Understanding charge transfer and recombination by interface engineering for improving the efficiency of PbS quantum dot solar cells

Chao Ding ab, Yaohong Zhang a, Feng Liu a, Yukiko Kitabatake c, Shuzi Hayase *de, Taro Toyoda ae, Ruixiang Wang f, Kenji Yoshino *eg, Takashi Minemoto *eh and Qing Shen *ae
aFaculty of Informatics and Engineering, The University of Electro-Communications, 1-5-1 Chofugaoka, Chofu, Tokyo 182-8585, Japan. E-mail: shen@pc.uec.ac.jp; Tel: 81-424435464
bChina Scholarship Council, Level 13, Building A3, No. 9 Che gong zhuang Avenue, Beijing 100044, China
cDepartment of Applied Chemistry, Chuo University, 1-13-27, Kasuga, Bunkyo, Tokyo 112-8551, Japan
dGraduate School of Life Science and Systems Engineering, Kyushu Institute of Technology, 2-4 Hibikino, Wakamatsu-ku, Kitakyushu, Fukuoka 808-0196, Japan. E-mail: hayase@life.kyutech.ac.jp
eCREST, Japan Science and Technology Agency (JST), 4-1-8 Honcho, Kawaguchi, Saitama 332-0012, Japan
fBeijing Engineering Research Centre of Sustainable Energy and Buildings, Beijing University of Civil Engineering and Architecture, No. 15 Yongyuan Road, Huangcun, Daxing, Beijing 102616, China
gDepartment of Electrical and Electronic Engineering, Miyazaki University, 1-1 Gakuen Kibanadai Nishi, Miyazaki 889-2192, Japan. E-mail: t0b114u@cc.miyazaki-u.ac.jp
hDepartment of Photonics, Faculty of Science and Engineering, Ritsumeikan University, 1-1-1 Noji Higashi, Kusatsu, Siga 525-8577, Japan. E-mail: minemoto@se.ritsumei.ac.jp

Received 8th February 2018 , Accepted 21st March 2018

First published on 26th March 2018


In quantum dot heterojunction solar cells (QDHSCs), the QD active layer absorbs sunlight and then transfers the photogenerated electrons to an electron-transport layer (ETL). It is generally believed that the conduction band minimum (CBM) of the ETL should be lower than that of the QDs to enable efficient charge transfer from the QDs to the collection electrode (here, FTO) through the ETL. However, by employing Mg-doped ZnO (Zn1−xMgxO) as a model ETL in PbS QDHSCs, we found that an ETL with a lower CBM is not necessary to realize efficient charge transfer in QDHSCs. The existence of shallow defect states in the Zn1−xMgxO ETL can serve as additional charge-transfer pathways. In addition, the conduction band offset (CBO) between the ETL and the QD absorber has been, for the first time, revealed to significantly affect interfacial recombination in QDHSCs. We demonstrate that a spike in the band structure at the ETL/QD interface is useful for suppressing interfacial recombination and improving the open-circuit voltage. By varying the Mg doping level in ZnO, we were able to tune the CBM, defect distribution and carrier concentration in the ETL, which play key roles in charge transfer and recombination and therefore the device performance. PbS QDHSCs based on the optimized Zn1−xMgxO ETL exhibited a high power conversion efficiency of 10.6%. Our findings provide important guidance for enhancing the photovoltaic performance of QD-based solar cells.



Conceptual insights

For the first time, we have demonstrated that in quantum dot heterojunction solar cells (QDHSCs) a spike structure at the interface, i.e., the energy level of the conduction band of the electron transporting layer (ETL) is higher than that of the QD absorber, can suppress the charge recombination at the interface. We find that efficient charge injection can still occur even when the spike structure is formed between the QD absorber and the ETL. We think that this is because that the shallow defect states in the ETL can serve as additional pathways to transport the photoexcited electrons from the QDs to the electron collecting electrode (FTO). By taking advantage of these characteristics, we can greatly improve the efficiency of the QDHSCs just by slightly tuning the conduction band energy offset between the ETL and the QDs. In addition, by probing free electron absorption in FTO using broadband transient absorption spectroscopy, we find that ultra-fast electron transfer occurs in a time scale of a few hundred femtoseconds from PbS QDs to FTO through the Zn1−xMgxO compact layer. We believe that our findings would provide guidance for enhancing the photovoltaic performance of QD-based solar cells further.

Introduction

Over the past few decades, quantum dot solar cells (QDSCs) have been widely studied as a promising next-generation photovoltaic technology to generate clean and renewable energy. Among the various kinds of QDSCs, PbS QDSCs have attracted much attention due to their significant advantages such as ease of solution processing, high stability, and potential multiple exciton generation.1–4 In addition, the band gap of PbS QDs can be readily tuned over a wide range (0.5–1.6 eV) by adjusting their particle size due to the quantum confinement effect, which enables a wide spectral response from the ultraviolet to the near-infrared region.5 Due to these unique properties, QDSCs based on PbS QDs have achieved impressive device performance, with a power conversion efficiency (PCE) exceeding 11%.6

Colloidal PbS QDs are often employed in depleted-heterojunction solar cells, which contain five different parts: a transparent conducting oxide (TCO) layer (e.g., fluorine-doped tin oxide (FTO) and indium-doped tin oxide (ITO)), an electron-transport layer (ETL) (or a hole-blocking layer (HBL), e.g., ZnO, SnO2, and TiO2), PbS QDs, a hole-transport layer (HTL) (or an electron-blocking layer (EBL), e.g., MoOx and organic small molecules), and contact electrodes (e.g., Au and Ag).7–11 Unlike perovskite solar cells, which can still function without the inclusion of a HTL or an ETL, in PbS QD heterojunction solar cells (QDHSCs), both an ETL and a HTL are necessary for achieving high efficiency, because they enhance charge extraction and suppress charge recombination at the TCO/QD and QD/gold interfaces.12–15 This may result because QDs have more surface defects than perovskites, which can cause severe back recombination of injected electron carriers at the FTO/QD interface (Fig. S1, ESI).

