Ultrafast and long-time excited state kinetics of an NIR-emissive vanadium(iii) complex I: synthesis, spectroscopy and static quantum chemistry

In spite of intense, recent research efforts, luminescent transition metal complexes with Earth-abundant metals are still very rare owing to the small ligand field splitting of 3d transition metal complexes and the resulting non-emissive low-energy metal-centered states. Low-energy excited states decay efficiently non-radiatively, so that near-infrared emissive transition metal complexes with 3d transition metals are even more challenging. We report that the heteroleptic pseudo-octahedral d2-vanadium(iii) complex VCl3(ddpd) (ddpd = N,N′-dimethyl-N,N′-dipyridine-2-yl-pyridine-2,6-diamine) shows near-infrared singlet → triplet spin–flip phosphorescence maxima at 1102, 1219 and 1256 nm with a lifetime of 0.5 μs at room temperature. Band splitting, ligand deuteration, excitation energy and temperature effects on the excited state dynamics will be discussed on slow and fast timescales using Raman, static and time-resolved photoluminescence, step-scan FTIR and fs-UV pump-vis probe spectroscopy as well as photolysis experiments in combination with static quantum chemical calculations. These results inform future design strategies for molecular materials of Earth-abundant metal ions exhibiting spin–flip luminescence and photoinduced metal–ligand bond homolysis.

The d 2 -vanadium(III) polypyridine complex [V(ddpd) 2 ] 3+ emits in the NIR (1100 nm), observed for the rst time even at room temperature in solution (Chart 1). 36 Its phosphorescent singlet state with a lifetime of 0.79 ms/8.8 ms (93%/7%; 77 K in butyronitrile glass) is populated within picoseconds aer excitation. Yet, the efficiency of the population transfer to the singlet states by ISC is rather small as conrmed by nonadiabatic molecular dynamics calculations. 36 Unexpectedly, and in contrast to the analogous chromium(III) spin-ip emitters, the decay of the phosphorescent spin-ip states of [V(ddpd) 2 ] 3+ is insensitive to ligand deuteration, in spite of the signicant spectral overlap of the NIR emission with the second aromatic C-H overtone n 3 CH of the ligand. This suggests that other non-radiative decay pathways are more relevant than the multiphonon relaxation involving high-energy C-H oscillators 57 in this particular case. 36 A profound understanding of the decisive excited states and the excited state dynamics of these polypyridine vanadium(III) chromophores on ultrafast (population of emissive states) and slow timescales (depopulation of emissive states) is lacking. To better understand the novel class of d 2 -spin-ip luminophores based on vanadium(III) with respect to the population and decay of the emissive spin-ip states, we selected the chlorido vanadium(III) complex VCl 3 (ddpd) (Chart 1) 58 for a detailed study of the photodynamics at ultrashort (sub-picosecond) to microsecond timescales (Chart 1).
Spin-orbit effects are weak in vanadium(III) complexes based on the lower intensity of the singlet transitions compared to the triplet bands by more than three orders of magnitude. 59 Our choice of molecular system is guided by two considerations: (i) the slow ISC rate dened by the small SOC constant of vanadium(III) (z z 210/206/220 cm À1 ) 11,60,61 can increase due to the inuence of the coordinated chlorido ligands with their higher SOC constant (z z 547 cm À1 ) 61 as compared to nitrogen (z z 76 cm À1 ). 61 (ii) The symmetry reduction by using different ligand types can relax Laporte's rule 62 and increase the radiative rate from metal-centred spin-ip states. 35 We report here that the heteroleptic VCl 3 (ddpd) 58 complex is NIR-emissive at room temperature (Chart 1). We undertook a detailed photophysical study using Raman spectroscopy, variable temperature and variable pressure steady-state photoluminescence spectroscopy, fs-transient absorption spectroscopy, time-resolved photoluminescence and variable temperature step-scan FTIR spectroscopy to cover the ground state splitting, as well as the ultrafast and slow time regimes of the excited state kinetics. To elucidate whether non-radiative relaxation of the luminescent singlet states via aromatic C-H overtones (C-H ar ) plays a signicant role for the nonradiative relaxation in this particular case, the deuterated complex VCl 3 (ddpd-[D 17 ]) was prepared and studied for comparison. The electronic structures of ground and excited states at the ground state geometry were described utilizing relativistic two-component time-dependent density functional theory (TDDFT) and CASSCF-NEVPT2 calculations (NEVPT2 ¼ N-electron valence state perturbation theory to second-order). A detailed kinetic model of the excited state dynamics will be derived by trajectory surface hopping simulations within a linear vibronic coupling model in the accompanying paper. 63

Results and discussion
The vanadium(III) complex VCl 3 (ddpd-[D 0 ]) has been prepared as reported from VCl 3 (solv) 3 (solv ¼ CH 3 CN or THF) 64 and the ligand ddpd-[D 0 ] 65 as poorly soluble orange coloured complex. 58 Its structure and magnetic properties in the solid state have been reported. 58 Yet, luminescence, ultrafast and slow excitedstate dynamics as well as temperature and pressure effects on the decisive ground and excited states and the excited state kinetics remained unexplored.
