Recent progress in the understanding of exciton dynamics within phosphorescent OLEDs


  • Sebastian Reineke,

    Corresponding author
    1. Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA 02139, USA
    • Phone: +1-617-253-0085, Fax: +1-617-324-5275
    Search for more papers by this author
  • Marc A. Baldo

    Corresponding author
    1. Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA 02139, USA
    • Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA 02139, USA.
    Search for more papers by this author

  • This article will be included, in edited form, as a chapter of the forthcoming book “Physics of Organic Semiconductors”, edited by W. Brütting and C. Adachi (Wiley-VCH, Weinheim, 2012), ISBN 978-3-527-41053-8.


Excited states of organic molecules (excitons) are the heart of any organic electroluminescent device. They mediate the conversion of injected charges – electrons and holes – into photons. Phosphorescent emission originating from triplet excitons is especially important, as it is to date the only general route to enable unity charge-to-photon conversion efficiencies. In this paper, we discuss the key aspects of excitons, following the excited state lifecycle. First, we review fundamentals of singlet and triplet exciton formation in organic semiconductors, followed by a discussion of concepts that aim to alter the singlet-to-triplet formation rates to enable higher electroluminescence yields in the fluorescence manifold. Subsequently, we focus on the exciton distribution within the organic semiconductor material during its lifetime. The processes involved ultimately determine organic light-emitting diode (OLED) performance and are especially key in the development of concepts for white emission, where precise balance of the exciton between different emitter species control the emitted color. We close this paper with discussion of non-linear effects at high excitation levels that, to date, limit the high brightness efficiency of phosphorescent OLEDs.

1 Introduction

Phosphorescence is now widely adopted in commercial organic light-emitting devices (OLEDs). It has reached its position as the premier high efficiency technology after a long development effort. Despite this, many of the underlying mechanisms remain unclear, attracting continuing research interest and the attention of regular scientific reviews 1–3. Here, we focus on recent developments in three areas of particular relevance to the operation of phosphorescent OLEDs. Following the lifecycle of the excited states, we first review advances in the understanding of exciton formation, with a particular focus on emerging competitors to phosphorescence. Then we review progress in the understanding and control of exciton transport and harvesting, before concluding with a summary of current knowledge about exciton quenching.

2 Exciton formation

2.1 Background

In an OLED, the conversion of electrical energy into light is mediated by excitons. It is the properties of the excitons that primarily determine the overall luminescent efficiency of the device.

An exciton in organic materials may be thought of as a two particle system: one is an electron excited into a unfilled higher energy orbital of a molecule or polymer while the second is a hole created in the ground state due to the excitation of the electron.

The exciton formation process in OLEDs begins with electrons and holes injected at the electrodes. As the charges drift together under the influence of an electric field, there are two principal interactions between them: Coulombic and Exchange. At separations of a few nanometers, the Coulombic attraction binds the charges together and they are no longer able to escape one another by thermal diffusion. Here, compared to inorganic semiconductors, the attraction is much stronger, enabling instantaneous exciton formation, because the dielectric constant of organic/polymer materials with εr ∼ 3 is much smaller (equation image, cf. e.g. εr,Si = 11.68). Typically, this capture occurs before the electron and hole form an exciton on a single molecule or polymer chain. For example, Fig. 1 shows a an electron and a hole on two adjacent molecules of the archetype OLED material tris(8-hydroxyquinoline) aluminum (Alq3) 4. This combination of an electron and hole is also bound and forms what is known as a charge transfer (CT) state.

Figure 1.

(online color at: The calculated charge density difference between the singlet and triplet CT states, and the ground state, in a δ-Alq3 dimer. The right (left) hand molecule is constrained to be negatively (positively) charged and gold (purple) surfaces enclose volumes where the CT state has more (fewer) electrons. Analogous plots for the triplet CT state are visually indistinguishable from the above plot. Reprinted from Ref. 4.

Exchange effects are weaker than Coulombic interactions but they have a profound effect on the operation of an OLED. The total spin of the two electron system may be either S = 0, or S = 1. The S = 0 state is antisymmetric under particle exchange, but the S = 1 contains three possible states, all symmetric under particle exchange. The degeneracies of each state are reflected in their titles, the S = 0 state is known as a singlet, and the S = 1 is a triplet. Crucially, the opposite sign of the exchange interaction in singlet and triplet states creates an energetic difference. The exchange splitting between singlets and triplets is usually substantial in excitons (S > 0.1 eV) 5. But exchange interactions are more controversial in the CT states. Measurements suggest appreciable splittings 4, 6, but theory suggests that the exchange interaction may also be very weak 7. Also, the ordering of the CT states is debated 6–8; in some cases the singlet CT state is predicted to have the lowest energy 8.

In addition to its effect on exchange interactions, the spin of the exciton controls its luminescence. The ground state of most molecules is a singlet state, and because the emission of a photon conserves spin, typically only singlet excited states can emit light. Radiation of singlet excitons is fast, often efficient, and is known as fluorescence. The probability of luminescence from the remaining triplet states is generally so low that almost all their energy is lost to non-radiative processes.

