**2. Results**

Pure CsPbBr3 was synthesized using an anti-solvent method and MIL-100(Fe) was synthesized by a modified non-hydrothermal method (see Section 3). The composite photocatalysts were obtained by an in situ growth method and were named after the amount of MIL-100(Fe) precursor (Fe(NO3)3·9H2O) added (see Section 3).

In the first step, X-ray diffraction (XRD) was used to investigate the crystallinity of the pure and composite materials (Figure 1a). The XRD patterns clearly show the orthorhombic CsPbBr3 structure and in the composite materials, additional diffraction peaks at 3.4◦ and 11◦ from MIL-100(Fe) appear (Figure 1b) [25,26]. With an increasing amount of MIL-100(Fe) loaded in the composites, we observe a gradual transformation of CsPbBr3 (PDF#18-0364) to CsPb2Br5 (PDF#25-0211). Figure 1c illustrates the perovskites' crystallographic structure transformation. This phase transformation could be ascribed to the excessive H2O in the Fe precursor, which partially converts the CsPbBr3. To evaluate the necessity of H2O during the in-situ growth of the MOF, the Fe precursor was dried ahead of the synthesis. With this dried Fe(NO3)3 precursor, the desired MIL-100(Fe) could not be generated (Figure 1b), indicating the critical role of water during the MIL-100(Fe) formation, in line with literature reports [10].

The Fourier-transform infrared (FTIR) spectra of all composites exhibit typical MIL-100(Fe) bands (Figure 2a) [27]. The bands at 1625 and 1380 cm<sup>−</sup><sup>1</sup> are related to the stretching vibrations of carboxyl groups [28], the band at 1446 cm<sup>−</sup><sup>1</sup> can be ascribed to the O–H stretching vibration, while the ones corresponding to the bending vibration of C–H (at 759 cm<sup>−</sup>1) and C=C (at 711 cm<sup>−</sup>1) originate from the benzene ring. As expected, the peak intensities of the Fe-O stretching vibration (at 491 cm<sup>−</sup>1) and free C=O stretching vibration from unreacted H3BTC (at 1716 cm<sup>−</sup>1) enhanced with increasing the amount of MIL-100(Fe) loaded in the composites [29]. Further, a broad peak at 3000–3500 cm<sup>−</sup><sup>1</sup> indicates the presence of a significant amount of adsorbed H2O in the composites.

Thermogravimetric analysis (TGA) was performed in O2 to quantify the MIL-100(Fe) amount in the composites based on the oxidation of the organic linker (Figure 2b). MIL-100(Fe) exhibits a clear two-step weight loss: (1) below 200 ◦C, the weight loss is associated with the removal of adsorbed H2O, and H2O coordinated to the iron trimers, (2) around 300 ◦C, H3BTC decomposes [30]. This two-step weight loss was identified in all the composites. CsPbBr3 shows a good thermal stability to about 500 ◦C. Hence, the weight loss due to H3BTC decomposition was used to determine the relative amount of MIL-100(Fe) in the composites. As listed in Table 1, the MIL-100(Fe) content in the composites varies from 9 wt% in p-30Fe to 53 wt% in p-180Fe.

**Figure 1.** (**<sup>a</sup>**,**b**) XRD patterns of the as-prepared photocatalysts, and (**c**) visualization of the perovskites' crystallographic structure transformation.

**Figure 2.** (**a**) FTIR spectra, and (**b**) TGA thermograms of the as-prepared photocatalysts.


**Table 1.** Weight ratio of MIL-100(Fe) in the composites.

The surface chemical composition and chemical states of the as-synthesized composites were further revealed by X-ray photoelectron spectroscopy (XPS). The XPS spectrum (Figure 3a) of p-90Fe shows distinct peaks from both of MIL-100(Fe) and CsPbBr3. Relevant high-resolution spectra of p-90Fe (Figure S1) display peaks of Cs 3d, Pb 4f, and Br 3d in the range of 720–745, 134–148, and 64–74 eV, respectively, which are well-matched with those of CsPbBr3 [31,32]. Two dominant peaks of Fe (Figure 3b) at 724.6 and 711.7 eV are attributed to Fe 2p1/2 and Fe 2p3/2, respectively. Additional satellite peak appearing at 717.2 eV corresponds to Fe3<sup>+</sup> in MIL-100(Fe) [33].