To better describe the split 3 T 1 ground state and the intraand intercongurational metal-centred transitions of the vanadium(III) ion, we resorted to CASSCF(6,12)-FIC-NEVPT2 calculations on the CPCM(acetonitrile)-RIJCOSX-UB3LYP-D3BJ-ZORA/def2-TZVPP optimised triplet ground state geometry (for details see Experimental section). The active space included ve 3d and ve 4d orbitals as well as two occupied V-N/Cl s-bonding orbitals occupied with six electrons (Table S5 and Fig. S3, ESI ‡). 68 Without including the dynamic correlation with NEVPT2, the 1 E/ 1 T 2 and 3 T 2 states are quite close in energy (Table S6, ESI ‡). Inclusion of the dynamic correlation with NEVPT2 lowers the energy of the 1 E/ 1 T 2 states by ca. 2550 cm À1 and raises the energy of the 3 T 2 states by ca. 1200 cm À1 . On this level of theory, the energy gap between the lowest singlet and triplet excited states amounts to ca. 3750 cm À1 (Fig. 1b). These calculations place the lowest excited triplet states of 3 T 2 parentage at 14 968/16 464/17 393 cm À1 above the split 3 T 1 ground state (0/755/1076 cm À1 ) (Fig. 1b). The lowest energy spin-allowed transition is calculated at 14 968 cm À1 in reasonable agreement with the experimental band maximum (14 245 cm À1 , Fig. 1a inset). The calculated splitting of all 3 T states is substantial reecting the low symmetry of the complex. The ve lowest excited singlet states of 1 E/ 1 T 2 parentage are calculated at 10 086/10 161/10 949/11 068/ 11 309 cm À1 . These spread over 1200 cm À1 (Fig. 1b). As the energy of the lowest singlet state is signicantly lower than the lowest excited triplet state by ca. 4900 cm À1 , NIR phosphorescence from the singlet state to the split ground state is conceivable (Fig. 1b). Extended calculations using an even larger active space will be presented in the accompanying paper. 63 Excitation of solid VCl 3 (ddpd-[D 0 ]) with 350 nm (ILCT) at 298 K results in the appearance of two NIR emission bands ( Fig. 2a). At 77 K, the emission bands increase in intensity and develop a characteristic ne structure (Fig. 2a). Discernible peaks occur at 1102, 1219 and 1256 nm (9074, 8203, 7962 cm À1 ). Considering the calculated ground state splitting, we assign these clearly visible band maxima to radiative transitions from the lowest excited singlet state(s) to the split ground state (Fig. 1b). The resulting experimental ground state splitting of ca. 800 and 1100 cm À1 excellently agrees with the CASSCF-NEVPT2 calculated splitting (755 and 1076 cm À1 ). The experimental ground state splitting refers to the geometry minimum of the singlet state, while the CASSCF-NEVPT2 calculated splitting refers to the ground state geometry. As spin-ip states are rather nested, the geometry differences should be marginal. Raman spectra of [V(H 2 O) 6 ] 3+ and [V(urea) 6 ] 3+ show broad electronic Raman transitions around 1900-2900 cm À1 and 1400 cm À1 , respectively, 51,52 due to the trigonal ground-state splitting. For VCl 3 (ddpd-[D 0 ]), we observe two broad electronic Raman transitions around 500 and 900 cm À1 in its Raman spectrum in accordance with its lower symmetry (Fig. 2b). These energies t well to the splitting assigned by luminescence spectroscopy (Fig. 2a) and obtained from the CASSCF-NEVPT2 calculations.
