Separation and Purification Technology 136 (2014) 233–240

Contents lists available at ScienceDirect

Separation and Purification Technology

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

Evaluation of lactic acid purification from fermentation broth by hybridshort path evaporation using factorial experimental design

http://dx.doi.org/10.1016/j.seppur.2014.09.0101383-5866/� 2014 Elsevier B.V. All rights reserved.

⇑ Corresponding author. Tel.: +55 19 35213971, +55 19 983764925; fax: +55 1935213910.

E-mail address: andrea_komesu@hotmail.com (A. Komesu).

Andrea Komesu a,⇑, Patrícia Fazzio Martins a,b, Betânia Hoss Lunelli a, Johnatt Oliveira a,Rubens Maciel Filho a, Maria Regina Wolf Maciel a

a School of Chemical Engineering, University of Campinas (UNICAMP), 13083-970 Campinas, SP, Brazilb Departamento de Ciências Exatas e da Terra, Universidade Federal de São Paulo (UNIFESP), 09972-270 Diadema, SP, Brazil

a r t i c l e i n f o

Article history:Received 9 January 2014Received in revised form 13 August 2014Accepted 9 September 2014Available online 17 September 2014

Keywords:Lactic acidPurificationFermentationEvaporation system

a b s t r a c t

This work describes the evaluation of lactic acid purification from fermentation broth by hybrid shortpath evaporation. The proposed hybrid purification process consists of an evaporation system composedby a cylindrical wiped film evaporator with two condensers, one located internally and other externallyto the evaporator. Through factorial experimental design, the influence of operation conditions as feedflow rate, agitation, condenser and evaporator temperature on residue and distilled percentages, lacticacid purity and recovery were studied. Models were developed in order to describe the response of inter-est as function of operating conditions. The results showed that with a high operating pressure (in termsof short path evaporation), with a pressure of 1000 Pa, and one step of separation, lactic acid purityaround 89.7% was obtained which was about 18 times lactic acid concentration higher than the initialcontent in raw material.

� 2014 Elsevier B.V. All rights reserved.

1. Introduction

An increasing interest for discovering new environment-friendly sources of chemicals has been observed due to the currentconcerns related to the cost and environmental impact of using tra-ditional petrochemical processes. One important technological bio-mass-based platform is the biotechnological process for lactic acidproduction by fermentation that potentially offers several advanta-ges: low substrate costs, production temperature and energy con-sumption [1].

Lactic acid has a wide variety of applications such as cosmetics,pharmaceutical products, chemistry, food and more recently in themedical area. The presence of two adjacent functional groups inthe lactic acid (hydroxyl and carboxyl) in a small molecule withonly three carbon atoms shows its high reactivity, as well as theirtendency to decompose at high temperatures.

The development of an efficient method of lactic acid separationand purification from fermentation broth is very important,because, these steps can reach up to 50% of the total costs [1,2]and it is still difficult to recover it with high purity for the reasons:high affinity with water, decomposition at elevated temperatures

and complex and energy intensive recovery technology [3]. A con-siderable number of methods for the recovery of lactic acid fromfermentation broth, such as solvent extraction [4–7], separationwith membranes [8–10], reactive distillation [11–13] and othershave been studied.

Conventional molecular distillation (or short path evaporation)had been used to recovery lactic acid with purity up to 95–96%[14–16]. The operating pressure is usually below 0.1 Pa and twoor more steps of refining are required. To keep the high-vacuuma mechanical force-pump and diffusion pump should be usedsimultaneously [14], which is in direct conflict with energy-saving.Each additional step in the downstream represents an increase inthe total operating costs.

Hybrid short path evaporation [17–19] is an alternative separa-tion process with potential for the recovery and concentration ofthermally unstable molecules such as lactic acid. It has been recog-nized as a promising technology mainly because of its low evapo-ration temperature and short residence time, which minimizeproblems with thermal decomposition [20].

In previous work, our research group studied the technical fea-sibility of lactic acid concentration from synthetic mixture ofwater: lactic acid (36 wt% of lactic acid) using hybrid short pathevaporation system [20]. The experimental results showed thatcarrying out the lactic acid concentration by using evaporativesystem is technically feasible and advantageous. Based on the

234 A. Komesu et al. / Separation and Purification Technology 136 (2014) 233–240

preliminary results with synthetic mixture, new experiments werecarried out with fermentation broth. By the fact that fermentationbroth is a mixture more complex than synthetic mixture, by thepresence of residual sugars and other organic acids, preparationand analysis of the raw material was different as well as the perfor-mance of the separation process.

Bearing all this in mind, the objective of this paper was to eval-uate the lactic acid purification from fermentation broth by hybridshort path evaporation system using factorial experimental design.It allows the determination and evaluation of the relative signifi-cance of operational parameters on the process, even in the pres-ence of complex interactions [19]. In the studied process higheroperating pressure (1000 Pa), compared to that usually employedin conventional molecular distillation, and one step of refiningwere used, which make this technique more suitable for lactic acidpurification than the literature published works.

