GCEP award #40654: High-Efficiency, Low-Cost Thin Film Solar Cells

Investigators Alberto Salleo, Assistant Professor, Materials Science and Engineering; Yi Cui,

Assistant Professor, Materials Science and Engineering; Peter Peumans, Assistant Professor, Materials Science and Engineering, Ludwig Goris, Post-doctoral Researcher, Stanford University; Rodrigo Noriega, Stephen Connor, Ching-Mei Hsu, Ben Weil, Graduate Researchers, Stanford University.

Abstract Several key advances were accomplished within this project in the last year. These

advances are subdivided in materials development, device development and modeling and simulation. Highly doped ZnO nanowires were synthesized and characterized. The goal is to use them as a solution-processable alternative to ITO for transparent electrodes. Appropriate characterization techniques for these nanomaterials were developed this year. We were able to independently determine carrier mobility and charge density in the nanowires. The most highly-doped material has a resistivity of the order of a few mΩ.cm. We found that the concentration of dopant atoms exceeds the carrier density by up to one order of magnitude suggesting that at high dopant density a compensating impurity sequesters the free charge. Simultaneously, Ag nanowire mats are being developed for the same purpose. A plating process was explored to improve the sheet resistance. Ag nanowire electrodes were demonstrated on flexible substrates owing to their low processing temperature. The last set of materials under development is a family of doped Cu-In-S nanowires, which will constitute the active layer of the cells. Initial devices were made with Ag nanowire electrodes. Single-junction organic cells performed as well as the ITO control cells. Simple tandem unconstrained cell stacks were fabricated as well. The two cells operated independently, as predicted, confirming the advantage of the unconstrained geometry. Finally, optical, electrical and device simulation tools were developed. An optical simulation program was written to rigorously solve for the trasmission through Ag nanowire meshes as a function of nanowire density. Furthermore, a separate program was developed to calculate the sheet resistance of Ag nanowire meshes with random geometry as a function of wire density. The latter program is being adapted to work with ZnO nanowires. Finally, a full device simulation tool has been developed to estimate what materials or cell architecture parameters limit performance. This program helps quantify the advantage of the unconstrained cell architecture compared to conventional current-matched architectures. A particularly attractive feature of the unconstrained architecture is its relative robustness with respect to spectrum (i.e. time of the day).

Introduction and Background The goal of the project is to develop materials and processing techniques that will

allow the fabrication of low-cost multi-junction solar cells entirely by solution-processing or other low-cost techniques amenable to roll-to-roll (R2R) fabrication such as lamination. Device modeling will help identify the best device architecture. A specific device design goal consists of developing an unconstrained multi-terminal multi-junction photovoltaic cell.

We propose to take advantage of nanostructured materials to fabricate high-efficiency multi-junction cells using solution-based processing techniques. As a result, our cells will be amenable to large-area fabrication, thereby dramatically reducing their cost. Finally, our materials will allow to operate the individual cells independently, rather than in series, thereby bypassing current matching requirements.

Results Materials Development ZnO nanowires

Figure 1: High-resolution TEM micrographs of ZnO nanowires doped with Al (a,b) and Ga (c,d). The wires shown in b and d were grown in the presence of a surfactant (oleic acid).

Degeneratively doped ZnO nanowires will be used as solution-processable

transparent conductors. ZnO is known to form nanostructures of varied shapes. A synthesis in non-aqueous solvent was adapted. ZnO nanowires were obtained by decomposing Zn acetate in tri-octylamine at 365°C. These wires can be doped with with Al or Ga during synthesis by introducing Ga nitrate and Al acetate in the synthesis bath. In order to study dopant incorporation, the nanowires are characterized by scanning electron microscopy (SEM), scanning Auger spectroscopy, x-ray diffraction and inductively-coupled plasma atomic emission spectroscopy (ICP-AES). The typical range of dopant concentration used is 1 to 5 atomic % in solution. ICP-AES allow to determine the Al:Zn and Ga:Zn ratio in the wires. It is found that the dopant incorporation rate in the nanowires is approximately 50%. As a result, extremely high Al and Ga

concentrations are obtained (up to 2x1021 cm-3), which is remarkable given the low temperatures involved in the synthesis. Within the range of concentration we investigated, no phase separation is observed by x-ray diffraction. The formation of amorphous phases however cannot be excluded. In fact, in the absence of a surfactant during the synthesis process, an amorphous shell is observed by high-resolution TEM around Al-doped ZnO nanowires (Fig. 1a). The surfactant stabilizes the growth of Ga-doped ZnO nanowires as well (Fig. 1d). Scanning Auger electron spectroscopy confirms that the dopants are distributed evenly within the nanowires.

