Performance enhancement of hybrid nanocrystal/polymer bulk heterojunction solar cells: aspects of device efficiency, reproducibility, and stability

Transkript

1 Dissertation zur Erlangung des Doktorgrades der Technischen Fakultät der Albert-Ludwigs-Universität Freiburg im Breisgau Performance enhancement of hybrid nanocrystal/polymer bulk heterojunction solar cells: aspects of device efficiency, reproducibility, and stability Michael Eck Albert-Ludwigs-Universität Freiburg im Breisgau Technische Fakultät Institut für Mikrosystemtechnik 2014

2

3 Dekan: Prof. Dr. Yiannos Manoli Referenten: Dr. Michael Krüger Prof. Dr. Margit Zacharias Autor: Michael Eck Geboren: , Arad (Rumänien) Datum der Abgabe: Datum der Disputation:

4

5 I Erklärung Ich erkläre, dass ich die vorliegende Arbeit ohne unzulässige Hilfe Dritter und ohne Benutzung anderer als der angegebenen Hilfsmittel angefertigt habe. Die aus anderen Quellen direkt oder indirekt übernommenen Daten und Konzepte sind unter Angabe der Quelle gekennzeichnet. Insbesondere habe ich hierfür nicht die entgeltliche Hilfe von Vermittlungs- oder Beratungsdiensten (Promotionsberaterinnen oder Promotionsberater oder anderer Personen) in Anspruch genommen. Niemand hat von mir unmittelbar oder mittelbar geldwerte Leistungen für Arbeiten erhalten, die im Zusammenhang mit dem Inhalt der vorgelegten Dissertation stehen. Die Arbeit wurde bisher weder im In- noch im Ausland in gleicher oder ähnlicher Form einer anderen Prüfungsbehörde vorgelegt. Ich erkläre hiermit, dass ich mich noch nie an einer inoder ausländischen wissenschaftlichen Hochschule um die Promotion beworben habe oder gleichzeitig bewerbe. Datum/date: Unterschrift/signature:

6 II Table of Content Abstract... VII Zusammenfassung... VIII Acknowledgement... IX 1. Motivation The potential of solar energy Solar cells for micro energy harvesting Thesis outline Background Working principle of hybrid heterojunction solar cells Device structure of hybrid heterojunction solar cells Donor and acceptor materials Semiconducting polymers Semiconducting nanocrystals Synthesis of colloidal semiconducting nanocrystals Post-synthetic nanocrystal treatment Historical background and state of the art Evolution of hybrid bulk heterojunction solar cells Hybrid nanostructured heterojunction solar cells Degradation of hybrid heterojunction solar cells Methods Nanomaterial characterization Photoluminescence (PL) spectroscopy UV-Vis absorption spectroscopy Fourier transform infrared spectroscopy (FTIR)... 40

7 III Dynamic light scattering (DLS) Transmission electron microscopy (TEM) Scanning electron microscopy (SEM) Atomic force microscopy (AFM) Solar cell fabrication Utilized solar cell designs Hot injection CdSe nanocrystal synthesis One-pot CdSe nanocrystal synthesis Nanocrystal dispersion formulation Formation of CdSe QD-TrGO hybrid material Preparation of conducting oxide substrate Formation of alumina and titania nanostructures Polymer/nanocrystal blend formulation and deposition Solar cell characterization Sun Simulator Solar cell measurement Resistive effects Determination of charge carrier mobility Results - Solar cell reproducibility Intra-batch NC variations Influence of post-synthetic NC treatment Influence on solar cell efficiency Summary Batch-to-batch NC variations Influence of the post-synthetic NC washing time on the solar cell performance... 87

8 4.4 Comparison of NC ligand sphere from HDA/TOPO and HDO synthesis Influence of post-synthetic NC washing time over thermal solar cell stability Conclusion Results - Efficiency enhancement Elongated CdSe nanocrystals for BHJ solar cells Strategies for efficiency improvement Variating the blend solvent Low-bandgap polymer as donor material Use of polymer chain interlinking additive Removal of NC aggregates Inter-batch differences Summary and conclusion of nanorod/polymer solar cell experiments Use of nanorod/quantum dot mixtures for hybrid BHJ solar cells CdSe NC/Graphene hybrid for BHJ solar cells TrGO-CdSe hybrid material TrGO-CdSe/polymer hybrid film TrGO-CdSe/PCPDTBT hybrid solar cell performance Charge extraction Discussion and conclusion Nanostructured inverted hybrid solar cells Manufacturing of titania nanotube arrays on ITO Inverted solar cells from TiO 2 nanotubes on ITO Results - Solar cell degradation Degradation of hybrid BHJ solar cells in inert atmosphere Utilized NCs and their post-synthetic treatment Initial performance IV

9 V Performance over time Summary Degradation of unsealed hybrid BHJ solar cells in air Initial performance Performance in air Summary Degradation of nanostructured inverted hybrid solar cells Summary Solar cell reproducibility Efficiency enhancement Solar cell degradation Outlook Appendix A 1. Temperature dependency of the NC photoluminescence A 2. Ligand exchange experiments of CdSe NCs A 2.1 In-situ ligand exchange mapping A 2.2 NC ligand and solar cell performance A 3. Photoluminescence quantum yield calculation A 4. FTIR measurements of CdSe QDs before and after HA washing A 4.1 FTIR spectrum of TOPO A 4.2 FTIR spectrum of HDA A 4.3 FTIR spectrum of HA A 4.4 FTIR spectrum of MeOH A 4.5 FTIR spectrum of CdSe QDs A 4.6 Conclusion

10 A 5. Solar cells containing CdSe NCs from a one-pot synthesis at 200 C A 6. Anodization of nm titanium films on ITO A 7. Towards nanostructured hybrid solar cells using an alumina nanopore scaffold A 7.1 Formation of alumina nanopores on ITO A 7.2 Polymer pore-filling experiments References Publications and conference contributions Table of abbreviations..200 VI

11 VII Abstract Hybrid nanocrystal (NC)/polymer bulk heterojunction (BHJ) solar cells consist out of semiconducting polymers as main electron donor and semiconducting nanocrystals as electron acceptor. BHJ hybrid solar cells utilize the advantage of a large interfacial area between the donor and the nanoscopic acceptor material upon mixing both components, to allow for an efficient dissociation of photoexcited excitons. However, this concept has a disadvantage of discontinuous pathways for the subsequent electron and hole extraction, and in the specific concept of NC/polymer solar cells has the drawback of the variation of quality in the utilized NCs and their surface constitution. The aim of this thesis is to elaborate how to enhance the performance including power conversion efficiency, reproducibility, and stability - of BHJ NC/polymer solar cells. Therefore, several approaches towards the increase of reproducibility, improvement of the charge extraction, and durability of hybrid nanocrystal/polymer solar cells are elaborated, whereby the focus is laid on the CdSe NCs synthesis, processing, and integration into the solar cell. Within this thesis it is described how to reproducible manufacture BHJ NC/polymer solar cells on a high, state-of the art efficiency level. For the attempt to improve the charge extraction elongated NCs, so called nanorods (NRs), were utilized in order to provide a more direct pathway for the electron transport out of the photoactive layer compared to quantum dots (QDs). With the same aim, CdSe QDs attached to reduced graphene oxide were used as electron acceptor, thereby reaching with 4.2% the up to date highest power conversion efficiency (PCE) of single layer BHJ solar cells using a polymer/qd donor-acceptor system. This thesis also emphasizes the importance of the correlation of several NC quality factors with the post-synthetic treatment of the NCs and with the PCE, but as well with the thermal and long-term stability of the resulting hybrid solar cell. Finally, also the performance stability of hybrid solar cells in air was investigated, which also shows a strong impact of the post-synthetic NC treatment.

12 VIII Zusammenfassung Hybride Nanokristall (NC)/Polymer Bulk-Heteroübergang (BHJ) Solarzellen bestehen aus halbleitenden Polymeren als Haupt-Elektronendonator und halbleitenden Nanokristallen als Elektronenakzeptor. Hybride BHJ Solarzellen nutzen den Vorteil einer großen spezifischen Oberfläche zwischen dem Donor und dem nanoskopischen Akzeptor Material, um eine effiziente Dissoziation der durch Photonen angeregten Exzitonen zu gewährleisten. Jedoch hat dieses Konzept den Nachteil diskontinuierlicher Pfade für die nachfolgende Extraktion von Elektronen und Löcher. Zusätzlich hat das Konzept der hybriden BHJ Solarzelle den Nachteil der Variation der Qualität und Oberflächenbeschaffenheit der verwendeten Nanokristalle. Ziel der vorliegenden Dissertation ist eine Verbesserung der Performance von BHJ NC/Polymer Solarzellen zu erzielen, und zwar sowohl durch Verbesserung ihres photovoltaischen Wirkungsgrades, als auch ihrer Reproduzierbarkeit und Haltbarkeit. Dabei wird der Fokus auf die Synthese, Prozessierung, und Integration von CdSe Nanokristallen in die hybride BHJ Solarzelle gelegt. In dieser Dissertation wird beschrieben, welche Prozessschritte nötig sind, um reproduzierbar BHJ NC/Polymer Solarzellen auf einem hohen, dem Stand der Technik entsprechenden, Wirkungsgrad herzustellen. Für den Versuch der Verbesserung der Ladungsträgerextraktion wurden elongierte Nanokristalle, sogenannte Nanorods (NRs), verwendet. Diese sollten einen direkteren Pfad für den Ladungsträgertransport aus der photoaktiven Schicht im Vergleich zu sphärischen Nanokristallen, sogenannte Quantenpunkte (QDs), erlauben. Zum selben Zweck wurde desweiteren ein Hybridmaterial aus an reduziertem Graphitoxid angehefteten CdSe QDs als Elektronenakzeptor in BHJ Solarzellen eingesetzt. Dadurch wurde mit 4.2% die bisher höchste Effizienz für BHJ Solarzellen, die aus einer einzigen photoaktiven QD/Polymer Schicht bestehen, erreicht. Weiterhin verdeutlicht diese Dissertation auch die Wichtigkeit der Betrachtung und Korrelation verschiedener NC Qualitätsparameter zur postsynthetischen NC Behandlung, sowie deren Einfluss auf die resultierende Effizienz und Langzeitstabilität sowie der thermischen Belastbarkeit von BHJ NC/Polymer Solarzellen. Abschließend wurde auch die Degradation der hybriden BHJ Solarzellen an Luft untersucht, auf die die post-synthetische Behandlung der Nanokristalle ebenfalls einen großen Einfluss hat.

13 IX Acknowledgement At first I like to thank Dr. Michael Krüger and Prof. Gerald Urban for their support by accepting me as Ph.D. student, thus allowing me to conclude with this Ph.D. thesis. I especially thank my group leader Dr. Michael Krüger for his constant moral support, and for lots of discussions, which helped to develop many innovative ideas only a few of which are presented within this thesis. Furthermore, I want to thank him for many hours and late evenings spent for prove reading of abstracts, manuscripts, and presentations, and thereby providing very important feedback throughout the time of my study. Next, I want to thank Dr. Yunfei Zhou for introducing me to the synthesis, post-synthetic NC treatment, solar cell fabrication, and solar cell manufacturing. His effort when I was still a foreigner to this field allowed me to be fastly able to conduct my own research. Furthermore, I thank Dr. Frank-Stefan Riehle and Chuyen Van Pham for numerous discussions, also besides the scientific area, which I greatly enjoyed as an origin of inspiration. My colleague Ying Yuan I want to thank for being always available when I needed help in the laboratory with the NC synthesis or with anything else. Also, I thank my coworker Simon Einwächter, who put a lot of effort in setting up the in-situ PL spectrometry of NCs. In continuation, I thank Dr. Ralf Thomann for introducing me to the SEM, and performing many TEM measurements for me. Dr. Yi Thomann I acknowledge for introducing me to the AFM, and always being spontaneously available for answering my questions regarding microscopy. I thank Sebastian Schütt for introducing me to XRD and performing XRD measurements even on short notice. I thank Dr. Alex Fauler for his kind support, helping me to overcome technical problems with vacuum deposition machines. Clemens Veit *, Hans- Freider Schleiermacher *, and Martin Sessler * I want to thank for performing solar cell measurements in their laboratory for confirmation of checking my own characterization setup, and being helpful with many aspects of the solar cell production and characterization. Also, I thank Dr. Franziska Scholz for performing and explaining me the FTIR spectroscopy. I thank Dr. Alina Chanaewa for providing me with ZnO NCs, and Andreas Schreiber for preparing protein coated NCs; I m sorry that our cooperation did not result in any publication. * Fraunhofer Institute for Solar Energy Systems (ISE), Heidenhofstr. 2, Freiburg, Germany

14 X Furthermore, I thank Dorothea Scheuneman for performing high quality 3D TEM measurements, Simon Züfle and Martin Neukom for performing CELIV and IS spectroscopy, which were all very helpful for my research. I also want to thank everybody mentioned before - and also Alfian Ferdiansyah Madsuha, Jan Waldmann, Rahul Valecha, Jyotika Chhibber Valecha, Feng Mei, Cong Men, Dr. Gregory Stevens, Dr. Andreas Menzel and others - for offering me their friendship and making my Ph.D. study at the Freiburg Materials Research Center an enjoyable time. Finally, I thank the DFG GR1322 Micro Energy Harvesting graduate school for supporting me with a scholarship. Department of Physics, Carl von Ossietzky University, Carl-von-Ossietzky-Str. 9-11, Oldenburg, Germany Institute of Computational Physics, Zürich University of Applied Sciences (ZHAW), Technikumstr. 9, Winterthur, Switzerland, & Fluxim AG, Technoparkstr. 2, Winterthur, Switzerland Meinen Eltern und Großeltern.

15 1 1. Motivation 1. Motivation 1.1 The potential of solar energy The general motivation for the development and use of solar cells is the avoidance of the use of dwindling and increasingly contested non-renewable energy sources. Solar cells can, broadly spoken, produce electricity from the sun. Thus, using an abundant energy source - accessible to everybody - by the conversion of photons into electric energy (photovoltaic effect). Thus, solar cells are able to deliver one of the most valuable goods of the modern world, already facing the consequences of climate change also induced by the exploitation of its fossil energy sources. In 2011 about 18% of the consumed energy worldwide was used as electricity, which is furthermore the fastest growing form of consumed energy nowadays [1]. However, still 68% of the produced electric energy is created from fossil energy sources (see Fig. 1.1). Figure 1.1. Left: Development of the world energy consumption by different energy sources between the years 1971 and Right: Development of the world electricity production divided into energy sources. (Images are based on graphs from Ref. [1] ). In contrast, electricity created by photovoltaic (PV) accounted for only 0.35% of the consumed electricity worldwide in 2011, but already for 0.5% in Germany s share in 2012 therein is of about 1 / 3 of the worldwide installed PV capacity. However, in recent years Over the period

16 1. Motivation 2 the raise of solar energy production mostly originates from increase of the installed PV capacity in China and America. Hence, PV is nowadays a fast growing source of electricity and its potential in contribution to the world energy consumption is far from being fully exploited. Only by the use of the solar irradiation on an area as large as Spain s, solar cells with a power conversion efficiency (PCE) of 10% could provide the complete world energy consumption. Moreover, the consumption of a large portion of fossil fuels could be substituted by the consumption of electricity, as well as the produced electricity could be produced of a higher fraction by renewable energies; thus, helping to reduce the emission of greenhouse gases. 1.2 Solar cells for micro energy harvesting Another beneficial aspect of PV is that it can provide energy independent of an energy distribution infrastructure, which is important for less developed countries. Moreover, these countries also are usually geographically situated in areas around the equator that benefit from a higher solar power irradiation than the world average. The grid independence of photovoltaic power supply is also important for applications of energy autonomous sensors, actuators, or displays. Herein, a PV power supply is reducing wiring and could even substitute batteries to reduce maintenance costs. However, in this case the PV power supply would have to provide enough energy, posses an adequate lifetime, and potentially also a module for energy storage would have to be included. For the purpose of empowering devices that don t necessarily require the highest PCE, solar cells produced directly from solution like the ones presented by this thesis are an alternative to bulk inorganic solar cells [2]. Their advantage lies in potentially lower manufacturing and module installation costs [3], and the possibility of manufacturing of flexible modules [4, 5]. The highest PCEs nowadays with solution processable solar cells are reached with organic bulk heterojunction (BHJ) solar cells with PCEs of over 10% [6, 7]. But, there are also purely inorganic solar cells from nanocrystals (NCs) that are processable from solution with PCEs of up to 7% [8]. Recently, Global Market Outlook for Photovoltaics , European Photovoltaic Industry Association (EPIA) Calculation assumes that the average solar irradiation is of 1000 W/m² with a daily average of 250 W/m². The resulting energy produced at a 10% PCE on an area of 505,370 km² would be of about PWh, and the world energy consumption in 2011 was of PWh [1].

17 3 1. Motivation a solar cell combining inorganic photoactive NC layers incorporating an additional NC/polymer blend layer resulted in a PCE of 5.5% [9]. However, for the solar cell type used in this thesis - consisting out of a single photoactive layer out of a polymer/nanocrystal mixture in the form of hybrid BHJ solar cells only a few reports claim PCEs exceeding 4% [10, 11]. Also, this solar cell type is far away from commercialization; but due to the similarities to allorganic BHJ solar cells they profit from developments towards commercialization that are made for organic PV cells [12]. Nevertheless, using inorganic NCs instead of organic acceptor materials, has the advantage of a higher dielectric constant (improving exciton dissociation), a higher intrinsic charge carrier mobility (improving electron extraction), tunable properties of the electronic structure of the inorganic semiconductors by change of the NC size (e.g. for adaptation to different donor polymers), and the possibility of creating different shaped NCs (i.e. quantum dots, nanorods, tetrapods, and multibranched) for facilitating the electron extraction. The specific drawbacks for hybrid nanocrystal/polymer BHJ solar cells lay in the needed integration of the inorganic NCs with organic polymer, which requires organic surfactants (ligands) around the NCs to enable a cosolvation with the polymer. These ligands are also needed for the stabilization of the NCs during their manufacturing by colloidal synthesis in solution. However, these ligands represent an insulation layer around the NCs [13], reducing the charge carrier mobility between the NCs, which is required to be high for effectively serving as electron extraction material. Other specific problems of NCs are surface charge traps [8], NC aggregation leading to shunts of the thin film solar cells and reduction of the output current, and a varying NC quality throughout the NC synthesis. Moreover, a general drawback of the bulk heterojunction concept is the uncontrolled nanomorphology in between the donor and acceptor phase. 1.3 Thesis outline This thesis deals with bulk heterojunction hybrid nanocrystal/polymer solar cells. This type of solar cell was first published by Neil C. Greenahm et al. [14] in Since then this solar cell concept has been subject of many research groups worldwide (see Chapter 2.6), greatly improving the initially obtained PCE. When I started my research on BHJ solar cells, my coworker at that time Yunfei Zhou, had already published a paper about 2% efficient CdSe

18 1. Motivation 4 quantum dot/polymer solar cells based on a non-ligand-exchange post-synthetic NC treatment [15] based on hexanoic acid (HA), conceived by Frank-Stefan Riehle [16], aimed to reduce the amount of insulating ligands around the NC. One of the points, which I intend to further elucidate within this thesis is, whether one of the assumptions made by Y. Zhou within his thesis [17] that the NC photoluminescence intensity (and thus the NC surface quality 18, 16] ) is dictating the power conversion efficiency of hybrid NC/polymer solar cells is true. The post-synthetic treatment based on HA has the advantage of gradual reduction of the initial NC ligand sphere, instead of a complete removal of the NC surface ligands with commonly used methods (e.g. pyridine ligand exchange). Thus, the post-synthetic treatment can and has to be adapted to NCs of different surface constitutions in such a way that colloidal stability of the NCs is still given, but sufficiently NC ligands are removed to allow an efficient electron extraction over the NCs within the hybrid BHJ solar cell. Implementing criteria for finding the optimal post-synthetic HA treatment for the NCs, thereby allowing the reproducible production of hybrid BHJ solar cells, was also part of my thesis. Therefore, I drew a correlation between the varying NC quality during their synthesis (observed by an insitu observation of the NC synthesis), the optimal post-synthetic NC treatment time (observed by photoluminescence and UV-Vis spectroscopy, and dynamic light scattering), and the resulting solar cell performance (Chapter 4.1). Thereby, it was found that hybrid solar cells perform best when using NCs from the point of highest NC homogeneity within their synthesis, and that the photoluminescence intensity is more important for indicating the time (or intensity) of the post-synthetic NC treatment. Subsequently, the prior obtained results were applied on an optimized CdSe quantum dot synthesis [19], to achieve solar cells with state of the art efficiencies of up to 3% (Chapter 4.2). Moreover, within that chapter is was illustrated that one can already from measuring the solar cell before and after a short thermal annealing roughly predict the optimal post-synthetic hexanoic acid NC treatment time. Also, several attempts were undertaken to improve the solar cell efficiency. The first attempt presented in this thesis describes the utilization of elongated NCs (Chapter 5.1). Therein, so-called nanorods (NRs) should provide a more directed pathway for electron extraction from the hybrid NC/polymer solar cell. Despite not resulting in PCEs higher than 1.9% this attempt is described within this thesis, since it illustrates problems and occurring in manufacturing of BHJ hybrid solar cells and strategies for solar cell performance improvement in practice, which might be of interest for other researchers within this field.

19 5 1. Motivation Furthermore, I developed the integration of a hybrid material of CdSe quantum dots (QDs) chemically attached to graphene via a thiol group [20] ) into the BHJ hybrid solar cell (Chapter 5.2). Thereby, the graphene is presumably directly accessing the attached nanocrystals for charge transfer and further directed extraction over the graphene. Thus, the PCE could be increased from 3% up to 4.2% [21], which represents the highest PCE of BHJ quantum dot/polymer solar cells to best of my knowledge. In addition, since very little has been reported about the degradation of BHJ NC/polymer solar cells, this aspect was also explored (Chapter 6), wherein the influence of the postsynthetic NC treatment on the long-term device stability and the behavior of the solar cells in air is discussed. Moreover, in order to further elucidate aspects of the hybrid solar cell degradation and to attempt a complete charge extraction of the photoexcited charge carriers from the solar cell, nanostructured interdigitated solar cells were built (Chapters 5.3 & 6.3).

20 2. Background 6 2. Background (Parts of this chapter are published in: Polymer-Nanocrystal Hybrid Solar Cells, M. Eck and M. Krüger, chapter 6 in Organic Photovoltaics, ISBN , Wiley-VCH, 2014, pp ) Within this chapter at first the working principle of hybrid heterojunction solar cells is introduced, followed by the introduction of its actual device structure. Subsequently, the properties of the materials responsible for the photo-energy conversion (i.e. semiconducting nanocrystals and polymers) are discussed. Within the then following section the principle of colloidal NC synthesis is explained, as it is necessary to understand the potential quality differences of the produced NC originating from their synthesis. Finally, an overview about the development and state of the art of BHJ hybrid solar cells is given, including nanostructured solar cells and solar cell degradation. 2.1 Working principle of hybrid heterojunction solar cells Hybrid NC/polymer solar cells are working by the principle of heterojunction solar cells (see Fig. 2.1), meaning that two different semiconducting materials are used, one as electron donor and the other as electron acceptor. In case of hybrid heterojunction solar cells, conjugated semiconducting polymers are usually used as organic donor materials, and semiconducting NCs usually act as acceptor materials. Together, both components form the active layer of the solar cell. As both materials are able to absorb light, the following figure is describing the physical processes inside the heterojunction solar cell both for the case that the donor material is absorbing (Fig. 2.1a) and in case the acceptor material is absorbing a photon (Fig. 2.1b).

21 7 2. Background a) b) Figure 2.1. Working principle of a heterojunction solar cell and physical processes taking place within the photoactive layer at the donor-acceptor interface. (a) In case the photon is absorbed by the donor material: (1) Photon absorption by a donor leads to exciton generation, (2) Exciton diffusion, (3) Charge separation between electron and hole at the donor-acceptor interface & electron transfer towards the acceptor, (4) Charge transport of electrons along the acceptor material, simultaneous transport of holes along the donor material & collection of electrons by the cathode and hole collection by the anode. (b) Working principle in case the photon is absorbed by the acceptor material. (Figure reused with permission from Ref. [22] ). In the following, the working principle of a hybrid solar cell is explained by division into four different steps, for the common case that light is absorbed in the donor material according to Figure 2.1a. 1. Photon absorption: When a photon is absorbed by the donor material with an energy that corresponds at least to its bandgap energy, an electron is excited from the HOMO (highest occupied molecular orbital) to the LUMO (lowest occupied molecular orbital) level. In conjugated polymers the excited electron has a reduced mobility, because it is bound by coulomb attraction to the generated hole with typical binding energies ranging between mev [23, 24]. Thereby this electron-hole pair is regarded as a quasi-particle, the so called exciton. 2. Exciton diffusion: The generated exciton diffuses along the polymer for a certain distance before recombination occurs, leading to the emission of light or/and the release of thermal energy. The average distance an exciton can dislocate from its place of generation within its lifetime is called exciton diffusion length, which in conjugated polymers, is typically in the range between nm [25 27], depending on the polymer.

22 2. Background 8 3. Charge separation & transfer: In case the exciton reaches the donor-acceptor interface, its electron is attracted by the energetically lower positioned LUMO of the acceptor material. Thereby the exciton can dissociate into a free electron and into a free hole, with the later one remaining in the HOMO of the donor material. This process requires energy on behalf of the overall power conversion efficiency. Therefore, it is important that the LUMO levels of the donor and acceptor are matching energetically which will be discussed in more detail in Section Charge transport & collection: The electrons are transported through the acceptor material towards the cathode, which is usually a metal with a low work function (e.g. Al), slightly below the energy level of the acceptor LUMO. In the case of semiconductor NCs as acceptor material, the electrons are moving by hopping along the individual interconnected network of NCs towards the cathode [28]. Holes are separately moving through the donor material, which is a major benefit of a heterojunction solar cell [29] over a single junction cell, thereby reducing electron-hole recombination. Finally, holes are collected by the anode, after they passed the highly hole conductive, but electron blocking PEDOT:PSS (poly(3,4-ethylenedioxythiophene): poly(styrenesulfonic acid)) layer. Inside the polymer phase the holes are moving by intra-chain conduction as along the chain, as well as by inter-chain hopping processes towards the anode. The electron and hole transport is driven by an internal electric field deriving from the Fermi level differences of the electrodes, also the charge carrier mobilities are an important criterion for efficient charge transport. Photons can also be absorbed by the NCs (Fig. 2.1b). In this case, the excitons must be generated near an acceptor material to be able to dissociate at the interface between NC and polymer. An exciton that would be generated in the center of a NC aggregation would therefore recombine. Holes are transferred to the polymer HOMO and the electron remains inside the NC. Then electrons and holes are transported, as described before, to the respective electrodes.

23 9 2. Background 2.2 Device structure of hybrid heterojunction solar cells The basic device structure of a hybrid solar cell is depicted in Figure 2.2 (left). The bottom layer of the solar cell consists out of a thin transparent conductive layer on a transparent substrate. Typically a thin layer of indium tin oxide (ITO) coated on a supporting glass or transparent polymer foil is used, because ITO is a semiconductor with a wide bandgap that is absorbing only highly energetic photons and therefore has a high light transmittivity in the visible range [30], which is important because the light absorption should occur only in the active layer of the solar cell. On top, a hole conductive, but electron blocking PEDOT:PSS layer (hole extraction layer) of about 50 nm is deposited e.g. by spin coating, which is also only absorbing in the deep UV region [31]. The PEDOT:PSS layer enhances the adhesion of the upper light absorbing layer to the relatively rough ITO surface and improves the overall device stability by hindering oxygen and indium diffusion into the photoactive layer [32 34]. After drying and hardening of this hole extraction layer, the active layer is deposited on top. It consists out of a blend of conjugated polymers and NCs, exhibiting a typical thickness of about 100 nm after film drying. On top a thin metallic electrode layer for the electron collection is deposited typically by thermal evaporation of e. g. Al, LiF/Al or Ca/Al. This top electrode layer is called cathode, although - contrary to the general definition of a cathode - electrons are not injected into the device over it, but are extracted from this top contact. However, the name cathode is generally utilized for this solar cell type for the electron extraction contact due to the similar device structure with organic light emitting diodes (OLEDs) - possessing the opposite working principle of a solar cell - for which electrons are injected over the top electrode into the electron transport layer. The use of the name cathode in solar cells inversely to the classical definition can more generally be explained by their close similarity to a diode structure. Both - diode and classical solar cell - consist out of a p-n junction. In case of a diode, in order to allow for a current- flow through the device, electrons must be injected into the n-type material in order to overcome the built-in potential of the depletion zone formed at the p-n interface. Thus, for a diode the cathode connection is defined to be at the side of the n-type material. Since for a photodiode the device setup is the same, identical names are used for the connections of cathode and

24 2. Background 10 anode. Although, in case of operating the diode as photovoltaic cell electrons are extracted from the device over the cathode connection. Figure 2.2. Left: Schematic structure of a hybrid solar cell. Right: different morphologies for the realization of the active layer: bilayer heterojunction (A), bulk heterojunction (B), and interdigitated heterojunction (C). (Figure reused with permission from Ref. [22] ). The different possible structures of the active layer of a heterojunction solar cell are depicted in Figure 2.2 (right). The first possibility is to create a bilayer (A) between a semiconducting donor and a different semiconducting acceptor material, thus creating a single heterojunction interface. The second possibility is to create a blend (B) of donor and acceptor by mixing the two components, and thereby achieving an increased donor-acceptor interfacial area after the drying process where phase separation of the donor and acceptor phases usually occurs depending on the solvents, the solvent evaporation conditions, the solubility of the materials and other factors [35 38]. Therefore, improved internal structures, where nanophase separation guarantees a sufficient charge separation and transport, are relaying on multiple parameters which have to be explored and optimized for each material system. This so-called bulk-heterojunction (BHJ) structure is now the most commonly used active layer structure. A lack of nanophase separation as well as a lack of a continuous network formation of donor and acceptor phases towards the respective electrodes has to

25 11 2. Background be avoided. When the acceptor and donor material phases form larger regions, the donoracceptor interface - where exciton dissociation takes place - is too far away, and the excitons are recombining and are lost for power conversion into electricity. If the donor acceptor phases are not continuous it can also happen that charges are trapped on their way to the electrode inside the active layer by so called dead ends of the acceptor or donor phase [39 41]. Therefore, the blend morphology is crucial and there is a clear structure-function relation between active layer morphology and solar cell performance. Methods for controlling the blend morphology to some degree are described in more detail later in Chapter 5.1 and 5.2. Figure 2.2c illustrates an idealized interdigitated structure (C) in which the donor and acceptor nano-phases are aligned in parallel within the exciton diffusion limit, where every donor-acceptor interface can be reached by the exciton before its recombination, and additionally both electron and hole have a continuous connection within the acceptor and donor phase for extraction towards their respective electrodes. Examples and methods towards the realization of such structures are described in Chapters 2.6.3, 3.2.7, 5.3, A6, and A Donor and acceptor materials Like previously described, the heterojunction solar cell consists of two different semiconducting materials with different LUMO and HOMO levels and possibly different bandgaps. The LUMO level of the donor must be energetically higher than the acceptor LUMO to be able to donate its electron. The acceptor HOMO level must lie under the donor HOMO, so that the hole is not attracted by the electron acceptor material and thus leading to the recombination with the electron. This is schematically shown in Figure 2.3. Calculated and measured HOMO and LUMO levels of some semiconducting polymers are mentioned in Table 2.1 and the respective energy levels of NC based acceptor materials are summarized in Table 2.2. For efficient charge separation and transfer at the donor-acceptor (D-A) interface, the band energy levels have to match. An optimum energy gap LUMO of about 0.3 ev was proposed between the LUMO level of the donor and the LUMO level of the acceptor in order to enable an effective charge separation [42 44]. At the same time a maximum possible open circuit voltage (V OC ) in the device must be granted [44, 45] for a high device efficiency. This is

26 2. Background 12 schematically described in the following Figure 2.3. The V OC is considered being proportional to the energy gap between the HOMO level of the donor and the LUMO level of the acceptor following empirical formula from Scharber et al. [44] for its calculation: V OC = 1 e E HOMO,don or E LUMO,acceptor 0.3V (2.1),while V OC is the open circuit voltage of the solar cell, e is the elementary charge, E HOMO, donor is the HOMO level of the donor in ev, and E LUMO, acceptor is the LUMO level of the acceptor in ev. Figure 2.3. Schematic illustration of the donor-acceptor band matching. A certain energy difference LUMO is necessary for efficient charge separation. The V OC is proportional to the energy difference between the acceptor LUMO level and the donor HOMO level. (Figure reused with permission from Ref. [22] ). The loss of V OC assumed in the above Formula (2.1) with 0.3 V was explained to originate from the recombination-driven decrease of difference between the quasi Fermi level of electrons of the acceptor material and the respective level of the holes of the donor material [52, 204]. Due to the use of two different semiconducting materials, a decrease of the polymer bandgap - for increasing the portion of the solar spectrum, which can be absorbed (described in the following section) - does not lead to the increase of the heterojunction solar cell PCE (see Fig. 2.3). With decreasing E g,donor, while still maintaining the 0.3eV LUMO, the V OC of the solar cell would be reduced for too low donor bandgaps. However, for different acceptor materials the optimal matching LUMO level of the donor materials can be calculated. For example Scharber et al. [44] calculated the highest possible PCE to exceed 10% for a heterojunction solar cell based on PCBM as acceptor. For hybrid polymer/inorganic semiconductor solar cells Xu and Qiao [46] made similar calculations for TiO 2, ZnO and CdSe NCs as acceptor materials. They calculated optimal PCE values of up to 12 % by using semiconductor NCs out of these materials as acceptor.

27 13 2. Background Semiconducting polymers Polymers are known as insulators, which is true for polymers with saturated compounds, forming single bonds along the carbon chain, like it is the case in polyethylene (see Fig. 2.4, left). But by use of alternating single and double bonds along a carbon chain, i.e. in case of polyacetylene (see Fig. 2.4, right), charges could propagate by flipping of the double and single bonds along the carbon chain. By removing or adding at least one electron to the polyacetylene chain (which can happen by photoexcitation or doping) the conductivity of the polymer can be controlled from semiconducting towards metallic behavior. For the discovery of the conductivity of polymers described in the mid 1970 s, A. J. Heeger, A. MacDiarmid, and H. Shirakawa received the Nobel Prize in Chemistry in 2000 [47]. In the following Figure 2.4 the difference between an electrically insulating and a semiconducting polymer is described. Figure 2.4. Left: Structural formula of the polyethylene monomer, and schematic of the valence electron orbitals (four sp3 orbitals) together with their energetic positions and bond types. Right: Structural formula of the polyacetylene monomer, and schematic of the valence electron orbitals (three sp2 orbitals and one 2p z orbital) together with their energetic positions and bond types. All four populated valence orbitals of the carbon atom are filled in the molecule with a second electron from two neighboring carbon and two hydrogen atoms in case of polyethylene, and with electrons from two neighboring carbon and one hydrogen atom in case of polyacetylene.

28 2. Background 14 As one can see from Figure 2.4 (left), the valence carbon electrons in the polyethylene are equally distributed along 4 sp3 hybridized molecular orbitals, and are covalently bound either to hydrogen (2 bonds) or to carbon (2 bonds). In contrast, for polyacetylene (see Fig. 2.4, right) each carbon atom shows 3 sp2 hybridized orbitals (forming covalent σ bonds), and one 2p z orbital which forms one π-bond to one neighboring carbon atom [48]. The π-bond thereby exhibits an energy lower than the σ sp2 orbitals. However, also an orbital for the antibonding π*-bond of the 2p z orbital is created on a higher energetic level (the LUMO level). The energy of a photon may be sufficient (depending on its wavelength) to excite an electron by absorption of the photon energy into the π*-2p z orbital, thus opening the π- bond. However, due to the remaining σ bond between the carbon atoms the polymer still remains intact, while the excited electron in case of having an external driving force [49] can further propagate along the carbon chain. Besides this simple example of a semiconducting conjugated polymer, plenty further semiconducting polymers were developed [50]. With addition of elements (e.g. the substitution of carbon with sulfur) or variation of the conjugated organic polymer structure the energy levels of HOMO and LUMO levels can be tuned, as well as their conductivity. Also by addition of organic side-chains the solubility in organic solvents of the polymers is enhanced. In Table 2.1 examples of conjugated polymer donor materials utilized in hybrid solar cells with respective HOMO-LUMO levels, and hole mobilities are summarized.

29 15 2. Background Table 2.1. Conjugated polymers utilized or mentioned in this thesis with respective HOMO- LUMO levels, exciton diffusion lengths and hole mobilities extracted from literature, sorted by increasing bandgap size. Structural Formula Short name Full name LUMO/HOMO [ev] Hole mobility [cm 2 V -1 s -1 ] PCPDT-BT poly[2,6-(4,4-bis-(2- ethylhexyl)-4hcyclopenta[2,1-b;3,4-b +dithiophene)-alt-4,7-(2,1, / -5.3 [51] [51, 52] benzothiadiazole)] poly[4,4-bis(2-ethylhexyl)- PCPDT-TBTT 4H-cyclopenta[2,1-b:3,4- b ]dithiophene-2,6-diyl-alt- 4,7-bis(2-thienyl)-2,1, / [53] benzothiadiazole-5,5 -diyl] PSiF-DBT poly[2,1,3-benzothiadiazole- 4,7-diyl-2,5- thiophenediyl(9,9-dioctyl- 9H-9-silafluorene-2,7-diyl) / [54] [54] 2,5-thiophenediyl] P3HT poly(3-hexylthiophene-2,5- diyl) -3.2/ -5.2 [55] [56] MEH-PPV poly(2-methoxy-5-(2- ethylhexoxy)-1,4-phenylene vinylene) -2.8/ -5.0 [57] [58] PEDOT:PSS poly(3,4- ethylenedioxythiophene): poly(styrenesulfonic acid) -2.2/ -5.2 [59] 20 [60] from measurements performed by Seyfullah Yilmaz, University of Wuppertal.

