Everything You Wanted to Know about Deep Eutectic Solvents but Were Afraid to Be Told

Are deep eutectic solvents (DESs) a promising alternative to conventional solvents? Perhaps, but their development is hindered by a plethora of mis-conceptions. These are carefully analyzed here, beginning with the very meaning of DESs, which has strayed far beyond its original scope of eutectic mixtures of Lewis or Brønsted acids and bases. Instead, a definition that is grounded on thermodynamic principles and distinguishes between eutectic and deep eutectic is encouraged, and the types of precursors that can be used to prepare DESs are reviewed. Landmark works surrounding the sustainability, stability, toxicity, and biodegradability of these solvents are also discussed, revealing piling evidence that numerous DESs reported thus far, particularly those that are choline based,lack


WHAT'S IN A NAME?
"What's in a name?That which we call a rose by any other name would smell just as sweet," would muse Juliet.Are names and definitions important?We, as a society, believe they are.Why would the use of some words be otherwise forbidden or eschewed to describe people, acts, or even things?Classification is at the core of a scientific tradition that goes back to the Greek philosophers and their attempt to categorize and rationalize nature.It is important to distinguish a rose, not only for its smell but to know how to treat it properly and what can one expect from it.To a layperson, a wild rose with its five plain petals can easily be mistaken for a cherry flower if the two are removed from the plant.This is not surprising given that both plants belong to the Rosaceae family.But one would find trouble by growing a wild rose while expecting to gather cherries.
The correct classification of a living being, mineral, or chemical should be based on unique attributes, usefulness, and/or applicability, so that the practitioner, regardless of skill, can make an informed decision and know what to expect from each category presented to them.That is why the definition of a concept such as deep eutectic solvent (DES) is important.It must have some meaning and ought to separate this class of solvents from others by some set of distinct characteristics.This section aims to achieve such a definition while dispelling some of the most prominent myths surrounding DESs.
The first, and perhaps easiest to dispel, myth about DESs is that they are a type of ionic liquid or ionic liquid-related compound.Abbott and coworkers (1, p. 11060) clearly refute this notion, stating, "the terms DES and IL have been used interchangeably in the literature though it is necessary to point out that these are actually two different types of solvent."There are significant differences between the two families of compounds, the central one being that ionic liquids are pure compounds, whereas DESs are mixtures.The most common are actually mixtures between salts and molecular compounds (type-III), or just mixtures of two molecular compounds (type-V) (2).Although type-I DESs could be composed solely of ionic species (type-II DESs are composed of water), they remain a mixture.Even cholinium chloride (also known as choline chloride, or simply choline), the most widely used salt to prepare DESs, although composed of two ions that can be often found in ionic liquids, is a salt that does not fit any possible definition of ionic liquid.
The second myth that must be dismissed is that only DESs display freezing point depressions.To meaningfully and usefully define DESs, we must acknowledge that in any mixture of two compounds that do not form a solid solution, each of these compounds undergoes a cryoscopic depression due to the presence of the second compound, leading in the simplest cases to a eutectic point and, thus, to a large range of concentrations in which the mixture has a melting point that is lower than those of the pure compounds (3).This is true irrespective of the nature of the compounds, their melting points (above or below room temperature, i.e., whether they are solid or liquid), and the type of interactions that they may establish.A eutectic point, and often a significant cryoscopic depression, is guaranteed even if the compounds do not establish favorable interactions.An excellent example of this are apolar mixtures, such as cyclohexanol/benzene or pentane/cyclohexanol, that display severe melting temperature depressions (>50 K) and the archetypical V-shaped solid-liquid phase diagram without establishing a single hydrogen bond across components (4).Another good example are mixtures of ethanol and water, which, despite the hydrogen bond they form, present positive deviations from thermodynamic ideality leading to a minimum-boiling azeotrope.Nevertheless, the freezing point of ethanol/water mixtures is depressed significantly, as can be witnessed by leaving a bottle of any liquor in the freezer.All these systems are examples of eutectic mixtures.Is it useful to categorize them in the same group as, say, choline chloride/urea or thymol/menthol, which are examples of DESs?Of course not.Hence, "deep eutectic solvent" must be a subclass of, rather than overlap with, "eutectic solvent" (1); otherwise, it becomes a hollow definition.
The current, most widely accepted (yet apocryphal) definition of DESs presents them as a new compound formed by complexation of a hydrogen bond donor (HBD) and a hydrogen bond acceptor (HBA), with their name often being an amalgamation of their constituents (e.g., ethaline, glyceline, reline, aquoline).Most authors are content to present some sort of evidence, often Fourier-transform infrared (FTIR) spectra, showing the presence of hydrogen bonding in the system.This simplistic approach has led to the proposal of senseless mixtures of HBDs and HBAs as DESs, such as binary mixtures composed exclusively of carboxylic acids (5).These mixtures, and those of fatty alcohols, have been characterized extensively by Costa et al. (6)(7)(8) and present a liquid phase with an expectable ideal thermodynamic behavior.On the other hand, Abbott and coworkers' (1, p. 11060) original definition of DESs is not based on HBDs or HBAs but states that "DESs are systems formed from a eutectic mixture of Lewis or Brønsted acids and bases which can contain a variety of anionic and/or cationic species."However, the differences in acidity between these systems must be small; otherwise, proton transfer would take place and protic ionic liquids would form rather than DESs.If DESs must be distinct from eutectic mixtures, they cannot also be confused with protic ionic liquids.The acidity level must lead to a hydrogen bonding where a proton is shared but proton transfer does not occur.Abbott and coworkers (1, p. 11060) further add that DESs "are usually obtained by the complexation of a quaternary ammonium salt with a metal salt or hydrogen bond donor (HBD)."This concept of DESs as "a eutectic mixture of Lewis or Brønsted acids and bases" is very important because it implies that the presence of a hydrogen bond between a donor and an acceptor is not enough to yield a DES, but that its formation requires some difference in acidity between the donor and the acceptor.This is the reason mixtures of carboxylic acids, or mixtures of alcohols, cannot be deemed DESs, because in these situations there is no "mixture of Lewis or Brønsted acids and bases." That a DES is "a eutectic mixture of Lewis or Brønsted acids and bases" brings two thermodynamic notions into play: The first, based on the concept of eutectic mixture, highlights the importance of the phase diagram describing its solid-liquid equilibrium (SLE).The second, based on the idea of a "mixture of Lewis or Brønsted acids and bases," requires a degree of thermodynamic nonideality to be present in the liquid phase of the eutectic mixture.The latter is further emphasized by the qualification of "deep" to these eutectic mixtures.The enhanced decrease of the melting points of a eutectic mixture through favorable interactions between the two compounds (as illustrated in Figure 1) requires that the liquid phase present negative deviations from thermodynamic ideality that are achieved by the interaction between the Lewis or Brønsted acids and bases of different acidities.This concept and definition, first proposed by Martins et al. (9), has since found favor among many authors (10)(11)(12)(13)(14).
