Adsorption of heterocyclic sulfur and nitrogen compounds in liquid hydrocarbons on activated carbons modified by oxidation: capacity, selectivity and mechanism

Abstract

In order to understand the adsorption mechanism for removing heterocyclic sulfur/nitrogen compounds from liquid hydrocarbons on activated carbon, the liquid-phase adsorption of quinoline, indole, benzothiophene, dibenzothiophene as well as naphthalene and fluorene on four activated carbons were conducted in both batch and fixed-bed adsorption systems by using model fuels. The four activated carbons with different surface chemistries were prepared by oxidation with ammonium persulfate (APS) and heat treatment, and their oxygen-containing functional groups (OCFGs) on the surface were characterized by the temperature-programmed desorption (TPD). The influences of surface OCFGs and co-existing compounds on the adsorptive removal of heterocyclic sulfur/nitrogen compounds were examined. The results showed that the OCFGs play a decisive role in the adsorptive removal of nitrogen compounds. For the carbon adsorbents with no OCFGs, the adsorption is conducted dominantly through the π–π stacking interaction and the affinity of adsorbates is dependent on the size of their π systems. For the carbon adsorbent with bound OCFGs, the adsorption is dominantly based on the acid–base interaction and the π–H interaction, resulting in the much higher capacity and selectivity for nitrogen compounds. The carbon adsorbent with phenol, carboxylic anhydride, and lactone groups, but no carboxylic groups gives the highest capacity and selectivity for indole, probably through the H-bond interaction. The effect of the co-existing compounds depends on their adsorption mechanism and affinity. Careful design and modification of carbon adsorbents are able to selectively remove the undesired compounds from the liquid hydrocarbons.

---

1. Introduction

Removal of nitrogen and sulfur-containing heterocyclic aromatic contaminants from liquid hydrocarbon streams has become a hot research topic due to stringent environmental regulations.^1,2 The EU and US have limited the sulfur content in diesel fuel to less than 10 and 15 ppm, respectively. On the other hand, due to the decline in crude oil quality and a large demand for transportation fuels, the refineries have to deal with heavier feedstock and other refinery hydrocarbon streams with much higher sulfur and nitrogen concentrations. The removal of the sulfur and nitrogen compounds is important to reduce the emission of SO_X and NO_X during the combustion of the fuels, as these species are responsible for haze, acid rain, and global warming.³ From a refining point of view, the conventional hydrodesulfurization (HDS) process on solid Co–Mo/Ni–Mo catalysts is capable of meeting the increasing restrictions of the sulfur content. However, the HDS process needs to be conducted at the severe conditions with high hydrogen consumption, resulting in high operation cost. It was found that the heterocyclic nitrogen compounds present in fuel, even in very low concentration, could strongly inhibit the ultra-deep HDS, especially for removal of the refractory sulfur compounds, by competitive adsorption or even poisoning the catalyst.^1,3,4 As a result, removal of the nitrogen compounds prior to HDS is critical in ultra-deep HDS.⁵

Many methods for removing the sulfur and nitrogen compounds from hydrocarbon streams have been investigated and reported in literature.⁶ Among them, the adsorption has been considered as one of the most competitive methods, especially when the S and N concentrations in the hydrocarbon streams are low. Recently, Laredo et al. have given a good review paper on adsorptive denitrogenation of middle distillates over various adsorbent materials.⁷ Several adsorbents have been reported over the years for effective removing the S and N compounds, such as the Cu⁺Y, MOFs, metal oxides, zeolites, and activated carbons.^5–12 A number of studies by using activated carbon as an adsorbent for removing the S and N compounds from both model fuels and real fuels have been reported in literature, as activated carbon has large surface area, well developed internal pore structure, rich surface functional groups and wide availability.^5,13–15 In addition, easy modification and tailoring of the functional groups on activated carbon also leads to the selective adsorption of some special guest molecules.^16,17 However, due to the complexity of the heterogeneous surface and well-developed pore structure of activated carbon, many mechanisms, even contradictory mechanisms, were proposed by different researchers.^5,18–22 Mochida and co-workers used activated carbon to remove sulfur and nitrogen compounds in real gas oil. They reported that the CO-releasing oxygen groups decomposing at 600–800 °C might be the most important factor for adsorptive denitrogenation (ADN), which as the CO₂-releasing groups seemed to inhibit the ADN. The surface area might be the keys for adsorption of the refractory S compounds.⁵ On the contrary, Seredych et al. reported that the adsorption of DBT and 4,6-DMDBT was mainly governed by the volume of micropores.¹⁸ Almarri et al. studied the adsorption capacity and selectivity of seven representative activated carbons for removing the S and N compounds, and found that the oxygen groups may play a more important role in ADN than the micropore surface area. They also highlighted the different oxygen groups might be crucial in determining their selectivity for different nitrogen and sulfur compounds.²⁰ Consequently, it is still unclear how the surface chemistry of carbon adsorbents and the co-existing compounds affects the adsorptive removal of heterocyclic sulfur/nitrogen compounds from liquid hydrocarbon streams, due to the absence of deep and systematic study and comparison of the one-component and multi-component adsorptions on activated carbons with different chemical properties on their surface.

