Comparative Study of APFO-3 Solar Cells Using Monoand Bisadduct Fullerenes as Acceptor

Transcription

1 Institutionen för fysik, kemi och biologi Examenarbete Comparative Study of APFO-3 Solar Cells Using Monoand Bisadduct Fullerenes as Acceptor Yu-Te Hsu LITH-IFM-A-EX--10/2320 SE Linköpings universitet Institutionen för fysik, kemi och biologi Linköping

2 Avdelning, institution Division, Department Chemistry Department of Physics, Chemistry and Biology Linköping University Datum Date Språk Language Svenska/Swedish Engelska/English Rapporttyp Report category Licentiatavhandling Examensarbete C-uppsats D-uppsats Övrig rapport ISBN ISRN: LITH-IFM-A-EX--10/2320--SE Serietitel och serienummer ISSN Title of series, numbering URL för elektronisk version Titel Title Comparative Study of APFO-3 Solar Cells Using Mono- and Bisadduct Fullerenes as Acceptor Författare Author Yu-Te Hsu Sammanfattning Abstract The urgent need for new, sustainable energy source intrigues scientists to provide the solution by developing new technology. Polymer solar cell appears to be the most promising candidate for its low cost, flexibility, and massive producibility. Novel polymers have been constantly synthesized and investigated, while the use of PCBM as acceptor seems to be the universal choice. Here, we studied the use of four different fullerene derivatives - [60]PCBM, [70]PCBM, and their bisadduct analogues - as acceptor in APFO-3 solar cells. A series of investigations were performed to study how the processing parameters - blend ratio, spin speed, and choice of solvent - influence the device performance. Using bisadduct fullerenes results in an enhanced Voc, as predicted by the up-shift of energy levels, but a strongly reduced Jsc, hence a poor PCE. Photoluminescence study indicates that all APFO-3:fullerene devices are limited by the inefficient dissociation of fullerene excitations, while it becomes more influential when bisadduct fullerenes were used as acceptor. The best device in this study was fabricated using [70]PCBM as acceptor and chlorobenzene as solvent, which exhibits a PCE of 2.9%, for the strong absorption, fine morphology, and comparatively strong driving force. Nyckelord Keyword Polymer solar cells, bisadduct fullerenes, photoluminescence quenching.

3 Contents Abstract 3 Motivation 4 1 Introduction: Polymer Solar Cells Development History Working Principle Material Properties Light Harvest Photovoltaic Parameters External Quantum Efficiency Short Circuit Current Density Open Circuit Voltage Fill Factor Power Conversion Efficiency Device Architectures Single Layer Bilayer Bulk Heterojunction Other Architectures Loss Mechanisms Optical Losses Exciton Losses Recombination Losses Collection Losses Optimization Strategies Electronic Structures Morphology Thickness

4 CONTENTS 2 2 Fabrication and Characterization Material System Donor Acceptor Device Fabrication Device Characterization Results and Discussions Specific Parameters Absorption Current Density Saturation Photocurrent Density Extraction Ratio Comparison of Experimental and Simulation Data Influence of Blend Ratio Influence of Spin Speed Influence of Solvent Origins of J sc Drop in bispcbm Devices Performance Summary of Best Devices Conclusions 33 Acknowledgement 34

5 Abstract The urgent need for new, sustainable energy source intrigues scientists to provide the solution by developing new technology. Polymer solar cell appears to be the most promising candidate for its low cost, flexibility, and massive producibility. Novel polymers have been constantly synthesized and investigated, while the use of PCBM as acceptor seems to be the universal choice. Here, we studied the use of four different fullerene derivatives - [60]PCBM, [70]PCBM, and their bisadduct analogues - as acceptor in APFO-3 solar cells. A series of investigations were performed to study how the processing parameters - blend ratio, spin speed, and choice of solvent - influence the device performance. Using bisadduct fullerenes results in an enhanced V oc, as predicted by the up-shift of energy levels, but a strongly reduced J sc, hence a poor PCE. Photoluminescence study indicates that all APFO-3:fullerene devices are limited by the inefficient dissociation of fullerene excitations, while it becomes more influential when bisadduct fullerenes were used as acceptor. The best device in this study was fabricated by using [70]PCBM as acceptor and chlorobenzene as solvent, exhibits a PCE of 2.9%, for the strong absorption, fine morphology, and comparatively strong driving force. 3

6 Motivation Recently, the search for new sustainable energy sources has become a global race. Solar energy has attracted great interests of scientists and become one of the most active research field today. Semiconductor solar cells are efficient, but expensive. Polymer solar cells, on the other hands, are cheap, which makes it appealing for household applications. Other properties, such as flexibility, light-weight, and massive-producibility also make this technology appealing. However, polymer solar cells are limited by the low power conversion efficiency (PCE) and short lifetime, due to the intrinsic properties of organic semiconductors. Approaches to extend the lifetime of polymer solar cells have been proposed, and it is believed that the lifetime issue will not be the obstacle for production. The main challenge is to improve the PCE over 10%, which is the criterion for polymer solar cells to become a viable technology. PCE of polymer solar cells has been greatly improved from 2.5% in 2001 to 7.9% in 2009, which gives strong confidence to scientists for commercialization. Various conjugated polymers, such as P3HT and MEH-PPV, have been extensively studied as electron donor in polymer solar cells, while the use of [60]PCBM or [70]PCBM as electron acceptor seems to be the universal choice. Although PCBM works well as acceptor, somehow it limits the potential of using other materials with different properties. Previously, Lenes et al. demonstrated improvement in PCE from 3.8% to 4.5% by using bis[60]pcbm, the bisadduct analogue of [60]PCBM, as acceptor in P3HT cell. This improvement is contributed by the significant increase in V oc of 0.15 V, for the higher LUMO level of bis[60]pcbm. This result demostrates a new pathway to improve the performance of polymer solar cells. In this work, four different fullerene derivatives - [60]PCBM, [70]PCBM, and their bisadduct analogues - were used as acceptor in APFO-3 solar cells. The influences of process parameters, such as blend ratio, spin speed, and choice of solvent, to device performances were investigated. This report is arranged in the following way: Chapter 1 introduces the development history, working principle, performance parameters, device architectures, losses mechanisms, and optimization strategies of polymer solar cells. Chapter 2 introduces the material system, device fabrication, and characterization techniques used in this work. Chapter 3 presents the results and discussions of device performances using different fullerenes and process parameters. Chapter 4 summarized the conclusions of this work. 4

