(ALK) organophosphite antioxidant, and Stabaxol® P200 (P200) anti-hydrolysis carbodiimide. The chemical structures of these stabilizers are shown in Figure 4.1.

4.1. PHB S

PHB stabilization was performed by the use of three commercial processing additives:

Anox 20

(ANX) phenolic antioxidant, Alkanox 240

(ALK) organophosphite antioxidant, and Stabaxol® P200 (P200) anti-hydrolysis carbodiimide. The chemical structures of these stabilizers are shown in Figure 4.1.

(a) (b)

(c)

Figure 4.1. Structures of commercial stabilizers Anox 20

(a), Alkanox 240

(b), and Stabaxol® P200 (c)

Phosphites compounds, due to their unshared pair of electrons, show generally a high degree of chemical reactivity. Experiments carried out by Jacques et al. [205] showed that triphenylphosphite (TPP) can promote chain extension during the processing of poly(ethylene terephtalate) (PET). The mechanism of chain extension occurs predominantly by reaction of the more reactive hydroxyl end groups with TPP, followed by the esterification of carboxylic acid end groups of PET with the phosphite groups previously bounded to PET chain ends.

Microbial PHB contains both hydroxyl and carboxyl end groups in its main chain. During

melt processing this polymer undergoes a non-radical random chain scission reaction,

leading to the formation of both carboxyl and vinyl (crotonate) ester groups [37]. At the

early stage of PHB thermal degradation, a polycondensation reaction between hydroxyl

and carboxyl end groups of PHB was observed, occurring until the total consumption of

the hydroxyl groups [37, 206, 207]. In this view, it was hypothesized that the addition of an organophosphite compound to PHB during its melt processing could promote a reaction similar to that proposed for the system PET-TPP, as indicated by the mechanisms reported in Schemes 4.1 and 4.2.

The hypothesized reaction between PHB hydroxyl end groups and ALK organophosphite could result in the substitution of the radical 2,4-di-t-butylphenyl and the production of the correspondent phenol as by-product. With the elimination of this by-product, a multi- substitution reaction could proceed. According to this hypothesis, both molecular weights should increase, and branching should occur by means of phosphite linkage (scheme 4.1).

Scheme 4.1. Hypothesized reaction mechanism between PHB hydroxyl end groups and organophosphitic compound

On the other hand, if PHB carboxyl end groups could react with the phenoxy groups of phosphite, phenyl ester should formed at the polymer chain end (Scheme 4.2a).

Alternatively, a new ester bond could be formed by transesterification between

phosphited and carboxylic acid chain ends of PHB (Scheme 4.2b). Also in this case, a

molecular weight increase should be observed.

(a)

(b)

Scheme 4.2. Hypothesized reaction mechanism between PHB carboxyl end groups and organophosphitic compound: substitution (a) and transesterification (b)

In addition, phosphites are known to be very easily hydrolyzed by water. As a consequence, another possible reaction during the processing could be the hydrolyzation of ALK by the water molecules present in PHB in the form of humidity and/or produced by esterification. In this case ALK could have a protective role on PHB hydrolytic degradation.

In order to perform a cooperative effect on polymer stabilization, the blending of two different classes of antioxidants is a commonly used industrial practice [208]. Therefore, some formulation containing ANX (a hindered phenolic antioxidant) were prepared, with the aim to verify the possibility of a synergistic effect with ALK on PHB stabilization.

Carbodiimide compounds are claimed to be effective against hydrolysis in formulations

containing easily hydrolysable polymers such as polyesters or polyurethanes in a

concentration ranging from 0.01 to 5 wt-% [209]. The hypothesized mechanism for

carboxyl-carbodiimide reaction is shown in Scheme 4.3. Initially, cerbodiimide (I) is

protonated (II), with the following attack of the cation by the acid anion (III) to form the

O-acylisourea (IV). From this intermediate, the reaction can proceed with two different

pathways: either (i) the arrangement by way of a cyclic electronic displacement forming

the stable N-acylurea (V), or (ii) O-acylisourea may be protonated to the cation (VI),

which is subsequently converted into N,N’-disubstituted urea (VII) and acid anhydride

(VIII) by the attack of a second anion. With aromatic carbodiimides, the formation of N-

acylurea greatly predominates [210, 211].

Scheme 4.3.Hypothesized reaction mechanism between PHB carboxyl end groupscarbodiimide-based compound

Two kinds of experiments were carried out aiming to PHB stabilization. The first approach consisted in the preparation, according to a statistical mixture design, of some formulations containing Anox 20

and Alkanox 240

. The proportion of the two stabilizers was considered to influence both molecular weight and tensile properties of PHB. In the second experiment PHB was processed in the presence of Stabaxol® P200 at three different concentrations, with a maximum of 5-wt-%.

4.1.1. PHB Stabilization with Anox 20

and Alkanox 240

The total amount of the additives in the formulations was fixed in 0.5 wt-%. Four proportions between the two additives (0, 1/3, 2/3, 1) were defined and used on modelling, according to Equations 1 and 2:

Linear model y = b

x

+ b

x

+ ε _[1]

Quadratic model y = b

x

+ b

x

+ b

x

x

+ ε _[2]

The central proportion 1/2 and other two points corresponding to an additives amount of 1% were included in order to test the model. Table 4.1 shows the formulations of the {2, 3} mixture design (including the model test points) and the respective identification codes. Pristine PHB was also processed in the same experimental conditions as reference.

Each formulation was prepared in replicate.

Table 4.1. { 2, 3 } mixture design for PHB stabilization

Design PHB Anox 20

Alkanox 240

Code Sample (wt-%) (wt-%) (wt-%)

F0 PHB 100.0 0.00 0.00

F1 50A 99.5 0.50 0.00

F2 33A17P 99.5 0.33 0.17

F3 17A33P 99.5 0.17 0.33

F4 50P 99.5 0.00 0.50

F5 25A25P 99.5 0.25 0.25

F6 50A50P 99.0 0.50 0.50

F7 100P 99.0 0.00 1.00

One of the reactions suggested for PHB additivated with organophosphitic compound consisted in a possible chain extension, with a consequent increase in molecular weight and melt viscosity. According to this hypothesis, this effect should be reflected on the value of the torque observed during melt mixing of the mixtures. Torque traces representative of pristine and ANX/ALK stabilized PHB recorded from internal mixer are shown in Figure 4.2. Average torque values of each duplicate formulation measured at mixing end (420 s) are reported in Table 4.2.

Figure 4.2. Internal mixer torque traces of pristine and ANX/ALK stabilized PHB

Peaks observed up to 180 s (Fig. 4.2) corresponded to the chamber feeding of the internal mixer. The presence of both stabilizers increased slightly the final value of torque (Fig.

4.2 and Tab. 4.2), supporting the hypothesis of PHB chain extension during melt mixing.

However, applying a statistical Dunnett’s test, consisting in the comparison of a mean with a control (in this case the mean of replicated PHB torque values), no significant differences at 95% of confidence were found between torque values of pristine and stabilized PHB. The high degree of dispersion between duplicate processing, represented by the standard deviation (SD) in Table 4.2, suggests that in the considered system a not controlled variable could be present. Before melt processing, materials were dried overnight in an oven with air circulation, and then processed randomly in relation to the order reported in Table 4.1. Hence, it is possible that the applied drying system was not able to ensure a complete drying of PHB and additives, leading to the presence of residual humidity during processing. Further, additional humidity could be adsorbed by the components during their handling before melt processing.

