This is an open access article published under an ACS AuthorChoice License, which permits copying and redistribution of the article or any adaptations for non-commercial purposes. Article Cite This: J. Phys. Chem. C 2019, 123, 29524−29532 pubs.acs.org/JPCC Cation Solvation and Physicochemical Properties of Ca Battery Electrolytes J. D. Forero-Saboya,† E. Marchante,† R. B. Araujo,‡ D. Monti,† P. Johansson,*,‡,§ and A. Ponrouch*,†,§ † Institut de Ciència de Materials de Barcelona (ICMAB-CSIC), Campus UAB, 08193 Bellaterra, Catalonia, Spain ‡ Department of Applied Physics, Chalmers University of Technology, SE-412 96 Göteborg, Sweden § ALISTOREEuropean Research Institute, CNRS FR 3104, Hub de l’Energie, 15 Rue Baudelocque, 80039 Amiens, France * S Supporting Information ABSTRACT: Divalent-cation-based batteries are being considered as potential high energy density storage devices. The optimization of electrolytes for these technologies is, however, still largely lacking. Recent demonstration of the feasibility of Ca and Mg plating and stripping in the presence of a passivation layer or an artificial interphase has paved the way for more diverse electrolyte formulations. Here, we exhaustively evaluate several Ca-based electrolytes with different salts, solvents, and concentrations, via measuring physicochemical properties and using vibrational spectroscopy. Some comparisons with Mg- and Li-based electrolytes are made to highlight the unique properties of the Ca2+ cation. The Ca-salt solubility is found to be a major issue, calling for development of new highly dissociative salts. Nonetheless, reasonable salt solubility and dissociation are achieved using bis(trifluoromethanesulfonyl)imide (TFSI), BF4, and triflate anion based electrolytes and high-permittivity solvents, such as ethylene carbonate (EC), propylene carbonate (PC), γ-butyrolactone (gBL), and N,N-dimethylformamide (DMF). The local Ca2+ coordination is concentration- dependent and rather complex, possibly involving bidentate coordination and participation of the nitrogen atom of DMF. The ionicity and the degree of ion-pair formation are both investigated and found to be strongly dependent on the nature of the cation, solvent donicity, and salt concentration. The large ion−ion interaction energies of the contact ion pairs, confirmed by density functional theory (DFT) calculations, are expected to play a major role in the interfacial processes, and thus, we here provide electrolyte design strategies to engineer the cation solvation and possibly improve the power performance of divalent battery systems. ■ INTRODUCTION Achieving sustainable and affordable energy sources and 1964.5 Mostly, thermal batteries, using electrolytes with very high melting points, and the calcium equivalent to the storage is one of the major challenges of our modern commercial Li thionyl chloride (Li-SOCl2) batteries were societynot the least to attack climate change and oil developed.6 However, in contrast to Mg, electrodeposition of dependency. Considering only the transportation contribution, calcium has for long been elusive and only primary cells were more than 1.2 billion vehicles are currently at use all around achieved. The lack of calcium plating upon cell reversal was the world, and less than 0.2% are electric. While Li-based even considered as a safety advantage for the Ca-SOCl2 cells as battery technologies undoubtedly will power more and more compared with its Li analogue.7 electric vehicles (EVs) in the next 5−10 years, the Moving to room-temperature and liquid electrolytes, no Ca controversial debate on sustainability and production/extrac- equivalents to Mg-based Grignard reagent electrolytes tion costs of lithium cannot be ignored.1,2 The development of (allowing for Mg plating and stripping) are available.8 Seminal alternative battery chemistries based on abundant elements is studies by Aurbach et al. using conventional aprotic organic thus urged for to accommodate a significant growth of the EV electrolytes similar to those used for lithium-ion batteries market. Calcium- and magnesium-based batteries are interest- (LIBs) allowed them to conclude that the electrochemical ing candidates as they hold promise for safe use of metal behavior of Ca metal anodes is surface-film controlled, as is the anodes, thus potentially presenting high energy densities and case for Mg.9,10 The lack of divalent cation permeability of the low cost. Indeed, Ca and Mg electrodes have theoretical formed passivation layer was suggested to be the origin of the specific capacities of 1340 and 2000 mAh g−1, standard redox impossibility to electrodeposit calcium or magnesium. While potentials of −2.76 and −2.38 V vs the standard hydrogen the electrolyte remains a major bottleneck for any divalent- electrode (SHE), and are the fifth and eighth most abundant elements on the Earth’s crust, respectively.3,4 Received: July 31, 2019 A myriad of Ca-based battery concepts have been developed Revised: November 6, 2019 since the first report of calcium as an electroactive element in Published: November 14, 2019 © 2019 American Chemical Society 29524 DOI: 