Supplementary Figure S1. The maximum possible short circuit current (J sc ) from a solar cell versus the absorber band-gap calculated assuming 100%

Transcription

1 Supplementary Figure S1. The maximum possible short circuit current (J sc ) from a solar cell versus the absorber band-gap calculated assuming 100% (black) and 80% (red) external quantum efficiency (EQE) at all photon energies above the band gap. EQE is limited to a maximum of 80% in QDSSCs and DSSCs due to reflection losses at the FTO/glass substrate of the photoanode. The AM 1.5G solar spectrum is shown (blue) for context (the x-axis becomes photon energy). See Supplementary Discussion below. S2

2 Supplementary Figure S2. Second NREL-certified QDSSC. (a) Current density versus voltage (J-V) characteristics of a QDSSC certified by NREL. (b) EQE spectra for the same device measured at NREL. The fabrication process for this device was the same as that of the device whose certification is shown in Figures 6a and 6b, except that the QDs used were from a different synthesis run, and 2 ML of ZnS was deposited by SILAR on the photoanode following mp-tio 2 sensitization. This device was tested inhouse over a period of months; the results are shown in Fig. 7c and 7d. We measured the efficiency of this device to be 5.51% after 71 days. S3

3 Supplementary Figure S3. Composition as a function of cation exchange temperature. (a) ICP measurements of the QD composition before (0 C) and after varying degrees of Zn cation exchange. (b) ICP measurements of the QD composition before (0 C) and after varying degrees of Cd cation exchange. Clearly, Cd-oleate is more reactive towards the QDs than Zn-oleate leading to a greater degree of cation exchange at low temperatures. Also, Cd seems to have a preference for replacing In in the CISeS QD. S4

4 Supplementary Figure S4: Cu x S cathode stability. (a) Optical images of the Cu x S cathode before and after device fabrication and operation with compositions measured by EDX shown. The higher x values were measured from cathode cross-sections (Fig. 6a) while the lower x values were measured from the top down configuration (Fig. 6b) via SEM. The higher copper content measured in cathode crosssections likely results from a thin layer of pure copper at the Cu x S/FTO interface. (b) Top down SEM images of 100 nm of evaporated copper (left) and Cu x S from the champion device after 71 days (right). S5

5 Supplementary Figure S5: mp-tio 2 wettability. We conducted a simple test of wettability by putting a droplet of water (H 2 O) (left) and methanol (MeOH) (right) onto mp-tio 2 /glass substrates. The water droplet covered < 50% of the surface and had a non-zero contact angle, indicating poor wetting of the mp-tio 2. The methanol droplet rapidly spread across the whole film with a ~0 contact angle, indicating good wetting of the mp-tio 2. S6

6 Supplementary Figure S6: Effect of adding Li + ions to the electrolyte on QDSSC performance. Current density versus voltage characteristics under simulated sunlight for QDSSCs with increasing concentrations of Li 2 S in the electrolyte. The addition of Li 2 S yields Li + counter ions in addition to the majority population of Na + counter ions. A table of the photovoltaic properties can be found in Supplementary Table S7. S7

7 Supplementary Figure S7: QDSSC stability. (a) Short circuit current density (J sc ), (b) open circuit voltage (V oc ), (c) power conversion efficiency, and (d) fill factor as a function of time since fabrication. We observe that the devices improve modestly in the first few weeks after fabrication, mostly due to an increase in the short circuit current, and then stabilize at their highest efficiency. The photovoltaic properties of this device can be found in Supplementary Table S8. S8

8 Supplementary Figure S8: Normalized device parameters under continuous illumination with simulated sunlight (AM 1.5G 100 mw/cm 2 ). After more than 24 hrs of continuous illumination the device efficiency was higher than the initial value. The device was kept under short-circuit (maximum current flow) in between I-V measurements. Each device attribute was normalized to the initial value. No adjustments were made to either the data or the lamp to compensate for changes in the lamp brightness over time. The temperature was not controlled and no efforts were made to cool the device during the test. S9

9 film in MeOH film in H 2 O film in electrolyte in soln % 0.036% 0.576% 20.8% Supplementary Table S1. PL QY of QD-sensitized mp-tio 2 films in solvents vs isolated QDs in solution (absent of TiO 2 ). The peak of the PL for these samples is at ~1.13 ev (1100 nm). The QY is much higher in the sensitized film soaked in the polysulfide electrolyte than in methanol or water because the electrolyte facilitates discharging of photoexcited QDs. This suggests that CISeS QDs which have become positively charged after electron injection into TiO 2 have very rapid non-radiative Auger recombination which can be effectively eliminated if holes are removed by the electrolyte. Further, the relatively high QY of QDs attached to mp-tio 2 in the polysulfide electrolyte indicates that electron transfer to TiO 2 may be sufficiently slow to compete with radiative recombination in the QDs. This result also demonstrates that methanol alone is not effective at removing holes from QDs. S10

10 CISeS 50 C 100 C 150 C SILAR V oc (V) J sc (ma/cm 2 ) FF Efficiency (%) R series (Ω cm 2 ) Supplementary Table S2. Effect of Zn 2+ cation exchange and post-treatment with ZnS by SILAR on QDSSC performance. Table of the photovoltaic properties of the J-V curves shown in Fig. 2b. Treatment with Zn-oleate at 100 C (green) results in the best performance and all cases of Zn-oleate treatment had better performance than SILAR deposition of ZnS. These devices used tba-recapped QDs and a 50% methanol electrolyte but no scattering layer. S11

