Loïc
Assaud‡
*a,
Johannes
Schumacher§
a,
Alexander
Tafel
a,
Sebastian
Bochmann
a,
Silke
Christiansen
bc and
Julien
Bachmann
*a
aFriedrich-Alexander University Erlangen-Nürnberg, Department of Chemistry and Pharmacy, Egerlandstrasse 1, 91058 Erlangen, Germany. E-mail: loic.assaud@ensiacet.fr; julien.bachmann@fau.de
bMax Planck Institute for the Science of Light, Günther-Scharowsky-Strasse 1, 91058 Erlangen, Germany
cHelmholtz-Center Berlin (HZB), Hahn-Meitner-Platz 1, 14109 Berlin, Germany
First published on 11th March 2015
We establish a procedure for the fabrication of electrocatalytically active, nanoporous surfaces coated with Pt and exhibiting a high geometric area. Firstly, the mechanism of the surface reactions between platinum(II) acetylacetonate and ozone is investigated by piezoelectric microbalance measurements. The data reveal that ozone oxidizes the metallic Pt surface to an extent which can exceed one monolayer depending on the reaction conditions. Proper reaction parameters yield a self-limited growth in atomic layer deposition (ALD) mode. Secondly, the ALD procedure is applied to porous anodic oxide substrates. The morphology and the crystal structure of the deposits are characterized. The ALD coating results in a continuous layer of Pt nanocrystallites along deep pore walls (aspect ratio 70). Thirdly, the Pt/TiO2 surfaces are shown to be electrochemically active in both acidic and alkaline media, in a way that qualitatively conforms to literature precedents based on Pt. Finally, we apply the anodization and ALD procedure to commercial Ti felts and demonstrate systematically how the electrochemical current density is increased by the large specific surface area and by the presence of the catalyst. Thereby, the catalyst loading, as well as its efficient utilization, can be optimized accurately. The preparative approach demonstrated here can be generalized and applied to the various electrocatalytic reactions of energy conversion devices.
A flurry of activity has been dedicated to nanostructured positive and negative fuel cell electrodes lately, with the prospects of large surface area and potentially efficient usage of low noble metal loadings.7–9 However, direct comparison between related systems and realistic modelling have been rendered difficult by the limited level of control over the geometry of the electrode surface. Indeed, transport is usually modeled at simple geometries. In the extreme case of fast electrochemical reactions at an array of closely packed, parallel nanowires (convex) or nanotubes (concave), the galvanic current is hardly affected by the nanowire or nanotube length (and thereby by the electrode's geometric surface area), whereas capacitive effects are enhanced by them.10 The situation is reversed when the surface electrochemical reaction steps are slower than diffusion: in this case, the current can increase with the specific surface area, in some cases even linearly.11,12 This is why technically relevant electrodes of fuel cells are always porous in order to enhance the geometric surface area at which the slow multielectron transformations take place.13,14 The cases in which the electrochemical characteristics were determined experimentally for a systematically varying length of an arrayed nanostructure at the surface are rare.12 Thus, the development of preparative methods tailored to the accurate definition of nanostructured surfaces, the geometry of which is simple, well defined and tunable, is a crucial determinant of experimental advances. Some model electrode systems have been proposed (such as melt-pulled Pt wires in glass capillaries, electroless coated metals in polycarbonate templates, and elongated metallic lines prepared by thin film techniques), but they do not provide order on a large scale,15–19 so that the diffusion profiles overlap in an ill-defined manner.15 Conversely, a very accurate and tunable distance between neighbors can be achieved by conventional micromachining techniques, but with a limitation on the size of the structures produced.20,21
We propose an ordered nanoporous template such as that provided by anodic oxides to offer both advantages of order and small sizes simultaneously.22–28 In addition to the fundamental aspect of an accurately defined geometry and wide tunability of pore diameter and length, coating such oxide surfaces with noble metal catalyst particles delivers materials systems akin to classical supported catalysts.29 Such catalyst coatings have been performed by electrodeposition into the pores,30,31 and by atomic layer deposition (ALD) onto the pore walls.32–34 In addition to being inherently suitable to conformally coating narrow pores based on its self-limited surface chemistry,35,36 ALD provides the additional advantage that it is independent of the electrical properties of the substrate.
