Zhehao
Zhu‡
ab,
Satish Kumar
Iyemperumal
b,
Kateryna
Kushnir
c,
Alexander D.
Carl
a,
Lite
Zhou
d,
Drew R.
Brodeur
a,
Ronald L.
Grimm
a,
Lyubov V.
Titova
c,
N. Aaron
Deskins
b and
Pratap M.
Rao
*de
aDepartment of Chemistry and Biochemistry, Worcester Polytechnic Institute, Worcester, MA 01609, USA
bDepartment of Chemical Engineering, Worcester Polytechnic Institute, Worcester, MA 01609, USA
cDepartment of Physics, Worcester Polytechnic Institute, Worcester, MA 01609, USA
dMaterials Science and Engineering Graduate Program, Worcester Polytechnic Institute, Worcester, MA 01609, USA. E-mail: pmrao@wpi.edu; Tel: +1 508-831-4828
eDepartment of Mechanical Engineering, Worcester Polytechnic Institute, Worcester, MA 01609, USA
First published on 26th September 2017
Bi2S3 is a non-toxic n-type semiconductor, which has been commonly synthesized in the form of quantum dots or nanocrystalline films by solution deposition methods. Despite a favorable optical band gap of ∼1.3 eV, such films have not achieved high solar energy conversion efficiencies to date. We hypothesize that this is in part due to the presence of sulfur vacancies that, according to our density functional theory calculations, form a deep trap state in the band gap of Bi2S3, which can act as a strong recombination channel for photoexcited charges. Here, we report a microcrystalline Bi2S3 thin-film synthesized by annealing solution-deposited nanocrystalline Bi2S3 in a sulfur vapor environment at 445 °C, which simultaneously increases the grain size and phase purity of Bi2S3, fills in sulfur vacancies, and improves optical absorption. Time-resolved terahertz spectroscopy (TRTS) reveals that sulfur annealing increases the photoexcited carrier lifetime from sub-picosecond to ∼30 picoseconds, while the internal quantum efficiency of a photoelectrochemical solar cell device is increased 4-fold from ∼10% to ∼40%. In addition, TRTS reveals that the intra-grain carrier mobility in the sulfur-annealed films is ∼165 cm2 V−1 s−1 and the long-range mobility is ∼111 cm2 V−1 s−1 at short times, indicating that carriers are able to hop across grain boundaries. These results indicate that annealing in sulfur vapor can produce simultaneously high light absorption and charge separation efficiencies by achieving carrier diffusion length that is comparable to the light absorption depth, leading to high solar energy conversion efficiencies in Bi2S3.
In this work, we synthesize microcrystalline Bi2S3 thin films by high-temperature annealing of solution-deposited nanocrystalline Bi2S3 in a sulfur vapor environment. The sulfur annealing increases grain size and phase purity, fills in sulfur vacancies, and improves optical absorption. Time-resolved terahertz spectroscopy (TRTS) reveals that sulfur annealing increases the photoexcited carrier lifetime from sub-picosecond to ∼30 picoseconds, while the internal quantum efficiency of a photoelectrochemical (PEC) solar cell device is increased from ∼10% to ∼40%. In addition, TRTS reveals that the intra-grain carrier mobility in the S-annealed films is ∼165 cm2 V−1 s−1 and the long-range mobility is ∼111 cm2 V−1 s−1 at short times, indicating that carriers are able to hop across grain boundaries. These results indicate that annealing in sulfur vapor can produce simultaneously high light absorption and charge separation efficiencies by achieving a carrier diffusion length that is comparable to the light absorption depth, leading to high solar energy conversion efficiencies in Bi2S3.
