Towards Efficient Carbon-Based Solar Cells

Transcription

1 Towards Efficient Carbon-Based Solar Cells Investigators 2014 GCEP Report - External Zhenan Bao, Professor, Chemical Engineering Huiliang Wang, Graduate Researcher Ghada Koleilat, Postdoctoral Researcher Myung Gil Kim, Postdoctoral Researcher Kwanpyo Kim, Postdoc Researcher Abstract Our work in the last year focused on improving the sorting of semiconducting carbon nanotubes and the enhancement of the conductivity and transparency of carbon based electrodes. The active layer of our carbon solar cell is based on semiconducting single-walled carbon nanotubes; the highly selective sorting of small diameter carbon nanotube is essential for photovoltaic applications. Ideally, the active layer of a solar cell comprises of a material with bandgap near 1.3 ev corresponding to the optimal single junction bandgap for highest possible efficiency. Thus, our efforts focus on the high yield sorting process to selectively disperse semiconducting nanotubes with bandgaps around 1.3 ev. Also, all efficient solar cells comprise of a highly conductive transparent electrode to enable light transmission into the active layer and efficient carrier extraction. In this report, we first describe a high-yield method to selectively disperse semiconducting CoMoCAT (CO disproportionation on Co-Mo Catalysts) single-walled carbon nanotubes (SWNTs) with regioregular poly (3-alkylthiophenes) polymers. We observed that the dispersion yield was directly related to the length of the polymer s alkyl side chains. Molecular dynamics simulations in explicit toluene (real 2014 GCEP Report - External 1

2 toluene molecules) indicate that polythiophenes with longer alkyl sidechains bind stronger to SWNTs, due to the increased overall surface contact area with the nanotube. Furthermore, the sorting process selectively enriches smaller-diameter CoMoCAT SWNTs with larger bandgaps, which is ideal for solar cell applications. Compared to the larger diameter sorted HiPco (High-Pressure CO) SWNTs, solar cells fabricated using our sorted CoMoCAT SWNTs demonstrated higher open-circuit voltage (V oc ) and infrared external quantum efficiency (EQE). The V oc achieved is the highest reported for solar cells based on SWNT absorbers under simulated AM1.5 solar illumination. Additionally, we employed the sorted CoMoCAT SWNTs to fabricate thin film transistors with excellent uniformity and device performance. We also conclude this section by listing ongoing efforts to improve film processing, electrode transparency and doping, and finally device performance. We also report a simple and scalable method to enrich large quantities of semiconducting arc-discharged single-walled carbon nanotubes (SWNTs) with diameters of 1.1nm~1.8nm using dithiafulvalene/thiophene copolymers. Stable solutions of highly individualized and highly enriched semiconducting SWNTs were obtained after a simple sonication and centrifuge process. Molecular Dynamics (MD) simulations of polymer backbone interactions with and without side chains indicated that the presence of long alkyl side chains gave rise to the selectivity towards semiconducting tubes, indicating the importance of the roles of the side chains to both solubilize and confer selectivity to the polymers. We found that by increasing the ratio of thiophene to dithiafulvalene units in the polymer backbone (from pdtff-1t to pdtff-3t), we can slightly improve the selectivity towards semiconducting SWNTs. These experimental and modeling results provide a better understanding for future rational design of polymers for sorting SWNTs. We conclude this section with insights on our future work regarding SWNTs sorting GCEP Report - External 2

3 Content 1 Introduction Small Diameter Single-Walled Carbon Nanotubes for Photovoltaic Applications High-Yield Sorting of Small-Diameter CoMoCAT SWNTs (Wang and Koleilat et al. ACS Nano 2014)... 5 Improving Yield by Side-Chain Modification... 6 Selective Dispersion of Small-Diameter SWNTs Small-Diameter CoMoCAT SWNTs for Solar and Transistor Applications (Wang and Koleilat et al. ACS Nano 2014) Photovoltaic Device Thin Film Transistors Ongoing Work on Material properties, Film processing and Electrode Doping Scalable and Selective Dispersion of Semiconducting Arc-Discharged Carbon Nanotubes (Wang et al., ACS Nano 2013) UV-Vis-NIR characterization of the large diameter SWNTs Molecular Dynamics (MD) Simulations Ongoing Work on SWNTs Material Processing Summary Published Work Contacts GCEP Report - External 3

