rapid adsorption of alcohol biofuels by high surface area mesoporous carbons

8
Rapid adsorption of alcohol biofuels by high surface area mesoporous carbons Thomas J. Levario, Mingzhi Dai, Wei Yuan, Bryan D. Vogt 1 , David R. Nielsen Chemical Engineering, Arizona State University, 501 E. Tyler Mall, ECG 301, Tempe, AZ 85287-6106, United States article info Article history: Received 28 June 2011 Accepted 1 August 2011 Available online 7 August 2011 Keywords: Biofuels Mesoporous carbon Adsorption In situ product recovery n-Butanol abstract Surfactant templated mesoporous carbons were evaluated as biofuel adsorbents through characterization of equilibrium and kinetic behavior for both ethanol and n-butanol. Variations in synthetic conditions enabled facile tuning of specific surface area (500–1300 m 2 /g) and pore morphology (hexagonally packed cylindrical or BCC spherical pores). n-Butanol was more effectively adsorbed than ethanol for all meso- porous carbons, suggesting a mechanism of hydrophobic adsorption. The adsorbed alcohol capacity increased with elevated specific surface area of the adsorbents, irrespective of pore morphology. While adsorption capacity of these mesoporous carbons is comparable to commercially-available, hydrophobic polymer adsorbents of similar surface area, the pore morphology and structure of mesoporous carbons greatly influenced adsorption rates, enhancing them by up to 1–2 orders of magnitude over commercial polymer adsorbents. Multiple cycles of adsorbent regeneration did not impact the adsorption equilibrium or kinetics. The high chemical and thermal stability of mesoporous carbons provide potential significant advantages over other commonly examined biofuel adsorbents, such as polymers and zeolites. Ó 2011 Elsevier Inc. All rights reserved. 1. Introduction Emerging environmental issues, ever-shrinking petroleum re- serves, and unstable fossil fuel costs continue to spur interest in the development of sustainable biofuels from renewable feed- stocks. Biomass-derived fuels presently provide only 10% of total global energy demand and represent an even smaller fraction of the US liquid transportation fuel demand (only 3% in 2009) [1]. Re- cent congressional mandates, however, will require annual biofuel production to be increased from 11 billion gal. in 2009 to 36 billion gal. by 2022 [2]. Meeting such an aggressive production target will require the development of next generation biofuels and novel tech- nologies to support their sustainable and economical production. Among conventional liquid biofuels, ethanol remains the most ac- tively pursued molecule as a result of its excellent physicochemical characterization and the technological maturity associated with its fermentative production. In recent years, however, a strong interest in biologically-derived n-butanol (a natural fermentation product of many Clostridium sp.) as a liquid transportation fuel alternative has re-emerged as a result of its favorable physical and thermodynamic properties [3,4]. Attributes of n-butanol include a higher energy density and lower water solubility relative to ethanol; these proper- ties provide greater compatibility with conventional engines and fuel distribution infrastructure (see Table 1). However, a critical challenge that limits the development and viability of all alcohol biofuel fermentations is feedback inhibition caused by product toxicity at relatively low concentrations [5–7]. For example, feedback inhibition limits maximal ethanol produc- tion to final titers of below approximately 21%, 12%, and 6% (w./ v.) for the ethanologenic microbes Saccharomyces cerevisiae [8], Zymomonas mobilis [9], and Escherichia coli [10], respectively. This inhibition is even worse for the more hydrophobic n-butanol (Ta- ble 1), where titers as low as 1.3% (w./v.) can induce feedback inhibition towards Clostridium acetobutylicum [5]. This feedback limits conventional biofuel fermentations to very dilute aqueous feeds for downstream product recovery and purification processes [11], which leads to inefficient and generally energy intensive product separation. This separation step is often a large contributor to poor economic viability of bioprocesses (behind only feedstock costs) [12]. The large energy demand and expense associated with this separation is driven by the use of distillation for conventional ethanol and n-butanol separations from fermentation broths; for example, the energy cost of distillative recovery can be 10% of the value of the produced ethanol [13]. Moreover, n-butanol recov- ery by distillation is further challenged by its low vapor pressure (3-fold less than water at 25 °C), which necessitates the use of multi-stage designs to ultimately achieve acceptable product pur- ity [14]. Furthermore, distillation is an inherently ill-suited tech- nology for integration with continuous bioprocesses due to the thermal sensitivity displayed by both cells and essential media components (e.g., carbohydrate substrates). Hence, the application of distillation for alcohol biofuel recovery is regulated to the sup- port of batch processes and is not a viable option for achieving 1387-1811/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.micromeso.2011.08.001 Corresponding author. Tel.: +1 480 965 4113; fax: +1 480 727 9321. E-mail address: [email protected] (D.R. Nielsen). 1 Present address: Department of Polymer Engineering, University of Akron, 250 South Forge Street, Akron, OH 44325-0301, United States. Microporous and Mesoporous Materials 148 (2012) 107–114 Contents lists available at SciVerse ScienceDirect Microporous and Mesoporous Materials journal homepage: www.elsevier.com/locate/micromeso

Upload: independent

Post on 29-Apr-2023

1 views

Category:

Documents


0 download

TRANSCRIPT

Microporous and Mesoporous Materials 148 (2012) 107–114

Contents lists available at SciVerse ScienceDirect

Microporous and Mesoporous Materials

journal homepage: www.elsevier .com/locate /micromeso

Rapid adsorption of alcohol biofuels by high surface area mesoporous carbons

Thomas J. Levario, Mingzhi Dai, Wei Yuan, Bryan D. Vogt 1, David R. Nielsen ⇑Chemical Engineering, Arizona State University, 501 E. Tyler Mall, ECG 301, Tempe, AZ 85287-6106, United States

a r t i c l e i n f o a b s t r a c t

Article history:Received 28 June 2011Accepted 1 August 2011Available online 7 August 2011

