Tailor-made preparation of Co–C, Co–B, and Co catalytic thin films using magnetron sputtering: insights into structure–composition and activation effects for catalyzed NaBH4 hydrolysis

Themagnetron sputtering (MS) methodology is a powerful tool for tailor-made fabrication of Co-based thin film catalysts with controlledmicrostructures and compositions for sodium borohydride (SBH) hydrolysis. In particular, Co–C catalysts were tested in this reaction and compared to Co–B and Co catalyst coatings. The microstructural and chemical analyses by X-ray diffraction (XRD), scanning and transmission electron microscopy (SEM and TEM), Rutherford back scattering (RBS) and X-ray photoelectron spectroscopy (XPS) were used to characterize a complete library of thin film catalysts. Pure Co materials were characterized by their nanocrystalline microstructure, and grain refinement was achieved via an increase in the deposition pressure. The incorporation of boron or carbon via co-deposition results in amorphization and dispersion of the active metallic Co phase. The composition can be tuned while keeping a controlled microstructure, and a comparison of activity at 25 C was performed on catalysts deposited on Ni foam substrates. A comparison of the initial activities showed that the Co–B samples were more active than the Co–C samples because of electronic effects. However, a strong activation was found for the Co–C catalysts after the first use. This effect was dependent upon the incorporation of cobalt boride (CoxB) species on the catalysts' surface, as shown by XPS. After the first several uses, the activity of the Co–C samples (values up to 2495 mL min 1 gcatalyst ) were as high as that of fresh Co–B, and the surface composition of both the catalysts was similar. This activation was not observed for the pure Co and was very weak for the Co–B catalysts. The use of polymeric (PTFE) substrates (flexible membranes) illustrated the versatility of the methodology to obtain catalytic membranes and allowed for a TEM microstructural analysis at the nanoscale. Catalytic activities at 60 C were as high as 16.7 and 20 L min 1 gCo 1 for the Co–C and Co–B membranes, respectively. We determined the optimized conditions to increase the catalytic activity of Co-based coatings prepared via magnetron sputtering.


Introduction
The search for applicable hydrogen storage materials is important because of the advantages of hydrogen energy. 1 The catalyzed hydrolysis of lithium and sodium borohydride, ammonia borane, hydrazine, and formic acid have been extensively investigated as promising hydrogen storage materials based on their relatively high hydrogen contents and low molecular weights. In particular, the catalyzed hydrolysis of sodium borohydride (SBH) 2-4 is a safe reaction that produces hydrogen according to eqn (1): Co-based catalysts have been widely investigated for this reaction, and the ammonia borane hydrolysis, as cost-effective materials [5][6][7][8] with the aim of replacing the more expensive and scarce noble metals catalysts. Pure Co, Co-B, Co-B-O, Co-P and bimetallic Co-based materials have been widely investigated. [5][6][7][8] Recently, metal-carbide catalysts have been proposed as alternative non-precious materials for fuel cells or electrolyzers in electrochemical energy conversion devices. 9,10 A volcano plot was presented for the different metals and metal carbides for their use in hydrogen evolution reactions (HER). 9 For technological/practical applications, i.e., systems and reactor designs, it is highly desirable to have catalysts in a supported form. 3 In a previous paper, 11 we proposed the magnetron sputtering (MS) deposition technique as a method to prepare pure Co catalysts supported on different substrates for the catalyzed hydrolysis of SBH and ammonia borane. MS is a technique currently used in industry that permits the deposition of a wide range of materials with varied compositions, microstructures, crystallinities or porosities on varied supports by varying the target composition and deposition conditions (nature and pressure of discharge gas, deposition power, temperature of substrate, etc.). [11][12][13][14] Supported Co catalysts have mainly been prepared using the impregnation-chemical reduction method. In most cases, SBH is employed as a reducing agent to form the supported Co-B materials. 15,16 Pure Co catalysts could be obtained for fundamental studies with boron-free targets using MS. Electrodeposition has also been presented as a convenient scale-up method for the production of supported Co catalysts. 17,18 However, the use of this method was limited to metallic (or bimetallic) lms and can only be used for conductive substrates. Pulsed laser deposition has also been used for the preparation of supported catalysts for SBH and ammonia borane hydrolysis. Co-B thin lms and Co nanoparticles on a B lm have been prepared using this method. 19,20 MS deposition has been used to fabricate pure Co catalysts supported by non-conductive substrates for water oxidation 21 and for the SBH hydrolysis reaction. 11 In the present work, we studied Co-C catalysts and compared them with Co and Co-B catalysts, and all the catalysts were prepared using MS. The superior adhesion of these coatings on porous substrates (Ni foam and polymeric membranes) is an additional advantage of the MS deposition in this work.
