Heterogeneous Microstructures of Spherulites of Lipid Mixtures Characterized with Synchrotron Radiation Microbeam X-ray Diffraction

spherulites were grown from neat liquid, and from solution containing 50% n-dodecane and 50% POP+OPO. SR-μ-XRD analysis revealed heterogeneous distributions of MCPOP:OPO and component TAGs in every spherulite; TAG compositions in the inner and outer areas differed when the relative ratios of POP and OPO were changed. In the 75POP:25OPO spherulites, MCPOP:OPO always occupied the inner areas and POP dominated the outer areas as a result of different rates of 15


Introduction
Lipids are major nutrients, along with proteins and carbohydrates, and are employed as lipophilic materials in 25 food, cosmetic, and pharmaceutical industries. 1 Many lipids, such as triacylglycerols (TAGs) and diacylglycerols (DAGs), consist of multiple components in two ways: (i) a lipid phase contains a variety of lipids differing in hydrophilic and hydrophobic structures, and (ii) each single lipid molecule 30 involves different types of fatty acid moieties. Therefore, the physicochemical properties of lipids must be studied not only in their pure systems but also in mixed systems. In particular, studies on binary mixture systems provide valuable information about molecular interactions among different 35 lipid materials. The same analysis is applied to TAGs, which are the main components of natural and industrial lipids.
Diversified molecular interactions among TAGs are found in polymorphism and mixing behavior. Multiple polymorphic forms occur in almost all TAGs. The number of individual 40 polymorphic forms and their molecular structures drastically change when fatty acid compositions included in the TAGs are changed. 1 Moreover, for the binary mixing behavior of TAGs, three phases have been confirmed: (i) solid solution phase, (ii) eutectic phase, and (iii) molecular compound (MC) 45 formation phase. When two component TAG molecules exhibit structural similarity and affinitive molecular interactions, a miscible solid solution phase is formed. However, when two component molecules are immiscible due to steric hindrance, the eutectic phase is formed. [1][2][3] An 50 example of immiscible mixing systems is the formation of MC phase, which is formed through specific molecular interactions between two component molecules. 4 The properties of polymorphism and mixing behavior of TAGs are strongly related because the chemical nature 55 (saturated or unsaturated) of the component TAG molecules results in complicated polymorphism and mixing behavior. For instance, referring to the chemical nature and the unsaturation effect in monoacylglycerols 5 , Kulkarni et al. 6 examined the influence of manipulating double bonds in the 60 phase behavior and the number density of channels in inverse bicontinuous cubic phases. As to lipid mixtures systems, Seddon et al. 7 studied the phase behavior of saturated diacyl phosphatidylcholine/fatty acid mixtures with different chain lenghts as a function of water content. Later, Templer et al. 8 35 glycerol), OOP (1,2-dioleoyl-3-palmitoyl glycerol), PPO (1,2dipalmitoyl-3-oleoyl glycerol), and OPO (1,3-dioleoyl-2-palmitoyl glycerol). The mixtures of PPP+POP 9 , PPP+OOO 10 , and POP+OOP 11 are eutectic, whereas the mixtures of POP+PPO and POP+OPO are MC-forming at ratios of including n-dodecane as a solvent 14 , as this binary mixture is significant for fractionation of palm oil in dry and solvent methods.
Currently, most research on the binary mixing behavior of TAGs has been conducted on a thermodynamic equilibrium 5 phase diagram. However, crystallization behavior is significant for applications: for example, fractionation of palm oil to produce high-melting, medium-melting, and lowmelting fractions for multiple applications is influenced by crystallization properties of high-melting fractions such as 10 PPP, POP, PPO, OOP, and OPO. 15 This property is related to kinetic aspects of the binary mixing systems. However, few studies of kinetic properties of TAG mixtures have been conducted.
