**Thermal and Catalytic Pyrolysis of Dodecanoic Acid on SAPO-5 and Al-MCM-41 Catalysts**

#### **Carolina Freitas 1, Marizania Pereira 2, Damari Souza 2, Noyala Fonseca 3, Emerson Sales 1,3 , Roger Frety 1,3, Camila Felix 4, Aroldo Azevedo Jr. <sup>5</sup> and Soraia Brandao 1,2,\***


Received: 25 February 2019; Accepted: 27 April 2019; Published: 3 May 2019

**Abstract:** In this study, dodecanoic acid was decomposed during fast pyrolysis experiments either thermally or in the presence of SAPO-5 and Al-MCM-41catalysts. The catalysts were synthesized by a hydrothermal route and subsequently characterized by XRD, TPD-NH3, and TGA, and dodecanoic acid was characterized by TGA and DSC. Analysis of the post-pyrolysis products was performed online by gas chromatography coupled with mass spectrometry (GC-MS). The results from pyrolysis at 650 ◦C indicated that the nature of the catalysts strongly influences the composition of the products. Linear alkenes were standard products for all pyrolysis experiments, but with Al-MCM-41, various alkene isomers with a linear and cyclic structure formed, as well as saturated and aromatic hydrocarbons. As a whole, Al-MCM-41 led to a much higher dodecanoic acid conversion and higher deoxygenation than SAPO-5. As these catalysts present small differences in strong acid site density, the difference in the global conversion of dodecanoic acid could be attributed to textural characteristics such as pore volume and surface area. In this case, the textural properties of the SAPO-5 are much lower when compared to Al-MCM-41 and, due to a lower accessibility of the reactant molecule to the acidic sites of SAPO-5, partially blocked for fatty acid molecules by the considerable amount of amorphous material, as detected by XRD.

**Keywords:** fast pyrolysis; SAPO-5; Al-MCM-41; dodecanoic acid

#### **1. Introduction**

The global energy matrix is still based on non-renewable sources and fossil fuels. The use of these sources generates pollution and emissions of greenhouse gases, leaving environmental problems for future generations. Since the second half of the 20th Century, with oil crises, rising fuel demand, and growing environmental awareness, research into alternative energy sources has grown. Renewable sources are now much favored due to their full availability, biodegradability, and low cost [1].

Vegetable oils have a large amount of triacylglycerides (about 90%) and a lower amount of free fatty acids, mono-, and di-glycerides (8–10%). Among the acids present in the various oily compounds, saturated fatty acids, such as palmitic (C16:0) and stearic (C18:0), and unsaturated, such as linoleic

(C18:2) and oleic (C18:1), are the most common, but other acids, such as lauric (C12:0) and myristic (C14:0), exist in specific oils and fats [2].

Brazil has several sources for the supply of oily compounds, such as soybean (whose predominant fatty acid is linoleic acid), babassu (lauric acid), and animal fat (stearic acid). Research into oils from other biomasses has been carried out, for example, castor oil, palm oil pulp, palm kernel almond, babassu coconut kernel, sunflower seeds, and other raw materials [3]. The country is the fourth largest producer of coconut in the world with almost three million tons of fruit per year, and the State of Bahia is the largest producer. Coconut and coconut oil are relevant natural sources of saturated fats, based mainly on dodecanoic acid (C12:0), also known as lauric acid. Coconut oil is obtained from the pulp of mature fresh coconut (*Cocos nucifera* L.) and contains more than 80% of saturated fatty acids, such as caproic, caprylic, capric, lauric, myristic, palmitic, and stearic acids. Coconut oil has a concentration of lauric acid above 40% [4].

Pyrolysis or thermal cracking of triglycerides is a well-known method for producing renewable fuels (Bressler, Maher). Pyrolysis operates in the absence of external O2, providing high yields in liquids when performed at elevated temperatures (400–600 ◦C), with vapor residence times lower than 5 s and heating rates between 10 and 200 ◦C/s [2]. Pyrolysis of triglycerides can be summarized by Reaction 1:

Thermal cracking produces saturated, unsaturated linear hydrocarbons, or both, via the deoxygenation of carboxylic acids, which can occur through two distinct reaction paths: decarboxylation and decarbonylation. In decarboxylation, the carboxyl group of the fatty acid is removed, forming a saturated hydrocarbon and CO2 (Reaction 2). The decarbonylation consists of the removal of the carbonyl of the fatty acid forming CO, H2O, and olefins (Reaction 3) [1].

In addition to the formation of the primary pyrolysis products, many parallel and sequential transformations can occur, leading to a series of oxygenated and deoxygenated molecules with a smaller or higher number of carbon atoms, compared to the initial carbon chains, such as olefins, paraffins, ketones, aldehydes, and aromatics, among others [5].

One way of making pyrolysis more efficient is to promote the reaction in the presence of a suitable catalyst. In recent years, studies dealing with the catalytic conversion of oily biomass feeds, through thermochemical reactions, have reported impressive results, and many reviews have summarized the positive and negative aspects of such transformations [6–8]. Further, current research has been done into the use of multi-element and multi-functional catalysts and complex mixtures of feeds in coprocessing strategies [9–12]. However, much essential work is still needed to understand the complexity of the reaction mechanism sequences, and therefore, studies with simpler systems, or model molecules, must be performed.

The interest of the petrochemical and chemical industries in the use of molecular sieves is due to the application of these types of materials in catalytic cracking, isomerization, and alkylation. Molecular sieves have a combination of important properties, such as surface area, large ion exchange capacity, and strong acidity, with high thermal and hydrothermal stability. Due to the regularity of their crystalline channels, molecular sieves allow higher selectivity of the reagents, products, and transition states, on a molecular scale [13]. Catalysts based on molecular sieves with different crystalline structures can induce changes in the product distribution, in the conversion of bio-oils and triglycerides. An example is the several zeolites, such as ZSM-5, Y, and beta, and silicoaluminophosphates studied in bio-oil cracking [3].

The SAPO-5 catalyst has one-dimensional 12-membered straight channel rings. It is a crystalline and microporous material, presenting excellent thermal and hydrothermal stability, and its acidity reaches intermediate values between the zeolite and the AlPO (aluminophosphate molecular sieves). Due to its topology, SAPO-5 provides potential catalytic applications, such as catalytic cracking, isomerization, and alkylation reactions [14].

The molecular sieve MCM-41 belongs to the class of materials known as M41S (Mobil 41 Synthesis). They are Si-based materials with highly-ordered mesopores and a high surface area. MCM-41, consisting of pure silica, presents low acidity. The incorporation of aluminum into the MCM-41 structure increases the number of acid sites, improving catalytic activity [15].

In this work, molecular sieves SAPO-5 and Al-MCM-41 were synthesized and evaluated as catalysts in the fast pyrolysis of dodecanoic acid, a simple model molecule of fatty compounds. The differences between thermal and catalytic pyrolysis, as well as the distribution of the products formed were analyzed.

#### **2. Results and Discussion**

Although the textural properties of the present samples were not determined, other samples of SAPO-5 and Al-MCM-41 using the same synthesis procedures have been prepared by some of us. In these cases, surface areas of SAPO-5 and Al-MCM-41 ranged from 247–261 and 485–510 m2/g, respectively, whereas pore volumes were in the order of 0.19 and 0.85 cm3/g, respectively. Therefore, Al-MCM-41 presented a higher surface area and a much larger pore volume than SAPO-5 [16].

#### *2.1. X-ray Di*ff*raction*

Figure 1a shows the diffractogram of the SAPO-5 molecular sieve. The material presented a characteristic diffractogram of the AFI structure, similar to that found in the literature [17]. The presence of the intense halo at about 2θ = 15–35◦ was associated with some extra frame silica.

The diffractogram of the Al-MCM-41 catalyst shown in Figure 1b presented a strong reflection at 2θ = 2.1◦, characteristic of the MCM-41 structure. The two peaks of lower intensity and greater width observed at 2θ = 3–4◦ suggest a loss of homogeneity and hexagonal symmetry of the crystalline structure due to the incorporation of aluminum in the structure of MCM-41 [18–20].

**Figure 1.** X-ray diffraction of SAPO-5 (**a**); X-ray diffraction of Al-MCM-41 (**b**)

#### *2.2. Thermodesorption of Ammonia*

Thermodesorption (TPD) of ammonia is often used to estimate the acidity of the molecular sieves and establish strong correlations both for the extent of the reactions and product selectivity [21]. TPD-NH3 profiles of the SAPO-5 and Al-MCM-41 catalysts in Figure 2a,b, respectively, are generally divided into three temperature ranges: (i) below 300 ◦C is associated with weak acid sites; (ii) between 300 and 500 ◦C with acidic sites of moderate strength; and (iii) above 500 ◦C corresponding to strong acid sites [22].

**Figure 2.** TPD ammonia analysis of SAPO-5 (**a**); TPD ammonia analysis of Al-MCM-41(**b**).

Table 1 presents the concentrations, in mmol/g, of the acid sites derived from the deconvolution of the experimental NH3 thermodesorption curves, using a Gaussian distribution function.

**Table 1.** Concentration (×10−<sup>2</sup> mmol/g) of SAPO-5 and Al-MCM-41 acid sites.


For the SAPO-5 catalyst, there was a predominance of weak and moderate acid sites with maximum desorption temperature respectively at 272 ◦C and 427 ◦C, in addition to the presence of a peak ascribed to strong acid sites, at a temperature of 654 ◦C, with a lower concentration. Al-MCM-41 had three desorption peaks at maximum temperatures of 184 ◦C, 470 ◦C, and 767 ◦C, respectively. The highest concentration, in mmol/g, found for the Al-MCM-41 catalyst was for moderate acid sites. The desorption peak at a low temperature (<300 ◦C) probably refers to weak Brønsted acid sites, which are bound to aluminum with tetrahedral coordination (AlO4−). Thus, a high temperature (>500 ◦C) desorption may be ascribed to ammonia adsorbed in Lewis acid sites on the surface of the molecular sieve, created by the dehydroxylation of the aluminum atoms substituting silicon atoms in the structure [21,22].

#### *2.3. Thermogravimetric Analysis*

The profiles of the thermogravimetric analysis curves of the Al-MCM-41 and SAPO-5 samples are shown respectively in Figure 3a,b.

**Figure 3.** TGA profile of SAPO-5 catalyst (**a**); TGA profile of Al-MCM-41 catalyst (**b**).

In Figure 3a, the thermogravimetric analysis of the SAPO-5 catalyst is presented, in which three regions of mass loss can be observed. The first 8% mass loss event, from room temperature up to 200 ◦C, was associated with the desorption of adsorbed water on the outer surface and occluded in the pores. The second step had a loss of 10% in the range of 200 ◦C–470 ◦C, with a maximum rate from 440 ◦C–470 ◦C, which was ascribed to the desorption of triethylamine (template) occluded in the channels. Finally, the third region, between 470 and 750 ◦C, with a loss of 2%, was related to the desorption of the occluded channel surfactant and to the decomposition of the protonated amine (TEA+) [23].

In Figure 3b, the first mass loss event (2%) was related to the loss of water physically adsorbed in the pores of the Al-MCM-41 catalyst. The second mass loss event (33%) was attributed to the degradation of the organic molecules of the template CTMABr (hexadecyltrimethylammonium bromide). The third event (10%) was associated with the removal of the residual template and also water resulting from the secondary condensation of the silanol groups, according to the literature [24,25].

The TGA/DTG curves of pure lauric acid, Figure 4, exhibited a region of significant mass loss (95%) between 150 ◦C and 250 ◦C, followed by small mass losses up to 600 ◦C. The excessive loss of mass was attributed to the volatilization of lauric acid and the losses at higher temperatures to progressive degradation of non-volatile residues.

**Figure 4.** TGA/DTG profile of lauric acid.

#### *2.4. Di*ff*erential Scanning Calorimetry*

The DSC profile of pure lauric acid, Figure 5, shows a succession of endothermic and exothermic events. The endothermic phenomenon appearing at around 50 ◦C, not associated with mass loss, according to Figure 4, was related to the fusion of lauric acid. At about 200 ◦C, a second weak endothermal phenomenon was observed, probably related to the initial volatilization of lauric acid, quickly followed by an exothermic event linked to the processes of cracking and autoxidation, possibly associated with the degradation of the fatty acid. The results obtained by the DSC and TGA/DTG analyses suggest that the lauric acid was fully converted to the vapor phase below 250 ◦C and transformed into products between 250 and 600 ◦C.

**Figure 5.** DSC profile of lauric acid.

#### *2.5. Fast Pyrolysis Reaction*

The fast pyrolysis experiments with pure lauric acid and lauric acid adsorbed on both molecular sieve catalysts was performed at 650 ◦C under helium flow and with an estimated heating rate of 1000 ◦C s <sup>−</sup>1. The global results of the products obtained from thermal and thermo-catalytic pyrolysis are shown in Figures 6–8. The deoxygenated compounds were grouped into different product families to facilitate the analysis of the products. The remaining reactant is also indicated in Figures 6 and 7, as the observed conversions were 9% and 12%, respectively. When using the Al-MCM-41 catalyst (Figure 8), no lauric acid was detected after pyrolysis, indicating a total conversion.

**Figure 6.** Compounds identified after fast pyrolysis of pure lauric acid at 650 ◦C.

**Figure 7.** Compounds identified after fast pyrolysis of lauric acid at 650 ◦C in the presence of SAPO-5.

**Figure 8.** The yield of fast pyrolysis products of lauric acid at 650 ◦C in the presence of Al-MCM-41.

In thermal pyrolysis, limited deoxygenation of lauric acid occurs via the decarbonylation route, mainly giving α-olefins (C3 to C11) [8], as observed in Figure 6. Pyrolysis of fatty acids can proceed through the breaking of C-C and C-O bonds. Two competing routes can explain the presence of products with fewer than ten carbon atoms: (i) deoxygenation of the starting acid, followed by breaking of the C-C bond to produce hydrocarbon radicals; and (ii) cracking of the C-C bond of the chain, followed by deoxygenation of a shorter chain carboxylic acid. With saturated compounds, such as lauric acid, the first route is favored [16,24].

In the catalytic pyrolysis of the acid adsorbed on SAPO-5 (Figure 7), a slightly larger number of α-olefins (from C3–C11) were formed, which consisted of potential feedstock to produce high octane gasoline. A similar study with the same catalyst (SAPO-5), using palmitic acid, i.e., a saturated fatty acid with a carbon chain longer than lauric acid, can be found in [16]. In both cases, the pyrolysis products presented great similarity since they showed a high distribution of linear mono-unsaturated hydrocarbons.

For pyrolysis of lauric acid in the presence of Al-MCM-41 catalysts (Figure 8), C3–C11 olefins (68%) including isomers with internal C=C bonds, both in cis and trans configuration, cyclic mono-unsaturated molecules, and methylated molecules, were identified. Cyclic and alkylated cyclic saturated products (14%) and aliphatic alkanes (4%), together with mono-aromatics (benzene, toluene, and p-xylene) (4%), were also found. The formation of both aromatic and saturated products can be attributed to the acidity of the Al-MCM-41 catalyst, which should promote hydrogen transfer reactions. The hydrogen liberated during aromatic formation could be in part used to saturate olefin-based products and possibly helped the alkylation reaction.

Figure 9 presents the normalized distribution of the α-olefins obtained (an area greater than 0.5% of the total area) after thermal pyrolysis of lauric acid at 650 ◦C and catalytic pyrolysis of lauric acid with SAPO-5 at the same temperature. The SAPO-5 catalyst did not alter the distribution of the products significantly, compared to the thermal pyrolysis. Only the conversion was slightly higher (12% and 9%, as seen in Figures 6 and 7), which is not visible in the figure because of normalization.

**Figure 9.** Normalized distribution of the α-olefins obtained after pyrolysis of lauric acid at 650 ◦C: (-) SAPO-5 catalyst and (-) no catalyst.

Table 2 presents the C6 (6 isomers) and C11 (13 isomers) olefins identified after dodecanoic acid pyrolysis at 650 ◦C in the presence of Al-MCM-41, limiting the identification of the products to those with a relative area greater than 0.5% of the pyrogram total area. Among the main products, linear, cyclic, and alkylated olefins were identified, as well as cis and trans isomers.

When considering the complete family of monounsaturated molecules obtained in the presence of Al-MCM-41, represented in Figure 10, the amount of isomers does not increase continuously with chain length; C8 olefins have both a lower number of identified isomers (two) and the most moderate content. It suggests a potential application for gasoline and kerosene fuels, after partial hydrogenation.


**Table 2.** C6 and C11 mono-unsaturated main products observed during dodecanoic acid pyrolysis at 650 ◦C using Al-MCM-41 catalyst.

**Figure 10.** Mono-unsaturated isomers' distribution from pyrolysis of lauric acid at 650 ◦C, in the presence of Al-MCM-41.

#### *2.6. Comparison between the SAPO-5 and Al-MCM-41 Catalytic Behavior*

The extent of lauric acid conversion between the three pyrolysis reactions was different: 9% was observed during thermal pyrolysis, 12% when lauric acid was adsorbed on the SAPO-5 catalyst, and about 100% when the fatty acid was adsorbed on Al-MCM-41 catalyst. The results are consistent and close to those found in the literature regarding the pyrolysis of saturated fatty acids [4]. However, the results obtained for pyrolysis of unsaturated fatty acids on Al-MCM-41-based catalyst, similar to those in the present work, showed that deoxygenation was not as effective as for saturated acids [19].

An essential point of the discussion is linked to the low conversion of lauric acid observed when using SAPO-5 catalyst compared to Al-MCM-41. The TPD of NH3 profiles are different for these two catalysts, but small differences in the number of strong acid sites were observed; this alone cannot explain the significant differences in conversion between both molecular sieves. Specific surface area and pore volume of SAPO-5 were much lower than the values for Al-MCM-41. Moreover, a large amount of amorphous material in SAPO-5 was revealed by XRD analysis, and this can cause lower accessibility of the dodecanoic acid molecule to acid sites in SAPO-5 catalyst. These two last characteristics are probably responsible for the observed differences in performances between both

catalysts. The access of large lauric acid molecules to the acidic sites located inside the porous structure of SAPO-5 could be limited for two main reasons: (i) a lower mean diameter of SAPO-5 pores compared to that of Al-MCM-41; (ii) the pore mouth in SAPO-5 can be partially blocked by the amorphous species. In other words, only a small proportion of SAPO-5 can interact with dodecanoic acid, limiting the conversion of the fatty acid during pyrolysis experiments.

#### **3. Materials and Methods**

#### *3.1. Catalyst Preparation*

The synthesis of the SAPO-5 catalyst was carried out following the methodology found in [4], and some modifications were made to adjust the Si/Al ratio. Pseudoboehmite and orthophosphoric acid (85%, Sigma-Aldrich) were used as sources of Si and P, respectively. Then, the triethylamine (Sigma-Aldrich, St. Louis, MO, USA) template was added to tetraethyl orthosilicate (98%, Sigma-Aldrich, St. Louis, MO, USA) and the surfactant hexadecyltrimethylammonium bromide (Sigma-Aldrich, St. Louis, MO, USA), in solution with hexanol (Sigma-Aldrich, St. Louis, MO, USA). Initially, 19.56 mL of orthophosphoric acid (previously diluted in 20 mL of deionized water) were placed in a beaker containing 22.4 g of pseudoboehmite (dispersed in 63 mL of H2O); this system was kept under stirring for two hours. After that, 38.64 mL of triethylamine were added, and stirring was continued for a further two hours. After this time, a solution prepared by the addition of 4.82 mL of TEOS (Tetraethyl Orthosilicate) + 78.82 mL of hexanol + 3.8 g of CTMABr (hexadecyltrimethylammonium bromide) was added and again stirred for another two hours. The mixture was partitioned into autoclaves and put into an oven preheated to 170 ◦C, where it remained for 24 h. The material obtained after this period was centrifuged, washed with alcohol and distilled water, respectively, and again routed to a drying oven at 100 ◦C for six hours.

A sample of Al-MCM-41 with a silica/alumina (SAR) ratio of 60 was prepared. Commercial silica (Aerosil 200) and sodium aluminate were used as sources of Si and Al, respectively. As a guideline of the mesoporous structure, hexadecyltrimethylammonium bromide (CTMABr) was used in the form of 25% *w*/*w* aqueous solution. Sodium hydroxide solution (50% *w*/*w*) was used to maintain the basic pH. The molar composition of the synthesis gel was: 1 SiO2: 0.025 Al2O3: 0.08 Na2O: 0.3 CTMA: 26 H2O.

