Ammonia can be used as fuel in internal combustion engines (ICEs). In this case, a flame accelerator, such as hydrogen, is needed. H2 can be produced on-board by partial decomposition of ammonia. In this work, ruthenium nanoparticles were embedded into a lanthanum-stabilized zirconia (LSZ) support to obtain active and stable heterogeneous catalysts for NH3 decomposition. The effects of the preparation of both Ru nanoparticles and LSZ support were investigated. The embedded catalysts present high metal dispersion and good metal accessibility. Despite the relatively low metal loading (3 wt %), activity was very high in the temperature range 400–600 °C. The activity of the reference catalysts prepared by using classical impregnation was significantly lower under the same working conditions. Although many factors contribute to the final catalyst performances, the data reported confirm that the embedding strategy minimizes the undesirable sintering of the Ru nanoparticles, leading to promising and stable catalytic activity.
Interest in nonconventional fuels has significantly increased over the last few years because of environmental issues, as well as the irreversible decrease in fossil-fuel reservoirs.1 Several alternatives to fossil fuels are under investigation worldwide, for example, renewable fuels, obtained from second- and third-generation biomasses, and H2 in combination with fuel cells. A third, less investigated alternative is ammonia (NH3).2–4 Ammonia is an interesting energy vector, as it can be used either for the production of H2 (NH3 has an H2 content of 17.7 wt %, higher than that of methanol) or as an ultimate fuel in internal combustion engines (ICEs). In this respect, the use of ammonia as a fuel for ICEs is attracting increasing interest. Ammonia-fuelled engines can operate by slightly modifying conventional spark ignition engines. However, owing to its high autoignition temperature (651 °C), ammonia needs a flame accelerator, such as hydrogen, to burn efficiently. Therefore, ammonia must be, in part, decomposed to H2/N2 mixtures, to maintain acceptable combustion in an engine. A minimum of 5 vol % of H2 is required in the gas stream for a normal ICE operation, whereas a higher H2 content is beneficial during the starting procedure.5
Hydrogen can be produced from ammonia by endothermic decomposition, accordingly a process favored at high- temperature and low-pressure operating conditions [Eq. (Spiel um Platz drei)]:
Notably, the waste heat from the ICE can be used to sustain the endothermic reaction of H2 production needed for the operation of the system.
In this context, the development of heterogeneous catalysts that are active in the temperature range of 400–600 °C is a must to design an efficient ammonia-fuelled ICE. In this way, the waste heat produced by the ICE could be used to heat the NH3 decomposition reactor. Many different metals have been tested as active catalysts for ammonia decomposition; particularly active are group VIII metals (Fe, Ni, Ru, Ir, Co, Pt).6 The catalytic activities are generally very low below 600 °C, even for Ru, which generates the most active systems. The performance of Ru-based nanosized catalysts is strongly affected by the nature of the support7–9 and by the mean crystallite size, with an optimum size around 2 nm.10, 11 Unfortunately, Ru-based catalysts usually exhibit quite low metal dispersion. For this reason, increasing efforts are being made to reduce the metal particle size.12
Graphitic carbons and carbon nanotubes (CNTs) are very attractive supports for Ru catalysts,9, 13 although their high cost may limit large-scale use. NH3 can be completely decomposed over K-promoted Ru(5 wt %)/CNT catalyst at relatively low temperature (450–500 °C) and even at high space velocity.14 The good performance of Ru/CNT catalyst is related to a) the presence of COOH and OH groups on the CNT surface, responsible for an effective anchoring of Ru entities, and b) the confinement effect of Ru nanoparticles inside the CNT pores (with an inner diameter of 3–10 nm). Although the catalyst shows a good stability of the catalytic activity, carbon stability against gasification has not been reported.9, 14
Basic oxides are more efficient as supports for Ru catalysts, as compared to acidic oxides.6 Zheng et al.10 reported that Ru(5 wt %)/Al2O3 facilitates the complete conversion of ammonia at 550 °C, whereas Choudhary et al.7 obtained the complete conversion of NH3 only at 650 °C by using Ru(10 wt %)/SiO2 under similar reaction conditions. Zhang et al.15 reported that a Ru(2.8 wt %)/MgO catalyst, prepared by the polyol reduction method, shows approximately 80 % NH3 conversion at 500 °C. The complete conversion was achieved by the use of Cs, as the promoter, deposited by impregnation. Yin et al.16 reported that high NH3 conversion can be achieved by using a superbasic support, such as ZrO(OH)2 gel digested by heating in a refluxing KOH aqueous solution (denoted as K-ZrO2-KOH). This support is amorphous and contains significant amount of K and Si. Ru (loading≈5 wt %) was deposited on this support, with a high metal dispersion, from either [Ru(acac)3] or RuCl3.32 This study provided evidence of the poisoning effect of residual chloride ions. Ru/K-ZrO2-KOH from [Ru(acac)3] shows a very good stability at 350 °C under severe reaction conditions (high space velocity) within a 300 h period, with rapid aging at higher temperatures (up to 600 °C).
