Bandgap engineering of a lead-free defect perovskite Cs3Bi2I9 through trivalent doping of Ru3+

Inorganic defect halide compounds such as Cs3Bi2I9 have been regarded as promising alternatives to overcome the instability and toxicity issues of conventional perovskite solar cells. However, their wide indirect bandgaps and deep defect states severely limit their photoelectronic conversion efficiency when implemented in devices. Trivalent cation substitution has been proposed by previous calculations allowing the engineering of their band structures, but experimental evidences are still lacking. Herein we use the trivalent cation Ru3+ to partially replace Bi3+ in Cs3Bi2I9, and reveal their structural and optoelectronic properties, as well as the environmental stability. The Ru-doped Cs3Bi2I9 shows a decreasing bandgap with the increasing doping levels and an overall up-shift of band structure, owing to the dopant-induced defect states and thus enhanced phonon–electron coupling. As a result, upon Ru3+ doping, the narrowed bandgap and the upward shift of the band structures might facilitate and broaden their applications in optoelectronic devices.


Introduction
Recently perovskite materials have attracted great attention for their promising applications in optoelectronic devices such as solar cells, light emitting diodes, and photodetectors. [1][2][3][4][5][6] In particular, the photoelectron conversion efficiency (PCE) of hybrid organic-inorganic perovskite solar cells (PSCs) has increased rapidly since 2009, 7 reaching more than 22% up to date. 8 However, the toxicity and instability of these conventional alkylamine-based PSCs have greatly restricted their larger scale production outside laboratories. In this context, photoabsorbers based on inorganic lead-free perovskites have been pursued as potential alternatives, including three major families: AM II X 3 , 9,10 A 2 M I M III X 6 [11][12][13] and A 3 M III 2 ,X 9 14,15 (A ¼ Cs, Rb; M II ¼ Sn, Ge; M I ¼ Ag, Na, K; M III ¼ Bi, Sb;, ¼ vacancy sites; X ¼ Cl, Br, I), each representing the halide perovskites, double halide perovskites and defect halide perovskites, respectively. With the choice of different elements for A, M and X, the bandgap of these compounds can be effectively tuned. 16 Further in each of the above perovskite families, the metal M of various chemical states can be further substituted with alloys containing multiple metals with the same total oxidation state (socalled cation transmutation), 17,18 adding an additional degree of freedom to regulate their bandgaps.
In the defect halide perovskite family A 3 M III 2 ,X 9 , Cs 3 Bi 2 I 9 has been mostly studied with the consideration of Bi 3+ having similar electronic structure and ion radius to Pb 2+ , but much less toxicity. [19][20][21] At room temperature, Cs 3 Bi 2 I 9 exhibits a distorted and defect-modulated hexagonal perovskite structure, in which a pair of [BiI 6 ] 3À octahedral share faces to form a [Bi 2 I 9 ] 3À bioctahedra, and every third layer of the octahedral Bi sites is depleted for charge neutrality, resulting in a zerodimensional (0D) molecular salt crystal structure (space group P6 3 /mmc). 22 The 0D structure with separated [Bi 2 I 9 ] 3À bioctahedra leads to low carrier mobility and therefore poor charge transport, which might be more benecial for optical properties, rather than photovoltaic applications. [23][24][25] For instance, colloidal Cs 3 Bi 2 I 9 nanocrystals have shown unique absorption spectra arising from the 0D crystal structure, with a very high excitonic binding energy of 300 meV. 23 Cs 3 Bi 2 I 9 possesses an average indirect optical bandgap of $2 eV, with variations reported from different studies. 24 DFT calculations indicated a higher effective mass of charge carriers and prominent deep-level defects for Cs 3 Bi 2 I 9 when compared with their Pb counterparts, and pointed out that defect passivation through external doping or nonequilibrium synthesis might lead to better photovoltaic performance. 25 Solar cells with Cs 3 Bi 2 I 9 as the photoabsorber has shown a power conversion efficiency over 1% with enhanced chemical stability and additional room for improvement. 27 Hence, in order to further promote the performance of Cs 3 Bi 2 I 9 based optoelectronic devices, bandgap engineering is essential to narrow the optical bandgap, tune the defect chemistry and optimize the electronic structure. 16,28,29 For example, DFT studies by Hong et al. inferred that the bandgap of Cs 3 Bi 2 I 9 can be reduced by applying dual trivalent metal cations to form Cs 3 BiM III I 9 , 18 although there has been no experimental evidence in this regard.
