Influences of Technological Parameters on Cross-Flow Nanofiltration of Cranberry Juice

The paper focused on the influence of operative conditions on the separation of benzoic acid from 10 °Brix cranberry juice by cross-flow nanofiltration with a plate and frame pilot scale (DDS Lab Module Type 20 system). Six kinds of commercial nanofiltration membrane were investigated. The results showed that the rejection of benzoic acid was significantly lower than that of other components in cranberry juice, including sugars and other organic acids. In a range of 2–7.5 L/min, feed flow rate slightly affected the performance of nanofiltration. Higher temperatures resulted in higher permeate flux and lower rejection of benzoic acid, whereas rejection of sugar and organic acid was stable at a high value. In a range of 2.5–5.5, pH also significantly affected the separation of benzoic acid and negative rejection against benzoic acid was observed at pH 4.5 with some of the membranes. This implies that pH 4.5 is considered as an optimum pH for benzoic acid separation from cranberry juice. The lower permeate flux caused a lower rejection of benzoic acid and negative rejection of benzoic acid was observed at the low permeate flux. Pretreatment by ultrafiltration with CR61PP membranes could improve the permeate flux but insignificantly influenced the efficiency of separation. The results also indicated that NF99 and DK membranes can be effectively used to separate benzoic acid from cranberry juice.


Introduction
Benzoic acid (M = 122.12, pKa = 4.21) has been used widely in the food and cosmetic industries as a preservative because of its anti-bacterial property. In nature, it occurs in some fruits and spices, especially in the cranberry (Vaccinium macrocarpon) [1]. The amount of benzoic acid in cranberry fruit is approximately 4.741 g/kg fresh weight, including about 10% in free state and 90% in bound state [2]. Fresh cranberry juice (prepared from the cranberry by squeezing) contains 41 ppm of free benzoic acid, and the total content (including free and bound states) is about 178 ppm [1,3]. The content of benzoic acid in 50 • Brix concentrated cranberry is approximately 500 ppm. This content implies that cranberry juice contains an excessive amount of benzoic acid which can be utilized as a natural preservative if it is separated with low cost. The De Danske Sukkerfabriker (DDS) "Lab Module Type 20" plate and frame system (Copenhagen, Denmark) was used to conduct the nanofiltration in this study ( Figure 1). The unit consisted of 6 couples of membrane sheets (0.018 m 2 /sheet). The unit was equipped with a high-pressure pump (Hydra-Cell pump, supplied by Wanner Engineering Inc., Minneapolis, MN, USA) to provide feed for the unit. Six kinds of commercial nanofiltration membranes (stated in Table 1) were installed in series (1 couple per membrane). During operation, permeate and retentate were fully circulated into the feed tank to remain in the feedwithout changing its chemical composition (total recirculation mode).

Membrane Apparatus
Six kinds of commercial nanofiltration membrane were investigated with the characteristics shown in Table 1. The De Danske Sukkerfabriker (DDS) "Lab Module Type 20" plate and frame system (Copenhagen, Denmark) was used to conduct the nanofiltration in this study ( Figure 1). The unit consisted of 6 couples of membrane sheets (0.018 m 2 /sheet). The unit was equipped with a high-pressure pump (Hydra-Cell pump, supplied by Wanner Engineering Inc., Minneapolis, MN, USA) to provide feed for the unit. Six kinds of commercial nanofiltration membranes (stated in Table 1) were installed in series (1 couple per membrane). During operation, permeate and retentate were fully circulated into the feed tank to remain in the feedwithout changing its chemical composition (total recirculation mode).

Analysis Methods
Glucose and fructose (180 g/mol) were analyzed by using YMC-Pack Polyamine II (250 × 4.6 mm ID) (supplied by YMC Co. Ltd., Kyoto, Japan) coupled to a refractive index detector and Waters 515 HPLC pump. The column was maintained at 35 °C and the mobile phase was acetonitrile/water (70/30) at the flow rate 1.0 mL/min.