There is still controversy regarding the nature of the active phase(s) and the role of boron and/or microstructure in the activity and durability of these Co-based catalysts. [5][6][7][22][23][24] Even though the Co x B phase has been widely proposed as active phase (also by its in situ formation at the surface in the presence of SBH 23 ); a Co core with a (poly)borate shell was proposed as well. 6,18 The tailor-made catalytic coatings presented here were used in a set of experiments to expand the knowledge on these issues. Supported catalysts with a controlled microstructure and/or composition were all prepared using the same experimental set-up and were investigated and compared to other reported catalysts in the literature.

Catalyst preparation and selection of materials and supports
On the basis of our previous work for pure Co in ref. 11, the deposition parameters selected for the coating fabrication in this paper are summarized in Table 1 with the deposition rates and thicknesses.
The coatings were prepared via co-deposition, and two magnetron sources from AJA (N Scituate, MA, USA) were employed for 2 00 diameter targets. One was operated in a magnetic target conguration under DC power for a pure cobalt target (Kurt J. Lesker, 99.95% pure, 1 mm thick). The second one was operated when appropriate under an RF conguration for pure boron (Kurt J. Lesker, 99.95% pure) or carbon (Kurt J. Lesker, 99.99% pure) targets. The base pressure before deposition was 10 À6 mbar, and the working pressure for Ar was 2.8 Â 10 À2 mbar for all the coatings, except for one Co sample grown at 4.5 Â 10 À2 mbar (named hp to indicate high pressure). Table 1 also summarizes the abbreviated names used for all the coatings, which were based on the target compositions, pressure and applied discharge power.
For the catalytic studies, the thin lms were deposited on a commercial Ni foam (Goodfellow, 1.6 mm thick, 95% porosity, 20 pores per cm) and on PTFE (polytetrauoroethylene) membranes (Pall Corporation, PTF002LH0A-SAMP, 0.02 mm pore size, polypropylene backed). For the chemical and microstructural characterization of the coatings, Si (100) pieces were employed as substrates in addition to the Ni foam and PTFE membranes. Prior to deposition, the Ni foam was cut into ca. 0.5 Â 0.5 cm 2 pieces and placed in groups to be used in a small reactor (see Section 2.3). Each group was weighed before and aer deposition to obtain the total mass of the deposited catalyst. Before each synthesis, the Ni foam pieces were washed in an ultrasonic bath successively with distilled water, ethanol/ acetone (1 : 1), 0.1 M HCl, distilled water, ethanol and acetone, and then it was dried in air for 24 hours. No previous treatment was performed on the PTFE membranes. The Si substrates were cleaned with acetone and dried in a nitrogen ow. When the support was different from the Ni foam, it will be clearly indicated in the text.

Catalyst characterization
X-ray diffraction measurements were performed using Cu Ka radiation in a Siemens D5000 diffractometer in a Bragg-Brentano conguration. The coatings grown on the PTFE membranes were used for these measurements.
Scanning electron microscopy (SEM) was used to study the morphology and microstructure of the samples in a high resolution FEG microscope HITACHI S4800. The coatings were analyzed directly on the Ni foam and PTFE membranes for top view observations. For the cross section views, samples were cleaved from coatings grown on Si. The thickness of the coatings was evaluated using the SEM cross-section measurements.
TEM (Transmission Electron Microscopy) and ED (electron diffraction) studies were performed on a FEI Tecnai G2 F30 FEG (eld emission gun) microscope equipped with a HAADF (High Angle Annular Dark Field) detector from Fischione Instruments. Images were obtained in STEM (scanning TEM) mode at 300 kV with HAADF detection. The thin lm samples were grown on PTFE membranes. They were removed from the polypropylene back-support, xed in a frame grid, and then submitted to ion thinning from the back side.
