COSMO-RS-Based Screening of Antisolvents for the Separation of Sugars from Ionic Liquids: Experimental and Molecular Dynamic Simulations.

The use of ionic liquids (ILs) in the biorefinery process has been increasing for the past few decades. In biorefinery, the separation process with respect to sugars needs to be evaluated for an efficient process design. Therefore, the present work aims to investigate the separation of sugars and ILs by means of a precipitation process using an antisolvent method. For this purpose, both theoretical and experimental studies were conducted. Initially, the conductor-like screening model for real solvents model was employed to screen the suitable antisolvents for the separation of sugars from the ILs. From the screening study, dichloromethane (DCM) and 1,2-dichloroethane were found to be the better antisolvents for the separation process. With the selected antisolvents, precipitation experiments were conducted for the mixtures involving four different sugars and three ILs at different experimental conditions. The process variables such as different antisolvents, sugars, ILs, antisolvent-IL molar ratios, and temperatures were examined in terms of their effect on sugar removal and IL recovery. DCM was found to be the most suitable antisolvent in this study with 90-99% of sugar removal and 80-98% of IL recovery. Further, molecular dynamics simulations were adopted to understand the structural properties of carbohydrates with ILs and antisolvents via interaction energies, hydrogen bonding, and coordination numbers. It was observed that the interaction energy between the sugars and IL plays a critical role in the removal of sugar. Higher the interaction energy between the sugars and IL, lower is the sugar removal.

Introduction

The continuous population growth has led to an increased consumption of fossil fuels for the production of energy and chemicals.[1] However, the changes made to harness the fossil resources have made a significant impact on the economic and environmental processes. To decrease the usage of fossil resources and nonrenewable resources, an alternative approach is required.[2,3] The utilization of lignocellulosic biomass is one such alternative which can act as a suitable energy source.[1,4] Biomass is an environmentally favorable and the most abundant renewable raw material that can be used as a fuel source.[5] The major components of lignocellulosic biomass are cellulose (30–50%), hemicellulose (25–35%), and aromatic polymer lignin (5–30%).[1,2,4] Currently, an extensive research is carried out for the conversion of lignocellulosic biomass to value-added chemicals.[6−8] The conversion of biomass consists of a set of chemical and biological processes, including separation, conversion of biopolymers into other valuable products, and fermentation reactions.[5,9−11] Because of the complex structure and strong intra- and intermolecular hydrogen bonding in the biomass, these processes require suitable solvents for the dissolution of its components.[5]

Over the past few decades, ionic liquids (ILs) are currently being explored as an alternative class of solvents in biorefinery processes.[12] ILs possess several significant properties such as negligible vapor pressure and high thermal and chemical stability which are useful for biorefinery processing.[7,12] Apart from the dissolution of carbohydrates, ILs also have been reported to solubilize the major polysaccharides (cellulose, hemicellulose, and lignin) and convert them into amorphous compounds which accelerate further enzymatic hydrolysis.[13,14] On the other hand, ILs can also be used as cocatalysts in the acidic hydrolysis of biomass for the production of sugars, thereby yielding a higher conversion at faster rates when compared with the process without ILs.[15] Several studies have reported the solubility of simple carbohydrates and sugar alcohols in ILs.[16−22] However, the recovery data of sugars and ILs from their mixtures are scarce.

The separation of carbohydrates from ILs as well as their further recycling is extremely important with respect to the process economics, particularly in biorefineries using ILs. The recovery of ILs is essential so as to retain the economic feasibility because of the high prices associated with the ILs. For the separation of ILs and polysaccharides, it is comparatively easy to precipitate polysaccharides from the IL solutions. However, the separation of simple carbohydrates such as monosaccharides and sugar alcohols (xylitol or sorbitol) from ILs remains an interesting task. A few attempts with techniques such as chromatographic method, aqueous biphasic systems (ABS), solid-phase extraction, and crystallization/precipitation using antisolvents have been made so far.[23−28]

Feng et al. (2011)[29] studied the separation of glucose from the dimethylphosphate anion-based ILs by applying liquid chromatography. They achieved more than 90% of IL and glucose recovery. Further, Mai et al. (2012)[30] separated glucose and xylose from 1-ethyl-3-methylimidazolium acetate ([Emim][OAc]) using simulated moving bed chromatography coupled with an ion-exclusion column. The amounts of 71.38% of glucose, 99.37% of xylose, and 98.92% of [Emim][OAc] were recovered from the aqueous mixtures. Despite being very effective, these chromatographic methods require very expensive columns which make the operation costs high. Recently, Tonova (2012)[23] reported the use of ABS to separate xanthan and maltose from water/1-hexyl-3-methylimidazolium tetrafluoroborate and 3-methyl-1-octylimidazolium chloride/K2HPO4. To separate maltose from IL, the solution mixture was cooled down to 0 °C. It resulted in the formation of two liquid phases, namely an upper layer containing maltose and a lower phase (IL-rich) for IL recycling and reutilization. ABS requires another additional nonvolatile solvent to completely recover the sugars from the salt/sugar mixtures. This can be achieved by using the antisolvent methods or the “solventing out” phenomenon. The antisolvent methods have already gained importance for drug recovery, separation of nanoparticles, separation of taurine, fractionation of biomass, and other applications.[26,31−33]

Liu et al. (2011) and Hassan et al. (2013) have reported the potential of the antisolvent method to separate glucose from different ILs at different conditions in the mixtures containing antisolvents and ILs. The effects of various experimental variables such as temperature, antisolvents, different ILs, antisolvent to IL molar ratios, and water content were reported.[34,35] In addition, Carneiro et al. (2014) studied the experimental separation of sugar and sugar alcohols from different IL–antisolvent mixtures at room temperatures by varying the antisolvent to IL molar ratio (R) from 5 to 35. At optimal conditions, more than 80% of IL recovery and 83% of sugar removal were observed (R = 20).[26] However, the screening of antisolvents and their interactions with sugars and ILs are not yet reported. Therefore, the application of antisolvents in separating the monomeric sugars from ILs has to be explored in terms of theoretical and experimental aspects.

