Hydrophobic (deep) eutectic solvents: from characterization and modeling to rational design and applications Authors: Yuxuan Zhang, Kailong Zhang, Juan Feng, Jian Shi Published: 2026 ABSTRACT (Deep) eutectic solvents (DESs) have emerged as versatile designer solvents because of their simple preparation, broad compositional flexibility, and highly tunable physicochemical properties. Hydrophobic (deep) eutectic solvents (HDESs), first reported in 2015, further extend the potential of eutectic solvents to aqueous and multiphase systems by exhibiting limited affinity toward water. This feature has generated growing interest in HDESs for separation, resource recovery, pollutant removal, gas capture, and reaction processes. Despite their rapid development, HDES research remains relatively young, and fundamental questions regarding HDES definition, characterization, and rational design remain unresolved. Moreover, while most previous reviews have largely focused on specific HDES formulations or applications, the experimental and computational tools required to systematically characterize, understand, and design these solvents have received comparatively limited attention. This review addresses this gap by providing a practical framework connecting HDES fundamentals, experimental characterization, computational modeling, rational design, and applications. Key physicochemical properties, including density, viscosity, thermal stability, phase behavior, and hydrophobicity/water stability, are discussed together with the information provided by different characterization methods, their advantages, limitations, and recommended uses. Computational approaches, including quantum chemical calculations, COSMO-RS, molecular dynamics simulations, and machine learning, are further reviewed as complementary tools for elucidating molecular interactions, predicting solvent behavior, and guiding solvent design. HDES applications are discussed according to the phase configuration and physical role of the solvent, including liquid–liquid extraction, solid–liquid extraction, gas–liquid absorption, and reaction media. Finally, key challenges and future research needs are discussed to support the advancement of HDESs toward rationally designed and practically viable solvent systems. Introduction The increasing emphasis on green and sustainable chemistry has stimulated substantial interest in developing alternative solvents that can reduce reliance on conventional volatile and potentially hazardous organic solvents. (Deep) eutectic solvents (DESs) have recently emerged as an important class of designer solvents with simple preparation, low cost, broad compositional flexibility, low vapor pressure, non-flammability, high solvating capability, and highly tunable properties.1,2 DESs are generally prepared by combining two or more components, commonly described as a hydrogen-bond acceptor (HBA) and a hydrogen-bond donor (HBD), to form a eutectic mixture with a melting temperature substantially lower than those of the individual components. Their formation is commonly associated with intermolecular interactions such as hydrogen bonding, electrostatic interactions, van der Waals forces, and mixing entropy.2 Additionally, DESs can be prepared by direct mixing or mild heating without additional solvents or extensive purification, making their preparation relatively straightforward with minimal generation of process waste.3,4 Moreover, DESs can be formulated from a wide range of naturally derived or readily available compounds, including sugars, organic acids, alcohols, phenolic compounds, and choline derivatives. Systems composed primarily of naturally occurring metabolites are often referred to as natural deep eutectic solvents (NADESs) and have attracted particular interest because their constituent sources and properties may offer potential advantages in terms of biocompatibility and environmental performance.5,6 This broad chemical diversity provides considerable flexibility for tailoring DESs toward specific functions and applications. Since their initial introduction in 2004,7 DES research has expanded rapidly across separation, extraction, catalysis, biomass processing, materials synthesis, electrochemistry, and other fields.8,9 However, this rapid development has also generated continuing debate regarding what should formally be classified as a “deep” eutectic solvent. In much of the literature, mixtures exhibiting substantial melting-point depression and remaining liquid near ambient temperature have been broadly described as DESs.4,10 More rigorous thermodynamic definitions have subsequently been proposed to distinguish DESs from ordinary eutectic mixtures. For example, Martins and co-workers proposed that DESs should exhibit significant negative deviation from thermodynamic ideality in their solid–liquid equilibrium behavior.11 Nevertheless, complete solid–liquid phase diagrams are unavailable for many reported DES formulations, making rigorous classification difficult. Furthermore, eutectic mixtures that do not satisfy a strict definition of “deep” may still possess useful solvent properties and perform effectively in practical applications.12,13 Accordingly, this review considers both “deep” and related non-deep eutectic solvent systems where they are relevant to the development and application of hydrophobic eutectic solvents. Additionally, this review generally retains the DES terminology used in the original literature while recognizing that some reported systems may be more rigorously described as eutectic mixtures until their phase behavior is fully established. Most initially reported DESs are hydrophilic due to their water-soluble components and potential solvating effects of internal hydrogen bonds with water.14,15 Additionally, the interactions between water and DESs could disrupt the structure and performance of hydrophilic DESs.16 As a result, the application of DESs in certain situations with high water contents is limited. Hydrophobic (deep) eutectic solvents (HDESs) have recently emerged as an important extension of DES chemistry to address this limitation. Since their first report in 2015, HDESs have attracted increasing interest because their limited affinity toward water enables the formation of distinct solvent-rich and aqueous phases, thereby expanding the use of eutectic solvents in aqueous extraction, separation, reaction, and other multiphase processes.10,17 Their hydrophobic character is commonly achieved through the use of poorly water-soluble constituents, including long-chain quaternary ammonium or phosphonium salts, fatty acids and alcohols, terpenoids, and hydrophobic phenolic compounds. This compositional flexibility has generated a rapidly growing library of both ionic and non-ionic HDESs with widely varying physicochemical properties and application potential. Despite this progress, HDES research remains a relatively young field and several fundamental issues remain unresolved. One important challenge is the lack of a universally accepted criterion for defining and evaluating “hydrophobicity” in eutectic solvent systems.18 In the literature, the term HDES is often applied to systems that display limited macroscopic miscibility with water. However, the presence of two liquid phases does not necessarily indicate negligible mutual solubility, because water uptake into the HDES phase or transfer of individual HDES components into the aqueous phase may still occur.19,20 Further efforts are therefore needed to establish widely accepted and quantitative criteria for evaluating the hydrophobic character and water stability of HDESs. More broadly, although numerous HDES formulations and applications have been reported, the experimental and computational tools required to characterize, understand, and rationally design these solvents have received comparatively limited systematic attention. Reliable comparison among reported HDESs can be difficult when fundamental properties such as density, viscosity, thermal stability, phase behavior, and water stability are measured using different procedures or reported without sufficient methodological detail. At the same time, computational approaches are increasingly providing complementary information on intermolecular interactions, phase behavior, solute partitioning, and solvent structure that may be difficult to obtain experimentally. Several previous reviews have summarized the development and applications of HDESs, often focusing on particular areas such as extraction, separation, environmental remediation, food-related applications, or analytical chemistry.10,18,21,22 Rather than providing another application-specific survey, this review focuses on the methodological framework needed to move HDES development from empirical formulation toward systematic characterization and rational solvent design. First, the fundamental physicochemical properties of HDESs are discussed together with the experimental techniques used for their characterization, with particular attention to the information provided by different methods as well as their advantages and limitations. Second, computational approaches, including quantum chemical calculations, conductor-like screening model for real solvents (COSMO-RS), molecular dynamics (MD) simulations, and machine learning (ML), are reviewed as complementary tools for elucidating molecular interactions, predicting solvent behavior, guiding component selection, and reducing reliance on empirical trial-and-error screening. Third, instead of organizing HDES applications solely according to individual industries or target compounds, their applications are classified according to the phase configuration and physical role of the solvent, including liquid–liquid extraction, solid–liquid extraction, gas–liquid absorption, and reaction media. This framework highlights common physicochemical principles across otherwise distinct applications and may facilitate knowledge transfer between research areas while inspiring new applications of HDESs. Finally, the major challenges associated with HDES definition, characterization, modeling, sustainability, and practical implementation are discussed together with priorities for future research. Collectively, this review aims to provide a practical framework connecting HDES discovery, characterization, molecular-level understanding, rational design, and application. Fundamentals and classification of HDESs Currently, DESs are commonly classified into five categories based on the chemical nature of their compositions. Types I and II DESs are formulated using a quaternary ammonium salt like choline chloride (ChCl) combined with a metal halide such as CuCl2, or hydrated metal halides like CuCl2·2H2O in the case of type II. Contrastingly, type III DESs are formed using a quaternary ammonium salt combined with a carboxylic acid, an alcohol, or an amide as the HBD, while type IV DESs are constructed by mixing a metal chloride salt with acetamide, urea or ethylene glycol.8,23 Type V DESs, on the other hand, are a novel class of DESs that do not contain any ionic salts, making them both ion- and halide-free.24–26 These solvents are generally formed through strong hydrogen bonding interactions between the HBA and hydroxyl groups in HBD, which are phenolic compounds. Importantly, HDESs do not constitute an additional category within this composition-based classification. Rather, the term HDES is commonly used to describe DES or eutectic systems that exhibit sufficiently low affinity toward water to form a distinct solvent-rich phase when contacted with an aqueous phase. The first preparation of HDESs was reported in 2015, where HDESs were formed by combining long-chain quaternary ammonium salts with decanoic acid to extract volatile fatty acids (VFAs) from aqueous solutions.19 Since then, the exploration of novel HDESs as well as their applications in various fields has gained significant momentum (Fig. 1). Evolution of HDES development, applications, and future directions. The hydrophobic character of an HDES is primarily associated with the molecular structures and water affinities of its constituent components. Accordingly, many commonly used HDES constituents, including long-chain quaternary ammonium or phosphonium salts, fatty acids and alcohols, and hydrophobic terpenoid or phenolic compounds, possess relatively low water solubility.18 However, the relationship between component hydrophobicity and the water stability of the resulting eutectic mixture is not always straightforward. Some reported HDESs contain relatively water-soluble constituents, such as acetic acid or lactic acid, together with a hydrophobic component and can nevertheless form a biphasic system upon contact with water.17,27 In such systems, the more hydrophilic component may preferentially partition into the aqueous phase, causing changes in the composition of the solvent phase as well as potential contamination of the aqueous phase. Therefore, the observation of two macroscopic liquid phases alone does not necessarily demonstrate that an HDES is compositionally stable in water. Accordingly, macroscopic phase separation, water uptake, component leaching, and overall compositional stability may represent related but distinct aspects of HDES – water behavior. This point is particularly important because no universally accepted quantitative criterion currently defines how hydrophobic a eutectic solvent must be to qualify as an HDES. In the literature, the term is often applied to systems that display limited macroscopic miscibility with water. However, further efforts are needed to establish widely accepted and quantitative criteria for defining the hydrophobicity of HDESs. HDESs can be further divided into ionic and non-ionic systems according to whether ionic species are present in their constituent components. Ionic HDESs commonly employ long-alkyl-chain quaternary ammonium or phosphonium salts, such as tetrabutylammonium bromide ([N4444]Br), methyltrioctylammonium chloride ([N8881]Cl), methyltrioctylammonium bromide ([N8881]Br), and trihexyl(tetradecyl)phosphonium chloride ([P14,666]Cl) as the HBA.28 Long-chain fatty acids like decanoic and dodecanoic acid, as well as alcohols such as hexanol and octanol, serve as HBDs.29 Undoubtedly, many commonly reported ionic HDESs broadly correspond to type III DES systems. In contrast, non-ionic HDESs consist entirely of neutral molecular constituents and therefore broadly correspond to type V DESs. Representative systems are based on terpenoids such as menthol and thymol combined with carboxylic acids, alcohols, phenolic compounds, or other functional organic molecules.21 Compared with ionic HDESs, their components are often more readily available. Moreover, non-ionic HDESs may allow simpler solvent recovery because sufficiently volatile components can potentially be separated by evaporation or distillation, whereas the recovery process for ionic HDESs remains challenging.30 These advantages, together with their generally lower viscosity and absence of ionic salts, have stimulated increasing interest in non-ionic HDESs. HDES properties One appealing advantage of DESs is their tailorable physicochemical properties, positioning them as “designer” solvents for task-specific applications. As the inherent properties of a DES are highly dependent on the constituent components applied during its formulation, researchers can strategically adjust these components to achieve desired characteristics.31 This flexibility, combined with the expansive variety of potential component combinations, makes DESs highly versatile and capable of meeting specific application requirements. The following subsections discuss several fundamental properties that are critical to understanding the behavior and potential of HDESs. In addition, representative values of these typical properties of HDESs reported in literature are summarized in Table 1. Table 1Representative physicochemical properties of ionic and non-ionic HDESs reported in the literatureCategoryComponentsMolar ratioDensity (g cm−3)Viscosity (mPa s)Melting/phase transition (°C)Thermal degradation/evaporation (°C)Ref.Ionic[N8881]Br + decanoic acid1 : 20.942576.59.0186.0 19 [N8881]Cl + decanoic acid1 : 20.896783.4−0.1199.3 19 Tetrabutylammonium chloride ([N4444]Cl) + decanoic acid1 : 20.917265.3−12.0194.6 19 Tetraoctylammonium bromide + decanoic acid1 : 20.930636.49.0186.2 19 [N8881]Cl + dodecanoic acid1 : 20.89573.815— 32 [N4444]Br + 1-dodecanol1 : 30.92366.51.0— 32 [N4444]Br + n-hexanol1 : 40.90–0.9129.3−85.0240 32 Tetraheptylammonium chloride + ibuprofen7 : 30.891029−72.8— 33 Tetraheptylammonium chloride + oleic acid1 : 20.87244.7−41.2; −34.7; −10.4— 33 [N8881]Cl + decanoic acid1 : 10.8932121413— 34 [N8881]Cl + gemfibrozil1 : 10.94173040−70— 34 [N8881]Cl + ketoprofen1 : 10.98884717−46— 34 Benzyltriethylammonium chloride + thymol1 : 40.91230−18.0— 35 Betaine + hexafluoroisopropanol (HFIP)1 : 21.47676−39.4— 36 l-carnitine + HFIP1 : 21.503698−18.7— 36 [N8881]Cl + methylparaben1 : 10.96371088−77.7— 37 [N8881]Cl + butylparaben1 : 31.046910−65.1242 37 [N8881]Cl + isobutylparaben1 : 10.94811525−37.0— 37 [N8881]Cl + n-octyl paraben1 : 20.96281045−78.7— 37 [N8881]Cl + 2-ethylhexyl 4-hydroxybenzoate1 : 10.94221680−76.4— 37 [N4444]Cl + perfluorooctanol2 : 11.19831912— 38 [N4444]Cl + perfluorooctanol1 : 11.1152647— 38 [N4444]Cl + perfluorooctanol1 : 21.098198−1— 38 Tetrabutylphosphonium bromide (TBPBr) + dodecanoic acid1 : 10.990413.95.5207; 369 39 TBPBr + octanoic acid1 : 11.012335.91.5155; 367 39 Non-ionicMenthol + octanoic acid1 : 20.9038.631−11114 39 Thymol + octanoic acid1 : 10.9399.151−2.4139 39 Thymol + oleic acid1 : 10.91523.520.585; 182 39 Menthol + dodecanoic acid2 : 10.89424.427.1; 13.8231.5 27 Menthol + pyruvic acid1 : 20.99529.95−58.8; −6.8218.5 27 Menthol + levulinic acid1 : 10.98530.10.2— 40 Menthol + thymol1 : 10.93338.080.1119.8 41 and 42Menthol + thymol2 : 10.92367.85—105.6 41 Menthol + 1-tetradecanol2 : 10.87243.8626113.2 41 and 43Thymol + coumarin2 : 11.05031.35—117.7 41 Thymol + lidocaine2 : 10.989122.1—139.8 41 Menthol + decanoic acid1 : 20.896212.75711.7— 44 Menthol + decanoic acid1 : 10.896115.81414.4— 44 Menthol + decanoic acid2 : 10.895921.11619.7— 44 Thymol + decanoic acid2 : 10.918612.3218−20.6240 45 Thymol + decanoic acid3 : 10.923013.8926−34.7240 45 Thymol + 2,6-dimethoxyphenol (Dmp)2 : 11.04130.60−60.4196.6 46 Thymol + Dmp1 : 11.07734.91−58.2201.4 47 Thymol + Dmp1 : 21.11335.44−57.4199.8 48 Dodecanoic acid + decanoic acid1 : 20.89410.7618.0— 13 Dodecanoic acid + decanoic acid1 : 30.9017.099.0— 49 Trioctylphosphine oxide (TOPO) + phenol1 : 20.93312.38−33250 50 Representative physicochemical properties of ionic and non-ionic HDESs reported in the literature Density Density is one of the most fundamental physicochemical properties of HDESs and plays a critical role in determining their practical performance, particularly in liquid–liquid extraction processes. In HDES – water biphasic systems, density determines whether the HDES phase forms the upper or lower layer after phase separation and strongly influences the ease of phase separation and solvent recovery. An HDES density close to that of water can be disadvantageous, as it may lead to the formation of emulsions and hinder phase disengagement .28 Conversely, a greater difference in density between HDES and water generally facilitates faster and cleaner phase separation, and thus, improves process efficiency. Therefore, density should be considered an important design parameter when developing HDESs for separation applications. To ensure an effective phase separation, a density difference greater than 0.05 g mL−1 is recommended.41 The density of HDESs in the literature typically varies from approximately 0.85 to 1.5 g mL−1, with a significant portion between 0.88 and 1.1 g mL−1. HDES density is associated with the chemical structures of its constituents, their molar ratio, intermolecular interactions within the liquid network, and other factors such as temperature and impurities.21,28 In many cases, the density of a HDES tends to lie between that of its individual components, and increasing the proportion of the higher-density component leads to a corresponding increase in overall density or vice versa.37,50 However, this relationship is not always strictly linear because the formation of an HDES may involve changes in molecular packing, free volume, and interaction strength. For example, in ammonium-based HDESs with the same HBD, shorter alkyl chains in the quaternary ammonium salt generally lead to higher density,19 whereas thymol-based HDESs have shown decreasing density with decreasing acid chain length.49 These observations highlight the complexity of the relationship between HDES density and chemical structure, indicating a need for further research to fully understand this relationship. External factors such as temperature and impurities also influence HDES density. Temperature is particularly important because density commonly decreases linearly with increasing temperature.21,49 Water uptake may also affect density, especially for HDESs that partially absorb water or undergo component leaching after contact with aqueous phases. In such cases, the measured density may shift toward that of water, although this effect is often minor unless substantial water incorporation occurs.17 Experimentally, HDES density is commonly determined by measuring the mass of a known volume of solvent.51,52 The simplest approach uses a precise analytical balance combined with a volumetric flask or graduated cylinder. This method is inexpensive, rapid, and requires only basic laboratory equipment, making it suitable for preliminary screening of large numbers of HDES formulations. However, its accuracy is strongly dependent on precise volume handling. This limitation is particularly relevant for HDESs because many of them are viscous, sticky, or prone to adhering to the wall of glassware or pipette tips. Air bubbles, incomplete transfer, meniscus-reading errors, and temperature fluctuations can all introduce significant uncertainty. Therefore, although this simple mass-volume method is useful for rapid screening, it is less suitable when high-precision density data are needed. Pycnometers provide a more accurate and reproducible method for HDES density measurement.53–55 A pycnometer has a precisely calibrated volume, and the density can be calculated from the mass difference between the empty and filled pycnometer.56 Compared with ordinary volumetric glassware, pycnometers reduce volume uncertainty and are well suited for accurate measurements. They are useful for viscous HDESs because the fixed-volume design minimizes errors associated with manual volume reading. However, pycnometer measurements can be time-consuming, and therefore, less convenient for high-throughput screening. Digital densitometers are another widely used tool for determining HDES density.34,41,57 Most densitometers operate based on an oscillating U-tube principle, where the density is calculated from the change in oscillation frequency after the tube is filled with the sample.58 Compared with manual mass-volume methods, digital densitometers offer several advantages, including high precision, automated temperature control, small sample-volume requirements, and rapid measurement. These features make them particularly attractive for systematic studies of density as a function of temperature or composition (Table 2). Table 2Comparison of experimental methods for characterizing key physicochemical properties of HDESsPropertiesMethodStrengthsWeaknessesRecommended useDensitySimple mass-volume methodSimple, rapid, inexpensive; requires only basic laboratory equipmentLower precision; sensitive to volume-handling errors such as air bubbles, incomplete transfer, and meniscus readingPreliminary screening of large numbers of formulationsPycnometerHigher accuracy and reproducibility; calibrated fixed volume reduces volume uncertaintyRelatively time-consuming and less convenient for high-throughput measurementsAccurate density determinationDigital densitometerHigh precision; rapid measurement; small sample requirement; automated temperature controlRequires dedicated instrumentationSystematic density measurements as a function of temperature or compositionViscosityCapillary viscometerSimple and inexpensiveNot ideal for highly viscous samplesRoutine measurement of HDES with low- to moderate-viscosityRotational viscometerBroad measurable viscosity range; straightforward operation and sample preparationProvides less detailed control and rheological information than a rheometerRoutine viscosity measurement, particularly for relatively viscous HDESsRheometerPrecise control of shear rate, shear stress, temperature, and oscillatory conditions; provides shear-dependent flow behaviorExpensive; requires method optimizationDetailed rheological characterization and investigation of flow behaviorThermal stabilityDynamic TGA/DTGSimple; small sample requirement; rapidly provides Tonset, Tpeak, residual mass, and mass-loss stagesMeasures only mass change rather than chemical identityRapid comparison and screening of relative thermal stabilityIsothermal TGAMeasures time-dependent mass loss under prolonged exposure at a fixed temperature; more representative of practical operationTime-consuming; still does not identify evolved speciesLong-term thermal stability and solvent-loss assessment under application-relevant conditionsTG-FTIR/TG-MSReal-time analysis of evolved species; helps distinguish evaporation of intact components from decomposition productsOverlapping spectral/mass signals may complicate identification; complementary interpretation may be requiredMechanistic investigation of thermal mass-loss pathwaysPY-GC/MS/HS-GC/MS combined with TGAProvides molecular-level identification of volatile or thermally generated compounds associated with specific mass-loss regionsNeed to combine with TG to identify the target temperature regions firstDetailed identification of evaporation/decomposition products and pathwaysPhase behaviorVisual/optical observation during heating and cooling, including hot-stage microscopySimple visual screening; directly observes crystallization, melting, turbidity, phase separation, and crystal morphology (with microscopy)Generally qualitative or semiquantitativePreliminary phase-transition screening and complementary visualization of DSC eventsDSCSmall sample requirement; sensitive; detects melting, crystallization, cold crystallization, and glass transitionsSensitive to heating rate and thermal history; broad/overlapping peaks and supercooling complicate interpretationRoutine quantitative characterization of thermal transitionsModulated DSCImproves detection of weak glass transitions; separates reversing and non-reversing heat-flow contributionsLonger analysis time; more complex method development and data interpretationAmorphous, partially crystalline, or complex HDES systemsSolid–liquid phase diagramProvides the strongest evidence for eutectic behavior; identifies eutectic composition/temperature, multiple solid phases, and crystallization-prone composition rangesLabor-intensive; requires preparation and precise analysis of many compositionsRigorous determination of eutectic behavior and composition-dependent phase stabilityHydrophobicity and water stabilityVisual phase observation/contact angle measurementRapid and inexpensive; useful for initial screening; contact angle provides a semiquantitative measure of water affinityTwo phases do not prove negligible mutual solubility; contact-angle interpretation is complicated because the HDES itself is liquidPreliminary screening of macroscopic phase separation and relative water affinityKarl Fischer titrationSelective and relatively accurate quantification of water in the HDES phasePossible chemical interference; does not measure HDES-component leachingQuantitative assessment of water uptake by the HDES phaseAqueous-phase analysis: pH, TOC, GC, IC, and NMREvaluates component leaching; TOC gives total organic transfer; GC/IC/NMR can provide component-specific informationTechnique-dependent limitations: pH/TOC are nonspecific; GC requires sufficiently volatile compounds; IC is limited to ionic/ionizable species; NMR has lower sensitivity and possible signal overlapQuantitative evaluation of compositional stability and HDES-component transfer into water Comparison of experimental methods for characterizing key physicochemical properties of HDESs Viscosity Viscosity is another key physicochemical property of HDESs because it strongly affects mixing, mass transfer, phase separation, and overall process efficiency. In applications, high viscosity can hinder proper mixing of HDES with the other phase, slow the diffusion, and decrease mass transfer rates, thereby leading to reduced performances.18 Highly viscous HDESs may also complicate solvent handling, pumping, and recovery, thereby increasing the energy demand and operational cost of a process. Therefore, although strong intermolecular interactions are often beneficial for forming stable eutectic liquids and enhancing solvation capacity, excessively high viscosity remains a major practical limitation for many HDES systems. The viscosity of HDESs reported in the literature varies greatly, ranging from only a few mPa s to tens of thousands of mPa s.18 This wide range is largely due to the variation of intermolecular interactions (e.g., hydrogen bonding, electrostatic interactions, and van der Waals forces), molecular structure, ion size, void volume, and free volume present in different HDESs.2,59,60 In general, strong intermolecular interactions can restrict molecular mobility and therefore increase viscosity.2,61 This trend is especially evident when comparing ionic and non-ionic HDESs. Ionic HDESs, which usually contain long-chain quaternary ammonium or phosphonium salts, often exhibit stronger electrostatic and ionic hydrogen-bonding interactions and therefore tend to have higher viscosities.62,63 In contrast, non-ionic HDESs primarily rely on moderate neutral hydrogen bonding interactions via hydroxyl (and/or carboxyl) groups, which generally result in lower viscosity and easier handling.60,61 Temperature is one of the most important external factors affecting HDES viscosity. Unlike density, which usually changes only moderately with temperature, viscosity often decreases dramatically as temperature increases. This behavior can be attributed to the Arrhenius principle and hole theory.28,59 In many cases, the temperature dependence of viscosity can be described using Arrhenius-type or Vogel–Fulcher–Tammann relationships.39,64 Therefore, mild heating is frequently used to reduce the viscosity of HDESs during preparation or application. Alternatively, viscosity may be lowered by introducing a co-solvent, such as water or a compatible organic solvent.23 However, it should be noted that the hydrophobic nature of HDESs generally limits the amount of water that can be incorporated, while the addition of organic solvents may disrupt the intrinsic hydrogen-bonding network of the HDES and potentially increase solvent toxicity.29 For experimental measurements, viscosity is commonly obtained using capillary viscometers, rotational viscometers, or rheometers. Capillary viscometers determine viscosity from the time required for a liquid to flow through a narrow capillary under gravity or applied pressure. This method is relatively simple, inexpensive, and capable of providing accurate viscosity values for Newtonian fluids. However, capillary viscometers are less suitable for highly viscous HDESs because the flow time can become very long and difficult to reproduce. Rotational viscometers are also widely used for HDES viscosity measurement. In this method, viscosity is determined from the torque required to rotate a spindle, cylinder, or cone in the liquid. Compared with capillary viscometers, rotational viscometers are easier to operate for viscous liquids and are more suitable for routine measurements of HDESs. They can handle a broader viscosity range and require relatively straightforward sample preparation. Compared to viscometers, rheometers provide the most comprehensive tool for evaluating the flow behavior of HDESs. Similar to rotational viscometers, rheometers measure the resistance of a fluid to deformation, but they offer much greater control over shear rate, shear stress, temperature, and oscillatory conditions. Rheometers can provide shear-rate-dependent viscosity profiles at different conditions, which are valuable for understanding HDES microstructure, hydrogen-bonding networks, and process behavior.37 Thermal stability Thermal stability demonstrates the ability of HDESs to withstand high temperatures without undergoing decomposition or significant changes in their composition. This property is particularly important for applications involving heating, such as biomass processing, extraction at elevated temperature, catalytic reaction, and solvent regeneration and recycling. A