LETTER pubs.acs.org/NanoLett
Solution-Processed Sintered Nanocrystal Solar Cells via Layer-by-Layer Assembly Jacek Jasieniak,*,†,§ Brandon I. MacDonald,‡,†,§ Scott E. Watkins,† and Paul Mulvaney*,‡ † ‡
CSIRO, Materials Science and Engineering, Bayview Avenue, Clayton, Victoria, 3168, Australia School of Chemistry and Bio21 Institute, The University of Melbourne, Parkville, Victoria, 3010, Australia
bS Supporting Information ABSTRACT: Solar cells made by high temperature and vacuum processes from inorganic semiconductors are at a perceived cost disadvantage when compared with solution-processed systems such as organic and dye-sensitized solar cells. We demonstrate that totally solution processable solar cells can be fabricated from inorganic nanocrystal inks in air at temperature as low as 300 °C. Focusing on a CdTe/ZnO thin-film system, we report solar cells that achieve power conversion efficiencies of 6.9% with greater than 90% internal quantum efficiency. In our approach, nanocrystals are deposited from solution in a layer-by-layer process. Chemical and thermal treatments between layers induce large scale grain formation, turning the 4 nm CdTe particles into pinhole-free films with an optimized average crystallite size of ∼70 nm. Through capacitance voltage measurements we demonstrate that the CdTe layer is fully depleted which enables the charge carrier collection to be maximized. KEYWORDS: Solar cell, nanocrystal, CdTe, sintered, layer-by-layer, solution processed
T
he development of cheap, efficient, and economically viable solar energy technologies is an important goal for managing the growing global energy demand while reducing greenhouse gas emissions.1 Thin-film solar cell technologies may allow these targets to be met through lower material consumption and faster deposition rates offered by roll-to-roll processing.2 To achieve this goal, materials must be developed which can not only efficiently collect solar radiation but also convert it to electricity, exhibit thermal stability and photostability, and be solution processable.3 Satisfying all these conditions has to date hindered the realization of any true, solution-processed photovoltaic (PV) systems that can compete with existing silicon technologies. Inorganic semiconductors are ideal for thin-film solar cells because they are thermally stable and photostable, are spectrally broad absorbers, and have excellent electronic properties compared to their organic counterparts. Unfortunately, one of their major disadvantages is that they are difficult to deposit directly from solution. In an attempt to overcome this limitation, the development of inks based on (i) metallic4 or semiconducting colloid particles5 7 and (ii) metal complexes8 or oligomers9,10 have been proposed. Postdeposition chemical and/or thermal treatment steps are required in both approaches to ensure that the resulting polycrystalline, inorganic thin films possess suitable light-absorption and electronic properties for developing efficient PV technologies. In this work we specifically focus on using inks composed of 1 10 nm sized inorganic semiconducting particles to fabricate efficient, solution processed solar cells. Advantageously, colloidal nanocrystals in this size range are easily synthesized, their surface chemistry can be readily modified, and they can be deposited as r 2011 American Chemical Society
thin films from solution. However, the increased exciton binding energy of small nanocrystals11 and the inherent dispersion in the charge conduction level caused by the ensemble polydispersity,12 do create potential drawbacks to their use in photovoltaic applications. Gur and co-workers were among the first to exploit the advantageous processing properties of semiconductor nanoparticles by employing additional chemical and thermal treatment steps.5 This allowed for the fabrication of bulk polycrystalline CdTe/CdSe sintered, inorganic solar cells from colloidal CdTe and CdSe nanorods with power conversion efficiencies as high as 2.9%.13 More recently, Olson and co-workers adapted this approach to show that colloidal CdTe nanorods can been used to fabricate single layered, sintered, inorganic Schottky solar cells with efficiencies as high as 5%.14 Our attempts to generalize this method to the more industrially friendly and synthetically generic spherical nanocrystallites have consistently failed. A key reason for this is the increased stresses that develop during the thermal treatment step within a film of spherical nanocrystallites as compared to a film of nanorods. The large scale grain growth which occurs in this step causes significant defect formation in the thin films, which then act to short-circuit any devices that are fabricated.15 By adopting a simple layer-by-layer approach, this problem can be overcome to produce highly reproducible and microscopically defect-free films of sintered nanocrystals (Figure 1). Received: April 16, 2011 Revised: May 18, 2011 Published: May 27, 2011 2856