The definition introduced above has several advantages.It restricts DESs to a subclass of eutectic mixtures, forcing these mixtures to at least have negative deviations from thermodynamic ideality due to favorable enthalpic interactions between the two precursors that result from the sharing of a proton through some sort of hydrogen bond or any other type of strong interaction [e.g., halogen bonds (15)].This precludes the systems in which the acidity difference between the Lewis or Brønsted acids and bases is so strong that proton transfer does take place, or mixtures whose nonideality is driven by entropic factors [e.g., mixtures with polymers where the nonideality is dominated by the combinatorial part of the excess Gibbs free energy as described by a Flory-Huggins type of contribution (16)].It also precludes any sort of system that has an ideal liquidphase behavior or even positive deviations to it (5,10).Moreover, although this definition imposes no restrictions on the state of matter of the mixture, part of the usefulness of eutectic solvents results from the liquefaction of solids substances.Thus, at least one component of a DES should be a solid that has its melting point depressed through DES formation, becoming a liquid at the desired operating temperature.To summarize the discussion thus far, DESs should be defined as follows: Schematic illustration of solid-liquid phase diagrams for hypothetical mixtures of components A and B displaying thermodynamic ideal behavior (dashed black line) or negative (dashed red line), positive (dashed green line), or asymmetric (dashed blue line) deviations from thermodynamic ideality.The dotted black line is physically meaningless.Two melting point depressions for 1:1 mixtures, taking the melting point of component A as a reference, are highlighted: (i) depression due to ideal liquid-phase behavior, seen for most mixtures of common organic compounds (example of a eutectic solvent), and (ii) depression due to liquid-phase negative nonideality (example of a deep eutectic solvent).
A eutectic solvent is a eutectic-type system that is a liquid at a given desired temperature where at least one of its components would, otherwise, be a solid unfit to be applied as a solvent.A deep eutectic solvent is a eutectic solvent whose components present enthalpic-driven negative deviations from thermodynamic ideality.
The fluidity and lack of precision around the definition of DESs, and the resulting misconceptions, have not helped the development of the field.Some of these misconceptions can be traced back to Abbott and colleagues' review (1), and two must be discussed here.One relates to the definition of ideality and the temperature decrease at the eutectic point.In figure 2 of their article, Abbott and colleagues (1) claim that the straight line connecting the melting points of the two pure compounds corresponds to the ideal behavior and that the temperature decrease induced by DES formation should be T f .As discussed above and elsewhere (3,9,17), and with more rigor in the next section, when two compounds mix, their melting points naturally decrease, and the actual ideal line is not a straight line but rather possesses a distinctive V-shape (Figure 1).Thus, T f (1) is meaningless, and the correct melting point depressions associated with (a) ideal behavior (main trait of a eutectic solvent) and (b) negative thermodynamic nonideality (main trait of a DES) are clearly marked in Figure 1.The other misconception is that "the term DES refers to liquids close to the eutectic composition of the mixtures, i.e., the molar ratio of the components which gives the lowest melting point" (1, p. 11061).Conventional wisdom in the DES community holds that the eutectic point must be found at fixed, magical stoichiometric proportions of 1:2, 1:1, or 2:1 that would correspond to a special arrangement of a complex occurring in the liquid phase that is responsible for the eutectic point formation.Actually, the eutectic point, which seldom corresponds to any of those stoichiometric ratios, is a characteristic of the solid phase and its fusion process, not of the liquid phase.It results from the intersection of two melting lines that are, to a great extent, independent.Looking for the eutectic point in the liquid phase is a futile task, as Alizadeh et al. (18) showed.In spite of this, it remains common practice to prepare eutectic mixtures that are deemed DESs at several of these magical stoichiometries.Because, as Alizadeh et al. (18) showed, nothing in particular happens in the liquid phase at these molar stoichiometries, the concept of DESs should not be restricted by fixed proportions between the HBD and HBA.Instead, the user should be free to optimize the composition of the eutectic solvent to maximize its performance.Martins et al. (19) provide a good example of this, showing how the solvatochromic parameters of a mixture of terpenes forming a type-V DES vary continuously between the two pure compounds and that the optimal composition for the particular application studied is found at an intermediate concentration unrelated to the eutectic point.This approach allows DESs to be much more tunable solvents than if they are restricted to use just at the eutectic point concentration or at any arbitrary stoichiometric ratio.

HOW TO BAKE A DES
The previous section discussed the definition of DESs and how it is tied to the thermodynamic nonideality of a system.These terms are now contextualized within a rigorous thermodynamic framework to shed light not only on qualitative approaches to designing DESs but also on how their properties and phase equilibria can be modeled and predicted accurately.

Thermodynamic Nonideality
The SLE phase diagram of a eutectic-type system is given by ( 3) where x i and γ i are the mole fraction and activity coefficient of component i in the liquid phase, T m,i and m h i are its melting temperature and enthalpy, T is the SLE temperature, and R is the ideal gas constant.Equation 1 is independently applied to each component of the mixture.That is, each component possesses its own liquidus curve, with their intersection marking the eutectic point.
