2.1 Characterization

Figure 1 displays the N₂ adsorption–desorption isotherms at 77 K for CNT, CNT-N, and the Mo₂N/CNT, Ni₃N/CNT, and Co₄N/CNT catalysts. All the samples exhibit type IV isotherms with H3-type hysteresis loops, indicative of mesoporous structures, in accordance with the IUPAC classification.^[38, 39]

Textural properties and metal loading are summarized in Table 1.

The decrease in BET surface area observed for CNT-N compared to pristine CNT is attributed to the partial degradation or shortening of the nanotubes during nitridation, as confirmed by transmission electron microscopy (TEM) analysis. The experimentally determined metal loadings (AAS) slightly exceed the nominal 5 wt%, likely due to CNT mass loss during NH₃ treatment, resulting in apparent enrichment of metal content.

XRD patterns (Figure 2) confirm the successful formation of metal nitride phases. The diffractograms were compared with patterns of CNT (JCPDS-ICDD, 75-1621), metal nitrides γ-Mo₂N (JCPDS-ICDD, 25-1366), Ni₃N (JCPDS-ICDD, 00-010-0280), Co₄N (JCPDS-ICDD, 41-0943), and metallic Ni (JCPDS-ICDD, 03-065-0380). CNT and CNT-N show characteristic peaks at 2θ = 25.7° and 42.8°, corresponding to the graphitic structure.

Additional peaks at 2θ = 53.1° and 77.9° indicate the presence of the graphene phase,^[40] suggesting that the CC bonds remain dominant over the CN formation during nitridation. Mo₂N/CNT exhibits peaks corresponding to γ-Mo₂N and β-Mo₂N, with no detectable Mo oxide signals, confirming complete nitridation. Ni₃N/CNT shows Ni₃N and metallic Ni° reflections, consistent with partial decomposition during nitridation.^[41] Co₄N/CNT shows well-defined Co₄N reflections, with no evidence of metallic Co or oxide species.

Thermogravimetric analysis in oxidizing conditions (Figure S1, Supporting Information) demonstrates the increased thermal stability of CNT-N compared to untreated CNT, confirming the structural integrity after nitridation. All metal nitride catalysts are thermally stable above 400 °C and suitable for catalytic applications.

TEM images and particle size histograms (Figure 3) show that Mo₂N particles are primarily confined within the CNT channels (≈5.3 nm), while Ni₃N and Co₄N are more externally dispersed with larger particle sizes (≈12.2 nm and 15.7 nm, respectively). This confinement effect may influence both stability and accessibility of active sites.

FTIR spectra (Figure 4) show O–H stretching at 3735 cm⁻¹ and 3436 cm⁻¹ in CNT and CNT-N, related to ambient moisture, respectively. The characteristic CC vibration of graphene appears at 1550 cm⁻¹. Nitrided samples show new bands at 1646 cm⁻¹ and 1186 cm⁻¹, attributed to the N–H and C–N bonds, respectively,^[42-44] confirming nitrogen incorporation. Mo₂N/CNT and Ni₃N/CNT exhibit moderate intensities for these bands, while Co₄N/CNT shows stronger absorption, suggesting more extensive nitride phase exposure.

For the nitride-metal catalysts, temperature-programmed reduction (H₂-TPR) profiles (Figure 5) show weak reduction signals for CNT and CNT-N at ∼300 °C and ≈770 °C, attributed to surface and internal CN bond reduction.

The difference in temperature for the appearance of the reduction peaks for CNT and CNT-N is attributed to the environment of the CN bonds, being the surface bonds reduced at lower temperature, and at higher temperatures, the CN bonds inside the CNT. Mo₂N/CNT displays two overlapping peaks at 366 °C and 405 °C associated with NH_x reduction in γ- and β-Mo₂N phases, and additional peaks at 637 °C and 758 °C related to bulk Mo₂N reduction and confined C = N bond reduction.^[45] Ni₃N/CNT shows surface NH_x reduction at 204 °C and Ni₃N to Ni° reduction at 544 °C.^[41] Co₄N/CNT presents a single reduction peak at 285 °C, associated with Co₄N to Co° transformation.^[46]

In Figure 6, the NH₃–TPD profiles indicate the strength of the acid sites classified as weak (<300 °C), medium (300 °C < T < 500 °C) and strong (>500 °C).^[47]

NH₃–TPD profiles (Figure 6a) show negligible ammonia desorption for CNT and CNT-N, indicating an absence of significant acidity. Mo₂N/CNT shows a strong desorption peak at 781 °C, confirming the presence of strong acid sites. Ni₃N/CNT and Co₄N/CNT exhibit broad, low-intensity peaks, suggesting fewer and weaker acid sites.