Discovered since the 1980s', Cs 3 Ru 2 Cl 9 is another compound of the A 3 M III 2 ,X 9 family that also belongs to the P6 3 /mmc space group with bioctahedra [Ru 2 Cl 9 ] 3À . 30,31 Inspired by this fact, herein we attempt to dope the Cs 3 Bi 2 I 9 with various concentrations of Ru 3+ through atomic substitution of Bi 3+ in order to regulate its bandgap and optical properties. As a result, crystals of Cs 3 Bi 2Àx Ru x I 9 with dimensions from several to hundreds of micrometers were obtained by a hydrothermal approach. Through changing the molar ratio of RuCl 3 and BiI 3 , various doping levels of Ru 3+ were realized without altering the crystal structure. Measurements using Raman, UV-Vis, Photoluminescence and Ultraviolet Photoelectron Spectroscopy illustrated that even with a small doping amount, the band structure and optical properties of Cs 3 -Bi 2Àx Ru x I 9 can change signicantly. Consequently, through effective tuning of the bandgap, defect states and electronic structure, Ru-doped Cs 3 Bi 2 I 9 perovskites exhibit a great perspective for future optoelectronic applications.

Materials and preparation
All chemicals were purchased from the Aladdin Company (China) with purity above 99.95% and used without further purication.
Cs 3 Bi 2 I 9 . Microcrystals of Cs 3 Bi 2 I 9 were synthesized by the hydrothermal method. In a 50 mL reactor, 1.5 mmol CsI and 1 mmol BiI 3 were dissolved in 20 mL concentrated HI acid. The reactor was then sealed and placed into a thermal oven with the temperature hold at 120 C. Aer 6 hours, the temperature was turned down at 10 C/30 min to the room temperature. Powders of red colour were precipitated from the precursor solution. The solids were then ltered out, washed repeatedly with ethanol and diluted HI, and nally dried under vacuum at 60 C for overnight.
Cs 3 Bi 2Àx Ru x I 9 (x ¼ 0.02, 0.04, 0.1, 0.2). Microcrystals of Cs 3 Bi 2Àx Ru x I 9 were also synthesized by the hydrothermal method. 1.5 mmol CsI, a mmol BiI 3 and b mmol RuCl 3 (a + b ¼ 2) were dissolved in 20 mL concentrated HI acid in a 50 mL reactor. The reactor was then sealed and placed into a thermal oven at 120 C for 6 h. Aer cooling down, powders of dark red colour were precipitated from the precursor solution. The solids were then ltered out, washed repeatedly with ethanol and diluted HI, and nally dried under vacuum at 60 C for overnight.

Characterization
The powder X-ray diffraction (XRD) measurements were conducted using a D8 Advance diffractometer (Bruker Corporation, Germany) operating at 40 kV and 40 mA with Cu Ka radiation (l ¼ 1.5406Å), and the scanning rate was xed at 4 min À1 . Scanning electron microscopy (SEM) images and Energy Dispersive X-ray spectroscopy (EDX) were taken by a FEI Scios Microscope (Thermo Fisher, USA). Transmission electron microscope (TEM) images and elemental mapping were taken by the FEI 200 kV G20 Microscope (Thermo Fisher, USA). X-ray photoelectron spectroscopy (XPS) and Ultraviolet Photoelectron Spectroscopy (UPS) were performed on an Escalab 250Xi (Thermo Fisher, USA). For UPS measurements, pristine Au was used to calibrate the Fermi level. Raman spectra were obtained on a Jobin Yvon HR Evolution Spectrometer (Horiba, Japan). Fluorescence Spectroscopy was taken with a FLS980 Fluorescence Spectrometer (Edinburgh instruments, UK). Thermogravimetric analysis (TGA) was conducted with the SII TG/DTA7200 (Japan) at a rate of 20 C min À1 ramping from room temperature to 700 C inside a crucible. UV-Vis spectroscopy was recorded with PerkinElmer Lambda 750S (USA).

