Separation and Purification Technology 185 (2017) 186–195

Contents lists available at ScienceDirect

Separation and Purification Technology

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

Reactive extraction and recovery of levulinic acid, formic acid andfurfural from aqueous solutions containing sulphuric acid

http://dx.doi.org/10.1016/j.seppur.2017.05.0361383-5866/� 2017 The Authors. Published by Elsevier B.V.This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

Abbreviations: 5-HMF, 5-hydroxymethylfurfural; HPLC, High Pressure LiquidChromatography; KF, Karl Fischer; DIPE, diisopropylether; TOA, trioctylamine;TOPO, trioctylphosphine oxide; TB, 4-tert-butylbenzenediol; TSBE, temperatureswing back extraction; GJ, gigajoule.⇑ Corresponding author.

E-mail address: b.schuur@utwente.nl (B. Schuur).

Thomas Brouwer a, Marek Blahusiak a, Katarina Babic b, Boelo Schuur a,⇑aUniversity of Twente, Faculty of Science and Technology, Sustainable Process Technology Group, PO Box 217, 7500AE Enschede, The NetherlandsbDSM ACES Ahead R&D, P.O. Box 1066, 6160 BB Geleen, The Netherlands

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

Article history:Received 11 November 2016Received in revised form 17 May 2017Accepted 20 May 2017Available online 23 May 2017

Keywords:Levulinic acidSulphuric acidFurfuralFormic acidSolvent screeningLiquid-liquid extractionTemperature swing back extraction

Levulinic acid (LA) can be produced from lignocellulosic materials via hydroxylation followed by anacid-catalyzed conversion of hexoses. Inorganic homogeneous catalysts are mostly used, in particularsulphuric acid, yielding a mixture of LA with sulphuric acid, formic acid (FA) and furfural. Significantattention has been paid to optimization of the yield, but purification of the LA is a challenge too. Thiswork focuses on the separation of LA from the complex aqueous mixtures by liquid-liquid extraction.Two aqueous product feeds were considered, reflecting two different processes. One aqueous productstream contains sulphuric acid and LA, while the second product stream also contains formic FA andfurfural. Furfural could be removed selectively via liquid-liquid extraction with toluene. For selectiveextraction of LA and FA without co-extracting sulphuric acid, 30 wt.% of trioctylphosphine oxide(TOPO) in methylisobutylketone (MIBK) was found most suitable, showing a high selectivity oversulphuric acid, and a high equilibrium partitioning of LA. When instead of MIBK, 1-octanol was appliedas diluent, the co-extraction of FA was enhanced, while hexanoic acid suppressed the acid extraction. Toobtain the LA pure, eventually a distillation is required, and the potential of temperature swing backextraction (TSBE) at 90 �C to pre-concentrate the acid solutions was evaluated for 30 wt.% TOPO inMIBK. This pre-concentration step increased the concentrations of LA and FA by a factor of 2.45 and2.45 respectively, reducing the distillation reboiler duty from roughly 31.5 to 11.3 GJ per ton LA, at a costof roughly 4.5 GJ heating duty per ton produced LA.� 2017 The Authors. Published by Elsevier B.V. This is anopenaccess article under the CCBY license (http://

creativecommons.org/licenses/by/4.0/).

1. Introduction

Levulinic acid (LA) is widely described as one of the high poten-tial platform chemicals which can be derived from lignocellulosicmaterials [1]. The predominant conversion route in production ofLA starts with hydrolysis of the polymeric lignocellulosic materialsinto the hexose- and pentose monomers under acidic conditions.LA is formed via an acid-catalyzed reaction of the hexoses via theintermediate 5-hydroxymethylfurfural (5-HMF). Formic acid (FA)is obtained as an unavoidable side product. In parallel, the pen-toses are converted into a second side product, furfural. The totalyield of useful products is reduced by the formation of black insol-uble particles, called humins, via a polymerization reaction [2].

This unwanted by-product can be removed together with the unre-acted lignocellulosic materials by in situ filtration of the reactionmixture [3]. Various acidic materials have been reported to facili-tate the reaction to LA. Usually homogeneous acid catalysts, suchas inorganic acids [4–7], enzymes [8], acidic functionalized ionicliquids [9], organic acids [5] and metal salts of such acids [8] areconsidered. Besides those, heterogeneous acid catalysts i.e. ionexchange resins, were mentioned too [8,10]. Despite purificationcomplications, inorganic acids are most often used in the conver-sion to LA, because very long reaction times are required to obtainsimilar yield when using a heterogeneous catalyst [11].

The homogeneous mineral acids hydrochloric acid and sul-phuric acid can be used both, with sulphuric acid as preferred acid[12], due to the potential risk of releasing chlorine in the biospherewhen using hydrochloric acid [11].

Research papers and patents focus mainly on the description ofthe pre-fractionation [4,5,13–15] or conversion [6,7,16–22] of var-ious types of feedstocks to LA. In acid-catalyzed reactions of ligno-cellulosic materials to LA, two approaches are commonly applied

T. Brouwer et al. / Separation and Purification Technology 185 (2017) 186–195 187

using strong inorganic acids, i.e. using high acid concentrations atmild temperatures (55–60 wt.% hydrochloric acid or 31–70 wt.%sulphuric acid at 20–50 �C), or diluted acids (<10 wt.%) at high tem-peratures of 170–240 �C [3,12,16,20,23–32]. The high reactiontemperature and diluted catalyst option is most often applied[6,7,12,16,20,22,29,32]. This prevents the use of expensive equip-ment which can withstand the high acid concentration, which isconsidered beneficial over the advantage of high acid concentra-tions that according to Mullen et al. [6] Reunanen et al. [12] andCuzens et al. [24] result in less char formation, a higher LA yieldand fewer by-products at lower temperatures. .

