A Supported Bismuth Halide Perovskite Photocatalyst for Selective Aliphatic and Aromatic C–H Bond Activation

Abstract Direct selective oxidation of hydrocarbons to oxygenates by O2 is challenging. Catalysts are limited by the low activity and narrow application scope, and the main focus is on active C−H bonds at benzylic positions. In this work, stable, lead‐free, Cs3Bi2Br9 halide perovskites are integrated within the pore channels of mesoporous SBA‐15 silica and demonstrate their photocatalytic potentials for C−H bond activation. The composite photocatalysts can effectively oxidize hydrocarbons (C5 to C16 including aromatic and aliphatic alkanes) with a conversion rate up to 32900 μmol gcat −1 h−1 and excellent selectivity (>99 %) towards aldehydes and ketones under visible‐light irradiation. Isotopic labeling, in situ spectroscopic studies, and DFT calculations reveal that well‐dispersed small perovskite nanoparticles (2–5 nm) possess enhanced electron–hole separation and a close contact with hydrocarbons that facilitates C(sp3)−H bond activation by photoinduced charges.


Introduction
Thes elective oxidation of sp 3 CÀHb onds using O 2 to valuable oxygenated products is one of the most important reactions in the chemical industry for fine chemicals and pharmaceuticals production. [1] In comparison with the partial oxidation of active and expensive alcohol substrates, [2] the direct conversion of hydrocarbons (for example,a liphatic alkanes) in air to commercial oxygenates remains challenging. This is due to the high bond dissociation energy of C(sp 3 )ÀH bonds (70-130 kcal mol À1 )a nd their unfavorable adsorption. [1b, 3] Theactivation of C(sp 3 ) À Hbonds generally demands specific expensive metal complexes or aggressive oxidants with harsh reaction conditions (high pressure and temperature). [4] Additionally,insome cases the conversion has to be suppressed (< 15 %) to avoid poor selectivity owing to overoxidation products like CO and CO 2 . [5] Photocatalytic organic synthesis is apromising and green approach that is performed under mild conditions with solar energy as driving force. [6] To date,s everal photocatalytic systems have been reported for aromatic oxidations. [7] For example,CdS, [7c] Bi 2 WO 6 /BiOCl, [7f] and TiO 2 /Bi 2 MO 6 [7g] have activated benzylic CÀHbonds under visible light illumination. Very recently,l ead-based perovskite NiO x /FAPbBr 3 /TiO 2 composite was reported to activate CÀHb ond under solar light irradiation;h owever,t he production rate dropped tenfold when only visible light was used as light source. [7i] Thep hotocatalysts developed so far are limited by low efficiencies (the highest reported conversion rate is 4388 mmol g cat À1 h À1 for toluene oxidation) [7f] and narrow substrate scopes,m ainly focusing on CÀHb onds at benzylic or allylic positions.T hus,t here is an eed to design more effective photocatalysts for selective C À Hbond activation.
Herein, we report an ovel class of ordered mesoporous SBA-15 silica supported halide perovskite photocatalysts. Halide perovskite nanoparticles with as ize of 2-5 nm are confined within pore channels of SBA-15, which provides large numbers of catalytically active centers.C onsequently, SBA-15 supported Cs 3 Bi 2 Br 9 nanoparticles show outstanding catalytic performance for selective aliphatic and aromatic C(sp 3 )ÀHb ond activation under visible light irradiation. Electronic structure calculations using density functional theory (DFT) indicate that clusters like Cs 12 Bi 14 Br 54 with as ize of af ew nanometers are stable and have bismuth-rich surfaces.T he optimized photocatalyst, 10 wt %C s 3 Bi 2 Br 9 / SBA-15, displays ah igh conversion rate of up to ca. 32 900 mmol g cat À1 h À1 ,g oing beyond state-of-the-art catalysts. Aq uantum efficiency (QE) of 11 %i sa chieved under 1sun illumination. By using toluene oxidation as amodel reaction, our experimental, spectroscopic,a nd computational studies reveal that supported perovskite nanoparticles show enhanced charge separation and close interactions with hydrocarbons that promote C À Hb ond activation.

