Nickel Phosphide Catalysts as Efficient Systems for CO2 Upgrading via Dry Reforming of Methane

This work establishes the primordial role played by the support’s nature when aimed at the constitution of Ni2P active phases for supported catalysts. Thus, carbon dioxide reforming of methane was studied over three novel Ni2P catalysts supported on Al2O3, CeO2 and SiO2-Al2O3 oxides. The catalytic performance, shown by the catalysts’ series, decreased according to the sequence: Ni2P/Al2O3 > Ni2P/CeO2 > Ni2P/SiO2-Al2O3. The depleted CO2 conversion rates discerned for the Ni2P/SiO2-Al2O3 sample were associated to the high sintering rates, large amounts of coke deposits and lower fractions of Ni2P constituted in the catalyst surface. The strong deactivation issues found for the Ni2P/CeO2 catalyst, which also exhibited small amounts of Ni2P species, were majorly associated to Ni oxidation issues. Along with lower surface areas, oxidation reactions might also affect the catalytic behaviour exhibited by the Ni2P/CeO2 sample. With the highest conversion rate and optimal stabilities, the excellent performance depicted by the Ni2P/Al2O3 catalyst was mostly related to the noticeable larger fractions of Ni2P species established.


Introduction
The reduction of atmospheric carbon dioxide, and thus the greenhouse effect, is one of the defining scientific and engineering challenges of our time. Progress is being made in reducing carbon dioxide emissions, and technologies such as carbon capture and storage are available [1]. However, the sequestering of carbon dioxide in oceans or mineral formations is expensive and provides no useful end product. One technology with great potential in this field is the dry reforming of methane (DRM, CO 2 + CH 4 = 2CO + 2H 2 ) to produce a mixture of carbon monoxide and hydrogen-syngas. Dry reforming was first studied by Fischer and Tropsch as early as 1928 [2], and today attracts ever increasing interest due to the rising importance of CO 2 mitigation. The advantages of the dry reforming of methane are two-fold. Firstly, methane is also a potent greenhouse gas, so the reduction of this pollutant in the atmosphere is highly desirable. Secondly, the product syngas, as well as being a viable fuel for internal combustion [3], is an important feedstock for the production of higher hydrocarbon fuels through the Fischer Tropsch process [4] and for methanol production [5].
The DRM reaction is endothermic, requiring low pressure and high temperatures, above 600 • C according to thermodynamics, to progress at a good rate [6]. A significant amount of research has focused on noble metal catalysts. These catalysts often present far superior resistance to deactivation compared to transition metals. Still, the high cost of noble metals has driven research towards cost-effective formulations, usually based on transition metals such as Cu, Fe and Ni [2]. For traditional supported catalysts, deactivation issues related to carbon deposits, metal sintering and the oxidation of active species have been reported [7]. Compared to Fe and Co systems, relatively higher coking endurances coupled to fair reaction rates have been described for Ni catalysts. Over the past few decades, a substantial body of research has focused on reducing the deactivation of nickel catalysts by employing different metal oxide supports and metal promoters [1,2,6]. On developing effective catalysts, the rapid surface decomposition of methane, coupled with low tendencies towards constituting carbon deposits, potentially blocking the active sites, are generally intended features. On this premise, competitive catalytic systems have been reported. For instance, the Ni-Sn/CeO 2 -Al 2 O 3 catalyst combined the higher reaction rates depicted by Ni-Sn alloys with the benefits provided by oxygen vacancies of cerium dioxide support [8]. It is known that the support strongly determines the performance displayed by dry reforming catalysts, the metal support interfaces being described as powerful actives sites. Depositing catalytic material on a porous support also increases the metal exposed area whilst the acid-base support promotes CH 4 /CO 2 activation, thus favouring the reaction rate [2]. For DRM reactions, it is known that relatively acidic supports such as SiO 2 favour the cracking of methane, which is usually considered as a rate limiting step. On the contrary, basic surfaces promote the activation of CO 2 molecules. In fact, oxygen vacancies present in partially reduced oxides have also been stated as effective sites towards CO 2 dissociative reductions [9]. Nevertheless, the acid base properties could also determine the extent of carbon deposits constituted over the catalyst surface [6].
When seeking novel catalyst formulations, transition metal carbides and phosphides have been proposed as appealing catalytic materials for a number of catalytic reactions, such as WGS and CO hydrogenation [10,11]. Among them, Mo 2 C catalysts are suggested as appealing alternatives, even being described as Pt-like materials [12,13]. In this context, Yao et al. [14] proposed molybdenum phosphide as a highly active and stable catalyst for dry reforming. Likewise, Guharoy et al. [15] proposed, via Density Functional Theory (DFT) calculations, nickel phosphide structures as active sites for Reverse Water Gas Shift (RWGS) reaction on the basis of the potential surface energies estimated for different reaction intermediates. Although scarcely analysed for CO 2 reduction processes, nickel phosphide clusters have proven successful as a catalyst for the hydrotreating of species such as thiophene [11]. For Ni 2 P surfaces, the strong oxygen interaction has been also related to promoted Water Gas Shift reaction rates [16] and hindered deactivation by coking in dry reforming applications [17]. Furthermore, the low barrier energy towards H 2 dissociation, along with the thermodynamic stability of H species over Ni 2 P surfaces under hydrogen rich environments, suggest Ni 2 P as an active phase for DRM reactions [18].
Considering the promising prospect envisaged for Ni 2 P active sites, this work aims to establish the influence of the support on Ni 2 P catalysts for DRM reactions. For that purpose, and in order to cover a wide range of chemical properties, 20% Ni 2 P was impregnated over Al 2 O 3 , SiO 2 -Al 2 O 3 and CeO 2 supports. From the obtained characterisation and activity outcomes, Ni 2 P/Al 2 O 3 catalysts could lead to a viable industrial CO 2 recycling process, with an impact on the pressing problem of rising atmospheric CO 2 levels.

