APPLIED PHYSICS LETTERS 98, 183113 共2011兲
Carbon nanotube semitransparent electrodes for amorphous silicon based photovoltaic devices S. Del Gobbo,1 P. Castrucci,1,a兲 M. Scarselli,1 L. Camilli,1 M. De Crescenzi,1 L. Mariucci,2 A. Valletta,2 A. Minotti,2 and G. Fortunato2 1
Dipartimento di Fisica, Università di Roma Tor Vergata, via della Ricerca Scientifica 1, 00133 Roma, Italy IMM-CNR, Via del Fosso del Cavaliere 100, 00133 Roma, Italy
2
共Received 24 February 2011; accepted 16 April 2011; published online 6 May 2011兲 Different amounts of single wall carbon nanotubes 共SWCNTs兲 have been sprayed on amorphous silicon substrates to form Schottky barrier solar cells. The measured external quantum efficiency showed a spectral behavior depending on the SWCNT network optical transparency, presenting a maximum up to 35% at a wavelength of about 460 nm. Ultrathin network of SWCNTs acts as semitransparent electrode and forms Schottky barrier with amorphous silicon, enabling new generation low cost amorphous silicon based solar cells. Numerical simulations show a poor efficiency of SWCNT contacts in collecting holes suggesting that improvement in contact quality is needed to further improve solar cell efficiency. © 2011 American Institute of Physics. 关doi:10.1063/1.3588352兴 Carbon nanotubes 共CNTs兲 have attracted more and more attention for application in photovoltaics.1–8 At the base of photovoltaic effect is the presence of p-n or Schottky junctions. Single wall CNTs 共SWCNTs兲 exhibit a semiconducting or metallic behavior depending on their chirality. Therefore, CNTs can form local Schottky junctions 共a兲 on the same nanotube due to the presence of twisting9 or bending, 共b兲 between adjacent tubes forming a bundle,10 and 共c兲 between tubes crossing each other.11 Moreover, the presence of van Hove singularities in the electronic density of states of CNTs provides optical absorption responses extending from the infrared to the near ultraviolet 共UV兲 energy range. Indeed, individual SWCNT showed an interesting photovoltaic effect when an ideal p-n diode junction has been induced by electrostatic doping.1 Also networks of entangled CNTs, where nanotubes touch and cross each other forming many heterojunctions, exhibited sizeable photocurrent and interesting photovoltaic effect.11–14 Recently, a few works appeared dealing with devices formed by CNTs covering a crystalline n-type silicon substrate. In these cases, both the CNT network and the Si underneath participate in the photocurrent activation process. In particular, upon illumination e-h pairs are generated either in CNTs or in the Si substrate and separated at the heterojunctions formed by silicon and CNTs.2–7 Devices with very interesting power efficiency values up to 10% have been reported5 making this system very promising for a new generation solar cell. It is well known that crystalline silicon is very costly and since Si-wafers are relatively thick, there is a certain waste of material and the support is rigid. A good alternative to it can be represented by amorphous silicon that is much cheaper and can be used in very thin layers. However, one of the problems with this material is the use of semitransparent electrodes such as indium-tinoxide 共ITO兲 or ZnO which are rather expensive and brittle. In this work we suggest to replace these materials with a network of SWCNTs, which acts mainly as semitransparent electrode and to a lesser extent as photoactive layer. We suc-
ceeded in obtaining an external quantum efficiency 共EQE兲 as high as 35% for an optimum SWCNT network thickness whose spectral behavior mimics closely that of the amorphous silicon based solar cells. A scheme of the SWCNT/a-Si:H photovoltaic device is reported in Fig. 1. In our experiment, the a-Si:H solar cell substrate has been deposited by plasma enhanced chemical vapor deposition. After depositing a 100 nm thick Cr as back side electrode, 30 nm n+ followed by 350 nm undoped a-Si:H were sequentially deposited at 300 ° C and at a pressure of 0.3 mbar. Then the substrate was passivated by photoresist 共1.4 m thick兲 and windows 共7 ⫻ 9 mm2兲 were opened by selective etching. The front electrode was finally formed by spraying SWCNTs over the vias and two silver paint spots were used to connect the back Cr ohmic contact and the front SWCNT electrode to the external circuit. As purchased SWCNTs 共Sigma-Aldrich兲 were dispersed in 1,2-dichlorobenzene 共99% Sigma-Aldrich兲 and sonicated for several hours until the suspension became fully transparent and that no apparent agglomeration and subsequent precipitation occurred. The as prepared suspension was then simultaneously deposited by airbrush on the a-Si:H substrates and on a 5 ⫻ 10 mm2 UV grade quartz 共film transparency reference兲 slices so to obtain the same SWCNT film thickness on the two substrates. During the spray deposition, both substrates were heated at about 80 ° C. This allows a fast evaporation of the solvent in order to improve the film homogeneity and to avoid the formation of material halos. Different film transparencies have been easily obtained by varying the deposition time. Scanning electron microscopy
a兲
FIG. 1. 共Color online兲 A cross-sectional view of the device design. The insulator part is as thick as 1.4 m.