7 Chapter 1 Introduction: Polymer Solar Cells 1.1 Development History With the discovery and development of conductive polymers [1], the field of organic electronics covering organic light emitting diode (OLED) [2], organic field effect transistor (OFET) [3], and organic photovoltaics (OPV) [4, 5] has been established. Conductive polymer was first applied in photovoltaic device, also known as solar cell, in 1980s, while the PCE was well below 0.1% [6]. A major breakthrough was made by Tang et al. in 1986 [7]. By bringing the electron donor of organic molecule with an acceptor in a bilayer architecture, he achieved PCE of 1%. The interface between the donor and acceptor layer is called heterojunction. In 1995, Yu et al. made cell of PCE up to 2.9% under monochromatic illumination (430 nm) by using blend of donor and acceptor to create enormous heterojunctions throughout the absorbing layer [4]. The idea of mixing donor and acceptor in solution, known as the bulk heterojuncion (BHJ), is dominating in this field today. In 2001, Shaheen et al. achieved PCE of 2.5% under white light illumination using the BHJ concept [8]. Since then, the field of polymer photovoltaics has attracted great attention of scientists Annual publications Efficiency records Solarmer UCLA 8 7 Annual publications Univ. Linz Univ. Linz Univ. Linz UCSB UCSB Efficiency record (%) 0 Univ. Linz Year Figure 1.1: Annual publications of OPV field and efficiency records with year from Scopus database. Corresponding research institutes are given in the proximity to the record-efficiencies. 5

8 1.2. WORKING PRINCIPLE 6 In Figure 1.1, the annual publications of OPV field and efficiency records are shown. It can be seen that with the increasing amount of researches, the efficiency has increased steadily. In 2009, PCE record boosted from less than 6% to almost 8%. The recent progress gives scientists strong confidence to convert OPV into a viable technology. Today, polymer solar cell is regarded as the most promising candidate for the next-generation energy source, for its properties of low-cost, light-weight, and manufacturability for mass production. 1.2 Working Principle Material Properties In analogy of using traditional semiconductor, typically Si, in inorganic solar cells, organic semiconductors are used in polymer solar cells to harvest sunlight. Organic semiconductors are conductive organic molecules or polymers with π-conjugation system. Chemical and electronic structures of some studied organic semiconductors are shown in Figure 1.2. O N S N S S N N S N S n O n n S S n LUMO O O O O HOMO P3HT MDMO-PPV PCDTBT APFO-3 [60]PCBM [70]PCBM Figure 1.2: Chemical and electronic structures of some studied organic semiconductors. Values shown above and below the bars indicate the LUMO and HOMO levels in ev [9, 10, 11]. The lengths of bars represent the width of bandgaps. Colors filled in bars denote the type of materials, blue for donors and red for acceptors. The type of organic semiconductor, p-type or n-type, is determined by the carrier transport property. Hole-conducting materials that work as electron donor, usually conjugated polymers, are considered as p-type, whereas electron-conducting materials that work as electron acceptor, usually fullerene derivatives, are considered as n-type. In analogy of the electronic structure of inorganic semiconductors, the highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) in organic semiconductors correspond to the valence band (E v ) and conduction band (E c ). The gap between HOMO and LUMO is denoted as the bandgap.

9 1.2. WORKING PRINCIPLE 7 The bandgap energy (E g ) of commonly used conjugated polymers in polymer solar cells is around 2.0 ev, higher than of Si (1.1 ev). This larger bandgap energy results in a greater optical losses in sunlight harvesting. For a better harvest of the solar spectrum, conjugated polymers with narrower bandgap have been synthesized and studied [12, 13]. Conjugated polymers usually have very high absorption coefficient (10 5 cm 1 ) [14] compared to Si (10 3 cm 1 ) [15] in the visible range, meaning that light can be absorbed very efficiently in polymer solar cells. Typically, thin film of thickness around 300 nm will be adequate to absorb most incident light [16]. However, the thicknesses of optimized polymer solar cells are usually around 100 nm [17], limited by the inefficient carrier transport. This limitation is due to the low carrier mobility of conjugated polymers, usually in the order of 10 4 cm 2 V 1 s 6, which is several order of magnitudes lower than Si (1400 cm 2 V 1 s 6 ). Researchers have put great efforts to obtain efficient collection of carriers via morphological control [18, 19, 20] Light Harvest Light harvest in polymer solar cells is a multiple-step process, as illustrated in Figure 1.3: (1) Upon illumination, a fraction of light with energy larger than E g is absorbed by the active layer, the organic layer where light is absorbed. Electrons are excited to the LUMO and leaving holes in the HOMO. In this illustration, excitation occurred in the donor. Due to the stronger Coulomb-interacting nature in conjugated polymers, electrostatic-bounded holeelectron pairs (dash-surrounded), or excitons, are generated at room temperature. (2) Excitons are dissociated by internal electric field, generating free electrons and holes. Internal electric field can be achieved by using electrodes of different work functions, or by creating interfaces of donor and acceptor with different electron affinities (D/A heterojunction). (3) Free electrons and holes are collected at cathode and anode respectively, by transporting through the pure acceptor and donor pathway, and generating current. For light to be able to enter the active layer, one of the electrodes must be transparent, usually indium tin oxide (ITO). The other electrode is highly reflective metal, usually Al. a (1) (2) (3) (3) b Φ ITO (1) χ D Eg (2) (2) χa LUMO (3) Φ Al E 0 (3) HOMO ITO D HJ A Al Figure 1.3: a, Microscopic and b, energetic illustration of light harvest process in polymer solar cell. E 0, χ, and φ denotes the vacuum level, electron affinity, and work function. (1) Light with energy exceeding E g can be absorbed and excitation occur upon illumination. (2) Photogenerated excitons diffuse to the donor/acceptor heterojunction (HJ) within their diffusion length and dissociate into free carriers. (3) Free electrons and holes transport to the electrodes via the pure acceptor and donor pathway.

10 1.3. PHOTOVOLTAIC PARAMETERS 8 The selective transport of free carriers is determined by the work functions of electrodes, hence the work function of respective electrode has to match the energy level of carriers. ITO matches HOMO levels of most conjugated polymers, makes it ideal to work as anode. On the other hand, Al usually works as cathode. 1.3 Photovoltaic Parameters In a photovoltaic device, carriers can only be collected at their corresponding electrodes, meaning that only one direction of current flow is allowed. Therefore, a photovoltaic device behaves like a diode and has J-V characteristics as Figure 1.4. Photovoltaic parameters of devices are extracted from the J-V characteristics. J Illumination Dark 0 V max V oc V J max J sc Figure 1.4: J-V characteristics of photovoltaic device. Solid and dash line represents device behavior under illumination and in the dark condition. J sc, V oc, J max and V max denotes short circuit current density, open circuit voltage, current density and voltage value corresponding to the maximum power point. The maximum power output is illustrated by the slashed area External Quantum Efficiency External quantum efficiency (EQE) describes how efficient the device is to convert incident photon into free carriers, defined as EQE(%) = Number of collected electron N umber of incident photon EQE can also be expressed by EQE = η abs η diff η diss η tr η cc where η abs, η diff, η sep, η tr, and η cc represents the efficiency of photon absorption, exciton generation and diffusion, exciton dissociation, carrier transport, and carrier collection [21]. The product of last four parameters is the internal quantum efficiency (IQE), the efficiency of converting absorbed photon into carriers.

