Selective Adsorption for Removal of Nitrogen Compounds From Liquid HC Streams Over Carbon- And Alumina- Based Adsorbents

Selective Adsorption for Removal of Nitrogen Compounds from LiquidHydrocarbon Streams over Carbon- and Alumina-Based Adsorbents

Masoud Almarri,, Xiaoliang Ma, and Chunshan Song*,

Clean Fuels and Catalysis Program, EMS Energy Institute, and Department of Energy and Mineral Engineering,PennsylVania State UniVersity, 209 Academic Projects Building, UniVersity Park, PennsylVania 16802 andPetroleum Refining Department, Kuwait Institute for Scientific Research, P.O. Box 24885, Safat, 13109, Kuwait

In order to explore the adsorptive denitrogenation of liquid hydrocarbon streams for producing ultracleanfuels, the adsorption performance of seven representative activated carbon samples and three activated aluminasamples was evaluated in a batch adsorption system and a fixed-bed flow adsorption system for removingquinoline and indole from a model diesel fuel in the coexistence of sulfur compounds and aromatics. Differentadsorbents show quite different selectivity toward basic and nonbasic nitrogen compounds (quinoline andindole) and sulfur compounds (dibenzothiophene and 4,6-dimethyldibenzothiophene). The activated carbonsgenerally show higher capacity than activated alumina samples for removing the nitrogen compounds. Theadsorption capacity and selectivity of the activated carbons for nitrogen compounds were further correlatedwith their textural properties and oxygen content. It was found that (1) the microporous surface area andmicropore volume are not a key factor for removal of the nitrogen compounds in the tested activated carbons;(2) the oxygen functionality of the activated carbons may play a more important role in determining theadsorption capacity for the nitrogen compounds since the adsorption capacity for nitrogen compounds increaseswith increase in the oxygen concentration of the activated carbons; and (3) the type of the oxygen-functionalgroups may be crucial in determining their selectivity for various nitrogen or sulfur compounds. In addition,regeneration of the saturated adsorbents was conducted by the toluene washing followed by the heating toremove the remained toluene. The results show that the spent activated carbons can be regenerated to completelyrecover the adsorption capacity. The high capacity and selectivity of carbon-based adsorbents for the nitrogencompounds, along with their good regenerability, indicate that the activated carbons may be promisingadsorbents for deep denitrogenation of liquid hydrocarbon streams.

1. Introduction

Because of the continuous decline in crude oil quality, therefining industry needs to deal with heavier feedstocks thatcontain higher concentrations of nitrogen and sulfur com-pounds.1 However, it has been reported that due to the increasingdemand for diesel fuel, the United States and Europe will facethe increasing shortages in diesel fuel in the future.2 Therefore,it is extremely important to blend more of other refineryhydrocarbon streams, such as light cycle oil (LCO), and cokergas oil (CGO), into the diesel pool. LCO and CGO typicallycontain much more nitrogen, sulfur, and aromatic compoundsin comparison with the straight run gas oil (SRGO).

In addition, other fossil energy resources such as oil sandsand oil shale3,4 as well as coal liquid5 are attracting increasingattention in the energy industry.4,5 Together, these fossilresources are expected to supply the liquid transportation fuels,since the supply of crude oil alone will not meet the increasingglobal demand in the near future. Gas oil fractions derived fromoil sand (OSGO) and coal liquid (CLGO) have higher nitrogenconcentrations that may be 2 orders of magnitude higher thanthat of the corresponding petroleum fractions.

The nitrogen compounds found in gas oil are generallydivided into two groups: basic compounds, such as aniline,pyridine, quinoline, acridine, and their alkyl substituent deriva-

tives, and nonbasic compounds, such as pyrrole, indole, carba-zole, and their alkyl substituent derivatives. Laredo et al.6determined the nitrogen compounds in SRGO and LCO derivedfrom Mexican crude oil using gas chromatography with a massspectrometer (GC-MS). They found that the most abundantnitrogen compounds in the SRGO sample were quinoline,indole, carbazole, and their alkyl substituent derivatives, whereasaniline, indole, carbazole, and their alkyl substituent derivativeswere the predominant nitrogen compounds in the LCO sample.

