J. S. Kim, M. Granström, and R. H. Friend Cavendish Laboratory, Madingley Road, Cambridge CB3 0HE, United Kingdom

Transcription

1 JOURNAL OF APPLIED PHYSICS VOLUME 84, NUMBER DECEMBER 1998 Indium tin oxide treatments for single- and double-layer polymeric light-emitting diodes: The relation between the anode physical, chemical, and morphological properties and the device performance J. S. Kim, M. Granström, and R. H. Friend Cavendish Laboratory, Madingley Road, Cambridge CB3 0HE, United Kingdom N. Johansson and W. R. Salaneck Department of Physics, IFM, Linköping University, S Linköping, Sweden R. Daik and W. J. Feast IRC Polymer Science and Technology, Durham University, South Road, Durham DH1 3LE, United Kingdom F. Cacialli a) Cavendish Laboratory, Madingley Road, Cambridge CB3 0HE, United Kingdom Received 14 May 1998; accepted for publication 15 September 1998 We report combined studies of the influence of chemical and physical treatments on the properties of indium tin oxide ITO thin films. The ITO films were also used as transparent anodes of polymeric light-emitting diodes LEDs incorporating poly p-phenylene vinylene PPV as the emitter material, with, or without, doped poly 3,4-ethylene dioxythiophene PEDOT as a hole-injection/transport layer. Structures based on a soluble green derivative of PPV, poly 4,4 -diphenylene diphenylvinylene were also tested. We studied chemical aquaregia, degreasing, RCA protocol and physical oxygen and argon plasmas, Teflon, and paper rubbing treatments and, in contrast to recently published work, we find that for Balzer Baltracon ITO, oxygen plasma and not aquaregia yields the highest efficiencies and luminances and the lowest drive voltages. For oxygen-plasma-treated anodes, the device efficiency clearly correlates with the value of the ITO surface work function, which in turn depends on the time of treatment. Interestingly, we find that work-function variations induced by our oxygen-plasma treatment are unchanged after long-term storage in air and in the dark. Unexpectedly, we also find that devices incorporating a PEDOT layer benefit from an appropriate treatment of the ITO surface, for both efficiency and lifetime. The results shed light on the physics of conjugated, organic semiconductors and related devices, and in particular on the presence and the role of an anodic energy barrier on the LEDs mechanism of operation. We also discuss the implications of our integrated experimental study in relation to the modification of the ITO sheet resistance, surface and bulk composition, and surface morphology American Institute of Physics. S I. INTRODUCTION The electroluminescence of conjugated polymers 1 and oligomers, 2 has motivated a vast research activity in the field of solid-state physics and in that of the technology of macromolecules. There is increased interest in studying device stability 3 13 during prolonged operation and in improving the luminous efficiency Lm/W. Although lifetimes in excess of 1200 h corresponding to more than 10 6 C/cm 2 charge density passed through the device for both dc and ac drive, 3 and several thousand hours have been reported, further improvements are needed, and interfaces are considered to be very critical to device performance. Work on light-emitting diodes LEDs consisting of a hole-transport layer and a tris 8-quinolinato aluminium (Alq 3 ) emissive layer has shown, for example, that the barrier height at the holeinjecting contact is a parameter which correlates to device lifetime better than the melting point or the glass-transition a Electronic mail: fc10004@cus.cam.ac.uk temperature of the hole-transport material. 7 Furthermore, the quality of both the anode and the cathode interface with the emissive layer is known to influence the occurrence of anomalous current/voltage (I/V) characteristics, which manifest in polymer LEDs PLEDs in the form of current spikes in the low-voltage region 5 V. 17 Although a large variety of transparent conducting oxides TCO are known, such as tin oxide, indium oxide, zinc oxide, and their combinations, the anode of organic LEDs OLEDs, is in the vast majority of cases a thin layer of a mixed indium tin oxide ITO 9 10 mol % tin oxide in indium oxide or fluorinated tin oxide FTO. ITO has been by far the common choice, due to the availability related to the large use made of it by the liquid crystal display LCD industry, its good transparency better than 90% at 550 nm, and low resistivity, and also to the ease with which it can be patterned. The electronic structure of the transparent conducting oxides has been studied for about 30 years, and is discussed in Refs. 18 and 19. Due to a large variety of preparation and postdeposition processes, however, it is not clear /98/84(12)/6859/12/$ American Institute of Physics

