ARTICLE

Removal and Recovery of Furfural,5-Hydroxymethylfurfural, and Acetic Acid FromAqueous Solutions Using a Soluble Polyelectrolyte

Brian Carter, Patrick C. Gilcrease, Todd J. Menkhausy

Department of Chemical and Biological Engineering, South Dakota School of Mines and

Technology, 501 East St. Joseph Street, Rapid City, South Dakota 57701;

telephone: þ1-605-394-2422; fax: þ1-605-394-1232Received 29 November 2010; revision received 3 March 2011; accepted 14 March 2011

Published online 31 March 2011 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/bit.23153

ABSTRACT: In the cellulosic ethanol process, furfural, 5-hydroxymethylfurfural (HMF), and acetic acid are formedduring the high temperature acidic pretreatment stepneeded to convert biomass into fermentable sugars. Thesecompounds can inhibit cellulase enzymes and fermentationorganisms at relatively low concentrations (�1 g/L). Effec-tive removal of these inhibitory compounds would allow theuse of more severe pretreatment conditions to improvesugar yields and lead to more efficient fermentations; ifrecovered and purified, they could also be sold as valuableby-products. This study investigated the separation of ald-hehydes (furfural and HMF) and organic acid (acetic acid)inhibitory compounds from simple aqueous solutions byusing polyethyleneimene (PEI), a soluble cationic polyelec-trolyte. PEI added to simple solutions of each inhibitor at aratio of 1mol of functional group to 1mol inhibitorremoved up to 89.1, 58.6, and 81.5 wt% of acetic acid,HMF, and furfural, respectively. Furfural and HMF wererecovered after removal by washing the polyelectrolyte/inhibitor complex with dilute sulfuric acid solution. Recov-eries up to 81.0 and 97.0 wt% were achieved for furfural andHMF, respectively. The interaction between PEI and aceticacid was easily disrupted by the addition of chloride ions,sulfate ions, or hydroxide ions. The use of soluble polymersfor the removal and recovery of inhibitory compounds frombiomass slurries is a promising approach to enhance theefficiency and economics of an envisioned biorefinery.

Biotechnol. Bioeng. 2011;108: 2046–2052.

� 2011 Wiley Periodicals, Inc.

KEYWORDS: biomass; inhibitory compounds; separations;polyelectrolytes; adsorption; flocculation

Introduction

The production of alternative liquid transportation fuels(e.g., ethanol) and chemical feedstocks from biorenewableresources (e.g., grasses, corn stover, and woods) has showngreat promise toward reducing our dependence onpetroleum. In recent years, biorefining process technologieshave been greatly improved, allowing biorenewable pro-ducts to become more economically competitive withpetroleum-derived fuels and chemicals. However, there arestill many areas that require further development to improveprocess efficiencies (Banerjee et al., 2010; Lin and Tanaka,2006). For instance, severe pretreatment conditions (e.g.,high temperature and the addition of mineral acids) are usedto disrupt the biomass structure and create more surfacearea for cellulose attack; this leads to accelerated down-stream enzymatic hydrolysis rates and higher sugar yields(Lee et al., 1999; Mosier et al., 2005; Tsao et al., 1982; Yangand Wyman, 2007). The higher temperatures and acidicconditions used in pretreatment (160–2208C, pH�2) resultin the partial breakdown of lignin to phenolic compounds(para-hydroxyphenyl, guaiacyl, and syrignyl) (Klinke et al.,2004), the near complete hydrolysis of hemicellulose toxylose, arabinose, mannose, galactose, the cleavage of aceticacid (Lin and Tanaka, 2006), along with some hydrolysis ofcellulose to glucose. Glucose and xylose can degrade furtherat pretreatment conditions to form 5-hydroxymethylfur-fural (HMF) and furfural, respectively (Mosier et al., 2005).Formation of soluble phenolic compounds, acetic acid,HMF, and furfural can be detrimental to downstreamoperations due to inhibition of enzymes (at concentrations>1 g/L) (Nakagame et al., 2010; Tengborg et al., 2001) andpoor cell growth/function caused by the inhibitors (atconcentrations >1 g/L) (Bothast et al., 1999; Klinke et al.,2004; Villa et al., 1992). Therefore, if the inhibitorycompounds could easily be removed from solution, thenmore severe pretreatment conditions and the associatedimprovement of enzymatic hydrolysis/fermentation would

yAssistant Professor.Correspondence to: T.J. Menkhaus

Contract grant sponsor: USDA Cooperative State Research, Education and Extension

Service

Contract grant number: 2007-35504-18344

Contract grant sponsor: South Dakota 2010 Center for Bioprocessing Research &

Development

2046 Biotechnology and Bioengineering, Vol. 108, No. 9, September, 2011 � 2011 Wiley Periodicals, Inc.

be advantageous. Recovery and purification of inhibitorycompounds could also provide valuable co-products in thebioethanol process.

