Abstract
Carbon-modified titanium dioxide nanoparticles (C:TiO2 NPs) have been synthesized by ultrasonic nebulizer spray pyrolysis (USP) and pneumatic spray pyrolysis (PSP) techniques. HRTEM on the NPs shows difference in lattice spacing in the NP structures prepared by the two methods—2.02 Å for the USP NPs and an average of 3.74 Å for the PSP NPs. The most probable particle sizes are 3.11 nm and 5.5 nm, respectively. Raman spectroscopy supported by FTIR confirms the TiO2 polymorph to be anatase with the intense phonon frequency at 153 cm−1 blue-shifted from 141 cm−1 ascribed to both carbon doping and particle size. A modified phonon confinement model for nanoparticles has been used to extract phonon dispersion and other parameters for anatase for the first time. Electronic measurements show “negative conductance” at some critical bias voltage, which is characteristic of n-type conductivity in the carbon-doped TiO2 NPs as confirmed by the calculated areas under the I-V curves, a property suited for solar cell applications. Practical solar cells built from carbon-doped TiO2 electrodes show up to 1.5 times improvement in efficiency compared to pure TiO2 electrodes of similar construction.
1. Introduction: Efficiency Improvement Efforts on Dye Solar Cells
Various tables reporting figures of efficiency for different types of solar cells have been released several times by different groups. But the most up-to-date chart has been released by the National Renewable Energy Laboratory [1]. The chart shows that efficiency values for all solar cell types are steadily increasing from the earliest reported levels in the 1970s and 1980s. The highest values from 16–41% are found in multijunction concentrators GaAs with the latest (2007) figure of 41% reported by Boeing. Crystalline silicon cells range from 16–24%, multicrystalline Si from 12–18%, thin-film solar cells containing either Cu(In,Ga)Se2 or CdTe or amorphous SiH or nano-Si or micro-Si or poly-Si or multijunction Si ranging from 1–17%. The rapid increase in efficiency in nano-Si from about 8% in 1997 to 14% in 2001 demonstrates the importance of nanotechnologies in the improvement of the solar cells level of efficiency. Dye-sensitized solar cells (DSSCs) with a range of 5% in 1991 to 11% in 2006 have nearly the lowest range with the partners—the organic solar cells—at the lowest end of 1% in 2001 to 5.4% in 2006.
Solar cell types displaying the highest efficiency figures are difficult to produce due to material processing costs such as in silicon and GaAs alloys or precursor toxicity such as in selenides and sulfides and hence expensive [2]. These disadvantages are not an issue in low efficient types such as the organic ones and the dye-sensitized cells. For the reasons of nontoxicity and the possibility of mass production from low-cost starting materials, it is wise to study the materials properties for the emerging solar cell types especially the dye-sensitized one. The discovery of the pertinent properties that would drastically raise the efficiency of DSSCs will be very indispensable. The DSSCs, the Grätzel cells, consist of thin films of n-type titanium dioxide (TiO2) coated with a ruthenium metal organic dye [3]. They offer easy processing and can attain new roles over conventional silicon solar cells. Generally, the ways explored to improve the efficiency of DSSCs have focused on upgrading the components of the solar cell to enhance photovoltage () by varying (a) the supporting oxide, (b) the electrolyte which stabilizes the solar cell, and (c) the dyes which enhance the photocurrent [4, 5]. Carbon in DSC electrodes has gained attention recently, for instance, flexible C to complete with Pt:FTOs [6], C:TiO2 for photocatalysis [7–9].
We report on the optical, structural, and electronic properties of TiO2 NPs when doped with carbon (C:TiO2 NPs) synthesized by spray pyrolysis and how that C:TiO2 NPs increase the visible light absorption coefficient. These C:TiO2 NPs are tested in DSSCs.
2. Experimental: Preparation and Characterization of Carbon Modified TiO2
The proposed reaction sequence for the production of C:TiO2 NPs, similar to that proposed by Livage [10] in the production of vanadium ethoxide and Mwakikunga et al. [11] in the production of tungsten ethoxide, is as follows:
C:TiO2 NPs were synthesized by USP and PSP. The schematics for USP and PSP are given in Figure 1.
