Cobalt Redox Mediators for Ruthenium-Based Dye-Sensitized Solar

Aug 25, 2011 - E-mail: [email protected]; Fax: +65 6776 3604; Tel: +65 6516 7118 (Q.W.). ... In situ near-IR transmittance measurements in conjunction ...
0 downloads 0 Views 3MB Size
ARTICLE pubs.acs.org/JPCC

Cobalt Redox Mediators for Ruthenium-Based Dye-Sensitized Solar Cells: A Combined Impedance Spectroscopy and Near-IR Transmittance Study Yeru Liu,† James R. Jennings,† Yao Huang,† Qing Wang,*,† Shaik M. Zakeeruddin,*,‡ and Michael Gr€atzel*,‡ †

Department of Materials Science and Engineering, Faculty of Engineering, NUSNNI-Nanocore, National University of Singapore, Singapore 117574 ‡ cole Polytechnique Laboratory of Photonics and Interfaces, Institute of Chemical Science and Engineering, Faculty of Basic Science, E Federale de Lausanne, CH-1015 Lausanne, Switzerland ABSTRACT: Dye-sensitized solar cells with power conversion efficiencies of up to 6.5% have been fabricated using a cobalt trisbipyridyl redox mediator with the cis-diisothiocyanato-(2,20 -bipyridyl-4,40 -dicarboxylic acid)-(2,20 -bipyridyl-4,40 -dinonyl) ruthenium(II) (Z907) sensitizer. This represents a significant improvement in efficiency compared with previous reports using ruthenium sensitizers. In situ near-IR transmittance measurements in conjunction with electrochemical impedance spectroscopy have been used to explain the difference in performance between DSCs using Z907 and another benchmark sensitizer cis-diisothiocyanato-bis(2,20 -bipyridyl-4,40 -dicarboxylic acid) ruthenium(II) bis(tetrabutylammonium) (N719). It is found that the small-perturbation electron diffusion length (Ln) is significantly longer in Z907 cells compared with that in N719 cells, which can explain most of the difference in performance. It is also shown that the longer Ln in Z907 cells is caused by inhibited recombination, as opposed to faster transport, and possible reasons for this are discussed. Our methodological approach is especially useful for accurately determining Ln when it is shorter than the TiO2 layer thickness, where standard dynamic techniques start to become unreliable.

I. INTRODUCTION Dye-sensitized solar cells (DSCs) have been recognized as a credible low-cost alternative to conventional silicon photovoltaics since the breakthrough in 1991,1 and recently power conversion efficiencies of up to 12% have been achieved.2 A key component of the DSC is the redox couple in the electrolyte. Although electrolytes of the state-of-the-art DSC have been highly optimized compared with those used in the original reports 20 years ago, they are still based on the same basic redox couple, that is, I3/I, which is known to have limitations.36 Recent reports about alternative inner sphere mediators such as disulfide/thiolate,7 2-mercapto-5-methyl-1,3,4-thiadiazole and its disulfide dimer (McMT/BMT),8 and tetramethyl formaminium disulfide/tetramethylthiourea (TMFDS2+/TMTU)9,10 have shown promising results compared with conventional I3/I. However, from a fundamental point of view, these systems are similar in complexity to the I3/I system because they involve the transfer of two electrons in their overall redox reactions. Compared with the mechanistic complications of multiple electron transfer redox reactions, one-electron outer-sphere redox shuttles such as ferrocene/ferrocenium, Ni(III)/(IV) bis(dicarbollide) or Co(II)/(III) complexes,1117 with the notable exception of TEMPO/TEMPO+,18 usually show simpler kinetics and may require a smaller energy expenditure for the dye regeneration process, reducing the loss of VOC. Among these alternative mediators, cobalt bipyridyl complexes have provided the best r 2011 American Chemical Society

performance to date. Besides the reduced visible light absorption and noncorrosion toward metals, the standard redox potential for the [Co(bpy)3]3+/2+ is more positive than that of I3/I by 210 mV (Figure.1a). In a DSC, this implies a reduction in the driving force for sensitizer regeneration and a possible increase in photovoltage and power conversion efficiency, provided that the rate of charge recombination for the [Co(bpy)3]3+/2+ redox couple is comparable to or slower than that for I3/I. According to previous work, Co(III)/Co(II) redox shuttles generally show inferior performance when used with N719, compared with I3/I. It was assumed that ion pairs can form between Co3+ and negatively charged dye molecules, which could block the approach of Co2+ thus retarding regeneration.13 It was also suggested that ion-pair formation may increase the probability of intercepting TiO2 electrons and consequently decrease the charge collection efficiency and limit the measured photocurrent and photovoltage.15 Bulky derivatives of Co(III)/Co(II) complexes can help to avoid the fast backward electron transfer from TiO2 to Co(III); however, in this case, mass transport limits the overall device performance.14,19 Recently, Ondersma et al. have used electrochemical impedance spectroscopy (EIS) to show that electron lifetimes and diffusion length decrease as the steric bulk of Received: May 15, 2011 Revised: August 14, 2011 Published: August 25, 2011 18847

