Electron transport and recombination in dye sensitized solar cells fabricated from obliquely sputter deposited and thermally annealed TiO2 films

Journal of

Electroanalytical Chemistry Journal of Electroanalytical Chemistry 605 (2007) 151–156 www.elsevier.com/locate/jelechem

Electron transport and recombination in dye sensitized solar cells fabricated from obliquely sputter deposited and thermally annealed TiO2 films Sebastian M. Waita a, Bernard O. Aduda a, Julius M. Mwabora a, Claes G. Granqvist b, Sten-Eric Lindquist b, Gunnar A. Niklasson b, Anders Hagfeldt c, Gerrit Boschloo c,* a Department of Physics, University of Nairobi, PO Box 30197, Nairobi, Kenya Department of Engineering Sciences, The A˚ngstro¨m Laboratory, Uppsala University, PO Box 534, SE-75121 Uppsala, Sweden Center of Molecular Devices, Department of Chemistry, Royal Institute of Technology, Teknikringen 30, SE-10044 Stockholm, Sweden b

c

Received 11 November 2006; received in revised form 14 March 2007; accepted 2 April 2007 Available online 12 April 2007

Abstract Dye sensitized solar cells based on annealed titanium dioxide films prepared by oblique reactive DC magnetron sputtering have been investigated in detail. Electron transport and recombination were studied using intensity-modulated photocurrent and photovoltage spectroscopy. Electron transport time as well as lifetime were found to increase upon lowering of the light intensity and to increase upon increasing the thickness of the TiO2 film. The properties are very similar to those observed for solar cells based on colloidal TiO2 films despite the morphologies being very different. In all cases, films are composed of a porous assembly of TiO2 nanocrystals. Grain boundaries with associated trap and/or energy barriers may explain the observed transport properties. Ó 2007 Elsevier B.V. All rights reserved. Keywords: Nanostructured TiO2; Mesoporous materials; Reactive sputtering; Electron trapping

1. Introduction Dye sensitized solar cells (DSSCs) based on nanocrystalline TiO2 have been named as a possible low-cost alternative to conventional silicon-based photovoltaics [1,2]. Solar to electrical power conversion efficiencies exceeding 11% have been achieved in several laboratories [3]. The structure and working mechanism of DSSCs differ completely from those of conventional photovoltaics and make use of thin films of porous nanocrystalline TiO2 onto which dyes are adsorbed with the object of harvesting solar radiation. On irradiation, photoexcited electrons are injected from the dye into to the TiO2 conduction band and are transported to the back contact and the external circuit. Iodide ions in the electrolyte reduce the oxidized dye. *

Corresponding author. Tel.: +46 8 790 8178. E-mail address: [email protected] (G. Boschloo).

0022-0728/$ - see front matter Ó 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.jelechem.2007.04.001

Finally the triiodide ions formed in this reaction are reduced to iodide at a platinized counter electrode. Charge collection in the DSSC is surprisingly efficient considering that injected electrons travel relatively slowly towards the substrate by diffusion and remain in close proximity to electron acceptors in the electrolyte. It is known that electron transport is strongly dependent on light intensity [4–7]. Electron transport properties are therefore best studied using techniques such as intensitymodulated photocurrent spectroscopy (IMPS), wherein a small sinusoidal light intensity is superimposed on a larger steady background level illuminating the solar cell [4–7]. A closely related technique, intensity-modulated photovoltage spectroscopy (IMVS), can be used to measure electron lifetime under open-circuit conditions [8]. Several studies have focused on the effect of the nanocrystalline TiO2 film on the electron transport. For instance, the porosity was found to play an important role

152

S.M. Waita et al. / Journal of Electroanalytical Chemistry 605 (2007) 151–156

as it affects the number of connections between neighboring particles and the number of dead ends [9]. Electron transport was therefore found to be slower in more porous films. Also the TiO2 particle size has been investigated; increasing size was found to lead to faster electron transport and slower recombination [10]. While most work has been done on films prepared from TiO2 colloids, other methods are also suitable to obtain films for DSSCs. Oblique deposition of TiO2 by sputtering [11,12] or by electron beam evaporation [13] yield porous TiO2 films. Highly porous ‘‘penniform’’ structures have been made by reactive sputter deposition onto rotating substrates [11,12]. We will show below that porous columnar TiO2 films can be formed by oblique sputtering without rotation and that such films are suitable as electrodes in DSSCs after thermal annealing. Furthermore we will report on electron transport and recombination in such films.

