**Controlling the Isothermal Crystallization of Isodimorphic PBS-***ran***-PCL Random Copolymers by Varying Composition and Supercooling**

#### **Maryam Safari 1, Agurtzane Mugica 1, Manuela Zubitur 2, Antxon Martínez de Ilarduya 3, Sebastián Muñoz-Guerra <sup>3</sup> and Alejandro J. Müller 1,4,\***


Received: 1 December 2019; Accepted: 18 December 2019; Published: 20 December 2019

**Abstract:** In this work, we study for the first time, the isothermal crystallization behavior of isodimorphic random poly(butylene succinate)-*ran*-poly(ε-caprolactone) copolyesters, PBS-*ran*-PCL, previously synthesized by us. We perform nucleation and spherulitic growth kinetics by polarized light optical microscopy (PLOM) and overall isothermal crystallization kinetics by differential scanning calorimetry (DSC). Selected samples were also studied by real-time wide angle X-ray diffraction (WAXS). Under isothermal conditions, only the PBS-rich phase or the PCL-rich phase could crystallize as long as the composition was away from the pseudo-eutectic point. In comparison with the parent homopolymers, as comonomer content increased, both PBS-rich and PCL-rich phases nucleated much faster, but their spherulitic growth rates were much slower. Therefore, the overall crystallization kinetics was a strong function of composition and supercooling. The only copolymer with the eutectic composition exhibited a remarkable behavior. By tuning the crystallization temperature, this copolyester could form either a single crystalline phase or both phases, with remarkably different thermal properties.

**Keywords:** isodimorphism; random copolymers; crystallization; nucleation; growth rate

#### **1. Introduction**

Biocompatible and biodegradable polymers are being developed for a wide range of applications due to their potential to solve the environmental concerns caused by traditional nondegradable plastics [1–4]. Among the biodegradable polymers that have been most intensively studied are aliphatic polyesters such as Poly(glycolide) (PGA), Poly(L-lactide) (PLLA), Poly(ethylene succinate) (PES), Poly(butylene succinate) (PBS), and Poly(ε-caprolactone) (PCL) [5,6]. Although aliphatic polyesters have been used for many years in industrial, biomedical, agricultural, and pharmaceutical applications, there is still room for many improvements [7,8]. The synthesis of random copolyesters, using biobased comonomers, can overcome some of the drawbacks of biodegradable polyesters, such as slow biodegradation rate (due to high crystallinity degrees) and undesirable mechanical properties [9–11].

The properties of crystallizable random copolymers constituted by two semicrystalline parent components have been recently reviewed [12]. Depending on their ability to share crystal lattices, three different cases have been reported [12–14]: (a) total comonomer exclusion occurs when the chemical repeat units are very different and the crystal lattice of each one of the components cannot tolerate the presence of the other; (b) total comonomer inclusion or isomorphic behavior can only be obtained in cases where the components can cocrystallize in the entire composition range (as their chemical structures are very similar), forming a single crystal structure [15,16]; (c) an intermediate and complex case, where a balance between comonomer inclusion and exclusion occurs, leading to isodimorphic copolymers.

In isodimorphic random copolymers, at least one of the two crystalline phases includes some repeat units of the minor component in its crystal lattice. When the melting point is plotted as a function of composition a pseudo-eutectic behavior is commonly observed, where, on each side of the pseudo-eutectic point, only the crystalline phase of the major component is formed, which may contain a limited amount of the minor comonomer chains included in the crystal lattice [12].

In our previous works [10,17], we have synthesized and studied the morphology and crystallinity of poly (butylene succinate-*ran*-caprolactone) (PBS-*ran*-PCL) copolyesters. In situ wide angle X-ray scattering (WAXS) indicated that changes were produced in the crystalline unit cell dimensions of the dominant crystalline phase. In addition, differential scanning calorimetry (DSC) measurements showed that all copolymers could crystallize, regardless of composition, and their thermal transitions temperatures (i.e., *Tc* and *Tm*) went through a pseudo-eutectic point when plotted as a function of composition. Therefore, all this evidence demonstrated an isodimorphic behavior. At the pseudo-eutectic composition, both PBS-rich and PCL-rich phases can crystallize [10,17].

In the current work, we perform a detailed isothermal crystallization study of PBS-*ran*-PCL copolymers to determine the nucleation and crystallization kinetics of the copolyesters and study the influence of composition on the crystallization kinetics. This information is very important as it allows tailoring the properties of random copolymers as well as their applications. The analysis of the isothermal crystallization kinetics of PBS-*ran*-PCL was performed using differential scanning calorimetry (DSC), polarized light optical microscopy (PLOM), and in situ wide angle X-ray scattering (WAXS).

#### **2. Materials and Methods**

#### *2.1. Synthesis*

The PBS-*ran*-PCL copolymers were synthesized by a two-stage melt-polycondensation reaction. First, transesterification/ROP reaction of dimethyl succinate (DMS), 1,4-butanediol (BD), and ε-caprolactone (CL), and then polycondensation at reduced pressure, as reported in detail previously [17]. Samples are denoted in an abbreviated form, e.g., BSxCLy, indicating the molar ratio of each component determined by 1H-NMR, as subscripts (x and y). Table 1 shows molar composition, number- and weight-average molar mass and thermal transitions of the isodimorphic random copolyesters under study in this work.


**Table 1.** Molar composition determined by 1H-NMR, number- and weight-average molar mass determined by gel permeation chromatography (GPC), and thermal transitions determined by differential scanning calorimetry (DSC) (at 10 ◦C/min) of the materials employed in this work.

#### *2.2. Polarized Light Optical Microscopy (PLOM)*

A polarized light optical microscope, Olympus BX51 (Olympus, Tokyo, Japan), equipped with an Olympus SC50 digital camera and with a Linkam-15 TP-91 hot stage (Linkam, Tadworth, UK) (coupled to a liquid nitrogen cooling system) was used to observe spherulites nucleation and growth. Films with around 100 μm thickness were prepared by melting the samples in between two glass slides. For the isothermal experiments, the conditions were very similar to those employed during DSC measurements. The samples were heated to 30 ◦C above their melting point to erase their thermal history, and then were rapidly cooled from the melt at 60 ◦C/min to the selected isothermal crystallization temperature, *Tc*. Then, the sample was kept at *Tc* while the spherulites appeared and grew. To better compare the compositions in a fixed common crystallization temperature, *Tc* values were chosen at the same supercooling degree for PBS-rich compositions at Δ*T* = 40, 38, 36, 34, and 32 ◦C and for PCL-rich compositions at Δ*T* = 40, 39, 38, 37, and 36 ◦C. The supercooling was calculated from the equilibrium melting temperatures determined by Hoffman–Weeks extrapolations after isothermal crystallization in the DSC, see below.

#### *2.3. Di*ff*erential Scanning Calorimetry (DSC)*

Isothermal differential scanning calorimetry experiments were performed using a Perkin Elmer 8500 calorimeter equipped with a refrigerated cooling system Intracooler 2P, under a nitrogen atmosphere (with a flow of 20 mL/min) and calibrated with high purity indium and tin standards. The weight of the samples was about 5 mg and samples were hermetically sealed in standard aluminum pans.

To investigate the overall crystallization kinetics, an isothermal protocol was applied. First, the minimum isothermal crystallization temperature *Tc,min* was determined by trial and error following Müller et al. [18,19]. Samples were quenched to *Tc* values (estimated from the nonisothermal DSC runs) at 60 ◦C/min and then immediately reheated at 20 ◦C/min up to temperatures above the melting point of the crystalline phase involved. If any latent melting enthalpy is detected, this means that the sample was able to crystallize during the cooling to *Tc*, therefore, this *Tc* value cannot be used as *Tc,min* and a higher *Tc* value is explored.

After the *Tc* range was determined, the isothermal crystallization experiments were performed, closely following the procedure suggested by Lorenzo et al. [19]: (I) heating from room temperature to 30 ◦C above their melting point at 10 ◦C/min; (II) holding the sample for 3 min at that temperature to erase thermal history; (III) quenching the sample to a predetermined crystallization temperature (*Tc*) at 60 ◦C/min. *Tc* was in the range between −3 and 90 ◦C depending on composition; (IV) isothermal

crystallization until maximum saturation; (V) heating from *Tc* to 30 ◦C above the melting point of the sample at 10 ◦C/min, to record the melting behavior after the isothermal crystallization. This final melting run provided the values of apparent melting points that were employed to perform the Hoffman–Weeks extrapolation to calculate the equilibrium melting temperature of each material.

#### *2.4. Simultaneous WAXS Synchrotron Measurements*

To study the crystal structure during isothermal crystallization at the pseudo-eutectic point, the BS45CL55 sample was examined by in situ WAXS performed at beamline BL11-NCD at the ALBA Synchrotron radiation facility, Cerdanyola del Vallés, Barcelona, Spain. The samples in DSC aluminum pans were placed in a Linkam THMS-600 stage coupled to a liquid nitrogen cooling system. WAXS scans were taken periodically every 30 s during the isothermal crystallization. The energy of the X-ray source was 12.4 keV (λ = 1.0 Å). In the WAXS configuration, the sample-detector, Rayonix LX255-HS with an active area of 230.4 <sup>×</sup> 76.8 mm (pixel size: 44 <sup>μ</sup>m2) distance employed was 15.5 mm with a tilt angle of 27.3◦. The scattering vector was calibrated using chromium (III) oxide for WAXS experiments.

#### **3. Results**

We have studied previously [17] the nonisothermal crystallization behavior of the same PBS-ran-PCL random copolymers employed in this work. The results demonstrated that these copolymers exhibit an isodimorphic behavior.

Figure 1 presents a phase diagram for the PBS-ran-PCL system. These random copolymers exhibit a single-phase melt and a single glass transition temperature, as expected for random copolymers. Upon cooling from the melt, the materials are capable of crystallizing in the entire composition range, in spite of being random, as demonstrated by NMR studies [17]. The copolymers display a pseudo-eutectic point at the composition BS45CL55. This BS45CL55 copolymer is the only one in the series that can form two crystalline phases upon cooling from the melt, i.e., a PBS-rich phase and a PCL-rich phase (as evidenced earlier by WAXS and DSC [17]), hence the two melting point values reported in Figure 1 for this composition. To each side of the pseudo-eutectic point, a single crystalline phase is formed, either a PBS-rich phase (i.e., left-hand side of the eutectic) or a PCL-rich phase (i.e., right-hand side of the eutectic), with crystalline unit cells resembling those of PBS and PCL respectively.

**Figure 1.** Phase diagram based on data published in [17] on the nonisothermal crystallization of the PBS-*ran*-PCL copolymers under study. Additionally, equilibrium melting temperatures obtained in the present work by Hoffman–Weeks analysis of isothermally obtained data, are also included. The dashed vertical line indicates the pseudo-eutectic point. The dashed horizontal line indicates an arbitrary room temperature value.

In the present work, we performed isothermal crystallization studies and calculated the equilibrium melting temperatures (*T*<sup>0</sup> *<sup>m</sup>*) of homopolymers and copolymers by employing the Hoffman–Weeks extrapolation. Examples of Hoffman–Weeks plots can be found in Figure SI-3, while Figure 1 reports the variation of the equilibrium melting temperatures obtained with composition. The *T*<sup>0</sup> *<sup>m</sup>* values show a similar trend with composition as the apparent melting peak temperatures determined by DSC during nonisothermal experiments, and they also display a pseudo-eutectic point. These *T*<sup>0</sup> *m* values will be employed throughout this paper, as they are needed to fit the Lauritzen and Hoffman nucleation and crystallization theory to analyze the experimental data.

The phase diagram shown in Figure 1 illustrates the versatility of isodimorphic copolymers. It is well known that the optimal mechanical properties in terms of ductility and toughness of thermoplastic semicrystalline materials are generally observed at temperatures in between *Tg* and *Tm*. Thanks to random copolymerization, the copolymers exhibit a single *Tg* value that is independent of the melting point of the phase (or phases in the case of the composition at the pseudo-eutectic point) that is able to crystallize. This remarkable behavior provides a separate control of *Tg* and *Tm* which cannot be obtained in homopolymers. Additionally, as Figure 1 shows, depending on composition, the samples can be molten at room temperature or they can be semicrystalline. Such wide range of thermal properties can lead to fine tuning mechanical properties and crystallinities to tailor applications.

