Improved Open- Circuit Voltage in ZnO–PbSe Quantum Dot Solar Cells by Understanding and Reducing Losses Arising from the ZnO Conduction Band Tail

Colloidal quantum dot solar cells (CQDSCs) are attracting growing attention owing to significant improvements in efficiency. However, even the best depleted-heterojunction CQDSCs currently display open-circuit voltages (VOCs) at least 0.5 V below the voltage corresponding to the bandgap. We find that the tail of states in the conduction band of the metal oxide layer can limit the achievable device efficiency. By continuously tuning the zinc oxide conduction band position via magnesium doping, we probe this critical loss pathway in ZnO–PbSe CQDSCs and optimize the energetic position of the tail of states, thereby increasing both the VOC (from 408 mV to 608 mV) and the device efficiency.


Section S1. Optimized Conditions for Synthesizing Zn 1-x Mg x O by Atmospheric ALD
Atmospheric Atomic Layer Deposition (AALD) system. [1,2] Afterwards, the substrate moves back through the gas channels in the opposite direction, and so forth to obtain the desired film thickness. The precursors were introduced to the gas phase by bubbling carrier gas (N 2 ) through the liquid precursors. The AFM images, obtained using an Agilent 5500 SPM in  [3][4][5] The ZnO (110) peak for x = 0 is not noticeable above the background because undoped ZnO is strongly c-axis oriented and few ZnO (100) and (110) planes (which run parallel to the c-axis) diffract. The ZnO (110) peak has a lower intensity than the ZnO (100) peak, which itself is barely noticeable. But these peaks become more intense with Mg doping as the grains become more a-axis oriented.

Section S2. Pulsed Laser Deposited Zn 1-x Mg x O
The AALD Zn The PLD films were highly crystalline and c-axis oriented ( Figure S2a), and had bandgaps that corresponded with those of the AALD films with the same Mg doping (as seen from the excitonic peaks in Figure S2b and the onset of absorption in Figure S2c). The presence of band tails, evident in Figure S2b The CQDSCs showed an increase in the V OC and efficiency with Mg doping, as expected, and the change in the J SC s and FFs did not exceed the uncertainties in those quantities (Table S2).  The PLD films were rougher than the AALD films, as can be seen from the topography images in Figure S3. The AALD film was highly conformal to the substrate (as seen by comparing Figure S3a with Figure S3c) and the roughness was mainly due to the ITO substrate. The high degree of smoothness would have led to the deposition of a uniform PbSe QD layer. By contrast, the PLD films had large vertical variations in thickness (~30 nm) that were significant compared with the 100 nm thick PbSe QD layer ( Figure S3b and S3d). The PbSe QD layers deposited onto these PLD films would have been less uniform, which may have affected light absorption and charge transport within the film, resulting in the observed reduction in device perfromance as compared with the devices using AALD Zn 1-x Mg x O.
Other factors, such as Zn 1-x Mg x O carrier properties, compactness, transmittance and the influence of high temperature processing on the ITO may also have influenced the The PDS measurements show absorption due to defect states within the bandgap, in addition to the tail of sub-bandgap states. To deconvolute these contributions, we fitted the Urbach model to the absorption front. The Urbach model is given by Eq. S1: [7] U where E U is the Urbach energy, α the absorption coefficient, E g the bandgap, hν the photon energy and α 0 a constant. [7] The Urbach fit was applied to the absorption front in the log-linear absorption plots ( Figure   S5a and S5b). A comparison of the fitted Urbach models to the absorbance data is given in Figure S5a-c, showing that there is a good fit. If we compare the Urbach models themselves to observe the absorption only due to the tail of sub-bandgap states, we can see that as the conduction band was raised, the energy required for sub-bandgap absorption also increased ( Figure S5d). This is in agreement with the photoluminescence spectra data in Figure 3b and suggests that the available electron acceptor levels due to any sub-bandgap states below the Zn 1-x Mg x O conduction band were shifted to higher energies as the conduction band was shifted upwards by Mg doping.

Section S5. Hall-Effect Measurements of AALD Zn 1-x Mg x O Films
The carrier concentration and mobility of the Zn 1-x Mg x O films were measured using the Van der Pauw method in a Hall effect rig. The sheet resistance was first measured with a Keithley 6220 current source and 2182A Nanovoltmeter. A 1 T magnetic field was then applied and the current source and the Nanovoltmeter used to measure the Hall voltage, from which the carrier concentration was calculated. The mobility was calculated from the carrier concentration and resistivity. [1,8]  The decrease in the carrier concentration found ( Figure S6) may have been due to a change in the formation of point defects in the ZnO with Mg doping. [9] The decrease in the carrier concentration observed here was smaller than that previously reported for N-doped ZnO. [10] This is because N has a higher valency than O, meaning that N would remove more free electrons than O, leading to a decrease in the carrier concentration. Mg, however, has the same valency as Zn and should not donate fewer free electrons to the lattice than Zn.
In Figure 2b in the manuscript, there is a decrease in J SC for x > 0.21. As discussed in the manuscript, this was due to a significant extent to the reduction in the accessible density of ZnO band-tail states as the conduction band was raised. The J SC s reduction can also be attributed in part to a reduced electron mobility in the Zn 1-x Mg x O upon doping ( Figure S6), which is expected to arise from an increase in the effective electron mass and alloy disorder scattering. [11] Section S6. Photointensity measurements