LIFETIME AND RELIABILITY OF POLYMER SOLAR CELLS

Transcription

1 LIFETIME AND RELIABILITY OF POLYMER SOLAR CELLS A DISSERTATION SUBMITTED TO THE DEPARTMENT OF MATERIALS SCIENCE AND ENGINEERING AND THE COMMITTEE ON GRADUATE STUDIES OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY Craig Homer Peters November 2011

2 2011 by Craig H Peters. All Rights Reserved. Re-distributed by Stanford University under license with the author. This dissertation is online at: ii

3 I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy. Michael McGehee, Primary Adviser I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy. Reinhold Dauskardt I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy. Alan Sellinger Approved for the Stanford University Committee on Graduate Studies. Patricia J. Gumport, Vice Provost Graduate Education This signature page was generated electronically upon submission of this dissertation in electronic format. An original signed hard copy of the signature page is on file in University Archives. iii

4 Abstract The power conversion efficiency of organic photovoltaic (OPV) cells has increased from 4-5% in 2005 to 8.3% in The goal of a 10% single junction OPV device seems attainable making the commercialization of OPV more realistic. With advances made on the efficiency front, the lifetime and reliability of OPV devices has come into focus. To date there has been considerable work done in understanding and quantifying the lifetime and degradation of bulk heterojunction solar cells (BHJs) based on poly-(paraphenylene-vinylene) (PPV) and poly(3-hexylthiophene) (P3HT) polymers. A comparison of OPV lifetime experimental results across different research groups has posed challenges due to the lack of standardized testing and reporting procedures; however, great strides have been made in this regard during the most recent International Summit on OPV Stability (ISOS-3). Modules based on P3HT/fullerene BHJs have shown lifetimes of 5,000 hours when state-of-the-art encapsulation with a glass-on-glass architecture is used. Assuming negligible degradation in the dark and 5.5 hours of onesun intensity per day, 365 days per year, this translates into an operating lifetime approaching three years. More recently P3HT/PCBM devices utilizing an inverted architecture have been shown to retain more than 50% of their initial efficiency after 4,700 hours of continuous exposure to one-sun intensity at elevated temperatures and have exhibited a long shelf life when stored in the dark in ambient conditions. However, results to date have yet to show polymer based OPV lifetimes greater than 3-4 years. ii

5 In my dissertation I present a detailed operating lifetime study of encapsulated solar cells comprised of poly[n-9'-hepta-decanyl-2,7-carbazole-alt-5,5-(4',7'-di-2-thienyl-2',1',3'- benzothiadiazole) (PCDTBT) in BHJ composites with the fullerene derivative [6,6]- phenyl C 70 -butyric acid methyl ester (PC 70 BM). PCDTBT/PC 70 BM solar cells achieved an efficiency greater than 6%, making this one of a small number of polymers able to achieve this level of performance. I describe an experimental set-up that is capable of testing large numbers of solar cells simultaneously, holding each device at its maximum power point while controlling and monitoring the temperature and light intensity. Using this set-up we were able to compare the PCDTBT/PC 70 BM system with the well-studied P3HT/PCBM system and demonstrate a lifetime for PCDTBT devices that approaches 7 years, which is the longest reported operating lifetime for a polymer based solar cell. I will further present a systematic study of the burn-in degradation mechanism behind PCDTBT:PC 70 BM solar cells. I will show that a photochemical reaction in the photoactive layer creates states in the bandgap of PCDTBT. These sub-bandgap states increase the energetic disorder in the system, which reduces the FF, V oc and to a lesser extent J sc. The photochemical reactions are shown to progress rapidly when first exposed to light but subsequently decrease in occurrence, which results in the stabilization of the V oc and FF. iii

6 Acknowledgements First and foremost I would like to thank Professor Michael McGehee. Mike looked at an older graduate student and had the foresight and guts to take me into his group. He provided an exceptional environment for me to develop my scientific skills and pushed me to think about my work in a bigger picture manner. Mike taught me how to give compelling scientific talks and how to write exceptional scientific publications. Finally, Mike became a close friend through the process. As for the research itself, the saying that, it takes a village to raise a child, should now read, it takes a village to perform a lifetime and reliability experiment. I have to thank the incredibly hard work that was put in by Toby Sachs- Quintana. Toby worked countless hours and helped in all aspects of the work presented. Jack Kastrop also worked for hours on end soldering leads, building reflectors and testing solar cells. There is no doubt that without the two of them the first lifetime study would not have been launched as early or as well. I also have to thank Billy Mateker and Thomas Heumueller for their hard work in getting the second paper to print. Both worked tirelessly on building testing equipment and taking measurements into the weekends. I would like to thank my labmates who offered countless hours of discussion and brainstorming. Zach Beiley and Eric Hoke have been co- author on one of my papers and have both helped me understand energetic disorder and models associated with these concepts as well as charge- transfer states and the implications on device parameters. Eric has built many of the characterization tools that were used to take the data presented in this thesis. George Burkhard has been a go- to iv

7 person for me when trying to tackle complex physics problems. George was also a consultant for most of what was constructed to perform the reliability experiments. I- Kang Ding was my neighbor and we had numerous discussions about solar cells and the greatness of badminton even though I wasn t great at badminton. The rest of the McGehee group, including G- 5, Nicky, Roman, Mark, Sam, Jon and all of the newest members, made the experience one of the most memorable in my life. I also have to thank Shawn Scully for developing my understanding of the physics of polymer solar cells. He was my mentor for the first 1.5 years of my PhD and provided me with incredibly valuable insight into organic PV. Numerous discussions with Michael Rowell were also critical to my understanding of OPV and I want to thank him for keeping me healthy by going off to the gym or volleyball. Brian Hardin was instrumental in keeping the pressure on the lab to push itself to new levels. His enthusiasm and creativity were valuable to everyone around him. He also initiated numerous Ab- offs which helped me continue to get in shape. He is now a trusted partner in our new business and will continue to be a life- long friend. I need to thank my amazing wife, Kathryn, for being so incredibly supportive in my final year of graduate school. She pushed me to be exceptional in every way and gave me the time and love that only comes from a true life partner. My family has been there for me at every step of this process. From going back to undergraduate studies at the age of 31 to completing a PhD at the age of 41 they have lifted me up the whole way through. I want to especially thank my mother for being my greatest cheerleader and Harrison Griswold, one of my fathers, who s love v

8 of science and engineering inspired me to enter this field. He has promoted my endeavor down this path from the beginning and been there for me through the good and tough times. Finally, Eric and Mara, my oldest brother and his wife, deserve a special thanks. They inspire others to follow their dreams and not the money. They have never let up their support of me and continue to push me to achieve greater things with my life. vi

9 Dedication This work is dedicated to my father, Harrison Griswold, who instilled in me a love of science and to anyone else who wants to pursue their dreams regardless of their age. vii

10 Table of Contents Abstract... ii Acknowledgements... iv Dedication... vii List of Figures... xi 1. Background and Motivation The reliability of organic solar cells comes into focus The similarity of OPV and organic light emitting diodes (OLEDs) A brief look at the lifetime of state-of-the-art OLEDs State-of-the-art lifetime studies for OPV Small molecule OPV Polymer OPV Fundamentals of Polymer Solar Cells BHJ morphology Optoelectronic operation of BHJ solar cells Light absorption Exciton diffusion and charge transfer Charge separation and extraction Degradation behavior in OPV Photobleaching Deep trap formation Phase separation Shallow trap formation Delamination General Decay Behavior of OPV Devices Experimental set-up for lifetime measurements Background viii

11 4.1.1 Overview of aging apparatus Light source and reflectors Heating Stage and Solar Cell Mount Electronics and Connectors Laser beam induced current State-of-the-Art Lifetime Study Background and methodology Efficiency decay over time and lifetime calculation Ordered lifetimes Decay of J sc, FF and V oc Decay of key device characteristics for each solar cell Laser beam induced current maps Summary of lifetime experiment Analysis of Burn-in Loss Background of burn-in loss Approach to analyzing the decay mechanism in solar cells Variation of electrodes Morphology: Grazing Incidence X-ray Diffraction (GIXD) Morphology: Sub-bandgap external quantum efficiency (EQE) Light or temperature induced chemical reactions Photochemical reactions in the polymer Background on charge transport models PCDTBT:PC 70 BM hole-only diode preparation and aging PCDTBT:PC 70 BM hole-only diode testing and analysis ix

12 6.5.4 Background: Photothermal deflection spectroscopy (PDS) Film fabrication and aging PDS results Summary of burn-in study Identifying the photochemical reaction pathway Background Infrared Fourier Transform Spectroscopy (FTIR) Sample preparation and aging Results of photo-oxidation study using FTIR FTIR results of PCDTBT films aged in nitrogen under one-sun intensity Impurity detection Conclusion Summary of work Future direction Closing thoughts x

13 List of Figures Figure Schematic of a) an organic solar cell and b) an organic light emitting diode Figure 1.2. a) The first OLED device architecture and b) a modern OLED architecture Figure 1.3. Reported lifetimes of P- LEDs developed by Sumitomo Chemical Figure 1.4 Efficiency over time for encapsulated small molecule solar cells. The devices were illuminated with one- sun intensity using a sulfur plasma lamp Figure 1.5 Efficiency over time for encapsulated polymer (P3HT:PCBM) solar cells. The devices were illuminated with one- sun intensity using a sulfur plasma lamp Figure 1.6 Chemical structure for P3HT and structures of newer, more efficient, polymers Figure 2.1 Schematic intercalated (left) and unintercalated (right) morphologies of the polymer- fullerene bulk heterojunction active layer Figure 2.2 Solar cell energy levels (increasing energy going up the page). A blue photon is incident on the solar cell and subsequently absorbed by the polymer Figure 2.3 (Left) Schematic of the exciton after diffusing to the donor/acceptor interface. (Right) The exciton after charge transfer, creating a charge- transfer state between the electron and hole Figure 2.4 Schematic of the extracted charges after charge separation and transport drives the electrons to the cathode and holes to the anode Figure 3.1 (Top) Schematic of a photon passing through the polymer solar cell. (Bottom) Photographs of a P3HT film on glass before (left) and after (right) exposure to light in air Figure 3.2 Schematic of an electron in a deep trap about to recombined with a hole in the HOMO of the polymer Figure 3.4 (a) Schematic of shallow traps in the polymer resulting in space- charge due to a trapped hole. (b) J- V curves showing the degradation in performance of a solar cell after shallow traps are gradually introduced into the polymer phase Figure 3.5 (a) Schematic of delamination of the cathode from the active layer. (b) Laser beam induced current maps of a fresh (left) and aged (right) P3HT:PCBM solar cell aged under one- sun intensity in air for 5 days Figure 3.6. Typical decay curve of a polymer solar cell (solid black line) employing a standard architecture with an organic hole transporting layer as the anode and a metal (e.g., Ca/Al) as the cathode. The lifetime (dashed black) is defined by the point at which the efficiency has dropped by 20% from the start of the linear decay period. Both axes are linear Figure 4.1 J- V curve of an organic solar cell. The key device characteristics are shown Figure 4.2 Diagram of an aging apparatus Figure 4.3 Comparison of light sources xi