The ner details of the NIR luminescence band structure can be tentatively assigned to population of the close-lying singlet states with the difference between the two lowest singlet states calculated as 75 cm À1 and to enabling vibrations around 120 cm À1 . Indeed, Cl-V-Cl deformation vibrations (125, 136, 159 cm À1 ; unscaled) appear in this energy region according to the DFT calculations (ESI, Fig. S4 ‡).
Upon pressurizing solid VCl 3 (ddpd-[D 0 ]) to 7 kbar, two very weak NIR emission bands at approximately 9100 and 9280 cm À1 (lowest energy detectable with the employed detector) shi to higher energy by z10 cm À1 kbar À1 . (Fig. S5, ESI ‡). This hypsochromic shi of the emission bands differs from the bathochromic behaviour encountered by the d 3 complex [Cr(ddpd) 2 ] 3+ at increasing pressure. 39 This unusual pressuresensitivity is probably a combined effect of energy changes of the emissive singlet states and of the ground state splitting under increasing pressure. The broad, electronic Raman bands assigned to transitions within the split 3 T 1g ground state experience variations in intensity and broaden strongly with increasing pressure (Fig. S6, ESI ‡). At pressures higher than 30 kbar the broadening, most likely due to effects of nonhydrostatic pressure, dominates and the bands can no longer be observed. These observations illustrate that luminescence shis different from those for spin-ip transitions with nondegenerate ground states are expected for vanadium(III) complexes.   17 ]), the NIR luminescence intensity strongly increases compared to that of the non-deuterated complex (Fig. 2c). Concomitantly, the luminescence lifetime at 298 K increases to s D 298 ¼ 3.3 ms (l obs ¼ 1106 nm) and 3.4 ms (l obs ¼ 1222 nm) ( Fig. S8 and S9, ESI ‡). The deuteration effect conrms that multiphonon relaxation (Fig. 1b) is substantial in VCl 3 (ddpd-[D 0 ]). The estimated spectral overlap integral (SOI) of the second C-H ar overtone n 3 CH (ref. 34) of the ligand at 8972 cm À1 is signicant, while the relevant third CD ar overtone n 4 CD (ref. 34) at 8755 cm À1 has a much lower SOI due to its lower extinction coefficient (ESI, Fig. S10-S12 ‡). Based on the vibrational overtone analysis and SOI calculation, the rate constant for this overtone-mediated nonradiative decay mechanism should diminish by a factor of 36 (ESI, Fig. S10-S16 ‡). This qualitatively matches the observed intensity enhancement upon deuteration. The observation of an isotope effect conrms that multiphonon relaxation is a major non-radiative decay path of the singlet states in this complex dominating other non-radiative decays. This nding contrasts the observations for the homoleptic complex [V(ddpd) 2 ][PF 6 ] 3 and its deuterated isotopologue where other decay pathways appear to dominate the non-radiative decay of the NIR-emissive states. 36 Cooling solid VCl 3 (ddpd-[D n ]) both as neat powder and as KBr pellet increases the luminescence intensity (Fig. 2a, d; and S17-S19, ESI ‡). For example, cooling VCl 3 (ddpd-[D 17 ]) from 290 to 200 K yields a 1.5-fold increased integrated NIR intensity, while further cooling to 5 K only has a minor effect ( Fig. 2d; and S17-S19, ESI ‡). This suggests the presence of a thermally activated non-radiative pathway accessible at temperatures above 200 K.