dx.doi.org/10.1021/jp204519s | J. Phys. Chem. C 2011, 115, 18847–18855

The Journal of Physical Chemistry C

ARTICLE

the possible influence of dye structure on their results is also obscured by the fact that a coadsorbent (1-decylphosphonic acid) was used with Z907 but not with N719. We demonstrate that power conversion efficiencies as high as 6.5% can be achieved with the well-known Z907 sensitizer and the nonbulky cobalt tris-bipyridyl mediator, which is significantly higher than previous reports where ruthenium dyes are used with cobalt mediators. We have made extensive use of EIS to compare differences in recombination rate and electron diffusion length between Z907 and N719. (Molecular structures are shown in Figure 1b.) The results show that electrons in N719 cells have a very short diffusion length (Ln), although precise determination of Ln is difficult as fitting becomes unreliable for such short Ln. In situ near-IR transmittance measurements were used to remedy this situation by providing independent information about the density of electrons in the TiO2,22 in parallel with EIS measurements, to confirm the reliability of fitting results. The causes of the differences in charge collection between N719 and Z907 cells when employing [Co(bpy)3]3+/2+ are discussed, and it is shown that the shorter Ln arises as a result of faster recombination and not slower electron transport.

II. EXPERIMENTAL SECTION

Figure 1. (a) Energetic diagram and (b) structures of the reduced form of the redox mediator and dye molecules used in this work. Potentials in part a are obtained from ref 4 or estimated from cyclic voltammetry measurements.

cobalt complex mediators decreases.20 Feldt et al. recently found that by building steric hindrance into organic dye molecules performance can be greatly improved, even when the comparatively nonbulky redox mediator cobalt (III/II) tris(2,20 -bipyridine) ([Co(bpy)3]3+/2+) is used.16 Even more recently, Tsao et al. modified the organic sensitizer developed by Feldt et al. to improve its red absorption and achieved an impressive efficiency of 9.6%.21 It was shown that these recent breakthroughs in efficiency are due to reduced recombination brought about by increasing the steric bulk of the dye. In the present work, we further test the hypothesis that incorporating bulky groups into the dye molecule can reduce recombination, using Z907 and N719 as model dyes, one bearing a bulky nonyl chain and the other not. Although Feldt et al. have already shown that the Z907 dye exhibits better performance than N719 when used with [Co(bpy)3]3+/2+, a detailed characterization of these DSCs was not presented, and

Materials. The cobalt complexes [Co(bpy)3 ](PF 6 )2 , [Co(bpy)3](PF6)3 (bpy = 2,20 -bipyridine) were synthesized according to a previously reported procedure. Tetrabutyl ammonium perchlorate (TBAP), lithium perchlorate (LiClO4), 4-tertButylpyridine (tBP), propylmethylimidazolium iodide (PMII), iodine (I2), guanidine thiocyanate, and solvents were used as received from Sigma-Aldrich. Synthesis of Co(bpy)3[PF6]2 and Co(bpy)3[PF6]3 Complexes. CoCl2 3 6H2O (0.25 g) was dissolved in 5 mL of water, and to this a methanolic solution of 2,20 -bipyridine (0.55 g) was added dropwise while stirring. After stirring for a couple of minutes, potassium hexafluorophosphate (1.2 g) in water was added to form a precipitate. The precipitated complex was filtered, washed with water, and dried under vacuum to isolate Co(bpy)3[PF6]2. CoCl2 3 6H2O (0.25 g) was dissolved in 5 mL of water and to this a methanolic solution of 2,20 -bipyridine (0.55 g) was added dropwise while stirring. Methanolic bromine solution was then added in equivalent molar concentration, and stirring continued for a couple more minutes. The solution was filtered to remove any precipitate present. The solvent was evaporated with a Rotavapor under vacuum and redissolved in methanol solution (15 mL) then filtered. Potassium hexafluorophosphate (1.2 g) in water was added to the filtrate to form a precipitate. The precipitated complex was filtered, washed with water, and dried under vacuum to isolate Co(bpy)3[PF6]3. Elemental analysis calculated for C30H24CoF12N6P2: C, 44.08; H, 2.96; N, 10.28%. Found: C, 43.81; H, 2.97; N, 10.07%. Elemental analysis calculated for C30H24CoF18N6P3 3 H2O: C, 36.75; H, 2.67; N, 8.57%. Found: C, 36.78; H, 2.74; N, 8.49%. Solar Cell Fabrication. Fluorine-doped tin oxide coated glass (FTO, Pilkington TEC-15) was cleaned by sequential sonication in 5% Decon 90 solution, distilled water, propan-2-ol, and absolute ethanol. A thin compact layer of TiO2 was deposited onto all FTO substrates by spray pyrolysis.23,24 Transparent layers of nanocrystalline TiO2 (5 or 10 μm thick) were deposited by successive screen printing using a commercial TiO2 paste (Dyesol, DSL 18NR). Where required, a 2 μm layer of larger TiO2 particles (ca. 400 nm, WER-O paste, Dyesol) was then deposited on top of the 18848