Analysis Methods
Glucose and fructose (180 g/mol) were analyzed by using YMC-Pack Polyamine II (250 × 4.6 mm ID) (supplied by YMC Co. Ltd., Kyoto, Japan) coupled to a refractive index detector and Waters 515 HPLC pump. The column was maintained at 35 • C and the mobile phase was acetonitrile/water (70/30) at the flow rate 1.0 mL/min.

Performance Parameters
The performance of nanofiltration was expressed in terms of permeate flux (L/m 2 /h) and the observed rejection of benzoic acid, sugars (being defined as the sum of glucose and fructose), organic acids (being defined as the sum of quinic acid, malic acid, and citric acid). Both of them were determined when permeate flux became stable (after approximately 30-40 min).
The observed rejection R o was calculated from the following formula where, C p and C f were the concentration (g/L) of the solutes in permeate and feed, respectively.

Influence of Feed Flow Rate
The effect of feed flow rate on permeate flux in the separation of benzoic acid from cranberry juice was shown in Figure 2a. It is apparent that feed flow rate slightly affected the permeate flux. Theoretically, the increase in cross-flow velocity reduces the concentration polarization on the membrane surface. When the cross-flow velocity reaches a critical value, the concentration polarization is considered to be eliminated and the relationship between solution permeability and applied pressure is linear at constant cross-flow velocity [18]. In our present work, with the investigated range of feed flow rate, during the nanofiltration of cranberry juice with "Lab Module Type 20" plate and frame system, concentration polarization might be diminished because the relationship between permeate flux and applied pressure was linear, with the correlation coefficient being approximately 1 (the result was not shown in this paper). Therefore, the possible explanation for the effect of feed flow rate is that an increase in velocity of fluid flow on the membrane surface reduced the reversible adsorption on the membrane surface, consequently reducing the resistance to permeate flux.
From the data in Figure 2a, DRA 4510 has almost half the flux of NF99 but NF99 has less permeate flux compared to DRA 4510. This is because NF99 is a hydrophilic membrane, while DRA 4510 is a hydrophobic membrane. Hydrophilic membranes are capable of forming gravitational interactions between water and membrane materials such as dipole-dipole interactions, hydrogen bonds and ionic-dipole interactions [19].
The effect of feed flow rate on the rejection of sugars ( Figure 2b) and organic acids ( Figure 2c) in cranberry juice fruit was inappreciable and approximately 1.0 for sugar and above 0.9 for organic acids, with the exception being the G5 membrane. Perhaps the high rejection concealed the effect of feed flow rate. Figure 2 also shows the effect of feed flow rate on the benzoic acid separation from cranberry juice. From data in Figure 2b-d, it is apparent that rejection of benzoic acid was lower than for sugars and other organic acids. The reasonable explanation for this is that the molecular weight of benzoic acid is lower than the others. Thus, benzoic acid goes through membrane more easily. Simultaneously, results also indicate that cross flow velocity slightly affected the rejection of benzoic acid. Maybe this phenomenon is related to the contribution of convection and selective layers made from reversible adsorption on the surface of membranes into benzoic acid separation. At a feed flow rate of 3 to 4 L/min, the rejection of benzoic acid in UF-treated cranberry juice decreased, then it increased. This is explained by the UF treatment, which helps cranberry juice reject large compounds so that the small compounds easily pass through the membrane. However, an increase in feed flow rate might promote fouling phenomenon or concentration polarization, leading to higher resistance at the membrane surface [20]. This prevents compounds from passing through the membrane and, consequently, an increase in rejection. From the data in Figure 2a, DRA 4510 has almost half the flux of NF99 but NF99 has less permeate flux compared to DRA 4510. This is because NF99 is a hydrophilic membrane, while DRA 4510 is a hydrophobic membrane. Hydrophilic membranes are capable of forming gravitational interactions between water and membrane materials such as dipole-dipole interactions, hydrogen bonds and ionic-dipole interactions [19].
