Exploiting trifluoromethyl substituents for tuning orbital character of singlet and triplet states to increase the rate of thermally activated delayed fluorescence

This work shows that trifluoromethyl (CF3) substituents can be used to increase the rate of thermally activated delayed fluorescence (TADF) in conjugated organic molecules by tuning the excitonic character of the singlet and triplet excited states. The synthesis and detailed photophysical characterization of four new functionalized donor–acceptor–donor (D–A–D) compounds using CF3-substitutedthioxanthene-S,S-dioxide and dimethylacridine units are presented. Several compounds are reported to exhibit rapid blue/green thermally activated delayed fluorescence with major delayed fluorescence lifetime components of E1–2 ms. Changes in the orbital character of singlet and triplet states (local vs. charge transfer) result in significant changes in rates of delayed fluorescence, despite similar DEST values. The change in orbital character is well supported by TD-DFT calculations. Electrochemical measurements reveal strong shifts in redox potentials that can be induced by s-electron withdrawing CF3 substituents. Trifluoromethyl substituted compounds with a wider DEST also perform more efficiently than might be expected due to the demonstrated changes in excited state character. This study reveals important photophysical and molecular design implications for the future development of TADF emitters.


Introduction
Light emission from organic materials is revolutionizing the color-display and lighting industries. [1][2][3] However, traditional fluorescent molecules have severe drawbacks in these applications, as only 25% of the excited states (singlets) that are generated under electrical excitation are harvested as emission; the remaining 75% of triplet excitons decay non-radiatively. 4 Phosphorescent molecules can address this issue and achieve internal quantum efficiencies (IQEs) approaching 100% due to light emission from both singlet and triplet excited states. 5 This has led to extensive studies on emitters containing heavy metal atoms, where strong spin-orbit coupling promotes intersystem crossing (ISC) and phosphorescence (PH) emission. 6,7 The main disadvantages of organometallic emitters are their high cost due to the necessity of precious and possibly toxic heavy metals (typically iridium and platinum) and the electrochemical and thermal instability of the blue and especially deep-blue emitting complexes. 8 E-type delayed fluorescence, 9 which has been re-named thermally activated delayed fluorescence (TADF) is an alternative mechanism that harvests triplet excited states in all-organic emitters for light emission. 10,11 The main application of TADF materials is in organic light-emitting diodes (OLEDs). [12][13][14][15] TADF materials have also been deployed in fluorescence imaging, 16,17 optical temperature sensing 18,19 and catalysis. 20 Specific molecular design features enable fast reverse intersystem crossing (rISC), which can convert up to 100% of triplet excited states to emissive singlet states, leading to TADF. 21 Despite intense global research efforts, full understanding of the specific molecular features that enable rapid rISC remain elusive. Indeed, several different rISC mechanisms with different design rules have been discovered, including spin-vibronic coupling in donor-acceptor-donor (D-A-D) charge transfer (CT) emitters, 22 upper triplet state crossings, 23 resonant non-overlapping orbital distributions in planar boron compounds, 24 exciplexes, 25,26 and hybrid D-A compounds. 