Template-Derived Mesoporous Carbons with Highly Dispersed

Department of Chemistry, The City College of New York, New York, New York 10031, and The Graduate School of the City University of New York, New York,...
1 downloads 0 Views 306KB Size
Langmuir 2007, 23, 6033-6041

6033

Template-Derived Mesoporous Carbons with Highly Dispersed Transition Metals as Media for the Reactive Adsorption of Dibenzothiophene Mykola Seredych† and Teresa J. Bandosz*,‡ Department of Chemistry, The City College of New York, New York, New York 10031, and The Graduate School of the City UniVersity of New York, New York, New York 10031 ReceiVed NoVember 10, 2006. In Final Form: March 14, 2007 Mesoporous carbons with highly dispersed copper, cobalt, and iron were obtained from an organic polymer within amorphous silica powder, alumina, and zeolite 13X. The materials were characterized using the adsorption of nitrogen, potentiometric titration, and elemental analysis. The small metal content (less than 1%) and its chelation in the precursor polymers ensure a high dispersion of metallic centers. The materials obtained are mainly mesoporous but differ significantly in their porosity and surface chemistry, which is linked to the effect of template constraints and chemistry and the kind of metal and is related to the differences in the carbonization mechanism. On the carbon obtained, the adsorption of dibenzothiophene (DBT) from hexane was carried out. The high capacities (up to 130 mg S/g) obtained were linked to the high volume of mesopores and specific interactions of DBT with surface acidic groups and strong interactions of metals with dibenzothiophene via S-M σ bonds or, in the case of copper, via interaction of metals with disturbed π electrons of aromatic rings of DBT.

Introduction Recently, great interest has been shown in the application of adsorption for the desulfurization of liquid fuels.1-19 The new requirements implemented in 2006 mandate the reduction in the sulfur level for gasoline and diesel fuel to 30 and 15 ppm, respectively.20 Besides detrimental environmental effects of sulfur compounds, they also poison both catalytic conversion automobile catalysts and fuel cell reformer catalysts. The majority of sulfur compounds (thioles and sulfides) are successfully removed from liquid fuel via a hydrodesulfurization * To whom correspondence should be addressed. E-mail: tbandosz@ ccny.cuny.edu. Tel: (212) 650-6017. Fax: (212) 650-6107. † The City College of New York. ‡ The Graduate School of the City University of New York. (1) Song, Ch. Catal. Today 2003, 86, 211-263. (2) Babich, I. V.; Moulijn, J. A. Fuel 2003, 82, 607-631. (3) Matsuzawa, S.; Tanaka, J.; Sato, S.; Ibusuki, T. J. Photochem. Photobiol., A 2002, 149, 183-189. (4) Jirsak, T.; Dvorak, J.; Rodriguez, J. A. J. Phys. Chem. B 1999, 103, 55505559. (5) Rodriguez, J. A.; Hrbek, J. Acc. Chem. Res. 1999, 32, 719-728. (6) Gerlock, J. L.; Mahoney, L. R.; Harvey, T. M. Ind. Eng. Chem. Fundam. 1978, 17, 23-28. (7) Shirashiri, Y.; Taki, Y.; Hirai, T.; Komasawa, I. Ind. Eng. Chem. Res. 2001, 40, 1213-1224. (8) Tam, P. S.; Kittrell, J. R.; Eldridge, J. W. Ind. Eng. Chem. Res. 1990, 29, 324-329. (9) Dhen, W. M. J. Am. Chem. Soc. 1940, 62, 3189-3190. (10) Hernandez-Maldonado, A. J.; Yang, R. T. Ind. Eng. Chem. Res. 2003, 42, 123-129. (11) Yazu, K.; Yamamoto, Y.; Feruya, T.; Miki, K.; Ukegawa, K. Energy Fuels 2001, 15, 1535-1536. (12) Yang, R. T.; Takahasi, A.; Yang, F. H. Ind. Eng. Chem. Res. 2003, 42, 3103-3110. (13) Salem, A. B. S. H. Ind. Eng. Chem. Res. 1994, 33, 336-340. (14) Salem, A. B. S. H.; Hamid, H. S. Chem. Eng. Technol. 1997, 20, 342347. (15) Haji, S.; Erkey, C. Ind. Eng. Chem. Res. 2003, 42, 6933-6937. (16) Jiang, Z.; Liu, Y.; Sun, X.; Tian, F.; Sun, F.; Liang, Ch.; You, W.; Han, Ch.; Li, C. Langmuir 2003, 19, 731-736. (17) Ania, C. O.; Bandosz, T. J. Langmuir 2005, 21, 7752-7759. (18) Ania, C. O.; Bandosz, T. J. Carbon 2006, 44, 2404-2412. (19) Ania, C. O.; Parra, J. B.; Arenillas, A.; Rubiera, F.; Bandosz, T. J.; Pis, J. J. Surf. Sci., in press. (20) Federal Register 64/92/1999/40CFR, 80/85/86/AMS-FRL: 6337-3, RIN 2060-A123.