The effect of feed flow rate on the rejection of sugars ( Figure 2b) and organic acids ( Figure 2c) in cranberry juice fruit was inappreciable and approximately 1.0 for sugar and above 0.9 for organic acids, with the exception being the G5 membrane. Perhaps the high rejection concealed the effect of feed flow rate. Figure 2 also shows the effect of feed flow rate on the benzoic acid separation from cranberry juice. From data in Figure 2b-d, it is apparent that rejection of benzoic acid was lower than for sugars and other organic acids. The reasonable explanation for this is that the molecular weight of benzoic acid is lower than the others. Thus, benzoic acid goes through membrane more easily. Simultaneously, results also indicate that cross flow velocity slightly affected the rejection of benzoic acid. Maybe this phenomenon is related to the contribution of convection and selective layers made from reversible adsorption on the surface of membranes into benzoic acid separation. At a feed flow rate of 3 to 4 L/min, the rejection of benzoic acid in UF-treated cranberry juice decreased, then it increased. This is explained by the UF treatment, which helps cranberry juice reject large compounds From the data in Figure 2a, DRA 4510 has almost half the flux of NF99 but NF99 has less permeate flux compared to DRA 4510. This is because NF99 is a hydrophilic membrane, while DRA 4510 is a hydrophobic membrane. Hydrophilic membranes are capable of forming gravitational interactions between water and membrane materials such as dipole-dipole interactions, hydrogen bonds and ionic-dipole interactions [19].
The effect of feed flow rate on the rejection of sugars ( Figure 2b) and organic acids ( Figure 2c) in cranberry juice fruit was inappreciable and approximately 1.0 for sugar and above 0.9 for organic acids, with the exception being the G5 membrane. Perhaps the high rejection concealed the effect of feed flow rate. Figure 2 also shows the effect of feed flow rate on the benzoic acid separation from cranberry juice. From data in Figure 2b-d, it is apparent that rejection of benzoic acid was lower than for sugars and other organic acids. The reasonable explanation for this is that the molecular weight of benzoic acid is lower than the others. Thus, benzoic acid goes through membrane more easily. Simultaneously, results also indicate that cross flow velocity slightly affected the rejection of benzoic acid. Maybe this phenomenon is related to the contribution of convection and selective layers made from reversible adsorption on the surface of membranes into benzoic acid separation. At a feed flow rate of 3 to 4 L/min, the rejection of benzoic acid in UF-treated cranberry juice decreased, then it increased. This is explained by the UF treatment, which helps cranberry juice reject large compounds From the data in Figure 2a, DRA 4510 has almost half the flux of NF99 but NF99 has less permeate flux compared to DRA 4510. This is because NF99 is a hydrophilic membrane, while DRA 4510 is a hydrophobic membrane. Hydrophilic membranes are capable of forming gravitational interactions between water and membrane materials such as dipole-dipole interactions, hydrogen bonds and ionic-dipole interactions [19].
The effect of feed flow rate on the rejection of sugars ( Figure 2b) and organic acids ( Figure 2c) in cranberry juice fruit was inappreciable and approximately 1.0 for sugar and above 0.9 for organic acids, with the exception being the G5 membrane. Perhaps the high rejection concealed the effect of feed flow rate. Figure 2 also shows the effect of feed flow rate on the benzoic acid separation from cranberry juice. From data in Figure 2b-d, it is apparent that rejection of benzoic acid was lower than for sugars and other organic acids. The reasonable explanation for this is that the molecular weight of benzoic acid is lower than the others. Thus, benzoic acid goes through membrane more easily. Simultaneously, results also indicate that cross flow velocity slightly affected the rejection of benzoic acid. Maybe this phenomenon is related to the contribution of convection and selective layers made from reversible adsorption on the surface of membranes into benzoic acid separation. At a feed flow rate of 3 to 4 L/min, the rejection of benzoic acid in UF-treated cranberry juice decreased, then it increased. This is explained by the UF treatment, which helps cranberry juice reject large compounds Data from Figure 2a also show that the permeation of UF cranberry juice was approximately two-fold higher than the one of fresh cranberry juice. The possible explanation for this is due to the absorption of high molecular weight compounds and the cake layer on a membrane's surface. This causes an increase in permeate resistance. Since the cake layer contributed to the separation, the presence of a reversibly adsorptive layer on the membrane surface also results in the higher rejection of benzoic acid in fresh cranberry juice than in UF-treated juice because the cake layer contributed to the separation. The reversibly adsorptive layer is constituted by interaction between high-molecular-weight compounds, such as anthocyanin and membrane material, which is attributed to molecular interactions, such as hydrophobic interactions and hydrogen bonds [21].

