Optimization of the Electrospinning Conditions by Box-Behnken Design to Prepare Poly(Vinyl Alcohol)/Chitosan Crosslinked Nanofibers

Electrospun poly(vinyl alcohol)/chitosan nanofibers had their solution and process parameters optimized using a Box-Behnken design and desirability function. Four factors (applied voltage, flow rate, distance tip-to-plate and amount of chitosan) were varied to produce electrospun mats with a low fiber diameter. An empirical model was developed for each response using response surface methodology (RSM), which revealed that flow rate had no significant influence on the assessed responses. With desirability function, the optimal conditions to produce the nanofibers were applied voltage of 13.1 kV, 30% chitosan concentration and distance tip-to-plate of 10 cm. The fiber diameter and standard deviation were 196.5 ± 28.3 nm, compared to the predicted values of 185.9 ± 26.8 nm. The desirability function allied with Box-Benhken design proved themselves important tools to predict process parameters for the development of nanofibers. The mats were crosslinked with glutaraldehyde for 24 h and 48 h and presented good water stability and enhanced mechanical properties.


Introduction
Electrospinning is a promising and versatile technique that is used to fabricate polymeric nanofibers for a wide variety of biomedical applications such as drug delivery systems [1], scaffolds for tissue engineering [2] [3] [4] [5], wound The aim of DoE is to maximize or minimize the responses of interest using a small number of experiments even with a wide range of factors [28]. Box-Behnken is a design used for the adjustment of quadratic functions. It is considered an economical and efficient tool for determining first and second order coefficients of regression models [29]. BBD is a 3-level (3k) statistical design where k is the number of factors to be tested. Its experimental matrix eliminates the vertices points of a cube and it is not possible to carry out experiments under extreme conditions. In other words, it is not possible to perform an experiment with all factors at the maximum or minimum level simultaneously, which may prove to be an advantage in some cases [30] [31]. Its experimental matrix eliminates the vertices points of a cube and it is not possible to carry out experiments under extreme conditions, that is, all factors at the maximum or minimum level simultaneously, which may prove to be an advantage in some cases [30] [31]. The use of Box-Behnken design in the production of nanofibers has been reported by several authors in order to understand the influence of solution and process parameters on the formation of materials, as well as the optimization of significant factors [32] [33] [34] [35].
Response Surface Methodology (RSM) is a mathematical technique that, when allied with DoE, is ideal for constructing models and find optimal conditions for one or more responses. One of the main advantages of RSM is the ability to exclude insignificant factors and interactions from the regression model [28] [31] [36]. In situations where the best conditions of a process are to be determined, the desirability method becomes a very useful tool for the simultaneous optimization of multiple responses. Thus, it is possible to obtain the best conditions of the multiple independent variables simultaneously, both to maximize and to minimize or even to obtain specific nominal values [37] [38]. For this, the expected result will depend on the desired objective (maximization, normalization or minimization), the specified limits and the weight assigned to each of the responses [39].
This work aims to investigate the influence of process parameters on the morphology of PVA/CS electrospun nanofibers. A four-factor three-level Box-Behnken design was used to determine the better conditions of electrospinning processing to optimize the production of PVA/CS nanofibers.

Rheological Characterization of Electrospinning Solutions
The rheological properties of the solutions were investigated at 25˚C within the linear viscoelastic range, at 6.28 rad•s −1 , on rheometer AR-G2 (TA Instruments, New Castle, DE, USA) using a concentric cylinder geometry. The variation of the storage modulus (G') and the loss modulus (G''), as a function of the oscillatory frequency, in the range from 0.1 to 100 rad•s −1 was investigated.