Much attention has been paid to the removal of nitrogencompounds because these compounds have to be removed fromvarious refinery streams before such streams can be furtherprocessed in subsequent processes, such as isomerization,reforming, catalytic cracking, and hydrocracking, where thecatalysts are very sensitive to nitrogen compounds.7 In particular,the basic nitrogen compounds can be adsorbed strongly on theacidic sites of various catalysts used in petroleum refiningprocesses, resulting in poisoning of the active sites. For example,in catalytic cracking, basic nitrogen compounds are adsorbedon active acid sites and reduce the cracking activity of thecatalyst. In addition, the presence of nitrogen compounds affectsthe stability of fuels.8 Recently, predenitrogenation for ultradeephydrodesulfurization (HDS) is capturing much attention, asnitrogen compounds inhibit the ultradeep HDS, especially HDSof the refractory sulfur compounds, such as 4-methyldiben-zothiophene (4-MDBT) and 4,6-dimethyldibenzothiophene (4,6-DMDBT).9

Recently, many new approaches have been made to efficientlyproduce ultra clean transportation fuel.9-12 Numerous effortsin the last 10 years have significantly advanced the knowledgein the inhibiting effect of nitrogen compounds on deep

* To whom correspondence should be addressed. E-mail: [email protected].

Clean Fuels and Catalysis Program, EMS Energy Institute, andDepartment of Energy and Mineral Engineering, Pennsylvania StateUniversity.

Petroleum Refining Department, Kuwait Institute for ScientificResearch.

Ind. Eng. Chem. Res. 2009, 48, 951960 951

10.1021/ie801010w CCC: $40.75 2009 American Chemical SocietyPublished on Web 11/26/2008

HDS.13-19 It is well-known that the removal of the refractorysulfur compounds in diesel fuel is difficult by the conventionalHDS.2,9-12 The coexisting nitrogen compounds make deepdesulfurization even more complicated and difficult, as thesenitrogen compounds strongly inhibit the HDS of the refractorysulfur compounds.13-19 In order to achieve ultradeep HDS, themost refractory sulfur compounds have to be removed, and thetotal sulfur concentration in diesel fuel must be reduced toaround the ten parts per million (ppm) level, which is compa-rable to the nitrogen concentration in the fuel. In this case,influence of the nitrogen compounds on HDS becomes signifi-cant. The influence of the nitrogen compounds, including indole16,17 and quinoline,18,19 on deep HDS of diesel and model dieselfuels has been extensively investigated. It has been concludedthat the removal of the nitrogen compounds prior to HDS canremarkably improve HDS performance.14,15,17-19

However, the reactivity of the nitrogen compounds inhydrotreatment is significantly lower than that of the corre-sponding sulfur compounds. For example, alkyl-substitutedcarbazoles appear to react at the rates only about 1/10 as fastas those of alkyl-dibenzothiophenes, which have a skeletonstructure13 similar to carbazole. Therefore, when the nitrogencompounds are adsorbed onto the active site on catalyst surface,they remain there due to their strong adsorption affinity andlow reactivity, blocking the adsorption of the sulfur compoundsonto the active sites. Moreover, intermediate products and finalproduct ammonia from both basic and nonbasic nitrogencompounds during the hydrotreatment process inhibit furtherthe deep HDS.13,20,21 As a result, removal of the nitrogencompounds prior to HDS is critical in ultradeep HDS.

As is well-known, hydrodenitrogenation of all heterocyclicnitrogen compounds in gas oil is complicated and cannotproceed through the direct N-C cleavage without saturationof the heteroaromatic ring, in contrast to the direct S-C cleavagein HDS of dibenzothiophene (DBT). HDN needs to proceedthrough a hydrogenation pathway, which involves the completehydrogenation of the heteroring before C-N cleavage. Forexample, removal of sulfur from a DBT molecule only needsconsumption of 4 H atoms, while the removal of nitrogen froma quinoline molecule requires at least 8 H atoms.22 Thus,hydrodenitrogenation (HDN) is not only more kineticallydifficult than HDS, but also higher in hydrogen consumption,which is a key factor in determining the capital and operationalcost of the hydrotreatment process.

