage ILs

Table 1 lists the structure of the eight ILs, coded IL1 to IL8, along with, the three major peaks seen in the mass spectra of the residual ILs after 30 min heating at 400 °C, in decreasing intensity order. The mass spectra of the collected volatiles were much richer in peaks than those of the residual heated ILs. Table 2 lists similarly the four major peaks seen in the spectra of the collected volatiles along with the bonds broken to form these fragments. Figure 1 compared the mass spectrum of the original IL1 (Fig. 1a) with that of the 400 °C heated IL1 residue (Fig. 1b), and the corresponding spectrum of the collected IL1 volatiles after 30 min of heating at 400 °C (Fig. 1c).

Table 1 MS peaks of residual (non-volatile) dicationic NTf2 ionic liquids seen after heating at 400 °C for 30 min under argon flow
Full size table
Table 2 MS peaks of the volatiles emitted by dicationic NTf2 ionic liquids after heating at 400 °C for 30 min under argon flow
Full size table
Fig. 1
figure 1

Mass spectra of IL1. a Original IL. b Residual IL1 after heating at 400 °C for 30 min. c Methanolic solution of collected IL1 volatiles after heating at 400 °C for 30 min. See ESM Fig. S6 for all possible IL1 fragments

Full size image

On visual inspection, apart from some darkening, IL1 is practically unchanged after heating at 400 °C for 30 min. The mass spectrum of the heated IL1 is practically identical to the initial spectrum showing a very small 152 m/z peak, likely the result of N-methyl elimination (Fig. 1a, b). A color change is observed (see ESM Fig. S3), and trace amounts of volatiles were collected giving the Fig. 1c mass spectrum. As listed in Table 2, the 207 m/z peak corresponds to the loss of a dimethylimidazole group plus a methylene group. The possible mechanism for the loss of alkyl substituents on the imidazole ring is shown in Fig. 3. The loss of alkyl substituents on the imidazole ring due to the nucleophilicity of an anion, known as the “reverse Menshutkin reaction” was observed in case of ammonium and imidazolium cations paired with halide ions [39, 46, 51, 52]. Since halides being more nucleophilic, the reaction can occur at lower temperatures (< 100 °C), decreasing the thermal stability of ILs. [39] The NTf2 anion is less nucleophilic compared to the halides, which provides higher thermal stabilities and the reverse Menshutkin reaction occurs at higher temperatures (> 350 °C). Further fragmentation was observed with loss of methylene groups from the alkyl chain with intervals of m/z 14 (Fig. 1c and see ESM Fig. S6). The sequential loss of CH2 groups is a characteristic feature of alkanes observed under electron ionization experiments [53].

Figure 2 shows the mass spectra of the original IL5 (Fig. 2a), the IL5 residue (Fig. 2b) collected after heating at 400 °C, and the collected IL5 volatiles after 30 min of heating at 400 °C (Fig. 2c). An important point to note is that the mass spectra of the collected volatiles of phosphonium ILs with alkyl (C9) linkage, IL5, and IL6 showed small peaks at respectively m/z 223 and 325 (Fig. 2c and ESM Figs. S18 and S21). They correspond to the intact dicationic IL demonstrating the possibility of volatile NIP formation for these two ILs. The mass spectrum of the collected volatiles was more complex compared to the residual IL and showed multiple fragments due to the breaking of C-P bonds and C-C bonds (Table 2, Fig. 2). The mass spectra of residual IL5 and collected volatiles showed peaks at m/z 419 and 199. The peaks were of interest, as they belonged to the formation of oxides by IL5 at high temperatures. The possible mechanism for the formation of phosphine oxides is shown in Fig. 3. Despite the inert atmosphere, trace moisture content of the argon may have led to the oxide formation. The formation of phosphine oxides was also observed during the thermal degradation study of quaternary phosphonium-based montmorillonites [54, 55]. Like IL5, IL6 also showed the formation of oxides at high temperatures (see ESM Figs. S19 and S21). The results for IL6 were similar to IL1, with a practically unchanged mass spectrum compared to the original IL, and a spectrum of the trace collected volatiles showing phosphorus oxidation and some thermal breakings (Table 2) (see ESM Figs. S19 and S21).