The samples were prepared from suspensions and solutions named as A, B, C. In Suspension A, 20.2 g of silica (SiO2) were mixed in 38.3 g of distilled water. For Solution B, 2.8 g of sodium hydroxide (NaOH) were added to 2.8 g of distilled water, and in Solution C, 36.8 g of hexadecyltrimethylammonium bromide (CTMABr) were dissolved in 110.5 g of distilled water.

In Solution B, the amount of 0.9 g of sodium aluminate and then Solution C were added under vigorous and constant stirring until complete dissolution of the sodium aluminate. Solution A was then added under vigorous stirring at room temperature. The mixture was then placed in autoclaves without stirring and heated at 150 ◦C for 48 h.

Then were added 0.9 g of sodium aluminate to Solution B and then Solution C added under vigorous and constant stirring until complete dissolution of the sodium aluminate. Suspension A was submitted to vigorous stirring at room temperature. The mixture was then placed in autoclaves without stirring and heated at 150 ◦C for 48 h.

After this procedure, the product was washed with distilled water separated using a centrifuge and oven dried at 100 ◦C for 4 h. After this step, the material was calcined to remove the remaining template in the porous structure.

Al-MCM-41 catalyst was calcined under an inert atmosphere (N2, 30 mL/min) to 370 ◦C for 1 h 30 min to remove the remaining template in the porous structure. After this step, the N2 flux was replaced by a synthetic air flow (30 mL/min) maintaining the temperature of 550 ◦C for over ten hours for remaining burning of the template.

The SAPO-5 catalyst was calcined in an inert atmosphere (N2, 100 mL/min) to 450 ◦C (10 ◦C/min), remaining for one hour at this temperature, to remove the remaining template in the structure. After

this stage, the N2 was replaced by synthetic air (100 mL/min), maintained at a temperature of 550 ◦C for 5 h to burn the coke formed with the decomposition of the template.

#### *3.2. Catalyst Characterization*

#### 3.2.1. X-ray Diffraction

The X-ray diffraction measurements were conducted on a Shimadzu apparatus, model XRD-6000, using Cu Kα radiation (λ = 1.5418 Å) (Shimadzu, Kobe, Japan). The diffractograms were collected in a range from 1–50◦, with a scanning speed of 0.25◦ min−1. They were obtained at a constant power source of 40 kV and 30 mA, at room temperature. The powder samples were analyzed without any previous treatment.

#### 3.2.2. Temperature Programmed Desorption of Ammonia (TPD-NH3)

The measurement of the density and strength of the acid sites of the samples was carried out using the temperature programmed desorption (TPD-NH3). Initially, the ammonia was adsorbed at room temperature for 1 h. The samples were then heated at a rate of 10 ◦C/min to 150 ◦C for 1 h under He flow to remove the poorly-adsorbed molecules. The ammonia desorption step was performed under a constant helium flow of 25 mL/min to 700 ◦C, using a heating rate of 10 ◦C/min. The equipment used was the Chemisorb 2720, Pulse Chemisorption System, Micromeritics, equipped with a thermal conductivity detector (TCD), a quartz reactor, and gas supply system.

#### 3.2.3. Thermogravimetric Analysis

The samples (2 mg of mass) were heated with a temperature ramp of 10◦C/min from ambient temperature to 1000 ◦C in a Shimadzu apparatus, Model TGA-50 (Shimadzu, Kobe, Japan), under a constant flow of 50 mL/min of nitrogen.

#### *3.3. Catalytic Test*

The pyrolysis experiments were performed on a multi-shot pyrolyzer model EGA/PY-3030D (Frontier Laboratories LTD, Fukushima, JPN) connected online with a GC-MS-5799 (Agilent, Santa Clara, CA, USA). In the thermal fast pyrolysis, 1.1 milligrams of pure lauric acid were used, whereas in the catalytic pyrolysis tests, 1.1 mg of the lauric acid/catalyst mixture were used.

The mixture of the lauric acid with the catalyst was carried out with the addition of the fatty acid to the catalyst in the proportion 1:10 (m/m), by mechanical mixing. Subsequently, the mixture was heated to 60 ◦C, a temperature higher than the melting point of lauric acid (43 ◦C), and continuously homogenized. This treatment allowed the lauric acid to migrate inside the pores of the catalysts. After this step, the mixture still in powder form was stored away from moisture, then added to the titanium sample holder, and covered with quartz wool.

The samples were dropped by gravity into the pyrolysis furnace preheated to 650 ◦C under a helium atmosphere and remained under these conditions for 20 s before an aliquot of the gaseous atmosphere was introduced into the analytical system. For the chromatographic analysis of the products, an HP-5MS column was used with a heating ramp comprising an initial temperature of 40 ◦C for 3 minutes, followed by a ramp of 20 ◦C/min to 320 ◦C, with the temperature maintained for 20 min. The rate of helium flow through the column was 2 mL/min. The temperature of the pyrolyzer interface was 320 ◦C. The products obtained by GC/MS were identified based on the retention times and by comparison with the standard mass spectral fragments of the National Institute of Standards and Testing database.

#### **4. Concluding Remarks**

The presence of the catalysts modified the quantity and distribution of the products. The fast pyrolysis in the presence of catalysts based on mesoporous and microporous molecular sieves allowed the production of olefins in the gasoline range. Using both thermal and catalytic pyrolysis with the SAPO-5 catalyst, the conversion of dodecanoic acid was limited, 9% and 12%, respectively, and the main products were terminal olefins (C3–C11). In the presence of the Al-MCM-41 catalyst, the total conversion of the lauric acid was observed, and the main products were internal and alkylated olefins, fundamental for both chemical and petrochemical industries.

**Author Contributions:** S.B. has conceived the project, funding acquisition, designed the experiments, analyzed the data and written the paper. C.F. (Carolina Freitas), M.P., D.S. and N.F. have designed and performed the experiments, analyzed the data and written the paper. E.S., R.F., C.F. (Camila Felix) and A.A.J. have designed the experiments, analyzed the data and written the paper.

**Funding:** FAPESB (Bahia Research Foundation), project DTE 0043/2011, and CNPq, project 552367/2011-7 funded this research.

**Acknowledgments:** The authors are grateful to CNPq, FAPESB, and CAPES for scholarships and financial support. **Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Dibenzothiophene Hydrodesulfurization over P-CoMo on Sol-Gel Alumina Modified by La Addition. E**ff**ect of Rare-Earth Content**

**José Escobar 1,\* , María C. Barrera 2, Jaime S. Valente <sup>1</sup> , Dora A. Solís-Casados 3, Víctor Santes <sup>4</sup> , José E. Terrazas <sup>2</sup> and Benoit A.R. Fouconnier <sup>2</sup>**


Received: 11 March 2019; Accepted: 9 April 2019; Published: 13 April 2019

**Abstract:** Alumina-lanthana (La at 1, 3, or 5 wt%) supports were prepared by sol-gel from Al alkoxide sol where La(NO3)3 was added. Annealed (550 ◦C) xerogels were characterized by N2 physisorption, thermal analysis (TG-DTA), X-ray diffraction (XRD), scanning electron microscopy- energy dispersive spectroscopy (SEM-EDS), CO2-adsorption studied in IR region, Raman and ultraviolet-vis (UV-vis) spectroscopies. The texture of amorphous binary matrices of high La dispersion was adequate to applications in catalysts for middle distillates hydrodesulfurization (HDS). Generally, the amount and strength of surface basic sites increased with La content in solids. Mo (at 2.8 at. nm−2) and Co (at Co/(Co+Mo) = 0.3) were deposited over carriers by one-pot simultaneous impregnation in the presence of PO4 <sup>3</sup><sup>−</sup> (P2O5/(NiO+MoO3) = 0.2 mass ratio). Calcined (400 ◦C) Co-Mo-P impregnated precursors had decreased basicity as to that of corresponding carriers, suggesting strong La-deposited species interaction. As La content in carriers increased Mo=O Raman stretching vibrations shifted to lower wave-numbers (949 to 935 cm−1) suggesting octahedral molybdates coordination change to tetrahedral. Although La at the lowest concentration (1 wt%) enhanced dibenzothiophene, HDS (~38% higher as to the Al2O3-supported formulation) desulfurization was significantly diminished at augmented content. Presence of hardly sulfidable tetrahedral Mo originated during impregnation at basic conditions in pores of La-modified carriers seemed to dictate observed behavior. Rare earth content in formulations enhanced selectivity to biphenyl.

**Keywords:** hydrodesulfurization; CoMo/Al2O3; basic additive; lanthanum

#### **1. Introduction**

Improved catalytic formulations of enhanced properties in hydrodesulfurization (HDS) reactions that allow for compliance with ever stricter environmental regulations on S content in internal combustion engine fuels remains a challenging research field. In this context, one of the most relevant lines of investigation involves preparation of improved catalyst supports, where conventional active phases used in formulations applied at the industrial hydrotreating scale (sulfided Mo or W promoted by either Co or Ni) could have enhanced properties, including better dispersion and sulfidability or more efficient promoter integration [1–3]. In this regard, the effects of using carriers with augmented surface basicity, as to that of conventionally used alumina, still remains a matter of debate. Some reports have shown that over carriers of increased basicity, by alkaline [4] or alkaline-earth metal [5] species addition, catalysts of enhanced selectivity to desulfurization and limited hydrogenating properties could be obtained. That effect has been attributed to several factors, for instance, decreased surface acidity by basic agent addition [4] or formation of oxidic species (magnesium molybdate and NiO-MgO solid solution, [5]) that resulted after sulfiding in MoS2 (from MgMoO4, for example) phases of increased slab length and diminished saturation ability. However, decreased activity for organo-S species desulfurization (thiophene) was also observed [4], although in a lower degree, as to that registered for alkenes hydrogenation. Formation of K-decorated and K-intercalated MoS2 phases supported on SBA-15 mesoporous materials of enhanced activity and selectivity to methanethiol from a H2S-CH4 mixture have been recently reported [6]. According to the authors, electron donation from K atoms could enhance the electron density of sulfided Mo species that facilitated formation of increased amounts of Mo-coordinatively unsaturated sites (CUS).

Environmentally friendly lanthanum is a cheap rare earth that has been added to catalyst carriers in order to increase their thermal resistance [7] and surface basicity [8]. For instance, decreased acidity of alumina-La oxides contributed to augmented stability in ethanol conversion, due to diminished coke formation [9]. Also, due to that diminished acidity, corresponding rare-earth-modified Al2O3 had enhanced selectivity to ethylene (instead of to carbonaceous species) at high ethanol conversion [10]. Augmented basicity provoked by La addition also resulted in improved activity in a rare-earth-modified Ni/alumina formulation tested in CO2 methanation, that fact being originated by stronger carbon dioxide adsorption as surface carbonates that acted as so-called reactant reservoirs [11]. Higher stability of La-modified HZSM-5 zeolite catalysts in methyl mercaptan decomposition, as to that of the non-doped material, has been attributed to decreased Brønsted acidity [12].

Regarding lanthanum-containing formulations applied as HDS catalysts carriers, to our best knowledge, the information available is rather scarce [13,14]. Diminished activity in thiophene HDS of sulfided CoMo/alumina catalysts modified by La has been attributed to enhanced proportions of isolated MoO4 <sup>2</sup><sup>−</sup> refractory and less-sulfidable species adsorbed on lanthana domains [14]. From their studies on molybdenum supported on La-modified alumina, Massoth et al. [13] found that strong interaction of rare earth oxidic domains with Al2O3 resulted in LaO monolayer dispersion. At high La loadings, however, a second lanthanum oxide layer could be formed right over the first, the latter being partially sulfidable under 10% H2S/H2 treatment (400 ◦C, 2 h). Higher activity, as to that of non-doped alumina-supported sulfided Mo catalysts, in both thiophene conversion and hexene saturation was found for formulations of high-La content (≥ 10 wt%). Different trends in those reactions, as a function of rare-earth concentration in tested formulations, were related to electronic support effects, which affected those reactions in distinctive ways. Also, a diminished proportion of disintegration reactions was observed, decreased cracking being originated by the neutralization of surface acidic sites over alumina, due to the basic agent deposition. Considering those contradictory results, the influence of basic La as a carrier additive on HDS catalyst properties deserves further investigation.

Respecting HDS catalysts' active phases (as opposed to corresponding supports), Chevrel phases (MxMo6S8, where M could be La) have been synthesized and tested in HDS reaction schemes [15]. Different to conventional MoS2-based hydrodesulfurization catalysts, where Mo is in 4<sup>+</sup> oxidation state, in Chevrel phases Mo is in either 2<sup>+</sup> or 2.666+, depending on the type of second metal M. In general, HDS properties of Chevrel phases are not as good as those of MoS2-based catalysts, which explains why studies on those species (and particularly on those where La is the second metal) are very scarce.

In this work, lanthana modified-alumina at various rare-earth contents (1, 3, and 5 wt%) is used as CoMo-based HDS catalyst support. Textural, structural, and surface properties of carriers and corresponding impregnated materials, in an oxidic phase, were characterized through various techniques. Sulfided materials were tested in dibenzothiophene (most representative organo-S species present in middle distillates) hydrodesulfurization, in batch reactor at conditions close to those used in commercial-scale hydrotreaters aimed at diesel fuel production from oil-derived feedstocks. The activity trends found were correlated to physicochemical characteristics of studied catalysts. Some advances from this investigation have been recently reported [16].

#### **2. Results and Discussion**

#### *2.1. N2 Physisorption*

N2 physisorption isotherms (Figure S1) of alumina and La-modified supports (see Section 3.1. *Materials Synthesis* for samples nomenclature) were intermediate between types II and IV, according to IUPAC (International union of pure and applied chemistry) classification [17], suggesting mesoporous materials. All solids had type H1 hysteresis characterizing solids with porous networks of uniform size and shape [17], where capillary condensation started at ~P/Ps = 0.5.

From Table 1, La addition was reflected in slightly increased (~10%) surface area (SgBET) as to that of the pristine alumina carrier. More noticeable improvements were observed in both pore volume (Vp) and average pore diameter, that effect being more evident in materials at a higher rare earth content (in ALa3, 40% and 53% higher, respectively, as to sol-gel alumina). The texture of prepared binary matrices was adequate for carriers of catalysts applied in oil-derived middle distillate hydrotreatment aimed at diesel fuel production [18].

**Table 1.** Textural properties (as determined by low-temperature N2 physisorption) of ALa*z* supports and corresponding CoMo-impregnated oxidic materials.


<sup>a</sup> 4 × Vp × SgBET<sup>−</sup><sup>1</sup> [19].

Considering Co-Mo-P oxidic impregnated materials (see corresponding isotherms in Figure S2), CM/A had ~35% lower surface area as to that of the Al2O3 carrier, commensurate with non-porous oxidic phase loading (~32%), which suggests well-dispersed deposited phases. On the other hand, La-modified supports had significant textural losses after Co-Mo-P impregnation (~60% diminished Sg in CM/ALa1, as to that of the corresponding carrier), pointing to a partial collapse of the binary support matrix. Strandberg anions (phosphopentamolybdates), present in highly acidic impregnating solutions (see Section 3.1. *Materials synthesis*), could probably be decomposed under basic conditions in the interior of pores of rare earth modified materials (as indeed reported in Section 2.6. *Raman spectroscopy*), releasing PO4 <sup>3</sup><sup>−</sup> anions. Due to their strong affinity, lanthanum atoms in the solid matrices could strongly interact with those phosphate anions [20,21] being partially leached from the carriers, thus provoking the mentioned partial porous network destruction. Additionally, it has been reported [14] that, when impregnated through originally acidic solutions (as in our case), Mo could partially extract lanthanum atoms deposited over an alumina matrix, forming defined La molybdate domains. It seemed that particle growth could cause an enhanced average pore size (Table 1, as determined by Gurvich´s law [19]) of impregnated Co-Mo-P materials over La-doped carriers. Others [22] have also found that, during HDS catalysts preparation, impregnating conditions could be crucial in preserving textural

properties of corresponding supports, avoiding then their collapse. Expectably, that fact could be reflected in improved catalytic activity.

Pore size distributions (PSD) were determined through Barrett-Joyner-Halenda (BJH) methodology. In order to determine which PSD profile, obtained from an either adsorption or desorption data branch, better fits the actual ones, corresponding cumulative surface area values were compared to those obtained by the BET (Brunauer-Emett-Teller method, Table 1). Values from adsorption branch data (SgBJHa) were closest to the BET ones as to those obtained from desorption data (SgBJHd). Thus, PSD profiles from the former data were deemed the most suitable in describing the actual porous networks. PSD maxima of La-containing sol-gel carriers shifted to higher diameters as to that of non-modified alumina (~6.6 nm), ALa3 showing the maximum average value (~8.5 nm) (Figure S3a). Considering Co-Mo-P impregnated samples (Figure S3b), the maxima of CM/A and CM/ALa5 profiles shifted to larger diameters, as to those of corresponding supports (from 6.6 to 8.6 nm and from 6.4 to 8.1 nm, respectively). The amount of pores with diameters ≤ 5 nm, in the PSD of all carriers, strongly diminished after Co-Mo-P deposition.

#### *2.2. Thermal Analyses (Thermogravimetric and Di*ff*erential Thermal Analyses, TG and DTA)*

Thermogravimetric (TG) and differential thermal analyses (DTA) profiles of various carriers at different La contents were similar, with just small differences (not shown) in either weight losses (TG) or signal intensity (DTA). For example, corresponding plots for ALa5 are included in Figure 1. From room temperature to 550 ◦C, weight losses at a more or less constant rate were observed. The first part of the TGA curve (until approximately 140 ◦C) could be related to the evaporation of physisorbed water and alcohol from the sol-gel technique used during materials synthesis [23] (see Section 3.1. *Materials synthesis*). Further losses until 550 ◦C could be originated by the elimination of both nitrates (from La salt used) and organic remains from the Al alkoxide utilized. The endothermal signal centered at ~140 ◦C in the ALa5 DTA profile corresponded to aforementioned evaporation of water, alcohol, and structural hydroxyls. Meanwhile, the exothermal peak, at ~220 ◦C, could be provoked by NO3 − anion decomposition [24]. The intensity of this signal increased with La content in mixed formulations. The strong exothermal peak, centered at 277 ◦C, could be provoked by organic residua (from Al alkoxide) combustion [23]. Due to crystallization phenomena (to any Al2O3, LaO or La2O3 defined phases), no further signals were observed at more severe treatment conditions. In full agreement with our findings, it has been reported [25] that lanthanum addition during sol-gel alumina synthesis contributes to obtaining highly stable binary matrices delaying crystallization to the gamma phase, with defined La phases detected just after high-temperature (1000 ◦C) annealing.

#### *2.3. X-ray Difracction (XRD)*

Diffractograms of various supports at various compositions (not shown) showed no crystalline phases. The amorphousness of carriers suggested homogeneous materials, where La was both well-integrated and well-dispersed in alumina matrices. In Figure 2, diffractograms of the oxide at the highest La content and the corresponding Co-Mo-P impregnated material are shown. Both patterns did not show any diffraction peak, pointing to either well-dispersed deposited molybdates on the support surface or the presence of disordered amorphous Mo domains, undetectable by the used technique [26].

**Figure 1.** Thermal analysis (TG and DTA) profiles of the ALa5 dried carrier.

**Figure 2.** X-ray diffraction patterns of ALa5 support and corresponding P-doped CoMo impregnated oxidic material.

#### *2.4. Chemical Analysis (Scanning Electron Microscopy-Energy Dispersive Spectroscopy (SEM-EDS)*

The chemical composition of the supports and the various impregnated materials prepared were determined from EDS (energy dispersive spectroscopy, equipment attached to a SEM, scanning electron microscope, with back-scattered electrons detector). Corresponding micrographs of La-modified carriers and Co-Mo-P impregnated materials are included in Figures S4 and S5, respectively.

No important differences in sample morphology were observed at this stage, all solids being constituted by faceted particles. Regarding Co-Mo-P impregnated solids, morphology of corresponding carriers is preserved (note higher mixrograph magnification in former samples). In Figure 3, representative EDS profiles of samples ALa5 and CM/ALa5 are shown. From Table 2, actual La loading (wt%) in sol-gel carriers nicely corresponded to nominal values, suggesting homogeneous binary matrices.

**Figure 3.** EDS profiles of ALa5 support and corresponding oxidic P-doped CoMo impregnated material. (**a**) ALa5; (**b**) CM/ALa5.

**Table 2.** Chemical analysis (by SEM-EDS) of La-modified sol-gel alumina supports used.


Regarding Co-Mo-P impregnated solids, phosphorus content in CM/A and CM/ALa1 was very similar (Table S1), being slightly lower in the rest of samples. Additionally, Co and Mo contents in samples were alike in all impregnated samples.

