Unravelling the Photocatalytic Behavior of All-Inorganic Mixed Halide Perovskites: The Role of Surface Chemical States

Within the most mesmerizing materials in the world of optoelectronics, mixed halide perovskites (MHP) have been distinguished due to the tunability of their optoelectronic properties, balancing both the light harvesting efficiency and charge extraction into highly efficient solar devices. This feature has drawn the attention of analogous hot-topics as photocatalysis for carrying out more efficiently the degradation of organic compounds. However, the photo-oxidation ability of perovskite does not only depend on its excellent light-harvesting properties, but also on the surface chemical environment provided during its synthesis. Accordingly, we studied the role of surface chemical states of MHP based nanocrystals (NCs) synthesized by hot-injection (H-I) and anion-exchange (A-E) approaches, on their photocatalytic (PC) activity for the oxidation of β-naphthol as a model system. We concluded that iodide vacancies are the main surface chemical states that facilitate the formation of superoxide ions, O2 ●─ , responsible for the PC activity in A-E-MHP. Conversely, the PC performance of H-I-MHP is related to the appropriate balance between band gap and a highly oxidizing valence band. This work offers new insights on the surface properties of MHP related to their catalytic activity in photochemical applications.


Introduction
All-inorganic halide perovskites have become extremely competitive optoelectronic materials due to their high sunlight harvesting efficiency and notable charge carrier generation/transfer capabilities. 1,2 These features have been recognized as the key factors for enhancing photoconversion efficiencies in solar photovoltaic devices. 3 Furthermore, their fascinating photoluminescence quantum yield (PLQY), induced by a low nonradiative carrier recombination effect, can explain the expansion of their applicability in the wide optoelectronic field. 4 A high and prolonged PLQY from halide perovskites is the indication of the defect-tolerant structure, which increases the lifetime of photoexcited charge carriers. 5 This feature has been successfully exploited in few works related with photocatalysis, for increasing the oxidizing power during the photodegradation of organic molecules. [6][7][8] Pure perovskite nanocrystals (NCs) such as CsPbBr 3 and CsPbI 3 exhibit fixed band gaps. 9 By tuning the chemical composition of the NCs, harnessing band engineering approaches, both the light harvesting and relative positions of valence and conduction bands (VB and CB, respectively) are controlled. 10,11 Therefore, the photo-oxidation and/or photo-reduction abilities of these materials can be maximized. The band engineering mediated by chemical composition has been useful to improve the photocatalytic (PC) properties of earth-abundant insulators such as BaSO 4 by incorporating Ba-vacancies. 12 Band gap of modified BaSO 4 was reduced to improve its light harvesting efficiency, achieving a highly oxidizing VB, and performing more efficiently the PC removal of NO molecules. In this context, CsPbBr 3-x I x mixed halide perovskites (MHP) NCs constitute highly promising materials due to the improved light-to-carrier conversion efficiency with tunable band gap, which can be optimized by adjusting the halide content in their structure. 9 This outstanding photophysical feature has been extremely useful for the fabrication of efficient optoelectronic devices as multijunction tandem solar cells 13 and multicolor light-emitting diodes. 14,15 Nevertheless, very few works have reported the use of CsPbX 3 NCs (X = Br, I or Br/I combinations) for solar synthesis (e.g. polymerization of 3,4-ethylenedioxythiophene), 16 CO 2 reduction to C 1 products as methane and carbon monoxide, 6, 17 solar fuel generation (e.g. hydrogen production), 18,19 and advanced oxidation processes. 8,11 For the synthesis of MHP NCs, some methods have been described in the literature: (1) hot-injection (H-I), (2) anion-exchange (A-E), and supersaturated recrystallization (S-R) approaches. 4,20,21 H-I is performed by pre-mixing certain amounts of lead bromide (PbBr 2 ) and iodide (PbI 2 ) salts. This fact ensures the complete reaction of both the precursors PbBr 2 /PbI 2 , predicting the right MHP stoichiometry. 9 In A-E, pure CsPbBr 3 and CsPbI 3 colloidal NCs are mixed together with different Br:I ratios, taking advantage of the high ion mobility of halides into the perovskite lattice. 20,22 In S-R, MHP NCs are obtained in the absence of inert gas and injection process. 21 However, for A-E method halide diffusion generates inhomogeneous composition across the perovskite nanostructure, producing domains with high iodide content at the surface with a small bromide content within the core. 5,23 Moreover, it has been observed that iodide vacancies are formed at the MHP surface, defined by the presence of metallic lead (Pb 0 ) detected by as X-ray photoelectron spectroscopy (XPS). The defect density tends to linearly increase with the amount of CsPbI 3 in the A-E synthesis, 5 facilitating the O 2 incorporation, and consequently inducing the perovskite degradation. 24 For all the materials, the typical cubic morphology was evidenced, achieving a particle size between 11.1 and 13.6 nm. The size distribution of each MHP is shown as Figure S1A-G.
The UV-Vis absorption onset of the MHP NCs ( Figure 1H) was red-shifted after iodine incorporation in H-I-CsPbBr 3-x I x and A-E-CsPbBr 3-x I x , due to the lower band gap (E g ). Consistently, the photoluminescence (PL) emission only exhibited a single symmetric peak, which was also red shifted ( Figure S2B).The narrow full width at half maximum (FWHM) ( Figure S2C), highlights that the preparation was effective to achieve homogeneous solid solutions with larger particle size for the iodine enriched MHP. This fact is in good agreement with TEM images. 4,9 From emission peak positions, the estimation of the E g values (Table S2), confirmed that H-I and A-E are suitable methods to modulate the band gap of NCs, thus covering a broad range of the visible spectra (from green to red). The influence of the halide content on the crystalline structure of the H-I-MHP and A-E-MHP was also assessed by XRD ( Figure S2A), where two main diffraction peaks are present for all studied materials. These signals were ascribed to the (100) and (200) planes from the perovskite lattice, crystallizing in a pure single cubic phase (JCPDS card # 00-054-0752). 33 Furthermore, the two representative peak positions at the XRD profiles were shifted to lower Bragg angles after increasing the amount of iodine into the perovskite NCs.
This fact was ascribed to the enlargement of the Pb-X bond by increasing the ionic radius of the halide anion during bromine-to-iodine substitution. 22,34 Through EDS analysis, the chemical compositions of H-I-MHP and A-E-MHP were estimated (Table S1). It was observed that the stoichiometry of the CsPbX 3type structure in the NCs remained unaltered after variation of the halide content. It has been proposed that this exchange process occurs by simultaneous halide migration between the perovskites (from bromide-to iodide-perovskite and vice-versa), 22  to the emission feature of pure CsPbBr 3 NCs ( Figures S4B,B'). These results strongly suggest that A-E-MHP exhibits more iodide vacancies exposed to the O 2 molecules compared to H-I-MHP, which is the main reason for self-degradation. 25,39,40 However, the fact that the A-E-CsPbBr 2.22 I 0.78 was still luminescent is an indication that the degradation of all-inorganic perovskite nanocrystals is slow at room conditions. In this context, if the iodide vacancies are found at the NC surface as mentioned above, these species would be the main surface chemical states dictating the reactivity of MHP. obtained. These signals represent the Cs 3d 5/2 and Cs 3d 3/2 core levels, respectively, associated to the presence of Cs + into the CsPbX 3 -type structure. 41 In addition, it was observed that the Cs 3d doublet changes its binding energy (BE) depending on the type of perovskite. For H-I-MHP, these peaks were shifted to lower BEs by increasing the iodide content, which was ascribed to the substitution of Brwith Iinto the Cs-(PbX 6 ) octahedra. 42 Conversely, the opposite trend was evidenced for A-E-MHP, where the BE of Cs 3d doublet was displaced to higher values by reducing the Br:I ratio. Here, the added Icontent into Cs-(PbX 6 ) octahedra was decreased after anion exchange. This could be possible if a high amount of iodide vacancies is formed at the NC surface to expose the bromide domain-based core. In order to corroborate this hypothesis, I 3d doublet analysis was carried out.
Then, an increase of the iodide content in A-E-MHP to Br:I ratio 1:3 led to a shift to higher BE values.
This trend nicely agreed with the chemical composition of each material (See Table 1 (Table 1). 23 The lower complexation affinity between Pb 2+ and Icompared to Pb 2+ and Brinteraction, 9, 37, 38 allows a faster iodide migration compared to bromide diffusion between CsPbI 3 and CsPbBr 3 NCs. Hence, by increasing the iodide content in the A-E-MHP, more Ispecies are available to diffuse into NCs, and in turn highly iodide deficient MHP are generated. Therefore, at this stage, we can claim that iodide vacancies constitute the main surface chemical states produced in the MHP.

