Unsupported Pt Nanoparticles: Synthesis, Deactivation, and Hydrogen Electrocatalysis in Unpurified Electrolytes

Our synthesis of platinum nanoparticles relies on the reduction of H₂PtCl₆ using ascorbic acid as the reducing agent and sodium polyacrylate as a capping agent. The synthesis procedure is adopted from prior literature reports, where Pt(IV) salts have been shown to undergo reduction to metallic Pt with two equivalents of ascorbic acid. ^44–49 Sodium polyacrylate polymer as a capping agent offers narrow particle size distribution and has been reported previously for Pt synthesis. ^50–55 Alkalizing the reaction mixture with sodium hydroxide destabilizes the colloid and facilitates removal of the polyacrylate from the NPs. ^56–58 The main advantage of this approach (over the use of commercial Pt/C catalyst, for example) is that it offers a route to synthesize catalytically active Pt NPs that can be deposited on any substrate and analyzed without the confounding effects of supports or polymer binders. Our synthetic approach also makes use of low-cost reagents (other than the Pt precursor itself) and equipment that is readily available in many wet chemistry laboratories.

We executed TEM, XRD, and FTIR measurements to verify the composition and crystallinity of the Pt NPs before and after removal of the polyacrylate capping agent. Figures 1a, 1b show representative TEM images of capped and decapped (based-treated) Pt NPs. The particles in panel a clearly remain encapsulated in a matrix of polymer several nm in diameter. In some cases the polymer encapsulates a single Pt particle and in others several distinct particles are visible, which suggests that the polyacrylate arrests Pt NP growth by encapsulating individual particles, and these core–shell assemblies further aggregate in solution or upon deposition on the TEM grid. By contrast, particles that have been purified with alkaline solution aggregate on the TEM grid. Manual analysis of the particle size distribution (see Supporting Information) shows a normal distribution on the basis of particle number with a mean diameter of 3.1 ± 0.05 nm and circularity index of 0.9. The mass-averaged particle diameter, which is the relevant metric for estimating mass-specific surface area, is 3.4 nm. D-spacings extracted from high resolution TEM imaging (insets of Figs. 1a, 1b) further confirm the particles as face-centered cubic (fcc) Pt.

Zoom In Zoom Out Reset image size

Figures 1c, 1d depict representative XRD and ATR-FTIR data for capped and decapped Pt NPs. XRD results further confirm fcc Pt as the only crystalline component in the reaction mixture before and after purification. ATR-FTIR data from capped Pt exhibit absorbance features at 1640 cm⁻¹ and 1723 cm⁻¹, which we attribute to carboxylic and/or alcoholic peaks that are blue-shifted relative to the aqueous sodium polyacrylate control. This shift has been previously attributed to the interaction between the particle and the polymer. ^59–61 The decapped Pt NPs also exhibited absorbance features in this range, but at a considerably lower intensity and without clear evidence for a blueshift relative to polyacrylate in aqueous solution. This further suggests that the base treatment substantially removes polyacrylate from the NPs, but we cannot rule out the possibility that some polymer remains in solution or weakly bound to Pt.

To further characterize the decapped NPs, we used surface voltammetry to assess the accessibility of Pt surface sites via hydrogen underpotential deposition (HUPD) in aqueous H₂SO₄. These measurements yield one estimate of the electrochemically accessible surface area (ECSA) via integration of the charge associated with hydrogen adsorption–desorption. ¹⁸ Figure 2 shows a representative surface voltammogram for a film containing decapped Pt nanoparticles in 0.5 M H₂SO₄ at 100 mV s⁻¹ and a loading of 130 μg cm⁻².

Zoom In Zoom Out Reset image size

Using the generally accepted HUPD surface charge density of 210 μC cm⁻², the mass-specific ECSA for decapped Pt NPs was found to be 37.13 ± 0.69 m²/g (Error bound is 95% CI, for 29 samples tested) in 0.5 M H₂SO₄. This is less than half the value of 82 m²/g predicted directly from the TEM-derived particle size as 6/ρ d, where ρ is the bulk density of Pt (21.5 g cm⁻³) and d is the mass-average particle diameter (3.4 nm). To address this discrepancy, we carried out a series of HUPD measurements as a function of Pt mass loading on a glassy carbon electrode, with the results shown in Fig. 3. These data show the apparent mass-specific ECSA increased with reduced loading to a lower bound in the μg/cm² range. ECSA values calculated at these very low mass loadings, which correspond to one or a few layers of ∼3 nm particles, were consistent with the value expected from particle geometry alone, albeit with a commensurate increase in statistical uncertainty attributable to the difficulty of obtaining such low mass loadings with high precision via solution-phase deposition. The results of further measurements to rule out contributions from organic contamination are included in the Supplementary Data.