10.1021/acs.jpcc.9b07308 J. Phys. Chem. C 2019, 123, 29524−29532 The Journal of Physical Chemistry C Article Figure 1. Physical properties of calcium-based electrolytes. Arrhenius plots for (a) viscosity and (b) conductivity of 0.1 M calcium salt in PC. Arrhenius plots for (c) viscosity and (d) conductivity for 0.1 M Ca(TFSI)2 in several solvents. Solid lines represent the Vogel−Tammann−Fulcher (VTF) fits. cation-based batteries and most research studies so far have focused on electrolytes for which no passivation layers are ■ EXPERIMENTAL SECTION Materials. The solvents used were propylene carbonate formed (mostly based on ether solvents), 8−10 recent (PC, Aldrich, anhydrous, 99.7%), ethylene carbonate (EC, demonstrations of Ca and Mg plating and stripping in the Aldrich, anhydrous, 99.0%), N,N-dimethylformamide (DMF, presence of a passivation layer or an artificial interphase11−13 Alfa Aesar, 99.7%), γ-butyrolactone (gBL, Aldrich, anhydrous, have paved the way for evaluation of new electrolyte 99.0%), tetrahydrofuran (THF, Aldrich, anhydrous, 99.9%), formulations by allowing decoupling of interfacial and 1,2-dimethoxyethane (DME, Aldrich, anhydrous, 99.5%), electrolyte requirements. The present study, thus, focuses on diethyleneglycol dimethyl ether (diglyme, Aldrich, 99.0%), important parameters in the design of new electrolytes,14,15 triethyleneglycol dimethyl ether (triglyme, Aldrich, 99.0%), such as cation mobility and desolvation energies, broadening and fluoroethylene carbonate (FEC, Aldrich, 99%). the choices of electrolyte component candidate salts and The commercially available calcium salts used were calcium solvents, potentially leading to formation of better-performing bis(trifluoromethanesulfonyl)imide (Ca(TFSI)2, Solvionic, passivation layers. 99.5%), calcium tetrafluoroborate (Ca(BF4)2, Alfa Aesar, As compared to Li+ and Na+ conducting electrolytes, hydrated), calcium trifluoromethanesulfonate (Ca(Tf) 2, divalent-cation-based electrolytes have a stronger tendency of Sigma-Aldrich, 99.9%), calcium sulfate (CaSO4, Sigma-Aldrich, ion-pair formation16 and larger cation desolvation energies.17 99.99%), calcium bis(6,6,7,7,8,8,8-heptafluoro-2,2-dimethyl- In addition, as the Li+ and Na+ mobilities are significantly 3,5-octanedionate) (Ca(FOD)2, Sigma-Aldrich, >99.9%), and lower than those of the anions1 and the desolvation energy calcium bis(2,2,6,6-tetramethyl-3,5-heptanedionate) (Ca- plays a major role in the kinetics of Li+ insertion into (TMHD)2, Sigma-Aldrich). LiTFSI (Solvionic, 99.9%) and graphite,18−21 it can be expected that the power performance Mg(TFSI)2 (Solvionic, 99.5%) were also used to create of any divalent-cation-based batteries will also depend greatly reference electrolyte systems. All salts and solvents were kept on the nature of the cation solvation shell. Here, we and used under a dry atmosphere (<1 ppm O2 and H2O) in an systematically investigate calcium-based electrolytes and report argon-filled Jacomex glovebox at 25 °C. on the relationship between the cation solvation, coordination Electrolytes. First, two families of Ca-based electrolytes number and ion pairing, and physicochemical properties, such were made by dissolving (i) six different Ca-salts in PC at a as ion conductivity and viscosity for different salts and concentration of 0.1 M and (ii) 0.1 M Ca(TFSI)2 in eight concentrations and organic solvents. Some parallels are different solvents. Subsequently, a third family was evaluated drawn with Li and Mg systems allowing for a better for the best solvent and 0.1−1.5 M Ca(TFSI)2, and this was understanding of the strategies to optimize divalent cation also further compared with LiTFSI- and Mg(TFSI)2-based mobility and desolvation energy. electrolytes. All electrolytes were prepared under a dry 29525 DOI: 10.1021/acs.jpcc.9b07308 J. Phys. Chem. C 2019, 123, 29524−29532 The Journal of Physical Chemistry C Article Figure 2. Ion conductivity plots for (a−c) Ca(TFSI)2, (d−f) Mg(TFSI)2, and (g−i) LiTFSI in (a, d, g) PC, (b, e, h) EC, and (c, f, i) DMF, at different concentrations and temperatures. atmosphere in the glovebox and remained there until analysis. local optimizations to explore the energy landscape where the Molecular sieves (Alfa Aesar, 3 Å) were used as drying agents temperature is used to escape any local minima. The AIMD for each electrolyte. Prior to any characterization, the water runs were performed in the microcanonical ensemble with a contents were measured using an 899 Karl Fisher Coulometer time step of 1 fs. The BP86 functional was employed together (Metrohm), resulting in <30 ppm for all electrolytes. with a SVP basis set as implemented in TURBOMOLE.25,26 Experimental Methods. Ion conductivity measurements The MH algorithm started at 300 K and an initial energy were performed using a FRA-based Multiplexed Conductivity threshold of 48 kJ mol−1. The subsequent calculations of gas- Meter MCM 10 (BioLogic Science Instruments) from −20 to phase geometries and Raman activities for the most stable +120 °C (10 °C steps with 20 min equilibration time). structures used the M06-2X functional and the 6-31G** basis Viscosity and