Introduction

Proton exchange membrane fuel cell (PEMFC) is known to have outstanding efficiencies, high power density, with the advantage of operating at low temperatures. Therefore, they have attracted huge attention for application in electrical vehicles. However, the major challenges for scaling up this technology and competing with conventional energy conversion technologies remain the high material costs and durability of platinum-based electrocatalysts [1,2,3]. The poor durability of PEMFC is mainly related to catalyst degradation in the oxygen reduction reaction (ORR) electrode. Until now, commercially available electrocatalysts used in PEMFCs were based on Pt nanoparticles (Pt) and carbon black supports. Catalyst degradation caused by the corrosion of carbon supports and platinum dissolution under fuel cell operation conditions leads to extensive degradation of fuel cell performance [4]. Therefore, highly stable electrocatalysts are required to increase PEMFC long-term performance.

One approach to improve catalyst durability is to replace the carbon support with ceramic-based materials such as titanium oxide [5], niobium-doped titanium oxide [6, 7], niobium oxide [8], titanium nitride [9,10,11], titanium oxynitride [12] and titanium–niobium nitride [13]. In general, ceramic supports, based on oxide materials, have high corrosion resistance, excellent thermal stability, and low cost. However, their low electrical conductivity is a barrier to their use in PEMFC [5, 6, 14]. On the other hand, titanium nitride-based materials have much higher conductivity than oxides and are more electrochemically stable and more robust than carbon supports. Therefore, they have been investigated as electrocatalyst supports [10, 11, 15, 16]. Avasarala et al. [4, 15, 16] evaluated the durability and stability of TiN as Pt catalyst support for PEMFC and demonstrated that TiN supports have higher corrosion resistance than carbon supports. However, the overall increase in durability of the electrocatalyst Pt/TiN compared to Pt/C is still inadequate due to Pt agglomeration [15] and, therefore, alternative catalysts are required.

One approach for improving Pt stability is a surface treatment of the TiN nanoparticles to avoid Pt agglomeration. For instance, the preliminary results of Avasarala et al. [15] suggested that a heat treatment in an N2 atmosphere between 70 and 120 °C increases the oxynitride layer thickness. The oxynitride layer can encapsulate the Pt catalyst reducing Pt particle mobility and agglomeration. This oxynitride layer formed at TiN nanoparticles has conductive nature but it dissolves in acidic electrolytes so that its contribution to decreased electron transport during the initial cycles was minimal. The thermally treated TiN showed an increased stability; however, a relatively lower electrochemical active surface area (ECSA). Shin et al. [17] improved conductivity and durability of TiN supports using a nanotube structure as support for Pt catalysts. The nanotube structure was designed as a grain-boundary free scaffold to reduce ohmic losses. According to Shin et al. [17], the strong metal support interaction (SMSI) positively affected the electrode durability once the electrons moved from the TiN support to Pt 5d band. Furthermore, the TiN surface barely changed after durability tests.

Platinum is the most commonly used electrocatalyst for PEMFC in part due to its excellent catalytic activity. However, the high cost of Pt and its limited availability has restricted the widespread use of PEMFC technology. Alloying Pt with a non-noble metal is an effective strategy for lowering Pt loading, decreasing PEMFC costs, as well as improving ORR activity. Recently, Goshal et al. [18], and Jia et al. [19] demonstrated that platinum and niobium form an alloy with a high potential to enhance ORR activity. Pt interacts with Nb and NbOx, and Nb contributes to the suppression of Pt dissolution due to strong ligand effects, which enhance catalytic activity toward ORR [19]. Pt–Nb/NbOx loaded on carbon was tested as an electrocatalyst and demonstrated high activity and durability for both ORR and hydrogen oxidation reaction (HOR) [18]. This new electrocatalyst has significant potential for further improvement but more studies are required to better understand the Pt–Nb and Pt–NbOx interactions and tailor them to improve the catalyst’s performance. A promising approach could be combining the Pt–Nb electrocatalyst with a non-carbon support such as TiN mentioned above.

Methods for the production of Pt-based electrocatalyst including precipitation of Pt [5, 15, 18], microwave-assisted synthesis [20], electrodeposition [21, 22], as well as thin film techniques like magnetron sputtering [8, 23], arc plasma deposition [19], and flame-based processes [24, 25] have been reported. Reactive spray deposition technique (RSDT) has been highlighted as an open-atmosphere, flame-based deposition method because it combines catalyst synthesis and electrode fabrication into a single step. RSDT is a breakthrough process for the deposition of very small catalyst particles (nanoscale) at significantly lower production costs and high efficiency [26, 27]. It also allows the control of the catalyst, support, and ionomer composition; therefore, it is suitable for cost-effective production of low-Pt-loading electrocatalysts [28, 29].