dx.doi.org/10.1021/nl201282v | Nano Lett. 2011, 11, 2856–2864
Nano Letters
Figure 1. A general schematic of the layer-by-layer process for developing sintered inorganic thin-film solar cells from colloidal nanocrystals. (A) As synthesized nanoparticles are purified and their surface chemistry is modified to ensure dispersion in a solvent which is compatible with multilayer deposition, examples of which include pyridine, alcohols, and water. (B) The surface functionalized nanocrystals are then deposited as a thin film onto a substrate by any suitable deposition method, in our case spin-coating. (C) The thin film is then exposed to first a chemical treatment (e.g., CdCl2 for CdTe) and then, if desired, a thermal treatment. The chemical treatment step is necessary to ensure that the surface chemistry of the nanoparticles is modified. In the thermal annealing step which follows, elevated temperatures cause crystal growth and sintering between nanoparticles. Both of these effects enhance the optical and electronic properties of the film for solar cell applications. The deposition of nanocrystalline thin film and the interlayer chemical and thermal treatment steps are repeated to achieve a desired device architecture. (D) A scanning electron microscope image of a typical polycrystalline CdTe/ZnO solar cell fabricated using the layer-by-layer approach. Aluminum was deposited by evaporation. (E) The electronic levels with respect to vacuum (Evac = 0 eV) of all components associated with the CdTe/ZnO devices explored in this work. Dashed lines indicate approximate values of the Fermi level in the given materials.
This is achieved through the use of interlayer chemical and thermal treatment steps, which alleviate stresses and effectively fill the residual defects that are formed within subsequent layers. To demonstrate the effectiveness of this approach, we have chosen to study the development of solution processed cadmium telluride (CdTe)/zinc oxide (ZnO) thin-film solar cells (Figure 1D). CdTe was selected because it has an almost ideal bulk optical band gap for solar energy conversion of 1.45 eV, it possesses a high absorption coefficient, and it can be intrinsically doped p-type.16 Meanwhile, ZnO is readily made n-type following UV exposure and possesses the right energy level configuration so that when paired with CdTe, it produces an abrupt p n heterojunction (Figure 1E).17 Used within highly optimized, non-solution processed solar cell configurations, record laboratory power conversion efficiencies (PCE) of up to 16.7% have to date been achieved with CdTe based heterojunctions.18 This suggests that CdTe is an ideal system from which to identify the generic requirements that need to be satisfied in order to develop highly efficient, solution-processed inorganic solar cells. As a manufacturing benchmark, commercially available CdTe modules with power conversion efficiencies of over 12% and manufacturing costs 60% at wavelengths less than 700 nm. For this reason, optimal device performance was seen for films with CdTe thicknesses above 250 nm, where power conversion efficiencies >6% were obtained. Considering the simple bilayer nature of these devices, further optimization of device architecture is envisaged to increase efficiencies to values approaching the 9.4% which has been achieved for highly optimized, sputtered multilayered CdTe cells of a similar thickness.29 To understand the efficiency of the photocurrent generation process in solar cells, it is vital to gauge how well absorbed photons are converted to free carriers and then collected. This figure of merit is defined as the internal quantum efficiency (IQE). Conventional reflectance and transmittance measurements are typically employed to determine the wavelengthdependent absorptance of the active layer. By integrating the product of the absorptance values and the equivalent current density of the AM1.5 spectrum, the maximum photocurrent under standard conditions is easily determined. Because these measurements implicitly include parasitic contributions, such as interference and absorption by adjunct layers (including the substrate and electrodes), when performed on multilayered thinfilm solar cells, they result in an overestimation of the true active layer absorptance.30 In order to remove these parasitic contributions, it is useful to model the system using a scattering matrix method. Using this approach, we have modeled our entire multilayered device structure and explicitly calculated the wavelength dependence of all the parasitic contributions (see Supporting Information for details).31 To confirm the validity of the modeling, we compared the calculated reflectance spectra of completed devices to that determined experimentally (Figure 3C). The excellent agreement between theory and experiment enabled a good approximation to the true CdTe absorptance within our device structure to be extracted. A
