This paper reports our latest results using colloidal CuInSe2 nanocrystal inks to prepare photovoltaic (PV) devices. Thus far, devices with nanocrystal layers processed under ambient conditions with no post-deposition treatment have achieved power conversion efficiencies of up to 3.1%. Device efficiency is largely limited by charge carrier trapping in the nanocrystal layer, and the highest device efficiencies are obtained with very thin layers—less than 150 nm—absorbing only a fraction of the incident light. Devices with thicker nanocrystal layers had lower power conversion efficiency, despite the increased photon absorption, because the internal quantum efficiency of the devices decreased significantly. The thin, most efficient devices exhibited internal quantum efficiencies as high as 40%, across a wide spectrum. Mott-Schottky measurements revealed that the active region thickness in the devices is approximately 50 nm.
© 2010 OSA
Various new approaches are being examined to produce “third generation” photovoltaic devices (PVs) with very low cost and high efficiency [1–3]. There is interest in enhancing performance using nanostructured materials via new physical effects, such as multi-exciton generation (MEG)  and intermediate band absorption , as well as lowering manufacturing costs by using new materials that can be deposited without the need for high temperature and vacuum processing . Photovoltaics made of organics (OPVs) are printable, do not require high temperature processing, and have demonstrated efficiencies of 7.4% for solution processed devices . However, the most efficient devices have so far required relatively expensive specialty chemicals to achieve high efficiencies. Furthermore, achieving adequate photostability remains a major challenge with these materials. An alternative to organic materials, which can still be printed and require only moderate processing conditions, are dispersions, or inks, of inorganic nanocrystals. With this approach, inorganic materials with proven and stable performance can be deposited using mild deposition processes typically only suitable for organic materials and not crystalline semiconductors . Cd and Pb based nanocrystal PV devices [9–13] have exhibited power conversion efficiencies as high as 5.1% . However, the use of Cd and Pb is not very desirable from an environmental perspective due to their toxicity. Other more environmentally-friendly semiconductor nanocrystals, such as Cu(In1-xGax)Se2 , Cu2ZnSnS4 [16,17] or Cu2S  have also been used to fabricate PVs, but with slightly more moderate efficiencies of up to 1.6% to date. These multicomponent semiconductors are difficult and expensive to prepare through conventional vapor deposition techniques. Nanocrystal solution processing provides an inexpensive and simple processing route that could be scaled to larger production lines and transferred to flexible and light-weight plastic substrates. For commercialization, the device efficiency must be improved by about a factor of 5 to 8.
The factors that limit nanocrystal-based PV device efficiency are still relatively unexplored and to date, the highest efficiency nanocrystal PVs have been made by time-consuming empirical optimization. The devices consist of several layers of different materials and the quality of each layer and the interfaces critically impact the performance of the devices. Here, we examine the factors related to the nanocrystal layer and the semiconductor heterojunction that limit the efficiency of CuInSe2 nanocrystal-based PVs.
We previously reported the synthesis of CuInS2, CuInSe2, CuGaSe2 and Cu(In1-xGax)Se2 nanocrystals and the implementation of CuInSe2 nanocrystals into functioning PVs . The device efficiency in Ref  was only 0.24%. Since that time, we have improved the efficiency significantly, now reaching 3.1% without the need for high temperature annealing. We report these results here, and examine the factors that are currently limiting device efficiency.
2.1 Nanocrystal inks
CuInSe2 nanocrystals were synthesized by arrested precipitation using a modification of previously reported procedures . Oleylamine (OLA; >70%), tributylphosphine (TBP; 97%), copper(I) chloride (CuCl; 99.995%), indium(III) chloride (InCl3; anhydrous 99.99%) and selenium (Se; 99.99%) were purchased from Aldrich Chemical Co. and used as received.
5 mmol of CuCl and 5 mmol of InCl3 was combined with 50 ml of OLA in a three neck flask in a nitrogen-filled glovebox. The flask was sealed by a condenser-stop cock setup and two septa; the flask was removed from the glovebox and mounted on a Schlenk line. The reaction mixture was heated up to 110 °C, degassed for 10 minutes by pulling vacuum on the vessel and purged with clean nitrogen for 5 minutes. While maintaining the nitrogen environment, the reaction vessel was heated to 180 °C and 10 ml solution of 1 M Se in TBP (0.79 g Se powder and 10 ml TBP) was injected into the reaction vessel. The reaction mixture temperature was raised to 240 °C as quickly as possible and the reaction was allowed to proceed for 10 minutes. The heating mantle was removed and the reaction was cooled slowly to room temperature.
