Separation of Sugars by Ceramic NanofiltrationMembranes

Catarina Ricardo Oliveiracatarina.r.oliveira@tecnico.ulisboa.pt

Instituto Superior Tecnico, Lisboa, Portugal

November 2018

Abstract

Sugar as sucrose, is one of the most consumed products nowadays and its extraction process isalready well studied. Molasses is a by-product of this process, which has a high content of sucrose,yet to be extracted. Recovery of sucrose from molasses is in development and it is necessary toinvestigate new methods suitable for such. It is in this context that the membrane separation processescan contribute to a procedural and economic development of this process. The aim of the project wasto evaluate the performance of a nanofiltration process using ceramic membranes to recover sucrosefrom sugar beet molasses. Two membranes with MWCO of 200 and 350 Da were tested under differentoperating conditions. Membrane LC1 was the most adequate to reach the project objective, presentinghigh permeate fluxes and high levels of sucrose retention. This membrane had a hydraulic permeabilityof 28.2 L/m2hbar. The permeate flux was around 250 L/m2h and the retention of sucrose was 80%,when operated at 70◦C, a CFV of 1 m/s and a TMP of 10 bar.Keywords: Sugar beet, Molasses, Sucrose, Nanofiltration, Ceramic Membranes.

1. Introduction

Sugar is an important form of carbohydrates that iswidely consumed. Due to its properties, sugar hasmany applications such as sweetening, preserva-tion and production of ethanol. The sugar sucrosecan be produced from sugar cane or sugar beet.In geographic regions with moderate climates, theproduction of sugar from sugar beet is preferreddue to the existence of more suitable climatic con-ditions for its cultivation. Nowadays, this processis well developed, but it is important to continuedeveloping new technologies that allow a good ef-ficiency and a lower energy consumption.

The sugar extracted from sugar beets is com-posed by different types of sugar molecules, mono-and disaccharides. So far, the bulk of sucrose isrecovered through crystallisation but the separa-tion of sugars from residues, such as molasses,has been accomplished through chromatography.However, that process presents a series of disad-vantages since it requires a large amount of time,eluent and energy.

In recent years, nanofiltration has been consid-ered as a promising alternative technology for theseparation of sugars. When compared with chro-matography, the nanofiltration process demandsno eluent, consumes less energy and is faster. An

important challenge for an efficient separation ofsugars by nanofiltration is the selection of suitablemembranes. There has been a great deal of com-mitment in the development and improvement ofceramic nanofiltration membranes, which are con-sidered to have great potential for the separationof sugars due to their high durability, resistance tohigh temperatures and bacterial resistance.

Therefore, this project will focus on testing ce-ramic nanofiltration membranes to perform theseparation of sucrose from fructose/glucose solu-tions which would enable purification of sucrosefrom molasses. In addition, an attempt will bemade to find suitable operating conditions.

1.1. Objectives

The project aimed for the chemical and physicalcharacterization of a sample of beet molasses, se-lection of a suitable membrane and operating con-ditions through parametric studies and analysis ofthe influence of temperature on the separation.

1.2. Nanofiltration for Recovery of SugarsIn the last years many studies have been per-formed in order to assess on the use of membraneprocesses to separation and recovery of differentsugars.

Nihal, et al., studied the effect of operating pa-

1

rameters on the separation of sugars by nanofil-tration. Fist, it was study the separation perfor-mance of sucrose and glucose as individual so-lutions and then the separation performance of abinary solution of sucrose and glucose. Two dif-ferent types of asymmetric polyamide membraneswith 500 MWCO (Berghof, BM5) and 200 MWCOwere used during nanofiltration experiments. Itwas observed that increasing the feed flow rate in-creased the permeate flux and rejection by mini-mizing the effect of concentration polarization. In-creasing the pressure did not affect rejection sig-nificantly. Only in the low range did an increase insolute concentration caused important decreasesin permeate flux and increases in rejection. Theresults indicated that a mechanism for the sepa-ration of organics by nanofiltration can be basedsolely on molecular size. Studies with binary solu-tions demonstrated that solute-solute interactionsaffected the performance of nanofiltration mem-branes. Rejection of a smaller solute (e.g., glu-cose) was improved by the presence of a large one(e.g., sucrose) [2].