The USP technique involves focusing ultrasound waves of frequency producing capillary waves of frequency at the surface of the precursor liquid. This produces ultrafine droplets of the precursor liquid. The ultrasonic nebulizer uses an ultrasound generator to drive a piezoelectric crystal at a fixed frequency, F. The surface of the liquid will break down when the longitudinal waves propagate from the crystal to the liquid-air interface. The PSP involves blowing the precursor liquid through fine nozzles to produce a mist of the precursor liquid at audible sound and ultrasonic frequencies, . In both cases, the precursor droplets were generated in a chamber that was housing the nebulizer. The vapors generated are transported to a heated zone through a quartz tube onto glass substrate precoated with a transparent conducting layer of F:SnO2 placed inside a furnace. The spray deposition is carried out for an hour. The wavelength of the capillary waves is expressed as where σ and ρ are the surface tension and density of the liquid, which, for high enough intensity of the forcing sound, erupt into droplets of mean diameter and respectively. After decomposition and deposition in the pyrolysis process, the solid-state particles obtained have mean diameter that can be estimated respectively as and where is the concentration of the precursor, and are the molar masses of the precursor and deposited particles respectively; and and are the densities of the deposited particles and precursor liquid respectively [12–15].
An aliquot of 0.1 M precursor solution of titanium ethoxide was prepared by adding 11.8825 g of titanium tetrachloride (TiCl4) and 0.2162 g of glucose into a 250 mL volumetric flask in absolute ethanol (99.99%. Aldrich). The solution was then mixed vigorously till all the glucose particles were not visible. The precursor solution was then decanted into a chamber that was housing the ultrasonic nebulizer operating at a frequency of 1.67 MHz for USP and into tube connected to a pneumatic pump for PSP. F:SnO2 transparent glass substrates were first washed with detergent then thoroughly rinsed with distilled water, isopropanol, distilled water and finally acetone. The glass substrates were then dried under hot air to evaporate the acetone. Nano-particle deposition was done on F:SnO2 glass substrate at furnace temperatures of 450–455°C, at various substrate positions inside the furnace using argon as the carrier gas at a flow rate of 5.5 mL/min. Structural studies of the as deposited nano-particles were done using a Philips XPert powder X-ray diffractometer with a CuKα wavelength of 0.154184 nm. High resolution transmission electron micrography (HRTEM) performed on a Jeol JSM 2100 was used to determine structure—lattice parameters—of the C:TiO2 NPs; the HRTEM was also equipped with the energy dispersive X-ray spectroscopy (EDXS) setup using a Thermo Fisher Scientific detector cooled at liquid nitrogen temperature. Optical characterization (Diffuse reflectance spectra) was done using a Varian Cary Uv-Vis-IR spectrometer in the wavelength range of 250–2000 nm.
Raman spectroscopy (RS) was carried out using a Jobin-Yvon T64000 Raman spectrograph with a 514.5 nm line from an argon ion laser. The power of the laser at the sample at the post-annealed samples was small enough (0.384 mW) in order to minimise localised heating of the sample. The T64000 was operated in single spectrograph mode, with the 1800 lines/mm grating and three objective lenses on the microscope with the following magnifications: 20x, 50x and 100x.
Fourier Transform Infrared (FTIR) spectra of both pure and carbon modified titanium dioxide samples were carried out in the 350–4000 cm−1 frequency range. The infrared absorption spectra were recorded on Nicolet Nexus 870 system FTIR spectrometer at room temperature. Dried solid samples were pressed with KBr (FT-IR grade Aldrich) and pellets were scanned 34 times using transmission mode with a resolution of 4 cm−1.
A colinear four-point probe system from Cascade Microtech, Inc., Oregon, USA was used to perform conductance measurements on the C:TiO2 NPs. The equidistant tungsten carbide probes have a separation distance, a, of 0.127 cm and a probe radius of 0.005 cm. The Keithley 4200 semiconductor characterization system (SCS), equipped with four supply-and-measure units (SMUs) and a pre-amplifier, was used to perform high-precision direct-current characterization capable of probing-sensitive voltages and injecting currents ranging from 1 fA–1 mA with a resolution of 0.1 fA. In this work, a current was injected through two extreme probes through the sample. The potential difference was measured across the middle probes.
3. Structural Properties of the Carbon-Modified TiO2 Anatase
The lower magnification HRTEM showed that for the PSP prepared sample the particles were mostly spherical and had mean diameter of 5.5 nm. The same can be said of the USP prepared sample only the particle size is smaller (Figure 2). In this case, the mean diameter is 3.11 nm. The EDXS of these PSP synthesized C:TiO2 NPs (not shown) showed presence of Ti, O and C only signifying the purity of the obtained material. Cu peaks and some C intensity came from the carbon-holey copper grids used in these experiments.