Crystalline phase and morphology
As reported previously, Cs 3 Bi 2 I 9 crystallizes in the P6 3 /mmc space group with a lattice constant of a ¼ 8.335Å and c ¼ 21.326 A (Fig. 1a). 16 When a portion of Bi 3+ sites in Cs 3 Bi 2 I 9 is substituted with Ru 3+ , the colour of the crystal powder changes from bright to dark red with the increasing doping level (Fig. 1b). The darkened colour is indicative of narrowed bandgap, as more visible light is absorbed by the compound. The PXRD data of Cs 3 Bi 2Àx Ru x I 9 with x ¼ 0, 0.02, 0.04, 0.1 and 0.2 are given in Fig. 1c (also amplied in Fig. S1 †), from which one can see that at low doping levels, the XRD patterns exhibit no obvious change from the original Cs 3 Bi 2 I 9 spectrum, matching well the standard prole of JCPDS no. 39-1346. This suggests the lattice structure of Cs 3 Bi 2 I 9 is not observably affected by low levels of Ru 3+ substitution, likely due to the smaller atomic radius of Ru 3+ than Bi 3+ (Fig. 1a). However, at a higher level of substitution (x ¼ 0.2), more broadened brag peaks can be observed at 26 , 28 , 31 and 44 , which could be resulted from possible structural corruption, as evidenced by morphological changes from the scanning electron microscopy (SEM) characterization (Fig. S2 †) and that none of these peaks is assignable to any raw materials. Fig. 2a and b present the characteristic SEM images of the asprepared Cs 3 Bi 2 I 9 and Cs 3 Bi 1.9 Ru 0.1 I 9 microcrystals, respectively. Particles of various size ranging from a few to hundreds of microns can be found throughout the samples. While small particles exhibit more irregular crystal shapes, those larger ones are typically hexagonal. Simultaneously, energy-dispersive X-ray spectroscopy (EDX) measurements were taken to verify the existence of Ru 3+ in the doped samples, all showing prominent Ru peaks (Fig. S3 †). Fig. 2c presents the TEM image of a small particle of Cs 3 Bi 1.9 Ru 0.1 I 9 , with the corresponding elemental mapping clearly showing that atoms of Bi and Ru are homogeneously distributed within the crystal.
XPS analysis of surface atomic chemical states reveals all Cs, Bi, I elements for Cs 3 Bi 2 I 9 and the additional Ru element for Cs 3 Bi 2Àx Ru x I 9 ( Fig. 3 and S4 †). No obvious shi of chemical states is observable for Cs, Bi, and I elements upon Ru-doping. There is a carbon peak at 284.9 eV in all spectra of tested samples, attributable to atmospheric carbon introduced during the sample preparation and testing process. Neighboring to the C peak, the Ru 3d peaks are composed of two doublets, assignable to Ru 3d 3/2 centered at around 286.0 eV and Ru 3d 5/2 at around 281.0 eV. The XPS analysis not only proves the successful doping of Ru into Cs 3 Bi 2 I 9 , but also conrms its trivalent chemical state. In addition, the intensity of the Ru 3d 5/2 doublet increases monotonically with the doping level. To further quantify the actual doping concentration of Ru 3+ in replacement of Bi 3+ in different Cs 3 Bi 2Àx Ru x I 9 samples, elemental analysis by Inductively Coupled Plasma-Atomic Emission Spectrometry (ICP-AES) was carried out on the obtained crystals and the results are listed in the Table S1. † For a theoretical doping concentration of 0, 1%, 2% and 5% based on the molar ratio of added RuCl 3 to BiCl 3 , the corresponding ICP-AES reading is 0, 0.9%, 1.9% and 4.3%, respectively. The discrepancy between the theoretical and actual Ru-doping levels could be caused by both incomplete reaction and weighting errors, but overall a good correlation between the two can be seen as long as the doping level isn't too high (>5%). For the purpose of convenience and clarication, we use the theoretical doping level to label all samples throughout this study.