2 6860 J. Appl. Phys., Vol. 84, No. 12, 15 December 1998 Kim et al. how the detailed physical properties differ from sample to sample, and in general, what properties would make a commercially available ITO glass more or less suitable for LEDs fabrication. Among other parameters, the work function and the surface roughness of ITO, which have a crucial effect on the charge injection in the active layers of organic LEDs, and are both significantly affected by thermal and surface treatments, have not been considered in great detail for LCD applications, where virtually no current is injected in the cell. ITO has a relatively large work function between 4 and 5 ev, depending on the treatment, which makes it suitable for injection of holes into the highest occupied molecular orbitals HOMO of the conjugated semiconductors, rather than for injection of electrons into the lowest unoccupied molecular orbital LUMO. For most polymers, however, the HOMO lies more than 5 ev below vacuum, so that there is a significant energy barrier to hole injection into the polymer, which must be responsible for an increase in the drive voltage of the devices. A possible strategy to ease the problem is to insert a hole-injection-transport organic layer, and in fact, dramatic improvement of the lifetime and of the luminous efficiency have been reported by several groups 15,20,21 upon insertion of a conducting polymer between the indium tin oxide anode and the emissive layer, and in particular, for poly 3,4-ethylene dioxythiophene PEDOT doped with poly styrene sulfonate PSS. The group at Kodak has also reported improvements upon incorporation of a very thin film of phthalocyanines 150 Å in contact with the ITO. 16 All these results suggest that the electronic and chemical properties of ITO need to be understood and controlled in a more accurate way than they are for use in LCD applications. Several studies have dealt with the effect of ITO chemical and physical treatment on the electroluminescent properties of organic LEDs, however, to the best of our knowledge none has considered the effect of these treatments when using a conducting polymer hole-injection layer. We present here results regarding these effects. In addition to work-function investigations we report the characterization of the sheet resistance and of the surface morphology of the treated substrates as measured by atomic force microscopy AFM, and the impact of the ITO treatments on the polymer films through optical spectroscopy, in particular, with the help of absolute photoluminescence PL efficiency measurements. Electroluminescent EL structures were assessed in terms of their maximum efficiency, brightness, and drive voltage. We also looked at the influence of the poly p-phenylene vinylene PPV precursor thermal conversion on the interface properties. HCl is released during the conversion and this is known to etch the ITO and possibly to dope the polymer with the side products of this reaction. 26,27 We have approached the problem by comparing results for singlelayer devices incorporating either PPV or a soluble green derivative of PPV which does not require thermal conversion from a precursor, namely, poly 4,4 -diphenylene diphenylvinylene PDPV as the emitting layer. For all the device structures ITO/PPV/Ca, ITO/PDPV/Ca, and ITO/ PEDOT/PPV/Ca, we find that the best performance in terms of efficiency at high current densities and maximum luminance, is achieved, among the single treatments, with an oxygen-plasma treatment of 10 min, which corresponds to the maximum measured value of the work function. Interestingly, a combined treatment oxygen-plasma/aquaregia helps to reduce even further the turn-on voltage for the current and to slightly raise the efficiency at low current density, but this is not a significant advantage, since these devices are less stable both at low and high current densities, at which the efficiency also drops more rapidly. II. EXPERIMENTS A. Device preparation 1. Surface treatment We used four different sets of processing techniques for treating the surface of the ITO substrates, before spin coating the polymer films, and these were: a mechanical cleaning, b wet cleaning, c dry cleaning, and d combined cleaning processes. The substrates were received from Balzer and cut into mm samples by UQG Cambridge. For LED fabrication we also etched 2 mm on each side by dipping the samples in a HCl bath 5 wt%,60 C, after protecting the central area with commercial resist for conventional photolithography. This is in order to avoid the formation of shorts subsequent to wiring of electrical contacts to the cathode by use of silver paint. a. Mechanical cleaning. The ITO substrates were first rubbed with ethanol impregnated paper or Teflon specimens for 5 min. They were then dipped in an ultrasonic bath of propan-2-ol IPA for 30 minutes in order to remove any surface contamination caused by paper or Teflon, and finally dried in a flow of nitrogen. b. Wet cleaning. We used three different processes: b1 Ultrasonic degreasing The substrates were washed in an ultrasonic bath of acetone high performance liquid chromatography HPLC grade for 30 min, followed by the IPA HPLC grade for 30 min at room temperature. b2 RCA treatment 28 We prepared the RCA protocol solution by adding NH 3 70 wt % and H 2 O 2 30 wt % into distilled water in a ratio of 1:4:20. The substrates were first dipped into the RCA solution at 60 C for 30 min, and then rinsed twice in distilled water. In order to eliminate the traces of water we then rinsed some of substrates in IPA HPLC grade. All substrates were then dried in a flow of nitrogen. b3 Aquaregia treatment A dilute aquaregia solution was made of HNO 3 70 wt %, HCl 37 wt % and distilled water with a ratio of 1:3:20. The substrates were dipped into the aquaregia ultrasonic bath at room temperature for times ranging from 10 to 30 min. The ITO substrates were then rinsed in distilled water and finally dried with nitrogen. c. Dry cleaning. We used two different plasmas: c1 Oxygen plasma Pure oxygen was fed to a commercial plasma etcher at a pressure of mbar in a 10 1 mbar vacuum, at room temperature. The applied forward and reflected powers were 400 W and less than 5 W, respectively. c2 Argon plasma This was used at an argon pressure of mbar after evacuating the chamber to a base level of less than mbar at room temperature.