Here, we have investigated Pt–NbOx supported on TiN as the electrocatalyst for ORR. To the authors’ knowledge, this is the first time that such a system is investigated as an electrocatalyst for ORR. The electrocatalyst nanoparticles were synthesized by RSDT. The physical properties were characterized by transmission electron microscopy (TEM), scanning electron microscopy (SEM), inductively coupled plasma optical emission spectroscopy (ICP-OES) and X-ray diffraction (XRD). X-ray photoelectron spectroscopy (XPS) was performed to evaluate the metal–metal and metal–metal oxide interactions on the electrocatalyst surface. The electrochemical performance was evaluated for ORR using rotating disk electrode (RDE).

Experimental

Synthesis of Pt–NbOx/TiN catalyst

Titanium nitride nanoparticles of an average particle size of 20 nm (97% basis, SkySpring Nanomaterials Inc.), niobium (V) ethoxide (99.95% basis, Sigma-Aldrich), and platinum acetylacetonate (98% basis, Colonial Metals Inc.) were used as precursor materials. The Pt–NbOx precursor solution was prepared such that the Nb:Pt molar ratio in the final solution is 1:1. The Nb and Pt precursors (0.64 wt% total concentration) were dissolved in a solvent mixture composed of 23.6 wt% methanol (ACS grade, Fisher Scientific) and 59.2 wt% xylene (ACS grade, Fisher Scientific) by sonicating in a bath sonicator for 60 min at room temperature. 16.6 wt% of liquid propane (2.0 grade, Airgas) was then added to the solution. The TiN slurry was prepared by dispersing the TiN nanoparticles along with 0.002 wt% Nafion™ in a solution of isopropanol (ACS, Fisher Scientific) and DI water (1:1 by volume). TiN slurry concentration was 0.5 wt% for 5.5% Pt–NbOx/TiN samples and 0.15 wt% for 26.6% Pt–NbOx/TiN samples.

TiN-supported electrocatalysts were synthesized by RSDT and characterized without any additional processing. A detailed description of the RDST set-up and process can be found elsewhere [27, 30]. Pt–NbOx loaded on TiN nanoparticles were deposited onto multiple substrates including Nafion™ 211 membranes (Ion Power), polished silicon wafers (Nova Electronic Materials), polypropylene membranes, and 5-mm-diameter hands-free glassy carbon inserts (Pine instruments). The substrates were taped on a 7.5-cm2 aluminum sample holder as described by Yu et al. [28]. The precursor solution was heated to a set point of 190 °C, and atomized through the primary nozzle of the RSDT at a constant flow rate of 4 mL min−1. The TiN slurry was sprayed through a set of two secondary spray nozzles (Nordson EFD) at constant flow rate of 1.5 mL min−1. A quench ring, flowing air at a flow rate of 100 SLPM was introduced at a distance of 9.7 cm from the primary nozzle to maintain the substrate temperature in the range of 80–100 °C during deposition. Deposition time was 30 min for 5.5% Pt–NbOx/TiN samples and 120 min for 26.5% Pt–NbOx/TiN. RSDT parameters are summarized in Table 1. For the deposition of TiN without the catalyst, a precursor solution was prepared using the same solvent composition as the Pt–NbOx/TiN catalyst, but without the addition of the Pt and Nb precursors to the solvent mix. Unsupported Pt–NbOx samples (100% Pt–NbOx in Table 1) were prepared under deposition conditions similar to those of the supported catalysts but without any secondary nozzles spraying TiN. Electrocatalysts deposited on Nafion™ membranes and silicon wafers were used for microstructural characterization, while the material deposited on polypropylene was used for compositional analysis. The film deposited on glassy carbon inserts was used directly for RDE experiments.

Table 1 Summary of experimental condition for RSDT
Full size table

Physical characterization

Sample microstructure was characterized by Scanning Electron Microscopy (SEM) using Thermo Fisher Quanta 250 FEG SEM. For the cross section, the analysis on the electrocatalyst layer deposited on Nafion™ membranes was freeze fractured using liquid nitrogen. Sample composition was analyzed by X-ray–energy-dispersive spectroscopy (EDS). Phase composition was investigated by X-ray diffraction (XRD) using Bragg–Brentano geometry (θ–2θ) in a Bruker D8 advance X-ray diffractometer. Lattice constant was calculated using Bragg’s law \(\left( {d_{{\left[ {111} \right]}} = \frac{{a_{o} }}{{h^{2} + k^{2} + l^{2} }}} \right)\) for the Bragg’s angle of the (1 1 1) diffraction plane of XRD spectra of TiN. The crystallite size was calculated using the following Scherrer equation:

$$ B\left( {2\theta } \right) = \frac{K\lambda }{{L\cos \theta }}, $$
(1)

where B is the width of half the diffraction peak, L is the crystallite size, λ is the diffracted wavelength of X-ray (0.154056 nm), K is the Scherrer constant (0.94 for spherical particles), and θ is the diffraction angle.

