Kinetics of Lifetime Changes in Bimetallic Nanocatalysts Revealed by Quick X-ray Absorption Spectroscopy.

Alloyed metal nanocatalysts are of environmental and economic importance in a plethora of chemical technologies. During the catalyst lifetime, supported alloy nanoparticles undergo dynamic changes which are well-recognized but still poorly understood. High-temperature O2 -H2 redox cycling was applied to mimic the lifetime changes in model Pt13 In9 nanocatalysts, while monitoring the induced changes by in situ quick X-ray absorption spectroscopy with one-second resolution. The different reaction steps involved in repeated Pt13 In9 segregation-alloying are identified and kinetically characterized at the single-cycle level. Over longer time scales, sintering phenomena are substantiated and the intraparticle structure is revealed throughout the catalyst lifetime. The in situ time-resolved observation of the dynamic habits of alloyed nanoparticles and their kinetic description can impact catalysis and other fields involving (bi)metallic nanoalloys.

Bimetallic catalysts can exhibit drastically increased selectivity, activity, and stability compared to their monometallic siblings owing to an interplay of electronic and geometric effects. They are the workhorses in widespread technologies, such as car exhaust converters,1 and fuel cells,2 and have acquired a pivotal role in chemical industry. Within the library of bimetallic catalysts, archetypal Pt‐based catalysts promote a vast array of reactions,3 such as NO reduction1 (e.g. Pt‐Rh) to alkane dehydrogenation4, 5, 6 (e.g. Pt‐In). In reaction, these catalysts operate at high temperature under rapidly changing reductive and oxidative environments.7, 8 For example, dehydrogenation catalysts require cyclic O2‐H2 redox treatments to regenerate (O2) and reactivate (H2) Pt‐In nanoalloys after coking in high‐temperature alkane flows.9, 10, 11

Whilst catalysts were originally assumed to be static substances, intense research in the past decades has witnessed that (bi)metal catalysts are dynamically changing nanomaterials under reaction conditions.12 During the catalyst lifetime, multiple atomic scale processes shape the working state of the catalyst and thereby influence its performance. It goes without saying that understanding these processes and detailing their kinetics is of crucial importance to control their occurrence and thus the catalyst performance. To this aim, techniques which provide high time resolution, but do not require long range order and can be applied in situ, are crucial for interrogating the kinetic behavior of (often disordered) bimetallic nanocatalysts under operating conditions. For example, in  situ, dispersive and quick X‐ray absorption spectroscopy (XAS) have established a role of utmost importance in uncovering the motifs of (bi)metallic nanoparticle restructuring.13, 14, 15, 16, 17, 18

Herein, we probe the dynamic lifetime changes in a Pt13In9 model nanocatalyst induced by high‐temperature O2–H2 redox cycling with quick X‐ray absorption spectroscopy (QXAS). Such Pt‐In catalysts show great promise for alkane dehydrogenation, hydroconversion of bio‐based oxygenates, and as direct methanol and ethanol fuel cells.3 Additionally, efforts have been undertaken recently by our groups19, 20, 21, 22 and others23, 24 to understand their formation and control their intermetallic composition. To monitor the kinetic habits of these catalysts in situ with high time resolution, QXAS spectra are recorded at the Pt L3‐edge with one second time resolution (Figure 1 a). While pioneering in situ QXAS studies on the formation and working state of nanocatalysts are available,7, 25, 26, 27, 28, 29, 30 the quantitative extraction of kinetic data and the construction of a full reaction mechanism has, to our best knowledge, not been reported for (bi)metallic nanoparticles. For this work, phase‐pure Pt‐In alloys were fabricated by atomic layer deposition19 (ALD) and their XAS spectra measured to benchmark the QXAS data in their Pt‐In composition during O2‐H2 cycling. This allows to probe Pt‐In nanoalloy‐segregation dynamics on a quantitative level and to extract kinetic information.

