Structural and spectroscopic studies of a rare non-oxido V(v) complex crystallized from aqueous solution

A non-oxido V(v) complex with glutaroimide-dioxime (H3L), a ligand for recovering uranium from seawater, was synthesized from aqueous solution as Na[V(L)2]·2H2O, and the structure determined by X-ray diffraction.


Introduction
The recovery of uranium from seawater has received considerable attention in the last few years because this untapped source contains 4.5 billion tons of uranium, 1 vastly more than the entire known terrestrial supply. Development of an efficient and economical technology for recovering uranium from seawater could therefore make the world's oceans a nearly limitless source of fuel for nuclear reactors. Currently, the most advanced technology for extracting the exceedingly dilute uranium (3.3 mg kg À1 , 14 nM) 2 from seawater involves the use of polymeric sorbents functionalized with the amidoxime moiety (-C(NH 2 ) NOH). 3,4 Promising marine test results have been reported in Japan over a decade ago in which the uranium uptake was 1.5 g U kg À1 sorbent aer 30 days 5 while more recently, marine tests conducted in the United States revealed that 3.3 g U kg À1 sorbent was obtained aer 8 weeks. 6 Although these results are promising, studies also reported signicant co-sorption of iron(III) and, in particular, vanadium(V), which is the most stable oxidation state under the conditions of seawater E h and pH. Sorption of these cations on poly(amidoxime) sorbents follows the order: vanadium(V) [ iron(III) > uranium(VI). 7 Interestingly, though the concentration of vanadium (1.9 mg kg À1 , 37 nM) 8 is approximately three times the uranium concentration in seawater, it occupies nearly twenty times as many sorption sites as uranium, essentially limiting the sorption capacity for uranium. Moreover, the stripping conditions required to elute the sorbed V(V) from the sorbent for reuse are much harsher than those used to elute uranium and other cations and ultimately destroy the sorbent. 9,10 These factors indicate that vanadium is a particularly problematic element that affects the economic viability of extraction of uranium from seawater using poly(amidoxime) sorbents. Therefore, a fundamental understanding of vanadium coordination to amidoxime-type sorbents could help optimize this extraction technology.
Structural studies can be used to provide valuable insights into the coordination behavior of vanadium and other metal cations with amidoxime ligands and can also help explain their subsequent sorption behavior with poly(amidoxime) sorbents. For example, the crystal structures and thermodynamic stability constants have been reported for U(VI) and Fe(III) complexes with glutaroimide-dioxime (Fig. 1), a cyclic imidedioxime moiety that can form during the synthesis of the poly(amidoxime) sorbent and is reputedly responsible for the extraction of uranium from seawater. 11,12 For both cations, two glutaroimide-dioxime ligands bind in a tridentate mode to the metal center. However, the ligands were found to bind Fe(III) much more strongly than U(VI) as manifested by the shorter Fe-O and Fe-N bond lengths relative to the corresponding U-O and U-N bond lengths (even aer taking into consideration the difference in ionic radii between Fe 3+ and UO 2 2+ ). The shorter bond lengths in the Fe(III) complex were attributed to the higher charge density of Fe(III) as well as its larger orbital participation in bonding relative to uranium. The higher thermodynamic stability and shorter bond lengths of the Fe 3+ / glutaroimide-dioxime complexes were postulated to be responsible for the higher sorption of Fe 3+ compared to UO 2 2+ in marine tests. Though the crystal structure of V(V) with glutaroimide-dioxime has not been reported, reasonable speculations about its structure can be made using information obtained from the known V(V) crystal structures. Based on the reported structures of V(V) complexes with organic ligands prepared from aqueous solutions (or ionic liquid equilibrated with water), it is known that the VO 2 + moiety with two short oxido V]O bonds (R V]O ¼ 1.60-1.63Å) usually remains intact. [13][14][15] Therefore, unlike the UO 2 2+ cation which possesses a linear trans dioxido conguration that allows two tridentate ligands to bind in the equatorial plane to form a strong 1 : 2 U(VI)/L complex, 11 the VO 2 + cation with its bent cis dioxido conguration cannot accommodate two such ligands due to steric hindrance and insufficient coordination sites.
