Simultaneous enhancement of thermally activated delayed fluorescence and photoluminescence quantum yield via homoconjugation

A fundamental problem facing thermally activated delayed fluorescence (TADF) is to overcome the paradox of efficient electronic transitions and a narrow singlet-triplet energy gap (ΔEST) in a single luminophore. We present a quinoxaline-based TADF iptycene as the first clear example that homoconjugation can be harnessed as a viable design strategy toward this objective. Homoconjugation was introduced in an established TADF luminophore by trimerization through an iptycene core. This homoconjugation was confirmed by electrochemistry. As a direct consequence of homoconjugation we observed synergistic improvements to photoluminescence quantum yield (ΦPL), radiative rate of singlet decay (krS), delayed fluorescence lifetime (τTADF), and rate of reverse intersystem crossing (krISC), while narrowing the ΔEST. The cooperative enhancement is rationalised with TD-DFT calculations including spin-orbit coupling (SOC). A facile synthesis of this system, and the ubiquity of the pyrazine motif in state-of-the-art TADF materials across the electromagnetic spectrum, leads to a great potential for generality.


Experimental Section
General All reactants and reagents were purchased from commercial suppliers and used without further purification unless otherwise stated. Column chromatography was carried out using silica gel 60, 40−60 μm mesh (Fluorochem) and Aluminium oxide 90 active neutral 0.063−0.200 mm (70−230 mesh ASTM, Merck). Analytical thin-layer chromatography was performed on precoated aluminum silica gel 60 F254 plates (Merck) and Macherey-Nagel ALUGRAM Alox N/UV254 Aluminum Sheets (Fisher), which were approximately 2 cm × 6 cm in size and visualized using ultraviolet light (254/365 nm).
NMR spectra were recorded on Jeol ECS 400 MHz and Jeol ECZ 500 MHz spectrometers. Chemical shifts are reported in ppm downfield of tetramethylsilane (TMS) using TMS or the residual solvent as an internal reference. NMR spectra were processed using MestReNova.
Multiplicities are reported as singlet (s), doublet (d), triplet (t), and multiplet (m). Melting points were determined in open-ended capillaries using a Stuart Scientific SMP10 melting point apparatus at a ramping rate of 1 °C/min. They are recorded to the nearest 1 °C and are uncorrected. IR spectra were collected on a Thermo Scientific Nicolet FTIR spectrometer.

Electrochemistry
Cyclic voltammetry was recorded using a Princeton Applied Research VersaSTAT 3. A glassy carbon disk, Pt wire, and Ag/Ag + (AgNO3 in acetonitrile) were used as the working, counter, and reference electrodes, respectively. Measurements were corrected to the ferrocene/ferrocenium redox couple as an internal standard. 1,2-Dichlorobenzene was used as the solvent with an analyte molarity of ca. 10 −5 M in the presence of 10 −1 M (n-Bu4N)(PF6) as a supporting electrolyte. Solutions were degassed with Ar and experiments run under a blanket of Ar.

