Meysam
Pazoki
ab,
Nima
Taghavinia
*a,
Yaser
Abdi
c,
Fariba
Tajabadi
a,
Gerrit
Boschloo
*b and
Anders
Hagfeldt
b
aPhysics Department, Sharif University of Technology, Tehran, 14588, Iran. E-mail: taghavinia@sharif.edu; Fax: +98 2166164119
bDepartment of Chemistry-Ångström Laboratory, Uppsala University, Box 532 SE 75120, Uppsala, Sweden. E-mail: Gerrit.Boschloo@kemi.uu.se; Fax: +46 184713633
cPhysics Department, Tehran University, Box 6455-14155, Tehran, Iran. Fax: +98 2166164119
First published on 9th October 2012
Chemical vapour deposition (CVD) at atmospheric pressure, using TiCl4 as a precursor, was used to grow nanostructured TiO2 films on glass substrates. At relatively low temperatures (∼245 °C) and using relatively high reactant concentrations, different nano-morphologies of TiO2 were formed simultaneously, such as spheres, nanowires and mesoporous structures. The TiO2 spheres were successfully applied as light-scattering particles in dye-sensitized solar cells, either by direct deposition onto electrodes in the reactor, or by preparation of a printing paste from the deposited particles. For dye-sensitized solar cells using the organic dye D35 as sensitizer and a cobalt-complex based redox electrolyte, a solar cell efficiency of 4.4% was obtained using a 5 μm transparent mesoporous TiO2 layer. Addition of a 5 μm light-scattering CVD-particle film increased the efficiency by 22% to 5.4%, which was similar to the result obtained with an equally thick commercially available light-scattering film (5.3%). Longer electron lifetime was found using CVD-based films, which is attributed to the presence of more traps in the bulk of the material.
DSCs as introduced by O'Regan and Grätzel at 199117 have potential as low cost, environmentally friendly solar cells. They work efficiently under indoor conditions and can be used on flexible substrates.5 Increasing the light harvesting efficiency is one of the subjects that is studied widely for increasing the efficiency of DSCs.5,18,19 The addition of a light-scattering layer on top of a transparent mesoporous TiO2 film increases the optical path length in the film, and thus the photocurrent of the DSC. Specifically, the IPCE is improved in regions where the light absorption by the dye is weak. The light-scattering film is usually composed of a blend of TiO2 particles several 100 nm in size and smaller TiO2 particles (20 nm).
The CVD process introduced here shows promise for new, quick, low-cost and low temperature (<250 °C) growth of TiO2 nanostructures. CVD-grown TiO2 particles can be introduced in two ways for use as light-scattering particles in dye-sensitized solar cells: they can be directly grown on the top of a mesoporous TiO2 film inside the CVD reactor (see the ESI†), or a paste can be prepared from CVD-grown particles and deposited onto mesoporous TiO2 films by printing methods. Using the latter approach, the CVD scattering paste increased the power conversion efficiency of a D35-sensitzed TiO2 solar cell with a cobalt-complex based electrolyte from 4.4% to 5.4%.
Fig. 1 Schematic of the CVD setup, illustrating 3 sample zones, A, B and C separated 3 cm from each other. Zone C is in the centre of the reactor. The corresponding SEM images of the deposited films at each sample zone are shown on top. |
It is noted that the distance between zones A–C is small and that the growth is sensitive to the precise position of the substrates. Careful control (cleanliness of the reactor, exact flow regulation and temperature settings, exact positioning of substrates) is essential to obtain reproducible results.
The morphological difference of the TiO2 nanostructures in zones A–C can be explained by differences in temperature, reactant concentrations and gas boundary conditions.24 A general description of the factors responsible for differences in growth is given in the ESI.†
In zone A, nearly mono-disperse, spherical particles with an average size of 450 nm are deposited on the substrate. These spheres are most likely formed in the gas phase, and subsequently deposited on the substrate. The high concentration of reactants in zone A (close to the inlet of reactants), compared to zones B and C, favors gas phase nucleation and the growth of TiO2 particles over surface growth. The gas phase growth of TiO2 spherical nanoparticles has been reported in low pressure reactors (∼mbar), resulting in particles of about 10 nm size.12,13 The higher pressure in our case is expected to considerably decrease the mean free path of monomers and increase the monomer collision rate as well as the residence time in the reactor, leading to the formation of larger particles.
