Thirty years of ionic liquids: a personal perspective

Introduction

There are two common themes in the scientific literature on ionic liquids (ILs) that appear frequently in research articles, reviews, communications, and conference presentations. The first is the “exponential growth” in the number of publications that mention “ionic liquids,” often shown as a histogram. The second is a bullet-point list highlighting the positive properties of ILs, such as low vapor pressure, non-flammability, high chemical stability, a large liquidous range, and a wide electrochemical window. These recurring themes in ILs literature serve as markers of the significance and popularity of this class of compounds. The scientific literature already offers detailed and comprehensive reviews on the fundamental aspects of ILs and their applications, ^{1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , 11 , 12} and this review does not intend to duplicate that work. Instead, the aim here is to reflect on the evolution – over the last thirty years – of certain concepts and achievements that, in the author’s subjective view, represent important, fascinating, and distinctive aspects underlying the success of ILs. Following a brief historical introduction leading to the so-called second generation of ILs, this work will focus on four key milestones in the study of ILs. First, the nanostructuring of ILs, which refers to the mesoscopic consequences of their molecular characteristics that make them much more than just molten salts. Second, the quantitative evaluation of intermolecular interactions that lead to nanostructuring. In this part, the significant development of the theory behind the nuclear Overhauser effect (NOE) applied to ILs will be critically examined, with the help of case studies. Third, the ability of ILs to dissolve gases, often in a selective manner. This property will be discussed in relation to local ordering, creating a connection with the previous section. Finally, the contribution of ILs to the advancement of 21st century chemistry, particularly in promoting renewable sources of carbon atoms, will be briefly highlighted.

Prehistory and history of ILS

If we consider the widely accepted definition of an ionic liquid (IL) as an ionic compound with a melting point below the boiling point of water at atmospheric pressure, the earliest examples of ILs can be traced back to 1888 and 1914. ¹³ Some remarkable examples are sketched in Scheme 1. In their pioneering work, Gabriel and Wiener ¹⁴ noted the unusually low melting point (52 °C) of ethanolamine nitrate, placing it within the general properties of propylamine derivatives. Around the same time, Paul Walden included ethylammonium nitrate (EAN), a salt with a surprising melting point of just 12 °C, in his broader investigation titled “On the molecular size and electrical conductivity of some molten salts”. ¹⁵ Chemistry at the turn of the 19th and 20th centuries was not yet advanced enough to anticipate the profound impact of these low-melting salts on the fundamental understanding of matter assembly and organization, nor their potential technological applications as early ILs. Indeed, it took 94 years after Walden’s publication to demonstrate that the network of intermolecular interactions in EAN resulted in the first instance of nanostructuring in an IL with an alkyl chain shorter than four carbon atoms. ¹⁶ The modern term “ionic liquids” and the emergence of the so-called first generation of ILs ¹ can be traced to 1978, when Osteryoung introduced butylpyridinium chloroaluminate. ¹⁷ A few years later, alkylimidazolium chloroaluminates were also introduced. ¹⁸ Early research on ILs was primarily focused on applications, largely overlooking the significant drawbacks of using strong Lewis acids like AlCl₃ and the high temperatures required for the practical use of chloroaluminate ILs in energy storage devices. The issue of ILs instability to air and humidity was successfully addressed in 1992 by Wilkes and Zaworotko. They used simple metathesis reactions on ethylmethylimidazolium iodide [EtMeim]I to prepare air and water stable ILs such as [EtMeim]NO₃, [EtMeim]BF₄, [EtMeim]MeCO₂, [EtMeim]₂SO₄·H₂O. ¹⁹ The authors not only paved the way to the second generation ILs (see Scheme 2 for a selection of the most common ILs), but also set the foundation for the rational understanding of the mechanisms underlying both liquidous range and nanostructuration: “It is clear from this study that EtMeim ⁺ is an ideal candidate for general use in ionic liquids. It is of moderate size (thereby lowering lattice energies), its shape appears to promote cation stacking, it is facile and inexpensive to prepare, chemically and electrochemically robust, and can only engage in the thermodynamically weak C–H⋯X hydrogen bonding”. ¹⁹ This conclusion anticipated the rigorous answer to the question “Why are ionic liquids liquid?” proposed in 2008 by Krossing et al. ²⁰ The authors assessed the Δ _fus G° of a collection of ILs through a Born-Fajans-Haber cycle (Fig. 1), and draw the clear-cut conclusion that “The negative Δ _fus G° values obtained for all ILs indicate that these salts are liquid at ambient temperatures […] due to the large size and conformational flexibility of the ions, which leads to small lattice enthalpies and large entropy changes that favor the liquid state.” ²⁰ The iconic pictures reported as Fig. 2 ²¹ summarizes the complex and appealing connection between molecular features and macroscopic behavior: The melting point trend of 1-methyl-3-alkylimidazolium PF₆ indicates the competition between the symmetry breaking effect of the alkyl chain – opposing to the coulombic attraction and destabilizing the crystal packing – and the attractive dispersive interactions, whose effect predominates over the “packing frustration” starting from C6. Overall, this is the preludio to the main theme of the local structuration of ILs.

Scheme 1:

Structures of compounds mentioned in the Prehistory section: (a) ethanolammonium nitrate, ref. 14]; (b) ethylammonium nitrate (EAN), ref. 15]; (c) butylpyridinium chloroaluminate, ref. 17]; (d) alkylimidazolium chloroalumiminates, ref. 18].

Scheme 2:

Sketch of the structure of the most common second generation ILs. Reprinted with permission from ref. 12] Copyright © 2008 Royal Society of Chemistry.

Fig. 1:

The Born-Fajans-Haber thermodynamic cycle to calculate Δ _fus G° of selected ILs. Adapted with permission from ref. 20]. Copyright © 2006 American Chemical Society.

Fig. 2:

Top: mp of methylalkylimidazolium PF₆ as a function of the number of C atoms in the alkyl chain. Bottom: optimized geometry of 1-methyl-3-octadecylimidazolium cation with the regions affecting the melting point trend. From ref. 21]. Copyright © 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim.