Zoom In Zoom Out Reset image size

These results can be interpreted as evidence that adhesion between unsupported Pt NPs is sufficiently weak that a fraction of the particles in the film spontaneously detach from the electrode upon submersion in the electrolyte (note that we used no polymer binders to adhere them to the carbon substrate), whereas films approaching a single particulate layer are more strongly bound. Another possibility is that reduced apparent ECSA for thicker Pt NP films is attributable to mass-transfer limitations, wherein the accessibility of protons to the inner layers of Pt particles is reduced as the total film thickness increases. We speculatively prefer the latter explanation on the basis that a volume occupied by spherical 3.4 nm Pt particles at a volume fraction of 50%, with the interstitial space containing 0.5 M H₂SO₄, will contain just 1 proton per ∼1000 surface Pt atoms, which illustrates the need for rapid proton diffusion through the dense porous film to accommodate H-adsorption on each Pt surface site. Nonetheless, further experiments were conducted with Pt loadings ≥100 μg cm⁻² to comport with common practice for voltammetric studies of Pt NP catalysts.

Prior literature reports have emphasized the susceptibility of Pt electrocatalysts to deactivation, which negatively impacts activity and lifetime. ^62–64 We were interested in understanding the mechanism or mechanisms by which these unsupported Pt NPs degrade. To do so, we quantified ECSA decay under continuous voltammetric cycling. Figure 4 presents the results as normalized ECSA versus cycle number in acid and alkaline electrolytes. For this analysis, 0.2 mg cm⁻² of Pt NPs were swept at 100 mV s⁻¹ over the potential range from 0 V to one of three positive potential limits: 0.5, 1.0, and 1.5 V vs RHE for a total of 50 cycles. A different Pt NP film was used for each potential window and 15 such films obtained from 5 different synthesis batches were tested for the analysis.

Zoom In Zoom Out Reset image size

Error bars representing 95% confidence intervals about the mean of 15 measurements illustrate remarkable consistency in the relationship between relative ECSA and cycle number. A monotonic decrease in ECSA was observed in all cases over 50 cycles. Moreover, we observed that the relative (normalized to initial ECSA) rate of ECSA loss per cycle was nearly identical when Pt nanoparticles were scanned at 0.5 V and 1 V vs RHE, suggesting similarity in their deactivation mechanisms. When scanned to 1.5 V vs RHE, the rate of deactivation was initially slower, but accelerated after several cycles such that the total fractional loss in surface area was ∼40% in all cases.

ECSA values also decreased monotonically with cycle number in alkaline electrolyte across all positive potential limits. The relative rates of deactivation in base were also found to be similar at 1 V and 1.5 V vs RHE, whereas the less positive potential limit resulted in marginally slower ECSA loss. These observations are consistent with the ability of aqueous base to remove residual capping agent from the Pt NP surface (which negates the benefit of positive potential cycling to remove contaminants) and suggests the presence of a second deactivation mechanism that dominates at positive potentials.

We consider it significant that these Pt NPs lost 25%–40% of their initial ECSA over just 50 voltammetric cycles, which required as little as ∼10 min. This extent of ECSA loss is large enough to conclude that the Pt particles are no longer pristine, thereby calling into question their use as experimental controls or activity benchmarks. Hence, it is reasonable to adopt as a general guideline that unsupported Pt NPs of this type should be freshly deposited onto a clean electrode at relatively short time intervals (every few minutes or perhaps between each individual measurement) if they are to be used to good effect in unpurified electrolytes.

