Thermally Activated Delayed Fluorescence (TADF) Path toward

Article Cite This: Acc. Chem. Res. XXXX, XXX, XXX−XXX

pubs.acs.org/accounts

Thermally Activated Delayed Fluorescence (TADF) Path toward Efficient Electroluminescence in Purely Organic Materials: Molecular Level Insight Xian-Kai Chen,† Dongwook Kim,‡ and Jean-Luc Brédas*,† †

Downloaded via UNIV OF THE WESTERN CAPE on August 25, 2018 at 01:32:23 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

Center for Organic Photonics and Electronics and School of Chemistry and Biochemistry, Georgia Institute of Technology, Atlanta, Georgia 30332-0400, United States ‡ Department of Chemistry, Kyonggi University, 154-42 Gwanggyosan-Ro, Yeongtong-Gu, Suwon 16227, Korea CONSPECTUS: Since the seminal work of Tang and Vanslyke in 1987 on small-molecule emitters and that of Friend and co-workers in 1990 on conjugated-polymer emitters, organic light-emitting diodes (OLEDs) have attracted much attention from academia as well as industry, as the OLED market is estimated to reach the $30 billion mark by the end of 2018. In these first-generation organic emitters, on the basis of simple spin statistics, electrical excitation resulted in the formation of ∼25% singlet excitons and ∼75% triplet excitons. Radiative decay of the singlet excitons to the singlet ground state leads to a prompt fluorescence emission, while the triplet excitons only lead to weak phosphorescence due to the very small spin−orbit couplings present in purely organic molecules. The consequence is a ca. 75% energy loss, which triggered wide-ranging efforts to try and harvest as many of the triplet excitons as possible. In 1998, Thompson, Forrest, and their co-workers reported second-generation OLED emitters based on coordination complexes with heavy transition metals (e.g., iridium or platinum). Here, the triplet excitons stimulate efficient and fast phosphorescence due to the strong spin−orbit couplings enabled by the heavy-metal atoms. Internal quantum efficiencies (IQE) up to 100% have been reported, which means that for every electron injected into the device, a photon is emitted. While these second-generation emitters are those mainly exploited in current OLED applications, there is strong impetus from both cost and environmental standpoints to find new ways of exploiting purely organic emitters, which in addition can offer greater flexibility to fine-tune the electronic and optical properties by exploiting the synthetic organic chemistry toolbox. In 2012, Adachi and co-workers introduced a promising strategy, based on thermally activated delayed fluorescence (TADF), to harvest the triplet excitons in purely organic molecular materials. These materials now represent the third generation of OLED emitters. Impressive photophysical properties and device performances have been reported, with internal quantum efficiencies also reaching nearly 100%. Our objectives in this Account are threefold: (i) to lay out a comprehensive description, at the molecular level, of the fundamental photophysical processes behind TADF emitters; (ii) to discuss some of the challenges facing the design of TADF emitters, such as the need to balance the efficiency of thermal activation of triplet excitons into the singlet manifold with the efficiency of radiative transition to the ground state; and (iii) to highlight briefly some of the recent molecular-design strategies that pave the way to new classes of TADF materials. delayed fluorescence.2 It has now been documented in several classes of organic molecules.3 In 2012, Adachi and co-workers proposed a simple molecular design principle to develop purely organic TADF emitters and successfully exploited such emitters in the active layers of organic light-emitting diodes (OLEDs).4 In section 2, we describe the RISC process from the T1 to the S1 state. The conversion between the singlet and triplet manifold depends on the S1−T1 energy gaps (ΔEST) and the spin−orbit S1T1 couplings (HSO ); indeed, in the framework of simple

1. INTRODUCTION Thermally activated delayed fluorescence (TADF) was first rationalized by Perrin in 1929.1 It is based on a thermally activated reverse (up-conversion) intersystem crossing (RISC) from the lowest triplet (T1) excited state to the lowest singlet (S1) excited state, see the energy diagram in Figure 1. The singlet excitons generated in this way from triplet excitons can then decay radiatively to the electronic ground state (S0). In the case of electroluminescence, they lead to a delayed fluorescence signal that complements the prompt fluorescence signal associated with the excitons directly generated into the singlet manifold. TADF in a purely organic molecule was first reported in the case of eosin and is also referred to as “E-type” © XXXX American Chemical Society

