Operating lifetime recovery in organic light-emitting diodes having an azaaromatic hole-blocking/electron-transporting layer

Viktor V. Jarikov

Supplemental Information

The “intrinsic” degradation of organic light-emitting diode (OLED)devices can be defined as an operation-driven, monotonic loss of electroluminescent (EL) efficiency (“EL fade”) that is uniform across the luminous area. It is proposed to occur via several mechanisms:1,2 (i) reversible or physical, e.g., diffusion of ions3-6 and reorientation of dipoles,7,8 and (ii) irreversible or chemical, e.g., reactivity of radical ions9 and electronically excited states10 and reactions of organic materials with oxygen and water.11-13 The EL fade rate of fluorescent OLEDs is usually approximately proportional to the drive current density (J)14,15 rather than the drive voltage (Vd). The light output and, hence, the exciton concentration are also approximately proportional to J. This suggests that the rate-limiting degradation step may involve a species generated by charge recombination, such as an excited state of an OLED material.10 Nevertheless, radical ions were also shown to be involved in degradation.1 Chemical reactions appear to take place inside or in the vicinity of the charge recombination zone and produce fluorescence quenchers9,16,17 and charge traps acting as nonradiative recombination centers (NRRC).18

TDPF / DCJTB / DBIP / C545T
DPQA / TBADN / TBP / NP

FIG 1. Continued.

TDPF versus DCJTB. It appears that compounds having functional groups or heteroatoms in their molecular structure often display shorter EL lifetimes than those of their parent materials or pure PAHs, and the bluer the EL, the stronger the lifetime shortening induced by substitution. In a vast majority of cases, such materials also have relatively weak bonds or relatively photochemically or electrochemically reactive moieties. Hence, they may be less stable merely because their electronically excited states or radical ion states undergo chemistry. The markedly longer lifetimes of the TDPF-doped devices versus those of the corresponding DCJTB-doped cells appear to fall along the lines of such observations and reasoning.

Even in the simple Alq-based devices, TDPF advantage is clearly observed. For example, the NPB|Alq+0.5%TDPF|Alq cell exhibits a half-life of ~15,000 h at 20 mA/cm2 and room temperature and 3500 h at the same J and 70 ºC versus 3300 h and 580 h, respectively, for the corresponding devices doped with 1%2% DCJTB.19 However, the TDPF cell has an inferior EQE of ~1.4% (1.8 cd/A) and poor color because of the relatively strong Alq EL contribution (CIEx,y 0.57, 0.40). The 1% and 2% DCJTB cells produce an EQE of 2% (2.5 cd/A) and 1.5% (1.7 cd/A), respectively, and CIEx,y of 0.61, 0.38 and 0.64, 0.36, respectively.

The half-life drastically extends with the increasing TDPF concentration in any host (albeit the EQE strongly decreases as a result of efficient aggregation and concentration quenching of this flat and rigid PAH). At 5% TDPF, the half-life (at a fixed J) extends by several times. With DCJTB, the half-life is insensitive in the ~0.5%3% range and starts to shorten at higher percentages. DCJTB also undergoes an efficient concentration quenching.

DBIP versus C545T and DPQA. A close analogous case is illustrated by comparing a very stable green dopant DBIP (which is a PAH similar to TDPF) with such known green dopants as coumarin dye C545T and quinacridone dye DPQA (Fig. 1). An improved synthetic procedure for DBIP has been described recently.20 DBIP has a solution f of near unity. According to solution electrochemistry, this PAH dispersed in Alq should be a ~0.2-eV hole trap and a deep ~0.55-eV electron trap.

When doped in Alq at the optimum concentration of 0.25% (NPB|LEL|Alq), DBIP displays yellow-green emission with a moderate EL yield of 7.4 cd/A at 20 mA/cm2.19,21 As a result of the concentration quenching, the EL efficiency of this flat and rigid PAH is reduced at higher concentrations, e.g., to 5.3 cd/A at 1%. Significantly, the operating half-life is greatly extended to ~5500 h at 40 mA/cm2 versus ~1000 h for the undoped control device (NPB|Alq). The EL spectrum peaks at 548 nm (CIEx,y 0.45, 0.53). The emission can be blue-shifted to 540 nm and CIEx,y improved to 0.41, 0.52 by employing a nonpolar host, such as TBADN. The EL yield drops to 5.3 cd/A while the lifetime remains long, ~5000 h at 40 mA/cm2 versus 550 h for the undoped NPB|TBADN|Alq control device.

