Pd/In-based catalysts for nitrate catalytic removal from water: synthesis designs aiming for better N₂ selectivity

Providing high-quality drinking water is one of the Sustainable Development Goals (SDGs) for 2030 established by the United Nations General Assembly in 2015. Despite this, water-body pollution constitutes one of the areas of major concern for research. The pollution of groundwater and surface water caused by nitrate has become a serious problem since about 90% of the world's drinking water supplies come from groundwater sources (Ramalingam & Subramania 2021), and due to the hazard it represents to human health and the environment (Kiani et al. 2022; Qasemi et al. 2022; Singh et al. 2022). Nowadays, according to the United Nations (UN) report (UNESCO 2019), nitrate is the most common chemical contaminant found in groundwater and the most widespread pollutant in groundwater due to its high solubility in water and the difficulty of fixing it in the soil. There is an increasing concern about this pollutant as a consequence of intensive anthropogenic activities, which are known as the main source of this pollutant. The excessive use of fertilizers and the discharge of liquid industrial waste without appropriate treatment represent the major fonts of nitrate pollution in groundwaters (Chauhan & Srivastava 2021).

High concentrations of nitrates may damage the quality of water resources and have significant adverse effects on human health, like methemoglobinemia (blue baby syndrome) as well as cancer and hypertension caused by the metabolic conversion of nitrite into N-nitroso compounds (Clark et al. 2020). In this sense, the removal of nitrates from water bodies is necessary to ensure the consumption of clean and safe drinking water.

Numerous methods have been reported for the removal of nitrates from water involving adsorption (Wang et al. 2021; Liu et al. 2022), reverse osmosis (Fuentes & Hwang 2021; Scholes et al. 2021), biological denitrification (Hurtado-Martinez et al. 2021; Pang & Wang 2021), electrochemical reduction (Fu et al. 2021; Fu et al. 2022), and catalytic reduction (Martínez et al. 2017; Guo et al. 2018; Lv et al. 2022). Among them, catalytic nitrate reduction from water has been proposed as an efficient and inexpensive method for meeting and standards and removing nitrate. It has also advantages such as no sludge production, low energy consumption, and high N₂ selectivity (Martínez et al. 2017; Tokazhanov et al. 2020).

For catalytic nitrate removal, it is critical to synthesize a catalyst with suitable metals and structures to obtain a good nitrate removal capacity. The catalytic composition, structure, and supports not only affect the reaction but also the reaction rate and the end-products (Shen et al. 2021). Bimetallic catalysts have been proven to be more efficient than monometallic catalysts in terms of conversion and selectivity to nitrogen gas in the catalytic denitrification of water (Shen et al. 2021). Sá & Vinek (2005) studied the catalytic hydrogenation of nitrate over a bimetallic Pd/Cu catalyst and compared its catalytic performance with the corresponding monometallic catalysts (Pd and Cu). The authors found that monometallic Pd catalysts could reduce ⁠, but its degradation was very slow. The monometallic Cu catalyst was only active at the beginning of the reaction, and the fast oxidation of Cu led to the deactivation of the catalyst, Pd being necessary to stabilize Cu in its lower oxidation state. On the other hand, Pd/Cu was demonstrated to be more efficient than Pd and Cu monometallic catalysts, and a better activity and selectivity in nitrate degradation was achieved by reducing the amount of the noble metal of Pd/Cu from 5 to 2 wt.%. Devadas et al. (2011) studied catalytic nitrate reduction using Pd/CeO₂ and PdSn/CeO₂ catalysts and found that although Pd/CeO₂ is more active than the bimetallic catalysts under pure hydrogen flow, the monometallic Pd/CeO₂ is inactive for nitrate reduction in the presence of CO₂. The bimetallic catalysts turn out to be more active than under pure hydrogen flow and also than the monometallic catalyst with a low selectivity toward ammonium ions, which are the undesired product of the reaction. Thus, the Pd–Sn/CeO₂ catalyst is comparatively the most suitable catalyst for nitrate reduction. One explication for the monometallic Pd/CeO₂ activity was given by Devadas et al. (2011), who investigated the performance of Pd/CeO₂ and reported that after reduction under hydrogen flow, Ce was in its low coordination form and vacancies of oxygen were created on the surface. Such reducible supports were capable of allowing nitrate to initiate the reduction reaction, but the main product of the reaction was ammonium due to the high availability of electrons to over-reduce NO(ads) – NO adsorbed in active sites.

In most of the reported works, the catalysts used in catalytic denitrification were composed of a combination of noble metals, usually Pd, Pt, or Rh, and promoter metals (In, Sn, Cu, Au, and Ni) deposited on different supports (Imura et al. 2014; Li et al. 2019; Hamid et al. 2020).