Figure 2: Electron mobility as a function of Al% in the nanowires (left) and carrier density in the nanowires as a function of Al% (right) obtained from optical absorption measurements.

It is important to determine the amount of free charge generated by the dopant atoms.

Indeed, compensating defects can substantially lower the free-carrier concentration expected from chemical analysis. The carrier concentration and mobility in semiconducting thin films are typically obtained using Hall effect measurements, whose application is problematic for extracting electrical characteristics from quasi-one-dimensional structures such as nanowires. As an alternative, single-wire field effect transistors (SWFET) have been used to measure carrier mobility. In this technique, only a single device (i.e. a single wire) can be measured at a time, making the collection of data from a set of nanowires a labor-intensive and time-consuming process. This technique is also hindered by additional issues associated with the deposition of electrical contacts on nanomaterials, including the effects of contact resistance and the alteration of the material’s properties (such as unintentional doping and/or damage caused by the ion or electron beam) that must be taken into account during measurement interpretation. The electrical properties of nanowire ensembles can be extracted from optical absorption measurements by applying a Drude model, provided the carrier density is high enough. Because nanowire films are rough and scatter light effectively, optical absorption measurements are nearly impossible. We used photothermal deflection spectroscopy

(PDS) to carry out such measurements. Single-wire field-effect transistors were fabricated by dispersing AZO nanowires onto heavily doped, thermally oxidized silicon substrates (with the substrate serving as the gate and the 200 nm thick SiO2 as the gate dielectric) to confirm our optical measurements. So far, we have applied this technique to Al-doped ZnO nanowires prior to thermal anneal (Fig. 2). We found that the amount of free charge is independent of Al concentration (for Al:Zn ratios higher than 0.01) around Ne~1020 cm-3, which suggests the existence of compensating acceptors in ZnO. Furthermore, the carrier mobility remain high at high dopant concentration (µ~10 cm2/V.s).

Figure 3: Summary of the electrical properties of ZnO nanowire cast films.

Films of doped nanowires were cast and thermally treated (Fig. 3). Films made with

undoped ZnO are virtually insulating, as expected for an intrinsic wide-bandgap semiconductor at room temperature. In the Al-doped materials family, only the films made with nanowires grown in a 1 at.% Al synthesis solution have a lower sheet resistance than that of the annealed undoped ZnO nanowires. The effect is particularly significant in the wires synthesized in the presence of oleic acid, which have a sheet resistance over three orders of magnitude lower than that of undoped wires. In the films made from wires synthesized in a 5 at. % Al solution, the sheet resistance prior to annealing is equal to or higher than that of undoped films. Furthermore, Al-doped wires synthesized in the presence of oleic acid always outperform those synthesized without oleic acid. The presence of the amorphous shell around the nanowires in the latter case is the likely cause of this observation. Moreover, the morphology of wires synthesized in a 5 at. % Al solution is more irregular than that of wires synthesized in a 1 at. % Al

solution. Thus, even though more Al may be incorporated in the wires synthesized in a 5 at. % Al solution, the sheet resistance of such a nanowire film is higher. Thermal annealing always decreases the sheet resistance of the nanowire films. This effect is large, with resistance decreases exceeding five orders of magnitude in two cases. Within the Al-doped materials, the lowest sheet resistance was measured in films made with wires synthesized in a 1 at. % Al solution in the presence of oleic acid (~10 kΩ/square after annealing). While this value is still two-to-three orders of magnitudes too high to be competitive with current technologies, it is encouraging since no materials or film optimization was undertaken here. Ga behaves quite differently from Al. All films made with Ga-doped nanowires exhibit significant sheet resistance decreases compared to the undoped materials even before annealing. In the case of the film made with wires synthesized in a 5 at. % Ga solution without oleic acid, this effect is particularly large (6 orders of magnitude). Thus, the electrical activation during synthesis is more efficient with Ga than with Al. Ga may be more easily incorporated in the ZnO lattice than Al because its atomic radius (1.3Å) is very similar to that of Zn (1.35Å) while Al is a substantially smaller atom (1.25Å). Contrary to the case of Al, Ga-doped nanowires synthesized without oleic acid systematically produce films with a lower sheet resistance than those made with nanowires synthesized in the presence of oleic acid by a factor of 20 to 100. Since in the case of Ga-doping, the crystallinity of the wires was not affected by the oleic acid, the increased resistance may be due to the presence of adsorbed insulating oleic acid molecules along the nanowires. This hypothesis is corroborated by the observation that after annealing at 600°C, which burns off the oleic acid, the sheet resistances of both types of films become similar. Within the Ga-doped materials, the lowest sheet resistance was measured in films made with wires synthesized in a 5 at.% Ga solution without oleic acid (~1 kΩ/square, one order of magnitude better than the best Al-doped films), eight orders of magnitude more conductive than the initial undoped films.

