Crystallization of ultrathin poly(3-hydroxybutyrate) films in blends with small amounts of poly(l-lactic acid): correlation between film thickness and molecular weight of poly(l-lactic acid)

The crystallization behavior of poly(3-hydroxybutyrate) (PHB) ultrathin films in blends with small amounts of poly(L-lactic acids) (PLLAs) was investigated using grazing incidence X-ray diffraction (GIXD) and infraredreflection absorption spectroscopy (IRRAS). Ultrathin films of PHB/PLLA with the same blend ratio of 80/20 (w/w) and two different film thicknesses (30 and 13 nm) were prepared using PHB (Mw 1⁄4 650 000 g mol ) and PLLA with Mw's ranging from 300 000 to 710 g mol 1 to explore the effects of molecular weight on the crystallization of PHB in different confined environments. In the 30 nm thick films, the intensity of PHB crystalline Bragg reflections was strikingly affected by the blended PLLA molecules. Middling molecular weight PLLAs (Mw 1⁄4 23 000–13 100 g mol ) significantly inhibited the crystallization of PHB compared with higher molecular weight PLLAs (Mw $ 50 000 g mol ). In the 13 nm thick films, thickness confinement was revealed to play an important role in crystallization inhibition as both higher and middling molecular weight PLLAs inhibited the crystallization of PHB. Furthermore, IRRAS showed that, for the 30 nm thick films, the addition of a small amount of PLLA (Mw $ 13 100 g mol ) only altered the crystalline structure of PHB in the highly ordered state. In contrast, such PLLAs greatly affected the PHB crystals in both intermediate and highly ordered states in the 13 nm thick films. Unlike the behavior of PLLA in bulk PHB, the lower molecular weight PLLAs (Mw # 3600 g mol ) showed limited effects on the crystallinity and crystalline structure of PHB in both thicknesses of thin films. Several factors, such as phase separation caused by free surface and interface effects, entanglement of PLLA chains, and molecular size of PLLA, are very likely to be responsible for the particular crystallization behavior of PHB in the studied blends.

Previously, we found that PHB ultrathin lms (thickness $52 nm) formed two kinds of crystalline structures, that is, a highly ordered structure and a less ordered structure (the latter has been recognized as the intermediate state), which yield infrared (IR) bands at 1723 and 1731 cm À1 , respectively. 45 The band at 1731 cm À1 , due to the less ordered structure, appears during the melt-crystallization process, which is usually hard to detect in bulk PHB. Similarly, the corresponding band shows up in miscible PHB/PLLA (M w of PLLA ¼ 13 100 g mol À1 ) ultrathin lms with similar thicknesses. 46 In addition, the crystallization of PHB is extremely reduced with the addition of $22 wt% PLLA. This result indicates that the addition of another component, even by a relatively small amount, can be quite effective at controlling the crystallization of polymer thin lms. Since the miscibility of PHB and PLLA depends on the molecular weight, varying the molecular weight of a small amount of PLLA in PHB/PLLA blend ultrathin lms should reveal a crystallization process of PHB that may be different from the bulk.
In the present study, we investigated the molecular weight dependence of a small amount of PLLA on the crystallization of ultrathin PHB lms and its correlation with lm thickness, using a wide range of PLLAs with different molecular weights (300 000-710 g mol À1 ). Based on a previous study, 46 the present study deals with only a small amount of PLLA (20 wt%) with the aim of studying the effect of PLLA molecular weight on the crystallization of PHB in ultrathin lms. Ultrathin lms generally have thickness up to 100 nm; however, lms where the thickness is close to the polymer coil size ($10 nm) show more specic connement effects than conventional ultrathin lms. 62 For distinguishing that geometrical connement effect and molecular weight dependence, we prepared blend lms with two different thicknesses: 30 nm and 13 nm. Two surfacesensitive techniques were employed to observe the crystallization behavior of PHB in the ultrathin blend lms: grazing incidence X-ray diffraction (GIXD) and infrared-reection absorption spectroscopy (IRRAS). We found that the inhibition of PHB crystallization in the blends strongly depended on the molecular weight of the PLLA and the lm thickness. Moreover, several important factors that inuenced the ability of a small amount of PLLAs to inhibit the crystallization of PHB are also discussed. Controlling the crystallinity of PHB lms using a small addition of another polymer, as presented in this study, would make a great advance for thin lm technology and will greatly expand the range of PHB applications.