Table 4.2. Torque values of pristine and ANX/ALK stabilized PHB at 420 s of mixing

Sample Torque ± SD (Nm)

PHB 16.00 ± 0.99

50A 17.05 ± 1.20

33A17P 16.05 ± 0.07

17A33P 15.75 ± 0.78

50P 18.65 ± 2.47

25A25P 16.00 ± 1.41

50A05P 17.50 ± 0.71

100P 17.10 ± 0.85

a) SD is the standard deviation

4.1.1.1 Structural Characterization

Transmission PHB FT-IR spectrum is shown in Figure 4.3. The assignments of the

significant peaks of PHB are reported in Table 4.3 [212, 215].

Figure 4.3. FT-IR spectrum of PHB

Table 4.3. Assignments of FT-IR peaks of PHB

Wavenumber (cm

) Assignment

3437 OH stretching (H bridges)

2976 CH

(C-H asymmetric stretching)

2934 CH

(C-H asymmetric stretching)

2875 CH

(C-H symmetric stretching)

1724 C=O stretching

1453 CH

(asymmetric)

1379 CH

symmetric wagging

1289 CH

wagging

1278 CH

wagging

1228 Conformational band of the helical chains

1184 C-O-C asymmetric stretching

1132 C-O-C symmetric stretching

1100 OH stretching

1057 C-O symmetric stretching

979 C-C stretching

929 CH

rocking

896 CH

rocking

825 CH

rocking

Transmission FT-IR spectra of ANX and ALK are shown in Figure 4.4. Both the additives showed absorption peaks typical of aromatic compounds, such as stretching of the C

C bond in the aromatic ring (1485 and 1600 cm

) and stretching of the aromatic C-H bond at 3030 cm

. Other characteristics absorptions were found at ca. 900 cm

(C-H bending out of the plane) and in the region 1100-1200 cm

(C-H bending in the plane).

Absorption at around 2960 cm

, 2910 cm

, and 2870 cm

were attributed to methyl groups. Due to their different structures, the additives presented also peculiar peaks. ANX showed adsorptions at 3640 cm

(stretching of the OH linked to the aromatic ring), 1741 cm

(stretching of C=O), and 1435 cm

(CH

scissoring), while ALK spectrum was characterized by the absorption peaks relative to P-O bond (1200 and 850 cm

).

Figure 4.4. FT-IR spectra of ALK and ANX

As discussed above, it was supposed that both carboxyl and hydroxyl end groups of PHB could react with ALK, resulting in the extension of PHB chains. If this reaction occurred, the formation of new P-O bonds should be observed (Schemes 4.1 and 4.2). FT-IR absorptions of P-O bonds are known to appear in the range 1250-800 cm

[213].

However, no significant changes in the spectra of stabilized PHB were observed in this

range in relation to pristine PHB (Fig. 4.5).

Figure 4.5. FT-IR spectrum of pristine and ANX/ALK stabilized PHB in the range 1250-800 cm

4.1.1.2. Molecular Weight Characterization

Average weight molecular weight (Mw) and polydispersity index (PDI) of pristine and ANX/ALK stabilized PHB were controlled at two processing stages: after internal mixer and after compression moulding. The respective GPC traces are shown in Figure 4.6. Mw and PDI values, calculated as the average of two replicated formulations, are reported in Table 4.4 with the relative standard deviation (DS). Average Mw of PHB processed in the internal mixer was calculated in 186 KDa with a PDI of 2.10. The value of not processed PHB powder was 230 KDa with a PDI of 2.07. Therefore, a melt processing of 420 s reduced PHB Mw of around 19%. Apparently, Mw of PHB was slightly increased by the addition of ANK and ALK, whereas an opposite behaviour was verified for PDI. Mw of stabilized PHB ranged from 191 KDa to 208 KDa for 50A and 50P respectively. On the other hand, the lower value of PDI was 1.86 for 33A17P, while the higher one was 1.99 for both 50P and 17A33P formulations. The higher stabilization effect was observed for the formulations containing ALK (208 KDa and 202 KDa for 50P and 100P respectively).

These results suggest that some reactions occurred, leading to the extension of PHB

macromolecular chains. As a consequence, a corresponding fraction with increased

molecular weight could be formed. This proposal can find a support in the GPC traces of

pristine and ANX/ALK stabilized PHB from internal mixer showed in Figures 4.6a and

4.7a.

Table 4.4. Weight average molecular weight (Mw) and polydispersity index (PDI) of pristine and ANX/ALK stabilized PHB

Internal mixer Compression moulding

Sample Mw ± SD (KDa) PDI ± SD Mw ± SD (KDa) PDI ± SD ΔMw (%)

PHB 186 ± 21 2.10 ± 0.07 160 ± 2 1.99 ± 0.11 14

50A 191± 13 1.92 ± 0.05 147 ± 1 2.07 ± 0.01 23

33A17P 199 ± 13 1.86 ± 0.13 161 ± 3 2.04 ± 0.08 19

17A33P 195 ± 9 1.99 ± 0.12 163 ± 2 2.00 ± 0.00 16

50P 208 ± 4 1.99 ± 0.11 145 ± 3 2.04 ± 0.01 30

25A25P 193 ± 8 1.91 ± 0.02 156 ± 12 1.99 ± 0.00 19

50A50P 198 ± 1 1.90 ± 0.03 158 ± 4 2.07 ± 0.13 20

100P 202 ± 3 1.92 ± 0.06 161 ± 2 2.04 ± 0.02 20

a) SD is the standard deviation, ΔMw is the difference in Mw recorded after internal mixer and compression moulding

The GPC peak of pristine PHB started at a retention time at around 10 min, with a maximum placed at ca. 11.5 min. As indicated by the dotted lines in Figure 4.6a, the onset and the peak of stabilized PHB started at slightly lower retention times (higher Mw side). The two small peaks appeared at retention times of ca. 17 and 18 min, corresponding to Mw values of around 1200 and 650 Da, were attributed to the presence of unreacted ANX and ALK. Observing in detail GPC traces in the range of 9.5-10.5 min of retention time (Fig. 4.7a), it was verified that the initial slope of stabilized PHB increased more rapidly in relation to that of pristine PHB, indicating the presence of a fraction of PHB chains with higher Mw in stabilized PHB mixtures.

When PHB formulations were melted for the second time in the compression moulding

stage, a further decrease in Mw was verified. The percentage values of Mw decrease,

indicated in Table 4.4 as ΔMw, ranged between 14 and 30% for PHB and 50P

respectively. These findings indicated that, differently to what previously observed, after

this second processing stage no improvements were observed in stabilized PHB

formulations. Mw ranged from 145 KDa (50P) to 163 KDa (17A33P), while PDI

increased up to the value of 2.07 for both 50A and 50A50P mixtures. This decrease in

Mw was confirmed by the general shift to higher retention times of the peak of GPC

traces (Fig. 4.6b), where the presence of unreacted ANX and ALK was still visible.

(a) (b)

Figure 4.6. GPC traces of pristine and ALK/ANX stabilized PHB from internal mixer (a) and from compression moulding after internal mixer (b)

(a) (b)

Figure 4.7. Zoom on GPC traces of ANX/ALK stabilized PHB from internal mixer in

the range 9.5-10.5 min (a) and from compression moulding after internal

On the other hand, the presence in some formulations of a polymer fraction with higher molecular weight was again indicated by the initial slope of the GPC traces (Fig. 4.7b).