The magic of ILs: nanostructuration

The ILs have distinctive aggregation motives fingerprinting their physical chemistry. An exhaustive survey can be found in refs. 9], 10], which also highlight the large variety of spectroscopic and theoretical methods used to investigate the microheterogeneity in ILs. In this section I will focus on some paradigmatic achievements reported by neutron scattering, X-ray scattering and molecular dynamics. Considering the second generation ILs, the milestone work on neutron scattering of symmetric dimethylimidazolium (dmim⁺) salts from the group of Hardacre highlighted similarity and differences between the crystallographic order detected in the solid state and the local order in the liquids state, largely modulated by size, shape and electron distribution of the anion, have been highlighted. ²² The discovery that the anions’ structure influenced the first and second solvation shells sets the bases of the tunability of ILs properties according to the structure of the components. Figure 3 summarizes some key points in the paradigmatic example of chloride, PF₆ and bis-trifluoromethylsulphonyl imide (TFSI) anions. The surfaces show the probability of finding an ion with respect to the central imidazolium cation, considering both the cation – anion and the cation – cation shells. The mutual disposition of chloride – dmim and PF₆ – dmim are opposite, showing the dominant point-charge behavior of Cl⁻ leading to the strong C(2)–H⋯Cl hydrogen bond on the one side, and the purely coulombic interaction of dmim and PF₆ on the other. The distribution obtained for the dmim TFSI is driven by the high degree of charge delocalization and increased size of the anion. The probability surfaces show an increasing contribution of anion – cation mutual exclusion in the order Cl < PF₆ << TFSI. A good similarity between the liquid structuration and some crystallographic features was reported in the case dmim Cl whereas, on the opposite, no correlation was observed and reported in the case of dmim TFSI. The role of the alkyl chain in imidazolium-based ionic liquids was the object of a milestone paper ²³ that introduced, in 2006, the nowadays consolidated concept of polar – apolar domains as source of microheterogeneity in isotropic ionic liquids. The investigation, carried out by molecular dynamics simulations, demonstrated the tendency of the polar parts of the IL – the imidazolium ring and the counterion – to aggregate via coulombic interactions, and the apolar tails to segregate into domains whose size and interconnection increased with the number of C atoms of the alkyl chains and driven by dispersive interactions. The visual description of the progressive formation of the polar – apolar domains is reported in Fig. 4, actually one of the most iconic pictures in the history of ILs, due to the communication efficacy associated to the color code selected by the authors to designate the opposite domains (red: polar, green: apolar), and coincident with the colors of their own country flag. Figure 4 shows how the polar components arrange in a three-dimensional network of ionic channels, while the apolar domains progressively evolve from a dispersed microphase with C2 alkyl chains to a continuous, interconnected phase for longer chains (C6 → C12), with C4 setting the border for the two regimes.

Fig. 3:

Top: probability surface for anion around central imidazolium cation. Surface encloses top 5 % anions. Bottom: probability surface of imidazolium cations around central imidazolium ion. The surfaces enclose the top 20 % of molecules each in the distance range 0–9 Å. Reprinted with permission from ref. 22]. Copyright © 2007 American Chemical Society.

Fig. 4:

Simulation boxes for [C_nmim][PF₆]. Panel (a) shows a snapshot of the MD of [C₂mim][PF₆] with CPK graphics and conventional colors. Panels (b) to (f) show a snapshot of MD simulations of [C_nmim][PF₆] with n = 2, 4, 6, 8, 12, in the order. The color code is green for apolar domanins and red for charged domains. The use of red/green colors guides the eye to the detection of the progressive increase of domain size with increasing n. Reprinted with permission from ref. 23]. Copyright © 2006 American Chemical Society.

Experimental evidence for nanoscale segregation in ILs were proposed in 2007 by using X-ray scattering. ²⁴ Figure 5 shows the experimental scattering profile obtained on 4 different methyalkylimidazolium chlorides with increasing alkyl chain length (Cn with n = 3, 4, 6, 8, 10). The featureless curve displayed in the case of C2 progressively transforms into a profile showing a clear interference peak, pointing toward the existence of a correlation length L introduced by the alkyl chain. The position of the maximum shifts toward smaller Q – i.e. longer correlation lengths – with increasing n, while, at the same time, the peak gets sharper and more intense. The L values extracted via the Bragg’s equation do provide an excellent linear correlation with the n values, consistent with nanometric ordering introduced by the growing alkyl chains. The generality if the L versus n correlation was tested with BF₄ ⁻ analogues (Fig. 5, right panel). Also, in that case the experiments confirmed n = 4 as the critical length for nanoscale segregation. The correlation distances calculated fitted the linear correlation of Fig. 5, left panel, confirming the role of the alkyl chain decoupled from the anion type.

Fig. 5:

Left: X-ray diffraction profiles for a series of methylalkylimidazolium chlorides: [Cnmim]Cl, with n = 3,4,6,8,10 at 25 °C. The inset reports the linear correlation between the alkyl chain length and the measured spatial correlation L = 2π/Q _max in Å. Q _max is the momentum transfer corresponding to the interference peak. The circles indicate the correlation points found for the [Cnmim]Cl while the squares refer to selected [Cnmim][BF₄] (squares). Right: X-ray scattering intensity plots for selected [Cnmim][BF₄] at 25 °C. Reprinted with permission from ref. 24]. Copyright © 2007 American Chemical Society.

In a sufficiently large Q range, ILs usually show three distinct features in the X-ray small/wide angle (SWAXS) profile: a prepeak in the range 0.2–0.6 Å⁻¹, a second peak between 0.5 and 1.0 Å⁻¹, and a third feature between 1.0 and 2.5 Å⁻¹. A crystal-clear physical interpretation of these interference peaks can be found in refs. 25], 26], while Fig. 6 shows the concept in a pictorial way. The key importance of the prepeak for the assessment of the local structuration of ILs was stressed by the authors. Indeed, the existence of a prepeak “is indicative of polar–apolar alternation […] linked to structural heterogeneity”, thus “a missing prepeak is often indicative of small apolar components and a lack of polar–apolar alternation. A missing intermediate peak is likely due to cancellations because at room temperature charge alternation is always present.” ²⁶ From a different viewpoint, an appearing prepeak is diagnostic of a structural transition. Figure 7 provides an example of this. ²⁷ The overlayed profiles follow a variable T observation carried out on a pyrrolidinium based IL with the perfluorinated anion IM₂₄ (TFSI analogue with perfluorinatd ethyl and n-butyl side chains). The reversible appearance of a feature at T = −55 °C and centered at Q = 0.48 Å⁻¹ points toward the formation of fluorous domains due to segregation of fluorinated tails once the translational dynamics has been dumped at low T. This research broads the repertoire of nanodomains within ILs, setting fluorophilic domains in the toolcase with polar and apolar domains. The applicative consequences of the possibility of fluorous domains in ILs will be clarified in the next sections.

Fig. 6:

Simulated SWAXS profiles of paradigmatic ILs highlighting the three common features. The peak observed at higher q values corresponds to adjacency correlations between nearby atoms, which arise from both intermolecular and intramolecular interactions. The intermediate peak reflects charge alternation, while the smallest q value peak results from polarity alternation, indicating structural heterogeneity. The lack of a charge alternation peak, as seen in [Im_1,8][BF₄], does not imply the absence of this symmetry; rather, it signifies complete cancellation of peaks and antipeaks at this specific q value. For a discussion of the origin of peaks and anti-peaks see ref. 25]. Adapted with permission from ref. 26]. Copyright © 2015 American Chemical Society.

Fig. 7:

Small and wide angle X-ray scattering (SWAXS) intensity profiles of 1-butyl,3-methylpyrrolidinium [IM24] in −80 °C < T < 20 °C temperature range. The peak at q = 0.48 Å^–1 is detectable for T < −55 °C and it is consistent with the formation of fluorous tails clusters. Adapted with permission form ref. 27] Copyright © 2013 Royal Society of Chemistry.