**Figure 3.** XPS spectra of p-90Fe: (**a**) survey and (**b**) Fe 2p.

The response of MIL-100(Fe) in the visible light region arises from the direct excitation of the Fe–O clusters. As shown in the absorption spectra in Figure 4a (as obtained by UV-vis diffuse reflectance, see Section 3), p-30Fe, p-60Fe, and p-90Fe basically maintain the pattern of CsPbBr3. p-120Fe and p-180Fe exhibit a pattern similar to MIL-100(Fe), due to the larger amount of MIL-100(Fe) included. The addition of MIL-100(Fe) enhances the composites' optical response in the visible light region, especially above 550 nm. The absorption edges around 550 nm gradually blue shift, corresponding to the increased bandgaps (Eg) (calculated by the Tauc plots, Figure 4b) of the composites consisting of more MIL-100(Fe) (Table 2). This blue-shift arises from the phase transformation from CsPbBr3 (Eg = 2.3 eV) to CsPb2Br5 (Eg = 3.0 eV) and the more dominant role of MIL-100(Fe) in the composite photocatalysts upon increasing the Fe content.

**Figure 4.** (**a**) UV-vis absorption spectra of parent compounds and composites, and (**b**) Tauc plots of MIL-100(Fe) and the composites. The band gap energies of the photocatalysts, listed in Table 2, were estimated following Kubelka-Munk transformation. Each bandgap is determined by the intersection point of the corresponding tangent line and horizontal axis.

**Table 2.** Bandgap data of the prepared samples.


Figure 5 and Figure S2 show the steady-state photoluminescence (PL) spectra of the photocatalysts at an excitation wavelength of 380 nm. As shown in Figure S2, CsPbBr3 has significantly higher peak intensity than the composites. The broad emission peak from 400 to 450 nm can be attributed to the 1,3,5-benzene tricarboxylic acid linkers in the MIL-100(Fe) structure [34]. The PL emission spectra of p-30Fe, p-60Fe and p-90Fe show a gradual blue shift and weakening intensity of the peak at around 530 nm. Different from 3D CsPbBr3, Cs+ ions separate the layers of 2D CsPb2Br5, and the excitons are slowed down by the layered structure [35]. Thus, with more CsPb2Br5 formed, the PL peaks of p-120Fe and p-180Fe ge<sup>t</sup> more blue shifted [36,37]. p-90Fe possesses the lowest PL intensity among all as-synthesized CsPbBr3/MIL-100(Fe) samples.

**Figure 5.** Steady-state PL spectra at an excitation wavelength of 380 nm of the composites. PL spectra of the parent compounds in Supplementary Materials.

As shown in Figure S3a, the pure CsPbBr3 sample displays nanoparticles with a uniform cubic shape. MIL-100(Fe) synthesized in HF-free conditions crystallizes into small nanoparticles with no clearly visible shape compared to the classic octahedral shapes (Figure S3b); this is in line with previous reports [38]. In the composite materials, e.g., p-90Fe, the cubic CsPbBr3 morphology is lost and only agglomerated particles are observed (Figure S3c). Selected location elemental analysis by energy-dispersive X-ray spectroscopy (EDS) confirms the co-existence of both CsPbBr3 and MIL-100(Fe) (Figure S3d).

The components' dispersion and interaction were studied by confocal laser scanning fluorescence measurements on p-90Fe, using a 375 nm laser. Figure 6d showed the wide field image of p-90Fe. Based on the PL spectra mentioned above, two emission channels at 430–470 nm and 505–540 nm were chosen to visualize the MIL-100(Fe) (red colored) (Figure 6a) and CsPbBr3 (green colored) (Figure 6b) distribution, respectively. Both channels were acquired with the same excitation power. As CsPbBr3 has a significantly stronger signal than MIL-100(Fe) upon excitation (Figure S2), Figure 6a was adjusted to be 25% brighter than Figure 6b.