Materials and sample preparation
Bacterially synthesized PHB with a weight-averaged molecular weight (M w ) of 650 000 g mol À1 , purchased from Aldrich Co., Ltd., was dissolved in hot chloroform, precipitated in methanol, and vacuum-dried at 60 C for several days to remove impurities. Eight kinds of PLLAs with different molecular weights were obtained from Polysciences Inc. (M w ¼ 300 000; 100 000; and 50 000 g mol À1 ) and Polymer Source Inc. (M w ¼ 23 000; 13 100; 6900; 3600; and 710 g mol À1 ). Blends of PHB/PLLAs were prepared by dissolving the prescribed amount of PHB and PLLA in hot chloroform.
Ultrathin lms were deposited on at Au and Si (100) substrates for IRRAS and GIXD measurements, respectively, by spin coating for 45 s. The lms were dried under low vacuum at 60 C for 12 h. To advance the crystallization, annealing was carried out by keeping the lms in a vacuum oven at 120 C which is close to the melting point of PHB ultrathin lms 45 for 5 h. The temperature was then slowly decreased to room temperature before the measurement of GIXD. Two types of lm thicknesses, 30 AE 3 nm and 13 AE 2 nm, were formed by changing the concentration of solutions and rotating speeds.

Infrared spectroscopy
All of the IRRAS spectra were measured using a Thermo Nicolet 6700 FTIR spectrometer equipped with an MCT detector at 2 cm À1 resolution over 256 scans to ensure a high signal-tonoise ratio. A reection attachment (Spectra-Tech. FT80 RAS) with an incident angle of 80 was attached to the spectrometer together with a rotatable wire-grid KRS5-substrate polarizer (ST Japan).

X-ray diffraction
GIXD proles were collected using a Rigaku Ultima IV X-ray diffractometer (Rigaku Co., Japan) where CuKa (l ¼ 1.542 A) radiation was generated at 40 kV and 40 mA. The system was equipped with a scintillation detector and cross beam optics for automatic alignment coupled with a thin lm attachment. In order to investigate the entire crystalline area in the lms, the angle of incidence was set to 0.19 (corresponding to 1.14q c ) for the incident X-ray to fully penetrate into the lms. GIXD proles were collected at room temperature with a scanning speed of 0.3 min À1 and interval of 0.1 . Film thicknesses were determined from the X-ray reectivity proles aer analysis with inhouse developed nonlinear curve-tting soware.

Results and discussion
3.1. Crystallinity and crystallization behavior of PHB as studied by GIXD Fig. 1(a) shows out-of-plane GIXD proles collected at room temperature from the 30 nm thick lms of neat PHB and PHB/PLLA blends (80/20), in which PHB was blended with eight molecular weights of PLLAs. An obvious Bragg reection, centered at 2q ¼ 13.6 , originated from the crystallites of PHB in the lms and was indexed to be the (020) reection. [42][43][44][45][46][47] Two other weak reections also appear at around 2q ¼ 16.6 and 20.3 in Fig. 1(a)(1)- (5), which are the (200/110) and (203) reections of crystalline PLLA, respectively. 46 However, several diffraction proles in Fig. 1(a)(6)-(8) did not exhibit any reections from PLLA, indicating that the PLLAs did not crystallize in those lms. The critical molecular weight for crystallizable PLLA may be exists. Moreover, the results displayed in Fig. 1(a) reveal that PHB and several kinds of PLLA (M w $ 13 100 g mol À1 ) crystallized separately in those blends.
