Intrinsic intermolecular photoinduced charge separation in organic radical semiconductors Biwen Li1, Petri Murto1,2,3, Rituparno Chowdhury1, Laura Brown2, Yutong Han4, Giacomo Londi5, David Beljonne6, Hugo Bronstein1,2* and Richard H Friend1* 1The Cavendish Laboratory, University of Cambridge, J. J. Thomson Avenue, Cambridge CB3 0HE, United Kingdom. 2Yusuf Hamied Department of Chemistry, University of Cambridge, Lensfield Road, Cambridge CB2 1EW, United Kingdom. 3Department of Chemistry and Materials Science, Aalto University, Kemistintie 1, 02150 Espoo, Finland 4Department of Chemical Engineering and Biotechnology, University of Cambridge, Cambridge CB3 0AS, United Kingdom 5Department of Chemistry and Industrial Chemistry, University of Pisa, 56124, Pisa 6Laboratory for Chemistry of Novel Materials, University of Mons, Place du Parc, 20, Mons 7000, Belgium *email: rhf10@cam.ac.uk, hab60@cam.ac.uk Abstract Organic radicals based on tris(2,4,6-trichlorophenyl)methyl radicals, TTMs, show efficient photoluminescence, PL, from excitons in the spin doublet manifold. We find that when TTMs are in contact, photoexcitation generates TTM anion – TTM cation pairs. These can decay radiatively or be fully separated under electric field bias. We use a triphenyl substituted TTM, P3TTM, where contacts via the phenyl end groups enhance intermolecular interactions. In dilute (5wt%) films in a wide energy gap organic semiconductor host we observe prompt PL from the excited radical at 645 nm, and a delayed component, beyond 1 µs, at 800 nm due to recombination of P3TTM anion - cation pairs. Measurements of photocurrent made with diode structures with 100% P3TTM showed close to unity charge collection efficiency in reverse bias. In summary, we have found ‘homo-junction’ intermolecular charge separation, made possible when the extra energy for double occupancy of the non-bonding radical level on the anion is lower than the energy of the doublet exciton. This opens new possibilities for light harvesting using molecular semiconductors. Interest in organic radical semiconductors is due to their potential applications in optoelectronic devices, biomedical technologies, quantum information systems, and chiral materials. 1-10 Among the organic radicals family, trityl systems are of interest due to their chemical stability and photoluminescence quantum efficiency (PLQE). 11,12 These support efficient organic light-emitting diodes, OLEDs, but are generally used as a dilute guest component in a molecular semiconductor ‘host’, since PL is generally quenched at high concentrations when inter-radical interactions are possible. TTM by itself is an alternant hydrocarbon, so the energy degeneracy between highest occupied molecular orbital (HOMO) to singly occupied molecular orbital (SOMO) and SOMO to lowest unoccupied molecular orbital (LUMO) transitions cause weak absorption and slow PL emission rates. Higher efficiencies are achieved with an attached electron donor, including carbazole and triphenylamine, 13-15 where the lowest transition is of intramolecular CT from donor to TTM. PLQEs reported for these systems which emit in the red and near-IR are often very high, particularly for red and near-IR emission, and it has recently been shown that multiphonon decay is suppressed for these structures due to reduced vibrational coupling to the exciton. 16 We have also reported a wide range of TTMs with substitution of the para-chlorine with phenyl-based-substituents to adjust the colour and efficiency of emission by means of steric bulk. 17-19 This provides a new platform for red and near-IR OLEDs. We consider in this paper the effects of intermolecular interactions between radical- radical (or radical-host). Inter-radical interactions have been reported using co-crystals of radicals and their hydrogenated radical precursors, allowing good control of inter- radical contacts. A strongly red-shifted PL band at room temperature is observed above 5% radical loading, and this shows strong magnetoluminescence below 20 K. 20 This is attributed to an excimer, but as noted in a recent review, the propeller shape of the TTM radical should not allow the strong π-π interactions generally required for strong excimer formation with red-shifted PL, and that further work is needed to explain this red-shifted emission. 