In a QDHSC, the p-type layer is a QD film, and the n-type layer is an ETL that accepts photogenerated electron carriers. The built-in electric field is mainly distributed near the ETL/QD interface.11,16 Usually, when a favourable cliff-like conduction band structure forms at the ETL/QD interface (i.e., the conduction band of the QDs is higher than that of the ETL), electron injection will occur. When the conduction band of the QDs is lower than that of the ETL, forming a spike structure, electron injection will be retarded.13,16 Such observations have been demonstrated in QD-sensitized solar cells and Cu(In,Ga)Se2 (CIGS) solar cells.13,17–19 However, in this work, we found that in magnesium (Mg)-doped ZnO/PbS QDHSCs the ZnO layer does not need to have a lower conduction band energy to realize efficient charge transfer; that is, although a spike structure is formed between the QDs and the “electron acceptor”, charge injection can still occur, and improved injection of photogenerated electrons can occur at a certain conduction band offset (CBO). Here, we used a thin Mg-doped ZnO film (∼30 nm) (Zn1−xMgxO, x = 0, 0.05, 0.10, 0.15, 0.20) as an ETL with tuneable characteristics. In fact, Zn1−xMgxO films with high transparency and tuneable band gaps have been employed in PbX (X = S, Se) QDHSCs as a novel electron-transport material, and greatly improved solar cell efficiencies have been achieved with enhanced open-circuit voltages (Voc).8,20,21 However, previous studies did not provide further insight into the role of Mg doping, for example, its impact on the carrier dynamics, including charge injection and recombination. Here, we explore the carrier injection and recombination mechanism in FTO/Zn1−xMgxO/PbS QDHSCs by varying the Mg doping level in ZnO. The effects of Mg doping on the intrinsic properties (e.g., defect states and carrier concentration) of the Zn1−xMgxO layer and the ensuing device performance were systematically studied by several spectral measurements, such as transient photovoltage (TPV) decay measurements and ultrafast transient absorption (TA) spectroscopy. Interestingly, we found that Mg doping can greatly alter the photophysical properties of ZnO and, in particular, that the resulting spike band structure of Zn1−xMgxO/PbS QDs can inhibit charge recombination, while the shallow defect states can serve as additional pathways to transport photoexcited electrons from the QDs to the electron-collecting electrode. More importantly, we also revealed the time constant of this transport process by TA measurements.

Results and discussion

A Zn1−xMgxO ETL was prepared using a sol–gel method as described previously.22,23 Briefly, the Zn–Mg mixed sol–gel solution was prepared by mixing zinc salt (Zn2+) and magnesium salt (Mg2+) (0, 5, 10, 15 and 20 mol% based on metal ions), and then, the mixed sol–gel solution was spin-coated onto pre-cleaned FTO substrates and sintered at 290 °C. A ZnO film without Mg doping was prepared according to ref. 23. X-ray photoelectron spectroscopy was carried out to verify the Mg doping. The binding energy (BE) scale was calibrated with the C 1s peak of carbon at 284.28 eV. Fig. 1 shows the XPS spectra of the Zn1−xMgxO (x = 0, 0.05, 0.10, 0.15, 0.20) films with different Mg doping levels. Typical BE peaks at 1021.3 and 1044.5 eV, assigned to Zn2+ 2p3/2 and 2p1/2, are observed in all Zn1−xMgxO spectra, as shown in Fig. 1(a), confirming the presence of divalent Zn ions in all samples.24 As shown in Fig. 1(b), the BE peak assigned to Mg 2p is observed at 49.7 eV in the Zn1−xMgxO (x = 0.5, 0.10, 0.15) spectra, while that of the x = 0.2 film is observed at 48.7 eV with enhanced intensity and area, which suggests that increased Mg doping in Zn1−xMgxO can lead to the formation of MgO for x ≥ 0.2. Fig. 1(c) shows BE components at 532.4 eV and 533.9 eV in the Zn1−xMgxO spectra, attributable to the O2− ions in Zn–O or Mg–O and the oxygen-deficient components of the films, respectively.25–27 The XPS results confirm the successful doping of Mg in the ZnO films.28,29
image file: c8nh00030a-f1.tif
Fig. 1 XPS spectra of the Zn1−xMgxO films (x = 0, 0.05, 0.10, 0.15, 0.20): (a) Zn 2p, (b) Mg 2p, and (c) O 1s with two resolved O bonding (red and green lines) components.

The optical absorption spectra of the Zn1−xMgxO-coated FTO samples, shown in Fig. 2(a), indicate a continuous blueshift of the absorption edges with increasing Mg content (x = 0–0.2). The optical band gaps, Eg, obtained by linearly extrapolating the Tauc plot [(αhν)2vs. photon energy],30 are presented in the inset of Fig. 2(a), and the corresponding band gap of the Zn1−xMgxO films continuously increases from 3.26 to 3.56 eV with Mg doping, which is in agreement with previous reports.8,29 According to other reports of Zn1−xMgxO,28,31 the suitable doping of Mg2+ at Zn2+ sites in ZnO induces an obvious change in the conduction band (CB) energy of the ternary oxide, but the influence on the valence band (VB) edge of Zn1−xMgxO is negligible. To verify these results, the VB edge energy levels of the five Zn1−xMgxO samples were measured by photoelectron yield spectroscopy (PYS), as shown in Fig. S2 (ESI). The VB edge energy levels were obtained from the intersection of the baseline and the tangent to the spectra, and we also found a negligible effect of Mg doping on the VB edge energy levels of the Zn1−xMgxO films. The energy band diagram of the Zn1−xMgxO films is schematically depicted in Fig. 2(b) by combining the above results of both the band gap and the VB edge level. From this energy band diagram, a continuous shift in the conduction band minimum (CBM) towards higher energy with increasing Mg doping is observed. The general trend of these values is in good agreement with previous reports.8,32


image file: c8nh00030a-f2.tif
Fig. 2 (a) Optical absorption spectra of the Zn1−xMgxO films. The inset shows an increase in the bandgap of Zn1−xMgxO films with Mg doping. (b) Experimentally determined diagram of energy levels (relative to the vacuum level) of the Zn1−xMgxO films (x = 0–0.20), the valence band maximum (VBM) and the conduction band minimum (CBM) are represented in eV.

Our above characterization suggests that ZnO films can be doped with Mg by a facile sol–gel method and the band gaps can be easily tuned by adjusting the Mg doping level. Next, we used Zn1−xMgxO (x = 0, 0.05, 0.10, 0.15, 0.20) thin films (∼30 nm) to prepare PbS QDHSCs. PbS QDs were synthesized using a modified hot-injection method,33 and QDs of different sizes were prepared by controlling the injection temperature. The excitonic peak of the PbS QDs was located at 1.19 eV, as shown in Fig. S3(a) (ESI), and the VB edge energy level of the PbS QDs was confirmed to be −5.24 eV by PYS measurements, as shown in Fig. S3(b) (ESI). By combining the results of both the bandgap and the VB edge level, the CB edge level was determined to be −4.06 eV, as seen in the inset of Fig. S3(b) (ESI). The corresponding transmission electron microscopy (TEM) image reveals that the average size of the PbS QDs is approximately 3.5 nm (Fig. S4(a), ESI), and the high-resolution transmission electron microscopy (HR-TEM) image shows that the PbS QDs exhibit a highly-crystalline structure with interplanar distances of 0.342 and 0.291 nm, corresponding to the (111) and (200) planes of rock-salt PbS, respectively (Fig. S4(b), ESI). The crystallinity of PbS QDs is further supported by the selected area electron diffraction (SAED) pattern (see insets in Fig. S4(b), ESI). Fig. 3(a) and (b) show the typical schematic structure and a cross-sectional scanning electron microscopy (SEM) image, respectively, of a Zn1−xMgxO/PbS solar cell, wherein the PbS QD layer (∼300 nm) serves as the light absorbing layer.


image file: c8nh00030a-f3.tif
Fig. 3 (a) Schematic illustration of a Zn1−xMgxO/PbS solar cell structure. (b) Cross-sectional SEM image of a typical Zn1−xMgxO/PbS solar cell (∼300 nm thick PbS QD active layer).