Nitrate reduction by bimetallic catalysts involves a multi-step redox process requiring an additional reducing agent, being H₂(g) being the most commonly used (Werth et al. 2021). The role of the transition metal is to reduce nitrate to nitrite by a redox process, leading to its oxidation; and the function of the noble metal is to stabilize the transition metal in its lower oxidation state by hydrogen spillover to prevent deactivation. This is the reason why bimetallic catalysts are more efficient than monometallic ones (Santos et al. 2020).

In this sense, metal disposition, interactions between metal–metal and metal–support, metal loading, atomic configuration, and size had a significant effect on the reduction rate of nitrate (Guo et al. 2018). Moreover, the co-production of ammonium during nitrate reduction is still the main drawback of the process when the aim is to obtain drinking water, as it causes adverse effects, such as toxicity, to some aquatic organisms in surface waters, increasing the need for disinfectants and/or biological nitrification in water distribution systems (Garcia-Segura et al. 2018). Even so, there is a lack of information on synthesizing materials that are both efficient in removing nitrate from water and highly selective for N₂. In this line, it is important to remark that Yin et al. (2018) showed that this bimetallic catalyst could be used in the catalytic reduction of oxyanions such as chromium, chlorine, and bromine as examples. Although this implementation is not the goal of the present study, it is possible to think of a practical, wide implementation of the material, which will give it the highest added economic value.

In this work, (Pd, In)-based catalysts deposited by the wet impregnation method onto different supports, alumina (Al₂O₃), silica (SiO₂), and titania (TiO₂) were prepared and evaluated for nitrate removal from water to study the influence of support on the catalytic performance and selectivity. Then, catalysts with the same formulation were synthesized by alternating the sequential steps of synthesis, including both the order of deposition of the noble and promoter metals on the support, as well as the calcination and reduction stages, to study how that affects the properties of the catalytic systems and to find the one that allows the highest selectivity to N₂. Thus, this study aims to determine whether the type of support used during the synthesis or the sequence of synthesis steps has a significant influence on the catalytic properties of the bimetallic system for the catalytic removal of nitrate from water and to report the conditions that allow obtaining a catalyst with high activity toward the removal of nitrate from water and high N₂ selectivity. Thus, it would be possible to attain a safer and more reliable water supply in the future by treating nitrate-contaminated water with high N₂ selectivity.

Gamma-Al₂O₃ (KEDJEN CK300, 180 m²/g, vol. pore 0.5 cm³/g), SiO₂ (Alfa Aesar 300 m²/g, vol. pore 1 cm³/g), and TiO₂ (Degussa, P25, 48 m²/g, pore 0.28 cm³/g) were used as support for the synthesis of the (Pd, In)-based catalysts. These oxides with well-differentiated acid properties, in terms of their surface sites and acid strength, were selected as supports for the catalysts’ conformation to evaluate their influence on the physicochemical and catalytic properties.

PdCl₂ (Sigma Aldrich, p.a.), In₂O₃·3H₂O (Sigma Aldrich, 99.9%) were employed as precursors for metallic Pd and In, respectively. Hydrazine chloride (Sigma Aldrich, p.a.) was used as the reducing agent for the synthesis. KNO₃ (Cicarelli, p.a.) and hydrogen gas (INDURA, 99.998%) were employed as the nitrate precursor and the reducing agent, respectively, for the catalytic assays. HCl (Cicarelli p.a., 37%) was used in the analytical determination of products and reactants.

PSCRIR catalyst was prepared as follows: (1) wet impregnation of Pd over the supporting material and (2) calcination at 500 °C for 4 h. In this step, Pd oxides were obtained on the surface of the catalyst. (3) Reduction of the material with hydrazine hydrate solution 0.8 M to obtain Pd⁰ (metallic form). (4) Wet impregnation of In precursor onto the material obtained in step 3. (5) Reduction of the catalyst obtained in hydrazine hydrate solution. Then, upon reduction in the aqueous phase, Pd was reduced by the action of hydrazine, forming strongly anchored Pd⁰.

PSCRI catalyst was synthesized as follows: (1) wet impregnation of the Pd precursor onto the support. (2) Calcination at 500 °C for 4 h, obtaining Pd oxides on the surface of the support. (3) Reduction of the material with hydrazine hydrate solution 0.8 M leading to the formation of Pd⁰. (4) Wet impregnation with the indium precursor onto the material obtained after step 3.

In the PSCRI catalyst, palladium was calcined on the support and then reduced, generating isolated metallic species on the support. Indium was then added, allowing its deposition on the surface of the metallic palladium.

The following is a description of the steps followed for the synthesis of PISCR catalyst: (1) wet impregnation of Pd and In precursors. (2) Calcination at 500 °C for 4 h. After calcination, it was expected to obtain the material surface coated with Pd oxides and In oxides, anchored to the support. (3) Reduction with hydrazine hydrate solution 0.8 M.

