Hydrogen is recognised as a clean and highly efficient energy source with enormous potential to meet energy needs while minimising greenhouse gas emissions, thus facilitating the global transition to renewable energy. The proton exchange membrane fuel cell (PEMFC) is a highly promising and advanced technology in energy development due to its exceptional efficiency and eco-friendly characteristics [1,2]. The most important issue facing PEMFC users and manufacturers is the durability of the fuel cell [3]. The stable operation of a single PEMFC is determined by its components, such as the bipolar plate and membrane–electrode assembly (MEA) [4,5]. Particular attention should be paid to the cleanliness of the fuel in PEMFC operation. Contaminants are known to be particularly harmful and affect performance even when present in trace amounts (ppm) [6].

Carbon monoxide (CO) impurities are considered to be particularly harmful [7]. The main mechanism for the reduction of PEMFC parameters in the presence of CO in the fuel is the poisoning of platinum catalysts due to the irreversible adsorption of CO [8]. CO competes with hydrogen for the active sites on the platinum catalyst at normal anode operating potentials. The surface available for the target process is reduced, while CO accumulates on the Pt surface [9]. Today, more than half of the world’s hydrogen is produced by steam reforming of natural gas, methane or other hydrocarbons [10]. Hydrogen gas - produced by steam reforming of natural gas has a typical composition of 40 to 70% H₂, 15 to 25% CO₂, 1 to 2% CO and impurities of inert gases. Hydrogen produced in this way is therefore not suitable for direct use in PEMFCs. Several stages of purification of the converted gas are required to reduce the CO concentration to a level acceptable for use in a PEMFC (less than 10 ppm).

Mitigation strategies the impact of CO impurities on PEMFC performance can be divided into three groups: purification of hydrogen from impurities [11], additional hydrogen purification systems as part of the PEMFC-based power system [12], increasing CO tolerance and restoring performance characteristics directly during PEMFC operation [9]. Most promising, especially for the use of PEMFCs in mobile power plants with strict volume and/or weight requirements, is the development of a strategy for restoring the PEMFC characteristics directly during operation. The resistance of the anodic catalyst to carbon monoxide poisoning is one of the most important properties of Pt/C electrocatalysts in PEMFCs, along with their stability.

Pt/C catalysts can be synthesized using the (electro) chemical bottom-up approach to optimize their performance. In this case, the synthesis of Pt nanoparticles (NPs) involves the electrochemical reduction of metal ions to a metallic state. The size of the electrodeposited Pt NPs can be controlled by varying the electrolysis parameters [13]. However, in this bottom-up electrochemical approach, the Pt NPs are produced only on the surface of the working electrode and not on the dispersed support. It should be noted that most industrial applications still require the use of dispersed catalytic systems, where the active catalytic phase is located on the surface of a support material. To obtain such catalytic systems, the impregnation method is the most commonly used approach [14]. The microstructural characteristics of the catalytically active phase (Pt NPs) are strongly influenced by the nature of the reactants (solid surface and liquid) and the reaction conditions. It is important to note that Pt NPs are thermodynamically metastable due to their large surface area [15]. In addition, dispersity stabilisers and capping agents are used during the synthesis process to prevent particle growth and agglomeration. However, these additives may contaminate the final product and negatively impact its catalytic properties. Furthermore, in accordance with the principles of green chemistry, Pt catalysts that are used in environmentally friendly technologies should also be synthesized using environmentally friendly methods. The use of electrochemical techniques in top-down approaches, specifically pulse electrolysis, can be a simple and environmentally-friendly method to disperse Pt electrodes into nanoscale Pt particles: compared to bottom-up electrochemical and liquid-phase chemical approaches for synthesising catalytic materials based on metal nanoparticles, the top-down electrochemical approach does not require the use of organic solvents or stabilisers during synthesis [16].

Previously, we have demonstrated the influence of the preparation technique of the Pt/C catalyst and the morphology of the carbon support in the composition of such catalysts on their resistance to carbon monoxide poisoning [17]. This paper investigates the carbon monoxide resistance of a platinum-based catalyst prepared by pulsed alternating current (AC) technique in a membrane electrode assembly (MEA) and develops a method to restore the MEA performance characteristics directly during PEMFC operation.

