Predicting the suitability of aqueous solutions of deep eutectic solvents for preparation of co-continuous porous carbons via spinodal decomposition processes

Spinodal decomposition (SD) processes have proved effective for the synthesis of macroand mesoporous materials. Despite the theoretical aspects of SD processes are well understood, finding the proper experimental conditions – both the components as well as the ratio in which they have to be combined – to attain bicontinuous structures is a non-predictable and quite tedious process, typically based on trial and error. The challenge is therefore finding both a “tool” capable to predict the suitability of a particular starting solution to undergo a SD process. Here in, we have used aqueous solutions of DESs for the preparation of bicontinuous porous carbons via SD. Interestingly, morphologies range from spinodalto aggregates-ofparticle-like just depending on dilution. More interestingly, we also studied the starting DES/H2O binary mixtures by Brillouin spectroscopy given the capability of this technique to observe local structure rearrangements resulting from non-ideal mixing. In our case, Brillouin spectra revealed the presence of a certain nanostructure in the macroscopically homogeneous DES/H2O binary mixtures used for carbons preparation. Moreover, we also found that the representation of hypersonic velocity obtained from Brillouin spectroscopy revealed the occurrence of nanostructural rearrangements in a dilution range that coincided with that where carbons morphology transitioned from spinodalto aggregates-of-particle-like. According to these results, Brillouin spectroscopy allows predicting not only the suitability of a particular solution for the preparation of carbons via spinodal decomposition processes but also the range of dilutions for which this morphology will be ultimately obtained. Introduction Porous carbons are lately attracting considerable attention due to their potential in applications such as membranes for adsorption, filtration and separation, catalytic supports, electrodes in fuel cells, supercapacitors or batteries, 4, 5 or controlled release media, among others. In practical terms, hierarchical materials combining small and large pores are required. It is well known that the former are the main responsible of the activity – the larger the surface area, the higher the activity – but it is also worth noting that the ultimate performance of these materials is largely dependent on the accessibility to such an active surface. Thus, efforts must focus on the design of synthetic processes promoting the formation of bicontinuous structures of interconnected pores with mean diameters in the nano/micrometer range that allow mass transport throughout the whole monolithic structure. A number of synthetic routes have been explored by using different carbonaceous precursors 8 and either hard or soft templates to modulate the porous texture of the resulting carbon structures. 10 In this regard, syntheses using hard-templates – where the pore structure is a replica of pre-synthesized mesoporous oxides and nanoparticles – typically suffer from a tedious procedure during the preparation and removal of templates – e.g. chemical etching. Thus, syntheses based on the use of soft-templates – via selfassembly of carbon precursors with block copolymers and surfactants that are sacrificed as porogens during carbonization – are considered as a more convenient approach. In this latter case, microphase separation induced by the well-ordered phases of self-assembled block copolymers and surfactants are responsible of the ultimate structure of the resulting materials. Nonetheless, the scale-up capability of these soft-template syntheses is questionable because the synthetic origin of most of block copolymers and surfactants make them costly and non-friendly in environmental terms. Within this context, more environmentally benign and cost-effective phase separation processes have been investigated. Among others, spinodal decomposition (SD) processes have proved effective for the synthesis of macroand mesoporous materials. 12, 13, 14, 15, 16 In this case, it is widely accepted that phase separations can be induced both thermally and chemically. Thermally induced phase separations are applied when the phase diagram of a certain polymer exhibits an upper critical solution temperature so that a homogeneous solution formed at a determined temperature is phase separated by thermal quenching. In chemically induced phase separations, the growth of oligomers/polymers makes them less and less soluble in the originally homogeneous starting solution and results in a phase-separation – i.e. between a polymerrich and a polymer-depleted phase – that is ultimately frozen upon gelation when the growing polymers reach a certain degree of cross-linking. Diverse structures – e.g. interconnected or open-cell pores – can be formed depending on the experimental parameters used for polymerization such as the solvent of choice and its volume fraction, and/or the temperature. Nonetheless, the transient structure that results from phase separation comes ultimately determined by the time relation between the onset of phase separation and gel formation – e.g. interconnected structures with a network-like morphology if the onset of both processes coincides, or with an aggregates-of-particles morphology if phase separation occurs earlier than gel formation so that the polymer-rich domains break-up and become spherical during the coarsening process of the SD in order to decrease the interfacial energy. 