The two PHB formulations that showed a decrease in the initial peak slope of GPC traces are 50A and 50P, confirming the lower values of Mw after compression moulding reported in Table 4.4. Further, these two samples presented higher PDI, probably due to an increase of number of PHB chains with lower Mn. The different behaviour found in Mw after the two processing stages (internal mixing and compression moulding) can be explained with the thermo-mechanical stress undergone by PHB in each processing stage [29]. During internal mixing, PHB can be degraded by the interdependent action of temperature and the mechanical action of the mixing rotors. In the compression moulding stage, degradation can be promoted by both temperature and pressure applied to the polymer in order to obtain the film. Hence, it can be supposed that, in the first processing step, a certain fraction of ANX and ALK can react with PHB, due to the high probability of contact between reactive groups promoted by the mixing action of the rotors. However, in compression moulding the absence of mixing action reduces the probability of a contact between the reactive groups. As a consequence, thermal degradation would predominate over the stabilization effect.

Linear and quadratic models based on Scheffé canonic polynomials (Eq. 1 and 2) were evaluated in order to explain the variance observed on Mw measured after internal mixer.

Figure 4.8a shows the dependence of PHB Mw from ANX/ALK proportion fitted by the linear model regression expressed in Equation 3:

Mw = 190.50*ANX + 204.70*ALK [3]

(± 4.02) (± 4.01)

The numbers indicated below the model coefficients are the respective standard

deviations. As pointed out above ALK thermo-stabilized PHB, being more effective

without ANX. This result was confirmed by the ALK coefficient of Equation 3, which

was higher than that of ANX. Analysis of variance (ANOVA) and graphical residual

analysis were used as statistical tools for model validation. Residuals from fitted model

were obtained by the differences between the responses observed and the corresponding

prediction of response. ANOVA for linear model of Equation 3 is reported in Table 4.5.

(a) (b)

Figure 4.8. PHB Mw changes as a function of ANX/ALK proportions: ( ) experimental points, (- - -) linear curve fitting (a), and (- - -) predicted Mw curve (b)

Table 4.5. ANOVA for the regression of Mw data as a linear function of ANX/ALK proportions

Variance Source DF SS MS F

Linear model 1 316.38 316.38 3.97

Residual 6 477.50 79.58

Total 7 793.88

a) DF = degrees of freedom, SS = sum of squares, MS = mean square, F_calc = comparison of variances

The value of tabled F (F

) at 95% of confidence with 1 and 6 degrees of freedom (F

) of 5.99 is higher than the value of calculated F (F

). This result indicated that there

was not a significant linear trend describing the experimental data. Analyzing normal

probability plot of residuals (Fig. 4.9a), it was found that the values were distributed only

approximately on a straight line. The formulation that showed a much larger residual was

17A33P, and probably this “outlier” contributed for distortions in the ANOVA. The

graphic representing the pattern of the residuals versus predicted Mw (Fig. 4.9b)

confirmed this probability, showing the higher value (in terms of absolute value)

calculated for the residual of the 17A33P formulation.

(a) (b)

Figure 4.9. Normal probability plot (a) and residuals versus predicted Mw (b) from linear model

In order to test the adequacy of the linear model, a new ANOVA including the data of formulation 25A25P was performed. The relation between PHB Mw and ANX/ALK proportions and the predicted values obtained from the new linear regression (Eq. 4) is presented in the Figure 4.8b. The respective ANOVA is reported in Table 4.6.

Mw = 189.60*ANX + 203.79*ALK [4]

(± 3.70) (± 3.70)

Table 4.6. ANOVA for the regression of Mw data as a linear function of ANX/ALK proportions

Variance Source DF SS MS F

Linear model 1 224.05 224.05 2.651

Residual 8 676.04 84.50

Lack of fit 3 126.54 42.18 0.384

Pure error 5 549.50 109.9

Total 9 900.10

a) DF = degrees of freedom, SS = sum of squares, MS = mean square, F_calc = comparison of variances

Since the value of F

was lower than F

= 5.32, it was confirmed that linear model was not adequate to describe the variations in PHB Mw as a function of ANX/ALK proportions after internal mixer. Further, the value of adjusted R-squared (R

) was 0.384, confirming the inadequacy of the linear model. R

, calculated according to Equation 5, can be interpreted as the proportion of variability of the dependent variable around the mean that can be explained by the respective model.

R

= 1 - (means square of residual)/(mean square of total) [5]

From the value F

relative to the lack of fit, which was lower than F

= 5.41, it was concluded that there was an uncontrolled environmental variable contributing to data dispersion. In the view of these results, a regression with a quadratic model was performed. Equation 6 expresses the first attempt. The fit of the curves with the initial and predicted data are presented in Figure 4.10a and 4.10b respectively. The ANOVA of regression is described in Table 4.7.

Mw = 191.92*ANX + 206.09*ALK - 12.44*ANX*ALK [6]

(± 6.05) (± 6.05) (± 28.1)

(a) (b)

Figure 4.10. PHB Mw changes as a function of ANX/ALK proportions: ( )

experimental points, (- - -) quadratic curve fitting (a), and (- - -) predicted

Mw curve (b)

Table 4.7. ANOVA for the regression of Mw data as a quadratic function of ANX/ALK proportions

Variance Source DF SS MS F

Quadratic model 2 316.38 158.19 1.656

Residual 5 477.50 95.5

Total 7 793.88

a) DF = degrees of freedom, SS = sum of squares, MS = mean square, F_calc = comparison of variances

The absolute value of quadratic coefficient, which was smaller than the relative standard deviation, and its negative value suggested an antagonistic effect between ANX and ALK. Other statistics made obvious the insufficiency of the quadratic model to describe Mw data. The value of F

= 5.79 was higher than F

indicating that quadratic model was not significant at 95% of confidence.

Following the test of the quadratic model, a new point corresponding to the 25A25P formulation was added. Equation 7 showed no modification in terms of signal of quadratic term and its standard deviation, if compared with Equation 6. The residual plots shown in Figure 4.11 suggested that the formulation 33A17P showed the higher deviation, and probably this sample was responsible for the failure of the fitting model.

Mw = 192.03*ANX + 206.20*ALK - 17.44*ANX*ALK [7]

(± 4.54) (± 4.54) (± 18.40)

The analysis of variance presented in Table 4.8 showed that F

of both regression and

lack of fit resulted lower than F

= 4.74 and F

= 5.79 respectively. Therefore, as

already stated in the case of the linear model, Mw data of PHB were strongly influenced

by the contribution of not explained variable(s). A hypothesis that can be formulated is

related with humidity uptake as a function of time. As formulation materials after drying

were processed in random order, it is possible that formulations were processed later

remaining more time in environment conditions. Another possibility is that a more

complex regression would explain the changes in Mw with stabilizers proportions. In this

case, a higher number of formulations and replicates would be necessary.

