Platinum(ii) complexes of mixed-valent radicals derived from cyclotricatechylene, a macrocyclic tris-dioxolene† †Electronic supplementary information (ESI) available: Experimental procedures for the physical and computational characterisation of the compounds in this work; crystallographic figures, 

The redox series [1]0/1+/2+/3+ has been characterised by UV/vis/NIR spectroelectrochemistry, cw EPR, ENDOR and HYSCORE spectroscopies and DF calculations.

Experimental details for the crystal structure determinations Table S1. Experimental data for the crystal structure determinations in this work.

Experimental details for the electrochemical, spectroscopic and computational studies
Figure S1 Electrospray mass spectra of 2 and 3. Figure S2 1 H NMR spectra of 2 in {CD 3 } 2 SO at 298 K, under an N 2 atmosphere and in air.                               Single crystal X-ray structure determinations Slow diffusion of methanol into a solution of 2 in dma under an N 2 atmosphere yielded a mixture of two different solvate crystals. Both these were analysed, using a Rigaku Saturn CCD diffractometer in station I19 at the UK Diamond Light Source synchrotron ( = 0.6889 Å). Experimental details of the structure determinations in this study are given in Table S1. All the structures were solved by direct methods (SHELXS97 1 ), and developed by full least-squares refinement on F 2 (SHELXL97 1 ). Crystallographic figures were prepared using XSEED. 2 The entire contents of the asymmetric unit of 2·H 2 O·8dma are disordered, across a crystallographic mirror plane. Metal ion Pt(1) lies on this mirror plane, but every other atom in the complex and the resolved solvent residues is half-occupied and on a general crystallographic site. The asymmetric unit contains ⅓ of the complex half-molecule, spanning a C 3 axis, and two dma half-molecules which lie near the same mirror plane as the complex. The four half-occupied phenyl groups in the asymmetric unit were refined as rigid hexagons, but no other restraints were applied to the model. There is no resolved organic solvent in the cavity of the complex. However, a Fourier peak lying on the same C 3 axis as the complex molecule, [⅓, ⅔, z], refined reasonably as a half-molecule of water. The contents of the cavity of the complex were further defined by a SQUEEZE analysis of the final model. 35 This identified only small cavities totalling 490 Å 3 per unit cell, which is 7% of the cell volume V, containing 192 electrons. That is equivalent to a void volume of 245 Å 3 and 96 electrons per molecule, which corresponds perfectly to two molecules of dma (48 electrons each). Therefore, the cavities of the complex are assumed to contain two additional, unresolved dma molecules, and this formula was used for the density and F(000) calculations. All non-H atoms in the model were refined anisotropically, while H atoms were placed in calculated positions and refined using a riding model. The asymmetric unit of 2·2H 2 O·1.3dma·0.5MeOH contains one molecule of the complex, two disordered solvent sites and some additional Fourier peaks that were also modelled as partial solvent environments. Six phenyl groups in the complex were obviously disordered, and refined over two sites as rigid hexagons. Two other phenyl groups in the model also deviate significantly from planarity which indicates disorder, but attempts to resolve that were unsuccessful in those cases. The dma molecule in the cavity of the complex was refined over two sites, with occupancies 0.5 and 0.3. Another solvent site was modelled as a mixture of dma and water, each with half-occupancy. A methanol molecule was also clearly resolved in the model, which was refined as halfoccupied. Lastly, three other Fourier peaks that were not bonded to any other atom were included as partial water sites. Fixed interatomic distance restraints were applied to all the partial dma molecules. All wholly occupied non-H atoms were refined anisotropically, while C-bound H atoms were placed in calculated positions and refined using a riding model.

EPR Spectroscopy
Multifrequency EPR measurements were carried out at the EPSRC National UK EPR Facility and Service in the Photon Science Institute at The University of Manchester. Samples were prepared by oxidation of the neutral complex by ferrocenium hexafluorophosphate in CH 2 Cl 2 solution at -78 °C with a concentration ranging 3-5 mM. A small amount of THF was added to the reaction mixture before samples were loaded into the appropriately sised quartz tubes. S-and X-band fluid solution continuous wave (cw) spectra were collected using a Bruker EMX Micro spectrometer. Frozen solution X-and Q-band cw spectra were measured using a Bruker EMX spectrometer. Simulations of cw data were performed using Bruker's Xsophe software package. 3 Pulsed X-and Q-band EPR measurements were performed on a Bruker ELEXSYS E580 spectrometer at 20 K. The standard dielectric ring Bruker EPR cavities (ER 4118S-MS5), (ER4118X-MD5) and (EN4118X-MD4) were used, which were equipped with an Oxford CF 935 helium flow cryostat. X-band ESE-detected EPR spectra were measured with a 16 -200 -32 ns Hahn echo pulse sequence; Q-band utilised a 22 -400 -44 ns sequence. Davies ENDOR were performed using the pulse sequence π inv − RF − π/2 − τ − π − τ -echo. 4 RF pulses of 16 μs were generated by the Bruker "DICE" system and amplified by a 60 dB gain ENI A-500 RF amplifier. The threepulse ESEEM (stimulated echo) experiments were performed using the pulse sequence π/2 − τ − π/2 − T − π/2 − τ − echo as a function of T at different, fixed time τ. 5 The 2D ESEEM spectra, so-called hyperfine sublevel correlation (HYSCORE) spectra, 6 were recorded employing the sequence π/2 − τ − π/2 − t 1 − π − t 2 − π/2 − τ − echo with mw pulses of length t π/2 = 16 ns and t π = 32 ns, starting times t 1,2 = 500/300 ns, and time increments Δt 1,2 = 20 ns. The intensity of the inverted echo following the fourth pulse is measured with t 2 and t 1 varied and constant τ (of 300 ns). Unwanted features from the experimental electron spin echo envelopes were removed by using a four-step phase cycle. 7,8 In both dimensions 256 data points were collected. The relaxation decay was subtracted using baseline corrections (fitting by polynomials of 3-6 degree) in both time domains, subsequently applying apodisation (Hamming window) and zero-filling to 1024 data points in both dimensions. After 2D fast Fourier transformation absolute-value spectra were obtained. Initial analysis of the cross-ridges in (݊ ଵ ଶ ) versus (݊ ଶ ଶ ) allows in many cases for simultaneous determination of the isotropic and anisotropic components of the hyperfine matrix. [9][10][11] Simulations of EPR, ENDOR and HYSCORE data were performed using Easyspin, 12,13 with the following spin-Hamiltonian (Eq. 1): where β is the Bohr magneton, β n is the nuclear magneton, g i n is the g-factor of the ith nucleus, and g represents the electronic g-matrix. The first and second terms in the expression correspond to the electron and nuclear Zeeman interactions with the external magnetic field; the third term describes the hyperfine interaction defined by matrix A. The last term describes the quadrupole coupling defined by tensor P, which is traceless. The representation in its principal axes system is defined by Eq. 2.
Here, we will use the two usual parameters to characterise the 14 N quadrupole coupling constants: K = e 2 qQ/4h and η = (P x -P y )/P z . In all calculations, the electron Zeeman interaction was assumed to be the dominant term. All other interactions were treated as a perturbation. The orientation of the A-matrix and P-tensor was defined with respect to the principal axes of the electronic g-matrix.