14 Figure 4.4 (a) Photograph of the 5300K sulfur plasma lamps and (b) the spectra of the lamps at various temperatures. For all experiments presented in this dissertation either a 5300k or 6200k lamp was used Figure 4.5 Photograph of the 5300K sulfur plasma lamp bulb and internal reflector (left) and the advertised uniformity of the lamp intensity (left) Figure dimensional plot of the lamp intensity versus position at a working distance of 20 from the lamp Figure 4.7 Diagram of the heated plate and solar cell mounting system Figure 4.8 Photograph of the heated plate and solar cell mount Figure 4.10 Photograph of LBIC apparatus Figure Plot of the efficiency versus time for the average of the eight PCDTBT and eight P3HT devices Figure Efficiency decay for PCDTBT (red) and P3HT (blue) solar cells. Data is plotted every 100 hours and averaged over 8 devices of each type. The linear fits use hourly data from 1300 to 4400 hours and lifetimes (T80) are extrapolated out to where the efficiency drops to 80% of the efficiency value at 1300 hours. The lifetimes are calculated by assuming 5.5 hours of one- sun intensity per day, 365 days per year Figure Ordered lifetimes of 16 devices. PCDTBT (red, 9-16) devices show substantially longer lifetimes but with a wider spread due to variation in Jsc over time while P3HT (blue, 1-8) solar cells show a narrower spread. Linear fits were used for each device using 3100 data points from 1300 to 4400 hours and the lifetime was defined by the point where the initial efficiency at the 1300 hour point drops by 20% Figure Device characteristics for PCDTBT (red) and P3HT (blue) solar cells over 4,400 hours of continuous testing. The curves are each normalized by the initial value at the start of the aging process. Each point represents the average of 100 hours of data for 8 solar cells of each type Figure Jsc (asterisk), Voc (open diamond), FF (open square) and efficiency (open circle) for eight P3HT/PCBM devices used in this study. The values of each have been normalized to the initial values at the start of the aging process Figure Jsc (asterisk), Voc (open diamond), FF (open square) and efficiency (open circle) for eight PCDTBT/PC70BM devices used in this study. The values of each have been normalized to the initial values at the start of the aging process Figure LBIC images of solar cells before and after aging under one- sun light intensity. Each image is a square area of 3.6 mm by 3.6 mm. (a) and (b) are images of a P3HT device before and after aging continuously for 4400 hours. The images show no loss of device area and the formation of smaller spots of low current. (c) and (d) are images of a PCDTBT solar cell before and after aging. No loss of device area is seen. (e) and (f) show the effects of encapsulation failure on a PCDTBT solar cell made identically to that in (c) Figure Efficiency decay for PCDTBT (red) and P3HT (blue) solar cells over 4400 hours of continuous testing with the burn- in period shown in darkened region. The curves are each normalized by the initial value at the start of the aging process. Each point represents the average of 100 hours of data for 8 solar cells xii

15 of each type. The error bars show the highest and lowest values at each point Figure Efficiency loss over time for PCDTBT:PC70BM devices using various electrodes Figure 6.4 Device characteristics over time for PCDTBT:PC70BM devices using various electrodes Figure 6.5 (a) Experimental set- up for grazing incidence x- ray diffraction. (b) 2- d image for a PCDTBT film on silicon showing the π- π stacking diffraction peak in the out- of- plane direction. The cake segment is shown around the peak Figure 6.6 (a) Integrated cake segment over different aging periods under one- sun intensity at 40 o C. The peaks have been normalized for comparison of peak width. (b) Integrated cake segment of PCDTBT films that have been annealed at different temperatures. The peaks have been normalized for comparison of peak width Figure 6.9 I- V curve of a fresh and aged PCDTBT solar cell Figure 6.10 EQE curves of a fresh and aged PCDTBT solar cell (left axis). The normalized absorption profiles for PCDTBT and PC70BM are shown at the bottom (right axis) Figure 6.11 Hole- only diode employing PCDTBT with high work function contacts on both sides of the device to enable hole- only injection Figure 6.12 (a) Hole- only diode employing PCDTBT with high work function contacts on both sides of the device to enable hole- only injection. The red crosses represent electronic states. (b) Density of states beginning at the HOMO and exponentially decreasing into the bandgap of the polymer with a characteristic width, Et. The Fermi energy is show by the dashed line Figure 6.13 (a) Log- log plot (circles) and fits (dashed lines) of the I- V curves for a PCDTBT hole- only diode at various temperatures. (b) Log- log plot (circles) and fits (dashed lines) of the I- V curves for PCDTBT hole- only diodes at various active layer thicknesses Figure I- V curves in forward bias (solid) and curve fits (dashed) for a hole- only diode as a function of aging time Figure The diamonds (left axis) represent the characteristic width of the exponential distribution of traps (Et) in PCDTBT:PC70BM hole- only diodes extending into the bandgap from the HOMO of the polymer. The hole- only diode was aged at 40 o C in a nitrogen filled box under one- sun intensity. The triangles (right axis) represent the efficiency loss of PCDTBT:PC70BM devices aged in a similar manner to the hole- only diodes Figure 6.16 (a) Experimental set- up for PDS. (b) Absorption spectra taken a various aging times. Samples were aged under one- sun intensity in a nitrogen filled chamber Figure 7.1 Possible reaction pathways for PCDTBT Figure 7.2 (a) FTIR spectrum of a fresh film of PCDTBT on silicon. (b) FTIR spectrum of the same film that has been photo- oxidized. (c) The spectrum of the aged minus the fresh xiii

16 Figure 7.3 FTIR spectrum of a fresh (purple) and aged (pink) film of PCDTBT on silicon after 10 days of exposure to one- sun intensity in a nitrogen filled chamber Figure 8.1 Summary of key findings in the understanding of the lifetime and degradation mechanisms in PCDTBT:PC70BM solar cells xiv

17 1. Background and Motivation 1.1 The reliability of organic solar cells comes into focus The power conversion efficiency of organic photovoltaic (OPV) cells has increased from 4-5% in 2005[1, 2] to 7.4%[3] and 8.3%[4] in The goal of a 10% single junction OPV device seems attainable[5] making the commercialization of OPV more realistic. With advances made on the efficiency front, the lifetime and reliability of OPV devices has come into focus[6, 7]. To date there has been considerable work performed in understanding and quantifying the lifetime and degradation of bulk heterojunction solar cells (BHJs) based on poly-(para-phenylene-vinylene) (PPV)[8-11] and poly(3-hexylthiophene) (P3HT) polymers[12-15]. A comparison of OPV lifetime experimental results across different research groups has posed challenges due to the lack of standardized testing and reporting procedures; however, great strides have been made in this regard during the most recent International Summit on OPV Stability (ISOS-3). Modules based on P3HT/fullerene BHJs have shown lifetimes of 5,000 hours when state-of-the-art encapsulation with a glass-on-glass architecture is used[16]. Assuming negligible degradation in the dark and 5.5 hours of one-sun intensity per day, 365 days per year, this translates into an operating lifetime approaching three years. More recently P3HT/PCBM devices utilizing an inverted architecture have been shown to retain more than 50% of their initial efficiency after 4,700 hours of continuous exposure to one-sun intensity at elevated temperatures[17] and have exhibited a long shelf life when stored in the dark in ambient conditions [18, 19]. However, results to date have yet to show polymer based OPV 1

18 lifetimes greater than 3-4 years. This has made the quest for a deeper understand of the degradation mechanisms in the higher efficiency organic solar cells important. 1.2 The similarity of OPV and organic light emitting diodes (OLEDs) In order to gain insight into the lifetime and reliability of OPV it is helpful to first look at the status of organic light emitting diodes (OLEDs). The reason for looking at OLEDs is in the similarity of device architecture and materials. Figure 1.1a shows a simplified schematic of a BHJ OPV device that employs a standard architecture. Under normal operating conditions light enters the solar cell through the glass and is subsequently absorbed by the organic light absorbing layer. The charges are then separated generating free charges in the device. The electrons and holes are driven to the cathode and anode, respectively, with the help of an internal electric field. The charges can then be extracted and used to generate power. More detail on this process will be provided in subsequent sections. Figure 1.1b shows a simplified standard architecture for an OLED. Under normal operating conditions an external power source injects electrons and holes from the electrodes into the light emitting layer. Once in the light emitting layer the charges are driven by an internal electric field until they encounter each other and recombine, a process which can emit light through the glass. 2

19 Figure Schematic of a) an organic solar cell and b) an organic light emitting diode. In this regard it is not unreasonable to say that an OLED is an OPV device that is operated under a different voltage bias. In the case of both an organic solar cell and organic LED, organic semiconductors are used to carry charges and similar electrode materials (e.g., aluminum, indium tin oxide) are used. Additionally, in both OPV and OLED devices the current densities are comparable. For these reasons, understanding the lifetime of state-of-the-art OLEDs can give us a rough guide as to what we can expect for a lifetime in organic solar cells in the future. 1.3 A brief look at the lifetime of state-of-the-art OLEDs The first OLED was developed in 1987 by Tang, et al[20]. Figure 1.2a shows a schematic of the first OLED, which is very similar to what was shown in Figure 1.1b. The early lifetimes were on the order of seconds to minutes before complete failure of the device and the conversion efficiency of injected charge carriers to emitted photos was 3

20 much less than 1%. Figure 1.2b shows a schematic of a modern OLED device. Though there have been significant enhancements to boost both the efficiency and lifetime of the device, the general operating principle is the same. Figure 1.2. a) The first OLED device architecture and b) a modern OLED architecture. Given the stage of development for OLEDs most of the current research and development efforts are performed by industry. Modern OLEDs are made from both small molecule and polymer semiconductors. Small molecule OLEDs (SM-OLEDs) have been commercialized while polymer OLEDs (P-LEDS) are still in the development stage. In a recent report Samsung claimed that state-of-the-art glass encapsulated SM-OLEDs can operate for >100,000 hours before dropping to 95% of their initial luminescence efficiency (T95). This equates to over 10 years of continuous operation. P-LEDs, however, are shorter lived. Figure 1.3 is a plot of the reported lifetimes of P-LEDs from Sumitomo Chemical, a leader in P-LED technology. The green and red emitters are claimed to last almost 200,000 hours before losing 50% of their initial luminescence 4

21 efficiency (T50 lifetime). This would equate to over 20 years of continuous operation to reach T50. The blue emitters, however, are relatively short-lived due to higher energy excitons being present in this system. Figure 1.3. Reported lifetimes of P-LEDs developed by Sumitomo Chemical. A full explanation as to the differences between SM-OLED and P-LED degradation is beyond the scope of this dissertation. However, the fact that polymers must be processed from solution, which can introduce both impurities as well as residual solvent molecules in the films, appears to be important. In any event, the results for both SM-OLEDs and P-LEDs are exciting and lend perspective on where we can expect the lifetimes of OPV to be in the coming years. 5

22 1.4 State-of-the-art lifetime studies for OPV Small molecule OPV Though the lifetime of polymer solar cells is the focus of this dissertation, it is interesting to take a brief look at the lifetime of state-of-the-art small molecule organic solar cells. Figure 1.4 is a plot of the lifetime versus power conversion efficiency for glass encapsulated tandem small molecular solar cells. The black circles represent the efficiency loss over time for solar cells made with a zinc phthalocyanine(znpc):c60 bulk heterojunction active layer. Figure 1.4 Efficiency over time for encapsulated small molecule solar cells. The devices were illuminated with one-sun intensity using a sulfur plasma lamp. Performing a linear fit and extrapolating the fit out to where 20% of the initial efficiency is lost indicates a lifetime of close to 30~40 years. This lifetime assumes 5.5 hours of one-sun intensity per day for 365 days per year. This is extremely promising, 6

23 though the initial efficiency is only ~4%. It should also be noted that the solar cells in this study were fabricated through thermal evaporation of ZnPc:C60 active layers under high vacuum. Some argue that is an expensive processing technique and one that does not utilize one of the key advantages that organic molecules possess, which is their ability to be processed at ambient pressure and temperature from solution Polymer OPV The lifetime and reliability of polymer OPV has been well studied for bulk heterojunction solar cells (BHJs) based on poly-(para-phenylene-vinylene) (PPV)[8-11] and poly(3-hexylthiophene) (P3HT) polymers[12-15]. Modules based on P3HT/fullerene BHJs have shown reasonable stability when stateof-the-art encapsulation with a glass-on-glass architecture is used[16]. Figure 1.5 shows the decay curves for a number of devices containing different cathode architectures used in that study. Extrapolation of the curves leads to lifetimes of 5,000 hours (the methodology for extrapolating and calculating lifetimes will be discussed later). Assuming negligible degradation in the dark and 5.5 hours of one-sun intensity per day, this translates into an operating lifetime approaching three years. 7