In the present study, we seek to analyze the microstructures 15 of spherulites of the mixtures of POP and OPO formed from neat liquid and solution including n-dodecane. Spherulites are the most typical crystal morphology of TAGs during cooling from neat liquid and solution. 16 Instead of growth of single crystal pieces separately, fat crystals tend to grow in the form 20 of spherulites at ambient temperatures and moderate cooling rates. 17, 18 The morphology and microstructures of spherulites are determined by crystallization kinetics, since the central region of a spherulite is formed by the nucleation of many tiny crystals, and post-nucleation crystal growth occurs 25 toward the external region. Thus, the microstructures of the spherulites of the mixture samples are determined by relative rates of nucleation and crystal growth of component materials. Therefore, detailed analysis of microstructures of spherulites provides deep insight into crystallization kinetics of long- 30 chain soft materials, as reported in polymer 19 and TAG crystals. 20 Our particular concern in this work was to examine how the microstructures of spherulites of the TAG mixtures are determined, when the MC crystals and POP or OPO 35 component crystal can be formed competitively. The polymorphic structures of POP, OPO, and MCPOP:OPO are summarized in Table 1. Figure 1 depicts phase diagrams of POP+OPO mixtures in neat liquid (Fig. 1a) 13 and n-dodecane (50%) solution ( Fig. 1(b)). 14 At a concentration ratio of The principle of SR-µ-XRD relies on X-ray focusing optics and a synchrotron radiation X-ray source, and it enables scanning in two dimensions of the sample, with steps on the order of the beam size. SR-µ-XRD has been used to examine 55 many soft materials (e.g., biological tissues, 22-31 starch, 32 proteins, and synthetic polymers 19,[33][34][35]. For example, Kajiura et al. 28 studied the keratin fibre arrangement of different types of human hair by SR-µ-XRD to determine the origin of curliness. Also, Seidel et al. 30 studied the orientation of chitin 60 fibres of the airflow sensors of crickets. SR-µ-XRD analysis of spherulites revealed different types of microstructures on a micrometer scale, which traditional diffraction techniques had not detected. 36 Our group was the first to apply SR-µ-XRD to lipid 65 samples, such as spherulite of trilaurin, 20 oil-in-water emulsion, 37,38 and granular crystals of palm-oil-based waterin-emulsion. 39 SR-µ-XRD analysis of trilaurin spherulites indicated that the lamellar planes of tiny crystals are aligned parallel to the radial direction from the central and peripheral 70 directions of the spherulite. Furthermore, the same study showed that the lamellar directions are not randomized after solid-state polymorphic transformation from β' to β forms.
In the present work, scanning SR-µ-XRD experiments of spherulites of the binary mixtures of POP and OPO clarified 75 the details of heterogeneous structures.

Experimental
Samples of POP and OPO (purity >99%) were purchased from Tsukishima Foods Industry (Tokyo, Japan) and used without further purification. The solvent, n-dodecane (99% pure), was 80 purchased from Nakalai Tesque (Kyoto, Japan). To prepare the binary mixtures of OPO and POP (wt%), two TAG samples were melted at 50ºC and mixed using a vortex. When n-dodecane was added to the POP+OPO mixtures, the ratio of n-dodecane/(POP+OPO) was 1/1 (50% solution).  Supplementary Information). The X-ray microbeam wavelength was 0.11nm, and the beam area was 5x5µm 2 . The sample was moved by an x-y-z stepping motor (1µm step) while being observed by an optical microscope. The X-ray data was detected by a CCD camera to produce two-95 dimensional (2D) patterns. Basically, it is possible to observe small-angle and wide-angle diffraction patterns simultaneously in the SR-µ-XRD experiments, as examined in O/W emulsion. 37 In the present study, however, we focused on only small-angle SR-µ-XRD patterns, so that diffraction peaks 100 corresponding to long spacing values (lamellar distances) could be observed with high resolution by expanding the 2D patterns at the small diffraction angle regions.
In order to determine the temperature at which spherulites that were large enough to be examined with the SR-µ-XRD 105 were grown, crystallization experiments were performed using a Linkam LK-600 PM stage (Tadworth, UK) mounted on an Olympus BX51 microscope (Tokyo, Japan). The system also consisted of an LNP liquid nitrogen cooling system and an L-600A temperature controller. Optical micrograph images were 110 taken with an Olympus DP12 Digital Camera.