Particle size and particle size distribution were evaluated by analyzing the transmission electron microscopy (TEM) images with ImageJ software [31]. TEM images were obtained in a Thermo Scientific Talos F200X S/TEM at 200 kV. The distribution of Pt and Nb was analyzed by STEM-EDS mapping using the Super-X EDS system (Bruker). TEM specimen were prepared by dispersing deposited material in spectroscopy grade isopropanol (Sigma-Aldrich) for 30 min using a bath sonicator and putting a 10 μL drop of the dispersed solution on a carbon-coated 3 mm Cu grid (300HD, Pacific Grid Tech). The droplet was covered and allowed to dry overnight under ambient conditions.

The platinum and niobium loadings were determined by ICP-OES using a Perkin Elmer Optima 7300DVICP-OES. The electrocatalyst surface was analyzed by X-ray photoelectron spectroscopy (XPS) using Phi 510 X-ray photoelectron spectrometer (radiation source: Al Kα, λ = 1486.6 eV). The binding energies reported in this study were charge-corrected using the adventitious carbon 1 s binding energy at 284.8 eV as reference. The XPS data was processed using CasaXPS software (Casa Software Ltd). Micro-Raman spectroscopy was used to examine the TiN support for evidence of partial oxidation. A Renishaw System 2000 with a 785-nm laser source was used for the micro Raman analysis at a spatial resolution of ~ 1 µm.

Electrochemical characterization

The ORR activity was evaluated using a rotating disk electrode (RDE). The electrochemical measurements were conducted in a custom-designed three-compartment electrochemical cell (Adams & Chittenden Scientific Glass, CA). Nitrogen- or oxygen-purged 0.1 M perchloric acid (double distilled, GFS Chemicals. Inc.) was used as the electrolyte. The potential was determined using the Hg/HgSO4 reference electrode with a potential of 0.72 V vs. RHE. A Pt wire was used as counter electrode. The working electrode was mounted on an MSR rotator (AFMSRCE, Pine Instruments). A Solartron 1287 potentiostat was used for all measurements. Prior to the measurement, the working electrode was electrochemically cleaned by cycling between 0.0 V and 1.4 V (vs. RHE) at a scan rate of 200 mV s−1 for 40 cycles. The polarization curves for ORR were recorded in linear sweep voltammetry scanning from 0.05 to 1.05 V vs. RHE with scan rate of 20 mV s−1 [32]. The performance was recorded with rotation speeds of 400, 900, 1600, and 2500 rpm. The background current was measured by running ORR sweep in a nitrogen-purged 0.1 M perchloric acid solution at same protocol applied to collect ORR curves in O2 statured electrolyte. The background current was subtracted from ORR to eliminate any capacitive current contribution. The high-frequency resistance (HFR) was measured using electrochemical impedance spectroscopy (EIS) recorded at 0.9 V vs RHE with 1600 rpm rotating speed to obtain the iR corrections. Cyclic voltammograms (CVs) were collected in nitrogen-purged cells between 0.05 and 1.2 V (vs. RHE) at a scan rate of 20 mV∙s−1 for 30 cycles until repeatable (the 5 last cycles overlapped). Only the last cycle is shown. Tafel plots (E versus log\(\left| {J_{{\text{k}}} } \right|\)) were obtained from corrected ORR data collected at 1600 rpm by calculating the kinetic current (Ik) according to the following equation:

$$ I_{{\text{k}}} \left( {\text{A}} \right) = \frac{{I_{{\lim}} X I({\text{A}})}}{{(I_{{\lim}} - I({\text{A}}))}}, $$
(2)

where Ilim is the measured limiting current (A) and I is the measured current (A).