process.1,2,21,22 Nevertheless, some sulfur species, especially dibenzothiphene and 4,6-dimethyldibenzothiophene, are very resistant to hydrodesulfurization.1 Various methods based on extraction and adsorption have been proposed to remove these compounds.1-4,6-8,10-19 Whereas in the former process sulfur species are first oxidized and then extracted using organic solvents such as acetonitrille,3,8 adsorption is based on an enhancement of either adsorption forces or selectivity, or on imposing a chemical reaction. So far, the enhancement of the removal of thiophene compounds was reported for materials where π complexation can occur, such as on Cu-Y zeolites10,11 or on alumina and carbon with highly dispersed sodium.6,23 In the latter case, mono- and disodium thiophene metalates are formed. Other desulfurization methods use the formation and subsequent precipitation of S-alkylsulfonium salts.7 So far, the adsorption of dibezothiophene species on carbonaceous adsorbents has been addressed in the literature to a limited extent. One factor that hinders the application of carbons is their low selectivity. On their hydrophobic surfaces, all aromatic components of liquid fuels are expected to be adsorbed in significant amounts. Nevertheless, some applications of activated carbons for deep desulfurization have been described in the literature.1,2,13-19,23 The comparison of the feasibility of the desulfurization of naphtha on activated carbons and zeolites 5A and 13X was described by Salem13 and Salem and Hamid.14 The results indicated that zeolite 13X has a high capacity for sulfur in low concentration ranges and should be used when the sulfur content in the organic fraction is less than 25 ppm. When the concentration increases, the capacity of activated carbon can be 3 times greater than that on a zeolite. To include the effects of the porosity of carbon adsorbents on the removal of dibenzothiophene from fuel oil, the adsorbents were treated with concentrated sulfuric (21) Stirling, D. The Sulfur Problem: Cleaning up Industrial Feedstocks; Royal Society of Chemistry: Cambridge, U.K., 2000. (22) Satterfield, C. N. Heterogeneous Catalysis in Industrial Practice, 2nd ed.; McGraw-Hill: New York, 1991; p 378. (23) Ania, C. O.; Bandosz, T. J. Energy Fuels 2006, 20, 1076-1080.

10.1021/la063291j CCC: $37.00 © 2007 American Chemical Society Published on Web 04/18/2007

6034 Langmuir, Vol. 23, No. 11, 2007

acid at temperatures between 150 and 270 °C.16 This process, besides leading to significant changes in the porosity toward the development of a high volume of mesopores, also caused changes in surface chemistry and incorporation of numerous acidic groups. A significant increase in the amount of DBT adsorbed was linked to the development of mesoporosity and an increase in the DBT adsorption rate.16 Partial support for the results described above was obtained on carbon with various degrees of oxidation containing oxygenor sulfur-based acidic groups.17,19 Whereas mesoporosity indeed increased the kinetics of adsorption, the amount adsorbed seems to be linked to the volume of micropores. Moreover, after a non-structure-affecting oxidation treatment, the increase in the amount adsorbed was attributed to the increase in the number of acidic groups. These groups are indicated as the specific centers for the adsorption of DBT via oxygen- or sulfur-sulfur interactions. Moreover, strong acidity was hypothesized as a factor promoting the carbocation formation and oligomerization of DBT on the surface.17 These results indicated that surface chemical modification of carbon can increase the selectivity of DBT adsorption (adsorption of aromatic hydrocarbons is expected to decrease with an increase in the extent of oxidation). Selectivity was also improved when porous carbons containing highly dispersed metals were used as DBT adsorbents.18 Besides microporosity, the presence of metals and their speciation were proven as important for DBT retention on the surface. Whereas in the case of silver π -complexation was possible to occur, an enhanced adsorption in the case of cobalt or copper was linked to the catalytic effects of cobalt and ability of copper to activate oxygen.24,25 The objective of this article is to evaluate the role of the surface chemistry of carbons further, including the presence of metals, and their porosity in the process of DBT reactive adsorption from hexane. For this purpose, the carbons obtained from metalcontaining polymer carbonized within inorganic matrices are studied. As templates, alumina, amorphous silica, and zeolite were used. The differences in the porous structure and chemistry of templates lead to unique carbons with a high volume of mesopores, a significant contribution of micropores, and a very high dispersion of metals. Their behavior as DBT adsorbents is analyzed, and the performance is linked to the surface features. Experimental Section Materials. Poly(styrene sulfonic acid-co-maleic acid) sodium salt (PS) was used as the organic precursor. As the template, silica gel (powder, 60-200 mesh), activated alumina (particle size >0.425 mm), and 13X (Na86[(AlO2)86(SiO2)106]‚H2O molecular sieves (particle size >0.425 mm) were used. The carbons obtained from PS were synthesized using the modified procedure described by Hines et al.26 and Ania and Bandosz.27 The dry inorganic sorbent was impregnated with an aqueous solution of PS polymer (18 wt %). The polymer was used as received (sodium salts) and after ion exchange with nitrates of Cu (II), Co (II), or Fe(III) (0.05 M).26,27 The metals were chosen on the basis of their ability to catalyze the oxidation of sulfur compounds (Cu and Fe28) and their catalytic effect on HDS (Co). The samples were immersed in the polymerwater solution and stirred for 72 h. Then were they filtered, dried, and carbonized under nitrogen at 800 °C for 40 min. After (24) Hafen, J. A.; Mahapatra, S.; Wilkinson, E. C.; Kaderli, S.; Young, V. C.; Que, L., Jr.; Zuberbuhler, A. D.; Tolman, W. B. Science 1996, 272, 1397-1400. (25) Hu, Z.; Williams, R. D.; Tran, D.; Spiro, T. G.; Gorun, S. M. J. Am. Chem. Soc. 2000, 122, 3556-3557. (26) Hines, D.; Bagreev, A.; Bandosz, T. J. Langmuir 2004, 20, 3388-3397. (27) Ania, C. O.; Bandosz, T. J. Microporous Mesoporous Mater. 2006, 89, 315-324. (28) Bandosz, T. J. In ActiVated Carbon Surfaces in EnVironmental Remediation; Bandosz, T. J., Eds.; Elsevier: Oxford, U.K., 2006; p 231.