After determining the optimal crystallization conditions, we prepared spherulites for the SR-µ-XRD experiments by the following process. A sample of molten mixture was placed   between two Mylar films to obtain thin layers in which crystallization occurred. The films were set on a temperaturecontrolled Linkam stage and quickly cooled from 50ºC to 25 crystallization temperature (Tc), at which the SR-µ-XRD experiments were carried out. The values of Tc are presented in Table 2. At least two different spherulites were measured in every case. In order to confirm the polymorphic forms of TAGs, 30 laboratory-scale X-ray diffraction (XRD) was also performed (RINT-TTR, Rigaku Co., Tokyo, Japan, λ = 0.154nm) using a rotator-anode X-ray beam generator.

35
Optical Characterization of Spherulites Figure 2 depicts polarized microscopic images of the spherulites of the mixtures of 75POP:25OPO and 25POP:75OPO grown from neat liquid and 50% n-dodecane solution. In Fig. 2(a) of 75POP:25OPO, large and many small 40 spherulites coexisted when grown from neat liquid. The diameters of the large spherulites were 150µm, and those of the small spherulites were less than 20µm. We observe aggregation of small spherulites between the large spherulites.
One may reasonably expect that the large spherulites 45 started growing later than the small spherulites. Although not shown here, we performed SR-µ-XRD analysis on the aggregated small spherulites. In these spherulites, the polymorphic occurrence and lamellar direction were quite random, and not every spherulite was distinguished from the 50 others. Therefore, SR-µ-XRD analysis was performed on only the large spherulites. It is evident that in every typical large spherulite of 75POP:25OPO depicted in Fig. 2(a), morphology in the central part differs from that in the peripheral part: a compact pattern is observed in the central area and needle 55 patterns appear in the peripheral areas. In contrast, for 25POP:75OPO grown from neat liquid (Fig. 2(c)), every spherulite has a uniform pattern of polarized-crossing without any separated inner textures. The spherulites grown from 50% n-dodecane solution have 60 uniform morphology in which needle patterns are arranged parallel to the radial direction from the central to the outer regions of the spherulites.

SR-µ-XRD Analysis
Spherulites grown from neat liquid  The samples were cooled from 50ºC to 16ºC for the 30 75POP:25OPO mixture, and 6ºC or 9ºC for the 25POP:75OPO mixture, where isothermal stabilization was applied to grow large spherulites before the SR-µ-XRD experiments.   three sharp arc peaks are identified, indicating the presence of 30 two different types of crystals having well-ordered lamellar orientations. These sharp peaks were analyzed in two ways: the 2θ extension provides information on the long spacing of crystals, and the χ extension indicates the lamellar direction of the crystals. 35 To precisely analyze the spatial distribution of MCPOP:OPO and POP crystals within the spherulite, Fig. 4 depicts enlarged 2D patterns and corresponding 2θ and χ extension patterns taken at three positions along the radial direction from the central to outer regions of the spherulite (denoted as positions 40 1, 2, and 3 in Fig. 3).
The 2θ extension pattern of position 1 indicates a single peak (denoted by ■) at 2θ = 2.0°, which corresponds to the 002 reflection of the 6.4nm long spacing. This value corresponds to the long spacing of βPOP-3. 13 The χ extension 45 has two sharp peaks at χ = 5° and 185°, although each peak is split into two. No peak is observed at any χ value except for these two peaks; therefore, we conclude that the lamellar planes of the crystals at this position are aligned along the direction denoted by broken lines in the figure. In this way, 50 we can interpret the 2D diffraction patterns of the other two positions.