2. Film preparation and characterization Film preparation for the working electrode was done in two stages: sputter deposition and subsequent thermal annealing. Reactive DC magnetron sputtering was carried out in a Balzers UTT 400 vacuum chamber with turbo molecular pumping. The target was a Ti disc, 5 cm in diameter and 0.625 cm thick, with a purity of 99.9%. A constant DC current of 750 mA at a power of 385 W were applied to the target. After pumping to 107 mbar, Ar gas (99.998%) was introduced into the sputtering chamber at 100 ml/min. To remove surface contaminants on the target, pre-sputtering was done for 10 min in pure argon. Oxygen (99.998%) was then introduced into the chamber. The Ar/O2 gas ratio was maintained at 0.02 while the working pressure was 12 mTorr. The average deposition rate was 0.3 nm/s. Transparent and electrically conducting fluorine-doped tin oxide (SnO2:F, denoted FTO) coated glass substrates (Hartford Glass Company, Inc.) with a sheet resistance of 8 X/square were used. Vacuum tape fixed the substrates onto the substrate holder and provided a step for film thickness determination. The substrates were positioned 11.5 cm from the target at 60° angle between the direction of the sputtered flux and the substrate normal. The substrate was not heated and was kept stationary during the deposition. The films were subsequently annealed in air at 450 °C for 4 h using a programmable furnace to obtain the desired stoichiometry and crystallinity of TiO2. The annealing temperature was above the amorphous-to-crystalline (anatase) transition temperature at 350 °C but below the softening point of the glass substrate at 550 °C. The samples were removed from the oven after cooling to room temperature. The film thickness was determined using a Tencor Alpha-Step 200 surface profilometer. A Siemens D5000 diffractometer with Cu anode and grazing incidence unit was used for X-ray diffraction (XRD). The morphol-

ogy of the film was studied using a Leo Gemini 1550 Field Emitting Gun scanning electron microscope. 3. Solar cell fabrication The TiO2 films were heated at 450 °C for 10 min to remove possible contaminants such as water vapor and were then allowed to cool to 80 °C before immersion into the dye complex (0.5 mM bis(tetrabutylammonium)cisbis(thiocyanato)bis(2,2 0 -bipyridine-4-carboxylicacid,4 0 -carboxylate)ruthenium(II)), commonly referred to as N719 (Solaronix S.A.) in ethanol solution for a day. Excess dye was removed by rinsing with ethanol. The counter electrode was thermally platizined FTO glass, prepared by spreading 10 ll of 5 mM H2PtCl6 in isopropanol on the glass and heating in air at 380 °C for 10 min. The working and counter electrodes were sandwiched using a frame of thermoplastic (Surlyn 1702) and were laminated for about one minute at 100 °C in vacuum. The electrolyte (0.5 M LiI, 0.05 M I2, and 0.5 M 4-tert-butyl pyridine in 3-metoxypropionitrile (3-MPN)) was introduced through a hole in the counter electrode that was sealed afterwards. Electrical contacts were made by silver paint. The active area of the cell was 0.785 cm2. Three solar cells were fabricated; they are designated 1, 2, and 3, and are characterized by their TiO2 film thicknesses being 3, 6.7, and 10 lm, respectively. IMPS and IMVS measurements were carried out using illumination from a diode laser with a wavelength of 635 nm. The solar cell was connected directly to a lock-in amplifier (Stanford Research Systems SR 830) in the case of IMVS, or via a low-noise current amplifier (Stanford Research Systems SR 570) in the case of IMPS. The intensity of the incident laser light was measured using a calibrated silicon diode. All measurements were performed in a closed black metallic box. Solar cells were illuminated from the photoanode side. 4. Results and discussion Fig. 1 shows XRD patterns for sputter deposited TiO2 films. As-deposited films are amorphous and the observed peaks are due to the FTO back contact. Films annealed at 450 °C are crystalline and show evidence only for the anatase phase of TiO2. These results are in agreement with literature data [14,15]. The mean size of the crystallites, calculated from the broadening of the anatase (1 0 1) peak using the Scherrer equation, was in the range of 33–37 nm. Scanning electron microscopy (SEM) was used to study the morphology of the TiO2 films. Fig. 2a is a top-down image showing a cauliflower-like surface with clearly defined particulates. Their size are 27–38 nm, i.e., consistent with those determined by XRD. Voids are clearly visible and are essential for efficient dye sensitized solar cells. The voids make the films porous and increase the surface area so that a sufficient adsorption of dye can take place throughout the films. The voids also can provide light scat-

S.M. Waita et al. / Journal of Electroanalytical Chemistry 605 (2007) 151–156

10.0 µm

2500

2

A (105) A (211)

Sn O :F

A (200)

2

SnO :F

3000

A (004)

A (101)

2

Intensity (Arb. units)

3500

SnO :F

4000

2000 6.7 µm

1500 1000

3.0 µm

500 as deposited

0 25

30

35

40

45

50

55

60

Diffraction angle, 2theta (degrees) Fig. 1. XRD patterns of as-deposited and annealed sputter deposited TiO2 films. The peaks are designated with the reflections of the anatase (A) phase; also shown are reflections from the SnO2:F transparent conductor. The as-deposited TiO2 films showed the same pattern irrespective of their thickness.