#### *3.1. Nucleation Kinetics Studied by PLOM*

Counting the number of spherulites in PLOM experiments is the usual way of obtaining nucleation data by assuming that each spherulite grows from one heterogeneous nucleus. In this work, we studied the nucleation kinetics by determining the nucleation density as a function of time by PLOM, from which nucleation rates can be calculated.

Figure 2 shows four examples of plots of the nucleation density ρnuclei (nuclei/mm3) as a function of time for neat PBS, neat PCL, and two sample copolymers. The rest of the data can be found in the Supplementary Information (Figure SI-1). The nucleation density increases almost linearly with time at short times, then it tends to saturate. The number of heterogeneous nuclei that are activated at longer times increases as nucleation temperature decreases, a typical behavior of polymer nucleation [20]. As expected, the nucleation density at any given time increases as *Tc* decreases, because the thermodynamic driving force for primary nucleation increases with supercooling [21].

Figure 3 shows plots of nucleation density versus temperature taken at a constant nucleation time of 100 s for neat PBS and PBS-rich copolymers (Figure 3a) and 10 min in the case of PCL and BS11CL89 copolymer (Figure 3b).

PBS exhibits the lowest nucleation density of all samples, therefore, the largest spherulites (see Figure 5 below). As the amount of CL comonomer increases in the PBS-rich copolymers (Figure 3a), the nucleation density increases, as well as the supercooling needed for nucleation. In the case of PBS-rich copolymers, Figure 3a (and Figure SI-1 in the SI) shows nucleation data for seven different copolymers, with compositions ranging from 91% to 45% PBS.

The dependence of nucleation density on supercooling, can be observed in Figure SI-2. The data presented in Figure 3a can be reduced to a supercooling range between 32 and 40 ◦C, i.e., only 8 ◦C. This means that a large part of the horizontal shift in the curves of Figure 3a (spanning nearly 60 ◦C in crystallization temperature) is due to changes in supercooling. These changes are caused by the variations in equilibrium melting temperatures with composition (see Figure 1).

PCL has a higher nucleation density than PBS when compared at equal supercoolings (see Figure SI-2b). When a small amount of BS comonomer is incorporated, as in random copolymer BS11CL89, the nucleation density increases significantly (Figure 3b). Due to the very high nucleation density of the other PCL-rich composition copolymers (with higher amounts of PBS), it was impossible to determine their nucleation kinetics. Examples of the microspherulitic morphologies obtained for such PCL-rich copolymers can be observed in Figure 5 below.

It is interesting to note than in both sides of the pseudo-eutectic point (i.e., the PBS-rich side represented in Figure 3a and the PCL-rich side represented in Figure 3b, see also Figure 1), the copolymers exhibit higher nucleation density than their corresponding homopolymers. This behavior could be somewhat analogous to what has been observed in long-chain branched polylactides (PLLAs) [22] or long-chain branched polypropylenes (PPs) with respect to linear analogs [23]. The interruption of crystallizable linear sequences with defects has been reported to increase nucleation density although the reasons are not clear. In the present case, the linear crystallizable sequence of PBS, for instance, is being changed by the introduction of randomly placed PCL repeat units. Even though the random copolyesters can form a single phase in the melt, there may be at the segmental level, some preference for PBS-PBS local chain segmental contacts in comparison to less favorable PBS-PCL contacts. We speculate that this may drive the enhancement of nucleation, but more in-depth studies would be needed to ascertain the exact reason for this behavior.

**Figure 2.** Nucleation kinetics data obtained by polarized light optical microscopy (PLOM): (**a**) PBS; (**b**) BS78CL22; (**c**) PCL; (**d**) BS11CL89. Nuclei density as a function of time at different crystallization temperatures for the indicated samples.

**Figure 3.** (**a**) Nuclei density during isothermal crystallization as a function of *Tc* at a constant time of 100 s for PBS-rich compositions and (**b**) at 10 min for PCL-rich compositions.

The Fisher–Turnbull nucleation theory [24] can be used to quantify the activation free energy of primary nucleation. This theory gives the steady-state rate of primary nucleation per unit volume and time, *I* = dN/dt, for a heterogeneous nucleation process on a preexisting flat surface (or heterogeneous nucleus) as:

$$\log I = \log I\_0 - \frac{\Delta F^\*}{2.3kT} - \frac{16\alpha \sigma\_c (\Delta \sigma) T\_m^{\prime 2}}{2.3kT \left(\Delta T\right)^2 \left(\Delta H\_v\right)^2} \tag{1}$$

where *I*<sup>0</sup> is related to the diffusion of polymeric segments from the melt to the nucleation site, Δ*F*\* is a parameter proportional to the primary nucleation free energy, and σ and σ*<sup>e</sup>* are the lateral and fold surface free energies, respectively. Δ*T* is the supercooling defined as Δ*T* = *T*<sup>0</sup> *<sup>m</sup>* <sup>−</sup> *Tc* and *<sup>T</sup>*<sup>0</sup> *<sup>m</sup>* is the equilibrium melting point. Δσ is the interfacial free energy difference, given by:

$$
\Delta \sigma = \sigma + \sigma\_{\text{s}\circ \text{\textquotedblleft}} - \sigma\_{\text{s}\circ \text{\textquotedblright}\prime} \tag{2}
$$

in which σ*s*/*<sup>c</sup>* is the crystal-substrate interfacial energy and σ*s*/*<sup>m</sup>* is the melt-substrate interfacial energy. Therefore, Δσ can be considered proportional to the surface tension properties of the substrate, polymer crystal and polymer melt. The interfacial free energy difference is a convenient way to express the nucleating ability of the substrate towards the polymer melt.

In this work, the values of *T*<sup>0</sup> *<sup>m</sup>* (listed in Table SI-1 and plotted in Figure 1) were obtained by isothermal crystallization DSC experiments followed by Hoffman–Weeks extrapolations (see Figure SI-3). Δ*Hv* is the volumetric melting enthalpy (J/m3) and it was estimated by Δ*HV* = Δ*H*<sup>0</sup> *<sup>m</sup>* × ρ, so that ρ = 1.26 g/cm3 and Δ*H*<sup>0</sup> *<sup>m</sup>* = 213 J/g for neat PBS [25] and ρ = 1.14 g/cm<sup>3</sup> and Δ*H*<sup>0</sup> *<sup>m</sup>* = 139.5 J/g for neat PCL [26].

In this work, we employed the value of Δ*H*<sup>0</sup> *<sup>m</sup>* = 213 J/g for neat PBS and PBS-rich phase composition that has been recently obtained by some of us [25]. This value was determined employing a combined DSC and X-ray diffraction method using isothermal crystallization data. This experimentally extrapolated value is higher than that of Δ*H*<sup>0</sup> *<sup>m</sup>* = 110 J/g, estimated empirically by the group contribution method [27], but very close to the value of 210 J/g reported by Papageorgiou et al. [28].

The values of the nucleation rate I were calculated from the initial slope (i.e., at short measurement times, where linear trends were obtained) of the plots shown in Figure 2 and Figure SI-1. Figure 4a shows log I as a function of CL-unit molar fraction for a constant supercooling of Δ*T* = 40 ◦C. The nucleation rate strongly depends on copolymer composition. Adding a comonomer randomly along the chain to either PBS or PCL largely increases the nucleation rate. In the PBS-rich composition side (to the left of the pseudo-eutectic point signaled by a vertical line in Figure 4a) the nucleation rate increases up to 7.5 times with respect to neat PBS, as the amount of PCL units in the random copolymer increases. Neat PCL nucleates faster than neat PBS. In the PCL-rich composition side, only one copolymer was measured (whose nucleation rate increased two-fold with respect to neat PCL), as increasing PBS content towards the pseudo-eutectic point increased nucleation rate so much that measurements were no longer possible.

**Figure 4.** Nucleation rate (*I*) data: (**a**) log I as a function of copolymer composition, expressed as mol % of ε-caprolactone (CL) units, taken at a constant supercooling of Δ*T* = 40 ◦C. The segmented vertical line is drawn to indicate the pseudo-eutectic composition. (**b**) Plot of *log I* versus 1/*Tc*(Δ*T*) <sup>2</sup> for PBS-rich compositions. The black lines represent fittings to the Turnbull–Fisher equation (Equation (1)). (**c**) Interfacial free energy difference (Δσ) as a function of composition. The segmented vertical line is drawn to indicate the pseudo-eutectic composition.

Figure 4b shows the Turnbull–Fisher plots for PBS and PBS-rich compositions based on Equation (1). Turnbull–Fisher plots for PCL and BS11CL89 copolymer are presented in the supplementary information (Figure SI-4). The nucleation data can be successfully fitted with the linearized version of Equation (1). From the slope, a value of the interfacial free energy difference (Δσ) can be obtained.

Small values of Δσ are indicative of good nucleation efficiency since a lower amount of interfacial energy is required to form the crystal–substrate interface. Table 2 reports a value of Δσ for PBS equal to 1.97 erg/cm2. As seen in Figure 4c (and Table 2), this interfacial free energy difference progressively decreases in the copolymers as the amount of CL comonomer increases, indicating that the primary nucleation process is facilitated by copolymerization with PCL until the pseudo-eutectic point is reached. On the right-hand side of the pseudo-eutetic point in Figure 4c, PCL has a Δσ value of 1.53 erg/cm2, which is, as expected, smaller than that of PBS, as PCL has a larger nucleation density at equivalent supercoolings than PBS. The copolymer B11CL89 shows an even smaller value of Δσ, as the incorporation of PBS in the copolymer increases its nucleation capacity.

**Table 2.** Primary nucleation and growth isothermal kinetics data parameters according to Equations (1) and (3) derived from experimental results obtained by PLOM.


*<sup>a</sup> R*<sup>2</sup> is the correlation coefficient for the fitting of the nucleation kinetics with the Turnbull–Fisher model (Equation (1)), *log I* vs. 1/*T*(Δ*T*) 2. *<sup>b</sup> R*<sup>2</sup> is the correlation coefficient for the fitting of the Lauritzen–Hoffman model (Equation (3)), *lnG* + *U*\*/*R*(*Tc* − *T*0) vs. 1/*f.Tc.*Δ*T*.

#### *3.2. Kinetics of Superstructural Growth (Secondary Nucleation) by PLOM*

PBS, PCL, and all the random copolymers prepared in this work exhibited spherulitic superstructural morphologies. Examples of the spherulites obtained at a constant supercooling value of 40 ◦C can be observed in Figure 5. Both PBS and PCL exhibited well-developed spherulites without banding. PBS-rich copolymers that contain more than 34% PCL exhibit clear banding. This is consistent with previous works indicating that the addition of diluents (for PBS-rich compositions, crystallization occurs while PCL chains are in the liquid state) to several polyesters induces banding [29,30].

Isothermal crystallization experiments were performed to follow the growth of spherulites as a function of time using PLOM. The growth rate was calculated from the slope of spherulite radius versus time plots, which were always observed to be highly linear [18,31].

The experimental growth rates are plotted as a function of the isothermal crystallization temperatures employed in Figure 6a with a linear scale and in Figure 6b with a log scale, so that differences in G values for PBS-rich samples with PCL contents larger than 22% are observed. The incorporation of PCL repeat units in the random copolymers have a dramatic influence on the growth rate of the PBS-rich phase spherulites, as G decreases up to 3.5 orders of magnitude (Figure 6b). The decrease in G values with comonomer incorporation for the PBS rich copolymers is due to two reasons. Firstly, as in any isodimorphic copolymer, there is a competition between inclusion and exclusion of repeat units within the PBS crystal lattice, where exclusion typically predominates. Secondly, incorporation of PCL repeat units in the copolymer chains reduces *Tg* values (as shown in Figure 1), thereby causing a plasticization effect on the PBS-rich phase. In the case of the PCL-rich

compositions, the spherulitic growth rate was determined for only one copolymer (i.e., BS11CL89), as in the other cases, as pointed out above, the nucleation rate and nucleation density were so high, that it was impossible to measure the extremely fast growth of very small spherulites. For this copolymer, the growth rate decreased in relative terms (see Figure 6c) by a factor of approximately 2.5 at a supercooling of 40 ◦C.

