SUPPLEMENTARY INFORMATION

Transcription

1 SUPPLEMENTARY INFORMATION doi: 1.138/nphoton Tandem Colloidal Quantum Dot Solar Cells Employing a Graded Recombination Layer Xihua Wang 1,, Ghada I. Koleilat 1,, Jiang Tang 1, Huan Liu 1, 2, Illan J. Kramer 1, Ratan Debnath 1, Lukasz Brzozowski 1, D. Aaron R. Barkhouse 1, Larissa Levina 1, Sjoerd Hoogland 1, and Edward H. Sargent 1,*. These authors contributed equally to this work 1. Department of Electrical and Computer Engineering, University of Toronto, 1 King s College Rd., Toronto, Ontario M5S 3G4, Canada 2. Department of Electronic Science & Technology, Huazhong University of Science & Technology, Wuhan 4374 P. R. China * ted.sargent@utoronto.ca Table of Contents SI1 Tandem CQD solar cell cross-sectional SEM image... 2 SI2 Materials... 2 Chemicals... 2 Colloidal Quantum Dot Synthesis and Purification... 2 SI3 Device modeling... 3 Spatial band diagrams of constituent devices and tandem device... 3 In the case of the bottom single-junction solar cell with PbS CQD (1.6 ev) film... 4 In the case of the top single-junction solar cell with PbS CQD (1 ev bandgap) film... 4 Estimation of the resistance of a suitably-designed GRL, and of a non-optimal RL... 5 SI4 Current matching... 6 SI5 I-V curves of tandem CQD solar cells... 6 SI6 Characterization of GRL materials...7 UPS results... 7 XPS results... 9 Optical absorption results Cyclic voltammetry results FET results nature photonics 1

2 supplementary information SI1 Tandem CQD solar cell cross-sectional SEM image doi: 1.138/nphoton SI2 Materials Chemicals Lead oxide (PbO) (99.9%), oleic acid (9%), bis(trimethylsilyl)sulphide (TMS, synthesis grade), 1-octadecene (9%), 3-mercaptopropionic acid (99%), terpineol, Triton-X and all solvents (anhydrous grade) were obtained from Sigma-Aldrich. Titanium dioxide (TiO 2 ) sputtering target, aluminium-doped zinc oxide (AZO) and indium tin oxide (ITO) sputtering target were obtained from Kurt J. Lesker, Inc. ITO-coated glass substrates were obtained from Delta Technology. MoO 3 was obtained from Alfa Aesar. Colloidal Quantum Dot Synthesis and Purification TMS (.18 g, 1 mmol) was added to 1-octadecene (1 ml), which had been dried and degassed by heating to 8 C under vacuum for 24 hours. A mixture of oleic acid (1.34 g, 4.8 mmol), PbO 2 nature photonics

3 doi: 1.138/nphoton supplementary information (.45 g, 2. mmol), and 1-octadecene (14.2 g, 56.2 mmol) was heated to 95 C under vacuum for 16 hours then placed under Ar. The flask temperature was increased to 12 C and the TMS/octadecene mixture was injected. After injection, the temperature dropped to ~95 C, and the flask was allowed to cool to 36 C. The nanocrystals were precipitated by adding 5 ml of distilled acetone and were centrifuged under ambient conditions. After discarding the supernatant, the precipitate was redispersed in toluene. The nanocrystals were precipitated again using 2 ml of acetone, centrifuged for 5 min, dried, and finally dispersed in toluene (~2 mg ml -1 ). The PbS nanocrystals were then brought into a N 2 -filled glove box. They were precipitated twice using methanol, and then finally redispersed at 5 mg ml -1 in octane. SI3 Device modeling Spatial band diagrams of constituent devices and tandem device To generate the spatial band diagrams of Figure 1c-e, we carried out simulation using the software PC1D [University of New South Wales]. We used electronic materials parameters (see below) obtained from direct measurement of our materials or, when values are well-accepted, from published references. In the table, values measured by us are shown in black; values measured by others are shown in blue. All values measured by others are accompanied by references to the literature. Band parameters in the table below are subject to a.1-.2 ev uncertainty in light of the measurement methods used. The band diagrams are subject to the same uncertainties as a result. We simulated the band diagram of individual single-junction semiconductor devices. We then drew the band diagram of tandem CQD solar cell under the premise that an effective GRL achieves alignment of the Fermi levels in the ITO with that in the TiO 2. Materials thickness dielectric constant electron affinity Ionization energy doping concentration Fermi-level Bottom TiO 2 5 nm ev 2 NA 1 x 1 16 cm ev 2 PbS CQD (1.6 ev) 2 nm ev 5.2 ev 2 x 1 16 cm ev MoO 3 1 nm ev 8.5 ev ~1 19 cm ev ITO 5 nm 4 4 NA NA > 1 2 cm ev 5 AZO 5 nm ev 7.4 ev ~5 x 1 19 cm ev TiO 2 4 nm ev 7.7 ev ~5 x 1 16 cm ev PbS CQD (1 ev) 3 nm ev 5 ev 2 x 1 16 cm ev nature photonics 3