Seredych and Bandosz carbonization, the inorganic matrices were removed using 48% HF (in the case of silica) or HF and HCl (in the case of alumina and molecular sieves). The resulting carbons were washed with water in a Soxhlet apparatus. They are referred to as C-Si-M, C-Al-M, and C-Z-M where Si, Al, and Z refer to a silica, alumina, or zeolite matrix, respectively, and M is either Cu, Co, or Fe. Methods. Adsorption of DBT from Solution. The adsorption of DBT was carried out at room temperature in a stirred batch system. Before these experiments, kinetic studies were performed to determine the equilibration time of the system. Different amounts of carbon (from 25 mg to 1 g) were weighed and added to 12 bottles containing 40 mL of the sulfur-containing solution with an initial concentration of 1000 ppmw DBT (ca. 178 ppmw S). All of the solutions were prepared in hexane. The covered bottles were placed in a shaking bath and allowed to shake for 72 h at constant temperature. After equilibration, the concentration in the liquid was determined using a UV spectrophotometer at the corresponding wavelength (312 nm). The amount adsorbed was calculated from the equation qe ) V(Co - Ce)/m, where qe is the amount adsorbed, V is the volume of the liquid phase, Co is the concentration of solute in the bulk phase before it comes into contact with the adsorbent, Ce is the concentration of solute in the bulk phase at equilibrium, and m is the amount of adsorbent. The equilibrium data were fitted to the so-called LangmuirFreundlich single-solute isotherm,29 which is described by the equation ϑt )

qe (KC)n ) qo 1 + (KC)n

where qe is the adsorbed amount of solute per unit gram of adsorbent, qo is its maximum adsorption per unit weight of adsorbent, K is the Langmuir-type constant defined by the van’t Hoff equation, and the exponential term n represents the heterogeneity of the site energies. The fitting range was from 0 to 250 mg of S per gram of activated carbon (recalculated from its content in DBT). Textural Characterization. Textural characterization was carried out by measuring the N2 adsorption isotherms at 77 K. Before the experiments, the samples were outgassed under vacuum at 393 K. The isotherms were used to calculate the specific surface area SBET, total pore volume VT, volume of micropores Vmic, volume of mesopores Vmes,30 and pore size distributions. The latter were obtained using density functional theory (DFT).31 Thermal Analysis. Thermal analysis was carried out using a TA Instruments thermal analyzer. The instrument settings were: a heating rate of 10 °C/min and a nitrogen atmosphere with a 100 mL/min flow rate. For each measurement, about 15 mg of a carbon sample was used. Potentiometric Titration. Potentiometric titration measurements were performed with a DMS Titrino 716 automatic titrator (Metrohm, Brinkmann Instruments, Westbury, NY). The instrument was set in equilibrium mode when the pH was collected. Approximately 0.1000 g samples were placed in a container thermostatted at 298 K with 50 mL of 0.01 M NaNO3 and equilibrated overnight. To eliminate any interference by dissolved CO2, the suspension was continuously saturated with N2. The carbon suspension was stirred throughout the measurement. Each sample was titrated with a 0.1 M NaOH titrant using 0.001 mL increments. Experiments were carried out in the pH range of 3-10.32,33 Surface pH. The pH of a carbon sample suspension provides information about the acidity and basicity of the surface. A sample of 0.4 g of dry carbon powder was added to 20 mL of water, and (29) Derylo-Marczewska, A.; Jaroniec, M. Chem. Scr. 1984, 24, 239. (30) Gregg, S. J.; Sing, K. S. W. Adsorption, Surface Area, and Porosity; Academic Press: New York, 1982. (31) Lastoskie, Ch. M.; Gubbins, K. E.; Quirke, N. J. Phys. Chem. 1993, 97, 4786-4796. (32) Bandosz, T. J.; Jagiello, J.; Contescu, C.; Schwarz, J. A. Carbon 1993, 31, 1193-1202. (33) Jagiello, J.; Bandosz, T. J.; Schwarz, J. A. Carbon 1994, 32, 1026-1028.