There is no abrupt change in the liquid structure of the system at this composition.It is merely the point at which the crystallization of one of the components becomes more favorable than that of the other.Note that, for the liquidus curve of a given component, the melting temperature of the system (T ) decreases as its (a) melting properties (T m,i or m h i ), (b) composition (underpinning the claim that all eutectic-type mixtures display melting temperature depressions, regardless of interaction strength), and (c) activity coefficient decrease.The latter is the quantity of interest when designing DESs.Equation 1 is the basis to study the nonideality of eutectic mixtures.For a mixture to be classified as a DES, its components must display enthalpy-driven negative deviations from ideality (γ i < 1).Once experimental melting temperatures for the system of interest are known, Equation 1 can be used to calculate the activity coefficients of the components (provided that their melting properties are available), which unequivocally establish whether a eutectic mixture can be classified as deep (9,17).Even easier, the ideal (calculated using Equation 1 and setting γ i = 1) and experimental solid-liquid phase diagrams can simply be compared, as illustrated in Figure 1; if the experimental melting temperatures are lower than those predicted by the ideal phase diagram, the system is a DES.Regretfully, most authors do not measure or report any SLE data, preventing this essential thermodynamic analysis.
In broad terms and for a hypothetical mixture of components A and B, negative deviations occur when the interactions across components (ensemble of A-B interactions) are stronger than the interactions of the components with themselves (A-A and B-B interactions).Illustrative solidliquid phase diagrams for different relative strengths of A-B, A-A, and A-B interactions are provided in Figure 1.Note how eutectic temperatures are lower than the ideal case only when A-B interactions are stronger than at least one of the pure-component interactions.Thus, when designing DESs, the aim should be to select precursors that have either positive or negative excess polarities, allowing them to interact favorably with the second component (increasing the strength of A-B interactions) but preventing strong interactions with themselves (decreasing the strength of A-A or B-B interactions).Such precursors include lone HBAs (molecules with HBA sites but no HBD sites, e.g., tertiary amines, aldehydes, or ketones) and asymmetric HBDs (molecules with a particularly acidic HBD site, e.g., phenols and carboxylic acids) (2).
An excellent example of a lone HBA is the zwitterionic compound betaine.Despite having a strong HBA site with a formal negative charge, its positive charge is shielded by methyl groups, preventing strong ion-ion interactions with itself.As such, betaine can form DESs with myriad different HBDs (20)(21)(22).In fact, Abranches et al. (20) demonstrated that the melting temperature depression of betaine-based DESs is well correlated against the excess positive polarity of the HBD used, reinforcing polarity asymmetries across eutectic mixture precursors as the key to achieve the desired negative nonideality.This correlation is depicted in Figure 2, along with the sigma surfaces and profiles of betaine, sorbitol, and thymol.Note that sigma surfaces and sigma profiles (which are concepts deeply connected to the COSMO-RS thermodynamics model) are molecular descriptors that fully describe the polarity of molecules and are excellent visual guides to anticipate their potential for intermolecular interactions, as thoroughly explained by Klamt (23).The sigma profiles of betaine and thymol (see 24 for an explanation regarding the polarity asymmetry of thymol) reveal excess negative and positive polarities, respectively, hence betaine/thymol being the mixture with the greatest melting point depression.On the other hand, despite its multiple HBD sites (sorbitol possesses far more positive polarity than thymol), sorbitol was the component least able to depress the melting temperature of betaine.
Given the previous paragraphs, it is no surprise that quaternary ammonium (and, to a lesser extent, quaternary phosphonium) salts behave as lone HBAs and are the main components of most studied DESs in the literature, particularly those of types I-III (1,25).Because quaternary ammonium cations often possess bulky alkyl chains that shield their positive charge, cation-anion interactions are hindered.When these salts are mixed with HBDs, a strong interaction is established between the anion of the quaternary ammonium salt (typically chloride) and the HBD, which in many cases leads to negative deviations from ideality.The strength of this interaction is related not only to the nature of the HBD but also to the availability of the anion.In other words, the bulkier the ammonium cation, the freer the anion is to interact with the second component.In fact, even mixtures composed solely of quaternary ammonium chlorides of different sizes display this type of anion transfer, which is very characteristic of type-I and -II DESs, leading to negative deviations from ideality (26).
Although it is a quaternary ammonium, the notion that cholinium chloride (henceforth, just choline for brevity) is a good DES-forming compound must be refuted.Because the cation of choline possesses a hydroxyl group, choline already establishes a strong ion-dipole interaction between this hydroxyl group and its chloride anion (OH• • • Cl), hence behaving ideally when mixed with most other substances (27) or even displaying positive deviations from ideality (28).For instance, whereas negative nonideality is seen in DESs composed of tetraalkylammonium chlorides Sigma surfaces (a) and sigma profiles (b) of betaine (left), sorbitol (middle), and thymol (right), along with the correlation between the melting temperature of betaine/HBD (1:1) mixtures and the excess positive polarity of the HBD (20).The positive (blue, P + ), apolar (green), and negative (red, P − ) regions of the sigma profiles are highlighted.Note the difficulty of finding HBDs with actual excess positive polarity (most HBD sites are also HBA sites, e.g., alcohols, carboxylic acids, and primary amines).Abbreviations: HBA, hydrogen bond acceptor; HBD, hydrogen bond donor.and fatty acids (29), mixtures of choline chloride with fatty acids or fatty alcohols all lead to positive deviations from ideality or near-ideal behavior (30).López-Porfiri et al. (31) measured the molar enthalpies of mixing of choline chloride or tetrabutylammonium chloride and ethylene glycol or glycerol and found all values to be positive.Not only does this hint at positive (rather than negative) deviations from ideality, but the enthalpy values for choline-based systems were almost double those of tetrabutylammonium-based systems.