21 We present evidence here that red-shifted PL may arise from a fully charge-separated anion-cation pair, rather than from a conventional excimer. The materials explored here are shown in Figure 1. For a regular closed shell semiconductor, electron-hole generation requires photoexcitation across the semiconductor band gap. In contrast, as we develop here, for radical semiconductors, electron-hole generation between neighbouring P3TTMs can occur just within the SOMO non-bonding orbitals, as illustrated in Figure 2(a), leading to spinless anions and cations. In the absence of on-site Coulomb repulsion energies, this transfer would be barrierless. However, there is an energy cost to do this; it is the charging energy to doubly occupy the anion SOMO, this onsite Coulomb energy is the Hubbard U and this is easily measured from the voltage difference for electrochemical oxidation, where the electron is removed from the SOMO to form a cation, versus reduction where a second electron is added to the SOMO to form an anion. It is large because the SOMO non- bonding orbitals are relatively localised. The reduction and oxidation potentials of P3TTM are - 1.05 and + 0.67 V, respectively, so the energy gap is 1.72 eV corresponding to the 720 nm PL which is close to the broad red-shifted emission band. 18 In this study we use the P3TTM shown in Figure 1(a) where the phenyl groups are in good conjugation with the TTM core (an average phenyl-phenyl dihedral of 34.8° is obtained from X-ray crystal structure), 18 giving PL with peak emission at 645 nm, and can provide effective inter-radical contact as illustrated in SI section 8. We used a series of hole transport materials widely used in OLEDs as host matrixes of P3TTM. These allow tuning of the host-dopant energy alignment (Figure 1(c)) and allow study of radical-radical and radical-host interactions. We also use solutions in toluene where different concentrations can be used to control radical-radical interactions. Figure 1. The chemical structures of investigated molecules and relative molecular orbital energy between radical dopants and host materials. The chemical structures of a, TTM based radical emitters and b, host materials investigated in this study. c, The corresponding molecular orbital energy alignment of P3TTM and host materials. The host materials only show the HOMO energy. (Energies are indicative and estimated from reported values18,22-25) Radical doped films were prepared via vacuum sublimation, as used for radical OLEDs. 5,13,14,26,27 In the doping range 3wt% to 8wt% differences in observed PL are small and most results presented here are for 5wt%. We denote films as P3TTM:host for range of hosts used, as shown in Figure 1(b), along with an overview of the relative molecular orbital energy alignment of dopant and host materials, based on previously reported values. 18,22-25 We note that these hosts allow tuning of the HOMO energy gap between dopant and hosts. c Figure 2. Schematic illustrations of CT and P3TTM anions and cations recombination process in TSPO1 and CBP. Note that a-d show occupancies of the single-particle states, but do not reflect the total ‘exciton’ energy changes that occur when occupancies of states are changed. a, The intermolecular CT directly takes place between the ground sate radical (R) and the excited state radical (R*) to generate cations (R+) and anions (R-). b, Photogenerated hole is quickly transferred to the CBP to form R- first, followed by the second CT process between R and CBP+ to generate R+. c, The red-shifted emission band is attributed to the electron-hole recombination from the ion pair. d, The magnetic field modulates the population of singlet molecular pairs which is allowed for intermolecular CT, while triplet pairs are not able to form R+ and R-. e, TDDFT electron-hole wave function for a singlet ion pair in the P3TTM a. b. d. c. e. single crystal. Note how the wave functions with major weight on the radical center extend over the conjugated phenyl groups. Time resolved PL with 400 nm excitation for P3TTM in dilute and concentrated toluene solutions and in TSPO1 films are shown Figure 3 and S3. In dilute solution (0.1 mM), the PL peak at 645 nm shows no spectral evolution and a mono-exponential lifetime of 9.1 ns. In contrast, the concentrated 10 mM toluene solution shows a red-shifted emission band that becomes dominant at late time (> ca. 40 ns) in Figure 3(a). At this concentration the rate of P3TTM- P3TTM collisions is high enough to ensure collision within the 9.1 ns exciton lifetime. 28 The associated reduction of the 645 nm PL lifetime with increased concentration follows Stern-Volmer kinetics indicating that this is a bimolecular process (SI section 9). Figure 3(c) shows PL results for P3TTM:TSPO1. TSPO1 with its wide bandgap provides an inert environment for the radical dopants so there is no host-dopant interface effect. At early times only the molecular emission peak at 645 nm from the molecular P3TTM exciton is observed. An additional broad emission band beyond 750 nm appears at later times. Figure 3(d) shows the integrated PL fraction of P3TTM:TSPO1. We see the PL has two main contributions, at 550-750 nm, PL is mainly from the molecular exciton with a lifetime of 11.5 ns, similar to the dilute solution. However, the red-shifted emission band at 750-840 nm is long-lived, around 50% of integrated PL is detected after 500 ns. In contrast to the 645 nm emission band, which changes little (<2%) the red-shifted emission band shows a strong magnetic field effect (MFE), shown in Figure 3(b). The PL is suppressed at 0.7 T at room temperature and the spectrum for the change of PL intensity, ∆PL = PL (0 T) - PL (0.7 T), shows a broad band extending from 700 nm to a peak at 800 nm. We note that this MFE has been reported for ‘excimer-like’ emissions of other TTM or (3,5-dichloro-4- pyridyl)bis(2,4,6-trichlorophenyl)methyl (PyBTM) based mono- or di- radicals, and low temperature (< 20K) magneto-photoluminescence is reported on PyBTM cocrystals. 