CO₂–TPD results (Figure 6b) follow a similar trend. Mo₂N/CNT shows a prominent peak at ∼800 °C, attributed to strong basic sites. Co₄N/CNT exhibits weak basicity, while Ni₃N/CNT shows negligible CO₂ desorption, indicating limited or undetectable basic sites. These results suggest that the acid–base properties depend more strongly on the metal–nitrogen bond than on the specific nitride phase. Mo₂N, with higher Mo–N interaction, exhibits stronger acid and basic site densities compared to Ni₃N and Co₄N.

To identify Lewis and Brønsted sites, the catalytic conversion of 2-propanol was carried out as an indirect method to evaluate the acid properties.

The dehydration reaction of 2-propanol to propylene occurs on Lewis acid sites and the dehydrogenation of 2-propanol to acetone occurs on the Brønsted acid sites;^[48] therefore, the product distribution of this catalytic reaction can be related to the amount of Lewis and Brønsted sites. 2-propanol conversion (Figure S3, Supporting Information) was used to distinguish Brønsted and Lewis acidity. Propylene formation (dehydration) indicates Lewis sites, while acetone (dehydrogenation) suggests Brønsted sites. Mo₂N/CNT shows a higher proportion of Lewis sites, while Ni₃N/CNT and Co₄N/CNT exhibit a more balanced acidity profile.

The XP spectra of the Mo 3d_5/2, Ni 2p_3/2, and Co 2p_3/2 core-level spectra are shown in Figure 7, the C1s and N1s in the supplementary information, and the binding energy (BE) values and surface distribution in Table 2. The deconvoluted Mo 3d_5/2 X-ray photoelectron spectroscopy (XPS) (Figure 7a) shows several oxidation states ranging from Mo²⁺ to Mo⁶⁺ at 228.7 eV assigned to surface Mo₂N, with Mo^δ+ (2 < δ < 4) at 229.9 eV to Mo⁴⁺ and at 232.5 eV to Mo⁶⁺ in molybdenum oxynitrides.^{[20, 23, 49]} For Ni₃N/CNT, the doublet for Ni 2p_3/2 at 854.0 and 856.0 eV in Figure 7b to surface nickel nitride and oxynitride.^[50] For Co₄N/CNT, Figure 7c shows the Co 2p_3/2 signal for BE at 779.6 eV attributed to cobalt–nitride species, and at 781.0 eV and 783.1 eV to Co³⁺ of cobalt oxynitrides.^[23, 51]

The surface extent of the nitride phase (area under curve highlighted in blue color in Fig. a, b, and c), indicates in Table 2, in line with TEM results, a surface of 31% of Mo₂N (228.7 eV), 39% of Ni₃N (854.0 eV), and 42% of Co₄N (779.6 eV).

The similar surface oxynitride phase contributions (≈70%) of the Mo₂N/CNT, Ni₃N/CNT, and Co₄N/CNT catalysts are in line with Tyrone Ghampson et al.^[23] who reported protected Mo₂N supported on the Al₂O₃ and MCM-41 surface by reduction and passivated treatments. Therefore, the large stability of the surface nitride phase in the Mo₂N/CNT, Ni₃N/CNT, and Co₄N/CNT catalysts can be attributed to the confinement effect of the CNT. Regarding the presence of the surface N, the XPS of N1s (Figure S4, Supporting Information) indicates only surface contributions of nitrogen assigned to the surface Mo₂N species at 394.9 eV and 397.5 eV,^[52, 53] not detected for Ni₃N/CNT and Co₄N/CNT. The XP spectra of C1s (Figure S5, Supporting Information) display the expected signals at 284.3 eV of the CC bond of the graphene sheets, 285.1 eV of the C–C bond, 286.0 and 287.2 eV of the C–O bonds, C–H, and C–N bonds, 289.0 eV C = O and at 291.0 eV the O–CO bonds characteristic of carboxylic acids, esters, and other organic functional groups^[54-57] in the surface of the Mo₂N/CNT, Ni₃N/CNT, and Co₄N/CNT catalysts.

3.1 Fresh Catalysts

The catalytic performance of Mo₂N/CNT, Ni₃N/CNT, and Co₄N/CNT catalysts in the conversion of LA was evaluated in terms of total conversion (X_t) and initial reaction rate (r_s), as summarized in Table 1 and Figure S6, Supporting Information. The blank experiment with CNT-N showed no detectable LA conversion, confirming the necessity of the active metal nitride phase.