Electronic and optical properties
Raman spectra were taken to illustrate the evolution of chemical bonding and lattice conguration inside the Cs 3 Bi 2Àx Ru x I 9 crystals with increasing Ru 3+ doping (Fig. 4). Within the wavenumber range from 50 cm À1 to 2000 cm À1 , the obtained spectrum for each compound at room temperature reveals similar crystal structure. No peaks were found above 200 cm À1 , and all compounds have six major peaks at the same positions without apparent peak shi. The position of these six peaks is in good agreement with what reported for Cs 3 Bi 2 I 9 before, 26,32 where the three peaks at higher wave number (high energy peaks) are   resulted from symmetric and asymmetric stretches of the terminal Bi-I bonds in the [Bi 2 I 9 ] 3À anion, and the middle two peaks at 105.8 cm À1 and 91.6 cm À1 are caused by symmetric and asymmetric stretches of the bridging Bi-I bonds, respectively. 33 On the far le side, the low energy peak at 59.4 cm À1 represents the bending modes of the Bi-I bonds. A prominent trend observed from the Raman spectra is the gradual disappearance of the antisymmetric peak at 121.1 cm À1 with the increasing Ru 3+ doping. Since both of the atomic mass and radius of Ru is smaller than Bi, one of the asymmetric stretches associated with it is likely to induce less inelastic scattering of the laser light, resulting in the weakened antisymmetric mode. As the peak at 121.1 cm À1 represents one of the two terminal Bi-I asymmetric stretches, this indicates the replacement of Bi with Ru has higher impact on the terminal Bi-I stretches than on the bridging B-I stretches, owing to the stronger force constants of the terminal Bi-I bonds as the bridging I atoms share their bonds with two Bi atoms. 32 The doping of Ru 3+ in replacement of Bi 3+ further causes local lattice distortions, as indicated by the Raman spectral evolution, and creates new defect states as well as enhanced electron-phonon coupling, both enabling to trap excitons upon irradiated absorption.
As shown in the optical images in Fig. 1b, the color of the compounds turns darker as the doping amount of Ru 3+ increases. To further quantify the light absorption and optical bandgaps of the Cs 3 Bi 2Àx Ru x I 9 compounds with various Ru 3+ doping, UV-Vis diffuse reectance spectra were taken as shown in Fig. 5a. All acquired UV-Vis spectra display similar curves with a stronger absorption observed with the increasing Ru 3+ amount. For all samples, the absorption starts from around 630 nm with an additional peak observed near 490 nm, corresponding to the n ¼ 1 excitonic transition as seen previously. 34 From the UV spectra, the optical bandgap can be estimated according to the equation below: where hn is the photon energy, A is a proportional constant, E g is the value of the band gap, n ¼ 2 for a direct transition or 1/2 for an indirect transition, and F(R) is the Kubelka-Munk function. By linearly extrapolating [F(R)hn] n ¼ 0, indirect optical bandgaps from 1.98 eV to 1.81 eV (Fig. 5b) and direct bandgaps from 2.07 eV to 1.98 eV (Fig. 5c) can be deduced for Cs 3 Bi 2Àx Ru x I 9 with increasing Ru 3+ doping. The bandgap value of pristine Cs 3 Bi 2 I 9 is in good agreement with previous reports, [34][35][36] and the shi of bandgaps is consistent with the optical color change observed for these compounds.