24 Figure 1.5 Efficiency over time for encapsulated polymer (P3HT:PCBM) solar cells. The devices were illuminated with one-sun intensity using a sulfur plasma lamp. More recently P3HT/PCBM devices utilizing an inverted architecture have been shown to retain more than 50% of their initial efficiency after 4,700 hours of continuous exposure to one-sun intensity at elevated temperatures[17] and have exhibited a long shelf life when stored in the dark in ambient conditions [18, 19]. However, results to date have focused on BHJs employing a fullerene combined with PPV derivatives or P3HT as the polymer and have yet to show lifetimes greater than 3 years. Figure 1.6 shows the chemical structure of P3HT together with the structures of more recently developed, higher efficiency polymers. The newer polymers are more complex in structure and tend to have a push-pull architecture where the lowest unoccupied molecular orbital (LUMO) and highest occupied molecular orbital (HOMO) 8

25 are localized on different regions of the polymer. This has led to better performing OPV devices with efficiencies greater than 8%[21]. However, with the exception of the studies presented in this dissertation, there have been no systematic lifetime studies on these higher efficiency polymers. Figure 1.6 Chemical structure for P3HT and structures of newer, more efficient, polymers. 2. Fundamentals of Polymer Solar Cells 2.1 BHJ morphology The polymer-fullerene BHJ is a critical part of the solar cell. In this layer light absorption, charge separation and charge transport occur. These processes will be discussed in more detail in section

26 Generally speaking, the polymer-fullerene films can assume two different morphologies. The first, depicted on the right in Figure 2.1, shows the polymer (purple lines) and fullerene (green circles) intimately mixed in some regions and phase separated in other regions. A key feature of this morphology is that in the pure polymer domains an electronic excitation (exciton) is not in immediate proximity to a fullerene molecule, which implies that exciton must diffuse some distance in order to transfer the electron to a fullerene molecule. If the polymer domain size is larger than the exciton diffusion length the excitons can be lost to recombination, making the domain size a critical aspect of this morphology. Though this morphology is relatively uncommon, the well-studied P3HT:PCBM BHJ assumes this morphology. The second type of film morphology is depicted on the left in Figure 2.1. In this case the fullerene molecules (green circles) intercalate between the side chains on the polymer (vertical purple lines) in very close proximity to the polymer backbone (horizontal purple lines). Excitons formed through light absorption are in very close proximity to a fullerene molecule negating the need for exciton diffusion. This leads to a charge transfer efficiency approaching unity. 10

27 Figure 2.1 Schematic intercalated (left) and unintercalated (right) morphologies of the polymer-fullerene bulk heterojunction active layer. 2.2 Optoelectronic operation of BHJ solar cells Light absorption In order to understand the various degradation pathways in polymer solar cells it is important to first understand the basic operating principles in these devices. Polymer- fullerene solar cells are comprised of semiconducting materials that are capable of absorbing near- infrared to ultraviolet photons due to band- to- band transitions in the material. Upon light absorption, the energy in the photon is transferred to an electron in the material, which is shifted from the ground state to an excited state. Figure 2.2 shows the energy levels of the polymer and fullerene together with the anode and cathode energy levels. In this schematic a blue photon incident on the solar cell is absorbed by the polymer, which transfers an electron from the ground state to the excited state of the polymer. 11

28 Figure 2.2 Solar cell energy levels (increasing energy going up the page). A blue photon is incident on the solar cell and subsequently absorbed by the polymer. An important aspect of organic solar cells is that the excited electron is still coulombically bound to the hole after light absorption, requiring an acceptor molecule (e.g., fullerene) to achieve efficient charge separation. Though this process is shown in the polymer an analogous process occurs in the fullurene Exciton diffusion and charge transfer After light absorption the exciton can diffuse for a period of time (typically on the order of 1 ns) before recombining, which results in a loss of the absorbed energy to heat. It is thus important to have an acceptor molecule within the exciton diffusion radius in order to have efficient charge transfer. The schematic on the left in Figure 2.3 shows the exciton after having diffused to the interface of the polymer- acceptor while the schematic on the right of Figure 2.3 shows the exciton after charge transfer occurs, creating a charge- transfer state between the electron and hole. 12

29 Figure 2.3 (Left) Schematic of the exciton after diffusing to the donor/acceptor interface. (Right) The exciton after charge transfer, creating a charge-transfer state between the electron and hole Charge separation and extraction After charge transfer occurs, the electron and hole are still coulombically bound in a charge- transfer state. Charge separation is achieved through the internal electric field formed due to the mismatch in work functions of the anode and cathode. The electric field drives the electrons and holes apart with the electrons traveling through the fullerene phase and holes traveling through the polymer phase. The electrons are subsequently extracted at the cathode while the holes are extracted at the anode (Figure 2.4). 13

30 Figure 2.4 Schematic of the extracted charges after charge separation and transport drives the electrons to the cathode and holes to the anode. 3. Degradation behavior in OPV 3.1 Photobleaching Photobleaching occurs when the chromophore is destroyed due to a photochemical reaction. This reaction is often caused by a photo-oxidative event where the polymer backbone experiences chain scission. This event breaks the conjugation along the polymer backbone, leading to a loss of light absorption in the polymer. The top schematic in Figure 3.1 shows a schematic of a photon passing through the polymer without being absorbed. In practice the photon would reflect off of the cathode, which is typically aluminum, and re-exit out the front of the solar cell but for simplicity it is shown passing through the solar cell. The bottom photographs show a fresh (left) and aged (right) film of P3HT after having been exposed to one-sun intensity light in air for 3 days. It is clear that the P3HT film has lost most of its light absorption capabilities. 14

31 Figure 3.1 (Top) Schematic of a photon passing through the polymer solar cell. (Bottom) Photographs of a P3HT film on glass before (left) and after (right) exposure to light in air. 3.2 Deep trap formation Deep trap formation can occur through photochemical reactions in organic materials or impurity inclusion either during fabrication or subsequently due to metal ion diffusion from the electrodes. The deep traps can result in a loss of excitons and/or charger carriers through Shockley-Read-Hall recombination. This can have deleterious effects on the performance of the solar cell resulting in a loss of open-circuit voltage (Voc), Fill factor (FF) and short-circuit current (J sc ). Figure 3.2 shows an electron in a deep trap about to recombine with a hole in the HOMO. In this case the deep trap is contained in the donor (polymer) but an analogous process can occur in the acceptor (fullurene). 15

32 Figure 3.2 Schematic of an electron in a deep trap about to recombined with a hole in the HOMO of the polymer. 3.3 Phase separation Phase separation can occur when the temperature of the film increases beyond the glass transition temperature (T g ) of the organic molecules. In this process the polymer and fullerene separate from one another, growing larger polymer-only and fullerene-only domains. If the domain sizes increase beyond the exciton diffusion radius it can lead to a loss of excitons through recombination (Figure 3.3). It is important to note that for many of the successful solar cell materials the glass transition temperature is >100 o C, which is much larger than normal operating temperatures for solar cells in the field. However, this becomes increasingly important if future solar cell materials have lower T g. 16

33 Figure 3.3 Schematic of an exciton about to recombine due to phase separation creating a polymer domain that is larger than the exciton diffusion radius. 3.4 Shallow trap formation Shallow traps can be introduced into the photoactive layer through photochemical events or molecular reorganization, both of which can result in an increase in energetic disorder in the system. Shallow traps can trap charge in the device resulting in spacecharge build-up, which can impact all device parameters. Figure 3.4a shows a schematic of shallow traps in the donor, which allows the electron to exit the device but traps holes as space-charge in the device. Figure 3.4b shows J-V curves of actual solar cells that were annealed to temperatures above the glass transition of the active layer. This led to molecular reorganization, which was shown to introduce shallow traps in the polymer[22]. The J-V curves show a loss of Js and FF with an eventual increase of Voc upon extreme degradation. 17

34 Figure 3.4 (a) Schematic of shallow traps in the polymer resulting in space-charge due to a trapped hole. (b) J-V curves showing the degradation in performance of a solar cell after shallow traps are gradually introduced into the polymer phase. 3.5 Delamination Delamination can occur at various interfaces in an organic solar cell (e.g., anode/active layer) caused by thermomechanical stresses in the device[23]. Delamination can lead to an energetic barrier for charge extraction resulting in a complete loss of efficiency in the regions of delamination. The top schematic in Figure 3.5 shows the delamination of the cathode from the active layer. The bottom images are laser beam induced current (LBIC) maps of a fresh (left) and aged (right) P3HT solar cell. Delamination of the electrode at the anode/active layer led to complete loss of efficiency in certain regions due to edge ingress and dead spot formation[24]. 18

35 Figure 3.5 (a) Schematic of delamination of the cathode from the active layer. (b) Laser beam induced current maps of a fresh (left) and aged (right) P3HT:PCBM solar cell aged under one-sun intensity in air for 5 days. 3.6 General Decay Behavior of OPV Devices The decay pathways previously discussed do not necessarily occur simultaneously. In fact, if devices suffered from all of the degradation mechanisms they would likely be short-lived. However, all optoelectronics do degrade and it is the rate of degradation on which we should focus with the goal of deeper insight into, and ultimately prevention of, the dominant decay mechanism(s). Figure 3.6 shows a typical efficiency decay pattern for polymer/fullerene BHJs employing a standard architecture with an organic hole transporting layer as the anode (e.g., PEDOT:PSS) and a metal cathode (e.g., Ca/Al)[14, 15]. One typically observes a burn-in period characterized by an exponential loss in efficiency whose magnitude and 19

36 duration can vary by polymer/fullerene system, followed by a linear decay period that sometimes ends abruptly when the packaging fails. Device lifetime is measured in the linear decay period once burn-in has ended. Lifetime is defined as the point at which the efficiency from the beginning of the linear decay period has fallen to 80% of this initial value (T80 point). Figure 3.6. Typical decay curve of a polymer solar cell (solid black line) employing a standard architecture with an organic hole transporting layer as the anode and a metal (e.g., Ca/Al) as the cathode. The lifetime (dashed black) is defined by the point at which the efficiency has dropped by 20% from the start of the linear decay period. Both axes are linear. 4. Experimental set-up for lifetime measurements 4.1 Background In order to perform statistically relevant lifetime measurements of OPV devices a robust experimental set-up is required. The set-up must be capable of testing large numbers of solar cells while controlling the key environmental parameters (e.g., 20

37 temperature). The key device parameters, shown in Figure 4.1, that should be monitored during any aging experiment are short-circuit current (J sc ), open-circuit voltage (Voc), fill factor (FF) and maximum power point (Pmax). In addition, if the solar cells are encapsulated the integrity of the encapsulation must be monitored to ensure proper comparison across devices. Finally, in order to ultimately uncover the degradation mechanism(s) advanced characterization must be performed. Figure 4.1 J-V curve of an organic solar cell. The key device characteristics are shown Overview of aging apparatus The essential features of an advanced aging apparatus are shown in Figure 4.2. A light source combined with a reflector is required for obtaining uniform light intensity over all of the solar cells being tested. The solar cells must be mounted on a heated stage in order to ensure uniform temperature across all samples. Temperature sensors with a 21

38 feedback mechanism must be in place to monitor and control the temperature of the solar cells as the ambient temperature in the testing facility will likely change throughout the year. Photodiodes must be used to monitor the fluctuations in the lamp intensity, which can be caused by natural oscillations in the lamp as well as degradation in the lamp over time. Electronics must be used to read and log the temperature and lamp intensity at regular intervals. In addition the electronics must take regular J-V curves to record the J sc, Voc and FF of each device. Finally, the electronics should be capable of holding each solar cell at the desired operating voltage such as the maximum point voltage (V max ). This last point is important since the aging behavior of solar cells is likely different under different voltage biases and a solar cell under normal operating conditions will be held at close to V max. Figure 4.2 Diagram of an aging apparatus. 22

39 4.1.2 Light source and reflectors In choosing the appropriate light source the lamp lifetime, cost, power output, area of uniformity and spectrum must be considered. Figure 4.3 compares various light sources across these categories. Tungsten halogen bulbs can be very cheap and long-lived but are typically low wattage and give very small areas of light uniformity. Metal halide lamps can be long-lived, relatively low cost with reasonable power and large areas of uniformity but have sharp spectral lines and tend to be weak in the infrared. The three light sources highlighted in red tend to be the favored lamps for lifetime experiments. Both xenon arc and metal halide global lamps have good overlap with the solar spectrum in the UV, visible and infrared but tend to be costly. Sulfur plasma lamps, on the other hand, are cheap and have excellent match to the solar spectrum in the visible wavelengths but have very little UV and IR radiation. Figure 4.3 Comparison of light sources. 23