Surface photovoltage of hot-injection and anion-exchange mixed halide perovskite nanocrystals
After establishing the iodide vacancies as the main surface chemical states on H-I and A-E-MHP NCs, it is also clear that PC performances for the oxidation of organic molecules can be significantly affected by these surface states. Therefore, it is essential to clarify their acceptor/donor nature at the MHP lattice.
The synthesized NCs were deposited as thin films onto a TiO 2 compact layer (C-TiO 2 )/FTO, in order to observe the surface electrical potential generated by band-to-band (VB maximum→CB minimum) and trap-to-band (intraband gap levels→CB minimum) transitions at the NCs surface under illumination. SPV analysis is commonly based on the band bending model of semiconductors as a function of the carrier accumulation at the solid surface. 48 Upon illumination, acceptor states (n-type semiconductors) or donor states (p-type semiconductors) migrate to the semiconductor surface (depletion layer), where they are balanced with countercharges to produce the space-charge region (SCR). This region has a thickness around 1-10 3 nm. 49  The contact potential difference (CPD) of these NCs was measured as a function of the illumination intensity. CPD is considered as the potential drop generated between two surfaces in close proximity (in our case, C-TiO 2 /FTO and perovskites). 48 By subtracting CPD light -CPD dark , SPV for each material can be obtained. According to Figure S6, the higher the light intensity, the lower the measured CPD. This tendency was ascribed to electron extraction from MHP to C-TiO 2 /FTO, and hole accumulation at the NC surface. 11, 50, 52 Therefore, n-type semiconducting behavior of the MHP could be clearly identified.
Nonetheless, the CPD is less negative upon decreasing the Br:I ratio in the NCs. In order to understand this behavior, we studied the SPV vs. light intensity plots obtained for the abovementioned NCs ( Figure   4A). Under low light intensity conditions (3 W m -2 ), we evidenced a rapid displacement of SPV to more negative values. Then, SPV was almost constant at higher light intensities. This fact indicates that the maximum capability of charge carrier separation into perovskite NCs can be reached at the lowest light intensity used in these measurements. This ability keeps stable up to the maximum irradiance conditions.
Additionally, the negative SPV obtained for the three different NCs involves the presence of surface acceptor states, associated to accumulated surface holes after electron extraction. 50,51 Comparing the corresponding magnitude of SPV for each nanocrystal, it is also clear that a more positive value was reached upon increasing the iodide content into the CsPbBr 3-x I x lattice. Considering that H-I-CsPbBr 2.37 I 0.63 and H-I-CsPbBr 1.56 I 1.44 were obtained by the same synthesis method, the slight difference in SPV was ascribed to the increase of particle size of MHP. Thus, the carrier diffusion length was increased and the carrier separation slowed down. 49 Conversely, the markedly increase of SPV magnitude for A-E-CsPbBr 0.42 I 2.58 strongly suggests that the surface chemical states can also influence the surface charge redistribution by trap-to-band transitions. 49 These findings together with the XPS analysis, according to which A-E-MHP produce a higher density of iodide vacancies (and consequently also metallic lead) compared to H-I-MHP, demonstrate that electron extraction from A-E-CsPbBr 0.42 I 2.58 is hindered by these surface states. This fact favors electron trapping. As it has been reported that iodide vacancies are produced by the generation of Pb dangling bonds, which are located close to the CB of the perovskite ( Figure 4B), 53,54 we attributed the trap-to-band transitions to electron photoexcitation from VB to the I vacancy levels and then to the CB under illumination. Therefore, we conclude that iodide vacancies act as surface donor states for electron trapping at the NCs.

