In situ photodeposition of ultra-small palladium particles on TiO2

An in situ approach to generate photocatalysts using a custom-made photocatalytic cell allowed, by synchrotron-based X-ray absorption spectroscopy, the different stages of Pd nucleation onto TiO2 samples to be followed yielding a highly homogeneous distribution of 1 nm palladium nanoparticles.


Introduction
Carlo Lamberti, whether as a teacher, a supervisor or a colleague of us, had led numerous research initiatives that our team is still actively pursuing.Those mainly include studies of catalysts represented by supported metal nanoparticles (Bugaev et al., 2013(Bugaev et al., , 2016;;Bugaev, Guda et al., 2019) or functionalized metal-organic frameworks (Braglia, Borfecchia, Martini et al., 2017;Braglia, Borfecchia, Lomachenko et al., 2017;Bugaev, Skorynina et al., 2019;Skorynina et al., 2024;Pnevskaya & Bugaev, 2023;Pnevskaya et al., 2021;Kamyshova et al., 2020;Lomachenko et al., 2018;Gallo et al., 2017;Usoltsev et al., 2024) under reaction conditions.Highlighting the role of in situ and operando spectroscopies for in depth understanding of functional materials, these approaches were successfully applied to real industrial catalysts (Fang et al., 2021;Groppo et al., 2018;Bugaev et al., 2021), and were further extended to investigate high-pressure industrially relevant processes (Janssens et al., 2022;Ramirez et al., 2021).In many of the aforementioned works, the catalysts themselves were also synthesized in situ, followed by the application of the catalytic reaction conditions, and Carlo was one of the key actors on the synchrotron-world scene, advancing the development of in situ and operando methodologies.During the 2018 Faraday Discussion conference (Adishev et al., 2018), he was suggesting pushing the development of a new direction, calling it 'construendo' or 'aedificando', analogous to operando spectroscopy, to focus on the formation steps of nano-systems during their synthesis.Although that proposal was made with a touch of Carlo's playful and adventurous spirit, the topic was, and remains, of significant importance, supported by numerous successful examples in the literature.
Application of X-ray absorption spectroscopy (XAS), one of Carlo's biggest passions ( van Bokhoven & Lamberti, 2016;Mino et al., 2013;Giannini et al., 2020), in heterogeneous catalysis has long become a widely used and standard tool, despite the need of using synchrotron radiation to follow fast processes (Abdala et al., 2012;Weckhuysen, 2003;Bugaev et al., 2024), while for slow processes, i.e. hours or days, laboratory XAS can be applied (Sherborne & Nguyen, 2015).The area of photocatalysis, however, remains more demanding for several reasons, among which are (i) the typical presence of a liquid phase, (ii) lower concentrations of active metallic species with respect to the substrate, and (iii) a more complicated design of the in situ/operando cells (Piccolo et al., 2020;Khare et al., 2020;Issa Hamoud et al., 2022).If the photocatalyst is suspended in a solution, the detection of XAS signal is complicated; therefore, an approach when the catalyst is deposited on a window of an in situ cell is more often exploited (Mu et al., 2023;Haselmann et al., 2020).However, due to the substantial difference in penetration depth between UV or visible light and hard X-ray photons, it is possible that only a minor fraction of the species probed by XAS are actually involved in the photocatalytic process, therefore making structure-activity correlations irrelevant (Jacobs et al., 1989).
This contribution describes the in situ synthesis of a Pd/TiO 2 catalyst in a home-made photocatalytic cell.The process of in situ photodeposition of Pd and the formation and evolution of different Pd species during synthesis are probed by synchrotron-based XAS under reaction conditions.Furthermore, the morphology of the in situ prepared catalyst and its counterpart, made in a conventional batch reactor with suspended TiO 2 , is investigated by high-angle annular darkfield scanning transmission electron microscopy (HAADF-STEM).The combination of in situ XAS spectroscopy with ex situ HAADF STEM has been demonstrated as a powerful workflow for detailed investigation of the formation mechanisms of the current catalysts (Pastor et al., 2024).