dx.doi.org/10.1021/jp204519s |J. Phys. Chem. C 2011, 115, 18847–18855

The Journal of Physical Chemistry C transparent layer to increase light scattering in the TiO2 layer. The FTO/TiO2 electrodes were gradually heated to 450 °C, where they were held for 15 min before being heated to 500 °C for a further 15 min. Electrodes were treated in 40 mM TiCl4 solution at 70 °C for 30 min, followed by rinsing in distilled water and then heated at ca. 450 °C in a hot air stream for 30 min. The electrodes were immersed in a 0.25 mM solution of sensitizer (either Z907 or N719) in 1:1 acetonitrile/tert-butanol and left for 16 h before being removed and rinsed in acetonitrile immediately before cell fabrication. Cells were fabricated by sandwiching the sensitized TiO2 electrode and a thermally platinized FTO counter electrode together with a hot-melt polymer (Surlyn, DuPont). The interelectrode space (ca. 25 μm) was filled with electrolyte by vacuum backfilling. Holes were sealed using a small piece of hot-melt polymer and a microscope coverslide. The cobalt complex electrolyte contains 0.2 M [Co(bpy)3](PF6)2, 0.02 M [Co(bpy)3](PF6)3, 0.1 M LiClO4, and 0.5 M 4-tert-butylpyridine in acetonitrile. The reference iodide/triiodide electrolyte contains 0.6 M PMII, 0.03 M I2, 0.1 M guanidine thiocyanate, and 0.5 M 4-tert-butylpyridine in a mixture of acetonitrile and valeronitrile (volume ratio 85:15). The following cell naming scheme is used throughout the article: N719I (I3/I electrolyte, N719 dye), Z907-I (I3/I electrolyte, Z907 dye), N719-Co ([Co(bpy)3]3+/2+ electrolyte, N719 dye), Z907Co ([Co(bpy)3]3+/2+ electrolyte, Z907 dye). Characterization of Dye-Sensitized Solar Cells. The formal potentials of the redox couples were determined by cyclic voltammetry experiments using a three-electrode cell and an Autolab potentiostat/galvanostat controlled by the Nova 1.6 software package. The working electrode was a 1.6 mm diameter Pt electrode, the auxiliary electrode was a platinum wire, and the reference electrode was Ag/Ag+. The supporting electrolyte consisted of 0.1 M TBAP in acetonitrile. Currentvoltage characteristics under simulated AM 1.5 illumination were measured using a Keithley Source Meter and the PVIV software package (Newport). Simulated AM 1.5 illumination was provided by a Newport class A solar simulator, and light intensity was measured using a calibrated Si reference cell. Incident photon-to-collected electron conversion efficiency (IPCE) spectra were measured with a spectral resolution of ca. 5 nm using a 300 W xenon lamp and a grating monochromator equipped with order sorting filters (Newport/Oriel). The incident photon flux was determined using a calibrated silicon photodiode (Newport/ Oriel). Photocurrents were measured using an autoranging current amplifier (Newport/Oriel). Control of the monochromator and recording of photocurrent spectra were performed using a PC running the TRACQ Basic software (Newport). VOCintensity characteristics and EIS spectra were measured using an Autolab potentiostat/galvanostat and the Nova 1.6 software package. EIS experiments were performed with cells under illumination provided by a red LED (center wavelength λ = 625 nm) and biased at the VOC induced by the illumination. A 15 mV rms voltage perturbation was used, and the frequency range was 105 to 0.1 Hz. Different illumination intensities were achieved using neutral density filters mounted in an automated filter wheel system (Newport), which was controlled by the Nova 1.6 software. EIS spectra were also measured at various different bias voltages in the dark. Cells used for EIS measurements did not employ TiO2 light scattering layers to avoid complications in applying the usual transmission line model to the data. For EIS measurements made under illumination, 10 μm thick semitransparent TiO2 layers were used. For EIS measurements made in the dark, 5 μm thick semitransparent TiO2 layers were used.

ARTICLE

Before and after the EIS experiments, the transmission of NIR light (λ = 940 nm) through the cell was measured. The NIR light used for transmission measurements was produced using an LED (RS OPE5594). The intensity of the light was modulated at a frequency of 10 kHz, and the light intensity after passing through the cell was detected using a photodiode and a fast current amplifier (Stanford SR570) connected to a lock-in amplifier (Stanford SR850). The photodiode was masked with an IR highpass filter (Newport, cut-on wavelength 850 nm) to avoid saturating the detector with the red background illumination used in the EIS measurements. To remove drift effects partially, the signal from the detector behind the cell was normalized to the signal recorded by a second photodiode, which was positioned so that it sampled a fraction of the NIR light before it passed through the cell. Both the detector and the cell were masked appropriately to avoid problems arising from stray NIR light reaching the detector behind the cell without first passing through the active layer of the cell. Before and after each individual EIS measurement, cells were biased to 0 V in the dark for 5 min, and the recorded NIR signal was used as a baseline for calculation of the change in transmission. Absorbance changes were then calculated, neglecting possible changes in reflectivity/ scattering.

