1.

Introduction

Thermally activated delayed fluorescence (TADF) of organic molecules has attracted tremendous attention owing to its potential application to organic light-emitting diodes (OLEDs) and has also renewed interest in photochemistry and photophysics.^1,2 Uoyama et al.³ established a design concept for the synthesis of highly efficient TADF molecules by connecting electron-donating and -accepting substituents with a large torsion angle. Such a structure can separate the wavefunction overlap between the highest occupied molecular orbital and the lowest unoccupied molecular orbital, leading to a significant reduction in the energy gap ($\mathrm{\Delta }{E}_{\mathrm{ST}}$) between the first excited singlet state ($S_1$) and the first excited triplet state ($T_1$). Notably, the $\mathrm{\Delta }{E}_{\mathrm{ST}}$ diminishes to thermal energy at room temperature, $\sim 25\mathrm{meV}$, and typically $<100\mathrm{meV}$. Thus, this approach affords a thermal pathway to convert from the $T_1$ state to the $S_1$ state via reverse intersystem crossing (RISC), thereby facilitating TADF. After Uoyama’s report,³ a number of TADF molecules were rapidly synthesized and exhibited a high internal photoluminescence quantum yield (PLQY) of TADF (${\mathrm{\varphi }}_{\mathrm{DF}}$) in various emission colors, including blue.^4–10 Additionally, the device performance of OLEDs, such as external electroluminescence quantum efficiency and operational stability, is being improved by this approach.^11–14

In addition to these exciting applications of TADF, the fundamental processes in TADF are also attracting considerable interest in both experimental and theoretical fields.^15–26 In this case, the goal is to achieve a sufficient understanding of these processes to realize a high RISC rate ($k_{\mathrm{RISC}}$) by judicious designing and optimizing the most suitable chemical structures, emission environments, and so on. Increasing the $k_{\mathrm{RISC}}$ suppresses triplet state quenching and is, therefore, beneficial for reducing roll-off in OLEDs and organic laser diodes.^27–29

One important question of the TADF mechanism concerns the rate of the various electron transfer processes. That is, although $\mathrm{\Delta }{E}_{\mathrm{ST}}$ values at thermal energy or even less have been attained, and intersystem crossing rates ($k_{\mathrm{ISC}}$) as high as ${10}^8{\mathrm{s}}^{-1}$ have been obtained, the corresponding $k_{\mathrm{RISC}}$ values were often limited to a maximum of ${10}^6{\mathrm{s}}^{-1}$. Moreover, ${\mathrm{\varphi }}_{\mathrm{DF}}$ values close to 80% were achieved, even with a relatively appreciable $\mathrm{\Delta }{E}_{\mathrm{ST}}$ value of $\sim 0.2\mathrm{eV}$.²⁴ Theoretical simulations have indicated that selecting certain molecular orbitals between the $S_1$ and $T_1$ states is essential to achieve a high $k_{\mathrm{RISC}}$; known as El-Sayed’s rule, it has been shown that high $k_{\mathrm{RISC}}$ values can be achieved by exploiting the singlet charge-transfer (${}^1\mathrm{CT}$) and triplet locally excited (${}^3\mathrm{LE}$) states.^18,19,21,25 In addition, some groups have proposed that second-order spin-orbit coupling enables utilization of higher order triplet states, which further facilitates RISC.^18,20,23 Conventional and time-dependent density functional theory calculations based on such orbital selection rules and/or second-order perturbation theory have succeeded in reproducing the experimentally determined $k_{\mathrm{RISC}}$ values of some, but not all, of the organic molecules investigated.²¹ Thus, a broad consensus of the TADF mechanism has yet to be reached.

Another aspect of the selection rules for RISC is our recent proposal based on experimental data that the energy gap between ${}^1\mathrm{CT}$ and ${}^3\mathrm{LE}$, so-called $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$, is an important parameter to signify TADF [Fig. 1(a)].²⁴ The carbazole benzonitrile (CzBN) derivatives 2CzBN, o-3CzBN, m-3CzBN, p-3CzBN, 4CzBN, and 5CzBN [Fig. 1(b)] have shown relatively similar $\mathrm{\Delta }{E}_{\mathrm{ST}}$ values of $\sim 0.2\mathrm{eV}$, as determined by their corresponding fluorescence and phosphorescence spectra, together with fairly different $k_{\mathrm{RISC}}$ values from zero to ${10}^5{\mathrm{s}}^{-1}$.²⁴ Assuming that their ${}^3\mathrm{LE}$ values are similar, which arise from structurally common Cz moieties (vide infra), the $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$ values estimated from the ${}^3\mathrm{LE}(\mathrm{Cz})$ and ${}^1\mathrm{CT}$ levels of each CzBN derivative varied from $-0.06$ to 0.23 eV because of different ${}^1\mathrm{CT}$ levels for each CzBN derivative. As a result, $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$ and $k_{\mathrm{RISC}}$, but not $\mathrm{\Delta }{E}_{\mathrm{ST}}$, were found to be correlated; specifically, decreasing values of $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$ lead to an increase in $k_{\mathrm{RISC}}$. For this reason, $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$ was proposed as an additional measure, in conjunction with $\mathrm{\Delta }{E}_{\mathrm{ST}}$, for determining whether molecules could exhibit TADF. However, further investigations are needed to verify the generality of this concept; in particular, it is desirable to establish whether tuning $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$ can indeed facilitate TADF and vice versa.