3 J. Appl. Phys., Vol. 84, No. 12, 15 December 1998 Kim et al The forward and reflected powers were 60 and 0 W, respectively. For both treatments, we varied the exposure time from 5 to 15 min. d. Combined treatments. We combined aquaregia and oxygen-plasma treatments, because these two treatments allowed higher current densities and EL efficiency in PPV single-layer LEDs when used alone. For these, we used the optimized process times found for b3 and c1 above. The ITO-coated glass substrates used in this study were purchased from Balzer Baltracon mm thick. 2. LED fabrication The substrates treated as described in Sec. II A 1 above were used to fabricate three different types of devices, namely, single-layer PPV and PDPV and double-layer PEDOT/PPV LEDs. PDPV is a green-emitting soluble derivative of PPV and soluble in common organic solvents such as chloroform or toluene due to the presence of the two phenyl groups on the two vinylene carbons. 29 For the doublelayer LEDs a modified water dispersion of poly 3,4-ethylene dioxythiophene doped with poly styrene sulfonate Bayer Corp., Germany 30,31 was used as a hole-transporting material. We first spin-coated PEDOT on the surface-treated ITO substrates, and then prepared PPV films spun on top of the PEDOT by conversion of tetrahydrothiophenium THT leaving the group precursor in standard condition 300 C, 10 h under vacuum. 1 The very poor solubility of our PEDOT in the PPV-precursor solvent methanol is crucial for this step. For all devices, calcium cathodes 100 nm with an additional encapsulating layer of aluminum 100 nm were thermally evaporated at 10 6 mbar, to give an active area of about 3 mm. 2 All processing of Ca-based LEDs was carried out in a dry box under a nitrogen atmosphere with less than 20 ppm residual water vapor and 5 ppm oxygen. The chemical structures of all polymers and LED configurations used in this study are shown in Fig. 1. FIG. 1. Summary of the device configurations and chemical structures of a poly p-phenylene vinylene PPV, b poly 4,4 -diphenylene diphenylvinylene PDPV, and c poly 3,4-ethylene dioxythiophene PE- DOT. B. Device characterization 1. Surface treatments The sheet resistance was measured with the four-point probe method. 32 The work function and the chemical composition of the samples were measured by ultraviolet photoelectron spectroscopy UPS and x-ray photoelectron spectroscopy XPS, respectively. 33 Samples were prepared in Cambridge and sent to Linköping in a nitrogen-filled box for work-function measurements. The UPS measurements were carried out in an instrument of our own design and construction while the XPS measurements were carried out in a Scienta 200 x-ray electron spectrometer, both with a base pressure better than mbar. For UPS, unfiltered ultraviolet helium I radiation 21.1 ev from a VG differentially pumped resonance lamp was used while for XPS monochromatized Al K ev radiation was used to photoionize the samples. Energy calibration was performed on gold, with the Au (4f 7/2 ) line at 83.8 ev below the Fermi level and the Fermi level at 0 ev in the UPS. The work function of the samples was estimated from the cutoff of the secondary electron energy distribution in the UPS spectra. 34,35 In the UPS measurements the resolution of the analyzer was set to 0.1 ev. The x-ray spectra were recorded at an x-ray power of 500 W 15 kv, 33 ma, using a 0.8 mm slit and a pass energy of 150 ev resulting in a full width at half maximum FWHM of the Au (4 f 7/2 ) line of 0.65 ev. A commercial atomic force microscope in contact mode NanoScope II, Digital Instruments Inc. was used to investigate the morphological features. We also probed the hardness of the ITO coating after treatment. These measurements were performed by oscillating the z-piezo in the microscope and measuring the deflection of the tip caused by this vertical movement of the sample. The tip and parameters such as setpoint and oscillation amplitude were identical for all samples. Note that for harder substrates the deflection is higher as the penetration of the tip into the surface is reduced. The deflection is finally expressed in terms of the force acting on the tip by taking into account the force constant and the geometry of the cantilever. Optical absorption spectra nm of the polymer films spun onto either quartz or ITO substrates were measured using a Perkin-Elmer Lambda 9 spectrophotometer. Photoluminescence and electroluminescence spectral characterization were performed by means of an Oriel charge-coupled device matrix spectrograph, with a spectral resolution better than 5 nm. For photoluminescence, the films were excited using the output of an argon-ion laser in the visible nm or UV range 351 and 364 nm at room temperature. The PL efficiency of the films was calculated by comparison of the absorbed laser light with photoluminescence emission, using an integrating sphere, purged with nitrogen to avoid photo-oxidation. 36

4 6862 J. Appl. Phys., Vol. 84, No. 12, 15 December 1998 Kim et al. TABLE I. Work function, sheet resistance, and root-mean-square rms surface roughness of ITO for the different surface treatments. Surface treatment Work function ev Sheet resistance /sq Surface roughness nm As-received a Mechanical Paper rubbing Teflon rubbing b Wet cleaning Ultrasonic RCA no IPA Aquaregia (10 /20 /30 ) 4.6/4.3/ /23.5/ /8.4/8.8 c Dry cleaning O 2 plasma (5 /10 /15 ) 4.35/4.75/ /15.0/ /1.4/2.1 Ar plasma (5 /10 /15 ) 4.5/4.5/ /17.3/ /15.4/23.0 d Combined* Aquaregia 20 /O 2 plasma O 2 plasma 10 /aquaregia LEDs Devices fabricated as described in Sec. II A 2 were transferred under nitrogen to a chamber which was then evacuated to a pressure of less than 10 2 mbar. Basic characterization involved measuring the device current and light output as a function of the applied voltage. The voltage was supplied by a Keithley 230 voltage source, and the current was measured by a Keithley 195A digital multimeter. The luminous output was measured with a calibrated silicon photodiode. films prepared on top of treated ITO or quartz with PPV or PEDOT/PPV layers. Here, the frontside denotes a measurement performed with laser light incident on the polymer side of the sample, while the backside refers to a measurement with laser light incident on the glass side of the sample. The two measurements probe different regions of the polymer III. RESULTS A. Properties of surface treated ITO Work function, root-mean-square rms surface roughness calculated over the whole imaged surface, and sheet resistance values are summarized in Table I. Note that the highest work function, and the lowest surface roughness and resistance are found with a 10 min oxygen-plasma treatment. The value of the sheet resistance is significantly increased for the oxygen-plasma 10 min /aquaregia 20 min treatment. Note that we did not measure appreciable differences in the average thickness of the ITO layers, subsequent to the treatments, as measured by profilometer Dektak IIa and by variation in the absorption coefficients. Figure 2 shows AFM images of differently treated ITO surfaces in comparison with as-received ITO. Significant changes in surface morphology are found only for aquaregia and argon-plasma samples, with increasing roughness as a function of treatment time. The highest roughness, 23 nm, is observed with argon-plasma treatment 15 min, which also determines a morphology dominated by columnar formations randomly scattered over the surface. The highest hardness was found for the oxygen-plasma-treated samples nn as compared to aquaregia 7.15 nn. We will comment on this in the light of the PL efficiency of the samples treated with combined treatments: oxygen-plasma/ aquaregia 9.38 nn and aquaregia/oxygen-plasma 9.54 nn. No significant differences in absorption and PL spectral shapes are observed for both PPV and PEDOT/PPV structures spin coated on either ITO or quartz. Table II shows the measured PL quantum efficiencies performed on polymer FIG. 2. AFM images of the ITO substrates for different surface treatments.

