Transient analysis of triplet exciton dynamics in amorphous organic semiconductor thin films

Organic Electronics 7 (2006) 375–386 www.elsevier.com/locate/orgel

Transient analysis of triplet exciton dynamics in amorphous organic semiconductor thin films N.C. Giebink a, Y. Sun a, S.R. Forrest a

b,*

Princeton Institute for the Science and Technology of Materials (PRISM), Department of Electrical Engineering, Princeton University, Princeton, NJ 08544, USA b Departments of Electrical Engineering and Computer Science, Physics, Materials Science and Engineering, University of Michigan, Ann Arbor, MI 48109, USA Received 1 March 2006; received in revised form 12 April 2006; accepted 20 April 2006 Available online 6 June 2006

Abstract We study triplet exciton diffusion in the archetype organic material 4,4 0 -bis(N-carbazolyl)biphenyl (CBP) commonly used as a conductive host in the emissive zone of organic light emitting devices. Using time-resolved spectral decay ensuing from the diffusion of an initially localized triplet population to a spatially separated phosphor doped region, we model the delayed fluorescence and phosphorescence decays based on non-dispersive triplet transport. Fits to the model yield a diffusion coefficient of D = (1.4 ± 0.3) · 108 cm2/s, and a triplet–triplet bimolecular quenching rate constant of KTT = (1.6 ± 0.4) · 1014 cm3/s. The results are extended by doping a wide energy-gap molecule into CBP that serves to frustrate triplet transport, lowering both the diffusion coefficient and annihilation rate. These results are used to model a recently demonstrated white organic light emitting device that depends on triplet diffusion in CBP to excite spatially separate fluorescent and phosphorescent doped regions of the emissive layer. We determine the extent to which diffusion contributes to light emission in this structure, and predict its performance based on ideal lumophores with unity quantum yield.  2006 Elsevier B.V. All rights reserved. Keywords: Exciton; Electrophosphorescence; Organic light emitting device; Triplet; Diffusion

1. Introduction Energy transport by exciton diffusion in organic thin films plays a significant role in many practical applications, including organic light emitting devices [1,2] (OLEDs), and organic photovoltaic * Corresponding author. Tel.: +1 734 936 2680; fax: +1 734 763 0085. E-mail address: [email protected] (S.R. Forrest).

cells [3]. For example, Sun et al. recently introduced a white light emitting device that depends on diffusive energy transport to achieve the desired color balance at high luminous efficiency [4]. Exciton diffusion also controls the performance of organic photovoltaics, linking exciton generation to the subsequent dissociation into a free electron and hole at a nearby donor/acceptor interface [3]. In both of these examples, efficient energy transport is required to achieve high device performance.

1566-1199/$ - see front matter  2006 Elsevier B.V. All rights reserved. doi:10.1016/j.orgel.2006.04.007

376

N.C. Giebink et al. / Organic Electronics 7 (2006) 375–386

The diffusive characteristics of spin singlet and triplet excitons are considerably different. The forbidden triplet exciton decay transition typically takes 106 times longer than the allowed singlet decay of 1–10 ns [5]. This leads to a triplet diffusion length >100 nm [6,7] in amorphous organic thin films that is often an order of magnitude larger than for singlets [8,9]. Triplet excitons thus provide a means for efficient energy transport within an organic solid. For example, intersystem crossing following photon absorption in C60-based organic solar cells is a source of triplets that leads to an exciton diffusion length (40 nm) that is comparable to the optical absorption length [3], resulting in a high device efficiency. Control of exciton diffusion is the basis for the white-emitting fluorescent/phosphorescent (F/P) hybrid OLED architecture recently reported by Sun et al. [4]. That device uses the different diffusive characteristics of singlet and triplet excitons to spatially separate the decay channels for each species. In constraining singlet decay to a blue fluorescent zone and triplet decay to spatially remote red and green phosphorescent regions, high efficiency white emission was observed. In principle, this device concept offers a lower operating voltage and a more stable white emission spectrum with drive current than conventional all-phosphorescent white organic light emitting devices (WOLEDs), while maintaining the possibility for 100% internal quantum efficiency. Improved understanding of the diffusive process in such a device is one motivation for this current study. Diffusion also plays an important role in both energy transport and triplet–triplet (T–T) annihilation [5,10], both of which are central to understanding the operation of practical devices. Several studies have examined triplet diffusion in organic molecular crystals [5]. Focus has also been given to thin films by employing both steady-state [7] and time-of-flight experiments under electrical excitation [11]. The simple experimental procedure and easily interpretable results presented here serve to complement these previous investigations. We employ optically pumped pseudo-time-offlight measurements to obtain both the triplet diffusion coefficient, D, and the T–T annihilation rate, KTT, for the commonly used OLED host material, 4,4 0 -bis(N-carbazolyl)biphenyl (CBP). These measurements use the delayed fluorescence from CBP and the delayed phosphorescence from a phosphor-doped ‘‘sensing layer’’ to monitor the evolu-