11 1.3. PHOTOVOLTAIC PARAMETERS Short Circuit Current Density Short circuit current density (J sc ) is the unbiased current density under illumination (see Figure 1.4), can be formulated as J sc = e λmax λ min EQE(λ)φ sun (λ) dλ where e, λ max and λ min, EQE(λ) and φ sun (λ) represents the elementary charge, absorption wavelength range, EQE and photon flux of sunlight spectrum in respect of wavelength. The carrier mobility does not show explicitly in this expression, but implicitly in EQE. If the mobility is not sufficient, carriers might recombine before reaching the electrodes and cannot be collected, hence a reduced EQE. Notice that carrier mobility in polymer solar cells is not a material parameter but a device parameter, it is sensitive to the morphology of active layer [22]. The BHJ architecture increases dissociation efficiency but sacrifices mobility due to the formation of numerous complicated heterojunctions, which is difficult to control and optimize Open Circuit Voltage Open circuit voltage (V oc ) is the voltage corresponding to zero current density under illumination. Empirically, V oc in bulk-heterojunction polymer solar cells can be estimated by [23] V oc = (1/e)( Edonor HOMO Eacceptor ) LUMO 0.3V Hence the electronic structure of components has a direct influence on V oc. The occupation of Edonor HOMO and Eacceptor LUMO is denoted as the charge-transfer state. Very recently, Vandewal et al. proposed a theoretical approach of V oc based the correlation of charge-transfer absorption and emission, using the detailed balance and quasi-equilibrium theory [24], formulated as V oc = kt e ln(j sc + 1), J 0 e J 0 = EQE P V (E)φ BB (E) EQE EL where J 0, EQE EL and EQE P V, and φ BB are dark saturation current density, the electroluminescent and photovoltaic EQE, and photon flux of blackbody spectrum. The reduction of J 0, by reducing the electronic coupling between the donor and acceptor or eliminating non-radiative recombination of carriers (increasing EQE EL ), will lead to an increase in V oc Fill Factor Fill factor (FF) defines the maximum output of device under illumination, formulated as: F F = J maxv max J sc V oc where J max and V max are the corresponding current density and voltage at the maximum power point. FF is determined by the number of carriers reaching the electrodes. To optimize FF, the drift distance of carriers, which is proportional to the product of carrier lifetime τ and mobility µ, has to be maximized [25]. FF is also affected by the serial and parallel resistance of devices. Lower serial resistance and higher parallel resistance will result in a higher FF. In practice, V oc and FF can be enhanced by coating additional hole/electron transport layers in contact with anode/cathode to improve the carrier injection efficiency. PEDOT:PSS and LiF are often used for this purpose [26, 27].

12 1.4. DEVICE ARCHITECTURES Power Conversion Efficiency Power conversion efficiency (PCE), which is the most important parameter of photovoltaic device, describes the ability of device to convert light into electricity, defined as P CE = J scv oc F F P in where P in is the power of incident light. To maximize PCE, the product of J sc, V oc, and FF has to be maximized. 1.4 Device Architectures Single Layer Figure 1.5 illustrates different architectures of polymer solar cells. Figure 1.5 a shows the simplest case: a polymer active layer is sandwiched by electrodes with different work functions. The difference in work functions creates an electric field to dissociate excitons; however, the field built by the asymmetrical work functions of electrodes is not strong enough to effectively dissociate excitons, results in very low EQE and PCE around 0.01% [6] Bilayer To deal with the dissociation issue in single layer device, the concept of heterojunction is introduced. In bilayer architecture (Figure 1.5 b), two layers of pure donor and acceptor stacked in sequence are sandwiched by asymmetrical electrodes. With this architecture, a donor/acceptor interface, or heterojunction, is created. The difference in electron affinity between donor and acceptor creates an electric field at the interface. Excitons, which are generated nearby, can diffuse to the interface and be dissociated. PCE can be significantly enhanced to 1% with this architecture [7]. The main limit of bilayer devices is the discrepancy between the exciton diffusion length ( 10 nm) [28] and the optimal device thickness ( 100 nm). With this discrepancy, excitons generated further away from the interface than their diffusion length cannot be dissociated. Hence EQE and PCE are still limited for bilayer devices. a b c d Dead ends Single Layer Bilayer Ideal Bulk Heterojunction Realistic Bulk Heterojunction Figure 1.5: Schematic diagram of different architectures of polymer solar cells. For all architectures, active layers are sandwiched by aluminum cathode (top) and ITO anode (bottom). Blue and red color represents donor and acceptor, respectively.

13 1.5. LOSS MECHANISMS Bulk Heterojunction To overcome the exciton diffusion issue, Yu et al. tailored numerous heterojunctions throughout the active layer, the bulk heterojunction (BHJ), by using blend of donor and acceptor in solution to fabricate active layer. With BHJ architecture, excitons can diffuse to their nearest interfaces and dissociate, and EQE can be significantly improved. While the dissociation efficiency is improved, the transport efficiency of carriers is sacrificed. Figure 1.5 c depicts the ideal structure of BHJ device: donor and acceptor form bi-continuous percolating pathway, and each domain is in the dimension of exciton diffusion length. Therefore, every photogenerated exciton can be dissociated and carriers can transport to the electrodes via the percolating pathway. In reality, however, it is very difficult to create the ideal structure. Domains that are not in contact with electrodes, the dead-ends, and larger dimension than exciton diffusion length are usually created (Fig 1.5 d). Therefore, the exciton dissociation and carrier transport are still limited in BHJ device. Besides, for the increase of interfaces, the probability of carrier recombination increases, further limits the carrier transport Other Architectures Novel architectures have been made to enhance the performance of polymer solar cells. One example is the hybrid planar-bulk heterojunction architecture. In this architecture, the active layer is composed of a BHJ layer inserted in a bilayer. The high carrier mobility of bilayer and the high exciton-dissociation efficiency of bulk heterojunction can be realized simultaneously with this architecture. A remarkable improvement in PCE from 3.5% to 5.0% with this concept has been reported [29]. Another example is using the diblock copolymers composing donor and acceptor segment in the polymerizing unit as building blocks [30]. Ideally, ideal bulk heterojunction can be achieved via the self-assembling of copolymers. The main challenge is to synthesize copolymers with the desired electronic, transport, and chemical properties, that are suitable for device fabrication. 1.5 Loss Mechanisms Light harvest in polymer solar cells undergoes multiple losses. According to Kirchartz et al. [31], losses in polymer solar cells can be classified into four categories: optical losses, exciton losses, recombination losses, and collection losses. The mechanisms of these losses are described in this section Optical Losses Due to the existence of bandgap, only photons with energy higher than E g can be absorbed by semiconducting materials, meaning that the photons with less energy than E g will be wasted. Also, the multilayer structure of devices leads to reflection at layer-layer interfaces and parasitic absorption, the absorption of light in regions other than active layer (electrodes and substrate), that cannot contribute to energy conversion. These losses arising from the optical properties of materials are classified as optical losses. For many polymer solar cells, which are limited by the wide bandgap and/or thin optimal thickness, the optical losses are more pronounced than in inorganic solar cells; in some cases, where the transport is not limiting and/or low bandgap polymer is used, the optical losses can be effectively reduced.

