N. C. Dyck and
G. P. Demopoulos*
Materials Engineering, McGill University, 3610 University Street, H3A 0C5, Montreal, QC, Canada. E-mail: george.demopoulos@mcgill.ca
First published on 9th October 2014
Near infrared-to-visible upconverting NaYF4: Yb3+,Er3+@TiO2 composites are engineered by controlled titanium dioxide coating of citrate-modified β-NaYF4:Yb3+,Er3+ sub-microcrystals of ∼300 nm in size and employed as light harvesting layers in dye-sensitized solar cell fabrication. The NIR-to-vis upconversion response of the as-prepared composites is enhanced by a 2 orders of magnitude increase in the upconversion luminescence as a result of annealing effects induced by the thermal processing steps involved in DSSC fabrication. In addition, the TiO2 coating suppresses the β to α phase transition in NaYF4 thereby maximizing the upconversion response. The as-prepared upconversion crystals as well as 2 types of composite materials are integrated as internal light-harvesting layers in dye-sensitized solar cells (DSSCs). The insertion of the as-prepared or TiO2 coated β-NaYF4:Yb3+,Er3 crystals with inadequate TiO2 thickness results in a decrease in photovoltaic performance due to increased charge recombination. By ensuring the β-NaYF4:Yb3+,Er3+ crystals are fully coated in TiO2 (“core@shell” configuration), their integration into the DSSC is optimized, resulting in 16% relative increase in power-conversion efficiency over the control devices without the NaYF4:Yb3+,Er3+@TiO2 composite internal layer. This increase, as determined by EIS and IPCE under 1 sun illumination, is attributed to the recovery of adequate charge recombination resistance and light harvesting enhancement via scattering.
Spectral engineering via upconversion is an approach that potentially can lead to increased efficiencies for these types of devices. Considering for example the standard dye-sensitized solar cell7 as case study, we note that the typical light absorbing dye, a ruthenium-based complex known as N-719, has a bandgap (HOMO/LUMO gap) of ∼1.6 eV. Most of the energy losses at this energy level come from sub-bandgap photons,12 making DSSCs sensitized with these types of dyes ideal candidates for upconversion. Upconversion is a process that converts two or more low energy photons into a single higher energy photon. While many upconversion mechanisms exist, energy transfer upconversion (ETU) via the Yb3+–Er3+ lanthanide pair is the most efficient.13 By far β-NaYF4 is the most efficient host material for Yb3+–Er3+ upconversion pair, with an optimal doping profile of 18% Yb3+ and 2% Er3+ replacing Y3+ in the host lattice.14 Recent work in our laboratory has demonstrated that via crystal engineering and annealing the upconversion efficiency of β-NaYF4:18% Yb3+,2% Er3+ can be enhanced as much as three orders of magnitude.15,16 Infrared-to-visible upconversion has attracted a great deal of attention with potential applications as biological markers,17–19 photonic devices,20–22 and solar cells.12,23–38
Recently, application of these materials as internal layers in DSSCs has been the subject of a number of studies18,23,27–33. Shan and Demopoulos investigated the effect of an internal LaF3:Yb3+,Er3+–TiO2 layer and demonstrated proof-of-concept NIR operation but ultimately saw drops in performance associated with charge recombination.23 Another study by Wang et al. used YOF:Yb3+,Er3+ upconversion materials in the TiO2 photoanode layer to increase overall conversion efficiency by a factor of 1.23.32 More recently, Zhang et al. fully replaced the TiO2 photoanode material with a NaYF4:Yb3+,Er3+@TiO2 core–shell composite material, attaining enhancement of the conversion efficiency by a factor of 1.23.33 Similarly, Liang et al. used double-shell β-NaYF4:Yb3+,Er3+/SiO2/TiO2 submicroplates and obtained 29.4% improvement in efficiency.28 In all these studies, however, insufficient effort has been placed on optimizing and understanding the best way to integrate the UC materials, and more importantly in analysing and quantifying the relative contributions of upconversion and internal light scattering in device's improved PV performance.