In this work, the aim is to clarify how the surface chemistry of the activated carbons and the co-existing compounds affect the selectively adsorptive removal of heterocyclic sulfur and nitrogen compounds from liquid hydrocarbon streams. The adsorptive removal of sulfur and nitrogen compounds as well as polyaromatics from model fuels was conducted on four activated carbons with different surface chemistry, which were prepared by ammonium persulfate (APS) oxidation and heat treatment of a commercial activated carbon to control the oxygen-containing functional groups (OCFGs) on their surface. The measured adsorption capacity and selectivity of the carbon adsorbents for various compounds were correlated to their surface chemical properties to build the relationship between the adsorption performance and the surface properties, and thus clarify the adsorption mechanism of different carbon adsorbents for different adsorbates. Furthermore, effects of the coexisting compounds and the solvents in the fuel on the adsorption performance were also examined and discussed.

2. Experimental section

2.1. Activated carbons and modification

A commercial activated carbon, denoted as AC, in which the ash content is less than 0.07 wt%, was used as an original carbon-based material. AC was washed with the distilled water and dried in a vacuum oven at 120 °C for 24 h for removing the absorbed contaminants before use. In order to introduce OCFGs on the surface, AC was oxidized by using an APS solution with APS concentration of 1.5 mol l⁻¹ at 60 °C for 3 h according to our previous papers,^23,24 then the oxidized AC was separated by filtering from the solution and dried at 120 °C under vacuum overnight. The obtained AC is denoted as OAC. A part of OAC was further heated to 450 °C or 650 °C, respectively, in a tube furnace under a nitrogen flow at a rate of 10 °C min⁻¹, and held at the target temperature for 120 min. These prepared adsorbents were denoted as OAC-450 and OAC-650, respectively, according to the final treatment temperature. APS used in this study was purchased from Sigma-Aldrich with purity over 98%.

2.2. Model fuels

In order to get the one-component adsorption isotherms for each compound, including naphthalene (Nap), fluorene (Flu), quinoline (Qui), indole (Ind), benzothiophene (BT) and dibenzothiophene (DBT), a set of the one-component solutions in heptane or toluene, respectively, were prepared. The structures and some properties of the compounds are shown in Table S1.†^19,25,26

A model fuel (MF) that contains six components in heptane was also prepared and used for flow adsorption test to examine the adsorption performance of the different carbon adsorbents for them. MF contained almost the same molar concentrations of Nap, Flu, Qui, Ind, BT and DBT, as shown in Table S2.† All of the aromatic, sulfur and nitrogen compounds were purchased from Sigma-Aldrich, and heptane with a purity of 97% was obtained from Sionpharm.

2.3. Adsorption test

For the one-component adsorption in a stirred batch system, about 0.05 g of the adsorbent and 5 g of the solution were added into a 40 ml glass vials with stirring at 25 °C for 3 h. After the adsorption time was reached, the mixture was separated by centrifugation and the solution was analyzed to estimate the adsorption capacity using the following equation:where q_e is the equilibrium adsorption capacity (mmol g⁻¹); C_o and C_e is the initial and equilibrium concentrations (mmol kg⁻¹), respectively in the solution; w is the weight of the solution (kg); and W is weight (g) of the adsorbent.

For the multi-component adsorption with MF, about 0.7–0.9 g adsorbent was loaded into a stainless steel column (diameter: 4.6 mm; length: 150 mm), and MF was fed into the column at 25 °C using a HPLC pump with a liquid hourly speed velocity of 4.8 h⁻¹. The treated MF was regularly sampled every 5–10 min for analysis.

2.4. Characterization of activated carbons

The textural characterization of the activated carbons was conducted by using ASAP 2020 surface area and porosimetry analyzer (Micromeritics). The adsorption of N₂ was performed at −196 °C. The surface area, pore volume and pore size data were calculated according to Brunauer–Emmett–Teller (BET), Dubinin–Radushkevitch (DR) and density functional theory (DFT) methods.²⁴

Elemental analysis of all adsorbents was obtained by using Vario EL-III equipment. The activated carbons were also characterized by the temperature-programmed desorption (TPD, Chem BET plus Quantachrome) with a mass spectrometer (Dycor, Model 2000). The total oxygen content was calculated on the basis of the released CO and CO₂. In addition, the spectra deconvolution method that was developed in our previous study on the basis of the modification of the method reported by Figueiredo and co-workers was used for identification and quantification of the OCFGs.^23,27,28

2.5. Analysis of treated fuels

The treated fuels were analyzed by GC 2010 plus gas chromatography (Shimadzu) with a low-polarity capillary column (Rtx®-5, 30 m × 0.32 mm × 0.25 μm, RESTEK) and a flame ionization detector (FID), using an external standard method for quantification. The temperature of injector and detector was kept at 280 and 300 °C, respectively. The oven temperature was initially set at 80 °C, ramped to 130 °C by 10 °C min⁻¹, and held for 3 min, and then was further ramped to 280 °C by 20 °C min⁻¹ and held at it for 2 min.

3. Results and discussion

3.1. Characterization of activated carbons

In order to examine the effect of carbon surface chemistry on the adsorption of different activated carbons, the four adsorbents, AC, OAC, OAC-450 and OAC-650, with different surface chemistry were prepared, and their textural and surface properties were characterized.

3.1.1. Textural properties of activated carbons. The nitrogen adsorption–desorption isotherms of AC, OAC, OAC-450 and OAC-650 were obtained, and the results are shown in Fig. S1a† and Table 1. The N₂ adsorption capacity on the activated carbon increased in the order of OAC-650 ≈ OAC-450 < OAC < AC. The carbon adsorbents after modification show significant reduction in the specific surface area (S_BET) in comparison with the initial one (AC). This may be blamed on the damage of the partial porous structure due to the over oxidation.²³ However, all carbon materials show similar pore size and pore size distribution (as shown in Fig. S1b†), indicating that the prepared adsorbents are good activated carbon samples for examining the effect of their surface functional groups on the adsorption performance, regardless of effect of pore size distribution.