(a) (b)

Figure 4.11. Normal probability plot (a) and residuals versus predicted Mw (b) from quadratic model

Table 4.8. ANOVA for the regression of Mw data as a quadratic function of ANX/ALK proportions

Variance Source DF SS MS F

Quadratic model 2 263.66 131.83 1.449

Residual 7 636.84 90.98

Lack of fit 2 87.34 43.67 0.397

Pure error 5 549.50 109.9

Total 9 900.10

a) DF = degrees of freedom, SS = sum of squares, MS = mean square, F_calc = comparison of variances

4.1.1.3. Thermal Characterization

Thermogravimetric parameters of pristine and ANX/ALK stabilized PHB, calculated as

the average of two replicated formulations, are reported in Table 4.9 with the relative

standard deviation (SD). Derivative TGA (DTG) traces of ANX/ALK stabilized PHB at

0.5 wt-% and 1.0 wt-% are shown in Figures 4.12 and 4.13 respectively.

Table 4.9. Thermogravimetric parameters of pristine and ANX/ALK stabilized PHB

T

± SD T

± SD T

± SD T

± SD R

± SD

Sample (°C) (°C) (°C) (°C) (wt-%)

PHB 276.4 ± 0.6 270.3 ± 2.3 303.2 ± 0.7 300.2 ± 1.2 0.26 ± 0.05 50A 275.9 ± 0.4 270.1 ± 3.5 300.8 ± 1.1 299.1 ± 2.5 0.40 ± 0.07 33A17P 275.8 ± 0.1 272.8 ± 0.6 303.1 ± 0.1 300.6 ± 0.0 0.28 ± 0.02 17A33P 274.6 ± 0.2 271.5 ± 1.9 301.3 ± 0.4 300.6 ± 0.1 0.34 ± 0.01 50P 274.9 ± 1.1 272.1 ± 1.1 301.9 ± 1.1 299.8 ± 1.3 0.26 ± 0.03 25A25P 274.1 ± 1.8 272.1 ± 0.4 301.4 ± 0.8 300.2 ± 1.1 0.28 ± 0.01 50A50P 275.1 ± 0.5 272.9 ± 0.3 302.2 ± 2.3 301.2 ± 0.0 0.28 ± 0.04 100P 274.5 ± 0.1 273.5 ± 0.2 302.2 ± 0.6 301.4 ± 1.3 0.35 ± 0.14

a) T_2% is the degradation temperature corresponding to 2% weight loss in the sample, T_onDSC is the onset degradation temperature measured by DSC, Tp is the temperature corresponding to the maximum weight loss rate after volatile, T_pDSC is the temperature corresponding to the maximum of degradation peak measured by DSC, R₄₅₀ is the residual weight measured at 490°C, SD is the standard deviation

Figure 4.12. DTG traces of pristine and ANX/ALK stabilized PHB at 0.5 wt-%

PHB decomposition occurred in a single weight loss step. The temperature of 2% weight loss (T

) in the sample was considered as the onset of degradation. This parameter in pristine PHB was measured in around 276°C, whereas the temperature of maximum degradation rate (T

) was 303°C. PHB residual weight, measured at 490°C, was 0.3 wt-

%. TGA parameters were not significantly affected by the addition of the two

antioxidants. Stabilized PHB samples decomposed in a single degradation step,

analogously to pristine PHB (Figs. 4.12 and 4.13).

Figure 4.13. DTG traces of pristine and ANX/ALK stabilized PHB at 1.0 wt-%

T

ranged from 274°C for 25A25P formulation to 276°C obtained for both 50A and 33A17P formulations. On the other hand, T

was comprised between 303°C (33A17P) and 301°C (50A, 17A33P, and 25A25P). Residual weight of stabilized PHB at 490°C was found to be slightly higher if compared with pristine PHB, but this difference may be within experimental error of the instrument.

The onset degradation temperature and the temperature corresponding to the maximum of degradation rate were also evaluated by DSC technique. With this aim, the 3

heating scan of pristine and ANX/ALK stabilized PHB was performed until the temperature of 350°C. The endothermic peak observed in the range 270-320°C was attributed to PHB thermal degradation (Fig. 4.14). Onset degradation temperature (T

) was calculated as the temperature at the intersection of the two lines tangent to the degradation peak, with a slope of around -0.08 W/gK. Temperature of maximum degradation rate (T

) was taken at peak temperature (Fig. 4.14). In good agreement with TGA measurements, T

and T

of pristine PHB were 270°C and 300°C respectively, and stabilized PHB samples did not show any appreciable variations (Tab. 4.9). Higher value of T

was observed for 100P (274°C), whereas T

was found to change of 1-2 °C.

Thermodynamic parameters of pristine and ANX/ALK stabilized PHB, calculated as the

average of two replicated formulations, are reported in Tables 4.10-4.12 with the relative

standard deviation (SD). The DSC traces are shown in Figures 4.15-4.17. The first

heating DSC scan was not considered because of the strong dependence from the thermal

history of the sample. However, the degree of crystallinity (Xc) of the mixtures,

Figure 4.14. DSC traces (3

heating scan) of pristine and ANX/ALK stabilized PHB in the range 230-330°C

In the 2

heating scan after quenching, glass transition temperature (T

), not reported in Table 4.10, was 7.0°C for both pristine and stabilized PHB samples. Pristine PHB showed two exothermic cold crystallization peaks. The peak of the first cold crystallization (T

) appeared at around 60°C, and its enthalpy (ΔH

) was 52 J/g, which represents ca. 61%

of the total crystallization energy. The peak temperature (T

) and the enthalpy (ΔH

)

of the second cold crystallization transition were measured at 80°C and 29 J/g

respectively. In addition, PHB showed a single endothermic melting peak, with an

enthalpy (ΔH

) of ca. 85 J/g and a peak temperature (T

) of 180°C. The values of

thermodynamic parameters of stabilized PHB from the 2

heating scan were similar to

those observed for pristine PHB. The data dispersion for the two replicates did not show a

significant difference at 95% of confidence.

Figure 4.15. DSC traces (2

heating scan) of pristine and ANX/ALK stabilized PHB

Table 4.10. DSC parameters (2

heating) of pristine and ANX/ALK stabilized PHB

T

± SD ΔH

± SD T

± SD ΔH

± SD T

± SD ΔH

± SD

Sample (°C) (J/g) (°C) (J/g) (°C) (J/g)

PHB 60.0 ± 0.7 52.3 ± 3.5 80.0 ± 1.4 28.9 ± 2.1 180.3 ± 0.1 84.7 ± 2.2 50A 59.5 ± 1.5 50.6 ± 2.3 80.0 ± 0.0 31.0 ± 0.6 180.2 ± 0.0 83.7 ± 0.8 33A17P 60.2 ± 0.8 52.6 ± 0.7 80.5 ± 0.7 29.5 ± 0.3 180.2 ± 0.1 84.3 ± 0.6 17A33P 60.1 ± 0.7 51.8 ± 0.1 80.0 ± 5.7 29.8 ± 0.6 180.1 ± 0.0 82.0 ± 3.8 50P 60.1 ± 0.7 54.9 ± 2.9 78.5 ± 0.7 30.8 ± 1.2 180.1 ± 0.0 86.1 ± 2.8 25A25P 60.0 ± 0.7 51.0 ± 2.2 82.5 ± 4.9 29.7 ± 1.3 180.3 ± 0.1 85.1 ± 0.9 50A50P 61.0 ± 0.7 53.4 ± 2.7 80.0 ± 0.0 30.5 ± 2.5 180.4 ± 0.1 84.1 ± 3.7 100P 61.7 ± 1.3 55.2 ± 1.8 78.0 ± 2.8 29.3 ± 0.8 181.6 ± 2.1 84.2 ± 1.3

a) T_cc is the cold crystallization temperature, T_m is the melting temperature, ΔHcc is the cold crystallization

In the 2

cooling scan at 10°C/min (Tab 4.11 and Fig. 4.16), PHB showed a single exothermic crystallization peak. Crystallization onset temperature (T

) was observed at 114°C, with a range (ΔT

) of 50°C. Crystallization temperature peak (T

) and enthalpy (ΔH

) were 90°C and 76 J/g respectively. Similarly, a single crystallization peak was observed also in the case of stabilized PHB mixtures. However, some differences were found in the values of the thermodynamic parameters. T

was decreased to ca. 110°C in the case of 100P formulation. On the other hand, samples containing ALK in the formulation showed that T

and ΔH

tends to decrease and ΔT

leans to increase with the increasing of its proportion and amount. The more relevant variations were found in 100P, where ΔT

, T

, and ΔH

and were measured in 72°C, 79°C, and 69 J/g respectively.