In the present study, the separation of carbohydrates from ILs using an antisolvent method was studied by applying both computational and experimental aspects. Initially, the conductor-like screening model for real solvents (COSMO-RS) model was employed to screen the suitable antisolvents for the separation of sugars from ILs. Thereafter, the selected antisolvents were used for the precipitation experiments for the mixtures containing four different sugars (glucose, xylose, fructose, and galactose) and three ILs at different experimental conditions. The experimental environments such as different antisolvents, antisolvent to IL molar ratios, and temperatures were also investigated. Furthermore, molecular dynamics (MD) simulations have been performed to understand the basic mechanism of sugar–IL separations using antisolvents.

Results and Discussion

In the present study, the mixture of carbohydrate + IL was separated by using an antisolvent method. The solubility data of carbohydrates in different ILs were taken from previous literature (Figure S1), and the antisolvents were screened by the COSMO-RS model. In our earlier work, we have successfully validated the COSMO-RS model with the available experimental infinite dilution activity coefficients (IDAC) of different solutes in ILs.[12] Nevertheless, a separate benchmark study was also performed for the solubility of monosaccharides in ILs using the COSMO-RS model.[22] Furthermore, the COSMO-RS model was again revalidated with the known experimental data of sugar (fructose and sucrose) separation from [Emim][EtSO4] using different antisolvents.[26] The summary of the benchmarked predicted results is presented in Figure a (fructose–[Emim][EtSO4]–antisolvents) and Figure b (sucrose–[Emim][EtSO4]–antisolvents) as a function of logarithmic IDAC (ln γ). From Figure , it is observed that lower the IDAC value (i.e., ln γ < 1) of an antisolvent in IL, higher is the recovery of IL. On the other hand, a higher IDAC value of sugar in the antisolvent effects a higher removal of sugar from IL. The results from the above validation study gave a good agreement with the experimental recovery data. After the successful benchmarking study, we shall now move to the screening exercise which will be done in two steps, namely (a) screening of antisolvents in IL and (b) screening of sugars in antisolvents. The results of both will then be used to choose the suitable antisolvent.

Figure 1

Correlation between the predicted activity coefficient (COSMO-RS) and experimental recovery of (a) fructose from [Emim][EtSO4] and (b) sucrose from [Emim][EtSO4] using four different antisolvents. The antisolvent experimental data were taken from Carneiro et al. (2014).[26]

Screening of Antisolvents by the COSMO-RS Model

In this section, 35 antisolvents, 11 different carbohydrates, and 27 ILs (composed of 1-ethyl-3-methylimidazolium [Emim]+ cation and 27 different anions) have been screened by the COSMO-RS model. The main aim of the COSMO-RS screening study is to provide a suitable antisolvent(s) for the separation of sugar and sugar alcohols from ILs by computing the IDAC values.

Figures and 3 represent the predicted logarithmic IDAC of 35 antisolvents. The logarithmic activity coefficients of antisolvents in ILs (Figure ) and carbohydrates in antisolvents (Figure ) at infinite dilution are considered together as a quantitative measurement for the solvation power of the antisolvent. The antisolvents were sorted according to their miscibility capacity and arranged in such a way that the antisolvents with high dissolving power (logarithmic negative values of γ, i.e., ln γ ≤ 1) are situated in the left portion of Figures and 3 and the weaker ones (positive values of ln γ, i.e., ln γ ≫ 1) are situated in the right portion of Figures and 3. Lower the activity coefficient values, greater is the tendency for the antisolvent to solvate the IL and carbohydrate. For the separation process using an antisolvent method, the antisolvent should be miscible with IL (ln γ ≤ 1), but immiscible with carbohydrate (ln γ ≫ 1). As can be seen from Figure , the antisolvents such as acetic acid, ethylene glycol, diethylene glycol, glycerol, chloroform, formamide, dichloromethane (DCM), and 1,2-dichloroethane (DCE) are predicted to be miscible with ILs more efficiently, whereas in the case of the carbohydrate–antisolvent system (Figure ), the antisolvents, namely acetonitrile, nitromethane, DCM, and DCE are predicted to be immiscible with the carbohydrates. Therefore, according to the precipitation concept, the chosen antisolvent should be miscible with IL and immiscible with the carbohydrate. Therefore, the following antisolvents, namely DCM and DCE have been preferred for both experiments and MD simulations.

Figure 2

Graphical representation of the IDACs of 35 antisolvents (organic solvents) in 27 different ILs at 298.15 K by COSMO-RS.

Figure 3

Graphical representation of the IDACs of 11 carbohydrates in 35 different antisolvents (organic solvents) at 298.15 K by COSMO-RS.

Moreover, from a close look at Figures and 3, nitromethane also seemed to be a potential antisolvent as similar to DCE; however, DCE was preferred owing to its lower boiling point and economic aspects. The boiling points of DCM, DCE, and nitromethane and other antisolvents are listed in Table S1. On the other hand, nitromethane gave less IDAC values of carbohydrates as compared to DCE (e.g., ln γ of glucose in DCE is 12.37, whereas in the case nitromethane, it is 6.98). Hence, we have excluded nitromethane from the separation experiments, although nitromethane could be used as a potential antisolvent in the carbohydrate separation process.

Experimental Separation of Sugars from ILs

After the successful screening of antisolvents using the COSMO-RS model, the precipitation experiments were performed for sugars such as glucose, fructose, xylose, and galactose from ILs ([Emim][EtSO4], [Emim][SCN], and [Emim][MeSO3]) using DCM and DCE as the antisolvents. As can be seen from Figure , in terms of IDAC values, lower the IDAC values from unity, that is, the blue shaded region, better is the IL. Thus, [Emim][OAc] is the best solvent. It should be noted that other than the solubility of IL and the antisolvent, the recovery of sugar is also a critical issue. This is primarily done by boiling the solution such that the antisolvent, because of its lower boiling point, separates out easily from the IL–antisolvent mixture. Therefore, for an overall selection of an IL and antisolvent pair, one should have negative values of ln γ (Figure ) and highly positive values of ln γ (Figure ) for the sugars in the antisolvent. The IL [Emim][OAc] is known to be a powerful solvent in the dissolution of biomass and its derived components, namely cellulose, hemicellulose, disaccharides, and monosaccharides.[7,34,36] When we attempted the separation of glucose from [Emim][acetate] using DCM as an antisolvent, the glucose removal (55%) and IL recovery (40%) percentages were found to be lower because of the strong interactions between [Emim][acetate] and glucose. In the separation of sugars from IL, the interaction between sugar and IL also plays a significant role (lower IDAC values). The solubility of the investigated sugars in [Emim][acetate] is higher than that in other ILs (Figure S2). Thus, the recovery of sugars and the [Emim][acetate] IL is lower. Figure S2 represents the predicted logarithmic IDAC of 11 sugars in 27 different ILs. Another set of ILs, namely, amino acid-based ILs, have shown higher interactions or solubility with sugars (Figure S2), which could lead to the solubilization of the biomass as well. Therefore, the amino acid-based ILs are selected for the screening study. It should be noted that the experimental solubility data for the investigated sugars are absent in the amino acid-based ILs. Further, the synthesis of amino acid-based ILs was economically not easy, which is the reason they were omitted. This is the very reason that the next set of ILs ([Emim][EtSO4], [Emim][SCN], and [Emim][MeSO3]) was used for the separation process.

