Contents lists available at SciVerse ScienceDirect
Journal of Membrane Science
jo u rn al hom epa ge: www.elsev ier .com/ locate /memsci
anofiltration as key technology for the separation of LA and AA
. Eckera,∗, T. Raabb, M. Haraseka
Vienna University of Technology, Institute of Chemical Engineering, Getreidemarkt 9/166, 1060 Vienna, AustriaSchool of Engineering and Environmental Sciences, Stelzhamerstrasse 23, 4600 Wels, Austria
r t i c l e i n f o
rticle history:eceived 8 September 2011eceived in revised form 31 October 2011ccepted 1 November 2011vailable online 9 November 2011
Nanofiltration as state-of-the-art technology was used for the separation of lactic acid (LA) and aminoacids (AA) in a ‘Green Biorefinery’ pilot plant. For this process, the performances of six different nanofiltra-tion membranes were compared by experiments in lab scale. In this work the focus was on the separationof the two products, LA and AA. Enhanced differences in the retentions were required to produce twopurified process streams, LA enriched permeate and amino acid enriched retentate. In the referenceexperiment, performed with original solution from the ‘Green Biorefinery’ pilot plant, the retention val-ues were about 60% for LA, and about 88% for AA, this hindered good performance in the separation of themain components. Process optimization with pH value variations and different diafiltration-modes wereinvestigated; one experiment was done with original solution, two tests dealt with varying pH-values,two with different diafiltration rates. A pH-variation from 3.9 (reference solution) down to 2.5 transferredthe chemical structure of LA, which reduced the retention of the LA significantly from 67% to 42% for the
membrane DL (Osmonics). Beside the separation, further attention was given to the flux behaviour. Allscreening scenarios were compared with a reference experiment done with original solution and stan-dard process parameters as used in the plant itself to evaluate the efficiency trends shown in the tests.It was shown that a nanofiltration unit allowed the separation of sufficient degree for further treatmenttechnologies between AA and LA, a membrane screening for the optimization of this process ensuredbest performance in practice.
. Introduction
Manufacturing of high potential products from renewable mate-ials is the main focus of a green biorefinery. This leads to anconomic and ecological upgrade of unused agriculturally avail-ble land, [1]. In this context, the ‘Green Biorefinery Upper Austria’eals with grass silage as raw material to upgrade it to a higher
evel of usage. In the project lactic acid (LA) and amino acids (AA)re recovered from this renewable resource. LA, which is manu-actured by fermentation by default, can be used in the food andhemical industry. AA production is also based on fermentationechnology in general. The acids play an important role in pharma-eutical, food and biotechnological processes, depending on theirurity level.
Downstream technologies are cost intensive steps in theanufacturing processes, hence efficiency and recovery rate
ave to be optimized. A separation of products without chemicalnput or phase changes offers purification at low conditioningosts. In contrast to standard separation processes, i.e. extraction,
ion-exchanger and distillation, membrane based technologiescombine selectivity with lower demand of chemicals (cleaning)and heat input [2]. Ultrafiltration, nanofiltration (NF) and reverseosmosis are alternative separation technologies. Novalin andZweckmair [1] discussed their application in the biorefinery sectorin detail. The industrial applications of membrane based processesallow a compact structure due to the modular design, and are veryflexible in scale. The processes combine production and purifica-tion in one operation unit [3]. In dependency of the componentsefficient separation with membrane technology enables the abilityto produce two valuable process streams, retentate and permeate,in one operation unit. Separation, purification and recovery of theproducts influence the effectiveness of the whole process chain.
There is a wide range of usage for nanofiltrations. Next to inor-ganic salts, organic components can be separated with this type ofmembrane filtration systems. The production of AA with nanofil-tration membranes was investigated in previous papers [4] and[5]. Timmer et al. determined the separation and purification ofan artificial amino acid solution [6]. As different AA have very sim-ilar retentions, the usage of membrane technology for separation
is not possible. Further enrichment and separation has to be doneapplying other technologies, e.g. ion exchanger.
Desalination and product recovery of LA or AA were done in for-mer papers [7,8]. Nevertheless, the combined separation of these
Table 1Nanofiltration membranes used in the study (hp: hydrophilic, hb: hydrophobic).
Name DK DL HLProducer Osmonics Osmonics OsmonicsLabel DK DL HLMembrane material hp hp hpMWCO 100–300 100–300 100–300Retention [%] 98 (MgSO4) 96 (MgSO4) 95(MgSO4)Pure water
permeability[L/(h m2 bar)]
2.81 3.88 3.18
Name HT FT KOProducer Hydranautics Filmtech KochLabel HT NF270 MPF36Membrane material hb hp hbMWCO 720 270 1000Retention [%] 97 (MgSO4) 10 (NaCl) 10 (NaCl)Pure water 1.22 9.76 1.96
90 J. Ecker et al. / Journal of Mem
wo main components from one feed solution with NF membraneshows high potential for application of this membrane based pro-ess. The use of NF membranes for the recovery of two products,ne in the permeate and the other in the retentate, requires selec-ive separation of the components. Former investigations paid theirttention to enrich LA in the retentate of a NF [9]. Thang andovalin [10], tested an ion exchanger process for the separationf LA from a silage juice. However, in this study the amount of therganic acid in the permeate was targeted. Due to their chemicaltructure AA determine high retentions, the retentate concentra-ion can be increased. The influence of the process parametersn the separation is an approach to the topic, not covered inhe literature so far. The pH value variation and the diafiltrationre two potential approaches for the influence on the separationerformance.
