Exploring the nanofiltration mass transfer characteristic and concentrate process of procyanidins from grape juice

Abstract In order to separate procyanidins from grape juice at room temperature, a separation prediction model was established based on nanofiltration. The mass transfer coefficient was positively correlated with the initial concentration. Nanofiltration performance of procyanidins was affected by filtration conditions, membrane properties, and molecular states. The correlation between mass transfer coefficient and initial concentration was established based on the linear equations of the rejection and mass transfer coefficient. The rejection of procyanidins predicted with the mass transfer model was in accordance with the experimental value, and the antioxidant activity was preserved effectively. The mathematical model could predict the rejection of procyanidins. The nanofiltration technology for procyanidin separation from grape juice was characterized by fast separation, low energy consumption, and zero oxidization loss. The nanofiltration technology could greatly improve the utilization efficiency of food products and decrease the energy consumption.

not discussed. In order to clarify the relationship between membrane transport mechanisms and molecular state, the mass transfer mathematic model was fitted and verified based on the solution-diffusion effect and Donnan steric partitioning pore model (Pérez, Escudero, Arcos-Martínez, & Benito, 2016;Wang et al., 2012). Procyanidins were selected as the indication of phenolic compounds in grape juices to evaluate the performance of a NF membrane under different concentrations and pH. The prediction model of nanofiltration separation provides the prediction basis for nanofiltration separation, especially for functional food with phenolic compounds.

| Preparation of grape juice
Fresh grapes were obtained from a local market, which were from Pakwachow Island in Nanjing. The grapes were washed with purified water, and fruit branches were cut off. The grapes were processed in a commercial juicer to yield the natural juice. The natural juice was kept at 4-7°C to prevent damage or degradation.

| Microfiltration pretreatment
In order to improve the clarity, grape juice was pretreated by microfiltration to remove suspended solids. In the microfiltration, a poly-

| Procyanidin content
The content of procyanidins was determined with an Agilent 1,100 HPLC system equipped with a reverse-phase column (Agilent C 18 , 4.6 mm Ø × 250 mm) at 30°C and a UV-visible detector (λ = 280 nm).
An isocratic mobile phase of 0.4% aqueous phosphoric acid: acetonitrile (15:85, ml/ml) mixture was used under a flow rate of 0.8 ml/ min. The injection volume was 10 μl. For the quantitative analysis, a standard calibration curve was obtained by plotting the peak area against different concentrations (5, 10, 50, 150, 300 µg/ml) of procyanidin standard compound. The curves showed a good linearity and followed Beer's Law (r 2 = 0.9987). Similarly, the final concentration of compounds in the samples in three consecutive injections was determined as the average content.

| The Nanofiltration system and operations
A laboratory bench scale cross-flow NF apparatus was used in all experiments. The apparatus consisted of a NF membrane, one variable-speed gear pump (Model JDB-12A, Tuozhu Corporation, China) for pressure and recirculation, a digital pressure gauge (Mettler Toledo, Germany) for the measurement of operating pressure, and tubings. The model of NF membrane was spiral, and the mode code was NFG-2B-1812. NF was carried out by using polyamide membrane with a molecular weight cutoff of 800 Da, a filtration area of 0.30 m 2 , max cross-flow operation pressure of 3.0 MPa, PEG800 minimum rejection of 95.0%, and permeation flux of 76.5-93.5 L/(m 2 h) (Synder Filtration, USA).
In order to ensure that the separation performance of the membranes was not changed during filtration experiments, first, remaining water was pumped from the NF apparatus. Second, microfiltered grape juice was used in NF system. Testing pressures were 0.2, 0.4, 0.6, 0.8, 1.0, and 1.2 MPa, and the permeate flux (J) was regulated by the variable-speed gear pump. The pipeline of feed solution, filtrate, and rejected solution was placed in the same tube. Before sampling analysis, membrane module was pressurized at the test pressure for minimum 2 hr to reach the steady-state conditions. When the adsorption-desorption equilibrium between solutes and membrane was reached, the concentrations of the feed and permeate were analyzed with high-performance liquid chromatography (Agilent 1100, USA). And the rejection was calculated according to Equation (1) (Qiu & Yang, 2010).
where C f and C p are the solute concentrations in feed and permeate solution. Each measurement was performed in triplicate.