We report herein the preparation and the functional and structural/morphological characterization of an active and stable lanthanum-stabilized zirconia (LSZ)-based catalyst containing 3 wt % Ru. The catalysts were prepared by an innovative synthetic method that allows the encapsulation of an active metal phase inside a porous oxide matrix, which generally avoids sintering of the metal, as well as catalyst deactivation, during the reaction. The adopted embedding strategy is largely reported in the literature for the preparation of heterogeneous catalysts with peculiar activity and stability.17 Heterogeneous catalysts comprising embedded Ru nanoparticles were previously described for a variety of supports, preparation methods, and test reactions.18–21 The embedding methodology was successfully employed by our group in other processes involved in the H2 production, such as methane partial oxidation22, 23 and ethanol steam reforming.24, 25
Two different embedding strategies were adopted in this study to prepare Ru@LSZ catalysts. Ru nanoparticles were synthesized by reduction of RuCl3 in the presence of two different protecting agents. In the first case, Ru nanoparticles were prepared by alcohol reduction in the presence of a polymer (polyvinyl pyrrolidone, PVP), obtaining a deep purple metal nanoparticle suspensions (DP method).26 In the second case, Ru nanoparticles, formed by NaBH4 reduction were protected by N-hexadecyl-N-(2-hydroxyethyl)-N,N-dimethyl ammonium bromide, a cationic surfactant denoted as HEAC16Br (NaBH4 method).27 The preparation conditions were chosen to obtain protected Ru nanoparticles with a mean dimension of 2.5 nm if the DP method26 was used and of 3 nm if the NaBH4 method27 was used. Around the nanoparticles, mixed La and Zr hydroxides were deposited by coprecipitation. After various post-synthesis treatments, Ru nanoparticles embedded in the ZrO2-based oxides were obtained. The performances of the embedded catalysts were compared to those of catalysts prepared by conventional impregnation methods (Ru/LSZ).
Results and Discussion
Catalytic activity for NH3 decomposition
The activities of prereduced embedded or impregnated Ru(3 wt %)-LSZ catalysts for NH3 decomposition are shown in Figure 1. Before catalytic activity experiments, the samples were fully reduced in H2 flow at 800 °C for 3 h. This protocol guarantees full metal reduction and serves as aging pretreatment of the samples, to avoid transient phenomena during activity experiments. Catalytic tests were performed at atmospheric pressure and with pure NH3. Both high and low space velocities were investigated. The samples prepared by the embedding method showed higher activity compared to the samples prepared by classical impregnation. Over the embedded catalysts, NH3 conversion started above 300 °C, whereas over the impregnated catalysts the reaction started at 400 °C. The NH3 conversion over Ru@LSZ-DP was slightly higher than over Ru@LSZ-NaBH4. One may readily observe significant differences in the ammonia conversion over the whole range of temperatures investigated: The light-off temperatures (T50—defined as the temperature corresponding to a 50 % NH3 conversion) for the embedded catalysts were at least 200 °C lower than those employed for the corresponding impregnated samples. Two reactivity cycles were performed on the various catalysts, without evidencing any appreciable deactivation.
Tests on other Ru-based catalysts under comparable reaction conditions have been reported (GHSV=30 000 mL g−1 h−1).9, 10, 15 Ru@LSZ-DP promotes ammonia conversion to the same extent as the K-promoted catalyst Ru(5 wt %)/CNTs reported by Yin et al.,9 which is one of the most active catalysts ever reported. Notably, the metal loadings in our embedded catalysts are lower than that of the catalysts usually employed for NH3 decomposition. An attempt to further decrease the Ru metal loading to 1 wt % was investigated using the DP method. Unfortunately, the activity was significantly lower with respect to Ru@LSZ-DP, with a NH3 conversion of 40 % at 500 °C (GHSV=30 000 mL g−1 h−1). On the other hand, no improvement was observed on increasing the Ru loading to 5 wt %.
An important issue for any new heterogeneous catalyst is the stability of the material under prolonged exposure to the reaction mixture. To this purpose, the performance of the embedded Ru@LSZ catalysts and of the corresponding reference impregnated Ru/LSZ-NaBH4 were investigated over a period of 120 h at 500 °C (Figure 2). This temperature can be easily reached during the operation of an NH3 decomposition reactor by using the waste heat from an ICE and, at the same time, allows us to obtain high NH3 conversion under the experimental conditions adopted in this study. The stability of the Ru/LSZ-DP catalyst was not studied, owing to its very low activity. All the catalysts showed NH3 conversion comparable to that during the run-up experiments at 500 °C (Figure 1). Analysis of Figure 2 reveals that the NH3 conversion remains almost constant for all the catalysts over the period of the stability test.
To investigate more realistic conditions, reactivity tests were also performed by using technical ammonia, that contains approximately 3000 ppm of H2O, 50 ppm of O2 and 10 ppm of hydrocarbons. Identical activity and stability results were obtained, indicating that small impurities in the feedstock have a negligible effect on the catalytic performance.
The reduction properties of the as prepared samples were examined by temperature-programmed reduction (TPR) under a H2(5 %)/Ar flow (Figure 3).
As a general characteristic, all the samples, both embedded and impregnated, revealed multiple reduction processes at low temperature (below 250 °C). The positions of the reduction maxima are strongly dependent on the reaction conditions (mass of catalysts, gas flow and composition, heating rate). However, it is usually accepted that the reduction at low temperature is ascribable to a highly dispersed RuOx species, whereas at higher temperature the reduction of well crystallized RuO2 particles is operative.28 Moreover, the reduction temperature is affected by the nature of the support.29, 30 Consistently, the different H2 consumption at low temperature of the samples based on identical supports can be justified by considering a different size and distribution of the RuOx particles.
A reduction peak at high temperature (600–700 °C) was detected, especially for the samples prepared by the DP method. The reduction of RuOx species in this temperature range has not to date been reported in the literature. Although this reduction process might be partially ascribed to an artifact (originated by an over-pressure on the sample, owing to the compression of the sample), the reduction of RuOx species strongly interacting with the support or almost fully embedded into the oxide matrix appears as a reasonable explanation for the H2 consumption at high temperatures. Indeed, the high temperature peak was more pronounced in the samples that have lower specific surface areas (see below) for which the contribution of such an over-pressure artifact should be minimal. Quantification of H2 consumed during TPR was performed by integrating the TPR peak and comparing the obtained areas, which are very similar in all cases, with the value obtained using CuO as standard. The results indicate that the H2 consumption is consistent with a Ru amount very close to the nominal 3 wt %, present in the form of RuO2, confirming that the reduction processes are essentially related to RuOx species. The presence of metal oxide species strongly interacting with the support has already been reported for catalysts prepared by the embedding approach.22
Consistent with the TPR results, a pretreatment in H2 flow (35 mL min−1) at 800 °C for 3 h was applied to all the materials to fully reduce the metal species before the reactivity test. This pretreatment serves as an activation treatment and as an aging process, because the samples were not subjected to any thermal treatment above 500 °C.