III. RESULTS AND DISCUSSION Photovoltaic Characteristics. Figure 2 and Table 1 shows the jV characteristics and IPCE spectra for DSCs employing either Z907 or N719 with [Co(bpy)3]3+/2+ electrolyte. For the I3/I reference electrolyte, slightly better performance was obtained when using the N719 dye compared with that obtained using the Z907 dye, which is consistent with the literature.25,26 For the [Co(bpy)3]3+/2+ electrolyte, device performance when using N719 is very poor,27,28 whereas device performance when using Z907 is much higher, as previously reported by Feldt et al.16 Unexpectedly, open-circuit photovoltage (VOC) for devices employing the [Co(bpy)3]3+/2+ mediator is not higher than that for devices using the I3/I mediator, despite the more positive redox potential of the [Co(bpy)3]3+/2+ mediator (Figure 1a). This indicates that recombination is faster when using the [Co(bpy)3]3+/2+ redox mediator either due to an increased recombination rate constant or due to a relatively low-lying conduction band edge, which increases electron concentration for any given cell voltage and thus the recombination flux. The IPCE spectra (measured with very low incident light intensity compared with the 1 Sun jV measurements) also indicate that Z907 is a far superior sensitizer compared with N719 when used with the [Co(bpy)3]3+/2+ mediator. This is mirrored in 1 Sun short-circuit photocurrent data and is presumably due to more significant charge collection losses occurring when using the [Co(bpy)3]3+/2+ redox mediator with N719. Charge Collection of DSCs Using Various Electrolytes and Sensitizers. To ascertain the reason for the lower photocurrent in cells using the [Co(bpy)3]3+/2+ redox couple, we have used impedance spectroscopy measurements to estimate the effective electron diffusion length under a variety of operating conditions. Figure 3a,b shows representative EIS spectra for cells employing different electrolytes and dyes without TiO2 scattering layers and with 10 μm thick semitransparent TiO2 layers. Spectra were recorded near open-circuit conditions with cells under illumination from 625 nm light of various intensities. (See the Figure caption for details.) 18849

dx.doi.org/10.1021/jp204519s |J. Phys. Chem. C 2011, 115, 18847–18855

The Journal of Physical Chemistry C

ARTICLE

Figure 2. (a) jV characteristics measured under simulated AM1.5 1 Sun illumination and (b) IPCE spectra for DSCs employing N719-I (black solid line), Z907-I (red dashed-dotted line), Z907-Co (blue dashed double-dotted line), and N719-Co (green dashed line). All cells employed 5 μm thick transparent TiO2 layers and 2 μm thick light scattering layers.

Table 1. Photovoltaic Parameters for DSCs Employing Different Electrolytes and Sensitizing Dyes, Measured under Simulated AM1.5 1 Sun Illuminationa jsc (mA cm2)

Voc (mV)

fill factor

η (%)

Z907-I

15.9

790

0.61

7.7

N719-I

16.8

758

0.63

8.0

Z907-Co

14.0

744

0.62

6.5

N719-Co

3.8

620

0.76

1.8

cell

All cells employed 5 μm thick transparent TiO2 layers and 2 μm thick light scattering layers. a

Data for all cells can be well-fitted by an equivalent circuit model (Figure 3) containing an impedance describing diffusion and recombination in the TiO2 layer, which is given by Bisquert as29 !1=2 "  # Rt Rct Rt 1=2 1=2 Z¼ coth ð1 þ iωRct Cμ Þ 1 þ iωRct Cμ Rct ð1Þ where Rt is the charge transport resistance, Rct is the charge transfer resistance, and Cμ is the chemical capacitance. The electron diffusion length can be defined by pffiffiffiffiffiffiffiffiffiffiffiffi ð2Þ Ln ¼ d Rct =Rt where d is the TiO2 layer thickness. Figure 3df shows the TiO2 conductivity (σ = d/RtA, where A is the projected area of the cell), Cμ and Ln, respectively, versus VOC. For the N719-Co cell, errors in fitted parameters are large in all cases because the impedance spectra resemble a Gerischer impedance, indicating that the electron diffusion length is shorter than the TiO2 layer thickness, whereas the other three combinations yielded spectra consistent with a diffusion length that is longer than the TiO2 layer thickness. Despite the possible errors in fitting the N719-Co data, we have included the results obtained from the best free fits to the spectra for comparison with the data obtained for the other cells, for which the parameters are believed to be reliable. Plots of σ or Cμ for each electrolyte lie close to one another, indicating that the conduction band edge lies at a similar energy for both sensitizers. The

shift between plots for the two different electrolytes is ca. 50 mV. This is smaller than the measured difference in electrolyte redox potential, implying that the band edge in the [Co(bpy)3]3+/2+ electrolyte lies about 160 mV more positive than that in the I3/ I electrolyte. This is very plausible because the [Co(bpy)3]3+/2+ electrolyte differs from the I3/I electrolyte in several respects; most importantly, it contains a higher concentration of Li+ ions, which are thought to shift the band edge to more positive potentials. For the Z907-I, N719-I, and Z907-Co cells, Ln was calculated from values of Rt with errors smaller than 10% and where Rt g RPt.27,30 For the N719-Co cell, Ln was found to be less than d. It has been suggested that for Ln < d it is not possible to fit EIS spectra to obtain a unique value of Ln.31 This is certainly true for extremely short Ln (i.e., Ln , d and thus Rt . Rct) because eq 1 simplifies to !1=2 Rt Rct ð3Þ Z¼ 1 þ iωRct Cμ In this case it is clear that an arbitrary number of combinations of Rct, Rt, and Cμ can fit simulated spectra provided that Rct , Rt and Cμ is chosen appropriately. It has been pointed out by Halme et al. that even for very short Ln the electron lifetime (τn = RctCμ) can still be obtained from EIS.31 It follows then that if Cμ can somehow be found independently, then Rct can be determined. Rt can then be determined by the DC limit of the photoelectrode impedance, which for the Gerischer-type impedance is given by (RctRt)1/2. Therefore, to ascertain whether the fitted results obtained for the N719-Co cell are meaningful, we have performed NIR transmission measurements with the aim of obtaining an independent estimate of the charge stored in the TiO2 layer as a function of open-circuit photovoltage and thus the capacitance. We should also note here that other techniques are suitable for determining electron diffusion length when it is short compared with the TiO2 layer thickness. These include back/ front IPCE measurements,3234 IMPS/IMVS,3537 or transients, either made at constant bias voltage or under different conditions if combined with appropriate charge extraction measurements. When using IMPS/IMVS or transients, full fits must be performed, taking proper account of the AC photocurrent generation efficiency and not simply relying on time 18850