The electrochemical active surface area (ECSA) was calculated using the CV data collected at a scan rate of 20 mV s−1 according to the following equation for hydrogen adsorption:

$$ {\text{ECSA}}_{{{\text{Pt}}}} \left( {{\text{m}}^{2} {\text{g}}_{{{\text{Pt}}}}^{ - 1} } \right) = \left[ {\frac{{Q_{{\text{H-adsorption}}} \left( C \right)}}{{210 \upmu {\text{Ccm}}_{{{\text{Pt}}}}^{ - 2} \left( {{\text{mg}}_{{{\text{Pt}}}} {\text{cm}}^{ - 2} } \right)A_{g} \left( {{\text{cm}}^{2} } \right)}}} \right]10^{5} . $$
(3)

The electron transfer number (n) was calculated by applying the Koutecky–Levich equation as follows:

$$ \frac{1}{i} = \frac{1}{{i_{k} }} + \left( {\frac{1}{{0.62nFAD^{2/3} v^{ - 1/6} w^{1/2} }}} \right), $$
(4)

where A is the geometric area of the disk (cm2), F is Faraday’s constant (C mol−1), D is the diffusion coefficient of O2 in the electrolyte (cm2 s−1), v is the kinematic viscosity of the electrolyte (cm2 s−1), C is the concentration of O2 in the electrolyte (mol cm−3), ω is the angular frequency of rotation (rad s−1) and the kinetic current (ik) is the inverse of the intercept of the linear Koutecky–Levich plot with the Y-axis.

The Pt area-specific activity (\(I_{{\text{s}}}\)) was estimated using the kinetic current (\(I_{{\text{k}}}\)) calculated from the linear Koutecky–Levich plots and normalization to the Pt electrochemical surface area using

$$ I_{{\text{s}}} \left( {\upmu A\;{\text{cm}}_{{{\text{Pt}}}}^{ - 2} } \right) = \frac{{I_{{\text{k}}} \left( {\text{A}} \right)}}{{Q_{{{\text{H}} = {\text{adsorptin}}}} /210 \upmu {\text{Ccm}}_{{{\text{Pt}}}}^{ - 2} }} . $$
(5)

The mass activity was calculating (\(I_{{\text{m}}}\)) according to the following equation:

$$ I_{{\text{s}}} \left( {{\text{A}}\;{\text{mg}}_{{{\text{Pt}}}}^{ - 1} } \right) = \frac{{I_{{\text{k}}} \left( {\text{A}} \right)}}{{{\text{mg}}_{{{\text{Pt}}}} }}. $$
(6)

More details about the methodology for ECSA analysis and Is calculation can be found in the literature [32].

Results and discussion

Synthesis and physical characterization of Pt–NbOx/TiN catalyst

We successfully synthesized Pt–NbOx/TiN catalyst in a one-step process using RSDT. RSDT was capable of depositing a homogeneous Pt–NbOx/TiN catalyst layer onto Nafion™ membranes (Fig. 1). A catalyst content of 5.5 wt% with a Pt loading of 0.032 mg cm−2 and a thickness of ca. 4.5 µm was obtained using a TiN slurry concentration of 0.5 wt% and a deposition time of 32 min (Fig. 1a, b). By increasing the deposition time to 60 min and decreasing the TiN slurry concentration to 0.15 wt%, an electrocatalyst layer of ca. 3.5 μm thickness, catalyst content of 26.6 wt% with a Pt loading of 0.077 mg cm−2 was obtained on Nafion™ membrane (Fig. 1c, d). RSDT deposition of the support TiN nanoparticles (without Pt–NbOx particles) on Nafion™ resulted in a homogenous layer of ca. 2.6 μm thickness (Fig. 1e, f). In the case of the 100% Pt–NbOx electrocatalyst deposited on Nafion™ membrane, a rough layer of < 500 nm thickness was obtained (Fig. 1g, h).

Fig. 1
figure 1

SEM image of a 5.5% Pt–NbOx/TiN surface and b cross section, c 26.6% Pt–NbOx/TiN surface and d cross section, e TiN surface and f cross section and g 100% Pt–NbOx surface and h cross section

Full size image

Bright-field (BF) TEM images of the Pt–NbOx (Fig. 2a, b) show a relatively homogeneous dispersion of Pt–NbOx catalyst particles on TiN support. The average particle diameter was determined to be 1.6 nm with size distribution ranging from 0.6 to 2.9 nm (Fig. 2c). HAADF imaging and STEM-EDS elemental mapping, as shown in Fig. 2d–f, revealed clear signs of overlap between Pt- and Nb-rich regions suggesting good mixing and uniform distribution of Pt and Nb on TiN support.