A model of exciton formation 3, 9 is summarized in Fig. 2. Excitons are formed from CT states, which in turn are formed from injected electrons and holes. Most importantly, the model predicts that in the absence of a process that mixes the singlet and triplet channels, only one quarter of the injected charges will form luminescent (singlet) excitons, irrespective of the relative formation rates of singlets and triplets.

Figure 2.

(online color at: A model of exciton formation 9. The formation rates of triplet and singlet excitons are kS and kT, respectively. The mixing rate at room temperature is kmix. G is a constant determined by the formation rate of CT states. η is the maximum quantum efficiency of light emission.

2.2 Spin mixing for higher efficiency

There are two opportunities for spin mixing in OLEDs: the exciton and the CT state. Below, we briefly summarize two methods of exciton mixing, phosphorescence and thermally activated delayed fluorescence (TADF), and we also review progress toward higher efficiencies by exploiting CT state mixing.

2.2.1 Exciton mixing and phosphorescence

It is well known now that heavy atoms such as platinum or iridium can enhance spin–orbit interactions within a molecule or polymer 10–12. This in turn can mix the excited singlet and triplet states such that the triplet gains some singlet character and the decay of the triplet state is partially allowed (cf. Fig. 2b). The emission of light is still significantly slower than fluorescence, but if singlet–triplet mixing in the exciton state yields a radiative decay rate that is faster than the non-radiative rate then the luminescence can be efficient. This emission of light from a disallowed transition is known as phosphorescence. We want to note here that, despite the extensive experimental validation of this phenomenon, it is still a challenge to correctly treat these organometallic complexes theoretically in order to explain the transitions in the triplet manifold. Recently, novel descriptions have been introduced 13 and tested against experimental data 14.

The exciton mixing required for efficient phosphorescence enhances intersystem crossing (ISC) from the singlet exciton to the lower energy triplet exciton. Thus, phosphorescence can harness 100% of the injected carriers 15. Also, by making the lowest energy state in the system emissive, phosphorescence is naturally robust to the presence of mixing anywhere in the formation path. This is not true for enhanced fluorescence following CT state mixing, where exciton mixing can generate non-emissive triplet excitons. Next we describe alternatives before returning to summarize the technological prospects of phosphorescence.

2.2.2 CT state mixing and enhanced fluorescence

If the emissive material does not possess sufficient exciton mixing to enable strong phosphorescence, it is still theoretically possible to obtain light from 100% of the injected charges. The mixing process can instead be performed in the CT state. Including CT mixing yields a maximum quantum efficiency (photons/electrons) of Ref. 7:

equation image((1))

There are two important parameters: the rate of singlet–triplet CT state mixing, kmix, and the ratio of exciton formation rates, kS/kT. If equation image then η becomes a function of the exciton formation rates, i.e., equation image. These rates have been calculated by numerous groups. While the predictions are highly material dependent 16, kS is generally expected to be faster than kT because the energetic separation between the CT state and singlet exciton is smaller than the gap between the CT state and the triplet exciton. Consequently, singlet exciton formation is less Marcus inverted than triplet exciton formation 5.

The general expectation that kS/kT > 1 suggests that η > 1/4 in many fluorescent OLEDs. These predictions, however, assumed that the mixing rate is fast (equation image). The rate of singlet–triplet CT mixing is governed by the strength of the coupling between singlet and triplet CT states, which is typically less than 10−4–10−5 eV in purely organic systems due to weak spin–orbit and spin–lattice interactions 7.

Thermally stimulated luminescence measurements by Kadashchuk et al. 17 have demonstrated that CT state mixing does occur over long time scales (seconds) 6. Perhaps the first measurement of the CT mixing rate at shorter time scales was performed by Reufer et al. 18. In that work, excitons in the prototypical conjugated polymer ladder-type poly(para-phenylene) (PhLPPP) were generated by an optical pulse, then separated into charges using an electric field pulse, before finally being allowed to reform and emit 18. The singlet exciton population is monitored by the fluorescence 18. But very weak Pd doping in the polymer enhances phosphorescence and also allows the triplet exciton population to be measured 18.

The data in Fig. 3 shows a decrease in both fluorescence and phosphorescence during the electrical pulse 18. Then both the singlet and triplet channels are observed to rebound as the charges separated by the field recombine. Crucially, the inset shows that the phosphorescent rebound scales approximately linearly with the length of the electrical pulse. This, together with the overall similarity of both singlet and triplet channel dynamics, confirms that there is no appreciable mixing of the CT states prior to recombination.

Figure 3.

Field modulated recombination of optically generated singlet and triplet charge carrier pairs; (a) at 4 K, (b) at 200 K (electric field: ∼1 MV cm−1; duration marked by vertical dashed lines). Dotted lines indicate the PL dynamics in the absence of a field. All spectra were recorded in 200-ns gate windows. The inset in (b) shows the effect of pulse length τ on the triplet overshoot post turn-off for 3 µs (circles), 7 µs (squares), 11 µs (triangles) long pulses. Reprinted from Ref. 18.