Catalyst Characterization
Theh alide perovskite Cs 3 Bi 2 Br 9 nanoparticles were formed within the mesopores of SBA-15 silica with different loadings (calculated 5, 10, 20, 40 wt %; Supporting Information, Figure S1 d, and Table S1 for elemental analysis) by the incipient wetness impregnation (IWI) method. [22] Fort he Cs 3 Bi 2 Br 9 /SBA-15 sample with 5wt% loading of perovskite nanoparticles,s canning transmission electron microscopy (STEM) in secondary electron (SE) mode ( Figure 1a)shows at ypical morphology of SBA-15. [23] As shown in Figure 1b, high-angle annular dark-field mode (HAADF) STEM clearly demonstrates the successful confinement of small Cs 3 Bi 2 Br 9 nanoparticles with as ize of 2-5 nm in the ordered channels (pore size ca. 8nm) of SBA-15 silica, and nanoparticles are well-separated. When the loading is increased to 10 wt %, the high dispersion of small perovskite nanoparticles is still maintained as illustrated by Figure 1c,d. Ah igher loading of 20 wt %m akes the pore filling by Cs 3 Bi 2 Br 9 much more compact, leading to poor utilization of available space in the support matrix and active sites on the perovskite surface. Some parts of the channels of SBA-15 were completely filled, resulting in more aggregated perovskite nanoparticles,asseen in Figure 1e.M eanwhile,t he element mapping results  demonstrate that the material confined in SBA-15 channels is halide perovskite,w hile its crystallinity was confirmed by powder X-ray diffraction (XRD) (Supporting Information, Figure S1 a,b). At the highest loading of 40 wt %, formation of bulk halide perovskites on the structure of SBA-15 silica is observed along with the pore confined nanoparticles (Figure 1k,l). With the IWI method, av ery homogenous distribution of perovskite crystals over mesoporous silica cannot be achieved, thus pores of SBA-15 stay unfilled in some areas (Supporting Information, Figure S1 c).
In agreement with the electron microscopy analysis,t he XRD results (Supporting Information, Figure S1 a,b) show that the samples with lower loadings (5 wt %a nd 10 wt %) presented weak and broad characteristic diffraction peaks of Cs 3 Bi 2 Br 9 ,i ndicating small perovskite nanoparticles.S BA-15 silica kept its ordered structure even at ah igh perovskite loading of 40 wt %. Additionally,b ased on the N 2 -physisorption analysis (Supporting Information, Figure S2), the samples with lower perovskite loadings (5 wt %a nd 10 wt %) have similar pore volumes (1.1 cm 3 g À1 )and surface areas (802 and 767 m 2 g À1 ,r espectively). Increasing the loading amount to 20 wt %a nd 40 wt %c auses ad rop in pore volumes (0.9 and 0.6 cm 3 g À1 )and surface areas (677 and 473 m 2 g À1 )for the composite materials owing to the gradual blockage of mesopores by aggregated perovskite particles,a ss een in STEM images in Figure 1.