14 1.5. LOSS MECHANISMS 12 Optical Losses Exciton Losses Recombination Losses Collection Losses Figure 1.6: Illustration of multiple losses in photon-electron conversion process Exciton Losses Studies have shown an energy difference exceeding the binding energy of excitons, commonly assigned to 0.3 ev, between LUMO levels of donor and acceptor is required to effectively dissociate the photogenerated excitons in polymer excitation [32]. Likewise, an offset between HOMO levels of donor and acceptor exceeding the binding energy is required for dissociation of fullerene excitation. This energy offset is also known as the driving force. If driving force is not sufficient, EQE will be strongly limited. Moreover, excitons will recombine before reaching the D/A interfaces, if they are generated further away from the interface than their diffusion length. These losses arising from inefficient dissociation of excitons is classified as exciton losses Recombination Losses Recombinations may occur between free carriers before they reach electrodes. Recombinations can be either geminate, electrons recombine with holes generated from the same exciton, or non-geminate, electrons recombine with the holes generated by other excitations. If the recombination is non-radiative, namely no out-coming photons, the excitation is wasted and causes recombination losses. In polymer solar cells, most recombinations are non-radiative [24], usually occured at interfaces or defects in the active layer, or defects at the electrodes [31]. Due to the lower mobility of organic semiconductors, the time for carriers to reach the electrodes is longer and recombination losses is more pronounced in polymer solar cells than in inorganic ones Collection Losses Due to the insufficient mobility, only a fraction of the free carriers can reach the electrodes in a given time. Besides, for non-optimal morphology, carriers generated in the dead-ends cannot be collected. The losses caused by the inefficient transport of carriers are classified as the collection losses. Moreover, the reduced parallel resistance due to undesired short circuit between the electrodes, the shunt resistance, and enhanced serial resistance will also cause collection losses, which are reflected in the decrease in FF.

15 1.6. OPTIMIZATION STRATEGIES Optimization Strategies To optimize the performance of polymer solar cells, there are mainly three aspects to be considered: electronic properties of components, morphology, and thickness. The influences of these factors and optimization strategies are described in this section Electronic Structures Figure 1.7 shows how the electronic structures of donor and acceptor affect the photovoltaic parameters of devices. As J sc decreases with the increase of bandgap (less absorption), V oc scales with the difference between Edonor HOMO and Eacceptor, LUMO and 0.3 ev offset of Edonor LUMO and Eacceptor LUMO is required for exciton dissociation, a proper combination of donor and acceptor has to be chosen to maximize the PCE. PCBM works well as acceptor, therefore the optimization mainly focus on designing donor material with the optimal electronic structure. The current most-studied donor material, P3HT, has the LUMO at 3.2 ev and HOMO at 5.2 ev, resulting an E g of 2.0 ev. The wide bandgap and large HOMO offset strongly limit the harvest of sunlight and V oc ( 0.6 V). Previous study has pointed out the ideal donor has the LUMO at 3.9 ev and HOMO at 5.4 ev, corresponding to a smaller E g of 1.5 ev and higher V oc of 0.9 V [23]. Device made from the ideal donor combined with PCBM has the optimal balance between the light absorption, exciton dissociation, and V oc. Additionally, using novel acceptor with higher LUMO level, such as bispcbm, would increase V oc and potentially the PCE. However, synthesis of materials with the ideal electronic, transport, and chemical properties, that are suitable for device fabrication is difficult. 3.2 P3HT Ideal Donor 5.4 V oc Driving Force 4.2 PCBM 6.0 Figure 1.7: Electronic structures of P3HT, ideal donor, and PCBM. LUMO and HOMO levels in ev are given by the values shown above and below the bars. The offset between LUMO levels of donor and acceptor is considered as the driving force for exciton dissociation in polymer excitation. The difference between HOMO of donor and LUMO of acceptor is proportional to V oc. The ideal donor for PCBM has LUMO at 3.9 ev and HOMO at 5.4 ev, corresponding to a low E g of 1.5 ev.

16 1.6. OPTIMIZATION STRATEGIES Morphology The morphology of active layer plays a crucial role in the performance of polymer solar cells. As mentioned, bi-continuous percolating pathway and dimensions of each domain corresponding to exciton diffusion length are desired. If the domain sizes are too large (coarse film), excitons cannot be dissociated effectively for the lack of interfaces; while if the domain sizes are too small (fine film), recombinations are prone to happen, results in a lower FF. Moreover, if the pathways are too short to contact with the electrodes, dead-ends will form and carriers cannot be collected. Studies have shown the morphology of active layer can be affected by various ways: thermal and solvent annealing [33], choice of solvent [19], spin speed, and blend ratio [34]. The proper treatment for optimal morphology differs from one material system to another Thickness In general, as the thickness of active layer increases, the absorption increases while the transport efficiency decreases. However, as the thickness of active layer is comparable to the wavelength of light, interference effect has to be considered. By modeling the optical properties of materials, the refraction and extinction indices of materials, thicknesses corresponding to the interference maximum and photocurrent can be calculated. With the optimal thickness, more light can be absorbed in a rather thin active layer without losing the transport efficiency.