Received: April 19, 2018

A

DOI: 10.1021/acs.accounts.8b00174 Acc. Chem. Res. XXXX, XXX, XXX−XXX

Article

Accounts of Chemical Research

the LUMO, and KHL denotes the corresponding electron exchange energy: JHL =

∫ ∫ d r1⃗ d r2⃗ H( r1⃗)H( r1⃗) r1 L( r2⃗ )L( r2⃗ )

(5)

KHL =

∫ ∫ d r1⃗ d r2⃗ H( r1⃗)L( r1⃗) r1 H( r2⃗ )L( r2⃗ )

(6)

12

12

Importantly, the electron exchange energy depends on the spatial overlap (= ∫ dr⃗ |H(r⃗)||L(r⃗)|) between the HOMO and LUMO wave functions; the greater [smaller] this spatial overlap, the greater [smaller] KHL. Equations 2 and 3 show that, while the electron exchange energy destabilizes the S1 state, it stabilizes the T1 state. This is the reason why, in actual situations where (as will be illustrated below) the T1 states are not defined by just a single HOMO-to-LUMO electronic configuration, it has been shown that the triplet state tends to have a more confined, local-excitation (LE) character, while the singlet state displays more charge-transfer (CT) excitation character.9,10 It is useful to point out that, in order to describe reliably the electronic structure of extended π-conjugated systems, especially in instances where the electronic states can possess a strong CT character, it is necessary, as we detailed earlier,11 to set aside popular DFT functionals such as B3LYP and to exploit modern functionals such as long-range corrected functionals (e.g., ωB97XD or LC-ωPBE) as well as to include dispersion corrections.12,13 If the S1 and T1 states both strictly correspond to a HOMO → LUMO charge-transfer transition with the HOMO and LUMO fully spatially separated, eq 6 shows that the exchange energy goes to zero, ES1 and ET1 become equal, and ΔEST vanishes. In reality, however, the T1 states are calculated to often possess much more complex electronic-configuration descriptions, which require to go well beyond the simple HOMO−LUMO picture. It then becomes useful to rely on natural transition orbitals (NTOs); these provide a compact orbital representation of the electronic excitations, especially in the case where the relevant excited states correspond to a single pair (one for the hole, one for the electron) of NTO orbitals.14 Figure 2 illustrates the NTOs calculated for the S1, T1, and T2 states in two representative TADF emitters, 10-(4(4,6-diphenyl-1,3,5-triazin-2-yl)phenyl)-10H-phenoxazine (PXZ-TRZ) and 10-phenyl-10H-spiro[acridine-9,9-fluorene]2,7-dicarbonitrile (ACRFLCN). In PXZ-TRZ, the electron and hole NTOs describing the S1 and T1 states are strongly separated in space due to the nearly 90° dihedral angle between the phenoxazine and phenyltriazine segments. Both S1 and T1 states have large CTexcitation characteristics. As a consequence, the measured ΔEST of the PXZ-TRZ is as low as 0.06 eV.15 On the other hand, a number of theoretical investigations have demonstrated that local-excitation triplet states can be more stable than their CT-dominated counterparts,9,10 which is consistent with the fact that the electron exchange energy increases with the spatial overlap of the relevant wave functions. This is exemplified by the ACRFLCN emitter: while the S1 state maintains a substantial CT-excitation character, see Figure 2, the hole and electron NTOs describing the T1 state are both localized on the cyano-substituted fluorene unit. As a result, the ΔEST (ca. 0.24 eV) measured for ACRFLCN is much larger than that in PXZ-TRZ.16

Figure 1. Sketch of a typical energy diagram illustrating thermally activated delayed fluorescence. S0 denotes the electronic ground state; S1, the first singlet excited state; T1, the first triplet excited state; ΔEST, the energy gap between the S1 and T1 states; SOC, spin−orbit coupling; F, prompt fluorescence; DF, delayed fluorescence; and RISC, reverse intersystem crossing.

perturbation theory, the mixing coefficient cS1T1 between the S1 and T1 states can be expressed as cS1T1 ≡

S1T1 HSO ΔEST

(1)

We also show how the T1 → S1 conversion can be strongly impacted by molecular vibrations and vibronic coupling effects. In section 3, we discuss the efficiency of both prompt and delayed fluorescence processes, which depend on the strength of the radiative decay to the ground state. Finally, in section 4, we highlight some of the recent molecular design strategies that aim to expand the range of TADF organic materials.