It is also a fact that the selected ILs have already proven to be the best in the dissolution of sugars.[20,22,34,37] For the COSMO-RS screening study, the sugars, namely monosaccharides, disaccharides, and sugar alcohols are selected for separation from ILs. In the sugar separation process, the monosaccharides and sugar alcohols are difficult to recover as compared to the disaccharides, oligosaccharides, and polysaccharides, owing to their higher solubility in ILs. From the results of Carneiro et al. (2014), the removal percentage of the disaccharide, that is, sucrose is ≥93–99%, whereas the removal percentages of glucose and fructose are 8–75.3% and 5.3–90%, respectively.[26] Therefore, we have given more attention for the recovery of monomeric sugars from ILs. The equilibrium solubility data of sugars in the respective ILs are taken from previous literature (Figure S1), except for the glucose–[Emim][MeSO3] system. The solubility of glucose in [Emim][MeSO3] was measured at 298.15 K (Table ). Table reports the percentage of IL recovered (% ILR) and percentage of carbohydrate removal (% CR) with different sugar–IL–antisolvent mixtures at 298.15 K and with different antisolvent to IL molar ratios (Rantisolvent). From Table , it can be observed that the solubility of fructose was higher than that of other sugars. This solubility difference was attributed to the difference in the phase transition properties (such as melting temperature and heat of fusion values) of sugars. Lower the melting temperature and heat of fusion value, higher will be the solubility of the solute in the respective solvents.[20,22] The melting properties of the investigated sugars are reported in the Supporting Information, Table S2.

Table 1

% ILR and % CR with Different Sugar–IL–Antisolvents at 298.15 K and Different Antisolvent to IL Molar Ratios (R)a

				IL		carbohydrate
IL–antisolvent	sugar	R	xsugarb	recovery (%)	σc	removal (%)	σc
[Emim][EtSO4]/DCM	glucose	5	0.161	76.45	2.01	96.86	1.14
		10	0.161	80.75	1.26	98.15	0.95
		15	0.161	88.46	1.47	98.46	0.74
		20	0.161	92.98	1.70	98.60	0.66
	fructose	20	0.348	86.02	1.42	90.94	1.90
	xylose	20	0.288	89.64	1.31	89.49	1.83
	galactose	20	0.088	98.12	0.87	90.44	1.19
[Emim][EtSO4]/DCE	glucose	20	0.161	81.79	1.18	99.55	0.40
	fructose	20	0.348	79.50	2.17	98.55	1.21
	xylose	20	0.288	77.78	2.74	98.35	0.91
	galactose	20	0.088	79.74	1.52	99.77	0.14
[Emim][SCN]/DCM	glucose	10	0.075	81.92	1.80	97.30	0.83
		20	0.075	94.30	1.62	99.23	0.28
	fructose	20	0.332	85.95	1.60	95.42	1.34
	xylose	20	0.096	87.07	1.72	92.25	2.05
	galactose	20	0.036	95.49	1.12	97.78	1.52
[Emim][MeSO3]/DCM	glucose	20	0.183	87.47	1.80	96.27	1.41

Standard uncertainty for temperature and pressure are u(T) = 0.1 K and u(p) = 1 kPa; the standard uncertainty for % ILR and % CR are U(ILR) = 1.58% and U(CR) = 0.82% at 95% confidence level; the standard uncertainty is calculated by using the following equation, .

Mole fraction of sugar in IL at 298.15 K.

Standard deviation.

Effect of Antisolvent to IL Molar Ratio in Sugar Separation

The effect of the antisolvent to IL molar ratio in sugar separation was studied by varying from 5 to 20, which is shown in Table . From Table (for glucose–[Emim][EtSO4]/DCM and glucose–[Emim][SCN]/DCM), the recovery and removal percentage of ILs and carbohydrates were found to increase with the increase in the antisolvent to IL molar ratio. In the recovery of IL, a significant difference was observed with the increase in the antisolvent molar ratio. At higher antisolvent molar ratio (R = 20), the recovery of [Emim][EtSO4] was about 93% for the glucose–[Emim][EtSO4]/DCM system. In the case of carbohydrate removal percentage, the difference in % CR was minor, that is, 98–99% from R = 10 to 20. As expected, higher the molar ratio of the antisolvent to IL, lower will be the solubility of the carbohydrate in the mixtures.

According to the results of Carneiro et al. (2014), a further increase in the antisolvent molar ratio (R > 20) did not show a significant effect in the recovery and removal percentages of both IL and the carbohydrate.[26] It was also reported that for the case of ethanol as an antisolvent, a lower molar ratio is preferable, as the solubility of carbohydrate in pure ethanol is higher as compared with that in DCM, acetonitrile, and DCE. Therefore, in the solubility aspects, R should not be too high as it may lead to a lower carbohydrate removal. For this purpose, ethanol can only be used as an antisolvent at lower molar ratios, that is, R < 3, for efficient carbohydrate removal.[26] Therefore, the optimum value of R should be a balance between the separation performance and antisolvent loading. Thus, R = 20 was considered as the optimal molar ratio for the separation of carbohydrates from ILs.

For further investigations, three important phases are involved in the experimental design for the separation of sugars from ILs. In the first phase, the separation experiments are performed with ILs [Emim][EtSO4] and [Emim][SCN] in the presence of the DCM antisolvent with four different sugars. This step provides the effect of sugar removal with different sugars in the separation process. In the second phase, the IL [Emim][MeSO3] was tested with glucose in the DCM antisolvent to understand the effect of the third IL. Finally, to understand the effect of different antisolvents, two different antisolvents, namely, DCM and DCE, were tested for four different sugars using [Emim][EtSO4] at R = 20. Hence, to avoid all the combinations, the representative combinations were chosen.