Choline chloride behaves ideally even in the prototypical choline/urea DES (32).Despite the "alphabet soup" (33) of interactions in this system, most authors (32)(33)(34)(35)(36)(37)(38) agree that the main choline-urea interaction occurs between the amino groups of urea and the chloride anion (NH 2 • • • Cl).However, because the strength of this interaction is similar to that of OH• • • Cl, choline chloride behaves ideally.The classification of choline/urea as a DES is retained, albeit in extremis, only due to the negative deviations of urea transpiring from the NH 2 • • • Cl interaction (ion-dipole), which is stronger than any other interaction that urea establishes with itself (e.g., NH 2 • • • O hydrogen bonding).In fact, the few choline-based eutectic mixtures that can be classified as deep owe their negative nonideality to the second component, being examples of the asymmetric behavior illustrated in Figure 1.Without refuting their usefulness or applicability, the sharp melting temperature depressions associated with most choline-based eutectic mixtures, particularly those richer in choline than in the second component, arise not from the nonideality of these systems but due to the unusually low enthalpy of fusion of choline, 4.3 kJ/mol (39), as revealed by a careful inspection of Equation 1.Because this is a solid-state phenomenon, poking the liquid phase of choline-based eutectic mixtures with an excessive number of experimental and computational techniques, expecting to find some sort of mystic complex or interaction, is unproductive.
Having discussed the formation of ionic DESs, we now shift focus to type-V (nonionic) DESs.We reviewed their design recently (2) and provide only a brief summary here.One of the main groups of HBDs used to prepare type-V DESs are phenolic compounds, in particular thymol (19,(40)(41)(42)(43)(44), which, due to resonance effects, possess a hydroxyl group that is an excellent HBD but a poor HBA (24).Lone HBAs used in the past include trioctylphosphine oxide (TOPO) (45)(46)(47), flavones (48), ketones (19), and tertiary amines (49).Finally, also of note is the recent work of Wu et al. (50), in which type-V DESs are cleverly designed by pairing compounds with their lone HBA homologs (i.e., molecules where the proton of the HBD site has been replaced by an alkyl chain).

Modeling DES Behavior
The most popular methods adopted to predict or describe DES properties and phase equilibria are presented in Figure 3 and can be divided into four main groups: computational simulations, excess Gibbs energy (g E ) models, equations of state (EoS), and machine learning (ML).Of those, computational simulations and g E are clearly the most widespread.Some other types of empirical models may also be found, but their number is rather small.g E models work by predicting or modeling activity coefficients, which are then coupled with Equation 1 to obtain the solid-liquid phase diagram of the mixture.Within this category, the different flavors of COSMO-based models, particularly COSMO-RS (51) and COSMO-SAC (52), are undoubtfully the most popular.Their widespread use stems from their accuracy and the fact that they are truly predictive models: No experimental data beyond the melting properties of the pure components are needed to estimate phase equilibria.They can be used not only to predict the phase diagram of eutectic mixtures (24,27,53,54) (although with lower accuracy than other models based on the correlation of experimental data) but also to predict the solubility of a third component in them (55)(56)(57) and to provide further information about the dominant interactions present in the system, leading to the design of novel solvents.However, the description of ionic species with COSMO-based models can be problematic (27,53,58), and care must be taken when applying the model to choose the right approach to properly treat them (meta-file approach, ion-pair approach, or electroneutral mixture approach).NRTL, which is used as a correlation of experimental data measured, has also been applied in the field (57).It can be used safely for process design but not for solvent design or selection.Some authors also use UNIQUAC (59) and Redlich-Kister (60) but to a much lesser extent.Surprisingly, the predictive UNIFAC model is seldom employed.
The use of g E models to describe SLE suffers from certain shortcomings associated with Equation 1.In particular, accurate melting properties for the DES precursors are required (61), and Equation 1 must be expanded when solid-solid transitions are present or cocrystals are formed between the DES precursors, as thoroughly explained by Minceva and coauthors (54,62).On the other hand, the heat capacity term that otherwise would be added to Equation 1 appears to have a negligible impact on the phase diagram of most mixtures (9,63).Curiously, choline chloride appears to be able to incorporate water in its crystal structure, though its impact on SLE is unknown (64).
Compared to g E models, the use of EoS is much less common.One of the major reasons is that most applications of DESs are found at low pressures, and thus, the use of an EoS is not required.In fact, their use in the literature is often associated with the modeling of gas solubility in DESs (65,66).Most authors seem to favor statistical associating fluid theory (SAFT)-type EoS, followed at long distance by Prigogine-Flory-Patterson, Peng-Robinson, and cubic-plus-association equations.