4 1 Introduction Our work in the last year focused on improving the sorting of semiconducting carbon nanotubes and the enhancement of the conductivity and transparency of carbon based electrodes. The active layer of our carbon solar cell is based on semiconducting single-walled carbon nanotubes; the highly selective sorting of small diameter carbon nanotube is essential for photovoltaic applications. Ideally, the active layer of a solar cell comprises of a material with bandgap near 1.3 ev corresponding to the optimal single junction bandgap for highest possible efficiency. Thus, our efforts focus on the high yield sorting process to selectively disperse semiconducting nanotubes with bandgaps around 1.3 ev. Not only, we expected to achieve higher open circuit voltage but also higher carrier separation efficiencies because of the enhanced band gap alignment between the carbon nanotubes and the electron acceptor C 60 employed. Also, all efficient solar cells comprise of a highly conductive transparent electrode to enable light transmission into the active layer and efficient carrier extraction. Herein, we list our achievements that bring us closer to our realization of an efficient all-carbon solar cell: The successful realization of a high-yield sorting process of small-diameter CoMoCAT singlewalled carbon nanotubes (SWNT) idea for single junction solar cells, published in ACS Nano 2014 and highlighted in several media outlets. Section 2 of this report describes the details of this work. We also achieved a scalable and selective sorting of large diameter SWNTs that absorb far into the infrared and thus access a region unavailable to traditional photovoltaic materials. We hope to utilize this material in multijunction application. This work was also published in ACS Nano and is detailed in section 3. We also have ongoing efforts regarding conductive and transparent graphene based electrodes, film deposition techniques influencing carbon nanotube packing, and sorting in various solvents media GCEP Report - External 4

5 2 Small Diameter Single-Walled Carbon Nanotubes for Photovoltaic Applications 2.1 High-Yield Sorting of Small-Diameter CoMoCAT SWNTs (Wang and Koleilat et al. ACS Nano 2014) Single-walled carbon nanotubes (SWNTs) have been widely studied in the field of electronics and optoelectronics due to their high carrier mobility, strong absorptivity and their direct bandgaps. Applications include thin film transistors, logic circuits, solar cells and photodetectors. However, the assynthesized SWNTs are typically mixtures of tubes with various chiralities, with approximately one-third metallic and the remaining two-third semiconducting. They also have a wide range of diameters and bandgaps, which have significant impact on device performance. As a reference, arc-discharge SWNTs have diameters ranging between 1.2~1.7 nm, HiPco SWNTs have diameters of 0.8~1.1 nm while CoMoCATs SWNTs have smaller diameters of 0.7~0.9 nm. Until now, great progress has been made on sorting SWNTs by solution-based non-covalent functionalization of SWNTs. This approach has the advantage of not significantly alter the SWNTs electronic properties. Examples of this approach include density gradient centrifugation, sorting by DNAs, gel chromatography, and selective dispersion by conjugated polymers. However, most of these processes suffer from low-yield, lengthy procedures or difficulty in removing insulating sorting reagents. Sorting of SWNTs by conjugated polymers is of particular interest because of its simplicity, high selectivity and high-yield. Herein, we report the selective dispersion of small-diameter semiconducting CoMoCAT SWNTs using poly(3-alkylthiophene)s (rr-p3ats). We observed that rr-p3ats selectively enriched CoMoCAT SWNTs with diameters less than 0.76 nm. By increasing the length of the alkyl side chains, we were able to increase the dispersion yield. As a result, thicker films can be prepared to allow more light absorption for solar cell devices. Molecular dynamics simulations in presence of toluene solvent molecules indicated 2014 GCEP Report - External 5

6 that the increased yield is due to the higher contact surface area between longer side-chain polymers and SWNTs. Improving Yield by Side-Chain Modification The sorting procedure was similar to our previously reported process which involved only a simple sonication followed by a centrifugation step. First, 5 mg of CoMoCATs and 5 mg of polythiophenes were mixed in 25 ml of toluene for the dispersion process. After the sonication and centrifugation steps, a very dark solution was obtained (Figure 1a). Ultraviolet-Visible-Near Infrared (UV-Vis-NIR) spectroscopy was used to measure dispersion yield for polymers of various alkyl side chain lengths (where R = 8, 10, 12, 13 shown in Figure 1b). Compared to unsorted SWNTs dispersed in N-methyl-2-pyrrolidone (NMP), the peaks representing SWNTs with different chiralities are well resolved in the rr-p3at dispersions, indicating well dispersed SWNTs. The yield estimated for P3OT (R = 8), P3DT (R = 10), P3DDT (R = 12) and P3TDT (R = 13) sorted SWNTs were 4%, 17%, 25% and 31%, respectively. These values were derived by comparing the semiconducting E S 11 peak resonance areas in the absorption spectrum with those from NMP dispersion with a known SWNT concentration. We ruled out the possibility that this increase in dispersion yield was due to the increase in molecular weight of the polymer because the molecular weight of the polymer did not show any particular trend when the length of the side chains were varied from C8 to C13. We speculate that regardless of the diameters of the SWNTs, polymers with longer side chains generally result in a higher dispersion yield of SWNTs GCEP Report - External 6

