Introduction

Quantum dot (QD, semiconductor nanocrystal) has been recently drawing great attention as a material for solar energy conversion due to high absorption coefficient, tunable band gap and multiple exciton generation (MEG) effect1,2,3,4. QDs have been actively studied in dye-sensitized solar cell (DSSC) as alternatives to conventional organic dyes5. As a result, remarkable accomplishments have been achieved in recent year, where a power conversion efficiency (PCE) of ca. 5% was achieved using metal chalcogenides6,7 and around 10% using organic-inorganic perovskite8. However, most of the studied QDs for DSSC were based on the relatively large band gap materials (>1.5 eV)6,7,8. Using these materials, it may not be possible to utilize the full solar spectrum, which leads to a limited short-circuit photocurrent density (JSC) of ca. 20 mA/cm2 (considering 20% loss from FTO glass)9. No successful examples have been reported on high JSC QD-sensitized solar cells based on low band gap QDs (<1.5 eV) although it has ability to generate extremely high JSC owing to near IR absorption.

Among the various low band gap QDs, PbS is one of the most intensively studied low band gap semiconductors since it has a high absorption coefficient of 1~5×105 cm−1 and wide range of tunable band gap owing to its large bohr exciton radius of 18 nm10,11. There have been a lot of efforts to use the PbS QD as a sensitizer. Antonio Braga et al. demonstrated a PCE of 2.21% with JSC of 10.91 mA/cm2 using PbS/CdS QD photoelectrode being in contact with polysulfide electrolyte12. Very recently, it was enhanced to 3.82% with JSC of 18.84 mA/cm2 by incorporating PbS/CdS QD in hierarchical porous TiO213. However, the observed JSCs are still far below the theoretical value of 38 mA/cm2 (after considering 20% light reflection at FTO) when considering the band gap of ca. 1.0 eV9.

Main drawback of PbS QD in QD-sensitized solar cell has been argued to be inefficient charge separation and collection due to slow electron transfer kinetics14. Electron injection from the conduction band of PbS QD to the conduction band of TiO2 was measured to be a ca. 100 ns, which is five times slower than CdSe QD (20 ns from colloidal in organic solvent)14. In addition, it was also argued that PbS QD served as critical recombination center, which means that injected electrons in conduction band of TiO2 tends to recombine readily with positively charged PbS QD and thereby efficient charge collection was seriously impeded12,15. For efficient charge separation and collection from the PbS QD, injection and recombination kinetics of PbS QD must be tuned. However, little attempt has been made to date.

Here, we report on a QD-sensitized solar cell with JSC approaching 30 mA/cm2 based on PbS:Hg QDs. Deposition of PbS:Hg QD on nanostructured TiO2 is implemented by simple wet chemical process, where the photovoltaic performance is evaluated by using polysulfide as a redox electrolyte. Noticeable conversion efficiency of 5.6% with unprecedented high JSC is demonstrated under AM 1.5G one sun illumination. Electron injection and charge recombination kinetics are studied by femtosecond transient absorption measurements. Chemical bonding nature surrounding Pb and Hg is investigated by extended X-ray absorption fine structure (EXAFS).