The application of pulsed AC to platinum electrodes in highly alkaline aqueous electrolytes containing alkali metal cations allows the production of Pt nanoparticles. The mechanism of Pt nanoparticles formation under pulsed AC conditions is due to alkali metal intercalation into the Pt crystal lattice, formation of platinum–alkali metal intermetallic compound, and decomposition of the intermetallic compound by water, resulting in formation of dispersed Pt. The formation of Pt nanoparticles under pulsed AC conditions was influenced by the formation of platinum oxide phase and its reduction, as well as gas filling of the near-electrode layer and thermokinetic effects. This mechanism has been discussed in detail in [16].

TEM images (Fig. 1a,b) showed that the Pt particles were uniformly distributed over the carbon black surface. During the synthesis process of Pt-containing catalytic systems under pulsed AC conditions, it is important to stir the suspension of support material in electrolyte solution. The support material can be of different natures, such as carbon black, carbon nanotubes, graphene, metal oxide, and metal oxide–carbon black hybrid carriers. The degree of agglomeration is largely determined by the duration of synthesis, Pt content in the catalyst and nature of support materials [18]. Simultaneously, the profiles of the primary electrical parameters of the electrochemical cell remained unchanged upon introduction of the support material into the electrolyte solution. The average size of platinum nanoparticles (<d>) for pulsed AC catalysts is 8–10 nm according to TEM data (Fig. 1c). The presence of agglomerates larger than 14 nm is observed. It is worth noting that the fraction of agglomerates larger than 14 nm is less than 1%, which does not affect the catalyst’s overall performance. However, the study conducted by Professor Feliu's research group [19] on platinum CO oxidation processes revealed the impact of platinum nanoparticle agglomeration on CO oxidation kinetics. It was demonstrated that the overvoltage of the CO oxidation reaction on platinum agglomerates decreases due to a distinct mechanism of interaction between CO and oxygen-containing species (intra-particle reaction and inter-particle reaction).

The X-ray diffraction patterns of the synthesised Pt/C catalyst are shown in Fig. 1d. Diffraction peaks of the XRD patterns corresponding to reflections of the face centred cubic (fcc) structure of platinum (Pt) can be assigned (JCPDS card 04-0802). The average size of the Pt nanoparticles and the unit cell parameters were determined by Rietveld refinement [20] (Table 1). The FullProf Suite was used to refine the powder diffraction pattern of Pt/Vulcan on a cubic cell in the Fm3m space group. The line shape of the diffraction peaks was generated using the Thomson–Cox–Hastings profile function [21]. The background was found by linear interpolation between successive breakpoints in the pattern. The particle shape was simulated using the symmetrized spherical harmonics, which describe the dependence of the Voigt function on the integral width. The Gaussian part of the profile was used to calculate the anisotropic strain. The Lorentzian component of the total Voigt function was used to calculate the crystallite size anisotropy. The shapes of the Pt nanoparticles were in the form of a truncated cube. The unit cell parameter of Pt nanoparticles is smaller than that of bulk platinum due to the size effect [22].

Power characteristics MEA before and after CO poisoning are shown in Fig. 2a. Fig. 2a shows that in the 0.9–0.6 V region there are losses associated with activation polarisation. This is mainly related to the kinetics of the oxygen reduction reaction. The region of ohmic losses (0.6–0.5 V region) is characterised by a linear dependence of the voltage on the current flowing. This is due to the influence of the resistance of the electrolyte and the electrodes on the transfer of ions and electrons. At high current densities, losses associated with mass transfer dominate, mainly due to the supply of gases through the porous structures of the gas diffusion and catalytic layers. The MEA power density at a cell voltage of 0.4 V was 160 mW cm^–2. The CO poisoning was carried out by introducing a large amount of CO into the anode line. This corresponds to several dozen hours of operation on dirty hydrogen. During long-term operation of the MEA, CO accumulates on the catalyst surface, which allows us to compare the volume of CO introduced with the volume of CO accumulated during long-term operation of the MEA on contaminated hydrogen. The CO was injected by opening the electromagnetic valve, the volume of injected CO was ~12·10^–4 L. The injection of this quantity of carbon monoxide corresponds to 100 hours of operation of this MEA on hydrogen fuel with a CO concentration of 25 ppm. Fig. 2b shows the effect of CO contamination on the MEA performance at a cell voltage of 0.4 V. When CO is injected into the fuel, a characteristic drop in power density is observed due to the accumulation of carbon monoxide on the catalyst surface and the blocking of the electrochemically active surface area of the platinum. The power density drop for each injection ranged from 4 to 10%.