7 Figure 1. (a) Sorted CoMoCAT carbon nanotube solutions by P3DDT, (b) Polymer structure of polythiophenes with different side chains. (c) Optical Absorbance spectra comparing unsorted CoMoCAT SWNTs dispersed in NMP to the ones sorted by polythiophene in toluene. To investigate the origin of observed increased dispersion yield with increased alkyl side-chain length, we performed molecular dynamics (MD) simulations using AMBER 12 and the general AMBER force field (GAFF). During the initial stages of the implicit solvation simulation, the polymer backbones gradually evolved from a linear arrangement to a more helical wrapping around the circumference of the nanotube. After approximately 200 ps, the polymer backbone extended to a more stable quasi-linear conformation and remained in contact with the nanotube surface. A greater propensity for formation of helical conformations of the polymer backbones was observed when the simulations were performed in explicit solvents. Strong contact of the alkyl side chains with the nanotube surface was observed throughout the simulations since most alkyl side chains remained bound to the nanotube surface. Snapshots of the structures of the polymers interacting with (6,5) SWNT obtained after 60 ns simulation in explicit solvent are shown in Figure 2a. The flexibility of the polythiophene backbone and the alkyl side chains, together with the good line-up of the alkyl side chains due to the regioregularity, collectively resulted in the tight wrapping of the polymer to the curved nanotube surface GCEP Report - External 7

8 Figure 2. (a) Representative snapshots of the MD simulations for (6,5) SWNT with P3OT, P3DT, P3DDT, and P3TDT in explicit toluene. (b) Calculated binding energies (E B ) vs. yield of sorted (6,5) SWNTs. (c) Calculated binding energies (E B ) vs. contact surface area (CSA) between the polythiophene polymers and (6,5) SWNTs. Energies and contact surface areas were averaged along 60 ns MD trajectories. CSA were calculated using UCSF Chimera 1.8. We computed the binding energy (E B ) of the SWNT/polythiophene complex as follows: E B = [E complex (E SWNT + E polythiophene )], where E complex, E SWNT and E polythiophene are the energies of the SWNT/polythiophene complex, the nanotube, and the polymer, respectively, based on the implicit generalized Born (GB) solvation model. The computed average binding energies during the 60 ns explicit solvation simulation of the four different polymers with (6,5) SWNT are shown in Figure 2. During this period, when the systems were completely equilibrated, the average binding energies of P3DDT (C12) and P3TDT (C13) with (6,5) CoMoCAT SWNT were similar and the greatest among the four polymers. This observation is in agreement with the measured amounts of SWNTs dispersed (Figure 2b). The origin of the greater binding energies of the P3DDT and P3TDT polymers was attributed to their longer side chains, which led to higher contact surface area (CSA) for the same length of polymer and thus promoted the binding with the nanotubes. For all the polymers studied, the computed binding energies correlated well with the CSA (Figure 2c) GCEP Report - External 8

9 Another factor that might have contributed to the efficient SWNT dispersion by polymers with longer alkyl side chains is that the stronger side chain/solvent interactions lead to increased solubility of the wrapped SWNT in toluene. As depicted in the representative structures, the shorter alkyl groups appeared to solvate better in toluene in comparison to implicit solvent (Figure 2a). This observation is somewhat counterintuitive and has been observed for high molecular weight alkanes. Longer alkyl chains seemed to have a larger SWNT/toluene partition coefficient than the shorter chains. These results reinforce the notion that the origin of greater SWNT dispersion yield with polymers with long alkyl side chains is the stronger interactions of these peripheral residues with the nanotubes, achieving more effective wrapping of SWNTs. The principle of increasing the CSA between the polymers and nanotubes is highly applicable to design more efficient polymers for sorting SWNT dispersion. Selective Dispersion of Small-Diameter SWNTs The sorting selectivity of the CoMoCAT SWNTs by rr-p3at polymers was characterized by UV-Vis- NIR, Raman resonance spectroscopy and photoluminescence measurements. First, as seen in the UV- Vis-NIR spectrum shown in Figure 1b, we observed identical peak positions and similar relative selectivity for SWNTs sorted by polymers with different side-chain lengths. In comparison with unsorted SWNTs in NMP, we observed that the absorption peaks for (7,5) and (7,6) SWNTs at a wavelength of 634 nm were largely suppressed, while new features appeared in the sorted SWNTs, e.g. a small (9,1) peak at 975 nm. We observed that the polymers with longer alkyl side chains are slightly less selective to SWNTs because their higher contact surface area allows them to bind with more types of SWNTs. Since polymers with different alkyl chain lengths vary so little in their selectivity towards SWNTs, the CoMoCAT SWNTs sorted by commercially available P3DDT will be examined further as an example for their composition and device performance. Using Raman spectroscopy, major peaks of CoMoCAT SWNTs were probed with lasers excitation energies 1.96 ev (633nm) and 1.58 ev (785nm). The radial breathing mode (RBM) of the SWNTs 2014 GCEP Report - External 9