The appearance of only the (020) reection of PHB indicates that the b-axis of the PHB crystallites was almost aligned normal to the lm surface (the ac planes are parallel to the lm surface). 42,45,47 This alignment also xed the type of PHB lamellae to be the edge-on lamellae. The fact that the (020) reection of PHB appears in all of the PHB/PLLA blends indicates that the preferred orientation of PHB crystallites in the blends is the same as that in neat PHB. Even though PLLA does not disturb the alignment and lamellae type of PHB in the blends, the intensity of the (020) reection of PHB showed a distinct variation as the molecular weight of PLLA varied. The (020) peak, which is sharp and strong and does not overlap with any other peaks, has oen been used to characterize the degree of PHB crystallinity. Since the blend ratio and illuminated area of the X-ray beam were xed in this study, the intensity variation of the (020) reection would simply indicate a change in the crystalline volume of PHB in the blends. Fig. 1(b) shows a plot of (020) intensity ratio of PHB as a function of the log ratio of PHB to PLLA M w . The (020) intensity ratio is calculated from the integrated intensity of (020) reection in the blends to that of neat PHB. As can be seen from Fig. 1(b), the intensity ratio slightly decreased with decreasing PLLA molecular weight (M w ¼ 300 000-50 000 g mol À1 , hereinaer called the higher molecular weight PLLAs) in the blends. The intensity ratio dramatically decreased in the blends with PLLA M w ¼ 23 000-13 100 g mol À1 (hereinaer called the middling molecular weight PLLAs). Furthermore, the intensity ratio increased again in the blends where M w of PLLA was #6900 g mol À1 (hereinaer called the lower molecular weight PLLAs). These results clearly indicate that the inhibition of PHB crystallization by the addition of a small amount of PLLA strongly depends on the PLLA molecular weight. The drastic decrease of intensity ratio for the middling molecular weight PLLAs blended PHB lms should be noticed. In the case of blends with the higher molecular weight PLLAs, it is quite probable that those PLLAs were also immiscible with PHB, even in the thin lms, and that the origin of the moderate decrease in the crystalline volume of PHB in those lms might be common in bulk PHB/PLLA blends. The lower molecular weight PLLAs also showed a slight effect on the PHB crystallization, demonstrating a sharp contrast to the behavior of those in bulk samples, in which a noticeable change is usually observed. [26][27][28] Several factors would affect such a PHB crystallization behavior in the PHB/PLLA ultrathin blend lms. Before we discuss those possible factors, we should investigate the effect of thickness connement using the thinner (13 nm) PHB/PLLA blend lms. Fig. 2 displays the out-of-plane GIXD proles of the 13 nm thick lms of neat PHB and PHB/PLLA blends with various molecular weights of PLLA. PHB is still able to crystallize in the lm as the (020) reection was clearly observed for the neat PHB, although the intensity of this (020) peak was much lower than that of the 30 nm thick lm; that is, the (020) intensity divided by 13 nm (thickness) (¼26 nm À1 ) was signicantly smaller than that of the 30 nm thick lm (¼68 nm À1 ). It has been widely accepted that the degree of crystallinity of a polymer oen decreases with reducing thickness. 40,[48][49][50] As shown in the GIXD proles of (1)-(5) in Fig. 2, no (020) reection of PHB was observed in the blends with higher and middling molecular weight PLLAs (M w $ 13 100 g mol À1 ), indicating that the  (2)  100 000 (3) 50 000; (4) 23 000; (5) 13 100; (6) 6900; (7) 3600; and (8)  crystallization of PHB ultrathin lms is fully inhibited by the addition of higher and middling molecular weight PLLAs. Thus, reducing the lm thickness seems to enhance the ability of higher and middling molecular weight PLLAs to inhibit PHB crystallization. The (020) reection also appears in the lower molecular weight PLLA blends, as depicted in (7) and (8), which may be similar to the results observed in the corresponding lms depicted in Fig. 1(a). These results indicate that the lower molecular weight PLLAs do not signicantly affect the crystallization of PHB ultrathin lms. However, it is noted that the (020) reection prole of (6) in Fig. 2 was found to be very weak if we compare it to that of (6) in Fig. 1(a). In this case, the effect of thickness connement in the 13 nm thick lm also seems to enhance the ability of PLLA with M w 6900 g mol À1 to inhibit the crystallization of PHB as occurred by the higher and middling molecular weights of PLLAs.
Furthermore, the higher molecular weight PLLAs still crystallized as the (200/110) reection appears in Fig. 2(1)-(5). The very weak (200/110) reections indicate that only a tiny fraction of the PLLA chains were able to crystallize in those blends. Similar to the results in Fig. 1(a)(6)-(8), the lower molecular weight PLLAs in Fig. 2(6)-(8) did not exhibit any PLLA reection peaks, suggesting that the lower molecular weight PLLAs did not crystallize, irrespective of thickness connement. Additionally, thermal annealing could not induce the crystallization of PHB in those blends in the 13 nm thick lms, as shown in Fig. 3, indicating strong inhibition of PHB crystallization by the small amount of PLLA in these blend lms. In addition, the crystallization of PLLAs was promoted to a certain extent by the thermal annealing as shown in Fig. 3(2)- (6) where increase in intensity from crystalline PLLA is noticed.