Effect of Different Sugars

The effects of different sugars such as glucose, xylose, fructose, and galactose were studied in [Emim][EtSO4] and [Emim][SCN] with DCM as an antisolvent. As can be seen from Table , the recovery percentage of ILs in the galactose-based system was higher than that of the other sugars. This is mainly due to the weaker interactions and lower solubility of galactose in ILs. Galactose exhibits higher melting properties than other sugars which lead to lower solubility. It should be noted that the lower solubility of a solute in the solvent causes weaker interactions. On the contrary, the recovery percentage of ILs in fructose-based systems was lower than that of other sugars. This is due to the lower melting properties of fructose (Table S2) which lead to higher solubility. Therefore, the higher solubility of sugar in IL yielded a less percentage of IL recovery (see Table ).

Further, the recovery percentage of [Emim][EtSO4] and [Emim][SCN] ILs with the mixture of different sugars was also assessed. Except for the glucose-based system, the recovery percentage of [Emim][EtSO4] IL was found to be higher than that of [Emim][SCN] at R = 20 (Table ). The solubility of carbohydrates in [Emim][SCN] was lower than that in [Emim][EtSO4] because of the weaker interactions between the carbohydrates and [Emim][SCN]. Despite this fact, the recovery percentage of [Emim][EtSO4] in DCM was higher when compared to that of [Emim][SCN] because of the disturbance in the hydrogen-bonding network between the carbohydrates and [Emim][EtSO4] (see Table S3). As can be seen from Table S3, with the addition of an antisolvent to the sugar–IL system, the H bonding and interaction energy (IE) values were found to be significantly decreased. Furthermore, the activity coefficient of [Emim][EtSO4] in DCM is also lower than that of [Emim][SCN] (Table ). This leads to a higher miscibility of [Emim][EtSO4] in DCM, resulting in a higher recovery percentage.

Table 2

COSMO-RS-Predicted Logarithmic IDACs (ln γ) of Sugar and IL Molecules in Different Antisolvents

sl. no.	IL–sugar	DCM	DCE
1	[Emim][MeSO3]	–3.58	–0.32
2	[Emim][EtSO4]	–2.42	0.05
3	[Emim][SCN]	–1.42	0.46
4	glucose	8.06	12.37
5	fructose	7.50	11.46
6	xylose	7.32	10.76
7	galactose	7.16	11.94

On the other hand, the removal percentage of glucose was higher than that of other sugars, even though glucose exhibits lower melting properties than galactose. Here, the COSMO-RS results also suggested that the ln γ values of glucose in the DCM and DCE antisolvents are higher than those of other sugars (i.e., as per Table , higher the IDAC value, lower are the interactions). Further observations from Table showed that the removal of xylose was lower than that of fructose, although fructose had higher solubility in ILs. This is due to the fact that two parameters actually effect the removal of carbohydrates from ILs, that is, the sugar–IL and sugar–antisolvent interactions. If the sugar–IL interactions are weaker and the sugar–antisolvent interactions are stronger, it leads to a lower removal of carbohydrate. Thus, lower the IE value, lesser is the removal of sugar. Xylose has weaker interactions with ILs and stronger interactions with the antisolvents (Table ) as compared to fructose. Therefore, xylose achieved a lower removal percentage.

Effect of Different ILs

ILs such as [Emim][EtSO4], [Emim][SCN], and [Emim][MeSO3] were used to separate glucose from a mixture in the presence of DCM as an antisolvent. The ILs [Emim][MeSO3] and [Emim][EtSO4] gave higher solubility (xsugar) of glucose at 298.15 K in the studied systems, indicating that these ILs have a strong dissolving capacity for glucose. However, [Emim][SCN] has a higher recovery percentage as compared to the other two ILs when they were mixed with glucose in the presence of DCM. On the other hand, the COSMO-RS results suggest that [Emim][MeSO3] and [Emim][EtSO4] have lower activity coefficient values in DCM as compared to [Emim][SCN] (Table ), resulting in a lower recovery percentage of IL. This is due to the fact that [Emim][MeSO3] and [Emim][EtSO4] have strong interactions with glucose than DCM. Therefore, the recovery percentages of [Emim][MeSO3] and [Emim][EtSO4] were lower than that of [Emim][SCN], whereas in the case of glucose removal percentage, it does not have a general trend. A lower removal percentage of glucose was obtained in [Emim][MeSO3] because of the higher solubility of glucose. However, the percentage of glucose removal in all the three different ILs gave a similar value (∼97–99%).

Effect of Different Antisolvents

Two different antisolvents, namely DCM and DCE were tested for the experimental separation of four different sugars from [Emim][EtSO4] at R = 20. From Table , it was observed that there is a remarkable difference on the ability of the antisolvents in the recovery of [Emim][EtSO4] from these mixtures. As compared with the recovery percentage of [Emim][EtSO4] in both antisolvents, a significant difference was not observed in the removal of carbohydrates. The results from Table indicate that the recovery percentage of [Emim][EtSO4] was higher in DCM as compared with that in the DCE antisolvent. This may be due to the higher interaction between [Emim][EtSO4] and DCM. On the other hand, the removal percentage of carbohydrate was higher in DCE when compared to that in DCM. This ascription was attained because of the weaker interactions between the carbohydrates and DCE, which cause a lower solubility of carbohydrates in the DCE antisolvent.

It should be noted that the investigated antisolvents DCM and DCE are not environmentally benign as compared to ethanol. However, the separation of these antisolvents from the IL phase can be easily performed by using distillation at a lower pressure. However, when ethanol is used as an antisolvent in the separation of carbohydrates from ILs, the removal percentage of carbohydrates is lower because of the higher interactions between ethanol and carbohydrates.[26] Recently, Hassan et al. (2013) performed the glucose dissolution in different ILs and further extracted glucose from ILs using the antisolvent method with different antisolvents.[35] Glucose (80%) was extracted from [Emim][SCN] in the presence of ethanol as an antisolvent, whereas in the case of acetonitrile, 90% of glucose was extracted from [Emim][SCN] at R = 20. In our case, 97.3 and 99.23% of glucose were recovered from [Emim][SCN] with the DCM antisolvent at R = 10 and 20, respectively, whereas for ethanol, at R = 10, 37.15% of glucose was recovered.