The diameter of the spheres can be tuned by the temperature and partial pressure of reactants (see section S2 in the ESI†). Decreasing the temperature and increasing the flow rate of reactants leads to larger particle sizes, in good agreement with other work25,26 and can be explained by reaction–coagulation–sintering model.27
The specific surface area of annealed nanostructured films deposited in different positions in the CVD reactor was obtained by dye adsorption/desorption measurements using the dye N719 and by weighing the mechanically removed film.28 The specific surface area of the spherical particle film was determined as 3.6 m2 g−1. The average particle size observed by SEM was 450 nm under the conditions used. Using this value a theoretical surface area S can be calculated using S = 6/ρd, with d the particle diameter and ρ the density of the pure material (3.84 g cm−3 for TiO2 anatase), yielding a value of 3.5 m2 g−1, in good agreement with the experimental value. This suggests that these particles are non-porous.
The spherical particles are most likely formed in the gas phase and subsequently deposited on the substrate. There are some indications that indicate some further growth takes place on the substrate: the fact that particles are highly packed in some areas of the film, and that the shape of some particles is distorted in order to accommodate close packing, suggests that particles have a degree of flexibility, when they are deposited. It was also observed that the thickness of the particle film does not increase beyond a certain value (∼4 μm), even though more particles are formed and found in other parts of the reactor. More study on the nature and precise control of particle deposition is required.
In zone B, TiO2 nanowires of ∼150 nm in diameter and ∼2 μm in length are formed. The growth rate is rather high: ∼200 nm min-1. The active surface area, measured by dye loading, for 1 cm2 of a 2 μm-thick film, was 21.9 cm2. The estimated roughness factor of cylindrical wires of 150 nm diameter and 2 μm length with 2.3 × 109 wires cm−2 (estimated from SEM pictures) is 22 cm2 cm−2, which is close to the measured value, and suggests that the wires are not porous inside. Most wires are oriented perpendicular to the substrate. The reason for the asymmetric growth of TiO2 in the form of wires is unclear. Upon close inspection of the SEM images it was found that some nanowires combine together at the top and make a single nanowire, growing from the intersection point (Fig. 2a). This implies that the nanowires grow from the tip. Furthermore, most wires have spherical particles on top and show some radial distortions in the whole area around the diameter, giving it a worm-like appearance (Fig. 2b).
Fig. 2 High resolution SEM pictures of CVD-grown TiO2 nanowires. (a) Nanowires combining on top to a single nanowire. This shows that the growth continues from the top. (b) Vertical growth of wires with a spherical tip and radial rings on the diameter. |
In zone C, at the centre of the reactor, the temperature is higher than in zones A and B, while the reactant concentration is lower. Here, clumps of aggregated nanoparticles are deposited, forming a mesoporous TiO2 structure, giving a uniform coverage of substrate and yielding films with high surface area. For a film of 200 nm thickness, the roughness factor was determined to be 83 cm2 cm−2 using the dye loading measurement. This is about 40 times higher than nanowire films of the same thickness. The film has features of ∼30 nm size in the SEM micrograph (Fig. 1), but the high roughness factor suggests that the actual particle size must be less than 10 nm. The growth rate of these structures is very low: a few nanometres per minute. Unlike the particle and wire films, the as-grown structures are crystalline with the anatase phase. The low growth rate, plus the strong adhesion of the films to the substrate, demonstrates that the films grow on the substrate by the surface-based CVD process.
