Hazard ranking classification for aquatic organisms.
Ionic liquids (ILs) have gained significant attention within the academic and industrial circle owing to their attractive and unique characters. However, the usual green image of the ionic liquids mainly associated with their low vapour pressure has become increasingly doubtful. Several recent studies have highlighted the underestimated ILs toxicity which has not been adequately addressed. Therefore, improving the understanding of the ionic liquids toxicity towards aquatic organisms will undoubtedly lead to formulation of right solutions to address the toxicity problem hence contributing towards the development of green and sustainable ILs‐based technology. The chapter provides a collective review of studies conducted on the effect of ILs structure on toxicity, specifically focussing on the various types of cations and anions, and the length of the alkyl chain attached. Based on the qualitative outcome from the review, a discussion on the development of statistical modelling on the impact of ILs structural features towards the overall toxicity is presented. The application of quantitative structure activity relationship (QSAR) for developing the predictive model for toxicity is highlighted.
- ionic liquids
- Vibrio fischeri
- structural features
One of the major developments made on green solvents focuses on the design of new and more environmental friendly solvents. From the green chemistry perspective, green solvent should be non‐toxic, readily biodegradable and is synthesized using the environmental friendly synthesis procedure, whilst at the same time able to meet the application target technologically and economically [1, 2]. For several years, ionic liquids (ILs) have been gaining significant attention as the candidate for future ‘green solvents’ from the scientific and industrial community. It has shown several advantages over the volatile organic solvents (VOCs), among others, covering three major aspects namely:
Extremely low vapour pressure in comparison to the VOCs resulting in insignificant vaporization losses to the atmosphere.
Inflammable as opposed to the flammable VOCs hence easier to handle and store.
Non‐toxic perception due to minute losses through vaporization into the atmosphere compared to the VOCs.
Whilst most studies have converged opinion on the first two aspects, the third has been increasingly contested. The work presented focuses on addressing the latter in order to provide further clarity pertaining to the issue of ILs toxicity. It has been generally agreed that the unique feature which enables ILs to capture significant interest is the ability to design them for specific application by changing the cation and anion coupling to meet specified physical, chemical and biological properties. To date, significant number of ILs has been developed and most of them are now commercially available. In addition, there are few more millions of possible ILs that could be theoretically synthesized .
ILs are mainly designed to be inflammable, non‐volatile, and non‐explosive media with a high thermal stability . Due to their hardly measurable vapour pressure, they are not expected to contribute towards atmospheric pollution. On the contrary, most of them display high aqueous solubility. Even the supposedly most hydrophobic IL was found to exert some degree of solubility hence allowing their possible dispersion into aquatic systems, raising concerns on their subsequent environmental impact [5, 6]. Given the almost unlimited combinations of possible ILs that could be developed, the toxicity determination could become highly laborious and extremely costly as well as time consuming. Developing predictive methods would require systematic understanding of the complex interactions between the cation and anion pairings leading to the toxicity properties, which are not easily done.
On another note, recent studies have pointed towards some possible draw backs on the use of ILs on an industrial scale. One of the major concerns highlighted was from aquatic toxicity studies showing potential drastic impact of some ILs which was considered as being green, on various aquatic organisms. A number of these ILs were found to possess higher toxicity than some of the acute organic solvents. Therefore, evaluating their overall toxicity has become of primary interest to the industries and public at large prior to their bulk application. This necessitates the development of a predictive method to substitute the laborious manual toxicity measurement in light of the increasing interest on ILs applications.