thermally unstable HDES may lose one or more components during high temperature operation, resulting in changes in composition and hydrogen-bonding structure. Therefore, thermal stability should not be considered only as a safety-related parameter, but also as an indicator of solvent durability and reusability. According to the literature, HDESs exhibit a degradation temperature ranging from 95 to 267 °C, which is relatively lower compared to ionic liquids and hydrophilic DESs.18 Similar to the density and viscosity, the thermal stability of HDESs is profoundly dependent on the chemical properties of their constituents and the interaction between them. For example, ionic HDESs generally display higher thermal stability than non-ionic HDESs. It is noteworthy to mention that non-ionic HDESs, especially those based on terpenes, phenols, and fatty acids, tend to evaporate rather than decompose at elevated temperatures.21 This phenomenon is likely ascribed to the volatility of their neutral components and the weaker intermolecular interactions present in these solvents.18,28 Although this volatility may raise concerns regarding solvent loss and environmental compatibility, the vapor pressures of non-ionic HDESs are reported to be hundreds of times lower than those of conventional organic solvents,65 potentially reducing evaporative solvent losses. Moreover, controlled volatility can be advantageous in certain cases because it enables solvent recovery through evaporation or distillation, which may be more difficult for ionic HDESs.25,30 Thermogravimetric analysis (TGA) is the most widely employed technique for evaluating the thermal stability of HDESs.41,46,66–68 In a typical TGA experiment, a small amount of HDES is heated at a programmed rate under an inert atmosphere, while the sample mass is continuously recorded as a function of temperature. The resulting thermogravimetric (TG) and derivative thermogravimetric (DTG) curves provide characteristic parameters such as the onset temperature of mass loss (Tonset), the temperature at maximum mass-loss rate (Tpeak), residual mass, and the number of mass-loss stages. TGA is simple, requires only a small sample quantity, and provides rapid comparison among different HDES formulations. It is therefore suitable for screening the relative thermal stability of HDESs with different compositions. In addition, isothermal TGA can provide additional information that is more relevant to practical applications.48,69 Instead of continuously heating the sample, the HDES is held at a fixed temperature for a certain period, and the mass loss is monitored over time, providing information on time-dependent mass loss during prolonged thermal exposure. This method may better reflect the actual operating conditions of HDES-based processes. However, TGA also has important limitations. TGA measures mass change rather than chemical structure; therefore, it cannot provide insights into the weight loss mechanisms. Evolved-gas analysis techniques may be able to provide deeper insight into the mechanisms of thermal mass loss. TG-FTIR and TG-MS couple thermogravimetric analysis with real-time analysis of volatile products released during heating. These methods are particularly useful for determining whether mass loss is caused by evaporation of intact DES components or decomposition into smaller molecules.70 However, both methods may face challenges when complex mixtures produce overlapping signals, and compound identification may require careful interpretation or complementary techniques. Pyrolysis-gas chromatography-mass spectrometry (PY-GC/MS) and headspace-gas chromatography-mass spectrometry (HS-GC/MS) can be further used in combination with TG information to identify volatile or thermally generated compounds from HDESs at the temperature related to specific weight loss regions, providing molecular-level information on evaporation or decomposition pathways.46,65,71 Phase behavior HDESs can undergo physical state transitions in response to temperature variations. This solid–liquid phase behavior of an HDES determines the temperature and compositional ranges over which it remains homogeneous and liquid. This property is essential for solvent preparation, storage, transportation, and low-temperature operation. An HDES that crystallizes or undergoes phase separation during use may experience substantial changes in viscosity, mass-transfer behavior, and extraction performance. Therefore, understanding its melting, crystallization, and glass-transition behavior is important not only for confirming eutectic formation but also for evaluating its practical operating window. The melting point of an HDES, defined as the lowest temperature at which the solvent remains in the liquid state,28 is a widely used parameter to characterize its physical properties. By the general definition, HDESs should exhibit a melting point significantly lower than that of their individual components.12 For most HDESs, this melting point depression allows them to remain in a liquid state under ambient conditions.18,28 On the molecular level, this phenomenon arises from strong interactions between HBA and HBD molecules within the HDES matrix, which disrupt the regular lattice structure that would otherwise form in pure components.72,73 Hydrogen bonding, electrostatic interactions, dispersion forces, molecular-size mismatch, and conformational disorder can disrupt the crystal lattices of the pure components and stabilize the liquid phase. Mixing entropy also contributes to the reduction in melting temperature. Generally, stronger intermolecular interactions may increase system heterogeneity, leading to a more pronounced melting point depression as a macroscopic outcome.46 In other words, the phase behavior of HDESs is highly dependent on component identity and molar ratio. Changes in composition alter not only the relative quantities of the components but also molecular packing, interaction networks, and the identity of the solid phase that crystallizes. For instance, in ionic HDESs, the melting point tends to increase with the length of the tetraalkylammonium or tetraalkylphosphonium chains.23 The simplest approach for evaluating HDES phase behavior is visual observation during controlled heating or cooling. A sample can be placed in a transparent vial or capillary and monitored for the appearance or disappearance of crystals or turbidity.13,74 These methods are inexpensive, require little specialized equipment, and are useful for preliminary screening of many compositions. However, visual measurements are generally qualitative or semiquantitative and may lack accuracy. Differential scanning calorimetry (DSC) is the most widely used technique for characterizing the solid–liquid phase behavior of HDESs.49,75–77 In DSC, a small sample is subjected to a programmed heating and cooling sequence while the difference in heat flow between the sample and a reference is recorded. DSC offers several advantages over visual methods. It requires only a small quantity of samples, provides relatively high sensitivity, and can detect transitions that are not visually apparent. It can also distinguish an endothermic melting event from an exothermic crystallization event and identify a glass transition through a step change in heat capacity. Nevertheless, DSC results require careful interpretation. The temperature at the maximum of an endothermic peak is often reported as the melting point, but this peak temperature may be influenced by many factors such as heating rate, sample mass, and heat-transfer conditions. The onset temperature may be a more suitable estimate of the melting temperature. However, for broad or overlapping transitions, even the onset may be difficult to define reproducibly. Additionally, cooling scans are particularly susceptible to kinetic effects. Many HDESs remain liquid below their equilibrium liquidus temperature because nucleation and crystal growth are slow. Consequently, the crystallization temperature observed during cooling may vary substantially with cooling rate and should not be treated as an equilibrium freezing point. A formulation may also show no crystallization during cooling but exhibit an exothermic cold-crystallization peak during subsequent heating, followed by melting at a higher temperature. Multiple heating and cooling cycles are therefore recommended. Slow heating and cooling rates improve thermal resolution and reduce kinetic shifts but considerably increase measurement time. Moreover, modulated DSC can be applied for complex HDES systems.78 It applies a small periodic temperature modulation superimposed on the underlying heating program, which allows the total heat flow to be separated into reversing and non-reversing components.79 Compared with conventional DSC, modulated DSC can improve the detection of weak glass transitions and provide greater insight into amorphous or partially crystalline HDESs. Its disadvantages include longer analysis times, more complex method development, and more complicated data interpretation. Hot-stage optical microscopy is another characterization method for HDESs that enables direct observation of changes in crystal morphology during controlled heating and cooling.80 This technique can visualize crystal nucleation, growth, melting, polymorphic transformation, and macroscopic phase separation. The main advantage of hot-stage microscopy is that it provides spatial information that DSC cannot supply. For example, it can help determine whether a broad DSC peak arises from gradual melting, multiple crystalline phases, or sample inhomogeneity. However, it is less quantitative than calorimetry, and the observed behavior may be affected by the small sample thickness, surface interactions, evaporation, and thermal gradients. Consequently, hot-stage microscopy is most effective as a complementary technique rather than as the sole method for establishing phase-transition temperatures. Importantly, a rigorous evaluation of phase behavior would require measurements across a range of HBA: HBD compositions rather than analysis of only one nominal formulation. These data can be used to construct a solid–liquid phase diagram, which provides a more comprehensive understanding of phase behavior and enables identification of the true eutectic composition and temperature.17 The principal advantage of a complete phase diagram is that it provides thermodynamic evidence of eutectic behavior and distinguishes a true eutectic composition from an arbitrary room-temperature liquid mixture.11,80 It can also reveal multiple solid phases and composition ranges susceptible to crystallization.81 However, the main disadvantage is the substantial experimental effort required as numerous compositions need to be prepared and analyzed precisely. It is worth noting that some HDESs may fail to crystallize during cooling periods and instead undergo vitrification. In these cases, DSC may show a glass transition as a step in the baseline rather than a distinct crystallization or melting peak.37,46,82 Such mixtures are sometimes described as low-transition-temperature mixtures.83 However, as the glass-transition is kinetic phenomenon rather than an equilibrium and its temperature may depend on many other factors such as cooling rate, heating rate, and water content, the absence of an observed melting peak alone does not establish that the material lacks an equilibrium crystalline state; it may simply indicate that crystallization is kinetically inhibited.80,84,85 Hydrophobicity and water stability Hydrophobicity is the defining characteristic of HDESs that distinguishes HDESs from conventional hydrophilic DESs and underpins many of their applications in liquid–liquid extraction. Ideally, an HDES should remain as a stable and compositionally intact liquid phase after contact with water, with minimal uptake of water into the HDES phase and negligible dissolution of its constituents into the aqueous phase. The hydrophobicity and water stability of an HDES are largely governed by the intrinsic water solubilities, molecular structures, and relative proportions of its constituents. HDESs formed exclusively from poorly water-soluble components generally exhibit greater stability in aqueous environments. In contrast, systems containing a relatively hydrophilic acid, alcohol, or other polar constituent may undergo substantial compositional changes after contact with water.28 For example, nuclear magnetic resonance (NMR) analysis has shown that menthol–acetic acid mixtures experience considerable transfer of acetic acid into the aqueous phase, whereas menthol–dodecanoic acid systems exhibit substantially lower component leaching.20 Molecular structure also strongly affects water stability. For quaternary ammonium- and phosphonium-based HDESs, increasing the length of the hydrophobic alkyl chains generally reduces water miscibility by increasing the nonpolar character of the ions.21 Moreover, ionic HDESs may exhibit greater water solubility than their separate components, which may result from the effects of ionic hydrogen bonding.33,86 In contrast, some non-ionic HDESs display reduced solubility in water relative to their individual constituents. For example, the water solubility of TOPO and phenol decreases from 0.15 mg L−1 and 82 g L−1 to an undetectable level and 1.34 g L−1, respectively, after they form a HDES.50 Currently, there are no universally accepted criteria for assessing the hydrophobicity of HDESs. Instead, several related but distinct aspects of HDES – water behavior should be considered separately. Macroscopic phase separation refers to the formation of distinct HDES-rich and aqueous phases, whereas water uptake describes the transfer of water into the HDES-rich phase and component leaching refers to the transfer of individual HDES constituents into the aqueous phase. Compositional stability more broadly reflects the extent to which the HDES retains its original composition after equilibration with water. Various methods have been developed to evaluate these related attributes directly or indirectly, which can be broadly classified into visual and chemical approaches. The simplest method for evaluating HDES hydrophobicity is visual observation after mixing the HDES with water.19,41 The appearance of two distinct liquid layers indicates limited macroscopic miscibility, while turbidity or formation of a single phase may indicate partial or complete miscibility. This approach is inexpensive, rapid, and useful for preliminary screening of large numbers of formulations. However, visual observation provides only qualitative information. The presence of two phases does not prove negligible mutual solubility, because water uptake or component leaching may occur without eliminating the phase boundary. Contact-angle measurements are another method used to assess the hydrophobicity of HDESs visually.37,87 In a typical experiment, a water droplet is placed on an HDES layer, and the apparent contact angle is recorded. A larger angle is generally interpreted as indicating a lower affinity between water and the HDES. Compared with simple visual observation, this method provides a more quantitative parameter. However, unlike conventional contact-angle analysis, which involves a liquid droplet on a solid surface, HDESs are liquids. As a result, both the measurement and its interpretation require caution. Chemical methods assess hydrophobicity or water stability of HDESs by analyzing potential cross-contamination between the HDES and aqueous phases after mixing. Karl Fischer titration is one of the most widely used methods for quantifying water uptake by an HDES after equilibration with an aqueous phase.13,47 The principal advantages of Karl Fischer titration are its selectivity for water, relatively high accuracy, and broad applicability. It provides a direct quantitative measure of the amount of water present in the HDES phase and is therefore more informative than visual observation. However, the technique also has limitations. It may be hindered by poor dissolution of viscous HDESs and chemical interference from certain components, and it measures only water uptake rather than HDES-component leaching into the aqueous phase. The leaching of HDES components into aqueous phases can be measured by methods including pH, total organic carbon (TOC), gas chromatography (GC), ion chromatography (IC), and NMR. Changes in the pH of the aqueous phase provide a simple indirect indication of component leaching when the HDES contains