Structural Characterization of the Samples
The textural properties of the as-prepared mesoporous samples were evaluated by N 2 -physisorption. Figure 1 displays the N 2 adsorption desorption hysteresis exhibited by the catalysts' series. According to IUPAC (International Union of Pure and Applied Chemistry) classification, the Ni 2 P/Al 2 O 3 and Ni 2 P/SiO 2 -Al 2 O 3 samples presented a type H4 hysteresis shape characteristic of mesoporous materials. Considerably different hysteresis was noticed for the Ni 2 P/CeO 2 system, which presented type H3 isotherm shapes usually associated to aggregated particles. Table 1 summarises the results of the BET and BJH analysis. Compared to Ni 2 P/Al 2 O 3 and Ni 2 P/SiO 2 -Al 2 O 3 catalysts, significantly lower surface areas were discerned for the Ni/CeO 2 catalyst. Besides, the metal incorporation resulted in decreased surface areas and pore volumes for all the catalysts' series. The lower surface and pore volumes observed with respect to the supports are most likely associated with the presence of non-porous Ni particles which, at the same time, might be partially blocking the surface pores [19].   Figure 2 displays the diffractograms obtained for the as-prepared samples where γ-Al 2 O 3 support (JCPDS No. 29-0063) is also included for clarity. For the Ni 2 P/Al 2 O 3 and Ni 2 P/SiO 2 -Al 2 O 3 samples, no diffraction peaks associated to AlPO 4 or NiAl 2 O 4 phases were observed. The constitution of NiAl 2 O 4 phases could be restricted due to the relatively lower calcination temperatures (500 • C). Concerning the Ni 2 P/SiO 2 -Al 2 O 3 sample, the relatively lower signal to noise ratio showed should be ascribed to the amorphous structure of silica present in the support (20% SiO 2 ) [20]. The characteristic peaks of fluorite CeO 2 (JCPDS No. 34-0394) were clearly observed for the Ni 2 P/CeO 2 system. Besides, no clear diffraction peaks associated to Ni phases were observed for any of the samples. Omar et al. [21] argued that nickel phosphate calcined up to 600 • C results in amorphous structures, whilst increasing the calcination temperature up to 900 • C permits the constitution of highly crystalline Ni 2 P phases. Thus, the lack of Ni 2 P peaks in XRD data should relate to the amorphous character of nickel phosphate deposits because of the employed calcination temperature.  Figure 3A shows the Ni 2p 3/2 spectra obtained for reduced samples where Ni δ+ and Ni 2+ species along with their corresponding satellites are identified. For the Ni 2 P/Al 2 O 3 and Ni 2 P/SiO 2 -Al 2 O 3 samples, the band contribution located at 852.7 eV was associated to Ni δ+ species. Slightly lower binding energies (852.5 eV) were observed for Ni δ+ species in the case of the Ni/CeO 2 catalyst. For Ni-supported ceria systems, peak shifts towards lower binding energies are usually related to the higher electron densities attained by Ni species in contact with partially reduced ceria support [22]. Besides, all samples exhibited peaks attributed to oxidised nickel species. Still, the support nature significantly influenced the electronic structure displayed by surface Ni 2+ species. Typically, bands located at 854.7 and 856 eV are assigned to NiO and Ni(OH) 2 species, respectively. For the reduced Ni 2 P/Al 2 O 3 and Ni 2 P/SiO 2 -Al 2 O 3 samples, XPS data showed a single Ni 2+ band located at 854.7 and 855.0 eV, correspondingly. The differences observed for the Ni 2+ binding energies between the Ni 2 P/Al 2 O 3 and Ni 2 P/SiO 2 -Al 2 O 3 samples underline the higher electronic density achieved for the Ni 2 P/Al 2 O 3 sample. Instead, two different peaks located at 854.7 and 856.8 eV were observed for Ni/CeO 2 catalyst, underlining two different Ni 2+ species interacting with partially reduced cerium dioxide.