Fig. 1

(a) Definition of $\mathrm{\Delta }{E}_{\mathrm{ST}}$ and $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$. (b) Chemical structures of CzBN derivatives.

In this paper, we investigated the solvent dependence of the photophysical properties of six types of CzBN derivatives. It is known that a host matrix can tune CT state energies and that large environmental dielectric fields stabilize CT states by orienting the polarization of the surrounding solvent. Notably, LE states are hardly affected by this phenomenon.³⁰ This means that selecting the type of host matrix (e.g., solvent) can modulate the energy levels of CzBN derivatives and give rise to different photophysical properties. We previously investigated the photophysical properties of CzBN derivatives in toluene, whose dielectric constant ($ϵ$) is $\sim 2$ at room temperature. In the present study, the strong polar solvent acetonitrile ($ϵ\sim 37$) was used to investigate the photophysical properties of the same six CzBN derivatives, the results of which are compared with the previous results determined in toluene.²⁴ The large $ϵ$ of acetonitrile is expected to stabilize the ${}^1\mathrm{CT}$ and ${}^3\mathrm{CT}$ states and, thus, tune the energy levels, in particular $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$, of the CzBN derivatives.

2.

Experimental

All the materials were synthesized according to Ref. 24 and purified by thermal sublimation. Solutions were prepared by dissolving the purified molecules in toluene or acetonitrile, both of which were of spectrochemical analysis grade (purity: 99.8%, Wako Pure Chemical Industries, Ltd.). The solution concentration ranged from ${10}^{-4}$ to ${10}^{-5}\mathrm{M}$ depending on the samples.

Optical properties of the solutions were characterized by measuring the steady-state ultraviolet-visible (UV-vis) absorption/PL spectra and transient PL decay curves (hereafter called transient PL). UV-vis absorption spectra were acquired using a UV-VIS-NIR scanning spectrophotometer (UV-3100PC, Shimazu), whereas PL spectra were recorded using a fluorescence spectrophotometer (LS50B, Perkin Instruments or Lambda 950-PKA Instrument, Perkin–Elmer). Transient PLs were measured using a fluorescence lifetime spectrometer (C11367-01, Hamamatsu Photonics or a FluoroCube, Horiba). The PLQY of the solutions was measured using an absolute PLQY measurement system (C11347-01, Hamamatsu Photonics) with an excitation wavelength of 337 nm under two conditions: as prepared and after degassing by dry argon gas to eliminate triplet state deactivation. Before the transient PL measurements, the solutions were also deoxygenated with dry nitrogen or argon gas for the same reason. The triplet state lifetime of 2CzBN in acetonitrile was determined by microsecond-transient absorption spectroscopy using the third harmonic of fundamental light (1064 nm) of a ${\mathrm{Nd}}^{3+}:\mathrm{YAG}$ laser as the pump light (wavelength, 355 nm; FWHM pulse, $<150\mathrm{ps}$; repetition, 10 Hz) and a xenon steady-state lamp as the probe light. Full details of this design are described in Ref. 31. The irradiated intensity of the pump laser was set to $0.7\mathrm{mJ}/{\mathrm{cm}}^2$.

3.