The morphologies, crystal structures, and chemical compositions of the Bi2S3 thin films were characterized by scanning electron microscopy (SEM, JEOL 7000F, 10 kV), transmission electron microscopy (TEM, JEOL 2010F, 200 kV), parallel beam X-ray diffraction (XRD, PANalytical Empyrean, Cu-Kα, 45 kV, 40 mA), and X-ray photoelectron spectroscopy (XPS, PHI 5600, Al-Kα, 13.5 kV, 300 W). The crystallite size of the un-annealed Bi2S3 nanocrystals was determined by TEM imaging of a single layer of Bi2S3 deposited directly onto a TEM grid using the above-described solution deposition method. The TEM grid consisted of ultrathin carbon film supported by a lacey carbon film on a 400 mesh copper grid (Ted Pella). The average crystallite size of the sulfur-annealed films was calculated from the measured XRD pattern using the Scherrer equation:19
(1) |
The wavelength-dependent optical absorption properties of the samples were obtained using illumination from a Xe lamp (Model 66902, Newport). Two spectrometers (USB 2000+ and Flame-NIR, Ocean Optics) were used to measure the incident, transmitted and reflected light at UV-visible and near-infrared regions, respectively. Bi2S3 thin films were prepared on quartz slides (1 mm thick, Ted Pella) for the optical measurements to minimize diffuse scattering by FTO substrates. For both the transmission and reflection measurements, light was incident at a 45° angle to the back-side (quartz) surface of the sample. For the transmission measurements, the spectrometers were aligned with the incident light to capture the transmitted light (T). For the reflection measurements, the spectrometers were placed at a 90° angle to the incident light to capture the reflected light (R). The absorption efficiency was calculated using
A(λ) = 100% − T(λ) − R(λ). | (2) |
The light absorption depth (δ) was then calculated as
(3) |
A PHI5600 XPS system acquired all photoelectron spectra. The instrument utilized a monochromated Kα Al source and a third-party data acquisition system (RBD Instruments, Bend Oregon). Analysis chamber base pressures were <1 × 10−9 torr. A hemispherical energy analyzer that was positioned at 90° with respect to the incoming X-ray flux and 45° with respect to standard sample positioning collected the photoelectrons. High-resolution XP spectra employed a 23.5 eV pass energy, 25 meV step size, and a 50 ms dwell time per step. High-resolution spectra quantified Bi 4f, S 2s (due to the spectral overlap and difficulty in resolving S 2p from Bi 4f), O 1s, and C 1s regions of the photoelectron spectrum. The XPS spectra were calibrated based on a binding energy of 284.8 eV for adventitious carbon. Post-acquisition data fitting of the spectral features utilized a Shirley-shaped background and GL(70) functional peak shapes for Bi 4f spectra while fits to the S 2s region utilized a linear background and GL(30) functional peak shapes.20,21 The Bi 4f doublets were assigned identical full width at half maximum (FWHM) peak widths with areas for 4f5/2 peaks containing 75% of the area of 4f7/2 peaks.22 To prevent mathematically optimized but physically unrealistic peak widths, all features in the S 2s spectral region were constrained to identical FWHM values.
The photoconductivity and photoexcited carrier lifetime were analyzed using the TRTS technique. The charge carriers were excited in the sample with an ultrafast optical pump pulse with photon energy above the band gap, and the transient photoinduced conductivity was monitored in transmission using a time-delayed THz pulse. Here we used 100 fs duration, 400 nm pulses derived from a 1 kHz amplified Ti:sapphire laser source for excitation. Absorption of both un-annealed and S-annealed Bi2S3 films was similarly high at 400 nm. Thus, comparing transient photoconductive response at this excitation wavelength allows us to focus exclusively on the effect of annealing on microscopic conductivity and carrier lifetime. THz probe pulses were generated by optical rectification of 800 nm, 100 fs pulses from the same laser source in a [110] ZnTe crystal, and coherently detected by free-space electro-optic sampling in a second [110] ZnTe crystal.