Several methods have been investigated for the removal ofinhibitory compounds from biomass slurries, includingoverliming, pH adjustment, and various forms of adsorp-tion and chromatography (Larsson et al., 1999; Martinezet al., 2001). For adsorption, ion exchange resins have beenevaluated to capture organic acids and separate them fromsugars, and sugar degradation products by chromatography(Han et al., 2006). Results using a strong base ion exchangefunctionality (quaternary amine) showed that acetic acidcould be captured and recovered from a corn stoverhydrolysate sample through an assumed ionic interaction byoperating above the pKa of acetic acid (Han et al., 2006;Maciel de Mancilha and Karim, 2003). Similarly, Nilvebrantet al. (2001), report a 96% removal (2.4–0.1 g/L) of aceticacid from a spruce hydrolyzate using a strong anionexchange resin. However, when adjusting the hydrolyzatewith NaOH to pH 10, the amount of anion exchangematerial needed to remove acetic acid increased from 3.6 to8.0 g, indicating possible ionic interferences (Nilvebrantet al., 2001). The molar selectivity sequence for anionexchange is generally SO2�

4 >Cl�> acetate>OH�;however, OH� may fall before or after acetate dependingon the base strength of the fixed ionogenic groups(Helfferich, 1962). Generally, as the strength of the baseincreases the selectivity for the hydroxide ion decreases. Thecaptured acetic acid (in the form of acetate anion) on theadsorbent could be recovered using an elution solution suchas NaCl or Na2SO4, or by shifting the pH to acidicconditions.

In contrast to acetate, HMF and furfural are unchargedaldehydes and thus would not be removed by an ionicinteraction. Thus, polymeric adsorbents, such as XAD-4 (ahydrophobic polystyrene–divinylbenzene copolymer bead)and XAD-7 (a hydrophobic methacrylic ester bead), havebeen employed in a packed bed to reduce the level of furfuralfrom 5 to <0.2 g/L (Jerabek et al., 1994; Weil et al., 2002).

The most common contacting mechanism to performadsorption/chromatography separations is with a packedbed of solid resin. While successful for purifying relativelysmall volumes of liquid feed, packed bed ion exchangeprocesses suffer from a number of limitations. The pressuredrop across the bed is generally high, especially with elevatedflow rates, and can increase during operation due todeformation of the solid phase. Pore diffusion is often slowleading to increased processing time and poor utilization ofion-exchange capacity. Scale up of packed-bed columns canalso be difficult.

Membrane adsorption is an alternative to packed bed ionexchange that reduces pressure drop and internal diffusionlimitations; however, this can be economically challengingdue to the high cost and low durability of many membranes.As a more elegant option, if adsorption could beaccomplished simultaneously with an existing operation,such as removal of solids with a soluble polymeric flocculant

(Burke et al., 2010; Menkhaus et al., 2010), the overallprocess may become easier and more economical.

Coagulants or flocculants are soluble chemicals thatpromote the adsorption and/or aggregation of solids, whichcan then agglomerate/precipitate out of solution and beremoved via settling/filtration. Flocculating agents may bepolycationic, polyanionic, or neutral allowing for a variety ofchemical interactions with solids and flocculant. Differenttypes of coagulants/flocculants can be organic (polyaminesor polyacrylamides) or inorganic (aluminum sulfate,chitosan, and ferric sulfate). Even though the agents aresoluble, the flocculation process operates on the samechemical principles as other adsorption systems; that is, aphysical/chemical interaction between the flocculant andparticle (i.e., opposite ionic charges or affinity attraction).Depending on the properties of the polymer (charge density,chain length, and degree of branching) and particle (chargedensity, charge distribution, and size), interactions betweenthe soluble polymeric flocculants and solution componentscan occur through various mechanisms (e.g., ionic,hydrophobic, affinity, etc.).