Keywords:BiofuelsMesoporous carbonAdsorptionIn situ product recoveryn-Butanol

1387-1811/$ - see front matter � 2011 Elsevier Inc. Adoi:10.1016/j.micromeso.2011.08.001

⇑ Corresponding author. Tel.: +1 480 965 4113; faxE-mail address: [email protected] (D.R. Nie

1 Present address: Department of Polymer EngineerSouth Forge Street, Akron, OH 44325-0301, United Sta

Surfactant templated mesoporous carbons were evaluated as biofuel adsorbents through characterizationof equilibrium and kinetic behavior for both ethanol and n-butanol. Variations in synthetic conditionsenabled facile tuning of specific surface area (500–1300 m2/g) and pore morphology (hexagonally packedcylindrical or BCC spherical pores). n-Butanol was more effectively adsorbed than ethanol for all meso-porous carbons, suggesting a mechanism of hydrophobic adsorption. The adsorbed alcohol capacityincreased with elevated specific surface area of the adsorbents, irrespective of pore morphology. Whileadsorption capacity of these mesoporous carbons is comparable to commercially-available, hydrophobicpolymer adsorbents of similar surface area, the pore morphology and structure of mesoporous carbonsgreatly influenced adsorption rates, enhancing them by up to 1–2 orders of magnitude over commercialpolymer adsorbents. Multiple cycles of adsorbent regeneration did not impact the adsorption equilibriumor kinetics. The high chemical and thermal stability of mesoporous carbons provide potential significantadvantages over other commonly examined biofuel adsorbents, such as polymers and zeolites.

� 2011 Elsevier Inc. All rights reserved.

1. Introduction

Emerging environmental issues, ever-shrinking petroleum re-serves, and unstable fossil fuel costs continue to spur interest inthe development of sustainable biofuels from renewable feed-stocks. Biomass-derived fuels presently provide only �10% of totalglobal energy demand and represent an even smaller fraction ofthe US liquid transportation fuel demand (only�3% in 2009) [1]. Re-cent congressional mandates, however, will require annual biofuelproduction to be increased from�11 billion gal. in 2009 to 36 billiongal. by 2022 [2]. Meeting such an aggressive production target willrequire the development of next generation biofuels and novel tech-nologies to support their sustainable and economical production.Among conventional liquid biofuels, ethanol remains the most ac-tively pursued molecule as a result of its excellent physicochemicalcharacterization and the technological maturity associated with itsfermentative production. In recent years, however, a strong interestin biologically-derived n-butanol (a natural fermentation product ofmany Clostridium sp.) as a liquid transportation fuel alternative hasre-emerged as a result of its favorable physical and thermodynamicproperties [3,4]. Attributes of n-butanol include a higher energydensity and lower water solubility relative to ethanol; these proper-ties provide greater compatibility with conventional engines andfuel distribution infrastructure (see Table 1).

ll rights reserved.

: +1 480 727 9321.lsen).ing, University of Akron, 250tes.

However, a critical challenge that limits the development andviability of all alcohol biofuel fermentations is feedback inhibitioncaused by product toxicity at relatively low concentrations [5–7].For example, feedback inhibition limits maximal ethanol produc-tion to final titers of below approximately 21%, 12%, and 6% (w./v.) for the ethanologenic microbes Saccharomyces cerevisiae [8],Zymomonas mobilis [9], and Escherichia coli [10], respectively. Thisinhibition is even worse for the more hydrophobic n-butanol (Ta-ble 1), where titers as low as �1.3% (w./v.) can induce feedbackinhibition towards Clostridium acetobutylicum [5]. This feedbacklimits conventional biofuel fermentations to very dilute aqueousfeeds for downstream product recovery and purification processes[11], which leads to inefficient and generally energy intensiveproduct separation. This separation step is often a large contributorto poor economic viability of bioprocesses (behind only feedstockcosts) [12]. The large energy demand and expense associated withthis separation is driven by the use of distillation for conventionalethanol and n-butanol separations from fermentation broths; forexample, the energy cost of distillative recovery can be �10% ofthe value of the produced ethanol [13]. Moreover, n-butanol recov-ery by distillation is further challenged by its low vapor pressure(�3-fold less than water at 25 �C), which necessitates the use ofmulti-stage designs to ultimately achieve acceptable product pur-ity [14]. Furthermore, distillation is an inherently ill-suited tech-nology for integration with continuous bioprocesses due to thethermal sensitivity displayed by both cells and essential mediacomponents (e.g., carbohydrate substrates). Hence, the applicationof distillation for alcohol biofuel recovery is regulated to the sup-port of batch processes and is not a viable option for achieving

Table 1Comparison of physiochemical properties of conventional alcohol biofuels.

Biofuel compound Ethanol n-Butanol

Molecular weight 46.1 74.1Specific gravity @ 20 �C 0.79 0.81Vapor pressure (atm) @ 25 �C 0.0312 0.0109Normal boiling point (�C) 78 117Log KO/W

a �0.26 0.8Water solubility @ 20 �C (g/L) 1 77Energy density (MJ/L) 19.6 29.2

a Higher values indicate greater hydrophobicity.

108 T.J. Levario et al. / Microporous and Mesoporous Materials 148 (2012) 107–114

in situ product recovery (ISPR) as a means to eliminate feedbackinhibition against producing microbes [15].