153

The columns are slightly tilted by 10° with respect to the substrate normal in the direction towards the sputtered flux during the film deposition. Inclinations of this type are expected for oblique-angle physical vapor deposition [16]. Fig. 3 shows current–voltage characteristics of dye sensitized solar cells recorded in simulated sunlight at two intensities, 100 mW cm2 (1 sun, a) and 10 mW cm2 (0.1 sun, b). The photocurrent increases with increasing film thickness, indicating that light absorption is a limiting factor in the solar cell. The roughness factor (i.e., the ratio between the actual surface area and projected area) of the obliquely sputtered TiO2 films appears to be significantly lower than that of films prepared from TiO2 colloids. Overall light to electric power conversion efficiencies are lower at 1 sun than at 0.1 sun. This is due to the poorer fill factor found at the higher intensity (0.5 vs. 0.7 at 0.1 sun) and due to the non-linear increase of the short-circuit photocurrent. The low fill factor can to a large extend be attributed to the non-ideal cell design, giving relatively large resistance losses in the conducting glass. Other factors may, however, also play a role. Poor

7 Cell 3 (10.0 µm) η=2.1%

6 5 Cell 2 (6.7 µm) η=1.7 %

4 3

Cell 1 (3.0 µm) η=1.3%

2 1 0

0

0.1 0.2 0.3 0.4

0.5

0.6 0.7

0.8

0.8 Cell 3 (10.0 µm) η=3.3%

0.7 0.6

Cell 2 (6.7 µm) η=2.6%

0.5 0.4 Cell 1 (3.0 µm) η=1.5%

0.3 0.2 0.1

Fig. 2. Top-down (a) and cross-sectional (b) SEM images of sputter deposited TiO2 films. Horizontal bars indicate magnifications.

tering and thereby lead to an increase of the solar cell efficiency [3]. Fig. 2b is a cross-sectional SEM image of the TiO2 film and gives a clear picture of a columnar structure.

0 0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Fig. 3. Current–voltage characteristics of dye sensitized solar cells based on sputter deposited TiO2 films in simulated sunlight. Intensities: 100 mW cm2 (a) and 10 mW cm2 (b). Film thicknesses are indicated.

154

S.M. Waita et al. / Journal of Electroanalytical Chemistry 605 (2007) 151–156

connectivity of the pores or too small pore size in the sputtered TiO2 may for instance occur, limiting both fill factor and short circuit current as the conduction of the redox pair in the electrolyte may be obstructed. The photovoltaic conversion efficiency increased with increasing film thickness and reached 3.3% for the thickest film at 0.1 sun. This efficiency is comparable to 4.1% reported for evaporated films [13] but somewhat lower than 7% reported for optimized films made by sputtering [17]. It is noted that we did not optimize the electrolyte with respect to efficiency. Studies of electron transport and lifetime were performed in order to compare the properties of the sputter deposited and colloid-based TiO2 films. Transport of photoinjected electrons in the sputtered TiO2 films was studied using IMPS at short-circuit conditions. Fig. 4a shows typical IMPS responses for cells with different TiO2 film thickness in a complex plane plot. The data form depressed semicircles that increase in radius with increasing TiO2 film

thickness. This is a direct result of the larger photocurrents obtained in thicker films. Fig. 4b displays real and imaginary parts of the modulated photocurrent as a function of frequency for one solar cell at different light intensities. It can be seen that the IMPS response shifts towards higher frequencies with increasing light intensities while the amplitude of the signals remains the same. The general shape and behavior of the IMPS signals are very similar to those recorded for cells prepared from TiO2 colloids. A single time constant can be extracted from the IMPS data. Fig. 4c shows the IMPS time constants (sIMPS) of the three cells as a function of light intensity. As will be shown later, sIMPS can be interpreted as the electron transport time. Electron transport becomes more rapid at higher light intensities, as is the case for dye sensitized solar cells based on colloidal TiO2. The transport times are similar to those found for colloidal TiO2 films under identical conditions [18,19]. This behavior can be explained from a mul-

