1. Introduction
Predicting catalytic activities is of vital importance in catalyst design, considering the large efforts often required for synthesis and testing. The underlying phenomena responsible for catalytic performance is sometimes poorly understood due to limited catalyst characterisation. Microscopic and spectroscopic techniques are very often applied in the characterisation of heterogeneous catalysts, of which a significant number is based on supported metal nanoparticles. High-resolution transmission electron microscopy (HR-TEM), X-ray photoelectron spectroscopy (XPS), and Fourier-transform infrared spectroscopy (FT-IR), among other techniques, help untangling the role of these metal nanoparticles and the impact of their dimensions, surface exposure, oxidation state and interaction with the support on the overall catalytic performance. However, the whole catalytic behaviour is not only determined by the structural properties of the catalytic material, but is rather the result of the complex interplay between catalyst, reactant and reaction medium [
1].
Some techniques allow studying the interaction between a molecule and the catalyst surface, as for example FT-IR spectroscopy of probe molecules [
2] that are able to adsorb on specific sites which can subsequently be detected, identified and quantified, based on the intensity and position of the absorption band. Busca et al. [
3], for example, studied the adsorption of formaldehyde on different oxides (TiO
2, ZrO
2, Al
2O
3) by FT-IR. This is of crucial interest for the understanding of many heterogeneous catalytic reactions, such as CO hydrogenation. However, FT-IR does not allow studying the competition between reactant and solvent in case of liquid-phase reactions. This is however possible by means of Attenuated Total Reflection spectroscopy (ATR-IR). For example, CO adsorption on Pt/Al
2O
3 and Pd/Al
2O
3 catalysts has been investigated with ATR-IR in gas and water, showing that water itself and pH have a large effect on the extent of CO interactions with the surface [
4]. Moreover, operando ATR-IR analyses were performed on Pd/Al
2O
3 catalysts in contact with a solution of benzyl alcohol in cyclohexane, monitoring the restructuring of the catalyst surface and revealing important information about the metal active sites [
5]. However, ATR-IR is limited to specific support and is not applicable to carbons.
A catalytic cycle is generally considered to be composed of different steps, where the adsorption/desorption of the reactant (R) (and the product (P)), as well as diffusion to the active site, are crucial in the observed activity of the catalyst.
Figure 1 reports a simplified scheme:
Too weak adsorption usually leads to poor catalytic activity, while an extensively strong interaction induces an irreversible adsorption of (R) with a consequent deactivation of the active site. Moreover, the adsorption step is dramatically affected by the presence of a solvent and by other reaction conditions such as temperature and pressure. An important point to stress is that this step could be the common first step of different reactions.
Recently, it has been shown that the extent of interaction between the reactant and the catalyst can be evaluated through Nuclear Magnetic Resonance spectroscopy [
6,
7,
8]. In particular, one of the NMR observables, the spin relaxation, is considered a sensitive probe for molecular dynamics. Relaxation times are divided into two types: longitudinal, which concerns change in magnetisation along the
z-axis (T1), and transverse, which concerns a change in magnetisation in the
x-
y plane (T2). Relaxation properties, such as T1 and T2, may be used to identify molecular dynamics processes since they both depend on the rotational correlation time of the reactant [
9]. Indeed, reduced T1 and T2 relaxation times are observed when liquid molecules adsorb on a solid surface due to a decrease in molecular mobility [
10]. The ratio between T1 and T2 has been used in
1H-NMR spectroscopy, as a probe for the molecule/surface interaction [
11,
12], the energy of which is related to the residence time of molecules on the surface [
13]. This methodology has been successfully used to study interactions between liquids and a variety of porous media [
14] and has recently been extended to molecule-surface interactions in supported metal catalysts. D’Agostino et al. [
8] studied the competition between substrate and solvent in the aerobic oxidation of 1,4-butanediol in CH
3OH over Pt/SiO
2, Pd/Al
2O
3 and Ru/SiO
2 catalysts using the T1/T2 ratio. They showed that the catalysts with the lowest activity presented a stronger affinity for CH
3OH than for 1,4-butanediol. It has also been shown that the addition of water even more inhibited the adsorption of the reactant over the catalyst surface, as the T1/T2 ratio of 1,4-butanediol decreased significantly when water was added [
7].
However, as a result of the differences in magnetic susceptibility at the liquid–solid interface, and the limited chemical shift range associated with the
1H nucleus, the line broadening of the spectral resonance that occurs for adsorbed species can prevent the identification of individual
1H resonances. In the present paper, an alternative approach based on
13C NMR is presented showing some practical advantages. The larger chemical shift range of
13C allows to discriminate the individual resonances [
15], thus the surface interaction of each individual carbon atom can be analysed. Additionally, concerning the coupling constant effect, proton decoupled
13C NMR does not suffer from spin coupling due to the very low natural abundance of
13C (1.1%).