Results and Discussion

Figures 2(a)–2(f) show the UV–vis absorption, fluorescence, and phosphorescence spectra of the six CzBN derivatives in toluene (upper) and acetonitrile (bottom). In the absorption spectra, $\pi -{\pi }^{*}$ transitions at a single Cz moiety $[\mathrm{Cz}(\pi -{\pi }^{*})]$ are commonly observed in the UV region (3.7 to 4.0 eV). Below this region, two types of CT transitions are observed depending on the molecules: CT1, which originates from a CT transition from one Cz moiety to the center BN moiety, located at $\sim 3.4$ to 3.6 eV, and CT2, which is a CT transition among two Cz moieties at para-positions [2 and 5 or 3 and 6, see blue-colored Czs in Fig. 1(b)] and the BN moiety at the lowest energy region ($\sim 3.0$ to 3.2 eV).²⁴ Both CT transitions are apparent in the absorption spectrum of 5CzBN [Fig. 2(f)]. Their formation in 5CzBN can be understood in terms of its chemical structure, which is composed of two para-Cz pairs and a single Cz moiety opposite the cyano group; collectively, these structural features fulfill the electronic condition of both CT transitions. By comparison, only the CT1 transition was present in the spectra of 2CzBN, o-3CzBN, and m-3CzBN, and only the CT2 transition was evident in the spectra of p-3CzBN and 4CzBN. In terms of their molecular structures, the former three have no such Cz pairs, and the latter two possess Cz pairs. The structure-dependent absorption of the CzBN derivatives in toluene was also observed viz. their corresponding transient absorption spectra, as reported in Ref. 24.

Fig. 2

UV–vis absorption, fluorescence, and phosphorescence (77 K) spectra of (a) 2CzBN, (b) o-3CzBN, (c) m-3CzBN, (d) p-3CzBN, (e) 4CzBN, and (f) 5CzBN in toluene (upper) and acetonitrile (bottom). Threshold (or partial peak top) energies in the fluorescence and phosphorescence spectra are depicted by blue or red vertical arrows, respectively.

The three distinct absorption features in Figs. 2(a)–2(f) [i.e., Cz ($\pi -{\pi }^{*}$), CT1, and CT2] indicate different photophysical behavior of the CzBN derivatives in toluene and acetonitrile. The peak positions of the Cz ($\pi -{\pi }^{*}$) transition remain almost unchanged, whereas both the CT1 and CT2 transitions exhibit a blue shift and a decrease in intensity in acetonitrile. Moreover, the intensity decrease is much stronger for CT1 compared with CT2; for instance, the peak structure of the CT1 transition changes to a shoulder-like feature in acetonitrile for 5CzBN. These results suggest that both the CT1 and CT2 transitions contain $n-{\pi }^{*}$ character, but the solvent effect differs between local CT (CT1) and delocalized (at para-Cz pairs) CT (CT2), presumably due to the distinct electron distributions in their excited states.

Solvent effects were also evident in the fluorescence and phosphorescence spectra of the six CzBN derivatives. Essentially, both the fluorescence and phosphorescence spectra show a red shift in acetonitrile, as seen in Figs. 2(a)–2(f), which we attribute to stabilization of both the ${}^1\mathrm{CT}$ and ${}^3\mathrm{CT}$ states. Only the phosphorescence spectrum of 2CzBN in toluene exhibits distinctly sharp vibronic progressions, which are attributable to emission from ${}^3\mathrm{LE}$ in the Cz moieties.²⁴ This result also suggests that the ${}^3\mathrm{LE}$ state of 2CzBN in toluene is lower in energy than the ${}^3\mathrm{CT}$ state. By comparison, the phosphorescence spectrum of 2CzBN in acetonitrile is rather broadened, suggesting an exchange of the $T_1$ from ${}^3\mathrm{LE}$ to ${}^3\mathrm{CT}$ or an increase of electronic coupling between the two states, i.e., the strengthening of state mixing, due to their energy proximity because of the ${}^3\mathrm{CT}$ stabilization.^32,33

On the basis of the fluorescence and phosphorescence results, the $\mathrm{\Delta }{E}_{\mathrm{ST}}$ and $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$ values of each CzBN derivative were estimated and are given in Table 1, where the $S_1$ and $T_1$ energies used for the estimation are also depicted. For the $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$, the energy of the ${}^3\mathrm{LE}$ state in the phosphorescence spectrum of 2CzBN was used, as with our previous report.²⁴ Here we note that both the $\mathrm{\Delta }{E}_{\mathrm{ST}}$ and $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$ values in the toluene samples are slightly differed from the previous reports owing to the use of different apparatus and measurement conditions. As seen in Table 1, the $\mathrm{\Delta }{E}_{\mathrm{ST}}$ values in toluene range from 0.16 to 0.27 eV, and these values are not appreciably reduced in acetonitrile, except for 5CzBN; in this case, 5CzBN showed a decrease in $\mathrm{\Delta }{E}_{\mathrm{ST}}$ from 0.16 to 0.07 eV. In this context, the $k_{\mathrm{RISC}}$ value of only 5CzBN is anticipated to increase according to the relation between $\mathrm{\Delta }{E}_{\mathrm{ST}}$ and $k_{\mathrm{RISC}}$: $k_{\mathrm{RISC}}\sim 1/\exp (\mathrm{\Delta }{E}_{\mathrm{ST}}/kT)$, where $k$ is the Boltzmann constant and $T$ is the temperature. However, as will be shown, this expectation is misleading because $k_{\mathrm{RISC}}$ is actually influenced significantly by $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$. As expected from the decrease in the ${}^3\mathrm{CT}$ energies, the $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$ values of all the CzBN derivatives decreased in acetonitrile. In particular, the $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$ values of o-3CzBN and m-3CzBN are reduced largely from 0.16 to 0.04 eV and from 0.20 to 0.09 eV, respectively. These values are less than that of p-3CzBN examined previously, which was the only isomer of the three that showed TADF in toluene.²⁴ On the basis of the present results, TADF is expected to be active for o-3CzBN and m-3CzBN in acetonitrile.