After the conversion step, producing the LA, a complex aqueousproduct mixture is obtained, which needs further fractionation.Although the LA concentration in the mixture is strongly depen-dent on the type of lignocellulosic material used and the reactionconditions applied, often 3–8 wt.% has been found, for both FAand furfural a concentration of 1–5 wt.% was found[7,20,22,24,31,32].

For dilute high boilers such as LA, due to the low concentrationof products in the aqueous stream, direct distillation is economi-cally less favorable, and liquid-liquid extraction has been men-tioned as potential technique for fractionating the productmixture [7,32].

Traditionally high molecular weight aliphatic amines [21,33–36] and organophosphorus extractants [12,37] have been used toextract carboxylic acids from diluted aqueous streams. In additiona large variety of physical solvents have been suggested such asalkylphenols [6,17], ketones [6], alcohols [6], fatty acids [6,12],esters [12], ethers [6,20], and halogenated hydrocarbons [6]. Veryrecently, a study has been reported using octanol, MIBK and fur-fural as physical solvents, and the heat duty to recover the LA fromthese solvents was compared [38]. Recently new extractants havebeen designed to further improve the extraction of carboxylicacids, such as aromatic amines [39] and ionic liquids [40–43]. Rey-hanitash et al. [44] recently showed that both aliphatic amines andionic liquids cause significant co-extraction of sulphuric acid, mak-ing them less suited for selective extraction of LA due to the pres-ence of sulphuric acid as catalyst. Since the raffinate of the acidextraction is recycled back to the reaction step, and the productshould not contain sulphuric acid, any sulphuric acid co-extraction results in the necessity of an additional catalyst recov-ery step.

It is thus important to find a solvent that shows a high selectiv-ity for LA over sulphuric acid, and here, we report a solvent screen-ing study focusing on the acid fractionation, which is a key step inthe process allowing product isolation without loss of catalyst. Pos-sibly, the acid fractionation is preceded by a furfural extraction sec-tion (but this might also be done afterwards). In the completefractionation scheme, as shown in Fig. 1, these are the secondand third sections. After the physical solvent screening for furfuralremoval, the composite solvent screening study was performed intwo stages, in the first stage only sulphuric acid and LA were pre-sent in the mixtures, while in the second stage the more complexmixtures were applied that also contain FA and furfural.

Next to selectivity and distribution in the extraction process,any economically feasible process should exhibit good recyclabilityof the solvent. Next to the solvent screening, also recyclability byback-extraction at elevated temperature was investigated.

2. Experimental

2.1. Chemicals

All chemicals were purchased by Sigma Aldrich and used asreceived without further treatment, i.e. levulinic acid (98%),

1-octanol (�99%), formic acid (�95%), heptane (99%), 4-tert-butylcatechol (�98.0%), 4-methyl-2-pentanone (�98.5%), dode-cane (�99%), sulphuric acid (95.0–98.0%), toluene (�99.9%),diisopropyl ether (�98.5%), furfural (99%), 1-pentanol (�99%),1-butanol (�99.7%), 1-hexanol (�99%), hexanoic acid (�98.0%),trioctylamine (99.6%) and trioctylphosphine oxide (99%).

2.2. Liquid – liquid extraction experiments

Liquid-liquid extraction experiments were conducted in 10 mLglass vials. An analytical balance was used to weigh all compoundsin both phases with an accuracy of 0.5 mg. The biphasic systems inthe vials was shaken rigorously using a mechanical mixer and con-secutively shaken in a shaking bath for at least 20 h at a constanttemperature. All the experiments have been performed at25 ± 0.02 �C, with the exception of experiments where the temper-ature is indicated otherwise. The mass of the aqueous phase waskept constant at 4 g with a solvent to feed ratio of 1 (mass based).The aqueous phase contained 8 wt.% LA and 10 wt.% sulphuric acid,and in case the more complicated feed was used, also 5 wt.% FA and5 wt.% furfural was added to the aqueous phase. These concentra-tions reflect achievable concentrations for the acid catalyzed con-version in various processes with various lignocellulosicfeedstocks.

2.3. Analytical procedures

2.3.1. High Pressure Liquid Chromatography (HPLC)The aqueous phase concentrations of LA, FA, furfural and sul-

phuric acid were determined with HPLC using an Agilent 1200 ser-ies apparatus, equipped with a Hi-Plex-H column (300 � 7.7 mm)and a refractive index detector (RID). 5 mM aqueous sulphuric acidsolution was applied as eluent at a flow rate of 0.6 mL/min. Theinjection volume was 10 lL and the column oven was isothermallyoperated at 65 �C.

2.3.2. Karl Fischer Titration (KFT)The water content of all the organic phases was analyzed with

Karl Fischer Titration using a 787 KF Titrino 730 TiStand of Metrohm,each sample was analyzed in duplo. A methanol/dichloromethane(volumetric ratio of 3:1) solution and HYDRANAL� were used astitrant. A standard deviation over all measurements was deter-mined to be 0.19 vol.%.

2.4. Definitions

The reported distribution coefficients are defined on molarbasis:

KD;i ¼ ½i�o½i�aq

mole=Lmole=L

� �ð1Þ

where i represents LA, FA, sulphuric acid or furfural. The organicphase concentrations were determined using the mass balance,and taking into account the mass of water leached to the organicphase and the mass of solvent leached to the aqueous phase.