density measurements were carried out from set as implemented in the Jaguar package.27−31 All obtained +10 to +90 °C using a rolling-ball viscometer (Lovis 2000 M/ structures appear to be minima, with no imaginary vibrational ME, Anton Parr). The cell temperature was regulated within frequencies. ±0.02 °C. Ultrapure water was used to calibrate the viscometer and densitometer. The uncertainties of the density and viscosity measurements were less than 5 × 10−5 g cm−3 and ■ RESULTS AND DISCUSSION Viscosity and Ion Conductivity. To separate out the 0.5%, respectively. relative influence of the solvent and the nature and Fourier transform infrared (FT-IR) spectroscopy was carried concentration of the salt, two different families of electrolytes out using a Jasco FT/IR-4700 Spectrometer with the universal were made. First, the six different Ca-salts were dissolved in PC ATR Pro One accessory (equipped with a diamond crystal) for aiming at a concentration of 0.1 M to target salt solubility the range 500−4000 cm−1 at a resolution of 2 cm−1. i.e., can we at all make electrolytes from these salts? Notably, Raman spectra were recorded using a Bruker MultiRAM FT- 0.1 M solutions in PC were easily prepared for the majority of Raman spectrometer with a spectral resolution of 2 cm−1 using salts under study, but CaSO4 and Ca(TMHD)2 presented poor a Nd:YAG laser (1064 nm, 500 mW) as the excitation source. solubilities, and therefore, their saturated electrolytes were Data were typically averaged over 500 scans and recorded at 40 used. While the viscosities of all electrolytes are relatively °C for the carbonate-based electrolytes and 25 °C for the similar, their ionic conductivities differ greatly for the different DMF-based electrolytes. All bands in the resulting spectra were salts under study (Figure 1a,b). The Ca(TMHD)2- and fitted as Voigt functions, typically with a unity Gaussian/ CaSO4-based electrolytes show the lowest ion conductivities, Lorentzian mix by width, using Origin Software. mainly due to the poor salt solubility, while Ca(TFSI)2, Computational Details. The Raman activities of the Ca(BF4)2, and Ca(Tf)2 all perform better with their ion complexes [Ca(DMF)6]2+, [Ca(PC)6]2+, and [Ca(FEC)6]2+ conductivities being 3 orders of magnitude higher (Figure 1b). were obtained by first creating the initial structures by This is consistent with the ease of ion−ion dissociation, employing the minima hopping global optimization algorithm facilitating solubility and creation of charge carriers. These (MH) as implemented in the Atomistic Simulation Environ- results highlight one of the major issues in the development of ment (ASE) coupled with TURBOMOLE.22−24 This method new divalent-cation-based electrolytes: the salt solubility and carries out short ab initio molecular dynamics (AIMD) and the need for new highly dissociating salts. 29526 DOI: 10.1021/acs.jpcc.9b07308 J. Phys. Chem. C 2019, 123, 29524−29532 The Journal of Physical Chemistry C Article Turning to the role of the electrolyte solvent, 0.1 M Ca(TFSI)2 electrolytes were studied (Figure 1c,d). It appears that the ion conductivity and the ability to dissolve Ca-salt both follow the solvent donor number (Table S1); highest for DMF and lowest for PC. This trend agrees well with the computed solubility of calcium salts in a wide variety of solvents.32 Tentatively, salt solubility, and not viscosity, seems to be the determining factor for ion conductivity as we find, for example, one of the highest ion conductivities for the EC-based electrolyte (cf. Figure 1d) while being the most viscous electrolyte (Figure 1c). The THF- and glyme-based electro- lytes show decreased ion conductivities as a function of temperature, an unusual behavior attributed to the presence and augmentation of strong ion pairing, as previously reported for Mg-glyme systems.33 Overall, the DMF-based electrolytes preliminarily seem to offer the best performance, and together with more standard PC- and EC-based electrolytes, maximum ionic conductivities are obtained for salt concentrations of 0.42−0.57 M (Figure 2), after which a steep decrease is recorded, especially at lower temperatures. The maxima correspond to the lowest activation energies (Ea) in the VTF fits (Figure S6 and Table S3). Notably, these maxima are found at much lower salt concentrations than those for standard Li+ and Na+ conducting electrolytes, with maxima typically at 1.0−1.2 M. Therefore, as far as macroscopic properties (viscosity and ionic conductivity) Figure 3. Walden plots of (a) different Ca-salts (0.1 M) in PC, (b) are concerned, moderate salt concentrations (around 0.5 M), 0.1 M Ca(TFSI)2 in different solvents and 0.1 to 1.5 M (c) highly dissociating anions (such as TFSI or BF4−), and solvents Ca(TFSI)2 in EC, (d) Ca(TFSI)2 in DMF, (e) LiTFSI in EC, and (f) with high donor numbers (DMF, EC, or PC in that order) are Mg(TFSI)2 in EC. The dashed line represents an aqueous KCl required for Ca-battery electrolytes. solution as a reference for an ideal, fully dissociated electrolyte. Ionicity. While viscosity, in general, limits ion motion in electrolytes, intermolecular and