27 Ideal TADF materials possess both high photoluminescence quantum yields (PLQYs) and fast rISC rates, the latter of which are enabled by a small energy difference between the lowest singlet and triplet excited states (typically DE ST o 0.25 eV). These properties can be found in heteroaromatic donoracceptor (D-A) molecules with charge-transfer (CT) excited states. In these D-A molecules, a large average dihedral angle between the planes of the D and A units ensures the highest occupied and lowest unoccupied molecular orbitals (HOMO and LUMO, respectively) are spatially separated. This minimizes electron exchange energy and thus also DE ST . A detailed analysis of the mechanism of TADF in D-A-D materials has revealed a second-order spin-vibronic coupling process, in which a triplet state of local excitonic character ( 3 LE) mediates rISC between the triplet and singlet charge transfer states ( 3 CT and 1 CT, respectively). [28][29][30] The key roles of D-A conjugation and dihedral angles, molecular rigidity, conformation, connectivity, and steric effects are also important factors that influence CT formation and hence TADF efficiency and emission color. [31][32][33][34][35][36][37] Despite this level of understanding, stable blue emission, which is invaluable for high-quality full-color OLED displays, remains especially challenging to achieve. Only a limited number of blue/ deep-blue TADF emitters possessing good color purity have been reported. 36,[38][39][40][41][42][43][44] In this context a systematic series of TADF molecules that are functionalized at specific sites with trifluoromethyl substituents are herein presented. Trifluoromethyl (CF 3 ) is a strongly inductive electron-withdrawing group which makes it a useful substituent for emission colour tuning via donor and acceptor strength manipulation in D-A-D CT emitters. CF 3 is not p-conjugative, which makes it distinct from the electronwithdrawing groups that are typically used in TADF emitters, for example: cyano, 10,45 sulfone, 46,47 ketone, 48 triazine 49 and oxadiazole. 50 Trifluoromethyl groups can also be very stable to chemical, electrochemical, thermal and photochemical degradation, which are valuable properties to impart to TADF materials. [51][52][53][54] Extensive systematic studies of TADF emitters with CF 3 substituents are scarce. Current CF 3 functionality in TADF emitters are typically incorporated in the 'Ar-CF 3 ' type motif. 55 Examples include CF 3 -phenyl [56][57][58][59] or CF 3 -pyridyl 52,60 as an acceptor unit, or CF 3 -carbazole as a donor unit. 61,62 This present study further diversifies the library of reported CF 3 functionalised TADF molecules and has led to new blue and green TADF emitters by manipulation of the electron-donating and electron-accepting abilities of the D and A subunits with CF 3 substituents. In particular, we have investigated the use of the CF 3 unit as part of the bridge of the acceptor unit as opposed to the typical 'Ar-CF 3 ' motif currently observed in the literature. There are few examples of CF 3 'bridges' being utilized in TADF molecules. 53,63 Swager and co-workers showed that geminal-CF 3 substituents on (non-TADF) polyfluorene derivatives were more photostable compared to geminal-CH 3 substituents. 53 Adachi and co-workers used a flexible bis(CF 3 )-substituted methylene bridge directly between donor and acceptor in TADF emitters. 63 The synthetic routes, X-ray crystal structures, solution electrochemical data and a thorough photophysical investigation of the materials are presented here. These results show that trifluoromethyl substituents can be used to impart large shifts in redox potentials and emission properties through inductive effects. Molecules 1 and 2 show highly similar DE ST values, but 2 shows faster delayed fluorescence. Detailed analysis and supporting hybrid-DFT calculations reveal how the tuning of orbital character with trifluoromethyl substituents leads to an increase in TADF emission rates. OLEDs were also fabricated, although these devices showed unexpected instability.