2. Experimental

2.1. Raw material preparation

Fermentation process was carried out in a 7 L New BrunswickScientific BioFlo 415 bioreactor. Sugarcane molasses (48% sucrosew/w) without pretreatment, typical of large scale industrial mills(from Costa Pinto Mill, Piracicaba, Brazil) was diluted with distilledwater in order to obtain an initial sucrose concentration of 32 g/Lapproximately. The fermentation medium was enriched with 4 g/L of yeast extract to attend the nutritional requirements of the bac-terium. The Lactobacillus plantarum inoculum, from Fundação Trop-ical de Pesquisa e Tecnologia André Tosello (Campinas, Brazil),with adequate preparation [21] was added to the fermenter. Thetemperature was maintained at 37 �C, pH at 5.0 by adding 4 MNaOH and agitation speed at 200 rpm. A pulse of diluted molasses(32 g/L) was carried out after the sucrose had been completely con-sumed in order to avoid the inhibition of the cell growth by highsucrose concentration as well as to increase of lactic acid end con-centration [21]. The total time of fermentation was approximately30 h. The fermentation product containing about 5% (w/w) lacticacid was vacuum filtered and centrifuged (5000 rpm for 15 minat room temperature) to removal of Lactobacillus. The sedimentwas discarded and in the fermentation product (supernatantliquid) was added sulfuric acid to convert sodium lactate to lacticacid and used as raw material for the investigation of the evapora-tion system.

Fig. 1. Schematic diagram of evaporator [20].

2.2. Chromatographic analysis

Analyses of the raw material and products were performed inan equipment of high performance liquid chromatography (HPLC),Agilent model 1260, equipped with UV detector (UV/vis) con-nected in series with the chromatography column Bio-Rad Aminex,model HPX-87H (300 � 7.8 mm). The equipment was controlledthrough OpenLab software. Sulfuric acid solution with 5 mM wasused as mobile phase at flow rate of 0.6 mL/min. The column tem-perature was kept constant at 37 �C. In each run, an injection vol-ume of 25 lL was used. For lactic acid detection and quantification,the wavelength of 215 nm was used in the UV detection system.The lactic acid concentrations were determined using the calibra-tion curve (regression coefficient of 0.99987) obtained with stan-dard solutions of DL-Lactic acid 90% supplied by Sigma–Aldrich(St Louis, Missouri, EUA). The identification of the substances peaksin the chromatogram profiles was performed by comparison oftheir retention times with standard substances.

2.3. Hybrid short path evaporation system for lactic acid concentration

Lactic acid was concentrated in an evaporation system com-posed by a short path evaporator, Model Pope 2 Wiped Film Still,manufactured by Pope Scientific Inc. (Saukville, WI, USA). An exter-nal condenser was associated to the evaporation system which wasnamed hybrid short path evaporation. The distance between theevaporator and the internal condenser is 17 mm. The evaporatorhas an evaporation surface of 0.033 m2 (or 0.35 ft2) and it is jack-eted with an electric heating system. A schematic diagram of theapparatus is shown in Fig. 1 [20]. From tests carried out, it wasrealized that for pressures lower than 1.33 kPa a considerableamount of volatile material moved toward the trap, hinderingthe equipment operation at this condition, because the separatedmaterial in the trap returned to the evaporator [18]. In order toallow the use of lower pressures, in this work an external con-denser was attached to the equipment.

During the experiments, the external condenser was fixed at�5 �C. A trap was coupled to an external condenser which was con-tinuously fed with liquid nitrogen (�196 �C), freezing and avoidingthe volatiles migration to the pump and its oil contamination. Thetransfer of raw material (about 40 g at room temperature) to theequipment was conducted by using a peristaltic metering pumpCole Palmer Masterflex model 77200-60. The vacuum systemwas composed of a mechanical pump, keeping the pressure at1 kPa. At lower pressures a considerable amount of volatile mate-rial moved toward the trap, which is undesired.

In this system, it was possible to collect 03 streams: lightstream, residue and distillate as identified in Fig. 1. The substancesof higher vapor pressure were collected predominantly in the lightstream, while substances with intermediate vapor pressure in thedistillate and substances with lower vapor pressure in the residue.

A. Komesu et al. / Separation and Purification Technology 136 (2014) 233–240 235

The streams (light stream, residue and distillate) were collected inglass flasks and analyzed by liquid chromatography to determinelactic acid concentration.

2.4. Experimental design

In this work, an experimental design 24 with four factors andthree replicates at central point was used to investigate the impactof main operational parameters on the process performance. Thefactors selected were: feed flow rate (FFR), agitation (Agit), con-denser temperature (Tcond) and evaporator temperature (Tevap),which were represented by dimensionless coded variables X1, X2,X3 and X4, respectively. Response variables were: residue (%R)and distilled percentages (%D), lactic acid purities at residue(%PR) and distillate streams (%PD) and lactic acid recoveries at res-idue (%RecR) and distilled streams (%RecD).

Table 1 shows coded variables and real values used in thematrix of experiments. Coded and real variable values are relatedby the general equation: Xi ¼ ni�n0

Dni, where Xi is the dimensionless

coded value of the ith factor, ni is the real value of the ith factor,n0 is the real value of the ith factor at the central point, and Dni

is the step change value of the real variable i [19].The experiments were carried out in a randomized order and

three replicates at central point of the design were performed toallow the estimation of the pure error (runs 17, 18 and 19).