3. Results and discussion

Eight physical solvents were evaluated, both as monomolecularsolvent, and as diluent when in a composite solvent with an extrac-tant. In addition the process side product, furfural, was evaluatedas described by Nhien et al. [38]. Though in presence of sulphuricacid, furfural was found to be unstable and formed humins.

Fig. 1. The conceptual design of the process which separates LA, FA and furfural from an acidic aqueous solution. In the acid section, LA and FA are extracted, and sulphuricacid (the catalyst), is fully recovered and recycled to the reaction section.

188 T. Brouwer et al. / Separation and Purification Technology 185 (2017) 186–195

3.1. Physical solvent evaluation

A series of 8 different physical solvents was evaluated forextracting the complex mixture containing all four componentsLA, FA, furfural and sulphuric acid. The obtained distribution coef-ficients for LA, FA, furfural and sulphuric acid are presented in

Fig. 2. The distribution coefficient of LA, FA, furfural and sulphuric acid at 25 �C for8 different physical solvents.

Fig. 2, whereas the mutual solubility’s of these physical solventsand water are given in Fig. 3.

As can be seen from Fig. 2, dodecane and toluene show onlyaffinity towards furfural, although the distribution in dodecane islow. In addition furfural has the highest distribution coefficientwith respect to the other compounds in all physical solvents. This

Fig. 3. Mutual solubility of water and organic solvents at 25 �C.

T. Brouwer et al. / Separation and Purification Technology 185 (2017) 186–195 189

can be attributed to the high activity coefficient of furfural in water[45], making the extraction of furfural from the aqueous phase intoan apolar phase highly energetically favorable. For the alcohols, itwas found that the extraction of LA and FA increases as the molec-ular size of the alcohol decreases. This trend can be assigned to theincreasing relative amount of functional hydroxyl groups inthe organic phase with decreasing molecular size of the alcohols.The alcohols are thus less suitable for selective removal of furfuralfrom the mixture. Sulphuric acid extraction is negligible in allinvestigated physical solvents, which is due to the unfavorablesolvation of the dissociated sulphuric acid in the hydrophobicenvironments.

Next to toluene, also MIBK showed good furfural extractioncapacity, however, as follows from Fig. 3, MIBK leaches signifi-cantly. Therefore the toluene may be considered most suitablefor the furfural removal prior to acid fractionation. Toluene wasalso described in the patent by De Jong et al. for furfural removal[46].

3.2. Composite solvent evaluation

The extraction capacity of composite solvents can benefit fromcomplexation interactions between the solute and the extractant.In order to investigate whether a beneficial interaction for LA ispossible, without too strong co-extraction of sulphuric acid, threedifferent extractants, trioctylamine (TOA), trioctylphosphine oxide(TOPO) and 4-tert-butylbenzenediol (TB), have been evaluated incombination with several diluents. TOA and TOPO are well knownextractants for acid extractions [12,33,37], while also alkylphenolsare known to extract LA [17]. The LA distribution into TB isexpected to benefit from TB’s two hydroxyl groups.

3.2.1. Simple mixtures with only LA and sulphuric acidThe experimental results concerning the extraction of LA and

sulphuric acid with TOA, TOPO and TB in various diluents are pre-sented in Fig. 4.

As can be seen in Fig. 4a and b, the extraction of LA and sul-phuric acid increases as function of the TOA concentration, exceptwhen hexanoic acid or 1-octanol are used as diluents. At 30 wt.% ofTOA, the molar distribution coefficient of TOA for sulphuric acidwas observed to be 0.77–0.84 for all diluents, except for hexanoicacid, for which the capacity at 30 wt.% was only 0.47, indicatinghexanoic acid is counteracting the complexation of TOA with sul-phuric acid. 1-Octanol as diluent also causes a slight decrease inthe distribution coefficient of LA as function of the TOA concentra-tion, though the molar capacity is not decreased relative to theother diluents.

Due to the low pH of the aqueous phase, LA is present in itsundissociated form. Therefore it is likely that the extraction is donevia H-bonding and solvation [47]. Senol et al. [47] also explain thehigh distribution coefficient of LA when using toluene as diluent byp-system–carboxylic group stabilization interaction during thecomplexation stage. In the case of hexanoic acid as diluent, adductsare potentially formed between the hexanoic acid and TOA, whichcan decrease the capability for interactions with the sulphuric acid,as is supported by the experimental observations.

The significant extraction of dissociated sulphuric acid can beexplained by ion pair formation with TOA. The highest selectivityof 4.4 was obtained with hexanoic acid and 10 wt.% of TOA. At30 wt.% of TOA independent of the diluent used, the selectivitydrops to approximately 0.5.

A negligible amount of sulphuric acid was observed in theextract when using TOPO as an extractant, as can be seen in Fig. 4d.This was also observed by Shuyun et al. [48] when using TOPOdiluted in heptane. The lack of sulphuric acid extraction may beexplained by the fact that TOPO is a solvatizing carrier [49], instead

of an anion-active carrier such as high aliphatic amines. Solvatizingcarriers form non-stoichiometric complexes with neutral solutes,thereby making sulphuric acid extraction unlikely. The solvationmechanism, extraction without the formation of complexes, willalso be present.