Fig. 2
figure 2

Bright-field (BF) TEM image of a 26.6% Pt–NbOx/TiN, b 100% Pt–NbOx electrocatalyst, c particle size distribution, d HAADF TEM image of Pt–NbOx nanoparticles, e EDS mapping of platinum and f EDS mapping of Nb

Full size image

XRD pattern (Fig. S1) confirmed that TiN particles were successfully deposited by RSDT. The absence of peaks related to Pt and NbOx phases can be related to their small particle size and the likely low crystallinity. Diffuse rings observed in selected area electron diffraction (SAED) pattern from TEM analysis (Fig. S2) confirmed the low crystallinity of Pt–NbOx nanoparticles produced by RSDT. The peak position of TiN peaks in the XRD pattern of Pt–NbOx/TiN deposited by RSDT on Si shifted slightly when compared to the as-received TiN precursor nanoparticles (Table 2). The lattice constant calculated from Bragg’s Law was 4.241 Å for the precursor TiN nanoparticles and 4.229 Å for TiN nanoparticles after RSDT deposition. The lattice constant of the precursor TiN nanoparticles is the same constant reported for the cubic TiN phase (4.241 Å), while the lattice constant of RSDT TiN nanoparticles is between the cubic TiN phase (4.241 Å) and TiO phase (4.185 Å) [33]. Titanium oxynitride (TiOxNy) is an intermediate phase and it can be considered a solid solution of titanium nitride (TiN) and a cubic titanium oxide (TiO) [4, 34]. The lattice constant of RSDT TiN nanoparticles indicates there is replacement of some nitrogen atoms for oxygen atoms in the surface of TiN nanoparticles, which corroborates to a formation of oxynitride near the particle surface during RSDT deposition. As the titanium oxide is more thermodynamically stable than titanium nitride, an oxide or oxynitride layer is expected to form at the surface of TiN nanoparticles at room temperature, therefore, even under the relatively mild RSDT deposition conditions (80 °C) the potential oxidation of TiN is expected.

Table 2 XRD peak position, crystallite size, and lattice constant
Full size table

Raman Spectroscopy analysis (Figs. S3 and S4) corroborated with XRD results. Raman bands of cubic TiN were shown in three regions: the first in between 150 and 250 cm−1 corresponding to transverse acoustic mode, the second in between 250 and 350 cm−1 corresponding to longitudinal acoustic mode and the third in between 450 and 700 cm−1 corresponding to transverse optical mode in accordance with literature [35,36,37,38]. Raman lines of TiN nanoparticles processed by RDST were slightly shifted to lower frequencies compared to as purchased TiN nanoparticles (Fig. S3), which is expected to be a result of a decrease in the lattice parameter [9].

Electronic structure of Pt–NbOx/TiN catalyst

XPS measurements were performed to evaluate the elemental composition and oxidation state of the surface of TiN and Pt–NbOx/TiN catalyst (Figs. 3, 4, 5, 6, 7, 8). The XPS survey of TiN nanoparticles proceeds by RSDT (Fig. 3a) which indicates that the majority of atoms near the surface are titanium (25.3 wt%), fluorine (24.7 wt%), carbon (15.3 wt%), oxygen (14.3 wt%) and nitrogen (6.1 wt%). Fluorine and carbon come from the slurry dispersion, which contains Nafion™ and titanium and nitrogen from TiN nanoparticles. High-resolution XPS spectra were collected for the binding energy of Ti 2p, N 1 s and O 1 s (Figs. 4, 5, 6) and the peak assignments are shown in Table S1. Ti 2p spectra showed that Ti is present as TiN, TiOxNy, and TiO2. The peak assignments of Ti 2p3/2 and Ti 2p1/2 shown in Table S1 were based on the literature [39,40,41]. This result suggests that the surface of TiN nanoparticles underwent oxidation during RSDT deposition, which is also supported by the results of XRD analysis (Table 2) and Raman spectroscopy (Fig. S3).

Fig. 3
figure 3

XPS survey scan spectra of a TiN, b 5.5% Pt–NbOx/TiN and c 26.6% Pt–NbOx/TiN electrocatalysts

Full size image
Fig. 4
figure 4

High-resolution XPS spectra showing Ti 2p of a TiN, b 5.5% Pt–NbOx/TiN and c 26.6% Pt–NbOx/TiN electrocatalysts

Full size image
Fig. 5
figure 5

High-resolution XPS spectra showing N s of a TiN, b 5.5% Pt–NbOx/TiN and c 26.6% Pt–NbOx/TiN electrocatalysts

Full size image
Fig. 6
figure 6

High-resolution XPS spectra showing O s of a TiN, b 5.5% Pt–NbOx/TiN and c 26.6% Pt–NbOx/TiN electrocatalysts

Full size image
Fig. 7
figure 7

High-resolution XPS spectra showing Pt 4f of a 5.5% Pt–NbOx/TiN and b 26.6% Pt–NbOx/TiN electrocatalysts

Full size image
Fig. 8
figure 8

High-resolution XPS spectra showing Nb 3d of a 5.5% Pt–NbOx/TiN and b 26.6% Pt–NbOx/TiN electrocatalysts

