Computational and experimental investigation of the effect of cation structure on the solubility of anionic flow battery active-materials

Recent advances in clean, sustainable energy sources such as wind and solar have enabled significant cost improvements, yet their inherent intermittency remains a considerable challenge for year-round reliability demanding the need for grid-scale energy storage. Nonaqueous redox flow batteries (NRFBs) have the potential to address this need, with attractive attributes such as flexibility to accommodate long- and short-duration storage, separately scalable energy and power ratings, and improved safety profile over integrated systems such as lithium-ion batteries. Currently, the low-solubility of NRFB electrolytes fundamentally limits their energy density. However, synthetically exploring the large chemical and parameter space of NRFB active materials is not only costly but also intractable. Here, we report a computational framework, coupled with experimental validation, designed to predict the solubility trends of electrolytes, incorporating both the lattice and solvation free energies. We reveal that lattice free energy, which has previously been neglected, has a significant role in tuning electrolyte solubility, and that solvation free energies alone is insufficient. The desymmetrization of the alkylammonium cation leading to short-chain, asymmetric cations demonstrated a modest increase in solubility, which can be further explored for NRFB electrolyte development and optimization. The resulting synergistic computational–experimental approach provides a cost-effective strategy in the development of high-solubility active materials for high energy density NRFB systems.


Introduction
Current technological improvements, growing global concern for CO 2 emissions, and the rapid increase of energy demand have contributed to a signicant lowering of the cost for renewable energy generation, with wind and solar sources being the main contributors. 1-3 Despite these improvements, the inherent intermittency of these sources still poses a considerable challenge towards the year-round reliability of these renewable energy sources, necessitating grid-scale energy storage. 2,[4][5][6][7] Redox ow batteries (RFBs) are considered one of the most promising electrochemical technologies for the largescale storage of renewable electrical energy. 8,9 A key advantage of RFB systems is the highly scalable, uncoupled power rating (area of the electrode) and energy capacity (amount of electrolyte), providing excellent exibility for a range of stationary storage applications. Of particular interest in the eld is the development of nonaqueous redox ow battery (NRFB) systems. Polar organic solvents have a wide electrochemical stability window, allowing higher cell voltages and energy densities than water. 8,[10][11][12] Despite having great potential for grid-scale application, NRFB technology is still in its infancy with technical hurdles that must be overcome, including low solubility of activematerials, which limits energy density, and poor stability, leading to low cyclability. [13][14][15][16] To date, most research has focused on improving the performance of individual components such as high-energy-capacity electrolytes. 17 Moreover, electrolyte viscosity, which is correlated with conductivity and diffusivity, is another important issue in the development of NRFB systems and is gaining attention in the community. [18][19][20] Electrolyte systems with low viscosity, high diffusivity, and high ionic conductivity are ideal. 12,21 The demand for higher energy density systems, pushing towards higher active material concentration, has resulted in higher viscosity in NRFB electrolytes which hinders practical applications due to pumping losses and decreased conductivity. 19 We recently elaborated on an NRFB active-material, leveraging a bio-inspired redox molecule 5,22,23 known as Amavadin (Fig. 1a), biosynthesis of which has evolved naturally in mushrooms of the Amanita genus under selection pressure for strong and specic vanadium binding. As a result, this compound and its analogs exhibit the highest stability constants ever measured for a V 4+ ion. 24 Using inexpensive reagents, vanadium (4+) bis-hydroxyiminodiacetic acid ([VBH] 2À ) (Fig. 1b) has been synthesized at a large scale which exhibits high chemical stability, even under cycling at high current and to deep states-of-charge. 23 More recently, we demonstrated a synthetic strategy to increase the solubility of VBH-based NRFB materials while still maintaining stability during deep cycling for extended time periods. 5 This highlights the potential of VBH-based NRFB materials as a molecular scaffold for the development of next-generation ow battery electrolytes.
