Investigations on the stability of poly(phenylene vinylene)-based organic light-emitting diodes Zur Erlangung des akademischen Grades Doktor-Ingenieur (Dr.-Ing.) genehmigte Dissertation von diplomi-insinööri Oili Pekkola aus Helsinki Juli 2017 — Darmstadt — D 17 Fachbereich Material- und Geowis- senschaften Elektronische Materialeigenschaften Investigations on the stability of poly(phenylene vinylene)-based organic light-emitting diodes Genehmigte Dissertation von diplomi-insinööri Oili Pekkola aus Helsinki 1. Gutachten: Prof. Dr.-Ing. Heinz von Seggern 2. Gutachten: Prof. Dr. Matthias Rehahn Tag der Einreichung: 05.12.2016 Tag der Prüfung: 25.04.2017 Darmstadt — D 17 Erklärung zur Dissertation Hiermit versichere ich, die vorliegende Dissertation ohne Hilfe Dritter nur mit den angegebenen Quellen und Hilfsmitteln angefertigt zu haben. Alle Stellen, die aus Quellen entnommen wurden, sind als solche kenntlich gemacht. Diese Arbeit hat in gleicher oder ähnlicher Form noch keiner Prüfungsbehörde vorgelegen. Darmstadt, den 25.04.2017 (Oili Pekkola) 1 Contents 1 Introduction 5 2 Theoretical background 9 2.1 Organic semiconductors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.1.1 Physical properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.1.2 Optical processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2.2 Organic light-emitting diodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 2.2.1 Device physics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 2.2.2 Electrical degradation of organic light-emitting diodes . . . . . . . . . . . . . . . 24 3 Experimental 27 3.1 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 3.2 Sample preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 3.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 3.3.1 Standard methods for material and device characterization . . . . . . . . . . . . 32 3.3.2 Electrical stressing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 3.3.3 Carrier extraction by linearly increasing voltage . . . . . . . . . . . . . . . . . . . 33 3.3.4 Photoinduced absorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 3.3.5 Secondary ion mass spectrometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 3.3.6 Infrared thermography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 4 Experimental techniques for the investigation of PPV-based devices 39 4.1 Removing the cathode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 4.2 Electrical stressing in accelerated conditions . . . . . . . . . . . . . . . . . . . . . . . . . . 41 5 Dark-CELIV investigations on OC3C8-PPV 45 5.1 Previous investigations with dark-CELIV . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 5.2 Origin of equilibrium charge in OC3C8-PPV . . . . . . . . . . . . . . . . . . . . . . . . . . 45 5.3 Oxygen doping of OC3C8-PPV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 5.3.1 Gradual increase of extraction current in present devices . . . . . . . . . . . . . 51 5.3.2 Dynamics of charges in oxygen-doped OC3C8-PPV . . . . . . . . . . . . . . . . . 54 5.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 6 Influence of triplet excitons on device lifetime 59 6.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 3 6.2 Singlet-to-triplet conversion in layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 6.3 Increased triplet exciton concentration in OLED devices . . . . . . . . . . . . . . . . . . 67 6.3.1 Bipolar devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 6.3.2 Unipolar devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 6.3.3 Consequences of a higher concentration of triplet excitons . . . . . . . . . . . . 78 6.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 7 Summary and outlook 91 List of Figures 95 Bibliography 97 Curriculum Vitae 110 4 1 Introduction In late 1980s, Tang and Van Slyke presented an efficient two-layer organic light-emitting diode (OLED) [1]. This device was based on small molecule organic semiconductors and started the development of organic thin-film electroluminescence [2]. A few years later, in 1990, Burroughes and co-workers introduced the first light-emitting diodes that were based on a conjugated polymer, a poly(p-phenylene vinylene) (PPV) [3]. This first polymer-based OLED was realized with precursor- based PPV, but soon afterwards different groups were able to use solution-processed polymers [4]. This enabled an easy processing of the active materials by wet chemical deposition. Since these early findings, the development of OLEDs has been fast. OLEDs reached the market in 1997, as Pioneer Corporation released the first commercial OLED product, a passive matrix display for car audio devices [5]. Nowadays the probably best-known OLED application, an active matrix OLED (AMOLED) mobile phone display, was first released by Samsung Mobile Display in 2007 [6]. Nowadays Samsung uses OLEDs in all their high-end smartphones and many other companies have AMOLED displays in their flagship models, too. OLED TVs reached the market in 2013, and LG is the leading player in the field. In general, OLED display technology has in recent years become a real alternative for LCD displays. Unlike LCDs, OLEDs emit light themselves and do therefore not need a backlight, which enables the realization of extremely thin and light-weight devices. The contrast of OLED displays is generally superior to that of LCDs, and their emission is independent of the viewing angle. In addition to displays, lighting is another area for OLED technology. OSRAM Opto Semiconductors was the first company to announce a commercial OLED white lighting panel in 2010 [7]. Due to their large-area, diffuse emission and the possibility for the realization of very thin, planar and flexible light sources, OLEDs offer plenty of possibilities for new kinds of illumination solutions. OSRAM OLED is currently working intensively with automobile industry, focusing on rear lights as well as interior lighting. Of course, OLEDs do not have only advantages. Up to date, they are more expensive than LCDs. Additionally, blue emitter materials suffer from lifetime issues: their lifetime is significantly lower than that of green and red emitters [8], which leads to changes in color balance. Due to the emissive nature of OLED displays, they are often difficult to use in direct sunlight [7]. The vast majority of the OLED products currently in market are based on small molecules. Small molecules are typically processed by evaporation, which enables the use of multilayered structures. Polymers are deposited by wet chemical methods like inkjet printing or spin coating, and the real- ization of several-layer structures is in general not possible due to the damaging of the underlying layers with the solvents. Additionally, in comparison to their small molecule-based counterparts, polymer-based OLEDs suffer from issues related to their lifetime and efficiency [7], which limits 5 their application in commercial products. Despite the dominance of small molecules in the OLED market, there are ongoing efforts on the development of polymer-based OLED devices for commer- cial use, and they are a lucrative alternative especially for low-cost, printed devices owned by their easy wet chemical processing. Although the intensive research during the last 30 years has already provided the field with extensive knowledge in topics related to the materials as well as device physics, and OLEDs have developed from small-scale laboratory prototypes to a mature, commer- cially available technology, especially polymer-based diodes are still in need of research in order to increase their lifetime to the level of small molecules. The aim of this thesis is to contribute to this research, increasing the knowledge of light-emitting diodes based on conjugated polymers. This work focuses on one of the most important conjugated polymers, poly(p-phenylene vinylene) and aims to gain new insights into the fatigue of PPV-based devices. The thesis begins with the introduction of the relevant theoretical background in the material class of organic semiconductors as well as of organic light-emitting diodes and their functional principles in Chapter 2. After this, special attention is paid to the electrical degradation of OLEDs in form of a short literature review on the most important degradation mechanisms. Chapter 3 focuses on the experimental side of the work. It begins with the introduction of the materials that were used in this thesis and continues with the sample preparation procedure. Finally, the utilized methods are introduced. The results of the thesis are divided into three main topics. First, two methods that were developed for the general investigation of PPV-based diodes are introduced in Chapter 4. Fatigue studies often require access to the active polymer layer after the operation of the device in order to detect possible morphological changes, for example. For this, it is necessary to remove the cathode from the device. Chapter 4 introduces a removal process in which the calcium cathode is etched away with acetic acid. The other topic of this chapter is the aging of OC3C8-PPV-based diodes under accelerated conditions. A scaling law for the operation at different current densities is determined, enabling the operation of the diodes at higher current densities that lead to faster degradation. Chapter 5 concentrates on the stability of OC3C8-PPV during storage in inert atmosphere. Many conjugated polymers are very sensitive towards oxygen, which is why careful encapsulation of the devices is necessary. The chapter focuses on the interaction between oxygen and OC3C8-PPV and shows that residual oxygen is found to diffuse in the polymer even during storage in the glovebox, leading to unintentional p-doping of the active material. The evolution of doping and the dynamics of the hole diffusion in the form of temporal establishment of steady-state conditions in a doped system are investigated with dark-CELIV. The findings suggest that even an unencapsulated storage in inert atmosphere with a low concentration of residual oxygen can lead to reactions in the sensitive polymers. Chapter 6 investigates the other main topic of the thesis, the influence of triplet excitons on de- vice lifetime. In fluorescent OLEDs, only singlet excitons participate in emission. This leaves the 6 triplet excitons - statistically the majority of all formed excitons - to dissipate their energy non- radiatively. The impact of a large density of triplet excitons on the stability of the OLEDs has not been extensively investigated until now. In the present work, the concentration of triplet excitons in the active polymer is increased through the incorporation of a triplet sensitizer material into the polymer. The influence of the increased triplet exciton concentration on the performance and sta- bility of the devices is then investigated; the triplet excitons are found to shorten the lifetime of the diodes significantly. The chapter then continues with the discussion of phenomena that influence the accelerated degradation. Finally, the results and conclusions of the thesis as well as an outlook for futher research are sum- marized in Chapter 7. 7 2 Theoretical background This chapter provides theoretical background on topics that are essential for the understanding of the results presented in this thesis. First, organic semiconductors are introduced, focusing on the general properties of this material class. The chapter then continues with organic light-emitting diodes, describing the physical processes that take place during operation and the different opera- tion schemes with the help of energy diagrams. Finally, special attention is payed to the electrical fatigue of the OLEDs. 