Integration of CdSe/CdSexTe1−x Type-II Heterojunction Nanorods into Hierarchically Porous TiO2 Electrode for Efficient Solar Energy Conversion

Semiconductor sensitized solar cells, a promising candidate for next-generation photovoltaics, have seen notable progress using 0-D quantum dots as light harvesting materials. Integration of higher-dimensional nanostructures and their multi-composition variants into sensitized solar cells is, however, still not fully investigated despite their unique features potentially beneficial for improving performance. Herein, CdSe/CdSexTe1−x type-II heterojunction nanorods are utilized as novel light harvesters for sensitized solar cells for the first time. The CdSe/CdSexTe1−x heterojunction-nanorod sensitized solar cell exhibits ~33% improvement in the power conversion efficiency compared to its single-component counterpart, resulting from superior optoelectronic properties of the type-II heterostructure and 1-octanethiol ligands aiding facile electron extraction at the heterojunction nanorod-TiO2 interface. Additional ~32% enhancement in power conversion efficiency is achieved by introducing percolation channels of large pores in the mesoporous TiO2 electrode, which allow 1-D sensitizers to infiltrate the entire depth of electrode. These strategies combined together lead to 3.02% power conversion efficiency, which is one of the highest values among sensitized solar cells utilizing 1-D nanostructures as sensitizer materials.


Optical Properties of NRs and HNRs. Absorption and photoluminescence (PL) spectra of NRs and
HNRs used in this study are shown in Fig. 1. Each sample is synthesized from CdSe NR seeds, with the second components grown at the tips. TEM images of NRs and HNRs are presented in Fig. 2, showing that HNRs synthesized from ~15 nm long CdSe NR seeds are ~25 nm long in average. The absorption spectrum for CdSe-only NRs shows a peak near 600 nm with a sharp PL peak at 615 nm. When CdTe is introduced as the second component, an absorption shoulder appears at 650 nm due to the smaller bandgap of CdTe. The absorption tail extends beyond 700 nm as a result of the charge transfer transition from the CdTe valence band to the CdSe conduction band. Recombination across the CdSe/CdTe interface occurs at energies lower than the bandgap of either component, shifting the PL peak to ~800 nm and broadening it considerably. Alloyed CdSe 0.4 Te 0.6 also forms a type-II junction with CdSe, therefore many of the absorption and PL features for these HNRs are similar. The difference between the PL peak position of the HNRs and that of the CdSe seeds, measured to be 0.37 eV, was used to determine the concentration of Te in the alloy following the calibration introduced in our previous work 34, 35 . The CdSe/CdSe 0.4 Te 0. 6 HNRs recapped with 1-octanethiol retain the absorption features seen with the native ligands, but have higher PL intensity and better stability in air, similar to the 1-octanethiol capped CdSe/ CdTe HNRs we reported recently 36 . For every heterostructured sample, the absorption spectrum extends beyond 700 nm, and should allow a greater portion of the incident solar spectrum to be collected in a photovoltaic device compared to single-component CdSe.