acidic or basic constituents.41 For example, dissolution of an organic acid from the HDES into water may cause a measurable decrease in aqueous-phase pH.88 This approach is inexpensive, rapid, and useful for comparative screening. However, pH measurements are nonspecific and generally cannot provide accurate concentrations of individual components. Additionally, neither pH nor conductivity is useful for neutral non-ionic components. These measurements should therefore be treated as qualitative or semiquantitative indicators. TOC analysis can be used to estimate the overall amount of organic material transferred from the HDES into the aqueous phase.41,89 This method is relatively rapid, sensitive, and broadly applicable to organic HDES components. However, TOC does not identify which component has leached into the aqueous phase. GC coupled with FID or MS can be used to identify and quantify volatile or semi-volatile HDES components in either phase.46 GC-based methods are especially suitable for HDESs containing terpenes, fatty alcohols, short-chain organic acids, phenols, or other sufficiently volatile constituents, providing robust and sensitive qualitative analysis and quantification. However, nonvolatile, thermally unstable, or ionic components may not be directly amenable to GC analysis. IC provides sensitive and compound-specific quantification of ionic HDES components transferred into the aqueous phase, making it useful for assessing leaching of the salt components of HDESs.19 However, it is limited to only ionic or ionizable species and may require cautions on calibration and interference from complex samples. NMR spectroscopy is another powerful method for identifying and quantifying individual HDES components transferred into the aqueous phase (Fig. 2).20,50 It can be used for nonvolatile HDES components that are difficult to analyze by GC. Its disadvantages include relatively high instrument cost, lower sensitivity than some chromatographic methods, and possible signal overlap in complex mixtures. Despite these limitations, NMR is one of the most informative techniques for evaluating the compositional stability of HDESs. Summary matrix illustrating how the water solubility of HDES components affects HDES – water miscibility. Adapted from ref. 14 with permission from JOHN Wiley and Sons under the license number 6355521012177, copyright 2026 and ref. 20 with permission from Elsevier under the license number 6355521191810, copyright 2026. Overall, no single parameter fully describes HDES hydrophobicity or water stability. A rigorous evaluation should distinguish macroscopic phase separation from water uptake and component leaching and should combine quantitative water-content measurements with analysis of individual HDES components in the aqueous phase to assess overall compositional stability. Simulations and modeling of HDES Computational simulations and predictive modeling have become important tools in the study of HDESs. Unlike conventional organic solvents, HDESs are not single molecular liquids but complex multicomponent systems whose properties arise from non-ideal interactions between HBA and HBD. Their macroscopic properties, including density, viscosity, hydrophobicity, phase stability, and solute partitioning, are therefore controlled by a balance of molecular interactions. Simulations and modeling can reduce this experimental burden by identifying promising component combinations, explaining observed extraction mechanisms, and predicting partitioning behavior.90–92 The simulation and modeling studies of HDESs include several approaches: quantum chemical calculations, COSMO-RS and thermodynamic screening models, MD simulations, and data-driven ML modeling strategies. These approaches differ in length scale, time scale, computational cost, and predictive capability. Quantum chemical calculations provide information about local electronic interactions; COSMO-RS and thermodynamic models predict phase behavior and solute partitioning; MD simulations describe liquid structure, dynamics, and interfaces; and data-driven ML methods support high-throughput screening and process-level prediction. Quantum chemical calculations and molecular interaction analysis Quantum chemical calculations are used to examine the electronic origin of interactions between HDES components and between HDESs and target solutes. These calculations are typically performed using density functional theory (DFT) or related electronic-structure methods on representative clusters containing one or more HBA and HBD molecules. The optimized geometries, interaction energies, charge distributions, electrostatic potential maps, and orbital-related descriptors obtained from these calculations provide molecular-level information that cannot be directly accessed from macroscopic experiments.93 In the context of HDES formation, quantum chemical calculations are useful for identifying the dominant interaction motifs responsible for eutectic behavior. In classical HDESs, melting-point depression is often associated with strong hydrogen bonding between HBA and HBD. However, many HDESs do not follow this simple picture. Systems based on menthol, thymol, terpenes, fatty acids, long-chain alcohols, quaternary ammonium salts, or phosphonium salts may be stabilized not only by hydrogen bonding but also by dispersion forces, hydrophobic association, steric mismatch, and conformational disorder.94 Quantum chemical analysis can help determine whether the interaction between two components is dominated by O–H⋯O hydrogen bonding, carboxylic acid dimer disruption, halide-mediated hydrogen bonding, electrostatic attraction, or nonpolar van der Waals contacts.95 For example, in menthol- or thymol-based HDESs, the hydroxyl group can participate in hydrogen bonding, while the bulky hydrophobic skeleton promotes low water miscibility and weak crystalline packing.96,97 In fatty-acid-based systems, carboxyl groups can form strong hydrogen-bonded networks, while long alkyl chains contribute hydrophobicity and liquid-phase structuring.96,98 In quaternary ammonium- or phosphonium-based HDESs, the cation alkyl chains provide hydrophobic character, whereas the anion or counterion may act as a strong hydrogen-bond acceptor.99 Quantum chemical calculations can clarify how these polar and nonpolar contributions coexist within the same solvent system. Several descriptors are commonly extracted from quantum chemical calculations. Binding energy or interaction energy is used to estimate the relative strength of HBA–HBD or solute – HDES association. Natural bond orbital analysis, Mulliken charge, Hirshfeld charge, or related charge-partitioning methods can identify electron-donating and electron-accepting sites. Molecular electrostatic potential maps can reveal regions susceptible to hydrogen bonding, electrostatic attraction, or nucleophilic/electrophilic interaction. These complementary descriptors are illustrated in Fig. 3. The COSMO-derived surface charge-density maps and σ-profiles in Fig. 3a describe the distribution of hydrogen-bond-donating, hydrogen-bond-accepting, and comparatively nonpolar surface regions of quercetin and several candidate DES constituents. Such profiles can be used to compare the qualitative complementarity between a target solute and the HBA or HBD components considered during solvent screening.100 In contrast, the electrostatic-potential maps in Fig. 3b identify localized electron-rich and electron-deficient regions of Eriochrome Black T (EBT), water, and a trioctylamine/2-hydroxybenzoic acid DES, thereby illustrating how quantum-chemical calculations can be used to examine solute binding and competition with water.101 Representative quantum-chemical descriptors used to characterize interactions in deep eutectic solvent systems. (a) COSMO-derived three-dimensional induced surface charge-density maps and corresponding σ-profiles of quercetin, choline chloride (ChCl), glycerol, xylitol, 1,2-propanediol, and oxalic acid. The negative-σ, approximately neutral, and positive-σ regions describe predominantly hydrogen-bond-donor, nonpolar, and hydrogen-bond-acceptor surface contributions, respectively. Adapted from ref. 100 with permission from Springer Nature under the license number 6356091213956, copyright 2026. (b) Electrostatic-potential maps of EBT, water, and DES1, composed of trioctylamine and 2-hydroxybenzoic acid, showing representative electron-rich and electron-deficient regions that may participate in DES – solute and DES – water interactions. Adapted from ref. 101 with permission from Elsevier under the license number 6356100081082, copyright 2026. Quantum chemical calculations are also important for understanding solute extraction mechanisms. In liquid–liquid extraction, the target molecule is transferred from an aqueous or organic phase into the HDES phase driven by molecular interactions such as hydrogen bonding, metal coordination, and hydrophobic effects.92,102 By calculating the binding energy between the target solute and individual HDES components, researchers can identify which component contributes most strongly to solute uptake. For aromatic pollutants and dyes, interactions with phenolic, terpene, or aromatic HDES components may be important. For phenolic compounds, hydrogen bonding and π-related interactions may dominate. For metal ions, the situation is more complex because extraction often depends on speciation, ligand coordination, counterion exchange, and hydration-shell disruption. An example is the extraction of Eriochrome Black T by the amine-based DES represented in Fig. 3b. In that study, the calculated interaction energies for DES–EBT and DES – water association were −75.24 and −6.50 kcal mol−1, respectively, indicating substantially stronger association of the DES with the dye than with water. Independent-gradient-model analysis further attributed the favorable DES–EBT interaction to a combination of strong hydrogen bonding and van der Waals interactions.101 A useful role of quantum chemical calculations is therefore to support mechanistic interpretation rather than to predict bulk extraction performance directly. A strong binding energy between a solute and a DES component may indicate favorable affinity, but it does not automatically guarantee high extraction efficiency. Despite their value, quantum chemical calculations have some limitations. Most studies use small clusters, often containing only one HBA and one HBD molecule. Such clusters do not fully represent the heterogeneous and dynamic liquid environment of an HDES. The selected initial cluster geometry may also strongly influence the optimized structure and calculated interaction energy. In addition, dispersion corrections are essential for HDESs because van der Waals interactions and alkyl-chain contacts often contribute significantly to stability. COSMO-RS and thermodynamic screening models COSMO-RS is among the most widely used modeling approaches for DES research because it provides a practical route for solvent screening before experimental preparation. The method combines quantum chemical calculations with statistical thermodynamics. First, a molecule is treated in a virtual conductor environment to obtain its surface charge density distribution. Then, the surface polarity information is used to estimate intermolecular interactions and thermodynamic properties in liquid mixtures. The resulting σ-profile and σ-potential provide information about HBD ability, HBA ability, polarity, and nonpolar surface contributions.103,104 For HDESs, COSMO-RS is particularly useful because it allows researchers to screen large numbers of possible component combinations. The high-throughput capability of COSMO-RS is illustrated by the screening study of Wang et al., in which 1768 NADES formulations were computationally evaluated as potential solvents for six triazole fungicides (Fig. 4a). The predicted-solubility heat map enabled rapid comparison of a large number of HBA–HBD combinations, whereas the corresponding σ-profiles allowed the polarity distributions of the fungicides, water, and selected NADES components to be compared. In this way, COSMO-RS provided both a numerical ranking of candidate solvents and a molecular-level interpretation of the polar, hydrogen-bonding, and nonpolar contributions governing fungicide solubilization.105 Representative applications of COSMO-RS for high-throughput solvent screening and interpretation of solute – DES interactions. (a) COSMO-RS analysis of NADES formulations for triazole fungicide solubilization, including a heat map of the predicted solubilities of six triazole fungicides in 1768 NADESs, σ-profile comparisons between the fungicides and water, and σ-profile comparisons of representative fungicide – NADES component combinations. Adapted from ref. 105 with permission from American Chemical Society under the license number 1799638-1, copyright 2026. (b) COSMO-RS analysis of DESs for rosemary-antioxidant extraction, showing the σ-profiles and σ-surfaces of carnosic acid, carnosol, and water, together with contour maps of the predicted infinite-dilution activity coefficients of carnosic acid and carnosol in equimolar DESs containing 30 wt% water at 35 °C. Adapted from ref. 104 with permission from American Chemical Society under the license number 1799629-1, copyright 2026. One major application of COSMO-RS is the prediction of solute partitioning between water and the HDES phase. In such systems, a good solvent must extract the target compound efficiently while remaining largely immiscible with water. COSMO-RS can estimate whether a solute has greater thermodynamic affinity for the HDES phase or the aqueous phase. It can also help compare different HDES compositions by ranking their predicted extraction capacities or selectivities.103 However, for HDESs, solvent selection should not be based only on target-solute partitioning. Water stability is equally important. A solvent may show excellent predicted affinity for a pollutant, dye, metal complex, or natural compound, but if one of the HDES components dissolves significantly in water, the solvent may lose its composition and extraction performance. Therefore, COSMO-RS studies of HDESs should ideally consider three partitioning processes simultaneously: the partitioning of the target solute into the HDES, the uptake of water into the HDES, and the leaching of individual HDES components into the aqueous phase. This is particularly important for systems containing moderately polar alcohols, organic acids, phenols, or short-chain components, which may not be fully water-immiscible.104 The influence of solvent composition and water content is further illustrated by the COSMO-RS design of DESs for extracting carnosic acid and carnosol from rosemary (Fig. 4b). Wojeicchowski et al. evaluated 1372 equimolar combinations formed from 28 HBAs and 49 HBDs and predicted the activity coefficients of both antioxidants at infinite dilution and 35 °C.104 The similar σ-profiles of carnosic acid and carnosol indicated that solvent formulations favorable for one antioxidant were generally also favorable for the other. The screening results identified hydrophobic ammonium salts as promising HBAs, particularly when combined with fatty acids, aromatic carboxylic acids, or alcohols as HBDs. When 30 wt% water was included, the predicted activity coefficients increased, indicating reduced thermodynamic solubility of the hydrophobic antioxidants, although the general ranking of the most promising DES components was largely retained.104 COSMO-RS can also be used to rationalize hydrophobicity trends. Increasing alkyl-chain length usually decreases water solubility and enhances hydrophobic character, but it may also increase viscosity and reduce mass transfer. Introducing aromatic or terpene-based components may improve extraction of aromatic pollutants due to favorable nonpolar and π-related interactions. Adding strongly hydrogen-bonding components may enhance solute affinity but may also increase internal cohesion and affect transport properties. Such predictions are useful for identifying composition – property relationships and narrowing the search space for experimental studies.105–107 Thermodynamic models other than COSMO-RS are also relevant. Activity-coefficient models, equations of state, group-contribution methods, and association-based models can be used to describe phase equilibria, density, viscosity, and solubility.91,108,109 In process design, these models are needed to simulate extraction