Table 1 Pore properties of original and modified activated carbons

Sample	SBET (m^2 g^−1)	SMIC (m^2 g^−1)	Vtotal (cm^3 g^−1)	VMIC (cm^3 g^−1)	VMIC/Vtotal (%)	Pore size (nm)
AC	2051.5	1057.3	0.99	0.44	44	1.94
OAC	1048.7	675.5	0.50	0.27	54	1.88
OAC-450	840.2	464.5	0.41	0.19	46	1.93
OAC-650	810.9	431.9	0.40	0.18	45	1.94

---

3.1.2. Measurement of OCFGs on activated carbons. As is well-known, the OCFGs on the activated carbon surface play a very important role in many adsorption processes, especially in liquid phase adsorption. TPD characterization and elemental analysis of the four activated carbons were conducted to get the total oxygen content and the information of OCFGs on the adsorbents. Fig. S2† shows TPD-CO and TPD-CO₂ profiles of different activated carbons. The CO and CO₂ signals are significantly different, which are related to the decomposition of the surface OCFGs within the temperature program.⁸ The total oxygen concentration of the activated carbons increased in the order of AC < OAC-650 < OAC-450 < OAC, which is different from the order of their surface area. It is clear that the oxidative modification introduced abundant OCFGs onto the surface, while the heat treatment modification at different temperature under N₂ made the oxygen concentration declined to different degrees. The total oxygen concentrations of the activated carbons, which were measured by the elemental analysis are also listed in Table 2, which are agreement with the results from TPD-CO and TPD-CO₂ analyses.

Table 2 Element analysis of activated carbons

Sample	C (wt%)	H (wt%)	N (wt%)	O^a (wt%)	O-TPD^b (wt%)
a Oxygen contents were calculated by the differences.b Oxygen contents were estimated by the evolved CO and CO2 amount through TPD analysis.
AC	91.9	0.0	0.0	7.5	0.03
OAC	69.4	3.4	0.0	27.5	27.3
OAC-450	77.4	1.9	0.0	19.9	15.5
OAC-650	91.5	0.0	0.0	8.3	3.8

---

In order to identify and quantify the various OCFGs on different activated carbons, the CO and CO₂ profiles were further deconvoluted according to the modified Figueiredo's method,²⁴ and the results are shown in Fig. 1. The estimated concentrations of various OCFGs on the different activated carbons are listed in Table 3. There were almost no OCFGs on the surface of AC. The oxidative modification by the APS solution introduced abundant OCFGs onto the surface of OAC, including strong carboxyl (SA, 1.18 mmol g⁻¹), weak carboxyl (WA, 1.47 mmol g⁻¹), carboxylic anhydrides (CA, 2.97 mmol g⁻¹), phenol (PH, 2.12 mmol g⁻¹), lactone (LC + LD, 1.08 mmol g⁻¹), and carbonyl or quinone groups (CQ1 + CQ2, 2.24 mmol g⁻¹). The concentration of the total acidic functional groups on OAC, including SA and WA, is around 2.65 mmol g⁻¹. The heat treatment at 450 °C under N₂ removed almost all of carboxyl groups (SA + WA) and most of CA, but kept others (PH, LD, LC, CQ1 and CQ2). The heat treatment at 650 °C removed all of SA, WA, CA, LC, and LD, and only kept a part of PH, and majority of CQ1 and CQ2. The results indicate that the activated carbons have quite different distribution of OCFGs on the surface, which provides an excellent base for examining the effect of OCFGs on their adsorption performance.

Fig. 1 Deconvolution of TPD-CO and TPD-CO₂ profiles of different activated carbons.

Table 3 Concentrations of various functional groups on different activated carbons, which were determined by the deconvolution of the TPD-CO, TPD-CO₂ (mmol g⁻¹)^a

Sample	SA	WA	CA	PH	LD	LC	CQ#1	CQ#2
a SA: strongly acidic carboxyls; WA: weakly acidic carboxyls; CA: carboxylic anhydrides; PH: phenols; LD and LC: lactones; CQ#1 and CQ#2: carbonyls or quinones.
AC	—	—	—	—	—	—	—	—
OAC	1.18	1.47	2.97	2.12	0.59	0.49	1.35	0.89
OAC-450	—	—	0.69	2.61	0.25	0.33	1.92	0.65
OAC-650	—	—	—	0.14	—	—	0.83	0.97

---

3.2. Effect of oxidation modification on adsorption of individual compounds

3.2.1. Effect of oxidation modification on adsorption of aromatic compounds. The one-component adsorption of the two-ring and three-ring aromatics, Nap and Flu, on the different activated carbons (AC OAC, OAC-450 and OAC-650) was conducted with the solutions containing different concentration of Nap or Flu in the batch adsorption system at 25 °C for 3 h. The adsorption isotherms of Nap and Flu over the four adsorbents are shown in Fig. 2a and b, respectively. In general, the adsorption capacity increased with increasing equilibrium concentration of the adsorbate in the solution, expressing the type I isotherm on all adsorbents. For Nap, the adsorption capacity of the different adsorbents decreased in the order of OAC > OAC-450 > AC > OAC-650. For Flu, the adsorption capacity decreased in the order similar to that of Nap. OAC gave the highest adsorption capacity for both Nap and Flu, indicating that the OCFGs have a positive effect on adsorption of aromatics, probably through the electron donor–acceptor interaction between the aromatic ring of the adsorbate and the OCFGs on the carbon surface.^25,26,29

Fig. 2 Adsorption isotherms of (a) naphthalene, (b) fluorene, (c) benzothiophene, (d) dibenzothiophene, (e) quinoline, and (f) indole in heptane on activated carbons.