Powder XRD analysis of the samples after different thermochemical treatments have given evidence for great differences among the various materials studied.
The XRD patterns of the samples after calcination at 500 °C are shown in Figure S1 (see the Supporting Information). The XRD patterns of the calcined samples are dominated by broad reflections attributable to the LSZ support. The presence of amorphous materials cannot be ruled out (especially in Figure S1, traces b and c). RuO2 reflections were generally weak and very sharp, indicating the presence of very large RuO2 particles (crystallite size 30–100 nm). In the case of the Ru/LSZ-DP sample, the presence of NaCl is evident, likely originated during the synthesis of the catalyst by reaction of chloride from RuCl3 with impurities of sodium.
XRD patterns of the samples after standard activation in H2 (800 °C for 3 h) are shown in Figure 4. Depending on the preparation method and treatment, different crystalline phases were observed. The analysis of each XRD pattern (Figures S2–S5 in the Supporting Information) allows for a semi-quantitative determination of the structure and composition of the various phases. Summarized in Table 1 are the crystal phases identified and the mean crystallite size for each phase, whereas the crystallographic parameters (from the Rietveld refinements) are reported in Tables S1–S4 (see the Supporting Information).
Table 1. Results from the analysis of the XRD patterns of embedded (Ru@LSZ) and impregnated (Ru/LSZ) samples.
After stability tests[c]
Loading [wt %]
Loading [wt %]
Loading [wt %]
[a] After standard activation pretreatment (reduction in H2 flow at 800 °C for 3 h); [b] activated sample after 2 cycles under NH3 decomposition conditions (GHSV=30 000 mL g−1 h−1); [c] activated sample after stability tests under NH3 decomposition conditions (GHSV=30 000 mL g−1 h−1) for 110 h at 500 °C; [d] CS=crystallite size calculated applying Scherrer’s equation; [e] Rietveld results for these samples are semi-quantitative, since the exact composition of the phases is not known and some reflections are not clearly identified.
tetragonal LSZ cubic LSZ Na2ZrO3
46.5 14.9 38.6
16 13 42
tetragonal LSZ cubic LSZ Na2ZrO3
54.2 27.7 18.1
17 12 37
tetragonal LSZ cubic LSZ Na2ZrO3
12.0 28.6 59.4
17 16 31
tetragonal LSZ Ru
tetragonal LSZ Ru
tetragonal LSZ Ru
tetragonal LSZ cubic LSZ monoclinic ZrO2 Ru
44.7 10.2 41.7 3.4
25 58 69 70
tetragonal LSZ cubic LSZ monoclinic ZrO2 Ru
44.5 12.7 39.9 2.9
25 58 64 79
tetragonal LSZ Ru
tetragonal LSZ Ru
tetragonal LSZ Ru
The samples prepared by the DP method (either embedded or impregnated) are composed of a complex mixture of phases. The XRD pattern of the reduced Ru@LSZ-DP sample (Figure S2 and Table S1 in the Supporting Information) is dominated by the tetragonal structure (space group P42/nmc) of the LSZ. An increase of the cell parameters with respect to those of pure tetragonal ZrO2 (JCPDS n° 50-1089: a=0.35984 nm, c=0.51520 nm) and the very low tetragonal distortion (from the cell parameters, c/(a) of 1.005 can be calculated) clearly prove the incorporation of La into the ZrO2 oxide, as a result of the larger ionic radius of LaIII with respect to ZrIV (for a 8-fold coordinated site, 0.130 nm and 0.098 nm, respectively).31 The XRD pattern of the reduced Ru@LSZ-DP reveals also the presence of monoclinic Na2ZrO3 and of a cubic LSZ phase with fluoritic structure. The cell parameters of Na2ZrO3 are close to those of the pure monoclinic phase (JCPDS no. 35-0770: a=0.56233 nm, b=0.97491 nm, c=1.11270 nm, β=99.983°). The formation of Na2ZrO3 is an indication of the presence of large amounts of Na+ adsorbed on the solid material as a consequence of the precipitation of the La/Zr mixed hydroxides with NaOH. The presence of Na+ in the material may also affect the structure of the tetragonal LSZ phase, reducing the tetragonal distortion.32 Finally, accordingly to the empirical formula proposed by Kim,33 from the value of the cell parameter of the cubic LSZ phase (0.5309 nm), it is possible to calculate the composition of Zr1−xLaxO2−x/2. (x=0.385), which indicates that an enrichment in La with respect to the nominal composition of the LSZ oxide has occurred. However, the inclusion of Na+ as an impurity in the La site, resulting in the formation of a cubic doped-ZrO2,32 cannot be ruled out. Notably, no reflection clearly related to Ru-based phases was detected, suggesting that Ru is present as highly dispersed nanoparticles. This feature is in good agreement with the high catalytic activity demonstrated by this sample. Zheng et al.10 reported that the maximum reaction rate for NH3 decomposition is observed when Ru nanoparticles of approximately 2 nm are present on the catalyst. Moreover, Na2ZrO3 or Na+ may act as promoters for the reaction, as is well known for basic oxide supports15, 16 and/or the presence of alkali metals as promoters9, 14–16 may have a beneficial effect on Ru catalysts. To demonstrate the effect of Na+ or Na2ZrO3 on the catalytic performance of Ru@LSZ-DP, several attempts were made to reduce the presence of Na+ by washing the gel with diluted HNO3 (1 %) and to reintroduce sodium by impregnating the washed material with NaNO3 (soaking the catalyst powder into a 2 M aqueous solution of NaNO3). After washing, a lower catalytic activity was observed (Figure S6 in the Supporting Information). The XRD pattern of the reduced sample (Figure S7 in the Supporting Information) shows the formation of a solid solution La2O3ZrO2 with tetragonal structure and of large Ru crystallites (167 nm). This observation can be attributed to the possible Ru redispersion during the washing treatment. After impregnation of the washed material with NaNO3, a further decrease of the catalytic activity was observed (Figure S6). The XRD pattern of the activated sample shows the presence of tetragonal LSZ, monoclinic ZrO2, metallic Ru (mean crystallite size of 124 nm) and of a cubic phase with patterns compatible with those of Na2O (although its cell parameter is strongly reduced with respect to the theoretical value of Na2O, 0.5392 vs 0.555 nm). Moreover, a strong sintering of the material is observed, as a result of the phase separation and/or pore blocking induced by NaNO3 impregnation.