Preparation of Electrospun and Crosslinked Mats
The environmental conditions used for electrospinning were 25˚C and 50% -

Box-Behnken Design (BBD) of Electrospinning Process
Box-Behnken design was used to verify and identify the effect of 4 process parameters (applied voltage, solution flow, chitosan concentration in the mixture and distance between the needle and the collecting plate) on the mean fiber diameter (Y 1 ) and in the standard deviation of the fiber diameter (Y 2 ). In the present study, the four-factor, three-level Box-Behnken experimental design was applied to determine optimal conditions to minimize the mean diameter and standard deviation of nanofibers. Factor levels were coded as 1 (low), 0 (center point), and 1 (high) as shown in Table 1.
In a system involving four significant independent variables (X 1 , X 2 , X 3 and X 4 ), the mathematical relationship of the responses about these variables may be given approximately by the quadratic polynomial Equation (1). Tip to needle distance (cm) 10 12. 5 15 where β 0 is the offset term, β i is the slope or linear effect of input factor X i , β ii is the quadratic effect of input factor X i and β ij is the linear-linear interaction effect between input factors X i and X j .
All experiments were performed in a random order to minimize the effect of unexpected variability on observed response due to external factors. Statistica 12.0 software was used for all statistical calculations.

Optimization of BBD by Desirability Function
Initially, each response (Y 1 and Y 2 ) was transformed into a dimensionless desirability function (d i ) within a range from 0 to 1 (the lowest and highest desirability). The value of d i increases as the i-th response approaches the limits imposed.

Water Stability
The nanofibers were soaked in distilled water and after 1 h they were carefully removed and dried with filter paper to remove surface water and then weighed. The materials were dried under vacuum to constant weight.

Characterization of PVA/CS Mats
The morphology of PVA/CS nanofibers was observed by scanning electron microscopy (SEM), using a Tescan model 212 Vega 3LMU (Brno-Kohoutovice, Czech Republic) equipment. The surfaces were vacuum coated with gold prior to measurements. SEM images were used to measure the average fiber diameters with the Size Meter Software. At least 50 nanofibers were randomly selected from each of the SEM images.
Structural characterization was performed by X-ray diffraction (XRD) operated with CuKα radiation (λ = 0.15418 nm). XRD analyses were performed in an Ultima IV diffractometer (Rigaku Corporation, Osaka, Japan) operating at the CuKα wavelength of 1.5418 Å, at 40 kV and 20 mA. The scattered radiation was detected at ambient temperature in the 0.6˚ to 80˚ (2θ) angular region at 0.5˚ (2θ)/min in the reflection mode. For a better visualization of the characteristic reflections, the diffractograms were smoothed using the software OriginPro 8.0 (Savitsky-Golay, polynome: 2, points of window: 10). The free software Fityk, downloaded at http://www.unipress.waw.pl/fityk, was used to determine the degree of crystallinity.
Infrared spectroscopy (FTIR) analyses were carried from 4000 cm −1 to 600 cm −1 , at ambient temperature, using a Perkin Elmer Frontier spectrometer (Waltham, MA, USA), equipped with an attenuated total reflection accessory (ATR), by averaging 60 scans with a resolution of 4 cm −1 in transmission mode. Dynamic mechanical analyses (DMA) were performed in a Q800 DMA from