As shown in Figure 6c, yellow shaded areas appear where both signals overlap. Based on particle count, ca. 80% particles were yellowish-green. These uniformly dispersed yellow shades reflect the close contact between MOF and MHP parts in the composite.

The complementary light absorption by CsPbBr3 and MIL-100(Fe) should result in an efficient photocatalytic activity under simulated solar irradiation. Here, CO2 photoreduction under visible light irradiation was chosen to test the photoactivity of the different catalysts. The reaction was performed under 1 bar hydrated CO2 atmosphere, at ambient temperature. CO was found as the

only photoreduction product generated from CO2, and no other carbonaceous products was detected. A series of control experiments were also conducted. No appreciable amounts of CO or other hydrocarbons were detected in the absence of light irradiation, or photocatalyst, or CO2 (under wet He atmosphere).

**Figure 6.** Confocal fluorescence scanning images of (**a**) 430–470 nm emission and (**b**) 505–540 nm emission, (**c**) overlapped image of (**<sup>a</sup>**,**b**), and (**d**) wide field image, using a 375 nm laser excitation of p-90Fe.

Figure 7a shows the time-dependent CO production on the as-synthesized samples during a 4 h experiment. The CO production rate over pure CsPbBr3 and MIL-100(Fe) are similar, about 4.5 μmol g<sup>−</sup><sup>1</sup> h−1. The composite materials show significantly higher activity, and p-90Fe exhibits a maximum CO production of 20.4 μmol g<sup>−</sup><sup>1</sup> <sup>h</sup>−1, which is about 4.5 times higher than the pure constituents. Upon increasing the load of MIL-100(Fe), lower photocatalytic activity is obtained. The decrease in photocatalytic activity may be caused by an increasing amount of CsPb2Br5 in the composites, which is not active upon visible light irradiation. The critical role of H2O was revealed by a test reaction on p-90Fe in high purity CO2 gas without H2O. The CO yield over p-90Fe dropped from 20.4 to 5.3 μmol g<sup>−</sup><sup>1</sup> h−1. The composites' stability was evaluated by four consecutive runs (4 h each, in total 16 h) on p-90Fe. As shown in Figure 7b, the composite exhibits no significant deactivation, and its crystal structure is well maintained after the 16 h photocatalytic reaction (Figure S4).

Two reference samples, p-post and p-mix, with the same amount of MIL-100(Fe) loaded as p-90Fe, were constructed. p-post was synthesized by the anti-solvent deposition of CsPbBr3 onto MIL-100(Fe) [39]. p-mix was prepared through ultrasonically mixing the CsPbBr3 and MIL-100(Fe) powders. Under visible light irradiation, the CO production rates over p-post and p-mix are only 8.7 μmol g<sup>−</sup><sup>1</sup> h−<sup>1</sup> and 6.8 μmol g<sup>−</sup><sup>1</sup> <sup>h</sup>−1, nearly one-third of that obtained via the newly introduced in situ growth route. A summary of the reported photocatalytic CO2 reduction performance on perovskite-based and traditional photocatalysts under various illumination conditions is listed in Table S1.

**Figure 7.** (**a**) Time-dependent CO generation over the synthesized MIL-100(Fe), CsPbBr3, and composites, (**b**) Stability test on p-90Fe for four consecutive runs (4 h each, in total 16 h), and (**c**) CO generation over the photocatalysts under different illumination conditions.