18, 19 SD processes have been typically used for the preparation of porous metal oxides via sol-gel, 18, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 vinyl and acrylate polymers via radical polymerization, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44 and epoxy, phenolic and furfuryl resins via condensation (Table 1). 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60 It is worth noting that submission of some of these polymers – e.g. vinyl polymers as well as phenolic and furfuryl resins – to a carbonization process is a common process to obtain macroporous carbons. 50-60 In practical terms, the achievement of bicontinuous structures via SD processes need of homogeneous starting solutions containing (1) the monomers/precursors that will form the final polymer/material and (2) a phase separation inducing agent – PSIA – that, at a certain stage of the process, segregate from the originally homogenous solution into a continuous phase. Eventually, the starting solution also contains (3) a co-solvent that helps to demix the phase separation inducing agent and the growing polymer into a bi-continuousphase-separated system. The PSIA most widely used – either by its own or with the aid of some additional co-solvents – with every of the monomers/precursors has been polyethylene glycol (PEG) but the use of other polymers is also common – see, for instance, PO, PC, PDMS, or PAAm in Table 1. In these cases, both the polymer mass ratio – referred to the monomer/precursor – and the molecular weight have been used to control the SD process and thus obtain different morphologies. More recently, bicontinuous structures via SD processes have been obtained using different mass fractions of single co-solvents – e.g. formamide, ethylenglycol, ethanol, or even water in some few cases – as PSIAs (see Table 1). Despite the theoretical aspects of SD processes are well understood, finding the proper experimental conditions – both the components as well as the ratio in which they have to combine among – to attain bicontinuous structures via SD processes is a non-predictable and quite tedious process, typically based on trial and error. Thus, finding a “tool” capable to predict the suitability of a particular starting solution would indeed be challenging. For this purpose, we hypothesized that a simplification of the starting solutions by using single co-solvents as PSIAs – e.g. those described above or, at least, similar to – would be also of help to get the rationale behind the experimental aspects of the process. Our group has recently reported on the preparation of bicontinuous porous carbons using deep eutectic solvents (DESs). 62, 63, 64, 65 DESs were first described by Abbot and coworkers in 2003 as supromolecular complexes formed between an hydrogen bond donor – HBD – and an hydrogen bond aceptor – HBA, typically an ammonium or phosphonium salt. More recently, similar eutectics have also received the name of low-melting eutectic mixtures – by König and coworkers – or low-transition–temperature mixtures – by Kroon and coworkers. DESs share many characteristics with conventional ionic liquids (ILs) – e.g. nonreactive with water, non-volatile, and biodegradable – but the decrease in the melting point in DES is ascribed to the charge delocalization that occurs between the HDB and the HBA. In our case, DESs were composed of phenol derivatives and ammonium salts, so that the materials resulting after polycondensation exhibited a bicontinuous porous structure obtained via a SD process where the formation of a polymer-rich phase by polycondensation was accompanied by the segregation of the non-condensed matter – e.g. the ammonium salts – creating first a polymer-poor phase that, ultimately, becomes a polymer-depleted phase. We found that this phase-separation process ends with the formation of either interconnected or aggregates-of-particles-like structures but, despite we have been recently capable to tailor the textural properties of the resulting porous carbons, we never rationalized the experimental conditions that let transitioning from one structure to another. This paper contains two-well-separated parts. We first studied whether carbons prepared from DESs could exhibit different morphologies upon the control of the SD. The DES of choice was based on a mixture of resorcinol, glycerol, phosphoric acid and choline chloride in a 1:1:1:1 molar ratio. DES was studied by H NMR spectroscopy and differential scanning calorimetry (DSC). This DES was used for the pre


Introduction
Porous carbons are lately attracting considerable attention due to their potential in applications such as membranes for adsorption, filtration and separation, 1 catalytic supports, 2 electrodes in fuel cells, supercapacitors or batteries, 3,4,5 or controlled release media, 6 among others. In practical terms, hierarchical materials combining small and large pores are required. It is well known that the former are the main responsible of the activity -the larger the surface area, the higher the activity -but it is also worth noting that the ultimate performance of these materials is largely dependent on the accessibility to such an active surface. Thus, efforts must focus on the design of synthetic processes promoting the formation of bicontinuous structures of interconnected pores with mean diameters in the nano/micrometer range that allow mass transport throughout the whole monolithic structure.