PSCIR catalyst was prepared as follows: (1) wet impregnation of Pd precursors onto the support. (2) Calcination at 500 °C for 4 h. (3) Wet impregnation of In precursors. (4) Reduction in hydrazine hydrate solution 0.8 M. In PSCIR, the palladium was anchored to the support by the calcination process, and the indium was deposited randomly on the catalyst surface.

For PSCICR synthesis, we proceeded as follows: (1) wet impregnation of Pd precursors onto the support. (2) Calcination at 500 °C for 4 h. (3) Wet impregnation of In precursors. (4) Calcination at 500 °C for 4 h. (5) Reduction in hydrazine hydrate solution 0.8 M.

Scanning electron microscopic (SEM) micrographs and EDX spectra were obtained with a Carl Zeiss Sigma microscope. Moreover, a JEOL scanning electron microscope model JSM 35C was used to elucidate the composition and distribution of Pd and In over the supports (SEM-EDS).

The elemental quantification of elemental Pd and In was carried out by employing a Shimadzu spectrometer, equipped with an energy-dispersive analytical system (EDX-720) with a high detection range (from Na11 to U92), a rhodium X-ray source, operated at 50 kV and 100 μA, using five X-ray primary filters.

C₀ is N-ppm of nitrates at the beginning of the reduction process, C is N-ppm of nitrates at time t, and C_A is the concentration of N-ppm products (nitrites or ammonia) at time t.

The initial rate (v_i, given in ppm/min) was estimated taking into account that up to 10% conversion, the system was not yet affected by its reaction products. Therefore, v_i was calculated as the slope of the nitrate concentration as a function of the time curve at the time at which 10% of nitrate was converted (the time when the concentration of N-NO₃ in the reactor was 90 ppm).

Pd and In atomic percentages, obtained by X-ray fluorescence (XRF), are summarized in Table 1. The Pd and In retention percentages were estimated taking into account the amount added in the synthesis procedure and the values obtained by XRF. A composition deviation from the theoretical values was observed after each synthesis procedure. The highest retained metal percentage was observed in samples supported on TiO₂ (85%–99%). Regarding SiO₂- and Al₂O₃-based catalysts, when calcination occurs after impregnation, the greater proportion of Pd and In was retained by the support for PIACR (49%) and PISCR (46%). It is believed that due to the lower interaction with Al₂O₃ and SiO₂, it is only possible to promote the anchoring of both metals at the support surface with a calcination step after the impregnation. This affirmation is corroborated by Wang et al. (2021), who calculated the adsorption energy of each of the Pd nanoparticles (NPs) over Al₂O₃, SiO_2, and TiO₂ and showed that a higher interaction occurred with TiO₂. Concerning TiO₂ support-based catalysts, the higher retention occurs when In was impregnated after Pd calcination and reduction (PTCRIR and PTCRI). Then, Pd reacted producing a strong metal–support interaction (SMSI) (Jia et al. 2017) with titania terminal OH and, after reduction, the reduced Pd reacted selectively with In, resulting in a stable bimetallic particle of Pd:In. This type of metal–support interaction (MSI) produced by the sequential method of preparation was recognized through the decrease in the capability to chemisorb molecules, minimizing the metal nanoparticles' superficial energy and consequently blocking the accessibility to H₂ to react dissociatively with the Pd particles (Lou et al. 2020). Chen et al. (2017) reported that the nitrogen-doping TiO₂ enhanced the MSI of the Pd–TiO₂ system due to an electronic promoting effect. In this work, hydrazine hydrate was used for the reduction PdO_x, and this reagent could interact with the acidic sites of the support (Wojcieszak et al. 2013) and remain on it providing N-doping TiO₂ support which would enhance the MSI with Pd particles. On the other hand, the higher retention of In in the catalysts supported on TiO₂, compared with the other two supports, suggested a preferential deposit of In on Pd during synthesis, and it could be associated with a lower MSI concerning Pd when SiO₂ and Al₂O₃ were used. This process is thermodynamically favored considering the possible redox mechanism. Metallic Pd can react with In(III) and this ion is reduced to In(0) over Pd. Then, the metallic form of the novel metal was reloaded in the next step with the reduction of H₂. Thus, In was deposited over Pd, and Pd had an electronic effect over In, and at a low wt.% In remained reduced over Pd (Guo et al. 2018). This process resulted in a bimetallic particle, bonded by the most stable energetic position and electrostatic forces, and leach resistant. Although energetic calculus is necessary to corroborate what we can appreciate experimentally due to the low lixiviation, the affirmation is supported by Wang et al. (2021), who calculated the adsorption energy of each of the Pd NPs over Al₂O₃, SiO_2, and TiO₂ and showed that a higher interaction occurred with TiO₂. Since, in this study, the temperature effect was only introduced by the calcination process, and the reduction stage occurred in mild conditions, the aggregation of the particles observed over the surface could have resulted as a consequence of the thermal treatment. It is well known that the extent of MSI depends on the nature of the metal but mostly on the size of the particles and the nature of the support (Coq 2000). The support acted as a supramolecular ligand and has been claimed to promote specific electronic properties and/or geometrical features of the nano-sized supported metal particles. The key point lies in the strength of interaction between the d-band of the metal and the molecular orbitals of reactants and products (Coq 2000). Also, electronic interactions between Pd NPs and the TiO₂ support rise by decreasing particle size, as a result, creating more admetals on the Pd NPs in contact with TiO₂ (Zhao et al. 2022).