Results

Fig. 1a shows deposition process of PbS:Hg QD on nanostructured TiO2 surface using successive ionic layer adsorption and reaction (SILAR) method. Pb2+ and Hg2+ ions are adsorbed on TiO2 surface during the first dipping, which is followed by chemical reaction with S2− ion in second bath to form PbS:Hg QD on TiO2 surface. Concentration of HgCl2 is varied from 0 mmol to 8 mmol to evaluate the effect of Hg2+ concentration. Fig. 1b shows TEM micrographs of bare TiO2 and PbS:Hg QD-adsorbed TiO2 with different HgCl2 concentration. It is clearly identified that semi-sphered QDs are sparsely formed on the surface of TiO2 nanoparticles regardless of HgCl2 concentration. Pb L3 and Hg L3 edges X-ray absorption near edge spectroscopy (XANES) spectra of bulk PbS and HgS with PbS:Hg QDs deposited on TiO2 nanoparticles are compared in Supplementary Fig. S1. To identify undesired impurity species, PbO, PbO2, Pb(NO3)2 and HgCl2 are also measured as references. All electrodes showed almost identical shape with bulk PbS and HgS, which means that lead and mercury ions in samples exist in the form of Pb2+ and Hg2+ bonded to S2− 16,17. Since it was reported that the peak area in X-ray photoelectron spectroscopic (XPS) data, representing the number of electrons, is assumed to be proportional to the number of atoms in a given state18, the doping concentration is estimated from the area of Pb 4f and Hg 4f peaks (Supplementary Fig. S2, Table S1). As expected, the ratio of Hg2+ to (Pb2+ + Hg2+) increases as the HgCl2 concentration increases. Doping concentration of Hg2+ in PbS:Hg is estimated to be 0, 5.4, 6.3, 14.6 and 16.7 at% for 0, 2, 4, 6 and 8 mmol of HgCl2, respectively. Relatively high amount of Hg content of 16.7 at% for 8 mmol compared to the nominal composition (7.4 at%) is due to the stronger oxidizing characteristic of Hg2+ than Pb2+ 19.

Figure 1
figure 1

Morphology and energetic properties of PbS and PbS:Hg QDs.

(a) Deposition process of PbS:Hg QDs on mesoporous TiO2 nanostructure using successive ionic layer adsorption and reaction (SILAR) method. (b) TEM micrograph of bare TiO2 and QD coated TiO2 nanoparticles using the cationic solution containing 0 mmol, 2 mmol, 4 mmol, 6 mmol and 8 mmol of HgCl2. (c) Tauc plot calculated by using Kubelka-Munk equation from reflectance spectra. (d) UPS spectra of PbS:Hg QD adsorbed TiO2 and extrapolation of low binding energy region. (e) Band edge alignment diagram for PbS and PbS:Hg QDs.

Optical band gap of the QDs is estimated from reflectance spectra using Kubelka-Munk equation20 and Tauc plot21 (Fig. 1c). The optical band gap of PbS:Hg QD increases from 0.94 eV to 1.11 eV as the doping concentration increases from 0 at% to 16.7 at%. UV-vis absorption spectra of PbS:Hg QDs look red-shifted compared to the pristine PbS, especially in visible region as can be seen in Supplementary Fig. S3a. However, the threshold for the PbS doped with Hg (equal or greater than 4 mmol) in the NIR region in inset of Fig. S3a is moved to shorter wavelength than the pristine PbS, which is consistent with the increase in band gap. From the detailed investigation on HgCl2 concentration change from 0 mmol to 2 mmol and from 4 mmol to 6 mmol, optical band gap is almost unchanged in spite of the increased Hg2+ in PbS:Hg, which is related to quantum size effect (ca. 2.4 nm for 0 mmol, ca. 4.0 nm for 2 mmol and 4 mmol and ca. 5.7 nm for 6 mmol and 8 mmol) as can be seen in Supplementary Fig. S4. As mentioned before, increase in Hg doping concentration is expected to enlarge the optical band gap, which, however, will be compensated with increase in QD size. For this reason, little change in band gap is observed between 0 mmol and 2 mmol or 4 mmol and 6 mmol. Valence band maximum (VBM) is determined by ultraviolet photoelectron spectroscopy (UPS)22. Secondary cut-off is fitted to energy of He I light source (21.2 eV), where extrapolation of low energy region corresponds to potential energy of VBM from the vacuum level (Fig. 1d)22. It is noted that VBM is hardly changed by Hg concentration, which indicates that VBM is mainly dominated by S ion. The position of conduction band minimum (CBM) is calculated based on VBM and optical band gap energy. Band edge alignment is demonstrated in Fig. 1e, where CBMs of the QDs move upward as the concentration of HgCl2 increases, leading to gradual increase in optical band gap. PbS doped with Hg is measured to be p-type semiconductor (Supplementary Table S2), where thin film of PbS:Hg deposited on glass substrate is used for the Hall measurement instead of using QD-adsorbed TiO2 film because deposited QD on TiO2 is discontinuous.