dx.doi.org/10.1021/jp204519s |J. Phys. Chem. C 2011, 115, 18847–18855

The Journal of Physical Chemistry C

ARTICLE

Figure 3. (a) Typical EIS spectra for a DSC with the N719 sensitizer and an I3/I electrolyte, which exhibits Ln . d, measured at VOC = 0.507 V (red circles) and for a DSC with the N719 sensitizer and a [Co(bpy)3]3+/2+ electrolyte which exhibits Ln < d, measured at VOC = 0.513 V (black circles). (b) Enlarged version of spectra shown in part a. Red solid lines show the fits. (c) Equivalent circuit used to fit EIS spectra. Circuit elements that are not described in the main text are the substrate resistance (Rs), the Warburg impedance due to diffusion of redox species in the electrolyte (Zd), the resistance to charge transfer at the Pt-coated counter electrode (RPt), and the capacitance of the electrical double layer at the counter electrode (CPt). (dg) Dependence of TiO2 conductivity (d), TiO2 chemical capacitance (e), interfacial charge transfer resistance (f), and electron diffusion length (g) on VOC for various electrolyte/dye combinations. The legend for dg is: Z907-I, black circles; N719-I, red triangles; Z907-Co, green squares; N719-Co, yellow diamonds. 18851

dx.doi.org/10.1021/jp204519s |J. Phys. Chem. C 2011, 115, 18847–18855

The Journal of Physical Chemistry C

ARTICLE

Figure 4. (a) Plots of ΔA versus Qxs for electrolyte/dye combinations, where Ln g d. (b) Plots of ΔA versus VOC for various electrolyte/dye combinations. The legend is: Z907-I, black circles; N719-I, red triangles; Z907-Co, green squares; N719-Co, yellow diamonds.

constants, which could also lead to inaccurate results if Ln is shorter than the TiO2 layer thickness.38 Our approach begins by calibrating NIR measurements so that we can relate the absorbance change at 940 nm, ΔA, to the excess charge stored in the cell, Q xs, relative to equilibrium in the dark. We obtain Q xs from EIS data using Z vOC Q xs ðVOC Þ ¼ Cμ ðV Þ dV ð4Þ 0

The most obvious alternative technique that could be used to obtain this information, known as “charge extraction”, could potentially suffer from unknown recombination losses during extraction. This potential error source does not affect Q xs determined by EIS measurements (provided that the spectra are not Gerischer-type), which separately accounts for the efficiency with which charge is extracted from throughout the TiO2 layer, through the Rt and Rct parameters. Figure 4a shows plots of ΔA versus Qxs calculated in the manner outlined above. Data for three cells where the inequality Ln > d is satisfied (so that EIS fitting results should be valid) are shown in Figure 4a: Z907-I, N719-I, and Z907-Co. For all cells, plots are in very close coincidence, indicating that changing the dye and electrolyte does not significantly impact the proportionality between absorbance and stored charge. If it is assumed that the absorption change originating from charge accumulation is caused by electrons in the TiO2, then for a constant electron absorption cross section we have ΔA ¼ ϕn dΔn

ð5Þ

where ϕn is the electron absorption cross section, d is TiO2 layer thickness, and n is the electron concentration. If we fit a straight line forcing it to go through the origin, then we obtain ϕn = 2.1  1018 cm2 (neglecting the porosity of the TiO2 layer), which is of the order of the value reported by Nguyen et al. (5.4  1018 cm2), who also used the same commercial TiO2 paste (Dyesol, DSL18NR-T).22 On the basis of the results presented in Figure 4a, we will proceed to use ΔA directly as an indicator of charge stored in the cell. Figure 4b shows plots of ΔA versus VOC for all cells. The plots of ΔA versus VOC for the N719-Co and Z907-Co cells overlap very well, but, unfortunately only a very limited range of ΔA is covered by the measurements for the N719-Co cell. This problem is caused

Figure 5. Plots of Ln versus VOC for the N719-Co cell. Fits were performed either with the TiO2 capacitance allowed to vary freely (red triangles) or with the capacitance fixed (black circles) to the value predicted from Z907-Co fitting results for each VOC.