The situation of the activated Ru/LSZ-DP (prepared by impregnation) is more complex than that of the corresponding embedded sample (Figure S4 and Table S3 in the Supporting Information). The oxide support is composed of a mixture of tetragonal LSZ (major phase), monoclinic ZrO2 and cubic LSZ. The latter phase is featured by a cell parameter shorter than that of the corresponding phase observed for the reduced Ru@LSZ-DP, suggesting a different degree of substitution of Zr with La and/or Na. Moreover, a non-stoichiometric NaCl- derived phase was included to better fit the XRD pattern, although a significant lengthening of the cell parameter was observed. Only minor reflections remain unidentified. Notably, in this sample, clear and sharp reflections related to metal Ru are present, indicating the presence of very large Ru particles (ca. 70 nm). The presence of unknown amounts of Na-based compound precludes the quantification of the other species.
The XRD analysis of the samples prepared following the NaBH4 method is significantly less complex (see the Supporting Information; Figure S3 and Table S2 for the embedded Ru@LSZ-NaBH4; Figure S5 and Table S4 for the impregnated Ru/LSZ-NaBH4). A solid solution of ZrO2 and La2O3 with a tetragonal structure (designed as tetragonal LSZ) constitutes the oxide support. No reflections related to cubic LSZ, monoclinic ZrO2 or monoclinic Na2ZrO3 were detected. This result suggests that the washing procedure adopted to remove the bromide ions and the excess of surfactant used for the preparation of the Ru nanoparticles is also effective for the elimination of the Na+ impurities. Notably, the catalytic activity of Ru@LSZ-NaBH4 was only slightly lower than that of Ru@LSZ-DP. Weak, but very sharp, reflections assigned to metallic Ru were also observed. In the case of the impregnated Ru/LSZ-NaBH4 sample, the Rietveld analysis accounts for a 3.2 wt % of Ru (instead of the nominal 3 wt %) with very large crystallites (ca. 78 nm). In contrast, the amount of Ru estimated for the embedded Ru@LSZ-NaBH4 is quite low (0.5 wt %). Furthermore, very large Ru crystallites were observed (ca. 120 nm). The low amount of Ru estimated by the Rietveld analysis, in conjunction with the good catalytic activity and the acceptable Ru accessibility (see H2 chemisorption section) suggests the presence of a large number of highly dispersed Ru nanoparticles that are not detectable by XRD. A partial redissolution of some small Ru nanoparticles during the washing procedure cannot be excluded. The redeposition of Ru on the external surface of the material might account for the presence of these large Ru particles. In fact, a significant sintering of the noble metal redeposited on the external surface of the support and, therefore not protected, may occur during the calcination and reduction steps.
After aging under NH3 decomposition, only minor modifications of the structural properties of the catalysts were evidenced. The XRD patterns of the samples after aging under run-up experiments are shown in Figure 5, left (the results of the Rietveld analysis are presented in Figures S8–S11 and Tables S5–S8 in the Supporting Information), whereas, after the stability test the samples exhibit the XRD pattern shown in Figure 5, right (the results of the Rietveld analysis are presented in Figures S12–S14 and Tables S9–S11 in the Supporting Information). Notably, no significant difference in the Ru crystallite size was observed. Only for Ru@LSZ-DP, we observed that, after run-up experiments (up to 800 °C), the amount of Na2ZrO3 decreased to form tetragonal and cubic LSZ. In contrast, after the stability test (prolonged exposure to NH3 at 500 °C), the amount of tetragonal LSZ decreased to form Na2ZrO3 and cubic LSZ. Nevertheless, the activity and stability results suggest that the evolution of the phases in the oxide does not appreciably influence the catalytic activity and the dimensions of the Ru nanoparticles.
The results of the textural characterization of the calcined samples are summarized in Table 2. All the samples gave rise to type IV isotherms with hysteresis loops typical of mesoporous materials.34 The t-plot analysis reveals that the microporous volume is always negligible. After all of the thermochemical treatments considered, the samples prepared by the DP method showed lower surface area and pore volume, as well as larger pores than the samples prepared by the NaBH4 method. In the latter samples, the supports are homogeneous and their better textural properties (Table 2) are attributed to the stabilizing effect of the formation of ZrO2La2O3 solid solution. Moreover, in the preparation of the NaBH4 samples, the precipitate was carefully washed to remove Br− and Na+ ions, as well as the excess of surfactant, whereas in the DP method the excess of PVP was removed only by filtration of the mother liquor. The deep sintering of the DP materials might be the result of the presence of a various crystallographic phases and of a possible local over-heating consequent to the combustion of the residual polymer. Finally, in the NaBH4 preparation method, the large amount of a cationic surfactant (HEAC16Br) can act as directing agent during the gel formation, leading to a material with high surface area similarly to that previously reported for ZrO2-based materials.35, 36
Table 2. N2 physisorption results for embedded (Ru@LSZ) and impregnated (Ru/LSZ) samples.