Two methods are considered for restoring the functionality of the MEA during operation to increase CO tolerance. The first method is based on varying the parameters of the load profile. Carbon monoxide can be removed from the catalyst surface by short pulse loading. In addition, brief exposure to short-circuit currents can restore the operating characteristics of an already activated MEA [23]. The second method is based on the injection of an oxidising agent (air) directly into the fuel flow during the operation of the MEA. This recovery method is based on the selective oxidation of carbon monoxide in the catalyst layer. Fig. 3 shows the restoration of MEA operating characteristics using a short circuit and oxidiser injection. The current density of the membrane–electrode assembly at a cell voltage of 0.4 V in nominal mode was 400 mA cm^–2, corresponding to a power density of more than 160 mW cm^–2 (Fig. 3a,c). When CO is added to the fuel, a decrease in specific power of about 25% is observed. At the same time, the process of reducing the power characteristics is stabilised (Fig. 3b,d).

CO electrooxidation on Pt is known to proceed through the Langmuir–Hinshelwood mechanism, i.e., the reaction occurs between adsorbed CO and adsorbed O-containing species.

Since at potentials above the ignition potential the oxidation current becomes diffusion controlled, also the transport of CO molecules from the bulk solution to the surface. The reaction scheme possesses three time-dependent quantities, the surface coverages CO, OH and the concentration of CO in front of the electrode [24].

During a short circuit, a sharp change in the potential and current density of the cell is observed. During short circuit, the current reaches a large value and much water and heat are produced. Excess water is generated in the catalyst layer and moisture evaporates from the gas diffusion layer with high heat generation [25,26]. Producing periods of high current density increases the anode potential and CO on the catalyst is oxidised to CO₂ [27]. The removal of CO from the Pt surface occurs in a short time (tens of milliseconds), so the anode potential is mainly in the hydrogen oxidation region rather than the CO oxidation region [28]. The pulse amplitude is selected according to the MEA configuration, as the potential at which CO oxidation occurs depends on the catalyst used. It is recommended to keep the pulse duration constant regardless of the CO concentration in the fuel. Unjustified increases in pulse duration will reduce the efficiency of the MEA [29]. The key parameter is the pulse frequency, which is selected according to the electrode size, flow rate and CO concentration in the fuel.

It should be noted that the use of a short circuit only partially restored the power characteristics of the MEA due to a number of factors (Fig. 3a,b). Exposure to a short circuit causes an increase in the anodic overvoltage, which affects the surface coverage of the CL and GDL with CO and OH species. According to [30], there is an increase in CO coverage close to the CL–PEM interface, a decreasing behaviour of adsorbed CO can be seen in the middle region of CL and close to the GDL–CL interface. On the contrary, the opposite trend can be concluded for the H and OH coverages. The main reason for this condition is that CO diffusion through CL and penetration of adsorbed CO is still ongoing and incomplete near the PEM interface. Therefore, the overpotential is high enough to overcome the diffusion effect and oxidise the small amount of adsorbed CO near the GDL interface. In contrast, CO diffusion is more dominant than the anode overpotential near the PEM interface. Therefore, only a limited amount of adsorbed CO can be oxidised by the production of OH at the peak overpotential.

At the same time, a brief injection of an oxidising agent (air) into the fuel flow allowed the MEA performance to be restored without interrupting its operation (Fig. 3d). When air is added to the fuel stream, oxygen is adsorbed on the CO-free areas of the Pt catalyst and a surface reaction between Pt–CO and Pt–O then occurs to form CO₂. This gas phase catalytic oxidation reduces the surface area covered by CO and increases the activity of the catalyst in the hydrogen electrooxidation reaction [31]. Carbon monoxide can be oxidised to CO₂ via O₂ dissociation and formation of the OC–COO complex. Concentrations of 2–5% oxygen in the fuel are sufficient to completely mitigate the effects of even high concentrations of carbon monoxide at MEA operating temperatures. About one in 400 oxygen molecules participates in the oxidation of CO [32]. The rest of the oxygen burns chemically with hydrogen. The main advantage of this process is that a small amount of CO reacts in the presence of a large amount of H₂. The process can be run continuously. In addition, as the oxygen required for the reaction comes from the air, little energy is required to supply it and the process can take place at atmospheric pressure [33].