Polyethylenimine (PEI) is a water soluble cationicpolymer that can be either linear or branched. Linear PEIcontains all secondary amines whereas branched PEIcontains primary, secondary, and tertiary amines. It iseasily dissolved in water and has a vapor pressure of only9mmHg at 208C. Polyethyleneimine is a commonflocculating agent which has been used in cell cultures toprecipitate cellular debris (Wickramasinghe et al., 2010),and to recover and purify proteins from plant extracts(Menkhaus et al., 2002, 2004). It has also been used toremove cell debris particles, nucleic acid polymers andproteins from Escherichia coli cell disintegrates by sedi-mentation (Agerkvistm and Eriksson, 1993). PEI has alsobeen utilized to remove soluble solutes from solution. Forinstance, Narkis and Rebhun (1977) performed experimentsto precipitate humic and fulvic acids from groundwaterusing polyethyleneimine and remove them by centrifuga-tion. They discovered a stoichiometric relationship betweenthe humic and fulvic acid content (containing carboxylateand phenolate ionic groups) in solution and the cationicpolyelectrolyte dose required for effective flocculation. Theyalso determined that the flocculant requirement was loweras the pH decreased from 8.0–8.2 to 5.8–6.7 because only thecarboxylic groups are dissociated, while the phenolates arebelow their pKa and thus associated. Finally, a process forseparating furfural from a middle distillate (diesel oil)solution by use of a branched polyimine membrane whichcontains primary, secondary, and tertiary amine groups hasbeen reported (Pasternak and Reale, 1991).

When using PEI as a soluble adsorbent for inhibitorremoval in biomass slurry processing, two distinctmechanisms can occur. First, the polyamine can act as apolycationic agent with potential ionic interactions, whichcould be used to adsorb organic acids. Second, the use of theprimary and secondary amine functionality offers thepossibility that the aldehydes, furfural, and HMF, could

Carter et al.: Furfural, HMF, and Acetic Acid Removal 2047

Biotechnology and Bioengineering

be reversibly complexed and removed through the Mannichreaction sequence (Thompson, 1968) as shown in Reaction1, specifically for furfural. Previous research has shown thatan intermediate step of the Mannich reaction is acidcatalyzed and proceeds more rapidly than in neutral media.The electrophilic attack of the amine on the furan may occureasily; this is favorable to the formation of the activatediminium ion, which complexes the free aldehyde with theamine polymer through a covalent linkage (Li and Xiao,1995). Addition of water to the overall process may also helpdrive the equilibrium to the left, resulting in releasedaldehyde and regenerating the amine. Decreasing the pH ofthe solution may provide a way to increase the reversereaction rate of the Mannich reaction and with the additionof water, recover the furfural and HMF.

R2NHþ 2C5H4O2 ! R2NCC4H3OC5H3O2 þH2O (1)

The use of soluble polyelectrolytes as a separationmedium may allow for novel processing strategies to easilyremove enzymatic and fermentation inhibitors from anaqueous solution by a combination of reaction mechanisms,including ion-exchange and reversible covalent interactions.These processing strategies may include removal ofinhibitors after pretreatment, removal of inhibitors afterenzymatic hydrolysis, and simultaneous solids and inhibitorremoval. In this study, we investigate the separationinteraction mechanisms required for removing and recover-ing acetic acid, furfural, and HMF from aqueous solutionsby studying simple single component solutions.

Materials and Methods

Materials

Simple aqueous solutions were prepared using glucose,xylose, acetic acid, diethylamine, furfural, and HMF (all�99%) obtained from Fisher Scientific (Pittsburgh, PA).Superfloc 329, a quaternary amine polymer with averagemolecular weight (MW) of 250,000, and charge density of1824.8 (�)/mole, was obtained from Kemira North America(Mobile, AL). Polyethyleneimine (50%w/v), obtained fromFisher Scientific, was diluted with distilled water to 50 g/Lfor addition to samples.

Sample Preparation

A 10mL aliquot of each simple solution (furfural, HMF, oracetic acid at different concentrations) was added to separate50mL beakers. An equimolar amount (molar concentrationof imine group) of the 50 g/L PEI was added to each 10mLsample at room temperature (�228C). Then water wasadded to each sample to a total of 20mL to ensure an equaldilution between all the samples. The solution was mixed byswirling the beaker for a minimum of 30 s and the pH was

measured. The polymer and adsorbed solutes were separatedfrom the sample with using 2mL of sample in a 10,000molecular weight cutoff (MWCO) Centricon centrifugalfilter (Millipore, Billerica, MA) and a Beckman Coulter(Brea, CA) J2-HS Centrifuge spinning at 5,000 rpm for30min. Following ultrafiltration, the concentration ofinhibitor was measured in the permeate by HPLC (discussedin detail below), and by difference (compared to the initialamount of inhibitor) the amount adsorbed to thepolyelectrolyte could be calculated. Control samples werealso evaluated by filtering solutions containing no polymerto ensure that the filter itself was not responsible forremoving the inhibitor. In all cases, no loss of inhibitorycompound was observed in the control samples.