Band structure of the hot-injection and anion-exchange mixed halide perovskites
The band gap of NCs plays a key role on the PC activity as a narrower band gap produce higher photon harvesting but the energetic band position also influences the PC activity as more positive energy position of the VB (vs. Fermi level), increases the oxidizing power of the MHP photocatalysts.
Consequently, the final PC activity is strongly determined by both band gap and energy position.
Therefore, to understand the correlation between electronic structure and PC activity for H-I-MHP and A-E-MHP NCs, their corresponding VB and CB relative positions were determined. By obtaining normalized XPS VB spectra ( Figures S7A,B), the valence band energy (E VB ) was determined for all the perovskites through the extrapolation method. Then, the equation E VB = E CB + E g was used to calculate the associated conduction band energy (E CB ) and thereby estimate the band structure. The obtained energy levels are in good agreement with band structures extracted from electrochemical measurements for these perovskites NCs. 11 As seen in Figure 5A, can provide a high PC activity for performing the β-naphthol (β-NPT) photodegradation but the different band position will influence which NC composition will produce the better PC activity depending on the synthesis process. In addition, the different density of surface chemical states of these materials suggests that diverse organic oxidation pathways could be followed, as we analyzed in the next section.

Figure 5. Band structure of (A) H-I-CsPbBr 3-x I x and (B) A-E-CsPbBr 3-x I x nanocrystals, obtained from
XPS VB spectra through the extrapolation method.  Figure S4B,B'). 40 However, the remained green luminescence of A-E-CsPbBr 2.22 I 0.78 also suggest that the incorporated O 2 is not enough to promote the complete degradation of the MHP.

Photocatalytic activity of mixed halide perovskite nanocrystals for β-naphthol degradation
Conversely, a higher density of iodide vacancies evidenced at the A-E-CsPbBr 0. 42  Nonetheless, short times would be required to use these materials in PC oxidation reactions before to evidence their fast-self-degradation. Conversely, the PC activity of A-E-MHP with lower density of iodide vacancies (slow-self-degradation) could be studied at longer times, which could compensate their   as a secondary reaction product in the solution. This reaction mechanism has been reported as a concerted double proton-transfer electron transfer reaction (CDPET). 57 Figure S10, supporting video 1 and 2) (See supporting information for more details). However, although the complete β-NPT photodegradation was not observed in the ATR-IR spectra after using A-E-MHP, we infer that the O 2 •─ species can also directly degrade the organic molecules to obtain non-radical products (reactions 6 and 7). 29,30 Hence, the decrease of C/C o relative concentration of β-NPT can be reached: Previous reports have reported that phenoxy-based radicals from the reaction of phenolic compounds and O 2 •─ are prone to form dimers, oligomers, or/and quinones as the main non-radical products in organic media. 29,30 Another useful strategy to analyze the effect of O 2 on the PC performance of MHP was by conducting the photodegradation measurements under inert atmosphere. Figures S11A,B  relative positions to produce a high density of highly-oxidizing photoexcited holes.

Conclusions
In