Koschuh et al. [5] tested cellulosic and ceramic membranesor the purification of silage juice as a pretreatment step previ-us to nanofiltration where the intrinsic separation of LA and AAake place. The tests showed the need of a cut off between 100nd 400 Da to gain sufficient results, which can be accomplishedsing polymeric nanofiltration membranes. Furthermore, treatingiological solutions require high pH stability, especially for theleaning procedure. Polymeric membranes can be operated in aide pH-range. Standard applications are done at pH 12 and pH 2,
espectively, to ensure the removal of all organic residues.The study was carried out with raw solution obtained from the
reen Biorefinery Upper Austria. The main components of this solu-ion are LA and a mixture of AA. The reference experiment presentshe proof of concept of the refinery and the specifications for theeparation of LA and AA using NF. In the pilot plant the NF is theain important separation unit, the performance influences the
urity and the recoveries of both product lines. For further treat-ent of the LA enriched stream, the concentrations of AA in the
ermeate should be minimized. The remaining LA in the secondroduct stream (NF-retentate), can be reused in this type of pilotlant after removing AA, as additional feed stream for the NF itself.
n contrast to results gained from artificial solutions, the tests focusn the interaction of the principal constituents.
In this paper the separation and purification of LA and AA inne step nanofiltration has been investigated. To obtain this goalifferent six membranes and different operation conditions like pHalue variation and diafiltration rates were also studied.
Fig. 1. Schematic expe
permeability[L/(h m2 bar)]
2. Materials and methods
2.1. Membranes and experimental setup
The screening of the membranes was done with a lab scaleexperimental setup. For process optimization six different poly-meric membranes, hydrophilic and hydrophobic, with varyingmolecular weight cut off (MWCO) were screened. Detailed prop-erties of all nanofiltration membranes are shown in Table 1. Threemembranes were tested simultaneously, a schematic experimentalsetup is shown in Fig. 1.
All membranes were implemented as flat sheet samples, theactive area of each test cell was 0.0127 m2. All experiments weredone at equal conditions with regard to value for recirculationrate, temperature and pressure levels (1.02 m3/h, 25 ◦C); the exper-iments were done at a pressure range between 15 and 25 bar. Toensure constant feed concentrations, the tank volume was period-ically reconditioned with fresh experimental raw solution.
The screening of the six membranes included one experiment
(I), which was done with the original solution from the pilot plantwithout any changes in the mixture or pH. Two tests were dealtwith varied pH-values (II and III) and two with different diafiltrationrates (IV and V). In Table 2 there is a summary of the experiments
rimental setup.
J. Ecker et al. / Journal of Membrane
Table 2Experimental description; feed: 20 g/L LA and 19 g/L AA.
Experiment pH Diafiltration Comment
I 3.9 1 Reference runII 2.5 1 90% Undissociated LAIII 5.5 1 90% Dissociated LA
wwaswc
bm
2
fitmwawswd(
2
aURBAu
bha
I
TC
IV 4.11 1.5V 4.23 2
hich are discussed in this paper. The membranes were first runith original solution for 2 h and then cleaned (alkaline run and
cidic run). This membrane treatment was done to ensure repre-entative sheets. The pure water permeabilities of the membranesere determined before and after the complete pretreatment pro-
edure, the average results are presented in Table 1.There was no significant reduction in the water permeability
efore and after the cleaning. After the conditioning the experi-ents were done.
.2. Feed preparation
The feed solution for the nanofiltration screening was a puri-ed silage juice produced in the pilot plant. After production ofhe juice, it was ultrafiltrated to remove impurities such as macro-
olecules and slurry. The most important components of the feedere measured (see Section 2.3) to be 20 g/L LA and 19 g/L AA. These
nd further components of interest are listed in Table 3 togetherith other characteristics. All experiments were done with the
ame raw solution. The pH-shift of the feed solution was doneith hydrochloric acid and sodium hydroxide, respectively; theiafiltration medium for the diafiltration runs was purified waterRO-water).
.3. Chemical-analytical methods
Analytics were focused on the concentrations of organic acidsnd AA. The detection of organic acids was done by a HPCL (Dionexltimate 3000) with a pre column Micro Guard Cartridge (Cation-Hefill Cartridge). The main column was an Aminex HPX-87H fromio-Rad Co., using a 5 mM H2SO4 solution as mobile phase at 60 ◦C.n UV detection system for organic acids and RI detection systemsed for sugars were utilized.
The concentrations of AA were chromatographically determinedy a BioChrom 30 apparatus. The detection was done with aydrolysate program with lithium salt. Using this apparatus 18 AA
nd ammonium can be detected.
For analysis of inorganic ions an ion chromatography (DIONEXCS-300) with a conductivity detector was used. Cations were
able 3hemical and physical properties of the feed material.
determined with IonPac CS12A column using methanesulfonicacids as mobile phase and anions were detected by AS9HC columnusing sodium carbonate as mobile phase. Dry matter content of thejuices was determined by drying the sample in a drying balanceat 105 ◦C. For better results the measurement was repeated threetimes and the average value was calculated.
2.4. Calculations
Retention R of the components in each experimental run wasone of the evaluation values. This parameter depends on the feedand the permeate concentration cF and cP. The retention Ri of com-ponent i is defined as:
Ri =(
cFi− cPi
cFi
)× 100 (1)
As 18 AA were measured in the feed, the sum of concentrationsof all AA were used for the calculation of retention. In additionimportant results of specific AA are presented. Permeate flux wasmeasured in weight per time and calculated via density to volumeper time.
Transmembrane pressure TMP was calculated based on Eq. (2)
TMP = p1 + p2
2, (2)
where p1 is the pressure in the feed line for the test cell and p2 isthe pressure of the retentate in the exit of the test cell.
A further parameter for the separation performance is given bythe transport rate factor TRF. Within this factor the ratio of spe-cific transport rates of LA and AA in one experimental setup wascompared. The factor is given by
TRF = transport rateLA
transport rateAA. (3)
For the evaluation of diafiltration experiments the compari-son of transport rates, gained at different diafiltration rates, wasrequired. Therefore, a transport rate factor ˛D is applied. It isdefined as the relation between the transport rate of the specificsolute in the reference experiment to a diafiltration experimentalrun
˛Di= txi
tIi, (4)
where ˛Diis the result of the calculation for solute i at a diafiltration
rate D, tIi is the transport rate in the reference experiment (I) forspecies i, txi
is the transport rate of experiment x for i at same pres-sure level. This factor enables to compare the influence of differentfeed preparations. Diafiltration rate D refers the volume dilutionwith reference volume (exp. I) of the feed F.