It has been known for some time that the structural feature of ILs may have different contributions on the ILs overall toxicity. Hence, a systematic study to assess the variation of the ILs structure on its overall toxicity has to be commissioned separately. The review attempted at investigating the influence of the changes in ILs structural features involves: (i) the cation core and the functional group substituent, (ii) the length of the alkyl chain substituent, and (iii) anion nature, on the ILs toxicity using bioluminescent
2. Ecotoxicity measurement using
|Hazard ranking||Concentration of ILs in mg L−1|
|Ionic liquid names||Abbreviation||EC50 in mg L−1|
2.1. Effect of the cation core
Numerous cations have been used to create ILs such as imidazolium and pyridinium, which have been appearing mostly in past ILs studies particularly for the room temperature ionic liquids (RTILs). In the study, the influences of the cation core on the ILs toxicity were investigated using imidazolium‐, pyridinium‐, pyrrolidinium‐, piperidinium‐ and morpholinium‐based cations. The structure with regards to the chain length variation on the cation core was kept within 1‐butyl‐(1 or 3)‐methyl (cation) bromide ILs , as shown in Table 2 , group A.
Although the EC50 values for the five ILs fall under the same category i.e. practically harmless, the impact of the cation variations on the overall toxicity are found to be obvious. The result on EC50 values highlighted that the imidazolium‐based ILs exhibited about 4–5 magnitude higher toxicity measurement compared to piperidinium‐ and pyrrolidinium‐based ILs respectively. The most toxic IL is the one based on pyridinium cation where a slightly higher toxicity values were observed compared to the imidazolium analogue. Meanwhile the morpholinium cation demonstrates far less toxicity behaviour than the other counterparts with EC50 value reaching as high as 66,729 mg L-1. The piperidinium and morpholinium exhibited almost similar cationic core structure where the latter can simply be established by replacing the carbon atom located opposite to the amine group in the piperidinium structure, with an oxygen atom. Despite the slight structural differences, the presence of the oxygenated atom in the morpholinium cation led to its significant toxicity reduction in the order of 17 times, compared to the piperidinium‐based IL. This finding augurs well with the earlier work reported by Samorì et al. [16, 17].
The toxicity of some starting reactant for the cation used in the ILs synthesized in the study were tested against
Generally, it was extremely difficult to establish sensible and systematic structure‐toxicity relationship with exception to the observation involving relationship between the position and the number of methyl groups attached to the pyridine cation and the effect on toxicity of the ILs as a whole. Also, the result indicates an important general trend which shows that the ILs with the cation structure containing aromatic, are always more toxic than the non‐aromatic ones. According to Ventura et al.  and Kurnia et al. , the aromatic cation are more soluble in water and therefore capable of directly exhibiting its high toxicity effect on the aqua environment compared to the non‐aromatic‐based ILs which are much less soluble. However, it is worth to note that the toxicity of the non‐cyclic cations such as sulphonium, ammonium and phosphonium has not been rigorously studied. Nevertheless, the present study still takes into account of the toxicity involving few ammonium and phosphonium ILs reported earlier and the results are included.
2.2. Effect of alkyl chain length
It is known that the alkyl chain length of the cation has a strong effect on the physical and chemical properties of the ILs. For example, the extension of the cation side resulting from a longer alkyl chain commonly results in lower density and solubility, slower diffusion rate and increases viscosity [19, 20]. In addition, it was also observed that the cation alkyl chain length has a pronounced effect on the ecotoxicity towards microorganisms. Most of the established ecotoxicity data covering not only
The type of cation also seems to have an important role in changing the toxicity effect of the ILs when its’ substituent chain length are further extended. For the shorter alkyl chain (butyl), the pyrrolidinium cation demonstrated lower toxicity compared to piperidinium and imidazolium, respectively. The recorded EC50 value of [C4mpyrr][Br] is 5525 mg L-1 and this marked a reduction of 109‐fold in the EC50 value, signifying a tremendous increase in toxicity when the side alkyl chain was extended from four to eight carbon atoms. For the piperidinium type, the recorded effect was even more drastic with the toxicity increased by 135‐fold from 3958 mg L-1 for [C4mpip][Br] to 29.3 mg L-1 for [C8mpip][Br]. Between the three cations, imidazolium recorded the lowest rate of increase in toxicity with the increase in the alkyl chain length, where the increment was only by 20‐fold from 1002 mg L-1 for [C4mim][Br] to 50.9 mg L-1 for [C8mim][Br]. The lowest rate of increase in toxicity could be explained by the fact that imidazolium is already possessing highest toxicity even with shorter alkyl chain length compared to the others.