These values correspond to an increase of ΔT

of ca. 46% and to a decrease of 12% and 9% for T

and ΔH

respectively. As observed in the discussion of molecular weight data, the samples containing a higher amount or proportion of ALK presented higher Mw and lower PDI. As crystallization depends on chain flexibility, it is possible that these variations in Mw and PDI decreased the possibility of a close packaging of the polymeric chains, rendering the formation of the crystals more difficult.

Table 4.11. DSC parameters (2

cooling) of pristine and ANX/ALK stabilized PHB

T

± SD ΔT

± SD T

± SD ΔH

± SD

Sample (°C) (°C) (°C) (J/g)

PHB 113.5 ± 1.3 49.5 ± 1.3 90.3 ± 1.7 76.0 ± 0.8

50A 115.2 ± 2.3 54.8 ± 1.3 91.0 ± 2.5 76.1 ± 2.0

33A17P 113.4 ± 0.7 60.0 ± 3.4 88.6 ± 0.8 74.4 ± 0.4

17A33P 116.6 ± 3.1 55.8 ± 5.4 91.6 ± 3.4 74.6 ± 7.1

50P 112.1 ± 1.3 63.2 ± 3.7 86.3 ± 2.5 74.5 ± 4.3

25A25P 115.4 ± 2.1 50.7 ± 1.8 91.5 ± 1.8 76.1 ± 0.2

50A50P 114.1 ± 1.7 63.0 ± 5.4 84.4 ± 1.8 70.8 ± 1.2

100P 110.3 ± 2.4 72.4 ± 6.4 79.1 ± 7.5 68.8 ± 6.8

a) Ton is the crystallization onset temperature, ΔTc is the crystallization range temperature, Tc is the crystallization temperature, ΔHc is the crystallization enthalpy, SD is the standard deviation

Figure 4.16. DSC traces (2

cooling scan) of pristine and ANX/ALK stabilized PHB

These observations were also verified for the data obtained from the 3

heating scan after

crystallization at 10°C/min (Tab. 4.12 and Fig. 4.17). In fact, during this scan, the cold

crystallization temperature (T

) tended to decrease and the associate enthalpy (ΔH

)

tended to rise at the increasing of ALK content. These results confirmed the difficult

crystallization process of PHB stabilized with ALK, and indicated that after

crystallization at 10°C/min there was a larger fraction of PHB that was still able to

crystallize before melting, suggesting a kinetic phenomenon. As a result, crystals with

different sizes and perfection could be developed, as confirmed by the presence of a

double melting peak in the 3

heating scan traces (Fig. 4.17). In particular, the enthalpy

of melting (ΔH

) of both peaks was strongly affected by ALK concentration, showing an

opposite behaviour. Increasing ALK content in the formulations, the enthalpy of first

melting peak decreased, and the second one increased. However, the sum of these two

Table 4.12. DSC parameters (3

heating) of PHB and ANX/ALK stabilized PHB

T

± SD ΔH

± SD T

± SD ΔH

± SD T

± SD ΔH

± SD

Sample (°C) (J/g) (°C) (J/g) (°C) (J/g)

PHB 103.4 ± 3.7 3.2 ± 0.8 172.3 ± 0.1 55.5 ± 0.4 179.8 ± 0.5 37.5 ± 3.4 50A 105.2 ± 1.7 3.5 ± 0.5 172.6 ± 0.4 59.1 ± 7.0 178.1 ± 1.1 34.5 ± 4.2 33A17P 102.0 ± 0.9 4.0 ± 0.7 171.9 ± 0.2 49.4 ± 1.3 179.2 ± 1.0 42.5 ± 0.4 17A33P 104.4 ± 4.9 3.9 ± 0.8 172.8 ± 2.1 54.3 ± 9.9 179.7 ± 0.7 36.4 ± 4.8 50P 102.4 ± 2.1 4.5 ± 0.1 172.4 ± 0.0 50.2 ± 2.1 179.6 ± 0.7 45.8 ± 2.1 25A25P 104.9 ± 1.6 4.6 ± 0.1 172.8 ± 0.8 57.6 ± 0.6 179.8 ±±0.8 35.5 ± 2.1 50A50P 99.5 ± 1.6 5.3 ± 1.7 172.1 ± 0.2 46.2 ± 1.8 178.9 ± 0.3 45.6 ± 5.6 100P 100.0 ± 1.5 9.5 ± 3.3 173.4 ± 1.4 43.2 ± 5.9 179.6 ± 0.8 47.8 ± 2.9

a) T_cc is the cold crystallization temperature, T_m is the melting temperature, ΔHcc is the cold crystallization enthalpy, ΔHm is the enthalpy of fusion, SD is the standard deviation

Figure 4.17. DSC traces (3

heating scan) of pristine and ANX/ALK stabilized PHB

The degree of crystallinity (Xc), calculated from the three heating scans as the ratio of ΔH

observed in the sample and the theoretical enthalpy of fusion of a 100% crystalline polymer (146 J/g) [27], is reported in Table 4.13. Xc of PHB in the 2

heating scan was calculated as 58%. As expected, this value was (around 9%) lower if compared with the value observed in the 1

heating scan. In fact, it can be considered that the transition in the 1

heating scan represents a situation of equilibrium, and therefore corresponds to the maximum degree of crystallinity achievable by the polymer. In the successive 3

heating scan after cooling at 10°C/min, values of Xc were very close to those found in the 1

heating. Although apparently a slight change of crystallinity with ANX/ALK proportions was observed, due to values dispersion these variations were statistically equivalent.

Table 4.13. Degree of crystallinity (Xc) of pristine and ANX/ALK stabilized PHB

Xc ± SD (%) Xc ± SD (%) Xc ± SD (%) Sample 1

heating 2

heating 3

heating

PHB 67.2 ± 0.4 58.0 ± 1.5 63.7 ± 2.0

50A 66.9 ± 2.8 57.3 ± 0.6 64.1 ± 1.9

33A17P 70.1 ± 3.0 57.8 ± 0.4 62.9 ± 0.7

17A33P 64.1 ± 3.8 56.2 ± 2.6 62.2 ± 3.5

50P 68.8 ± 1.3 59.0 ± 1.9 65.7 ± 2.8

25A25P 67.3 ± 1.6 58.3 ± 0.6 63.8 ± 1.8

50A50P 68.3 ± 5.7 57.6 ± 2.5 62.9 ± 2.6

100P 69.4 ± 0.4 57.7 ± 0.9 62.3 ± 2.1

a) SD is the standard deviation

4.1.1.4. Mechanical Characterization

Young modulus (YM) and tensile strength (TS) of pristine and ANX/ALK stabilized

PHB are reported in Table 4.14. 12 specimens were tested for each formulation and each

replicate. Confidence limits (SE) of tensile properties were calculated from Student’s t

distribution according to Equation 8:

where t is the tabled upper critical value (t of Student) at the desired significance level and degrees of freedom (ν = n - 1), SD is the standard deviation, and n is the number of specimens. For n = 10 and 95% of confidence, t

is 2.262.