A number of synthetic routes have been explored by using different carbonaceous precursors 7,8 and either hard or soft templates to modulate the porous texture of the resulting carbon structures. 9,10 In this regard, syntheses using hard-templates -where the pore structure is a replica of pre-synthesized mesoporous oxides and nanoparticles -typically suffer from a tedious procedure during the preparation and removal of templates -e.g. chemical etching. Thus, syntheses based on the use of soft-templates -via selfassembly of carbon precursors with block copolymers and surfactants that are sacrificed as porogens during carbonization -are considered as a more convenient approach. In this latter case, microphase separation induced by the well-ordered phases of self-assembled block copolymers and surfactants are responsible of the ultimate structure of the resulting materials. Nonetheless, the scale-up capability of these soft-template syntheses is questionable because the synthetic origin of most of block copolymers and surfactants make them costly and non-friendly in environmental terms.
Within this context, more environmentally benign and cost-effective phase separation processes have been investigated. Among others, spinodal decomposition (SD) processes have proved effective for the synthesis of macro-and mesoporous materials. 11,12,13,14,15,16 In this case, it is widely accepted that phase separations can be induced both thermally and chemically. Thermally induced phase separations are applied when the phase diagram of a certain polymer exhibits an upper critical solution temperature so that a homogeneous solution formed at a determined temperature is phase separated by thermal quenching. In chemically induced phase separations, the growth of oligomers/polymers makes them less and less soluble in the originally homogeneous starting solution and results in a phase-separation -i.e. between a polymerrich and a polymer-depleted phase -that is ultimately frozen upon gelation when the growing polymers reach a certain degree of cross-linking. Diverse structures -e.g. interconnected or open-cell pores -can be formed depending on the experimental parameters used for polymerization such as the solvent of choice and its volume fraction, and/or the temperature. Nonetheless, the transient structure that results from phase separation comes ultimately determined by the time relation between the onset of phase separation and gel formation -e.g. interconnected structures with a network-like morphology if the onset of both processes coincides, or with an aggregates-of-particles morphology if phase separation occurs earlier than gel formation so that the polymer-rich domains break-up and become spherical during the coarsening process of the SD in order to decrease the interfacial energy. 17,18,19 SD processes have been typically used for the preparation of porous metal oxides via sol-gel, 17,18,20,21,22,23,24,25,26,27,28,29,30,31 vinyl and acrylate polymers via radical polymerization, 32,33,34,35,36,37,38,39,40,41,42,43,44 and epoxy, phenolic and furfuryl resins via condensation (Table 1). 45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60 It is worth noting that submission of some of these polymers -e.g. vinyl polymers as well as phenolic and furfuryl resins -to a carbonization process is a common process to obtain macroporous carbons. [38][39][40][50][51][52][53][54][55][56][57][58][59][60] In practical terms, the achievement of bicontinuous structures via SD processes need of homogeneous starting solutions containing (1) the monomers/precursors that will form the final polymer/material and (2) a phase separation inducing agent -PSIA -that, at a certain stage of the process, segregate from the originally homogenous solution into a continuous phase. Eventually, the starting solution also contains (3) a co-solvent that helps to demix the phase separation inducing agent and the growing polymer into a bi-continuousphase-separated system. The PSIA most widely used -either by its own or with the aid of some additional co-solvents -with every of the monomers/precursors has been polyethylene glycol (PEG) but the use of other polymers is also common -see, for instance, PO, PC, PDMS, or PAAm in Table 1. In these cases, both the polymer mass ratio -referred to the monomer/precursor -and the molecular weight have been used to control the SD process and thus obtain different morphologies. More recently, bicontinuous structures via SD processes have been obtained using different mass fractions of single co-solvents -e.g. formamide, ethylenglycol, ethanol, or even water in some few cases -as PSIAs (see Table 1). Despite the theoretical aspects of SD processes are well understood, finding the proper experimental conditions -both the components as well as the ratio in which they have to combine among -to attain bicontinuous structures via SD processes is a non-predictable and quite tedious process, typically based on trial and error. Thus, finding a "tool" capable to predict the suitability of a particular starting solution would indeed be challenging. For this purpose, we hypothesized that a simplification of the starting solutions by using single co-solvents as PSIAs -e.g. those described above or, at least, similar to -would be also of help to get the rationale behind the experimental aspects of the process.
Our group has recently reported on the preparation of bicontinuous porous carbons using deep eutectic solvents (DESs). 61,62,63,64,65 DESs were first described by Abbot and coworkers in 2003 66 as supromolecular complexes formed between an hydrogen bond donor -HBD -and an hydrogen bond aceptor -HBA, typically an ammonium or phosphonium salt. More recently, similar eutectics have also received the name of low-melting eutectic mixtures -by König and coworkers 67 -or low-transition-temperature mixtures -by Kroon and coworkers. 68 DESs share many characteristics with conventional ionic liquids (ILs) -e.g. nonreactive with water, non-volatile, and biodegradable -but the decrease in the melting point in DES is ascribed to the charge delocalization that occurs between the HDB and the HBA. In our case, DESs were composed of phenol derivatives and ammonium salts, so that the materials resulting after polycondensation exhibited a bicontinuous porous structure obtained via a SD process where the formation of a polymer-rich phase by polycondensation was accompanied by the segregation of the non-condensed matter -e.g. the ammonium salts -creating first a polymer-poor phase that, ultimately, becomes a polymer-depleted phase.