Iph (real) (µA)

0

-0.2 -0.3 -0.4

Iph (imag) (µA)

Imag photocurrent (µA)

-0.1

-0.5 -0.6 Cell 1 (3.0 µm) Cell 2 (6.7 µm) Cell 3 (10.0 µm)

-0.7 -0.8 0

0.2

0.4

0.6

0.8

1

1.2

1.5 3.2 1.8 -2 0.83 mW cm

1

0.5

0

-0.5 0.1

1.4

1

10

100

Frequency (Hz)

Real photocurrent (µA) -5

10

6×10

0.08

-5

3

IMPS

(s)

5×10 0.04

-5

4×10

-2

QSC (C cm )

τ

Transport time (s)

1

1000

0.1

0

0

50 2

100 2

d (μm )

2

-5

3×10

-5

2×10

0.01

1

-5

1×10

Cell 1 (3.0 µm) Cell 2 ( 6.7 µm) Cell 3 (10.0 µm)

0.001 0.1

0

1

Light Intensity

10

(mW/cm2)

100

-5

10

10

-4

10

-3

-2

JSC (A cm )

Fig. 4. (a) Complex plane plot of measured IMPS spectra for solar cells under monochromatic irradiation at 0.83 mW cm2. (b) Imaginary and real photocurrents as a function of frequency for cell 2 (6.7 lm thick) irradiated with different light intensities as indicated. (c) Electron transport time as a function of light intensity for the solar cells. The inset shows the dependence of sIMPS at a current density of 0.1 mA cm2 as function of the square of the TiO2 film thickness. (d) Accumulated charge in solar cells under short-circuit conditions as a function of short-circuit current density. The dotted lines indicate a logarithmic fit.

S.M. Waita et al. / Journal of Electroanalytical Chemistry 605 (2007) 151–156

Electron lifetime (s)

1

0.1 Cell 1 (3.0 um) Cell 2 (6.7 um) Cell 3 (10.0 um)

0.01 0.1

1

10

Light intensity (mW cm-2)

Cell 1 (3.0 µm)

0.8

Cell 2 ( 6.7 µm) Cell 3 (10.0 µm)

Electron lifetime (s)

tiple-trapping model in which an exponentially increasing density of traps towards the conduction band is assumed [7,20]. The transport time increases with film thickness. If diffusion is the driving force for electron transport one would expect sIMPS to scale with the square of the film thickness. It appears, however, that sIMPS for the solar cells based on sputter deposited TiO2 is proportional to the film thickness. The reason for this apparent disagreement is that the current densities of the films differ significantly. If sIMPS is plotted against current density instead, sIMPS scales well with the square of the film thickness (see inset in Fig. 4c). The number of electrons accumulated in TiO2 under short-circuit conditions (QSC) was determined by charge extraction. Fig. 4d displays QSC as short-circuit current density. QSC increases with light intensity and with increasing TiO2 film thickness. The extracted charge is approximately proportional to the film thickness. The extracted charge for the sputter deposited films are comparable with those recorded for films based on colloidal TiO2 [18,19]. Interestingly, the extracted charge appears to follow a logarithmic dependence on the current density rather than a power-law dependence as is found in films based on colloidal TiO2. The lifetime of the electrons in sputter deposited TiO2 films was investigated at open circuit conditions using IMVS. Spectra obtained using this technique form semicircles in the complex plane (not shown) and are similar to results from dye-sensitized solar cells based on colloidal TiO2 [8]. The time constant extracted from the IMVS data corresponds to the electron lifetime se and is shown as a function of light intensity in Fig. 5a. There is a powerlaw dependence between se and light intensity, as has been found for solar cells based on colloidal TiO2 [7,8]. The magnitude of se is also similar to that found in colloidal TiO2 solar cells. The electron lifetime is determined by recombination of electrons in the TiO2 with the oxidized part of the redox couple, I 3 . Interestingly, it was found that se is increased at increasing film thickness. Plotting the lifetime against the open-circuit potential, as done in Fig. 5b, it can be seen that thicker films yield longer lifetimes at a given voltage. As the voltage determines the electron concentration in the TiO2, the change in se is unexpected. A possible explanation can be that the FTO film on the substrate contributes to the back transfer of electrons to the redox couple [21]. This effect will become less important if thicker TiO2 films are present. It is noted that IMVS times are up to one order of magnitude higher than IMPS times at the same light intensity. As the internal potential in the sputter deposited TiO2 will be lower under short-circuit conditions than at open circuit it is clear that the electron lifetime under short-circuit condition will be much longer. Recombination of electrons with the triiodide is therefore negligible under short circuit conditions. The relatively low short-circuit photocurrents for the sputter deposited TiO2 films are therefore not caused by electron recombination losses during electron transport.