20,29 Figure 3. Transient photoluminescence and magneto-photoluminescence of P3TTM in solution and TSPO1. The time resolved PL spectra of a, concentrated P3TTM toluene solution (10 mM) and c, P3TTM:TSPO1 (5wt%). b, PL spectra of P3TTM:TSPO1 (5wt%) under 0 T and 0.7 T field at room temperature and PL change under magnetic field with respect to the emission wavelength, normalised to the peak emission at 645 nm. d, The time evolution of integrated PL fraction (obtained from the time of the PL spectra) of P3TTM:TSPO1 (5wt%) for molecular emission and red- shifted emission band. We instead associate the red-shifted emission with an inter-P3TTM CT recombination (Figure 2(c)). First evidence for this comes from quantum-chemical calculations performed on close molecular pairs of the P3TTM single crystal, see SI section 7 for details. Time-dependent Density Functional Theory (TDDFT) calculations indeed show the presence of intermolecular singlet CT states lying ~0.4 eV below the localized a b c d P3TTM excitons, in excellent agreement with the energy difference (~0.3-0.5 eV) inferred from the two main measured PL features in Figure 3. As we hypothesized, there is significant extension of the SOMOs onto the peripheral phenyl rings (Figure 2(e)) that provides the needed electronic couplings for the generation of CT pairs from the optically excited excitons and the (non)radiative recombination of these CTs to the ground state. Applying Marcus-like rate theory to the results of TDDFT calculations performed for close pairs in the crystal yields charge generation rates up to 1 ps-1 and CT lifetimes approaching the µs range, see SI section 7. Thus, our calculations suggest fast formation of long-lived CT pairs that, in absence of competing decay mechanisms, should enable efficient free charge generation, as confirmed below. Transient optical absorption, TA, measurements are shown in Figure 4. For dilute toluene solution in the visible probe region (Figure 4(a)), a broad photoinduced absorption, PIA, signal peaked at 675 nm and a sharp PIA in the region of 480-550 nm corresponding to the radical D1 state are probed. 5,12 There is no spectral shape change observed in the late time, matching the result of time resolved PL. However, for P3TTM:TSPO1 as shown in Figure 4(b), two distinct PIA peaks (570-630 nm and 730- 850 nm) grow, while the PIA of radical D1 state reduces with the time. We have carried out steady-state spectroelectrochemistry measurements, shown in Figure 4(c). We observe absorptions of anions at 580 nm and cations at 760 nm and a bleaching of the ground state absorption at 407 nm for D0-D2 transition. 13 These features from spectroelectrochemistry map well with the TA features and we associate these PIA states to the closed shell anions (570-630 nm) and cations (730-850 nm). The concentrated P3TTM solution (10 mM) shows similar long time TA, in Figure S9 and S10, though the CT kinetics are slower in the solution since molecular diffusion and collision need to be considered. 28 We note that CT can only proceed for overall spin singlet P3TTM*-P3TTM pairs (Figure 2(d)). This is consistent with the strong magneto- photoluminescence shown in Figure 3(b) and we consider the applied magnetic field modulates the population of triplet and the singlet sublevels. The reduction in singlet population with applied field is consistent with level crossing at the exchange energy, noting that we expect an antiferromagnetic ground state. Larger values of magnetoluminescence are reported at low temperatures and higher magnetic fields. 20 We emphasise that the CT between P3TTM radical pairs is very different to the standard electron-hole transfer in closed shell semiconductors. As set out in Figure 2(a), the electron and hole of are both in the SOMO levels of two radicals, rather than HOMO and LUMO with an energy gap. We note that this inter-molecular CT exciton can be seen as an inter-molecular analogue of zwitterionic excited state of di- or poly- radicals.30,31 Figure 4. Transient absorption measurements and spectroelectrochemistry of P3TTM in solution and TSPO1. The ps visible TA spectrum of a, diluted P3TTM toluene solution (0.1 mM) and b, P3TTM:TSPO1 (5wt%). (𝜆!" = 400 nm, 17-27 µ𝐽𝑐𝑚#$ per pulse) c, The spectroelectrochemistry of P3TTM in degassed tetrahydrofuran solution. We have previously reported host-dopant intermolecular CT state between mesitylated TTM, M3TTM, and a donor CBP32 host which gives long-lived emission, in which the TTM acts as an electron acceptor. 