Photovoltaic Properties of NR-and HNR-Sensitized PV Devices. Photocurrent-voltage (J-V)
characteristics, absorbance spectra of sensitized films, incident photon-to-current efficiency (IPCE), and absorbed photon-to-current efficiency (APCE) spectra for various NR-and HNR-sensitized PV devices are presented in Fig. 3. Photovoltaic parameters of the devices are given in Table 1. Every device exhibits similar open-circuit voltages (V oc ) and fill factors (FF), but short-circuit current density (J sc ) values show remarkable differences. As the time constant of the electron transfer from CdTe (or CdSe 0.4 Te 0.6 ) tips to CdSe in HNRs (<400 fs) is much shorter than that of the electron transfer from CdTe to TiO 2 (~1 ns) 34,36,37 , we believe that most of the photocurrent would come from the electron transfer at the CdSe-TiO 2 interface (for CdSe NRs and HNRs). If a major portion of HNRs are anchored to TiO 2 only by their tips, electron extraction at the CdTe/TiO 2 interface should compete with more efficient electron Scientific RepoRts | 5:17472 | DOI: 10.1038/srep17472 transfer to CdSe. Moreover, electron transfer from CdSe to CdTe is not favored considering the band diagram in Fig. 1b, leading to poor fill factor as we previously reported in the case of organic-inorganic hybrid solar-cell structure utilizing curved CdSe/CdTe HNRs as inorganic light harvesters 38 . Such a loss of fill factor was not observed in HNR-sensitized PV devices, and therefore we believe that most of HNRs incorporated in the mp-TiO 2 frame have some portion of CdSe directly anchored to the TiO 2 nanoparticles. The TEM images of HNRs anchored on mp-TiO 2 in Fig. 2f partially support this argument.
The TOPO-CdSe/CdTe HNR device delivers slightly higher J sc compared to the TOPO-CdSe device. Considering that the photocurrent generated from single-component CdTe sensitizers is generally far lower than the single-component CdSe sensitizers presumably due to the recombination loss from the charge-carrier trapping [38][39][40][41] , higher J sc of the TOPO-CdSe/CdTe HNRs than that of the TOPO-CdSe NRs (with the same dimension of ~25 nm) cannot be explained simply by summing up the photocurrent generated from individual CdSe and CdTe components in HNRs. Due to the unique feature of type-II band offset in HNRs, where photoexcited electron-hole pairs are innately separated, the electron extraction process at the TiO 2 -sensitizer interface becomes more efficient with HNRs than with the single-component NRs 39 . This explains the increase of IPCE over the entire wavelength region for the TOPO-CdSe/CdTe HNR-SSCs compared to the case of TOPO-CdSe single-component NRs working as light harvesters. Furthermore, the type-II interface of TOPO-CdSe/CdTe HNR leads to a charge-separated state (CSS) absorption 39 , enabling the utilization of less energetic photons close to the near-infrared region. The effect of CSS absorption inherent for the HNR sensitizers is reflected in the extended absorption tail of the sensitized films ( Fig. 3b) compared to that of the TOPO-CdSe NR device.
In terms of the performance of TOPO-CdSe/CdTe HNR device, the stability of CdTe component in polysulfide electrolyte is a major concern, and therefore should be discussed. It has been well known that in the case of CdTe-sensitized SSCs using polysulfide electrolyte, anodic corrosion due to the ineffective hole scavenging from CdTe leads to the degradation of CdTe and low photovoltaic performance 42,43 . This degradation can be lessened by forming a semiconductor shell preventing CdTe cores from directly facing the polysulfide electrolyte 25 . Even though the CdTe component of the CdSe/CdTe HNRs does not have such a core-shell structure, our devices are finally treated to have ZnS passivation layer and therefore CdSe/CdTe HNR devices are not much affected by the anodic corrosion and exhibit reasonable performance.