17 Chapter 2 Fabrication and Characterization 2.1 Material System Donor Alternating polyfluorene copolymer, poly(2,7-(9,9-dioctyl-fluorene)-alt-5,5-(4,7 -di-2-thienyl- 2,1,3 -benzothiadiazole)) (APFO-3), was used as donor in all devices. APFO-3 has a decent carrier mobility and high solubility in many solvents, which makes it suitable for photovoltaic application. S N S N S LUMO APFO-3 n O O O O [60]PBCM O O O bis[60]pcbm O [70]PCBM O bis[70]pcbm O O O 5.8 APFO [60]PCBM 6.0 HOMO bis[60]pcbm 6.1 [70]PCBM 6.0 bis[70]pcbm Figure 2.1: Chemical and electronic structures of materials used in this study. LUMO and HOMO levels in ev are given by the values shown above and below the bars. For the higher solubility, higher energy levels, and lower mobility of bisadduct PCBMs, finer morphology, higher V oc, and lower J sc are expected Acceptor Four different acceptors - [6,6]-phenyl-C 61 -butyric acid methyl ester ([60]PCBM), [6,6]-phenyl- C 71 -butyric acid methyl ester ([70]PCBM), and their bisadduct analogues - were used in this study. In the context, PCBM(s) describes monoadduct PCBMs ([60]PCBM and [70]PCBM), and bispcbm(s) describes bisadduct PCBMs (bis[60]pcbm and bis[70]pcbm), if not specified. 15

18 2.2. DEVICE FABRICATION 16 Chemical and electronic structures of APFO-3 and various PCBMs are shown in Figure 2.1. Due to the less symmetrical structure of [70]PCBM, the absorption is stronger than [60]PCBM in the visible range, so as the bispcbms; therefore, higher photocurrent is expected for devices made from [70]fullerene-derivatives. Comparing with PCBMs, bispcbms have higher HOMO and LUMO levels, up-shifted by 0.1 ev. This up-shift in energy levels will result in an increase in V oc but a decrease in driving forces. With the additional functionalization group, solubility of bispcbms are higher than PCBMs and a finer morphology is expected. However, previous study has shown the mobility of bispcbms is one order of magnitude lower than PCBMs [35]. Besides, the second functionalization group could attach to different positions on the fullerene cage, results in a mixture of isomers of the material with slightly different LUMO levels. This additional disorder may cause a negative influence in the charge-carrier transportation [35]. Therefore, lower J sc is expected for devices made of bispcbms. 2.2 Device Fabrication APFO-3:fullerene bulk heterojunction solar cells were fabricated in this study. The structure of devices can be described as [Glass/ITO/PEDOT:PSS/APFO-3:fullerene/LiF:Al], as illustrated in Figure 2.2. Patterned ITO substrates were cleaned by acetone, detergent, and boiled in TL1 solution (H 2 O : NH 3(aq) (25%) : H 2 O 2 (25%) = 5:1:1 (vol.)) for five minutes. A thin layer of PEDOT:PSS (Baytron P VP Al 4083, EL grade) was spin-coated onto the cleaned ITO substrate with the spin speed of 3000 r.p.m., followed by thermal annealing at 120 C for 15 min. The active layer was spin-coated from blend solutions in a nitrogen glovebox, then transferred into the vacuum chamber of thermal evaporator. LiF (0.6 nm) and Al (80 nm) were deposited as cathode, under the pressure below torr. Different concentrations were used for different solvents. For pure toluene (TO), chloroform (CF), and CF-additive blends, a total concentration of 15 mg/ml were used. For chlorobenzene (CB), 1,2-dichlorobenzene (DCB), and 1,2,4-trichlorobenzene (TCB) blends, higher concentration (from 30 to 60 mg/ml) were used. Devices were fabricated from neat CF if not specified. Varying spin speeds, from 500 to 4000 r.p.m., were applied for different devices. a b LiF:Al cathode Active layer 4.7 ev PEDOT:PSS 5.0 ev 3.5 ev APFO ev 4.1 ev Fullerene 4.2 ev Al PEDOT:PSS ITO 5.8 ev 6.0 ev 6.1 ev ITO anode on glass Figure 2.2: a, Architecture and b, energy diagram of devices in this study. Energy levels are slightly different for different fullerenes, indicated by the dashed-line. The use of PEDOT:PSS can block electron injection from APFO-3 to ITO for the deeper-lying LUMO of PEDOT:PSS, so to improve the current injection efficiency of devices.

19 2.3. DEVICE CHARACTERIZATION Device Characterization Absorption Spectra The absorption spectra of devices were recorded using a Perkin Elmer Lamda 9 Spectrophotometer. The scan range was set from 300 to 700 nm, which probes the absorption edge of the fullerenes. Reflection mode was used, for simulating the actual optical path of incident light in the devices. Transmission mode was used for films of different materials. J-V Characteristics Devices were characterized using a Keithley 2400 Source Meter under illumination of AM 1.5 condition (1000 Wm 2 ) from a solar simulator (Model SS-50A, Photo Emission Tech.). Photovoltaic parameters were recorded by the accompanying software. Calibrations were done by measuring the actual active area of devices using an optical microscope and recalculating the parameters if necessary. External Quantum Efficiency EQE of devices were measured under short circuit condition, using a Keithley 485 picoammeter with a monochromatic light source of halogen lamp. The following equation was used to calculate EQE: EQE(%) = 1240J sc λp in Thickness Profiles The thickness profiles were recorded using a Sloan DEKTAK 3030 surface profilemeter. Total thicknesses of PEDOT:PSS and active layers were obtained by scanning the scratched surface of devices. Thicknesses of active layers were then calculated by subtracting the thickness of PE- DOT:PSS layer, approximately 45 nm, from the total thicknesses. Thicknesses of PEDOT:PSS and active layer in different positions may differ slightly for different fabrication sequences and solvent choice. Although thicknesses of different positions on the samples were measured and averaged, experimental error may still be arisen. Photoluminescence Photoluminescence (PL) spectra were recorded using a blue laser with pumping wavelength of 405 nm as the excitation source. An Oriel optical liquid guide was located to the excitation cite as close as possible. Measurements were done at eleven different positions of the samples and representative data were chosen. PL intensities were calibrated by the absorption of different samples for the thickness variations. A Newton electron multiplying CCD Si array detector cooled to -60 C in conjunction with a Shamrock sr 303i spectrograph from Andor Technology was used as the emission-detection system. Atomic Force Microscopy The surface morphology of films were investigated using an atomic force microscope (AFM) with a Dimension 3100 system (Digital Instruments/Veeco) operating in tapping mode. Silicon cantilevers with a force constant of N/m, a resonance frequency of khz, and a tip curvature radius of 10 nm were used. The scan size was 1 µm 1 µm for all images.