We found that this phase-separation process ends with the formation of either interconnected or aggregates-of-particles-like structures but, despite we have been recently capable to tailor the textural properties of the resulting porous carbons, 69 we never rationalized the experimental conditions that let transitioning from one structure to another. This paper contains two-well-separated parts. We first studied whether carbons prepared from DESs could exhibit different morphologies upon the control of the SD. The DES of choice was based on a mixture of resorcinol, glycerol, phosphoric acid and choline chloride in a 1:1:1:1 molar ratio. DES was studied by 1 H NMR spectroscopy and differential scanning calorimetry (DSC). This DES was used for the preparation of porous carbons via polycondensation with formaldehyde, and subsequent carbonization. As mentioned above, SD processes typically need of a PSIA that, at a certain stage of the process, segregate from the originally homogenous solution into a continuous phase. In our case, we just used water as co-solvent and different mass fractions of it -i.e. first, that originally coming from the formaldehyde solution and then, increasing this one upon further additions -to tailor the morphology of the resulting bicontinuous structure.
The degree of condensation of the polymers resulting after polycondensation was studied by 13 C-NMR and FTIR spectroscopies. The morphology of the resulting materials was studied by scanning and transmission electron microscopies. Further insights about the textural properties of the resulting carbons were obtained from N 2 and CO 2 adsorption/desorption isotherms, and Hg porosimetry. In the second part, we used Brillouin spectroscopy to study the starting DES/H 2 O binary solutions aiming to find any physical insight that, in advance, may help to predict the suitability of these solutions in SD processes.

Experimental Section
Materials: Resorcinol (R), phosphoric acid (P), Glycerol (G), choline chloride (C) and formaldehyde (F, 37 wt% in aqueous solution) were purchased from Sigma-Aldrich and used as received. Water was distilled and deionized.

Preparation of carbons:
The DES RPGC was obtained by physical mixing of the individual components (R, P, G and C) and posterior thermal treatment at 90°C overnight. The molar ratio among the components was 1:1:1:1. Phenolic resins were obtained from DES RPGC following the procedure described in previous reports. 65 Briefly, F (37 wt% in aqueous solution) was added to DES RPGC at a F to R molar ratio of 2:1. Afterwards, samples were thermally treated for 6 h at 60 °C and then for 4 days at 90 °C to become phenolic resins. The resin thus prepared was named as R RPGC80 where "80" indicates the DES mass fraction considering the sum of the mass of DES and water as 100%. R RPGC70 , R RPGC60 , R RPGC50 , R RPGC40 , R RPGC30 , and R RPGC20 were obtained upon the addition of the corresponding amount of water to every DES prior addition of F (see Table 2). After polycondensation, R RPGC were submitted to a carbonization process -i.e. 800 °C for 4 hours in nitrogen atmosphere, with a heating ramp of 1.0 °C/min -to obtain the respective carbons -e.g. C RPGC80 , C RPGC70 , C RPGC60 , C RPGC50 , C RPGC40 , C RPGC30 , and C RPGC20 .
Sample characterization: DES RPGC and its respective dilutions -e.g. DES RPGC80 , DES RPGC70 , DES RPGC60 , DES RPGC50 , DES RPGC40 , DES RPGC30 , and DES RPGC20 , see Table 1 -were studied by 1 H NMR spectroscopy using a Bruker spectrometer DRX-500. The samples were placed in capillary tubes, using deuterated water (D 2 O) for dilution and deuterated chloroform (CDCl 3 ) as an external reference. DES RPGC and its respective dilutions were also studied by high resolution Brillouin spectroscopy. Brillouin spectra were recorded using a Sandercock 3+3 Pass Fabry-Pérot interferometer 70 as Brillouin spectrometer and the light source was an Ar+ laser working at a wavelength (λ 0 ) of 514.5 nm. In this case, the liquid samples were placed in optical cuvettes (Starna) with 1 mm in optical path length. Experiments were performed using the 90A scattering geometry that is independent of the refractive index. 71 The acoustic wave vector was q 90A = [4 π sin(π/4)] / λ 0 . The hypersonic sound propagation velocity ( H ) was obtained from the relation between the Brillouin frequency shift (f) and q 90A , and expressed as  H = (2 π f) / q 90A . Differential scanning calorimetry (DSC) was performed with a TA Instruments Model DSC Q-100 system, under a nitrogen atmosphere. The DSC scan was run in an aluminum pan in a sealed furnace, stabilized for 30 min at 50 °C, and then cooled to -90 °C before heating at rates of 1 and 10 °C/min. Fourier transform infrared (FTIR) spectroscopy of R RPGC -resins was performed in a FTIR spectrometer Bruker Model IFS60v. Solid-state 13 C CPMAS NMR spectra of R RPGC -resins were obtained using a Bruker Model AV-400-WB spectrometer, by applying a standard cross-polarization pulse sequence.