After cooling, the reaction mixture was transferred to a glass centrifuge tube. 10 ml of ethanol was added to the centrifuge tube and the mixture was centrifuged at 4500 RPM for 10 minutes. The supernatant was discarded and the solid precipitate was dissolved in 10 ml of toluene. The new solution was centrifuged at 4500 RPM for 10 minutes to remove larger and poorly capped nanocrystals. The supernatant was transferred to a new glass centrifuge tube and the solid precipitate discarded. Ethanol was added drop-wise to the particle solution until a turbid mixture was achieved. The reaction was centrifuged again at 4500 RPM for 10 minutes. The supernatant was discarded and the precipitate dissolved in toluene to achieve a 20 mg/ml solution.
2.2 Materials characterization
TEM was performed on a Phillips 208 operated at 80 kV accelerating voltage or a JEOL 2010F TEM at 200 kV accelerating voltage. TEM samples were prepared on a 200 mesh nickel grid (Electron Microscopy Sciences) by dropcasting a dilute solution of nanocrystals in chloroform. EDS was carried out using an Oxford INCA EDS detector mounted on the JEOL TEM or a Bruker Quantax 200 detector mounted on a Hitachi S-5500 STEM. SEM images were collected using a Zeiss Supra 40 VP SEM operated at 10 kV accelerating voltage. Images were collected through the in-lens detector. SEM samples were prepared by depositing a thin layer of the nanocrystals on a conductive surface. XRD data was collected on a Bruker-Nonius D8 advance θ−2θ powder diffractometer equipped with a Bruker Sol-X Si(Li) solid state detector and a rotating stage. 1.54 Å radiation (Cu Kα) was used to collect at 0.02 increments of 2θ at a scan rate of 12 °/min while the sample was rotating at 15 RPM. XRD samples were prepared by depositing a relatively thick (≈10 μm) film of nanocrystals on a glass substrate. UV-vis-NIR absorbance spectra were measured using a Varian Cary 500 spectrophotometer.
2.3 PV device fabrication
PV devices were fabricated with a layered CuInSe2/CdS/i-ZnO structure, similar to reported by Contraras . Sodalime glass substrates (Delta Techology) were cleaned by sonication in an acetone/isoproponol mixture, followed by rinse with DI water, and drying under a nitrogen. Back contact layers of 5 nm thick Cr (99.999%, Lesker) adhesion layer with 60 nm of Au (99.95%, Lesker) were deposited by thermal evaporation. Nanocrystal dispersions were deposited on the Au back contacts by spray deposition of dispersions in toluene with concentrations of 20 mg/ml nanocrystals using a commercial spray gun (Iwata Eclipse HP-CS) operated at 50 psig head pressure. The nanocrystal layer thickness was obtained by profilometry or from SEM images of cross-sectioned devices. Following nanocrystal deposition, a CdS layer was deposited with the method reported by McCandless and Shafarman . Then, 50 nm of ZnO was AC sputtered (99.9% Lesker, 5 ppm O2 in Ar sputtering gas) followed by another 600 nm layer of sputtered ITO (99.99% Lesker, UHP Ar sputtering gas). The final active area of the device is 8 mm2, a 4 mm by 2 mm rectangle. Completed devices were placed in a vacuum oven at 200 °C for up to 40 minutes to improve the device performance.
Mott-Schottky measurements were performed on PVs with slightly different structures. 300 nm of AC sputtered ITO was used as substrate to deposit CuInSe2 nanocrystal films as described above. The top contact was a 40 nm layer of Al (Lesker, 99.99%) thermally evaporated from a tungsten boat. The active device area was 8 mm2.