#### *2.5. Surface Bascity (CO2-Adsorption Studied In IR Region, CO2-FTIR)*

The surface basic properties of the studied materials were estimated by integrating the area of peaks registered in the mid-infrared region, between 1800–1200 cm<sup>−</sup>1, observed after CO2 room-temperature adsorption [27], followed by thermo-desorption at various temperatures (Table 3). CO2 has been applied as a probe in determining the density and strength of surface basic sites of La-containing materials [28,29]. Different surface species could be originated by carbon dioxide adsorption, namely, unidentate and bidentate carbonates and bicarbonates [30] (Scheme S1). Carbonates, related to CO2 adsorbed on surface sites with intermediate and high basic strengths, respectively, requires of surface basic oxygen atoms [31]. Unidentate carbonates, characterized by symmetric and asymmetric O-C-O stretching at 1360–1400 cm−<sup>1</sup> and 1510–1560 cm−1, respectively, could be originated on isolated low-coordination surface O2<sup>−</sup> anions in corners or edges. On the other hand, bidentate carbonates that could be formed on Lewis acid-base pairs (M-O2<sup>−</sup> pair site, where M is La or Al cation, Scheme S1) are characterized by absorptions due to symmetric and asymmetric O-C-O stretching at 1320–1340 cm−<sup>1</sup> and 1610–1630 cm<sup>−</sup>1, respectively [31]. Finally, bicarbonates involve CO2 adsorption on low-strength basic surface hydroxyls [31], showing a C-OH bending mode at 1220 cm−<sup>1</sup> and symmetric and asymmetric O-C-O stretching vibrations at 1480 cm−<sup>1</sup> and 1650 cm−1, respectively (Scheme S1). We arbitrarily classified the strength of various types of sites according to the temperature at which they could retain CO2 (Td as weak (200 ◦C < Td), medium (200 ◦C > Td < 400 ◦C) and strong (Td > 400 ◦C).


**Table 3.** Relative surface basicity (CO2-FTIR, integrated area of peaks in the 1200–1800 cm<sup>−</sup><sup>1</sup> region) of supports used and corresponding Co-Mo-P oxidic impregnated materials. Td: Desorption temperature. ND: Non-determined.

The number of surface basic sites of alumina augmented, by La addition, being that especially evident in samples of higher rare-earth concentration (see, for instance, values from room temperature spectra of ALa3 and ALa5, Table 3). Cui et al. [13] found a linear relationship between the amount of adsorbed CO2 and La content (until ~20 wt%) in sulfided Mo/alumina samples prepared by rare-earth nitrate impregnation, although non-doped Mo/Al2O3 had some surface basicity. In addition, the strength of sites as a function of CO2, retained after a progressively increased desorption temperature, was clearly enhanced in aforementioned solids. Most of the sites on sol-gel alumina, ALa1 and ALa5, were weak, mainly forming bicarbonates. In the opposite, ALa3 contained the strongest surface basic sites (those related to unidentate carbonates formation) among the studied materials, preserving around the third part of the total number of sites retaining CO2 at room temperature and after treatment at 500 ◦C. Additionally, the relative amount of strong sites was three-fold to those over pristine alumina. In full agreement, it has been reported [32] that, on La-modified Al2O3 at contents close to monolayer (~5.12 at. La3<sup>+</sup> nm−<sup>2</sup> [33]) population of strong sites retaining CO2, was significantly enhanced.

For Co-Mo-P impregnated samples, the amount of adsorbed CO2 was clearly diminished as to those of corresponding carriers, with the sole exception of the alumina-supported solid. In this case, significantly decreased basicity (~50%) was registered just in the case of weak sites, which could be related to the well-known strong interaction between molybdate anions and the most basic hydroxyl groups on Al2O3 surface [34] (those related to bicarbonates formation during CO2 adsorption experiments). Additionally, CM/ALa1 lost around 62% of sites adsorbing CO2 at room temperature. The observed trend was even more marked regarding strong sites (~80% loss). Those facts could probably be related to strong deposited phases-La interaction. In the opposite, CM/ALa5 retained between 72%–83% of medium strength sites (mainly related to bidentate carbonates) of those on corresponding support. Due to strong Co [35], Mo [14] and P [20,21] interaction with La, it is probable that after impregnation the rare-earth could be lixiviated, to some extent, from the sol-gel matrix, with that fact explaining the partial textural collapse (Table 1). According to Cui et al. [13], during impregnation, Mo could preferably interact with La domains then with the alumina carrier. In addition, Ledford et al. [35] found that, in materials of La/Al ≥ 0.0075 atomic ratio (as in the case of our carriers with 3 and 5 wt% La), a Co-La phase could be formed that, being reflected in, lessened cobalt reducibility. The effects of lanthanum extraction could be more significant in the sample of the lowest rare earth content (CM/ALa1), rationalizing the important surface basicity lost. Conversely, the Co-Mo-P impregnated samples of enhanced La concentration had around 33% more sites retaining CO2 at 400 ◦C as to the alumina-supported solid, pointing out the retention of mid-strength basic sites.

#### *2.6. Raman Spectroscopy*

Al2O3 does not show any Raman band due to the low polarizability of light atoms and the ionic character of Al-O bonds [36]. Characteristic peaks of lanthana at 104, 191 [37], 310, 350, and 415 cm−<sup>1</sup> [38] were not evidenced, pointing to the absence of La2O3 definite domains in our oxidic carriers. Raman signals at 225, 359, 911 (shoulder), and 949 cm−<sup>1</sup> were observed in the CM/A spectrum (Figure 4a). The former two signals could be assigned to Mo-O-Mo and O-Mo-O vibrations [39]. Meanwhile, those at higher wavenumbers were related to Mo=O bands. They were ascribed to Anderson species Al(OH)6Mo6O18<sup>3</sup><sup>−</sup> [40], whose presence suggested partial support dissolution during contact with the acidic Co-Mo-P impregnating solution (pH ~2.5). Considering the Mo/P ratio in those solutions (~2.0) and their pH, H2P2Mo5O23<sup>4</sup><sup>−</sup> anions (two PO4 tetrahedrons surrounded by five interconnected octahedral MoO6 species) could, very probably, be preferentially present [40,41]. Although, due to their stability, those Strandberg heteropolyanions (HxP2Mo5O23(6−x)−) [42] could exist in various protonation states over a wide pH range [41], they could be decomposed to phosphates and molybdates owing to their strong interaction with both basic hydroxyls and Al3<sup>+</sup> coordinatively unsaturated (CUS) Lewis acid sites on alumina surfaces [43].

**Figure 4.** Raman spectra of oxidic Co-Mo-P impregnated materials. (**a**) CM/A and CM/ALa5; (**b**) CM/ALa1 and CM/ALa3.

In spite of some fluorescence interference that decreased sensitivity of the technique used, CM/ALa1 and CM/ALa3, Raman spectra (Figure 4b) evidenced that the signal related to M=O stretching vibrations progressively shifted to a lower wavelength as the rare earth concentration increased, being registered at 935 cm−<sup>1</sup> for CM/ALa5 (as compared to 949 cm−<sup>1</sup> observed in CM/A spectrum, Figure 4a). That fact, accompanied by the apparition of a signal at 321 cm−<sup>1</sup> (more evident in CM/ALa5), strongly suggested the co-existence of both octahedral and tetrahedral Mo species in La-containing samples. In full agreement, Payen et al. [14] reported that the Mo/ La ratio determined the type of oxomolybdenum entities adsorbed on lanthanum-modified alumina. Monomeric tetrahedral species (MoO4 <sup>2</sup>−) could be preferably formed by an interaction with La-OH groups, whereas polymeric octahedral molybdates could be deposited on Al2O3 domains. For materials at a high La concentration (Mo/La~1), just tetrahedral molybdates were observed [14]. In the same line, tetrahedral Mo species were mainly found [44] on basic magnesium-modified sol-gel alumina supports. It seemed that the high point of zero charge values of carriers with basic properties determined that behavior, as in contact with water those solids could produce high pH impregnating conditions where MoO4 <sup>2</sup><sup>−</sup> species are favored. Additionally, coordination state modifications of Mo6<sup>+</sup> species, from octahedral to tetrahedral (as determined by corresponding Raman shifts), by basic additive doping of alumina-supported unpromoted Mo (Mg, [5]) and CoMo formulations have been reported by others (K [45], Mg-Li [46]).

Definite domains (although not large enough to be detectable by X-ray diffraction) of lanthanum molybdate have also been registered [14] by impregnating La-modified alumina with acidic solutions prepared from ammonium heptamolybdate. In our case, contributions of characteristic bands at 915 and 940 cm−<sup>1</sup> could not be discarded, although, due to the low intensity of the registered peaks, they could remain masked. If existent, La2(Mo4)3-like domains could also be responsible for tetrahedral molybdenum presence [47]. According to the peaks' intensity in spectra, in Figure 4a,b, it seemed that samples CM/ALa1 and CM/ALa5 were the ones with the highest Mo dispersion (although with a high

proportion of tetrahedral species). Small signals, at 1119 cm<sup>−</sup>1, could be related to C-C bond stretching vibrations [48], probably originated by organic remains from either aluminum alkoxide used during carrier synthesis or due to acetate utilized during Co impregnation.

#### *2.7. UV-Vis Spectroscopy*

Being an insulator alumina does not absorb in ultraviolet-visible (UV-vis) region [49]. Absorptions in the 267–306 nm range (more notable for ALa1, Figure 5) could be related to charge transfer from <sup>O</sup>2<sup>−</sup> to framework La3<sup>+</sup> ions [50]. Regarding Co-Mo-P impregnated materials, the O−<sup>2</sup> <sup>→</sup>Mo6<sup>+</sup> charge transfer transition bands of tetrahedral (Mo(Th)<sup>6</sup><sup>+</sup>) and octahedral (Mo(Oh)<sup>6</sup>+) molybdenum species were centered at ~280 and ~314 nm, respectively [51] (Figure 6). That fact suggests that, although basic pH conditions that prevailed during Co-Mo-P impregnation of rare-earth-containing carriers strongly contributed to formation of isolated monomeric Mo(Th)<sup>6</sup><sup>+</sup> species (Figure 4a,b), Mo(Oh)<sup>6</sup><sup>+</sup> species co-existed with them, as already suggested by our Raman spectroscopy studies (Section 2.6). As proposed by Payen et al. [14], Mo(Th)<sup>6</sup><sup>+</sup> species, related to molybdenum in strong interaction with the support, could be related to La domains, whereas Mo(Oh)<sup>6</sup><sup>+</sup> ones could be deposited on alumina surfaces. As the former species were absent in the impregnating solutions (containing octahedral Mo6<sup>+</sup> heteropolyanions) their presence strongly pointed to the decomposition of the Strandberg species through their interaction with basic OH groups and CUS (Lewis sites) on the Al2O3 carrier surface [43]. As aforementioned, basic conditions prevalent in the pores of La-modified carriers could also play a major role in tetrahedral Mo6<sup>+</sup> species formation. The bathochromic shift of the Mo(Oh)<sup>6</sup><sup>+</sup> low-energy absorption edge strongly suggested an augmented oxomolybdates polymerization degree [52] in La-containing formulations as to that in CM/A following the order: CM/ALa1 > CM/ALa3 > CM/ALa5 > CM/A. The diminished surface area of lanthanum-modified CM materials (Table 1) could contribute to lowered octahedral Mo species dispersion. Indeed, a shoulder at ~370 nm, whose intensity increased according to the order CM/ALa1 > CM/ALa3 > CM/ALa5, pointed to the formation of MoO3 domains [53].

**Figure 5.** UV-vis spectra of La-modified alumina supports.

**Figure 6.** UV-vis spectra of Co-Mo-P impregnated oxidic formulations on La-modified Al2O3. Alumina-supported material included as reference.

The triplet band related to 4A2(F) <sup>→</sup> 4T1(P) transitions of tetrahedral cobalt Co(Td)2<sup>+</sup> at 540, 580, and 625 nm [51] was augmented in rare earth-modified solids, but with no definite trend regarding La concentration. Those cobalt species could be originated in lanthanum cobaltate-like entities. Interestingly, the enhanced proportion of tetrahedral Co2<sup>+</sup> in alumina-supported CoMo oxidic formulations has been observed in the past by progressively increasing the basic additives concentration (Mg and Li) [46], with augmented CoAl2O4 formation being invoked in that case. For La-containing samples, shoulders at ~495 and ~563 nm could be related to octahedral Co2<sup>+</sup> species cobalt (Co(Oh)<sup>2</sup>+) [54]. These signals progressively increased in the order CM/ALa1 > CM/ALa3 > CM/ALa5, being essentially absent in the alumina-supported formulation. Additionally, a small shoulder at 518 nm (probably from 4T1g(F) <sup>→</sup> 4T1g(P) electronic transition) suggested Co(Oh) <sup>2</sup><sup>+</sup> entities [55] in the CM/ALa1 spectrum.

#### *2.8. HDS Reaction Test*

La content in supports of sulfided Co-Mo-P (see 3.3. HDS Reaction Test) catalysts clearly affected their dibenzothiophene HDS activity (Figure 7). The formulation with the lowest rare-earth content had the highest pseudo first order kinetic constant value (~38% enhanced as to that of the non-doped Al2O3-supported material). However, significant diminution in HDS properties was observed as lanthanum concentration augmented in tested materials. Payen et al. [14] reported progressive diminution in thiophene HDS activity as La content in alumina- supported CoMo formulations increased. Very probably, enhanced proportions of refractory tetrahedral MoO4 <sup>2</sup><sup>−</sup> (see Section 2.6. *Raman spectroscopy*), as rare earth content in formulations augmented, could be primarily responsible for that behavior. Those species are characterized by being hardly sulfidable under the used conditions [14], then precluding MoS2 phase formation.

**Figure 7.** Pseudo first order kinetic constant (DBT HDS) of various tested catalysts. Batch reactor, *n*-hexadecane as solvent, P = 5.67 MPa, T = 320 ◦C, 107 rad s−<sup>1</sup> (1030 rpm) mixing speed.

The increased amount of MoO4 <sup>2</sup><sup>−</sup> species by the addition of basic agents in alumina-supported HDS catalysts has been reported in the past [14,56]. For instance, Malet et al. [56] found augmented proportions of hardly-reducible tetrahedral molybdenum entities in Mo/TiO2 solids doped with Na (through NaOH impregnation) at various concentrations. The decreased reducibility of tetrahedral Mo6<sup>+</sup> species on Al2O3-supported oxidic Mo and CoMo formulations modified by basic agents (Mg and Li) was confirmed by temperature-programmed reduction studies [5,46]. Regarding sulfided alumina-supported CoMo formulations modified by K addition (2.7 wt%), Maugé et al. [4] observed (by FTIR) decreased intensity in bands related to adsorbed CO as to that over the counterpart with a non-doped Al2O3 carrier. That could mean a lower proportion of sulfided phases, although the authors mainly attributed that to adsorption sites poisoning provoked by augmented sulfided phase electron density that could be reflected in diminished CO adsorption strength, under tested conditions (−173 ◦C). Similar to our case, those authors identified enhanced surface basicity of K-doped solids by formation of hydrogen carbonates and bidentate carbonates (weak and medium strength sites, respectively [30]), as studied by CO2 adsorption followed by FTIR). Decreased conversion in the 2-methylthiophene in *n*-heptane was observed in the case of K-doped CoMo/alumina catalysts when tested in a model FCC (fluid catalytic cracking) naphtha mixture also containing 2,3-dimethylbut-2-ene and orthoxylene. Regarding unpromoted sulfided alumina-supported Mo catalysts, Cui et al. [13] observed unaltered activity in thiophene HDS for materials at low La content (2 wt%), which was followed by decreased desulfurating properties in the solid modified with 5 wt% rare earth loading. Regarding the effect of basic additives on dibenzothiophene (DBT) HDS activity of alumina-supported materials, strongly decreased activity was observed [44] in the case of NiMo formulations with a MgO-modified (5 mol%) Al2O3 carrier. As the HDS activity of sulfided Mo has often been related to surface acidity [57], it seems reasonable to assume that diminished Brønsted or Lewis ones (or both) in formulations doped with basic additives could provoke limited organo-S species conversion.

The high activity in DBT HDS of our sulfided alumina-supported Co-Mo-P formulation modified with lanthanum at the lowest content could result from a combination of factors. Our UV-vis results (Figure 6) suggested that, in CM/ALa1, extant octahedral molybdenum species were at lower interaction with the support. Although, according to Raman characterization (Figure 4), amount of refractory tetrahedral Mo entities augmented in that formulation, as to those over the non-doped material with an Al2O3 carrier, the proportion of those monomeric species seemed to be not large enough to significantly affect proportion of sulfidable phases. In addition, the increased proportion of octahedral Co2<sup>+</sup> (as also observed by UV-vis, Figure 6) in the corresponding oxidic solid could be beneficial as that species is considered a precursor of cobalt that could efficiently form the highly active "CoMoS" phase during catalyst activation by sulfiding (see Section 3.3. *HDS Reaction Test*) [58]. A linear relationship between Co(Oh)<sup>2</sup><sup>+</sup> concentration in oxidic materials and HDS activity, in both gas-phase thiophene conversion and liquid phase real feedstock (vacuum gas oil and deasphalted oil) desulfurization, was found in that case. Interestingly, improved DBT HDS activity (as to that of corresponding non-modified catalysts) was observed in CoMo/alumina modified by basic agent addition, but just at a low doping agent content (1.2 wt% Mg or that solid with additional 4 wt% potassium loading) [59]. However, and similarly to our case, HDS activity significantly diminished by augmenting basic species concentration in corresponding sulfided catalysts. Additionally, Cao et al. [60] reported decreased DBT HDS activity on magnesium modified (20 wt% MgO) CoMo/Al2O3 materials.

Very interestingly, it has been reported [61] that activity in the SO2 conversion to sulfur, using coal gas as a reductant, of La-modified sulfided (at similar conditions to those used in the present work, Section 3.3. *HDS Reaction Test*) Co-Cu/γ-Al2O3 catalysts was significantly improved over formulations at low rare earth content, as to that of the reference non-doped formulation. Activity was augmented by adding 0.5 wt% rare earth, the maximum being found at 1 wt% La. However, diminished SO2 conversion was observed by augmenting dopant concentration (4 wt%). The authors attributed that to excessive La2O3 loading that could accumulate around crystals of active components, generating an undesired phase change by embedding corresponding particles. Those facts clearly show that La-modified sulfided catalysts at low rare-earth contents deserve further studies, not just in HDS reaction schemes. It is also worth mentioning that recently [62] we reported on the increased activity in naphthalene conversion of Pt (1 wt%) supported on alumina modified with a basic additive (Mg in this case) at low content, as compared to the non-doped formulation with the Al2O3 carrier. However, and similarly to what found during the present investigation, activity diminished in materials with a higher magnesium content (8 wt%).

Regarding selectivity to various products, DBT conversion could be carried out through direct desulfurization (DDS, to biphenyl, BP) and hydrogenation (HYD, firstly to hydrodibenzothiophenes, HDBT's) [63] reaction pathways (Scheme S2). At our reaction conditions (see Section 3.3. *HDS Reaction Test*), bicyclohexyl (BCH) saturation, from both aromatic rings, was observed just in trace amounts. Taking into account that BP hydrogenation to cyclohexylbenzene (CHB) could be strongly inhibited by DBT competitive adsorption under our HDS conditions, all CHB produced must have predominantly come from the HYD route through the sulfur removal of partially saturated HDBTs [64]. Selectivity to BP was clearly enhanced with La content in sulfided formulations at isoconversion (*x* ~17%) (Figure 8). Previous studies in HDS of DBT [65,66] and benzothiophene (BT) [67] strongly evidenced that an augmented DDS/HYD ratio could be related to more efficient MoS2 promotion by proper Co integration, then enhanced "CoMoS" phase formation. In this direction, Kaluža et al. [67] proposed that the amount of dihydrobenzothiophene (DHBT, formed from BT aromatic ring partial saturation) that could be then subsequently eliminated by C-S hydrogenolysis could be a measure of relative hydrogenation/hydrogenolysis selectivity. During their studies on CoMo sulfided formulations with various carriers, it was found that a very high promotion degree (17.9, over MgO-supported catalyst) was accompanied by a significantly decreased DHBT formation.

In the case of CoMo materials supported on alumina modified by basic agents, significantly augmented DDS/HYD selectivity in DBT HDS has been reported [46] in the past. For instance, by adding magnesia (0.05 mol ratio) to an Al2O3 carrier CoMo catalysts producing strongly increased (100% enhancement at 30% conversion) DDS/HYD ratio were obtained. Even more, selectivity to biphenyl was further progressively augmented by additional lithium doping of the aforementioned

formulation. Others [44] have also found decreased DBT transformation through the HYD route over sulfided NiMo catalysts supported on alumina modified by 5 mol% magnesia.