17 Figure 5 shows the effects of CBP host matrixes on intermolecular CT. The PL of P3TTM:CBP system is shown in Figure 5(a), and shows similar behaviour to the P3TTM:TSPO1 system, with a magnetic field dependent red-shifted band centered near 790 nm. This is also time delayed as shown in Figure S5. This indicates broadly similar behaviour as for the P3TTM:TSPO1 system, but as we a b c explore below, there is an intermediate time regime where CBP has transferred an electron to the P3TTM. This can then be followed by either back transfer from the P3TTM anion (causing a longer molecular emission lifetime of 13.7 ns, see Figure S8) and later electron transfer to CBP from a neutral P3TTM. The visible TA spectrum in Figure 5(b) of CBP film has two similar broad PIA peaks in the same region of 570-630 nm and 730-850 nm as TSPO1 film, as the same radical anions and cations are generated, but the 730-850 nm band (associated with the P3TTM cation) grows in more slowly over the 100 ps. Given the 5% P3TTM fraction, the CT process most accessible will be from CBP to P3TTM, and we consider the slower build up of the P3TTM cation population is due to the subsequent CT to bring CBP back to zero charge. Access to the CBP-P3TTM CT process allows faster initial growth of the P3TTM anion population compared to the P3TTM:TSPO1 system. We also note the first CT process between CBP and excited P3TTM, P3TTM*, is not spin selective, the spinless CBP-P3TTM* pairs give quicker CT kinetics, but the following CT between positively charged CBP, CBP+, and P3TTM is dependent on the spin state of the CBP+- P3TTM pairs. Therefore, the PL of P3TTM:CBP also has a MFE for the red-shifted emission band shown in Figure 5(a). In both TSPO1 and CBP, the growth rate of anion and cation PIA increases for a higher doping concentration or closer contact between radicals. The CT population build-up times are shortened to several ps in the 100wt% doped film (without hosts) due to a short inter-radical distance (Figure S30 and S31). Figure 5. Magneto-photoluminescence and transient spectroscopies of P3TTM in CBP. a, PL spectra of P3TTM:CBP (5wt%) under 0 T and 0.7 T field at room temperature and PL change under magnetic field with respect to the emission wavelength, normalised to emission at 645 nm. b, The visible TA spectrum of P3TTM:CBP. The ultra-violet TA spectra of c, diluted P3TTM toluene solution (0.1 mM) and d, P3TTM:CBP (5wt%). The spectrum break is due to the pump laser scattering. (𝜆!"= 400 nm, 12-13 µ𝐽𝑐𝑚#$ per pulse) TA of the ground state bleach, GSB, for P3TTM (360-430 nm) is shown for dilute solutions in Figure 5(c) and for films of P3TTM:CBP in Figure 5(d). For the dilute solution the GSB is largest at early times and its later decay is consistent with the 9.1 ns PL decay. In contrast, the P3TTM:CBP film shows GSB growth up to 2 ns. The early time GSB is due to P3TTM exciton evolving to P3TTM anions, accompanied by CBP cations. The GSB growth is due to conversion of CBP cations to P3TTM cations as the a c d b full P3TTM anion-cation population develops. In the first 2 ns, whereas the GSB for the dilute solution falls by 27%, for the P3TTM:CBP film the GSB rises by 32%. This indicates that the quantum yield for charge photogeneration is remarkably high, which we estimate to be up to 40%, see SI Section S5. Figure 6. The photocurrent measurements of P3TTM and rubrene based devices. a, The device architecture and schematic illustration of charge separation process in P3TTM device (left) and rubrene device (right). b, Photocurrent density under 395 nm excitation at 160 mW/cm2 and dark current density (J) comparison of P3TTM device (left) and rubrene device (right). Our studies of P3TTM in solution and in solid hosts indicate clear evidence for a fully charge-separated anion-cation intermolecular state, which shows PL near 800 nm. Excitation fluence measurements, Figure S7, indicate that this anion-cation pair is mostly bound at room temperature. We have investigated solid films of P3TTM without a host material in standard diode device structures and do find that it is possible to fully a b separate electrons and holes with an applied field. The films show significantly reduced PL, see Table S4, and as tracked through significantly reduced photoexcited state lifetimes, see Figures S30 and S31. The TA spectra shown in Figure S32 show clear evidence for early time charge separation, with anion (600 nm) and cation (800 nm) bands fully formed by 4 ps, but rapid decay to the ground state within a few ns. In spite of these reduced lifetimes, we are still able to get long range charge separation in biased diode structures. We fabricated standard multilayer diode structures using PEDOT:PSS on ITO as hole injection/extraction electrode, and C60/BCP/Al (C60: fullerene, BCP: bathocuproine, Al: aluminium) as electron injection/extraction layers. These were selected to give band alignment with the P3TTM radical. We also made devices using rubrene as a control, since this has similar redox potentials. The