The NOE saga: measuring the contacts

In 1952, a rather sectorial and visionary paper titled “Polarization of nuclei in metals” ²⁸ paved the way to an unthinkable set of applications in nuclear magnetic resonance spectroscopy, from protein structure to the most recent dynamic nuclear polarization (DNP), and spotlighted on a scientist whose name is now ubiquitous in the never-ending catalogue of NMR acronyms: Overhauser. ²⁹ Indeed, the tremendous technology advancement of NMR instrumentation and software, along with the capillary educational efforts for the dissemination of NMR culture in the chemical community, made easily available sophisticated experiments such as NOESY, ROESY, HOESY (all of them embedding the “O” of Overhauser) and variants even to undergraduate students as irreplaceable physical method for the assessment of three-dimensional structure of molecules in synthetic chemistry ^{30 , 31} and biochemistry. ³²

A “nuts and bolts” illustration of what the chemical applications of the Nuclear Overhauser Enhancement (NOE) relies on the basic concept of mutual relaxation (“cross relaxation”, here henceforth indicated with σ and expressed in s⁻¹) occurring among nuclei close in space and thus connected with configurational, conformational and assembly features of the molecular architecture. The observation of NOE – as, for instance, correlation peaks in a 2D NOESY experiment or in its rotating frame variant, ROESY – is due to dipolar interactions. These, in turn, decay rapidly with increasing distance r of the interacting nuclei (r ⁻⁶ distance dependency), and are sensitive to the dynamic regime, generally expressed by the product ω τ _c of the nuclei Larmor frequency ω and the rotational correlation time τ _c, in first approximation the time needed for a molecular rotation of one radian under random thermal tumbling. The distance dependence of NOE intensity, in particular, is the conceptual gate to the measurement of internuclear distances in solution. The limiting factor is assumed to be the distance threshold of 5 Å for vanishing NOE. This property paves the way to measure internuclear distances in the liquid, i.e., to use the NOE as a “proximity ruler”.

Whitin these premises, the first attempt of assessing the nanostructuration of ILs using the NOE proximity ruler came from the group of Osteryoung in 1995. ³³ The authors studied the ROESY pattern in 1-ethyl-3-methylimidazolium (Emim)⁺ chloroaluminate in the melt state. For the sake of clarity, and for the non-NMR-specialist reader, also the rotating frame NOE will be referred to as NOE (and not as ROE, with no comment in this article on the similarities and differences between NOE and ROE). In order to distinguish intramolecular short contacts – i.e. intramolecular NOEs, linked to the imidazolium geometry – and the intermolecular contacts – i.e. intermolecular NOEs, originated by cation-cation assembly due to nanostructuration – the authors synthesized the homologue IL based on the fully deuterated cation (Emim-d ₁₁)⁺ and mixed the pristine and the deuterated ILs in molar ratio 5:95. In this way, each (Emim)⁺ was totally surrounded by (Emim-d ₁₁) species, resulting in the complete quench of the intermolecular NOEs. Figure 8 shows the two-dimensional contour plot of the ROESY experiment carried out on the pristine IL and the 95:5 deuterated mixture. The comparison of the correlation peaks (cross-peaks originated by NOE between H nuclei in space proximity) detectable in the pure IL and in the mixture of protium and deuterated IL allowed the authors to distinguish intermolecular and intramolecular NOE. On this basis, the authors gave the first evidence of local structuration of an IL in the liquid state without solvent added, presumably due to the H-bond network connecting the imidazolium cations and the chloride anions. ³³ In the subsequent years, a different ROESY-based strategy to investigate the network of interactions driving imidazolium-based ILs’ local ordering was proposed. The emblematic IL 1-butyl-3-methylimidazolium BF₄ [Bmim][BF₄] was added of known amount of water used as competing agent to the intermolecular interactions. ^{34 , 35} The structural changes on the bulk IL were monitored by volume integration of the ROESY cross-peaks. The observed trends are reported in Fig. 9. The histogram displays two different regimes: the H pairs corresponding to intermolecular contacts (as worked out in ref. 33]) showed negative bars, indicating weaker NOE intensity with increasing water content, whereas the positive bars highlighted the opposite behaviour affecting the H pairs characterized by intramolecular NOE. The two opposite response to the growing water concentration point towards water molecules interacting with the polar domain, making the imidazolium-imidazolium association looser, possibly by competing for the C–H⋯F hydrogen bond. In turn, the augmented polarity experienced by the butyl chain with growing water concentration seems to reinforce the chain-chain interaction within the apolar domain, leading to increased NOE intensity. ^{34 , 35}

Fig. 8:

Left panel: molecular formula, atom number of the Emim⁺ cation with the contour plot of ROESY experiment. Peak assignment is reported for the one-dimensional projection and on the diagonal. On top of the right panel are reported the molecular formulae of protium and deuterated Emim⁺ cations, used by the authors in 95:5 M ratio. The orange circle and the arrows indicate the cross peaks that disappeared after dilution of the protium IL with the deuterated analogue, consistent with the intermolecular short contacts. Adapted with permission from ref. 33] Copyright © 1995 American Chemical Society.

Fig. 9:

Relative intensities of intermolecular ROESY cross-peaks in samples 1b–g. The samples were prepared from pure 1a (x _water 1a = 0) by adding a suitable amount of water to obtain the following water molar fractions x _i: 0.09, 0.17, 0.27, 0.36, 0.45, 0.52 for 1b to 1g, in the order. The relative intensity (%) is defined as I _rel = [(V − V ₀)/V ₀] × 100, with V ₀ is the integrated volume of a specific cross-peak in sample 1a, and V is the integrated volume of the same cross-peak 1b–1g. All volumes are internally normalized against the H2–H10 cross-peak, which is assigned a value of 1.00 (arbitrary units). A negative bar for a given cross-peak indicates a reduction in relative intensity compared to the corresponding cross-peak in the pure ionic liquid. Reprinted from ref. 34] with permission and under the terms of the Creative Commons Attribution (CC BY).

An attempt of measuring intermolecular dintances inside the polar domain of representative 1-butyl-3-methylimidazolium BF₄ [Bmim][BF₄] and 1-butyl-2,3-dimethylimidazolium BF₄ [Bdmim][BF₄] was reported in 2006. ³⁶ The methodological approach was based on the measurement of the NOE kinetics by using an array of NOESY experiments with incrementing mixing time τ (i.e. the time interval within NOESY the pulse sequence in which the dipolar interaction leading to NOE is allowed to occur) and extracting the cross-relaxation rate σ from the slope of the NOE build-up curve for τ = 0. The data showed unexpected resemblance of some of the intermolecular distances measured in the pure liquid with the corresponding ones in the solid state via single-crystal X-ray diffraction. This finding confirmed the existence of local structure in the pure ILs and alligned with early discoveries from X-ray scattering [see previous section]. In 2012 a sistematic investigation from the group of Giernoth and based on heteronuclear ¹H–¹⁹F NOE data highlighted an impressive matching of the NOE-derived intermolecular contacts with those predicted via molecular dynamics. ³⁷ These works assumed more or less tacitely that the formation of suitable aggregates of the IL’s components – from ion-pairs to more complex supramolecular assemblies – were stable enough to allow the treatment of the intermolecular NOE as intramolecular, i.e. characterized of a single correlation time of the vector connecting the space-close nuclei and rigorously adhering to the r ⁻⁶ distance dependence. ^{30 , 31} Although such an assumption fueled – and in part still fuels – the debate on the issue of the actual formation of long-lived aggregates within ionic liquids, at the time of the pubblication of the NOE-extracted intermolecular dinstances, “that was the state of the art for ILs”. ³⁸ Indeed, the theoretical framework underlying intermolecular NOEs is more intricate than for intramolecular cases. Comprehensive treatments of the theory, which can be conceptually challenging, are detailed in works by Ayant, ³⁹ Furo, ^{40 , 41} Gerig, ^{42 , 43} Halle, ⁴⁴ and Bagno. ^{45 , 46} Reduced to the bare minimum, the assumptions and main conclusions of the model can be summarized as follows: i) the intermolecular vector r connecting the nuclei undergoing cross-relaxation is modulated by translational diffusion. As a consequence, the randomization time of r is longer compared to the intramolecular case; ii) the predicted distance dependency is r ^–n , with 1 < n < 6, depending on the system dynamics; iii) also long-range interactions may give rise to observable intermolecular NOEs. In other words, contacts beyond the first solvation shell may provide observable, long-range NOE. The application of these theoretical framework to model [Bmim][BF₄] was presented in 2013 in a disruptive paper by Gabl, Steinhauser and Weingärtner. ⁴⁷ On the bases of their simulations, the authors showed how the distance dependency of intermolecular NOE was a function of the Larmor frequency of the interacting nuclei. This conclusion prompted the authors to state that, when measuring and interpreting NOE, “frequency matters.” ⁴⁷ Figure 10 shows one of the graphic outputs of their simulations. To fully understand the disruptive meaning of the curve, it is necessary to introduce here the spectral density function J(ω), a mathematical object containing the cross-relaxation σ _IS and describing the modulation of the dipole-dipole interactions of the nuclei developing NOE, here indicated as I and S. The authors pointed out that the overall J(ω) was a linear combination of J(ω _I  + ω _S ) and J(ω _I  − ω _S ), in turn associated to short-range and long-range interactions, respectively, and that the term (ω _I  − ω _S ) overpowered (ω _I  + ω _S ). In the case of ¹H–¹H or ¹H–¹⁹F NOE, the corresponding (ω _I  − ω _S ) are very small, due to the similarity of the Larmor frequencies of I and S, heading to interactions beyond the first solvation shell. On the contrary, heteronuclear NOEs such as ¹H–⁷Li or ¹H–³¹P are expected to truly spot on short contacts, due to the large difference or the Larmor frequency of the nuclides, thus providing the researcher with a unique spectroscopic tool to assess an atomistic picture of solvation.