Whereas Reufer et al. found mixing rates slower than 105 s−1 in PhLPPP 18, a later study by Ford et al. established a minimum rate of 2 × 106 s−1 in blends of poly(9,9-dioctylfluorene-co-benzothiadiazole) (F8BT) with poly(9,9-dioctylfluorene-co-bis-N,N′-(4-butylphenyl)-bis-N,N′-phenyl-1,4-phenylene-diamine) (PFB) 19. Although the latter measurement may be consistent with much faster mixing rates, if the results of Reufer et al. hold generally, then fluorescent materials do not possess sufficient mixing in their CT states to achieve η > 1/4. As discussed previously 3, 16, direct measurements of η are contradictory. Some measurements suggest that η > 1/4 and others that η = 1/4 3, 16. Studies have also suggested that η > 1/4 in fluorescent polymers but not small molecules 20. Shuai et al., based on a molecular orbital perturbation approach, calculated the formation cross section for singlets and triplets in the polymer polyparaphenylene vinylene (PPV), where they found that the interchain bond-charge correlation has a strong influence 21. They report that generally η > 1/4, i.e., the spin-degeneracy limit can be surpassed in polymeric systems.

Unfortunately, no one measurement technique appears unimpeachable, with comparisons between electroluminescence and photoluminescence prone to underestimation of the photoluminescent efficiency because 70–80% of the emitted light is waveguided and subject to self-absorption 22 and OLED-based measurements subject to the uncertainty of charge quenching and exciton formation in traps 16. The quantum efficiency of fluorescent OLEDs is the ultimate measure, and to date, it has remained substantially below that of phosphorescent OLEDs.

The apparent absence of sufficient CT mixing in some fluorescent materials suggests that to enhance fluorescence, an OLED could be engineered to specifically mix CT states but not excitons 4. It is important not to mix the exciton state because that results in triplet exciton formation, which lowers the fluorescent efficiency. Figure 4a shows an OLED with a selective CT mixing layer (X-OLED). As in other heterostructure OLEDs, excitons are formed on the lower energy side of the interface between the hole transport layer (HTL) and the electron transport layer (ETL). The emissive material is the red fluorophore 4-(dicyanomethylene)-2-methyl-6-julolidyl-9-enyl-4H-pyran (DCM2). It is inserted between the HTL/ETL interface, doped into a wide band-gap host material, in a narrow layer just 50 Å thick, to minimize the possibility of efficiency artifacts caused by shifts in the exciton formation zone 23.

Figure 4.

(online color at: Performance of the enhanced fluorescent OLED to the control device. The EQE of the OLED with mixing reaches a maximum of 3.4%, or 2.8 times larger than the control. An OLED identical to the X-OLED, but with the FIrpic layer spaced from DCM2 by 100 Å of BCP, does not show enhanced efficiency. This is consistent with the enhancement in fluorescence resulting from spin mixing at the exciton formation interface. Reprinted from Ref. 4.

To obtain CT state mixing and enhanced fluorescence (cf. Fig. 2c), the ETL consists of a thin film of iridium(III) bis[(4,6-difluorophenyl) pyridinato-N,C20] picolinate (FIrpic) 24, followed by an additional conventional ETL (BCP). The presence of iridium in FIrpic enhances spin–orbit coupling and mixes the spin state of the electron it carries. The spin of the CT state consisting of an electron on FIrpic and a hole on DCM2 is therefore also mixed. Further, calculations similar to those described above give a singlet–triplet CT gap of 60 meV for a FIrpic−/DCM2+ heterodimer 4. Thus the interfacial CT states should be appreciably split. FIrpic will not quench DCM2, as FIrpic phosphoresces in the blue–green. It is employed here, however, purely as an ETL. Indeed, its electroluminescent quantum efficiency is only 0.2% in a neat film. In addition, FIrpic's spin mixing effect on neighboring molecules is reduced by the bulky side groups which surround its central heavy metal atom, reducing ISC effects in DCM2. In a control device, the FIrpic ETL is replaced by an ETL with low spin–orbit coupling: 2,9-dimethyl-4,7-diphenyl-1,10-phenanthroline (BCP) 16. Figure 4b shows the external quantum efficiency (EQE) of control- and X-OLED together with the corresponding improvement obtained by employing CT-mixing.

Based on this study, CT mixing could be used to enhance the efficiency of OLEDs without appreciably lengthening the excited state lifetime. Much work remains, however, to extend the approach to other fluorescent emitters and demonstrate an improvement at high brightness.

2.2.3 Thermally activated delayed fluorescence (TADF)

Another route to enhance the overall luminescence efficiency from fluorescent materials makes use of excitons that are mixed back to the fluorescent channel from a non-radiative reservoir in the molecular triplet manifold by means of thermal activation (cf. Fig. 2d) 25. For this mixing process to be efficient, the energetic splitting of singlet and triplet states (ΔEST) must be comparable to kBT. The first observation of TADF in OLEDs was reported by Endo et al. 26 discussing various tin(IV) fluoride-porphyrin complexes, which posses energetic splittings of approximately 0.4 eV, which still is too large for efficient reverse ISC from the triplet back to the singlet.