Thus, the present report establishes a preparative method for generating nanoporous Pt surfaces of accurately tunable geometry, demonstrates the effect of pore length on electrocatalytic current density at model electrodes systematically, and applies the methods to a substrate of technical significance. First, the growth mechanism of Pt ALD using a Pt(acac)2 + O3 reaction is investigated in situ by piezoelectric microbalance, and the material is characterized by X-ray diffraction (XRD), energy-dispersive X-ray spectroscopy (EDS), and scanning electron microscopy (SEM). Secondly, the ALD procedure is applied to a variety of anodized substrates (planar Al, planar Ti, and Ti fleece). Finally, the electrochemical performance of the catalytic systems is compared by cyclic voltammetry (CV) for the electrooxidation of ethanol and the hydrogen evolution reaction. Despite this being a model system fundamentally, we believe that further reductions in catalyst loadings are possible and could become of relevance to applied systems.
Fig. 1 shows the GCM growth rate determined in steady state in three different conditions, as a function of the Pt precursor dosage. Two sets of conditions demonstrate ALD (self-limiting) behavior, with a curve that approaches an asymptote at large dosage: the 150-degree growth with short exposure durations, and the 130-degree growth. The saturated growth rates are 0.8 and 0.6 Å per cycle, respectively, comparable to the value reported by Hämäläinen et al.38 The third set of conditions stands in stark contrast to the other results: when large exposure durations are used between the micropulses at 150 °C, the growth rate increases with dosage in a linear manner, inconsistent with self-limiting surface chemistry. The reason for these fundamentally distinct behaviors is provided by investigating the details of the GCM curves. In Fig. 2, the mass gain curves obtained at 150 °C with short and long exposure durations exhibit contrasting shapes. The commonality is that mass increases during the ozone pulse. This can only be rationalized if ozone oxidizes the metallic surface, as has been observed in similar conditions in the past.39,46 In the short exposure conditions, Fig. 2a, the Pt(acac)2 half-cycle results in a further mass uptake, consistent with the self-limited immobilization of one monolayer of it by the reactive, oxidized surface.47–49
The long exposure conditions (Fig. 2b) differ in that exposure to Pt(acac)2 results in large mass variations, and in an overall mass loss. Given that mass increases overall, Pt uptake from the gaseous precursor must be correlated with the loss of oxygen from the solid by decomposition of the platinum oxide to the metallic state. The absolute mass loss at this stage exceeds the mass gain over each cycle, therefore it cannot be due to combustion of the ligands alone. This indicates a ‘deep’ oxidation of the Pt surface to a metastable oxide, followed by its decomposition triggered by the Pt(acac)2 – again a precedented phenomenon.46,50–54
In other words, the ‘short exposure’ conditions do not leave sufficient time for the alternating transformations of the film between metallic and oxidic beyond the topmost layer, so that a well-behaved ALD growth ensues. In the ‘long exposure’ conditions, the film oscillates between states in which it is metallic and oxidic. In this case, the amount of reactive oxide, capable of immobilizing Pt(acac)2, is not under experimental control, which results in non-ALD growth. Note that this mechanism of non-self-limiting growth is fundamentally distinct from the simple thermal decomposition of the metal–organic precursor. In fact, we can exclude the thermal decomposition of Pt(acac)2 in these conditions by a control experiment conducted in the absence of O3, which yielded no growth.
The two contrasting cases of Fig. 2a and b are close to each other, as demonstrated by the GCM data recorded during nucleation in the ‘short exposure’ case, Fig. 3. During the first approximately 30 cycles in the ALD conditions, mass oscillations are observed reminiscent of those that prevail in the non-self-limited conditions. The growth then slowly reaches the well-behaved steady state. Such a nucleation stage is often observed when elemental metals are grown on oxide surfaces.55 The transition to steady-state growth can be associated with coalescence of the isolated nuclei to a continuous, more thermodynamically stable, film. This effect has been demonstrated previously.38
Fig. 3 Mass evolution measured by piezoelectric crystal microbalance during the first 65 ALD cycles of Pt carried out at 150 °C with 5 s exposure duration. |
At a deposition temperature of 130 °C, mass fluctuations are consistently observed at all exposure durations (Fig. 4), but with a smaller absolute amplitude. We interpret this observation as indicating the formation upon O3 reaction of an oxide species exclusively confined to the surface, which then reacts with a limited amount of Pt(acac)2 in a fast manner.
