Electrodeposition of palladium-dotted nickel nanowire networks as a robust self-supported methanol electrooxidation catalyst

All solutions were freshly prepared and all glassware was cleaned with boiling aqua regia prior to use. For the preparation of all solutions, purified water (Milli-Q, > \({18.2}\,\hbox{M}\Omega\,\hbox{cm}\)) and the following chemicals were used without further purification: borane dimethylamine complex (Sigma-Aldrich, \({97}\%\)), boric acid (Sigma-Aldrich, > \({99.5}\%\)), (DCM) (Merck, > \({99.8}\%\)), ethanol (Roth, > \({99.8}\%\)), hydrochloric acid (\({37}\%\), AppliChem, for analysis; \({32}\%\), J.T. Baker, Baker Analyzed purity grade), methanol (AppliChem, > \({99.5}\%\)), Nafion solution (\({5}\,{\hbox {wt.}\%}\), EC-NS-05, ElectroChem Inc.), nickel(II) chloride (Aldrich, \({98}\%\)), nitric acid ( \({70}\%\), Sigma-Aldrich, > \({99.999}\%\)), palladium on carbon (\({10}\,{\hbox {wt.}\%}\), activated carbon support, Aldrich), palladium(II) chloride (Alfa Aesar, \({99.9}\%\) metal basis), sodium hydroxide (Grüssing, > \({99}\%\)).

Commercially available ipPORE polycarbonate membranes from it4ip with a thickness of \({25}\,\upmu \hbox {m}\), an ion fluence of \({1.5\times 10^8}\,\hbox {cm}^{-2}\), and pore diameters of \({400}\,\hbox {nm}\) were first activated by means of swelling activation [26]. Therefore, the membranes were immersed in a solution of \({0.5}\,\hbox {M}\) borane dimethylamine complex (DMAB) in methanol for \({2}\,\hbox {min}\) (swelling step). Afterwards, they were rinsed with water two times and transferred to a solution of \({59}\,\hbox {mM}\) PdCl\(_2\) and \({108}\,\hbox {mM}\) HCl in water for another \({2}\,\hbox {min}\) (activation step) followed by another rinsing step in water. This procedure of swelling and activation was repeated up to six times to achieve different Pd loadings. After this nanoparticle deposition process, the template surface was gently rubbed with an ethanol-containing tissue to remove Pd nuclei, which have formed on the outer template surfaces.

For the electrodeposition of Ni, a Ni back electrode was sputtered onto one surface of the membranes using a Quorum Q300T D sputter coater. The sputter current was adjusted to \({100}\,\hbox {mA}\) and applied for \({120}\,\hbox {s}\) to form a conductive layer. Electrodeposition was carried out at a potential of \({-1.0}\,\hbox {V}\) vs. Ag\(\vert\)AgCl (Meinsberg, in \({3}\,\hbox {M}\) KCl, \({+210}\,\hbox {mV}\) vs. SHE) in an electrolyte composed of \({0.5}\,\hbox {M}\) NiCl\(_2\) and \({0.5}\,\hbox {M}\) H\(_3\)BO\(_3\) in purified water. The deposition was stopped when the current started to rapidly increase after the template pores were completely filled. After electrodeposition, the Ni back electrode was removed by dabbing and rubbing the surface with a cotton bud impregnated with aqua regia so that only the polymer template with the nanowire network inside remained.

For scanning electron microscopy (SEM), samples were immersed in fresh DCM three times for \({30}\,\hbox {min}\) each to remove the polymer template. Afterwards, the plain nanostructures were glued to an SEM sample holder with Planocarbon N650 carbon glue to ensure electric conductivity. SEM measurements were performed on a Philips XL30 FEG using an accelerating voltage of \({25}\,\hbox {kV}\). For energy-dispersive X-ray spectroscopy (EDX) measurements, an EDAX CDU LEAP detector was used.

For the transmission electron microscopy (TEM) measurements, a small piece of each sample (still located in the polymer template) was embedded in Araldite^® 502 resin. This resin was cured at \({60}\,^\circ \hbox {C}\) for \({16}\,\hbox {h}\). Afterwards, the obtained resin block was cut into thin slices of approximately \({70}\,\hbox {nm}\) thickness by a Reichert-Jung Ultracut E ultramicrotome equipped with a diamond knife by Delaware Diamond Knives. The cut slices were collected on a Cu grid and then investigated with a FEI CM20 operating at \({200}\,\hbox {kV}\). EDX measurements were performed using an Oxford X-Max detector.