From the quantum mechanical viewpoint, the energy difference ΔEST is proportional to the exchange integral K between spatial overlap of the highest occupied and lowest unoccupied molecular orbital. Thus, the splitting can be reduced by minimizing K (it is worth noting that fullerenes also show TADF 25, because of moderate ΔEST, which is caused by the molecules symmetry rather that by a small K). The same group (Endo et al.) succeeded in developing a hybrid molecule 2-biphenyl-4,6-bis(12-phenylindolo[2,3-a] carbazole-11-yl)-1,3,5-triazine (PIC-TRZ) as shown in Fig. 5a containing donor (indolocarbazole) and acceptor (triazine) units, respectively, resulting in vanishing spatial overlap [cf. Fig. 5a)] and strongly reduced splitting of ΔEST = 0.11 eV [cf. Fig. 5b] 27. This leads to a very high reverse ISC efficiency of 29%. Figure 5c gives clear evidence for TADF showing that both the prompt and the delayed emission coincide. Further promising results about TADF-based electroluminescence have been discussed by Deaton et al. for a bis(phosphine)diarylamido dinuclear copper(I) complex with ΔEST = 0.10 eV 29. At low excitation levels, OLEDs based on this copper complex reach an EQE of 16%, clearly outperforming the 5% limit for conventional fluorescent materials. Very recently, Goushi et al. reported on TADF in a system that is designed to emit from the exciplex state at a hetero-interface 30. Here, a very high reverse ISC efficiency of 86.5% is realized, boosting the electroluminescence efficiency.

Figure 5.

(online color at: (a) Molecular structure and calculated HOMO and LUMO orbital distributions as obtained from a Gaussian 03 (B3LYP/cc-pVDZ) calculation of PIC-TRZ. (b) Photo-physical properties of PIC-TRZ: fluorescence and phosphorescence are obtained at 5 K for a 6 wt% mixed film of 1,3-bis(9-carbazolyl)benzene:PIC-TRZ. (c) Streak-camera image of the samples. Reprinted from Ref. 27.

The triplet reservoir that feeds the luminescent singlet state is the major source of loss in this concept because triplet excitons pile up if the thermally activated ISC is slow. Thus, triplet–triplet 31 and triplet-polaron 28 annihilation become severe problems that strongly reduce the device efficiency at high excitation levels. All reports on TADF to date exhibit a pronounced roll-off in efficiency with increasing current density. For instance, Fig. 6 shows the quantum efficiency versus current density of a device based on the copper complex of Deaton et al. dropping to half of its initial value at only a few mA cm−2 29. The data of Fig. 6b show the drastic impact of triplet-polaron quenching. In contrast, a similar drop in efficiency is observed at roughly two orders of magnitude higher current density for phosphorescent devices 32, 33.

Figure 6.

(online color at: (a) EQE and (b) photoluminescence spectra, obtained on a hole-only single carrier device, both as a function of current density of a TADF device based on the copper complex of Deaton et al. The initial EQE drops to half of its value at only a few mA cm−2. The data in (b) clearly indicates triplet-polaron quenching 28. Reprinted from Ref. 29.

2.2.4 Summary: Comparison between phosphorescence, extrafluorescence, and TADF

Although extrafluorescence from CT state mixing and TADF have shown promise, it seems unlikely that either approach will soon replace phosphorescence in mainstream applications. The challenge in extrafluorescence is to design systems where the CT state mixing competes with rapid exciton formation rates while also not mixing the exciton. TADF will require smaller singlet–triplet splittings to increase the ISC rate and reduce the density of triplet excitons. Thus, phosphorescence appears likely to remain the preferred technology in high efficiency OLEDs. Its only important weaknesses are the relatively long lifetime of the excited state (see Section 4), which enhances quenching processes and a reliance at present on expensive metals such as iridium or platinum to perform the spin mixing.

3 Distributing excitons in the organic layer(s)

Once an exciton is formed in the organic semiconductor within an OLED, it is a challenge to efficiently direct it to the desired emissive state. Here, we will address key requirements for efficient phosphorescence and point out strategies to exploit the nature of triplet excitons to manage the exciton distribution, especially for white light-emitting devices.

3.1 Excitonic confinement: Host–guest systems

Many phosphorescent emitters exhibit a noticeable reduction in luminescence quantum yield for bulk layers 11, 34. On the other hand, strongly improved quantum yields are obtained whenever the phosphor is diluted into another material – the host material (cf. Fig. 7). The phosphorescent dopant is accordingly referred to as guest material. This effect is called concentration or aggregation quenching 35, 36. High quantum yields are typically achieved with a concentration of the phosphor in the host–guest system in the range of 1–10 wt%.

Figure 7.

(online color at: Concentration dependence of three different phosphorescent emitters Ir(ppy)3, BtpIr2(acac), and FIrpic as dispersed into a proper host material. For the blue emitter FIrpic, data is shown for an endothermic and an exothermic system, respectively. Reprinted from Ref. 34.

The host material must be selected carefully with respect to its and the phosphors energy levels. It is a prerequisite for efficient emission that the excitation created on host sites be transferred to the phosphor, unless emission from the host is desired (cf. Section 3.4). Here, singlet excitons are mainly transferred via dipole–dipole energy transfer (Förster mechanism 37) while triplet excitons migrate from site to site, ultimately reaching a phosphorescent molecule, based on Dexter-type 38 transfer steps (hopping). Once the excitation is transferred to guest sites, it is necessary to confine the excitation on the phosphor so that radiative recombination can occur efficiently 32, 39. In case the phosphorescent emitter is in its lowest T1 state, migration to other molecules proceeds via highly efficient Dexter-type triplet–triplet energy exchange steps. Therefore, the relative position of the matrix triplet level with respect to the phosphor T1 state is important. The highest phosphorescence quantum yield of the mixed system can be expected for an exothermic system, where the triplet level of the matrix is higher in energy compared to the phosphor. By reducing the energy of the host triplet level to a resonant or endothermic system, respectively, the phosphorescent quantum yield of the mixed film will decrease accordingly as more excitations remain on host sites 40.