Next, a solution of either distilled water or sulfuric acid inwater at varying pH (pH 2 or 3, and distilled water at pH4.46) was prepared for the desorption of furfural and HMF.One milliliter of this solution was added to the 2mLcentrifugal filter (which still contained the bound PEI andinhibitor) after it had completed the filtration. It should benoted that the pH distilled water was an initial approxima-tion and was only used for comparison to lower pH acidwashes. The filter and solution were vortexed for 10 s toresuspend the complex and centrifuged again at 5,000 rpmfor 30min to recover the now unbound inhibitors. Bymeasuring the concentration and volume of the inhibitorycompound in the ultrafiltration permeate the total mass ofinhibitor recovered could be compared to the initial mass ofthe compound and total percentage recovery calculated.

The quaternary amine, Kemira Superfloc 329 (averageMW of 250,000 and charge density of 1824.8 (�)/mole), wassubjected to ‘‘conditioning’’ to associate its positive chargewith either Cl� or OH� ions. For this conditioning, anamount of Kemira Superfloc 329 (1 monomer MW of 137)equimolar to the acetic acid solution was added directly tothe 2mL centrifugal filter. Then 0.5mL of a 2M solution ofeither NaCl or NaOH was added. The solution was vortexedfor 10 s to mix and centrifuged for 30min at 5,000 rpm. Theacetic acid solution was then added to the flocculant retainedon the filter and vortexed again to mix before centrifugation.

UV/Vis

Five solutions using furfural, PEI, and diethylamine wereprepared to be analyzed by a Beckman Coulter DU 640 UV/Vis Spectrophotometer. Each individual component wasmixed separately with distilled water to a concentration of30mmol/L. A binary mixture of furfural/PEI and a binarymixture of furfural/diethylamine with water at concentra-tions of 30mmol/L of each component were also prepared.Then a UV/Vis scan of each solution was measured from 200to 800 nm in a 1 cm quartz cell with distilled water as theblank. The scan of each single component solution wassubtracted from the two binary mixtures to determine anynew bonds created from chemical reactions taking placebetween furfural/PEI or furfural/diethylamine.

2048 Biotechnology and Bioengineering, Vol. 108, No. 9, September, 2011

Concentration Analyses

Sugars and inhibitor compositions in different sampleswere analyzed by HPLC using a Beckman Coulter systemequipped with a Biorad (Hercules, CA) Aminex HPX-87Hion exclusion column. Sample injection volumes were 50mLwith a 5mM sulfuric acid mobile phase at 0.6mL/min. Thecolumn was heated to 658C with a Timberline InstrumentsModel 105 column heater. The refractive index of the eluantwas measured by a Jasco RI-1530 Intelligent refractive indexdetector. Five different standard concentrations were chosento cover the range of the analyzed samples (0–20 g/L forsugars, 0–10 g/L for inhibitors).

Results and Discussion

Acetic Acid Removal With a Quaternary Amine Polymer

Kemira Superfloc 329, a quaternary amine polymer withMW of 250,000 and a charge density of 1824.8 (�)/mol wasused to evaluate the potential of using ionic interactions forremoval of acetic acid. Since the quaternary amine had apermanent positive charge regardless of pH, it couldpotentially form an ionic bond with an acetate anion aslong as the pH of the solution was above the pKa of aceticacid (pKa¼ 4.76). The ionically bound complex consistingof high MW polymer and bound acetate could then beremoved by filtration with a 10 kDa MWCO membrane.Kemira Superfloc 329 was ‘‘conditioned’’ with either OH�

or Cl� as the exchangeable anion. When acetic acid wastitrated with OH� to a final pH of 7, only 8.6% and 11.9%removal of acetic acid was observed when either OH� or Cl�

ions, respectively, were used as the exchangeable anion onthe polymer. Since Cl� has a higher selectivity for thequaternary amine than the acetate ion, the equilibriumfavored acetate in free solution rather than bound to thepolymer. This also indicated that for the base strength of thisquaternary amine, OH� also had a stronger affinity thanacetate. This could pose some problems when attempting toremove acetic acid from a real biomass slurry in which manyanions may be present and suggested that an alternativepolymer conditioning method, a different polymer chem-istry, or an entirely different operation to separate acetic acidwould be needed.