The precise amounts of Ni and Pd were determined by inductively coupled plasma optical emission spectroscopy (ICP-OES). Analyses were conducted on an iCAP 7400 Duo MFC ICP-OES Analyzer by Thermo Scientific. After dissolving the polymer template, the samples were dissolved in \({4.5}\,\hbox {mL}\) aqua regia diluted with \({5}\,\hbox {mL}\) of H\(_{2}\)O. This solution was then further diluted with \({3}\,{\hbox {wt.}\%}\) HNO\(_{3}\) prior to the measurements.

Electrochemical measurements were carried out on a Gamry Reference 600+ potentiostat. Prior to the measurements, the polymer template was completely dissolved by repeatedly rinsing the samples with DCM similar to SEM preparation. The remaining nanowire network was then weighed on a balance, placed on a glass slide, and electrically contacted by a Ni wire with \({0.5}\,\hbox {mm}\) diameter in the measurement cell. For the reference catalyst, \({15.64}\,\hbox {mg}\) of Pd@C powder and \({50}\,\hbox {mL}\) of \({5}\,{\hbox {wt.}\%}\) Nafion™ solution were mixed with \({1}\,\hbox {mL}\) of water. This suspension was then ultrasonicated for approximately \({10}\,\hbox {min}\). Subsequently, \({2.5}\,{\upmu \hbox {L}}\) of the suspension (containing \({3.5}\,{\upmu \hbox {g}}\) of Pd) was deposited on the glassy carbon electrode and left to dry in air for \({15}\,\hbox {min}\). Both preparation techniques are also schematically illustrated in the Supporting Information, Figs. S1 and S2.

The so prepared electrodes were then connected to the potentiostat as working electrode. A platinised Ti mesh was used as a counter electrode and an Ag\(\vert\)AgCl electrode (Meinsberg, in \({3}\,\hbox {M}\) KCl, \({+210}\,\hbox {mV}\) vs. SHE) as reference electrode, respectively. All measurements were carried out in unstirred \({0.1}\,\hbox {M}\) NaOH solution. For electrocatalytic measurements, methanol was added in various concentrations. The scan range was set to \({-1.0}\,\hbox {V}\) to \({0.5}\,\hbox {V}\) while the scan rate was adjusted to \({50}\,\hbox {mV}\,\hbox {s}^{-1}\). Also, automated iR drop-correction via current interrupt technique was applied to all measurements [27, 28]. For the long-term measurements, the alcohol concentration was fixed at \({0.25}\,\hbox {M}\) and the scan rate was increased to \({100}\,\hbox{mV}\,\hbox {s}^{-1}\).

Figure 3a shows SEM images of the fabricated nanowire networks. Although the single nanowires are stochastically aligned, the mechanical stability of the network is remarkably high, as seen in Fig. 2. The height of the network amounts to approximately \({25}\,\upmu\hbox{m}\). In combination with the lack of cap growth, i.e. Ni deposition occurring on top of the polymer membrane after completely filling the template pores, this indicates a good control over the electrodeposition process with homogeneous wire growth along the whole sample. The back electrode removal by careful dabbing with aqua regia also proves to be a precise process, since no remains of the back electrode are visible, while no damage caused by etching could be observed on the edges of the obtained nanowires. The detailed view shown in the inset of Fig. 3a indicates continuous and smooth-walled nanowires. This suggests a good integration of the Pd nanoparticles into the nanowire surface. Furthermore, since no voids in nanowire growth can be observed, it can be deduced that the density of Pd seeds was low enough and their size sufficiently small to not provide any extended conductive paths and thus ensure void-free nanowire growth along the template pores. Similar results were obtained for the two and four times activated samples. An EDX spectrum of a nanowire network that was activated six times is shown in Fig. 3b. The Pd signal is very low as expected and corresponds to approximately \({4.9}\,{\hbox {wt.}\%}\) of Pd when quantifying only Ni and Pd. For samples with fewer activation steps, the Pd signal becomes even weaker and can no longer be distinguished from the background spectrum.

To obtain information about the spatial distribution and nanostructure of the Pd particles on the six times activated sample, TEM was performed. An image of a cross section of one single nanowire is shown in Fig. 4a. Due to the disordered alignment of the single nanowires, it cannot be deduced, whether this cross section was deformed during the TEM preparation or whether the nanowire was not aligned orthogonally to the cutting plane. From the image itself it can be observed that there are no voids inside the nanowire. Also, there is no clear visual indication of Pd nanoparticles which are embedded in the outer nanowire surface.