Template-DeriVed Mesoporous Carbons

Langmuir, Vol. 23, No. 11, 2007 6035

Table 1. Yields of Materials, Ash and Metal Contents, Ignition Temperatures, and Surface pH

sample C-Si C-Si-Cu C-Si-Co C-Si-Fe C-Al C-Al-Cu C-Al-Co C-Al-Fe C-Z C-Z-Cu C-Z-Co C-Z-Fe

yield after final yield ash metal carbonization (after HF/HCl) content content Tign (%) treatment (%) (wt %) (%) (°C) 56.3 64.5 62.1 63.6 54.8 68.3 62.5 57.2 57.8 56.4 55.9 57.1

9.8 10.2 10.6 11.2 11.4 11.0 15.4 10.7 14.6 15.1 14.9 12.3

0.0 1.1 2.5 3.7 0.5 1.1 6.7 0.7 3.4 6.5 5.5 1.2

0 0.49 0.38 0.0063 0 0.71 0.32 0.013 0 0.055 0.33 0.19

590 521 575 577 551 506 520 564 464 481 520 517

pH 3.67 3.82 3.70 3.55 3.80 3.49 3.43 3.92 4.22 4.44 4.97 5.27

the suspension was stirred overnight to reach equilibrium. Then the sample was filtered, and the pH of the solution was measured.

Results and Discussion Characterization of Adsorbents. The yields of materials and their ash contents, metal contents, ignition temperatures, and surface pH values are listed in Table 1. The small yields between 10 and 15% are related to the significant contribution of an inorganic matrix to the weight of the samples after carbonization. Acid treatment, although removed most of the templates, still left a few percent of an inorganic matter, especially in the case of zeolite. In fact, only small fractions of that inorganic matter can be linked to metals because their amounts are less than 1%. This should ensure a high dispersion of metal within the porous carbon matrix. It is interesting that in the case of alumina- and silica-derived samples washing with acid removed almost all of the iron with much smaller effects on copper and cobalt. This may be linked to less iron than copper and cobalt present in the initial polymers as a result of its higher oxidation state.26 It is also likely that owing to the difference in chelation, copper and cobalt are dispersed in smaller pores and thus are less accessible to acids. This is not true in the case of zeolite-derived samples where high contents of inorganic matter are left after washing and the porosity is rather secondary, formed between washed out graphene layers.34 The ignition temperatures are about 500 °C, and the addition of metal generally decreases them in the case of silica- and alumina-derived carbons, whereas in the case of zeolite-derived materials an increase is observed. Although it is difficult to establish any direct relationships at this stage in our study, these trends can be caused by the sizes of crystallites, the content of oxygen (chemisorbed or in functional groups as pH values), and the reactivity of metals. Because the presence of copper clearly decreases the ignition temperature, this can be caused by its activation of oxygen.24,25 The acidic pH of the carbon surfaces is related to the presence of sulfonic groups having their origin in the carbon precursor18,19,26 and oxygencontaining acidic groups, which are formed when the samples are exposed to air. (In some cases, strong exothermic reactions were observed when samples were exposed to the atmosphere.) The surface acidity of the carbons studied was estimated using potentiometric titration experiments. From the titration curves, the proton binding curves, Q, were derived (Figure 1).32,33 The location of curves with relation to Q ) 0 is related either to proton uptake (over zero), and thus the basic character of the surface, or to proton release (negative values), and thus the acidic character of the surface. Whereas all samples obtained in alumina and silica matrices are acidic, in the case of carbons derived from (34) Seredych, M.; Bandosz, T. J. Microporous Mesoporous Mater. 2007, 100, 45-54.

Figure 1. Proton uptake curves for carbons obtained in silica, alumina, and zeolite matrices.

zeolites, especially those containing iron and cobalt, some basicity is present. This is in agreement with the pH values reported in Table 1. Generally, the presence of metals decreases the number of acidic groups, and this effect is strongly seen in the case of iron-containing samples. The changes are more pronounced for the species that dissociate in a pH range greater than 8. From the proton uptake curves, the pKa distributions of the species present on the surface were calculated using the SAIEUS procedure.35They are presented in Figure 2. For all samples, (35) Jagiello, J. Langmuir 1994, 10, 2778-2785. (36) Weast, R. C. Handbook of Chemistry and Physics, 62nd ed.; CRC Press: Boca Raton, FL, 1981-1982.