The effect of each process variable and their interactions in theresponse variables were calculated using the software STATISTICA7.0 from Statsoft Inc. (2004). The relationship between factors andeach response variable was modeled by fitting the polynomialequation given by Eq. (1). The quality of the fitted models was val-idated by analysis of variance (ANOVA).

Y ¼ b0 þ b1X1 þ b2X2 þ b3X3 þ b4X4 þ b12X1X2 þ b13X1X3

þ b14X1X4 þ b23X2X3 þ b24X2X4 þ b34X3X4 ð1Þ

In which X1, X2, X3 and X4 are the independent coded variables indimensionless form, b0, b1, b2, b3, b4, b12, b13, b14, b23, b24 and b34

are the regression coefficients, and Y is the response function.

Table 1Two level factorial design matrix and experimental results at residue and distillate stream

Runs Coded variables Real variables

X1 X2 X3 X4 FFR (mL/min) Agit (rpm) Tcond (�C)

1 �1 �1 �1 �1 8 250 72 +1 �1 �1 �1 20 250 73 �1 +1 �1 �1 8 1250 74 +1 +1 �1 �1 20 1250 75 �1 �1 +1 �1 8 250 136 +1 �1 +1 �1 20 250 137 �1 +1 +1 �1 8 1250 138 +1 +1 +1 �1 20 1250 139 �1 �1 �1 +1 8 250 7

10 +1 �1 �1 +1 20 250 711 �1 +1 �1 +1 8 1250 712 +1 +1 �1 +1 20 1250 713 �1 �1 +1 +1 8 250 1314 +1 �1 +1 +1 20 250 1315 �1 +1 +1 +1 8 1250 1316 +1 +1 +1 +1 20 1250 13

17 (C) 0 0 0 0 14 750 1018 (C) 0 0 0 0 14 750 1019 (C) 0 0 0 0 14 750 10

FFR (mL/min), feed flow rate; Agit (rpm), agitation; Tcond (�C), condenser temperature; Tpurity at residue; %RecR, lactic acid recovery at residue;%D, mass percentage of distillate

3. Results and discussion

3.1. Raw material component analyses

After the fermentation, the resulting product was analyzed byHPLC. This step is important to determine the concentration ofthe substances obtained in each stream (light stream, residueand distillate) and thus, it will be used to evaluate operating con-ditions and separation process performance. The chromatogramprofile is presented in Fig. 2. It allowed the identification of 04 sub-stances, to know: sucrose (7.619 min), glucose (8.709 min), fruc-tose (9.593 min) and lactic acid (12.686 min). By calibrationcurve, the lactic acid mass concentration in the fermentation prod-uct was about 5% (w/w).

3.2. Separation process results

Table 2 shows the experimental results for the residue (%R) andfor the distilled (%D) mass percentages, lactic acid purities (%PRand %PD) and lactic acid recoveries (%RecR and %RecD) at residueand distillate streams. Mass percentage was defined as the ratiobetween mass of distilled or residue stream and the sum of resi-due, distillate and light mass after evaporation. The lactic acidrecovery was calculated by Eq. (2).

Recð%Þ ¼ mi �%LAiP3

i¼1mi �%LAi

� 100 ð2Þ

In Eq. (2) i is the index for residue or distillate or light; m is the res-idue, distillate or light mass (g) and %LA is the fraction of lactic acidin residue, distillate and light.

3.3. Residue and distilled percentage analyses

Table 2 shows the effects of operational hybrid separation pro-cess variables on residue and distillate percentage with a confi-dence level of 95%. According to Table 2, the feed flow rate,agitation and evaporator temperature were statistically significantfor residue percentage. For distilled percentage, only the agitationwas statistically significant. Preliminary studies by the author’sresearch group with other raw material showed that the variables

.

Residue Distillate

Tevap (�C) %R %PR %RecR %D %PD %RecD

80 82.91 7.34 93.88 12.32 1.79 3.4080 86.80 6.40 93.52 4.39 7.93 5.8680 24.09 20.33 81.34 19.16 2.49 7.9480 57.86 10.09 84.45 11.03 8.01 12.7880 85.54 7.70 82.69 2.03 34.50 8.7980 97.84 7.01 99.10 0.00 0.00 0.0080 25.32 22.56 88.37 0.61 32.70 3.0780 38.99 12.92 63.51 12.81 18.71 30.23

120 73.75 7.71 73.80 7.67 21.31 21.21120 93.77 8.05 91.30 1.78 31.48 6.76120 4.97 31.47 31.22 18.41 13.30 48.84120 30.33 12.33 67.99 13.82 9.18 23.08120 71.61 7.98 84.93 0.00 0.00 0.00120 96.54 8.08 84.58 2.21 54.12 12.98120 2.34 34.35 13.58 9.52 31.21 50.20120 29,06 15.18 54.54 7.90 28.55 27.90100 31.08 14.01 63.54 5.06 30.04 25.50100 23.85 14.56 73.26 7.06 35.91 34.47100 21.64 14.50 65.01 11.67 26.78 21.79

evap (�C), evaporator temperature; %R, mass percentage of residue; %PR, lactic acid; %PD, lactic acid purity at distillate; %RecD, lactic acid recovery at distillate.