Selective adsorption, so-called solid phase extraction (SPE), pro-vides an alternative, low energy approach to separate biofuel com-pounds from dilute aqueous fermentation broths [16–19]. Alcoholsare adsorbed from aqueous media via hydrophobic interactions(i.e. Van der Waals forces), which occur between the surface ofthe sorbent matrix and the alkyl segment of the alcohol [20]. Priorstudies demonstrate that increasing the hydrophobicity of theadsorbent increases its affinity for alcohol adsorption [21–24]. Thisprinciple has been explored with regards to numerous differentclasses of hydrophobic sorbents including: synthetic polymer res-ins [21,22], zeolites [23,24], silica gels [25], and activated carbons[26]. To improve performance, there have been efforts to increaseadsorbent hydrophobicity through selecting more hydrophobicpolymers [21] or increasing the Si content in zeolites [23]. In addi-tion to hydrophobicity, the specific surface area of the adsorbentsignificantly impacts its efficacy by controlling the number ofhydrophobic adsorption ‘sites’ available for the biofuel[21,22,24,27,28]. For example, by comparing two adsorbents (Dia-ion HP-20 and Dowex™ Optipore™ L-493) with the same polymercomposition (poly(styrene-co-divinylbenzene): pSDVB), it was pre-viously shown that as the specific surface area increased from 500to 1100 m2/g, respectively, the corresponding achievable specificloading capacity for n-butanol increased from 1200 to4100 mmol/kg at a fixed aqueous n-butanol concentration of135 mM (or 1% (w./v.)) [22]. However, limitations exist regardingthe use of such polymeric adsorbents; in particular, with respectto their ease of regeneration and prospects for continued reuse.For instance, low thermal stability of pore structures can lead todegradation in re-use performance due to the requisite tempera-tures associated with recovery and purification of the alcohol. Inprior studies, the recovery of adsorbed n-butanol from pSDVB poly-mer resins [21] was found to be limited by the low glass transitiontemperature of pSDVB (95 �C), whereas the normal boiling point ofn-butanol is 117 �C (Table 1). Thus, the recovery temperature forremoving n-butanol from the pores must be less than its normalboiling point, which could adversely impact rates and extent ofrecovery. Conversely, although silicalite based adsorbents notablypossess high thermal stability [23], high silica content hydrophobiczeolites have been shown to suffer from hydrolysis and degrada-tion in aqueous systems [29], a shortcoming which diminishesthe long term prospects of zeolites for biofuel recovery from aque-ous solutions.

One possible solution to obtaining both high thermal and chem-ical stability with the desired attributes of high hydrophobicity andsurface area lies in the use of activated carbon, which presentlyfinds commercial applications in water purification. Activated car-bons have frequently been applied for the removal of organic pol-lutants from aqueous solutions including, for example, phenol [30]and methylene blue [31]. However, initial reports on the use ofactivated carbon for the separation of biofuel alcohols

demonstrated only marginal performance [26]. The observed poorbehavior was attributed to either: (1) low specific surface areas at-tained at low activation of the material, or (2) the broadening ofpore sizes at high activation. These observations suggest that thelack of precise control over specific surface area and nanopore sizelimit the adsorptive capacity of alcohol biofuels, such as ethanol,with activated carbon. However, mesoporous carbons can be tai-lored to possess well-defined and controllable pore sizes throughsynthetic and post-treatment methods. One such study showedthat polymer-derived carbon membranes can attain sub-angstromcontrol in pore size through altering the pyrolysis temperatureduring synthesis [32]. With these considerations in mind, it is pos-tulated that ordered mesoporous carbons (MPCs) with well-de-fined pores would exhibit improved performance for biofuelseparations since the hydrophobicity, pore connectivity, and sur-face area of MPCs can be engineered through processing and choiceof template [33–35]. In past works, the specific surface area for softtemplated, ordered MPCs has been reported to be as high as2580 m2/g [36]. Although ordered MPCs have been shown to beeffective for the separation of organic dyes from water [36,37], or-dered MPCs have still not been specifically developed and investi-gated as biofuel adsorbents.

In addition to achieving high specific loadings, effective adsor-bents must also rapidly separate biofuels from the aqueous phase.This is particularly true when developing adsorbents to effectivelyserve in ISPR applications, wherein biofuel molecules must be ad-sorbed from the culture medium at rates equal to or greater thantheir rate of biosynthesis to prevent the concentration from reach-ing inhibitory thresholds. For porous adsorbents, biofuel adsorp-tion kinetics are typically controlled by diffusion within thepores [24]. Self-assembly of colloids and surfactants enable thesynthesis of porous carbons with well-defined pore sizes andgeometries [38–41]. In particular, soft templating routes involvingcooperative assembly of carbonizable material with a surfactant[34] are particularly attractive as the obtained lyotropic morphol-ogy is controllable throughout the material composition. Heatingthe self-assembled organic composite in a non-oxidizing environ-ment leads to carbonization of the precursor and degradation ofthe surfactant to yield well-defined pores. The size of the hydro-phobic segments in the surfactant typically controls the pore size,but addition of a swelling agent allows for further tuning of thepore size with the same surfactant [42]. Furthermore, these meso-porous carbons have been shown to exhibit low cytotoxicity, sothey should not interfere with cellular processes [43]. These syn-thetic attributes make ordered MPCs highly desirable as adsor-bents for biofuel recovery and ISPR.

In this work, the utility of ordered MPCs as adsorbents for theefficient recovery of biofuels (specifically, ethanol and n-butanol)is systematically examined by variation of the surface area andpore morphology. The impact of these parameters on biofueladsorption equilibria and kinetics is explored to determine the effi-cacy of these MPC adsorbents under biologically-relevant condi-tions. The performance of these MPCs are directly compared toconventional and previously characterized polymeric adsorbentsto determine if they are viable alternatives for biofuel separations.

2. Materials and methods

2.1. Chemicals

Ethanol (95%, HPLC grade) and n-butanol (99.5%, reagent grade)were purchased from Fisher Scientific (Pittsburgh, PA). For the syn-thesis of the mesoporous carbons, triblock copolymer, PluronicF127 (PEO106-PPO70-PEO106, BASF, Florham Park, NJ), wasutilized to template the carbon-based composite mesostructures.