Fig. 4(a) shows the typical photocurrent density–voltage (JV) characteristics of the five types of QDHSCs upon standard 100 mW cm−2 AM 1.5G illumination. We measured 24 devices for each type of QDHSC. The average values of photovoltaic parameters such as the open-circuit voltage (Voc), short-circuit current (Jsc), fill factor (FF), series resistance (Rs), shunt resistance (Rsh), and PCE are summarized in Table 1. Compared with the undoped device (ZnO), the Zn1−xMgxO films significantly improved the device performance. As expected, the Voc was continuously enhanced with increasing Mg doping level due to the reduced loss of electrons, as observed in other reports.8,20 We consider the enhanced Voc to result from the higher CBM of Zn1−xMgxO, which is shown in Fig. 4(b). However, when the Mg doping level exceeds 10%, the Voc becomes saturated due to dominant charge recombination at the Zn1−xMgxO/QD interface caused by the high level of Mg doping, as demonstrated previously.32,34,35 Surprisingly, the Jsc of the device continued to increase at high Mg doping levels. For example, the Jsc of the device with a 15% Mg doping level was greatly improved by 30%. The control photovoltaic device based on pure ZnO showed a PCE of 5.51 ± 0.11%, with a smaller Voc of 419.06 ± 5.8 mV and a lower FF of 49.94 ± 0.85%. The solar cell based on the optimized Mg doping level achieved the highest efficiency of 7.55 ± 0.21%, with a Voc of 437.62 ± 6.95 mV, Jsc of 33.59 ± 1.32 mA cm−2 and FF of 51.82 ± 1.55%, as shown in Table 1, providing an improvement in efficiency of over 36% compared to the ZnO control devices.


image file: c8nh00030a-f4.tif
Fig. 4 (a) JV characteristics of PbS QDHSCs with Zn1−xMgxO (x = 0–0.20) under simulated AM 1.5G illumination. (b) Band alignment of PbS QDs and Zn1−xMgxO with different Mg contents (x = 0–0.2).
Table 1 Statistical averages of the photovoltaic performance parameters of 24 devices for each type of QDHSCa
Sample J sc (mA cm−2) V oc (mV) FF (%) R s (Ω cm−2) R sh (Ω cm−2) PCE (%)
a Results for the device with the highest PCE are shown in parentheses.
x = 0 26.34 ± 0.75 419.06 ± 5.8 49.94 ± 0.85 64.33 ± 3.02 99.86 ± 2.97 5.51 ± 0.11
(27.64) (412) (49.3) (67.35) (102.83) (5.62)
x = 0.05 29.33 ± 0.88 434.18 ± 6.35 50.53 ± 1.52 55.66 ± 4.78 204.93 ± 4.23 6.43 ± 0.08
(30.53) (428) (49.9) (60.44) (209.16) (6.51)
x = 0.10 30.91 ± 1.27 437.53 ± 4.19 51.05 ± 1.21 52.33 ± 3.71 220.27 ± 3.14 6.79 ± 0.09
(31.36) (436) (50.3) (48.92) (223.41) (6.88)
x = 0.15 33.59 ± 1.32 437.62 ± 6.95 51.82 ± 1.55 43.52 ± 2.96 174.03 ± 3.99 7.55 ± 0.21
(34.86) (437) (51.1) (40.56) (178.02) (7.76)
x = 0.20 34.23 ± 0.93 434.12 ± 6.05 47.18 ± 1.55 59.33 ± 4.11 74.17 ± 3.67 7.01 ± 0.34
(35.05) (434) (48.3) (57.75) (78.44) (7.35)


To elucidate the difference in the photovoltaic performance when changing the Mg doping level in ZnO, we first investigated the band alignment between the PbS and Zn1−xMgxO layers. As shown in Fig. 4(b), the CBO values between the PbS QDs (Eg = 1.19 eV) and Zn1−xMgxO were controlled from +0.07 eV to −0.2 eV by changing the Mg doping level in ZnO. When the Mg doping level was below 10%, the CB of Zn1−xMgxO was lower than that of the PbS QDs; i.e., the CBO was positive, thus forming a cliff structure. When the Mg doping amount was above 10%, the conduction band of Zn1−xMgxO was higher than that of the PbS QDs; i.e., the CBO was negative, thus forming a spike structure. Generally, the larger the conduction band offset of the cliff structure (CBOcliff), the better the injection of the photogenerated electrons. However, when the ETL/QD interface forms a spike structure (CBOspike), the injection of photogenerated electrons is retarded.13,16,20 However, in our case, a large CBOspike in a certain range instead led to better electron injection and an enhanced Jsc, which may be related to the presence of shallow defect states in the Zn1−xMgxO films and the decreased carrier concentration due to Mg doping, which will be discussed further below.

To investigate the effect of Mg doping on the solar cell performance, the room-temperature photoluminescence (PL) was measured. Fig. 5(a) shows the PL spectra of the Zn1−xMgxO samples at room temperature. The PL spectrum of the ZnO film was fit by a Gaussian function, as shown in Fig. S5(a) (ESI), which gives five peaks centred at 380, 400, 478, 529, and 572 nm. The narrow ultraviolet emission at 380 nm is related to the near band edge transition in ZnO. Furthermore, the edge of this near band edge-related emission shows a continuous shift towards a shorter wavelength with increasing doping level, which reflects the band gap broadening due to Mg doping and is consistent with the results of the optical absorption spectra in Fig. 2(a). The 400–600 nm emission regions are attributed to shallow and deep level defects (as shown in Fig. S5(b), ESI) within the ZnO crystal, such as interstitials and vacancies of zinc and oxygen.36 In general, the violet (400 nm) and blue (470 nm) emissions in the PL spectra of ZnO are related to shallow defects, while the green (529 nm) and yellow (572 nm) emissions are related to deep defects.37 Therefore, the PL spectra can be divided into sections I, II, and III, representing band, shallow and deep defect-dominated emission, respectively. Fig. 5(b) shows the integral area of the shallow defect-related emission range in the PL spectra (i.e., section II). It is well known that the integrated area of the PL emission from defect states is proportional to the defect state density of the corresponding energy level. Interestingly, we found that the change in the integral area for section II (shallow defect state emission) with Mg doping level was in good accordance with the change in Jsc, suggesting that the electron-transfer process may have some relation to the shallow defect states. This can be understood by the fact that in the absence of external excitation charge-carrier transfer from low to high energy levels is difficult. In addition, although the thickness of Zn1−xMgxO is low (∼30 nm), it is not low enough for electrons to be transported directly from the QDs to FTO by tunnelling. Therefore, when a spike band structure is formed, the shallow defect states in the Zn1−xMgxO layer may act as additional pathways for charge carriers to be transferred from the QDs to the FTO electrode. As shown in Fig. 5(c), in our PbS QDHSCs, when Zn1−xMgxO and the PbS QDs form a cliff band structure, the photogenerated electrons can be transferred from the QD CB to the Zn1−xMgxO CB via pathways ① and ② (intragap states). When Zn1−xMgxO and the PbS QDs form a spike band structure, transfer of the photogenerated carriers to Zn1−xMgxO through pathways ① and ② is difficult due to the energy barrier. However, due to the favourable energy level alignment between the shallow defect level of Zn1−xMgxO and the CB or intragap states of the PbS QDs, photogenerated electrons can be injected into Zn1−xMgxO through pathways ③ and ④ and then collected at the FTO electrode. In addition, the charge injection rate from the absorber to the ETL depends on the density of accepting states (i.e., sub-band-gap surface states) in the ETL.38–41 The Mg-doped ZnO ETL provides a greater density of accepting states for electron injection due to the lower carrier concentration in Zn1−xMgxO (∼1014 cm−3), as shown in Fig. S6 (ESI). The broadened optical band gap of Mg-doped ZnO enhances the light absorption of the PbS QD layer, which also improves the Jsc.