Fig. 10:

Graphic representation of the exponent n of the r ^–n distance dependency of intermolecular NOE intensity as a function of operating frequency. The simulation is obtained for a basic spectrometer frequency of 400 MHz. The reference points ω ₁ to ω ₄ are set at 1 MHz, 0.15 GHz, 2.5 GHz, and 4.85 GHz, returning n exponents of 1.17, 2.5, 7.2, and 8.2, in the order. The data clearly show the transition from long-range (small exponent) to short range (large exponent) distance dependency of NOE. Reprinted from ref. 47] with permission and under the terms of the Creative Commons Attribution (CC BY).

After 2013, many efforts have been done to refine the acquisition and processing of intermolecular data according to the GSW approach. It is worth noting that the refined fitting of appropriately normalized HOESY build-up curves, using a modified model that incorporates both longitudinal relaxation times and diffusion coefficients, enables accurate determination of absolute intermolecular cross-relaxation rates. This approach, elegant although not straightforward, facilitates comparative analyses across various ionic liquids, concentrations, and temperature conditions. ^{48 , 49 , 50} The GSW model received some validations, partly based on the comparison of the NOE data and the intermolecular distances available from other methodologies, such as single crystal X-ray crystallography ⁵¹ or molecular dynamics. ^{52 , 53} The group of Schröder proposed a full simulation of Emim trifluoromethanesulphonate (triflate, TfO) as pure liquid and containing LiTfO. ⁵³ The authors simulated the systems using both fully atomistic and polarizable models, as nonpolarizable force fields tend to exaggerate directed electrostatic interactions in ionic liquids (ILs). In reality, electronic charge distributions adjust during molecular interactions, leading to reduced viscosity in ILs. Consequently, polarizable force fields are essential for capturing realistic pair diffusion dynamics critical to the NOE. Additionally, the NOE were directly extracted from the simulation trajectory, bypassing traditional models and thus avoiding their inherent assumptions. The comparison of experimental and calculated ¹H–¹H, ¹H–¹⁹F and ¹H–⁷Li NOEs was exceedingly good and the simulations of the distance dependency of NOE on the Larmor frequency difference of the interacting nuclei fully confirmed the long-range nature of ¹H–¹H, ¹H–¹⁹F NOE, and the short-range contribution of ¹H–⁷Li NOEs, as displayed in Fig. 11.

Fig. 11:

Plot of the exponent n of the r ^–n power law for NOE intensity as a function of the magnetic field (in frequency units). The H−Li spins show a distance dependency nearly independent of the magnetic field and consistent with r ^–6 power law. The H−F spin-pairs curves show a marked magnetic field dependency and long-range character at the typical spectrometer frequency. Reprinted from ref. 53] with permission and under the terms of the Creative Commons Attribution (CC BY).

Finally, two critical considerations must be highlighted at the conclusion of this section. First, while the GSW theory has undeniably prompted NMR specialists to reconsider the use and interpretation of NOE in cases of solvation and intermolecular aggregation, the authors may have overlooked its potential to broaden, rather than restrict, the scope of intermolecular NOESY experiments. The long-range nature of certain important classes of intermolecular NOEs, such as ¹H–¹H and ¹H–¹⁹F, should not deter chemists from measuring them. Their interpretation contributes to an atomistic understanding that complements the microheterogeneity revealed by WAXS, SAXS, or SANS techniques. As demonstrated previously, the combined application of NOEs, scattering methods, and molecular dynamics (MD) simulations is a powerful investigative tool for studying nanostructured liquids. ^{52 , 54} Conversely, the strictly short-range nature of specific heteronuclear NOEs (e.g., ¹H–⁷Li NOEs) offers accurate and valuable distance constraints for MD simulations of systems of contemporary interest, such as ionic liquid electrolytes containing dissolved lithium salts. ⁵⁵

Second, despite the significant validation of GSW theory through a comparison with experimental data, the debate regarding its general applicability is still open. Notable examples of excellent concordance between NOE-derived structural models and experimental or calculated data exist in the literature without relying on the GSW framework. ^{56 , 57} Further research and theoretical advancements are essential to reach a conclusive resolution to this ongoing scientific discussion.

The Horror Vacui of gases in ionic liquids

Two primary motivations have driven research into gas solubility in ionic liquids: the need for novel technologies for gas capture, particularly CO₂, and a deeper understanding of how the nanostructuration of ILs influences the mechanisms of gas dissolution. The growing interest in gas capture is largely tied to environmental concerns related to climate change caused by CO₂ emissions. However, research in the early 2000s focused mainly on the engineering aspects of raw gas purification, as CO₂ is the primary contaminant in natural gas streams. Gas-sweetening technologies, based on CO₂ reactions with alkanolamines to form carbonates and carbamates, offer excellent efficiency and (apparently) low costs. Nonetheless, this approach has significant drawbacks, including reagent loss, water transfer to the gas stream, formation of corrosive by-products due to amine degradation, and high energy consumption for regeneration, estimated at 4 GJ per tonne of CO₂. ^{58 , 59}