To probe the structure, the vibrational signature and possible distortions of the long-lived excited singlet states at high and low temperature, step-scan FTIR spectra of VCl 3 (ddpd-[D 0 ]) and Step-scan FTIR spectra recorded for VCl 3 (ddpd-[D 0 ]) and VCl 3 (ddpd-[D 17 ]) in a KBr pellet with l exc ¼ 355 nm collected over 0-500 ns at 290 and 20 K are depicted in Fig. 3a, S20, S21 and S23 (ESI ‡). The long-lived excited singlet states give rise to positive and negative bands corresponding to the population of the excited singlet states and ground state bleach, respectively. The excited state spectra of VCl 3 (ddpd-[D 0 ]) and VCl 3 (ddpd- [D 17 ]) at 290 and 20 K are derived from the respective step-scan and the ground state spectra (Fig. 3b; and S22, ESI ‡). Temperature has only a minor inuence on the excited state spectra of VCl 3 (ddpd-[D 0 ]) ( Fig. 3b), but modies the relative excited state IR intensities of the deuterated derivative VCl 3 (ddpd-[D 0 ]) (Fig. S22, ESI ‡).
The evolution of prominent IR bands aer excitation of VCl 3 (ddpd-[D 0 ]) over time was tted in a global analysis giving a monoexponential decay with s H 290K ¼ 0.6 ms at 290 K (Fig. S24, ESI ‡) excellently tting to the decay observed by photoluminescence spectroscopy at 298 K. Cooling to 20 K approximately doubles the excited state lifetime to s H 20K ¼ 1.3 ms (Fig. S25, ESI ‡). This conrms that thermally activated nonradiative pathways are operative at room temperature in addition to the multiphonon relaxation via C-H oscillators, 57 which takes place at all temperatures. Surprisingly, the step-scan data of the deuterated complex  the emissive singlet states but fails to probe the second emissive singlet state. According to the CASSCF-NEVPT2 calculations, the two lowest singlet states derive from terms with essentially 1 E and 1 T 2 character, respectively, with a very small energy difference of only 75 cm À1 (Table S6, ESI ‡). As the orbital population of the 1 T 2 -derived term matches that of the lowest term of the split 3 T 2 ground state (Fig. S3, ESI ‡), this excited state possesses the same equilibrium nuclear conguration as the ground state (nested state). Consequently, step-scan FTIR spectroscopy would not be able to detect this excited state. Clearly, a model of the excited state decay of the two lowest energy singlet excited states to the split ground state requires at least ve electronic states. For a kinetic model of the non-radiative decay via highand low-frequency modes 57 these comparably close-lying electronic states (Fig. 1b) would have to be combined with the different anharmonic vibrational C-H/C-D modes as well as the pyridine ring vibrational ladders. The latter modes are also affected by deuteration according to the ground state FTIR spectra of VCl 3 (ddpd-[D 0 ]) and VCl 3 (ddpd-[D 17 ]) ( Fig. 3a and S21, ESI ‡).
As VCl 3 (ddpd) is only poorly soluble in typical solvents, a detailed reliable study of its weak NIR luminescence in solution is unfortunately impeded, especially when exciting at the very weak 3 MC band. Furthermore, we noted a follow-up reaction upon irradiating VCl 3 (ddpd-[D 0 ]) at 350 AE 5 nm (ILCT) in acetonitrile solution. The absorption spectrum changes and an emission band at ca. 400 nm grows in over time ( Fig. S28 and S29, ESI ‡). The photostability is much higher under irradiation at 400 AE 5 nm including consideration for absorption and light intensity of the light source. (Fig. S30-S32, ESI ‡). This suggests that the low energy 3 T 2 , 1 T 2 / 1 E and ddpd / V 3 LMCT states are not responsible for the photoreactivity (Fig. 1b). At the higher excitation energy and with the assumption that LMCT states are likely involved (Fig. 1), we speculate that V-Cl homolysis could occur in solution. The well-known fact, that M-Cl bonds of reducible metal ions are prone to photoinduced homolysis has regained considerable interest in organic photoredox catalysis in particular for copper [69][70][71] and nickel. 