40 The lamp choice can have very significant implications for a lifetime experiment. Xenon arc and metal halide bulbs have sort lifetimes, which necessitates frequent bulb changes for experiments that often last more than 5,000 hours. Xenon arc and MHG lamps also emit large amounts of IR radiation requiring a larger working distance between the lamps and solar cells, for thermal reasons. Xenon arc lamps are capable of emitting ozone, which requires proper ventilation, though non-ozone emitting Xenon arc lamps are available. Sulfur plasma lamps, on the other hand, solve the problems with the xenon arc and MHG lamps by having long lifetimes, little emission in the IR and no ozone emission. However, sulfur plasma bulbs emit very little in the UV wavelengths. The energy contained in UV photons is on the order of the bonding energies for organic materials, which often results in accelerated degradation. Therefore care must be taken when interpreting results of lifetime experiments that employ sulfur plasma bulbs. The results presented in this dissertation were obtained through the use of LG sulfur plasma lamps (PSH0731A) as the illumination source. The bulb temperature of the lamps can shift the wavelength at which the power of the lamp peaks. Figure 4.4 shows a picture of the lamps (left) and the spectra of the lamps at different bulb temperatures (right). The lamps have excellent spectral match to AM 1.5 solar spectrum in the visible with little power in the UV and IR. The lamps were mounted on industrial racks above the sample stage. 24

41 Figure 4.4 (a) Photograph of the 5300K sulfur plasma lamps and (b) the spectra of the lamps at various temperatures. For all experiments presented in this dissertation either a 5300k or 6200k lamp was used. Though the lamps are sold with internal reflectors to give uniformity over a large area we found this not to be the case in reality. Figure 4.5 shows a picture from the manufacturer of the bulb and internal reflector as well as the expected angle over which uniformity is to be expected. Given our working distance between the lamp and substrate, a 14 diameter circle of uniformity was anticipated. 25

42 Figure 4.5 Photograph of the 5300K sulfur plasma lamp bulb and internal reflector (left) and the advertised uniformity of the lamp intensity (left). Intensity maps were made using a filtered photodiode and a Keithley source meter. Figure 4.6 is a 3d plot of the lamp intensity as a function of position. From direct observation the peak shape is completely different from the advertised peak. Figure dimensional plot of the lamp intensity versus position at a working distance of 20 from the lamp. 26

43 In order to create one-sun intensity that was uniform to within ±5% over an 8 x8 area a conical reflector was built using ReflecTech Mirror Film that was laminated onto plastic sheets and shaped using a box frame. The idiosyncratic behavior of each lamp made it necessary to shape each reflector separately. We generally found conical reflectors with a 16 upper diameter and 18 lower diameter and 14 in height to be ideally suited to these lamps Heating Stage and Solar Cell Mount Controlling and monitoring the temperature of the solar cells during extensive testing is important on a number of levels. Temperature can have a large impact on the lifetime of solar cells with higher temperatures leading to faster rates of degradation[12, 25]. In addition, in order for research groups around the world to be able to compare results the exact conditions of the experiment should be stated. Finally, if the temperature is not accurately controlled each solar cell may experience different temperatures leading to a spread in performance, which makes later interpretation more challenging. As a heating stage, we chose aluminum plates (Figure 4.7) with a resistive heating element (Omega CSi32 Series) that was able to control the temperature with an accuracy of ± 0.5 o C. The resistive heating element was placed into a precisely milled hole in the aluminum plate allowing for intimate contact between the heating element and plate. In order to make good thermal contact between the solar cells and Al plate, small Al spacers together with thermal gap pads and an aluminum strapping system were used. Aluminum spacers (1 x 1 x 0.25 ) were machined with a cavity for a one-wire temperature sensor (Maxim DS18B20+PAR). Thermal epoxy (OMEGA Thermcoat) was 27

44 used to secure the sensors into the cavity. The Al spacers were placed on the Al plate using a thermal pad (Digi-Key BER161-ND) to ensure good contact and proper thermal conduction. Solar cells were secured to the top of the Al spacers using thermal pads (Digi-Key Ber222-ND) and aluminum straps that were used to gently press the solar cells into the thermal gap pad for intimate contact. The Al straps were covered with a soft pad to prevent cracking of the glass and to prevent shorting of the alligator clips that were used to connect to the solar cell electrodes. The temperature was read out from the onewire sensors via a custom Python program based on DigiTemp v3.5 by Brian C. Lane, which was last available at The lamp intensity was constantly monitored using an NREL calibrated KG5 filtered silicon photodiodes next to each solar cell. The solar cells were contacted using alligator clips to the leads that extended out from beneath the encapsulation. Figure 4.8 shows an actual photo of the Al plate and mounting system. Figure 4.7 Diagram of the heated plate and solar cell mounting system. 28

45 Figure 4.8 Photograph of the heated plate and solar cell mount Electronics and Connectors Custom electronics (Figure 4.9) were developed for Stanford University by Science Wares Inc. to allow individual control and monitoring of 32 solar cells per tester (Stanford has two testers). Each channel of the tester operates as an independent computer controlled four-wire voltage source with separate force and sense lines. Each voltage source is able to operate in all four quadrants and report the current supplied to, or sourced by, the device under test. Each channel can also monitor light intensity and temperature. The tester operates through a LabVIEW interface that facilitates control of individual channel parameters and graphical monitoring of results while testing is in progress. 29

46 Figure 4.9 Photograph of the 32-channel tester. For the experiments presented in this dissertation, testing began by simultaneously obtaining IV curves for each device in the power quadrant, scanning from Jsc to Voc at approximately 0.25 V/sec. After locating Vmpp for each cell from the first IV curve, the cell was held at its maximum power point (Mpp) using a standard perturb and observe (P&O) method implemented in software. During the first step of the P&O algorithm, the cell was driven at a small voltage step above Vmpp (e.g., 1% increase), and the current at this new voltage was measured. If the resulting cell power had increased from its previous value, the new Vmpp was held and in the next iteration of the algorithm the voltage would be increased again. If the resulting cell power had decreased from its previous value, then in the next iteration of the algorithm the voltage would be decreased. The tracking adjustments were made simultaneously every five seconds on each channel. The voltage, current, power, and photodiode current for each channel were logged in text data files at five-second intervals. The Mpp tracking algorithm was interrupted once every hour to obtain new IV curves simultaneously on all channels. 30

47 Each solar cell connects to the tester via a DB-15 connector (Figure 4.8). As it is important to be able to disconnect and reconnect the solar cells from the tester periodically, electrical boxes fitted with DB-15 connectors were securely attached to the steel shelf around the solar cell heating stage. This allowed the solar cells to remain connected to the electrical box while the DB-15 plugs from the tester were disconnected from the back side of the electrical boxes. The tester outputs large amounts of data during long term tests. For one study that will be presented in later sections each device had 4,400 data files, which amounted to over 70,000 data files. We have developed custom code at Stanford using Matlab that has automated the data analysis process Laser beam induced current Laser beam induced current (LBIC) mapping is a powerful tool to monitor changes in local current patterns. In particular, failure of the encapsulation or delamination with the solar cells can be easily monitored. In this system (Figure 4.10), solar cells are mounted to a 2-axis motorized translation stage (Standa 8MT173 20). Light from a continuous wave laser source (Spectra- Physics Stabilite 2017 argon ion laser) is focused down to a beam diameter of approximately 10 µm through a long working distance infinity corrected objective lens (Mitutoyo M Plan APO 20 /0.42). The laser beam is optically chopped (Stanford Research Systems SR540 Optical Chopper) while the sample stage moves in a two-dimensional pattern. The anode and cathode of the solar cell are connected to a transimpedance amplifier (Oriel), where the signal is converted to a voltage and subsequently sent to a lock-in amplifier (Stanford Research Systems SR830 DSP Lock-In Amplifier) for detection. The entire system is 31

48 controlled via a LabVIEW interface. Figure 4.10 Photograph of LBIC apparatus. 5. State-of-the-Art Lifetime Study 5.1 Background and methodology In the following sections I present a detailed operating lifetime study of encapsulated solar cells comprised of poly[n-9'-hepta-decanyl-2,7-carbazole-alt-5,5(4',7'-di-2-thienyl-2',1',3'-benzothiadiazole) (PCDTBT) in BHJ composites with the fullerene derivative [6,6]-phenyl C70-butyric acid methyl ester (PC70BM). PCDTBT/PC70BM solar cells achieved an efficiency greater than 6% in 2009[26], making this one of a small number of polymers able to achieve this level of performance. In this study the experimental equipment previously described is used to age a large number of solar cells simultaneously, holding each device at its maximum power point while controlling and monitoring the temperature and light intensity. Using this set-up we were able to compare the PCDTBT/PC70BM system with the well-studied P3HT/PCBM system and demonstrate a lifetime for PCDTBT devices that approaches 7 years, which is the longest reported operating lifetime for a polymer based solar cell. 32

49 Since testing and environmental conditions as well as sample preparation can vary greatly between laboratories, it is important for any lifetime study to use a sufficiently large sample size and to compare any new system against a well studied system under identical aging conditions. In the current experiment, eight PCDTBT/PC 70 BM and eight P3HT/PCBM solar cells were prepared with average initial device efficiencies of 5.5 ± 0.15% and 4 ± 0.05%, respectively. PCDTBT devices (St-Jean Photochemicals) were fabricated on glass substrates with the following structure: indium tin oxide (ITO)/poly(3,4- ethylenedioxythiophene)(pedot:pss)/pcdtbt:pc 70 BM/Ca/Al. The ITO (8 Ohm/sq from Thin Film Devices) coated glass substrate was ultrasonically cleaned in detergent, acetone and isopropyl alcohol, and subsequently dried overnight in an oven. The substrates were placed in a UV ozone chamber for 20 minutes prior to the deposition of PEDOT:PSS (AI 4083, HC Starck)via spin-casting from aqueous solution to form a 25 nm thick film. The substrate was annealed for 10 min at 140 o C in air and then transferred into a glove box to deposit the active layer and counter electrode. A solution containing a mixture of PCDTBT:PC 70 BM (1:4) (from M. Leclerc and Nano-C, respectively) in 1,2- ortho dichlorobenzene solvent with a concentration of 7 mg/ml was heated to 60 o C and subsequently spin-cast on top of the PEDOT/PSS layer to achieve an active layer thickness of ~80 nm. The film was slow dried overnight in a covered Petri dish in the glove box. Finally a Ca/Al (7 nm/100 nm) was deposited by thermal evaporation in a vacuum of about 1x10-6 mbar. To fabricate the P3HT/PCBM devices, P3HT (Plextronics) and PCBM (Nano-C) in a 1.5:1 solution were dissolved in 1,2-ortho dichlorobenzene. Substrates were 33

50 prepared similarly to the PCDTBT devices. However, PV1000 HTL was used in place of PEDOT:PSS as the hole transport layer since this provides better device stability for the P3HT system. PV1000 HTL is spun in air to a thickness of 60 nm and is then baked on a hotplate for 15 min at 170 o C in a glovebox. The P3HT:PCBM active layer is then spin cast in the glovebox to a thickness of 200 nm and baked on a hotplate in the glovebox at 175 o C for 30 min. Finally, a Ca (20 nm)/al (200 nm) cathode is deposited in a thermal evaporation chamber at a base pressure of 1x10-6 mbar. It is well known that UV radiation can induce defects and even chain scission in conjugated polymers[27]. To remove the harmful effects of UV, which will likely be filtered when OPV is commercialized, an LG sulfur plasma lamp, with little UV power but strong spectral match to the AM 1.5 G solar spectrum in the visible wavelengths was used. It remains to be determined whether devices held at open-circuit voltage (Voc) or short circuit current (J sc ) exhibit differing rates of degradation. However, solar cells in the field will be operating close to or at their maximum power point (M pp ). For this experiment each cell was held at its M pp, which was dynamically adjusted every 5 seconds using a standard perturb and observe method implemented in the software. The solar cells were fabricated, characterized and stored in the dark for one week before being placed under the lamp. All devices experienced modest decay in J sc while being stored in the dark. The solar cells were then aged at M pp continuously for 4,400 hours in air under one-sun intensity. Reflectors were constructed to provide uniform light intensity (± 4%) over all devices. The intensity was calibrated using an NREL certified KG5 filtered silicon photodiode. The temperature was held at 37 C using a resistive 34