Elemental chemical analysis (ECA) of R RPGC -resins was performed in an LECO Elemental Analyzer CHNS-932.
Total Reflexion X-Ray Fluorescence (TXRF) analysis of R RPGC resins was performed in a TXRF S2 PicoFox (Bruker), after a resins digestion in open vessel in nitric acid. The morphology of R RPGC -resins and C RPGCcarbons was studied via scanning electron microscopy (SEM), using a SEM Hitachi S-3000N system. N 2 adsorption/desorption isotherms were conducted using an ASAP 2010 from Micromeritics system on C RPGCcarbons thermally treated overnight at 250 °C prior to nitrogen gas adsorption. Brunauer-Emmett-Teller (BET) theory and the Dubinin-Radushkevich method were used to calculate the specific surface areas (SBET) and micropore volumes, respectively. CO 2 adsorption-desorption isotherms were performed at 0 and 25 °C in a Micromeritics Tristar 3020 instrument in the pressure range of 0.1-900 mbar.

Preparation of Porous Materials from DES/H 2 O binary mixtures
As mentioned in the introduction, we first prepared the DES RPGC composed of resorcinol (R), glycerol (G), phosphoric acid (P), and choline chloride (C) in a 1:1:1:1 molar ratio. This DES was obtained by simple thermal treatment at 90 °C of the physical mixture of the components. The formation of eutectic mixtures was first revealed by the liquid nature of the mixture at room temperature ( Figure 1). Neither the melting (T m ) nor the crystallization temperatures (T c ) were clearly displayed in the DSC trace of the DES RPGC ( Figure 2), a common feature observed for non-easily crystallizable ILs and DESs. 72 The formation of H-bond complexes was also confirmed by 1 H NMR spectroscopy -i.e. see the upfield chemical shift of the signals ascribed to R, G and C in DES RPGC as compared to the chemical shifts of the individual components (see Figure S1 and Table   S1 at Supporting Information). 73,74,75,76 Polycondensation was promoted by addition of F to DES RPGC with a fixed F to R molar ratio of 2 ( Table 2).
As described above, we prepared a set of samples upon addition of different amounts of water -e.g. R RPGC80 , R RPGC70 , R RPGC60 , R RPGC50 , R RPGC40 , R RPGC30 , and R RPGC20 . R RPGC80 was the sample with the minimum water content, coming from the aqueous F solution supplied by Sigma-Aldrich that we used for polycondensation. R RPGC70 , R RPGC60 , R RPGC50 , R RPGC40 , R RPGC30 , and R RPGC20 samples were obtained upon addition of further water fractions to the aqueous F solution ( Table 2). Condensation between R and F was assessed by FTIR and solid-state 13 C CPMAS NMR spectroscopies. The FTIR spectra of every resin revealed the presence of bands ascribed to methylene (C-H stretching and bending modes at ca. 2953 and 1475 cm -1 , respectively) and to methylene oxide groups (C-O benzyl ether groups at about 1091 and 1220 cm -1 ), both indicative of a significant number of linkages between R rings ( Figure S2). 77 The  3 and  1 modes of phosphoric acid typically produce bands at 1183, 1003, and 882 cm -1 so their appearance in the spectra indicated the presence of phosphoric acid entrapped into the phenolic resin. 65 Solid-state 13 C CPMAS NMR spectroscopy provided further insights with regard to the polycondensation issue ( Figure S3). The signal at 151 ppm corresponded to aromatic phenolic carbons and its intensity could be used as an internal reference to estimate the polycondensation degree.