VD = VF + Vw, D = VD
VF, (5)
where VD is the diluted volume, VF is the original feed volume andVw is the amount of solvent added to fix the diafiltration rate.
3. Results
To evaluate the membrane behaviour, first, the permeate fluxesthrough the membrane samples were measured for all experi-ments. Second, the specific transport rates of LA and AA (sum) weredetermined.
3.1. Pure water permeability
At the beginning pure water fluxes were determined for ref-erencing the filtration systems, results are shown in Table 1. In
392 J. Ecker et al. / Journal of Membrane Science 389 (2012) 389– 398
Table 4Reference experiment (I), retentions (%) of LA and AA; feed: 20 g/L LA and 19 g/L AA;T = 25 ◦C.
Component Membrane
DK DL HL HT FT KO
LA15 bar 56.1 52.7 57.5 53.9 51.3 50.0LA20 bar 65.4 61.9 64.4 68.5 64.7 57.9LA25 bar 71.5 67.4 69.4 67.3 66.1 65.0
AA 89.9 84.4 81.2 68.0 80.2 55.2
gpp1pFraeK
3
tiLNtAasRi
eFaflb5od
FD
0
10
20
30
40
50
60
70
80
2 4 6 8 10 12
spec
ific
trans
port
rate
, [g/
hm²]
total permeate flow, [L/hm²]
FT
Fig. 3. Reference experiment (I), comparison of membrane performances(Filmtech), LA (+), AA (×); feed: 20 g/L LA and 19 g/L AA, p = 20 bar, T = 25 ◦C.
Table 5Reference experiment (I), absolute specific transport rates [g/(m2 h)]; feed: 20 g/LLA and 19 g/L AA, p = 20 bar, T = 25 ◦C.
Component Membrane
DK DL HL HT FT KO
LA 35.8 49.6 29.7 10.6 51.9 26.8
1.8 L/(m h) permeate flux at 20 bar, was not satisfactory at all.The hydrophobic character of the membrane material explained
15 bar
AA20 bar 95.2 92.5 89.7 81.2 88.6 64.7AA25 bar 96.1 94.8 91.1 77.7 92.7 71.1
eneral water fluxes refer to the cut off of the membrane sam-les, but do not predict the flux behaviour with real solution. Theermeate flux measurements were done at a pressure range from5 to 25 bar, which presented the desired pressure range in theilot plant. Highest water fluxes were measured with membraneT (Filmtech), resulting from the high cut off of the membrane. Theesults gained from the Osmonics membranes were close together,ll three membranes presented similar cut off characteristics. Low-st pure water flux was observed for the hydrophobic membranesO and HT.
.2. Reference experiment
For reference, experiment I was done with original solution fromhe pilot plant, the main ingredients of the raw solution are listedn Table 3. It was shown that for all membranes the retentions ofA and AA at pH 3.9 were high and similar in general, see Table 4.evertheless, enhanced differences in the retentions were required
o produce two purified process streams, LA enriched permeate andA enriched retentate. The retention values were about 60% for LA,nd about 88% for AA in the reference experiment (I). The resultshow the complexity and difficulty of the separation requirement.esults concerning the retentions gained in experiment I are listed
n Table 4, flux results are shown in Fig. 2.Membrane FT gained highest permeate fluxes with the refer-
nce solution because of the high cut off of the membrane, seeig. 3. Higher pressure caused higher fluxes, the increase in thebsolute LA transport rate correlated positively with the risingux. However, the transport rate of the AA was not influenced
y flux increase. For membrane FT the transport rate of LA was1.9 g/(m2 h) and 17.9 g/(m2 h) for AA at 20 bar. The absolute lossf the AA to the permeate indicated lower separation performanceue to the high cut off. A comparison of the transport rates can be
done with the TRF value, Eq. (3). For membrane FT the value is 2.9,with indicated low differences in the separation performance.
A comparison of the transport rates of all membranes was doneat 20 bar, given in Table 5. The higher the differences in the specifictransport rates the better separation between the main compo-nents was observed. The ratio of the transport rates was used toevaluate the results.
For membrane HT low permeate fluxes were measured, alsoabsolute specific transport rates were least for both main com-ponents, see Fig. 4. The permeate performance of membrane HT,
2
reduced permeate fluxes, a high cut off (720 Da) resulted inincreased specific transport rates compared to the permeate flux.
0
5
10
15
20
25
30
35
1 2 3 4 5 6
spec
ific
trans
port
rate
, [g/
hm²]
total permeate flow, [L/hm²]
KO
HT
Fig. 4. Reference experiment (I), comparison of membrane performances (Koch andHydranautics), KOLA(+), KOAA (×), HTLA(∗), HTAA(�); feed: 20 g/L LA and 19 g/L AA,p = 20 bar, T = 25 ◦C.
