Color reduction of raw sugar syrup using hydrogen peroxide

A commercial H2O2 solution (35%, v/v) was evaluated as a clarifying agent for raw, type VHP (very high polarization) sugar syrup, using an experimental design applying artificial neural networks (ANN). Fifteen experimental runs were carried out and the samples were taken at the following time intervals: 0, 5, 10, 15, 20, 25, 30, 40, 50, 60, 75 and 90 min. The treatments were carried out using an experimental design consisting of three variables: H2O2 (X1: 0; 202.4; 500; 797.6 and 1.000 mg L-1); pH (X2: 3.32, 5, 7.5, 10 and 11.68) and temperature (X3: 16.4, 30, 50, 70 and 83.6°C). The theoretical and measured values fitted the analysis by artificial neural networks (ANN) well. A reduction in colour (ICUMSA method) was observed between 60 and 75 min, except for treatments # 11 (pH= 11.68; 50°C and 500 mgH2O2 L-1), # 13 (pH=7.5; 83.6°C and 500 mgH2O2 L-1) and #15 (pH= 7.5; 50°C and 1.000 mgH2O2 L-1), which showed a colour reduction after 30 min. In the treatments at pH 3.32 or 11.68, temperature of 83.6°C and H2O2 dose of 1.000 mg L-1, an average sucrose degradation of 55% was observed. The best colour reduction result was obtained with treatment # 9 (pH 7.5, 50 °C and 500 mgH2O2 L-1), although sucrose degradation of 26% was observed.


Introduction
Hydrogen peroxide (H2O2) has a standard reduction potential (E 0 ) of 1.77 V, higher than that of other oxidants such as chlorine (E 0 = 1.36 v), chlorine dioxide (E 0 = 1.50 v), molecular oxygen (E 0 = 1.23 v) and potassium permanganate (E = 1.67 v). In a reaction medium, H2O2 can be converted to hydroxyl radicals (•OH), a highly reactive species (E 0 = 2.80 V) (Üzer et al., 2017;Mattos et al., 2013), making it of interest in the treatment of drinking water and industrial effluents (Guo et al., 2017;Russo et al., 2017;Spasiano et al., 2016). Stout et al. (2017), Jervis et al. (2015) and Fairbanks (2009) reported that hydrogen peroxide has been used in the bleaching of different types of food. Since 1979, H2O2 has been recognized as a GRAS (Generally Recognized as Safe) product by the Food and Drug Administration (FDA, 2015).
Studies on the use of H2O2 in sugarcane juice or sugar beet clarification have been carried out (Mane et al., 2000;Sartori et al., 2015a;Sartori et al., 2015b;Mandro et al., 2015), and the action of H2O2 on compounds such as melanoids, melanins, caramels, polyphenols, starch and amino acids, which are present during the processes used to obtain sugar, reported. H2O2 can also act during the purification/clarification step of Very High Polarization (VHP) or Very Very High Polarization (VVHP) raw sugar syrup during sugar refining . Nowadays, the sugar refining process consists of the dissolution of crystal sugar in hot water, and its passage through ion exchange resins and activated carbon to remove the impurities (Crema, 2012). The high costs are related to the regeneration pof the resins, due to a progressive loss of ion exchange capacity (Rodrigues, 1998;Baccar et al., 2009).
Thus, by way of an analysis in artificial neural networks, this study aimed to evaluate the clarification of sugar syrup (VHP type) using H2O2 as the clarification agent.

Sugar syrup preparation
The sugar syrup was prepared by dilution in ultrapure water (18 MΩ cm) up to a final concentration of 66 Brix (% soluble solids). Each treatment used 400 mL of 66 Brix syrup, whose concentration was measured in a refractometer Mod. RFM-712 (Bellingham+Stanley Co., UK).

Treatment with hydrogen peroxide
The conical flasks (500 mL) were maintained under different reaction conditions depending on the experimental design: temperature (16.4 to 83.6 °C); pH (3.32 to 11.68) and H2O2 doses (between 0 and 1000 mg L -1 ) ( Table 1). Each experimental run was monitored for 90 min, and samples were collected at the following time intervals 0,5,10,15,20,25,30,40,50,60,75 and 90 min. Colour reduction was analysed by the ICUMSA method and the sucrose contents by HPLC. The results were expressed as the average ± standard deviation of three analytical replicates. Bovine liver catalase (freeze-driedpowder, 2000 -5000 U/mg protein; Sigma-Aldrich) was added at the end of each treatment to disrupt the oxidation reactions of the hydrogen peroxide.

ICUMSA colour analysis of the sugar syrups
The colour analyses of the samples treated with hydrogen peroxide were carried out using method GS1/3-7 (International Commission for Uniform Methods of Sugar Analysis, 2011). Initially, the sample Brix was adjusted to 30 Brix using a bench refractometer and they were then vacuum filtered through a PTFE 0.45 μm ϕ pore membrane (Millipore Co., Brazil). The pH was then adjusted to 7±0.05 with HCl 0.1 mol L -1 or NaOH 0.1 mol L -1 and the absorbance measured in a UV-Mini 1240 spectrophotometer (Shimadzu Co., Japan) at 420 nm. The ICUMSA colour of the samples was expressed as the result obtained from the Expression (1): Where: ABS: absorbance at 420 (nm); b: Cell optical path (cm); c: Sucrose concentration (g mL -1 ) in solution in Brix at 20 °C.