Fig. 2. Fermentation broth chromatographic profile in HPLC. Legend: (1) sucrose,(2) glucose, (3) fructose, and (4) lactic acid.

236 A. Komesu et al. / Separation and Purification Technology 136 (2014) 233–240

evaporator temperature and feed flow rate were the ones that hadgreatest effect on the response variable in the short path evapora-tion process [18,19]. The results obtained in this work confirm this.

The regression models for residue (%R) and distilled percentage(%D), considering only the statistically significant variables, aregiven by Eqs. (3) and (4), respectively. The factors X1, X2 and X4 rep-resent coded values of feed flow rate, agitation and evaporatortemperature, respectively.

%R ¼ 51:4889þ 10:0413� X1 � 29:7375� X2 � 6:0613� X4 ð3Þ%D ¼ 7:76053þ 3:92875� X2 ð4Þ

Table 2Estimated effects on residue percentage and on distilled percentage with a confidence lev

Factor Residue

Regression coefficient Standard error t(2) p

Mean 51.4889 1.132729 45.4557 0.00048(1)FFR 10.0413 2.468726 8.1348 0.01477(2)Agit �29.7375 2.468726 24.0914 0.00171(3)Tcond �0.4525 2.468726 �0.3666 0.74907(4)Tevap �6.0613 2.468726 �4.9104 0.03905(1) ⁄ (2) 2.3987 2.468726 1.9433 0.19144(1) ⁄ (3) �0.3388 2.468726 �0.2744 0.80950(1) ⁄ (4) 2.0875 2.468726 1.6912 0.23287(2) ⁄ (3) �2.2400 2.468726 �1.8147 0.21123(2) ⁄ (4) �3.8838 2.468726 �3.1464 0.08790(3) ⁄ (4) 0.0437 2.468726 0.0354 0.97494

Table 3ANOVA of residue and distilled percentage model (confidence level of 95%), lactic acid purlactic acid recovery at residue and distillate model (confidence level of 95%).

Source of variation Sum of squares Degrees of freedom Mean square Fcalculated Ftabul

Residue percentageRegression 16350.15 3 5450.05 26.00 F3,15

Residues 3143.71 15 209.58 9.77 F13,2

Lack of fit 3094.95 13 238.07Pure error 48.76 2 24.38Total 19493.86 18

Lactic acid purity at residueRegression 1156.79 7 165.26 54.47 F7,11

Residues 33.37 11 3.03 40.51 F9,2 =Lack of fit 33.19 9 3.69Pure error 0.18 2 0.09Total 1190.17 18

Lactic acid recovery at residueRegression 6599.10 4 1649.77 11.74 F4,14

Residues 1966.85 14 140.49 5.80 F12,2

Lack of fit 1911.95 12 159.33Pure error 54.90 2 27.45Total 8565.95 18

The models were evaluated through ANOVA. Table 3 shows theANOVA for the residue and distilled percentages. In order to evalu-ate if the models are statistically significant, one criterion is toattend F-test. F values are calculated by the ratio between the meansquare of regression and the mean square residual (Eq. (5)) and bythe ratio between the mean square of lack of fit and the meansquare of pure error (Eq. (6)). Then, these values are compared withtabulated F values considering the same confidence level. If theFcalculated by Eq. (5) is higher than Ftabulated and the Fcalculated byEq. (6) is lower than Ftabulated, these are indication that the modelrepresents well experimental data.

MQR

MQr� Fn;m ð5Þ

where MQR is the mean square of regression; MQr is the meansquare of residues; F is the F-distribution; n and m are the freedomnumber of regression and residues, respectively.

MQLOF

MQPE� Fs;t ð6Þ

where MQLOF is the mean square of lack of fit; MQPE is the meansquare of pure error; F is the F-distribution; s and t are the freedomnumber of lack of fit and pure error, respectively.

According to Table 3, the F3,15 calculated (26.00) was higherthan F3,15 tabulated (3.29) and 96.2% of variation is explained bythe model, showing thus, that the linear model (Eq. (3)) to the

el of 95%.

Distilled

Regression coefficient Standard error t(2) p

4 7.76053 0.777672 9.97918 0.0098937 �0.98625 1.694897 1.16379 0.3645729 3.92875 1.694897 4.63598 0.0435148 �3.34375 1.694897 3.94567 0.0586409 �0.06500 1.694897 0.07670 0.9458442 0.71875 1.694897 0.84813 0.4856800 2.33125 1.694897 2.75091 0.1106427 �0.25000 1.694897 0.29500 0.7957974 �0.60375 1.694897 0.71243 0.5500980 0.82000 1.694897 0.96761 0.4353206 0.58750 1.694897 0.69326 0.559835

ity at residue (confidence level of 99%) and distillate model (confidence level of 95%),

ated Sum of squares Degrees of freedom Mean square Fcalculated Ftabulated

Distilled percentage= 3,29 246.96 1 246.96 10.53 F1,17 = 4.45= 19,42 398.64 17 23.45 2.18 F15,2 = 19.43