brane Science 389 (2012) 389– 398 393
TlMsmaafbatsssbt
sbgAwfdidtfwopt
tamtartbdm
tei
3
AildvpIptvowsa
0
10
20
30
40
50
60
2 3 4 5 6 7 8 9 10
spec
ific
tran
spor
t rat
e, [
g/hm
²]
total permeate flow, [L/hm²]
DK
0
10
20
30
40
50
60
2 3 4 5 6 7 8 9 10
spec
ific
tran
spor
t rat
e, [
g/hm
²]
total permeate flow, [L/hm²]
DL
0
10
20
30
40
50
60
2 3 4 5 6 7 8 9 10
spec
ific
tran
spor
t rat
e, [
g/hm
²]
total permeate flow, [L/hm²]
HL
J. Ecker et al. / Journal of Mem
he transport rate factor was 1.6 for membrane HT. Therefore,ow selectivity between LA and AA was observed. The membrane
PS-36 from Koch interacted also hydrophobically with the feedolution. This was caused by the membrane material again. The per-eate performance was 3.5 L/(m2 h) at 20 bar. The transport rates
nd the corresponding retentions, gained with membrane KO, werelike for both components (26.8 g/(m2 h) for LA and 23.6 g/(m2 h)or AA at 20 bar). Hence, the separation performance of the mem-rane KO for experiment I was low, i.e. similar retentions of LAnd AA (57.9% for LA and 64.7% for AA). The transport rate fac-or was calculated to 1.1 for membrane KO, which indicated leastelectivity. Membrane HT and KO showed in general unsatisfactoryeparation performances, differences in the retentions were toomall. By trend, the separation performances were not influencedy increased permeate fluxes, see Fig. 4. Based on these results thesewo membranes were not used in further experiments.
The obtained results from the three Osmonics membraneshowed similar permeate fluxes, see Fig. 2. The different cut offecame noticeable at high pressure levels. Thus, membrane DLained highest specific transport rates for LA and AA, Table 5.lso highest permeate fluxes were observed with this membrane,hereas membrane HL showed the lowest permeate flux per-
ormance of these three membranes. However, about 40% of alletected AA in the permeate are alanine and glycine for DK and DL,
n contrast to only 30% for HL. This was a further proof of the moreense character of membrane HL. For all three membranes the LAransport rates increased with the permeate fluxes, see Fig. 5, butor the AA there was no increase observed. Thus, the transport of AAas not controlled by flux, however, by the chemical composition
f the feed solution and the membrane material. The separationerformance resulted from the interactions of the charged AA andhe membranes itself.
Increased differences in the retentions were required to producewo purified process streams, LA enriched permeate and aminocid enriched retentate. Best separation results were obtained fromembrane DL and DK for experiment I. Although higher specific
ransport rates with FT were observed in experiment I for LA, themino acid loss to the permeate was higher compared with DL. Theatio of the transport rates, TRF, was best for membrane DK, theransport of LA was 7 times higher than the AA transport, for mem-rane DL the factor was 4.8 and 2.9 for membrane FT. Further, theifferences in retention of the resources were high for these threeembranes, which indicated good separation performances.In reference experiment I, at pH 3.9, two mechanisms affected
he recovery of LA and AA. A combination of size effects andlectrostatic interactions between the solute and the membranenfluenced the specific transport rates, see also [11].
.3. pH-shift
Due to the pH dependency of the dissociation degree of LA andA a pH-change presents a good method for improving the selectiv-
ty. In case of a drop below the pI value (zero net charge) of LA at 3.8eads to the presence of more undissociated LA than lactate. Thus,ecreasing the pH value results in a reduced retention of LA. At a pHalues higher than the pI, more LA is dissociated. With increasingH value charge effects dominate the recovery of the organic acid.
n contrast the separation of AA (sum) could not be improved byH-change in general. Each amino acid has a specific pI value. Thus,he decision of the pH value variation was done according to the pIalue of LA. Experiment II was carried out with pH 2.5, where 90%
f LA were undissociated and experiment III was done at pH 5.5,here 90% of the organic acid was in its salt form lactate. The pH
hifts were done with raw solution as used in experiment I with original pH value of 3.9. About 20 g/L of hydrochloric acid were
Fig. 5. Reference experiment (I), comparison of membrane performances (GEOsmonics), LA (+), AA (×); feed: 20 g/L LA and 19 g/L AA, p = 20 bar, T = 25 ◦C.
used for the pH value reduction, exp. II, and about 35 g/L of sodiumhydroxide tablets were use for the pH value increase, exp. III.
In Table 6 the results gained at different pH values are presented.A reduction of the pH value, exp. II, caused significant increases inthe LA transport rates for all membranes. The rates were doubledfor membrane DK, and DL. For membranes HL and FT the transportrates increased up to three times. The AA losses to the permeatewere decreased in experiment II. For measurements done with highpH value, experiment III, the AA transport rates were similar to the
reference run, experiment I.
All tested membranes were negatively charged. Hence,uncharged or positive charged components could pass through themembranes more easily [2]. Although 90% of LA were dissociated
394 J. Ecker et al. / Journal of Membrane Science 389 (2012) 389– 398
Table 6pH variation, comparison of experiments I, II and III, absolute specific transport rates[g/(m2 h)]; feed: 20 g/L LA and 19 g/L AA, p = 20 bar, T = 25 ◦C; pHI 3.9, pHII 2.5, pHIII
o lactate at pH 5.5 the transport rates of LA were higher than inxp. I, done at pH 3.9. It was assumed that the bulk charge becameore dominant than the specific charge repulsion of lactate, the
o-called Donnan equilibrium between the bulk solution and thepecific membrane was achieved [11,12]. An explanation for thats the saturation of the membrane charged layer, which enablesigher transport rates for lactate. The increased presence of otheralts, e.g. sodium hydroxide, reduced the influence of lactate [13].herefore, higher transport rates of LA were observed in exp. IIIhan in exp. I. Results are shown in Table 6. This phenomenon wasbserved for membrane DL, HL and KO.