The solubility of active-materials is a critical factor in determining the energy density of NRFBs. 2,25 However, synthetically exploring a large parameter space to identify soluble species empirically is an inefficient approach. Solubility can be calculated by comparing the stabilization of the gasphase ions upon solvation ðDG * sol Þ with that of the organization of the ions into a crystal lattice ðDG latt Þ (see Fig. 2). In the well-known thermochemical Born-Fajans-Haber cycle (Fig. 2), 26 the latter factor can be expressed in terms of free energy of sublimation ðDG * sub Þ; which describes the energetics of a material's phase change from solid to gas, and is approximately the negative of DG latt (see ESI, † eqn (2)). This thermochemical cycle has been utilized in previous solubility studies in pharmaceutics, [27][28][29] druglike molecules, 30,31 and systems like LiO 2 and Li 2 O 2 . 32 These works involved neutral molecules and disregarded some thermochemical corrections. With full thermochemical corrections, a similar approach was used by Chen and Bryantsev to predict the melting points of ionic liquids. 33 Despite the straightforward description of solubility, accurate calculations of DG latt have proven challenging and computationally expensive. 33 As such, recent computational approaches to determine RFB active-material solubility have focused only on DG * sol : 5,25 In our previous work, 5 we demonstrated that solvation energies are insufficient to describe the experimental solubility trend for VBH with long-chain alkylammonium counter cations. This implies a critical role for DG latt in determining the solubility of active materials in NRFB electrolytes and highlights the importance of its accurate calculation in any quantitively useful prediction of solubility.
In this paper, we report a computational framework designed to predict the solubility trends of electrolytes, tracking both the thermodynamic contributions due to free energy of active-material/solvent interactions and free-energy of the active-material crystal lattice. These calculations make use of extensive information on the solid-state geometries of VBH with various cations, provided by eight crystal structures, for input geometries. These structures, correspond to the reduced and oxidized forms of [VBH] with symmetrical alkylammonium cations, having one to four carbon atoms in each chain. Two of these are newly reported, herein (see Fig. 3b, c and Table S6 †), and ve of these have been reported previously. 5,22,23 One of the eight structures, corresponding to [N 3333 ] 2 [VBH], was of high enough quality to obtain structural information (unit cell, Cartesian coordinates, and bonding parameters for starting    (Table 1). Solubility is also affected by the oxidation state, with +4 states generally exhibiting greater solubility than their +5 counterparts, with notable variability in various solvents. Because experimentally exploring all possible combinations of alkyl substituents on the supporting cations and their interactions with solvents to identify highly soluble species empirically is an inefficient approach, our results suggest that a predictive theoretical framework could be benecial to NRFB electrolyte development and optimization. This led us to explicitly explore the interplay between competing trends in solvation energy and lattice enthalpy, which was implied by our previous work, 5 using a combined computational and synthetic strategy.
2.2 Alkyl ammonium cation substitution to explore steric and electrostatic effects on solubility To explore other possible soluble alkylammonium cations, we hypothesized that short-chain, asymmetric alkylammonium cations would exhibit better solubility due to the destabilization of the crystal structure ði:e:; less negative DG latt Þ: We employed desymmetrization by substitutions on the smallest ([N 1111 ] + ) and largest ([N 4444 ] + ) cations that were used in experiment ( Fig. 3a). At most two substitutions were performed in each set, generating groups of monosubstitutions, symmetric disubstitutions, and asymmetric disubstitutions. To probe the steric effect of alkyl chain length, we used ethyl (2), n-propyl (3), and n-butyl (4) for [N 1111 ] + . For the larger cation, [N 4444 ] + , the steric effects were also probed using a methyl (1) substituent and the relatively bulkier sec-butyl (s) and tert-butyl substituents (t). To investigate electrostatic effects, we modied the [N 4444 ] + with triuoromethyl (c) substituent. Fig. 3e-h shows representative structures of the symmetric and desymmetrized [N 4444 ] + .