Chemicals
Ex situ Pd/TiO 2 samples were prepared using P25-type TiO 2 (Evonik Operations GmbH) as a support.Palladium atomic absorption standard solution (1 mg ml À 1 Pd in 5% HCl, Sigma-Aldrich) was used as a metal source.Both the support and the metal source, in a ratio to obtain 2 wt% Pd/TiO 2 , were dissolved in 0.1 M solution of formic acid in Mili-Q water with a total volume of 100 ml inside a batch-type quartz reactor cell, which was permanently flushed with N 2 .After 15 min of flushing, the solution was irradiated by UV light (� = 365 nm) for 20 min with irradiance E = 30 W m À 2 .
In situ Pd/TiO 2 samples were prepared directly during the XAS experiment (see below).First, the P25 powder was dissolved in water and the resulting TiO 2 ink (� 15 mg of dry powder, measured for each sample) was deposited on the window of the home-made operando photocatalytic cell (Kozyr et al., 2023) and left drying.The cell was then filled with a mixture of 15 ml of Mili-Q water and 10 mL of formic acid.The amount of Pd source (1 mg ml À 1 Pd in 5% HCl, Sigma-Aldrich) was calculated for every sample (� 300 mL) depending on the amount of TiO 2 , in order to reach the theoretical maximal loading of 2 wt% Pd/TiO 2 .The cell was tightly sealed and permanently flushed with He at 20 ml min À 1 , the outlet being monitored by a Cirrus 3-XD mass spectrometer (MKS Instruments).Then, the cell window was irradiated by a remotely controlled UV LED source (� = 365 nm).After 20 min of total exposure to UV radiation, an additional 1 ml of formic acid was added to the cell as a hole scavenger and the cell was continuously exposed to UV irradiation.

In situ XAS measurements
In situ Pd K-edge XAS data were collected at the NOTOS beamline of ALBA synchrotron (Spain).The NOTOS beamline source is a 1.42 T bending magnet with 8.5 keV critical energy and a maximum flux of 10 11 photons s À 1 at this critical energy.The beamline optics consists of a cylindrical mirror (M1) with downward orientation, with a Si and a Rh stripe, that collimates the beam onto a double-crystal monochromator (DCM) made by CINEL.The DCM is a direct drive monochromator with fixed exit and two pairs of crystals, Si(111) and Si (220), enabling an energy range between 4.7 and 30 keV.The beam coming from the DCM impinges on an Rh-coated double-channel toroidal mirror (M2, with 34 and 55 mm sagittal radii of curvature), which allows for horizontal focusing of the beam on the sample.
For this particular experiment, the beamline optics was optimized for the Pd K-edge (24.35 keV) by using the Rhcoated stripe of M1, the Si(111) crystals pair and the 34 mm channel of M2.The beam size dimensions were 250 mm � 370 mm (H � V) approximately.The DCM was operated in continuous scanning mode and the energy was scanned within the 24.2-24.9keV range (� 1 min per spectrum).The cell window was oriented perpendicularly to the incoming X-ray beam, allowing the collection of both fluorescence and transmission signals.The transmission measurements were acquired with three ion chambers as detectors with a proper gas mixture, optimized to improve the measured signal (absorbing � 15%, 85% and 85%, respectively).The homemade cell was placed between the first and second ionization chambers, and a Pd metallic reference foil was placed between the second and the third to check for energy calibration.The fluorescence detector, a 13-channel Si drift detector from Mirion Technologies (Canberra Olen), was oriented at � 45 � with respect to the incoming beam, from the same side as the UV source (see Fig. 1).
x-ray spectroscopy for functional materials 1072 Elizaveta Kozyr et al. � In situ photodeposition of ultra-small palladium particles Among different cell windows materials, cellulose acetate film demonstrated the highest stability under the beam, while for other types of windows (quartz glass, Mylar film) beaminduced implantation of palladium into window material was observed during beam damage tests.To further reduce the beam effects, the cell was automatically moved by 0.5 mm in the horizontal direction (and when the row is finished, by 0.5 mm vertically) after each XAS spectrum.This procedure limited the maximum exposure of any fraction of the sample to the X-ray beam to � 1 min and showed that no palladium nanoparticles were formed on the window in the absence of UV light.XAS data processing was carried out using the Demeter package (Ravel & Newville, 2005).

HAADF-STEM measurements
HAADF-STEM micrographs and energy-dispersive X-ray spectroscopy (EDX) elemental maps were acquired in a double-corrected Spectra 300 STEM microscope (Thermo-Fisher Scientific) operated at 300 kV.Histograms of particle size distribution were obtained by measuring 150-250 particles for each sample.For sample deposition, powder samples were dispersed in water, sonicated and deposited on lacey carbon Cu grids with an ultrathin carbon layer.