**Figure 8.** Selectivity to various products (DBT HDS, ~17% conversion), of various tested catalysts. HDBT: hydrodibenzothiophenes; BP: biphenyl; CHB: cyclohexylbenzene; BCH: bicyclohexyl. Batch reactor, *n*-hexadecane as solvent, P = 5.67 MPa, T = 320 ◦C, 107 rad s−<sup>1</sup> (1030 rpm) mixing speed.

Following this line of reasoning, the enhanced DDS/HYD ratio, in our case, could be originated by well-promoted (by Co) MoS2 slab edges, whose number could be restricted due to limited Mo6<sup>+</sup> sulfidation provoked by extant tetrahedral species. That could justify low DBT HDS activity of formulations of higher La content where cobalt could yet be properly integrated to molybdenum sulfide edges. In those catalysts, however, a certain proportion of the sulfided promoter could be as partially segregated CoxSy domains (of marginal HDS activity [66]) due to the reduced number of MoS2 layer edges, where they could be properly incorporated to produce the highly active "CoMoS phase".

Well-dispersed molybdenum sulfide phases (thus composed of short slabs) could be very efficient in aromatic ring hydrogenation [68]. HYD sites have been related [46] to corner sites in Co-MoS2 slabs, whose proportion decreases as slab length grows. Those longer molybdenum sulfide layers could be originated by oxidic molybdates of lower dispersion, similarly to that proposed by Halachev et al. [69] in the case of alumina-supported P-modified NiW materials where the ratio W=O/W-O-W, considered as a measure of dispersion of oxidic tungsten species, was related to naphthalene hydrogenation activity of corresponding sulfided formulations. Solids with a higher proportion of terminal M=O bonds resulted (after sulfiding) in catalysts with improved dispersion, and, thus, with enhanced saturation properties. In the opposite, materials with diminished dispersion resulted in decreased hydrogenating capabilities by generating longer MoS2 with an increased edge sites/corner sites ratio. In this regard, it should be mentioned that, in our case, La-addition resulted in materials with lower oxidic Mo6<sup>+</sup> phases dispersion (Figure 6). Interestingly, sulfided CoMo/alumina catalysts with limited dispersion (longer slabs and augmented stacking) have also shown [70] enhanced HDS of thiophenic species, accompanied by decreased olefins hydrogenation in the selective hydrotreating of FCC naphtha focused on lowering octane number losses.

Bataille et al. [65] proposed that the main role of the promoter in DBT HDS is to enhance the reaction rate through the DDS pathway, or the C-S bond breakage activity in general. The involved mechanism could consist of attacking a hydrogen atom, in a β position relative to the S atom in a DBT molecule, by a basic sulfur anion. Primarily, both reaction routes (DDS and HYD) require a dihydrodibenzohiophene intermediate [65]. Cleaving the C-S bond to produce two phenyl rings species (BP) could be carried out by hydrogenating one of the double bonds (in any of the aromatic rings) in the sulfur atom vicinity, resulting in the dihydrogenated DBT. Then, the C-S bond in the partially saturated ring could be opened by an elimination process. The second C-S bond cleavage, leading to BP, could possibly occur through a similar mechanism [65].

As the promoter (either Ni or Co) increases electron density of basic S anions, they could favor C-S bonds cleavage. The possibility of electron-rich alumina-lanthana supports contributing to increased sulfur anion basicity (then enhancing the DDS reaction pathway), as recently reported in the case of K-doped MoS2 phases applied in methanethiol synthesis from a H2S-CH4 mixture [6], remains a question that deserves to be answered. Mechanistic studies on DBT HDS over La-modified alumina supports are clearly needed, as information on reaction routes and kinetic parameters could provide valuable parameters that could be useful in investigations focused on applying that kind of catalyst under more realistic conditions in the presence of oil-derived middle distillates [71].

Regarding CoMo alumina formulations modified by La at various loadings, Payen et al. [14] found a similar promotion degree in the thiophene HDS as to that observed over a non-modified solid with an Al2O3 carrier. Different trends, as to those reported by others, for corresponding catalysts doped with basic agents (in BT HDS, for instance [67]) could be originated in distinctive rate-limiting steps in tested reactions. Unfortunately, no data on selectivity through DDS and HYD reaction pathways were provided in that case [14].

From the results of the present investigation we consider that sulfided CoMo formulations, supported on alumina doped with La at low concentration (up to 1 wt%), clearly deserve deeper studies. These ongoing investigations will be the subject of further reports.

Finally, the development of HDS catalysts with enhanced activity and increased DDS/HYD selectivity ratio (as our CM/ALa1 formulation) results is promising considering their limited consumption of expensive hydrogen, mainly in the case of H2-constrained refineries.

#### **3. Experimental**

#### *3.1. Materials Synthesis*

Al2O3-La2O3 mixed oxides at three compositions (1, 3, and 5 wt%, ALa1, ALa3, and ALa5, respectively) were prepared by a sol-gel method. First, aluminum alkoxide was dissolved at 80 ◦C in corresponding alcohol (ROH/alkoxide = 60), under stirring and refluxing being kept during 2 h. Then, water and HNO3 (3 N) hydrolysis catalysts [23] were added (H2O/alkoxide= 1, HNO3/alkoxide = 0.03, respectively), the mixture being kept under stirring and refluxing during approximately 1 more h until sol formation. Pertinent amounts of La(NO3)3·6H2O in an alcoholic solution were added to the prepared alumina sol. Additional amounts of distillated water were supplied until total aluminum alkoxide hydrolysis. The mixtures were kept under stirring and reflux during 1 more h at room temperature until gelification. The obtained gels were aged, dried (~100 ◦C), and further calcined (550 ◦C, 4 h).

Over annealed mixed oxides Mo and Co were deposited (at 2.8 atoms Mo nm−<sup>2</sup> and (Co/(Co+Mo)) = 0.3, respectively). Pore-filling simultaneous impregnation was carried out over previously dried supports (120 ◦C, 2 h) with an acidic solution (pH ~2.5) prepared from digestion (at ~80 ◦C, in water and under vigorous stirring, 4 h) of MoO3 99.5 wt% (PQM) in the presence of H3PO4 85.3 wt% (Tecsiquim, Mexico City, Mexico). H2O in excess (typically 250 ml of starting solution to impregnate 5 g of carrier) was used to accelerate molybdenum salt dissolution and digestion. A yellow transparent solution was obtained after hydrolysis. (CH3COO)2Ni·4H2O (Sigma-Aldrich, Darmstadt, Germany) was then added, with the stirring and processing temperature being maintained for 2 h. A P2O5/(NiO+MoO3) = 0.2 mass

ratio was fulfilled [72]. A transparent dark wine solution was thus finally obtained. The solution volume was then reduced by evaporation until reaching the suitable one for pore-filling impregnation of the given mass of support. The described one-pot impregnation method was chosen as it constitutes a readily scalable methodology during catalyst preparation at the commercial scale. After impregnation materials were left aging overnight at room temperature for the diffusion of the impregnated species into the carrier porous network. Then, the materials were dried (2 h at 120 ◦C) and further calcined (400 ◦C for 5 h). Impregnated solids were identified by using the CM/ALa*z* key, where "*z*" represents La loading in corresponding supports.

#### *3.2. Materials Characterization*

Textural properties of prepared materials were determined by N2 physisorption (at −198 ◦C, nitrogen saturation temperature at Mexico City barometric pressure) in a Micromeritics (Norcross, GA, USA) ASAP 2000 apparatus after ultra-high vacuum (133.32 <sup>×</sup> <sup>10</sup>−<sup>5</sup> Pa) degassing at 300 ◦C for 2 h to eliminate adsorbed molecules. The surface area (BET, Brunauer-Emett-Teller method), pore volume, and average pore size of the materials (BJH, Barrett-Joyner-Halenda method) were determined from the corresponding isotherms.

Thermal analyses (thermogravimetric and differential thermal analysis (TG and DTA, respectively) of dried (non-calcined) samples were carried out with a Netzsch Thermische Analyze (Burlington, MA, USA) STA 409 EP apparatus, under a static air atmosphere. The crystallographic order of the studied samples was determined by powder X-ray diffraction (XRD, Siemens (Munich, Germany) D-500 Kristalloflex, copper anode, CuK<sup>α</sup> radiation, λ = 0.15406 nm, 35 kV, 25 mA), in the 5–80◦ 2θ range.

Scanning electron microscopy pictures and chemical analysis by energy dispersive spectroscopy (EDS) of the prepared samples were performed in an environmental scanning electron microscope XL30 with an attached energy dispersive X-ray spectroscope (EDAX, Berwyn, PA, USA).

The surface basicity of alumina, alumina-lanthana supports, and the corresponding impregnated materials were characterized by CO2 adsorption studied in the mid-infrared region. As an acid probe, carbon dioxide interacts with basic catalysts' surfaces, forming bidentate carbonates, unidentate carbonates, and bicarbonates (Scheme S1, [30]), depending on surface basic site strength. Previously, to corresponding studies, all materials (carefully weighed) were submitted to thermal treatments (500 ◦C), under an inert atmosphere, to eliminate adsorbed species. Samples were cooled down to room temperature with CO2 being then fed to the system (at room temperature). Once samples were saturated, the flow of probe gas was suspended, with excess CO2 being flushed by an inert gas stream. The corresponding spectra were acquired by using a Nicolet (Waltham, MA, USA) Magna 560 FTIR spectrometer at 4 cm−<sup>1</sup> resolution (50 scans) in diffuse reflectance mode. The analyzed samples were treated at various progressively higher temperatures (25–500 ◦C) and detected peaks in the 1800–1200 cm−<sup>1</sup> wavenumbers infrared range (corresponding to formed surface bicarbonates and carbonates) were then registered and integrated.

The Raman spectra of various impregnated samples were obtained using a Jobin Yvon Horiba (Northampton, UK) T64000 spectrometer, equipped with a CCD camera detector. As an excitation source, the 514.5 nm line of a Spectra Physics 2018 Argon/Krypton ion laser system was focused through an Olympus (Tokyo, Japan) BX41 microscope equipped with a 100× magnification objective. Laser power on the sample never exceeded 5 mW to avoid thermal effects on the samples studied.

#### *3.3. HDS Reaction Test*

Prepared catalysts were tested in sulfided form (see below) by using a batch reactor, as the main goal was to determine the effect of La content on the HDS properties of otherwise conventional P-modified CoMo/Al2O3 catalysts. That approach allowed a simple, rapid comparison among the catalytic activity of various studied materials [44,46], although commercial-scale HDS applications in refineries are carried out in fixed-bed plug-flow reactors. Sulfided catalysts were obtained by submitting impregnated precursors to treatment at 400 ◦C (heating rate 6 ◦C min<sup>−</sup>1), under H2/H2S (Praxair, Mexico City, Mexico) flow at 50/6 (ml min−1)/(ml min−1) during 2 h. The HDS activity of sulfided catalysts was studied in a tri-phasic slurry batch reactor (Parr 4575, Moline, IL, USA). The reaction mixture was prepared by dissolving ~0.3 g of dibenzothiophene (representing S-bearing species present in oil-derived middle distillates) in 100 cm<sup>3</sup> of *n*-hexadecane (98 mass % and 99+ mass %, respectively, both from Aldrich, Darmstadt, Germany). *N*-hexadecane (cetane) well-represented the hydrocarbons extant in aforementioned real feedstocks from which diesel fuel is produced. Approximately ~0.2 g of sieved sulfided catalyst (80–100 Tyler mesh, ~0.165 mm average particle diameter) were also added, avoiding contact with atmospheric air, precluding sulfates formation. The HDS test reactions were carried out at T = 320 ◦C, P = 5.67 MPa, and 1000 rpm (~105 rad s−1, mixing speed). Catalysts' particle size and operating conditions were carefully chosen to avoid internal/external diffusional limitations [73]. Liquid samples were taken from the reactor, periodically, then analyzed by gas chromatography (Agilent (Santa Clara, CA, USA) 6890N with flame ionization detector and Econo-Cap-5 capillary column length: 30 m, 0.53 mm diameter, and film thickness of 1.2 μm, from Alltech (Nicholasville, KY, USA). The HDS kinetic constants were calculated assuming pseudo-first order model kinetics referred to DBT concentration and zero order with respect to excess H2, as follows:

$$k = \frac{-\ln(1-x)}{t\_r} \tag{1}$$

where *x* is DBT conversion and *tr* refers to reaction time. The *k* values for various catalysts were normalized by considering the reaction volume and the mass of the catalyst used (*k* expressed in m<sup>3</sup> kgcat−<sup>1</sup> s−1). It should be mentioned that H2S, a byproduct from S-C bond scission reactions, contributed in maintaning tested catalysts in sulfided form and then avoiding their reduction to corresponding metallic phases, characterized by being prone to sintering [74].

#### **4. Conclusions**

Alumina-lantana (1, 3, or 5 wt% La) mixed oxides of suitable texture, to be applied as supports of catalysts for hydrotreatment of oil-derived middle-distillates, were prepared by a sol-gel method. In general, the amount and strength of surface basic sites increased with rare-earth content in binary carriers. La at low content (1 wt%) was beneficial in sulfided P-doped CoMo formulations where enhanced dibenzothiophene hydrodesulfurization (HDS) activity (~38%), as to that over the Al2O3-supported counterpart, was observed. However, increased La concentration in supports was detrimental on HDS properties. Hardly sulfidable tetrahedral Mo species, originated during impregnation at basic conditions, in pores of La-modified carriers seemed to dictate that behavior. Augmented rare earth concentrations in mixed supports of sulfided CoMo catalysts favored dibenzothiophene conversion through the direct desulfurization (DDS) reaction pathway to biphenyl, over the hydrogenation (HYD) one. That fact, presumably originated in enhanced MoS2 promotion by efficient Co integration, seems to be promising in the development of alumina-supported HDT formulations at low La content of limited expensive hydrogen consumption.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4344/9/4/359/s1, Figure S1: N2 adsorption isotherms (at −198 ◦C) of alumina support (A) and La-modified carriers at various rare earth contents (ALa*z*). Closed symbols: adsorption branch; open symbols: desorption branch. Figure S2: N2 physisorption isotherms (at −198 ◦C) of oxidic P-doped CoMo materials impregnated over sol-gel alumina (A) and corresponding La-modified (ALa*z*) carriers. Closed symbols: adsorption branch; open symbols: desorption branch. Figure S3: Pore size distributions of various prepared supports (a) and oxidic Co-Mo-P impregnated materials (b), as calculated by Barrett-Joyner-Halenda methodology with data from adsorption branch of corresponding N2 adsorption isotherms. Figure S4: SEM micrographs of La-modified supports. At 1000× magnification, back-scattered electrons detector. (a) ALa1; (b) ALa3; (c) ALa5. Figure S5: SEM micrographs of Co-Mo-P impregnated oxidic samples on alumina and La-modified supports. At 4000× magnification, back-scattered electrons detector. (a) CM/A; (b) CM/ALa1 (c) CM/ALa3; (d) CM/ALa5. Table S1: SEM-EDS chemical analysis of Co-Mo-P impregnated oxidic materials prepared. Scheme S1: Absorption bands in the infrared region of CO2 species adsorbed on basic sites [from ref. S1]. Scheme S2: Dibenzothiophene HDS reaction network over sulfided CoMo/Al2O3 [from ref. S2]. HDBT´s: hydrodibenzothiophenes; BP: biphenyl; CHB: cyclohexylbenzene; BCH: bicyclohexyl.

**Author Contributions:** Funding acquisition, J.E.; Investigation, J.E., M.C.B., J.S.V., D.A.S.-C., V.S., J.E.T., and B.A.R.F.; Project administration, J.E.; Validation, J.E. and M.C.B.; Writing, review and editing, J.E. and M.C.B.

**Funding:** 117086 SENER-CONACYT-Hidrocarburos grant and Y.00105 Project from IMP.

**Acknowledgments:** J. Escobar acknowledges financial support from IMP (Y.00105) and SENER-CONACYT-Hidrocarburos (115086) fund.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*

### **Assessment of Ag Nanoparticles Interaction over Low-Cost Mesoporous Silica in Deep Desulfurization of Diesel**

**Rafael V. Sales 1, Heloise O. M. A. Moura 1, Anne B. F. Câmara 1, Enrique Rodríguez-Castellón <sup>2</sup> , José A. B. Silva 1, Sibele B. C. Pergher 1, Leila M. A. Campos 3, Maritza M. Urbina 4, Tatiana C. Bicudo <sup>1</sup> and Luciene S. de Carvalho 1,\***


Received: 5 July 2019; Accepted: 18 July 2019; Published: 30 July 2019

**Abstract:** Chemical interactions between metal particles (Ag or Ni) dispersed in a low-cost MCM-41<sup>M</sup> produced from beach sand amorphous silica and sulfur compounds were evaluated in the deep adsorptive desulfurization process of real diesel fuel. N2 adsorption-desorption isotherms, X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), high-resolution transmission electron microscopy (HRTEM) and scanning transmission electron microscopy coupled to energy-dispersive X-ray spectroscopy (STEM-EDX) were used for characterizing the adsorbents. HRTEM and XPS confirmed the high dispersion of Ag nanoparticles on the MCM-41 surface, and its chemical interaction with support and sulfur compounds by diverse mechanisms such as π-complexation and oxidation. Thermodynamic tests indicated that the adsorption of sulfur compounds over Ag(I)/MCM-41M is an endothermic process under the studied conditions. The magnitude of ΔH◦ (42.1 kJ/mol) indicates that chemisorptive mechanisms govern the sulfur removal. The best fit of kinetic and equilibrium data to pseudo-second order (R2 > 0.99) and Langmuir models (R<sup>2</sup> > 0.98), respectively, along with the results for intraparticle diffusion and Boyd's film-diffusion kinetic models, suggest that the chemisorptive interaction between organosulfur compounds and Ag nanosites controls sulfur adsorption, as seen in the XPS results. Its adsorption capacity (*q*m = 31.25 mgS/g) was 10 times higher than that obtained for pure MCM-41<sup>M</sup> and double the *q*<sup>m</sup> for the Ag(I)/MCM-41C adsorbent from commercial silica. Saturated adsorbents presented a satisfactory regeneration rate after a total of five sulfur adsorption cycles.

**Keywords:** desulfurization; chemisorption; MPI silica; MCM-41; Ag nanoparticles; XPS assessment

#### **1. Introduction**

The demand for clean energy has attracted great attention in recent years and has generated an increasing focus on sulfur compounds removal from fuels [1]. The combustion of these compounds results in liberating sulfur oxides (SOx), which poison catalytic converters and are corrosive to fuel cell electrodes [2], in addition to polluting the environment; therefore, governments worldwide have adopted standards to decrease sulfur concentration to 10 ppm for diesel fuel [3]. The current technique for sulfur removal in oil refineries is the highly expensive hydrodesulfurization (HDS), performed at hard conditions (>573 K, >4 MPa) using sophisticated Co-Mo/Al2O3 or Ni-Mo/Al2O3 catalysts. However, some recalcitrant organosulfur such as thiophene derivatives are not removed by this method [4].

Methods based on membrane separation [5], catalytic oxidation [3], biological desulfurization [6], and adsorption [1,7–11] have been developed for desulfurizing fuels. Among these techniques, adsorptive desulfurization stands out due to its simplicity, efficiency and low-cost according to the solid chosen as the adsorbent. Activated carbon [1], MOFs [9], KIT-6 [10], SBA-15 [8], and MCM-41 [7,11] are currently being investigated for sulfur adsorption from liquid fuels, among other materials.

These molecular sieves, especially MCM-41, are extensively applied as supports for adsorbents and heterogeneous catalysts due to their high surface area, ordered structure, and uniform pore size [12]. In general, heterogeneous micrometric size catalysts present low catalytic activity due to slow diffusion of reagents [13]. Therefore, active metal nanoparticles (NPs), which have high catalytic activity due to their high surface-volume ratio and chemical reactivity, have been incorporated into solid supports for increasing their efficiency, recyclability and stability, thus providing more sustainable processes [14–16]. When unsupported, the high surface energy of nanocatalysts increases system instability and leads to aggregation, thus resulting in the loss of its catalytic properties. The use of molecular sieve supports is fundamental to prevent the undesirable aggregation and loss of nanocatalysts [17,18].