Theo ptical properties of as-prepared Cs 3 Bi 2 Br 9 /SBA-15 were further characterized by UV/Vis diffuse reflectance (DRS) and photoluminescence (PL) spectroscopy.F igure 2a shows that all samples have visible light absorption extended to the wavelength of 500 nm. As expected, increase of loading enhances light absorption since Cs 3 Bi 2 Br 9 has the capability to absorb visible light as indicated by the DRS results of bulk Cs 3 Bi 2 Br 9 and bare SBA-15 ( Figure 2a). Theoptical band gap (E g )i ncreased gradually as the loading decreased due to the quantum confinement effect for smaller nanoparticles (Supporting Information, Figure S4 a). [24] TheP Ls pectra of supported samples ( Figure 2b)c learly show ar ed-shift of main emission peaks as the loading increased, matching well with the gradually decreased E g .Moreover,the samples with lower loadings (5 and 10 wt %) showed am uch lower PL intensity,i ndicating the suppressed radiative recombination of photoinduced electron-hole pairs. [12a] This may originate from the smaller perovskite nanoparticles with shorter diffusion distance of charges in excited state. [25] Thus, Cs 3 Bi 2 Br 9 /SBA-15 samples with lower loadings seems to have better charge separation. After adsorption of hydrocarbons (for example,t oluene), the supported samples with low perovskite loadings (5 and 10 wt %) display an additional absorption band in the visible range (from 500 to 750 nm; Supporting Information, Figure S4 b), which indicates strong interactions between aromatics and Cs 3 Bi 2 Br 9 /SBA-15 composite with Bi atoms as Lewis acid sites (Supporting Information, Figure S3). [20] These types of interactions might promote the catalytic efficiency of the photocatalysts.T he FTIR spectroscopy showed no strong intramolecular interactions between hydrocarbons and composite material (Supporting Information, Figure S4 g). [26]

Photocatalytic Oxidation of Aliphatic and Aromatic Hydrocarbons
After catalyst characterizations,weemployed as-prepared Cs 3 Bi 2 Br 9 /SBA-15 as photocatalysts to activate C(sp 3 )ÀH bonds of hydrocarbons in air under visible light irradiation. By using toluene as am odel substrate,t he catalytic performance of several photocatalysts were systematically evaluated in ah ome-made setup (Supporting Information,Figure S5). Thec ontrol experiments show that the absence of light or the halide perovskite phase (no catalyst or bare SBA-15 support) led to no production of oxygenates from toluene.
As shown in Figure 3a,f or the supported Cs 3 Bi 2 Br 9 /SBA-15 photocatalysts,t he sample with al oading of 10 wt %h as the highest conversion rate of 12 600 mmol g cat À1 h À1 and high selectivity of 90 %t owards benzaldehyde (benzyl alcohol as the main by-product). Meanwhile,noincrease of CO/CO 2 was detected in the gas phase using gas chromatography (GC) and mass spectrometry (MS;Supporting Information, Figure S6). In contrast, the 5wt% Cs 3 Bi 2 Br 9 /SBA-15 photocatalyst showed lower activity owing to al ower number of active centers and weaker light absorption (Figure 2a). In the cases of higher loadings (20 and 40 wt %) and bulk Cs 3 Bi 2 Br 9 ,t he dramatic drop of activities (down to 140 mmol g cat À1 h À1 )w as observed mainly due to limited accessible active sites and serious charge recombination with aggregated or large nanoparticles.  Thec atalytic survey of 40 wt %s upported and bulk perovskite samples suggests that the specific interactions with toluene indicated by UV/Vis DRS analysis are not essential to initiate C À Hb ond activation. Meanwhile,t he toluene conversion rate could be promoted when the strong interactions are present (Supporting Information, Figure S4 b). This correlation also applies for other halide perovskites (for example,C s 2 AgBiBr 6 and CsPbBr 3 ). Although supported Cs 2 AgBiBr 6 and CsPbBr 3 nanoparticles on SBA-15 can absorb more photons due to smaller band gaps (2.28 and 2.33 eV respectively), they showed lower photocatalytic activity (2100 and 200 mmol g cat À1 h À1 respectively), which corresponds well with their weaker interactions with toluene (Supporting Information, Figure S4 d). Furthermore,t he CsBr/SBA-15 sample presented no activity,w hile the BiBr 3 counterpart was active for the activation of C À Hb onds with ac onversion rate of 2100 mmol g cat À1 h À1 (Figure 3a), suggesting Bi and Br atoms dominantly contribute to the catalytic reaction. BiBr 3 /SBA-15 also displayed interactions with toluene (Supporting Information, Figure S4 d), but its poor activity could be related to its wider band gap (E g of 3.1 eV; Supporting Information, Figure S7). Other common Bi-based photocatalysts BiOBr and Bi 2 WO 6 showed negligible activities (< 100 mmol g cat À1 h À1 ), which agrees with the previous reports.