In the case of TOPO-CdSe/CdSe 0.4 Te 0.6 HNR device, ~11% extra enhancement of J sc compared to that of the TOPO-CdSe/CdTe NR device is observed. CdSe/CdSe 0.4 Te 0.6 HNRs allow faster charge separation compared to the CdSe/CdTe HNRs and have lower valence band position in the alloyed tip 34 , which might lead to the J sc improvement by aiding facile charge extraction to TiO 2 . Furthermore, in the case of HNRs with alloyed tips, enhanced chemical stability of alloyed tips is also thought to contribute to achieve the improved photocurrent 39 . A noticeable further enhancement in the photocurrent density is observed by exchanging the surface ligands on the CdSe/CdSe 0.4 Te 0.6 HNRs, from native TOPO, ODPA, and TOP (simply referred to here as TOPO) ligands to 1-octanethiol (OT). As the absorption spectra obtained from mp-TiO 2 films sensitized with TOPO-CdSe/CdSe 0.4 Te 0.6 and OT-CdSe/CdSe 0.4 Te 0.6 are quite similar (Fig. 3b), the J sc upsurge may be attributed to the enhancement of electron injection/collection efficiency rather than the difference in the light harvesting from both sensitizers. Such an interpretation is further supported by the trend shown in the APCE spectra ( Fig. 3d): APCE is represented as the product of charge collection efficiency (η col ) and charge injection efficiency (η inj ), which can be calculated from the following relation (equation (1)) 44 : where the term LHE represents light harvesting efficiency. By comparing Fig. 3c,d, it is found that the APCE onset of the OT-CdSe/CdSe 0.4 Te 0.6 device is extended to ~760 nm compared to that of the TOPO-capped HNR devices. This result is in line with the previous report of the photocurrent spectra extended over ~700 nm for the 1-octanethiol capped CdSe/CdTe HNRs 36 . Moreover, the difference of IPCE among the TOPO-capped sensitizer devices almost vanishes from the ~580-nm wavelength in the APCE spectra, showing that the charge collection and injection efficiency of the TOPO-capped samples are quite similar in this wavelength region. However, APCE near ~480 nm corresponding to the X 2 excitonic band of CdSe still shows composition dependence, which we might attribute to the aforementioned effect from the faster charge-separation kinetics of type-II band offset and higher photocurrent obtainable from alloyed tips than CdTe tips. However, clear interpretation for this phenomenon is not established yet and further study is needed.
It is readily deduced that η col can be improved from the 1-octanethiol capping, as it has been well known that the thiol recapping of CdTe reduces the surface defect sites responsible for any unwanted recombination of photo-generated charge carriers 36,40,41 . This feature is also found in Fig. 1a showing higher PL intensity of OT-CdSe/CdSe 0.4 Te 0.6 than TOPO-CdSe/CdSe 0.4 Te 0.6 , which is presumably resulted from the suppressed recombination by 1-octanethiol capping. Improvement of carrier collection from enhanced electric-field throughout the HNR absorber layer is another possible factor leading to the improved J sc , but it may hardly be the case in our devices considering photocurrent saturation under the reverse-biased condition in the J-V curve (Fig. 3a) and our sensitized solar cell structure utilizing monolayer HNRs as light absorber layer 45 .
However, the difference of η inj from 1-octanethiol recapping is not self-evident and further investigation on the TiO 2 -OT-HNR interface is needed. To elucidate the interface properties, electrochemical impedance spectroscopy (EIS) was carried out 46,47 . The mp-TiO 2 electrodes sensitized with TOPO-capped CdSe/CdSe 0.4 Te 0.6 HNR and OT-capped CdSe/CdSe 0.4 Te 0.6 HNR sensitizers were chosen for the EIS measurement, to examine the capping ligand effect by excluding any side effects from the HNR composition and different loading amount of HNRs on mp-TiO 2 electrodes. Figure 4 presents the   52 , faster electron injection is predicted for this case as the CBM offset serves as the driving force for the electron transfer at the TiO 2 -CdSe interface [53][54][55] . Less energetic electrons generated in the HNRs may become extractable by this larger injection driving force, which also explains the origin of red-shifted APCE onset for the OT-CdSe/CdSe 0.4 Te 0.6 device. The higher recombination resistance is believed to result from the reduced recombination by thiol-capping, as explained in the previous section. Therefore, it is concluded that both η col and η inj (equation (1)) are improved through the 1-octanethiol recapping, and the J sc increase from the OT-CdSe/CdSe 0.4 Te 0.6 device with the red-shifted IPCE onset compared to the TOPO-CdSe/CdSe 0.4 Te 0.6 can happen without remarkable increase of the light absorption. The OT-CdSe/CdSe 0.4 Te 0.6 HNR-SSC shows J sc of 6.238 ± 0.212 mA cm −2 with 2.202 ± 0.131 overall PCE, approximately 33% improvement over the single-component CdSe NR devices with native ligands.