Proton-elastic back-scattering spectrometry (p-EBS) was used to determine the composition of the catalytic coatings grown on the Si substrates. This technique is well suited for characterization of the coatings as it provides simultaneous determination of Co and the light elements C, O and B. 25 Measurements were carried out at the National Center for Accelerators (CNA, Sevilla, Spain) using a 3 MV tandem accelerator. The spectra were obtained using a 2 MeV H + beam and passivated implanted planar silicon (PIPS) detector set at 165 . To obtain the thickness and composition of the lms, the spectra were simulated using the SIMNRA code. 26 XPS spectra were recorded using a SPECS electron spectrometer equipped with a PHOIBOS 150 9MCD analyzer using Al Ka radiation with a 35 eV pass energy at a normal emission take off angle. The spectra were calibrated to the position of the C (1s) signal (from contamination) at 284.6 eV and validated using the characteristic position of the surface oxidized Co (2p 3/2 ) at 780.9 eV. Deconvolution was performed using the CasaXPS program provided by the SPECS Company. The coatings grown on the PTFE membranes were used for these measurements. The samples were investigated aer deposition and aer use for the SBH hydrolysis reaction.

Catalytic activity: hydrogen generation tests
The Ni foam pieces were grouped (10 to 15 pieces), cleaned and weighed before and aer deposition, as described in Section 2.1. The pieces with the supported catalyst (1-6 mg) were placed at the bottom of a three necked heart-shaped ask. The ask was immersed in a water bath maintained at 25 AE 0.5 C and connected to a 100 mL gas burette. The reaction (1) was started by injecting 38 mg of SBH dissolved in 1 mL of a 4.5 wt% NaOH solution. The amount of hydrogen generated was measured by reading the displacement of the piston in the gas burette (gastight with a mercury o-ring) as a function of time. No additional stirring was used for the experiments, except for the stirring created by the evolving hydrogen. The HGR (Hydrogen Generation Rate, mL min À1 ) was obtained from the slope of a plot of the volume of hydrogen evolved vs. time in a linear regime. In this paper, the catalytic activity (expressed in mL min À1 g cat À1 ) was the slope from the plot of HGR (mL min À1 ) as a function of the mass of the supported catalyst. The activity was also expressed per gram of cobalt considering the composition of the catalysts determined using p-EBS spectrometry. Finally, the catalytic activities were measured for selected samples at different temperatures, and the Arrhenius plots were analyzed to determine the apparent activation energies. Cycling experiments were also conducted. Aer each test, the supported catalyst was extracted from the reaction medium, washed with distilled water and ethanol and then dried for one day under atmospheric conditions before the activity test was repeated.
The activities of the samples supported on the PTFE membrane were tested at 60 C (in a thermostatic bath) using the same experimental set-up described for the Ni foam samples. Typically, 4 pieces of 1 cm 2 each were cut into small pieces and placed in the reaction ask. The procedure was similar to the one described above except the mass of the catalyst was not determined by weight. Activity was expressed per gram of cobalt. For Co quantication in the membrane deposited catalysts, the colorimetric nitroso-R salt (NRS) method 27 was used. A 1 cm 2 piece of the supported catalyst was treated with 1 M HCl for complete dissolution of the Co-based catalysts. The amount of Co determined using the photometric method was evaluated by comparing the results with the results obtained using plasma ICP measurements for selected samples. We found that the colorimetric method agreed with the ICP results with less than 12% error.
The catalysts supported on the PTFE membranes were more convenient for XPS analysis (thinner and atter samples compared to the Ni foam substrate). The samples were investigated aer deposition and aer use. In the latter case, the catalytic surface of the membranes was kept in contact with a 19 wt% SBH solution in a 4.5 wt% NaOH solution for 1.5 or 24 hours. The samples were rinsed with water and dried under atmospheric conditions prior to XPS analysis.

The supported catalysts and their microstructural and chemical characterization
The following samples (see Table 1) were prepared starting with 50 W of DC power at the Co target: two pure Co catalysts at medium (2.8 Â 10 À2 mbar) and high (4.5 Â 10 À2 mbar) pressures of the process gas (Ar), and two Co-B catalysts using 50 and 100 W RF power at the metalloid (B) target with medium process gas pressure. Four samples were prepared with different combinations of power at the Co and C targets for the newly investigated Co-C catalysts. The medium Ar pressure was used when applying the DC and RF power at the Co and C targets in all of the samples.
