Accepted Manuscript

Title: Clarification of crude extract of yerba mate (Ilex

paraguariensis) by membrane processes: analysis of fouling
and loss of bioactive compounds

Author: Isabel B.B. Gerke Fabiane Hamerski Agnes P. Scheer

Vitor R. Silva

PII: S0960-3085(16)30183-3
DOI: http://dx.doi.org/doi:10.1016/j.fbp.2016.12.008
Reference: FBP 815

To appear in: Food and Bioproducts Processing

Received date: 11-2-2016

Revised date: 23-11-2016
Accepted date: 14-12-2016

Please cite this article as: Gerke, Isabel B.B., Hamerski, Fabiane, Scheer, Agnes P.,
Silva, Vitor R., Clarification of crude extract of yerba mate (Ilex paraguariensis) by
membrane processes: analysis of fouling and loss of bioactive compounds.Food and
Bioproducts Processing http://dx.doi.org/10.1016/j.fbp.2016.12.008

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
 1

1 Clarification of crude extract of yerba mate (Ilex paraguariensis) by membrane processes: analysis

2 of fouling and loss of bioactive compounds

4 Isabel B. B. Gerke*a, Fabiane Hamerskib, Agnes P. Scheerb, Vitor R. Silvab

5
a
6 Graduate Program in Food Engineering, Federal University of Paraná, P.O. Box 19011, Curitiba,

7 PR, Brazil
b
8 Department of Chemical Engineering, Federal University of Paraná, P.O. Box 19011, Curitiba, PR,

9 Brazil

10
*
11 Corresponding author: Tel +55 (41) 3361-3584

12 Email address: isabelbubans@yahoo.com.br (Isabel B. B. Gerke)

13 Graphical Abstract

14

15
16
17
 2

1 HIGHLIGHTS
2
3 Yerba mate extract, a non conventional feed stream, was clarified by MF and UF.
4 Effect of different membrane materials on the loss of bioactive compounds.
5 The pore blocking model was used to identify the major fouling mechanism.
6 Resistances in serie model was used to analyze the permeate flux resistances.
7 The lowest bioactive compounds rejection was for MF with PEI.
8

10

11 ABSTRACT

12 The clarification of a crude yerba mate extract using membrane processes was evaluated as a

13 treatment to reduce the turbidity of the aqueous extract prior to its use in food products. The variations

14 in the permeate flux and the concentrations of some phenolic and methylxanthine compounds were

15 investigated during crossflow filtration with -alumina (MF-alumina) and polyetherimide (MF-PEI)

16 microfiltration membranes and a polyethersulfone ultrafiltration membrane (UF-PES). All

17 membrane systems reduced the initial turbidity (207 ± 25 NTU) to zero. The loss of bioactive

18 compounds was lowest for the MF-PEI membrane (11.0-15.5 %). The highest permeate flux (36.33

19 ± 0.71) was observed for the MF-PEI at 0.02 MPa. Also, in the case of the MF-PEI, the permeate

20 flux of the clarified yerba mate extract showed an initial sharp decrease followed by a gradual decline

21 and subsequent stabilization, resulting in the final permeate flux being approximately 65.0 % lower

22 than the initial permeate flux. Almost 66 % of the total resistance to permeation was attributed to the

23 polarized layer. The predominant fouling mechanism is due the cake layer, indicative of superficial

24 fouling.

25

26 Keywords: yerba mate; clarification; phenolic compounds; methylxanthine compounds;

27 ultrafiltration; microfiltration.

28

29
30
 3

1 1. Introduction
2

3 Yerba mate (Ilex paraguariensis) is a plant native to South America which grows in Brazil,

4 Argentina and Paraguay and is used in the preparation of mate (hot drink), “tererê” (iced tea) and a

5 toasted mate tea infusion (Bracesco et al., 2011; Burris et al., 2012). Hot or cold infusions are prepared

6 mainly from the leaves of mate, as well as the smaller branches (< 10 mm diameter), and they are

7 traditionally known as tonic stimulants and digestive drinks. The per capita consumption of mate in

8 Brazil is estimated to be 1.2 kg per year, while in Argentina and Uruguay an average of almost 7 kg of

9 dried mate per year is used per person for the preparation of teas. More recently, mate-based products

10 have no longer been limited to their producing countries, with external markets emerging, such as the

11 USA and Europe (Heck and De Mejia, 2007; Pagliosa et al., 2010).

12 The growing popularity of yerba mate around the world is ascribed to the variety of bioactive

13 compounds present, such as methylxanthines (notably caffeine and theobromine) and phenolic

14 compounds (in particular caffeic acid and chlorogenic acid), all known for their bioactivity (Dutra et al.,

15 2010; Silveira et al., 2016). These compounds exhibit pharmacological activity, such as an antioxidant

16 effect and protection against induced DNA damage and osteoporosis (Anesini et al., 2012; Conforti et

17 al., 2012). Other functional properties, including diuretic, antifungal, and antiobesity effects, have also

18 been reported, which are associated with the bioactive compounds present in yerba mate (Bracesco et

19 al., 2011; Filip et al., 2010; Kim et al., 2015; Pang et al., 2008).

20 Recent reports describe the development of new food products with the addition of yerba mate

21 extracts, such as jellies, soft drinks and teas (Berté et al., 2011; Godoy et al., 2013). However, prior to

22 the use of yerba mate in the preparation of food products a step for the clarification of the crude aqueous

23 extract needs to be carried out to reduce the suspense solids content and the turbidity. Conventional

24 clarification techniques include filtration with diatomaceous earth, and the addition of an agglomeration

25 agent has been applied to crude fruit juice. However, clarification and stabilization processes are not

26 only complex but also expensive (Espamer et al., 2006).

27 Alternatives to the traditional methods include the porous membrane processes of microfiltration

28 (MF) and ultrafiltration (UF), which are associated with lower energy requirements and capital costs.
 4

1 Porous membrane processes can be used for industrial-scale non-thermal separation, without altering

2 the physical state of the treated solution and allowing minimal losses of proteins, vitamins, sugars and

3 salts (Habert et al., 2006). MF and UF have been applied successfully in juice clarification, for the

4 treatment of highly thermosensitive juices, resulting in a microbiologically stabilized clarified product,

5 while preserving the original flavor and aroma (Cassano et al., 2007).