The PEC measurements were performed in a three-electrode configuration, using a potentiostat (Model SP-200, BioLogic) under back-side broadband illumination from a Xe lamp. The incident light intensity from the Xe lamp at each wavelength was measured by a spectrometer. The integrated power of the Xe lamp output at wavelengths shorter than 950 nm (1.3 eV) was 81.7 mW cm−2, as compared to 71.0 mW cm−2 for the standard AM 1.5G spectrum (Fig. S1a†). Current density–voltage curves (J–V curves) were measured at a scan rate of 10 mV s−1. J–V curves in aqueous electrolytes were measured in a three-electrode configuration with the Bi2S3 photoanode as the working electrode, a Pt wire (0.5 mm diameter, 99.99%, Sigma Aldrich) as the counter electrode, and a saturated calomel (SCE) reference electrode (Gamry). The aqueous electrolyte used was 0.3 M Na2S electrolyte (pH ≈ 13). Potentials (in volts) in aqueous electrolytes are reported versus the reversible hydrogen electrode (RHE) using
VRHE = VSCE + 0.244 + [0.059 × pH]. | (4) |
The light absorption efficiency (ηabs) is the fraction of incident photons that are absorbed. The charge separation efficiency (ηsep) is the fraction of absorbed photons that reach the semiconductor/electrolyte interface. The product of light absorption efficiency (ηabs) and charge separation efficiency (ηsep) was calculated at each potential V by
(5) |
The incident photon-to-current efficiency (IPCE), also known as external quantum efficiency (EQE), was measured at 0.6 VRHE, which is the onset potential for sulfide oxidation in the dark. A Xe lamp equipped with a monochromator (Cornerstone 130 1/8 m, Newport) was used as the illumination source. The spectral irradiance of monochromatic light at each wavelength was measured by a spectrometer and reported in Fig. S1b.† The IPCE was calculated using
(6) |
(7) |
Density functional theory (DFT) calculations were implemented by the Vienna ab initio simulation package (VASP) code.23,24 We used the generalized gradient approximation (GGA) exchange and correlation functionals as parameterized by Perdew, Burke, and Ernzerhof (the PBE functional).25,26 The electron–ion interactions were treated within the framework of the standard frozen-core projector augmented-wave (PAW) method with valence configurations of 6s26p35d10 for Bi and 3s23p4 for S.27,28 An energy cut-off of 400 eV was used in the plane-wave basis-set expansion. Gaussian smearing with a width of 0.2 eV was used for ionic relaxation and the tetrahedron method with Blöchl corrections was used for density of states (DOS) calculations. The Grimme D3 correction method was used to account for dispersion interactions between layers of Bi2S3.29 For calculations of the pristine Bi2S3 bulk unit cell (1 × 1 × 1), a 6 × 2 × 2 Monkhorst–Pack30k-point sampling was used for ionic relaxation. For defect calculations, a bulk supercell (3 × 1 × 1) containing 60 atoms was constructed and spin-polarized DFT calculations were performed (Fig. S2†). For calculations of the bulk supercell, a 2 × 2 × 2 k-point sampling was used for ionic relaxation and a higher 8 × 8 × 8 k-point sampling was used for density of states calculations. Electronic band structure calculations were performed with 50 k-points in each high symmetry direction within reciprocal space of the crystal. Defect formation energies were calculated by
ΔEdefective = Edefective − Estoichiometric + ∑ndefectμdefect, | (8) |
The light absorption efficiency of the Bi2S3 thin film significantly increases at longer wavelengths after sulfur vapor annealing (Fig. 2a). From the measured optical absorption spectra, the indirect band gaps can be determined from the (αhν)1/2vs. hν Tauc plot (Fig. 2b). The S-annealed Bi2S3 was thus determined to possess an indirect band gap of ∼1.24 eV. The experimentally determined band gap corresponds well to the theoretically predicted fundamental band gap, which is found to be indirect and occurs in the ΓX region of the Brillouin zone,32 with energy of 1.25 eV (Fig. 2c). However, the un-annealed Bi2S3 appears to have a larger band gap of ∼1.37 eV, which may be due to quantum confinement caused by nanoscale grain size. The S-annealed Bi2S3 has higher absorption at longer wavelengths (red-shifted) due to the increased crystallite size and lack of quantum confinement. The sub-bandgap absorption of the un-annealed Bi2S3 reaches a minimum value of ∼2.6% at 950 nm (1.31 eV), which suggests that the diffuse scattering by quartz substrate is negligible. Moderate sub-bandgap absorption is observed for the S-annealed Bi2S3, which is attributed to electronic transitions from the defects states to the conduction band, as will be discussed later along with IPCE. Overall, the S-annealed films absorb 67.0% of above-gap photons, compared to only 50.1% for the un-annealed films, despite the films having nearly identical thickness.