comparison of the experimentally determined absorptance of our entire solar cells to that of the extracted CdTe layer shows that the standard reflection and transmission measurement only slightly overestimate the true absorptance at wavelengths below that of the CdTe optical band gap (∼845 nm); at higher wavelengths this overestimation significantly worsens. By assuming that 100% of the absorbed photons in the CdTe layer resulted in free-carrier generation and were then collected, the maximum short-circuit current of the device was calculated. A comparison of the experimentally determined Jsc values as a function of CdTe thickness to the expected Jsc values calculated using the scattering matrix model is shown in Figure 3D. The error bars represent the 2.5% variation in measured device area from the nominal value of 0.205 cm2. Good agreement between the experimental and predicted values is observed across the entire thickness range examined. As the model assumes 100% of the absorbed light within the CdTe layer in the device is converted to photocurrent, we are able to determine the internal quantum efficiency (IQE) of our devices by IQE(%) = Jsc,exp/ Jsc,calc 100%. For all thicknesses examined, our layer-by-layer cells exhibit IQEs of greater than 80%, with many devices possessing an IQE of >90%. This implies that under short-circuit conditions nearly all photons absorbed within the device are collected as photocurrent. Depleted Heterojunction Mechanism. Highly efficient polycrystalline thin-film solar cells are generally fabricated with either p n or p i n device configurations.32 In the latter, the intrinsic (i) region is particularly important for ensuring that the built-in field decreases homogeneously across the layer. This is vital for effectively collecting photocarriers from systems in which the diffusion coefficients are small.33 For our system, initial inspection of the band structure would suggest the formation of a p n junction. This would naturally arise because CdTe that is treated with CdCl2 and annealed is intrinsically p-type due to the formation of Cd vacancies.22 Meanwhile, ultraviolet irradiation of ZnO nanocrystal films leads to desorption of surface bound oxygen or carbon dioxide, which in turn creates free electron densities of approximately 1 1019 cm 3 and renders the ZnO highly n-type.17 However, as we will now show, this system operates not like a p n junction, but more like an i n junction due to full depletion of the CdTe layer in our devices. Capacitance voltage (C V) analysis is a useful approach for identifying the existence of a depletion layer stemming from a p n junction, as well as providing an estimation of the bulk doping densities and the built-in field.33 When the majority carriers in a layer are fully depleted, C V analysis yields a value of 1/C2 that is largely independent of the applied voltage. Simulations of our system through the freely available SCAPS software package demonstrate this effect for a ITO/CdTe (183 nm)/ZnO (60 nm)/Al device with a constant CdTe trap state density of 2 1014 cm 3, a constant ZnO donor density of 1 1019 cm 3, and a varying bulk CdTe acceptor density (Figure 4A).34 Full depletion in this system is observed at 0 V for bulk doping densities below ∼2.5 1016 cm 3. Experimental results agree well with the simulated trends indicating that our system is fully depleted at 0 V. The best fit yields a bulk CdTe doping density close to 1.5 1016 cm 3 (Figure 4A). In these simulations, the low trap density ensures that the main depletion mechanisms stem from the CdTe/ZnO junction and the ITO/CdTe interface. In reality, the polycrystalline nature of this system would result in large surface trap densities, which 2860
dx.doi.org/10.1021/nl201282v |Nano Lett. 2011, 11, 2856–2864
Nano Letters
LETTER
Figure 4. (A) Mott Schottky plot of ITO (125 nm)/CdTe (200 nm)/ZnO (60 nm)/Al (100 nm) solar cells. The experimental data shown here were obtained at 1 kHz and analyzed according to a Randles equivalent circuit. Simulations of the system were performed using the SCAPS 2.9.02 software package by using energy levels described within the main text, a CdTe trap density of 2 1014 cm 3, and a ZnO donor density of 1 1019 cm 3. To obtain comparable bulk capacitance values, the CdTe layer was modeled with a slightly lower thickness of 183 nm. We have found that this deviation in thickness is within the experimental error of using profilometry to measure the film thickness. Comparison of the simulated and experimental data suggests that for the entire range of CdTe thicknesses studied here, the CdTe layer was fully depleted at 0 V. (B) Schematic representations of partially and fully depleted abrupt heterojunctions based on the device architecture explored in this work. (C, D) The spatial and wavelength-dependent optical field density that has been normalized by the incoming field and the simulated exciton generation rate (cm 3 s 1) within the CdTe layer of an optimized device with 400 nm CdTe and 60 nm ZnO layers. The exciton generation rate is simulated with the standard AM1.5 100 mW/cm2 irradiance as the incoming light source.