Surface Characterization of the Samples: XPS Analysis
In the P2p region ( Figure 3B), bands associated to slightly negatively charged phosphorous species (P δ− ) were discerned for all samples. Considering the Ni δ+ species noticed in the Ni 2p 3/2 spectra, the presence of Ni 2 P surface clusters with the consequent electrons transfer from Ni to P species could be advocated [23]. Once again, the support nature influenced the electron properties exhibited by P δ− surface species. Thus, while the Ni 2 P/Al 2 O 3 sample exhibited bands located at 129.2 and 130.4 eV, associated to P δ− species, the larger binding energies noticed for P δ− species in the Ni 2 P/SiO 2 -Al 2 O 3 sample (129.4 and 130.6 eV) suggest that P δ− species presented lower electron densities. This suggests that stronger Ni-P interactions and enhanced electron donations from Ni to P were achieved for the Ni 2 P/Al 2 O 3 sample. For the Ni 2 P/CeO 2 system, P δ− species appeared (129.5 and 133.0 eV) due to, in agreement with previous outcomes, efficient charge redistribution processes at the Ni 2 P-CeO 2 interface [24].
Furthermore, the bands showing higher binding energies for all samples are usually ascribed to highly oxidized P species. In this sense, the bands located at 134.6 and 135.0 eV noticed for the Ni 2 P/Al 2 O 3 and Ni 2 P/SiO 2 -Al 2 O 3 samples were ascribed to P 5+ species. For the Ni 2 P/CeO 2 samples, P 5+ species appeared at significantly lower binding energies. For the latter systems, bands placed at 135.8 and 136.4 eV were ascribed to the constitution of AlPO 4 [25].
Therefore, all reduced catalysts displayed both Ni 2+ and Ni slightly oxidized (Ni δ+ ) species constituting Ni 2 P surface clusters, in agreement with the Ni 2p 3/2 and P 2p spectra. The larger electron density displayed by P δ− for the Ni 2 P/Al 2 O 3 catalyst suggested strongly interacting nickel phosphide structures. Besides, Ni particles interacting with cerium dioxide result in the evolution of: (i) rich-electron density Ni δ+ species and (ii) two different Ni 2+ species with relatively low electronic densities. The dispersion of the catalyst material on the supports was approximated by the atomic ratio of nickel species to support material on the catalysts' surfaces ( Table 2). This calculation revealed that the ceria supported catalyst had the greatest proportion of its surface occupied by nickel species. Although strong Ni-ceria interactions and high metal dispersions are usually reported, this result can also be significantly influenced by the lower specific surface area of the ceria support. When comparing the alumina and silica-alumina supported catalysts, the higher Ni surface coverage found for the Ni 2 P/SiO 2 -Al 2 O 3 sample suggests that the SiO 2 -Al 2 O 3 dispersing matrix favours, at least a priori, the Ni metal dispersion. For Ni supported catalysts, the optimal Ni loading over the Al 2 O 3 and ZrO 2 dispersing matrix was found at 20 and 30%, respectively [26]. In agreement, the obtained outcomes suggest that silica-based supports lead to greater dispersion of the nickel species over the surface. From XPS data, the Ni/Ni δ+ surface ratio displayed by the catalyst series was also estimated. Qualitatively, the higher Ni δ+ /Ni 2+ ratios found for the Ni 2 P/SiO 2 -Al 2 O 3 sample might be associated to the higher dispersions. Besides, the significantly higher P δ− /[P 5+ +P-O] ratios found over the Ni 2 P/Al 2 O 3 sample corroborated the favourable constitutions of Ni 2 P phases already intuited from the higher P δ− electron densities noticed in the P 2p spectra.  Figure 4 shows the catalyst performance exhibited by the samples at 30 L/g·h and 700 • C. For all reaction tests, the CH 4 conversion was lower than the CO 2 conversion. This is congruent with the current literature on the kinetics and equilibrium of dry reforming of methane where the activation of CH 4 on the catalysts' surface is often cited as the rate limiting step for DRM [27]. Hence, the equilibrium conversion of CO 2 should be ca. 10% higher than for CH 4 [2]. Initially, all systems achieved close to 1 H 2 /CO ratios and, in the case of the Ni 2 P/CeO 2 sample, decreased proportionally to the conversion rate. The reaction rate observed for the catalysts' series decreased according to the sequence: Ni 2 P/Al 2 O 3 > Ni 2 P/CeO 2 > Ni 2 P/SiO 2 -Al 2 O 3 . Thus, the highest CO 2 conversion, combined with good catalyst stabilities, revealed the Ni 2 P/Al 2 O 3 system as the best-performing catalyst. On the contrary, the worst catalyst behaviour of the series was exhibited by the Ni 2 P/SiO 2 _Al 2 O 3 catalyst. The lower conversion rates might be associated to important cooking processes over such acidic surfaces. Rather unexpected poor catalytic behaviours were displayed by Ni 2 P/CeO 2 catalysts. Thus, even though relatively optimal catalyst performances were initially depicted, the Ni 2 P/CeO 2 catalyst depicted an important loss of catalytic activity, with the CO 2 conversion dropping from around 80 to 50% and the CH 4 conversion dropping from around 70 to 35% during the catalytic test. For this sample, the significantly lower catalyst surface area could be related to the deprived catalyst performance.