because larger charges/photovoltages cannot be reached for the N719-Co cell at the light intensities available experimentally because of the particularly fast charge recombination in this cell. Further measurements at lower light intensities are also not useful because of an unacceptably poor signal-to-noise ratio. We assume, based upon the data presented in Figure 4b, that the TiO2 band edge in the N719-Co cell lies at the same energy as that for the Z907-Co cell. This assumption is also broadly consistent with Figure 3d,e, which implies no large shift in band edge upon switching from Z907 to N719 for each electrolyte. (Note that for the I3/I data the EIS fitting is almost certainly reliable.) Assuming that the variation of capacitance with voltage for both cells is also the same, the relationship between capacitance and photovoltage for the Z907-Co cell can be used to predict, by interpolation or fitting, the capacitance of the N719Co cell at each photovoltage. Therefore, EIS spectra for the N719-Co cell can be fitted while holding the values for the capacitance fixed to those predicted by the Z907-Co fitting results for each VOC. Figure 5 shows the dependence of electron diffusion length on photovoltage for the N719-Co cell derived from the best free fits to the EIS spectra and from the fit obtained 18852

dx.doi.org/10.1021/jp204519s |J. Phys. Chem. C 2011, 115, 18847–18855

The Journal of Physical Chemistry C

ARTICLE

Figure 6. (a) Plots of Rct versus Cμ derived from EIS measurements performed in the dark for thin layer cells employing a [Co(bpy)3]3+/2+ electrolyte with 5 μm thick bare TiO2 (yellow tilted squares), N719-sensitized TiO2 (red triangles), Z907-sensitized TiO2 (black circles), and Z995-sensitized TiO2 (green squares) electrodes. (b) Plots of Cμ versus bias (corrected by removing iR drop across series resistance Rs for each combination in part a). (c) Structure of the Z995 dye.

while holding the capacitance fixed, as previously described. The error bars are calculated from the error estimates reported by the nonlinear least-squares fitting algorithm (ZView 3.1c). It is evident that the results are in excellent agreement. Error estimates for the free fit of EIS spectra are obviously much larger. They are, however, less than the parameter values themselves in all cases, and it would therefore appear that the results from the free fits are meaningful, even though Ln is only about one-third of the TiO2 layer thickness. Influence of Sensitizer Molecular Structure on Charge Collection. In principle, the low jsc of the N719-Co cells compared with Z907-Co cells could occur because of less efficient light absorption, electron injection, dye regeneration, or charge collection. It is highly unlikely that differences in light absorption can account for the ca. 73% decrease in jsc upon switching from Z907 to N719 because for the I3/I electrolyte there is only a small difference in jsc between the two dyes. This also implies that differences in electron injection efficiency between the two dyes are not large in the I3/I electrolyte; the high jsc found for the Z907 dye in the [Co(bpy)3]3+/2+ electrolyte then also implies reasonably efficient electron injection. It has been shown that with a similar cobalt redox couple, dye regeneration with N719 is much faster than that with Z907.15 Therefore, the most likely cause for the low jsc of the N719-Co

cells is less efficient charge collection. Plots of σ and Cμ versus VOC reveal that the TiO2 conduction band edge is likely to be at the same energy for the two different dyes and also that the free electron mobility (or diffusion coefficient) is unaffected by the choice of sensitizing dye. This means that the poor charge collection efficiency arises from faster charge recombination in the N719-Co system as compared with the Z907-Co system. To ascertain the reason for faster charge recombination in the N719-Co cells, we studied the charge transfer kinetics by EIS measurements in the dark to exclude possible effects caused by different dye regeneration kinetics. Testing devices were thin layer cells employing a [Co(bpy)3]3+/2+ electrolyte, Pt-coated counter electrodes, and 5 μm thick TiO2 working electrodes, which were either unmodified or had been sensitized with N719 or Z907. A similar ruthenium dye Ru(dcbpy)(dmbpy)2 (dcbpy = 2,20 -bipyridyl-4,40 -dicarboxylate, dmbpy = 4,40 -dimethyl 2,20 bipyridine, coded as Z995) has also been employed in this experiment for comparison. Thin TiO2 layers were chosen so that Ln > d under all conditions, making interpretation of EIS spectra simpler. Rct data are plotted versus Cμ in Figure 6a. We note that this approach is equivalent to plotting data versus TiO2 electron concentration or TiO2 density of states39 and that Cμ is a good indicator of the position of the electron quasi-Fermi level relative 18853