**Figure 5.** PLOM micrographs after isothermal crystallization at Δ*T* = 40 ◦C for the indicated samples.

The data presented in Figure 6a are plotted as a function of supercooling in the Supplementary Information (Figure SI-5). The PBS-rich growth rate data is shifted horizontally and but there is no overlap in the y axis values. If we were dealing with a simple solvent effect, the growth rate curves at different compositions should completely overlap in a master curve when plotted as a function of supercooling. The lack of superposition is due to the fact that PCL repeat units are randomly incorporated and covalently bonded with the PBS repeat units. The interruption of crystallizable PBS repeat units (by the majority of PCL repeat units that are excluded from the crystals) makes more difficult the secondary nucleation process.

Figure 6c shows how the growth rate depends on composition at a constant supercooling of 40 ◦C. The trend is the opposite as that obtained for primary nucleation (compare Figure 6c with Figure 4a). In order to quantify the restrictions imposed by the comonomer on the crystallization of the major component, we employed the Lauritzen and Hoffman theory, as it allows the calculation of energetic terms related to the secondary nucleation process (i.e., growth process).

The Lauritzen and Hoffman (LH) nucleation and growth theory [32] was used to fit the spherulitic growth rate data as a function of isothermal crystallization temperature, according to the following equation:

$$G = G\_0 \exp\left[\frac{-lI^\*}{R(T\_c - T\_0)}\right] \left[\frac{-K\_\%}{fT\left(T\_m^0 - T\_c\right)}\right] \tag{3}$$

where *G*<sup>0</sup> is the growth rate constant that includes all the terms that are temperature-insensitive, *U*\* is the transport activation energy which characterizes molecular diffusion across the interfacial boundary between melt and crystals (in this work, we employ a constant value of 1500 cal/mol). *Tc* is the crystallization temperature and *T*<sup>0</sup> is a hypothetical temperature at which all chain movements freeze (taken as *T*<sup>0</sup> = *Tg* <sup>−</sup> 30 ◦C); *<sup>T</sup>*<sup>0</sup> *<sup>m</sup>* is the equilibrium melting temperature and *f* is a temperature correction factor given by the following expression: *f* = 2*Tc*/(*Tc* + *T*<sup>0</sup> *<sup>m</sup>*).

The equilibrium melting temperatures *T*<sup>0</sup> *<sup>m</sup>* were estimated by the Hoffman–Weeks linear extrapolation (Figure SI-3 and Table SI-1). The parameter *K<sup>G</sup> <sup>g</sup>* is proportional to the energy barrier for secondary nucleation or spherulitic growth and is given by:

$$K\_{\%}^{G} = \frac{jb\_{0}\sigma\sigma\_{c}T\_{m}^{0}}{k\Delta h\_{f}},$$

where *j* is assumed to be equal to 2 for crystallization in the so-called Regime II, a regime where both secondary nucleation at the growth front and the rate of spread along the growing crystal face are comparable [26]. The other terms in the equation are the width of the chain *bo*, the lateral surface free energy σ, the fold surface free energy σ*e*, the Boltzman constant k, and the equilibrium latent heat of fusion, Δ*H*<sup>0</sup> *m*.

Plotting *lnG* + <sup>−</sup>*<sup>U</sup> <sup>R</sup>*(*Tc*−*T*0) versus 1/*Tc*(Δ*T*)*<sup>f</sup>* (i.e., the Lauritzen and Hoffman plots) gives a straight line and its slope and intercept are equal to *K<sup>G</sup> <sup>g</sup>* and *G*<sup>0</sup> respectively. Examples of LH plots can be found in the Supplementary Information, Figure SI-6. Having the value of *K<sup>G</sup> <sup>g</sup>* , the magnitude of σσ*<sup>e</sup>* can be calculated from Equation (5). In order to calculate separately the values of σ and σ*e*, the following expression can be used [33]:

$$
\sigma = 0.1 \Delta h\_f \sqrt{a\_0 b\_0} \tag{5}
$$

where *a*0*b*<sup>0</sup> is the cross sectional area of the chain. To obtain the parameters of the LH theory, the following values were used for neat PBS and BS-rich compositions [34,35]: *a*<sup>0</sup> = 5.25 Å and *b*<sup>0</sup> = 4.04 Å, and for neat PCL and CL-rich compositions [36]: *a*<sup>0</sup> = 4.52 Å and *b*<sup>0</sup> = 4.12 Å.

Finally, *q*, the work done by the macromolecule to form a fold is given by [33]:

$$q = 2a\_0 b\_0 \sigma\_{\iota^\*}.\tag{6}$$

The solid lines in Figure 6a,b correspond to fittings to Equation (3). Table 2 shows that *K<sup>G</sup> <sup>g</sup>* values (which are proportional to the energy barrier for spherulitic growth) for the PBS-rich crystal phase tend to increase as PCL repeat units are incorporated in the random copolymers until a maximum value is reached at the pseudo-eutectic point. Similar trends are observed for the fold surface free energy and for the work done to form folds.

A plot of fold surface free energy versus composition can be found in the Supplementary Information (Figure SI-7). These results quantitatively measure how comonomer incorporation makes

difficult the spherulitic growth of the PBS-rich phase. A similar interpretation can be done to the mirror values presented in Table 2 for PCL and the BS11CL89 copolymer with respect to the PCL phase.

The results presented in the two sections above can be summarized by comparing Figure 4 with Figure 6. The incorporation of comonomers at each side of the eutectic causes an increase in the nucleation density and nucleation rate but at the same time a decrease in spherulitic growth rate. These two processes, primary nucleation and growth are combined when a semicrystalline polymer is crystallized from the melt. Their simultaneous effect can be ascertained by determining overall crystallization kinetics by DSC.

**Figure 6.** Spherulitic growth rates determined by PLOM. (**a**) Growth rate, G, as a function of *Tc*. (**b**) Same data as in (**a**) but G is plotted on a logarithmic scale. The black solid lines are fits to the experimental data performed with the Lauritzen and Hoffman theory (L-H). (**c**) G versus CL-unit content at Δ*T* = 40 ◦C. The black line is an arbitrary polynomial fit drawn to guide the eye.

#### *3.3. Overall Crystallization Kinetics Studied by DSC*

The overall isothermal crystallization kinetics considers both nucleation and growth, and can be conveniently determined by isothermal DSC experiments. Figure 7a,b shows the experimental overall crystallization rate expressed as the inverse of the crystallization half time (τ50%). By DSC, we were able to determine the overall isothermal crystallization kinetics for both homopolymers and all copolymers (five copolymers where only the PBS-rich phase crystallized and four copolymers where only the PCL-rich phase can crystallize).

For the special composition at the eutectic point (i.e., BS45CL55) that shows two crystalline phases, PBS-rich phase and PCL-rich phase, we performed isothermal crystallization using different protocols. For the PBS-rich phase crystallization, isothermal DSC experiments were performed at temperatures where the PCL-rich phase is in the melt and cannot crystallize, while in the PCL-rich phase, a special protocol was adopted to previously crystallize the PBS-rich phase to saturation (see experimental part).

Figure 7a,b shows the strong dependence of the overall crystallization rate and the temperature range where measurements were possible on copolymer composition. In the case of the PBS and all PBS-rich compositions, the overall crystallization proceeds from a single-phase melt. Upon increasing PCL content, the amount of the crystallizable PBS-rich phase decreases and there will be more molten PCL component causing a plasticization ("solvent effect"). In addition, the effect of PCL exclusion in the PBS-rich crystal lattice may cause some further reduction in crystallization rate. Figure 7a shows that the temperature needed for crystallization decreases as PCL content in the copolymer increases, while the overall crystallization rate measured at the minimum *Tc* value possible tends to decrease.

**Figure 7.** Overall crystallization versus isothermal crystallization temperature for neat PBS and PBS-rich compositions (**a**) and for neat PCL and PCL-rich compositions (**b**) versus *Tc*. Continuous lines correspond to the fitting of the Lauritzen–Hoffman theory with the parameters in Table 3. Changes of *Tc* versus CL-unit content in a constant rate (1/τ50% = 1 min<sup>−</sup>1) (**c**). Changes of inverse of half-crystallization time, 1/τ50% versus CL-unit content in a constant supercooling degree, Δ*T* = 45 ◦C (**d**).

Figure 7c plots the crystallization temperature needed to obtain the same overall crystallization rate of 1 min<sup>−</sup>1. These Tc values monotonically decrease with PCL content until the pseudo-eutectic region is reached. On the PCL-rich side, Figure 7b,c shows similar results, as the crystallization temperatures needed to crystallize the PCL phase decrease as PBS repeat units are added to the copolymer.

To check if the supercooling is playing a major role upon changing composition, we plot the data contained in Figure 7a,b as a function of Δ*T* in Figure 8a,b, respectively. Surprisingly, the trends are quite different depending on the phase under consideration, or the composition range.

Figure 8a shows that the curves of PBS-rich overall growth rate data that originally spanned a *Tc* range of approximately 90 ◦C (in Figure 7a) are now within 30 ◦C in supercooling, attesting for the thermodynamic compensation of the solvent effect, as the PCL-rich phase is in the melt. In fact, the curves of PBS and BS91CL9 completely overlap, while that of BS78CL22 is relatively close to that of neat PBS. However, beyond 22% CL incorporation in the copolymer, the samples require much larger supercooling to crystallize. It is clear that the dominant factor to the left of the eutectic point is the growth rate, as the results presented in Figures 6a and 7a imply an overall crystallization rate reduction with PCL incorporation in the copolymer, both in terms of crystallization temperature or supercooling. In spite of the increase in nucleation density and nucleation rate (see Figures 3 and 4) with PCL incorporation in the copolymer, it is the very large decrease in growth rate (of up to three orders of magnitude, see Figure 6b) that dominates, leading to a decrease in overall crystallization rate (Figures 7a and 8a).

**Figure 8.** Overall crystallization for neat PBS and PBS-rich compositions (**a**) and for neat PCL and PCL-rich compositions (**b**) versus supercooling temperature.

Figure 7d represents the overall crystallization rate as a function of composition for a constant supercooling of 45 ◦C. In the PBS-rich side of the pseudo-eutectic region (left-hand side of Figure 7d), only three data points are plotted, as they are the only ones that could be measured at such constant supercooling value (check Figure 8b). The trend clearly shows a significant decrease in overall crystallization rate as CL unit content increases, as expected from Figures 7a and 8a and the discussion above. Even though the value of supercooling is not exactly the same (with only 5 ◦C difference), a comparison with Figures 4a and 6c clearly indicates that the PBS-rich phase overall crystallization is dominated by growth rate.

Figure 8b shows remarkable results for the PCL-rich phase overall crystallization. The 1/τ50% curves versus temperature span a temperature range of 65 ◦C (Figure 6b). When they are plot as a function of supercooling, they only span 15 ◦C. However, they do not overlap, as would be expected for a simple solvent effect. In fact, the supercooling needed for crystallization of the PCL-rich phase remarkably decreases as PBS repeat units are included in the copolymer. The results indicate an acceleration of the overall crystallization rate (at constant supercooling) that can only be explained by the increase in both nucleation density and nucleation rate. We were only able to measure the increase in nucleation density and nucleation rate in Figures 3 and 4 for neat PCL and BS11CL89, as further incorporation of BS units increased the nucleation density so much that measurements by polarized

optical microscopy of nucleation rate became impossible. Hence, we are convinced that primary nucleation enhancement upon PBS repeat unit incorporation in the copolymers is the reason behind the acceleration of the overall crystallization kinetics, when this is considered in terms of supercooling. In the right-hand side of the pseudo-eutectic point in Figure 7d, the increase in overall crystallization rate at a constant supercooling of 45 ◦C can be appreciated.

The Lauritzen and Hoffman theory can also be applied to fit the overall crystallization data presented above. Equation (3) has to be modified to employ, as a characteristic rate, the inverse of the half-crystallization time determined by DSC, as follows [37]:

$$\mathbf{M}^{\prime}\!/\tau\_{90\text{v}} = 1 \cdot \left[\tau\_{90\text{v}}\exp\left[\frac{\mathcal{U}}{R(T\_c - T\_0)}\right]\right] \frac{-K\_{\rm g}^{\rm r}}{fT\left(T\_m^0 - T\_c\right)}\bigg|\_{\rm r} \tag{7}$$

where all the terms have been defined above, except for *K*<sup>τ</sup> *<sup>g</sup>*, which now represents a parameter proportional to the energy barrier for both primary nucleation and spherulitic growth. The superscript τ is used to indicate its origin (coming from DSC data, and hence from fitting 1/τ50% versus crystallization temperature). In this way, it is different from *K<sup>G</sup> <sup>g</sup>* , defined in Equation (3), derived from growth rate data and therefore proportional just to the free energy barrier for secondary nucleation or growth. The solid lines in Figure 7a,b and Figure 8a,b represent the fits to the Lauritzen and Hoffman theory. Table 3 on the other hand reports all the relevant parameters.