In order to achieve deep denitrogenation, use of adsorbentsto selectively remove the nitrogen compounds has attracted greatattention.23-26 The selectively removal of the nitrogen com-pounds and refractory sulfur compounds from liquid hydrocar-bons by adsorption is a promising approach,23,24 as theadsorption can be conducted at ambient temperatures withoutusing hydrogen. As well-known, liquid hydrocarbon streamsusually contain not only the nitrogen and sulfur compounds,but also a large amount of structurally similar aromaticcompounds. Thus, a great challenge is to identify an adsorbentthat can selectively adsorb the nitrogen compounds but notcoexisting aromatic compounds. Recently, several types ofadsorbents have been reported for adsorptive denitrogenationof liquid hydrocarbon fuels, including zeolite,23,27 activatedcarbon,15,24,26 activated alumina,15,24,28 and silica gel.25,29,30 Forexample, SK Company has developed a pretreating adsorptionprocess that removes over 90% of nitrogen compounds from adiesel fuel using silica gel as an adsorbent.25,30 The total amountof fuel adsorbed is approximately 2% of the total diesel feed,which comprises a large quantity of polar compounds that

contain functional groups, such as -COOH (naphthenic acids),-OH (phenols), -N (pyridines), and -NH (pyrroles), andaromatic compounds. It has been shown that the degree ofimprovement in the subsequent HDS is directly proportional tothe degree of nitrogen removal.

Several studies have shown that some activated carbons canhave much higher adsorption capacities for the nitrogencompounds than activated alumina15,24 and silica gel.15 Mochidaand co-workers published several papers on adsorptive deni-trogenation of a real gas oil over activated carbon materials.15,26They found that MAXSORB-II, an activated carbon preparedfrom petroleum coke through KOH activation with an apparentsurface area of 3000 m2/g, was effective for adsorptive deni-trogenation of gas oil at ambient temperatures. They attributedthis to the oxygen functional groups on the carbon surface inparticular those evolve CO in the temperature range of 600-800C. However, the absence of detailed analytical data for theconcentration of the coexisting sulfur and aromatic compoundsin the paper26 makes it difficult to compare the adsorptiveselectivity of different adsorbents for various nitrogen, sulfur,and aromatic compounds in the fuel, which is important in orderto clarify the adsorptive denitrogenation mechanism on carbonsurface.

It seems that the activated carbon is a potential adsorbent,which has shown higher adsorption capacity for removing thenitrogen compounds among all adsorbent materials reported inliterature. Adsorption capacity, selectivity, and regenerabilityof the adsorbent are the three critical factors for its practicalapplication in industry, in which adsorption selectivity andregenerability of the adsorbent are especially important incommercialization of a successful adsorption process. However,there is relatively limited information of adsorption selectivityand regenerability of activated carbons for the nitrogen com-pounds, in the available literature in comparison with theiradsorption capacity. The adsorption mechanism of the nitrogencompounds on activated carbon is also unclear.

In the present study, adsorption denitrogenation of a modeldiesel fuel (MDF) was conducted over seven commercialactivated carbon samples, as well as three activated aluminasamples for comparison purpose, in both batch and flowadsorption systems. The adsorption capacity and selectivity ofdifferent adsorbents for different compounds were quantitativelymeasured. The regeneration method and regenerability of thespent activated carbons were also explored. The adsorptionperformance and regenerability of various activated carbonswere further correlated with their physicochemical propertiesof the surface to understand the contribution of each propertyto the adsorption performance.

2. Experimental Section

Materials. Different types of commercial activated carbonsamples that were produced from a variety of carbon sourcematerials by different activation methods were examined in orderto study the structure-performance relationship. Table 1 showsthe sources and some manufacturing parameters for all activatedcarbons used in this study.

Three alumina samples, strong acid alumina, weak acidalumina, and basic alumina, were purchased from AldrichChemical Co. These activated alumina samples had the similarphysical properties. The average surface area and pore diameterof these samples were 160 m2/g and 5.8 nm, respectively.

Before use in experiments, all activated carbon samples werewashed by deionized water, and then heated at 110 C in avacuum oven overnight for drying.