2.1 Organic semiconductors Organic semiconductors are commonly divided into two material classes, low-molecular weight small molecules and conjugated polymers. Examples of both types are presented in Fig. 2.1. An important difference between small molecules and conjugated polymers lies in the processability. Small molecules are typically deposited from the gas phase via physical vapour deposition (PVD), which enables the preparation of multilayer structures of desired layer order and thickness. In contrast to this, the thermal stability of conjugated polymers is limited and they are processed out of solution by spin-coating, dip-coating or printing, for example. The advantage of solution- processing is the simple and cheap layer preparation. However, the deposition of multiple layers is challenging as the underlying layers are easily dissolved by the solvent if orthogonal solvents are not available. This limits the use of conjugated polymers to simple device structures. The commercially available OLEDs are therefore typically based on small molecules. In this work, only polymers were used and the focus of this chapter is thus set on those materials. Figure 2.1: Examples of organic semiconductors: polymers a) polyfluorene, b) poly(p-phenylene viny- lene) and the small molecule c) tris-(8-hydroxyquinoline) aluminum. 9 2.1.1 Physical properties Organic semiconductors are carbon-based materials with a conjugated π electron system, where single and double bonds alternate. Fig. 2.2 a) shows a schematic picture of the formation of a double bond. The carbon atoms are sp2 hybridized and the three sp2 hybrid orbitals lie planarly with an angle of 120 ° between them, leaving one unhybridized pz orbital perpendicular to the plane. The sp2 orbitals of adjacent carbon atoms form σ bonds that build the backbone of the material. During the bond formation, the hybrid orbitals split into one occupied, bonding σ and one unoccupied, antibonding σ∗ molecular orbital. Additionally, the unhybridized pzorbitals form a π bond with the p orbital of a neighboring atom. Similar to the formation of the σ bond, also the pzorbitals split into a bonding π and antibonding π∗ molecular orbitals. The σ bond is strong and the energetic difference between the σ and σ∗ molecular orbitals is high, whereas the weaker interaction between the pzorbitals leads to a significantly smaller splitting of the π and π∗ orbitals. Figure 2.2: a) Formation of a double bond in ethene, the simplest conjugated π system. The π bonds are formed between pz orbitals whereas the sp2 orbitals form the σ bond. b) Energy diagram of the bonding and anti-bonding σ and π molecular orbitals. The antibonding molecular orbitals are denoted with an asterisk. HOMO and LUMO are formed by the π orbitals. When several carbon atoms with alternating single and double bonds are joined together in a molecule, their pzorbitals interact, forming new molecular orbitals. They are filled with the pzelectrons starting from the lowest energy so that the bonding π molecular orbitals are completely filled and the antibonding π∗ molecular orbitals completely empty. The π electrons delocalize over the complete conjugation length and a delocalized π electron system is formed. The bond- ing π orbital with the highest energy is called the highest occupied molecular orbital (HOMO) and the antibonding π∗ orbital with the lowest energy the lowest unoccupied molecular orbital (LUMO). The energy gap between HOMO and LUMO in conjugated molecules is typically between 1.5 and 10 3 eV [9], which leads to absorption and emission in the visible range. The energy gap decreases with increasing conjugation length. This offers possibilities to synthetize materials with desired optoelectronic properties by tuning the conjugation length. In polymers, the conjugation length does not cover the complete chain length. Because the overlap of the π orbitals is dependent on the geometry of the macromolecule, defects and distortions in the chain geometry limit the extend of the conjugation. For example, an effective conjugation length of 5-10 monomer units has been reported for MEH-PPV [10]. In addition to the structure of the polymer, the effective conjugation length is affected by synthesis and processing of the material [11]. The interaction between single molecules is van der Waals type and is therefore much weaker than between covalently bonded inorganic semiconductors. As a consequence, the optical and energetic properties of isolated molecules do not differ greatly from molecules in the solid. The weak bonding between molecules also results in lower melting points and reduced hardness. Due to the weak interaction between molecules, the electronic wavefunction is not as extended as in inorganic semiconductors. In polycrystalline and amorphous systems, the charge carriers are localized and no band transport takes place [12]. These systems are characterized by local differences in the conformation and morphology. Due to coulomb interaction, the presence of an excess charge leads to polarization of its environment. Because of the lack of translational symmetry, this polarization will be random, and charges will see different dipole orientations in their environment. This leads to a statistical variation in the energy levels. The energetic distribution of the localized states (the density of states, DOS) is assumed to be Gaussian with a width of approximately 100 meV and was first introduced by Bässler in the Gaussian disorder model [13]: DOS(E) = N p 2πσ2 · exp � − EHOMO/LU MO 2 2σ2 � (2.1) where N is the density of molecules in a solid, σ the width of the Gaussian density of states (the standard deviation) and EHOMO/LU MO the energetical position of the maximum of the HOMO or LUMO density of states. The Gaussian distribution of HOMO and LUMO is illustrated in Fig. 2.3. 2.1.2 Optical processes Upon an excitation of a molecule, an electron is lifted to an excited state and a hole is left on the ground state. The exciton, the resulting electron-hole pair, is bound by Coulomb interaction. There are three types of excitons: Wannier, Frenkel and charge transfer (CT) excitons. The excitons in organic semiconductors, especially in organic light-emitting diodes, are typically Frenkel-type. They have a radius on the order of few nanometers and are localized on one molecule. The Frenkel excitons are strongly bound and have a binding energy as high as 0.5 to 1 eV [9]. Frenkel excitons 11 Figure 2.3: Energetic distribution of electronic states in an organic semiconductor. The distribution of HOMO and LUMO is taken to be Gaussian. have well-defined singlet and triplet states. The properties of these states are introduced in the following. Charge transfer excitons, on the other hand, are excitations in which the electron and the hole lie on adjacent molecules. This type of excitons is of importance in organic solar cells, in which the charge separation takes place at a donor-acceptor interface. The third type of excitons, the Wannier exciton, is typically found in inorganic semiconductors. The basic optical processes in an organic molecule are depicted in Fig. 2.4. The electronic energy levels of the molecule are illustrated with thick horizontal lines. The ground state is the singlet state S0, whereas S1 and S2 represent the first and second excited singlet states. Correspondingly, the first and second excited states of the triplet manifold are denoted with T1 and T2. The thin horizontal lines correspond the vibronic sublevels of each energy level. Radiative processes are illustrated with straight vertical arrows and non-radiative with wavy arrows. Absorption of a photon takes place from the ground state to one of the excited singlet states. There are non-radiative and radiative ways to dissipate the energy that was gained through absorption. Vibrational relaxation and internal conversion (IC) are non-radiative processes. Vibrational relax- ation, shown as wavy downward arrows in Fig. 2.4, takes place within vibrational levels of one electronic state. Vibrational relaxation is a very fast process and is thus very likely to take place immediately after absorption. The relaxation is called internal conversion if vibrational levels of two electronic states overlap; in this case the excited electron transfers from the vibrational level of a higher electronic state to a vibrational level of a lower one. Emission of photons from the first excited singlet state S1 to the ground state, called fluorescence, is a radiative process. Because of the fast relaxation of excitations to the lowest vibrational S1 state, fluorescence generally takes place only from this state. Its characteristics do therefore not depend on the excitation wavelength [14]. Transition between singlet and triplet systems involves a spin flip. A nonradiative transition is called intersystem crossing (ISC), whereas a radiative transition from T1 to the ground state S0 has the name phosphorescence. 12 Figure 2.4: Perrin-Jablonski diagram showing the energetic level scheme of an organic semiconduc- tor, adopted from [9, 12, 14]. IC stands for internal conversion and ISC for intersystem crossing. For details see text. The spins of the two electrons that are forming a singlet state are anti-parallel (see Fig. 2.5). The singlet states have a total spin quantum number of S = 0. An excited state with parallel spins is called the triplet state. It has a total spin of S = 1. As its name denotes, the triplet state is degenerate with three quantum states and a spin multiplicity of 3. The energy of the triplet states is always lower than the corresponding singlet energy. Figure 2.5: Singlet and triplet states on an organic molecule. The singlet state has a zero total spin, whereas the three-fold degenerate triplet state has a spin of S = 1. 13 Transitions that involve a change in spin multiplicity are forbidden. However, the quantum mechan- ical selection rules are weakened by spin-orbit coupling. The coupling is very weak in materials consisting of pure hydrocarbons. In these systems, emission from the triplet state is very weak. However, if heavier atoms like Pt or Ir are involved, the spin-orbit interaction becomes stronger and the spin selection rules that forbid the S − T transition are weakened [12]. Transitions between singlet and triplet states become allowed and the rates for intersystem crossing as well as phospho- rescence increase. Molecules with intense emission from the triplet state, called triplet emitters, are often organo-transition metal compounds [15]. They are utilized in phosphorescent OLEDs. 2.2 Organic light-emitting diodes The following section focuses on the the functional principles of organic light-emitting diodes. The main processes that take place are introduced and explained. The section ends with a short review of the electric fatigue of organic light-emitting diodes. 