In Fig. 3 SEM micrographs of CVD and JGC light-scattering films are shown. The films prepared using CVD-particles (collected from zone A) have bigger particles and also have bigger holes compared to the JGC film. Both electrodes consist of a mix of big particles that act as light scatterers, and smaller particles that are required for a good connection between all particles. The latter also increases the surface area of the resulting electrodes. Fig. 4 shows the XRD pattern of a sintered light-scattering CVD film. The phase of the TiO2 particles was mainly anatase, but there is also a small fraction of rutile particles. The anatase crystallite size of the CVD paste, calculated using the Debye–Scherrer formula, was 10.7 nm. This suggests that the large particles observed in SEM are aggregates built up from much smaller crystals. This is confirmed by high resolution SEM (see Fig. S3 in the ESI†). Raman spectra of the large particles display anatase peaks (Fig. S4 in the ESI†). From the full width at half maximum of the Eg1 peak a crystallite size of about 14 nm was determined.1 X-Ray photoelectron spectroscopy (XPS) spectra of the CVD films suggest pure TiO2 material, e.g., no nitrogen peaks were found (see Fig. S5 in the ESI†).
Fig. 3 SEM pictures of the light-scattering TiO2 films (a) prepared from CVD particles and (b) JGC_PST-400C. |
Fig. 4 The XRD pattern of the sintered CVD light-scattering film. |
Fig. 5 shows the total transmission spectra of the different TiO2 films on the FTO substrates. The transparent film shows a high transmittance of about 82%. Addition of the CVD scattering layer decreased the transmittance to 33% in the 600 to 700 nm wavelength range, while addition of the JGC layer lowered the transmittance to about 22%. Since TiO2 is not absorbing light at these wavelengths, the decrease in transmittance is directly attributed to the light-scattering properties of the CVD and JGC films. This is confirmed by the pure white appearance of these films. The total reflectance at these wavelengths is thus 67% and 78% for the CVD and JGC film, respectively.
Fig. 5 Total (specular + diffuse) transmission of T, T + CVD and T + JGC films on TEC15 substrates. |
Fig. 6 and Table 1 show the IV curves and characteristics for fabricated DSCs comparing the effect of the CVD paste and JGC paste as scattering layers. T, T + CVD and T + JGC are representative of DSCs with 1 layer of transparent Dyesol paste, 1 layer of Dyesol paste plus 1 layer of CVD paste, and 1 layer of Dyesol paste plus 1 layer of JGC_PST-400C scattering paste, respectively. Addition of a light-scattering layer improved the current density and the overall efficiency of the solar cells. The T + CVD has slightly better power conversion efficiency in comparison to T + JGC. The main difference between T + CVD and T + JGC is the better fill factor of the former. The short circuit current varies linearly with light intensity (see Fig. S8 of the ESI†), demonstrating that the mass transport of cobalt mediators in the electrolyte is not limited by film thickness or porosity.
Fig. 6 (a) Current–voltage characteristics of D35-sensitized solar cells with cobalt tris-bipyridine based electrolyte, with 5 μm transparent TiO2 (T), T plus 5 μm CVD-based TiO2 (T + CVD), and T plus 5 μm JGC-based TiO2. (b) IPCE spectra of the same solar cells. |
Fig. 6b shows the IPCE spectra of DSCs with and without scattering layers. The IPCE is increased by the presence of the light-scattering film, due to the increased overall thickness of the electrode as well as the significantly improved light scattering, which will reflect light back into the transparent TiO2 part of the electrode. In red region T + JGC has a slightly better IPCE than the CVD paste. Dye adsorption/desorption experiments showed a slightly higher amount of adsorbed dye for the T + JGC films (∼8% more than for T + CVD), see Table 1. The enhancement of the photocurrent due to the light-scattering layer was much more pronounced for solar cells sensitized with N719, which has a broader spectrum and a lower extinction coefficient than D35. See the ESI† section S3 for IV curves and further information.
Electron transport, recombination and accumulation in the different solar cell devices were investigated using light modulation techniques.22,23Fig. 7 shows the results of the electron transport time and the electron lifetime. The electron lifetime in devices with the CVD scattering paste was longer than in devices with the JGC films. At the open circuit voltage (870 mV) of the T + JGC cell, the electron lifetime was 11 milliseconds but in the T + CVD cell at the open circuit voltage (860 mV) it was 22 milliseconds. The longer electron lifetime of the T + CVD cell implies a slower recombination of electrons and is likely to contribute to a higher fill factor in the solar cell. Also the presence of larger pores in the CVD film can improve the diffusion of cobalt redox mediators inside the electrode leading to the enhancement of the fill factor of the T + CVD cell.