Despite their popularity, computational simulations are used much differently than g E models or EoS.Their main goal is usually not to describe the phase behavior or thermophysical properties of DESs but to elucidate the molecular interactions between the DES precursors, or between DES and water or target solutes that may enhance understanding of the system and improve its design.Molecular dynamics simulations are by far the most widely adopted (67), followed by density functional theory calculations (68).The use of Monte Carlo simulations seems to be lagging in this field, which according to Salehi et al. (69, p. 2) is due to "the strong intermolecular interactions, including hydrogen bonding, that result in a high viscosity of most common DESs, and may cause slow equilibration, difficult molecule insertions and inefficient sampling of phase-space." Finally, the use of ML-type methodologies in the DES field is still uncommon, perhaps owing to the scarcity of available SLE experimental data, as well as the restriction of mixtures to fixed stoichiometric ratios.Nevertheless, several excellent efforts in the literature are worthy of mention.Using extensive property data sets, Wu and coauthors have developed group contribution models for the density (70), viscosity (71), and melting temperature (72) of DESs, achieving average absolute relative deviations of 1.56%, 8.64%, and 5.67%, respectively (the classification of group contribution models as ML is debatable).Zamora et al. (73) used support vector machines and

Sustainability
The sustainability and natural origin of the precursors most widely used to prepare DESs are questionable.Although the concept of natural deep eutectic solvents (NADES) is very attractive and saleable, and may be relevant in the context of plant biology or biochemistry (75,76), it has been abused in the field of novel green solvents.For example, choline can be found in biochemical pathways and isolated from living beings, being an essential nutrient that was once, but no longer is, considered part of vitamin B complex.However, choline is a cheap compound only because it is produced in large scale by reacting hydrogen chloride with trimethylamine and ethylene oxide to be used as, e.g., chicken feed or on fracking as a clay-controlling additive.Choline is thus not a natural compound from renewable resources.The same is true of urea.Used by animals for nitrogen excretion, urea was the first biomolecule to be produced synthetically, destroying the concept of vitalism, and is nowadays produced on a large scale for use as fertilizer and raw material in chemical industries.Its major production process relies on the reaction of ammonia with carbon dioxide.The list goes on, and in fact, most natural compounds that could be used to prepare DESs are either not available on a large scale or not economically competitive with available solvents.One of the few DES precursors that could be claimed to be natural and renewable is betaine, obtained from sugar beets, but it has been seldom used in the literature.Moreover, most NADES Dai et al. (76) proposed either are not stable (mixtures of choline, sugars, or polyols with acids), as discussed below; are simply metastable (mixtures of sugars); or possess viscosities that prevent their application as solvents.
In this field, authors often make claims such as, "DESs can be prepared from single components via heating treatment, whereas ILs are normally synthesized using synthetic routes involving various reagents and solvents" (77, p. 2).This ignores that most DES precursors, even if present in nature, are rarely obtained from extraction of natural products.Instead, to be available on a large scale and at a low price, most will be an output of the petrochemical industry, with the exception of sugars, some alcohols, simple organic acids, and a few terpenes.In particular-and because DESs are based largely on quaternary ammonium salts, whose synthesis relies on ammonia-those interested in the green aspect of these compounds must remember that more than 90% of ammonia is still produced from fossil fuels through the century-old Haber-Bosch process (78).This process is highly energy intensive, accounting for 1-2% of global energy consumption, 3% of global carbon emissions, and 3-5% of natural gas consumption (79)(80)(81).Thus, it cannot be claimed that DES preparation is a simple, effortless process with an atom economy of 100%.This disregards the entire process of precursor preparation and begins the analysis at the mixing step.

Stability
Regarding the preparation and stability of DESs, the most common misconception is not just that by being a mixture, "DESs are typically easy to prepare, and the method of preparation is usually determined by personal preference, equipment available, and ability to minimize water content," but rather that the preparation method involves heating and stirring the constituents of the DES together under an inert atmosphere until a homogeneous liquid is formed.No additional solvent is needed, and no reaction in the traditional sense occurs.Consequently, no purification steps are needed, contributing to their promise as economically viable alternatives to conventional organic solvents and ILs.(25, pp. 1233-34) Sadly, the reality is far more complex than most authors admit.
That DES preparation is not a trivial affair and, depending on the approach used, chemical degradation could take place was first the object of a detailed study by Marrucho and coworkers (82).In their seminal article that is now almost a decade old, and that should have attracted far more attention and changed attitudes in the DES community, the authors compared the most common method of DES preparation by heating and stirring, described in the previous paragraph, with grinding.The results revealed the poor chemical stability of the DESs prepared via the heating method.When heating choline together with organic acids, the authors could highlight the formation of hydrogen chloride, esters, and cholinium ionic liquids.Impurity concentrations ranged between 5% and 30%.Gull et al. (83) were probably the first to report an esterification of DESs with an acid, but because the acid was phosphoric and not a precursor of the DES, they garnered little attention.
Van den Bruinhorst et al. (84) further showed that other HBAs besides choline-in their study, proline-would also lead to ester formation.They used three methods to prepare DESs (heating, grinding, and freeze drying) and observed esterification in all of them, including the formation of oligomers of the hydroxy acids.Later, in a study of DES formation between choline and dicarboxylic acids, Gontrani et al. (85) observed the decarboxylation of malic acid above 60°C and the release of CO 2 .Rodriguez Rodriguez et al. (86) reported the ultimate study of the stability of DESs with acids as HBDs.In this decisive work, the authors clearly demonstrated that DESs prepared from choline and acids are not stable, with the esterification of the hydroxyl group of choline taking place even at low or moderate temperatures and being accelerated by increasing temperature.This chemical reaction makes the long-term stability of these mixtures impossible to achieve.Despite this evidence, every month a flurry of articles using acidic, choline-based DESs are reported without discussing their inherent instability.Other works like that of Zhang et al. (87), while acknowledging previous failures, still try to develop strategies to prepare DESs made of choline chloride and organic acids, as if the esterification resulting from combining two reactive compounds in liquid phase could be prevented.
Acidic, choline-based DESs are not alone in having stability problems during their preparation.Crawford et al. (88) showed how sugar-based DESs, even prepared under gentle heating conditions, display a brown color characteristic of their degradation.In reality, unlike several authors (e.g., 75), Silva et al. (89) established the phase diagrams of binary and ternary mixtures of sugars and of sugars with choline, clearly showing that their melting points are well above room temperature.The apparently stable liquid phases observed correspond to metastable phases, and to prepare these liquid phases, the required temperatures often compromise the compounds' stability.
But the "high thermal stabilities, low volatility, low vapor pressures" (25, p. 1233) are also questionable.Notwithstanding the chemical instability and thermal degradation described in the previous paragraphs (which is more severe at high temperatures), compounds or mixtures composed solely of ions, such as ionic liquids or type-I DESs, can present vapor pressures so low as to be considered, in practical terms, nonvolatile.In contrast, mixtures containing molecular compounds, such as type-II, -III, -IV, and -V DESs, are inherently volatile, and thus thermally unstable, leading to the vaporization of at least one of the compounds of the mixture.As early as 2016, Shahbaz et al. (90) reported the vapor pressures of DESs based on mixtures of chloride and bromide salts with urea and glycerol, which given the HBDs were understandably low but far from negligible, with values at 373 K of the order of 10 Pa.Baker and coworkers (91) later repeated these measurements with an alternative technique, essentially confirming them and adding information for the vapor pressure of choline:ethylene glycol, which as expected is more volatile than urea-or glycerol-based DESs.