Han et al. (2006), reported a higher acetic acid capacity ofa quaternary amine ion exchange membrane (Sartobind Q)compared to a polystyrene resin (Amberlyst A21). Using a0.1mol/L HCl wash, they were able to concentrate the eluentup to nine times greater than the loading acetic acidconcentration. Their breakthrough curves, however, showalmost immediate breakthrough of acetic acid in thepermeate. This may be comparable to our low removalpercentages for Kemira Superfloc 329, again due to theunfavorable ion exchange equilibrium coefficient of acetatecompared to competing ions present in a biomasshydrolysate. Similarly, it has been speculated that removal

of aliphatic acids from a biomass hydrolysate can becomplicated by the presence of phenolate compounds,especially under basic pH conditions, and requires addi-tional adsorbent for effective removal of all compounds(Nilvebrant et al., 2001).

Acetic Acid Removal With Polyethyleneimene

Polyethyleneimene (PEI), a secondary amine and weakbase, was used to determine if acetic acid could be removedby inducing a cationic site free of competing ions on thepolymer. In the previous experiment, a polymer with apermanent cationic charge was used to attempt removalof the acetate anion. With PEI, an acid/base neutralizationreaction coupled with an ionic interaction was investigatedto remove acetic acid from solution. PEI is a weak baseand can react with acetic acid, a weak acid (Reaction 2).The resulting acetate anion can then complex with thenewly created amine cation sites associated with thepolymer. The complex can then be removed usinga filtration membrane with MWCO below that of thepolymer MW.

R�NH�R0 þ CH3OOH ! RNþH2R0 þ CH3OO

� (2)

Two acetic acid solutions were prepared for thisexperiment at a concentration of 5 g/L. The pH of thesolutions was either adjusted with NaOH to 7.0 (to representa lignocellulosic hydrolysate where the acetate is in itsanionic form) or left unadjusted at a pH of 3.4. The amountof PEI mixed with the acetic acid solutions was varied todetermine the effect of dosage level on acetate removal. Theacid/base neutralization reaction between PEI and aceticacid was expected in the un-adjusted acetic acid solution(initially at pH 3.4) but not the pH adjusted sample. Anincrease in the dose of PEI was expected to drive the aceticacid into its anionic form, which could then form an ionicbond with PEI. As expected, no appreciable amount of aceticacid was removed from the solution with a pH of 7.0,because it was neutralized with NaOH instead of PEI. Thepresence of OH� may have interfered with the ionexchange of acetate onto the polymer due to the molarselectivity. Figure 1 shows the results for the acetic acidsolution initially at a pH of 3.4. At the highest dose of PEIused, a molar ratio of 1:1 (PEI cation groups:acetic acid),89.1 wt% of the acetic acid was removed. The resultingsolution had a pH shift from 3.4 to 11.2. Based on the weakacid/base chemistry of acetate and PEI, addition of waterwith either an acid or base to the separated polymer complexmay also provide a means to reverse the reaction and recoveracetic acid. Or, alternatively, based on the molar selectivitysequence, adding other competing anions to the solutioncould be used to recover any bound acetate, as describedbelow.

Carter et al.: Furfural, HMF, and Acetic Acid Removal 2049

Biotechnology and Bioengineering

Effect of Other Ions on Acetic Acid Removal Using PEI

An acid–base neutralization reaction (Reaction 2) shouldstill occur in the system if starting below the pKa of aceticacid (pKa¼ 4.76), to create a cation exchange site on PEI.However, other anions in solution (i.e., Cl� or SO2�

4 ) maycompete with the formed acetate ion for the cationic bindingsite developed on PEI. For this analysis, simple solutionswere prepared with an acetic acid concentration of 80mmol/L (4.8 g/L), which had a pH of 3.5, to which sodium salts ofchloride or sulfate were added. In Figure 2, the PEIconcentration was constant at a molar ratio of 1:1(PEI:acetic acid), except for the control which had noPEI, while the concentration of NaCl was varied from 0 to160mM (2M equiv.). The chloride anions in solutioncompeted with the acetate ion for the cationic sites on thePEI, decreasing the removal of acetic acid. At a chlorideconcentration of 160mM, only 27% of the acetic acid wasremoved, compared to 89.1% removal with no chloride

present. This indicated that some acetate binding stilloccurred and the binding with Cl� was not exclusive.