Author to whom correspondence should be addressed. Electronic mail:
[email protected]. Tel.: ⫹39 06 7259 4545. FAX: ⫹39 06 2023507.
0003-6951/2011/98共18兲/183113/3/$30.00
98, 183113-1
© 2011 American Institute of Physics
Downloaded 05 Sep 2011 to 141.108.253.56. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions
183113-2
Del Gobbo et al.
FIG. 2. 共Color兲 EQE spectra of the SWCNT/a-Si:H device recorded as a function of the incident light wavelength for several SWCNT spraying times. In the inset the EQE spectrum of a 10 nm Au film covering the same a-Si:H device.
共not reported here兲 showed that the sprayed SWCNTs are randomly distributed on the surfaces and form agglomerates of nanotube bundles alternating with empty areas. This characteristic makes the evaluation of the thickness of the SWCNT networks very difficult. In order to average the morphological irregularities of the SWCNT networks, optical transmittance measurements have been used and different samples have been characterized by the transparency of the deposited random network on the UV grade quartz substrate. Optical transmittance decreases as a function of the deposition time from about 90% to 68%, thus indicating the increasing of the average thickness of the nanotube network. Due to this close correspondence, in the following we more easily refer to the SWCNT spraying time. The photocurrent spectra have been measured using an optical set-up made up of a xenon lamp equipped with a monochromator, focusing and collecting optics, a reflecting chopper and lock-in electronics. The light spot has a rectangular shape and a size of 7 ⫻ 5 mm2. The EQE has been calculated starting from the measured photocurrent density, I共兲 according to the equation reported in ref. 15. The incident light power density, P, is of about 9 mW/ cm2 and two neutral density filters have been used to reduce the light intensity down to 3% and 12% of its initial value. Finally, using a Keithley 2602A sourcemeter, we recorded the current-voltage curves in dark and under white light illumination. In Fig. 2 the EQE is reported for several thicknesses of the SWCNT network, corresponding to increasing spraying time. It is noticeable that for very low SWCNT amount covering the a-Si:H substrate the EQE resembles closely the CNTs optical absorbance, with a decreasing trend from UV to red wavelength range. For longer and longer deposition times, the EQE spectral shape changes dramatically up to resemble the typical spectrum recorded for the same a-Si:H device covered by a 10 nm Au film 共inset of Fig. 2兲. In addition the EQE value increases over all the wavelength region investigated by increasing the SWCNT amount. The maximum EQE value is of about 35% at 460 nm and it is reached for a spraying time of 500 s. For longer exposure a marked decrease in the EQE value is observed, though EQE line shape does not change any more. Moreover, for exposure times longer than 250 s, a marked decrease in the EQE spectra is visible at about 720 nm 共⬃1.7 eV兲. Interestingly,
Appl. Phys. Lett. 98, 183113 共2011兲
FIG. 3. 共Color兲 I-V characteristics recorded, for the device obtained after a SWCNT spraying time of 500 s, in dark 共black, open diamonds兲 and under a white light illumination 共red, open triangles兲 of about 9 mW/ cm2. The blue, open circle and the green, open square curves refer to the I-V plots when neutral density filters have been used to reduce the light intensity down to 3% and 12% of its initial value, respectively. The corresponding simulated I-V characteristics 共solid lines兲 are also shown. In the inset, the short circuit currents values are shown as a function of the SWCNT spraying time.