X-ray photoelectron spectroscopy was used to quantify the chemical compositions of both un-annealed and S-annealed Bi2S3 films. Fig. 3 reports the XP spectra for the (a) Bi 4f and S 2p regions and the (b) S 2s region of the photoelectron spectra both for (top) un-annealed and for (bottom) S-annealed Bi2S3 films. Prior studies ascribed peaks at 158.9 eV to the Bi 4f7/2 feature from Bi2S3 and 159.3 eV to the corresponding feature from Bi2O3.33 For the Bi 4f region of the un-annealed film (Fig. 3a, top), fits reveal a feature at 158.6 eV (red) and a feature at 159.4 eV (blue). In agreement with prior studies, we ascribe the fit of the red doublet to Bi2S3 and the fit of the blue doublet to Bi2O3 in both the un-annealed and sulfur-annealed Bi 4f spectra. Notably, both Bi2S3 and a large amount of Bi2O3 are present on the surface of the un-annealed Bi2S3, which we interpret to indicate film oxidation either during the solution deposition process or by exposure to the air ambient. For the S 2s region of the un-annealed film (Fig. 3b, top), fits are dominated by a feature at 225.8 eV (red) that we attribute to S2− in the Bi2S3. Additionally, the S 2s spectrum for the un-annealed film contains a smaller feature at 228.1 eV (green) that we attribute to neutral sulfur and two features above 231 eV (blue) that we attribute to highly oxidized sulfur species, denoted SOx. The spectra at the bottom of Fig. 3 show each respective region following the sulfur annealing. The Bi 4f region of the S-annealed film (Fig. 3a, bottom) demonstrates significantly smaller Bi 4f features due to bismuth oxide as compared to the oxide features in the un-annealed film. Concomitant with the sulfur treatment, the S 2s region in the bottom of Fig. 3b demonstrates higher concentrations of the neutral sulfur (S0) relative to the un-annealed film. We attribute the excess S0 feature to trace sulfur from the annealing process that may be incorporated as interstitial sulfur in the Bi2S3 crystal and/or residual surface sulfur. Interestingly, the sulfur annealing step did not attenuate the SOx features (blue) relative to the un-annealed film.
The influence of defects on the electronic structure was then analyzed by calculating the DOS of pristine and defect-containing Bi2S3, along with their formation energies (Fig. 4 and S3†). We hypothesize that the sulfur vapor annealing fills in sulfur vacancies of the un-annealed Bi2S3. The defect states related to a sulfur vacancy (Sv) are occupied electronic states found deep in the band gap, at 0.63 eV above the valence band maximum (VBM). This finding is consistent with previous reports that Sv creates deep hole trapping states that allow electron–hole recombination.34–37 Additionally, Bi2S3 containing two Sv possesses a higher DOS population for the mid-gap charge trapping states, which further shows that Sv in Bi2S3 can act as recombination sites (Fig. 4). Therefore, the un-annealed Bi2S3 may contain a higher concentration of mid-gap Sv states, which increases recombination and decreases photoexcited carrier lifetime. Moreover, the S-annealed Bi2S3 likely contains more sulfur interstitials (Si) due to the low formation energy of this defect in a S-rich environment,36 which is consistent with the greater amount of elemental sulfur species observed on the surface of the film by XPS. However, Si only creates shallow hole trapping states that are about 0.12 eV above the VBM. In contrast to the Sv defect states, the trapped holes at Si are less likely to recombine with the photoexcited electrons due to the smaller energy required to remove the holes to the valence band and avoid recombination (Fig. 4). On the other hand, the un-annealed Bi2S3 likely contains more oxygen impurities in the forms of oxygen substituting for sulfur (OS) and oxygen interstitial (Oi), the formation of which is thermodynamically favorable due to the negative formation energies, −0.93 and −0.57 eV, respectively (Fig. S3†). OS–Bi2S3 exhibits similar electronic structure to pristine Bi2S3. This is expected, since oxygen has the same number of valence electrons as sulfur and is more electronegative. Oi–Bi2S3 has a similar DOS to Si–Bi2S3, resulting in shallow electronic states 0.12 eV above the VBM. Therefore, the oxide impurities in the un-annealed Bi2S3 seem to only cause shallow defect states, which may not contribute to the poor performance of the film.
We then analyzed the impact of sulfur annealing on photoconductivity and photoexcited carrier lifetime using TRTS. We interrogated the pump-induced changes in the sample conductivity with sub-picosecond time resolution by varying the delay between the pump pulse and THz probe pulse. The change in transmission of the main peak of the THz probe pulse, –ΔT/T, as a function of time after excitation with 378 μJ cm−2 pulse, for both un-annealed and S-annealed films is shown in Fig. 5a. It is proportional to the transient photoconductivity.38–40 At the same excitation conditions, peak photoinduced conductivity is more than twice higher in the S-annealed film as a result of a less prevalent fast trapping and recombination of photoinjected charge carriers on a timescale that is shorter than our instrumental response time of ∼300 fs. Moreover, photoconductivity of un-annealed film is very short-lived and decays on sub-picosecond time scale due to both rapid carrier recombination inside the small Bi2S3 crystallites and fast trapping of photoexcited carriers in trap states at crystallite boundaries. Larger grain size, lower concentration of sulfur vacancies, and reduced number of interface defects in S-annealed Bi2S3 film translates into a significantly longer lifetime of mobile carriers (Fig. 5a). Photoconductivity in the S-annealed sample follows a bi-exponential decay with a fast (∼3 ps) and a slower (30 ps) component. The fast decay time is dependent on excitation fluence, increasing from 2.6 ps to 4.4 ps as excitation is decreased 6-fold to 63 μJ cm−2 (Fig. S4†). This suggests that the process responsible for this fast photoconductivity decay is recombination of mobile carriers. On the other hand, the slower, 30 ps component is fluence-independent and represents trapping of carriers at sulfur vacancies and interface states that do not become saturated in the studied fluence range.