5 J. Appl. Phys., Vol. 84, No. 12, 15 December 1998 Kim et al TABLE II. Measured PL efficiency for the various samples studied. Measured PL efficiency % PPV PDPV PEDOT/PPV Surface treatment ITO front ITO back Quartz ITO front ITO back Quartz ITO front ITO back Quartz As-received Ultrasonic Aquaregia (20 ) O 2 plasma (10 ) Ar plasma (15 ) Aquaregia 20 /O 2 plasma O 2 plasma 10 /aquaregia and serve to understand the extent of the interaction of the emitting dipoles with the ITO layer. For each case, ITO and quartz are treated in the same way. In Table III we report selected data on the surface and bulk relative composition of the ITO as from photoemission experiments. B. EL properties of LEDs The EL of single-layer ITO/PPV/Ca and ITO/PDPV/ Ca devices and double-layer ITO/PEDOT/PPV/Ca diodes, fabricated with different ITOs were characterized by measuring the current density and light output as a function of the applied voltage, and by calculating the efficiencies in Cd/A and in Lm/W luminous efficiency. The results are summarized in Tables IV and V and Figs We also consider the diodes turn-on voltage, here defined as the voltage required for a current density of 10 ma/cm 2. Figure 3 a shows the EL efficiency versus current density J, and the luminance L versus drive voltage (V), and inset curves for different exposure times to oxygen plasma. The corresponding luminous efficiency versus luminance curves are shown in Fig. 3 b. Note that the 10 min oxygen plasma leads to the highest luminance and the maximum efficiencies 0.24 Cd/A and 0.06 Lm/W, whereas the optimum treatment time for argon plasma is 15 min. The IVL characteristics of aquaregia-treated devices not reported here, show maximum efficiencies for 20 min treatment time 0.2 Cd/A and Lm/W, although the maximum current density for a fixed voltage is observed with 30 min treatment time and corresponds to a maximum of the surface roughness and resistance. For samples treated according to the RCA protocol, we find that a final IPA rinse after the RCA degrades significantly the EL performance. Compared to RCA without IPA rinse, the current density and luminance are decreased by a factor of 3, leading to lower efficiencies. Figure 4 shows the current density versus drive voltage a and the EL efficiency Cd/A versus current density b characteristics of single-layer ITO/PPV 240 nm /Ca LEDs fabricated with mechanical, wet, and dry cleaning of the ITO substrates. Figure 5 shows similar characteristics of single-layer ITO/PPV 260 nm /Ca devices for the combined cleaning processes oxygen-plasma/aquaregia and aquaregia/oxygenplasma. Note that there is a significant difference in the EL performance depending on the sequence of the two treatments. If the oxygen plasma precedes the aquaregia oxygenplasma/aquaregia, the EL performance is dramatically improved, with luminance up to two orders of magnitude higher than that of the samples treated with aquaregia first and later with oxygen plasma aquaregia/oxygen-plasma. We also find, however, that stability at high current densities is not particularly good for these oxygen-plasma/aquaregiatreated samples, and we observe a rapid reduction in effi- TABLE III. The chemical composition in the bulk and on the surface of the studied samples. The work function of the surface is also reported for convenience of comparison. Sample Work function ev In/Sn bulk In/Sn surface In Sn /O bulk In Sn /O surface Contaminations As-received Ca 3%, C 40% traces of N Paper rubbing Ca 2%, C 41% traces of N Teflon rubbing Ca 0.7%, C 38% traces of N Aquaregia (10 ) C 35% traces of N Aquaregia (20 ) C 41% traces of N Aquaregia (30 ) C 30% traces of N O 2 plasma (10 ) C 30% Ar plasma (5 ) C 17.3% Ar plasma (10 ) C 19% Ar plasma (15 ) C 18% Aquaregia 20 /O 2 plasma C 15% O 2 plasma 10 /aquaregia C 19.5%

6 6864 J. Appl. Phys., Vol. 84, No. 12, 15 December 1998 Kim et al. TABLE IV. Turn-on voltage (V), maximum EL efficiencies of ITO/PPV 240 nm /Ca 100 nm singlelayer devices for different surface treatments of the ITO substrates. The turn-on voltage is defined as the voltage required for a current density of 10 ma/cm 2. ITO/PPV 240 nm /Ca ITO surface treatment Turn-on V V Cd/A Efficiency Lm/W a Mechanical Paper rubbing Teflon rubbing Ultrasonic b Wet cleaning RCA no IPA Aquaregia (20 ) c Dry cleaning O 2 plasma (10 ) Ar plasma (15 ) d Combined a Aquaregia 20 /O 2 plasma O 2 -plasma 10 /aquaregia a Indicates a PPV layer of 260 nm thickness. ciency for current densities of 200 ma/cm 2 or greater, as shown in Fig. 5 b. The characteristics of ITO/PDPV/Ca LEDs are reported in Fig. 6. Note that the drive voltages necessary for EL 20 V are now significantly higher than for PPV. The aquaregia/ oxygen-plasma treatment gives the highest current density and luminance, in contrast with PPV-based devices Fig. 5 where the oxygen-plasma/aquaregia treatment shows the best performance. However, we find that the 10 min oxygenplasma treatment still gives the highest maximum EL efficiencies of PDPV-based single-layer LEDs 0.1 Cd/A and 0.01 Lm/W. The characteristics of PEDOT 100 nm /PPV 240 nm double-layer