It is observed that CsPbBr3 gradually transforms to CsPb2Br5 in the composites, and CsPb2Br5 is a large bandgap material without visible light response. Therefore, full-spectrum (300–800 nm) measurements were performed on the selected samples to investigate the photocatalytic contribution of CsPb2Br5 in the composites. As shown in Figure 7c, only slight improvements in CO generation were found on CsPbBr3, p-post, and p-mix, compared to that under visible light irradiation (420–800 nm). The CO generation significantly enhanced from 20.4 to 29.6 μmol g<sup>−</sup><sup>1</sup> h−<sup>1</sup> on p-90Fe and from 13.6 to 16.6 μmol g<sup>−</sup><sup>1</sup> h−<sup>1</sup> on p-120Fe. Hence, CsPb2Br5, used to be seen as the undesirable byproduct of CsPbBr3, can contribute to the composites' photoactivity.

It is acknowledged that the photocatalysts' performance can be influenced by the specific surface area. First, gas sorption measurements were performed (Figure S5a). N2 physisorption revealed a negligible surface area for the pure CsPbBr3 microcrystals. The composite materials show a tremendously increased specific surface area between 130 and 400 m<sup>2</sup>/g (Table 3); the specific surface area increases with the amount of MIL-100(Fe). This surface area enhancement favors the exposure of active sites and the adsorption of CO2 molecules. For the composites, a type IV isotherm is observed related to the existence of a mesoporous structure. Further, Figure S5b shows the CO2 adsorption isotherms of the as-obtained samples. The CO2 uptakes in the composites are 15 to 30 times higher than that in pure CsPbBr3, which benefits the photocatalytic efficiency. Figure S6 shows the N2 physisorption isotherms and related pore size distribution of p-90Fe before and after the reaction. After the reaction, the surface area reduced from 201 m<sup>2</sup>/g to 155 m<sup>2</sup>/g, the average pore size decreased from 4.5 nm to 3.7 nm, and the pore volume dropped from 0.21 cm<sup>3</sup>/g to 0.11 cm<sup>3</sup>/g. As can be seen in Figure S6b, p-90Fe has dominant peaks at 1.2 and 2.0 nm, which is the typical pore size distribution of MIL-100(Fe) [40]. After the reaction, the portion of the mesopores around 10 nm decreased, which may be due to the influence of H2O on the particle size and shape during the reaction.

**Table 3.** The specific surface area of the prepared composites.


As shown in Figure 8, the PL decay plots of CsPbBr3 and composite p-90Fe are fitted with a biexponential decay function. The short and long PL lifetimes can be assigned to two different physical origins. The short (τ1) and long (τ2) lifetimes are related to the trap-assisted and exciton recombination, respectively [13]. The average lifetime (τ) of CsPbBr3 exhibits an obvious decrease from 4.4 to 1.6 ns after adding MOF, resulting from the suppressed exciton recombination. MIL-100(Fe) here functions as a quencher of CsPbBr3, endowing the composite with a more effective electron extraction [41].

**Figure 8.** Time-resolved PL spectra of pure CsPbBr3 and p-90Fe fitted with a biexponential decay kinetic, including the corresponding fitting parameters.

The photocatalytic activity of MIL-100(Fe) mainly originates from the direct excitation of the Fe3–μ3–oχ<sup>o</sup> clusters inside the structure. Under light irradiation, Fe–O clusters in MIL-100(Fe) can be excited, transferring an electron from the O2− to Fe3<sup>+</sup> for the formation of Fe2+, which is responsible for CO2 reduction over pure MIL-100(Fe) [22,42]. The combination of the bandgap data with valence band measurements (Figure S7 and Table 2) allow to determine the VB and CB edge potentials of CsPbBr3 to be 0.82 and −1.45 eV, respectively. The VB and CB edge potentials of MIL-100(Fe) are calculated to be 1.75 and −0.93 eV, respectively. Therefore, a typical type II heterojunction was formed by the perfect band structure matching of CsPbBr3 and MIL-100(Fe) [42].

A possible mechanism for the visible-light driven photocatalytic CO2 reduction over the composite is proposed, as shown in Figure 9. Photo-induced electron and hole pairs are generated on CsPbBr3 and MIL-100(Fe) and tend to transfer. The electrons in the conduction band of CsPbBr3 will transfer to that of MIL-100(Fe), where CO2 would be reduced to CO. The holes on the valence band of MIL-100(Fe) would migrate to that of CsPbBr3, where H2O will be trapped to generate O2.

