Structurally-Tuned Nitrogen-Doped Cerium Oxide Exhibits Exceptional Regenerative Free Radical Scavenging Activity in Polymer Electrolytes

Four different CeO₂ samples were employed in this study: (i) commercial CeO₂ (ii) high-surface-area CeO₂ (iii) N-doped commercial CeO₂ and (iv) N-doped high-surface-area CeO₂. Commercial CeO₂ was obtained from Sigma-Aldrich and high-surface-area CeO₂ was synthesized in-house via a hydrothermal precipitation route using aqueous ammonia and Cerium (III) chloride heptahydrate.^4,60,61 See Figure S1(a) under supporting information (SI). The commercial CeO₂ and as-prepared high-surface-area CeO₂ were doped via exposure to a nitrogen rich atmosphere. During the N-doping process, a polyaniline-CeO₂ mixture was prepared and the mixture was annealed at 900°C under flowing argon to yield N-doped CeO₂ samples⁵⁷ See Figure S1(b) under SI. Detailed procedures describing the synthesis of the high-surface-area CeO₂ and the N-doping of both variants of CeO₂ are provided in the SI section (I.a and I.b).

The BET surface area of all CeO₂ samples was measured using a Quantachrome NOVA 2200 BET analyzer. The Brunauer–Emmet–Teller (BET) method was used to calculate specific surface area from nitrogen sorption measurements performed at −196°C.

Transmission electron microscopy (TEM) was used to determine the particle size, lattice spacing and d-spacing (from small angle electron diffraction- SAED) of all CeO₂ samples. The nanoparticles were dispersed in ethanol, deposited onto carbon coated copper grids and dried at room temperature. TEM micrographs were obtained using a JEOL- 3010F electron microscope at an acceleration voltage of 300 kV equipped with a LaB₆ filament. Elemental analysis was performed for all CeO₂ samples using X-Ray Energy Dispersive Spectroscopy (XEDS) and the details are discussed under SI section IV (Figure S3).

X-ray photoelectron spectroscopy measurements were performed using an XPS spectrometer model Axis 165 (Kratos Analytical, Shimadzu Co.), with a monochromatic X-ray source of MgKα. Raman spectra were recorded from 100 to 1200 cm⁻¹ using a Renishaw Ramascope 2000. The calculation of Ce³⁺ surface concentration and surface oxygen vacancy concentration are presented under SI sections IV and V respectively.

The X-ray powder patterns of all CeO₂ samples were measured on a Bruker D2 Phaser diffractometer equipped with a LynxEye position-sensitive detector (15–100° 2θ, 0.0202144° steps, 0.5 sec/step, 0.6 mm divergence slit, 2.5E Soller slits, 3 mm scatter screen height). Quantitative phase analysis of the crystalline phases was carried out by the Rietveld method using the General Structure Analysis System (GSAS).⁶²

X-ray absorption spectroscopy measurements - which include XANES (X-ray absorption near edge structure) and EXAFS (extended X-ray absorption fine structure) - were recorded for all samples at room temperature in the transmission mode using synchronton radiation with a Si (111) double crystal monochromator at the 10 BM beam line of the Advanced Photon Source (APS) at Argonne National Laboratory. Fine powder samples were uniformly spread on a 50 μm thin Kapton tape and used for the measurements. The spectra were scanned in the range of 5500 eV to 6200 eV, which covered the L3 edge XAS region of cerium atoms. The photon energy was initially calibrated with the inflection point of the Ce L3 edge of commercial CeO₂. By satisfying the edge step (Δμx) of ≤ 1, the effect of sample thickness on the measurement was avoided. The thickness of the sample layers in the Kapton film was varied in each case to get a reasonable edge step.

In-situ fluorescence spectroscopy was employed to determine the regenerative ROS scavenging efficacy of all CeO₂ samples. Details of the in-situ fluorescence method and setup have been described in detail in our previous work, and are reproduced in the SI section (I.c - I.f).²⁴ The membrane electrode assemblies (MEAs) described in Section I.f were used in the in-situ experiments.