columns, solvent regeneration, gas absorption, or separation processes. However, applying conventional thermodynamic models to HDESs is challenging because DESs are not ordinary binary mixtures. Their properties often depend on strong specific interactions, non-ideal association, composition-dependent structure, and sometimes partial dissociation or complexation. A modeling question is whether the HDES should be treated as a pseudo-component or as a mixture of individual constituents. The pseudo-component approach is convenient for process simulation because the DES is represented as a single solvent with effective properties. This simplifies calculations but can obscure the role of individual components, especially when water contact changes the DES composition through selective leaching. In contrast, a constituent-based model is more physically realistic because it treats each HBA, HBD, water molecule, and solute explicitly. However, this requires more parameters and more experimental data for validation. COSMO-RS should be viewed as a powerful screening and ranking tool rather than a definitive replacement for experiments. Its strongest use is in narrowing the number of candidate HDESs, identifying likely interaction mechanisms, and estimating trends in solubility or selectivity. Its predictions become more reliable when combined with experimental validation and molecular dynamics simulations. Molecular dynamics simulations MD simulations provide detailed information about the structure, dynamics, and interfacial behavior of HDESs at the molecular level. In contrast to quantum chemical calculations, which are usually limited to small clusters, MD simulations can represent hundreds or thousands of molecules in a liquid phase. This makes MD particularly important for understanding liquid structure, hydrogen-bond networks, nanosegregation, diffusion, viscosity, water uptake, interfacial behavior, and solute extraction mechanisms.110 Radial distribution functions (RDFs) are commonly used to identify preferred intermolecular distances and coordination environments. For example, RDFs between hydroxyl hydrogen atoms and oxygen atoms can reveal hydrogen-bonding interactions, while RDFs involving alkyl-chain carbon atoms can show hydrophobic aggregation or chain packing. Spatial distribution functions provide three-dimensional information about where one component prefers to locate around another. Coordination numbers quantify the average number of neighboring molecules within a given interaction shell. These analyses help describe the local organization of the HDES. Hydrogen-bond analysis is especially important. Many HDESs contain both polar functional groups and bulky hydrophobic regions. The polar groups may form transient hydrogen-bond networks, while nonpolar regions may form separate domains or clusters. Therefore, HDES structure can be heterogeneous, with local polar and nonpolar environments coexisting in the same liquid. This heterogeneity may influence solute selectivity. Polar solutes may associate with hydrogen-bonding sites, whereas hydrophobic solutes may preferentially occupy nonpolar domains. MD simulations are well suited to reveal this local segregation, which is difficult to infer from bulk measurements alone.111 MD simulations can also clarify the molecular basis of viscosity and diffusion. Many DESs, including HDESs, suffer from high viscosity, which can limit their use in extraction and separation because mass transfer becomes slow. In MD, self-diffusion coefficients can be obtained from mean square displacement calculations. Rotational correlation functions, residence time analysis, and hydrogen-bond lifetime analysis can provide additional information about molecular mobility. Strong and persistent hydrogen-bond networks usually reduce mobility, while bulky hydrophobic components may slow diffusion through steric hindrance and chain entanglement. However, weaker interactions or increased free volume may enhance molecular mobility. By comparing different HDES compositions, MD can help explain why some systems are more fluid than others.112 Water interaction is one of the most important HDES-specific issues that can be studied by MD. Hydrophobic DESs are often used in aqueous environments, but their stability is not guaranteed. Water may remain mostly in the aqueous phase, dissolve partially into the HDES, disrupt HBA–HBD interactions, form water clusters, or preferentially solvate one component and cause leaching. MD simulations can reveal how water molecules distribute within the HDES phase and whether they interact mainly with polar functional groups, ionic sites, carboxyl groups, hydroxyl groups, or halide anions. They can also show whether water breaks existing hydrogen bonds between HDES components or becomes incorporated into the hydrogen-bond network.113 Interfacial MD simulations are particularly valuable for HDES – water systems. In liquid–liquid extraction, mass transfer occurs across the water/HDES interface. The structure of this interface influences solute transfer, water uptake, and component loss. MD can show whether hydrophobic alkyl chains orient toward the HDES interior while polar groups face the aqueous phase, or whether certain components migrate preferentially to the interface. Such interfacial enrichment can strongly affect extraction mechanisms. For example, if the HBD accumulates at the interface, it may interact with the target solute before transfer into the bulk HDES. If one component is partially water-soluble, the interface may become a pathway for leaching and solvent degradation.114,115 A representative application of interfacial MD analysis is shown in Fig. 5 for aqueous systems containing thymol/Dmp (Thy–Dmp) and menthol/Dmp (Men–Dmp) HDESs. The initially mixed configurations at 0 ns evolved during the 500 ns simulations into spatially separated water-rich and HDES-rich domains. The corresponding one-dimensional number-density profiles quantitatively locate water and the individual HDES components across the simulation cell, thereby providing information on phase separation, water penetration, interfacial width, and component-specific distributions. Comparison of different Thy–Dmp and Men–Dmp molar ratios further demonstrates that interfacial organization is sensitive to HDES composition.115 Molecular dynamics snapshots and number – density profiles of aqueous lignin-derived hydrophobic deep eutectic solvent systems. The systems comprise (a) Thy–Dmp at a 1 : 1 molar ratio, (b) Thy–Dmp at 1 : 2, (c) Thy–Dmp at 2 : 1, (d) Men–Dmp at 1 : 1, and (e) Men–Dmp at 2 : 1. For each composition, representative simulation snapshots at 0 and 500 ns are presented above the corresponding one-dimensional number – density profiles. Blue, green, and red represent Dmp, water, and the second HDES component—Thy or Men—respectively. The final configurations and density profiles show the formation of water-rich and HDES-rich domains and reveal composition-dependent differences in their interfacial organization. Adapted from ref. 115 with permission from Elsevier under the license number 6356070191850, copyright 2026. The results in Fig. 5 also illustrate why interfacial stability cannot be inferred solely from the nominal hydrophobicity of the HDES components. Changing the HBA–HBD molar ratio alters the relative widths, overlap, and spatial positions of the component-density distributions. Consequently, simulations of only one composition may not adequately represent the range of water – HDES behavior. Comparing several compositions using quantities such as water penetration depth, density-profile overlap, interfacial thickness, and component-specific enrichment can provide a more rigorous assessment of HDES stability under aqueous extraction conditions.115 The quality of MD simulations depends strongly on force-field reliability. This is one of the major limitations in HDES modeling. Many HDES components are natural or bulky organic molecules for which standard force-field parameters may not be fully validated. Ionic HDESs may require charge scaling to avoid overestimating electrostatic interactions and viscosity. Nonpolarizable force fields may fail to capture polarization effects near ions, interfaces, or highly polar functional groups. Hydrogen-bond lifetimes, diffusion coefficients, and viscosity predictions can vary significantly depending on force-field choice. Therefore, MD simulations should be validated against experimental density, viscosity, diffusion, spectroscopy, surface tension, and phase behavior whenever possible.116 Time-scale and size limitations are also important. Many HDESs are viscous liquids with slow relaxation. Short MD simulations may not capture long-time diffusion, structural rearrangement, or phase separation. Water uptake and solute transfer across an interface may be rare events that require enhanced sampling methods. Viscosity calculations can require long trajectories and careful statistical treatment. Furthermore, conventional MD cannot easily model bond breaking, proton transfer, chemical reactions, or changes in metal speciation. These limitations do not reduce the value of MD, but they mean that MD results should be interpreted as part of a broader modeling and experimental framework.111 ML, molecular representation, and data-driven HDES design ML and data-driven modeling are emerging as important tools for accelerating the design of HDESs. The chemical space of HDESs is large because they can be prepared from many possible combinations of HBA andHBD at different molar ratios. As a result, purely experimental trial-and-error screening is inefficient and often cannot reveal the underlying structure–property relationships. ML offers a route to learn from existing experimental and computational data, predict promising HDES formulations, and guide the rational design of solvents with balanced physicochemical, separation, and sustainability properties. Data sources and molecular representation of HDESs The first requirement for reliable ML is high-quality data. In HDES research, data can be obtained from experimental measurements, computational simulations, and curated literature databases. Experimental data remain the most important source because they directly reflect real solvent behavior. Common experimental properties used for data-driven modeling include density, viscosity, melting point, thermal stability, water content, solubility, partition coefficient, extraction efficiency, recyclability, and phase behavior. For applications in liquid–liquid extraction, useful data may also include target solute concentration, phase ratio, extraction time, pH, ionic strength, temperature, and number of reuse cycles. However, data quality remains a major challenge. Published HDES datasets are often small, scattered across different application areas, and inconsistent in units, temperature ranges, component ratios, and measurement protocols. Negative results, such as mixtures that fail to form a stable liquid or show significant leaching, are rarely published. This creates a selection bias because models trained only on successful formulations may overestimate the likelihood of finding stable and effective HDESs. Future progress will therefore require curated databases that include both successful and unsuccessful formulations, standardized property reporting, uncertainty estimates, and clear information on experimental conditions. The molecular representation determines how chemical information is provided to a ML model. For HDESs, this step is more complex than for single molecules because the solvent is a mixture rather than a single compound. Several representation strategies can be used, including simplified molecular-input line-entry system (SMILES) or self-referencing embedded strings (SELFIES) embeddings, handcrafted molecular descriptors, molecular fingerprints, graph-based representations, quantum chemical descriptors, COSMO-RS descriptors, and MD-derived descriptors. A common starting point is the use of handcrafted molecular descriptors calculated from cheminformatics packages such as RDKit.117 These descriptors may include molecular weight, topological polar surface area, number of hydrogen bond donors, number of hydrogen bond acceptors, number of rotatable bonds, number of rings, aromatic atom count, heteroatom count, formal charge, molar volume-related descriptors, and calculated hydrophobicity such as log P. Molecular fingerprints provide another widely used representation. Fingerprints encode molecular substructures as bit vectors or count vectors. Common examples include Morgan fingerprints.118 SMILES-based and SELFIES-based embeddings are increasingly important in molecular ML. In these approaches, molecular structures are represented as strings and processed using sequence models such as recurrent neural networks, convolutional models, or transformer-based chemical language models. ChemBERTa is a particularly relevant example because it is a transformer-based model pretrained on large molecular SMILES datasets, allowing molecular structures to be converted into continuous latent vectors that capture substructural and chemical-context information.119 Graph-based molecular representations offer a more chemically expressive approach. In graph neural networks, atoms are represented as nodes and bonds as edges, allowing the model to learn structural features directly from molecular connectivity. For HDESs, separate molecular graphs can be constructed for the HBA, HBD, additional components, and target solute. These component embeddings can then be combined through pooling, attention, or composition-weighted fusion.120 For HDESs, these descriptors can be calculated separately for each component and then combined into a mixture-level representation. One simple approach is to concatenate the descriptor vectors of the HBA and HBD and add the molar ratio and temperature as additional features. Another approach is to calculate mole-fraction-weighted averages of component descriptors. Descriptor differences, ratios, products, maximum values, and minimum values can also be included to capture interactions between components. For example, the difference in polarity or hydrogen-bonding ability between the HBA and HBD may be more informative than the absolute value for either component alone. However, simple averaging may miss specific noncovalent interactions, local structuring, and eutectic effects, which are central to DES behavior. Table 3 summarizes a recommended dataset structure for machine-learning and data-driven HDES studies. In addition to component identities and mixture compositions, the dataset usually retains experimental conditions, unsuccessful formulations, water-contact behavior, transport information, recyclability, and uncertainty estimates. Where available, quantum-chemical, COSMO-derived, and MD-derived descriptors can be included as physically informed features, provided that the corresponding computational methods and modeling assumptions are reported consistently. Table 3Typical dataset structure for machine-learning and data-driven HDES studiesData categoryEssential data fields for each observationTypical prediction targets or derived descriptorsEssential metadata and reporting requirementsChemical identity and formulationUnambiguous HBA and HBD names; canonical SMILES or international chemical identifier (InChI); component roles; molar or mass ratio; number of componentsComponent descriptors; fingerprints; molecular embeddings; mixture-level representationReport whether ratios are molar, mass or volume based; retain the original identity of every constituent rather than only an HDES namePreparation and initial conditionPreparation method; mixing time; preparation temperature; drying procedure; initial water content; purity of each component; storage timeFormation success; phase state; initial viscosity or densityWater content should be measured rather than described only as “dry”; unsuccessful or phase-separating formulations should be retained as negative dataThermophysical propertiesDensity; dynamic or kinematic viscosity; melting/freezing point; glass-transition temperature; surface tension; conductivity; heat capacity; thermal stabilityContinuous-property prediction; temperature-dependent correlations; process-property modelsInclude units, temperature, pressure, water content, replicate number and uncertainty for every valueHydrophobicity and water stabilityWater uptake by the HDES; water solubility; aqueous concentration