Despite their different efficiencies, both methods form the basis of the mitigation strategy of CO-poisoning for the PEMFC. The strategy consists of several stages (Fig. 4а). The first stage consists of cyclic humidification of the membrane–electrode assembly by alternating high power (short-circuit) and nominal power phases. In this case, a brief switch to high power mode should occur whenever the nominal power of the PEMFC drops by more than 5%. During a short-circuit, the PEMFC works at full power (water and heat production, gas consumption, etc.). Therefore, morphological changes are accelerated: restructuring of the ionomer and the catalyst. The cyclic oscillation of the cathode voltage is favourable for the reduction and oxidation of the various species blocking the surface, and the high gas flow is ideal for their evacuation. Cyclic swelling/shrinking of the membrane may also be more effective than conditions of constant high humidity [34]. If it is not possible to restore the nominal power of the PEMFC due to a short circuit as a result of CO accumulation on the catalyst surface, then it is necessary to proceed to the second stage. The second stage consists of selective oxidation of the CO accumulated on the catalyst surface. Oxidation is achieved by introducing an oxidant (air) directly into the fuel stream while the fuel cell is operating. It should be noted that the introduction of oxygen into the fuel presents a risk of forming explosive mixtures with hydrogen. The amount of gaseous oxygen in hydrogen has been limited to 0.5–1% [35] due to Low Flammability Level (LFL) limits [36].

This strategy was investigated by operating the MEA on hydrogen with CO contamination (Fig. 4b). When the MEA is operated on contaminated fuel, a systematic power drop is observed due to the blocking of the catalyst surface with carbon monoxide. Once the nominal power drop exceeded 5%, the first stage of power recovery began, significantly reducing the rate of loss of MEA productivity and stabilising its nominal power. To further reduce productivity, air was injected into the fuel stream to selectively oxidise the CO on the catalyst surface. Regular use of this procedure restored the MEA to its previous nominal power levels and stabilised its productivity in the range of 100% to 95% when operating on contaminated fuel. It is worth noting that in a strategy’s to reduce CO poisoning [27,37–40] based on similar processes, the level of power recovery is 70–95%. However, the amount of oxidizer introduced was higher, which will have a negative impact on the durability of the MEA. Besides despite the fact that carbon monoxide poisons the MEA, its oxidative desorption can contribute to the activation of the MEA and also has a positive effect on the operation of the PEMFC [41].

From the results and analyses discussed above, it is clear that this strategy is an effective approach to relieving CO poisoning for PEMFCs. However, all the tests were conducted over a period of several hours. The effects of short-circuit and injection of an oxidising agent (air) on durability and stability are still unclear. Therefore, the present study also conducted a stability study. Stability test is based on cyclic introducing a large amount of CO and restoring of the characteristics of the MEA using the developed method. Experimental results are shown in Fig. 4c. At the end of the 50-cylces, the MEA subjected to short-circuit and injection of an oxidising agent (air) had recovered up to 96% of its original power density. During the testing period, the cell current density was maintained in a fairly stable range (380–399 mA cm^–2). The degradation in the average current density during the test was very small (<4%). Although no reduction in MEA performance was observed during the tests, it should be noted that prolonged exposure to oxygen can affect the components of the MEA. Sintering of the catalyst due to the highly exothermic reaction between H₂ and O₂ [35]. Membrane degradation due to production of hydrogen peroxide, which results from a two-electron oxygen reduction reaction pathway and is related to the fraction of catalytic sites covered by CO at the anode [42,43]. However further minimising the percentage of required oxygen can significantly diminish the adverse effects of this strategy and enhance the overall mitigation efficiency.

In this study we investigated the resistance to carbon monoxide poisoning of platinum catalysts prepared by the pulsed AC technique. The catalyst obtained was studied as part of a membrane–electrode assembly. The CO poisoning was carried out by introducing a large amount of CO into the anode line. This corresponds to several dozen hours of operation on dirty hydrogen. The power density drop for each injection ranged from 4 to 10%. A method for restoring the operating characteristics of the membrane–electrode assembly of proton exchange membrane fuel cells has been developed, based on short-circuiting and selective oxidation of CO. The strategy consists of several stages based on alternating short-circuit and oxidiser injection. Regular use of this procedure restored the MEA to its previous nominal power levels and stabilised its productivity in the range of 100% to 95% when operating on contaminated fuel. The data obtained indicate that the use of reformed hydrogen as a fuel for PEMFCs is possible with the developed electrocatalysts.