Starting from the first report accounting for the unexpected CO₂ solubility in ILs, ⁶⁰ ILs have been successfully tested as absorption materials using various strategies. These include the dissolution of gas in bulk ILs for selective CO₂ retention, chemical reactions of CO₂ with task-specific ILs (TSILs), and selective gas separation using supported IL membranes (SILMs). ^{61 , 62 , 63} This non-exhaustive list also includes the use of polymerized ILs (polyILs) and polyIL-IL composite materials. ⁶⁴ One particularly intriguing application involves switchable ILs for CO₂ sequestration. In 2005, it was demonstrated that a mixture of the superbase 1,8-diazabicyclo-[5.4.0]-undec-7-ene (DBU) and an alcohol could be reversibly transformed into an IL through the in situ formation of a DBUH carbonate with the alcohol. The reaction could be reversed by bubbling N₂ or Ar through the IL, regenerating the starting materials (DBU and alcohol). ⁶⁵ A smart application of this concept emerged in 2010, where the alcohol functional group was incorporated into the N-alkyl chain of the alkylimidazolium cation. For example, 1-(2-hydroxyethyl)-3-methylimidazolium bis(trifluoromethylsulfonyl)imide ([Im₂₁OH][Tf₂N]), Fig. 12) captures CO₂ via its reaction with DBU to form a liquid amidinium alkylcarbonate. The system’s recyclability, a key feature for the greenness of ILs, was achieved simply by raising the temperature or bubbling N₂. The key mechanism involves the interaction of a strong proton acceptor (DBU) with a proton donor (the OH functional group on the modified alkylimidazolium cation), forming an IL. ⁶⁶

Fig. 12:

Scheme of the reaction of superbase 1,8-diazabicyclo-[5.4.0]undec-7-ene (DBU) with hydroxylated imidazolium derivative [Im₂₁OH][Tf₂N] in the presence of CO₂. The deprotonated imidazolium forms reversibly the imidazolium carbonate. [Im₂₁OH][Tf₂N] and DBU are easily recovered by N₂ bubbling. Reprinted with permission from ref. 66]. Copyright © 2010 Royal Society of Chemistry.

This method was further generalized by the same research group, leveraging the well-known acidity of the C(2)–H site of imidazolium cations. ⁶⁷ As shown in Fig. 13, the superbase 1,3,4,6,7,8-hexahydro-1-methyl-2H-pyrimido[1,2-a]pyrimidine (MTBD) in its protonated state MTBDH⁺ has a counterion derived from the functionalization of the imidazolium ring at C(2) with a carbonate functional group. The formation of such a zwitterion involves a nucleophilic attack on the CO₂ carbon by the deprotonated, electron rich C(2) of the imidazolium ring, forming a carbonate functional group. Chemisorption of CO₂ is thereby achieved, in a reversible way, without prior functionalization of the IL, relying instead on the intrinsic properties of the C atom flanked by N atoms in the imidazolium ring.

Fig. 13:

Deprotonation of C(2) on the imidazolium cation operated by the superbase generates in situ the nucleophilic intermediate (not shown in the picture) immediately reacting with CO₂ and generating the final doubly charged ionic liquid. Reprinted with permission from ref. 67]. Copyright © 2010 Royal Society of Chemistry.

Beyond the undeniable practical interest in gas entrapment in ILs, the quest for fundamental knowledge about the mechanism of gas solubility has sparked curiosity among many research teams. With the metaphorical “Hamletic question” – “Why is CO ₂ so soluble in imidazolium-based ionic liquids?” – A landmark paper ⁶⁸ explored this topic in depth for the first time. Using molecular dynamics (MD) simulations, the authors explored the local structure of [Bmim][PF₆], [Bmmim][PF₆], and their homologues with the NTf₂ anion, both as pure liquids and in the presence of CO₂. Their findings, pivotal for our understanding, can be summarized as follows (see Fig. 14): i) By defining clever distance descriptors for calculating radial distribution functions g(r), indicative of local order, the study revealed that CO₂ solubility is predominantly driven by CO₂-anion interactions, with the cations playing only a negligible stabilizing role. ii) In an intriguing twist, the presence of 10 mol% dissolved CO₂ in [Bmim][PF₆] or [Bmmim][PF₆] left the anion-cation arrangement unchanged. For instance, the g(r) profile based on the P⋯C(2) distance between anion and cation remained identical whether CO₂ was present or not. ⁶⁸ iii) Coupled with calculated densities, this lack of structural perturbation aligns with experimental observations showing no volume expansion of ILs upon CO₂ dissolution. The initial conclusion seemed almost too straightforward: CO₂ occupies the liquid’s interstitial voids.

Fig. 14:

Top left: The radial distribution functions g(r) and number integrals I(r) illustrate the spatial relationship between the phosphorus atom of the [PF₆]⁻ anion and the C2 carbon of the cation. The data are represented as follows: g(r) for [bmim] (○), g(r) for [bmmim] (□), I(r) for [bmim] (●), and I(r) for [bmmim] (■). Bottom left: The same analysis is presented for a system containing 10 mol% CO₂. The results indicate that CO₂ does not significantly alter the structural arrangement of cations and anions. Top right: Radial distribution functions and number integrals describing the spatial correlation between the carbon atom of CO₂ and the phosphorus atom in [PF₆]⁻. The symbols represent: g(r) for [bmim] (○), g(r) for [bmmim] (□), I(r) for the phosphorus atom in [bmim][PF₆] relative to CO₂ carbon (●), I(r) for the phosphorus atom in [bmmim][PF₆] relative to CO₂ carbon (■), I(r) for the CO₂ carbon relative to the phosphorus in [bmim][PF₆] (▲), and I(r) for the CO₂ carbon relative to the phosphorus in [bmmim][PF₆] (♦). Bottom Right: Radial distribution functions and number integrals illustrating the spatial correlations between CO₂ and the N1 nitrogen of the cation. The symbols represent: g(r) for [bmim] (○), g(r) for [bmmim] (□), I(r) for the nitrogen in [bmim] relative to the CO₂ carbon (●), I(r) for the nitrogen in [bmmim] relative to the CO₂ carbon (■), I(r) for the CO₂ carbon relative to the nitrogen in [bmim] (▲), and I(r) for the CO₂ carbon relative to the nitrogen in [bmmim] (◆). Adapted with permission from ref. 68]. Copyright © 2004 American Chemical Society.

While this “void-filling” argument appears captivating and intuitive, further structural analysis contradicted its simplicity. Huang et al. ⁶⁹ tackled the void conundrum using Voronoi polyhedral analysis. ⁷⁰ They demonstrated that the dimensions of existing voids are too small to accommodate CO₂ molecules. Instead, the CO₂ dissolution mechanism in [Bmim][PF₆] involves a subtle angular displacement of the anion relative to the imidazolium ring plane (Fig. 15). This tiny rearrangement does not visibly alter g(r), consistent with, ⁶⁸ yet successfully creates voids large enough to host CO₂ molecules.

Fig. 15:

Proposed model for CO₂ placement [Bmim][PF₆] according to ref. 68]. The left panel sketches the initial arrangement of cation, anion and CO₂. The arrows indicate the displacement of CO₂ and anions to reach the final positioning of CO₂, shown in the right panel. Adapted with permission from ref. 69]. Copyright © 2005 American Chemical Society.

The narrative emerging from these studies paints ILs as liquids teeming with small voids, which rearrange into larger cavities upon the introduction of CO₂. This unique rearrangement enables significant gas dissolution without causing volume expansion. More recently, Shelepova and Medvedev refined this model by conceptualizing interstitial spaces as “empty spheres inscribed between atomic surfaces”. ⁷¹ Their simulations identified large empty spheres (radius > 0.15 nm) capable of hosting CO₂ molecules. In an unexpected twist, they observed that these large spheres tended to cluster around CO₂ molecules, suggesting that dissolving CO₂ does not merely exploit pre-existing voids – it actively generates additional empty spaces in the IL.