Feng, et al., studied the separation of galacto-oligosaccharides mixture by nanofiltration. Fourspiral wound nanofiltration membranes were ap-plied to examine the separation performance ofcommercial galactose-oligosaccharide (GOS) mix-ture with the specification of 18.8 wt% monosac-charide (17.9 wt% glucose and 0.9 wt% galac-tose), 44.8 wt% lactose and 36.4 wt% galactose-oligosaccharides. The four tested commercialmembranes were 1812 spiral wound membranemodules. The membranes NF-2 (500-600 Da) andNF-3 (800-1000) were supplied by Sepro Co(USA),assembled by Shanghai Mosu Co.(China). Themanufacturer and module manufacturer of NF-1812-50 (150-300 Da)were Dow Chemical (USA)and Beijing Ande Co. (China), respectively. HBRO-1812-2 (800-1000 Da) was made and assembledby Hebei R.O.Environment Tech. Co. (China) Itwas concluded that high pressure can induce se-rious fouling due to the compaction of the mem-brane layer, which inevitably results in high en-ergy cost. Hence, operation pressure of lowerthan 8 bar was adopted to test the separation per-formance of the membranes with sugar solutions.The permeate fluxes and apparent rejection of sug-ars increased with increasing pressure for a givenmembrane. With increase of temperature, the per-meate fluxes increased and rejections decreasedwhen other parameters were fixed. On the otherhand, both permeate fluxes and rejections of sug-ars decreased with increasing feed concentration.The selection of optimum NF membrane and op-eration conditions for a separation process was acompromise between the permeate fluxes and re-

jections of sugar [1].Qi, et al., investigated the removal of furfural

and concentration of monosaccharides by us-ing two commercial nanofiltraton (NF) membraneswith synthetic glucose–xylose–furfural solution asmodel. Two commercially available NF membranesobtained from Dow FilmTech, NF90 (90 Da) andNF270 (150 Da), were employed in the presentstudy. The effects of main operating parameterssuch as feed, pH, permeation flux, temperatureand feed concentration on the rejections of thethree solutes, were studied. For both membranes,the rejections of the three solutes were significantlyinfluenced by the operating conditions, such asfeed pH and solutes concentration, permeation fluxand temperature [4].

The possibility of nanofiltration in a demandingseparation of a pentose sugar, xylose, from a hex-ose sugar, glucose, was studied by Sjoman, et al.Xylose is an intermediate product in xylitol produc-tion and glucose interferes in the process. Feed so-lutions were made of xylose and glucose in differ-ent mass ratios and total monosaccharide concen-trations. The experiments were made with a DSSLabStak M20 (Alfa Laval Copenhagen A/S, Den-mark) plate and frame filtration equipment. Themembrane stack contained three different mem-branes (listed from bottom to top): Desal-5 DK(150-300 Da), Desal-5 DL (150-300 Da)(GE Os-monics, USA) and NF270(150-200 Da)(Dow Liq-uid Separations, USA). The results indicated thatthe separation of xylose from glucose by nanofil-tration is possible to a limited extent. The observedmonosaccharide retentions depend highly on per-meate flux, and retentions increased to certain re-producible level as pressure and consequently fluxis increased. The observed xylose retentions werefrom 0 to 80% and the glucose retentions werefrom 10 to 90%. The effect of total monosaccharideconcentration on the observed retention is smallerthan the effect of flux. The ratio of xylose to glucosein the feed had an influence on permeate flux andon xylose retentions. Xylose retentions decreasedas the proportion of glucose increased in the feed.The higher the proportion of xylose in the feed thehigher was the total permeate flux [5].

2

2. Materials and Methods2.1. Molasses and Sugar Solutions

The sugar beet molasses were supplied by NordicSugar, Ortofta, Sweden. Different sugar solu-tions were prepared to be tested in the membranescreening step and their composition is presentedin Table 1.

Table 1: Composition in %wt of the solutions prepared duringthe studies.

Solution/Composition

Sucrose Glucose Fructose

Solution A 0.5% - -

Solution B - 0.5% -

Solution C - - 0.5%

Solution D 0.5%

According to the molasses supplier, the sugarcontent on them was around 0.5-10%wt. Thus, itwas decided that all the solutions should have aconcentration of 0.5%wt of sugars. The filtrationof Solution D aimed to reproduce the filtration ofa molasses solution. Considering the informationgiven in [3], sucrose represents 98% of the sugarcontent in the sugar beet molasses and 2% is glu-cose and fructose. Thus, Solution D has the samesugar composition.