4 supplementary information doi: 1.138/nphoton In the case of the bottom single-junction solar cell with PbS CQD (1.6 ev) film The device modeled using PC1D consisted of 5 nm TiO 2, 2 nm PbS CQD (1.6 ev bandgap), 1 nm MoO 3, and 2 nm ITO. The ITO-TiO 2 contact was treated as ohmic. We found that, for these materials parameters, the entire device was fully depleted, including both TiO 2 and PbS film. A further simulation of expanded PbS CQD (1.6 ev) film to 5 nm thick shows that more than 3 nm of PbS is depleted. The electric field is plotted below for the simulated 5 nm thick cell. In the case of the top single-junction solar cell with PbS CQD (1 ev bandgap) film The device modeled consisted of 3 nm ITO, 5 nm AZO, 4 nm TiO 2 and 3 nm PbS CQD (1 ev bandgap). The ohmic contact condition was applied to the PbS CQD film. We found that the entire device was fully depleted, including both TiO 2 and PbS film. A further simulation of expanded PbS CQD (1 ev) film to 5 nm showed that more than 35 nm of PbS is depleted. The electric field is plotted below for the simulated 5 nm thick cell. The above analysis of depletion width agrees with our previous report (results obtained from capacitance-voltage measurement) for a similar type of PbS CQD (1.3 ev) film 2. 4 nature photonics

5 doi: 1.138/nphoton supplementary information Estimation of the resistance of a suitably-designed GRL, and of a non-optimal RL We sought to assess whether the energetic barriers to electron flow in the GRL device could reasonably be expected to be compatible with the flow of solar current densities without imposing excessive resistive loss. To begin we obtained spatial band diagrams using the same methods as described above. The graded recombination layer (left figure) consisted of 1 nm MoO 3, 5 nm ITO, 5 nm AZO, and 4 nm TiO 2. Spikes appeared in the conduction band at the MoO 3 /ITO and ITO/AZO interfaces (Figure below, left panel). The larger barrier along the path for electrons to flow from right to the left was a.5 ev triangular barrier of width ~ 5 nm. An abrupt, or nongraded, recombination layer, shown on the right below, consists of 1 nm MoO 3, 5 nm ITO, and 4 nm TiO 2. Omitting the shallow work function heavily-doped AZO caused a wide depletion region in TiO 2. The barrier height here of close to 1 ev was expected to impede significantly the flow of appreciable electrical current via thermionic emission. The calculations below show that, for the abrupt recombination layer, a parasitic voltage drop of multiple tenths of an ev would be required to support solar current densities. This would lead to a significantly lowered operating voltage V m at the maximum power point. The calculations below show that, for the GRL, solar current densities can be supported at the expense of a minimal (less than kt) voltage drop. We now provide a more detailed quantitative estimation of the current that can flow in the presence of a ~.5 ev triangular barrier. The tunnel current is usually described with Fowler-Nordheim type equation 7. J= a b -1 F 2 exp(- b b 3/2 /F) Here a ~ 1.5x1-6 A ev V -2 and b ~ 6.8 ev -3/2 V nm -1, =1, =(m e */m e ) 3/2, and ef= b /x d. We use the reference value 8 m e *=.2m e for AZO. For the case of b =.5 ev high and x d =5 nm as obtained from our spatial band diagram calculations above, the barrier can readily support a nature photonics 5