6036 Langmuir, Vol. 23, No. 11, 2007

Seredych and Bandosz Table 2. Structural Parameters Calculated from the Adsorption of Nitrogen SBET (m2/g)

VT (cm3/g)

Vmeso (cm3/g)

Vmic (cm3/g)

Vmic/Vt

Si

255

1.162

1.061

0.101

0.09

C-Si C-Si-DBT C-Si-Cu C-Si-Cu-DBT C-Si-Co C-Si-Co-DBT C-Si-Fe C-Si-Fe-DBT

958 604 592 480 807 550 717 454

3.166 2.316 1.936 1.813 2.316 1.868 2.097 1.625

2.773 2.080 1.695 1.625 1.987 1.640 1.792 1.448

0.393 0.236 0.241 0.188 0.329 0.228 0.305 0.177

0.12 0.10 0.12 0.10 0.14 0.12 0.14 0.11

Al

318

0.376

0.258

0.118

0.31

C-Al C-Al-DBT C-Al-Cu C-Al-Cu-DBT C-Al-Co C-Al-Co-DBT C-Al-Fe C-Al-Fe-DBT

661 370 391 218 398 237 376 202

0.888 0.665 0.531 0.422 0.632 0.451 0.603 0.487

0.625 0.527 0.369 0.341 0.475 0.365 0.459 0.413

0.263 0.138 0.162 0.081 0.157 0.086 0.144 0.074

0.30 0.21 0.30 0.19 0.25 0.19 0.24 0.15

Z

486

0.340

0.096

0.244

0.72

C-Z C-Z-DBT C-Z-Cu C-Z-Cu-DBT C-Z-Co C-Z-Co-DBT C-Z-Fe C-Z-Fe-DBT

410 115 551 112 456 118 414 108

0.626 0.441 0.681 0.412 0.654 0.447 0.598 0.411

0.435 0.395 0.418 0.367 0.445 0.401 0.403 0.368

0.191 0.046 0.263 0.045 0.209 0.046 0.195 0.043

0.30 0.10 0.39 0.11 0.32 0.10 0.33 0.10

sample

Figure 2. pKa distributions for carbons obtained in silica, alumina, and zeolite matrices.

numerous peaks are revealed, representing the high degree of surface chemical heterogeneity. On the basis of the number of peaks and their intensity, broadness, and location, the coppercontaining samples, especially C-Al-Cu, are found to be the most acidic whereas iron-containing samples are found to be the least acidic ones. That acidity can be certainly, but not exclusively, linked to the presence of surface functional groups. Metal species such as oxides can also contribute to the samples’ acidity. The decomposition of various inorganic salts during carbonization, accompanied by the release of gases and the migration of metals, must have an effect on the development of porosity in template-derived carbons.26,34 The structural parameters

calculated from the adsorption of nitrogen are collected in Table 2. Whereas the materials obtained from the sodium form of the polymer in silica or alumina matrices have the largest surface areas and pore volumes, the structural parameters of corresponding carbons obtained from transition-metal salts are much smaller, especially in the case of the alumina matrix. As mentioned previously when silica is used as a template, the high-temperature reaction of sodium with silica walls significantly increases the volume of pores, especially large mesopores.27,34 It is interesting that when copper is present about a 50% decrease in the structural parameters is noticed whereas for Co- and Fe-containing samples only about a 20% decrease is found compared to that for C-Si. Similar behavior was observed for carbons obtained form bulk polymers; however, their structure was much more microporous.26 In the case of alumina templating, the presence of transition metals results in a similar, about 50% decrease in the structural parameters compared to that for the sample obtained from the initial sodium form of the polymer, C-Al. These differences in the behavior between silica- and alumina-derived materials are caused by template constrains during the carbonization process that affect the mechanism of carbonization and pore formation. As hypothesized elsewhere,26 sodium as a one-valent metal should lead to the high volume of small pores formed as a result of the release of a large number of sodium atoms from the structure and their migration to the surface. An increase in the metal charge leads to fewer metal ions needed for chelation and thus a smaller contribution of pores formed as a result of metal migration. Another factor is a reduction potential that is low in the case of copper.18 Whereas other metals can form sulfates, which further decompose, resulting in pore formation, copper, if reduced to metallic copper, is likely not involved in this process. The differences in porosity between alumina- and silica-derived samples are also caused by the fact that initially larger pores in silica enable more polymer incorporation and thicker layers of carbons, especially when the pore walls are destructed via their

Template-DeriVed Mesoporous Carbons

Langmuir, Vol. 23, No. 11, 2007 6037

Figure 3. Comparison of pore size distributions for carbons obtained in silica, alumina, and zeolite matrices.