Influence of Temperature
The effect of temperature on the performance of cranberry juice nanofiltration is shown in Figure 3. It is apparent that the permeate flux increased and rejection of benzoic acid decreased with the increase in operating temperature (Figure 3a,c). Rejection of organic acid also slightly reduced. The same effect of temperature on solutes was also reported in some of the literature [22][23][24][25]. This can be explained by the expansion of the active layer structure because the higher temperatures increase the mobility of polymer chains, causing the membrane's porosity and pore size to increase. In addition, higher temperature tends to a more even distribution of organics between the solution and membrane phases, which means less selective partitioning, and as a result, lower rejection. Rejection of sugar did not change because it was concealed by high rejection (approximately 1.0). Although benzoic acid rejection at 40 • C was lower than that at 25 • C, the reduction was not large (Figure 3b). Moreover, nanofiltration at 40 • C spends more energy than at 25 • C. Thus, ambient temperature is suitable for separating benzoic acid by nanofiltration from cranberry juice.
pounds, such as anthocyanin and membrane material, which is attributed to molecular interactions, such as hydrophobic interactions and hydrogen bonds [21].

Influence of Temperature
The effect of temperature on the performance of cranberry juice nanofiltration is shown in Figure 3. It is apparent that the permeate flux increased and rejection of benzoic acid decreased with the increase in operating temperature (Figure 3a,c). Rejection of organic acid also slightly reduced. The same effect of temperature on solutes was also reported in some of the literature [22][23][24][25]. This can be explained by the expansion of the active layer structure because the higher temperatures increase the mobility of polymer chains, causing the membrane's porosity and pore size to increase. In addition, higher temperature tends to a more even distribution of organics between the solution and membrane phases, which means less selective partitioning, and as a result, lower rejection. Rejection of sugar did not change because it was concealed by high rejection (approximately 1.0). Although benzoic acid rejection at 40 °C was lower than that at 25 °C, the reduction was not large (Figure 3b). Moreover, nanofiltration at 40 °C spends more energy than at 25 °C. Thus, ambient temperature is suitable for separating benzoic acid by nanofiltration from cranberry juice.

Influence of pH
The influence of pH on the nanofiltration performance to separate benzoic acid from cranberry juice is shown in Figure 4. The permeate flux was slightly affected by pH value. The change in permeate flux relates to the change in pore size and electroviscosity. The change in pH leads to the change in charge on pore walls, which takes account of changes in electrostatic interaction, not only between charged groups in membrane materials but also between these groups and water. Consequently, these interactions lead to the increase in charge on the pore wall, pore size (because of swelling) [26] and electroviscosity [27]. While an increase in pore size increases permeate flux, the increase in electroviscosity makes permeate flux decrease. If the increase in electroviscosity is predominant, permeate flux decreases. On the contrary, if the increase in pore size is predominant, permeate flux increases.