Photophysics
UV−vis absorbance spectra were measured using a UV-1800 UV−vis spectrophotometer (Shimadzu) and UVProbe version 2.33 software. Photoluminescence (PL) spectra of solutions were recorded using a QePro compact spectrometer (Ocean Optics), FluoroMax-3 fluorescence spectrometer (Jobin Yvon) or a SPEX Fluoromax luminescence spectrometer (Jobin Yvon). Solid-state emission spectra and photoluminescence quantum yield (PLQY) were obtained using an integrating sphere (Labsphere) coupled with a 365 nm LED light source and QePro (Ocean Optics) detector. Photoluminescence decays in film and solution were recorded using nanosecond gated luminescence and lifetime measurements (from 400 ps to 1 s) using the third harmonic of a high-energy pulsed Nd:YAG laser emitting at 355 nm (EKSPLA) as an excitation source. The emitted light was focused onto a spectrograph and detected with a sensitive gated iCCD camera (Stanford Computer Optics) with subnanosecond resolution. Time-resolved measurements were performed by exponentially increasing gate and integration times. Further details are available in reference 1 . Low temperature experiments were conducted using a liquid nitrogen cryostat VNF-100 (sample in flowing vapour, Janis Research) under nitrogen atmosphere, while measurements at room temperature were recorded under vacuum in the same cryostat. Solutions were degassed using five freeze-pump-thaw cycles. Thin films were deposited from toluene solutions through drop cast or spin coat and dried under vacuum at room temperature. Å. The molecules form into thick bands/layers in the a/b plane. For 2·7CHCl3 there was some evidence of partial de-solvation, so the total number of solvent molecules of crystallisation should be taken as approximate. One of the four chloroform molecules was refined at half weight. The molecule lies on a symmetry element, so half is unique.         Cyclic voltammetry reveals a single oxidation wave and a single reduction wave for 1. Both of which are essentially reversible. The current response for the oxidation wave is twice that of the reduction but no narrowing of the peak separation occurs. This indicates that both phenoxazine moieties are oxidised simultaneously to radical cations with no significant electronic coupling between the two donors. The reduction occurs over the quinoxaline acceptor to form a radical anion.

Electrochemistry
The voltammetry of 2 is much more complex. Similarly to 1 and in line with previous observations, as the voltage increases beyond the oxidation potential a multi-electron oxidation occurs. Upon reversal of polarity the reverse wave is cathodically shifted and displays an even larger peak current response which tails off sharply. This cathodic stripping peak indicates that upon oxidation the molecule undergoes electrodeposition onto the electrode surface via the phenoxazine rings of either two or three fins of the triptycene. While this could be eliminated by introducing substituents to the reactive 3-and 7-positions of the phenoxazine rings, it indicates that such scaffolds have the potential for surface functionalisation of electrodes.
The reduction wave for 2 presents itself as three closely overlapping single electron reductions, one for each fin. This is further experimental proof of homoconjugation in the LUMO manifold as without any significant electronic communication between the fins one single, well-defined reduction would be expected.
The oxidation potential for both molecules is similar and the onset of oxidation identical showing that the position of the HOMO has not changed between the single fin and the iptycene. The first reduction potential is marginally higher for iptycene 2 so the LUMO is slightly more accessible. This indicates that 2 has a narrower band gap which is replicated by the redshift in the onset of absorbance and through DFT analysis.

Electron-hole analysis
Natural transition orbitals for states calculated in the ground state geometry with TD-PBE0/def2svp are shown in figures S6.1-6.32. Hole densities are in blue and particle densities in green. The results from electron-hole analysis used to determine excited state character are shown in Table S6.1 Hole-electron analysis was conducted using Multiwfn. [7][8][9] . In electron-hole analysis the Sr index is a measure to characterize the overlapping extent of holes and electrons by integration of the geometric mean of their charge densities over all space, which approaches zero as CT character increases. The D index is the total magnitude of CT length as measured by the distance between the centres of mass of electrons and holes, hence D typically increases with CT character. The t index represents the difference between the D index and the average spatial extension degree of hole and electron distribution in the CT direction, characterizing the separation degree of holes and electrons in the CT direction. Negative t indices imply that the holes and electrons are not substantially separated due to CT.
It is important to consider all parameters when assigning state character. For example, a CT state spread symmetrically across all the three iptycene fins of 2 (e.g., the T3 state - Figure  S6.8) will demonstrate a small D index as the hole and electron centroids are each technically over the centre of the molecule. However, hole and electron overlap is still low, and so the state can be assigned as CT based on the Sr index being close to zero. During analysis, we also noticed that when spread across multiple fins of 2, negative t indices could be observed even for states of clear CT character based on Sr and D indices e.g. S1 and T1.
The S1 state of 2 is assigned as CT in nature due to the low Sr index and high D index. Both the hole and electron are primarily localised over fin A. A marginally negative t index is observed for S1, although not as negative as for states of clear LE character such as T13 ( Figure S6.30). The slightly negative t index of S1 is expected to be related to the molecular symmetry of 2, which complicates assignment of a clear CT direction.
The electron-hole analysis for T1 is near-identical to that of S1, meaning both states are the same in configuration (CT) with the same spatial distribution.
The difference in the localisation of the CT state T8 compared to S1 is indicated by electronhole analysis -T8 has a significantly larger D index than S1, and a more positive t index. Calculated energy gap between S1 and Tn; b Calculated spin-orbit coupling matrix element between S1 and Tn; c CT = charge transfer, LE = locally excited.