6 The major problem during porous membrane clarification processes is the continuous permeate

7 flux decay, due to the polarized layer and fouling resistances (Silva et al., 2012; Habert et al., 2006).

8 The polarized layer is the concentration boundary layer adjacent to the membrane surface, formed at the

9 beginning of the process, resulting in a sharp decrease in the flux due to retrodiffusion from the

10 membrane surface to the feed solution, which is stabilized by the surface renewal effect promoted by

11 the cross-flow flux (Koltuniewicz et al., 1995; Neggaz et al., 2007). The physical and chemical

12 interactions (solute adsorption, pore seal, gelling/precipitation) between the solute rejected by the

13 membrane and the surface and internal pores of the membrane result in fouling. Fouling limits the flux

14 by continuous blocking of the membrane pores, thus gradually reducing the permeate flux (Barros et al.,

15 2003; Gimenes et al., 2014).

16 Juices and crude leaf extracts contain cell wall polysaccharides (such as lignin, cellulose and

17 pectin) which interact with the membrane surface or the inside of the pores during the clarification

18 process, inducing fouling (Cassano et al., 2007; Razi et al., 2012). In the case of juice production fouling

19 can lead to a considerable wastage of raw material and energy, as well as a reduction in the production

20 rate and juice quality (Barros et al., 2003; Razi et al., 2012). The identification of the phenomena

21 associated with the operating conditions allows an assessment of the technical viability of the process.

22 The research described in this paper was undertaken to study the application of the porous

23 membrane process to the clarification of crude yerba mate extract. The influence of different membrane

24 materials and transmembrane pressures (TMPs) was evaluated in terms of the permeate flux and the loss

25 of bioactive compounds. The membrane resistance and the identification of the predominant fouling

26 mechanisms during the clarification process were also investigated.

27
28
 5

1 2. Materials and methods

2

3 2.1. Materials

4 Yerba mate was obtained from Vier ltda. (Palmeirinha, Paraná State, Brazil). Deionized water

5 was used as the solvent for the extraction process. The chemicals 5-O-caffeoylquinic acid (5-CQA),

6 caffeine (Ca), theobromine (Th) and rutin (Ru) were purchased from Sigma–Aldrich Chemical Co. (St.

7 Louis, MO, USA). Methanol (chromatography grade) was purchased from Vetec Ltda. (São Paulo, São

8 Paulo State, Brazil). All other chemicals used were of analytical grade.

10 2.2. Yerba mate crude extract

11 The crude aqueous extract of yerba mate was obtained by batch liquid-solid extraction, adapted

12 from Jensen and Zanoelo (2012). A weighed sample (140 g) of yerba mate leaves was placed in a beaker

13 with 2 L of deionized water and heated to 80 °C. The liquid-solid extraction system was maintained at

14 80 °C in a thermostatic batch (Quimis, MODEL) under mechanical stirring (Quimis, model Q25M) at

15 200 rpm for 30 min. The leaves and larger solid particles were removed using a 200 mesh stainless steel

16 sieve (pore size 0.075 mm; Chemist Ltda., São Paulo, São Paulo State, Brazil). The solution recovered

17 after the filtration was made up to 2 L with the addition of deionized water.

18

19 2.3. Analytical procedure

20 The turbidity of the yerba mate solutions was determined using a portable turbidimeter

21 (PoliControl, model AP 2000) and the results are reported in nephelometric turbidity units (NTU).

22 The concentrations of the most representative bioactive compounds present in yerba mate were

23 determined in this study: rutin (Ru), caffeine (Ca), theobromine (Th) and chlorogenic acid (5-O-

24 caffeoylquinic acid, 5-CQA). These bioactive compounds were determined by high-pressure liquid

25 chromatography (HPLC) in according with the method described by Dutra et al. (2010) with

26 modifications.

27 Analyses were performed with the use of an Agilent 1200 Series HPLC system (Wilmington, DE,

28 USA) equipped with a diode array detector (DAD), controlled by the EZChrom Elite (Agilent) software,
 6

1 with an automatic liquid sampler (ALS) and quaternary pump. The sample injection volume was 5 L.

2 The compounds were separated in a Zorbax Eclipse XDB-C18 column (4.6 × 150 mm, 5 m), using a

3 gradient system of two mobile phases consisting of water/formic acid (99.55/0.45; v/v - eluent A) and

4 methanol/formic acid (99.55/0.45; v/v - eluent B).

5 The gradient program, with a flow of 1 mL.min-1, began with an isocratic chromatographic run

6 for 9 min composed of 80 % and 20% of mobile phases A and B, respectively. The gradient of the

7 mobile phases then started, reaching 58 % of mobile phase A at 11 min. This composition was

8 maintained in isocratic mode until a total run time of 20 min. A new gradient was then started, reaching

9 80 % of mobile phase A at 22 min and this was maintained up to a total run time of 25 min, corresponding

10 to the total time of the chromatographic run. Chromatograms were recorded at 272 nm for Th and Ca,

11 at 325 nm for 5-CQA and at 370 nm for Ru. The concentrations of all compounds analyzed were

12 determined by external calibration and the identification was performed on the basis of retention times

13 and the UV absorption spectra obtained for the standard references.

14

15 2.4. Membrane Systems

16 This study was performed using two types of membrane systems:

17 a) Ceramic pilot unit, development as previously reported (Silva et al., 2012). This unit has an

18 multi-channel tubular -alumina ceramic membrane (Fairey Ceramic IncTM, London, England) with a

19 nominal pore size of 0.44 m, length of 60 cm, external diameter of 2.0 cm, 7 channels with internal

20 diameter of 0.26 cm and total surface area of 0.06 m². The system was operated with TMP varying

21 between 0.04 and 0.10 MPa, at a feed rate of 600 L h-1 and at room temperature.