The Effect of Polystyrene Bead-Induced Percolating Pores on the PV Performance.
Even though the power conversion efficiency of the HNR device is much enhanced by utilizing both the type-II HNR sensitizers and ligand exchange, utilization of such 1-D long nanostructures as light harvesters for SSCs is still limited by their spatial incompatibility with the TiO 2 nanoparticle-based porous photoelectrodes. This type of electrode commonly provides ~20-nm sized pores (BET scale) 56 , which is smaller than the length of these 1-D nanostructures (~25 nm long). However, strategies to circumvent such limitations have been investigated only by a few researchers 28,57 .
To render the internal pore structures of mp-TiO 2 electrode more suitable for the infiltration of 1-D sensitizers, we utilized polystyrene (PS) microbeads as sacrificial additives for the conventional TiO 2 paste. Figures 5 and 6b show cross-sectional SEM images and EDS profiles of mp-TiO 2 electrodes derived from the pastes of different TiO 2 :PS weight ratios. For the electrode without PS beads, Cd/Ti ratio is not homogeneous throughout the electrode cross-section, and abruptly decreases at the ~2-μ m depth, suggesting that ~20-nm sized pores in mp-TiO 2 electrodes are easily clogged by the NR or HNR sensitizers. However, when the PS beads are incorporated into the TiO 2 pastes and calcined to leave large pores (Fig. 7a), the Cd/Ti ratio at every given depth of the electrode cross-section is highly increased, and the distribution of Cd/Ti values becomes much more homogeneous throughout the entire electrode depth. Figure 6a shows the J-V characteristics of HNR devices with different TiO 2 to PS ratios, the photovoltaic parameters of which are presented in Table 2. Incorporation of PS beads leads to a significant increase in both J sc and PCE. From these results, we presume that the incorporation of PS beads leads to It is predicted that with ~32.5% volumetric occupation of 200-nm sized PS beads in the TiO 2 nanoparticle electrode, percolation of beads extends throughout the entire electrode depth (8 μ m). Assuming the porosity of mp-TiO 2 films to be in the range of 0.5-0.6 56 , the PS beads account for the 27.5-32.1% of the total volume in the electrodes with TiO 2 :PS = 4:1 after calcination, which is close to 32.5%. Therefore, the homogeneous Cd/Ti ratio throughout the electrode depth and the improvement of PV performances from the PS-derived hierarchically porous TiO 2 electrodes can be explained by the percolation of large pores and subsequently facilitated infiltration of HNRs. Furthermore, dS/dz values increase even further after the formation of electrode-penetrating percolation channel, as seen in Fig. 7b, meaning that the infiltration of HNRs would be more promoted by the addition of more PS beads. Therefore, the Cd/  Ti ratio exhibits a further increase for the TiO 2 :PS = 2:1 electrode, and the optimal PV performance is obtained at this ratio. We also believe that this strategy can benefit from the Mie-type light-scattering effects produced by spherical pores filled with electrolytes, as previously explored theoretically (in the case of dye-sensitized solar cells) and experimentally 58,59 . However, with a larger amount of PS beads the electrode structure collapses from the aggregation of PS beads forming large voids (Fig. 5d), leading to inferior J sc and FF due to the loss of active mp-TiO 2 volume to accommodate HNRs and deteriorated recombination at the direct electrolyte-FTO contact.
Our simple approach utilizing sacrificial spherical additives results in ~32% additional enhancement in J sc compared to the OT-CdSe/CdSe 0.4 Te 0.6 HNR-SSC from the mp-TiO 2 electrode without polystyrene   microbead-induced percolating pores, yielding a 3.02% efficient PV device. This is one of the highest values among the SSCs with 1-D sensitizers, and a promising result proving that multi-composition variants of 1-D nanostructures can be fully utilized for this type of photovoltaic device with sensible engineering.