Photovoltaic performances of PbS:Hg QD-sensitized solar cells are evaluated from current density-voltage (J-V) and external quantum efficiency (EQE) measurements (Fig. 2). Schematic structure of the QD-sensitized solar cell is depicted in Fig. 2a. Dependence of photovoltaic performance on the SILAR coating cycle number is first investigated at the given TiO2 thickness of 6.5 μm (Supplementary Fig. S5). Optimum coating cycle for best efficiency is found to be 2 cycles for PbS (0 mmol of HgCl2), 4 cycles for both 2 mmol and 4 mmol HgCl2 and 6 cycles for both 6 mmol and 8 mmol HgCl2 (Supplementary Fig. S5 and S6 and Table S3). In case of 0 mmol HgCl2 condition (PbS), efficiency reaches maximum at 2 coating cycles (Supplementary Fig. S6a) although absorption is further increased after 2 cycles (Supplementary Fig. S7a). This might be related to decreased band gap due to large sized PbS QD formed after several coating cycles14 (ca. 5.7 nm for 6 cycle in Supplementary Fig. S4). No further increase in short-circuit current density (JSC) is observed after 2 coating cycle (Supplementary Fig. S6b). On the other hand, PbS:Hg (2~8 mmol HgCl2) shows significant increase in JSC as the coating cycle is increased, which is due to the increased absorption (Supplementary Fig. S6b and Fig. S7b). Fig. 2b shows J-V curves of the PbS:Hg QD-sensitized solar cells depending on HgCl2 concentration and photovoltaic parameters are summarized in Table S4, where ca. 10 μm-thick TiO2 film is used since optimal efficiency is obtained with around 10 μm-thick film from the investigation on dependence of TiO2 film thickness on photovoltaic performance (Supplementary Fig. S8 and Table S5). JSC of QD-sensitized solar cell is substantially improved after Hg2+ doping. By adding 2 mmol of HgCl2 in Pb2+ precursor solution, conversion efficiency is significantly improved from 2.38% to 3.78%, which is mainly due to remarkable increase in JSC from 13.68 mA/cm2 to 22.27 mA/cm2 (63% increment). For the case of Hg2+ doped PbS, as the concentration of HgCl2 increases from 2 mmol to 8 mmol, no significant change in JSC is observed, whereas VOC is gradually increased. Since the fill factors are degraded as photocurrent increase, shunt and series resistances are calculated from slope of J-V curves at short-circuit and at open-circuit region, respectively, using K2400 I-V system program based on LABVIEW (PV Measurements Inc) and plotted in Supplementary Fig. S9. Change in shunt resistance and series resistance are investigated as function of TiO2 film thickness and number of coating cycle. Upon increasing TiO2 film thickness or number of coating cycle, photocurrent is increased, while both shunt and series resistances are decreased. As can be seen in Tables S3 and S5, fill factors are decreased with increasing TiO2 film thickness or number of coating cycle, which is due to the decreased shunt resistance despite the decreased series resistance.

Figure 2
figure 2

Photovoltaic performance and femtosecond transient absorption spectra of PbS and PbS:Hg QDs.

(a) Schematic device structure of QD-sensitized solar cell, (b) J-V curves and (c) Normalized JSC as a function of time for PbS:Hg QD-sensitized solar cells with different HgCl2 concentration. Inset is EQE spectra for PbS and 14.6 at% Hg2+ (6 mmol) doped PbS. (d) Transient absorption spectra at 0.5 ps delay. The wavelength of pump light was 490 nm. (e) Transient absorption time profile and (f) Normalized ΔOD at 2 (± 0.1) ps (red), 20 (± 1) ps (blue) and 300 (± 15) ps (black) with standard deviation. The wavelength of probe was 870 nm. (g) Intensity averaged decay time constants (blue) and rise time constants (red) with error bars.