22 b) Commercial polymeric cross-flow system (PAM MembranasTM, Brazil) with hollow fiber

23 microfiltration membranes (MF-PEI), made of polyetherimide, with a nominal pore size of 0.44 m,

24 length of each fiber 20 cm, internal diameter of each fiber of 4 mm, ratio of membrane surface area and

25 housing volume of 300 m² m-3 and surface area of 3.6 cm2 and ultrafiltration membranes (UF-PES),

26 made of polyethersulfone with a nominal molecular weight limit (NMWL) of 50 kDa and a surface area
 7

1 of 3.6 cm2. This system was operated at 150 L h-1, at room temperature and with TMP varying between

2 0.01 and 0.07 MPa.

3 Initially, all types of membrane configurations were tested in the total recycling mode with initial

4 crude yerba mate of 2 kg, i.e., the retentate and permeate were continuously recycled to the feed tank to

5 ensure steady-state conditions in relation to the volume and composition of the feed. This configuration

6 was adopted to evaluate the effect of the TMP on the permeate flux and the bioactive compounds content

7 for each membrane. Such compounds were evaluated in each membrane configuration by collecting 2

8 g for each system on the permeate stream.

9 The membrane configuration with the lowest loss of bioactive compounds was selected for the

10 dynamic study, carried out in batch mode, to evaluate the effect of transient operation and determine the

11 predominant fouling mechanism which occurs during the application of the crossflow process to the

12 aqueous extract of yerba mate. The permeate flux was evaluated by the gravimetric method (Silva et al.,

13 2012), using Equation 1:

M Permeate
14 J P (t )  (1)
tS

15 Where JP is the permeate flow rate (kg m-2 h-1), MPermeate is the mass of clarified yerba mate

16 accumulated in permeate stream (kg) during the filtration time t (h), and S is the membrane surface area

17 (m²). The percentage of permeate flux decline (J) was estimated by Equation 2, adapted by Luo et al.

18 (2013),

19  J (t )  J (t i )  (2)
J   i 1   100
 J ( t i 1 ) 

20 Where J(ti) and J(ti-1) are the permeate flux evaluated at a specific time (i) and at a time (i-1) prior

21 to (i), respectively. The rejection coefficient for each bioactive compound was calculated according to

22 Equation (2) (Silva et al., 2012).

23  C  (3)
CR  1  P  100
 CF 

24 where CR is the bioactive compounds rejection coefficient (%), CP is the concentration of

25 bioactive compounds in the permeate stream (g L-1), and CF is the concentration of bioactive compounds
 8

1 in the feed stream (g L-1). The concentration factor for the batch mode process was assessed using

2 Equation (4) (Ushikubo et al., 2007).

MF
3 CF  (4)
M F  M Permeate

4 where CF is the concentration factor and MF is the initial batch mass (kg) and MPermeate is the mass

5 of clarified yerba mate accumulated in permeate stream (kg) during the filtration time t (h).

7 2.5. Theory

8 2.5.1. Resistance effects during the crossflow membrane process

9 The resistance effects were estimated using the resistance-in-series model (Razi et al., 2012;

10 Ushikubo et al., 2007). This model indicates the overall contribution of the main resistance components

11 which can interfere with the permeate flux. The permeate flux can be described as a function of the total

12 resistance (RT), based on Equation (5):

TMP
13 JP  (5)
  RT

14 where TMP is the transmembrane pressure and  is the permeate viscosity.

15 The total resistance (RT) is comprised of the membrane resistance (RM), the resistance due to

16 concentration polarization (RP ), and the resistance due to fouling (RF ), Equation (6):

17 RT  RM  RP  RF (6)

18 The hydraulic resistance of the membrane (RM) was determined, based on Equation (7), using

19 deionized water, cleaning the membrane at a TMP of 0.04 MPa and applying a water temperature of 25

20 °C.

TMP
21 RM  (7)
W  J W

22 where JW is the water permeate flux during the process with clean membrane and W is the water

23 viscosity.
 9

1 Fouling resistance (RF) was determined with a flux of deionized water after the treatment of the

2 yerba mate extracts and a washing solution of system. After washing, the only components of the total

3 resistance were RF and RM. The fouling resistance was determined using Equation (8):

TMP
4 RF   RM (8)
W  J W '

5 The resistance due to concentration polarization (RP) was determined based on Equation (5) using

6 the (RT), (RM) and (RF) values previously obtained.

8 2.5.6. Identification of the fouling mechanism

9 The predominant fouling mechanism was identified using the model proposed by Hermia (1982)

10 adapted for the crossflow filtration system (Field et al., 1995; Gimenes et al., 2014). The general pore

11 blocking model can be described by Equation (9):

dJ P
12
dt
 
  kn  J P  J *  J P
2 n
(9)

13 where kn and n represent the phenomenological coefficient and the general fouling index,

14 respectively, and J* is the ideal critical flux for which fouling does not occur. The integration of Equation

15 (8) gives the distinct pore-blocking mechanisms as a function of the index (n).

16 For n = 2.0, the complete pore-blocking model is defined in Equation (10). In this model, the

17 particles are larger than the pore opening, a part of the membrane surface covered by the particles is

18 sealed, and permeation is prevented by a reduction in the surface area.

19 J p (t)  J LIM  JO  J LIM   exp k2  t  (10)

20 where Jp(t) is the permeate flux value at time t, Jo is the permeate flux value at the initial time

21 and JLIM is the steady-state value for the permeate flux.

22 The value n = 1.5 gives the internal pore-blocking model. This model, Equation (11), considers

23 that the particle size is smaller than the pore opening and that the particles tend to be retained inside the

24 pores due to adsorption on the pore walls or by deposition in the internal cavities. Since the blocking

25 occurs internally, the reduction in the pore opening by blocking becomes independent of the flux

26 conditions, that is, the flux decays to a null value.

 10

Jo
1 J P (t )  (11)
Jo  Jo 1/ 2
 k1.5  t 
2

2 The value n = 1.0 gives the partial pore-blocking model, Equation (12). In this case, the particles

3 are close to the pore size and they tend to cluster in specific regions of the pore, without closing it.

Jo
4 J P (t )  (12)
J LIM  Jo  expk1  J LIM  t   1

5 For n = 0, the resulting model corresponds to cake formation, Equation (13). Considering the

6 effect of larger particles on the pores and the fact that they agglomerate on the membrane surface

7 forming a filter cake, they constitute an additional resistance to the process.