reaction with sodium.27,34 In such a material, a greater volume of pores can be formed because either metals or gases have to migrate through a thick deposit of char. In the case of zeolitederived materials, the degree of microporosity, Vmic/Vt, is the highest from all samples, but mesopores are still the predominant pore sizes. It is likely that the porosity of zeolite-based samples is the secondary one formed between carbon crystallites removed from small channels of zeolites.34 In such a case, the porosities of all samples, regardless of the metal, are more or less similar. The smallest pore sizes are in those samples, although they are not replicas of the zeolite matrix. Details of the structural parameters are presented in Figure 3 as pore size distributions (PSDs). Whereas the silica-derived samples have almost all pores with sizes between 20 and 200 Å, the alumina-derived samples are much more heterogeneous

Figure 4. DBT adsorption isotherms. Solid lines indicate the fit of the L-F equation.

with two predominant pore ranges: between 10 and 20 Å and between 20 and 400 Å. The volume of the latter significantly depends on the metals used in the polymers and can be linked to the chelating properties26 and the redox potential. As mentioned above, all samples derived form zeolites have similar pore structures. Mechanism of DBT Adsorption. The carbons obtained were used as adsorbents of DBT from hexane. The isotherms along with the fit to the Langmuir-Freundlich equation are presented in Figure 4. Table 3 shows the fitting parameters and the limiting capacity in milligrams of sulfur per gram of adsorbents. Compared to the data that have been reported so far,15-19 the carbons used in this study have a very high capacity for DBT removal. The

6038 Langmuir, Vol. 23, No. 11, 2007

Seredych and Bandosz

Table 3. Fitting Parameters of the L-F Equation sample

qm (mg S/g)

K (L/mg)

n

R2

C-Si C-Si-Cu C-Si-Co C-Si-Fe C-Al C-Al-Cu C-Al-Co C-Al-Fe C-Z C-Z-Cu C-Z-Co C-Z-Fe

138.6 114.3 120.2 103.1 81.0 99.7 89.4 66.5 51.4 69.8 75.5 68.0

0.06 0.05 0.03 0.04 0.04 0.04 0.02 0.04 0.05 0.07 0.05 0.06

0.48 0.44 0.43 0.43 0.40 0.34 0.46 0.38 0.33 0.37 0.36 0.39

0.9986 0.9907 0.9926 0.9902 0.9813 0.9963 0.9962 0.9971 0.9985 0.9975 0.9991 0.9980

Table 4. Normalized Metal Content and Normalized DBT Adsorption

sample C-Si C-Si-Cu C-Si-Co C-Si-Fe C-Al C-Al-Cu C-Al-Co C-Al-Fe C-Z C-Z-Cu C-Z-Co C-Z-Fe

metal content (mmol/g)

metal content × 10-5 (mmol/m2)

0.080 0.070 0.001

0.135 0.087 0.0016

0.110 0.054 0.002

0.286 0.136 0.0080

0.009 0.056 0.034

0.016 0.123 0.082

DBT adsorbed (mmol/g)

DBT adsorbed × 10-3 (mmol/m2)

4.3 3.6 3.8 3.2 2.5 3.1 2.8 2.1 1.6 2.2 2.4 2.1

4.5 6.0 4.7 4.5 3.8 8.0 7.0 7.2 3.9 4.0 5.2 5.1

lowest capacity is found for the samples obtained from the iron salt, and the highest capacity is found for the copper salt, with the exception of the samples derived in silica matrices where C-Si and C-Si-Co perform better than the sample containing copper. Knowing the significant differences in the porosity and surface chemistry of these carbons, such a comparison of the sample activity would be oversimplified. To compare the performance of the adsorbents properly, their surface features have to be taken into account. Table 4 lists the metal contents in millimoles per gram and in millimoles per unit surface area of the adsorbent. The latter quality provides information about the degree of metal dispersion on the surface. For the sake of comparison, the DBT limiting capacity was also recalculated in the same units. Figure 5 shows the relationship between the capacity and the metal contents. Although the points are scattered, especially for low metal content, which might be related to the accuracy of the measurements of low iron content, the trend of an increasing capacity with an increasing metal content can be noticed. The correlation is especially good for copper. For zero metal content, the capacity estimated using the linear trend is about 4 mmol/m2, which in fact is the value found for carbons without a transition metal present. This suggests that about a 90% increase in surface activity toward the retention of DBT can be attributed to the transitionmetal species highly dispersed on the surface. Because pores in our materials are rather large, the diffusion limitations are not expected to be significant, and the metals should be easily accessible to DBT molecules. Nevertheless, certain pore sizes, especially in carbons without metals, should attract DBT molecules.17-19 By taking into account that the critical size of the DBT molecule is similar to that of benzene (6.8 Å), the strongest adsorption should occur in pores smaller than 7 Å. Contrary to the results reported for “classical” activated carbons,17