375.66 15 25.0422.98 2 11.49

645.60 18

Lactic acid purity at distillate= 4.89 1666.93 3 555.64 3.35 F3,15 = 3.2999.39 2488.86 15 165.92 8.79 F13,2 = 19.42

2446.04 13 188.1642.81 2 21.41

4155.78 18

Lactic acid recovery at distillate= 3.11 2198.36 2 1099.18 8.75 F2,16 = 3.63= 19.41 2011.00 16 125.69 3.24 F14,2 = 19.42

1925.99 14 137.5785.00 2 42.50

4209.36 18

A. Komesu et al. / Separation and Purification Technology 136 (2014) 233–240 237

residue percentage is statistically significant to a confidence levelof 95%. In addition, the F13,2 calculated (9.77) was lower thanF13,2 tabulated (19.42), it means that the model given by Eq. (3)can be used to make predictions in the range studied.

Observing Eq. (3), the amount of residues decreased withincrease in temperature because the high temperature promotesthe evaporation of a larger amount of material. Therefore, the per-centage of other streams increased. This is in accordance with pre-liminary results with synthetic mixture of water: lactic acid [20].Similarly, the higher agitation promotes a reduction of the percent-age at residue stream. The agitation system is designed to provide aturbulent flow of the material inside the evaporation system. Then,higher agitation favors the mechanism of mass and heat transfersand consequently, the percentage of light stream and distillateincrease. On the other hand, with a relatively high feed flow rate,the residue percentage increases, since there is not enough timefor volatilization of the material. By lowering the feed flow rate,the exposure time was prolonged which might increase the riskof decomposition of thermally sensitive components [14].

Table 3 shows a F1,17 calculated (10.53) higher than F1,17

tabulated (4.45) and the F15,2 calculated (2.18) lower than F15,2

tabulated (19.43), attending F-test. However, for the distillate per-centage, only 38.3% of the total variation is explained by the model.From these results, it can be concluded that the linear model is notadequate to represent the variation of the distillate percentage. Asecond-order polynomial equation could be used to it, but anotherexperimental design is required, such as central composite design.

3.4. Lactic acid purity analysis

Table 4 shows the effects of different process operational vari-ables on lactic acid purity at residue and distillate streams. Consid-ering a confidence level of 99% at residue, all main effects and theinteractions (1) ⁄ (2), (1) ⁄ (4) and (2) ⁄ (4) are statistically signifi-cant. Considering a confidence level of 95% at distillate, condensertemperature, evaporator temperature and the interaction (1) ⁄ (4)are statistically significant.

Models for residue and distillate purities are represented byEqs. (7) and (8), where only the statistically significant factors wereconsidered. Factors X1, X2, X3 and X4 represent coded values of feedflow rate, agitation, condenser temperature and evaporator tem-perature, respectively.

%PR ¼ 13:81947� 3:71125� X1 þ 6:18500� X2

þ 0:75375� X3 þ 1:92500� X4 � 3:56250� X1

� X2 � 1:02250� X1 � X4 þ 1:50375� X2 � X4 ð7Þ%PD ¼ 20:42158þ 6:51875� X3 þ 5:18875� X4

þ 5:89625� X1 � X4 ð8Þ

Table 4Estimated effects on lactic acid purity at residue (confidence level of 99%) and distillate (c

Factor Residue

Regression coefficient Standard error t(2) p

Mean 13.81947 0.069219 199.6494 0.00002(1)FFR �3.71125 0.150859 �49.2017 0.00041(2)Agit 6.18500 0.150859 81.9973 0.00014(3)Tcond 0.75375 0.150859 9.9928 0.00986(4)Tevap 1.92500 0.150859 25.5206 0.00153(1) ⁄ (2) �3.56250 0.150859 �47.2296 0.00044(1) ⁄ (3) 0.03625 0.150859 0.4806 0.67824(1) ⁄ (4) �1.02250 0.150859 �13.5557 0.00539(2) ⁄ (3) 0.59500 0.150859 7.8882 0.01569(2) ⁄ (4) 1.50375 0.150859 19.9359 0.00250(3) ⁄ (4) 0.00000 0.150859 0.0000 1.00000

In order to evaluate the models, F-test was applied. Table 3 showsthe ANOVA for lactic acid purity at residue and distilled, respec-tively. For Eq. (7), the F7,11 calculated (54.47) was higher thanF7,11 tabulated (4.89) and the F9,2 calculated (40.51) was lower thanF9,2 tabulated (99.39), attending F-test. The model explains 97.2% ofthe total variation, which shows that the model is statistically sig-nificant to confidence level of 99%. Observed and predicted valuesfor lactic acid purities at residue stream are given in Fig. 3 and goodmodel fitness to experimental data can be observed.

From Eq. (7), it can be seen that, increasing the feed flow rate,the purity decreases, since there is not enough time for lactic acidto be separated. By high agitation, the mechanism of mass and heattransfers are potentially increased, so the molecules are volatizedand the purity increases. Similarly, the purity increases with evap-orator temperature, which was expected, because the increment oftemperature enlarged the evaporation rate [14].