1   J (t ) J  J LIM   1 1 
8 k0  t  2
 ln  P  O   J LIM     (13)
J LIM   Jo J P (t )  J LIM   J P (t ) J O 

9 The fouling models were adjusted by non-linear least-squares regression analysis at a 95%

10 confidence interval. The Levenberg–Marquardt algorithm was employed as the interactive method, and

11 StatSoft STATISTICA software (version 10.0) was used for all of the calculations. The sum of squared

12 residuals (SSR) and coefficient of determination (R2) were determined as the evaluation criteria for the

13 models. The mean experimental values were adopted for parameters (JLIM) and (Jo), adjusting only the

14 parameter kn. The predominant fouling mechanism is identified by the best fitting model.

15 All experiments were made in triplicate and the results were expressed in terms of means and

16 standard deviations. The analysis of variance was then performed. Tukey’s test was used to compare

17 means with a 5% significance level. The software StatSoft STATISTICA (version 10.0) was used.

18

19 3. Results and discussion

20

21 3.1. Effect of membrane type on the clarification process

22 Figure 1 shows the variation in the permeate flux for each membrane system: hollow fiber

23 microfiltration with PEI membrane (MF-PEI); hollow fiber ultrafiltration with PES membrane (UF-

24 PES); and tubular microfiltration with -alumina membrane (MF-alumina), for the different

25 transmembrane pressures applied. Table 1 reports the results for the turbidity and bioactive compounds
 11

1 content of the crude yerba mate extract and the clarified yerba mate extract for each membrane type.

2 The retention of bioactive compounds and turbidity results for the clarified yerba mate extract are given

3 in Table 2 for each membrane.

4 The turbidity of the crude yerba mate extract, obtained from the batch extraction of 140 g of yerba

5 mate leaves in 2 L of water followed by pre-treatment using a 200 mesh (0.075 mm) stainless steel sieve,

6 was found to be around 207 ± 25 NTU. All types of membranes evaluated enabled an efficient

7 clarification of the crude aqueous extract of yerba mate, reducing the initial turbidity of the aqueous

8 extract to close to zero. This finding was expected since porous membrane separation (both

9 microfiltration and ultrafiltration) is widely used to obtain clarified juices (Jiao et al., 2004; Vaillant et

10 al., 2001) and teas (Subramanian et al., 2014).

11 The highest permeate flux values obtained using the MF-PEI and UF-PES were 36.33 ± 0.71 and

12 19.51 ± 0.47 kg m² h-1, respectively, achieved under lowest transmembrane pressure evaluated (0.02

13 MPa) and feed flow of 150 L h-1. For the MF-alumina the maximum permeate flux was of 50.19 ± 0.54

14 kg m-2 h-1 for 0.07 MPa with a feed flow of 600 L h-1. With polymeric membranes, the increasing the

15 TMP promoted a reduction in the permeate flux due to an increase in the compacting effect and

16 enhancement of the fouling mechanisms, as commonly found in other studies on polymeric membranes

17 (Barros et al., 2003; Persson et al., 1995; Razi et al., 2012).

18 Although the -alumina ceramic membrane provided the highest permeate flux value, the

19 rejection of bioactive compounds (22.8-27.8 %) was higher than the values obtained for the other

20 membranes. In a clarification process it is desirable to obtain an extract without turbidity while retaining,

21 as far as possible, the nutrient content, i.e., the coefficient of rejection for the permeate stream should

22 be close to zero. In the case of the polymeric membranes, the rejection of bioactive compounds varied

23 between 18.3 and 35.5 % for UF-PES while for MF-PEI values ranged from 11 to 15.5 %. The bioactive

24 lost observed in all kind of the membrane system evaluated can be explained by the solid suspended

25 rejection into membrane and the adsorption of each bioactivity compounds on membrane surface. The

26 turbidity retrieved on membrane contain small particle of yerba mate, these particles have some amount

27 of bioactive compounds.
 12

1 Although there are few studies reported in the literature addressing the treatment of yerba mate

2 extract using a membrane process, tea infusions have been successfully clarified using membrane

3 processes. Murakami et al. (2011) applied a nanofiltration process (Osmonics membranes, HL2521TF,

4 molecular weight cut-off 150 Da) to concentrate the phenolic compounds present in a crude yerba mate

5 extract and the authors obtained a total phenolic retention of almost 99 %, but the concentrated yerba

6 mate extract still contained a large amount of suspended solids. In a study on green tea clarification

7 applying porous membrane processes, Kawakatsu et al. (1995) evaluated different types of membranes

8 to clarified a green tea extract and reported an almost 100 % rejection of suspended solids using

9 polyethersulfone (DDS Dow Danmark A/S) and polysulfone (Nihon Millipore, Ltda) ultrafiltration

10 membranes, but they observed caffeine losses of 15.7 and 19.9 % (wt), respectively, which are similar

11 to the caffeine losses observed in our study.

12 The polymeric system with a polyetherimide microfiltration membrane (MF-PEI) was selected to

13 evaluated the batch process for the clarification of the crude yerba mate extract since the loss of bioactive

14 compounds was lower compared with the UF-PES and MF-alumina and the permeate flux was higher

15 than that of UF-PES. The transmembrane pressure adopted was 0.02 MPa, as explained above.

16

17 3.2. Batch process operation

18 The microfiltration process carried out in batch mode resulted in a clarified extract with a 100.0%

19 reduction in turbidity, ensuring a product free of suspended colloids, which is desirable since these can

20 cause technological problems (precipitation) in the beverage industry. Liang and Xu (2001) showed that

21 the average size of the particles responsible for the “creaming” or "fog" phenomenon, which can occur

22 during the cooling of ready-to-drink teas through the formation of polyphenol complexes, is 0.50 µm.

23 Since the pore size of the MF-PEI membrane used was 0.40 m, it can be assumed that this phenomenon

24 is also minimized by clarifying the extract.