The same devices doped with C545T or DPQA are not nearly as stable. For instance, an average ITO|10Å CFx|750Å NPB|400Å Alq+1%C545T|350Å Alq|Mg:Ag cell exhibits a 7.3 V drive voltage, 2.9% EQE (10 cd/A), and CIEx,y of 0.29, 0.65, all at 20 mA/cm2 (versus 7 V, 1% EQE, 3 cd/A, and CIEx,y of 0.34, 0.55 for the undoped NPB|Alq control device). The half-life is 550 h at 40 mA/cm2, i.e., only about half of that for the control device (~1000 h). The same device but doped with 0.7% DPQA exhibits a 7.8 V drive voltage, 2.5% EQE (9 cd/A), and CIEx,y of 0.31, 0.65, all at 20 mA/cm2. The half-life is essentially the same as that of the control device (~1000 h).

Thus, DBIP devices are ~10 times and ~5.5 times more stable that the equivalent C545T and DPQA cells, respectively. Again, as with the TDPF versus DCJTB comparison, the increase in the DBIP percentage in any host extends the lifetime dramatically while increasing C545T or DPQA concentration dramatically shortens the lifetime.

Curiously, the PL of the C545T and DPQA devices, as well as DCJTB ones, fades markedly faster compared to that in the undoped cells. In one admittedly weak interpretation, this may hint that the dopants themselves degrade in the operating OLEDs. At least, this signals that the dopants affect the pathways of operating degradation of OLEDs. This is described in the following section.

PL fade versus EL fade. It has been shown that the degradation products act as fluorescence quenchers9,16,17 and nonradiative charge recombination centers (NRRC).18 The PL loss of OLED devices is usually smaller than the EL loss [Fig. 2(a) and (b)]. It is possible that this disparity arises because the EL and PL originate from different parts of the LEL and the distribution of fluorescence quenchers is nonuniform.17 Also, the accumulation of NNRCs is expected to result in EL loss being larger than PL loss.16-18 It is conceivable that the PL loss and EL loss are effected via different mechanisms. However, there appears to exist a direct linear correlation between the EL fade and PL fade, both of which are also linearly correlated to the accumulation of the deeply trapped positive charge in operating devices.22

FIG. 2. (a) Evolution of Alq EL and PL, NPB PL, and drive voltage for an average NPB|Alq device driven at 40 mA/cm2. (b) The dependence of EL and PL on the time of operation at 40 mA/cm2 for an average NPB|TBADN|Alq device, where TBADN is either undoped or doped with 2% TBP. (c) Comparison of normalized EL spectra for an average NPB|Alq device at two current densities: fresh cell versus the one aged for 11,250 h at 40 mA/cm2 [plot (a)]. (d) Comparison of normalized EL spectra for an average NPB|TBADN|Alq device at two current densities: fresh cell versus the one aged for 1030 h at 40 mA/cm2 [plot (b)]. (e) Comparison of normalized EL spectra for an NPB|Alq+1%DCJTB|Alq device at two current densities: fresh cell versus the one aged for ~10,000 h at 40 mA/cm2.

A typical device ITO|10Å CFx|750Å NPB|750Å Alq|Mg:Ag has a 7 V voltage, 1% EQE (3 cd/A), and ~2500 h half-life, all at the current density (J) of 20 mA/cm2. At the half-life time, the Alq PL loss averages ~13%. After being driven for 11,250 h at 40 mA/cm2, it loses 92% of EL while Alq PL fades by 46% and NPB PL fades by 8% [Fig. 2(a)]. The PL fade is insensitive to the excitation wavelength for either material (350–470 nm in the Alq case). The comparison of the EL spectra before (528 nm, FWHM 100 nm, and CIEx,y of 0.33, 0.56) and after the aging (532 nm, FWHM 112 nm, and CIEx,y 0.36, 0.55) indicates that the spectrum is slightly red-shifted and the red edge is elevated, as usual [Fig. 2(c)].