2.2. Membranes and Module

During the project it was study the performance oftwo tubular ceramic nanofiltration membranes withdifferent MWCO: Membranes LC1 (200 Da) andLC2 (350 Da). All the experiments for both mem-branes were performed in the same tubular mod-ule.

2.3. Equipments

A schematic illustration of the experimental set-upis shown in Figure 1. The set-up consisted of a 15L feed tank where the experimental solutions wereheated by an electric heater. Inside the tank, twoPt-100 temperature sensors (Pentronic, Gunnebo,Sweden) were placed and connected to a temper-ature regulator, one to control the temperature inthe tank and the other to record the temperature.Transmembrane pressure was regulated by twomanual valves, one on the retentate side and theother on the permeate side. Three pressureme-ters (Trafag DCS40.0AR, Regal Components AB,Uppsala, Sweden) were place at the inlet, the per-meate outlet and retentate outlet. The perme-ate flow was measured with an electronic balance(PL6001-S, Mettler-Toledo Inc., Columbus, USA.).A displacement pump (Hydra-cell D25XL, Wan-ner, Minneapolis, MN, USA), connected to a fre-quency converter (ELEX 4000, Bergkvist & Co.,

AB, Gothenburg, Sweden), was used to controlthe cross-flow velocity. The pressure indicators,the temperature transmitters, the flowmeter, andthe balance were connected to a computer andall the data were recorded using the LabVIEW R©2009 software (National Instruments Co, Austin,TX, USA).

Figure 1: Schematic representation of the experimental set-up.

2.4. Operating Procedures2.4.1 Membrane Cleaning and PWF Measurement

Membrane LC1 was cleaned with an alkalinecleaning agent, P3-Ultrasil 110 (Ecolab AB, Alvsjo,Sweden), and an acidic cleaning agent, P3-Ultrasil73 (Ecolab AB, Alvsjo, Sweden). The membranewas first cleaned with the alkaline solution, fol-lowed by the acidic solution and finally again withthe alkaline solution, each for 1 h. For MembraneLC2, it was used a stronger acidic cleaning agent,P3-Ultrasil 75 (Ecolab AB, Alvsjo, Sweden). Bothmembranes were cleaned at 50◦C, TMP of 3 barand CFV of 3 m/s. The cleaning agent was usedwith a concentration of 0.5%wt and with a volumeof 13 L.

The PWF was first measured before carrying outany experiment with the new membranes (Jwi).After performing each experiment, the PWF ofthe fouled membrane (Jw∗) was again measured.Hereafter, the membrane was cleaned and thePWF of the cleaned membrane (Jw) was one lasttime measured. The measurement was carried outat 30◦C and with a CFV of 3 m/s, for a TMP rangefrom 1 to 9 bar with a step size of 2 bar. The PWFwas measured for 5 min at each pressure.

2.4.2 Membrane Screening

The membrane screening consisted of carrying outparametric studies with each membrane, leading tothe conclusion of the most promising membrane tothe desired separation. Primarily, both membraneswere tested with solutions A, B and C, in order toverify how the membrane behaves when it comes

3

to the separation of single sugars. Afterwards, So-lution D was tested, this solution aims to reproducethe sugar composition of a molasses solution.

Parametric StudiesThe parametric studies allow the assessment ofthe influence of the operating conditions on the per-meate flux and on the retention levels. The testswere performed at 70◦C, with a CFV of 1 m/s and3 m/s, at TMP of 2, 5 and 10 bar and both the re-tentate and the permeate were recycled to the feedin order to maintain a constant concentration.

The tests were always started on the highestCFV and the lowest TMP, the TMP was then in-creased to the highest value. After this, the newCFV was set and the TMP was decreased back tothe lowest value.

All the measurements were carried out for 15min at each operating conditions and samples ofthe permeate were collected for each pressure.The composition of the samples was subsequentlyanalysed and the retention evaluated.

2.4.3 Membrane Performance

The membrane screening step concluded thatMembrane LC1 is the most suitable to achieve de-sired separation. In order to assess on the effectof temperature on the flux and retention, the sameparametric studies were performed at 60◦C and80◦C. These tests were only performed for SolutionA. These last test, together with those performedin the screening phase at 70◦C, will lead to theconclusion about the best operating conditions toachieve the best possible results.