In another study, Carneiro et al. (2014) investigated the separation of carbohydrates from ILs using an antisolvent method.[26] They reported a recovery of 5.3% of fructose from [Emim][EtSO4] and 8.1% of glucose from [Emim][TFA] when using ethanol as an antisolvent (at R = 20). Moreover, when acetone and acetonitrile were used as antisolvents, 67 and 76.7% of fructose were recovered from [Emim][EtSO4], respectively. On the other hand, when DCM and DCE were used as antisolvents, 90% of fructose and 99% of glucose were extracted from [Emim][EtSO4], respectively. Further, our COSMO-RS predictions (Figures and 4) also prove the fact that the lower ln γ value of sugars in ethanol results in a higher solubility of sugar in ethanol. Therefore, in the present study, DCM and DCE solvents have been used to separate carbohydrates from ILs.

Figure 4

Effect of temperature on the separation of glucose and [Emim][SCN] in the presence of DCM at R = 20.

Effect of Temperature on Sugar Separation

Furthermore, the effect of temperature in the separation of glucose from the [Emim][SCN]/DCM (at R = 20) mixture was studied at different temperatures (298.15, 303.15, and 313.15 K). As can be seen from Figure , the recovery percentage of [Emim][SCN] and the removal percentage of glucose were found to decrease with increasing temperature. An increase of temperature will promote the molecular thermal motion of the entire system and the volatility of the nonvolatile solute, resulting in an increase in the solubility of glucose in the [Emim][SCN]/DCM mixtures. Hence, low temperature is favorable for the separation of carbohydrates from ILs. A similar observation was also reported by Liu et al. (2011)[34] during their study on the extraction of glucose from different ILs using ethanol as an antisolvent.

Results of MD Simulations

Apart from the experimental evidence, the MD simulations were also carried out to understand the structural properties of carbohydrates with ILs and antisolvents via IEs, average hydrogen bonding (hydrogen bonds (HBs)), and coordination numbers (CN). For the MD simulation system, the carbohydrates, namely, glucose, xylose, and fructose are considered. Figure S3 shows the optimized geometries of glucose, xylose, fructose, [Emim]+, anions, and two antisolvents (DCM and DCE).

Nonbonded IE

The MD-simulated nonbonded IE is the sum of the electrostatic and van der Waals energies. Table reports the IEs of glucose with three different IL–antisolvent mixtures at 298.15 K and R = 20. The IE results of glucose are compared with the other two carbohydrate systems, namely, xylose–[Emim][EtSO4]/DCM and fructose–[Emim][EtSO4]/DCM.

Table 3

Nonbonded IEs (kJ mol–1) for Different IL–Sugar–Antisolvent Systems Obtained from MD Simulations at T = 298.15 K and R = 20a

energy type	[Emim]+–sugar	[anion]−–sugar	[Emim]+–antisolvent	[anion]−–Antisolvent	sugar–antisolvent
Glucose–[Emim][EtSO4]–DCM
electrostatic (Eelec)b	–8.4	–41.07	36.23	–94.18	–3.98
van der Waals (EvdW)c	–10.39	–2.02	–33.13	–32.52	–27.03
total energy (Etotal)d	–18.79	–43.08	3.09	–126.69	–31.02
Glucose–[Emim][MeSO3]–DCM
electrostatic (Eelec)	0.35	–59.06	44.17	–100.1	–4.73
van der Waals (EvdW)	–10.88	0.02	–33.92	–21.32	–27.94
total energy (Etotal)	–10.53	–59.04	10.25	–121.42	–32.67
Glucose–[Emim][SCN]–DCM
electrostatic (Eelec)	–2.4	–21.8	44.59	–96.62	1.61
van der Waals (EvdW)	–6.11	–0.08	–33.55	–12.36	–20.97
total energy (Etotal)	–8.51	–21.88	11.04	–108.98	–19.36
Glucose–[Emim][EtSO4]–DCE
electrostatic (Eelec)	–6.83	–43.86	7.57	–51.93	–2.34
van der Waals (EvdW)	–11.1	–2.75	–29.12	–31.7	–25.45
total energy (Etotal)	–17.93	–46.61	–21.54	–83.63	–27.8
Fructose–[Emim][EtSO4]–DCM
electrostatic (Eelec)	–17.99	–76.33	36.63	–91.1	–3.16
van der Waals (EvdW)	–20.58	–5.44	–31.18	–31.65	–25.9
total energy (Etotal)	–38.58	–81.77	5.44	–122.75	–29.06
Xylose–[Emim][EtSO4]–DCM
electrostatic (Eelec)	–8.10	–54.77	31.75	–94.09	–5.22
van der Waals (EvdW)	–12.31	–3.87	–34.79	–33.55	–23.70
total energy (Etotal)	–20.41	–58.64	–3.04	–127.64	–28.92

Antisolvent to IL molar ratio.

Electrostatic IE (Eelec) of the system in kJ mol–1.

van der Waals IE (Evdw) of the system in kJ mol–1.

Sum of Eelec and Evdw in kJ mol–1.

Interactions of Different IL–Antisolvents with Glucose

The nonbonded IEs of anions, [Emim]+, and antisolvents with glucose were computed, as shown in Table 3. As can be seen from Table , the anions possessed a higher IE with glucose and DCM when compared to the [Emim]+ cation. This higher IE is obtained because of the formation of a stronger intermolecular hydrogen bonding between the glucose–anion (1.75 Å) and anion–DCM (≤2.15 Å) pairs (Figure S4). For the glucose–[Emim]+, glucose–DCM, and [Emim]+–DCM systems, the intermolecular distances were found within the range of 2.20–3.05 Å (Figure S4), indicating a weak HB formation. Therefore, it reveals that the anionic part of the IL plays a predominating role as compared to the cation. The IE between glucose and [Emim][MeSO3] was found to be higher than that between glucose and [Emim][EtSO4] and glucose and [Emim][SCN]. From the experimental data, the solubility of glucose (x) was reported to be higher in [Emim][MeSO3] than in the other two ILs (Table ). A similar result was reflected in their IEs, that is, higher the IE value, higher is the solubility. However, the removal percentage of glucose was lower in [Emim][MeSO3] (96%) because of the stronger IE between glucose and [Emim][MeSO3]. Further, the IE between glucose and DCM was higher in the glucose–[Emim][MeSO3]/DCM system; as a result, the recovery of glucose was lower. Therefore, the IE between glucose and IL and that between glucose and DCM is crucial in the removal of glucose from IL. On the other hand, the removal percentage of glucose was higher in [Emim][SCN] (99.23%) because of lower solubility and weaker interactions between glucose and [Emim][SCN] and glucose and DCM (Figure ).