| Nanofiltration separation prediction model
The solution-diffusion model of the NF assumes that the solute contacts the solvent and is dissolved on the membrane surface (Murthy & Gupta, 1997;Geraldes, Semião, & Pinho, 2001). Then, the solute passes through NF membrane pore under chemical potential differences. The model can be expressed as: (2)

Practical application
Thermal breakage of phenolic ingredients was a common problem to which attention should be paid in the application of food and chemistry industry. It has been evidenced that NF separation was an effective technique for the concentrate of procyanidins from the grape juice. Given today's green separation demand over the world, it is important for the researchers to understand this method and its benefits for food and chemistry industry.
where J V is permeate flux, L/(m 2 h); L p is the pure water permeability, L/(m 2 h Pa); p is operating pressure, Pa; Δπ is the osmotic pressure difference across the membrane, Pa; K is partition coefficient; δ is membrane thickness, cm; DK/δ is the mass transfer performance of a membrane, cm/s; N A is the volume flux of solute, mol/(cm 2 s); C m is solute concentrations in NF membrane surface, mol/L.
The rejection of solutes can be divided into apparent rejection R o and real rejection R r , which can be, respectively, expressed as: C o is the original solute concentration. Based on the solution-diffusion model and Equation (2) -Equation (5), the relationship between R o and mass transfer coefficient k can be expressed as: According to Equation (6)

| Antioxidant activity determinations
Antioxidant activity is one of the important indexes to evaluate the quality of procyanidins. The ABTS method was selected to detect the antioxidant activity of samples (Arend et al., 2017). The ABTS radical-scavenging activity of the samples was measured by the method described by Sachindra (Sachindra et al., 2007). ABTS radical solution was prepared by mixing 5 ml of ready-to-use ABTS solu-

| Nanofiltration separation prediction model verification
The procyanidin concentration in a new grape juice sample was detected by high-performance liquid chromatography (Agilent 1100, USA). A series of procyanidin concentrations were treated in the NF system under the operating pressures of 0.2, 0.4, 0.6, 0.8, 1.0, and 1.2 MPa to establish the NF separation prediction model. The k was calculated by the equations of Table 2 with the series of procyanidin concentrations, and then, the predicted rejections were fitted by Equation (6) with the value of k.
The experimental R o was calculated according to Equation (1) and compared with the predicted value to analyze the applicability of NF separation prediction model.

| Membrane morphology analysis
The NF membrane was washed with 25 mmol/L sodium hydroxide aqueous solution to remove contaminants. The polluted and cleaned membranes were detected by scanning electron microscope (SEM).
From ZEISS MERLIN Compact ultra-high-resolution field emission scanning electron microscopy, test parameters were Mag 20.00 kx, WD 7.2 nm, EHT 10.00 kv, and scale 1 µm. Prior to SEM analysis, the membrane samples were air-dried and subsequently coated with an ultrathin layer of carbon. Extreme care was taken when preparing the fouled and scaled membrane samples to ensure that the fouling and scaling layer remained intact.

| Nanofiltration Permeate flux
The membrane permeate flux directly relates to the production ef- (4)  The effects of different pH on membrane permeate flux were analyzed (Figure 1). Procyanidins in grape juice existed in two states: ionic status and dissociative state. With the increase in pH, the proportion of the dissociative state increased accordingly, but the membrane permeate flux decreased because it was difficult for the solutes in the ionic status to pass through the NF membrane due to the charge effect (Ryzhkov & Minakov, 2016).

| Effect of operating pressure on rejection
When the operating pressure of NF increased from 0.2 to 1.2 MPa, the rejections increased insignificantly. Meanwhile, the membrane flux was increased linearly on the whole and the NF concentration efficiency (the amount of water removed per unit time) was increased (Nakari, Pihlajamäki, & Mänttäri, 2016).

| Effect of concentration on rejection
The effects of different concentrations from 10 to 200 µg/ml on the rejections during NF process were investigated. The rejection of procyanidins slowly decreased with the increase in the concentration ( Figure 2). This result was consistent with the solution-diffusion theory (Paul, 2004;Wijmans & Baker, 1995). In NF process, procyanidin molecules accumulated in the boundary layer, so the local concentration of procyanidins in the boundary layer was much higher than that in the bulk. The increase in the solute concentration increased the permeable pressure difference and the solute could pass through the membrane pores, thus resulting in the decrease in the rejection. The solution-diffusion effect was increased under higher concentrations, which enhanced the membrane pollution and greatly affected the further separation. The solute rejection increased with the increase in the solution pH due to the Donnan effect between procyanidins and membrane surface charge.