These in-situ experiments were performed with the following materials added to the Nafion PEM: i) 2 wt% of undoped commercial CeO₂ (ii) 5 wt% of N-doped commercial CeO₂ (iii) 0.5 wt% of undoped high-surface-area CeO₂ and (iv) 1 wt% of N-doped high-surface-area CeO₂. The weight fractions of CeO₂ in the membrane were varied since the CeO₂ samples employed in this study had significant differences in their BET surface area, and it was desired to maintain the total surface area of CeO₂ in the membrane constant at . The distribution of CeO₂ nanoparticles in the Nafion membrane were estimated by obtaining optical micrographs of the CeO₂ impregnated membranes (refer SI section II). The CeO₂ nanoparticles observed in all membrane configurations were well dispersed. These images provide the required evidence to showcase that the scavenging activity was derived from all the CeO₂ added to the Nafion membrane, and that there was no significant agglomeration of CeO₂ nanoparticles in the membrane. Furthermore, to ensure that the availability of CeO₂ active sites exceeded the total number of ROS generated in-situ, the ratio of total number of FRS active sites available in the CeO₂ samples to the total amount of ROS generated in-situ (per second) was estimated to be on the order of 10⁵. The calculation is shown in SI section III. Given that the FRS was regenerative, it was concluded that an adequate number of active scavenging sites were present in the membrane. Hence, the regenerative ROS scavenging efficacy inferred from in-situ fluorescence experiments are an intrinsic property of the respective CeO₂ samples. Two sets of in-situ experiments were performed: (i) The first set of experiments were conducted at 80°C and 75% RH, with H₂ and O₂ flowing at the rate of 0.1 lpm at the anode and cathode, respectively; and (ii) The second set of experiments were performed under much accelerated conditions at 90°C and 50% RH wherein the rate of ROS generation in-situ was much higher than at 80°C and 75% RH.

An in-situ fluorescence technique was chosen to monitor PEM degradation. We have established that the in-situ fluorescence method yields results that have a direct correlation with the more conventional fluoride emission rate (FER) measurements across all fuel cell operating conditions.⁶³

Ex-situ experiments were conducted to elucidate the stability of Ce³⁺ active clusters upon exposure to ROS (generated using Fenton's reagent). In the first set of experiments, both the N-doped CeO₂ samples were exposed to ROS, and the surface Ce³⁺ concentration and surface oxygen vacancy concentration were estimated subsequent to this treatment using XPS and Raman spectroscopy. See Figure S9. In the second set of experiments (which mimicked the conditions existing in an operating fuel cell), all the CeO₂ samples were exposed to ROS generated by Fenton's reagent in the presence of Nafion. Subsequent to exposure, the samples were analyzed by XAS to determine changes in the electronic structure and microstructure of the samples upon exposure to and interaction with ROS. A detailed experimental description is provided under SI (Section VI).

Transmission electron microscopy

TEM images and corresponding SAED patterns were recorded for all CeO₂ samples to probe their micro and nano-structure. See Figure 2. The average particle size (HRTEM), estimated lattice fringe spacing (HRTEM) and the d-spacing (SAED) obtained have been marked in the corresponding images in Figure 2. The highlights of the observations were as follows: Commercial CeO₂ and its N-doped version had particle sizes of 12 nm and 20 nm respectively. Both nanowire and nanocluster formations were seen in the high-surface-area CeO₂ as expected based on past reports.⁶⁴ The nanowires had an average diameter of 25 nm and the nanoactive clusters had an average particle size of 4 nm. It has been postulated in the literature that the presence of nanowires among the nanoactive clusters in other metal oxide systems improved the stability of the entire oxide structure.^65,66 It was hypothesized that the combination of nanoactive clusters and nanowires minimized the further ripening of CeO₂ particles upon high temperature annealing during the N-doping process.⁶⁶ The N-doped high-surface-area CeO₂ exhibited exclusively a nanocluster morphology, with an average particle size of 15 nm, which was 5 nm less than that of N-doped commercial CeO₂. The calculation of lattice fringe spacing using HRTEM images showed that commercial CeO₂ and its N-doped version had the same lattice fringe spacing of 3.0 Å. The high-surface-area CeO₂ and its N-doped version showed a larger lattice fringe spacing of ∼3.5 Å. The observed expanded lattice structure of the high-surface-area CeO₂ samples correlated well with the increase in lattice parameter seen in XRD analysis (discussed below). SAED patterns were analyzed for all CeO₂ samples and the d-spacing for the different lattice planes were indexed to the corresponding lattice planes of standard CeO₂ (JCPDS# 81–0792). Refer to Table S1 (under SI, Section V) for the list for parameters extracted from the TEM analysis. The correlation of d-spacing for different lattice planes in the CeO₂ crystal structure confirmed the crystallinity of all CeO₂ samples, in-line with XRD observations.

