Influence of ligand encapsulation on cobalt-59 chemical-shift thermometry† †Electronic supplementary information (ESI) available. See DOI: 10.1039/c9sc01689a

This manuscript details the first investigation of ligand encapsulation on thermometry by cobalt-59 nuclear spins.


Introduction
The structural exibilities of metal complexes are key design principles for applications in the areas of reactivity, 1,2 medicine, 3 photophysical properties, 4 and magnetic information storage. 5 Flexibility engenders stimuli-dependent changes in the coordination geometry of a metal, hence impacting d-orbital energies and any properties stemming from electronic structure. 6 Thus, the control of exibility is potentially a powerful way for targeting applications for metal complexes. One such application is biomedical thermometry by magnetic resonance imaging (MRI), 7,8 where the temperature-dependent structure of a exible complex induces highly temperature-dependent spin-Hamiltonian parameters or relaxation times. If this variation could be harnessed to develop an imaging technique, such an application would circumvent many of the challenges associated with invasive thermometry, e.g. the point-like nature of the measurement.
One promising system for such thermometry by magnetic resonance is the cobalt-59 nucleus in low-spin cobalt (III) complexes. This NMR-active nucleus is 100% naturally abundant, I ¼ 7/2, and has a receptivity of ca. 30% that of 1 H. Furthermore, the nucleus displays a wide reported chemical shi (d) window (20 000 ppm) as a result of a paramagnetic contribution to d that is directly tied to the ligand eld splitting, D o . [9][10][11] For complexes that contain cobalt-59 nuclei, changes in solution structure, such as lengthening Co-ligand bonds, can impact D o , imparting changes in d, and providing a mechanism for thermometry. In principle, then, the 59 Co chemical shi could be used to spatially map temperature through a technique known as chemical shi imaging. [12][13][14] Initial studies reveal sensitivities (Dd/DT) on the order of 1-3 ppm C À1 , 15-20 order-ofmagnitude upgrades to the possibilities for conventional 1 H NMR thermometry. 21,22 Hence, these species may be useful to develop as new probes for chemical-shi imaging of temperature. 12 However, fundamental insight about the factors to govern that sensitivity is lacking (this manuscript explores one factorencapsulationas depicted in Fig. 1). As a result, design principles for enhancing Dd/DT values are absent, and the true potential of cobalt-59 NMR thermometers for MRI remains unrealized.
This manuscript details the rst systematic exploration of the molecular factors that govern Dd/DT in a family of cobalt-59 NMR thermometers ( Fig. 1 and 2). As a rst step, we sought to explore the role of ligand encapsulation on the temperature sensitivity of the ligand eld and cobalt-59 NMR properties.
Encapsulation is known to afford enhanced stability for metal complexes via the chelate and related macrocyclic effects. 25 Such stability is an important property for any imaging agent, as release of the metal can both induce toxicity and deactivate the magnetic species being used as a sensor. At the same time, a rigid, encapsulated ion can be readily envisioned to lack the exibility needed for thermometry via structural change. Hence, encapsulation, while affording signicant chemical stability, might simultaneously subdue the ability to sense temperature via cobalt-59 NMR.
To test this hypothesis, we investigated Dd/DT for the 59 Co nuclei of the low-spin Co(III) complexes (Fig. 2) 29 [Co(diNOsar)]Cl 3 (5, ¼ dinitrosarcophagine), 23 and K 3 [Co(CN) 6 ], the 59 Co NMR standard. This series of complexes was selected to enable an investigation of both molecular and electronic structure on Dd/DT. First, we hypothesized that the increasing connectivity between the nitrogen donor atoms in 1-5 would engender an increasingly rigid coordination environment and, hence, suppress Dd/DT. Thus, we expected that sensitivity to temperature would be compromised in favor of chemical stability. Indeed, the fullyencapsulating sarcophagine scaffold 23,25,30 will only surrender its NMR-active Co(III) ion under harsh conditionsheating in concentrated cyanide solution or acidic media. 31,32 The second investigation enabled by this set of complexes is the test of whether Dd/DT directly correlates with D o . The 59 Co chemical shi is proportional to 1/D o , 10 hence, d should be more sensitive to tiny uctuations in D o at lower D o . 10 These studies are the rst to reveal three key facts about Dd/DT. Firstly, in contrast to our expectations, encapsulation enhances Dd/DT. That isthe "rigid" ligand frameworks in 5 and 4 induce a stronger temperature-dependence in D o (and d) than the lessencapsulated species 1-3. Indeed, variable-temperature UV-Vis and 59 Co spin-lattice relaxation studies indicate that encapsulation counterintuitively supports higher temperature dependence in the coordination geometry. Second, our studies show that D o alone does not correlate to the magnitude of Dd/DT. Finally, third, Raman spectroscopy studies suggest molecular vibrational lifetimesprolonged by high interconnectivity among donor atomsare important factors governing Dd/DT. Together, the data highlight a new implication for rigidity in molecular magnetism.