2.2.1 Device physics A simple polymer-based organic light-emitting diode consists of a thin (on the order of 100 nm) polymer layer sandwiched between two electrodes. At least one of the electrodes, normally the bottom electrode, is transparent, which allows for light outcoupling. Fig. 2.6 shows a schematic picture of the processes related to the operation of an OLED. The functional principle of OLEDs can be divided into five processes: injection of charge carriers from the contacts into the organic semiconductor (1), their transport in the organic material (2), formation of excitons (3), their diffusion (4), and the (radiative) recombination (5). In the following, the single steps are explained in more detail. Charge carrier injection The density of intrinsic charge carriers in organic semiconductors is low. Consider an organic semi- conductor with a bandgap of 2.5 eV and an effective density of states N0 = 1021 cm−3. The density of charge carriers can be estimated with Boltzmann distribution ni = N0 exp � −EG 2kB T � to be ni < 1 cm−3. Such a density of intrisic charge carriers leads to an extremely low conductivity. In order to over- come this limitation, the charge carrier density can be increased by different means. Charge carriers can be created optically, as is done in the photo-CELIV technique (Chapter 3.3.3), by electrochemi- cal doping or injection from the contacts. The latter is the process that enables device operation in organic light-emitting diodes. During operation, an external voltage is applied and holes are injected from the anode and elec- trons from the cathode. Fig. 2.7 illustrates the contact formation between a metal and an organic 14 Figure 2.6: A schematic illustration showing the working principle of an OLED. The individual pro- cesses are the injection of charge carriers from the contacts into the organic semiconductor (1), their transport in the organic material (2), formation of excitons (3), their diffusion (4), and the (radiative) recombination (5). semiconductor. The energetic states of the metal and the semiconductor are shown in Fig. 2.7a for the case prior to contact. Ev ac denotes the vacuum level; ΦM and EF,M are the work function and the Fermi level of the metal, respectively. Correspondingly, the organic semiconductor has the Fermi level EF,SC and the work function ΦSC . EI is the ionization potential and EA the electron affinity of the semiconductor. When the metal and the semiconductor are brought into electric contact (Fig. 2.7b), their Fermi levels become equal through diffusion of electrons from the lower work function material into the material with the higher work function. The diffusion over the interface leads to the formation of a charge carrier reservoir that is observed as the bending of the band structures. Due to the high charge carrier density, the reservoir in the metal reaches only few monolayers, whereas its size can exceed 100 nm in the organic semiconductor that has a low intrinsic charge carrier density. The formation of the space charge caused by the charge carrier reservoir generates an electric field that accounts for a drift current in the opposite direction to the diffusion current. In thermodynamic equilibrium, the drift and diffusion currents compensate each other and the overall current equals zero. If interface effects are neglected, the injection barrier for electrons ΦB,e is defined as the difference between the work function of the electron injecting contact ΦM and the electron affinity of the organic semiconductor EA; for holes, the barrier ΦB,h equals the energetic difference between the work function of the contact metal ΦM and the ionization potential of the organic material EI : ΦB,e = ΦM − EA and ΦB,h = EI −ΦM (2.2) 15 Figure 2.7: A schematic illustration of contact formation between a metal and an organic semiconduc- tor. The relevant variables are ΦM and ΦSC for the work functions of the metal and the semiconductor, respectively, EF,M and EF,SC for the Fermi levels of the metal and the semi- conductor, EA the electron affinity and EI the ionization potential of the semiconductor and Ev ac the vacuum level. a) The situation prior to electric contact. b) In electric contact, the Fermi levels of both materials become equal through a charge transfer between the materi- als; in this case, electrons are transferred from the metal into the semiconductor. The band bending in the organic semiconductor is induced by the formation of a space charge region. Injection barriers for electrons and holes are denoted with ΦB,e and ΦB,h. For an optimal injection, the injection barriers for both charge carrier types (denoted with ΦB,e and ΦB,h for electrons and holes, respectively) should be kept as low as possible. This requirement influences the choice of the electrodes. As the work function of the cathode should match the LUMO of the organic material, low work function materials such as calcium or barium are often used as negative electrodes in OLEDs. Due to the high reactivity of these materials, however, more stable systems like LiF/Al [16, 17] or Ag have gained importance as cathode materials. The anode, on the other hand, is a high work function material in order to obtain a good energetical alignment with the HOMO of the semiconductor. Since one of the electrodes has to be transparent to ensure efficient light outcoupling from the OLED, transparent conductive oxides (TCO), most often ITO (indium tin oxide), are typically used as anodes. Due to the nearly endless possibilities for tailoring the energetic positions of HOMO and LUMO levels of the organic materials as well as choosing suitable contacts for both anode and cathode side of the device, injection barriers can often be optimized to a large extent. However, the actual barrier height can vary significantly from the value that would be expected from the alignment of vacuum levels at the interface [18, 19] and is strongly dependent on processing conditions like deposition sequence and morphology [19]. In ohmic metal-semiconductor contacts, the energetical alignment between the contact and the semiconductor is good and the injection barrier is so small that charge carriers are injected into the semiconductor already without an applied voltage. The injection is a diffusion process, and the charge carriers build a reservoir close to the contact. The inner electric field in the semiconductor, caused by the asymmetric contacts, prevents the carriers from drifting out of the reservoir until an external voltage high enough is applied to turn the direction of the electric field. 16 Unfortunately, it is not possible to find suitable electrode materials for all semiconductor systems. In such case, an ohmic contact cannot be established and the charge carriers must overcome an injection barrier. The effective barrier height is affected by the so-called Schottky effect [12, 20] that is illustrated in Fig. 2.8 for the case of the injection of electrons from a metal into a semiconductor. An injected charge carrier in the semiconducor at the distance x from the electrode surface induces an image charge of opposite sign in the metal at −x . There is an Coulomb interaction between the injected charge and its image charge. The field of the image charge gives rise to an attractive force, the so called image force, and its potential is called the image potential Φimage:. Φimage =− e2 16πεrε0 x (2.3) where e is the elementar charge, εr the relative permittivity, ε0 the vacuum permittivity and x the distance of the charges from the metal surface. Now, if an external bias is applied, an electric field F is induced in the semiconductor, leading to a potential Φ f ield : Φ f ield =−eF x (2.4) Figure 2.8: A schematic illustration of the Schottky effect for the injection of an electron from a metal into a semiconductor at an applied electic field F . The injection barrier ΦB is reduced to an effective barrier ΦB,e f f due to the influence of image charge. xm is the distance of the barrier maximum from the metal-semiconductor interface. The electric potential of the applied field is superpositioned with the image potential. This leads to the lowering of the injection barrier ΦB by the amount ∆Φ: 17 ∆Φ= È e3F 4πεrε0 (2.5) Additionally, the maximum of the barrier is shifted from the metal-semiconductor interface into the semiconductor to the distance xm as illustrated in Fig. 2.8. Models that describe the injection of charge carriers into a semiconductor over an existing injection barrier are illustrated schematically in Fig. 2.9. The charge carriers can overcome the injection barrier by thermionic emission, field emission and thermionic field emission, a combination of the two. Figure 2.9: A schematic illustration of different models to describe the injection of charge carriers into a semiconductor in the case of an existing injection barrier. In thermionic emission, the charge carrier is excited thermally to overcome the injection barrier. It is assumed that the injection barrier ΦB is much smaller than the thermal energy kB T . The model was originally calculated by Richardson for the emission of electrons from a metal into vacuum. Taking the Schottky effect into account, the current density is given by the Richardson-Dushman equation [12]: jRD =−A∗T 2 exp � − ΦB,e f f kB T � (2.6) A∗ = 4πem∗k2/h3 is the effective Richardson constant that is proportional to the effective mass m∗. If the applied electric field F is high enough, the triangular injection barrier ΦB,e f f is thin enough for the charge carriers to tunnel through it. This injection process is called field emission. The tunneling current was calculated by Fowler and Nordheim [21] as follows: 18 j ∝ F2 · exp � 2α(ΦB)3/2 3eF � (2.7) where α= 4π p 2m2/h, F is the electric field and ΦB the height of the injection barrier. In reality, both thermionic emission and field emission are borderline cases that do not occur sepa- rately. Injection of charge carriers is therefore a mixture of both processes, thermionic field emission. The injection barrier is partly overcome by thermal activation and partly by tunneling. The domi- nating process is determined by the injection barrier, temperature and electric field. Charge transport After the charge carriers are injected from the contact into the active material, they are driven through the active layer by the electric field that is applied to the diode. Depending on the degree of order, there are two possible mechanisms for charge transport in organic materials. In highly purified molecular crystals, band transport is observed [9]. It should be noted that the bands are narrow due to the weak electronic delocalization. In polycrystalline and amorphous organic materials, on the other hand, the charges are transferred from one site to another via a hopping mechanism. Because the focus of this thesis is on amorphous polymers, only hopping transport will be discussed. It can essentially be described as a sequence of redox reactions. In case of electron transport, the electron moves from a donating molecule, which is a radical ion, to a receiving one, where a new radical ion is created. During electron transfer, the molecule that receives the electron is reduced to a singly charged anion, whereas the donating molecule is oxidized. Similarly, a molecule that receives a hole during hole transport is oxidized to a singly charged cation, while the hole donating molecule is reduced. In a hopping process, the charges tunnel between localized states. The frequency of the jumps between states i and j νi, j can be described in a simplified form with the Miller-Abrahams model [22]: νi, j = ν0 · exp � −2αri, j � · ( exp � Ei−E j kB T � for Ei > E j 1 for Ei ≤ E j (2.8) where ν0 is the so called attempt-to-escape frequency (in other words the maximum hopping rate), α the inverse localization radius or the spatial decline of the electronic wavefunction, expressing how well the charge carriers can tunnel between states i and j, ri j the distance between the states i and j, Ei and E j their energies, kB the Boltzmann factor and T the temperature. 19 Tunneling is an isoenergetic process. Jumps from state i to state j require therefore an additional absorption or emission of a phonon. The tunneling probability is expressed with the second term in Eq. 2.8. The third term corresponds to the probability of the absorption or emission of a phonon. Thermal activation is required for upward jumps. The charge carrier must move to a higher energy by absorbing a phonon prior to tunneling. The probability for an upward jump is of Arrhenius type. Downward jumps (Ei ≤ E j) are independent of temperature and their probability is considered to be one. Hopping between localized states results in low mobilitiy of the charge carriers. Charge carrier mobility is defined as the proportionality factor between the mean drift velocity v and the electric field F ; v = µF (2.9) Charge carrier mobility is dependent on the electric field. The mobility at a certain electric field F follows the Poole-Frenkel type dependence [23] µ (E) = µ∗ exp � γ p F � (2.10) where µ∗ is the so-called zero field mobility, the mobility in the abscence of an electric field. γ, the field strength factor, is a material constant that depends on the electric disorder of the environment and increases with the broadening of the Gaussian DOS, which is manifested with an increasing width σ. At best, amorphous organic semiconductors reach mobilities on the order of 10−3 cm2/Vs [9], but in many cases the values are much lower. The mobilities are therefore orders of magnitude lower than those of inorganic semiconductors. The transport of charge carriers is affected by electronic traps. They are energetic sites that lie within the band gap of the material. Trapped charge carriers are not available for charge transport without thermal activation. An activation energy EA is needed to free a trapped charge carrier. The depth of traps and thus the activation energy is defined as the energetic difference between the trap energy and the corresponding transport level. Based on their energetic position, traps are divided into shallow and deep traps. If a trap lies energetically below the Fermi level EF , one speaks of a deep trap. Shallow traps lie above the Fermi level and have activation energies of the same magnitude than the thermal energy (EA ≤ kB T). Owing to their low energy, charge carriers in shallow traps are easily released, whereas carriers in deep traps are nearly immobilized once trapped [24]. Traps can be created by impurities or structural defects, for example [25]. Impurities are molecules with HOMO or LUMO (or both) located whithin the energy gap of the host, whereas structural defects in the material lead to fluctuation in the surroundings or the molecule and the conjugation length. These fluctuations create tail states below the transport level that act as traps [26]. 