2.5. Analytical Methods2.5.1 Total Solids and Ash

To determine the TS content, samples were driedin an oven at 105◦C for 72 h and then cooled toroom temperature for 1 h in a desiccator. Then,the weight of the samples was measured. The drysamples were then moved to a furnace, where theywere heated to 105◦C for 30 min. After, they wereagain heated to 250◦C at a heating rate of 5◦C/minand this temperature was maintained for 30 min.Finally, the temperature was raised to 575◦C at aheating rate of 10◦C/min and it was held for 3 h.The ash content was calculated from the weight ofthe remaining samples after cooling to room tem-perature in a desiccator for 1 h.

2.5.2 Lactic Acid

The molasses samples were acidified to a pH of5 by using 95% sulphuric acid, then diluted 20times and filtered with a 0.2 µm filter (Scheicher

& Schuell, Germany). The samples were anal-ysed by high-performance liquid chromatography(HPLC) using a Shimadzu HPLC system (Shi-madzu Corp., Kyoto, Japan) which was equippedwith an RID-10A refractive index detector (Shi-madzu Corp., Kyoto, Japan).

2.5.3 Density, Conductivity and pH

The conductivity of the molasses was measuredusing an HI 99301 EC/TDS meter equipped withan HI 76306 probe (Hanna Instruments Inc.,Woonsocket, RI, USA.) The pH of the molasseswas determined using an InLab Expert pH (Mettler,Toledo.) The density of the molasses was deter-mined using an analytical hydrometer (Kebo Grave,Spanga, Sweden).

2.5.4 Sugar Content

The molasses samples were first filtered using a0.2 µm filter (Scheicher & Schuell, Germany) andthen diluted. The sugar content was determinedthrough high-performance anion exchange chro-matography coupled with pulsed amperometric de-tection (HPAE-PAD), using for that a ICS-3000chromatography system (Dionex Corp., USA). Thesystem was equipped with a Carbo Pac PA1 ana-lytical column, a ICS-3000 SP gradient pump andan AS-AP autosampler. The standards used wereD-sucrose, D-glucose, D-fructose and D-raffinose.

2.5.5 Molecular Mass Distribution

The molecular mass distribution of molasses wasdetermined by size exclusion chromatography(SEC) using a Waters 600E chromatography sys-tem (Waters, USA). The system was equippedwith a refractive index (RI) detector (model 2414,Waters) and UV detector (model 486, Waters).The analytical column was packed with 30 cm Su-perdex30 and 30cm Superdex200 (GE Healthcare,Sweden). The injection volume was 500 µm anda solution of 0.5 wt% NaOH was used as elu-ent at a flow-rate of 1mL/min. The system wascalibrated with polyethylene glycol standards withpeak molecular masses of 0.4, 4, 10 and 35 kDa(Merck Schuchardt OHG, Germany.) Before per-forming the test, all samples were filtered using a0.2 µm filter (Scheicher & Schuell, Germany).

4

3. Results and Discussion3.1. Molasses Characterization

The analysis methods performed were describedin the previous section and the results related withthe first three tests are exposed in Table 2.

Table 2: Molasses characteristics.

pH 7.14

Conductivity (mS/cm) 1.19

(g/cm3) 1.852

TS (%) 70.6 ± 6.1

Ash (%) 22.7 ± 0.5

Lactic Acid (g/L) 2.5 ± 0.1

Sugar ContentThe molasses sample has a sugar content of 470g sucrose/g molasses, <1 g glucose/g molasses,<1 g fructose/g molasses and 10 g raffinose/g mo-lasses.

According to [3], it would be expected that themolasses were 98% composed of sucrose, the re-mainder being mainly glucose and fructose andsome trace amounts of raffinose, yet the obtainedresults showed that in the tested sample, the raf-finose concentration was higher than the glucoseand fructose. The deviation of results to [3] is prob-ably related with two factors: (i) place of cultiva-tion and (ii) sugar extraction method. Molasses un-der study in [3] were produced in Germany, whilethe ones tested in this project were produced inSweden. Different cultivation sites suggest differ-ent soil properties, climate, and then beets withdifferent sugar content and sugar composition. Itis also necessary to consider that, being producedin different places, the method to extract the sugarsfrom the beets can also be different, which it will in-fluence the properties of its by-products, includingthe molasses.