51 heating element. The temperature varied by less than 0.5 C from sample to sample. Both the light intensity and temperature were monitored every 5 seconds for each device. Current-voltage curves were taken every hour for the duration of the experiment. 5.2 Efficiency decay over time and lifetime calculation Figure 5.1 shows the efficiency decay over time for PCDTBT and P3HT devices. The curves are normalized to their initial values at the start of the aging process and the data points represent the average of every 100 hours worth of data for the 8 devices for each polymer type. The error bars for each point represent the maximum and minimum values for the devices at each of the data points. Figure Plot of the efficiency versus time for the average of the eight PCDTBT and eight P3HT devices. From Figure 5.1 we see that PCDTBT devices experience a more rapid decay during burn-in than do P3HT devices. However, in the linear decay region PCDTBT 35

52 devices have a less severe slope than the P3HT devices, which has significant implications for the lifetime of each device type. A more detailed analysis of the initial severe loss of efficiency for PCDTBT devices will be presented in later sections. Figure 5.2 is a plot of the efficiency versus time for the average of the eight PCDTBT and eight P3HT devices. Every 100th data point has been plotted with the maximum and minimum values for the 8 devices at those points, similar to what was presented in Figure 5.1. However, I have ignored the data points surrounding the points shown (i.e., no averaging around the data points has been performed as was performed in the plot in figure 5.1) in order to highlight the PCDTBT decay behavior between 1,400 and 2,300 hours. We see a leveling out and even a very slight rise in the average efficiency, which was caused by variations in J sc among samples. The demarcation for the end of the burn- in period was chosen to be 1,300 hours. Linear regression was performed on the 3,100 data points between 1,300 and 4,400 hours for each set of devices (dashed lines). The linear fits are extrapolated and terminate at the lifetimes for each of the device types. The plot shows the average lifetime for the PCDTBT devices to be over 6 years, which is twice that of the P3HT devices, assuming 5.5 hours of one- sun intensity per day, 365 days per year. 36

53 Figure Efficiency decay for PCDTBT (red) and P3HT (blue) solar cells. Data is plotted every 100 hours and averaged over 8 devices of each type. The linear fits use hourly data from 1300 to 4400 hours and lifetimes (T80) are extrapolated out to where the efficiency drops to 80% of the efficiency value at 1300 hours. The lifetimes are calculated by assuming 5.5 hours of one-sun intensity per day, 365 days per year. An important point is that there is no set guideline for choosing the time at which the burn- in period ends. Therefore it is important to analyze the lifetimes assuming various starting points for the linear decay period. Varying the end of the burn- in period between 1,000 and 2,000 hours had almost no impact on the expected lifetime of the PCDTBT devices but made the lifetime of the P3HT devices range between 2.5 and 3.8 years. In any case the lifetime of the PCDTBT system was shown to be substantially longer than that of the P3HT system. Finally, this lifetime is meant as a rough estimate. Correlated studies that relate laboratory lifetimes with 37

54 outdoor conditions, which naturally involve thermal cycling, aging in the dark as well as variations in light exposure and humidity, are important. 5.3 Ordered lifetimes Figure 5.3 shows the ordered lifetime of each device used in this study. The P3HT devices have a very narrow spread in lifetimes demonstrating the reproducibility of P3HT devices. Every PCDTBT device had a longer lifetime than the P3HT devices, though the PCDTBT devices had a larger spread in expected lifetimes due to variations in J sc over time (discussed in a later section). One PCDTBT device displayed remarkably stable characteristics over the entire experiment leading to an expected lifetime approaching 11 years. It is important to point out that a reasonable amount of effort by the scientific community has been directed toward understanding and improving the stability and lifetime of the P3HT system[6] while very little effort has been directed toward optimizing the stability of the PCDTBT system. We expect optimization to narrow the spread in lifetimes seen in this experiment and increase the average value. 38

55 Figure Ordered lifetimes of 16 devices. PCDTBT (red, 9-16) devices show substantially longer lifetimes but with a wider spread due to variation in J sc over time while P3HT (blue, 1-8) solar cells show a narrower spread. Linear fits were used for each device using 3100 data points from 1300 to 4400 hours and the lifetime was defined by the point where the initial efficiency at the 1300 hour point drops by 20%. 5.4 Decay of J sc, FF and V oc Figure 5.4 shows the decay of the key devices characteristics (J sc, FF, Voc) over the entire experiment. The Voc and FF of the PCDTBT devices remained remarkably stable for more than 4,000 hours after a burn-in period of ~400 hours. The decay was then dominated by a slow decline in J sc over the duration of the experiment. The abrupt drop in the PCDTBT upper errorbar for the FF at 2,100 hours is due to one device experiencing a drop in FF over the course of several days. P3HT devices show similar behavior for J sc and FF but experience a slow decline in Voc throughout the experiment. Similarly made state-of-the-art encapsulated P3HT devices that were made into modules showed a drop in efficiency of 9.82% per 1,000 hours of continuous exposure, which is 39

56 higher but comparable to what we have seen in this experiment[16]. The decline in both J sc and Voc of the P3HT devices results in a faster loss of efficiency when compared to PCDTBT devices. Figure Device characteristics for PCDTBT (red) and P3HT (blue) solar cells over 4,400 hours of continuous testing. The curves are each normalized by the initial value at the start of the aging process. Each point represents the average of 100 hours of data for 8 solar cells of each type. 40

57 5.5 Decay of key device characteristics for each solar cell It is insightful to examine the individual performance of the 16 solar cells used in the experiment. Figures 5.5 and 5.6 display the behavior of the J sc, Voc, FF and efficiency for the P3HT and PCDTBT solar cells used in this study, respectively. The data points represent the average value of 100 hours of data and the plots are normalized to the initial value for each parameter at the start of the aging process. Figure J sc (asterisk), V oc (open diamond), FF (open square) and efficiency (open circle) for eight P3HT/PCBM devices used in this study. The values of each have been normalized to the initial values at the start of the aging process. 41

58 Figure Jsc (asterisk), Voc (open diamond), FF (open square) and efficiency (open circle) for eight PCDTBT/PC70BM devices used in this study. The values of each have been normalized to the initial values at the start of the aging process. It is worth making a few observations concerning the behavior of the individual solar cells. The first is the uniformity of the degradation behavior among the P3HT samples. This was previously mentioned but visually it is clear from Figure 5.5 that the devices are remarkably similar. This speaks to reproducibility of P3HT:PCBM solar cells, which is important for future commercialization. The second observation is the variability of the PCDTBT:PC 70 BM devices (Figure 5.6). In particular the current (J sc ) rises for a period of time in certain devices before continuing its decay. This directly impacts the efficiency of the devices and has significant implications for using a linear fit to the data points in the linear decay region. In certain cases the rise in J sc occurs at the start of the linear decay period and creates a small rise in efficiency that shortens the 42

59 estimated lifetime for the device by creating a more severe slope for the linear fit. This is the main reason for the large variation in device lifetimes in PCDTBT solar cells used in this study. 5.6 Laser beam induced current maps Laser beam induced current maps (LBIC) provide valuable insight into spatial distribution of J sc and its subsequent degradation behavior by determining whether the loss is occurring uniformly over the sample or locally as in the case of pinhole formation[28-32]. Figure 5.7 shows LBIC images for P3HT and PCDTBT devices before and after aging. Figures 5.7a-b show a uniform loss of current for a P3HT device. No loss of device area is seen, though the start of pinhole formation can be identified. Figures 5.7c-d show a fresh and aged PCDTBT device. Similar to the P3HT device there is a uniform loss of current with no loss of device area. However, fewer pinholes are seen. Finally, Figures 5.7e-f show the rapid formation of pinholes and a loss of device area for a PCDTBT device within 200 hours of aging as the result of encapsulation failure. The encapsulation failure was observable under an optical microscope. This device was not used in this study but is included here to show the importance of encapsulation on device lifetime. 43

60 Figure LBIC images of solar cells before and after aging under one-sun light intensity. Each image is a square area of 3.6 mm by 3.6 mm. (a) and (b) are images of a P3HT device before and after aging continuously for 4400 hours. The images show no loss of device area and the formation of smaller spots of low current. (c) and (d) are images of a PCDTBT solar cell before and after aging. No loss of device area is seen. (e) and (f) show the effects of encapsulation failure on a PCDTBT solar cell made identically to that in (c). 5.7 Summary of lifetime experiment Upon completion of this experiment we were able to show that PCDTBT/PC 70 BM solar cells have a lifetime approaching seven years, which is the longest reported lifetime for a polymer-based solar cell and approximately twice that of the well-studied P3HT/PCBM system. The Voc and FF of the PCDTBT system experience a rapid initial decay but then show remarkable stability for over 4,000 hours of continuous testing. Given the recent development of this polymer, little optimization of the architecture has been performed to promote the stability and enhance the lifetime of PCDTBT based solar 44

61 cells. Through a deeper understanding of the decay mechanisms and further optimization it seems reasonable to be able to reduce the burn-in loss and extend the lifetime, which would make OPV commercially viable. 6. Analysis of Burn-in Loss 6.1 Background of burn-in loss Figure 6.1 presents the efficiency decay of the devices used in the previously described experiment. A key finding was that the PCDTBT-based solar cells experienced a loss of efficiency (~25%) during burn-in, which was significantly higher than the 10% loss experienced by P3HT-based solar cells. The burn-in loss for PCDTBT solar cells, however, is followed by a remarkably stable performance that continues throughout the duration of the 4,400 hour experiment. A deeper understand of this loss mechanism is important and can lead to more efficient devices in the future. Figure Efficiency decay for PCDTBT (red) and P3HT (blue) solar cells over 4400 hours of continuous testing with the burn-in period shown in darkened region. The curves are each normalized by the initial value at the start of the aging process. Each point represents the average of 100 hours of data for 8 solar cells of each type. The error bars show the highest and lowest values at each point. 45

62 In the subsequent sections I present a systematic study of the burn-in degradation mechanism behind PCDTBT:PC 70 BM solar cells. I show that a photochemical reaction in the photoactive layer creates states in the bandgap of PCDTBT. These sub-bandgap states increase the energetic disorder in the system, which reduces the FF, V oc and to a lesser extent J sc. The photochemical reactions are shown to progress rapidly when first exposed to light but subsequently decrease in occurrence, which results in the stabilization of the V oc and FF. 6.2 Approach to analyzing the decay mechanism in solar cells In general, degradation can occur at the electrodes or within the photoactive layer of OPV devices (Figure 6.2). Electrodes in organic electronic devices have been shown to degrade via oxidation[19, 33, 34], delamination[24, 34, 35], de-doping[36] and interfacial organometallic chemistry[37-41]. Photoactive materials can suffer from photochemical reactions[42-46], thermochemical reactions[47, 48], morphological changes[49, 50] and impurity inclusion such as metal ion diffusion from the electrodes[41, 51-54]. In the following sections we describe a systematic study of the aging of PCDTBT:PC 70 BM devices in which we deduce the probable pathway for device degradation. The studies were all performed in a nitrogen environment, under one-sun intensity with illumination provided by LG PSH K sulfur plasma lamps (to limit the amount of UV radiation) and temperatures in the o C range, unless otherwise stated. These conditions are similar to those used in the original lifetime study of PCDTBT:PC 70 BM devices[55]. 46