There are no results in the literature regarding the isothermal crystallization of PBS-ran-PCL copolymers, therefore, we cannot compare the parameters reported in Tables 2 and 3 with literature values. In the case of the homopolymers, Wu et al. [38] and Papageorgiou et al. [39] reported *K*<sup>τ</sup> *<sup>g</sup>* values for neat PBS equal to 1.157 <sup>×</sup> <sup>10</sup><sup>5</sup> and 2.64 <sup>×</sup> 105 K2, respectively, which are close to the value obtained in this work, i.e., 2.04 <sup>×</sup> <sup>10</sup><sup>5</sup> K2. A value for *<sup>K</sup><sup>G</sup> <sup>g</sup>* for neat PBS reported [39] equal to 1.88 <sup>×</sup> <sup>10</sup><sup>5</sup> has been reported, that is somewhat higher than that obtained in this work, i.e., 0.87 <sup>×</sup> 105 K2. For neat PCL, the energetic parameters previously reported based on fits of the Lauritzen and Hoffman theories [40] are in the same order of magnitude as those reported here [33].

Figure 9 plots both Kg values, obtained by PLOM (*K<sup>G</sup> <sup>g</sup>* ) and DSC (*K*<sup>τ</sup> *<sup>g</sup>* ) as a function of CL-unit molar content. As expected, all *K*<sup>τ</sup> *<sup>g</sup>* values are larger than *K<sup>G</sup> <sup>g</sup>* values, as DSC measurements take into account both nucleation and growth, while PLOM measurements considered only growth (see more details in [37]).

In the case of *K<sup>G</sup> <sup>g</sup>* values, the trends observed are expected in view of the results obtained in Figure 6. The energy barrier for crystal growth increased with CL-unit molar content, since growth rate decreased as comonomer incorporation increased. On the other hand, when we analyzed the results obtained for *K*<sup>τ</sup> *<sup>g</sup>* in Figure 9, we noticed that there is a clear asymmetry depending on which side of the pseudo-eutectic region the material is. On the PBS-rich side (left-hand side of Figure 9), *K*<sup>τ</sup> *<sup>g</sup>* values rapidly increased upon CL units addition. This is expected from the results presented in Figure 7d, where a large decrease in overall crystallization rate for the PBS-rich side of the composition range can be observed.

In the case of the PCL-rich composition range, we would have expected a decrease in *K*<sup>τ</sup> *<sup>g</sup>* values with PCL content increases according to Figure 7d. Instead, we observe in Figure 9 that the energy barrier for both nucleation and growth does not significantly change with composition (see right-hand side of Figure 9). We have to remember that for the PCL-rich copolymers the situation is particularly complicated as the nucleation density and nucleation rate increase with CL-unit content but the growth rate decreases. Hence, even though according to Figure 7d the overall crystallization rate at constant supercooling seems to be dominated by primary nucleation, the values of *K*<sup>τ</sup> *<sup>g</sup>* are obtained from the slope of the Lauritzen and Hoffman plots that take into account the full range of supercoolings where the measurements were taken. Therefore, it seems that when the overall energy barrier is considered, there is a balance between nucleation and growth which keeps the *K*<sup>τ</sup> *<sup>g</sup>* values constant with composition.

**Figure 9.** Kg versus CL-unit molar fraction that obtained for PLOM experiments (*K<sup>G</sup> <sup>g</sup>* ) and DSC experiments (*K*<sup>τ</sup> *<sup>g</sup>* ).

**Table 3.** Parameters obtained from fitting the DSC data presented in Figure 7a,b to the Lauritzen and Hoffman model (Equation (7)).


*<sup>R</sup>*<sup>2</sup> is the correlation coefficient for the Lauritzen–Hoffman (Equation (7)) linear plots ln(1/τ 50%) <sup>+</sup> *<sup>U</sup>*\*/*R*(*Tc* − *<sup>T</sup>*0) vs. 1/*f*·*Tc*·Δ*T*.

#### *3.4. Double Crystallization at the Pseudo-Eutectic Point*

For the copolymer whose composition is within the pseudo-eutectic point, i.e., BS45CL55, we performed isothermal crystallization in a wide range of crystallization temperatures *Tc*, to find the temperature region where only one of the phases, PBS or PCL, is able to crystallize. Figure 10 shows the heating DSC scan recorded at 10 ◦C/min for BS45CL55 sample after it was isothermally crystallized at the indicated *Tc* values.

At least five different endotherms can be found upon close examination of Figure 10 and we indicated with dashed lines how these endotherms approximately shift depending on the *Tc* values employed before heating the samples. The first melting peak *T*m1 is present in all melting curves and its location is almost at 7 ◦C higher than the crystallization temperature. This peak has been traditionally regarded as the melting of thin crystals formed during the secondary crystallization process [41]. The second peak, labeled *T*m2, appeared at *Tc* values lower than −6 ◦C and corresponds to the melting of PCL-rich crystals. The third peak or *T*m3 labeled peak in Figure 10 highly depends on the isothermal crystallization temperature and corresponds to the melting of the PBS-rich crystals, which were formed during the isothermal crystallization.

In addition, a melting peak (*T*m5) at around 50 ◦C and another one just below it (*T*m4) were observed. The melting peak labeled *T*m5 corresponds to the melting of PBS crystals that have reorganized during the heating scan, and have a melting point which is almost constant at around 50 ◦C, regardless of the crystallization temperature [42,43]. The *T*m2 and *T*m3 peaks increase almost linearly with increasing

*Tc*. As shown in Figure 10, *T*m3 disappeared in the DSC heating curves where the crystallization temperature is less than −9 ◦C.

The morphologies obtained after isothermal crystallization at three selected temperatures can be observed in Figure 10b–d. As it will be shown below, WAXS experiments confirmed that at very low *Tc* values including −12 ◦C, only PCL-rich crystals can be formed. Figure 10d shows small spherulites that were formed at *Tc* = −4 ◦C with spherulites size around 10 μm. At *Tc* = −8 ◦C, where both PCL and PBS crystals can form, there are two crystals sizes, one with 4 μm radii (PBS crystals) and another with around 1.5 μm size (PCL crystals), see Figure 10d. Figure 10b shows only PCL crystals with small spherulites size (less than 1 μm) at *Tc* = −12 ◦C.

We performed in situ synchrotron WAXS experiments for the sample at the pseudo-eutectic point, to clarify the temperature range of crystallization of the PBS-rich and the PCL-rich phases and corroborate the assignment of the thermal transitions in Figure 10. These experiments were performed during isothermal crystallization (for 20 min) at three different *Tc* values chosen from three different crystallization regions in Figure 10.

Figure 11a–c shows selected real-time WAXS diffractograms for BS45CL55 (i.e., the sample at the pseudo-eutectic point) measured during isothermal crystallization at −12, −9, and −6 ◦C. If the sample shows characteristic reflections at q = 13.9 and 16.1 nm<sup>−</sup>1, they correspond to the PBS (020) and (110) crystallographic planes. If the sample exhibits reflections at q = 15.3 and 17.4 nm<sup>−</sup>1, they belong to the PCL (110) and (200) planes [10].

**Figure 10.** (**a**) DSC heating runs (at 10 ◦C/min) after isothermal crystallization at different temperatures. See text for the explanation of the color code employed. PLOM micrographs after isothermal crystallization at −12 ◦C (**b**), at −8 ◦C (**c**), and at −4 ◦ C (**d**) for the BS45CL55 sample.

Changes in the crystallization temperature strongly affect the diffraction pattern at the pseudo-eutectic point. As can be seen in Figure 11, at −12 ◦C only the PCL-rich phase is able to crystalize (Figure 11a) while at −6 ◦C (Figure 11c) only the PBS-rich phase crystallizes. On the other hand, at the intermediate *Tc* value of −9 ◦C, both PBS-rich and PCL-rich phases can crystallize. If the DSC curves of Figure 10 are considered again, the WAXS assignments are consistent with the heating runs after crystallization for all samples crystallized at −9 ◦C and higher. In the case of low crystallization temperatures, i.e., below −9 ◦C, it should be noted that WAXS indicate that only the PCL-rich phase can crystallize. The DSC heating runs shown in Figure 10 also show melting transitions

corresponding to the melting of PBS-rich phase. These PBS-rich phase crystals must be formed by cold-crystallization during the heating scan for the samples crystallized at −10, −12, and −14 ◦C in Figure 10. In fact, upon close examination of Figure 10, the end of a cold crystallization process can be observed just after the melting peak of the PCL-rich phase crystals.

Taking into account the WAXS and DSC results presented in Figures 10 and 11, the DSC curves in Figure 10 were plotted with a color code to indicate which phases can crystallize during isothermal crystallization depending on the *Tc* values employed. If the *Tc* values are −10 ◦C or lower, only the PCL-rich phase can crystallize, and the curves were arbitrarily plotted in red in Figure 10a. If the *Tc* values are between −9 and −7 ◦C (including these two temperatures), both the PCL-rich and the PBS-rich phases can crystallize (green curves in Figure 10). Finally, if the *Tc* temperatures are −6 ◦C and above, only the PBS-rich phase can crystallize (blue curves in Figure 10).

The pseudo-eutectic sample, BS45CL55, exhibits a very interesting phase behavior, as depending on the crystallization conditions, one or both phases can be formed. We have studied previously the nonisothermal crystallization of the same copolymers employed here [17]. It is interesting to note that under nonisothermal conditions, the cooling rate employed determines which phase can crystallize and also if one or two phases are formed. In this work, on the other hand, we show that one or two phases can be formed depending on the isothermal crystallization temperature chosen. Therefore, the properties of this isodimorphic copolyester with pseudo-eutectic composition can be tailored by varying both nonisothermal or isothermal crystallization conditions, a remarkable and novel behavior, as far as the authors are aware.

**Figure 11.** Wide angle X-ray diffraction (WAXS) diffraction patterns of BS45CL55 registered during isothermal crystallization at −12 ◦C (**a**), −9 ◦C (**b**), and −6 ◦C (**c**).

#### **4. Conclusions**

The complex isothermal nucleation, growth and overall crystallization of isodimorphic PBS-*ran*-PCL copolyesters were studied for the first time. The equilibrium melting temperatures show a very clear pseudo-eutectic point at a composition of BS45CL55. To the left of the pseudo-eutectic point (in a plot of *Tm<sup>0</sup>* versus CL-unit molar content) only the PBS-rich phase is able to crystallize, at the pseudo-eutectic point both PBS-rich and PCL-rich phases can crystallize and to the right of the pseudo-eutectic point only PCL-rich crystals are formed. With respect to the parent homopolymers, any comonomer incorporation on either side of the pseudo-eutectic point causes as increase in nucleation density and nucleation rate, as well as a decrease in spherulitic growth rate.

As a result, the overall crystallization rate determined by DSC was a strong function of composition and supercooling. For PBS-rich copolymers, the PBS-rich phase overall crystallization rate-determining-step was the spherulitic growth rate. On the other hand, for PCL-rich copolymers, the nucleation rate (which was always larger for mirror compositions) gained more importance in the control of the overall crystallization rate.

The crystallization of the isodimorphic copolyester with pseudo-eutectic composition can be tailored by varying the isothermal crystallization temperature, depending on which, either one or both phases are able to crystallize. Such remarkable property control allows the possibility of having a single copolyester with only PCL crystals, only PBS crystals, or both types of crystals, thus exhibiting very different thermal properties.