Influence of pH
The influence of pH on the nanofiltration performance to separate benzoic acid from cranberry juice is shown in Figure 4. The permeate flux was slightly affected by pH value. The change in permeate flux relates to the change in pore size and electroviscosity. The change in pH leads to the change in charge on pore walls, which takes account of changes in electrostatic interaction, not only between charged groups in membrane materials but also between these groups and water. Consequently, these interactions lead to the increase in charge on the pore wall, pore size (because of swelling) [26] and electroviscosity [27]. While an increase in pore size increases permeate flux, the increase in electroviscosity makes permeate flux decrease. If the increase in electroviscosity is predominant, permeate flux decreases. On the contrary, if the increase in pore size is predominant, permeate flux increases.

Influence of pH
The influence of pH on the nanofiltration performance to separate benzoic acid from cranberry juice is shown in Figure 4. The permeate flux was slightly affected by pH value. The change in permeate flux relates to the change in pore size and electroviscosity. The change in pH leads to the change in charge on pore walls, which takes account of changes in electrostatic interaction, not only between charged groups in membrane materials but also between these groups and water. Consequently, these interactions lead to the increase in charge on the pore wall, pore size (because of swelling) [26] and electroviscosity [27]. While an increase in pore size increases permeate flux, the increase in electroviscosity makes permeate flux decrease. If the increase in electroviscosity is predominant, permeate flux decreases. On the contrary, if the increase in pore size is predominant, permeate flux increases.  The influences of pH on the rejection of sugars are shown in Figure 4b. From pH 2.5 to 4.5, the separation of sugars is approximately 1. The high rejection of sugar concealed the effect of pH in this range. The rejection tended to decrease slightly when pH increased from 4.5 to 5.5. This can be explained by the increase in pore size because of charge on the pore wall increasing [28,29].
From the data in Figure 4c, the rejection of organic acids tends to increase with an increase in pH, especially in the case of G5 membrane. This phenomenon relates to the repulsion between charged groups in the membrane and solutes. With the increase in pH value, the charge of membranes becomes more negative and organic acid also dissociates more. At pH 5.5, the dissociation of malic acid, citric acid and quinic acid is approximately 100%. Thus, it is difficult for these acids to move through membranes because of the repulsive force between negatively charged groups (on the membrane surface and pore wall) and negatively charged solutes (the dissociated organic acids). Consequently, they were still retained in the retentate side.
With regard to the separation of benzoic acid, from pH 2.5 to 3.5 the rejection decreased, then it increased with an increase in pH. There are some attributes contributing to the change in benzoic acid rejection by nanofiltration membranes under the effect of pH (Figure 4d). Firstly, the pH of cranberry juice was adjusted by NaOH. Thus, the content of sodium ion in the juice increased with the increase in pH. The presence of sodium From the data in Figure 2a, DRA 4510 has almost half the flux of NF99 but NF99 has less permeate flux compared to DRA 4510. This is because NF99 is a hydrophilic membrane, while DRA 4510 is a hydrophobic membrane. Hydrophilic membranes are capable of forming gravitational interactions between water and membrane materials such as dipole-dipole interactions, hydrogen bonds and ionic-dipole interactions [19].