Studies of DES thermal stability remain scarce, probably due to the generally accepted idea that they are highly thermally stable and nonvolatile.Haz et al. (92) reported the first work with dynamic thermogravimetric analysis of DESs.The authors showed that choline:organic acids had an onset decomposition temperature between 135°C and 200°C.Subsequent studies adopted longterm isothermal stability as more reliable measurements to assess DES stability.For the same type of DES composed of choline and organic acids (citric, malic, oxalic, and glycolic), Jablonsky et al. (93) reported mass losses of ∼5% in 10 h.They reported a similar mass loss for choline:glycerol.In a definitive study in this field, Delgado-Mellado et al. ( 94) used thermogravimetric analysis-FTIR/ATR (attenuated total reflection) to again study the long-term isothermal stability of a wide range of choline-based DESs, with acids (malic, levulinic, phenylacetic and phenylpropionic), ethylene glycol, glycerol, urea, and glucose.The systems with acids show, even at 50°C, mass losses between 3% and 10% at the end of 20 h, whereas the DES based on ethylene glycol suffers losses of 8.5%.DESs based on urea and glycerol start showing mass losses above 1% at 70°C and glucose above 100°C.FTIR/ATR analysis confirms that in general the mass loss is due to HBD volatility.More recent works confirm DESs' volatility (95), including Chen et al.'s (96, p. 9493) surprising result showing that "PEG-based DESs could volatilize at room temperature and atmospheric pressure.The mass loss from long-term evaporation of PEG-based DESs at 298.15 K for 1000 min could be as high as 30.5%." Type-V DESs are, predictably, volatile to the limit that their precursors are also volatile.In one of their pioneering studies of type-V DESs, Schaeffer et al. (97) reported the vapor-liquid equilibrium phase diagram for mixtures of thymol:menthol, showing it to be, unlike the SLE phase diagram, near ideal, with vapor pressures of 0.05 MPa at temperatures between 460 K and 480 K. Data for other type-V DESs were also reported and modeled with PC-SAFT by Dietz et al. (98) and more recently by Xin et al. (99).Although long-term volatility studies have not been reported for type-V DESs, their volatility is expected to lead to considerable mass losses.They should be treated as regular organic solvents.Ironically, Potticary et al. (100) proposed the concept of volatile DESs, whereby phenol's inherent volatility is exploited to obtain specific polymorphs of paracetamol and benzamides through the evaporation of phenol from phenol/paracetamol and phenol/benzamide eutectic mixtures.

Biodegradability
DESs are generally assumed to be biodegradable because their precursors are often naturally occurring compounds.However, for one of the foundational characteristics of these solvents, which has been accepted as common sense in all articles in this field, surprisingly little information and very few studies discuss the actual biodegradability of DESs and the impact that environmental variables (e.g., water) may have.We found only four articles dealing with this topic (101)(102)(103)(104).They present preliminary results in that only a few conventional DES were studied and contradictory results were observed.
Most authors (81-83) use the closed-bottle test recommended as Organisation for Economic Co-operation and Development (OECD) guideline 301 D. Only Lapeña et al. (104) used the manometric respirometry test (OECD 301 F).Radošević et al. (103) report degradations above 60% for choline:glucose, choline:glycerol, and choline:oxalic acid, although in this case only 66% degradation was observed after 14 days.Juneidi et al. (101) presented similar results.They studied several DESs based on both choline and N,N-diethyl ethanol ammonium chloride (EAC) and found all of them to be biodegradable, although the biodegradability of EAC-based DESs was worse than that of choline-based DESs.Moreover, the DES formed by malonic acid showed just 61% degradation, which the authors attributed to the acidity of the media conferred by this DES.However, Wen et al. (102) reported less favorable results.They compared the biodegradability of DESs based on choline and cholinium acetate, finding that only choline:urea and choline:acetamide could be considered readily biodegradable.They also reported cholinium acetate-based DESs to be less biodegradable than choline-based DESs.The most recent work on this subject (104) remeasured the biodegradability of choline:urea, choline:glycerol, and choline:ethylene glycol, confirming that they can be considered readily biodegradable, as previously reported by others, although the observed degradation was significantly lower than in previous works.Moreover, this work brings another variable into play: the water content of the DES.Because DESs are in general highly hygroscopic mixtures, it is especially important to analyze their fate in the presence of water.Lapeña et al. ( 104) observed a significant impact of water on their biodegradability.Whereas the biodegradability of choline:glycerol and choline:ethylene glycol decreased significantly, that of choline:urea increased.
The available results can be summarized quickly.The more fundamental DESs seem to be readily biodegradable, whereas use of organic acids as HBDs negatively impacts biodegradability, as does replacing choline with another HBA (although only two others have been studied so far).Water seems to significantly impact biodegradability, although it is unclear how.Many aspects remain to be studied.A more extensive database would be required to create heuristics that could help design toward biodegradability.This information is available for the DES precursors, but how DES biodegradability compares with that of its precursors is unclear.Can it be estimated from that information?Or will antagonistic or synergistic phenomena take place, as is common with toxicity?If it affects the microorganisms responsible for the degradation through toxicity, such a phenomenon could be expected.And finally, the effect of environmental parameters must be considered, firstly that of water, given not only the hygroscopic character of DESs but also its presence in the environment.For one of the most touted characteristics of these solvents, it is surprisingly poorly studied.