Ag nanowire meshes

Figure 4: Ag nanowire film on kapton bent to a radius of ~1.4 cm with a sheet resistance of the order of 10 Ω/square (left). Sheet resistance vs. transmission of plated and annealed Ag nanowire films (right).

We have developed a low-cost method to fabricate transparent, nanostructured metal electrodes by casting suspensions of solution-synthesized metal nanowires. To fabricate transparent electrodes using nanowire suspensions, a volume of the suspension was dropped on a substrate with 100nm-thick prepatterned Ag contact and was allowed to dry in for 15 minutes. Annealing of the meshes at a temperature of 180 °C results in a steep drop of the sheet resistance. Ag nanowires were also deposited on flexible polyimide substrates reaching sheet resistances of the order of 10 Ω/square with bending radii smaller than 1 cm (Fig. 4, left). Electroless plated and annealed Ag nanowire meshes were fabricated in an attempt to reduce the wire-to-wire contact resistance (Fig. 4, right). Sheet resistances as low as 5 Ω/square were obtained at the expense of a slight drop in transmission (T~60%).

Active material: CuInS2 CuInS2 nanostructures were synthesized by colloidal chemistry. In order to control

the bandgap, the material is doped with Ag, Fe, Zn and Ga. The dopant percentage was varied between 5 and 20%. Films of these materials can be cast but the characterization of their optical properties is still incomplete. Optical charaterization is challenging since absorption and scattering must be separated.

Process Development A crucial aspect of the work proposed is the fabrication of an unconstrained

photovoltaic cell using the materials developed here, and the transparent conducting materials in particular. As a starting point, a single-junction organic solar cell was fabricated with a Ag nanowire electrode (Fig. 5). The performance of the cell was comparable to that of a control cell made with the same active materials and an ITO electrode.

Figure 5: Organic solar cell made with a transparent Ag nanowire electrode (left). I-V characteristics of the cell compared to an ITO standard (right).

Furthermore, a lamination process was developed to be able to “stamp” Ag nanowire films on delicate substrates, such as organic thin films. This process allowed the fabrication of an organic solar cell on flexible metal foil. The same process will be used to deposit intermediate electrode layers in multijunction solar cells as it can be combined with a flattening process to reduce surface roughness. Hence, a prototype

semitransparent cell (with both layers being identical) was fabricated using an Ag nanowire intermediate electrode (Fig. 6). As expected, the top and bottom cell performed independently thereby providing a very encouraging proof-of-principle of the unconstrained cell design.

Figure 6: Schematic of a semitransparent unconstrained tandem cell (left) and I-V characteristics of the two sub-cells (right).

Because they are very dense networks, in-plane alignment of the ZnO nanowires may be important. To this effect, a rubbing technique was used to induce controlled in-plane alignment of the ZnO nanowires and study its effect quantitatively. While alignment makes a difference of a factor of approximately 3 or 4, the result of this experiment is still being quantitatively analyzed with the help of modeling tools.

Modeling and simulation A substantial component of this project involves the use of computer simulations and

device modeling.

Modeling of Ag nanowire meshes The Ag nanowire mesh electrodes are the result of a tradeoff: a higher connectivity of

the network leads to lower sheet resistance but also a lower transparency. In order to understand this tradeoff, two simulation tools were developed. A rigorous coupled-wave analysis (RCWA) simulator allows to predict the transmitted intensity as a function of nanowire density. An electrical network simulation tool allows to calculate the sheet resistance of a network of nanowires having the topology of the network (density and connectivity) and the resistance of each wire as well as the contact resistance between the wires as inputs. Both models agreed well with experimental data (Fig. 7). In fact, the electrical network simulation tool was used to determine that the interwire contact resistance limits the performance of the Ag nanowire electrodes, especially in the low-density regime. This simulation tool is currently being adapted to the ZnO nanowire electrodes.