The crystallization behavior of PHB in the PHB/PLLA ultrathin lms shown in Fig. 1 and 2 demonstrates a striking contrast with the behavior observed in bulk PHB/PLLA blends. In bulk PHB/PLLA blends, it has been claimed that low molecular weight PLLAs are miscible with PHB and easily affect the crystallization of PHB. In contrast, the high molecular weight PLLAs were immiscible with PHB because of the phase separation between PHB and PLLA. [26][27][28] However, in the present study on ultrathin lms, a small amount of the lower molecular weight PLLAs did not show signicant effects on the crystallization of PHB, indicating that those two components are very likely to be immiscible. The higher molecular weight PLLAs also showed phase separation with PHB in the 30 nm thick lms, which was similar to the behavior in the bulk. Reducing the lm thickness could be considered to make the connement effect stronger, such that it can overcome the phase separation between high molecular weight PLLA and PHB. The present results are discussed and analyzed more completely in the following sections. The next several factors would be worth considering to understand the complex behavior of ultrathin PHB/PLLA lm crystallization.
3.1.1. Phase separation caused by the free surface effect. In the 30 nm thick lms, phase separation would occur between PHB and higher molecular weight PLLAs. In those lms, a free surface effect on the phase separation would possibly be more dominant than the effect from the interface between the polymer and the substrate. 36 The presence of a free surface (polymer-air interface) may promote the phase separation of the two polymers in the blends as the mobility of the molecular chain is increased at the free surface. 51 Therefore, PHB and higher molecular weight PLLA would crystallize separately from each other. On the other hand, reducing the thickness to 13 nm is  likely to enhance the polymer-substrate interface effect; the mobility of molecules would be very restricted in the interface region to only the interaction between molecules and the substrate. In other words, PHB and PLLA would be forced to be miscible in ultrathin lms where the interface effect would be dominant.
3.1.2. Entanglement of PLLA molecular chains. PLLA has a critical molecular weight for entanglement at around 16 000 g mol À1 . 52,53 Long PLLA chains, which can be entangled, would effectively trap PHB molecules and disturb the crystallization of PHB chains. The PLLA chains with M w $ 13 100 g mol À1 may be adequate for the entanglement with PHB molecules and inhibit crystallization. On the other hand, the chain length of lower molecular weight PLLAs (M w # 3600 g mol À1 ) should be too short for entanglement.

Size of PLLA molecules
Since the crystallization of PHB was revealed to have a complex thickness dependence, the relationship with the molecular size of PLLA should be considered; the molecular size of each PLLA is listed in Table 1 in the ESI. † Some might consider that small PLLA molecules (lower molecular weight PLLAs) would hinder the crystallization of PHB through interlamellar segregation because such shorter PLLA molecules are expected to easily diffuse into the PHB matrix and act as an effective obstacle for lamellae folding of PHB. However, our GIXD results have proved that such small molecules are ineffective at hindering lamellae folding. Small molecules of PLLAs most probably aggregate during the crystallization process. As a consequence, being unable to affect the crystallization of PHB. Similarly, big molecules of high molecular weight PLLAs would also be ineffective at reducing the crystallinity of PHB if the phase separation (as mentioned in point 1) between PHB and PLLA occurs. However, the molecules, whose sizes are comparable or exceed half of the lm's thickness, might effectively inhibit the crystallization of PHB because phase separation would not be expected if part of the PLLA molecules that can inltrate through the lms are pinned by the interaction with the substrate.
The crystallization behavior of PHB in the PHB/PLLA ultrathin blend lms of 30 and 13 nm thick lms observed in the present study is summarized in Fig. 4 and compared with bulk blends. As shown in Fig. 4, the inhibition effect by a small amount of PLLA on the crystallization of PHB is quite different from bulk blends. The long chains of high molecular weight PLLAs do not hinder the crystallization of PHB in bulk but effectively inhibit that in 13 nm thick lms. Similarly, the shorter chains of low molecular weight PLLAs effectively inhibit the crystallization of PHB in the bulk but do not have much effect in ultrathin lms.

Conformation and crystalline structure of PHB as studied by IRRAS
In order to explore the chain conformation and crystalline structure of PHB in the PHB/PLLA ultrathin lms on the molecular scale, IRRAS measurements were conducted.  45 With the addition of PLLAs, the intensity of the C]O crystalline band at 1726 cm À1 is signicantly reduced in the spectra of the blends where the molecular weight of PLLA $ 6900 g mol À1 (see spectra (1)- (6) in the bottom panel of Fig. 5). These results reveal that the presence of PLLA appreciably decreases the crystallization of PHB. According to the IRRAS selection rule, 60,61 the weak absorbance of the C]O crystalline band of PHB in the IRRAS spectra of the blends indicates that the C]O group is aligned more parallel to the substrate's surface. In the previous studies, the crystalline C]O groups of PHB in the ultrathin lms were also reported to be nearly parallel to the substrate surface (along the a-axis). 45,54 Further reduction of the molecular weight of PLLA (M w # 3600 g mol À1 ) in the blends yielded spectra that were more similar to that of neat PHB, as seen from (7)-(8) in Fig. 5. This indicates that these low molecular weight PLLAs neither stopped the crystallization nor changed the preferred alignment of the PHB crystallites. This result was consistent with the results obtained from the GIXD measurements (proles (7) and (8) in Fig. 1(a) and 2).