Calculations
Geometry optimisations of all complexes were performed at the BP86 level using Gaussian 09 14 for energy calculations and the DL-FIND algorithm 15 implemented within the ChemShell package 16,17 for optimisation. The geometries of all complexes were fully optimised by spin-unrestricted DFT method with dichloromethane as solvent. The all-electron basis sets were those reported by the Ahlrichs group. 18,19 Triple--quality basis sets with one set of polarisation functions (def2-TZVP) were used for all atoms. The stability of all solutions was checked by performing frequency calculations: No negative frequencies were observed. Electronic properties were calculated on the optimised coordinates at the B3LYP 20,21 level of theory using ORCA. 22 The conductor like screening model (COSMO) was used for all calculations (except [3 • ] + ), 23 and RIJCOSX approximation 24 combined with appropriate Ahlrichs auxiliary basis sets were routinely employed to speed up the calculations. 25,26 Relativistic effects were accounted for using the zero order regular approximation (ZORA), 27 and enhanced integration accuracy was used for platinum (SPECIALGRIDINTACC 12). The exchange coupling constants J were calculated on broken-symmetry 28-32 geometries using Eq. 3, 33,34 and assuming the spin-Hamiltonian Eq. 4 is valid.
TD-DFT calculations were performed using the B3LYP functional with dichloromethane as a solvent. The first 40 states were calculated, whereas the maximum dimension of the expansion space in the Davidson procedure (MAXDIM) was set to 400. The full width at half maximum (FWHM) was set to 3500 cm -1 . Molecular orbitals and spin density maps were visualised via the program Molekel. 35

Other measurements
Elemental microanalyses were performed by the University of Leeds School of Chemistry microanalytical service. Electrospray mass spectra (MS) were obtained on a Bruker MicroTOF spectrometer, from MeCN feed solutions. All mass peaks have the correct isotopic distributions for the proposed assignments. Alkali metal cations and formate anions in the molecular ion assignments originate from calibrants in the spectrometer feed solutions. NMR spectra were run using a Bruker Avance 500 spectrometer operating at 500.1 MHz ( 1 H) or 125.6 MHz ( 31 P). UV/vis/NIR spectra for redox titrations were run on a Perkin Elmer Lambda900 spectrophotometer using 1 cm quartz cells.
Electrochemical measurements were carried out using an Autolab PGSTAT20 voltammetric analyser, under an argon atmosphere, in predried CH 2 Cl 2 containing 0.5 M [ n Bu 4 N]BF 4 as supporting electrolyte. Voltammetry experiments used a Pt disk working electrode, a Pt rod counter electrode and an Ag/AgCl reference electrode. All potentials quoted are referenced to internal ferrocene and were obtained at a scan rate (ν) of 100 mV s -1 . The Fc +/0 couple under these conditions was observed at +0.42 ≤ E 1/2 ≤ 0.48 V vs. Ag/AgCl. UV/vis/NIR spectroelectrochemistry experiments were conducted using a Cary 5000 spectrophotometer, fitted with an optically transparent thin layer electrode (OTTLE) cell. The working electrode was platinum gauze in the path of the beam, the counter electrode was platinum wire, and silver wire was used as the reference electrode. The electrolysis was controlled by an EG&G 273A potentiostat/galvanostat.   The weak signal at g = 2.03 indicates that the NMR line-broadening observed in air-exposed solutions of the compound (Fig. S4) reflects partial oxidation of the [ctc] 6ligand.    This orientation of the dma molecule, and the other disorder orientation (not shown), both have a methyl group pointing into the hydrophobic cavity of the complex. That closely resembles that seen for other solvents included into the cavity of H 6 ctc and its complexes, including 1. [1,2] 85.53(10) 85.41(10)  is a bond-valence sum parameter giving the oxidation state of dioxolene groups, which takes the values of 0, -1 and -2 for the q, sq and cat levels respectively. 36 Although there is some scatter, all the values in the                   Figure S25. MO energy level scheme of frontier Kohn-Sham orbitals for 1 with C 3v symmetry labels.