device architecture used is shown in Figure 6(a), with ITO (150 nm)/PEDOT:PSS (40 nm)/photoactive layer (P3TTM or rubrene) (80 nm)/C60 (20 nm)/BCP (5 nm)/Al (100 nm). Figure 6(b) shows the photocurrent and dark current under the bias for both the P3TTM and rubrene devices (both made with and without the C60 layer. Photoexcitation at 395 nm was at 160 mW/cm2). The rubrene device shows expected behaviour for charge photogeneration at the rubrene/C60 heterojunction giving a short-circuit current density of 0.25 mA/cm2 giving a quantum yield of 4.9%, considering 90% transmission rate of the rubrene thin film under ultra-violet light. This is consistent with charge photogeneration within an exciton diffusion range of 6-8 nm in the rubrene. 33 In reverse bias these devices showed little further current increase, rising to 4 mA/cm2 at -20 V. For diodes made without the C60 layer the photocurrents are much lower, similar in magnitude to the dark current, showing that the rubrene/BCP heterojunction does not generate free charge carriers. In contrast, the P3TTM diode showed a low quantum yield at short circuit, but a very strong increase in reverse bias that saturates at around 45 mA/cm2 indicating a quantum yield for charge collection close to 100%, considering 15% transmission rate of the P3TTM thin film under ultra-violet light. Behaviour was very similar for devices made with and without the C60 layer, indicating that charge photogeneration was not controlled by the P3TTM/C60 interface. Rather, this close to unity charge collection efficiency is due to bulk charge photogeneration, as we have described above. The ability to fully separate electron and hole following photoexcitation in a diode structure made with a single non-polar semiconductor is routine in mainstream inorganic semiconductors such as silicon where the photogenerated exciton is large and has low binding energy (<10 meV), but for molecular systems this has rarely been previously reported. We note that there is an interesting literature on ‘symmetry- breaking’ charge transfer at an intramolecular level, often in the presence of a polar solvent.34-39 We also note that our demonstrated high-efficiency of charge- photogeneration precludes any ‘excimer’ formation which typically do not undergo charge-separation. In summary, the intermolecular interaction and excited state dynamics of neutral 𝜋 radicals are studied by various time-resolved techniques and steady-state spectroscopies. In both solutions and films, we identified an intermolecular symmetry- breaking charge separation process between P3TTM SOMOs to form an excited ion- pair state comprising a pair of closed-shell anion and cation. The GSB growth of P3TTM provides further evidence for the intermolecular process. We also demonstrated the red-shifted PL has a MFE, as the intermolecular CT depends on the spin state of P3TTM*-P3TTM and CBP+-P3TTM intermediates. This mechanism can be generalised to other open-shell materials, for example the long-lived and red-shifted emission generally observed for TTM- and PyBTM- based radical can be attributed to the recombination of this CT exciton. We also explored the effect of host-dopant interface on the intermolecular CT channels. This work demonstrates that photoinduced charge separation can be driven intermolecularly in neutral radicals and that the unpaired electron can be mobile and not bound to the radical centre. Homojunction charge separation as we observe here, has been a long sought after goal in organic photovoltaics. 40 This work provides a new avenue for the exploration of power generation and solar-driven chemistry in both solution and solid-state using only a single component. Experimental methods Film Preparation. Radicals were synthesised as previously reported. 18 Host materials were obtained from Ossila without further purification. Thin films were prepared by thermal evaporation under vacuum (ca. 10-7 torr, Angstrom Engineering EvoVac 700 system). 100 nm of 5wt% radical doped CBP, TSPO1 and TAPC films were deposited on UV fused silica substrates. The PMMA films were prepared by spin-coating in a nitrogen fill glove box. All films were encapsulated in the nitrogen filled glovebox. The doping concentration stated in this study denotes weight percentage. Device fabrication and characterization. Organic transport layer materials were obtained from Ossila without further purification. The indium tin oxide (ITO) substate was cleaned in an ultrasonic bath of detergent, deionized water, acetone, and isopropanol for 10 mins each. It was then UV-treated in a UV-ozone chamber for 12 mins. A thin layer of PEDOT:PSS (poly(3,4-ethylene dioxythiophene):poly(styrene sulfonate)) (Clevios™ P VP AI 4083) was prepared by spin-coating the PEDOT:PSS solution at 3,000 rpm for 40 s on the ITO substrate and annealed at 150 °C for 15 mins in the air. The following layers, P3TTM, C60, BCP and Al of devices were deposited by thermal evaporation under