The largest conversion level obtained for Ni₃N/CNT, the same trend of initial rate, can be explained considering the distribution of the active phase inside and outside graphene sheets of the CNT, allowing accessibility of LA, promoting the catalytic activity. To explain the lowest catalytic performance of Mo₂N/CNT, it is proposed that the confinement effect favors the stability of the active phase, promoting a smaller particle size, and hinders the access of LA to the active nitride sites. The catalytic performance of Co₄N/CNT is attributed to the surface aggregates of large particle size, decreasing the surface availability of the active sites. This catalytic performance indicates that under the used reaction conditions, the active sites in the conversion of LA are the surface nitride species with not a large effect of the particle size.

Product distribution analysis (Figure 8a) reveals that the main reaction intermediates are angelica lactone (AL) and 4-hydroxypentanoic acid (4-HPA), with GVL as the principal final product. These results are consistent with reported pathways^[4, 5, 58] in which LA conversion proceeds via parallel dehydration (to AL) and hydrogenation (to 4-HPA) routes, followed by hydrogenation or dehydration to GVL.

Selectivity data at 10% LA conversion suggest that all catalysts initially favor the formation of AL via acid-catalyzed dehydration, followed by its hydrogenation to GVL. The Mo₂N/CNT catalyst exhibited a higher proportion of 4-HPA, indicative of its strong hydrogenation capability. However, the slower overall kinetics likely result from limited accessibility to active sites due to confinement effects.^[59] Final product yields at 240 min (Figure 8b) show GVL as the dominant product and AL as the only remaining intermediate. This behavior correlates with the ∼70% surface oxynitride species identified by XPS for all catalysts,^[60, 61] highlighting the importance of the Brønsted and Lewis acid site strength in determining the rate and extent of the LA conversion. These findings are supported by the 2-propanol dehydration results, which confirm the presence of acid sites, with a slight Brønsted acidity predominance in Ni₃N/CNT and Co₄N/CNT. Overall, the superior GVL yield (∼90%) for Ni₃N/CNT and Co₄N/CNT compared to Mo₂N/CNT (≈70%) demonstrates the critical role of accessible hydrogenation sites and surface composition. The catalytic behavior observed is comparable to that of the noble metal-based systems and can be explained by the interaction between metal d orbitals and nitrogen sp orbitals, leading to stabilized metal/nonmetal bonds and modulated electronic properties.^[62]

3.2 Characterization Post Reaction

Postreaction XRD patterns (Figure 9) confirm the structural stability of the nitride phases. For Mo₂N/CNT, the characteristic reflections of γ-Mo₂N and β-Mo₂N remain, along with the appearance of a peak at 2θ = 48.8°, corresponding to MoO₂, suggesting partial surface oxidation. For Ni₃N/CNT and Co₄N/CNT, the respective reflections for Ni₃N/Ni° and Co₄N remain unchanged, confirming high-phase stability during reaction.

AAS results (Table 1) indicate metal leaching during the catalytic tests, with the highest loss for Co (53%), followed by Mo (41%), and Ni (21%). The lower leaching observed for Ni₃N/CNT supports the hypothesis of more stable metal–support interactions and better nitride phase anchoring in this system.

3.3 Catalysts Recycling

Based on its superior catalytic performance and lower metal loss, Ni₃N/CNT was selected for recycling tests. After each reaction, the catalyst was recovered by vacuum filtration, washed with dioxane, and dried at 120 °C. The catalytic data for fresh catalyst and after the first (R1) and second (R2) reuse cycles are presented in Table 1 and Figure S7, Supporting Information. A progressive decline in X_t was observed from 70% (fresh) to 50% (R1) and 43% (R2). The initial rate dropped by ≈40% after the first cycle and remained relatively constant thereafter. Selectivity trends (Figure 10a) show increased AL and reduced GVL formation in the recycled samples, suggesting a loss of hydrogenation functionality. However, product yields at 240 min (Figure 10b) remain consistent between R1 and R2, indicating that the nature of the active sites remains unchanged. XRD analysis of fresh and recycled Ni₃N/CNT samples (Figure 11) confirms the structural integrity of the Ni₃N phase throughout both recycling steps. The diffraction patterns show consistent peak intensity and width, indicating that the phase crystalline is preserved. These findings suggest that the decline in the catalytic performance is due to a decrease in the number of active sites, primarily from Ni leaching, rather than deactivation or structural degradation of the nitride phase.