A similar study has been conducted for pyridinium‐based ILs by Docherty and Kulpa  who studied the effect of butyl, hexyl and octyl substituents on the toxicity of 1‐alkyl‐3‐methylpyridinium bromide‐based ILs. A similar increasing trend in the toxicity associated with increasing alkyl chain length was observed. As expected, the least toxic compound of this cation type is the butyl‐based ILs with EC50 value of 130.48 mg L-1. The EC50 value reduces to 29.99 mg L-1 through the addition of two carbon atoms to the alkyl chain (hexyl) resulting in a fourfold increase in toxicity. Switching the hexyl substituent to an octyl causes an increase in toxicity i.e. by 17‐fold compared to the hexyl‐based IL and up to 73‐fold compared to the butyl‐based IL.
Generally, it can be stated that increase in the side chain for the pyrrolidinium‐ and piperidinium‐based ILs produces more pronounced effect than those of the aromatic‐based cation such as imidazolium‐ and pyridinium‐based ILs. Although there were almost zero data reported for much longer alkyl chain except for imidazolium‐type ILs which will be further discussed below, the increasing trend of the toxicity with respect to the increase in the side alkyl chain reveals that longer extension on the alkyl chain beyond C9 and above, will produce highly toxic ILs and therefore should be avoided.
For the longer alkyl chain, Ranke et al.  studied the toxicities for 1‐alkyl‐3‐methylimidazolium tetrafluoroborate with alkyl chain varied from C3 to C10. Based on their EC50 values, as can be seen from Table 2 group B, the tetrafluoroborate‐based ILs having alkyl chain length up to heptyl, can still be classified as practically harmless. Further extension of the alkyl chain will lead to more toxic effect on the
Despite the above highlighted findings, a temporary reverse effect was observed for the ILs when the alkyl chain reaches decyl, where a noticeable increase in the EC50 value indicating reduction in the toxicity of the ILs. The observed effect continues until the chain length reaches C17 before reversing back to the earlier trend. As a result, the hazard ranking of [C18mim][Cl] changes back to a slightly toxic effect, similar to the earlier hazard ranking for the ILs when the alkyl chain was at C8. Hence it can still be concluded that for long alkyl side chains beyond C8, the dependence between the chain length and toxicity is still valid.
The toxicity of hydrophobic [Cnmim][NTf2]‐based ILs was also discussed by Ventura et al. . For this ILs, the alkyl chain is varied from C2 to C8 (see Table 2 group B). The reported EC50 values are found to be comparable and it follows a similar trend observed for the chloride and tetrafluoroborate‐based ILs. The exception was only observed for the short alkyl chain, where [NTf2]‐based ILs exhibited higher toxicity than their corresponding [BF4] and [Cl] anions. For instance, the EC50 value for [C2mim][NTf2] is 330.23 mg L-1 which is 5.6‐fold more toxic than [C3mim][BF4] despite having shorter alkyl chain attached to the cation, and 9.5‐fold more toxic than [C2mim][Cl]. On the contrary, for the [C8mim]‐based ILs, much smaller differences in EC50 values were reported for the three different anions, where the values for [C8mim][ Cl], [C8mim][NTf2] and [C8mim][BF4] were 2.36 mg L-1, 6.44 mg L-1 and 7.25 mg L-1, respectively. This could be explained based on earlier reported work that the
Phosphonium‐based ILs also showed a similar behaviour as discussed earlier where its toxicity reduces as the alkyl chain length grew longer than decyl. For instance, the reported EC50 value for trihexyl(tetradecyl)phosphonium bromide is 1449.09 mg L-1, eightfold greater than the EC50 value of tetrabutylphosphonium bromide i.e. 174.09 mg L-1 , as presented in Table 2 group B. The phenomenon is well reported in the literature for highly lipophilic substances (log
The increasing trend in toxicity with alkyl chain length was also confirmed for quaternary ammonium‐based ILs . As can be seen from the tabulated data shown in Table 2 group B, tetramethylammonium bromide and tetraethylammonium bromide show non‐toxic behaviour with EC50 values greater than 5000 mg L-1. The toxicity of this ILs increases i.e. EC50 value reduces to 600.28 mg L-1, for tetrabutylammonium bromide. Further noticeable increase in toxicity was observed for hexyltriethylammonium bromide with EC50 value reduces to 64.65 mg L-1.