**Supplementary Materials:** The following materials are available online at http://www.mdpi.com/2073-4360/12/ 1/17/s1, Table SI-1. Equilibrium melting temperatures for CoP(BSxCLy) compositions and their corresponding homopolymers, Figure SI-1. Nucleation kinetics studies by PLOM. Nuclei density as a function of time at different crystallization temperature for PBS-rich phase samples: (a) BS91CL9, (b) BS66CL34, (c) BS62CL38, (d) BS55CL45, (e) BS51CL49, and (f) BS45CL55. *Tc* employed are chosen so that Δ*T* = 40, 38, 36, 34, 32 ◦C for all samples, Figure SI-2. Nuclei density during isothermal crystallization as a functional Δ*T* for PBS-rich (a) and for PCL-rich (b) copolyesters, Figure SI-3. Hoffman–Weeks plots for PBS-*ran*-PCL compositions. The black solid line represents the thermodynamic equilibrium line *Tm* = *Tc*, Figure SI-4. Plot of *log I* versus 1/*T*(Δ*T*) <sup>2</sup> and fitting to Turnbull–Fisher equation (Equation (1)) for PBS-rich (a) and PCL-rich (b) compositions, Figure SI-5. Spherulitic growth rates *G* determined by PLOM for neat PBS and PBS-rich (a) and for neat PCL and PCL-rich (b) copolymers as a function of supercooling, Figure SI-6. The fits to the Lauritzen–Hoffman equation using the free Origin plug-in developed by Lorenzo et al. and the experimental data for the (a-a') PBS, (b-b´) BS78CL22, and (c-c´) BS45CL55, Figure SI-7. The σ*<sup>e</sup>* value versus CL-unit molar fraction that obtained for PLOM experiments (σ*<sup>G</sup> <sup>g</sup>* ) and DSC experiments (στ *<sup>g</sup>* ).

**Author Contributions:** The general conceptualization of the work described here was performed by A.J.M. S.M.-G. and A.M.d.I. designed the synthetic route of the copolymers and their molecular characterization. The experiments were performed by M.S. The general supervision of experimental measurements at the UPV/EHU labs was done by A.M. and at UPC by A.M.d.I. M.Z. and A.J.M. supervised calculations and fittings of experimental data. The paper was written by M.S. and A.J.M. and revised by all co-authors. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by (a) MINECO through project MAT2017-83014-C2-1-P, (b) ALBA synchrotron facility through granted proposal 2018082953 (c) the Basque Government through grant IT1309-19 and (d) European Union´s Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement No 778092

**Acknowledgments:** MS gratefully acknowledges the award of a PhD fellowship by POLYMAT Basque Center for Macromolecular Design and Engineering.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

## **Assessment of the Mechanical and Thermal Properties of Injection-Molded Poly(3-hydroxybutyrate-***co***-3 hydroxyhexanoate)**/**Hydroxyapatite Nanoparticles Parts for Use in Bone Tissue Engineering**

#### **Juan Ivorra-Martinez 1, Luis Quiles-Carrillo 1, Teodomiro Boronat 1, Sergio Torres-Giner 2,\* and José A. Covas 3,\***


Received: 18 May 2020; Accepted: 17 June 2020; Published: 21 June 2020

**Abstract:** In the present study, poly(3-hydroxybutyrate-*co*-3-hydroxyhexanoate) [P(3HB-*co*-3HHx)] was reinforced with hydroxyapatite nanoparticles (nHA) to produce novel nanocomposites for potential uses in bone reconstruction. Contents of nHA in the 2.5–20 wt % range were incorporated into P(3HB-*co*-3HHx) by melt compounding and the resulting pellets were shaped into parts by injection molding. The addition of nHA improved the mechanical strength and the thermomechanical resistance of the microbial copolyester parts. In particular, the addition of 20 wt % of nHA increased the tensile (Et) and flexural (Ef) moduli by approximately 64% and 61%, respectively. At the highest contents, however, the nanoparticles tended to agglomerate, and the ductility, toughness, and thermal stability of the parts also declined. The P(3HB-*co*-3HHx) parts filled with nHA contents of up to 10 wt % matched more closely the mechanical properties of the native bone in terms of strength and ductility when compared with metal alloys and other biopolymers used in bone tissue engineering. This fact, in combination with their biocompatibility, enables the development of nanocomposite parts to be applied as low-stress implantable devices that can promote bone reconstruction and be reabsorbed into the human body.

**Keywords:** P(3HB-*co*-3HHx); nHA; nanocomposites; mechanical properties; bone reconstruction

#### **1. Introduction**

Bone fracture is one of the most common injuries. Bone regeneration encompasses three stages, namely inflammation, bone production, and bone remodeling [1]. During the latter, it is extremely important to expose the bone to the natural load-bearing conditions associated to its function [2]. Currently, titanium alloys such as Ti-6Al-4V are the most used for the manufacture of orthopedic fixing devices and bone implants due to their excellent biocompatibility and high mechanical resistance [3]. However, they prevent the bone from being subjected to the required mechanical loadings [4]. Indeed, while natural bone has a modulus ranging between 8 to 25 GPa, metals have a modulus of 110–210 GPa, which results in the load being imparted onto the device rather than the bone which then causes a localized decrease in bone mineral density [5]. Meanwhile, metal ion leaching increases inflammation

and irritation around the implant [6]. As a result, there is often a need for a second surgery to remove the fixation device, leading to higher medical costs and greatly increased patient discomfort. A current alternative is the use of fixation devices that metabolize in the human body after fulfilling their function [7]. In particular, the use of biopolymers with biocompatibility and reabsorption capacities is very promising [8]. Biocompatibility involves the capability of a given substance to perform with a suitable host response in a particular use. Furthermore, no substance or material can be "biocompatible" if it releases cytotoxic substances. The degradation process of a given biopolymer within the human body consists of two phases. First, the biopolymer chains break, either as a consequence of hydrolysis or due to the action of a body enzyme. Thereafter, the human body assimilates the fragments. For this purpose, either a phagocytosic or metabolic process develops [9]. Surface porosity, shape, and tissue environment, including chemical build-up of the materials, play a significant role in biocompatibility [10,11].

For the past few decades, polymers of the polyhydroxyalkanoates (PHAs) family have been paving the way for the development of new biomedical products. These microbial biopolyesters degrade when exposed to marine sediment, soil or compost. A vast number of microorganisms secrete extracellular PHA-hydrolyzing enzymes, so-called PHA depolymerases, to degrade PHA into their oligomers and monomers, which subsequently act as nutrients inside the cells [12]. Their potential as alternatives for the manufacture of a wide range of medical devices, such as absorbable sutures, surgical pins or staples, is well recognized on account of their biodegradable nature as well as disintegration by surface erosion [13]. Broadly, the biocompatibility of PHA materials can be differentiated into two categories, immunocompatibility and nonallergic response. The former involves the extent of antigenic resemblance between the tissues of various individuals that determines the acceptance or rejection of allografts. PHAs are essentially immunocompatible for use in medical applications, that is, their materials should not elicit harsh immune responses upon introduction into the soft tissues or blood of a host organism [14]. Indeed, 3-hydroxybutyrate (3HB), the main monomeric constituent of most PHAs, is a result of cellular metabolism that is formed by oxidation of fatty acid within the liver cells and it is a usual component of human blood [15]. Other previous studies have also revealed that PHA did not elicit an allergic response or any hypersensitive immune reaction [10,16].

Depending on the number of carbon atoms in the monomers, PHAs can be classified as short-chain-length PHAs (*scl*-PHAs; 3−5 C-atoms) and medium-chain-length PHAs (*mcl*-PHAs; 6−14 C-atoms). Generally, *scl*-PHAs are rigid and brittle, while *mcl*-PHAs have higher flexibility and toughness [17]. Poly(3-hydroxybutyrate) (PHB) is the simplest and most common member of the PHA family. However, the high brittleness of PHB and other *scl*-PHAs, such as poly(3-hydroxybutyrate-*co*-3-hydroxyvalerate) (PHBV) with less than 15 mol % fraction of 3-hydroxyvalerate (3HV), restricts their application in bone fixing devices [18]. In this regard, poly(3-hydroxybutyrate-*co*-3hydroxyhexanoate) [P(3HB-*co*-3HHx)], also referred as PHBH, represents a recent addition to the group of PHAs for biomedical applications. The introduction of the *mcl* 3-hydroxyhexanoate (3HHx) co-monomer into the polymer backbone of PHB significantly increases the flexibility and reduces stiffness [19]. Therefore, the macroscopic properties of P(3HB-*co*-3HHx) vary with the proportion of each monomer in the copolyester [20], in which the higher the 3HHx content, the higher the ductility [21]. Apart from the changes in the mechanical properties, the most remarkable transformation that P(3HB-*co*-3HHx) brings along is its ability to undergo enzymatic degradation by lipase [22], which is not seen in either PHB or PHBV. Prior experiments have shown that materials based on P(3HB-*co*-3HHx) and other *mcl*-PHAs have good biocompatibility for chondrocytes [23], nerve cells [24] as well as osteoblast and fibroblast cells [25,26]. This property should make P(3HB-*co*-3HHx) a suitable choice for several tissue engineering applications since it adds a further variable that can be used to tailor its degradation [27].

While PHAs are biocompatible substrates for cell propagation and are potentially an effective template for the repair of osseous and chondral defects, there is still a need to improve the mechanical strength, thermal resistance, and biological response of these biomaterials in order to make them more suitable for bone tissue engineering. Osteoconductive fillers can be introduced into polymer matrices with the aim of improving the mechanical properties and also accelerating the bone repair process by favoring the growth of bone cells inside the pores [28,29]. For example, calcium orthophosphates (CaPO4) have bioactive properties that increase bone cell proliferation, the so-called osteoinduction [30]. As a rule, both the mechanical resistance and bioactivity of composites prepared with collagen, chitin and/or gelatin, increase with increasing CaPO4 content [31]. Hydroxyapatite, Ca5(PO4)3OHCa5(PO4)3OH, which is the principal crystalline constituent of bone, shows a high degree of biocompatibility and good osteoconductive and osteoinductive properties. Therefore, hydroxyapatite nanoparticle or nanohydroxyapatite (nHA) is the most widely used "bioceramic" for the manufacture of medical devices and dental implants [32]. This fact is exemplified by the production of prostheses for cranial reconstruction using poly(methyl methacrylate) (PMMA)/nHA composites [33]. Indeed, nHA exists in the human bone in the form of nanometer-sized threads, thus ensuring biocompatibility. At present, it is mostly used to produce surface coatings, as its biomimetic mineralization enables the production of biomaterials with biomimetic compositions and hierarchical micro/nanostructures that closely mimic the extracellular matrix of native bone tissue [34,35].

Due to the well-known high bioactive properties in terms of bone regeneration of PHA- and nHA-based composites, this study aims to determine the physical properties of injection-molded parts made of P(3HB-*co*-3HHx)/nHA composites, for potential use as bone resorbable devices. To this end, different contents of nHA were incorporated into P(3HB-*co*-3HHx) and the mechanical, thermal, and thermomechanical properties were analyzed and compared to some metal alloy-based solutions currently available in the biomedical field. As a first, the parts showed sufficient dimensional and thermal stability for bone tissue engineering and their elasticity was nearer to that of the natural bone when compared to the metal alloys used for bone implants.

#### **2. Materials and Methods**

#### *2.1. Materials*

P(3HB-*co*-3HHx) copolymer was supplied by Ercros S.A. (Barcelona, Spain) as ErcrosBio PH110. The ratio of 3HHx in the copolyester is ~10 mol % and its number average molecular weight (Mn) is 1.22 <sup>×</sup> 105 g/mol. It shows a melt flow index (MFI) of 1 g/10 min (2.16kg/160 ◦C) according to the ISO 1133-2 standard and a true density of 1.20 g/cm<sup>3</sup> following the UNE EN ISO 1183-1 standard. Hydroxyapatite synthetic nanopowder was procured from Sigma-Aldrich S.A. (Madrid, Spain) with commercial reference 677418. According to the manufacturer, it presents the following properties: particle size <sup>&</sup>lt; 200 nm, surface area <sup>&</sup>gt; 9.4 m2/g by Brunauer-Emmett-Teller (BET) analysis, purity <sup>≥</sup> 97%, and molecular weight (MW) of 502.31 g/mol.

#### *2.2. Preparation and Processing of P(3HB-co-3HHx)*/*nHA Parts*

Both P(3HB-*co*-3HHx) pellets and nHA powder were dried separately for at least 6 h at 80 ◦C in a dehumidifying oven from Industrial Marsé S.A. (Barcelona, Spain). The materials were then pre-mixed manually in closed zip-bags at the ratios presented in Table 1.

**Table 1.** Code and composition of the samples prepared according to the weight content (wt %) of poly(3-hydroxybutyrate-*co*-3-hydroxyhexanoate) [P(3HB-*co*-3HHx)] and hydroxyapatite nanoparticles (nHA).