Results and discussion
Understanding the temperature sensitivity of the chemical shi requires rst establishing the electronic structures of the cobalt(III) ions in 1-5. UV-Vis electronic absorption spectra of compounds 1-5 and K 3 [Co(CN) 6 ] in H 2 O reproduce reported results for the individual complexes (Fig. 3), wherein the lowest energy peak indicates the 1 A 1g / 1 T 1g transition and the higher energy peak indicates the 1 A 1g / 1 T 2g transition. 10,23 The energies of these two peaks and a Tanabe-Sugano diagram permit quantitation of D o , which increases from 3 (22 376 cm À1 ) to 5 (22 754 cm À1 ) and 1 (23 018 cm À1 ) to 4 (23 276 cm À1 ) to 2 (23 321 cm À1 ) (Table S1 †). These values are consistent with literature values for 1-5 and stand in contrast to the strong ligand eld of K 3 [Co(CN) 6 ] that engenders a D o of 38 000 cm À1 . 33 59 Co resonant frequencies were observed for 1-5 over the range of 6800 to 8400 ppm (referenced to K 3 [Co(CN) 6 ]). According to the 59 Co chemical shis, the magnitude of D o increases in the order 3 < 1 < 4 < 2 < 5 < K 3 [Co(CN) 6 ]. This order is at odds with the trend obtained from electronic absorption spectroscopy Anions and hydrogens omitted for clarity where necessary. Purple, blue, red, grey, and light grey spheres correspond to cobalt, nitrogen, oxygen, carbon, and hydrogen atoms, respectively.   6 ] in H 2 O at room temperature. The lower-energy peak is the 1 A 1g to 1 T 1g transition while the higher-wavenumber peak is 1 A 1g to 1 T 2g . Bottom: 500 MHz 59 Co NMR spectra for 1-5 in H 2 O at room temperature. measurements (see Fig. S1 †). However, reported correlations between UV-Vis peak position and d are only approximate, not quantitative. 9,10 Nevertheless, these measurements provide (i) two points of reference to test for a correlation between Dd/DT and D o , and (ii) the location of the 59 Co NMR resonances for variable-temperature analyses.
Variable-temperature 59 Co NMR spectra were collected for 1-5 and K 3 [Co(CN) 6 ] in H 2 O from 10-60 C (see Fig. 4, 5 and S2-S7 †) to explore the temperature dependence of d. With increasing temperature, peaks for 1-5 and K 3 [Co(CN) 6 ] shi downeld to higher d. This temperature-dependent shi of peaks is consistent with varying coordination geometry in solution. 18,34 As temperature is increased, energy is introduced into the vibrational modes of the cobalt complex, expanding M-L bond distances and engendering generally weaker D o . 18,35 Precise determination of the sensitivity of the 59 Co NMR peak to temperature (Dd/DT) was achieved via linear regression of the temperature-dependent data (see Fig. 5 (1), and 2.04(2) ppm C À1 . These values are within the ranges of sensitivity reported for the few 59 Co NMR thermometers, 15-18,20 but it's worth noting that, to the best of our knowledge, the Dd/DT of 5 is eclipsed only by Co(acac) 3 , a molecule that is completely unsuitable for aqueous (e.g. physiological) applications. 16,20 Most importantly (and surprisingly), these data indicate that the highest sensitivity to changes in temperature is held by the completely encaged complex 5.
The values of Dd/DT follow an opposing trend to the initial hypothesis, in that 5, with the highest degree of encapsulation, displays the strongest Dd/DT. Complex 4, with the second highest degree of encapsulation, displays the second highest sensitivity of our studied complexes. Yet, a comprehensive trend for all complexes on the basis of encapsulation is not indicated by these data. For example, in 1 and K 3 [Co(CN) 6 ], the ligand donor atoms are not connected in any manner. Yet, these species demonstrate higher Dd/DT than both 2 and 3, which contain bidentate chelates. Furthermore, the collected data show that electronic structure considerations alone (specically, D o ) do not govern sensitivity. Here, neither the 6 ]) reproduce the trend in Dd/DT (see Fig. S9 and Table S3 †).