Commented [PPH1]:
Shouldn't this be called "spin-orbit coupling matrix element (SOCME)"? I use term SOCME after Tom Penfold and others, but perhaps simply spin-orbit coupling is sort of clear enough?

Calculation of kinetic parameters of luminescent decay of TADF emitters
While there exist models to describe kinetics of TADF emitters the typical assumptions used require high (>0.9) triplet formation yield. Given a relatively modest contribution of TADF to overall emission in 1 and 2 it can be reasonably assumed that triplet formation yield is much lower than 0.9 and cannot be approximated with unity. The aim of the reasoning below is to establish a model with the minimal possible number of assumptions. Yet such assumptions should be reasonable in the context of the particular emitter.
In case of 1 and 2 the triplet formation yield, < 1. On the other side, both luminophores present strong fluorescent properties and invariance of prompt fluorescence lifetime with temperature. This suggests the non-radiative decay rate of the S1 state might be relatively low and thus can be neglected.
We know that: One can then establish a ratio between and from the luminescence decay and then calculate the partial luminescence yields of each decay component.
Assuming the only non-radiative process affecting singlet decay to be intersystem crossing, the can be estimated as follows: Now, using the known 10 relation below it is possible to give an estimate of : The value of can then be used to calculate 10 as all other parameters are known. to show only the ICT band. The intensity of the spectrum for 1 is multiplied by a factor of three to illustrate that the maximum extinction coefficient of the ICT band for 2 is greater than three times that of 1.

Photophysics in Zeonex matrix
Photoluminescent behaviour of 1 and 2 in Zeonex matrix is generally not very complex and is reminiscent of the properties in toluene solution. However, in this case the emission spectrum is significantly blue shifted due to lesser stabilisation of the ICT by the aliphatic Zeonex matrix ( Figure S9.1). Interestingly, the TADF photoluminescence spectrum is not identical with the prompt fluorescence, but nevertheless they are very similar. We believe that Zeonex restricts vibrations of luminophores leading to formation of a distribution of ICT energy, as opposed to toluene, where all molecules have identical energy of the ICT state. As a result of this distribution prompt and delayed fluorescence no longer follow monoexponential decay laws so that each of the exponential regions can be approximated with two or more monoexponential terms. Further to that, molecules with slightly lower CT energy (i.e. CT is more stabilised) show a shorter TADF decay than those with higher CT energy. This leads to an apparent blue shift of emission spectrum with increasing time delay in the TADF decay region ( Figure S9.3). On the other side, in the prompt fluorescence region the luminescence spectrum appears to red shift at larger delay after excitation ( Figure S9.4). This behaviour can be explained by a smaller transition oscillator strength of the stabilised CT states leading to their smaller singlet radiative rate. This description is in a good agreement with behaviour of 1 and 2 in Zeonex matrix.
Zeonex provides poor solubilisation of dispersed molecules 2 leading to aggregation at 1 wt % load ( Figure S9.6). This leads to an emergence of another emissive species which show distinctive luminescent behaviour ( Figure S9.7 and Figure S9.8). As we aim to study the intrinsic properties of 1 and 2 we choose to use conditions at which both present characteristics typical of individual molecules, thus 2 is studied at reduced load, 0.01% wt. The results obtained from 0.01 wt % films with 2 are used for comparison with the 1 wt % loaded films with 1. Importantly, both 1 and 2 demonstrate TADF at room temperature confirmed by laser fluence experiment ( Figure S9.9). It is worth to point out as a side note that another evidence for the intramolecular mechanism of delayed fluorescence in 2, i.e. TADF, is the invariance of the decay characteristics of the monomolecular emission with concentration ( Figure S9.6).
The aggregate emission identified in 2 1 wt % Zeonex films shows an interesting behaviour which deserves an extended discussion ( Figure S9.6 and Figure S9.7). At 295 K the decay of the aggregated species appears to involve a single exponential component visible at ~100ns -1 μs delay which can be attributed to prompt fluorescence. At 80 K apart from the prompt fluorescence component appearing in the same delay of ~100ns -1 μs as at 295 K, there exists another, longer lived exponential component at ~10 μs -1 ms delay. Given the conditions of the experiment this decay component might be attributed to phosphorescence emerging from the aggregate. However, given a great degree of similarity between time resolved spectra at the two distinct decay regions it is more likely the longer-lived component is related with delayed fluorescence of the aggregate. The apparent lack of this delayed fluorescence component at 295 K may be related to non-radiative triplet decay on one side. On the other side, the emission appears to be generally weaker than the monomolecular delayed fluorescence of 2, therefore it may not be clearly visible in the decay nor time-resolved spectra. As a final note, we believe the longer-lived decay component of aggregate photoluminescence to be most likely related with triplet-triplet annihilation due to facilitated triplet-triplet interactions in the aggregated state.