Figure 1

a) Synchrotron source producing an X‐ray beam by a quick‐scanning channel‐cut monochromator; the X‐ray beam irradiates the catalyst bed, contained in a capillary by which cyclic H2‐O2 pulses are flown to induce structural changes in the Pt‐In nanocatalyst. The X‐ray absorption spectrum is measured by the X‐ray detector with 1 s time resolution. b) Pt L3‐edge XANES white line (WL) analysis yielding the (top) WL height in terms of normalized absorption and (bottom) WL energy (at the WL maximum) in eV, as a function of the number of H2‐O2 redox cycles. The experiments are performed at variable temperature, ranging from 923 K down to 723 K. c) XANES reference spectra of PtO2, Pt and Pt13In9. HAADF‐STEM images of Pt‐In nanoparticle catalysts after d, e) 1 and f, g) 60 redox cycles. e) HAADF‐STEM of Pt‐In nanoparticle after the first H2‐O2 cycle, that is, after re‐oxidation and alloy segregation, showing a bright contrast Pt core surrounded by a dark contrast In oxide shell. f) HAADF‐STEM image of sintered nanoparticles after 60 redox cycles in oxidized (alloy segregated) state, showing Pt nanoworm network formation, surrounded by In oxide, together with magnified region (red square) and corresponding FFT.

The white line (WL) height and energy position of the X‐ray absorption near edge structure (XANES) spectra are used to probe the state of Pt in situ during repetitive H2‐O2 redox cycling (Figure 1 b, Supporting Information, section S4). Reference XANES spectra show that 1) Pt oxidation mainly results in an increase in the WL height (PtO2), while 2) Pt alloying towards Pt13In9 primarily results in a blue shift of the WL energy (Figure 1 c; Supporting Information, sections S2,4). The WL height reversibly alternates from low to high values, respectively during H2 and O2 pulses, and vice versa for the WL energy. This suggests that Pt reversibly undergoes PtOx reduction and Pt‐In alloying (H2), followed by Pt‐In alloy segregation and re‐oxidation into PtO (O2).

The WL height and energy position remain constant at the end of each H2 pulse, implying that the alloy composition is stable for at least 60 H2‐O2 cycles. In contrast, the WL height strongly decreases in O2 with increasing number of redox cycles as a result of Pt nanoparticle (NP) sintering. HAADF‐STEM confirms that initially well‐dispersed NPs (ca. 2 nm, Figure 1 d) sinter into larger ones (ca. 10 nm, Figure 1 g), steered by H2‐O2 cycling. A magnified view of a small NP obtained after one H2‐O2 redox cycle, that is, in oxidized state, shows that the NP is composed of a metallic Pt core, surrounded by an In2O3 crust (Figure 1 e). Large NPs after 60 H2‐O2 redox cycles reveal a nanocomposite consisting of a Pt nanoworm network surrounded by In2O3, as confirmed by EDX (Supporting Information, section S7) and fast Fourier transformed (FFT) analysis showing face centered cubic (fcc) Pt (Figure 1 f). Such nanocomposite structure suggests that nanoalloys decompose spinodally, as caused by kinetically rapid segregation.

The dynamic changes of Pt are now studied at the single cycle level. Figure 2 a displays the Pt L3‐edge wavelet transformed (WT) quick extended X‐ray absorption fine structure (EXAFS) magnitude at specific points in redox cycle 21.20 At the end of the H2 pulse, the WT‐EXAFS magnitude shows both Pt and In around Pt and mix to form a Pt‐In alloy (Figure 2 a,1; 116 s in H2). In the middle of the O2 pulse both Pt and O are observed, but no In, indicating that the Pt‐In alloy has decomposed by segregation (Figure 2 a,3; 60 s in O2). This results in phase‐pure Pt with some degree of surface oxidation (suggested by XANES). After the start in H2, only In is visible in WT‐EXAFS, while Pt is barely observed (Figure 2 a,5; 4 s in H2), showing that the Pt‐In alloy is still evolving to a local equilibrium state and might be disordered. At the very start of the O2 and H2 pulses, WT‐EXAFS plots show intermediate states with combined contributions from Pt, In and O (Figure 2 a,2 and 2.a,4; 0 s).