After stability tests[d]
SSA[e] [m2 g−1]
CPV[g] [mL g−1]
SSA[e] [m2 g−1]
CPV[g] [mL g−1]
SSA[e] [m2 g−1]
CPV[g] [mL g−1]
SSA[e] [m2 g−1]
CPV[g] [mL g−1]
[a] After calcination in air at 500 °C for 5 h; [b] after standard activation pretreatment (reduction in H2 flow at 800 °C for 3 h); [c] after 2 cycles under NH3 decomposition conditions (GHSV=30 000 mL g−1 h−1); [d] after stability tests under NH3 decomposition conditions (GHSV=30 000 mL g−1 h−1) for 110 h at 500 °C; [e] SSA=specific surface area; [f] dM=maxima of the pore distribution calculated by BJH analysis of the desorption branch of the N2 isotherm; [g] CPV=cumulative pore volume; [h] bimodal distribution.
After standard activation pretreatment (reduction in H2 flow at 800 °C for 3 h), all the samples underwent a significant decrease of the surface area and of the pore volume, as a result of sintering phenomena induced by the reduction at high temperature. Despite its good catalytic activity (Figure 1), the embedded Ru@LSZ-DP sample exhibits a low surface area (Table 2). On the other hand, the embedded Ru@LSZ-NaBH4 sample shows a lower decrease of the surface area and of the pore volume, owing to the higher thermal stability of the tetragonal ZrO2-La2O3 solid solution. Comparable surface areas were observed for the embedded catalysts and the respective impregnated samples. These results suggest that the catalytic activity is mainly influenced by the Ru crystallite size, and not by the surface area of the support.
Finally, the textural properties of the samples were only marginally affected by the aging under NH3 decomposition, during both run-up tests and long term stability tests.
Static and dynamic H2 chemisorption
H2 chemisorption experiments were performed with the aim of further investigating the accessibility of the metal nanoparticles to NH3. The measurements were performed both under static and dynamic conditions in order to better evaluate the accessibility of Ru. It is important to underline that, unlike most metals, the determination of the Ru dispersion by chemisorption is particularly difficult, as the H2 adsorption kinetics are quite slow. Many attempts to establish the best experimental conditions for the determination of the Ru dispersion by chemisorptions have been reported in the literature.37, 38 Okal et al.37 demonstrated that, operating at 100 °C and with an equilibration time of at least 1 h, static H2 chemisorption experiments provide reliable values of the Ru dispersion. Similar experimental conditions were used in this study (Table 3).
Table 3. H2 chemisorption results for embedded (Ru@LSZ) and impregnated (Ru/LSZ) samples after standard activation pretreatment (H2 reduction at 800 °C for 3 h).
Static H2 chemisorption[a]
Dynamic H2 chemisorption[b]
Area[d] [m2 g−1]
Area[d] [m2 g−1]
[a] Static chemisorption was measured on the samples at 100 °C in the 25–400 torr pressure range. The reversible adsorption was subtracted by the double isotherm method; [b] dynamic chemisorption was performed at 100 °C using pulses of H2 (73 μL) with N2 as the carrier gas (75 mL min−1); [c] calculated using a H/Ru=1:1 stoichiometry; [d] Metal surface area per gram of catalyst; [e] PS=article size.
The sample Ru@LSZ-DP shows a relatively high static H2 chemisorption capability, suggesting the presence of small Ru nanoparticles, in agreement with the absence of reflections related to Ru in the XRD pattern (Figure 4, trace a). Nevertheless, a lower H/Ru ratio was shown by a dynamic H2 chemisorption experiment. In all the other cases, the results obtained using the two techniques were comparable and in good agreement with the indication obtained from the XRD patterns of the corresponding reduced samples. The relatively high H/Ru ratio for the activated Ru@LSZ-NaBH4 is another indication of the presence of an appreciable amount of small Ru nanoparticles, partially embedded into LSZ, together with a fraction of large Ru particles (see the XRD data).
Static H2 chemisorption experiments have shown that the Ru nanoparticles are accessible to the gas phase reagents, evidencing that the embedding preparation does not result in a complete occlusion of the metal phase. Furthermore, the embedded catalysts present the highest Ru dispersion and the highest catalytic activity. On the other hand, the impregnated samples exhibit a very low metal surface area (Table 3), large Ru crystallites (from XRD), and the worst catalytic performance.
In the case of the Ru@LSZ-DP sample, for which static H2 chemisorption suggests the presence of very small Ru nanoparticles, dynamic H2 chemisorption reveals a H/Ru ratio significantly lower than that obtained by the static technique. Despite the presence of large pores in this sample, this result could be attributed to gas diffusional limitations. On the other hand, an effect of the small Ru particle size in the H2 adsorption kinetics cannot be ruled out. Rather, a better agreement between the two techniques was obtained when very large Ru particles were present, as in the impregnated samples. In the case of Ru@LSZ-NaBH4, for which H2 can be easily adsorbed on the large particles and transferred to the smaller ones by spillover through the support, a realistic determination of the Ru particle size can be achieved also by dynamic H2 chemisorption.
Temperature-programmed surface reaction
Temperature-programmed surface reactions (TPSR) of adsorbed ammonia were conducted to further investigate the differences in the samples considered (Figure 6). After saturation of the surface with NH3 at 110 °C, the desorption of various species was observed. The main species detected during the heating ramp were H2, N2, NH3, and H2O (from the support).
