A Paramagnetic NMR Spectroscopy Toolbox for the Characterisation of Paramagnetic/Spin‐Crossover Coordination Complexes and Metal–Organic Cages

Abstract The large paramagnetic shifts and short relaxation times resulting from the presence of a paramagnetic centre complicate NMR data acquisition and interpretation in solution. As a result, NMR analysis of paramagnetic complexes is limited in comparison to diamagnetic compounds and often relies on theoretical models. We report a toolbox of 1D (1H, proton‐coupled 13C, selective 1H‐decoupling 13C, steady‐state NOE) and 2D (COSY, NOESY, HMQC) paramagnetic NMR methods that enables unprecedented structural characterisation and in some cases, provides more structural information than would be observable for a diamagnetic analogue. We demonstrate the toolbox's broad versatility for fields from coordination chemistry and spin‐crossover complexes to supramolecular chemistry through the characterisation of CoII and high‐spin FeII mononuclear complexes as well as a Co4L6 cage.


Introduction
NMR spectroscopy is indispensable for the solution structural characterisation of macromolecules from proteins [1] to supramolecular architectures, [2] including interlocked structures, [3] metal-organic cages [4] and topologically complex molecules. [5] With the standard suite of 1D and 2D NMR methods,s tructural assignment of diamagnetic compounds and complexes is straightforward but NMR spectroscopy in the presence of paramagnetic centres is more difficult.
Paramagnetic NMR spectroscopy [6] is central to many fields from chemical and structural biology for studying the structure,d ynamics and interactions of proteins [1b,c,7] to probing spin-state populations in spin-crossover compounds [8] and the structural characterisation of paramagnetic complexes [9] and supramolecular architectures. [4b, 10] However, NMR data acquisition and interpretation in the presence of aparamagnetic centre presents anumber of challenges due to the large paramagnetic shifts,s hort relaxation times and broad linewidths:pulse programs with long or multiple pulses are not suitable since relaxation can occur before data acquisition takes place; [6b] uniform excitation is more difficult over the larger spectral range;s ome signals may be lost completely in the case of very short relaxation times and broad linewidths; [6e] structural information usually extracted from the chemical shift and J-coupling is lost. [6e,9a, 10a] Furthermore,t here are limitations to current methods for spectral assignment;t hey often rely on the availability of accurate theoretical models [6b,9b, 11] or single crystal X-ray structures for correlation of T 1 relaxation times to metalproton distances using the Solomon equation. [4b, 10b,d] As ar esult, solution characterisation is,i nm any cases,l imited to a 1 HNMR spectrum only where complete and unambiguous assignment may not be possible.
Nevertheless,p aramagnetic NMR spectroscopy also has advantages compared to diamagnetic NMR spectroscopy:the large paramagnetic shifts result in reduced likelihood of signal overlap from dispersion of the NMR signals over aw ider chemical shift range; [10c] the fast relaxation times in comparison to diamagnetic compounds enables reduction of the acquisition times and recycle delays,t hereby significantly reducing the demand on instrument time. [4b, 6b,12] Alternatively,t his can be exploited for ag reater sensitivity of detection through extensive scan averaging. In supramolecular chemistry this has allowed detection of guest binding within paramagnetic cages, [4b, 10c] even when the guest is present as atrace impurity. [4b] Despite these advantages,t he full potential of paramagnetic NMR spectroscopy is still to be realised;i n comparison to the wealth of 1D and 2D NMR methods for diamagnetic compounds,the number of methods suitable for paramagnetic complexes is limited by fast relaxation although paramagnetic DOSY has been recently reported. [6e, 13] Furthermore,t he data acquisition/interpretation difficulties [6c] still need to be overcome to enable straightforward structural characterisation by paramagnetic NMR spectroscopy.
We report at oolbox of 1D ( 1 H, proton-coupled 13 C, selective 1 H-decoupling 13 C, steady-state NOE) and 2D (COSY,NOESY,HMQC) paramagnetic NMR methods that have proven particularly robust towards fast relaxation and enable unprecedented in-depth structural analysis of para-magnetic complexes in solution. We demonstrate the general applicability of this selection of robust experiments by characterising paramagnetic complexes from various fields of chemistry:Co II mononuclear complexes 1a-7a( Figure 1a) as representative examples of paramagnetic coordination complexes;F e II mononuclear complex 1b (Figure 1a), whose perchlorate and tetrafluoroborate salts are known to undergo spin-crossover in the solid state, [14] to represent the high-spin state of as pin-crossover complex;a nd metal-organic cage 8 ( Figure 1b)asanexample of aparamagnetic supramolecular architecture.