30 2. Background 16 The development and utilization of low bandgap polymer donor materials has been one of the main reasons for substantial performance improvement of organic photovoltaic (OPV). With decreasing polymer bandgap the absorption of a higher portion of the sun spectrum becomes possible, as additional low energetic photons can be absorbed. This enables a higher current generation in case that the same fraction of absorbed photons is converted into electrons and extracted from the cell. In Figure 2.5 the improved light absorption from the low bandgap polymer PCPDT-BT over MEH-PPV compared to the AM 1.5G sun spectrum is illustrated. Figure 2.5. Absorption spectrum of MEH-PPV and PCPDT-BT compared to the AM1.5G photon flux (orange graph). The absorption of the two polymers is given in arbitrary units. The x-axis is given in logarithmic scale. However, from a polymer monolayer (in contrast to inorganic semiconductors) the photovoltaic PCE is very low, due to low relative permittivity of the polymer, leading to a high recombination between hole and electron (geminate recombination) due to a coulomb attraction of about 0.4 ev [61], depending on the polymer. The thereby created electron-holepair due to excitation is therefore regarded as a quasiparticle and called exciton. Precisely, excitons of conjugated polymers but also of small semiconducting nanocrystals (i.e diameter smaller than ca. 5 nm for CdSe [62] ) - are called Frenkel-excitons due to their high exciton binding energy. By adding a second semiconducting material with a LUMO level of a lower energy, a driving force for the dissociation of the exciton is introduced forming a heterojunction. This second material is called acceptor material, as it accepts the excited electron from the polymer (the donor material). By extracting the electron from the donor material, the electron-hole recombination is further reduced by providing separate materials for electron and hole extraction. As acceptor material at first a second polymer was

31 17 2. Background used [29, 63, 64]. Later, more efficient electron extraction was achieved by using a blend of semiconducting polymer and the C 60 Buckminsterfullerene [65, 66] derivative PCBM (Phenyl- C61-butyric acid methyl ester) [67]. However, in this thesis semiconducting nanocrystals are used as acceptor material, which were introduced for heterojunction solar cells in 1996 [14] Semiconducting nanocrystals In the nanoscale, materials undergo changes in their fundamental properties as function of their size. In particular, semiconductors show the remarkable effect of increasing bandgap by decreasing their size from the macroscopic bulk crystal below the Bohr radius down to nanocrystals. However, these NCs still have the same crystal lattice structure as the bulk material [68], and are consisting out of thousands down to several hundreds of atoms [69, 70]. In the early 1980s the first nanocrystals for which a larger bandgap compared to bulk materials was observed were nanoscopic II-VI crystallites (i.e. CdS [71 74] and ZnS [75] ). Louis E. Brus [68] used for the illustration of this so-called quantum size effect first described for nanocrystals by Alexey I. Ekivov in 1981 [76] - the example of the conjugated benzene ring, which has discrete molecular orbitals (MO) for the delocalized π-electrons. As more benzene rings are joined together - i.e. in case of pentacene - the possibilities for delocalization of electrons are increasing, with the HOMO level increasing and the LUMO decreasing its energy. Thus, leading to a reduction of the bandgap, which is observable by a change in the optical absorption of the material; in the special case of graphene even leading to a bandgap of 0 ev. More in depth, Armand P. Alivisatos applied in an overview article the particle in a box model to describe the physical cause for the quantum confinement in nanocrystals by the relation between the position and momentum in free and in confined particles [70]. While for a free particle (electron or hole) - like it is the case inside the bulk material - the position cannot be defined precisely, its energy and momentum can be defined. In contrast, for a spatially confined particle, its position can be defined more and more precise with higher confinement, but its momentum cannot - according to the Heisenberg principle - even though its energy might still be known. The wavelengths of the electron and holes are more and more spatially confined with decreasing size on the nanoscale, resulting in an increased

32 2. Background 18 coulomb energy of the electron hole pair compared to bulk material of the same composition and lattice structure, where electrons and holes can often move freely within the semiconductor bulk crystals. This resulting increased electron-hole excitonic binding energy is mainly responsible for the bandgap widening of NCs compared to bulk material, affecting the absorption and luminescence properties significantly. Thus, due to their molecule like discretization of the energy levels, for nanocrystals instead of using the terms valence band (VB) and conduction band (CB) the terms HOMO and LUMO are used. In a first simple effective mass approximation the relation between the bandgap energy of the bulk material E g,bulk and the bandgap energy E g,nc of a spherical nanocrystal with a radius r is given by following formula [77] : E g,nc (r) = E g,bulk + h 2 8r 2 1 m e + 1 m h (2.2),while h represents the Planck constant, r is the particle radius and m * e and m * h are the effective masses of the electron and the hole, respectively which depend on the semiconducting material. More detailed theories and approximations exist taking into account additional effects such as coulomb interactions between electron and hole [77, 68], influencing effects by the crystal field [78] or the deviation from particle sphericity [79]. For the calculated energy levels E HOMO (r) and E LUMO (r) of the NCs with a certain size and radius r following formulas developed by Brus [68], using the effective mass approximation and additionally considering Coulomb interaction of the electron-hole pair, have been applied: E HOMO r = E VB h2 8m h r 2 1.8e2 4πrε r ε 0 (2.3) E LUMO r = E CB + h2 8m e r 2 (2.4),while E VB and E CB are the energy levels of the bulk material respective to vacuum level, h being the Planck constant, m h the effective hole mass, m e the effective electron mass, e the elementary electric charge, ε r the relative permittivity and ε 0 the vacuum permittivity. An example of the change of HOMO and LUMO levels by NC size is given in Figure 2.6 for CdSe NCs. For the calculations an effective hole mass of m* h =0.45 m e [80], an effective electron mass of m* e =0.13 m e [80], a bulk bandgap of 1.75 ev [81], and a dielectric constant of

33 19 2. Background ε r =4.86 [82], and a valence band position at -5.6 ev [83] were assumed. The figure shows the increase of the bandgap size with decreasing NC diameter, while the HOMO level is changing less than the LUMO level due to the higher m* h compared to m* e. Figure 2.6. Diagram of the development of the CdSe bandgap from the bulk value to quantum dots (QDs) with decreasing particle size from right to the left. For NCs the energy levels split into discrete levels while the bandgap increases with decreasing particle size. The values for the HOMO and LUMO energy levels of CdSe QDs of diameters of 3 nm, 5 nm and 7 nm are calculated using the effective mass approximation model. The size dependency of HOMO, LUMO energy levels, and of the bandgap energy of TiO 2, which is also used in this PhD thesis, from its crystallite size is given in Figure 2.7. Therefore, m* h =0.8 m [84] e, m* e =9 m [85] e, a bandgap of 3.2 ev [86], and ε r =34 [87], were assumed for anatase TiO 2. Also, a bulk valence band energy of -7.4 ev [88, 89] was taken for the calculation. Figure 2.7. Left: Calculated dependency of the bandgap of anatase TiO 2 from crystal size. Right: Calculated dependency of HOMO and LUMO values of TiO 2 from crystal size.

34 2. Background 20 The experimentally extracted HOMO and LUMO values can vary between different applied measurement methods. There are several experimental methods for determining the absolute values of the HOMO and LUMO levels, such as cyclic voltammetry (CV) [90, 91], ultraviolet photoelectron spectroscopy (UPS) [92] and X-ray absorption spectroscopy (XAS) [93]. This methods are often combined with UV-Vis spectroscopy, with which optical bandgaps are determined, by adding this value the HOMO level - determined by for example CV also the LUMO value can be obtained [94]. In Table 2.2 examples of measured LUMO-HOMO energy levels of different NCs are summarized and compared to bulk and calculated values based on the effective mass approximation with Formulas (2.2), (2.3), and (2.4). Table 2.2. Measured and calculated energy values for selected bulk and NC type semiconductors utilized or mentioned in the thesis. HOMO- and LUMO levels of the NCs are calculated based on reported bulk values, effective electron and hole mobilities of the bulk material and the relative permittivities following the formula for the effective mass approximation from Brus [69]. HOMO and LUMO levels of PC 60 BM and PC 70 BM are also added to this overview for comparison. Material Bulk CB / VB [ev] NC Diameter [nm] Measured LUMO / HOMO [ev] Calculated LUMO / HOMO [ev] CdSe / -5.6 [83] / [95] / CdS / [96] / -6.6 [10] / CdTe / [96] / [97] / ZnO -4.4 / -7.9 [98] / [99] / CIS (CuInS 2 ) -4.1 / -5.6 [100] / -6.0 [100] / TiO / -7.4 [101] / -6.9 [102] / PbSe / [94] / [103] / PbS / [94] / [104] / PC 60 BM / [105] PC 70 BM / [105]

35 21 2. Background 2.4 Synthesis of colloidal semiconducting nanocrystals Since the semiconducting polymers used in this thesis were either obtained within collaborations from the Polymer-Electronics group of Prof. Dr. Ullrich Scherf from the University of Wuppertal (Germany), or obtained from commercial suppliers (i.e. 1-Material, and Sigma-Aldrich), the synthesis of these polymers is not a subject of this thesis. However, the utilized semiconducting NCs were synthesized by myself, largely according to a protocol elaborated by Yuan, Riehle et al. [19] and with constant input of these two authors. The nowadays most common method for colloidal synthesis of cadmium chalcogenide nanocrystals is the wet-chemical hot injection synthesis in organic solvents, described by Murray et al. in 1993 [106]. Therein, both organo-metallic cadmium and chalcogenide precursors are injected at a high temperature of 300 C into a high-boiling point organic coordinating solvent. This solvent leads as NC surfactant (also called ligand) stabilizing the NC surface from decomposition during the synthesis. This method improved the monodispersity and NC surface quality of the prior published non-injection (one-pot) synthesis [107], from which one can nowadays however also obtain high quality semiconducting nanocrystals [108, 109, 16]. For understanding the principle advantage of the hot injection method, in the following some theoretical aspects of the bottom-up concept of colloidal NC formation from monomer building blocks are shortly elaborated in the following. LaMer et al. [110] describes the NC formation by means of the monomer concentration. In the beginning of the NC synthesis process the monomer concentration is increasing by thermal decomposition of the precursors. After reaching a critical concentration nuclei are formed by the monomers. By a further formation of nuclei and the attachment of monomers to existing nuclei leading to the growth of NCs the monomer concentration is decreasing again under the critical concentration under which no more nuclei are formed. At that stage the nucleation phase is finished, however the growth phase has already begun. After all precursors decomposed, there are no new monomers available and the monomer concentration is stable, determined by the dissociation rates of monomers from the NC surface into the solution and from the solution onto the NC surface. During this last phase the NC surface in annealed by the described de-attachment and attachment process of monomers from and onto the NC surface. This phase can however be followed by an

36 2. Background 22 Ostwald ripening regime, during which larger NCs are growing by incorporating material from smaller NCs. In Figure 2.8 the prior described colloidal NC synthesis is visually depicted. Figure 2.8. Schematic of the occurring processes during a wet-chemical colloidal NC synthesis from the starting point at which the precursor decomposition creates a critical monomer concentration over the formation of nuclei and NC growth towards NC size focusing and a possibly occurring final size defocusing process (Ostwald ripening). According to the prior description, a rapid decomposition of the precursors into monomers would induce a short nucleation phase, as a large number of nuclei could be formed at a similar time point. The subsequently starting growth phase would then start from similar NC seed sizes, leading to a narrow size distribution in the growth phase. This fast precursor decomposition is achieved by the injection of precursors in to a hot coordinating solvent. The method of injecting only one precursor, while the other is already inside the heated surfactant is further accelerating the nucleation process, as a monomer from the decomposed precursor of the injected species more likely meets is metal-atom counterpart, which is already released from the pre-decomposed precursor. However, also this hot injection method cannot completely separate the nucleation phase from the growth phase, thus still leading to a certain NC size distribution. 2.5 Post-synthetic nanocrystal treatment The as synthesized NCs cannot be used for highly efficient hybrid solar cells, as their ligand sphere is acting as insulator, hindering charge transfer to the NC and charge transport over

37 23 2. Background the NCs [13]. Thus, the long alkyl chain containing synthesis ligands of NCs were removed by ligand exchange with small molecules such a pyridine prior to the integration of NCs into the hybrid polymer-nc blend, leading to substantial progress of NC based hybrid solar cells [111]. This pyridine treatment is believed to replace the insulating synthesis ligand with shorter and more conductive pyridine molecules and is up to date widely used. However, NCs tend to aggregate after ligand exchange by small molecules and precipitate, making it challenging to obtain stable mixtures of NCs and polymer [112]. Therefore, a method allowing for a controlled reduction of the NC ligands was developed in our group [15]. This post-synthetic treatment is based on the assumption of a protonation of the NC ligand by hexanoic acid (HA). Figure 2.9 shows this procedure in case of the hexadecylamine (HDA) ligand, to which the protonation of the NH 2 group introduces a strongly polar group, thus allowing for the subsequent solvation of HDA by a polar solvent, like i.e. methanol. Figure 2.9. Description of the dissolution of the NC synthesis ligand hexadecylamine (HDA) by methanol, favored by the protonation with hexanoic acid (HA): HA transfers a proton to HDA, thus introducing a positive charge onto HDA; subsequently the polar solvent methanol can dissolve the resulting salt consisting out of the protonized positively charged HDA and the negatively charged HA ions. Based on a model described by F.-S. Riehle [16], the synthesis ligands form a sphere around the NCs, which is not only consisting out of one monolayer of synthesis ligands around the NC surface; but also out of additional ligand molecules, which are coordinated around the initial ligand layer forming a ligand shell. Thus, around the NC, there is first the inner core of

38 2. Background 24 the ligand shell, i.e. the ligands directly bound to the NC surface, followed by an outer shell of loosely bound ligand molecules forming the outer shell of the ligand sphere. The existence of this ligand sphere was first visualized by transmission electron microscopy (TEM), i.e. by superposition of the dark and bright field image [15], and later proved by thermo gravimetric analysis-mass spectroscopy (TGA-MS) [113]. The later study divided the ligands within the ligand sphere according to thermo gravimetric measurements into an inner layer, strongly bound to the NC surface, a surrounding second ligand layer, and a large third layer of weakly associated ligands forming an onion-like structure. Within this thesis the approximate hydrodynamic diameter of the ligand shell was determined by dynamic light scattering (DLS). Also, the decrease of photoluminescence (PL) intensity of the NCs was used as qualitative feedback of the ligand sphere reduction, since the PL correlates with the NC surface quality. More exactly, ligands are passivating dangling bonds on the NC surface, thus removing energetic states within the NC bandgap (deep traps) which lead to non-radiative exciton recombination [114]. Hence, a strong reduction of the ligand sphere can lead to the induction of surface traps, and additionally to the reduction of the NC dispersivity in common solvents, representing the drawback of this method. In consequence, one has to carefully adjust the post-synthetic NCs treatment in such a way that the ligand sphere is sufficiently reduced to allow an efficient charge transport, but on the other hand NC dispersivity has to be maintained and the number of potentially induced NC surface traps must be minimized. It should be also mentioned here that the HA based ligand sphere reduction treatment was originally developed for the protonation of the amine group of hexadecylamine of HDA/TOPO capped CdSe QDs. However, the procedure proved also to be applicable for oleic acid (OA)/trioctylphosphine (TOP) capped CdSe QDs, and for OA/tetradecylphosphonic acid capped NRs [115], which is presented within Chapter 5.1 of this thesis. Additionally, the treatment is also applicable on hexadecanol (HDO) capped NCs, which is shown in Chapter 4.1 of this thesis. Moreover, In the Appendix A4 an infrared spectroscopy study is presented, which is leading to the indication of a strong reduction of the initial QD ligands and the incorporation of HA molecules on the CdSe QD surface after treatment with HA.

39 25 2. Background 2.6 Historical background and state of the art Within this section the evolution of hybrid heterojunction nanocrystal-polymer solar cells is presented, as well as approaches to control the nanomorphology of the solar cell towards enhancing the charge extraction. Finally, also an overview over current research on the degradation of hybrid solar cells is presented Evolution of hybrid bulk heterojunction solar cells Towards the first hybrid BHJ solar cell: After already in the late 1950s the photovoltaic effect of conjugated organic molecules (i.e. anthracene) had been discovered [116], organic polymer solar cells consisting out of a single layer of conjugated polymer had been subject of constant improvement [117], reached PCEs of up to 0.7% in the late 1970s [118]. In 1986 a major improvement for organic solar cells was achieved by Tang et al. [29] using a bilayer of two different organic polymers, one (donor material) for the hole extraction and a distinctive layer (acceptor material) for the electron extraction after exciton dissociation at their interface and reduction of electron-hole recombination during the charge extraction, thus creating the first heterojunction organic solar cell. A further breakthrough was achieved in 1995, by blending of two polymers, thereby increasing the interfacial area between the donor and the acceptor material [63, 64], whereby the first bulk heterojunction solar cell was created. By using the C 60 Buckminsterfullerene [65, 66] derivative PCBM [67] as electron acceptor material in the blend [39], in the same year the efficiency of organic solar cells was already enhanced to 2.9% (under 0.2 sun). One year later, in 1996 the first hybrid bulk heterojunction solar cell using semiconductor NCs as acceptor material was reported by Greenham et al. [14], using a blend of CdSe QDs and MEH-PPV. In the first reported hybrid solar cell PCEs were very low of about 0.1%. At a high concentration of NCs of around 90% by weight (wt%) external quantum efficiencies (EQEs) up to 10% were already achieved indicating a reasonable charge separation taking place at the polymer/nc interface.

40 2. Background 26 Efficiency improvement by change of NC shape and ligand exchange: One of the main challenges was already realized: the inefficient electron transport between individual NCs [14]. This challenge was addressed in the following years by pursuing strategies to overcome these limitations, including ligand-exchange of NC synthesis ligands with pyridine (see Chapter 2.5), whereby mainly NCs out of CdSe as best established NC acceptor material were used. In 2002 Huynh at al. presented blends of elongated CdSe nanorods (NRs) and regioregular P3HT as light absorbing layer reaching efficiencies up to 1.7% [119]. With increasing NR length, improved electron transport properties were demonstrated resulting in an increase of hybrid solar cell performance. One year later Sun et al. [111] demonstrated in CdSe/OC 1 C 10 -PPV devices an improved performance of hybrid solar cells incorporating CdSe tetrapods (TPs), reaching a PCE of 1.8%. He attributed this enhancement to an improved electron transport perpendicular to the active layer, since TPs always have an extension towards the cathode, thus facilitating electron extraction. Later on, numerous works based on pyridine treated CdSe-P3HT hybrid solar cells were reported, where extended NC structures led to increased device efficiencies reaching up to 2.6% for CdSe NR-P3HT based solar cells in 2006 [120]. The later result was realized by using a solvent with a high boiling point (i.e. trichlorobenzene) during the deposition and drying of the active layer, which was generally found to lead to devices with better performance. By combining post-synthetic washing of CdSe NRs and pyridine ligand exchange improved NR-PCPDTBT hybrid solar cells with PCEs approaching 3.5% were reported by Celik et al. [113]. The reduction of a ligand shell around the NRs depending on the post-synthetic washing procedure was detected by a combination of thermal gravimetric analysis coupled with a mass spectroscopy (TGA-MS) and optimized post-synthetic treatment and ligand exchange protocols could be elaborated for the NCs for integration into high efficient hybrid solar cells. However, the highest PCE for BHJ NC/polymer solar cells of 4.7% was recently reached by R. Zhou et al. [11] by combining the advantage of long CdSe NRs with a post-synthetic ethanedithiol (EDT)/acetonitril (ACN) treatment of the photoactive layer in addition to a prior performed pyridine ligand exchange on the NCs. Thereby EDT is binding to the NC surface, passivates it and can interlink the NCs due to the two existing thiol groups in EDT, leading to a doubling of the electron mobility within respective solar cell devices.

41 27 2. Background Improvement of QD/polymer solar cells: Up to recent years, the utilization of spherical QDs for hybrid solar cells did not progress very much. In 2006 Han et al. reported a PCE of 1.1% for CdSe QD-P3HT based hybrid solar cells, by using pyridine ligand exchange [121]. Another improvement for CdSe QD-P3HT systems was reported 2009 by Olson et al. reaching PCEs up to 1.77% [122]. Here, butylamine instead of pyridine was used as a shorter capping ligand, subsequent to a pyridine ligand exchange procedure, before integration of the QDs into the photoactive hybrid film. On year later, Zhou et al. introduced a novel post-synthetic surface modification method, where NCs were treated with a simple and fast hexanoic acid treatment leading to a significant reduction of the ligand sphere derived from colloidal synthesis [15]. Devices of hybrid solar cells with hexanoic acid treated spherical CdSe QDs and P3HT exposed PCE values reaching up to 2.1% [123]. A significant step towards high efficient hybrid solar cells was the utilization of lower bandgap donor polymers, which were developed for pure OPV. Solar cells utilizing the low bandgap polymer PCPDT-BT with an optical bandgap of about 1.4 ev and a high hole mobility of up to cm 2 V -1 s -1[52], combined with CdSe TPs reached PCEs of up to 3.13% [124]. Similarly, using the BCPDTBT polymer with spherical CdSe QDs, led to hybrid solar cells with a PCE of 2.7% [125] and reaching up to 3.1% by using larger CdSe QDs [115]. Furthermore, the concept of using an additional hole blocking layer (also serving as optical spacer [126] ) out of ZnO NCs in between the active layer and the top metal contact is reported by R. Zhou et al. [127] to improve the PCE of CdSe QD/PCPDTBT solar cells to 3.7%. A similar PCE of 3.78% was reached by Seo et al. [128] utilizing a TiO 2 hole blocking layer and additionally utilizing low bandgap NCs out of PbS. However, the highest PCE of BHJ QD/polymer solar cells was reached by Ren et al. [10], with a solar cell which used P3HT nanofibers decorated with CdS QDs. Thereby, the exciton separation was improved by the intimate contact between QDs and polymer and also by providing a direct pathway for the hole extraction. Moreover, the organic hole blocking layer bathocuproine was introduced to reach PCEs of up to 4.1%.

42 2. Background 28 NC/polymer solar cells using QD/NR mixtures: Together with Yunfei Zhou, I demonstrated that a mixture of hexanoic acid treated QDs and NRs integrated into the hybrid film lead to relatively better device performances than QD or NR only based devices [115] (see Chapter 4). The QD-NR mixture within the PCPDT-BT films seems improve the interconnectivity of the inorganic network, since in NR only based devices the NRs align predominately in horizontal position [129]. This benefit of a pyridine treated QD-NR mixture integrated into PCPDT-BT was additionally confirmed by Jeltsch et al. [130], leading to devices with PCEs up to 3.5%. The following Table 2.3 summarizes the development of hybrid bulk heterojunction NCpolymer solar cells, based on selected results for CdSe NC-polymer devices. Table 2.3. Evolution of hybrid bulk heterojunction CdSe NC-polymer solar cells, illustrated on selected solar cells. NC shape Pyridine treatment Polymer Solvent(s) Max. EQE (monochromatic) [%] PCE (under AM 1.5G) [%] Year QD no MEH- PPV CHCl [14] QD NR not mentioned P3HT CHCl 3 +pyridine 19* 55*? [119] NR TP 24 h OC1C10- PPV CHCl 3 +pyridine 23 $ 45 $? [111] TP 24 h OC1C10- PPV CHCl 3 TCB / [131] CHCl # NR 24 h P3HT thiophene TCB # 2.6/2.9 # 2006 [120] QD ca. 12 h P3HT CB +pyridine [121]

43 29 2. Background QD CHCl NR ca. 12 h P3HT +pyridine [132] TP +TCB TP ca. 12 h PCPDTBT QD NR QD/NR mixture NR no (hexanoic acid) pre washing + ca. 12 h pyridine pre washing + 8 h pyridine 24h + postdeposition treatment: EDT/ACN P3HT PCPDTBT CHCl 3 +pyridine +TCB CB + DCB CB + TCB [124] [125] PCPDTBT CB + TCB [113] PCPDTBT CB [130] PCPDTBT CHCl [11] *under mw/cm², $ under 0.39 mw/cm², + under 66 mw/cm², $ average PCE, # no spectral mismatch, + under 0.67 mw/cm² Hybrid nanostructured heterojunction solar cells This subchapter is discussing concepts and state of the art of hybrid solar cells with an active layer of an interdigitated donor-acceptor design, theoretically representing the most promising design as discussed in Chapter 2.2. More precise, the creation and further perspective of usage of nanostructures, acting as scaffold or as acceptor material for such nanostructured solar cells is discussed in this thesis in Chapters 5.3, 6.3, A6, and A7. For creating these solar cells, two approaches were followed. First, a thin layer of regular, vertically aligned nanopores of aluminum oxide (Al 2 O 3 ) was created by electrochemical anodization from a thin film of aluminum, deposited on an ITO coated glass substrate. And secondly, an array of regular, vertically aligned TiO 2 nanotubes was also created by the same method on ITO.

44 2. Background 30 The first approach using anodic aluminum oxidation can lead to nanostructured solar cells by filling of the alumina pores by a semiconducting polymer (or with CdSe NCs), the alumina would then be removed by KOH etching and the created voids could be filled with CdSe NCs (or with polymer) to form an interdigitated donor-acceptor structure (see Fig. 2.10). Figure Approach towards a nanostructured interdigitated heterojunction solar cell by using a template of vertically aligned alumina pores as scaffold. (Figure reused with permission from Ref. [22] ). A second concept based on anodic aluminum oxidation (AAO) is to coat the alumina pores by a thin TiO 2 layer by atomic layer deposition (ALD), to form the electron acceptor. The pores would then be filled by a semiconducting polymer to form a heterojunction solar cell. However, during this thesis the second concept was already implemented by Lee et al. [133], with a PCE of 0.5 %. And for the first concept, a similar procedure was published, using the AAO template by filling with TiO 2 NCs from solution, followed by removal of alumina and covering of the free standing TiO 2 nanopillars with P3HT, resulting in a PCE of 0.5 % as well [101]. And even 0.6 % PCE were reached by obtaining free standing sol-gel process created TiO 2 nanopillars molded from an AAO template, subsequently covered with P3HT [134]. An AAO template was also used to create P3HT nanopillars by nanoimprint lithography (NIL) that are subsequently filled with a different semiconducting polymer to reach an efficiency of 1.9% [135]. However, the second followed approach (see Fig. 2.11) is a more direct one. Therefore, a thin layer of titanium is deposited on an ITO coated glass substrate which is transformed into TiO 2 nanotubes (NTs) by anodization. The created TIO 2 NTs would act as electron acceptor

45 31 2. Background and can subsequently be filled with a semiconducting polymer or with a polymer/semiconductor NC mixture to form a heterojunction solar cell. Figure Approach towards a nanostructured interdigitated heterojunction solar cell by using a TiO 2 nanotube array as electron acceptor, filled with conjugated polymer to form a donor-acceptor heterojunction. The concept of creating nanostructures for solar cells directly on ITO is very appealing, there are very few publications dealing with the creation of AAO on ITO [136, 137, 137] compared to the overall number of publications of AAO formation. And there are even fewer reports about the creation of ATO on ITO [138, 139] or on fluorine-doped tin oxide (FTO) [140]. Nevertheless, for dye sensitized solar cells (DSSCs) PCEs of up to 6.9% have been reached with 2 µm high TiO 2 nanotube structures, and for solid state DSSCs using the semiconducting polymer P3HT instead of the liquid electrolyte the PCE of 3.8% was reached [141]. However, such efficiencies could by far not been reached by simple TiO 2 nanotube/polymer solar cells. For an efficient charge transfer between the semiconducting polymer and the metal oxide an interlayer is needed, passivating the metal oxide surface and significantly improving the charge transfer. Therefore, for example the ruthenium dye N719 or pyridine [142], 4-mercaptopyridine [143], Sb 2 S [144] 3, or Sb 2 S 3 together with decyl-phosphonic acid (DPA) [145] have been used Degradation of hybrid heterojunction solar cells Sources of solar cell degradation: In general, there are several sources causing the degradation of BHJ hybrid solar cells during its operation, whose influence on the solar cell need to be studied. However, most studies on this subject are covering purely organic solar cells [146], some are dealing with all-inorganic

46 2. Background 32 QD solar cells, and only very few are reporting on the degradation of hybrid BHJ solar cells. However, since hybrid BHJ solar cells have many components in common with all-organic and all-inorganic solar cells, some findings on the later two solar cell types are also applicable on hybrid solar cells. One source of degradation is the UV-light being a part of the solar spectrum, causing degradation of the chemical structure of the conjugated polymer [147, 148], leading by time to increased charge recombination and lower charge carrier mobility. These defects introduced into the conjugated polymer by photo-oxidation are however reported to be at least partially reversible by the charge carrier flux through the device [149], unless no polymer chain scissoring has occurred as reported upon formation of photo-induced singlet oxygen formation [150]. Also, oxygen present in air is causing the degradation of several components of hybrid solar cells. Such is the decrease of charge carrier mobility of conjugated polymer upon O 2 uptake [ ], which is explained by oxygen inclusion within the thiophene ring of the conjugated polymer leading to an inhibition of charge transport [153] and by inclusion of O 2 into the side and main chain of the polymer [151]. At least the later chain-inclusion of oxygen is reported to be reversible by applying heat for de-trapping [151]. Oxygen is further decreasing the solar cell performance by causing a reduced electron injection into the metal top electrode from the active layer by formation of Al 2 O 3 at the Al/active layer interface. Oxygen is believed to mainly enter the solar cell through grain boundaries or microscopic pinholes of the evaporated aluminum top electrode [154]. For all inorganic NC solar cells it is reported that the use of aluminum as top electrode is causing 5 times faster PCE decrease in air compared to silver due to its fast oxidation leading to a fast increase rate of the solar cells series resistance [155]. This performance decrease could however be greatly slowed down (by a factor of 16) upon utilization of a LiF interlayer between the active layer and the aluminum electrode. Moreover, oxygen is also causing oxidation of the NCs utilized in the solar cell. For all-inorganic NC solar cells the oxidation of NCs reportedly represents a drawback [156], leading to creation of trap states inside the metal chalcogenide NC bandgap, thus diminishing the charge extraction efficiency [157]. This oxidation is reported to be strongly reduced for PbS QD sensitized (QD decorated TiO 2 nanocrystals with P3HT as hole extraction layer) solar cells by passivation of the PbS NC surface with dodecanethiol [158] and for allinorganic PbS QD solar cells by passivation with trimethylphenyl-methyldithiocarbamate [157].

47 33 2. Background Furthermore, the use of smaller NCs is also reported to exhibit an improved stability in air over the use of large NCs in all-inorganic NC solar cells due to a relatively higher ligand density on the NC surface [156]. A third source of solar cell degradation is water, which is also present in the atmosphere. H 2 O is reported to fastly deteriorate the performance of organic solar cells, if a PEDOT:PSS layer is used as electron blocking layer [159]. The decrease is indentified to be caused by a decrease in the solar cells series resistance due to a supposed reduction of hole injection from the active layer to PEDOT:PSS. Thus, hole transport is hindered, and a hole injection towards ITO is reduced by reaction of PSS with H 2 O. Krebs et al. [5] measured a much faster decrease of the performance of a ITO/PEDOT:PSS/P3HT:PCBM/Al BHJ solar cell in presence of humid nitrogen than for dry oxygen, further supporting the thesis of the solar cell performance deterioration by water uptake inside the PEDOT:PSS layer. Moreover, water uptake into the PEDOT:PSS layer might accelerate the etching of the underlying ITO layer, which is reported to occur fastest during the spin coating of the acidic PEDOT:PSS solution on ITO [160]. Prevention of BHJ CdSe polymer solar cell degradation: Hybrid BHJ CdSe/polymer solar cells can be of cause prevented from degradation from environmental sources by encapsulation of the whole device. However, for improving the intrinsic stability of the solar cell the introduction of protective layers, or the avoidance of the use of materials especially susceptible to degradation is reported. Such approaches for the protection of the active layer and of the contact between the active layer and the charge extraction electrodes are realized by the use of metal-oxide interlayers and inverted solar cells. Namely, Zhou et al. [127] report in their publication about CdSe/PCPDTBT BHJ solar cells a dramatic increase of the solar cell lifetime in air when using a ZnO NC layer between the active layer and the aluminum top electrode. The PCE in air without the ZnO interlayer was reported to drop by over 90% within several hours. However, with the ZnO interlayer the PCE dropped much slower (mainly due to a decrease of shortcut current density (J SC ), accompanied by a FF decrease and a slight V OC increase) still exhibiting after 60 days about 60% of the initial PCE value. Furthermore, also a BHJ hybrid solar cell without use of the

48 2. Background 34 hygroscopic PEDOT:PSS and of the easy oxidizable aluminum layers is reported by Kwon et al. [161]. For the realization they utilize an inverted BHJ CdSe/P3HT design with a ZnO:Cs interlayer between ITO and the active layer, and a protective MoO 3 interlayer between the active layer and the silver top electrode. Thereby, they improve the performance stability of the solar cell in air from a PCE decrease after 1 day by 99% for the normal ITO/PEDOT:PSS/(P3HT:CdSe)/Al solar cell to a decrease of only 2% for the inverted solar cell. Investigations of hybrid BHJ CdSe polymer solar cell degradation: Reports of solar cell degradation without any use of metal-oxide like the ones used in this thesis - are scarce. Thus, within this thesis the degradation of hybrid BHJ CdSe/polymer solar cells in inert atmosphere (also under high temperature) and in air is observed, degradation causes specific for NC/polymer solar cells are detected, and strategies for improvement of the solar cell stability are elaborated. However, one report already exists from Yang et al. [162]. They reported on a CdSe/P3HT hybrid solar cell (of the same design as the solar cells used in this thesis) an increase to 1.5% PCE after 25 min in air from the initial 0.59%. The increase was visible by a rise in V OC by 60% to 0.68 V and a rise in J SC by about 50% to 4.7 ma/cm². Also, an improvement (lowering) of the solar cells series resistance (R S ) (not mentioned in the publication but visible from to the presented J-V measurement) was measured. The rise in V OC is accompanied by a decrease of the dark saturation current, a finding which is also reported in all-inorganic NC solar cells [163] and organic solar cells [164]. The improvement was explained by the adsorption of oxygen on the CdSe surface, passivating Se vacancies on the CdSe surface by binding to Cd atoms. Yang et al. further assume that the trapping and recombination on the NC surface might be reduced, leading to an increase in charge carrier mobility and a more efficient charge carrier collection. However, a further measurement 85 minutes later revealed a strong decrease to 0.45% PCE mainly due to a reduction of J SC and simultaneous increase of R S. This later effect might have the same reason as concluded by Krebs et al. [154] for purely organic solar cells, namely originating from the creation of an insulating aluminum oxide layer at the interface between the active layer and the top aluminum electrode.

49 35 3. Methods 3. Methods Within this chapter, at first characterization methods utilized for the characterization of nanomaterials are presented, which were mainly applied to CdSe NCs, but also to nanoporous TiO 2 and Al 2 O 3 structures and conjugated polymers. Subsequently, detailed descriptions of all fabrication processes needed for the solar cell manufacturing are given. Finally, utilized methods for determining relevant solar cell parameters are presented. 3.1 Nanomaterial characterization Photoluminescence (PL) spectroscopy Photoluminescence (PL) is the general term for photon emission from excited energetic states. After the excitation of an electron from its ground state to an excited state, the relaxation can follow different pathways during which photons are eventually emitted at different rates. Therefore, PL is further categorized into fluorescence and phosphorescence [165] (see Fig. 3.1). Figure 3.1. Jablonski diagram explaining the occurrence of fluorescence and phosphorescence after excitation of an electron from the singlet ground state (S 0 ) to the first excited singlet state (S 1 ). The above Figure 3.1 explains the processes leading to fluorescence and phosphorescence by a Jablonski diagram [166] to which the direction of the electron spin was added for illustration reasons. Fluorescence occurs from an electron, previously excited from the

50 3. Methods 36 singlet ground state (S 0 ) into the first excited singlet sate (S 1 ). Despite its excitation, the electron still forms an electron pair with the electron of opposite spin from the S 0 ground state. An excited electron typically remains in this S 1 state for a time of about 10 ns [167]. Afterwards the electron relaxes to S 0, thus emitting a photon of the wavelength ʎ with an energy E corresponding to the difference between S 1 and S 0 ; this fast photon emission is called fluorescence. The wavelength of the emitted photon can be calculated from the Planck relation by: ʎ = h c E eV nm E (3.1),with h being the Planck constant ( x10-15 ev s) and c being the speed of light ( x10 18 nm/s). While fluorescence occurs over relaxation from S 1 directly back to S 0, phosphorescence occurs after the electron undergoes an intersystem crossing to a triplet state (T 1 ), thereby switching its spin. Intersystem crossing is a process with a material dependent likelihood [168], therefore some materials favor phosphorescence over fluorescence and vice versa. Subsequently, the relaxation from T 1 back to the S 0 state again requires a spin change and is therefore unlikely to happen, thereby delaying the relaxation process. Hence, phosphorescence occurs in ms [167] or even up to several minutes [169]. In CdSe NCs the S 1 state is populated within 1-2 ps, while the fluorescence lifetime of the [170, 171] excited electrons is of about 5-20 ns (depending on NC size and fluorescence efficiency). In CdSe NCs relaxation from triplet states (causing phosphorescence) is not detectable by a photoluminescence spectrometer, since the energy loss is of only a few mev [172] and thus the phosphorescence spectrum is overlapped by the fluorescence spectrum. However, one would be able to distinguish between both emission mechanisms by transient measurements which were not performed within this thesis.. Thus, the photoluminescence spectra of the investigated CdSe NCs within this thesis will not be named fluorescence or phosphorescence spectra but PL spectra. In photoluminescence spectroscopy (the wavelength resolved recording of photoluminescence) the investigated sample is excited by a specific wavelength, which should be of sufficient energy to overcome the energetic difference between S 0 and S 1, i.e. the bandgap size of a semiconductor. This excitation signal is created from a lamp with a

51 Photoluminescence [CPS] Methods broad spectral photoemission (i.e. a xenon lamp) from which a single wavelength is selected by a monochromator. The PL spectrum emitted by the sample is then recorded in a 90 angle with respect to the excitation source. For PL spectra acquisition a PL spectrometer from J&M Analytik AG and the FluoroLog 3 PL spectrometer from Horiba Jobin Yvon were utilized. A sample of a PL spectrum recorded from a CdSe NC sample is presented in the following Figure M 0.5M PL peak position FWHM PL (peak) intensity Wavelength [nm] Figure 3.2. Example of a recorded PL spectrum of a CdSe NCs. Extractable are the PL peak position, the PL peak intensity, and the full width at half maximum (FWHM). As depicted in Figure 3.2, various parameters can be extracted from the measured PL signal of NCs. First, the NC size distribution can be roughly indicated from the full width at half maximum (FWHM). Since the PL signal is the accumulation of all PL emissions of the NCs inside the sample, the accumulative signal is not a discrete one because of a certain size distribution of the investigated NCs. Like mentioned in Chapter 2.3.2, the NC bandgap is dependent on the NC size. Therefore, a broader size distribution is also causing a broader PL spectrum around the peak position. Hence, the FWHM is an indication for the size distribution of the NC sample. Accordingly, the average NC diameter can be derived from the PL peak position. Moreover, from the PL peak intensity one can get some insight about the relative nanocrystal surface quality within the same ligand system [18]. An absolute measurement of the PL intensity is the PL quantum yield (explained on a practical example in Appendix A3). Furthermore, since the NC bandgap is temperature dependent [173], also the PL signal is temperature dependant. For PL spectra recorded in-situ at an elevated temperature during the microwave synthesis this had to be considered in order to compare spectra at different temperature (see Appendix A1). Also, it must be stated that the PL intensities

52 Absorption [A.U.] 3. Methods 38 presented throughout the thesis are not all intercomparable in between experimental series, since different PL spectrometers and different signal integration settings have been used. Determining absolute intercomparable PL quantum yield values takes either a great effort or specialized equipment. The PL spectra were measured inside Hellma fluorescence cuvettes out of Suprasil quartz glass, and were obtained from solutions of 0.1 absorbance units (measured at the 1 st excitonic peek) if not otherwise stated UV-Vis absorption spectroscopy According to the previous Section 3.1.1, the absorption is dependent on the energy difference between the ground state of the electron and its first excited state. According to Figure 3.1 the absorption peak energy (respectively absorption peak wavelength) of the lowest energy (respectively the highest wavelength) called the 1 st excitonic peak - corresponds to the sum of the bandgap energy and the energy invested in excitation of vibrational modes. The UV-Vis (ultraviolet-visible) absorption spectrum is recorded by a similar setup as mentioned before for the PL spectrometer. The difference is that the incident light is a series of single wavelengths - instead of a single wavelength, and that the detected light lies on the pathway of the incident light instead of a 90 angle. In absorption spectroscopy the change of intensity of the incident light is recorded after passing the sample (see Fig. 3.3, left) st excitonic peak Wavelength [nm] Figure 3.3. Left: Scheme of optical absorption spectrometry of a sample with the thickness d, the wavelength dependant absorption coefficient ε(ʎ) and the concentration c. The incident light of a given intensity I 0 (ʎ) at a series of wavelengths. The transmitted light intensity I(ʎ) for each utilized wavelength is observed by a photo-detector after the sample. Right: Example of a recorded UV-Vis absorption spectrum of CdSe NCs, together with the indication for the position of the 1 st excitonic peak.