Experimental section
The synthesis and structural characterization of the molecules are given in the ESI. † X-ray diffraction experiments: full data (including structure factors) in CIF format is available as ESI † and have been deposited with the Cambridge Structural Database, CCDC 1887606 (3) and 1887607 (4). † Optical and optoelectronic methods are given in the ESI. †

Molecular design and synthesis
Starting from the well-characterized blue TADF emitter 2,7-bis-(9,9-dimethylacridin-10-yl)-9,9-dimethylthioxanthene-S,S-dioxide (DDMA-TXO2) 1, (Fig. 1), the motivation was to explore the effect on photophysical properties that arise from tuning the donor and acceptor abilities of the sub-units using systematic CF 3 substitution. 64 Fig. 1 presents the chemical structures of 1 and the new analogs 2-5. The methyl groups of the thioxanthene acceptor in 1 are replaced by CF 3 groups in 2-4 ( Fig. 1), whereas CF 3 groups are added directly on to the aryl rings of the D unit in 5 (with both D and A modifications in 4). The acceptor CF 3 groups in 2-4 are on the bridging sp 3 carbon atom and are not directly attached to the p-electron system. The potential ability of these CF 3 groups to alter the singlet and triplet levels (and DE ST ) are therefore of both fundamental and practical interest. The donor within the series 1-4 is systematically changed from the 9,9-dimethylacridine units (in 1 and 2) to the 3,6-dimethoxycarbazole (in 3) and 2,7-bis(trifluoromethyl)-9,9-dimethylacridine (in 4 and 5).
The synthetic routes to 2-5 proceeded via palladiumcatalysed Buchwald-Hartwig coupling of suitably functionalized D and A fragments. 66 The identity and high purity of 2-5 were confirmed by a combination of NMR spectroscopy, high resolution mass spectrometry, and X-ray crystallography (see the ESI †). The synthesis of molecule 4 is shown in Scheme 1, as a representative example.
To prepare CF 3 substituted compound 4, the acceptor precursor 13 with a bis-CF 3 methylene bridge was required. Using modified literature procedures, acid chloride-containing 8 was synthesized using copper-mediated thioether synthesis, followed by conversion to acid chloride with oxalyl chloride. Pleasingly, 8 could be converted to bis-CF 3 -alcohol 9 using NMe 4 F and TMSCF 3 , 67 avoiding the use of gaseous hexafluoroacetone typically used to synthesize bis-CF 3 functionality. 53 Attempts to close alcohol 9 to thioxanthene 11 with standard reagents did not yield the desired product. 68 Therefore, 9 was converted to the mesylate 10 to introduce a more suitable leaving group. Treatment of 10 with a solid supported sulfonic acid resin (Amberlyst-15) resulted in formation of bis-CF 3 substituted thioxanthene 11. Interestingly GCMS analysis of the reaction revealed that 10 was also partly converted back to 9. Upon complete conversion of mesylate 10, no additional thioxanthene 11 was formed. This analysis shows that the mesylate leaving group was essential to form thioxanthene 11. Bromination of 11 to give dibromothioxanthene 12 was achieved with bromine in AcOH, but required a higher temperature to proceed than the corresponding dimethylthioxanthene derivative used to prepare the acceptor precursor of 1. 69 Typical oxidation of 12 gives the thioxanthene-S,Sdioxide acceptor precursor 13.
To prepare the 2,7-bis-CF 3 substituted acridine donor 18, the rings were constructed using starting materials with CF 3 functionality already installed. Ester formation yielded compound 15 from 14. Installation of the dimethylmethylene bridge was achieved by reaction of MeMgBr with 15 to give alcohol 16. Successful utilization of a recently reported protecting-group-free copper-catalyzed C-N coupling gave diarylamine compound 17. 70 Compound 17 was cyclized to give new bis-CF 3 -acridine donor 18 using Amberlyst-15. Compound 4 was synthesized from the coupling of 13 and 18 using typical Buchwald-Hartwig conditions, which generally tolerate a wide range of substituents for donor-acceptor couplings. 47,65,71,72 Compounds 2, 3 and 5 were also coupled using Buchwald-Hartwig conditions, with 2-5 prepared with yields ranging from 56 to 73%.

X-Ray crystal structures
The crystal structures of 3 and 4 are shown in Fig. 2. Structure 3 is modulated along the [1 1 1] direction: three symmetrically independent molecules are related roughly by pseudo-translations