The two bands at 1764 and 1778 cm À1 in all of the blends were ascribed to the C]O stretching mode of amorphous PLLA and that of fairly structural defect of PLLA, respectively. 23,54,55 The intensity of the amorphous band at 1764 cm À1 of PLLA also decreased in the blends with decreasing the molecular weight of PLLA and eventually became very weak, as observed for the blend with the lowest molecular weight PLLA (see (8) in Fig. 5). The lowest molecular weight PLLA eventually aggregated in the lms as described above, and aggregated short PLLA chains are irrelevant to crystallization.
In the previous study, we revealed the behavior of two crystalline structures in PHB ultrathin lms with different structural orderings, that is, highly ordered and less ordered (intermediate) structures that can be characterized by bands at 1724 and 1731 cm À1 , respectively. They can be recognized aer deconvolution of the IRRAS spectra of PHB ultrathin lms where several overlapped bands coexist in the C]O stretching region of PHB due to the coexistence of the two crystalline structures. 45 In this study, the C]O stretching band region of the IRRAS spectra shown in Fig. 5 can be decomposed into several components using the GRAMS soware. Fig. 6(a) displays the result of decomposition, showing the bands of the highly ordered structure (1724 cm À1 ), the less ordered (intermediate) state (1731 cm À1 ), two amorphous states arising from different conformations of PHB (1739 and 1748 cm À1 ), one amorphous band of PLLA (1764 cm À1 ), and another band arising from structural defects of PLLA (1778 cm À1 ). 46 The integrated intensity of each crystalline structure at 1724 and 1731 cm À1 is plotted in Fig. 6(b) aer normalization by the sum of all fractions in each blend. In comparison with the intensity of the corresponding bands of neat PHB, the intensity of both the intermediate (1731 cm À1 ) and the highly ordered crystalline (1724 cm À1 ) bands became weaker in the blends, suggesting that the presence of PLLAs suppresses the formation of PHB crystals in the blends. While the integrated intensity of the intermediate state looked almost unchanged among the blend samples, the plot of the highly ordered crystalline band  showed a gradual increase in the blends with decreasing molecular weight of PLLA. This indicated that the variation in the molecular weight of PLLA does not affect the crystal formation of the intermediate state, but the higher molecular weight PLLA disturbs the crystal formation of the highly ordered state. Fig. 7 shows the IRRAS spectra (bottom) and their second derivatives (upper) in the C]O stretching region of the 13 nm thick lm of PHB and PHB/PLLA (80/20) blends with various molecular weights of PLLA. PHB still shows two bands due to the crystalline and amorphous C]O bands at 1726 and 1749 cm À1 , respectively. However, the intensity of the 1726 cm À1 band was lower than that of the 30 nm thick lms, presumably because the volume of crystallites decreased. Since a small amount of PLLAs (M w $ 13 100 g mol À1 ) greatly suppressed the intensity and altered the band shape of the C]O crystalline band of PHB, as noticeably observable in (1)-(5) of Fig. 7, we conclude that those PLLAs disturb both the crystallinity and the crystalline structure of PHB.
The high and middling molecular weight PLLAs caused the C]O crystalline band of PHB at 1726 cm À1 to split into two bands at 1723 and 1731 cm À1 , which presumably correspond to the highly ordered and intermediate crystalline structures of PHB, respectively. However, the IRRAS results would be inconsistent with the GIXD proles shown as (1)-(5) in Fig. 2, in which no reection from PHB crystals was recognized in the corresponding blend samples. The inconsistency between GIXD and IRRAS may arise from the different sensitivity of these techniques in detecting very tiny crystallites. 46 In the GIXD samples, the PHB crystals were too small and the amount of crystallites was also too small to be detected by the X-ray diffractometer equipped with a conventional X-ray tube. The intensity of C]O crystalline bands observed in the blends with low molecular weight PLLAs was found to be rather similar to that of neat PHB at 1726 cm À1 ( (7) and (8) in Fig. 7). This clearly indicates that the low molecular weight PLLAs are unable to alter the crystallization of PHB, which is independent of the lm thickness. Furthermore, the intensity of the PLLA amorphous band at 1765 cm À1 , which can be used as an indicator of PLLA chain entanglement, decreased as the molecular weight of PLLAs decreased, which agrees with the scheme discussed above.