tion of the triplet spatial profile. In addition, we demonstrate that the diffusion coefficient is decreased by doping into CBP a wide energy-gap molecule that scatters triplets and frustrates their transport. The results are employed to understand the mechanisms that control the performance of fluorescent/phosphorescent WOLEDs, and to determine the limitations of such an architecture using CBP as a host material. In Section 2 we develop a theory describing triplet exciton migration and annihilation processes. Experimental details are given in Section 3, and in Section 4 we present the results from transient measurements. In Section 5 we fit and discuss these results in terms of the theory in Section 2, and in Section 6 these results are used to analyze the fluorescent/phosphorescent WOLED. Section 7 provides a summary. 2. Theory Intermolecular triplet transfer is fundamentally different from that for singlets. Transfer of a singlet exciton is often dominated by long-range (5 nm) Fo¨rster dipole–dipole coupling [9]; however, this mechanism contributes negligibly to triplet transfer since both donor and acceptor transitions are disallowed. Triplets, instead, transfer by the short range (<1 nm) Dexter process [12], an exchange coupling between nearest-neighbor molecules that is permitted by a simultaneous interchange of spin on both the donor and acceptor molecules. Triplet migration thus proceeds as a series of incoherent hops between adjacent molecules that have considerable intermolecular electronic orbital overlap. Both Fo¨rster and Dexter processes are modeled as diffusive in the continuum limit [9]. The morphological disorder present in an amorphous organic solid leads to a distribution of both molecular site energies and intersite exchange couplings [13,14]. This results in an inhomogeneously broadened density of states (DOS) within which triplet transport occurs. Relaxation of triplets into the low-energy tail of this DOS gives rise to dispersive transport [13,14], which is typically described by a time-dependent diffusion coefficient. At room temperature, however, thermal energy is generally sufficient to ensure that equilibrium is rapidly approached [14]. Thus, transport quickly moves from the dispersive to the classical regime, at which point it can be described by a time-independent diffusion coefficient.

N.C. Giebink et al. / Organic Electronics 7 (2006) 375–386

The encounter of two triplet excitons (T1) may result in an annihilation reaction that proceeds as K TT

T 1 þ T 1 ! fðT n þ S 0 Þ or ðS n þ S 0 Þg; where the reaction products are a ground state (S0), and a triplet (Tn) or singlet (Sn) in the nth electronic manifold (n P 1) [5]. The creation of a singlet exciton in the latter reaction is responsible for delayed fluorescence [5,15]. The triplet–triplet annihilation rate, KTT, is related to the diffusion coefficient since the triplet mobility controls their encounter probability. That is [15,16]: K TT ¼ 4pDReff ;

ð1Þ

where D is the non-dispersive diffusion coefficient, and Reff is the effective radius for quenching such that triplets approaching within this distance are assumed to annihilate with unity probability. In this work, we combine the results of several experiments to study the one-dimensional evolution of the CBP triplet density, TH(x, t). Assuming nondispersive transport, TH(x, t) obeys the following diffusion equation: oT H o2 T H T H ¼D   K TT T 2H : ot ox2 sT

ð2Þ

Here, triplets are lost via natural decay in time, sT, as well as via T–T annihilation at a rate, KTT. The annihilation reaction can yield singlets that give rise to delayed fluorescence. In this case, the singlet density, SH(x, t), is oS H SH 1 ¼ þ K TT T 2H ; ot sS 2

ð3Þ

where the generation term (second on the righthand side) has a prefactor of 1/2 since annihilation of two triplets produces one singlet. We assume for simplicity that all T–T annihilation reactions yield singlets that decay naturally in time, sS. There is no explicit spatial dependence in Eq. (3) since the effects of singlet diffusion are not relevant on the time scales of our experiments. Similar equations are obtained for the case of a doped film. The dopant provides an additional decay route for host triplets – namely Dexter transfer to the guest dopant molecules at a rate, KTG. Hence, in the presence of a guest:   oT H o2 T H 1 2 ¼D  K T  K þ ð4Þ T H: TT H TG sT ot ox2 For exothermic TH ! TG transfer, back-transfer from guest to host is not energetically favorable,

377

resulting in triplet localization on the dopant. In this case, the guest triplet density, TG(x, t), is oT G TG ¼ þ K TG T H : ot sG

ð5Þ

Eqs. (2)–(5) provide the required framework for the analysis of the experimental results that follow. 3. Experimental A schematic of the sample structure, as well as the processes leading to luminescence, are shown in ˚ thick layer of CBP Fig. 1. For all samples, a 200 A doped at 10 wt.% with a metallorganic phosphor is thermally evaporated onto the surface of a freshly cleaned and deoxidized Si substrate. The phosphors used are fac-tris(2-phenylpyridine) iridium (Ir(ppy)3) [17] emitting in the green, and iridium(III) bis(2-phe0 nyl quinolyl-NC2 ) acetylacetonate (PQIr) [18] that emits in the red. The doped layer functions as a triplet ‘‘sensor’’ since CBP triplets transfer to, and localize on the lower energy phosphor triplet state. Phosphor emission thus serves to indicate the presence of CBP triplets in the doped region. Since the rate of triplet transfer depends strongly on energetic resonance between host and guest, the use of Ir(ppy)3 and PQIr, which have different triplet energies (see Table 1) allows host–guest triplet transfer in the sensing layer to be isolated from that of diffusion in CBP. A CBP spacer layer of thickness, X, varied ˚ and 1200 A ˚ , is deposited on the sensbetween 100 A ing layer surface. Prior to deposition, all organics are purified at least once by train sublimation [19], and all deposition is carried out at a growth rate ˚ /s in vacuum at a base pressure of approximately 3 A 7 of 10 Torr. Immediately following deposition, samples are placed in a closed-cycle He cryostat, and the chamber is evacuated to roughly 50 mTorr. We focus the output of a pulsed N2 laser (<1 ns pulse width; wavelength k = 337 nm; pulse energy from 0.9 to 4.1 lJ) onto a 2 mm diameter spot incident on the spacer layer surface. Each pulse generates an exponentially decaying singlet density spatial profile according to the measured CBP absorption coefficient of a = 2.0 · 105 cm1 at the pump laser wavelength. Neither interference effects nor absorptive saturation are expected to significantly change this excited state profile. Using the above parameters at a pulse energy of 4.1 lJ, we estimate an initial singlet density of (2.9 ± 0.4) · 1019 cm3 at the film surface on which the beam is

378

N.C. Giebink et al. / Organic Electronics 7 (2006) 375–386

200 (Å) Phosphor doped sensing layer (Region 2)