In situ photodeposition of Pd on TiO 2
Two general strategies can be exploited when performing photocatalytic reactions (Wei et al., 2022).In the first one, the photocatalyst is dispersed or suspended in a solution, which is believed to result in a better mass transfer.However, the distribution of light from the external source through the volume of such reactors is typically inhomogeneous.An alternative approach uses an immobilized catalyst, typically deposited on a UV-transparent window, which can be easily illuminated in a constant manner.The latter is also preferable for XAS studies, as the amounts of catalyst and/or the cell dimensions are typically incompatible with data collection in transmission geometry, while deposition of the catalyst on the window of a photocatalytic cell allows for fluorescence data collection.Still, another problem remains due to the fact that the penetration depth of X-rays within TiO 2 varies from a tenth or hundreds of micrometres at K-edges of 3d elements or L 3 -edges of 5d elements to millimetres for 4d elements, while the penetration depth of UV light into TiO 2 is typically in the nanometre to micrometre range.As a result, the majority of metal species, that are easily probed by the X-ray beam, may not be accessible by UV radiation and remain inactive for the photocatalytic reactions.This issue can be solved by precise control of the thickness of the photocatalyst or by in situ photodeposition of metals on TiO 2 .
Here, we focus on the second approach, in which Pd was in situ photodeposited on TiO 2 .The procedure was performed using a home-made operando cell shown in Fig. 1.A thin layer of TiO 2 was first applied on the front cell window and the cell was then filled with a solution of Pd-precursor under continuous stirring and flushing with He.The UV radiation was applied from the outside of the cell to the front window by setting the irradiance to 400 W m À 2 .The photodeposition of Pd onto the TiO 2 layer was probed by XAS in fluorescence mode from the same side.The advantages of this approach are: (i) the homogeneous irradiance of the catalyst over time; (ii) the distribution of the Pd species only in the UV-accessible part of TiO 2 , i.e. all Pd/TiO 2 species can be further considered to be photoactive; and (iii) the minimization of the possible direct interaction of UV light with Pd species in solution.
To follow the progressive growth and evolution of Pd species on TiO 2 upon photodeposition, the UV irradiation was switched on for 10 s, 20 s, 30 s, 4 min and 15 min, resulting in cumulative irradiation times of 10 s, 30 s, 1 min, 5 min and 20 min, respectively.After each step, XAS spectra were collected to probe the state of the photodeposited Pd.Fig. 2 shows the evolution of XANES and EXAFS data indicating a gradual formation of metallic Pd species, as compared with the reference Pd foil.Indeed, the photodeposition process was accompanied by the progressive formation of distinct XANES features, characteristic of metallic Pd, a decrease of the firstshell Pd-Cl contribution in EXAFS and a growth of the Pd-Pd one.Due to the nanometric dimensions, the intensity of the latter was smaller with respect to the Pd reference foil.The average Pd-Pd coordination number obtained through EXAFS fitting was 7.9 � 0.8, corresponding to an average particle size of � 1.2 nm, assuming their face-centred-cubiclike structure.
At the same time, a decrease of the edge jump at the Pd K-edge in the transmission signal was observed.This was explained by the fact that the transmission signal was initially dominated by the Pd species homogeneously distributed in the solution; however, when Pd was being deposited on the TiO 2 layer, which was inhomogeneous both in its thickness and distribution over the window, this inhomogeneity led to a lower edge jump.The trend estimated based on this fact is shown in Fig. S1 of the supporting information while the results of elemental composition analysis from EDX data are reported in Table S1.
Interestingly, already after 10 s the edge position in XANES corresponded to the Pd(0) oxidation state [Fig.2(a)].Moreover, this spectrum could not be reproduced by a linear combination of the reference spectra of H 2 PdCl 4 (the spectrum of this reference was identical to that of the precursor as shown in Fig. S2) and Pd foil, which is shown by the absence of isosbestic points in these spectra.This gives an important insight into the formation mechanism of Pd/TiO 2 catalysts, suggesting that the photoreduction of Pd started with formation of isolated Pd(0) species, followed by a subsequent growth of Pd nanoparticles, most likely around those initial Pd(0) sites.The signals of H 2 and CO 2 , due to the decomposition of formic acid, was observed in mass spectrometry data (Fig. S3) during each UV illumination.Finally, we observed that the average Pd-Pd interatomic distance in the formed nanoparticles was 2.81 � 0.01 A ˚, which is significantly higher than the value of 2.74 � 0.01 A ˚determined for Pd foil.The shaping of XANES spectra, namely the positions and relative intensities of the first two maxima, indicated that such elongation was caused by the formation of palladium hydride (Bugaev et al., 2013(Bugaev et al., , 2016)), while the presence of a carbide like phase was unlikely (Bugaev, Guda et al., 2019;Usoltsev et al., 2024).
The sample prepared by conventional batch synthesis was also investigated and was applied to the window of the photocatalytic cell in a similar way, but already as Pd/TiO 2 .The average Pd-Pd coordination number for this sample was higher (11.6 � 0.6) indicating that bigger particles were present.Upon the addition of the formic acid solution, the typical XANES features of palladium hydride also appeared [Fig.3(a)] and the average Pd-Pd interatomic distances increased from 2.73 � 0.01 A ˚to 2.84 � 0.01 A ˚, which is visually shown in Fig. 3(b).The fact that the elongation of the interatomic distance upon formation of the palladium hydride phase was larger for this sample, compared with the one generated in situ, correlates with the assumption of a larger  particle size.Interestingly, the hydride features in XANES and increased Pd-Pd distances remained even after the solution was removed.