this value closely approaches the a-Si:H energy gap. The present EQE behavior can be interpreted as following. For very small amount of SWCNTs 共up to 250 s spraying time兲 the number of SWCNT/a-Si:H heterojunctions is so small that upon illumination nanotubes dominate the photocurrent generation, separation and transport mechanisms. In fact, the EQE spectral shape resembles closely the SWCNTs absorbance and no hint of the a-Si:H energy gap decrease can be detected. This could mean that the e-h pairs are mostly generated in SWCNTs, splitted at the CNT-CNT or CNT-Si heterojunctions and transported through nanotubes or silicon. When the amount of SWCNT/a-Si:H heterojunctions increases, these junctions are able to separate more and more e-h pairs generated in the a-Si:H substrate so that the contribution of the a-Si:H substrate becomes more and more important. Indeed, the EQE spectral shape changes showing a sharp decreases for wavelengths shorter than 350 nm and longer than 720 nm. The former one can be associated to the a-Si:H surface recombination while the latter to the a-Si:H energy gap. At the same time, the value of the EQE strongly increases passing from about 10% to 20%–25% at 400 nm. This suggests that current photogeneration is much more efficient in silicon than in SWCNTs. When the number of SWCNT/a-Si:H heterojunctions reaches its optimum value 共for a deposition time of 500 s兲, the EQE value gets to 35% at 460 nm and its spectral shape is very similar to that of the same a-Si:H cell covered by a 10 nm Au film 共inset of Fig. 2兲. These results indicate that for such a high SWCNT/a-Si:H heterojunction density the role of SWCNT film as semitransparent conductive electrode is highly enhanced with respect to its photogeneration properties thus confirming the Schottky barrier nature of these heterojunctions. This could be due to the wide spatial extension of the depletion region inside the a-Si:H substrate allowing a great number of e-h pairs coming from silicon to be separated. Finally, the EQE behavior for the spraying time of 600 s can be interpreted in terms of SWCNT partial optical absorption that prevents light to reach the a-Si:H substrate underneath. In Fig. 3 typical current-voltage curves measured in dark and under white light illumination conditions are shown. For the maximum
Downloaded 05 Sep 2011 to 141.108.253.56. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions
183113-3
Appl. Phys. Lett. 98, 183113 共2011兲
Del Gobbo et al.
FIG. 4. 共Color online兲 I-V characteristics recorded, for the device obtained after a SWCNT spraying time of 500 s in dark 共open circles兲 and under a white light illumination 共open triangles兲 of about 8 mW/ cm2. Simulated I-V characteristics are reported as solid lines. In the inset, the density of state in the amorphous silicon bandgap used for the simulations is shown.
value of the light incident power 共open triangles兲, we observe an increase in about seven order of magnitude of the short circuit current, Isc, and an open circuit value, Voc, of 0.5 V. The inset of Fig. 3 exhibits the Isc values measured as a function of the SWCNT spraying time. It is easy to note that the Isc maximum value corresponds to the SWCNT amount giving the EQE maximum value. Furthermore, the Isc and Voc decrease greatly when neutral density filters are applied. From these curves it is evident that for highest photon density no saturation is present for reverse bias. The triangle curve in Fig. 4 represents the typical I-V characteristic under white light illumination in a linear scale. It does not follow the well-known rectifying behavior, shows a fill factor of about 0.19 and a kink around 0.5 eV can be observed. In order to explain such an anomalous behavior, we analyzed the electrical characteristics by using numerical simulations, adopting a drift-diffusion model and introducing a density of state in the amorphous silicon bandgap,16 illustrated in the inset of Fig. 4. The SWCNT/a-Si:H interface has been modeled as a Schottky contact, assuming a workfunction of 4.8 eV for the SWCNT, while the optical generation has been taken into account by introducing in the simulations the presence of a photon beam with wavelength equal to the peak wavelength of the Xe lamp spectrum 共around 400 nm兲. We were able to reproduce both the kink observed in the characteristics measured under illumination 共see Fig. 4兲 as well as the dependence upon the light intensity 共see Fig. 3兲, by introducing a reduced recombination velocity