Simplied crystal models
In order to calculate the solubility by rst-principles, we rst developed a simplied model for the crystal structures of Table 1 Experimental solubilities of the [N xxxx ] y [VBH] active materials in the +4 (where y ¼ 2) and +5 (where y ¼ 1) states in various solvents. Values correspond to the maximum concentration, in mol L À1 , of VBH in the electrolyte as determined by UV-Vis spectroscopy a Solubilities from ref. 5. interest. The unit cells of the crystallized active materials, resolved from X-ray diffraction, ranged from two (Z ¼ 2) to 16 (Z ¼ 16) formula units corresponding to system sizes of 126 and 736 atoms, respectively, which poses expensive computational cost. Solvent molecules were also observed in most crystal structures which adds complexity to the systems (see Table S1 † for parameters). To reduce the complexity and the cost of the calculations, we simplied the systems into single formula unit cells (Z ¼ 1) and monitored the response of the corresponding energetics of the system to this modication.
To benchmark our calculations, we compared the calculated molar volumes of the simplied crystals to the molar volumes of the experimentally crystallized active materials for [N xxxx ] y [VBH] (where x ¼ 1, 2, 3, 4 and y ¼ 1, 2). We note that bulk morphological variation is neglected in this approach resulting in systematically high lattice energies, especially for crystals with large Z. Nevertheless, the simplied approach employed here, which is empirically calibrated by solubility measurements, provides a middle ground for both computational cost and accuracy. It is an excellent tool for exhaustive screening and exploration of materials to a depth that would not be possible by synthetic approaches. Table S7 † lists the formula unit Cartesian coordinates for each of the simplied crystals in both +4 and +5 redox states.

Lattice free energy ðDG
latt Þ and free energies of sublimation ðDG * sub Þ Sublimation of crystalline materials is an endergonic process under standard conditions ðDG * sub . 0Þ: As the magnitude of DG * sub increases, sublimation becomes less favorable leading to a more stable crystal. Eqn (1) implies that the greater the magnitude of DG * sub ; the lower the solubility. Since DG * sub is approximately the negative of the DG latt (ESI eqn (2) †), a material would become more soluble as the DG latt becomes less negative. Fig. 4a and b illustrate the performance of the models in predicting the trends of DG latt (data shown in Table S3    196 kJ mol À1 and 15-45 kJ mol À1 lower than the other symmetric compounds, respectively (Table S3 †). This difference between the scale of the relative DG * latt as a response to the change of cation can be attributed mainly to the electrostatic potential of the V 4+ and V 5+ states. Increasing cation size destabilizes the coulombic interactions in the lattice by increasing ion separation and interactions with neighboring ions, which has a greater effect in [VBH] 2À .
In general, both the alkyl chain length and bulkiness promoted lattice destabilization. However, because of the greater electrostatic potential of the dianionic [

Free energies of solvation, DG * sol
To determine and quantify the role of DG * sol on the solubility prediction and compare with the magnitude of DG latt ; we calculated the DG * sol of the symmetric and asymmetric VBH active materials in several solvents ( Fig. 5 and Table 2, full list of solvents in Table S2, † and heat maps of V 4+ and V 5+ compounds in Tables S4 and S5 †). Of the three solvents (THF, MeCN, and DMSO) that were used in the experimental solubilities, our calculations predicted that the solvent with the most favorable DG * sol for V 4+ compounds is MeCN, while for V 5+ compounds, it is DMSO. These are consistent with the general experimental solubility trends (i.e., the solubilities for V 4+ are highest in MeCN, while V 5+ solubilities are highest in DMSO, Table 1). However, DG * sol alone could not explain the solubility trends with respect to the size of the alkylammonium chain. For example, for the four symmetric V 4+ compounds in MeCN with increasing chain length ([N 1111 ] 2 [VBH] to [N 4444 ] 2 [VBH]), DG * sol were determined to be À1107, À1060, À1045, À1050 kJ mol À1 (Table 2), contrary to experimental solubilities of 0.004, 0.330, 1.090, and 0.800 (Table 1), respectively.