Solution electrochemistry
Solution state cyclic voltammetry (CV) was performed to assess the effect of the trifluoromethyl substituents on the redox properties of 1-5. All CV measurements were performed with acetonitrile (MeCN) as solvent due to its wide electrochemical window. Trifluoromethyl substituents significantly improve the solubility of 2-5 compared to 1 in MeCN, making it possible to obtain good signal-to-noise CV traces for 2-5. The solubility of 1 is poor in MeCN but the measurements are in agreement with the reported oxidation potential of 1. 73 Table 1 and Fig. 3 summarize the CV data.
Several clear trends can be observed by systematic CF 3 substitution on 1. Addition of CF 3 groups in the para-position on the donor shifts the oxidation to significantly higher potential in 4 and 5 compared to unsubstituted 1 (E+0.55 -E+0.9 V). This shift is in agreement with the strongly s-withdrawing effect of the CF 3 substituent, destabilizing any resulting cation. Addition of CF 3 groups on the donor makes the oxidation wave reversible in 4 and 5 compared to 1 (see Fig. S4b and c for multiple CV cycles, ESI †). The stability imparted is similar to that seen by introduction of t Bu substituents in other acridine analogs, 23 but CF 3 groups are electron withdrawing in contrast to the inductive electron donating t Bu group. Interestingly, 5 has a highly reversible oxidation and reduction despite a wide electrochemical window of 3.16 V.
Addition of CF 3 groups on the acceptor in 2-4 reduces the reduction potential compared to unsubstituted 1 (E-2.3 V -E-2.1 V). This shift in reduction is quite substantial considering that the CF 3 groups on the acceptor in 2-4 are on an sp 3 hybridized bridge which is not in direct conjugation with the p-system, and highlights the strongly s-withdrawing effect of the CF 3 substituent. Unexpectedly, 2-4 all have irreversible reduction waves in contrast to 5. While the exact origin of this irreversibility is unclear, it is suggested that the s-bridge of the acceptor is likely to be highly d+ due to the CF 3 functionality and the sulfone group. These multiple electron withdrawing groups in combination could reduce the strength of neighbouring C-C alkyl-aryl bonds on the s-bridge resulting in bond cleavage upon reduction.

Optical and photophysical properties
All the molecules in this study have been characterized using UV/Vis absorption and a range of photoluminescence techniques.

Solution-state absorption
The UV/Vis absorption spectra of molecules 1-5 were obtained in DCM solution and are shown in Fig. 4.
All molecules in the series have p-p* transitions in the 275-320 nm region. Molecule 2 has an additional shoulder at 325 nm. Compounds 1-5 have charge transfer (n-p*) transitions at E350-375 nm. The dimethoxycarbazole moiety in 3 induces intense absorption bands in the 360-375 nm region, 74 which is due to a smaller torsion angle between the D-A units in carbazole functionalized systems (see Fig. 2: o601 dihedral angle for molecule 3 and 4701 for molecule 4). The 360-375 nm band shape in 3 is not pure CT in character, i.e. not classically  Gaussian; it appears as a normal pp* transition with some vibronic structure. However, this is very similar to what was previously observed in other carbazole systems, 75 where the carbazole units are partially conjugated giving a highly mixed LE/CT ground state and had CT character as shown by the electroabsorption. The previous observations are in agreement with 3 as the D and A are more conjugated in the ground state in 3 with a smaller D-A torsion angle than typically observed in carbazole units (Fig. 2). The highly mixed LE/CT nature of the state here gives rise to the far stronger absorbance on excitation in a polar solvent, the CT character is fully stabilized and emission from 3 falls in sequence with the other compounds.

Steady-state solution photoluminescence
The emission spectra of all compounds were obtained in solution to probe for solvatochromic behaviour. All molecules 1-5 show solvatochromic behavior in solution, with a red-shift in onset of emission ranging from 17-33 nm upon switching from toluene to DCM as solvent ( Table 2). Solvatochromism in combination with the broad Gaussian band shapes establishes predominately CT character in the emissive states of molecules 1-5. 1 CT emission is an important type of transition character typically required for efficient vibronically coupled TADF. The CF 3 functionality is electron withdrawing via inductive effects, yet it has a profound effect of the emission properties of 2-5 when compared with unsubstituted 1.   stronger CF 3 substituted acceptor (as in 2) is offset by the blueshift associated with weaker donor (as in 5), leading to a similar emission profile as in 1. Unsubstituted 1 has slightly bluer emission in toluene and the emission of 4 is slightly bluer in dichloromethane, suggesting weaker CT emission character in 4 compared to 1. This difference is reflected in the solvatochromic shifts from toluene to dichloromethane in molecules 1 and 4.
All molecules 1-5 show some degree of red-shift in emission from apolar zeonex to DPEPO and UGH3 hosts. This red-shift is mainly due to packing effects in the solid state, as can be seen through the consistent similarity between UGH3 and DPEPO film spectra. Molecule 4 shows a significantly smaller shift in emission color between hosts. It is suggested that the weak CT strength makes 4 less sensitive to polarity of the surrounding medium, as found in solution state emission measurements (Fig. 5). Interestingly, the high degree of hydrophobicity imparted by the more extensive trifluoromethylation on molecule 4 may limit spectral shifts associated with modifications in molecular packing, considering that 5 (with a weaker acceptor than 4) and subsequently weaker charge transfer still significantly shifts from zeonex to DPEPO or UGH3 hosts.