952 Ind. Eng. Chem. Res., Vol. 48, No. 2, 2009

Characterization of Samples. Textural characterization ofthe activated carbon samples was performed by the adsorptionof N2 at 77 K using the Autosorb-1 MP system (QuantachromeCorp.). Before the N2 adsorption, the samples were subjectedto degassing at 200 C under turbomolecular vacuum. Thesurface areas were obtained from the N2 adsorption data atrelative pressures 0.05 < P/P < 0.2 using the BET equation.The total pore volumes (Vtotal) were estimated from the volumeof N2 (as liquid) held at a relative pressure (P/P) of 0.98. Thepore-size distributions were evaluated using nonlocal densityfunctional theory (DFT)31 dedicated to nitrogen (77.4 K)adsorption on carbon materials with slit-like pores.

Elemental analysis of all samples was determined using aLECO CHN-600 instrument. The activated carbons were alsoanalyzed by the temperature-programmed desorption (TPD) ata temperature increase rate of 10 C/min from 40 C to 950 Cunder a He flow of 50 mL/min. The total organic oxygen contentin various activated carbons was estimated by the total evolvedamount of CO and CO2. This method has been reported tocorrelate well with the oxygen content obtained by elementalanalysis.32

Model Diesel Fuel. In order to compare the adsorptionselectivity of various activated carbon samples for nitrogen,sulfur, and aromatic compounds, a model diesel fuel (MDF),containing the same molar concentration (10.0 mol/g) ofdibenzothiophene (DBT), 4,6-dimethyl-dibenzothiophene (4,6-DMDBT), indole, quinoline, naphthalene (NA), 1-methylnaph-thalene (1-MNA), and fluorene, was prepared using decane asa solvent. The detailed composition of the MDF is listed in Table2. The total nitrogen and sulfur concentrations of the MDF were280 and 641ppmw, respectively. In order to determine the timeneeded to reach the adsorption equilibrium, a model fuel (MF)containing equimolar concentration (10.0 mol/g) of quinolineand indole, corresponding to a total nitrogen concentration of280 ppmw, was used. All chemicals for preparation of the MDF

and MF were purchased from Aldrich Chemical Co. and usedwithout further purification. The purity of each chemical is alsolisted in Table 2.

Adsorption Experiments. For adsorption in a stirred batchsystem, about 20 g of MDF (or MF) and 0.20 g of the testedadsorbent were added into a glass tube. The tube was cappedand placed in an Omni-Reacto Station batch system (BarnsteadInternational, USA) at room temperature with electromagneticstirring at 300 rpm. After the desired time was reached, themixture was filtered, and the treated MDF samples wereanalyzed to estimate the adsorption capacity and selectivity ofthe adsorbents for various compounds in the fuel. The amountadsorbed, q (mmol/g), was calculated according to the followingequation:

q)L(Ci -Ce)

M

where L is the liquid fuel weight (g), Co, and Ce are the initialand equilibrium concentrations of the solute in the liquid fuel(mmol/g), respectively, and M is the amount of the adsorbentused (g).

In order to quantitatively estimate the adsorption selectivityof various adsorbents for each compound in MDF, a selectivityfactor was used,33 which is expressed as follows:

Ri-r )qi qr

Ce,i Ce,r

where qi and qr are the adsorption capacities of the compoundi and the reference compound r at equilibrium, respectively.Ce,i and Ce,r are the equilibrium concentrations of compound iand the reference compound r, respectively. A two-ring aromaticcompound, naphthalene (NA) was selected as a referencecompound in this study.

On the basis of the screening adsorption experiments, threerepresentative activated carbon samples; AC3, AC4, and AC6,which had quite different adsorption performance, were selectedfor further examination in a fixed-bed flowing system. Theseactivated carbons were packed respectively in a stainless steelcolumn (diameter: 4.6 mm; length: 150 mm). The packedcolumns were pretreated by passing N2 gas at 200 C for 2 hfor drying. After pretreatment, the column temperature wasdecreased to room temperature, and the MDF was then fed intothe adsorbent column using an HPLC pump in a flow-up modeat a liquid hourly space velocity (LHSV) of 4.8 h-1. The treatedMDF was periodically sampled every 15-20 min, until thesaturation point was reached.