Photophysics in OLED host
Luminescent behaviour of 1 and 2 in PVK:PBD matrix is more complex than in toluene and somewhat more similar to Zeonex films. Similarly to other studied media, 1 and 2 show TADF also in the PVK:PBD matrix used as host in solution processed OLEDs 1 and 2 (Figure S11.4 and Figure S11.6). In contrary to films cast in Zeonex matrix the films obtained with PVK:PBD matrix show a much more significant distinction between early prompt fluorescence and delayed fluorescence (Figure S11.2. Time-resolved photoluminescence spectra of 1 in PVK:PBD matrix at 5 wt % emitter load in the prompt fluorescence region. Note the thicker black line indicates delayed fluorescence spectrum at 2 μs (left) and 1 μs (right) delay., Figure S11.3 and Figure S11.5). Most importantly, delayed fluorescence is red shifted in respect to prompt fluorescence at any time delay with the difference being larger in 1 than in 2. We believe this behaviour indicates that in the distribution of excited states only the ones with the most stabilised ICT produce any TADF at all, while others emit only through prompt fluorescence. This description would seem to indicate that triplet states produced through intersystem crossing in molecules with high ICT energy might be lost, increasing non-radiative losses. However, we believe these triplet states will most likely be scavenged by those molecules able to form low energy ICT through triplet hopping. In other words, molecules with low ICT energy will show behaviour reminiscent of triplet trapsdespite the triplet energy being identical to other luminophores in the blend they show accelerated triplet decay via. TADF.
While there is observed a distribution in energy of the S1 ICT state the same does not apply to the T1. At 80 K prompt fluorescence behaves in a similar fashion as it does at 295 K with the distribution of S1 energy reflected by the time-resolved spectra (Figure S11.3 and Figure  S11.5). Phosphorescence of 1 and 2 can be identified as a long-lived emission appearing in the 10-100 ms region of the decay. For both compounds the phosphorescence spectrum does not undergo any shifts or changes with time delay. This important observation indicates there is no distribution in T1 energy in both molecules and all emit phosphorescence from a state with identical energy. Such behaviour of S1 and T1 states suggests the effect on excited state energy is not related to strong ground state interactions, such as aggregation. Similar behaviour in solid film has been observed earlier. 11,12 Figure S11.1. Steady-state photoluminescence spectra of 1 and 2 in PVK:PBS matrix at 5 wt % emitter load.