Results and Discussion
We initially investigated the mononuclear complexes to optimise the toolbox and test its limits for spectral assignment. Since the mononuclear complexes could form amixture of meridional (mer)and facial (fac)isomers,uptofour sets of NMR signals could, in principle,b eo bserved based on symmetry considerations:1set for the fac isomer (represented as green ligands in Figure 1a)a nd 3s ets for the ligand in the three different environments in the mer isomer (represented as black, red and blue ligands,F igure 1a). The2pyridylquinoline (pq) coordination motif was chosen to investigate the influence of steric bulk on the fac/mer ratio since its coordination chemistry with labile octahedral metal ions has been underexplored in solution. [11b] Co II complex 1a with the parent pq [15] ligand (SI, Section 2.1) was studied first rather than the literature-known Fe II complex 1b, [14] which could undergo spin-crossover complicating NMR analysis.Complex 1a was prepared either in situ or as crystals by mixing Co(BF 4 ) 2 ·6 H 2 Oa nd three equivalents of the ligand in CD 3 CN or EtOH, respectively. Single crystal X-ray analysis revealed mer-1a crystallised (Figure 2a nd SI, Section 3.1.1.1). [18] Like the analogous crystal structure of [Fe(pq) 3 ](BF 4 ) 2 (1b), [14a] adoption of the mer configuration is attributed to the ligands steric bulk and p-p stacking interactions are observed between two of the ligands (black and red ligand environments in Figure 1a) causing distortion of the N-Co II -N angles from the ideal 908 8 octahedral geometry (Table S3).
In solution, the 1 HNMR spectra of the complex prepared in situ and by redissolving the crystals in CD 3 CN were similar ( Figure S32). The 1 Hs ignals were spread over a2 50 ppm range with 24 relatively sharp signals of equal intensity (linewidths up to 70 Hz, Table S5) and 4b roader signals ( Figure S34). This suggests that the mer isomer is not only the solid-state structure but also the only structure in solution;the deviation from the expected 30 signals is attributed to either overlapping or broad signals.
Assignment of the 1 Hs pectrum was initially attempted using the Solomon equation, which has been successfully applied to assign the spectra of highly symmetric cages by correlating T 1 relaxation times to the metal-proton distances from the single crystal X-ray structure. [4b, 10b,d] Forcomplex 1a the T 1 relaxation times varied from 0.7 to 80 ms (Table S4) but alimitation of this currently available assignment method was highlighted during analysis;o nly partial assignment was possible since not only protons d and g (Figure 3a)b ut also, and more importantly,t he three different ligand environments cannot be distinguished on the basis of T 1 relaxation times/Co II -proton distances alone.Ward also encountered this limitation in the assignment of al ower symmetry metalorganic cage. [10c] Therefore,wesought to remove the reliance of assignment on the Solomon equation by optimising atoolbox of paramagnetic NMR experiments with broad applicability for the straightforward characterisation of avariety of Figure 1. Paramagnetic a) mononuclear Co II and Fe II complexesb ased on sterically bulky 2-pyridylquinoline (pq) motifs and b) Co 4 L 6 cage for characterisation by the toolbox of paramagnetic NMR techniques. paramagnetic complexes.T he following description of the structural assignment of complex 1aand related mononuclear complexes is used to illustrate how the paramagnetic NMR toolbox was optimised to overcome commonly encountered data acquisition and interpretation difficulties.Aninstruction manual for application of the toolbox to the characterisation of other paramagnetic complexes and cages is provided in Section 1.1.1 of the SI.
We initially turned our attention to COSY since this would allow grouping of the resonances within the different ligand environments.Encouraged by the observation of crosspeaks in the COSY spectra of several paramagnetic complexes [9a] and supramolecular architectures [4b, 10a,b] despite the broad linewidths and short relaxation times,w et ested and optimised several COSY variants (SI, Section 1.1.1.2). Reduction of the recycle delays and acquisition times due to the fast relaxation times enabled the acquisition of more data in ashorter amount of time using these optimised paramagnetic COSY parameters (typically 5.5 mins for 4s cans,T able S1) compared to the standard COSY parameters (8 min for 1scan, Table S1).