25 The behavior of the permeate flux on treating the crude yerba mate extract using the MF-PEI is

26 shown in Figure 2 and Table 3. Initially, there was an abrupt decrease, followed by a gradual and

27 continuous decline, resulting in a final permeate flux reduction of approximately 65.0% compared with

28 the initial value. The high values of the permeate flux decline were observed at the beginning, due the
 13

1 polarized layer formation, with a reduction in order of 36 %. After the polarized layer resistance was

2 stabilized, the permeate flux show a continuous decrease, with percentage of flux decline in order of 9

3 %, indicated the operational conditions adopted induced a permeate flux above of the hypothetic critical

4 flux.

5 This behavior is typical of a crossflow system, as observed by Kawakatsu et al. (1995) in the

6 clarification of green tea by ultrafiltration with different types of membranes. The amount of feed in

7 batch operation was reduced by 50.0 % after 45 min of operation, but the increase in the CF did not

8 significantly affect the permeate flux during the process.

9 The concentration of bioactive compounds and the coefficient of rejection (%) were determined

10 during the application of the MF-PEI batch mode operation to the clarification of the yerba mate extract

11 and the values obtained are showed in Table 3 and Figure 3, respectively. According with the ANOVA

12 and Tukey’s test, the bioactive compounds concentration in the clarified yerba mate stream showed

13 lower than the crude yerba mate extract feed, but these concentrations in permeate stream show

14 statistically equal during the clarification process in batch mode operation. After 45 minutes, the caffeine

15 and theobromine concentrations were 577.25 ± 14.90 and 90.32 ± 1.73 mg L-1, respectively, while the

16 rutin and chlorogenic acid concentrations were 30.54 ± 1.45 and 389.50 ± 15.31 mg L-1, respectively.

17 These differences occur due to the interaction of these compounds with the polyetherimide

18 leading to the polarized layer and fouling phenomena. As previously mentioned, the rejection of solutes

19 during the clarification process suggests their loss from the clarified extract and, according to Figure 3,

20 the rejection of the bioactive compounds during the microfiltration process varies between 8 % and 18

21 % for concentrate factor of 50 % of initial feed mass, indicating that the loss of bioactive compounds

22 during the process is low.

23

24 3.3. Analysis of resistance phenomena during the clarification process

25 Table 4 reports the contribution of each effect to the total resistance to permeation of the yerba

26 mate extract through the polyetherimide membrane. According to the resistance-in-series analysis, the

27 polarized layer can be considered as the major resistance phenomenon, accounting for around 66 % of
 14

1 the total resistance to permeation, while the fouling and hydraulic membrane resistances represent

2 around 20 % and 14 % of the total resistance, respectively.

3 Argyle and Bird (2014) also reported that the resistance due to concentration polarization is

4 predominantly forward resistance in the clarification of black tea (Camellia sinensis) using a

5 microfiltration system with a polysulfone membrane (pore size varying between 0.5 and 1.55 m). The

6 authors concluded that this is due to the high concentration of solids in the feed, which favors the

7 formation of a concentration gradient at the membrane surface.

8 To identify the predominant fouling mechanism that occurred during the clarification of the crude

9 yerba mate extract, a comparative study, considering the experimental values obtained for the permeate

10 flux and the pore blocking model development by Hermia (1982) and adapted by Field et al. (1995),

11 was carried out. The coefficient of determination (R²) and the sum of squared residuals (SSR) were the

12 criteria used to compare the numerical predictions and the experimental data and to select the best fit

13 model for each assay evaluated. The isotherm parameters and statistical analysis results are summarized

14 in Table 5 and Figure 4 shows the permeate flux prediction obtained applying each model.

15
16 The fouling mechanism described as cake filtration showed the best fit with the highest R2 value

17 of 99.74% and lowest SSR value of 157.93. For all models considered the precision observed in the

18 fitted results is initially poor, as shown in Figure 4, since these models evaluate only the fouling aspects

19 and at the beginning of the process the resistance mechanisms are not fully formed, in particular the

20 dynamic resistances, such as the polarized layer.

21 This fouling mechanism, cake deposition in membrane surface, is commonly observed in juice

22 clarification processes, as reported by many authors, for instance, Barros et al. (2003), in a study on

23 pineapple juice clarification by ultrafiltration (polysulfone membrane, cut-off 100 kDa), Razi et al.

24 (2012), in a study on tomato juice clarification using microfiltration (polyvinylidenefluoride membrane,

25 nominal pore size 0.45 m) and Qin et al. (2015) in study on kiwi juice clarification with a

26 microfiltration system (synthetic fly-ash-based inorganic ceramic, nominal pore size 0.30 m).

27 Although the MF-PEI operate at a low transmembrane pressure (0.02 MPa), the permeate flux

28 showed a high decrease during the batch mode, as showed in Table 3. This behavior can be explained
 15

1 by the operational condition adopted: the feed flow rate was 150 L h-1, that is the highest flow capacity

2 of the pump used in the polymeric unity of this study. For a initial feed flow of 150 L h-1, the estimated

3 tangential velocity in the membrane surface and Reynolds number were 0.13 m s-1 and 2,570,

4 respectively, that is, laminar flow and the feed flow exerts a low renewal effect onto the suspended

5 solids deposited onto membrane surface. Since the predominant fouling mechanism is cake filtration, a

6 physical procedure for the cleaning of the surface of the membrane, such as backflushing or ultrasound,

7 should be investigated.

8 Luo et al. (2013) evaluated the performance of high-shear dynamic filtration in chicory juice

9 clarification by UF and MF with rotating membrane disk and observed that the shear rate increased with

10 the membrane disk rotation and promoted the reduction of polarized layer and fouling phenomena.

11 Tangential system, like the adopted in this study, showed a limited shear rate and low renewal effect

12 onto membrane surface, as indicated by high flux decline during the 45 minutes of batch process and

13 reduction of the initial feed mass by almost 50 % (Table 3). To improve the shear rate in a classical

14 tangential membrane system will be necessary to evaluate some hydrodynamic techniques as pulsating

15 flow, pressure pulsing or periodic backwashing. These techniques can be applied to improve the renewal

16 effect of the tangential feed flow, reducing the external fouling and dynamic polarized layer, resulting

17 in a low permeate flux decline, approaches of critical flux concept.

18
19 4. Conclusions
20

21 This study verified the potential for the application of a porous membrane process in the

22 clarification of crude yerba mate extract, with minimal loss of bioactive compounds.