On the other hand, AC and OAC-650 also gave the significant adsorption capacity for Nap and Flu, although their capacity was lower than those of OAC and OAC-450. Since there is almost no OCFGs on AC and only a little of PH, CQ1 and CQ2 on OAC-650, it implies that adsorption of the aromatics on the carbon surface may also be through a π–π stacking interaction between the π electrons in the aromatic ring of the adsorbates and the carbon surface. It should be mentioned that the different groups have divers effect on the adsorption of aromatic compounds.^30,31 For example, some researchers believed that the COOH and OH groups attached on the edge of the carbon might inhibit the adsorption for Nap and Flu, since the oxygen atom can bound electrons, and thus reduce the interaction between the aromatics and the π system of the carbon plane. In contrast, the C=O may improve the adsorption for aromatic compounds, which might be through the electron donor–accepter system between the aromatic ring of the adsorbate and the C=O on the carbon surface.³¹

In order to further discuss the adsorption performance, the Langmuir isotherm model was used to treat the liquid-phase adsorption data, as we did before:³²

where K is the adsorption equilibrium constant; q_m is the maximum loading corresponding to complete coverage of the surface by the adsorbate; and C_e is the equilibrium concentration of adsorbate in liquid phase. The estimated parameters with R² value are listed in Table 4. The all R² values are higher than 0.98, indicating that the adsorption isotherms fit well the Langmuir isotherm.

Table 4 Adsorption parameters for different compounds over different activated carbons on the basis of Langmuir isotherms

Adsorbate	Adsorbent	Langmuir
		K (g μg^−1)	qm (mmol g^−1)	R^2
Nap	AC	0.007	1.082	0.991
	OAC	0.009	1.429	0.994
	OAC-450	0.006	1.299	0.995
	OAC-650	0.004	1.116	0.995
Flu	AC	0.039	1.171	0.999
	OAC	0.018	1.488	0.994
	OAC-450	0.025	1.348	0.998
	OAC-650	0.014	1.171	0.998
Qui	AC	0.019	1.296	0.997
	OAC	0.089	2.857	0.998
	OAC-450	0.027	1.253	0.999
	OAC-650	0.010	1.035	0.985
Ind	AC	0.015	1.553	0.998
	OAC	0.074	2.857	0.999
	OAC-450	0.058	1.323	0.999
	OAC-650	0.020	1.099	0.998
BT	AC	0.011	0.952	0.977
	OAC	0.021	1.520	0.997
	OAC-450	0.012	1.152	0.989
	OAC-650	0.025	0.605	0.999
DBT	AC	0.038	2.747	0.999
	OAC	0.062	2.976	0.999
	OAC-450	0.040	2.646	0.999
	OAC-650	0.026	2.232	0.996

---

In comparison of q_m values for Nap and Flu, it was found that q_m values for Nap on the different adsorbents were similar to those for Flu on the different adsorbents, implying that the both Nap and Flu may be adsorbed at the same sites. The q_m values for both Nap and Flu on OAC were higher than those on AC, indicating that the introduction of OCFGs on the surface increased the total number of adsorption sites, although the type of the increased adsorbent sites may be different from that on AC. One the other hand, it was also found that the K values for Flu over the all adsorbents were higher than those for Nap by 2 to 5 times. It means that the interaction between Flu molecule and the adsorption site is stronger than that between Nap molecule and the adsorption site. It can be contributed to that Flu contains higher number (12) of C (sp²) than Nap (10), and also contains additional two no-aromatic hydrogen atoms,^26,33 resulting in the high adsorption capacity for Flu than Nap.

3.2.2. Effect of oxidation modification on adsorption of sulfur compounds. The adsorption isotherms of BT and DBT in heptane over the four activated carbons are shown in Fig. 2c and d. All these adsorptions are also the type I isotherm. The estimated Langmuir adsorptive behavior of these two representative sulfur compounds over all the adsorbents was similar to their adsorption behavior for the corresponding two-ring aromatics (Nap) and three-ring aromatics (Flu). The adsorption capacity of the activated carbons decreased in the order of OAC > OAC-450 > AC > OAC-650 for BT and OAC > OAC-450 ≈ AC > OAC-650 for DBT. The similar order to the adsorption for the aromatics, it indicates that the sulfur compounds may be adsorbed on the same adsorption sites as those for the aromatics through the similar adsorption mechanism.^17,32 For AC with no or less oxygen groups, the π–π stacking interaction might govern their adsorption capacity, while for OAC with rich OCFGs, the electron donor–acceptor interaction may be major.³³ For all carbon adsorbents, the adsorption capacity for DBT is higher than that for BT by more than 2 times, indicating that the size of the adsorbate π-system is important for such adsorption, similar to the cases for Flu and Nap.