TEM EDX was performed on the two spots marked in Fig. 4a. The corresponding spectra are shown in Fig. 4b. The spectrum that was recorded in the centre of the nanowire shows only signals for Ni, Cu, and Si. Since the Cu and Si signals are caused by the TEM grid which held the sample slices and the surrounding TEM equipment, respectively, it can be stated that the centre of the nanowire consists only of pure Ni. The spectrum recorded in the outer parts of the nanowire shows additional signals for C, O, and Pd. While it is very likely that the signals for C and O can be attributed to the surrounding polymer resin and polymer template, the Pd signals are caused by the particles that were formed in the swelling activation process. Considering only the Ni and Pd signal, a quantification of this local EDX signal yields approximately \({0.85}\,{\hbox {wt.}\%}\) Pd and \({99.15}\,{\hbox {wt.}\%}\) Ni. This result verifies the successful implantation of Pd in the outer surface of the Ni nanowire network. To further fortify this, a TEM mapping was performed on another cross section of the same sample (see Supporting information, Fig. S6), which again shows the presence of Pd on the edge of the cross section.

ICP-OES measurements were performed on the two, four, and six times activated samples. The obtained fractions of Ni and Pd in \({\hbox {wt.}\%}\) are shown in Table 1. The values for Pd are very low (in the range of \({0.6}{\hbox {wt.}\%} \; \hbox{to} \; {2.4}{\hbox {wt.}\%}\)).

In Fig. 5, the iR drop-corrected cyclic voltammetry (CV) curves for two, four, and six times activated dotted samples in \({0.1}\,\hbox {M}\) NaOH solution are depicted. Furthermore, an iR drop-corrected CV curve of the prepared commercial Pd@C reference catalyst is shown. Data normalisation was performed by dividing the measured currents by the mass of Pd contained in each sample as derived from the sample weight and ICP-OES measurements.

A total of six different peaks can be observed for the dotted structures. The peak at \({-706}\,\hbox {mV}\) for the two times activated sample, \({-596}\,\hbox {mV}\) for the four times activated sample, and \({-745}\,\hbox {mV}\), \({-637}\,\hbox {mV}\) and \({-556}\,\hbox {mV}\) for the six times activated sample (peaks I–III) can be associated with the desorption of hydrogen species and the adsorption of oxidic/hydroxidic species at the Pd surface [17, 33, 34]. In the range from \({+426}\,\hbox {mV}\) to \({+460}\,\hbox {mV}\), a reduction peak which can be associated with the reduction of NiOOH species to Ni(OH)\(_2\) (peak IV) is found [17, 35, 36]. The fifth peak (peak V) that can be identified occurs in the range from \({-250}\,\hbox {mV}\) to \({-228}\,\hbox {mV}\) and is commonly attributed to the reduction of PdO species to Pd [34, 37, 38]. The corresponding oxidation peak of Pd, which can be found in anodic scan direction at values around \({0}\,\hbox {mV}\) according to literature, is not observed in this case [34, 37]. At approximately \({-640}\,\hbox {mV}\) (peak VI) the adsorption of hydrogen species occurs [34].

When comparing this to the Pd@C reference catalyst, it can be seen that the reference catalyst shows two more peaks in close vicinity to the Pd reduction peak in both anodic (\({-264}\,\hbox {mV}\)) and cathodic (\({-326}\,\hbox {mV}\)) scan direction (marked with an asterisk). It remains unclear to which chemical reactions these peaks correspond to. However, similar peaks are found for Pd@C catalysts in the literature [39].

The same set of samples was then exposed to various concentrations of methanol and again measured by means of CV. The obtained curves for a methanol concentration of \({0.25}\,\hbox{mol}\,\hbox {L}^{-1}\) are shown in Fig. 6. At first glance, two new peaks can be noticed: one is located at approximately \({-40}\,\hbox {mV}\) in anodic scan direction, the other one is found between \({-214}\,\hbox {mV}\) and \({-230}\,\hbox {mV}\) in cathodic scan direction. Both peaks can be associated with the oxidation of methanol species at the Pd surface [34, 37]. Further investigation of the curves reveals an attenuation of the H desorption and OH\(^-\) adsorption peaks and the vanishing of the H adsorption peak.