In Fig. 4f, it can be seen that also a negligible amount of sul-phuric acid was found in the extract when TB was used as anextractant, which is comparable to the extraction with TOPO inFig. 4d. In contrary to the results for TOPO and TOA, hexanoic acidas diluent results in the highest distribution coefficient for LA, ascan be seen in Fig. 4e. This may be explained by the absence of acomplexation interaction and solely the solvation mechanismwas predominant. The dimer formation between hexanoic acid,LA and TB were not disrupting complexation interactions, but werepossibly enhancing the solvation mechanism.

Due to the low selectivity arising from the preferential extrac-tion of sulphuric acid. TOA was considered an unsuitable extrac-tant to isolate LA from an acidic aqueous mixture. For thisreason, TOA was not evaluated in the following experiments whereadditionally FA and furfural was added to the complex acidic aque-ous mixture.

3.2.2. Complex mixtures containing LA, FA, furfural and sulphuric acid3.2.2.1. Trioctylphosphine oxide (TOPO). The experimental resultsconcerning the extraction of LA, FA, furfural and sulphuric acidwith TOPO in various diluents are presented in Fig. 5.

In Fig. 5a an increasing distribution coefficient of LA withincreasing TOPO concentration can be observed for MIBK andtoluene, whereas for hexanoic acid the LA distribution coefficientdrops with increasing TOPO concentration, and for 1-octanol asdiluent the distribution coefficient is constant. In Fig. 5b, it canbe seen that with 1-octanol as a diluent also the formic acid distri-bution coefficient is more or less stable, at values just over 2. Withhexanoic acid as diluent, FA extraction was not very significant,while for toluene and MIBK an increasing distribution withincreasing TOPO concentration was measured, similar to LA. InFig. 5c significant distribution coefficients for furfural can beobserved, which coincides with statements made during the phys-ical solvent screening. It was however observed that the selectivitybetween FA and LA can be significantly influenced through selec-tion of the appropriate diluent. Sulfuric acid extraction is negligiblefor all diluents with TOPO.

3.2.2.2. 4-tert-butylbenzenediol (TB). The experimental results con-cerning the extraction of LA, FA, furfural and sulphuric acid withTB in various diluents are presented in Fig. 6. There were no exper-iments done using toluene as diluent, because the solubility of TBin toluene was found to be very low.

It follows from Fig. 6a that the distribution coefficient of LAincreases with hexanoic acid as diluent up to a value of about1.2, i.e. higher than the highest value observed with TOPO. How-ever, FA is not extracted very well using hexanoic acid as diluent,as follows from Fig. 6b. In all cases, just as with TOPO, the furfuralhas a higher distribution coefficient than the acids, and will be co-extracted when it is not removed prior. Based on the extractionresults at 25 �C, TOPO with toluene and with MIBK, as well as TBin hexanoic acid were selected for studies on temperature swingback extraction.

3.3. Temperature swing back extraction equilibrium

The solvent screening in the previous sections confirmed thepossibility of selectively removing furfural from an aqueous mix-ture containing LA, FA and sulphuric acid with toluene, followedby selective extraction of the carboxylic acids FA and LA with com-

Fig. 4. The distribution coefficient of LA and sulphuric acid at 25 �C for 7 different diluents containing TOA (A & B), TOPO (C & D) and TB (E & F).

190 T. Brouwer et al. / Separation and Purification Technology 185 (2017) 186–195

posite solvents containing TB and TOPO as extractant without theco-extraction of sulphuric acid.

In order to obtain a pure LA product, the carboxylic acids needto be recovered from the solvent, and fractionated. For this solventregeneration and acid fractionation, either a direct distillation from

the solvent could be applied, or a back-extraction followed with adistillation from an aqueous stream. For the last option, tempera-ture swing back extraction (TSBE) was attempted. Back extractionexperiments at 50 �C, 75 �C and 90 �C, seen in Fig. 7, showed adecreasing distribution coefficient of LA with increasing tempera-

Fig. 5. The distribution coefficient of LA (A), FA (B), furfural (C) and sulphuric acid (D) at 25 �C for 4 different diluents containing TOPO.

Fig. 6. The distribution coefficient of LA, FA, furfural and sulphuric acid at 25 �C for 3 different diluents containing TB.

T. Brouwer et al. / Separation and Purification Technology 185 (2017) 186–195 191

192 T. Brouwer et al. / Separation and Purification Technology 185 (2017) 186–195

ture for 30 wt.% of TOPO in MIBK and toluene, however no signifi-cant temperature dependency was observed using 30 wt.% of TB inhexanoic acid.

The absence of a temperature dependency in the mixture con-taining TB limits the applicability of TSBE to zero, and this solventcan thus not be used for this application. The composite solvent30 wt.% TOPO in MIBK was found to obtain the highest distributioncoefficient for LA and in addition showed a strong temperaturedependency, it was therefore chosen as most appropriate solventfor this recovery strategy, and the TSBE was studied in more detailfor this solvent.