vacuum (ca. 10-7 torr, Angstrom Engineering EvoVac 700 system) at a rate of 0.1-1 Å/s–1. The whole device has a structure of ITO (150 nm)/PEDOT:PSS (40 nm)/P3TTM (80 nm)/C60 (20 nm)/BCP (5 nm)/Al (100 nm). A Keithley 2635 A source with a 395 nm light source was used to measure the current- voltage plot. The voltage was applied to the device between −20 to 3 V at a sweep speed of 0.5 V/s. The effective electrode overlap area was 4.5 mm2 which was used to calculate the current density. Steady state photophysics measurements. The absorption spectra were measured by a commercially available Shimadzu UV-2550 spectrophotometer and a Shimadzu UV- 1800 spectrophotometer. Photoluminescence was measured by a home-build setup by providing a continuous photo-excitation at 405 nm from a laser diode. The photoluminescence from samples was collected in a collimating 2-lens apparatus and directed into an optical fiber which supplies the photons into a calibrated grating- spectrometer (Andor SR-303i) and finally into a Si-camera where it is recorded. Transient photoluminescence spectroscopy. Transient photoluminescence (time resolved photoluminescence) spectra at nanosecond-microsecond timescales were recorded using an electronically gated intensified charge-coupled device (ICCD) camera (Andor iStar DH740 CCI-010) connected to a calibrated grating spectrometer (Andor SR303i). A narrowband non-colinear optical parametric amplifier pumped with a frequency doubled output of a 1 kHz 800 nm laser pulse (100 fs duration) from a Ti:sapphire amplifier (Spectra Physics Solstice Ace) was used to generate a tuneable excitation pulse. A 400 nm excitation can be achieved by the second harmonic of 800 nm output, generated using a β-barium borate crystal. A 425 nm long-pass filters (Edmund Optics) were used to prevent scattered laser signals from entering the spectrometer. Temporal evolution of the emission was obtained by stepping the ICCD delay with respect to the excitation pulse, with a minimum gate width of 5 ns. Recorded data was corrected to account for filter transmission and camera sensitivity. Transient absorption spectroscopy. The ps TA spectroscopies were carried out either on a home-built up TA set up or a commercialised TA setup. The home-built TA set up has an output from Ti:sapphire amplifier (Spectra Physics Solstice Ace) that generated 100-fs-duration pulses centred at 800 nm with a 1 kHz repetition rate. The pulse was split to the pump and probe beams. The 400 nm excitation pump was generated by passing the second harmonic of 800 nm output through using a β-barium borate crystal. The pump light was chopped at 500 Hz by a chopper wheel. The visible probe light was generated via non-collinear optical parametric amplifiers. The pump-probe time delay was provided by a mechanical delay stage (Thorlabs DDS300-E/M). The probe pulses were split into two beams by a 50/50 beam splitter to provide a reference beam which increases the signal:noise ratio. The probe pulses were detected by silicon (Hamamatsu S8381-1024Q) dual-line array with a custom-built board from Stresing Entwicklungsbüro. Fundamental laser beam of commercialised ps TA set up at 1030 nm was provided by Pharos. The Orpheus system which was pumped by ytterbium doped solid state based chirped pulse amplifier generated pump lights using 90% of the fundamental laser (Light Conversion). The wavelength of pump can be varied from 350 nm to 2000 nm with a 10 kHz repletion rate and a ~ 100 fs pulse duration. The rest of 10% of fundamental beam was injected into Harpia-TA to generate probe light. The pump-probe time delay was provided by a mechanical delay stage. The probe pulses were detected by a spectrograph with dual outputs to cover UV-NIR regions (Kymera- 193i-B2). Magneto-photoluminescence measurement. The encapsulated film sample was positioned between magnet cores (GMW 3470 electromagnet). Its PL spectrum was recorded with an Andor spectrometer (Andor SR-303i) with and without a magnetic field. The magnetic field was swept from 0 to 0.7 T with a ramping step of 0.01 T at room temperature. The continuous photo-excitation at 405 nm was provided by a laser diode. Spectroelectrochemistry. A commercially available PalmSens EmStat4S potentiostat was connected to a commercially available Shimadzu UV-1800 spectrophotometer. The measurements were carried out in a home-built three-electrode setup using a quartz cuvette (0.1 mm pathlength) as the spectroelectrochemical cell, a coil of platinum wire as the working electrode (in the light path), a platinum wire as the counter electrode and a freshly activated silver wire as the Ag/Ag+ reference electrode. The silver wire was activated by immersing in concentrated HCl solution to remove any silver oxides or other impurities, then rinsed with water and acetone and dried prior to measurements. The reference electrode was calibrated against ferrocene/ferrocenium (Fc/Fc+) redox couple. For this setup, Fc/Fc+ half-wave potential, E1/2, was determined at 0.40 V vs. Ag/Ag+. The supporting electrolyte was 0.1 M solution of Bu4NPF6 in anhydrous THF. The electrolyte was bubbled with Ar gas before measurements to remove any dissolved oxygen and the measurements were carried out under an Ar atmosphere. The sample concentration was adjusted to keep the optical density below 1.0 a.u. for neutral radical. Absorption spectra of the oxidized (cationic) and reduced (anionic) species were measured by applying a constant potential of +1.1 V and –1.4 V vs. Fc/Fc+, respectively. These are ca. 0.4 V higher than the corresponding half-wave potentials determined for P3TTM radical. 