No significant evolution of species other than H2O (not reported) was observed for the Ru/LSZ-DP sample, which possesses a low surface area (5 m2 g−1) and very low metal dispersion (H/Ru=0.026 by static H2 chemisorption). These results suggest that the amount of NH3 adsorbed on this material is negligible, in agreement with the very low activity of this catalyst. In contrast, the samples with either a high surface area (Ru/LSZ-NaBH4), or a high metal dispersion (Ru@LSZ-DP), or both (Ru@LSZ-NaBH4), showed significant desorption of H2, N2, and NH3. The relative amounts of the desorbed compounds reflect the characteristics of the investigated samples, suggesting that H2 and N2 are produced from the decomposition of ammonia adsorbed on the Ru nanoparticles, whereas the desorbed NH3 is derived from ammonia adsorbed on the support. Notably, TPSR experiments on the embedded samples show the prevalence of the reaction products (H2 and N2), in agreement with their high metal dispersion and catalytic activity. Moreover, Ru@LSZ-NaBH4 desorbs larger quantities of the products, owing to its higher surface area with respect to Ru@LSZ-DP. Finally, Ru/LSZ-NaBH4 shows the prevalence of NH3 as desorbed product, accordingly with the high surface area and low metal dispersion.
The trend in the peak temperature of the different species (corresponding to the temperature of the maximum of the desorption peak) is consistent with the catalytic performance of the materials, as well as with their morphological characteristics. The samples prepared by the NaBH4 method showed peak temperatures of 385 °C for H2, 325 °C for N2, and 250 °C for NH3. The highest intensity observed for Ru@LSZ-NaBH4 can be ascribed to the higher metal dispersion, which favors the NH3 adsorption. On the other hand, Ru@LSZ-DP, which is the most active, showed lower peak temperatures (345 °C for H2, 270 °C for N2, and 260 °C for NH3). The decrease of the desorption temperatures for Ru@LSZ-DP compared to Ru@LSZ-NaBH4 may be related to the presence of Na2ZrO3, in view of the well known promotional effect of either basic supports15, 16 or alkali metal promoters9, 14–16 on the activity of Ru catalysts. In a study of the dissociative chemisorption and associative desorption of N2 on Ru/MgO catalysts, Hinrichsen et al.39 showed that, after impregnation with CsNO3, the temperature for N2 desorption from a surface previously saturated with N atoms (from N2 dissociation) strongly decreased. Moreover, studying Ru/CNT catalysts for NH3 decomposition, Wang et al.40 reported that the promotional effect of a modifier can comprise i) an enhancement of the combinative desorption of nitrogen atoms and ii) a decrease of the apparent activation energy of the decomposition reaction. The best promoters for Ru/CNT catalysts are K, Na, and Li from nitrates, hydroxides or carbonates40 and their action is an electronic effect on the Ru nanoparticles, like reported for other catalysts for NH3 synthesis.41–46 Conversely, electron-withdrawing groups (such as halogens or sulfates) can act as inhibitors.40 In this respect, it is not surprising that the catalyst Ru/LSZ-DP, which comprises a very complicated mixtures of phases, including chloride ions, exhibits a very low activity and neither H2 nor N2 desorption was observed in the TPSR experiments.
The influence of the synthetic method on the catalytic activity of Ru-based catalysts for ammonia decomposition has been investigated. Active and stable catalysts were prepared by embedding preformed Ru nanoparticles (protected by a polymer or a cationic surfactant) into a La-doped ZrO2 (LSZ) support formed by coprecipitation. The activity of the embedded catalysts was significantly higher than that of reference catalysts prepared by Ru impregnation of similar supports. Depending on the preparation method, the oxide support was composed either of a ZrO2La2O3 solid solution, with tetragonal structure and high surface area, or of a complex mixture of phases (such as tetragonal and cubic LSZ, monoclinic ZrO2 and Na2ZrO3), with small surface areas, derived from the presence of considerable amounts of a Na-based impurity (owing to the precipitation step with NaOH). Notably, the factor governing the activity of the catalysts was the dispersion of the Ru nanoparticles. Catalysts prepared by embedding preformed nanoparticles contained small Ru nanoparticles (as shown by XRD and chemisorption). As expected, the catalysts prepared by the impregnation method underwent a significant sintering of the metal phase, leading to significantly lower conversions. Temperature-programmed surface reaction experiments with ammonia were consistent with the catalytic activity results, indicating that high H2 and N2 desorption took place only when Ru was dispersed as very small nanoparticles.
Preparation of the catalysts
The catalyst preparation was designed to optimize the performance for the ammonia decomposition reaction. A nominal metal loading of 3 wt % was used for all samples. The preparation of Ru@LSZ involved with the preparation of colloidal Ru nanoparticles as first step. LSZ was deposited around the Ru nanoparticles by a precipitation technique. The two preparation methods differ for preparation of colloidal Ru nanoparticles.
Deep purple (DP) Method: Preparation of Ru@LSZ-DP
Colloidal Ru nanoparticles were prepared by a reflux method using poly(N-vinyl-2-pyrrolidone) (PVP) as protecting agent.26 The chosen methodology allows modulation of the mean dimension of the Ru nanoparticles by varying the experimental conditions, such as the Ru concentration, the PVP/Ru molar ratio and, most importantly, the molecular weight of the boiling alcohol. In fact, the increase of the molecular weight and the boiling point of the solvent alcohol resulted in a decrease of the mean dimension of the Ru nanoparticles. On the other hand, high boiling alcohols are hardly removable by evaporation and the solubility of metal oxide precursors are low. For the preparation of these samples, ethanol was selected as solvent, leading to very small colloidal Ru nanoparticles as a deep purple (DP) suspension.