LEDs are reported in Fig. 7. Note that the current density for both aquaregia and aquaregia/oxygenplasma increases more slowly than that for the oxygenplasma treatment, in the high-voltage regions. Correspondingly, the highest current density 3700 ma/cm 2 is obtained with the oxygen-plasma treatment, in concomitance with the highest luminance 2000 Cd/m 2 at 14 V. Although the oxygen-plasma/aquaregia treatment gives the highest maximum EL efficiencies, the 10 min oxygen plasma shows relatively stable and slowly decreasing efficiencies up to a current density of 2500 ma/cm 2. IV. DISCUSSION First, we discuss the properties of the treated surfaces work function, sheet resistance, and morphology, and of the optical properties of the polymers on top of those PL efficiency,, before considering the properties of the LEDs. A. The work function and sheet resistance of treated ITO The results we obtained confirm that the electronic properties of ITO can be significantly modified by different surface treatments, which then play an important role in determining the EL performance of LEDs. This is because the hole injection into the active layer is mainly influenced by the properties of the ITO anode, whereas electron injection is determined by the cathode. This is to be related to the electrode work function, although there are indications that other parameters are important as well. 37 It is, therefore, important to note that not only does the ITO work function depend strongly on treatment methods but also on treatment times. For example, the highest ITO work function is obtained by treating the substrates for 10 min with oxygen plasma and is 0.25 ev higher than that of untreated ITO. This increase produced by oxygen plasma is in qualitative agreement with TABLE V. Turn-on voltage (V), maximum EL efficiencies for both ITO/PDPV 210 nm /Ca 100 nm single-layer devices and ITO/PEDOT 100 nm /PPV 240 nm /Ca 100 nm double-layer devices, for the most significant treatments of ITO. ITO/PDPV/Ca ITO/PEDOT/PPV/Ca ITO surface treatment Turn-on V V Efficiency Efficiency Turn-on V Cd/A Lm/W V Cd/A Lm/W Ultrasonic Aquaregia (20 ) O 2 plasma (10 ) Aquaregia 20 /O 2 plasma O 2 plasma 10 /aquaregia

7 J. Appl. Phys., Vol. 84, No. 12, 15 December 1998 Kim et al FIG. 3. EL characteristics of ITO/PPV 240 nm /Ca single-layer devices with oxygen-plasma-treated ITO substrates as a function of the exposure time: a EL efficiency vs current density and luminance vs drive voltage inset, and b luminous efficiency vs luminance. that reported by Wu et al. 22 and Fujita et al., 23 although the values are different because of different ITO and plasma parameters, such as power and treatment time. The effect has been attributed to removal of organic surface residues, changes of surface chemical compositions, 22 or adsorption of oxygen at the surface. 15 It is, in fact, well known that appropriate adsorbed molecules on the surface of a metal can change its work function. 38 To the best of our knowledge, however, it is not unambiguously known that the work function of oxygen-plasma-treated ITO depends on the treatment time, as shown in Table I. We observe that several factors contribute to the final value of the work function. These include the concentration of the different species forming the ITO namely, metallic In and Sn, SnO, SnO 2, and In 2 O 3 ), but also the relative amount of the crystalline and the amorphous phases, as reported by Ishida and co-workers. 39 Although several authors have interpreted their results by focusing on the variation of the oxygen concentration, 15,22 the data we report in Table III show that such an analysis is not able to provide a full explanation and that a complete interpretation of the effects FIG. 4. a Current density vs drive voltage and b EL efficiency vs current density curves of ITO/PPV 240 nm /Ca single-layer devices for mechanical, wet, and dry cleaning of the ITO substrates. reported here requires a detailed structural and compositional analysis beyond the scope of this work. Furthermore, we note that: a Despite the short-term instability of the surface potential of oxygen-plasma-treated samples reported by the Philips group, 15 our samples maintained significant differences over several months when kept in nitrogen or air in the dark. In particular, the work function of the 10 min oxygen-plasma-treated samples was found to be constant over this period. This may be due to the extreme conditions of our treatment 400 W, several minutes and possibly to more permanent alterations of the structural features in particular, the relative concentration of crystalline and amorphous phases. b There must be more than one mechanism involved, since the work-function variation is not monotonic with treatment time. Note that the fact that the work-function evolution is not monotonic with time must not be confused with lack of reproducibility of the results. c The rubbing procedure with either Teflon or paper produces a substantial lowering of the work function and a concomitant increase of the tin and oxygen concentrations both in the surface and in the bulk.