Generally it can also be seen that the quaternary ammonium‐based ILs exhibits lower toxicity against
2.3. Effect of the anions
The anion chemistry has a great impact on the alteration of the ILs properties. Most of ILs properties such as melting point, hydrophobicity, chemical and thermal stabilities, ability to dissolve organic and inorganic solutes and miscibility with organic solvent rely mainly on the type of the anion [27–29]. Although there is no clear pattern that could be drawn for the anion influence on the ILs toxicity, recent studies have given more attention towards the impact of anions type on ILs toxicity. The data reported for the anion effect are tabulated in Table 2 group C. Ventura et al.  investigated the toxicity of 10 ILs with 9 of them comprising the cholinium cation with different anions. Cholinium‐based ILs has received significant attention due to its non‐toxic and biocompatible nature [31–33]. Using cholinium as the cation, the study on the impact of various anions on toxicity of the ILs was conducted. The bicarbonate anion was found to be the least toxic whilst the dihydrogen citrate being the most toxic. In fact, there was no EC50 value reported for cholinium bicarbonate due to its unnaturally high value. The maximum luminescence inhibition caused by this IL was as low as 35% at a very high concentration of 20,000 mg L−1.
Judging on the anion structure, the butanoate anion which corresponds to the addition of propanoate and acetate anions with one and two methyl group, respectively is expected to possess higher toxicity due to longer alkyl chain on its structure. However, the toxicity reported for the cholinium cation did not show any consequence leading to noticeable toxicity increase with increase in the alkyl chain on the anion side. Except for cholinium bitartrate and cholinium dihydrogencitrate which are classified as moderately toxic with EC50 values lower than 100 mg L-1, the other seven compounds reported fell under practically harmless class. The reported ecotoxicity data also demonstrated that some classic ILs, pairing the imidazolium or pyridinium cation with alkyl chain varied from C1 to C6 with similar anions, may possess lower toxicity than the one exerted by the cholinium cation. Hence, contrary to the effect of alkyl chain length on the IL cation towards ILs toxicity, the same effect is proven to be inconclusive for the anion.
Overall, the order of toxicity sequence reported in increasing order is bicarbonate < butanoate < acetate < dihydrogenphosphate < propanoate < chloride < salicylate < bitartrate < dihydrogencitrate.
Another recent work concerning the anion impact on IL toxicity is the study conducted by Peric et al.  where they investigated the ecotoxicity of compounds based on substituted amines as the cations including monoethanolamine, [2‐HEA], diethanolamine [2‐HDEA] and triethanolamine [2‐HTEA], paired with organic acids with different numbers of carbon atoms (formic, propionic, butanoic, isobutanoic and pentanoic acid) as the anions. From the reported data, it is apparent that the alkyl chain length did show some influence on the ILs toxicity. The 2‐hydroxydiethanolamine pentanoate [2‐HDEA][Pe] was found to be the most toxic with EC50 values of 350 mg L-1, followed by 2‐hydroxytriethanolamine pentanoate [2‐HTEA][Pe] with EC50 of 461 mg L-1. The two ILs have pentanoic as its anion, indicating the presence of five carbon atoms in the anion side chain. The toxicity exerted by the two ILs demonstrated stronger influence by the cation structure although [2‐HDEA][Pe] displayed a 1.3‐fold higher toxicity than the [2‐HTEA][Pe] despite the latter having a larger cation size. The lower toxicity of [2‐HTEA] is attributed mainly to the presence of three hydroxyl groups (-OH) in the cation structure as opposed to only two in [2‐HDEA]. As a matter of fact, the observation agrees well with the conclusion stated earlier on the influence of the oxygenated atom within the cation structure which tends to lower the ILs toxicity. Interestingly, this is not always true for the butanoate anion as the toxicity trend displayed different results. The 2‐hydroxydiethanolamine butanoate [2‐HDEA][B] was reported to be 1.6‐fold less toxic than the 2‐hydroxytriethanolamine butanoate (2‐HTEAB) despite having more hydroxyl (-OH) group attached to it. In this respect, the authors argued that the noticeable increase in the toxicity is due to the more dominant effect contributed from the longer alkyl chain length of the anion. The argument is rather controversial as it was earlier explained that the trend on the effect of the alkyl chain length of the anion on toxicity is inconclusive and lesser compared to the cation. Nevertheless, all the studied ILs were non‐toxic and fell under the class of practically harmless.