[7e,f] Among all the reported active catalysts (even including thermal systems with harsh reaction conditions), our supported Cs 3 Bi 2 Br 9 /SBA-15 photocatalyst demonstrates the highest reported catalytic activity (Supporting Information, Figure S8).
Halide perovskite-based materials are known for their thermal and moisture instabilities,w hich are the main drawbacks for their use in wide-range applications.A ss een in Figure 3b,t he high benzaldehyde production rate was maintained up to 4h of irradiation with visible light and the catalyst showed negligible structural changes after reaction (see XRD and UV/Vis DRS in the Supporting Information, Figure S9, and element analysis of leached species in the Supporting Information, Table S2). Furthermore,f rom the action spectrum with monochromatic light irradiation (Supporting Information, Figure S10), the amount of benzaldehyde varied in the same trend with the light adsorption profile,s uggesting light energy as the driving force for C À H bond activation. Notably,aQE of 11 %w as achieved under 1sun irradiation by using as olar simulator.H owever,t he production rate clearly slowed down after another 4hof light illumination and the reaction stopped when the irradiation time was extended to 12 h. After the reaction, the yellow photocatalyst had turned to white in color as BiOBr was formed (Supporting Information, Figure S11) because of water generation. Thed eactivation of the photocatalysts can be prevented by addition of anhydrous Na 2 SO 4 to absorb the produced water (Figure 3b).
Thea pplication scope of supported perovskite nanoparticle photocatalysts was extended to other hydrocarbons from C 5 to C 16 including aliphatic and aromatic alkanes.A s shown in Table 1a nd the Supporting Information, Tables S3,  S4, in all cases their corresponding oxygenates (aldehydes/ ketones and alcohols) were formed with excellent selectivities (> 99 %) without formation of any over-oxidation products like CO 2 .Even though very low yields of oxygenates (< 0.3 %; Supporting Information, Table S3) were obtained owing to the limited reaction time and very low catalyst concentration, the benzaldehyde yield can be enhanced to 3.8 %w ith modified reaction conditions like longer reaction time,higher catalyst concentration and higher light irradiation intensity (Supporting Information, Figure S12).
Thea romatic substrates presented much higher conversion rate (Table 1, entries 7-12, > 10 000 mmol g cat À1 h À1 ) than the aliphatic ones (Table 1, entries 1-6, < 2500 mmol g cat À1 h À1 ), which generally have more robust C À Hb onds with higher bond dissociation energies (BDEs; Supporting Information, Table S5). [27] Fort he toluene derivatives,the substrate with an electron-donating substituent (for example, À CH 3 )s howed ah igher reaction rate than the one with an electron-withdrawing substituent (for example, À NO 2 ), even though both have similar BDEs (Supporting Information, Table S5). However,t he oxidation activities positively correlate with interactions between aromatics and perovskite nanoparticles,which could be indicated by UV/Vis DRS analysis (Supporting Information, Figure S4 f). Notably, for the linear aliphatic alkanes (from C 5 to C 8 ), the reaction rate increased as the carbon chain length increased, which could be ascribed to the more stable alkyl radical intermediates.F or the longer alkanes (C 10 and C 16 ), the dramatically decreased activity might be originated from the diffusion limitations and steric hindrances owing to larger size and higher viscosity of substrates.B esides,l inear ketones appeared as the dominant products with the secondary C À H bonds cleaved, which generally have lower BDEs than the terminal ones. [28] Interaction of Catalysts with Hydrocarbons ADFT study was conducted to investigate the electronic structure of halide perovskite and its interactions with hydrocarbons (Supporting Information, Figures S13-S17 and Tables S6, S7). Figure 4s hows modeling of as table Cs 12 Bi 14 Br 54 cluster with BiBr 3 and BiBr 5 motifs on the surface (size ca. 1.5 nm;F igure 4b), and optimized geometries for toluene adsorption on the nanoparticle (Figure 4c,d). Specifically,this model, with ahigh surface to bulk ratio,ismainly designed to simulate the surface environment of Cs 3 Bi 2 Br 9 nanoparticles with aw ide variety of motifs. Furthermore,a ll adsorbed toluene molecules exhibit an interaction pattern with their benzene rings either on top of Bi or Cs atoms (average distance of 3.0 and 3.5 ,r espectively) (Figure 4c), resulting in Bi-arene (Cs-arene) interacting through dispersion interactions and charge transfer (from the p system to Bi s*orbitals). [29] Notably,hydrogen atoms of ÀCH 3 groups are located in the vicinity of surface Br atoms (distance around 2.9 ;F igure 4d). Our calculations also indicate that the presence of toluene molecules at the surface induces low-energy absorption peaks (from 500 nm to 650 nm) as shown in the Supporting Information, Figure S16. Thea nalysis of computational results (Supporting Information, Figures S13-S17 and Tables S6-S7) shows that lowenergy transitions mostly correspond to the charge transfer from lone electron pairs of aB ra tom with as mall contribution of the p system of toluene to an empty orbital of aB i atom. An additional work needs to be conducted to inves-tigate the interactions between halide perovskites and aliphatic hydrocarbons,which will be scope of another study.