In this sense, the In ability to deposit preferentially over Pd during synthesis was reflected by the higher retention of indium in the catalysts supported on titanium, compared with the other two supports, which presented a lower MSI with respect to Pd. Two situations could produce the preferential deposition and anchoring effect. In the first place, it is known that Al₂O₃ and TiO₂ have a higher density of reactive hydroxyl groups compared with SiO₂ and therefore provide more anchoring sites for Pd and In (Yang & Bent 2017). In the second place, as titanium is a semiconductor support, the availability of electrons to create a chemical interaction between Pd and In is higher than that in the other two isolating supports (alumina and silica). Wang et al. (2021) showed that the adsorption energy of a Pd cluster increased in the sequence SiO₂ (0.85 eV) < Al₂O₃ (2.12 eV) < TiO₂ (3.05 eV), which was accompanied by a decrease in the distance between the cluster and the surface. Those properties may contribute to justify why a more stable Pd cluster is produced on TiO₂, and due to this energy, In is preferentially linked to Pd in this material. No significant differences were observed between SiO₂ and Al₂O₃ in metal retention, in spite of the significant energetic difference, which could be associated with the almost two-fold higher surface area of SiO₂ (300 m²/g). Durkin et al. (2018) showed that the ratio of Pd to In measured in different nanoparticle regions was more uniform than the one observed in Pd–Cu catalysts. It was consistent with a uniform distribution, which is compatible with a selective deposition process induced by a reduced Pd or by an electron-rich zone near Pd. It is noteworthy that Pd (2.20 eV) had a higher electronegativity than In (1.78 eV) and that this difference (0.42 eV) favored the formation of a very polar interaction between both elements. The ratio of Pd to In measured in different nanoparticle regions appeared to be more uniform than the one observed in the Pd–Cu catalysts from our previous study, consistent with the XPS and energy-dispersive spectroscopy (EDS) data that showed that the Pd–In composition throughout the fibers was uniform.

Byun et al. (2020) found that the active phase/support atomic ratio decreased in a similar order as the ability to retain metal decreased. The lower the active phase/support atomic ratio, the lower the interaction with the support, and a less active phase is retained after the process. The order proposed is Al₂O₃ > TiO₂ > SiO₂.

In this study, the order in the active phase/support atomic ratio was TiO₂ > Al₂O₃ > SiO₂ and was preferably modulated by the sequential synthesis process more than by the MSI. Probably, TiO₂ can retain a higher proportion of metals than the other supports since metals had the strongest interaction with the semiconductor support. Moreover, TiO₂ and interactions are weaker with isolator supports such as SiO₂ and Al₂O₃ (Coq 2000). These findings are in line with several of the aforementioned authors who stated that superficial energy (or the enhanced electron accumulation at the surface) contributed to stabilize the Pd cluster over the support surface depending on its semiconductor properties. Thus, as the TiO₂ had the highest energy, the particles were more established in this surface (Li et al. 2022). In addition, Wang et al. (2021) confirmed, with optimized geometric configurations of Pd/Support (where the supports are SiO₂, Al₂O₃, TiO₂, and CeO₂), that the adsorption energy increased in this order, suggesting that the strength of the MSI also increased in this order: Pd/SiO₂ < Pd/Al₂O₃ < Pd/TiO₂.

A SEM-EDS analysis of PSCIR catalyst (Figures S1 and S2) was carried out to analyze the active phase distribution and quantification. Pd and In wt.% are summarized in Table S1. Pd agglomerates of about 7 μm were distributed heterogeneously over the SiO₂ support, which do not include In on their surface. However, In particles were detected on the support surface in much smaller particles than the Pd agglomerates. In a region containing Pd and In (Figure S2), a weight ratio of Pd/In = 2 was found, a value equal to half the nominal weight ratio, indicating that In particles were highly dispersed onto the silica surface since XRF measurements indicated a Pd/In ratio close to the nominal value.

In Figure 4, micrographs of TiO₂-supported catalysts are shown. In the highest magnification of the micrograph of the support (Figure 4(B2)), a hierarchical aggregated cauliflower-like structure, which is typical of the rutile phase (Shi et al. 2019; Ren et al. 2022), can be seen. At higher magnifications (Figure 4(B4) and 4(B7)), it can be seen that the particles present a cauliflower-like structure, following the structure of the support (TiO₂).