EQE spectra are demonstrated in inset of Fig. 2b. Maximum EQE of PbS:Hg QD sensitized solar cell reaches 74.6% at 470 nm, which is improved by 78% compared to the maximum EQE of 41.8% at 400 nm for PbS QD sensitized one. EQE is substantially enhanced from visible to near IR region, where 123% at 500 nm (from 33.1% to 74%), 472% at 700 nm (from 9.41% to 53.8%) and 440% at 1000 nm (from 1.89% to 10.2%) are improved. Stability of the devices is evaluated by measuring JSC with respect to time after electrolyte injection (Fig. 2c). JSC of PbS QD sensitized solar cell sharply decreases immediately after electrolyte injection, where almost 80% of JSC degrades during the initial 180 seconds. Instability of PbS QD in polysulfide electrolyte is also observed in previous research12. However, degradation of JSC surprisingly is overcome as the HgCl2 concentration increases. Only 8% degradation of JSC is observed for the 6 mmol sample. This is considered to be attributed to change in chemical bonding nature of Pb-S owing to Hg2+ doping, which will be discussed in detail based on EXAFS analysis.

Femtosecond transient absorption (TA) spectra are measured to analyze the carrier dynamics in PbS:Hg QD-TiO2 system along with pristine PbS QD-TiO2 system. Fig. 2d shows femtosecond TA spectra in the near IR (NIR) region from 840 nm to 1200 nm, where the source of absorption is related to electrons in conduction band of TiO223,24. When considering the wavelength of pump light (490 nm) which is only absorbed by the QD, NIR absorption by electrons in TiO2 conduction band can be ascribed to the injected photoelectrons from PbS and PbS:Hg QD to TiO2. It can be noticed that ΔO.D. increases as the HgCl2 concentration increases and is highest for HgCl2 concentration of 6 mmol. Since ΔO.D. is related to the injected electrons, the higher ΔO.D. is indicative of more populated electrons in TiO2 conduction band, which is responsible not only for higher JSC but also for higher VOC after doping of Hg2+ in PbS. It is expected that increase in ΔO.D. leads to upward shift of Fermi level, resulting in increase in VOC25. The observed ΔO.D. is well correlated with change in VOC measured in Table S4, where VOC increases from 0.387 V to 0.426 V as the HgCl2 increases from 0 mmol to 6 mmol and slightly decreases to 0.424 with 8 mmol. Decreased ΔO.D. for 8 mmol HgCl2 condition might be related to blue shift of absorption caused by the increased band gap. TA time profile in this region exhibits triply exponential decays with an ultrafast rising component as a result of numerical convolutions with cross-correlation function (FWHM = 200 fs). In order to compare the carrier dynamics, ΔO.D. is normalized as a function of time (Fig. 2e). The decay dynamics slow down monotonically with an increased concentration of HgCl2 from 0 mmol to 8 mmol. In Fig. 2f, ΔO.D. at arbitrary time delays of 2 ps, 20 ps and 300 ps are elucidated with different HgCl2 concentration. The ΔO.D. at arbitrary time increases with increased HgCl2 concentration, which have the same trend with the shorter time delay (−1 ps~13 ps). Numerically fitted parameters for decay time profile are listed in Table S6 where triply exponential model (eq 1) is used and average decay time constants (τd) are estimated by eq 226,

