Influence of Metal Deposition and Activation Method on the Structure and Performance of Carbon Nanotube Supported Palladium Catalysts

Abstract

The effects of the metal deposition and activation methods on metal particle size and distribution were investigated for carbon nanotube supported Pd catalysts. The Pd precursor was loaded by incipient wetness impregnation, ion adsorption, and deposition precipitation and was activated by thermal treatment under a nitrogen atmosphere or in the liquid phase by reduction by formaldehyde or sodium borohydride. Regardless of the metal precursor loading method, activation under a N₂ atmosphere at 500 °C led to homogeneously distributed 4 nm Pd particles. Liquid‐phase reduction by sodium borohydride provided a bimodal distribution with particle sizes of approximately 1 and >10 nm. A somewhat weaker reducing agent, formaldehyde, yielded particles approximately 1 nm in size. The activities of the catalysts for the hydrogenation of cinnamaldehyde correlated with the particle sizes.

Carbon supported palladium (Pd/C) catalysts are widely used for hydrogenation in the fine‐chemicals industry. Numerous studies have been performed investigating the influence of the support material,1, 2, 3, 4, 5 the surface functional groups of the support,6, 7 metal precursor,2, 8 and activation method2, 9, 10, 11 on the structure and activity of the final Pd/C catalyst. Ideally, the metal particles should have a narrow size distribution and be uniformly distributed across the support to arrive at increased activity and stability. However, particles are often distributed nonuniformly or the particle‐size distributions are broad or multimodal.12, 13, 14

Carbon supports that are often used include activated carbon, carbon black, graphite, and carbon nanomaterials such as carbon nanofibers and carbon nanotubes (CNTs). In research, CNTs are of particular interest, as they have a well‐defined structure with a rather uniform surface, high purity, and no micropores. The structure and porosity are largely maintained even if they are exposed to aggressive media.15, 16 The well‐defined structure of CNTs makes them suitable for analysis by transmission electron microscopy, with high contrast between the supported metal nanoparticles and the support. Although CNTs are relatively inert, functional groups can be introduced onto the surface by functionalization treatment such as liquid‐phase oxidation (LPO) and gas‐phase oxidation (GPO). Surface chemistry can have a strong influence on metal dispersion6, 17 and interactions between the metal and the support7, 18 and, thus, on the performance of Pd/C catalysts.7, 17, 19

Various metal deposition methods such as impregnation,2 deposition precipitation,3, 6 deposition reduction,20 ion adsorption,21, 22 and colloidal routes7, 23 have been studied. The deposited metal precursor is commonly reduced in the gas phase, and this typically yields well‐distributed Pd nanoparticles.24, 25 However, as Pd/C catalysts are often used in the liquid phase,26 performing activation in the liquid phase is highly desirable.

In this work, we systematically investigated the influence of the different metal deposition methods and subsequent activation methods with varying conditions thereof on the structure of Pd/CNT catalysts. The (NH₃)₄Pd(NO₃)₂ precursor was deposited from aqueous solution onto LPO‐treated CNTs by incipient wetness impregnation (IWI), deposition precipitation (DP), or ion adsorption (IA) with a nominal Pd loading of 5.0–6.0 wt %. The deposited Pd precursor was converted into metallic Pd particles in the gas phase by thermal treatment under a N₂ atmosphere at 500 °C.19 Catalyst precursors prepared by ion adsorption were also reduced in the liquid phase by sodium borohydride or formaldehyde, both known to act as reducing agents in the preparation of Pd/C catalysts,9, 10, 24, 27 with varying conditions such as rate of addition of the reducing agent, temperature, and atmosphere under which the addition is performed. Details on the deposition and reduction methods can be found in the Supporting Information.

As expected, ion adsorption followed by gas‐phase activation under a N₂ atmosphere yielded uniformly distributed Pd particles with a unimodal particle‐size distribution and an average size of approximately 5 nm (Figure 1 a). Comparable structures of Pd/C catalysts were observed if the deposition method was IWI or DP (Figure S1 in the Supporting Information). Ion adsorption was chosen to investigate further and to compare liquid‐phase to gas‐phase activation. In contrast to N₂ activation, reduction by sodium borohydride led to a bimodal particle‐size distribution with 1 nm Pd particles uniformly distributed across the CNT support and 50 nm Pd particles probably located at the external surface of the CNT skeins (Figure 1 b).28 Similarly, reduction by formaldehyde led to 1 nm Pd particles uniformly distributed over the support, but with only a negligible amount of the Pd ending up in 10–50 nm particles (Figure 1 c).

Figure 1.

Representative high‐angle annular dark‐field scanning transmission electron microscopy (HAADF‐STEM) images and particle‐size distribution histograms of the Pd/CNT catalysts activated by a) thermal treatment at 500 °C under a N₂ atmosphere, b) slow addition of NaBH₄ at room temperature in air, c) instant addition of formaldehyde at room temperature under a N₂ atmosphere. Only sub‐10 nm particles are included in the particle‐size distribution of the NaBH₄‐reduced sample.

X‐ray diffraction (XRD) and H₂ chemisorption were used to determine and compare the average Pd particle size from the bulk of a sample with that obtained by transmission electron microscopy (TEM) analysis (Table 1). Details on the characterization methods can be found in the Supporting Information. The ion‐adsorption sample activated by N₂ treatment shows that the particle sizes determined by XRD, H₂ chemisorption, and TEM are similar. For the formaldehyde‐reduced sample, the absence of Pd signals in the X‐ray diffractograms and very low H₂ chemisorption‐derived Pd particle sizes further corroborate that the 10–50 nm sized Pd particles observed by TEM are rather negligible. For the NaBH₄‐reduced sample, the heterogeneity was quantified by calculating the contributions of the 1 and 50 nm particles to the average size determined by H₂ chemisorption. It was calculated that approximately 50 % of the Pd ended up as 1 nm particles and approximately 50 % of the Pd ended up as 50 nm particles.