53 39 3. Methods The absorption is thus obtained by comparing the intensity of light transmitted through the sample (I) with the intensity of the incident light (I 0 ). A monochromator is used select incident light I 0 (ʎ) of different wavelengths of the UV-Vis spectrum. The absorption spectrum is defined accordingly to the Beer Lambert law [174] written in dependence of the wavelength by: A(ʎ) = log 10 I ʎ I 0 ʎ = ε(ʎ) c d (3.2),with A(ʎ) being the wavelength dependent absorbance, I(ʎ) being the wavelength dependent intensity of the light transmitted through the sample, I 0 (ʎ) being reference spectrum of the incident light, ε(ʎ) being the wavelength dependent extinction coefficient of the sample, c being the sample concentration, and d being the distance that the light is passing through the sample. Moreover, it should be stated that the extinction coefficient and thus the absorption spectrum are also dependent on the measurement temperature; however the temperature dependency is specific for each material [175, 176]. One can convert from the absorbance given in Absorbance Units (A.U.) - which inversely logarithmically scales the transmittance - to transmittance by using the correlation between logarithm and exponentiation by: T ʎ = I ʎ I 0 ʎ = 10 A(ʎ) (3.3) For UV-Vis absorption spectroscopy two different TIDAS TM spectrometers from J&M Analytik AG were utilized. Experimentally, first a dark scan was done to record the signal of the photo-detector without illumination, then spectrum of the light intensity distribution over different wavelengths was recorded. Thus including the lamp spectrum, the absorption spectrum, and potentially also the emission spectrum of the solvent (i.e. chloroform inside a fused silica cuvette) used for the sample - in case of a solution or dispersion. This second scan is called the reference scan. Now, the sample was placed in the light path (e.g. CdSe NCs dispersed in chloroform inside a Hellma fluorescence cuvette out of Suprasil quartz

54 3. Methods 40 glass) and a third scan is performed to determine the absorption of the investigated material. An example of a UV-Vis absorption spectrum thereby recorded for CdSe NCs is given in Figure 3.3 (right) Fourier transform infrared spectroscopy (FTIR) Like in UV-Vis spectroscopy, by Fourier transform infrared spectroscopy (FTIR) the absorption of a sample over different frequencies is detected, just in the infrared wavelength range. For FT-IR detection, however an absorption spectrum is obtained by Fourier transformation of an interferogram obtained by one single excitation of the sample by polychromatic light [177]. Thereby, in contrast to the energy dispersive spectroscopy the spectral information can be fastly obtained by a single, short excitation signal, allowing for the increase of accuracy by the accumulation of multiple measurements within the time of a single dispersive measurement. FTIR is used to detect different functional groups of molecules, based on the characteristic absorbed energy. FTIR is based on the effect that atoms within a molecule absorb infrared irradiation based of an energy related to their specific bond energy within the molecule. Thereby, stretching (symmetric, asymmetric) or bending (scissoring, rocking, wagging, twisting) movement of the involved atoms is caused [178]. For illustration of the possible induced molecular vibrations [179], in the following Figure 3.4 the example of the methylene group (CH 2 ) is given, with carbon as reference atom. Figure 3.4. Possible molecular vibrations, illustrated on a methylene group, with the blue sphere representing the carbon atom, the red spheres representing the hydrogen atoms and the gray lines representing chemical bonds.

55 41 3. Methods The thus created features in the absorption spectrum, based on characteristic energies of the specific molecular vibration and the functional group or interatomic bond, can mostly be detected within the mid infrared [180] ( µm, or 4, cm- 1 ) region. They can be regarded as fingerprints of the functional chemical groups of the investigated sample. Nevertheless, some molecular vibrations are also induced by near IR and far IR (terahertz) irradiation, and molecular rotation is induced by microwave irradiation [180]. In FTIR, the absorbed energies are given in wavenumbers instead of wavelengths, describing the units of cycles per centimeter ( inverse centimeters, cm -1 ) [180]. Hence, the wavenumber represents the inverse of the wavelength, and thus has the advantage that a higher value represents a higher energy - see Planck relation of Formula (3.1). From a FTIR spectrum one can not only read energies at which the IR irradiation is absorbed, but also like in UV-Vis spectroscopy, which fraction of the specific energy is absorbed. Hence, one can draw from the transmittance also a conclusion of the quantity of absorbing material (although for quantitative values calibration curves need to be created first). The FTIR spectra presented in this thesis were recorded by a Nicolet/Thermo Magna-IR 760 spectrometer with a probe chamber using a diamond attenuated total reflection (ATR)-unit with a detection range of cm Dynamic light scattering (DLS) Briefly, dynamic light scattering (DLS) is calculating the hydrodynamic radius of particles inside a dispersion by detecting the time resolved intensity deriving from an incident laser beam, scattered in a certain angle (e.g. 90 ) by a particle, and by application of correlation models on the acquired signal developed by time, especially since the possibility of laser scattering in the 1960 s [181]. The size detection is based on the fact that particles within the sample undergo Brownian motion. Hence, the intensity of the scattered laser beam is changing by time, which will occur less frequent for larger and more frequent for smaller particles. For DLS measurements the system Zetasizer Nano ZS of Malvern Instruments Ltd was utilized. The investigated NCs were usually dispersed in Chloroform to obtain a dispersion with an absorbance not higher than 0.1 A.U. at the first excitonic peak. For the

56 3. Methods 42 measurements a Hellma fluorescence cuvette out of Suprasil quartz glass was utilized. The DLS measurements resulted in the size distribution of the hydrodynamic diameter. The hydrodynamic diameter consists not only out of the inorganic NC diameter, but also includes the thickness of the surfface-bound molecules, the NC ligand shell [182] Transmission electron microscopy (TEM) In microscopy the resolving power of the system is according to Ernst Abbe [183, 184] ultimately limited by the numerical aberration (NA) [184] -which is depending on the degree of perfection of the optical system - and by the wavelength ʎ of the particle used for detecting an object: d = ʎ 2 NA (3.4), where d is the minimal distance that two objects might have to be still distinguishable from each other. In electron microscopy the resolving power of the microscope is enhanced (d is diminished) by using instead of photons in the visible range, electrons witch depending on their kinetic energy might have a much lower wavelength. The wavelength of an electron, like of any other particles, can in principle be calculated according to Louis de Broglie [185] by: ʎ= h p (3.5), where h is the Planck constant and p is the impulse of the particle. Thus, the wavelength of the electron can be diminished when increasing its impulse, achievable by increasing the acceleration voltage applied on the electron. In TEM microscopy acceleration voltages of up to a few hundreds of kv are applied, leading to wavelengths of the electrons of several tenths of picometers. However, due to the aberrations of the subsequently utilized electromagnetic lens system, obtained resolutions are only of a few angstroms, underlining the importance of the lens system, for a highperformance TEM.

57 43 3. Methods When finally hitting the sample, electrons are scattered by the atom nucleus according to Rayleigh scattering. Thus, from the investigated sample first, a diffraction pattern can be obtained by irradiating only a small spot of the sample (selected area diffraction SAD [186] ) from which a potentially present crystal structure of the sample can be found, and secondly a bright field picture showing the transmitted intensity through the sample is obtained. The brightness of different areas of the bright field image thus depend on the density of atoms of the specimen in the respective area, with areas of less density appearing brighter then areas of higher density. For TEM imaging of two dimensional bright field images the Zeiss LEO 912 Omega was utilized, operating at acceleration voltages of up to 120 kv. For sample preparation investigated nanocrystals were washed at 110 C for 5 min in hexanoic acid, precipitated with methanol and a subsequent centrifugation, and redispersed in chloroform at a concentration of about 0.1 mg/ml. The NCs were collected on a carbon film coated 300 mesh copper TEM grid (Quantifoil Micro Tools GmbH, Germany) by a single dip of the grid into the NC dispersion. In case of the bright field images recorded from the active layer of the solar cell, the samples were prepared by dissolving the PEDOT:PSS layer of a hybrid BHJ solar cell in a water bath, thereby delaminating the active layer from the ITO substrate after about 30 seconds. Subsequently, the layer floating on water - was collected as a planar film on the TEM grid. The TEM tomography was recorded with a Jeol JEM2100F electron microscope, operated at 200 kv. All tilt series were obtained in an automatic fashion by using TEMography microscope control software in a tilt angle range of approximately -60 to 60 in steps of 2. The alignment and reconstruction of the data series and visualization of the 3D reconstructed volume was carried out by using the TEMography software packages Composer and Visualizer-kai (System in Frontier Inc., Tokyo, Japan) Scanning electron microscopy (SEM) By scanning electron microscopy (SEM) surface images of an investigated sample can be obtained. Therefore, inside the scanning electron microscope electrons are accelerated inside a vacuum on a small spot of the sample, whose diameter is essentially determining

58 3. Methods 44 the imaging resolution [187]. This spot is moved over the sample surface in a raster pattern to scan the surface. For topographical surface imaging the occurring secondary electrons (SE) are recorded. Secondary electrons are electrons that left the original atom, ionized by inelastic scattering of the primary electron beam. By scanning over the surface a surface image of the sample is thus created, showing an intensity image of the detected secondary electrons which is dependent on the density and conductivity of the sample. Also, the SEM was utilized in a second mode within this thesis for detecting the elements of the sample, which is called energy-dispersive X-ray (EDX) spectroscopy. Therefore, an X-Ray detector is utilized to detect X-Rays emitted by electrons of the sample, which were excited by the primary electron beam into higher energetic states. Then, the energies of the emitted X-Rays can be assigned to characteristic X-Ray irradiation of specific elements. A third common SEM analysis method, which was however not utilized within this thesis, is the detection of elastically back-scattered electrons (BSE). This method allows for a detection of differences in the chemical composition of the sample, but results in a lower resolution compared to images recorded in the SE mode [188]. For this thesis a FEI Quanta 250 FEG SEM was used for SE and EDX mode measurements with typical electron acceleration voltages of 5-20 kv and a spot size of 2-3 nm (thus being also the optimal resolution of the microscope). Information about the containing elements of the sample were obtained in the EDX mode by the INCA energy system (Oxford Instruments Analytical, Halifax, UK), connected to the utilized SEM Atomic force microscopy (AFM) Atomic force microscopy (AFM) is a scanning probe method based on nuclear interactions between the sample and the probe, resulting ideally in images of an atomic resolution in z- direction. Within this thesis atomic force microscopy was undertaken in tapping mode, also called amplitude modulation AFM (AM-AFM) [189]. The tapping mode AFM is an intermediate operation mode compared to the two other AFM modes: the contact mode in which the AFM tip is in permanent contact with the surface, and the non contact mode in which the AFM tip is prevented to touch the sample surface by a feedback-loop. The tapping mode is less destructive than the contact mode and also faster than the non-contact mode, because

59 45 3. Methods less feedback loops are needed during the scanning, since occasional contact with the sample surface must not be avoided. For the operation a small cantilever - typically out of silicon or silicon nitrite - with a sharp silicon tip (of a radius <10 nm) is excited at or near its resonance frequency perpendicular to the investigated sample surface. Thereby, large amplitudes of up to 100 nm are reached. Both the frequency and the amplitude of the cantilever are tried to be held constant, while the tip scans over the surface. Therefore, the cantilever is excited by a piezoelectric source and its frequency and amplitude are measured by a laser beam, which is reflected from the cantilever to a photodetector. When the tip touches or arrives near the sample surface during the scanning process, the amplitude of the cantilever is changing (first increasing due to Van der Waals forces, then decreasing due to electrostatic forces and finally (occasionally) decreased by collision on the sample surface). This amplitude change is detected and the cantilever is retracted in order to reestablish the original amplitude. Now, both the needed distance of retraction and the resulting phase shift between the external excitation signal and the actual measured cantilever vibration are recorded. Thereby, information about the amplitude change and phase shift between the excitation amplitude and the cantilever amplitude are obtained, which is used for the sample surface imaging. After retraction, the cantilever is again approached to the surface, until an amplitude change is sensed again. This feedback-loop is constantly running during the sample scanning [190]. For the utilized measurements the cantilever was driven with frequencies of about 200 khz (depending on the utilized cantilever), over an area of 1x1 µm up to 10x10 µm with a speed of about 1 µm/s. The resolution of the record was of 515x512 data points. Interestingly, from the tapping mode one does not only obtain topographic information, but from the recorded phase shift, also material properties (i.e. distinguishing between organic and inorganic material [191] ) can be obtained. The utilized Veeco Multimode AFM was operated in air. The sample (usually the solar cell active layer on an ITO/glass substrate of ca 1x1 cm) was fixed by a double sided tape on a circular magnetic disc, which was used to fix the probe inside the AFM.

60 3. Methods Solar cell fabrication For this thesis hybrid solar cells were built as bulk heterojunction NC/polymer and graphene- NC/polymer solar cells. Additionally, nanostructured heterojunction hybrid solar cells were built based on a nanostructured inorganic template and framework material. Therefore, now first the synthesis methods utilized to produce the used NCs, graphene-nc hybrids, and the nanostructured aluminia and titania scaffolds for the interdigitated solar cells are described. Subsequently, the integration of these materials to form hybrid solar cells is described Utilized solar cell designs The design of the utilized BHJ solar cells is adopted from the work of Yunfei Zhou [17] and is depicted schematically in Figure 3.5. Three solar cells are built upon one 2x2 cm ITO coated glass substrate. Thereby, the structured ITO (see Section 3.2.6) is forming the anode, a 70 nm PEDOT:PSS layer serves as electron blocking layer, a ca. 80 nm thick hybrid CdSe NC/polymer or CdSe nanocrystal-(thiolated reduced graphene oxide)/polymer layer is the photoactive active layer, and finally a ca. 80 nm thick aluminum top contact serves as cathode. The working principle of this solar cell design is described in Chapter 2.1. Figure 3.5. Top view of the design of the utilized hybrid BHJ device containing three individual solar cells (left middle, right) on one substrate. The active layer consists either of a blend of CdSe NCs and PCPDTBT polymer, or of a blend out of a CdSe NC- thiolated reduced graphene oxide (TrGO) hybrid material and PCPDTBT. (Figure taken from Ref. [21] - reproduced by permission of the PCCP Owner Societies).

61 47 3. Methods The design of the inverted nanostructured TiO 2 nanotube based hybrid solar cells (see Fig. 3.6) had to be adapted to the restrictions given by the anodization process. Figure 3.6. Structure and manufacturing steps of an inverted nanostructured hybrid heterojunction solar cell, built upon a glass/ito/tio 2 nanotube substrate. At first the underlying ITO layer has been separated into 2 parts (1 st step), leading to the creation of a large part on which the ATO is later formed and a smaller stripe to support the later evaporated gold contact. The 2 nd step consists out of titanium sputtering, covering nearly the whole area exposed by the electrolyte during the TiO 2 pore formation (during the 3 rd step), despite the side over which later the top contact is evaporated (during the 4 th step) to ensure that there is no ITO left at the edge of the exposed area what would cause a shortcut between the then underlying ITO and the evaporated gold contact. The 3 rd step consist out of the formation of anodized titanium oxide inside the before mentioned reactor. The final procedures (4 th step) are the spin coating of a conjugated polymer, or of a CdSe NC/polymer solution, followed in some cases by a spin coating of PEDOT:PSS and the evaporation of the gold top-contact, which in this case is representing the anode. Thereby a nanostructured, inverted solar cell is formed, because the TiO 2 is acting as electron acceptor and is directly in contact with the ITO which is therefore representing the anode of the solar cell Hot injection CdSe nanocrystal synthesis The synthesis of II-VI-type semiconductor NCs is using precursors as providers of the NC s molecular building blocks and a solvent acting additionally as ligand for the synthesized NCs.

62 3. Methods 48 An established NC synthesis method is the so called hot injection method [106] where the solvent and surfactant together with one of the precursors are first heated up to the desired synthesis temperature and then the second complementary precursor is rapidly injected. This procedure is enabling a low size distribution of the synthesized NCs, like described in Section 2.4. In this thesis a CdSe NC synthesis based on the publication of Yuan et al. [19] was conducted. At first, the precursor synthesis is described. The cadmium precursor - cadmium stearate - was synthesized by mixing of 10 mmol (1.284 g) of CdO ( %, ABCR) and a 1.75x excess of 35 mmol (9.957 g) of stearic acid ( 98.5%, Sigma-Aldrich) or respectively 35 mmol (7.011g) of lauric acid ( 99%, Fluka) together with 50 mg of succinic acid ( 99.5%, Fluka) acting as catalyst. The educts are weight in into a three neck flask (needed for inserting the temperature sensor through one arm), which is 3 times successively evacuated and then refilled with nitrogen. Then, under vacuum the flask is heated by a heating mantle to 190 C under stirring. In the first 3-5 minutes there is a strong formation of water visible by bubbles, which is removed by vacuum. After 5 min the atmosphere is periodically exchanged with nitrogen. Then, as the water formation is already slowing down, the reaction is continued in nitrogen atmosphere with sporadically evacuations of the flask. During the reaction the color is changing from a strong dark redbrown to lighter red-brown, that is appearing after 1h as light rose. After additional 20 min the solution appears as light yellow. The heating is then stopped and the solution is left for cool-down. When reaching 130 C the three neck flask is opened and the solution is poured into a mortar. After hardening, the color changes into light beige and the cadmium precursor is fine-grinded to a white appearing powder inside the mortar. The selenium precursor - trioctylphosphine selenide - was also synthesized in a three neck flask in which first 10 mmol (790 mg) of selenium ( 99.5%, Sigma-Aldrich) were weight in. The flask was then 3 times successively evacuated and refilled with nitrogen. Subsequently, 10 ml of TOP (trioctylphosphine, 97%, ABCR) were injected into the flask, which was then heated under stirring to 200 C. After 18h the heating was stopped. After cooling down to room temperature, the solution was centrifuged to remove residual selenium and to obtain a clear solution.

63 49 3. Methods The NC synthesis was conducted in a nitrogen flooded three neck flask. First, 2898 mg (12 mmol) of HDA (hexadecylamine, 95%, Merck Schuchardt), 3092 mg (8 mmol) of TOPO (trioctylphosphine oxide, 99%, Sigma-Aldrich), and 444 mg (0.4 mmol) of Cd-stearate were heated up under nitrogen atmosphere inside a 25 ml three neck flask to 300 C. After reaching 300 C, 400 µl (0.4 mmol) of a 1M solution of Selenium in TOP was rapidly injected. The synthesis was continued at 300 C under stirring and stopped after 30 min, unless stated otherwise One-pot CdSe nanocrystal synthesis Besides a hot injection synthesis, also a one pot synthesis of nanocrystals in a laboratory microwave reactor was used, where both precursors and the solvent and ligand are already inside the reaction vessel from the beginning. This method requires less experimental effort, but the reaction is starting with a larger NC size distribution (see Section 2.4). The reason for using a microwave reactor for the NC synthesis within the investigations presented in this thesis are the rigid experimental conditions of an automated microwave synthesis protocol, and the possibility to analyze the NC growth procedure together with several relevant NC parameters in-situ during the synthesis. The in-situ NC characterization synthesis used for these investigations were setup by my coworker Simon Einwächter [192]. For the conducted one-pot microwave synthesis of CdSe NCs a molar ratio of 100:1:1 between HDO (hexadecanol) - used as ligand and as solvent - and the cadmium and selenium precursors is used. First, 5 mmol ( mg) of HDO ( 99%, Sigma-Aldrich) are weighted into a 10 ml glass vial, specially shaped to fit into the laboratory microwave reactor (CEM Corporation, Discover S-Class). Then 0.05 mmol (40.5 mg) cadmium-laureate (Cd-LA) and 0.05 mmol (50 µl) of trioctylphosphine-selenid (TOP-Se) are added. TOP-Se is stored under nitrogen atmosphere and added into the reaction vessel together with a magnetic stirring bar inside a glove box. There, the reaction vessel (10 ml) is afterwards sealed by a snap cap (both, vessels and caps were purchased from CEM). The vial is then placed inside the reaction chamber of the laboratory microwave. The microwave is used in the so-called dynamic power mode, wherein the microwave power is adjusted to reach the set temperature as fast as possible, but without exceeding it, and later the power is adjusted to keep the temperature as

64 3. Methods 50 constant as possible to the set temperature. The maximum power is set to 300 W, the maximum stirring speed for the magnetic stirring bar is set, and target temperatures were programmed to values between 180 C and 220 C for different NC syntheses Nanocrystal dispersion formulation The CdSe QDs, as obtained from the synthesis, were dissolved with a concentration of 1 mg CdSe NCs per 2.5 ml of hexanoic acid and stirred for several minutes at 110 C. The optimum time for the NC washing in HA had to be adapted to each synthesis type (e.g. one-pot synthesis with HDO ligands, or hot injection synthesis with HDA & TOPO as ligands). The optimum treatment time did not only differ between hot-injection or one-pot syntheses, but also more slightly between syntheses performed with the same settings, especially for the less controlled manual hot-injection method. After an adequate time of washing in HA, the double volume (respective to HA) of methanol (MeOH) was added to the solution that was stirred for half the time of the HA treatment, but due to its lower boiling point of 65 C compared to 202 C of HA not longer than 15 min, at 110 C in order to precipitate the CdSe QDs. To separate the QDs from the liquid, the dispersion was centrifuged by the Eppendorf MiniSpin plus centrifuge for 1 min at 14.5 krpm. Hereby one must be careful to prevent the cool-down of the NC/HA/MeOH mixture in order to prevent the re-crystallization of the organic ligands around the NCs, which would drastically reduce the washing effect. Therefore, the centrifugation tube holder is heated in an oven to 90 C prior to usage and the NC/HA/MeOH solution is filled into the centrifugation tubes by a Pasteur pipette as fast as possible. To remove remaining free hexanoic acid, the NCs were redispersed in chloroform (CHCl 3 ) with a concentration of 2 mg/ml and stirred at 105 C for 1 min. Consequently, the triple volume of methanol was added, and the NCs were further stirred for 3 min at 105 C for precipitation. Afterwards, the NCs were collected by centrifugation for 30 s at 14.5 krpm. Chlorobenzene (CB) or other solvents or solvent mixtures were then added to obtain a CdSe QD dispersion with concentrations of typically 24 mg/ml, if not stated otherwise.

65 51 3. Methods Formation of CdSe QD-TrGO hybrid material The formation of a hybrid material out of CdSe NCs linked to thiolated reduced graphene oxide (TrGO) was developed in our group by Pham et al. [20]. Therefore, at first graphite oxide is simultaneously reduced and thiol-functionalized, followed by the decoration of ligandreduced CdSe NCs. Graphene oxide (GO) was synthesized from graphite by a modified Hummers Method [193]. Subsequently, GO was simultaneously reduced and thiol-functionalized with phosphorus pentasulfide (P 4 S 10 ) in DMF (dimethylformamide, 99.8%, Carl Roth) at 120 C under vacuum for 12 h. The final product was collected with a Whatman NL 16 polyamide 0.2 µm membrane filter. Subsequently, a washing procedure of the TrGO on the membrane filter was conducted. Therefore, first DMF ( 99.8%, Carl Roth) was added in the recipient over the membrane filter and left 30 min for soaking. Afterwards the DMF was actively soaked through the filter membrane by applying a vacuum on the flask below the membrane. The same procedure was repeated a second time to further clean the product. TrGO was then redispersed at a high concentration of 1 mg/ml in DMF, for easily weighting in the light material, and homogenized for 10 min in an ultrasonic bath. Afterwards, the TrGO concentration was reduced by dilution down to 50 µg/ml. But, the actual concentration in the TrGO dispersion still decreased by precipitation of larger TrGO aggregates (see Fig. 3.7, left). Therefore, the TrGO concentration must be stated to be <50 µg/ml. However, after about 1 day the non-precipitated particles dispersed in the DMF were forming a quasi stable dispersion. Scanning electron microscopy images reveal diameters of these TrGO flakes of a wide range between ca. 3 µm down to 80 nm. Larger particles seem to be composed out of aggregated sheets of about 1 µm in length (see Fig. 3.7, right). Now, the TrGO-CdSe hybrid was formed as subsequently described. The CdSe QDs, as obtained from the synthesis, were dissolved with a concentration of 1 mg CdSe NCs per 2.5 ml of hexanoic acid and stirred for 15 min at 110 C. In the following, the double volume of methanol was added to the solution that was stirred for another 7 min in order to precipitate the CdSe QDs. To separate the QDs from the liquid, the dispersion was centrifuged by the Eppendorf MiniSpin plus centrifuge for 1 min at 14.5 krpm. To remove remaining free hexanoic acid, the NCs were redispersed in CHCl 3 with 2 mg/ml and stirred at

66 3. Methods 52 Figure 3.7. Left: 12 h reduced and thiol-functionalized TrGO with a concentration of <50 µg/ml one day after dispersion in DMF. Fallout of large, indispersable TrGO aggregates is visible. Right: Scanning electron microscopy (SEM) image of the same TrGO solution dropcasted on an ITO covered glass substrate reveals diameters of the TrGO particles of a wide range between ca. 3 µm down to under 80 nm. Larger particles seem to be composed out of aggregated sheets of about 1µm in length, according to the SEM image. (Figure 3.7 (right) taken from Ref. [21] - reproduced by permission of the PCCP Owner Societies). 105 C for 1 min. Consequently, the triple volume of methanol was added, and the NCs were further stirred for 3 min at 105 C, after that they were collected by centrifugation for 30 s at 14.5 krpm. Chlorobenzene was added to obtain a CdSe QD concentration of 24 mg/ml. For 1 mg of CdSe QDs, 50 µl of the <50 µg/ml TrGO dispersion in DMF (thus containing <2.5µg TrGO ** ) were taken and centrifuged in a 2 ml Eppendorf centrifugation tube for 3 min. Afterwards, all DMF solution was carefully removed by beating the centrifugation tube with its opening showing downwards on a paper towel. Finally, the CdSe QDs/CB solution was added into the centrifugation tube with the collected TrGO inside, resulting in a weight ratio of 1000:<2.5 (CdSe QDs:TrGO) and stirred for 45 s at maximum speed by a vortex mixer. ** Only about 0.01 µg of single layer graphene would be needed for a 50µl QD dispersion of mentioned concentration, if all QDs would be decorated on the perfectly horizontally and parallel stacked graphene sheets within the later produced active layer; given a C-C atom distance of 1.42 Å within graphene.

67 53 3. Methods Preparation of conducting oxide substrate The hybrid solar cells were built on 2 cm x 2 cm ITO coated glass substrates from Bte Bedampfungstechnik GmbH with a measured sheet resistance of R < 35 Ω(sq) and on ITO substartes from Präzisions Glas & Optik GmbH with a measured sheet resistance of R < 10 Ω(sq). The difference in sheet resistance caused only negligible differences of the solar cell performance (by a difference in series resistance), especially for solar cells of high PCEs. The ITO layers of the substrates were structured according to Figure 3.5 and Figure 3.6, allowing for the creation of separated contact areas on the same substrate. The structuring was realized by first covering the ITO substrate by adhesive tape. The adhesive tape was structured by cutting with a scalpel, and areas that should be removed by the later ITO etching were uncovered by removal of the tape with tweezers. The etching of the now uncovered ITO areas was realized by 37% hydrochloric acid (HCl). After 10 minutes immersion in HCl, the substrates were taken into deionized (DI) water to stop the etching. The portions of adhesive tape over the protected ITO areas were manually removed and the substrates were dried under a stream of nitrogen. In order to remove residual adhesive material, the substrates were manually cleaned by acetone with lens cleaning tissue. Subsequently, the substrates were immersed in acetone and placed in an ultrasonic bath for cleaning for 5 min. Afterwards, the substrates were immersed in isopropanol and also placed in an ultrasonic bath for 5 minutes. The same cleaning procedure was also applied with DI water 3 times for 5 minutes. The substrates were then dried under a nitrogen stream and subsequently placed in an oven at 120 C for 20 min for drying. Finally, the surface was further cleaned and hydrophilized by oxygen plasma in a Diener Electronic Femto (Version 5) plasma oven, set to maximal power, for 5 minutes. Now, an aqueous solution of PEDOT:PSS (Baytron AI4083, HC Starck) was spun cast by the WS400-6NPP-Lite spin coater from Laurell Technologies at 2000 rpm for 30 s and dried for 20 min at 160 C, to form a 70 nm thick hole blocking layer Formation of alumina and titania nanostructures For creating a porous aluminum oxide (Al 2 O 3 ) layer (manufacturing procedure was mostly inspired by Ref. [136,234,236,238,267,271] ), first the surface of ITO coated glass substrates from Bte

68 3. Methods 54 Bedampfungstechnik GmbH with a measured sheet resistance of R < 35 Ω (sq) were cleaned by oxygen plasma in a Diener Electronic Femto (Version 5) plasma oven for 5 minutes. Subsequently, a 200 nm aluminum layer was thermally evaporated under a pressure of 1x10-6 mbar from % pure Al procured from Testbourne Ltd by the Edwards AUTO 306 at a rate of 1 nm/s. Afterwards, the substrates were annealed in air for 2 h at 450 C. Then, the substrate was placed inside the anodization reactor (see Fig. 3.8) in which a 0.3 molar aqueous solution of oxalic acid (H 2 C 2 O 4 ) was filled in. A copper disk electrode was placed inside the solution with a distance of ca. 1.7 cm in parallel to the substrate. A Keithley 2602 source-meter was connected to the reaction chamber over which a voltage of 40 V was applied. The current was monitored via a GPIB card by a LabView computer program. After the anodization process carried out at room temperature - was complete (discussion about the reaction stopping time is included in Appendix A7) the substrate was taken out of the reactor, rinsed with DI water and dried under a nitrogen stream. Subsequently, the substrate was soaked for 60 min in an aqueous 5 wt% phosphoric acid (H 3 PO 4 ) solution for widening the pores formed by the anodization process. Figure 3.8. Reactor used for the anodization of aluminum and titanium on an ITO substrate. Part A: Substrate holder with 20 mm x 20 mm x 1 mm mold for fixing the glass/ito/metal substrate. Part B: Aluminum foil (with a round spare in the center, for uncovering the reacting aluminum/titanium area) for contacting the substrate with the anode. Part C: Teflon vessel for filling with electrolyte solution surrounded by an aluminum ring for fixing the vessel to the substrate holder. An o-ring (brown ring in the scheme and in the picture) is placed at the bottom of the Teflon vessel to prevent leaking. Part D: Cap serving as holder for the copper cathode.

69 55 3. Methods TiO 2 nanotubes (manufacturing procedure was mostly inspired by Ref. [138,140,240,268,269] ) were created by anodic titanium oxidation (ATO) on titanium coated ITO substrates in the same reactor. After cleaning the ITO substrate in the same way as described before for Al 2 O 3, titanium (99.995% pure, FHR Anlagenbau GmbH) was sputtered with the RF sputtering function of the Edwards AUTO 306 at a rate of about 0.8 nm/s to create a titanium layer 200 nm in thickness. The sputtering was conducted under a pressure of 1.3x10-2 mbar by creating an argon plasma by a 200 W, MHz RF signal. Subsequently, the substrates were placed into the anodization reactor, which was filled unless stated otherwise - with 0.3 wt% ammonium fluoride (NH 4 F) dissolved in a 90:10 vol%/vol% mixture of ethylene glycol and H 2 O. The anodization was carried out at a voltage of 40 V at room temperature. After finishing the anodization process (a discussion about the reaction stopping time is given in Appendix A6) the substrates were rinsed with isopropanol and DI water and dried under a nitrogen stream Polymer/nanocrystal blend formulation and deposition Ink formulation: The hybrid solar cells presented in Figure 3.5 were formulated from a CdSe/CB solution (according to Section 3.2.4), respectively from a TrGO-CdSe/CB solution (according to Section 3.2.5) with a CdSe NC concentration of 24 mg/ml (unless stated differently in the text). The CdSe QD/CB solution was mixed in a weight ratio of 88:12 (similar the optimal ratio reported by Zhou et al. [125] ) with a 20 mg/ml solution of PCPDTBT (M n =10-20 kda, 1- Material) in chlorobenzene, DCB, TCB or other solvents, or solvent mixtures mentioned in the text of this thesis. In comparison, the TrGO-CdSe QD/CB solution was mixed with a PCPDTBT/CB solution in a weight ratio of 85:15, which was found to be the optimum ratio according to experiments shown in Figure Ink deposition: About 45 µl of the final ink was spun cast onto PEDOT:PSS coated, structured ITO substrates (see Section 3.2.6) with 800 rpm for 30 s, followed by a 60 s drying step at 1800 rpm, resulting in an average active layer thickness of about 80 nm. Subsequently, the substrate

70 3. Methods 56 was cleaned on two sides by a cotton tips soaked in ethanol. Thereby, the spun cast layers (including the underlying PEDOT:PSS layer) were removed from both sides of the substrate - ca. 5 mm. inwards - in order to free the anode and cathode contact area. Thus, leaving only a central ca. 1 cm wide stripe of PEDOT:PSS and NC/polymer blend left on the substrate (see Figure 3.5). Nanostructured inverted hybrid solar cells were built according to Figure 3.6 on a glass/ito/tio 2 -nanotube substrate (see Section 3.2.7). Therefore, first ca. 45 µl of CB solution with 20 mg/ml of PCPDTBT or ca. 45 µl of the prior described CdSe-QDs/PCPDTBT mixture was spun cast with 1000 rpm for 60 s. Subsequently, all areas besides the titanium/titania layer were cleaned by ethanol soaked cotton tips. However, a second drying step was not performed on the spin coater (like for the solar cells created on the ITO substrate), but instead under vacuum (ca. 0.1 mbar) for 30 min at 145 C. On ATO/polymer solar cells also a PEDOT:PSS solution with added 0.5vol% of Triton X100 were spun cast at 3000 rpm for 30 seconds, then again all areas aside from the titania layer were cleaned with an ethanol soaked cotton tip. Subsequently the PEDOT:PSS layer was dried for 10 minutes at 145 C. Top electrode deposition: Finally, for the BHJ solar cells an approximately 80 nm thick aluminum (99.999%, Testbourne Ltd.) layer was evaporated as electron extraction layer by a self assembled thermal evaporator (based on Leybold Vacuum UNIVEX 300) from a tungsten boat, or out of a boron nitrite crucible (which exhibits a longer lifetime), at a rate of 0.6 nm/s. For the nanostructured solar cells a ca. 70 nm thick gold (99.99%) layer was evaporated as top electrode, working as hole extraction layer, by an Edwards AUTO 306 thermal evaporator out of a tungsten boat (gold evaporation from a tungsten boat does not corrode the boat like aluminum does) with a rate of about 1nm/s. The resulting active area was typically in between 0.08 cm² and cm² and was always estimated to be slightly higher than the actual size. If the actual size was determined (i.e. for solar cells of high PCEs), photos of the solar cells were taken to exactly determine the area of the active layer.

71 57 3. Methods 3.3 Solar cell characterization Sun Simulator The main application area up to today for solar cells is the conversion of solar irradiation on earth into electric energy. The sun spectrum on earth is different from that in space due to scattering and absorption by the earth s atmosphere. The longer sunlight travels through the atmosphere, the more its intensity spectrum is differing from the original spectrum. However, in order to be able to compare solar cells throughout the world a standard sun spectrum was defined. Nowadays, the standard intensity spectrum representing the sun illumination is defined by the American Society for Testing and Materials (ASTM) [194]. This reference irradiance spectrum is defined from the spectrum observable on a plane that is facing the sun at an angle of 48.2 relative to a plane at the equator. This inclination of the sun is approximately corresponding to the average sun s position in industrialized countries of the northern hemisphere and has been adopted as the standard irradiance spectrum for solar cell characterization (see Fig. 3.10). This spectrum is called the air mass 1.5 global spectrum (in short AM 1.5 G). Air mass 1.5 referring to the path the sunlight is traveling at 48.2 through the atmosphere compared to the path it would travel perpendicular to the earth s atmosphere; and global means that both direct and diffuse irradiation are considered [195]. Moreover, an integrated irradiation density of 1000 W/m² is used for solar cell characterization, as well as a cell temperature of 25 C. For characterization of solar cells under this standardized AM 1.5 G irradiation spectrum sun simulators are used. The sun simulators are categorized in classes depending on their degree of matching with the standardized spectrum and by the stability of the spectrum and intensity [196]. A comparison of the measured irradiation spectrum of the sun simulator (rated class B in spectral match by LOT-Oriel) utilized in this thesis, and the measured development of the irradiation intensity after switch-on are given in Figure 3.9.

72 Irradiation [W/m²/nm] Relative Intensity [a.u.] 3. Methods 58 3 AM 1.5G Sun Simulator Wavelength [nm] Time [min] Figure 3.9. Left: Comparison of the defined AM 1.5 G irradiation spectrum with the measured spectrum of the sun simulator utilized in this thesis with strong characteristic xenon peaks between 800 and 1000 nm. Right: Measured development of the irradiation intensity of the utilized sun simulator after switch-on. As one can see from above Figure 3.9 there is a mismatch between the AM 1.5 G spectrum and the actual spectrum of the sun simulator. In this example a solar cell which would absorb photons only up to 350 nm would perform much worse than illuminated by the defined sun spectrum. Thus, for an accurate solar cell efficiency determination a mismatch factor [197] must be introduced, which furthermore also considers the mismatch between the spectral response (shape of the current output spectrum) of the reference solar cell used for the sun simulator intensity calibration, and the spectral response of the tested solar cell. This is necessary because the reference solar cell was not calibrated under an exact AM 1.5 G spectrum, but an approximate spectrum, and thus shows in case of a different spectral response then the device under test (DUT) the wrong output current. A very comprehensive explanation of the spectral mismatch correction is given in an application note by Newport Inc Solar cell measurement Solar cells were measured in our laboratory inside a nitrogen filled glovebox by a computer controlled Keithley 2602A source-meter in a 2-point-probe setup. The cells were individually Application Note 51 - The Spectral Mismatch Factor, Newport Inc. (2013).

73 J [ma/cm²] P [mw/cm²] Methods illuminated by a LS0400 LOT-Oriel Sun Simulator, housing a 300 W xenon lamp and using an AM 1.5 G filter. The light is coupled to a solar cell device holder inside the glovebox by a liquid light guide from Lot-Oriel. The light intensity is adjusted by a calibrated silicon reference solar cell to match 100 mw/cm². Selected solar cells were transferred inside a sealed flask to the group of dye and organic solar cells of the Fraunhofer Institute for Solar Energy Systems (ISE) for re-testing and mismatch corrected measurements. Spectral response measurements were conducted by first illuminating with monochromatic light the device under test and recording its current spectrum by a computer-program controlled Keithley 2400 source-meter. Subsequently the same procedure was applied on a calibrated Hamamatsu S1336 photodiode to obtain the spectral response of the DUT. The solar cells were measured in a 4 point probe setup through computer controlled Keithley 2400 sourcemeter. The solar cells were illuminated with a sun simulator from K. H. Steuernagel Lichttechnik GmbH, and the light intensity was adjusted to spectral mismatch corrected intensity by a calibrated reference silicon solar cell V max V OC J P max max J SC Voltage [V] Figure Recorded current density-voltage (J-V) and power density-voltage (P-V) diagram of a manufactured bulk heterojunction CdSe QD/PCPDTBT solar cell. Marked are for the J-V diagram (red curve) the short circuit current density (J SC ), open circuit voltage (V OC ). For the P- V diagram, which is resulting by multiplying the measured current density with the applied voltage and plotting it over the applied voltage, the maximum power density (P max ) is marked together with the corresponding current density (J max ) and voltage (V max ) in that maximum power point.