72-75 VCl 3 itself is photoreduced to vanadium(II) in alcoholic solutions via excitation into LMCT states (chloride-to-vanadium or alkoxide-tovanadium charge transfer). 76 A 3 LMCT state with chloride-tovanadium character was found by TDDFT at 324 nm (state 14 shied hypsochromically by 3400 cm À1 , Tables S3, S4, and Fig. S2, ESI ‡). This 3 LMCT state could qualify as excited state with V-Cl dissociative character. To probe the conceivable V III/II reduction process, a cyclic voltammogram of VCl 3 (ddpd-[D 0 ]) was recorded in CH 3 CN. The cathodic scan reveals an irreversible reduction wave at E p ¼ À1.11 V versus ferrocene/ ferrocenium with an oxidative follow-up wave at E p ¼ À0.25 V and a reductive wave at À0.83 V (Fig. S33, ESI ‡). This behaviour can be associated with chloride loss upon electron capture, similar to the reported preparation of VCl 2 (py) 4 from VCl 3 , pyridine and zinc dust as reductant. 77 Consequently, we consider V-Cl bond homolysis as a viable reaction path under UV light photolysis in uid solution. In contrast to this photoreactivity of the chlorido complex, the homoleptic complex [V(ddpd) 2 ][PF 6 ] 3 appears comparably photostable in solution, which can be ascribed to the absence of suitable dissociative LMCT states. 36 Finally we explored the reaction path from the Franck-Condon excited triplet state to the long-lived singlet states by ultrafast transient absorption spectroscopy in CH 3 CN. To diminish the dissociative processes assigned to high-energy 3 LMCT states with Cl / V character, 400 nm pulses were employed populating essentially 3 LMCT states with NMe(ddpd) / V character (Tables S3, S4 and Fig. S2, ESI ‡). Aer excitation, a broad excited state absorption (ESA) from 470-700 nm appears in addition to an ESA around 410 nm (Fig. 4a). The ground state bleach ts to the dip in the transient absorption spectrum around 463 nm (Fig. 1a and 4b).
The broad ESA evolves with s 1 ¼ 1.5 ps to a long-lived state with a maximum at 543 nm (Fig. 4b). This state persists much longer than the time window of 1.4 ns of the pump-probe experiment, which is given by the length of the motorized delay stage in the setup. Since electronic relaxation between electronically excited states of the same spin multiplicity is typically rather fast, one of the lowest excited states of the different spin multiplicities should be responsible for the long- lived state, i.e. the 3 T 2 state or the 1 E/ 1 T 2 states (Fig. 1). Significant population in a long living electronically excited triplet state should result in uorescence, since the radiative transition to the ground state would be spin allowed and Laporte's rule is relaxed in the complex. However, no luminescence is observed at wavelengths below 1050 nm as would be expected for triplet states (Fig. 1). This excludes that a signicant excited state population is trapped in any triplet state. A partitioning of excited state population in triplet and singlet states, as it was observed in [V(ddpd) 2 ] 3+ (ref. 36) does not occur with VCl 3 (dppd). The persistent TA component is therefore assigned to the long-lived 1 E/ 1 T 2 states. Clearly, ISC to the singlet manifold and vibrational cooling proceed to completion within a few ps. The rate constant for ISC k ISC is at least 1/s 1 ¼ 6.7 Â 10 11 s À1 . Trajectory surface hopping simulations within a linear vibronic coupling model will derive a detailed kinetic model of the initial dynamics and the efficiency of the ISC processes (s ISC,simulation ¼ 1.7 AE 0.3 ps) in the accompanying paper. Photolysis experiments were carried out in CH 3 CN using an Asahi Spectra Max-303 Xenon Light Source (300 W, Fig. S32, ESI ‡), together with 350 AE 5 nm and 400 AE 5 nm lters, respectively.
UV/Vis photoluminescence spectra during photolysis experiments were collected on a Varian Cary Eclipse spectrometer.
UV/Vis/NIR spectra were recorded on a Varian Cary 5000 spectrometer using 1.0 cm cells.
Raman and luminescence spectra under pressure at wavelengths up to 1050 nm were measured with a Renishaw InVia microscope (488 and 785 nm laser wavelengths) and a HPDO diamond anvil cell.