Fig. 2
figure 2

Mass spectra of IL5. a Original IL. b Residual IL5 after heating at 400 °C for 30 min. c Methanolic solution of collected IL5 volatiles after heating at 400 °C for 30 min. See ESM Fig. S18 for all possible IL5 fragments

Full size image
Fig. 3
figure 3

Postulated mechanism for the decomposition of ILs. a Imidazolium-based ILs. b Phosphonium-based ILs

Full size image

The six other ILs had different behaviors. For IL2, IL5, and IL8, the original IL peak, observed before heating, was still visible in the spectrum of the heated residual IL (see ESM Figs. S8, S17, and S26). However, it was not the highest peak and was accompanied by several peaks denoting partial but significant thermal decomposition. Indeed, the corresponding spectra of the collected fumes showed a wealth of IL thermal fragments (Table 2) (see ESM Figs. S7, S16, and S25).

IL3, IL4, and IL7, the ILs with a three-unit PEG spacer, were completely destroyed by 30 min heating at 400 °C. Table 1 lists the m/z values of major peaks seen in the crowded MS spectra obtained by infusing the methanol solution that washed the black carbonaceous material remaining in the vial. The PEG-linked ILs started decomposing at about 300 °C as fragments due to the breaking of C-O bonds (possible mechanism shown in the Fig. 4) were observed in the mass spectra of collected volatiles at 300 °C. The decomposition at lower temperatures can also be seen with lower Td temperatures of all three PEG ILs (Table 2).

Fig. 4
figure 4

Postulated mechanism for the thermal decomposition of PEG-linked ILs

Full size image

To complete this study, the 400 °C-resistant ILs, i.e., IL1, IL2, IL5, IL6, and IL8, were submitted to a 440 °C test. For each IL, an ampoule containing 0.2 mL of IL was sealed under argon and placed in an oven able to generate a 440 °C regulated temperature. A second ampoule was similarly prepared in the air (no argon purging) and 440 °C heating. IL2 and IL8 were completely decomposed by heating at 440 °C for 30 min leaving black solid material in the vials. The IL5 and IL6 MS base peaks, corresponding to the intact phosphonium dicationic ILs with C9 spacer were observed (see ESM Fig. S16). The intensity of the m/z 419 (IL5) peak (oxidized IL) was, not surprisingly, much higher under air than under argon heating (see ESM Figs. S16 and S17). IL1 showed some darkening color but seemed to mostly withstand the 440 °C temperature both under argon and air (see ESM Figs. S4 and S5). There was more degradation under air than under argon as seen by the darker color of heated IL1 and the lower solubility of the residual material in methanol. However, the two mass spectra of the extracted methanolic solutions still showed the intact IL1 peak and some degradation fragments (see ESM Figs. S4 and S5).

The decomposition studies on dicationic ILs led to some specific observations: the ether bond present in the PEG linkage is the most thermally fragile. The next bond that seems to break under thermal stress is the C-N bond followed by C-C bonds. This order does not follow the strength of chemical bonds that is 358, 348, and 293 kJ/mol for the C-O, C-C, and C-N bond, respectively [56]. The listed strength of the C-P bond, 264 kJ/mol, is even weaker than the C-N bond. However, it seems that the phosphorus-based dicationic ILs are thermally stable, being able to form NIPs with two NTf2 anions. The probability of cleavage of a particular bond not only depends on the bond strengths but also on the stability of the fragments formed in the fragmentation process [53]. The rupture of the C-P bond may be initiated by oxidation by trace amounts of oxygen present in our set-up, inducing the thermal breakdown of phosphonium-based dicationic IL5 and IL6. It is worth mentioning that the thermal decomposition temperature is the upper limit of the liquid range of ILs and is much higher than the boiling points and decomposition temperatures of starting materials (N-methyl pyrrolidine: bp = 81 °C, tripropyl phosphine: bp = 187.5 °C, dibromoalkanes: bp range = 167 to 286 °C, tetraethylene glycol: bp = 330 °C with extreme decomposition) [37, 38, 57]. The lower volatilities of ionic liquids (salts) are due to the increased electrostatic and intermolecular interactions as is well-known [37, 39, 58].

Analysis of branched chain linkage ILs

To understand the effect of substituted linkage chains on the thermal stabilities of dicationic ILs, the degradation study was extended to a selection of C3- and C5-branched chain dicationic ILs. Table 3 summarizes the data obtained by the degradation study of six branched chain ILs and one linear analog, IL9 to IL15.