13C NMR has been used to investigate oligomers grafted onto silica surfaces [
16], ionic surfactants adsorbed on silica [
17], and within catalysis, for example to study the adsorption of hydrocarbons on zeolites [
18].
Moreover, as reported by Gladden et al. [
15], in
13C NMR the longitudinal relaxation time T1 by its own can be a suitable probe for surface interactions, omitting the use of T2 which is more likely to suffer from resolution issues. The relatively slow molecular motion of the adsorbate on the adsorbent surface causes a decrease in T1 compared to its bulk phase. In this case, the ratio of T1 relaxation times for surface adsorbed (ads) to free diffusing (bulk) molecules (T1
ads/T1
bulk) can be used as an indicator of the relative strength of surface interaction. The lower the T1
ads/T1
bulk, the higher the adsorption strength.
In this study, we focus on Pd nanoparticles supported on a carbon support, commercially named GNP, which is a mesoporous graphitic carbon. This catalyst (1% Pd/GNP) has been used for the hydrogenation of aromatic (benzaldehyde) and aliphatic (n-octanal) aldehydes. Indeed, the catalytic hydrogenation/hydrogenolysis of aldehydes is of high interest from an industrial point of view and particularly relevant in the field of biomass transformation. Indeed, the high oxygen content present in the biomass must be removed to develop biomass-based processes and hydrogenation/hydrogenolysis catalytic processes represent one of the most appealing alternatives. Benzaldehyde and n-octanal can be considered model molecules that can be used to establish the influence of the aldehyde structure on the behaviour of the catalyst.
Alike, carbon-based catalysts are widely used in the industry because of their high stability in all the conditions required for the biomass transformation, and because they can combine hydrophobic (graphitic) and hydrophilic (oxygen functional groups) domains which could allow a better absorption of amphiphilic molecules.
While benzaldehyde could be readily converted at a temperature as low as 50 °C, negligible activity was observed with n-octanal. Here, we proved that by using 13C T1 NMR, this difference in catalytic behaviour can be predominantly attributed to the contrasting adsorption behaviour of both substrates on the catalytic surface which, in turn, is strongly affected by the choice of solvent. In fact, we disclosed that the presence of the solvent is critical in modulating the extent of adsorption, thus affecting catalysis, and our NMR results were able to guide the selection of more suitable reaction conditions.
2. Materials and Methods
2.1. Carbon Supports and Synthesis of Pd-Supported Catalyst
A commercial carbon xGnP® (purchased by XG Sciences Inc., Lansing, MI, USA), with a BET surface area of 490 m2 g−1 and a total pore volume of 0.84 mL g−1 was used as support.
For the synthesis of Pd-supported catalysts, we used Na
2PdCl
4 (≥99.99%, Sigma-Aldrich, St. Louis, MO, USA) as the precursor, NaBH
4 (powder, ≥98.0% Sigma Aldrich) as the reducing agent and Poly(vinyl alcohol) (PVA) (Mw 9000–10,000, 80% hydrolysed, Sigma Aldrich) as the protecting agent. All the reagents were used without further purification. The Pd/C catalyst was prepared by the sol-immobilization technique [
19]. To an aqueous solution of Na
2PdCl
4 (0.5 mmol L
−1) under constant stirring, we added a PVA solution (1% wt.) with a PVA-to-metal weight ratio of 0.5. A freshly prepared aqueous solution of NaBH
4 (NaBH
4-to-metal molar ratio of 8) was added to form a dark brown sol. After 30 min of sol generation, the colloid was immobilized by adding the carbon support (GNP) and acidified at pH 2 by sulfuric acid under continuous vigorous stirring. A metal loading of 1% wt. was targeted. After 1 h of stirring, the slurry was filtered, washed and dried at 80 °C for 2 h.
2.2. Catalytic Reaction and Conditions
The reference catalysts (1% Pd/GNP) were tested in the hydrogenation of both benzaldehyde (ReagentPlus®, ≥99% Sigma-Aldrich) and octanal (for synthesis, Sigma-Aldrich).
The reactions were carried out in a 100 mL stainless steel autoclave. We used 10 mL of a 0.3 M solution of substrate (benzaldehyde or octanal) in the solvent (p-xylene or dodecane, anhydrous, ≥99% Sigma-aldrich) and an amount of catalyst to reach a reactant-to-metal molar ratio of 1000:1. The reaction occurred at 50 °C and 2 bar of H2 pressure. Samplings were carried out by stopping the stirrer and quenching the reaction under cold water. In order to separate the catalyst, 200 µL of reaction mixture were withdrawn and centrifuged. Then, 100 µL of the supernatant solution was diluted with a solution of 1-dodecanol in p-xylene (external standard) for GC measurement.