Average Young modulus of PHB replicates was 1.75 GPa, whereas the average tensile strength was measured in 31.81 MPa. YM of stabilized PHB formulations ranged from 1.88 (replicate I of 33A17P) to 1.60 (replicate II of 50A50P), indicating that there were no appreciable variations in relation to the values of pristine PHB. The same observations were made in the case of TS, which varied from 32.60 MPa (replicate I of 25A25P) to 27.66 MPa (replicate I of 50P).

Table 4.14. Tensile properties of pristine and ANX/ALK stabilized PHB

Replicate I Replicate II

Sample YM ± SE (GPa) TS ± SE (MPa) YM ± SE (GPa) TS ± SE (MPa)

PHB 1.76 ± 0.05 30.54 ± 1.21 1.74 ± 0.04 33.08 ± 1.97

50A 1.82 ± 0.04 29.74 ± 0.95 1.63 ± 0.05 32.49 ± 1.87

33A17P 1.88 ± 0.04 29.06 ± 1.76 1.71 ± 0.04 31.29 ± 2.11 17A33P 1.73 ± 0.03 28.79 ± 2.43 1.70 ± 0.04 31.05 ± 1.41

50P 1.68 ± 0.03 27.66 ± 1.20 1.76 ± 0.03 29.21 ± 1.72

25A25P 1.62 ± 0.06 32.60 ± 1.43 1.84 ± 0.05 28.41 ± 1.40 50A50P 1.71 ± 0.05 30.65 ± 1.48 1.60 ± 0.06 30.04 ± 0.91

100P 1.78 ± 0.04 29.58 ± 0.40 1.66 ± 0.06 27.98 ± 1.62

a) YM is the Young modulus, TS is the tensile strength, SE is the confidence limit at 95%

In order to verify the homogeneity and to compare the mean values, analysis of variance

(ANOVA) was performed on the Young modulus. Each formulation was treated as a

single factor, and as a source of variation. Tables 4.15-4.22 present the ANOVA for each

formulation. The tabulated F value at 95% confidence level with 1 and 18 degrees of

freedom is 4.41. It was observed that F

was not significant on formulations PHB and

17A33P, meaning that only for these two samples the variance within each replicate was

nearly equivalent to the variance between them. These results indicated that, in the

processing of these two formulations, no other variables contributed to the dispersion of

their YM values.

Table 4.15. ANOVA for YM of PHB

DF SS MS F

Between 1 0.002645 0.002645 0.7624

Within 18 0.6245 0.003469 —

Total 19 0.0651 0.006114 —

a) DF = degrees of freedom, SS = sum squares, MS = mean square, F_calc = comparison of variances

Table 4.16. ANOVA for YM of 50A

DF SS MS F

Between 1 0.1748 0.1748 38.89

Within 18 0.08093 0.004496 —

Total 19 0.2558 0.1793 —

a) DF = degrees of freedom, SS = sum squares, MS = mean square, Fcalc = comparison of variances

Table 4.17. ANOVA for YM of 33A17P

DF SS MS F

Between 1 0.1462 0.1462 47.77

Within 18 0.05509 0.003061 —

Total 19 0.2013 0.1493 —

a) DF = degrees of freedom, SS = sum squares, MS = mean square, F_calc = comparison of variances

Table 4.18. ANOVA for YM of 25A25P

DF SS MS F

Between 1 0.2354 0.2354 41.25

Within 18 0.1027 0.005707 —

Total 19 0.3382 0.2412 —

a) DF = degrees of freedom, SS = sum squares, MS = mean square, F_calc = comparison of variances

Table 4.19. ANOVA for YM of 17A33P

DF SS MS F

Between 1 0.003125 0.003125 1.321

Within 18 0.04257 0.002365 —

Total 19 0.0457 0.00549 —

a) DF = degrees of freedom, SS = sum squares, MS = mean square, F_calc = comparison of variances

Table 4.20. ANOVA for YM of 50P

DF SS MS F

Between 1 0.003444 0.003444 19.06

Within 18 0.3253 0.001807 —

Total 19 0.06697 0.03625 —

a) DF = degrees of freedom, SS = sum squares, MS = mean square, Fcalc = comparison of variances

Table 4.21. ANOVA for YM of 50A50P

DF SS MS F

Between 1 0.0605 0.0605 11.9

Within 18 0.09148 0.005082 —

Total 19 0.152 0.06558 —

a) DF = degrees of freedom, SS = sum squares, MS = mean square, F_calc = comparison of variances

Table 4.22. ANOVA for YM of 100P

DF SS MS F

Between 1 0.06498 0.06498 13.14

Within 18 0.8902 0.004946 —

Total 19 0.154 0.06993 —

a) DF = degrees of freedom, SS = sum squares, MS = mean square, F_calc = comparison of variances

A different behaviour was observed in the other samples, for which the variances between the replicates were significantly higher that within them. As pointed out in the analysis of the results of molecular weight, one of the not controlled variables could be the humidity adsorbed by the dried components of the formulation prior to their processing. This result was confirmed by applying Newman Keuls paired comparison test. Compared pairs of replicate formulations are reported in Table 4.23. As previously mentioned, the hypothesis explaining the significant differences between duplicate samples can be represented by the residual humidity present in the samples before melt processing.

Table 4.23. Newman Keuls test PHB and ANX/ALK stabilized PHB

DM SE SRT Prob

PHB 0.023 0.02634 1.235 0.3941

50A 0.187 0.02999 8.819 1.893e-05**

33A17P 0.171 0.02474 9.775 1.386e-05**

17A33P 0.025 0.02175 1.626 0.2654

50P 0.083 0.01901 6.174 0.0003837**

25A25P 0.217 0.03379 9.083 1.67e-05**

50A50P 0.11 0.03188 4.879 0.002866**

100P 0.114 0.03145 5.126 0.001949**

a) DM = difference of means, SE = standard error, SRT = Studentized Range statistic, Prob = probability that a value of SRT as high might happen by chance (* = significant at 5% level, ** = significant at 1%

level)

Values of YM (replicate I – Tab. 4.14) were used for modelling. ANOVA models are a versatile statistical tool for studying the relation between a dependent and one or more independent variables. Linear model of a {2, 3} mixture design is presented in Equation 9 and graphically represented in Figure 4.18a. ANX coefficient was higher than that calculated for ALK, indicating that the contribution of ANX on Young modulus was higher. This result was apparently in contrast with what observed in the case of the molecular weight (Eq. 3).