Through the SEM-EDS map of the PTCRI catalyst (Figure S3 in Supplementary Material), a Pd/In ratio equal to 3 was obtained (Table S1), which was lower than the theoretical value (4). This behavior could be related to the strong interaction of Pd with the support, as seen by XRF, promoting that most Pd is in bulk, and In preferentially on the surface of the catalyst.

Regarding alumina-supported catalysts (see Figure 5), C1–C3 show alumina micrographs, C4–C7 show micrographs of the PACRIR catalyst, and C8–C11 show micrographs of the PIACR catalyst.

Figure 5(C10) reveals the presence of repetitive prism- or conical-like structures, attributed to Pd agglomerates since micrographs of the support do not show this tendency and SEM-EDS analysis has shown a good distribution of In on the support. Also, it is clear from the micrographs C7 and C10 that PIACR maintains a rough surface like the Al₂O₃ support, while PACRIR seems to have smoother surfaces. Moreover, the agglomerates formed in PIACR are much smaller (about five times smaller) than those observed in the PACRIR catalyst, in agreement with what was observed by SEM-EDS (Figures S4–S8).

Figures S4–S8 show SEM-EDS analysis of alumina-supported catalysts. Although the Pd/In ratio in these catalysts was lower than nominal, indicating that there was lower retention of Pd with respect to In, the effect of the variation of the synthesis steps was mostly seen in the distribution of the active phases. In the catalysts in which the Pd and In precursors were simultaneously impregnated (PIACR), the atomic distribution of both elements followed the same pattern, indicating the coexistence of both phases in different regions. This would indicate that the introduction of intermediate stages, either calcination or reduction, considerably affected not only the distribution but also the loading of the active material.

However, not only observed was the effect of the order of the synthesis steps on the catalytic loading or distribution of active elements but also the influence of the support on the retention of both Pd and In. As mentioned above and in agreement with the silica-supported catalysts, the Pd and In retention values obtained by XRF in alumina-supported catalysts were lower than those observed for the titania-supported systems.

A strong interaction between Pd and In leading to the formation of intermetallic compounds was reported in Pd,In systems supported on Al₂O₃ and SiO₂ (Krawczyk et al. 2011). Moreover, in the presence of the intermetallic compound, a decrease in selectivity to nitrogen and, at the same time, an increase in the amount of ammonia formed was observed during the reaction (Witońska et al. 2008). On the other hand, it was reported that, for titanium support, an SMSI with noble metals could be related to the partial reduction of TiO₂. It means that some oxygen from the TiO(δ+) lattice was replaced by Pd, as demonstrated by Wang et al. (2021). In this sense, assuming the form of a reduced titania state, a promoting effect of TiO₂ can be expected in nitrate reduction (Krawczyk et al. 2011). The authors also found, after 1 h of reaction, the presence of trace amounts of Pd in the reaction solution when the catalyst was prepared with Al₂O₃ and SiO₂ as support, while no Pd was detected in the reaction medium when TiO₂ was used.

All the prepared Al₂O₃-supported catalysts resulted in being active in nitrate conversion (Figure 6(a)). Total conversions and conversions close to 100% were obtained, with PIACR being the catalyst with the best catalytic activity. The best catalytic activity is defined as the one which reaches the highest conversion and selectivity in the lowest time. When the catalysts were compared at the same conversion (90%), PIACR was the one that obtained the best selectivity at the lowest time with the same conversion (see Table 2). In the wet co-impregnation method, Pd and In cations were atomically mixed and their dispersion over the support was extensive before being calcinated (Frei et al. (2019). With this calcination step: (1) Pd and In fixed to the support and between them, a metal/metal oxide interaction (MMOI) was produced (Jia et al. 2017); (2) then, after reduction, the metallic promoted particles could react with the nitrate to convert it in nitrite; (3) nitrite reacted with the hydrogen adsorbed on Pd atoms and continued the reduction reaction to produce nitrogen or ammonia. Additionally, this behavior was reported and associated with the positive charge developed by the alumina surface, which introduced an attractive force between the positive alumina surface and the negative nitrate and nitrite ions leading by opposite charges (alumina isoelectric point = 8, at reaction pH = 5, the alumina surface is positive) (Marchesini et al. 2012).