2.4 PV device testing
Current-potential (I-V) characteristics were collected using a Keithley 2400 general purpose sourcemeter and a Xenon lamp solar simulator (Newport) equipped with an AM1.5 optical filter. Intensity of the light source was calibrated using a NIST calibrated Si photodiode (Hamamatsu, S1787-08). Different fractions of solar spectrum were generated by placing colored glass cutoff filters (Newport) directly in the path of light beam emanating from the solar simulator. Incident photon-to-electron conversion efficiency (IPCE) measurements were performed at zero bias between 300 and 1300 nm in 10 nm steps using an in-house fabricated spectrophotometer. Monochromatic light was generated using a commercial monochromator (Newport Cornerstone 260 1/4M). Generated light was chopped at 213 Hz and was focused to a spot size of 1 mm in diameter on the active region. The response of the device was recorded using a lock-in-amplifier (Stanford Research Systems, model SR830). The light intensity was calibrated using calibrated photodiodes of silicon (Hamamatsu) and germanium (Judson).
Impedance characteristics were measured by applying a 50 mV A-C waveform at frequencies between 0.1 Hz and 107 Hz with 5 steps per decade using a Solartron 1260A Frequency Response Analyzer coupled with a Solartron 1296 Dielectric Interface.
3. Results and discussion
3.1 CuInSe2 nanocrystal inks
Figure 1 shows TEM images and XRD data for the CuInSe2 nanocrystals used to fabricate the PV devices. The nanocrystals are crystalline, with irregular, faceted shapes and an average diameter of 14 nm ± 4 nm. XRD (Fig. 1C) and EDS confirmed that the nanocrystals have chalcopyrite crystal structure with stoichiometric (1:1:2) Cu:In:Se composition. Figure 1C also shows an illustration of the chalcopyrite unit cell of CuInSe2. The nanocrystals are coated with a monolayer of oleylamine, which stabilizes their size and provides dispersibility in organic solvents.
3.2 PV device fabrication
PV devices were fabricated by spray-coating the CuInSe2 nanocrystal ink onto Au back contacts on glass substrates. Nanocrystal films with uniform thickness with few pinholes or cracks are obtained using this method. Devices have been made using this approach with power conversion efficiencies under AM1.5 illumination of up to 3.1%. Figure 2 shows the dark and light I-V curves for the device with highest power conversion efficiency (PCE).
3.3 Diode behavior in the dark and light
The device in Fig. 2 exhibits a “crossover” between the dark and the light I-V curves at forward bias, which is commonly observed in our devices. This crossover is undesirable as it leads to a decrease in device efficiency, and ideally should be prevented. The reason for this crossover can be deduced by modeling the device current density, J, based on a single junction diode model:Eq. (1), J0 is the reverse bias saturation current density, A is the device area, n is the ideality factor of the diode, Rs is the series resistance of the diode, Rsh is the shunt resistance of the diode, Jph is the photogenerated current density, k is Boltzmann’s constant and T is temperature. Table 1 lists the device parameters obtained by fitting Eq. (1) to the device data in Fig. 2. The large value of n, greater than 3, suggests that the devices are dominated by recombination current and illumination with light increases the non-ideality of the diode. Rs and Rsh both decrease under illumination and J0 increases by two orders of magnitude, indicating that the crossover is an outcome of photoconductivity of the materials in the device—mostly likely the CdS layer (see discussion in next paragraph). A reduction in Rs is desirable, as it lowers the barrier for current extraction, but reduced Rsh and increased J0 are undesirable and result from higher recombination within the device.
The origin of the photoconductivity effect leading to the crossover of the light and dark curves in the device discussed above was examined by shining light on the device with different ranges of wavelengths. The CuInSe2 nanocrystals have a band gap of about 1 eV, corresponding to a wavelength of 1236 nm. ZnO and CdS have much wider band gaps of 3.3 eV and 2.4 eV, corresponding to wavelengths of 375 nm and 515 nm, respectively. Figure 3A shows the I-V response of the device when it was illuminated with light with the high energy photons filtered out, as shown in Fig. 3B.
As shown in Fig. 3A, the amount of crossover between the light and dark curves decreased when the shorter wavelength light was filtered, and it was completely eliminated when the illumination had wavelengths larger than the absorption edge of the CdS (515 nm). Similarly, the shunt current in the reverse bias is reduced as longer wavelength light is used. These data show that it is the CdS buffer layer that leads to the high leakage current under illumination.