where Ai and τi are triply exponential decay components constant. Average decay time constant increases from 151 ps to 186 ps as the HgCl2 concentration increases from 0 mmol to 8 mmol (Fig. 2g). Absorption decay dynamics of electrons in conduction band of TiO2 result from recombination of injected electrons with holes in QDs in absence of electrolyte. Therefore it is obvious that charge recombination in TiO2 is substantially suppressed as the HgCl2 concentration increases, which is responsible for the increased VOC. Also, considering that charge collecting efficiency is inversely proportional to recombination kinetics, suppressed charge recombination may lead to improvement of charge collecting efficiency (Фc)27. Ultrafast rising of ΔO.D. signal in Fig. 2e is originated from injection of photo-excited electrons from QD to TiO223,24. Therefore, it is possible to estimate the electron injection kinetics by measuring the time for ΔO.D. signal to reach its maximum. Fitted rise time constants are listed in Table S6 and Fig. 2g. Rise time constant decreases from 340 fs to 270 fs as the HgCl2 concentration increases from 0 mmol to 8 mmol, which implies an improved electron injection efficiency (Фinj)28. It may not be ruled out that upward shift of PbS QD conduction band upon Hg2+ doping provides a driving force for faster electron injection in PbS:Hg QD than in PbS QD29. However, since the band shift of CBM of QDs seems favoring electron injection toward TiO2, one hypothesis for the improved functional properties of the cell and the mechanism for the inhibition of charge recombination seems to be related to structural reinforcement by Hg doping.

The effect of Hg2+ doping on the local structure of Pb2+ ions in the PbS:Hg QD adsorbed on TiO2 surface is investigated with EXAFS analysis at Pb and Hg L3-edge. As can be seen in Fig. 3a, regardless of Hg doping, all the present PbS:Hg QDs adsorbed on TiO2 show typical EXAFS oscillation of the reference PbS, confirming the formation of rocksalt-structured PbS phase. A closer inspection on the experimental spectra reveals that the wavelength of EXAFS oscillation becomes greater with the increase of Hg content, strongly suggesting the shortening of (Pb-S) bonds upon Hg doping30. The decrease of (Pb-S) bond distance upon the Hg doping is obviously verified by the quantitative EXAFS fitting analysis showing the gradual decrease of (Pb-S) distance with the increase of HgCl2 concentration (Fig. 3c, Fig. 3e and Table 1). This result provides strong evidence for the reinforcement of the (Pb-S) bonds by the doping of Hg2+ ions. In addition to the systematic variation of bond distance, there is an overall decreasing trend of Debye-Waller factor (σ2) with the increase of HgCl2 concentration, indicating the depression of structural disorder around Pb ions (Table 1)31. Such a decrease of structural disorder is also attributable to the reinforcement of (Pb-S) bonds upon the Hg doping. The strengthening of chemical bonds between Pb and S leads not only to the increase of the optical band gap of PbS QD but also to the upward shift of the conduction band of PbS with antibonding character. For local structure of mercury ion, detailed analysis on the Hg L3-edge EXAFS spectra (Fig. 3b, 3d and 3f) is performed as follows. Since the value of coordination number obtained from the EXAFS fitting analysis corresponds to the product of coordination number (CN) and amplitude reduction factor (S02), it is necessary to get the knowledge of amplitude reduction factor for a Hg-S pair, prior to determining the exact coordination number of Hg ion. In the present analysis, the S02 value obtained from the pure HgS reference is used to calculate the exact CN of mercury in the present PbS:Hg QDs. As listed in Table 1, this fitting analysis yields the CN of ~3.7–4 for mercury ions in the PbS:Hg QDs, strongly suggesting the stabilization of Hg ions in the interstitial tetrahedral sites. The tetrahedral symmetry of Hg ions in the PbS:Hg QDs is further confirmed by the comparison of their bond distance with that of pure (bulk) HgS material. In comparison with the reference HgS where the Hg ions are stabilized in the tetrahedral sites with the (Hg-S) bond distance of 2.49 Å, the present PbS:Hg QDs show a shorter bond distance of ~2.36–2.38 Å. Since the decrease of CN shortens the bond distance, the shorter (Hg-S) bond distance of the PbS:Hg QDs than that of the pure HgS underscores that the CN of Hg ion is smaller than or similar to four (i.e. the CN of Hg in the reference HgS). This result provides strong evidence for the stabilization of mercury ions in interstitial tetrahedral sites. If the Hg ions exist in octahedral symmetry (Pb site) with CN = 6, the PbS:Hg QDs should show a longer (Hg-S) bond distance than that of the reference HgS (2.49 Å) with CN = 4. The tetrahedrally coordinated interstitial Hg2+ ions contribute to contraction of lattice and consequently causing a reinforcement of Pb-S bond (Fig. 3g). Theoretically calculated (Hg-S) bond distance for 6 mmol HgCl2 case is 2.386 Å, which is well consistent with the observed bond distance of 2.369 Å. Both the enhancement of (Pb-S) bond strength and the improvement of structural ordering around Pb2+ ions strongly suggest the enhanced structural stability of PbS QD via the Hg doping. In addition, the enhanced structural ordering is likely to lead to reduction of deep trap density32,33, which contributes to suppression of charge recombination. Thus, the structural reinforcement by Hg addition is playing major role in improving the functional properties of the cell and is the main mechanism for the inhibition of charge recombination compared to the role of band shift.