Cross-peaks were observed in spectra recorded using the pulse programs cosygpqf (Figure 3a), cosyqf90 and cosygpmfqf ( Figure S35), although intense diagonal peaks [6b] and commonly observed artefacts,s uch as T 1 noise streaks and anti-diagonal peaks,w ere present to varying degrees dependent on the pulse program. Thea vailability of three suitable paramagnetic COSY pulse programs will enable broad applicability to av ariety of paramagnetic complexes and facilitate the implementation of paramagnetic COSY as astandard characterisation method.
Thet hree COSY spectra of complex 1a display al arge number of cross-peaks ( Figure 3a,F igure S35). Notably, COSY [19] cross-peaks expected on the basis of 3 J coupling were observable facilitating identification of neighbouring protons and thus,grouping of the proton resonances to one of the three ligand environments (black, red and blue circles, Figure 3a). Unexpectedly,a dditional cross-peaks were observed (orange and purple boxes in Figure 3a,F igure S35), although the COSY spectrum of [Co(bpy) 3 ](BF 4 ) 2 as ar eference complex showed only the two expected cross-peaks arising from 3 J coupling ( Figure S96). Theo rigin of these additional cross-peaks was investigated using NOESY since Wimperis and Bodenhausen [16c,d] as well as Bertini [16a,b] have reported the presence of additional relaxation-allowed crosspeaks in the COSY spectra of paramagnetic complexes that result not from through-bond coupling but rather crosscorrelation from through-space (NOE) coupling between the nuclei as well as between the nuclei and the paramagnetic centre.
Thes tandard NOESY pulse program was modified for application to paramagnetic complexes (SI, Section 1.1.1.3). Foraninitial experiment, amixing time of 20 ms was chosen as ac ompromise between the short relaxation times and comparatively long NOE cross-relaxation rates.P leasingly, cross-peaks for the sharp signals (linewidth < 70 Hz) were observed ( Figure 3b). Given the range of T 1 relaxation times within the complex, optimisation of the mixing time was investigated. As eries of NOESY spectra were measured varying the mixing time from 1mst o2 0msa nd the crosspeaks were integrated ( Figure S36). Forp rotons with T 1 relaxation times significantly longer than the mixing time, the exchange integral approached am aximum as the mixing  4 ) 2 (1a)w here through-bond( COSY) cross-peaks within the three ligand environments are represented by the red, black and blue circles, respectively.Additional structurali nformation in the form of exchange (EXSY) cross-peaks (orange boxes) and through-space(NOE) cross-peaks (purple boxes) is present due to chemical exchange and cross-correlation, [16] respectively.N ote:t he numbers 1, 2, 3o nthe proton labels represent the three sets of coupled protons within asingle spin system (i.e. protons b-d). The absence of NOE cross-peaks between protons h and j prevented assignmento fthese spin systems to aparticular ligand environmentand therefore, spin system j-l was arbitrarily labelled with black, red and blue labels in decreasing chemical shift order of proton j to represent the three ligand environments. b) 1 H-1 HN OESY NMR spectrum (600 MHz, CD 3 CN, 298 K) of mer-[Co(pq) 3 ](BF 4 ) 2 (1a). Exchange( EXSY) cross-peaks are represented by the orange boxes and exchange cross-peaks between the complex and excess ligand are shown by the green boxes. time increased. However,f or protons with shorter T 1 relaxation times (e.g.protons d and g), the exchange integral reached am aximum before decreasing as relaxation began competing with exchange when the mixing time increased. A mixing time of 10 ms was found to be agood compromise for maximising the exchange cross-peak of all protons despite their differing T 1 relaxation times.