63 Figure An extensive, but not exhaustive, list of degradation pathways that are present in polymer solar cells. 6.3 Variation of electrodes In order to determine the effect of the electrodes and the electrode/active layer interface on the degradation mechanism of PCDTBT:PC 70 BM solar cells, devices were fabricated using five different cathode/ anode combinations. PCDTBT devices (St-Jean Photochemicals, Mw = 112 kda) with an active area of 0.1 cm 2 were fabricated on indium tin oxide coated glass substrates (8 Ohm/sq from Thin Film Devices). The substrate was ultrasonically cleaned in detergent, acetone and isopropyl alcohol, and subsequently dried overnight in an oven at 110 o C. The substrates were placed in a UV ozone chamber for 20 minutes prior to the deposition of the anode. Each substrate contained 4 active devices. Films of poly(3,4-ethylenedioxythiophene) poly(styrenesulfonate) (PEDOT,CLEVIOS P VP AI 4083, work function ~5.0 ev) were 47

64 deposited via spin-casting (4000 RPM for 45 sec) from aqueous solution to a thickness of 25 nm. The substrate was annealed for 10 min at 140 o C in air and then transferred into a nitrogen-filled glove box (oxygen and moisture content < 3 ppm) to deposit the active layer and counter electrode. Vanadium oxide (V 2 O 5, Sigma-Aldrich 99.99% purity) and molybdenum oxide (MoO 3, Alfa Aesar 99.95% purity) were thermally evaporated with thickness of ~7nm. For the active layer, a solution containing a mixture of PCDTBT:PC 70 BM (1:4, by weight) (from St. Jean Photochemicals and Nano-C, respectively) in 1,2-ortho dichlorobenzene with a concentration of 25 mg/ml. The solution was stirred overnight at 90 o C then cooled to 60 o C before depositing the active layer by spin coating at a spin speed of 1500rpm for 45 s to achieve an active layer thickness of ~80 nm. The film was slowly dried overnight in a covered Petri dish in the glove box. Finally the cathode a Ca/Al (7 nm/100 nm) was deposited by thermal evaporation in a vacuum of about 1x10-6 mbar. In the case where 2,2,2 -(1,3,5- benzenetriyl)tris-[1-phenyl-1h-benzimidazole] (TPBI) was used in place of Ca, a 10 nm layer of TPBI was thermal deposited at 1x10-6 mbar prior to the deposition of the Al cathode. Solar cells were encapsulated in a glass-on-glass architecture and aged under the sulfur plasma lamps as previously described[55]. Aging studies were performed and the device characteristics monitored over time. Additionally, LBIC mapping was used to confirm that there was no failure of the encapsulation barrier. The average efficiency decay for 16 solar cells of each electrode type over the first 70 hours of operation is shown in Figure 6.3. This time period is where the majority of burn-in occurs in this system. In all cases ITO was used as the transparent conducting electrode. With the exception of the devices without calcium in the electrodes, the entire 48

65 set of devices had approximately the same initial efficiency of ~5.5% (V oc ~0.87V, FF~0.63, J sc ~10mA/cm 2 ); those made without calcium had an initial power conversion efficiency ~4.5% (V oc ~0.83V, FF~0.61, J sc ~9mA/cm 2 ). The samples that employ V 2 O 5 and MoO 5 had fewer sampling points but were aged under the same conditions as the other devices. All electrode configurations led to a similar loss of efficiency. Figure Efficiency loss over time for PCDTBT:PC 70 BM devices using various electrodes. Figure 6.4 shows the device characteristics during burn-in as shown in the previous figure. In every case burn-in for PCDTBT:PC 70 BM solar cells entails a large loss of fill factor, a moderate loss of open-circuit voltage and a smaller loss of shortcircuit current. This trend was reproduced hundreds of times using various electrodes and aging conditions (e.g., different light sources). The magnitude of the loss of each can vary for each polymer batch but this general trend is reproducible. In Figure 6.4 it is interesting to note that substituting V 2 O 5 or MoO 3 in place of PEDOT:PSS leads to a 49

66 slightly larger decrease in the fill factor but smaller loss of V oc relative to the cells that employ PEDOT:PSS. However, given the variation in the magnitude of the loss of FF, V oc and J sc in each polymer batch we are hesitant to claim this as a trend. Efficiency (norm) Efficiency (norm) Fill Factor (norm) Aging time (hours) Aging time (hours) CaAl Al TPBI_Al V205 MoO3 ITO/PEDOT:PSS/BHJ/Ca/Al ITO/PEDOT:PSS/BHJ/Al ITO/PEDOT:PSS/BHJ/TPBI/Al ITO/V 2 O 5 /BHJ/Ca/Al ITO/MoO 3 /BHJ/Ca/Al Jsc (norm) Voc (norm) Aging Aging time time (hours) (hours) Aging time (hours) Figure 6.4 Device characteristics over time for PCDTBT:PC 70 BM devices using various electrodes. In summary, the efficiency loss in all cases was dominated by a loss of FF and V oc and a smaller loss of J sc with slight differences in the magnitude of the decay depending on the electrode used. Though it is possible that the cause of degradation is in the ITO or Al layers, as these were common to all devices, we will presently show that the dominant degradation process occurs in the photoactive layer. 50

67 6.3.1 Morphology: Grazing Incidence X- ray Diffraction (GIXD) Another factor that can significantly impact the performance of organic solar cells is the morphology of the photoactive layer[50]. Crystallite size and orientation[56, 57], π-π stacking coherence length[22, 58-60] as well as interaction between the electron donating material and fullerene[61, 62] are important for charge separation and transport and can change with time depending on the conditions under which the films are aged. PCDTBT has been shown to have local order in the π-π stacking direction[22, 63]. Figure 6.5a shows a schematic of the experimental set-up. Incident x-rays, which impinge on and subsequently diffract off the film, are collect by an image plate. Figure 6.5b shows a typical diffraction image we obtained from a PCDTBT film on silicon. The highlighted region of intensity suggests that PCDTBT is a highly amorphous polymer with short-range order in the π-π stacking direction, which is in line with what has been reported in the literature[22, 63]. 51

68 Figure 6.5 (a) Experimental set-up for grazing incidence x-ray diffraction. (b) 2-d image for a PCDTBT film on silicon showing the π-π stacking diffraction peak in the out-of-plane direction. The cake segment is shown around the peak. Our group recently reported that when PCDTBT is heated above its glass transition temperature, the X-ray coherence length along the π-π stacking direction decreases[22]. This shortening of the coherence length along a critical direction for charge transport is correlated with the creation of hole traps in PCDTBT and a reduction in the solar cell efficiency. In the present study, X-ray diffraction experiments were conducted to determine the effect of aging on film morphology. Films of PCDTBT were aged at 40 o C in nitrogen in the light for up to 200 hours and GIXD patterns were obtained at beamline 11-3 at the Stanford Synchroton Radiation Lightsource. To prepare the samples, films of PCDTBT 52

69 were spin coated onto silicon substrates using the same method as for solar cells previously described. The solution contained 7mg/mL of PCDTBT in 1,2-ortho dichlorobenzene. X-ray diffraction was performed at the Stanford Synchrotron Radiation Lightsource (SSRL) on beam line 11-3 (2-D scattering with an area detector, MAR345 image plate, at grazing incidence) with an incident energy of 12.7 kev. For grazing incidence experiments, the incidence angle was slightly larger than the critical angle, ensuring that we sampled the full film depth. Scattering data are expressed as a function of the scattering vector q=(4*pi/λ)sinθ where θ is half the scattering angle and λ is the wavelength of the incident radiation. Here qxy (qz) is the component of the scattering vector parallel (perpendicular) to the substrate. Analysis of the 2D-GIXD patterns was performed with the software WxDiff provided by S.C.B. Mannsfeld[64]. Figure 6.6 is a plot of the integrated cake segment at different aging times. There is very little change in the peak width, which implies that morphological changes do not play a dominant role in the degradation process. As a comparison, Figure 6.6b is a plot of the integrated cake segments of PCDTBT films that have been annealed for 10 minutes at temperatures above the glass transition temperature of PCDTBT. Our group previously showed that annealing the films at temperatures above the glass transition temperature for PCDTBT resulted in a broadening of the integrated peak by as much as 24% at Full- Width Half- Maximum (FWHM). This broadening was later correlated to a decrease in the local order in π- π stacking direction, which led to an increase in the number of shallow traps above the HOMO of PCDTBT. The increase in traps then led to a degradation in the performance of PCDTBT solar cells. 53

70 Figure 6.6 (a) Integrated cake segment over different aging periods under one-sun intensity at 40 o C. The peaks have been normalized for comparison of peak width. (b) Integrated cake segment of PCDTBT films that have been annealed at different temperatures. The peaks have been normalized for comparison of peak width. In our study, we observed a broadening in the peak width of <3%, which lends further support to our claim that morphology plays at most a minor role in degradation of the solar cells. With this said, it is still possible that nano-scale phase separation between the polymer and fullerene is occurring. However, we expect this change to be minimal because the polymer glass transition temperature[65] is much higher than the operating temperatures used in this study. 54

71 6.3.2 Morphology: Sub- bandgap external quantum efficiency (EQE) EQE measurements provide the number of electrons extracted from the device per photon incident on the device. Intuitively photons with energies below the bandgap of the polymer and fullerene should have an EQE approaching zero since they should not be absorbed through a band-to-band transition. However, due to the existence of a chargetransfer state, as shown in Figure 6.7a, photons with energies below the bandgaps of the polymer and fullerene can be directly absorbed by an electron transition from the HOMO of the polymer to the LUMO of the fullerene. This charge-transfer state absorption can then lead to current generation and thus a non-zero EQE. Furthermore, it has been shown that the charge-transfer state energy is correlated to the open-circuit voltage of bulk heterojunction solar cells and ultimately provides a limit to the achievable open-circuit voltage[66]. A lower charge-transfer state energy necessarily leads to a lower realizable V oc. By probing the charge-transfer state energy as a function of aging time we can then gain insight into whether or not a shift in this energy is ultimately causing the V oc of the solar cell to decrease as it ages. Figure 6.7b shows a plot of the EQE curves for a fresh and aged PCDTBT:PC 70 BM solar cell. The curves were fit with a model derived by Vandewal et al.[67] and the charge-transfer state energy was extracted. The extracted charge-transfer energy for the fresh device matched well with that reported in the literature for PCDTBT:PC 70 BM solar cells and was shown to change very little as the solar cell was aged. Together with the previously described GIXD data, this provides further evidence that morphological changes play only a minor role in the degradation process that is seen in PCDTBT solar cells. 55

72 Figure 6.7 (a) Charge-transfer state of electron on the LUMO of the acceptor and hole on the HOMO of the donor. (b) External quantum efficiency of a fresh and aged PCDTBT solar cell with the theoretical fit. 6.4 Light or temperature induced chemical reactions Chemical reactions in conjugated polymers have been well-studied because of their importance in the more mature technology of organic light-emitting diodes. In some cases LED materials degrade when excitons are present[42, 68-70] and in other cases they degrade when molecules are in the charged state[71, 72]. Organic semiconductors could also potentially degrade at elevated temperatures even when excitons or polarons are not present. In order to probe chemical reactions in the initial degradation of PCDTBT:PC 70 BM solar cells, the V oc under 1 sun illumination was periodically 56

73 measured after aging under three different experimental conditions: 1) aging at a constant temperature of 50 o C in the dark with no applied bias, 2) aging at a constant temperature of 50 o C under one-sun illumination (using a white-light LED) and operating at opencircuit, and 3) forward bias in the dark at a constant current of 10mA/cm 2, which is a typical current experienced for PCDTBT:PC 70 BM solar cells under normal operating conditions. For testing, the devices were loaded into a cryostat from inside a glovebox and subsequently held under high vacuum to remove external influences on device performance. For the device aged at 50 o C under illumination at open-circuit conditions, we observe an immediate decay in the V oc, which continues throughout the 40 hour experiment (Figure 6.8). A loss of FF and V oc is observed in I-V curves taken before and after aging (Figure 6.9), similar to the behavior observed during burn-in. The device aged in the dark at 50 o C shows no change in open-circuit voltage and the I-V curves taken after aging were unchanged. The device with 10 ma/cm 2 of current in the dark showed no change in V oc or in the I-V curves during aging. The lack of degradation while operating at a constant current in the dark suggests the degradation is mediated by excitons and not polarons. We hypothesize that the loss of V oc and FF during burn-in are the results of photochemical reactions in PCDTBT that lead to degradation of the photoactive layer and adversely affect its charge transport properties. 57