The different P(3HB-*co*-3HHx) and nHA mixtures weighing 800 g were melt-compounded using a co-rotating twin-screw extruder from Dupra S.L. (Castalla, Spain). It features two screws with a diameter (D) of 25 mm and a length-to-diameter ratio (L/D) of 24, while the modular barrel is equipped with 4 individual heating zones coupled to a strand die. Further details of the extruder can be found elsewhere [36]. Extrusion was performed with a screw speed of 20–25 rpm to prevent material degradation due to shear-induced viscous dissipation, a feed of 1.2 kg/h, and a barrel set temperature profile of 110–120–130–140 ◦C from hopper to die. The extruded filaments were cooled down in an air stream and pelletized using an air-knife unit.

Test parts for characterization were obtained by injection molding. The equipment (Meteor 270/75, Mateu & Solé, Barcelona, Spain) was operated with a barrel set temperature profile of 115–120–125–130 ◦C from hopper to nozzle, with the mold kept at 60 ◦C. An injection time of 1 s was used to avoid material degradation by shear-induced viscous dissipation. The clamping force was 75 tons and the cooling time was set at 60 s. Parts with a thickness of approximately 4 mm were obtained for characterization. Since P(3HB-*co*-3HHx) develops secondary crystallization with time, the parts were allowed to age for 14 days at room temperature prior to characterization.

#### *2.3. Mechanical Tests*

Uniaxial tensile tests were performed according to the ISO 527-2: 2012 standard using a universal testing machine ELIB-50 (Ibertest S.A., Madrid, Spain) fitted with a load cell of 5 kN and using a 3542-050M-050-ST extensometer from Epsilon Technology Corporation (Jackson, WY, USA). Flexural properties were determined following the ISO 178: 2011 standard using the same equipment. Both tests were carried out at 5 mm/min using 150 mm × 10 mm × 4 mm parts. Charpy impact tests were performed following the ISO 179-1: 2010 standard. Samples with a V-shaped notch with a radius of 0.25 mm and dimensions 80 mm × 10 mm × 4 mm were subjected to the impact of a 1-J pendulum impact tester from Metrotec S.A. (San Sebastián, Spain). Shore hardness was measured with a 673-D durometer (J. Bot Instruments, Barcelona, Spain), following the ISO 868: 2003 standard. At least six parts were tested for each mechanical test.

#### *2.4. Thermal Tests*

Samples weighing 5–10 mg were analyzed by differential scanning calorimetry (DSC) in a Q200 from TA Instruments (New Castle, DE, USA) to study the thermal transitions. The samples were subjected to a three-stage thermal cycle in which the samples were first heated from −50 to 200 ◦C and cooled down to −50 ◦C in order to eliminate the thermal history and then reheated to 200 ◦C. All the heating and cooling scans were performed at 10 ◦C/min. Testing was performed under inert atmosphere using a nitrogen flow of 50 mL/min. The degree of crystallinity (*XC\_max*) was calculated using Equation (1) [37]:

$$Xc\_{\rm max} = \left[\frac{\Delta H\_{\rm m}}{\Delta H\_{\rm m} \cdot (1 - w)}\right] \cdot 100\% \tag{1}$$

where Δ*H*<sup>m</sup> (J/g) corresponds to the melting enthalpy of P(3HB-*co*-3HHx), Δ*H*m<sup>0</sup> (J/g) is the theoretical value of a fully crystalline of P(3HB-*co*-3HHx), taken as 146 J/g [38], an 1 − *w* indicates the weight fraction of P(3HB-*co*-3HHx) in the sample.

Thermogravimetric analysis (TGA) was performed to determine the thermal stability of the injection-molded parts. Samples weighing 10–20 mg were heated from 30 to 700 ◦C at a heating rate of 20 ◦C/min in a TGA 100 from Linseis Messgeräte GmbH (Selb, Germany) under nitrogen atmosphere with a flow rate of 25 mL/min. All thermal tests were carried out in triplicate.

#### *2.5. Thermomechanical Tests*

Injection-molded parts sizing 10 mm × 5 mm × 4 mm were subjected to a temperature sweep from −70 to 100 ◦C at a heating rate of 2 ◦C/min using a DMA-1 from Mettler-Toledo S.A. (Barcelona, Spain). Dynamic thermomechanical analysis (DMTA) was carried out in bending mode with a maximum bending strain of 10 μm at a frequency of 1 Hz and a force of 0.02 N.

The dimensional stability of the parts was studied by thermomechanical analysis (TMA) in a Q400 thermomechanical analyzer from TA Instruments (New Castle, DE, USA). The applied force was set to 0.02 N and the temperature program was scheduled from −70 to 70 ◦C in air atmosphere (50 mL/min) at a constant heating rate of 2 ◦C/min. All thermomechanical tests were performed in triplicate.

#### *2.6. Microscopy*

The fracture surfaces of the injection-molded parts after the Charpy impact tests were analyzed by field-emission scanning electron microscopy (FESEM) (Oxford Instruments, Abingdon, UK) with an electron acceleration voltage of 2 kV. A gold-palladium coating was applied through sputtering (SC7620, from Quorum Technologies Ltd, East Sussex, UK). Additionally, to visualize the dispersion of nHA in the P(3HB-*co*-3HHx) matrix, the fracture surfaces were attacked with 6M hydrochloric acid (HCl) (37% purity, Panreac AppliChem, Barcelona, Spain) for 12 h to selectively remove nHA prior to observation [39].

#### *2.7. Statistical Analysis*

Statistical evaluation of the mechanical, thermal, and thermomechanical properties of P(3HB-*co*-3HHx)/nHA parts was carried out with the open source R software (http://www.r-project.org) with a Shapiro–Wilk test regarding a normal distribution for n < 1000. Tukey tests were performed to determine significant differences between the data on normally distributed data. In order to establish the non-parametric relationship between mechanical properties and nHA content in the parts, the Spearman's correlation test was followed. The number of tested samples for each test is included in Table 2 and the level of significance was established as *p* < 0.05 in all cases.


**Table 2.** Number of tested samples (*n*) for each injection-molded poly(3-hydroxybutyrate-*co*-3 hydroxyhexanoate) [P(3HB-*co*-3HHx)]/hydroxyapatite nanoparticles (nHA) parts and the type of statistical test performed for each testing method with level of significance (*p*).

DSC = differential scanning calorimetry; TGA = thermogravimetric analysis; DMTA = dynamic thermomechanical analysis; TMA = thermomechanical analysis.

#### **3. Results and Discussion**

#### *3.1. Mechanical Characterization of the P(3HB-co-3HHx)*/*nHA Parts*

The data collected for the mechanical properties from the tensile, flexural, hardness, and impact Charpy tests of the neat P(3HB-*co*-3HHx) and P(3HB-*co*-3HHx)/nHA composite parts produced with the different compositions is summarized in Table 3. Figures 1 and 2 display the effect of nHA incorporation on the tensile and flexural properties, respectively, whereas Table 4 shows the correlation coefficient (*rs*) and *p* for each mechanical property according to the Spearman's test.



\* Indicates a significant difference compared with the previous sample (*p* < 0.05). Level of significance (*p*) values are included in Table S1.

**Figure 1.** Evolution of the maximum tensile stress (σmax), tensile modulus (Et), elongation at break (εb) in the injection-molded parts of poly(3-hydroxybutyrate-*co*-3-hydroxyhexanoate) [P(3HB-*co*-3HHx)] with the content of hydroxyapatite nanoparticles (nHAs). \* Indicates a significant difference compared with the previous sample (*p* < 0.05)

**Figure 2.** Evolution of the maximum flexural stress (σf) and flexural modulus (Ef) in the injection-molded parts of poly(3-hydroxybutyrate-*co*-3-hydroxyhexanoate) [P(3HB-*co*-3HHx)] with the content of hydroxyapatite nanoparticles (nHAs). \* Indicates a significant difference compared with the previous sample (*p* < 0.05).


**Table 4.** Spearman's test correlation coefficient (*rs*) and level of significance (*p*) for each mechanical property.

The tensile properties of the injection-molded P(3HB-*co*-3HHx) parts were relatively similar to those reported by Giubilini et al. [40], although the here-prepared materials were slightly less mechanically resistant and more ductile. These differences could be related to the 3HHx monomer content in the copolyester as well as to differences in processing. One can observe in both Table 3 and Figure 1 that the values of σmax and ε<sup>b</sup> decreased, while those of Et increased with increasing nHA concentration in the nanocomposite parts. The Spearman's test confirmed the existence of a trend between the tensile properties of the nanocomposites and the nHA content, showing a negative *rs* trend (inversely proportional correlation) for σmax and ε<sup>b</sup> and a positive trend (directly proportional correlation) for Et, while in all cases *p* < 0.05. In particular, the addition of 20 wt % of nHA produced a slight decrease of σmax from 17.7 to 14.4 MPa, but an increase of nearly 64% in Et (from approximately 1 to 1.7 GPa) accompanied with a significant loss of ductility (ε<sup>b</sup> was reduced from 19.4 to 6.5%). The reduction in stress was probably caused by the poor interface adhesion between biopolymer and nanofiller. Higher interfacial adhesion can probably be promoted through the pretreatment of nHA with silanes [41], but it could negatively affect the biocompatibility of the parts. The increase in Et was anticipated, since nHA forms highly rigid structures. Furthermore, as it will be discussed during the thermal characterization, the addition of nHA could promote higher degrees of crystallinity and, hence, higher stiffness. Although similar results have been reported earlier [42,43], the here-prepared parts showed higher ductility due to the use of a more flexible PHA. The decrease observed in stiffness with increasing nHA content can be attributed to insufficient wetting and impregnation of the nanoparticles by the polymer matrix, mainly due to particle agglomeration during manufacture or processing of the materials [44]. However, melt-mixing methodologies using co-rotating twin-screw extruders, as adopted here, can generally yield well-dispersed nanocomposites [45]. Ductility loss was expected since the presence of nHA can prompt polymer crystallinity, hindering chain mobility due to adsorption of biopolymer chains on the surface of the nanoparticles [46,47].

In Figure 2, it can be seen that the addition of nHA to P(3HB-*co*-3HHx) increased both σ<sup>f</sup> and Ef, particularly the latter. The former increased up to a content of 5 wt % of nHA and then became insensitive to higher nanoparticle contents, since the values showed no significant differences. Indeed, the Spearman's test showed a positive correlation (*rs* > 0) for both Ef and σf, however, for the latter, the statistical hypothesis should be rejected as p was higher than 0.05. Contrarily, the addition of 20 wt % of nHA caused an increase of approximately 60% of Ef, as similarly observed above for Et. The resultant increase in mechanical strength can be related to the intrinsic high values of compressive strength and modulus of nHA, which are in the ranges of 500–1000 MPa and 80–110 GPa, respectively [48,49].

In comparison with the mechanical values of other degradable and non-degradable materials, the P(3HB-*co*-3HHx)/nHA parts produced in this study showed intermediate values to most biodegradable polymers and metal alloys. For instance, the Et values of poly(ε-caprolactone) (PCL) and PLA materials range between 400–600 MPa [50] and 2−3 GPa [51,52], respectively, while other biodegradable copolyesters such as poly(butylene adipate-*co*-terephthalate) (PBAT) show significantly lower values [53]. However, PLA is a brittle polymer, which can limit its application in bone fixation devices, or any other biomedical device that would be subjected to local flexural stress or impacts. The values attained are relatively similar to those of poly(lactic-*co*-glycolic acid) (PLGA), that is,

1.4−2.8 GPa [54]. Indeed, PLGA is widely used in biomedical and pharmaceutical applications, but it shows longer degradation times, which can extend up to 12 months [55]. Regarding metal alloys, the Et values of the most widely used stainless steels for implant fixing devices and screws, that is, SUS316L stainless steel and cobalt-chrome (Co-Cr) alloys, are around 180 GPa and 210 GPa, respectively [56]. Lower values have been reported for titanium (Ti) and its light alloys, such as Ti-6Al-4V ELI, which are also widely used for making implant devices, having a value of around 110 GPa [57]. As shown above, in comparison to metal alloys, the elasticity of the P(3HB-*co*-3HHx)/nHA composites prepared in this study is nearer to that of the natural bone, which is in the 8–25 GPa range [5]. Thus, from a mechanical point of view, their use in bone scaffolds and resorbable plates or screws looks promising.