6 supplementary information doi: 1.138/nphoton current density J of order 1 6 A/cm 2 at a cost of less than kt/q of applied bias (corresponding to well under kt/q parasitic voltage drop). Negligible tunneling current density is supported in case of the abrupt (nongraded) device in view of its extremely wide depletion in the TiO 2. We now examine the themionic emission current density: J=A ** T 2 exp(-e b /kt)(exp(v/nkt)-1) A ** is the Richardson constant, b is the barrier height, n is the ideal factor of the diode. Using A**=12 Acm -2 K -2 (an optimistic value intended to give an upper bound on supported current density), b =.9 ev for the ITO/TiO 2 junction, we find only 1-7 ma/cm 2 of current density are supported for kt/q applied bias. We conclude that, from these calculations, the GRL concept is necessary to achieve flow of solar current densities at minimal cost to operating voltage. SI4 Current matching In order to determine the proper film thickness to achieve current matching in our structure, we considered the following in our evaluation: We began by fixing the thickness of the front cell at 2 nm. We had found this thickness to provide the best single-junction visible-cell performance experimentally. The average internal quantum efficiency was up to ~7% in the visible region. We determined the total expected current density available from the single-junction visible-cell (with transparent top contact) to be ~9.2 ma/cm 2. From the absorption coefficient of the large bandgap CQDs (figure 2a), we determined the remaining AM1.5 flux illuminating the back cell by employing the single-junction visiblecell (with transparent top contact) as the visible light filter. We found the average internal quantum efficiency of ~35% for the 1. ev cell under the remaining AM1.5 flux. Thus we estimated the expected current density for a double-pass for various thicknesses of the smallbandgap CQD film and plotted the calculated values in figure 2b. Current matching occurs for 25-3 nm thick films SI5 I-V curves of tandem CQD solar cells Our tandem devices as well as single-junction solar cells need 5-1 minutes of light soaking to achieve their maximum efficiency. Similar phenomena were observed by Grätzel et al. in dyesensitized solar cells 9. We provide in the figure below the forward and reverse scan of a typical device. The curves overlap closely, evidencing minimal hysteresis. 6 nature photonics

7 doi: 1.138/nphoton supplementary information Below we plot the I-V characteristics of a typical device on a log-linear scale (left panel) and show the full I-V scan beyond open-circuit voltage (right panel). SI6 Characterization of GRL materials UPS results Ultraviolet photoelectron spectroscopy (UPS) allows determination of the absolute value of work function (Fermi level, E f ) and ionization potential (equivalent to valence band edge, E v ) of semiconductor materials. UPS was carried out using He I (21.22 ev) photon lines from a discharge lamp. The GRL materials are deposited on a commercial ITO substrate (from Delta Technology). The thickness of MoO 3, AZO and TiO 2 were all 5 nm in order to eliminate background signal from the ITO substrate. The full UPS spectra for MoO 3, AZO and TiO 2 are shown in Figure 4a of the main text. The following plots show the regions of interest. The E f is extracted by subtracting the cut-off value nature photonics 7

8 supplementary information doi: 1.138/nphoton of the curve from the kinetic energy of He I (21.22 ev) photon. The E v is extracted from the cutoff value of the curve, and it represents the energy below Fermi level of the material. MoO 3 MoO 3 Intenstiy (arb. u.) Intenstiy (arb. u.) energy w.r.t. Fermi level (ev) energy w.r.t. Fermi level (ev) The red lines in the above two plots show the cut-off position, and UPS analysis gives values of ev and 3.13 ev. We conclude that MoO 3 has E f of ~5.4 ev and E v of ~8.5 ev. AZO AZO Intensity (arb. u.) Intensity (arb. u.) energy w.r.t. Fermi level (ev) energy w.r.t. Fermi level (ev) The red lines in the above two plots show the cut-off position, and UPS analysis gives the value of ev and 3.34 ev. We conclude that AZO has E f ~4.1 ev and E v ~7.4 ev. TiO 2 TiO 2 Intensity (arb. u.) Intensity (arb. u.) energy w.r.t. Fermi level (ev) energy w.r.t. Fermi level (ev) 8 nature photonics