Several deactivation mechanisms are understood to proceed for Pt catalysts under electrochemical conditions. These include catalyst poisoning via surface adsorption or electrodeposition of impurities; oxidative dissolution, which could be chemically or electrochemically induced; particle growth resulting from Ostwald ripening or similar processes; and particle detachment from the electrode substrate. In acidic electrolyte, unsupported Pt nanoparticles are known to undergo dissolution by repeated cycling across the potential range required to generate and reduce Pt-oxides. ^63–68 The dissolution proceeds primarily during the reductive scan when platinum oxide is nominally reduced back to metallic Pt; this reaction proceeds in parallel with dissolution of Pt²⁺ ions. Under continuous cycling, these ions may remain in the electrolyte or they may be redeposited as the electrode potential is swept further negative. This dissolution-redeposition process may then induce particle growth via electrochemically accelerated Ostwald ripening. ⁶⁹ Nevertheless, it is difficult to entirely eliminate oxidative cycling in practice, as this treatment is also highly effective at removing organic and/or inorganic contaminants from the Pt surface, either through catalytic oxidation or as a direct result of the dissolution of Pt surface layers. ⁷⁰

By contrast, the deactivation of Pt nanoparticles has not been as thoroughly characterized in alkaline conditions as in acid. Deposition of transition metal impurities on the electrode surface and analogous dissolution processes have each been implicated as the root causes of reduced catalytic activity. ¹⁹ The ECSA variation versus cycle number data in Fig. 4 yield puzzling results where the relative rates of deactivation (20%–40% over 50 cycles) are similar across nearly all conditions studied. These observations leave open the question of whether the dominant mechanism of ECSA loss is similar or different across all potential windows and pH conditions.

Figure 5 compiles the results of an experiment directed at differentiating between reversible catalyst poisoning and irreversible loss of surface area, wherein Pt NP films were first subjected to 50 cycles to a positive limit of 0.5 V vs RHE, followed by 50 cycles to 1.5 V vs RHE, and then a third set of 50 cycles to 0.5 V. The ECSA decay on scanning at 0.5 V vs RHE for the first 50 cycles in acid and alkaline solutions again resulted in a 20%–40% loss in ECSA. When the positive potential was shifted to 1.5 V vs RHE, ECSA values in acid fully recovered over 5–10 cycles and in some cases even exceeded the initial values. This indicates that the Pt surface was renewed after scanning at oxidizing potential, which strongly suggests catalyst poisoning dominates at the positive limit of 0.5 V vs RHE. Nonetheless, after the initial recovery at 1.5 V vs RHE, the ECSA decayed again by nearly the same amount as in the initial set of 50 cycles. Returning the potential window to a 0.5 V positive limit had little further impact on the ECSA, which continued to diminish at a reduced rate through a final set of 50 cycles. These data suggest the mechanism by which ECSA diminishes when cycling to a more positive potential limit is irreversible—dissolution, particle growth, or detachment.

Zoom In Zoom Out Reset image size

By contrast to the results in acid, when cycled in alkaline electrolyte, the ECSA did not recover when the voltage range was modulated from 0.5 to 1.5 V and back to 0.5 V. Instead we observed a nearly continuous decrease in surface area over 150 cycles, with a systematic negative offset in apparent ECSA for data collected to 1.5 V vs RHE attributable to systematic errors resulting from the Pt surface oxide reduction feature in the voltammogram (see Supplementary Data).

Our observations agree with prior reports that electrochemical oxidation is highly effective at removing catalyst poisons from Pt in acid. ⁷⁰ Many organic (e.g. hydrocarbons) and inorganic (e.g. sulfate, chloride) species from the synthesis mixture, the electrolyte, or the atmosphere may adsorb to the Pt surface and can be removed by cycling to positive potentials. ^70–72 By contrast, oxidative electrochemical cleaning did not renew the catalyst surface in base. There are at least three possible explanations for this latter observation. First, the catalyst surface may become contaminated with an intractable poison over all potential windows. A second explanation is that alkaline conditions facilitate detachment of nanoparticles from the electrode surface, and this may proceed at some relatively constant rate regardless of cycling conditions. A third explanation is that the Pt NPs undergo continuous particle growth in alkaline conditions under potential cycling that again depends weakly, if at all, on the positive scan limit.

To further differentiate between loss of surface area via particle growth, dissolution, and detachment, we performed identical location transmission electron microscopy (IL-TEM) experiments on catalyst thin films before and after they were subjected to CV cycling over a range of potential windows for 50 cycles at 100 mV s⁻¹. Figure 6 compiles the results of these experiments in acid electrolyte. At a positive limit of 0.5 V vs RHE, the TEM images before and after cycling show minimal changes in particle morphology. The overall particle size distribution also did not change before and after cycling. Similar results were observed when Pt NPs were scanned to 1 V vs RHE in acid. However, a stark difference was observed when the particles were cycled to 1.5 V vs RHE, where the TEM images clearly show particle growth. The particle size distribution data confirmed that the mean particle diameter increased by slightly more than two-fold, in rough agreement with the nearly 50% loss in ECSA for NP films on glassy carbon substrates cycled under the same conditions. A similar analysis was also performed in alkaline electrolyte, where scanning at 0.5 V and 1 V vs RHE did not result in a significant increase in particle size but cycling at 1.5 V vs RHE resulted in a ∼30% increase in mean particle diameter (see Supplementary Data), also in rough agreement with the ECSA loss observed for Pt NP films on Au.