To understand this better, within the implicit solvation model based on density (SMD) approach, we can decompose the DG * sol further into two main contributions: (1) the DG ENP and (2) DG CDS . The DG ENP takes into account the bulk-electrostatics from the solute electronic kinetic and electronic-nuclear coulombic energies in the presence of the solvent (EN) and the solution polarization (P) free energy. 34,35 The DG CDS accounts for the non-electrostatic effects of cavitation (C, energy required to make room in the solvent for the solute), dispersion (D, change in dispersion energy upon dissolution), and changes in the solvent structure (S, energetic and entropic effects from structural changes in the solvent). 34,35 By evaluating DG ENP and DG CDS separately, we can gain insights into the effects of the structural modications of the cations to the DG * sol : As mentioned earlier, in all solvents and for both +4 and +5 states considered, DG * sol increases (more positive) with cationic size ( Table 2, Fig. 5a and b), indicating that solvation becomes less favorable. This could be attributed mainly to the reduced bulk-electrostatic interactions (DG ENP ) as the cation size is increased (Fig. 5c). The smaller cations have more favorable bulk-electrostatic interactions (Fig. 5c). Electrostatic potential maps from natural population analysis charges reveal that the unit charge of the cations is distributed over the hydrogens of the carbon atoms adjacent to the nitrogen atom (Fig. S2 †). In  Fig. 5g and h). An inverse trend is observed for the DG CDS of the symmetric cations (Fig. 5d) where increasing the cation size improves stabilization from non-electrostatic interactions. However, unlike the DG ENP , there is no apparent diminishing effect for the DG CDS aer three carbon atoms (see Fig. 5g (Fig. S1a and b †) as the symmetric ones with a similar diminishing DG ENP penalty beyond three carbon atoms. For the [N 44xx ] + modications ( Fig. S1c and d †), the use of the bulkier sec-and tert-butyl appears to restrict solvent access further, leading to less favorable DG * sol when compared to the symmetric [N 4444 ] + . In contrast, the smaller methyl substituent allowed for better solvent access, which resulted in the greatest improvement, followed by the -CF 3 , of DG * sol in the [N 44xx ] + group.
The effect of hydrogen bonding interactions between the VBH anions and protic solvents signicantly improved the DG ENP (Fig. 5e). As expected, the response of DG ENP towards changing the solvent is similar for both [VBH] 2À and [VBH] À , which differed only in the scale due to the charge difference. Also, preference towards MeCN over DMSO is observed for the VBH anions and is due to the slight acidity of the hydrogen atoms of MeCN (Abraham's hydrogen bond acidity, a MeCN ¼ 0.07) despite being classied as aprotic. Almost identical values of DG CDS (Fig. 5f) were observed for both oxidation states of the VBH anion. While solvents with acidic protons should be considered for NRFB applications only with caution, since they would be expected to affect the electrochemical stability windows negatively, this is nevertheless an interesting observation and an important consideration in optimizing the solvent to be used in an electrolyte formulation.

Solvent proticity and viscosity considerations
The choice of solvent is a crucial aspect of NRFBs as it affects not only its chemical but also its operational and physical properties. Our DG * sol results suggest that alcohols could improve the solubilities of both V 4+ and V 5+ crystals. The more polar solvents N,N-dimethylacetate and N,N-dimethylformamide, and the less polar propanonitrile, are also promising solvents for the V 5+ crystals (Fig. 8). Although alcohols appear to be promising solvents based on solubility, their protic nature could pose unwanted degradation reactions in prolonged cycling. Their affinity towards absorbing atmospheric moisture could also prove to be problematic in longer operations. 36 To demonstrate the viscosity behavior of VBH electrolytes, we measured the viscosity of [N 4444 ] 2 Fig. 9). Our results show that viscosity generally increases with solution concentration. Moreover, the viscosity of DMSO electrolyte depends more steeply on concentration than the MeCN electrolyte, which could considerably impede the ow performance at the concentrations that we wanted to achieve with increased pumping power requirements and slower reaction kinetics. Our ndings demonstrate the vital role of the viscosity of the electrolyte solution in choosing the appropriate solvent. A rough approximation of the resulting viscosities can be obtained from the kinematic viscosities (Table S2 †) of all the solvents used in the calculations. Although we predicted that N,N-dimethylacetate, N,N-dimethylformamide, and propanonitrile are favorable solvents from the perspective of activematerial solubility, these solvents' viscosities are signicantly higher than that of MeCN, which will likely hinder their practical applications.