According to Table 3, for Eq. (8), the F3,15 calculated value (3.35)is very close to F3,15 tabulated value (3.29). If the Fcalculated is at leastthree times higher than Ftabulated this is an indication that themodel represents well experimental data [18]. In this case,Fcalculated is 1.02 times higher than Ftabulated and only 40.1% of thevariation is explained by the model, therefore, the linear model(Eq. (8)) is not satisfactory to fit experimental data of lactic acidpurity at distillate stream.

The response surface of lactic acid purity at residue (%PR) infunction of agitation and feed flow rate (Fig. 4a) and evaporatortemperature and condenser temperature (Fig. 4b), consideringonly statistically significant variables, is given in Fig. 4. Consider-ing good separation performance in order to obtain higher lacticacid purity at residue, it is favorable to work with higher agita-tion (1250 rpm), lower feed flow rate (7 mL/min), higher evapora-tor temperature (120 �C) and higher condenser temperature(13 �C).

The results showed that it is possible to increase lactic acid pur-ity at residue and distillate changing some variables of the process.Then, in future works, the maximum lactic acid purity can beachieved through the optimization of the process variables.

3.5. Lactic acid recovery analysis

One important process parameter to evaluate the separationprocess is the lactic acid recovery. Table 5 shows the variableeffects on lactic acid recovery at residue and distillate. Accordingto Table 5, the agitation, evaporator temperature and the interac-tions (1) ⁄ (4) and (2) ⁄ (4) are statistically significant on lactic acidrecovery at residue stream to a confidence level of 95%. However,on the distillate stream, only agitation and evaporator temperatureare statistically significant on lactic acid recovery to a confidencelevel of 95%.

onfidence level of 95%).

Distillate

Regression coefficient Standard error t(2) p

5 20.42158 1.061451 19.23930 0.0026913 1.29250 2.313380 1.11741 0.3800389 �0.43625 2.313380 �0.37715 0.7423186 6.51875 2.313380 5.63569 0.0300722 5.18875 2.313380 4.48586 0.0462728 �3.19875 2.313380 �2.76543 0.1096667 �0.92125 2.313380 �0.79645 0.5092908 5.89625 2.313380 5.09752 0.0363964 3.25500 2.313380 2.81406 0.1064877 �2.64750 2.313380 �2.28886 0.1492860 �1.69250 2.313380 �1.46323 0.280953

Fig. 3. Predicted and observed values for lactic acid purity at residue.

Fig. 4. Response surface of lactic acid purity at residue (%PR) in function of (a)agitation (Agit) and feed flow rate (FFR) and (b) evaporator temperature (Tevap) andcondenser temperature (Tcond).

238 A. Komesu et al. / Separation and Purification Technology 136 (2014) 233–240

The regression models to lactic acid recovery at residue and dis-tillate streams are given by Eqs. (9) and (10), respectively, consid-ering only factors statistically significant. Factors X1, X2 and X4

represent coded values of feed flow rate, agitation and evaporatortemperature, respectively.

%RecR ¼ 73:10900� 13:6750� X2 � 11:5575� X4 þ 6:2863� X1 � X4 � 7:2350� X2 � X4 ð9Þ

%RecD ¼ 16:43862þ 9:06480� X2 þ 7:43017� X4 ð10Þ

ANOVA for lactic acid recovery at residue and distilled are givenin Table 3. For Eq. (9), the F4,14 calculated (11.74) was higher thanF4,14 tabulated (3.11) and the F12,2 calculated (5.80) was lower thanF12,2 tabulated (19.41), satisfying F-test. In addition, 77.0% of thevariation is explained by the model, showing that the model is ade-quate to representing the experimental data.

Analyzing Eq. (9), the lactic acid recovery at residue decreasedwith the increase of evaporator temperature. This is in accordancewith preliminary results with synthetic mixture of water: lacticacid [9]. The same behavior is observed increasing the agitation.

Table 3 shows a F2,16 calculated (8.75) higher than F2,16 tabu-lated (3.63) and F14,2 calculated (3.24) lower than F14,2 tabulated(19.42). However, only 52.2% of the variation is explained by themodel. Thus, the linear model (Eq. (10)) used to represent the lacticacid recovery not adjust well the experimental data.

The relationship between the independent and dependent vari-ables is shown in a three-dimensional representation of theresponse surface. Fig. 5 shows the response surface of lactic acidrecovery at residue (%RecR) in function of agitation and evaporatortemperature. The figure was constructed considering only statisti-cally significant variables. From Fig. 5, it can be seen that higherlactic acid recovery at residue was observed at lower agitationand lower evaporator temperature.

3.6. Process evaluation

The results showed that while the mass percentage range variedfrom 2.34% to 97.84% at residue, this value varied from 0.00% to19.16% for distillate. This means that more material tends tomigrate to the residue stream at low temperatures. Similar behav-ior was obtained with synthetic solution lactic acid: water in pre-vious work [20].

In terms of highest lactic acid purity produced in the streams:34.35 ± 0.30% was reached in run 15 in residue and 54.12 ± 4.62%was reached in run 14 in distillate. Considering that the percentage(2.34% at residue and 2.21% at distillate) and recovery in each ofthe runs was similar (13.58% at residue and 12.98% at distillate),lactic acid recovery at distillate is more interesting.