10 selected before and after sorting by P3DDT is shown in Figure 3. When the SWNTs were irradiated using the 633 nm laser, we observed that the peaks of the metallic SWNTs that were originally present in the unsorted SWNTs were almost completely eliminated after sorting. In the regime for semiconducting SWNT resonance, the peaks for (10,3), (11,1), (7,6), (7,5) and (8,3) SWNTs were all suppressed, which accords with the reduction of (7,5) and (7,6) peaks observed by UV-Vis-NIR measurements. On the other hand, the smallest diameter (6,5) and (6,4) SWNT peaks were clearly enriched as evidenced from the increased relative intensity. With the 785 nm laser excitation, the peaks of the relatively large-diameter (8,6), (9,5) and (12,1) SWNTs were also suppressed significantly after the sorting process. In comparison with (8,6)/(9,5) peaks in UV-Vis-NIR spectrum in Figure 1c, the reduction of (8,6)/(9,5) peaks at Raman 785nm laser are much more distinct. This might be due to the residuals of the burned polymers that shift the SWNT E S 22 resonance spectra from the laser excitation energy. Figure 3. Raman spectra of unsorted CoMoCAT SWNTs dispersed by NMP and sorted ones dispersed by P3DDT: (a) with 633nm laser excitation (b) with 785nm laser excitation. Using the UV-Vis-NIR and Raman spectroscopy features, we summarize the enrichment of the semiconducting CoMoCAT SWNT species in the Sorted Chirality Map shown in Figure 4. The orange hexagons represent the relative enrichment in SWNT types while the light green hexagons represent the relative reduction in SWNT types. The polythiophene polymer sorting process clearly dispersed semiconducting SWNTs with diameter <0.76 nm, which are the smallest-diameter species among all the CoMoCAT SWNTs. Fortunately, the selected SWNTs possess first inter-band transition (S11) larger 2014 GCEP Report - External 10

11 than 1.1 ev, which correspond to desirable bandgaps for materials integrated in a solar cell. In fact, the optimal bandgap for a single junction photovoltaic cell is 1.3 ev considering the Shockley-Queisser limit. Figure 4. Sorted Chirality Map showing the enrichment of CoMoCAT SWNTs by P3DDT: orange hexagons and light green circles are showing the increase or decrease in the abundance of SWNTs after sorting with P3DDT respectively from UV-Vis-NIR and Raman spectroscopy. 2.2 Small-Diameter CoMoCAT SWNTs for Solar and Transistor Applications (Wang and Koleilat et al. ACS Nano 2014) In comparison with poly(3-dodecylthiophene-2,5-diyl) (P3DDT) sorted larger-diameter HiPco SWNTs, P3DDT-sorted CoMoCAT SWNTs enabled an increase in open-circuit voltage from 0.39V to 0.44V when used as the active material in solar cells. In addition, the external quantum efficiency (EQE) of the solar cell reached 16% at an excitation wavelength of 1020 nm, compared to 4% for HiPco SWNT solar cells at its peak wavelength of 1100nm. Furthermore, sorted small-diameter CoMoCAT SWNT thin film transistor devices demonstrated an average mobility of >1cm 2 /Vs and an average on/off ratio >10 4 with excellent uniformity. Photovoltaic Device As shown in Figure 5a, the CoMoCAT SWNTs film possess absorption peaks closer to the 1.3 ev bandgap (~950 nm) than the HiPco SWNTs, which have peaks up to 0.85 ev (~1450 nm). By 2014 GCEP Report - External 11