The foregoing results highlight the need for deeper studies to derive fundamental insight. An important implication of the foregoing results is the concept that the encaged complex counterintuitively demonstrates the highest uxionality in the inner-coordination structure. Four key experiments were applied to further test this rationale.
If the molecular structure of [Co(diNOsar)]Cl 3 is truly more temperature-dependent than 1-4, then D o for 5 should show the greatest temperature dependence. Variable-temperature UV-Vis Fig. 4 Variable-temperature 59 Co NMR for 4 at 500 MHz in H 2 O, collected in increments of 5 C. The system was allowed to equilibrate for at least 5 minutes between each temperature point prior to measurement.  spectra for 1-5 show slight shis to lower energy with increasing temperature (Fig. 6 and S10-S12 †). Analyses of these data reveal a change in D o as a function of temperature, DD o /DT. Over 1-5, DD o /DT assumes values of À2.78(4), À1.36 (17), À2.91(5), À3.70 (17), and À5.65(32) cm À1 C À1 for 1-5 respectively. These spectral changes are consistent with studies probing temperature-dependent UV-Vis spectra for metal complexes wherein spin-state changes are absent 36,37 (versus systems displaying spin-crossover 38 or valence tautomerization 39 ). These data trend in a manner (particularly for 2-5) that seems opposed to an association between encapsulation and increased rigidity of the coordination environment. Indeed, 5 exhibits the largest change in temperature, followed by 4, then 3 and 1, and nally 2. Hence, these data point to a more dynamic inner coordination sphere.
If the inner coordination sphere is less rigid upon encapsulation, then cobalt-59 spin lattice relaxation times should reect that point. Indeed, the I ¼ 7/2 cobalt-59 nucleus is quadrupolar, and, hence, its spin-lattice relaxation rate (1/T 1 ) is dominated by uctuations in the local electric eld gradient. 40 Hence, the anticipated higher uxionality in the CoN 6 environment of 4 and 5 versus 1-3 should correspondingly engender shorter T 1 . Analysis of the inversion recovery traces for 1-5 (Fig. 7) reveal T 1 parameters for 1-5 that follow a trend with encapsulation, wherein [Co(NH 3 ) 6 ]Cl 3 (1) displays the longest T 1 (48.47(5) ms), followed by 2 (9.09(2) ms) and 3 (2.73(1) ms). In contrast, the species of highest encapsulation, 4 and 5, have the shortest T 1 values (346(1) and 323(1) ms, respectively). For 1 and 2, these values match previously reported results. [40][41][42] Quadrupolar relaxation is also enhanced in systems with a higher 59 Co quadrupole coupling constant, and this constant is smaller for high-symmetry complexes. 9 Compound 1 is clearly higher symmetry (O h ) than 2-5 (D 3 ). This symmetry difference is likely an important contributor to the T 1 of 1 versus 2-5, but quadrupolar couplings in this latter set of compounds are similar (when known). 42,43 Moreover, solution-phase rotational rates for the series extrema, 1 and 5, are similar, 42 suggesting rotational correlation is also not driving the difference in T 1 across the series. Together, these points suggest that considerations beyond symmetry/rotation dene T 1 for these compounds. In light of the other data in this paper, we propose that the T 1 trend evidences a more dynamic coordination environment upon encapsulation, though deeper investigations are needed to test this hypothesis.
Vibrational spectra in the 100-650 cm À1 window, wherein metal-ligand vibrations typically occur, ought to vary with rigidity as well. 44 To test this concept, microcrystalline powders of 1-5 were analyzed via Raman spectroscopy. As the probed molecules increase in structural complexity, so do the Raman spectra, with compound 1 exhibiting 8, 2 displaying 12, and 5 producing 21 bands below 650 cm À1 (Fig. 8). Previous reports identify symmetric Co-N bond stretches at 500 and 486 cm À1 for 1, and 526, 444, and 476 cm À1 for 2. 45,46 For 3-5, in contrast, no Raman spectra are reported to the best of our knowledge. Closer inspection of the Raman spectra reveals a general sharpening of transitions with increasing encapsulation. This sharpening is most noticeable when comparing the spectra of lesser encapsulated compounds 1-3 with the completely encapsulated species 5. Linewidth analyses of the observed vibrations for 1-5 permitted relative quantitation of the general degree of sharpness of these spectra (Fig. S13-S17 and Table S4 †). The averages of the peak linewidths for the spectra are ordered from 5 < 2 < 4 < 1 < 3, where the fully encapsulated species, 5, exhibits the smallest average peak width.