As explained before, the composition in sugarsof Solution D was decide according with informa-tion provided in [3]. The aim was to have a solu-tion with the same sugar composition of sugar beetmolasses but, at that point, it was unknown thecomposition of the molasses that were provided byNordic Sugar. Considering the difference in sugarcomposition of the Solution D and the molasses,the results obtained when performing experimentswith Solution D will not be as informative as ex-pected.

Molecular Mass DistributionThe SEC diagram presented in Figure 2 shows themolecular mass distribution of the molasses sam-ple and a sugar standard.

Figure 2: Molecular mass distribution of molasses and sugarstandard (measured as refractive index).

The RI curve showed that both samples con-tained a high amount of the same material, around0.5 kDa. This component was certainly sucrose,which has a MW of 0.34 Da. The SEC results alsocorroborate the results obtained by the sugar con-tent tests (Table 2), where it was shown that su-crose is the main component in the molasses sam-ple.

Figure 3 shows the UV response of the molasseswhere several peaks of high molecular mass (100-10000 kDa) can be seen, consisting of colloidalparticles.

Figure 3: Molecular mass distribution of molasses (measuredas UV absorbance).

3.2. PWF and Hydraulic Permeability

The influence of TMP on the PWF of each mem-brane is shown in Figure 4.

Figure 4: Influence of TMP on the pure water flux, Jw, for eachmembrane. T=30◦C and CFV=3 m/s.

5

Hydraulic permeability, Lp, represents the perme-ation capacity of a membrane and it is the slop ofEquation (1) which describes the linear variation ofthe pure water flux (Jwi) as a function of TMP.

Jwi = Lp · TMP (1)

Membranes LC1 and LC2 have an hydraulic per-meability of 28.2 L/m2hbar and 20.5 L/m2hbar.

3.3. Membrane Screening3.3.1 Permeate Fluxes

The permeate fluxes were measured during theparametric studies. For a better follow-up of theresults and their discussion, this section is dividedinto three parts: (i) Membrane LC1, (ii) MembraneLC2 and (iii) Membrane LC1 vs Membrane LC2.

Membrane LC1The effect of TMP and CFV on the permeate fluxesof each tested solution, when using MembraneLC1, is presented in Figure 5.

Figure 5: Influence of the TMP and CFV on the permeate fluxesof tested solutions using the Membrane LC1. T=70◦C.

Figure 5 shows that the permeate fluxes of allsolutions increased with increasing TMP. The per-meate fluxes of Solution A were considerably lowerthan those of the other solutions. Solution A iscomposed only of sucrose which has a higher MW(342.3 Da) than glucose and fructose (180.2 Da)and this value is close to the membrane MWCO(200 Da), which may indicate that the membranewas retaining the sucrose as desired.

For Solution B, the results show that CFV did notinfluenced glucose permeation to lower TMP val-ues but may had a heavier impact on fluxes whenperforming the experiments at a TMP above to 10bar.

The permeate fluxes of Solution C were influ-enced by this CFV. Furthermore, it is noticeablethat Solutions B and C had similar results whichit is related to their close molecular weight.

The permeate fluxes of Solution D were lowerthan for Solutions B and C and higher than for So-lution A, when operating at a CFV of 1 m/s. At-tending to the composition of Solution D (Table X),

it was expected that the membrane had been re-taining sucrose, but not as much as in Solution Awhich is composed only of sucrose. For a CFV of3 m/s, Solution D had the highest permeate fluxeswhich was not in line with the previous results. Thecurrent results might had been caused by the highCFV, that forces the sucrose molecules trough themembrane pores and hinder their retention.

Membrane LC2The effect of TMP and CFV on the permeate fluxesof each tested solution, when using MembraneLC2, is presented in Figure 6.

Figure 6: Influence of the TMP and CFV on the permeate fluxesof tested solutions using the Membrane LC2. T=70◦C.

Figure 6 shows that the TMP enhanced the per-meate fluxes of all solutions. The permeate fluxesof Solution A were not influenced by CFV. Besidesthis, the permeate fluxes of Solution A were lowerthan for Solutions B and C, which denotes that themembrane was probably retaining sucrose, eventhough the Membrane LC2 has a higher MWCO(350 Da) than the MW (342.3 Da) of sucrosemolecules.

Regarding Solution B, the permeate fluxes in-creased with CFV and were higher than those ofSolution A, which means that the membrane waspermeating glucose molecules.