The rigorous definition of “free volume” or “voids” in ionic liquids (ILs) is not as straightforward as it is in polymers, where parameters such as size, shape, size distribution, topology, and connectivity can be explicitly defined. ⁷² For ILs, Beichel and Krossing ^{73 , 74 , 75} reported a collection of experimental average local free volume (“hole” volume, denoted as ⟨v _h⟩) using positron annihilation lifetime spectroscopy (PALS). This data provides a valuable reference for geometric features, particularly when combined with other physical methods. For instance, ¹²⁹Xe NMR spectroscopy, known for its sensitivity to compression effects experienced by gas atoms in a porous medium, serves as a molecular probe to explore molecular voids. ^{76 , 77}

Measurements on non-ideal IL mixtures, such as [C₄C₁im][Me₂PO₄]_x[NTf₂]_1–x and [C₄C₁im][OTf]_x[NTf₂]_1–x, have shed light on the nature of free volume in ILs, as illustrated in Fig. 16. ⁷⁸ The average excess hole volume v h E (see caption to Fig. 16 for definition) and the excess volume V ^E , derived from density measurements, showed a fair correlation, suggesting that the excess mixing volume originates from the average cavity size in the liquid. Similarly, the excess δ ^E (see caption to Fig. 16 for the definition) correlated well with both V ^E and  v h E .

Fig. 16:

Top: Plot of the measured excess hole volume v h E as a function of the excess molar volume of mixing V ^E for the [C₄C₁im][Me₂PO₄]_x[NTf₂]_{1_x} mixtures. The values of v h E were obtaind from PALS measurements as difference between the measured hole volume and the ideal value predicted by linear relatinship between the hole volumes determined in the pure liquids. V ^E are determined by density measurements. Bottom left: correlation between ¹²⁹Xe NMR excess chemical shift and v h E from PALS measurements from the same mixtures as above. δ ^E is defined as δ ^E  = δ − x ₁ δ ₁ − x ₂ δ ₂, where x _1,2 is mole fraction of IL 1 and IL 2 respectively, and δ _1,2 is the ¹²⁹Xe chemical shift in pure IL 1 and 2, respectively. Bottom right: correlation between ¹²⁹Xe NMR excess chemical shift and excess molar volume of mixing V ^E for the same mixtures. Reprinted from ref. 78] with permission and under the terms of the Creative Commons Attribution (CC BY).

This correlation between PALS and NMR data, unprecedented at the time of publication, validated the use of ¹²⁹Xe NMR as an atomistic probe for free volume in liquids. Despite the vast difference in timescales between the two techniques – seconds for NMR versus nanoseconds for PALS – the results highlight a strong connection between static (PALS) and dynamic (NMR) free volumes. This finding suggests that fluxional processes have a limited influence on the free volume of ILs. Ultimately, the free volume in ILs appears to arise from more persistent structural effects, such as Coulombic and steric interactions, which contribute to the ionic superstructure characteristic of these liquids. ⁷⁸

In the final part of this section, the spotlight shifts to a less common but equally important gas: molecular nitrogen N₂. Traditionally, N₂ was regarded as a contaminant in natural gas, similar to CO₂, with its separation presenting a significant challenge due to its well-known chemical inertness and lack of distinct acid-base properties. Recently, however, this perspective has been entirely reversed, as N₂ is now viewed as a virtually limitless and cost-free resource. The renewed interest in N₂ is closely linked to the sustainability challenges associated with NH₃ production, which remains predominantly reliant on the Haber–Bosch process. It is helpful reminding here that the 80 % of worldwide production of ammonia is addressed to fertilizers, the remaining 20 % is used for detergents, drugs, explosives and other products. Additionally, the growing importance of ammonia is also connected to its potential application as energy vector at zero CO₂ emission.

The primary sources of hydrogen for large-scale NH₃ production include hydrocarbons (e.g., methane, biomass, fossil fuels) and water. While using water as a hydrogen source demands significantly more energy than methane (+613.4 kJ/mol N₂), it is more environmentally friendly, avoids greenhouse gas emissions, and conserves finite hydrocarbon resources. These advantages are maximized when the required energy is derived from clean, renewable sources such as solar or wind power. ⁷⁹

One of the major challenges in nitrogen reduction reactions (NRR), i.e., the activation of the inert molecular nitrogen to a more reactive N-species, lies in the extremely low solubility of N₂ in aqueous electrolytes (∼0.00061 M at 25 °C and 1 atm). This is due to the nonpolar nature of N₂, its strong triple bond, and low polarizability, all of which significantly limit the amount of N₂ available for reaction. Among the strategies developed to enhance N₂ dissolution, the use of non-aqueous electrolytes, such as aprotic ionic liquids, has shown promising results. ⁸⁰ Specifically, it was highlighted how the presence of fluorous domains in the IL can be a determining factor for good nitrogen solubility. The flourination can be chemically inserted in the alkyl chain of the cation or in the anion or in both. Some authors reported on the role of the cations’ structure on the solubility of CO₂ and N₂, both increasing in ionic liquids with fluorinated cations due to the rigidity of the fluorinated chains, which results in a larger free volume. ⁸¹ Considering the N₂ solubility as the main target, rather than achieving good CO₂/N₂ selectivity, more recent papers underlined the basic strategy of exploiting the known solubility of N₂ in fluorinated solvents to host N₂ in the fluorinated domains of suitable ILs. ^{82 , 83} The rationale behind this is in the structural and electronic properties of N₂ – low polarizability, strong covalent bonds, lack of specific intermolecular interactions – indicating that a system with scarce tendency to establish van der Waals interactions could fit the principle similia similibus solvuntur. Figure 17 offers a graphic summary of these concepts and the results. The authors considered the N₂ solubility in two reference benchmarks (water and [C₂mim][NTf₂]) and compared it to those achieved with the ILs shown in the scheme above the histogram. The pale- and dark blue bars indicate the molar solubility of N₂ in two ILs made of a tetraalkylphosphonium cation with fully hydrogenated chains and the anion of perfluorobutylsulphonic acid. The histogram points out that the N₂ solubility is largely exceeding that of the benchmarks and that elongating the perfluoroalkyl chain length from C4 to C8 further improves the solubility. The adjacent two green bars show that the result is practically independent of the type of perfluorinated acid (sulphonic vs carboxylic), thus highlighting the crucial role of the fluorine pool on the anion. This conclusion is confirmed by the last set of bars: the comparison between [P₄₄₄₈] (fully hydrogenated cation) and [P_444(8f)] (partially fluorinated cation) shows that the fluorination of the cation is not the determinant factor for N₂ solubility. ⁸² Overall, the formation of fluorous domains in the ILs paves the way to new classes of non-aqueous electrolytes for up-to-date electrochemical activation of molecular nitrogen.

Fig. 17:

Top: Molecular formulae of fluorinated ILs for N₂ capture. The histogram shows the N₂ solubility (in mmol L⁻¹) of selected ILs compared to benchmark [C₂mim][NTf₂] (grey bar, left). The dotted line is referred to the N₂ solubility in water. Adapted with permission from ref. 82]. Copyright © 2018 American Chemical Society.

What are the key factors determining the ability of fluorinated functional groups to influence N₂ solvation and solubility? This question remains open to investigation. A recent study by Li et al. ⁸⁴ explored this issue using molecular dynamics simulations and free energy perturbation methods to examine N₂ solubility in 20 ionic liquids (ILs), including both traditional and fluorinated types. Using [Emim]NO₃ as the benchmark for N₂ solubility, the study reported a remarkable enhancement – up to 118 times – for the fluorinated IL [Emim]FAP.