Discussion
In this work, CdSe/CdSe x Te 1−x HNRs were effectively utilized as a light absorber material for semiconductor-sensitized solar cells. Tailored nano-heterostructures and surface passivation with suitable ligand proved to be crucial factors to achieve the enhanced photocurrent density, by utilizing the superior optoelectronic properties of type-II heterostructures and modifying the conduction band level with surface ligands. Furthermore, to circumvent the spatial incompatibility of 1-D sensitizers with the conventional mp-TiO 2 electrodes, spherical PS bead additives were incorporated to render a hierarchical pore structure in the mp-TiO 2 electrode. Electrodes with PS-modified pore structures showed highly enhanced PV performance by the formation of large percolating pores inside the electrode, which was also supported by a Monte-Carlo simulation. Consequently, 3.02% efficient HNR-SSC was achieved by integrating the aforementioned approaches, exhibiting that 1-or higher-dimensional nanostructures and their multi-composition variants can potentially be fascinating alternatives as light harvesters for sensitized PVs. during the growth of seeds, limiting the average overall length of the HNRs to ~25 nm. After 5 additional minutes of growth, the reaction mixture was cooled by removing the heating mantle. The NR/ HNR suspensions were cleaned once by precipitation with chloroform and methanol, and redissolved in hexanes. Any insoluble precipitates were discarded, and the NR/HNRs were precipitated again by the addition of chloroform and methanol, dissolved in anhydrous toluene, and stored under N 2 . Recapping with 1-octanethiol followed our previous work on the CdSe/CdTe HNRs 36 . Briefly, 12 mL of 1-octanethiol was added to the flask containing approximately 4 mL of reaction mixture under Ar atmosphere. The reaction vessel was heated to 110 °C, and allowed to cool slowly to 60 °C. After stirring overnight, the HNRs were purified and stored as above. Each immersion took 1 min, and the films were rinsed with DIW after each immersion to remove any remaining ions. This cycle was repeated twice to complete the ZnS passivation. To fabricate cuprous sulfide counter electrodes, thoroughly polished brass foil was first etched by hydrochloric acid for 20 min in an 80 °C oven. Etched foil was then sulfurized by adding a polysulfide solution (2 M Na 2 S and 2 M S in DIW) droplet, immediately after which the foil turned black 62 . The NR-or HNR-sensitized mp-TiO 2 electrode and cuprous sulfide counter electrode were finally assembled into a sandwich-type cell using a binder clip and 60 μ m thick scotch tape as a spacer. A solution of 1 M Na 2 S, 1 M S, and 0.2 M KCl in MeOH:DIW = 7:3 solvent was used as the working electrolyte.

Preparation of Polystyrene
Characterization. Transmission electron microscopy (TEM) was carried out on a JEOL 2100 TEM operating at 200 kV with samples that were prepared by drop-drying a dilute solution of NRs or HNRs in chloroform onto a Cu grid with a thin carbon film (Electron Microscopy Sciences). The UV-vis absorption spectra were collected with an Agilent 8453 photodiode array spectrometer, and photoluminescence spectra were collected with a Horiba Jobin Yvon FluoroMax-3 fluorometer. The morphologies of mp-TiO 2 electrodes were analysed using a scanning electron microscope (JSM-6360: Hitachi). The photocurrent-voltage (J-V) curves of SSCs were obtained with a potentiostat (CHI 608C: CH Instrumental Inc., Austin, USA) under AM 1.5 illumination (K3000: McScience, Korea, intensity at 100 mW cm −2 ). An incident photon-to-current conversion efficiency (IPCE) measurement system (K3100: McScience, Korea) was used to obtain IPCE spectra. Electrochemical impedance spectra (EIS) were obtained by a potentiostat with 20 mV sinusoidal perturbation and frequencies ranging from 10 −1 to 10 5 Hz.