Figure 2

a) Pt L3‐edge wavelet transformed (WT) quick EXAFS magnitude maps at five points in time (1–5) at the single cycle level after 21 H2‐O2 redox cycles. The plots reveal the element type (k‐axis) and position (R‐axis) of neighbors around Pt, namely In, O or Pt. The color bar represents the normalized intensity (0–1) of the WT QEXAFS magnitude. b) The WL height and energy versus time within the 21st redox cycle. Light blue (navy blue) corresponds to H2 (O2) gas environment resp. in panels (b–d). c) The transient evolution of the WL maximum during H2‐O2 redox cycle 21 at 893 K, evolving from Pt13In9 to a segregated PtO state (O2) and back to Pt13In9 (H2) via the Pt3In intermediate. The WL positions of Pt, Pt3In and Pt13In9 references are shown in the squares. d) Mechanism of reversible Pt‐In segregation, Pt oxidation and Pt‐In re‐alloying at the single H2‐O2 redox level: (2–5) simultaneous Pt13In9 segregation, Pt surface oxidation, (5–8,1) consecutive Pt reduction and Pt‐In alloying.

The XANES WL height and energy position in Figure 2 b uncover the kinetics of the processes occurring within one H2‐O2 redox cycle. During the O2 pulse, the decomposition of the initial Pt‐In nanoalloy into fcc Pt and In (oxide) is rapid (WL energy shift), clarifying the spinodal nanostructure observed by HAADF‐STEM for Pt‐In2O3 nanonetworks. Simultaneous to alloy segregation, a slower Pt oxidation process takes place at the Pt surface, which is not entirely stabilized even within the 120 s O2 pulse (WL height). In contrast to Pt oxidation, Pt reduction at the start of the H2 pulse is completed in 2 s. Thereafter, Pt‐In alloying takes place, which in contrast to Pt oxidation stabilizes within the H2 period.

Figure 2 c shows the evolution of the WL in time for both the O2 (navy blue) and H2 (light blue) pulses. Additionally, the white line positions of ALD‐derived phase‐pure Pt, Pt3In, and Pt13In9 are depicted (Supporting Information, S2). Notably, the final WL position in H2 matches to a stoichiometric Pt13In9 alloy. There is a marked difference between the alloy formation and decomposition process. Pt‐In alloy formation sets in after full reduction of PtO into Pt by a further drop in WL height (consecutive processes). The evolution from fcc Pt to Pt13In9 presumably occurs via the fcc Pt3In intermediate, established by the gradual but rapid incorporation of In into fcc Pt. Thereafter, the further uptake of In transforms Pt3In (75 % Pt) into intermetallic Pt13In9 (59 % Pt), which is kinetically slower. In contrast, Pt‐In alloy decomposition occurs simultaneously with Pt surface oxidation to PtO (parallel processes). Instead of evolving via Pt3In and Pt references towards PtO, the WL height shows a monotonic, upward trend that is intermediate to the phase‐pure reference positions. Figure 2 d illustrates the full redox process.

To perform a full kinetic analysis, the 60 H2‐O2 redox cycles are executed at temperatures ranging from 923 K down to 723 K in steps of 50 K for each 10 cycles, except for 823 K which is kept for 20 cycles (Figure 1 b). The WL height and energy are used for kinetic modeling of the degree of Pt oxidation and Pt‐In alloying over time, respectively. First, WL‐time data are averaged over all cycles at a given temperature to reduce noise (Supporting Information, section S6). Next, the WL time evolution is modelled by a zero‐ and/or first‐order rate law. This yields rate coefficients for each temperature, from which Arrhenius plots are constructed and apparent activation energies E a are estimated for the reduction/alloying and segregation/ oxidation processes (Figure 3 a,b). Additionally, 1) an estimated content of Pt in Pt‐In alloys during segregation and alloying (resp. Figure 3 c,d), and 2) the extent of Pt oxidation during H2/O2 oxidation/reduction (resp. Figure 3 e,f, S6) are extracted.

Figure 3

a) The Arrhenius plots for (green) Pt‐In segregation, (red) Pt surface oxidation, and (blue) Pt diffusion‐controlled oxidation process; c)  green, and e) red and blue, respectively, are their modeled curves to WL XANES data (darker colors correspond to lower temperatures). b) Arrhenius plot for (green) Pt‐In alloying and (red) Pt reduction, and their concomitant fits to the WL XANES during d) alloying (modeled) and f) reduction (Supporting Information, section S6) as a function of time. g) Ea versus ln(k) for (red, top) Pt oxidation, (green, top) Pt‐In segregation in O2 and (green, bottom) Pt‐In alloying and (red, bottom) PtO reduction in H2. The rectangle width represents the range that ln(k) varies across the measured temperature range, while its height is the 66 % confidence interval of E a. h) reaction mechanism derived from Arrhenius plots in (a) and (b). In2O3 activation energies are obtained from redox data collected at different temperature at the In K‐edge as further detailed at the end of Suporting Information, section S6.

During the O2 pulse, the majority of In dealloys by segregating out of the Pt13In9 alloy in 5 seconds by a first‐order process (Figure 3 c, green). The much slower Pt surface oxidation can only be modelled by combining a first‐order rate law (Figure 3 e, red) with a zero‐order reaction with constant rate (Figure 3 e, blue). The latter leads to a continuous increase in Pt oxidation state throughout the entire O2 pulse, and is presumably caused by diffusion controlled redispersion of Pt into smaller NPs after spinodal decomposition. The first‐ order processes, namely Pt13In9 dealloying and initial Pt surface oxidation, show strong temperature dependence. Their Arrhenius plots (Figure 3 a, red) exhibit linear correlations with relatively high slopes, yielding apparent activation energies of 67±1 kJ mol−1 for Pt13In9 dealloying and 64±4 kJ mol−1 for initial Pt surface oxidation. The zero‐order process approaches a zero activation energy (Supporting Information, section S6). Notably, the order of magnitude of the rate coefficients of Pt surface oxidation accords with an earlier study,31 assessing Pt NP oxidation only.

During the H2 pulse, surface oxidized Pt is fully reduced to metallic Pt with a high rate (in 1.5 s; Figure 3 f). The activation energy retrieved in a qualitative way from the Arrhenius plot approaches zero (for more information see Supporting Information, section S6,and Figure 3 b). Similar to Pt13In9 segregation, Pt‐In alloying is described by a first‐order rate law with now an activation energy of only 12±3 kJ mol−1 (Figure 3 d).

Figure 3 g and h present an overview of E a‐ln(k) space and a reaction mechanism for the Pt13In9‐PtO‐Pt13In9 redox process. The E a–ln(k) plot shows that the apparent activation energies for Pt13In9 segregation and Pt oxidation in O2 are significantly higher compared to PtO reduction and Pt‐In alloying in H2, and therefore require thermal activation. While Pt surface oxidation (Figure 3 g, O2‐red) is the rate‐determining step in O2, In2O3 reduction (Supporting Information, section S6, last page) together with Pt‐In alloying (Figure 3 g, H2‐green) are kinetically the slowest processes in H2. In O2 flow, the stability of In2O3/PtO drives the segregation of Pt13In9 to segregated In/Pt, while the driving force for In2O3/PtO reduction and alloying in H2 is the stability of the Pt13In9 nanoalloy (Figure 3 h).

In summary, we have demonstrated that kinetic modeling of in situ quick X‐ray absorption spectroscopy data can result in the unrivaled identification of the reaction mechanism describing the dynamic restructuring of nanoparticles. The approach allowed to elucidate the steps involved in alloying‐segregation of Pt13In9 alloys into In2O3/PtO nanocomposites during high‐temperature H2–O2 redox cycling. In O2, Pt13In9 decomposition and Pt surface oxidation are discerned as simultaneous processes exhibiting high activation energy in which Pt oxidation is rate‐determining. In contrast, the reverse processes in H2 drive the equilibrium state from In2O3/PtO back to Pt13In9 through In2O3 and PtO reduction followed by Pt‐In alloying, both exhibiting low apparent activation energies. This QXAS approach can be applied to a wide variety of (bi)metallic nanocatalysts and by extension to any field studying dynamic changes in bimetallic nanoparticles and nanomaterials.

Conflict of interest

The authors declare no conflict of interest.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supplementary

Click here for additional data file.