Moreover, MCM-41 has gained attention for its flexibility regarding the silica feedstock for synthesis. Recently, it has been produced from greener sources of silica such as coal fly ash (CFA) [12], rice husk ash [19], iron ore residues [20], wheat stem ash (WSA) [21], and MPI silica from beach sand [22,23], used in this study. This material has been functionalized by adding active sites to its surface for promoting new chemical interactions in the support-site-adsorbate system and improving the adsorptive treatment [24]. Researchers have observed that adsorbents functionalized with transition metals are capable of capturing aromatic sulfur compounds refractory to HDS processes via π-complexation [25,26]. Furthermore, Ag species dispersed onto metal and semi-metal oxide adsorbents generate materials described as high activity catalysts for oxidation of several toxic compounds such as formaldehyde [27], benzene [28], toluene [29] and carbon monoxide [24], among others.

The Ag(0) sites present high reactivity both by adsorbing and activating atmospheric O2 at ambient conditions (300 K, 1 atm) [30] and by increasing the mobility of lattice oxygens from the adsorbent structure. This increase in mobility can occur by bridging bonds between Ag(0) and the metal oxide molecules [28], but it is also reported by covalent bonding in the case of a more oxidized state of silver as Ag(I), and in a cation exchange of hydrolyzed silica silanol groups (Si-O-H) [31]. Thus, an instrumental technique such as X-ray photoemission spectroscopy (XPS) is required to identify the chemical species in the sample and to indicate what kind of interactions could be involved in impregnation and chemisorptions processes. This analysis is able to capture chemical changes in the adsorbent surface after metal deposition and adsorption by recording displacements in binding energies of the atoms and ions involved in the mechanisms [32].

This paper proposes a low-cost method for mitigating sulfur content in diesel fuel based on the adsorption of organosulfur compounds over Ag and Ni impregnated MCM-41 supports from beach sand silica, in a complementary process to the hydrodesulfurzation (HDS) applied in refineries with a greener and effective approach. Furthermore, high-efficiency characterization techniques such as X-ray photoelectron spectroscopy (XPS) and high-resolution transmission electron microscopy (HRTEM) were used to enable the more accurate analysis of the different species and interaction mechanisms in the metal-support-organosulfur system, and were associated with a deep evaluation of the kinetic and thermodynamic features.

#### **2. Results and Discussion**

#### *2.1. Preliminary Sulfur Adsorption Results*

The results of sulfur adsorption in columns for AgNO3/MCM-41 produced from both MPI and commercial silicas interacted more strongly with organosulfur compounds and reached higher desulfurization results. The adsorbent generated from MPI silica presented lower performance in comparison to the commercial material; however, its efficiency is significant and it stands out for being extracted from a renewable and low-cost source. There was no relevant variation in adsorption by changing metal concentration from 2% to 8%. Taking into consideration the performance and cost-benefits, the concentration of 2% salts in MCM-41<sup>M</sup> was selected for the next steps of characterization and adsorption studies. The sulfur adsorption results for all materials applied in this research are depicted in Figure S1 of the Supplementary Materials. The XRD patterns for the adsorbents with 8% metal (Ni or Ag) in salt and oxide forms are presented in Figure S2 of the Supplementary Materials.

#### *2.2. Characterization of the Adsorbents*

Low-angle XRD patterns of the samples are shown in Figure 1a. The presence of diffraction peaks indexed by (100), (110) and (200) confirms the well-ordered structure with hexagonal *p6mm* symmetry, indicating a relative perfect structure of MCM-41 samples [33]. Less ordered materials were obtained after impregnation; however, the deposition of transition metals in the adsorbent via wet impregnation did not compromise the ordered porous structure of MCM-41M.

**Figure 1.** (**a**) Low-angle and (**b**) wide-angle X-ray diffractograms of the adsorbents.

Wide-angle XRD patterns of all samples (Figure 1b) exhibited broad diffraction peaks centered at 22.8◦, originated from amorphous silica (MPI silica). AgNO3/MCM-41<sup>M</sup> and Ni(NO3)2/MCM-41M adsorbents have not shown any reflection referent to deposited species, supporting the idea that they are highly dispersed and have a high interaction with the mesoporous support [34]. Refraction peaks corresponding to the cubic phase of NiO at 2θ of 37.16, 43.31, 62.88 and 79.31 (JCPDS 01-073-1523) and the face-centered-cubic (FCC) phase from metallic Ag at 2θ of 38.16, 44.30, 64.30 and 77.44 (JCPDS 04-0783) were detected in NiO/MCM-41M and Ag2O/MCM-41M, respectively (Figure 1b).

HRTEM micrographs for pure MCM-41<sup>M</sup> (Figure 2a,b) confirm the hexagonal high-organized pore matrix, typical of this material, as seen in XRD patterns [35]. The AgNO3/MCM-41<sup>M</sup> sample was analyzed by STEM-EDX mapping images and HRTEM. The STEM-EDX mapping profile for deposited species (Figure 2c–e) shows highly uniform distribution of silver nanoparticles (red spots), with some agglomerations, and silica particles were also observed in green. According to a HRTEM micrograph (Figure 2f), the MCM-41<sup>M</sup> mesostructure was preserved after its modification with Ag<sup>+</sup> ions.

**Figure 2.** (**a**,**b**) HRTEM micrograph for pure MCM-41M; (**c**) STEM, (**d**) EDX, (**e**) STEM-EDX and (**f**) HRTEM analysis for AgNO3/MCM-41M.

The N2 adsorption-desorption isotherms and pore size distribution are depicted in Figure 3. The isotherms for all the samples are type IV according to the IUPAC classification, which is characteristic of MCM-41 [36]. As depicted in Figure 3a, the N2 adsorption–desorption isotherms for all the samples were correspondent to type IV according to the IUPAC classification, which is associated to mesoporous materials with H3-type hysteresis loop due to capillary condensation [34]. The adsorption-desorption in intermediary relative pressures (0.3 < P/P0 < 0.4) correspond to N2 capillary condensation of uniform mesopores. The metal incorporation to the mesoporous sieve has not affected the isotherms profile. A slight inflection for the Ag2O/MCM-41<sup>M</sup> sample can be noticed, which suggests a less-organized mesoporous structure [11], as observed in the XRD data shown in Figure 1a. All samples presented an increase in adsorbed volume due to the macropore filling resulting from interparticle spaces and untransformed amorphous silica in the relative pressure range 0.90–0.99 [35].

**Figure 3.** (**a**) N2 adsorption-desorption isotherms at 77 K (**b**) and the corresponding BJH adsorption pore size distributions for the adsorbents.

The pore size distribution (Figure 3b) estimated by the BJH method indicates that the materials have uniform mesoporosity with a maximum distribution in the range of 23.54–24.82 Å. An exception was the Ag2O/MCM-41<sup>M</sup> sample, which exhibited a smaller maximum pore size centered at about 20.20 Å and heterogeneous distribution [4]. In this case, Ag2O species are possibly deposited inside the pores, while the species were preferentially deposited outside the pore system for the other materials.

Textural parameters of the adsorbents are listed in Table 1 and presented a reduction in specific area (SBET) and pore volume (VP) compared to the initial silica material, indicating that the deposited particles caused the occlusion of some pores, reducing the amount of adsorbed N2. In addition, the embedded species are not porous, which reduces porosity and consequently the specific area of the materials [37].


**Table 1.** Textural properties of the studied samples.

<sup>1</sup> Specific surface area calculated by the BET method. <sup>2</sup> Total pore volume recorded at P/P0 = 0.993.

XPS analysis was employed to obtain information about the surface composition, species identification and chemical state of the transition metal over MCM-41<sup>M</sup> surface. The XPS spectrum of the pure MCM-41M sample is shown in Figure 4. The binding energy (BE) of the Si 2p orbital in all the samples was about 103.3 eV, which is characteristic of mesoporous silicates such as MCM-41. Additionally, a single peak centered at 532.7 eV would be assigned to the O 1 s photoemission of oxygen atoms from the siliceous support [38].

**Figure 4.** XPS spectra for pure MCM-41M.

For AgNO3/MCM-41<sup>M</sup> (Figure 5a), broad peaks in between the characteristic BE for Ag(I) species (367.5 and 373.8 eV) and Ag (0) nanoparticles (369.7 and 375.9 eV) are observed for the Ag 3d5/<sup>2</sup> and 3d3/<sup>2</sup> orbitals at 368.1 and 374.1 eV, respectively [39]. This phenomenon suggests the interaction of impregnated AgNO3 Ag(I) ions with other atoms such as adsorbed atmospheric oxygen (Ag-Oads), lattice oxygen (Ag-Olat) and silanol groups (Ag-O-Si) of MCM-41M, now presenting a BE state in between the ionic and metallic forms (Agδ+, 0 < δ < 1) [31,40].

**Figure 5.** XPS spectra for (**a**) AgNO3/MCM-41M and (**b**) Ag2O/MCM-41<sup>M</sup> before and after desulfurization.

Smaller peaks at 369.0 and 375.0 eV are assigned to the typical BE region for Ag(0) nanoparticles highly dispersed in the support [30,41]. The presence of the metallic electronic state of silver may be related to the photolysis of the AgNO3 molecule in the presence of light; furthermore, it is known in the literature that some oxygenated metal-adsorbents can undergo autoreduction processes in the presence of lattice oxygen and water [42]. Ag(0) can also interact with oxygen (Ag-Oads and Ag-Olat). All these oxygenated species (Ag-O) are highly active for oxidation of aromatic pollutants [34] and even for weak π-complexation interactions with aromatic organosulfurs, as well as the non-bonding Ag(0) nanodomains [36].

After sulfur adsorption was performed with model diesel, the displacement to higher BE values indicates more positive electronic density of Ag species [27]. The increase in BEs of oxygenated Ag species (368.4 and 374.5 eV) can be assigned to oxygen loss after oxidation reactions, while the broad peaks which emerged at 369.4 and 375.6 eV are ascribed to the formation of an Ag-S-R bridging configuration via π-complexation (S-R = DBT) performed by Ag(0) and also by active Ag-O species [26,43]. In this mechanism, Ag(0) ([Kr]4d105s1) and Ag(I) ([Kr]4d105s0) species are able to form bonds with their empty or semi-filled *s* orbitals and their *d* orbitals can retrograde electronic density to the π anti-ligands (π\*) of aromatic organosulfur rings [25].

For the sample that was calcined in an oxidizing atmosphere, the XPS spectra (Figure 5b) show broad peaks at 368.5 and 374.6 eV, which can be ascribed to the production of stable Ag2O species that can emerge in this BE region [44]; since the spectral resolution for all XPS analysis was 0.8 eV, significant peak separation was not possible. Highly dispersed Ag nanodomains were also observed at 369.4 and 375.9 eV. The spectra recorded after this material was used for desulfurization shows that the Ag2O/Ag-O peak BE remained almost unchanged; however, the increase in Ag(0) BE (369.8 and 375.4 eV) demonstrates the same phenomenon observed for the non-calcined sample, where metallic and active silver species were complexed by DBT molecules.

The atomic concentrations in the adsorbents surface were measured by the XPS technique and are summarized in Table 2. Ni, C, O and Si contents in Ni(NO3)2/MCM-41M and NiO/MCM-41M adsorbents were similar; however, the Ag-impregnated sample after calcination presented some significant variations in the atomic concentrations, with a increase in the C atomic percentage. As the penetration depth of XPS radiation is about 10 nm [44], X-rays are only able to reach the atoms near the surface, and the higher amount of carbon in this region promotes a decrease in other atoms' concentrations, as seen for silver and oxygen. The AgNO3/MCM-41<sup>M</sup> adsorbent presented higher sulfur concentrations after adsorption tests among all samples.


**Table 2.** Atomic concentrations (%) in the adsorbents surface and Si/metal molar ratio.

#### *2.3. Kinetic Study*

Sulfur adsorption rate performed with real diesel on AgNO3/MCM-41<sup>M</sup> and pure MCM-41M was determined via kinetic analysis. Contact time effect data presented in Figure 6a show that the organosulfur contaminants were rapidly adsorbed by AgNO3/MCM-41M (about 5 min of contact) and reached equilibrium in approximately 120 min. No significant adsorption of these compounds was observed after 5 min for the pure MCM-41M, which may be related to the fewer available active sites [45]. Figure 6 shows the fitting and Table 3 brings the summary of the corresponding adsorption parameters and determination coefficients (R2).

**Table 3.** Kinetic parameters fitted to pseudo-first order, pseudo-second order, Elovich, intraparticle diffusion and Boyd models. Experimental conditions: room temperature and pressure, 500 ppm sulfur real diesel.


<sup>1</sup> B is the Boyd's constant, which is equivalent to the angular coefficient of Boyd's plot (t × Bt).

Experimental results of sulfur removal on AgNO3/MCM-41M adsorbent were better fitted to the pseudo-second order model, with R<sup>2</sup> > 0.99 and similar calculated and experimental *qe* values (7.31 mg/g and 7.26 mg/g, respectively), followed by Elovich (R<sup>2</sup> = 0.977) and pseudo-first order (R<sup>2</sup> = 0.814) models, respectively. The best data fitting to the pseudo-second order and Elovich models is a clue that the chemisorption process is the rate determining step of adsorption [46]. Furthermore, the higher value of parameter α obtained by Elovich equation (Table 3) suggests that there is a strong affinity between the recalcitrant compounds containing S and the active sites of AgNO3/MCM-41M adsorbent, strengthening the hypothesis that this adsorption is highly influenced by Ag sites and its different active forms (Ag(0) and Ag-O), and mainly occurs via π-complexation and oxidation [47,48]. The slower adsorption rate after the initial minutes may be a reflection of concurrent kinetic mechanisms [46]. For pure MCM-41M, the experimental results were better fitted to the pseudo-second order model, with R2 > 0.89 and calculated *q<sup>e</sup>* (1.60 mg/g) close to the experimental result (1.64 mg/g). Low R<sup>2</sup> values indicate that the pseudo-first-order and Elovich kinetic models did not adequately describe the adsorption process performed by this sample.

**Figure 6.** Plots of (**a**) contact time effect, (**b**) pseudo-first order, (**c**) pseudo-second order, (**d**) Elovich, (**e**) intraparticle diffusion and (**f**) Boyd models, where represents the kinetic data obtained for AgMCM-41<sup>M</sup> and represents the kinetic data obtained for MCM-41M.

Considering molecular structures, critical diameters of sulfur compounds in model diesel are smaller than those in commercial diesel [49], whose main composition is listed in Table S1 of the Supplementary Materials. Thus, performing desulfurization tests with the real fuel is the best way to evaluate the performance of the adsorbents. For the effective adsorption of larger molecules in terms of physisorption, not only should the pore size of the material be at least larger than the critical diameter of the adsorbate, but it should also be large enough to reduce diffusional resistance during adsorption [50]. Therefore, the high sulfur adsorption performed by AgNO3/MCM-41 can also be related to its larger pore diameter (24.82 Å) in relation to the other synthesized impregnated materials (as shown in Table 1), favoring adsorption due to both the more effective reduction of the diffusional resistance of aromatic

organosulfurs (whose critical diameters are shown in Table S2 of the Supplementary Materials) and the stronger chemical interactions promoted by the more active Ag species over this adsorbent.

This fact can be verified by the results of intraparticle diffusion and film diffusion of Boyd models. Since the data fitting curves for intraparticle diffusion did not cross the axes origin (Figure 6e), it can be assumed that intrapore diffusion is not the limiting step of the adsorption process, and therefore other mechanisms must be acting simultaneously. This is confirmed by the Boyd model (Figure 6f), since the graphs have non-zero intercepts. The non-linearity of Boyd model graphs indicates that pore diffusion is not the rate control step and that there is a mass transfer resistance step in the outer film [51]. Thus, although both intraparticle diffusion and film diffusion are involved in the adsorptive process, these physical phenomena do not govern desulfurization.

#### *2.4. Adsorption Equilibrium Isotherms*

Langmuir, Freundlich, Temkin, and Dubinin-Radushkevich equations were used to fit the experimental sulfur adsorption results with real diesel. The isotherm parameters and the correlation coefficients (R2) obtained from the linear fitting of experimental data are summarized in Table 4. The non-linear adjustments and variations in the separation factor (RL) as a function of initial sulfur concentration (C0) are displayed in Figures 7 and 8, respectively. Based on R<sup>2</sup> values, Langmuir and Freundlich mathematical models show good fitting to the adsorption equilibrium data for all the studied adsorbents. Experimental adsorption results for MCM-41<sup>M</sup> were better adjusted to the Freundlich model (R<sup>2</sup> = 0.9904) in comparison to the Langmuir model (R<sup>2</sup> = 0.9802), indicating that the process preferably occurs through physisorption [52]. However, as shown in Figure 7, the adsorption on materials modified with transition metals better fit the Langmuir model (R2 > 0.98), suggesting that the physisorption mechanism was preferably substituted by chemisorption [26].

The values of parameter n in the Freundlich isotherm are higher than 1 for all adsorbents, indicating favorable adsorption. The Temkin model (R<sup>2</sup> < 0.93) could be used to describe the experimental data. D-R model (R<sup>2</sup> < 0.47) was not able to fit the S-compounds adsorption processes performed in this work.

For the Langmuir isotherm, RL reached values between 0 and 1 (Figure 8), indicating that the sulfur adsorption is favorable. The maximum adsorption capacity of S-compounds (*qm*) for this model, in mg/g, was estimated in the following descending order: AgNO3/MCM-41<sup>M</sup> > Ag2O/MCM-41M > Ni(NO3)2/MCM-4M > NiO/MCM-41<sup>M</sup> > MCM-41M. This result is a clue that AgNO3/MCM-41<sup>M</sup> was able to promote the best adsorption performance among the tested adsorbents, considering the same study conditions. MCM-41M modification with Ag species from AgNO3 can offer a high quantity of new active sites over the silica surface to chemically interact with sulfur compounds [53].

The adsorption equilibrium results of the materials synthesized from commercial silica are shown in Table 4 and Figure 7f–h. As expected, the experimental sulfur adsorption data of pure MCM-41<sup>C</sup> were better fitted to the Freundlich model, and the results of the samples impregnated with non-calcined salts were better described by the Langmuir mathematical model. AgNO3/MCM-41<sup>C</sup> synthesized from commercial silica obtained the highest maximum adsorptive capacity among them (qm = 15.41 mg/g), approximately half of the content adsorbed by AgNO3/MCM-41M synthesized from amorphous silica MPI, emphasizing the higher sulfur adsorption efficiency of the renewable material even when compared to other adsorbents applied for desulfurization in the literature (Table S3 of the Supplementary Materials).

**Figure 7.** Non-linear adjustments of Langmuir, Freundlich, Temkin and Dubinin-Radushkevich (D-R) equilibrium isotherms of S-compounds adsorption over (**a**) AgNO3/MCM-41M, (**b**) Ag2O/MCM-41M, (**c**) Ni(NO3)2/MCM-41M, (**d**) NiO/MCM-41M, (**e**) MCM-41M, (**f**) AgNO3/MCM-41C, (**g**) Ni(NO3)2/MCM-41C and (**h**) MCM-41C. Experimental conditions: room temperature and pressure, real diesel with sulfur concecntration ranging from 25 to 1000 ppm.



**Figure 8.** Plots of RL versus initial concentrations.

#### *2.5. Adsorption Thermodynamics*

A thermodynamic assessment was performed for the sulfur adsorption over the AgNO3/MCM-41M material. The properties observed at 298, 308 and 318 K are given in Table 5. The values of ΔH◦ and ΔS◦ can be obtained from the slope and intercept of Van't Hoff plot of ln(KL) against the reciprocal of T (1/T, K<sup>−</sup>1) (Figure 9). The Langmuir model fitting for sulfur adsorption results with this material at 298, 308 and 318 K is depicted in Figure S4 of the Supplementary Materials.

**Table 5.** Thermodynamic parameters for sulfur adsorption on AgNO3/MCM-41M. Experimental conditions: room pressure, real diesel with sulfur concentration ranging from 25 to 500 ppm.


<sup>1</sup> KL is the Langmuir equilibrium constant calculated by the equation *Ce qe* <sup>=</sup> *Ce qm* <sup>+</sup> <sup>1</sup> *qmKL* .

**Figure 9.** Van't Hoff plot.

The positive value for ΔH◦ indicates an endothermic process and its magnitude reaches the region defined for chemisorption in literature (≥40 kJ/mol) [54,55], thus suggesting predominantly chemical interactions between S-compounds and AgNO3/MCM-41M adsorbent. Although the process is not spontaneous, it can be noticed that there is a tendency of increasing spontaneity by increasing the adsorption temperature, since the total free energy (ΔG◦) of the system decreases. Finally, the positive entropy (ΔS◦) confirms the affinity of the adsorbent to S-compounds [46,56].