Figure 7: Geometry of the nanowire simulation (a). Transmissivity calculated as a function of nanowire density compared with experiments (b). Sheet resistance calculated as a function of nanowire density, with three values of the wire-to-wire contact resistance as a parameter, compared with experimental results (c).

Device simulation In order to design correctly unconstrained multijunction cells, predict their

performance and compare to conventional multijunction cells, a device simulation tool is being developed. To study trends and quantify the advantage of the unconstrained architecture, well-known materials are currently being used in the simulations (the optical and electrical properties of the semiconductor nanowires and nanoparticles are unknown). These initial simulations confirm the advantage of an unconstrained 2-cell architecture vs. a series-connected 3-cell architecture, where 2 of the 3 cells of the 3-cell architecture are made of the same materials as the 2-cell architecture. In particular, the unconstrained cell is robust with respect to spectral changes occurring during the day.

Progress The proposed technology is game-changing inasmuch as the introduction of cheap,

highly efficient solar-cells would allow to eliminate the biggest obstacle to the widespread adoption of solar energy conversion, namely the high cost per unit energy produced (currently $0.27/kW.hr for crystalline Si solar cells vs. $0.06/kW.hr for grid power). As such, this project if successful could have a very substantial impact in the reduction of greenhouse gases. The proposed program hinges on two key technologies: the development of solution-processed highly-conducting transparent electrodes and the development of solution-processed active layers. Not only will the use of solution-dispersed materials enable the use of low-cost manufacturing technologies borrowed from the printing industry. It may also allow us to take advantage of phenomena typical of the liquid phase, such as capillarity, immiscibility and preferential adsorption, to drive the self-assembly of parts of the devices. In this context, this year’s activities are well in line with the desired progress direction. One material for transparent electrodes is now well-developed and integrated in devices (Ag nanowires) while the other is still in the development phase but has experienced substantial improvements in the last year (over 8 orders of magnitude conductivity

increase). It will therefore be integrated in devices in the next few months. Progress in the synthesis and characterization of the active layer is also been made. On the device side, the fabrication of tandem unconstrained cells is particularly encouraging as this technology promises to produce a substantial benefit in terms of W/$. The simulation tools that have been developed and tested in the last year, finally, are of great help in planning experiments and designing devices.

Future plans Degeneratively doped ZnO nanowires

• Understand the defect chemistry of doped ZnO • Increase carrier density in doped ZnO nanowires by chemical or thermal

treatments • Explore film-forming processes • Study the relationship between film conductivity and wire connectivity • Study interwire contact resistance

Ag nanowire meshes • Develop plating and annealing processes • Tuning contact resistance • Explore composites with ZnO nanowires

Semiconductor nanocrystals

• Characterize optical and electronic properties of materials • Study effect of dopants on bandgap • Study sintering behavior

Device fabrication and characterization

• Fabricate multi-junction cell using lamination, solution processing or a combination

• Further develop simulation codes for unconstrained cells • Optimize device design

Publications

1. S. T. Connor, C.-M. Hsu, B.Weil, Y. Cui, “ Mechanistic Studies of CuInS2 Nanocrystal Growth”, J. Am. Chem. Soc. (in press).

2. L. Goris, R. Noriega, M. Donovan, J. Jokisaari, G. Kusinski and A. Salleo, “Intrinsic and doped zinc oxide nanowires for transparent electrode fabrication via low temperature solution synthesis”, J. of Electronic Materials 38, 586 (2009).

3. Rodrigo Noriega, Jonathan Rivnay, Ludwig Goris, Daniel Kälblein, Hagen Klauk, Klaus Kern, Linda Thompson, Aaron Palke, Jonathan Stebbins, Alberto Salleo, “Probing the electrical properties of highly-doped Al:ZnO single nanowires”, submitted to Nanoletters.

Contacts

Alberto Salleo: asalleo@stanford.edu Yi Cui: yicui@stanford.edu Peter Peumans: ppeumans@stanford.edu