On the basis of the variable-temperature NMR and UV-Vis data, the highest exibility is observed for 5. However, the lifetime of the NMR experiment is much longer than that of vibrational spectroscopy. 47 Hence, one may therefore expect the greater structural variability from the NMR/UV-Vis analyses to result in greater inhomogeneous broadening of the vibrational peaks for 5. The observations from the Raman spectra are in contrast to this expectation, as 5 demonstrates the sharpest peaks. One alternative mechanism that governs peak linewidths is homogeneous broadening, which causes sharper peaks for  excitations that have longer lifetimes. 48 This mechanism is acknowledged as dominant in studies of M(CO) n at room temperature in solution. 49,50 If operative and dominant in powders of 1-5, this admittedly simplistic model of broadening would suggest that the lifetimes of the vibrations of the coordination sphere are enhanced by encapsulation. Translation of the average linewidths of 1-5 into average vibrational, spectroscopic lifetimes (via the relationship FWHM ¼ 1/ps) yields lifetimes of 0.4(2), 0.6(3), 0.4(2), 0.8(4), and 1.3(4) ps for 1-5, respectively.
The foregoing linewidth interpretation should be treated with caution owing to three specic factors. First, differences in microcrystalline environment can have an important impact on Raman linewidths (e.g. ref. 51 and 52). We note, however, that a preliminary powder diffraction analysis of the same samples measured by Raman spectroscopy did not reveal a noticeable trend of crystallinity correlating to the observed lifetimes (Fig. S18 †). Second, modes of differing symmetries can yield different linewidths, 49 as is likely evidenced here in the spread of linewidths in the deconvoluted peaks. Third, true elucidation of the vibration lifetimes requires time-resolved methods, which would also help differentiate inhomogeneous versus homogeneous broadening mechanisms. 50 These data clearly motivate further solution-phase, time-resolved vibrational studies, a critical component of planned follow up work. Nevertheless, the obtained lifetimes are in the general picosecond range expected for metal complexes. 49,50 If encapsulation affects Dd/DT via modulating vibration lifetimes, that insight would provide a new design principle for vibrational control of molecular spin. Variable-solvent studies of 2 were performed as one nal test of this concept. In particular, as the polar N-H bonds of the coordinated nitrogen atoms in 1-5 likely interact with the aqueous environment, this interaction should mediate the vibrations and structure of the [Co(en) 3 ] 3+ moiety, potentially imparting large differences to Dd/ DT. Indeed, such hydrogen bonding interactions are demonstrated to enable modulation of M-N and M-O bonds in other molecular systems. 53,54 Here, this concept is being tested for temperature-dependent magnetic effects.
Initial studies focused on one member of the series, [Co(en) 3 ]Cl 3 (2), dissolved in four additional solvents: dimethylformamide (DMF), hexamethylphosphoramide (HMPA), § dimethylsulfoxide (DMSO) and d 6 -dimethylsulfoxide (d 6 -DMSO). The solvents DMF, HMPA, and DMSO were selected to test polarity, and d 6 -DMSO chosen to test the impact of environmental deuteration. While solvent/deuteration impacts on 59 Co d are reported, 55 their role on thermometry is not yet understood. 59 Co NMR spectra collected at 25 C reveal a peak position that shis over a range of 200 ppm as a function of solvent identity (see Fig. 9). This solvent-dependent effect is known for the ClO 4 À salt of the [Co(en) 3 ] 3+ cation, stemming from modulation of the N-atom ligand eld via hydrogen bonding between the solvent and N-H protons. 56 Furthermore, only a tiny shi in d is observed between DMSO and d 6 -DMSO, also in line with expected results. 55 Variable-temperature analyses tested the impact of these differing solvent cages on Dd/DT (Fig. S19-S22 and Table S5 †).
Here, analysis of the variable-temperature 59 Co NMR peak positions as a function of solvent demonstrate a noticeable impact of solvent identity on Dd/DT (Fig. 9). As in H 2 O, all 59 Co NMR chemical shis move downeld with increasing temperature. Linear regression of these temperature-dependent data reveal Dd/DT values of 1.19(2), 1.23(1), 1.27(1), and 1.28(1) ppm C À1 , respectively, for DMF, HMPA, DMSO, and d 6 -DMSO.