TiO2 exists in the following known polymorphs: anatase (a = 3.784 Å, c = 9.514 Å, α = β = γ = 90° PDF 78–2486) with the strongest d-spacing of 3.52 Å [16], rutile (a = 4.593 Å, c = 2.958 Å, α = β = γ = 90° PDF# 89-8304) with a d-spacing of 3.25 Å [17] brookite (a = 9.174 Å, b = 5.449 Å, c = 5.138 Å, α = β = γ = 90° PDF# 76-1934) with a d-spacing of 3.51 Å [18] and hongquite (a = 4.173 α = β = γ = 90° PDF# 89-3660) with a d-spacing of 2.09 Å [19].
The higher magnification HRTEM clearly reveals that USP TiO2 NPs have a very small d-spacing compared to the normal anatase spacing of 0.352 nm. The d-spacing found in this sample is about 0.202 nm as shown Figure 3.
However, the PSP C:TiO2 NPs have a distribution of interplanar d-spacings whose distribution is given in Figure 4(b) indicating that some particles have their d-spacings less or more than 0.352 nm. In fact, most of the particles have most probably d-spacing of 0.374 nm in agreement with the anatase TiO2 as found by Bersani et al. [17]. It has been observed, especially in ZnO, that carbon doping, where C substitutes O, tends to reduce the d-spacing of the C:ZnO. It can, therefore, be concluded that those C:TiO2 NPs that show larger d-spacings than the anatase observed in PSP samples have interstitial carbon doping as presented in Figure 4. The lower d-spacings suggest substitutional C doping as seen in some particles in the PSP sample and in almost all the particles in the USP sample. We can, therefore, speculate that USP is a more efficient technique for incorporating carbon uniformly in the TiO2 matrix. On the other hand, such a large change in lattice spacing, compounded with the large distribution shown in Figure 4(b) cannot be entirely due to carbon doping. This can be attributed to the fact there are various types of crystallites in the intrinsic TiO2 material itself. Therefore, we can conclude this is a polycrystalline material.
(a)
(b)
(c)
The fast Fourier transform (FFT) of the images, synonymous to the selected area electron diffraction, displays the fact that only the main diffraction (reflection) plane of Miller indices (101) in TiO2 [17] is responsible for diffraction pattern. In the PSP sample, the particles with the same Miller indices are oriented in different direction. This is why their spots on the FFT image (Figure 4 inset a) give a circle of uniform radius, which is indexed to the (101) plane. However, in the USP sample, the particles tend to show similar orientation with one direction on the FFT image (Figure 3(b)).
4. Phonons in the Alloy
The FTIR spectra for pure unannealed TiO2 (Pure TiO2 spot 1 in Figure 5), pure annealed TiO2 at 500°C (pure TiO2 spot 2 in Figure 5), and carbon modified TiO2 samples display significant differences: the splitting of the main Ti-O phonon (at 664 cm−1) into two—one at 659 cm−1 and another at 536 cm−1 and the emergence of new absorption peaks in the range from 1400 to 1600 cm−1. The phonon splitting can be assigned to surface phonons due to small particle size whereas the new peaks in the range from 1400 to 1600 cm−1 can be due to the presence of carbon in the particle as shown in the enlarged inset of Figure 5. There is no indication of the structure of the carbon dopant with the present FTIR spectra. However, these peaks are ably captured in Raman spectroscopy presented herein.
TiO2 is known to exhibit six Raman active bands characteristic of anatase phase at 144, 197, 399, 515, 519, and 639 cm−1 with symmetries of , , , , , and respectively [20–22]. However, it has been shown [23] that, when TiO2 are coated with carbon as core-shell particles, the main mode of 144 cm−1 blue shifts to ~153 cm−1. This is what we see in our typical Raman spectrum for the PSP sample where this phonon mode has shifted to 152 cm−1. Raman spectroscopy in this case suggests carbon doping in our samples. We have extended our RS to include the carbon bands as shown in Figure 6(a). We observed the D-band of the disordered carbon at 1354 cm−1 and the perfect graphite peak at 1583 cm−1 [23, 24]. This is a clear indication of the presence of carbon in these samples, which supports the HRTEM results in this work. Raman spectroscopy of the C:TiO2 NPs also shows broadening and blue shift of the anatase phonon lineshape (Figure 6(b)) when compared to bulk TiO2 of mean size of 50 nm. This lineshape is also asymmetrical (Figure 6(c)) which is an indication that either there is optical phonon confinement or the inhomogeous laser heating effects on the TiO2 nanostructures [24, 25] or both.