The molar concentrations of LA and FA are related to the molarratio units which are used in the McCabe-Thiele method [50];

XLA ¼ LA½ �aq1� LA½ �aq

; XFA ¼ FA½ �aq1� FA½ �aq

YLA ¼ LA½ �o1� LA½ �o

;

YFA ¼ FA½ �o1� FA½ �o

ð2Þ

where X is the molar ratio in the aqueous phase, Y is the molar ratioin the organic phase and the concentrations of the acids are inweight fractions. Because the acid concentrations are relativelylow, in Eq. (2) only the acids themselves and the solvent are consid-ered. The equilibrium isotherms of LA and FA have been determinedat 25 �C and 90 �C and are shown in Fig. 8. Three regions of LAextraction behavior were observed, as can be seen in Fig. 8a. RegionI is at low concentrations of LA, where a significant fraction of LA islikely in the dissociated form, and at increasing acid concentration,more of the weak acid will be in the undissociated form. This buffer-ing equilibrium causes the line to flatten off in the region II, wherethe undissociated LA becomes predominant. In region III an over-load of LA in the organic phase is seen. The shape of the LA iso-therms are consistent with a type II adsorption isothermdescribed by Brunauer [51] and is associated with multimolecularBET adsorption. [52] Due to the complex nature of the BETisotherms and the relative small amount of data points. Theisotherms are fitted via a 5th degree polynomial function (seeTables 1 and 2):

YLA ¼ a5X5LA þ a4X

4LA þ a3X

3LA þ a2X

2LA þ a1XLA þ b ð3Þ

Tamada et al. [53] also observed Type II isotherms of carboxylicacids in amine extractants and assigned the overload to higherorder stoichiometric complexations between the carboxylicacid and amine extractant. TOPO most likely forms neutral

Fig. 7. The temperature dependency of three composite solvents; TOPO (30 wt.%) inMIBK, TOPO (30 wt.%) in toluene and TB (30 wt.%) in hexanoic acid. The deviation inthe distribution coefficient is ± 10%.

non-stoichiometric complexes with LA. Hence, a specific stoi-chiometry could not be coupled to these isotherms. The FA behav-ior could be described by a Freundlich isotherm [54], as can be seenin Fig. 8b:

YFA ¼ K � X1nFA ð4Þ

3.4. Process design

A schematic overview of a Temperature Swing Bach Extraction(TSBE) is given in Fig. 9. Two Counter-Current Extraction columnare used, where a solvent is circulated between both columns.The feed of the process was defined to contain; 8.4 wt.% LA,5.3 wt.% FA and 10.5 wt.% sulphuric acid in water. It has beenassumed that all furfural is already removed via extraction withtoluene, see Fig. 1. As can be seen in Fig. 9, there are four operationpoints that can be distinguished, A1, A2, B1 and B2.

Under the condition of minimal solvent to feed ratio, point A1

describes the equilibrium of the aqueous feed streamwith the exit-ing organic stream, the extract. To really reflect a minimum solventto feed ratio, A2 for the exiting aqueous stream (the raffinate) isalso drawn on the equilibrium line. The raffinate is in equilibriumwith the organic stream coming from the back extraction column.Both equilibria are in the forward extraction column at 25 �C. ThePoint B1 represents the point where the loaded organic streamenters the back-extraction. The organic phase composition is thusequal to that of A1, and the aqueous phase composition is selectedsuch that the operating line of the back extractor does not cross theequilibrium line at 90 �C. At minimal Wash to Extract ratio, pointB1 indicates equilibrium between the water wash stream leavingthe back extraction column and the organic stream entering. Theorganic phase composition of point B2 in the back-extractor islinked to the composition of the raffinate in the extractor, sincethe solvent is cycled back to the extraction with that composition.The maximum driving force for washing the acids out of the sol-vents is obtained when clean wash liquor is entering the back-extractor, hence, X(B1) = 0.

The following procedure is applied to design the TSBE opera-tion, see Fig. 10;

(1) Point A1 is set on the 25 �C equilibrium line, at the knownacid feed concentration.

(2) A2 is an operator choice which is on the 25 �C equilibriumline. The choice of A2 may be limited by the 90 �C backextraction equilibrium. By choosing A2, also the B2 operationpoint will be defined. Any acids left in the raffinate will notbe lost, but may be recycled back in the process, see Fig. 1,thus depending on the back-extraction considerations, thechoice for a relatively high raffinate concentration thatmay be necessary for technical feasibility is not necessarilydetrimental for the process feasibility.

(3) Point B1 is again an operator choice. An aqueous concentra-tion as high as possible was chosen here, so the highestobtainable acid concentration in the product stream wasobtained.

(4) From point B1 to B2 the operation line of the back extractionwill hit the equilibrium line of 90 �C at a minimum wash toextract ratio, (W/E)min.

The solvent to feed ratio, (S/F), and the wash to extract ratio,(W/E), can be determined from operational lines resp. A1-A2 andB1-B2 by the following equation [50];

Yout ¼ FS

� �Xout þ Yin

FS

� �Xin

� �ð5Þ

Fig. 8. The isotherm extraction profiles of LA and FA at 25 �C and 90 �C between water and the composite solvent TOPO (30 wt.%) in MIBK.

Table 1The fitting parameters of the LA isotherms at 25 �C and 90 �C for a 5th degree polynomial function. The error in YLA is given as twice the standard deviation, e.g. a 95% confidenceinterval.

a5 a4 a3 a2 a1 b error

25 �C 389.6 �365.3 134.8 �24.00 2.297 7.992 � 10�3 3.001 � 10�3

90 �C 207.1 �174.5 60.43 �11.14 1.382 4.181 � 10�3 3.292 � 10�4

Table 2The fitting parameters of the FA Freundlich isotherms at 25 �C and 90 �C. The error inYFA is given as twice the standard deviation, e.g. a 95% confidence interval.

K n error

25 �C 0.1329 3.202 5.997 � 10�3

90 �C 0.1105 2.775 3.425 � 10�3

T. Brouwer et al. / Separation and Purification Technology 185 (2017) 186–195 193

From the operating equation, see Eq. (4), the solvent to feedratio, or wash to extract ratio, can be determined by taking theinverse of the slope. If the operating line simulate a limited case,than the minimum solvent to feed ratio, (S/F)min, or minimumwash to extract ratio, (W/E)min, can be obtained.