Figure 5

Effect of different ILs on the recovery of glucose from ILs in sugar–IL separations by using the DCM antisolvent at 298.15 K and R = 20. (a) MD-simulated IEs and (b) experimental % ILR and % CR.

Furthermore, on comparing the IL recoveries, the recovery percentage of [Emim][SCN] was higher than the other two ILs. It was also noted that the recovery percentage of [Emim][SCN] was found to be higher than that of [Emim][EtSO4] at R = 10 and 20. According to the definition, the antisolvent should be miscible or marginally miscible with IL. However, the IE between [Emim][SCN] and DCM was weaker as compared to that between [Emim][EtSO4] and DCM and [Emim][MeSO3] and DCM (Figure ). The recovery of [Emim][EtSO4] and [Emim][MeSO3] was lower because of the stronger IE between glucose and [Emim][MeSO3] and glucose and [Emim][EtSO4] (see Figure ). Hence, the IE between glucose and IL and IL and antisolvent has a dominating influence on the recovery of IL. It was also worthwhile to note that higher the difference in IE between IL and sugar and IL and antisolvent, higher is the recovery percentage of IL. The magnitude of difference in IE between glucose–[Emim][SCN] and [Emim][SCN]–DCM (−67.55 kJ mol–1) is higher than [Emim][EtSO4] and [Emim][MeSO3] (−61.73 and −41.60 kJ mol–1). Therefore, the recovery of [Emim][SCN] is higher than that of [Emim][EtSO4].

In addition, the effects of the antisolvent (DCM and DCE) interactions on glucose–[Emim][EtSO4] were also investigated and are shown in Table and Figure . From Figure , it is seen that the recovery of [Emim][EtSO4] in DCM was higher as compared with that in DCE. It is due to the stronger interactions between [Emim][EtSO4] and the DCM antisolvent. On the other hand, the higher removal percentage of glucose was found in the DCE antisolvent. It should be noted that DCE possessed a lower IE (−27.80 kJ mol–1) with glucose when compared to DCM (−31.02 kJ mol–1). Even though the difference between the IEs is −3.22 kJ mol–1, it highly alters the glucose removal percentage from [Emim][EtSO4].

From Table , it is noted that higher the difference in IEs between IL and sugar and that between IL and the antisolvent, higher is the recovery percentage of IL. However, a contrary trend is obtained for the sugar removal percentage. Here, as the energy difference between IL and sugar and that between sugar and antisolvent decreases, the removal percentage of sugar is found to increase for [Emim][SCN], [Emim][EtSO4], and [Emim][MeSO3] with the DCM antisolvent. For the systems with [Emim][EtSO4]–DCM and [Emim][EtSO4]–DCE, the energy difference between IL and glucose and that between glucose and antisolvent increases with an increasing removal percentage of glucose. This attribution may be due to the difference in the physiochemical properties of the antisolvents.

IEs of Different Sugars with [Emim][EtSO4]/DCM

The IEs of glucose, xylose, and fructose with [Emim][EtSO4]/DCM are reported in Table and shown in Figure . The IL recovery and sugar removal percentage were higher in glucose–[Emim][EtSO4]/DCM as compared with the xylose- and fructose-based systems (Figure ). The IE between fructose and IL (−120.35 kJ mol–1) was much stronger than that between xylose and IL (−79.05 kJ mol–1) and glucose and IL (−61.87 kJ mol–1), which leads to a higher removal percentage of glucose (98.60%). Further, it should be noted that higher the IE between the molecules, stronger is the formation of HBs between them. This makes it difficult to disturb the HBs formed between them. In addition, the IE difference between glucose and [Emim][EtSO4] and that between glucose and DCM is lower than that between fructose and xylose, whereas in the case of IL recovery percentage, the IE difference between the sugar and [Emim][EtSO4] and that between [Emim][EtSO4] and DCM is lower in the fructose-based system (3.04 kJ mol–1) than in the xylose (−51.63 kJ mol–1)- and glucose (−61.73 kJ mol–1)-based systems (see Figure ). Consequently, the recovery percentage of IL was less in the fructose-based system because of their lower interaction between IL and DCM and stronger interaction between fructose and IL.

Figure 6

Effect of different ILs on the recovery of glucose from ILs in IL–sugar separations by using the antisolvent method at 298.15 K and R = 20. (a) MD-simulated IEs and (b) experimental % ILR and % CR.

Hydrogen Bonding and CN

Table reports the average number of HBs per glucose and IL molecule established between sugar–IL–antisolvent. The criteria for the calculation of HBs are as follows: the distance between the acceptor and donor molecules is kept at 3.2 Å, and the cutoff angle for acceptor–H–donor is 120°. For reliability, we have used the same distance and angle for all simulated systems. Table also reports the CN and the total nonbonded IE between the simulated molecules. From Table , it is observed that the average number of HBs formed was higher with increasing CN. However, this trend excludes the [Emim]+–sugar and [Emim]+–antisolvent systems. It should be noted that HB and CN are two different descriptors. HBs are a type of dipole–dipole interactions formed between the proton of a donor group X–H (X: electronegative atom) and one or more other electronegative atoms (Y) having a pair of nonbonded electrons.[38,39] CN is computed from the integration of radial distribution function (RDF) peaks and provides an estimate of the number of molecules in the surrounding environment of the reference molecule. In these calculations, along with the height and width of the RDF peaks, the density of the system is also accounted.[40,41] Therefore, HB and CN in Table do not provide a direct correlation.