YM = 1.862*ANX + 1.691*ALK [9]

(± 0.050) (± 0.050)

(a) (b)

Figure 4.18. PHB YM changes as a function of ANX/ALK proportions: ( ) experimental points, (- - -) linear curve fitting (a), and (- - -) predicted YM curve (b)

The dependence of mechanical properties from molecular weight is known to be directly proportional. In the view that the contribution of ANX on Mw was minor than that of ALK, it could have been supposed that the contribution on YM given by the formulations containing ALK would have been higher. One hypothesis could be formulated on the basis of the presence of a higher fraction of PHB with a lower molecular weight acting as a plasticizer. However, PDI values reported in Table 4.4 do not support this hypothesis.

Another explanation could be associated with the degree of crystallinity of the formulations (Tab. 4.13), but a relation between this property and YM was not found.

Then, a further hypothesis to explain the higher contribution of ANX on YM of stabilized PHB can be related with ANX structure, which can be more rigid if compared with ALK.

In fact ALK, reacting with PHB, can loose its rigidity without contributing as a “filler”, as ANX may be doing.

Table 2.24 reports ANOVA analysis of the linear model described in Equation 9. The model F value (F

) of 94.931 implied that the model was significant (F

= 4.098).

Residual-normal probability and residual-predicted YM plots illustrated in Figure 4.19

suggested that the sample 33A17P did not fulfil the normal distribution. However, its

residual was lower than 2SD, which had a low significancy. The value of R

(Eq. 5) was

0.708, meaning that the linear model was able to explain ca. 70% of the results relative to

the Young modulus.

Table 4.24.ANOVA for the regression of YM data as a linear function of ANX/ALK proportions

Variance Source DF SS MS F

Linear model 1 0.23353 0.23353 94.931

Residual 38 0.09338 0.00246

Total 39 0.32691

a) DF = degrees of freedom, SS = sum of squares, MS = mean square, F_calc = comparison of variances

(a) (b)

Figure 4.19. Normal probability plot (a) and residuals versus predicted YM (b) from linear model

In order to check the adequacy of the model, a new linear regression was performed by the addition of the data calculated for the formulation 25A25P. The new model is described in Equation 10:

YM = 1.830*ANX + 1.660*ALK [10]

(± 0.076) (± 0.076)

ANOVA analysis of this regression, presented in Table 4.25, showed that F

was higher

However, the value of R

indicated that this model was able to explain only 26% of YM data, and the formulations 33A17P and 25A25P contributed for the lack of fit (F

= 2.812).

Table 4.25.ANOVA for the control of regression of YM data as a linear function of ANX/ALK proportions

Variance Source DF SS MS F

Linear model 1 0.16234 0.16234 18.302

Residual 48 0.42591 0.00887

Lack of fit 3 0.26964 0.08988 25.902

Pure error 45 0.15627 0.00347

Total 49 0.58825

a) DF = degrees of freedom, SS = sum of squares, MS = mean square, F_calc = comparison of variances

Equation 11 describes the behaviour of YM as a quadratic canonical polynomial of ANX and ALK variables. Figure 4.20 shows the relationship for experimental and predicted values of YM. ANOVA analysis of the regression of the quadratic model is reported in Table 4.26.

YM = 1.836*ANX + 1.666*ALK + 0.231*ANX*ALK [11]

(± 0.065) (± 0.065) (± 0.304)

(a) (b)

Figure 4.20. PHB YM changes as a function of ANX/ALK proportions: ( )

experimental points, (- - -) quadratic curve fitting (a), and (- - -) predicted

YM curve (b)

SD for the coefficient related to the formulations containing both stabilizers (0.304) was higher that its value, meaning that this contribution was not considerable. By the elimination of this coefficient, a linear model is obtained. However, F

value (3.252) was lower than the calculated one, indicating the significance of the model. In fact, from R

this regression could explain ca. 70% of fit.

Table 4.26.ANOVA for the regression of YM data as a quadratic function of ANX/ALK proportions

Variance Source DF SS MS F

Quadratic model 2 0.23353 0.11676 46.333

Residual 37 0.09338 0.00252

Total 39 0.32691

a) DF = degrees of freedom, SS = sum of squares, MS = mean square, F_calc = comparison of variances

In order to confirm the findings obtained by the quadratic model, a new regression was performed including YM data of the formulation 25A25P. The results are presented in Equation 12, and the relative ANOVA is reported in Table 4.27.

YM = 1.844*ANX + 1.673*ALK - 0.098*ANX*ALK [12]

(± 0.112) (± 0.112) (± 0.452)

Table 4.27.ANOVA for the control of the regression of YM data as a quadratic function of ANX/ALK proportions

Variance Source DF SS MS F

Quadratic model 2 0.16849 0.0842 9.429

Residual 47 0.41976 0.00893

Lack of fit 2 0.26349 0.1317 37.954

Pure error 45 0.15627 0.00347

Total 49 0.58825

a) DF = degrees of freedom, SS = sum of squares, MS = mean square, F_calc = comparison of variances

The F

value of regression was not much higher than the tabled value (F

= 3.195), as was observed for the models presented above, suggesting that the significance of the coefficients was not sufficient to allow the acceptance of the model. Further, the F

of lack of fit was higher than the table value (F

= 3.204), confirming with the value of R

(0.2558) the inadequacy of the quadratic model. Residual plots in Figure 4.21 showed again that 33A17P and 25A25P were the formulations that contributed most to the variance of YM data, affecting negatively the success of the models. Finally, it can be concluded that the accuracy of the developed models could be improved by including a higher number of parameters and levels.

(a) (b)

Figure 4.21. Normal probability plot (a) and residuals versus predicted YM (b) from quadratic model

4.1.2. PHB Stabilization with Stabaxol® P200

The degradation rate of dried and undried polyester melt was found to be significantly

different [216]. The earlier degradation stage showed to be the more sensible to moisture,

while the later one did not show any dependence. As previously mentioned, during

thermal degradation PHB macromolecules undergo a non-radical random chain scission

reaction, followed by a polycondensation reaction between hydroxyl and carboxyl end

groups [37, 206, 207]. Therefore, the hydroxyl groups of water can also take part to this reaction, leading to a decrease in the molecular weight. As a consequence, the presence of humidity during melt processing of PHB based materials can contribute with hydrolysis reaction.

In the previous study, PHB stabilization with phenolic and organophosphitic antioxidants was investigated. However, from statistical analysis, it was concluded that the results were affected by an uncontrollable variable, which was supposed to be the moisture content of the samples. Carbodiimide-based compounds have been proposed as stabilizer against hydrolysis for ester-group containing materials [209, 210, 217, 218] at amounts up to 5 wt-%. In view of this, a screening experiment consisting in PHB processing in presence of the commercial carbodiimide-based compound Stabaxol® P200 (P200) was performed. P200 amount was fixed at 1 wt-% (SL1), 3 wt-% (SL3), and 5 wt-% (SL5).

Torque traces representative of pristine and P200 stabilized PHB recorded from internal mixer are shown in Figure 4.22. Peaks observed up to 60 sec corresponded to the chamber feeding of the internal mixer. No significant differences were observed on torque behaviour of pristine and stabilized PHB at 420 sec, suggesting that the addition of P200 in this range of concentrations did not affect PHB melt viscosity.

Figure 4.22. Internal mixer torque traces of pristine and P200 stabilized PHB 4.1.2.1. Morphological Characterization

POM micrographs of pristine and P200 stabilized PHB are shown in Figures 4.23-4.25.

quenched to 95°C and then kept at this temperature until complete crystallization. The temperature of 95°C was chosen on the base on DSC crystallization traces. PHB showed the characteristic large spherulites containing a Maltese cross-birefringent pattern, as shown in Figure 4.23a and 4.23b at 83x and 208x of magnification respectively.