Table 1.

Weight loading, average crystallite sizes from XRD, and particle sizes from H_₂ chemisorption and TEM.

Activation method	Pd loading [wt %]	Pd particle/crystallite size [nm]
		XRD	H2	TEM
N2 treatment	4.0	2.6	3.6	3.2±1.4
NaBH4	4.4	2.9	2.0	1/50[a]
formaldehyde	3.7	–	0.8	1.1±0.3

[a] Bimodal distribution.

One possible cause for the larger particles in the NaBH₄‐reduced sample might be a higher loading of Pd in the sample. However, inductively coupled plasma (ICP) analysis confirmed that in all three ion‐adsorption samples, as well as in all other samples prepared in this study, the achieved Pd loading was rather similar ranging from 3.7 to 4.4 wt % (see Table 1 and Table S1). Comparing the nominal loading of 5.0–6.0 wt % with the achieved loading of 3.7–4.5 wt % in the ion‐adsorption samples, the extent of Pd deposition was approximately 70 %, which is slightly lower than the amounts of Pd deposited by IWI (85 %) and DP (95 %).

To investigate further how liquid‐phase reduction might have affected the distribution of the metal nanoparticles over the support, experiments were performed under different atmospheres (air or N₂), different temperatures at which the reducing agent was added (RT and 2 and 80 °C), and different addition rates (fast or slow). Relative to the samples shown in Figure 1 b, c, none of the listed conditions seemed to impact the final catalyst structure significantly (Figures S2 and S3). The amount of NaBH₄ used also had no significant influence on the catalyst structure, as observed by TEM (Figure S2 f, g). Analysis of the metal loadings and Pd particle sizes determined by XRD, H₂ chemisorption, and TEM for all samples can be found in Table S1. These results exclude effects of concentration gradients as well as Ostwald ripening resulting from oxidationreduction cycles.

An important difference between the NaBH₄‐reduced and formaldehyde‐reduced catalysts lies in the nature of the reducing agent. The standard reduction potentials (at pH 0) of Pd²⁺/Pd, formic acid/formaldehyde, and boric acid/borohydride are 0.9, −0.25, and −0.5 V, respectively.29, 30, 31 The more negative boric acid/borohydride potential may explain the observed differences by which the Pd particles grow to even 50 nm. This could be a result of immediate reduction of the metal precursor from the solution onto the support, whereas milder formaldehyde may reduce only precursor species adsorbed on the support.

All three catalysts were active in the hydrogenation of cinnamaldehyde (Table 2 and Figure S4). Normalized to the amount of Pd, the formaldehyde‐reduced catalyst was much more active than the N₂‐activated and NaBH₄‐reduced catalysts. This difference could be explained by the particle sizes of the spent catalysts. The average particle sizes, determined by TEM, were (3.8±1.4), (2.3±1.3), and (2.0±1.0) nm for the spent N₂‐activated, NaBH₄‐reduced (sub‐10 nm particles only), and formaldehyde‐reduced catalysts, respectively (Figure S6). The turnover frequencies calculated by using these sizes were similar.

Table 2.

Catalytic activity and selectivity towards hydrocinnamyl aldehyde.

Activation method	Activity[a] [mmol gPd −1 s−1]	TOF[b] [s−1]	Selectivity [%]
N2 treatment	2.0	0.7	24
NaBH4	2.4	1.0	63
formaldehyde	3.5	0.7	75

[a] Activity at 50 % conversion, determined from a linear fit of ln(x _{cinnamaldehyde}) versus t (see Figure S5). [b] Turnover frequency calculated as the number of cinnamaldehyde conversions per second, normalized to the number of Pd surface atoms calculated from the average Pd particle size of the spent catalysts, determined by TEM.

The selectivity to hydrocinnamaldehyde was 75 % for the formaldehyde‐reduced catalyst and 62 % for the NaBH₄‐reduced catalyst; the remaining amounts were hydrocinnamyl alcohol. A low selectivity of 24 % was found for the N₂‐activated catalyst. No cinnamyl alcohol was formed, as expected,32 and no other products were observed in the GC traces. The difference in selectivity could be caused by the larger Pd particles in the NaBH₄‐reduced sample.33 Alternatively, the selectivity difference could be explained by support effects,34 which could be introduced by a change in the nature of the functional groups caused by the activation method.

Preparation of Pd/CNT by using (NH₃)₄Pd(NO₃)₂ as a metal precursor and activation under a N₂ atmosphere yielded catalysts with particle sizes of approximately 4–5 nm distributed uniformly across the support regardless of the metal deposition method. Pd/CNT preparation by ion adsorption followed by reduction in the liquid phase by NaBH₄ yielded catalysts with a nonuniform metal‐particle distribution and a bimodal size distribution, whereas reduction by formaldehyde yielded a catalyst with uniformly distributed metal particles approximately 1 nm in size. Various liquid‐phase reduction conditions appeared to have no influence on the final catalyst. It was the nature and probably the strength of the reducing agent that determined the final structure of the catalyst, with weaker reducing agents promoting narrower particle‐size distributions and a uniform particle distribution across the support.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

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

W. S. Lamme, J. Zečević, K. P. de Jong, ChemCatChem 2018, 10, 1552.