Fig. 10 simulated a TSBE operation wherein a minimum solventto feed ratio, (S/F)min = 0.78, and a minimum wash to extract ratio,(W/E)min = 0.39, is used. Though each isotherm has been measuredindependently, the co-extraction of both acids is simulated by

Fig. 9. A temperature swing back extraction (TSBE) Operation where the forward extraperformed with water at 90 �C.

coupling both McCabe-Thiele approaches seen in Fig. 10a and b.The coupling is done by keeping the slopes of the operation linesbetween A1 and A2, and B1 and B2, equal for both acids. By keepingthe slopes equal in both approaches, the same extraction operationis simulated simultaneously, thereby reflecting the co-extraction ofLA and FA. Resulting from the TSBE operation, two aqueous raffi-nate streams are obtained starting with the feed concentration, F.R1 is the recycle stream which is lean on LA and FA, while R2 isthe aqueous stream where LA and FA are concentrated. The extrac-tion of LA was observed to be the limiting factor, due to the factthat during the forward extraction FA cannot achieve equilibriumat 25 �C due to the equilibrium line of the back extraction at90 �C. The result of the TSBE operation is that LA and FA are con-centrated with a factor of resp. 1.69 and 1.58.

An increase in solvent to feed ratio or wash to extract ratio,causes the operational lines to become less steep. If the acid feedconcentration is set and both operational lines become less steep,

ction is performed with 30 wt.% TOPO in MIBK at 25 �C and the back extraction is

Fig. 11. McCable-Thiele graphic approach to simulate a second temperature swing back extraction between 25 �C and 90 �C of (a) LA and (b) FA. From the operational linesfollow a (S/F)min and (W/E)min of resp. 0.48 and 0.28.

Fig. 10. McCable-Thiele graphic approach to simulate a temperature swing back extraction between 25 �C and 90 �C of (a) LA and (b) FA. From the operational lines follow a(S/F)min and (W/E)min of resp. 0.78 and 0.39.

194 T. Brouwer et al. / Separation and Purification Technology 185 (2017) 186–195

the raffinate, A1, must be become richer in acid. Also the product,B1, must become leaner in acid in order to keep the operationallines within the border of the equilibrium lines. For this reason itwould be most beneficial to operate the TSBE operation as to the(S/F)min and (W/E)min as possible.

Until now, a single TSBE operation was simulated. This methodcan be repeated to concentrate the acid to a further extent, hence adouble TSBE is obtained in Fig. 11. The aqueous concentration of R2

from the initial TSBE operation is equal to the aqueous feed con-centration of the second TSBE operation.

By applying a second TSBE operation, LA and FA are additionallyconcentrated with a factor of resp. 1.45 and 1.55. Combining bothTSBE operations the LA and FA concentration are concentratedwith a factor of resp. 2.45 and 2.45. During both TSBE operationsan aqueous raffinate stream was obtained lean on LA and FA.Though, both raffinate streams decreases the total amount of LAand FA in the product stream, these raffinate streams may be recy-cled to the reaction section, see Fig. 1, preventing any loss of LA andFA.

A rough estimation of the heater duty for pre-dehydrating theproduct stream via a double TSBE is made in AspenPlus. Heatingthe organic phase and the water wash stream to 90 �C will be thesources of the heater duty. In addition a heat exchanger is presentto minimize the energy requirement. A minimum temperature dif-ference of 10 �C over the heat exchanger over the cold outlet and

hot inlet stream was assumed. Roughly a heater duty of 4.5 GJper produced tons of LA was necessary to concentrate LA and FAwith a factor of resp. 2.45 and 2.45. Still, distillation has to be doneto isolate LA from the aqueous FA mixture. Via the heat of vapor-ization of FA and water, this evaporation operation was estimatedto require roughly 11.3 GJ per tons of produced LA on heater duty.Without the double TSBE operation, this evaporation would haverequired roughly 31.5 GJ per tons of produced LA on heater duty.

4. Conclusions

From an exploratory solvent study selective extraction of LAfrom an aqueous solution containing solely sulphuric acid wasobserved using polar solvents. In addition TOA was deemed unsuit-able due to the significant extraction of sulphuric acid. The possi-bility of selective removal of furfural using apolar solvents, inparticular toluene was likewise observed. MIBK may also be usedto extract furfural, though a lower selectivity to furfural isobserved. Both TB as TOPO were suitable extractants for the selec-tive extraction of both LA and FA. It was observed that 1-octanolcan enhance the FA extraction, while hexanoic acid suppresses thisextraction. A double TSBE was performed to concentrate LA and FAwith a factor of resp. 2.45 and 2.45 using a heater duty of 4.5 GJ perproduced tons of LA. The heater duty of the final distillation col-umn drops due to the TSBE operations from approximately 31.5

T. Brouwer et al. / Separation and Purification Technology 185 (2017) 186–195 195

to 11.3 GJ per produced tons of LA. These rough calculations showthe potential energy savings by pre-dehydrating an aqueous feedvia TSBE operations.

Raising the environmental and toxicity issue of using toluenefor the extraction of furfural. An alternative multistage fractionalextraction column could also be designed to fractionate LA, FAand furfural from an aqueous solution containing sulphuric acidby the same composite solvent as in the LA and FA extraction,30 wt.% TOPO in MIBK.