A reduction of the pH value (exp. II) determined better sep-ration of LA through the membranes, mostly resulted from thendissociated form of the acid. Due to pH decrease, done in exper-
ment II, the amount of transported LA to the permeate wasncreased drastically, see Table 6. With membrane DL and FT high-st transport rates of LA were observed. Consequently the permeateoncentrations of LA in experiment II were highest with DL and FT,6.9 g/L and 16.6 g/L, respectively. In contrast, the transport ratesor AA were decreased for all membranes with experiment II. AAetentions increased slightly with low pH values for DK and DL,hich results from the side charged components. As discussed pre-
iously in Table 6 no significant differences for AA in retentionetween experiments I and III were observed. This is also shown
n Fig. 6.The calculated TRF values resulted in best separation perfor-
ances in experiment II, see Table 7. Comparing the transportates of both main components, membrane DK and DL showedigh selectivity in the purification. The TRF value was highest forembrane DK at 37.3. This result was strongly influenced by the
ecreased loss of the AA to the permeate in experiment II. Gooderformance was also achieved with membrane DL, where the TRFalue was about 29. For this membrane the AA transport to theermeate was reduced by two thirds.
Due to the mentioned high cut off of membrane KO, there waso significant change in TRF value observed. For all experiments theeparation of the components was poor, referred by a low TRF valuef 1.1 and 1.6, respectively.
The membranes of Osmonics DK, DL and HL showed different
eparation behaviour at pH value variations. Although all threeembrane were negatively charged, the changes in pH value influ-
nced the separation differently, see Fig. 6. Permeate fluxes of
able 7H variation, comparison of TRF in different experiments; feed: 20 g/L LA and 19 g/LA, p = 20 bar, T = 25 ◦C, pHI 3.9, pHII 2.5, pHIII 5.5.
membrane DK and DL decreased with both pH variations to 70%of initial flux at experiment I. Adding acid or base caused a higheramount of ions in the raw solutions. Thus the osmotic pressurefor these solutions was raised, the driving force for the permeateflux was reduced. However, the transport rate of LA was increasing24% for membrane DL in experiment III compared to experiment I.This effect, combined with the reduced permeate fluxes, resultedin lower retentions for LA, from 62% (exp. I) to 43% (exp. III).
Divergent flux behaviour was observed with membrane HL,permeate fluxes increased for both pH change experiments. Themembrane showed a minimum permeate flux of 4.61 L/(h m2) atpH 3.9 (20 bar) with the reference solution. Flux increases of 14%
J. Ecker et al. / Journal of Membrane Science 389 (2012) 389– 398 395
0
20
40
60
80
100
2 4 6 8 10 12
Ret
entio
n, [%
]
total permeate flow, [L/hm²]
FT
Fig. 7. pH-dependency of LA and AA retentions, Filmtech membrane FT; exp. ILA (+),ea
itc6cmco
sFawtcttCpLttsiso3nwi
eapattvpwwAma
0
20
40
60
80
100
0 1 2 3 4 5
Ret
entio
n, [
%]
total permeate flow, [L/hm²]
HT
Fig. 8. pH-dependency of LA and AA retentions, Hydranautics membrane HT; exp.
lysine. The retention decreased with increasing the pH value. ForpH 2.5 the retention was about 98.4%, about 96% for exp. I andlowest with experiment III at 88%. The acidic amino acid aspartic
n experiment II and 25% in experiment III were observed. Further,he transport rate of LA was doubled, 29.7 g/(m2 h) to 45.9 g/(m2 h),omparing experiments I and III, the retention decreased from4% (exp. I) to 54% (exp. V). This resulted in higher permeateoncentrations of LA detected with membrane HL in experi-ent III (permeate: 8.5 g/L at 20 bar), although higher pH values
omparing to experiment I (permeate: 6.45 g/L at 20 bar) werebserved.
With membrane FT the pH variations influenced the flux andeparation performances in expected directions, shown in Fig. 7.or this membrane similar fluxes were observed for experiments Ind II. The slightly reduced flux in experiment II can be explainedith higher osmotic pressure due to acid addition. Nevertheless,
he high flux decline of 30% occurred in experiment III, was mostlyaused by membrane material shrinking [14]. However, comparinghe three experiments, see Table 6, the enhanced lactate concentra-ion at pH 5.5 determined lowest transport rates for experiment III.harge repulsion was the dominant mechanism. Highest LA trans-ort rates were obtained in experiment II with this membrane. TheA transport rate increased up to 145.7 g/(m2 h) in experiment II;his was an ascent of three times compared to experiment I. Never-heless, the amino acid loss in the permeate was slightly reduced,ee Table 6. Although deep decline in the permeate flux in exper-ment III, the concentration of AA in the permeate (2.11 g/L) wasimilar with experiment I. The transport rate of the AA was reducedf the same ratio as the permeate flux, which decreased in value of0% with pH 5.5. Although the pore size was shrinking, there waso specific separation based on size exclusion [15]. This combinedith the highest LA transport rate, membrane FT gained best results
n experiment II.Membranes HT and KO, which showed poor performances in
xperiment I, were also tested for pH-variation. Selectivity and sep-ration of LA increased for both membrane types with decreasedH, experiment II. Both membrane samples showed low retentionsccording to the AA for all three pH-tests, see Figs. 8 and 9. Withhe pH variation experiments there was no influences in the selec-ivity observed for the membrane KO. For all three tests the TRFalue was similar. Only the permeate fluxes were influenced by theH variations. Due to the hydrophobic character of membrane HT,hich was enhanced with increased pH, poor total permeate fluxesere achieved. This caused increased permeate concentrations of
A. Further the retention of AA decreased significantly in experi-ent III. Due to the fact, that LA and AA were in their charged forms
t pH 5.5, the selectivity between the components was drastically
reduced, see [16]. Therefore, same separation characteristics for LAand AA were observed with this membrane, see Fig. 8.
Membrane HT, which performed least in experiment I, showeddecreased fluxes for both pH variation experiments. Here theincreased osmotic pressure caused the permeate performance.