To explore the crystallization behavior in the intermediate and highly ordered states, the spectra shown in Fig. 7 were also decomposed. A typical result is shown in Fig. 8(a) and displays the deconvoluted IRRAS spectra in the C]O stretching region of the 13 nm thick lms of PHB/PLLA (M w ¼ 300 000 g mol À1 ).  Plots of the integrated intensity of the 1724 cm À1 (highly ordered) and 1731 cm À1 (intermediate) bands, normalized by the sum of the integrated intensity in each blend, are depicted in Fig. 8(b). Similar to Fig. 6(b), the intensity of both intermediate and highly ordered crystalline structures decreased markedly aer the addition of PLLAs. However, the intensity of the intermediate state (1731 cm À1 ) showed a small increase with decreasing PLLA molecular weight. Furthermore, the intensity of the highly ordered crystalline state (1724 cm À1 ) greatly reduced at rst and subsequently increased with decreasing PLLA molecular weight. These results suggest that the molecular weight dependence of PLLA on the crystallization of PHB is greater for thinner lms.
In the C-H stretching region, weak intermolecular hydrogen bonding between carbonyl and methyl groups (CH 3 /O]C) of PHB was found at an anomalously high frequency region. [56][57][58][59] The weak hydrogen bonding bands of PHB were still recognizable in the 30 nm and 13 nm thick lms, as shown in Fig. 9. In the 30 nm thick lm, the weak hydrogen bonding band appeared at 3011 cm À1 as a weak shoulder band, while in the 13 nm thick lm, it shied to 3009 cm À1 and the peak became more apparent. These results may indicate that the hydrogen bonds in the 13 nm thick lm were weaker than the hydrogen bonds formed in the 30 nm thick lm. However, the slightly higher intensity of the 3009 cm À1 bands (indicated in Fig. 9(b)) would reect that, in the 13 nm thick lm, the number of bonds oriented nearly perpendicular to the lm surface seemed to be larger than that in the 30 nm thick lms.

Conclusions
We investigated the crystallization behavior of PHB in PHB/ PLLA ultrathin lms with thicknesses of 30 and 13 nm as a function of PLLA molecular weight using GIXD and IRRAS measurements. Eight PLLAs with different molecular weights (ranging from 300 000 to 710 g mol À1 ) were added to PHB (M w ¼ 650 000 g mol À1 ) in order to explore the effect of a small amount of PLLA on the crystallization behavior of PHB in ultrathin lm blends. The PHB/PLLA ratio was xed at 80/20 (w/w) for all of the blends studied. The crystallization of PHB showed a strong dependency on the molecular weight of PLLA and lm thickness. In the 30 nm thick lms, phase separation occurred between PHB and higher molecular weight PLLAs (M w $ 50 000 g mol À1 ) in the blends; therefore, those PLLAs were less affected the crystallization of PHB. The miscible PHB and PLLA blends were obtained when the middling molecular weight PLLAs (M w ¼ 23 000-13 100 g mol À1 ) were used, yielding dramatic decreases in the crystallinity of PHB. It was quite noteworthy that the crystallization of PHB was fully inhibited in the M w $ 6900 g mol À1 PLLA blends when the lm thickness was reduced to 13 nm.
IRRAS demonstrated that, for the 30 nm thick lms, the addition of a small amount of PLLA signicantly suppressed the intensity of the PHB crystalline band at 1726 cm À1 . However, the PLLAs did not affect the crystalline structures of PHB signicantly. Furthermore, for the 13 nm thick lms, both the intensity and the crystalline structures of PHB were remarkably altered by a small amount of PLLA (M w $ 6900 g mol À1 ). Both GIXD and IRRAS results consistently showed that the lower molecular weight PLLAs (M w # 3600 g mol À1 ) did not affect the crystallinity and crystalline structures of PHB. Furthermore, several factors, such as the free surface and interface effects, entanglement of PLLA chains, and molecular size of PLLA, must be taken into account to comprehend the complex crystallization behavior of PHB in PHB/PLLA ultrathin lms.

Conflicts of interest
The authors declare no competing nancial interest.