The data from the kinetic, equilibrium and thermodynamic tests reinforce the results observed in XPS spectra (Figure 5) for AgNO3/MCM-41M and Ag2O/MCM-41M materials before and after desulfurization, where stable Ag2O species apparently did not interact with DBT molecules and only bridging bonds occurred between Ag (0) sites and the organosulfur rings (Ag-S-R), while active Ag-O species and Ag(0) nanodomains concurrently participate for adsorbing the contaminants via oxidation reactions and bridging interactions via π-complexation on the non-calcined adsorbent, visibly increasing the process efficiency. The same pattern occurs for Ni-impregnated MCM-41M, where the non-calcined sample presents higher sulfur adsorption capacity than the calcined material; however, their results are not significant in comparison to those obtained with Ag species.

The sulfur adsorption results obtained for the MCM-41<sup>M</sup> and AgNO3/MCM-41<sup>M</sup> materials regenerated by eluting a chloroform/hexane/acetone solution in the 40:30:30 ratio are depicted in Figure 10. Both materials presented a satisfactory reuse rate with just a slight decrease in sulfur adsorption after five regeneration cycles, thus indicating that the methodology was successful in removing the adsorbed organosulfurs without compromising their adsorption efficiency.

**Figure 10.** Sulfur adsorption results for the regenerated materials. Experimental conditions: room temperature and pressure.

#### **3. Materials and Methods**

#### *3.1. Synthesis*

Amorphous silica (MPI) was synthesized from beach sand following a new methodology described in our previous publication [23]. This silica presents isotherm type III, typical of some porous solids with meso and macroporosity, 33.54 m2/g surface area (SBET), 130.3 Å pore diameter and 0.18 cm<sup>3</sup>/g pore volume. In addition, this material presents a chemical composition of 96.05% SiO2 and main impurities such as K2O, Fe2O3 and CaO with average contents of 1.84%, 1.12% and 0.81%, respectively.

MCM-41 was synthesized by the hydrothermal method described in [22,23] using MPI silica, with some modifications. The reaction system was composed of two initial solutions: (I) a mixture containing 14.75 g of MPI silica, 4.66 g of sodium hydroxide (PA, Vetec|Sigma-Aldrich, Duque de Caxias, RJ, Brazil) and 105.0 mL of distillated water, which was stirred for 2 h at 333 K; and (II) a mixture containing 21.68 g of cetyltrimethylammonium bromide (CTAB) (98.0%, Vetec|Sigma-Aldrich, Duque de Caxias, RJ, Brazil) and 105.0 mL of distillated water, stirred for 1 h at room temperature. Then, mixture (I) was added to (II) and stirred for 1 h at room temperature. The final gel had a molar composition of 1.0 CTAB:4.0 SiO2:2.0 NaOH:200.0 H2O, and was added into a teflon autoclave at 373 K for 96 h. The pH was adjusted daily with a 30% acetic acid (99.8%, Proquímios, Bangu, RJ, Brazil) solution until achieving a range of 9.5–10.0 pH. Then, the gel was washed with distilled water and 37% HCl (PA, Proquímios, Bangu, RJ, Brazil) solution in ethanol (99.5%, NEON, São Paulo, SP, Brazil) at 2%. The material was recovered by vacuum filtration and then dried at 393 K for 2 h. Finally, the solid was calcinated in air flow at 823 K with 278 K/min heating rate for 5 h. MCM-41 was also synthesized with commercial silica (silica gel 60, Macherey-Nagel, Düren, Germany), MCM-41C, by following the same methodology for comparison with MCM-41<sup>M</sup> in sulfur adsorption tests.

#### *3.2. Adsorbent Modification*

The mesoporous MCM-41 silica was modified with Ag<sup>+</sup> and Ni2<sup>+</sup> cations from silver nitrate (99.8%, Vetec|Sigma-Aldrich, Duque de Caxias, RJ, Brazil) and hexa-hydrated nickel nitrate (99.999%, Sigma-Aldrich, St. Louis, MO, USA) salts via wet impregnation. Ethanolic solutions of AgNO3 (0.03 mol/L for 2% (*w*/*w*) and 0.12 mol/L for 8% (*w*/*w*) concentrations) and Ni(NO3)2·6H2O (0.05 mol/L for 2% (*w*/*w*) and 0.22 for 8% (*w*/*w*)) salts were prepared and added to 3.0 g of calcinated MCM-41. The functionalizations were performed in rotary evaporators at 120 rpm for 1 h at room temperature and 3.5 h under progressive heating (283 K/30 min) in a water bath until the solvent completely evaporated. The modified adsorbents were dried in an oven at 393 K for 2h [57]. Next, 50% of the material was separated and denominated AgNO3/MCM-41 and Ni(NO3)2/MCM-41, while the other 50% was calcinated at 773 K during 4 h in an oven at 278 K/min heating rate and denominated Ag2O/MCM-41 and NiO/MCM-41.

#### *3.3. Adsorbent Characterization*

X-ray diffraction analyses (XRD) were performed using a Bruker D2 Phaser (Bruker AXS, Madison, WI, USA) with CuKα radiation (λ = 1.5406 Å), 30 kV filament, 10 mA current, Ni filter and a LYNXEYE detector. The XRD patterns were obtained in the range of 1◦–10◦ (low-angle) and 10◦–90◦ (wide-angle) 2θ.

The textural parameters were evaluated via nitrogen adsorption–desorption isotherms at 77 K, as determined by an automatic ASAP 2420 system from Micrometrics (Micrometrics, Norcross, GA, USA). Prior to the measurements, the samples were previously degassed at 473 K and 10−<sup>4</sup> mbar. The pore size distribution was calculated by applying the Barrett–Joyner–Halenda (BJH) method to the desorption branch of the N2 isotherm.

MCM-41M and AgNO3/MCM-41<sup>M</sup> were evaluated by high-resolution transmission electron microscopy (HRTEM) using a Philips CCCM 200 Supertwin-DX4 microscope. Scanning transmission electron microscopy (STEM) analysis and mapping data were recorded in a Helios Nanolab 650 (FEI, Brno, Czech Republic) instrument with a high-angle annular dark field (HAADF) detector, at 200 kV and 200 nA. The microanalysis was carried out with energy dispersive X-ray (EDAX) spectroscopy Super-X system provided with four X-ray detectors and an X-FEG beam.

X-ray photoelectron spectra (XPS) was collected using a Physical Electronics PHI 5700 (Physical Electronic, Minneapolis, MN, USA) spectrometer with non-monochromatic Al Kα radiation (95.2 W, 15 kV, and 1486.6 eV) and a multi-channel detector. C1s peaks were used as an inner standard calibration peak at 284.8 eV. The Multipack software version 9.6.0.15 was used for data analysis. The recorded spectra were fitted using Gaussian–Lorentzian curves to more accurately determine the binding energies of the different element core levels.

#### *3.4. Real and Model Fuels*

Diesel fuel samples containing 1234.9 ppm (high-sulfur) and 5.1 ppm (low-sulfur) sulfur were kindly donated by the Clara Camarão Potiguar Refinery (Guamaré/RN, Brazil). This material was used for equilibrium, kinetic and thermodynamic evaluation of sulfur adsorption tests performed in a batch system. Moreover, model diesel solutions prepared with dibenzothiophene (98.0%, Sigma-Aldrich, Co., St. Louis, MO, USA) and n-decane (99.0%, Sigma-Aldrich, Co., St. Louis, MO, USA) in a concentration of around 2000 ppm sulfur were used in a set of adsorption tests in fixed bed columns, exclusively for evaluating the composition of the adsorbents after desulfurization via XPS analysis, since real diesel has many interferents in its composition and would generate noisy spectra.

#### *3.5. Preliminary Adsorption Tests*

All materials prepared in this work were subjected to initial sulfur adsorption tests with real diesel fuel samples (1234.9 ppm sulfur). The tests were performed in adsorption columns (0.61 cm × 30.0 cm) with downward flow. Each adsorbent was added to the glass columns up to 9.0 cm in height (about 0.5 g of material), and a volume of 5.0 mL of diesel was kept constant during the experiments. The first aliquots (0.3 mL) were collected and the sulfur content data were recorded using a total sulfur analyzer via ultraviolet fluorescence (UVF) spectrometry (Antek Multitek, PAC, L.P, Houston, TX, USA).

#### *3.6. Batch Adsorption*

The adsorptive desulphurization tests were performed using the finite bath method at ambient temperature and pressure. An amount of 0.3 g of adsorbent and 8.0 mL of diluted real diesel prepared by a solution of the high and low-sulfur samples were placed in erlenmeyer flasks under constant stirring (100 rpm) in a stirring table (SL 180/DT, Solab, Piracicaba, SP, Brazil). Different proportions of real diesel samples were used to produce the solutions with distinct sulfur concentrations which were applied in each test. The solids were separated by centrifugation and the residual sulfur contents in the liquid phase were collected for quantification following ASTM D5453 methodology. Adsorption effectivity was measured by the UVF sulfur analyzer and the sulfur adsorption capacity per gram of adsorbent at equilibrium (*qe*) was calculated using the following expression:

$$q\_c = (\mathbb{C}\_i - \mathbb{C}\_c)V/W \tag{1}$$

where *V* is the diesel solution volume (L), *W* is the adsorbent mass (g) and *Ci* and *Ce* are the initial sulfur content in the fuel (mg/L) and at equilibrium, respectively.

#### 3.6.1. Kinetic Tests

Kinetic tests were performed for MCM-41<sup>M</sup> and AgNO3/MCM-41M samples following the methodology described for batch adsorption using a real diesel solution with 500 ppm sulfur content. The supernatant fluid was collected at predetermined times between 5 and 180 min of stirring. The generated models are used to explain the adsorption mechanism characteristics. Pseudo-first-order (Equation (2)), pseudo-second-order (Equation (3)), Elovich (Equation (4)), intra-particle diffusion (Equation (5)) and the film-diffusion model of Boyd (Equation (6)) were used to analyze the experimental data using the following equations [46]:

$$
\ln(q\_t - q\_t) = \ln q\_t - k\_1 t \tag{2}
$$

$$\frac{t}{q\_t} = \frac{1}{k\_2 q\_c^2} + \frac{t}{q\_c} \tag{3}$$

$$q\_t = \frac{1}{\beta} \ln \alpha \beta + \frac{1}{\beta} \ln t \tag{4}$$

$$q\_t = k\_{\rm id} t^{\frac{1}{2}} + \mathbb{C} \tag{5}$$

$$F = 1 - \frac{6}{\pi^2} \sum\_{n=1}^{\infty} \frac{1}{n^2} \exp\left(-n^2 B\_t\right) \tag{6}$$

where *qt* and *qe* are the amounts of sulfur adsorbed (mg/g) at time *t* (min) and at equilibrium, respectively, *k*<sup>1</sup> is the pseudo-first-order constant (min<sup>−</sup>1), *k*<sup>2</sup> is the pseudo-second-order adsorption constant (mg/g min), the parameter α is the initial adsorption rate of Elovich equation (mg/g min) and β is the desorption constant (g/mg) and *kid* is the rate constant of the intra-particle diffusion models (mg/g min1/2). *F* is fractional uptake (*qt*/*qe*) at a certain time and *Bt* is a mathematical function of *F* calculated by Equations (7) and (8) [58]:

$$\text{For } F > 0.85, \ B\_l = f(F) = -0.4977 - \ln(1 - F) \tag{7}$$

$$\text{For } F < 0.85, \ B\_t = f(F) = \left(\sqrt{\pi} - \sqrt{\pi - \sqrt{\frac{\pi^2 F}{3}}}\right)^2 \tag{8}$$

#### 3.6.2. Adsorption Equilibrium Tests

The evaluation of sulfur adsorption equilibrium was performed in a batch system during 24 h with samples of real diesel containing sulfur concentrations in the 25–1000 ppm range. Langmuir, Freundlich, Temkin and Dubinin-Radushkevich mathematical models were used to fit the experimental data of sulfur adsorption in order to explain the adsorption on the adsorbent/adsorbate system, as well as predict their equilibrium parameters. The adsorption equilibrium of MCM-41<sup>C</sup> produced from commercial silica and impregnated with non-calcined salts was also evaluated for comparison with the renewable material. The Langmuir model (Equations (9) and (10)) was used to quantify and contrast the performance of various adsorbents [59]:

$$q\_c = \frac{q\_m K\_L \mathcal{C}\_c}{1 + K\_L \mathcal{C}\_c} \tag{9}$$

$$\frac{C\_{\epsilon}}{q\_{\epsilon}} = \frac{1}{q\_{m}K\_{L}} + \frac{C\_{\epsilon}}{q\_{m}} \tag{10}$$

where *Ce* (g/L) and *qe* (mg/g) are the equilibrium concentration of the adsorbate and adsorption capacity of the adsorbent, respectively; Langmuir *qm* is a constant expressing the maximum absorption of the adsorbate (mg/g) and *KL* is the Langmuir constant (L/g), also related to the adsorption energy and the affinity of the adsorbent. Dimensionless constant commonly known as a separation factor (*RL*) defined by Weber and Chakravorti (1974) [60], can be given by the following equation:

$$R\_L = 1/\left(1 + K\_L C\_0\right) \tag{11}$$

where *C*<sup>0</sup> in mg/L is the initial concentration of the analytes and *KL* in L/mg is Langmuir constant. *RL* > 1 indicates an unfavorable process, *RL* = 1 indicates linearity, 0 < *RL* < 1 indicates a favorable process and *RL* = 0 signifies an irreversible adsorption.

The Freundlich model (Equations (12) and (13)) describes non-ideal and reversible adsorption, not being restricted to the formation of a monolayer [61]. This empirical model can be applied to the adsorption of multiple layers, with a non-uniform distribution of adsorption heat and affinities of the heterogeneous surface:

$$q\_{\mathfrak{e}} = \mathbb{K}\_{\mathbb{F}} \mathbb{C}\_{\mathfrak{e}}^{\frac{1}{n}} \tag{12}$$

$$
\ln q\_{\varepsilon} = \ln \mathcal{K}\_{\mathcal{F}} + \frac{1}{n} \ln \mathcal{C}\_{\varepsilon} \tag{13}
$$

The Freundlich expression is an exponential equation and therefore assumes that the adsorbent concentration in the surface increases with the concentration of adsorbate. With the use of this expression, an infinite amount of adsorption can theoretically occur [62]. The equation is widely used in heterogeneous systems, where *KF* (L/g) and n are characteristic constants of Freundlich, which indicate the adsorption capacity and adsorption intensity, respectively, where values of n in the range 1 < *n* < 10 indicate favorable adsorption.

The Temkin isotherm model takes into consideration the effects of indirect adsorbate/adsorbate interactions which suggest that the adsorption heat of all adsorbed molecules in the layer decreases linearly with coverage [47]. The model can be expressed according to Equation (14):

$$
\eta\_{\varepsilon} = \beta \ln \alpha \mathbb{C}\_{\varepsilon} \tag{14}
$$

where β is related to the heat of adsorption (J/mol) and α is the Temkin isotherm constant (L/g). The Temkin linearized equation is given in Equation (15):

$$
\eta\_{\varepsilon} = \beta l n \alpha + \beta l n \text{C}\_{\varepsilon} \tag{15}
$$

The Dubinin-Radushkevich (D-R) isotherm expresses the adsorption mechanism on a heterogeneous and porous surface with variable parameters [63]. The D-R model is described by Equation (16), and the linearized equation is presented in Equation (17):

$$q\_{\iota} = q\_{\iota} \varepsilon^{-\beta \iota^2} \tag{16}$$

$$
\ln q\_{\varepsilon} = \ln q\_{\varepsilon} - \beta \varepsilon^2 \tag{17}
$$

where *qs* is the theoretical saturation capacity (mg/g), β is the D-R constant (mol2/kJ2) and ε is the Polanyi potential, and can be expressed as follows:

$$
\varepsilon = RTln(1 + 1/\mathbb{C}\_{\varepsilon}) \tag{18}
$$

where *R* is the universal gas constant (8.314 J/mol K), and *T* is the absolute temperature (K).

The β constant is related to the mean free energy of adsorption which is computed through Equation (19):

$$E = 1/\sqrt{2\beta} \tag{19}$$

where *E* is the mean adsorption energy (kJ/mol). The values of *E* < 40 kJ/mol implies physical adsorption while *E* > 40 kJ/mol suggests chemical adsorption [64].

Origin 2018 software (Origin Labs) was used to adjust the linearized equations of the models for calculating the equilibrium parameters.

#### 3.6.3. Sulfur Adsorption Thermodynamics

The effect of temperature in S-compounds adsorption was studied by varying the temperature from 298, 308 and 318 K. The obtained thermodynamics parameters describe the variation or the transformation of a system. The standard free energy (Δ*G*◦, kJ/mol), standard enthalpy change (Δ*H*◦, kJ/mol) and standard entropy change (Δ*S*◦, kJ/mol K) can be calculated from the following Equations [54]:

$$
\Delta G^\circ = -RT\ln(K\_L) = \Delta H^\circ - T\Delta S^\circ \tag{20}
$$

$$\ln(\mathcal{K}\_L) = -\frac{\Delta H^\circ}{RT} + \frac{\Delta S^\circ}{R} \tag{21}$$

where *KL* is the Langmuir isotherm constant, *R* is the gas constant (8.314 J/K mol) and *T* is temperature in Kelvin.

#### 3.6.4. Regeneration Tests

The regeneration methodology consisted of packing 1 g of residual adsorbent collected from the sulfur adsorption tests in a column with ~10 mm diameter and 50 cm length, forming a fixed bed with about 8 cm adsorbent. Next, 25 mL of a chloroform/hexane/acetone solution in a ratio of 40:30:30 (chosen after some preliminary tests using various proportions) at 323 K were added to the column. When the entire solution was completely eluted, the washed adsorbent was heated to 343 K in a rotary evaporator for 1 h to evaporate the solvent residue and replaced in a column for a new adsorption test. This evaluation was repeated for 5 regeneration cycles. A process flow diagram of the overall methodology is presented in Scheme 1.

**Scheme 1.** Process flow diagram for the methodology used in this work.

#### **4. Conclusions**

The MCM-41<sup>M</sup> adsorbent obtained from MPI silica presented the typical characteristics of this material, and its modification with metallic ions did not alter the material structure. Preliminary desulfurization results indicated better adsorption performance of AgNO3/MCM-41M and Ag2O/MCM-41M samples in comparison to the Ni-impregnated materials. High-resolution images of STEM-EDX and XPS spectra showed the presence of highly dispersed Ag on MCM-41M support in the forms of Ag(0) nanoparticles and active Ag-O species, which were able to interact with organosulfur rings via oxidation and π-complexation in the non-calcined adsorbent (AgNO3/MCM-41M).

The results from the kinetic, equilibrium and thermodynamic tests reinforced the phenomena observed in XPS data and suggest that the chemisorptive interaction between organosulfur compounds and active Ag nanosites controls sulfur adsorption. The maximum adsorption capacity calculated by the Langmuir equation was about ten times higher in comparison to pure MCM-41<sup>M</sup> and double the *qm* for the AgNO3/MCM-41C sample from commercial silica. Thus, the AgNO3/MCM-41M material synthesized from MPI silica presents significant efficiency for adsorptive desulfurization and can be used as a complementary process to the expensive HDS method for its low-cost, sustainability and efficient removal of recalcitrant organosulfur compounds. Furthermore, the adsorbents presented a satisfactory regeneration rate by using a chloroform/hexane/acetone mixture (40:30:30 ratio) after a total of five sulfur adsorption cycles. This allows for efficient reuse of this material with certain longevity, ensuring even more reduction in operating costs.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4344/9/8/651/s1, Figure S1: Results for preliminary sulfur adsorption tests, Figure S2: XRD results for the adsorbents with 8% metal, Figure S3: Langmuir model fitting for sulfur adsorption results over AgNO3/MCM-41<sup>M</sup> material at different temperatures, Table S1: Sulfur compound distribution in a straight run diesel oil sample (Hua et al., 2003), Table S2: Critical diameter data for the main aromatic organosulfurs in real diesel, Table S3: Sulfur adsorption performance of some materials studied in literature for comparison.

**Author Contributions:** Conceptualization, R.V.S. and L.S.d.C.; methodology, R.V.S., A.B.F.C. and J.A.B.S.; formal analysis, E.R.-C., L.M.A.C., M.M.U. and S.B.C.P.; investigation, R.V.S., L.S.d.C., H.O.M.A.M., and A.B.F.C.; resources, L.M.A.C., J.A.B.S. and T.C.B.; data curation, E.R.-C. and S.B.C.P.; writing—original draft preparation, R.V.S., H.O.M.A.M., A.B.F.C., J.A.B.S. and L.S.d.C.; writing—review and editing, R.V.S., H.O.M.A.M., A.B.F.C., E.R.-C. and L.S.d.C.; supervision, R.V.S. and L.S.d.C.; project administration, L.S.d.C.; funding acquisition, S.B.C.P. and E.R.-C.