Table 1 Results of non-linear least-squares curve fitting analysis for the Pb and Hg L3 edge spectra of the PbS and PbS:Hg QDs adsorbed on TiO2 with different HgCl2 concentration. Bulk PbS and HgS are references. CN, R and σ2 represent coordination number, interatomic distance and Debye-Waller factor
Figure 3
figure 3

Pb and Hg L3-edge EXAFS analysis of PbS and PbS:Hg QDs.

EXAFS data of Pb (a, c, e) and Hg (b, d, f) L3-edge region. (a),(b) Experimental k3X (k) EXAFS oscillations, (c),(d) Fourier transformed EXAFS data and (e)(f) the Fourier-filtered EXAFS data of (1) bulk PbS, (2) 0 mmol, (3) 2 mmol, (4) 4 mmol, (5) 6 mmol, (6) 8 mmol HgCl2 and (7) bulk HgS. In the (c), (d), (e) and (f), empty circles and solid lines represent the experimental and fitted data, respectively. (g) PbS and PbS:Hg quantum dot unit cell using ball-stick model, showing contraction of Pb-S bond via interstitial occupation of Hg ion.

Best performance of PbS:Hg QD sensitized solar cell in this study is obtained by combination of 6 mmol HgCl2 condition, scattering layer and CsOH additive in electrolyte. Addition of 0.3 M CsOH into the electrolyte improves fill factor of the device by 13% (Supplementary Fig. S10 and Table S7), which is probably due to decrease in interfacial resistance at electrolyte/Cu2S counter electrode34. Light scattering layer increases further EQE at longer wavelength (EQE of 56% at 800 nm and 32% at 1000 nm). As a result, JSC reaches 29.98 mA/cm2, leading to a PCE as high as 5.58% at AM 1.5G one sun illumination (Fig. 4).

Figure 4
figure 4

Photovoltaic performance of PbS:Hg QD-sensitized solar cell.

J-V curve and EQE spectrum (inset) of PbS:Hg QD-sensitized solar cell. HgCl2 concentration was 6 mmol and number of coating cycle was 6. Total TiO2 film thickness was 25 μm, comprising 5 μm with 20 nm TiO2 particles and 20 μm with 40 nm and 500 nm TiO2 particles.

Discussion

We presented a mercury doped PbS QD-sensitized solar cell with extremely high JSC of 30 mA/cm2. Such a high JSC was realized by reinforcing Pb-S bond via incorporation of Hg2+ ion into the interstitial site of PbS lattice. The improved ordering and covalency surrounding Pb ion led to faster electron injection and suppressed charge recombination. It is therefore concluded that structural reinforcement by Hg addition plays a major role in improving the functional properties of the cell and is the main mechanism for the inhibition of charge recombination. Based on our findings, optoelectronic and photovoltaic properties of QD materials are expected to be significantly improved by careful investigation on chemical bonding nature. Furthermore, such a preponderant enhancement in QD photovoltaic materials is believed to provide new insight into super high efficiency solar cell technology such as third generation MEG photovoltaics.