20 Chapter 3 Results and Discussions 3.1 Specific Parameters In this study, we defined some specific parameters to better analyze the results. Up to our knowledge, these parameters have not been used elsewhere. The concepts and physical significance of these parameters are described in this section Absorption Current Density The absorption current density (J abs ) is calculated by converting the absorbed photon flux, the integral of the product of device absorption and photon flux of AM 1.5 solar spectrum over the absorption range, into electrical current (Figure 3.1). Although some photons are absorbed in the electrodes and substrate, other than the active layer, J abs still enable us to compare absorption efficiency among devices. 6.00E E Photon flux (s^-1*cm^-2) 4.00E E E E Device absorption 0.00E Wavelength (nm) Figure 3.1: Scheme of J abs calculation. Absorbed photon flux is first calculated by integrating the product of device absorption, a(λ), and AM 1.5 photon flux, φ AM1.5 (λ), over the absorption range, 300 to 700 nm, then converted into electrical current. 18

21 3.1. SPECIFIC PARAMETERS Saturation Photocurrent Density Figure 3.2 shows the typical device behavior of APFO-3 solar cell in reverse bias. The dashed and dotted curve represents J-V characteristics of devices under illumination and in the dark condition. The solid curve is obtained by subtracting the dark current density from the illuminated current density, namely the net photocurrent density (J net ). When a small bias is applied (-3 V oc ), J net increases monotonically with the reverse bias. In this region, denoted as the limited-extraction region, carrier extraction is limited by the dissociation efficiency and carrier mobility. As the reverse bias increases, the photogenerated carriers experience a stronger field, and become more dissociated and mobile, hence the number of carriers collected at electrodes increases. In a dissociation-efficient device, current extraction is mainly limited by the mobility; in a transport-efficient device, current extraction is mainly limited by the dissociation efficiency. When a strong bias is applied (-6-3 V), denoted as the saturation region, J net starts to saturate. In saturation region, collection of carriers is no longer limited by mobility or dissociation, nearly all the photogenerated carriers can be collected at the electrodes, namely a high IQE is obtained. In this study, we define the saturation current density (J sat ) as the J net value at -5 V. We infer J sat is the maximum extractable current of the device. When an even stronger bias (< -6 V) is applied, devices break down and become not functional. 10 Illum. Dark Net Current density (ma/cm2) 0-10 Jsat Saturation Region Limited-Extraction Region Voltage (V) 0 Figure 3.2: Typical device behavior of APFO-3 solar cell under illumination (dashed) and in the dark condition (dotted). The solid curve represents the net photocurrent density (J net ), obtained by subtracting the dark current density from the illuminated current density. Two regions are specified: limited-extraction region (-3 V oc ) and saturation region (-6-3V). J net at -5 V defines the saturation current density J sat Extraction Ratio In this study, we introduced the concept of extraction ratio to evaluate the dissociation and transport efficiency of devices. Extraction ratio can be calculated by three ways: J sc /J sat, J sat /J abs, or J sc /J abs, J sc /J sat illustrates how much the the maximum possible current is extracted at short circuit condition, which provides insight to the dissociation and transport efficiency of devices. More interfacial area, higher mobility, and sufficient driving force will lead

22 3.1. SPECIFIC PARAMETERS 20 to higher J sc /J sat. J sat /J abs refers to the overall IQE when there is no dissociation or collection loss in devices. In some cases, where J sat is not accessible, J sc /J abs is used Comparison of Experimental and Simulation Data Experimental data of J sc, J sat, J abs, and simulated data are compared in Figure 3.3. All the experimental data are collected from APFO-3:[70]PCBM (1:4) devices. Closed and open symbols represents data collected from devices cast from neat chloroform (CF) and solvents mixtures (mix). Simulation is based the optical modeling of APFO-3:[60]PCBM (1:4) solar cell and multiplied by a factor of 1.5, based on the fact that [70]PCBM devices show 50% more absorption than [60]PCBM devices. IQE of 100% is assumed for the simulation. By calculating the ratio between the peak of experimental J sat and the peak of simulated data, IQE of 75% is obtained for J sat. This observation is consistent with our previous argument that J sat corresponds to a high IQE. By a simple scaling process, J sc and J abs are fitted with simulation. The trend of experimental J abs and J sat fits with simulation confirms the validity of our defined parameters. We observed that the data collected from CF devices fits better to the simulation than devices cast from solvent mixture; this is due to the choice of solvent which can significantly vary the morphology and dissociation efficiency. Notice that the experimental maximum occurs at 70 nm, while the simulation maximum occurs at 80 nm. This shift of 10 nm may arise from the difference between [70]PCBM and [60]PCBM or systematic experimental error, for instance of thickness measurement. Photocurrent (ma/cm2) Simulation Max. (80 nm) Exp. Max. (70 nm) Jsc (CF) Jsat (CF) Jabs (CF) Jsc (mix) Jsat (mix) Jabs (mix) Simulation Fit (Jsc) Fit (Jsat) Fit (Jabs) Thickness (nm) Figure 3.3: Comparison of J sc, J sat, J abs, and simulated data. Closed and open symbols represents data collected from devices using neat chloroform (CF) and solvent mixture (mix) as the solvent.

23 3.2. INFLUENCE OF BLEND RATIO Influence of Blend Ratio A change in blend ratio, or stoichiometry, will induce a change in device absorption and film morphology. Previous study has also shown the carrier mobility in APFO-3:[60]PCBM blend increases with the acceptor loading [36]. APFO-3:[70]PCBM devices of three different stoichiometries - 1:1, 1:4, and 1:9 - have been fabricated. Absorption spectra, J-V characteristics, and surface morphology of devices with different stoichiometries are shown in Figure 3.4. Devices of different stoichiometries exhibit slightly different spectral responses in absorption, due to the different absorption maximum of APFO-3 and [70]PCBM and thickness variation, but no essential difference in the total absorption. From Figure 3.4 b, it can be seen that 1:1 device exhibits a higher slope in the bottom of J-V curve than 1:4 or 1:9 devices, hence a lower FF. We refer this increase in slope to the higher recombination rate of 1:1 device for the finer morphology (Figure 3.4 b), and lower mobility for the low fullerene loading. Similar slopes of 1:4 and 1:9 devices are observed, indicates good transport in both cases, while J sc of 1:4 device is significantly higher. We refer the higher J sc to the increase in interfacial area as polymer loading increases. Therefore, for the balance in transport and charge generation, 1:4 is the optimal stoichiometry for APFO-3:fullerene solar cells. All devices in the following content were fabricated with 1:4 stoichiometry. a b Absorption (norm.) :1 (115 nm) 1:4 (72 nm) 1:9 (80 nm) Current density (ma/cm2) :1 1:4 1: Wavelength (nm) Voltage (V) c d e 5.0 nm 2.5 nm 0.0 nm Figure 3.4: Influence of blend ratio on device performance and film morphology. a, absorption spectra and b, J-V characteristics of devices of different stoichiometries. Thicknesses of devices are given in the parenthesis in a. c-e, AFM images (1 µm 1 µm) of APFO-3:[70]PCBM films of (c) 1:1, (d) 1:4, and (e) 1:9 stoichiometry. Spin speed of 1000 r.p.m. was used for all films.