Zoom In Zoom Out Reset image size

Combining the ECSA and IL-TEM analysis shows that when the potential is cycled up to 1 V vs RHE, catalyst poisoning is the dominant mechanism by which ECSA diminishes. This is evidenced by the fact that we saw no changes in particle size before and after cycling. But when samples were cycled to 1.5 V vs RHE, the particle size increased significantly in acid and base. We therefore attribute the growth of Pt nanoparticles to Ostwald ripening, which may be facilitated by electrochemical Pt dissolution and redeposition. ³³

It is striking that, despite evidence that at least two different potential and pH-dependent mechanisms are involved, the observed rate of Pt NP deactivation was remarkably consistent—tens of percent loss in ECSA over tens of minutes—across all conditions studied. Absent evidence to the contrary, we conclude that this consistent rate of ECSA loss is coincidental, and in fact the comparative rates of catalyst poisoning and particle growth could be different in more highly purified or more severely contaminated electrolytes. Indeed, prior work on Pt NPs under scrupulously clean conditions showed that stable cycling can be achieved when the potential is maintained negative of 1 V vs RHE. ^43,54 However, it is challenging and costly to obtain this level of cleanliness; hence, it may be preferable in many cases to simply refresh the Pt NP film by removing "spent" particles from the electrode substrate and replacing with a fresh NP film. This approach is directly analogous to the ubiquitous use of abrasive polishing to clean polycrystalline Pt electrodes between experiments, with the added benefit of the ability to control the Pt loading, and hence the areal density of active sites.

Alongside efforts to study degradation processes in unsupported Pt NPs, we assessed their suitability for routine measurements associated with HER/HOR catalyst benchmarking. Specifically, we sought to identify experimental timescales over which catalyst degradation translates to diminished activity toward hydrogen evolution/oxidation in reagent-grade electrolytes.

Figure 7 shows representative HER/HOR polarization curves of Pt nanoparticles compared to a commercial Pt/C catalyst at Pt mass loading of 0.2 mg/cm². Note that these electrodes initially exhibit roughness factors (electroactive surface area normalized to superficial area of the electrode) of ∼50, assuming accessibility of surface sites in the porous film matches that implied by HUPD measurements. We also included a simulated concentration overpotential curve (simulation details can be found in the Supplementary Data), which represents the current-overpotential response of an ideal nonpolarizable electrode whose activity is limited by mass transfer of H₂(aq) to and from the electrode surface. In acid, the HER activity of Pt/C and our unsupported Pt NPs are nominally identical and overlay with the concentration overpotential curve. This means that the hydrogen evolution/oxidation reaction on Pt is purely transport limited, and the kinetic behavior of the catalyst is not directly measurable. We then cycled the unsupported Pt NPs in the HER/HOR region for ∼30 min from −0.15 to +0.1 V vs RHE at 100 mV s⁻¹ and measured the HER/HOR performance again. No change was observed in the current-overpotential data after cycling, which suggests that any loss in catalytic activity was not sufficient to induce a kinetic limitation.

Zoom In Zoom Out Reset image size

This result illustrates several important properties of Pt catalysts when they are used for hydrogen evolution/oxidation measurements in acid. First, a clean, high surface area Pt catalyst film should generally yield transport-limited activity for the HER/HOR in aqueous acid unless steps are taken to greatly increase the mass transfer coefficient beyond what is achievable using conventional techniques like RDE voltammetry. ^73,74 Such a nanostructured Pt catalyst will continue to exhibit transport-limited activity even after it becomes poisoned or otherwise deactivated, provided the remaining active site density remains sufficiently high. A recent report showed that areal loadings of unsupported 3.8 nm Pt NPs needed to be reduced to <1 μg cm⁻² (i.e., sub-nm equivalent thickness) to reduce the observed rate of HER per superficial electrode area. ²¹ This would be roughly equivalent to a >99.5% loss in ECSA at the 0.2 mg cm⁻² loading we used here. Accordingly, unsupported Pt NPs and commercial Pt/C catalysts can certainly be used to benchmark the activity of a new HER/HOR catalyst, even in unpurified reagent-grade electrolyte, provided steps are taken to avoid catalyst contamination by dissolved Pt. ⁷⁵ However, if a novel catalyst is found to exhibit higher activity toward the HER or HOR than a Pt control in acid solution, it is in fact very unlikely that the catalyst of interest is "better than Pt," but rather the Pt control has likely been severely contaminated by electrolyte impurities or another source. ²¹