18 Quantum-chemical calculations. Molecular pairs in close contact selected from the P3TTM’s X-ray crystal structure were used for the calculation of excited-state properties (vertical excitations, transition dipole moments, and state dipoles) by means of TDDFT within the Tamm-Dancoff approximation (TDA), in conjunction with the LC-ωhPBE functional and the 6-311G(d,p) basis set. 41 Furthermore, to implicitly take into account dielectric screening effects, the screened range-separated hybrid procedure (SRSH) was applied. 42 In such scheme, the range-separation parameter ω was set to 0.100 Bohr-1, while the other parameters (𝛼 and 𝛽) were modified according to the dielectric constant of solvent/environment. In this work we chose a dielectric constant typical of toluene, ε = 2.37. Regarding the electronic couplings calculations and the (non)radiative charge dissociation/recombination rates, we refer to SI for further details. All the calculations were performed within the Gaussian16 suite package. Acknowledgements We thank Dr Andrew Bond for carrying out the X-ray crystallography measurements and data analysis at the Yusuf Hamied Department of Chemistry, University of Cambridge. P.M. and R.C. have received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreements No. 891167 and No. 859752. R.H.F., B.L., R.C. and P.M. acknowledge funding from the European Research Council under the European Union’s Horizon 2020 research and innovation programme grant agreement No. 101020167. G.L. acknowledges the Italian Ministry of University and Research for funding provided by the European Union-NextGenerationEU-PNRR, Missione 4, Componente 2, Linea di investimento 1.2. Computational resources in Mons were provided by the FNRS “Consortium des Equipements de Calcul Intensif-CECI” program Grant No. 2.5020.11. D.B. is Research Director of the Belgian National Fund for Scientific Research (FRS- FNRS). Author contributions B.L. performed the transient spectroscopy measurements and fabricated and measured the photodiodes used in this work. P.M. and L.B. synthesized the materials. P.M. carried out the spectroelectrochemistry measurements. B.L. and R.C. measured magneto- photoluminescence. B.L. and Y.H. performed the photocurrent measurements. G.L. and D.B. performed the quantum-chemical calculations. R.H.F. and H.B. supervised the work. B.L. and R.H.F. wrote the manuscript with input from all authors. Competing interests The authors declare no competing interests. Additional information Supplementary information is available at https://doi.org/ Correspondence and requests for materials should be addressed to R.H.F. or H.B. Reprints and permissions information is available at www.nature.com/reprints. Reference 1 Peng, Q., Obolda, A., Zhang, M. & Li, F. Organic light‐emitting diodes using a neutral π radical as emitter: the emission from a doublet. Angew. Chem. 127, 7197-7201 (2015). 2 Ji, L., Shi, J., Wei, J., Yu, T. & Huang, W. Air-Stable Organic Radicals: New-Generation Materials for Flexible Electronics? Adv. Mater. 32, 1908015 (2020). 3 Cui, X. et al. Stable π-radical nanoparticles as versatile photosensitizers for effective hypoxia-overcoming photodynamic therapy. Mater. Horiz. 8, 571-576 (2021). 4 Gorgon, S. et al. Reversible spin-optical interface in luminescent organic radicals. Nature 620, 538-544 (2023). 5 Ai, X. et al. Efficient radical-based light-emitting diodes with doublet emission. Nature 563, 536-540 (2018). 6 Mayorga Burrezo, P. et al. Organic free radicals as circularly polarized luminescence emitters. Angew. Chem. Int. Ed. 58, 16282-16288 (2019). 7 Mas-Torrent, M. et al. Organic radicals on surfaces: towards molecular spintronics. J. Mater. Chem. 19, 1691-1695 (2009). 8 Wu, S. et al. Toward two-dimensional p-conjugated covalent organic radical frameworks. Angew. Chem. Int. Ed. 57, 8007 –8011 (2018). 9 Gu, Q., Abdurahman, A., Friend, R. H. & Li, F. Polymer light emitting diodes with doublet emission. J. Phys. Chem. Lett. 11, 5638-5642 (2020). 10 Teki, Y. Excited‐State Dynamics of Non‐Luminescent and Luminescent π‐Radicals. Chem. Eur. J. 26, 980-996 (2020). 11 Murto, P. & Bronstein, H. Electro-optical π-radicals: design advances, applications and future perspectives. J. Mater. Chem. C 10, 7368-7403 (2022). 