Regeneration of Spent Adsorbents. In order to explore theregenerability of the activated carbons, the activated carbon aftersaturation in the fixed-bed adsorption experiment was subjectedto a regeneration test. Toluene was used as a solvent to washout the adsorbates from the spent activated carbons at 80 Cand 4.8 h-1 LHSV. The washing continued until the nitrogenand sulfur concentrations in the eluted toluene were close to

Table 1. Source and Some Manufacturing Parameters of the Studied Activated Carbonscarbon ID source activation commercial ID maker particle size, D50

AC1 coconut steam PCB-G calgon 45 mAC2 wood chemical nuchar AC20 WESTVACO ultrafineAC3 wood chemical nuchar AC1500 WESTVACO ultrafineAC4 pet. coke KOH maxsorb1 kansai ultrafineAC5 lignite coal steam darco-G60 norit 40 mAC6 coal steam norit SA 4 PAH norit 38 maAC7 wood WESTVACO ultrafine

a AC7 was obtained by heat treatment of AC3(Nuchar AC1500) with N2 at 550 C for one hour.

Table 2. Composition of the Model Diesel Fuel (MDF)purity concentration molar concentration

chemicals (wt%) (wt%) S or N (ppmw) (mol/g)sulfur compounds

DBT 0.98 0.184 320 10.04,6-DMDBT 0.97 0.212 320 10.0total 641

nitrogen compoundsindole 0.99 0.117 140 10.0quinoline 0.98 0.129 140 10.0total 280

aromaticsNA 0.99 0.128 10.01-MNA 0.95 0.142 10.0fluorene 0.99 0.166 10.0

paraffindecane 0.99 97.850n-Tetradecane 0.99 0.057 2.9others 1.014total 100.000

Ind. Eng. Chem. Res., Vol. 48, No. 2, 2009 953

zero. After washing with toluene, the system was purged withN2 at 200 C for 8 h to remove the remaining solvent, and thetemperature of the adsorption bed was reduced to roomtemperature for adsorption test of the regenerated adsorbent.

Analysis of Treated MDF Samples. All treated MDFsamples were analyzed by a gas chromatography, Varian CP3800, equipped with a flame ionization detector (FID) and aCP-8400 autosampler. The compounds were separated by aVF-5 ms capillary column (30-m length, 0.25-mm internaldiameter, and 0.25-mm film thickness) (Varian). The oventemperature was initially set to 100 C and ramped immediatelyat 5 C/min to 170 C, followed by a ramp at 20 C /min to290 C and held at this temperature for 2 min. Helium gas wasused as a carrier gas at a flow rate of 1.0 mL/min. Thetemperature of both injector and detector was 290 C. In thisanalysis, n-tetradecane was used as an internal standard. Forquantitative analysis of total nitrogen and sulfur concentrations(ppmw) in the treated MDF, an Antek 9000 series nitrogen andsulfur analyzer was used. A detailed analysis of the methodhas been previously reported.34

3. Results and Discussion

3.1. Effect of Adsorption Conditions on Adsorption Ca-pacity. Adsorptive denitrogenation of model fuel containingequimolar concentration of quinoline and indole was conductedat 25 C and at a fuel/adsorbent weight ratio of 100 in the batchsystem over two representative activated carbons, AC1 and AC3.AC1 had the highest percentage of microporosity (81%),whereas AC3 had the highest percentage of mesoporosity (52%)in porous distribution among the tested activated carbon samples(See Table 3). As shown in Figure 1, the adsorption capacityfor nitrogen on both carbons increased sharply within the first5 min. In fact, over 95% of the saturation adsorption capacitywas achieved within 20 min for AC1, and only half of this timewas needed for AC3 to achieve 95% of the saturation adsorption

capacity. It indicates that the activated carbon with 52% ofmesoporosity favors the diffusion of the nitrogen compoundsinto the pores in comparison with one with 19% of mesoporosityin the liquid phase adsorption. After the initial 10-20 min, theadsorption uptake for both carbons increased slowly and thenremained nearly constant after 1 h. The results imply that theadsorption approached the equilibrium after 1 h.

The effect of adsorption temperature on the adsorptioncapacity of AC3 was also examined at an adsorption time of4 h and at a fuel/adsorbent weight ratio of 100 in the batchsystem in a temperature range of 25-100 C, as shown in Figure2. The adsorption capacity decreased slightly at a temperaturerange of 25-75 C. Beyond 75 C, the adsorption capacitydecreased significantly with increasing temperature, indicatingthat the adsorption capacity was determined by the thermody-namics under this condition.