Toxicity
Although a chemical compound's toxicity can be quantitatively defined and established, in both cytotoxic and ecotoxic terms, in vitro or in vivo, and using model cell lines or organisms, more often than not it is a matter of perception.People, including researchers, tend to consider familiar compounds as nontoxic and unfamiliar compounds as suspect.This is particularly notorious in the DES field, where most articles start by declaring DESs as nontoxic compounds, although no systematic studies support this point of view.DESs can be prepared with nonnatural compounds [e.g., imidazolium (105), quaternary ammonium (106), or quaternary phosphonium ( 107) salts], but the trend is toward using DES precursors that even when synthetically produced (e.g., choline chloride) can be found as natural compounds.This creates the false perception that, because they are present in living organisms, these organisms are prepared to cope with them, and thus they are not toxic.Giner and coworkers (77) recently reviewed the scattered information about DES toxicity, and the pattern resulting from the available data does not support this green and idyllic version of DESs.
The in vitro cytotoxic studies depend greatly on the cell lines used, and thus may seem sometimes contradictory, but in general they show that organic acids are enhancers of toxicity (pH strongly affects toxicity).They also show that synergistic effects are important, the DESs generally being more toxic than their precursors, hinting that the molar ratio of the DES precursors also significantly impacts its toxicity.The few in vivo assays reveal comparable results, with DESs showing higher toxicity than their precursors in mice and important toxicity being reported even when the compounds used are based on natural metabolites such as fructose, glucose, or choline.
The ecotoxicity studies being reported are focused essentially on the aquatic compartment along the trophic chains.Although reasonably little attention is devoted to air, because most DESs have low vapor pressure, the terrestrial compartment should also be addressed.
The most widely used models for aquatic ecotoxicity assays are Aliivibrio fischeri, Daphnia magna, and Artemia salina.Studies in the aquatic environment are much affected by the presence of water, as it affects DES structure and stability.Concerning crustaceans, although no significant toxicity was observed for D. magna, the results for A. salina are much more worrisome, disclosing important toxicity for both phosphonium-and choline-based DESs.In particular, the observation that DESs are more toxic than their precursors has been confirmed.
The results for A. fischeri are far more extensive and confirm the observations of the cytotoxicity studies.De Morais et al. (108) provided an interesting comparison between choline-organic acid DESs and the equivalent ionic liquids, showing that the DESs have a much higher toxicity due to the presence of the acid and the pH that they confer to the media.In two articles, Macário et al. (109,110) study the toxicity of DESs as mixtures instead of looking at them as pseudopure compounds and disclose both synergistic and antagonistic behaviors, thus supporting the recurrent observations and rep orts that DESs present more toxicity than do their individual precursors.
The results summarized here support Hayyan et al.'s (111, p. 2193) vision in their seminal work on DES toxicity, where they stated, "the toxicity and cytotoxicity of DESs varied depending on the structure of components.Careful usage of the terms nontoxicity and biodegradability must be considered.More investigation on this matter is required."These ideas remain valid a decade later, and progress has been slow and often scarce.A recent comprehensive review on this subject (112) shows that DES toxicity in many model cell lines or organisms cannot be dismissed, even when DESs are based on familiar natural metabolites, and that the toxicity of mixtures cannot be extrapolated from the toxicity of the pure compounds of these mixtures because synergistic effects are often present.

DES! WHAT ARE THEY GOOD FOR?
Countless reviews in the literature describe the applications of DESs (1,25,44,(113)(114)(115).In fact, quite curiously, approximately 10% of all works published in the field are review articles (856 reviews out of 8,206 articles within the Web of Science Core Collection containing the keyword "Deep Eutectic").A nonextensive list of applications includes extraction and separation (116), particularly of natural materials such as lignin (117) and phenolic compounds (118); gas capture and separation (119); solubilization of metal oxides (120); electrodeposition techniques (121); and redox flow batteries (122).Of course, because the distinction between eutectic and deep eutectic solvents is rarely made, it is difficult to draw general conclusions from these reviews.This is not to say that eutectic solvents are not useful.However, their liquid phase is markedly distinct from that of a DES; thus, the scope of their applications will also be different.
Rather than seeing eutectic and deep eutectic mixtures as solvents sharing a supernatural ability to solvate solutes, their true remarkableness stems from their ability to liquefy a compound with a target property or applicability that would otherwise be a solid and unfit for use as a liquid at a desired temperature.Such properties can include, for instance, complexation ability or therapeutic potential.This liquefaction framework, which requires a sound knowledge of SLE and the design considerations raised in the sections above, has been gaining traction in the literature and thus is the focus of this section.For example, Ünlü et al. (123) recently reviewed the feasibility of liquefying Lewis and Brønsted acids with catalytic properties via DES formation (particularly of types I and II) and their applications.Despite the lack of a proper examination of the stability and thermodynamic ideality of these mixtures, their usefulness is unquestionable.
Another example of this framework is liquefaction of compounds with extraordinary complexation abilities.For instance, Gilmore et al. (46, p. 17323) liquefied TOPO by forming a DES with phenol (note the use of a lone HBA and an asymmetric HBD); confirmed the nonideality of the mixture by measuring its phase diagram; and acknowledged that "the strategy of liquefying the active extracting agent as a eutectic liquid produced liquids that contain an intrinsically high concentration of TOPO (. ..) and renders the use of an organic (hydrocarbon) diluent redundant."Many other examples of TOPO liquefaction through DES formation have been reviewed elsewhere (2).Regarding gas capture, Wu et al. (50) designed type-V DESs for sulfur dioxide absorption by first identifying solid compounds with potential to favorably interact and capture sulfur dioxide and then liquefying them with their lone HBA homologs (i.e., molecules where the proton of the HBD site has been replaced by an alkyl chain).Similar examples exist for carbon dioxide capture; for example, Cao et al. ( 124) used an asymmetric HBD (phenol) to liquefy a tetramethylpropanediamine chloride.