Development of Nanofibers
The combined effects of the independent variables (applied voltage, polymeric solution flow rate, chitosan concentration in the mixture and distance between plate and needle) on nanofiber diameter and homogeneity were evaluated by Box-Behnken design. According to statistical theory, a Box-Behnken design with four independent factors comprises 27 experiments as shown in Table 2 Table 3. The p-value was used as a measure of statistical significance to assess the impact of each factor on the response, where p values greater than 0.05 do not significantly affect response within a confidence interval greater than 95%. Also, the lower the p-value, the more important that factor would be. For both responses, voltage was the factor with the greatest impact, followed by the distance between the plates and the chitosan concentration.
The insignificant coefficients (p > 0.05) were taken from the complete quadratic model in order to obtain a more refined mathematical model, and again through an ANOVA test, the coefficients were recalculated. The reduced response surface model for the mean diameter (Y 1 ) containing only statistically V. R. Viana et al.
where X 1 is the applied voltage (kV), X 3 is the concentration of chitosan solution in the mixture (%) and X 4 is the distance between the collecting plate and the needle. The mathematical model for nanofiber standard deviation (Y 2 ) after the elimination of insignificant terms is presented in Equation (4).
The p-value was also used to measure the statistical significance related to estimated coefficients of the presented models. It was observed that the equations obtained for the responses Y 1 (mean diameter) and Y 2 (standard deviation) were considered significant, with no lack of fit in reduced models. The p-value observed for mean diameter and standard deviation in reduced models was 0.1014 and 0.1256, respectively. The adequacy of the predicted values by the model with the experimental data was also confirmed by the analysis of the coefficient of determination (R 2 ), which represents the proportion of the total variability that can be explained by a regression model. After excluding non-significant terms, the calculated value of R 2 was 0.84002 for the average diameter model, indicating that the model can explain 84% of the average diameter variation of nanofibers. The standard deviation model can explain 83% of the variation in the experimental observations. The adjusted R 2 value for the mean diameter (0.781) and standard deviation (0.7677), after withdrawal of insignificant terms also showed a good correlation between the observed and predicted values, showing that the quadratic models obtained are significant and favorable for the representation of the relationship between the response and the independent variables.  [41]. Also, it was possible to observe the curvature on the plots, which confirms the need for a quadratic regression model. The models show that applied voltage has significant interactions with tip-to-collector distance (X 1 X 3 ) and chitosan content on blend (X 1 X 4 ). Figure 1(a) and Figure 1(c) shows the impact of tip-to-collector distance and applied voltage in nanofibers with flow rate and chitosan content kept constant at center point. A minimum distance was required to give the fibers enough time to evaporate the solvent before reaching the collecting plate, thus avoiding the appearance of failures if they are too close or too far away [42]. The impact of tip-to-plate distance is given by competition between long-and short-range effects. While shorter distances led to the formation of thinner fibers, due to increased electric field, longer distances provided longer jet stretching, due to longer flight times [43] [44]. Figure 1(b) and Figure 1(d) shows the effect of increasing chitosan concentration on PVA/CS mixture. As shown in Figure 2(a), because of increased entanglements and intermolecular interactions between polymer molecules the solution viscosity was enhanced.

Assessing 3D Surface Plots
The liquid-like behavior of the solutions is shown in Figure 2(b), in which the loss modulus (G'') is larger than the storage modulus (G') for all studied solutions. Thus, a higher voltage was required to charge the solutions and start the electrospinning process as shown in Figure 1(b). It is possible to observe that in low voltage regions, regardless of the chitosan concentration, the average fiber diameter increased. However, the increase of the two factors caused a decrease in the responses, as already observed in the model, probably because of the significance of the term (X 1 X 4 ).

Optimization of Nanofibers
The desirability function was used to optimize the influence of the production variables on the process of obtaining PVA/CS nanofibers, seeking to minimize the mean diameter and standard deviation. Figure 3 shows the conditions  obtained after the test application, in which the smallest diameter and standard deviation nanofibers were obtained by optimizing all independent variables. The conditions for lower mean fiber diameter and lower standard deviation by desirability function were Applied Voltage = 13.1 kV, Chitosan concentration = 30% and distance tip-to-plate = 10 cm. As the flow rate was not statistically significant, it was kept at the center point value (0.5 mL/h). The predicted values for mean fiber diameter and the standard deviation were calculated based on the obtained model. The observed values for mean fiber diameter and standard deviation were (196.5 ± 28.3) nm, compared to the predicted values of (185.9 ± 26.8) nm. The error was less than 5% and acceptable.

Stability of PVA/CS Mats in Water
Since PVA is water soluble, a simple contact with water can immediately destroy PVA/CS mats (Figure 4(a)) compromising their wound dressing application as shown in Figure 4(b). A crosslink treatment was performed to guarantee water resistance, maintaining the nanofibrous structure with contact with water. After a 24 h treatment with GTA, water stability was increased. Glutaraldehyde treatment was able to increase stability in water, regardless of exposure time, as shown in Figure 4(c) and Figure 4(e). After a 24 h treatment, it was possible to observe a decrease in nanofiber diameter to 126.5 ± 25.7 nm. A 35.5% reduction compared to fiber diameter without crosslinking. Regarding morphology, it can be observed that after 1 h of immersion, the fibrous morphology was not lost; however, there was swelling of the nanofibers with visible coalescence that prevented the correct measurement of the average diameter (Figure 4(d)). The 48 h treatment (Figure 4(e) and Figure 4(f)) showed an even larger diameter reduction, 44.8% (108.4 ± 15.4) nm, with fibers slightly altered after exposure to water, Journal of Materials Science and Chemical Engineering with an average diameter of (128.4 ± 22.4) nm, maintaining its homogeneous appearance, confirming the efficiency of the vapor-phase crosslinking reaction