The observed effects of adsorption time and temperature onadsorption capacity imply that at 25 C and adsorption timelonger than 1 h, effect of the diffusion on the adsorption capacityfor the nitrogen compounds is negligible. According to theseresults, the adsorption at 25 C, for 4 h with the fuel/adsorbentweight ratio of 100 was selected as the standard test conditionfor all batch adsorption experiments.

3.2. Comparison of Adsorption Performance of VariousAdsorbents. In order to compare the adsorption performanceof the various adsorbents, all activated carbon samples andactivated alumina samples were tested in the batch system atthe standard condition by using MDF.

Figure 3 shows the adsorption capacity of the three activatedalumina adsorbents, including strong acidic, weak acidic, andbasic activated aluminas. It is clear that all three activated

Table 3. Textural Properties of the Activated Carbons

carbonSBET

(m2/g)Smicro

(m2/g)Smeso

(m2/g)Vtotal

(cm2/g)Vmicro

(cm2/g)Vmeso

(cm2/g)mean poresize (nm)

AC1 993 807 186 0.560 0.455 0.105 2.26AC2 1603 926 677 1.227 0.709 0.518 3.06AC3 2320 1125 1195 1.638 0.794 0.844 2.82AC4 2263 1655 608 1.206 0.882 0.324 2.13AC5 650 435 215 0.321 0.215 0.106 1.98AC6 1151 799 352 0.637 0.442 0.195 2.21AC7 2190 1360 830 1.559 0.968 0.591 2.85

Figure 1. Effect of adsorption time on adsorption capacity of two activatedcarbons, AC1 and AC3, for nitrogen removal at 25 C.

Figure 2. Effect of adsorption temperature on adsorption capacity of AC1fornitrogen removal, adsorption time 4 h.

Figure 3. Adsorption capacity for various activated aluminas; at 25 C,4 h adsorption time, and fuel/adsorbent weight ratio of 100 g-MDF/g-A.

954 Ind. Eng. Chem. Res., Vol. 48, No. 2, 2009

alumina adsorbents, regardless of their acidic nature, showhigher adsorption capacities for both quinoline and indole thanother compounds in MDF.

The acidity/basicity of the activated aluminas also played animportant role in determining their adsorption capacity. Theadsorption capacity for quinoline increased with increase inadsorbent acidity, while the adsorption capacity for indoleincreased with the increase in adsorbent basicity. The former isconsistent with the basicity of quinoline and the latter indicatesthat indole has a weakly acidic character. However, it shouldbe pointed out that the adsorption capacity does not simplydepend on the acid-base interaction, as the weakly acidicactivated alumina even has higher adsorption capacity fornonbasic indole than the basic quinoline.

Figure 4 presents the adsorption capacity of the activatedcarbons. In comparison with the activated aluminas, most ofthe activated carbon samples, except AC5 and AC6, havesignificantly higher adsorption capacity for the nitrogen com-pounds. AC3 has even about 2.5 higher capacity than thoseof the activated aluminas. The higher adsorption capacity maybe related to much higher surface area (2320 m2/g) and morefunctional groups on the surface of AC3 than those (160 m2/g)of the activated aluminas. The adsorption capacity of the variousactivated carbons for the nitrogen compounds are quite different,decreasing in the order of AC3 > AC4 > AC2 > AC1 > AC5 AC6, regardless quinoline or indole.

Figure 5 shows the adsorption selectivity of various activatedcarbon samples for each compound in comparison with the weakacidic activated alumina. The results show that the activated

alumina has the moderate adsorption selectivity for the nitrogencompounds (15-20), but very lower adsorption selectivity forthe sulfur and aromatic compounds (

Three of these factors work together to determine the adsorptionperformance of the adsorbents. Fundamental understanding ofthe roles played by these physical and chemical properties ofthe activated carbons is essential in development of novelcarbon-based adsorbents for adsorptive denitrogenation.

In order to understand the effect of surface textural structureon adsorption capacity, the textural properties of the activatedcarbon samples were characterized by the N2 adsorption, andthe results are listed in Table 3. AC2, AC3, and AC7 combinethe micropores and mesopores with an average pore sizessignificantly larger than others. The microporosity is dominantin AC1, AC4, AC5, and AC6, particularly in AC1, in whichthe micropore volume accounted for over 81% of the total porevolume. The surface areas of AC3 and AC4 were nearly twicethat of AC1 and AC6. The micropore surface areas and themicropore volumes for AC1 and AC6 were almost identical.