Morphology of Pd/TiO 2 samples
As the next step, we investigated the series of Pd/TiO 2 samples using HAADF-STEM.The samples prepared in situ and collected from the window of the photocatalytic cell after XAS investigation will be referred to as Pd insitu /TiO 2 and the sample synthesized in the conventional batch procedure as Pd batch /TiO 2 .As evident from Figs. 4 and 5 and additional micrographs shown in the supporting information, the in situ generated sample exhibited a homogeneous distribution of palladium over TiO 2 with a narrow particle size distribution centered around 1 nm [Fig. 4(c)], in agreement with EXAFS data.In contrast, a very inhomogeneous distribution of Pd was observed for Pd batch /TiO 2 .This sample had bigger particles, which correlates with the EXAFS results, and a broader particle size distribution [Fig.5(c)].Moreover, particles or particle agglomerates with sizes up to 100 nm were detected, which were almost absent for Pd insitu /TiO 2 (Fig. S8).
Further insights were gained from two additional samples that were generated in situ during XAS experiments following a similar procedure as for Pd insitu /TiO 2 but using a lower irradiance of 80 and 200 W m À 2 .Unfortunately, a smaller degree of Pd incorporation into TiO 2 limited the interpretation of XAS data (Fig. S9), although the presence of Pd( 0  spectra.HAADF-STEM data of these samples highlighted the presence of ultra-small Pd clusters (<1 nm) or single-atom Pd sites (Fig. S10), confirming the photodeposition mechanism proposed in the previous section.Bigger particles obtained for Pd batch /TiO 2 can be explained by the fact that part of Pd was reduced in the liquid phase, acting as nucleation centers that facilitated the growth of larger particles that were then deposited on TiO 2 .This correlates with our recent study on Ptbased samples (Kozyr et al., 2023), where photoinduced Pt reduction and formation of Pt nanoparticles were observed even in the absence of TiO 2 .Conversely, in Pd insitu /TiO 2 , the interaction of the solution with UV light was minimized since the radiation was mainly absorbed by the TiO 2 layer.Thus, photoexcited TiO 2 could act by providing the binding sites for Pd(0) single atoms while the maximum particle size could be limited due to the interaction with the support.

Conclusions
We have presented an efficient approach for in situ generation of Pd/TiO 2 photocatalysts that allows all Pd sites to be equally accessible by both X-ray and UV radiation.In situ generation of the Pd/TiO 2 sample using TiO 2 immobilized on the front window of the photocatalytic cell was shown to produce a homogeneous distribution of Pd over the photoactive support, with a narrow particle size distribution and an average particle size of � 1 nm.In contrast, bigger particles and a more inhomogeneous distribution were observed for batch synthesis with similar parameters.In situ XAS spectroscopy complemented by HAADF-STEM data revealed the formation of either single-atom Pd(0) sites or ultra-small Pd(0) clusters, as soon as Pd species were photodeposited on TiO 2 , that subsequently acted as nucleation sites for the growth of nanometresized Pd particles.The state of Pd in the presence of water and hole scavenger (formic acid) was identified as Pd hydride.

Figure 1
Figure 1 Schematic (a) and photograph (b) of the operando photocatalytic cell.
Figure3XANES (a) and phase-uncorrected FT-EXAFS (b) data for the sample, synthesized using a conventional batch procedure, in the initial state (black), after addition of water and formic acid to the photocatalytic cell (red) and after performing UV irradiation and removing the liquid (blue).

Figure 2 XANES
Figure 2 XANES (a) and phase-uncorrected FT-EXAFS (b) data collected during in situ photodeposition of palladium on TiO 2 .Colored lines correspond to different irradiation times.Dashed black and gray lines correspond to the reference H 2 PdCl 4 and Pd foil, respectively.
Figure 4 STEM characterization of the Pd insitu sample.Low (a) and high (b) magnification HAADF-STEM images.(c) Histogram of Pd particle size distribution.(d) EDX elemental maps corresponding to the area shown in part (a).

Figure 5 STEM
Figure 5 STEM characterization of the Pd batch sample.Low (a) and high (b) magnification HAADF-STEM images.(c) Histogram of Pd particle size distribution.(d) EDX elemental maps corresponding to the area shown in part (a).