X (Å) Neat CBP spacer layer (Region 1)

λ= 337nm

SH

We ak ISC

Triplet diffusion TH

TH

TH

TH

TH

KTG

TG

hv Triplet-Triplet Annihilation TH + T H

KTT

SH

hv

Prompt fluorescence Phosphorescence

hv Delayed fluorescence

Fig. 1. Schematic illustration of the light emission process in the time-of-flight experiments. The sample structure is also defined in this ˚ ) is deposited on top of a 200 A ˚ phosphor doped (10 wt.%) triplet ‘sensing’ layer diagram. A spacer layer of neat CBP with thickness X (A on a Si substrate. Pulsed laser excitation is incident on the spacer surface, generating singlets. Most singlets decay immediately (prompt fluorescence), while a small fraction intersystem cross to the triplet state. The triplets diffuse through the spacer to the sensing layer where they emit from the guest phosphors (phosphorescence). As they diffuse, they also undergo triplet–triplet annihilation, which can result in delayed fluorescence.

Table 1 Fitting results: CBP triplet dynamics Parameter 2

D (cm /s) KTT (cm3/s) TH0 (cm3) sH (ms)

KTG (s1)

CBP (1.4 ± 0.3) · 108 (1.6 ± 0.4) · 1014 (7 ± 2) · 1017 (14 ± 8) CBP/Ir(ppy)3 (DET  0.2 eV)

CBP/PQIr (DET  0.4 eV)

(0.8 ± 0.4) · 107

(0.5 ± 0.2) · 107

incident. The resulting luminescence from the samples is focused into a fiber and channeled to a Hamamatsu C4334 streak camera that monitors the time-resolved spectral decay. The majority of singlets decay within nanoseconds following the optical pulse, while a small fraction cross into the long-lived triplet manifold [20]. The concentration gradient stemming from the large optical absorption coefficient results in net diffusion towards the sensing layer. During diffusion, triplets may annihilate to produce delayed fluorescence. Fig. 1 schematically illustrates the decay progression. This is a pseudo-time-of-flight (TOF) experiment since varying the spacer layer thickness changes the time that it takes for the bulk of the triplet popula-

tion to diffuse into the sensing layer to produce phosphorescence. We also vary the incident laser intensity by a factor of 5 to control the initial triplet excitation density. This affects the amount of T–T annihilation (and hence delayed fluorescence intensity). 4. Results The results of the TOF transient experiments are shown in Figs. 2 and 3 for two different sets of samples employing Ir(ppy)3 and PQIr as the sensing layer dopant, respectively. These plots show the decay of the delayed fluorescence and the sensing layer phosphorescence for different spacer thicknesses, X. No significant change in the shape of either the delayed fluorescence or phosphorescence ˚ to transient decay occurs as X varies from 1200 A ˚ 600 A, while the decay for both accelerates for ˚ . This reflects a balance reached between X 6 300 A the losses that occur in the spacer and the increased decay rate in the sensing layer due to the presence of the phosphor molecules. The thickness of the spacer layer in relation to both the optical absorption and triplet diffusion lengths is one factor that determines the balance between the various decay channels. For a thick

N.C. Giebink et al. / Organic Electronics 7 (2006) 375–386 6

6

10

10

PL Intensity (a.u.)

4

3

10

1200Å 2

10

200Å

1

5

10

4

10

3

10

2

10

1

4

10

0.0

0.2

0.4 Time (ms)

0.6

600Å

200Å 0.4 Time (ms)

1200Å 0.6

0.8

Fig. 2. Spectrally resolved photoluminescence (PL) transients for Ir(ppy)3 doped in CBP as the sensing layer. The spacer layer ˚ to 1200 A ˚ . The large peaks at the thickness ranges from 100 A start of each transient are due to prompt fluorescence and phosphorescence. The delayed fluorescence transient from CBP (k < 400 nm) is shown in (a), with the corresponding Ir(ppy)3 phosphorescence (k > 450 nm) in (b). Transients for all samples are fit (solid lines) using the theory in the text to give a diffusion constant of D = (1.4 ± 0.3) · 108 cm2/s, a triplet–triplet annihilation rate of KTT = (1.6 ± 0.4) · 1014 cm3/s, and an initial triplet density of TH0 = (7 ± 2) · 1017 cm3.

spacer layer, the majority of triplets are generated too far from the sensing layer to diffuse to it before they annihilate or otherwise naturally decay. In that case, the delayed fluorescence is intense while the phosphorescence is weak due to the small fraction of triplets that reach the sensor. Conversely, samples with thin spacers show increased phosphorescence intensity and fast decays for both luminescence signals, since diffusion to the sensing layer efficiently competes with annihilation and natural decay. The intensity of the exciting optical pulse is also a factor in determining the decay transients, since the annihilation rate is proportional to the square of the

300Å 0.0

0.8

100Å

0.2

1200Å

2

10

1

(b)

0.0

600Å

3

10

100Å

0

10

10

10

600Å

PL Intensity (a.u.)

PL Intensity (a.u.)

5

10

10

PL Intensity (a.u.)