Product analysis was carried out with a GC-MS (Thermo Scientific, Waltham, MA, USA, ISQ QD equipped with an Agilent VF-5 ms column, Santa Clara, California, USA) and the resulting fragmentation peaks were compared with standards present in the software database. Product quantification was carried out through a GC-FID equipped with a non-polar column (Thermo Scientific, TRACE 1300 equipped with an Agilent HP-5 column).
2.3. NMR Analyses
1,4-Dimethylbenzene-d10 (p-Xylene-d10), dodecane and chloroform-d (CDCl3) has been purchased from Sigma-Aldrich and used as solvents. The NMR tube has been prepared as follows. First, ~15 mg of the catalyst was loaded into the tube. Then, ~700 μL of the benzaldehyde or octanal solution (in p-Xylene-d10 or CDCl3) was added using a micropipette. The concentration of the aldehyde solution was set at a concentration as much as possible similar to that used in the catalytic tests.
NMR analyses were carried out on a Bruker Advance 600 MHz spectrometer at 298 K. For compound characterisation and 13C chemical shift identification, the J-modulated spin-echo sequence was used with 64k points in the time domain and 256 scans.
T1 measurements were obtained using standard inversion recovery from Bruker library (T1 measurements using inversion recovery with power gated decoupling).
One of the assumptions made in the T1 calculations is that the equilibrium is approached exponentially, and therefore, the magnetisation along the
z-axis is represented by Equation (1).
Herein, M
0 is the magnetization at thermal equilibrium, t is the time elapsed, and T1 is the time constant that is obtained by plotting M
z as a function of time. The pulse sequence used in the inversion-recovery experiment is shown below:
In this experiment, the nuclei are first allowed to relax to equilibrium. A 180° pulse (p1) is then applied to invert the signals. The signals are then allowed to relax for a length of time (d2) that is varied in each experiment. After the variable d2 (recovery delay), a 90° pulse (p2) was applied, and the FID was recorded. The FID records the spectrum intensity as a function of the variable delay d2. The signal will have relaxed more with longer d2. The peak intensity will reflect the extent to which each signal has relaxed during the d2 period.
14 T1 recovery delays were used ranging from 0.05 ms to 100 s. We used 8 repeat scans and a relaxation delay (d1) of 50 s between each scan to ensure a maximum signal (and to ensure the reach of the equilibrium, the d1 delay in the pulse sequence should be set to ~5 × the longest T1 of interest in the molecule) was maintained at all times.
The analysis of the T1 measurements was performed with the standard Bruker routine for T1/T2 calculation and with the Bruker Dynamic Centre software version 2.5.6, using the following fitted function:
where I
0 is the equilibrium magnetization and the parameter a determine the magnetization at time zero, that thus corresponds to I
0(1 − a).
The errors associated with fitting the bulk and adsorbed T1 were all within ±1%.
4. Conclusions
The different activity of 1% Pd/GNP catalyst towards aromatic and aliphatic aldehydes hydrogenation has been explored by 13C NMR relaxation.
By studying the relaxation time ratio T1ads/T1bulk, we can correlate the different catalytic behaviour of benzaldehyde and n-octanal with their adsorption strength on the catalyst.
The choice of the solvent is crucial in the adsorption of the reactant, which is a pre-requisite for catalytic activity. When p-xylene was used as the solvent, benzaldehyde can be adsorbed on the catalyst surface and subsequently converted, whereas octanal resulted inert due to p-xylene governing the adsorption step. Further NMR analysis demonstrated that octanal can adsorb and react on Pd on GNP catalyst when a non-aromatic solvent is used. Therefore, as final proof of concept, a catalytic test changing the solvent from p-xylene to dodecane has been performed showing that 1% Pd/GNP was able to hydrogenate 40% of octanal in 2 h. It was also disclosed that the selectivity of the reaction can be affected also by the solvent: when dodecane was used as the solvent, n-octanal was converted to dioctyl-ether, whereas benzaldehyde showed a faster production of toluene.
At present, T1ads/T1bulk measurements do not allow to differentiate physical vs. chemical adsorption but were able to clearly indicate the role of the Pd active site in the adsorption of the substrate on the catalyst surface, which in turn should be considered a pre-requisite for having catalytic activity. Further studies are ongoing in order to fine-tune the technique in an attempt to disentangle chemical from physical adsorption.