Conclusion
We demonstrated a rst-ever, rst-principles prediction of solubilities of nonaqueous ow-battery active-materials, incorporating both the lattice and solvation energies. The computational predictions were compared and validated with experimental data. We found that the solvation energy alone is insufficient and can even be misleading as a tuning parameter to improve solubility, especially in ionic systems. Lattice energy has a more signicant effect on solubility. The desymmetrization of the alkylammonium cation leading to short-chain, asymmetric cations demonstrated a modest increase in solubility, which can be further explored for NRFB electrolyte development and optimization. Based on the framework developed herein, investigations on structural modication of the active-material, with the goal of synergistic improvements to solubility, are on-going. While changes in the reduction potential arising from varying the counter ion are expected to be minimal, based on previous investigations, 37 we believe that tuning the structure of the active-material will allow simultaneous improvements to solubility and redox properties. Our ndings demonstrated the critical role of the viscosity of the electrolyte solution in choosing the appropriate solvent.
Design strategies should take into account the solid-state energetics. In our efforts to nd VBH compounds with high solubility, we found that, while desymmetrization of the cation can be important, it does not always lead to a more soluble active material. One design strategy for improving VBH solubility using alkyl ammonium cations is by increasing the alkyl chain lengths, which is favorable with respect to lattice free energy, but unfavorable with respect to solvation free energy because of the steric hindrance imparted by the larger cations. In this, and in previous work, 5 we demonstrate that the former effect outweighs the latter, giving rise to the overall improvement to solubility observed for longer alkyl-chain cations. With the computational investigations reported herein, we elaborate that the electrostatic interaction between the cations and the solvent becomes less effective as the alkyl chain length is increased up until 3-carbon atoms, then plateaus. This implies that, as a design strategy for improving solubility, the use of alkyl ammonium cations with chainlengths of 4-carbon atoms, or more, should exhibit a reduced lattice free energy with a decreasingly signicant penalty to solvation free energy. Another design principle for improving VBH solubility is the use of bulky substituents. This effect is particularly pronounced in the V 5+ crystals where lattice electrostatic interactions are lower than in the V 4+ crystals. Compared to their straight-chain counterparts, sec-and tertbutyl substituents showed appreciable reduction of DG * sub and slight increase in DG * sol : Nevertheless, these responses from bulky substituents, together with alkyl chain length, are promising, and will be explored in our future efforts.
This work is an in-road toward VBH active materials with very high solubility and potentially those that are liquid at operating temperature. If viscosity concerns could be mitigated, the production of such liquid active materials would signicantly increase energy density and could eliminate the need for supporting electrolytes. Furthermore, the establishment of this theoretical framework, coupled with experimental verications, opens avenues for machine learning models and other computational screens of solution properties (e.g., diffusion rates, viscosities, and ionic conductivities), both of which are parts of our in-road strategy for designing high energy density NRFB active materials. Because the methods involved in this computational protocol are available in most computational chemistry programs, this strategy can be applied to any RFB chemistry including organic aqueous systems, with the cost and complexity of such calculations largely depending on the composition and size of the system.

Computational
The intrinsic solubility (S 0 ) of a substance is directly related to the difference in its stabilities in both the solid and solvated states. This relationship is demonstrated from the well-known Born-Fajans-Haber thermochemical correlation 26 (Fig. 2), which translates to where DG * dis is the free energy of dissolution and is the sum of the free energies of sublimation ðDG * sub Þ and solvation ðDG * sol Þ while R, T, and V m are the gas constant, temperature, and the solid's molar volume, respectively. In this cycle, a substance rst sublimes into the gas phase, then gets solvated into the solution phase and the free energies associated with each process determine its dissolution. The free energy of sublimation is roughly the negative of the lattice free energy ðDG latt Þ. The details of the theory and corresponding equations are given in ESI †. [30][31][32][38][39][40] Starting geometries for single formula unit crystals of the systems with symmetric cations ([N 1111 ] + , [N 2222 ] + , [N 3333 ] + , and [N 4444 ] + ) were extracted from the X-ray crystal structures for both reduced and oxidized states. Five of these have been reported previously 5,22,23 and two are newly reported herein (Fig. 3 and Table S6 †). These crystals were then equilibrated using the method discussed below. The resulting optimized crystal structures of the [N 1111 ] + and [N 4444 ] + systems were then subjected to cation modication and re-optimization.