In a pretreated lignocellulosic feedstock, there may beseveral other anions that compete for PEI cation sites,specifically the sulfate anion, as sulfuric acid is used duringdilute acid pretreatment and then neutralized with NaOH orlime prior to enzymatic digestion. The competing anionexperiment was repeated using sodium sulfate as the addedsalt in place of NaCl. The PEI to acetic acid molar ratio was1:1 except for the control, while the amount of sodiumsulfate added was increased. Figure 2 shows an 85.6%removal of acetic acid with no sulfate and only 4.3% removalas the sulfate was increased to 160mM (2M equiv.). SO2�

4

has a stronger affinity than acetic acid and interferes morethan the Cl� ion, in part due to its (�2) charge. Acetic acidremoval from dilute-acid hydrolysate may require anothermeans of separation due to these ionic interferences.

Furfural and 5-Hydroxymethylfurfural Removal WithPolyethyleneimene

Two pure solutions of furfural and HMF were prepared atconcentrations of 2 and 3 g/L, respectively (representative ofa biomass hydrolysate after dilute acid pretreatment).The pH of the neat furfural solution was 4.25, while the pHof the neat HMF solution was 5.33. As displayed in Figure 3,at a PEI to furfural molar ratio of 1:1, 81.5 wt% of thefurfural was removed and at a PEI to HMF molar ratio of1:1, 58.6 wt% of the HMF was removed.

For the Mannich reaction product shown in Eq. (1), thecovalent imine bond should exhibit a strong UV absorbancenear 370 nm due to the n–pi� transition (Jain and Singh,2006). To help confirm the proposed Mannich-style reaction(Reaction 1) between aldehydes (furfural and HMF) and PEI,a UV/Vis wavelength scan of the PEI/furfural reacted complexwas measured. For comparison, furfural was also reacted withdiethylamine, a simpler and smaller secondary amine. Figure 4compares the UV/Vis wavelength scans of the PEI/furfural anddiethylamine/furfural reacted complexes after subtracting theabsorbances of the pure components. The matched peaks in

Figure 1. Acetic acid removal with PEI. The initial solution was at pH 3.4 before

the addition of PEI. The percentages indicate the amount of acetic acid removed. At

the highest PEI dosage the final solution was pH 11.2.

Figure 2. Effect of NaCl and Na2SO4 on acetic acid removal with PEI. The initial

80 mmol/L acetic acid solution had a pH of 3.5. PEI was added to a concentration of

80mmol/L, except for the control, which contained no PEI, NaCl, or Na2SO4. After

addition of PEI, the solution pH was 6.7 at 160mM of either NaCl or Na2SO4.

Figure 3. Furfural and HMF removal with PEI. No pH adjustment was made prior

to addition of PEI to the pure solutions of aldehyde.

2050 Biotechnology and Bioengineering, Vol. 108, No. 9, September, 2011

the 350–370nm range of the scans for these two complexessuggest the same imine bond formation between a long chainsecondary amine and furfural (PEI/furfural) and the muchsimpler single molecule diethylamine and furfural. Apossibility for the broad tailing peak of the PEI/furfural scanmay be the longer amine chain wrapping around on itself andmaking aminol or carbinolamine formations. The resultsshown here are consistent with the idea that there is atemporary covalent bonding mechanism that allows for theremoval of furfural and HMF via a complex formation withthe secondary amine followed by molecular filtration of thePEI polymer complex.

Furfural and HMF Recovery

Based on the Mannich reaction mechanism (Reaction 1), itwas believed that the reaction could be reversed with theaddition of water catalyzed with acid which would drive thereaction toward the reactants. Following removal with 1Mequiv. of PEI, a membrane water wash with pH 2.0 allowedfor recovery of 81.0 and 96.9 wt% of the furfural and HMF,respectively, of the amounts removed from solution in thefirst step, as shown in Figure 5. This provides a potentialmechanism to remove fermentation inhibitory compoundsfrom solution and recover them in purified form as avaluable by-product within a biorefinery.

Effect of PEI on Glucose and Xylose

An ideal separation would remove inhibitors from alignocellulosic hydrolysate while retaining fermentablesugars in solution. From the results shown in Figure 6,glucose remains in solution following PEI addition andsubsequent ultrafiltration, even at PEI molar equivalentdosages of 1:1. Xylose behaves similarly to glucose and alsoremains in solution.