Fixing the pump-probe delay, and detecting the pump-induced changes in the amplitude and phase of the THz pulse transmitted through the sample allows us to extract the complex-valued, frequency-resolved THz photoconductivity spectrum at a given time following the photoexcitation. Fig. 5b shows the real (σ1) and imaginary (σ2) conductivity of the S-annealed Bi2S3 film at 2, 3, 5 and 10 ps after excitation with a 378 μJ cm−2 pulse. The real component of conductivity decreases with time while the imaginary one stays almost unchanged and close to zero. We analyze this progression by fitting both real and imaginary conductivity to the Drude–Smith model, a phenomenological model of microscopic THz conductivity in granular materials where the grain size is comparable to the carrier mean free path.38,41–48 It allows the extraction of the instantaneous photoexcited carrier density N, the effective scattering time τDS, and a measure of carrier localization within individual grains at a given time after excitation:
(9) |
1/τDS = 1/τbulk + 1/τboundary. | (10) |
Lines in Fig. 5b are Drude–Smith fits to the experimental complex conductivity. The resulting carrier density is plotted as circles in Fig. 5a, and it follows the same trend as the bi-exponential decay of photoconductivity, indicating that trapping and recombination of mobile carriers is responsible for the observed transient reduction in photoconductivity while the intrinsic carrier mobility stays unchanged. Effective carrier scattering time τDS is unchanged over the first 10 ps and equal to 23 ± 5 fs, corresponding to the carrier mean free path of ∼3 nm. As the mean free path is significantly smaller than the ∼45 nm average grain size, contribution of grain boundary scattering to the effective relaxation time is negligible, and intrinsic, intra-grain mobility (μint) can be estimated as
(11) |
However, long-range, inter-grain mobility decreases as carriers, while free to move within the individual grains, becomes more localized inside those grains. This phenomenon is reflected in the time dependence of the c parameter, which changes from ∼−0.43 ± 0.04 at 2 ps to −0.63 ± 0.04 at 10 ps after excitation, as has been observed in other polycrystalline systems.38,43 There are two possible explanations for this behavior, which we are not presently able to distinguish between. One possibility is that, as carriers get trapped at inter-grain boundaries, the electrostatic field associated with those carriers increases the height of the potential barrier experienced by the free carriers.51 Another possible explanation is that the semiconductor goes towards a flat-band condition under illumination due to the large concentration of photoexcited carriers, resulting in a decrease in an already-existing potential barrier at the grain boundaries, and that this potential barrier subsequently returns to its original height as the photoexcited carriers recombine.52 Regardless of the mechanism, the long-range, dc-mobility of the film, μdc = (1 + c)μint decreases as a function of time after excitation from 111 cm2 V−1 s−1 at 2 ps to 61 cm2 V−1 s−1 at 10 ps. Based on a carrier lifetime (t) of 30 ps and an inter-grain mobility (μdc) of 61 cm2 V−1 s−1, the carrier diffusion length (LD) can be calculated as
(12) |
This carrier diffusion length is similar to the light absorption depth at ∼700 nm (Fig. S5†), which suggests that simultaneously high light absorption and charge separation efficiencies can be achieved in the S-annealed Bi2S3 thin films, leading to high overall solar energy conversion efficiencies.53 On the other hand, for un-annealed Bi2S3 films, low signal/noise ratio in the complex conductivity spectra did not permit accurate measurement of mobility.