T.J. Levario et al. / Microporous and Mesoporous Materials 148 (2012) 107–114 109

Sodium hydroxide (NaOH), phenol (99.0%, ACS reagent grade), andformaldehyde (37 wt.% in H2O) were purchased from Sigma–Al-drich (St. Louis, MO) and used as received. Tetraethylorthosilicate(TEOS, Sigma–Aldrich, St. Louis, MO) served as a silica precursorand HCl (Sigma–Aldrich, St. Louis, MO) was used as a catalyst forthe condensation of TEOS. Resol, a low-molecular-weight, watersoluble phenolic resin, was prepared from the polymerization ofphenol and formaldehyde using a base (NaOH) catalyst as de-scribed previously [34].

2.2. Adsorbents

Dowex™ Optipore™ L-493 was purchased from Sigma–Aldrich(St. Louis, MO). The mesoporous carbons studied were synthesizedin house. Four different mesoporous carbons were investigated andfollow the naming protocol from our previous work [44]. Hexago-nally packed (P6mm) cylindrical mesopores were obtained by dis-solving resol and Pluronic F127 at molar composition of phenol/formaldehyde/NaOH/F127 = 1:2:0.1:0.012 in ethanol and subse-quent evaporation of the ethanol yielded a polymer powder la-beled FDU-15-800 [34]. Body centered cubic packed (Im3m)spherical mesoporous were obtained by dissolving resol and Plu-ronic F127 at molar composition of phenol/formaldehyde/NaOH/F127 = 1:2:0.1:0.006 in ethanol and subsequent evaporation ofthe ethanol yielded a polymer powder labeled FDU-16-800 [34].The phenolic resin in these polymeric powders (FDU-15-800 andFDU-16-800) was then thermally crosslinked at 120 �C for 24 h.Carbonization was performed in tubular furnace under nitrogenatmosphere with a flow rate of 140 cm3/min at 800 �C for 2 h withheating rates of 1 �C/min below 600 �C, and 5 �C/min above 600 �C.To increase the surface area of the mesoporous powders, silica viacondensation of TEOS was added to the synthesis solution and fol-lowed the procedures listed above. The silica was subsequently re-moved in 1 M NaOH after carbonization. Two differentcompositions were examined and labeled as CS-68-800 (precursorsolution: 0.208 g TEOS, 0.1 g resol, and 0.1 g Pluronic F127) and CS-81-800 (precursor solution: 0.416 g TEOS, 0.1 g resol, and 0.14 gPluronic F127).

2.3. Characterization of mesoporous carbons

The ordered mesostructure of the carbons was determined fromX-ray diffraction (XRD) in h/2h geometry with Cu Ka source (PAN-alytical X’Pert PRO, Almelo, The Netherlands). A parallel plate col-limator (PPC) was used in combination with an incident beamoptical module providing an X-ray beam with very low divergence.The angle of incidence, h, was varied from 0.25� to 1.5�. The poresof the carbon materials were visualized via transmission electronmicroscopy (TEM) using a JEOL 2010F microscope operating at200 kV. Nitrogen adsorption–desorption isotherms were measuredwith Tristar II 3020 (Micromeritics Instrument Corporation, Nor-cross, GA) at 77 K. Before the measurement, the samples were de-gassed at 300 �C for at least 1 h. Specific surface areas wereestimated using the Brunauer–Emmett–Teller (BET) method in arelative pressure range of P/P0 = 0.05–0.25. The pore size distribu-tion (PSD) and pore volume were calculated from the adsorptionbranch of the isotherm by using the Barrett–Joyner–Halenda(BJH) model.

2.4. Analytical methods

Aqueous alcohol solutions were analyzed via high-pressure li-quid chromatography (HPLC, 1100 series, Agilent, Santa Clara,CA). Separation was achieved on a ZORBAX Eclipse XDB-C18 col-umn (Agilent, Santa Clara, CA) operated isothermally at 50 �C usingwater as the mobile phase at a flow rate of 1.0 mL/min. Analytes

were detected using a refractive index detector, with concentra-tions determined through the use of external standards.

2.5. Isotherm determination

Equilibrium adsorption experiments were performed in sterile,2 mL glass HPLC vials containing 1 mL of aqueous alcohol solutionand 20 mg of carbon powder. Solutions were prepared in steriledeionized water at initial concentrations ranging from 1–120 g/L(22–2600 mM) for ethanol and from 1–70 g/L (13–944 mM) forn-butanol. Samples were equilibrated for 24 h at 37 �C while beingstirred at 250 rpm. After equilibration, the supernatant was re-moved for HPLC analysis via pipetting through plastic pipet tipspacked with DMCS treated glass wool (to eliminate carryover ofMPCs). The specific loading capacity (q) was then determined bythe following material balance relationship:

q ¼ ðCaq;o � CaqÞVaq

mð1Þ

where Caq,o and Caq are the initial and equilibrated concentrations ofalcohol in the aqueous phase, respectively, Vaq is the volume of theaqueous phase, and m is the mass of carbon powder. All experi-ments were performed at least in triplicate, with experimental errorbeing reported as one standard deviation.

2.6. Modeling adsorption isotherms

Adsorption data for each alcohol-adsorbent pair were subse-quently fit to the Freundlich isotherm model:

q ¼ kFðCaqÞ1=n ð2Þ

where kF is the Freundlich adsorption coefficient, and n is the Fre-undlich exponent. Model parameters were estimated via nonlinearleast squares regression of the experimental equilibrium data to Eq.(2) using MATLAB� and the intrinsic function nlinfit.

2.7. Adsorption kinetics: experiments and modeling

Dynamic adsorption experiments of n-butanol were performedusing the same experimental setup described above; however, afixed initial concentration of 10 g/L (equivalent to 135 mM) wasused to elucidate the kinetic parameters. Experiments were initi-ated by addition of 1 mL n-butanol solution to 20 mg of MPC. Foreach MPC, nine samples were prepared and performed in parallelto obtain time dependent data. These parallel samples facilitatedanalysis by minimizing disruption due to sampling time and elim-inating volume changes imposed by sampling. Samples were vigor-ously mixed at maximum speed on a Vortex Genie 2.0 (ScientificIndustries, Bohemia, New York) for up to 24 h. Preliminary exper-iments indicated that mass transfer limitations associated withbulk transport through the aqueous solutions were not rate limit-ing under these experimental conditions (data not shown). Aque-ous samples were removed from the MPCs at time intervalsranging between 15 s to several hours to obtain the kinetic data.