There were small effects on the retentions of the AA due to thepH value variations observed. The pI of most AA is in the rangeof 6, at this level the acids are changed to their uncharged stateand the effective loss through the membrane increases, see [4].Although experiment III was done with a pH near the pI of mostAA (e.g. tyrosine 5.66, asparagine 5.41) there was no significantchange in the retention (sum AA). For the sake of completeness,the retentions of an acidic, a neutral and a alkaline amino acidare discussed shortly. The retention of leucine, representative forneutral AA (no net charge) is not influenced by the pH changes. Bytrend the rejection of leucine was constant for all membranes. Forexample, the retention was 97% with membranes DK and about95% at 20 bar for membrane DL in experiments I, II and III. Therewas a different behaviour observed for the alkaline amino acid
396 J. Ecker et al. / Journal of Membrane Science 389 (2012) 389– 398
0
2
4
6
8
10
12
14
16
0.8 1 1.2 1.4 1.6 1.8 2 2.2
spec
ific
trans
port
rate
, [g/
hm²]
diafiltration rate D, [−]
FbA
awpe
3
aaoHbtsbtd
tataHAstTcia2
i
TDV
0
20
40
60
80
100
0.8 1 1.2 1.4 1.6 1.8 2 2.2
Ret
entio
n, [
%]
total permeate flow, [L/hm²]
enhanced transport of small AA, e.g. alanine, through the mem-branes, too. Although more of the small AA, like alanine and glycine,were transported through the membrane, the sum value of the AA
20
25
ig. 10. Diafiltration experiments (IV and V), flux performance of tested mem-ranes; IV: D = 1.5; V: D = 2, DK (+), DL (×), HL (∗), FT (�); feed: 20 g/L LA and 19 g/LA, p = 20 bar, T = 25 ◦C.
cid, behaved the other way around. Here the lowest retentionas observed with membrane HL at pH 2.5, exp. II. Due to the lowI value of this AA (1.99) there was no change in the retentionxpected. It was constant at 92% exps. I and III.
.4. Diafiltration
An ability to increase the purification and the separation in nanofiltration step was a diafiltration mode. Within this modedding solvent reduced the membrane loading, which can bebserved in higher permeate fluxes at constant pressure levels [17].igher fluxes resulted in higher transport rates through the mem-ranes, in this operation mode more LA can be transported throughhe membranes, [18]. This supported the purification of the feedtream. However, the corresponding AA loss to the permeate has toe considered. To observe the influence of the feed concentrationso the fluxes and the separation performance two experiments withifferent diafiltration rates were done, experiments IV and V.
The diafiltration experiments, exps. IV and V, showed similarrends in increased permeate fluxes, see Fig. 10. For all membranes
proportional increase in permeate fluxes with increasing diafil-ration rates occurred. The fluxes were doubled for membranes FTnd DL but increased three times with D = 2 for membrane DK andL. Resulting from the high fluxes the concentrations of LA andA in the permeates were low and thus the retentions increased,ee Eq. (1). Further, for both experiments rising transport rateso the permeate of the interesting components were observed.he increased transport rates compensated higher retentions. Thisan be observed for membrane FT, where the retention of AAncreased from 88.6% to 96.8% at a diafiltration rate D = 1.5, although
n enhanced transport rate from 17.5 g/(m2 h) to 20.5 g/(m2 h) at0 bar was observed, see Table 8.
Despite high retentions the absolute transport rates of LAncreased with increased D value. But the positive effect had a
able 8iafiltration experiments (IV and V), retentions (%) of LA and AA (sum); IV: D = 1.5;: D = 2, feed: 20 g/L LA and 19 g/L AA, p = 20 bar, T = 25 ◦C.
Fig. 11. Diafiltration, experiment (IV and V), LA transport rate related to diafiltrationrate D; IV: D = 1.5; V: D = 2, DK (+), DL (×), HL (∗), FT (�); feed: 20 g/L LA and 19 g/LAA, p = 20 bar, T = 25 ◦C.
maximum at a diafiltration rate of D = 1.5, higher rates deter-mined smaller absolute LA transport rates because of high solventfluxes. This behaviour was observed for all Osmonic membranes,however, membrane FT did not show this maximum. Thus, themaximum diafiltration rate for this membrane was not achieved.Higher solvent addition could further increase the LA transportrate for membrane FT, Fig. 11. However, the specific transport rateof AA (sum), showed a different trend for membrane FT. The AAloss was stable up to D = 1.5, by further increasing the diafiltrationrate, the loss of the AA to the permeate decreased significantly.It was assumed, that the reduced feed concentrations determinedreduced transport rates, Fig. 12.
All three membranes from Osmonics showed the same trendin LA transport rates. Highest transport rates were observed withmembrane DL. The maximum was achieved with 77.5 g/(m2 h) andD = 1.5. For membrane DK and HL, the same maxima in LA transportrates were observed, although the values were significantly smallercompared to membrane DL.
On the one hand high diafiltration rates reduced the feed con-centrations. On the other hand a high diafiltration rate facilitated
0
5
10
15
0.8 1 1.2 1.4 1.6 1.8 2 2.2
Ret
entio
n, [%
]
total permeate flow, [L/hm²]
Fig. 12. Diafiltration, experiments IV and V, AA transport rate related to diafiltrationrate D; DK (+), DL (×), HL (∗), FT (�); IV: D = 1.5; V: D = 2, feed: 20 g/L LA and 19 g/LAA, p = 20 bar, T = 25 ◦C.
J. Ecker et al. / Journal of Membrane
Table 9Comparison of diafiltration experiments (IV and V), ̨ value, IV: D = 1.5; V: D = 2;feed: 20 g/L LA and 19 g/L AA, p = 20 bar, T = 25 ◦C.