Plane-wave-based periodic density functional theory (DFT) calculations for the sublimation free energies were performed using Quantum Espresso v6.5 (QE). 41,42 The valence electronic states were expanded based on plane waves, and the corevalence interaction was described using the ultraso pseudopotential approach. The Perdew-Burke-Ernzerhof (PBE) 43 generalized gradient approximation functional with a 1225 eV and 9796 eV basis set cutoffs for the plane wave kinetic energy and the electron density, respectively, were used. Spinpolarization was used for the open-shell systems (V 4+ crystals; d 1 ), while non-polarized calculations were done for the closedshell systems (V 5+ crystals; d 0 ). For the bulk calculations, the Brillouin zone was sampled using a Monkhorst-Pack k-point mesh of 2 Â 2 Â 2 with grid offsets, resulting in less than 0.0001 eV change in the total energy compared to 3 Â 3 Â 3. The two-body dispersion interactions in the crystals were included using Grimme's DFT-D2 (ref. 44 and 45) method implemented in QE. The atomic coordinates and cell parameters were fully optimized to a 1 Â 10 À5 eV and a 1 Â 10 À5 eVÅ À1 force threshold. Given the size of the unit cells (containing 46 to 135 atoms), the phonon frequencies were only calculated at the Gamma point. 33,46 The free energy of solvation ðDG * sol Þ 47,48 was obtained using the implicit solvation model based on density (SMD) method 34 as implemented in Orca v4.2.1. 49 Each ion was rst optimized in the gas phase using the PBE functional and Grimme's DFT-D3 (ref. 50) dispersion correction with the Becke-Johnson damping (D3BJ) 51 with a def2-TZVP 52 basis set. SMD was then employed on the gas-phase structures to obtain the standard free energy of solvation. 47,53,54 A list of battery-relevant solvents used in this study is listed in Table S2. †

Experimental
Physical methods. UV-Vis spectra were collected using the Evolution 220 UV-visible spectrophotometer with a quartz cuvette of 1 cm path length. The molar extinction coefficients for oxidized and reduced species in all solvents except for DMSO were used as determined in our previous study. 5 Molar extinction coefficients for oxidized species in DMSO were determined from a calibration curve. Infrared spectra were collected on a Thermo Scientic Nicolet iS5 equipped with iD7ATR module and a diamond crystal. 1 H-and 13 C-NMRs were performed on Bruker AVANCE III HD 400 MHz High-Performance Digital NMR spectrometer operating at 400 MHz for 1 H NMR and 101 MHz for 13 C NMR. Data acquisition was performed on IconNMR 5.0.3, and spectra were processed in TopSpin 3.5 and Mnova. High-Resolution Mass Spectrometry (HRMS) was performed on a Waters ACQUITY UPLC Xevo QTOF high resolution mass spectrometer using electrospray ionization. X-ray crystallographic experiments were performed on a Bruker D8 Venture X-instrument, using Mo Ka radiation at 200 K. Data were corrected for absorption using SADABS. The structures were solved by direct methods. All non-hydrogen atoms were rened anisotropically by full matrix leastsquares on F 2 and all hydrogen atoms except those on water were placed in calculated positions with appropriate riding parameters. Further renement and molecular graphics were obtained using Bruker Suite of structural programs, 55 OLEX2, 56 and Mercury. 