Analysis of the NOESY spectrum revealed groups of three cross-peaks (orange boxes, Figure 3b), corresponding to chemical exchange between the three different ligand environments of the mer isomer.T hus,the spectrum has no NOE cross-peaks but is an EXSY spectrum since the mixing time was so short. Furthermore,E XSY cross-peak intensities can be close to 100 %, [6c] whereas NOE intensities are,ingeneral, small for small molecules and reduced even further by the fast relaxation from coupling to the paramagnetic center. [6d] Since some excess ligand was also present in the complex solution, exchange cross-peaks were also observed between the excess ligand and ligand in the complex (green boxes,F igure 3b), enabling assignment of protons b-l in the complex using the free ligand assignments (Table S6) despite broadening and small shifts between free and excess ligand signals due to the presence of Co II .
TheC OSY spectrum ( Figure 3a)w as then reanalysed with the proton assignments to determine the origin of the additional cross-peaks beyond the expected COSY crosspeaks (black, red and blue circles). These additional crosspeaks correspond to structural information that is not typically observable in the COSY spectra of diamagnetic compounds;r elaxation-allowed through-space (NOE) crosspeaks between protons d and g (purple squares) were observed due to cross-correlation [16] and EXSY cross-peaks (orange boxes) were observed due to exchange between the three ligand environments,a sc onfirmed by the exchange cross-peaks in the NOESY spectrum ( Figure 3b).
Thus,a lmost complete assignment of the 1 Hs pectrum of 1a was possible using COSY and NOESY with the exceptions of the assignment of:i )spin systems g-h and j-l to aparticular ligand environment since relaxation-allowed (NOE) crosspeaks were not observed between protons h and j in the COSY spectrum;ii) the broad signals,which are proposed to be protons a and m due to their close proximity to the paramagnetic Co II centre.T OCSY ( Figure S37) and steadystate NOE experiments (Figures S38, S39) were carried out but exchange cross-peaks rather than long-range coupling and NOE cross-peaks,respectively,dominated these experiments. Therefore,t he unambiguous assignment of spin-system containing protons j-l to aparticular ligand environment was not possible and the spin system was arbitrarily labelled with black, red and blue labels according to decreasing chemical shift of proton j to represent the three different ligand environments.S teady-state NOE experiments,h owever,d id allow assignment of the broad signals above and below À22 ppm as protons a and m,r espectively ( Figure S40). The complete proton assignment of complex 1a was independently corroborated by T 1 relaxation measurements (Table S4) as well as exchange cross-peaks between the excess ligand present in the sample and the complex.
Having successfully assigned the 1 HNMR spectrum of 1a, we investigated assignment of the 13 CNMR spectrum using paramagnetic analogues of heteronuclear 1D and 2D techniques (e.g.H SQC and HMBC). Thep roton-coupled 13 C spectrum contained signals over almost a9 00 ppm range ( Figure S41), making uniform excitation over this very wide spectral range difficult. Therefore,o verlapping spectra of smaller spectral widths were acquired to cover the entire range.T he quaternary carbons could be distinguished from the tertiary carbons on the basis of the multiplicity and initially,the tertiary carbons were assigned using selective 1 Hdecoupling 13 Ce xperiments where one after another,e ach proton signal was selectively irradiated during repeated acquisitions of the 13 CNMR data (Figures S42-S46). However,t his method is time-consuming due to the number of signals and the sensitivity of 13 CNMR measurements and therefore,w ei nstead chose to investigate 2D heteronuclear experiments.
We focused on the HMQC pulse program as an alternative to HSQC because of its simple four pulse sequence and pleasingly,cross-peaks were observed for the sharp 1 Hsignals (linewidth < 70 Hz, Figures S47-S49). However,t he acquisition of two HMQC spectra was necessary to cover the large spectral range in both dimensions as non-uniform excitation resulted in ad ecrease in the intensity or complete loss of cross-peaks at the extremes of the spectral range ( Figure 4). Nevertheless,anoverlay of the two HMQC spectra confirmed the assignments made using the selective 1 H-decoupling 13 C experiments ( Figures S42-S46). HMBC spectra were also acquired in an attempt to assign the quaternary carbons and the three spin systems within aparticular ligand environment; however, no cross-peaks were observed due to the long delays resulting from the small magnitude of 3 J coupling constants in contrast to the large 1 J coupling constants utilised in the HMQC experiments.T herefore,t he quaternary carbons and carbon a were tentatively assigned through comparison to the reference complex [Co(bpy) 3 ](BF 4 ) 2 ( Figures S97-S99).