11 CISeS 50 C 100 C 150 C V oc (V) J sc (ma/cm 2 ) FF Efficiency (%) R series (Ω cm 2 ) Supplementary Table S3. Effect of Cd 2+ cation exchange on QDSSC performance. Table of the photovoltaic properties of the J-V curves shown in Fig. 2d. Treatment with Cd-oleate at 50 C (green) results in the best performance. These devices used tba-recapped QDs and a 50% methanol electrolyte but no scattering layer. S12

12 tba nba sba pyridine MPA 1-T (%) at 1.5eV V oc (V) J sc (ma/cm 2 ) FF Efficiency (%) R series (Ω cm 2 ) Supplementary Table S4. Effect of passivation ligand of the QDs on QDSSC performance. Table of 1 - T (T is transmittance) data at 1.5 ev shown in Fig. 3b and the photovoltaic properties of the J-V curves shown in Fig. 3d. Recapping with tba (green) results in the best performance. These devices used QDs treated with Cd-oleate at 50 C and a 50% methanol electrolyte but no scattering layer. S13

13 tba nba nba 2 sba sba 2 MPA V oc (V) J sc (ma/cm 2 ) FF Efficiency (%) R series (Ω cm 2 ) Supplementary Table S5. Effect of passivation ligands of the QDs on QDSSC performance after 60 days. Table of the photovoltaic properties of the J-V curves shown in Fig. 3e. Recapping with tba (green) results in the best stability and performance. These devices used QDs treated with Cd-oleate at 50 C and a 50% methanol electrolyte but no scattering layer. S14

14 0 % 25 % 50 % 75 % V oc (V) J sc (ma/cm 2 ) FF Efficiency (%) R series (Ω cm 2 ) Supplementary Table S6. Effect of adding methanol to the electrolyte on QDSSC performance. Table of the photovoltaic properties of the J-V curves shown in Fig. 5a. Dilution of the polysulfide electrolyte with methanol is shown to increase J sc (orange) and reduce R series (yellow) (resulting in a higher FF), both of which contribute to the higher efficiency. These devices used QDs treated with Cdoleate at 50 C then recapped with tba but no scattering layer. S15

15 0.0 M 0.05 M 0.10 M 0.20 M V oc (V) J sc (ma/cm 2 ) FF Efficiency (%) R series (Ω cm 2 ) Supplementary Table S7. Effect of adding Li + ions to the electrolyte on QDSSC performance. Table of the photovoltaic properties of the J-V curves shown in Supplementary Figure S6. Addition of Li 2 S to the electrolyte improves J sc slightly (orange) but reduces V oc (yellow) and FF. The V oc drops more readily than J sc increases resulting in the efficiency being highest without any Li 2 S (green). These devices used QDs treated with Cd-oleate at 50 C then recapped with tba and a 50% methanol electrolyte but no scattering layer. S16

16 Day 3 NREL Day 7 Day 9 Day 45 Day 59 Day 71 V oc (V) J sc (ma/cm 2 ) FF Efficiency (%) R series (Ω cm 2 ) Supplementary Table S8. QDSSC stability. Table of the photovoltaic properties of the J-V curves shown in Fig. 7c. We observe that the devices improve modestly in the first few weeks after fabrication then stabilize at their highest efficiency (green). We also note that NREL found our J sc to be higher than we measured (orange), but also found the FF to be lower due to R series being roughly double (yellow), probably due to poor contact. This device used QDs treated with Cd-oleate at 50 C then recapped with tba, had a 75 % methanol electrolyte, had a scattering layer, and had silver paint at the contact points. S17

17 V oc (V) J sc (ma/cm 2 ) FF Efficiency (%) R series (Ω cm 2 ) Supplementary Table S9. Sequential current-voltage measurements, the light soaking effect. Table of the photovoltaic properties of the J-V curves shown in Fig. 7d. We observe that the devices improve modestly with light soaking (green) mostly due to an increase in J sc but also a reduction in R series, which more than make up for the small drop in V oc. This effect is attributed to heating of the electrolyte which results in improved wetting and lower viscosity (i.e. higher ionic mobility). This device (same device as Figure S8) used QDs treated with Cd-oleate at 50 C then recapped with tba, had a 75 % methanol electrolyte, had a scattering layer, and had silver paint at the contact points. S18

18 Supplementary Discussion Comments on the importance of certification. Characterization of the photocurrent, specifically the short-circuit current (J sc ), of a photovoltaic (PV) cell is often trivialized but can be challenging to do accurately for various reasons. These include improperly accounted-for spectral mismatch between the calibration solar cell and the test cell (inherent with simulated sunlight), transient charging effects (from an inappropriate I-V sweep rate), capture of stray or scattered light (from inadequate masking), and inaccurate measurement of the device active area (aperture area). Some of the typical challenges with characterizing J sc can be magnified in quantum dot (QD) sensitized solar cells (QDSSCs) due, for example, to their ability to capture diffuse light and the particularly slow charging and discharging of trap states. In a given study, these factors would remain constant, and thus meaningful conclusions can be drawn; however, when comparing between reports from different groups, these factors may not necessarily be constant. A simple way of validating that a J sc measurement is at least reasonable is to compare the J sc of a cell, along with the cell s external quantum efficiency (EQE, from the photocurrent spectrum), to the theoretical maximum based on the solar spectrum. We calculate the maximum possible J sc (Supplementary Figure S1) as a function of absorber band-gap under AM 1.5 illumination, considering both the total limit (black) and the case of a realistic QDSSC considering the losses due to the top electrode (red). Because of the non-triviality of measuring the true J sc, the traditional PV community relies on a small number of laboratories, such as the National Renewable Energy Laboratory (NREL), to certify the efficiency of solar cells and thus give reliable benchmarks for the progress of cell development. Two of our best cells were certified by the PV Characterization Team at NREL; the results of one device are shown in the main text (Fig. 7a), the other device is shown in Supplementary Figure S2. S19