Methods

Sensitization of PbS:Hg QD on mesoporous TiO2 layer

TiO2 nanoparticles and films were prepared according to the method described elsewhere35. Fluorine-doped tin oxide (FTO) glass (Pilkington, TEC-8,8 Ω/sq ) was cleaned with ethanol using ultrasonic cleaner for 15 min and subsequently UV/Ozone cleaning was conducted for 15 min to remove organic contaminants. A dense blocking layer was prepared on the cleaned FTO glass using 0.1 M Ti(IV) bis (ethyl acetoacetato) diisopropoxide (Aldrich, 98%) in 1-butanol (Aldrich, 99.8%) solution. TiO2 nanoparticles were hydrothermally synthesized using titanium isopropoxide (Aldrich, 97%) in autoclave at 230°C. TiO2 paste was prepared by mixing the synthesized nanocrystalline TiO2, terpineol (Aldrich, 99.5%), ethyl cellulose (Aldrich, 46 cp) and lauric acid (Fluka, 96%) with nominal composition of 1.25:4:0.3:0.1. The paste was further treated using three-roll-mill for 40 min to disperse nanoparticles sufficiently. The TiO2 paste was coated on the blocking layer coated FTO substrate using doctor-blade method and annealed at 550°C for 1 h in air. Thickness of the annealed mesoporous TiO2 layers was measured by an alpha-step IQ surface profiler (KLA Tencor). Successive ionic layer adsorption and reaction (SILAR) method was used to sensitize the TiO2 nanoparticles with PbS:Hg QDs as described in Fig. 1a. Briefly, mesoporous TiO2 coated electrode was first dipped in aqueous solution of 0.1 M Pb(NO3)2 (Aldrich, 99.99%) for 1 min, followed by dipped in 0.1 M Na2S (Aldrich) for 1 min. Between each dipping, the electrode was thoroughly washed with DI (deionized) water. For PbS:Hg, 2~8 mmol of HgCl2 (Sigma-Aldrich, 99.5%) was added to cationic precursor solution. These processes were defined as one cycle and several coating cycle was repeated to obtain the optimum photovoltaic performance in each condition (2 cycles for 0 mmol, 4 cycles for 2 mmol and 4 mmol and 6 cycles for 6 mmol and 8 mmol) (Supplementary Fig. S5, Fig. S6 and Table S3). To enhance the photovoltaic performance, ZnS passivation layer was deposited after PbS:Hg deposition on TiO2 by using aqueous solution of 0.1 M Zn(CH3COO)2 (Aldrich, 99.99%) and 0.1 M Na2S36.

Solar cell fabrication

Cu2S formed on a brass sheet was used for counter electrode. Brass foil (Alfa aesar, 0.25 mm thick) was etched in hydrochloric acid (SAMCHUM, 35.0~37.0%) at 80°C for 15 min and subsequently washed with DI water and dried using air gun. Polysulfide solution composed of 1 M Na2S (Aldrich) and 1 M S (Sigma-Aldrich, 99.5%) in DI water was dropped on unmasked area of etched brass and black colored Cu2S immediately formed on the brass foil. The PbS:Hg QD-sensitized working electrode and the Cu2S-brass counter electrode were assembled using double-side silicon-binder-coated polyimide tape (ca. 80 μm) as a spacer and sealant. Polysulfide redox electrolyte composed of 1 M Na2S (Aldrich) and 1 M S (Sigma-Aldrich, 99.5%) in DI water was injected through the 0.75 mm-diameter hole at working electrode and the holes were sealed using polyimide tape. For the best performance, 0.3 M of CsOH (Aldrich, 99.9%, 50 wt% solution in water) was added to electrolyte.