electrostatic interactions are electrolytes have data in close vicinity to the ideal KCl also important as they affect both the solvation, i.e., nature and standard line, implying a high ion−ion dissociation degree. radius of the diffusing (cationic) species, and ion pairing, i.e., The Walden plot slopes are 1.01, 0.95, 0.90, and 0.87 for EC, concentration of charge carriers. All of these interactions are PC, gBL, and DMF, respectively, which shows that while the taken into account in the ionicity of the electrolyte within the DMF electrolytes have the highest ion conductivities and framework of the empirical Walden’s rule.34 While originally lowest viscosities, they do seem slightly less dissociated than formulated for infinitely diluted aqueous electrolytes with the gBL-, EC-, and even PC-based electrolytes. solvated ions of the same radius, it can be used to compare Increasing the salt concentration results in very slight shifts nonideal, nonaqueous electrolytes. In a Walden plot, the toward the ideal KCl line and decreased slopes for both EC logarithm of the molar conductivity, Λ, is plotted against the and DMF electrolytes (Figure 3c−f). This is clearly observed fluidity, i.e., the inverse of the viscosity, 1/η, and most often an for Ca(TFSI)2 and Mg(TFSI)2 but is much subtler for the ideal (infinitely diluted) aqueous KCl solution is chosen as a LiTFSI electrolytes. It can be understood in terms of ionicity reference system.35,36 Deviations below the ideal KCl line as increasing salt concentration for divalent cation electrolytes indicate that other actions than viscosity hinder the ionic promotes extensive ion−ion interactions. motion, e.g., ion pairing. In contrast to the macroscopic view on electrolyte The Walden plots for the Ca, Mg, and Li electrolytes show performance presented in the previous section, ionicity is a all data to fall below the ideal KCl line, and hence they are more local criterion, and an electrolyte with a high dielectric subject to ion pairing. The anion influence is clear as the constant shows a better performance: EC > PC > gBL > DMF. ionicity (Figure 3a) and the ion conductivity (Figure 1b) co- Therefore, there is a clear discrepancy between the micro- and vary: Ca(TFSI)2 > Ca(BF4)2 > Ca(Tf)2 >≫ Ca(FOD)2 > macroscopic points of view. To understand this discrepancy CaSO4 > Ca(TMHD)2, in clear agreement with how better, the molecular interactions and speciation were further delocalized the negative charge is for the three former anions investigated by vibrational spectroscopy. and the poor salt solubility and ion−ion dissociation for the Cation−Anion and Cation−Solvent Interactions. As three latter salts. the cation−solvent interactions compete with the cation− For the solvents, the ionicity of the electrolytes increases anion interactions, we turn to vibrational spectroscopy to with increasing dielectric constant (Figure 3b and Table S1): assess the local coordination in more detail. Starting with EC > PC > gBL > DMF > triglyme > diglyme > DME > THF. infrared spectroscopy, the very intense IR absorption band of For THF and the glymes (incl. DME), the unusual behavior, the carbonyl CO stretching mode should be an interesting illustrated by the negative slope in Figure 3b, can be attributed probe for EC, PC, and DMF as solvents. However, for cyclic to their low dielectric constants and the formation of ionic carbonates, multiple mechanisms prevent any straightforward aggregates.37 In contrast, the EC, PC, gBL, and DMF analysis. The fundamental stretching frequency of CO (ν2) 29527 DOI: 10.1021/acs.jpcc.9b07308 J. Phys. Chem. C 2019, 123, 29524−29532 The Journal of Physical Chemistry C Article at 1793 cm−1 couples strongly with the first overtone of the skeletal breathing of the ring (2ν7) and produces an intense band at 1767 cm−1; a Fermi resonance phenomenon (Figure 4).38 Upon coordination with a cation, these bands can, thus, Figure 5. IR spectra of DMF-based electrolytes as a function of Mg(TFSI)2, Ca(TFSI)2, and LiTFSI salt concentrations. Figure 4. Carbonyl stretching region IR spectra of Ca(TFSI)2 in EC- coordination.41,42 As compared to formamide, the nitrogen and PC-based electrolytes as a function of salt concentrations. atom in DMF has a higher electronic density due to the electron donor character of the methyl groupsfurther shift due to different mechanisms: losing the Fermi coupling supported by computational studies showing that its electron condition would return the mode to its “unperturbed” lone-pair occupies a π nonbonding molecular orbital, which frequency (gas-phase IR39 ca. 1870 cm−1, upward shift), also involves the methyl groups.43 The distinct 20 cm−1 shift of decreased electron density of the CO bond (downward the symmetric bending of the −CH3 groups upon coordination shift), and/or reduced intensity, and possibly a shift, of the 2ν7 by Ca2+ or Mg2+ points to a direct participation of the nitrogen overtone band due to the lack of Fermi resonance (upward atom in the cation solvation (Figure S7b). shift). This complex combination of mechanisms prevents any For the cation solvation number determination, we turn to unambiguous