Fig. 7 (a) Electron lifetime as a function of the open-circuit potential. (b) Electron transport time as a function of the short-circuit photocurrent of the fabricated solar cells. |
As expected, the electron transport time is shortest in devices with the thinnest TiO2 film, i.e. device T. There were some differences between the T + JGC and T + CVD devices, where electron transport was 20% faster in T + JGC.
Charge extraction experiments were performed from open circuit and short circuit solar conditions to determine the amount of charge stored in the different TiO2 films, see Fig. 8. In device T the lowest amount of charge is found, which is expected on the basis of the thinnest TiO2 layer. Addition of a scattering layer doubles the accumulated charge at a given voltage (Fig. 8a) for the T + CVD device, while the increase in charge is less for the T + JGC device. We attribute the presence of a larger number of trap states in the CVD film to the polycrystalline nature of the large light-scattering aggregates, compared to the JGC film that has large single crystalline scattering particles.
Fig. 8 Extracted charge from (a) open circuit and (b) short circuit conditions. |
The larger amount of trap states in the CVD film is also evident from the extracted charge under short-circuit conditions (Fig. 8b), where the charge is about two times larger for T + CVD compared to T + JGC at the same current density.
The exponential behaviour of the charge versus open circuit voltage is consistent with an exponential distribution of trap states below the conduction band edge. For a trap distribution g(V) = g0exp(V/m), the density of trap states nt is given by: 7
(1) |
The recombination resistance and chemical capacitance of the cells was studied with electrochemical impedance spectroscopy (EIS) in the dark at an applied potential of 0.85 V. Fig. 9 shows the Bode plots of the measured data, along with the fitted curves. The high-frequency part is nearly identical for all devices. This part is attributed to the resistance and capacity of the counter electrode–electrolyte interface. The low-frequency part differs, and is dominated by the recombination resistance (Rrec) and chemical capacitance (CTiO2) of the porous TiO2 electrode. As expected Rrec is smaller for the thinnest electrode (T). The model developed by Fabregat-Santiago et al. was used for fitting and interpretation of the EIS data.29
Fig. 9 Real and imaginary part of the impedance as a function of the frequency for the different devices. The points show the measured data and the solid lines show the fitted curves. |
The fitted values presented in Table 3 show that both the recombination resistance and the chemical capacitance is higher in T + CVD than in the T + JGC device, resulting in a higher electron lifetime (= Rrec × CTiO2) in these cells. The determined electron lifetimes from the EIS data recorded in the dark were systematically higher than the lifetimes from the photoinduced voltage transient measurements. This can be attributed to the fact that under illumination there is additional recombination to oxidized dye molecules. Furthermore, the local concentration of Co(III) species will be higher than under dark conditions. Finally, there are slight differences in the Fermi-level in the porous TiO2 under the different conditions.
The trend in the capacitances agrees with that found with charge extraction data at the open circuit (Fig. 8a). The large CTiO2 for T + CVD is attributed to large density of bulk traps. These traps will not contribute to recombination. This is consistent with the fact that the recombination resistance of T + CVD is decreased compared to T, but much less than the decrease for T + JGC.
CVD grown TiO2 spheres are efficient for use as scattering structures in DSCs. The scattering properties of electrodes prepared using CVD-grown particle paste were found to be similar to that of JGC_PST-400C paste and resulted in similar solar cell efficiencies. The presence of bulk traps in the CVD particles, probably related to the small grains from which they are composed, resulted in a higher trap density in the electrodes with these particles, but this did not have a detrimental effect on the solar cell performance. Using the CVD scattering paste with cobalt-based electrolyte and D35 dye improved the efficiency of a 5 μm thick transparent mesoporous TiO2 from 4.4% to 5.4%. The CVD growth of micrometre TiO2 spheres is a very quick and low cost method for the fabrication of efficient scattering particles for DSCs.
The temperature for the deposition of the scattering layer was as low as 250 °C, which may be compatible with plastic substrates. This suggests that full CVD processes may be designed for the fabrication of the mesoporous photoanode of dye-sensitized solar cells.
Footnote |
† Electronic Supplementary Information (ESI) available: See DOI: 10.1039/c2ra21361c |
This journal is © The Royal Society of Chemistry 2012 |