Polycondensation was confirmed by the presence of methylene ether and methylene groups -e.g. CH 2 -O-CH 2 at ca. 68 and 53 ppm and CH 2 at ca. 31 and 24 ppm -linking adjacent aromatic rings through 4-4' and 2-4' positions. 78,79 The signal at ca. 120-121 ppm was also interesting in terms of condensation since it corresponds to aromatic carbons of mono-and disubstituted R molecules bearing CH 2 groups in all ortho positions relative to the phenolic OHs. Non-substituted aromatic CH groups in meta positions appeared at and 131 ppm exhibited similar intensities for all the samples, thus revealing basically identical polycondensation degrees in every case.
Resins were thermally treated at 800 °C in N 2 atmosphere for the formation of the respective carbons with good conversions -i.e. in range to those reported for previous DES-assisted condensations 61-65 - (Table   3) as anticipated by FTIR and 13 C CPMAS NMR spectroscopies. The carbonization of unwashed resins produced carbons with significant nitrogen and phosphorus contents ( Table 3). The textural properties of C RPGC80 , C RPGC40 and C RPGC20 were studied by gas adsorption -N 2 at -196 °C and CO 2 at 0 and 25 °C -and Hg porosimetry ( Figure 3 and Figure S4, respectively). N 2 adsorption-desorption isotherms were type I with a sharp increase in adsorption at low relative pressures that is characteristic of microporous materials ( Figure   3a). Table 4 summarizes the Brunauer-Emmett-Teller surface area, the micropore volume and the pore size distributions (PSDs) of micro-and macropores -i.e. obtained from the analysis of the isotherms using the Barrett-Joyner-Halenda (BJH) method in the former case, and from Hg porosimetry in the latter one. CO 2 adsorption-desorption isotherms also confirmed the presence of a narrow microporous network throughout every RPGC carbon ( Figure 3b). Moreover, the CO 2 adsorption capacity was remarkable at both 0 and 25 °C, with uptakes at 760 mbar that were in range to or even better than some of the most remarkable sorption capacities found for microporous carbons of different natures -see in Table 5 examples of non-doped, doped, and co-doped carbons in either non-activated or activated form. 80,81,82,83,84,85,86 At this stage, we investigated the morphology of the different carbons by SEM. SEM micrographs revealed how the morphologies changed remarkably with the increase of the mass fraction of water during polycondensation. Thus, C RPGC80 exhibited a fine bicontinuous structure ( Figure 4). Meanwhile, C RPGC70 exhibited a bicontinuous structure as well, but with a thicker skeleton as compared to R RPGC80 (Figure 4). The SEM micrographs of C RPGC70 also revealed the presence of both small spherical pores -of about 0.5-1 μm -in the coarse skeleton and small particles -of about 0.5-1 μm -within the open porosity that ultimately landed on the surface of the skeleton after the removal of the non-condensed matter -i.e. by drying and subsequent carbonization. The spherical pores were voids left by the solvents that, during polycondensation, separated out in the polymer-rich domain as a result of the secondary phase separation. Secondary phase separation is the precipitation of a phase within a phase, and can occur when a region of a phase-separated microstructure is physically isolated and develops further immiscibility. 29,30,50,51 Secondary phase separation may occur in the polymer phase, in the pore phase or in both, as it actually occurred in our case. Further increase of the mass fraction of water during polycondensation provided morphologies to C RPGC60 , C RPGC50 , C RPGC40 , C RPGC30 , and C RPGC20 characterized by loosely packed isolated spheres, the mean size of which became larger along with the increase of water content because of the occurrence of further particles growth for long gelation times (Figures 4 and Figure S5).

Study of starting DES/H 2 O binary mixtures
As described above, the aim of this work was also studying the starting DES/H 2 O binary solutions so that one could predict the suitability of these solutions in SD processes. The above-described results with regard to SD processes during the polycondensation of DES/H 2 O binary mixtures reconciled with previously described chemically induced phase separations where immiscibility is observed to develop concurrently with polymerization, and the ultimate morphology of the resulting polymer depends on the onset between phase separation and gel formation. In every SD process reported to date, the fluid phase resulting after phase separation is a solvent mixture that contains the PSIA, the co-solvent that helps to demix the PSIAnote co-solvents might be used or not -and any additional compound that may result as a by-product of the reaction. 14 In our particular case, the fluid phase was composed of the non-polymerizable DES componentse.g. P, G and C -and H 2 O. The former were originally part of the DES, the rupture of which during polycondensation resulted in their segregation to the fluid phase. With regard to H 2 O, a minor fraction of it was the by-product of RF polycondensation but most of it was added at the starting homogeneous solution and promoted/favoured phase separation. The use of water as PSIA has been rarely described in papers dealing with SD processes 41,42,43,50 despite the obvious interest that its use offers in environmental and sustainable terms as compared to other co-solvents more commonly used -e.g. non-friendly solvents, polymers and/or surfactants. As described previously, 50 the mass fraction of water controls the resulting morphology in terms of reaction kinetics. Thus, in diluted samples, phase separation occurs earlier than gel formation so that the polymer-rich domains break-up into spherical colloids that, afterwards, become a geli.e. sooner for pseudo-diluted samples and later for highly diluted ones ( Figure S6).