| Fitting mass transfer model
The correlation between J V and ln[(1-R o )·J v /R o ] was fitted by Equation (6) Table 1 (pH 3.0). The k of procyanidins increased with the concentration. The tendency was consistent with the solution-diffusion theory.
As the dissociative state of procyanidins was transformed into the dissociative and ionic coexistence, the rejections were changed dynamically. Dissociative procyanidins have the priority to enter NF membrane interface and are then dissolved to pass through NF membrane pores under the intermembrane pressure difference.
Ionic procyanidins with NF membrane showed the charge effect (Table 1), and it was difficult for ionic procyanidins to pass through NF membrane, thus decreasing the mass transfer coefficient of procyanidins (pH 5.5). With the increase in the procyanidin concentration, the mass transfer coefficient was increased accordingly due to the solution-diffusion effect and the charge repulsion effect (Weng et al., 2016).
The NF membrane surface carries negative charge (Synder Filtration, USA). It is difficult for procyanidin anions to pass through the NF membrane due to the electrostatic repulsion between anions and the NF membrane. Therefore, the rejection increased accordingly. Based on the data in Table 1, at pH 8.0, the mass transfer coefficient of procyanidins in the ionic state was significantly lower than that of procyanidins in the dissociative state.
In addition, ln[DK/δ] value was independent of the initial concentration of procyanidins, but it was related to the existence state of procyanidins.

| Model verification
The correlation between the procyanidin concentration and k was Therefore, it was difficult for procyanidin molecules to penetrate the membrane.
At pH 8.0, the experimental rejection was slightly lower than the predicted value, unlike the results at the pH 3.0 and 5.5. Ionic state was the main existence state of procyanidins in solution, but its ionization level might be lower than other phenolic acids. Therefore, it is easy for the procyanidin molecule to penetrate the membrane.

| Antioxidant activity determination
ABTS assays indicated that the antioxidant activity in the concentrate was significantly increased (p < 0.01). The concentration factor of procyanidins was 3.8. NF could efficiently separate the main bioactive compounds including phenolic compounds from grape juice.
Phenolic compounds determine the quality of grape juice.

| Membrane fouling of grape juice
The morphology of the scaling layer confirmed the deposition of grape juice on the membrane surface ( Figure 5a). The cake layer of grape juice on the membrane surface was easily cleaned (Figure 5b), and the membrane flux increased rapidly with washing time, suggesting that membrane fouling has redissolved. The good separation of grape juice was achieved by polyamide NF membranes, while the concentration efficiency was maintained.

| CON CLUS ION
Heat-sensitive ingredients were concentrated by NF technology at normal temperature (15-27°C). The interaction force between procyanidins and NF membrane mainly involved the solution-diffusion effect and charge repulsion. Therefore, the rejection of procyanidins could be adjusted by changing the existence states of procyanidins.
Mass transfer model was established on the basis of the solutiondiffusion theory and Donnan effect to demonstrate the relationships between rejection and molecular existence parameters. C o and pH were the main factors of the rejection. The k of procyanidins was directly related to the concentration under a fixed pH.
In recent years, NF studies were focused on the separation of glucose and ionic components based on the solution-diffusion theory and Donnan effect (Pérez et al., 2016;Wang et al., 2012). The exploration of the mass transfer mechanism for the NF separation of procyanidins provides the basis for improving NF separation of procyanidins. In the mass transfer model, electrical properties of procyanidins are the important parameter of k. The NF membrane surface carries negative charges. It is difficult for procyanidin anions to pass through the NF membrane due to the electrostatic repulsion between anions and the NF membrane. Therefore, the mass transfer coefficient increased accordingly. Then, the targeted rejection can be achieved by changing the existence state, concentration, and operating pressure.
Nanofiltration separation is an effective technique for the concentration of procyanidins from grape juice. In addition, the NF technology increases the utilization of agricultural products greatly and decreases the energy consumption.

CO N FLI C T O F I NTE R E S T
The authors have declared no conflict of interest.

E TH I C A L A PPROVA L
The study did not involve any human or animal experimentation.