The adsorption capacity of the AC samples as a function oftotal surface area, mesoporous area, and microporous area,respectively, is shown in Figure 6A, and as a function of totalporous volume, mesoporous volume, and microporous volume,respectively, is shown in Figure 6B. No good relationshipbetween the adsorption capacity and microporous surface area(or mesoporous surface) was observed, indicating that themicropore in the studied samples may not play an importantrole in determination of their adsorption capacity for the nitrogencompounds. There is an increasing trend of the adsorptioncapacity with increase of total surface area. However, it wasnoted that such correlation is not perfect, as the accessiblesurface is not the sole factor that determines the adsorptioncapacity. The adsorption capacity is also unlikely determinedby the pore volume or micropore volume, as the evaluatedcoverage of AC surface at the saturate adsorption is less than70%. For example, the surface area of AC1 is even 15% lessthan AC6 and both have almost the same micropore volume.However, the adsorption capacity of AC1 is almost twice thatof AC6. The results imply that the factors of both, the site

density and the property of the adsorption site, rather thansurface physical property, may play a more important role indetermining the adsorption capacity of the adsorbent for thenitrogen compounds.

3.4. Effect of Surface Chemistry on Adsorption Perfor-mance. It is well-known that the activated carbon surfacechemistry plays an important role in many selective adsorptionprocesses. In order to study the effect of surface chemistry ofactivated carbon on the adsorption performance, the elementalanalysis and TPD characterization of the activated carbonsamples were conducted and the results are listed in Table 4.The oxygen concentration of the activated carbon samplesincreased from 2.4 to 8.4 wt % in the order of AC6 < AC5 67%) of microporous structure, such as AC1, AC4, AC5,and AC6, give excellent correlation between the adsorptioncapacity and the oxygen concentration with R2 value of 0.99,whereas the activated carbon samples with lower percentages(

than 100% (C/Co ) 2.0), which was also observed in ourprevious study.24 After reaching this maximum value, the outletconcentration gradually decreases to the initial concentrationwhen the subsequent breakthrough compound reaches thesaturation point (C/Co ) 1.0). This phenomenon was observedfor most compounds in MDF over AC3 and AC4 except forquinoline and indole. This interesting phenomenon may beexplained as follows: NA, 1MNA, FLRN, DBT, 4,6-DMDBT,and quinoline were adsorbed, at least partially, on the sameadsorption sites. The later breakthrough compound may have astronger adsorption affinity on the sites than the prior break-through compounds, thus the later breakthrough compoundkicks off the prior adsorbates from the adsorption site throughthe competitive adsorption, resulting in higher concentration ofthe prior breakthrough compounds at outlet than the initial one.As observed in Figure 8, 1-MNA displaces the adsorbed NApartially; FLRN displaces 1-MNA partially; DBT displacesFLRN partially; 4,6-DMDBT displaces DBT partially andquinoline partially displaces 4,6-DMDBT. It needs to point outthat such displacement extent is only about 10-60%, dependingon the adsorption affinity of the compounds. Interestingly, indoledoes not appear to displace the adsorbed quinoline, althoughthe breakthrough of indole was after the breakthrough ofquinoline. This finding suggests that the adsorption sites forindole maybe different from those for quinoline and/or theinteraction between quinoline and the adsorption site is toostrong to be displaced by indole. Further investigation isnecessary to understand the findings.

The breakthrough curves for different compounds over AC4are shown in Figure 9. The breakthrough amount for differentcompounds increased in the order of NA < 1-MNA < fluorene< quinoline DBT indole , 4,6-DMDBT, which is quitedifferent from that over AC3. The corresponding breakthroughcapacity (mmol-compounds/g-A) is 0.23, 0.29, 0.44, 0.53, 0.57,0.60, and 0.96, respectively, for NA, 1-MNA, fluorene, quino-line, DBT, indole, and 4,6-DMDBT. In comparison with theresults for AC3, a significant difference is that AC4 has muchhigher adsorption capacity and selectivity for 4,6-DMDBT thanAC3, while AC3 has a much higher adsorption capacity andselectivity for indole. Considering the similar oxygen contentin AC3 (7.3 wt %) and in AC4 (8.4 wt%), the result clearlyindicates that the type of the oxygen functional groups on surfaceis critical in determining the adsorption capacity and selectivityof the activated carbon for different compounds. As well-known,4,6-DMDBT is a major refractory sulfur compound existing inthe commercial diesel fuel due to its low HDS reactivity. Thus,

AC4 is a promising adsorbent for removing this type of sulfurcompounds for ultradeep desulfurization of diesel fuel.