12 employing the heterojunction structure in all our devices, we proceed to evaluate the contribution of the sorted CoMoCAT SWNTs to the overall efficiency in comparison with the sorted HiPco SWNTs. Figure 5. (a) Comparison of the UV-VIS-NIR absorption spectrums between P3DDT sorted CoMoCAT and HiPco SWNTs films. (b) AFM image of the sorted CoMoCAT SWNTs layer. (c) Device structure of the carbon based solar cell structure. (d) Schematic diagram showing the band structure predicted open circuit voltage and charge transfer of SWNTs/C 60 heterojunction. (e) J-V characteristics of the devices measured under standard illumination AM1.5 for CoMoCAT SWNTs and HipCO SWNTs dispersed by P3DDT. (e) External quantum efficiency for the comparison of HipCO and CoMoCAT SWNTs in the IR regime. The active layer consists of a bilayer of 70 nm of C 60 on top of a spincoated thin layer of SWNTs (~3-5 nm thick). We used ITO as the anode and poly (3,4- ethylenedoxythiophene):poly(styrenesulfonate(pedot:pss)) layer as a hole transporting layer. 70 nm of Ag was used as the top reflective cathode. We present an AFM image of the rr-p3ddt dispersed 2014 GCEP Report - External 12

13 CoMoCAT spincoated film in Figure 5b, and both the schematic of the device structure and their corresponding energy band level diagrams are shown in Figures 5c and 5d, respectively. As depicted in Figure 5e, CoMoCAT SWNTs based solar cells exhibited systematically higher open circuit voltages of 0.44 V, compared to the 0.39 V open circuit voltage of solar cells fabricated from HiPco SWNTs. We attributed this increase in open circuit voltage to the smaller diameter and lower bandgap tubes in the sorted CoMoCAT SWNTs. The open circuit voltage is generally related to the difference between the HOMO of SWNTs and the LUMO of C 60, which are labeled in Figure 6d. As mentioned previously, employing CoMoCAT SWNTs in solar cells should enable a more efficient charge separation which should be directly translated into a higher short-circuit current. However, as depicted in the J-V curve (Figure 5e), we observed that the currents generated from both types of solar cells are approximately equivalent. We thus analyzed the amount of infrared and visible photons absorbed by the SWNTs and the polymers, respectively, in the solar cells and compared it to the amount of charges they each contributed to the current (Figures 5e, f). In Figure 5a, we observed that the 5 nm thick P3DDT wrapped HiPco SWNTs films had higher absorption peak intensities not only in the near-infrared (at 1150 nm), but also in the visible at 550 nm indicating a slightly higher polymer content in the spin-coated films. Thus, for HiPco based solar cells, the current generated is dominated by the contribution of the P3DDT absorption at the visible spectrum, rather than the infrared contribution of the SWNTs. On the other hand, for the CoMoCAT based films, there was a significant contribution from the SWNTs due to the more efficient charge transfer from the small diameter SWNTs to the C 60 (larger LUMO offsets between SWNTs and C 60 ), which was further highlighted by the high EQE peak of 16% at 1050 nm (Figure 5f). We clearly observed a lower nearinfrared carrier separation efficiency from the large diameter HiPco SWNTs to the C 60 with the EQE peaks at 1100 nm being suppressed reaching a maximum of only 4% and the peaks above 1200 nm were almost completely eliminated. It was previously reported EQE peaks higher than 30% at 1050 nm from PFO sorted (7,5) CoMoCAT based solar cells. Their higher EQE value maybe due to the inclusion of a 2014 GCEP Report - External 13

14 buffer layer to minimize charge recombination and the use of optimized thickness for their SWNT film to maximize interference effects at 1050 nm. Interestingly, the V OC (0.44 V) of our device is higher than their V OC of 0.38 V at AM1.5 spectrum. This result was due to our utilization of sorted smaller diameter CoMoCAT tubes, predominantly (6,5) rather than the (7,5), which highlights the benefit of using even smaller diameter SWNTs. Thin Film Transistors We successfully demonstrated the merits of our rr-p3ddt sorted CoMoCAT SWNTs for solar cell applications. In this section, we applied these tubes for thin film transistor applications. We fabricated bottom-gate bottom-contact thin film transistors (TFTs) with highly doped silicon substrate as a gate and 300 nm SiO 2 as the dielectric with the sorted CoMoCAT SWNTs. The device structure is shown in Figure 6a and an AFM image of the SWNT films in the channel is shown in Figure 6b. The transfer and output curves of a typical device are shown in Figures 6c, d. Our fabricated devices were observed to exhibit excellent uniformity and have an average charge carrier mobility of 1.18 ± 0.28 cm 2 /Vs and average on/off ratio of (2.36 ± 4.52) x 10 4, as computed from more than 20 randomly chosen devices. The high on/off ratio at this high SWNT density (Figure 6b) also supports the high selectivity of the semiconducting SWNTs by polymers. For random network SWNT TFTs, the contacts between the smalldiameter SWNTs and metals are not the limiting factor as is the case for single SWNT devices. Indeed, we observed negligible contact resistances as shown from the output curves in Figure 6d. Rather, the limiting factor of our thin-film transistor can be attributed to tube-to-tube junctions between the SWNTs, which under percolation transport has been shown to dominate the resistance of the channel. Therefore, by implement alignment techniques to optimally set the tube-to-tube junctions or by reducing the tube-totube junction resistance can potentially improve the performance of our transistor GCEP Report - External 14