It was assumed that adsorption followed pseudo-first orderkinetics, which allowed the rate of n-butanol uptake to be closelyapproximated by:

dqdt¼ k1ðqeq � qtÞ ð3Þ

where k1 is the pseudo-first order kinetic constant, qeq is the equi-librium specific loading capacity, and qt is the specific loadingcapacity at time t. To facilitate comparisons between differentmaterials, a normalized scale represented as the fractional satura-tion (FS) was used, and could be calculated as:

110 T.J. Levario et al. / Microporous and Mesoporous Materials 148 (2012) 107–114

FS ¼ qt=qeq ð4Þ

FS values range between 0 and 1, representing no adsorption (theinitial condition) and full saturation (the final, equilibrium condi-tion), respectively.

2.8. Sorbent regeneration

Mesoporous carbon sorbents were routinely regenerated (up to10 times throughout the course of this study). Regeneration wasachieved utilizing a two-step protocol that consisted of: (1) dryingadsorbents at 110 �C for 24 h, and (2) heating at 170 �C for an addi-tional 24 h to further release any adsorbed species. It should benoted that these conditions were not optimized. An excessivetreatment was used to ensure full regeneration of the adsorbentsbefore their subsequent reuse in the equilibrium and kineticstudies.

3. Results and discussion

3.1. Physical properties of adsorbents

The relevant physical characteristics for each of mesoporouscarbon are summarized in Table 2. These materials exhibit surfaceareas between 500 and 1300 m2/g, which are consistent with priorreports in the literature [36]. The surface area for the CS materialsis significantly larger due to the presence of micropores in the car-bon wall that are generated by the removal of silica that is presentduring the synthesis of these materials. The pore size distributionfrom BET N2 sorption is relatively narrow for each material withthe average (peak) pore size listed in Table 2. The diameter forCS-68-800 is largest due to the decreased shrinkage of the resolduring carbonization from the reinforcing co-continuous silica inthe wall. However, further increasing the silica concentration(CS-81-800) actually leads to a decrease in the average pore sizepresumably due to the volume contraction associated with thecondensation of the TEOS precursor. To better illustrate the well-defined structure of these MPCs, Fig. 1 shows TEM micrographsof the different powders. Ordered arrays of monodisperse poresare clearly visible in the micrographs except for the CS-81-800(Fig. 1D). This decrease in the visible order could be a result ofthe high initial silica loading in the wall that has been removed.For a control and reference, hydrophobic, macroporous, polymeric(poly(styrene-co-divinylbenzene)) adsorbent Dowex™ Optipore™L-493 (hereafter referred to as L-493), with physical characteristicslisted in Table 2, is used as this adsorbent has been well character-ized for biofuel separation [21,22].

3.2. Equilibrium adsorption

Upon comparison with other commonly employed adsorptionmodels (e.g., Langmuir isotherm), the Freundlich isotherm model(Eq. (2)) best captures the qualitative concentration dependenceof biofuel adsorption by all MPCs examined (results not shown).

Table 2Characteristic attributes of adsorbents investigated.

Adsorbent ABET

(m2/g)Vp

(cm3/g)dp

(nm)Mesostructure

FDU-15-800 538 0.028 5.0 P6mm (cylindrical)FDU-16-800 671 0.14 5.8 Im3m (spherical)CS-68-800 1287 1.39 8.2 Im3m (spherical)CS-81-800 1307 1.26 7.2 Im3m (spherical)Dowex™ Optipore™ L-493 >1100a 1.16a 4.6a Disordered random

a As available and reported by supplier.

Therefore, all equilibrium adsorption data are fit to the Freundlichisotherm model, as illustrated in Fig. 2. Meanwhile, the corre-sponding ‘best-fit’ parameter estimates are provided and com-pared in Table 3. As is consistent with prior studies which havesimilarly explored the use of hydrophobic adsorbents for biofuelrecovery [22], the adsorption potential (i.e. the magnitude of theadsorbent-aqueous phase partitioning ratio) is observed to in-crease as the carbon chain length increases from ethanol (2C) ton-butanol (4C). For example at an aqueous phase concentrationof 27 mM (approximately 1 g/L ethanol or 2 g/L n-butanol), theadsorbent-aqueous phase partitioning ratio (defined as qeq/Caq) ofCS-81-800 increases more than 4-fold from 13 (mmol/kg)/(mM)for ethanol to 61 (mmol/kg)/(mM) for n-butanol. As n-butanol bylog KO/W is over 11-times more hydrophobic than ethanol (Table 1),the specific adsorption driving forces (i.e. Van der Waals interac-tions) for n-butanol are significantly larger than that for ethanoland leads to the larger partitioning ratio. Additionally, the relativemagnitudes of kF and n positively correlate with alcohol carbonchain length. This relationship is also consistent with prior studiesusing polymeric adsorbents such as L-493 [21,22] as shown in Ta-ble 3. Overall, the efficacy of n-butanol adsorption by MPCs is sub-stantially greater than that for ethanol; a finding which supportsthe hypothesis that alcohol biofuels are adsorbed by MPCs accord-ing to hydrophobic interactions, as well as the growing conclusionthat alcohol biofuel molecules of increasing carbon chain length(so-called ‘second generation biofuels’) show the greatest potentialfor adsorptive recovery [17,21,22,45]. The n-butanol isotherm(Fig. 3) illustrates that MPC adsorbents nearly match the adsorp-tion performance of polymer adsorbents with comparable specificsurface areas, such as L-493 (Table 2). Thus the observed similari-ties between the adsorption isotherms (as well as the kF values inTable 3) of CS-68-800, CS-81-800 and L-493 can be attributed totheir similar specific surface areas (Table 2). The importance of sur-face area can be elucidated by further comparison of the Freundlichisotherm behaviors of these materials. For instance, the Freundlichexponent is similar for FDU-15-800 and FDU-16-800, where thepores are solely templated by the surfactant and the surface areasare comparable. This relationship between the Freundlich expo-nent and surface area also holds for CS-68-800 and CS-81-800,where secondary pores are induced by etching of silica. Addition-ally, the relative magnitudes of the Freundlich exponent, whichquantifies the sensitivity of the equilibrium loading capacity tochanges in the aqueous phase concentration, for both FDU-15-800 and FDU-16-800 are greater than those found for both CS-68-800-800 and CS-81-800-800. The large kF values of both FDU-15-800 and FDU-16-800 result from the high nonlinearity of theiradsorption isotherms manifested as a ‘sharp bend’ in the isotherm.Specific surface area (Table 2) limits the number of alcohol mole-cules that can be adsorbed on the surface and, as a consequence,FDU-15-800 and FDU-16-800 adsorbents saturate at lower concen-trations of alcohol in the aqueous phase. To illustrate this effect, itcan be seen, for example, that increasing the aqueous n-butanolconcentration from 67 to 270 mM (equivalent to 5–20 g/L) in-creases the specific loading capacity of FDU-16-800 by �35% whilethe specific loading capacity of CS-68-800 increases by �106% overthe same range of equilibrated concentrations. This result indicatesthat micropore inclusion can improve biofuel adsorption perfor-mance by significantly increasing the specific surface area of MPCadsorbents. As the extent of microporosity in MPC adsorbents syn-thesized and examined here was neither maximized nor opti-mized, no substantial benefit with respect to biofuel adsorptionefficacy is observed (relative to L-493). However, as it has beenshown that MPCs can be synthesized with specific surface areasreaching up to 2580 m2/g (which is �200% greater than CS-81-800), the future ability to enhance surface area by tuning the pore