74 3. Methods 60 The solar cell performance was determined from its current density-voltage (J-V) characteristic (see Fig. 3.10). Therefore, the short circuit current (J SC ), the open circuit voltage (V OC ), and the power density in the maximum power point (P max ) were determined. By division of P max by the illumination power density (P i ) the solar cell power conversion efficiency (PCE) can be determined [198] : PCE = P max P i = V max J max P i (3.6) Moreover, by dividing P max by the product of J SC and V OC, the fill factor (FF) - a solar cell quality criterion - can be calculated: FF = V max J max V OC J SC (3.7) The importance of the fill factor is described in the following subchapter Resistive effects The simple solar cell equivalent circuit (see Fig. 3.11) is composed out of a current source accounting for the generated photocurrent I Ph, a diode modeling the recombination current I D (also called dark saturation current I 0 ), a resistance), and a series resistance R S [199]. parallel resistance R p (also called shunt Figure Simple equivalent circuit of a solar cell with a connected resistive load. The diode is modeling the recombination occurring by applying a voltage V out at the solar cell output, which leads in the model to an increased diode voltage V D by increasing output

75 61 3. Methods voltage (realized by an increased load resistance R L ). The increased diode voltage is leading to an increase of the diode current (recombination current), thus reducing the output current. When the connected load resistance is high enough, the recombination current (together with the shunt current) are compensating the photocurrent, leading to zero output current I out and thus to the open circuit case. The series resistance is representing the active layer (bulk) resistivity and the contact resistances of the solar cell [200]. The parallel resistance R P - also called the shunt resistance - is representing losses of the photocurrent due to shunts through the solar cell s active layer. The output current of the above presented equivalent solar cell circuit can be expressed by the following formula: I out = I Ph I D I P = I Ph I 0 exp q V D nkt 1 V D R P (3.8) According to Figure 3.14, the solar cell output voltage V out can be expressed by: V out = V D V Rs (3.9) So, the diode voltage V D, which is controlling the diode current, can be expressed by: V D = V out + V Rs = V out + R S I out (3.10) By inserting Equation 3.10 into Equation 3.8 the output current can be expressed with: I out = I Ph I 0 exp q (V out +R s I out ) nkt 1 V out +R s I out R P (3.11) So, one can see from the above equations that an increased series resistance is leading to an increased diode voltage (rising the diode current I D ) and to an increased shunt current I Rp, both reducing the output current. And according to Equation 3.11, also a high parallel resistance is needed for a high output current. However, at open circuit (I out =0), the series resistance plays no role, being eliminated from the diode equation:

76 Current [ma] Current [ma] 3. Methods 62 0 = I Ph I 0 exp q V OC nkt 1 V OC R P (3.12) Thereby, the V OC (V out in open circuit conditions) is not influenced by the series resistance. However, the V OC is influenced by the parallel resistance, according to Equation For visualizing the influence of the series resistance and parallel resistance on the shortcircuit current (I out for V out =0) and on the open-circuit voltage, the program PSpice (of Cadence Design Systems Inc.) was used to perform simulations on the equivalent circuit presented in Figure The obtained results for increasing series resistance at a constant high R P and of decreasing parallel resistance at a constant low R S are presented in Figure R P = R S = R S : -6 R P : Voltage [V] Voltage [V] Figure Left: PSpice simulation results of the equivalent solar cell circuit for a fixed parallel resistance and varying series resistances. Right: Respective simulation results for a fixed series resistance and varying parallel resistances. Like stated before, an increased series resistance R S is decreasing the output current, and thus also the short circuit current I SC. Also, like predicted, the series resistance shows no influence on the open circuit voltage (see Fig. 3.12, left). But a low parallel resistance R P does decrease the V OC (see Fig. 3.12, right), and also lead to decreased short circuit current. A series resistance of R S =0 and a parallel resistance of R P would result in a rectangular, ideal diode curve and thus in a fill factor of FF=1. However, the real case the resistances are different from zero respectively from infinite. The parallel and the series resistance can be extracted from the current (or current-density) voltage plot of the solar cell by calculating

77 J [ma/cm²] Methods the slope at the point of J SC (R P can be extracted from here) and at the point of V OC (R S can be extracted from here) according to the following Figure V=0 V R P V = J J V OC V J R = S V J Voltage [V] Figure Extraction of the parallel resistance (R P ) and of the series resistance (R S ) from the measured J-V curve of a solar cell from the tangent in the short-circuit point, respectively the tangent in the open-circuit point Determination of charge carrier mobility For the determination of the charge carrier mobility within a solar cell, a charge extraction by linear increase of voltage (CELIV) measurement can be conducted on a solar cell. Therefore, during the CELIV measurement (see Figure 3.14) a voltage ramp A in reverse bias is applied to the solar cell. Extracted charges are visible as a current in an external sensing circuit. The faster the charges are extracted the faster their mobilities are [201, 202]. Figure Schematic of the measured CELIV current signal, and resulting signal parameters.

78 3. Methods 64 From the time t max of the current peak one can calculate the mobility µ of the fastest charge carrier. In case that the current density caused by the geometric capacitance of the solar cell j(0) is in a similar range as the maximum of the extracted current density Δj the following formula can be applied to calculate the charge mobility: µ = 2d² 3At max 2 [ j j 0 ] (3.13),with d being the thickness of the active layer and A being the applied voltage slope. The current density due to the capacitive response j(0) of the solar cell can be calculated as follows: j t = C du t dt 1 du (t) = ε 0 ε r d dt du (t) =A dt A j = J 0 = ε0 ε r d (3.14) and then subtracted from the measured current density j to obtain Δj, which is representing the actual targeted current density to measure. So, one has to first obtain the capacity C of the solar cell to make a statement about the mobility. However, only in case that hole mobility µ h and electron mobility µ e would be very different from each other and the number of both positive and negative charges being similar, both current peaks would be visible during the extraction. But, if both mobilities are in a similar range, one will not be able to distinguish between the charge carriers. For the determination of the charge carrier mobility (results mentioned in Chapter 5.2), solar cells were transported to FLUXIM AG, Winterthur (Switzerland) for CELIV examinations as well as impedance and C-V (capacitance voltage) measurements. The solar cells used for the measurements were sealed before the transportation against degradation in air with a 1 mm thick glass plate, fixed by melting for 3 min at 145 C of a 25 µm thick thermoplastic sealing film (purchased from Solaronix SA, Aubonne, Switzerland), placed in between the solar cell and the glass plate. The measurements were performed using the PAIOS (version 1.0) measurement system for steady-state, transient and AC measurements by FLUXIM AG. This system has the advantage of performing a variety of measurements within a few seconds, thus being able to minimize changes of the sample in between different characterization methods.

79 65 4. Results - Solar cell reproducibility 4. Results - Solar cell reproducibility There are two main problems when using NCs as acceptor material in comparison to PCBM and small molecules. One is the existence of insulating surfactants (ligands) deriving from the synthesis of NCs, which are necessary for obtaining solution processable nanoscopic crystals and enabling their colloidal stability. Like mentioned in previous chapters, these NC surfactants are hindering the charge transfer from the polymer towards the NCs at the organic-inorganic interface and the charge extraction over the NC network towards the cathode. The other drawback is the variation in NC quality. While in comparison PCBM and small molecules have a well defined molecular structure, NCs can differ in size and size distribution during their synthesis (see Chapter 2.3.2). Also, depending on the synthesis procedure, NCs have different surface ligand compositions and constitutions as well as surface traps and crystal impurities which can alter the energy levels of the NCs. These differences influence the optical properties of NC, which can be detected by UV-Vis and photoluminescence (PL) spectroscopy, but also the ligand sphere constitution whose diameter can be approximated by DLS measurements [182]. Within this chapter it will be demonstrated that both the NC properties variations and the post-synthetic ligand sphere reduction procedure are interconnected, and that the control of both is required in order to obtain highly efficient hybrid BHJ solar cells. Within this chapter at first the influence of the intra-batch variations during the NC synthesis on the post-synthetic treatment and on the resulting solar cell performance were investigated. Therefore, a low temperature one-pot CdSe NC synthesis was performed within a laboratory microwave reactor with an attached PL-spectrometer, allowing for in-situ monitoring and recording of the NC synthesis. Subsequently, NCs of specific properties from certain stages of the synthesis could be compared with respect of colloidal stability after post-synthetic treatment and the resulting performance after incorporation into hybrid BHJ solar cells. Secondly, the influence of typical batch-to-batch variations on the post-synthetic treatment and on the solar cell performance is discussed. This later discussion is made for CdSe NCs synthesized by a manual hot-injection method, since the NC quality reached with the more reproducible automated one-pot microwave synthesis was inferior.

80 4. Results - Solar cell reproducibility Intra-batch NC variations Within this section, by measuring the PL signal of the NCs during their synthesis, the development of several NC quality parameters (surface quality, size, size distribution) could be mapped in their evolution. Thereby, points within the synthesis with a specific set of NC properties could be determined and reproduced due to the chosen method of an automated one-pot synthesis. And subsequently, the influence of these NC properties on the postsynthetic treatment and on the hybrid solar cell performance could be studied. Also, post-synthetic ligand exchange experiments were conducted on the synthesized NCs, aspiring for an easy, mildly executable ligand substitution for hexadecanol capped CdSe NCs. In case of HDO the oxygen atom of the alcohol group is coordinating towards the Cd atom (which is partially positively charged due to the Se atom in CdSe), while the aliphatic tail provides the solubility in organic solvents. The CdSe synthesis was executed inside the microwave reactor, wherein the PL signal development was observed in-situ. Furthermore, post-synthetic PL and UV-Vis spectroscopy, and DLS investigations were execute from externally performed measurements. Also, ligand exchange following a precipitation of NC, and finally the manufacturing of hybrid solar cells were performed. The ligand-exchange experiments gave insight on the impact of different substituent molecules on the colloidal stabilization of the NCs and on the solar cell performance, but did not result in improved PCE. Thus, the results are mentioned and discussed in Appendix A2. The CdSe NCs investigated within this chapter are synthesized in a microwave reactor, connected to a PL spectrometer. Hence, the PL signal of the NCs during the synthesis can be detected (see Figure 4.1). Therefore, optical waveguides (one for excitation and one for detection) were brought in close distance to the reaction vessel within the microwave reactor and connected to PL spectrometer (HORIBA JobinYvon FluroLog). Details about this setup are described in a publication of my coworker Simon Einwächter [192]. Since, the NC PL signal is quenched by temperature (see Appendix A1), a low temperature synthesis method which is also suitable for one-pot synthesis inside the microwave reactor, was chosen. Specifically, a synthesis utilizing hexadecanol (HDO) as solvent and ligand was chosen,

81 67 4. Results - Solar cell reproducibility allowing for a one-pot CdSe NC synthesis at a low temperature around 200 C, which was studied within our group in a master thesis by Lina Martinsson [203] for CdS NCs. Figure 4.1. : Nanocrystal synthesis and in-situ monitoring system, composed by a laboratory microwave synthesis reactor and a PL spectrometer, enabling an in-situ observation of the PL development of the NCs during their synthesis. Like mentioned in Chapter 3.1.1, from the measured PL signal the PL peak position (correlated to the average NC size), the full width at half maximum (FWHM) (correlated to their size distribution), and the PL intensity (proportional to the relative NC surface quality [18] ) can be extracted. At first, the CdSe NC syntheses were mapped for 4 different synthesis batches at each of the three chosen temperatures (180 C, 200 C, and 220 C) utilizing in-situ PL spectroscopy. Different temperatures were chosen in order to demonstrate that the findings are not valid only for a single set of conditions, but can be applied more generally. In the following figures (Fig. 4.2, 4.3, 4.4) the extracted temperature corrected values calculated from the synthesis temperature down to room temperature for better comparability (see Appendix A1) are presented. The broad lines represent the average of the 4 different recorded syntheses. The points indicated by filled circles are time points at which in subsequent syntheses CdSe NC products were synthesized for post-synthetic treatment and for hybrid solar cell manufacturing.

82 PL Intensity [CPS] FWHM [nm] PL Peak Position [nm] NC Concentration [mg/ml] PL Intensity [CPS] FWHM [nm] PL Intensity [CPS] FWHM [nm] PL Peak Position [nm] 4. Results - Solar cell reproducibility k 150k 100k 220 C A B C D A B C D 50k C NC Synthesis Time [min] NC Synthesis Time [min] Figure 4.2. Left: Photoluminescence (PL) Intensity and full width at half maximum (FWHM) extracted from the in-situ monitoring of 4 different microwave syntheses carried out at 220 C. Right: Extracted PL peak position of the respective synthesis. The bold lines are indicating the arithmetic mean values of individual measurements (thin lines), and the dotted borders of the colored area are indicating the calculated standard deviation. 200k 200 C A B C D k 100k 60 50k NC Synthesis Time [min] 30 Figure 4.3. Left: Photoluminescence (PL) Intensity and full width at half maximum (FWHM) extracted from the in-situ monitoring of the microwave synthesis carried out at 200 C. Right: Extracted PL peak position of the respective synthesis. 200k 150k 100k 180 C k C NC Synthesis Time [min] NC Synthesis Time [min] 0 Figure 4.4. Left: Photoluminescence (PL) Intensity and full width at half maximum (FWHM) extracted from the in-situ monitoring of the microwave synthesis carried out at 180 C. Right: Extracted PL peak position of the respective synthesis (orange curves) and post-synthesis determined NC concentrations (red triangles).

83 69 4. Results - Solar cell reproducibility One can observe that despite the continuous changes of the recorded NC properties during the synthesis, the batch-to-batch reproducibility is decent. The following Table 4.1 summarizes the average PL intensity, PL peak position and the time when the minimum FWHM is reached for all 3 synthesis temperatures. Table 4.1. Average synthesis time at which the point of minimum FWHM is reached for the CdSe NC one-pot microwave assisted syntheses, conducted at 180 C, 200 C, and 220 C, together with the PL intensity, PL peak position in that points, converted from the synthesis temperature to room temperature for direct comparability. Temperature [ C] Time [min] FWHM [nm] PL intensity [CPS] PL peak position [nm] k k k From the above comparison one can see, that the one-pot microwave assisted CdSe NC syntheses are producing more similar NC sizes (lower FWHM) at higher temperatures such as 200 C and 220 C compared to 180 C. This might be due to a better separation between the initial NC nucleation process and the NC growth (see Chapter 2.4). Moreover, the point of the lowest FWHM (which is almost coinciding with the point for the maximum PL intensity the bright point) is occurring after longer synthesis times for lower synthesis temperature, which is corresponding to the reduced precursor reactivity for lower temperatures. Interestingly, the values for the resulting PL intensity are nearly the same for all 3 tested synthesis temperatures. However, NCs from the synthesis at 180 C are smaller (according to the smaller PL peak position), which might be an indication that a longer nucleation phase used more precursor material to create a higher number of NC nuclei leading to an overall reduced size for the resulting NCs.

84 Absorption [AU] 4. Results - Solar cell reproducibility 70 TEM pictures of the synthesis products from 220 C were taken. In the following Figure 4.5 UV-Vis absorption spectra and TEM pictures of the utilized synthesis products from 220 C are shown min (A) 9 min (B) 24 min (C) 50 min (D) C Wavelength [nm] Figure 4.5. Left: UV-Vis absorption spectra of CdSe NCs synthesized at 220 C for 4 min, 9 min, 24 min and 50 min recorded in chloroform. Right: TEM pictures of the respective NCs. UV-Vis spectra of the CdSe NCs synthesized at 220 C reveal a clearly distinguishable 1 st excitonic absorption peak which is most prominent for the shortest synthesis time. The absorption tail at wavelengths above is flat, indicating a good dispersibility of the NCs in chloroform. TEM pictures show that the NCs defer from a spherical shape and are sometimes slightly elongated with aspect ratios between 1:1 up to 1.5:1, and rarely even as high as 2:1. Determinable diameters of the NCs are of 4.9 nm to 5.3 nm Influence of post-synthetic NC treatment In the following the development of UV-Vis absorption, PL intensity, and hydrodynamic diameter (determined by DLS) as a function of post-synthetic hexanoic acid washing time is described; giving some insight of the changes that are occurring during the NC ligand reduction process. Therefore, in the following the example of the behavior of the NCs synthesized for 9 min at 220 C (bright point, point B) is given in detail. Later, the experiments are repeated also for NCs from other synthesis times at 220 C.

85 Absorption [AU] PL Intensity [CPS] Results - Solar cell reproducibility UV-Vis absorption spectroscopy & PL intensity: From the following Figure 4.6 one can see that the absorption spectra of the NCs washed for 1 min and 5 min in hexanoic acid (HA) are apparently unchanged with respect to the untreated sample (dashed line) C 9 min (B) 0 min HA 1 min HA 5 min HA 7 min HA 9 min HA 11 min HA 100k 10k 220 C 9 min (B) Wavelength [nm] 1k HA Washing Time [min] Figure 4.6. Left: UV-Vis absorption of CdSe NCs synthesized at 220 C for 9 min and postsynthetically treated by different HA washing times at 100 C followed by precipitation in MeOH at 100 C for 3 min. Right: PL Intensity of the same NCs at different HA washing times. Nevertheless, from 7 min of HA treatment on the absorption begins to increase, and the dispersion is becoming increasingly turbid, visible clearly from the increased absorption for wavelengths after the 1 st excitonic peak. However, from PL spectrometry one can extract a strong decrease of the PL peak intensity of 95% already after 1 min of hexanoic acid washing. Dynamic light scattering: From the following Figure 4.7 one can observe that untreated CdSe NCs exhibit a hydrodynamic diameter of 22 nm in chloroform, as measured by DLS.

86 Diameter [nm] 4. Results - Solar cell reproducibility <2 min after HA washing 80 min later C, 9 min (B) HA washing time [min] Figure 4.7. Hydrodynamic diameter determined by dynamic light scattering of CdSe NCs washed with hexanoic acid (HA) for different times at 100 C. The filled circles represent the average hydrodynamic diameter directly after HA washing, precipitation with methanol, and re-dispersion in chloroform. The void circles represent the measured hydrodynamic diameter measured 80 min after the washing procedure, indicating the colloidal stability of the NCs. Furthermore, until a HA washing time of 5 min the initial hydrodynamic diameter remains nearly unchanged. Only after a HA washing of 7 min the measured hydrodynamic diameter is changing, showing a rise to an average diameter of 30 nm. Executing the HA washing for 9 min results in a strong increase to an average of 410 nm, and 11 min of HA washing results in an average hydrodynamic diameter of 629 nm. By repeating the DLS measurements of the NC dispersions 80 min after the first measurement, the colloidal stability of the NCs was examined. The untreated, as-synthesized, NCs showed no change after 80 min. However, NCs washed for 3 min showed an increase by 15 nm of the hydrodynamic diameter, while the 5 min HA washed NCs exhibited a doubling of the average diameter. The 7 min washed NCs exhibited even a more then quintuple increase of the measured hydrodynamic diameter after 80 min in dispersion. The measurements for the 9 min and 11 min washed NCs indicated similar high values as directly after the washing. Hence, in general one can say that the longer the NCs were washed in HA, the stronger the recorded increase of the hydrodynamic diameter after 80 minutes is. From the increasing measured hydrodynamic diameter for the 80 min old dispersions one might conclude that the increasing diameter is corresponding to an occurring aggregation of the NCs. Accordingly, also increasing hydrodynamic diameters measured directly after HA washing indicate an aggregation of the NCs. Interestingly, the increased hydrodynamic diameter, measured by DLS for HA treatment times of 7 min and longer, is also visible from the UV-Vis absorption signal (see Fig. 4.6 left),

87 PL Intensity [CPS] Diameter [nm] Diameter [nm] Results - Solar cell reproducibility which is starting to show an increased baseline due to the increased turbidity of the solution with increased HA treatment time. A conclusion, meaningful for the application in solution-based NC containing solar cells, is that one can derive from both DLS and roughly also by UV-Vis absorption spectroscopy the maximum washing time applicable to the NCs. This maximal washing time is shortly before the NCs exhibit the strong, clearly visible decrease in colloidal stability (which in the prior described case is of about 5 minutes). Influence of different synthesis times: The same experiments - measuring the PL intensity and hydrodynamic diameter by DLS - were also executed for CdSe NCs synthesized at 220 C for 4 min, 24 min, and 50 min. A comparison of the results of all measured NCs from the 220 C one-pot microwave-assisted synthesis is depicted in the following Figure k 220 A 220 B 220 C 220 D A 220 B 220 C 220 D A 220 B 220 C 220 D 10k k HA Washing Time [min] HA washing time [min] min later HA Washing Time [min] Figure 4.8. PL intensity development (left), hydrodynamic diameter directly after washing (middle), and hydrodynamic diameter 80 min after washing (right), for different hexanoic acid washing times of CdSe NCs synthesized at 220 C. Now, if comparing the behavior of the NCs from different time points of the synthesis, one can see that the HA washing is altering the colloidal stability of the NCs with shorter washing times for longer NC synthesis times. The 50 min synthesized NCs are not colloidal stable even untreated, as their average hydrodynamic diameter is increasing after 80 minutes in dispersion from 25 nm to 37 nm. Conclusively, one can say that the optimum HA washing

88 Diameter [nm] 4. Results - Solar cell reproducibility 74 time for 4 min synthesized NCs at 220 C (220 A) is of more than 11 min, for the NCs synthesized for 7 min (220 B) the optimum washing time would be of 7 min, for the 24 min (220 C) synthesized NCs the optimum HA treatment time would be of slightly less than 3 minutes, and the CdSe NCs synthesized for 50 min would require no HA washing for reduction of the insulating ligand sphere, but only a precipitation in methanol. For estimating the reliability of the measurements and the statement from the DLS, the measurements were repeated with a second NC synthesis batch (see Fig. 4.9) at a synthesis time of 9 min C, 9min ( ) +80 min 220 C, 9min ( ) +80 min HA Washing Time [min] Figure 4.9. Hydrodynamic diameter determined by DLS of CdSe NCs synthesized for 24 min at 220 C washed for different times with HA. The values were determined for NCs from 2 different NC syntheses right after the washing (filled circles) and 80 min later (void circles). The results of the DLS measurements are very similar, especially for the values measured right after the hexanoic acid washing Influence on solar cell efficiency First, it must be noted that in order to create intercomparable solar cells, the NC concentration from each synthesis point has to be determined. The NC concentration in the unpurified synthesis solution was following the trend of the NC growth curve; with a fast growing NC concentration in the beginning of the synthesis and a saturation of the NC concentration value for longer synthesis times (example see Fig. 4.4, right). Whereby a reaction yield of 80% was reached near the bright point of the NC synthesis for all 3 synthesis temperatures and did not further change afterwards. Solar cells were built only Based on dividing the mass of the weighted ligand stripped CdSe NCs (after HA washing, precipitation with MeOH, redispersion in CHCl 3, and evaporation of the solvent by a heating gun at about 170 C) by the mass resulting from the sum of all Cd and Se atoms from the precursors, calculated to be inside the sample volume.

89 75 4. Results - Solar cell reproducibility from NCs of syntheses at 220 C and 200 C, as they exhibited a lower size distribution and a faster synthesis (see Table 4.1). However, the focus was laid on the synthesis at 220 C, because at this temperature NCs of very different constitution (i.e. different PL intensity, FWHM, and absorption peak position) can be achieved within short synthesis times (see Fig. 4.2). Nevertheless, also solar cells from CdSe NCs synthesized at 200 C were built in order to support the presented findings; these experiments are shown in Appendix A5. Properties of utilized NCs: At first, the solar cell performance of CdSe NC/PCPDTBT bulk heterojunction solar cells was investigated for solar cells containing NCs synthesized at 220 C. Hence, at a temperature at which the NC synthesis already exhibits its bright point after 9 min and Ostwald Ripening (visible by increasing FWHM) already begins at an early stage of the synthesis after about 18 min (see Fig. 4.2, left). Thus indicating that HDO is not a suitable ligand for a stable NC synthesis at the elevated temperature of 220 C. However, like stated before, it is interesting for this study, since NC batches with very different properties can be obtained within a short synthesis time. So, NCs from syntheses at 220 C for 4 min, 9 min, 24 min and 50 min (properties see Table 4.2) were utilized, applying the same HA washing time of 3 minutes to all NCs. Table 4.2: FWHM, PL intensity and PL peak position for CdSe NCs synthesized for different times at 220 C (values calculated from measurement temperature to room temperature). NC Sample Synthesis time [min] FWHM [nm] PL intensity [CPS] PL peak position [nm] 220 A k B k C k D k According to the prior experiments, 3 min of HA washing proved to be short enough to allow for colloidal stability of the treated NCs - except for the 50 min synthesized NCs. Therefore, for these NCs (220 D), also a very short HA washing of 1 min was applied in order to achieve a similar colloidal stability. However, one have to keep in mind that for the 4 min and 9 min

90 J SC [ma/cm 2 ] V OC [V] 4. Results - Solar cell reproducibility 76 synthesized NCs a washing time even longer than 3 min would be indicated, according to the prior description, to have exactly the same colloidal stability for all compared NC batches. Solar cell fabrication: After the post-synthetic treatment with HA and precipitation with methanol, the NCs are dispersed in chlorobenzene, establishing a concentration of 20 mg/ml. The NC solution was mixed with a 20 mg/ml solution of the conjugated polymer PCPDTBT in a 88:12 weight ratio prior to the spin casting on a PEDOT:PSS coated ITO/glass substrate inside the glove box. The last fabrication step is a thermal Al evaporation step of the 40 nm thick aluminum cathode at a rate of 5 Å/s in a vacuum of around mbar. Solar cell results: The results of the hybrid solar cells manufactured from NCs of the synthesis points 220 A, 220 B, and 220 C (with 9 solar cells from 3 different substrates from each point) are presented in the following Figures 4.10 & The graphs display the evolution of the solar cell parameters during annealing at 145 C inside a glovebox A 220 B 220 C Annealing Time [h] Annealing Time [h] Figure Measurement results of CdSe NC / PCPDT-BT bulk heterojunction solar cells during annealing at 145 C, containing HDO capped NCs from each point (A,B,C) of the synthesis at 220 C, washed in HA for 3 min. The bold lines are indicating the arithmetic mean values of 9 individual measurements (thin lines), and the dotted borders of the colored area are indicating the calculated standard deviation.

91 FF PCE [%] Results - Solar cell reproducibility From the J SC graphs one can extract that the short-circuit current is rising during annealing, and that it is rising fastest for solar cells manufactured from NCs of the longest synthesis times. The highest peak J SC value is reached during 52 hours of annealing for solar cells from NCs of the point of the lowest FWHM and highest PL intensity (220 B). The open circuit voltages are very similar with starting values of 700 mv for short annealing times saturating down to 660 mv for annealing times of up to 52h. However, there are two cells that are an exception, both from solar cells containing NCs from the synthesis point 220 C (24 min). Solar cells with NCs from this synthesis time show for 2 out of 3 cells a strong V OC decrease by annealing, down to about 350 mv. This decrease might be explained by occurring NC aggregation during annealing, inducing NC surface defects which increase recombination rates and thereby reduce the V [204] OC or (or additionally) by NC aggregates forming paths for shorting anode and cathode of the device A 220 B 220 C Annealing Time [h] Annealing Time [h] Figure Measurement results of CdSe NC / PCPDT-BT bulk heterojunction solar cells during annealing at 145 C, containing NCs from each point (A,B,C) of the synthesis at 220 C, washed in HA for 3 min. The initial average fill factors of all solar cells have a value of about However, only FFs for solar cells from the two longer synthesized NCs (220 B, 220 C) are initially increasing upon annealing. The solar cells containing NCs from the shortest synthesis (220 A, 4 min) show an exponential decrease of FF down to 0.2 after annealing for about 2 days at 145 C. As it can be seen from Figure 4.12, the decrease is partially due to a decrease of the parallel

92 R P [ cm²] R S [ cm²] 4. Results - Solar cell reproducibility 78 resistance (R P ), but mostly the consequence of a strong and steady increase of series resistance (R S ). In comparison, the series resistance of the solar cells of the hybrid solar cells containing NCs from a later point of the synthesis, i.e. 220 C (24 min synthesis time), stays almost constant during annealing. The very strong increase of the series resistance of hybrid solar cells from the milder washed (in the sense of causing less NC aggregation) NCs of sample 220 A might be explained by an accumulation of organic ligands from the NCs ligand sphere accumulating and forming a barrier for charge transfer e.g. at the interface between active layer and top electrode, for which there is however no direct evidence. 4k 220 A 3k 220 C 2k 1k Annealing Time [h] 4k 3k 2k 1k A 220 C Annealing Time [h] Figure Parallel and series resistance values extracted from J-V characteristics of hybrid CdSe NC/PCPDTBT BHJ solar cells during annealing at 145 C, containing NCs synthesized at 220 C for 4 min (220 A) and 24 min (220 C), both washed by hexanoic acid for 3 min. The parallel resistance of the solar cells made from NCs that were relatively stronger washed causing stronger NC aggregation even for the same HA treatment time according to the prior described DLS measurements namely the solar cells containing NCs from the synthesis point 220 C is already lower before annealing, and always remains lower during annealing. This is indicating a stronger NC aggregation in the solar cells containing a longer acid-washed NCs. since aggregation might indirectly increase the density of surface defects on the NCs by removal of the original NC surface ligand, and thus increases the charge recombination inside the active layer of the solar cell, leading to a decreased parallel resistance.

93 V OC [V] PCE [%] J SC [ma/cm²] FF Results - Solar cell reproducibility Influence of different HA washing times on the same NC batch: In the following Figure 4.13, also the performance of solar cells made from NCs synthesized for the same time (50 min, point 220 D) is depicted. The washing time of the NCs was 3 min and also for a reduced short time of 1 min. Thus, testing the influence of a more stable NC dispersion on the hybrid solar cell, resulting for the shorter HA washing time according to the DLS results D min HA 1 min HA Annealing Time [h] Annealing Time [h] Annealing Time [h] Annealing Time [h] Figure Measured performance of CdSe NC / PCPDTBT bulk heterojunction solar cells incorporating NCs synthesized at 220 C for 50 min (220 D) for 3 min and for 1 min HA washing during annealing at 145 C. Solar cells from NCs with stronger reduced ligand sphere (due to the longer washing time of 3 min) show a higher shortcut current (J SC ) right after beginning of the thermal solar cell annealing, which however is decreasing with increasing annealing time while the shorter washed NCs cause the solar cell s J SC to run into a saturation trend. Moreover, the open circuit voltage (V OC ), of the solar cells containing shorter washed NCs is also considerably more stable during thermal annealing, leading ultimately to more stable PCE values over

94 4. Results - Solar cell reproducibility 80 annealing time. Thus indicating that for the CdSe NCs taken from this long synthesis time of 50 min only the very short HA washing of 1 min is able to produce stable solar cells operating under elevated temperatures. Now, in the following the results for the best cells and the average values of solar cells manufactured from CdSe NCs of the four selected times of the microwave-synthesis at 220 C are summarized in Table 4.3. Table 4.3. Best and average results (indicated in brackets) from all solar cells after the indicated annealing time from of the four selected time points of the 220 C microwave synthesis. The NCs were all washed for 3 min in HA. NC sample Annealing C [h] J SC [ma/cm 2 ] V OC [V] FF PCE [%] 220 A (2.41) 0.64 (0.66) 0.20 (0.20) 0.56 (0.31) 220 B (3.93) 0.68 (0.68) 0.31 (0.35) 1.51 (0.92) 220 C (2.97) 0.66 (0.45) 0.37 (0.37) 1.06 (0.50) 220 D (3.60) 0.66 (0.65) 0.37 (0.38) 0.95 (0.88) Summary In summary, from the solar cell results presented above (incorporating NCs synthesized at 220 C) and in Appendix A5 (incorporating NCs synthesized at 200 C) one can conclude that both the NC quality (i.e. NC size homogeneity, NC surface quality) as well as the postsynthetic NC treatment have an impact on the solar cell performance. For example, from the average short-circuit current over synthesis time plot (see Fig. 4.14), one can see that that the plot follows the PL bright point behavior of the NC synthesis and therefore also the development of the FWHM value of the utilized NCs.

95 J SC [ma/cm²] Results - Solar cell reproducibility C 220 C A A B NC Synthesis Time [min] C B C D bright point D Figure Development of the maximum average short-circuit current density from 3 solar cells in each point over the NC synthesis time from NCs synthesized at 200 C and 220 C and washed in HA for 3 min. The single red circle represents the average J SC for solar cells from NCs synthesized at 220 C for 50 min, and 1 min HA washing time (for which the colloidal NC stability is more comparable after HA washing to the used NCs from earlier points of the synthesis). The upper Figure 4.14 indicates that NCs taken from the bright point of the synthesis, where the surface annealing is in its optimum point, before being dominated by Ostwald ripening, leading to increased FWHM, result in hybrid NC/polymer solar cells of the highest J SC. Moreover, the size of the NCs might have an influence on the V OC, with small NCs resulting in higher V OC (even though this behavior might derive from a different degree of ligand reduction - see Table A2). Therefore, a tracking of the NC synthesis is needed for obtaining the best quality NCs (i.e. minimal FWHM) for the use in solar cells. However, the post-synthetic treatment of the NCs is at least of similar importance compared to the properties of the utilized NCs, as it is strongly influencing important solar cell parameters like fill factor and thermal stability of V OC and J SC. The optimum post-synthetic NC washing must be strong enough to remove enough ligands from the NC surface to prevent the intake of an excess of ligands into the hybrid solar cell, which later lead to high series resistance of the solar cell during thermal annealing. At the same time, the washing should not be too strong to still allow colloidal stability of the NCs for the integration into a NC/polymer dispersion.

96 4. Results - Solar cell reproducibility Batch-to-batch NC variations Since the one-pot synthesis presented in the previous Section 4.1 did not produce very homogenous CdSe NCs, leading to rather low short circuit currents compared to the current state of the art, the following experiments will use CdSe NCs synthesized by a manual hotinjection synthesis using a different ligand system (HDA/TOPO instead of HDO), allowing for production of nearly monodisperse CdSe QDs [19], which already led to hybrid BHJ solar cells of up to 2.7% PCE [125]. Nevertheless, by choosing this synthesis method, on the one hand NCs with higher size and shape homogeneity can be produced but on the other hand higher batch-to-batch fluctuations are resulting due to the manual synthesis. Hence, in the following, methods will be shown how state of the art solar cells can efficiently be obtained despite the possible batch-to-batch variations of the utilized NCs. Achieving smaller NC size distribution: The hot injection colloidal NC synthesis (performed according to Chapter 3.2.2) already has the advantage over a one-pot synthesis to generally result in a lower NC size distribution due to a better separation between nucleation and growth (as described in Section 2.4). However, due to results obtained within our group showing a decrease of the minimum FWHM by using a higher precursor concentration in the hot-injection CdSe QD synthesis in HDA/TOPO [16]. It was tested whether the reported results could be reproduced, and the CdSe NCs were additionally utilized in BHJ hybrid solar cells. Hence, the PL development of two hot injection syntheses at 300 C for (HDA/TOPO):(Cd-SA):(TOP-Se) ratios of 100:1:1 and 100:2:2 was mapped by manually extracting a series of samples, which is depicted in the following Figures 4.15 & Moreover, by these experiments, also the influence of a fluctuation of the precursor-to-ligand ratio has been investigated.

97 FWHM [nm] Absorption [A. U.] PL Peak Position [nm] PL Intensity [CPS] Results - Solar cell reproducibility k k :1:1 100:2: Time [min] Time [min] Figure Left: Development of the PL peak position during hot-injection synthesis of (HDA/TOPO):(Cd-SA):(TOP-Se) at 300 C for ratios of 100:1:1 and 100:2:2. The solid line is connecting the average values recorded from two different syntheses of the same ratio. Right: Respective development of the PL intensity :1:1 100:2: :1:1 (#1 & #2) 100:2:2 (#1 & #2) Time [min] Wavelength [nm] Figure Left: Development of the FWHM during hot-injection synthesis of (HDA/TOPO):(Cd-SA):(TOP-Se) at 300 C for ratios of 100:1:1 and 100:2:2. The solid line is connecting the average values recorded from two different syntheses of the same ratio. Right: UV-Vis absorption signal of NCs extracted from 4 different syntheses after 30 min. Indeed, like described by F.-S. Riehle [16], the minimum FWHM of the synthesis is decreasing while the average final NC size is decreasing as well for higher precursor concentration, pointing towards an improved separation between nucleation and growth. Moreover, the 1 st excitonic peak is also more prominent in the UV-Vis absorption spectrum (see Fig. 4.16, right). The benefit of a lower FWHM of the NCs used in hybrid solar cells, found in the previous chapter, caused the decision to now use NCs synthesized at a higher precursor concentration for the solar cell fabrication.

98 PL Intensity [CPS] PL peak position [nm] 4. Results - Solar cell reproducibility 84 Achieving increased NC PL intensity: Inspired by the observation of Yunfei Zhou [17] that NCs of a higher PL intensity result in higher hybrid solar cell performance, it was tried to achieve a higher PL intensity by increasing the Se:Cd precursor ratio, as described by Qu et al. [205]. Like depicted in Figure 4.17, also the impact of a higher Cd:Se ratio was investigated Cd:Se 3:2 Cd:Se 2:2 Cd:Se 2: Time [min] Time [min] Figure Left: Development of the PL intensity during hot-injection synthesis of (HDA/TOPO):(Cd-SA):(TOP-Se) at 300 C for ratios of 100:3:2, 100:2:2, and 100:2:3. The dashed lines represent an interpolation over the average results of two different syntheses by a B-spline. Right: Respective development of the PL peak position. As described by Qu et al. [205] the increase of the selenium precursor concentration led to a higher maximal PL intensity and to a later occurrence of the PL bright point. The higher PL intensity when using a higher selenium:cadmium precursor ratio might be originating from a lower reactivity of the selenium precursor, hence resulting to an imbalance between Cd and Se atoms within the NC. Thereby, creating dangling bonds on the NC surface, especially the cationic ones (being the Cd dangling bonds in CdSe), which might not all be passivated by the surface ligands. Since cationic dangling bonds act as electron traps located within the bandgap [272], they reduce the PL intensity of the NCs by favoring non-photon-emitting exciton recombination. Thus, a higher Se precursor concentration can reduce these free Cd dangling bonds. Also the NC sizes observed from the PL peak position seem to be slightly lower for the Se-rich synthesis, which might be an indication that more nuclei are formed in the initial stage of the synthesis due to the more balanced release of Cd and Se. Also, in the following Figure 4.18 the development of the FWHM is depicted for the 3 different precursor ratios.

99 Absorption [A. U.] FWHM [nm] FWHM [nm] Results - Solar cell reproducibility Cd:Se 3:2 Cd:Se 2:2 Cd:Se 2: Time [min] Time [min] Figure Left: Development of the FWHM during hot-injection synthesis of (HDA/TOPO):(Cd-SA):(TOP-Se) at 300 C for ratios of 100:3:2, 100:2:2, and 100:2:3. The dashed lines represent an interpolation over the average results from two different syntheses by a B-spline. Right: The same FWHM development, but displayed only until ¼ of the recorded synthesis time to better visualize the development at the beginning of the synthesis. From the development of the FWHM one cannot indentify a clear difference of the PL halfwidth for the first 30 minutes. However, the NCs synthesis at an increased Se:Cd ratio tends to exhibit a shorter time of low FWHM, with the half-width already strongly increasing after the PL bright point. In contrast, NCs from an increased Cd:Se ratio tend to have a longer time of smaller PL half-width. The CdSe NC synthesis of an equal Cd:Se precursor ratio tends to have (from the average of two syntheses) the lowest FWHM in the PL bright point. The UV- Vis absorption spectrum of NCs from the bright points of the syntheses at the 3 different Cd:Se ratios (see Fig. 4.19) show only minor differences: only slight size differences are visible and a slightly stronger pronounced 1 st excitation peak from the Cd-rich synthesis is visible :3:2 100:2:2 100:2: Wavelength [nm] Figure 4.19 UV-Vis absorption spectra from CdSe NCs of different Cd:Se precursor ratios in their PL bright-point. Bold lines are the average from 2 synthesis batches shown in thin lines.