(a)
(b)
(c)
In order to pursue phonon confinement, it is important to lay a brief background. In the 1980s, it became apparent that Raman spectral line shapes for microcrystals could not follow the perfect Gaussian profile [26]. It was then realised that there was need for a deeper understanding of the origin of this asymmetry. Spatial correlation modelling, as it was called, then became one way to extract information from RS data. It was stated in our previous work [27] that the justification for this new behavior is that, for a perfect (bulk) crystal of diameter having vibrational modes of momentum q, the Heisenberg’s uncertainty principle means that as (bulk) then δq → 0 (the “q = 0” selection rule applies). As the diameter, (nanometric) then the vibrational momentum, q →∞. In this case, the “q = 0” selection rule breaks down; a contribution from the q ≠ 0 phonons determined by the dispersion relation ωq is allowed [27]. This accounts for the asymmetric broadening of the peaks in a Raman spectrum. The Richter equation for confined phonons in spherical nanoparticles [26] was modified to include Gaussian distribution of the phonon momenta and particle size by Bhattacharyya and Samui [28] and given here as In the equation, is the full-width-at-half-maximum (FWHM) for bulk material Raman peak (for instance = 7.0 cm−1 for bulk TiO2 [29]), ωq is the phonon dispersion curve relation (PDR) for the material (e.g. A = 145 cm−1, a = 0.32 nm for the 144 cm−1 phonon branch). The Richter et al. model is valid for spherical nanocrystals of mean diameter for which the Fourier coefficient was assumed to be Gaussian of the form with a phonon amplitude at the boundary of . When geometry deviates from a sphere, then adjustment has to be carried out. And so, we fit the following equation to the 6 nm mean diameter nanoparticles:
The fitting was performed by activating a “Statistics NonlinearFit” routine in Mathematica 4.1 and is presented graphically in Figure 6(c). From this fit, the following parameters are extracted = 10−17, α = 3 π2 (~ 30), cm−1, B = 0.005 cm−2, and = 20 cm−1. Note that the value of 20 cm−1 for is 13 cm−1 above the value for bulk TiO2. Also the size distribution integral in (3) has been replaced by an expression + mω = 0.25 + to take into account the background noise inherent in the fluorescent nanostructures. These results show that the blue shift of the central Brillouin zone phonon frequency from 143.4 cm−1 for particles of mean diameter of 50 nm of pure TiO2 to 150.4 cm−1 for particles of mean diameter of 6 nm is not only due to reduction of particle size but also could be due to the presence of carbon in the TiO2 structure. We are yet to determine the singular contribution of carbon doping to this blue shift.
5. Optical Properties of the Carbon-Modified TiO2
It is well known that pure TiO2 can only absorb light below the wavelength of 387.5 nm because its bandgap is 3.2 eV. In order to extend the absorption region of TiO2 to use the majority of the sunlight of the solar spectrum, doping of titanium dioxide with nonmetal, transition metals, and composite transition metal oxide has been employed [30–35]. Doping of TiO2 inhibits recombination of photogenerated electrons and holes by increasing the charge separation and, therefore, enhances the efficiency of photon generation [30]. Carbon has been found to be a suitable dopant for TiO2, in applications that involve solar energy conversion. The presently carbon-doped TiO2 samples showed a strong and broad absorption band around 637 nm in the UV-Vis-IR spectra (not shown). The solar emission spectrum ranges from 250 nm to 2500 nm. The shift in the absorption from below 387.5 nm (pure TiO2) to 637 nm (C:TiO2 NPs) is a positive indication that the C:TiO2 samples will perform different in a solar cell.
6. Current-Voltage Output Characteristics of the Alloy
Figure 7 shows the I-V characteristics of the C:TiO2 NPs. The four-point probe I-V curve characteristic of F:SnO2 on the glass substrate is shown in Figure 7(b) with a voltage sweep from −0.4 mV to 0.4 mV. The I-V curve shows that there is a linear relationship between current and voltage where I ~ σ. V (σ being the conductance) and it also shows that the F:SnO2 on the glass substrate is indeed a transparent ohmic conductor. Typical I-V characteristics of C:TiO2 NPs deposited on FTO glass substrate are shown in Figure 6(a) with a current sweep in the range of −60 μA to +60 μA. The measurements from different spots gave slightly different values, which is caused by inhomogeneous deposition of the nanoparticles onto the glass substrates during spray pyrolysis. However, there is a general similarity of the I-V curves for the different spots.
On all spots of the C:TiO2 NPs film, the I-V curves clearly suggest that the NPs are semiconducting with a bandgap () determined from the width of the I-V curve when current is nearly zero. The bandgap here expressed in volts is found to be about 4 × 10−4 V. Extracalibration is required to convert this reading into electron volts (eV). This is beyond the scope of this paper. However, this value could be comparable to the band gap of TiO2 of 3.2 eV found by other methods. The deviation of the from 3.2 eV should be proportional to the amount of carbon doping and also particle size.