Acknowledgements

This has been an ISPT project. ISPT is Institute for SustainableProcess Technology, headquartered in Amersfoort, theNetherlands.

References

[1] T. Werpy, G. Petersen, Top Value Added Chemicals from Biomass: Results ofScreening for Potential Candidates from Sugars and Synthesis Gas, U.S.,Department of Energy, 2004.

[2] S. Bertarione, F. Bonino, F. Cesano, S. Jain, M. Zanetti, D. Scarano, A. Zecchina,Micro-FTIR and micro-Raman studies of a carbon film prepared from furfurylalcohol polymerization, J. Phys. Chem. B 113 (2009) 10571–10574.

[3] B. Girisuta, L.P.B.M. Janssen, H.J. Heeres, Kinetic study on the acid-catalyzedhydrolysis of cellulose to levulinic acid, Ind. Eng. Chem. Res. 46 (2007) 1696–1708.

[4] J.G. De Vries, J.A. Kroon, R.F.M.J. Parton, P.L. Woestenborghs, A. De Rijke,Process for the production of furfural and levulinic acid from lignocellulosicbiomass, 2013, EP2872495 A1.

[5] T. Retsina, V. Pylkkanen, K. Nelson, R. O’connor, Processes and apparatus forproducing furfural, levulinic acid, and other sugar-derived products frombiomass, 2014, WO2014105289 A1.

[6] B.D. Mullen, D.J. Yontz, C.M. Leibig, Process to prepare levulinic acid, 2013,US9073841 B2.

[7] S.W. Fitzpatrick, Production of levulinic acid from carbohydrate-containingmaterials, 1997, US5608105 A.

[8] M.A. Lake, S.W. Burton, W.C. Fuller, R. Sasser, M.E. Lindstrom, J.T. Wheless,Production of levulinic acid and levulinate esters from biomass, 2010,US20100312006 A1.

[9] N.A.S. Ramli, N.A.S. Amin, A new functionalized ionic liquid for efficientglucose conversion to 5-hydroxymethyl furfural and levulinic acid, J. Mol.Catal. A: Chem. 407 (2015) 113–121.

[10] B.C. Redmon, Process for the production of levulinic acid, 1956, US2738367 A.[11] J. Ahlkvist, S. Ajaikumar, W. Larsson, J.-P. Mikkola, One-pot catalytic

conversion of Nordic pulp media into green platform chemicals, Appl. Catal.A 454 (2013) 21–29.

[12] J. Reunanen, P. Oinas, T. Nissinen, A process for recovery of formic acid, 2009,WO2009130386A1.

[13] R. Jansen, J.A. Lawson, N. Lapidot, B. Hallac, R. Perry, Methods for preparingthermally stable lignin fractions using various biomass types, 2014,WO2014179777A1.

[14] C.G. Yoo, C.-W. Lee, T.H. Kim, Two-stage fractionation of corn stover usingaqueous ammonia and hot water, Appl. Biochem. Biotechnol. 164 (2011) 729–740.

[15] M.S. Tunc, J. Chheda, E. van der Heide, J. Morris, A. van Heiningen, Two-stagefractionation and fiber production of lignocellulosic biomass for liquid fuelsand chemicals, Ind. Eng. Chem. Res. 52 (2013) 13209–13216.

[16] V.M. Ghorpade, M.A. Hanna, Method and apparatus for production of levulinicacid via reactive extrusion, 1999, US5859263A.

[17] V. Badarinarayana, M.D. Rodwogin, B.D. Mullen, I. Purtle, E.J. Molitor, Processesto prepare levulinic acid, formic acid and/or hydroxymethyl furfural fromvarious biomass materials, 2014, WO2014189991A1.

[18] S.W. Fitzpatrick, Lignocellulose degradation to furfural and levulinic acid,1990, US4897497 A.

[19] D. Hayes, Report on optimal use of DIBANET feedstocks and technologies, EU’sSeventh Framework Programme, 2013.

[20] A. De Rijke, G.W.A. Hangx, R.F.M.J. Parton, B. Engendahl, Process for theisolation of levulinic acid, 2014, WO2014087013 A1.

[21] G.-T. Jeong, S.-K. Kim, D.-H. Park, Detoxification of hydrolysate by reactive-extraction for generating biofuels, Biotechnol. Bioprocess Eng. 18 (2013) 88–93.

[22] J.A. Dumesic, D.M. Alonso, E.I. Gürbüz, S.G. Wettstein, Production of levulinicacid, furfural, and gamma valerolactone from C5 and C6 carbohydrates inmono- and biphasic systems using gamma-valerolactone as a solvent, 2013,US8399688 B2.

[23] W.W. Moyer, Preparation of levulinic acid, 1942, US2270328 A.[24] J.E. Cuzens, W.A. Farone, A method for the production of levulinic acid and its

derivatives, 1998, EP0937024 A1.[25] J.Y. Cha, M.A. Hanna, Levulinic acid production based on extrusion and

pressurized batch reaction, Ind. Crops Prod. 16 (2002) 109–118.[26] J. Dumesic, J. Bond, D. Alonso, T. Root, Method to produce and recover levulinic

acid and/or gamma-valerolactone from aqueous solutions using alkylphenols,2013, WO2012162028 A1.

[27] E. Ramos-Rodriguez, Process for jointly producing furfural and levulinic acidfrom bagasse and other lignocellulosic materials, 1972, US3701789 A.

[28] C.P. Sassenrath, W.L. Shilling, Preparation of levulinic acid from hexose-containing material, 1966, US3258481 A.