However, when comparing the Langmuir adsorption parameters (see Table 4), it clearly shows that the value of the adsorption equilibrium constant (K) for the sulfur compounds is greater than that of the aromatics, indicating that the adsorption affinity between the sulfur compound and the adsorption site is stronger than that between the aromatic compound and the adsorption site.³³ As well knows, for BT and DBT, there are two lone electron pairs on the sulfur atom. One of them is devoted to the π electrons and delocalized into the aromatic ring to create the aromatic system, while the other is the alone pair (no conjugation). This alone-pair electrons may increase the interaction between adsorbate and active sites, especially for the adsorbents with high concentration of OCFGs. However, it is still not clear why the sulfur atom in DBT was found to have significant contribution to the increase of the q_m value for DBT in comparison with Flu, while the sulfur atom in BT to have almost no contribution in increase of the q_m value for BT in comparison with Nap, probably the size of the DBT π system (number of C (sp²) + S/N atoms: 13) is larger than that of Flu (number of C (sp²) + S/N atoms: 12), while the size of BT π system (number of C (sp²) + S/N atoms: 9) is less than that of Nap (number of C (sp²) + S/N atoms: 10), seeing Table S1.†

When comparing the Langmuir adsorption parameters for BT and DBT (see Table 4), both K value and q_m value for DBT are higher than those for BT by more than two times for all adsorbents except K value of OAC-650. It indicates that the adsorbents have higher number of the adsorption sites and stronger interaction between the adsorbate and the adsorption site for DBT than for BT due to the larger aromatic system in the former than the later, resulting in significantly higher adsorption capacity for the former.

3.2.3. Effect of oxidation modification on adsorption of nitrogen compounds. The adsorptive isotherms for Qui and Ind in heptane on the four adsorbents are shown in Fig. 2e and f, and the estimated Langmuir adsorption parameters are also listed in Table 4. In comparison with the adsorption of aromatics and sulfur compounds, OAC showed dramatically high adsorption capacity for both Qui and Ind. The adsorption capacity of OAC for Qui was about 2.65 mmol g⁻¹ at an equilibrium Qui concentration of 10 mmol g⁻¹ (seeing Fig. 2e), which is about 3 times higher than that of AC (0.90 mmol g⁻¹). After the heat treatment at around 450 °C, the capacity of the adsorbent (OAC-450) decreased to 0.90 mmol g⁻¹, which is 66% decrease. Since OAC has surface area and pore distribution similar to those of OAC-450, it is believed that the carboxylic acid groups (SA and WA), which were removed by the heat treatment at 450 °C, is crucial for Qui adsorption.^21,24 As expected, the capacity of OAC-650, on which the CA, PH, LC and LD groups were removed by the heat treatment at 650 °C, decreased further.

In the case for Ind, the adsorption capacity also decreases with the decrease of the OCFGs on the adsorbents. OAC gave the adsorption capacity around 2.6 mmol g⁻¹ (being about 2.6 times of AC) at the equilibrium concentration of 10 mmol kg⁻¹, seeing Fig. 2f. The teat treatment at 450 °C and 650 °C decreased the adsorption capacity to 1.15 and 0.75 mmol g⁻¹, respectively, at the corresponding equilibrium concentration. Since OAC, OAC-450 and OAC-650 have similar surface area and pore size distribution, but quite different OCFGs, which provide a perfect series of adsorbent base for discussing the effect of OCFGs. The results indicate that SA and WA groups on the surface play a key role in the adsorption of Ind, while CA, PH and CQ groups also take the certain responsibility in the adsorption of Ind.

In comparison of the estimated q_m and K values, it is found that the q_m values for Qui on different adsorbents are similar to those for Ind, indicating that the adsorption sites for both Qui and Ind may be similar. When compare the K values, it is interestingly found that the K value for Qui on OAC was 0.089 g μg⁻¹, which is significantly higher than the value (0.074 g μg⁻¹) for Ind over OAC. It implies that the OCFGs on OAC have stronger interaction with Qui than with Ind, as OAC contains abound SA and WA groups which can interact with Qui through the acid–base interaction.³² On the other hand, the K values (0.058 g μg⁻¹) for Ind over OAC-450 is higher than that (0.027 g μg⁻¹) for Qui by more than 2 times, indicating that the adsorption sites on OAC-450 have stronger interaction with Ind than with Qui. Considering that OAC-450 contains no SA and WA, but CA, PH, LD and others, it implies that these remained OCFGs on OAC-450 have stronger affinity with Ind than with Qui, probably through the H-bond interaction between N–H in Ind and the OCFGs.

3.3. Comparison of adsorption capacity of each activated carbon for different adsorbates

In order to facilitate the comparison of adsorption capacity of each activated carbon for different adsorbates, the adsorption isotherms of each carbon adsorbent for different adsorbates are shown in Fig. 3. Both AC and OAC-650 show the similar adsorption selectivity for different adsorbates, as both contain almost no OCFGs on the adsorbent surface. The adsorption capacity for different compounds increased in the order of BT < Nap < Qui < Ind < DBT ≈ Flu. Considering that there is almost no OCFGs on the surfaces of AC and OAC-650, the major interaction is through the π–π stacking interaction between the aromatic π system on the carbon surface and the adsorbates. That is why AC and OAC-650 gave the highest adsorption capacity for three ring aromatic compounds and DBT. For adsorbates of OAC, the adsorption selectivity is significantly different from those for AC and OAC-650. The adsorption capacity increased in the order of BT < Nap < FIu < DBT < Qui ≈ Ind. Introduction of OCFGs onto the surface (such as OAC) increased dramatically the adsorption capacity for Qui and Ind, but just slightly for sulfur and aromatic compounds. It can be ascribed to the high concentration of OCFGs on OAC, which interact strongly with Qui and Ind adsorbates through the acid–base interaction and/or H-bond interaction, but have relatively weaker interaction with the sulfur and aromatic compounds, resulting in the different effects on the adsorption for nitrogen, sulfur and aromatic compounds.³² In summary, the oxidation modification of carbon adsorbents significantly improves the selective removal of the nitrogen-containing compounds from the liquid hydrocarbon stream, but only slightly increases the adsorption capacity for thiophenic compounds (BT and DBT).