57 General. Hydroxylamine hydrochloride (Alfa Aesar), sodium hydroxide (Acros organics), chloroacetic acid (BTC), zinc acetate dihydrate (Acros Organics), calcium chloride (VWR), tetramethylammonium uoride tetrahydrate (Matrix Scientic), tetraethylammonium uoride (BTC), triethylamine (Acros Organics), n-butylbromide (Alfa Aesar), n-propylbromide (BTC), ferrocenium hexauorophosphate (Sigma Aldrich), and vanadyl (iv) acetylacetonate (BTC) were purchased from commercial sources and used as received. Information on the characterization of new materials and those that were reported previously, and were used for solubility studies herein, are provided below. Solvent used in solubility measurement were of anhydrous grade and purchased from Sigma Aldrich. Synthetic methods Zinc hydroxyiminodiacetate (ZnHIDA) ligand synthesis. Zinc hydroxyiminodiacetate (ZnHIDA) was synthesized utilizing a method we previously reported. 5 Hydroxylamine hydrochloride (34.7 g, 0.500 mol) was neutralized with 5 M NaOH (100 mL, 0.500 mol) in an ice bath maintaining the temperature of the reaction mixture between 0-4 C. In another ask, chloroacetic acid (94.5 g, 1.00 mol) was neutralized by dropwise addition of 5 M NaOH (200 mL, 1.00 mol) in an ice bath. The neutralized mixture of chloroacetic acid was then added dropwise to hydroxylamine solution. At the end, additional 200 mL 5 M NaOH was added dropwise to the reaction mixture and the solution was stirred for 72 h in an ice bath. Then, the pH of the mixture was brought to 4.01 by acidifying with 6 M HCl to which zinc acetate dihydrate (110 g, 0.500 mol) was added to the solution while stirring. Upon addition, the ligand precipitated immediately as a zinc salt. A white precipitate of ZnHIDA was then ltered, washed with cold water several times, and dried in vacuo. Calcium(II)vanadium(IV)-bis-hydroxyiminodiacetate (CaVBH). Calcium(II)vanadium(IV)-bis-hydroiminodiacetate (CaVBH) was synthesized using a method previously reported. 5 A suspension of ZnHIDA (89.0 g, 0.420 mol) in 500 mL deionized water was prepared, to which vanadyl(IV)acetylacetonate (55.7 g, 0.210 mol) was added and mixed well using a magnetic stir bar. Hydrochloric acid (80 mL, 6.0 M) was added dropwise to the stirring mixture. Calcium chloride dihydrate (30.8 g, 0.500 mol) was added to the solution and dissolved. 2-Propanol (2000 mL) was added to the resulting blue solution and stirred vigorously to facilitate the precipitation of CaVBH. The product was ltered and washed with 20 mL isopropanol followed by acetone and dried in vacuo for 24 h at room temperature (Yield: 90.0 g, 0.190 mol, 90%). UV-Vis (DMSO): l max ¼ 578 nm (3 ¼ 27.5 mol À1 cm À1 ). IR (n/cm À1 ): 3541 (w), 3321 (w), 2988 (w), 1589 (C]O, s).
Tetramethylammonium vanadium(IV)-bis-hydroxyiminodiacetate [N 1111 ] 2 VBH (1). CaVBH (7.15 g, 0.0151 mol) was added to a stirring mixture of tetramethylammonium uoride tetrahydrate (5.00 g, 0.0300 mol) in 40 mL ethanol. Aer 3 h of stirring, the solution was ltered to remove calcium uoride and unreacted CaVBH. Tetrahydrofuran (15 mL) was added to the ltrate and cooled in the refrigerator. Crystals were obtained aer 24 h in the refrigerator, which was then isolated via vacuum ltration and subsequently dried in vacuo for 24 h at room temperature (yield: 6.80 g, 88%). UV-Vis (DMSO): l max ¼ 579 nm (3 ¼ 30. General procedure for preparation of asymmetric alkylammonium VBH. First, asymmetric quaternary ammonium bromides were prepared using Menshutkin reaction 58 (Scheme 1). Thus prepared quat-bromides were converted to uoride salts using halogen exchange reaction as described in the literature. 59 Fluoride salts prepared in situ were directly reacted with CaVBH to afford corresponding quaternary ammonium VBH, as shown in reaction Scheme 1.