Table 4

Average HBs, CN, and Nonbonded IEs (Etotal, kJ mol–1) for Different IL–Sugar–Antisolvent Systems Obtained from MD Simulations at T = 298.15 K and R = 20a

type of measurement	[Emim]+–sugar	[anion]−–Sugar	[Emim]+–antisolvent	[anion]−–antisolvent	sugar–antisolvent
Glucose–[Emim][EtSO4]–DCM
HBb	0.67	0.91	0.13	1.13	0.61
CNc	1.86 (3.65)	1.22 (2.35)	4.62 (4.25)	5.20 (3.65)	3.24 (3.55)
Etotald	–18.79	–43.08	3.09	–126.69	–31.02
Glucose–[Emim][MeSO3]–DCM
HB	0.66	1.06	0.13	1.04	0.71
CN	1.81 (3.65)	1.38 (2.35)	4.23 (4.15)	4.33 (3.45)	3.59 (3.55)
Etotal	–10.53	–59.04	10.25	–121.42	–32.67
Glucose–[Emim][SCN]–DCM
HB	0.27	0.29	0.12	0.21	0.35
CN	1.07 (3.95)	0.94 (2.65)	4.65 (4.25)	2.05 (3.75)	2.16 (3.55)
Etotal	–8.51	–21.88	11.04	–108.98	–19.36
Glucose–[Emim][EtSO4]–DCE
HB	0.65	0.92	0.10	0.77	0.48
CN	1.82 (3.65)	1.21 (2.35)	2.86 (4.15)	2.88 (3.75)	2.26 (3.65)
Etotal	–17.93	–46.61	–21.54	–83.63	–27.80
Fructose–[Emim][EtSO4]–DCM
HB	1.06	1.51	0.11	0.93	0.51
CN	3.66 (3.65)	2.26 (2.35)	4.33 (4.25)	4.70 (3.65)	3.14 (3.55)
Etotal	–38.58	–81.77	5.44	–122.75	–29.06
Xylose–[Emim][EtSO4]–DCM
HB	0.78	1.00	0.13	1.13	0.50
CN	2.46 (3.65)	1.73 (2.35)	4.98 (4.25)	5.37 (3.65)	2.53 (3.55)
Etotal	–20.41	–58.64	–3.04	–127.64	–28.92

Antisolvent to IL molar ratio.

HB cutoff distance––3.2 Å and angle cutoff––120°.

The values in parentheses denote the maximum distance of the RDF first peak (r, Å) for the first solvation shell.

Sum of Eelec and Evdw.

From Table , it is observed that the anion of IL forms a higher number of HBs with the sugar molecule than either the cation of IL or the antisolvent. Contrary to HBs, the CN between the anion and antisolvent is higher than that between the anion and sugar molecules. This is due to the fact that the cutoff distance between the acceptor and donor molecules is only 3.2 Å. On the other hand, the first solvation peak for anion–sugar was obtained at around 2.35–2.65 Å, whereas in the case of anion–antisolvent, the solvation peak was attained at 3.45–3.75 Å (see Table ). Furthermore, the anions possess a higher IE with the sugar molecules as a result of the stronger HBs formed between the anion and sugar molecules. It is interesting to note that in the DCE-based system, the IE between the cation and antisolvent is higher than that in the DCM-based system. Nevertheless, DCE-based systems had lower H-bonding and CN values than the DCM-based systems (Table ). The trend is due to the lower electrostatic interactions of [Emim]+–DCE. The IE, average HBs, and CNs for glucose–[Emim][EtSO4]/DCM was lower than those for xylose–[Emim][EtSO4]/DCM and fructose–[Emim][EtSO4]/DCM, which revealed a higher recovery of IL and glucose. Furthermore, a close look at Table S3 reveals that the IE between IL and sugar is much stronger in a pure IL–sugar system. However, with the addition of antisolvent to the sugar–IL system, the H-bonding and IE values were lowered. In summary, the antisolvent was indeed efficient in disturbing the hydrogen-bonding network between the carbohydrates and IL.

Correlation between Theoretical and Experimental Separation Data

Figure depicts the relationship between the predicted activity coefficient (COSMO-RS), MD-simulated IE, and experimental fructose removal from [Emim][EtSO4] in the presence of different antisolvents. The experimental separation data for [Emim][EtSO4]/ethanol and [Emim][EtSO4]/acetonitrile were taken from the literature.[26] From Figure , it is observed that an increase in IE between fructose and antisolvent lowers the removal percentage of fructose. The higher IE between the molecules leads to a higher solubility of solute in the solvent. The COSMO-RS calculations suggest that lower the ln γ value, lower is the removal of fructose (Figure ). It was reported that the solubility of monomeric sugars in ethanol was higher as compared to that in DCM and acetonitrile.[34] Therefore, the removal percentage of fructose was less in ethanol as compared to that in the other two antisolvents. Overall, this agrees well with the sugar removal, but fails in few cases for IL recovery. In Table S4, DCM shows a lower ln γ value with [Emim][EtSO4] than with ethanol and acetonitrile, as DCM gave less recovery of IL. This is mainly attributed to the fact that although COSMO-RS predictions are performed on an IL–antisolvent mixture only, it neglects the presence of sugar. Hence, the predicted activity coefficient follows a separate trend. On the contrary, the MD simulations are conducted with the entire mixture, namely, the sugar–IL/antisolvent, and the IEs showed a good correlation with the experimental recovery (Table S4). Further, to elucidate the strength of the HBs formed between fructose and antisolvents, separate quantum chemical calculations have been performed and are presented in the Supporting Information.

Figure 7

Correlation among the predicted activity coefficient (COSMO-RS), MD-simulated IE, and experimental fructose recovery from [Emim][EtSO4] using the different antisolvents. The antisolvent experimental data for ethanol and acetonitrile were taken from Carneiro et al. (2014).[26]

Conclusions

In the current study, 35 antisolvents were screened by the COSMO-RS model for the separation of sugars and ILs by means of the precipitation process. From the COSMO-RS screening study, DCM and DCE were found to be the efficient antisolvents for the separation process. Thereafter, the precipitation experiments were conducted with the selected antisolvents for the separation of four different sugars from ILs at different process variables. From the experimental results, 90–99% of carbohydrate removal and 80–98% of IL recovery were achieved with the DCM antisolvent at an optimal antisolvent molar ratio (R = 20). At higher temperatures, the recovery percentage of [Emim][SCN] and the removal of glucose were found to decrease. Hence, low temperature is favorable for the separation of carbohydrates from ILs. Furthermore, MD simulations were performed to elucidate the structural properties of carbohydrates with the IL–antisolvent mixtures. The IE between glucose and IL was found to be weaker than that between xylose and IL and fructose and IL. This led to a higher removal percentage of glucose than xylose and fructose. The higher the difference in IE between IL and sugar and that between IL and antisolvent, higher is the recovery percentage of IL. A higher removal percentage of glucose was also found in the DCE antisolvent. DCE possessed a lower IE (−27.80 kJ mol–1) with glucose when compared to DCM (−31.02 kJ mol–1). The IEs between carbohydrate and IL and that between carbohydrate and antisolvent hence have a dominating influence on the removal of carbohydrate from IL. Overall, a higher IE (MD) and lower activity coefficient (COSMO-RS) yield a lower removal of carbohydrates (experimental). In summary, these results prove that the usage of the DCM antisolvent for the separation of carbohydrates from ILs is feasible.