Elimination of the shorter wavelength light and higher series resistance also significantly changed the fill factor of the device. The devices performed well under AM1.5 illumination, with a fill factor of 0.56. With a 515 nm cutoff filter, the fill factor decreased to 0.29, and with a 630 nm cutoff filter, it decreased to 0.24. Low fill factors in vapor-deposited CuInSe2  have been attributed to type-I band alignment between the CuInSe2 and CdS layers. Figure 4 shows the expected band alignment at the CuInSe2/CdS/ZnO heterojunction . The CuInSe2 nanocrystals are p-type, and there is expected to be a “spike” in the conduction band alignment with the CdS buffer layer that creates a barrier to electron extraction under forward bias. The CdS buffer layer has a significant concentration of low energy donors that leads to its n-type behavior. There are additional deep electron traps, however, that reduce the concentration of mobile carriers present in the n-type CdS layer. When excitons are generated in the CdS layer, the deep traps are compensated by “photo-doping” , which increases the number of mobile carriers, reduces the barrier to electron transport across the CdS layer and leads to an increased junction conductance (Fig. 4) [21,24]. Increased conductance is the reason for the observed crossover of the dark and light I-V curves. Additionally, as the concentration of mobile carriers in the CdS layer is increased, the Schottky barrier between the n-type CdS layer and the Au back contact is reduced, which can allow higher leakage current through pinholes and cracks in the nanocrystal film when illuminated.
3.4 Device performance limitation
Device efficiencies of 3% are too low for commercialization and need to be improved . The highest efficiency devices are actually composed of relatively thin nanocrystal layers that are only about 150 nm thick. We have found that increasing the nanocrystal layer thickness enhances light absorption but it does not improve device efficiency. Figure 5 shows I-V characteristics of devices made with nanocrystal films of increasing thickness. Jsc actually decreased when the nanocrystal films were made thicker, even though more electrons and holes are being photogenerated. This indicates that the photogenerated carriers cannot be extracted from the nanocrystal layer unless they are relatively close to the junction.
Measurements of the incident photon-to-electron conversion efficiency (IPCE) provide additional insight into how well the devices are performing and what the limiting factors are. In IPCE measurements, the short circuit current is measured as a function of the wavelength of the incident illumination. Figure 6A shows IPCE measurements for devices with varying nanocrystal layer thickness. The IPCE data is essentially an external quantum efficiency (at zero bias) that does not account for how much light is absorbed by the device—it is a measure of charge carriers extracted based on the number of photons that are illuminating the device. Another useful quantity is the internal quantum efficiency, which provides an accounting of the photon absorption and tells what fraction of the photogenerated carriers are actually extracted from the device.
The internal quantum efficiency of the devices, IQE(λ), is the ratio of the wavelength-dependent IPCE, IPCE(λ), to the fraction of the incident light at that wavelength that is absorbed by the CuInSe2 nanocrystal films, f(λ). f(λ) is determined from the transmittance of the top window layer, Ttop(λ); the transmittance of the CuInSe2 nanocrystals layer, T1(λ); and the reflectivity of the back contact, RBC(λ).
It should be noted that this estimate of the IQE(λ) does not account for internal reflection or optical interference effects that may also contribute to f(λ) and represents an upper bound. Fig. 6B shows the device IQE(λ) for devices with different nanocrystal layer thickness. Consistent with the reduced Jsc for devices with thicker nanocrystal films, the thinner devices have much higher IQE, indicating that they are much better at extracting photogenerated carriers, across a wide range of wavelength, than the thicker devices.
The higher IQE and more efficient device performance of the thinner devices is also enhanced by light reflection from the back contact. Especially, the thinner films benefit from a “second pass” of light reflected off the back contact. This is evident in the IPCE measurements at longer wavelegths (600 nm to 1200 nm) where only a very small fraction of the incident light is absorbed by the thinner layers on the first pass. As the films get thicker, a large fraction of the incident photons are absorbed deeper in the nanocrystal layer and the resulting photogenerated carriers are unable to be efficiently extracted. This data also indicates that the photogenerated carriers can only be extracted efficiently when they are generated close to the CuInSe2/CdS/ZnO heterojunction.