Fig. 1. TEM micrographs of (A) FDU-15-800, (B) FDU-16-800, (C) CS-68-800, and (D) CS-810800.

Fig. 2. Experimental and Freundlich model predictions of ethanol (closed symbols) and n-butanol (open symbols) adsorption equilibria using (A) FDU-15-800,(B) FDU-16-800, (C) CS-68-800, and (D) CS-81-800 as adsorbents.

T.J. Levario et al. / Microporous and Mesoporous Materials 148 (2012) 107–114 111

structure and morphology of MPCs will lead to improvements inperformance as hydrophobic biofuel adsorbents.

3.3. Adsorption kinetics

As n-butanol displays the greatest extent of recovery via hydro-phobic adsorption on MPCs, a series of dynamic adsorption exper-

iments were subsequently performed to characterize the kineticsof its adsorption. The dynamic adsorption data for each of the syn-thesized MPCs are illustrated in Fig. 4, together with the predictedisotherms obtained from the ‘best-fit’ of the pseudo-first orderadsorption model (Eq. (3)). The resultant ‘best-fit’ estimates ofthe kinetic constants are listed in Table 4. As is illustrated inFig. 4, the observed specific rates of adsorption are comparable

Table 3Freundlich adsorption model ‘best-fit’ parameter estimates.

Adsorbent Ethanol n-Butanol

kF

(mmol/kg)n kF

(mmol/kg)n

FDU-15-800 115 ± 32 2.41 ± 0.23 371 ± 86 3.65 ± 0.49FDU-16-800 158 ± 28 2.53 ± 0.16 708 ± 66 4.61 ± 0.34CS-68-800 62.9 ± 9.4 1.82 ± 0.07 245 ± 59 1.92 ± 0.14CS-81-800 69.7 ± 10 1.88 ± 0.07 446 ± 95 2.39 ± 0.19Dowex™ Optipore™ L-493 23 ± 12a 1.25 ± 0.29a 446 ± 115a 2.22 ± 0.26a

a As reported previously [22].

Fig. 3. Experimental isotherms and Freundlich model predictions of n-butanoladsorption at titers relevant to n-butanol fermentation using FDU-15-800 (square),FDU-16-800 (circle), CS-68-800 (triangle), and CS-81-800 (inverted triangle).Freundlich model prediction of n-butanol adsorption equilibria with Dowex™Optipore™ L-493 is shown for comparison (solid line).

Fig. 4. Experimental and model predictions of n-butanol adsorption dynamics withFDU-15-800 (square), FDU-16-800 (circle), CS-68-800 (triangle), and CS-81-800(inverted triangle).

Table 4Pseudo-first order kinetic model ‘best-fit’ parameter estimates.

Adsorbent k1 (min�1)

FDU1 1.8 ± 0.43FDU-16-800 0.9 ± 0.12CS-68-800 4.5 ± 0.29CS-81-800 7.5 ± 0.12Dowex™ Optipore™ L-493 0.05 ± 0.02

Fig. 5. Experimental and model predictions of n-butanol adsorption dynamics withFDU-16-800 (circle) and Dowex™ Optipore™ L-493 (square).

112 T.J. Levario et al. / Microporous and Mesoporous Materials 148 (2012) 107–114

for each of the synthesized carbon powders, corresponding to esti-mated kinetic constants that differ by less than an order of magni-tude and are all greater than 0.9 min�1 (Table 4). In comparison,and as illustrated in Fig. 5, the specific rate of n-butanol adsorptiondisplayed by L-493 is much less than that of even the slowest MPC(FDU-16-800); the kinetic constant for L-493 is 18- to 150-timeslower than those of the studied MPC sorbents. For example, fromthese pseudo-first order kinetic constants, one can predict thatFDU-16-800 will reach 99% of its equilibrium specific loading afterjust 6 min, whereas L-493 would require 116 min (or nearly 13-times longer) to reach the same extent of equilibration.