23 The turbidity of the clarified extract approaches zero for all types of membranes evaluated. The

24 lowest coefficient for the rejection of bioactive compounds was observed for the MF carried out with a

25 polyetherimide membrane (11 and 15.5 %), which suggests that the clarified yerba mate extract contains

26 more bioactive compounds than the clarified yerba mate extract obtained applying MF with an alumina

27 membrane or UF with a polyethersulfone membrane.

 16

1 The major resistance effect during the clarification of the crude yerba mate extract with MF-PEI

2 at 0.02 MPa is the polarized layer, accounting for around 66 % of the total resistance.

3 Based on the analysis applying the blocking models it is concluded that the dominant resistance

4 mechanisms are the surface effects, notably the formation of a filtration cake on the membrane surface.

NOMENCLATURE

5-CQA – 5-O-caffeoylquinic acid

Ca - Caffeine
CF - Concentration of bioactive compounds in the feed stream (g L-1)
CF - Concentration factor
CP - Concentration of bioactive compounds in the permeate (g L-1)
CR - Coefficient of rejection (%)
CS(t) - Concentration of bioactive compounds in the solution as a function of time (g L-1)
J* - Critical permeate flux (kg m-2 h-1)
JO - Initial permeate flux (kg m-2 h-1)
JLIM - Permeate flux at the end of the process (kg m-2 h-1)
JP - Permeate flux (kg m-2 h-1)
J(ti) - Permeate flux at a specific time (i) (kg m-2 h-1)
J(ti-1) - Permeate flux at a time prior to (i) (kg m-2 h-1)
JW - Water permeate flux during the process with clean membrane (kg m-2 h-1)
JW’ - Water permeate flux through the blocked membrane (kg m-2 h-1)
kn- Phenomenological coefficient of pore blocking
MF - Initial batch mass (kg)
MF - Microfiltration
MF-alumina – α-alumina microfiltration membrane
MF-PEI – Polyetherimide microfiltration membrane
MPermeate - Accumulated permeate mass (kg)
n - General fouling index
R2 – Coefficient of determination
RF - Resistance due to fouling (m2 kg-1)
RM - Membrane resistance (m2 kg-1)
RP - Resistance due concentration polarization (m2 kg-1)
RT- Total resistance (m2 kg-1)
Ru - Rutin
S- Surface area of the membrane (m²)
SSR - Sum of Square residuals
t- Operational time (h)
Th - Theobromine
TMP - Transmembrane pressure (MPa)
UF - Ultrafiltration
UF-PES - Polyethersulfone ultrafiltration membrane
 - Permeate viscosity (Pa.s)
W - Water viscosity (Pa.s)
ΔJ - Permeate flux decline (%)

ACKNOWLEDGEMENTS
 17

The authors are grateful for the scholarship provided by CAPES, the research grant and support

of the Graduation Program of Food Engineering (Federal University of Paraná, Curitiba, Brazil) and the

yerba mate supplied by Vier Ltda.

REFERENCES

Anesini, C., Turner, S., Cogoi, L., Filip, R., 2012. Study of the participation of caffeine and
polyphenols on the overall antioxidant activity of mate (Ilex paraguariensis). LWT - Food
Sci. Technol. 45, 299–304. doi:10.1016/j.lwt.2011.06.015

Argyle, I.S., Bird, M.R., 2014. Microfiltration of high concentration black tea streams for haze
removal using polymeric membranes. Desalin. Water Treat. 53, 1516–1531.
doi:10.1080/19443994.2014.953401

Barros, S.T.D., Andrade, C.M.G., Mendes, E.S., Peres, L., 2003. Study of fouling mechanism
in pineapple juice clarification by ultrafiltration. J. Memb. Sci. 215, 213–224.
doi:10.1016/S0376-7388(02)00615-4

Berté, K.A.S., Izidoro, D.R., Dutra, F.L.G., Hoffmann-Ribani, R., 2011. Desenvolvimento de
gelatina funcional de erva-mate. Ciência Rural 41, 354–360. doi:10.1590/S0103-
84782011000200029

Bracesco, N., Sanchez, A.G., Contreras, V., Menini, T., Gugliucci, A., 2011. Recent advances
on Ilex paraguariensis research: Minireview. J. Ethnopharmacol. 136, 378–384.
doi:10.1016/j.jep.2010.06.032

Burris, K.P., Harte, F.M., Michael Davidson, P., Neal Stewart Jr, C., Zivanovic, S., 2012.
Composition and Bioactive Properties of Yerba Mate (llex paraguariensis A. St.-Hil.): A
Review. Chil. J. Agric. Res. 72, 268–275. doi:10.4067/S0718-58392012000200016

Cassano, A., Donato, L., Drioli, E., 2007. Ultrafiltration of kiwifruit juice: Operating
parameters, juice quality and membrane fouling. J. Food Eng. 79, 613–621.
doi:10.1016/j.jfoodeng.2006.02.020

Conforti, A.S., Gallo, M.E., Saraví, F.D., 2012. Yerba Mate (Ilex paraguariensis) consumption
is associated with higher bone mineral density in postmenopausal women. Bone 50, 9–13.
doi:10.1016/j.bone.2011.08.029

Dutra, F.L.G., Hoffmann-Ribani, R., Ribani, M., 2010. Determination of phenolic compounds
by isocratic high performance liquid chromatografic method during storage of yerba-mate.
Quim. Nova 33, 119–123.

Espamer, L., Pagliero, C., Marchese, J., Ochoa, A., 2006. Clarification of lemon juice using
membrane process. Desalination 200, 565–567.