Fig. 4 Mass evolution measured by piezoelectric crystal microbalance during Pt deposition carried out at 130 °C. (a) Overall mass evolution, (b) enlarged view over one ALD cycle. |
The ALD reaction between Pt(acac)2 and ozone enables us to coat deep pores conformally in both sets of self-limiting conditions defined above. Fig. 5 shows SEM micrographs of a Pt deposit performed at 130 °C on anodic alumina (AAO) membranes featuring parallel pores of 200 nm diameter. The close-up view (Fig. 5a) shows near the opening of the pores that the Pt deposit is continuous but with a significant roughness. The low-magnification cross-section (Fig. 5b) demonstrates a conformal Pt deposit over 25 μm long pores (aspect ratio >100).
Depositions carried out at 150 °C using short metal exposure durations yield films with a larger roughness and the thickness of which decreases over depth in the pores, so that coating is limited to an aspect ratio of 50 (Fig. 6). Note that in freestanding membranes (with the barrier oxide layer removed), precursors can access more easily from both sides of the membrane, resulting in even coating at an aspect ratio of 70.
The crystalline structure of Pt films supported on AAO membranes was analyzed by X-ray diffraction. The diffractograms for both films grown at 150 °C and 130 °C are shown on Fig. 7. The three peaks located at 2θ values of 40.0°, 46.4° and 67.7° can be assigned to the [111], [200] and [220] reflexes of the fcc Pt lattice. This polycrystalline structure of Pt is likely the cause of the observed film roughness. Note that no crystalline platinum oxide is evident. Furthermore, no shift of the peaks with respect to the pure metallic reference (possibly due to incorporation of oxidized species) is apparent.56
The presence of Pt in the nanoporous samples is confirmed by energy-dispersive X-ray spectroscopic chemical microanalysis (peaks at 2.0, 9.5 and 11.0 keV in Fig. 9a). The X-ray diffractogram of the resulting Pt/TiO2 system (Fig. 9b) shows that the new substrate does not influence the polycrystalline nature of the Pt deposit. Peaks at 40.0°, 46.4° and 67.7° correspond to the [111], [200] and [220] fcc Pt reflexes. Moreover, the peak at about 26° corresponds to the anatase phase of the crystalline TiO2 tubes, whereas the metallic Ti substrate gives rise to recognizable sharp peaks.
Fig. 9 (a) Energy-dispersive X-ray spectroscopy analysis and (b) X-ray diffractogram performed on Pt deposited by ALD on annealed TiO2 nanotubes. |
The electrochemical characteristics of our Pt/TiO2 nanotubular samples are qualitatively similar to those of planar systems,57 as investigated by cyclic voltammetry. Starting from the open-circuit potential (OCP) measured at +0.05 V (vs. Ag/AgCl), the cyclic voltammogram (CV) recorded in aqueous acidic medium (0.5 M H2SO4, Fig. 10a) shows two reduction peaks located at −0.10 V and −0.19 V on the cathodic scan (with corresponding oxidation maxima at −0.08 V and −0.14 V). They correspond to the reductive adsorption of protons as hydrogen adatoms on the Pt surface. At −0.27 V, bulk dihydrogen evolution sets on. This shape of the cyclic voltammogram is fully consistent with reports of planar Pt electrodes in the literature.57
Fig. 10 Cyclic voltammograms of Pt clusters supported on TiO2 nanotubes carried out in acidic medium (0.5 M H2SO4 (a), and 0.5 M H2SO4 + 1 M EtOH (b)). Scan rate: 50 mV s−1. |
The same conclusion is reached when ethanol (1 M) is added to the mixture. The resulting CV curve (Fig. 10b) displays an oxidation onset at +0.40 V, an anodic maximum at 0.68 V, followed by a new anodic maximum on the reverse scan. This behavior, which is characteristic of CO poisoning and subsequent surface reactivation, is again known for the planar Pt system.57
The curves recorded in basic conditions (Fig. 11) display very similar features, shifted along the potential axis due to the difference in pH. Once again, the surface electrochemistry of our nanotubular Pt/TiO2 system conforms to that known for pure, planar Pt.
Fig. 11 Cyclic voltammograms of Pt clusters supported on TiO2 nanotubes carried out in 1 M KOH and in 1 M KOH + 1 M EtOH. Scan rate: 50 mV s−1. |
Although the curves qualitatively conform to the known planar case, they differ from it in a quantitative sense. Indeed, the current densities obtained are significantly higher, up to tens of mA cm−2. Taking into account the geometry of the samples, this corresponds to a specific current on the order of 50 mA (mg Pt)−1, comparable to the state of the art.58 This highly efficient utilization of the noble metal catalysts suggests that our preparative method might be of interest in electrochemical energy conversion applications.