100 4. Results - Solar cell reproducibility 86 Solar cell results: Various hybrid BHJ NC/PCPDTBT solar cells were made including the NCs of the 3 different precursor ratios (according to Chapter 3.2, except a different HA washing time). In Table 4.4 the optimum found HA washing times for solar cells made from NCs of the above described batches are mentioned together with the resulting achieved average PCE values. Additionally, the optimum HA washing times and the thereby achieved PCEs from further CdSe NC batches (of 3:2 and 2:2 Cd:Se ratios) are mentioned in brackets. Table 4.4. PCEs of hybrid CdSe/PCPDTBT BJH solar cells from NCs of different Cd:Se precursor ratios together with the found optimal HA washing time. In brackets HA washing times and PCEs for further CdSe NC batches synthesized and used in hybrid NC/polymer BHJ solar cells throughout the thesis are given, for which PCEs of over 2.0% were reached. Cd:Se precursor ratio Min./max. QY [%] (of the bright point) 3:2 18.9/29.2 2:2 19.8/30.1 Optimum HA washing time [min] 26 (10-26) 23 (16-24) Average PCE [%] 2.0 ( ) 2.1 ( ) 2:3 28.5/ From the above Table 4.4 one can see that the CdSe NCs from 3:2 and 2:2 Cd:Se precursor ratios exhibit similar PL quantum yield (QY) - calculation procedure see Appendix A3 - in their bright point, and accordingly also their optimal HA washing time is similar with similar resulting PCEs. In case of using NCs from the stronger fluorescent selenium-rich synthesis, the optimal washing time is considerably longer, and the resulting PCE is slightly smaller in the direct comparison to solar cells from higher Cd:Se ratios. This shows that a higher PL quantum yield might indicate a better NC passivation by synthesis ligands if comparing different synthesis batches of different synthesis methods (one-pot, hot injection or different surfactants), but monitoring the PL quantum yield seems only to be useful for mapping the NC quality evolution within one individual synthesis. A higher NC PL quantum yield from one synthesis does not necessarily imply a better resulting solar cell performance than NCs with a lower QY from a different synthesis. However, one can say that NCs with a higher PL QY require a longer post-synthetic treatment. More important is the comparison

101 V OC [V] FF J SC [ma/cm 2 ] PCE [%] Results - Solar cell reproducibility of the size distribution of the NCs, as indicated by the results of Chapter 4.1. A conclusion that may also be drawn by comparing the PCEs of Chapter with a maximum of 1.5% - to the PCEs of up to 3% presented within this chapter, and considering at the higher FWHM values (min. 34 nm in Chapter 4.1) compared to the here presented values (min. 29 nm) resulting from the hot injection synthesis. Furthermore, from the above Table 4.4, one can see that the differences in optimal post-synthetic treatment time (for obtaining solar cells of more than 2% PCE) fluctuates even for the (presumably) same synthesis conditions by a factor of 2 or more in between different NC batches. And only small deviations from the optimal NC treatment time result in large differences in the solar cell efficiency (see Fig. 4.20). Hence, one has to establish guidelines for efficiently finding the optimal post-synthetic treatment time, using as less experiments as possible. These guidelines will be elaborated within the next subsections, and will be presented in a conclusion at the end of this chapter. 4.3 Influence of the post-synthetic NC washing time on the solar cell performance In the following, also an example of how the HA washing time influences the performance of the hybrid CdSe/polymer BJH solar cells is given in Figure Therein, the experiments for finding the optimum post-synthetic HA treatment time are presented for hybrid CdSe/PCPDTBT solar cells containing the high PL QY exhibiting CdSe NCs resulting from the 2:3 Cd:Se precursor ratio synthesis HA Washing Time [min] HA Washing Time [min] Figure Dependency of V OC, fill factor, J SC and PCE from the HA washing time at 105 C of hybrid BHJ solar cells from CdSe NCs from a 100:2:3 (HDA/TOPO):(Cd-SA):(TOP-Se) ratio wetchemical hot-injection synthesis at 300 C.

102 PL Intensity [CPS] Diameter [nm] 4. Results - Solar cell reproducibility 88 From the above figure one can also see how the HA washing time of the NCs is influencing the solar cell performance. The fill factor is steadily increasing by a longer HA washing time, probably by the stronger removal of initial synthesis surfactants, presumably leading to higher NC conductivity in the photoactive layer of hybrid solar cells. Also, the short circuit current is increasing with a longer post-synthetic NC washing, but for too long washing the J SC is decreasing, probably by increasing shortcuts throughout the active layer due to increased NC aggregation by a lower colloidal stability. This assumption is also supported by the decreasing V OC for longer HA washing times, indicating increasing number of shortcuts between anode and cathode. Also, there is a trend of decreasing series resistance and also a decreasing parallel resistance of the hybrid solar cell with increasing post-synthetic NC treatment time, which was however not intensely studied, and would need further investigation to draw solid conclusions. 4.4 Comparison of NC ligand sphere from HDA/TOPO and HDO synthesis The supposedly larger ligand sphere of CdSe QDs with higher PL QY, as described before, is in the following proven with DLS measurements. The following Figure 4.21 shows a comparison of CdSe NCs synthesized in HDA/TOPO by the prior described hot-injection synthesis and of CdSe NCs from the HDO microwave-synthesis described in the previous Chapter CdSe (HDA/TOPO) CdSe (HDO) HA Washing Time [min] HA Washing Time [min] Figure Left: Development of PL Intensity over the HA washing time for CdSe NCs synthesized in HDA/TOPO (100:3:2, 30 min at 300 C, hot-injection) and in HDO (100:1:1, 30 min at 200 C, one-pot microwave synthesis). Right: Respective development of hydrodynamic diameter for CdSe NCs measured by DLS. Black data points represent CdSe NCs synthesized in HDA/TOPO, red data points represent CdSe NCs synthesized in HDO.

103 89 4. Results - Solar cell reproducibility The initial PL intensity of the NCs from the HDO synthesis is about 8 times lower than the intensity of the NCs from the HDA/TOPO synthesis. At the same time, also the measured initial ligand sphere diameter is about 5 times lower. By washing with hexanoic acid, the PL intensity is decreasing for NCs from both syntheses. However, in contrast to NCs synthesized in HDO, where the lowest diameter of 10 nm is measured already after 5 min HA washing, and further washing is already inducing NC aggregation, for NCs synthesized in HDA/TOPO the hydrodynamic diameter is decreasing down to 18 nm after 18 min of HA washing. Furthermore, in contrast to NCs from the HDO NCs there was no NC aggregation measurable, even one day after the washing procedure. Moreover, the exponential decrease of the PL intensity of CdSe NCs from the HDA/TOPO synthesis (see Fig. 4.21, left) with a linear decrease of hydrodynamic NC diameter d (see Fig. 4.21, right), indicated that the PL intensity is proportional to the volume of the ligand sphere, which seems to scale with (d/2)³. 4.5 Influence of post-synthetic NC washing time over thermal solar cell stability Like mentioned before, conducting the post-synthetic hexanoic acid treatment for a short time results in a large remaining ligand sphere around the NCs. Meaning, that the expected colloidal stability is higher, but on the other side more organic ligands are introduced into the active layer of the solar cell. For this comparison CdSe NCs washed with HA for a shorter and for a longer time, i.e. 10 min and 20 min, were used to create hybrid CdSe/PCPDTBT BHJ solar cells. The solar cells were measured right after manufacturing, again after thermal annealing at 150 C for 5 minutes, and also after a subsequent 5 min of annealing (see Fig. 4.22).

104 4. Results - Solar cell reproducibility 90 Figure Influence of HA washing time (10 min and 20 min) on the J SC, V OC, FF, PCE, R P and R S of CdSe/PCPDTBT BHJ solar cells during annealing at 150 C. The used NCs are from a hotinjection synthesis for 30 min at 300 C using HDA/TOPO surfactants and a molar ratio of 100:3:2 between surfactant, Cd-precursor and Se-precursor. The longer HA treated NCs resulted in a 4.5 times higher initial short-circuit current, which is however assimilating with thermal annealing, maybe due to desorption of synthesis ligands from the NCs. The open circuit voltage is nearly the same in both cases. However, the development of the V OC under thermal treatment of the solar cell is inverse for short NC washing compared to long NC washing. The V OC of solar cells made from 10 min HA washing time is increasing from about 0.6 V to about 0.7 V after 5 min at 150 C, while the V OC of solar cells from 20 min HA washed CdSe NCs is strongly decreasing to about 0.4 V. The fill factor of the solar cell from longer washed NCs is already as high as 0.55 before annealing, but it is strongly decreasing after annealing. On the same time the FF is constantly increasing during annealing for hybrid solar cells made from shorter washed NCs. This increase of FF is mainly due to a strong decrease in series resistance (R S ) during the thermal annealing. This decrease of R S might be a second prove for the desorption of synthesis ligands from the NCs besides the before mentioned increase of J SC. The parallel resistance (R p ) is more than double as high for solar cells containing 10 min washed NCs. However, it is decreasing in both cases by annealing, caused probably by thermally induced NC aggregation that might increase shunts between anode and cathode. In summary, hybrid BHJ solar cells containing longer washed

105 91 4. Results - Solar cell reproducibility NCs exhibit an initial high PCE that is strongly decreasing upon thermal annealing, in contrast solar cells including NCs that received a shorter post-synthetic treatment show an increasing PCE within the first minutes of the thermal treatment. However, solar cells containing shorter HA washed NCs, which exhibit a higher stability under heat, perform worse at room temperature in long-term stability tests, which are described and discussed in Chapter Conclusion In conclusion, it was demonstrated that with highly homogenous QDs also high power conversion efficiencies are possible for hybrid BHJ QD/polymer solar cells reaching PCE values of 2-3%. A higher PL quantum yield of the QDs does not principally result in higher PCEs given a similar size distribution. In fact, the QY seems to influence the duration (or intensity) of the optimal post-synthetic treatment. This finding is consistent with the logical conclusion of Frank-Stefan Riehle [16] that NCs of a higher PL quantum yield posses a larger NC ligand sphere, when assuming that a larger ligand sphere also needs a longer post-synthetic treatment time to be reduced. Also, it is interesting to see that a short post-synthetic NC treatment results in solar cells of a high open circuit voltage (as long as the washing time is not much too short), but a low short-circuit current. Which is indicating that longer postsynthetic treatment leads besides the reduction of the NC ligand sphere also to an induction of NC surface traps, which are reportedly leading to decreased V [206] OC. On the other hand a too long post-synthetic treatment also results in a low short-circuit current, and a low open circuit voltage. After a short thermal annealing one can clearly indentify the solar cell containing shortly washed NCs by its increasing J SC, while the J SC of the solar cell containing longer washed NCs is decreasing. Also the fill factor is an indication of the state of the NCs ligand shell. The solar cell s FF is constantly rising for a longer post-synthetic NC treatment time; and a long washing results in high initial fill factors, which are however decreasing faster during thermal annealing. Thus, it is possible from the initial solar cell measurement and the solar cell performance after a short thermal treatment to conclude in which direction from the optimal post-synthetic treatment time the incorporated NCs deviate. This allows for decreasing the number of experiments and material needed for finding the optimal post-synthetic NC treatment time.

106 4. Results - Solar cell reproducibility 92 Nevertheless, an optimum NC synthesis would always produce the same NCs of the same high quality, and necessitating exactly the same post-synthetic treatment. Therefore, ideally, one would combine the first attempted one-pot microwave-assisted synthesis method with the possibility of in-situ synthesis monitoring described in Chapter 4.1 with the HDA/TOPO synthesis method presented in the subchapters thereafter. This is in principle possible, since F.-S. Riehle recently reported that also nearly monodisperse CdSe QDs can be synthesized in a controlled way by the one-pot synthesis method [16].

107 93 5. Results - Efficiency enhancement 5. Results - Efficiency enhancement Within this chapter three attempts towards efficiency enhancement of hybrid NC/polymer heterojunction solar cells are described. The first attempt is the use of elongated CdSe NCs for a more efficient charge extraction. The second attempt is based on the attachment of CdSe NCs to graphene, which should also provide a more directed charge extraction. And finally, the third presented attempt is based on an interdigitated donor-acceptor structure, which should allow efficient exciton dissociation and charge extraction. 5.1 Elongated CdSe nanocrystals for BHJ solar cells Elongated CdSe NCs, so called nanorods (NRs), are used in hybrid NC/polymer solar cells in the attempt to enhance the electron extraction [119, 111, 120, 207, 132, 113, 11]. In the year 2002 Huynh et al. [119] were the first reporting an enhanced PCE upon use of CdSe NRs over QDs in NC/polymer BHJ solar cells. Moreover, the solar cell power conversion efficiency improved with increasing NR length up to 1.7%. Like depicted in Figure 5.1, the idea behind is that by using NRs the electrons can be extracted faster out of the active layer, since less hopping events [28] between individual NCs are needed, and thus the electron transport can occur for longer distances inside the NC. Figure 5.1. Scheme of electron extraction along nanorods (left) compared to quantum dots (right): less hopping events are needed in the first case for charge extraction to the top electrode, given the assumption that NRs might be partially orientated along the vertical axis. At the time the following described experiments were conducted, the highest published PCEs of hybrid BHJ solar cells were reached with 2.4% [120] upon use of CdSe NRs. The PCEs of QD containing cells were lower with a maximum PCE of 2.0% [15], a record at that time held by our group. Thus, we believed that utilizing NRs has the potential to result in higher solar cell efficiencies than the ones reached by us with QDs.

108 5. Results - Efficiency enhancement 94 The attempt within this thesis was to use the hexanoic acid treatment - which was applied to QDs for removing access NC synthesis ligands [15] - to NRs and thereby improve the solar cell PCE by the previously described geometrical advantages of NRs. The nanorods for these experiments were provided by Bayer Technology Services (BTS) GmbH and were synthesized according to the method of Liu et al. [208] by using a mixture of TDPA (tetradecylphosphonic acid), trioctylphosphine oxide and trioctylphosphine as surface ligands by a hot injection method at 300 C. The utilized NRs had a diameter of 7-8 nm and a length of nm, and were provided untreated and dispersed in toluene. A TEM image of the NRs utilized in the initial experiments is provided in the lower left inset of Figure 5.2. Furthermore, Figure 5.2 shows a TEM image of the active layer of a NR/polymer solar cell. Figure 5.2. TEM image of CdSe NRs (NR 153) in toluene (lower left inset); TEM image of active layer of BHJ NR/polymer solar cell spun cast from min hexanoic acid washed NRs (NR 153) and PCPDTBTT in a weight ratio of 6.5:1 and a NR concentration of 18.5 mg/ml in a 97.5:2.5 vol% mixture of DCB and DIO, annealed for 30 min at 145 C. The hexanoic acid treatment, as described by Zhou et al. [15], was applied to the prior mentioned NRs; and solar cells like described in Chapter 3.2 were built. However, the obtained PCE of the thereby obtained CdSe NR/P3HT solar cells with chlorobenzene as the common solvent was of only 0.005% (resulting from two independent experiments), compared to 2.1% obtained by Zhou et al. with QDs [115]. The HA washing step was carried out with a NC concentration of 0.2 mg/ml for 5 min at 105 C, like found to be optimal in case of HDA/TOPO capped QDs. Therefore, further optimization of the post-synthetic NC treatment and further improvements of the NC/polymer ink composition needed to be performed.

109 95 5. Results - Efficiency enhancement Strategies for efficiency improvement The most obvious measure to improve the PCE is the increase of the post-synthetic hexanoic acid NC treatment time to reduce the original NC ligand shell, end hence to improve the conductivity of the NCs. However, this is not the only here applied method to increase the solar cell performance. The effect of adding pyridine to the NC/polymer blend was already described to improve solar cell efficiency by an improved NC dispersibility. Thereby, the intermixing between NCs and polymer is improved, which is leading to a higher donoracceptor interfacial area. However, the pyridine content cannot be increased over vol%, as it is also reported to decrease the solar cell performance by precipitation of the polymer in that case P3HT [119]. Furthermore, the use of higher boiling point solvents is reported to result in higher PCEs due to a stronger vertical phase segregation, leading to a higher NC content near the top electrode, thus reducing shunts throughout the active layer [131], an effect also observed for pure organic PCBM/polymer solar cells [209]. The slower evaporation allows for an increased fraction of π-π stacking along the polymer chains, leading for P3HT to even nano-fibrous structures [120]. Also, using slower evaporating solvents is regarded to be generally beneficial due to resulting in more stable, thermodynamically favored blend structures [131]. In the following sections also polymers with a bandgap lower than for P3HT are used in order to be able to absorb a higher fraction of the solar spectrum (see Fig. 5.3), thus leading to higher short-circuit current densities. Figure 5.3. Absorption spectra of the conjugated polymers P3HT, PCPDTBTT and PCPDTBT compared to the AM 1.5G sun spectrum between 300 nm and 1000 nm.

110 5. Results - Efficiency enhancement 96 Also, a linker molecule for a better interconnection of the polymer, i.e. 1,8-diiodooctane (DIO), was used as additive within the NC/polymer blend. Thereby, Lee et al. [210] could demonstrate for PCPDTBT/PC 71 BM solar cells upon addition of 2.5 vol% DIO in CB a PCE enhancement from 3.4% to 5.1% Variating the blend solvent The following Table 5.1 shows the power conversion efficiencies of BHJ CdSe NR/P3HT solar cells obtained by increasing the NC HA washing time up to 150 min. Additionally, also solar cells using the higher boiling point solvent DCB and using pyridine as co-solvent were conducted. Table 5.1. Power conversion efficiencies of NC/polymer BHJ solar cells for different NC solvents and different hexanoic acid washing times of NRs. The solar cells included NRs with a 3:1 length to diameter ratio and a 87:13 weight ratio of NRs to P3HT in the active layer. Efficiencies are average values from 3 solar cells after post aluminum-deposition annealing for 10 min at 145 C. NR solvent Hexanoic acid washing time 5 min 60 min 120 min 150 min PCE [%] CB CB +2.5% Pyridine DCB DCB +2.5% Pyridine From the experiments summarized in the above Table 5.1 one may conclude that longer washing times in hexanoic acid are leading to higher PCEs, as well as the addition of pyridine as co-solvent (2.5 vol% of total NC/polymer blend). Moreover, the use of the higher boiling point solvent dichlorobenzene (DCB) is also leading to increased performance.

111 Absorption [A. U.] Results - Efficiency enhancement In the conducted experiments NRs displayed a superior dispersibility upon addition of pyridine in the DCB solvent. What is visible from UV-Vis absorption measurement (Fig. 5.4, left) and from DLS measurements (Fig. 5.4, right). From 5 consecutive DLS measurements, the average measured hydrodynamic diameter measured from a NR dispersion in DCB was of nm. However, when adding 2.5 vol% of pyridine the measured diameter decreased to nm. Moreover, the diameter values remained stable during the approximately 10 min of the 5 subsequent size measurements performed for the NRs solved in the DCB/Pyridine mixture. In contrast, for the NRs solved in DCB the measured size steadily increased from nm to nm during the measurements, indicating a continuous NC aggregation DCB DCB+Pyridine Wavelength [nm] Figure 5.4. Absorption (left) and diameter distribution measured by dynamic light scattering (right) for NRs after hexanoic washing, solved in DCB and in a DCB/pyridine mixture. Therefore, one can conclude from absorption and DLS measurements, that by addition of pyridine the dispersibility of the NRs and the stability of the dispersion against NC aggregation is increased. Interestingly, these characteristics also have a positive effect of the resulting solar cell performance, like seen from Table Low-bandgap polymer as donor material Despite the better solar cell performance when carrying out the post-synthetic treatment for longer times, the HA washing time cannot be further increased. This is because HA evaporates for longer washing times, leading to the precipitation of NCs on the wall of the

112 5. Results - Efficiency enhancement 98 flask the treatment is conducted in. Hence, a different approach was introduced by repeating the hexanoic acid washing of NRs for a second time after precipitating them with methanol. Moreover, now the lower bandgap polymer PCPDTTBTT, allowing for the absorption of longer wavelength photons (see Fig. 5.3) was used instead of P3HT, which should lead to a higher exciton generation. Table 4.2. Power conversion efficiency of NR/PCPDTTBTT solar cells after 10 min annealing at 145 C with NRs being treated by hexanoic acid for 60 min, respectively min at 115 C. Hexanoic acid washing time NR solvent DCB Polymer 60 min min PCE [%] DCB +2.5% Pyridine PCPDTTBTT CB Indeed, by the use of PCPDTTBTT for a HA washing time of 60 min the PCE is more than doubled to 0.077%, compared to 0.032% achieved with P3HT. Moreover, when repeating the HA treatment for another 60 min - after a preceding NR precipitation by methanol the PCE could further be increased to 0.128%. Also, the addition of pyridine was investigated, but resulted even in a slight decrease of PCE instead of showing the strong improvement observable in the P3HT/NR system. It was also investigated whether the use of DCB instead of CB also results in higher PCE for the PCPDTTBTT [211] /NR system like it proved to be if using P3HT - and indeed using CB as solvent resulted in a slightly lower PCE of 0.11% Use of polymer chain interlinking additive Now, after concluding that the PCPDRTBTT/NR system is providing higher PCEs without addition of pyridine, using DCB as solvent, and by repeating the HA washing for a second subsequent time, the first HA treatment time was further increased to 150 min and subsequent HA washing steps of 30 min were introduced. Additionally, 1,8-diiodooctane (DIO) was added to the NC/polymer blend, which like described in Section is reported

113 99 5. Results - Efficiency enhancement to increase the charge extraction over the polymer. The results of these investigations are displayed in the following Table 5.3. Table 5.3. NR/PCPDTTBTT solar cells from min and min HA treatment at 115 C, with different fractions of DIO in DCB, with the same NR concentration in the resulting blend. Mentioned are PCEs after 10 min (in brackets) and after 30 min solar cell annealing at 145. HA washing NR solvent: DCB DCB +2.5% DIO DCB +5.0% DIO DCB +7.5% DIO DCB +10% DIO (0.594) 2x PCE [%] 3x - (0.762) (0.478) (0.772) (0.230) (0.254) (0.167) Indeed, the PCE is increasing by over 50% from 0.625% to 0.987% when adding 5 vol% of DIO into the DCB/NR-PCPDTTBTT dispersion. The improvement is mainly due to an increase of J SC, but also of V OC. Hence, the enhancement can be assigned to an improved charge extraction and reduced recombination; which is the expected effect of the improved polymer chain interconnection when utilizing the DIO additive [210]. Using a higher fraction of DIO of 10% resulted in a strong decrease of solar cell efficiency. Also, a third repetition of the HA washing for 150 C resulted in lower PCEs, probably due to the observed increased NR aggregation which impeded the redispersion of the NRs in DCB Removal of NC aggregates Now, three further supposed improvements were introduced. First, instead of DCB the even higher boiling point solvent trichlorobenzene (TCB) was utilized. Also, the even lower bandgap polymer PCPDTBT (see Fig. 5.3) and an additional centrifugation step were introduced. The centrifugation is conducted after dispersing the NRs in TCB, for removal of aggregated NCs from the dispersion representing a kind of filtration procedure. Therefore, the NR dispersion is centrifuged for 20 s at 4 krpm, and subsequently the NRs still in solution are being decanted. For being able to determine the correct remaining NR concentration,

114 5. Results - Efficiency enhancement 100 from the decanted NR dispersion 5% are taken as aliquote for an absorption measurement. Prior to the absorption measurement a concentration series was recorded, showing a linear behavior between concentration and absorption at the 1 st excitonic peak, showing however different slopes for different NR batches. So, one can conclude from the absorption at the 1 st excitonic peak of the NR aliquot on the NR concentration in the decanted NR dispersion. Usually, by this step 40%-70% of the NRs were lost. As mentioned in the following Table 5.4, NR/PCPDTBT solar cells using TCB as NR solvent resulted in 0.908% PCE for 2x HA washing without centrifugation. When applying the centrifugation step, the average PCEs increase to 1.53% for 2x HA washing and still 1.17% is reached for 1x HA washing. However, the addition of pyridine results in a decreased PCE, like also use of DIO was resulting in a strongly decreased PCE when added to TCB. Therefore, one may conclude, that the use of DIO is either not useful when using TCB as solvent, or is not beneficial for PCPDTBT containing solar cells. The later assumption however has already been proved being wrong like mentioned before - for PCPDTBT/C71-PCBM solar cells. Table 5.4. Average efficiencies from NR/ PCPDTBT solar cells, containing NRs treated for 150 min respectively min at 115 Cwith hexanoic acid with and without a final centrifugation filtration and dispersed in different solvent mixtures. Mentioned are PCEs after 10 min (in brackets) and after 30 min solar cell annealing at 145. HA washing NR solvent: TCB TCB +centrifugation TCB +5% Pyridine TCB +5% DIO 1x 2x PCE [%] (0.511) (0.681) (1.167) (1.440) (0.536) (0.022) In order to support the assumption of removing NR aggregates by the centrifugation procedure, atomic force microscope (AFM) measurements in tapping mode were conducted by a Veeco Multimode AFM on the surface of two NR/polymer solar cells, one built from a NR dispersion on which no centrifugation step was applied to and one on which the centrifugation procedure was performed. The resulting height profiles of the two cells are presented in Figure 5.5.

115 Results - Efficiency enhancement Figure D and 3D height profiles from 3x3 µm surfaces of CdSe NR/PCPDTBT solar cells, recorded by an AFM (Veeco Multimode AFM) in tapping mode. The left topography is recorded for a solar cell for which no centrifugation procedure was conducted on the NR dispersion, the right topography is recorded from a solar cell on which the centrifugation step has been applied. From the AFM data one can clearly observe, for the solar cell on which the centrifugation procedure has not been conducted, among fine grained areas also strongly elevated areas, which are probably corresponding to NR aggregates. On the other side, for the solar cell on which the centrifugation step was carried out, such aggregates are very rarely detectable. Calculations of the surface roughness result in higher values for the solar cell on which the filtration by the centrifugation procedure has not been conducted (see Table 5.5). Table 5.5. Arithmetic average roughness (R a ) and root mean squared roughness (R RMS ) measured from AFM measurements over an area of 3x3µm and of 10x10µm. 3x3µm 10x10µm R a [nm] R RMS [nm] R a [nm] R RMS [nm] no centrifugation with centrifugation

116 5. Results - Efficiency enhancement Inter-batch differences All experiments presented so far have been conducted with the same NR batch (NR 153) provided by BTS in order to allow for comparing the obtained experimental results. However, different NR batches are exhibiting different dispersivities, what can be firstly observed by the fact that some provided NR batches are already showing precipitations in the toluene solution. Also, different NR batches require a differential post-synthetic treatment in order to obtain best solar cell results with the individual NR batch (see Table 5.6). Table 5.6. Best CdSe QD/PCPDTBT solar cell results with different NR batches and the found optimum found post-synthetic treatment procedure with hexanoic acid (HA). For the batches NR 153 and NR 258 2x HA treatment resulted in best results, whereas foe batch NR 263 a simple HA washing led to the best solar cell performance. NR batch 1 st HA treatment 2 nd HA treatment J SC [ma/cm 2 ] V OC [V] FF PCE [%] NR min 60 min NR min 30 min NR min 70 min min As can be seen from the above Table 5.6, there are batches that require longer HA treatment times (i.e. NR 258) and batches that require considerable shorter post-synthetic treatment time (i.e. NR 263) in order to reach the optimum solar cell performance. Interestingly, one can extract from the above table, that NRs which require a shorter HA treatment time also exhibit the best solar cell performance after solar cell integration due to higher V OC and FF.

117 Results - Efficiency enhancement Summary and conclusion of nanorod/polymer solar cell experiments In following Table 5.7 the values for the best measured pixels of solar cells from each of the three utilized polymer types are listed. For the PCPDTBT containing solar cells, additionally the best result utilizing the previously described centrifugation step is also mentioned. Table 5.7. Average values of shortcut current density (J SC ), open circuit voltage (V OC ), fill factor (FF), and power conversion efficiency (PCE) of the 9 best solar cells (on 3 different substrates) of BHJ NR/polymer devices incorporating each polymer type (P3HT, PCPDTBTT, and PCPDTBT). For the P3HT containing solar cells only 1x washing was utilized. Polymer J SC [ma/cm 2 ] V OC [V] FF PCE [%] P3HT (only 1x HA) PCPDTTBTT PCPDTBT PCPDTBT (+ centrifugation) The direct comparison of the solar cells without centrifugation step, devices from PCPDTTBTT (with DIO additive) and PCPDTBT reach nearly the same J SC and FF, only the V OC is slightly higher upon utilization of PCPDTTBTT. This higher V OC might be described by the 0.08 ev lower lying LUMO level of ev compared to ev for PCBDTBT reported by BTS. By conducting the centrifugation procedure the considerable higher PCE is resulting from a strong increase in J SC and an increasing FF as well. In summary, considerably longer hexanoic treatment is needed for NRs then for QDs. The optimal treatment time was increased from 5 min for QDs to 150 min plus an additional 30 min second HA washing step after a preceding NC precipitation. Moreover, the highest average PCE of 1.533% was reached when using the low bandgap polymer PCPDTBT and conducting a centrifugation step for removal of NR aggregates from the final NR/polymer blend. In the following Figure 5.6 the evolution of the PCEs throughout the previously discussed experiments is represented in a graph, and current-voltage characteristics of the

118 PCE [%] Current density [ma/cm²] 5. Results - Efficiency enhancement 104 best solar cells obtained from use of the three polymers P3HT, PCPDTTBTT and PCPDTBT are shown as well P3HT 1x HA PCPDTBTT 1x HA 2x HA PCPDTBT 1x HA 2x HA 1x HA + centr. 2x HA + centr HA washing time [min] NR/P3HT NR/PCPDTTBTT NR/PCPDTBT Voltage [V] Figure 5.6. Left: Power conversion efficiencies of BHJ NR/polymer solar cells using three different polymers (i.e. P3HT, PCPDTTBTT, PCPDTBT) without additive, different postsynthetic washing times in hexanoic acid (HA), one (1x HA) or two (2x HA) HA washing steps and an eventual added final centrifugation step for removal of aggregated NRs. Right: Current-density vs. voltage characteristics of the best solar cells obtained from use of the three polymers P3HT, PCPDTTBTT and PCPDTBT. With respect to the finding that NRs need longer post-synthetic treatment time for resulting in best performing BHJ hybrid solar cells an observation published by Schädel et al. [212] in 2012 might be supportive. He measured in cooperation with our group that the remaining ligand sphere thickness, after applying the same hexanoic acid treatment, is of a factor 3.6 larger for NRs compared to QDs. This the finding that indicates that ligands on NRs are harder to remove, what might be explained by a higher density of the ligands as they are supposedly oriented parallel to each other in the direction along the NR, compared to the exclusively radial orientation on the spherical quantum dot (see Fig. 5.7). Figure 5.7. Supposed orientation of the first ligand layer around colloidal NRs (left) and QDs (right). The long aliphatic tails of the NC ligands (drawn as lines perpendicular to the NC surface) align to each other through Wan-der-Waals forces.

119 Results - Efficiency enhancement Use of nanorod/quantum dot mixtures for hybrid BHJ solar cells The highest CdSe NR/polymer solar cell efficiency reached and presented in the above section of about 1.9% is lower than the top efficiency of 2.7% reached in our group in the meanwhile by use of CdSe QDs blended with the same PCPDTBT polymer [125]. Nevertheless, the addition of NRs to the QD/polymer cell may in principle improve the electron extraction of the high performance QD/polymer solar cell by improved charge extraction. Especially, considering, that a mixture between NRs and QDs might elevate some NRs from their predominately horizontal orientation within the active layer found by Hindson et al. [129]. Therefore, now solar cells with NR/QD/polymer mixtures were investigated. The QDs utilized in this experiment were synthesized in the presence of TOP and OA as ligands in a microfluidic reaction chamber similar to the one presented by Chan et al. [213]. The utilized quantum dots exhibited an average size of 4.7 nm (see Fig. 5.8, right) and were treated by a hexanoic acid washing at 105 C for 10 min by my coworker Yunfei Zhou. Figure 5.8. TEM images (taken by Bayer Technology Services GmbH) of the utilized CdSe NRs (left), and CdSe QDs (right) utilized for the NR/QD/polymer solar cells. The NRs were taken from batch NR 263 and a HA washing for 50 min at 115 C was applied to them. QD and NRs were mixed in the common solvent CB, representing a compromise, as NR based solar cells performed better if DCB was used, however QDs were showing the opposite behavior. Hence, CB was used as nanocrystal solvent for the NR/QD mixtures for avoiding the reduction of the high QD solar cell based performance on the downside of a reduced performance of NR only devices. In the following Figure 5.9 the results of the four executed experimental series of QD/NR/PCPDTBT mixtures are given.

120 J SC [ma/cm²] V OC [V] FF PCE [%] 5. Results - Efficiency enhancement NR [wt%] NR [wt%] NR [wt%] NR [wt%] Figure 5.9. Shortcut current (J SC ), open circuit voltage (V OC ), fill factor (FF), and power conversion efficiency (PCE) of BHJ nanocrystal/pcbdtbt solar cells with different QD:NR ratio from 4 different experimental series (indicated by different colors) performed using NRs and QDs from the respectively same synthesis batches. The graphs mention the wt% of NRs of the total NC weight. At 0 wt% NRs 100% QDs, and at 100 wt% NRs 0 wt% QDs are present in the NC/polymer cell. For the NCs CB was used as solvent, whereas TCB was used as polymer solvent. The utilized NC to polymer ratio was of 7:1. There are two noticeable improvements upon the integration of NRs into the NC/polymer blend. One is the increase of shortcut current (J SC ), with a maximum average gain of 21% at a weight ratio of 66:33 QD:NR. The other improvement of the QD/NR mixture is the increase of open circuit voltage (V OC ) with a maximum average gain of 6.4% at projected 72 wt% NR content. The fill factor (FF) is improved compared to a pure NR/polymer solar cell and is not declining much under the initial value of a QD/polymer solar cell until very high NR contents. In contrast, in case the QD/polymer solar cell is showing a lower FF than the NR/polymer solar cell, the QD solar cell does not already profit from small addition of NRs, the improvement is rather linear than exponential. Summing up, the PCE of QD:NR/polymer solar cells is mostly benefitting from the increased J SC, exhibiting an average increase of 21.2%, with power conversion efficiencies of up to 2.86%. The QD/NR mixture experiments were jointly executed with Yunfei Zhou who contributed the HA treated QDs and myself who contributed the HA treated NRs. The results of the best experimental series were incorporated into a publication [115] about the benefit of using larger QDs and QDs/NR mixtures in BHJ solar cells through a supposed resulting improved nano-composition. Shortly afterwards Jeltsch et al. [130] also published experiments about QD/NR mixtures, using the same polymer and NCs, also provided by BTS, but on which a pyridine ligand exchange has been applied. Therein, the QD/polymer solar cell resulted in

121 Results - Efficiency enhancement an average PCE of 2.42%, the NR/polymer device resulted in a PCE of 1.72% and the mixture of QD/NR/polymer resulted in an average PCE of 3.51%. Interestingly, the best performance resulted from the inverse QD:NR weight ratio (i.e. 33:66), compared to the optimal ratio found in our experiments. Like also found by our experiments J SC strongly increased and the FF showed only very little improvement. However, it is reported that the V OC was not profiting from the QD/NR mixture, unlike observed by our experiments. The dissimilarities might derive from the different composition of the active layer induced by the differently treated QDs. While pyridine treated QDs which form strongly aggregated areas of QDs spread within the hybrid active layer of the solar cell, our hexanoic acid treatment results in a more homogenous QD distribution due to a better dispersibility of the treated QDs. Thus, one might conclude that trap-assisted recombination is reduced for HA-treated QDs, hence [52, allowing a higher V 204] OC. However, in both cases the PCE increase is supposedly mainly attributed to an improved charge extraction by a better interconnection of the QDs by the NRs (respectively a better interconnection of the NRs by the QDs), leading to a higher J SC. In respect of the alignment of NCs in the NR:QD mixture it was speculated by Zhou et al. [115] that the addition of QDs might elevate one end of a NR from its predominately horizontal position (see Fig. 5.10b). Jeltsch et al. mentioned a presumably better interconnection of the in plane aligned NRs (see Fig. 5.11a). Indeed, it is reported by Barnes et al. and Adams et al. [214, 215] that in mixtures of colloidal spheres and rods, the two alike species tend to align with each other, meaning that the result might most likely be µm scale phases of aligned NRs connected by QD phases on all sides. But indeed within the individual NR phases there might be some in which the NRs are not aligned horizontally (see Fig. 5.10c). However, all three models depicted in Figure 5.10 remain speculative, as no study of the real NC alignment was carried out by for example TEM tomography. Figure Different models of NR and QD orientation in binary mixtures a): QDs act as interconnection for horizontally aligned NRs. b): QDs are elevating NRs from their horizontal alignment. c) NRs align with respect to each other; however, the orientation of the ordered NR domains is disturbed by QDs leading to non horizontal orientation of some NR domains.

122 5. Results - Efficiency enhancement CdSe NC/Graphene hybrid for BHJ solar cells (Parts of this chapter are included in the paper: Improved Efficiency for Bulk Heterojunction Hybrid Solar Cells by utilizing CdSe Quantum Dot - Graphene Nanocomposites by M. Eck et al., published at Phys. Chem. Chem. Phys., 2014, 16, ) In this chapter the usage of CdSe NCs attached to reduced graphene oxide (rgo) flakes - a hybrid material successfully developed in our group [20] - into the photoactive layer of hybrid BHJ solar cells is described. By the usage of graphene as a more direct charge percolation pathway within the active layer of the solar cell, a better charge extraction is expected (see Fig. 5.11). Thus avoiding relying solely on the electron extraction over NC nearest neighbor hopping [28], which is limited due to NC surfactants. Figure Concept of using graphene sheets (dark gray) decorated with CdSe QDs for an improved electron extraction from the active layer of a hybrid BHJ NC/polymer solar cell. In purely organic BHJ solar cells there are several reports upon utilization of graphene as electron acceptor in the donor-acceptor blend [ ] or as electron extraction layer when decorated with PCBM (Phenyl-C61-butyric acid methyl ester) [219]. These reports resulted only in relative improvement of purely organic BHJ solar cells, and usually could not reach PCEs in the range of state of the art cells, with the exception of physical inclusion of graphene QDs (small graphene flakes of around 5-15 nm in diameter) [218]. However, shortly before submitting this thesis a promising result about the integration of graphene in hybrid NC/polymer BHJ solar cells was published by Tong et al. [220], describing a PCE of 1.4% using CdTe tetrapod decorated oleylamine functionalized rgo, blended within the polymer matrix. This result could even be improved to a 3.3% PCE by using type II CdTe/CdSe TPs which are claimed to improve the exciton separation. Nevertheless, the investigations within this

123 Results - Efficiency enhancement chapter describe the first time that CdSe QD-decorated graphene nanocomposites were introduced into the active layer of hybrid BHJ solar cells using a thiol functionalized rgo- CdSe QD hybrid material. This approach was applied on CdSe QD/PCPDTBT solar cells for which PCEs of up to 3% could be reached (see Chapter 4.2). At first, the created rgo-cdse hybrid material is characterized with TEM, UV-Vis spectroscopy, PL spectroscopy, and electron paramagnetic resonance spectroscopy, to prove the rgo decoration. Subsequently, the morphology of the created BHJ hybrid rgo/qd/polymer layer was characterized with AFM, SEM, and TEM (including TEM tomography), to elucidate the induced differences. And finally the solar cell performance and charge extraction of QD/polymer and rgo-qd/polymer BHJ solar cells were compared TrGO-CdSe hybrid material Thiolated reduced graphene oxide (TrGO) decorated with CdSe QDs was obtained according to a procedure recently developed in our group [20]. Its formation is described in Chapter and consists, in brief, by the creation of graphite oxide (GO) from graphite, the simultaneous reduction and thiol functionalization of GO to TrGO, and the subsequent decoration with CdSe NCs. A recorded TEM picture of the CdSe QD-TrGO hybrid material is depicted in Figure 5.12, showing a large TrGO flake, densely decorated with HA washed CdSe QDs. Figure Left: TEM image of a TrGO flake decorated with CdSe QDs, from a hot-injection synthesis at 300 C with a molar HDO/TOPO:Cd-SA:TOPSe ratio of 100:2:2. Upper right inset: zoom in on the same CdSe QD-TrGO hybrid revealing the individual QDs. Right image: Schematic of the CdSe QD-TrGO hybrid material. (Figure taken from Ref. [21] - reproduced by permission of the PCCP Owner Societies).