Reaction Mechanism
To further investigate the photocatalytic process,aseries of control experiments have been conducted in the absence of O 2 or with addition of various scavengers,which demonstrated different effects on the conversion rate of toluene oxidation as shown in Figure 5a.A lmost no benzaldehyde and benzyl alcohol production happened under an Ar atmosphere,d irectly proving the necessity of O 2 for oxygenated compounds formation. However, in anaerobic conditions the coupling reaction occurred with formation of 1,2diphenylethane and change in photocatalyst color from yellow to black (Supporting Information, Figure S18).
Moreover,u nder Ar atmosphere the addition of 2,2,6,6tetermethylpiperidine-N-oxyl (TEMPO) as ar adical trapper captures the benzyl radical with the corresponding adduct formed (Supporting Information, Figure S19 a). Meanwhile, 1-hydroxy-2,2,6,6-tetramethyl-piperidine (TEMPO-H;S upporting Information, Figure S19 b) was also generated owing to the reaction with reducing species (H + /e À )a nd the photocatalyst color was maintained. [30] These results verify formation of carbon-centered radicals as intermediates after the cleavage step of C(sp 3 ) À Hbonds.The abstracted Hatoms may be transferred to the perovskite surface (see the Supporting Information, solid-state 1 Hn uclear magnetic resonance spectroscopy in Figure S20 and in situ MS in Figure S21), probably leading to the color change of the photocatalyst after irradiation in Ar. [2b] Besides,n or educed Bi species (such as Bi 2+ )w ere observed by the X-ray photoelectron spectroscopy (XPS) analysis on the excited photocatalyst (cleaned and dried under vacuum;S upporting Information, Figure S22 a,b). Thea ddition of ammonium oxalate (NH 4 ) 2 C 2 O 4 as hole scavenger and butylated hydroxytoluene (BHT) as carbon-centered radical scavenger both terminated the reaction, while CCl 4 as electron acceptor slightly affected the activity.Clearly,photoinduced holes (h + ) are significant to activate C(sp 3 )ÀHbonds in toluene.
An isotopic labeling experiment was conducted in order to gain more insights into the reaction mechanism. When deuterated toluene was used as reactant, an apparent kinetic isotope effect (KIE) was observed with a k H /k D value of 4.4 (Supporting Information, Table S8), suggesting C(sp 3 )ÀH bond activation as the rate-determining step.I nc ontrast, the further reaction of benzyl radicals with O 2 or superoxide anions CO 2 À (from reduction of O 2 by photoinduced electrons) took place very rapidly,coinciding with the previous report. [31] Besides,hydroxyl radicals had no contribution, as tert-butanol addition did not suppress activity.