Thus,w es uccessfully assigned the 1 Ha nd 13 CNMR spectra of complex mer-1a using ac ombination of 1D ( 1 H, proton-coupled 13 C, selective 1 H-decoupling 13 C, steady-state NOE) and 2D (COSY,NOESY,HMQC) NMR paramagnetic experiments.W et hen sought to test the limits of this paramagnetic NMR toolbox in complexes where proton coupling within aspin system is interrupted by substitution as well as complexes with broader linewidths.W ep repared aseries of 2-pyridylquinoline ligands substituted in the 5'-or 6-positions (Figure 1a)a nd their corresponding Co II complexes 2a-7a in situ by mixing Co(BF 4 ) 2 ·6 H 2 Oa nd three equivalents of the ligand in CD 3 CN.
These complexes fulfil both criteria, as revealed by their NMR spectra ( Figures S93-S94, Table S5), and they also allowed investigation of the influence of substitution on the fac/mer isomerism. Based on comparison to the 1 HNMR spectrum of the parent complex 1a,itappeared that the mer isomer also predominated for complexes 3a-7a but additional species were also present ( Figures S93-S94). In the case of the complex with pq-5'-Br, as et of broader signals consistent with only one ligand environment was also as ignificant species within the mixture ( Figure S51).
Initially,assignment of the proposed mer isomer in the 1 H and 13 CNMR spectra of complexes 2a-7a was investigated. In comparison to the parent complex 1a,t he 1 Hs ignals of complexes 2a-4a with 5'-substituted ligands had the largest linewidths (> 90 Hz, Table S5) followed by complexes 5a-7a containing 6-substituted ligands (< 130 Hz, Table S5). Nevertheless,c ross-peaks were still observable in the 2D spectra (COSY,N OESY and HMQC) including exchange crosspeaks in the COSY spectra to varying degrees for complexes 3a-7a ( Figures S59, S66, S72, S79, S87). The 1 Ha ssignments for these complexes confirmed that in each case the major species in solution is the mer isomer ( Figures S93-S94). 1 H assignment of mer-2awas also possible despite the absence of many cross-peaks in the COSY spectrum since exchange cross-peaks were still observable in the NOESY spectrum. We attribute the incompleteness of cross-peaks in the COSY spectrum to the presence of broad linewidths (> 200 Hz, Table S5). Assignment of the 13 CNMR spectra of complexes 2a-7a was not as straightforward as the 1 Hs pectra, most likely due to the lower sensitivity of 13 CNMR spectroscopy compared to 1 HNMR spectroscopy,t he broadness of the signals and the presence of multiple species.T hus,i ns ome cases only partial assignment was possible ( Figures S62-S63, S68-S69, S74-S75, S81-S83, S89-S91).
Ac omparison of the 1 Ha ssignments for mer-2a-7a to those of the parent complex mer-1ashowed that the signals of the 2-pyridylquinoline backbone do not shift significantly upon substitution and as expected, the 1 Hs ignals for the 5'and 6-positions are not present in the spectra of complexes 2a-4a and 5a-7a,r espectively,d ue to substitution (Figures S93-S94). Theprotons in these spin systems with substituents could be assigned by the exchange peaks in the NOESY spectra but not to aparticular ligand environment, due to the disruption of proton coupling by substitution in the COSY spectrum and absence of suitable TOCSY and HMBC pulse programs for paramagnetic complexes.
We then investigated the assignment of the species other than the mer isomer in the spectra of 2a-7a.Aset of sharp signals consistent with only one ligand environment was visible in the spectra of complexes 3a and 5a ( Figures S93-S94), and another set of broader signals,a lso consistent with only one ligand environment, was visible for all complexes with the exception of 6a (Figures S93-S94). We propose these two species to be the fac isomer and asymmetric CoL 2 -based species.W hile these species cannot be distinguished on the basis of the number of NMR signals,w ea ttribute the set of broader signals to asymmetric CoL 2 -based species;this set of broader signals was significant for the complex with pq-5'-Br yet decreased in intensity upon addition of afourth equivalent of ligand while the mer-2a signals increased ( Figure S51). Furthermore,intwo samples of the complex with pq-6-Ph, the chemical shifts of the proposed symmetric CoL 2 -based species were sensitive to the differing water content of the samples whereas those corresponding to mer-7a were not (Figure S86). We attribute the observation of as ymmetric CoL 2based species as well as mer-2a to the steric bulk of pq-5'-Br from not only the quinoline ring but also the bromine substituent since this could reduce the efficiency of p-p stacking interactions between two of the ligands.Incontrast, the predominance of mer-5a based on the pq-6-Br ligand is likely due to the reduced steric influence of the bromine substituent in the 6-position compared with the 5'-position in complex 2a.