Conclusions

Results have shown that the inhibition aldehyde compoundsfurfural and HMF, formed during pretreatment of biomasswithin a biorefinery, can be selectively removed fromsolution using the soluble secondary amine polyelectrolytePEI. Compared to traditional packed bed adsorptionsystems, the use of a soluble polyelectrolyte for adsorptionwould allow for reduced mass transfer limitations and mayallow for the possibility of completing suspended solidsflocculation concurrently with inhibitor compoundremoval. However, it does require a step to remove thepolymer and adsorbed species (either ultrafiltration ifcompleted in absence of flocculation, or a solid/liquidclarification such as centrifugation or microfiltration ifcompeted simultaneously with flocculation of suspendedsolids). The reversible covalent interaction formed via aMannich-style reaction allowed for capture and recovery ofthese aldehydes. Removal of acetic acid from solution with apolycationic polymer was difficult by an ion-exchange

Figure 4. UV/Vis wavelength scan of PEI/furfural and diethylamine/furfural

reacted complexes. The matched peaks at 345 nm suggest a similar imine bond

formation between aldehyde and secondary amine.

Figure 5. Furfural and HMF recovery with the addition of H2SO4 at different pH

values to the PEI/aldehyde complex after removal from solution.

Figure 6. Effect of varying doses of PEI on glucose and xylose removal at pH

6.66 and 5.21, respectively. One molar equivalent of PEI to glucose and xylose equals

11.3 and 8.0 g/L, respectively.

Carter et al.: Furfural, HMF, and Acetic Acid Removal 2051

Biotechnology and Bioengineering

mechanism due to the unfavorable equilibrium betweenacetate and cationic amine. Chloride and sulfate anionsoutcompeted acetate for adsorption to the chargedpolymers, reducing the ability to exchange and removeacetate from solution, which would complicate acetic acidseparation from a dilute acid pretreated biomass slurry. Inall cases no significant levels of sugars were lost duringremoval of the inhibitory compounds using polyelectrolytes.The use of soluble polyelectrolytes to remove inhibitorycompounds from complex biomass hydrolysates providesanother alternative for removing and capturing thesecompounds during a biorenewable production process.Additional studies have been completed where the poly-electrolyte adsorption technology was applied to a realbiomass slurry for removal of fermentation inhibitorycompounds, and the effects on downstream operations wereevaluated.

The authors gratefully acknowledge financial support provided by the

National Research Initiative, grant 2007-35504-18344 from the USDA

Cooperative State Research, Education and Extension Service and the

South Dakota 2010 Center for Bioprocessing Research & Develop-

ment. The authors also wish to thank Dr. David Boyles for the valuable

discussions contributing to this work.

References

Agerkvistm I, Eriksson LB. 1993. Selective flocculation with chitosan in E.

coli disintegrates: Effects of ionic strength and multivalent metal ions. J

Dispersion Sci Technol 14(3): 269–293.

Banerjee S, Mudliar S, Sen R, Giri B, Satpute D, Chakrabarti T, Pandey RA.

2010. Commercializing lignocellulosic bioethanol: Technology bottle-

necks and possible remedies. Biofuels Bioprod Biorefin 4(1): 77–93.

Bothast RJ, Nichols NN, Dien BS. 1999. Fermentations with new recombi-

nant organisms. Biotechnol Prog 15(5): 867–875.

Burke D, Anderson J, Gilcrease P, Menkhaus T. 2011. Enhanced solid-liquid

clarification of biomass slurries using polyelectrolyte flocculating

agents. Biomass Bioenergy 35(1): 391–401.

Han B, Carvalho W, Canilha L, da Silva S, eSilva J, McMillan J, Wickra-

masinghe SR. 2006. Adsorptive membranes vs. resins for acetic acid

removal from biomass hydrolysates. Desalination 193(1–3): 361–366.

Helfferich F. 1962. Ion exchange. New York: McGraw-Hill.

Jain M, Singh RV. 2006. Synthesis, characterization, and biotoxicity of N-N

donor sulphonamide imine silicon(IV) complexes. Bioinorg Chem

Appl 2006:13743–13753.

Jerabek K, Hankova L, Prokop Z. 1994. Post-cross-linked polymer adsor-

bents and their properties for separation of furfural from aqueous

solutions. React Polym 23(2–3): 107–112.

KlinkeHB, Thomsen AB, Ahring BK. 2004. Inhibition of ethanol-producing

yeast and bacteria by degradation products produced during pre-

treatment of biomass. Appl Microbiol Biotechnol 66(1): 10–26.