15 Figure 6. Electrical transport properties of thin film transistors made of P3DDT dispersed CoMoCAT SWNTs in toluene. (a) Schematic diagram of the device structure (L = 20 µm, W = 400 µm) (b) AFM image of P3DDT sorted CoMoCAT SWNTs (c) Transfer curve of a typical device (V SD = -30 V) (d) Output curve of the device in (c) at different gate voltages. 2.3 Ongoing Work on Material properties, Film processing and Electrode Doping With the demonstration of the small diameter SWNTs based solar cell, we are currently aiming to improve the overall performance of the solar cell to showcase its viability as a novel solar technology able to harvest the near-infrared spectrum of the solar irradiance as well as its visible region. To that end, we are mainly pursuing the following directions: Comparing various film deposition techniques to improve carbon nanotube packing Employing oxide dopants for graphene based electrodes 3 Scalable and Selective Dispersion of Semiconducting Arc- Discharged Carbon Nanotubes (Wang et al., ACS Nano 2013) Here, we report a facile and efficient process for selective dispersion of large quantity of semiconducting AD-SWNTs via the use of a series of polymers based on poly(dithiafulvalene-fluorene-co-m-thiophene) 2014 GCEP Report - External 15

16 (pdtff-mt) (Figure 7a). This method is one of the few methods that can greatly enrich the largediameter semiconducting SWNTs but also result in a high yield with such a simple one-step process. Dithiafulvalene (DTF) units are chosen in the polymer because they have large conjugated planar surfaces to form a strong interaction with the SWNT walls as well as the possible interaction through charge transfer. By choosing different number of thiophene units (m) in the polymer, the polymer conformation and rigidity can be tuned. This will affect the capabilities of each polymer to effectively bind to the wall of the SWNT and hence influence the yield of sorting. In this work, we investigated the dispersion and sorting effects by systematically varying the number of thiophene repeating units in the polymer backbone and fabricated high on/off ratio thin film transistors with the sorted arc-discharged sc-swnts. Figure 7. (a) Chemical structure of poly(dithiafulvalene-fluorene-co-m-thiophene) (b) A picture of a SWNT dispersion obtained from pdtff-3t polymer in toluene (c) AFM image of pdtff-3t dispersed large-diameter SWNTs network and a height line profile to show that SWNTs are all individualized. (d) Height distribution of SWNTs dispersed pdtff-3t. 3.1 UV-Vis-NIR characterization of the large diameter SWNTs Ultraviolet-Visible-Near Infrared (UV-Vis-NIR) absorption spectroscopy was used to measure the amount of SWNTs dispersed and to estimate the yield of the dispersion process. The polymer-sorted SWNT solutions showed well-resolved peaks in contrast to the broad peak observed for unsorted SWNTs 2014 GCEP Report - External 16

17 in NMP. This indicates that SWNTs were well dispersed without large bundles, typically observed as broadened spectrum features. Since the same polymer to SWNT weight ratio and the same amount of solvent were used for preparing the dispersions, the higher absorption peak intensity for pdtff-1t indicates that it exhibits the highest dispersion efficiency for SWNTs. By comparing the areas of the second inter-band transition of semiconducting SWNT (S22) peak between the N-methyl-2-pyrrolidone (NMP) dispersed SWNTs (no sorting, all SWNTs dispersed) and pdtff-mt sorted SWNTs, the yields for the semiconducting SWNTs were estimated as 44%, 32% and 28% for pdtff-1t, pdtff-2t and pdtff-3t respectively. The higher dispersion efficiency for pdtff-1t is attributed to the larger ratio of DTF derivative to thiophene units in the pdtff-1t, which results in stronger π-π interactions with SWNTs, as well as the greater ratio of alkyl side chains, which could also stabilize the polymer/swnt interactions by C H π interactions. UV-Vis-NIR spectra can also be used to determine the types (metallic or semiconducting) of SWNT dispersed by pdtff-mt polymers. For comparison, the semiconducting (S22) peak intensity for unsorted SWNTs dispersed in NMP was normalized to the same intensity as the sorted peak (Figure 8a). Compared to the unsorted AD-SWNTs dispersed in NMP, there was a significant decrease in the first inter-band transition of metallic SWNT (M11) peaks for pdtff-mt polymer dispersed SWNTs. The M11 peaks for pdtff-2t and pdtff-3t are more significantly suppressed than for pdtff-1t, indicating more selective wrapping of semiconducting SWNTs by these polymers. Since UV-Vis-NIR spectrum shows the absorption from all metallic and semiconducting tubes, comparing the areas under the metallic (M11) and semiconducting peaks (S22) after baseline substraction, the percentage of semiconducting SWNTs in the sorted tubes can be estimated to be 94% for pdtff-1t and 96% for both pdtff-2t and pdtff-3t. In order to reduce the interference of the polymer peak in the UV-Vis-NIR spectra, the polymers were removed by annealing at 500 C for 1h in an Ar atmosphere After normalizing to semiconducting SWNTs (S22) peak, we observed much greater reduction in the M11 peak for the polymer sorted SWNTs, again indicating an efficient removal of the metallic tubes. The polymer-sorted SWNTs dispersion is stable for at least 6 months without visible aggregation or significant reduction/broadening in UV-Vis-NIR absorption (Figure 8b) GCEP Report - External 17