dx.doi.org/10.1021/jp204519s |J. Phys. Chem. C 2011, 115, 18847–18855

The Journal of Physical Chemistry C to the conduction band edge. Plots of Cμ versus bias (corrected by removing iR drop across the series resistance Rs) are also shown in Figure 6b to get a rough estimation of the conduction band edge difference between these devices. The conduction band of bare TiO2 seems to be very close to that of N719- or Z907-sensitized TiO2 devices in [Co(bpy)3]3+/2+ electrolyte, which is different for the case of a I3/I electrolyte, where the conduction band of TiO2 is more positive than that of sensitized TiO2 due to the adsorption of protons released from carboxylic acid groups.22 This is plausible because our electrolyte composition differs in many ways compared with typical I3/I electrolytes. From Figure 6a it is evident that at any given Cμ, Rct for Z907TiO2 is much bigger than bare TiO2 and N719- or Z995sensitized TiO2, which accounts for the greatly decreased charge recombination in Z907-Co cells. Rct values for N719-TiO2, Z995-TiO2, and bare TiO2 are quite close to each other at any given Cμ. Because Z995 and N719 are of similar size (Figure 6c) and are both smaller than Z907, which possesses two nonyl chains, it is expected that N719 and Z995 are worse than Z907 at sterically hindering the approach of [Co(bpy)3]3+ toward the surface of TiO2. However, this cannot explain why the plots of Rct versus Cμ for N719-TiO2 or Z995-TiO2 lie close to plots for bare TiO2, and the steric effect seems not applicable. A possible explanation for this is that the fully deprotonated N719 possesses four negative charges, and compared with Z907, which has only two negative charges, the Coulombic interaction with [Co(bpy)3]3+ is presumably much stronger, which leads to a higher possibility of electron interception. This may explain the similar Rct values of N719-TiO2 and bare TiO2 (which may be less negatively charged than N719-TiO2) and accounts for another possible reason for the much faster charge recombination in N719-Co cells compared with Z907-Co cells. However, the results obtained with Z995 seemed to contradict this explanation because it has a different charge to N719 yet yields almost identical plots of Rct versus Cμ. Another more likely explanation is simply that N719 and Z995, which are mostly conjugated molecules, are ineffective at attenuating electronic coupling between the TiO2 conduction band and the empty molecular orbitals on the cobalt acceptor species. The insulating nonyl chains on Z907 are likely to be very effective at attenuating coupling, thus resulting in increased recombination resistance.

IV. CONCLUSIONS DSCs employing [Co(bpy)3]3+/2+ redox shuttles show potential for replacement of the traditional I3/I couple even when used in conjunction with the well-known Z907 rutheniumbased dye, which incorporates nonyl chains. With an optimized electrolyte composition, a power conversion efficiency of 6.5% was achieved, which is the highest AM 1.5, 1 Sun efficiency reported to date for a ruthenium sensitizer and a cobalt redox mediator. EIS shows that the rate of recombination in this system is decreased by several orders of magnitudes compared with the system using the N719 dye. NIR transmittance measurements made in parallel with EIS measurements have enabled us to determine an accurate value for the diffusion length of electrons in the N719/[Co(bpy)3]3+/2+ system, which is much shorter than in the Z907 system. The presence of an insulating nonyl chain on Z907 appears to be the most likely reason for its superior performance in DSCs employing Co-based mediators, whereas simple considerations of molecular size or charge cannot adequately explain our data.

ARTICLE

Our methodological approach demonstrates that the additional information provided by NIR transmittance measurements made in parallel with EIS measurements enables accurate determination of electron diffusion length, even if it is much shorter than the TiO2 layer thickness. This may prove to be useful in accurately quantifying recombination losses in systems that are in the early stages of development and are far from being optimized, without the need to fabricate very thin TiO2 layers specially. Our results also show that high efficiencies with Cobased redox mediators can be obtained with ruthenium-based sensitizers, provided that appropriate steric hindrance is built into the dye molecule. We expect that other ruthenium-based sensitizers with appropriate bulky groups that are able to shield the TiO2 from electron acceptors may produce even higher power conversion efficiencies than the efficiency presently obtained with Z907. It may also be beneficial to increase the absorption coefficient of the dye, permitting thinner TiO2 layers to be used. These are more promising strategies than trying to reduce recombination by increasing the bulk of the cobalt complexes, which would lower their diffusion coefficients, thus limiting the cell performance as a result of mass transport overpotential.19

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected]; Fax: +65 6776 3604; Tel: +65 6516 7118 (Q.W.). E-mail: shaik.zakeer@epfl.ch (S.K.), michael. Gr€atzel@epfl.ch (M.G.); Fax: +41 21 6933112; Tel: +41 21 6933115.

’ ACKNOWLEDGMENT This work was financially supported by NUS startup grant no. R-284-000-064-133, URC grant no. R-284-000-068-112, and NRF CRP grant no. R-284-000-079-592. ’ REFERENCES (1) Oregan, B.; Gr€atzel, M. Nature 1991, 353, 737. (2) Gr€atzel, M. Acc. Chem. Res. 2009, 42, 1788. (3) Rowley, J. G.; Farnum, B. H.; Ardo, S.; Meyer, G. J. J. Phys. Chem. Lett. 2010, 1, 3132. (4) Boschloo, G.; Hagfeldt, A. Acc. Chem. Res. 2009, 42, 1819. (5) O’Regan, B. C.; Walley, K.; Juozapavicius, M.; Anderson, A.; Matar, F.; Ghaddar, T.; Zakeeruddin, S. M.; Klein, C.; Durrant, J. R. J. Am. Chem. Soc. 2009, 131, 3541. (6) Matar, F.; Ghaddar, T. H.; Walley, K.; DosSantos, T.; Durrant, J. R.; O’Regan, B. J. Mater. Chem. 2008, 18, 4246. (7) Wang, M. K.; Chamberland, N.; Breau, L.; Moser, J. E.; Humphry-Baker, R.; Marsan, B.; Zakeeruddin, S. M.; Gr€atzel, M. Nat. Chem. 2010, 2, 385. (8) Tian, H. N.; Jiang, X. A.; Yu, Z.; Kloo, L.; Hagfeldt, A.; Sun, L. C. Angew. Chem., Int. Ed. 2010, 49, 7328. (9) Li, D. M.; Li, H.; Luo, Y. H.; Li, K. X.; Meng, Q. B.; Armand, M.; Chen, L. Q. Adv. Funct. Mater. 2010, 20, 3358. (10) Liu, Y. R.; Jennings, J. R.; Parameswaran, M.; Wang, Q. Energy Environ. Sci. 2011, 4, 564. (11) Daeneke, T.; Kwon, T.-H.; Holmes, A. B.; Duffy, N. W.; Bach, U.; Spiccia, L. Nat Chem 2011, 3, 213. (12) Li, T. C.; Spokoyny, A. M.; She, C. X.; Farha, O. K.; Mirkin, C. A.; Marks, T. J.; Hupp, J. T. J. Am. Chem. Soc. 2010, 132, 4580. (13) Nusbaumer, H.; Moser, J. E.; Zakeeruddin, S. M.; Nazeeruddin, M. K.; Gr€atzel, M. J. Phys. Chem. B 2001, 105, 10461. 18854