124 5. Results - Efficiency enhancement 110 Upon the decoration on TrGO the CdSe NCs no changes in the PL or UV-Vis spectra are indicating any change in the bandgap size of the CdSe QDs (see Fig. 5.13). However, a strong quenching of the CdSe fluorescence occurs (see Fig. 5.13, right). A decreased PL intensity to 1/3 of the original value of the CdSe QDs was measured, when mixing TrGO with CdSe to form the hybrid material, what might be already an indication for charge or energy transfer from the CdSe QDs to the reduced graphite oxide. Additionally the transmission spectrum of a TrGO solution of the same concentration without CdSe NCs was measured, to show that the strong PL quenching is not originating from the absorption of the TrGO, which is of less than 10% in the wavelength-range of the CdSe phosphorescence (see right inset in Fig. 5.13). Figure Left: Transmission spectra of a 66 µg/ml dispersion of CdSe NCs in CHCl 3 and of a mixture between CdSe of the same concentration and TrGO of 0.66 µg/ml DMF. Right: PL spectra of the same two solutions directly after mixing. Right inset: Transmission spectrum of TrGO dispersion in DMF with a concentration of 0.66 µg/ml. (Figure taken from Ref. [21] - reproduced by permission of the PCCP Owner Societies). Besides the strong PL quenching indicating an energy or electron transfer from CdSe to TrGO, a direct chemical binding of CdSe QDs to TrGO could be proven by electron paramagnetic resonance (EPR) spectroscopy (performed by Emre Erdem and Suyan Tue, Institute of Physical Chemistry, University of Freiburg). Therein, it is observed that the EPR signal, arising from unpaired electrons and deriving from defects originally present in TrGO, is completely quenched in the CdSe-TrGO hybrid material. This can be explained by the formation of a chemical bond between TrGO and CdSe QDs. Further details about the prove of chemical attachment and electronic coupling between CdSe QDs and TrGO are given in two recently published papers by Pham et al. [221, 221]. Moreover, similar to this finding,

125 Results - Efficiency enhancement Lightcap and Kamat recently reported an electron and energy transfer from photoexcited CdSe QDs towards rgo [222] TrGO-CdSe/polymer hybrid film The morphology of the active layers out of CdSe/PCPDTBT and TrGO-CdSe/PCPDTBT was investigated by transmission electron microscopy (TEM), scanning electron microscopy (SEM), atomic force microscopy (AFM) (Figure 5.14a-c). The TEM images were obtained by dissolving the PEDOT:PSS layer in water and collecting the active layer on a TEM grid. Figure TEM images of the active layer of a CdSe/PCPDTBT (upper image) and of a TrGO- CdSe/PCPDTBT solar cell (lower image). (b) SEM surface images of the active layer of a CdSe/PCPDTBT (upper image) and of a TrGO-CdSe/PCPDTBT solar cell (lower image). (c): AFM topographical images of the active layer surface of a CdSe/PCPDTBT (upper image) and of a TrGO-CdSe/PCPDTBT solar cell (lower image) recorded in tapping mode. (Figure taken from Ref. [21] - reproduced by permission of the PCCP Owner Societies). The TEM images (see Fig. 5.14a) reveal in case of the CdSe/PCPDTBT sample a relatively homogenous distribution of the NCs within the active layer in the nanoscale in the x-y plane. For the TrGO containing cells, in contrast, a coarse organic-inorganic phase segregation is observed. SEM images (see Fig. 5.14b) reveal a rough surface in case of TrGO integration,

126 5. Results - Efficiency enhancement 112 showing elevated, hill like structures distributed at the same distance from each other like the aggregations found from the TEM images. From AFM micrographs (see Fig. 5.14c) one can find that the surface of the CdSe/polymer solar cells is much smoother then the surface of the CdSe-TrGO polymer solar cells. One can determine for an area of 2x2 µm² an average roughness (R a ) for the TrGO containing cells of R a =6.7 nm, while the CdSe/polymer solar cells show a much lower roughness with R a =1.15 nm. To get deeper insight into the three-dimensional morphology of the active layers, an electron tomography analysis, visualizing the spatial distribution of the CdSe QDs, was exemplarily performed (see Fig. 5.15). Experimental details are described in Section Figure (a) Top-view with 150x150 nm in the horizontal x-y plane of the calculated 3D reconstruction, obtained from TEM tomography measurements for the active layer of a CdSe QD/PCPDTBT and of a CdSe QD-TrGO/PCPDTBT solar cell. (b) Vertical x-z-cut (x=150 nm, y=150 nm, z=80 nm) through 3D reconstructions of the respective active layers obtained by TEM tomography. The bright regions correspond to the inorganic volumes filled with CdSe. The analysis of TEM slices parallel to the active layer confirmed the coarser phase-separation of the TrGO-containing solar cell (see Fig. 5.15a). Furthermore, electron tomography also revealed that the QD-TrGO hybrid nanocomposites are not randomly distributed in the film. Slices through the x-z plane (see Fig. 5.15b) provide evidence for the formation of structures

127 Results - Efficiency enhancement resulting in an improved interconnection of electron-transporting domains along the z-axis within the photoactive layer of CdSe QD-TrGO hybrid solar cell TrGO-CdSe/PCPDTBT hybrid solar cell performance Bulk heterojunction hybrid CdSe-QD/PCPDTBT solar cells, and BHJ CdSe QD-TrGO/PCPDTBT cells were prepared according to the description given in Chapter 3.2, using the solar cell design presented in Figure 3.5. At first, the optimum CdSe QD to polymer ratio in the CdSe/polymer and TrGO-CdSe/polymer devices was determined from 114 individual solar cells on 38 substrates (see Fig. 5.16). Figure Short-circuit current density, fill factor, open circuit voltage and power conversion efficiency from 19 CdSe/PCPDTBT and 19 TrGO-CdSe/PCPDTBT devices with 3 individual solar cells on each device. The data points represent the average value of each device, bars represent the distance between the lowest and highest measured value, centered in the graph on the point of the mean value. Dashed trend lines are inserted for easier visualization for both solar cell types. (Figure taken from Ref. [21] - reproduced by permission of the PCCP Owner Societies).

128 5. Results - Efficiency enhancement 114 As result, it was determined that the TrGO containing cells reach their highest short-circuit current at a slightly lower QD/polymer weight ratio of 85%, compared with the standard CdSe QDs containing hybrid solar cells with their optimum at 88 wt% QD. This difference is already an indication of a better interconnection of the acceptor material in TrGO containing cells, allowing for more efficient electron percolation pathways at a lower NC fraction. The solar cell performances in the respective optimum CdSe:PCPDTBT ratio obtained with both solar cell types in direct comparison is summarized in the following Table 5.8. Table 5.8. Average device parameters and standard deviations from 42 CdSe/PCPDTBT and TrGO-CdSe/PCPDTBT solar cells manufactured in the respective optimum CdSe QD:polymer ratio. The values for the two outer and the inner solar cells are given separately. CdSe/PCPDTBT TrGO- CdSe/PCPDTBT J SC [ma/cm²] FF V OC [V] PCE [%] outer solar cells (std. dev.) inner solar cell (std. dev.) outer solar cells (std. dev.) inner solar cells (std. dev.) From the above table one can see that there is a consistent trend of a higher open circuit voltage (V OC ) for the TrGO-CdSe/polymer solar cells, which persists also for other CdSe:Polymer ratios (see Fig. 5.16). The V OC difference is of about 0.15 V, with an average value of 0.56 V for the QD only/polymer cells and 0.71 V for the TrGO containing cells. Thereby, even though CdSe-TrGO/polymer solar cells tend to have a smaller average J SC of about 8 ma/cm² compared to ca. 9 ma/cm² of the TrGO free solar cells, they exhibit higher power conversion efficiencies with an average of 3.3% compared to 2.7% mainly due to a higher V OC. For verification of our results, measurements have been performed in an external laboratory, more specialized on solar cell characterization. Our best TrGO- CdSe/PCPDTBT solar cell with 85 wt% CdSe and CdSe/polymer solar cell with 88 wt% CdSe were characterized 20 days after manufacturing at the group of dye and organic solar cells of the Fraunhofer Institute for Solar Energy Systems (ISE). At first, the spectral response of the

129 Results - Efficiency enhancement solar cells was measured, allowing for spectral mismatch corrected solar cell PCE measurements [197]. Thereby, the correct illumination intensities could be applied for the following current-voltage measurements of the two examined solar cells. Also, photos of the two solar cells have been taken to exactly determine the active area represented by the overlapping region between the ITO substrate and the aluminum top electrode. The measurements showed similar results as measured in our lab - by taking into account the correct determined active area and the now obtained mismatch factor. A comparison of the solar cell measurement from our lab and from the external laboratory is given in Figure 5.17 and Table 5.9 for the two best solar cells. Figure Current-density/voltage diagram for the best CdSe QD/PCPDTBT and TrGO-CdSe QD/PCPDTBT solar cell measured under AM 1.5G illumination in our laboratory (solid lines), and the same two cells measured 20 days later at the group of dye and organic solar cells of the Fraunhofer Institute for Solar Energy Systems (dashed lines). (Figure taken from Ref. [21] - reproduced by permission of the PCCP Owner Societies). Table 5.9. Results of the best solar cells from measurements performed in our and at an external laboratory. The active area of the measured solar cells is of cm² for the CdSe/polymer and of cm² for the CdSe-TrGO/polymer device. Also, spectral mismatch for the AM 1.5G illumination was considered. CdSe/ PCPDTBT TrGO-CdSe/ PCPDTBT J SC [ma/cm²] our laboratory V OC FF [V] PCE [%] external laboratory J SC V OC FF [ma/cm²] [V] PCE [%]

130 5. Results - Efficiency enhancement 116 Hence, the best solar cell on the tested substrate for the CdSe/polymer device resulted in an efficiency of about 3% in our lab and of 2.9% measured in the external lab. The performance of the TrGO-CdSe/polymer solar cell resulted in a PCE of about 4.2% compared to 4.1% measured in our lab. Also, the observed higher open circuit voltage could be confirmed, with the TrGO containing solar cell displaying a ca. 30% higher V OC. Moreover, also a 13% higher J SC - compared to an only 5% higher short-circuit current determined in our lab - was measured; a difference which might be attributed to different aging of the TrGO free solar cell, although the differences are within the supposed margin of error of the measurement. It is noteworthy to mention that the TrGO containing solar cells have a higher fluctuation of their short circuit current, observable by a lower current of the central solar cell on each substrate a behavior also observed in our lab (see Table 5.8). This behavior might derive from an inhomogeneous distribution of TrGO within the active layer due to the centrifugal force during spin coating. Thus, a different coating technique, i.e. doctor-blading, might be more suitable for the manufacturing of the CdSe-TrGO hybrid containing cells. The external quantum efficiency (EQE) spectra of TrGO-QD and QD based devices determined at the group of dye and organic solar cells of the Fraunhofer ISE are depicted in Figure 5.18a. Figure (a): External quantum efficiency (EQE) spectra from which spectral mismatch factors of for the TrGO solar cell and for our standard solar cell were determined. (b): Absorption spectra of the active layers out of CdSe QD/PCPDTBT with a 88:12 weight ratio, TrGO-CdSe QD/PCPDTBT with a 85:15 weight ratio of CdSe:PCPDTBT, and of solutions in chloroform from CdSe QDs and PCPDTBT. (c) Normalized internal quantum efficiency (IQE) for both solar cell types, obtained by calculating the quotient between the respective percentage of the EQE and absorption spectrum of the active layer, without consideration of the reflection spectrum. (Figure taken from Ref. [21] - reproduced by permission of the PCCP Owner Societies).

131 Results - Efficiency enhancement In addition, UV-Vis absorption spectra of CdSe QDs and PCPDTBT in solution, as well as absorption spectra of the CdSe QD/PCPDTBT and TrGO-CdSe QD/PCPDTBT active layers are displayed in Figure 5.18b. From Figure 5.18b one can see that the TrGO containing cell exhibits a much lower CdSe absorption, observable from the less prominent absorption peak at 640 nm and a lower absorption at wavelengths below, which can be explained by the observed stronger QD aggregation in the active layer. On the other hand, the polymer absorption is only slightly stronger, observable from a higher absorption over 650 nm, probably additionally deriving from the slightly higher polymer fraction used for the optimized TrGO-CdSe/polymer solar cell. The resulting EQE (see Fig. 5.18a) demonstrates a much higher current output per incident photons for wavelengths between nm, a region of strong polymer absorption. Given the only slightly higher polymer absorption, the resulting IQE (see Fig. 5.18c) - without consideration of reflection - is therefore considerably higher in this region. This demonstrates a superior extraction of electrons excited in the polymer phase for the TrGO containing solar cells, with the CdSe QDs apparently contributing less to the photocurrent generation compared to the TrGO free solar cell Charge extraction To examine whether the hybrid TrGO-CdSe nanocomposite has a positive influence on the electron mobility (µ e ) in the hybrid solar cells, CdSe QD/PCPDTBT solar cells on 2 different substrates, and TrGO-CdSe QD/PCPDTBT solar cells on one substrate were taken to FLUXIM AG for CELIV, impedance spectroscopy, and capacitance-voltage measurements by the PAIOS (Platform for All-in-one characterization of Solar cells) system, whose theoretical and practical significance of enabling the execution of multiple measurement methods within a short time on a thereby nearly unchanged sample has been demonstrated by Neukom et al. [223]. At first, a CELIV measurement was performed in the dark and under 100 mw/cm² illumination by a white-light LED. In the CELIV measurement a voltage ramp in reverse bias is applied to the solar cell. Extracted holes and electrons are visible as a current in an external sensing circuit. The faster the charges can be extracted, the higher their extracted mobility values are [201, 202, 224]. A detailed description of the measurement procedure is given in Section The average CELIV graph from two CdSe QD/polymer solar cells and one TrGO-

132 5. Results - Efficiency enhancement 118 CdSe/polymer solar cell is presented in the following Figure The mobilities determined by the CELIV measurements of electrons and holes of QD based BHJ solar cells with and without TrGO are listed in Table Figure CELIV measurement results for a CdSe QD/polymer and a TrGO-CdSe QD/polymer solar cell, executed with a slope A of 400 mv/µs under illumination and in dark. Respective capacitive current (j(0)) values calculated from capacitance-voltage measurements are indicated by arrows (red arrow for TrGO containing cell and blue arrow for a cell without TrGO). (Figure taken from Ref. [21] - reproduced by permission of the PCCP Owner Societies). Table Electron and hole mobilities extracted from CELIV measurements for CdSe QD/PCPDTBT and TrGO-CdSe QD/PCPDTBT solar cells. µ e [cm²/vs] µ h [cm²/vs] CdSe/PCPDTBT 1.2x10-5 8x10-5 TrGO CdSe/PCPDTBT 2.3x10-5 CELIV measurements reveal that the resulting electron mobilities are approximately twice as high for the CdSe QD-TrGO hybrid materials containing cells. The determined values of the electron mobilities are in the same range as measured by Ginger et al. [13] who measured 1x10-4 cm²/vs - 1x10-6 cm²/vs for CdSe NC films. The determined hole mobilities of 8x10-5 cm²/vs are the same in both hybrid solar cell types but nearly one order of magnitude lower than measured from PCBM/PCPDTBT blends in ortho-dichlorobenzene (odcb) by

133 Results - Efficiency enhancement Morana et al. [52] with 4x10-4 cm²/vs - 7x10-4 cm²/vs. Besides the higher electron mobility in TrGO containing cells, a noticeable aspect of the CELIV measurements is that fewer charges (86% fewer in dark and 78% fewer under illumination) are extracted from the TrGO containing cells (see Table 5.11). The extracted charges Q from the solar cells can be calculated - after subtraction of the calculated capacitive current density j(0) derived from the obtained capacitance C (see Formula 3.14 in Section 3.3.4) - from the measured CELIV current integrating the remaining extracted current density Δj ( Δj = j CELIV - j(0) ) over time. In Table 5.11 the extracted charges for CELIV in dark and under illumination are displayed. Table Extracted charges in dark (Q D ), under illumination (Q L ), calculated amount of additionally extracted charges due to illumination (Q L -Q D ) and the ratio of these charges to charges extracted in dark (Q L -Q D ) / Q D of CdSe/polymer and TrGO-CdSe/polymer solar cells from CdSe QDs. CdSe QD/ PCPDTBT TrGO-CdSe QD/ PCPDTBT Extracted Q D [e/cm 3 ] in dark Extracted Q L [e/cm 3 ] under illumination (Q L -Q D ) [e/cm 3 ] (Q L -Q D ) / Q D 1.97 x x x x x x The results displayed in Table 5.11 reveal that although from the TrGO containing cell a lower quantity of charges can be extracted, the ratio between charges extracted due to illumination (Q L -Q D ) and charges extracted in dark Q D is with 3.19 about double as high for the TrGO containing cells, being an indication for a more efficient extraction free charges in the TrGO-CdSe/polymer solar cell. For obtaining the geometric capacitance of the solar cell, which was needed to calculate the capacitive current j(0) for the CELIV measurements, impedance spectroscopy (IS) measurements were performed with PAIOS. The following Figure 5.20 shows capacitance density/frequency plots were obtained under illumination and in dark.

134 5. Results - Efficiency enhancement 120 Figure Capacitance density vs. frequency of CdSe/polymer and TrGO-CdSe/polymer solar cells from CdSe QDs under illumination and in dark obtained by impedance spectroscopy. (Figure taken from Ref. [21] - reproduced by permission of the PCCP Owner Societies). From the frequency dependent capacitance one can notice that the capacitance for low frequencies is increasing with a higher slope for cells without TrGO than with TrGO. Ideally, the capacitance should run into saturation, forming a plateau, at low frequencies. Thus, the value after the knee point would indicate the geometric capacitance. According to Knapp et al. [225], a steadily increasing capacitance for low frequencies indicates the presence of slow trap states for charges in the device. Thereby one can conclude from the impedance spectroscopy that in cells containing the TrGO-CdSe hybrid material charge trapping is generally reduced, fitting to the less extracted charges in the CELIV experiment for TrGO containing cells. This finding is also supported by a former publication by Barkhouse et al. [226], showing that thiol passivation is reducing the number of deep surface traps in colloidal quantum dots. Moreover, he reports an increase of the built-in voltage for thiol capped QD solar cells. This effect was also seen from the capacitance voltage measurements performed by FLUXIM AG on the TrGO containing solar cells, as from the capacitance voltage measurement depicted in Figure 5.21 one can see from the capacitance peak position the built-in voltage [227], and which is of 0.15 V higher for TrGO containing solar cells. On the same time Barkhouse et al. observed an increased exciton dissociation efficiency, due to decreased nonradiative electron-hole recombination, leading to increased V OC and J [226] SC. This positive effect of a decreased recombination on the V OC was also shown in a more recent work by Maurano et al. [204].

135 Capacitance (R S corrected) [nf/cm²] Results - Efficiency enhancement CdSe/PCPDTBT TrGO-CdSe/PCPDTBT 80 khz Offset Voltage [V] Figure Capacitance vs. DC offset voltage at a modulation frequency of 10 khz for CdSe QD/PCPDTBT, and TrGO-CdSe QD/PCPDTBT solar cells show different built in voltages. In addition, a further indication for a decreasing density of trap states in the TrGO containing device was found by measuring the light intensity dependence of the V OC for both solar cell types (see Fig. 5.22). A lower slope-representing the solar cell ideality factor - of 1.22 compared to 1.33 for the CdSe only/pcpdtbt cell was found, indicating less trap states for charges [228, 229] in the TrGO containing hybrid solar cells. The same method was also recently applied by Gao et al. [206], demonstrating that reduced ideality factors, which correspond to less deep traps, are the cause of higher V OC of CdSe QD/polymer BHJ solar cells due to reduction of trap-mediated recombination. Figure Dependence of the open circuit voltage from the illumination intensity in a CdSe/PCPDTBT solar cell (left) and in a TrGO-CdSe/PCPDTBT solar cell (middle). The resulting slope of the VOC vs. the light intensity results in the ideality factor of the respective solar cell (right). (Figure taken from Ref. [21] - reproduced by permission of the PCCP Owner Societies).

136 5. Results - Efficiency enhancement Discussion and conclusion Now, a discussion about the observed improved performance due to the integration of TrGO-CdSe nanocomposites into hybrid BHJ solar cells should lead to a conclusion about the origin of the observed improvement. First, the morphological changes in the BHJ hybrid solar cell containing TrGO-CdSe are addressed. As it can be observed from AFM and SEM measurements (see Fig. 5.15b,c), TrGO containing cells exhibit a six times higher surface roughness of the active layer compared to TrGO free solar cells. This could be attributed by TEM imaging to coarser phase segregation (see Fig. 5.15a & Fig. 5.16), induced by the TrGO. Thereby, the nanomorphology in the active layer changes from more equally distributed QDs in CdSe QD/PCPDTBT solar cells to vertically aligned CdSe QD-TrGO nanocomposites in TrGO- CdSe QD/PCPDTBT hybrid solar cells as observed by TEM tomography (see Fig. 5.16). There have already been several attempts to control the morphology of the BHJ films by solvent variation, showing that usually a finer phase segregation between donor and acceptor material - typically observed by AFM surface roughness measurements - leads to higher J SC which is supposed to occur due to an increased exciton dissociation by the increased donor acceptor interfacial area [35, 36]. Hence, by the coarser phase segregation, in case of TrGO- CdSe QD integration, one would expect less free charges forming initially, due to the reduced donor/acceptor interfacial area, which is in accordance with the less extracted charges observed by CELIV measurements (see Fig & Table 5.11). Moreover, the observed two times increase of electron mobility (see Table 5.10) and the double ratio between charges extracted after illumination compared to charges extracted in dark (see Table 5.11), could to some extent be attributed to the observed vertically aligned CdSe QD- TrGO nanocomposites, proving a more efficient extraction of the generated free charges throughout the active layer. The CdSe-TrGO composites might have a different dispersibility compared to CdSe QDs in the solvent. Hence, after spin coating, the occurring solvent evaporation leads to the different coarser - phase segregation. The observed reduction of trapped charges in the TrGO solar cell, especially seen in impedance spectroscopy (see Fig. 5.20), might be attributed to the reduction of dead ends, leading generally to improved electron percolation pathways for electron extraction (and therefore increase of electron mobility). Both might also contribute to an overall reduction of charge recombination, which would result in higher V OC values. This cause of V OC enhancement is described for all-

137 J SC [ma/cm²] Results - Efficiency enhancement inorganic NC solar cells to derive from passivation of NC recombination centers [163]. However, the V OC enhancement achieved by reduced recombination is reported to also decrease the dark saturation current [164, 163], a behavior that could not be confirmed (see Fig. 5.23). The observed dark saturation current is in the same range for both measured TrGO containing and TrGO free devices. Therefore, the assumption of a lower recombination by the surface trap passivation to be confirmed as the cause of a higher V OC, at least not by following the argumentation in the previously given literature. So, in the following further factors which might contribute to the improved V OC values observed in TrGO containing hybrid solar cells are discussed E-3 1E-4 1E-5 CdSe/PCPDTBT CdSe-TrGO/PCPDTBT 1E Voltage [V] Figure Dark current density vs. voltage for a solar cell from a CdSe QD/PCPDTBT and a CdSe QD-TrGO/PCPDTBT solar cell. An electronic coupling due to the formed chemical bond of CdSe to TrGO via a thiol-bridge - demonstrated by the EPR quenching of free TrGO electrons in the CdSe-TrGO hybrid (see Fig. 5.14) and also to some extent by CdSe QD PL quenching in a solution containing TrGO (see Fig. 5.13, right) - might favor a better charge separation of the electron-hole pair and improved light induced electron transfer, which would also contribute to an increase of V OC in the TrGO containing cells. However, it is generally believed that in BHJ solar cells the [45] difference between the donor HOMO and the acceptor LUMO is proportional to the V OC (see Chapter 2.2). Hence, either a HOMO shift of the polymer or a LUMO shift of the NC would also result in a higher V OC. But, according to UV-Vis absorption and PL spectroscopy measurements of the CdSe-TrGO hybrid (Fig. 5.13) and UV-Vis absorption spectroscopy of a PCPDTBT-TrGO mixture (see Fig. 5.24), the optical bandgap is not changed; neither for the NCs nor for the polymer.

138 J [ma/cm²] 5. Results - Efficiency enhancement 124 Figure Transmission spectra of a PCPDTBT solution in CHCl 3 and of a mixture between PCPDTBT in CHCl 3 and TrGO in DMF, both directly after mixing and 24 hours later. (Figure taken from Ref. [21] - reproduced by permission of the PCCP Owner Societies). Therefore, a shift of the energy level positions (implying a simultaneous shift of HOMO and LUMO) of the acceptor material remains as explanation for the higher V OC in the TrGO-CdSe containing solar cells. Indeed, it has already been reported that thiol ligand exchange on CdSe NCs is shifting the energy levels towards the vacuum level without affecting the bandgap size [230, 231]. For comparison and clarification a control experiment comparing a PCBM/PCPDTBT organic solar cell with a PCBM/TrGO/PCPDTBT device was carried out, using the same procedure for adding TrGO like for the hybrid solar cells. And it resulted in nearly no change in V OC (see Fig. 5.25), which is indicating that the TrGO is affecting neither the PCPDTBT bandgap position, nor the PCBM band edge position PC 61 BM(71w%) + PCPDTBT: V OC =0.704V, J SC =5.28 ma/cm², FF=38.6%, PCE=1.40% TrGO/PC 61 BM(71w%) + PCPDTBT: V OC =0.696V, J SC =4.50 ma/cm², FF=37.9%, PCE=1.18% Voltage [V] Figure Current density vs. voltage for a solar cell from PC 61 BM/PCPDTBT and a PCBM/TrGO/PCPDTBT solar cell. (Figure taken from Ref. [21] - reproduced by permission of the PCCP Owner Societies).

139 Results - Efficiency enhancement Thus, one can conclude that the V OC enhancement is only originating from the CdSe-TrGO hybrid nanocomposite. Nevertheless, the blend morphology is also reported to influence the BHJ solar cell open circuit voltage. But, the effect on the V OC is usually considered to be small. Although, there are reports of higher V OC observable for coarser phase segregation [37, 38], which might be explained by a decreasing numbers of direct pathways within the donor and acceptor material, reducing shortage between anode and cathode [232] ; with the optimum V OC being reached for a structure where the donor material is only in contact with the anode and the acceptor, and the acceptor material is only in contact with the donor and the cathode. 5.3 Nanostructured inverted hybrid solar cells Within this subchapter the attempt of obtaining solar cells of a nanostructured interdigitated hybrid heterojunction donor-acceptor design (see Fig. 5.26) is presented. By these nanostructured solar cells in principle each exciton, created within the donor phase, is enabled to arrive at the donor-acceptor interface within the exciton diffusion range. On the same time, continuous pathways for the extraction of electrons and holes from the active layer are provided. Moreover, by the utilized solar cell structure the use of PEDOT:PSS and aluminum can be avoided. Thus, allowing further studies on the hybrid solar cell behavior on air, which are described later in Chapter 6.3 of this thesis. Figure Schematic comparison in between a hybrid bulk heterojunction solar cell (left half) and a nanostructured inverted hybrid heterojunction solar cell (right half).

140 5. Results - Efficiency enhancement 126 The nanostructured solar cells were built by utilizing nanostructures out of anodized aluminum oxide (AAO) and anodized titanium oxide (ATO). However, no functional solar cells incorporating the vertical nanopores out of Al 2 O 3 resulted. Thus, the creation of these pores by AAO and their filling with semiconducting polymer is reported in Appendix A7. Nevertheless, the creation of vertical TiO 2 nanotubes by ATO on ITO, and the manufacturing of hybrid solar cells based on titania nanotube arrays was realized and is described within this chapter. For the nanostructured solar cells presented in this thesis, aiming towards the fabrication of interdigitated solar cells, arrays of nanoporous metal oxide were fabricated on the glass/ito substrate. The theory behind their formation is subsequently described. The natural oxidation of metals in air can be further accelerated by immersing in an aqueous acidic solution. By applying an electric field perpendicular to the metal surface with the anode connected to the metal and the cathode connected to an electrode placed in parallel to the metal surface, the oxidation can be even further accelerated in vertical direction along the electric field. The beginning of the pore formation occurs at surface imperfections of the original this aluminum oxide layer like grain boundaries, nonmetallic inclusions or in general pits of a rough surface. The formation of porous aluminum oxide by anodic oxidation has been investigated for more than 70 years [ ]. However, until now the formation of the highly ordered vertical pores (see Figure 5.27, left) is still not completely understood. Figure Left: Schematic top and side view of pores of anodized aluminum oxide (AAO) with the interpore distance (a) and the oxide layer thickness (d). Right: Schematic top and side view of pores of anodized titanium oxide with the inner layer of TiO 2 and the outer layer of titanium oxide hydroxide; with the later one decreasing in volume by dehydration, thus leading to free standing TiO 2 tubes with an outer layer of β-tio 2.

141 Results - Efficiency enhancement However, the electrochemistry behind the process is well described in literature [234]. The process begins at the water/metal-oxide interface, where Al 2 O 3 is dissolved into Al 3+ cations and O 2- anions and water is dissociating into OH - and H +. The formed cations are extracted by the electric field into solution, a process which is further accelerated by the presence of acid. The formed anions are diffusing into the metal-oxide layer and are reaching the metaloxide/metal interface, where they oxidize aluminum to form Al 2 O 3. Thereby, the original alumina layer is dissolved and aluminum is further transformed into alumina. Until now many publications described the experimental results obtained with different experimental settings, like the observation of increasing interpore distance by increased applied voltage [ ], however there have been few models to explain the formation of the regular, hexagonal ordered pores. One coherent model of the formation mechanism is presented by Su et al. [239], which can as well be applied for the explanation of the occurring free standing nanotube array (see Figure 5.27, right) from anodic titanium oxidation. Su et al. are introducing the variable n for the ratio between dissociation of water and the dissolution of Al 2 O 3 : Al 2 O 3 + nh 2 O Al 3+ + O 2 + OH + H + (5.1) The variable n is meaningful, because a faster dissociation of water is increasing the OH - concentration and thus the rate at which new Al 2 O 3 is formed at the metal-oxide/metal interface, so that despite the dissolution of Al 2 O 3 the oxide layer thickness d is increasing. This oxide layer thickness is governing the electric field strength E=U/d at a given applied potential over the metal-oxide layer, which is determining the speed at which anions are accelerated towards the metal-oxide/metal interface. Therefore, the initially started metal oxidation at a pit in the surface would continue isotropic throughout the metal layer, simply enlarging the hemispheric pits until the oxide thickness has eventually reached a critical value d c at which the electrical field strength is too weak (E c ) to further accelerate anions through the metal-oxide layer. However, because several such pits are enlarged over the surface, the metal oxidation cannot continue in the horizontal plane only until different advancing pits meet each other and can only further continue the metal oxidation in the vertical direction.

142 5. Results - Efficiency enhancement 128 The same also applies to anodic titanium oxidation for which Su et al. expressed the dissolution of TiO 2 and dissociation of water, also using the variable n, by the following formula: TiO 2 + nh 2 O + F [TiF 6 ] 2 + O 2 + OH + H + (5.2) In contrast to the anodic aluminum oxidation (AAO), for anodic titanium oxidation (ATO) the constitution of the inner wall of the advancing pore is of a different constitution then the outer wall. The composition of the outer wall is called β-tio 2 and could be identified as deriving from TiO 2 xh 2 O (with unknown variable x). This is leading to the conclusion that titanium oxide hydroxide (Ti(OH) 4 ) layer is formed by OH - anions migrating through the TiO 2 layer towards the metal-oxide/metal interface. Thus the advancing front of the pore consists out of TiO 2 and titanium oxide hydroxide. When all metal material is used up a layer of titanium oxide hydroxide will remain as an outer shell of the TiO 2 pores. The same phenomenon is also considered to occur for anodic aluminum oxidation, but the layer of aluminum oxide hydroxide is much thinner as the dissociation energy of aluminum oxide hydroxide is over 30 times lower than that of Ti(OH) 4. Ultimately, the free standing TiO 2 tubes are formed because of volume reduction of the titanium oxide hydroxide by dehydration, forming β-tio 2. And finally, Su et al. were able to theoretically correlate their introduced variable n (which determines the ratio between metal-cation dissolvation and H 2 O dissociation) to the obtained porosity P P = 2π 3 r a 2 (5.3),with the pore radius r and the interpore distance a of the anodic metal oxidation process. Which is determined as P AAO =3/n+3 in case of AAO and P ATO =2/n+2 in case of ATO. Therefore, by knowing the porosity of an anodic oxidation reaction, also n for the Formulas (5.1) & (5.2) can be determined. Here it is also noteworthy to mention that in order to obtain the semiconducting anatase TiO 2 crystal structure from the created amorphous TiO 2, a further thermal annealing of the TiO 2 array over 400 C is needed [240].

143 Results - Efficiency enhancement Manufacturing of titania nanotube arrays on ITO In order to create a nanostructured electron acceptor material directly on ITO, the method of anodic titanium oxidation is used (experimental details see in Section 3.2.7), leading to the formation of a TiO 2 tube array on ITO was investigated with the aim of building interdigitated hybrid solar cells. The anodization of titanium layers of thicknesses between 500 nm and 1000 nm, sputtered on an ITO substrate, in an ammonium fluoride solution (0.5 wt% NH 4 F in 95:5 vol%:vol% ethylene glycol:h 2 O at 40V for 50 min) resulted in the formation of 3 different layers (marked with arrows in Figure 5.28). All three were porous, however the top layer is very irregular, only the middle layer shows homogenous well ordered nanotubes. Figure SEM image of anodized titanium, realized by 0.5 wt% NH 4 F in 95:5 vol% Ethylene Glycol:H2O at 40V for 50 min. Therefore, after 30 min of anodization the process was stopped, the substrate was taken out of the electrolyte solution and a drop (10 µl) of sulfuric acid (H 2 SO 4, 95%) were dispensed on the area prior exposed the electrolyte to remove observed irregular top layer. After 5 minutes at room temperature, the substrate was put into a glass of H 2 O, which was placed into an ultrasonic bath at room temperature for 5-15 min. Afterwards, the substrate changed again to a smooth surface layer (when observed by eye) like shown in Figure 5.29.

144 5. Results - Efficiency enhancement 130 Figure Picture of the glass/ito/titanium/titania substrate before (left) and after treatment with H 2 SO 4 and ultrasonic bath (right). Thereafter, the substrate was placed again into the reaction chamber to continue the anodization with the same settings like before for further 30 min. Afterwards, a semitransparent gray layer was formed (Fig. 5.30, left) which strongly increased its transparency after annealing at 450 C for 2h in air (Fig. 5.30, middle). According to the SEM images (Fig. 5.30, right) of this glass/ito/tio 2 substrate the irregular porous top layer has disappeared by this two-step anodization. Figure The created ATO sample (left) shows an increase of transparency after annealing at 450 C (middle), thus indicating a change from amorphous to crystalline TiO 2. SEM image of the ATO sample tilt by 45. The change in crystal structure by the prior described thermal annealing was investigated by X-ray diffraction (XRD) spectroscopy. The result is depicted in the following Figure 5.31.

145 Intensity [a.u.] Results - Efficiency enhancement Ti ITO ITO ITO ITO ITO A ITO Ti Ti before annealing ITO A ITO A ITO R A ITO after annelaing (2h 450 C) [degree] Figure XRD pattern of non-annealed and annealed sample of ATO created on an ITO substrate. The sample before annealing shows species of low crystallinity and characteristic ITO peaks. After annealing crystalline species of Ti and TiO 2 (mostly anatase, but also rutile) are detectable. The non-annealed ATO sample showed peaks from the underlying ITO and additionally two broad peaks that show a species of a low crystallinity (probably originating from titanium) respectively small crystallite sizes according to the Scherrer equation [241]. In this region after annealing two strong peaks characteristic for Ti are occurring with an additional Ti peak appearing at about 2θ=40. In the annealed ATO sample, also a very strong peak is occurring at 2θ=25.3 [242], which is the most prominent peak for anatase TiO 2. Also, smaller peaks for the anatase structure of TiO 2 and even a small rutile TiO 2 peak at about 35.6 are observed. So, by the recorded XRD pattern one con conclude that a prior amorphous titanium layer was transformed into a layer containing high fractions of anatase TiO 2. The Ti peaks, which are still visible in the annealed sample, might derive from not oxidized material under the TiO 2 tubes Inverted solar cells from TiO 2 nanotubes on ITO In order to further prove the formation of the semiconducting anatase species, the currentvoltage characteristic of a prior presented ATO sample was recorded with a Keithley 2603 source-meter under room light before and after annealing. Therefore, one contact tip was

146 Current [µa] 5. Results - Efficiency enhancement 132 placed on the ITO and the other contact tip was placed on top of the anodized titanium oxide layer in contact with the underlying ITO. The prior presented ATO arrays displayed an obvious diode behavior from I-V measurements after thermal annealing (see Fig. 5.32), further proving the formation of the semiconducting anatase TiO 2 crystal structure. Hence, the TiO 2 NT array might be used as electron acceptor in a solar cell after annealing before annealing Voltage [V] Figure I-V characteristic of an ATO sample measured under room light before and after thermal annealing for 2h at 450 C in a furnace. Solar cells were built by spin-coating a CdSe QD/PCPDTBT mixture in chlorobenzene over the TiO 2 nanotube layer and annealing the cells at 145 C for 8 minutes, followed by thermal evaporation of gold. The use of gold as top contact is necessary, because the material combination of the solar cell requires the holes to be collected at the top electrode by a material with a work function higher then ITO to allow the electron extraction over TiO 2, but not higher than the polymer HOMO in order to facilitate the hole extraction over the top contact (see Fig. 5.33, left). The so formed solar cell is thus an inverted solar cell compared to the solar cells presented in the previous chapters. The best so manufactured solar cell exhibited a short-circuit current of 0.65 ma/cm², a V OC of 0.34 V and a fill factor of 32.4%, to result in a PCE of 0.072% (see Fig. 5.33, right). Despite the low achieved power conversion efficiency, the hybrid solar cell built upon a TiO 2 nanotube array proved to work as inverted solar cell.

147 J [ma/cm²] Results - Efficiency enhancement 0.0 dark sun Voltage [V] Figure Left: Scheme of energy levels of the materials used in the inverted nanostructured solar cell, showing the conduction band (CB) positions of ITO and TiO 2, the LUMO and HOMO levels of CdSe NCs and of the semiconducting polymer, and the work function of gold. Right: Current density-voltage characteristics of a hybrid solar cell consisting out of TiO 2 nanotubes grown on ITO, covered by a mixture of CdSe and the semiconducting polymer PCPDTBT and gold as top contact. However, the titanium anodization process is relatively complex. Nevertheless, from experiments it was found that the titanium oxidation of thinner layers below ca. 300 nm does not result in the irregular porous top layer, allowing for a one-step anodization on ITO (see Appendix A6), with which however solar cells of only lower PCEs could be produced.

148 6. Results - Solar cell degradation Results - Solar cell degradation Herein, at first the degradation of hybrid CdSe QD/PCPDTBT BHJ solar cells with differently strong ligand reduced QDs, stored in nearly inert atmosphere, is described (Chapter 6.1). Subsequently, the behavior of hybrid and organic BHJ solar cells in air is compared (Chapter 6.2). And finally, the behavior of an inverted nanostructured hybrid solar cell in nitrogen, air, and oxygen atmosphere is investigated to get more insight about the degradation causes specific for hybrid NC/polymer solar cells (Chapter 6.3). 6.1 Degradation of hybrid BHJ solar cells in inert atmosphere Within this chapter the investigation of the long term stability of CdSe-QD/PCPDTBT BHJ solar cells over up to 1.5 years is reported. Moreover, the difference in long term stability of hybrid solar cells containing NCs from weak and strong post-synthetic hexanoic acid treatment is investigated and discussed. The solar cells were kept after manufacturing inside a nitrogen filled glovebox (H 2 O: <5 ppm, O 2 : flactuating around 25 ppm with typically +/- 10 ppm) and were periodically measured under approximate AM 1.5G conditions. In between the measurements, the solar cells were stored in the dark to prevent degradation by UV-illumination. Thereby, it was attempted to investigate the intrinsic stability of the hybrid solar cells, excluding as many external influences as possible after the solar cell manufacturing Utilized NCs and their post-synthetic treatment The utilized CdSe NCs were synthesized like described in Section at 300 C for 30 min with a molar ratio of 100:3:2 between surfactants (HDA & TOPO), cadmium-precursors, and selenium-precursors. A TEM image of the used NCs, QDs with an average diameter of 7 nm, is depicted in Figure 6.1.

149 Results - Solar cell degradation Figure 6.1. TEM image of CdSe NCs used for the investigation from a 100:3:2 (HDA/TOPO):(Cd-SA):(TOP-Se) ratio, synthesized for 30 min at 300 C by wet-chemical hot injection NC synthesis with an average diameter of 7 nm. Hybrid BHJ CdSe/PCPDTBT solar cells on 5 different substrates were manufactured containing NCs post-synthetically treated with hexanoic acid for different times. From CdSe NCs treated for 10 min with HA, solar cells on two different substrates were fabricated. Solar cells on one substrate contained 12 min HA washed NCs, and solar cells from NCs washed for 22 min were fabricated on two other substrates. The utilized NC concentration was of 30 mg/ml, and the PCPDTBT concentration was of 20 mg/ml. For the NC:polymer blend a ratio of 88:12 wt%/wt% was utilized Initial performance The solar cells containing 10 min and 12 min washed NCs were annealed at 150 C for 30 min. The solar cells made with 22 min washed NCs received no thermal treatment, as their PCE is thereby expected to strongly decrease, like reported in Chapter 4.5. The performances of the solar cells by the start of the long term measurements are summarized in Table 6.1.

150 V OC [V] J SC [ma/cm²] 6. Results - Solar cell degradation 136 Table 6.1. Average performance from hybrid CdSe/PCPDTBT BHJ solar cells on the 5 investigated substrates before and after thermal annealing ( C) at the start of the long term investigation. For substrates on which thermal annealing was performed, the initial values are listed in brackets. Substrate # HA washing time of CdSe QDs [min] V OC [V] (0.607) (0.453) (0.673) J SC [ma/cm 2 ] (1.34) 5.73 (1.40) 5.56 (2.57) 6.93 FF [%] (44.1) 48.9 (47.6) 47.6 (47.3) 50.5 PCE [%] (0.36) 1.86 (0.30) 1.64 (0.82) From the above Table 6.1 one can see that solar cells containing longer washed NCs tend to have a lower open circuit voltage. Moreover, the same tendency as shown in Figure 4.21 of a higher FF for solar cells from longer washed NCs is observed. Also, by thermal annealing the initially low short-circuit currents for solar cells containing shortly washed NCs were increased, but did not reach the high values of the solar cells from stronger washed NCs Performance over time The periodically measured solar cell parameters of V OC, J SC, FF, and PCE are depicted in the following Figures 6.2 & Time [months] 10 min HA 12 min HA 21 min HA Time [months] Figure 6.2. V OC and J SC periodically measured over up to 19 months for BHJ CdSe- QD/PCPDTBT solar cells stored inside a glovebox on 5 different substrates containing CdSe NCs treated with hexanoic acid for 10 min, 12 min, and 21 min.

151 R S [ cm²] J SC [ma/cm²] FF Efficiency [%] Results - Solar cell degradation min HA 12 min HA 21 min HA Time [months] Time [months] Figure 6.3. FF and power conversion efficiency periodically measured over up to about 1.5 years for BHJ CdSe-QD/PCPDTBT solar cells stored inside a glovebox on 5 different substrates containing CdSe NCs treated with hexanoic acid for 10 min, 12 min, and 21 min. In order to visualize also the early performance changes of the solar cells, the time axis was now logarithmically scaled (see Fig. 6.4). Also, in order to get a deeper insight into the solar cell performance, the series resistance (R S ) and parallel resistance (R P ) of the solar cells were calculated from the J-V curves, by the method described in Chapter min HA 12 min HA 21 min HA Time [h] Time [h] Figure 6.4. Development of and R S and J SC over time with logarithmic x-axis, and also logarithmic y-axis for the R S development in order to also show the change of the initially small RS values for the solar cells containing 10 min and 12 min washed NCs. Solar cell measurements shown inside the red-hatched area were obtained during annealing of the solar cells at 145 C. From the above Figure 6.4 (left) one can first notice that R S is decreasing for solar cells containing 10 min and 12 min washed NCs during annealing, which might be like also

152 R P [ cm²] V OC [V] 6. Results - Solar cell degradation 138 discussed in Chapter 4.5 due to thermally induced ligand desorption from the NCs. However, during the operation of these solar cells at room temperature, the series resistance is constantly increasing from an average value of 24 Ω/cm² right after annealing to 618 Ω/cm². At the same time the short-circuit current, which is also increasing during thermal annealing for solar cells containing NCs from shortly post-synthetically treated NCs, is as well constantly decreasing by time at room temperature inside the nitrogen filled glovebox from an average value of 6.1 ma/cm² to 2.6 ma/cm². In contrast, for solar cells containing NCs washed for 21 min in HA the initial average series resistance of 15 Ω/cm² increased only to 31 Ω/cm². At the same time J SC remained unchanged over 1.4 years at 7.9 ma/cm². This observed increase of the solar cells series resistance with accompanied by a strong J SC decrease for solar cells from shortly post-synthetically treated NCs might derive from an excess of NC surfactants present in the active layer and slowly accumulating below the aluminum top electrode. The lower amount of stabilizing NCs surfactants for longer post-synthetically treated NCs is also indirectly visible by the lower parallel resistance (Fig. 6.5, left) of an average value of 375 Ω/cm², decreasing down to 167 Ω/cm² for solar cells containing 21 min HA washed NCs. Due to ligand deattachment, the parallel resistance is also decreasing for solar cells containing shorter washed NCs. However, the values are always higher, with 990 Ω/cm² before thermal annealing, 463 Ω/cm² right after thermal annealing, and a final average value of 320 Ω/cm² after storage at room temperature of about 1.5 years min HA 12 min HA 21 min HA Time [h] Time [h] Figure 6.5. Development of V OC and R P over time with logarithmic x-axis, and also logarithmic y-axis for the R P development in order to also show the change of the later small R P values. Solar cell measurements shown inside the red-hatched area were obtained during annealing of the solar cells at 145 C.

153 FF Efficiency [%] Results - Solar cell degradation The open circuit voltage is decreasing for the shorter washed NCs during annealing; probably due to induced NC aggregation creating shunts between the two electrodes and also inducing surface defects which may lead to deep traps. This decrease of V OC is consistent with the observed decrease of R p, which indicates an increased recombination [204]. However, the V OC is subsequently recovering by time at room temperature (see Fig. 6.5, right), while the parallel resistance still shows a decreasing trend. Hence, this observed V OC increase might have a different cause, which could be according to Figure 6.7 uptake of residual oxygen inside the glovebox. Changes in the prior described R P and R S are together proportional to changes in the FF (see Fig. 6.6, left). The combination of V OC, J SC, and FF is resulting in the PCE, whose development is shown in Figure 6.6 (right) min HA 12 min HA 21 min HA Time [h] Time [h] Figure 6.6. Development of FF and PCE over time with logarithmic x-axis. Solar cell measurements shown inside the red-hatched area were obtained during annealing of the solar cells at 145 C. Solar cell parameter dependency from residual oxygen concentration: It seems that the shorter washed NCs are more sensitive to oxygen, with the V OC fluctuating stronger than solar cells containing longer washed NCs - even under the low O 2 concentration of about 25 ppm present in the glovebox (see Fig. 6.7). This low V OC increase of solar cells containing long washed NCs could have two reasons: one might be the tighter contact of the NCs with the polymer - compared to NCs with more surfactants around them;

154 V OC [V] O 2 concentration [ppm] 6. Results - Solar cell degradation 140 the second cause might be the stronger NC aggregation which is reducing their reactive surface Time [months] 0 Figure 6.7. Evolution of the open circuit voltage of 3 CdSe-QD/PCPDTBT solar cells from one substrate containing 10 min HA washed NCs, and of the oxygen concentration inside the glovebox. During the investigation the water concentration was of <5 ppm. In contrast, it is reported that non-encapsulated purely organic BHJ solar cells exposed to air show no increase of V OC by time, but rather a slow initial decrease, followed by a strong V OC decline [159, 243]. Whereas hybrid BHJ solar cells taken into air, without a protection layer against oxygen-induced degradation, show an initial V OC increase which has been reported by Yang et al. [162]. This result is now confirmed by the here presented work. As discussed in Section 2.6.3, the cause of this V OC rise might be the passivation of the NCs surface with oxygen. Like the open circuit voltage, also the short-circuit current shows a dependency on the O 2 concentration (see Fig left), exhibiting a decreasing J SC for an increased oxygen concentration. Furthermore, a decreased J SC can be recovered by decreasing the oxygen concentration, like it is clearly visible in the example of Figure 6.8 (left) at about 3 months. This observed reversibility is not occurring immediately, but rather slow, indicating that the origin of recovery is oxygen depletion within the active layer by O 2 diffusion, when reducing the ambient oxygen concentration. Also, this J SC reducing effect might not be linked with the adsorption of O 2 to the NC surface. Since, a decreasing J SC is also observed for purely organic solar cells when taken into air, like it is shown in the following Chapter 6.2. Thereafter, the reason for the reduced J SC might be the reversible uptake of oxygen within the polymer phase.

155 O 2 Concentration [ppm] FF O 2 Concentration [ppm] Results - Solar cell degradation J SC Time [months] Time [months] Figure 6.8. Left: Evolution of short-circuit current of 3 CdSe-QD/PCPDTBT solar cells from one substrate containing 10 min HA washed NCs, and of oxygen concentration inside glovebox. Right: Corresponding FF development. Herein, only the first 8 months are shown, for a better recognition of the dependency of the performance parameters with O 2 concentration. Also the fill factor shows a dependency from the oxygen concentration Figure 6.8 (left), even though the relative FF change is only very low compared to the observed changes in J SC Summary In summary, all hybrid solar cells show a strong initial decrease of their PCE within the first 1.5 months by about 25%, which is due to an increase of the series resistance by a factor of 2-3 to Ω/cm². Nevertheless, the initial PCE decrease is less pronounced for solar cells containing shortly washed NCs due to a lower initial decrease in R P, which might be explained by the higher colloidal stability of these NCs originating from the larger remaining ligand sphere. Furthermore, the hybrid CdSe-QD/PCPDTBT solar cells display an increasing V OC by time, which is also dependent on the O 2 concentration, and thus partially reversible with decreasing atmospheric O 2 concentration. However, solar cells containing NCs to which a longer post-synthetic treatment was applied show a higher stability of their V OC. Hybrid solar cells containing longer HA washed NCs also display a nearly stable R S inside the glovebox after the initial decay. In contrast, the series resistance of hybrid solar cells containing shortly HA washed NCs show a constant exponential increase of their series resistance by a factor of about 10 within 1.5 years. It is remarkable that the J SC of solar cells containing long washed NCs shows only a very slight loss of J SC throughout the about 1.5

156 6. Results - Solar cell degradation 142 years kept inside the glovebox at low O 2 and H 2 O concentration, and also nearly no change in PCE after the initial decay. Thus, hybrid solar cells, long term stable for at least 1.5 years at a PCE of about 1.75% inside the glovebox could be built, in contrast to the hybrid solar cells from shorter HA washed NCs, which displayed a constant PCE decay within the same time to about 0.5% PCE. In conclusion, the post-synthetic NC treatment time (here performed by a combined HA washing with a subsequent methanol NC precipitation and simultaneous dissolution of the protonized ligand), influences both the immediate performance (less or no thermal annealing is necessary) and the long term performance of BHJ hybrid NC/polymer solar cells. By including as less ligands as possible, achievable by a HA washing which removes NC surfactants down close to the edge of colloidal instability, solar cells can be stable under N 2 protection and by excluding UV degradation for over 1.5 years and can even withstand O 2 concentrations of up to 30 ppm without a considerable efficiency decay. 6.2 Degradation of unsealed hybrid BHJ solar cells in air Initial performance Now, the behavior of unsealed hybrid BHJ solar cells in air (manufactured like described in Chapter 3.2 incorporating short and long HA washed CdSe QDs) was investigated, and compared to a purely organic BHJ solar cell. Therefore, two different CdSe-QD/PCPDTBT solar cells containing CdSe QDs washed for two different HA washing times (15 min and 22 min) were compared with one organic PC 61 BM/PCPDTBT solar cell. The organic solar cell was manufactured like the hybrid solar cells, just that the active layer was composed out of a 2.5:1 weight ratio between PC 61 BM (Phenyl C 61 butyric acid methyl ester, Sigma- Aldrich, >99.5%) and PCPDTBT. The performance values of V OC, FF, J SC, and PCE of the 3 solar cells after thermal annealing (on hotplate at 145 C, inside glovebox c(o 2 ) 13 ppm and c(h 2 O)<5 ppm) and just prior to transfer of the cells into air is shown in the following Table 6.2. Therein, the performance of the solar cells after annealing is presented in brackets.

157 Results - Solar cell degradation Table 6.2. Open circuit voltage, short-circuit current, fill factor, and power conversion efficiency of BHJ solar cells of different composition right after annealing (in brackets) and before transferring the substrates into air, together with their annealing time, and the times for which the cells were stored inside the glovebox prior to air exposure. Substrate # Active layer 145 C Time in glovebox [months] V OC [V] J SC [ma/cm 2 ] FF [%] PCE [%] 1 CdSe QD (15 min HA) /PCPDTBT 4 h 0.13 (0.521) ( 6.13) 5.03 (45.7) 38.7 (1.39) CdSe QD (22 min HA) /PCPDTBT 20 min 6.5 (0.591) (9.03) 7.85 (58.9) 45.9 (3.14) PC 61 BM /PCPDTBT 12 min 6 (0.704) (4.84) 4.65 (38.1) 34.4 (1.30) 1.14 From the above Table 6.2 one can see that the solar cell containing shorter HA washed QDs needed a long thermal annealing of 4h to achieve the optimal PCE. During annealing the J SC rose, but the V OC decreased from 0.6 V to 0.52 V (development not shown). In contrast, the solar cell containing longer HA washed QDs reached its peak PCE after a much shorter annealing time of 20 min. The PCE of both hybrid solar cells was 27% lower prior to taking the substrates into air (due to decrease of J SC and FF), compared to the efficiency directly after annealing. The organic solar cell showed only a decrease in PCE of about 12%, mainly due to a fill factor decrease Performance in air Now, the 3 substrates were subsequently taken out of the glovebox for measurements in air under approximate AM 1.5G illumination. The oxygen concentration in air is known to be of about 209,420 ppm, the relative air humidity during the measurements was of 40-60%. In the following, the results of the 3 different solar cells are given. For easier comparability the performance values are plotted and normalized with respect to the value measured inside

158 J SC, norm. R S [ /cm²] 6. Results - Solar cell degradation 144 the glovebox right before taking the cell into air. The values of the FF and PCE are also shown, but their behavior can be derived from the also presented values for R P, R S, J SC, and V OC hybrid - short w. hybrid - long w. organic Time [min] Time [min] Figure 6.9. Development of short-circuit current density and series resistance for hybrid BHJ CdSe/PCPDTBT solar cells with short HA washed and long HA washed NCs in air, in comparison with a BHJ PC 61 BM/PCPDTBT solar cell in air. Short-circuit current (J SC ) development: The above Figure 6.9 (left) shows that the slowest decay of short circuit current within the first 150 min after air exposure is exhibited by the PCBM:PCPDTBT solar cell. The hybrid solar cell containing the shorter washed CdSe NCs shows a very strong J SC decay of over 75% within the first minute due to a strong increase of the series resistance R S (see Fig. 6.9, right) which was diminished during the following 20 min, thus leading to a partial recovery of J SC. In Figure 6.9 (left) therefore for the CdSe solar cell containing 15 min HA washed CdSe NCs, a black dashed trend line indicates the presumptive J SC behavior without this initially occurring R S peak. This R S peak is probably deriving from reduced conductivity by initial oxidation of one of the electron conducting parts of the solar cell which, in absence of the occurrence of this behavior in the other two cells, should derive from the usage of shorter washed NCs. One explanation might be an easier oxidation of the shorter HA washed CdSe QDs, which still possess a higher fraction of their original ligand shell. In contrast, the other hybrid solar cell - containing 22 min washed NCs - exhibits a decrease of the J SC after 150 min, which is only slightly stronger than the decrease of the organic solar cell. Moreover, after about one day in air this hybrid solar cell still provides about 56% of the original J SC, while the short-circuit current of the organic solar cell is at less than 3% of the original value after this time.

159 V OC, norm. R P [ /cm²] Results - Solar cell degradation Series resistance (R S ) development: The lowest initial series resistances (see Fig. 6.9, right) are reached by the hybrid solar cells with R S =22.2 Ω/cm² for the solar cell containing longer HA washed NCs, and R S =30.4 Ω/cm² for the solar cell containing the shorter HA washed NCs. The organic solar cell exhibited the highest initial R S value with 45.9 Ω/cm². The subsequent development shows an increase of R S for all 3 cells with the highest slope for the solar cell from shortly HA treated NCs. After 1 day in air the hybrid solar cell containing the longer HA washed CdSe QDs still possessed a R S of 180 Ω/cm², while the organic solar cell already exhibited a very high R S of ca. 2.6 kω/cm². Thus, also with respect of the series resistance the hybrid solar cell performs better for a longer air exposure time. Hence, the use of NCs does not represent a drawback in respect to the series resistance development of the BHJ solar cell in air, and like shown in the previous paragraph also not for the short-circuit current. On the contrary, an even better stability of these parameters might be expected upon use of inorganic QDs as electron acceptors k 1.1 hybrid - short w. hybrid - long w. organic Time [min] Time [min] Figure Development of open circuit voltage and parallel resistance for hybrid BHJ CdSe/PCPDTBT solar cells with short HA washed and long HA washed NCs on air, in comparison with a BHJ PC 61 BM/PCPDTBT solar cell on air. Open circuit voltage (V OC ) development: From the development of the open circuit voltage of the 3 cells (see Fig. 6.10, left) one can notice the strong increase of the V OC of the hybrid solar cell built from shortly washed CdSe NCs by about 23% reached after 30 min. This behavior was already measured occurring with a lower rate also inside the glovebox, in presence of low O 2 and H 2 O concentrations

160 6. Results - Solar cell degradation 146 (see Chapter 6.1). The hybrid solar cell from longer washed CdSe NCs also exhibits an increase of V OC in air. However, this increase occurs considerably slower, with a V OC increase of about 10% only being measurable after about 22 h. In contrast, the V OC of the purely organic solar cell slightly decreases in air, but by only 3%, even for drastically reduced short circuit current after about one day in air. The V OC increase, thus is apparently related to the use of NCs and being especially strong for NCs which presumably still poses a large fraction of their original synthesis ligand sphere (in case of the shortly HA washed QDs). Parallel resistance (R P ) development: The highest initial R P vales are measured for the hybrid solar cells, with 347 Ω/cm² (cell containing shortly washed QDs) respectively 296 Ω/cm² (cell containing longer washed QDs), while the organic solar cell possesses the lowest R P of 236 Ω/cm². Subsequently, the parallel resistance is initially increasing for all 3 solar cells in air. For comparison, the solar cells kept inside the glovebox showed a constant decrease of R P (see Fig. 6.5, left), except the solar cells containing shortly HA washed QDs, which exhibited a slight increase of R P after about 4 months. Nevertheless, this measured R P increase inside the glovebox might have the same reason like the very strong and fast R P increase in air for the solar cell containing shortly washed QDs, which is within the first 150 min by far the strongest increase among the 3 solar cells measured in air. The parallel resistance increase of the other hybrid solar cell and of the organic solar cell are relatively similar, but with the organic solar cell still showing the lowest increase in air for the first 150 min. Thus, the measured increase of parallel resistance seems not to be related to the use of NCs, but it can be amplified by applying a shorter postsynthetic treatment to QDs. The following Figure 6.11 shows the development of the fill factor and of the PCE of the solar cells taken into air, which can be at least qualitatively (in case of the FF) derived from the above presented parameters and are not further commented.

161 FF, norm. PCE, norm Results - Solar cell degradation hybrid - short w. hybrid - long w. organic Time [min] Time [min] Figure Development of open circuit voltage and parallel resistance for hybrid BHJ CdSe/PCPDTBT solar cells with short HA washed and long HA washed NCs on air, in comparison with a BHJ PC 61 BM/PCPDTBT solar cell on air Summary In summary, like observed in the previous Section 6.1 for the BHJ hybrid solar cells inside the glovebox, the post-synthetic NC treatment is also determining the behavior of the solar cells in air. More exactly, the use of longer HA washed CdSe QDs is strongly reducing the R S and J SC decrease of the solar cell in air maybe to a reduced oxygen permittivity of the active layer by an increased density due to the reduction of included ligand molecules. Thereby, the hybrid solar cell is allowed to have a very similar development of the PCE in air like the organic solar cell, and to even perform better after about one day exposure to air (see Fig. 6.11, right). Also, one may conclude that the decrease of J SC of the solar cells in air is not related to the use of NCs, but is probably dominated by the oxidation of the other solar cell components. Hence, the two most likely reasons are the oxidation of aluminum at the interface between active layer and cathode [155] and the loss in conductivity of the semiconducting polymer by inclusion of oxygen into its thiophene rings [153]. 6.3 Degradation of nanostructured inverted hybrid solar cells Although the fabricated inverted, nanostructured solar cells exhibit a very low PCE (see Chapter 5.3), they proved to be outstandingly stable in their performance in the glovebox

162 Normalized Performance Normalized Performance J [ma/cm²] J [ma/cm²] 6. Results - Solar cell degradation 148 over a time period of nearly 2 years (see Fig. 6.12, left). Therefore, it was also tested how this solar cell behaves in air (see Fig. 6.12, right) compared to the behavior of standard PEDOT:PSS/(CdSe QD:PCPDTBT)/Al solar cells in air, which was presented in the previous Chapter months inside GB 9 months inside GB 22 months inside GB stored inside glovebox 0.25 h in air 17 h in air 17 h in air h in O 2 17 h in air + 60 h in O Voltage [V] Voltage [V] Figure Left: Performance of a TiO 2 NT/(CdSe QD:PCPDTBT)/Au solar cell inside the glovebox, measured under AM 1.5 G illumination up to 22 months after its manufacturing. Right: Performance of the same solar cell taken into air and later into oxygen. From the above measurements presented in Fig (right), the following Figure 6.13 was extracted, showing the development of the solar cell performance values (measured under AM 1.5G, respectively in dark in case of the dark saturation current I 0 ) by time under nitrogen, air, and oxygen atmosphere. V OC J SC J O PCE R P R S FF N 2 air O N 2 air O Time [h] Time [h] 100 Figure Normalized development of the performance parameters of the nanostructured solar cell under different atmospheric conditions. The maximum values for the respective parameters are: J SC =0.183 ma/cm², V OC = V, FF= , J 0 =0.45 µa/cm², PCE= %, R P =9770 Ω/cm², R S =2017 Ω/cm².

163 Results - Solar cell degradation For this inverted solar cell type - without the presence of the hygroscopic PEDOT:PSS and easy oxidizable aluminum - the V OC, R S, and R P are rising as well when taking the solar cell from nitrogen into air, like it was also observed for the standard BHJ hybrid solar cells in the previous Chapter 6.2. However, the V OC is again decreasing, along with the parallel and series resistance when taking the cell into oxygen, while the dark saturation current being proportional to crystal deep traps [164, 163] is not decreasing. Hence, the V OC increase is probably not increasing due to oxygen exposure, but due to adsorption of water - present in air - on the NC surface. Hence, the V OC enhancement is not (or only of a small fraction) because of O 2 saturation of vacancies of TiO 2 (like described in literature for metaloxide/polymer solar cells [ ] ), because when the solar cell is taken into oxygen atmosphere its V OC is even decreasing (see Fig. 6.12, right and 6.13, left).

164 7. Summary Summary Within this thesis the subjects of reproducibility, efficiency enhancement, and performance stability of hybrid NC/polymer bulk heterojunction solar cells were addressed. 7.1 Solar cell reproducibility The reproducibility of bulk heterojunction solar cells was investigated regarding the influence of the nanocrystals and their post-synthetic treatment. Therefore, it was shown that one has to consider the observed effect of variating inter-batch and batch-to-batch nanocrystal properties. These variations influence the optimal PCE of the hybrid solar cells and the optimal post-synthetic treatment procedure of the NCs, thus making it difficult to reproduce results for hybrid solar cells for a new NC batch. The batch-to-batch fluctuation of the semiconducting nanocrystals is a general problem of hybrid BHJ nanocrystal/polymer solar cells, which is originating from their colloidal wetchemical synthesis from organo-metallic precursors within organic ligands. During the NC synthesis the average NC diameter is constantly increasing, furthermore the NC size distribution, and the size of the NC ligand sphere and the related photoluminescence quantum yield varies. In order to eliminate these variations, one typically attempts to obtain identical NC batches by using the same synthesis time, and further fix every synthesis parameter. Nevertheless, the NC properties may still be different in between supposedly identically fabricated NC batches. These later batch-to-batch fluctuations (of e.g. NC size distribution or PL quantum yield) may derive from different precursor purity (e.g. water content), dissimilarities in the synthesis procedures (which can be minimized using an automated microwave synthesis see Chapter 4.1), and different ratios in between the educts (see Chapter 4.2).While different NC size distributions and NC sizes induce different solar cell power conversion efficiencies (see Chapter 4.1), different PL quantum yields cause differences in the required post-synthetic NC ligand reduction treatment time (or intensity) due to different ligand sphere constitution (see Chapter 4) even for the same ligand system and NC shape. Regarding this effect, it was shown that even if NCs of different initial PL quantum yields were utilized, the final power conversion efficiency of the hybrid solar cells

165 Summary can be similar, given that a shorter post-synthetic treatment time has been applied to the weaker luminescent NCs, compared to the stronger luminescent NCs (see Chapter 4.2). The later finding could be explained with the fact found for CdSe QDs that the PL intensity is linear dependent on the NC ligand sphere volume (see Chapter 4.4). Thus, a larger initial ligand sphere, which also exhibits a higher PL quantum yield, needs a longer post-synthetic treatment to be reduced far enough to allow for an efficient charge extraction and on the same time still provide colloidal stability to the NCs. The following Figure 7.1 shows schematically the relation found between the PL quantum yield and the ligand sphere size (diameter) - a linear correlation between PL quantum yield and ligand sphere volume (see Chapter 4.4) - together with the point of optimal NC ligand reduction, which is reached after different post-synthetic treatment times due to the fluctuating starting conditions. Figure 7.1. Varying optimal post-synthetic treatment time, due to different starting condition of the NC ligand sphere. During the treatment, the photoluminescence and ligand sphere diameter are constantly decreasing, as is the colloidal stability of the NC. The difference in the NC ligand sphere constitution throughout the synthesis was shown in Chapter 4.1. Therein the colloidal NC stability, visible in the degree of aggregation detectable by DLS for a fixed time after the post-synthetic NC treatment, decreases with increasing NC synthesis time, a behavior which is progressively more pronounced after a NC synthesis passes the PL bright point. Moreover, it was demonstrated that the initial colloidal stability of the NCs is decreasing during the NC synthesis. Furthermore, hybrid solar cells incorporating CdSe NCs of a low colloidal stability (induced by post-synthetic treatment)

166 7. Summary 152 tend to exhibit a low thermal stability of their open circuit voltage (V OC ), and a low parallel resistance (R P ), but exhibit a better thermal stability of the short circuit current (J SC ) and series resistance (R S ) - see Chapter 4.5. Also, it was demonstrated that solar cells from longer HA treated NCs, exhibit an increasingly lower initial V OC and parallel resistance, but show up to a certain treatment time a higher J SC and series resistance (see Chapter 4.3). By these studies it was also enabled to predict from the initial solar cell performance and a second subsequent measurement after a short thermal treatment, whether the utilized NCs received too short or too long post-synthetic treatment (see Fig. 7.2) enabling the fast detection of the optimal NC treatment time. Figure 7.2. Schematic for finding the optimal post-synthetic NC treatment time from solar cell parameters before and after thermal annealing: If the V OC and J SC are increasing after annealing (change indicated by arrows), the washing time in HA of the CdSe NCs was not sufficiently long. This increase is stronger for the same annealing time, the shorter the NCs were treated. Inversely, the same is also valid for too long HA washing times. Moreover, the FF prior to thermal annealing is increasing with increased NC washing time, providing an additional hint for finding the optimum post-synthetic NC treatment time. Also, different NC shape (NRs compared to QDs see Chapter 5.1) and different NC ligands (HDA/TOPO, HDO see Chapter 4.4) require different post-synthetic treatment. Finally, the influence of the post-synthetic hexanoic acid treatment on the NC surface was investigated via FTIR spectroscopy, showing indications for ligand exchange between the initial HDA/TOPO ligand and hexanoic acid (see Appendix A4).

167 Summary 7.2 Efficiency enhancement Within this thesis several approaches for improving the charge extraction in hybrid NC/polymer solar cells were followed, based on geometric considerations using elongated NCs (see Chapter 5.1), and regarding the NCs surface trap reduction and improved charge extraction utilization of a NC-graphene hybrid (see Chapter 5.2). But also, an attempt of building nanostructured solar cells with an interdigitated donor-acceptor interface for efficient exciton dissociation and charge extraction have been made (see Chapter 5.3) But also first investigations of the limitations for the hybrid solar cell s lifetime and its degradation have been made. The results of these attempts reached from a slight efficiency improvement from 2.7% to 2.86% (in case of the NR/QD mixtures presented in Chapter 5.1), over a considerable improvement of up to 4.24% (upon use of the QD/graphene hybrid presented in Chapter 5.2), to a very low efficiency of under 0.1% PCE (for nanostructured solar cells presented in Chapter 5.3). At first, in Chapter 5.1 elongated NCs (so called nanorods) were used as electron extraction material from the solar cell s active layer. However, only PCEs of up to 1.9% were reached (compared with 2.7% reached with quantum dots [125] ) due to the necessity of a relatively long post-synthetic treatment time, leading to a strong aggregation of the nanorods and requiring a filtering step to remove large aggregates. Nevertheless, by mixing the NRs with QDs, a PCE of up to 2.86% was obtained. The efforts in the hybrid BHJ solar cell optimization culminated in the utilization of a CdSe-QD/thiolated reduced graphene oxide hybrid material, which is chemically attached to the QDs and enabling an efficient charge extraction from the QDs. Therein, the thiol functionalized reduced graphene oxide is acting as ligand of the QDs, passivating its surface and reducing surface charge traps. Thereby a hybrid material was formed, displaying also geometrically more favored charge extraction paths in the hybrid BHJ solar cell s active layer, resulting in PCEs of up to 4.2% (see Chapter 5.2). The last presented effort for PCE enhancement of Chapter 5.3 consisted in the formation of a nanostructured interdigitated donor-acceptor interface. Hence, allowing exciton - in principle - each to reach the donor-acceptor interface, and thus to be splitted into an electron and hole, which would directly be extracted within the respective material out of the active layer. Therefore, at first vertical aluminum-oxide nanopores were successfully

168 7. Summary 154 created as nanostructuring scaffold for a photoactive material by anodic aluminum oxidation of a thin aluminum film, but without resulting in working solar cells (see Appendix A7). Yet, the approach of creating horizontal TiO 2 nanotubes on an ITO substrate by anodization of a thin titanium film as electron acceptor material resulted in working hybrid solar cells (see Chapter 5.3). However, probably mostly due to a large number of defects, visible in the absorption spectrum of the created TiO 2 nanotube layer (see Appendix A6), the reached PCE was of less than 0.1%. Nevertheless, these inverted, nanostructured hybrid solar cells proved to degrade very slow in air. 7.3 Solar cell degradation Finally, in Chapter 6.1 the degradation of hybrid BHJ solar cells was investigated during 1.5 years inside the glovebox (under nearly inert atmosphere, prevented from degradation due to UV-irradiation or heat), and as well as in air (see Chapter 6.2). Thereby, an improved long-term stability was found for hybrid BHJ solar cells, which incorporate NCs for which the original ligand sphere was strongly reduced (see Chapter 6.1). These solar cells exhibited a faster initial decrease of their parallel resistance, probably induced by NC aggregation leading to creation of surface defects favoring charge recombination, but showed a much lower and on the long term nearly stable series resistance, which might be explained by either a better passivation by hexanoic acid or by the more intimate contact between the NCs, compared to weaker washed NCs, which hinders the diffusion of oxygen or water to the NC surface. Hence, these solar cells showed after the initial PCE decrease from about 2.5% to about 1.8% within 1.5 months nearly no efficiency decrease for the next 15 months. In contrast, the decrease of hybrid BHJ solar cells containing shorter washed NCs was after the initial decrease from about 2% to 1.5% within 1.5 months much stronger and resulted in a decrease down to 0.5% PCE until the end of the long term investigation. Hybrid BHJ solar cells containing stronger and weaker ligand reduced CdSe QDs were also taken into air and compared with a BHJ PCBM:polymer solar cell (see Chapter 6.2). Thereby, an increase of V OC was detected for hybrid BHJ solar cells, but not for the purely organic solar cell. This V OC increase was assumed to be due to a passivation of the NC surface defects

169 Summary by oxygen. Whereby the simultaneous decrease in J SC and series resistance was attributed to the oxidation of the NC surface and to the oxidation of the aluminum top electrode, thus deteriorating the acceptor/metal interface (see Section 6.2.3). Moreover, the hybrid BHJ solar cells containing stronger ligand reduced CdSe QDs exhibited a strongly improved efficiency stability in air, which was then similar to the purely organic solar cell, and even slightly surpassed it. Nevertheless, this utilized non-encapsulated solar cell design incorporating the hygroscopic PEDOT:PSS layer and the easily oxidizable aluminum top electrode results in a fast PCE decrease in air. With the PCE decreasing down to 66% within the first 2.5 hours both for the more stable hybrid and for the organic solar cell, and down to 33% respectively to 2% of the initial efficiency within about one day in air for the hybrid respectively the organic solar cell. The cause of PCE decrease was mainly an increase of series resistance for all solar cells, which supports the prior supposed oxidation of the aluminum top-electrode but also the deteriorating effect of oxygen over the semiconducting polymer, as these materials were utilized in hybrid and as well purely organic solar cells. Moreover, upon use of inverted nanostructured PEDOT:PSS-free solar cells - utilizing a gold instead of an aluminum electrode - incorporating the same semiconducting polymer (i.e. PCPDTBT) and CdSe QDs, the increase of the V OC was found not to originate primarily from oxygen, but probably originating from passivation of the NC surface with water molecules present in air (see Chapter 6.3).

170 8. Outlook Outlook Despite leading to record PCE values for BHJ QD/polymer hybrid solar cells, the use of the CdSe-TrGO hybrid material brought up many questions regarding the interaction between CdSe and TrGO and also on the impact on the interaction with the semiconducting polymer inside the active layer. Herein further studies must be performed, like stated in the conclusion of the respective Chapter 5.2. This solar cell concept certainly has a high potential for greatly surpassing the already achieved performance by for example adding a hole blocking layer (either an organic [10] or metal-oxide [126, 128] layer) between the active layer and the metal cathode. This might further increase the PCE and also air stability of the solar cell. But also if a finer phase segregation would be achieved in the TrGO containing solar cell by choice of the suitable solvent or solvent mixture, the PCE of this new hybrid BHJ solar cell concept might further be improved, especially by increasing the short circuit current. Already, first experiments using a ZnO NC interlayer were performed and resulted in small efficiency enhancements of the solar cell, but the ZnO layer roughness and uniformity has to be enhanced in order to reach the full potential and reproducibility of utilization of this interlayer. Also, the inclusion of carbon nanotubes promises to improve charge extraction from the BHJ solar cell. First experiments with carbon nanotubes (CNTs) cut in an ultrasonic bath to avoid shortcuts throughout the active layer and thiol functionalized and CdSe NC decorated resulted in working solar cells, but exhibited PCEs of only slightly over 0.1% (not shown). Herein, further studies for improving the CNT dispersibility and of the CNT:NC:polymer ratio must be performed. Also important is to test hybrid solar cells in respect to potential applications. Therefore, the increase of the active area, investigations about the performance under indoor conditions (including different light intensities and spectra) must be performed. As well, the enhancement of the performance in air by use of protecting interlayers (i.e. metal-oxide layers) and inverted (PEDOT:PSS and aluminum free) solar cell design or suitable encapsulation processes must be explored. Some of this work was already done during this thesis; like presented in Figure 8.1, which shows the behavior of the hybrid BHJ solar cell performance with increasing active area. Therefore, solar cells incorporating strongly washed CdSe NCs were investigated. The J SC and FF are constantly falling by gradually

171 J SC [ma/cm²] FF V OC [V] PCE [%] Outlook increasing the active area size from 0.1 cm² to 1 cm². Only the open circuit voltage remains stable until up to 0.45 cm². The PCE is falling from 2.5% for the smallest tested solar cell to 0.7% for the 1 cm² active layer Active Area [cm²] Active Area [cm²] Figure 8.1. Left: Short-circuit current and fill factor of CdSe QD/PCPDTBT BHJ hybrid solar cells with active layer areas of 0.1 cm², 0.2 cm², 0.45 cm², 0.7 cm², and 1 cm². Middle: Open circuit voltage and power conversion efficiency of the same solar cells. Right: Image of a substrate utilized for this investigation with one solar cell of 1 cm² and one of 0.1 cm². This decrease might originate from an accumulation of shunts over the active layer due to inclusion of large NC aggregates which are more probable to be included for larger active areas. Thus, the film quality becomes very important for large area solar cells, and the importance to avoid NC aggregations might therefore even increase. Maybe large area solar cells would have to use a thicker active layer in order to decrease shunts, but certainly an additional hole blocking layer under the top electrode would also lead to a performance improvement. Also, a first measurement of the light intensity dependency of a hybrid BHJ CdSe QD/PCPDTBT solar cell under the sun spectrum was performed (see Fig. 8.2). Thereby, the J SC displayed a linear increase with the solar cell illumination intensity, and the V OC displayed a linear increase with the logarithm of the illumination intensity (not shown), like it is also known from literature [247]. More interesting is the development of the PCE, which is slightly increasing with decreasing illumination intensity down to 0.15 mw/cm² from 2.02% to 2.09% (see Fig. 8.2, left). The measured increase is due to an increase of the fill factor with decreasing light intensity (see Figure 8.2, right), which is also observed for fully organic PCBM/polymer solar cells [248].