3.5 Impedance spectroscopy
The thickness of the depletion width in the nanocrystal layer was determined by measuring the impedance of the device. Figure 7C shows typical impedance data on a CuInSe2 nanocrystal PV device with slightly modified structure. The device geometry (shown in Fig. 6A) was devised to ensure that carrier depletion was limited to the spray deposited CuInSe2 film. The circuit model shown in Fig. 7B was found to provide the best fit to the impedance data. The capacitance of the space charge region Csc, was extracted to determine the majority carrier density and an effective depletion width in the nanocrystal layer using a Mott-Schottky analysis. Csc is related to the doping level NA, and applied voltage V:Eq. (3), Vbi is the built-in voltage of the junction, q is the elementary charge of an electron, ε0 is the vacuum permittivity and εs is the relative permittivity of CuInSe2 (≈10). Figure 7D shows Csc −2 plotted against V. Values of NA and Vbi were determined by fitting Eq. (3) to the data. The depletion layer width can be estimated from the relation:Fig. 7A, depletion occurs only in the p-type nanocrystal layer, and Eq. (4) simplifies to
In a typical device the depletion region thickness was found to be 55 nm in the dark. When the device was illuminated, the depletion region thickness was found to decrease to 45 nm (under AM1.5 illumination). The change in doping level in the CdS layer under light leads to a noticeable change in the device properties, as discussed above. Further work is underway to gain a more detailed understand about the band alignment between layers in the nanocrystal devices.
Power conversion efficiencies above 3% under AM1.5 are demonstrated for ambient processed CuInSe2 nanocrystal-based PVs. The extraction of photogenerated carriers from deep within the CuInSe2 nanocrystal film remains a major challenge. The high concentration of crystal interfaces leads to high recombination. A Mott-Schottky analysis of the space-charge capacitance in the device revealed that the active region of the device is only about 50 nm thick, which is consistent with IPCE and IQE measurements on devices with varying nanocrystal film thickness. Future efforts must focus on increasing the thickness of the space charge region to extract carriers deeper in the nanocrystal layer in order to improve device efficiency.
We thank Al Bard, Heechang Ye, Hyun Park, Paul Barbara, Ananth Dodabalapur and Chris Lombardo for insightful discussions. We also acknowledge financial support by the Robert A. Welch Foundation (Grant no. F-1464) and the Air Force Research Laboratory (Grant no. FA8650-07-2-5061).
References and links
1. M. Grätzel, “Photovoltaic and photoelectrochemical conversion of solar energy,” Philos. Trans. R. Soc. Lond. 365(1853), 993–1005 (2007). [CrossRef]
2. R. Po, M. Maggini, and N. Camaioni, “Polymer solar cells: recent approaches and achievements,” J. Phys. Chem. C 114(2), 695–706 (2010). [CrossRef]
3. L. L. Kazmerski, “Solar photovoltaics R&D at the tipping point: A 2005 technology overview,” J. Electron Spectrosc. Relat. Phenom. 150(2-3), 105–135 (2006). [CrossRef]
4. P. T. Landsberg, H. Nussbaumer, and G. Willeke, “Band-band impact ionization and solar cell efficiency,” J. Appl. Phys. 74(2), 1451–1452 (1993). [CrossRef]
5. A. Luque and A. Marti, “Increasing the efficiency of ideal solar cells by photon induced transitions at intermediate levels,” Phys. Rev. Lett. 78(26), 5014–5017 (1997). [CrossRef]
6. K. E. Knapp, and T. L. Jester, “Energy balances for photovoltaic modules: status and prospects” Proc. of 28th IEEE Photovoltaic Specialist Conf, Piscataway (2000).