20 Exciton formation and diffusion When the holes and electrons approach each other, they form an exciton, an excited state on a single molecule. The area in which the charge carriers meet to form excitons is called the recom- bination zone. Its position depends on the injection, transport and recombination properties of the material. As an example, the PPVs are hole conductors. Hole mobilities are thus some orders of magnitude higher than electron mobilities and provided that the injection properties are similar on both contacts, the recombination zone lies close to the cathode rather than in the middle of the active layer. Once formed, the excitons can migrate in the active layer by diffusion. The migration of an exciton can take place by three transfer mechanisms [27]: radiative energy transfer, Förster transfer and Dexter transfer. The radiative energy transfer, also called re-absorption, involves an emission and a subsequent absorption of a photon. It is a relevant process only if the absorption and emission spectra of the material overlap. The non-radiative transfer mechanisms, Förster and Dexter energy transfer, are presented in Fig. 2.10. Förster energy transfer [28] is a non-radiative long-range (up to 10 nm) transfer process that takes place via resonant energy transfer of neighboring molecules. A spectral overlap between the emission of the donor and absorption of the acceptor molecule is required. Here, donor and acceptor do not necessarily have to refer to different materials. In the case of exciton migration in a single material, donor and acceptor are adjacent molecules of the same type. For Förster-type transfer, the transition between ground and excited states have to be allowed for both donor and acceptor molecule. The Förster mechanism is therefore limited to singlet excitons. Figure 2.10: Exciton energy transfer mechanisms. a) Förster energy transfer. The transitions on both donor and acceptor molecules have to be spin-allowed, which limits the Förster transfer to singlet excitons only. b) Dexter energy transfer. The mechanism allows the transfer of both singlet and triplet excitons. 21 The Dexter energy transfer [29] involves a direct exchange of electrons and requires an overlap of the electronic wavefunctions of the involved molecules. The Dexter transfer is therefore a short- distance process with a typical range of 0.1 - 1 nm [27]. The mechanism allows the transfer of both singlet and triplet excitons and is considered to be the main transfer mechanism for triplets [30]. Recombination The radiative decay of an exciton leads to an emission of a photon. The energy of the photon and therefore the wavelength of the emitted light depends on the energetic difference between the LUMO and the HOMO of the organic semiconductor reduced by the exciton binding energy. The emission color can therefore be manipulated by tuning the energetic positions of the HOMO and LUMO of the emitter material. However, not all excitons decay radiatively, leading to electrolu- minescence. Statistically, singlet and triplet excitons are formed with a ratio of 1:3 [2, 31]. As discussed in Section 2.1.2, the radiative transition from the excited triplet to the ground singlet state is forbidden by the spin selection rules. This leads to low efficiencies in fluorescent OLEDs, because the majority of the excitons does not participate in electroluminescence. Additionally, there are also non-radiative recombination pathways for singlet excitons. For example, in cases where charge transport is strongly unbalanced and the recombination zone lies near one of the contacts, the excitons can be quenched at the contact [32]. Excitons can also be quenched through exciton- exciton annihilation at high exciton densities [33–35] or be dissociated into a geminate pair of charge carriers by deep traps or other non-radiative recombination centers [9, 36]. The quenching of excitons can be reduced by the use of multilayer OLED stack structures: with the use of electron and hole blocking layers, the recombination zone can be confined to the desired position away from possible quenchers like the electrodes. Device Operation Fig. 2.11 shows the energetic schemes of a prototypical, single-layer OLED during different stages of operation. For clarity, the localized energetic states of the organic semiconductor are presented as straight lines and band bending is omitted. The situation prior to electric contact is depicted on top left (1). The rectifying behavior of an OLED is provided through the choice of different contact materials. They are characterized by the work function φ, the difference between the Fermi level of the electrode material and the vacuum level. As the electrodes are contacted (2), their Fermi levels equalize due to diffusion of charge carriers. The potential difference between the anode and the cathode, called built-in potential eUBI , causes an internal field in the device. This field causes a drift current in the opposite direction to the diffusion. The drift and diffusion currents compensate each other and there is no net flow of charge carriers. If a negative voltage U < 0 is applied (3), meaning the anode is negative with respect to the cathode, the inner field caused by the difference in the Fermi levels of the contact materials is increased by the 22 Figure 2.11: Operation schemes for an OLED. (1) Situation prior to electric contact of the electrodes. (2) As the electrodes are contacted, the Fermi levels equalize and an inner field develops within the organic semiconductor. The potential difference between the electrodes is called built-in voltage UBI . (3) Reverse bias leads to large injection barriers for both charge carrier types. (4) If the applied voltage equals the built-in voltage, the inner field in the diode is compensated by the external bias and a flat-band condition is reached. (5) Increasing the voltage further in forward direction leads to the injection of charge carriers and a net drift current. applied bias. The voltage U∗ across the organic layer is now expressed as U∗ = U + UBI . Electrons are now injected from the anode and holes from cathode; due to the high injection barriers the resulting current is very small. This situation is called reverse bias and the contacts are reffered to as blocking contacts. In order to facilitate net drift current in the device, a voltage is applied in forward direction so that the anode is positive with respect to the cathode, U > 0. With voltages smaller than the built- in voltage (U < UBI), there is no current flow. If the applied voltage U equals UBI , the so-called flat band condition (4) is reached and the inner field in the diode is compensated by the external 23 bias. Applied voltages higher than UBI (5) result in charge injection from the contacts into the semiconductor layer and a rapid increase in current density. 2.2.2 Electrical degradation of organic light-emitting diodes This section offers a short overwiev on the relevant processes that are related to the electrical fatigue of polymer-based OLEDs. During continuous electrical operation, the performance of OLEDs deteriorates. The devices are often operated at a constant current; during driving, a decrease in luminance intensity is observed. It is mostly accompanied by a simultaneous increase in impedance that is manifested by an increasing operating voltage. An example of such a temporal evolution is presented in Fig. 2.12. The diode with the structure ITO / PEDOT:PSS / OC3C8-PPV/ Ca / Al was operated at a constant current density of 50 mA/cm2. At the beginning of the operation, the luminance decreases rather steeply before leveling off to a more moderate decay. The increasing driving voltage results from the increasing impedance of the device. As for this exemplary device, in many cases the luminance decay is more or less a mirror reflection of the voltage rise in polymer- based OLEDs [37]. The lifetime of a diode is generally defined as the half-life t50, which means the time it takes for the luminance to decrease to half of its initial intensity. The t50 of the exemplary diode in Fig. 2.12 is approximately 150 h. 4 , 0 4 , 5 5 , 0 5 , 5 6 , 0 6 , 5 7 , 0 7 , 5 0 2 0 0 4 0 0 6 0 0 8 0 0 1 0 0 0 1 2 0 0 1 4 0 0 0 , 0 0 , 2 0 , 4 0 , 6 0 , 8 1 , 0 Vo lta ge [V ] Lu mi ne sce nc e [ a.u .] T i m e [ h ] t 5 0 Figure 2.12: Exemplary lifetime measurement of an OLED with a structure ITO / PEDOT:PSS / OC3C8- PPV/ Ca /Al and a PPV layer thickess of 120 nm. The device was driven at a constant current density of 50 mA/cm2. The t50 lifetime of the device is approximately 150 h. The degradation of OLEDs is a complex topic and it is by no means possible to assign the deteriation to a single process. Today, some topics that were seriously problematic for the first OLEDs have lost 24 their importance owing to the technological progress in the field. Dark spots are a typical example. The first OLEDs were not encapsulated, which lead to problems regarding the stability of the low work function cathodes. Exposing the devices to ambient atmosphere leads to degradation of the cathode that manifests itself as non-emissive areas that were investigated in numerous publications in 1990s and 2000s [38–42]. Dark spots have been generally attributed to localized delamination of the cathode on areas with pre-existing particle defects on the substrate [37]. Another degradation process that is largely irrelevant in the state-of-the-art OLEDs is the catas- trophic failure of the devices, also called the sudden death. It is related to the development of electrical shorts in the device and leads to a sudden, often complete loss of luminance. Catastrophic failure has been connected to pre-existing morphological defects in the organic semiconductor or in the electrodes that lead to the formation of microscopic conduction paths in the organic material and to short cuts between the electrodes [43, 44]. Like dark spots, also the catastrophic failure can be prohibited to a large extent with the present know-how. It is essential to control the film formation and morphology of the materials as well as the preparation processes. In contrast to the two degradation mechanisms introduced above that are related to fabrication conditions and exposure to ambient atmosphere and can be avoided by controlling of these factors, intrinsic degradation is a far more complex issue. It is defined as the progressive and uniform decrease of luminance without any obvious change in device appearance that takes place over time during operation [37, 38, 45–47]. The processes leading to intrinsic degradation are manifold. One way of classifying them is to divide them into the morphological stability of the organic layer, the formation of traps and luminescence quenchers, interface degradation and electrode instability [37]. The thermal and morphological stability of the organic layer has been subject to discussion espe- cially with low Tg materials; hole transport materials, particularly TPD, have mostly been in focus of attention [45, 48–50]. Crystallization of the active material has been linked to shorter device lifetimes and defect formation [51–53]. The formation of traps affects both the luminance and the operating