**Funding:** This research was funded by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES—Brazil), grant number 001.

**Acknowledgments:** The authors thank the Energetic Technologies Laboratory (LTEN) and the Clara Camarão Potiguar Refinery (RPCC—Petrobras) for the diesel samples. The authors thank the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES—Brazil). ERC thanks to Ministerio de Ciencia, Innovación y Universidades, Project RTI2018-099668-B-C22 and FEDER funds.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **A New Tool in the Quest for Biocompatible Phthalocyanines: Palladium Catalyzed Aminocarbonylation for Amide Substituted Phthalonitriles and Illustrative Phthalocyanines Thereof**

#### **Vanessa A. Tomé, Mário J. F. Calvete \* , Carolina S. Vinagreiro , Rafael T. Aroso and Mariette M. Pereira \***

Centro de Química de Coimbra (CQC), Department of Chemistry, University of Coimbra, Coimbra 3004-535, Portugal; vanessalmeida\_97@hotmail.com (V.A.T.); carolina\_svinagreiro@hotmail.com (C.S.V.); fortilap@gmail.com (R.T.A.)

**\*** Correspondence: mcalvete@qui.uc.pt (M.J.F.C.); mmpereira@qui.uc.pt (M.M.P.); Tel.: +351-966-174744 (M.J.F.C.); +351-239-854474 (M.M.P.)

Received: 1 October 2018; Accepted: 17 October 2018; Published: 20 October 2018

**Abstract:** The amide peptide bond type linkage is one of the most natural conjugations available, present in many biological synthons and pharmaceutical drugs. Hence, aiming the direct conjugation of potentially biologically active compounds to phthalocyanines, herein we disclose a new strategy for direct modulation of phthalonitriles, inspired by an attractive synthetic strategy for the preparation of carboxamides based on palladium-catalyzed aminocarbonylation of aryl halides in the presence of carbon monoxide (CO) which, to our knowledge, has never been used to prepare amide-substituted phthalonitriles, the natural precursors for the synthesis of phthalocyanines. Some examples of phthalocyanines prepared thereof are also reported, along with their full spectroscopic characterization and photophysical properties initial assessment.

**Keywords:** peptide bond; phthalonitriles; phthalocyanines; aminocarbonylation; palladium catalysts

#### **1. Introduction**

Molecules of the tetrapyrrole family (e.g., porphyrin and phthalocyanine derivatives) are probably the most appealing chromophores for a vast array of photoactivated processes, such as phototherapy [1–3], photodiagnosis [4–6], photocatalysis [7,8], solar energy conversion [9,10], and also as photomaterials [11–13]. In particular, phthalocyanines largely fulfil a crucial optical requisite, necessary for photomedicinal applications, which is a strong absorption in near-infrared (NIR) spectral region (600–900 nm), as light in this region affords the deepest penetration in soft tissue. These highly stable compounds, possessing high molar absorptivity, high quantum yields of fluorescence and structural versatility [4,13,14], can be conveniently modified to grant suitable biological solubility [15–18], by introduction of hydrophilic moieties in the structure. The majority of the moieties used so far are negatively charged, such as sulfonates or carboxylates [15,19] or positively charged, like quaternized amines [15,20]. These ionic features present some important drawbacks like low cellular uptake and/or cellular internalization, due to the negatively-charged character of the cell plasma membranes, in the case of anionic phthalocyanines [21], or exaggerated phospholipid affinity, leading to phospholipidosis, in the case of cationic phthalocyanines [22]. Thus, in the search for biocompatibility, the amide peptide bond type linkage is one of the most natural conjugations available, present in many biological synthons, such as peptides, proteins, or amino acids [23], as well in

pharmaceutical drugs [24,25]. Nevertheless, amide-substituted phthalocyanines are rare [26–35], when compared with other functionalities, including a few reports of phthalocyanines conjugated with amino acids [27–30] and peptides [31–35]. The main reason for the scarcity in phthalocyanine-amino acid conjugates arises from the difficult synthetic manipulation, which relies in troublesome transformations and purification procedures using highly polluting chemicals [25,27–30,36,37].

Apparently, phthalocyanine post-synthetic modulation would not be a very straightforward option, due to the chemical stability owned by phthalocyanines, which are quite stable against this type of structural variation. On the other hand, modification of precursory phthalonitriles bearing carboxylic acids is also demanding, given the sensitiveness of nitrile functions. It is worth mentioning that we have tested the strategies ourselves, to explain our points, and results were as described in the text. Whether in case of post-synthetic phthalocyanine modification or phthalonitrile modulation, no reproducible results could be obtained, always leading to cumbersome work-up approaches.

Palladium-catalyzed carbonylation reactions were first described by Heck almost 40 years ago [38]. Since then, many developments have been reported [6,39–46] and nowadays carbonylation has become an indispensable alternative to the classic organic synthesis of carbonyl compounds, including carboxylic acid derivatives (e.g., amides, esters) with valuable application in both industrial and fine chemistry. Among these reactions, aminocarbonylation [47–50], carried out using Ar–X substrates (X=I, Br, Cl, OTf, OTs, etc.), in the presence of *N*-nucleophiles, emerges as a sustainable, one-step synthetic approach, for the efficient, selective and mild synthesis of amides.

Herein we disclose a new strategy for direct modulation of phthalonitriles, inspired by an attractive synthetic strategy for the preparation of carboxamides based on optimized palladium-catalyzed aminocarbonylation of aryl halides in the presence of carbon monoxide (CO) [38,51] which, to our knowledge, has never been used to prepare amide-substituted phthalonitriles, the natural precursors for the synthesis of phthalocyanines. Furthermore, transformation thereof to the desired phthalocyanines is also described.

#### **2. Results and Discussion**

Modification of phthalonitriles is usually the chosen methodology when attempting to introduce significant changes at the phthalocyanine periphery, instead of phthalocyanine post-synthetic modulation [52], due to the known chemical stability owned by phthalocyanines. A conceivable example would be, for instance, to synthesize a phthalocyanine bearing peripheral four carboxylic acid groups, followed by acyl chloride formation, using a hazardous chlorinating agent, and then functionalization with an amine. The main issue regarding this strategy would be the proneness to form mixtures of mono-, di-, tri-, and tetra-amide substituted phthalocyanines requiring the use of excessive amounts of nucleophile, giving raise to cumbersome purification and low yields.

Our herein envisaged strategy uses 4-iodophthalonitrile (**1**) [53] as substrate and a range of amines as nucleophiles, in presence of a palladium catalyst formed in situ by addition of palladium(II) acetate to triphenylphosphine (in 1:2 molar ratio), together with Et3N as base and carbon monoxide as reagent (Table 1) [54,55]. Our studies began with the aminocarbonylation of 4-iodophthalonitrile (**1**) using glycine methyl ester hydrochloride (**2a**) as model nucleophile, to optimize reaction conditions (temperature, pressure of CO and time reaction parameters) in the palladium-catalyzed aminocarbonylation reaction (Table 1).

The first reaction conditions employed (PCO = 10 bar and T = 65 ◦C) (Table 1, entry 1) afforded only 25% substrate conversion, after 24 h. Then, we investigated the effect of increasing temperature, keeping the CO pressure at 10 bar, and it was found that, at 85 ◦C, the reaction proceeded faster, and full conversion of substrate **1** within 24 h was obtained (Table 1, entry 2). Keeping CO pressure at 10 bar and temperature at 85 ◦C, the reaction was not complete when reaction time was decreased to 12 h, reaching only 70% conversion of **1** (Table 1, entry 3). Conversely, when the reaction temperature was increased to 100 ◦C, keeping the pressure at 10 bar and reaction time at 12 h, substrate **1** was totally transformed into the desired amide (Table 1, entry 4). In addition, when the CO pressure was

reduced to 5 bar, keeping the temperature at 100 ◦C, after 12 h, the conversion of substrate **1** was >97% (Table 1, entry 5). However, keeping CO pressure at 5 bar and reducing the temperature to 65 ◦C, it required 70 h until full conversion of substrate **1** was observed (Table 1, entry 6). Thus, this indicates that the temperature plays the most important role on the activity of the catalyst. To evaluate the effect of solvent, an additional experiment was performed using DMF instead of toluene and, regardless of the high conversion obtained using the same conditions, this reaction yielded a complex mixture of products (Table 1, entry 7), as checked and compared using thin layer chromatography-TLC, which may be attributed to decomposition of DMF. Summing, it was found that a temperature of 100 ◦C, a CO pressure of 5 bar and a reaction time of 12 h were the optimal reaction parameters selected to extend the scope of 4-iodophthalonitrile functionalization.


**Table 1.** Optimization of reaction conditions a.

<sup>a</sup> General reaction conditions: 2.5 mol % Pd(OAc)2, 5 mol % PPh3, 8 equiv. Et3N, 1.1 equiv. **2a**. <sup>b</sup> Substrate conversion determined by 1H-NMR on the reaction mixture obtained after evaporation of the solvent; <sup>c</sup> gave a complex mixture of products.

Hence, our next step was to promote the catalytic aminocarbonylation reaction between 4-iodophthalonitrile (**1**) with a wide range of amines as nucleophiles (**2a**–**2g**) to obtain the corresponding carboxamides (Table 2). Several structurally different amines as *N*-nucleophiles were used: three amino acid methyl esters (methyl glycinate (**2a**), methyl leucinate (**2b**) and methyl phenylalaninate (**2c**)), *tert*-butylamine (**2d**), *N*-BOC-ethylenediamine (**2e**), chalcone (*E*)-1-(4-aminophenyl)-3-(3,4,5-trimethoxyphenyl)prop-2-en-1-one (**2f**) [56], and piperazine (**2g**) (Table 2). Each reaction's progress was followed by TLC of aliquots taken from the reactor via *cannula*. After complete conversion of **1** to the corresponding carboxamides, the reaction mixture was then subjected to work-up and chromatographic purification procedures (See SI), yielding amide substituted phthalonitriles **3a**–**3g**, in good isolated yields (54–80%) at optimized reaction conditions (Table 2) (see also Figures S1–S21, SI).

When aminoesters were used as nucleophiles (**2a**–**2c**), 12 h were necessary for the complete conversion of the substrate, leading to carboxamides **3a**, **3b** and **3c** in 65%, 54%, and 59% isolated yields, respectively (Table 2, entries 1–3). It is worth mentioning that the yields obtained were higher than the ones reported for the model substrate iodobenzene using similar aminoesters as nucleophiles [57,58]. This may be attributed to the presence of cyano electron-withdrawing groups in 4-iodophthalonitrile, which enables an easier oxidative addition step in the catalytic cycle (A in Figure 1).


**Table 2.** Palladium catalyzed aminocarbonylation of 4-iodophthalonitrile using several amines as nucleophiles a.

<sup>a</sup> General reaction conditions: 5 bar (CO), 2.5 mol % Pd(OAc)2, 5 mol % PPh3, 8 equiv. Et3N. Reactions were carried out in toluene (0.1 M, concerning substrate **1**). <sup>b</sup> Isolated yield.

Simpler aliphatic amines such as *tert*-butylamine (**2d**) were also used as nucleophiles. In this case, using 3.3 equivalents of **2d**, the aminocarbonylation reaction of **1** proceeded in the presence of palladium catalyst formed in situ by addition of palladium(II) acetate to triphenylphosphine (in 1:2 molar ratio), using toluene as solvent, under a CO pressure of 5 bar. Complete conversion of **1** was obtained in just 4 h, yielding carboxamide **3d** in 74% isolated yield (Table 2, entry 4).

Moreover, an *N*-mono-protected-ethylenediamine (**2e**) was also used. We had to prepare the mono-protected amine since, under the same reaction conditions, when unprotected ethylenediamine as nucleophile was used, a complex mixture of *N*-mono and *N,N'*-bis-substituted ethylenediamine, along with degradation products was formed, according to 1H-NMR analysis. To overcome this problem, we then prepared **2e**, using *tert*-butyloxycarbonyl protecting group (BOC group) following a literature procedure [59]. Next, we promoted the aminocarbonylation reaction of **1** with nucleophile *N*-BOC-ethylenediamine (**2e**), yielding **3e** in 80% isolated yield after just 3 h (Table 2, entry 5). Next, the aminocarbonylation of **1** with chalcone **2f**, which is a potential anti-microbial agent [60], yielded carboxamide **3f** in 70% isolated yield, under standard reaction conditions, after 25 h (Table 2, entry 6). Since the chalcone is an aromatic amine, it is expected to be less nucleophilic and, consequently

a prolonged period of time was necessary for the complete conversion of the substrate **1** into the corresponding carboxamide.

**Figure 1.** Simplified catalytic cycle describing the formation of 4-amide substituted phthalonitriles. L = PPh3.

Using similar conditions, we also investigated the use of cyclic diamines in the palladium catalyzed aminocarbonylation reaction for the synthesis of *N*-mono-substituted diamines. Unprotected diamine piperazine (**2g**) is quite useful and interesting because the presence of two amine groups could enable the conjugation with bioactive molecules or functionalization with other relevant chemical groups. In order to attain the desired *N*-mono-substituted diamine we have selected an excess of 6 equiv. of the diamine **2g**. In the presence of palladium catalyst formed in situ by addition of palladium(II) acetate to triphenylphosphine (in 1:2 molar ratio), together with Et3N as base in toluene solvent, under a CO pressure of 5 bar, complete conversion of **1** was obtained, after 7 h, yielding carboxamide **3g** after work-up and purification in 77% isolated yield (Table 2, entry 7).

According to previously described [49,51,61–64], a simplified mechanism for the formation of 4-amide substituted phthalonitrile is proposed in Figure 1. The catalytic cycle begins with the oxidative addition of the in situ formed Pd(0)Ln active species to the 4-iodophthalonitrile, resulting in an arylpalladium(II) intermediate **A**, which is able to coordinate to carbon monoxide, leading to intermediate **B**. Then, this complex undergoes a nucleophilic attack by the desired amine (*N*-nucleophile), affording **C**. Through HI elimination with the aid of Et3N, intermediate **D** is formed, yielding the desired 4-amide substituted phthalonitrile, upon reductive elimination.

All carboxamide substituted phthalonitriles were characterized by 1H, 13C-NMR and mass spectrometry and their structures confirmed. It is worth mentioning that, under the reaction conditions employed (100 ◦C and 5 bar), 100% chemoselectivity toward the mono-carboxamide products was obtained, since no double carbon monoxide insertion product was observed, using these amines as nucleophiles [55,65].

Having established a methodology for the synthesis of several carboxamide-containing phthalonitriles **3a**–**3g**, we have then prepared, as selected examples, phthalocyanines **4a**, **4c**, and **4d**, starting from the corresponding phthalonitriles **3a**, **3c**, and **3d** (Table 3) (see also Figures S22–S27, SI). We have used an approach where the tetramerization of the phthalonitriles was carried out in pentan-1-ol at 140 ◦C, in the presence of zinc(II) acetate, for 20 h, with all reactions progress being followed by TLC and UV–VIS spectroscopy. Phthalocyanines **4a**, **4c** and **4d** were obtained, after purification and isolation by column chromatography on silica gel, in 58, 65 and 68% yields, respectively.

We have observed that the purification procedure for phthalocyanines **4a** and **4c** was considerably more demanding than for phthalocyanine **4d**. We have found that, even after repeated recrystallization from methanol/diethyl ether, pentan-1-ol remained coordinated with the waxy phthalocyanine molecules **4a** and **4c**, as observable on their corresponding 1H-NMR spectra. On the other hand, solid phthalocyanine **4d**, bearing *tert*-butyl carboxamide groups, was easily recrystallized from methanol. We assume this occurrence to the nature of the carboxamide substituent, as amino acid derivatives are more prone to establish interactions with alcohol molecules, in our case pentan-1-ol [66,67]. This was also corroborated by the elemental analysis of **4a** and **4c**, which agreed with the presence of two molecules of pentan-1-ol per molecule of phthalocyanine. All the other typical metallophthalocyanine characteristics in terms of 1H-NMR, mass spectrometry and UV–VIS spectroscopy were met, in agreement with the structures.

**Table 3.** Synthesis of zinc (II) metallophthalocyanines **4a**, **4c**, and **4d** and their spectral fundamental/excited state properties, studied in THF.

<sup>a</sup> Relative to unsubstituted ZnPc in DMSO (ΦF = 0.18) [68].

Initial photophysical assessment was carried out for the synthesized metallophthalocyanines. Absorption, emission and fluorescence quantum yields for the phthalocyanines **4a**, **4c**, and **4d** were recorded, using THF as solvent and the results are presented in Table 3.

The electronic absorption spectra of **4a**, **4c**, and **4d**, whose values of molar absorptivity coefficients (ε) are in the typical of range for zinc(II) metallophthalocyanines (Table 3), showed monomeric behavior evidenced by a single and sharp Q band, typical of non-aggregated metallated phthalocyanine complexes, with a maximum at respectively 676, 675 and 676 nm in THF, and a Soret band (the B-band) being observed at around 350 nm, as shown in Table 3 and Figure 2a. The B-bands are broad due to the superimposition of the B1 and B2 bands in the 350 nm region. Moreover, the absorption spectra, Figure 2a, shows that the introduction of the different substituents at the periphery of the phthalocyanine, does not disturb the UV–VIS spectrum, since the absorption bands maximum are similar.

**Figure 2.** UV–VIS spectra of metallophthalocyanines **4a, 4c**, and **4d** in THF (**a**); normalized UV–Vis of studied phthalocyanines with absorption (black solid line) and emission spectra (red dashed line) in THF of: Zn(II)-**4a** (**b**); Zn(II)-**4c** (**c**); Zn(II)-**4d** (**d**). Fluorescence quantum yields (ΦF) of the zinc phthalocyanines **4a** and **4c**–**d,** are presented in Table 3, were determined by the comparative method (Equation (1)) using the unsubstituted Zn phthalocyanine in DMSO as standard (Φ<sup>F</sup> = 0.18) [68], and both the samples and the standard were excited at the same wavelength (640 nm). The Φ<sup>F</sup> were calculated as 0.26, 0.31 and 0.38 for **4a**, **4c** and **4d**, respectively. The Φ<sup>F</sup> value of zinc phthalocyanine complexes functionalized with the amino acid esters **4a** and **4c** have the same order of magnitude (Φ<sup>F</sup> = 0.26–0.31) and are lower than non-biocompatible zinc phthalocyanine **4d** (Φ<sup>F</sup> = 0.38).

The steady-state fluorescence emission spectra of the compounds in THF are shown in Figure 2 and the related data were listed with Stokes shifts in Table 3. Upon excitation at 640 nm, **4a**, **4c**, and **4d** showed fluorescence emission at 685, 686, and 685 nm, respectively. Again, and as expected, the fluorescence emission spectra of all phthalocyanines were similar, as all zinc metal complexes have maximum emission at the same wavelength (λmax = 685–686 nm). It should be noted that the absorption spectra of all phthalocyanines were mirror images of the fluorescent spectra in THF, and that the emission is observed in the region of NIR, a pre-requisite for applications in fluorescence imaging within the important therapeutic window (λ = 650–900 nm) [2,5,69]. The observed Stokes shifts, were within the region ≈9–11 nm are typical of β-substituted phthalocyanines, which is a consequence of the rigidity of the macrocyclic ligand [70].

#### **3. Experimental**

#### *3.1. Materials and Methods*

Commercially available reagents were purchased from Aldrich (Lisbon, Portugal) and Fluorochem (Derbyshire, UK) and used as received. All solvents were pre-dried according to standard laboratory techniques. UV–VIS absorption spectra were recorded on *a Hitachi U-2010* (Hitachi Corporation, Tokyo, Japan) using quartz cells. The molar absorption coefficients were determined using THF as solvent. The fluorescence spectra for the determination of fluorescence quantum yields were acquired on a Spex Fluorolog 3 spectrofluorimeter (Horiba Instruments Incorporated, Edison, NJ, USA). 1H and 13C-NMR

spectra were recorded on a *Bruker Advance III* spectrometer (Bruker, Karlsruhe, Germany) (400.13 for 1H, and 100.61 MHz for 13C). Chemical shifts for 1H and 13C are expressed in ppm, relatively to an internal pattern of TMS. The MALDI-TOF mass spectra were acquired using a Bruker Daltonics Flex Analysis apparatus (Bruker, Madrid, Spain). High-resolution mass spectrometry analysis was carried out with a Bruker Microtof apparatus (Bruker, Madrid, Spain), equipped with selective ESI detector. Elemental analyses were acquired using a FISONS model EA 1108 (Thermo Scientific, Waltham, MA, USA). Column chromatography was performed with silica gel grade 60, 70–230 mesh. 4-Iodophthalonitrile (**1**) was prepared according to the literature procedure [53] starting from 4-nitrophthalonitrile. The nucleophiles (*E*)-1-(4-aminophenyl)-3-(3,4,5-trimethoxyphenyl)prop-2-en-1-one (**2f**) [56] and *N*-BOC-ethylenediamine (**2e**) [59] were prepared as described in the literature.