Time-resolved photoluminescence
Time-resolved emission spectroscopy was performed on films of 1-5 in DPEPO host to determine the effect of trifluoromethyl substituents on the TADF properties of 2-5 in comparison to 1. Fig. 7 shows the time resolved emission decays and Fig. 8 shows snapshots of the emission at various time delays during the time resolved decay.
Compounds 1-4 all show charge transfer (CT) emission in the early ns region with a slight red-shift of the emission during in the first 80 ns of decay ( Fig. 8 and Table 4). The CT emission at 80 ns representing the prompt fluorescence (PF) closely resembles the delayed fluorescence (DF) emission at 2 ms. This is typical of emission originating from a TADF process, where intersystem crossing to the triplet excited state, and reverse intersystem crossing back to the 1 CT state has occurred. Molecules 1-4 have DE ST gaps r 0.22 eV with varying DF emission lifetimes. Temperature dependence measurements on 2-4 show increased emission intensity in the ms delayed fluorescence region ( Fig. S5a and b, ESI †). Power dependence measurements (Fig. S5c, ESI †) rule out multi-exciton emission processes with a slope close to one for 2-4, confirming that vibronically coupled TADF processes are occurring.
The lifetimes for PF and DF emission for 1-4 were determined by fitting with dual exponential decay curves (Fig. 7). Dual exponential curves gave good R 2 values (40.98, ESI †) for all fittings with o10% standard error in all cases (Table S1, ESI †). The emission for 3 and 5 was very weak and consequently the error in fitting the emission is larger. The start of the fitting for the DF was carefully considered to minimize interference  1  406  407  409  441  452  503  32  51  2  428  427  446  479  515  572  33  57  3  413  416  427  457  502  554  30  52  4  391  391  406  423  462  492  17  30  5  374  374  373  392  410  441  19  31 a Measured in DCM solvent. b Measured in toluene solvent. c Shift in emission onset or l max from toluene to DCM solvent.

Research Article Materials Chemistry Frontiers
Open Access Article. Published on 29 July 2020. Downloaded on 9/24/2020 10:10:52 PM. This article is licensed under a Creative Commons Attribution 3.0 Unported Licence.

View Article Online
from the overlapping PF signal as much as possible at approx. 100 ns. For 1, 2 and 4 both components of the dual exponential decays contribute significantly to the total PF and DF emission (Table 3). This kinetic analysis for 1, 2 and 4 suggests that there   Table S1 (ESI †). Lifetimes and corresponding weightings are shown in Table 3.

Materials Chemistry Frontiers Research Article
Open Access Article. Published on 29 July 2020. Downloaded on 9/24/2020 10:10:52 PM. This article is licensed under a Creative Commons Attribution 3.0 Unported Licence.