In the present study, β-NaYF4:Yb3+,Er3+ sub-microcrystals were prepared using a citrate-stabilized hydrothermal approach to control particle size and morphology. These sub-microcrystals were coated with TiO2 (“core@shell” configuration type) to aid with integration into DSSCs as a scattering layer in such way that prevents charge recombination losses. The upconversion properties of the composite material are investigated, with attention given to the effect of the DSSC thermal processing steps and the true magnitude of upconversion contribution to light harvesting enhancement in the solar cell devices. The cells prepared were tested using a host of photovoltaic and electrochemical testing procedures, including performance under simulated solar illumination as well as NIR laser illumination.
The photoanode was then sensitized using a 0.5 M N-719 dye solution in ethanol for 24 hours. In the case of cells prepared using the NaYF4:Yb3+,Er3+ and NaYF4:Yb3+,Er3+@TiO2 (DIW:
TEOT, 3
:
1) scattering pastes, a longer sensitization time of 72 hours was needed. A platinized counter-electrode was produced using pre-drilled FTO glass cleaned in a similar fashion to the photoanode. A drop (∼8 μL) of H2PtCl6 (50 mM in isopropanol) was spread across the surface of the counter electrode and annealed at 400 °C for 30 minutes.
The cells were then sealed in a sandwich type configuration using a hot melt ionomer film (Surlyn, 30 μm) using a cell assembly machine (E002-1233, Dyesol). The cells were injected with an I3−/I− liquid electrolyte (EL-HPE, Dyesol) through the pre-drilled holes, filling via capillary forces. The holes were sealed using a coverslip and the hot melt ioner film. Details relating to cell characterization can be found in the ESI.†
Ti(OR)4 + 3H2O → Ti(OR)(OH)3 + 3ROH |
Ti(OR)(OH)3 → TiO2·xH2O + (1 − x)H2O + ROH |
The rate of reaction and therefore the amount of TiO2 deposited as a shell depends on the molar ratio of DIW:
TEOT and therefore the extent of the hydrolysis reaction. TEM images are presented in Fig. 1, showing the difference between the 3
:
1 DIW
:
TEOT ratio (UC@T1) and the 6
:
1 DIW
:
TEOT ratio (UC@T2) materials.
![]() | ||
Fig. 1 TEM images of the UC@TiO2 composite materials showing UC@T1 particles produced with 3![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
As can be seen, the particles in UC@T1 are covered by an incomplete shell of amorphous TiO2, while the particles in UC@T2 are either completely covered by a thicker amorphous TiO2 shell or embedded into a larger amorphous TiO2 matrix on the order of ∼1 μm.
NaYF4:Yb3+,Er3+ upconversion materials convert infrared light to both red and green visible light. Both the shape and intensity of the upconversion spectrum is important for proper technological application. The phase of the UC material is also very important in determining both these aspects. The β-hexagonal phase of NaYF4 is a much better upconverter compared with its α-cubic counterpart.14 In addition, the β-phase emits predominantly green light while the α-phase emits predominantly red. This is an important aspect when engineering these materials for DSSCs as the dye absorbs green light more strongly than red and as a result, the DSSC performs better under green illumination.43,23 Since the red and green upconversion processes compete for upconverted energy, it is important to ensure that the amount of α-phase is minimized in order to maximize the coupling of the upconverted light to the solar cell through green upconversion. The upconversion fluorescence spectra of the as-prepared UC sub-microcrystals and the composite UC materials are shown in Fig. 2(a).