Fig. 3 Adsorption isotherms of (a) AC, (b) OAC, (c) OAC-450, and (d) OAC-650 for different adsorbates.

3.4. Effect of solvent on adsorption performance

Presence of major monoaromatics may affect the adsorption capacity and selectivity of the adsorbents for different adsorbates. In order to examine and clarify this effect, the adsorption of various compounds in toluene on different activated carbons was also conducted in comparison with above study in heptane. The adsorption isotherms of Nap, Flu, Qui, Ind, BT and DBT in toluene and heptane, respectively, on different adsorbents are compared in Fig. 4. It shows clearly that the aromatic solvent has a significant effect on the adsorption behavior. The abundant presence of toluene in the model fuels has greatly negative effect on the adsorption for polyaromatics and thiophenic compounds. The adsorption capacity in the presence of toluene was decreased by more than 85%. Since the adsorption of aromatics and thiophenic compounds on the carbon surface is dominantly through the π–π stacking and/or H–π interactions, the competitive adsorption of abound toluene molecules through the same or similar mechanism definitely reduces the adsorption of polyaromatics and thiophenic compounds on the sites, although the π system size of toluene molecule is less than that of polyaromatics (Nap and Flu) and thiophenic compounds (BT and DBT).

Fig. 4 Adsorption isotherms of (a) naphthalene, (b) fluorene, (c) benzothiophene, (d) dibenzothiophene, (e) quinoline, and (f) indole in toluene and heptane solvents on activated carbons.

The solvent effect on the adsorption for nitrogen compounds is quite different to the aromatics and sulfur compounds. The presence of the abound toluene molecules was also found to have a strong effect on the adsorption capacity of AC and OAC-650 for Qui and Ind (seeing Fig. 4e and f), similar to the effect for aromatics and thiophenic compounds, as such adsorption is also dominantly through the π–π stacking and/or H–π interactions due to almost no OCFGs on the surface of AC and OAC-650. However, the presence of the abound toluene molecules also had a significant, but less effect on the adsorption capacity of OAC for nitrogen compounds, reducing the adsorption capacity of OAC for Qui by about 50% and for Ind about 80%, as shown in Fig. 4e and f. Since the adsorption of Qui and Ind on OAC is dominantly through the acid–base and H-bond interactions, the effect of toluene on the adsorption of Qui and Ind on OAC is weaker than those on AC and OAC-650. According to the estimated adsorption equilibrium constants (seeing Table 4), the K value (0.089 g μg⁻¹) for Qui on OAC is significantly higher than that (0.074 g μg⁻¹) for Ind on OAC, indicating the stronger interaction between Qui and the adsorption site than that between Ind and the adsorption site. Consequently, the presence of toluene has less effect on the adsorption for Qui.

On the other hand, the presence of major toluene in the liquid phase also enhances the interaction between the adsorbates and solvent (toluene) molecules due to higher polarity of toluene than that of saturate hydrocarbons (heptane), and thus relatively weakens the interaction between the adsorbates and adsorbents.^29,30 The results show clearly that the presence of major monoaromatics in liquid hydrocarbons has a strong effect on the adsorption of polyaromatics and thiophenic compounds, and a certain effect on adsorption of the nitrogen compounds over the activated carbons. This is probably a reason why many activated-carbon-based adsorbents showed the poorer adsorption performance for removing sulfur and nitrogen compounds from the real hydrocarbon fuels in comparison with those from the model fuels.^5,34

3.5. Effect of co-existing compounds

In order to examine and clarify how the co-existing adsorbates affect each other and change the adsorption capacity of different activated carbons for them, the multi-component adsorption on the activated carbons was conducted in a fixed-bed system at 25 °C with a LHSV of 4.8 h⁻¹ by using the model fuel (MF), composition of which is listed in Table S2.† The breakthrough curves for different compounds over these four adsorbents are shown in Fig. 5. The corresponding breakthrough capacity, saturation capacity, equilibrium capacity and selectivity were calculated on the basis of the breakthrough curves.²⁵ The results are summarized in Table 5.

Fig. 5 Breakthrough curves of different compounds on activated carbons.

Table 5 Adsorption capacities and selectivity of activated carbons on the basis of the test in the fixed-bed flow system (mmol g⁻¹)

Adsorbate	Adsorbent	Breakthrough	Saturate	Net^a	Selectivity-B^b
a Net adsorptive capacity was calculated by subtracting the amount of replaced (kicked off) adsorbate by other coexisting compounds from the saturate capacity.b The relative selectivity at the breakthrough point.
Nap	AC	0.174	0.400	0.005	1.000
	OAC	0.300	0.386	0.022	1.000
	OAC-450	0.315	0.373	0.096	1.000
	OAC-650	0.230	0.277	0.054	1.000
Flu	AC	0.450	0.569	0.238	2.586
	OAC	0.490	0.543	0.086	1.633
	OAC-450	0.462	0.547	0.294	1.467
	OAC-650	0.448	0.473	0.153	1.948
Qui	AC	0.450	0.659	0.659	2.586
	OAC	1.110	1.652	1.652	3.700
	OAC-450	0.527	1.145	1.145	1.673
	OAC-650	0.312	0.435	0.435	1.357
Ind	AC	0.529	0.715	0.715	3.040
	OAC	1.077	1.505	1.505	3.590
	OAC-450	0.772	1.221	1.221	2.451
	OAC-650	0.524	0.880	0.880	2.278
BT	AC	0.157	0.361	0.022	0.902
	OAC	0.300	0.386	0.022	1.000
	OAC-450	0.228	0.335	0.068	0.724
	OAC-650	0.143	0.249	0.026	0.622
DBT	AC	0.675	1.083	1.083	3.879
	OAC	0.621	0.739	0.228	2.070
	OAC-450	0.614	0.743	0.486	1.949
	OAC-650	0.524	0.741	0.741	2.278