While all MPCs studied demonstrate high specific rates of n-butanol adsorption, a further comparison of key structural differ-ences between the distinct classes of MPCs examined provides im-proved mechanistic understanding of the adsorption behavior.Table 2 compares the pertinent differences in mesostructure ofthe four MPCs examined. Of particular interest is the comparisonof FDU-15-800 and FDU-16-800, wherein FDU-15-800 notablycontains isolated cylindrical mesopores whereas FDU-16-800 pos-sesses a cage-like array of spherical mesopores. These dissimilarmorphologies lead to significant differences in the estimated pseu-do-first order kinetic constants for FDU-15-800 and FDU-16-800(1.8 versus 0.9 min�1, respectively). We postulate that the cylindri-cal mesopores of FDU-15-800 more readily facilitate the intrapar-ticle transport of n-butanol, improving rates of its adsorption.Meanwhile, the inclusion of micropores in the walls of the MPCmatrix (as in both CS-68-800 and CS-81-800) not only results ina doubling of specific surface area, but also more profoundly im-pacts adsorption kinetics than does differences in pore geometry.For example, CS-68-800 and CS-81-800 are found to have pseu-do-first order kinetic constants of 5.2- to 8.7-times greater (4.5and 7.5 min�1, respectively) than that of FDU-16-800 (0.9 min�1),despite having the same mesopore connectivity. There are two po-tential reasons for the improved kinetics: (1) microporosity resultsin perforation of the pore walls which in turn can provide a greaternumber of ‘paths’ for fluid transport, and (2) inclusion of the silicaduring carbonization decreases pore contraction, yielding largermesopores and correspondingly reduced transport resistances.The results suggest that transport through the micropores, whichexist in the thin walls of the mesopores, is responsible for the en-hanced kinetics of CS materials as the kinetic constant of CS-68-800 is lower than that of CS-81-800 despite its larger mean porediameter. In contrast to this behavior, L-493 is a macroporous,pSDVB resin whose random pore structure likely produces a moretortuous path for intraparticle diffusion, leading to its significantlylower pseudo-first order kinetic constant. The geometry andinterconnectivity of mesopores can significantly influence the rate

Fig. 6. Adsorption (closed symbols) and desorption (open symbols) isotherms forCS-81-800 in pristine condition (square) and after multiple separation andregeneration cycles (circle).

T.J. Levario et al. / Microporous and Mesoporous Materials 148 (2012) 107–114 113

of n-butanol uptake, but the rate of adsorption for MPCs in all casesis still significantly greater than that for commercial polymericadsorbents.

3.4. Sorbent regeneration

Throughout this study, both virgin and regenerated MPCs havebeen utilized for all of the biofuel adsorption characterizationexperiments performed, with no distinguishable differences ob-served. The use of regenerated adsorbents illustrates that thesematerials could be applied in an industrial application whereinmultiple cycles of adsorption/desorption and adsorbent reusewould be required to minimize material costs. To further confirmthat the MPC adsorbents are not altered by regeneration or reuse,the nitrogen adsorption/desorption behavior of both virgin andregenerated samples is examined. In all cases, the average pore sizeremained unchanged for each material, but minor increases anddecreases in the surface area did occur. These changes are likelya result of the elevated experimental sensitivity due to the limitedquantity of regenerated materials available (typically only500 mg). As no common trend can be resolved with regards tochanges in surface area and no change in the pore size is observed,it can be concluded that the mesoporous carbons neither degrade,nor irreversibly adsorb n-butanol. Fig. 6 illustrates the representa-tive nitrogen adsorption/desorption isotherms for CS-81-800 ob-tained using samples of the virgin material as well as thosehaving undergone multiple (no less than 8) cycles of regenerationand reuse.

Although the adsorbent polymer L-493 was not routinely regen-erated and reused within this study, prior works have demon-strated the prospects of its thermal regeneration in support ofrepeated n-butanol adsorption cycles [21]. Although pSDVB meltsat �250 �C and has a glass transition temperature of �95 �C, theupper operating temperature limit for L-493 is reported as110 �C. In contrast, MPC adsorbents can remain stable at tempera-tures well beyond 1400 �C in non-oxidating environments [34] andat temperatures less than 325 �C in air, making them ideal for usewith thermal cycling in support of biofuel recovery andpurification.

4. Conclusion

With high specific loading capacities and rapid adsorptionkinetics, highly-ordered mesoporous carbons possess great poten-tial as biofuel adsorbents. The adsorption kinetics of lower alcohol

biofuels upon mesoporous carbons can be enhanced through incor-poration of highly-ordered and uniform pore structure, and isgreatly influenced by both mesopore geometry and by inclusionof micropores. In contrast to traditional biofuel adsorbents, MPCspossess ultrahigh thermal and chemical stability, greatly promot-ing their facile regeneration and reuse. No loss of adsorption per-formance was observed as a result of material regenerationthroughout the duration of this study, a promising feature for fu-ture industrial applications.

Acknowledgments

T.J.L. was supported by financial assistance from the US Depart-ment of Energy, Office of ARPA-E (Award No. DE-AR0000011).Facilities supported by the Center for Solid State Science were usedfor the characterization of materials. This work is partially sup-ported by the National Science Foundation under Grant No.CBET-0746664.

References

[1] EIA, Annual energy review 2009, Annual Energy Review 2009, US EnergyInformation Administration, Washington, 2010.

[2] EIA, Energy independence and security act of 2007, Energy Independence andSecurity Act of 2007, Energy Information Administration, US Department ofEnergy, Washington, 2007.

[3] J.D. Keasling, H. Chou, Nat. Biotechnol. 26 (2008) 298.[4] M.R. Connor, J.C. Liao, Curr. Opin. Biotechnol. 20 (2009) 307.[5] D.T. Jones, D.R. Woods, Microbiol. Rev. 50 (1986) 484.[6] L.K. Bowles, W.L. Ellefson, Appl. Environ. Microbiol. 50 (1985) 1165.[7] L.O. Ingram, Crit. Rev. Biotechnol. 9 (1990) 305.[8] G. Walker, in: G. Walker (Ed.), Yeast: Physiology and Biotechnology, Wiley,

New York, 1998, p. 101.[9] P.L. Rogers, K. Lee, D. Tribe, Biotechnol. Lett. 1 (1996) 165.