In a 250 mL bottom flask, PVP K30 (0.989 g, 8.9 mmol as monomeric unit) and RuCl3⋅x H2O (0.233 g, 0.89 mmol Ru) were dissolved in ethanol (120 mL) under stirring to form a dark red solution. After heating at reflux for 24 h, a deep purple colloidal suspension was obtained, according to the procedure reported by Zhang et al.26 Under the experimental conditions adopted (VP/Ru=10, PVP molecular weight≈40 000 g mol−1, ethanol as solvent and reducing agent, reflux time of 24 h), Ru nanoparticles with a mean diameter of 2.5 nm were obtained. Before the deposition of LSZ around the nanoparticles, ethanol was evaporated at reduced pressure and the PVP-protected nanoparticles were resuspended in water (50 mL).
The La2O3-stabilized ZrO2 (LSZ) was deposited around the metal nanoparticles by a co-precipitation method. The aqueous Ru nanoparticles suspension was added under vigorous stirring to an aqueous solution of ZrO(NO3)2⋅x H2O (5.77 g, 99 %, Aldrich) and La(NO3)3⋅6 H2O (1.01 g, 99.9 %, Aldrich) in water (250 mL). The metal solution comprising the protected Ru nanoparticles was added dropwise to an aqueous solution of NaOH (220 mL, 2.5 M). The obtained precipitate was stirred for 2 h and then filtered and briefly washed with water on the filter. The solid obtained was suspended in 2-propanol (300 mL) and heated at reflux for 4 h to stabilize the texture of the oxide matrix.47, 48 After filtration, the solid was dried at 120 °C overnight and calcined in a static oven at 500 °C for 5 h with heating and cooling rates of 1 °C min−1.
NaBH4 method: Preparation of Ru@LSZ-NaBH4
Metal nanoparticles were synthesized by reduction of RhCl3 with NaBH4 in an aqueous solution containing cationic surfactants. The Ru nanoparticles were protected against aggregation by entrapment inside the hydrophobic environment of the micelles formed by the surfactants. N-Hexadecyl-N-(2-hydroxyethyl)-N,N-dimethyl ammonium bromide (HEAC16Br) was used as protecting agent. The use of HEAC16X (X=Cl or Br) as protecting agent for metal nanoparticles preparation has been extensively reported for different noble metals, such as Rh,49, 50 Ir,51 Pt,52 Pd,53 and Ru.27 Moreover, Rh nanoparticles protected by HEAC16Br have already been used as building blocks for the preparation of embedded catalysts.22, 24, 25
The cationic surfactant HEAC16Br was prepared by reaction of 1-bromohexadecane (99 %, Fluka) with a 30 % excess of N,N-dimethylethanolamine (99 %, Fluka) in absolute ethanol (Aldrich) at reflux for 1 h,54 and purified by crystallization in absolute ethanol,55 yield 70–80 %, m.p. 200–210 °C. The 1H and 13C NMR spectra are in agreement with those reported in Ref. 50).
The preparation of HEAC16Br-protected Ru nanoparticles was carried out as follows. A solution of surfactant HEAC16Br (3.262 g, 8.3 mmol) and NaBH4 (84 mg, 2.2 mmol) in water (210 mL) and a solution of RuCl3⋅x H2O (0.233 g, 0.89 mmol Ru) in water (52 mL) were thermostated at 20 °C for at least 1 h. Then, the solution containing HEAC16Br and NaBH4 was quickly added under vigorous stirring to the Ru solution to obtain an aqueous Ru colloidal suspension. The reduction occurs instantaneously, revealed by the immediate color change from yellow-brown to black. The suspension is then stirred for 2 h to decompose the excess of the reducing agent. Ru nanoparticles with a mean dimension of 3.0 nm are reported following this preparation.27
The La2O3-stabilized ZrO2 (LSZ) was deposited around the metal nanoparticles by a co-precipitation method, as previously described. The aqueous Ru nanoparticles suspension was added under vigorous stirring to an aqueous solution of ZrO(NO3)2⋅x H2O (5.77 g, 99 %, Aldrich) and La(NO3)3⋅6 H2O (1.01 g, 99.9 %, Aldrich) in water (250 mL). The metal solution comprising the protected Ru nanoparticles was added dropwise into an aqueous solution of NaOH (220 mL, 2.5 M). The obtained precipitate was stirred for 2 h and then filtered. In this case, the precipitate was washed by suspending the solid into a NH4NO3/NH4OH buffer solution (200 mL, pH 10), in order to remove the bromide ions involved in the synthesis. Finally, the solid obtained was suspended in 2-propanol (300 mL) and heated at reflux for 4 h in order to stabilize the texture of the oxide matrix.47, 48 After filtration, the solid was dried at 120 °C overnight and calcined in a static oven at 500 °C for 5 h with heating and cooling rates of 1 °C min−1.
For comparison, reference impregnated samples Ru/LSZ were prepared. The LSZ supports were synthesized using the already described procedures, without the Ru nanoparticles. After calcination at 500 °C for 5 h, the metal was deposited by conventional impregnation, by dispersing the support into an ethanol solution containing the appropriate amount of RuCl3⋅x H2O. After stirring for 1 h and evaporation of the solvent under reduced pressure, the samples were dried at 120 °C overnight and finally calcined at 400 °C for 5 h (heating rate=3 °C min−1).
Catalytic activity evaluation
Catalytic experiments were conducted in a U-shaped 8 mm ID quartz reactor under atmospheric pressure. Pure ammonia (transistor grade purity) was used as reactant. The mass of catalysts and the gas flow rates were adjusted in order to test the samples under different conditions: GHSV=4000 mL g−1 h−1 (250 mg of catalyst and 16.1 mL min−1 of NH3) and GHSV=30 000 mL g−1 h−1 (100 mg of catalysts and 51.0 mL min−1 of NH3).