image file: c8nh00030a-f5.tif
Fig. 5 (a) Room temperature PL spectra of Zn1−xMgxO films (x = 0, 0.05, 0.10, 0.15 and 0.20). (b) The integral area of the section II range of PL spectra of Zn1−xMgxO films and Jsc of Zn1−xMgxO/PbS QDHSCs. (c) Carrier-injection processes for photogenerated electrons (red arrows) and holes (blue arrow) in Zn1−xMgxO/PbS QDHSCs (cliff and spike structures) under short circuit conditions. ① Photogenerated electrons can be injected from the QD CB to the Zn1−xMgxO CB. ② Captured photo-generated carriers can be injected from the QD intragap state to the Zn1−xMgxO CB. ③ Photogenerated electrons can be injected from the QD CB to the Zn1−xMgxO shallow defects. ④ Captured photo-generated carriers can be injected from the QD intragap state to the Zn1−xMgxO shallow defects. (d) The integral area of the section III range of PL spectra of Zn1−xMgxO films and Voc of Zn1−xMgxO/PbS QDHSCs.

Notably, the properties of the ETL also have a significant effect on charge recombination in the device. Fig. 5(d) shows that as the Mg doping level increased, the integral area of section III (deep defect state emission) first decreased from x = 0 to 0.10 and then increased from x = 0.10 to 0.20, suggesting that with an increase in the Mg doping level, the deep defects caused by oxygen vacancies in Zn1−xMgxO gradually decrease, but when the Mg doping level exceeds 10%, the excessive Mg doping may introduce new deep defects in ZnO, e.g., MgInterstitial (as seen in Fig. S5(b), ESI), which can easily become carrier recombination centres,37 leading to greater charge recombination at the Zn1−xMgxO/PbS QD interface. This also explains the saturated Voc observed under excessive Mg doping conditions, as shown in Fig. 5(d). However, our JV results show that when the Mg doping level exceeds 10% the Voc tends to decrease to a saturated value rather than continuously decrease, possibly because the formed spike band structure alleviates interfacial recombination to a certain extent, as demonstrated in other types of heterojunction solar cells.19,32,35 Therefore, the change in Voc with increased Mg doping level in our QDHSCs is considered to be a combined effect of defect states in Zn1−xMgxO and the formed CBO.

To gain further insight into the effect of Mg doping in Zn1−xMgxO on charge recombination in solar cells, the diode performance of the solar cells was studied by examining the dark JV characteristics. As seen in Fig. 6(a), the dark current of the PbS QDHSCs decreased with increasing Mg doping level. The continuous decrease in the dark current of the solar cells may be due to the increased built-in potential in the Zn1−xMgxO layer, attributed to the increased Femi level of Mg-doped ZnO.41 To better understand the effect of Mg doping on device performance, we next analysed the diode behaviour of our devices in the dark by introducing the traditional equivalent circuit model (Fig. S7, ESI).42 In this model, the JV behaviour of the solar cell includes four constituent parts: a photocurrent source, a diode, a series resistor, and a shunt resistor. This model can be mathematically represented by the following equation:43

 
image file: c8nh00030a-t1.tif(1)
where J0 is the reverse bias saturation current density, q is the elementary charge, Rs is the series resistance, n is the diode ideality factor, kB is Boltzmann's constant, T is the temperature, and Rsh is the shunt resistance. Fig. 6(b) shows the dark curves derived from Fig. 6(a) using the semi-log scale, which can be divided into regions I, II, and III, corresponding to the three parts of eqn (1), which describe how the different components of the solar cell equivalent circuit (Fig. S7, ESI) affect the JV response of the cell at different voltages. At low voltages (region I: the 3rd term in eqn (1)), the JV curve is mainly related to Rsh (shunt current); at intermediate voltages (region II: the 1st term in eqn (1)), the JV curve is related to diode parameters J0 and n (recombination current); while at high voltages (region III: the 2nd term in eqn (1)), the JV curve is determined by Rs.43 These regions provide important information when evaluating the JV response curve. For instance, a steep slope in region III generally indicates a low Rs.43 Thus, according to the observation of the dark current density in regions I, II and III in Fig. 6(b), we found that the changes in Rsh and Rs of the devices with increasing Mg doping level are in agreement with the Rsh and Rs measured from JV characterization under illumination (Table 1). When the Mg doping level was increased from 0 to 10%, the current leakage (region I) of the device continuously decreased, but when the Mg doping level was above 20%, significant current leakage was observed. This could be attributed to the fact that the appropriate Mg doping level (x = 0.1) leads to better crystallization of the ZnO film and further decreases the defects and current leakage of the thin films.44


image file: c8nh00030a-f6.tif
Fig. 6 (a) Typical JV characteristics of PbS QDHSCs with Zn1−xMgxO (x = 0–0.20) ETLs measured in the dark. (b) Semilog plots of JV curves measured in the dark and using the absolute values of current density. The three regions indicate three different effects in the solar cell: region I accounts for leakage (shunt) currents, region II accounts for recombination currents, and region III accounts for series resistance.

The n and J0 values of the PbS QDHSCs with Zn1−xMgxO were determined by least-squares fitting the dark JV curves in section II to eqn (1), and the values are summarized in Table 2.45 The n value is related to the recombination mechanism in the solar cell. When n is close to 1, direct recombination (band-to-band) dominates; otherwise, indirect recombination mechanisms, such as interfacial recombination and trap-assisted recombination (n = 2), dominate.9 Our results show that the changes in the n value of the five different devices are consistent with the changes in the density of deep defects in Zn1−xMgxO with increasing Mg doping level (as shown in Fig. 5(d)). The decrease in the ideality factor from 1.73 (x = 0) to 1.35 (x = 0.10) with an increase in the Mg doping level in ZnO indicates that the contribution of indirect recombination (back-transfer recombination of the injected electrons through trapping by deep defects in Zn1−xMgxO) is relatively reduced under open-circuit conditions due to the removal of the Zn1−xMgxO deep trap sites. When the Mg doping level was above 10%, with an increase in deep level defects in Zn1−xMgxO, the n value increased rapidly to become larger than that of the undoped device of 1.73. Although Zn1−xMgxO and PbS QDs will form a spike band structure when the Mg doping level is above 10%, this does not lead to reduced indirect recombination of the charges. Previously, the Greenham group reported that reducing the ZnO carrier concentration by nitrogen doping can prevent interfacial recombination in CQDSCs;41 however, our results show that interfacial recombination is not suppressed despite the much lower carrier concentration of the Mg-doped devices compared with the undoped device (as shown in Fig. S6, ESI). These results indicate that indirect recombination of charges in the PbS QDHSCs strongly depends on the deep level defect density in Zn1−xMgxO. In addition, the J0 values displayed a similar trend for the device doped with 10% Mg, showing the smallest J0 value and the lowest charge recombination.