These values are all lower than in H 2 O (Dd/DT ¼ 1.38(1) ppm C À1 ) and indicate nearly no role for solvent deuteration on Dd/ DT in the present compound.
Comparison of the solvent-dependent Dd/DT results for 2 against measures of solvent-solute interaction potentially provide deeper insight into the role of the solvent cage (Fig. 9, S23 and Table S5 †). In particular, the trend in Dd/DT was contrasted against (1) the solvent acceptor and donor numbers, 57,58 (2) the p* solvent polarity scale, 59,60 and (3) the b and a hydrogen-bonding donor/acceptor scales. [60][61][62] There may be an approximate correlation between Dd/DT and acceptor number, whereas there is clearly none for donor number and little, if any with b (see Fig. S23 †). Analysis with a values is complicated as a is 0 for all solvents here except H 2 O. 60 However, the p* scale clearly reveals a correlation (R 2 ¼ 0.9, Fig. 9, bottom).
The foregoing data point toward a coordination environment that is counterintuitively more exible and dynamic with increasing encapsulation. Electron transfer studies of sarcophagine-like ligands provided the rst assertions of rigidity in encapsulated Co(III) complexes based on a conformationally inexible environment. [63][64][65][66][67] We propose that this conformation-based description of rigidity is insufficient for understanding the trend of Dd/DT. Instead, we tentatively posit an alternative, spin-relevant interpretation in this context. Here, the enhanced connectivity in 4 and 5 ensures a higher rigidity in the coordination environment, except in this case the rigidity permits vibrations of the encapsulated coordination environment to persist longer. Such longer lifetimes ultimately sustain a change in the coordination sphere by lengthening the equilibrium Co-N bond distances. Hence, there is a temperature dependence of D o and d in 4 and 5 that is larger than 1-3. This tentative interpretation of the data is also consistent with the solvent dependence of Dd/DT in 2, as the solvent cage is known for impacting vibration lifetimes in coordination complexes. 49,68,69 The fundamental argument we propose here is an analogue to the justication of long phonon lifetimes in materials like diamond, 70,71 except here related to the molecular vibrations of a complex in solution. These studies clearly motivate future investigations to evaluate the validity of this picture of vibration-controlled spin properties.
Furthermore, the solvent-dependent data hint at a rich area of inquiry into the role of the second coordination sphere and counterions. When considering the [Co(en) 3 ] 3+ unit, interactions with the solvent are most easily intuited via the N-H protons accepting electron density from solvent molecule lone pairs. The association of a higher Dd/DT with a higher p* index of solvent polarity 59 would mesh with this intuited picture. This model would also be consistent with the match between a lower b value and a higher Dd/DT, as a low b occurs when a solute will only weakly accept a proton. 62 Thus, these data suggest that the N-H interactions are key to understanding Dd/DT. However, the interpretation isn't without some uncertainty. To the extent that there is any correlation of Dd/DT with solvent properties, it is with their acceptor number, not donor number, meaning that [Co(en) 3 ] 3+ acts as a donor. This argument only makes sense if one also considers the lone pairs of a bound Cl À counterion, not the N-H bonds. Indeed, earlier studies of [Co(en) 3 ] 3+ and [Co(diNOsar)] 3+ demonstrate a close association between these species and their Cl À counterions that persists in solution. 43,67,72 Noted reservations about generalizing the acceptor/donor number scale lend caution to the second explanation of the solvent-dependent data. 73 Nevertheless, the conict between these two interpretations underlines the necessity of further investigations into the role of the counterion and solvent cage on Dd/DT.

Conclusions and outlook
The foregoing results are the rst evidence of synergy between ligand encapsulation and enhanced temperature-dependent magnetic changes in metal-ion nuclear spins. Such knowledge is of broad impact, as exploiting molecular rigidity to control magnetism is an emerging trend in designing molecules for other spin-based technologies, e.g. molecular quantum bit development. [74][75][76] Importantly, the presented arguments potentially tie vibration lifetimes to nuclear magnetismnecessitating future time-resolved measurements to test the validity of this analysis. Finally, our studies reveal that in addition to the ligand, manipulations of the counterion and solvent cage are the next stage for understanding the mechanisms that control Dd/DT. Beyond the targeted applications in thermometry, the concepts herein could be extended to understanding the impacts of molecular rigidity on other spinbased applications, for example, designing electron paramagnetic resonance imaging probes, 77 rigid systems for dynamic nuclear polarization [78][79][80] (particularly with metal ions), 81,82 or molecular quantum sensors. 83,84

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