would act to deplete each particle and lead to formation of a Schottky barrier extending from the grain boundaries.32 Under these circumstances, depletion would arise from both grain boundary and junction contributions. This additional mechanism would not, however, drastically alter the trend in the 1/C2 vs voltage plot (Figure S4, Supporting Information). Thus, while the bulk doping density may be varied from that determined above, the fully depleted heterojunction picture would still be valid. If we assume no surface state pinning, the built-in voltage of a fully depleted heterojunction is governed by the Fermi level of the photocathode and the heavily doped n-type material with an Ohmic contact (Figure 4B). The work function of cadmium chloride treated ITO, as determined by photoelectron spectroscopy in air (PESA), lies between 4.8 and 5.0 eV, depending on the exact processing conditions. Using the literature value of 4.4 eV for the electron affinity of ZnO,35 and considering the high donor doping density, the expected built-in voltage would be of the order of 0.4 0.6 V. These are in good agreement with experimentally determined open-circuit voltages that we have determined from the numerous devices tested (Figure S5, Supporting Information). It is evident that increasing the work function of the ITO photocathode beyond 5.0 eV would increase the built-in field and consequently improve device performance.
Scattering matrix modeling of the normalized optical intensity permits the determination of the spatial and wavelength dependent exciton generation rate within the device (see Supporting Information for details). The contour plots of both are shown in panels C and D of Figure 4, respectively, for a CdTe layer thickness of 400 nm. From these plots it is evident that a major decrease of the electric field in the optical frequency range occurs at smaller wavelengths due to the strong absorption in this spectral region. Assuming charge separation and collection efficiencies of unity for photogenerated carries, 50% of the photocurrent is generated within the first 100 nm of the CdTe layer (Figure S6, Supporting Information). The existence of a partially depleted heterojunction could be detrimental for attaining high photocurrents due to the nonideal nature of the ITO/ CdTe contact and the existence of a flat-band region (Figure 4B). Under idealized conditions, depletion ensures a linear gradient of the electric field across the device. This permits free carriers to be efficiently transported through both diffusion and electric field assisted drift contributions. To conclude, we have used nanocrystal inks to fabricate CdTe/ZnO solar cells with power conversion efficiencies that approach 7%. To attain such high efficiencies, a significant increase in crystallite size was necessary. This was achieved by controlling the surface chemistry of the as-deposited crystallites, which then permitted recrystallization and large grain growth at 2861
dx.doi.org/10.1021/nl201282v |Nano Lett. 2011, 11, 2856–2864
Nano Letters temperatures as low as 300 °C. A layer-by-layer approach with intervening chemical thermal treatments was found necessary to ensure pinhole-free films suitable for use in solar cells. Significantly, all layers were deposited from solution in air. Capacitance voltage analysis suggests that the CdTe layers in all devices fabricated here were fully depleted. This factor is believed to be the driving force for the near unity internal quantum efficiencies observed. These results unambiguously demonstrate that a planar heterojunction thin-film solar cell based on inorganic semiconductors can be readily fabricated with extremely high internal quantum efficiencies and that complex morphologies and device architectures are not necessary for this to be achieved. Methods. Chemicals. Cadmium oxide (CdO, Aldrich, 99.99%), tellurium powder (Aldrich, 99.9%), oleic acid (Aldrich, 90%), trioctylphosphine (TOP, Aldrich, 90%), 1-octadecene (ODE, Aldrich, 90%), tetramethylammonium hydroxide (Aldrich, 20% in MeOH), and zinc acetetate dihydrate (BDH Laboratory Supplies, 99.5%) were used in the preparations described here. All solvents were of analytical grade and purchased from Univar. All chemicals and solvents were used without further purification. Nanocrystal Synthesis. CdTe nanocrystals were prepared according to an adapted method described by Yu and Peng.36 In a typical CdTe synthesis 0.48 g of CdO, 4.24 g of oleic acid, and 60 g of octadecene (ODE) were heated under vacuum to 80 °C at which point the flask was purged with nitrogen. The solution was heated to 260 °C and maintained at this temperature until it turned clear. At this point a solution of 240 mg of Te dissolved in 5.3 mL of trioctylphosphine and 5 g of ODE was rapidly injected. The resulting CdTe nanocrystal (NC) solution was allowed to cool to room temperature naturally. ZnO NCs were synthesized in a similar manner to that reported by Wood et al.37 Briefly, in a typical synthesis 0.44 g of Zn acetate dihydrate was dissolved in 40 mL of ethanol at 60 °C. After 30 min of heating, 2 mL of tetramethylammonium hydroxide (20% in MeOH) in 10 mL of ethanol was added dropwise to the solution over 5 min. The ZnO nanoparticle solution was heated at 60 °C for 30 min to attain the ZnO nanoparticles of ∼5 nm in size. ZnO NCs dispersed in their growth solution were precipitated with hexane and centrifuged. The supernatant was discarded, and the precipitated nanoparticles were redispersed in 1-propanol at an appropriate concentration. Washing and Ligand Exchange. The as-prepared CdTe NCs were washed by twice precipitating with ethanol and redispersing in toluene. For ligand exchange the NCs were precipitated with ethanol and redispersed in pyridine. This solution was placed under nitrogen and stirred at 85 °C overnight. Pyridine capped CdTe NCs were precipitated with hexane and redispersed in a 1:1 (v/v) solution of pyridine:1-propanol at 40 mg/mL. Both CdTe and ZnO nanocrystal solutions were filtered prior to use through a Whatman Puradisc 25 GD filter. Device Fabrication. ITO coated glass (Kintec, 15 Ω/cm2) was cleaned by standing in a stirred solution of 5% (v/v) Deconex 12PA detergent at 90 °C for 20 min. The ITO was then successively sonicated for 10 min each in distilled water, acetone, and isopropanol. The substrates were then exposed to a UV ozone cleaning (at room temperature) for 10 min. The solution of CdTe nanocrystals was spin-cast onto ITO-coated glass slides at 800 rpm for 30 s. The substrates were placed on a hot plate at 150 °C for 2 min, immediately dipped into a 60 °C solution of saturated CdCl2 in methanol, rinsed gently with 1-PrOH, and finally dried under a nitrogen stream. The CdTe NCs were sintered by placing the substrates onto a hot plate at
LETTER
elevated temperatures (150 450 °C). This process was repeated a number of times until the desired thickness was achieved. Details of the optimized sintering times are given in the Supporting Information. A ZnO layer (typically 60 nm in thickness) was spin-cast on the CdTe and then heated at 300 °C for 2 min. All film deposition and annealing steps were performed in air. Aluminum contacts, 100 nm thick, were deposited via thermal evaporation through a shadow mask at pressures below 2 10 6 mbar. Using both profilometry and optical microscopy, we determined the total measured device area to be 0.21 cm2, a value that was approximately 5% larger than that obtained based on our shadow mask pattern. A connection point for the ITO electrode was made by manually scratching off a small area of the active layers. A small amount of silver paint (Silver Print II, GC electronics, Part no. 22-023) was then deposited onto all of the connection points, both ITO and Al. The completed devices were then encapsulated with glass and a UV-cured epoxy (Summers Optical, Lens Bond type J-91) by exposing to 365 nm UV light inside a glovebox (H2O and O2 levels both