of each HDES component; residual HDES composition; phase-volume changePhase-stability class; mutual solubility; individual component partition coefficients; leaching riskRecord equilibration time, phase ratio, temperature, pH, ionic strength and analytical method; avoid using visual phase separation as the only labelSolute extraction and separation performanceTarget identity and structure; initial and equilibrium concentrations in both phases; extraction efficiency; distribution coefficient; selectivity; solvent-to-feed ratioSolubility, partition coefficient, extraction efficiency, capacity and selectivityRecord pH, temperature, contact time, mixing intensity, phase ratio, matrix composition and analytical uncertaintyTransport and kinetic informationExtraction time profile; diffusion coefficient; viscosity during extraction; interfacial tension; droplet size or mixing conditions where relevantMass-transfer coefficient; equilibrium time; kinetic model parametersDistinguish equilibrium performance from rate-controlled performance; use consistent sampling intervals and equilibrium criteriaRecyclability and composition driftCycle number; extraction and stripping efficiency per cycle; mass recovery; water content; viscosity; component loss; impurity accumulationPerformance-retention rate; solvent lifetime; regeneration efficiencyReport the complete trajectory over cycles rather than only the first and final valuesSafety, sustainability and feasibilityToxicity endpoints; biodegradability; persistence; volatility; component price; renewable content; regeneration energySustainability score; toxicity prediction; cost-performance objectiveDo not assign a “green” label from component origin alone; specify the endpoint, organism, test protocol, source year and uncertainty Typical dataset structure for machine-learning and data-driven HDES studies Model architectures and learning strategies The choice of model depends on the size and quality of the dataset, the type of molecular representation, and the target property. For many HDES problems, datasets are still relatively small, so classical ML models remain highly relevant. Ensemble methods such as random forest and gradient boosting are particularly useful because they can capture nonlinear relationships and rank feature importance. For example, they may identify whether alkyl-chain length, log P, hydrogen-bond donor count, polar surface area, or temperature has the strongest influence on viscosity, water uptake, or extraction efficiency. Gaussian process regression is also attractive for HDES design because it provides uncertainty estimates, which can be used in active learning and Bayesian optimization. Deep learning models become more relevant when larger datasets or pretrained molecular embeddings are available. Feedforward neural networks can be trained on descriptor vectors, while recurrent neural networks or transformer models can process SMILES or SELFIES strings. Graph neural networks and message-passing neural networks can learn directly from molecular graphs. Multi-input neural network architectures are especially suitable for HDESs because separate branches can encode the HBA, HBD, target solute, and process conditions before combining them into a shared prediction layer. However, deep learning models require careful validation because they can easily overfit small HDES datasets and may not outperform classical ML when data are limited. Active learning and Bayesian optimization are highly relevant for HDES discovery. In an active-learning workflow, an initial model is trained on available data and then used to predict the properties of a virtual library of candidate HDESs. The model also estimates uncertainty or identifies regions where new data would be most informative. Selected candidates are then tested experimentally or computationally, and the new data are added to the training set. This iterative cycle is well suited to HDES research because experimental measurements are time-consuming and the candidate space is large. Bayesian optimization can further guide the search toward formulations that satisfy multiple objectives, such as high extraction efficiency, low viscosity, low water uptake, and low component leaching. Model interpretability is also important. In solvent design, the goal is not only to make accurate predictions but also to understand why a formulation is promising. Feature importance analysis, SHapley Additive exPlanations (SHAP) values, partial dependence plots, attention weights, and sensitivity analysis can help identify the molecular features controlling HDES behavior.121 For example, interpretability tools may reveal that water stability is governed mainly by hydrophobic surface area and hydrogen-bonding capacity, while extraction efficiency depends on the match between solute polarity and HDES functional groups. Such insights can guide the rational selection of new components. Prediction targets for HDES properties and applications ML models can be trained to predict a wide range of HDES properties, but the most useful targets are those directly connected to solvent design and application performance. Physicochemical properties include density, viscosity, melting point, freezing point, thermal stability, surface tension, conductivity, and diffusion coefficient. Among these, viscosity is particularly important because high viscosity can limit mass transfer and reduce extraction efficiency, even when thermodynamic affinity for the target solute is favorable.122 Hydrophobicity and water stability should be treated as central prediction targets for HDESs. Relevant outputs include water uptake into the HDES phase, water solubility of individual components, component leaching into the aqueous phase, water/HDES partition coefficients, and long-term phase stability. These properties are often more important than simple binary miscibility. A mixture may appear hydrophobic because it forms two phases with water, but one component may still dissolve into the aqueous phase and change the composition of the remaining HDES. Therefore, ML models should be developed to predict not only whether phase separation occurs but also whether the solvent maintains its composition after water contact. For extraction and separation applications, possible targets include extraction efficiency, distribution coefficient, partition coefficient, selectivity, solute solubility, gas solubility, metal extraction efficiency, and recyclability.123 In pollutant extraction, models may include both the HDES components and the target pollutant representation. In metal extraction, models may also need to include pH, counterions, ligand identity, and metal speciation. In gas capture, pressure and temperature become important input variables. For natural product extraction, solute polarity, molecular size, hydrogen-bonding capacity, and matrix effects may need to be considered. Sustainability and safety properties are also important for practical HDES design. These include toxicity, biodegradability, environmental persistence, cost, renewability, volatility, and ease of regeneration. HDESs are often described as green solvents, but this label should not be assumed without supporting data. Some hydrophobic components may be bio-based and biodegradable, whereas others may be persistent, toxic, or expensive. ML could help integrate performance and sustainability metrics into solvent selection, provided that sufficient data are available.124,125 Taken together, experimental characterization and computational modeling provide complementary information that can be integrated into an application-driven framework for rational HDES design. As illustrated in Fig. 6, target applications first define the required solvent properties, which can then guide the selection of molecular and compositional variables and computational screening. Candidate HDES formulations are also characterized experimentally to assess these properties. The resulting HDESs are further evaluated against application-specific performance criteria, which provides feedback for iterative solvent optimization. Application-driven framework for the rational design, characterization, and validation of HDESs. HDES applications HDESs have been investigated for a broad range of separation and reaction processes. Because numerous previous reviews have provided detailed discussions of specific HDES applications,10,18,21,126–128 this section provides a representative overview of the major and emerging application areas, with emphasis on recent developments and transferable design principles rather than an exhaustive catalogue of individual studies. The separation applications are organized according to the phase configuration of the treated system, namely liquid–liquid extraction, solid–liquid extraction, and gas–liquid absorption. Applications in which HDESs serve as reaction media are discussed separately because their primary role is to facilitate chemical or biochemical transformations rather than the transfer of a target compound between phases. To facilitate comparison across these phase configurations, Table 4 summarizes the major solvent properties, molecular and compositional design variables, performance criteria, and practical limitations relevant to application-oriented HDES design. Table 4Application-oriented design considerations for HDESApplicationRequired HDES propertiesMajor design variablesKey performance criteriaPractical considerationsLiquid–liquid extractionHigh target affinity and selectivity; sufficient hydrophobicity/water immiscibility and limited component leaching; suitable density difference for phase disengagement; moderate viscosity; regenerabilityHBA/HBD identity and molar ratio; solvent hydrophobicity and polarity; functional groups for hydrogen bonding, coordination, or other specific interactionsExtraction efficiency; distribution coefficient/capacity; selectivity or separation factor; extraction kinetics; phase-separation behavior; solute recovery/stripping; solvent regeneration and recyclabilityHigh viscosity may limit mass transfer; small density differences may hinder phase separation; water uptake or component leaching may alter solvent composition; coexisting species and matrix effects may reduce selectivity; strong solute binding may complicate recovery; composition drift may occur during recyclingSolid–liquid extractionHigh target solubility and affinity; effective wetting and penetration of the solid matrix; chemical compatibility with the matrix; sufficient thermal and compositional stabilityHBA/HBD identity and molar ratio; solvent polarity; functional groups for specific interaction; thermal stabilityExtraction yield and rate; selectivity; solvent-to-solid ratio; product recovery; solvent reuseEffects of complex matrix; penetration and diffusion issue for solids; solvent retention in solids; difficult product separation and solvent regenerationGas–liquid absorption and captureHigh gas affinity and capacity; strong selectivity; low volatilityHBA/HBD identity and molar ratio; functional groups for specific interactionAbsorption capacity; selectivity; absorption/desorption kinetics; regeneration efficiency; cyclic stabilityHigh viscosity may slow gas–liquid mass transfer; strong gas-solvent interactions may hinder desorption; multicomponent-gas competition; long-term solvent reuse and stabilityReaction media and catalystsSubstrate solubility; suitable polarity; catalyst/enzyme compatibility; chemical and thermal stability; facile product separationHBA/HBD identity and molar ratio; acidity/basicity; solvent hydrophobicity and polarity; catalytic functionality; functional groups for specific interactionConversion/yield; selectivity; reaction rate; catalyst/enzyme stability; product recovery; solvent/catalyst reuseMass-transfer limitations; catalyst or enzyme deactivation; product separation and solvent regeneration; composition changes during reaction or recycling Application-oriented design considerations for HDES Liquid–liquid extraction Recovery of value-added organic compounds The recovery of value-added organic compounds from aqueous streams is an important application of HDES-based liquid–liquid extraction. Representative targets include volatile fatty acids (VFAs), bioalcohols, and other organic products derived from fermentation, biomass conversion, and food-processing streams. VFAs are short-chain monocarboxylic acids, including formic, acetic, propionic, and butyric acids, that can be produced through the fermentation of organic wastes.129 They have considerable potential as renewable platform chemicals and can serve as precursors for the production of pharmaceuticals, biodegradable polymers, flavoring agents, and other chemicals and materials.130 These compounds are often present at relatively low concentrations in complex aqueous matrices, making their selective and energy-efficient recovery challenging.131,132 Therefore, efficient recovery of VFAs from aqueous streams, particularly fermentation broths, is important for improving the economic viability of waste valorization processes. Liquid–liquid extraction is among the commonly investigated techniques for VFA recovery.131 However, conventional organic extractants may present concerns related to toxicity, volatility, and environmental impact. HDESs have consequently been explored as alternative extractants for VFA recovery. The first reported HDES application employed long-chain quaternary ammonium salts as HBAs and decanoic acid as the HBD.19 These HDESs exhibited favorable extraction of VFAs compared with the conventional extractant trioctylamine. Among the formulations evaluated, the HDES containing [N8881]Cl achieved extraction efficiencies of approximately 38% for acetic acid, 70.5% for propionic acid, and 89.8% for butyric acid. The extraction efficiency increased with increasing VFA chain length and decreasing water solubility, indicating that the hydrophobicity of the target acid strongly influences its partitioning into the HDES phase. Similar trends have been reported for other HDESs used as VFA extractants.133,134 Bioalcohols, particularly ethanol and 1-butanol, represent another important group of fermentation-derived products. These renewable fuels and platform chemicals can contribute to reduced dependence on fossil resources.135 However, their recovery from fermentation broths remains challenging because they are generally produced in dilute aqueous solutions.136 Conventional recovery commonly relies on distillation, which can be highly energy-intensive, particularly for ethanol because of its strong affinity for water and azeotrope formation.136,137 Moreover, products such as 1-butanol can inhibit fermentative microorganisms, limiting their achievable concentration in the broth and further increasing the energy demand of downstream separation.137 HDES-mediated liquid–liquid extraction may therefore provide an alternative strategy for product recovery and potentially facilitate in situ product removal during fermentation. Menthol : dodecanoic acid at a 2 : 1 molar ratio and menthol : palmitic acid at a 12 : 1 molar ratio have been investigated for extracting ethanol, 1-propanol, and 1-butanol from aqueous solutions.138,139 Similar to the trend observed for VFAs, extraction performance generally increased with increasing alcohol-chain length and decreasing water solubility. This behavior indicates that hydrophobic interactions and target partitioning play important roles in the extraction process, although specific interactions between the alcohol and HDES components may also contribute. Beyond VFAs and bioalcohols, HDESs have also been investigated for the recovery of other value-added organic compounds from aqueous processing streams, particularly phenolic antioxidants and biomass-derived platform chemicals. Phenolic compounds constitute an important group of recoverable bioactive molecules. Wastewater from wine, food, and pharmaceutical production may contain phenolic acids, flavonoids, and related compounds with antioxidant and other biological activities. In one study, 14 HDESs composed of ammonium salts, menthol, and fatty acids were screened for the recovery of phenolic compounds from aqueous gallic acid solutions, wine samples, and winery wastewater. Recoveries of 64.14–84.10% were obtained from winery wastewater, while values of 74.86–92.31% were achieved from aqueous gallic acid solutions, demonstrating the potential of HDESs for coupling wastewater treatment with antioxidant recovery.140 A similar approach