Figure 5. Dependence of the normalized limiting DBT adsorption capacity on the normalized content of metals.

no dependence of capacity on the volume of pores was found in the case of template-derived carbons. When changes in the pore volume for the samples after DBT adsorption and in the pore size distributions are compared to those for the initial samples, the distribution of the DBT adsorption sites looks much more complex (Table 2, Figures A-C in Supporting Information). In the case of silica-templated samples, besides a significant decrease in the volume of micropores, mesopores are also affected, however to a much lesser extent. The largest changes in pore volume (20-25% in mesopores and 40% in micropores) are noticed for the carbons without transition metals and those obtained from iron salt where the metal content is very small. In the case of copper- and cobalt-containing samples, especially for the former one, changes in the pore volume are very small, taking into account the large amount of DBT adsorbed on those carbons. These results indicate differences in the mechanism of adsorption. Whereas in the case of carbons without or with a low content of metals, small pores (micropores and small mesopores) are very active in the adsorption, and their entrances can be blocked by DBT interacting with acidic groups.17-19 In the case of metal-containing samples, the metals located in the pores can be the active centers, and the DBT molecule can position itself parallel or perpendicular to the surface depending on the type of sulfur-metal interaction. The former geometry should result in a smaller decrease in the pore volume than the latter one. It is well known that thiophenic species can interact with metal either via donation of a lone pare of electrons, which is considered to be a direct S-M σ bond, or via delocalized electrons of the aromatic ring, where a π-type complex with the metals is formed.1,37 The latter arrangement should lead to a parallel location of DBT on the surface, provided that the pore geometry and size allow it. Because the smallest changes in the structural parameters are noticed for the copper-containing samples, the possibility of π-type interaction is a plausible option here. Because copper has a low reduction potential, the copper species that are likely to exist are Cu0 and Cu+1 or Cu+2 with the majority being Cu0. The reduced copper metal possibly binds to the π system of dibenzothiophene and thus causes (37) Yang, X.; Erickson, L. E.; Hohn, K. L.; Jeevanandam, P.; Klabunde, K. L. Ind. Eng. Chem. Res. 2006, 45, 6169-6174.

Template-DeriVed Mesoporous Carbons

Langmuir, Vol. 23, No. 11, 2007 6039

Figure 6. Comparison of DTG curves for all templated samples before and after only hexane adsorption (H) and DBT adsorption from hexane (DBT).

strong adsorption of DBT on the surface.37 In the case of cobalt, where a greater decrease in the pore volume is found, the S-M σ bond is more likely to form. It has to be mentioned here that these hypothesized scenarios occur at low surface coverage. With an increase in surface coverage, the surface acidic groups and dispersive interactions in the pore system become important. The adsorption of DBT looks different in the case of aluminaderived materials. In these, micropores are the most active in the process of DBT adsorption, and a 45-50% decrease in their volume is found for all samples. In fact, the sample derived from the sodium form of the polymer, C-Al, was indicated as the most acidic on the basis of our surface chemistry analyses. It is

possible that acids present on the surface (oxygen and sulfur based) interact in the same way as metals, via a sulfur active-site bond. The small decrease in the volume of mesopores for coppercontaining samples, with the high amount adsorbed on this sample, once again may suggest the π complexation of DBT aromatic ring electrons with metallic copper present on the walls of mesopores. In the case of samples obtained in zeolite templates, regardless of their metal content and the kind of metal, the same changes in the structural parameters are found after DBT adsorption: about an 80% decrease in the volume of micropores and about a 10% decrease in the volume of mesopores. As indicated previously, this is the result of a similar porous structure. With

6040 Langmuir, Vol. 23, No. 11, 2007

Seredych and Bandosz Table 5. Amount of DBT Adsorbed (from the Isotherm Measurements) and Weight Loss for Samples after Hexane and DBT Adsorption (from TA Analyses)

Figure 7. Relationship between DBT adsorbed and the weight loss caused by its decomposition/removal from the surface. Open symbol represent the samples for which more decomposition products are detected compared to the situation when only weak interactions are present (disturbed carbonization/aromatization process via strong DBT surface interactions).