To better understand the radiative exciton recombination of the Ru-doped Cs 3 Bi 2 I 9 compounds, photoluminescence (PL) measurements were performed (Fig. 5d). All powder samples exhibit relatively broad PL emissions in the wavelength range of 500-900 nm, with the unsubstituted Cs 3 Bi 2 I 9 showing a similar spectrum to previous literature reports. 12,27 The broad PL emission is well-known characteristic of the defect halide perovskites such as A 3 M III 2 ,X 9 due to the strong electronphonon coupling and defect-mediated recombination, which is evidenced by the observed high emission wavelength ($700 nm) in contrast to the absorption edge ($600 nm). In addition, the electron-phonon interaction causes local distortion of the lattice, generating so-called polarons to couple with carriers and increase their effective mass, and therefore severely broaden the optical spectra. 26 Both of the lattice defects and polarons enable to trap excitons, intercepting the radiative recombination. With the Ru 3+ doping, all emission peaks of Cs 3 Bi 2Àx Ru x I 9 display a higher intensity and a slight red-shi towards higher wavelength, which further indicates the emission centers belong to some shallow defect-type states, rather than the intrinsic bandedge states. 12,37 The red-shi of PL emissions (from 696 to 706 nm) with increasing Ru 3+ doping resonates with the narrowed bandgaps as observed in the UV-Vis studies, and might also indicate the defect states become relatively deeper with the increased doping level. This argument is further supported by the fact that the intensity of PL emissions does not linearly increase with the doping level of Ru 3+ , with the highest intensity observed for Cs 3 Bi 1.98 Ru 0.02 I 9 , and then decreased thereaer. Similar phenomena of PL decreasing with increasing dopant has also been seen for In-alloyed Cs 2 AgBiBr 6 (ref. 38) and Mndoped Cs 2 AgInCl 6 (ref. 39) samples.
To examine the effect of Ru-doping on modifying the band structure of Cs 3 Bi 2Àx Ru x I 9 , we employed Ultraviolet Photoelectron Spectroscopy (UPS) to measure the position of the valence band maximum (VBM) with respect to the vacuum energy for Cs 3 Bi 2 I 9 and Cs 3 Bi 1.9 Ru 0.1 I 9 . Previous protocol developed by Lehner et al. was followed to capture and analyze the data and Au was used to calibrate the Fermi level. 19 As shown in Fig. 6, the onset of the UPS spectrum starts from 1.70 eV for Cs 3 Bi 2 I 9 and 1.10 eV for Cs 3 Bi 1.9 Ru 0.1 I 9 , with the corresponding cut-off binding energy being 17.02 and 16.82 eV, respectively. Therefore, by calculation the work function of Cs 3 Bi 2 I 9 and Cs 3 Bi 1.9 -Ru 0.1 I 9 are 4.20 and 4.40 eV, and their corresponding VBMs are À5.90 and À5.50 eV, respectively. Further based on the optical bandgaps measured from the previously acquired UV-Vis spectra (1.98 eV for Cs 3 Bi 2 I 9 and 1.80 eV for Cs 3 Bi 1.9 Ru 0.1 I 9 ), we were able to deduce the conduction band minimum (CBM) for Cs 3 Bi 2 I 9 and Cs 3 Bi 1.9 Ru 0.1 I 9 , being À3.92 and À3.70 eV respectively. The band positions of Cs 3 Bi 2 I 9 are within the upper and lower bounds of previously reported range (À3.40 to À4.20 eV for CBM, À5.60 to À6.30 eV for VBM). 19 What's more, by comparing Fig. 6a with Fig. 6b, the counts of photoelectrons within the valence band are higher on Cs 3 Bi 1.9 Ru 0.1 I 9 , indicative of a higher photoactivity, which is consistent with previous UV and PL results. Fig. 6c shows the column diagram for comparing the band positions of Cs 3 Bi 2 I 9 and Cs 3 Bi 1.9 Ru 0.1 I 9 , with the classic perovskite photoabsorber (MAPbI 3 ) and two commonly used hole-transportation (PEDOT:PSS) and electrontransportation (TiO 2 ) materials as references. Upon doping with $5% Ru, the valance band position of Cs 3 Bi 1.9 Ru 0.1 I 9 shis up about 0.40 eV (vs. Cs 3 Bi 2 I 9 ) with a narrowed bandgap of 1.80 eV, and the CBM is close to that of MAPbI 3 . As a result, both the narrowed bandgap and the upshi of the overall band structure might facilitate their wider optoelectronic applications.