The efficiency of the upconversion process in bulk NaYF4:Yb3+,Er3+ is modest, and decreases further when the particle size of the crystals is smaller than a few microns.13 Other studies have increased the upconversion fluorescence intensity by using a core–shell approach to eliminate non-radiative pathways by separating the upconverting core from the quenching surface states.44,45 In the current study, a decrease is observed in Fig. 2(a) with addition of a TiO2 shell in the composite materials. UC@T1 has about 2/3 the upconversion intensity of the as-prepared UC material, while UC@T2 only has about 1/10 the upconversion intensity. This decrease is largely attributed to the decreasing absolute mass of UC crystals being excited due to the lower fraction of UC material in the composite powder as a whole. While it is possible to decouple these effects to see how the presence of the TiO2 affects upconversion in an equivalent number of excited crystals, in the context of solar cell integration, especially that of internal DSSC application, there are limitations placed on the amount of material that can be used within the cell itself based on the volume of the internal chamber, meaning that the primary concern should be the upconversion of an equivalent volume of scattering material rather than volume of upconverting material. Therefore examining the upconversion fluorescence of the composite powder as a whole is justified.
Another important consideration with regards to upconversion is the presence of organic molecules on the surface of the upconversion particles. The as-prepared particles were modified with citrate in order to control both the particle shape and size.15,16 The presence of organic molecules on the surface of upconversion particles has been shown to quench the upconversion fluorescence due to high-energy vibrational modes,46 and is something that should be avoided for solar cell integration. Annealing is a very effective way to increase the upconversion fluorescence of citrate-modified UC particles both through removal of organics as well as removal of internal defects in the crystal lattice that otherwise quench the upconverted energy states.46,16 In the case of the DSSCs assembled in this study, a standard DSSC annealing process (at 450 °C) was used to sinter the TiO2 sub-microcrystalline layer and remove the organic additives from the pastes used for screen-printing (Fig. S1, ESI†). Very advantageously, the same sintering process also serves to anneal the printed upconversion scattering layer, resulting in an impressive increase of the upconversion fluorescence of the UC@T2 material by over 2 orders of magnitude as can be seen in Fig. 2(b).
Although annealing clearly increases the upconversion fluorescence in the material, it can also cause partial transformation from the β-phase to the α-phase at temperatures as low as 400 °C.15 In order to investigate possible phase transformations, XRD was performed on the as-prepared and annealed UC and UC@T2 materials, as shown in Fig. 3.
From the XRD traces it can be seen that both UC and UC@T2 as-prepared materials are pure β-NaYF4. When sample UC is annealed at 450 °C, a partial transformation to α-phase is observed as indicated in Fig. 3, however for the UC@T2 material, we see that the phase transition is suppressed, possibly because the formation of crystalline TiO2 is more kinetically favoured and energy is initially directed towards the TiO2 material transformation. This preserves the β-phase NaYF4 of the core particles and helps to maintain a high green-to-red ratio.14 From the XRD results we also see appearance of crystalline TiO2 peaks (anatase) in the annealed UC@T2 material, indicating that the amorphous shell is converted to crystalline TiO2. TiO2 has a number of polymorphs, although anatase has been shown to have the best performance in DSSC devices due to its preferable bandgap and conduction band edge,47 and its presence is important for internal application of upconverter materials.33 In summary, the annealing process serves as a multifunctional tool, sintering the sub-microtitania-based photoanode, increasing the upconversion fluorescence, and converting the amorphous TiO2 to anatase, while the presence of the TiO2 shell helps to suppress the β-to-α phase transformation, maintaining a high green-to-red ratio.