9 doi: 1.138/nphoton supplementary information The red lines in the above two plots show the cut-off position, and UPS analysis gives the value of 17.1 ev and 3.55 ev. We conclude that TiO 2 has E f ~4.1 ev and E v of ~7.7 ev. XPS results X-ray photoelectron spectroscopy (XPS) is a quantitative spectroscopic technique to ascertain the elemental composition and chemical state of thin films. 2 x 15 q11459.spe 1.8 -O1s Mo3d5 -Mo3d3 -Mo3d c/s Mo3p1 -Mo3p3.8 -O KLL.6 -Mo3s.4 -C1s.2 -Mo4p -O2s Binding Energy (ev) The above XPS survey of MoO 3 shows the existence of Mo and O. Further quantitative analysis gives the atomic concentration of Mo/O=28%/72%. nature photonics 9

10 supplementary information doi: 1.138/nphoton x 15 q11453.spe 8 -Zn2p Zn2p1 c/s O KLL The above XPS survey of AZO shows the existence of Zn, O, and Al. Further quantitative analysis gives the atomic concentration of Zn/Al=98%/2%. 18 x 14 q11462.spe -Zn LMM2 -Zn LMM3 -O1s -Zn LMM -Zn LMM1 6 5 Binding Energy (ev) 4 -C1s 3 2 -Zn3s -Al2s c/s -O1s -Zn3p -Al2p -O2s -Zn3d -Ti2p3 -Ti2p1 -Ti2p 8 -Ti LMM1 6 -Ti LMM -O KLL -Ti2s 4 2 -C1s -Ti3s -Ti3p -O2s Binding Energy (ev) The above XPS survey of AZO shows the existence of Ti and O. Further quantitative analysis gives the atomic concentration of Ti/O=31%/69%. 1 nature photonics

11 doi: 1.138/nphoton supplementary information Optical absorption results Optical absorption measurements were carried out using a Varian Cary 5 UV-Vis-IR Scan spectrophotometer. Each material was deposited onto a transparent glass substrate. The same thickness of 5 nm was used for each of the following three samples; MoO 3, AZO and TiO 2. The results were shown in the graphs below. The spectra show that each material is highly transparent. We determined the optical bandgap of these materials to be 3.1 ev, 3.3 ev, and 3.4 ev for MoO 3, AZO and TiO 2, respectively. We observed a.3 ev discrepancy in the bandgap of TiO 2 obtained using UPS and cyclic voltammetry. Cyclic voltammetry results Cyclic voltammetry (CV) is a type of potentiodynamic electrochemical measurement. It has been applied to obtain the LUMO and HOMO levels of organic materials and quantum dots 1, as well as the electron affinity of semiconductors 11. Here we use it to measure the electron affinity (equivalent to conduction band edge E c ) of AZO and TiO 2. The results are shown in the following figures. nature photonics 11

12 supplementary information doi: 1.138/nphoton Current ( A) TiO E vs. Ag/AgNO 3 (V) Current ( A) AZO E vs. Ag/AgNO 3 (V) We use the Ag/AgNO 3 (.1M acetonitrile) reference electrode in the measurement, which has the absolute value of -4.7 ev. The position of the reduction peak reflects the E c of the materials. For AZO and TiO 2, we obtained electron affinities of 4.1 ev and 4. ev respectively. The left graph in below shows the same sample went through multiple scans in CV measurement. In order to test the TiO 2 surface response to chemical treatment, we broke a sample into two pieces and applied chemical (MPA/methanol) treatment only on one of them. In the right graph in below, we indicated the two pieces from the same sample as before chemical treatment and after chemical treatment. Since the sample was contaminated after CV measurement, there is a limit to test exactly the same sample before and after chemical treatment. The graph shows the electron affinity of the surface of the TiO 2 is not significantly affected by the MPA/methanol treatments which we use to fabricate our devices. FET results We built field effect transistor (FET) test structures with the goal of estimating the order of magnitude of the free carrier densities within each layer of the GRL. 12 nature photonics