(a)

(a)

5

10

10

379

10

5

10

4

10

3

10

2

10

1

0.2

0.4 Time (ms)

0.6

0.8

(b)

300Å

600Å

1200Å 0.0

0.2

0.4 Time (ms)

0.6

0.8

Fig. 3. Spectrally resolved photoluminescence (PL) transients for samples using a PQIr doped CBP sensing layer, as in Fig. 2. The ˚ to 1200 A ˚ . The delayed spacer layer thickness ranges from 300 A fluorescence transient from CBP (k < 400 nm) is shown in (a), with the corresponding PQIr phosphorescence (k > 450 nm) in (b). Transients for all samples are fit (solid lines) using the theory in the text to give D = (1.4 ± 0.3) · 108 cm2/s, KTT = (1.6 ± 0.4) · 1014 cm3/s, TH0 = (7 ± 2) · 1017 cm3.

triplet density. To examine this effect, we vary the pulse energy incident on a sample with a spacer ˚ deposited on a PQIr sensing thickness of X = 300 A layer. The transient results are plotted in Fig. 4 for pulse energies ranging from 0.9 lJ to 4.1 lJ. Both luminescent transients decay more slowly and become less intense with decreasing pump fluence, although this trend is most evident in the delayed fluorescence signal. Scattering or trapping of triplet excitons at defect sites in a thin film are expected to reduce the triplet diffusion coefficient as compared to a defect-free film. We demonstrate this effect by doping into CBP the wide energy gap molecule p-bis(triphenylsilyly)benzene (UGH2) whose triplet energy [21] is 3.5 eV; well in excess of the 2.6 eV CBP triplet [22]. Since no energy transfer from the CBP to

380

N.C. Giebink et al. / Organic Electronics 7 (2006) 375–386 6

10

5

(a)

( 4.1 ± 0.3) μ J ( 2.7 ± 0.2 ) μ J

4

10

( 0.9 ± 0.2 ) μ J

3

10

2

10

0

Delayed fluorescence

4

10

3

10

2

0.0

0.2

0.4 Time (ms)

1

10 0.6

Delayed fluorescence

0.0

0.8

0.1

0.2 0.3 Time (ms)

0.4

0.5

5

10

4

10

3

10

2

10

1

(b)

( 4.1 ± 0.3) μ J

PL Intensity (a.u.)

PL Intensity (a.u.)

10

10

(a)

10

1

10 10

0wt% UGH2 3wt% UGH2 10wt% UGH2 18wt% UGH2

5

PL Intensity (a.u.)

PL Intensity (a.u.)

10

( 2.7 ± 0.2 ) μ J ( 0.9 ± 0.2 ) μ J

PQIr phosphorescence 0.0

0.2

0.4 Time (ms)

10

5

10

4

10

3

10

2

10

1

0wt% UGH2 3wt% UGH2 10wt% UGH2 18wt% UGH2

PQIr phosphorescence

0.0 0.6

0.8

Fig. 4. Spectrally resolved photoluminescence transients obtained for several excitation pulse intensities for samples with PQIr doped into CBP as a sensing layer. Subplots (a) and (b) show the delayed fluorescence of CBP, and the phosphorescence of PQIr, respectively. Solid lines show fits calculated for different TH0 (listed in Table 2) while holding D and KTT constant at the values given in Table 1.

the UGH2 triplet state is possible, the UGH2 molecules scatter triplets and frustrate their movement through the CBP matrix, decreasing the diffusion coefficient. Transient decays are shown in Fig. 5 for samples having 0, 3, 10 and 18 wt.% UGH2 doped into a ˚ thick CBP spacer, using a PQIr-doped sens300 A ing layer. Both delayed fluorescence and phosphorescence trend towards lower intensity and slower decays with increasing UGH2 concentration. The reduced delayed fluorescence indicates that the T–T annihilation rate decreases in the presence of UGH2, whereas the decreased and extended phosphorescence signal suggests that triplets take longer to reach the sensing layer. Both of these observations are consistent with a reduced triplet mobility in the spacer layer.

(b)

0.1

0.2

0.3

0.4

0.5

Time (ms) Fig. 5. Spectrally resolved delayed fluorescence (a) and PQIr phosphorescence (b) transient data obtained for samples using a ˚ thick CBP spacer layer co-doped PQIr sensing layer with a 300 A with varying concentrations of the wide-gap molecule UGH2. Solid lines indicate fits to the theory in the text by varying D and KTT, holding all other parameters constant at the values listed in Table 1.

We also determined the natural CBP triplet lifetime, sT, at temperatures from 5 K to 170 K by opti˚ thick film deposited on cally pumping a neat, 500 A Si. The phosphorescence decay rates shown in Fig. 6 are obtained from monoexponential fits to the phosphorescence transient for times t > 5 ms following the N2 laser pulse to eliminate effects from T–T annihilation. The green CBP phosphorescence with a high-energy edge at k = 480 nm is apparent in the inset of Fig. 6, along with the delayed fluorescence signal centered at k = 410 nm. As shown in Fig. 6, the natural triplet decay is independent of temperature below 50 K, and is only weakly activated at higher temperatures, with an energy, Ea = 9 ± 1 meV. Since the phosphorescent decay of a single molecule is typically independent of temperature [20], the lifetime of triplets at

N.C. Giebink et al. / Organic Electronics 7 (2006) 375–386

50 45 40 35

Delayed fluorescence

CBP Phosphorescence 5K 20K 40K 60K 100K 140K 170K

1/τΗ(s-1)

30 25 20

350

400

450

500

550

600

650

15

0

10

20

30 40 50 1000/T (K-1)

190 200 210

Fig. 6. Arrhenius plot of the natural CBP triplet decay rate (1/sH) obtained for temperatures in the range from 5 K to 170 K. Inset: Spectra of CBP phosphorescence and delayed fluorescence at 1 ms following the initial optical excitation pulse.