Temperature-dependent steady-state NIR luminescence experiments down to 5 K were conducted on a Horiba Jobin Yvon Fluorolog 3-22 s spectrometer equipped with a 450 W xenon lamp and a DSS -IGA020L NIR detector (850 nm < l em < 1550 nm). Spectral selection was realized with double and single grating monochromators in the excitation and emission paths, respectively (excitation: 1200 grooves per mm; near-IR emission 600 grooves per mm). A combination of two long-pass lters (FELH0500 Thorlabs, transmission $92% above 500 nm and FELH0850 Thorlabs, transmission $98% above 1000 nm) was used in the emission channel to avoid higher order excitation light. For the preparation of KBr pellets, the compounds (ca. 1.0 mg for VCl 3 (ddpd-[D 0 ]) and 0.5 mg for VCl 3 (ddpd-[D 17 ])) were mixed with dry KBr (ca. 200 mg, stored in a compartment dryer at 80 C, purchased from Merck) and ground to a homogenous mixture. This mixture was lled into an evacuable pellet die with a diameter of 13 mm and sintered at a pressure of 0.75 GPa. Measurements on neat powders were conducted by homogenous spreading of the neat sample between two CaF 2 windows (13 mm diameter, 1 mm thick). Experiments at temperatures between 5 K and 290 K were performed using a closed-cycle helium cryostat (ColdEdge, 101J cryocooler). The cryocooler was equipped with a pellet holder (copper) and CaF 2 windows.
Steady-state NIR luminescence experiments on neat samples down to 77 K were conducted on a Horiba Fluorolog-3 spec-trouorimeter equipped with a 450 W xenon lamp for steadystate measurements. Emitted light was detected either by a Hamamatsu R2658P PMT detector (200 nm < l em < 1010 nm) or by a Hamamatsu H10330-75 PMT detector (950 nm < l em < 1700 nm). Spectral selection in the excitation path was accomplished by a DFX monochromator (double gratings: 1200 grooves per mm, 330 nm blaze) and in the emission paths in the visible/NIR spectral region (l em < 1010 nm) by a spectrograph iHR550 (single gratings: either 1200 grooves per mm, 500 nm blaze or 950 grooves per mm, 900 nm blaze) and in the NIR spectral region (l em > 950 nm) by a spectrograph iHR320 (single grating: 600 grooves per mm, 1000 nm blaze).
Near-IR luminescence lifetimes of the phosphorescent transitions were determined at 298 K (solid samples in standard NMR tubes under argon) with a PTI Quantamaster QM4 spectrouorimeter equipped with a 75 W continuous xenon short arc lamp as excitation source (Hamamatsu L4633: pulse width ca. 1.5 ms FWHM). Emission was monitored using a PTI P1.7R detector module (Hamamatsu PMT R5509-72 with a Hamamatsu C9525 power supply operated at 1500 V and a Hamamatsu liquid N 2 cooling unit C9940 set to À80 C). For the measurements above 1000 nm, a long-pass lter RG-850 (Schott, 3.0 mm thickness, transmission >98% above 970 nm) was used in the emission channel in order to avoid higher order excitation light. Spectral selection was achieved by single grating monochromators (excitation: 1200 grooves per mm, 300 nm blaze; Vis emission: 1200 grooves per mm, 500 nm blaze; near-IR emission: 600 grooves per mm, 1200 nm blaze) and an additional UG11 bandpass lter (Schott, 3.0 mm thickness) in the excitation channel. Lifetime data analysis (deconvolution, statistical parameters, etc.) was performed using the soware package FeliX32 from PTI. Lifetimes were determined by deconvolution of the decay proles with the instrument response function, which was determined using an empty NMR tube as scatterer. Estimated uncertainties for the lifetimes of the near-IR emission determined with this setup are 20%.