use of the carbonyl region. Raman spectroscopy and a few relatively intense bands for Instead, the out-of-phase wagging of the CH2 groups (ν15, each solvent, which are minimally affected by other vibration 1390 cm−1) and the fundamental skeletal breathing of the ring modes but highly sensitive to cation coordination (Figure 6): (ν7, 892 cm−1) are used, and they both shift to higher ring breathing mode for EC (at 894 cm−1), γCO for PC wavenumbers as the Ca-salt concentration increases (Figure (sum of two bands at 706 and 712 cm−1), and δOC−N for S7a). At 1.5 M, these “free” solvent bands are almost gone, DMF (at 658 cm−1). At this point, we also include suggesting that almost all ECs are coordinated by the Ca2+ fluoroethylene carbonate (FEC) as a model solvent with a cations. As compared to Li-salt-based electrolytes, where free very low donor number, and the band analyzed is the ring EC is found even at 3 M,35 the Ca2+ cations are indeed breathing mode (at 905 cm−1). expected to have larger first solvation shells. For all regions, the free solvent band shifts toward higher The carbonyl region of the IR spectra of PC-based wavenumbers upon cation coordination, and by deconvolution electrolytes is laden with the same problems as those for EC, of the spectra (Figure S8), the solvation numbers (SNs) of the ji zyzjij Cs zyz but with the additional disadvantage that the lower symmetry cations were calculated using eq 1 SN = jjj zj z j Abound + θA free zzjj CCa 2+ zz of the PC produces, in general, broader bands.36 For the PC Abound k {k { electrolytes, we indeed discern that all of the same modes are affected by Ca2+ coordination, but the shifts are not sufficiently (1) large to allow us to correctly determine the contributions from where CS and CCa2+ represent the molar concentrations of the free and coordinated PC, respectively. solvent and Ca-salt, respectively. The parameter, θ, is the ratio In the IR spectra of DMF electrolytes (Figure 5), the CO of the Raman activity of free and coordinated solvent stretching mode gets clearly shifted toward lower wave- molecules as obtained by density functional theory (DFT) numbers upon coordination by Li+, Mg2+, or Ca2+. However, calculations (θ = RAcoord./RAfree). the δ OC−N vibration at 658 cm−1 is a better option to Similarly, the number of TFSI anions coordinating each analyze the cation−DMF interaction, with a significantly larger cation, as a measure of the contact ion-pair concentration, was shift obtained for the Mg2+-containing electrolytes, attributed ij 2Acoord. yz calculated using eq 2 TFSI coordinated by M2 + = jjj zz to the more polarizing character of the magnesium cation, j Acoord. + A free zz resulting in a tighter first solvation shell. Significant differences k { in terms of solvation shell dynamics and desolvation energies (2) are, thus, expected between the Mg-, Ca-, and Li-based electrolytes. This is derived from eq 1 and considers the 2:1 TFSI/Ca2+ Ohashi et al. recently reported that formamide coordinates relation and, furthermore, assumes that the TFSI all breathing Ca2+ cations only through the oxygen atom,40 but the nitrogen mode (free at 740 cm−1) is equally Raman-active when of the amide group might also take part in the cation “coordinated” (θ = 1, confirmed for Li-based electrolytes44). 29528 DOI: 10.1021/acs.jpcc.9b07308 J. Phys. Chem. C 2019, 123, 29524−29532 The Journal of Physical Chemistry C Article Figure 6. Raman spectra of Ca(TFSI)2 in (a) PC, (b) EC, (c) DMF, and (d) FEC, at different salt concentrations. * denotes free solvent, while ° denotes Ca2+-coordinated solvent. The band assigned to cation−anion contact ion pairs appears in all cases at 748 cm−1 and that of free TFSI at 740 cm−1. The number of solvent molecules in the first solvation shell highly dependent on the electrolyte solvent (Figure 7a), where (SNsolvent) decreases upon increasing the salt concentration DMF is most resilient to be displaced, followed by EC, PC, (Figure 7a), which can be explained by the displacement of and finally FECwhich correlates with their donor numbers solvent molecules by anions, thus forming contact ion pairs. (Table S1). The concentration threshold, when this process is triggered, is The formation of contact ion pairs can also be clearly observed in the Raman spectra of the electrolytes by the appearance of a band at 748 cm−1 attributed to Ca2+−TFSI complexes (Figure 6). Indeed, DMF is proven to be a very good solvent as no contact ion pairs were detected below 1.2 M salt concentration, while for carbonate solvents (EC and PC) they appeared already at 0.5 M. The formation of contact ion pairs also heavily depends on the salt employed; as TFSI is flexible and its charge is highly delocalized, it is less prone to contact ion-pair formation and is, thus, preferred over other anions (such as BF4−), which are involved in the solvation shell at lower concentrations (Figure S9). For Mg electrolytes, however, the polarizing character of Mg2+ promotes the formation of contact ion pairs even for 0.1 M Mg(TFSI)2 in EC (Figure 7b). While the formation of contact ion pairs is moderated by the solvent’s donor