On average, a prototypical red-emitting device 750Å NPB|300Å Alq+1%DCJTB|300Å Alq has a 7 V voltage (note smaller LEL+ETL thickness), 2% EQE (2.5 cd/A), and ~3300 h half-life, all at 20 mA/cm2. When this devices loses 92% of its PL after being driven for ~10,000 h at 40 mA/cm2, the DCJTB PL fades by 89% if excited directly (500 nm) and by 84% if excited via Alq (430 nm). The Alq PL exhibits a 14% gain (observed at 500 nm) while NPB PL excited at 350 nm loses 7% (observed at 400 nm). Thus, the loss of DCJTB PL in the red cell exceeds the loss of Alq PL in the standard cell by about two times. The comparison of the EL spectra before (624 nm, FWHM 90 nm, and CIEx,y of 0.61, 0.38) and after the aging (612 nm, FWHM 100 nm, and CIEx,y 0.54, 0.43) indicates that the relative intensity of the Alq spectral component (presumably emanating from the ETL) strongly increases causing an effective blue shift of the total spectrum [Fig. 2(e)].

Thus, the addition of 1% DCJTB to the NPB|Alq device decelerates the EL fade ~1.3 times (t50 3300 h versus 2500 h) but it accelerates the PL fade ~4 times (t54PL ~3000 h versus ~11,000 h). This may hint that the presence of DCJTB changes the fade mechanism. Perhaps dopant degradation interferes and gives rise to more efficient quenchers or a larger amount thereof. DCJTB collects the excitation energy by acting both as a charge trap and a Förster-type energy acceptor. DCJTB is 0.50 V easier to oxidize and 0.53 V easier to reduce than Alq in solution.

An average prototypical green-emitting device 750Å NPB|450Å Alq+0.5%C545T|300Å Alq has a 7.4 V voltage, 2.9% EQE (10 cd/A), and ~1400 h half-life, all at 20 mA/cm2. When this device loses 45% of its EL after being driven for ~500 h at 40 mA/cm2, the C545T PL fades by 26% if excited directly (505 nm) and by 23% if excited via Alq (430 nm). Thus, the loss of C545T PL in the green cell (at t55) exceeds the ~8% loss of Alq PL in the standard cell (at ~t70) by ~3.3 times while the EL losses differ only by ~1.5 times (at 500 h of aging). As reported before, the addition of the so-called stabilizers, such as naphtho[2,3-a]pyrene (NP, Fig. 1), to the LEL extends the half-life ~10 times (at 8% NP) and brings the C545T PL loss (9%) closer to the EL loss (11% after 500 h at 40 mA/cm2).23 The initial EQE is 1.8 times lower at 1–8% NP.

The addition of 2%–15% NP to the undoped NPB|Alq device lacking C545T increases the EQE 1.5 times but the PL and EL behavior persists. Without NP, the EL fades by 32% and PL fades by 9% after 500 h at 40 mA/cm2. As the NP percentage increases from 2% to 15%, the EL loss decreases from 14% to 9% and the PL loss decreases from 7% to 5% (the same 500-h test).

Thus, the addition of 0.5% C545T to the NPB|Alq device accelerates the EL fade ~1.8 times (t50 1400 h versus 2500 h) but it accelerates the PL fade ~7 times (t74PL 500 h versus 3500 h). On the other hand, the addition of ~10% NP to both undoped and C545T-doped NPB|Alq devices decelerates the EL fade ~10 times but it decelerates the PL fade only ~5 times.