The authors identified two key, interconnected factors driving this solubility boost: (i) the presence of at least 10 C−F bonds per IL formula, and (ii) exothermic solvation. In contrast, ILs with fewer C−F bonds and endothermic solvation exhibited significantly lower N₂ solubility. These observations were encapsulated in a newly proposed metric, fluorine densification energy (FDE), which quantifies interaction strength in fluorine domains and correlates with C−F bond count. Lower FDE reduces N₂ -anion dissociation energy while increasing free volume, thereby enhancing solubility. Medium-to-long fluorinated alkyl chains within a polar environment form distinct domains, further emphasizing FDE’s role in improving N₂ solubility. ⁸⁴

These findings provide a valuable basis for designing ILs tailored for efficient N₂ capture and conversion. In the context of this review, the cited work highlights the critical interconnection between understanding the nonstructural organization of ILs, elucidating molecular interactions at the atomic level, and applying this fundamental knowledge to develop fit-for-purpose solvents.

Ionic liquids and the 21st-century chemistry

One of the major concerns in modern chemistry and chemical engineering is the urgent need to rethink chemical processes to ensure a more sustainable and responsible use of natural resources. A formal acknowledgment of this necessity can be found in Principle No. 7 of the Green Chemistry Decalogue, ^{85 , 86 , 87} which advocates for the use of renewable feedstocks. This principle is particularly impactful in organic and polymer chemistry, where fossil-based carbon sources remain the primary raw materials.

A virtually limitless reservoir of renewable chemical feedstocks for a redesigned, sustainable chemistry is represented by lignocellulosic waste – such as sugarcane bagasse, wheat straw, rice straw, corn husks, and stover – which is produced in vast quantities each year but is only partially utilized for value-added products, aside from biofuel production. ⁸⁸ Using cellulose, for instance, as a carbon source in chemical processes instead of petroleum derivatives could contribute to carbon-neutral chemistry, given that CO₂ is reabsorbed during the natural regeneration of cellulose. However, despite its potential, this shift is hampered by cellulose’s intrinsic recalcitrance to dissolution and chemical transformation, necessitating a multi-step approach to effectively process biomass-derived materials. This chemical stability arises from the complex three-dimensional architecture of lignocellulosic materials, where cellulose, hemicellulose, and lignin form microfibrillar units that provide both chemical and mechanical robustness. ⁸⁹ A key step in utilizing cellulose as a chemical precursor involves its degradation into furan derivatives (Fig. 18), ⁹⁰ with particular attention given to 2-hydroxymethylfurfural (HMF). HMF has recently been described as a “sleeping giant” ⁹¹ due to its transformative potential in the shift toward non-petroleum-based chemistry – one of the most promising scenarios for 21st-century chemical innovation. ⁹²

Fig. 18:

Strategies for the conversion of ligno-cellulosic biomass into furane derivatives. Reprinted from ref. 90] with permission and under the terms of the Creative Commons Attribution (CC BY).

The first challenge in cellulose processing is its solubilization, which traditionally relies on derivatizing agents – such as NaOH and CS₂ in the “Viscose” process, which produces cellulose xanthogenates – or non-derivatizing solvents, like N-methylmorpholine-N-oxide in the “Lyocell” process. ⁹³ While extensively studied, ^{93 , 94} these methods often suffer from high environmental impact and/or excessive energy consumption, making the search for more sustainable alternatives imperative. A major breakthrough came in 2002, when a selected set of imidazolium-based ionic liquids (ILs) was proposed as efficient, non-derivatizing solvents for cellulose. ⁹⁵ This discovery marked the beginning of a strong connection between ILs and biomass valorization, paving the way for more efficient utilization of natural waste. Systematic tests on alkylimidazolium ILs with varying alkyl chain lengths and counterions, ranging from highly hydrogen-bond-accepting chloride to large, non-coordinating anions such as PF₆ ⁻, BF₄ ⁻, Br⁻, and SCN⁻, were proposed. Results revealed that chloride-based ILs with medium alkyl chain lengths – specifically [Bmim][Cl] – exhibited the highest cellulose solubility, especially when combined with microwave heating. This approach enabled the preparation of solutions containing up to 10 wt% cellulose (viscous liquid phase) and even 25 wt% cellulose (semi-solid, nematic phase).

The underlying mechanism behind this remarkable solubility was attributed to chloride anions disrupting the extensive network of hydrogen bonds that hold cellulose chains together in a compact, fibrillar structure. A conceptual sketch of this process is shown in Fig. 19. By breaking these intermolecular hydrogen bonds, the chloride ions enable individual polysaccharide chains to interact with the solvent, effectively dissolving the cellulose. ⁹⁶ This solvation mechanism was elegantly confirmed through multinuclear NMR relaxation studies. ⁹⁷ The approach involved monitoring the dynamic behavior of pure ILs and IL-cellulose solutions using ¹³C and ³⁵Cl NMR relaxation times (T ₁ and T ₂). Since ¹³C data reflect the dynamics of the imidazolium cation, while ³⁵Cl NMR selectively probes the chloride environment, this method provided key insights into solvent-solute interactions. As shown in Fig. 20 (top panel), the ¹³C relaxation data indicate a transition in cation mobility from slow tumbling (T < 70 °C, spin-diffusion regime) to fast tumbling (T > 70 °C, extreme narrowing limit), as evidenced by the minimum of the T ₂ curve. Meanwhile, the ³⁵Cl relaxation times decrease with temperature due to viscosity changes and ion pair formation. The central panel highlights the contrasting responses of the cation (left) and anion (right) to increasing cellobiose concentration – a disaccharide chosen as a simple model for cellulose solutes. While ¹³C relaxation times remain largely unaffected, the³⁵Cl data exhibit a pronounced concentration dependence, providing clear evidence of strong chloride-solute interactions.

Fig. 19:

Scheme of the competitive hydrogen bond action of chloride ions on the intermolecular, interchains hydrogen bonds of cellulose causing the disruption of the supramolecular organization of the cellulose strands. Reprinted with permission from ref. 96]. Copyright © 2008 Elsevier B.V.

Fig. 20:

Top left: ¹³C NMR T ₁ (•) and T ₂ (○) relaxation times as a function of T (°C) for the N-methyl groups (C4′, full line) and methyl groups at the end of the butyl chain (C1″, dashed line) of [Bmim][Cl]. Top right: ³⁵Cl NMR T ₁ (•) and T ₂ (○) relaxation times for chloride ions as a function of T (K) of [Bmim][Cl]. Middle: Same as above at constant T = 90 °C as a function of cellobiose concentration mimiking the presence of cellulose (in brakets the corresponding viscosity). Bottom: Lineshape of ³⁵Cl NMR signal of chloride ions at 90 °C with increasing concentration of cellobiose (left panel), glucose (middle panel), and glucose pentaacetate (right panel) in [C₄mim]Cl. Reprinted with permission from ref. 97]. Copyright © 2006 Royal Society of Chemistry.