numbers, we propose that its permittivity (dielectric constant) plays a major role in the formation of solvent-separated ion pairs. In the Raman spectra of DMF electrolytes (a low-polarity solvent), the band of free TFSI (740 cm−1) grows more asymmetric in the range from 0.3−1.2 M. This asymmetry evidences a certain cation−anion interaction even if there is no direct contact. The presence Figure 7. SN of (a) solvent and (b) TFSI as a function of M(TFSI)2 of such solvent-separated ion pairs could, at least partially, salt concentration (M = Ca and Mg). explain why DMF-based electrolytes do not provide the 29529 DOI: 10.1021/acs.jpcc.9b07308 J. Phys. Chem. C 2019, 123, 29524−29532 The Journal of Physical Chemistry C Article highest ionicities (Figure 4b). On the other hand, the high degree of contact ion-pair formation and mono- vs bidentate permittivities of EC and PC effectively shield the cation charge, coordination were both found to evolve with salt concentration decreasing the tendency to form solvent-separated ion pairs, and solvent donor number. The amount of free solvent and hence, result in higher ionicities. Yet, as the donor molecules evolves accordingly, and divalent cation based numbers of EC and PC are lower, they are more easily electrolytes enter the domain of highly concentrated electro- displaced by the TFSI anions as the salt concentration lytes at concentrations as low as 1.2 M. We found that a high increases, forming contact ion pairs and possibly higher donor number solvent, such as DMF, allows for both low aggregates. viscosities, high ion conductivities, as well as few contact ion We also observe that the total SN (solvent + anion) pairs (although solvent-separated ion pairs can be envisaged, decreases as a function of salt concentration, and hence, each resulting in low ionicity). The lower probability for contact ion anion entering displaces more than one solvent molecule. If the pairs, together with the lower interaction energies within their coordination sites of the cation are to be constant, bidentate solvation shell, is fundamentally promising for better kinetics of coordination can be expected by either TFSI or the solvent. metal plating and cation insertion. For cyclic carbonates, bidentate coordination to Li+ and Na+ has been proposed for highly concentrated electrolytes,45 and as the band for the coordinated solvent is actually composed of ■ * ASSOCIATED CONTENT S Supporting Information two contributions shifted by ca. 5 cm−1 (Figure S10), bidentate The Supporting Information is available free of charge at coordination could also be the case for Ca2+. https://pubs.acs.org/doi/10.1021/acs.jpcc.9b07308. The evolution of the amount of free solvent molecules with salt concentration also appears to be solvent-dependent. For Physicochemical properties of salts, solvents, and the PC- and DMF-based electrolytes almost no free solvent calcium electrolytes; Vogel−Tammann−Fulcher (VTF) band can be detected (Figure 6a,c) for salt concentrations parameters of viscosities and conductivities; infrared and above 1.2 M, while they can still be detected for higher salt Raman spectra of calcium electrolytes; deconvolution of concentrations for both EC and FEC, most likely due to the Raman spectra; interaction/binding energies; Raman bidentate coordination of the solvent in the first case and to parameters for fitting (PDF) ■ the high degree of ion-pair formation in the latter. For Mg(TFSI)2 in EC, a larger free solvent feature was obtained as AUTHOR INFORMATION compared to the Ca system, in agreement with the higher degree of ion-pair formation. In all cases, the threshold Corresponding Authors concentration needed to reach the highly concentrated *E-mail: patrik.johansson@chalmers.se (P.J.). electrolyte domain is much lower for systems with divalent *E-mail: aponrouch@icmab.es (A.P.). cations as compared to monovalent cation based electrolytes ORCID (>3 M), in agreement with the higher coordination number of J. D. Forero-Saboya: 0000-0002-3403-6066 Ca and Mg (typically > 6) than that of Li (typically 4).12,16 P. Johansson: 0000-0002-9907-117X Understanding the solvation structure of battery electrolytes A. Ponrouch: 0000-0002-8232-6324 is a crucial step in the quest for electrolyte optimization. Indeed, the formation of cation−anion pairs effectively Notes The authors declare no competing financial interest. ■ diminishes the amount of charge carriers, and a cation in solution needs to get rid of its solvation shell before plating or insertion can take place at the anode and cathode, respectively. ACKNOWLEDGMENTS The ease with which the cation desolvates is related to its Funding from the European Union’s Horizon 2020 research binding energy with other species. Binding energies were and innovation program H2020 is acknowledged: European calculated for the M−X complexes (Table S6), and much Research Council (ERC-2016-STG, CAMBAT grant agree- larger energies were obtained for the ion pairs; M−TFSI with ment No. 715087), FETOPEN-1-2016-2017 (CARBAT, grant 1162 and 1442 kJ mol−1 for M = Ca and Mg, respectively, as agreement No. 766617), and H2020-MSCA-COFUND-2016 