These observations raise questions about why V(V) is sorbed much more strongly than U(VI) by the amidoxime sorbents. One hypothesis that could explain the much stronger complexation of V(V) is that V(V) exists in the glutaroimide-dioxime complex as a non-oxido, "bare" V 5+ ion coordinated with the ligand(s). A nonoxido V 5+ cation could have a very high affinity for O and N donor ligands due to its high charge density and could easily accommodate two tridentate ligands in a mode similar to that in the Fe 3+ /glutaroimide-dioxime complex. 12 However, crystal structure data in the Cambridge Structural Database (CSD) 16 indicate that, while there are non-oxido V 4+ complexes with ligands such as 1,3,5-triamino-1,3,5-trideoxy-cis-inositol (taci) 17 or N-hydroxyiminodiacetate 18 that have been crystallized from aqueous solutions, crystals of non-oxido V 5+ complexes from aqueous solutions are extremely rare. One non-oxido V 5+ complex, [PPh 4 ][D-V((S,S)-HIDPA) 2 ]$H 2 O (HIDPA 3: ¼ fully-deprotonated 2,2 0 -(hydroxyimino)dipropionic acid, H 3 HIDPA), was crystallized as the oxidized analogue of the naturally-existing Amavadin [19][20][21][22][23] from aqueous solution through the oxidation of a V(IV) complex by Ce(IV). 24 To the best of our knowledge, there have been no "bare" V 5+ complexes directly synthesized from oxido V(V) species ([O]V]O] + or vanadates) and crystallized from aqueous solution. In addition, the formation of non-oxido V 5+ complexes in aqueous solutions via the displacement of the oxido V]O bonds by chelating ligands (e.g., the trishydroxamate derivative deferoxamine 25 ) was only postulated but has not been demonstrated.
Although complexation of vanadium with Schiff bases such as glutaroimide-dioxime is problematic for the extraction of uranium from seawater, such complexes are currently of great interest for a variety of biological applications. For example, the V(V) complex with 4-hydroxy-dipicolinic acid (4-hydroxy-2,6pyridinedicarboxylic acid, H 2 Dpa-OH), a ligand that is structurally similar to glutaroimide-dioxime, was shown to exhibit insulin mimetic behavior in vivo. 14 However, though a signicant reduction of glucose levels was observed in animal studies, newer vanadium complexes need to be designed to further enhance mimetic behavior. Since 4-hydroxy-dipicolinic acid is structurally similar to glutaroimide-dioxime, structural comparisons of their respective V(V) complexes could prove useful for the design of improved insulin mimetic compounds.
In an effort to provide structural insights into vanadium complexation with amidoxime ligands, the present work has been conducted to synthesize crystals of V(V)/glutaroimidedioxime complexes and characterize their crystal-and solution structures by single-crystal X-ray diffraction (XRD), multinuclear ( 51 V, 17 O, 1 H, and 13 C) nuclear magnetic resonance (NMR), electrospray ionization mass spectrometry (ESI-MS), and electron paramagnetic resonance (EPR). This work represents the synthesis and identication of the rst non-oxido V(V) complex that was directly synthesized from an oxido V(V) species and crystallized from aqueous solution. The displacement of oxido V]O bonds by chelating ligands that leads to the formation of a non-oxido V(V) complex in aqueous solution has been unprecedentedly demonstrated by concurrent 51 V/ 17 O NMR experiments. Results from this work provide important insights into the strong sorption of vanadium on poly(amidoxime) sorbents in the recovery of uranium from seawater. Fig. 1 Glutaroimide-dioxime. *Glutaroimide-dioxime was denoted as H 2 L in previous publications 11,12 without taking into consideration all three dissociable protons.