Before any experiment, the calcined materials were activated by reduction in H2 flow (35 mL min−1) at 800 °C for 3 h. After reduction, H2 was removed by Ar flow (35 mL min−1) for 30 min and the temperature was decreased to 250 °C. Then, NH3 was introduced into the reactor, and the system was stabilized for 30 min. The catalytic activity was evaluated every 50 °C from 250 to 800 °C. The gaseous mixture was analyzed three times at each temperature. The catalyst was subjected to two activity cycles to verify the stability after the reaction at high temperature. For the stability test, after activation by H2 reduction at 800 °C, the samples were cooled at 500 °C under Ar flow (35 mL min−1). After stabilization of the temperature, NH3 was introduced into the reactor and the gaseous mixture released by the reactor was analyzed every hour.
On-line GC analysis was performed with a Shimadzu 2010 gas-chromatograph equipped with a TCD (AGC Instruments) equipped with Au filaments to avoid the corrosion by NH3. Two packed columns with Ar as carrier were used. A Molsieve 5 A column was used for the separation of H2 and N2 while a Porapak Q column was used for the separation of NH3 from permanent gases (H2 and N2). The analysis of the gases was performed by the alternative injection of the mixture into the columns by use of a 10-way valve equipped with 2 loops.
The NH3 conversion was calculated on the bases of the following reaction:
2 NH3→N2+3 H2
The BET surface areas, pore volumes, and average pore diameters of the catalysts were measured by N2 physisorption at liquid- nitrogen temperature by using a Micromeritics ASAP 2020. Approximately 200 mg of the catalyst, previously degassed at 350 °C for 18 h, were used for each analysis.
The metal dispersion was estimated by static H2 chemisorption at 100 °C by using a Micromeritics ASAP 2020 C instrument according to the indication reported by Okal et al.37 Prior to the H2 adsoprtion, the catalysts were reduced at 800 °C for 3 h flowing H2 (35 mL min−1) and degassed at 800 °C for 5 min and at 400 °C for 3 h. The H2 adsorption isotherms were recorded in the 3.3–53.3 kPa pressure range with an equilibration time of at least 1 h for each point. The total H2 adsorption was evaluated in the first isotherm. After that, the sample was evacuated at 100 °C for 5 min and a second isotherm was recorded. The chemisorbed H2 was determined by extrapolation to zero pressure of the linear part of the adsorption isotherm after removing the so called reversible hydrogen adsorption (“double isotherm” procedure) and assuming a chemisorption stoichiometry of H/Ru=1:1.
Hydrogen chemisorption experiments in dynamic conditions were performed at 100 °C by using a ChemBET Pulsar Automatic Chemisorption Analyzer from Quantachrome Instruments. Before H2 adsorption measurement, the catalyst (150 mg) was placed in a tubular quartz reactor and pretreated by drying at 450 °C in He for 1 h, then reducing in 5 % H2/Ar at 800 °C for 2 h, evacuation at the same temperature in He for 1.5 h, and finally cooling to 100 °C in He. The pulse adsorption of hydrogen was monitored by using a thermal conductivity detector (TCD). To ensure optimum sensitivity, nitrogen was used as carrier gas. Prior starting the analysis, the TCD detector was stabilized in 75 mL min−1 N2 flow for 2 h at 100 °C. Then, pulses of pure H2 (73 μL) were introduced until saturation. Metal surface area and average crystallite size were calculated assuming the stoichiometry H/Ru=1:1.
Powder X-ray diffraction (XRD) patterns of the samples after various treatments were recorded by using a computer-controlled Philips X′Pert diffractometer using CuKα radiation (λ=0.15406 nm). The data were collected in the 2 θ range 10–100° with steps of 0.02°. The Rietveld analysis were performed using the software PowderCell. Owing to the complexity of some samples, the mean crystallite sizes were calculated applying the Scherrer’s equation to the main reflection of each phase.
Temperature programmed reduction (TPR) experiments were performed on approximately 35 mg samples of the calcined materials. The samples were pretreated at 400 °C for 1 h by pulsing O2 in an Ar flow every 75 s, then purged with Ar at 400 °C for 15 min and cooled to room temperature. H2(5 %)/Ar was admitted into the reactor and the flow was allowed to stabilize for 30 min, before increasing the temperature to 1000 °C with a heating rate of 10 °C min−1. The H2 uptake was monitored using a thermal conductivity detector (TCD).
Temperature programmed surface reactions (TPSR) were conducted in a home-made flow apparatus by using a mass spectrometer Hiden HPR20 as analyzer. In a typical NH3-TPSR experiment, the sample (ca. 0.25 g) was loaded in U-shaped quartz microreactor. The samples were reduced in flowing H2 (35 mL min−1) at 800 °C for 3 h. After this treatment, the adsorbed H2 was removed by purging the system with Ar at 800 °C for 30 min. Afterwards, the samples were cooled at 110 °C under the inert gas flow. For NH3 adsorption, the reduced samples were saturated flowing NH3(10 %)/Ar (50 mL min−1) at 110 °C for 30 min. After NH3 adsorption, the sample was flushed in Ar flow at 110 °C for 1 h to remove physically adsorbed NH3. The NH3 TPSR profile for each sample was recorded by increasing the temperature from 110 to 800 °C at a heating rate of 20 °C min−1 under a flow of Ar (50 mL min−1). The desorbed products were analyzed by means of a mass spectrometer operating in the electron impact mode with a ionization energy of 35 eV. The desorbed species were identified on the basis of the intensity of various mass fragments: m/z=2 for H2, m/z=28 for N2 and m/z=16 for NH3 (the parent peak, m/z=17, is superimposed with the cracking pattern of desorbed water).
Prof. M. Graziani (University of Trieste) is gratefully acknowledged for the kind collaboration and the helpful discussion. BL thanks Acta SpA for financially support her Ph.D. position. University of Trieste, Acta SpA, ICCOM-CNR, INSTM, CENMAT and PRIN2007 “2nd generation sustainable processes for the hydrogen production from renewable sources” are gratefully acknowledged for financial support.