As expected, hardness increased with the presence of nHA that, due to its ceramic nature, is highly rigid. The increase was significant at nHA contents higher than 2.5 wt % and this effect was statistically corroborated by Spearman's test, showing a positive trend with an *rs* value of ~0.98. In addition, molecular mobility could be reduced due to the presence of the nanoparticles [58]. In particular, the incorporation of 20 wt % of nHA yielded an increase of 8% in hardness. A similar increase in Shore D hardness was reported by Ferri et al. [39] for PLA after the incorporation of nHA. In particular, it increased from 73.9, for neat PLA, up to 78.4, for the PLA composite containing 30 wt % of nHA. As also anticipated, the impact strength of the nanocomposites diminished significantly with increasing nHA content with significant differences between the samples, which was confirmed by the negative correlation obtained by the Spearman's test (*rs* −0.84). For instance, the nanocomposite parts containing 20 wt % of nHA revealed an impact strength approximately three times lower than that of the neat P(3HB-*co*-3HHx) part, that is, it reduced from 5.1 to 1.7 kJ/m2. Lower values of impact strength were reported for V-notched injection-molded pieces of PLA, that is, 2.1 kJ/m<sup>2</sup> [51]. In addition, significantly higher values have been described for Ti-6Al-4V, with a Rockwell hardness C (HRC) of 38 and approximately 112 kJ/m2 impact strength [59]. In the case of natural bone, toughness varies widely with age and type. For instance, the impact strength of the femora ranges from 4 to 70 kJ/m2 [60]. Therefore, the various mechanicals tests revealed a clear tendency towards a decrease in ductility and an increase in stiffness of the injection-molded parts with increasing nHA content, which are closer to those of the natural bone.

In summary, the here-developed P(3HB-*co*-3HHx)/nHA parts showed an improvement of the stiffness determined in terms of Et and Ef, in which a positive trend was observed in both cases (*rs* > 0). The ductile properties, that is, ε<sup>b</sup> and impact strength, showed negative trends (*rs* < 0), which was ascribed to a chain mobility reduction that also contributed to a hardness increase of the nanocomposite, showing a positive trend in the Spearman's test.

#### *3.2. Thermal Characterization of the P(3HB-co-3HHx)*/*nHA Parts*

Figure 3 displays the DSC curves for the neat P(3HB-*co*-3HHx) part and the P(3HB-*co*-3HHx)/nHA composite parts with different nanoparticle contents. Table 5 presents the thermal properties obtained from the second heating scan, after erasing the thermal history of the sample. At approximately 0 ◦C, one could observe a step change in the base lines, which corresponded to the glass transition temperature (Tg) of P(3HB-*co*-3HHx). This second-order thermal transition was located at −0.3 ◦C for the neat biopolymer and it was significantly unaffected by the presence of nHA. The exothermic peaks located between 40 and 70 ◦C corresponded to the cold crystallization temperature (Tcc) of P(3HB-*co*-3HHx). In the case of the neat biopolymer part, this peak was located at 49.8 ◦C. It could be observed that the values of Tcc increased with increasing nHA content until 10 wt %, and then slightly decreased at the highest content tested, that is, 20 wt %. These results suggested that low nHA contents impaired the movement of P(3HB-*co*-3HHx) chains and, hence, hindered the crystallization process. A similar thermal behavior during the analysis of the second heating curves was recently observed by Senatov et al. [61], who associated the presence of nHA to a decrease in the molecular chain mobility of the biopolymer that impeded the crystallization process. Finally, the crystalline P(3HB-*co*-3HHx) domains melted in the thermal range from 100 to 150 ◦C in two peaks. Furthermore, the occurrence of a broad melting region suggested the presence of heterogeneous crystallites with different degrees of perfection, commonly produced in PHAs with relatively high comonomer contents [62]. The thermogram of neat P(3HB-*co*-3HHx) revealed two melting temperatures (Tm1 and Tm2) at approximately 113 and 140 ◦C. Similar thermal properties were reported by Zhou et al. [63] for P(3HB-*co*-3HHx) with 11 mol % content of 3HHx, who also observed a double-melting peak phenomenon in the DSC heating curves of this copolyester. The presence of two melting peaks have been previously ascribed to the melting–recrystallization–melting process of P(3HB-*co*-3HHx) [64]. During this process, imperfect crystals melt at lower temperatures and the amorphous regions order into packed spherulites with thicker lamellar thicknesses that, thereafter, melt at higher temperatures. Alternatively, the melting peaks attained at low temperatures, that is, 110–115 ◦C, could also relate to the crystalline phase of the 3HHx-rich fractions. Lastly, one could observe that the melting profile of P(3HB-*co*-3HHx) was nearly unaffected by the nHA presence, indicating that the nanoparticles did not significantly influence the crystallization process.

**Figure 3.** Differential scanning calorimetry (DSC) thermograms taken during second heating of the injection-molded poly(3-hydroxybutyrate-*co*-3-hydroxyhexanoate) [P(3HB-*co*-3HHx)]/hydroxyapatite nanoparticle (nHA) parts.

**Table 5.** Thermal properties of the injection-molded poly(3-hydroxybutyrate-*co*-3-hydroxyhexanoate) [P(3HB-*co*-3HHx)]/hydroxyapatite nanoparticle (nHA) parts in terms of glass transition temperature (Tg), cold crystallization temperature (Tcc), melting temperatures (Tm1 and Tm2), cold crystallization enthalpy (ΔHcc), melting enthalpy (ΔHm), and maximum degree of crystallinity (χc max).


\* Indicates a significant difference compared with the previous sample (*p* < 0.05). Level of significance (*p*) values are included in Table S2.

In addition to the characteristic values of Tg, Tcc, and Tm, the enthalpies corresponding to the cold crystallization (ΔHcc) and melting (ΔHm) enthalpies were collected from the DSC curves. The latter parameter was used to determine the maximum degree of crystallinity, that is, *XC\_max*, which gives more information about the effect of the additives on the biopolymer, since it does not consider the crystals formed during cold crystallization. It can be seen that P(3HB-*co*-3HHx) showed a maximum degree of crystallinity of 21.4%. One can also observe that crystallinity varied significantly with nHA content. In particular, as nHA was gradually incorporated in higher percentages, the crystallinity increased steadily up to a maximum of nearly 29% at 5 wt % of nHA and then it slightly decreased to values close to 25% for nHA contents of 10 and 20 wt %. This result, in combination with the slightly higher Tcc and Tm values, suggests that the nanoparticles hindered the formation of crystals at low temperatures, but the crystals formed were slightly more perfect and more mass crystallized. This is in agreement with previous studies that concluded that the introduction of nHA into biopolyesters has an effect on the ordering of their molecular chains by acting as a nucleating agent [61,65].

Figure 4 presents the thermogravimetric data for all the materials, while Table 6 gathers the main thermal stability parameters obtained from the TGA curves. Thermal degradation of P(3HB-*co*-3HHx) was observed to occur through a one-step process, which is in agreement with the values reported by Li et al. [20], who showed that the thermal stability of the microbial copolyester was as high as 225 ◦C with almost no mass loss. The temperature at 5% mass loss (T5%) showed no significant differences with nHA contents of up to 5 wt %, but a significant decrease was observed for higher loadings. The temperature at which the maximum mass loss rate occurred (Tdeg) increased from 296.7 ◦C, for the neat P(3HB-*co*-3HHx) part, to 300.9 ◦C, for the part of P(3HB-*co*-3HHx) filled with 2.5 %wt of nHA. This increase in thermal stability has been previously ascribed to the formation of strong hydrogen interactions and Van der Walls forces between the inorganic nanoparticles and the biopolymer chains during the melt-mixing process [66]. The values of Tdeg remained nearly constant, showing no significant differences for nHA contents from 2.5 to 10 wt %, but it significantly decreased to 295.6 ◦C in the part filled with 20 wt % of nHA. The onset of degradation was also reduced for the most filled sample, showing a T5% value of 254.8 ◦C, which represents a reduction of approximately 18 ◦C in comparison to the unfilled P(3HB-*co*-3HHx) sample and its nanocomposites at low contents. These results further indicate that the nanoparticles formed aggregates at high contents, which created volumetric gradients of concentration [66]. In this regard, Bikiaris et al. [65] suggested that when high amounts of nanosized filler aggregates are formed, the structure shifts from nanocomposite to microcomposite and, thus, the shielding effect of the nanosized particles is lessened. In addition, Chen et al. [67] reported that high loadings of nHA in PHBV lower the onset degradation temperature since they can catalyze thermal decomposition. In any case, low nHA loadings (<10 wt %) slightly improved the thermal stability of P(3HB-*co*-3HHx) parts and their thermal stability is considered to be high enough for bone tissue engineering and biomedical applications, which can require thermal sterilization methods such as dry heat sterilization (160 ◦C for 2 h) and steam sterilization (121 ◦C for 20–60 min) [68]. However, the relatively low Tm of P(3HB-*co*-3HHx) would limit the use of these techniques for sterilization and the resultant implantable biomedical devices should be sterilized at low temperatures using ethylene oxide (EO) gas, gamma radiation or ozone. Finally, it can be observed that the residual mass at 700 ◦C increased gradually with the nHA content due to the high thermal stability of the mineral nanoparticles.

**Table 6.** Main thermal degradation parameters of the injection-molded poly(3-hydroxybutyrate*co*-3-hydroxyhexanoate) [P(3HB-*co*-3HHx)]/hydroxyapatite nanoparticle (nHA) parts in terms of onset temperature of degradation (T5%), degradation temperature (Tdeg), and residual mass at 700 ◦C.


\* Indicates a significant difference compared with the previous sample (*p* < 0.05). Level of significance (*p*) values are included in Table S3.

**Figure 4.** (**a**) Thermogravimetric analysis (TGA) and (**b**) first derivate thermogravimetric (DTG) curves of the injection-molded poly(3-hydroxybutyrate-*co*-3-hydroxyhexanoate) [P(3HB-*co*-3HHx)]/ hydroxyapatite nanoparticle (nHA) parts.

#### *3.3. Thermomechanical Characterization of the P(3HB-co-3HHx)*/*nHA Parts*

DMTA was carried out on the injection-molded composite parts in order to understand the role played by nHA on the viscoelastic behavior of P(3HB-*co*-3HHx)/nHA. Figure 5 illustrates the DMTA curves of the neat P(3HB-*co*-3HHx) part and the P(3HB-*co*-3HHx)/nHA composite parts with different nanoparticle contents. Figure 5a gathers the evolution of the storage moduli (E') in the temperature sweep from −40 to 80 ◦C at a frequency of 1 Hz. The Tg values and the corresponding values of E' at −40, 37, and 70 ◦C are presented in Table 7, since the first and last temperatures are representative of the stored elastic energy of the amorphous phase of P(3HB-*co*-3HHx) in its glassy and rubber states, respectively, whereas the middle one corresponds to the actual temperature of the human body. It can be observed that all the P(3HB-*co*-3HHx)-based parts presented a similar thermomechanical profile. In particular, the samples showed high E' values, that is, high stiffness, at temperatures below 0 ◦C and then E' sharply decreased. This thermomechanical change was produced because the temperature exceeded the alpha (α)-relaxation of the biopolymer, which is related to its Tg. One can also observe that the rate of decrease of E' reduced somewhat when the temperature reached approximately 40 ◦C due to the occurrence of cold crystallization. The values of E' at −40, 37, and 70 ◦C of the neat P(3HB-*co*-3HHx) part were 1909.9, 519.2, and 210.5 MPa, respectively. The E' value attained at 37 ◦C was in accordance with the mechanical data presented in Section 3.1, which indicated that only the P(3HB-*co*-3HHx) parts filled with the highest nHA contents, that is, 15 and 20 wt %, showed significantly higher values. However, the results also indicated that the parts crystallized during the ageing process since the thermomechanical changes during and after cold crystallization were relatively low. As expected, the E' values progressively increased with increasing the nHA content, given the high stiffness of the nanoparticles. It is worth noting that the reinforcing effect was more noticeable at higher temperatures since the amorphous phase of P(3HB-*co*-3HHx) was in the rubber state. Indeed, at higher temperatures, the thermomechanical response of all the P(3HB-*co*-3HHx) composite parts was significantly different, dependent upon the nHA content. For instance, at −40 ◦C the E' value increased from 1935.2 MPa for the nanocomposite part containing 2.5 wt % of nHA, to 2100.4 MPa for the part filled with 20 wt % of nHA, whereas these values increased from 212.3 MPa to 333.1 MPa at 70 ◦C.