---

As shown in Fig. 5a for adsorption on AC, the breakthrough capacity for different compounds increased in the order of BT < Nap < Flu = Qui < Ind < DBT with the breakthrough capacity of 0.157 mmol g⁻¹ for BT, 0.174 mmol g⁻¹ for Nap, 0.450 mmol g⁻¹ for Flu, 0.450 mmol g⁻¹ for Qui, 0.529 mmol g⁻¹ for Ind and 0.675 mmol g⁻¹ for DBT. As AC contain almost no OCFGs on the surface, which benefit the adsorption through the π–π stacking interaction. For adsorption on OAC-650, the breakthrough capacity for various compounds increased in the order of BT < Nap < Qui < Flu < Ind ≈ DBT, which is similar to that on AC.

OAC-450 showed an increasing breakthrough capacity order of BT < Nap < Flu < Qui < DBT < Ind with the highest breakthrough capacity of 0.772 mmol g⁻¹ for Ind. As OAC-450 contains CA, PH and CQ groups on the surface, which are favorable for the adsorption of Ind through the H-bond interaction.

The breakthrough capacity for various compounds on OAC is quite different from others, which increased in the order of BT ≈ Nap < Flu < DBT < Ind ≈ Qui with a breakthrough capacity of 0.300 mmol g⁻¹ for both BT and Nap, 0.490 mmol g⁻¹ for Flu, 0.621 mmol g⁻¹ for DBT, 1.077 mmol g⁻¹ for Ind, and 1.110 mmol g⁻¹ for Qui. Among the adsorbents, OAC gave the highest adsorption capacity for Qui and Ind, which is in agreement with the results from the one-component adsorption.

When carefully examining the breakthrough curves, it was interestingly found that after the breakthrough the relative concentration (C/C₀) of Nap, BT, and Flu at outlet increased sharply to over 1.0, even reaching to 2.5 on OAC. After passing the peak value, the relative concentration returned gradually to 1.0. However, the same phenomenon was not found for Ind and Qui. The similar phenomenon has also reported in our previous studies.^19,33 These interesting results might be explained as follows: the different compounds might be adsorbed, at least partially, on the same adsorption sites, but with different adsorption affinity. The later eluted compounds that have stronger adsorption affinity “kick off” the compounds that have been adsorbed on the same adsorption sites with lower adsorption affinity, leading to a higher concentration of the eluted compounds in the model fuel at outlet than that of the initial one.³³ For the cases on AC and OAC-650, due to almost no OCFGs on the surface, all compounds might be adsorbed at the same, or similar, adsorption sites through the π–π stacking interaction. As a result, DBT with the largest π system size “kicked off” the majority of Nap and BT, and part of Flu, Qui and Ind, as the result shown in Fig. 5a and d. In the case on AC, about 98.8% of the adsorbed Nap, 94.0% of the adsorbed BT and 58.2% of the adsorbed Flu were replaced by the later eluted compounds (seeing Table 5). However, for the case on OAC, since OAC contains abound OCFGs and the adsorption is dominantly conducted through the acid–base and H-bond interactions, Ind and Qui showed the highest adsorption affinity on the sites. As a result, 93.0% of the adsorbed BT, 94.4% of the adsorbed Nap, 84.2% of the adsorbed Flu and 69.1% of the adsorbed DBT were “kicked off” in turn by Ind and Qui, as shown in Fig. 5b and Table 5. The co-existing sulfur and aromatic compounds with the similar concentration have almost no effect on the adsorption of the nitrogen compounds on all activated carbons, as the nitrogen compounds have higher adsorption affinity than others. The results indicate that the different surface chemistry results in the different adsorption mechanism. For the carbon adsorbents with no OCFGs on the surface, the adsorption breakthrough curves showed the different pattern from those with rich OCFGs. This finding allows ones to be able to selectively remove some compounds through certain modification of activate carbon adsorbents.

4. Conclusions

Four representative activated carbons with different surface chemistry from a commercial activated carbon were prepared by oxidation modification and heat treatment. The adsorption performance of these activated carbons for removing nitrogen, sulfur and aromatic compounds were examined in a batch adsorption system and a flowing fixed-bed adsorption system. The APS-oxidation modification increased significantly the amount of the surface OCFGs, which are mainly comprised of carboxylic acid groups, phenol groups and others. The heat treatments at 450 °C and 650 °C can remove selectively OCFGs. The prepared carbon adsorbents provided a good base for examining how the surface chemistry of activated carbons affects the adsorption capacity and selectivity for removing nitrogen, sulfur and aromatic compounds from the liquid hydrocarbon stream. For the carbon adsorbents with no, or almost no, OCFG on the surface, the adsorption is conducted dominantly through the π–π stacking interaction and the adsorption affinity of adsorbates is dependent on the size of their π system, resulting in the high capacity and selectivity for DBT and Flu. For the carbon adsorbents with abundant OCFGs on the surface, the adsorption is dominantly based on the acid–base interaction and the π–H interaction, resulting in the high capacity and selectivity for Qui and Ind. For the carbon adsorbents with phenol, carboxylic anhydrides, lactone, and carbonyl/quinone groups, but no carboxylic groups on the surface, the adsorption may be conducted through both H-bond and π–π stacking interactions, resulting in the highest capacity and selectivity for Ind and DBT. The effect of the co-existing compounds depends on their adsorption mechanism and affinity. The co-existing aromatics, especially the polyaromatics, have great effect on the adsorption through the π–π stacking interaction and significant effect on the adsorption through the π–H interaction, while the co-existing heterocyclic nitrogen compounds show the great effect on the adsorption through the acid–base interaction, H-bond, π–H interaction as well as π–π stacking interaction. The careful design and modification of the carbon adsorbent are able to selectively remove the undesired compounds from the liquid hydrocarbon stream. The oxidative modification of carbon surface can greatly enhance the adsorption capacity and selectivity for removing the nitrogen-containing compounds.