It is also worth noting a peculiarity of these DES-assisted SD processes with regard to the role of water as PSIA. As in any regular SD processes, the macroscopic appearance of our starting solutions -aqueous dilutions of DES -was homogeneous ( Figure 1). In our case, all the DES components are soluble in water so, in these cases, it is commonly assumed that the resulting DESs are entirely water miscible as well. 87 However, recent reports describe how the scenery that results from aqueous dilution of DES is not yet fully understood by the research community. 88 For instance, DESs prepared from water-soluble components will tend to become a simple solution of the individual components at high dilutions 89  With regard to the first question, we have demonstrated in some of our previous works that 1 H NMR spectroscopy is a useful tool to get insights about DES rupture upon dilution. [73][74][75][76] As mentioned above, DESs result from the formation of the halide ion-hydrogen-bond-donor supramolecular complexes between two or more molecules, being at least one of them a HBD and another one a HBA. The formation of H-bond complexes typically occurs through the most labile protons that exist in the component acting as HBD. For instance, H-bond complex in one of the most studied DES -e.g. DES UC , composed of urea (U) and C, and combined in a 2:1 molar ratio -are suggested to occur via interactions between the NH 2 groups of U and the chloride ion of C. 90,91,92 More recently, different studies based on neutron diffraction experiments and atomistic modelling have confirmed that each chloride is H-bonded to two U molecules. 93,94 Interestingly, these works also described how, once chloride ion is stabilized by U and to better accommodate U within the H-bond complex, most chloride ions are positioned near the OH group of the choline cation. Thus, the Hbond complex involves one choline, one chloride and two urea molecules, being the H-bonding between the NH 2 groups of U and the chloride ion stronger than that between the OH group of the choline cation and the chloride ion.
The above-described different role that each particular labile proton plays in the H-bond complex is actually in agreement with some 1 H NMR spectroscopic results we published in a previous work. 75 In neat DES UC , we could see how labile protons at the OH group of C and at the NH 2 groups of U were not exchangeable among themselves and produced their respective signals in the spectrum -i.e. at 4.7 and 5.6 ppm. Interestingly, subsequent experiments in D 2 O-diluted DESs revealed that labile protons at the OH group of the choline cation were fully exchanged with D 2 O for DESs contents above 86 wt% whereas labile protons at the NH 2 groups of U needed of further dilution to become fully exchangeable -i.e. for DESs contents below 43 wt% (Figure 5a). In view of these results, we also hypothesized that DES UC fully breaks into its respective counterparts -so it become a solution of the individual components -when all the labile protons are fully exchanged with D 2 O as it happens in any regular solution.
The 1 H NMR spectra of DES RPGC in both its neat and diluted form was also illustrative with regard to the dilution range at which DES is fully broken into its respective components. Nonetheless, DES RPGC revealed a different behaviour than that described above for DES UC . In neat form, all the labile protons -e.g. coming from OH groups of resorcinol and glycerol, as well as from H 3 PO 4 -exchanged among themselves -as revealed their appearance as a single signal at 7.5 ppm -indicating an equal H-bonding participation for every one in the H-bond complex ( Figure S2 and Table S2 at Supporting Information). Upon dilution with D 2 O, the signal assigned to labile protons of DES components evolved -in an up-field-shifting fashionaiming to meet the signal of HDO located at ca. 4.7 ppm (Figure 5b). In structural terms, it is plausible correlating this shifting with the size of H-bond domains. 95,96 In our case, the occurrence of proton exchange must indeed be favoured as DES domains become smaller because of the increase of the surface interface with the surrounding D 2 O. Interestingly, two separate signals yet remained for highly diluted samples -ca.
10 wt% -so we could assume that DES RPGC was not fully broken and some tiny DES-domains yet remain even at that dilution.