74 0.84 Open-circuit voltage (V) Dark at 50 o C 1-sun intensity at at o C C Dark with 10 ma/cm 2 of current Aging time (hours) Figure sun intensity, open-circuit voltage (Voc) of devices aged for 40 hours under different conditions. The diamonds (blue) are the Voc of a device held at 50 oc in the dark. The triangles (red) are the Voc of a device held at 50 oc under illumination. The squares (black) are the Voc of a device held in the dark with 10mA/cm2 of current in forward bias. The dashed line is only a guide to the eye and does not represent data points. Figure 6.9 shows the current-voltage curves of a fresh and aged device represented by the red triangles in Figure 6.8. The cell was aged under 1-sun light intensity using a white light LED (thorlabs MWLED, at 50 o C for 40 hours. The I-V curves were taken at room temperature before and after the aging process. The loss in V oc, FF and J sc is apparent and is representative of burn-in in the PCDTBT:PC 70 BM system.. 58

75 I (ma/cm 20 2 ) 15 Fresh Aged V Figure 6.9 I-V curve of a fresh and aged PCDTBT solar cell. 6.5 Photochemical reactions in the polymer External quantum efficiency measurements were performed on fresh and aged PCDTBT:PC 70 BM solar cells. Figure 6.10 shows the EQE curves in addition to the absorption curves for both the PCDTBT and PC 70 BM. The EQE drops more dramatically in the wavelengths between nm, which is where the PCDTBT absorbs more strongly than the PC 70 BM. This suggests that either photobleaching is occurring in the PCDTBT molecule or excitons that are formed on PCDTBT molecules are being lost to recombination. More tests are required to determine which of these effects is occurring but it appears that it affects PCDTBT more than it affects PC 70 BM and would point to a photochemical reaction in PCDTBT. Importantly, this result was seen for all solar cells made by Plextronics so has statistical relevance and thus cannot be ignored. 59

76 Figure 6.10 EQE curves of a fresh and aged PCDTBT solar cell (left axis). The normalized absorption profiles for PCDTBT and PC 70 BM are shown at the bottom (right axis) Background on charge transport models Probing the charge transport properties of PCDTBT:PC 70 BM devices can allow for a deeper understanding of the effects of photochemical reactions on the device performance. In order to probe the hole transport properties through the polymer, a holeonly diode shown in Figure 6.11 was used. The device employs contacts that create a barrier to injection for electrons due to the large energy offset between the LUMO of the polymer and the work function of the contacts used on either side of the device. This allows holes to be efficiently injected and extracted, while preventing electrons from entering the device. The contacts are also solution deposited to avoid depositing metals 60

77 (e.g., palladium) that can penetrate the polymer layer and affect charge transport through defect states or subsequent photochemical reactions. Figure 6.11 Hole-only diode employing PCDTBT with high work function contacts on both sides of the device to enable hole-only injection. Space-charge limited current models that use an analogue of Child s Law[73] (Equation 1) have been widely used to study the charge transport properties of conjugated polymers. A key assumption in applying Child s Law to the current-voltage curves is that all of the injected holes are able to flow through the device. However, our group has shown that current-voltage measurements made using hole-only diodes comprised of PCDTBT:PC70BM are not well described by Child s Law[22]. Equation 1: 61

78 Instead, the J-V curves follow a modified charge transport model that assumes an exponential distribution of trap states that extend into the bandgap. Figure 6.12a shows the energy levels for the hole-only diode, similar to what was shown in Figure 6.11, but with the addition of electronic states and trapped holes in the bandgap of the polymer. Figure 6.12b shows a plot of the density of electronic states starting at the HOMO of the polymer and exponentially decreasing with increasing energy (toward vacuum) into the bandgap of the polymer. The model assumes this exponential distribution of states in the bandgap of the polymer, with a characteristic width of E t, and further assumes that any charges (holes in this case) that lie above the Fermi level are trapped and thus immobile. Figure 6.12 (a) Hole-only diode employing PCDTBT with high work function contacts on both sides of the device to enable hole-only injection. The red crosses represent electronic states. (b) Density of states beginning at the HOMO and exponentially decreasing into the bandgap of the polymer with a characteristic width, E t. The Fermi energy is show by the dashed line. 62

79 Equation 2 shows the analytical expression for the current-voltage relationship that can be derived under the new assumption of an exponential distribution of states extending into the bandgap of the polymer. The important parameters in this expression are the current (J), voltage (V), thickness (L) and E t /kt (m). This last parameter (m) is the width of the exponential divided by the thermal energy of a carrier at the temperature of interest (kt). A larger value for m implies a greater number of potential trap states in the bandgap of the polymer and thus a larger potential for space-charge to build up in the device. Equation 2: Important to note is the fact that the current (J) now depends on the voltage (V) and thickness (L) of the active layer with a very different dependency than seen in Child s Law. If we assume that the width of the exponential, Et, is equal to kt then we arrive back at Child s Law, which implies that all of the injected charge is free to move throughout the device, which makes sense. By examining a log-log plot of the J-V curve in forward bias the characteristic width, Et, can be directly extracted from the slope in forward bias (typically between 3 V~10 V). This model has been used successfully to describe hole transport in organic crystals[74], as well as electron[75-80] and hole[81-83] transport in certain conjugated polymers as a function of voltage, temperature and film thickness. By fitting the dark I-V curves of hole-only diodes in forward bias to this trap-mediated transport model, one can 63

80 directly extract the characteristic energetic breadth of the exponential trap profile extending from the edge of the HOMO into the bandgap of the polymer. Using this model, thermal processing was shown to increase the energetic breadth of the trap distribution in PCDTBT (increasing the number of deep trap states) as a result of morphological changes [22]. Figure 6.13(a) shows a log-log plot of the I-V curves for a PCDTBT hole-only diode at various temperatures. The device was loaded into a cryostat in a nitrogen filled glovebox and the cryostat was subsequently pumped down to 10-6 Torr. J-V curves were acquired using a Keithley sourcemeter. The circles represent the actual data while the dashed lines represent the theoretical fit using equation 2. Figure 6.13b shows a log-log plot of the I-V curves for 4 PCDTBT hole-only diodes with different active layer thicknesses. The experimental data overlaps the theoretical fits extremely well over a wide range of temperatures, thicknesses and voltages, giving us confidence that the model is applicable to PCDTBT devices. Figure 6.13 (a) Log-log plot (circles) and fits (dashed lines) of the I-V curves for a PCDTBT hole-only diode at various temperatures. (b) Log-log plot (circles) and fits (dashed lines) of the I-V curves for PCDTBT hole-only diodes at various active layer thicknesses. 64

81 6.5.2 PCDTBT:PC70BM hole- only diode preparation and aging To examine trap state formation in PCDTBT:PC 70 BM due to photochemical reactions, the previously described model was used to analyze the current-voltage curves of aged PCDTBT:PC 70 BM hole-only diodes. A set of 16 identical devices were fabricated as described below and then aged under one-sun intensity at 40 o C for over 300 hours. Glass substrates coated with patterned tin-doped indium oxide (8 Ohm/sq from Thin Film Devices) were cleaned and treated in a similar manner to solar cells. An aqueous solution of PEDOT:PSS was spun on top, then baked on a hotplate for 15 minutes at 140 o C to drive off any remaining solvent, resulting in a 25 nm film. The substrates were then transferred to a dry nitrogen glove box for the active layer deposition. The devices were fabricated on six separate substrates with up to five active devices per substrate to ensure at least 16 active devices throughout the entirety of the experiment. Active layer solutions were prepared in the glove box by dissolving a 1:4 weight ratio of PCDTBT: PC 70 BM in 1,2-ortho dichlorobenzene in a concentration of 35 mg/ml. The solutions were stirred overnight at 90 o C then cooled to 60 o C before depositing the active layer by spin coating at a spin speed of 700 rpm for 45 s. Films were allowed to dry slowly in covered Petri dishes at room temperature. A top contact of CA (Plextronics, diluted by 50% in ethanol, work function of ~5.5 ev) was spin-cast in air in the dark, and the remaining solvent was driven off by heating on a hotplate at 65 o C for 15 minutes. Following deposition of the CA-1914, the hole-only devices were held under vacuum in an antechamber for 30 minutes prior to being returned to the glove box 65

82 where 200 nm of Al was deposited by thermal evaporation. The area for all hole-only devices was 0.1 cm 2. To ensure that exposure to oxygen in the dark during the deposition of CA-1914 did not introduce new degradation phenomena, solar cells were fabricated and exposed to the same conditions as the hole-only diodes prior to cathode deposition. Aging of these solar cells along with solar cells that were not exposed to oxygen prior to cathode deposition showed identical burn-in behavior. Hole-only diodes were loaded into an airtight aluminum chamber with a glass window inside a nitrogen filled glovebox and subsequently aged under sulfur plasma lamps with one-sun light intensity. Current-voltage characteristics for hole-only devices were measured in an evacuated, liquid nitrogen cooled Janus ST-100 cryostat using a Keithley 2400 source meter. Devices were loaded into the cryostat in the glove box prior to measurement to avoid exposure to air. The voltages shown in plots of hole-only current measurements in the subsequent section were corrected for the potential lost due to series resistance, V = V applied I*R series. They were not corrected for any built-in potential due to a difference in electrode work functions, because in the measurement range (3 V 15 V) this correction is small, and the actual built in voltage is not necessarily the difference in nominal electrode work functions PCDTBT:PC70BM hole- only diode testing and analysis Throughout the aging process, the diodes (shown in the inset of Figure 6.14) were periodically loaded into a cryostat and I-V measurements were made in the dark at temperatures between K in 20 K increments. Using the curve fits we determined the breadth of the exponential trap distribution (E t ) for each diode at each aging time 66

83 (Figure 6.14). All diodes exhibited very similar results. Fresh devices exhibited an energetic trap width of ~40 mev, which is consistent with the trap width of unannealed PCDTBT: PCBM blends[22]. Over the course of aging for 350 hours, the energetic breadth of the trap distribution increased to ~130meV. Glass Glass ITO PEDOT (5.0eV) PCDTBT:PC 70 BM Plexcor PEDOT HTL (5.5ev) Al LogJ (A/cm 2 ) logj (ma/cm 2 ) Fresh µn HOMO = 3.99e+015 cm -1 /V-s E t = 43.4meV C:\MATLAB701\work\SCLC_ \T K-dark E t = 43.4 mev N t = 3.11e+018 cm K 240 nm logv (V) C:\MATLAB701\work\SCLC_ \T K-dark µn HOMO = 3.40e+013 cm -1 /V-s E t = 123meV E t = 123 mev N t = 1.09e+017 cm -3 LogV (V) 240 hours 300K 280K 220K 240 nm 240K 260K 240 nm 260K nm 280K 240K nm 220K LogJ (A/cm 2 ) logj (ma/cm 2 ) hours µn HOMO = 1.70e+013 cm -1 /V-s E t = 95.6meV C:\MATLAB701\work\SCLC_ \T K-dark E t = 95.6 mev N t = 1.13e+017 cm K 240 nm 220K 240 nm 240K 240 nm 260K 240 nm 280K 240 nm logv (V) LogV (V) 350 hours µn HOMO = 2.26e+012 cm -1 /V-s E t = 131meV C:\MATLAB701\work\SCLC_ \T K-dark E t = 131 mev N t = 7.53e+016 cm -3 LogJ (A/cm 2 ) logj (ma/cm 2 ) K 240 nm 240K 240 nm 260K 240 nm 280K 240 nm 300K 240 nm logv (V) LogV (V) LogJ (A/cm 2 ) logj (ma/cm 2 ) K 240 nm K 240 nm 240K 240 nm K 240 nm 280K 240 nm logv (V) LogV (V) Figure I-V curves in forward bias (solid) and curve fits (dashed) for a holeonly diode as a function of aging time. Figure 6.15 is a plot of the loss of efficiency during the burn-in period from our previous report on the long term aging of PCDTBT:PC 70 BM solar cells[55] versus the increase in Et from the curve fits previously described. The hole trap distribution widened significantly in the first 120 hours of operation (from 40meV to 100meV), which correlates to a loss of device efficiency of almost 18%. Over the subsequent 230 hours the increase in the width of the hole trap distribution began to saturate, which 67