155

0.6

0.4

0.2

0 0.5

0.55

0.6

0.65

Open circuit voltage (V) Fig. 5. (a) Electron lifetime under open-circuit conditions as a function of light intensity. (b) Electron lifetime as a function of open-circuit voltage.

5. Conclusion The properties of dye sensitized solar cells based on porous TiO2 films deposited by oblique angle DC magnetron sputtering have been studied in detail. Electron transport and recombination were found to be very similar to those observed for solar cells based on TiO2 films prepared by colloid-based methods such as doctor blading or screen printing. The methods yield films with different morphologies, but it is noted that they all are composed of a porous assembly of TiO2 nanocrystals. Grain boundaries with associated trap and/or energy barriers are the likely cause of the observed transport properties. Acknowledgement The authors thank the International Science Programme at Uppsala University for support. References [1] B. O’Regan, M. Gra¨tzel, Nature 353 (1991) 737. [2] M. Gra¨tzel, J. Photochem. Photobiol. C: Photochem. Rev. 4 (2003) 145.

156

S.M. Waita et al. / Journal of Electroanalytical Chemistry 605 (2007) 151–156

[3] (a) M. Gra¨tzel, Inorg. Chem. 44 (2005) 6841; (b) Y. Chiba, A. Islam, Y. Watanabe, R. Komiya, N. Koide, L. Han, Jpn. J. Appl. Phys. 45 (2006) L638. [4] F. Cao, G. Oskam, P.C. Searson, J. Phys. Chem. 100 (1996) 17021. [5] P.E. de Jongh, D. Vanmaekelbergh, Phys. Rev. Lett. 77 (1996) 3427. [6] L. Dloczik, O. Ileperuma, I. Lauermann, L. Peter, E. Ponomarev, G. Redmond, N. Shaw, I. Uhlendorf, J. Phys. Chem. B 101 (1997) 10281. [7] A.C. Fisher, L.M. Peter, E.A. Ponomarev, A.B. Walker, K.G.U. Wijayantha, J. Phys. Chem. B 104 (2000) 949. [8] G. Schlichtho¨rl, S.Y. Huang, J. Sprague, A.J. Frank, J. Phys. Chem. B 101 (1997) 8139. [9] K.D. Benkstein, N. Kopidakis, J. van de Lagemaat, A.J. Frank, J. Phys. Chem. B 107 (2003) 7759. [10] S. Nakade, Y. Saito, W. Kubo, T. Kitamura, Y. Wada, S. Yanagida, J. Phys. Chem. B 107 (2003) 8607. [11] J. Rodrı´guez, M. Go´mez, J. Lu, E. Olsson, C.G. Granqvist, Adv. Mater. 12 (2000) 341.

[12] M. Go´mez, J. Lu, J.L. Solis, E. Olsson, A. Hagfeldt, C.G. Granqvist, J. Phys. Chem. B 104 (2000) 8712. [13] G.K. Kiema, M.J. Colgan, M.J. Brett, Sol. Energy Mater. Sol. Cells 85 (2005) 321. [14] S.B. Amor, L. Guedri, G. Baud, M. Jacquet, M. Ghedira, Mater. Chem. Phys. 77 (2002) 903. [15] B. Karunagaran, R.T.R. Kumar, D. Mangalaraj, S.K. Narayandass, G.M. Rao, Cryst. Res. Technol. 37 (2002) 1285. [16] H.J. Leamy, G.H. Gilmer, A.G. Dirks, in: E. Kaldis (Ed.), Current Topics in Materials Science, vol. 6, North-Holland, Amsterdam, The Netherlands, 1980, p. 309. [17] M. Go´mez, J. Lu, E. Olsson, A. Hagfeldt, C.G. Granqvist, Sol. Energy Mater. Sol. Cells 64 (2000) 385. [18] G. Boschloo, A. Hagfeldt, J. Phys. Chem. B. 109 (2005) 12093. [19] H. Paulsson, L. Kloo, A. Hagfeldt, G. Boschloo, J. Electroanal. Chem. 586 (2006) 56. [20] J. van de Lagemaat, A.J. Frank, J. Phys. Chem. B 104 (2000) 4292. [21] P.J. Cameron, L.M. Peter, J. Phys. Chem. B 109 (2005) 7392.