[29] W.A. Frey, J.W. Brown, J.P. Kelly, M.E. Carroll, P.C. Guion, Method for producinglevulinic acid from lignocellulosis biomass, 2013, US20150052806 A1.

[30] C. Chang, P. Cen, X. Ma, Levulinic acid production from wheat straw, Biores.Technol. 98 (2007) 1448–1453.

[31] L.J. Carlson, Process for the manufacture of levulinic acid, 1962, US3065263 A.[32] A.P. Dunlop, J.P.A. Wells, Process for producing levulinic acid, 1957,

US2813900 A.[33] H. Uslu, S.I. Kırbaslar, Equilibrium studies of extraction of levulinic acid by

(trioctylamine (TOA)+ ester) solvents, J. Chem. Eng. Data 53 (2008) 1557–1563.

[34] R. Canari, A.M. Eyal, Selectivity in monocarboxylic acids extraction from theirmixture solutions using an amine-based extractant: effect of pH, Ind. Eng.Chem. Res. 42 (2003) 1301–1307.

[35] Y.K. Hong, W.H. Hong, Removal of acetic acid from aqueous solutionscontaining succinic acid and acetic acid by tri-n-octylamine, Separat.Purificat. Technol. 42 (2005) 151–157.

[36] M.E. Marti, T. Gurkan, Selective recovery of pyruvic acid from two and threeacid aqueous solutions by reactive extraction, Sep. Purif. Technol. 156 (2015)148–157.

[37] M. Wisniewski, M. Pierzchalska, Recovery of carboxylic acids C1–C3 withorganophosphine oxide solvating extractants, J. Chem. Technol. Biotechnol. 80(2005) 1425–1430.

[38] L.C. Nhien, N.V.D. Long, S. Kim, M. Lee, Design and assessment of hybridpurification processes through a systematic solvent screening for theproduction of levulinic acid from lignocellulosic biomass, Ind. Eng. Chem.Res. 55 (2016) 5180–5189.

[39] A. Krzyzaniak, B. Schuur, A.B. de Haan, Equilibrium studies on lactic acidextraction with N,N-didodecylpyridin-4-amine (DDAP) extractant, Chem. Eng.Sci. 109 (2014) 236–243.

[40] F.S. Oliveira, J.M.M. Araujo, R. Ferreira, L.P.N. Rebelo, I.M. Marrucho, Extractionof L-lactic, L-malic, and succinic acids using phosphonium-based ionic liquids,Sep. Purif. Technol. 85 (2012) 137–146.

[41] M. Blahusiak, S. Schlosser, J. Martak, Extraction of butyric acid with a solventcontaining ammonium ionic liquid, Sep. Purif. Technol. 119 (2013) 102–111.

[42] A.I. Pratiwi, T. Yokouchi, M. Matsumoto, K. Kondo, Extraction of succinic acidby aqueous two-phase system using alcohols/salts and ionic liquids/salts, Sep.Purif. Technol. 155 (2015) 127–132.

[43] K. Tonova, I. Svinyarov, M.G. Bogdanov, Biocompatible Ionic Liquids in Liquid-Liquid Extraction of Lactic Acid: A Comparative Study, Int. J. Chem. Mol. Nucl.Mat. Metallurg. Eng. 9 (2015) 558–562.

[44] E. Reyhanitash, B. Zaalberg, S. Kersten, B. Schuur, Extraction of volatile fattyacids from fermented wastewater, Sep. Purif. Technol. 161 (2016) 61–68.

[45] M.O. Hertel, K. Sommer, Limiting separation factors and limiting activitycoefficients for 2-furfural, c-nonalactone, benzaldehyde, and linalool in waterat 100 �C, J. Chem. Eng. Data 51 (2006) 1283–1285.

[46] W. De Jong, G. Marcotullio, Overview of biorefineries based on co-productionof furfural, existing concepts and novel developments, Int. J. Chem. React. Eng.8 (2010).

[47] A. Senol, Influence of conventional diluents on amine extraction of picolinicacid, Sep. Purif. Technol. 43 (2005) 49–57.

[48] X. Shuyun, Y. Yonghui, Y. Yanzhao, S. Sixiu, B. Borong, Extraction of sulfuricacid with TOPO, J. Radioanal. Nucl. Chem. 229 (1998) 161–163.

[49] M. Siebold, P.V. Frieling, R. Joppien, D. Rindfleisch, K. Schügerl, H. Röper,Comparison of the production of lactic acid by three different lactobacilli andits recovery by extraction and electrodialysis, Process Biochem. 30 (1995) 81–95.

[50] A.B. De Haan, H. Bosch, Liquid-liquid extraction, in: Industrial SeparationProcesses: Fundamentals, Walter de Gruyter, 2013.

[51] S. Brunauer, L.S. Deming, W.E. Deming, E. Teller, On a theory of the van derWaals adsorption of gases, J. Am. Chem. Soc. 62 (1940) 1723–1732.

[52] J.D. Seader, E.J. Henley, D.K. Roper, Adsorption, ion exchange, chromatography,and electrophoresis, in: Separation Process Principles, 3rd ed., John Wiley &Sons, 2010, p. 821.

[53] J.A. Tamada, A.S. Kertes, C.J. King, Extraction of carboxylic acids with amineextractants. 1. Equilibria and law of mass action modeling, Ind. Eng. Chem. Res.29 (1990) 1319–1326.

[54] H. Freundlich, The impact of the adsorption in the precipitation of suspensioncolloid, Zeitschrift für Physikalische Chemie 73 (1910) 385–423.