View Article Online
are two 'groups' of varying conformation/packing effects in the films that result in either a faster or a slower emission lifetime. In most cases the faster component of the DF or PF is the major emission component (Table 3). Unexpectedly, CF 3 -substituted compound 2 has a faster DF emission lifetime than 1 for both exponential components. This is a highly thought-provoking  This raises the question: why is 2 emitting delayed fluorescence faster than 1 when the DE ST values and structures are so similar? The emission spectrum of 2 is red-shifted compared to 1 (Fig. 5 and 6), highlighting a more stabilized CT state. The triplet (T 1 ) energies of 2-4 are lower compared to 1, showing that the CF 3 groups on the acceptor are responsible for the reduction in T 1 energy. The discussed analysis suggests that the lower T 1 energy is 3 CT in character and less localized due to the significantly increased acceptor strength. This change in triplet state character in 2 and 4 is likely to adjust the vibrational modes to the system, and it is proposed that in this case there is efficient mixing of 3 LE and 3 CT to allow rISC to occur. It is suggested that the differences in triplet state character improve the vibronic mixing of states, resulting in faster delayed fluorescence in 2. This may also explain why 4 exhibits DF with a similar lifetime to 1 despite a wider DE ST in 4 (Table 4).  Fig. 7 at the specific times given each figure panel. A separate measurement was performed at 80 K to obtain the phosphorescence spectrum at 80 ms delay times to give an unambiguous phosphorescence spectrum. Table 3 Dual exponential lifetime fittings of time resolved emission decay data in Fig. 7 for the prompt fluorescence (PF) and delayed fluorescence (DF). Major components of the PF and DF exponentials are in bold. The prompt fluorescence in compound 5 could be fitted with a single exponential, and no DF was detected  12.9 7.5 Â 10 4 (100) -----a % contribution of each PF exponential to the total PF contribution based on A 1 and A 2 values. b % contribution of each DF exponential to the total DF contribution based on B 1 and B 2 values. c The signal for the delayed fluorescence in 3 is very weak in comparison to molecules 1, 2 and 4 (see intensity in Fig. 7 and 8), and consequently only one lifetime from the fitting of 3 is meaningful due to strong overlapping PF signal at earlier times.