The rst characterization of the samples was performed using XRD, and the data are shown in Fig. 1 for the catalysts supported on the PTFE membranes. The pure Co 50 materials show small and broad peaks that can be assigned to a nanocrystalline hcp Co phase (ICDD 00-005-0727), but the cubic phase (ICDD 00-015-0806) cannot be disregarded. 17 As we demonstrated in our previous publication, 11 higher deposition powers are needed for higher crystallization of the cobalt phase. The incorporation of B and C results in amorphization of the catalysts at similar deposition powers. Based on research with powder materials, the presence of boron is generally associated with the formation of amorphous black solids. 5,6 In previous work, nanocrystalline and amorphous phases appeared to be more active catalytically. 5,6 In the present work, the addition of carbon produces a similar amorphization effect on the microstructure. Only when the highest deposition powers were used was a certain degree of nanocrystallization observed in Co 150 C 300 . Data obtained from the electron diffraction (ED) analyses are shown in Section 3.3.
To gain additional microstructural information, Fig. 2 and 3 show the SEM top and cross sectional views of the coatings grown on Si substrates for all of the investigated samples. The SEM analysis also permits determination of the thin lm thickness and deposition rate (Table 1). In general, all catalysts have a columnar growth, which is typical of magnetron sputtered deposited thin lms under low adatom mobility conditions, and surface shadowing governs the lm growth. 28 The top views and details of the columns, presented in Fig. 2 and 3, reveal a microstructure of mesocolumns (between nano and micro), which are typically formed via the association of nanocolumns. 29 A certain degree of intra-and inter-columnar porosity is also observed. The mesocolumn sizes were evaluated based on the direct observation of the columns in the cross sectional views and the grain sizes measured from the top view images. The results are summarized in Table 2. In the case of the two nanocrystalline, pure Co samples (Co 50 and Co 50 -hp), a reduction in the mesocolumns/grain sizes was found for the sample growth under high pressure (clearly shown in the top view images). For Co-B and Co-C, the mesocolumns/grain sizes were similar to the ones for Co 50 -hp. It was concluded that Co 50hp, Co-B and Co-C samples have a similar microstructure and grain and mesocolumn sizes at the selected deposition conditions. Only the Co 50 sample showed a clear increase in the grain/mesocolumn size. The use of high pressure during the magnetron sputtering for pure Co deposition produced grain renement, as described in our previous paper. 11 The introduction of B and C produces the same effect through enhanced amorphization. The Co-B and Co-C samples are directly comparable as they have a similar amorphization degree (XRD data) and similar thicknesses and mesocolumn sizes (Tables 1  and 2). This allows for a new fundamental understanding of the role of metalloid addition through comparison to pure Co samples with similar and controlled microstructures.
The composition determination for the Co, Co-B and Co-C coatings is crucial to quantify the incorporation of B and C in the samples and the oxidation degree. Proton-EBS is well established for these analyses of thin lms and coatings and for simultaneous quantication of light elements. 25 Table 2 Fig. 1 X-ray diffraction patterns for the as prepared Co, Co-B and Co-C coatings. The PTFE substrate is included as a reference. summarizes the results obtained for the samples grown on Si substrates. For the Co-C catalysts, the samples studied cover Co/C atomic ratios ranging from 6.1 to 1.9. The measured values for Co 50 B 100 are compatible with the typical stoichiometries previously reported for Co-B powders (i.e., Co 3 B). 30 An additional composition is given for a lower B content. The presence of a certain amount of oxygen in the Co deposited coatings is unavoidable because of the presence of oxygen and water in the residual vacuum. 31 It is also characteristic for columnar coatings to have a certain degree of inter-columnar porosity, which also favors a post-deposition oxidation. 32 Nevertheless, the presence of oxygen is a characteristic of powdery Co-based catalysts for this reaction. 5,6,23 In conclusion, the tailoredmade coatings can be used as a representative library for these types of catalysts. Fig. 4 shows the XPS surface analysis performed on selected, representative, fresh samples deposited on the PTFE membranes (Co 50 -hp, Co 50 B 100 and Co 50 C 100 ). The Co (2p) binding energies at 777.8 (Co 2p 3/2 ) and 792.9 eV (Co 2p 1/2 ) are consistent with those of metallic cobalt, 5,23,33,34 cobalt borides (Co x B) 35,36 or cobalt carbides, 37 which cannot be clearly differentiated using XPS analysis. Oxidized species are also detected on the surface, corresponding to Co (2p) peaks at higher binding energies of 780.9 (Co 2p 3/2 ) and 796.4 eV (Co 2p 1/2 ). 5,23,34 The shape of the peaks and satellites are compatible with the CoO phase. 23,34 The B (1s) peak in the Co-B sample shows two peaks at binding energies of 187.6 and 191.6 eV. The former peak at 187.6 eV can be assigned to elemental boron or boron in Co x B compounds, which is in agreement with literature data. 33,36,38,39 The second peak at 191.6 eV can be assigned to oxidized boron species at the surface. 5,22 For the Co-C sample, the analysis of the C (1s) peak is not presented because of strong overlapping with the adventitious carbon peak, which was a result of contamination and air exposure. For oxygen, the O (1s) peak corresponded to the characteristic Co-OH and Cooxide species at the surface. 40 The versatility of magnetron sputtering has been demonstrated for catalyst design. Composition and microstructure can be controlled while the catalysts grow directly on the substrates of interest.