The effect of feed flow rate on the rejection of sugars (Figure 2b) and organic acids (Figure 2c) in cranberry juice fruit was inappreciable and approximately 1.0 for sugar and above 0.9 for organic acids, with the exception being the G5 membrane. Perhaps the high rejection concealed the effect of feed flow rate. Figure 2 also shows the effect of feed flow rate on the benzoic acid separation from cranberry juice. From data in Figure 2b-d, it is apparent that rejection of benzoic acid was lower than for sugars and other organic acids. The reasonable explanation for this is that the molecular weight of benzoic acid is lower than the others. Thus, benzoic acid goes through membrane more easily. Simultaneously, results also indicate that cross flow velocity slightly affected the rejection of benzoic acid. Maybe this phenomenon is related to the contribution of convection and selective layers made from reversible adsorption on the surface of membranes into benzoic acid separation. At a feed flow rate of 3 to 4 L/min, the rejection of benzoic acid in UF-treated cranberry juice decreased, then it increased. This is explained by the UF treatment, which helps cranberry juice reject large compounds From the data in Figure 2a, DRA 4510 has almost half the flux of NF99 but NF99 has less permeate flux compared to DRA 4510. This is because NF99 is a hydrophilic membrane, while DRA 4510 is a hydrophobic membrane. Hydrophilic membranes are capable of forming gravitational interactions between water and membrane materials such as dipole-dipole interactions, hydrogen bonds and ionic-dipole interactions [19].
The effect of feed flow rate on the rejection of sugars (Figure 2b) and organic acids (Figure 2c) in cranberry juice fruit was inappreciable and approximately 1.0 for sugar and above 0.9 for organic acids, with the exception being the G5 membrane. Perhaps the high rejection concealed the effect of feed flow rate. Figure 2 also shows the effect of feed flow rate on the benzoic acid separation from cranberry juice. From data in Figure 2b-d, it is apparent that rejection of benzoic acid was lower than for sugars and other organic acids. The reasonable explanation for this is that the molecular weight of benzoic acid is lower than the others. Thus, benzoic acid goes through membrane more easily. Simultaneously, results also indicate that cross flow velocity slightly affected the rejection of benzoic acid. Maybe this phenomenon is related to the contribution of convection and selective layers made from reversible adsorption on the surface of membranes into benzoic acid separation. At a feed flow rate of 3 to 4 L/min, the rejection of benzoic acid in UF-treated cranberry juice decreased, then it increased. This is explained by the UF treatment, which helps cranberry juice reject large compounds The influences of pH on the rejection of sugars are shown in Figure 4b. From pH 2.5 to 4.5, the separation of sugars is approximately 1. The high rejection of sugar concealed the effect of pH in this range. The rejection tended to decrease slightly when pH increased from 4.5 to 5.5. This can be explained by the increase in pore size because of charge on the pore wall increasing [28,29].
From the data in Figure 4c, the rejection of organic acids tends to increase with an increase in pH, especially in the case of G5 membrane. This phenomenon relates to the repulsion between charged groups in the membrane and solutes. With the increase in pH value, the charge of membranes becomes more negative and organic acid also dissociates more. At pH 5.5, the dissociation of malic acid, citric acid and quinic acid is approximately 100%. Thus, it is difficult for these acids to move through membranes because of the repulsive force between negatively charged groups (on the membrane surface and pore wall) and negatively charged solutes (the dissociated organic acids). Consequently, they were still retained in the retentate side.
With regard to the separation of benzoic acid, from pH 2.5 to 3.5 the rejection decreased, then it increased with an increase in pH. There are some attributes contributing to the change in benzoic acid rejection by nanofiltration membranes under the effect of pH (Figure 4d). Firstly, the pH of cranberry juice was adjusted by NaOH. Thus, the content of sodium ion in the juice increased with the increase in pH. The presence of sodium ion can cause the swelling in membrane, which affects mechanical properties of the membrane simultaneously, altering their ability to recover adsorbed substances [14,30,31]. Thus, solutes can move through membrane more easily and cause a decrease in the rejection. Moreover, the augmentation of sodium content in the feed leads to an increase in sodium content in the permeate side and reinforces movement of the dissociated benzoic acid through the membrane (Donnan effect) [32].