Scatter Layer | Thickness (μm) | Jsc (mA cm−2) | Voc (V) | FF | η (%) | Dye Loading (10−7 mol cm−2) | |
---|---|---|---|---|---|---|---|
tra | scb | ||||||
a Transparent.b Scattering. | |||||||
None | 12 | — | 12.6 ± 0.6 | 0.7 ± 0.02 | 0.71 ± 0.03 | 6.21 ± 0.5 | 1.64 |
UC | 12 | 4 | 4.1 ± 0.4 | 0.76 ± 0.01 | 0.79 ± 0.01 | 2.48 ± 0.2 | 1.50 |
UC@T1 | 12 | 4 | 4.4 ± 0.5 | 0.76 ± 0.02 | 0.76 ± 0.05 | 2.5 ± 0.2 | 1.36 |
UC@T2 | 12 | 4 | 13.4 ± 0.2 | 0.73 ± 0.005 | 0.74 ± 0.01 | 7.22 ± 0.1 | 1.75 |
WER2-O | 12 | 4 | 14.6 ± 0.1 | 0.72 ± 0.01 | 0.73 ± 0.01 | 7.73 ± 0.1 | 1.70 |
The transparent control device without a scatter layer has an efficiency (η) of 6.21%, and it is immediately apparent that both the UC and UC@T1 scatter layers result in a severe drop in efficiency to 2.48% and 2.5%, respectively. This difference is due to a drop in the Jsc of the devices from 12.6 mA cm−2 in the control to 4.1 and 4.4 mA cm−2 in the UC and UC@T1 devices, respectively. Part of this drop can be attributed to a lower dye loading in the UC and UC@T1 devices, as can be seen in Table 1. Both the UC and UC@T1 devices required longer sensitization times, and despite this, the dye loading is still not as high as in the control devices, possibly because the additional layers hinder diffusion of the dye into the porous TiO2 layer. It is clear that the presence of these layers interferes with the sensitization process, and contributes to the decrease in the Jsc. On the other hand, the UC@T2 devices show efficiencies of 7.22%, a 16% relative increase in efficiency over the transparent control device without scatter layer. This increase comes from both a higher Jsc and Voc, both of which are characteristic of scattering layers. NIR performance of the device is demonstrated as proof-of-concept with 980 nm laser illumination and is shown in the inset of Fig. 5. The difference between the dark current and the current under illumination is small, however this can be partially attributed to the very small spot size of the laser. Although an increase in performance is seen in the UC@T2 devices, it does not match the performance of the commercially available scattering layer, WER2-O, which gave the highest efficiencies observed in this study at 7.73%.
To further investigate the effect of the UC@TiO2 scattering layers, the incident photon–electron conversion efficiency (IPCE) was measured, giving the spectral response of the cells. These results are shown in Fig. 6.
From the plot, it can be seen that the UC and UC@T1 devices have a much lower IPCE than that of the transparent control. The UC@T1 device has a slightly higher IPCE than the UC device, which is attributed to stronger scattering in the UC@T1 device. This was observed visually as a higher degree of transparency observed in the UC scatter layer devices (Fig. S2, ESI†). The refractive indices of the DSSC electrolyte, NaYF4, and TiO2 are about 1.4, 1.5, and 2.0, respectively. Since scattering relies on large changes in the material refractive index, TiO2 is a much more effective scattering agent than NaYF4 and the coated UC@T1 particles scatter more light than the uncoated UC particles.
UC@T2 provides higher IPCE response than the control device without scatter layer in the visible range and even extending slightly into the NIR from 300 nm to 800 nm. The IPCE of the commercial paste shows the highest response.
It is very important to note that the IPCE spectrum shows essentially no response near the absorption range of the Yb3+ ion as shown in the inset of Fig. 6. If a response is present, it does not exceed the noise of the measurement, corresponding to ±0.05%. Even if the contributions of upconversion were equivalent to the noise level, it would be negligibly small compared to the benefit coming from scattering. It should be noted that there was also no observable NIR response from the UC and UC@T1 devices (results not shown).
The efficiency, ηUC, of an n-photon upconversion process is as follows,48
ηUC ∝ irradiancen−1 | (1) |
While NIR laser illumination can been used to show proof-of-concept upconversion coupling in other studies27,28,33 as well as in the current study, its use should be avoided unless the technological application warrants the power densities used. In the case of photovoltaic application, in the absence of light concentration, this is clearly not the case. Other studies have additionally used an IR-pass filter for demonstration of NIR DSSC performance32,27 however this type of measurement is often unable to decouple the scattering effect from the upconversion if the cut-off wavelength of the filter is not selected properly. IPCE measurements under anticipated illumination provide the best assessment of the true contribution from upconversion. As has been seen from the results presented here, that contribution is negligible.