Larsson S, Reimann A, Nilvebrant N-O, Jonsson LJ. 1999. Comparison of

different methods for the detoxification of lignocellulose hydrolyzates

of spruce. Appl Biochem Biotechnol 77(1–3): 91–103.

Lee Y, Iyer P, Torget R. 1999. Dilute-acid hydrolysis of lignocellulosic

biomass. Adv Biochem Eng Biotechnol 65:93–115.

Li YM, Xiao HM. 1995. Studies on the mechanism of mannich reaction

involving iminium salt as potential mannich reagent. III. Furan as

pseudo acid component. Int J Quantum Chem 54:293–297.

Lin Y, Tanaka S. 2006. Ethanol fermentation from biomass resources:

Current state and prospects. Appl Microbiol Biotechnol 69:627–642.

Maciel de Mancilha, I, Karim MN. 2003. Evaluation of ion exchange resins

for removal of inhibitory compounds from corn stover hydrolyzate for

Xylitol fermentation. Biotechnol Prog 19:1837–1841.

Martinez A, Rodriguez ME, Wells ML, York SW, Preston JF, Ingram LO.

2001. Detoxification of dilute acid hydrolysates of lignocellulose with

lime. Biotechnol Prog 17:287–293.

Menkhaus TJ, Eriksson SU, Whitson PB, Glatz CE. 2002. Host selection as a

downstream strategy: Polyelectrolyte precipitation of beta-glucuroni-

dase from plant extracts. Biotechnol Bioeng 77(2): 148–154.

Menkhaus TJ, Pate C, Krech A, Glatz CE. 2004. Recombinant protein

purification from pea. Biotechnol Bioeng 86(1): 108–114.

Menkhaus TJ, Anderson J, Lane S, Waddell E. 2010. Polyelectrolyte

flocculation of grain stillage for improved clarification and water

recovery within bioethanol production facilities. Bioresour Technol

101(7): 2280–2286.

Mosier N, Wyman C, Dale B, Elander R, Lee YY, Holtzapple M, Ladisch M.

2005. Features of promising technologies for pretreatment of ligno-

cellulosic biomass. Bioresour Technol 96:673–686.

Nakagame S, Chandra RP, Saddler JN. 2010. The effect of isolated lignins,

obtained from a range of pretreated lignocellulosic substrates, on

enzymatic hydrolysis. Biotechnol Bioeng 105(5): 871–879.

Narkis N, RebhunM. 1977. Stoichiometric relationship between humic and

fulvic acids and flocculants. Water Technol 13:325–328.

Nilvebrant N-O, Anders R, Larsson S, Jonsson LJ. 2001. Detoxification of

lignocellulose hydrolysates with ion-exchange resins. Appl Biochem

Biotechnol 91–93:35–49.

Pasternak M, Reale J. 1991. Patent No. 4,982,051. United States.

Tengborg C, Galbe M, Zacchi G. 2001. Reduced inhibition of enzymatic

hydrolysis of steam-pretreated softwood. Enzyme Microb Technol

28:835–844.

Thompson BB. 1968. The mannich reaction. Mechanistic and technological

considerations. J Pharm Sci 57(5): 715–733.

Tsao GT, Ladisch M, Voloch M, Bienkowski P. 1982. Production of ethanol

and chemicals from cellulosic materials. Process Biochem 17:34–38.

Villa GP, Bartroli R, Lopez R, Guerra M, Enrique M, Penas M, Rodriquez E,

Redondo D, Iglesias I, Diaz M. 1992. Microbial transformation of

furfural to furfuryl alcohol by Saccharomyces cerevisiae. Acta Biotechnol

12(6): 509–512.

Weil JR, Dien B, Bothast R, Hendrickson R, Mosier NS, Ladisch MR. 2002.

Removal of fermentation inhibitors formed during pretreatment of

biomass by polymeric adsorbents. Ind Eng Chem Res 41(24): 6132–

6138.

Wickramasinghe, Leong Y-K, Mondal S, Liow J-L. 2010. Influence of

cationic properties on the flocculation of yeast suspensions. Adv

Powder Technol 21(4): 374–379.

Yang B, Wyman CE. 2007. Pretreatment: The key to unlocking low-cost

cellulosic ethanol. Biofuels Bioprod Biorefin 2(1): 26–40.

2052 Biotechnology and Bioengineering, Vol. 108, No. 9, September, 2011