Crystal structure of Na[V(L) 2 ]$2H 2 O(cr)
The asymmetric unit of Na[V(L) 2 ]$2H 2 O(cr) consists of a "bare" V 5+ center bound to two fully deprotonated glutaroimide-dioxime ligands (L 3À ), through one nitrogen and two oxygen atoms of each ligand, along with a sodium ion and two water molecules (Fig. 2a). The binding of the ligands around the vanadium center results in a highly distorted octahedral coordination environment in the triclinic space group P 1 ( Fig. 2b) with unit cell parameters a compounds obtained from non-aqueous solutions (1.8-2.0Å), 16 and much longer than those of the V]O double bonds ($1.6Å). 13,14 Crystal structure of Na[VO 2 (HL)](cr) The 1 : 1 V(V)/glutaroimide-dioxime complex (Fig. 3) possesses a distorted square pyramidal structure with s ¼ 0.35 in the monoclinic space group P2 1 /c: a ¼ 15.543 (8) (3) . The doubly deprotonated ligand (HL 2À ) coordinates to the V center through a k 3 binding motif via the imide N atom (R V-N6 ¼ 1.9885(17)Å) and the oxime O atoms (R V-O2,V-O5 ¼ 1.8931 (14), 2.0054(13)Å). Notably, the 1 : 1 complex (Fig. 3) is not a "bare" V 5+ complex unlike the 1 : 2 complex (Fig. 2). Instead, the 1 : 1 complex has the VO 2 + moiety with two short oxido bonds (V-O3 and V-O14) with bond distances of 1.6781 (15) Fig. 3), but differs in the location of one proton. In the crystal, the proton (H1) is located on the nitrogen (N1), probably due to the lattice interaction with Na + . Nevertheless, whether the 1 : 1 complex is  terization of V(V) complexes in solution due to its wide chemical shi range, high sensitivity, and high natural abundance. 26,27 On the other hand, oxygen-17, with I ¼ 5/2, is an NMR-active isotope of oxygen with a very low natural abundance and low NMR sensitivity, so isotopic enrichment is usually necessary for its detection and study. Indirect scalar spin-spin coupling between 17 O and 51 V can also be observed by 17 O and 51 V NMR if both atoms are bound directly. 28,29 As shown in Scheme 1, starting with 17 Table S3. † As different equivalents (1, 2, and 3) of glutaroimide-dioxime were added to the vanadate solution, both the 51 V and 17 O signals for vanadates (>) disappeared. In addition, a new 51 V signal in the 51 V spectra began to appear at d ¼ À410 ppm (7) and achieved maximum intensity at was increased to 2 ( 51 V spectrum c), and nearly disappeared as [L]/[V] was further increased to 3 ( 51 V spectrum d). Concurrently, a new peak appeared in the 17 O spectra around d ¼ 905 ppm (7) and achieved maximum Based on the changes in the peak intensities with the increase of The intensity of the 51 V NMR signal for the nal complex at [L]/[V] > 2 remained unchanged beyond 12 days, which suggests that vanadium remained in the V(V) oxidation state in the solution at neutral to slightly alkaline pH. If reduction of V(V) to the paramagnetic V(IV) species were to occur, it would diminish and eventually "wash-out" the 51 V NMR signal. Further reduction to V(III) is very unlikely: V(III) is generally much less stable in aqueous solutions, and no signals were observed in the lower 51 V chemical shi range of below d ¼ À1000 ppm. 28 51 V/ 17 O NMR experiments in acidic solutions were not performed in this study because (1) [V(L) 2 ] À may not be the dominant and most stable complex in acidic regions and (2) preliminary experiments suggested that redox reactions could occur between V(V) and glutaroimide-dioxime in more acidic solutions. The redox reactions between V(V) and the ligand are the subject of a future study.
To summarize, concurrent 51 V/ 17  Importantly, the 1 H spectra of the complexes showed that the equivalencies of the H atoms in the free ligand remain unchanged in the 1 : 1 and 1 : 2 complexes (Fig. 5). In other words, the same number of 1 H resonances (two) with the same spin-spin coupling ne structures is observed for the complex and the free ligand, which agrees with the coordination modes of the ligand in the complexes hypothesized in Scheme 1 and conrms the structure of a non-oxido V 5+ /glutaroimide-dioxime complex. The same analysis can be made with the 13 C NMR spectra (ESI, Fig. S1 †).

ESI-MS
The negative mode ESI-MS spectra for two aqueous solutions  Fig. S2. † According to the manufacturer's specications, the 10% 17 Oenriched water also contains at least 25% 18 O (see "Method" section). Consequently, the initial vanadate (Scheme 1) was actually labelled with 17 O as well as 18 O with the latter in a much higher yield. Therefore, unnatural isotopic patterns, particularly an (m + 2) peak corresponding to an isotopologue containing one 18 O, should be observed if the vanadium complex still contains an oxido V]*O bond from the vanadate and, more importantly, the (m +2) peak should be absent if all oxido V]*O bonds of the vanadate are displaced by the glutaroimidedioxime ligand.