24 3.3. INFLUENCE OF SPIN SPEED Influence of Spin Speed Using higher speed for spin coating will result in a thinner film and finer morphology. Figure 3.5 shows influence of spin speed on absorption, J-V characteristics, and film morphology of APFO-3:[70]PCBM devices. Device absorption increases with spin speed, from 1000 to 4000 r.p.m., for the thickness approaches to the corresponding value of the absorption maximum (70 nm). The increase in spin speed also results in a higher J sc and PCE, which can be related to the stronger absorption and/or the finer morphology of faster-spun film. Comparing the slopes of devices of different spin speeds, a slightly higher slope in faster-spun device is observed. This increase in slope can be explained by the photo-induced shunt resistance, which is more pronounced in thinner device, or higher recombination rate for the finer morphology. Similar trend is also observed in APFO-3:[60]PCBM devices. a b Absorption (norm.) r.p.m. (92 nm) 2000 r.p.m. (79 nm) 4000 r.p.m. (72 nm) Wavelength (nm) Current Density (ma/cm2) r.p.m r.p.m r.p.m Voltage (V) c d 5.0 nm 2.5 nm 0.0 nm Figure 3.5: Influence of spin speed on device performance and film morphology. a, Device absorption and b, J-V characteristics of APFO-3:[70]PCBM devices using different spin speeds. c,d, AFM images (1 µm 1 µm) of APFO-3:[70]PCBM films using spin speed of (c) 1000 r.p.m. and (d) 2000 r.p.m. A slightly different trend is found in APFO-3:bisPCBM devices. Figure 3.6 shows the influence of spin speed on J-V characteristics of APFO-3:bis[60]PCBM and APFO-3:bis[70]PCBM devices. Still, J sc increases with spin speed. However, the change in slope is more pronounced than APFO-3:PCBM devices, indicates a more mobility-limited behavior of APFO-3:bisPCBM devices. Notice that J sc of APFO-3:bisPCBM devices are much lower than for APFO-3:PCBM devices. The reasons for the severe drop in J sc of APFO-3:bisPCBM devices are discussed in detail in the subsequent section.

25 3.3. INFLUENCE OF SPIN SPEED 23 a b Current density (ma/cm2) r.p.m r.p.m r.p.m. Current density (ma/cm2) r.p.m r.p.m r.p.m Voltage (V) Voltage (V) Figure 3.6: Influence of spin speed on J-V characteristics of (a) APFO-3:bis[60]PCBM and (b) APFO-3:bis[70]PCBM devices. Influence of blend ratio and spin speed on extraction ratios of APFO-3:PCBM devices are shown in Figure 3.7. For nearly all devices, J sc /J sat increases with spin speed, which can be explained by the increase in dissociation efficiency with interfacial area or the better transport of thinner device. In Figure 3.7 a, as previously observed, APFO-3:[70]PCBM (1:4) devices exhibit the best extraction among all stoichiometries, for the better transport and more efficient dissociation. The extraction of 1:1 device is better than 1:9 device, which may due to the formation of large fullerene domains for the high fullerene loading and therefore a strong field is required to dissociate excitons in the pure fullerene domains. The drop in J sc /J sat of 1:1 device with 4000 r.p.m. may refer to the intermixing of polymer/fullerene phase due to the inherent fine morphology. On the other hand, in Figure 3.7 b, J sat /J abs of all devices remain constant around 0.5, indicates that the same portion of absorbed light can be converted into current in all devices providing a strong enough field. a 0.8 [60]PCBM 1:4 [70]PCBM 1:4 [70]PCBM 1:1 [70]PCBM 1:9 b 0.8 [60]PCBM 1:4 [70]PCBM 1:4 [70]PCBM 1:1 [70]PCBM 1: Jsc/Jsat 0.4 Jsat/Jabs Spin Speed (r.p.m.) Spin Speed (r.p.m.) Figure 3.7: Influence of blend ratio and spin speed on extraction ratios (a) J sc /J sat and (b) J sat /J abs of APFO-3:PCBM devices.

26 3.3. INFLUENCE OF SPIN SPEED 24 Notice that extraction ratios of APFO-3:bisPCBM devices are not included in Figure 3.7. In APFO-3:bisPCBM devices, the net photocurrent increase with reverse bias without reaching saturation, then leads to device breakdown, makes the J sat inaccessible. We referred this behavior to the ineffective dissociation and/or low mobility of bispcbms. a Absorption (norm.) Wavelength (nm) [60]PCBM [70]PCBM bis[60]pcbm bis[70]pcbm b Current density (ma/cm2) Voltage (V) [60]PCBM [70]PCBM bis[60]pcbm bis[70]pcbm Figure 3.8: Influence of the choice of acceptor on a, absorption spectra and b, J-V characteristics of APFO-3:fullerene devices. To compare the performance of devices using different fullerenes as acceptor, absorption spectra and J-V characteristics are shown in Figure 3.8. In Figure 3.8 a, small spectral shifts in absorption are observed when replacing [60]PCBM or [70]PCBM by its bisadduct analogue, but no significant difference in the total absorption is observed. Table 3.1 summarized photovoltaic parameters of these devices. The use of bispcbms leads to an enhanced V oc, as expected by the higher LUMO levels. Compared to [60]fullerene-derivatives, using [70]fullerene-derivatives results in an enhanced absorption of 50%. With this increase in absorption, APFO-3:[70]PCBM device exhibits an increase in J sc of 18% compared to APFO- 3:[60]PCBM device; however, APFO-3:bis[70]PCBM device exhibits a decrease in J sc of 15% compared to APFO-3:bis[60]PCBM device. The reason for using a strong-absorbing acceptor in APFO-3:bisPCBM devices does not contribute to higher J sc is the ineffective dissociation of the fullerene excitation, as proven in the subsequent section. For the dramatic drop in J sc, PCE of APFO-3:bisPCBM devices are much lower than APFO-3:PCBM devices despite the increase in V oc. APFO-3:[70]PCBM device exhibits the best performance among all devices, for the higher absorption and efficient dissociation. Table 3.1: Photovoltaic parameters of devices using different fullerenes as acceptor. Notice that J sat of bispcbm devices are calculated assuming J sc /J abs equals to 0.5. Acceptor J sc V oc FF PCE J sat J abs (ma/cm 2 ) (V) (%) (ma/cm 2 ) (ma/cm 2 ) [60]PCBM [70]PCBM bis[60]pcbm * bis[70]pcbm * 16.08