Table 2

Selectivities of those catalysts that achieved 90% nitrate reduction

Catalyst .	t(min) X = 90% .	S[] (%) .	S[] (%) .	S[N2] (%) .	vi (ppm/min) .	X (%) at 120 min .
PACIR	100	2.3	13.7	84.0	4.9	94.6
PACRIR	84	1.2	4.4	94.4	2.3	100
PIACR	44	0.9	20.8	78.3	4.4	100
PACRI	96	0.9	11.6	87.5	5.2	98.5
PACICR	98	0.3	13.0	86.7	2.8	98.1
PIACR(H2)a	30	2.2	22.2	75.6	7.1	100
PTCRIR	96	0.6	40.4	59.0	7.8	97.8
PTICR	97	0.7	24.2	75.1	5.5	80.1
PTCRI	83	0.6	17.0	82.4	4.2	100
PTCICR	100	0.3	37.9	61.8	4.2	98.8
PITCR(H2)b	35	2.7	16.1	81.2	3.5	100

Catalyst .	t(min) X = 90% .	S[] (%) .	S[] (%) .	S[N2] (%) .	vi (ppm/min) .	X (%) at 120 min .
PACIR	100	2.3	13.7	84.0	4.9	94.6
PACRIR	84	1.2	4.4	94.4	2.3	100
PIACR	44	0.9	20.8	78.3	4.4	100
PACRI	96	0.9	11.6	87.5	5.2	98.5
PACICR	98	0.3	13.0	86.7	2.8	98.1
PIACR(H2)a	30	2.2	22.2	75.6	7.1	100
PTCRIR	96	0.6	40.4	59.0	7.8	97.8
PTICR	97	0.7	24.2	75.1	5.5	80.1
PTCRI	83	0.6	17.0	82.4	4.2	100
PTCICR	100	0.3	37.9	61.8	4.2	98.8
PITCR(H2)b	35	2.7	16.1	81.2	3.5	100

^aMarchesini et al. (2008a).

^bMarchesini (2008).

View Large

This expression represents an apparent initial velocity, where k′ is the reaction constant and it includes the reductant agent concentration due to the H₂ in excess. The catalyst's behavior was not affected by reaction products when it was calculated at less than 5%–10% nitrate conversion, and it was possible to compare the initial rate of each catalyst as a representation of the maximal catalytic activity (Hall et al. 1976; Casado et al. 1986). The catalyst initial rate, calculated (ppm/min), can be observed in Table 2.

In addition, it was reported that the catalytic performance of Pd,In/Al₂O₃ systems in nitrate and nitrite reduction was influenced by the formation of the intermetallic phase and/or the increase in indium surface concentrations in those bimetallic systems (Krawczyk et al. 2011). Moreover, it is widely reported that the selectivity to N₂ is higher for catalysts with lower dispersion (Yoshinaga et al. 2002; Gauthard et al. 2003; Garron et al. 2005; Zhang et al. 2008a). The lowest availability of active sites saturated with hydrogen avoided the over-reduction of nitrites and increased the selectivity to nitrogen (Tokazhanov et al. 2020).

Regarding nitrite production, the highest nitrite concentration was detected (Figure 6(b)) at the beginning of the reaction with PIACR, the most active catalyst. The same behavior was observed with the formation of ammonium (Figure 6(c)), where the PIACR catalyst produced the largest amount of this reaction by-product in line with our previous work (Marchesini et al. 2012). It was reported that the amounts of produced depended on the mutual interactions between Pd and In and also on the porosity of the support (Krawczyk et al. 2011). A mutual interaction promoted a high recuperation of the reduced active sites which participated in the over-reduction from nitrite to ammonia, and porosity was involved in the generation of microenvironments which were inaccessible for the buffering system to neutralize, which increased the generation of ammonia pathway by-products. This behavior could be additionally related to the fact that Pd and In were co-impregnated, inducing a higher metal–metal interaction as revealed by the SEM-EDS analysis since the support used in all catalysts was the same (Figure 6). In terms of composition, particles with Pd and In were more active for nitrate conversion. However, there was more H-Pd availability when the surface was richer in Pd, allowing over-reduction. This situation favored reduction and higher selectivity to ammonia. Porosity had a great effect on the reaction because the system was not able to buffer the OH generated by each nitrate converted, so the environment near the active site became an alkaline region. It was reported that the alkaline pH promoted the increase of ammonia as the final product (Santos et al. 2020)

Nevertheless, no linear trend was observed between catalytic activity and ammonium production. For example, the second most active catalyst supported on alumina was PACRIR, but it produced the least amount of ammonium at the end of the reaction. There is an intense growth in knowledge about this reaction pathway (Sanchis et al. 2022). Since nitrate is converted to nitrite through a very quick reaction, it is necessary to control the average pH and the availability of H adatoms as quickly as possible. There was an exquisite balance between these variables that favored one or the other reaction pathway. In the case of the PACRIR catalyst, the Pd was calcinated and then reduced before depositing In, so the interaction generated between Pd and In was probably more intense. As mentioned above, the presence of In slightly prevented Pd from dissociatively adsorbing H, so the reaction occurred as quickly as possible, but there was no excess of H adatoms to over-reduce NO(ads) and to generate more ammonia.