HPLC-ELSD analysis of sucrose contents
Sucrose was determined by HPLC equipped with a low temperature evaporative light scattering detector (ELSD-LT). The mobile phase consisted of acetonitrile and water (85:15, v/v) previously filtered through a 0.45 µm pore diameter PTFE membrane. The samples were passed through a Kromasil 100 NH2-50 (250 mm × 46 mm, 5 µm) column, maintained at 30 °C, under the following analytical conditions: flow rate of 1.0 mL min -1 , detector temperature of 35 °C and nitrogen as the nebulizer gas at a pressure of 350 kPa. The sucrose concentrations (injection volume = 5 µL) were determined in triplicate by comparison with a calibration curve for sucrose of from 0.1 to 0.5 g L -1 (Sigma-Aldrich, HPLC grade).

Data analysis of the sugar syrup treated by hydrogen peroxide
An artificial neural network (ANN) was structured for the empirical prediction of the results. In the input layer, the values added were divided between the dependent (ICUMSA colour responses and sucrose contents of the sugar syrup samples) and controllable independent values (temperature (°C); pH; hydrogen peroxide doses (mg mL -1 ) and Brix). The validation test of the networks was carried out using 4 to 10 neurons. The software used for data preparation, adjustment and simulation of the neural networks was developed at CESQ/DEQ-EPUSP (Nascimento et al., 2000) and the relative importance of the variables was assessed using the Holdback Input Randomization Method (HIPR) proposed by Kemp et al. (2007). This method is based on a comparison of the errors caused in each output of the model by random disturbances imposed on each of the input variables.

Results and discussion
The reductions of the ICUMSA colour and sucrose contents of the sugar syrup samples were analysed with an ideal number of neurons for each treatment by way of the Learning-set (LS) and Test-set (TS). The values obtained for the angles and determination coefficients of the Learning-set (LS) and Test-set (TS) from 4 to 10 neurons can be seen in Figures 1a and 1b, respectively. For the ICUMSA colour and sucrose, the value closest to 1 for the Learning-set (LS) and Test-set (TS) coefficients was 8 neurons. Thereby, a representation of 8 neurons was chosen for a better fit of the artificial neural networks (ANN). The simulated values were similar to those obtained in the experiment. The artificial neural networks obtained good adjustment and were able to predict the results of the ICUMSA colour ( Figure 2). Treatments #9 (pH = 7.5; 50 °C and 500 mgH2O2 L -1 ) and #12 (pH = 7.5; 16.4 °C and 500 mgH2O2 L -1 ) presented greater similarity between the real and simulated values ( Figure 2). Treatment #9 showed a greater reduction in ICUMSA colour (82% reduction) (Table 2), reaching its peak between 60 and 75 min. On the other hand, treatment #14 (pH= 7.5; 50 °C; 0 mgH2O2 L -1 ) showed the lowest reduction in ICUMSA colour, obviously due to the absence of the clarifying agent.  Treatments #11 (pH = 11.68; 50°C and 500 mgH2O2 L -1 ), #13 (pH = 7.5; 83.6 °C and 500 mgH2O2 L -1 ) and #15 (pH = 7.5; 50°C and 1.000 mgH2O2 L -1 ) showed greater reductions in ICUMSA colour up to 30 min.
According to Lange et al. (2006), the pH of the reaction is important for H2O2 stability, because the speed and efficiency of oxidation are affected since H2O2 rapidly decomposes producing oxygen and water in alkaline pH values.
The constant speed for each sugar syrup treatment required to cause sucrose degradation was calculated in 15 min reactions of sugar syrup with H2O2.
Of the treatments, #9 ( Figure 5) presented the best colour × sucrose relationship, because it showed greater colour reduction (about 82%) with lower sucrose degradation (about 33%). In all the treatments analysed, the results were similar when comparing the theoretical and experimental values, and the best fits were observed for treatments #8 (pH= 10.0; 70 °C and 797.6 mgH2O2 L -1 ) and #9 (pH= 7.5; 50 °C and 500 mgH2O2 L -1 ). The calculation of the relative importance of the variables showed that the factor with the greatest effect on sucrose degradation was the initial Brix, with a percentage of 54.38% (Figure 6). For Sartori et al. (2017), in studies with sugarcane juice, sucrose was the most abundant soluble component in sugarcane juice and thus, any changes in Brix values affect the sucrose concentration. Hydrogen peroxide represented 17.50% followed by pH with 15.31% and temperature with 12.79%.

Conclusion
In all cases an ICUMSA colour reduction occurred in the sugar syrup after treatment; however, it was more pronounced under conditions of pH = 7.5; 50 °C and 500 mgH2O2 L -1 . The initial Brix was indicated as the variable with the greatest influence on ICUMSA colour reduction and sucrose degradation. The profile of sucrose degradation up to 90 min showed that the first 15 min were the most critical for degradation. The models presented a good fit, with the simulated values very close to the experimental values.