13 doi: 1.138/nphoton supplementary information We use the similar method as in the reference 12 to fabricate FET devices. In the following figures, our FET devices show good modulation along with applied gate bias. I d ( A) 1.6 V 1.4 g = V =.5 V 1.2 = 1. V V 1. g = 1.2 V = 1.5 V.8 = 2. V = 2.5 V.6 = 3. V MoO V d (V) I d ( A) = V =.5 V = 1. V = 1.2 V = 1.5 V = 2. V = 2.5 V = 3. V AZO V d (V) I d ( A) = V =.5 V = 1. V = 1.2 V = 1.5 V = 2. V = 2.5 V = 3. V V d (V) TiO 2 We calculated mobility from the measured transconductance g m I V d g Vd const WCiV L d by taking of slope of the I d vs. curve at linear region. Here, the channel length L =2.5 m, the channel width W = 2 mm, the capacitance per unit area of the insulating layer is C i = 1 F/cm 2. We saw appreciable hysteresis in FET measurements that render our mobility measurements and therefore our extracted free carrier densities accurate to within somewhat better than one order of magnitude. We comment below to the effect that this is sufficient for the purposes of this work (eg providing reasonable accuracy in spatial band diagrams, and in supporting the assertion that MoO 3 and AZO are both essentially degenerately-doped). From the measured.2 S/cm conductivity of AZO and the extracted mobility, we estimated AZO to have a free carrier density in the mid-1 19 cm -3 range. From the measured 6 x 1-5 S/cm conductivity obtained for MoO 3 and the extracted mobility, we estimated MoO 3 to have free carrier density in the low 1 19 cm -3 range. nature photonics 13

14 supplementary information doi: 1.138/nphoton From the measured 1.4 x 1-7 S/cm obtained for TiO 2 and the estimated mobility, we estimated a doping in the mid-1 16 cm -3 range. In semiconductor physics, E f =E c +ktln(n d /N c ), where N d is close to the doping density n, N c is the density of states at conduction band and has the typical value of ~1 19 cm -3 for wide-band gap metal-oxides. Considering the doping of our TiO 2 in the mid-1 16 cm -3 range, the Fermilevel is ~.15 ev below the conduction band edge. 1 Matsunami, H. & Fuyuki, T. Electonic properties of the interface between Si and TiO2 deposited at very low temperatures. Jpn. J. Appl. Phys. 25, (1986). 2 Pattantyus-Abraham, A. G. et al. Depleted-heterojunction colloidal quantum dot solar cells. ACS Nano 4, (21). 3 Simmons, J. G. & Nadkarni, G. S. Electrical properties of evaporated molybdenum oxide films. J. Appl. Phys. 41, (197). 4 Kulkarni, A. K. & Knickerbocker, S. A. Estimation and verification of the optical properties of indium tin oxide based on the energy band diagram. J. Vac. Sci. technol. A 14, (1996). 5 Kim, J. Y. et al. Efficient tandem polymer solar cells fabricated by all-solution processing. Science 317, (27). 6 Takahashi, Y., Kanamori, M., Kondoh, A., Minoura, H. & Ohya, Y. Photoconductivity of ultrathin zinc oxide films. Jpn. J. Appl. Phys. 33, (1994). 7 Forbes, R. G. Physics of generalized Fowler-Nordheim-type equation. J. Vac. Sci. Technol. B. 26, (28). 8 Jayaweera, P. V. V., Perera, A. G. U. & Tennakone, K. Why Gratzel s cell works so well. Inorganica Chimica Acta 361, (28). 9 Hagfeldt, A., Bj rkstén, U. & Grätzel, M. Photocapacitance of nanocrystalline oxide semiconductor films: band-edge movement in mesoporous TiO 2 electrodes during UV illumination. J. Phys. Chem. 1, (1996). 1 Hyun, B.-R. et al. Electron injection from colloidal PbS quantum dots into titanium dioxide nanoparticles. ACS Nano 2, (28). 11 Kim, J. Y. et al. New architecture for high-efficiency polymer photovoltaic cells using solution-based titanium oxide as an optical spacer. Adv. Mater. 18, (21). 12 Kang, M. S., Sahu, A., Norris, D. J. & Frisbie, C. D. Size-dependent electrical transport in CdSe nanocrystal thin films. Nano Lett. 1, (21). 14 nature photonics