172 PCE [%] FF 8. Outlook Light Intensity [mw/cm²] Light Intensity [mw/cm²] Figure 8.2. Left: Power conversion efficiency for different light intensities measured by the sun simulator spectrum for a CdSe QD/PCPDTBT BHJ solar cell. Right: Development of the fill factor of the same solar cell under the same conditions. However, the optimization for low light and indoor application also requires the characterization of the solar cell under different lamp spectra, utilized for indoor lighting and testing under much lower light intensities than 0.15 mw/cm² [249]. Also, a solar cell which performs well under low light conditions must have a high parallel resistance, as the leakage current gains a proportionally higher importance for low photocurrents. Furthermore, the reproducibility of highly efficient hybrid BHJ solar cells might further be improved by utilizing a more controlled setup for the NC synthesis, like the automated laboratory microwave rector setup shown in Chapter 4.1, connected with an in-situ determination of NC size and size distribution. However, a synthesis procedure, which would allow the creation of highly homogenous NCs would be desirable for obtaining high solar cell efficiencies. Hence, the one-pot CdSe synthesis in HDA/TOPO recently proposed by F.-S. Riehle [16] could be utilized. Moreover, it would be desirable to also automate the synthesis of the NC precursors in order to eliminate further uncertainties, causing variations in the NC quality. Thereby, one could further reduce the effort for adapting the post-synthetic treatment for each NC synthesis batch, easily enabling the creation of highly efficient hybrid NC/polymer solar cells. Thus, more time could be utilized in hybrid BHJ solar cell research for improvement of other parameters of the solar cell - like optimization of layer thicknesses for optimal absorption and charge extraction, solvent variation, influence of interlayers, active layer size increase, and other innovative approaches - for accelerating the performance improvement in this field.

173 159 Appendix Appendix A 1. Temperature dependency of the NC photoluminescence For measuring the temperature dependency of the photoluminescence (PL), a CdSe synthesis at 220 C was carried out for 10 min. During the (unaided) cool down of the sample the PL signal and the temperature were measured every 30 s. In Figure A.1 the extracted results for the PL peak position (Fig. A.1, left), PL Intensity (Fig. A.1, center) and the full width at half maximum (FWHM) (Fig. A.1, right) together with the respective calculated fitting curves are shown. Figure A.1. Left: Measured and fitted temperature dependence of the PL peak position; Center: Measured and fitted PL intensity; Right: Measured and fitted FWHM. The following Formulas (A.1, A.2, & A.3) and some of their fitting parameters for the temperature dependency of PL peak position, PL intensity and FWHM are taken from publications about the investigation of the temperature dependency of the respective parameters for semiconductors in the quantum confinement regime [250, 251], which are based on temperature dependency of the bandgap described by Y. P. Varshni [173]. For the PL peak position correction the following formula is taken for describing the temperature dependency of the bandgap (which is like previously mentioned linearly proportional to the PL wavelength) according to Varshni: E g T = E g 0 T2 (T+ ) (A.1),with E g being the bandgap, E g (0) being the bandgap a 0 K, T being the temperature in K, and α and Θ being fitting constants.

174 Appendix 160 Because measuring the PL peak position E g (0) at 0 K was not possible with our equipment, the values of = ev/k and =100 were taken from values published for CdSe NCs [250], the value for E g (0) was then chosen to fit the measurements. For the PL intensity correction the following formula was used: I T = I(0)/[1 + C exp( E a k B T )] (A.2),with I (0) being the PL intensity at 0 K, E a being the activation energy of a free exciton, k B being the Boltzmann constant, T being the temperature in K, and C being a fitting parameter. For the thermal activation energy of free exciton E a a value of 0.31eV was found to best fit the measurements. The intensity I(0) at 0 K was estimated at being 1.5 times higher than the values at 180 C. Then the constant C is presumed to be 140,000. For the FWHM correction the following formula was used: (T) = inh + T + LO (exp( E LO k B T ) 1) 1 (A.3),with Γ being the FWHM of the energy bandgap size distribution, σ being the excitonacoustic phonon coupling coefficient, Γ LO representing the strength of exciton LO (longitudinal optical)-phonon coupling, E LO being the LO-phonon energy, and k B being the Boltzmann constant. Based on values published for CdSe NCs [250] the following values were used to best fit the measurements: inh =48 mev, =10 µev/k, LO =132 mev and E LO =30 mev.

175 PL Intensity [CPS] 161 Appendix A 2. Ligand exchange experiments of CdSe NCs A 2.1 In-situ ligand exchange mapping The exchange of the original hexadecanol ligands of CdSe NCs synthesized by a one pot approach inside a microwave reactor (like described in Chapter 3.2.3) was attempted. The thereby occurring changes in PL intensity and dispersibility of the NCs, and the performance of CdSe-NC/polymer solar cells were investigated. At first, a ligand exchange inside the microwave reactor was performed. Thereby, the development of the PL intensity was recorded in-situ during the treatment process at 90 C inside the microwave reactor. Various solvents were used in which the NCs were dispersed at a concentration of 0.5 mg/ml (see Fig. A.2). 250k 200k 150k Butanoic Acid Hexanoic Acid Octanoic Acid Hexanol Octanol C 100k 50k 1 10 Time [min] Figure A.2. Left: Development of PL intensity of originally HDO capped CdSe NCs dispersed in different solvents at a concentration of 0.5 mg/ml and heated inside a microwave reactor at 90 C over 60 minutes. Right: PL intensities of the NCs directly after dispersion and 60 min later treated under stirring at 90 C for different solvents. Surprisingly, there was no PL decay observable for a treatment in hexanoic acid, like it was measured in the prior described HA washing (see Section 4.1.1) with a subsequent precipitation by methanol. This might mean that the usually conducted precipitation of the NCs by methanol (see Section 3.2.4) might additionally be needed to remove the original ligand sphere. Thus, reducing the PL intensity, and the prior HA treatment is only amplifying the impact of the MeOH washing by protonizing the HDO ligand, and making it soluble in the polar MeOH solvent, like described by Zhou et al. [15]. Also, for the other acids no PL decrease was observable during washing, but rather a constant increase of PL intensity. However, the

176 Appendix 162 original PL intensity displayed the highest increase when using octylamine or alcohols (hexanol and octanol) as solvents for the CdSe NCs. A 2.2 NC ligand and solar cell performance Also ligand exchange experiments with the usually applied approach of first dispersing the CdSe NCs in a solvent, and later precipitating them with methanol after a certain treatment time were conducted. Therefore, the CdSe NCs were dispersed with a concentration of 2 / 3 mg/ml in various solvents (see Fig. A.3), heated for 10 min at 90, and subsequently the double volume of methanol was added, followed by further stirring and heating at 90 C for 1 minute. Finally, the dispersion was centrifuged in order to precipitate the NCs, which were redispersed in chloroform for PL measurements (see Fig. Fig. A.3, left). Figure A.3. Left: Photoluminescence intensities of CdSe NCs dispersed in chloroform after 10 min at 90 C in various solvents and precipitation with methanol, compared to untreated CdSe NCs. Right: Subjective evaluation of solubility of CdSe NCs in chloroform (CHCl 3 ) and in chlorobenzene (CB) after treatment for 10 min at 90 C with various solvents. From the above graph one can see that when comparing the PL intensities of CdSe NCs treated with solvents of different functional groups (i.e. alcohol, amine, carboxyl, thiol) at the same alkyl chain length (i.e. of eight) different values are measured and compared to the untreated sample (see Fig. Fig. A.3, left). CdSe NCs treated with octylamine exhibit an increase of about 75% in their PL intensity, NCs treated with octanol show an increase of

177 163 Appendix about 30%, octanoic acid treated NCs show a decrease by about 15%, and CdSe NCs treated with octanethiol show a decrease by approximately 95% compared to the PL intensity of untreated CdSe NCs. These results are not surprising and are in accordance with other studies about the influence of the NC capping molecules on the CdSe NC PL intensity [ ]. For example, the very strong PL quenching achieved by octanethiol is both a result of the stronger binding of the thiol group to the Cd atom than that of TOPO [255] (implying a fast ligand exchange), but also due to the strong PL intensity decay (by 50%) of CdSe NCs already by a single octanethiol molecule on the surface of one CdSe NC [253]. Besides the octanethiol also pyridine induces a strong PL decay. However, pyridine treated NCs show a worse dispersibility (see Fig. A.3, right) at a higher PL intensity compared to ethanethiol. Thus, from Figure A.3 (right), which displays a subjective judgment of the dispersibility, one cannot directly conclude on the general ability for providing colloidal stability of a certain surface molecule, since the displayed dispersibility values derive from a fixed treatment time (10 min) and different ligands have different ligand exchange velocities. Now, also hybrid CdSe/PCPDTBT BHJ solar cells were built with the differently treated NCs. The following comparative values of the solar cells in Figure A.4 do not show the full potential of hybrid solar cells manufactured from CdSe NCs of the various surface treatments, but rather display the solar cell performance after 10 min of the specific treatment. Thus, for example the low V OC displayed by solar cells manufactured from octanethiol and pyridine treated NCs points towards a problem with shunts due to strong NC aggregation that occurred by a in this cases too long post-synthetic treatment rather than a general problem of the specific ligand molecule. Zillner et al. [256] have shown that e. g. the ligand exchange with pyridine is decreasing the PL intensity of the CdSe NCs, and is improving the charge separation within CdSe NC thin films, but the total overall extracted charges are diminished by created NC surface traps. Hence, ligand exchange is always a tradeoff between gaining NC conductivity and creating surface traps leading to charge recombination within the active layer. However, in case of octylamine, the V OC is on the same level like of hybrid solar cells from alcohol and acid treated NCs (approximately 0.7 V), like also the FF is at a similar level like devices containing alcohol treated CdSe NCs (ca. 0.25), but the short-circuit current is at a much lower level (see Fig. A.4, upper-left). This could be an indication that capping NCs with octylamine is reducing the conductivity of the

178 Appendix 164 aggregated nanoparticle network. The comparison in between the different alkyl chain lengths of the investigated alcohols and acids show a slight trend of longer alkyl chains (i.e. octanoic acid in direct comparison with hexanoic acid see Fig. A.4) leading to higher hybrid solar cell performance, even though the differences are low despite the butanoic acid treatment which led to bad NC dispersibility (see Fig. A.3, right) and no measureable photovoltaic performance (see Fig. A.4). Figure A.4. Solar cell performance values for CdSe NC/PCPDTBT BHJ hybrid solar cells containing CdSe NCs post-synthetically treated with various solvents for 10 min at 90 C. Bars of the same color represent direct comparisons from a solar cell batch manufactured under very similar conditions on the same day.

179 Absorption [AU] Photoluminescence [CPS] 165 Appendix A 3. Photoluminescence quantum yield calculation From UV-Vis absorption and PL spectroscopy the relative quantum yield (QY) of semiconducting QDs can be determined. The quantum yield is expressing the number of emitted photons relative to the number of absorbed photons. If one knows the quantum yield of a reference substance, the quantum yield of the examined substance can be determined using the following formula [167] : I OD QY = QY R R I R OD n 2 n R 2 (A.4),where QY is the quantum yield, I is the integrated PL intensity, OD is the optical density (also called absorption) for the PL excitation wavelength, and n is the refractive index. R in the index stands for reference. An exemplary measurement is described in the following. As reference, the Sulforhodamine 101 dye was used. It has a quantum yield of 1.0 at 25 C when solved in ethanol [257]. At first, its absorption and PL spectra were recorded (see Fig. A.5), while the maximum absorption was kept below 0.1 AU in order to avoid concentration quenching of the measured PL signal. The PL spectrum was recorded, with an excitation-wavelength of 575 nm corresponding to the absorption maximum of the reference dye M 2M M Wavelength [nm] Figure A.5. UV-Vis absorption spectrum (blue line) and photoluminescence spectrum recorded by excitation in the absorption peak of 575 nm (red line) of Sulforhodamine 101. Afterwards, the absorption of CdSe NCs in chloroform was recorded. And although it is not essential, the absorption at 575 nm was tuned to be the same as for the reference sample

180 Absorption [AU] Photoluminescence [CPS] Appendix 166 (i.e. of 0.1 AU). Thereby, the correction factor of the optical density in Equation (A.4) can be neglected. Here, it can be mentioned that the observed shift of the PL peak compared to the 1 st excitonic peak to a higher wavelength, results from an amount of energy lost due to vibrational relaxation of electrons after their excitation. This observed energy shift in between the absorption and the PL signal is called the Stokes shift [167]. Now, also a PL spectrum of the CdSe NCs was recorded by excitation at the same wavelength and intensity as the Sulforhodamine 101 dye (see Figure A.6). 3M 0.2 2M 0.1 1M Wavelength [nm] Figure A.6. UV-Vis absorption spectrum (blue line) and photoluminescence spectrum recorded by excitation of 575 nm (red line) of CdSe NCs synthesized by a hot injection synthesis at 300 C for 60 min using HDA, TOPO, CdSA, and TOP-Se in the molar ratio of 60:40:2:3. By integration over the PL signal of the reference dye and of the CdSe NCs the respective values can be inserted into Equation (A.4) together with the refractive index of ethanol and of chloroform, and the optical densities of the two solutions at 575 nm: QY CdSe = 1 42,127, ,545, = 0.45, resulting in a quantum efficiency of 45%.

181 167 Appendix A 4. FTIR measurements of CdSe QDs before and after HA washing With the herein described FTIR measurements it was attempted to get some insight into the influence of the hexanoic acid treatment to the CdSe NC surface. Therefore, first IR absorption spectra of the CdSe ligands (i.e. TOPO and HDA), and of hexanoic acid (HA) and methanol, which are used for the post-synthetic NC treatment are analyzed, in order to find out their specific fingerprints on the absorption spectrum. The structure of these molecules is given in the following Figure A.7. Figure A.7. Structural formula of TOPO (trioctylphosphine oxide), HDA (hexadecylamine), HA (hexanoic acid), and MeOH (methanol). Then, the recorded IR spectra of non-treated CdSe NCs and of HA treated CdSe NCs, precipitated with methanol are searched for indication of the presence of one of the prior analyzed molecules or for newly formed bonds with CdSe. Thereby the questions should be answered whether HA or MeOH bind to the surface of the CdSe QDs, and whether the original synthesis ligands are still detectable on the HA treated QDs.

182 Transmission [%] Appendix 168 A 4.1 FTIR spectrum of TOPO C-H C-H 3 2 P-CH 2 C-C C-H 2 P=O Wavenumber [cm -1 ] Figure A.8. IR spectrum of trioctylphosphine oxide (TOPO) and the most prominent detectable vibrations of its functional groups. The above TOPO IR spectrum shows at 2952 cm -1 [258] asymmetric stretching of C-H 3. Furthermore, it shows at 2919 cm -1 asymmetrical stretching, and at 2850 cm -1 symmetrical stretching of C-H [258] 2. The peak at 1462 cm -1 corresponds to P-C-H 2 deformation [259] or to C- H 2 scissoring [260], while the very prominent peak at 1147 cm -1 corresponds to P=O stretching [260]. The series of peaks between cm -1 belong a combination of C-H 2 rocking, C-C stretching and C-C-C deformation [259], while the peak at 723 cm -1 belongs to C- H 2 rocking along the carbon chain [260, 258]. For TOPO, a shift towards lower wavenumbers and strong signal broadening of the P=O stretching is expected upon binding to Cd on the CdSe NC, like it is reported for ZrO 2 NCs [260].

183 Transmission [%] 169 Appendix A 4.2 FTIR spectrum of HDA N-H 2 C-H 3 C-N N-H N-H 2 2 C-H 2 C-H Wavenumber [cm -1 ] Figure A.9. IR spectrum of hexadecylamine (HDA) and the most prominent detectable vibrations of its functional groups. For HDA the stretching peaks of C-H 3 and C-H 2, are at exactly the same positions like for TOPO. Only with the difference that symmetrical stretching of C-H 2 (2850 cm -1 ) is stronger than in TOPO, nearly leveling both peak intensities at 2919 cm -1 and 2850 cm -1. However, now characteristic peaks for the NH 2 groups appear. Such is the peak at 3250 cm-1, corresponding to N-H 2 antisymmetric stretching [261], and the peak at 3170 cm-1 corresponding to N-H 2 symmetric stretching [261]. Furthermore, the peak at 1467 cm-1 correspond to N-H 2 scissoring [261]. The peak at 1057 cm-1 corresponds to C-N stretching, and the peaks from cm-1 are assigned to N-H 2 wagging. Finally, again the sharp peak at 723 cm -1 appears, belonging to C-H 2 rocking along the carbon chain [260, 258].

184 Transmission [%] Appendix 170 A 4.3 FTIR spectrum of HA O-H C-H 2 C-H 3 OH... O C-O C-OH C=O Wavenumber [cm -1 ] Figure A.10. IR spectrum of hexanoic acid (HA) and the most prominent detectable vibrations of its functional groups. In Hexanoic acid a very broad peak can be observed at around 3000 cm -1 representing the O- H stretching [259]. HA also shows the 3 characteristic peaks of C-H 2 and C-H 3 stretching between cm -1, only with a relatively low intensity due to the shorter alkyl-chain compared to TOPO and HDA. For HA, the peak of the symmetric C-H 2 stretching is shorter than the asymmetric C-H 2 stretching peak. The by far most characteristic peak visible in the hexanoic acid IR spectrum is visible at 1705 cm -1 and originates from asymmetric stretching of C=O [259, 262]. The peak at 1415 cm -1 can be assigned to C-OH scissoring [259]. The peak at 1290 cm -1 represents the stretching vibration of the C-O bond. And finally, the strong broad peak around 929 cm -1 represents wagging of the OH group and the oxygen atom of the carboxylic group [259].

185 Transmission [%] Transmission [%] 171 Appendix A 4.4 FTIR spectrum of MeOH C-O-H O-H O-H C-H 3 C-O Wavenumber [cm -1 ] Figure A.11. IR spectrum of methanol (MeOH) and the most prominent detectable vibrations of its functional groups. For methanol the broad peak around 3670 cm -1 can be assigned to O-H stretch vibration [259]. The broad peak around 1370 cm -1 correlates with C-O-H bending(scissoring and rocking) [259]. The broad peak around 2950 cm -1 correlates to O-H stretching, and the peak at 2840 cm -1 corresponds to vibration of C-H 3. Finally, the broad peak around 1035 cm -1 correlates to C-O stretching [259]. A 4.5 FTIR spectrum of CdSe QDs 100 CdSe HDA/TOPO CdSe HDA/TOPO - HA treated C=O C-O Cd... O Wavenumber [cm -1 ] Figure A.12. IR spectrum of TOPO/HDA capped CdSe QDs, before (blue line) and after HA treatment (orange line), and the supposed vibrations of the most prominent differences between the two spectra.

186 Appendix 172 When comparing the IR signal of untreated CdSe QDs, synthesized in a 3:2 ligand mixture of HDA:TOPO by the hot injection method at 300 C (see Section 3.2.2), with HA treated NCs, precipitated with methanol, the first strong difference starting from low wavenumbers is the strong reduction of a peak around 700 cm -1, which might be attributed to the oxygen atom of TOPO, bound to Cd, thus forming a metal-oxide bond. The peak would fit to the assumption made in Section A7.1 that the sharp C=O peak would disappear and form a broad peak at a lower wavenumber. The reduction of this peak would thus mean the reduction of TOPO from the CdSe QD surface. The next large difference is the occurrence of a broad series of peaks around 1268 cm -1 in the HA treated QD sample, which might correlate with the C-O stretching, which was observed on a similar position for HA. Also, this peak did not occur for any other tested molecule on the IR spectrum near this position. Now, with increasing wavenumber the IR spectra of the HA treated and untreated CdSe sample are very similar, showing a strong peak at 1403 cm -1 and 1524 cm -1, until the occurrence of a double peak at 1637 cm -1 and 1725 cm -1. Within this region (i.e. at 1705 cm -1 ) there was the most prominent peak of hexanoic acid measured, namely the C=O stretching from the carboxylic group of the HA. Thus, this is another indication for the presence of hexanoic acid in the HA treated CdSe QD sample. Moreover, the broadening of the peak might indicate an interaction of the carboxylic group of HA with the surface of CdSe. The next obvious difference (besides a lower absorption of a very broad peak between cm -1 ) is the lower absorption of the two C-H 2 stretching vibrations at 2851 cm -1 and 2917 cm -1, for the HA treated CdSe sample, indicating a reduction of TOPO and HDA, which were the molecules with the strongest signals at these points due to their long alkyl chains. Thus, by this measurement, there is the indication of a reduction of the alkyl chains present around the CdSe NCs. Also, before the HA treatment, the two C-H 2 stretching peaks were of relatively similar intensity, which corresponds to the IR-signal of HDA, and is in contrast to the TOPO IR-absorption, wherein these two peaks were of clear unequal intensity. Thus, demonstrating, that more HDA is present in the untreated CdSe QD sample, like also expected from the fact that for the synthesis HDA and TOPO were mixed in a molar ration of 3:2. However, after HA treatment, these two C-H 2 stretching peaks are of unequal intensity, what is also the case for the IR spectrum of TOPO and of HA. Thus, one might conclude that

187 173 Appendix the HA treatment removes the HDA stronger than TOPO, especially, because the peak intensity of the peaks is higher than what one would expect from HA acid only. The last obvious difference between the IR spectra of HA treated and untreated CdSe sample is the very broad peak between cm -1 which was also observed for TOPO coated ZrO 2 NCs [260], but not commented in that publication. However, this broad peak completely disappeared for HA treated CdSe QDs, which might lead to the conclusion of a complete elimination of TOPO. But, which is very unlikely given the evidence present in the IR spectrum for a presence of TOPO also in the HA treated QD sample. With respect of methanol, there is no clear indication that methanol is present in both the treated or untreated CdSe samples, especially since all MeOH peaks might overlap with the peaks of the other investigated molecules. Moreover, exactly the most distinct MeOH peak, around 3670 cm -1 cannot be detected on the IR-spectra of both CdSe samples. A 4.6 Conclusion In conclusion, the FTIR spectroscopy measurements clearly show the presence of hexanoic acid on in the precipitated sample of HA treated CdSe QDs. There is also some indication for an interaction of the HA with the NC surface by the assumed broadening of the C=O stretching peak of the carboxylic group at around 1700 cm -1, and by the disappearance of the wagging of the OH group and the oxygen atom of the carboxylic group of the HA, which might derive from restriction of movement due to binding or coordination of the functional groups with the NC surface. Moreover, there is the indication of a strong reduction of TOPO from the reduced intensity of the supposed Cd-O peak at 700 cm -1. However, precise statements about the extent of HDA or TOPO reduction cannot be made, since the interactions of the multiple molecules included on the NC sample with each other and with the NC surface have unknown influence on the IR spectra. Nevertheless, the undertaken experiments and speculations might help for a future more detailed investigation of the NC surface ligands present after post-synthetic HA treatment.

188 Appendix 174 A 5. Solar cells containing CdSe NCs from a one-pot synthesis at 200 C Herein, shortly also the results of solar cells containing NCs of the 200 C one-pot microwavesynthesis in HDO are presented. For this synthesis the bright point occurred later - after 31 min, thus representing a more stable synthesis compared to the one-pot synthesis in HDO at 220 C presented in Chapter 4.1. NC properties: The properties of the different NC bathes, synthesized for different times are listed in the following Table A.1. Table A.1. FWHM, PL intensity and PL peak position for CdSe NCs synthesized for different times at 200 C (values calculated from measurement temperature to room temperature). NC Sample Synthesis time [min] FWHM [nm] PL intensity [CPS] PL peak position [nm] 200 A k B k C k D k Solar cell results: Without performing any of the experiments mentioned before concerning the finding of the maximal possible washing time, the NCs from the 200 C synthesis were also treated by a 3 min HA washing. The performance of the resulting hybrid solar cells is depicted in the following Figures A.13 & A.14 and in Table A.2.

189 FF Efficiency [%] Jsc [ma/cm2] Voc [V] 175 Appendix A 200 B 200 C 200 D Annealing Time [h] Annealing Time [h] Figure A.13. J SC and V OC development over thermal annealing at 145 C of hybrid solar cells containing CdSe NCs from a 200 C microwave one-pot synthesis washed for 3 min in HA and mixed with PCPDTBT. The short-circuit current is again the highest for solar cells manufactured from NCs taken from the synthesis bright point being at the same time the point of lowest FWHM. Noticeable is that the fastest J SC increase of PCE over thermal annealing is reached by NCs synthesized for the longest time (sample 220 D). For these NCs the conducted 3 min washing is probably already conducting to a strong reduction of the ligand sphere and inducing a low colloidal stability. An assumption also supported by displaying the fastest decrease of V OC during annealing. Interestingly, the final V OC values are inverse-proportional to the NC synthesis time, with NCs from short synthesis showing the highest V OC values. This could be both an effect deriving from a lower deterioration of the original surface ligands by the postsynthetic treatment, and maybe also from a smaller NC size, which would result in higher V OC due to the higher lying LUMO-level of the CdSe NC A 200 B 200 C D Annealing Time [h] Annealing Time [h] Figure A.14. Fill factor and power conversion efficiency development over thermal annealing at 145 C of hybrid solar cells containing CdSe NCs from a 200 C microwave one-pot synthesis washed for 3 min in HA and PCPDTBT.

190 Appendix 176 The fill factors of all solar cells are all of 0.36 before annealing, like the solar cells built from NCs of the 220 C NC synthesis (see Section 4.1.2). Thereafter, by annealing the FFs of solar cells containing the relatively weakly washed NCs are decreasing the most. Only the FF of the solar cells built from NC sample 200 D is increasing over thermal annealing, resulting also in the highest PCE. From these results one might conclude that the optimum washing times for all NCs from the synthesis points 200 A-C were not reached, but must be increased in order to obtain a higher FF and thereby a higher PCE. Hence, the PCE values reached with NCs of the 220 C synthesis (see Section 4.1.2) might also be achieved. Result summary: The best and average results of the solar cells containing NCs synthesized at 200 C are listed in the following Table A.2. Table A.2. Best and average results (indicated in brackets) from all solar cells after the indicated annealing time from of the four selected time points of the 200 C microwave synthesis. The NCs were all washed for 3 min in HA. NC sample Annealing C [h] J SC [ma/cm 2 ] V OC [V] FF PCE [%] 200 A (2.82) 0.71 (0.71) 0.25 (0.24) 0.75 (0.48) 200 B (5.27) 0.66 (0.68) 0.31 (0.30) 1.42 (1.10) 200 C (4.18) 0.66 (0.66) 0.36 (0.32) 1.15 (0.87) 200 D (4.01) 0.60 (0.62) 0.52 (0.50) 1.34 (1.25)

191 177 Appendix A 6. Anodization of nm titanium films on ITO Using titanium layers sputtered on ITO with thicknesses between 200 nm and 300 nm, resulted for the same anodization settings as used for the thicker films (see Chapter 5.3) in an inhomogeneous distribution of the top pore sizes (see Fig. A.15, left). However, now the resulting titania layers consist out of 2 stacked nanotubes layers (see Fig. A.15, right) Figure A.15. Left: SEM image of the surface of a 200 nm titanium film sputtered on ITO after anodization for 7 min at 40 V with 0.5 wt% NH 4 F in an ethylene glycol solution containing 5 vol% H 2 O. Right: SEM image of a sample created with the same settings, covered by a thin layer of polymer, broken into two pieces through the area exposed to the electrolyte, and tilt by 55 from the horizontal position. The red arrow marks the ITO layer, the blue and yellow layer mark the two formed nanoporous titania layers. When reducing the anodization voltage to 20 V, the pore distribution at the top surface is more homogenous (see Fig. A.16, left). However, from one of the very few cracks on the very homogenous layer surface one can see that there is still a second layer of nanotubes under the top layer (Fig. A.16, right).

192 J [ma/cm²] Appendix 178 Figure A.16 Left: SEM image of the surface of a 200 nm thick sputtered titanium layer anodization for 8 min at 20 V with 0.5 wt% NH 4 F in an ethylene glycol solution containing 5 vol% H 2 O. Right: SEM image of the same layer where a local delamination of the top titania layer uncovered a second TiO 2 nanotube layer below. For the now presented thinner titanium layers anodization has to be stopped earlier, i.e. the anodization process has to be stopped already after about 6-8 minutes for titanium layers of nm. Otherwise the complete dissolution of ITO follows, making the sample useless for solar cell application. Figure A.17 shows the current development during the anodization at room temperature. The current at the constant voltage of 40 V drops continuously with the transformation of titanium into titania. When the first pores reach the ITO substrate the current is increasing and subsequently rapidly dropping, supposedly after completing anodization of all titanium material in contact with ITO Anodization Time [s] Figure A.17. Left: Current development during anodization of a 200 nm thick titanium layer on an ITO substrate, anodized at 40 with 0.5 wt% NH 4 F in an ethylene glycol solution containing 5 vol% H 2 O for 3 different samples, stopped in different phases of the reaction. Right: Photos of the samples after the respective anodization before thermal annealing.

193 179 Appendix Energy-dispersive X-ray (EDX) spectroscopy measurements were performed on the three samples prior presented in Figure A.17 (see Table A.3). Table A.3. Results of EDX measurements of a 200 nm thick titanium layer on ITO, of 3 samples of the same initial titanium thickness anodized for different times (presented in Figure A.17) before and in case of sample no 2 - also after annealing for 2 h at 450 C, and of an untreated titanium layer on ITO. Given are weight percentages of relevant elements. Oxygen [wt%] Indium [wt%] Titanium [wt%] Fluor [wt%] Ti on ITO (annealed) The EDX measurements confirm the presence of oxygen in the anodized samples and also the reduction of the titanium fraction compared to an unanodized titanium sample on ITO. Flour, originating from the titanium oxidation process is present in the sample before annealing, but is removed after annealing. However, also the measured indium fraction is reduced compared to the untreated sample which might be an indication for dissolution of the ITO layer. Moreover, also the absorption spectrum in the center of the substrates no. 1-3 has been measured before and after annealing (see Fig. A.18).

194 Current [µa] Transmittance [%] Appendix before annealing after annealing Wavelength [nm] Figure A.18. Transmittance of TiO 2 nanotubes before and after annealing for 2 h at 450 C, taken from samples that were anodized for 3 different grades. 3 The first peak observed for the TiO 2 nanotubes (at about 320 nm) derives from the TiO 2 bandgap. The peaks at higher wavelengths probably originate from impurities, as TiO 2 at no crystallite size exhibits such a low bandgap to fit those peaks (see Fig. 2.7), but they are reported to occur by carbon doping [263]. It must be stated that theses described impurities also occurred in UV-Vis absorption spectra from TiO 2 nanotube arrays prepared by the two step anodization described in Section with different defect peak positions and also like for the one step anodization with varying intensity. A substrate for which the annealing was stopped just in the occurring current peak (like sample #2 in Fig. A.17 is given as example of optical transparency enhancement by thermal annealing (see Fig. A.19, left & middle). Also, besides the transparency enhancement, the annealed ATO substrate displays a diode behavior after thermal annealing (see Fig. A.19, right) Voltage [V] Figure A.19. Anodized TiO 2 nanotube substrate on glass/ito before (left) and after (middle) annealing. Right: Current-voltage curve measured over the annealed substrate.

195 J [ma/cm²] 181 Appendix Also, the TiO 2 nanotube walls were decorated with CdSe NCs prior to polymer filling (or to filling with NC/polymer mixture). The decoration by a mercapto-carboxylic acid would provide a binding site to the TiO 2 and to the CdSe. The functionalization on TiO 2 is reported to be more efficient when mixed with a carboxylic acid [264]. Shorter Acids would probably allow for a better charge transfer between CdSe and TiO 2. However, the acid treatment diminished the solar cell performance and a short treatment time results in poor CdSe decoration (measured by UV-Vis absorption). A performance improvement was found when adding a PEDOT:PSS layer on top of the active layer. The spin coating could be done by addition of 0.5 vol% Triton-X100 to the PEDOT:PSS solution to allow wetting of the aqueous solution over the polymer layer. The performance is improved by increasing the parallel resistance R p. The greatest improvement was achieved by filling the polymer solution on which the drying step after spin coating was not performed, but instead heating under vacuum in nitrogen atmosphere was applied. However, the highest achieved PCE was of 0.024% of the TiO 2 /polymer device (see Fig. A.20, left) manufactured with the one step anodization (see Fig. A.20, right). 0.1 Voc (V) Jsc(mA/cm2) FF PCE (%) No Vac Vacuum No vacuum Vacuum Voltage [V] Figure A.20. Left: Picture of a glass/ito/tio 2 nanotube/pcpdtbt/pedot:pss/au solar cell. Right: Current-density voltage graph for the same solar cell for which a vacuum filling step was applied (red curve) and for the solar cell for which no vacuum filling was applied (blue curve).

196 J [ma/cm²] Appendix 182 A 7. Towards nanostructured hybrid solar cells using an alumina nanopore scaffold A 7.1 Formation of alumina nanopores on ITO Vertical nanopores out of Al 2 O 3 are formed by anodic aluminum oxidation from aluminum deposited by thermal evaporation on ITO substrates, like described in Chapter 5.3 (theory) and Section (experimental execution). For the creation of porous alumina, a high deposition rate of 1 nm/s compared to 0.2 nm/s and a high annealing temperature resulted in a better film quality for the anodization process. This conclusion could be drawn from a higher final anodization current density, meaning that more pores reached the underlying ITO layer and from a higher probability of success of horizontal pore formation (observed by SEM imaging) by using an aluminum film annealed at 450 C and above instead of 310 C. An explanation for this finding could be the higher grain sized obtained by higher deposition rate from thermal aluminum evaporation [265] and also the increased grain size from thermal annealing [266], thereby limiting the possible pathways for the pore growth in other directions than the vertical one along the grain boundaries. In the following Figure A.21 a typical anodization curve of the optimized anodic aluminum oxidation procedure is shown. The result was obtained from a 0.3 M aqueous solution of oxalic acid under a constant voltage of 40 V applied to a 200 nm thick aluminum film thermally evaporated with 1nm/s and annealed for 500 C for 1h Time [s] Figure A.21. Anodization curve of a 200 nm thick aluminum film on ITO by using a 0.3 M oxalic acid solution and an applied voltage of 40 V.

197 183 Appendix The anodization current is decreasing by the transformation of aluminum into Al 2 O 3, decreasing the conductivity, once some pores reach the ITO substrate the current is again increasing. When a peak current is reached most of the pores reached the ITO substrate with only an alumina barrier layer remaining between the electrolyte and the ITO solution. Once the current drops the underlying ITO layer is also dissoluted at some spots. Here the anodization process should be stopped preserve the original ITO layer, needed for the functioning of the later manufactured solar cells. In the following figures SEM images of Al 2 O 3 nanopores samples prepared by the above mentioned settings are presented. Figure A.22. SEM images: top view of Al 2 O 3 nanopores created at 40 V in a 0.3 M oxalic acid solution before (left) and after (right) pore widening for 60 min by 5 wt% H 3 PO 4 at room temperature. After the anodization process, the measured average interpore distance was extracted to be 81 nm, and the average pore diameter was determined to be 23 nm. The thereby obtained porosity before pore widening is of 7.3%, which means that the ratio between metal-cation dissolvation and H 2 O dissolution, introduced in Chapter 5.3 by the variable n, is of 44. At a lower water dissociation rate, i.e. by decreased anodization voltage, a higher porosity could be reached in theory. However, a subsequent pore widening, in order to open the bottom of the alumina pores towards the underlying ITO layer, is conducted removing the need of higher initial porosity. The pore widening by immersion into a solution of 5 wt% phosphoric acid was conducted for 60 minutes, increasing the pore diameter to an average value of 48 nm, corresponding to an alumina dissolution rate during pore widening of 0.2 nm/min, and thereby increasing the porosity to 41.3%.

198 Appendix 184 Also SEM images - showing the cross-section of the obtained Al 2 O 3 nanopore layer on ITO - were taken (see Figure A.23) by breaking the anodized substrate into two pieces through the anodized area. Figure A.23. SEM images of 25 (left) and 45 (middle) tilt samples of Al 2 O 3 nanopores on ITO after pore widening for 60 min by 5 wt% H 3 PO 4 at room temperature. Right: Photo showing the Al 2 O 3 nanopore layer on an ITO substrate. From the above exposed images apparently the pores were opened by the pore widening process towards the ITO substrate. Thus, allowing direct contact with the underlying ITO layer for materials, which are subsequently filled into the vertical pores. A 7.2 Polymer pore-filling experiments After creating the nanostructured alumina pores, filling with a P3HT polymer solution was attempted by heating a spin coated polymer solution at 90 C for 5 hours. The taken SEM images of a substrate broken into 2 pieces (see Fig. A.24, left) revealed no clear evidence for pore filling from a cross sectional image. However, from a part of the polymer layer which delaminated from the nanopore array during the breaking of the substrate into two pieces and folded upwards so that the side formerly showing towards the nanotube array could be examined, revealed a rough surface with little spikes with heights of about 50 nm with distances to each other, corresponding to the Al 2 O 3 interpore distances (see Fig. A.24, right).

199 185 Appendix Figure A.24. SEM images of 25 (left) and 45 (right) tilt samples of Al 2 O 3 nanopores on ITO after pore widening for 60 min by 5 wt% H 3 PO 4 at room temperature. The work on the alumina nanopore array was stopped here after an experiment in which the pores were coated with TiO 2 by ALD (deposition done by Dr. Firat Gülder, Laboratory for Nanotechnology, IMTEK, University of Freiburg) and only an ohmic contact could be measured after subsequent polymer filling. Also, at the same time the creation of titania nanotubes arrays was advancing and started giving more promising results (see Chapter 5.3). Moreover, in the meanwhile a work of a nanostructured solar cell of 0.5% PCE was published by another group [133] utilizing the same attempted concept of Al 2 O 3 pores coated by ALD by TiO 2 and filled by P3HT. Nevertheless, the creation of nanostructured solar cells based on the alumina nanopore array still has potential for further developments of a variety of designs and process approaches. One large disadvantage of the Al 2 O 3 pore formation on ITO is the poor adhesion of the layer, which leads to regular occuring delamination of nearly spherical areas of several µm in diameter throughout the whole array. A problem, which might be solved by deposition of conducting interlayers between ITO and aluminum.