7. Y. Liang, Z. Xu, J. Xia, S.-T. Tsai, Y. Wu, G. Li, C. Ray, and L. Yu, “For the bright future-bulk heterojunction polymer solar cells with power conversion efficiency of 7.4%,” Adv. Mater. 22(20), E135–E138 (2010). [CrossRef] [PubMed]
8. A. J. Nozik, “Quantum dot solar cells,” Physica E 14(1-2), 115–120 (2002). [CrossRef]
10. J. M. Luther, M. Law, M. C. Beard, Q. Song, M. O. Reese, R. J. Ellingson, and A. J. Nozik, “Schottky solar cells based on colloidal nanocrystal films,” Nano Lett. 8(10), 3488–3492 (2008). [CrossRef] [PubMed]
12. G. I. Koleilat, L. Levina, H. Shukla, S. H. Myrskog, S. Hinds, A. G. Pattantyus-Abraham, and E. H. Sargent, “Efficient, stable infrared photovoltaics based on solution-cast colloidal quantum dots,” ACS Nano 2(5), 833–840 (2008). [CrossRef]
13. J. J. Choi, Y. F. Lim, M. B. Santiago-Berrios, M. Oh, B. R. Hyun, L. Sun, A. C. Bartnik, A. Goedhart, G. G. Malliaras, H. D. Abruña, F. W. Wise, and T. Hanrath, “PbSe nanocrystal excitonic solar cells,” Nano Lett. 9(11), 3749–3755 (2009). [CrossRef] [PubMed]
14. A. G. Pattantyus-Abraham, I. J. Kramer, A. R. Barkhouse, X. Wang, G. Konstantatos, R. Debnath, L. Levina, I. Raabe, M. K. Nazeeruddin, M. Grätzel, and E. H. Sargent, “Depleted-heterojunction colloidal quantum dot solar cells,” ACS Nano 4(6), 3374–3380 (2010). [CrossRef] [PubMed]
15. M. G. Panthani, V. Akhavan, B. Goodfellow, J. P. Schmidtke, L. Dunn, A. Dodabalapur, P. F. Barbara, and B. A. Korgel, “Synthesis of CulnS2, CulnSe2, and Cu(InxGa(1-x))Se2 (CIGS) nanocrystal “inks” for printable photovoltaics,” J. Am. Chem. Soc. 130(49), 16770–16777 (2008). [CrossRef] [PubMed]
16. C. Steinhagen, M. G. Panthani, V. Akhavan, B. Goodfellow, B. Koo, and B. A. Korgel, “Synthesis of Cu(2)ZnSnS(4) nanocrystals for use in low-cost photovoltaics,” J. Am. Chem. Soc. 131(35), 12554–12555 (2009). [CrossRef] [PubMed]
19. L. Stolt, J. Hedström, J. Kessler, M. Ruckh, K.-O. Velthaus, and H.-W. Schock, “ZnO/CdS/CuInSe2 thin-film solar cells with improved performance,” Appl. Phys. Lett. 62(6), 597–599 (1993). [CrossRef]
20. B. E. McCandless, and W. N. Shafarman, U.S. Patent 6,537,845, 2003.
21. A. O. Pudov, J. R. Sites, M. A. Contreras, T. Nakada, and H.-W. Schock, “CIGS J-V distortion in the absence of blue photons,” Thin Solid Films 480–481, 273–278 (2005). [CrossRef]
22. U. Rau and H. W. Schock, “Electronic properties of Cu(In,Ga)Se2 heterojunction solar cells–recent achievements, current understanding, and future challenges,” Appl. Phys., A Mater. Sci. Process. 69(2), 131–147 (1999). [CrossRef]
23. A. Niemegeers, M. Burgelman, and A. De Vos, “On the CdS/CuInSe2 conduction band discontinuity,” Appl. Phys. Lett. 67(6), 843–845 (1995). [CrossRef]
24. M. Burgelman, F. Engelhardt, J. F. Guillemoles, R. Herberholz, M. Igalson, R. Klenk, M. Lampert, T. Meyer, V. Nadenau, A. Niemegeers, J. Parisi, U. Rau, H.-W. Schock, M. Schmitt, O. Seifert, T. Walter, and S. Zott, “Defects in Cu(In,Ga)Se2 semiconductors and their role in the device performance of thin film solar cell,” Prog. Photovoltaics. 5(2), 121–130 (1997). [CrossRef]
25. V. A. Akhavan, B. W. Goodfellow, M. G. Panthani, and B. A. Korgel, “Nanocrystal Inks: towards a new generation of low cost photovoltaics,” Mod. Energy Rev. In press.