Secondly, this could be due to the effect of pH on the dissociation. The dissociation of organic acid increases with an increase in pH value (Figure 4c). At pH 2.5, only 2% of benzoic acid is dissociated. However, at pH 5.5, 95% of benzoic acid is dissociated. The investigated membranes are made from polyamide or a mixture of amide and the others. Polyamide is amphoteric and its charge depends on pH. The charge is positive with a pH lower than the isoelectric point (pI) and negative with a pH higher than pI. The pI of polyamide membranes often ranges from 4.0 to 6.0 [33,34]. If charge of membrane is positive, then this reinforces dissociated acids to move through the membrane. Conversely, the negative charge in the membrane repulses dissociated acids and increases the rejection. Therefore, from pH 2.5 to 3.5, the charge of membrane was positive and reinforced dissociated benzoic acid to move through the membrane and make rejection decrease. However, with pH rangng from 4.5 to 5.5, membrane charge may be negative, consequently, the repulsive force between membrane and dissociated benzoic acid increased making rejection increase.
Finally, as stated above, changes in pH can lead to a change in pore size [35]. The intensity of influence depends on the intrinsic membrane, such as its material or structure. These changes in pore size have obvious effects on rejection.
From pH 2.5 to 3.5, the factors which reinforced the movement of benzoic acid through membranes might be predominant. Therefore, the rejection of benzoic acid decreased. However, with the higher pH, the factors which hinder the transport of benzoic acid through nanofiltration membranes were predominant, especially the repulsive force between membrane and solutes. Consequently, rejection increased with an increase in pH value.
The results also showed that DK and NF99 membranes were the most suitable to separate benzoic acid from cranberry juice because of the high performance of nanofiltration.There were larger differences between rejections of benzoic acid and the others in cranberry juice, and high permeate flux. Besides, the pretreatment by ultrafiltration can improve permeate flux in nanofiltration.
The investigation into the influence of permeate flux on the rejection of benzoic acid from cranberry juice was carried out with NF99 and Desal DK membranes, and the results are shown in Figure 5c,d. The results showed that benzoic acid rejection increased with increases in permeate flux and tended to reach a critical rejection with both NF99 and Desal DK membrane at investigated pH values. As stated above, the transport of solutes through nanofiltration membranes is conducted by three mechanisms: convection, diffusion and electromigration. The proportion of their contribution depends on the attributes of the membrane (for example, material, structure, electrical property), solutes and operative conditions. Based on the extended Nernst-Planck equation and experimental data which investigated model solutions, many authors showed that rejection increased with increases in permeate flux and reached a critical value, so-called reflection rejection, when permeate flux moved towards the infinite, because of the domination of convection [29,33,34,36]. sion and electromigration. The proportion of their contribution depends on the attributes of the membrane (for example, material, structure, electrical property), solutes and operative conditions. Based on the extended Nernst-Planck equation and experimental data which investigated model solutions, many authors showed that rejection increased with increases in permeate flux and reached a critical value, so-called reflection rejection, when permeate flux moved towards the infinite, because of the domination of convection [29,33,34,36].