Assignment of these species was more difficult than the mer isomer since cross-peaks for these species were not typically observable in the COSY spectra, most likely due to the broadness of the signals in the case of the CoL 2 -based species and low concentration in the case of the fac isomer (estimated to constitute less than 1% of the complex mixture). However,N OESY appears to be less sensitive to signal broadness than COSY since exchange cross-peaks between the CoL 2 -based species and the mer complex were still observed ( Figures S54, S60, S67, S73). Furthermore,these  (1a). Twospectra with differing offsets (represented by the dashed lines) were recorded since uniform excitation could not be achieved with standard square pulses over the large spectral range resulting in absent cross-peaks or cross-peaks with reduced intensities (green circles) at the extremes of the spectral range.
cross-peaks were even visible when the CoL 2 -based species was not detectable in the 1 HNMR spectrum due to signal broadness and/or the low concentration of this species as seen in the spectrum of complex 6a (Figures S78, S80).
To further investigate the applicability of the paramagnetic NMR toolbox we extended our studies to the characterisation of ah igh spin/spin-crossover Fe II complex. Complex 1b was prepared in aglovebox by mixing Fe(OTf) 2 and three equivalents of the pq ligand in dry CD 3 CN.A tr oom temperature the 1 HNMR spectra of the complex contained broad signals for the complex and therefore,detailed analysis was not possible.Weattributed the broad signals to fast ligand exchange processes and therefore,v ariable temperature experiments were carried at lower temperatures where ligand exchange would be slower.
Upon cooling the solutions from 298 Kt o2 48 K, the signals sharpened and displayed Curie-Weiss behaviour ( Figure S106-S109). [17] There was no evidence of spin-crossover over this temperature range,c onsistent with previous studies on the tetrafluoroborate salt of complex 1b in acetone. [14b] At 248 Ka lmost complete assignment of the 1 H and 13 CNMR spectra was possible for mer-1b since the signals were relatively sharp and cross-peaks were observable in the COSY,N OESY and HMQC spectra (Figures S101-S105). However,t he carbon signals for d and g could not be assigned on the basis of the HMQC spectrum since the crosspeaks were absent or very weak, attributed to the increased influence of the paramagnetic Fe II ion on the relaxation times at lower temperature.I na ddition, at least one other species was present at equilibrium as broader signals were also seen in the 1 HNMR spectrum but they could not be assigned in the absence of cross-peaks in the 2D NMR spectra.
Given the large change in the linewidth of the 1 HNMR signals between 248 Ka nd 298 K ( Table S7), we carried out temperature-dependent COSY,N OESY and HMQC experiments to investigate whether cross-peaks were also observable at higher temperatures ( Figures S110-112). HMQC and COSY appeared to be more sensitive to temperature than NOESY since by 266 Km ost cross-peaks were no longer observable (Figures S110, S112). Nevertheless,assignment of the 1 Hsignals was still possible exploiting the exchange crosspeaks in the NOESY spectra ( Figure S111) and the assignments from lower temperatures to assign coupled protons within asingle spin system when COSY cross-peaks were not observable at that temperature (Figures S110).
Finally,weapplied the paramagnetic NMR toolbox to the characterisation of paramagnetic metal-organic cages.T he rational design of lower symmetry metal-organic cages is challenging and therefore,w ep repared and characterised instead the highly symmetric tetrahedral Co II 4 L 6 cage 8 as proof-of-principle for paramagnetic cage characterisation (Figures S113-S117). As olution of four equivalents of Co(NTf 2 ) 2 and six equivalents of ligand was heated at 50 8 8C in acetonitrile and the cage was isolated by precipitation with diethyl ether.