Catalytic activity: a comparative study
The activity for the SBH hydrolysis reaction was measured for each prepared Ni foam supported catalyst in 4.5 wt% NaOH and 3.8 wt% SBH solutions. The results are summarized in Table 3. Fig. 5 shows the hydrogen evolution curves (Fig. 5a), and hydrogen generation rate plot (Fig. 5b) as a function of the catalyst mass for the Co 50 C 100 sample measured at 25 C. For all of the tested catalysts, the hydrogen evolution curves followed a straight line, which indicated zero-order kinetics for the SBH concentration. No induction periods were detected in the   Fig. 5c for the Co 50 C 100 catalyst.
In previous work with Co and Ni based solids, 6,41 a certain activation of the catalysts upon rst uses (approximately 10% from the initial activity) was found. In addition, a repetitive cycling produced a progressive deactivation at the end. 6,42 For these reasons, each catalyst was cycled up to 7 times to determine the initial and maximum catalytic activity. Table 3 summarizes these values expressed in mL min À1 per gram of catalysts and per gram of cobalt (according to the stoichiometry given by p-EBS) for all of the investigated samples. Long life durability experiments will be presented in a future study. The data corresponding to the rst 7 cycles are included as ESI (Fig. S1 †), and the data show the initial activation effects and tendency towards deactivation aer cycling. Fig. 6 represents the catalytic activity data (both initial and maximum in mL min À1 g Co À1 ) as a function of the Co content (at%) for the investigated materials indicated in the gure.
As discussed above, the prepared Co 50 -hp, Co-B and Co-C samples are comparable in terms of microstructure and grain and mesocolumn size. We found that the initial activity of the Co 50 -hp sample is on the order of the Co 50 B 50 activity, as shown in Table 3. An increase in the activity is observed if the amount of boron is increased, which occurs with the Co 50 B 100 . This is consistent with the idea that the presence of boron is crucial for the high activity of the Co catalysts. 5,6 In the presence of boron, cobalt is electron-enriched and benets catalytic activity apart from increasing the dispersion of the metallic phase. 5,6 This hypothesis is conrmed because the Co-C catalysts show low initial activities with respect to the pure Co and Co-B samples. The addition of carbon increases the cobalt dispersion but no electronic effect was found in this case. However, a strong activation (up to a 220% increase in the activity with respect to the initial) is observed for the Co-C samples aer the rst uses (transition from points in a blue ellipse to points in a green ellipse in Fig. 6, see also Fig. S1 †). The activity of the Co-C samples aer use can be as high as that of Co-B (2495 vs. 2685 mL min À1 g cat À1 for Co 100 C 300 and Co 50 B 100 , respectively). This effect is not observed for the pure Co catalysts and is very weak for the Co-B samples. Considering the maximum activities for the Co-C catalysts, there is a correlation between the activity  Activity initial (mL min À1 g Co À1 ) Activity maximum rst 5 cycles (mL min À1 g cat À1 ) Activity initial (mL min À1 g cat À1 ) Co 50  87  96  1667  1667  1600  1600  Co 50 -hp  87  96  2083  2083  2000  2000  Co 50 B 50  75  93  2204  2005  2050  1865  Co 50 B 100  65  90  2983  2674  2685  2407  Co 50 C 100  79  94  1648  1222  1594  1149  Co 50 C 200  64  92  2229  1179  1984  1050  Co 100 C 300  59  86  2901  1337  2495  1150  Co 150 C 300  70  91  2186  978  1990  890 a Thicknesses as indicated in Table 1. and cobalt dispersion. An improvement in the activity was found when the dispersion of Co increased and was mediated by the higher incorporation of carbon (4 points inside the green ellipse in Fig. 6). A similar behavior is observed for the two Co-B samples (2 points inside the red ellipse) although the effect is not accompanied by a signicant activation upon use. To understand the activation process occurring in the Co-C catalysts, an XPS analysis was conducted on a representative Co-C sample supported on PTFE membranes aer use (see details in the Experimental section). The results can be found in Fig. 7 for the B (1s) region for as prepared samples and samples aer use.