The anodized felts are subsequently coated by a thin (5 nm) catalytic layer of Pt by ALD carried out at 130 °C using the parameters described above. For comparison purposes, other anodized felts were submitted to a deposition of 10 nm Pt by magnetron sputtering. The comparison of the electrocatalytic activity (in a 1 M H2SO4 solution) of Ti felts submitted to various treatments is shown on Fig. 13. The potentials are measured with respect to an Ag/AgCl reference electrode. Starting from the open circuit potential measured at about +0.6 V, the potentials are at first swept anodically up to +1.65 V, after which the CV is continued down to −0.45 V (at 100 mV s−1). For Ti felts without any electrochemical treatment nor Pt layer, the current density remains low (below 2 mA cm−2 in absolute value), no electrochemical activity is observed (gray curve). Providing a larger specific surface area by anodizing the Ti felts for 2 h (blue curve) increases the current somewhat, especially in the H2 evolution region (the most cathodic potentials). However, the absolute current density remains low, a few microamperes per cm2 of felt (measured macroscopically), due to the absence of an appropriate catalyst.
Deposition of 10 nm Pt by magnetron sputtering on the commercial Ti felts causes a significant increase in electrolytic current density (pink curve), reaching maximum current densities of −20 mA cm−2 and 5 mA cm−2 at −0.45 V and +1.65 V, respectively. These values remain quite modest for a felt (as opposed to a foil). Indeed, magnetron sputtering has limited ability to coat non-planar substrates. Being a directional deposition method, it only generates a film on the uppermost surfaces of the felt instead of coating each fiber. The ALD technique provides precisely the advantage of conformal coating. Depositing 5 nm of Pt by ALD, one obtains yet a further increase in current density (red curve) with respect to the sputter-coated sample. Moreover, the curve shape is now clearly reminiscent of that known for platinum (Fig. 11), with the anodic formation of surface hydroxide and oxide at positive potentials and the cathodic generation of surface hydrides in the negative potential region. In the bulk H2 generation regime, for potentials below −0.3 V, the current density reaches −35 mA cm−2 at −0.45 V.
Finally, we consider a Ti felt combining all favorable treatments: anodization (2 h) for the large surface area, followed by ALD coating with 5 nm Pt for catalytic proficiency. The result, shown as the green curve of Fig. 13, is quite dramatic. The qualitative curve shape of a Pt surface is clearly recognizable, but the quantitative values of the current density are very high. Both extremities of the cyclic voltammogram (in the bulk hydrogen and oxygen evolution regions) reach current densities on the order of 100 mA cm−2. This represents an enhancement by a factor >50 with respect to the naked, commercial Ti felt. Much of the success of our preparative strategy is attributable to the generation of a large surface area. The catalytically active Pt surface area can be evaluated by integrating the hydride generation current and comparing it to the established value of 0.21 mC cm−2 of planar Pt.59 This approach delivers a roughness factor (microscopic catalytically active surface area over macroscopically defined sample area) of approximately 300. This experimentally derived area combined with the known nominal Pt film thickness also yields an upper bound for the Pt loading of 3 mg cm−2 felt (value calculated for a hypothetically smooth film). Taking into account film roughness would yield lower loading values. Thus, the electrocatalytic water reduction efficiency is on the order of 30 mA (mg Pt)−1 at −0.45 V or better. Note that the capacitive contribution to the current measured at this potential is insignificant, as indicated by the light blue curve of Fig. 13. Also, the fact that the pink, red and green curves differ drastically in magnitude without large shifts of the peaks demonstrates that the ohmic resistance of the TiO2 material is low. This puts our ALD-based system on par with the state of the art,58 and leaves some room for improvement as the ALD method enables us to lower the Pt loading quite systematically in the future.
The method established here is quite general, and is applicable to various noble metals and various electrochemical applications. Based on it, further development is possible by exploitation of binary or ternary catalytic materials based for example on the combination of platinum and iridium, in elemental and/or oxidized form. Such systems, which can be obtained by ALD, are well established as reversible fuel cell electrode catalysts with properties that can be optimized in terms of chemical stability and electrocatalytic throughput.60–63
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ta00205b |
‡ Current address: University of Toulouse, Laboratory of Chemical Engineering, ENSIACET/INP Toulouse/UMR CNRS 5503, 4 allée Émile Monso, BP 44362, 31432 Toulouse Cedex 4, France. |
§ Current address: University of Göttingen, Institute for Organic and Biomolecular Chemistry, Tammannstr. 2, 37077 Göttingen, Germany. |
This journal is © The Royal Society of Chemistry 2015 |