higher temperatures is likely limited by quenching with impurities encountered in the course of migration in the film. As temperature increases, levels of higher energy in the triplet density of states (with typical Gaussian width r  30 meV [14], similar in magnitude to the value of Ea determined here) are populated. This increases the exciton mobility, and leads to more frequent encounters with quenching sites, thus decreasing the lifetime. This effect is intrinsic to the CBP films grown in these experiments, and extrapolation of the fit to room-temperature yields the natural triplet lifetime of sT = 14 ± 8 ms at T = 295 K. 5. Discussion We analyze the TOF transient data by applying Eqs. (2) and (3) to region 1 of our samples (see Fig. 1), and Eqs. (4) and (5) to region 2. Eq. (2) describes the CBP triplet density in the spacer layer, TH(x, t), while Eq. (4) quantifies it in the sensor layer. These two equations are coupled at the spacer/sensor layer interface, with both TH(x, t) and its spatial derivative taken to be continuous across the boundary. The non-linear partial differential equations (2) and (4) are solved using finite element methods [23]. The solutions for the spacer layer singlet population, SH(x, t) (Eq. 3), and the sensing layer guest triplet population, TG(x, t) (Eq. 5), are determined once TH(x, t) is known. By spatially integrating SH(x, t) and TG(x, t) across their respective regions,

381

we find the total luminescent populations S 0H ðtÞ and T 0G ðtÞ. These functions are then least-squares fit with statistical weighting to the corresponding TOF luminescence transient data. We fit the data in Figs. 2 and 3 for a given sensing layer using Eqs. (2)–(5) by varying the parameters D, KTT, KTG, and TH0, (the initial triplet density). For these fits, we use the measured CBP triplet lifetime in Section 4, the CBP singlet lifetime of ss = 0.7 ns [24], and the known triplet lifetimes of Ir(ppy)3, sG = 0.8 ls [11] and PQIr, sG = 2.0 ls [25]. The fitting results are summarized for both Ir(ppy)3 and PQIr samples in Table 1, and are shown by the solid lines in Figs. 2 and 3. Within experimental uncertainty, we obtain the same values for the parameters D, KTT, and TH0, for both sets of samples, as would be expected. However, KTG is nearly a factor of two greater for CBP ! Ir(ppy)3 transfer than for CBP ! PQIr, presumably due to the stronger resonance between Ir(ppy)3 and CBP triplet energies resulting in more efficient and rapid transfer. Accounting for T–T annihilation at TH0 = 7 · 1017 cm3, the effective diffusion length is LD  ˚ , which is considerably less than LD ¼ 250 A p ffiffiffiffiffiffiffiffi ffi ˚ found in the absence of T–T anniDsH  1400 A hilation. This is similar to the case of tris(8-hydroxy˚ was quinoline) (Alq3), where [26] LD > 1400 A ˚ at found at low excitation, and reduced to 140 A high excitation [11]. Note that the effective diffusion ˚ calculated for these TOF results is length of 250 A consistent with the previous observation that significant changes in the transients only occur as the ˚ . We similarly fit the data spacer is thinned to 300 A of Fig. 4 by fixing all parameters at the values given in Table 1, and allowing only TH0 to vary. In Table 2, the fitted values for TH0 are nearly proportional to the excitation intensity, as expected. Fitting the transient data for the UGH2 doped samples of Fig. 5 confirms that the diffusion coefficient decreases with increasing UGH2 concentration. In this case, D and KTT are varied while holding all other parameters constant. Summarizing the fitting results in Fig. 7, we find that both D and KTT decrease at the same rate, consistent with Eq. (1). At an UGH2 concentration of 18 wt.%, corresponding to less than two CBP molecules between each UGH2 molecule, the diffusion coefficient in the spacer layer drops by a factor of four. This effect is analogous to diffusion through porous media in that UHG2 molecules act as scattering centers for triplets diffusing in CBP [27].

382

N.C. Giebink et al. / Organic Electronics 7 (2006) 375–386

Table 2 Fitting results: initial triplet density vs. pulse energy

6. Analysis of the fluorescent/phosphorescent WOLED

Initial triplet density TH0 · 1017 cm3

4.1 lJ (100%) 2.7 lJ (65%) 0.9 lJ (22%)

9 ± 1 (100%) 6 ± 1 (66%) 2 ± 1 (22%)

2.0

2.0

1.8

1.8

1.6

1.6

1.4

1.4

1.2

1.2

1.0

1.0

0.8

0.8

0.6

0.6

0.4

0.4 -2 0 2

4 6 8 10 12 14 16 18 20 UGH2 Concentration (%)

The concept of the fluorescent/phosphorescent white organic light emitting device (WOLED) recently introduced by Sun et al. [4], involves separating the radiative decay channels available to singlet and triplet excitons. In that device, whose structure and emission properties are shown in Fig. 8, the extended triplet diffusion length provides the exciton separation mechanism. Excitons form on CBP in a blue fluorescent-doped region within the WOLED emissive layer. The singlets rapidly Fo¨rster transfer [5] to the blue fluorophore, while the long-lived triplet excitons diffuse to a spatially separated phosphor-doped region that provides the green and red components of the white emission spectrum. Fig. 9a summarizes these energy transfer paths. Hole transport from the indium-tin-oxide (ITO) anode to the emissive layer is facilitated by a ˚ thick layer of 4,4 0 -bis[N-(1-naphthyl)-N-phe400 A ˚ thick layer nyl-amino]-biphenyl (NPD). A 200 A of 4,7-diphenyl-1,10-phenanthroline (BPhen) fol˚ thick Li doped BPhen layer at a lowed by a 200 A 1:1 molar ratio serves to transport electrons from the cathode while confining excitons within the emissive layer (EML). In the center of the EML is ˚ thick region consisting of Ir(ppy)3 and PQIr a 200 A

Annihilation Coefficient, KTT x10-14 (cm3/s)

Diffusion Coefficient, D x10-8 (cm2/s)

Excitation level (% of full)

Fig. 7. Diffusion coefficient (D) and annihilation rate (KTT) obtained from the data of Fig. 5. Both D and KTT decrease in proportion to one another with increasing UGH2 concentration, consistent with Eq. (1) in the text. This is a result of triplet scattering by UGH2 molecules which frustrates transport through the CBP matrix.