Experiment Membrane
DK DL HL FT
IVLA 1.66 1.56 2.15 0.96VLA 1.3 1.2 1.12 1.45
weofHfadtdfw
bfldrafIp(i
ottfcmmdtolbrp
tmtd
ebidpoolmwq
IVAA 0.56 0.66 1.21 1.17VAA 0.16 0.36 0.86 0.34
as declining with increased diafiltration rates (D = 2), Fig. 12. As anxample, the separation behaviour of the amino acid alanine wasbserved in detail. The transport rate of the AA (sum) decreasedrom 17.5 g/(m2 h), exp. I, to 5.9 g/(m2 h) with experiment V, D = 2.owever, the specific transport rate of alanine was 4.13 g/(m2 h)
or exp. I and 4.05 g/(m2 h) for exp. V done with membrane FT. Thebsolute specific transport rates of small AA were not affected byecreased feed concentrations. The relative amount of alanine inhe permeate increased from 17%, exp. I to 21% exp. V. Separationue to size exclusion could be excluded for this small amino acidor the high cut off membrane FT. Consequently, the elution effectas the dominant mechanism for the transport to the permeate.
By contrast, membrane DK showed different separationehaviour. Despite decreased feed loading and higher permeateuxes, the absolute specific transport rates of alanine and glycineecreased. The permeate flux was doubled with D = 1.5. The cor-esponding transport values for alanine were 1.02 g/(m2 h) (exp. I)nd 0.5 g/(m2 h) (exp. IV), for glycine the transport rate decreasedrom 1.38 g/(m2 h) in experiment I to 0.69 g/(m2 h) in experimentV. The rejections resulted from the steric resistance for the trans-ort of alanine to the permeate. The relative amount of small AAalanine and glycine) in the permeate was constant at 30%, whichndicated no elution effects with increasing solvent addition.
Concerning the AA, there was no maximum of transport ratesbserved for the Osmonics membranes. For membrane DL and DKhe transport rates decreased significantly with increasing diafiltra-ion rate, see Fig. 12. The decline was proportional to the reducedeed loading and the water addition rate. Consequently, the feedoncentrations had highest influence on the separation perfor-ance of the AA. But different behaviour was observed for the thirdembrane of Osmonics HL. It showed similar transport rates at all
iafiltration steps, 9.0 g/(m2 h) at D = 1 and 7.7 at D = 2, exp. V. Forhis membrane no elution effects with increasing water additionccurred. Due to the low LA transport rate and the constant AAoss to the permeate with varying diafiltration rates, this mem-rane showed unsatisfactory separation performance. However,educed losses of AA to the permeate were desired, and the besterformances were achieved with membrane DL and DK.
For better description of the influence on the diafiltration rateo the specific separation of LA and AA and the purification of these
ain components a new factor was defined to evaluate the diafil-ration results, see Eq. (4). Hence, a quantitative analysis can beone. Results are presented in Table 9.
High ̨ values implied increased transport rates compared toxperiment I. A good separation with the nanofiltration mem-ranes was achieved with high ˛LA and low ˛AA values. As shown
n Figs. 11 and 12 also the ̨ values presented best results withiafiltration rate D = 1.5. Due to the very small amino acid trans-ort in experiment V with membrane DK, also good ̨ values werebserved with this experiment. Best relative improvement wasbserved for membrane HL with D = 1.5 for LA, although the abso-
ute LA transport rate was lower than for membranes DK and DL. As
entioned before the AA loss to the permeate was also increasedith this membrane and experiment. Thus, membrane HL was dis-
ualified for usage in diafiltration modes.
Science 389 (2012) 389– 398 397
Results of experiment V, gained with membrane FT, showedthe potential of higher diafiltration rates. Absolute transport rateswere best for this membrane with D = 2. Although the feed loadingwas reduced in this experiment, the ̨ value was increased for LAVLA = 1.17 and decreased for AA, VAA = 0.34, using this membranein a diafiltration run, high solvent addition rates were necessary togain improved results. However, higher diafiltration rates shouldbe tested for this membrane.
4. Conclusions
It was the objective of this paper to optimize the separation ofLA and AA with a NF step by variation of the membrane materialand process parameters.
Results of experiment I, done with the original solution from thepilot plant, show minor differences in the retentions between LAand AA for the tested membranes. However, for the membranes DKand DL partial separation was observed. Also the results gained withmembrane FT constitute potential for the separation. The perfor-mances of the membranes HT and KO were poor in all experiments;additionally lowest total permeate fluxes were observed for mem-brane HT. Although decreasing the pH value increased the retentiondifference, the observed results of membrane HT and KO were notsatisfying for a further investigation of these membranes in the pilotplant. As shown in this paper the performances of all membraneswere highly dependent on the feed preparation (diafiltration) andalso the operation parameters (pH value).
For improving the process performance pH reduction seemedto have the highest potential according to process optimization.The retentions and the separation of LA were highly dependenton the pH. Decreasing the pH increased the retention differences,which allowed to increase the recovery of the LA drastically (dupli-cation of the absolute specific transport with membranes DL andDK). AA were not influenced by pH variation in this range. Secondly,the process variations with diafiltration experiment, done up toD = 1.5, enhanced the absolute specific transport rates. The higherpermeate fluxes increased the transport of LA into the permeate.Adding more water (D > 1.5) downgraded the separation processand increased the loss of AA to the permeate. Further process man-agement could include partial recycling of nanofiltration permeate.Thus the required fresh water amount would be reduced. An addi-tional benefit would be, that remaining AA in the permeate couldbe recycled into the NF loop. However, the purification, gained withdiafiltration, is connected with lower product concentrations in thepermeate stream, which would require a concentration step afterthe filtration, e.g. a reverse osmosis plant.
Influences of the bulk ingredients on the fluxes or the reten-tions were not reviewed because tests in the pilot plant presentconstant fingerprints of the surrounding. Doubtlessly, the impactof the background matrix is not negligible but these tests have pre-sented the dominant influence of the pH on the separation of LAand AA. Concerning the fluxes, especially the amounts of sulfate-and magnesium-ions have to be considered. Pilot plant tests hadshown that they are depending on the origin of the raw mate-rial. The higher these amounts the more scaling occurred at themembranes.