With regard to the second question, it is worth reminding the idea of cooperativity of H-bonding in water to ultimately form 3-D H-bonded networks. Briefly, the formation of a first H-bond results in a change in the charge distribution within the first participating unit in such a way that the hydrogen acceptor molecule becomes potentially an even better H-bond donor than before and it is thus capable of forming a stronger second bond because of the existence of the first one. The same is true for the proton donor, which has an enhanced ability to accept a proton as a result of the bond that it has already formed. 97 More interestingly, it is widely accepted the occurrence of amorphous polymorphism -polyamorphism -at ambient conditions, this is liquid water (with average density 1 g/cm 3 ) fluctuating permanently between lowdensity liquid (LDL, density 0.94 g/cm 3 ) and high-density liquid (HDL, density 1.14 g/cm 3 ) clusters. 98 Phase separation between them and liquid-liquid polyamorphous transition have been predicted by numerical modelling 99 but its experimental observation has only been possible in aqueous solutions of salts 100,101 or Hbond-forming molecules -e.g. polyols, sugars, etc. 102,103 It is worth noting that liquid-liquid transitions without macroscopic phase separation has also been observed in water-glycerol mixtures. 104 Protic ILs are reminiscent of water and hence capable to form 3-D H-bonded networks in both neat form and aqueous dilutions. 105 Actually, aqueous solutions of hydrophilic ILs can also result in the formation of aqueous biphasic systems upon the addition of a kosmotropic -water structuring -salt. 106 Elucidating whether this was the situation in the aqueous solutions of RPGC described in this work was performed by Brillouin light-scattering experiments. This experimental technique focuses on the analysis of the light inelastically scattered on the thermally induced sound waves (phonons), that propagate through a medium at a velocity that depends on the sample density and rigidity (expressed in terms of the storage longitudinal modulus). Brillouin spectroscopy has been used to determine acoustic velocities and elastic properties of a number of crystalline solids, glasses, and liquids. Within the context of this work, particularly interesting is the use of Brillouin spectroscopy for studying liquid mixtures of H-bond co-solvents -e.g. methanol (MeOH), ethanol (EtOH), propanol (PrOH), or tert-butanol (t-BuOH), among others -and even polymers -e.g. polyethylenglycol (PEG) -with either water or organic solvents. 111,112,113,114,115,116,117,118,119 The spectrum associated with a binary solution consists of three peaks; the central one resulting from Rayleigh scattering and lacking of interest in this case, and the two Brillouin peaks located at frequencies  = ± k H (f) -referred to the Rayleigh one, where k denotes the absolute value of the wave vector transfer, f is the frequency of the sound wave and  H the hypersonic velocity. Interestingly, a deviation from the ideal behaviour has been observed in binary liquid mixtures when representing the  H obtained from Brillouin scattering measurements versus the molar fraction () of one of the components, thus revealing that these binary mixtures are inhomogeneous on a microscopic molecular level regardless their fully miscible appearance in macroscopic terms.
Local structure rearrangements resulting from non-ideal mixing in binary mixtures have been also observed by molecular simulations and, experimentally, by 1

Conclusions
We have demonstrated the capability of aqueous solutions of DESs for the preparation of porous carbons via SD -and subsequent carbonization -so that the resulting structure just depends on the DES dilution. In particular, SEM micrographs allowed distinguishing the transition from bicontinuous morphologies -this is those in which both phases are smoothly interconnected in three dimensions -to bicontinuous with secondary phase separation -with both isolated pores in the polymer phase and particles in the pore phase -up to aggregates-of-particle morphologies -which can nonetheless be interconnected due to packing considerations -when the mass fraction of water increased. The microporosity of some of these carbons was particularly effective for CO 2 capture with absorptions of up to 4.7 mmol/g at 0 °C and 760 mbar. More interestingly, we have also investigated whether Brillouin spectroscopy can be used to predict the suitability of a particular aqueous solution of DES to undergo the SD process. Brillouin spectroscopy was chosen given its capability to observe local structure rearrangements resulting from non-ideal mixing. In practical terms, we observed that the range of DES contents for which an aqueous solution of DES demix into a cocontinuous structure coincided with a discontinuity of the linearity in the plot of  H versus DES content.
According to these results, Brillouin spectroscopy was capable of predicting not only the suitability of a particular solution for the preparation of carbons via SD processes but also the range of dilutions for which every morphology -e.g. bicontinuous-or aggregates-of-particle-like -will be ultimately obtained.     As compared to regular SD processes, Brillouin spectroscopy has revealed that the macroscopically homogenous starting solution exhibit certain heterogeneities at the nanoscale e i.e. DES nanodomains, first isolated for low DES contents and then forming a bicontinuous structure for increased DES contents. Considering that DES nanodomains are rich in resorcinol, first nucleation and growth of colloidal particles will occur preferentially at these positions so that phase separation is anticipated as compared to regular SD processes.