(a)

(b)

LiF/Al (500Å)

400

Emissive Layer (x 0)

BCzVBi:CBP 5wt% (100Å) CBP (60Å) Ir(ppy)3:CBP 5wt% (120Å) PQIr:CBP 4wt% (80Å) CBP (40Å) BCzVBi:CBP 5wt% (150Å) NPD (400Å) ITO/Glass

Wavelength (nm) 500 550 600

650

700

0.8

0.6

0.4

PQIr

0.2

Ir(ppy)3

Spectral Intensity (a.u.)

BPhen (200Å)

Forward Emitted EQE (%)

Li : BPhen 1:1 mol (200Å)

12 11 10 9 8 7 6 5 4 3 2 1 0 -3 10

450

BCzVBi -2

10

-1

0

1

2

10 10 10 10 Current Density J (mA/cm2)

0.0 3 10

Fig. 8. (a) Schematic structure of the fluorescent/phosphorescent (F/P) WOLED of Ref. [4]. (b) The forward-emitted external quantum efficiency of the F/P WOLED (open squares) as a function of current density (J), and the emission spectrum at J = 10 mA/cm2 (open circles). This spectrum is fit (solid, bold line) using the individual dopant photoluminescent spectra (solid, thin lines).

N.C. Giebink et al. / Organic Electronics 7 (2006) 375–386

(a)

S1

Förster transfer

S1

hv

S1

Energy

X T1 CBP Host

X

Diffusive transfer

T1 BLUE fluorescent dopant

Interface exciton formation

(b)

hv

T1

RED and GREEN phosphorescent dopants

Right

Trap exciton formation

Left

dEX x=0 W PH

x = x0

Fig. 9. (a) The energy transfer routes available to singlet and triplet excitons for the device of Fig. 8. ‘X’ denotes an energy transfer route that is unlikely to occur due to the spatial separation and concentration of the fluorescent and phosphorescent dopants. (b) Exciton density within the emission layer of the device in Fig. 8 (dashed line). Trap recombination is assumed to be spatially uniform in the phosphor region, while formation on the CBP host follows an exponential profile at both edges of the emissive layer. The ratio of exciton densities at the two edges of the emission layer is dLeft/dRight, with dLeft + dRight = 1.

doped into CBP. This phosphorescent region is sandwiched between two undoped CBP spacers ˚ and the other 60 A ˚ thick, see Fig. 8a), (one 40 A which in turn are sandwiched between two layers ˚ and the other 100 A ˚ thick, see of CBP (one 150 A Fig. 8a) doped with 5 wt.% of the blue fluorophore, 4,4 0 -bis(9-ethyl-3-carbazovinylene)-1,1 0 -biphenyl (BCzVBi). Exciton formation has been shown to occur in both blue fluorophore-doped regions [4]. Roughly 75% of the resulting phosphor emission is due to triplet diffusion from the EML edges followed by Dexter transfer to the dopants [4], with the remainder due to excitons formed by direct charge trapping on the phosphors. Here, we quantitatively analyze the operation of this device in steady-state using the values of D and KTT obtained in Section 5. We assume negligible diffusion of singlets away from their fluorescent zones at both EML edges due to their comparatively shorter diffusion lengths [8]. All quenching processes are neglected except T–T annihilation since it is expected to be the dominant source of loss for the diffusing triplets. In addition, we assume that trip-

383

lets do not transfer to the blue fluorescent dopant due to the poor guest–host orbital overlap that results from their different molecular conformations [28]. Finally, the exciton distribution is assumed to exponentially decrease with distance from each edge of the EML [29] with characteristic length, dEX (see Fig. 9b). In this structure, nearly every injected charge results in an exciton since the transport layers (NPD and BPhen) confine charge within the emissive layer, thereby providing balanced electron and hole injection [30]. The component of the external quantum efficiency due to blue emission from BCzVBi is directly proportional to the number of singlets present in the two fluorescent zones. Using the fluorescent quantum yield of BCzVBi (/FL) and the optical out-coupling efficiency through the glass substrate [18,31] (/OC = 0.2), we can estimate the singlet density, S, as a function of injection current density, J. By equating S to the integral of the singlet exciton formation profile (see Fig. 9b) over the two distinct fluorescent zones, we can determine the total exciton density distribution, since three triplets are formed for every singlet [26]. In addition, we allow for a fraction, aT, of the total exciton population to form by trapping directly on the phosphors. Using this profile, the rate equations for the singlet and triplet densities in the emissive layer are oT o2 T T ¼ DT 2   k TT T 2 ot ox sT   3J ð1  aT Þ gBlue þ qd EX ð1  exp½x0 =d EX Þ /FL /OC      x x  x0  dLeft exp þ dRight exp d EX d EX   3J ðgBlue ÞaT  TK TG ; þ PH q/FL /OC W PH

ð6Þ

Rgn

and

  oS S J ð1  aT Þ gBlue ¼ þ ot sS qd EX ð1  exp½x0 =d EX Þ /FL /OC      x x  x0  dLeft exp þ dRight exp d EX d EX J ðgBlue ÞaT : ð7Þ þ q/FL /OC W PH Here, dLeft/Right are the weighting fractions for exciton formation at the left/right side of the emissive layer (see Fig. 9b), and sG and sS are the phosphor and BCzVBi lifetimes, respectively. We find that /FL  0.6 for BCzVBi in toluene solution, using