Fluorescence quantum yields (ΦF) were determined in DMSO using a comparative method with the Equation (1), using unsubstituted zinc(II) phthalocyanine (ZnPc) in DMSO (Φ = 0.18)[68] as standard:

$$\Phi\_{\rm F} = \left. \Phi\_{\rm F} \right. \frac{\rm F \, A\_{\rm Std} \, \eta^2}{\rm F\_{\rm Std} \, A \, \eta\_{\rm Std}^{2}} \tag{1}$$

where F and FStd are the areas under the fluorescence curves of the samples and the standard, respectively; A and AStd are the corresponding absorbances of the samples and standard at the excitation wavelengths, respectively; η<sup>2</sup> and η<sup>2</sup> Std are the refractive indices of solvents used for the sample and standard, respectively. The absorbance of the solutions at the excitation wavelength was around 0.1.

#### *3.2. General Procedure for Synthesis of CARBOXAMIDE Substituted Phthalonitriles* **3a**–**g**

In a typical aminocarbonylation reaction, the catalyst precursor Pd(OAc)2, triphenylphosphine (PPh3) ligand, substrate 4-iodophthalonitrile and the nucleophile were directly introduced in a high pressure reactor having a magnetic stirrer inside. The reactor was sealed and three vacuum/CO gas cycles were performed. Under vacuum, the reaction solvent was then added (toluene) via cannula, followed by triethylamine as base. The reactor was then pressurized using 5 bar CO and the reaction mixture maintained at 100 ◦C for the required period of time. After this period, the reactor was cooled to room temperature and depressurized. Palladium particles were filtered, the solvent rotary evaporated, and the crude product was then purified according to the corresponding procedure. All new compounds were characterized by means of 1H-, 13C-NMR, and mass spectrometry and presented in ESI.

Methyl 2-(3,4-dicyanobenzamido)acetate (glycine substituted phthalonitrile) (**3a**). Following the above described procedure, 6.75 mg (0.030 mmol) of Pd(OAc)2, 15.74 mg (0.060 mmol) of PPh3, 300 mg (1.18 mmol) of 4-iodophthalonitrile, 163.4 mg (1.30 mmol) glycine methyl ester hydrochloride (**2a**), and 1.1 mL Et3N were dissolved in 10 mL of toluene. The reaction was pressurized and maintained at 100 ◦C for 12 h. The residue was dissolved in dichloromethane (20 mL), washed with brine (3 × 20 mL) and water (3 × 20 mL). The organic phase was dried with sodium sulfate and the solvent evaporated. The product was purified by recrystallization with ethyl acetate/*n*-hexane yielding **3a** in 65% yield (158 mg). 1H-NMR (400.13 MHz, CDCl3) *δ* 8.26 (s, 1H), 8.15 (d, *J* = 8.1 Hz, 1H), 7.93 (d, *J* = 8.1 Hz, 1H), 6.77 (s, 1H), 4.27 (d, *J* = 4.9 Hz, 2H), 3.84 (s, 9H). 13C-NMR (100.61 MHz, CDCl3) *δ* 170.0, 163.7, 138.4, 134.1, 132.4, 131.6, 118.6, 116.9, 114.8, 114.7, 53.0, 42.1. HRMS (ESI-TOF) *m*/*z* calcd for [M + Na]+: C12H9N3NaO3 266.0536; found 266.0532.

(S)-Methyl 2-(3,4-dicyanobenzamido)-4-methylpentanoate (leucine substituted phthalonitrile) (**3b**). Following the above described procedure, 6.75 mg (0.030 mmol) of Pd(OAc)2, 15.74 mg (0.06 mmol) of PPh3, 300 mg (1.18 mmol) of 4-iodophthalonitrile, 236.6 mg (1.30 mmol) leucine methyl ester hydrochloride (**2b**) and 1.1 mL Et3N were dissolved in 10 mL toluene. The reaction was pressurized and maintained at 100 ◦C for 12 h. The residue was dissolved in dichloromethane (20 mL), washed with brine (3 × 20 mL) and water (3 × 20 mL).The organic phase was dried with sodium sulfate and the solvent evaporated. The product (**3b**) was purified by column chromatography on

silica gel (stationary phase) first using chloroform and then a mixture of chloroform/ethyl acetate (20/1) and obtained in 54% yield (120 mg), after being washed with *n*-hexane. 1H-NMR (400.13 MHz, CDCl3) *δ* 8.24 (d, *J* = 8.1 Hz, 1H), 8.14 (dd, *J* = 1.7 Hz*, J* = 8.1 Hz, 1H), 7.91 (d, *J* = 8.1 Hz, 1H), 6.77 (br s, 1H), 4.86–4.80 (m, 1H), 3.79 (s, 3H), 1.80–1.66 (2 m, 3H), 1.00–0.97 (m, 6H). 13C-NMR (100.61 MHz, CDCl3) *δ* 173.4, 163.5, 138.6, 134.1, 132.4, 131.9, 118.3, 116.6, 114.9, 52.9, 51.8, 41.6, 25.1, 22.9, 22.0. HRMS (ESI-TOF) *m*/*z* calcd for [M + Na]+: C16H17N3NaO3 322.1162; found 322.1153.

(S)-Methyl 2-(3,4-dicyanobenzamido)-3-phenylpropanoate (phenyl alanine substituted phthalonitrile) (**3c**). Following the above described procedure, 6.75 mg (0.030 mmol) of Pd(OAc)2, 15.74 mg (0.06 mmol) of PPh3, 300 mg (1.18 mmol) of 4-iodophthalonitrile, 280.4 mg (1.30 mmol) phenyl alanine methyl ester hydrochloride (**2c**) and 1.1 mL Et3N were dissolved in 10 mL toluene. The reaction was pressurized and maintained at 100 ◦C for 12 h. The residue was dissolved in dichloromethane (20 mL), washed with brine (3 × 20 mL) and water (3 × 20 mL). The organic phase was dried with sodium sulfate and the solvent evaporated. The product (**3c**) was purified by column chromatography on silica gel (stationary phase) first using chloroform and then a mixture of chloroform/ethyl acetate (10/1) and obtained in 59.0% yield (192 mg), after being washed with *n*-hexane. 1H-NMR (400.13 MHz, CDCl3) *δ* 8.12 (d, *J* = 1.7 Hz, 1H), 8.00 (dd, *J* = 8.1, 1.7 Hz, 1H), 7.88 (d, *J* = 8.1 Hz, 1H), 7.30–7.29 (m, 3H), 7.10–7.08 (m, 2H), 6.64 (br s, 1H), 5.09–5.04 (m, 1H), 3.82 (s, 3H), 3.34–3.21 (m, 2H). 13C-NMR (100.61 MHz, CDCl3) *δ* 171.7, 163.3, 138.6, 135.4, 134.1, 132.4, 131.6, 129.3, 128.9, 127.6, 118.4, 116.6, 114.8, 114.8, 54.0, 52.9, 37.7. HRMS (ESI-TOF) *m*/*z* calcd for [M + Na]+: C19H15N3NaO3 356.1003; found 356.1006.

N-Tert-butyl-3,4-dicyanobenzamide (**3d**). Following the above described procedure, 4.4 mg (0.020 mmol) of Pd(OAc)2, 10.5 mg (0.040 mmol) of PPh3, 200 mg (0.79 mmol) of 4-iodophthalonitrile, 0.28 mL (2.6 mmol) of *tert*-butyl amine (**2d**) and 0.8 mL Et3N were dissolved in 6 mL of toluene. The reaction was pressurized and maintained at 100 ◦C for 4 h. The residue was dissolved in dichloromethane (20 mL), washed with brine (3 × 20 mL) and water (3 × 20 mL). The organic phase was dried with sodium sulfate and the solvent evaporated. The product (**3d**) was purified by column chromatography on silica gel (stationary phase) using a mixture of dichloromethane/ethyl acetate (20/1) and obtained in 74% yield (132.9 mg). 1H-NMR (400.13 MHz, CDCl3) *δ* 8.14 (bs, 1H), 8.06 (d, *J* = 8.1 Hz, 1H), 7.88 (d, *J* = 8.1 Hz, 1H), 5.94 (br s, 1H), 1.49 (s, 9H). 13C-NMR (100.61 MHz, CDCl3) *δ* 163.1, 140.6, 134.0, 132.1, 131.6, 117.8, 116.4, 115.0, 53.0, 28.8. HRMS (EI) *m*/*z* calcd for [M]+: C13H13N3O 227.1059; found: 227.1060.

N-BOC-Ethylenediamine-3,4-dicyanobenzamide (**3e**). Following the above described procedure, 4.4 mg (0.020 mmol) of Pd(OAc)2, 10.5 mg (0.040 mmol) of PPh3, 200 mg (0.79 mmol) of 4-iodophthalonitrile, 151 mg (0.94 mmol) of *N*-BOC-ethylenediamine (**2e**) and 0.8 mL Et3N were dissolved in 6 mL toluene. The reaction was pressurized and maintained at 100 ◦C for 3 h. The product (**3e**) precipitated in the middle of the reaction and then was washed with *n*-hexane and obtained in 80% yield (198.5 mg). 1H-NMR (400.13 MHz, CDCl3) *δ* 8.31 (sl, 1H), 8.21 (dd, *J* = 8.1, 1.1 Hz, 1H), 8.17 (br s, 1H), 7.89 (d, *J* = 8.1 Hz, 1H), 5.18 (br s, 1H), 3.56–3.53 (m, 2H), 3.43–3.41 (m, 2H), 1.43 (s, 9H). 13C-NMR (100.61 MHz, CDCl3) *δ* 163.6, 158.7, 138.9, 133.9, 132.4, 131.7, 117.9, 116.4, 115.0, 114.9, 80.9, 43.6, 39.7, 28.4. HRMS (ESI-TOF) *m*/*z* calcd for [M + H]+: C16H18N4NaO3 337.1271; found 337.1271.

(E)-3,4-Dicyano-N-(4-(3-(3,4,5-trimethoxyphenyl)acryloyl)phenyl)benzamide (**3f**). Following the above described procedure, 4.4 mg (0.020 mmol) of Pd(OAc)2, 10.5 mg (0.040 mmol) of PPh3, 200 mg (0.79 mmol) of 4-iodophthalonitrile, 296 mg (0.94 mmol) of (*E*)-1-(4-aminophenyl)-3- (3,4,5-trimethoxyphenyl)prop-2-en-1-one (**2f**), and 0.8 mL Et3N were dissolved in 6 mL toluene. The reaction was pressurized and maintained at 100 ◦C for 25 h. The product (**3f**) precipitated in the middle of reaction and was washed with methanol and cyclohexane and obtained in 70% yield (256 mg). 1H-NMR (400.13 MHz, acetone-*d*6) *δ* 10.23 (s, 1H), 8.64 (d, *J* = 1.5 Hz, 1H), 8.51 (dd, *J* = 8.2, 1.6 Hz, 1H), 8.26 (d, *J* = 8.2 Hz, 1H), 8.18 (d, *J* = 8.7 Hz, 2H), 8.01 (d, *J* = 8.7 Hz, 2H), 7.84 (d, *J* = 15.5 Hz, 1H), 7.73 (d, *J* = 15.5 Hz, 1H), 7.19 (s, 2H), 3.91 (s, 6H), 3.79 (s, 3H). 13C-NMR (100.61 MHz, acetone-*d*6) *δ*188.7, 163.9,

154.9, 145.3, 143.8, 141.8, 140.6, 135.4, 133.9, 133.8, 131.8, 130.7, 122.1, 120.8, 120.7, 118.9, 117.0, 116.3, 107.4, 60.9, 56.8*.* HRMS (ESI-TOF) *m*/*z* calcd for [M]+: C27H21N3O5 468.1554; found 468.1555.

N-Piperazine-3,4-dicyanobenzamide (**3g**). Following the above described procedure, 4.4 mg (0.020 mmol) of Pd(OAc)2, 10.5 mg (0.040 mmol) of PPh3, 200 mg (0.79 mmol) of 4-iodophthalonitrile, 407 mg (4.72 mmol) piperazine (**2g**), and 0.8 mL Et3N were dissolved in 6 mL toluene. The reaction was pressurized and maintained at 100 ◦C for 7 h. The product (**3g**) was purified by column chromatography on silica gel (stationary phase) using ethanol as eluent and obtained in 77% yield (146 mg). 1H-NMR (400.13 MHz, CDCl3) *δ* 7.87 (d, *J* = 8.0 Hz, 1H), 7.83 (d, *J* = 1.3 Hz, 1H), 7.75 (dd, *J* = 8.0, 1.3 Hz, 1H), 3.75 (s, 2H), 3.32 (s, 2H), 2.90 (d, *J* = 49.8 Hz, 4H). 13C-NMR (100.61 MHz, CDCl3) *δ* 166.1, 141.1, 134.0, 132.2, 131.7, 116.8, 116.7, 114.9, 114.8, 49.01, 46.6, 45.9, 43.6. HRMS (ESI-TOF) *m*/*z* calcd for [M + H]+: C13H13N4O 241.1084; found 241.1081.

#### *3.3. General Procedure for Synthesis of Carboxamide Substituted Phthalocyanines*

In a typical experiment, the desired phthalonitrile and Zn(OAc)2·2H2O were dissolved in high boiling solvent pentan-1-ol and the mixture heated to reflux temperature for the required time for total consumption of the substrate (checked by TLC) under nitrogen atmosphere. After distilling off most of the solvent, the mixture was cooled to room temperature, and *n*-hexane was added to precipitate the crude compound. The solid was filtered, washed with water and purified according to the corresponding procedure. All new compounds were characterized by means of 1H-NMR, UV–VIS, fluorescence and mass spectrometry and presented in ESI.

2(3)-Tetra-(keto-N-glycinyl) phthalocyaninato zinc(II) (**4a**). Following the procedure described above, 100 mg of phthalonitrile **3a** (0.41 mmol) and 29.7 mg Zn(OAc)2·2H2O (0.14 mmol) were dissolved in 1 mL of pentan-1-ol. The mixture was heated to 140 ◦C and stirred for 20 h. After workup procedure, the zinc(II) phthalocyanine complex **4a** was purified by column chromatography on silica gel first using dichloromethane/ethyl acetate (1/1) and then a mixture of dichloromethane/ethanol (20/1) as eluent to obtain 62 mg of **4a** (58% yield), as a waxy dark blue solid. UV–VIS (THF) *λ*max (log *ε*) 350 (4.52), 611 (4.22), 676 (4.87). 1H-NMR (400.13 MHz, acetone-*d*6, 30 ◦C) *δ* 8.46 (br s, 4H), 8.34 (d, *J* = 7.7 Hz, 4H), 8.29 (s, 4H), 7.94 (d, *J* = 7.7 Hz, 4H), 4.04–3.94 (m, 8H), 2.76 (s, 12H). MS (MALDI-TOF-INFUSION) *m*/*z* calcd for [M + Li]+: C48H36N12O12LiZn 1043.2023; found 1043.2050. EA calcd for C48H36N12O12Zn·2C5H12O·2H2O C, 55.70; H, 5.16; N, 13.44; found C, 55.55; H, 5.35; N, 13.50.

(*S,S,S,S*)-2(3)-Tetra-(keto-*N*-phenyl alaninyl) phthalocyaninato zinc(II) (**4c**). Following the procedure described above, 45 mg of phthalonitrile **3c** (0.14 mmol) and 11 mg Zn(OAc)2·2H2O (0.05 mmol) were dissolved in 0.5 mL of pentan-1-ol. The mixture was heated to 140 ◦C and stirred for 20 h. After workup procedure, the zinc(II) phthalocyanine complex **4c** was purified by column chromatography on silica gel first using dichloromethane/ethyl acetate (5/1) and then a mixture of dichloromethane/ethanol (20/1) as eluent to obtain 32 mg of **4c** (65% yield), as a waxy dark blue solid. UV–VIS (THF) *λ*max (log *ε*) 350 (4.14), 610 (3.81), 675 (4.48). 1H-NMR (400.13 MHz, acetone-*d*6) *δ* 8.34 (br s, 4H), 8.24 (d, *J* = 7.8 Hz 4H), 8.20 (s, 4H), 7.89 (d, *J* = 7.7 Hz, 4H), 7.27 (2m, 20H), 4.93 (m, 4H), 3.31 (m, 4H), 3.19 (m, 4H), 2.86 (s, 12H). MS (ESI-TOF-INFUSION) *m*/*z* calcd for [M]+: C76H60N12O12Zn 1396.3745; found 1396.3754. EA calcd for C76H60N12O12Zn·2C5H12O·H2O C, 64.84; H, 5.44; N, 10.55; found C, 64.59; H, 5.75; N, 10.83.

2(3)-Tetra-(tert-butyl-carboxamidyl) phthalocyaninato zinc(II) (**4d**). Following the procedure described above, 100 mg of phthalonitrile **3d** (0.44 mmol) and 32.9 mg Zn(OAc)2·2H2O (0.15 mmol) were dissolved in 0.5 mL of pentan-1-ol. The mixture was heated to 140 ◦C and stirred for 20 h. After workup procedure, the zinc(II) phthalocyanine complex (**4d**) was purified by column chromatography on silica gel using a mixture dichloromethane/methanol (20/1) as eluent to obtain 74 mg of (**4d**) (68% yield) as a dark blue solid, after recrystallization from methanol. UV–VIS (THF) *λ*max (log *ε*) 351 (4.41), 610 (4.49), 676 (5.10). 1H-NMR (400.13 MHz, acetone-*d*6) *δ* 8.26–8.20 (br s, 8H), 7.88–7.86 (br s, 4H), 7.58 (s, 4H), 1.48 (sl, 36H). MS (MALDI-TOF) *m*/*z* calcd for [M]+: C52H52N12O4Zn 972.3; found 972.3; [M + Na]+, *<sup>m</sup>*/*z*: 995.3. EA calcd for C52H52N12O4Zn·2H2O C, 61.81; H, 5.59; N, 16.63; found C, 62.06; H, 5.50; N, 16.43.

#### **4. Conclusions**

In conclusion, we have established a straightforward methodology to prepare carboxamide substituted phthalonitriles, using the well-known palladium-catalyzed aminocarbonylation of aryl halides in the presence of carbon monoxide (CO) to our advantage. In virtue of this direct modification of phthalonitriles, a more accessible preparation of biocompatible phthalocyanines is, hence, achieved. Current efforts are being devoted to extending the methodology to other phthalonitriles and phthalocyanines thereof. Initial assessment of the photophysical properties led us to conclude that this type of phthalocyanines may be usable in medicinal applications, namely optical fluorescence imaging, given the high fluorescence quantum yields (Φ<sup>F</sup> = 0.31 for biocompatible amino acid ester substituted phthalocyanine **4c**) and acceptable Stokes shifts.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4344/8/10/480/s1, 1. Experimental procedures for the synthesis of phthalonitriles **3a–g** and copies of 1H, 13C NMR and Mass Spectra, 2. Experimental procedures for the synthesis of phthalocyanines and copies of 1H NMR and Mass Spectra.

**Author Contributions:** Conceptualization: M.J.F.C. and M.M.P.; investigation: V.A.T., M.J.F.C., C.S.V. and R.T.A.; methodology: M.M.P.; supervision: M.J.F.C.; validation: V.A.T., C.S.V. and R.T.A.; writing—original draft: V.A.T., M.J.F.C. and M.M.P.; writing—review and editing: M.J.F.C. and M.M.P.

**Funding:** This research was funded by FCT-Portugal (Portuguese Foundation for Science and Technology) and FEDER (ERDF)—European Regional Development Fund through the COMPETE Programme (Operational Programme for Competitiveness) with grants PEst-OE/QUI/UI0313/2014, to PT2020 POCI-01-0145-FEDER-027996, PTDC/QEQ-MED/3521/2014 and PTDC/QUI-OUT/27996/2017. M.J.F.C. post-doctoral grant SFRH/BPD/99698/2014 funded by FCT-Portugal. C.S.V., V.A.T. and R.T.A. PhD grants (PD/BD/128317/2017, PD/BD/128318/2017 and PD/BI/135341/2017, respectively) funded by FCT-Portugal. APC waived by MDPI.

**Conflicts of Interest:** There are no conflicts to declare.

#### **References and Notes**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