Table 2 Diode ideality factor (n) and reverse saturation current density (J0) of the PbS QDHSCs with Zn1−xMgxO (x = 0–0.20)
Sample n J 0
x = 0 1.73 7.374 × 10−3
x = 0.05 1.39 3.867 × 10−4
x = 0.10 1.35 1.974 × 10−4
x = 0.15 1.82 1.865 × 10−3
x = 0.20 2.54 3.727 × 10−2


For detailed analysis, the effect of increased Mg doping on the recombination mechanism in Zn1−xMgxO/PbS QDHSCs was examined by transient photovoltage (TPV) decay analysis. When a PbS QDHSC is illuminated by a laser pulse under open-circuit conditions, an open-circuit photovoltage is generated across the full device. When the incident laser is switched off, all photogenerated carriers recombine, and the photovoltage decays.9,46–48Fig. 7(a) shows the TPV decay curves of the solar cells with increasing Mg doping level in ZnO. All decay curves were well fitted by a three-exponential-function equation, and the corresponding parameters are summarized in Table 3. The three-component fitting implies that three recombination processes are present in this case,9 similar to our previously reported results.9,46–48 First, the ultrafast process (<0.1 ms, constant τ1 in Table 3, in the high-Voc regime) can be related to intrinsic trapping-assisted recombination in PbS and the Zn1−xMgxO films. In Table 3, the decay time constant τ1 was almost the same (approximately 0.02 ms) for all devices, considering the ultrafast time scale and fitting uncertainty. Notably, the weight of A1 decreases from the x = 0 device to the x = 0.1 device; this change reflects the reduction in the trap state density of the Zn1−xMgxO films and thus reduced recombination through trap states. For the second process (i.e., in the medium-Voc regime), the related constant τ2 is more than 0.2 ms, and this process can be related to interfacial recombination at the Zn1−xMgxO/PbS interface. Table 3 shows that the trends in the values of A2 and the decay time constant τ2 are in good agreement with the trend of deep level defect intensity (Fig. 5(d)) in Zn1−xMgxO with increasing Mg doping. This indicates that a reduction in the Zn1−xMgxO deep level defect density can mitigate interfacial recombination between electrons trapped in the Zn1−xMgxO layers and holes in the VB (or intragap states near the VB) of the PbS QDs. In the third decay process (i.e., in the low-Voc regime), Voc disappears completely. This process represents the direct recombination of electrons in the CB of FTO and holes in the VB of the PbS layer, and the related decay time constant τ3 is on the order of a few ms. The weights of A3 and τ3 significantly increased for Zn0.9Mg0.1O. We found that solar cells based on Zn0.8Mg0.2O have the longest recombination lifetime (τ3) in process 3, probably due to the larger CBO (0.2 eV) between Zn0.8Mg0.2O and the PbS QDs, which forms an energy barrier that slows the direct recombination of free electrons and holes. The effective carrier lifetimes (τeff) were determined from the TPV decay curves using the defined equations (see details in the ESI). As shown in Fig. 7(b), the dependence of the effective carrier lifetime on the photovoltage can also be divided into three sections, corresponding to the three photovoltage decay processes. Throughout the Voc regime, the τeff values of the Zn0.9Mg0.1O-based device were 2–3 times higher than those of the other devices and were in the order of 10% > 5% > 0% > 15% > 20% Mg doping, which can be understood through the above discussions. Although the Zn0.9Mg0.1O-based cell showed the longest recombination lifetime, the Zn0.85Mg0.15O-based solar cell had the largest Voc and the fastest Voc decay. This indicates that the recombination of photogenerated carriers in the device is related to both factors, i.e., deep level defects in Zn1−xMgxO and the formed CBO.


image file: c8nh00030a-f7.tif
Fig. 7 (a) Open-circuit photovoltage decay curves for Zn1−xMgxO/M-PbS QDHSCs, showing three dominant decay processes. (b) The effective carrier lifetimes calculated from the voltage decay curves for Zn1−xMgxO/M-PbS QDHSCs.
Table 3 Time constants and weights obtained by fitting the open-circuit photovoltage decay curves of the PbS QDHSCs with a three-exponential function
Sample

image file: c8nh00030a-t4.tif

A 1 (A1/(A1 + A2 + A3)) τ 1 (ms) A 2 (A1/(A1 + A2 + A3)) τ 2 (ms) A 3 (A1/(A1 + A2 + A3)) τ 3 (ms)
x = 0 0.440 (47.5%) 0.023 ± 0.0006 0.426 (46.0%) 0.285 ± 0.004 0.060 (6.5%) 2.275 ± 0.004
x = 0.05 0.323 (32.4%) 0.015 ± 0.001 0.294 (29.5%) 0.408 ± 0.001 0.381 (38.1%) 3.46 ± 0.01
x = 0.10 0.243 (24.3%) 0.025 ± 0.0005 0.259 (26.0%) 0.411 ± 0.003 0.495 (49.7%) 5.17 ± 0.03
x = 0.15 0.381 (38.4%) 0.019 ± 0.0005 0.366 (36.9%) 0.273 ± 0.005 0.245 (24.7%) 1.438 ± 0.006
x = 0.20 0.269 (27.1%) 0.023 ± 0.001 0.658 (66.2%) 0.256 ± 0.001 0.067 (6.7%) 8.166 ± 0.005