The factorial experimental design allowed the determinationand evaluation of the relative significance of parameters: feed flowrate, agitation, condenser and evaporator temperature on the pro-cess based on the responses: percentage, purity and recovery. Thepresence of complex interactions between the operating condi-tions was observed in the lactic acid purity at residue and distillateand lactic acid recovery at residue.

The agitation and temperature evaporator were the parametersthat had higher influence on the process in the range studied. Inaddition, percentage and lactic acid purity at residue were theresponses that had more variable statistically significant.

3.7. Improvement purity at distillate

In previous work [20], the purification process of lactic acidfrom a synthetic mixture was considered, varying only the evapo-rator temperature. The preliminary study was important becauseshowed that it is possible to concentrate the lactic acid using the

Table 5Estimated effects on lactic acid recovery at residue and distillate with a confidence level of 95%.

Factor Residue Distillate

Regression coefficient Standard error t(2) p Regression coefficient Standard error t(2) p

Mean 73.1900 1.201979 60.8913 0.000270 18.14737 1.495629 12.13360 0.006724(1)FFR 5.5737 2.619652 4.2553 0.051034 �1.49125 3.259648 �0.91498 0.456792(2)Agit �13.6750 2.619652 10.4403 0.009050 9.06500 3.259648 5.56195 0.030838(3)Tcond �2.8875 2.619652 �2.2045 0.158308 0.20625 3.259648 0.12655 0.910874(4)Tevap �11.5575 2.619652 �8.8237 0.012602 7.43125 3.259648 4.55954 0.044887(1) ⁄ (2) 1.4238 2.619652 1.0870 0.390599 �0.51625 3.259648 �0.31675 0.781438(1) ⁄ (3) �1.5538 2.619652 �1.1862 0.357352 2.62250 3.259648 1.60907 0.248877(1) ⁄ (4) 6.2863 2.619652 4.7993 0.040778 �4.70000 3.259648 �2.88375 0.102155(2) ⁄ (3) �2.7375 2.619652 �2.0900 0.171792 2.13875 3.259648 1.31226 0.319811(2) ⁄ (4) �7.2350 2.619652 �5.5236 0.031247 4.56875 3.259648 2.80322 0.107184(3) ⁄ (4) �0.4475 2.619652 �0.3416 0.765173 �1.30750 3.259648 �0.80223 0.506594

Fig. 5. Response surface of lactic acid recovery in function of agitation (Agit) andevaporator temperature (Tevap) at residue (%RecR).

A. Komesu et al. / Separation and Purification Technology 136 (2014) 233–240 239

evaporation system. Lactic acid was concentrated in residuestream in about 2.4 times with a recovery of 94%.

Lactic acid from fermentation broth is a mixture more complexthan synthetic mixture water: lactic acid, as can be seen in Fig. 2. Inaddition, lactic acid initial concentration in fermentation broth(5 wt% of lactic acid) is lower than synthetic mixture (36 wt% oflactic acid) used in previous work. Because these it was expectedthat the performance of the separation process would be different.

The factorial experimental design with fermentation brothshowed that lactic acid recovery at distillate was more interestingthan recovery at residue stream, and condenser and evaporatortemperature were significant variables of the process. In order toimprove the purity at distillate, experiments were performed atfeed flow rate of 14 mL/min, stirring of 750 rpm, pressure of

Table 6Experimental ranges and results at distillate stream.

Runs Tcond (�C) Tevap (�C) %PD %RecD

20 13 115.9 22.18 29.6221 16 131.8 55.33 71.4922 19 147.8 79.94 37.3623 22 163.7 89.71 14.27

Tcond (�C), temperature condenser; Tevap (�C), evaporator temperature; %PD, lacticacid purity at distillate; %RecD, lactic acid recovery at distillate.

1 kPa and varying the condenser and evaporator temperature.The experimental ranges and the results are shown in Table 6.

As expected, an increase in purity was obtained with evaporatortemperature. The maximum lactic purity of 89.71% was obtained inrun 23 which was about 18 times lactic acid concentration higherthan the initial content in raw material. Working with higher evap-orator temperature is not recommended because lactic acid ther-mal decomposition in many decomposition products such aswater, hydroxyl, propionic acid, acrylic acid, acetaldehyde, carbondioxide and formate. On the other hand, lactic acid recovery inrun 23 was low. The maximum recovery of 71.49% was obtainedin run 21. A possible alternative to allow high purity and recoveryis make multiple-pass distillation in run 21.

4. Conclusions

The influence of feed flow rate, agitation, condenser and evapo-rator temperature on the purification process of lactic acid fromfermentation broth was studied in this work. Lactic acid with pur-ity of around 89.71% at distillate was obtained.

Based on the results obtained in this work, it can be concludedthat carrying out the lactic acid concentration by using hybridshort path evaporation system is technically feasible and advanta-geous because the use of high operating pressure (1000 Pa) andone step of refining, significantly reduces the process costs.

Acknowledgements

The authors are grateful to the financial support of CAPES, CNPqand FAPESP.

References

[1] R. Datta, M. Henry, Lactic acid: recent advances in products, processes andtechnologies – a review, J. Chem. Technol. Biotechnol. 81 (2006) 1119–1129.