(a) (b)

Figure 4.23. POM images of PHB: 83x (a) and 208x (b)

POM images of SL1 and SL3 (Fig. 4.24a and 4.24b respectively) at 83x of magnification showed that the spherulitic morphology of P200 stabilized PHB was similar to that of the pristine polymer. However, some variations in the band spacing and birefringence colour were verified. These modifications were probably due to some changes in the configuration of the PHB segments during melt processing in presence of the stabilizer.

(a) (b)

Figure 4.24. POM images (83x) of SL1 (a) and SL3 (b)

In order to verify this hypothesis, a physical mixture corresponding to stabilized PHB

SL5 formulation was prepared by solution casting technique, and the POM images were

compared at 208x of magnification (Fig. 4.25). In the physical mixture (Fig. 4.25b) some yellow inclusions dispersed in the PHB matrix were observed, which probably consisted of P200 particles. On the other hand, no inclusions were observed in the case of melt processed SL5 (Fig. 4.25a). Therefore, these results could represent a confirmation of the introduction of P200 in the PHB chains, with a consequent modification of their configuration and/or conformation and hence of the birefringence colour.

(a) (b)

Figure 4.25. POM images (208x) of SL5 (a) and a physical mixture PHB/P200 95/5 wt-

% (b)

4.1.2.2. Structural Characterization

13

C NMR spectra of PHB and P200 are shown in Figure 4.26. PHB signals were found at

around 19.9 ppm (CH

), 40.9 ppm (CH

), 67.7 ppm (CH), and 169.2 (C=O). The

characteristic signals relatives to P200 were found at 32.1 ppm (C-CH

), 55.3 ppm (C-

CH

), 59.0 ppm (O-CH

), 60.2 ppm (CH

-CH

-O), 70.5 ppm (CH

-CH

-O), 121.5, 123.3,

128.0, and 147.1 ppm (aromatic), 138.8 ppm (N=C=N), and 154.4 ppm (C=O). Being

P200 a commercial product, other small signals were attributed to the presence of

impurities. The strong signal at 77 ppm was attributed to chloroform and used as

reference. According to the mechanism hypothesized in Scheme 4.3, the reaction between

PHB and P200 should lead to the formation of N-acylurea, with the characteristic group

N-CO-N. In literature it is reported that the

C NMR signal of ureic carbonyl is

commonly found at around 150-155 ppm [213].

C NMR spectra of SL5 and a physical

mixture PHB/P200 95/5 wt-% in the ranges 200-0 ppm and 165-145 ppm are shown in

Figure 4.27a and 4.27b respectively.

Figure 4.26.

C NMR spectra of PHB and P200 (CDCl

solution)

The spectra of SL5 and the physical mixture (Fig. 4.27a) were substantially identical, showing all the signals relative to PHB carbon atoms without any significant shift with respect to the pristine polymer.

(a) (b)

Figure 4.27.

C NMR spectra of SL5 and a physical mixture PHB/P200 95/5 wt-% in

the ranges 200-0 ppm (a) and 165-145 ppm (b) (CDCl

solution)

Two small peaks were observed at 29.5 ppm and 70.5 ppm corresponding to the two signals 32.1 ppm (C-CH

) and 70.5 ppm (CH

-CH

-O), which were found to be the two more intense peaks in pristine P200. In the view that it was observed also in the physical mixture, the shift of the peak relative to C-CH

carbon from 32.1 to 29.5 ppm could be most likely attributed only to a physical interaction between P200 and PHB in the CDCl

solution. On the other hand, the appearance of new signals in the SL5 spectrum in the range 150-155 ppm was not observed (Fig. 4.27b). Although these results seem to be in contrast with the hypothesis formulated in the morphological analysis, suggesting that no reaction occurred between PHB and P200, it is also possible that the reaction occurred in a relatively low extent. In this case, the low amount of formed N-acylurea could not be detected by the NMR apparatus.

The assignments of the significant transmission FT-IR absorptions of PHB were already reported in Table 4.3. FT-IR spectrum of P200 is shown in Figure 4.28. P200 was characterized by the presence of two strong absorptions at 1112 and 2116 cm

, attributed to the stretching of the C-O bonds in the ethylene glycol chain and to N=C=N groups respectively [211, 213]. Absorbance relative to the C=O bond was found at 1729 cm

. Weaker absorption peaks typical of aromatic compounds were found at 1487 cm

and 1603 cm

(stretching of the C

C bond in the aromatic ring), 3024 cm

(stretching of the aromatic C-H bond), ca. 900 cm

(C-H bending out of the plane) and in the region 1100- 1200 cm

(C-H bending in the plane) [213]. The absorptions at around 2972 cm

, 2926 cm

, and 2870 cm

were attributed to the ethylene glycol aliphatic groups [213].

Figure 4.28. FT-IR spectrum of P200

The FT-IR spectra of pristine and P200 stabilized PHB in the range 2200-2000 cm

is shown in Figure 4.29. The persistence in the spectra of stabilized PHB of the peak at around 2116 cm

, whose intensity increases with increasing of P200 amount, was observed. This peak in the stabilized PHB formulations indicated the presence of unreacted P200.

Figure 4.29. FT-IR spectra of pristine components and P200 stabilized PHB in the range 2200-2000 cm

4.1.2.3. Molecular Weight Characterization

Weight average molecular weight (Mw) and polydispersity index (PDI) of pristine and P200 stabilized PHB measured by GPC traces from refractive index (RI) detector are reported in Table 4.28. The values were calculated as the average of two replicated formulations, with the relative standard deviation (SD).

Pristine PHB showed a molecular weight of 159 KDa, with a PDI of 2.22. The addition of 1, 3, and 5 wt-% of P200 promoted a regular increase of PHB Mw of 3%, 6%, and 10%

respectively. However, the presence of unreacted P200 was clearly detected by GPC UV

detector (not shown here). It is interesting to notice that, plotting PHB Mw data against

P200 content, a linear relation was found (Fig. 4.30). Concerning PDI, no dependence

from P200 content was observed.

Table 4.28.Weight average molecular weight (Mw) and polydispersity index (PDI) of pristine and P200 stabilized PHB

Sample Mw ± SD (KDa) PDI ± SD

PHB 159 ± 20 2.22 ± 0.02

SL1 164 ± 15 2.18 ± 0.13

SL3 168 ± 27 2.16 ± 0.18

SL5 176 ± 20 2.30 ± 0.01

a) SD is the standard deviation

Figure 4.30. Dependence of PHB molecular weight from P200 content

GPC traces of pristine and P200 stabilized PHB obtained from the refractive index detector are shown in Figure 4.31. Overall retention peaks in the range 8.5-16.5 min (Fig.

4.31a) clearly showed that both the onset and the peak retention times, located at around

10.3 min and 12.1 min respectively, were shifted to lower values when the content of

P200 was increased. Another interesting information was obtained by the behaviour of

the initial slope of the traces, in the range of retention time 9.5-11.0 min (Fig. 4.31b). In

fact, the slope rised at the increasing of P200 content, indicating that the addition of the

carbodiimide-based stabilizer promoted the formation of a fraction of PHB chains with a

higher molecular weight.