However, it is important to take into account that the nitrate reduction rate and selectivity toward final products were affected by the dispersion of particles on the support, the size of the particles, and the formation of agglomerates. Dispersion and size directly affected activity and selectivity. Guo et al. (2018) found that at a low In loading, the element remained dispersed and in the metallic form, but In particles must be in contact with Pd to produce two types of sites with two different roles: (1) adsorb nitrate and (2) reduce nitrate. The role of Pd sites is to regenerate the In sites oxidated during nitrate reduction. Consequently, large In particles (due to low dispersion or agglomeration) block the Pd sites from dissociatively adsorbing H₂, and this leads to a lower reaction rate. In the opposite case, when PdH is abundant, over-reduction is predominant and selectivity leads to ammonia generation.

These factors, in turn, depend on the preparation method and the metal loading (Martínez et al. 2017). On the other hand, Pd particle size considerably affects the selectivity to N₂, with better selectivity being obtained in the presence of metallic agglomerates (Yoshinaga et al. 2002; Garron et al. 2005; Zhang et al. 2008; Soares et al. 2021). Marchesini (2008) and Zhang et al. (2008) found that selectivity was greatly correlated to the size of active metal particles. Pd–In particles tended to agglomerate in reducing aqueous media, and bigger particles increased the N(ads)/H(ads) ratio, so it is more possible that two N(ads) find each other generating N₂, resulting in a better selectivity (see the mechanism in the Supplementary Material, Equations S1–S5).

Regarding the support structure in TiO₂, Kim et al. (2013) prepared 2.5 wt.% Pd/1 wt.% Cu bimetallic catalysts supported on TiO₂ with different anatase-/rutile-phase compositions for the reduction of nitrate in water. Catalysts containing a higher anatase-phase composition were shown to have superior catalytic performance. The maximum nitrate conversion obtained was 100% in 40 min of reaction, with the catalyst with the highest content of the anatase phase (98%). In general, the TiO₂-supported catalysts prepared in this work obtained better results than the Pd,Cu/TiO₂ reported by Kim et al. (2013), with some of them reaching total conversion. It is important to highlight that this author used an almost twice-higher metallic concentration (2.5:1 wt.% Pd:Cu), so our catalyst resulted in being more sustainable. The conditions slightly favored our material since it had a lower metal loading (1:0.25 wt.% Pd:In), starting with a higher contaminant concentration (7.4 mM) and then obtaining total conversion and similar selectivity at the same time.

Thus, according to the results obtained, we can conclude that the most active catalysts in the catalytic reduction of nitrate were those prepared on alumina, followed by titania and then by the least active one supported on silica. The results obtained showed that nitrites were present in high concentrations in the reaction mixture only in the initial stage of the reaction for Pd,In/Al₂O₃ and Pd,In/SiO₂ catalysts. A lower concentration of this product was not observed when using Pd,In/TiO₂.

On the other hand, taking into account the specific surface of supports used (Al₂O₃, SiO₂, and TIO₂), it is widely known that the order in increasing values of the surface area is TiO₂ > Al₂O₃ > SiO₂. However, the catalytic performances found resulted in the following increasing order of activity: Al₂O₃ > TiO₂ > SiO₂. Therefore, we can assume that the catalytic performance was mostly affected by the type of interaction between the active phases (Pd and In) and between the active phases and the support (Sanchis et al. 2022).

It is important to consider that generally, supports that can disperse the In₂O₃ aggregates with high In stabilization could give an increase to active catalytic systems. In this work, alumina and titania supports of the catalysts with the highest catalytic activity have been reported as supports with high In dispersion. It is known that a balance between basic and acidic sites on the support surface plays an important role in the stabilization of In₂O₃ particles. Silica is a very low acid support which favors the agglomeration of the In particles (Gervasini et al. 2006), decreasing the contact with Pd which is essential for the development of the reaction. This behavior was observed by SEM-EDS analysis. Comparing the distribution of In in the catalysts prepared on SiO₂ (Figures S1 and S2) with those supported on Al₂O₃ (Figures S4–S8), it can be observed that alumina significantly favored the distribution of In on its surface, while silica generated In agglomerates and a less homogeneous distribution.

Table 2 summarizes the selectivities toward nitrite, nitrogen gas, and ammonium with those catalysts that reached 90% conversion. Also, a comparison with the three types of supports for the catalysts used in this work, in the same nitrate conversion, is shown in the Supplementary Material (Table S2).

Firstly, it is important to mention that the catalysts that promoted catalysts with a catalytic activity greater than 90% were those supported by Al₂O₃ and TiO₂. Regarding N₂ selectivity, the most efficient for the alumina group was PACRIR, with 94.4% selectivity to N₂. Even so, the Pd 1%, In 0.25% catalysts already reported in the literature showed lower selectivity to N₂ at 90% conversion.