For Solution C, the permeate fluxes were higherat 1 m/s, as opposed to what was expected. As ex-plained in Section 2.4.2 , the parametric tests werestarted at the highest CFV and then decreased tothe lowest CFV, so it is possible that the membranehas been fouled during the first part thereby af-fecting the fructose permeation by the membranewhen the lower CFV was tested.

Finally, the permeate fluxes of Solution D in-creased with CFV and were lower than for the othersolutions, suggesting that the membrane may havebeen retaining some sucrose molecules as de-sired.

6

Membrane LC1 vs Membrane LC2Comparing the results showed above, it is possibleto conclude that Membrane LC1 provided higherpermeate fluxes for both CFVs. As previouslyshown Membrane LC1 has a higher hydraulic per-meability than Membrane LC2, and thus it will givehigher permeate fluxes.

Overall, it is expected that Membrane LC1 wasretaining more sucrose molecules than MembraneLC2 but these can only be confirmed by theanalysing the sugar content of each permeatesample taken during the parametric studies. Theretention results are presented in the next section.

3.3.2 Retention Levels

Samples were collected during the parametricstudies for further analysis of their sugar contentand determination of the retention levels of thecomponents of each solution. To a better follow-up of the results and their discussion, this sectionis divided into two parts: (i) Membranes LC1 andLC2 and (ii) Solution A vs Solution D.

Membranes LC1 and LC2In Figures 7 and 8, is shown the retention of sugarwhen filtering solutions A, B and C by MembranesLC1 and LC2 at CFVs of 1 and 3 m/s, respectively.

Figure 7: Influence of TMP and CFV on retention of compo-nents of solutions A, B and C using Membrane LC1 during theparametric studies.

Regarding Membrane LC1, Figure 7 shows anincrease in retention with increasing TMP for bothSolutions A and B, while retention of Solution Cwas approximately zero. This may be caused bythe concentration polarization phenomenon thatexplains the increase in solute concentration at themembrane surface and therefore a higher resis-tance to permeation of molecules of Solutions Aand B. In addition, the retention of Solution A wasalways greater than the retention of Solutions Band C, which was probably due to the fact that theMW of the sucrose molecules is higher than the

MWCO of the membrane. When operating at aCFV of 3 m/s, it was seen the opposite effect onthe retention of Solution A, while the results for theother solutions remained similar. The decrease ofthe retention of Solution A was due to the increaseof the CFV during the tests, forcing the passage ofthe molecules through the pores of the membrane.

Figure 8: Influence of TMP and CFV on retention of compo-nents of solutions A, B and C using Membrane LC2 during theparametric studies.

Regarding Membrane LC2, Figure 8 shows thatthe retention of all solutions was very low and it wasnot influenced by TMP. The current results wereexpected for Solutions B and C, due to their lowMW compared to MWCO of the membrane but notfor Solution A that presents a MW quite similar tothe membrane’s MWCO. Thus, a higher retentionof Solution A would be expected. When operat-ing at a CFV of 3 m/s, the retention of all solutionswas higher. Solution A had a decrease in reten-tion levels with increasing TMP while Solutions Band C had not. For Solution C at 10 bar, the re-tention level was higher than for Solution A, whichis highly unlikely to occur and it has no theoreticalexplanation. However, this particular result may bedue to a high fouling of the membrane resulting inblockage of the pores and consequent retention offructose molecules.

Finally, it can be concluded that Membrane LC1provided better results: high sucrose retention (So-lution A) and low retention of glucose and fructose(Solutions B and C). These results indicate that theMembrane LC1 is most appropriate to achieve theretention of sucrose through membrane filtration.

7

Solution A vs Solution DFigures 9 and 10 present a comparison betweenthe sucrose retention in Solution A and Solution Dfor Membranes LC1 and LC2, respectively. Thegoal is to understand how the membrane behaveswhen filtrating sucrose molecules in a single sugarsolution (Solution A) and in a mixture sugar solu-tion (Solution D), i.e., how the presence of othersugars in solution, such as glucose and fructose,influences the permeation of sucrose.

Figure 9: Comparison of the influence of TMP and CVF onretention of sucrose in Solution A and Solution D. MembraneLC1.

Figure 10: Comparison of the influence of TMP and CVF onretention of sucrose in Solution A and Solution D. MembraneLC2.