8 6866 J. Appl. Phys., Vol. 84, No. 12, 15 December 1998 Kim et al. FIG. 5. a Current density vs drive voltage and b EL efficiency vs current density curves of ITO/PPV 260 nm /Ca single-layer devices for the combined treatments oxygen-plasma/aquaregia or aquaregia/oxygen-plasma of the ITO substrates. d All the argon-plasma-treated samples show a significant enrichment in the In content both in the bulk and, in particular, on the surface. e All the samples are affected by significant carbon contamination, of probable atmospheric origin. This, however, has no apparent effect on the work function of the samples. f Note that the ability of the oxygen-plasma treatment to induce a very high work function depends also on the history of the samples. Compare, for example, the aquaregia/ oxygen-plasma, with the oxygen-plasma/aquaregia. In the first case, where the samples are first exposed to the action of the aquaregia bath, the work function is only 4.6 ev, even though the final treatment is 10 oxygen plasma, whereas if the samples are first treated with oxygen plasma the final work function is much closer 4.7 ev to the value of the single-treatment oxygen-plasma samples 4.75 ev. This indicates in the first instance that an even higher work function of the ITO may be attainable if starting from an asreceived layer with different chemical or structural composition, and in a second instance also that the aquaregia bath reduces the potential for raising the work function by further treatments. FIG. 6. a Current density and luminance vs drive voltage and b EL efficiency vs current density curves of ITO/PDPV 210 nm /Ca single-layer devices with different surface treatments of the ITO substrates. As for the sheet resistance, we note from Table I that both RCA with no IPA final rinse and aquaregia increase the sheet resistance to 20 and 29 /sq, respectively compared to 16 /sq for as-received ITO, as expected for a progressive etching of the ITO. In contrast, the impact of both ultrasonic and oxygen-plasma treatments is not as significant and these procedures can even decrease slightly the sheet resistance, leading to an increase in the conductivity of ITO. The slight decrease in the sheet resistance by oxygen plasma is also in contrast to that obtained by Fujita and co-workers, 23 who report an increase of about 1.5 times with respect to that of their as-received ITO. B. Morphology The most significant changes in surface morphology are found for aquaregia and argon-plasma samples see Fig. 2, which also show a roughness increase with time of treatment. This indicates that the procedures produce some selective etching of the surface though with unappreciable variation of the average thickness, and are then likely to alter the surface composition. The roughness increase of aquaregiatreated samples is considerable, ranging from 4 nm 10 min to9nm 30 min, and it is even more remarkable following argon-plasma treatment from 11 nm 5 min to 23 nm 15

9 J. Appl. Phys., Vol. 84, No. 12, 15 December 1998 Kim et al FIG. 7. a Current density vs drive voltage and b EL efficiency vs current density curves of ITO/PEDOT 100 nm /PPV 240 nm /Ca double-layer devices with different surface treatments of the ITO substrates. min. Although tin oxide is generally more resistant than In oxide to wet etches such as HCl-based ones, the XPS data of Table III seem to indicate that the columnar formations can be interpreted as indium-oxide formations evident in Fig. 2, less efficiently removed than tin oxide by the dry etch. The negligible modifications of the work function only 0.05 ev for the 15 min argon plasma: see Table I, which accompany this surface etching, however, could also be attributed to clusters or crystalline grains with an improved resistance to etching by virtue of the specific crystallographic orientation the argon-plasma bombardment is, in fact, an anisotropic process. In contrast to the argon plasma, the oxygen plasma produces the smoothest surfaces, with a roughness of only 1.4 nm 5 and 10 min. Note also that these samples show the best EL performance in our LEDs in terms of both efficiencies and luminance. We consider that this improvement is to be related to both the roughness reduction and also the changes in work function and sheet resistance. C. The effect on the PL efficiencies The radiative efficiency of a chromophore depends on the environment it is embedded in, and in particular, on the joint photon density of states available for the radiative transition, which can influence the intrinsic radiative rate. The nonradiative rate can also be modified, depending on the number and the importance of the nonradiative channels accessible by the excited atom or molecules, with vibrational deactivation and energy transfer to other molecules or nonradiative centers being the most probable mechanisms. It is, then, natural to expect a variation of the efficiency with which excited polymer segments will reemit, for the different treatments. The PL efficiency photons emitted per photon absorbed of PPV on ITO 15% is considerably lower than that on quartz 34%. This is due to both the chemical interaction between the hydrogen chloride released during the thermal conversion of the PPV precursor and ITO, and also to the quenching of the radiating dipoles by the conducting surface of ITO. The PL efficiency for the excitation incident on the surface of the polymer at the air interface is also higher than for the case where the excitation is entering the PPV layer from the other side i.e., passing through the ITO. This is due to the increased distance, and hence, reduced quenching of the emitting molecules, in qualitative agreement with Greenham et al. 40 The use of a PEDOT layer in between the ITO and the emissive polymer does not improve the result significantly 13%. Note that we used very thick layers of PPV nm so that only a very small fraction of the incident beam is absorbed in the PEDOT, as confirmed by the very small reduction of the PL efficiency of the same stack PEDOT/ PPV on quartz ( 22%). The lack of improvement upon incorporation of the PEDOT layer is unexpected since this should increase the distance of the emitting chromophores from the ITO, and therefore, reduce both chemical and physical quenching interactions. This effect may be due to either a doping of the PPV with PSS, or to the residual though very small solubility of the PEDOT in methanol. We envisage here the possibility of significant improvements on these results through optimization of the PEDOT layer. We observe, however, that the LEDs performance is improved by the incorporation of the PEDOT. Note that we used a pure PPV emissive layer for our studies, and therefore, our PL efficiencies are not directly comparable with those reported earlier by Cambridge Display Technology CDT, 14 relative to a high efficiency PPV copolymer. Although in principle it is also possible that the presence of the PEDOT layer induces a different state of order or disorder and morphology in the PPV layer, which affects the radiative or nonradiative decay of excitons, we did not find evidence for this. We notice from Table II that different surface treatments also affect the PL efficiency of PPV on quartz. For example, for a PPV/PEDOT/quartz structure, the PL efficiency decreases from 22% for untreated samples to 16% for aquaregia samples. This is likely to be related to the presence of contaminants in the spectrosil B substrates we used, which are typically in the range of the several tens of ppb parts per billions in the case of K, Na, Al, Ca, Mg, Cu, Zn, Zr, Ti, Cr, and Li, and typically, 50 ppm parts per million for Cl data courtesy of UQG Ltd., Cambridge, U.K..