12 Guo, H. et al. High stability and luminescence efficiency in donor–acceptor neutral radicals not following the Aufbau principle. Nat. Mater. 18, 977-984 (2019). 13 Abdurahman, A. et al. Understanding the luminescent nature of organic radicals for efficient doublet emitters and pure-red light-emitting diodes. Nat. Mater. 19, 1224-1229 (2020). 14 Cho, H.-H. et al. Efficient near-infrared organic light-emitting diodes with emission from spin doublet excitons. Nat. Photon. 18, 905-912 (2024). 15 Lu, C. et al. Achieving Nearly 100% Photoluminescence Quantum Efficiency in Organic Radical Emitters by Fine‐Tuning the Effective Donor‐Acceptor Distance. Adv. Funct. Mater. 34, 2314811 (2024) 16 Ghosh, P. et al. Decoupling excitons from high-frequency vibrations in organic molecules. Nature 629, 355-362 (2024). 17 Murto, P. et al. Mesitylated trityl radicals, a platform for doublet emission: symmetry breaking, charge-transfer states and conjugated polymers. Nat. Commun. 14, 4147 (2023). 18 Murto, P. et al. Steric Control of Luminescence in Phenyl-Substituted Trityl Radicals. J. Am. Chem. Soc. 146, 13133-13141 (2024). 19 Ohisa, S. & Honda, S. Luminescence enhancement by symmetry-breaking in the excited state in radical organic light-emitting diodes. Commun. Chem. 6, 238 (2023). 20 Kimura, S. et al. Magnetoluminescence in a photostable, brightly luminescent organic radical in a rigid environment. Angew. Chem. 130, 12893-12897 (2018). 21 Mizuno, A., Matsuoka, R., Mibu, T. & Kusamoto, T. Luminescent radicals. Chem. Rev. 124, 1034-1121 (2024). 22 Yadav, R. A. K., Dubey, D. K., Chen, S.-Z., Liang, T.-W. & Jou, J.-H. Role of molecular orbital energy levels in OLED performance. Sci. Rep. 10, 9915 (2020). 23 Vasilopoulou, M. et al. High efficiency blue organic light-emitting diodes with below- bandgap electroluminescence. Nat. Commun 12, 4868 (2021). 24 Kim, B. S. & Lee, J. Y. Engineering of mixed host for high external quantum efficiency above 25% in green thermally activated delayed fluorescence device. Adv. Funct. Mater. 24, 3970-3977 (2014). 25 Cho, H. H. et al. Efficient and bright organic radical light‐emitting diodes with low efficiency roll‐off. Adv. Mater. 35, 2303666 (2023). 26 Gu, Q. et al. Fast transfer of triplet to doublet excitons from organometallic host to organic radical semiconductors. Adv. Mater. 36, 2402790 (2024). 27 Cho, H. H. et al. Near‐Infrared Light‐Emitting Diodes from Organic Radicals with Charge Control. Adv. Opt. Mater. 10, 2200628 (2022). 28 Walker, B. J., Musser, A. J., Beljonne, D. & Friend, R. H. Singlet exciton fission in solution. Nat. Chem. 5, 1019-1024 (2013). 29 Kimura, S. et al. A ground-state-dominated magnetic field effect on the luminescence of stable organic radicals. Chem. Sci. 12, 2025-2029 (2021). 30 Yu, C. P. et al. Near-infrared luminescent open-shell π-conjugated systems with a bright lowest-energy zwitterionic singlet excited state. Sci. Adv. 10, eado3476 (2024). 31 Zhou, Z. et al. Sulfone-Functionalized Chichibabin’s Hydrocarbons: Stable Diradicaloids with Symmetry Breaking Charge Transfer Contributing to NIR Emission beyond 900 nm. J. Am. Chem. Soc. 146, 6763-6772 (2024). 32 Guo, J. et al. Novel triazine derivatives with deep LUMO energy levels as the electron- accepting components of exciplexes. J. Mater. Chem. C 9, 939-946 (2021). 33 Halls, J. J., Pichler, K., Friend, R. H., Moratti, S. & Holmes, A. Exciton diffusion and dissociation in a poly (p‐phenylenevinylene)/C60 heterojunction photovoltaic cell. Appl. Phys. Lett. 68, 3120-3122 (1996). 34 Kaul, N. & Lomoth, R. The carbene cannibal: Photoinduced symmetry-breaking charge separation in an fe (iii) n-heterocyclic carbene. J. Am. Chem. Soc. 143, 10816-10821 (2021). 35 Ramirez, C. E. et al. Symmetry-breaking charge separation in the solid state: Tetra (phenoxy) perylenediimide polycrystalline films. J. Am. Chem. Soc. 142, 18243-18250 (2020). 36 Wega, J. & Vauthey, E. Bimolecular photoinduced symmetry-breaking charge separation of perylene in solution. Photochemical & Photobiological Sciences 23, 93-105 (2024). 37 Markovic, V., Villamaina, D., Barabanov, I., Lawson Daku, L. M. & Vauthey, E. Photoinduced symmetry‐breaking charge separation: The direction of the charge transfer. Angew. Chem. Int. Ed. 50, 7596-7598 (2011). 38 Whited, M. T. et al. Symmetry-breaking intramolecular charge transfer in the excited state of meso-linked BODIPY dyads. Chem. Commun. 48, 284-286 (2012). 49 Trinh, C. et al. Symmetry-breaking charge transfer of visible light absorbing systems: zinc dipyrrins. J. Phys. Chem. C. 118, 21834-21845 (2014). 40 Dong, Y. et al. Orientation dependent molecular electrostatics drives efficient charge generation in homojunction organic solar cells. Nat. Commun. 11, 4617 (2020). 41 Hirata, S. & Head-Gordon, M. Time-dependent density functional theory within the Tamm– Dancoff approximation. Chem. Phys. Lett. 314, 291-299 (1999). 42 Refaely-Abramson, S. et al. Gap renormalization of molecular crystals from density- functional theory. Phys. Rev. B Condens. Matter. 88, 081204 (2013).