Notably,t he addition of aniline as ab ase prohibited the reaction completely,w hile no influence appeared in the case  of benzoic acid. This could be explained by the poisoning of Bi atoms serving as weak Lewis acid sites. [20] Thep eroxy species (ROOC)c an be generated from the fast reaction of alkyl radicals with O 2 . [31] In our perovskite system, ROOC react with reducing species (H + /e À )o np erovskite surface to form ROOH, which can be further decomposed to carbonyl products (Figure 5c). [32] This dominant reaction path was supported by in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTs) experiments as shown in Figure 5b.Inthe presence of toluene and O 2 gas mixture,one peak at 1691 cm À1 corresponding to the carbonyl group of benzaldehyde appeared after the light irradiation of Cs 3 Bi 2 Br 9 /SBA-15 powders,while no peaks assigned to benzyl alcohol could be detected in the whole irradiation process. Moreover,the energy barrier for C À Hbond activation step on Cs 3 Bi 2 Br 9 nanoparticle can be regarded as the energy difference between reactants (toluene on excited surface À Br*[h + ]) and intermediates (PhCH 2 C with ÀBr À ÀH + surface). Clearly, this CÀHb ond activation induced by holes (on Br atoms), which is analogous to NO 3 C radical-driven oxidation as ac oncerted proton-electron transfer process, [33] is thermodynamically feasible as indicated by our calculations (see calculated potential energy of possible intermediates in the Supporting Information, Figure S23).
Based on the above experimental data and electronic structure calculations,wecan put forward ahypothesis for the reaction mechanism for hydrocarbon oxidation using supported halide perovskite nanoparticles as photocatalysts.A s shown in Figure 6a,u nder visible light irradiation of Cs 3 Bi 2 Br 9 nanoparticles,t he photoinduced electrons and holes are generated in the conduction band minimum (CBM, À0.57 eV vs.S HE) and valence band maximum (VBM, + 2.17 eV vs.S HE) positions,r espectively (Supporting Information, Figure S22 c,d). TheBiporbitals dominantly contribute to the CBM, while Br 4p orbitals correspond to VBM. [8e,34] Subsequently,w ith promoted electron-hole separation on Cs 3 Bi 2 Br 9 nanoparticles,t he electrons reduce the electron acceptors (for example,O 2 and H + )a nd the holes drive the challenging oxidation of C(sp 3 )ÀHb onds to form alkyl radical intermediates (RC). As illustrated in Figure 6b (with toluene as am odel), before light illumination, the preferential geometry of toluene adsorption results in the proximity between Hatoms of the À CH 3 group and Br atoms, which simultaneously serve as oxidation points on the perovskite.F inally,the close contact between reactive oxidation sites (Br atoms) and Ha toms from hydrocarbons promotes the activation of C(sp 3 ) À Hb onds.

Conclusion
As eries of photocatalysts consisting of lead-free Cs 3 Bi 2 Br 9 nanoparticles confined in am atrix of mesoporous SBA-15 silica have been prepared. Thewell-dispersed halide perovskite nanoparticles (2-5 nm) lead to better charge separation, more accessible active sites,a nd close contact with hydrocarbons,w hich facilitate the activation of CÀH bonds in hydrocarbons.U nder visible light irradiation, the supported Cs 3 Bi 2 Br 9 /SBA-15 photocatalysts efficiently oxidize C(sp 3 ) À Hbonds of various hydrocarbons (from C 5 to C 16 including aromatic and aliphatic alkanes) to their corresponding oxygenates (mainly aldehydes/ketones) with aconversion rate of up to 32 900 mmol g cat À1 h À1 and high selectivity of > 99 %. We believe this work could provide ap romising direction for exploring halide perovskite photocatalysis in more challenging organic transformation reactions. Figure 6. Energy diagram and proposed mechanism of supported Cs 3 Bi 2 Br 9 nanoparticle for C(sp 3 )ÀHbond activation. a) Band positions of Cs 3 Bi 2 Br 9 nanoparticle including the proposed redox reaction paths based on concerted proton-electron transfer process. b) Possible catalytic mechanism of Cs 3 Bi 2 Br 9 nanoparticle as photocatalyst for hydrocarbono xidation under visible light irradiation in air.The blue dotted circle indicates the production of an intermediate benzyl radical after cleaving CÀHbond.