27 3.4. INFLUENCE OF SOLVENT 25 Since J sat are not measurable in APFO-3:bisPCBM devices, J sat of those devices are calculated by assuming J sat /J abs equals to 0.5, as observed in Figure 3.7 b. Based on this assumption, J sc /J sat of devices are plotted in Figure 3.9. For all devices, a higher spin speed results in an enhanced J sc /J sat. This trend again indicates the more efficient dissociation and better transport for the faster-spun film. Comparing with devices using [70]-derivatives as acceptor, a better current extraction of [60]-derivatives devices is observed. This observation is explained by the less efficient dissociation of the fullerene excitation than the polymer excitation, hence a less absorbing fullerene will lead to an overall higher dissociation efficiency. Also, much higher J sc /J sat of APFO-3:PCBM devices than APFO-3:bisPCBM devices is observed, due to the very inefficient dissociation of fullerene excitations in APFO-3:bisPCBM devices. 0.8 [60]PCBM [70]PCBM bis[60]pcbm bis[70]pcbm 0.6 Jsc/Jsat Spin Speed (r.p.m.) Figure 3.9: Extraction ratio J sc /J sat of APFO-3:fullerene devices. For all devices, J sc /J sat increases with spin speed. Using [60]-derivatives and monoadduct fullerenes lead to higher J sc /J sat. 3.4 Influence of Solvent Six different solvents and their combinations were used to examine the best process condition for our devices. Vapor pressure and solubility data of monoadduct fullerenes are given in Table 3.2. Solvent of higher vapor pressure has higher evaporation rate, hence faster film formation; solvent of higher PCBM solubility leads to finer morphology. For the very low solubility of PCBM in xylene (XY), no devices was fabricated from neat XY. Vapor pressure, PCBM solubility of 1,2,4-trichlorobenezene (TCB) and solubility data of bisadduct fullerenes in all solvents have not been reported, but higher solubility and finer morphology are expected. Previous report has shown a small amount of solvent additive can lead to great influence in film morphology and PCE [19]. In this study, devices were made from both solvent mixture (additive devices) and neat solvents (neat devices) to investigate their influences. For solvent mixtures, additives were added into CF with the ratio of 1:80 in volume. Influence of solvent on performance of APFO-3:[70]PCBM devices are shown in Figure Thicknesses of additive devices are controlled in the range of nm, while thicknesses of neat devices are controlled in the range of nm, except one thick device of 180 nm is made from neat chlorobenzene (CB180). No spectral difference is observed for both additive and neat devices; however, for the

28 3.4. INFLUENCE OF SOLVENT 26 Table 3.2: Vapor pressure and PCBM solubility in different solvents [34, 37]. Solvent Vapor pressure Solubility 20 (mmhg) [60]PCBM [70]PCBM chloroform (CF) chlorobenzene (CB) ,2-dichlorobenzene (DCB) toluene (TO) xylene (XY) thickness variation induced by the change of solvent, a slight variation in absorption is observed. As expected, CB180 shows a considerably stronger absorption than other devices. Notice that devices made from TO-, XY-additive, and neat TO exhibit a much stronger absorption, due to the strong light scattering of rough surfaces, which cannot contribute to photocurrent generation. a 1.0 b Absorption CF (65) CFCB (75) CFDCB (65) CFTCB (70) CFXY (55) CFTO (60) Absorption CF (80) CB (75) CB180 (180) DCB (90) TCB (70) TO (75) c Wavelength (nm) d Wavelength (nm) 0 0 Current density (ma/cm2) Voltage (V) CF CFCB CFDCB CFTCB CFXY CFTO Current density (ma/cm2) Voltage (V) CF CB CB180 DCB TCB TO Figure 3.10: Influence of solvent on performance of APFO-3:[70]PCBM devices. CFCB and CB represents devices made from solvent mixture of CF:CB = 80:1 (vol.) and neat CB. a,b, Absorption spectra of (a) additive devices and (b) neat devices. c,d J-V characteristics of (c) additive devices and (d) neat devices. Thicknesses of active layer in nm of each device are given in the parenthesis in a and b. J-V characteristics of devices (Figures 3.10 c,d) show the use of high [70]PCBM solubility solvents (CB, DCB, and TCB), either by additives or neat solvents, can improve the J sc significantly. Consequently, J-V curves become steeper and FF are decreased. On the other hand, the use of low [70]PCBM solubility (TO and XY) lead to a decrease in J sc but similar slopes. The change in J-V curves is explained by the change in morphology induced by solvents. Using

29 3.4. INFLUENCE OF SOLVENT 27 a b c 10.0 nm 10.0 nm 50.0 nm 5.0 nm 5.0 nm 25.0 nm 0.0 nm 0.0 nm 0.0 nm d 10.0 nm e 10.0 nm f 10.0 nm 5.0 nm 5.0 nm 5.0 nm 0.0 nm 0.0 nm 0.0 nm Figure 3.11: Influence of solvent on morphology of APFO-3:[70]PCBM and APFO- 3:bis[70]PCBM films. a-c, AFM images of APFO-3:[70]PCBM film cast from (a) CF, (b) CB, (c) TO. d-f AFM images of APFO-3:bis[70]PCBM film cast from (d) CF, (e) CB, (f) TO. Sizes of all images are 1 µm 1 µm. solvents of higher [70]PCBM solubility leads to a finer morphology and more interfacial area (Figure 3.11 b), therefore a more efficient dissociation of devices; oppositely, using solvents of lower [70]PCBM solubility leads to a coarser morphology and less efficient dissociation (Figure 3.11 c). The increase in interfacial area results in the higher J sc but also higher recombination rate, hence the lower FF. Photovoltaic parameters of devices are summarized in Table 3.3. The influence of solvent become more pronounced in neat devices than in additive devices. A slight decrease in V oc is observed in additive devices of CB, DCB, or TCB, while the variation is less significant in neat devices. A possible explanation for this variation is the interaction between the matrix and additive solvents. Interestingly, CB180 still exhibits higher J sc than CF device, but a much lower FF. This indicates J sc in APFO-3:[70]PCBM devices is mainly limited by the dissociation, while FF is limited by the transport. CB and TCB devices exhibit the best performance of PCE around 2.6%, significantly higher than CF device. Table 3.3: Photovoltaic parameters of APFO-3:[70]PCBM devices made from different solvents. Units for J sc, V oc and PCE are ma/cm 2, V and %, respectively. Additive J sc V oc FF PCE Neat J sc V oc FF PCE CF CF CFCB CB CFDCB DCB CFTCB TCB CFTO TO CFXY CB For a better insight of carrier generation, influence of solvent on extraction ratios of APFO- 3:[70]PCBM devices are shown in Figure From Figure 3.12 a, it can be seen that solvent additives have considerable influence in J sc /J sat. The use of higher [70]PCBM solubility solvent leads to finer morphology and more efficient dissociation. As observed in J-V characteristics, a similar but more pronounced effect is observed in neat devices than additive devices. Among all devices, CB device exhibits the highest J sc /J sat of 0.9, nearly twice of CF device (0.5).