We also performed HER/HOR measurements in alkaline electrolyte, which gave considerably different results. The activity of Pt/C and Pt nanoparticles was similar, and both were lower than the simulated concentration overpotential. This is consistent with numerous prior reports showing that Pt is less active toward hydrogen evolution/oxidation under alkaline conditions than in acid. ^15,18,74 Cycling Pt nanoparticles in the HER region from −0.2 V to +0.2 V vs RHE for ∼30 min also resulted in ∼30% reduction in current density. This agrees well with ECSA data showing that accessible Pt surface sites decreased by tens of % over tens of minutes of continuous cycling. It is also consistent with the expectation that Pt HER/HOR electrocatalysis is kinetically limited in base, even for electrodes with relatively high Pt surface areas; hence, catalyst deactivation manifests as a proportional decrease in activity. These results collectively suggest that Pt-based control measurements in alkaline electrolytes are considerably more sensitive to sample history than those in acid. Comparisons between Pt and novel HER/HOR catalysts in alkaline conditions are also more relevant, since Pt remains limited by reaction kinetics over a wide range of catalyst loadings and it is indeed possible to obtain catalytic activities exceeding that of pure Pt. ^76–78

As a final assessment of the utility of our Pt NPs for catalyst benchmarking, we used open-circuit potential (OCP) measurements to calibrate Ag/AgCl reference electrodes to the reversible hydrogen electrode (RHE) potential. This procedure is crucial in characterizing catalysts for reactions involving proton-electron transfer because it enables accurate determination of catalyst overpotentials relative to an empirically measured, pH-dependent formal reduction potential. Figure 8 depicts representative RHE calibrations using Pt NPs alongside polycrystalline Pt disk electrodes in acid and base. For consistency, these measurements were initiated in electrolytes that were first purged with atmospheric air for several minutes (∼1000 seconds) until a stable potential was attained. The purge gas was then switched to hydrogen at the same flow rate. The open-circuit potential was found to decrease over several minutes as dissolved oxygen was replaced with hydrogen until it asymptotically approached a stable potential in the expected range for a reversible hydrogen electrode (near −0.25 V vs Ag/AgCl in 0.5 M H₂SO₄ and −1.0 V in 0.5 M KOH). Figures 8c, 8d show detailed views of a narrow range of potentials in the asymptotic region.

Zoom In Zoom Out Reset image size

In acid, OCP values for the Pt disk electrode and unsupported Pt NPs remained stable within ±2 mV over the last 10 min of the measurement. The minimum OCP values, which can be interpreted as an estimate of 0 V vs RHE, were −0.266 and −0.269 V vs Ag/AgCl; the modest difference between these values constitutes uncertainty in the RHE calibration attributable to, e.g. reference electrode drift or incomplete exclusion of atmospheric air during the hydrogen purge. These data show that clean Pt electrodes of varying surface roughness remain sufficiently catalytic over the measurement interval to obtain a stable equilibrium between hydrogen and protons.

In alkaline solution, Pt nanoparticles gave a similarly stable final OCP value, but the Pt disk was found to drift positive by ∼30 mV over several minutes. This indicates that the surface composition this electrode is sufficiently perturbed on the timescale of the RHE calibration to exhibit a mixed potential likely involving equilibration to the HER/HOR equilibrium and one or more additional redox reactions associated with impurities in solution or adsorbed on the electrode. ⁷⁹ These data again exemplify the greater sensitivity of Pt to deactivation in base, which results from its lower catalytic activity toward hydrogen evolution/oxidation. Moreover, the ability to ameliorate this instability by increasing the areal density of Pt surface sites suggests the deactivation mechanism involves surface chemistry where the rate of the deactivation is not directly proportional to the areal density of surface sites. For example, transport-limited surface poisoning reactions involving low concentrations of impurities with high sticking coefficients would be expected to give this behavior.