2. REVERSE INTERSYSTEM CROSSING 2.1. Singlet and Triplet Excited States

The initial design of TADF emitters was based either on charge-transfer (CT)-type molecules, which consist of an electron-rich (donor, D) moiety and an electron-poor (acceptor, A) moiety and generally display a large dihedral angle between these moieties, or on D/A complexes.5−7 In such systems, the wave functions of the highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) can be expected to be spatially separated, with the former essentially localized on the D unit and the latter on the A unit. The original molecular-design strategy4 was based on the following principle: provided that the main electronic configuration describing both S1 and T1 states corresponds to an electronic transition from the D-localized HOMO to the A-localized LUMO, the electron exchange energy vanishes and ΔEST is very small. Indeed, in such a simple two-electron two-state model, the energy (ES1 [ET1]) of the S1 [T1] state and ΔEST can be written as8 ES1 = hH + hL + JHL + KHL

(2)

E T1 = hH + hL + JHL − KHL

(3)

ΔEST = 2KHL

(4)

Here, hH [hL] denotes the one-electron energy of the HOMO (H) [LUMO (L)] orbital, while JHL is the Coulomb repulsion energy between electron 1 on the HOMO and electron 2 on B

DOI: 10.1021/acs.accounts.8b00174 Acc. Chem. Res. XXXX, XXX, XXX−XXX

Article

Accounts of Chemical Research ∞

n 2π 1 ∑ e −S S |⟨S1|Ĥ SO|T1⟩|2 ℏ n! 4πλMkBT n = 0 É ÅÄÅ Ñ Å (ΔEST + λM + nℏωeff )2 ÑÑÑ ÑÑ expÅÅÅÅ− ÑÑ ÅÅÅ λ k T 4 ÑÑÖ (7) M B Ç where ⟨S1|Ĥ SO|T1⟩ denotes the spin−orbit coupling (SOC) matrix element between the S1 and T1 states, λM the Marcus reorganization energy related to intermolecular and intramolecular low-frequency vibrations, kB the Boltzmann constant, T temperature, ℏωeff the effective energy of a mode representing the relevant nonclassical high-frequency intramolecular vibrations (ℏωeff/(kBT) ≫ 1), and S the effective Huang−Rhys factor corresponding to this mode. Note that eq 7 is equivalent to the expression that has been extensively used to describe charge transport in organic semiconductors, with the difference that the electronic couplings between adjacent molecules are replaced with spin−orbit couplings.18 Equation 7 highlights that, in addition to the singlet−triplet energy gaps we have discussed so far, the spin−orbit couplings also play a critical role in determining the rates of the RISC processes from triplet to singlet manifold. In second-generation OLED coordination-complex emitters such as iridium-trisphenylpyridine, spin−orbit coupling is evaluated to be about 150 cm−1.19 Given the purely organic nature of the TADF emitters, the spin−orbit couplings there are calculated to be much weaker, commonly ≤1 cm−1.10 However, it must be borne in mind that, in spite of these smaller values, an increase in SOC by 1 order of magnitude results in an increase in RISC rate by 2 orders of magnitude. In twisted CT-type TADF emitters in which the S1 and T1 states both show a substantial CT-excitation character, the positive outcome is a small ΔEST; however, a drawback is that the spin−orbit coupling between these excited states tends to vanish.10 The main reason is that, within the one-electron approximation, the spin−orbit operator acts on both the spin magnetic quantum number of the electron and its spatial angular momentum quantum number. Consequently, spin− orbit couplings between singlet and triplet states with the same spatial orbital occupation are formally zero, as any change in spin cannot be compensated by a corresponding change in orbital angular momentum and thus the total angular momentum would not be conserved. For example, in the ACRFLCN molecule, see Figure 2, the S1 and T2 states have a predominant CT nature and their SOC is as small as 0.01 cm−1. On the other hand, T1 is predominantly a localexcitation state; the difference in the nature of the S1 and T1 excited states gives rise to a much more substantial SOC value of 0.46 cm−1. Interestingly, PXZ-TRZ exhibits a very significant SOC value of 1.54 cm−1 between the S1 and T2 states, which again comes from their marked differences in character, see Figure 2. Such a large spin−orbit coupling between the S1 and T2 states can also facilitate RISC processes in TADF materials, as discussed in the following section. In addition, several investigations have underlined that the introduction of atoms such as bromine can lead to an enhancement in the spin−orbit couplings between the S1 and T1 states, as expected from the heavy-atom effect.20 Overall, these results highlight the role of variations in the characteristics of the excited states involved in the RISC processes. We also note that the enhancement in spin−orbit coupling facilitates not only the reverse ISC from T1/T2 to S1, but also the forward ISC from S1 to T1/T2. If the rate of nonradiative