The surface of the Co-C catalyst aer use shows the two characteristics peaks of the B-O (191.6 eV) and B-Co (187.6 eV) species. In fact, the surface becomes similar to that of the Co-B catalysts (see Fig. 4), which explains the activation of the Co-C material during the rst uses. The presence of borate reaction products on the surface of the catalysts aer use is well known; 5,6 however, the incorporation of boride was observed in our experiments for the case of the Co 50 C 100 sample, which showed a clear activation effect upon rst use. These results reinforce the idea that there is an electronic effect of the boride to improve catalytic activity in addition to a dispersion effect. An electronic effect coming from B incorporation as boride was also previously proposed for the Co-B black powder, 5,43 which is generally obtained via the reduction of oxidized Co with SBH. The nanocrystalline pure cobalt samples are also highly active (orange ellipse in Fig. 6). However, the lower dispersion does not seem to favor incorporation of B in the form of cobalt borides in this case, and the samples do not show strong activation during the rst uses (see data in Table 3 and Fig. S1 †). In agreement with our previous paper, ref. 11, we found that an increase in the Ar deposition pressure produced an increase in catalytic activity for the Co coatings prepared at the same DC deposition power of 50 W. This effect is associated 11 with a higher exposed surface due to a smaller mesocolumn size. It   was then assumed that a further increase in the Ar pressure during MS deposition of pure Co coatings would improve the catalytic activity. However, the use of higher pressures produced powdery coatings 44 with a loss of adhesion. Based on this, the use of dispersant agents such as B and C is suitable for achieving grain renement without a loss of adhesion.
Our results identied two catalytic phases: the amorphous cobalt boride for the Co-B and Co-C (in situ formed) catalysts and the nanocrystalline Co particles (with their corresponding surface oxide layer) in the case of pure Co. In the pure Co catalysts, there was no evidence of the formation of Co x B species on the surface. However, the formation of surface cobalt borides upon operation cannot be completely disregarded. Cobalt borides are unstable and would transform into metallic cobalt (which would then become oxidized) and borates in aqueous solutions during reaction quenching. One possible interpretation is that cobalt borides do form when in contact with SBH on the surface of any cobalt precursor (even in the form of oxides) and then decompose during the reaction quenching. 45 This would explain the small amounts of borides in comparison to the borates detected using XPS. The catalytic activity would correlate with the amount of cobalt boride phase formed in situ. 45 The highly dispersed (and thus more reactive) Co-C should incorporate enough surface cobalt boride to be detected aer quenching. This result shows that a high dispersion of cobalt is required for high catalytic activity, and the nature of the dispersant atom (B or C) is indistinct aer the rst uses. We suggest that in operando studies would help clarify these points.
The apparent activation energies (in the 15-35 C range) were 54, 44, 57 and 53 kJ mol À1 , for the Co 50 , Co 50 -hp, Co 50 C 100 and Co 50 B 100 samples, respectively, which correlated with the activity trend at 25 C.
The activity of the prepared samples can be compared to reference materials in the literature. For the Co-B powdery materials, the typically reported activities (at 20-25 C) are in the range of 400-3350 mL min À1 g cat À1 , depending on the composition and microstructure. 7 Our results (i.e., 2685 mL min À1 g cat À1 for the Co 50 B 100 sample) are in the same order of magnitude. However, new advances in the dispersion of Co-B nanoparticles on silica substrates 46 or in the design of new alloys, including P and Cr or Mo, as well as B, 7,47 have resulted in higher activities (up to 30 800 mL min À1 g cat À1 ). Some of these alternative nanostructures and compositions could be implemented using MS in multi-target chambers in future work and possibly prove the dispersion and electronic effect mechanism proposed here.