for the holes 共v = 102 cm/ s兲 at the Schottky contact between SWCNT/a-Si:H, if compared to the electron recombination velocity 共106 cm/ s兲, similarly to what observed in organic bulk heterojunction solar cells.17 This implies a slow charge transfer from the a-Si:H to the SWCNTs which induces the observed I-V behavior as a function of the incident power density: namely, the more carriers are photogenerated, the higher is the number of charges accumulating at the blocking contact. The non-optimal contact between CNTs and the amorphous silicon underneath or between CNTs and silver paste could be at the origin of the reduced hole extraction velocity at the Schottky contact. On the basis of this model we are working to improve the SWCNT/a-Si:H contacts in
order to maximize the Isc value and power conversion efficiency of this device. In conclusion, we demonstrated that there exists an optimum thickness value of SWCNT network giving rise to an EQE of 35% when deposited on an a-Si:H substrate. In this case, the EQE spectral shape is very similar to that of the corresponding a-Si:H based solar cells that are centered at about 470 nm. Therefore SWCNTs can act as a good semitransparent electrode that can substitute ITO or ZnO in the a-Si:H based solar cells, allowing simple preparation, excellent flexibility and low cost. Numerical simulations of the solar cells show a poor efficiency of SWCNT contacts in collecting holes thus suggesting that improvement in contact quality is needed to further improve this solar cell efficiency. In brief, we are confident that SWCNT/a-Si:H based devices are promising and that further studies will allow to achieve better performances making them competitive, in terms of costs, efficiency and flexibility, with the more complex, expensive and brittle conventional p-i-n a-Si:H based solar cells 共power conversion efficiency by 10% and EQE of 100%兲 covered by ITO or ZnO semitransparent electrodes. The authors acknowledge the financial support of the Australian Queensland Government Smart Futures Fund NIRAP: “Solar Powered Nano-Sensors For Data Acquisition And Surveying In Remote Areas.” J. U. Lee, Appl. Phys. Lett. 87, 073101 共2005兲. J. Wei, Y. Jia, Q. Shu, Z. Gu, K. Wang, D. Zhuang, G. Zhang, Z. Wang, J. Luo, A. Cao, and D. Wu, Nano Lett. 7, 2317 共2007兲. 3 Y. Jia, J. Wei, K. Wang, A. Cao, Q. Shu, X. Gui, Y. Zhu, D. Zhuang, G. Zhang, B. Ma, L. Wang, W. Z. Wang, J. Luo, and D. Wu, Adv. Mater. 共Weinheim, Ger.兲 20, 4594 共2008兲. 4 Z. Li, V. P. Kunets, V. Saini, Y. Xu, E. Dervishi, G. J. Salamo, A. R. Biris, and A. S. Biris, ACS Nano 3, 1407 共2009兲. 5 P. Wadhwa, B. Liu, M. A. McCarthy, Z. Wu, and A. G. Rinzler, Nano Lett. 10, 5001 共2010兲. 6 M. A. El Khakani, V. Le Borgne, B. Aïssa, F. Rosei, C. Scilletta, E. Speiser, M. Scarselli, P. Castrucci, and M. De Crescenzi, Appl. Phys. Lett. 95, 083114 共2009兲. 7 V. Le Borgne, P. Castrucci, S. Del Gobbo, M. Scarselli, M. De Crescenzi, M. Mohamedi, and M. A. El Khakani, Appl. Phys. Lett. 97, 193105 共2010兲. 8 P. Castrucci, F. Tombolini, M. Scarselli, E. Speiser, S. Del Gobbo, W. Richter, M. De Crescenzi, M. Diociaiuti, E. Gatto, and M. Venanzi, Appl. Phys. Lett. 89, 253107 共2006兲. 9 P. Castrucci, M. Scarselli, M. De Crescenzi, M. A. El Khakani, F. Rosei, N. Braidy, and J.-H. Yi, Appl. Phys. Lett. 85, 3857 共2004兲. 10 D. A. Stewart and F. Léonard, Phys. Rev. Lett. 93, 107401 共2004兲. 11 M. S. Fuhrer, J. Nygård, L. Shih, M. Forero, Y.-G. Yoon, M. S. C. Mazzoni, H. J. Choi, J. Ihm, M. Steven, G. Louie, A. Zettl, and P. L. McEuen, Science 288, 494 共2000兲. 12 S. Lu and B. Panchapakesan, Nanotechnology 17, 1843 共2006兲. 13 D. H. Lien, W. K. Hsu, H. W. Zan, N. H. Tai, and C. H. Tsai, Adv. Mater. 共Weinheim, Ger.兲 18, 98 共2006兲. 14 C. A. Merchant and N. Markovic, Appl. Phys. Lett. 92, 243510 共2008兲. 15 EQE is the ratio between the number of electrons generated by the device and the number of photons impinging it. For every wavelength 共兲, the photocurrent density generated by the sample I共兲 and the lamp power density P共兲 are measured and used as follows: EQE共%兲 = 100关hcI共兲兴ⴱ关P共兲q兴−1 where c is the speed of light, h is the Planck constant, and q is the electronic charge. 16 M. Hack and M. Shur, J. Appl. Phys. 58, 997 共1985兲. 17 M. Glatthaar, M. Riede, N. Keegan, K. Sylvester-Hvid, B. Zimmermann, M. Niggemann, A. Hinsch, and A. Gombert, Sol. Energy Mater. Sol. Cells 91, 390 共2007兲. 1 2
Downloaded 05 Sep 2011 to 141.108.253.56. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions
