Direct hydrogen peroxide synthesis Hydrogen peroxide is synthesized industrially without direct contact of hydrogen and oxygen in order to achieve high concentrations. For many applications, only dilute aqueous solutions are needed. Freakley et al. report an improvement in the direct synthesis of hydrogen peroxide over using palladium-tin alloys. This catalyst still achieves selectivities of >95%, like palladium-gold alloys, but is cheaper and can suppress reactions that decompose the product. Science, this issue p. 965

Abstract The direct synthesis of hydrogen peroxide (H 2 O 2 ) from H 2 and O 2 represents a potentially atom-efficient alternative to the current industrial indirect process. We show that the addition of tin to palladium catalysts coupled with an appropriate heat treatment cycle switches off the sequential hydrogenation and decomposition reactions, enabling selectivities of >95% toward H 2 O 2 . This effect arises from a tin oxide surface layer that encapsulates small Pd-rich particles while leaving larger Pd-Sn alloy particles exposed. We show that this effect is a general feature for oxide-supported Pd catalysts containing an appropriate second metal oxide component, and we set out the design principles for producing high-selectivity Pd-based catalysts for direct H 2 O 2 production that do not contain gold.

Currently, the demand for H 2 O 2 is met by an indirect process, which produces H 2 O 2 through the sequential hydrogenation and oxidation of a substituted anthraquinone (1). For economic reasons, the process is operated at large scale and produces concentrated H 2 O 2 . In reality, many applications, such as disinfection and water purification, require only dilute H 2 O 2 , which means that concentrated H 2 O 2 must be diluted at the point of use. Research into the direct synthesis of H 2 O 2 from H 2 and O 2 as a more suitable solution to small-scale, on-site H 2 O 2 production has focused on palladium (Pd)–based catalysts (2–4). However, H 2 O 2 is itself highly reactive, and the presence of H 2 favors hydrogenation and decomposition reactions that form water. The addition of strong acids and halides to the reaction medium can suppress the sequential hydrogenation and degradation in supported Pd catalysts (5) but can also promote metal leaching and requires further purification of the H 2 O 2 before use.

Bimetallic Au-Pd alloy catalysts have been extensively studied as catalysts for the direct H 2 O 2 synthesis reaction on a number of support materials, including TiO 2 , SiO 2 , and activated carbon (6–9). Yields comparable to monometallic Pd catalysts can be achieved without the need for acid and halide additives in the reaction mixture, and 95% selectivity to H 2 O 2 can be achieved with Au-Pd alloy nanoparticles (NPs) dispersed on an acid-pretreated activated carbon support material (10). Hydrogen peroxide hydrogenation could be decoupled from H 2 O 2 synthesis with an acid pretreatment that blocked sites on the carbon support material responsible for H 2 O 2 degradation (10). Although this approach was very successful on an activated carbon support material, the same blocking effect could not be fully achieved on other commercial support materials such as SiO 2 and TiO 2 .

Because O 2 dissociation is undesirable in the direct synthesis of H 2 O 2 , the reaction can be treated as a selective hydrogenation of O 2 . We explored other Pd-metal combinations that are used for selective hydrogenation reactions as potential catalysts for H 2 O 2 synthesis, focusing on nonprecious metals to lower costs. Tin (Sn) has been used to modify hydrogenation catalysts in reactions such as the selective hydrogenation of 1,3-butadiene (11). Further examples have been reported for the liquid-phase hydrogenation of hexa-1,3-diene and hexa-1,5-diene (12) as well as the hydrogenation of unsaturated alcohols (13). The addition of Sn to Pd or Pt can alter the behavior of the catalyst during hydrogenation reactions and, in particular, may have an effect on subsequent reactions of the products with the catalyst.

We report the development of Sn-containing Pd catalysts on commercially available TiO 2 and SiO 2 supports that can achieve >95% selectivity toward direct H 2 O 2 synthesis. These catalysts, after being subjected to an appropriate heat treatment regimen, obviate the need for pretreating the support with acids and contain far less precious metal than Au-Pd catalysts. We also present the general principles whereby high-selectivity catalysts can be obtained with other Pd-metal combinations.

Simple impregnation of Au and Pd metal salts onto many catalyst supports has been shown to generate highly active catalysts for direct H 2 O 2 synthesis. In addition, high-temperature calcination or reduction treatments are known to be crucial to improve the stability of the catalyst. As a starting point, we used this simple catalyst preparation methodology to prepare a 2.5 weight percent (wt%) Pd–2.5 wt % Sn/TiO 2 catalyst as well as its monometallic analogs (8, 10). A synergistic effect toward higher H 2 O 2 productivity was observed when both metals were present after calcination in static air at 500°C for 3 hours (table S1). The measured activity of this 2.5 wt % Pd–2.5 wt % Sn/TiO 2 catalyst (62 mol kg−1 hour−1) is similar to that of a 2.5 wt % Pd–2.5 wt % Au/TiO 2 catalyst (8) (i.e., 64 mol kg−1 hour−1). We then optimized the Sn/Pd ratio while maintaining a total metal content of 5 wt% (table S2) and found an optimum nominal composition, 3 wt % Pd–2 wt % Sn/TiO 2 , that exhibited an H 2 O 2 productivity of 68 mol kg−1 hour−1 (Table 1, entry 1). By comparison, the H 2 O 2 degradation activity of the optimized Sn-Pd catalyst was very low relative to that reported for Au-Pd systems (65 mol kg−1 hour−1 for 3 wt% Pd–2 wt% Sn/TiO 2 versus 235 mol kg−1 hour−1 for 2.5 wt % Pd–2.5 wt % Au/TiO 2 ); this result indicates that Sn is also playing a beneficial role in preventing the overhydrogenation and decomposition of H 2 O 2 . However, the Sn-Pd catalysts calcined at 500°C for 3 hours under static air were unstable to multiple reaction cycles (Table 1, entry 1).

Table 1 Direct H 2 O 2 synthesis and degradation testing. Catalytic testing results of the optimized 3 wt % Pd–2 wt % Sn/TiO 2 , 1 wt % Pd–4 wt % Sn/SiO 2 , and 2.5 wt% Pd–2.5 wt % Au/TiO 2 catalysts after being subjected to various heat treatment regimens. H 2 O 2 productivity was determined under the following reaction conditions: 5% H 2 /CO 2 (2.9 MPa) and 25% O 2 /CO 2 (1.1 MPa), 8.5 g solvent (2.9 g HPLC water, 5.6 g MeOH), 0.01 g catalyst, 2°C, 1200 rpm, 30 min. Second-use values are in parentheses. H 2 O 2 degradation was calculated from the loss of H 2 O 2 using standard reaction conditions: 5% H 2 /CO 2 (2.9 MPa), 8.5 g solvent (5.6 g MeOH, 2.22 g H 2 O, and 0.68 g 50% H 2 O 2 ), 0.01 g catalyst, 2°C, 1200 rpm, 30 min. n.d., not determined. View this table:

The nature of the catalyst surface, in particular the oxidation state of the active metal, is crucial to obtaining high selectivity. Therefore, we characterized this 3 wt % Pd–2 wt % Sn/TiO 2 catalyst after calcination by x-ray photoelectron spectroscopy (XPS). This showed that the majority of the surface Pd was present as Pd2+, whereas the Sn Auger parameter showed that SnO 2 was present in the calcined catalyst (fig. S1 and table S3). Analysis of the nanostructure of this catalyst using high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) and electron energy loss spectroscopy (EELS) revealed that the SnO x was present as a thin amorphous film (thickness <2 nm) coating the TiO 2 support particles (Fig. 1, A to D, and fig. S2). A population of 5- to 10-nm NPs was also present in the sample that contained a homogeneous mixture of both Pd and Sn (Fig. 1, A to D, and fig. S3). Lattice fringe fitting of these particles strongly suggests a metallic Pd-Sn alloy structure rather than segregated metal/metal oxides or mixed oxides, which would mean that only the surface of the particles is oxidized. Many sub–2-nm Pd-rich NPs were also observed and were often associated with the SnO x thin films, and because these species were much less numerous in a 5 wt% Sn-only/TiO 2 sample (fig. S4), they were primarily associated with the PdO x component in the bimetallic catalyst.

Fig. 1 Microstructural analysis of 3 wt % Pd–2 wt % Sn/TiO 2 . (A to D) Representative STEM-HAADF grayscale images and STEM-EELS (RGB) maps of 3 wt % Pd–2 wt % Sn/TiO 2 catalysts at the oxidized stage. (E to H) Images and maps as in (A) to (D) for the same catalyst at the oxidized-reduced-oxidized (O-R-O) stage. From the STEM-HAADF images, three distinct supported species are found in both these catalysts: (i) relatively large (i.e., 3 to 10 nm) nanoparticles (white arrows), (ii) smaller clusters on the nanometer or subnanometer scale (white circles), and (iii) continuous film covering the TiO 2 support surface. The qualitative elemental distribution of Pd, Sn, and Ti is represented by red, green, and blue, respectively, in the STEM-EELS maps. As shown in (C) and (H), the continuous film mainly contains Sn, which either supports or embeds the smaller Pd-rich species; (D) and (G) show that the larger particles are Pd-Sn alloys. Scale bars in the images and maps represent 1 nm unless noted otherwise.

A limiting factor in achieving high selectivity toward H 2 O 2 with Au-Pd/TiO 2 catalysts prepared by the wet impregnation method is that the catalyst nanoparticles exhibit a variation in composition with particle size, with the smallest particles being Pd-rich (9). These small Pd-rich NPs are likely to be highly active for H 2 O 2 synthesis and also for its subsequent hydrogenation and decomposition, as has been shown when Au-Pd catalysts are prepared by colloidal techniques with particle sizes typically 2 to 4 nm (14). In the case of the Sn-Pd system, the small Pd-rich NPs are often associated with the amorphous SnO x films. We postulated that it might be possible to further decrease the H 2 O 2 degradation activity of the catalyst by inducing encapsulation of the ultrasmall Pd-rich NPs by this SnO x film. We therefore used subsequent thermal treatments in an attempt to induce a strong metal-support interaction (SMSI) between the Pd and SnO x layer (15–20). We first added a low-temperature reduction step (200°C, 2 hours, 5% H 2 in Ar), which made the catalysts stable to multiple reaction cycles (Table 1, entry 2). However, the H 2 O 2 degradation activity increased markedly, from 65 mol kg−1 hour−1 to 300 mol kg−1 hour−1; this rate increase was associated with the reduction of Pd2+ to metallic Pd, as shown by XPS (table S3 and fig. S1). Metallic Pd is known to be a more effective H 2 O 2 hydrogenation catalyst (7).

Detailed STEM analysis of this reduced sample was carried out to investigate any structural changes in the sample on reduction (figs. S5 and S6). Analysis identified the presence of thin SnO x films, sub–2-nm Pd NPs, and 5- to 10-nm NPs whose lattice fringe spacings and angles were consistent with Pd-Sn metallic alloy phases. We then reoxidized this reduced catalyst to restore Pd2+ as the predominant surface species (as confirmed by XPS; fig. S1 and table S3), thereby completing an oxidation-reduction-oxidation (O-R-O) cycle, with the aim of encapsulating the small Pd species and regenerating the oxidized Pd-Sn surface on the larger NPs. Reoxidizing the reduced catalyst for various time periods at 400°C under static air had little effect on the H 2 O 2 synthesis productivity but markedly decreased the H 2 O 2 degradation activity (Fig. 2A). After 4 hours of reoxidation treatment, the catalyst showed no activity toward H 2 O 2 degradation and could produce H 2 O 2 with a H 2 selectivity of >95% (Table 1, entry 3). This catalyst was stable to multiple reaction cycles and showed negligible leaching of Sn (0 ppb) or Pd (2 ppb) during a 30-min reaction, as measured by inductively coupled plasma mass spectrometry. This reoxidized 3 wt % Pd–2 wt % Sn/TiO 2 catalyst was subjected to further H 2 O 2 testing, including multiple sequential H 2 O 2 synthesis tests (Fig. 2B). After running the reaction five times consecutively, the H 2 O 2 concentration increased linearly to 0.53 wt %, retaining both the high H 2 O 2 synthesis and zero H 2 O 2 degradation rates. This result implies that no subsequent decomposition or hydrogenation reactions of H 2 O 2 took place with this catalyst. The reoxidized 3 wt % Pd–2 wt % Sn/TiO 2 catalyst was also tested for H 2 O 2 degradation with varying concentrations of H 2 O 2 under a pressure of 5% H 2 /CO 2 (Fig. 2C) and showed no degradation of H 2 O 2 in solutions of up to 8 wt%, whereas the corresponding Au-Pd/TiO 2 catalyst showed substantially higher H 2 O 2 degradation activity at all concentrations studied.

Fig. 2 H 2 O 2 direct synthesis and degradation testing of 3 wt % Pd–2 wt % Sn/TiO 2 . (A) Effect of reoxidation time under static air at 400°C on H 2 O 2 synthesis and H 2 O 2 degradation activity for 3 wt % Pd–2 wt % Sn/TiO 2 catalyst subjected to oxidation-reduction. (B) Sequential H 2 O 2 synthesis reactions over the 3 wt % Pd–2 wt % Sn/TiO 2 material after O-R-O treatment. (C) Degradation activity of optimized O-R-O–treated 3 wt % Pd–2 wt % Sn/TiO 2 catalyst compared to 2.5 wt % Au–2.5 wt % Pd/TiO 2 (8). Experimental conditions are reported in Table 1.

The nanostructure of the catalyst after the O-R-O treatment was characterized to identify any structural changes that could be responsible for the observed high selectivity. Three structures were again revealed to be present in the catalyst: amorphous SnO x films on the TiO 2 particles, small Pd species associated with these films, and larger Pd-Sn NPs (Fig. 1, E to H, and figs. S7 and S8).

Detailed EELS analysis of the thin film regions (fig. S9) after various heat treatments indicated the presence of SnO x films, which can be either reduced or oxidized depending on the final heat treatment stage, as indicated by the absence or presence of the O K edge in the EELS spectrum. In contrast to the thin films, EELS analysis of Sn-Pd nanoparticles (fig. S10) at different stages of the heat treatment cycle showed no clear O K edge. This confirms these nanoparticles to be metallic Pd-Sn alloys, at least in the bulk, which is also consistent with the lattice fringe fitting. Some slight oxidation of the surface of these particles is possible, as detected by our XPS measurements, but at such a level that it is below the detectability limit of the EELS measurements. The 3 wt % Pd–2 wt % Sn/TiO 2 catalyst after the O-R-O cycle shows evidence that the small Pd-rich NPs responsible for high hydrogenation activity often appear to be encapsulated in the amorphous SnO x layer (fig. S11). We suggest that by generating SMSIs that can effectively “bury” the population of smaller Pd-rich NPs, the H 2 O 2 degradation activity of the catalyst is limited (Fig. 3A). The larger uncovered Pd-Sn alloy NPs are then mainly responsible for the H 2 O 2 production and are inherently less prone than Pd-Au NPs to cause subsequent H 2 O 2 degradation. The monometallic Pd/TiO 2 catalyst that was subjected to the optimized O-R-O cycle (table S4, entries 1 and 2) did not induce this effect, confirming that the amorphous layer responsible for SMSI is not TiO 2 -based.

To investigate whether this effect was unique to the 2 wt % Sn–3 wt % Pd/TiO 2 catalyst, we applied similar heat treatment protocols to the analogous 2.5 wt % Au–2.5 wt % Pd/TiO 2 system. The catalytic results (Table 1, entries 4 to 6) show that the same suppression of H 2 O 2 degradation was also not observed for a corresponding O-R-O Au-Pd/TiO 2 catalyst. Results obtained for monometallic Pd and bimetallic Sn-Pd catalysts (table S4, entries 3 to 5) also indicate that the reduction step is crucial, as three oxidative treatments failed to induce the effect. Note that the 200°C reduction temperature used in this study is much lower than that typically reported to induce SMSI effects in TiO 2 -only supported catalyst systems (21). To demonstrate the proposed SMSI effect with SnO 2 , we also prepared a “model” 5 wt % Pd/SnO 2 catalyst and studied it by electron microscopy. In the oxidized state, EELS mapping showed that the PdO x particles appear to be clean (Fig. 3B and figs. S12 and S13), whereas after the O-R-O treatment, the Pd particles showed clear evidence of an SnO x overlayer (Fig. 3C and figs. S14 and S15).

Further evidence that the suppression of the H 2 O 2 degradation does not originate from encapsulation of small Pd particles by TiO 2 but is an SMSI effect arising from the secondary SnO x component was obtained by using a nonreducible SiO 2 support in place of the TiO 2 . When the Sn/Pd ratio on SiO 2 was optimized (table S5) and the same O-R-O heat treatment regimen was applied (Table 1, entries 7 to 9), the resulting 1 wt % Pd–4 wt % Sn/SiO 2 catalyst produced H 2 O 2 at a rate of 50 mol kg−1 hour−1 (Table 1, entry 9) and showed no activity toward subsequent H 2 O 2 degradation. This 1 wt % Pd–4 wt % Sn/SiO 2 catalyst showed no propensity to decompose or hydrogenate H 2 O 2 even in solutions containing up to 12 wt % H 2 O 2 (fig. S16). Furthermore, the catalytic performance of the 1 wt% Pd–4 wt % Sn/SiO 2 material was stable through multiple uses (table S5). XPS analysis (fig. S1 and table S4) and electron microscopy characterization (figs. S17 and S18) of the 1 wt % Pd–4 wt % Sn/SiO 2 catalyst showed compositional and structural features (i.e., amorphous SnO x films and associated ultrasmall Pd-rich NPs, as well as larger Pd-Sn alloy particles) analogous to its 3 wt % Pd–2 wt % Sn/TiO 2 counterpart.

Our approach of encapsulating small Pd-rich species generated by wet impregnation preparations with secondary oxides can be generalized if the second metal oxide added to the Pd/primary oxide (e.g., TiO 2 /SiO 2 ) support system (i) shows no decomposition activity toward H 2 O 2 when in oxide form, (ii) forms an alloy or mixed oxide phase with Pd, and (iii) can encapsulate small Pd-rich particles by SMSI. Following these design rules with the use of a nominal composition of 0.5 wt % Pd–4.5 wt % M/TiO 2 that had been subjected to the optimized O-R-O treatment, we synthesized a series of bimetallic catalysts where M = Ni, Zn, Ga, In, and Co. All of these catalysts showed activity for H 2 O 2 synthesis (30 to 64% of that displayed by the Sn-Pd catalyst) and no activity for H 2 O 2 degradation (table S6). Thus, our approach opens up the possibility of designing reusable catalysts with greatly reduced precious metal content while still retaining high selectivity to H 2 O 2 .

Fig. 3 Evolution of catalyst through oxidation-reduction-oxidation cycle. (A) Proposed mechanism for switching off H 2 O 2 hydrogenation by small Pd-rich NPs through a strong metal-support interaction (SMSI). The secondary metal must both form an alloy with Pd and oxidize to form a secondary support (i.e., SnO x ) that can encapsulate the relatively small, poorly alloyed, Pd-rich NPs after an O-R-O cycle. This step prevents these NPs from decomposing and hydrogenating the H 2 O 2 product. (B and C) STEM-EELS mapping of a 5 wt % Pd/SnO 2 model catalyst at the oxidized (B) and O-R-O (C) stages, showing partial encapsulation of the Pd NP (red) by SnO x (green) after the O-R-O heat treatment cycle. The Sn intensities in the SnO 2 support area were deliberately saturated to reveal any relatively weak signals in the particle region. Scale bars, 1 nm.

Supplementary Materials www.sciencemag.org/content/351/6276/965/suppl/DC1 Materials and Methods Figs. S1 to S18 Tables S1 to S6 References (22–24)

Acknowledgments: Supported by NSF Major Research Instrumentation Program grant MRI/DMR-1040229 (C.J.K.); the U.S. Department of Energy, Office of Science, Basic Energy Sciences, Materials Sciences and Engineering Division (Q.H. and A.Y.B.); a user project supported by Oak Ridge National Laboratory’s Center for Nanophase Materials Sciences, sponsored by the Scientific User Facilities Division, Office of Science, Basic Energy Sciences, U.S. Department of Energy; and European Research Council grant ERC-2011-ADG, grant agreement no. 291319, acronym “AFTERTHEGOLDRUSH” (G.J.H.). The data contained in this paper are archived at DOI: 10.17035/d.2016.0008119507.