384

N.C. Giebink et al. / Organic Electronics 7 (2006) 375–386

standard procedures [32]. Also, gBlue is the forwardemitted external quantum efficiency (EQE) for an optimized, purely fluorescent device using a 5 wt.% BCzVBi:CBP emissive layer, and ranges from 2.7% at J = 1 mA/cm2 to 2.3% at 100 mA/cm2. Any deviation in the blue component of the F/P WOLED efficiency from gBlue is a result of the fraction, aT, of excitons formed by direct charge trapping on the phosphor molecules outside of the fluorescent zones, as described by the last term on the right-hand side of Eqs. (6) and (7). Solving Eqs. (6) and (7) in steady-state for S(x) and T(x), we can express the fluorescent (gFL) and phosphorescent (gPH) components of the total WOLED efficiency as Z J 1 ðg Þ ¼ SðxÞ dx; ð8Þ q/OC /FL FL sS Fluor Regions

and J 1 ðg Þ ¼ q/OC PH sG

Z PH Region

½T ðxÞ þ SðxÞ dx;

ð9Þ

Forward Viewing EQE (%)

where the phosphor quantum yield is taken to be unity [33]. Fig. 10 summarizes the results, showing lines of constant component efficiency (solid = fluorescent; dashed-dotted = phosphorescent) at a current density of 10 mA/cm2 for the device in Fig. 8, as determined from the weighted contributions of

the dopant photoluminescent spectra to the device emission (see Fig. 8). These are overlaid on contours that indicate the percentage of total excitons formed by trapping on the phosphors (dotted line). Note that the total exciton fraction formed in the phosphor-doped region, indicated by the contours, is slightly greater than the trapping fraction, aT, since the exponential tails of the edge formation distributions also extend into the phosphor-doped region. Although neither dEX nor aT are known, the device operating point must correspond to the crossing of the fluorescent and phosphorescent efficiency lines. The crossings at all current densities (1 and 100 mA/cm2, not shown) lie within the shaded region of Fig. 10, indicating a characteristic exciton ˚ , and a total exciton formation length of dEX  75 A trapping fraction of 20–30%. In Fig. 11 we simulate the external quantum efficiency at this operating point. For the fluorescent component, we find good agreement at all current densities with the device data from Ref. [4], which has been reproduced for comparison. At low current densities, however, the model overestimates the phosphorescent emission component of the EQE. This discrepancy is attributed to losses, such

Fig. 10. Model results for the fluorescent/phosphorescent WOLED calculated using Eqs. (6)–(9) in text. The dotted contours indicate the percentage of excitons that are formed by direct charge trapping on the phosphor dopants. Superimposed on this contour plane are lines of constant component dopant efficiency at a current density of J = 10 mA/cm2 for the fluorophore (solid line) and phosphor (dashed-dotted line) that are obtained by fits to the WOLED device data of Ref. [4].

24 22 20 18 16 14 12 10 8 6 4 2 0 0.0

0% Trapping 30% Trapping

Device of Ref. [4] Model fit Ideal lumophores

Phosphor component Fluorescent component

0.1 0.2 0.3 2 Current Density (mA/cm )

0.4

Fig. 11. Forward-emitted external quantum efficiency (EQE) vs. current density obtained from the data of Ref. [4] (squares) and the fit (circles) given by the model in the text. The separation of the device EQE data into individual fluorescent and phosphorescent contributions is obtained by weighting to the photoluminescent spectra of each lumophore in the fit to the WOLED emission spectrum as shown in Fig. 8, making sure to account for the intensity to quantum yield conversion. Inverted triangles indicate predictions for the structure in Fig. 8 using ‘‘ideal’’ emitters having unity quantum yield for the cases of 0% and 30% direct exciton formation on the phosphors.

N.C. Giebink et al. / Organic Electronics 7 (2006) 375–386

as exciton–polaron and singlet–triplet quenching, not accounted for in the model. However, at higher current densities where T–T annihilation becomes the dominant source of loss, the model prediction falls close to the device data. Balanced white emission with a high color rendering index [34] (CRI) can be achieved with a smaller blue emissive component than the singlet spin fraction (vS = 1/4) affords. Charges that form excitons by trapping increase the phosphor emission at the expense of blue fluorescence, since they are prevented from otherwise forming BCzVBi singlets. Trap formation is thus beneficial to device performance since the trapped exciton fraction is not subject to diffusive transport losses, and the reduced triplet density at the EML edges decreases T–T quenching. From the photoluminescent spectrum of each dopant (see Fig. 8b), we calculate that CRI > 80 can be maintained for up to 35% exciton formation via direct trapping. Thus, the device of Ref. [4] operates close to the optimum point, representing a tradeoff between high efficiency and white color balance. We can extend this model to predict the performance for this same structure using an ‘‘ideal’’ fluorescent dopant with unity luminescent quantum yield. As shown in Fig. 11, the maximum forwardemitted efficiency increases from 17% (CRI = 86) in the absence of trapping to 18.5% (CRI = 81) for an optimum trapping fraction of 30%. Accounting for all white light emitted [35] the total (integrated) external quantum efficiency for this optimized device is 31%. In addition, the roll-off in EQE with increasing current density is lessened (as compared to the device without trapping) since T–T annihilation is reduced, as discussed above. 7. Conclusions We have studied triplet exciton diffusion in amorphous thin films of the commonly used organic semiconductor, CBP, using optically pumped timeof-flight techniques. The diffusion of triplets through a CBP spacer layer to a phosphor doped sensing layer was used to determine a diffusion coefficient of D = (1.4 ± 0.3) · 108 cm2/s, and a triplet–triplet annihilation rate KTT = (1.6 ± 0.4) · 1014 cm3/s. Both D and KTT are similar in magnitude to corresponding results obtained for other amorphous, small-molecule films [11,36]. We also demonstrated that D can be reduced by a factor of four from its value in neat CBP by doping with