3.2 Exciton generation zone

The way excitons are spatially distributed within the device is vastly determined by the electrical (transport) properties of the various materials used. And, equally true for fluorescent and phosphorescent OLEDs, excitons are typically formed in a thin slab of an organic layer in the proximity of an interface to an adjacent material layer. This is because truly ambipolar materials, which transport electrons and holes equally well, are rare. Furthermore, both carrier species often need to pass different energy barriers on their way to the emission layer. Typically, the width of the generation zone is below 5 nm 28, and even its approximation as a delta-distribution often holds 41. Especially for multi-color devices, as adding thickness or layers to the device will not automatically broaden the exciton generation zone, it remains a key challenge addressing more spatially displaced emissive states.

One way of increasing the width of exciton generation zone is to bring two host materials together in a way that one is predominately hole-, the other electron-conductive, thus forming a double emission layer (DEML). Here, excitons are formed on either side of the interface between these two layers. This concept has successfully been used for demonstrating high efficiency monochrome 43 and white 40 phosphorescent devices. The exciton generation zone can also be broadened by mixing materials with altering transport properties to form a blend with ambipolar character 44. Ultimately, inherently ambipolar materials are desired which spatially distribute charges efficiently, enabling exciton formation throughout the layer. Currently, much synthetic effort is spent on developing these materials.

3.3 Exciton migration

Owing to the relatively long excited state lifetime of phosphors – typically in the range of microseconds – the exciton is likely to be transferred to other energetically accessible states. Because the donor (D) is an excited phosphor in its triplet state, it can transfer its energy efficiently to the singlet or triplet state of an acceptor (A) molecule via Förster-type 37,

equation image((2))

or Dexter-type 38,

equation image((3))

energy transfer, respectively. Here, the asterisks indicate excited states. The latter triplet–triplet transfer requires two simultaneous spin-flips, which are only possible incorporating exchange interactions. Energy transfer between one species of phosphorescent molecules should solely follow Dexter transfer-mediated hopping steps, because the fluorescent state of the identical acceptor it too high in energy. In contrast, however, Kawamura, et al. investigated the intra-species energy-transfer (triplet–triplet transfer) of the phosphorescent systems shown in Fig. 7 based on Försters theory with good agreement 35. For these emitters, they derived corresponding Förster radii of 1.4, 0.8, and 1.1 nm for fac-tris(2-phenylpyridine) iridium [Ir(ppy)3], bis[2-(2-benzothienyl)pyridinato-N,C3]acetylacetonato)iridium(III) [BtpIr2(acac)], and FIrpic, respectively 35. It is argued that multiple dipole–dipole transfers that lead to a dampening of energy are the cause of the observed concentration quenching (reduction of PL efficiency with increasing phosphor content cf. Fig. 7). This interpretation is in contrast to the publication of Kobayashi et al. where the authors introduce a non-emissive state 121 meV above the lowest Ir(ppy)33MLCT (metal-to-ligand CT) state with much higher decay rates compared to the triplet sublevels to explain the concentration quenching 36. Even so results published to date cannot establish a secure interpretation of the mechanism behind the concentration dependence of phosphors, the above stated results of Kawamura et al. clearly indicate that the energy exchange between phosphorescent molecules occurs efficiently up to distances of 2 nm. This coincides with the limit of the interaction range of the Dexter-type hopping steps.

The consequence of this discussion for energy migration is simple. While phosphors that differ in their emission wavelength will undergo cascade energy transfer from high to low energy sites with efficiencies approaching unity when in close proximity to each other, spatial distances down to 2 nm can decouple these excited states. Figure 8 shows a set of samples comprising three phosphors, i.e., FIrpic, Ir(ppy)3, and Ir(MDQ)2(acac) [iridium(III)bis(2-methyldibenzo[f,h]quinoxaline) (acetylacetonate)]. Each emitter is doped with 10 wt% into a wide-gap material 4,4,4-tris(N-carbazolyl)-triphenylamine (TCTA) having a T1 level of 2.83 eV above the respective emitter levels, forming a 1 nm thin layer. The different slabs of emitter doped films are intermitted by TCTA interlayers (see Fig. 8 for the sample layout). Thus, in addition to a spatial separation, the TCTA interlayer is used as an energetic barrier for triplet exciton movement based on Dexter-type hopping steps. Figure 8 shows that even a 2 nm spacing of different emitters can maintain strong emission from all emitters. In contrast, without the interlayers the emission from the high energy phosphors [FIrpic and Ir(ppy)3] is fully quenched, resulting in solely red emission.

Figure 8.