has been applied to olive vegetation water, a liquid by-product of olive-oil production containing phenolic compounds such as hydroxytyrosol, tyrosol, and caffeic acid.141 Biomass-derived compounds, such as furfural, 5-hydroxymethylfurfural (5-HMF), levulinic acid, lactic acid, vanillin, and vanillic acid, represent another promising class of targets. These compounds are important intermediates for producing fuels, polymers, solvents, and other renewable chemicals,142 and their efficient removal from water may also be beneficial during biomass conversion because prolonged exposure to acidic and high-temperature conditions can promote subsequent degradation reactions.143 HDES-based liquid–liquid extraction has been shown to be a promising route for both product recovery and in situ removal from reactive aqueous systems for these compounds.144–146 Extraction and separation of valuable and critical metals The growing demand for metals used in electronics, energy-storage systems, catalysts, and other advanced technologies has stimulated increasing interest in recovering valuable and critical metals from primary and secondary resources. Hydrometallurgical processing commonly transfers metals from ores, industrial residues, spent batteries, electronic wastes, and other solid feedstocks into aqueous leachates, after which selective separation and purification are required.147,148 Liquid–liquid extraction is widely used for this purpose, but conventional processes generally rely on molecular organic diluents together with extractants and modifiers.149,150 HDESs provide an alternative extraction phase whose composition and functionality can be tailored to promote metal coordination, ion exchange, or ion-pair formation while maintaining phase separation from the aqueous feed. Precious metals were among the first high-value metals investigated for recovery using HDESs. Geng et al. developed quaternary-ammonium-bromide-based HDESs containing n-hexanoic acid for Au(iii) extraction from hydrochloric acid media. Among the investigated systems, the HDES based on [N8881]Br achieved an Au(iii) extraction efficiency of 96.8% from a 1 mM solution and exhibited favorable tolerance to salinity. Spectroscopic analysis suggested that extraction proceeded primarily through an anion-exchange mechanism. The loaded Au could subsequently be recovered using NaBH4, and the extraction capability was retained over five extraction cycles.151 More complex precious-metal separations have also been demonstrated. Liu et al. employed a lidocaine – thymol HDES for the selective separation of Au(iii), Pd(ii), and Pt(iv) from acidic chloride solutions. By combining differences in extraction behavior with acidity control and selective stripping, recovery efficiencies of 97.8%, 99.9%, and 89.5% were obtained for Au, Pd, and Pt, respectively.152 Lithium (Li) recovery represents another rapidly developing application of HDES-based liquid–liquid extraction. Li-containing process streams are particularly challenging because Li+ is often accompanied by much higher concentrations of competing alkali-metal ions.153 Chen et al. developed HDESs composed of [N4444]Cl and oleic acid for recovering Li from the mother liquor generated during Li2CO3 production. In combination with tributyl phosphate as a coordinating extractant, the [N4444]Cl: oleic acid system achieved a single-stage Li+ extraction efficiency of 76.8% and a Li/Na separation factor of 20.5 despite the high Na+ concentration in the feed. The extraction was attributed to cation exchange coupled with coordination of Li+ by the phosphoryl group of tributyl phosphate.154 HDESs are also attracting increasing attention for the separation and recovery of rare-earth elements (REEs), whose similar chemical properties make their mutual separation particularly demanding. One strategy is to incorporate established coordinating groups directly into the HDES. For example, Ni et al. designed ternary HDESs containing bis(2-ethylhexyl)amine, carboxylic acids, and 1-decanol for the separation of yttrium from heavy rare-earth elements. By varying both the carboxylic-acid component and the proportion of 1-decanol, the extraction environment and Y/heavy REE (HREE) selectivity could be systematically adjusted, illustrating the potential of multicomponent HDES design for difficult metal separations.155 TOPO-based HDESs have likewise shown strong affinity for REEs. Guidugli and Reza evaluated HDESs formed from TOPO with 1-propanol, menthol, or octanoic acid for Ce(iii) extraction from water. The TOPO:propanol system achieved approximately 93.8% Ce extraction under the investigated conditions, demonstrating that phosphine-oxide-containing HDESs can provide efficient coordination of trivalent REEs.156 More importantly for resource recycling, HDES extraction has begun to be integrated with upstream leaching and downstream metal recovery, such as the extraction of critical metals from the leachate of spent batteries, electronic wastes, and used magnets. For example, Yu et al. applied HDES solvent extraction to an aqueous leachate derived from spent NdFeB magnets and subsequently coupled the extraction step with oxalate stripping and calcination to recover rare-earth oxides.157 Other studies have demonstrated that changing an apparently “inert” HBD can substantially alter metal selectivity. In TOPO-based HDESs containing different phenolic HBDs, the extraction of Sm(iii) relative to Co(ii) could be tuned by changing the phenolic structure, suggesting that the second HDES component affects not only physical properties but also the local coordination environment around the extractive component.158 Such observations are particularly important for HDES design because they indicate that extraction performance cannot necessarily be predicted from the properties of the nominal metal-binding component alone. Recent studies have further expanded HDES-based extraction toward transition metals and complex battery-derived streams. Organophosphorus-containing HDESs are particularly attractive because the solvent components can directly participate in metal coordination or ion-exchange processes, providing opportunities to tune both extraction efficiency and selectivity. For example, di(2-ethylhexyl) phosphoric acid (DEHPA)-based HDESs have been applied to the recovery of Co from spent lithium-ion battery leachates, with extraction attributed primarily to a cation-exchange mechanism.159 Similarly, menthol-based HDESs incorporating organophosphorus extractants have been investigated for the selective extraction of Fe, Al, and Cu from spent lithium iron phosphate battery leaching solutions.160 These studies illustrate the growing use of functional HDESs for selective metal recovery from chemically complex secondary-resource streams. Representative applications of HDESs for the extraction, separation, and recovery of valuable and critical metals are summarized in Table 5, highlighting the diversity of target metals, feed streams, solvent formulations, and recovery strategies investigated to date. Table 5Representative applications of HDESs for metal extraction, separation, and recoveryTarget metal(s)Feed/sourceHDES/extraction systemRepresentative performance and design featureRecovery, regeneration, or process relevanceRef.Au(iii)HCl solutionQuaternary ammonium bromide + n-hexanoic acid; [N8881]Br-based HDES showed the best performance96.8% Au extraction; extraction mainly attributed to anion exchange; favorable tolerance to salinityAu was stripped/recovered using NaBH4; extraction performance was maintained over five cycles 151 Au(iii), Pd(ii), Pt(iv)Acidic chloride leach solution of secondary resourcesLidocaine – thymol HDESAcidity-dependent selective extraction; recovery efficiencies of 97.8% Au, 99.9% Pd, and 89.5% PtMultistep separation using acidity control and selective stripping; regenerated HDES retained extraction performance 152 Li(i)Mother liquor from Li2CO3 production[N4444]Cl–oleic acid HDES with TBPSingle-stage Li extraction of 76.8%; Li/Na separation factor of 20.5 despite high Na concentrationDemonstrates coupling of ion exchange with TBP coordination for selective Li recovery 154 Y(iii)REE containing aqueous solutionTernary HDES containing bis(2-ethylhexyl)amine, carboxylic acid, and 1-decanolComposition-dependent synergistic extraction; the HBD and 1-decanol content could be adjusted to tune Y/HREE separationDemonstrates multicomponent HDES design for difficult rare-earth separations without conventional organic diluent 155 Ce(iii)Aqueous nitrate solutionTOPO with 1-propanol, menthol, or octanoic acidTOPO-1-propanol (1 : 1) achieved 93.8% Ce extraction; substantial differences in mass-transfer coefficients were observed among HBDsIllustrates simultaneous effects of HBD identity on extraction affinity and transport behavior 156 REEs, mainly Nd/Pr/DySpent NdFeB magnet leachateTOPO-based HDES; optimized ternary decyl alcohol: tetradecyl alcohol: TOPO systemRecovery reached ∼99% with strong selectivity against Fe; final product purity >99%Oxalate stripping and calcination integrated with extraction; HDES remained stable over five reuse cycles 157 Sm(iii)/Co(ii)SmCo-magnet-recycling model systemTOPO + phenolic HBDsSm extraction varied systematically with HBD structure, demonstrating that the nominally “inert” HBD can alter selectivityShows that HBD structure can be deliberately used to tune metal separation and solvent properties 158 Co(ii)Waste lithium-ion battery leachateDEHPA + alcohol-based HDES; 1-butanol/DEHPA gave the best performanceCo extraction reached 95.64% under optimized conditions; cation-exchange mechanism supported by spectroscopy and quantum calculationsExtraction efficiency remained >80% after five cycles 159 Fe(iii), Al(iii), Cu(ii), Li(i)Spent LiFePO4 battery leachateMenthol-based HDESs incorporating TBP and/or D2EHPAFe extraction >99% over much of the investigated acidity range; Al and Cu extraction could be controlled through acidity and HDES compositionSequential extraction scheme proposed; stripping conditions were investigated for solvent regeneration and production of individual metal solutions 160 La(iii), Ce(iii)NiMH battery waste leachateTOPO-decanoic acid HDES96% La and 98% Ce extraction; high La/Ni and Ce/Ni separation factorsLanthanides were stripped with HCl and recovered as La oxalate and Ce hydroxide, demonstrating an integrated recovery route 161 Fe(iii)/Ti(iv)Fe/Ti-containing leach solutionDodecanoic acid–menthol (1 : 3) HDES used as a diluent for TBPContinuous mixer – settler extraction achieved >95% of the theoretical extraction efficiency; Fe was concentrated approximately fourfold in a closed-loop processHDES/TBP system was reused for 10 cycles, demonstrating movement toward continuous-process operation 162 Representative applications of HDESs for metal extraction, separation, and recovery Overall, these studies demonstrate the potential of HDESs for the recovery and separation of valuable metals ranging from precious and critical metals to REEs. Their major advantage lies in the ability to combine hydrophobic phase formation with tunable chemical functionality, allowing solvent properties and metal affinity to be adjusted simultaneously, although future investigations are still needed for practical hydrometallurgical applications. Removal and determination of environmental contaminants The widespread release of inorganic and organic contaminants into aquatic environments has created increasing demand for efficient separation methods for both water remediation and environmental monitoring. Depending on their sources and applications, these contaminants encompass toxic metal and metalloid species, pesticides, pharmaceuticals, and other persistent or emerging organic pollutants. Liquid–liquid extraction can remove these compounds from aqueous matrices or preconcentrate them prior to instrumental analysis. Owing to the low water miscibility, highly tunable chemical functionality, and potentially low environmental impacts, HDESs have therefore attracted considerable interest as alternative extraction media for the removal and determination of environmental contaminants.18,126 Importantly, the classification here is based on the purpose of the separation: metals with certain health and environmental concerns extracted from dilute contaminated waters for remediation or monitoring are considered environmental contaminants, whereas the recovery of valuable metals for resource recycling is discussed separately in Section 5.1.2. HDES-mediated extraction of metal and metalloid contaminants represents one of the earliest applications of these solvents in water treatment. For example, Shi et al. developed a family of HDESs based on [N8881]Cl and parabens for Cr(vi) removal.37 Extraction efficiencies exceeding 90% were obtained, while relatively high concentrations of coexisting ions and organic compounds had limited influence on the extraction performance. The removal of other heavy metals such as Cd(ii), Pb(ii), and Cu(ii) has also been widely investigated.163,164 Besides metals, studies have also focused on organic contaminants, for which hydrophobic interactions, hydrogen bonding, π-related interactions, and other specific solute–solvent interactions can promote transfer from water into the HDES phase. Early work by Florindo et al. investigated menthol-based HDESs for the extraction of several neonicotinoid pesticides, including imidacloprid, acetamiprid, nitenpyram, and thiamethoxam, with extraction efficiencies of up to approximately 80%.20 Menthol-based HDESs have similarly been applied to phenolic pollutants. Sas et al. reported extraction efficiencies above 85% for compounds such as o-cresol and 2-chlorophenol using fatty-acid/menthol eutectic mixtures.165 Other HDES formulations have been investigated for the extraction of pharmaceutical products,166,167 endocrine-disrupting compounds such as bisphenol A (BPA),168,169 micro\nano-plastics,170,171 and per- and polyfluoroalkyl substances (PFAS).172,173 Collectively, these studies illustrate the broad applicability of HDESs toward organic contaminants. In addition to direct contaminant removal, HDESs have become increasingly important as extraction and preconcentration media for analytical determination. Many environmental contaminants occur at trace or ultra-trace concentrations and therefore require enrichment before chromatographic or spectrometric analysis. HDES-based microextraction has been extensively investigated for both inorganic and organic analytes. Among these methods, dispersive liquid–liquid microextraction (DLLME) and related miniaturized extraction techniques are particularly attractive because only small volumes of HDES are required and the large interfacial area created during dispersion can accelerate analyte transfer. For inorganic analytes, HDES-based DLLME has been used for the preconcentration and determination of Se as well as trace Cd and As in aqueous or real samples.174,175 For organic-contaminant analysis, Makos et al. incorporated thymol-based HDESs into ultrasound-assisted DLLME for the enrichment and determination of polycyclic aromatic hydrocarbons (PAHs) in water,52 while Deng et al. developed HFIP-based HDESs for DLLME of pyrethroid pesticides from tea beverages and fruit juices.36 More recently, Volpatto and Vitali developed a DLLME method using a hydrophobic natural deep eutectic solvent coupled with liquid chromatography-mass spectrometry (LC-MS) for the simultaneous determination of 14 emerging organic contaminants in surface water, illustrating the transition from single-class extraction toward multiclass environmental analysis.55 Overall, HDESs offer a versatile platform for both the removal and analytical determination of environmental contaminants. Their performance can be tailored through the selection of constituent molecules and molar ratios to control hydrophobicity, hydrogen-bonding capability, ionic interactions, and affinity toward specific analytes. Solid–liquid extraction Solid–liquid extraction is another important application of HDESs, in which target compounds are transferred directly from a solid matrix into the HDES phase. Compared with liquid–liquid extraction, the performance of solid–liquid extraction depends not only on the thermodynamic affinity between the target compound and the solvent but also on solvent penetration into the solid matrix, dissolution or desorption of the target, and subsequent mass transfer from the matrix into the bulk solvent. The relatively low polarity and tunable