Full size image

The XPS survey of 5.5% Pt–NbOx/TiN (Fig. 3b) indicates that the majority of atoms near the surface are titanium (29.7 wt%), with catalysts Pt and Nb present in smaller amounts: 2.5 and 2.1 wt% respectively. High-resolution XPS spectra were collected for the binding energy of Ti 2p, N 1 s, O 1 s, Pt 4f and Nb 3d, (Figs. 4, 5, 6, 7, 8) and the peak assignments are shown in Table S1. The Pt 4f spectra (Fig. 7) were fit with peaks corresponding to the 4f7/2 and 4f5/2 states. The binding energy of the Pt 4f7/2 dominant peak was 72.4 eV, which is close to the value reported for PtO [42], the binding energy of the smaller peak (73.9 eV) is slight lower than the 74.4 eV reported for PtO2 [42, 43]. The formation of Pt–O bonds is a result of the electron transfer from platinum to Nb2O5 [44]. The geometrical mismatch between Pt and Nb2O5 and the small size of Pt nanoparticles contribute to the strong metal–metal oxide interactions [19, 45]. The 5.5% Pt–NbOx/TiN electrocatalyst has higher ratio of TiN support to Pt–NbOx particles and the Pt–TiN interaction has major role. As a result, B.E. shifts towards lower energy due to charge transfer from the TiN support to platinum nanoparticles as reported by Roca-Ayats et al. [46, 47].

Nb 3d spectrum is a simple spin–orbit doublet with peaks corresponding to Nb3d5/2 and Nb3d3/2 (Fig. 8). The binding energy of the main peak at 208.06 eV agrees with Nb5+, which has a binding energy of 207.8 eV, as reported by Morris et al. [48]. The small size of Pt–NbOx nanoparticles contributes to shift Nb3d5/2 peak to higher energy due to the screening effect of core holes and lattice strain as previously reported in the literature [49, 50]. Nb5+ was the dominant oxidation state, which suggests that Nb is mainly present as Nb2O5 near the surface. Ti 2p spectra showed that Ti is present as TiN, TiOxNy, and TiO2 similar to the observation of the TiN nanoparticles processed by RSDT (Fig. 4).

As expected, the XPS survey scan of 26.6% Pt–NbOx/TiN (Fig. 3c) showed a higher concentration of Pt and Nb atoms compared to 5.5% Pt–NbOx/TiN. The absence of titanium and nitrogen peaks indicates that the Pt–NbOx catalyst is covering the surface of the TiN support particles. High-resolution XPS spectra of 26.6% Pt–NbOx/TiN were also collected for the binding energies of Ti 2p, N 1 s, O 1 s, Pt 4f and Nb 3d (Figs. 4, 5, 6, 7, 8). Ti 2p (Fig. 4) and N 1 s (Fig. 5) spectra had a significant decrease in intensity when compared to 5.5% Pt–NbOx/TiN, which was related to the higher catalyst to support ratio and, as mentioned previously, to the fact that Pt–NbOx catalyst are covering the TiN support particles. Furthermore, there was a displacement of peak position of Ti 2p, O 1 s, and N 1 s spectra of 26.6% Pt–NbOx/TiN catalyst compared to 5.5% Pt–NbOx/TiN and TiN (Table S1). The changes in O 1 s spectra are likely due to the higher amount of Nb2O5 in 26.6% Pt–NbOx/TiN, which leads to higher Nb2O5 signal. The change in Ti 2p and N 1 s is an indicative of increased concentrations of TiO2 and TiOxNy. During RSDT deposition, the substrate is heated in the range of 80–90 °C. This temperature is enough to promote surface oxidation of TiN nanoparticles [15]. As 26.6% Pt–NbOx/TiN catalysts were deposited for a longer time, the TiN particles were subjected to elevated temperatures for a longer duration and were expected to have formed thicker surface oxide layers. This was confirmed by the shift of the Raman lines to low frequencies in RSDT processed TiN nanoparticles (Fig. S3). The formation of the very stable oxynitride layer at TiN surface is expected to improve corrosion resistance and prevent further oxidation of TiN support during PEMFC operation. Therefore, TiN supports have the potential to improve support durability when compared to commercially available carbon-based supports [15]. The stability of the oxynitride layer on TiN was confirmed by analyzing samples after they had undergone electrochemical testing. Raman spectra of the samples after electrochemical testing showed no indication of significant change in oxide content (Fig. S4).