XRD of the Spent Samples
Considering the importance of deactivation issues related to sintering or coking processes described for DRM catalysts [27], a glimpse into the catalyst structural variations attained over the different samples under reaction conditions was attempted through XRD analysis. Figure 5 shows the diffractograms obtained for the post-reacted samples. For Ni/CeO 2 catalysts, no diffraction lines associated to Ni phases were discerned. In agreement with XPS data, ceria support seems, indeed, to favour higher Ni metal dispersions [6]. For Ni/Al 2 O 3 and Ni 2 P/SiO 2 -Al 2 O 3 catalysts, diffraction lines associated to the Ni 0 phase were clearly evidenced. Still, despite the good dispersions observed for the reduced sample, the spent Ni 2 P/SiO 2 -Al 2 O 3 sample showed prominent Ni 0 diffraction peaks, indicating that significantly larger Ni crystal sizes were constituted by the Ni 2 P/SiO 2 -Al 2 O 3 catalyst under reactions conditions. The higher Ni particles observed for the SiO 2 -Al 2 O 3 catalyst suggest that, despite the relatively high metal dispersions obtained with silica-based catalysts, poor resistances against sintering issues are presented. Moreover, the Ni/Al 2 O 3 sample exhibited clear diffraction peaks at 41.5 • , ascribed to Ni 2 P phases (JCPDS No. 01-074-1385), whilst a small Ni 2 P contribution could also be intuited for the Ni 2 P/SiO 2 -Al 2 O 3 catalyst sample. The larger concentration of Ni 2 P species, constituted for the Ni/Al 2 O 3 catalyst, agrees with the findings provided by XPS data. Comparatively, significantly larger amounts of C deposits were constituted over the silica supported catalyst. Given the higher surface acidity of silica-doped alumina [27], promoted deactivation processes related to coking phenomena are, indeed, expected [28,29]. In any case, carbon-forming reactions are, indeed, thermodynamically favoured for large nickel crystallites [30,31]. Therefore, the larger Ni particle sizes, combined with the intrinsic acidity of silica surfaces, account for the significantly higher amounts of carbon deposits constituted and for the observed conversion drop during the catalytic test. On the contrary, the absence of C related peaks for the Ni/CeO 2 catalyst suggest the active role of ceria support against the constitution of carbon deposits.
Overall, the behaviour observed for the catalysts' series evidences Al 2 O 3 support as the right choice for Ni 2 P catalysts in DRM reactions. For CO 2 methanation, higher concentrations of Ni 2 P were also related to improved reaction rates over Al 2 O 3 supported systems [24]. Along with promoted CO 2 activations, the perturbed electronic structure achieved within Ni 2 P has been proposed to increase the resistance against deactivation effects related to Ni oxidation [32] and carbon deposits [33]. Thus, the optimal activity and stability depicted by the Ni 2 P/Al 2 O 3 catalyst could be associated to the significant highest fractions of Ni 2 P species present on the catalyst surface, coupled with fair and relatively stable Ni dispersions. In a similar manner, the lower catalytic behaviour observed for the Ni 2 P/SiO 2 -Al 2 O 3 system should be associated with larger Ni agglomerates along with the lower Ni 2 P concentrations observed over the catalyst surface. Besides, the larger coke deposits found on the Ni 2 P/SiO 2 -Al 2 O 3 catalyst surface might also deplete its catalyst performance and propensity towards the development of carbon deposits due to the large Ni agglomerates dispersed over highly acidic surfaces. Moreover, the lower DRM performance observed for the Ni/CeO 2 catalyst might be strongly affected by the lower surface areas depicted by this sample. For the ceria supported sample, the strong interaction established between Ni and CeO 2 supports might hinder the effective constitution of Ni 2 P phosphide species. In any case, Ni-O-Ce sites are suggested to be active for the total oxidation of methane [34], with the evolution of CO 2 molecules formed through the reaction between hydroxyls groups constituted over Ce 3+ sites and surface methyl groups [35]. The relatively lower catalytic performances displayed by the Ni/CeO 2 sample could also be associated with favoured methane oxidation reactions. The absence of carbon deposits and Ni sintering effects evidenced in the XRD analysis of the post-reacted samples suggest the constitution of Ni 2+ species, inactive for CO 2 activation, as the main deactivation cause.