Environmental and thermal stability
Stability is a key factor when evaluating perovskite materials for potential optoelectronic applications. It was reported that Cs 3 Bi 2 I 9 is more stable than Pb-based perovskites upon exposure to the ambient atmosphere in both light and moisture conditions. To further investigate the stability of Ru-doped Cs 3 Bi 2 I 9 , we selected the Cs 3 Bi 1.9 Ru 0.1 I 9 as an example to examine its environmental stability. XRD spectra were acquired aer 60 and 120 days of full exposure to the room light and moisture ($60% RH) of lab environments. As shown in Fig. 7a, no obvious change is observable among these spectra, demonstrating excellent environmental stability of the Ru-doped Cs 3 Bi 2 I 9 compounds. Furthermore, we checked the thermal stability of the Cs 3 Bi 1.9 Ru 0.1 I 9 compound under extreme conditions using the thermogravimetric analysis (TGA). A twostep weight loss was seen for both doped and un-doped Cs 3 Bi 2 I 9 , with the rst weight loss around 30%, likely due to the partial sublimation of one molecule of BiI 3 (bp z 500 C) from the parent compound. 26 For the pristine Cs 3 Bi 2 I 9 , the rst decomposition starts from 450 C, whereas for Cs 3 Bi 1.9 Ru 0.1 I 9 this temperature is lowered to 325 C. A linear decrease of the rst decomposition temperature was seen with the increasing Ru-doping (Fig. S5 †). This provides extra evidence for the substitution of Ru 3+ into the [Bi 2 I 9 ] 3À bioctahedra, deteriorating its thermal stability. For all compounds, the second decomposition starts from about 650 C, coincident with the melting point of CsI (mp z 650 C) and evaporation of the rest BiI 3 . Based on the TG curves, the decomposition products of Cs 3 -Bi 1.9 Ru 0.1 I 9 from the two weight-loss stages were further examined by XRD (Fig. S6 †), showing the remained product aer heating at 600 C contains both CsI and BiI 3 , whereas that aer 800 C contains CsI and elemental Cs. It is worth to note that although the thermal stability of Cs 3 Bi 2Àx Ru x I 9 deteriorates with increasing Ru-doping, they can still tolerate high temperature up to 300 C, sufficient for most real-world applications.

Conclusions
In summary, although Cs 3 Bi 2 I 9 has been viewed as a potential candidate for the photoabsorber of lead-free perovskite solar cells, its relatively high bandgap and deep intrinsic defect states are still the limiting factors for achieving high photoelectric conversion efficiency. In the current study, we have successfully doped Ru 3+ into Cs 3 Bi 2 I 9 up to 4.3% using a hydrothermal method, and obtained homogeneously dispersed crystals without damaging its crystalline structure. Optical measurements using Raman, UV-Vis and photoluminescent spectroscopy illustrated that even with a small doping amount, the optoelectronic properties of Cs 3 Bi 2Àx Ru x I 9 changes signicantly, including narrowed optical bandgap, induced shallow defect states, and more radiative recombination centers. UPS measurements revealed that doping of Ru 3+ upshis the overall band structure of Cs 3 Bi 2 I 9 with a higher work function. Stability tests further conrmed the good environmental and thermal stability of Cs 3 Bi 2Àx Ru x I 9 . All these results indicate that doping of Ru 3+ into Cs 3 Bi 2 I 9 allows engineering the band structure by regulating its defect states as well as electronic structure. Consequently, this study shed new light on regulating the optoelectronic properties of the defect halide perovskite family through cation substitution.