Further confirmation comes from the bottom panel of Fig. 20, which highlights the sensitivity of the ³⁵Cl NMR linewidth profile to solvation shell symmetry, a direct consequence of the quadrupolar nature of ³⁵Cl relaxation. The spectral bands shown in Fig. 20 (bottom) exhibit significant broadening with increasing solute concentration when the solute acts as a hydrogen bond donor toward chloride. In contrast, when using peracetylated glucose – incapable of donating hydrogen bonds – the spectral lines remain sharp. These findings reinforce the pivotal role of chloride anions in cellulose dissolution, further supporting the potential of ILs as effective solvents for biomass-derived materials. ^{97 , 98} From a methodological perspective, the cellulose–alkylimidazolium chloride system serves as a paradigmatic example of the unique capability of NMR spectroscopy to selectively monitor individual components within IL-based systems by appropriately tuning the observation channel (see also the previous section on NOE).

The anion’s function as a hydrogen bond acceptor was thus identified as the key structural feature enabling cellulose dissolution. This insight laid the foundation for the true “rational design” of ILs – a concept frequently emphasized in the introductory sections of research articles on ILs (see Introduction). To further validate this model, Xu et al. ⁹⁹ conducted a systematic study of anion basicity, providing valuable insights for the scientific community. They were the first to report Kamlet-Taft (K-T) β values for a broad set of imidazolium-based ILs, establishing a clear relationship between anion structure and its H-bond acceptor capability, as quantified by β values. Notably, they proposed a simple spectroscopic criterion to assess how efficiently the anion acts as hydrogen bond acceptor, leveraging the well-known sensitivity of the C(2)–H ¹H NMR chemical shift in imidazolium cations to the presence of a hydrogen bond acceptor. ¹⁰⁰ Their study revealed a compelling linear correlation between cellulose solubility data and K-T β values, offering a practical criterion for selecting suitable ILs for cellulose dissolution. A graphical summary of this method is presented in Fig. 21.

Fig. 21:

Top: Correlation plot for cellulose solubility in imidazolium-based ILs and their Kamlet-Taft basicity parameter β. Bottom: Correlation between cellulose solubility imidazolium-based ILs and their ¹H NMR chemical shift at C(2)–H. Reprinted with permission from ref. 99]. Copyright © 2010 Royal Society of Chemistry.

Building upon this rational design approach, the same authors explored an additional strategy by leveraging the ability of Li⁺ ions to bind to Lewis base sites, such as the hydroxyl functional groups of cellulose. ⁹⁹ This interaction further weakens the intermolecular hydrogen bonds that maintain cellulose’s compact structure, thereby enhancing its solubility. Ultimately, the combination of a basic anion – such as acetate – with the addition of Li salts creates optimal conditions for cellulose dissolution, with the added benefit that the IL can be easily recovered without decomposition after precipitation in water, used here as an antisolvent.

Now, the final piece of the puzzle comes into play: IL recyclability, a crucial factor for the sustainable application of ILs. ¹⁰¹ A compelling case study is provided by Sixta’s research group, which investigated a recycling process for waste cotton fibers. In 2016, they successfully demonstrated a two-step process for producing man-made cellulosic fibers (MMCFs) using a Lyocell-type dry-jet wet spinning method. This approach started with a solution of post-consumer cotton fibers dissolved in a non-imidazolium-based IL. ¹⁰² The IL chosen for producing spinnable (dope) solutions was 1,5-diazabicyclo[4.3.0]non-5-enium acetate ([DBNH][OAc]), whose chemical structure is illustrated in Fig. 22. While the regenerated fibers displayed superior textile strength compared to both natural fibers and other MMCFs obtained via Lyocell-type processes, a critical limitation emerged: [DBNH][OAc] was not recyclable. After a single dissolution cycle, the IL completely lost its ability to dissolve cellulose pulp, rendering it unsuitable for reuse. To address this issue, in 2022, the researchers replaced [DBNH][OAc] with 7-methyl-1,5,7-triazabicyclo[4.4.0]dec-5-enium acetate ([mTBDH][OAc], see Fig. 22). This modification allowed quantitative solvent recovery from the coagulation bath across five consecutive cycles, without any loss of performance. ¹⁰³ The optimization of solvent recycling enabled the full integration of IL technology into a circular economy framework, as depicted in Fig. 22. This study serves as a paradigmatic example of how ILs can be effectively utilized in sustainable industrial production.

Fig. 22:

Left: superbases used by the authors of refs. 102], 103] to generate ILs for cellulose fibers recycling. Right: scheme of the circular economy loop after solvent recovery. Adapted with permission from ref. 103]. Copyright © 2020 American Chemical Society.

Conclusions and outlook

In recent years, ionic liquids (ILs) have transitioned from academic curiosities to viable industrial tools. A comprehensive 2020 review of patents and applications emphasizes that while only a few full-scale IL-based processes have been commercialized, a substantial number are in pilot or demonstration stages – totaling over 60 patents and nearly 300 scientific reports covering chemical synthesis, gas storage, environmental remediation, and more. ¹⁰⁴ The study highlights that economies of scale and improved production methods – such as continuous-flow synthesis and halide-free routes – are reducing costs and making ILs increasingly competitive with conventional solvents, especially when improvements in recyclability, selectivity, and safety are taken into account. ¹⁰⁵

What challenges lie ahead for ionic liquids in the near future? Among the many possible answers, I will focus on one that is particularly relevant today: sustainability. The early months of 2025 have seen an unpredictable yet remarkable trend in the Western world – a move to scale back the so-called Green Deal. This shift raises critical questions about the feasibility of reconciling environmental goals with labor and market dynamics (for instance, the ongoing controversy surrounding the transition from combustion engines to electric vehicles). In the context of this article, and in my personal and admittedly subjective view, the most significant challenge for ILs lies in the sustainability of the processes in which they are used.

One important example of this recent shift is the simplification package introduced by the European Commission in February 2025. This initiative postponed or reduced the scope of several sustainability-related regulations, including reporting obligations under the Corporate Sustainability Reporting Directive (CSRD), and narrowed the number of companies required to comply. ¹⁰⁶ These changes were presented as necessary to reduce the administrative burden on businesses, but they also reflect increasing tension between environmental objectives and economic priorities. As reported by the Financial Times, this political retreat – driven in part by pressure from certain EU member states – may slow down the adoption of green technologies and weaken long-term investment signals. ¹⁰⁷ For researchers working with ionic liquids, these developments suggest that regulatory support for sustainability can no longer be taken for granted and must be continuously reinforced by demonstrating robust, quantifiable environmental performance.

The modern vision of sustainability places two fundamental demands on the chemical community, whether focused on fundamental research or on more applied and engineering-oriented work. The first is the recognition that sustainability arises from the intersection of environmental, techno-economic, and social dimensions, ¹⁰⁸ as illustrated in Fig. 23. The second is that sustainability must be supported by quantitative descriptors, which compels all those working with ILs – whether in the lab or in industrial development – to engage with tools such as life cycle assessment and green metrics. ¹⁰⁹

Fig. 23:

The boundaries of sustainable chemistry: techno-economic, environmental and social sustainability. Reprinted with permission from ref. 104]. Copyright © 2022 John Wiley and Sons.

This Copernican revolution requires a strongly multidisciplinary approach, both from individual researchers and from research teams. A paradigmatic example helps illustrate this point: when evaluating the greenness of a process based on ILs, data on toxicity and biodegradability alone are no longer sufficient to certify it as sustainable. The accidental release of ILs into the environment can disrupt ecosystems, leading to consequences that are not only economic but also social – reaffirming the threefold nature of sustainability described above. Therefore, the design of an IL-based process must incorporate eco-toxicological tests and models, ¹¹⁰ which have traditionally been distant from core chemical research, in order to assess not only the effects on individual organisms but also the broader consequences for entire ecosystems.