[2] K.L. Wasewar, A.B.M. Heesink, G.F. Versteeg, V.G. Pangarkar, Reactiveextraction of lactic acid using alamine 336 in MIBK: equilibria and kinetics, J.Biotechnol. 97 (2002) 59–68.

[3] M. Järvinan, L. Myllykoski, R. Keiski, J. Schlo, Separation of lactic acid fromfermented broth by reactive extraction, Bioseparation 9 (2000) 163–166.

[4] E. Alkaya, S. Kaptan, L. Ozkan, S. Uludag-demirer, G.N. Demirer, Recovery ofacids from anaerobic acidification broth by liquid–liquid extraction,Chemosphere 77 (2009) 1137–1142.

[5] M.-T. Gao, T. Shimamura, N. Ishida, E. Nagamori, H. Takahashi, S. Uemoto, T.Omasa, H. Ohtake, Extractive lactic acid fermentation with tri-n-decylamine asthe extractant, Enzyme Microb. Technol. 44 (2009) 350–354.

[6] M. Matsumoto, T. Takahashi, K. Fukushima, Synergistic extraction of lactic acidwith alkylamine and tri-nbutylphosphate: effects of amines, diluents andtemperature, Sep. Purif. Technol. 33 (2003) 89–93.

[7] G. Malmary, J. Albet, A. Putranto, H. Hanine, J. Molinier, Recovery of aconiticand lactic acids from simulated aqueous effluents of the sugar-cane industrythrough liquid–liquid extraction, J. Chem. Technol. Biotechnol. 75 (2000)1169–1173.

240 A. Komesu et al. / Separation and Purification Technology 136 (2014) 233–240

[8] P. Pal, J. Sikder, S. Roy, L. Giorno, Process intensification in lactic acidproduction: a review of membrane based processes, Chem. Eng. Process. 49(2009) 1549–1559.

[9] M.I. González, S. Alvarez, F.A. Riera, R. Álvarez, Lactic acid recovery from wheyultrafiltrate fermentation broths and artificial solutions by nanofiltration,Desalination 228 (2008) 84–96.

[10] V. Hábová, K. Melzoch, M. Rychtera, B. Sekavová, Electrodialysis as a usefultechnique for lactic acid separation from a model solution and a fermentationbroth, Desalination 163 (2004) 361–372.

[11] L. Mo, J.S. Shao-Tong, P. Li-Jun, Z. Zhi, L. Shui-Zhong, Design and control ofreactive distillation for hydrolysis of methyl lactate, Chem. Eng. Res. Des. 89(2011) 2199–2206.

[12] E.A. Edreder, I.M. Mujtaba, M. Emtir, Optimal operation of different types ofbatch reactive distillation columns used for hydrolysis of methyl lactate tolactic acid, Chem. Eng. J. 172 (2011) 467–475.

[13] R. Kumar, H. Nanavati, S.B. Noronha, S.M. Mahajani, A continuous process forthe recovery of lactic acid by reactive distillation, J. Chem. Technol. Biotechnol.81 (2006) 1767–1777.

[14] L. Chen, A. Zeng, H. Dong, Q. Li, C. Niu, A novel process for recovery and refining ofl-lactic acid from fermentation broth, Bioresour. Technol. 112 (2012) 280–284.

[15] S.-L. Xu, T. Zheng, S.-M. Xu, The study of refining L-Lactic acid by moleculardistillation, J. Chem. Eng. Chin. Univ. 18 (2004) 246–249.

[16] Q. Wei, Z.W. Han, D.M. Bai, Z.H. Yan, X.M. Zhao, Study on recovery andpurification of l-lactic acid, in: Proceedings of the 4th International alConference on Separation Science and Technology: Frontiers on SeparationScience and Technology, Nanning, Guangxi, China, 2004, pp. 717–723.

[17] P.F. Martins, H.H.R. Medeiros, P. Sbaite, M.R. Wolf Maciel, Enrichment ofoxyterpenes from orange oil by short path evaporation, Sep. Purif. Technol.116 (2013) 385–390.

[18] P.F. Martins, C. Carmona, E.L. Martinez, P. Sbaite, R. Maciel Filho, M.R. WolfMaciel, Short path evaporation for methyl chavicol enrichment from basilessential oil, Sep. Purif. Technol. 87 (2012) 71–78.

[19] P.F. Martins, C. Carmona, E.L. Martinez, P. Sbaite, R. Maciel Filho, M.R. WolfMaciel, Evaluation of methyl chavicol concentration by different evaporationprocesses using central composite experimental design, Sep. Purif. Technol. 98(2012) 464–471.

[20] A. Komesu, P.F. Martins, B.H. Lunelli, A.T. Morita, P.L.A. Coutinho, R. MacielFilho, M.R. Wolf Maciel, Lactic acid purification by hybrid short pathevaporation, Chem. Eng. Trans. 32 (2013) 2017–2022.

[21] B.H. Lunelli, R.R. Andrade, D.I.P. Atala, M.R. Wolf Maciel, F. Maugeri Filho,R. Maciel Filho, Production of lactic acid from sucrose: strain selection,fermentation and kinetic modeling, Appl. Biochem. Biotechnol. 161 (2010)227–237.