Figure 9 shows that at a CFV of 1 m/s, the reten-tion of sucrose during filtration of Solution A wasgreater than for Solution D and both increased withTMP. The difference between these retentions in-dicate that the presence of glucose and fructosein Solution D was influencing sucrose retention.Once again, the increase of the retention levelswith TMP may be due to the concentration polar-ization phenomenon, as already explained. Fora CFV of 3 m/s, the retention levels of each so-lution show opposite results: the retention of su-

crose during filtration of Solution A decreased withincreasing TMP while in Solution D increased.

For Membrane LC2, Figure 10 shows that forboth CFVs tested there was a reduction of sucroseretention with increasing TMP, when filtering Solu-tions A and D. Sucrose retention in both solutionswas quite similar, suggesting that Membrane LC2may had not been so affected by concentration po-larization phenomena when filtering a solution withvarious sugars as Solution D.

Overall, the results presented above allow toconclude that Membrane LC1 is the most suitablemembrane, as it not only retains the sucrose in asingle sugar solution, but as well as in a solutionwith different types of sugars. Furthermore, the re-tention results also support the findings obtainedfrom the analysis of the permeate fluxes.

3.4. Membrane Performance

The results provided in the membrane screeningphase concluded that the Membrane LC1 is theone providing the best fluxes and, more impor-tantly, the greater retention of sucrose molecules.Hence, the permeation of Solution A was testedunder 60◦C and 80◦C to assess the influence oftemperature on permeate flux and retention levelsgiven by Membrane LC1.

Permeate FluxesIn Figure 11 is shown the permeate fluxes when

performing the parametric studies at 60◦C, 70◦Cand 80◦C.

Figure 11: Influence of the temperature on the permeate fluxesof Solution A.

Figure 11 shows that, regardless of the temper-ature tested, the permeate fluxes increased withTMP and CFV. The highest permeate fluxes wereobtained at 80◦C which is explained by the reducedviscosity at higher temperatures. As previouslymentioned, sucrose is a disaccharide that whensubjected to high temperatures will degrade intomonosaccharides. These molecules have a lowerMW than sucrose molecules and much lower thanthe MWCO of the membrane, therefore they will be

8

more easily transported through the pores, therebyincreasing the permeate flux. This hypothesis willbe confirmed by the retention levels presented be-low.

Retention LevelsFigure 12 shows the influence of temperature on

the retention levels when filtering Solution A.

Figure 12: Influence of the temperature on the retention levelsduring filtration of Solution A.

Experiments carried out at 80◦C showed thelowest retention values, which is in agreement withthe hypothesis that sucrose molecules in SolutionA started to degrade into smaller ones. The reten-tion levels at 60◦C were barely influenced by TMPand CFV and were lower than those at 70◦C. Thehighest retention of sucrose molecules, around80%, was obtained when performing the experi-ments at 70◦C, a CFV of 1 m/s and a TMP of 10bar.

3.5. Membrane Cleaning

The measurement of the PWF at different stagesof the experiments allowed to determine the hy-draulic permeability before and after the membranewas cleaned, Lpw∗ and Lp. These current param-eters allowed to asses on the degree of fouling ofthe membrane before and after the cleaning, Fb(%)and Fa(%), presented in Table 3.

%Fb =

(1− Lpw∗

Lp

)· 100 (2)

%Fa =

(1− Lpw

Lp

)· 100 (3)

Table 3: Degree of fouling before and after the cleaning, Fb(%)and Fa(%).

Membrane LC1 Membrane LC2Fb Fa Fb Fa

Solution A 63% 44% 77% 61%Solution B 33% 22% 51% 48%Solution C 26% 0% 66% 59%Solution D 42% 0% 65% 5%

First, it is important to refer that all the hydraulicpermeability values were corrected to 25◦C. Theresults shown in Table 3 indicate that both Mem-branes LC1 and LC2 were fouled after the filtra-tion of each solution. The cleaning procedure de-creased the degree of fouling in both membranes,but in different proportions and with different effi-ciency depending on the solution that was filtered.The effect of the cleaning procedure was higher forMembrane LC1, since the difference between Fb

and Fa was high. The lower values of Fa in Mem-brane LC1 indicate that the cleaning method andcleaning agent are suitable to be use in this mem-brane.Regarding Membrane LC2, the Fb values werevery high, indicating that the membrane was proneto fouling. After cleaning Membrane LC2, the de-gree of fouling did not decrease as must as itshould, as the Fa values were in most cases higherthan 50%. The low decrease of degree of foulingshows that the cleaning was not effective whichmay be related to the cleaning agent itself or theprocedure. Thus, a new cleaning procedure andcleaning agent should be considered.