18 Figure 8. (a) Optical absorption spectrum of carbon nanotubes before and after sorting with pdtff-mt. The unsorted AD-SWNTs were dispersed in N-methyl-2-pyrrolidone (NMP). (b) Absorption spectrums of the SWNTs dispersed by pdtff-3t as-prepared and after 6 months. Molecular Dynamics (MD) Simulations To investigate the wrapping geometry of the polymer-swnt complexes and further understand the selectivity achieved with different polymers, MD simulations of the wrapping process were performed in implicit toluene solvent (ε = 2.84). Two representative SWNTs observed by Raman spectroscopy, the metallic (11,8)-SWNT and the semiconducting (16,6)-SWNT, and three polymers (pdtff-mt, m = 1-3) were used in the modeling. During the very initial stages of the 100 ps simulation, the polymer backbones gradually evolved from a linear arrangement in the initial structure to a more helical wrapping around the circumference of the tube. Eventually, after approximately 200 ps, the polymer backbone subsequently extended to a more stable quasi-linear conformation and remained in contact with the nanotube surface (snapshots in Fig. 9a,b). This final linear conformation is reproduced regardless of the starting orientation of the polymer with respect to the SWNT. Most alkyl side chains also remained bound to the nanotube surface during the 2014 GCEP Report - External 18

19 100 ns simulations. Snapshots of the structures of the polymers interaction with metallic and semiconducting SWNTs obtained after 60 ns simulation are shown in Figure 9a and 9b, respectively. Figure 9. Representative snapshots of the MD simulation for (a) (11,8) m-swnt and (b) (16,6) sc-swnt with pdtff-1t, pdtff-2t, and pdtff-3t. Calculated absolute (c) and normalized (d) binding energies (E B ) as a function of the contact surface area (A 2 ) between the pdtff-mt polymers and both (11,8) m- and (16,6) sc-swnts. The computed binding energies and surface areas of polymers without alkyl side chains (labeled as w/o sc) are also shown to illustrate the influence of alkyl side chains. Contact surface areas were averaged along the last 20 ns of 100 ns MD trajectories and were calculated using UCSF Chimera 1.6. We computed the normalized binding energy (E B ) of the SWNT/pDTFF-mT complex as E B = [E complex (E SWNT + E pdtff-mt )/l pdtff-mt, where E complex, E SWNT, and E pdtff-mt are the energies of the SWNT/pDTFFmT complex, the nanotube, and the polymer, respectively, based on the implicit generalized Born (GB) solvation model, and l pdtff-mt is the length of the polymer in an idealized extended conformation (in Å). The binding energies of our polymers with metallic and semiconducting SWNTs are shown in Table 1. Interestingly all three polymers exhibited stronger binding to the semiconducting nanotube than to the metallic one. This is attributed to the more planar surface of the larger-diameter (16,6) semiconducting nanotube, which interacts more preferably with the polymer backbone and side chains. Within the last 20 ns of simulation, when the systems are completely equilibrated, the average normalized binding energies 2014 GCEP Report - External 19