It is worthy to highlight that it may be worthwhile to use some of the carboxylic acid anions such as decanoic and undecanoic acid which have longer alkyl chains, to synthesize novel ILs for undertaking the study to confirm the effect of the alkyl chain of the anion to the ILs toxicity.
More recently, Montalbán et al.  studied the toxicity of 1‐ethyl‐3 methylimidazolium‐based ILs using six different anions including PF6, TfO, NTf2, Cl, Ace and EtSO4. Due to the short alkyl chain used, the six ILs possess very low toxicity and hence all of them were classified as practically harmless. The reduction in toxicity of these ILs follow the trend of Ace > TfO > NTf2 > EtSO4 > PF6 > Cl. Contrary to the above finding, Alvarez‐Guerra and Irabien  studied the toxicity of tetrachloroferrate(III) anion with the same cation and found a very low EC50 value i.e. 9.99 mg L-1 indicating highly toxic behaviour. Clearly, the [FeCl4] anion contributes to the extreme toxicity effect when compared to the earlier six anions and the alkyl chain length used. The [FeCl4] found to possess toxicity of 132‐fold higher compared to [Ace] which was the most toxic among the former anions. According to the QSAR model developed by Alvarez‐Guerra and Irabien , it was similarly suggested that the highly toxic behaviour exerted by the anion was due to the presence of Fe in its structure, exerting a significant influence on the toxicity.
Based on past reported work, introducing amino acid anion to the structure of the ILs can be considered as a convenient approach to reduce its toxicity. Interestingly, the claim has yet to be experimentally proven. In our recent work , the variation of the toxicity of 1‐(2‐hydroxylethyl)‐3‐methylimidazolium, [C2OHmim] by pairing it with four different types of amino acid anions namely glycinate, alaninate, serinate and prolinate was studied towards the
Generally, it can be argued that the impact contributed by the anion on the toxicity did not show any strong systematic relationship involving its structure i.e. alkyl chain length as demonstrated in the case of the cation. However, changing the type of the anion would be crucial as it can change the ILs’ toxicity significantly.
2.4. Toxicity of starting material and organic solvent
Volatile organic solvents (VOCs) are considered as a major source of atmospheric pollution. They exert high vapour pressure hence have high volatility leading to their significant losses to the atmosphere. Their vapour can be highly toxic depending on the type of component and its respective concentration thus posing toxic exposure to process operators and the surrounding community. This is where ILs, having significantly low volatility, was promoted by many researchers as the ideal replacement to VOCs for many of the industrial application. However, several studies have reported comparable EC50 values towards
3. Quantitative structure activity/property relationship (QSAR/QSPR)
Quantitative structure activity/property relationships (QSAR/QSPR) are models which can be used to predict the relationship between the chemical compound structure and a desired end measure which could refer to any type of physical, chemical or biological activities/properties . Reliable experimental data is crucial at the model development stage in order to produce good prediction model. However, the more extensive the experimental data used, the more time and resources required causing higher cost.
The main aim of the QSAR development is to develop reliable predictive models with minimum possible experimental data thus reducing the time and resources required. The basic principles of the development of the QSAR/QSPR as outlined by Todeschini and Consonni  are;
The property of interest of the studied compound must have some strong form of relationship with their molecular structure.
Similar compounds judging from the orientation of the molecular structure, must behave in a similar fashion.