84 corresponds well with the leveling out of the device efficiency. These results support our hypothesis of a degradation mechanism in which photochemical reactions cause an increase in the average depth of hole traps in the PCDTBT:PC 70 BM layer and result in a loss of device efficiency. Figure The diamonds (left axis) represent the characteristic width of the exponential distribution of traps (E t ) in PCDTBT:PC 70 BM hole-only diodes extending into the bandgap from the HOMO of the polymer. The hole-only diode was aged at 40 o C in a nitrogen filled box under one-sun intensity. The triangles (right axis) represent the efficiency loss of PCDTBT:PC 70 BM devices aged in a similar manner to the hole-only diodes Background: Photothermal deflection spectroscopy (PDS) To further support our observation of the increase in states in the bandgap in aged PCDTBT:PC 70 BM films, we measured the change in sub-bandgap absorption in films of PCDTBT:PC 70 BM. Sub-bandgap absorption features typically have low absorption 68

85 coefficients[84] and their observation requires the use of sensitive measurement techniques such as photothermal deflection spectroscopy (PDS), which has been previously employed to measure absorption features in the sub-bandgap region in polymer-polymer[85] and polymer-fullerene blend films[84, 86]. PDS relies on the complete or fractional conversion of absorbed electromagnetic radiation by the material of interest into heat via nonradiative de-excitation processes. This conversion process causes a temperature rise in the material itself and its surroundings. The sample to be studied is submerged in an inert fluid (Fluorinert TM ) with a very large change in refractive index as a function of temperature, so that a small amount of heat leads to a localized change in the index of refraction in the liquid. Such change in refraction index is measured and correlated with the absorption coefficient of the material of interest. During a measurement, a modulated monochromatic pump beam (arranged perpendicular to the plane of the substrate) is absorbed by the sample. A second (transverse) probe laser beam is deflected by the localized change in the refractive index of the surrounding deflection medium, and a position sensitive detector records the periodic deflection by a lock-in technique. The measured deflection is proportional to the absorption coefficient of the measured thin film Film fabrication and aging In the present study, films of PCDTBT:PC 70 BM were fabricated by drop casting from the same solution as that used for solar cell fabrication on quartz substrates, without any electrodes, and allowed to dry for 5 days in a glovebox in the dark. Films were then held under low vacuum for up to 2 hours to further remove any remaining solvent. Films 69

86 were then immersed in a cuvette filled with Fluorinert. The pump beam used in PDS was obtained from a halogen lamp, then monochromated and focused onto the sample. The probe beam is a commercial He:Ne laser, and the deflection signal is measured with a position sensitive detector. Absorption measurements were then made as previously described by Goris et al[84]. For a typical time constant of 10 s in the detection lock-in amplifier 30 measurements are made for each energy; the plots show the average value with the error bars representing the standard deviation. In the present study films were aged under one-sun intensity in a similar manner to the hole-only diodes previously discussed. To ensure the deflection medium (Fluorinert) did not affect the sample being tested, multiple PDS spectra were taken on the same sample before aging and shown to be identical. In addition, films that were stored in the dark in nitrogen showed no change in absorption over the course of the experiment PDS results Figure 16a shows the experimental set-up used for taking PDS absorption spectra. Spectra were periodically taken on the same region of the sample and compared as shown in Figure 16b. The absorption features between 0.8 and 1.1 ev have previously been attributed to the vibrational overtones of the C-H bond stretch, while absorption at energies above 1.2 ev is dominated by the charge-transfer state[84]. A comparison of the fresh and aged curves shows an increase in absorption in the sub-bandgap region over time. The largest increase in absorption is observed within the first 120 hours of aging. The increase in sub-bandgap absorption slows over the subsequent 240 hours. The 70

87 timescale of sub-bandgap absorption increase is very similar to that of trap formation seen in hole-only diodes and efficiency decay in solar cells. The degradation seen in these films also reinforces the notion that degradation in this system is in the active layer and is independent of the choice of electrodes, since these samples were spun on bare quartz and had no electrodes. Figure 6.16 (a) Experimental set-up for PDS. (b) Absorption spectra taken a various aging times. Samples were aged under one-sun intensity in a nitrogen filled chamber Summary of burn- in study As a brief summary, PCDTBT:PC 70 BM solar cells experience a burn-in over the first few hundred hours of operation where up to 25% of the initial efficiency is lost, predominantly through a decrease in FF and V oc and to a lesser extent the J sc. The origin 71

88 of the loss is a photo-induced reaction in the active layer that leads to the formation of sub-bandgap states. These states increase the energetic disorder in the active layer and may reduce device performance by a number of mechanisms, including trap-mediated recombination (i.e. Shockley-Reed-Hall recombination), reduced hole mobility and the build-up of space-charge in traps, which diminishes the electric field available to drive charge carrier separation. An important aspect of the aging process during burn-in is that it initiates rapidly but then slows down and appears to stop, suggesting that the reaction species have been depleted. While it is too early to be conclusive, it is likely that the polymer, impurities within the polymer or the chain ends photo-oxidize due to trace amounts of oxygen within the films. Evidence for the formation of hole traps upon photooxidation has been seen in solar cells comprised of poly(2-methoxy-5-[3,7 - dimethyloctyloxy]-p-phenylene vinylene) (MDMO-PPV) with PCBM[46]. However, chemical reactions occurring in the fullerene are also possible with recent work out of NREL shedding some light on the role of fullerenes in the degradation process[87]. Further investigations should help to isolate the specific reaction pathway and assist in mitigating the burn-in loss in the future, improving the commercial viability of polymerbased photovoltaic devices. 7. Identifying the photochemical reaction pathway 7.1 Background Reaction pathways are complex due to the number of reaction pathways that exist and the detection limit of analytical existing techniques, which often are limited to the parts per thousands. Figure 7.1 shows the chemical structure of PCDTBT and possible 72

89 reactions that may occur. Based on discussions with the polymer chemist Dr. Mario Leclerc, who first synthesized this polymer, end group cleavage, as shown in Figure 7.1a, was identified as a likely reaction pathway. In addition, the literature has suggested that crosslinking, as shown in Figure 7.1b, is also possible via an oxygen bridge. The nitrogen-carbon bond connecting the side chain to the carbazole group has also been shown to be a weak point in this monomer[45]. Impurities, as shown in Figure 7.1c, are also possible and can lead to radical formation and subsequent photochemistry. These are only some of the many possible reactions that can occur and are meant to highlight the complexity of the issue. Figure 7.1 Possible reaction pathways for PCDTBT. 73

90 7.2 Infrared Fourier Transform Spectroscopy (FTIR) FTIR is a powerful tool used by synthetic chemists to identify bond formation in organic molecules. The detection limit of FTIR is generally one part per thousand, making detection of smaller reactions challenging. However, in that we currently do not know the extent to which photochemical reactions are occurring during burn-in it is useful to perform studies using FTIR. 7.3 Sample preparation and aging Films of pure PCDTBT (70nm) were deposited on silicon substrates. One set of samples was aged in air under one-sun intensity to understand the photo-oxidative pathways under extreme circumstances. Another set of samples was aged in a nitrogen filled box, identical to the one used to aged the hole-only diodes, under one-sun intensity. FTIR spectra were taken periodically. 7.4 Results of photo-oxidation study using FTIR Figure 7.2 shows the absorption spectra of a fresh (a) and photo-oxidized (b) PCDTBT film together with the difference spectrum (c) of the aged minus the fresh spectra. The fresh film shows strong absorption in the alkyl side chain region (~3000 cm - 1 ) and aromatic ring region (~1500 cm -1 ). The region below 1350 cm -1 is more complex, and unnecessary for the present discussion, but involves the vibrational coupling of various bonds on the polymer chain. The photo-oxidized spectrum shows a loss of absorption by the alkyl side chains and aromatic rings and new features in the hydroxyl group region (~3200 cm -1 ) and carbonyl region (~1700 cm -1 ). As the side chains photo- 74

91 oxidize it is expected that the hydroxyl groups will form and as the aromatic rings photooxidize it is common for the carbonyl groups to form. The difference spectrum makes it easier to see the loss of absorption peaks (dips in the difference spectrum) and gain of new absorption features (peaks in the difference spectrum). Figure 7.2 (a) FTIR spectrum of a fresh film of PCDTBT on silicon. (b) FTIR spectrum of the same film that has been photo-oxidized. (c) The spectrum of the aged minus the fresh. 7.5 FTIR results of PCDTBT films aged in nitrogen under one-sun intensity Figure 7.3 shows the absorption spectra of a fresh and aged PCDTBT film on silicon. The film was aged in a nitrogen filled chamber for 10 days. The overlay highlights many of the challenges presented by attempting to see a small signal with a large background. Reproducibility is very important and only after a large number of samples have been analyzed can any conclusions be drawn. One of the key challenges is 75

92 in taking spectra on the same film on different days. The lamp intensity of the FTIR spectrometer changes with time as does the beam alignment. This accounts for many of the differences seen in the spectra in Figure 7.3. At present we do not feel comfortable drawing any conclusions from the multiple attempts at using FTIR to see small changes in absorption due to photochemical reactions. Figure 7.3 FTIR spectrum of a fresh (purple) and aged (pink) film of PCDTBT on silicon after 10 days of exposure to one-sun intensity in a nitrogen filled chamber. 7.6 Impurity detection Impurities in polymer solar cells can come from the solvent used to spin films, the synthesis of the polymer itself or subsequent diffusion of metal ions into the films from the electrodes. During the synthesis of PCDTBT the metals Pd, Sn and Br are used. The polymer ends are terminated with Br and often left in this state. It is well known that Br- C bonds are weak, which is why this is used as a catalyst in the synthesis of polymers. In 76

93 the case of PCDTBT the brominated end groups are replaced by phenyl derivatives with the intention of removing the metal from the film. However, it is likely that this reaction is incomplete. To confirm this we used inductively couple plasma to measure the metal concentrations and found that the PCDTBT used in the previous studies have 1ppm Pd, 255ppm Br, and <1 ppm Sn. The Pd and Sn concentrations are well below the required limit for good device lifetimes while the Br content is quite high. Through discussions with manufacturers of organic light emitting diodes we found that metal traces above 10 ppm were found to affect the lifetime of the diodes. In the case of PCDTBT we found 25x this limit for BR, which is likely due to incomplete exchange of the phenyl groups. Further investigations are underway to determine whether or not the Br impurities are cause of the photochemical reactions that we have detected. 8. Conclusion 8.1 Summary of work Figure 8.1 provides a compact summary of the key results of this dissertation. The lifetime study resulted in lifetimes for the high efficiency polymer, PCDTBT, approaching 7 years. However, an efficiency tax of almost 25% was paid during the burnin period. Investigations of the loss mechanism ruled out the electrodes and pointed to a photochemical reaction in the active layer. A deeper analysis of the photochemical reaction provided evidence of a reaction in the polymer. The extent of trap formation due to the photochemical reactions was observed through the use of a novel charge transport model and sensitive sub-bandgap absorption measurements. Though it is preliminary, it would appear that remaining impurities in the film (i.e., Br) are a likely cause of the 77

94 photochemical reactions and further investigations are underway to confirm this and ultimately improve the reliability and lifetime of PCDTBT solar cells. Figure 8.1 Summary of key findings in the understanding of the lifetime and degradation mechanisms in PCDTBT:PC 70 BM solar cells. 8.2 Future direction If OPV is to become a viable technology and participate in solving the energy problems that face the world today the issues of lifetime and reliability must be solved. The more recent high efficiency polymers tend to have higher degrees of functionality and are thus more complex structures. Heteroatoms are common and lower degrees of crystallinity are the rule rather than the exception. This can lead to very different degradation patterns than those seen in the well-studied P3HT system. This will require detailed studies, as were presented in this dissertation, on numerous polymer systems. 78