Zoom In Zoom Out Reset image size

It is well-known that doping CeO₂ with nitrogen increases the amount of Ce³⁺ sites, which are critical for scavenging ROS.⁵⁶ The concentration of Ce³⁺ sites reflects the availability of oxygen vacancies (i.e., defects) for ROS scavenging activity.^18,55 The change in the surface concentration of Ce³⁺ and in the surface oxygen vacancy concentration of CeO₂ before and after N-doping was studied using XPS and Raman spectroscopy.

(i) X-ray photoelectron spectroscopy

(ii) Raman spectroscopy

The surface oxygen vacancy concentration varies with the Ce³⁺ surface concentration. Raman spectroscopy was used to estimate the concentration of surface oxygen vacancies in each sample (see SI, Section VII, for details of the calculations). Raman spectroscopy was recorded for all CeO₂ samples in the interval 300 cm⁻¹ to 600 cm⁻¹ (See Figure 3a). A Raman active mode at 460 cm⁻¹ was observed in all samples, and was attributed to the symmetrical stretching mode of the Ce-O8 vibrational unit.^55,67,68 Figure 3b shows the Raman spectrum of N-doped high-surface-area CeO₂ scanned from 600 to 2400 cm⁻¹. The peak observed around 1300 cm⁻¹ was assigned to the longitudinal optic mode of Ce³⁺ from the excited 5d state in the conduction band to the 4f state of Ce; an additional peak at 1600 cm⁻¹ was attributed to nitrogen atom occlusion on the CeO₂ surface.⁵⁶

Zoom In Zoom Out Reset image size

Undoped commercial CeO₂ had a surface oxygen vacancy concentration of 1.9 × 10¹⁹ vacancies cm⁻³ while N-doped commercial CeO₂ had an enhanced surface oxygen vacancy concentration of 3 × 10²⁰ vacancies cm⁻³. The surface oxygen vacancy concentration of high-surface-area CeO₂ was 3 × 10²⁰ vacancies cm⁻³, which was equivalent to the surface oxygen vacancy concentration observed in N-doped commercial CeO₂. The N-doped high-surface-area CeO₂ showed a further enhancement in the surface oxygen vacancy concentration to 1.01 × 10²¹ vacancies cm⁻³. This analysis confirmed that N-doping of CeO₂ enhanced its surface oxygen vacancy concentration.

X-ray diffraction of CeO₂ samples

Figure 4 presents the X-ray diffractograms of the four CeO₂ samples evaluated. The diffraction peaks of all CeO₂ samples could be completely indexed to the cubic FCC structure of CeO₂ (JCPDS no. 81–0792). The average value of the cubic lattice parameter (a, Å) of stoichiometric CeO₂ is 5.4112 Å.⁶² The commercial CeO₂ sample and its N-doped version had lattice parameters slightly higher than this value (5.4123 Å and 5.4125 Å respectively). The high-surface-area CeO₂ and its N-doped version had significantly larger lattice parameters (5.4189 Å and 5.4192 Å respectively). See Table I. The expanded unit cell observed in high-surface-area CeO₂ was attributed to the increase in the concentration of Ce³⁺ active clusters and a corresponding increase in Ce-O distance (as ascertained from XPS and XAS analysis, respectively). The crystallinity of the CeO₂ samples was quantified using the scale factor determined by least squares refinement during Rietveld analysis. The scale factor is the constant factor that scales the calculated intensities of all reflections in the XRD pattern up to the observed intensities. The scale factor for commercial CeO₂ was the largest, and this sample was assigned a crystallinity of 100%. The crystallinity of all the other samples was compared against this benchmark. See Table I. It was postulated that the annealing process at 900°C (during N-doping) induced an increase in N-atoms introduced into the CeO₂ structure.⁶⁹ The increase in the N-atoms in CeO₂ surface resulted in an increase in the number of defects, leading to loss of crystallinity. The loss of crystallinity was more significant in N-doped high-surface-area samples, implying that phase transformation occurred on a more significant scale during the N-doping process on high-surface-area CeO₂.