hydrogen-bonding characteristics of HDESs make them particularly attractive for extracting hydrophobic or weakly polar compounds that are poorly recovered by aqueous-based solvents. Consequently, most early applications of HDES-mediated solid–liquid extraction have focused on lipophilic bioactive compounds from natural matrices.21 Plant-derived natural products represent one of the most established examples. For instance, Cao et al. designed a ternary HDES composed of [N8881]Cl, 1-octanol, and octanoic acid for the extraction of polyprenyl acetates from Ginkgo biloba leaves.176 Under optimized conditions, an extraction yield of approximately 84.1 mg g−1 was obtained, exceeding those achieved using conventional hydrophobic solvents such as n-hexane and petroleum ether. The same group also designed a HDES for the extraction of artemisinin directly from Artemisia annua leaves, demonstrating that solvent composition can be tailored according to the polarity and functional groups of a target phytochemical.177 HDESs can also be incorporated into more complex separation schemes to achieve selective fractionation of compounds with different polarities. For example, a two-phase system combining hydrophilic and hydrophobic DESs was developed for the simultaneous extraction and separation of multiple bioactive compounds from Ginkgo biloba leaves.178 Flavonoids, terpene trilactones, and procyanidins were preferentially enriched in the hydrophilic DES phase, whereas the more lipophilic polyprenyl acetates partitioned into the HDES phase. This approach is particularly interesting because the HDES does not merely replace a conventional organic extraction solvent but becomes part of an integrated fractionation system in which compounds with different physicochemical properties can be recovered simultaneously. The applicability of HDESs also extends beyond plant materials to other biological matrices. Khare et al. screened 39 HDES formulations for extracting ergosterol, a precursor of vitamin D2, from mushrooms.179 More recently, HDESs have also been investigated for lipid extraction from wet microalgal biomass.180 Such applications highlight the potential of HDESs for directly recovering hydrophobic products from water-rich biological feedstocks. Beyond natural products, solid–liquid HDES processes are emerging as potential tools for the recovery of valuable materials from solid wastes. One representative example is the recycling of spent lithium-ion battery cathodes. Unlike the liquid–liquid extraction of metals from pre-generated leachates, HDESs can offer the possibility of integrating metal dissolution and separation while reducing the number of processing steps. For example, in a process using reusable TOPO-based hydrophobic eutectic solvents to directly extract metal species from cathode materials, the metal components were first transferred from the solid cathode into the eutectic-solvent phase, after which an aqueous oxalic acid treatment was used to separate Li from Co, Ni, and Mn while facilitating recovery and reuse of the solvent.181 In this respect, HDES-based solid–liquid extraction can potentially combine conventional leaching and solvent-extraction functions and reduce reliance on separate aqueous leaching and organic-solvent extraction stages. HDESs have also been increasingly explored for the extraction and determination of contaminants from complex solid and semisolid matrices, including soils, sludge, and food products.182–184 In these applications, HDESs can serve as alternative extraction media for isolating and enriching trace organic contaminants prior to instrumental analysis, extending HDES-based sample preparation from aqueous samples to complex solid or semisolid environmental matrices. Gas–liquid absorption and capture Gas–liquid absorption represents an emerging application of HDESs for the capture of major gaseous components and trace contaminants from industrial gas streams. CO2 capture represents one of the earliest and most extensively investigated gas-separation applications of HDESs.10 Zubeir et al. demonstrated appreciable CO2 solubility in decanoic-acid-based HDESs, supporting their potential as physical absorbents.185 Functionalization can further enhance CO2 affinity. For example, Gu et al. developed polyamine hydrochloride – thymol HDESs with high and reversible CO2 uptake.53 Computational tools are also increasingly being used to accelerate solvent selection; Lemaoui et al. applied ML to predict CO2 capture performance and screen more than 1000 DES combinations.186 Beyond CO2, menthol-based eutectic solvents have recently been investigated for both CO2 and H2S absorption, suggesting potential applications in natural-gas and biogas sweetening.187 HDESs have also shown considerable promise for volatile organic compound (VOC) capture. Fatty-acid-based HDESs were investigated for low-concentration toluene absorption, with longer hydrocarbon chains generally enhancing VOC affinity through stronger hydrophobic interactions.188 Interestingly, HDESs are not limited to hydrophobic VOCs. Thymol-based HDESs were subsequently shown to effectively absorb acetone, where hydrogen bonding between acetone and the solvent components contributed substantially to absorption.189 These studies illustrate that VOC affinity can be tailored through a combination of hydrophobic domains and specific polar functional groups. More recent work has extended HDES-based capture to highly nonpolar compounds such as n-hexane,190 further demonstrating the broad range of VOCs potentially accessible through solvent design. The ability of HDESs to capture organic vapors is particularly relevant to biogas purification, where several trace contaminants can interfere with downstream utilization. Monoterpene-carboxylic-acid eutectic solvents have been evaluated for the removal of benzene, toluene, ethylbenzene, and xylenes (BTEX), with eucalyptol – octanoic acid showing favorable absorption performance.191 Hydrophobic fatty-acid-based DESs have also been investigated for representative biogas contaminants including toluene, limonene, and volatile siloxanes.192 These examples extend the role of HDESs from bulk-gas capture toward the selective removal of trace contaminants from complex gas mixtures. Despite these promising results, HDES-based gas purification remains at an early stage. Future studies are needed to investigate dynamic and multicomponent gas systems under realistic operating conditions, while integrating experimental measurements with COSMO-RS, molecular simulations, and ML tools to support rational HDES design. HDESs as reaction media and catalysts Beyond separation processes, HDESs are increasingly being explored as alternative reaction media. Their low volatility, tunable polarity, and ability to dissolve hydrophobic substrates can provide advantages over aqueous or conventional organic media, particularly for reactions involving poorly water-soluble compounds. Moreover, HDES components can participate directly in a reaction or provide catalytic functionality, allowing the solvent to serve multiple roles within the same process. Biocatalysis represents one promising application of HDESs as reaction media. Hümmer et al. employed menthol–fatty-acid HDESs for lipase-catalyzed esterification, where the HDES simultaneously served as the reaction medium and substrate pool.193 Addition of a small amount of water further improved conversion, highlighting the importance of controlling water content for maintaining enzyme activity. Menthol–decanoic-acid HDES was subsequently applied to the enzymatic synthesis of glucose monodecanoate, providing higher product yields than hydrophilic DESs while allowing enzyme reuse.194 More recently, Zhang et al. extended HDES-based biocatalysis to redox reactions using lidocaine/oleic-acid and lidocaine/decanoic-acid systems for alcohol-dehydrogenase-catalyzed carbonyl reduction.195 The HDESs enhanced the availability of hydrophobic substrates while maintaining sufficient enzyme hydration, enabling substantially higher substrate loading than conventional aqueous conditions. HDESs are also beginning to emerge in non-enzymatic chemical synthesis. Pouget et al. developed a thermo-switchable lidocaine–decanoic-acid HDES for copper-catalyzed azide–alkyne cycloaddition, in which a mild temperature change induced phase separation after reaction, facilitating product and catalyst separation and solvent reuse.196 In another approach, cetyltrimethylammonium bromide (CTAB)-based acidic HDESs were used directly as catalysts for esterification of biomass-derived levulinic acid with 1-butanol, demonstrating that HDES constituents can provide both reaction-medium and catalytic functions.197 More recently, a camphor–decanoic-acid HDES was employed as a water-resistant medium for ring-opening polymerization of amino-acid N-carboxyanhydrides, achieving rapid and controlled polypeptide synthesis even under humid conditions.198 These studies demonstrate that the role of HDESs in reaction engineering can extend beyond simple solvent replacement to include substrate delivery, enzyme stabilization, catalysis, and integrated reaction – separation. Nevertheless, this application remains relatively underdeveloped, and future studies should systematically evaluate reaction kinetics, catalyst or enzyme stability, product recovery, solvent regeneration, and the influence of water and impurities. Overall, the optimal HDES formulation is application-dependent and requires balancing target affinity with transport properties, phase and compositional stability, product recovery, regeneration, and long-term solvent performance. Conclusion and perspectives In this review, we surveyed the development of HDESs since their first introduction in 2015, highlighting how reduced affinity toward water can mitigate some of the water-instability challenges associated with traditional (hydrophilic) DESs. This review has summarized the fundamental properties of HDESs, including density, viscosity, thermal stability, phase behavior, and water stability, with particular emphasis on the experimental methods used for their characterization. Computational approaches, including quantum chemical calculations, COSMO-RS, MD simulations, and ML, further provide complementary tools for understanding intermolecular interactions and accelerating solvent design. Despite this progress, HDES research remains relatively immature, and several fundamental and practical challenges should be addressed to facilitate its transition toward rational design and large-scale application. (1) A clearer and widely accepted definition of HDESs is urgently needed. The broader DES field continues to debate how a “deep” eutectic solvent should be distinguished from an ordinary eutectic solvent or mixture (ES/EM) and from more broadly defined low-transition-temperature or low-melting mixtures (LTTMs/LMMs).199,200 The term “hydrophobic” introduces an additional ambiguity. At present, no universally accepted quantitative threshold defines how much water uptake or component leaching is permissible for a DES to be classified as hydrophobic. Although screening criteria based on component water solubility, partition coefficients, phase separation, and compositional stability have been proposed,41 these are not universal definitions. Future studies should therefore, where feasible, separately quantify both water uptake into the HDES phase and transfer of individual HDES components into the aqueous phase after equilibration. Establishing community-agreed criteria for DES formation and hydrophobicity would substantially improve the consistency and comparability of HDES research. (2) Standardized or broadly accepted characterization protocols are critical for HDES studies. Considerable variation currently exists in the preparation, equilibration, measurement, and reporting of HDES properties, making direct comparison among studies difficult. At minimum, HDES studies should clearly report component purity, molar ratio, preparation procedure, initial water content, and measurement temperature. Determination of phase behavior across several compositions using DSC and, where possible, solid–liquid phase diagrams would provide stronger evidence of eutectic behavior. Likewise, assessment of hydrophobicity should extend beyond visual observation of two phases and combine quantitative water-content measurements with analysis of HDES-component leaching. Standardized procedures for density, viscosity, and thermal stability would further enable reliable cross-study comparison and provide higher-quality datasets for predictive modeling. (3) AI-guided HDES discovery represents an important opportunity for moving beyond trial-and-error screening. The enormous combination space created by different HBAs, HBDs, and molar ratios makes exhaustive experimental screening impractical. Machine-learning models have already been developed to predict properties such as eutectic temperature, viscosity, and CO2 solubility, demonstrating the potential of data-driven solvent design.122,201,202 Future AI frameworks should increasingly integrate experimental data with molecular descriptors obtained from COSMO-RS, quantum chemistry, and MD simulations to simultaneously predict multiple properties, including phase stability, hydrophobicity, viscosity, extraction performance, and toxicity. Importantly, further progress will depend on standardized and openly accessible datasets that include experimental conditions, uncertainty, and unsuccessful formulations rather than only positive results. Ultimately, closed-loop strategies combining AI prediction, experimental validation, and iterative model updating could enable task-specific HDESs to be designed rather than discovered empirically. (4) Solvent recovery, process integration, and scale-up require substantially greater attention. Many HDES studies report high extraction or absorption efficiencies while providing limited information on downstream solute recovery and solvent regeneration. Anti-solvent precipitation, evaporation, distillation, membrane separation, and back-extraction can recover products or regenerate HDESs, but these additional operations may introduce significant material and energy requirements. High viscosity also remains a practical limitation for mixing, pumping, mass transfer, and phase separation. Consequently, future work should evaluate HDES performance at the process level, including repeated recycling, solvent loss, and mass and energy balances. Stimuli-responsive or switchable HDESs that facilitate product separation and solvent regeneration may be particularly promising. (5) The molecular basis of HDES formation and performance should be connected more systematically with macroscopic properties and applications. Hydrogen bonding alone may not be able to fully describe the behavior of many HDESs. Integrating spectroscopy and thermodynamic measurements with DFT, MD, and COSMO-based modeling can help establish reliable composition–structure–property–performance relationships. Such understanding is particularly important for designing task-specific solvents. (6) Finally, the sustainability of HDESs should be demonstrated rather than assumed. The use of naturally derived or nominally low-toxicity components does not necessarily guarantee that the resulting mixture is environmentally benign. Available studies have shown that DES toxicity can depend strongly on both the individual constituents and their interactions.203–205 For HDESs to fulfill their green promise, thorough toxicological profiling is required for both individual components and the combined mixtures. This includes evaluating acute and chronic toxicity, biodegradation rates in the environment, and bioaccumulation potential. The lack of standardized protocols for evaluating DES toxicity and environmental fate is a gap that needs to be addressed by the research community and regulatory agencies. Additionally, integrating life-cycle assessment and techno-economic analysis in HDES research is essential for gaining a holistic understanding of their environmental impact and overall sustainability compared to conventional solvents. Author contributions Yuxuan Zhang: conceptualization, investigation, visualization, writing – original draft. Kailong Zhang: investigation, visualization, writing – review & editing. Juan Feng: visualization, writing – review & editing. Jian Shi: supervision, writing – review & editing, funding acquisition. Conflicts of interest There are no conflicts to declare.