For the above-studied PbS QDs (denoted “M-PbS”, Eg = 1.19 eV), the effect of the CBO on suppressing the interfacial recombination of the charge carriers is not obvious, which we attribute to the relatively small value of the CBO between Zn1−xMgxO and the QDs. Thus, large-sized PbS QDs (denoted “L-PbS”) were prepared to increase the CBO between Zn1−xMgxO and the PbS QDs. Fig. S8(a) (ESI) shows the first exciton absorption peak of the L-PbS QDs, located at 1.11 eV. The VB edge energy level of the L-PbS QDs was confirmed to be −5.22 eV by PYS measurements, as shown in Fig. S8(b) (ESI). By combining the results of both the band gap and the VB edge level, the CB edge level was determined to be −4.11 eV (inset of Fig. S8(b), ESI). The TEM and HRTEM images show that the average size of the L-PbS QDs was approximately 3.9 nm and the QDs were highly crystalline (Fig. S9, ESI). The photovoltaic parameters of the QDHSCs based on L-PbS are summarized in Fig. S10 (ESI). Fig. 8(a) shows both the cliff and spike band structures of the CBO in the PbS QDHSCs. The CBO values between the L-PbS QDs and Zn1−xMgxO were controlled from +0.10 eV to −0.25 eV by changing the Mg doping level in ZnO. In the QDHSCs, the Voc loss strongly depended on charge recombination, which was mainly affected by the junction characteristics.49 The Voc loss of all devices is shown in Fig. 8(b). The Voc loss was defined as Eg/qVoc, where q is the elementary charge. Compared with the effect of Zn1−xMgxO, the influence of the change in the size of the QDs on the Voc loss is more obvious. With an increase in QD size, the PL lifetime of the QDs increased from 276 ns to 985 ns, as shown in Fig. S11 (ESI). The increase in the PL lifetime is believed to mainly result from the reduction of surface defects in the PbS QDs. In addition, compared with the smaller PbS QDs, the defect states are mainly distributed near the CBM in the large-sized QDs.50 Therefore, devices based on larger QDs showed less Voc loss. We found that Voc loss in the L-PbS QD-based devices decreases with increasing deep defect density in Zn1−xMgxO when the Mg doping content is above 10%. We suggest that this is because the positive effect of the larger CBOspike value on carrier recombination is greater than the negative effect of deep defects in Zn1−xMgxO. We also measured the effective carrier lifetimes (τeff) of the L-PbS-based devices by the TPV decay, as shown in Fig. 8(c). As expected, for L-PbS QD-based devices, the variations in carrier lifetime in the high-voltage region are in line with the change in the deep defect density in the Zn1−xMgxO (x = 0–0.2) films. Note that for the L-PbS QDHSCs the τeff values of devices based on x = 0.15 and 0.20 are longer than that of the undoped device in the low-voltage regime, which is different from what was observed in M-PbS QDHSCs. According to the previous analysis, the medium-voltage regime is related to interfacial recombination, but the defect density of Zn1−xMgxO (x = 0.15 and 0.20) is higher than that of ZnO (see Fig. 7(a)). Therefore, we confirm that the reduced charge recombination is related to the increased CBOspike. In the medium-voltage regime, the τeff value increases with increasing CBOspike. However, from the photovoltaic performance of the QDHSCs based on L-PbS QDs (Fig. S10, ESI), we found that the electron injection is impeded when CBOspike exceeds 0.2 eV, which also results in a reduced Jsc and FF.


image file: c8nh00030a-f8.tif
Fig. 8 (a) The conduction band offset values for both PbS QDs (i.e., M-PbS: 1.19 eV and L-PbS: 1.11 eV) with Zn1−xMgxO, respectively. (b) The Voc losses of both M-PbS QDHSCs and L-QDHSCs, respectively. (c) The effective carrier lifetimes calculated from the voltage decay curves for Zn1−xMgxO/L-PbS QDHSCs.

To investigate the role of Mg doping in electron injection from the M-PbS QDs to Zn1−xMgxO and the FTO electrode at the heterojunction, we performed femtosecond (fs)-TA spectroscopy, which has broadband capability (470–1600 nm) and a 150 fs temporal resolution, as schematically shown in Fig. S12 (ESI). Because free-carrier absorption in transparent conductive glass, e.g., FTO, ZnO, and ITO, is mainly located in the infrared range (1400–4100 nm), we therefore focused on this range to investigate the dynamics of the injected electrons.51 The dotted black line in Fig. 9 shows the TA kinetics of the PbS QD layer on FTO, ZnO/FTO and Zn1−xMgxO/FTO (from bottom to top). The FTO/Zn1−xMgxO/PbS QD sample was excited at a wavelength of 470 nm and probed over a wavelength range of 1400 to 1600 nm, which was the same as that used for a PbS QD solution and PbS QD films deposited on glass and Zn1−xMgxO (without FTO). No TA signal was observed in this wavelength range (1400–1600 nm) for either the PbS QD solution or the PbS QD films on glass (as shown in Fig. S13(a), ESI). In addition, no TA signal was observed for the Zn1−xMgxO/PbS sample on a glass substrate (as shown in Fig. S13(b), ESI). Therefore, we confirm that the TA response shown in Fig. 9 is attributed to the absorption of electrons transferred to the FTO electrode from the PbS QDs. The TA response of the FTO/Zn1−xMgxO/PbS QD composite can be well fitted using the following single exponential functions image file: c8nh00030a-t2.tif (for the injection process) for time scales up to 1 ps and image file: c8nh00030a-t3.tif (for the recombination process) for time scales longer than 1 ps. The bottom part of Fig. 9(a) shows the injection dynamics for photogenerated electrons transferred from the PbS QDs to FTO. The time constant (410 ± 49 fs) is much faster than that for the transfer of photoexcited electrons from PbS QDs to TiO2 (<1 ns).52 By introducing a ZnO thin film (∼30 nm) between FTO and the PbS QD layer, the electron transfer rate decreases. We found that electron injection from the PbS QDs to the FTO electrode through the ZnO compact layer occurs on a time scale of a few hundred femtoseconds (here, 685 ± 47 fs), which is in good agreement with previous reports.53 In addition, electron transfer from the photoexcited PbS QDs to FTO through the Zn0.9Mg0.1O compact layer is faster (587 ± 61 fs) than that from the photoexcited PbS QDs to FTO through the ZnO compact layer. This result suggests that the photogenerated electrons are transported to the FTO electrode through the shallow defect states in Zn1−xMgxO when the CBs of Zn1−xMgxO and the QDs form a spike band structure. As can be seen in Fig. 9(b), the FTO/PbS QD sample shows an ultrafast recombination time of 125 ± 22 ps, but when a ZnO film was introduced between FTO and the PbS QDs, the recombination time significantly increased (895 ± 31 ps). Therefore, we conclude that the ZnO layer serves as a HBL to prevent recombination of the injected electrons in FTO with photogenerated holes in the PbS QDs, as shown in Fig. S1 (ESI). However, on the other hand, deep defects in ZnO can serve as recombination centres for the photogenerated electrons and holes. When Zn0.9Mg0.1O, which has fewer deep defects, was introduced as a HBL, no obvious carrier recombination was observed for up to 1 ns, which means that recombination occurs on a much longer time scale than observable by our experiments.


image file: c8nh00030a-f9.tif
Fig. 9 TA responses of the different samples of FTO/M-PbS QDs, FTO/ZnO/M-PbS QDs, and FTO/Zn1–xMgxO/M-PbS QDs, measured with a pump light wavelength of 470 nm and a probe wavelength of 1500 nm with a pump light intensity of 15 μJ cm−2. (a) Injection process of the photogenerated electrons from PbS QDs into the FTO electrode (red lines are fitted lines). (b) Decay process of the injected electrons in the FTO electrode (red lines are fitted lines).