voltage. Firstly, traps can act as non-radiative recombination centers [25, 54, 55] that decrease luminance intensity [36, 56, 57]. Secondly, the formation of traps increases the operating voltage by decreasing the effective mobility of the charge carriers [37]. Furthermore, the change in the effective mobility of one charge carrier type influences the balance of charge carriers. Polymer- based light emitting diodes are simple, single- or double-layer devices in which the confinement of excitons within the desired recombination zone by charge blocking layers is not possible. A strongly unbalanced transport of electrons and holes moves the recombination zone close to the electrodes and increases the quenching of excitons at the contacts, limiting the device efficiency. Degradation products of the organic materials as well as defects that originate from synthesis act as traps and non-radiative recombination centers [25, 46, 58]. The chemical degradation in the bulk of efficient state-of-the art OLEDs is often charge carrier induced [46]. In this type of degradation, the radical cations or anions of the organic materials are not stable and the degradation is caused by 25 the transport of charges through the organic material. An example is the instability of the cationic species of the small molecule Alq3 [45, 58–60]. For PPV, the two most relevant defects that originate from synthesis are the tolane-bisbenzyl defect [61, 62] and the halogene vinyl defect [63, 64]. Owing to the simple structure of polymer-based OLEDs, degradation of organic-organic interfaces is not as critical as in small molecule OLEDs that consist of multiple layers. Here, especially the interface between hole transport layer and emission layer is considered to be of importance [37, 45]. In contrast to this, the instability of the electrode-semiconductor interfaces is an important issue for both small molecule and polymer-based devices. At the cathode-organic interface, diffusion of Ca and Ba into a PPV layer has been observed [65]. The commonly used ITO has been reported to act as the source of both indium and oxygen that diffuse into the organic material [37]. Indium has been detected in the organic layers of stressed devices [66] but it is not considered to be of great importance for the stability of OLEDs [37, 47]. Oxygen seems to be more problematic. Diffusion of oxygen out of the ITO anode has been reported to affect the lifetime of polymer-based OLEDs [67]. The diffusion stems from the surface treatment of ITO with oxygen or UV-ozone plasma that forces oxygen atoms to the ITO surface. The oxygen is not stable on the surface and diffuses readily into the organic layer [37]. Inserting a conducting polymer (mostly PEDOT:PSS) as hole injection layer between ITO and the organic semiconductor has been reported to greatly increase the stability of the diodes [67, 68]. The injection of holes directly from PEDOT:PSS makes the stability of the ITO no longer critical for the lifetime of the device. However, the PEDOT:PSS layer as such has been related to the degradation of polymer-based OLEDs as well. PEDOT:PSS is sensitive to moisture and has been attributed to the formation of dark spots [69]. 26 3 Experimental This chapter begins with the introduction of the materials used throughout this thesis. After that, the procedures for sample fabrication are described, including the cleaning of the substrates, deposition of the active layers and the electrodes. The chapter then continues with the standard characteri- zation of the samples by spectroscopy, recording of current-voltage-luminance characteristics and electrical stressing of the devices. To finish, the experimental techniques applied for the investiga- tion of charge transport and the influence of the triplet excitons on the devices are described. These methods include charge extraction by linearly increasing voltage (CELIV), photoinduced absorption, secondary ion mass spectrometry and infrared thermography. 3.1 Materials OC3C8-PPV The active polymer used in this work is poly(2-propoxy-5-(2’-ethylhexyloxy)-phenylene vinylene) (OC3C8-PPV). It belongs to the group of poly(phenylene vinylenes), or in short PPVs. PPVs are an important group of organic semiconductors and have been utilized in the field of polymer-based OLEDs from the very beginning: the first polymer-based OLED was prepared with unsubstituted PPV as the active material [3]. The unsubstituted PPV is insoluble and difficult to process due to its strong interchain π - π stacking interactions [70]. Adding side chains to the polymer backbone increases the conformal mobility of the polymer chains and make the material soluble. The side chains are mostly long alkyl chains that are attached to the benzene ring of the polymer. The molecular structures of the unsubsituted PPV as well as OC3C8-PPV are shown in Fig. 3.1. OC3C8-PPV has unsymmetric alkoxy side chains and dissolves in common organic solvents such as toluene. The OC3C8-PPV used in this work was synthetized by Dr. Nicole Vilbrandt from the group of Prof. Matthias Rehahn in the Ernst-Berl-Institut für Technische und Makromolekulare Chemie at the Tech- nische Universität Darmstadt. The synthesis was performed using the Gilch route [71] that is out of scope of this work; for details of the synthesis, the interested reader is referred to the thesis of Dr. Nicole Vilbrandt [72]. The absorption and electroluminescence spectra of OC3C8-PPV are plotted in Fig. 3.2. OC3C8-PPV has an absorbance maximum at 506 nm within the broad fundamental absorption. The electro- luminescence spectrum has three peaks that correspond to the 0-0, 0-1 and 0-2 transitions. The emission peaks are connected to the formation of aggregates during the annealing of the polymer 27 Figure 3.1: Molecular structures of unsubstituted PPV (left) and OC3C8-PPV (right). films. The highest-energy 0-0 transition is thought to originate from the single-chain chromophore, whereas the other peaks are caused by aggregated parts [72]. A high degree of chain aggregates enhances the intensity of the 0-1 peak with respect to the 0-0 electronic transition [73, 74]. The orange-colored emission of OC3C8-PPV is red-shifted with respect to the absorption; this is due to the Stokes shift. The positions of the HOMO and LUMO levels of OC3C8-PPV were obtained by cyclovoltammetry measurements performed by Dr. Nicole Vilbrandt. The HOMO and LUMO lie at 5.2 and 2.6 eV, respectively. 4 0 0 5 0 0 6 0 0 7 0 0 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 1 . 2 1 . 4 Ab so rba nc e W a v e l e n g t h ( n m ) 5 0 0 6 0 0 7 0 0 8 0 0 9 0 0 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 No rm ali ze d ele ctr olu mi ne sce nc e W a v e l e n g t h ( n m ) Figure 3.2: Left: Absorption spectrum of a 140 nm thick OC3C8-PPV film. Right: Electroluminescence spectrum of a device with the structure ITO / PEDOT:PSS / OC3C8-PPV/ Ca / Al (right). PtOEPK Platinum (II) octaethylporphyrine ketone (PtOEPK) belongs to the group of metalloporphyrines and is often used as an oxygen sensor [75]. In this work, it was applied as the triplet sensitizer for OC3C8-PPV. The molecular structure of PtOEPK is shown in Fig. 3.3. Due to the strong spin- orbit coupling caused by the presence of the central platinum atom, PtOEPK has a high intersystem crossing efficiency. It is a phosphorescent material; no fluorescence emission from the material has been detected [75]. 28 Figure 3.3: Molecular structure of PtOEPK. The absorbance and emission spectra of PtOEPK are plotted in Fig. 3.4. The absorption spectrum is typical for porphyrines [76] and has two distinct peaks. The transition to the second excited state at 403 nm is called the Soret band; the Q band at 599 nm originates from the transition to the first excited state. The phosphorescence peaks at 758 nm. The HOMO and LUMO values of PtOEPK are not known in literature. As orientation, the HOMO and LUMO of PtOEP, a very similar molecule, lie at 2.8 - 3.2 eV and 5.3 - 5.6 eV [77–79]. Based on the optical spectra, the energy gap of PtOEP is approximately 0.2 - 0.3 eV larger than that of PtOEPK. 4 0 0 5 0 0 6 0 0 7 0 0 8 0 0 9 0 0 0 , 0 0 , 2 0 , 4 0 , 6 0 , 8 1 , 0 1 , 2 a b s o r p t i o n p h o t o l u m i n e s c e n c e No rm ali ze d a bs orp tio n, em iss ion W a v e l e n g t h ( n m ) Figure 3.4: Absorption and photoluminescence spectra of PtOEPK. Contact materials Indium doped tin oxide (ITO) is a standard choice for the anode in OLEDs. ITO-coated glass sub- strates were obtained from VisionTek Systems. In order to prevent the diffusion of alkali metal ions from the glass into the ITO layer, an SiO2 barrier layer had been deposited between the substrate and ITO. After purification, the ITO substrates were treated with ozone in order to remove carbon contaminations from the surface and to increase the work function of ITO. After the treatment, the work function has been reported to be 4.75 - 4.8 eV [80, 81]. A layer of poly(3,4-ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS) between ITO and the active polymer has been observed to improve the diode performance. PEDOT:PSS smooths the 29 ITO surface and improves the hole injection. The work function of PEDOT:PSS is 5.2 eV [82] and is therefore well matched with the HOMO of OC3C8-PPV. The PEDOT:PSS was obtained from Heraeus as an aqueous dispersion with the product name CleviosTM P VP Al 4083. The PEDOT:PSS ratio was 1:6 by weight. Alternatively to ITO / PEDOT:PSS, holes can also be injected from a gold electrode. In this thesis, Au was used as the top electrode in hole-only devices. The work function of Au has been reported to vary depending on the deposition method and the degree of contamination on the surface [83, 84]; values between 4.8 and 5.4 eV have often been reported [19, 84–86]. For an efficient injection of electrons, low work function electrodes are required. Calcium has a work function of 2.9 eV and matches the OC3C8-PPV LUMO well. Calcium is highly reactive and oxidizes fast in contact with oxygen. It was therefore routinely protected with a capping aluminum layer in the devices used in this work. In cases like electron-only devices where a low work function bottom electrode is needed, the significantly more inert LiF / Al is a good alternative to Ca. This electrode consists of a thick aluminum layer and a thin lithium fluoride interlayer with a thickness of 0.7 nm. The work function of LiF / Al is 2.9 eV [87]. 3.2 Sample preparation The sample layout is illustrated in Fig. 3.5. The ITO anode was structured by photolithography and it consisted of four L-shaped areas. The PEDOT:PSS as well as the PPV layers were spin-coated on top of the ITO. The stripe-formed cathode was deposited via physical vapor deposition. The diodes were the areas on which anode and cathode overlap. There were four diodes on each substrate, each with an area of 10 mm2. The use of multiple small diodes instead of one large one had two benefits. Firstly, it increased the yield through the smaller probability of both short cuts caused by dust particles and uneven emission due to variations in the quality of the spin-coated polymer layer. Secondly, the simultaneous preparation of multiple diodes enabled the direct comparison of the devices without unwanted changes in preparation parameters and conditions. For the diodes, it is crucial to have clean and dustfree substrates. Due to the thin films with a thickness on the order of some hundred nm, even smallest dust particles can cause short cuts. In order to compensate the lack of clean room environment, the sample preparation that took place outside the glovebox was performed in a flow box, in which a continuous flow of filtered air guaranteed dust-free conditions. The standard cleaning procedure of the ITO substrates was as follows: 1. Rinsing with deionized water, drying with nitrogen flow 2. 15 min ultrasonic bath at 60 °C in 5 % deconex (an alkaline cleaning concentrate) 3. Rinsing with deionized water, drying with nitrogen flow 4. 