(online color at: Photoluminescence spectra of an emission layer consisting of multiple thin phosphor-doped layers (1 nm and 10 wt% each) that are intermitted by intrinsic layers of the host material TCTA of various thickness (2, 3, or 4 nm). The samples have the following general sequence (with R = Ir(MDQ)2(acac), G = Ir(ppy)3, B = FIrpic, and I = TCTA): TCTA:R/I (x nm)/TCTA:G/I (x nm)/TCTA:B/I (x nm)/TCTA:B/I (x nm)/TCTA:G/I (x nm)/TCTA:R. Adapted from Ref. 42.

These finding go hand in hand with known concepts for white light-emitting devices incorporating phosphors. D'Andrade et al. made use of the same properties dispersing three different emitters into one wide-gap material at different concentration varying from 0.5 to 20 wt% which effectively assures inter-species spacing 46. Another way is to unite the concept of the DEML with intrinsic interlayers to suppress complete energy transfer to low energy emitters 40.

3.4 Triplet harvesting

The combination of a fluorescent blue emitter with phosphorescent green and red phosphors to form a hybrid emission layer is attracting much interest because it provides an alternative to the use of blue phosphors. The latter are the current bottleneck in white devices because they noticeably increase the operating voltage and remain chemically unstable. Sun et al. first proposed white OLEDs with hybrid fluorescent/phosphorescent emission layers to efficiently utilize both singlets and triplets 47. However, the authors used a fluorescent emitter with a triplet level below the corresponding energies of the phosphors 48, 49. Thus, the blue fluorophore still remains a potential triplet exciton trap within the system. In order to avoid exciton losses in the triplet manifold of the blue fluorophore, special emitters have been introduced possessing a small singlet–triplet splitting 49–51.

The hybrid fluorescence–phosphorescence concept dates to the first paper on phosphorescent OLEDs, where it was employed to demonstrate different singlet and triplet exciton diffusion length and triplet exciton harvesting by phosphors 10. For N,N-di-1-naphthalenyl-N,N-diphenyl-[1,1:4,1:4,1-quaterphenyl]-4,4-diamine (4P-NPD), the first material known to be used for white OLEDs in this approach, the triplet diffusion length in the bulk material has been determined to 11 nm 41. Figure 9a illustrates the working principle of triplet harvesting: the exciton generation zone is fixed to one interface within the blue bulk fluorophore. This is followed by host–guest system formed by the blue fluorescent and a red phosphorescent emitter. By increasing the thickness of the intrinsic blue layer it also acts as a spacer that decouples the red phosphor from the site of exciton generation. While singlet excitons, as a result of their short lifetime, recombine at the position of the generation interface, the triplet excitons diffuse further into the spacer layer ultimately being harnessed by the phosphor. This effect can be seen in Fig. 9b 45: for a large thickness of the blue spacer (30 nm), almost only blue fluorescence is observed. By reducing the spacing, the contribution of the red phosphor increases noticeably while the intensity of the fluorescence remains constant. This is a clear indication that the phosphorescent material diffusively harvests triplets of the blue fluorophore. Finally with no spacer, solely red phosphorescence is observed, indicating that singlet excitons are also transferred to the phosphor.

Figure 9.

(online color at: (a) Diagram illustrating the working principle of triplet harvesting. Formed at a defined interface within the device, triplet excitons will diffuse further into the adjacent emission layer compared to singlets. A lower energy phosphor is harvesting the triplet excitons. (b) Electroluminescence of a set of OLEDs varying in the blue fluorescent spacer layer (4P-NPD) thickness. Adapted from Ref. 45.

4 High brightness effects in phosphorescent devices

Phosphorescent OLEDs pay for their fourfold increase in internal quantum efficiency with at least one order of magnitude longer excited state lifetime compared to conventional fluorescent dyes. Thus, these excitons are highly exposed to non-linear quenching processes that limit the inherently high device efficiency at high excitation levels. Generally, three processes account for triplet exciton quenching: (i) triplet-polaron quenching, (ii) electric field-induced exciton dissociation, and (iii) triplet–triplet annihilation (TTA) 52, 53. Among those, TTA is the only process that scales with the square of the exciton density n, for which it dominates the decrease in efficiency at high exciton densities 28. It is worth noting that field-induced dissociation, as suggested by Kalinowski et al. to be the dominant effect 53, is not observed for state-of-the-art OLEDs 28. Furthermore, the overall effect of triplet-polaron quenching on the efficiency roll-off easily can vary over orders of magnitude as it strongly depends on the device charge carrier balance 28.

In the following discussion, we solely discuss the dynamics of guest triplet excitons. This is valid because the systems under study are designed to have efficient host–guest energy transfer 34, 54 and are confining the excitons to the guest molecules using high triplet energy host materials 32, 39 (cf. Section 3). Generally, the dynamics of the guest triplet exciton density n(t) upon short, pulsed excitation at t = 0 is governed by the following rate equation:

equation image((4))

where f accounts for the number of triplet states that are deactivated. Consequently, it can have values of 1/2 or 1. Typically, f = 1/2 is assumed, i.e., the acceptor state will ultimately keep its excited configuration. The first term represents the monomolecular deactivation of the triplet exciton density n that is inversely proportional to the excited state lifetime τ. Furthermore, kT(t) describes the TTA rate. With the boundary condition equation image and the subsequently introduced TTA rate constants, the above equation can be solved to describe the time evolution of equation image.