Table I. Parameters calculated from analysis of XRD patterns of all CeO₂ samples.

Parameters	Commercial CeO2	N-doped commercial CeO2	High-surface-area CeO2	N-doped high-surface-area CeO2
a, Å	5.41233(9)	5.41258(7)	5.41899(37)	5.41920(38)
Scale	21.09(6)	19.94(7)	20.61(9)	11.46(11)
% Crystallinity.	100	94.7	97.7	54.3
Crystallize size, Å	330(1)	770(10)	156(3)	133(3)
Ce Uiso, Å2	0.0019(2)	0.0049(2)	−0.0001(4)	0.0055(7)

Zoom In Zoom Out Reset image size

The electron density at the Ce site was probed using the isotropic displacement coefficient (U_iso) calculated after Rietveld refinement.⁷⁰ See Table I. The value of U_iso for commercial CeO₂ was smaller than the corresponding value for the N-doped sample, indicating the formation of a lower electron density region in the vicinity of the nitrogen rich sites in the N-doped sample (likewise for the high-surface area samples). The formation of lower electron density regions in the lattice upon N-doping was in line with expectations as the nitrogen atoms inside the CeO₂ lattice replaced the more electronegative oxygen atom. This phenomenon was previously observed and reported in nitrogen-doped on carbon nanotubes.⁷¹ The difference Fourier maps were calculated using GSAS in an attempt to map the nitrogen atoms in the crystal structure of CeO₂. The possibility of interstitial inclusion of nitrogen atoms within CeO₂ lattice was completely ruled out as no acceptable coordinates for the interstitial placement of nitrogen atom could be obtained from the Fourier mapping calculation. (The interstitial placement of nitrogen atom based on the coordinates found by Fourier mapping yielded a Ce-N distance of less 2 Å which was not feasible).⁷⁰ The elimination of interstitial nitrogen suggested that the most likely structure involved the substitution of lattice oxygen atoms with nitrogen upon doping.

X-ray absorption spectroscopy

XAS was employed to probe the electronic transitions in the local atomic states of CeO₂. The spectra obtained from XAS can be subdivided into two regions, namely XANES and EXAFS. The XANES region primarily probes the low-lying discrete and extended electronic states, which provides insights into the local geometry and valence states of CeO₂. The EXAFS region probes the continuum states above the vacuum limit with the help of higher energy irradiation; this provides insights into the global geometry and the unoccupied electronic states in the material. The normalized XANES spectra of the cerium L3 edge and Fourier transformed EXAFS signals for all CeO₂ samples are shown in Figure 5. The proximity of the L2 edge of cerium to the L3 edge limited the utilizable energy range to 440 eV above the L3 edge. This complication limited the information that could be derived based on structure-dependent oscillations (e.g. Ce-neighbor distance). In this study, the main focus was placed on monitoring the electronic transitions observed in the different CeO₂ samples as a consequence of N-doping. The XANES spectra of all CeO₂ samples showed two distinct peaks (marked as I and II in Figure 5a) arising from the electronic transition of 2p_3/2 → 5d_5/2 and 5d_3/2, with the final state configuration describable as [*]4f¹L^n-15d¹ and [*]4f¹Lⁿ 5d¹.⁶⁹ Qualitative observation of the XANES spectra suggested that the first electron transition (2p_3/2 → 5d _5/2) was favored in the N-doped versions of CeO₂. This in turn suggested that more electronic transitions were occurring from the Ce³⁺ state, confirming the abundant existence of active Ce³⁺ active clusters in the N-doped samples. The decreasing participation of the 4f¹Lⁿ 5d¹ component in the ground state was correlated to the decrease of the crystal field acting on the Ce ion.⁶⁹ This was consistent with our observations about the crystallinity of the various samples as gathered from XRD. The 2p_3/2 → 5d_5/2 transition was larger in the N-doped high-surface-area CeO₂ than in the N-doped commercial CeO₂, which implied that a larger numbers of catalytically active Ce³⁺ active clusters resided in the high-surface-area samples.