**Figure 9.** Schematics of the CO2 photoreduction process on CsPbBr3/MIL-100(Fe) under visible light irradiation.

With H2O involved in the synthesis, CsPbBr3 gradually converts to CsPb2Br5, which has no visible light response. Revealed by the XRD and UV-vis absorption spectra, p-90Fe has the highest amount of MIL-100(Fe) in the composite while maximally retaining CsPbBr3. MIL-100(Fe) greatly increases the surface area and enhances the visible light absorption ability of CsPbBr3. Upon visible light irradiation, the separation and transfer of the photogenerated charge carriers is promoted in the composites, resulting in enhanced photocatalytic performance. Further increasing the amount of MIL-100(Fe) increases the specific surface area and CO2 uptake, but at the expense of CsPbBr3.

#### **3. Materials and Methods**

### *3.1. Catalyst Synthesis*

#### 3.1.1. Synthesis of CsPbBr3

CsPbBr3 was synthesized by the anti-solventmethod. 2.5mmol cesium bromide (CsBr, 99.9%, Alfa Aesar, Kandel, Germany) and 2 mmol lead (II) bromide (PbBr2, 99.999%, Alfa Aesar, Kandel, Germany) were dissolved in 15 mL dimethyl sulphoxide (DMSO, ≥99.9%, ACS reagent, Sigma-Aldrich, Overijse, Belgium) and stirred for 12 h. The solution was quickly added into 150 mL toluene under stirring. The obtained product was collected by centrifugation, washed with toluene, and dried in a vacuum oven at 80 ◦C.

#### 3.1.2. Synthesis of Pure MIL-100(Fe)

In a typical procedure, 2.02 g iron (III) nitrate nonahydrate (Fe(NO3)3·9H2O)(99%, ACROS Organics, Geel, Belgium), and 0.7 g 1,3,5-benzenetricarboxylic acid (H3BTC, 95%, Sigma-Aldrich, Overijse, Belgium) were added to 5 mL H2O and stirred for 30 min at RT. This mixture was then heated to 95 ◦C and maintained at this temperature for 12 h. After cooling down the mixture, the obtained orange solid product was collected by centrifugation, washed with distilled water, and dried at 80 ◦C.

#### 3.1.3. Synthesis of the CsPbBr3/MIL-100(Fe) Composites

The CsPbBr3/MIL-100(Fe) composites were obtained by in situ growth. The Fe precursor solution for MIL-100(Fe) was made first, by adding 30, 60, 90, 120, and 180 mg Fe(NO3)3·9H2O into 3 mL 1-propanol (99.5%, ACS agent, Fisher Chemical, Merelbeke, Belgium). Next, 0.1 g CsPbBr3 was added to this Fe precursor solution and the mixture was stirred for 12 h at RT. The H3BTC powder was then added with a 3:1 molar ratio of H3BTC to Fe(NO3)3·9H2O. The mixture was heated to 95 ◦C and kept at this temperature for 12 h. The resulting orange solid product was collected by centrifugation, washed by toluene, and dried at 100 ◦C in a vacuum oven. The obtained samples were named as p-xFe (x = the weight of Fe(NO3)3·9H2O added).

For comparison, CsPbBr3 was loaded onto MIL-100(Fe) by anti-solvent deposition [39]. This sample is named as p-post. Furthermore, a physical mixture of the pure CsPbBr3 and MIL-100(Fe) was prepared by ultrasonically mixing the powders, named as p-mix. The weight ratios of MIL-100(Fe) in both p-post and p-mix are 18 wt%. Finally, 90 mg Fe(NO3)3·9H2O was dehydrated in a vacuum oven at 80 ◦C for 4 h before the composite synthesis, and the obtained sample was named as p-90Fe-dehydrated.