Catalyst Synthesis
All three catalysts were prepared by impregnation of the relevant support with a nickel phosphate solution. For that purpose, γ-Al 2 O 3 and 20% SiO 2 -80% Al 2 O 3 (Puralox SCFa-230 and Siralox 20/380, Sasol, Johannesburg, South Africa) commercial supports were employed. A third home-made ceria support was also employed. Then, CeO 2 support was synthesised by the reaction of urea and cerium nitrate. Thus, adequate amounts were dissolved in 100 mL of deionised water. After being stirred at 85 • C for 24 h, the solution was then filtered to collect the precipitate, which was dried overnight at 80 • C. The obtained solid was calcined in air at 400 • C for 4 h to produce a CeO 2 powder.
The nickel phosphide catalysts were prepared by a procedure similar to that seen in the literature [36,37]. Thus, the intended nominal contents were 20%Ni and had a Nickel:Phosphorous molar ratio of 3:2. For that purpose, a nickel phosphate solution was prepared by mixing solutions of Ni(NO 3 ).6H 2 O and (NH 4 ) 2 HPO 4 in deionised water, producing a nickel phosphate precipitate. The nickel phosphate solution was impregnated over the chosen supports and the excess of solvent was removed by rotary vacuum evaporation. The obtained samples were in an oven at 80 • C and calcined in air at 500 • C for 4 h. Over Puralox SCFa-230, Siralox 20/380 and CeO 2 supports, the obtained catalysts were labelled Ni 2 P/Al 2 O 3 , Ni 2 P/SiO 2 -Al 2 O 3 and Ni 2 P/CeO 2 , correspondingly. In agreement with previous works [36,37], Ni 2 P structures were constituted after subsequent reductions in H 2 presence.