In order to extract the various operational parameters of the solar cell, an equivalent circuit originally developed by Fabregat-Santiago et al.,50,51 was used to fit the impedance data. The model, shown in Fig. 7 includes a number of useful performance parameters including the series resistance of the solar cell, Rs, the transport resistance, rtr, the chemical capacitance, Cμ, and the charge transfer resistance from the TiO2 to the electrolyte, rCT. For the fitting, all capacitors were treated as constant phase elements.
![]() | ||
Fig. 7 Equivalent circuit used for extraction of solar cell performance parameters, originally developed by Fabregat-Santiago et al.49,50 |
The chemical capacitance of the film, Cμ (= cμL) is plotted in Fig. 8(a) against the Fermi voltage (VF) of the film. VF is proportional to the Fermi level of the electrons in the sub-microcrystalline TiO2 and is calculated by removing the voltage drop due to series resistance (Rs) in the solar cell by using the following equation:50
VF = Vapp − VS = Vapp + jRS | (2) |
The primary pathway for loss of excited carriers is via electron recombination from the conduction band of the TiO2 to the electrolyte, which is dictated by the recombination resistance, Rrec (= rCT/L). The recombination resistance is proportional to the position of the TiO2 conduction band, which can be displaced through the use of different dyes, electrolytes, and pastes.50 As such, a voltage shift, ΔEC, must be introduced, with Rrec being plotted against an “equivalent conduction band” potential, Vecb = VF − ΔEC/q.
This is accomplished by shifting the Cμ plot so that all curves overlap in relation to a reference sample (in this case, the control without scatting layer), as is seen in Fig. 8(b).
Rrec can now be plotted vs. Vecb as shown in Fig. 8(c). The recombination resistance is similar in the control device without scattering layer, the UC2@T2, and the WER2-O devices, while it is lower in the UC and UC@T1 devices indicating increased recombination rates. These increased recombination rates are the primary reason for the Jsc decreases seen in the UC and UC@T1 devices (see Fig. 5 and Table 1). Similarly, the electron lifetime can be calculated using the equation, τn = RrecCμ and is plotted in Fig. 8(d), showing lower electron lifetimes in the UC and UC@T1 devices. These results agree with other similar studies that have suggested the presence of internal NaYF4 upconversion particles contributes to electron recombination and therefore a reduction in overall device efficiency.28 The UC device has the lowest efficiency (2.48%), which only marginally increases with the UC@T1 device (2.5%). Both the UC and UC@T1 devices have poor recombination resistance, and the UC@T1 device actually has lower electron lifetimes, so the marginal increase in efficiency must be primarily attributed to the difference coming from scattering as already discussed. It is only in the UC@T2 devices, when the UC particles are fully coated/embedded into TiO2, that the recombination resistance and electron lifetime of the control device is recovered.
Integration of the as-prepared UC sub-microcrystals as an internal layer causes a dramatic decrease in efficiency that is attributed to an increase in electron recombination from the TiO2 to the electrolyte caused by the presence of the NaYF4 upconversion crystals. Coating the upconversion particles with a layer of TiO2 (“core–shell” configuration) helps to eliminate the charge recombination as well as resulting in an overall efficiency of 7.22%, a relative increase of 16% over a control device without a UC@TiO2 internal layer (6.21%). This increase is attributed entirely to scattering, with no measurable IPCE response (>0.05%) in the 950–1000 nm region, where upconversion via the Yb3+–Er3+ couple should, in principle, be occurring. Having demonstrated the importance of core–shell sub-microstructuring for integrating upconversion crystals in hybrid solar cells, future efforts should be geared towards the deployment of UC material blends for wider and more efficient NIR light harvesting. Similarly, UC@TiO2 core–shell sub-microcrystals can be evaluated in combination with quantum-dot semiconducting or plasmonic metal light harvesting centres.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c4ra08775e |
This journal is © The Royal Society of Chemistry 2014 |