Field, R.W., Wu, D., Howell, J.A., Gupta, B.B., 1995. Critical flux concept for microfiltration
fouling. J. Memb. Sci. 100, 259–272. doi:10.1016/0376-7388(94)00265-Z
 18

Filip, R., Davicina, R., Anesini, C., 2010. Antifungal activity of the aqueous extract of Ilex
paraguariensis against Malassezia furfur. Phyther. Res. 24, 715–719. doi:10.1002/ptr

Gimenes, M.L., Silva, V.R., Hamerski, F., Scheer, A.P., 2014. Pretreatment of Aqueous Pectin
Solution by Cross-Flow Microfiltration: Study on Fouling Mechanism. Int. J. Chem. Eng.
Appl. 5, 281–286. doi:10.7763/IJCEA.2014.V5.394

Godoy, R.C.B., Deliza, R., Gheno, L.B., Licodiedoff, S., Frizon, C.N.T., Ribani, R.H., dos
Santos, G.G., 2013. Consumer perceptions, attitudes and acceptance of new and traditional
mate tea products. Food Res. Int. 53, 801–807. doi:10.1016/j.foodres.2013.02.054

Habert, A.C., Borges, C.P., Nobrega, R., 2006. Processos de Separação por Membranas, 1st ed.
E-papers, Rio de Janeiro.

Heck, C.I., De Mejia, E.G., 2007. Yerba mate tea (Ilex paraguariensis): A comprehensive
review on chemistry, health implications, and technological considerations. J. Food Sci.
72, 138 –151. doi:10.1111/j.1750-3841.2007.00535.x

Hérmia, J., 1982. Constant pressure blocking filtration laws. Applications to power-law non-
Newtonian fluids. Trans. Instituition Chem. Eng. 60, 183–187.

Jensen, S., Zanoelo, É.F., 2012. Kinetics of aqueous extraction of mate (Ilex paraguariensis)
leaves. J. Food Process Eng. 36, 220–227. doi:10.1111/j.1745-4530.2012.00675.x

Jiao, B., Cassano, A., Drioli, E., 2004. Recent advances on membrane processes for the
concentration of fruit juices: A review. J. Food Eng. 63, 303–324.
doi:10.1016/j.jfoodeng.2003.08.003

Kawakatsu, T., Kobayashi, T., Sano, Y., Nakajima, M., 1995. Clarification of Green Tea
Extract by Microfiltration and Ultrafiltration. Biosci. Biotechnol. Biochem. 59, 1016–
1020. doi:10.1271/bbb.59.1016

Kim, S.-Y., Oh, M.-R., Kim, M.-G., Chae, H.-J., Chae, S.-W., 2015. Anti-obesity effects of
Yerba Mate (Ilex Paraguariensis): a randomized, double-blind, placebo-controlled clinical
trial. BMC Complement. Altern. Med. 15, 338. doi:10.1186/s12906-015-0859-1

Koltuniewicz, A.B., Field, R.W., Arnot, T.C., 1995. Cross-flow and dead-end microfiltration
of oily-water emulsion. Part I: Experimental study and analysis of flux decline. J. Memb.
Sci. 102, 193–207. doi:10.1016/0376-7388(94)00320-X

Liang, Y., Xu, Y., 2001. Effect of pH on cream particle formation and solids extraction yield
of black tea. Food Chem. 74, 155–160. doi:10.1016/S0308-8146(01)00108-X

Luo, J., Zhu, Z., Ding, L., Bals, O., Wan, Y., Michel, Y., Jaffrin, M. Y., Vorobiev, E., 2013.
Flux behavior in clarification of chicory juice by high-shear membrane filtration: Evidence
for threshold flux. J. Membr. Sci. 435, 120–129, doi: 10.1016/j.memsci.2013.01.057

Murakami, A.N.N., De Mello Castanho Amboni, R.D., Prudêncio, E.S., Amante, E.R., de
Moraes Zanotta, L., Maraschin, M., Cunha Petrus, J.C., Teófilo, R.F., 2011. Concentration
of phenolic compounds in aqueous mate (Ilex paraguariensis A. St. Hil) extract through
nanofiltration. LWT - Food Sci. Technol. 44, 2211–2216. doi:10.1016/j.lwt.2011.06.002
 19

Neggaz, Y., Vargas, M.L., Dris, A.O., Riera, F., Alvarez, R., 2007. A combination of serial
resistances and concentration polarization models along the membrane in ultrafiltration of
pectin and albumin solutions. Sep. Purif. Technol. 54, 18–27.
doi:10.1016/j.seppur.2006.08.017

Pagliosa, C.M., Vieira, M.A., Podestá, R., Maraschin, M., Zeni, A.L.B., Amante, E.R., Amboni,
R.D. de M.C., 2010. Methylxanthines, phenolic composition, and antioxidant activity of
bark from residues from mate tree harvesting (Ilex paraguariensis A. St. Hil.). Food Chem.
122, 173–178. doi:10.1016/j.foodchem.2010.02.040

Pang, J., Choi, Y., Park, T., 2008. Ilex paraguariensis extract ameliorates obesity induced by
high-fat diet: Potential role of AMPK in the visceral adipose tissue. Arch. Biochem.
Biophys. 476, 178–185. doi:10.1016/j.abb.2008.02.019

Persson, K.M., Gekas, V., Tr??g??rdh, G., 1995. Study of membrane compaction and its
influence on ultrafiltration water permeability. J. Memb. Sci. 100, 155–162.
doi:10.1016/0376-7388(94)00263-X

Razi, B., Aroujalian, A., Fathizadeh, M., 2012. Modeling of fouling layer deposition in cross-
flow microfiltration during tomato juice clarification. Food Bioprod. Process. 90, 841–
848. doi:10.1016/j.fbp.2012.05.004

Silva, V.R., Hamerski, F., Scheer, A.P., 2012. Pretreatment of aqueous pectin solution by cross-
flow microfiltration: Analysis of operational parameters, degree of concentration and
pectin losses. Int. J. Food Sci. Technol. 47, 1246–1252. doi:10.1111/j.1365-
2621.2012.02965.x

Silveira, T.F.F., Meinhart, A.D., De Souza, T.C.L., Teixeira Filho, J., Godoy, H.T., 2016.
Phenolic compounds from yerba mate based beverages - A multivariate optimisation. Food
Chem. 190, 1159–1167. doi:10.1016/j.foodchem.2015.06.031

Subramanian, R., Chandini, S.K., Sharma, P., 2014. Membrane Clarification of Tea Extracts.
Crit. Rev. Food Sci. Nutr. 54, 1151–1157. doi:10.1007/s11947-012-0847-0

Ushikubo, F.Y., Watanabe, A.P., Viotto, L.A., 2007. Microfiltration of umbu (Spondias
tuberosa Arr. Cam.) juice. J. Memb. Sci. 288, 61–66. doi:10.1016/j.memsci.2006.11.003

Vaillant, F., Millan, A., Dornier, M., Decloux, M., Reynes, M., 2001. Strategy for economical
optimization of the clarification of pulpy fruit juices using crossflow microfiltration. J.
Food Eng. 48, 83–90. doi:10.1016/S0260-8774(00)00152-7
 20

Figure 1: Effect of membrane material and transmembrane pressure on permeate flux:

(a) Polymeric system: --■-- MF-PEI, —○— UF-PES; (b) Ceramic system: —◊— MF-alumina.