4. Conclusions

The current project consisted in the study of ananofiltration membrane process to recover su-crose from sugar beet molasses. It was se-lected two tubular ceramic membranes with differ-ent MWCO, Membranes LC1 (200 Da) and LC2(350 Da). Four solutions with different sugar com-position were tested. Parametric studies were per-formed in order to assess on the most suitablemembrane and operating conditions to achieve thedesired separation. Furthermore, a detailed char-acterization of a sugar beet molasses sample pro-vided by Nordic Sugar was executed.

The molasses sample has a sugar content of470 g sucrose/g molasses, <1 g glucose/g mo-lasses, <1 g fructose/g molasses and 10 g raffi-nose/g molasses.

The pure water flux of each membrane was mea-sured allowing the determination of the hydraulicpermeability. Membranes LC1 and LC2 have anhydraulic permeability of 28.2 and 20.5 L/hm2bar.

From the parametric studies were obtained per-meate fluxes and retention results. Regarding

9

Membrane LC1, all solutions show an increase oftheir permeate fluxes with TMP and CFV. SolutionA shows the lowest permeate fluxes possibility dueto the high retention of sucrose. Solutions B and Cshow similar permeate fluxes and the highest ones.Solution D has intermediate permeate fluxes dueto its composition. The retention of sucrose whenfiltering Solution A goes up to 80%, and the re-tention of glucose and fructose when filtering Solu-tions B and C is low, indicating that Membrane LC1was permeating these molecules. Regarding thefiltration of Solution D, the sucrose retention wasnot as high as for Solution A but still around 50%.About Membrane LC2, the permeate fluxes of So-lution A are not influenced by CFV and are higherthan Solution D fluxes, indicating that retention ofsucrose is probably higher in Solution D. SolutionsB and C show the highest permeate fluxes due tohigh permeation of their components, glucose andfructose. The filtration of each solution by Mem-brane LC2 provides low retention levels, not higherthan 50%.

Comparing both membranes, Membrane LC1provides higher permeate fluxes and higher reten-tion of sucrose when filtering Solutions A and Dthan Membrane LC2. These specific results makeMembrane LC1 most suitable to recover sucrosefrom sugar beet molasses.

After selecting Membrane LC1 as the most suit-able, its performance was tested under tempera-tures of 60◦C and 80◦C, to compare with the re-sults already obtained at 70◦C. At 80◦C permeatefluxes were the highest but retention levels werethe lowest. At 60◦C, both permeate fluxes andretention levels were not influenced by TMP andCFV. Permeate fluxes at 70◦C were the lowest butit was obtained the best retention levels confirmingthe results of the screening phase. Furthermore,the highest retention of sucrose was around 80%when operated at 70◦C, CFV of 1 m/s and TMP of10 bar.

Different cleaning methods and cleaning agentswere selected for each membrane. The clean-ing of Membrane LC1 was more effective, havingthe degree of fouling decreased down to 0% insome experiments. Regarding Membrane LC2, anew cleaning method/agent should be consideredsince degree of fouling after cleaning exceed 50%in most cases.

References[1] Y. Feng, X. Chang, W. Wang, and R. Ma. Sep-

aration of galacto-oligosaccharides mixture bynanofiltration. Journal of the Taiwan Institute ofChemical Engineers, 40(3):326 – 332, 2009.

[2] A. Nihal, G. Turker, and Y. Levent. Effect ofoperating parameters on the separation of sug-

ars by nanofiltration. Separation Science andTechnology, 33(12):1767–1785, 1998.

[3] Organization for Economic Co-operation andDevelopment. Consensus Document on Com-positional Considerations for New Varieties osSugarbeet: Key Food and Feed Nutrients andAnti-Nutrients.

[4] B. Qi, J. Luo, X. Chen, X. Hang, and Y. Wan.Separation of furfural from monosaccharidesby nanofiltration. Bioresource Technology,102(14):7111 – 7118, 2011.

[5] E. Sjoman, M. Manttari, M. Nystrom,H. Koivikko, and H. Heikkila. Separationof xylose from glucose by nanofiltration fromconcentrated monosaccharide solutions. Jour-nal of Membrane Science, 292(1):106 – 115,2007.

10