10 6868 J. Appl. Phys., Vol. 84, No. 12, 15 December 1998 Kim et al. The highest values of the PL efficiency front excitation were obtained with the oxygen-plasma/aquaregia treatment, which for example, increases the efficiency from 15% for as-received ITO to a maximum of 18% 20% for PPV on ITO. In contrast, the aquaregia/oxygen-plasma treatments gave lower efficiencies of 13% and 12% for PPV and PEDOT/PPV, respectively. It is also interesting to consider the difference in the PL efficiency between the front- and backside as this gives indications on the strength of the interaction between the PPV and the ITO. We observe, in particular, the increases of this difference for the aquaregia treatments except the oxygenplasma/aquaregia treatment, which indicates that the aquaregia-treated surface is more reactive than others. Our data also suggest, however, that oxygen-plasma treatments are very effective at protecting the ITO layer from the effect of aquaregia etching. We relate this to the increased hardness of the oxygen-plasma-treated substrates and propose that the oxygen-plasma effectively passivates the surface against oxygen on the basis of the constant UPS work function versus storage time in air and dark and aquaregia also on the basis of the difference in the average roughness of aquaregia/oxygen-plasma and oxygen-plasma/aquaregia samples. In order to eliminate the influence on the PL efficiency of the conversion conditions of the PPV, and in particular, the In doping of the luminescent polymer due to the hightemperature etch of ITO by the HCl produced as a byproduct of the thermal conversion, we used a PPV derivative, PDPV, which is soluble in the conjugated form, and hence, is not affected by the serendipitous doping mentioned above. For this, we observe, in fact, that the quenching of the luminescence on ITO substrates relative to silica is less important than for PPV Table II. D. EL properties of diodes fabricated with the treated ITO 1. ITO/PPV/Ca single-layer devices An important point to ascertain in the understanding of the physics of polymer LEDs, with a view to the electrode surface engineering, is whether the diode performance correlates with the surface work function. The charge injection process is very sensitive to the chemical nature of the interfaces, and therefore, we did not expect to be able to rationalize the results of treatments involving very different chemical processes, in terms of the work function alone, as also proposed by Andersson and co-workers. 37 However, we propose that this should be possible for treatments of the same nature where the only parameter varying is the time of the treatment as for oxygen plasma. Furthermore, we notice that there is a fundamental difference between mechanical or wet processes such as degreasing or aquaregia, which are likely to leave traces of contaminants hard to remove as evident from the last column of Table III and dry treatments, which allow the preparation of surfaces with fewer contaminations self-cleaning, and even capable of passivating the surface with respect to some chemical reactions. In particular, we expect that a better quality of the surface passivation should allow a better correlation of the diode performance to the work-function data. The results reported in Table IV confirm that among the single treatments the oxygen-plasma treatment 10 min yields the lowest turn-on voltage and the highest EL efficiency, and that these correlate with the highest measured value of the work function 4.75 ev below the vacuum level. We consider that this is a good indication that the injection barrier is reduced for this type of surface modification, even though the difference between the ITO work function and the highest occupied molecular orbital, ev for PPV, cannot be directly interpreted as the injection energy barrier for the possible occurrence of polymer doping phenomena in the vicinity of the interface. We also observe from Fig. 3 that both EL efficiencies correlate with the work function for the different times of processing in plasma, with higher values corresponding to the higher work function. For the other noncombined treatments, we observe the best performance for devices fabricated with the 10 min oxygen plasma: the turn-on voltage at 10 ma/cm 2 is reduced from 13 V for the ultrasonic degreasing to only 7 V. Although this may seem a high voltage in absolute terms, it is relative to a thick 240 nm active layer, and corresponds to an average field of only 0.42 MV/cm. The maximum efficiencies are also improved with respect to the as-received or ultrasonically degreased samples, from to 0.24 Cd/A and from to 0.06 Lm/W, respectively. A luminance of 1000 Cd/m 2 was obtained at 11.5 V and at a current density of 700 ma/cm 2. In addition to the increased work function 4.75 ev, we attribute these improvements to the decrease in both sheet resistance and roughness 15 /sq and 1.4 nm, respectively. The increase of the work function provides a reduced energy barrier at the ITO/PPV interface, and results in an increased hole injection in the polymer layer. The enhanced conductivity of ITO also reduces the voltage drop over the interface. The low roughness provides the smoothest interface, allowing the PPV layer to adhere more uniformly to the ITO. As mentioned, it is difficult to analyze the role of surface roughness on the EL properties here because of the variation of other parameters, namely, work function and sheet resistance. Argon-plasma-treated samples, however, show a negligible variation of the work function upon treatment, and may then be suitable for such an analysis Table I. Although an increased surface roughness could lead, in principle, to an enhanced electric field and surface area, and hence, to more effective hole injection, our data provide evidence that increasing the roughness degrades the EL performance, at least beyond a certain threshold. We think that this is the result of poor adhesion of the polymer film to the substrate, resulting in a decrease of the effective area available for charge injection. We note here that although the precursor polymer may be fluid enough in solution to form a good interface with the ITO, the thermal elimination of the THT group reduces the volume to about one half, producing a concomitant stretching of the PPV, which is also responsible for a certain degree of in plane order and crystallinity of the layer. 41 We propose that the same phenomenon acts here