Both the vanishing of the H adsorption peak and the attenuation of the other two peaks can be explained by the presence of methanol in the alkaline electrolyte, since methanol also adsorbs on the Pd surface and thus reduces the number of available sites for other ad-/desorption reactions [33]. The same attenuation can also be observed for the Pd@C reference catalyst.

One feature that is unique to the Pd@C reference catalyst is two loops in the alcohol CV, both in close vicinity to the alcohol oxidation reaction peaks. These loops are caused by the automatic iR drop correction during the measurements, which was performed using the current interrupt technique. By adjusting the applied voltage from one data point to the next, this technique stretches and compresses the x-axis of the cyclovoltammograms depending on the voltage drop of the previous data point, which itself is depending on the measured electrical current [27]. For rapid changes in current followed by a slower current change, this can result in a strong deformation of the x-axis followed by its relaxation. This overshooting of the artificial control circuit imposed by the current interrupt technique thus leads to the formation of measurement loops.

For all samples discussed so far, CV measurements were performed with further concentrations of methanol. This is exemplarily shown in Fig. 7 for the two times activated sample. For the other curves, the reader is referred to the Supporting Information, Figs. S7–S9. For the dotted samples, the two peaks associated with the methanol oxidation reaction continuously increase in intensity with increasing methanol concentration, while the peaks associated with H ad-/desorption and OH\({^-}\) adsorption become attenuated. This fortifies the aforementioned correlation between the concentration of methanol in the electrolyte and the intensity of OH\({^-}\) adsorption due to the limited amount of available Pd catalyst sites.

A comparison of the mass activities (i.e. the normalised peak currents) of the dotted samples and the reference catalyst indicates higher mass activities for the latter one for alcohol concentrations up to \({1.00}\,\hbox {M}\) (see Fig. 8). Since the obtained currents were normalised to the mass of the Pd content of the samples, this indicates a higher amount of exposed surface area for the reference catalyst. On the one hand, this can be caused by a higher surface-to-volume ratio of the Pd nanoparticles of the reference catalyst. Since the surface-to-volume ratio of spherical particles decreases with \(r^{-1}\) with increasing radii r and thus also activity decreases [2, 37], this would indicate that the reference catalyst contains smaller catalyst particles than the dotted samples. On the other hand, the dotted structures have their catalytic particles not attached to the substrate surface but embedded into it, which leads to a decrease in exposed area for catalytic nanoparticles of identical dimensions.

Thus, a high-surface-to-volume ratio and a high amount of small Pd nanoparticles become even more important.

The latter was tried to be achieved by increasing the amount of activation cycles. In Fig. 9, various possibilities of nanoparticle growth during activation are shown schematically. While the first scheme column shows well separated nanoparticles with roughly the same dimensions, the second column shows an agglomeration of small nanoparticles and the third column shows the growth of one large nanoparticle instead of nucleation and formation of multiple nanoparticles. When comparing these three possibilities with regard to the expected mass-activities of the exposed catalyst, the mass activity would stay constant for the first schematic and growing numbers of nanoparticles. Only the absolute currents would increase. For the second and third case, mass activities would decrease with an increasing amount of deposited material.

However, mass activities increase with an increasing number of activation treatments, as derived from Fig. 6. This mismatch can either be attributed to compositional differences between the samples used in ICP-OES and the samples applied in half cell measurements or a not yet discovered parameter during swelling activation. Although compositional differences cannot be completely excluded, these also boil down to some experimental variance. We believe that the rinsing step between the swelling step and the activation step during nanoparticle deposition might cause these variances. When rinsing the sample between the two steps, there is a conflict of rinsing for too long, i.e. allowing too much of the absorbed reducing agent to leave the polymer again, and rinsing too short, i.e. not properly cleaning the template pores. While the first option would lead to lower amounts of reducing agent available for metal reduction, the latter option can lead to remains of the swelling solution inside the template pores and thus an increased amount of reducing agent for metal reduction above the intended value. Also, this might lead to nucleation inside the template pores followed by sedimentation of the so-formed nanoparticles at the pore walls. This would then lead to a difference in composition although the same procedure was carried out and thus would have an impact on the applied data normalisation.