Also from the I-V curves one could speculate the type of conductivity one gets in the C:TiO2 NPs. TiO2 is naturally n-type semiconductor. Doping with carbon is expected to increase the electron density and make it more n-type.
The I-V characteristics in the (−x,−y) quadrant of Figure 7 clearly indicates a negative slope which suggests “negative conductance.” This apparent negative conductance could be due to the so-called Gunn effect which was discovered in III–V compounds such as GaAs and InP and reported by Gunn in 1963 [33]. Gunn saw that when an electric field reaches a threshold value in some materials the mobility of electrons decreases as the electric field is increased further, thereby producing negative resistance. Other authors such as Shikwambana et al. [36] and Mwakikunga et al. [37, 38] have reported S-shaped I-V curves in VO2 which show negative conductance. At a critical current, a core of the low resistivity phase was seen to be formed and this was attributed to the falling of the voltage across the device while the current increased to a value dependent on a specific load line. With further increases in the circuit current, the voltage across the device was seen to decrease slightly. Also the role of doping on the negativity of the slope in the I-V curves have been reported in boron-doped diamond [38] pointing this phenomenon to the n-type conductivity of these type of diamonds.
Furthermore, by inspection, the I-V curves are not symmetrical about the V = 0 point but rather shifted to negative voltages such that if one fits some analytical function such as Shockley’s diode equation among others and calculates the area under the I-V curves in the negative voltage and positive voltage sections, one can state that The integral of calculates the total electrical energy dissipated on charge carriers. The principle in (4) states that if more electrical energy is spent on driving the negative charge carriers in a semiconductor then such a material is intrinsically n-type and vice versa. From the curves in Figure 7, 20 nW are spent on electrons whereas only 5 nW are spent on the holes. We sufficiently concluded that our carbon-doped TiO2 NPs are indeed n-type which is a necessary precursor for planned DSSC electrode.
7. TiO2 Electrodes of Carbon-Dopant Concentration Used in the Dye-Sensitised Solar Cells
With the carbon doped TiO2 electrode results, we continued to build dye-sensitized solar cells containing TiO2 electrodes of varying carbon-dopant concentrations. This concentration is calculated from the amount of sucrose in mg for every litre of the total volume of the precursor material. The solar cell area was kept constant at 1.8 cm2. The current-voltage characteristics of each individual cell are presented in Figure 8(a). Efficiency values were calculated from open-circuit voltage values and short-circuit current obtained from these current-voltage output characteristics. The efficiency values at varying carbon concentration are given in Figure 8(b). The current efficiency values (~0.3–1.1%) are admittedly much lower than those obtained for dye sensitized solar cells (~5–11%). This can be improved by employing more careful fabrication processes. However, the current results, at the laboratory scale, show that carbon concentration does increase the efficiency of the dye-sensitized solar cell by about 1.5 times over the solar cell with prestine TiO2 layer.
(a)
(b)
8. Conclusion
We have successfully synthesized C:TiO2 NPs using ultrasonic nebulizer spray pyrolysis (USP) and pneumatic spray pyrolysis (PSP) techniques. These spherical nanoparticles have a modal particles size of 3.11 nm and 5.5 nm, respectively. The lattice fringe separations are 2.017Å for USP and 3.74 Å for PSP. This suggests substitutional carbon doping for USP and a mixture of substitutional and interstitial carbon doping for PSP. In both processes, the TiO2 structure changes with the most predominant reflection plane of (101) of the TiO2 observed. The Raman active phonon of 141 cm−1 from anatase is shifted to 150.4 cm−1 due to both size effect and carbon doping in agreement with most previous findings. The carbon doping enhanced the TiO2’s n-type conductivity with the I-V characteristics manifesting “negative conductance” especially when negative voltage is used. The n-type conductivity has also been confirmed by calculating areas under the I-V curves. The C:TiO2 NPs produced in this study show properties that would not only make the materials suitable for photocatalytic activity, but also for electrodes in Grätzel-type of solar cells. Efficiency values for solar cells of varying carbon concentrations in their TiO2 electrodes show up to 1.5 times improvement over the pure TiO2 electrode-based solar cells.
Acknowledgments
Financial support from our sponsors, SANERI, ESKOM, and Govan Mbeki R&D Centre at the University of Fort Hare, is acknowledged. The authors are thankful for infrastructural support from Ithemba LABS (Gauteng,) and DST/CSIR National Centre for Nano-Structured Materials. Additional financial support from the India-Brazil-South Africa (IBSA) programme is also acknowledged.