Notably, the base peak at m/z ¼ 330.8 does not show the unnatural (m + 2) isotopic pattern that could indicate the presence of one 18 O atom (or two 17 O atoms with a much lower probability) in the 1 : 2 complex (Fig. 6, lower spectrum). This is because all of the oxido V]*O bonds of the initial 17 Table S3. † the initial 17,18 O-labelled vanadate in the intermediate 1 : 1 complex, in agreement with Scheme 1. It should be remarked that, for the 1 : 1 complexes, the intensities of the (m + 1) peaks include the contributions from the natural 13 C/ 15 N abundances, and the additional contribution from the isotopologue containing one 17 O atom.
Two ESI-MS spectra obtained by using a different diluent (methanol) on a different spectrometer (Finnigan LTQ FT mass spectrometer) are shown in ESI, Fig. S3. † Again, the spectra show the peak at 331.0 without the (m + 2) feature that corresponds to the non-oxido complex, [V(L) 2 ] À , and a peak at 238.0 with a prominent (m + 2) feature that corresponds to a 1 : EPR EPR spectra of powdered crystals were recorded at 300 K and 4 K (ESI, Fig. S4a †). At 4 K, only a weak signal with g ¼ 2.00 and no hyperne coupling was observed, which is due to the presence of organic radicals. This signal is frequently observed due to the high sensitivity of EPR spectroscopy. The lack of hyperne coupling and the fact that the g value is quite different from that typical for V(IV), 1.95, strongly suggest that only V(V) is present at low temperature. 30 Unlike the low temperature spectrum, the spectrum recorded at 300 K displays evidence for hyperne coupling typical of V(IV) (ESI, Fig. S4b †). However, the 300 K spectrum is still very weak, which indicates that V(IV) is only a minor component at this temperature. Overall, the EPR spectra are consistent with a V(V) ground state and indicate the potential presence of a low lying charge transfer state that could be populated at high temperatures.

Discussion
As previously mentioned, the sorption of V(V) to poly (amidoxime) sorbents in marine tests was reportedly much higher than that of Fe(III) and U(VI), following the order: V(V) [ Fe(III) > U(VI). Useful structural insights into the higher sorption of V(V) can be gained by comparing the structural parameters and coordination modes of the glutaroimide-dioxime complexes with V(V), Fe(III), and U(VI), as shown in Table 1 31 these structural data indicate that V 5+ forms a stronger complex with glutaroimide-dioxime than Fe 3+ assuming a predominantly ionic bonding model. The formation of stronger V 5+ complexes is most probably responsible for the higher sorption of V(V) than Fe(III) by poly(amidoxime) sorbents.  Since each batch of crystals was obtained from solutions prepared at or near neutral pH where the ligand had the same protonation state (H 3 L), it is evident that V(V) competes the most effectively with protons for the ligand under these conditions. In conjunction with the parallel trend in bond lengths discussed above, this observation corroborates the suggestion that vanadium(V), in the form of the "bare" V 5+ ion, forms the strongest complex with glutaroimide-dioxime by complete deprotonation of the ligand.
In summary, the extremely strong sorption of V(V) by the poly(amidoxime) sorbents is probably due to the formation of the very stable non-oxido V 5+ complex with glutaroimide-dioxime. To improve the selectivity of the sorbent for U(VI) over V(V), an ideal ligand would be the one(s) with a binding ability that is sufficiently high for U(VI) but not high enough to displace the oxido V]O bond(s) in the V(V) species. Starting with the cyclic glutaroimide-dioxime platform, adding electron-withdrawing groups to the platform could reduce the basicity of the imide and oxime groups and "ne-tune" the binding ability of the ligand(s).
The isolation of a "bare" non-oxido V(V) complex from aqueous solution is a very rare occurrence. To the best of the authors' knowledge, only one other non-oxido V(V) complex has been reported. The non-oxido V(V) complex, [PPh 4 ][D-V(HIDPA) 2 ]$H 2 O(cr), can be synthesized by oxidizing Amavadin, which is itself a non-oxido V(IV) complex of (S,S)-2,2 0 -(hydroxyimino)dipropionic acid (H 3 HIDPA), and subsequently precipitating it from an aqueous solution containing a tetraphenylphosphonium (PPh 4 + ) salt. 24 Interestingly, the V(HIDPA) 2 À complex contains two tetradentate ligands that coordinate via a central nitrogen and three oxygen donors unlike [V(L) 2 ] À , in which tridentate bonding is observed. As Table 2 shows, the average V-O and V-N bonds for Na[V(L) 2 ]$ 2H 2 O(cr) are signicantly shorter than the analogous bonds for the V(HIDPA) 2 À complex by 0.075Å and 0.055Å, respectively.