385

the wide energy-gap molecule UGH2 at 18 wt.%, demonstrating that this dopant introduces scattering centers that frustrate exciton transport. The model of exciton transport was extended to analyze the operation of a recently reported fluorescent/phosphorescent WOLED that depends on triplet diffusion to obtain high efficiency and a balanced white color emission. The model confirms that triplet diffusion is the dominant mechanism contributing to the total phosphor emission of this device, although charge trapping is also an important factor in obtaining optimum performance and color balance. Our analysis suggests that this particular device architecture can yield a total external quantum efficiency EQE = 31% using optimized dopants in a CBP host. Acknowledgements The authors thank Universal Display Corporation and the US Department of Energy for partial financial support of this work. In addition, we thank Dr. H. Kanno for sharing experimental data, as well as Prof. M. Thompson and Mr. R.J. Holmes for helpful discussions. References [1] B.W. D’Andrade, M.E. Thompson, S.R. Forrest, Adv. Mater. (Weinheim, Fed. Repub. Ger.) 14 (2002) 147–150. [2] B.W. D’Andrade, S.R. Forrest, J. Appl. Phys. 94 (2003) 3101–3109. [3] P. Peumans, A. Yakimov, S.R. Forrest, J. Appl. Phys. 93 (2003) 3693–3723. [4] Y. Sun, N.C. Giebink, H. Kanno, B. Ma, M.E. Thompson, S.R. Forrest, Nature 440 (2006) 908–911. [5] M. Pope, C. Swenberg, Electronic Processes in Organic Crystals and Polymers, Oxford University Press, New York, NY, 1999. [6] M.A. Baldo, S.R. Forrest, Phys. Rev. B 66 (1999) 14422– 14428. [7] N. Matsusue, S. Ikame, Y. Suzuki, H. Naito, J. Appl. Phys. 97 (2005). [8] C.W. Tang, S.A. Vanslyke, C.H. Chen, J. Appl. Phys. 65 (1989) 3610–3616. [9] R.C. Powell, Z.G. Soos, J. Lumin. 11 (1975) 1–45. [10] M.A. Baldo, C. Adachi, S.R. Forrest, Phys. Rev. B 62 (2000) 10967–10977. [11] M.A. Baldo, S.R. Forrest, Phys. Rev. B 62 (2000) 10958– 10966. [12] D.L. Dexter, J. Chem. Phys. 21 (1953) 836–850. [13] B. Movaghar, M. Grunewald, B. Ries, H. Bassler, D. Wurtz, Phys. Rev. B 33 (1986) 5545–5554. [14] R. Richert, H. Bassler, J. Chem. Phys. 84 (1986) 3567–3572. [15] D. Hertel, H. Bassler, R. Guentner, U. Scherf, J. Chem. Phys. 115 (2001) 10007–10013.

386

N.C. Giebink et al. / Organic Electronics 7 (2006) 375–386

[16] S. Chandrasekhar, Rev. Modern Phys. 15 (1943) 1–84. [17] M.A. Baldo, S. Lamansky, P.E. Burrows, M.E. Thompson, S.R. Forrest, Appl. Phys. Lett. 75 (1999) 4–6. [18] B.W. D’Andrade, R.J. Holmes, S.R. Forrest, Adv. Mater. (Weinheim, Fed. Repub. Ger.) 16 (2004) 624–627. [19] S.R. Forrest, Chem. Rev. 97 (1997) 1793–1896. [20] N.J. Turro, Modern Molecular Photochemistry, University Science Books, Menlo Park, CA, 1991. [21] X.F. Ren, J. Li, R.J. Holmes, P.I. Djurovich, S.R. Forrest, M.E. Thompson, Chem. Mater. 16 (2004) 4743–4747. [22] R.J. Holmes, B.W. D’Andrade, S.R. Forrest, X. Ren, J. Li, M.E. Thompson, Appl. Phys. Lett. 83 (2003) 3818–3820. [23] COMSOL_Multiphysics, Burlington, 2005. [24] Y. Kawamura, H. Yamamoto, K. Goushi, H. Sasabe, C. Adachi, H. Yoshizaki, Appl. Phys. Lett. 84 (2004) 2724– 2726. [25] S. Lamansky, P. Djurovich, D. Murphy, F. Abdel-Razzaq, R. Kwong, I. Tsyba, M. Bortz, B. Mui, R. Bau, M.E. Thompson, Inorg. Chem. 40 (2001) 1704–1711. [26] M.A. Baldo, D.F. O’Brien, M.E. Thompson, S.R. Forrest, Phys. Rev. B 60 (1999) 14422–14428.

[27] M.H. Abbasi, J.W. Evans, I.S. Abramson, AIChE J. 29 (1983) 617–624. [28] C. Hosokawa, H. Higashi, H. Nakamura, T. Kusumoto, Appl. Phys. Lett. 67 (1995) 3853–3855. [29] J. Kalinowski, N. Camaioni, P. Di Marco, V. Fattori, A. Martelli, Appl. Phys. Lett. 72 (1998) 513–515. [30] C. Adachi, M.E. Thompson, S.R. Forrest, IEEE J. Selected Topics Quant. Electron. 8 (2002) 372–377. [31] G. Gu, D.Z. Garbuzov, P.E. Burrows, S. Venkatesh, S.R. Forrest, M.E. Thompson, Opt. Lett. 22 (1997) 396–398. [32] J.V. Morris, M.A. Mahaney, J.R. Huber, J. Phys. Chem. 80 (1976) 969–974. [33] Y. Kawamura, K. Goushi, J. Brooks, J.J. Brown, H. Sasabe, C. Adachi, Appl. Phys. Lett. 86 (2005) 071104. [34] In Method of measuring and specifying colour rendering properties of light sources, Commission Internationale de L’eˆclairage (CIE), 1974. [35] B.W. D’Andrade, S.R. Forrest, Adv. Mater. 16 (2004) 1585– 1595. [36] J. Kalinowski, W. Stampor, J. Mezyk, M. Cocchi, D. Virgili, V. Fattori, P. Di Marco, Phys. Rev. B 66 (2002).