kRISC =

Figure 2. (a) Chemical structures of PXZ-TRZ and ACRFLCN. (b) Natural transition orbital analysis for the S1, T1, and T2 excited states in PXZ-TRZ and ACRFLCN, as calculated at the optimally tuned long-range corrected LC-ωPBE/6-31+G(d) level. The predominant hole (h) and electron (e) wave functions with the largest weight, v, are shown. Adapted with permission from ref 10. Copyright 2017 American Chemical Society.

At this stage, it is important to underline that solid-state polarization effects are expected to play a strong role in stabilizing the excited states with the largest CT character. For instance, in the 9H-thioxanthen-9-one-10,10-dioxide-triphenylamine emitter (TXO-TPA, see Figure 3), optimally tuned long-range corrected ωB97X calculations lead to a ΔEST value as large as 0.66 eV in vacuum (ε = 1); in such a case, the T1 → S1 reverse intersystem crossing would be negligible. 13 However, when the solid-state dielectric environment is implicitly taken into account via the polarizable continuum model (PCM), the calculated ΔEST comes down to 0.08−0.06 eV in the ε = 3−4.5 range, which is in close agreement with the experimental value (ca. 0.05 eV).17 While the solid-state polarization effects confer an increasing CT character to both S1 and T1, the intrinsically larger CT contribution to the S1 state makes its energy decrease more quickly and to a larger extent with an increase in dielectric constant, which acts to reduce ΔEST, as illustrated in Figure 3.13 2.2. Spin−Orbit Coupling

The rate of the thermally activated RISC process from T1 to S1 can be cast in the framework of Marcus electron-transfer theory. In the context of its Marcus−Levich−Jortner formulation, it can be expressed as10 C

DOI: 10.1021/acs.accounts.8b00174 Acc. Chem. Res. XXXX, XXX, XXX−XXX

Article

Accounts of Chemical Research

Figure 3. (a) Chemical structure of TXO-TPA. (b) Hole and electron NTOs in the S1 and T1 states for ε = 1 and ε = 3.0, as calculated at the PCM-tuned ωB97X/cc-pVDZ level. Superscripts “1” or “3” indicate a singlet or triplet state. (c) Calculated vertical excitation energies E0(S1), E0(T1), and ΔEST calculated for TXO-TPA as a function of dielectric constant. Adapted with permission from ref 13. Copyright 2017 American Chemical Society.

forward ISC is faster than the fluorescence rate, a strong forward ISC can become detrimental to the fluorescence quantum yield. Thus, the magnitude of the spin−orbit coupling must be such as to achieve the right balance among the fluorescence, forward ISC, and reverse ISC rates. 2.3. Dynamical Features of Reverse Intersystem Crossing

To gain an overall appreciation of the impact of both spin− orbit coupling and ΔEST, the T1 → S1 reverse intersystem crossing rates were evaluated as a function of these parameters, on the basis of eq 7 for two reasonable values of reorganization energy, see Figure 4.10 When ΔEST < 0.1 eV, the calculated RISC rates are fast, on the order of >105 s−1; this is consistent with the experimental data reported for efficient TADF emitters, for example, PXZ-TRZ (we note that a rate of 105 s−1 represents the longest time scale acceptable for most display applications). Figure 4 also indicates that when ΔEST ≥ 0.3 eV, the RISC rates are expected to be very small,