As shown in Fig. 4(a), the solar cells based on Zn0.85Mg0.15O exhibit the best PCE. According to our previous report,9 interfacial recombination at the PbS/Au interface is important for the performance of QDHSCs. At present, high-efficiency PbS QDHSCs typically employ Bawendi's structure, i.e., an EDT-passivated QD layer is used as an efficient EBL (or HTL). Thus, we used Bawendi's method to introduce the EDT-treated PbS QD layer as a HTL to prevent interfacial recombination at the PbS/Au interface. We prepared five batch devices with a FTO/Zn1−xMgxO/M-PbS (∼260 nm)/PbS-EDT (∼40 nm)/Au structure, and the JV characteristics are shown in Fig. S14 and Table S1 (ESI). We found that solar cells based on Zn0.85Mg0.15O showed the best PCE when this HTL was included. By combining all these advantages, Zn0.85Mg0.15O/M-PbS QD-based solar cells achieved a high PCE of 10.6%, with a Jsc of 31.5 mA cm−2 and Voc of 590 mV, as shown in Fig. 10(a) and Table 4, which represents a significant improvement over the undoped device (8.3%). The external quantum efficiency (EQE) spectrum was integrated with the AM 1.5G solar spectrum of the best QDHSCs to obtain a Jsc of 30.3 mA cm−2, which is in good agreement with the measured JV result (31.52 mA cm−2), as shown in Fig. 10(b).


image file: c8nh00030a-f10.tif
Fig. 10 (a) JV curves and (b) the external quantum efficiency of the best performing photovoltaic devices under AM 1.5G illumination. The structure of the solar cell device is FTO/Zn0.85Mg0.15O (∼30 nm)/M-PbS/EDT-PbS/Au.
Table 4 Solar cell performance obtained with Zn1−xMgxO (x = 0, 0.15) ETLs
Sample J sc (mA cm−2) V oc (mV) FF R s (Ω cm2) R sh (Ω cm2) PCE (%)
x = 0 27.37 571 0.534 16.18 107.57 8.35
x = 0.15 31.52 590 0.570 101.46 310.91 10.60


Conclusions

In summary, for the first time, we have systematically demonstrated that in QDHSCs the formed spike structure between the ETL and QDs can inhibit charge recombination, and shallow defect states in the ETL can transport photoexcited electrons from the QDs to the electron-collecting electrode. More importantly, we also revealed that this transport process occurs on the ultrafast time scale of a few hundred fs. Considering the above results, the optimized Zn0.85Mg0.15O/M-PbS QD-based device exhibited a maximum PCE of 10.6% (Jsc increased from 27.3 to 31.5 mA cm−2 and Voc increased from 571 to 590 mV). Our study demonstrates the importance of optimizing the energy level alignment and physical properties of the ETL layer towards the overall performance.

Author contributions

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

Conflicts of interest

The authors declare no competing financial interest.

Acknowledgements

This research was supported by the Japan Science and Technology Agency (JST) CREST and PRESTO program and MEXT KAKENHI grant numbers 26286013 and 17H02736, and by the China Scholarship Council (CSC) under grant number 201608050109.