Catalyst Characterisation
The textural properties of the samples were analysed using N 2 adsorption at −196 • C on an Autosorb iQ Station 2 instrument (Anton Paar QuantaTec Inc., Boynton Beach, FL, USA). The weight-specific surface area of the catalysts was determined using the Brunauer-Emmett-Teller equation [11]; the average pore size and weight-specific pore volume were calculated using the Barett-Joyner-Halenda method [22]. X-ray diffraction analysis was carried out on Panalytical X'Pert Pro Powder equipment (Marvern Panalytical) using Cu Kα radiation at 30 mA and 40 kV. X-ray photoelectron spectra were collected in order to determine the composition of the catalysts' surfaces using a K-Alpha Thermo Scientific instrument. The XPS analysis was carried out on freshly reduced catalysts. The spectra were recorded using Al-K radiation at 1486.6 eV, which was monochromatised by a twin crystal monochromator to produce a focused elliptical X-ray spot with a major axis length of 400 µm. The initial survey was carried out from a binding energy of 1350 eV to 0 eV, with higher resolution detailed scans being taken of the most relevant sections.

Catalyst Reaction Testing
The catalysts were tested at atmospheric pressure in a 10 mm internal diameter quartz tube reactor. In these tests, the reactor was loaded with 0.1 g of catalyst and fed with 100 mL/min of 50 vol.% N 2 , 25 vol.% CO 2 and 25 vol.% CH 4 ; neglecting the inert nitrogen, the Weight Hourly Space Velocity (WHSV) was 30 L·g − 1·h −1 . Prior to reaction testing, the catalysts were reduced in a 10 vol.% hydrogen stream with nitrogen as an inert carrier at 800 • C for 1 h. The gas mixture compositions were evaluated with an ABB continuous gas analyser (model ABB AO2020). The CO 2 and CH 4 conversions were estimated according to Equations (1) and (2), respectively.

Conclusions
In this work, the catalytic performance of supported nickel phosphide catalysts was investigated for dry reforming of methane. By analysing Al 2 O 3 , SiO 2 -Al 2 O 3 and CeO 2 supports, the prominent impact of the support nature on the catalytic behaviour of Ni 2 P systems was clearly established. The reaction rate observed for the catalysts' series decreased, following the trend Ni 2 P/Al 2 O 3 > Ni 2 P/CeO 2 > Ni 2 P/SiO 2 -Al 2 O 3 . The observed catalytic trend should be explained considering different factors. Thus, lower conversion rates, depicted by the Ni 2 P/SiO 2 -Al 2 O 3 system, were related to the higher Ni sintering and coking rates, coupled with hindered constitutions of Ni 2 P species. For the Ni 2 P/CeO 2 system, the lower surface areas coupled with low concentrations of Ni 2 P constituted over the catalyst surface might account for the exhibited catalytic performance. In any case, the strong deactivation issues displayed by the Ni 2 P/CeO 2 catalyst under reaction conditions should most likely relate to the partial oxidation of Ni species, since no significant C deposits nor Ni sintering effects were discerned for the post-reacted sample. For the ceria supported catalysts, the oxidation of methane and C intermediate species might also play a role.
Therefore, the obtained outcomes suggest the unsuitability of Ni 2 P/CeO 2 and Ni/SiO 2 -Al 2 O 3 catalysts for DRM processes. Within the analysed series, the Ni 2 P/Al 2 O 3 catalyst showed the highest catalyst performance as well as excellent stabilities. At 700 • C, the alumina supported catalyst consistently produced stable CO 2 conversions only 12% below equilibrium and a favourable H 2 /CO ratio of 1. The observed catalyst behaviour could be associated to the larger concentrations of Ni 2 P phases and fair Ni dispersions. Overall, the Ni 2 P/Al 2 O 3 catalyst has been proven to be an effective and easily synthesised novel catalyst with remarkable activity and stability for DRM reaction.

Funding:
The team at Surrey acknowledges the financial support provided by the EPSRC grant EP/R512904/1 as well as the Royal Society Research Grant RSGR1180353. This work was also partially sponsored by the CO2Chem UK through the EPSRC grant EP/P026435/1.

Data Availability Statement:
The data described in this article are openly available in the Open Science Frameworks at Digital Repositories of the University of Surrey (ExLibris) and Branden-burg University of Technology (OPUS).