Influence of Permeate Flux
Data in Figure 5c,d also showed that, at low permeate flux and high pH, the rejection of benzoic became negative for the NF99 and DK membranes. This result indicates that, at low permeate flux, diffusion and electromigration considerably contributed to the transport of benzoic acid through investigated nanofiltration membranes. This result accords with those of Szymczyk et al., 2003, obtained by solving theoretical equations [37]. In the case of our study, this can be explained by the influence of sodium ion on the separation of benzoic acid by nanofiltration membranes. When pH was adjusted by NaOH, the sodium content in feed increased with an increase in pH. Consequently, the content of sodium ion in the permeate side increased, and dissociated benzoic acid has to move more to permeate the side to neutralize electricity. Therefore, the content of benzoic acid in the permeate side could become greater than in the retentate side and rejection was negative. However, more research needs to be undertaken to more clearly investigate the transport of benzoic acid through nanofiltration membranes.
Rejections of sugar and organic acid under the effects of pH were also observed ( Figure 6). In both NF99 and Desal DK membranes, rejection of sugars and organic acids was over 0.9 and increased with increases in operation pressure.

Influence of Permeate Flux
Data in Figure 5c,d also showed that, at low permeate flux and high pH, the rejection of benzoic became negative for the NF99 and DK membranes. This result indicates that, at low permeate flux, diffusion and electromigration considerably contributed to the transport of benzoic acid through investigated nanofiltration membranes. This result accords with those of Szymczyk et al., 2003, obtained by solving theoretical equations [37]. In the case of our study, this can be explained by the influence of sodium ion on the separation of benzoic acid by nanofiltration membranes. When pH was adjusted by NaOH, the sodium content in feed increased with an increase in pH. Consequently, the content of sodium ion in the permeate side increased, and dissociated benzoic acid has to move more to permeate the side to neutralize electricity. Therefore, the content of benzoic acid in the permeate side could become greater than in the retentate side and rejection was negative. However, more research needs to be undertaken to more clearly investigate the transport of benzoic acid through nanofiltration membranes.
Rejections of sugar and organic acid under the effects of pH were also observed ( Figure 6). In both NF99 and Desal DK membranes, rejection of sugars and organic acids was over 0.9 and increased with increases in operation pressure. more to permeate the side to neutralize electricity. Therefore, the content of benzoic acid in the permeate side could become greater than in the retentate side and rejection was negative. However, more research needs to be undertaken to more clearly investigate the transport of benzoic acid through nanofiltration membranes.
Rejections of sugar and organic acid under the effects of pH were also observed ( Figure 6). In both NF99 and Desal DK membranes, rejection of sugars and organic acids was over 0.9 and increased with increases in operation pressure.

Conclusions
The influence of technical parameters on the separation of benzoic acid from cranberry juice by cross-flow nanofiltration with DSS "Lab Module Type 20" plate and frame system was investigated. The pretreatment by UF made permeate flux two-fold higher and separation lower. In the range from 2 L/min to 7.5 L/min, the effect of feed flow rate on performance of separation was slight. Higher temperatures led to higher permeate and lower rejection of benzoic acid, and the suitable temperature for separation of benzoic acid from cranberry juice was ambient. pH strongly affected the performance of nanofiltration, especially benzoic acid rejection, and the lowest rejection of benzoic acid was observed at pH 4.5. When compared to other membranes, the UTC 60 has a lower efficiency in separating benzoic acid from cranberry juice. As the rejection of reducing sugar and organic acid is not significantly different between membranes, rejection of benzoic acid of UTC60 is higher. As a result, the recovery efficiency and purity of benzoic acid are lower. With NF99 and DK membranes, in suitable conditions, the rejection of benzoic acid can be negative; this indicated that benzoic acid has high permeability and the permeate flux is low, while a reduction in sugar and organic acid is retained in the retentate flow. The results showed that cross-flow nanofiltration with NF99 and Desal DK membranes can be applied for the effective separation of benzoic acid from cranberry juice.

Conclusions
The influence of technical parameters on the separation of benzoic acid from cranberry juice by cross-flow nanofiltration with DSS "Lab Module Type 20" plate and frame system was investigated. The pretreatment by UF made permeate flux two-fold higher and separation lower. In the range from 2 L/min to 7.5 L/min, the effect of feed flow rate on performance of separation was slight. Higher temperatures led to higher permeate and lower rejection of benzoic acid, and the suitable temperature for separation of benzoic acid from cranberry juice was ambient. pH strongly affected the performance of nanofiltration, especially benzoic acid rejection, and the lowest rejection of benzoic acid was observed at pH 4.5. When compared to other membranes, the UTC 60 has a lower efficiency in separating benzoic acid from cranberry juice. As the rejection of reducing sugar and organic acid is not significantly different between membranes, rejection of benzoic acid of UTC60 is higher. As a result, the recovery efficiency and purity of benzoic acid are lower. With NF99 and DK membranes, in suitable conditions, the rejection of benzoic acid can be negative; this indicated that benzoic acid has high permeability and the permeate flux is low, while a reduction in sugar and organic acid is retained in the retentate flow. The results showed that cross-flow nanofiltration with NF99 and Desal DK membranes can be applied for the effective separation of benzoic acid from cranberry juice.