Notes and references

  1. S. A. McDonald, G. Konstantatos, S. Zhang, P. W. Cyr, E. J. D. Klem, L. Levina and E. H. Sargent, Nat. Mater., 2005, 4, 138–142 CrossRef CAS PubMed .
  2. P. V. Kamat, J. Phys. Chem. C, 2008, 112, 18737–18753 CAS .
  3. J. M. Luther, M. Law, M. C. Beard, Q. Song, M. O. Reese, R. J. Ellingson and A. J. Nozik, Nano Lett., 2008, 8, 3488–3492 CrossRef CAS PubMed .
  4. A. J. Nozik, M. C. Beard, J. M. Luther, M. Law, R. J. Ellingson and J. C. Johnson, Chem. Rev., 2010, 110, 6873–6890 CrossRef CAS PubMed .
  5. I. Kang and F. W. Wise, J. Opt. Soc. Am. B, 1997, 14, 1632–1646 CrossRef CAS .
  6. M. Liu, O. Voznyy, R. Sabatini, F. P. Garcia de Arquer, R. Munir, A. H. Balawi, X. Lan, F. Fan, G. Walters, A. R. Kirmani, S. Hoogland, F. Laquai, A. Amassian and E. H. Sargent, Nat. Mater., 2017, 16, 258–263 CrossRef CAS PubMed .
  7. J. Khan, X. Yang, K. Qiao, H. Deng, J. Zhang, Z. Liu, W. Ahmad, J. Zhang, D. Li, H. Liu, H. Song, C. Cheng and J. Tang, J. Mater. Chem. A, 2017, 5, 17240–17247 CAS .
  8. L. Hu, D.-B. Li, L. Gao, H. Tan, C. Chen, K. Li, M. Li, J.-B. Han, H. Song, H. Liu and J. Tang, Adv. Funct. Mater., 2016, 26, 1899–1907 CrossRef CAS .
  9. C. Ding, Y. Zhang, F. Liu, N. Nakazawa, Q. Huang, S. Hayase, Y. Ogomi, T. Toyoda, R. Wang and Q. Shen, ACS Appl. Mater. Interfaces, 2017 DOI:10.1021/acsami.7b06552 .
  10. C.-H. M. Chuang, P. R. Brown, V. Bulović and M. G. Bawendi, Nat. Mater., 2014, 13, 796 CrossRef CAS PubMed .
  11. X. Lan, S. Masala and E. H. Sargent, Nat. Mater., 2014, 13, 233 CrossRef CAS PubMed .
  12. A. G. Pattantyus-Abraham, I. J. Kramer, A. R. Barkhouse, X. Wang, G. Konstantatos, R. Debnath, L. Levina, I. Raabe, M. K. Nazeeruddin, M. Grätzel and E. H. Sargent, ACS Nano, 2010, 4, 3374–3380 CrossRef CAS PubMed .
  13. H. Liu, J. Tang, I. J. Kramer, R. Debnath, G. I. Koleilat, X. Wang, A. Fisher, R. Li, L. Brzozowski, L. Levina and E. H. Sargent, Adv. Mater., 2011, 23, 3832–3837 CrossRef CAS PubMed .
  14. S. M. Willis, C. Cheng, H. E. Assender and A. A. R. Watt, Nano Lett., 2012, 12, 1522–1526 CrossRef CAS PubMed .
  15. W. Ke, G. Fang, J. Wan, H. Tao, Q. Liu, L. Xiong, P. Qin, J. Wang, H. Lei, G. Yang, M. Qin, X. Zhao and Y. Yan, Nat. Commun., 2015, 6, 6700 CrossRef CAS PubMed .
  16. J. Tang, H. Liu, D. Zhitomirsky, S. Hoogland, X. Wang, M. Furukawa, L. Levina and E. H. Sargent, Nano Lett., 2012, 12, 4889–4894 CrossRef CAS PubMed .
  17. I. Mora-Seró and J. Bisquert, J. Phys. Chem. Lett., 2010, 1, 3046–3052 CrossRef .
  18. A. Braga, S. Giménez, I. Concina, A. Vomiero and I. Mora-Seró, J. Phys. Chem. Lett., 2011, 2, 454–460 CrossRef CAS .
  19. T. Minemoto, Y. Hashimoto, W. Shams-Kolahi, T. Satoh, T. Negami, H. Takakura and Y. Hamakawa, Sol. Energy Mater. Sol. Cells, 2003, 75, 121–126 CrossRef CAS .
  20. R. L. Z. Hoye, B. Ehrler, M. L. Böhm, D. Muñoz-Rojas, R. M. Altamimi, A. Y. Alyamani, Y. Vaynzof, A. Sadhanala, G. Ercolano, N. C. Greenham, R. H. Friend, J. L. MacManus-Driscoll and K. P. Musselman, Adv. Energy Mater., 2014, 4, 1301544–n/a CrossRef PubMed .
  21. X. Zhang and E. M. J. Johansson, J. Mater. Chem. A, 2017, 5, 303–310 CAS .
  22. D. C. Olson, S. E. Shaheen, M. S. White, W. J. Mitchell, M. F. A. M. van[thin space (1/6-em)]Hest, R. T. Collins and D. S. Ginley, Adv. Funct. Mater., 2007, 17, 264–269 CrossRef CAS .
  23. Z. Yin, Q. Zheng, S.-C. Chen and D. Cai, ACS Appl. Mater. Interfaces, 2013, 5, 9015–9025 CAS .
  24. C. J. Raj, K. Prabakar, S. N. Karthick, K. V. Hemalatha, M.-K. Son and H.-J. Kim, J. Phys. Chem. C, 2013, 117, 2600–2607 CAS .
  25. S. Bang, S. Lee, Y. Ko, J. Park, S. Shin, H. Seo and H. Jeon, Nanoscale Res. Lett., 2012, 7, 290 CrossRef PubMed .
  26. S.-H. Liao, H.-J. Jhuo, Y.-S. Cheng and S.-A. Chen, Adv. Mater., 2013, 25, 4766–4771 CrossRef CAS PubMed .
  27. T. Yang, W. Cai, D. Qin, E. Wang, L. Lan, X. Gong, J. Peng and Y. Cao, J. Phys. Chem. C, 2010, 114, 6849–6853 CAS .
  28. Z. Yin, Q. Zheng, S.-C. Chen, D. Cai, L. Zhou and J. Zhang, Adv. Energy Mater., 2014, 4, 1301404–n/a CrossRef .
  29. M. Wei, R. C. Boutwell, J. W. Mares, A. Scheurer and W. V. Schoenfeld, Appl. Phys. Lett., 2011, 98, 261913 CrossRef .
  30. J. Tauc, R. Grigorovici and A. Vancu, Phys. Status Solidi, 1966, 15, 627–637 CrossRef CAS .
  31. H. H. Zhang, X. H. Pan, B. Lu, J. Y. Huang, P. Ding, W. Chen, H. P. He, J. G. Lu, S. S. Chen and Z. Z. Ye, Phys. Chem. Chem. Phys., 2013, 15, 11231–11235 RSC .
  32. T. Minemoto, Y. Hashimoto, T. Satoh, T. Negami, H. Takakura and Y. Hamakawa, J. Appl. Phys., 2001, 89, 8327–8330 CrossRef CAS .
  33. J. Tang, K. W. Kemp, S. Hoogland, K. S. Jeong, H. Liu, L. Levina, M. Furukawa, X. Wang, R. Debnath, D. Cha, K. W. Chou, A. Fischer, A. Amassian, J. B. Asbury and E. H. Sargent, Nat. Mater., 2011, 10, 765–771 CrossRef CAS PubMed .
  34. T. Törndahl, C. Platzer-Björkman, J. Kessler and M. Edoff, Prog. Photovoltaics Res. Appl., 2007, 15, 225–235 CrossRef .
  35. T. Minemoto and M. Murata, Sol. Energy Mater. Sol. Cells, 2015, 133, 8–14 CrossRef CAS .
  36. R. B. M. Cross, M. M. D. Souza and E. M. S. Narayanan, Nanotechnology, 2005, 16, 2188 CrossRef CAS PubMed .
  37. S. Vempati, J. Mitra and P. Dawson, Nanoscale Res. Lett., 2012, 7, 470 CrossRef PubMed .
  38. K. Tvrdy, P. A. Frantsuzov and P. V. Kamat, Proc. Natl. Acad. Sci. U. S. A., 2011, 108, 29–34 CrossRef CAS PubMed .
  39. B.-R. Hyun, A. C. Bartnik, L. Sun, T. Hanrath and F. W. Wise, Nano Lett., 2011, 11, 2126–2132 CrossRef CAS PubMed .
  40. P. Tiwana, P. Docampo, M. B. Johnston, L. M. Herz and H. J. Snaith, Energy Environ. Sci., 2012, 5, 9566–9573 CAS .
  41. B. Ehrler, K. P. Musselman, M. L. Böhm, F. S. F. Morgenstern, Y. Vaynzof, B. J. Walker, J. L. MacManus-Driscoll and N. C. Greenham, ACS Nano, 2013, 7, 4210–4220 CrossRef CAS PubMed .
  42. K. K. N. Simon and M. Sze, Physics of Semiconductor Devices, Wiley-Interscience, New York, USA, 3rd edn, 2006 Search PubMed .
  43. J. D. Servaites, M. A. Ratner and T. J. Marks, Energy Environ. Sci., 2011, 4, 4410–4422 CAS .
  44. H. Saitoh, A. Takano, S. Kawaguchi, T. Washio, S. Ohshio and H. Akasaka, J. Ceram. Soc. Jpn., 2009, 117, 780–782 CrossRef CAS .
  45. X. Zhang, D. Yang, Z. Yang, X. Guo, B. Liu, X. Ren and S. Liu, Sci. Rep., 2016, 6, 35091 CrossRef CAS PubMed .
  46. J. Chang, Y. Kuga, I. Mora-Sero, T. Toyoda, Y. Ogomi, S. Hayase, J. Bisquert and Q. Shen, Nanoscale, 2015, 7, 5446–5456 RSC .
  47. Y. Zhang, C. Ding, G. Wu, N. Nakazawa, J. Chang, Y. Ogomi, T. Toyoda, S. Hayase, K. Katayama and Q. Shen, J. Phys. Chem. C, 2016, 120, 28509–28518 CAS .
  48. Y. Zhang, G. Wu, I. Mora-Seró, C. Ding, F. Liu, Q. Huang, Y. Ogomi, S. Hayase, T. Toyoda, R. Wang, J. Otsuki and Q. Shen, J. Phys. Chem. Lett., 2017, 8, 2163–2169 CrossRef CAS PubMed .
  49. C.-H. M. Chuang, A. Maurano, R. E. Brandt, G. W. Hwang, J. Jean, T. Buonassisi, V. Bulović and M. G. Bawendi, Nano Lett., 2015, 15, 3286–3294 CrossRef CAS PubMed .
  50. D. Bozyigit, S. Volk, O. Yarema and V. Wood, Nano Lett., 2013, 13, 5284–5288 CrossRef CAS PubMed .
  51. T. Minami, in Semiconductors and Semimetals, ed. B. G. Svensson, S. J. Pearton and C. Jagadish, Elsevier, 2013, vol. 88, pp. 159–200 Search PubMed .
  52. H. C. Leventis, F. O’Mahony, J. Akhtar, M. Afzaal, P. O’Brien and S. A. Haque, J. Am. Chem. Soc., 2010, 132, 2743–2750 CrossRef CAS PubMed .
  53. M. Eita, A. Usman, A. a. O. El-Ballouli, E. Alarousu, O. M. Bakr and O. F. Mohammed, Small, 2015, 11, 112–118 CrossRef CAS PubMed .

Footnote

Electronic supplementary information (ESI) available. See DOI: 10.1039/c8nh00030a

This journal is © The Royal Society of Chemistry 2018