dx.doi.org/10.1021/jp204519s |J. Phys. Chem. C 2011, 115, 18847–18855

The Journal of Physical Chemistry C

ARTICLE

(14) Sapp, S. A.; Elliott, C. M.; Contado, C.; Caramori, S.; Bignozzi, C. A. J. Am. Chem. Soc. 2002, 124, 11215. (15) Nusbaumer, H.; Zakeeruddin, S. M.; Moser, J. E.; Gr€atzel, M. Chem.—Eur. J. 2003, 9, 3756. (16) Feldt, S. M.; Gibson, E. A.; Gabrielsson, E.; Sun, L.; Boschloo, G.; Hagfeldt, A. J. Am. Chem. Soc. 2010, 132, 16714. (17) Cameron, P. J.; Peter, L. M.; Zakeeruddin, S. M.; Gr€atzel, M. Coord. Chem. Rev. 2004, 248, 1447. (18) Zhang, Z.; Chen, P.; Murakami, T. N.; Zakeeruddin, S. M.; Gr€atzel, M. Adv. Funct. Mater. 2008, 18, 341. (19) Nelson, J. J.; Amick, T. J.; Elliott, C. M. J. Phys. Chem. C 2008, 112, 18255. (20) Ondersma, J. W.; Hamann, T. W. J. Phys. Chem. C 2009, 114, 638. (21) Tsao, H. N.; Yi, C.; Moehl, T.; Yum, J.-H.; Zakeeruddin, S. M.; Nazeeruddin, M. K.; Gr€atzel, M. ChemSusChem 2011, 4, 591. (22) Nguyen, T. T. O.; Peter, L. M.; Wang, H. X. J. Phys. Chem. C 2009, 113, 8532. (23) Kavan, L.; Gr€atzel, M. Electrochim. Acta 1995, 40, 643. (24) Cameron, P. J.; Peter, L. M. J. Phys. Chem. B 2003, 107, 14394. (25) Zhang, Z. P.; Ito, S.; Moser, J. E.; Zakeeruddin, S. M.; Gr€atzel, M. ChemPhysChem 2009, 10, 1834. (26) Mori, S. N.; Kubo, W.; Kanzaki, T.; Masaki, N.; Wada, Y.; Yanagida, S. J. Phys. Chem. C 2007, 111, 3522. (27) Wang, H.; Nicholson, P. G.; Peter, L.; Zakeeruddin, S. M.; Gr€atzel, M. J. Phys. Chem. C 2010, 114, 14300. (28) Klahr, B. M.; Hamann, T. W. J. Phys. Chem. C 2009, 113, 14040. (29) Bisquert, J. J. Phys. Chem. B 2002, 106, 325. (30) Wang, H.; Peter, L. M. J. Phys. Chem. C 2009, 113, 18125. (31) Halme, J.; Vahermaa, P.; Miettunen, K.; Lund, P. Adv. Mater. 2010, 22, E210. (32) Halme, J.; Boschloo, G.; Hagfeldt, A.; Lund, P. J. Phys. Chem. C 2008, 112, 5623. (33) Barnes, P. R. F.; Anderson, A. Y.; Koops, S. E.; Durrant, J. R.; O’Regan, B. C. J. Phys. Chem. C 2009, 113, 1126. (34) Papageorgiou, N.; Gr€atzel, M.; Infelta, P. P. Sol. Energy Mater. Sol. Cells 1996, 44, 405. (35) Dloczik, L.; Ileperuma, O.; Lauermann, I.; Peter, L. M.; Ponomarev, E. A.; Redmond, G.; Shaw, N. J.; Uhlendorf, I. J. Phys. Chem. B 1997, 101, 10281. (36) Fisher, A. C.; Peter, L. M.; Ponomarev, E. A.; Walker, A. B.; Wijayantha, K. G. U. J. Phys. Chem. B 2000, 104, 949. (37) Jennings, J. R.; Ghicov, A.; Peter, L. M.; Schmuki, P.; Walker, A. B. J. Am. Chem. Soc. 2008, 130, 13364. (38) Halme, J. Phys. Chem. Chem. Phys. 2011, 13, 12435. (39) Jennings, J. R.; Liu, Y.; Wang, Q.; Zakeeruddin, S. M.; Gr€atzel, M. Phys. Chem. Chem. Phys. 2011, 13, 6637.

18855

dx.doi.org/10.1021/jp204519s |J. Phys. Chem. C 2011, 115, 18847–18855