Zoom In Zoom Out Reset image size

The EXAFS region the XAS spectra were analyzed after processing the data as follows: the data were initially processed for background removal and the normalized EXAFS spectra obtained using the following equation.⁷²

Where k is the wavenumber, μ(k) is the measured absorption, μ_○(k) is the background, and Δμ_○(0) is the edge step. After backround substraction, Fourier transforms of χ(k) to r space with k³ weighting factor and a Hanning window function (Dk1 = 0.2 and Dk2 = 4) were performed using the Athena data analysis system. From the Fourier transformed EXAFS data in r space, the qualitative change in the coordination number (number of Ce neighbors) and bond distances such as Ce-O (first shell), Ce-Ce (second shell) and Ce-O (third shell) were extracted from Figure 5b (the corresponding shells (Ce-O, Ce-Ce and Ce-O) are labeled in the figure). A significant increase in the first shell Ce-O distance and a corresponding decrease in intensity of the Fourier transformed EXAFS signals were observed in the N-doped samples. The increase in the Ce-O distance was attributed to lower orbital mixing (which causes lower covelency and higher ionicity) and to lower crystal field formation; the corresponding decrease in intensity was attributed to the decrease in the coordination number.^69,73 It was postulated that the greater ionic nature of the Ce-O bond strengthened the CeO₂ lattice and improved stability during prolonged ROS scavenging.

In-situ fluorescence spectroscopy experiments were performed to determine the regenerative ROS scavenging capacity of all samples.²⁴ The results obtained from the experiments performed at 80°C and 75% RH are shown in Figure 6. Briefly, any decay in fluorescence is indicative of ROS action (and hence lack of efficacy of ROS scavenging). The undoped samples exhibited high free radical scavenging efficacy for the first 40 hours, but subsequently began to lose scavenging ability as evidenced by the decay in the fluorescence signal emanating from within the PEM. On the other hand, the PEMs containing N-doped samples (both commercial and high-surface-area variants) evidenced no decay in the fluorescence signal for at least 60 hours, confirming superior regenerative scavenging of the free radicals generated in-situ. This demonstrated that N-doping enabled CeO₂ to retain its regenerative ROS scavenging capacity for a longer time period.

Zoom In Zoom Out Reset image size

To distinguish between the two nitrogen-doped samples (i.e. to establish whether the surface area of the N-doped cerium oxide and resultant changes in microstructure played a role in ROS scavenging efficacy), a similar in-situ fluorescence experiment was conducted under much accelerated conditions at 90°C and 50% RH, wherein the rate of ROS generation in-situ would be much higher than at 80°C and 75% RH. The results obtained are shown in Figure 7. Under these conditions, the N-doped commercial CeO₂ lost its regenerative ROS scavenging ability after 40 hours. However, the N-doped high-surface-area CeO₂ showed no change in ROS scavenging efficacy even after 100 hours, at which point the experiment was stopped. The exceptional regenerative scavenging efficacy of the N-doped high-surface-area CeO₂ was attributed to to the higher quantity and higher stability of the Ce³⁺ active clusters as well as to the higher stability of Ce-O bond in its lattice, as discussed under microstructure characterization. Further ex-situ experiments were performed to confirm this hypothesis.