11 J. Appl. Phys., Vol. 84, No. 12, 15 December 1998 Kim et al to lift the layer so that only a partial, and possibly unstable contact, is in fact available between the PPV and the anode. We also expect this to be critically dependent on the height and the shape of the features on the surface. Interestingly, in the case of the aquaregia treatment, which produces surfaces with a much smaller roughness, and features which are not as sharp as for the argon-plasmatreated samples, the EL efficiencies increase and the threshold voltage decreases with increasing time and roughness. We attribute these results to the increased surface area of the surface, now forming a good contact with the PPV. When we combine oxygen-plasma 10 min and aquaregia 20 min treatments, we find a significant difference depending on the sequence used. Namely, if the ITO is first treated by oxygen plasma and then by aquaregia, the EL performance is significantly better than for the opposite sequence. Note that these results are consistent with the improvement of the PL efficiency shown in Table II. A disadvantage of this treatment, however, is in the occasional occurrence of current anomalies in the low-voltage region of the IV current voltage characteristics. Furthermore, the efficiency decreases rapidly in the high current density region, 200 ma/cm 2, where oxygen-plasma samples still show high values Fig ITO/PDPV/Ca single-layer devices In these devices, we find again that the oxygen plasma 10 min increases EL efficiencies 0.1 Cd/A and 0.01 Lm/ W, compared with the ultrasonic and aquaregia treatments, as shown in Table V. Although the devices fabricated with the aquaregia 20 min substrates show higher current density in the low-voltage regions, both efficiencies Fig. 6 b are reduced significantly with respect to the oxygen plasma, with low luminance and large current anomalies. Note that turn-on and drive voltages are generally higher than for PPV. The maximum efficiencies 0.09 Cd/A, Lm/W by the aquaregia/oxygen-plasma treatment are three and seven times lower than for PPV devices. Interestingly, the oxygen-plasma/aquaregia treatment, which gives the highest performance for the PPV devices, is relatively unstable in terms of current anomalies, and at high current densities also produces the poorest and most unstable performance for the PDPV devices. This indicates that the use of aquaregia-treated surfaces in direct contact with PDPV makes devices less efficient for light emission. We confirm from the results that oxygen plasma 10 min is the most effective surface treatment for improving the EL performance of both PPV and PDPV single-layer devices. It shows an increase in EL efficiencies, and maintains these increased efficiencies into high current density regions. Most importantly, it also removes low-voltage current anomalies. 3. ITO/PEDOT/PPV/Ca double-layer devices The relatively high value ev of the ionization potential of PEDOT Ref. 33 makes it suitable as a hole-injection/transport layer between ITO and the active polymer of LEDs, and a concomitant improvement of the operational device stability has been demonstrated. 42 Regarding the optimization of the PEDOT anode, a question of considerable importance, and to the best of our knowledge still not addressed, is if and how the surface treatment of the ITO influences the performance of the LEDs, in the presence of a PEDOT layer. Addressing this question helps to understand if the PLEDs efficiency is solely determined by the properties of the interfaces of the active material with the transport layers/electrode, or if instead, the interfaces between the transport materials and the electrode are of some importance as well. The importance of this question, therefore, bears on the physics of the structures and extends beyond the technological interest for improved devices. The data reported in Fig. 7 and in Table V show, indeed, that the EL performance of ITO/PEDOT/PPV/Ca devices is significantly affected by the nature of the ITO surface. For example, the turn-on voltage is reduced from 8 V for the ultrasonic treatment to 4 V for the aquaregia/oxygen-plasma treatment, whereas the maximum EL efficiencies are obtained by the oxygen-plasma/aquaregia treatment 0.33 Cd/A and 0.11 Lm/W, respectively. We consider that this is partly related to differences in the PL efficiencies Table II, and partly to the electronic and transport properties of the ITO/ polymer interface such as the relative position of the relevant energy levels for charge injection and transport, and also the value of the surface sheet resistance. Although the oxygen-plasma/aquaregia-treated samples show again the maximum EL efficiencies, we observed that they have the same problems already found in single-layer diodes, namely: a the efficiency rapidly decreases for current densities 150 ma/cm 2, and b the low-voltage high current density anomaly. Oxygen-plasma treatments 10 min are preferable since they improve the EL performance more effectively with no associated current anomaly and the enhanced efficiencies 0.26 Cd/A and 0.08 Lm/W. As mentioned earlier, these values are not directly comparable with those of CDT 14 because of the different materials used. We note that the performance is improved with respect to single-layer PPV/Ca devices, as already noted in the literature. 21,43,44 This can be attributed to a significant reduction of chemical and physical interaction of the active layer with the anode. No significant differences in the EL spectral shapes are observed between PPV single- and PEDOT/PPV double-layer devices. 4. Current anomalies Many of the polymer LED structure described within this article, and elsewhere, can, for some devices, demonstrate high current peaks at low drive voltages. These current peaks, often referred to as current anomalies, are not associated with EL and are also inconsistent with the usual mechanisms of charge injection at a metal polymer interface. They are characterized by a significant increase in current above a certain threshold voltage, typically, 5 V.The features of this phenomenon appear to be relatively insensitive to the particular polymer being used, and current anomalies occur across a range of different electrode materials. Empirically, it has been observed that the probability of a