Supported catalysts on exible polymeric membranes
To illustrate the versatility of the deposition methodology, we studied polymeric (PTFE) membranes as substrates for the catalytic coatings. Flexible Teon membranes (commercially available) consist of expanded bers, as shown in the SEM image in Fig. 8. Catalyst depositions were conducted under the conditions reported in Table 1 but with shorter deposition times to reduce the coating thickness (Data are included in Table 4.). As shown in the SEM image in Fig. 8 for the Co 50 C 100 sample, a good coverage of the polymer bers was achieved while keeping the characteristic open pores of the membrane structure. A good adhesion while also maintaining the membrane exibility was also obtained for all of the investigated Co-based catalysts. The TEM microstructural analysis at the nanoscale is shown in Fig. 9 (STEM-HAADF images) for samples Co-hp, Co 50 B 100 and Co 50 C 100 . At this magnication, we were able to visualize the nanocolumns that compose the mesocolumns. The sizes of the nanocolumns range from 4 to 12 nm, and no signicant differences were found among the three samples, as expected from the SEM analysis. The electron diffraction patterns (insets in Fig. 9) clearly show the formation of nanocrystals in the pure Co sample, whereas alloying with B or C produces amorphization. For the Co-hp sample, rings were identied that corresponded to the interplanar distances of 2.17, 2.06 and 1.93Å, which were assigned to the (100), (002) and (101) planes of metallic hcp Co, respectively (ICDD 00-05-0727) (The cubic phase cannot be completely disregarded.). In addition, rings at spacings of 2.46 and 1.51Å were measured and assigned to the (111) and (220) planes of the CoO phase (ICDD 00-0481719) (The Co 3 O 4 phase can also form.). The surface oxidation of metallic Co is expected and is more important in the thin samples used for TEM analysis because additional crystallization can occur under the 300 kV electron beam.
The catalytic activities at 60 C are presented (in liters per minute and per gram of Co) for selected coatings in Table 4. For practical applications, taking into account the exothermic character of the SBH hydrolysis reaction, it is common in reference papers to report the activity at 60 C. For the Co-C and Co-B catalytic coatings on porous polymeric membranes,  activities as high as 16.7 and 20 L min À1 g Co À1 were measured, respectively. The nanostructured character of the coatings associated with the dispersion and electronic effect of the boron lead to very thin lms (100-600 nm) with activities in the range of the highest reported values for catalysts with similar compositions. 7

Conclusions
The magnetron sputtering deposition methodology is presented here as a versatile tool for the fabrication of Co-based supported catalysts for hydrolytic hydrogen generation from SBH. In addition to the adaptability of the technique to support catalysts on porous, metallic or polymeric substrates, we have shown its capacity to create tailor-made catalysts by controlling both the microstructure and composition. Novel Co-C catalytic coatings were investigated and compared to Co-B and pure Co. The composition can be controlled by using a dual head MS deposition chamber and tuning the power applied to the Co and the metalloid (B or C) targets. The microstructure can be controlled by increasing the deposition pressure, which produces grain renement for the nanocrystalline pure Co coating. The codeposition of C or B also induced grain renement through amorphization and dispersion of the Co active phase. A library of Co-C, Co-B and Co catalysts deposited onto Ni foam were tested for their SBH catalytic hydrolysis reaction activity. We determined the optimized conditions to increase the catalytic activity of Co-based coatings prepared using magnetron sputtering. We were able to prepare a series of samples with comparable microstructures, grain and crystal sizes to study the effect of the addition of carbon and boron to pure Co samples. We found a correlation between the degree of dispersion in the active Co phase and the activity for the Co-C and Co-B catalysts, as well as an electronic effect of B in the Co-B catalyst that improved catalytic activity. The Co-C catalysts showed evidence of in situ formation of Co x B on their surfaces, which explains the increase in activity aer the rst cycles (up to 220% with respect to the initial). Nanocrystalline pure Co and amorphous Co x B were proposed as catalytic phases.
The versatility and potential of this methodology was shown by the fabrication of highly active catalytic membranes, which was performed by depositing thin lms on exible polymeric substrates.