Table 1

ΔEST and ΔEST(LE) of CzBN derivatives in toluene and acetonitrile.

To examine the $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$ values in the context of TADF properties, we next discuss the results of transient PL for all the sample solutions. As seen in Figs. 3(a)–3(f), the transient PL measured for toluene (red line) are indicated in the bottom axes as a function of their delay time, whereas those of acetonitrile (blue line) are indicated in the upper axes. Note that the time scale for the three 3CzBN isomers differs by orders of magnitude in toluene ($\mu \mathrm{s}$) and acetonitrile (ms). As expected from the results of $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$, o-3CzBN and m-3CzBN, which are TADF inactive in toluene, become TADF active in acetonitrile. Delayed PL components with time constants (${\tau }_{\mathrm{DF}}$) of 66 and $220\mu \mathrm{s}$ for o-3CzBN [Fig. 3(b)] and m-3CzBN [Fig. 3(c)], respectively, were confirmed. These ${\tau }_{\mathrm{DF}}$ values are comparable with that of p-3CzBN in acetonitrile [$103\mu \mathrm{s}$, Fig. 3(d)]. Importantly, ${\mathrm{\varphi }}_{\mathrm{DF}}$ of the three 3CzBN isomers are similar: 21% for o-3CzBN, 18% for m-3CzBN, and 24% for p-3CzBN. These values were calculated by subtracting the PLQY in air (${\mathrm{\varphi }}_{\mathrm{PL},\text{air}}$) from the PLQY measured after argon bubbling (${\mathrm{\varphi }}_{\mathrm{PL},\text{degassed}}$). After analyzing the photophysical properties of all the solution samples according to Ref. 27, the values of $k_{\mathrm{RISC}}$ are obtained as $3.0\times {10}^4{\mathrm{s}}^{-1}$ for o-3CzBN, $6.4\times {10}^3{\mathrm{s}}^{-1}$ for m-3CzBN, and $2.1\times {10}^4{\mathrm{s}}^{-1}$ for p-3CzBN. The $k_{\mathrm{RISC}}$ of p-3CzBN is slightly improved from $1.3\times {10}^4{\mathrm{s}}^{-1}$ in toluene, but that of o-3CzBN and m-3CzBN improved at least 1 to 2 orders of magnitude, based on the assumption that ${10}^2{\mathrm{s}}^{-1}$ is a lower limit of our estimation. These results clearly indicate that a smaller $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$ increases the value of $k_{\mathrm{RISC}}$, thereby facilitating TADF. Therefore, TADF can be triggered from being inactive to active by environmental effects. Note that the $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$ value of 0.20 eV for 2CzBN in acetonitrile is still too large for TADF to develop.

Fig. 3

Comparison of transient PL of CzBN derivatives in toluene (red line, bottom axis) and acetonitrile (blue line, upper axis): (a) 2CzBN, (b) o-3CzBN, (c) m-3CzBN, (d) p-3CzBN, (e) 4CzBN, and (f) 5CzBN.

It seems that the effect of reducing $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$ on $k_{\mathrm{RISC}}$ is limited, as the $k_{\mathrm{RISC}}$ values of 4CzBN and 5CzBN improved only slightly from $1.8\times {10}^5{\mathrm{s}}^{-1}$ to $2.1\times {10}^5{\mathrm{s}}^{-1}$ and from $2.4\times {10}^5{\mathrm{s}}^{-1}$ to $2.6\times {10}^5{\mathrm{s}}^{-1}$ in acetonitrile, respectively. This observation may be deduced from the fact that their $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$ values become more negative, which means that the ${}^3\mathrm{LE}$ state lies farther from the ${}^1\mathrm{CT}$ state in acetonitrile. This negative shift in $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$ does not impact the ${\mathrm{\varphi }}_{\mathrm{DF}}$ value of 4CzBN, although it does significantly decrease the ${\mathrm{\varphi }}_{\mathrm{DF}}$ value of 5CzBN from 76% in toluene to 24% in acetonitrile. This decrease is caused by an increase in the nonradiative decay rate from $T_1$ to the ground state ($S_0$) ($k_{\mathrm{nr},\mathrm{T}}$) from $4.2\times {10}^3{\mathrm{s}}^{-1}$ (in toluene) to $9.3\times {10}^4{\mathrm{s}}^{-1}$ (in acetonitrile). As mentioned above, however, $k_{\mathrm{RISC}}$ does not decrease in acetonitrile, indicating that coupling between the ${}^3\mathrm{LE}$ and ${}^1\mathrm{CT}$ states is important for RISC.