Figure 2: Permeate flux and concentration factor for crude yerba mate extract treated by MF-PEI in batch mode
operation. …○… Permeate flux; —■— Concentration factor (CF)
 21

Figure 3: Coefficient of rejection of bioactive compounds determined during the treatment of the crude yerba mate
extract using MF-PEI in batch mode operation. —●— Rutin; ---□--- Caffeine; ….♦…. Chlorogenic acid; —+—
Theobromine

Figure 4. Fitting of the pore blocking models. ● experimental data; — complete pore blocking; — internal pore
blocking; …… partial pore blocking; ------ cake filtration
 22

Table 1: Turbidity and bioactive compounds content for yerba mate extracts, crude and clarified using different
membranes.

Parameters Crude yerba mate extract MF-PEI1 MF-alumina2 UF-PES1

Turbidity (NTU) * 207 ± 25 n.s. n.s. n.s.

Chlorogenic Acid (mg L-1) 449.2 ± 23.4A 392.6 ± 11.3AB 334.7 ± 1.0B 350.5 ± 18.9B

Rutin (mg L-1) 40.0 ± 3.9A 33.8 ± 2.4AB 28.9 ± 0.4AB 25.8 ± 3.9B

Caffeine (mg L-1) 639.8 ± 13.2A 569.5 ± 13.6B 494.0 ± 7.6C 522.7 ± 18.4BC

Theobromine (mg L-1) 103.6 ± 4.6A 90.2 ± 0.8B 75.9 ± 0.2C 82.1 ± 2.7BC
Means with same letters in the same line represent not statistically differences according to the Tukey’s test (P < 0.05).
1 Operational parameters in polymeric system: TMP = 0.02 MPa, T = 20 °C, feed flow of 150 L h-1.
2 Operational parameters in ceramic system: TMP = 0.07 MPa, T = 20 °C, feed flow of 600 L h-1.

* n.s.: not significant.

Table 2: Coefficient of rejection for bioactive compounds content and turbidity for each membrane material.

Coefficient of Rejection, CR (%)

Membrane material
Turbidity Chlorogenic Acid Rutin Caffeine Theobromine

MF-PEI 100 12.6 ± 0.7 15.5 ± 1.9 11.0 ± 0.3 12.9 ± 0.6

MF-alumina 100 25.5 ± 1.3 27.8 ± 2.7 22.8 ± 0.6 26.7 ± 1.2

UF-PES 100 22.0 ± 1.6 35.5 ± 6.4 18.3 ± 0.7 20.8 ± 1.1
1 Operational parameters in polymeric system: TMP = 0.02 MPa, T = 20 °C, feed flow of 150 L h-1.
2Operational parameters in ceramic system: TMP = 0.07 MPa, T = 20 °C, feed flow of 600 L h-1.

Table 3: Clarification of yerba mate extract with MF-PEI in batch mode: bioactive compounds concentration at
permeate stream, permeate flux and flux variation during the batch process.
Time Theobromine Caffein Chlorogenic acid Rutin Permeate flux Flux decline
(min) (mg L-1) (mg L-1) (mg L-1) (mg L-1) (kg m-2 h-1) J (%)
0* 102.5 ± 8.7A 679.9 ± 16.4A 463.3 ± 30.7A 37.3 ± 1.4A 82.0 ± 11.4
1 90.9 ± 6.1A 574.0 ± 34.8AB 390.2 ± 24.6A 31.4 ± 1.4B 62.3 ± 13.9 24.4 ± 6.4
5 92.0 ± 3.3A 581.6 ± 21.9AB 395.3 ± 11.5A 32.6 ± 1.4B 38.9 ± 4.8 36.9 ± 6.4
10 91.3 ± 3.3A 585.7 ± 27.7B 402.3 ± 14.3A 32.4 ± 1.4B 32.2 ± 2.2 17.0 ± 4.4
20 93.6 ± 3.6A 589.0 ± 27.5B 404.0 ± 21.2A 31.2 ± 1.4B 27.5 ± 1.5 14.4 ± 1.4
30 93.2 ± 1.1A 581.3 ± 20.4B 394.6 ± 16.2A 31.1 ± 1.4B 24.8 ± 1.1 9.9 ± 1.0
45 90.3 ± 1.8A 577.3 ± 14.9B 389.5 ± 15.3A 31.5 ± 1.4B 22.4 ± 0.7 9.6 ± 1.1
Means with same letters in the same column represent not statistically differences according to the Tukey’s test (P < 0.05).
* - Bioactive compounds concentration refers at crude yerba mate extract, and the permeate flux estimated at the beginning of batch process
 23

Table 4: Contribution of resistance phenomena during clarification of crude yerba mate extract with MF-PEI.
Resistance
Resistance effects
(m2.kg-1 x 10-6)
Hydraulic membrane (RM) 0.43±0.02
Fouling (RF) 0.65±0.09
Polarized layer (RP) 2.12±0.15
Total resistance 3.20±0.13

Table 5: Permeate flux data modeling.

Complete pore Internal pore Partial pore
Model Cake filtration
blocking blocking blocking
parameters n=0
n=2.0 n=1.5 n=1.0
Kn 9.85±0.51 0.18±0.01 0.20±0.01 5.4x10-3±2.0x10-4
JLIM 24.40±0.79 0 24.40±0.79 24.40±0.79
Jo 68.20±15.68 68.20±15.68 68.20±15.68 68.20±15.68
R² 0.9374 0.644 0.9802 0.9974
SSR 522.12 2972.39 164.92 157.93