Computations
Geometry optimizations on the ground state (S 0 ) geometries of 1-5 at the hybrid DFT functional CAM-B3LYP/6-31G(d) revealed D-A orientations to be orthogonal in all molecules except for 3 where the averaged D-A dihedral angle of approximately 501 is obtained (Table 5). Good agreements between computed and experimental geometries were found for 3 and 4 (  Table S5, ESI †). The perpendicular D-A energy barrier in 3 is estimated to be 2 kcal mol À1 thus a range of conformers of 3 in solution would be present. Excited state geometries of 1-5 were also optimized here by TD-DFT for S 1 states and by unrestricted DFT (UDFT) for T 1 states with D-A dihedral angles of the optimized geometries listed in Table 5. For all molecules except 3, the D-A orientations remain orthogonal on going from S 0 to S 1 so a relatively small geometrical rearrangement would take place after excitation and geometry relaxation to the most stable S 1 state for each geometry. To gain insight into the observed different emission data for 1-5, results from TD-DFT data carried out on the optimized S 1 geometries are analyzed here with the assumption that these S 1 geometries are responsible for the charge-transfer emissions and are summarized (Table S8, ESI †).
Table S8 (ESI †) and Fig. 9 show the energies of the S 1 , T 1 and T 2 states and their orbital/excitonic characters for 1 and 2. The 3 LE state is necessary to allow ISC/rISC processes to occur since the 1 CT and 3 CT exchange is formally forbidden. 28,29 Large spin-orbit couplings (SOC) between 1 CT and 3 LE (Table S9, ESI †) with vibronic coupling (VC) between 3 CT and 3 LE do take place if the 3 LE state, 3 CT and 1 CT are close together. Compound 2 has a fast DF rate and T 1 with 3 CT character, whereas compound 4 has a slower observed DF rate and T 1 with 3 LE character like 1 ( Table 3, and Fig. S7d, ESI †). Comparing compounds 1, 2 and 4, where very small systematic substitutions are made to the structure, it appears that the different T 1 character predicted is related to the DF rate where a faster rate occurs when T 1 has 3 CT character. Fig. 9 highlights the differences between 1 where the 1 CT-3 LE gap dominates whereas in 2 the 3 CT-3 LE gap dominates. Thus in 1 rIC to populate T 2 is the rate limiting step, whereas rIC is not rate limiting in 2, and 2 has faster rISC than 1. 76 In the case of 3, there is considerable early prompt fluorescence 1 CT (Fig. 8) and relatively low DF intensity, making it difficult to make quantitative comparisons in rates between 3 and the other compounds (Fig. 7). If the geometry in 3 contains D-A dihedral angles constrained at 501 in the S 1 state, the TD-DFT data provide an entirely different picture where the S 1 , T 1 and T 2 states are up to 0.59 eV apart and no rISC process would be expected ( Fig. S7e and Table S4, ESI †). For 4, the S 1 , T 1 and T 2 states are all within 0.17 eV and the larger energy gap suggests a less efficient rISC process in 4. The nature of the 1 CT character in 4 with only 74% CT (and thus 26% local) is much less than the 1 CT characters in the other molecules (1-3 and 5) a Two averaged dihedral angles are reported because of the D-A-D architecture. When 1-5 are excited and charge transfer occurs, only one electron is transferred from a single donor to the central acceptor. In this excited state the other donor is acting as a 'pendant' substituent that will play a role in the electronics, but in a more passive fashion compared to the donor that transfers an electron. The 'pendant' substituents in the excited state geometries (S 1 , T 1 ) contain similar dihedral angles to those in the ground state geometries (S 0 ).  (Fig. 6). b Onset of relaxed CT emission at approx. 75 ns as indicated in Fig. 8 for 1-4. For 5 the relaxed onset was taken at 11 ns due to little signal beyond this time. c Onset of phosphorescence emission at 80 K with a time delay of 80 ms (Fig. 8). d Calculated from steady state emission in DPEPO as in a. T 1 energy taken from c. e Calculated from relaxed CT emission as in b. T 1 energy taken from c. f Taken from steady state emission in DPEPO host (Fig. 6). g Taken from evaporated DPEPO film data from literature. 73 h Measured at 25 wt% in DPEPO evaporated film. Evaporated film at higher emitter concentration required in the blue region to obtain strong signal due to scattering and lower detector sensitivity in the blue region of the spectrum. i 1 wt% in zeonex matrix. Signal was too low to be measured in DPEPO film as shown by the signal to noise ratio in Fig. 7.
with 93-95% CT. This is in accord with the much weaker solvatochromic effect on the emissions of 4 compared to those of 1-3 and 5. Compound 5 has no observable TADF as expected with the large S 1 -T 1 gap of 0.31 eV from TD-DFT (Fig. S7f, ESI †) and 0.35 eV from experimental measurements. In general, the TD-DFT data is in good agreement with the experimental results and supports the proposed differences in TADF rates.

Device fabrication
OLEDs using compounds 2 and 4 in various hosts (Section S5, ESI †) were found to be unstable, with brightness dropping below half its initial value within seconds at constant voltage driving. Simultaneous drops in device current were observed at constant voltage, and despite the reasonable PLQY values, device EQEs were below 1%. This emitter instability is in accordance with the irreversibility of the reduction in the electrochemical measurements (Fig. 3). Considering that 1 produces more stable devices, 65 it is likely that the acceptor is responsible for the instability in the OLED devices of 2 and 4. The instability is likely to be due to the very high acceptor strength which results in cleavage of a C-C or C-S bond of the acceptor on reduction, rather than instability of the CF 3 functionality itself.

Conclusions and outlook
The CF 3 -substituted acceptor in 2-4 has been an invaluable structural tool for probing the effects of orbital/exciton character on the rates of TADF emission to further understand the TADF mechanism and improve future molecular design. A key finding is that changing the character of the TADF-active singlet and triplet states through tuning with trifluoromethyl substituents improves the rate of TADF with rapid delayed emission components in the early microsecond regime. This work highlights how changes in the orbital character of singlet and triplet states To further enhance the rate of delayed fluorescence, and thereby reduce the probability of triplet interactions which cause detrimental annihilation effects, it is necessary to gain a deeper understanding of how singlet and triplet state character and orbital distribution affect emission rates. This will be a focus of future studies.

Author contributions
The

Conflicts of interest
The authors declare no competing financial interest.