Acknowledgements

The financial support from the National Nature Science Foundation of China (No. 21306121), the Priority Academic Program Development (PAPD) of Jiangsu Higher Education Institutions and the China Postdoctoral Science Foundation (No. 2013M541723) are gratefully acknowledged.

Notes and references

1. A. Stanislaus, A. Marafi and M. S. Rana, Catal. Today, 2010, 153, 1–68.
2. C. Song, Catal. Today, 2003, 86, 211–263.
3. H. Yang, J. Chen, C. Fairbridge, Y. Briker, Y. J. Zhu and Z. Ring, Fuel Process. Technol., 2004, 85, 1415–1429.
4. D. Dong, S. Jeong and F. Massoth, Catal. Today, 1997, 37, 267–275.
5. Y. Sano, K.-H. Choi, Y. Korai and I. Mochida, Energy Fuels, 2004, 18, 644–651.
6. M. F. Ali, A. Al-Malki, B. El-Ali, G. Martinie and M. N. Siddiqui, Fuel, 2006, 85, 1354–1363.
7. G. C. Laredo, P. M. Vega-Merino, F. Trejo-Zárraga and J. Castillo, Fuel Process. Technol., 2013, 106, 21–32.
8. A. J. Hernández-Maldonado and R. T. Yang, Angew. Chem., 2004, 43, 1004–1006.
9. M. Maes, M. Trekels, M. Boulhout, S. Schouteden, F. Vermoortele, L. Alaerts, D. Heurtaux, Y. K. Seo, Y. K. Hwang and J. S. Chang, Angew. Chem., 2011, 123, 4296–4300.
10. L. Wang, B. Sun, F. H. Yang and R. T. Yang, Chem. Eng. Sci., 2012, 73, 208–217.
11. R. T. Yang, A. J. Hernández-Maldonado and F. H. Yang, Science, 2003, 301, 79–81.
12. A. Koriakin, K. M. Ponvel and C.-H. Lee, Chem. Eng. J., 2010, 162, 649–655.
13. Y. Sano, K.-H. Choi, Y. Korai and I. Mochida, Appl. Catal., B, 2004, 49, 219–225.
14. J. Wen, X. Han, H. Lin, Y. Zheng and W. Chu, Chem. Eng. J., 2010, 164, 29–36.
15. Y. Wang and R. T. Yang, Langmuir, 2007, 23, 3825–3831.
16. M. Seredych and T. J. Bandosz, Carbon, 2011, 49, 1216–1224.
17. Z. Jiang, Y. Liu, X. Sun, F. Tian, F. Sun, C. Liang, W. You, C. Han and C. Li, Langmuir, 2003, 19, 731–736.
18. M. Seredych, D. Hulicova-Jurcakova and T. J. Bandosz, Langmuir, 2009, 26, 227–233.
19. J. H. Kim, X. Ma, A. Zhou and C. Song, Catal. Today, 2006, 111, 74–83.
20. M. Lillo-Ródenas, D. Cazorla-Amorós and A. Linares-Solano, Carbon, 2005, 43, 1758–1767.
21. M. Seredych, J. Lison, U. Jans and T. J. Bandosz, Carbon, 2009, 47, 2491–2500.
22. C. O. Ania and T. J. Bandosz, Langmuir, 2005, 21, 7752–7759.
23. N. Li, X. Ma, Q. Zha, K. Kim, Y. Chen and C. Song, Carbon, 2011, 49, 5002–5013.
24. N. Li, J. Zhu, X. Ma, Q. Zha and C. Song, AIChE J., 2013, 59, 1236–1244.
25. M. S. Almarri, PhD thesis, The Pennsylvania State University, 2009.
26. J. Xiao, C. Song, X. Ma and Z. Li, Ind. Eng. Chem. Res., 2012, 51, 3436–3443.
27. J. L. Figueiredo and M. F. R. Pereira, Catal. Today, 2010, 150, 2–7.
28. J. L. Figueiredo, M. F. Pereira, M. M. Freitas and J. J. Órfão, Ind. Eng. Chem. Res., 2007, 46, 4110–4115.
29. H. A. Arafat, F. Ahnert and N. G. Pinto, Sep. Sci. Technol., 2005, 39, 43–62.
30. C. Ania and P. Parra, Adsorpt. Sci. Technol., 2007, 25, 155–167.
31. R. T. Yang, Sep. Sci. Technol., 2003, 280–381.
32. M. Almarri, X. Ma and C. Song, Energy Fuels, 2009, 23, 3940–3947.
33. M. Almarri, X. Ma and C. Song, Ind. Eng. Chem. Res., 2008, 48, 951–960.
34. K. A. Cychosz, A. G. Wong-Foy and A. J. Matzger, J. Am. Chem. Soc., 2008, 130, 6938–6939.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra06108g

---