The 1 HNMR spectrum of the redissolved cage in CD 3 CN contained 5sharp and 2broad signals,reflecting the presence of one ligand environment due to fac coordination around the metal centres.O nly the expected through-bond cross-peaks were observed in the COSY spectrum ( Figure S114). Full and unambiguous assignment of the 1 Hand 13 CNMR spectra was possible using the paramagnetic NMR toolbox, with the exception of the quaternary carbons and the broad signals corresponding to a and j (Figures S113, S115).
Following the characterisation of mononuclear complexes 1a-7a, 1b and cage 8,w ep ropose aw orkflow including troubleshooting experiments for the application of the paramagnetic NMR toolbox to the structural characterisation of other paramagnetic complexes and cages ( Figure 5). Am ore detailed instruction manual for each toolbox experiment is provided in the SI (Section 1.1.1). Thew orkflow begins with the acquisition of a 1 HNMR spectrum to establish the spectral width as well as the linewidths of the signals since large spectral widths often necessitate the acquisition of several 2D spectra with smaller spectral widths and the observation of cross-peaks in 2D spectra is,i nm any cases, dependent on the linewidth. With broad linewidths (typically > 100 Hz), variation and optimisation of the temperature is recommended in an attempt to reduce the linewidths and increase the likelihood of cross-peak observation in 2D experiments.
COSY is recommended as the second toolbox experiment to identify coupled protons within each spin system. However, additional relaxation-allowed exchange and NOE cross-peaks may be observable complicating assignment and therefore, either NOESY (toolbox experiment 3a) or steady-state NOE (toolbox experiment 3b) experiments are recommended to complete the 1 HNMR spectrum assignment. These two experiments provide similar structural information but NO-ESY has the advantage that the exchange and NOE crosspeaks can be observed in asingle 2D spectrum compared with multiple 1D spectra for steady-state NOE experiments.F or this reason, steady-state NOE experiments may only be necessary in the case of troubleshooting (Sections 1.1.1.3 and  1.1.1.4).

Conclusion
We report at oolbox of 1D ( 1 H, proton-coupled 13 C, selective 1 H-decoupling 13 C, steady-state NOE) and 2D (COSY,NOESY,HMQC) paramagnetic NMR methods that enables the straightforward characterisation of paramagnetic complexes.T his toolbox overcomes the data acquisition challenges due to the presence of paramagnetic centres,such as large paramagnetic shifts and short relaxation times,a nd also removes the reliance of data interpretation on theoretical models [6b, 9b,11] or the Solomon equation. [4b,10b,d] We demonstrated the general applicability of this toolbox for fields from coordination chemistry to spin-crossover complexes and supramolecular chemistry through the characterisation of Co II and high-spin Fe II mononuclear complexes as well as aCo 4 L 6 cage.Furthermore,wedemonstrated the toolbox can be successfully applied to structural characterisation in av ariety of situations:t he assignment of complexes with multiple ligand environments (e.g. mer complexes), complexes with ar ange of signal linewidths (including broad signals in the case of the CoL 2 -based species) and mixtures of complexes (e.g. mer-a nd fac-CoL 3 isomers as well as CoL 2based species).
This study also shows the advantages of paramagnetic versus diamagnetic NMR spectroscopy;t he short relaxation times of paramagnetic complexes enable reduction of the repetition delays and acquisition times,t hereby significantly reducing the experiment times or enabling the acquisition of more data in asimilar time (Table S1). In addition to reduced signal overlap and increased sensitivity,s tructural information can be observed by paramagnetic NMR spectroscopy that would not be observable in the diamagnetic analogue;in the COSY spectra of the Co II mer mononuclear complexes, relaxation-allowed through-space (NOE) cross-correlation peaks and exchange (EXSY) cross-peaks were observed in addition to the expected through-bond (COSY) cross-peaks. Furthermore,t he sensitivity of the exchange NOESY technique enabled the identification of additional species present at equilibrium that were not visible in the 1 HNMR spectra due to broad linewidths and/or their low concentration.
While solution characterisation of paramagnetic complexes and cages was previously typically limited to a 1 HNMR spectrum only,w ed emonstrate that in-depth structural analysis comparable to that for diamagnetic compounds is now possible using the paramagnetic NMR toolbox. We are now extending the use of this toolbox to the characterisation of more complex as well as mixtures of supramolecular architectures.