TTA in solid mixed films can have different underlying mechanisms 42. One of them is a single-step long-range interaction (dipole–dipole coupling), based on Förster-type energy transfer (cf. Fig. 10a). It marks the limit for every phosphorescent mixed system as it solely depends on the spectral properties of the emitting guest. Interestingly, Staroske et al. recently introduced this mechanism as the intrinsic limit for any phosphorescent system 54. TTA mediated by Förster-type energy transfer is possible if the donor is a phosphorescent material, because then transition moment for T1 → S0 is non-zero. Moreover, the transition of the acceptor molecule, i.e., T1 → Tn, lies within the triplet manifold and is therefore an allowed transition. According to Försters theory of energy transfer 37, the rate of TTA is proportional to the spectral overlap of phosphorescent emission of the donor and the excited triplet state absorption of the acceptor. Within the Förster framework, the rate constant kT,F(t) can be approximated to 55:

equation image((5))

with the Förster radius RF for the transfer to an excited guest molecule. Note that there is no regime for which the transfer rate is time independent. Furthermore, this model suggests that TTA does not depend on the dopant concentration and that TTA breaks down if the intermolecular spacing between the phosphorescent molecules exceeds RF.

Figure 10.

(online color at: Scheme illustrating TTA based on (a) single-step long-range energy transfer for an exothermic host–guest system. In the scenario (b), an additional TTA channel is shown mediated by hopping-assisted migration of triplet excitons in clusters of guest molecules. Adapted from Ref. 42.

If the host–guest system is exothermic, i.e., the triplet level of the host is higher than the one of the guest, the single-step long-range mechanism should be the only channel for TTA for typical guest concentrations ranging from 1 to 10 mol% – otherwise exciton confinement is not given, leading to host–guest and host–host TTA contributions 56, 57. In their analysis, Staroske et al. pointed out that the phosphorescent systems comprising the archetype emitter Ir(ppy)3 suffer from stronger TTA than predicted by their model. They suggested that efficient triplet migration, taking place in locally dense clusters of guest molecules by means of exciton hopping (Dexter-type energy transfer), leads to the unexpected impact of TTA (cf. Fig. 10b) 54.

The corresponding TTA rate constant for hopping assisted exciton motion reads 58:

equation image((6))

with the interaction distance equation image and the diffusion constant D. For times much greater than the nearest neighbor hopping time equation image, this rate constant can be approximated to an time-independent form: equation image 55. Note that TTA within this model is – in contrast to the single-step long-range model – concentration dependent, which is accounted for in the diffusion constant D 42.

In the following, experimental evidence will be provided that supports the picture of guest aggregates in the mixed film that ultimately leads to strongly enhanced TTA. Time resolved spectroscopy is commonly used to investigate TTA. Here, the sample of interest is excited with a short laser pulse (337 nm) and the luminescence response of the sample is resolved as a function of time. When visualized in semi-logarithmic plot, TTA leads to a curvature in the transient signal directly after the excitation pulse, which will transform into a monoexponential decay at longer times/lower excitation levels (cf. Fig. 11c). Now solely from analyzing the decay curve it is not possible to distinguish which mechanism (single-step long range and/or hopping assisted) contributes to TTA. Thus, indirect methods need to be applied to differentiate between those two.

Figure 11.

(online color at: (a) Guest triplet exciton density versus pump exciton density (initial singlet states formed on the matrix material) for three different Ir(ppy)3 concentrations in the matrix TCTA. (b) Transient signal of the phosphorescence of the highly diluted sample at highest triplet density in the linear regime, as indicated in (a). Additionally shown are calculated fits according to Eqs. (4) and (5). (c) Decay curves of two samples with 9.3 and 1.1 mol% guest concentration, respectively (split in time for clarity). Fits according to the single-step long-range model [cf. Eqs. (4) and (5)] are plotted for various Förster-radii. Adapted from Ref. 42.

For this purpose, three mixed films of TCTA:Ir(ppy)3 were prepared with a thickness of 20 nm and varying guest concentrations from 0.1 to roughly 10 mol% 59. The highly diluted sample is used to assure spatial separation of the guest molecules (taking into account their expected tendency to aggregate). In Fig. 11a, the green box indicates the triplet exciton densities that can be created in the system without seeing a TTA signature in the transient signal, i.e., the decay curve is fully monoexponential. Now, the monoexponential data with highest triplet density can be used to determine an upper limit for the Förster radius RF. Based on Eqs. (4) and (5), calculated fits can be used to approach the monoexponential decay shown in Fig. 11b from large RF (with stronger TTA impact) to small values. Within experimental error, 6 nm was found to be an upper limit, while ≤3 nm showed best agreement with this set of data 59. In comparison, based on photophysical properties of Ir(ppy)3, Staroske et al. estimated the Förster radius for this TTA process to be 2.9 nm 54.