15 min ultrasonic bath at room temperature in acetone 30 Figure 3.5: The sample layout consists of an L-formed anode structure (left), the active polymer layer covering the substrate homogeneously and a stripe-formed cathode (right). Due to the L- formed anode structure, there are four diodes on one substrate. The anodes are contacted separately with pins (denoted with dots) and the cathode contact is shared between all diodes (the dark square in the middle at the bottom). 5. Rinsing with deionized water, drying with nitrogen flow 6. 15 min ultrasonic bath at room temperature in isopropanol 7. Rinsing with deionized water, drying with nitrogen flow 8. 15 min UV-ozone treatment in a UV-ozone photoreactor UVP100 from Ultra-violet Products Ltd. Both PEDOT:PSS and PPV layers were deposited on the substrate via spin coating. PEDOT:PSS is delivered as an aqueous solution and must therefore be spin coated outside the glovebox. Like substrate cleaning, also the spin coating took place in a flow box in order to avoid the contamination of substrate surface by dust particles. In order to obtain as homogeneous films as possible, the PEDOT:PSS solution was mixed with isopropanol with a ratio of 2:1. The mixture was filtered with a 0.45µm polyvinylidene fluoride (PVDF) filter prior to spin coating to avoid having dry or agglomerated particles on the substrate. The PEDOT:PSS films were spin coated at 3000 RPM for 30 s, resulting in a layer thickness of approximately 30 nm. The substrates were subsequently annealed on a hot plate at 110 °C for 5 min in order to remove residual water from the layer. The spin coating of the active PPV layer as well as all following preparation steps took place in a glovebox in nitrogen atmosphere with oxygen and air concentrations on the order of 1-3 ppm. OC3C8-PPV was dissolved in toluene at a concentration of 7.5 mg/ml for solutions with both pristine OC3C8-PPV and OC3C8-PPV blended with PtOEPK. After preparing the solutions, they were left stirring overnight and filtered with a 5µm polytetrafluorethylen (PTFE) filter. The solutions were spin coated at approximately 3000 RPM and annealed on a hot plate for 5 min at 130 ° C. The resulting layer thickness of 130 nm was controlled with a Dektak XT profilometer. The metal electrodes were deposited by physical vapor deposition in the vacuum deposition system UNIVEX350G from Oerlikon Leybold Vacuum. The evaporation system is integrated in the glovebox. The samples do therefore not need to leave the inert atmosphere during the preparation process. 31 The deposition setup allows the use of four substrates at once. With four diodes per substrate, alltogether 16 diodes can be prepared simultaneously. In this way unwanted differences between substrates caused by e.g. changes in evaporation parameters can be avoided and a direct comparison of the diodes is possible. The materials were deposited in vacuum at a pressure on the order of 10−6 mbar at typical rates of 2-5 Å/s. As sources, resistively heated tungsten boats (for Au), tungsten spirals (for Al) and molybdenum boats (for Ca and LiF) were used. The form of the evaporated areas was defined by shadow masks. For an even thickness of the deposited material on all substrates, they were rotated at 5 RPM during the evaporation process. The film thickness was controlled by an oscillating crystal. 3.3 Methods This section introduces all methods used in this thesis. First, the focus is set on the standard material and device characterization and electrical stressing of the OLEDs, one of the fundamental methods that were applied in this work. After this, the section continues with the other applied techniques: carrier extracion by linearly increasing voltage (CELIV), photoinduced absorption, secondary ion mass spectrometry and infrared thermography. 3.3.1 Standard methods for material and device characterization Optical spectroscopy Transmission measurements were performed with a UV-VIS spectrometer Lambda 900 from Perkin Elmer. The sample layers were deposited on glass substrates. For each sample, an empty glass substrate was first measured as a reference. Photoluminescence spectra were measured with a Cary Eclipse Fluorescence Spectrophotometer from Varian. For electroluminescence measurements, a Maya 2000 Pro fiber spectrometer from Ocean Optics was used. It enabled the direct measurement of samples inside the glovebox without the need of encapsulation or sealed sample holders with an inert atmosphere. Current-voltage-luminance characteristics Each set of prepared diodes was first characterized by recording the current-voltage-luminance characteristics with a HP4515A from Hewlett Packard. The applied voltage was varied and the current density flowing through the device as well as the luminance were measured. The luminance was recorded as the current of a photodiode that was positioned directly over the measured diode. For converting this current to luminance, a photometric measure of the luminous intensity with the unit cd/m2, each material was calibrated with a spot photometer CS 100 from Minolta. 32 3.3.2 Electrical stressing The electrical stressing of the devices was performed with a constant current density as is standard practice in the field. The self-built lifetime measurement setup enabled the simultaneous measure- ment of six substrates. Because each substrate has four diodes, altogether 24 diodes could thus be measured simultaneously. Each diode was connected to a separately controlled current source and the desired current density was applied. The voltage that was required for the chosen current density as well as the luminance were recorded. The luminance was measured by a photodiode positioned directly over the diode. The diodes were measured sequentially one after each other. In order to be able to record the fast changes in the early stages of device operation, each diode was measured every 30 seconds for the first 60 minutes, every 60 seconds for the next 60 minutes and after that every 120 seconds. 3.3.3 Carrier extraction by linearly increasing voltage Carrier extraction by linearly increasing voltage was first introduced by Jus̆ka et al. in 2000 [88]. The technique was originally developed for the determination of charge carrier mobility and con- centration in microcrystalline silicon and doped conjugated polymers, in which thermally generated equilibrium charge carriers can be investigated [88, 89]. For undoped organic semiconductors, which have a very low intrinsic charge carrier density, nonequilibrium charge carriers are created by the absorption of a laser pulse; in this case, the method is referred to as photo-CELIV [90–92]. For the sake of clarity, in this work the investigations with and without optical excitation are referred to as photo-CELIV and dark-CELIV, respectively. Since its introduction, CELIV has attracted considerable interest. The technique is simple, the data analysis straightforward and, most notably, it can be applied to thin films with a thickness on the order of or below 100 nm. It is therefore possible to investigate charge transport phenomena di- rectly in the actual device geometries. This is of importance especially for solar cells with mixed donor/acceptor blends, in which the morphology of the layers plays a crucial role and cannot be reproduced reliably for large film thicknesses as required for time-of-flight measurements. A photo-CELIV experiment is illustrated schematically in Fig. 3.6. The basic principle is simple: first, the sample is illuminated by a short laser pulse through a transparent contact. The pulse penetrates the sample and creates (in an ideal case) a homogeneous charge carrier density in the active layer. During the complete measurement, an offset bias corresponding the built-in voltage of the diode is applied to the sample in order to minimize the inner electric field in active layer. The photogenerated charge carriers experience flat band conditions and remain in the semiconductor. After an adjustable delay time tdel after the laser pulse, the external bias is increased linearly in reverse direction with the voltage rate U ′ = dU/d t. This voltage ramp leads to the extraction of the 33 charge carriers. Due to the reverse bias, both contacts are blocking and the injection of new carriers is prohibited. Figure 3.6: a) Schematic illustration of a photo-CELIV measurement. Top: the applied voltage ramp in reverse direction. The ramp begins after a delay time tdel after the laser pulse. During the measurement, an offset voltage is applied to compensate the built-in voltage of the device. Bottom: the current response. j0 denotes the displacement current that is induced by the linear voltage ramp. If the sample is excited with the laser, an additional photocurrent with the maximum ∆ j appears. The maximum is reached at a time tmax . b) The charge carrier density distribution and electric field in the sample for a hole-dominated device, in which µh >> µe, adapted from [93]. The obtained current transient can be divided into two parts. The capacitive part j0 is the current induced by the linear voltage ramp alone, without the laser excitation. The linear voltage rise results in a rectangular current response with a plateau value corresponding to the capacitive displacement current j0 = 1 A dQ dt = C A dU dt = εε0U ′ d (3.1) where A is the area of the electrodes, Q the electric charge at the electrodes, C the capacitance of the OLED, U the applied voltage, ε the dielectric constant of the organic semiconductor, ε0 the vacuum permittivity and d the thickness of the organic layer. In addition to the displacement current induced by the linear voltage ramp, the laser pre- illumination creates another contribution to the current response. The photogenerated charge carriers that have not recombined during the delay time between the laser pulse and the begin- ning of the voltage ramp are extracted from the device during the triangular voltage pulse. This extraction creates a photocurrent with a maximum value ∆ j, which is reached after a time tmax 34 after the beginning of the voltage ramp. The extracted charge Q can be obtained by integrating the extraction current ∆ j over time[94]: Q = A ∫ ∆ jdt (3.2) The mathematical expression for both current components was derived by Jus̆ka et al. [88]. It is assumed that the densities of the photogenerated electrons and holes are equal, and that the charge carrier density ne = nh = n is homogeneously distributed in the whole bulk. Furthermore, it can be assumed that the charge carrier transport in the organic layer is dominated by one charge carrier type. Because in PPVs µh >> µe holds true, the holes can safely be treated as the mobile species. In reverse bias conditions, the positively charged cathode lies at z = 0 and the negatively charged anode at z = d; d is the thickness of the organic layer. As a result of the voltage ramp, at a time t the holes are extracted from the bulk up to an extraction depth l(t). The situation is illustrated in Fig. 3.6(b). The front of the extracted holes moves through the semiconductor layer with a velocity dl d t = µE(l(t), t). Because of the low mobility of the electrons, they are considered immobile. The total charge density ρ(z, t) is ρ(z, t) = ( −en for 0≤ z ≤ l(t) 0 otherwise (3.3) The electric field is coordinate dependent in the depletion (i.e. hole extracted) region 0 < z < l(t) and constant at l(t) < z < d. Now, combining continuity, current and Poisson equations [88] gives a mathematical expression for the current response j(t) = εε0U ′ d + en � 1− l(t) d � � µU ′ t d − enµ 2εε0d l2(t) � (3.4) Numerical calculations [89] lead to an expression for mobility µ= 2d 3U ′ t2 max � 1+ 0.36 ∆ j j0 � (3.5) The possibilities offered by dark-CELIV and photo-CELIV are not limited to mobility investigations. For example, it is possible to study charge relaxation phenomena by varying the delay time tdel between the laser pulse and the beginning of the voltage ramp [92, 95–97]. The concentration and field dependence of the mobility can be determined by changing the incoming light intensity [91, 92, 98] and the maximum of the applied voltage pulse [91, 99], respectively. In case of a system 35 with equilibrium charge carriers (CELIV without laser illumination), the recovery of equilibrium can be investigated by varying the time between two consecutive voltage ramps [88]. In a recent publication, the trap density as a function of escape energy has been evaluated using an application of the photo-CELIV technique [100]. The dispersion of charge transport (manifested in how fast the transient rises before and decays after reaching the maximum value) can be reflected with the ratio of extraction current at full width half maximum t1/2 and tmax [90]. The photo-CELIV setup is depicted in Fig. 3.7. The charge carriers were photoexcitated with laser pulses from an optical parameter oscillator (OPO) pumped by a Nd:YAG laser with a repetition frequency of 10 kHz. The wavelength of the excitation pulses was 532 nm, which lies within the fundamental absorption of OC3C8-PPV. The samples were excited through the transparent anode. The delay time between the laser excitation and the voltage ramp was 5µs. The triangular voltage ramps were created with an Agilent 3320A pulse generator and the extraction current transients were recorded with a Tektronix TDS 5052 digital oscilloscope using a Femto DHPCA-100 current- voltage amplifier. The current transients were averaged over 300 measurements in order to increase the signal-to-noise ratio. Figure 3.7: Photo-CELIV setup. 