The loss tangent or dynamic damping factor (*tan* δ) curves are shown in Figure 5b. Since the position of the *tan* δ peak gives an indication of the biopolymer's Tg, these values were also included in Table 7. In the case of the neat P(3HB-*co*-3HHx) part, the *tan* δ peak was located at 10.7 ◦C, which is similar to that reported by Valentini et al. [69]. It is worth mentioning that, in all cases, the *tan* δ peaks were approximately 10 ◦C higher than the Tg values. Since *tan* δ represents the ratio of the viscous to the elastic response of a viscoelastic material, this indicates that part of the applied load was dissipated by energy dissipation mechanisms such as segmental motions, which are related to Tg, but part of the energy was also stored and released upon removal of the load at higher temperatures. One can observe that the incorporation of nHA shifted slightly the position of the *tan* δ peaks and also reduced their intensity for the highest nHA loadings, that is, 10 and 20 wt %. Decreasing *tan* δ peaks intensity indicated that the nanocomposite parts showed a more elastic response and, hence, presented more potential to store the applied load rather than dissipating it [70]. This reduction is directly related to the higher E' values attained due to nanoparticle reinforcement and it confirmed that nHA imposed restrictions on the molecular motion of the P(3HB-*co*-3HHx) chains, resulting in a material with more elastic behavior [71]. It also correlated well with the DSC results shown above, indicating that P(3HB-*co*-3HHx) developed more crystallinity in the nanocomposite parts due to the nucleating effect of nHA and, thus, the less amorphous phase underwent glass transition.

**Figure 5.** Evolution as a function of temperature of the (**a**) storage modulus and (**b**) dynamic damping factor (*tan* δ) of the injection-molded hydroxybutyrate-*co*-3-hydroxyhexanoate) [P(3HB-*co*-3HHx)]/ hydroxyapatite nanoparticles (nHA) parts.



\* Indicates a significant difference compared with the previous sample (*p* < 0.05). Level of significance (*p*) values are included in Table S4.

The effect of temperature on the dimensional stability of the P(3HB-*co*-3HHx)/nHA parts was also determined by TMA. The coefficient of linear thermal expansion (CLTE), both below and above Tg, was obtained from the change in dimensions versus temperature and it is also included in Table 7 along with the Tg values. In all cases, lower CLTE values were attained in the parts below Tg, due to the lower mobility of the P(3HB-*co*-3HHx) chains of the amorphous regions in the glassy state. As anticipated, both below and above Tg, the CLTE values decreased significantly with increasing nHA content due to the increasing replacement of the soft biopolymer matrix by a ceramic material with a considerably lower CLTE value, that is, 13.6 μm/m· ◦C [72]. As a result, the CLTE value below Tg was reduced from 64.3 μm/m· ◦C for the neat P(3HB-*co*-3HHx) part, to 56.7 μm/m· ◦C for the nanocomposite part filled with 20 wt % of nHA. Similarly, above Tg, it decreased from 177.2 to 159.1 μm/m· ◦C, respectively. This thermomechanical response was slightly better than that of the PLA/nHA composites, in which the CLTE values below Tg decreased from 73 to 71 μm/m· ◦C after the incorporation of 20 wt % of nHA into PLA μm/m· ◦C [73]. These results point out that the nanocomposite parts prepared herein show excellent dimensional stability against temperature exposition. However, it is also worth mentioning that, as expected, the CLTE of Ti-based materials was significantly lower, having a mean value of 8.7 μm/m· ◦C [72].

#### *3.4. Morphological Characterization of the P(3HB-co-3HHx)*/*nHA Parts*

Figure 6 shows the samples before and after the various processing steps. Combining melt compounding and injection molding represents a cost-competitive melt-processing methodology to produce a large number of parts using nanocomposites. According to this route, the P(3HB-*co*-3HHx) pellets and the nHA powder were pre-mixed and fed together to the co-rotating twin-screw extruder. In this way, pellets of nanocomposites containing different contents of dispersed nHA particles were obtained. They were subsequently injection molded into dumbbell bars. All parts were defect-free and had a bright surface, the nHA content influencing their color; neat P(3HB-*co*-3HHx) parts were yellow pale, typical of microbial PHA, while the presence of nanoparticles induced a whiter color.

**Figure 6.** Processing steps carried out to prepare the poly(3-hydroxybutyrate-*co*-3-hydroxyhexanoate) [P(3HB-*co*-3HHx)]/hydroxyapatite nanoparticle (nHA) parts; from left to right: as-received P(3HB-*co*-3HHx) pellets and nHA powder, compounded pellets of the nanocomposite, injection-molded parts.

Figure 7 shows the FESEM image, taken at 10,000×, of the nHA powder. The nanoparticles show a flake-like morphology based on plates with sizes 60–120 nm and mean cross-sections of approximately 30 nm. This particular morphology of nHA has been reported to occur at pH values below 9, due to the solution environment changes by the OH− ions during synthesis using polyethylene glycol (PEG) as a template [74].

**Figure 7.** Field-emission scanning electron microscopy (FESEM) images of the hydroxyapatite nanoparticles (nHA) powder. Image was taken at 10,000× with scale marker 150 nm.

Micrographs obtained by FESEM of the fracture surfaces of the injection-molded parts of P(3HB-*co*-3HHx) and the various P(3HB-*co*-3HHx)/nHA composites after the Charpy impact tests are gathered in Figure 8. The fracture surface of the neat P(3HB-*co*-3HHx) part, shown in Figure 8a, indicated that the material presented a relatively high toughness, since it yielded a rough surface with the presence of multiple microcracks and some holes. Some microparticles could be seen in the inset FESEM micrograph taken at higher magnification, which could be related to the presence of nucleating agents and/or fillers added by the manufacturer, such as boron nitride (BN). In this regard, Türkez, et al. [75] have recently demonstrated that BN nanoparticles show slight cytotoxicity potential. In particular, contents below 100 mg/L did not lead to lethal response on human primary alveolar epithelial cells (HPAEpiC), suggesting their safe and effective use in both pharmacological and medical applications. Figure 8b–e gather the fracture surfaces of the P(3HB-*co*-3HHx)/nHA composite parts. The morphological characteristics of the fracture surfaces for the nanocomposites filled with low nHA contents, that is, 2.5 and 5 wt %, remained very similar to that of neat P(3HB-*co*-3HHx)/nHA. In all cases, the nanoparticles were relatively well dispersed and distributed within the biopolymer matrix. However, at higher contents, the nanoparticles tended to form some microaggreagates and the resultant fracture surfaces were smoother, indicating that the nanocomposites were more brittle.

Due to the low nHA particle size and the presence of BN and/or additives in the P(3HB-*co*-3HHx) matrix, selective separation was carried out on the fracture surfaces of the nanocomposite parts, in order to better evaluate the dispersion of the nanoparticles. Figure 9 presents the FESEM images of the fracture surfaces subjected to treatment with 6 M HCl for 12 h. The voids and holes formed in the surfaces were related to removed/dissolved nHA and the overall void size and distribution gave an indication of the original particle dispersion. The micrographs revealed that some microholes were produced after the selective attack on the P(3HB-*co*-3HHx) parts filled with 10 and 20 wt % of nHA, which should correspond to nHA aggregates, whereas the nanocomposites containing low nanoparticle loadings showed nano-sized holes well distributed along the biopolymer matrix, which suggested an

efficient dispersion. Agglomeration was particularly noticeable for the nanocomposite part containing 20 wt % of nHA, thus indicating that the presence of aggregates could induce particle debonding during fracture, as a result of the dissimilar mechanical strength and rigidity of the ceramic nanoparticles and biopolymer matrix. Therefore, the present results correlate well with the mechanical and thermal properties described above, in which nHA loadings of up to 10 wt % increased the mechanical and thermal performance of the P(3HB-*co*-3HHx) parts, whereas the highest nHA content impaired the overall properties due to nanoparticle aggregation.

**Figure 8.** Field-emission scanning electron microscopy (FESEM) images of the fracture surfaces of the injection-molded poly(3-hydroxybutyrate-*co*-3-hydroxyhexanoate) [P(3HB-*co*-3HHx)]/hydroxyapatite nanoparticle (nHA) parts of: (**a**) neat P(3HB-*co*-3HHx); (**b**) P(3HB-*co*-3HHx) + 2.5 nHA; (**c**) P(3HB-*co*-3HHx) + 5 nHA; (**d**) P(3HB-*co*-3HHx) + 10 nHA; (**e**) P(3HB-*co*-3HHx) + 20 nHA. Images were taken at 500× and with scale markers of 10 μm. Inset image showing the detail of the microparticles was taken at 2500× with scale marker of 2 μm.

**Figure 9.** Field-emission scanning electron microscopy (FESEM) images of the fracture surfaces of the injection-molded poly(3-hydroxybutyrate-*co*-3-hydroxyhexanoate) [P(3HB-*co*-3HHx)]/hydroxyapatite nanoparticles (nHA) parts after selective attack with 6M hydrochloric acid (HCl) for 12 h: (**a**) P(3HB-*co*-3HHx) + 2.5 nHA; (**b**) P(3HB-*co*-3HHx) + 5 nHA; (**c**) P(3HB-*co*-3HHx) + 10 nHA; (**d**) P(3HB-*co*-3HHx) + 20 nHA. Images were taken at 1000× with scale marker of 5 μm.

#### **4. Conclusions**

One of the most exciting areas of new material development in the biomedical device community is resorbable polymers. As bone scaffolds, biodegradable and biocompatible polymers will maintain their strength until the liquid in contact begins the dissolution process, eventually leading to their complete elimination from the body, thus avoiding a second surgery for their removal. The herein-prepared injection-molded composite parts of P(3HB-*co*-3HHx)/nHA showed a better matching of mechanical and thermomechanical performance than metal alloys to replace natural bone. While natural bone has a modulus ranging from about 8–25 GPa, the herein-prepared injection-molded parts showed Et values from approximately 1 up to 1.7 GPa and ε<sup>b</sup> values ranging from 6.5 to 19.4%. The incorporation of up to 10 wt % of nHA also improved slightly the thermal stability of the P(3HB-*co*-3HHx) parts and their thermal stability was considered to be high enough for bone tissue engineering, taking into account that nonthermal sterilization methods would be required. These balanced properties in terms of strength and ductility offer the biomedical industry a material that can accomplish different applications in bone reconstruction, for which high-stress materials are not needed, such as bone screws and small orthopedic plates or rods. Future works will explore the potential use of the P(3HB-*co*-3HHx)/nHA composites as drug delivery systems.

**SupplementaryMaterials:**The following tables are available online at http://www.mdpi.com/2073-4360/12/6/1389/s1, Table S1: Level of significance (*p*) values for Table 1; Table S2: Level of significance (*p*) values for Table 2; Table S3: Level of significance (*p*) values for Table 3; Table S4: Level of significance (*p*) values for Table 4.

**Author Contributions:** Conceptualization, S.T.-G. and L.Q.-C.; methodology, J.A.C. and J.I.-M.; validation, J.A.C., J.I.-M. and L.Q.-C.; formal analysis, L.Q.-C. and J.I.-M.; investigation, J.A.C., T.B., L.Q.-C. and J.I.-M.; data curation, L.Q.-C.; writing−original draft preparation, L.Q.-C. and T.B.; writing−review and editing, S.T.-G., J.I.-M., J.A.C.; supervision, S.T.-G., J.A.C. and T.B.; project administration, T.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research work was funded by the Spanish Ministry of Science and Innovation (MICI) project numbers RTI2018-097249-B-C21 and MAT2017-84909-C2-2-R and the POLISABIO program with grant number 2019-A02.

**Acknowledgments:** L.Q.-C. wants to thank GVA for his FPI grant (ACIF/2016/182) and the Spanish Ministry of Education, Culture, and Sports (MECD) for his FPU grant (FPU15/03812). S.T.-G. acknowledges MICI for his Juan de la Cierva–Incorporación contract (IJCI-2016-29675). J.I.-M. wants to thank Universitat Politècnica de València for his FPI grant (PAID-2019- SP20190011). Microscopy services of the Universitat Politècnica de València (UPV) are acknowledged for their help in collecting and analyzing the FESEM images. Authors also thank Ercros S.A. for kindly supplying ErcrosBio® PH110.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