Table 3 MS peaks of volatile dicationic NTf2-branched ionic liquids seen after heating at 400 °C for 30 min under argon flow
Full size table

C3-branched ILs

The α-(IL9) and β-(IL10) substituted C3-linked branched ILs showed similar mass spectra for the collected volatiles at 400 °C (see ESM Figs. S28 and S30). The peak at m/z 151 was the highest in intensity and belonged to the fragment formed from of breaking of C-N bonds. The m/z 151 peak was also present in the mass spectrum obtained from the analysis of volatile fragments of IL9 collected at 300 °C. However, the same fragment was absent in the mass spectrum of collected volatiles of IL10 at 300 °C. This shows that the presence of methyl group on the α-carbon provided extra stability to the carbocation of the broken fragment. This is also supported by the decreased Td temperature of IL9 (406 °C) compared to IL10 (438 °C). In the case of IL11 with an external double bond on the C2 carbon, the fragment with m/z 137 was the base peak, and the fragment due to breaking of C-N bond (m/z 151) was about one third the intensity of the base peak (see ESM Fig. S32). The double bond on the C2 carbon may have accelerated the breaking of the C1-C2 bond.

C5-branched ILs

Four ILs with different substitutions patterns on the C5-branched linkages were analyzed. IL13, IL14, and IL15 contain a methyl, dimethyl, and a tert-butyl substituent on the central carbon (C3) of the C5 linkage, respectively. The fragmentation pattern of IL12 was similar to IL1 (both straight chain-linked ILs), with the breaking of the C-N bond (base peak at m/z 165) and further loss of -CH2 groups in series (see ESM Figs. S34 and S35). IL13 and IL15 (both with single substituents on the C3 carbon) showed similar fragmentation pattern with major peaks belonging to the breaking of the C-N bond, C1-C2 bond, and C2-C3 bond (see ESM Figs. S36, S37, S40, and S41). The presence of methyl and tert-butyl substituent on the C3 carbon may have accelerated the breaking of the C2-C3 bond as fragments at m/z 165 and 207 were base peaks (see ESM Figs. S36 and S40). These fragments must be the daughter fragments of a parent fragment obtained by the breaking of a C-N bond (fragments at m/z 179 and 221) of the intact IL. The Td temperatures of IL12 and IL13 are close (438 and 440 °C, respectively) which means that the C2-C3 bond must have broken after the cleavage of the C-N bond. The breaking of tert-butyl substituent was also observed in the case of IL15 as it is a good leaving group, and this might be the reason for the lower Td temp of IL15 (429 °C) compared to IL12. Interestingly, IL14 showed a different degradation pattern compared to the IL13 and IL15. The mass spectrum of the volatiles of IL14 (Td = 423 °C) was cleaner than the mass spectra of IL13 and IL15 (see ESM Fig. S38). The base peak observed at the m/z 193 was due to the breaking of C-N bond and a less intense peak observed at m/z 179 was may be due to the breaking of one of the methyl substituents on the C3 carbon. Removal of one methyl group leads to the formation of stable tertiary carbocation which might be the reason for its cleavage and the lower Td temperature of IL14 compared to IL13.

Conclusions

The results of the degradation study led to the following observations: (1) only the phosphonium-based ILs formed volatile NIPs found in the collected “volatile fractions.” All other tested dicationic ILs have undetectable volatility at 400 °C with no NIP formation. (2) All the studied ILs show at least some decomposition at 400 °C, but with considerable differences in magnitude. The dicationic ILs decompose even when their observed TGA temperature (Td) is much higher than 400 °C (e.g., 467 and 464 °C for IL1 and IL5, respectively; Table 1). (3) All three studied dicationic ILs with a PEG spacer were completely destroyed in less than 30 min at 400 °C with the prolific breaking of C-O bonds. (4) The phosphonium-based dicationic ILs are sensitive to P-oxidation at higher temperatures. (5) The bis-dimethylimidazolium IL1 with a nonyl (C9) spacer was remarkably thermally resistant even in the presence of oxygen. (6) The position and number of substituents on the linkage chain affects the degradation pattern and stabilities of ILs, as the presence of one substituent on the central carbon of the C-5 linkage accelerated the breaking of the C2-C3 bond. Whereas, the presence of two substituents on the C3 carbon leads to removal of one of the substituents and a further decrease in thermal stability. (7) The breaking of bonds under thermal stress studies did not follow the order of bond strengths and may be dependent on other factors such as the stability of fragments and the influence of substituents. (8) The heteroatom-carbon single bonds were more prone to thermal decomposition, and the order of thermal breaking was observed to be C-O < C-N < C-P in the studied ILs. In high-temperature applications, to be sure that dicationic ILs are stable and remain in a condensed state, it is recommended that their maximum working temperature be at least 100 °C or in some cases even lower than their experimental TGA decomposition temperatures.