The shorter bond lengths coupled with the fact that the oxime moieties of glutaro-imidedioxime are more basic (pK a z 11-12) than the carboxylate moieties (pK a z 4-5) in HIDPA implies that [V(L) 2 ] À is likely a stronger complex.
In addition to helping improve the extraction of uranium from seawater, the structural information for both the 1 : 1 oxidovanadium(V)and 1 : 2 non-oxidovanadium(V)-glutaroimide-dioxime complexes could help to understand and develop vanadium(V) compounds that mimic the effects of insulin in the treatment of diabetes. It is known that vanadium plays very important roles in biological systems 20,34,35 and that some V(V) organic complexes, such as the aforementioned K [VO 2 (Dpa-OH)] complex, exhibit insulin mimetic behavior. 14 Since glutaroimide-dioxime is structurally similar to Dpa-OH, has a similar binding motif (O,N,O), and forms similarly charged complexes, useful insights can be gained by comparing the structures of these complexes. Table 2 compares the bond lengths of K[VO 2 (Dpa-OH)]$H 2 O(cr) and the two V(V)-glutaroimide-dioxime complexes.
As shown in the table, the V-N bond and the average V-O bond distances in Na[VO 2 (HL)](cr) are shorter than the analogous bond distances in K[VO 2 (Dpa-OH)]$H 2 O(cr) by 0.10Å and 0.06Å, respectively, implying stronger bonding in the glutaroimide-dioxime complex. Interestingly, the oxido V-O bonds in Table 2 Comparison of bond lengths (Å) for V(V) complexes with glutaroimide-dioxime (H 2 L), 4-hydroxydipicolinic acid (H 2 Dpa-OH), and (S,S)-2,2 0 -(hydroxyimino)dipropionic acid (H 3 HIDPA)  VO 2 (HL) À are slightly longer than the oxido bonds in the Dpa-OH complex, which implies weaker V]O bonds and may explain the ability of the second glutaroimide-dioxime ligand to subsequently displace the two oxido oxygens. In fact, the nonoxido [V(L) 2 ] À complex formed upon addition of a second ligand to VO 2 (HL) À leads to an even more signicant reduction of bond lengths in the Na[V(L) 2 ]$2H 2 O crystal. The V-N and average V-O bonds in [V(L) 2 ] À are the shortest of all three complexes by 0.13 A and 0.12Å, respectively, compared to VO 2 (Dpa-OH) À . In this case, the higher charge density of V 5+ (compared to the VO 2 + moiety) coupled with the very short bond lengths indicate that [V(L) 2 ] À is a much stronger complex than VO 2 (Dpa-OH) À . Concurrent 51 V/ 17 O NMR experiments in aqueous solution (Fig. 4) showed that, at pH 7.5 and a 1 : 1 V : L ratio, the VO 2 (HL) À complex predominates. This implies that the VO 2 (HL) À complex is stable and intact at physiological pH (pH 7.4), which is a desired property of organovanadium compounds in order to minimize the in vivo toxicity. Although the speciation of the V(V)-(Dpa-OH) system at physiological pH is not known, it is known that the structurally similar VO 2 (Dpa) À complex, which also exhibits insulin-mimetic behavior, dissociates above pH 5. 36 Based on the structural similarities between VO 2 (Dpa) À and VO 2 (Dpa-OH) À , the Dpa-OH complex should also dissociate at physiological pH. Therefore, if Na [VO 2 (HL)] exhibits insulin mimetic behavior in vivo, the mechanism of action could be different from that of the dissociated VO 2 (Dpa-OH) À complex, making Na[VO 2 (HL)] a worthy candidate for further investigation.