In conclusion it was not possible to produce two highly purifiedproduct streams. But the quality of the separation, done with NF,would be sufficient enough for specific further product treatmenttechnology. Subsequent state of the art technologies are availableto handle the remaining LA in the retentate or AA in the perme-
ate, respectively. Due to the integration of the nanofiltration in acomplex process chain, separation efficiencies may be lower thanfor a final clean up step. For the AA finishing an ion exchange pro-cess is used in the biorefinery application. Pilot plant experiments
3 brane
hDnptmessp
A
eaoeRrd
R
[
[
[
[
[
[
[
98 J. Ecker et al. / Journal of Mem
ave shown that the remaining LA does not interact with the resin.epending on the further usage of the LA enriched solution, theeed for subsequent purification of the LA may vary. Tests in theilot plant with different separation technologies can be done, e.g.wo stage – electrodialysis. Other process combinations may be
ore suitable depending on product quality requirements. How-ver, NF-with optimized operating parameters (pH, D, T, �p) – mayhow sufficient retention differences that an integration of NF in aeparation chain may be economically visible. E.g. multistage NFrocesses could be applied to improve the separation.
cknowledgements
The authors want to express theirs thanks to the Green Biorefin-ry Upper Austria F & E GmbH for providing the membrane sheetsnd the analytics. Further, we want to give our thanks to the Schoolf Engineering and Environmental Science Wels for providing thequipment. We also express our gratitude to the sponsors of theefinery, the Linz AG, the RAG, the Ferngas OOE, Land Oberoester-eich, the Energie AG and the research framework program Fabriker Zukunft.
eferences
[1] S. Novalin, T. Zweckmair, Renewable resources green biorefinery: separationof valuable substances from fluid fractions by means of membrane technology,Biofuels, Bioproducts and Biorefining 3 (1) (2009) 20–27, ISSN 1932-1031.
[2] M.I. Gonzalez, S. Alvarez, F.A. Riera, R. Alvarez, Lactic acid recovery fromwhey ultrafiltrate fermentation broths and artificial solutions by nanofiltration,Desalination 228 (1–3) (2008) 84–96, ISSN 0011-9164.
[3] P. Pal, J. Sikder, S. Roy, L. Giorno, Process intensification in lactic acid production:
a review of membrane based processes, Chemical Engineering and Processing:Process Intensification 48 (11–12) (2009) 1549–1559, ISSN 0255-2701.
[4] Z. Kovacs, M. Discacciati, W. Samhaber, Modeling of amino acid nanofiltra-tion by irreversible thermodynamics, Journal of Membrane Science 332 (1–2)(2009) 38–49, ISSN 0376-7388.
[[
Science 389 (2012) 389– 398
[5] W. Koschuh, V.H. Thang, S. Krasteva, S. Novalin, K.D. Kulbe, Flux and retentionbehaviour of nanofiltration and fine ultrafiltration membranes in filtrating juicefrom a green biorefinery: a membrane screening, Journal of Membrane Science261 (1–2) (2005) 121–128, ISSN 0376-7388.
[6] J. Timmer, M. Speelmans, H. van der Horst, Separation of amino acids by nanofil-tration and ultrafiltration membranes, Separation and Purification Technology14 (1–3) (1998) 133–144, ISSN 1383-5866.
[7] V.H. Thang, W. Koschuh, K.D. Kulbe, S. Novalin, Detailed investigation of an elec-trodialytic process during the separation of lactic acid from a complex mixture,Journal of Membrane Science 249 (1–2) (2005) 173–182, ISSN 0376-7388.
[8] Z. Kovacs, W. Samhaber, Nanofiltration of concentrated amino acid solutions,Desalination 240 (1–3) (2009) 78–88, ISSN 0011-9164.
[9] C. Umpuch, S. Galier, S. Kanchanatawee, H.R. de Balmann, Nanofiltration as apurification step in production process of organic acids: selectivity improve-ment by addition of an inorganic salt, Process Biochemistry 45 (11) (2010)1763–1768, ISSN 1359-5113.
10] V.H. Thang, S. Novalin, Green biorefinery: separation of lactic acid from grasssilage juice by chromatography using neutral polymeric resin, BioresourceTechnology 99 (10) (2008) 4368–4379, ISSN 0960-8524.
11] A. Bouchoux, H.R. de Balmann, F. Lutin, Investigation of nanofiltration as apurification step for lactic acid production processes based on conventionaland bipolar electrodialysis operations, Separation and Purification Technology52 (2) (2006) 266–273, ISSN 1383-5866.
12] F.G. Donnan, Theory of membrane equilibria and membrane potentials in thepresence of non-dialysing electrolytes. A contribution to physical–chemicalphysiology, Journal of Membrane Science 100 (1) (1995) 45–55, ISSN 0376-7388.
13] V. Freger, T.C. Arnot, J.A. Howell, Separation of concentrated organic/inorganicsalt mixtures by nanofiltration, Journal of Membrane Science 178 (1–2) (2000)185–193, ISSN 0376-7388.
14] M. Nilsson, G. Trägardh, K. Östergren, The influence of pH, salt and temperatureon nanofiltration performance, Journal of Membrane Science 312 (1–2) (2008)97–106, ISSN 0376-7388.
15] A.E. Childress, M. Elimelech, Effect of solution chemistry on the surface chargeof polymeric reverse osmosis and nanofiltration membranes, Journal of Mem-brane Science 119 (2) (1996) 253–268, ISSN 0376-7388.
16] S. Platt, M. Nyström, A. Bottino, G. Capannelli, Stability of NF membranes underextreme acidic conditions, Journal of Membrane Science 239 (1) (2004) 91–103,
ISSN 0376-7388.
17] T. Melin, R. Rautenbach, Membranverfahren, Springer-Verlag, 2007.18] A. Roman, J. Wang, J. Csanadi, C. Hodur, G. Vatai, Partial demineralization
and concentration of acid whey by nanofiltration combined with diafiltration,Desalination 241 (1–3) (2009) 288–295, ISSN 0011-9164.