The solar energy conversion efficiency of the 10 layers of Bi2S3 thin films with and without sulfur vapor annealing at 445 °C was evaluated by photoelectrochemical (PEC) measurements of the wavelength-dependent photon-to-current efficiencies and the potential-dependent light absorption and charge separation efficiencies (Fig. 6). The film thickness and annealing temperature were optimized for maximum photocurrent under white light illumination (Fig. S6†). The IPCE of the S-annealed films reaches a plateau of ∼40% at short wavelengths, compared to a plateau of ∼5% for the un-annealed films, and is higher than that of the un-annealed films at all wavelengths (Fig. 6a). The IPCE of the S-annealed films shows a sharp rise at wavelengths shorter than 1000 nm, which is consistent with the measured band gap of 1.25 eV, while the IPCE of the un-annealed films show a rise at wavelength shorter than 900 nm, which is consistent with the measured band gap of 1.37 eV. As mentioned earlier, the smaller band gap of the S-annealed Bi2S3 is consistent with the larger grain size of the films and a resulting lack of quantum confinement. In addition, IPCE shows sub-bandgap photon-to-current conversion up to 1100 nm for the S-annealed Bi2S3 film and up to 1050 nm for the un-annealed Bi2S3. The sub-bandgap IPCE for the un-annealed Bi2S3 is likely attributed to the electronic transition from the Oi defect states to the CBM (1.10 eV; see Fig. S3†) due to the expected abundance of Oi in the un-annealed film. The sub-bandgap IPCE for the S-annealed Bi2S3 is likely due to transition from the Si defect states to the CBM (0.93 eV; see Fig. 4) due to the large amount of Si expected in the film. Moreover, the APCE of the S-annealed films reaches a value of ∼50% at short wavelengths, compared to a maximum value of ∼15% for the un-annealed films, and is higher than that of the un-annealed films at all wavelengths (Fig. 6b). It should be noted that the measured quantum efficiencies do not represent the optimum performance of the Bi2S3 thin films as the applied voltage is lower than the built-in voltage that could be achieved in a photovoltaic device.11 Additionally, the product of ηabs and ηsep is calculated from the J–V curves under white-light illumination (Fig. 6c), and reaches a value of ∼35% for the S-annealed films but only ∼5% for un-annealed films at 1 VRHE. At the same voltage, ηsep (Fig. 6d) reaches a value of ∼55% for the S-annealed films, but only ∼10% for un-annealed films. The enhanced quantum efficiency and charge separation efficiency after sulfur annealing can be explained by an increased photoexcited carrier lifetime, consistent with the results of the THz spectroscopy characterization. Although the charge-carrier mobility was measured by THz spectroscopy only for the S-annealed films, it is very likely that the mobility is also higher than that for un-annealed films, further explaining the increased performance.
The Bi2S3 thin films exhibit n-type conductivity according to these PEC measurements, which is consistent with other reports.6,9,12 The n-type conductivity of Bi2S3 has been attributed to the presence of Sv and Si donor defects.36 However, for defect states to behave as effective donor levels, the energy difference between the defect states and the conduction band minimum (CBM) should be comparable to the thermal energy at room temperature (0.026 eV). Therefore, Sv and Si defect states, which have energies that are 0.52 and 0.93 eV below the CBM, respectively, cannot act as effective donor levels, and cannot be responsible for the n-type conductivity of Bi2S3. We additionally calculated the formation energies and density of states for Bi2S3 containing hydrogen substituted for sulfur (HS), or hydrogen interstitials (Hi). The DOS of HS and Hi– Bi2S3 (Fig. 7) show occupied states within the conduction band of Bi2S3, which suggests that these hydrogen impurities can create effective donor levels in Bi2S3 without introducing mid-gap charge trapping states. In fact, hydrogen impurities have previously been to found to cause the n-type conductivity of other semiconductors, such as BiVO4.54 Moreover, the formation energy of Hi is negative and the formation energy of HS is smaller than those of Sv and Si, which indicates that the incorporation of hydrogen is energetically favorable. Hydrogen may be incorporated into Bi2S3via the decomposition of precursors during the solution deposition process and/or from water vapor in the annealing environment. Although this requires further investigation, we propose that hydrogen impurities may play a role in causing the n-type conductivity of Bi2S3.
Fig. 7 DFT-calculated density of states (DOS) and formation energies of Bi2S3 containing hydrogen substitution of sulfur (HS) and hydrogen interstitial (Hi). |
Footnotes |
† Electronic supplementary information (ESI) available: Spectral output of illumination sources, optimized geometries of defect-containing Bi2S3 supercell, additional results on DOS, TRTS, light absorption, and PEC performance. See DOI: 10.1039/c7se00398f |
‡ Present addresses: Department of Materials Science and Engineering, Northwestern University, Evanston, IL 60208, USA. |
This journal is © The Royal Society of Chemistry 2017 |