Computational Details

COSMO-RS Model

COSMO-RS is a quantum chemical-based solvation model which can be used for screening of a large number of solvents or potential cation and anion combinations in ILs. The model can be used to predict the activity coefficient and chemical potential of any solute in the mixture. It requires the chemical structure of the component as the only information.[42,43] Initially, the structures of sugars, antisolvents (organic solvents), anions, and cations were drawn with the help of Avogadro freeware software.[44] The molecular geometries of all molecules were fully optimized by B3LYP/6-31G* via Gaussian09 package.[37,45] After molecular geometry optimization, the next step is to generate the COSMO file using the SCRF = COSMO-RS keyword with the final optimized structure. The detailed methodology of the COSMO-RS calculations is described in our previous work, hence not discussed here.[12,22,37,46]

MD Simulations

The microscopic interactions between the sugars (glucose, xylose, and fructose) and IL–antisolvent mixtures were studied by employing the MD simulations. The partial charges of all the studied molecules were attained by the restrained electrostatic potential charge derivation method.[47] For all the molecules, the generalized AMBER force-field (GAFF) parameters were employed.[48] The GAFF parameters were developed by Antechamber.[49]

All the MD simulations were performed with the NAMD ver2.10[50] at constant temperature (298.15 K) and atmospheric pressure using a Langevin thermostat and a Nose–Hoover Langevin barostat.[51,52] The SHAKE algorithm was implemented to constrain the lengths of all the covalent bonds to hydrogen.[53] The Particle Mesh Ewald method was applied to account for the long-range electrostatic interactions at the cutoff distance of 12 Å.[54] The initial configuration of the sugar–IL–antisolvents is prepared by PACKMOL in a cubic box with a random distribution of molecules.[55] Initially, the simulation system was minimized for 1 ns and then gradually heated to its desired temperature (298.15 K) for 0.5 ns. At the desired temperature, the system was then equilibrated for 10 ns under an isothermal–isobaric ensemble (NPT) to get the system converge to its experimental condition. Consequently, the production run lasted for 40 ns with a canonical (NVT) ensemble. At every 5 ps, the production run data were saved for structural analysis from the simulated trajectories using VMD software.[56] In the mixture of sugar–IL–antisolvent, the IE and HBs between IL and sugar and those between IL and antisolvent are calculated per mole of IL, and for the sugar–antisolvent system, the IE and HBs are calculated per mole of sugar.

Materials and Methods

Materials

Carbohydrates such as glucose, fructose, xylose, and galactose were purchased from Sigma-Aldrich. All the carbohydrates have purities greater than 99% as per the supplier specification. The ILs 1-ethyl-3-methylimidazolium thiocyanate [Emim][SCN] (≥95%), 1-ethyl-3-methylimidazolium ethylsulfate [Emim][EtSO4] (≥95%), and 1-ethyl-3-methylimidazolium methanesulfonate [Emim][MeSO3] (≥95%) were supplied from Sigma-Aldrich, Germany. These chemicals were used without further purification. The antisolvents such as DCM (≥99.5%) and DCE (≥99.5%) were purchased from Merck, India.

Separation of Sugars from ILs

The experimental solubility data of sugars, namely, glucose, xylose, fructose, and galactose in [Emim][EtSO4] and [Emim][SCN] were taken from the available literature where the experiments were conducted at various temperatures in different ILs. The results are presented in Figures S5 and S6 (Supporting Information).[20,22] The saturated solubility data of sugars in ILs ([Emim][EtSO4] and [Emim][SCN]) at 298.15 K are reported in Figure S1 and herein used for the separation process. The mixture containing sugar + IL was used without any further pretreatment or drying. For the separation process, the specified amount of sugar + IL (2 g) mixture with a known composition was mixed with a desired amount of antisolvent in a 30 mL flask. The flask was sealed and the mixture was magnetically stirred for 8–10 h to make sure that all the carbohydrate was precipitated. The precipitation experiments were conducted at different sugar–IL–antisolvents, different antisolvent to IL molar ratios (R = 5, 10, 15, 20), and temperatures. All the experiments were performed in duplicate.

After precipitation, the carbohydrates were collected by vacuum filtration and dried at 45 °C for 48 h. Subsequently, the vacuum-filtrated liquid portion was collected, and the solution was submitted to evaporation of antisolvent by heating to its boiling point and continued till the constant weight of IL was achieved. The recovered IL fraction was weighed and the final concentration of carbohydrates in IL was measured by the high-performance liquid chromatography (HPLC) analysis as described in our previous work.[41] The final liquid solution was filtered with a 0.2 μm nylon membrane filter paper (diameter: 25 mm, purchased from Axiva, India) through a 0.22 μm syringe filter prior to HPLC analysis. The quantitative estimation of carbohydrates was achieved by HPLC (PerkinElmer Series 200, USA) with the Hi-Plex H column (7.7 × 300 mm) connected to the guard column (Agilent, USA). The column oven temperature was kept at 65 °C with 0.5 mL min–1 flow rate. The samples were analyzed with a refractive index detector using 0.01 M H2SO4 as the mobile phase. The sampling and quantification of the liquid phase were performed in triplicate, and the average values are reported here. The % CR was calculated according to the following equationwhere, m0 and mf are the mass of the initial sample (IL + sugar) and mass of the recovered IL, respectively. C0 and Cf are the concentrations of sugar in the initial and recovered IL, respectively. The % ILR is defined as