the relatively high volume of small pores, the apparent adsorption seems to have the same mechanism on all zeolite-derived carbons. To see the effects of DBT and hexane adsorbed on carbons on the changes in the materials’ chemistry while they are heated in nitrogen, the DTG patterns for the initial samples and those exposed only to hexane and DBT in hexane are analyzed. The DTG curves are presented in Figure 6. The weight loss seen as peaks between 200 and 600 °C for the initial samples represent the decomposition of both the acidic groups and hydroxides formed when metals oxides were exposed to atmospheric moisture.26 After hexane adsorption on silica-derived samples, the new peaks appear at T < 100 °C and about 220 °C representing adsorbed hexane17-19 in the relatively large and small pores, respectively. The complexity of this peak is related to the pore size distribution. Because all silica-derived samples except for C-Si-Cu have similar micropore volumes, the similar amounts of hexane adsorption were detected as a second peak. After DBT adsorption, a large peaks appears between 100 and 600 °C, which is related to the decomposition of DBT adsorbed on the surface. The complexity of this peak is linked to the strength of DBT interactions with the surface. The main peak that is centered at about 250 °C represents the removal of physically adsorbed DBT. For some samples, such as those containing copper and cobalt, a well-defined shoulder is seen between 400 and 600 °C that represents the decomposition of DBT strongly adsorbed via S-M σ bonds or π complexation. This effect is especially well pronounced for the copper-containing sample. A similar trend is seen for the alumina-derived samples. A striking difference between alumina- and silica-derived samples is a decrease in the weight loss between 200 and 400 °C for the samples exposed to hexane in comparison with the initial samples. Whereas in the case of DBT adsorption the bond between the surface acidic groups and sulfur can be formed by changing the stability of functional groups, the lack of decomposition or a shift toward lower temperature in the case of hexane cannot be explained at this stage of our study. For the zeolite-derived samples, the most active centers for hexane adsorption are those in the relatively large pores (low-temperature peak). After DBT adsorption, all

sample

DBT adsorbed (%)

mass loss as DBT (%) (100 - 600 °C)

mass loss as hexane (%) (100 - 600 °C)

C-Si C-Si-Cu C-Si-Co C-Si-Fe C-Al C-Al-Cu C-Al-Co C-Al-Fe C-Z C-Z-Cu C-Z-Co C-Z-Fe

51 42 38 35 28 34 28 24 18 25 25 25

22.3 17.6 17.8 22.2 23.2 24.8 19.9 15.3 14.6 20.8 17.7 15.7

8.7 5.1 9.3 9.8 14.6 16.1 13.4 6.3 10.6 8.96 9.4 10.1

metal-containing samples derived in zeolites show the welldefined shoulder between 400 and 600 °C attributed to specific metal-DBT interactions. A decrease in the intensity of the first hexane removal peak must be related to the replacement of hexane by DBT molecules, which are more strongly adsorbed on the carbon surfaces. The chemistry of thermal decomposition of DBT on the carbonaceous surface should be related to its engagement in specific interactions with that surface. If no specific interactions are involved, then the fraction of the weight loss calculated from the amount of DBT adsorbed should be the same for all samples and is caused by the same mechanism. Figure 7 correlates the weight loss between 100 and 600 °C after DBT adsorption corrected for the amount of hexane adsorbed and the weight loss due to the release of gases from the decomposition of acidic groups with the amount of DBT adsorbed. The numerical values are reported in Table 5. As seen, some values are located on the straight line with the slope equal to 0.24. These samples represent those where strong interactions with the carbon surface are not present. In this case, the decomposition, carbonization, and aromatization of DBT are not disturbed. Some points are located over the y ) 0.24x line. They represent more decomposition products than those expected when the weaker interactions are present. The strong bond of DBT sulfur with metals disturbs the decomposition of DBT and the condensation of its rings, and as a result of this, more hydrocarbons are released compared to the situation when those bonds are not present. It is important to mention that the points located over the decomposition line are those for which the well-defined shoulder on DBT decomposition DTG patterns is noticed (Figure 6). As indicated by Ania and co-workers after heating the DBT-saturated samples in nitrogen, for which reactive adsorption was noticed, a significant amount of sulfur has left the surface19 and only a fraction of adsorbed DBT, similar to that found in this study, was removed from the surface during heating.

Conclusions The results described in this article present template-derived carbons with highly dispersed metals as efficient media for DBT removal. A combination of surface acidity and the high dispersion of metals and their specific properties along with the high volume of mesopores enhance the physical adsorption via reactive adsorption on surface active centers. That reactive adsorption is based on sulfur-oxygen, sulfur-sulfur, and sulfur-metal interactions. In the case of copper, the π complexation of DBT electrons with reduced metals is possible, and it significantly enhances the amount adsorbed. The exceptionally high volume

Template-DeriVed Mesoporous Carbons

of mesopores in the carbons studied (up to 2.7 cm3/g) is a result of the combined effects of chemistry of the carbonaceous precursor and the porous structure and chemistry of the templates. Acknowledgment. The work was supported by PSC CUNY grant no. 67284-0036. T.J.B. is grateful to Dr. Jacek Jagiello for the SAIEUS software.

Langmuir, Vol. 23, No. 11, 2007 6041

Supporting Information Available: Supporting information includes the comparison of pore size distributions for silica-templated samples before and after DBT adsorption, comparison of pore size distribution for alumina-templated samples before and after DBT adsorption, and the comparison of pore size distribution for zeolitetemplated samples before and after DBT adsorption. This material is available free of charge via the Internet at http://pubs.acs.org. LA063291J