Devices having only a thin region of the Alq layer doped with C545T, as in 750Å NPB|50Å Alq+C545T|700Å Alq, behave very similarly to those with thick Alq+C545T LEL. As usual, their half-lives shorten with the increasing dopant concentration, e.g., from 450 h at 1% C545T to 255 h at 2% and 150 h at 4% (40 mA/cm2). Importantly, the dopant PL loss in such cells actually exceeds the 50% EL loss and grows with the increasing concentration from 63% to 68% and 85%, respectively (C545T is excited directly).24 Alq can be excited instead but it is difficult to separate Alq and C545T emission. Nevertheless, the dopant PL loss appears to increase from 63% at 1%–2% C545T to 71% at 4% C545T. This prompts one to conclude that (i) again, C545T may be directly involved in the degradation process (an increase in C545T concentration from 1% to 4% accelerates the EL and PL fade 3 and ~6 times, respectively) and (ii) the closer the dopant is situated to the NPB|Alq interface, where most of the charge recombination and emission takes place, the faster the degradation.

The situation with the DPQA-doped Alq and TBADN devices is analogous24 to the C545T case while others, e.g., TBP, furnish an example of a stable dopant: both EL and PL fade are decelerated ~2 times in the presence of 2% TBP in the TBADN host [Fig. 2(b)].

1H. Aziz and Z. D. Popovic, Chem. Mater. 16, 4522 (2004).

2Z. D. Popovic and H. Aziz, IEEE J. Select. Top. Quantum Electron. 8, 362 (2002).

3S. T. Lee, Z. Q. Gao, and L. S. Hung, Appl. Phys. Lett. 75, 1404 (1999).

4J. Shen, D. Wang, E. Langlois, and J. Yang, Proc. SPIE 4105, 236 (2001).

5J. Shen, D. Wang, E. Langlois, W. A. Barrow, P. J. Green, C. W. Tang, and J. Shi, Synth. Met. 111-112, 233 (2000).

6D. Zou, M. Yahiro, and T. Tsutsui, Synth. Met. 91, 191 (1997).

7M. Yahiro, D. Zou, and T. Tsutsui, Synth. Met. 111-112, 245 (2000).

8T. Yamada, D. Zou, H. Jeong, Y. Akaki, and T. Tsutsui, Synth. Met. 111-112, 237 (2000).

9H. Aziz, Z. D. Popovic, N.-X. Hu, A.-M. Hor, and G. Xu, Science (Washington, D. C.) 283, 1900 (1999).

10D. Y. Kondakov, W. C. Lenhart, and W. F. Nichols, J. Appl. Phys. 101, 024512 (2007).

11F. Papadimitrakopoulos, X. M. Zhang, D. L. Thomsen, III, and K. A. Higginson, Chem. Mater. 8, 1363 (1996).

12J. C. Scott, J. H. Kaufman, P. J. Brock, R. DiPietro, J. Salem, and J. A. Goitia, J. Appl. Phys. 79, 2745 (1996).

13B. H. Cumpston, I. D. Parker, and K. F. Jensen, J. Appl. Phys. 81, 3716 (1997).

14S. A. Van Slyke, C. H. Chen, and C. W. Tang, Appl. Phys. Lett. 69, 2160 (1996).

15M. Ishii and Y. Taga, Appl. Phys. Lett. 80, 3430 (2002).

16Z. D. Popovic, H. Aziz, N.-X. Hu, and P. N. M. dos Anjos, SID Digest, 333 (2000).

17Z. D. Popovic, H. Aziz, N.-X. Hu, A. Ioannidis, and P. N. M. dos Anjos, J. Appl. Phys. 89, 4673 (2001).

18D. Y. Kondakov, J. R. Sandifer, C. W. Tang, and R. H. Young, J. Appl. Phys. 93, 1108 (2003).

19C. T. Brown and B. Owczarczyk, unpublished results.

20C. T. Brown, D. Shukla, K. P. Dockery, and J. R. Lenhard, (Eastman Kodak Company, USA), US 2005222417, 2005.

21C. T. Brown and D. Y. Kondakov, (Eastman Kodak Company, USA), EP 1357613, 2003.

22V. V. Jarikov and D. Y. Kondakov, J. Appl. Phys., submitted (2008).

23V. V. Jarikov, J. Appl. Phys. 100, 014901 (2006).

24R. H. Young, unpublished results.

1