There is a slight change in the peak profile and a displacement of peaks positions to higher binding energy in the Pt 4f spectra of the 26.6% Pt–NbOx/TiN electrocatalyst compared to the 5.5% Pt–NbOx/TiN electrocatalyst (Table S1). The 26.6% Pt–NbOx/TiN electrocatalyst has smaller ratio of TiN support to Pt–NbOx nanoparticles and the Pt–TiN interaction has no significant effect on the peak position. On the other hand, the Pt–NbOx interaction and small size of Pt nanoparticles play a major role resulting in the Pt 4f spectra shift towards higher energy [49, 50].

Electrochemical performance

RDE test was performed to study the electrochemical behavior of the Pt–NbOx/TiN catalyst synthesized by the RSDT process. As the new catalyst has the potential for cathode catalyst application, ORR was investigated. ORR polarization curves for Pt–NbOx/TiN electrocatalysts are shown in Fig. 9a. A half-wave potential in the ORR polarization curves of 790 mV was achieved for 5.5% Pt–NbOx/TiN samples (Pt loading of 0.032 mg cm−2) and 910 mV for 26.6% Pt–NbOx/TiN (Pt loading 0.077 mg cm−2). As expected, the half-wave potential increased with increased Pt loading, and a higher potential of those reported for Pt/C (880–900 mV) [32, 51,52,53] was achieved. The specific activity calculated with kinetic current ik at 0.9 V vs RHE for the 26.6% Pt–NbOx/TiN electrocatalyst was found to be 283 μA∙cmPt−2, which is also higher than those reported for Pt/C produced by RSDT at similar processing parameters [51] (Table 3). The higher half-wave potential and specific activity indicate that the modification in the structural and electronic state of Pt due to strong metal–metal oxide interactions among Pt, NbOx and that TiN positively affects catalytic activity toward ORR, which is in good alignment with the findings from Zheng et al. [8], Goshal et al. [18], Jia et al. [19] and Chinchila et al. [54] for the Pt–NbOx interactions and Ding et al. [9] and Zheng et al. [55] for Pt–TiNxOy interactions.

Fig. 9
figure 9

ORR polarization curves a and Tafel plots b versus RHE recorded at 1600 rpm, at room temperature in O2-saturated 0.1 M HClO4

Full size image
Table 3 Activity for oxygen reduction reaction obtained by RDE
Full size table

The mass activity calculated from ORR curves and Pt loading of the catalyst (Table 3) was found to be 0.0194 and 0.120 A∙mgPt−1 for the 5.5% Pt–NbOx/TiN and 26.6% Pt–NbOx/TiN electrocatalysts, respectively, which remain smaller than the 0.157 A∙mgPt−1 and 0.21 A∙mgPt−1 reported for Pt/C [32, 51]. In spite of the lower mass activity, it is expected that the Pt–NbOx/TiN electrocatalyst will be more durable than the Pt/C catalyst and will keep its mass activity after many cycles. Because TiN oxidation forms TiO2, which is solid, while C oxidation forms a gas CO2, which leads to severe loss of the carbon support and the mass activity decreases due to the isolation of the Pt catalyst particles [15].

Two distinct Tafel regions, which are typical for the ORR at Pt electrocatalysts in acidic media [56], were observed (Fig. 9b). For the 5.5% Pt–NbOx/TiN electrocatalyst, Tafel slopes of 72 mV dec−1 and 114.8 mV dec−1 were found at high (0.9–0.94 V) mV dec−1 and low (0.82–0.86 V) potentials, respectively. While, for the 26.6% Pt–NbOx/TiN electrocatalyst, Tafel slopes of 62 mV dec−1 and 122 mV dec−1 were found at high (0.9–0.94 V) and low (0.80–0.84 V) potential, respectively. The ORR on Pt in acid media follows a dissociative adsorption of O2, which is the rate-determining step. According to Banham et al. [56], the two distinct Tafel regions indicate that the rate-determining step (RDS) has changed from a rapid first electron transfer step at low potential to a slow first rapid first electron transfer step at high potential. At low potentials, the first electron step is the adsorption of reaction intermediates following Temkin isotherm. At high potential, the adsorption of reaction intermediates follows Langmuir isotherm as the catalyst surface is oxide free; thus, a different Tafel slope is measured.

The mass-corrected transport Tafel plots (Fig. 9b) and Tafel slopes of Pt–NbOx/TiN electrocatalysts were similar to those reported for Pt/C (Table 3), indicating no change in the ORR mechanism. The total number of electrons transferred (n) during ORR was determined from Koutecky–Levich plots (Fig. S5) and it was found to be 3.9 at 0.4–0.7 V for 5.5% Pt–NbOx/TiN and 3.7 at 0.3–0.85 V for 26.6% Pt–NbOx/TiN. The values of n were close to 4, which confirms the ORR mechanism is similar to the Pt/C-based electrocatalyst.