3.3.4 Photoinduced absorption Photoinduced absorption (PIA) is a pump-probe method that is often used to investigate the dy- namics of excited states in organic semiconductors. The triplet absorption measurements presented in this thesis were performed with a quasi-steady-state setup (see Fig. 3.8). The sample is excited by a laser or an LED, which is modulated by an optical chopper. This excitation is called the pump light. The transmission of the sample in ground and excited state is measured by the probe light, a monochromated light beam of a white light source. In other words, the probe light measures the changes in sample transmission induced by the pump excitation (which explains why the method is named photoinduced absorption). The measured signal is analyzed with a lock-in amplifier which 36 uses the modulation frequency of the chopper as a reference. The change in sample transmission is monitored as a function of wavelength. The ground state transmission is recorded prior to the PIA measurement; in addition to this, the measurements are corrected for photoluminescence. The measurements are performed at 80 K; at higher temperatures the lifetimes of the triplet states are too short for the method. Figure 3.8: Photoinduced absorption measurement setup. The measured signal is expressed as ∆T/T; that is, the change in transmission normalized to the ground state transmission. Additionally, the results are normalized to the optical density of the samples to ensure comparability between different materials. The ∆T/T signals can be either pos- itive or negative; the T1 → Tn absorption is observed as a negative signal. The pump light excites the first triplet state which then absorbs the probe light. This is observed as a negative change in transmission. The PIA measurements were performed with the help of Fabian Etzold at the Max Planck Insti- tute of Polymer Research in Mainz in the Research Group of Organic Optoelectronics supervised by Dr. Frédéric Laquai. As the pump light, a Newport LED (LED-527-HP) with an excitation wave- length of 527 nm and 100 mW/cm2 was used. The pump frequency was 317 Hz. The probe light consisted of a 100 W tungsten-halogen lamp and a LOT-Oriel Omni-λ 300 monochromator. An additional monochromator was placed behind the sample to suppress false signals due to sample photoluminescence. The transmitted probe light was detected with three detectors depending on the wavelength: for 450-1100 nm, an amplified Si photodetector (Thorlabs PDA 100A) and for 800-1800 nm an amplified Ge photodetector (Thorlabs PDA 50B) were used, whereas for the range 1700-4500 nm a liquid nitrogen cooled InSb photodetector was chosen. The changes in transmis- sion were recorded by a lock-in amplifier from EG&G Princeton Applied Research, model 5210. The samples were mounted in a nitrogen-cooled optical cryostat (Oxford Instruments Optistat CF) at 80 K in a helium atmosphere. 37 3.3.5 Secondary ion mass spectrometry Seoondary ion mass spectrometry is applied for the investigation of the composition of surfaces and thin films. In SIMS, the sample is bombarded with primary ions, leading to the ejection of secondary ions from the material. These are sent to a mass spectrometer, where they are analyzed qualitatively based on their masses and detected quantitatively. SIMS offers a depth dependent analysis of a sample when the rate of arrival of given secondary ions (measured as intensity in counts per second) is plotted against sputtering time. The SIMS measurements were performed in collaboration with Dr. Stefan Flege from the Material Analytics group lead by Prof. Ensinger at TU Darmstadt. 3.3.6 Infrared thermography The lifetime of an OLED is influenced by its operation temperature [101, 102]. The temperature distribution on the OLED surface can be determined with infrared thermography. Infrared thermo- grams were recorded with the ImageIR 8300 infrared camera from InfraTek GmbH and analyzed with the IRBIS 3 software. The available objective was able to focus on an area that was ap- proximately 1/4 of the diode surface. For long-term measurements during electrical stressing, a thermogram was captured every 10 minutes. 38 4 Experimental techniques for the investigation of PPV-based devices This chapter presents two techniques that were developed for the investigation of PPV-based diodes. First, the removal of the cathode with acetic acid is introduced, enabling the analysis of the polymer layer after the operation of the device. This is necessary for example in situations in which the morphology of the active layer is of interest. The second topic of this chapter is the accelerated aging of the diodes. In many cases, it is necessary to accelerate the fatigue of the devices by operating them at higher current densities. A scaling law for OC3C8-PPV-based diodes was determined, enabling the comparison of lifetime measurements performed at different current densities. 4.1 Removing the cathode In some situations the direct investigation of the polymer layer in the diode structure is of interest. Investigations of the structure or morphology of the polymer after degradation are examples of such cases. In order to examine the polymer layer in the diode structure, it is necessary to remove the cathode after the device has been operated. With small molecules, this can conveniently be done with tape [103]. The adhesion between the organic material and the cathode is low enough so that the cathode can be stripped off. However, this approach proved not to be applicable for the PPVs. Due to the strong adhesion between the polymer and the cathode, the polymer film was always stripped away with the cathode. It was therefore necessary to develop another approach for the removal of the cathode. Etching the metal stripe away instead of stripping it off proved to be a reasonable alternative. First experiments were done with hydrochloric acid (HCl) diluted with water with the ratio of 1:4. The diluted acid was dropped directly on the cathode stripe. The etching of the Ca layer could be easily observed by eye, and after the original layer was not visible, the acid was immediately washed away with deionized water and dried in nitrogen. After this, the sample was dried for 2 h in a vacuum oven at 100 ° C in order to remove residual water from the layer. The etching of the cathode with HCl lead to faster degradation, however, as seen in Fig. 4.1. The temporal evolution of luminance is plotted for a continuously operated reference sample with an original cathode and for a sample, the cathode of which was removed and replaced by a new one after ca. 70 h of operation. It is clear that the degradation is accelerated after the removal of the original cathode. 39 0 5 0 1 0 0 1 5 0 2 0 0 2 5 0 0 , 0 0 , 2 0 , 4 0 , 6 0 , 8 1 , 0 No rm aliz ed lu mi na nc e ( a.u .) T i m e ( h ) r e f e r e n c e 1 . c a t h o d e 2 . c a t h o d e Figure 4.1: Temporal evolution of luminance of devices with the structure ITO / PEDOT:PSS / OC3C8- PPV/ Ca. Black: the reference diode with an original cathode operated without interrup- tions. Red/blue: A diode that was operated for ca. 70 h with the first cathode (red), after which the cathode was removed with diluted HCl and replaced with a freshly deposited one (blue). One possible reason for the shortened lifetime is the presence of residual chlorine in the polymer layer, originating from the treatment with HCl. Secondary Ion Mass Spectrometry (SIMS) mea- surements were performed in order to clarify whether residual chlorine was present in the samples after the etching procedure. Three samples were measured: one neat OC3C8-PPV film on an ITO substrate without a top electrode, one OC3C8-PPV film that was treated with the diluted HCl, also without a top electrode, and one complete device with the structure ITO / OC3C8-PPV/ Ca, etched with diluted HCl. The results are shown in Fig. 4.2, where the intensities of relevant elements are plotted agains the sputtering time. It is possible to detect the different layers of the sample: through- out the polymer layer, the intensity of carbon stays constant. At the interface between polymer and the ITO substrate, the C intensity drops and the intensity of indium reaches its maximum plateau. The intensities of the other elements, Ca and Cl, are low throughout the scans in all samples. The reference sample with the untreated OC3C8-PPV layer does not contain these elements, nor does the etched reference OC3C8-PPV layer contain Ca. As the Ca intensity of the sample with an etched Ca cathode is not considerably higher than that of the other two samples, it is assumed that the cathode is etched away with the acid treatment. The intensity of the chlorine secondary ions does not differ considerably between the untreated and the treated OC3C8-PPV layers. A slightly higher Cl secondary ion intensity is seen in the etched OC3C8-PPV/ Ca sample, but the difference is not large enough to lead to a sure conclusion that Cl is present in the HCl-treated sample. Although it could not be explicitely shown that the presence of chlorine in hydrochlorid acid is harmful, the removal of the cathode with HCl did damage the device, leading to faster degradation after the removal of the cathode. In an optimal case the removing of the cathode has no influence on the performance of the device; additionally, for investigations of the underlying polymer layer one must be certain that the etching process does not damage the layer of interest. Therefore an 40 OC 3 C 8 -PPV + Ca, etched with HCl C Cl Ca In OC 3 C 8 -PPV as is OC 3 C 8 -PPV etched with HCl Sputtering time 10 -1 10 0 10 1 10 2 10 3 10 4 10 5 10 6 10 7 In te ns ity (c ou nt s/ s) a) b) c) Figure 4.2: Etching away the Ca cathode with diluted HCl: SIMS experiment. a) A reference sample with the structure ITO/OC3C8-PPV left as is. b) A sample with the structure ITO/OC3C8- PPV, etched with HCl. c) A sample with the structure ITO/OC3C8-PPV/Ca, the metal layer etched with HCl. alternative etching media was searched. Acetic acid proved to be a good alternative for hydrochlorid acid. Fig. 4.3 plots the temporal evolution of luminance of a reference diode with an original cathode as well as of a diode, the cathode of which was removed and replaced with a new one directly after preparation. No substantial differences between the two samples can be observed: the evolution of the luminance is very similar. The initial luminances were very similar as well. The operation of both samples was paused at 140 h for the recording of IVL characteristics; this explains the small-scale discontinuity in the luminance of both samples. The optimized procedure of removing the cathode with acetic acid is summarized in Fig. 4.4. It should be noted that for such samples, the normally used Al shielding layer is left out and the thickness of the Ca is increased to 100 nm. The calcium cathode is etched away by dropping few drops of acetic acid directly on the cathode stripe. After a waiting time of ca. 2 min, the acid is quickly washed away with deionized water and dried with nitrogen. The samples are subsequently dried in a vacuum oven at 100 ° C for 2 h. 4.2 Electrical stressing in accelerated conditions Typical t50 lifetimes for OC3C8-PPV are some hundreds of hours, which makes the electrical stress- ing very time-consuming. The possibility to shorten the electrical stressing is therefore desired in situations in which the available time is limited. This can be achieved by driving the devices at a higher current density, which leads to faster degradation. If a correlation between the lifetimes obtained at different current densities exists, the electrical stressing can be accelerated. 