As the QSAR/QSPR modelling involves computational work, it reflects the benefits as below;
low costs and high productivity levels especially dealing with data from large chemical libraries,
more environmental friendly approach leading to reduction in necessary chemical experiments and/or animal testing, which could be further reduced with selection of good descriptor linking the molecular structure to the property of interest,
possibility to predict properties of newly synthesized compounds based on its chemical structure without the need to conduct any experimental or test procedure.
In this study, ILs toxicity is the property of interest and an attempt was made to develop efficient relationship with the ILs molecular structure. There have been several QSAR models reported for predicting ILs ecotoxicity against
The evaluation of the model accuracy and stability are highly important. The correlation coefficient (
Different methods were adopted to validate the QSAR/QSPR models such as internal, external and cross‐validation . Most of the developed models used the multiple linear regression technique (MLR) which was widely employed due to its simplicity, transparency and reproducibility as well as easy interpretability . It also provides useful statistical parameter for evaluating the significance of the selected descriptors (i.e.
In most cases reported, the end measure for ecotoxicity of ILs towards
|No of IL||Method||Molecular descriptor||Descriptor type||Ref|
|25||GFA a||4||Electronic, spatial, structural, thermodynamic, and topological descriptors||0.78|||
|148||PLS‐DA b||94||Negative or positive effect||Tr = 0.963|
|10 & 19||MCA c||1||Number of aliphatic carbons & number of carbons in the cation||0.934 & 0.861|||
|147||MLR||12||DRAGON||Tr = 0.936|
Te = 0.757
There were two aspects clearly noticeable from the comparison made across all the published models. These are the variation in the dataset size and the number and type of molecular descriptors used. It is worth highlighting that adding new ILs which pair different elements in their structure, not considered in the earlier models, would require new set of descriptors for better molecular representation and good applicability of the QSAR model. For instance, Viboud et al.  developed two linear QSAR models for relatively small datasets containing 10 and 19 ILs. The first dataset which comprised pyridinium bromide‐based ILs was expanded by the addition of 9 ILs pairing different cations (imidazolium, pyrrolidinium and piperidinium). Although a single descriptor was used in the model construction, good correlation coefficients was achieved i.e. 0.934 and 0.861, respectively. However, it is clear that expanding dataset size would affect the model accuracy resulting in the accuracy drop. Therefore, many QSAR models developed to cover relatively huge dataset were constructed using higher number of molecular descriptors to cover the significant variation in all the molecular structure. So far, the largest dataset used comprised 157 ILs covering 74 cations and 22 anions, studied by Yan et al. . They used linear regression to propose a predictive model with good correlation coefficient i.e.
Overall, it can be concluded that the application of proper method for the selection of molecular descriptor, and the model validation method used, become the key factors in influencing the outcome of the QSAR/QSPR model developed. With the right selection of the molecular descriptors, the accuracy and reliability of the predictive models developed could be enhanced significantly. With such capability, the design of ILs for any application could be made to consider its toxicity thus enabling greener ILs developed for industrial application right from the design stage.
The collective study on the relationship between ILs ecotoxicity towards luminescent marine bacterium
There seems to be consensus from the past literature that the alkyl chain length appeared to be the dominant parameter controlling the ILs toxicity towards different aquatic organisms including
Besides the above factors highlighted on ILs cation affecting its toxicity, the proper selection of the anion moiety could also have impact on controlling its toxicity. With the exception of the FeCl4 anion which showed very high toxicity behaviour towards
Overall, from the aquatic toxicity point of view, ILs may not seem to necessarily perform better when compared to the organic solvent, which it supposed to replace for many industrial applications. However, considering their negligible impact on the atmosphere as a result of extremely low vapour pressure as well as being non‐flammable, and coupled with the unique tuning ability to meet specific industrial requirements, ILs can still be largely considered as promising class of greener material. In view of the need to perform the toxicity assessment to confirm fully green behaviour, the QSAR/QSPR method can be the key towards providing the predictive ability which could guide the design of novel greener ILs for industrial application.