Lastly, in a detailed study carried out by Yoshikawa and coworkers of six different crystalline non-oxido V(IV) complexes, very compelling evidence was provided suggesting that only those complexes that transformed to their vanadyl (oxido) form at physiological pH exhibited insulin mimetic behavior. 37 However, such a study could not be carried out with V(V) complexes partly because very few non-oxido V(V) complexes have been identied. The non-oxido and oxido V(V) complexes with glutaroimide-dioxime could provide a unique opportunity to investigate the in vivo behavior of intact oxido-and non-oxido V(V) complexes containing the same ligand, binding motif, and overall charge at physiological pH. The results of these studies could help corroborate the hypothesis of Yoshikawa et al. regarding the requirement for an oxido (or dioxido) vanadium moiety to observe insulin mimetic behavior.

Conclusions
A rare, non-oxido V(V) complex with glutaroimide-dioxime (H 3 L), Na[V(L) 2 ]$2H 2 O(cr), was crystallized from aqueous solution and characterized via X-ray diffraction. The complex was found to contain two fully deprotonated L 3À ligands bound to the bare V 5+ cation via two oxime oxygens and the imide nitrogen. An intermediate complex, Na[VO 2 (HL)](cr), was also isolated and found to contain the typical VO 2 + moiety present in many V(V) complexes. Further characterizations using 51 V, 17 O, 1 H, and 13 C NMR spectroscopy unprecedentedly demonstrated the stepwise displacement of the oxido oxygens to form the bare V(V)-glutaroimide-dioxime complex. ESI-MS studies of V(V)-glutaroimide-dioxime solutions allowed the identication the intermediate 1 : 1 M : L complex as well as the bare [V(L) 2 ] À complex at m/z ¼ 330. 8.
Structural insights into the much higher sorption of V(V) to amidoxime-based sorbents relative to U(VI) and Fe(III) were gained by comparing the structural parameters of the V(V)glutaroimide-dioxime complex with the analogous U(VI)-and Fe(III)-glutaroimide-dioxime complexes. For these complexes, the degree of protonation of the ligand was found to decrease from U(VI) to V(V). In conjunction with the substantially shorter bond lengths observed for the V(V) complex relative to the other complexes, this implies stronger bonding in the V(V) complex and higher thermodynamic stability. In fact, the trend in binding strengths parallels the observed trend in sorption of these cations to poly(amidoxime) sorbents in marine tests.
Lastly, as there are ongoing studies to synthesize vanadium(V) compounds suitable for the treatment of diabetes, the structural studies with glutaroimide-dioxime are useful for aiding the development of new, highly stable organic V(V) compounds. In fact, the high solubility of Na[V(L) 2 ]$2H 2 O in aqueous and ethanol solutions coupled with its stability at physiological pH could make it a potential candidate for use in diabetic treatment studies.

Synthesis and single-crystal XRD of Na[V(L) 2 ]$2H 2 O(cr)
Single crystals of Na[V(L) 2 ]$2H 2 O(cr) were prepared at Lawrence Berkeley National Laboratory (LBNL). The glutaroimide-dioxime ligand was synthesized, and its purity was veried as described previously. 38 A two milliliter aliquot of an aqueous stock solution at pH 8 containing NaVO 3 (0.2 mmol), NaCl (12 mmol), and 0.5 mmol glutaroimide-dioxime was slowly evaporated over the course of a week to generate shiny, dark brown/ black acicular crystals. The crystals are very soluble in water, fairly soluble in ethanol, and less soluble in acetonitrile and methanol. Interestingly, it was observed that prolonged heating of aqueous Na[V(L 2 )] solutions at $50-60 C resulted in the apparent decomposition of the complex as evidenced by the fading color of the solution from dark brown to a yellow-orange color. However, no further efforts were made to ascertain whether the apparent decomposition was due to either partial oxidation of glutaroimide-dioxime by V(V) or to other mechanisms.