20 of pdtff-1t with both (16,6) sc-swnt and (11,8) m-swnt are the greatest among the three polymers (-5.3 and -5.0 kcal/mol, respectively), and the binding energies of pdtff-2t (-4.4 and -4.2 kcal/mol) are slightly larger than those of pdtff-3t (-4.2 and -3.9 kcal/mol), in agreement with the measured amounts of SWNTs dispersed. A possible explanation is that the greater number of side chains per length unit in pdtff-1t leads to higher surface contact area for the same length of polymer and thus gives rise to higher binding energies with the nanotubes. The computed binding energies are in good linear relationship with the surface contact area (Figure 9c and 9d). On the other hand, the difference in computed selectivity of different polymer is rather small (0.2~0.3 kcal/mol/å), even though in qualitative agreement with the trend observed experimentally. Interestingly, when the same MD simulations were performed using only the polymer backbones without the side chains, i.e. the alkyl side chains were replaced with H atoms, the binding energies between the three polymers with the semiconducting and metallic SWNTs were all similar and the computed selectivities between semiconducting and metallic SWNTs were diminished (Table 1b). This suggests that the side chains promote polymer/swnt binding and can also contribute to the observed selectivity. Table 1. Average binding energies of polymers pdtff-mt (a) with side chain and (b) without side chains with (11,8) and (16,6)-SWNTs. Average binding energies computed from the last 20 ns of simulation. Absolute energies are in kcal/mol and normalized energies are in kcal/mol/å. (a) Length (Å) (11,8) m-swnt Absolute binding energies (16,6) sc-swnt ΔE B Normalized binding energies (11,8) m-swnt (16,6) sc-swnt pdtff-1t pdtff-2t pdtff-3t ΔE B (b) pdtff-1t (w/o sc) pdtff-2t (w/o sc) pdtff-3t (w/o sc) Length (Å) (11,8) m-swnt Absolute binding energies (16,6) sc-swnt ΔE B Normalized binding energies (11,8) m-swnt (16,6) sc-swnt ΔE B 2014 GCEP Report - External 20

21 3.2 Ongoing Work on SWNTs Material Processing We are pursuing efforts in sorting various diameter carbon nanotubes employing different solvents to compare selectivity, dispersion yield and influence on carbon solar cell performance. 4 Summary We summarize our work from the past year and our ongoing efforts towards an improved carbon solar cell as follows: 1. We have demonstrated the enrichment of smaller-diameter semiconducting SWNTs from CoMoCAT SWNTs using simple sonication and centrifugation processes. We observed that the yield of dispersion increased upon the lengthening of polymer s alkyl side chain. Molecular dynamics simulations indicated that the stronger interaction is due to the higher surface contact area between the longer alkyl side chain of the polymers and the SWNT. We also demonstrated that the sorted, highly concentrated smaller-diameter wider bandgap CoMoCAT SWNTs have great potential in solar cell applications, due to their improved open-circuit voltage and their improved near-infrared carrier extraction over the sorted HiPco SWNTs. Finally, thin film transistors fabricated with our sorted CoMoCAT SWNTs demonstrated excellent uniformity, mobility and on/off ratio. The high charge carrier mobility, together with the improved charge extraction efficiency, indicates that it would be highly advantageous to employ polymer sorted small-diameter SWNTs for photovoltaic applications. a. Our ongoing efforts focus on improving film quality and carbon based electrodes to enable higher performing photovoltaic performance. 2. We synthesized a new series of polymers (pdtff-mt) that selectively dispersed large-diameter arc-discharged semiconducting SWNTs in high yields by a simple sonication and centrifugation process. We found that the amounts of SWNTs dispersed were proportional to the available contact areas from the polymers and the increased polymer flexibility tends to improve 2014 GCEP Report - External 21

22 selectivity. Spectroscopic characterizations by UV-Vis-NIR confirms the selectivity for semiconducting SWNTs. The sorted, concentrated and stable large-diameter semiconducting SWNT solutions have great potential for applications as the semiconducting active layer in solar cells enabling absorption in the infrared which remains uncaptured by traditional photovoltaic materials. a. Our ongoing efforts in this domain remains to improve SWNTs sorting in terms of selectivity and yield by comparing the effects of employing various solvents during the sorting process. Published Work Wang, H., Mei, J., Liu, P., Schmidt, K., Jiménez-Osés, G., Osuna, S., Fang, L., Tassone, C.J., Zoombelt, A.P., Sokolov, A.N., et al. (2013). Scalable and Selective Dispersion of Semiconducting Arc-Discharged Carbon Nanotubes by Dithiafulvalene/Thiophene Copolymers for Thin Film Transistors. ACS Nano 7, Wang, H., Koleilat, G.I., Liu, P., Jiménez-Osés, G., Lai, Y.-C., Vosgueritchian, M., Fang, Y., Park, S., Houk, K.N., and Bao, Z. (2014). High-Yield Sorting of Small-Diameter Carbon Nanotubes for Solar Cells and Transistors. ACS Nano. Contacts Zhenan Bao Ghada Koleilat 2014 GCEP Report - External 22