On the other hand, the ammonium selectivity, which is the main drawback in this process, of the alumina group was between 1.6% and 16% at the same nitrate conversion (42%), the PACRIR catalyst being the one that achieved the lowest ammonium selectivity (see Table S2). The group of catalysts supported on SiO₂ presented ammonium selectivity in the range of 11.2%–17.1%, and the catalysts supported on TiO₂ presented higher ammonium selectivity values between 7.1% and 32.1%.

In a previous work (Marchesini et al. 2008b), a selectivity to N₂ of 75.6% was obtained by achieving 90% nitrate elimination. In that work, the reduction step during the synthesis was carried out with H₂ at 450 °C and the active phase was added in a single impregnation, comparable to the PIACR of this work S(N₂) = 78.3%). Since the operating conditions were the same, the main changes involved the interactions between the active phase and support during the synthesis. On the one hand, the results of the present work may show an advantage in using liquid-phase reduction compared with gas-phase reduction. Moreover, the addition of Pd and In in different steps represented a distribution of the particles on the surface in a way that favored the selectivity toward N₂. In this sense, it is reasonable that the PIACR catalyst presented the lowest selectivity to N₂ since it showed good distribution of Pd and In on the support surface, and generally, the selectivity to N₂ was higher for catalysts with lower dispersion (Yoshinaga et al. 2002; Garron et al. 2005; Zhang et al. 2008; Soares et al. 2021).

Regarding TiO₂ catalysts, the best selectivity to N₂ (82.4%) was obtained with PTCRI. The catalysts supported on TiO₂ showed a higher selectivity to compared with alumina and PACRIR (17.0% and 4.4%, respectively), indicating that the interaction between the active phase and support was different. In the case of TiO_2, as mentioned in the discussion of the SEM-EDS results, there may be more formation of Pd species with low coordination numbers, i.e., corner and edge positions due to the strong interaction of Pd with the support (Sá & Vinek 2005). These terminations have a higher hydrogenation capacity, facilitating the reduction of nitrite to at these sites. These results are contrary to those obtained by Krawczyk et al. (2011), where the Pd5%/In2%/TiO₂ catalysts with the lowest ammonium production were those supported on titania, compared with those supported on alumina, and at the same time showed the highest selectivity to N₂. The physicochemical properties of TiO₂ in Krawczyk et al. (2011) were different. It had a range of pore diameters of 100–300 Å, so the capacity to neutralize the OH produced during the nitrate reduction was higher than the one from this work. In this sense, the local pH developed inside the pores must be higher than the one produced on titania in Krawczyk et al. (2011).

Regarding the initial rates, no direct connection was found with the Pd content (see Table S4 in the Supplementary Material), indicating that other factors had a greater impact on this variable. As for alumina-supported catalysts, those in which In was added at the end (PACRI) showed a higher v_i than the same catalyst in which reduction took place after the addition of In (PACRIR). Among these catalysts, reduction after In deposition induced a small surface loss of Pd (Table 1) compared with the catalyst in which no reduction step was performed after In deposition (see Table S2).

Conversely, PTCRI presented a lower initial rate than PTCRIR with the TiO₂ support. Thus, although both catalysts showed similar values of N₂ selectivity and nitrate conversion at 120 min, the fact that there was a higher surface Pd (PACRI) led to higher conversion at the beginning of the reaction.

(Pd,In)-based catalysts deposited by wet impregnation onto alumina (Al₂O₃), silica (SiO₂), and titania (TiO₂) were prepared and evaluated for nitrate removal from water to study the influence of supports on catalytic performance and selectivity. In comparison with Al₂O₃ and SiO₂, a strong interaction between the TiO₂ support and its active phase was observed. Catalysts supported onto Al₂O₃ and SiO₂ showed a weaker MSI which favored N₂ selectivity and less retention of Pd and In in the bulk. The alumina-supported catalysts generally showed better catalytic activity, followed by the titanium- oxide-supported catalysts. Different sequential processes were applied to obtain catalysts that were more stable, active, and selective to nitrogens. The PACRIR catalyst, whose process includes the deposition of Pd over alumina + calcination + reduction + In impregnation and final reduction, was the sequence that produced the better sum of properties. The best compromise between activity and selectivity was obtained with the alumina-supported catalyst, which is synthesized by the PACRIR method. These results suggest that bimetallic In on Pd-supported catalysts can reduce efficiently the nitrate content in water, and a sequential synthesis is a powerful tool for improving this capability and increasing selectivity to N₂.

The authors wish to acknowledge the financial support of FIQ-UNL, INCAPE, CONICET, ANPCyT, PICT 2019 2019-02970, CAI + D 2020 50620190100148LI, of Argentina, Red CYTED Nro. 318RT0551, CNPq and CAPES from Brazil. Thanks are also given to translator Yanina Burkett Beliz for editing the English-language manuscript.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.