Materials and photovoltaic characterizations

The active area was measured by a digital microscope camera (DCMe 500) equipped with an image analysis program (Leopard 2009). Current density and voltage were measured using solar simulator (Oriel Sol 3A class AAA) equipped with 450 W Xenon lamp (Newport 6279NS) and a Keithley 2400 source meter. Light intensity was adjusted with the NREL-calibrated Si solar cell having KG-2 filter for approximating one sun light intensity (100 mW/cm2). While measuring current and voltage, the cell was covered with an aperture mask. External quantum efficiency (EQE) was measured by a specially designed EQE system (PV measurement Inc.). A 75 W Xenon lamp (USHIO, Japan) was used as a white light source to generate monochromatic beam. Morphology and size of PbS:Hg QDs adsorbed on the nanocrystalline TiO2 surface were investigated using a high-resolution transmission electron microscope (HR-TEM, Jeol, JEM-2100F) at an acceleration voltage of 200 kV. Average diameter of QDs was determined from TEM micrograph using image J program (version 1.45s). Stoichiometry of QDs was analyzed using X-ray photoelectron spectroscopy (VG microtech, ESCA 2000). X-ray source was Al Kα and x axis was fitted for C 1s peak to have 284.6 eV. Removal of background signal and integration of peak was conducted using VGX900-W system. Ultraviolet photoelectron spectroscopy (UPS) spectra were measured with an AXIS-NOVA using a He I (21.2 eV) gas discharge lamp. Absorbance and reflectance was measured by UV/Vis spectrometer (PerkinElmer, Lambda 1050).

Hall measurement

To prepare samples for Hall measurement, chemical bath deposition of PbS and PbS:Hg thin film was conducted as described elesewhere37. 1 M Pb(CH3COO)2, 1 M thiourea and 0.5 M NaOH in DI water were used for deposition of PbS while 0.85 M Pb(CH3COO)2 and 0.15 M HgCl2 were used for deposition of PbS:Hg film. Both of the deposition was performed for 10 min at room temperature, which led to 300 nm thick-PbS and 235 nm thick-PbS:Hg on glass substrate. Hall voltage was measured under bias current of 1 μA for PbS film and 3 μA for PbS:Hg film and applied field was adjusted from 1 to 5 kG using LaKeShore 7500 series Hall Measurement System at room temperature. Resistivity, type of conductivity, sheet carrier density and mobility of majority carrier were calculated by van der Pauw method.

Femtosecond transient absorption measurement

Pump-probe transient absorption (TA) experiments were carried out using commercial Helios system (Ultrafast Systmes). The pump beam of 490 nm with a pulse energy of 1.6 mJ was obtained from an optical parametric amplifier (Light conversion, TOPAS-C), pumped by a regeneratively amplified Ti:Sapphire laser system (Coherent, Libra) with 1-k Hz repetition rate. The probe pulse was a white light continuum generated by focusing a small portion of the amplifier output though a c-axis sapphire window.

XANES and EXAFS measurements

Local structure of QD was analyzed by XANES (X-ray absorption near edge structure) and EXAFS (extended X-ray absorption fine structure). The Pb L3-edge XANES/EXAFS data were collected at room temperature in a fluorescent mode at the beam line 10C at the Pohang Accelerator Laboratory (PAL), Korea. The spectra were calibrated by measuring the spectrum of Pb foil simultaneously. Data analysis for the experimental spectra was performed with the standard procedure by using UWXAFS 2.0 program38. For the fitting of the reference, the amplitude reduction factor (S02) was varied while the CN was fixed to the crystallographic value. All the bond distances (R), Debye-Waller factors (σ2) and energy shifts (ΔE) were allowed to vary. In order to exactly estimate the CNs of lead ions in the present PbS:Hg QDs adsorbed on TiO2, S02 was fixed to the value from the reference PbS.