Characterization of PVA/CS Mats
FTIR spectra were used to assess structural changes that occurred after crosslinking with GTA. Figure 5 The crystallinity of uncrosslinked and crosslinked 70:30 PVA/CS mats were evaluated by X-ray diffraction ( Figure 5(b)). According to the literature, the characteristic chitosan crystalline peaks are not visualized when the PVA composition is high in PVA/CS mats [46]. For the uncrosslinked mat ( Figure 5(b), trace I), the only maximum is in the 19˚ -20˚ (2θ) region, and was related to the combined 101 and 200 reflections of semicrystalline PVA [50]. The intensity of this characteristic PVA reflection decreased significantly for the crosslinked mat ( Figure 5(b), trace II and III). Therefore, the degrees of crystallinity found were 22.5% for the uncrosslinked and 9.5% and 4.7% for the 24 h and 48 h crosslinked materials, respectively. This behavior is in accordance with the formation of crosslinks, which hinder the mobility of PVA molecules, and this phenomenon depends on the exposure time to GTA.

Mechanical Behavior
Dynamic mechanical analyses were carried out for uncrosslinked and crosslinked optimized PVA/CS mats. The variation of the loss factor (tanδ), given by the ratio between loss modulus (E'') and storage modulus (E') values, as a function of temperature are shown in Figure 6(a), within the -60˚C to 310˚C range.
This factor is used as an indicator of stiffness or flexibility. In Figure 6(a), the peak at 10.5˚C, observed for the uncrosslinked sample, was attributed to the γ relaxation of CS. No significant change was observed for the 24 h crosslinked mat. The peak at 110.3˚C, observed for the uncrosslinked mat may be attributed to the PVA α relaxation, associated with its dynamic glass transition temperature (T g ), and to encompass the β relaxation peak for CS [51]. At a higher temperature, the peak with its maximum at 201.8˚C may be attributed to the α transition of CS. The presence of two α relaxations for the PVA/CS uncrosslinked mat revealed lack of miscibility between PVA and CS. On the other hand, for the crosslinked samples only one broad peak was observed in the ~60˚C to 235˚C temperature range for both crosslinked products. However, it is possible to observe a shift of the tanδ peak with increasing exposure time to GTA. The increase in tanδ observed for the product obtained after 48 h indicates a higher crosslinking density. In the crosslinked structure, a higher crosslinking density means a shorter distance between molecular chains. Such an arrangement forces the segments of the chains to be in a more intimate contact improving energy dissipation [52] [53].  Table 4. The decrease in elongation at break was assigned to the reduction in stretchiness after crosslinking reactions, which involved -OH and −NH 2 groups of PVA and CS.

Conclusion
In this study, PVA/CS nanofibers were fabricated. Process parameters and PVA/CS blend composition were optimized to determine their influence on the formation of small size nanofibers and good homogeneity, observed through standard deviation. A response surface methodology, combining Box-Behnken planning and desirability function, was employed to obtain and optimize a mathematical model for this system. A second order regression model was obtained and proved adequate to predict the behavior of PVA/CS fibers within the limits of the studied factors. The minimum diameter with minimum standard deviation was (196.5 ± 28.3) nm under conditions of 13.1 kV for applied voltage, 30% CS and 10 cm distance plate-to-tip. The flow rate was not significant for the studied limits. The optimized fibers were crosslinked and analyzed for stability in water, characterized by X-ray diffraction and Fourier transform infrared spectroscopy. The results revealed that 48 h exposure to the crosslinking agent (GTA vapor) was efficient to promote water stability and enhanced tensile strength. In this way, it was possible to optimize the electrospinning process and the water resistant PVA/CS of nanofibers, which enables the use of these materials as drug release systems.