compared to the cation−solvent interactions; M−EC with 380 (DOC-FAM, grant agreement No. 754397). A.P. is grateful to and 544 kJ mol−1 for M = Ca or Mg, respectively. Thus, as far the Spanish Ministry for Economy, Industry and Competitive- as interfacial processes are concerned, Ca-based systems ness Severo Ochoa Program for Centers of Excellence in R&D present prospects for better power performance than Mg. (SEV-2015-0496). P.J. acknowledges financial support from For both systems, a solvation shell composed mainly of solvent the Swedish Energy Agency (contract #37671-1, Next molecules is preferred (absence of ion pairs), which can be Generation Batteries) and Chalmers Battery Initiative, part of achieved by using high donor number solvents and moderate the profile Materials for Energy Applications jointly managed salt concentrations. by the Areas of Advance Materials Science and Energy at the ■ CONCLUSIONS Ca-based electrolytes were studied, and the influence of salt, Chalmers University of Technology. The stimulating dis- cussions within the ALISTORE-ERI network are also highly appreciated. This work was done within the framework of the concentration, and the solvent was investigated in terms of Doctoral Degree Program in Materials Science by the ionic conductivity, viscosity, ionicity, and solvation structure Universitat Autònoma de Barcelona. We are grateful to Nicolas and in some cases compared with Li and Mg solutions. One of Monti for the elaboration of the TOC graphical abstract. the main issues for further development of divalent-cation- based electrolytes, in general and also true for Ca batteries, is the salt solubilityresulting in only a few Ca-salts possible and ■ REFERENCES (1) Xu, K. Nonaqueous Liquid Electrolytes for Lithium-Based only when combined with high-permittivity solvents. The Rechargeable Batteries. Chem. Rev. 2004, 104, 4303−4418. 29530 DOI: 10.1021/acs.jpcc.9b07308 J. Phys. Chem. C 2019, 123, 29524−29532 The Journal of Physical Chemistry C Article (2) Li, H.; Zhou, H. Enhancing the Performances of Li-Ion Batteries (24) Goedecker, S. Minima Hopping: An Efficient Search Method by Carbon-Coating: Present and Future. Chem. Commun. 2012, 48, for the Global Minimum of the Potential Energy Surface of Complex 1201−1217. Molecular Systems. J Chem. Phys. 2004, 120, 9911−9917. (3) Yoo, H. D.; Shterenberg, I.; Gofer, Y.; Gershinsky, G.; Pour, N.; (25) Schäfer, A.; Horn, H.; Ahlrichs, R. Fully Optimized Contracted Aurbach, D. Mg Rechargeable Batteries: An on-Going Challenge. Gaussian Basis Sets for Atoms Li to Kr. J. Chem. Phys. 1992, 97, Energy Environ. Sci. 2013, 6, 2265−2279. 2571−2577. (4) Canepa, P.; Sai Gautam, G.; Hannah, D. C.; Malik, R.; Liu, M.; (26) Becke, A. D. Density-Functional Exchange-Energy Approx- Gallagher, K. G.; Persson, K. A.; Ceder, G. Odyssey of Multivalent imation with Correct Asymptotic Behavior. Phys. Rev. A 1988, 38, Cathode Materials: Open Questions and Future Challenges. Chem. 3098−3100. Rev. 2017, 117, 4287−4341. (27) Zhao, Y.; Truhlar, D. G. The M06 Suite of Density Functionals (5) Selis, S. M.; Wondowski, J. P.; Justus, R. F. A High-Rate, High- for Main Group Thermochemistry, Thermochemical Kinetics, Energy Thermal Battery System. J. Electrochem. Soc. 1964, 111, 6−13. Noncovalent Interactions, Excited States, and Transition Elements: (6) Arroyo-de Dompablo, M. E.; Ponrouch, A.; Johansson, P.; Two New Functionals and Systematic Testing of Four M06-Class Palacín, M. R. Achievements, Challenges, and Prospects of Calcium Functionals and 12 Other Functionals. Theor. Chem. Acc. 2008, 120, Batteries Chem. Rev. 2019, DOI: 10.1021/acs.chemrev.9b00339. 215−241. (7) Staniewicz, R. J. A Study of the Calcium-Thionyl Chloride (28) Hehre, W. J.; Ditchfield, R.; Pople, J. A. SelfConsistent Electrochemical System. J. Electrochem. Soc. 1980, 127, 782−789. Molecular Orbital Methods. XII. Further Extensions of Gaussian (8) Muldoon, J.; Bucur, C. B.; Gregory, T. Quest for Nonaqueous Type Basis Sets for Use in Molecular Orbital Studies of Organic Multivalent Secondary Batteries: Magnesium and Beyond. Chem. Rev. Molecules. J. Chem. Phys. 1972, 56, 2257−2261. 2014, 114, 11683−11720. (29) Francl, M. M.; Pietro, W. J.; Hehre, W. J.; Binkley, J. S.; (9) Aurbach, D.; Skaletsky, R.; Gofer, Y. The Electrochemical Gordon, M. S.; DeFrees, D. J.; Pople, J. A. Self-Consistent Molecular Behavior of Calcium Electrodes in a Few Organic Electrolytes. J. Orbital Methods. XXIII. A Polarization-Type Basis Set for Second- Electrochem. Soc. 1991, 138, 3536−3545. Row Elements. J.Chem. Phys. 1982, 77, 3654−3665. (10) Lu, Z.; Schechter, A.; Moshkovich, M.; Aurbach, D. On the (30) Bochevarov, A. D.; Harder, E.; Hughes, T. F.; Greenwood, J. Electrochemical Behavior of Magnesium Electrodes in Polar Aprotic R.; Braden, D. A.; Philipp, D. M.; Rinaldo, D.; Halls, M. D.; Zhang, J.; Electrolyte Solutions. J. Electroanal. Chem. 1999, 466, 203−217. Friesner, R. A. Jaguar: A High-Performance Quantum Chemistry (11) Ponrouch, A.; Frontera, C.; Bardé, F.; Palacín, M. R. Towards a Software Program with Strengths in Life and Materials Sciences. Int. J. Calcium-Based Rechargeable Battery. Nat. Mater. 