The breakthrough curves for different compounds over AC6are shown in Figure 10. The breakthrough curve for differentcompounds increased in the order of NA < 1-MNA < fluorene quinoline < indole < DBT , 4,6-DMDBT, which is differentfrom the breakthrough curves of both AC3 and AC4. Thecorresponding breakthrough capacity (mmol-compounds/g-A)is 0.08, 0.09, 0.13, 0.13, 0.17, 0.22, and 0.39, respectively, forNA, 1-MN, fluorene, quinoline, indole, DBT, and 4,6-DMDBT.In comparison with the results from AC3 and AC4, thesignificant difference is that AC6 has much low adsorptioncapacity for all tested compounds and has much low selectivityfor quinoline. In addition, over 28% of the quinoline adsorbedover AC6 was displaced by indole and others, while almost noadsorbed quinoline over AC3 and AC4 was displaced by indole.It indicates that the interaction between quinoline and the surfaceof AC6 is very weak. All of these can be contributed to thelow concentration of oxygen functional groups on the surfaceof AC6, as indicated in Table 4.

3.6. Regeneration of Spent Activated Carbons. The re-generability of adsorbent is crucial in a practical adsorptionprocess. In the present study, regeneration of the spent adsor-bents was conducted by toluene solvent washing in a fixed bedat 80 C and at 4.8 h-1 LHSV, followed by heating of theadsorbent bed to 200 C under a N2 flow to remove theremaining solvent. In order to estimate the required amount ofthe solvent for the regeneration, the measured total nitrogencontent and sulfur content in the effluent as a function of theamount of the used effluent for AC3, AC4, and AC6 is shownin Figure 11. In comparison of the three spent AC samples, theadsorbed sulfur and nitrogen compounds are easier to beremoved from the spent AC6 than AC3 and AC4, as AC6contains less oxygen function groups on the surface, and thushas less adsorption affinity for the sulfur and nitrogen com-pounds. For the spent AC3 and AC4, about 15 g of toluenesolvent is needed to remove the majority of the adsorbednitrogen compounds from 1 g of the spent adsorbent. The mostdifficult is to remove the adsorbed sulfur compounds from AC4.It is probably because of the higher adsorption affinity of AC4for 4,6-DMDBT and/or high diffusion barrier of 4,6-DMDBTin desorption due to the dominant micropores in AC4.

The evaluation of the first and second regenerated adsorbentswas conducted in the same fixed-bed adsorption system. Figures12, 13, and 14 show the breakthrough curves for total nitrogenand total sulfur over the first and second regenerated activated

Figure 9. Breakthrough curves of various compounds in MDF over AC4at 25 C and 4.8 h-1 LHSV.

Figure 10. Breakthrough curves of various compounds in from MDF overAC6 at 25 C and 4.8 h-1 LHSV.

958 Ind. Eng. Chem. Res., Vol. 48, No. 2, 2009

carbons of AC3, AC4, and AC6, respectively, in the comparisonwith the fresh ones. The adsorption performance of all theregenerated activated carbons coincided well with the fresh ones.It clearly indicates that all three activated carbons, regardlessof their physical and chemical nature, can be successfullyregenerated by using the toluene washing following by theheating to remove toluene.

4. Conclusions

Adsorption performance of seven commercial activatedcarbon samples and three activated alumina samples forremoving nitrogen compounds, including quinoline and indole,were evaluated using a model fuel in a batch adsorption systemand a flowing fixed-bed adsorption system. The activated carbonsamples show higher capacity than the activated aluminasamples for removing the nitrogen compounds.

Some activated carbons were found to show much higheradsorption capacity and selectivity for indole and quinoline, as

reflected by the increasing adsorption selectivity of AC3 in theorder of NA < 1-MNA < fluorene < DBT < 4,6-DMDBT