Figure 11c plots decay curves of the two samples with higher guest concentration, of which the 9.3 mol% sample is closest to application in real devices 28, 52. The red solid lines correspond to fit calculations on the single-step long-range model yielding radii of 9 and 10.5 nm for 1.1 and 9.3 mol% of guest concentration, respectively. Two important facts prove that an additional TTA channel in this system must be present: (i) In contrast to model predictions, these fits show concentration dependence (because RF is increased from 9 to 10.5 nm). Note that this is a noticeable difference as the TTA rate constant kT,F is proportional to the third power of the Förster radius RF. (ii) The derived values by far exceed both the upper limit and the estimated value of 2.9 nm. To illustrate this discrepancy, Fig. 12 plots the EQE of a standard OLED structure comprising this TCTA:Ir(ppy)3 emitter system. Going to higher brightness, the efficiency strongly decreases. However, using the data of the decay curves from Fig. 11c and a simple expression, equation image, to relate the exciton density n to current density j (or luminance) 60, it is possible to determine the impact of Förster-based TTA. Here, ν is the probability of exciton formation, e the elementary charge, and w the width of exciton generation zone. Assuming 3 nm as the Förster radius, which yields almost perfect monoexponential curves in Fig. 11c, the EQE of this TCTA:Ir(ppy)3 based OLED should remain constant up to an approximate device brightness of 30 000 cd m−2, as indicated by the green solid line.

Figure 12.

(online color at: EQE of a state-of-the-art green phosphorescent OLED comprising Ir(ppy)3 as emitter molecule (for device details refer to, e.g., Ref. 28). Green solid line indicates that the EQE would remain constant up to a brightness of 30 000 cd m−2, if only the single-step long-range model would contribute to TTA with a Förster radius of 3 nm, as suggested by Staroske et al. 54 and Reineke et al. 59.

The best proof to support the picture of guest aggregates is their visualization. Therefore, a 50 nm thick TCTA:Ir(ppy)3 mixed film with a high concentration of roughly 10 wt% was analyzed by high resolution transmission electron microscopy (TEM) (for details refer to Ref. 59). The TEM image is shown in Fig. 13. Here, the dark contrast refers to atoms with higher mass, i.e., the iridium atoms in the mixed film. Clearly, clustered features with sizes up to 10 nm are visible, resulting in a strong non-uniformity of the material mixture.

Figure 13.

High angle annular dark field (HAADF) TEM image of a 50 nm mixed film of a TCTA:Ir(ppy)3 with a concentration of 10 wt% 59. Higher resolution images are taken for the regions (a), (b), and (c). Image (c) is additionally shown with a further zoom-in on the bottom right. Dark contrast refers to atoms with high mass, i.e., iridium atoms. Arrows are pointing to single iridium atoms in the focal plane of the image. Adapted from Ref. 42.

It is worth mentioning that there is more experimental evidence supporting the picture of guest aggregation and efficient triplet migration in these clusters that is beyond the scope of this paper. Relevant data includes the suppression of triplet motion by inserting high T1 material acting as a barrier 33 and the spectral red-shift of phosphor emission with increasing guest concentration 34, 61. The latter effect can be used to reduce the aggregation and thus TTA by using phosphorescent emitter molecules with smaller dipole moments 60.

To conclude, the discussion of TTA annihilation and guest aggregation may give an answer to a very general question regarding phosphorescence: considering the exciton densities needed for bright OLEDs and the sufficiently short excited lifetime of phosphors that are compatible with the RC-time of the devices 62, why are high guest concentrations typically ranging from 5 to 20 mol% necessary for optimal OLED efficiency? Note here that the photoluminescence efficiency of phosphorescent emitters at these concentrations are already up to 10% lower than their maximum values at low densities (cf. Fig. 7) 34. The clusters of phosphorescent molecules leave regions that are well below the average guest concentration. Thus, whenever an exciton is formed on the host material in these regions, it may be unable to find a phosphorescent site for recombination, ultimately reducing the efficiency of the system. Hence, a system with spatially well-distributed emitters may be used at lower emitter concentration and consequently yield higher internal efficiencies.


This work was supported as part of the Center for Excitonics, an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences under Award Number DE-SC0001088 (MIT). S. R. was supported by the Deutsche Forschungsgemeinschaft (DFG).

Biographical Information

Sebastian Reineke is a post-doc in the research group of Prof. Marc Baldo (MIT, Cambridge), where he works on tailoring organic and organic/inorganic hybrid systems for energy harvesting and luminescent applications. Sebastian received his Ph.D. in physics from the Technische Universität Dresden, where he worked under the supervision of Prof. Karl Leo. There, his main focus was on the photo-physics of organic materials and their implementation in highly efficient organic light-emitting diodes. He is recipient of the Emanuel-Goldberg-Preis 2009 and the Professor-Schwabe-Preis 2006, acknowledging his Ph.D. and diploma work, respectively.

original image

Biographical Information

Marc Baldo is Professor of Electrical Engineering and Computer Science, Associate Director of the Research Laboratory of Electronics, and Director of the Center for Excitonics, a Department of Energy supported by Energy Frontier Research Center. Marc received his B.Eng. from the University of Sydney in 1995 with first class honors and university medal. He received his Ph.D. from Princeton University in 2001, where he helped to develop phosphorescent organic light-emitting devices – now the efficiency standard for organic displays and solid state lighting. He has been at MIT since 2002. At MIT he has worked on organic solar cells, fundamental improvements to the efficiency of organic light-emitting devices, luminescent solar concentrators, and singlet exciton fission.

original image