Additionally, it is instructive to discuss the impact of the solvent effect for other rate constants relevant to the electron transfer processes in the six CzBN molecules examined. In acetonitrile, the radiation decay rate from $S_1$ to $S_0$ ($k_{\mathrm{r},\mathrm{S}}$) and $k_{\mathrm{ISC}}$ tends to decrease (Table 2). However, this is not the cause of the increase in the nonradiative decay rate for the $S_1$ to $S_0$ transition ($k_{\mathrm{nr},\mathrm{S}}$), since the ${\mathrm{\varphi }}_{\mathrm{PL},\text{air}}$ of the CzBN derivatives, which assumes a PLQY of prompt fluorescence, remains unchanged or even increases, except for 2CzBN and o-3CzBN. Although $k_{\mathrm{nr},\mathrm{T}}$ of 5CzBN is increased in acetonitrile, the other CzBN derivatives showed a decrease of $k_{\mathrm{nr},\mathrm{T}}$. Overall, it is clear that ${\mathrm{\varphi }}_{\mathrm{PL},\text{degassed}}$ of the three 3CzBN isomers and 4CzBN increased in acetonitrile.

Table 2

Photophysical properties of the CzBN derivatives.

The change in the rate constants in acetonitrile is qualitatively similar to 4CzIPN,³⁴ implying that solvent effects are a general phenomenon for TADF molecules. The only difference is that the TADF efficiency in toluene, the most nonpolar solvent used by Ishimatsu et al.,³⁴ demonstrates the best value, suggesting that the absolute value of $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$ is a minimum in toluene for 4CzIPN, as for 5CzBN. In fact, the ${}^3\mathrm{LE}$ state of 4CzIPN is already above the ${}^1\mathrm{CT}$ and ${}^3\mathrm{CT}$ states (Fig. 4). This result further supports the solvent effect for TADF efficiency and suggests that the most appropriate “solvent environment”, that will couple the relevant electronic states, ${}^1\mathrm{CT}$, ${}^3\mathrm{CT}$, and ${}^3\mathrm{LE}$, for TADF efficiency depends on a combination of both the emitter molecules and the solvent.

Fig. 4

Energy level diagram of 4CzIPN in toluene. The $S_1$ and $T_1$ energies and the $\mathrm{\Delta }{E}_{\mathrm{ST}}$ value are taken from Ref. 24. The ${}^3\mathrm{LE}$ state at IPN is reported in Ref. 35.

4.

Conclusion

This study examined the solvent dependence of the photophysical properties of a series of CzBN derivatives. Compared with toluene (weakly polar), acetonitrile (strongly polar), which lowers the $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$, facilitates TADF for o-3CzBN and m-3CzBN and improves the TADF efficiency for p-3CzBN and 4CzBN. Kinetically, these results can be explained by both an increase in $k_{\mathrm{RISC}}$ and a decrease in $k_{\mathrm{nr},\mathrm{T}}$. However, a strongly polar solvent is not always beneficial for TADF efficiency, as 5CzBN (and also 4CzIPN in Ref. 34) showed a larger decrease in TADF efficiency due to the consequent increase in $k_{\mathrm{nr},\mathrm{T}}$. As a result, the polar host environment has an advantage for $k_{\mathrm{RISC}}$, but the appropriate polarity is required for respective molecules to maximize the effect in TADF efficiency. Overall, since close proximity of the ${}^3\mathrm{LE}$ and ${}^1\mathrm{CT}$ states facilitates TADF, its efficiency can be improved by tuning both the $\mathrm{\Delta }{E}_{\mathrm{ST}}$ and $\mathrm{\Delta }{E}_{\mathrm{ST}(\mathrm{LE})}$, which are realized by a combination of (1) the design of the donor-accepter geometry (for ${}^1\mathrm{CT}$ and ${}^3\mathrm{CT}$) and (2) the selection of donor and accepter moieties (for ${}^3\mathrm{LE}$).