A single crystal was selected, removed from Paratone oil with a MiTiGen microloop, and mounted on to a Bruker goniometer equipped with a PHOTON100 CMOS detector and Oxford Systems Cryostream 800 series on beamline 11.3.1 of the Advanced Light Source at LBNL. The data were collected at 100K using the Bruker APEX2 soware 39 in shutterless mode using u rotations at a wavelength of 0.7749Å. The intensity data were integrated using SAINT v.8.34A 40 and the absorption and other corrections were applied using SADABS 2014/5. 41 The appropriate dispersion corrections for C, H, N, O, and V at l ¼ 0.7749Å were calculated using the Brennan method in XDISP 42 run through WinGX. 43 The structure was solved with intrinsic phasing using SHELXT 2014/4 and rened using SHELXL 2014/7 (ref. 44). All non-hydrogen atoms were rened anisotropically. Hydrogen atoms were found in the difference map and allowed to rene freely. Detailed crystallographic data and structure renement for Na[V(L) 2 ]$2H 2 O(cr) are provided in ESI, Table S1. †

Synthesis and single-crystal XRD of Na[VO 2 (HL)](cr)
Single crystals of Na[VO 2 (HL)] were prepared at Pacic Northwest National Laboratory (PNNL). Glutaroimide-dioxime 11,38 (30 mg, 0.21 mmol) was suspended in deionized water (1 mL). NaVO 3 (25 mg; 0.21 mmol) was added, resulting in a dark brown solution immediately. Aer stirring for 5 h, the solution was ltered to remove any undissolved solids (e.g., unreacted glutaroimide-dioxime or by-product salts) prior to removing the solvent. The residue was then re-dissolved in ethanol and ltered as before. Orange crystals were obtained from vapor diffusion of hexane into the ethanol solution. Note that the undissolved solids remaining aer either ltration were not characterized.
A Bruker-AXS Kappa Apex II CCD diffractometer with 0.71073 A Mo Ka radiation was used for data collection. Crystals were mounted on a MiTeGen MicroMounts pin using Paratone-N oil. Data were collected at 100 K. The soware used for data analysis includes Brüker APEX II 39 to retrieve cell parameters, SAINT-Plus 40 for raw data integration, and SADABS 41 to apply the absorption correction. The structures were solved using either direct methods, charge ipping methods or the Patterson method and rened by a least-squares method on F2 using the SHELXTL program package. Space groups were chosen by analysis of systematic absences and intensity statistics. Detailed crystallographic data and structure renement for Na[VO 2 (HL)](cr) are provided in ESI, Preparation of vanadium/glutaroimide-dioxime solutions.  Table S3. † NMR data collection. All NMR spectra were acquired at 20-22 C. The 51 V spectrum of the D 2 O solution of Na[V(L) 2 ]$ 2H 2 O(cr) was acquired at LBNL on a Bruker AV-300 spectrometer referenced to an external standard of VOCl 3 in C 6 D 6 . All other 17 O, 51 V, and 13 C NMR spectra were acquired at UCB on a Bruker DRX-500 spectrometer equipped with a Z-gradient broadband probe. The 1 H spectra were acquired at UCB on a Bruker AV-500 spectrometer equipped with a Z-gradient triple broadband inverse detection probe using WATERGATE solvent suppression. The 1 H, 13 C, and 51 V spectra were referenced to an external standard of VOCl 3 in C 6 D 6 and the 17 O spectra were referenced to the H 2 17 O water resonance.

Electrospray ionization-mass spectrometry
Two sets of ESI-MS experiments were performed using different spray solutions (an ethanol/water mixture and methanol, respectively) on two different instruments. The ESI-MS experiments with the ethanol/water mixture were performed using an Agilent 6340 quadrupole ion trap mass spectrometer with a micro-ESI source at LBNL. Aliquots of the solutions with [L]/ [V] at 1 : 1 and 2 : 1 were diluted in (90/10) ethanol/water and injected into the instrument and sprayed in the negative ion mode at 1 mL min À1 . The ESI-MS experiments with methanol spray were conducted on a Finnigan LTQ FT mass spectrometer (Thermo) at the QB3/Chemistry Mass Spectrometry Facility (UCB). Aliquots of the 1 : 1 and 2 : 1 [L]/[V] samples were taken and diluted in methanol. The samples were injected directly via a syringe at a ow rate of 5 mL min À1 with a spray voltage of 3.5 kV.
Electron paramagnetic resonance spectroscopy EPR spectra were obtained at LBNL at room temperature and at 4 K with a Varian E-12 spectrometer equipped with liquid helium cryostat, an EIP-547 microwave frequency counter, and a Varian E-500 gaussmeter, which was calibrated using 2,2diphenyl-1-picrylhydrazyl (DPPH, g ¼ 2.0036).