Time-resolved FTIR experiments were performed with the FTIR spectrometer Bruker Vertex 80v, operated in the step-scan mode. A liquid-nitrogen-cooled mercury cadmium telluride (MCT) detector (Kolmar Tech., Model KV100-1-B-7/190) with a rise time of 25 ns, connected to a fast preamplier and a 14 bit transient recorder board (Spectrum Germany, M3I4142, 400 MS s À1 ), was used for signal detection and processing. The laser setup used for the measurements includes a Q-switched Nd:YAG laser (Innolas SpitLight Evo I) generating pulses with a pulse duration of about 6 ns at a repetition rate of 100 Hz. The third harmonic (355 nm) of the Nd:YAG laser was used directly for sample excitation. The UV pump beam was attenuated to about 1.8 mJ per shot at a diameter of 9 mm. The beam was directed onto the sample and adjusted to have a maximal overlap with the IR beam of the spectrometer. The sample chamber was equipped with anti-reection-coated germanium lters to prevent the entrance of laser radiation into the detector and interferometer compartments. The KBr pellets were prepared as described in the section on luminescence spectroscopy, however, with a smaller amount of sample of ca.  17 ]) with a spectral resolution of 4 cm À1 resulting in 1110 and 555 interferogram points, respectively. An IR broadband lter (850-1750 cm À1 ) and CaF 2 windows (no IR transmission <1000 cm À1 ) prevented problems when performing a Fourier transformation (i.e. no IR intensity outside the measured region should be observed). FTIR ground state spectra were recorded systematically to check if there is no sample degradation. Estimated uncertainties for the excited state lifetimes are on the order of 10%.
Transient absorption spectra of VCl 3 (ddpd-[D 0 ]) were recorded applying a pump-probe setup with an excitation wavelength of 400 nm. The setup is pumped by a Ti:sapphire laser system (Spectra-Physics, Spitre Pro) which provides ultrashort laser pulses centred at 800 nm with a repetition rate of 1 kHz. By frequency doubling its output with a BBO-crystal pump pulses with a centre wavelength of 400 nm and a pulse duration of 200 fs were obtained. For probing, a white light continuum generated with a CaF 2 crystal was used. Both beams, with polarizations arranged in magic angle, were focused onto the sample leading to pump and probe spots with diameters of 170 mm and 80 mm, respectively. Transient absorption spectra were recorded by dispersing the probe beam aer the sample with a prism and detecting its spectral intensity distribution with a CCD array. The metal complex was dissolved in acetonitrile under argon atmosphere and the obtained sample was lled into a 1 mm fused silica cuvette. The concentration was 1.5 Â 10 À3 M resulting in an optical density of 0.18 at 400 nm. Signicantly higher concentrations were not accessible because of the moderate solubility of the compound in acetonitrile.
Quantum chemical calculations. The characterization of the absorption spectrum was done employing two types of quantum chemical calculations: (i) density functional theory in its unrestricted form and (ii) multicongurational theory with an active space tailored to predict the MC states. The rst method is labelled as "Unrestricted Kohn-Sham" orbitals DFT (UKS), the second as "SOC-CASSCF(6,12)-FIC-NEVPT2". These two methods are complementary to each other, as the rst gives energies of the CT states, while the second one provides the energies of the MC states and the ground state splitting. 78 Unrestricted Kohn-Sham orbitals DFT (UKS): All calculations were performed using the quantum computing suite ORCA 4.0.1.12. 79 Geometry optimization (Tables S1 and S2 ‡) was performed using unrestricted Kohn-Sham orbitals DFT (UKS) and the B3LYP functional [80][81][82] in combination with Ahlrichs' split-valence triple-z basis set def2-TZVPP for all atoms. 83,84 Tight convergence criteria were chosen for DFT-UKS calculations (keywords tightscf and tightopt). All DFT-UKS calculations make use of the resolution of identity RIJ (Split-RI-J) approach for the Coulomb term in combination with the chain-of-spheres approximation for the exchange term (COSX). 85,86 The zero order relativistic approximation was used to describe relativistic effects in all calculations (keyword ZORA). 87 Grimme's empirical dispersion correction D3(BJ) was employed (keyword D3BJ). 88,89 To account for solvent effects, a conductor-like screening model (keyword CPCM) modelling acetonitrile was used in all calculations. 90,91 TDDFT-UKS calculations were performed at the same level of theory using unrestricted Kohn-Sham orbitals (UKS). Fiy vertical spin-allowed transitions were calculated (Tables S3 and S4 ‡).
Harmonic frequency calculations for the IR assignments were performed using Turbomole 7.4 (ref. 92 and 93) on the optimized geometry (RIJCOSX-UB3LYP-D3BJ/def2-TZVP). The vibrational frequencies were scaled by a factor of 0.98, which is typical for the chosen method and basis set, to minimize the differences between the experimental and calculated frequencies. A Gaussian convolution with a full-width at half-maximum of 15 cm À1 was applied to the calculated vibrational transitions.