41 0 5 0 1 0 0 1 5 0 2 0 0 2 5 0 3 0 0 0 , 0 0 , 2 0 , 4 0 , 6 0 , 8 1 , 0 No rm aliz ed lu mi na nc e ( a.u .) T i m e ( h ) o r i g i n a l c a t h o d e n e w c a t h o d e Figure 4.3: The influence of the cathode removal with acetic acid. Black: the development of luminance of a reference sample with an original Ca cathode during operation. Red: a sample from the same preparation run; the original cathode was removed and replaced with a new one directly after the preparation of the samples. The operation of both devices was paused at 140 h for the recording of IVL characteristics. Both devices had the structure ITO / PEDOT:PSS / OC3C8-PPV/ Ca and were operated at a constant current density of 50 mA/cm2. Figure 4.4: Removing the cathode with acetic acid. 42 The initial luminance is proportional to the applied current density. In experiments with differ- ent organic materials, the t50 lifetime has been experimentally observed to scale with the initial luminance [49, 104, 105]. This can be expressed as the so-called scaling law: t50× Ln 0 = const (4.1) where L0 is the initial luminance and n a system specific acceleration factor. It was investigated whether the scaling law that was first introduced by Van Slyke and co-workers [49] and later by Popovic et al. [104] can be applied for OC3C8-PPV-based diodes. A set of diodes with the structure ITO / PEDOT:PSS / OC3C8-PPV/ Ca / Al were prepared and stressed electrically at current densities of 50, 100, 200 and 400 mA/cm2, leading to different values for initial luminance. The corresponding t50 lifetimes were then extracted. The t50 lifetimes of the diodes are plotted against the initial luminance on a double-logarithmic scale in Fig. 4.5. Despite the scattering of the lifetime values, there is a clear correlation between the initial luminance and the t50. 1 The parameters for the scaling law can be extracted by fitting the presented data with a linear fit, which is shown in the Figure as the dotted line. For the tested OC3C8-PPV-based devices, the fit yields an acceleration factor of approximately n = −4.2. It could be shown that the scaling law can be applied to OC3C8-PPV-based devices. It is therefore possible to perform the electrical stressing under accelerated conditions if needed. 1 0 - 2 1 0 - 1 1 0 0 1 0 1 1 0 2 1 0 3 1 0 2 1 0 3 C u r r e n t d e n s i t y 5 0 m A / c m 2 1 0 0 m A / c m 2 2 0 0 m A / c m 2 4 0 0 m A / c m 2 I n i t i a l l u m i n e s c e n c e ( c d / m 2 ) t 50 lif eti me (h ) Figure 4.5: t50 lifetime against initial luminance of diodes operated at current densities of 50, 100, 200 and 400 mA/cm2. The data sets were fitted using Eq. 4.1 and the fit is shown as the dotted line. The diode structure was ITO / PEDOT:PSS / OC3C8-PPV/ Ca / Al. 1 The scattering of the lifetimes as well as the initial luminance values is a common phenomenon. The I-V-L char- acteristics as well as the electrical stressing results presented in this thesis are therefore averaged from several devices. 43 5 Dark-CELIV investigations on OC3C8-PPV This chapter utilizes dark-CELIV for the investigation of equilibrium charge carriers in OC3C8-PPV. Due to the low intrinsic charge carrier density in PPVs, no extraction current is observed in dark- CELIV measurements of freshly prepared devices. However, it will be shown that during aging in the glovebox, an unexpected density of equilibrium charges is observed to develop. This storage-related charge carriers will be attributed to doping by residual oxygen in the glovebox, and the process will be monitored by dark-CELIV measurements. Additionally, the time scale of the establishment of the thermal equilibrium in doped OC3C8-PPV after charge carrier extraction by CELIV will be monitored experimentally. 5.1 Previous investigations with dark-CELIV Due to their low density of intrinsic charge carriers, organic semiconductors have typically been investigated with photo-CELIV, in which charge carriers are created by exciting the material with a laser pulse. However, in some cases the density of equilibrium charge carriers is large enough so that the photoexcitation of the semiconductor is not necessary and dark-CELIV can be utilized. In organic semiconductors, such equilibrium charge carriers can originate from doping, impurities or space charge near the contacts, for example [106]. In recent publications, dark-CELIV has been applied for the investigation of the influence of postproduction treatment of P3HT on the density of intrinsic charge carriers and the charge transport [106], a charge carrier reservoir at the inter- face TiO2/P3HT:PCBM [94] as well as the oxygen-induced doping concentration in P3HT:PCBM bulk heterojunction solar cells [107–110]. Additionally, one work connects equilibrium charges in P3HT:PCBM to injection from the contacts due to energy level alignment caused by Fermi level pinning [111], and a very recent publication utilizes dark-CELIV for the investigation of the time dependence of the build-up of a space charge at the interface ITO / P3HT [112]. 5.2 Origin of equilibrium charge in OC3C8-PPV To begin, Fig. 5.1 illustrates the energetic situations at different stages of a dark-CELIV measurement as performed throughout this chapter. The measured samples had the structure ITO / PEDOT:PSS / OC3C8-PPV/ Ca / Al. For clarity, band bending and the presence of charge carriers are omitted. On top left, the energetic levels of the used materials are illustrated prior to contact, in absence of thermal equilibrium. On top right, the situation is depicted after the equalization of the Fermi levels of the contacts, i.e. in thermal equilibrium, with zero volt applied bias. The work function 45 difference of the electrode materials induces a built-in voltage UBI to the sample. The dark-CELIV ramp sequence is illustrated on bottom left. Before the linear voltage ramp, a constant offset voltage Uo f f set = 2 V is applied in order to compensate the built-in voltage (bottom middle). The dark-CELIV voltage ramp is then applied in reverse bias (bottom right), meaning a negative applied voltage at the ITO / PEDOT:PSS contact. The reverse bias is important in order to prevent injection of charge carriers during the voltage ramp. Figure 5.1: A schematic illustration of the dark-CELIV measurement as performed throughout this chapter. Top left: The energetic levels of the materials being not in thermal equilibrium. Top right: The situation in thermal equilibrium with zero volt applied voltage. The equalization of the Fermi levels of the contact materials induces a built-in voltage UBI that equals the work function difference of the electrodes. Bottom: The dark-CELIV measurement sequence. The dark-CELIV voltage ramp is illustrated on the left; before the ramp, an offset voltage Uo f f set = 2 V is applied (1). The offset voltage approximately compensates the built-in voltage of the device, leading to flat band conditions. The linear voltage ramp in reverse bias (2) means a negative applied voltage on the ITO / PEDOT:PSS contact; the reverse bias prevents the injection of charge carriers from the contacts. Without a significant intrinsic charge carrier concentration, the linear dark-CELIV voltage ramp results in a constant rectangular current response that corresponds the capacitive displacement current (see Section 3.3.3). Due to the low charge carrier density in undoped OC3C8-PPV, this is expected to be the case for the dark-CELIV current response of a standard diode with the structure ITO / PEDOT:PSS / OC3C8-PPV/ Ca / Al as was used in this work. Indeed, diodes that are mea- sured in a fresh state shortly after preparation show only the rectangular displacement current, as shown in Fig. 5.2 a). In both graphs in Fig. 5.2, CELIV measurements are plotted for samples with 46 the standard structure ITO / PEDOT:PSS / OC3C8-PPV/ Ca / Al at Uo f f set = 2 V. In photo-CELIV measurements the samples were excited optically at 506 nm, i.e. at the absorption maximum of OC3C8-PPV, 5µs prior to the beginning of the triangular voltage ramp; the photo-CELIV responses of both samples exhibit a significant peak originating from the extraction of the photogenerated charge carriers. Additionally and contrary to the expectations, in some measurements as plotted ex- emplarily in Fig. 5.2 b), a distinct extraction current signal originating from the extraction of mobile charge carriers is present in the dark-CELIV response in addition to the displacement current. Figure 5.2: Exemplary current transients of two CELIV measurements without (dark-CELIV, rectangles) as well as with optical excitation (photo-CELIV, triangles). An offset voltage Uo f f set = 2 V was applied in both measurements. In photo-CELIV, the sample was excited with a laser pulse at 506 nm 5µs prior to the beginning of the triangular voltage ramp. a) A fresh sample measured directly after preparation; b) a sample that was measured after a longer (some weeks) storage in the glovebox. Detailed investigations showed that such an extraction current without the optical generation of charge carriers is primarily observable when the samples are measured after storing them in the glovebox for several days. It was possible to conduct successful CELIV measurements without the appearance of extraction current in the dark on freshly prepared as well as electrically stressed samples (the results will be shown in Chapter 6), provided that the measurements were started immediately after preparation without a storing time in the glovebox. The extraction current in dark-CELIV did not increase with electrical stress and is therefore not directly related to electrical fatigue. It is thus a storage effect that is related to aging during storage in the glove box. The similarity of the form of the extraction current peaks in Fig. 5.2 is obvious. The extraction current in the photo-CELIV measurement is caused by holes, which are known to have a significantly higher mobility than electrons in PPV [113–115]. They are created by the optical excitation of the sample and stem from the bulk of the polymer layer. The similarity of the form of the peaks as well as the positions of the current maxima give hints on the type of charge carriers that cause the dark- CELIV signal. In CELIV, the mobility is calculated utilizing Eq. 3.5. The calculated mobilities for the photo- and dark-CELIV responses in Fig. 5.2 a) and b) are 5 ·10−6 cm2 Vs and 4 ·10−6 cm2 Vs , respectively. 47 The mobility of electrons in PPV derivates is known to be considerably smaller than that of holes [115], and the obtained values are both typical hole mobilities in OC3C8-PPV. It is concluded that the dark-CELIV voltage ramp extracts holes from OC3C8-PPV. Next, the location of these holes is of interest. In general, in present devices there are three possi- bities for the location of the equilibrium charge that is extracted with the dark-CELIV voltage ramp: the interface PEDOT:PSS / OC3C8-PPV, the interface OC3C8-PPV/ Ca and the bulk of the polymer. As depicted in Fig. 5.1, the dark-CELIV voltage ramp in reverse bias extracts holes at the PEDOT:PSS contact and electrons at the Ca contact. As discussed above, it is known that the extracted charge carriers are holes. A hole accumulation region at the interface PEDOT:PSS / OC3C8-PPV would be extracted immediately at the contact and would therefore not cause a CELIV extraction current re- sponse as observed in the present devices. Accordingly, if the extraction current stems from a space charge region at one of the contac