[10] L.P. Yomano, S.W. York, L.O. Ingram, J. Ind. Microbiol. Biotechnol. 20 (1998)132.

[11] A.J. Straathof, Biotechnol. Prog. 19 (2003) 755.[12] K. Schugerl, J. Hubbuch, Curr. Opin. Microbiol. 8 (2005) 294.[13] M. Galbe, P. Sassner, A. Wingren, G. Zacchi, Adv. Biochem. Eng. Biotechnol. 108

(2007) 303.[14] W.L. Luyben, Energy Fuels 22 (2008) 4249.[15] K. Schugerl, Biotechnol. Adv. 18 (2000) 581.[16] V. Garcia, J. Pakkila, H. Ojamo, E. Muurinen, R.L. Keiski, Renew. Sust. Energy

Rev. 15 (2011) 964.[17] T.C. Ezeji, N. Qureshi, H.P. Blaschek, Chem. Rec. 4 (2004) 305.[18] A. Oudshoorn, L.A.M. van der Wielen, A.J.J. Straathof, Ind. Eng. Chem. Res. 48

(2009) 7325.[19] M. Kumar, K. Gayen, Appl. Energy 88 (2011) 1999.[20] F.A. Carey, R.J. Sundberg, Advanced Organic Chemistry, Kluwer Academic/

Plenum Pub., New York, 2000.[21] D.R. Nielsen, K.J. Prather, Biotechnol. Bioeng. 102 (2009) 811.[22] D.R. Nielsen, K.L.J. Prather, G.S. Amarasiriwardena, Bioresour. Technol. 101

(2010) 2762.[23] V. Saravanan, D.A. Waijers, M. Ziari, M.A. Noordermeer, Biochem. Eng. J. 49

(2010) 33.[24] A. Oudshoorn, L.A.M. van der Wielen, A.J.J. Straathof, Biochem. Eng. J. 48 (2009)

99.[25] V.K. Saini, M. Andrade, M.L. Pinto, A.P. Carvalho, J. Pires, Sep. Purif. Technol. 75

(2010) 366.[26] J. Silvestre-Albero, A. Silvestre-Albero, A. Sepulveda-Escribano, F. Rodriguez-

Reinoso, Microporous Mesoporous Mater. 120 (2009) 62.[27] M. Hartmann, A. Vinu, G. Chandrasekar, Chem. Mater. 17 (2005) 829.[28] C.D. Nunes, J. Pires, A.P. Carvalho, M.J. Calhorda, P. Ferreira, Microporous

Mesoporous Mater. 111 (2008) 612.[29] T.E. Cook, W.A. Cilley, A.C. Savitsky, B.H. Wiers, Environ. Sci. Technol. 16 (1982)

344.[30] Q.R. Qian, Q.H. Chen, M. Machida, H. Tatsumoto, K. Mochidzuki, A. Sakoda,

Appl. Surf. Sci. 255 (2009) 6107.[31] M.U. Dural, L. Cavas, S.K. Papageorgiou, F.K. Katsaros, Chem. Eng. J. 168 (2011)

77.[32] P.S. Tin, H.Y. Lin, R.C. Ong, T.S. Chung, Carbon 49 (2011) 369.[33] C.D. Liang, Z.J. Li, S. Dai, Angew. Chem. Int. Ed. 47 (2008) 3696.[34] Y. Meng, D. Gu, F.Q. Zhang, Y.F. Shi, L. Cheng, D. Feng, Z.X. Wu, Z.X. Chen, Y.

Wan, A. Stein, D.Y. Zhao, Chem. Mater. 18 (2006) 4447.[35] R. Ryoo, S.H. Joo, M. Kruk, M. Jaroniec, Adv. Mater. 13 (2001) 677.[36] X. Zhuang, C.M. Feng, Y. Shen, D.Y. Zhao, Y. Wan, Chem. Mater. 21 (2009) 706.[37] Z.X. Wu, Y. Meng, D.Y. Zhao, Microporous Mesoporous Mater. 128 (2010) 165.[38] K.P. Gierszal, M. Jaroniec, J. Am. Chem. Soc. 128 (2006) 10026.[39] C.D. Liang, S. Dai, J. Am. Chem. Soc. 128 (2006) 5316.

114 T.J. Levario et al. / Microporous and Mesoporous Materials 148 (2012) 107–114

[40] R.L. Liu, Y.F. Shi, Y. Wan, Y. Meng, F.Q. Zhang, D. Gu, Z.X. Chen, B. Tu, D.Y. Zhao,J. Am. Chem. Soc. 128 (2006) 11652.

[41] A. Stein, Z.Y. Wang, M.A. Fierke, Adv. Mater. 21 (2009) 265.[42] Y.H. Deng, J. Liu, C. Liu, D. Gu, Z.K. Sun, J. Wei, J.Y. Zhang, L.J. Zhang, B. Tu, D.Y.

Zhao, Chem. Mater. 20 (2008) 7281.

[43] Y. Fang, D. Gu, Y. Zou, Z.X. Wu, F.Y. Li, R.C. Che, Y.H. Deng, B. Tu, D.Y. Zhao,Angew. Chem. Int. Ed. 49 (2010) 7987.

[44] L.Y. Song, D. Feng, C.G. Campbell, D. Gu, A.M. Forster, K.G. Yager, N. Fredin, H.J.Lee, R.L. Jones, D.Y. Zhao, B.D. Vogt, J. Mater. Chem. 20 (2010) 1691.

[45] T.C. Ezeji, N. Qureshi, H.P. Blaschek, Appl. Microbiol. Biotechnol. 63 (2004) 653.