Trimethylpropylammonium bromide [N 1113 ]Br. In a round bottom ask containing 50 mL ethanol, trimethylamine (59.1 g, 50 wt% aqueous solution, 0.500 mol) and 1-bromopropane (30.8 g, 0.250 mol) were added. The mixture was reuxed at 40 C for 12 h under N 2 atmosphere. The solvent was removed using a rotatory evaporator. The resulting solid was repeatedly washed with ethyl acetate and diethyl ether and dried in Schlenk line (yield: 31.0 g, 0.170 mol, 68%  (3.43 g, 1.3 equiv., 0.0590 mol) and deionized water (0.135 mL, 4 wt% of KF) was added to the solution. The mixture was vigorously stirred for 1 h and ltered to remove the solid. Potassium uoride was added, stirred, and ltered two more times. Finally, the ltrate was concentrated, and any solid that appeared was ltered again. The yellowish oil (trimethylpropylammonium uoride) was dissolved in 10 mL acetonitrile to which CaVBH (6.80 g, 0.0144 mol) was added and stirred for 30 min. The mixture was centrifuged, and the blue solution layer was concentrated under reduced pressure. Addition of diethyl ether afforded blue crystals of the title compound  Butyltrimethylammonium bromide (7.16 g, 0.0365 mol) was dissolved in 20 mL methanol. To the solution, potassium uoride (2.65 g, 0.0511 mol) and 0.105 mL deionized water (4 wt% of KF) was added, stirred for an hour, and ltered to separate the solid. The same amount of potassium uoride was added to the ltrate and stirred for 1 h before ltering out the solid. The process was repeated one more time. Finally, the ltrate was concentrated to 10 mL using a stream of nitrogen gas. The solution was ltered again to remove any solid precipitated. The brownish-yellow oil ([N 1114 ]F$xH2O) was dissolved in 10 mL acetonitrile to which CaVBH (4.75 g, 0.0100 mol) was added and stirred for 2 h. The mixture was centrifuged, and the liquid layer was concentrated under reduced pressure. Diethyl ether was added to aid precipitation. Blue solid was ltered and dried in vacuo. (Yield 5.20 g, 0.00904 mol, 90%) single crystal suitable for X-ray diffraction was prepared by layering diethyl ether into a solution of compound in propylene carbonate. UV-Vis (propylene carbonate): l max ¼ 579 nm (3 ¼ 27.5 mol À1 cm À1 ). IR (n/cm À1 ): 3444 (w), 3029 (w), 2961 (w), 1606 (s), 1493 (w), 1378 Trimethylpropylammonium vanadium(V)-bishydroxyiminodiacetate [N 1113 ]VBH (7). Compound 3 (2.19 g, 0.00401 mol) was dissolved in 3 : 1 MeCN/EtOH to make 20 mL solution. FcPF 6 (1.33 g, 0.00401 mol) was added to the solution and stirred for 20 min. 20 mL diethyl ether was added to aid precipitation of the complex and the precipitate was ltered which was then washed with 10 mL portion of diethyl ether ve times until trace of ferrocene was removed. Finally, red crystalline powder was dried in vacuo (yield 1.53 g, 87%). UV-Vis (MeCN): l max ¼ 495 nm (3 ¼ 241.9 mol À1 cm À1 ). IR (n/cm Viscosity measurement. The viscosity measurements were collected using the Cannon Instruments Semi-Micro Viscometer. Concentrations of 0.01 M-0.55 M were used when collecting the viscosity data at ambient temperature (25 C). The U-tube style viscometer allowed for the collection of kinematic viscosity in centiStokes (cSt) which was then converted to dynamic viscosity in centiPoise (cP) using the electrolyte density.

Data availability
Data associated with this research can be found in the ESI † of this article.

Conflicts of interest
There are no conicts to declare.