The voltammetric profile of Pt–NbOx/TiN electrocatalysts is shown in the cycle voltammograms (CV) in Fig. 10. The catalyst with a 5.5% Pt–NbOx/TiN composition had limited hydrogen adsorption and desorption, so that they behaved similarly to pure TiN [39]. The poor catalytic activity is due to the low catalytic surface area which is suspected to result from the presence of relatively fewer number of Pt–NbOx particles which was evident from the XPS measurements. Nevertheless, the catalyst with a 26.6% Pt–NbOx/TiN composition exhibits a typical CV with distinct H adsorption/desorption peaks (Pt + H3O +  ↔ Pt − Had + H2O) between 0.05 and 0.35 V. The higher activity is related to higher Pt loading in 26.6% Pt–NbOx/TiN catalyst.

Fig. 10
figure 10

Cycle voltammograms of 5.5% Pt–NbOx/TiN and 26.6% Pt–NbOx/TiN versus RHE collected at room temperature in N2 statured 0.1 M H ClO4 at sweep of 20 mV s−1

Full size image

The electrochemical surface area (ECSA) derived from H adsorption was found to be 42.4 m2gPt−1 for 26.6% Pt–NbOx/TiN (Table 3). ECSA was not calculated for 5.5% Pt–NbOx/TiN because there is no hydrogen desorption peak in CV (Fig. 10). The lower ECSA of 26.6% Pt–NbOx/TiN in comparison with the values of Pt/C reported in the literature [51,52,53] is most likely due to the presence of niobium oxides in the surface as well as the oxidation of TiN support. The formation of an oxide layer in TiN support can partially cover the catalyst particles. On the other hand, this oxide layer may encapsulate Pt particles and thus avoid Pt migration and agglomeration during PEMFC operation thus increasing the catalyst’s lifetime. Long-term tests are planned to investigate the durability of the catalyst. A compromise must be found between surface-active area and durability. The higher ORR activity in spite of low ECSA may be attributed to Pt and NbOx interactions as discussed by Goshal et al. [18]. Nb oxide is expected to affect the activity of Pt due to strong metal–metal oxide interaction, thus increasing electrochemical stability.

The electrochemical performance (half-wave potential in ORR curves and H desorption in the CV curves) increases after running a few RDE tests. One hypothesis is that oxides, formed on the catalyst surface, dissociate at certain potentials, which increase Pt surface area and, therefore, catalytic activity. The remaining challenge is the TiN oxidation, which tends to decrease electrical conductivity as previously reported by Qiu et al. [60]. One alternative to improve conductivity is optimizing support particle shape. For instance, in the literature, the use of nanorods [61] or nanotubes [9, 17] is reported as a strategy to improve conductivity of the Pt/TiN electrocatalyst by increasing contact between TiN particles.

Conclusions

In this work, we reported a new electrocatalyst based on Pt–NbOx supported on TiN nanoparticles. The capability of Pt–NbOx/TiN catalyst fabrication in a single step by RSDT was demonstrated. TEM and XRD confirmed the synthesis of a well-dispersed Pt–NbOx catalyst with an average particle size of 1.6 nm diameter supported on TiN particles with a 20 nm diameter. The electrochemical performance was evaluated for ORR using the RDE method. Strong metal–metal oxide interactions between Pt and Nb oxide as well as between Pt and TiNxOy were observed in XPS analysis and may contribute to avoid Pt dissolution. A half-wave potential of 910 mV and a specific activity of 283 μA∙cmPt−2 at 0.9 V was achieved in the new Pt–NbOx/TiN electrocatalyst with a composition of 26.6% Pt–NbOx/TiN which is higher than those reported for the state-of-the-art Pt/C. However, the ECSA of Pt–NbOx/TiN electrocatalyst is still lower than the state-of-the-art Pt/C electrocatalyst. Therefore, based in previous studies about Pt/TiN systems, future work on the use of TiN particles with nanotubes or nanowire shapes is suggested as an alternative to increase ECSA.

This is the first report on a Pt–NbOx/TiN electrocatalyst. The results achieved so far demonstrate that Pt/Nb–Nb2O5 loaded on TiN supports is a promising electrocatalyst for ORR in PEMFC. Further investigations to evaluate long-term performance as well as the performance of this catalyst in PEMFC tests will be presented in an ongoing study. As TiN is a more robust and stable support than carbon supports, it is expected that the Pt–NbOx/TiN electrocatalyst will have higher durability than the carbon-supported electrocatalyst. Furthermore, the strong metal–metal oxide interactions between Pt and NbOx and TiNxOy also require further analysis to optimize electrocatalyst performance and durability.