SOC-CASSCF(6,12)-FIC-NEVPT2: calculations of ground-and excited-state properties with respect to metal-centered (MC) states were performed using the complete-active-space selfconsistent eld method in conjunction with the fully internally contracted N-electron valence perturbation theory to second order based on a fully internally contracted (FIC) wave function (FIC-NEVPT2) 94,95 in order to recover missing dynamic electron correlation. In order to accurately model the ligand eld, active spaces were chosen to encompass the dominating bonding/antibonding orbitals formed between vanadium and the ligand. An active space of (6,12) along with 10 triplet roots and 9 singlet roots was selected (Tables S5 and S6 ‡). In addition to the minimal active space of (2,5) comprising the 3d orbitals, two occupied V-N s bonding orbitals and a second d shell 96 were included in these calculations.

Conclusions
The pseudo-octahedral vanadium(III) complex VCl 3 (ddpd) with the strong-eld ligand ddpd shows spin-ip phosphorescence at room temperature at 1102, 1219 and 1256 nm aer excitation into charge-transfer bands. Several factors are relevant for this emission from a 3d transition metal complex to occur: (i) The ligand eld splitting in VCl 3 (ddpd) is large enough to place the emissive singlet states 1 E/ 1 T 2 below the distorted metal-centred triplet excited states 3 T 2 .
(ii) Lower temperature disables thermally activated nonradiative pathways increasing the photoluminescence, yet even at room temperature a weak emission is still observed.
(iii) Deuteration of the ddpd ligand reduces non-radiative energy transfer to C-H overtones increasing the photoluminescence.
(iv) The radiative rate might be higher in noncentrosymmetric pseudo-octahedral vanadium complexes, although this effect of Laporte's parity rule was not experimentally conrmed in this particular case.
(v) The ISC rate constant from the triplet to the singlet manifold is high (k ISC > 6.7 Â 10 11 s À1 ) as conrmed by the molecular dynamics simulations in the accompanying paper. 63 This high rate could be an effect of the heavier chloride atoms (heavy atom effect), efficient vibronic coupling and/or enhanced SOC due to large differences in orbital type between the two states ( 3 LMCT / 1 MC), 97 although other ultrafast pathways might still compete with ISC.
Challenges with the emission from excited states of d 2 -VCl 3 (ddpd) arise from the large ground state splitting which spreads the emission bands over ca. 2400 cm À1 . This range is larger by almost two orders of magnitude than the corresponding spin-ip emission of d 3 -metal complexes with orbitally non-degenerate ground states, a very signicant difference for transitions involving essentially nested states. The considerable ground state splitting further reduces the already small energy gap between the emissive state and the ground state enabling a higher non-radiative decay according to the energy gap law. A second aspect of VCl 3 (ddpd) as phosphorescent emitter concerns the excited state reactivity of LMCT states with chloride / vanadium charge-transfer character in solution. As these states can lead to V-Cl homolysis in solution reducing the photoluminescence and nally decomposing the complex, solution photostability is a particularly important aspect for future applications of vanadium(III) complexes in solution.
This study emphasises that design strategies toward efficient d 2 -NIR emitters require a particular attention to the ISC efficiency from the triplets to the singlet states and potential dissociative unimolecular reactions at ultrafast timescales as well as on the radiative and non-radiative relaxation of the singlets at longer times. Details of the ultrafast excited state dynamics of VCl 3 (ddpd) up to 10 ps are discussed in the accompanying paper. 63

Data availability
Experimental and computational data are available as ESI.

Author contributions
MD prepared the complex, performed all ground state characterization and photolysis experiments and the quantum chemical calculations, JK and MS measured and analysed the luminescence and lifetime data of the neat complex, measured and analysed the NIR absorption data and performed the SOI calculations, CD and CR measured and analysed the Raman spectra and the spectra under pressure, PB, GNS and MG measured and analysed the step-scan FT-IR spectra and the temperature dependent luminescence spectra of the complex in KBr pellets, AK and SL measured and analysed the transient absorption spectra, MS and KH devised the concept. KH supervised the project and wrote the manuscript.

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