2016, 15, 169. Quantum Chem. 2013, 113, 2110−2142. (12) Wang, D.; Gao, X.; Chen, Y.; Jin, L.; Kuss, C.; Bruce, P. G. (31) Jaguar, version 9.1; Schrödinger, LLC: New York, 2019. Plating and Stripping Calcium in an Organic electrolyte. Nat. Mater. (32) Araujo, R. B.; Johansson, P., Manuscript in Preparation. 2018, 17, 16. (33) Samuel, D.; Steinhauser, C.; Smith, J. G.; Kaufman, A.; Radin, (13) Son, S.-B.; Gao, T.; Harvey, S. P.; Steirer, K. X.; Stokes, A.; M. D.; Naruse, J.; Hiramatsu, H.; Siegel, D. J. Ion Pairing and Norman, A.; Wang, C.; Cresce, A.; Xu, K.; Ban, C. An Artificial Diffusion in Magnesium Electrolytes Based on Magnesium Borohy- Interphase Enables Reversible Magnesium Chemistry in Carbonate dride. ACS Appl. Mater. Interfaces 2017, 9, 43755−43766. Electrolytes. Nat. Chem. 2018, 10, 532−539. (34) Walden, P. Ü ber Organische Lôsungs- Und Ionisiernngsmittel. (14) Ponrouch, A.; Palacin, M. R. On the Road toward Calcium- III. Teil: Innere Reibnng Und Deren Zusammenhang Mit Dem Based Batteries. Curr. Opin. Electrochem. 2018, 9, 1−7. Leitvermögen. Z. Phys. Chem. 1906, 55, 207−249. (15) Ponrouch, A.; Bitenc, J.; Dominko, R.; Lindahl, N.; Johansson, (35) Seo, D. M.; Reininger, S.; Kutcher, M.; Redmond, K.; Euler, W. P.; Palacin, M. R. Multivalent Rechargeable Batteries. Energy Storage B.; Lucht, B. L. Role of Mixed Solvation and Ion Pairing in the Mater. 2019, 20, 253−262. Solution Structure of Lithium Ion Battery Electrolytes. J. Phys. Chem. (16) Tchitchekova, D. S.; Monti, D.; Johansson, P.; Bardé, F.; C 2015, 119, 14038−14046. Randon-Vitanova, A.; Palacín, M. R.; Ponrouch, A. On the Reliability (36) Jankowski, P.; Grzegorzewska, K.; Szabłowska, A.; Piszcz, M.; of Half-Cell Tests for Monovalent (Li+, Na+) and Divalent (Mg2+, Dranka, M.; Ż ukowska, G. Z.; Kalita, M. Role of Propylene Carbonate Ca2+) Cation Based Batteries. J. Electrochem. Soc. 2017, 164, A1384− Chirality on Physicochemical Properties of the Corresponding Ion A1392. Conductors. Electrochim. Acta 2015, 175, 240−246. (17) Okoshi, M.; Yamada, Y.; Komaba, S.; Yamada, A.; Nakai, H. (37) Schreiner, C.; Zugmann, S.; Hartl, R.; Gores, H. J. Fractional Theoretical Analysis of Interactions between Potassium Ions and Walden Rule for Ionic Liquids: Examples from Recent Measurements Organic Electrolyte Solvents: A Comparison with Lithium, Sodium, and a Critique of the So-Called Ideal KCl Line for the Walden Plot. J. and Magnesium Ions. J. Electrochem. Soc. 2017, 164, A54−A60. Chem. Eng. Data 2010, 55, 1784−1788. (18) Xu, K.; von Cresce, A.; Lee, U. Differentiating Contributions to (38) Fortunato, B.; Mirone, P.; Fini, G. Infrared and Raman Spectra “Ion Transfer” Barrier from Interphasial Resistance and Li+ Desolva- and Vibrational Assignment of Ethylene Carbonate. Spectrochim. Acta, tion at Electrolyte/Graphite Interface. Langmuir 2010, 26, 11538− Part A 1971, 27, 1917−1927. 11543. (39) Angell, C. L. The Infra-Red Spectra and Structure of Ethylene (19) Xu, K. Electrolytes and Interphases in Li-Ion Batteries and Carbonate. Trans. Faraday Soc. 1956, 52, 1178−1183. Beyond. Chem. Rev. 2014, 114, 11503−11618. (40) Ohashi, K.; Hikiishi, N.; Takeshita, H. Infrared Spectroscopic (20) Gauthier, M.; et al. Electrode−Electrolyte Interface in Li-Ion and Computational Studies on Formamide Solutions of Ca2+. Batteries: Current Understanding and New Insights. J. Phys. Chem. Vibrational Frequencies of Formamide and Modes of Coordination Lett. 2015, 6, 4653−4672. to Ca2+. Spectrochim. Acta, Part A 2019, 206, 112−119. (21) Minato, T.; Abe, T. Surface and Interface Sciences of Li-Ion (41) Bukowska, J. Raman and Infrared Studies of Interactions Batteries: -Research Progress in Electrode−Electrolyte Interface. Prog. between Amides and Ions. J. Mol. Struct. 1983, 98, 1−10. Surf. Sci. 2017, 92, 240−280. (42) Bukowska, J.; Miaskiewicz, K. Infrared, Raman, and Cndo (22) Ahlrichs, R.; Bär, M.; Häser, M.; Horn, H.; Kölmel, C. Studies of the Effect of Ions on the Electron Density Distribution in Electronic Structure Calculations on Workstation Computers: The Formamide. J. Mol. Struct. 1981, 74, 1−10. Program System Turbomole. Chem. Phys. Lett. 1989, 162, 165−169. (43) Shastri, A.; Das, A. K.; Krishnakumar, S.; Singh, P. J.; Sekhar, B. (23) Larsen, A. H.; et al. The Atomic Simulation Environmenta N. R. Spectroscopy of N,N-Dimethylformamide in the VUV and IR Python Library for Working with Atoms. J. Phys.: Condens. Matter Regions: Experimental and Computational Studies. J. Chem.Phys. 2017, 29, No. 273002. 2017, 147, No. 224305. 29531 DOI: 10.1021/acs.jpcc.9b07308 J. Phys. Chem. C 2019, 123, 29524−29532 The Journal of Physical Chemistry C Article (44) Lassègues, J.-C.; Grondin, J.; Aupetit, C.; Johansson, P. Spectroscopic Identification of the Lithium Ion Transporting Species in LiTFSI-Doped Ionic Liquids. J. Phys. Chem. A 2009, 113, 305−314. (45) Flores, E.; Åvall, G.; Jeschke, S.; Johansson, P. Solvation Structure in Dilute to Highly Concentrated Electrolytes for Lithium- Ion and Sodium-Ion Batteries. Electrochim. Acta 2017, 233, 134−141. 29532 DOI: 10.1021/acs.jpcc.9b07308 J. Phys. Chem. C 2019, 123, 29524−29532