Another prominent case of rational DES design and application regards extracting compounds through in situ DES formation.Shi et al. (125) demonstrated this concept by extracting FWA52, a coumarin-based compound, from aqueous solutions with the aid of parabens.Because FWA52 is a lone HBA and parabens may be phenolic HBDs, and because both are hydrophobic, a type-V DES is formed by adding a paraben to an aqueous solution of FWA52, resulting in the formation of an immiscible DES phase that can be easily recovered via mechanical extraction.Pang et al. (126) could remove phenols from model oils using choline chloride.The simple addition of the latter to the former led to the formation of a second, immiscible liquid phase containing choline chloride and the target phenol.
The liquefaction of active pharmaceutical ingredients (APIs) through DES formation has also attracted much attention (127,128).The fact that the API is formulated as a liquid mitigates shelf-life issues, such as the formation of undesirable polymorphs, and enhances pharmacokinetics, because the dissolution step of the solid API is eliminated.Wolbert et al. (129) provided a seminal example of this concept; they attempted to liquefy three model APIs (lidocaine, ibuprofen, and phenylacetic acid) using several potential excipients (thymol, vanillin, lauric acid, para-toluic acid, benzoic acid, and cinnamic acid).In a remarkably thermodynamically consistent manner, and guided by the predictive g E model UNIFAC(Do), they identified several combinations that not only yielded liquid mixtures at room temperature but also presented negative deviations from ideality.The use of API-based DESs also allows for transdermal applications (130).For example, Al-Akayleh et al. (131) attained liquefaction of the drug risperidone using fatty acids.In this case, and although the nonideality of the mixture was not studied, the drug itself appeared to be the lone HBA.
Finally, we must acknowledge the trade-off between liquefying a target compound by forming a DES or a eutectic solvent.Aiming to form DESs (rather than ideal eutectic mixtures) using asymmetrical HBDs and lone HBAs greatly increases the chances of obtaining severe melting temperature depressions and, thus, liquids at room temperature in a wide composition range.However, the nonideality of the system may decrease its ability to solvate solutes (although this has not been an impediment thus far and may even be a positive trait when desorption or back extraction is desired).Although such a study does not seem to exist for DESs, an analogous one can be found for ionic liquids.Moya et al. (132, p. 182) related the ability of ionic liquid mixtures to dissolve carbon dioxide with their thermodynamic nonideality, finding the mixtures that presented negative deviations from ideality to be the worst performing and concluding that "synergistic CO 2 solubility can be obtained by combining two ILs which present mixtures with positive deviations from ideality, i.e., those system with unfavorable IL-IL intermolecular interactions.It should be emphasized that this kind of mixture usually presents favorable transport properties, with negative viscosity deviations."Neglecting pressure effects and using the simplest Margules activity coefficient model for a ternary system, this phenomenon can be interpreted for gaseous solutes dissolving in a binary solvent mixture using (3) ln H S,M = x 1 ln H S,1 + x 2 ln H S,2 − A 12 x 1 x 2 , 2.
where subscripts S, 1, and 2 represent the solute and the two hypothetical components of the solvent mixture; H S,M , H S,1 , and H S,2 are the Henry's constants of the solute in the mixture; the pure components x 1 and x 2 represent the composition of the binary solvent mixture; and A 12 is the Margules coefficient of the binary solvent mixture.Note how negative values of A 12 , which represent negative nonideality in the binary solvent mixture, lead to an increase of H S,M and, hence, decreased solubility.

CONCLUSIONS
Although the usefulness of both eutectic and deep eutectic solvents is unquestionable, care must be taken to properly define and design them and understand their applicability scope and limitations.
To this end, this work discussed the definition of DESs, their design and preparation, their green character, and their applications.The Future Issues section highlights the main conclusions of this critical review, as they entirely overlap with our view for the field of DESs.

FUTURE ISSUES
1.The first instinct should not be to seek answers in the liquid phase: Most mixtures of organic chemicals display melting temperature depressions.To understand whether these arise from solid-or liquid-state phenomena, the SLE phase diagram of the system should be measured and its thermodynamic behavior assessed.
2. The use of asymmetric HBDs and lone HBAs as DES precursors should become commonplace in the literature, as these mixtures greatly improve the chances of obtaining negative nonideality and, thus, severe melting temperature depressions and wide composition ranges.
3. An entire design dimension is missing in the literature due to the faulty assumption that DESs are characterized by fixed stoichiometric ratios; instead, their tunable potential is boosted by optimizing the relative composition of each precursor, provided the system remains a liquid at the desired temperature.
4. Better HBAs must be explored.Not only do most choline-chloride-based eutectic mixtures present severe sustainability issues, but choline chloride also seldom forms DESs when mixed with common HBDs, as its melting temperature depression stems from its low enthalpy of fusion.This undermines the idea that choline-based eutectic solvents are mixtures containing abnormally strong hydrogen bonds.
5. Greater attention must be given to the sustainability, stability, toxicity, and biodegradability of DESs, as limited data exist on the subject; the data that do exist seem to disprove any notion of DESs being green solvents.
6. Without a rational design, DESs are just mixtures of strongly hydrogen-bonded compounds, whereas eutectic solvents are just mixtures of organic compounds without any special trait; instead, they should be designed within a liquefaction framework, in which a solid compound with a desirable property is liquefied, enabling its use as a liquid solvent.
7. Given all the points above, more physicochemical data are necessary for eutectic systems at different compositions, not only to better outline the border between eutectic and deep eutectic solvents but also to better understand how to design these solvents and to feed data-driven, machine-learning approaches.

DISCLOSURE STATEMENT
The authors are not aware of any affiliations, memberships, funding, or financial holdings that might be perceived as affecting the objectivity of this review.

Figure 3
Figure 3Number of publications in the WoS Core Collection containing the keywords "Deep Eutectic" and each of the methodologies highlighted in each column section.Abbreviations: AI, artificial intelligence; CPA, cubic-plus-association; DFT, density functional theory; GC, group contribution models; ML, machine learning; NN, neural network; PFP, Prigogine-Flory-Patterson; PR, Peng-Robinson; SAFT, statistical associating fluid theory; WoS, Web of Science.