When looking at the CV curves of the dotted structures in comparison to the Pd@C reference catalyst in general, it becomes obvious that the latter one provides higher mass activities than each of the dotted structures for alcohol concentrations up to \({1.00}\,\hbox {M}\). At higher concentrations, the measured mass activity of the reference catalyst dramatically decreases and is then surpassed by all dotted structures. The increased mass activities for low alcohol concentrations of the reference catalyst can be explained by the exposed surface of the Pd nanoparticles for both catalyst types and the applied normalisation. The dotted structures, however, prove to be more tolerant to high alcohol concentrations as indicated by the measured mass activities at alcohol concentrations of \({5.00}\,\hbox {M}\). One possible reason for this might be a synergistic effect between the Pd catalyst and the less noble Ni, where the catalyst is responsible for oxidising methanol, while the surrounding Ni substrate can provide additional OH\(^-\) ions for facilitated oxidation of methanol, even at high concentrations of alcohol [19]. This is supported by An et al. who report an activity loss caused by a high concentration of ethanol and thus blocking of the active sites by adsorbed alcohol species [11].

To further investigate the performance of the dotted nanowire networks, long-term cycling in the same measurement range as before and with the addition of \({0.25}\,\hbox {M}\) methanol to the electrolyte was performed at a scan rate of \({100}\,\hbox{mV}\,\hbox{s}^{-1}\) after the aforementioned concentration studies. The obtained results are compared against the commercial Pd@C reference catalyst and illustrated in Fig. 10. For evaluation, the charge correlated with the methanol oxidation peak in anodic scan direction was determined. Normalisation was then performed with respect to the maximum charge obtained during the whole measurement.

The Pd@C reference catalyst shows an exponential decrease in activity over the whole measurement. Only approximately 4% of the maximum charge remain after 1000 cycles. The two and four times activated dotted structures behave differently in the first 100 measurement cycles, but then show a much weaker decay from 100% to 92% of the maximum charge and from 84% to 79% of the maximum charge, respectively. Interestingly, the six times activated sample does neither show a decrease in performance, nor a constant value, but indicates an increase in performance over time. Given the possibility of methanol evaporation during the measurement and thus a decrease of fuel concentration in the analyte solution, this is even more remarkable. However, due to the numerous factors which may influence long-term stability testing and the fact that an investigation of these would already be an own study, the reasons for this behaviour shall not be discussed here. Instead, it shall be pointed out, that the overall long-term stability of the dotted nanostructure catalysts is much higher than the one of the Pd@C reference catalyst.

To further examine the long-term stability of both the dotted nanowire networks and the Pd@C reference catalyst, SEM was carried out on each sample before and after the long-term stability test. The dotted samples did not show any change in morphology (see Supporting Information, Fig. S10). Their appearance was the same as shown in Fig. 3a. However, the Pd@C reference catalyst significantly changed its shape during the test, as shown in Fig. 11. The image on the left shows a dried droplet of the reference catalyst suspension before the test. Here, single platelets of carbon decorated with Pd can be observed and clearly distinguished from each other. After the test, the single platelets can hardly be discriminated. Instead, they changed their morphology and formed one large agglomerate with probably decreased surface area. This might explain the drop in performance of the reference catalyst observed in the long-term stability test (see Fig. 10).

Also, the electrolytes of the long-term cycling tests were investigated by means of ICP-OES to determine any metal losses during the cycling test and to further evaluate the possible degradation mechanisms of the catalyst. For all dotted samples, no Pd could be detected in the electrolytes. Thus, particle detachment and dissolution which are both severe degradation mechanisms for noble metal nanoparticle-based catalysts do not contribute to the change of the performance of the dotted catalyst. Also, Ni species could not be detected, which indicates that the nanowire network did not corrode, either. The same analysis was then performed on the electrolyte of the Pd@C reference catalyst. Here, no Pd could be detected in the used electrolyte. From this, it can be concluded that the decrease in performance of the Pd@C reference catalyst cannot be caused by catalyst particle detachment, or particle dissolution but must be caused by the enormous change in catalyst surface and morphology, probably accompanied by the agglomeration of Pd nanoparticles. Also, methanol evaporation during measurement or electrolyte depletion in close vicinity of the electrode remains possible.

For the dotted structure, no changes in the overall morphology were observed and also, Pd particle agglomeration is strongly suppressed due to the embedding of the particles in the Ni nanowire surface [40]. Thus, methanol evaporation, electrolyte depletion, and a surface modification of the structures on the atomic scale remain as explanations for any changes in measured performance.