Zoom In Zoom Out Reset image size

The CeO₂ samples were studied using ex-situ experiments designed to elucidate the stability of Ce³⁺ active clusters upon exposure to ROS (generated using Fenton's reagent). In the first set of experiments, both the N-doped CeO₂ samples were exposed to ROS generated using Fenton's reagent. This ex-situ experiment did not provide any tangible insights into the mechanism underlying the observed difference in the regenerative ROS scavenging capacity in the N-doped CeO₂ samples. See Figure S7. A second set of ex-situ experiments were performed by mimicking the conditions existing during fuel cell operation, wherein XAS spectra were recorded after sample exposure to ROS (Fenton's reagent) in Nafion. The normalized XANES spectra of the cerium L3 edge and the Fourier-transformed EXAFS signals for all treated CeO₂ samples are presented in the Figure 8. From the normalized XANES spectra of the commercial CeO₂ sample, a decrease in the number of active Ce³⁺ active clusters upon exposure to ROS could be discerned. A corresponding decrease in the Ce-O distance was also observed in the Fourier transformed EXAFS spectra of this sample (See Figure 8a). The N-doped commercial CeO₂ showed a small increase in the number of Ce³⁺ active clusters, but a decrease in the Ce-O distance upon ROS exposure in Nafion (See Figure 8b). The normalized XANES spectra and Fourier transformed EXAFS spectra of high-surface-area CeO₂ showed a small decrease in both the number of Ce³⁺ active clusters and in Ce-O distance upon exposure to ROS. (See Figure 8c). The normalized XANES spectra of N-doped high-surface-area CeO₂ revealed that there was an increase in the number of active Ce³⁺ active clusters upon exposure to ROS. Likewise, an increase in the Ce-O distance was observed from the Fourier transformed EXAFS spectra obtained with this sample. (See Figure 8d).

Zoom In Zoom Out Reset image size

It is reported in the literature that there exist two possible events that can occur during ROS scavenging by CeO₂: (i) First, CeO₂ changes its valence state from +3 to +4 upon scavenging ROS. The presence of a larger Ce³⁺ active clusters after the ROS scavenging reaction is advantageous in that this allows the CeO₂ to beter retain/regenerate its ROS scavenging ability.^74,75 (ii) Second, the increase in Ce-O bond distance during the ROS scavenging reaction leads to an expansion of the CeO₂ lattice structure. This CeO₂ lattice expansion favors the creation of additional oxygen vacancies in CeO₂ at the end of the ROS scavenging reaction.⁶⁹ To maintain its regenerative ROS scavenging ability, the CeO₂ has to either retain its expanded lattice structure (i.e. retain the larger Ce-O bond distance) or, at most, revert back to its normal lattice structure (i.e. retain the Ce-O bond distance before ROS exposure).^74,75 Based on these insights and from our observations, we propose that a contracted lattice structure (i.e. shorter Ce-O bond distance) at the end of the ROS scavenging reaction leads to loss of regenerative ROS scavenging ability in CeO₂. Furthermore, we propose that the sustained regenerative ROS scavenging ability of CeO₂ depends on the simultaneous occurrence of both of the above mentioned events (i.e. both enhancement in Ce³⁺ clusters and increase in Ce-O bond distance) at the end of the ROS scavenging reaction. The rationale is presented below.

Our XAS experimental results demonstrated that there was a loss of Ce³⁺ active clusters and a decrease in Ce-O bond distance in the commercial CeO₂ at the end of the ROS scavenging reaction; this correlated well with the loss of in regenerative ROS scavenging ability of this material. Similarly, while N-doped commercial CeO₂ showed an increase in Ce³⁺ active clusters, it also exhibited the decrease in Ce-O bond distance at the end of the ROS scavenging reaction, ultimately leading to loss of regenerative ROS scavenging ability. High-surface-area CeO₂ exhibited a small decrease in Ce³⁺ active clusters and Ce-O bond distance upon exposure to ROS. Though these changes were small relative to commercial CeO₂, the high-surface-area CeO₂ still lost its regenerative ROS scavenging ability. N-doped high-surface-area CeO₂ evidenced a clear increase in both the number of Ce³⁺ active clusters and in the Ce-O bond distance at the end of the ROS scavenging reaction. This was the only material to also demonstrate quantitative regenerative ROS scavenging for over 100 hours under harsh conditions (at which point the experiment was stopped).

Based on these observations and the above discussion, the in-situ fluorescence spectroscopy results correlated well with the results obtained from ex-situ ROS scavenging experiments. The superior regenerative ROS scavenging activity capacity of the N-doped high-surface-area CeO₂ was attributed to: a) the facilitation of a large number of active Ce³⁺ active clusters and, importantly, the retention of these active clusters even upon exposure to ROS; and b) maintaining an expanded lattice (Ce-O bond distance) upon exposure to ROS.