Position 2 indicates two sets of arc peaks, providing three peaks in the 2θ extension (denoted by ■). The peaks at 2θ = 1.3° and 2.7° correspond to the 001 and 002 reflections 55 of the 4.6nm long spacing, which corresponds to βMC-2. The peak at 2θ = 2.0° corresponds to the 002 reflection of the 6.4nm long spacing of βPOP-3. From these results, we can conclude that two kinds of crystals (βMC-2 and βPOP-3) exist at this position. Interesting results are indicated for the 60 χ extension, since the arrangements of lamellar planes of the βMC-2 and βPOP-3 crystals differ. In the χ extension patterns of the βMC-2 crystals, broad peaks appear at maximum χ values of 15° and 195°; however, sharp peaks appear at 45° and 225° for the βPOP-3 crystals. This result indicates that the 65 lamellar directions of the βMC-2 and βPOP-3 crystals differ by 30° at this position.
Finally, at position 3, which is near the central part of the spherulite, only βMC-2 crystals exist and have a randomly oriented lamellar arrangement. This conclusion could be 70 drawn by observing that the peaks at 2θ = 1.3° and 2.7° correspond to the 4.6nm long spacing and that the diffraction peaks have almost the same intensity at all    the azymuthal angles, although the patterns were highly noisy. From the results presented in Fig. 4, it was evident that the predominant TAG crystals varied from βMC-2 to βPOP-3 when the microbeam scanning positions changed from the central 25 regions to the outer regions in the same spherulite. Figure 5(a) illustrates the spatial distribution of βPOP-3 and βMC-2 crystals in the spherulite depicted in Fig. 3 and indicates their lamellar directions (Fig. 5(b)). The βMC-2 crystals with random lamellar orientation occupy the central 30 part of the spherulite (area of 30 x 40µm 2 ). However, predominant crystals vary from βMC-2 to βPOP-3 when the position moves from the central to the peripheral areas of the spherulite, as explained above. In addition, the lamellar plane directions of the βPOP-3 crystals are arranged parallel to the 35 radial direction of the spherulite. Between the periphery and the central areas of the spherulite are some areas where βMC-2 and βPOP-3 crystals co-exist. The lamellar directions of the βMC-2 and βPOP-3 crystals are almost parallel, with some exceptions (e. g., position 2 in Fig. 4). 40 Figure 5(a) also depicts many crystals whose lamellar planes make right angles at the periphery areas of the spherulite (green lines). These areas overlap with the neighbouring spherulites (see Fig. 2(a)). Interestingly, the crystals of the neighboring spherulites are all βPOP-3. This 45 result indicates that perhaps the heterogeneous structure of the spherulite examined in Figs. 3 through 5 also exists in the neighboring spherulites that have βPOP-3 crystal in their peripheral zones. Figure 6 presents the results of SR-µ-XRD experiments of 50 the spherulites of the 25POP:75OPO mixture grown from neat liquid at 9°C and 6°C. Two spherulites were analyzed, following the same methodology than in the case before, over the whole area by scanning SR-µ-XRD studies at each temperature, and the same results were obtained. It must be 55 noted that microbeam X-ray scanning was performed at 10µm distance in both cases. For spherulites grown at 9°C (Fig. 6(a)), concurrent crystallization of βMC-2 and βOPO-3 was observed at almost every position. However, the concentration of βOPO-3 crystals 60 was always higher than that of βMC-2 at the positions where the two crystals co-existed. Furthermore, only βOPO-3 crystals were observed at some places. Two typical examples are the 2θ extension patterns at positions 1 and 2 in Fig. 6(a). There are three peaks at position 1 (denoted by ■). The peaks at 65 2θ = 1.3° and 2.7° correspond to the 001 and 002 reflections of the 4.6nm long spacing of βMC-2, whereas the peak at 2θ = 2.0 ° corresponds to the 002 reflection of the 6.4nm long spacing of βOPO-3. The relative concentration ratio of the βOPO-3 and βMC-2 crystals should be compared by considering 70 the peak intensity of the 002 reflections. It is clear that the peak intensity of βOPO-3 is much higher than that of βMC-2. As an extreme case, position 2 has only one peak of βOPO-3. Thus, the predominant crystallization of βOPO-3 over βMC-2 was confirmed at 231 of the 241 positions in Fig. 6(a) 20 other positions, either βOPO-3 and βMC-2 or only βMC-2 was predominant, but this feature was not common to all the positions of this spherulite. The predominant crystallization of βOPO-3 thus covered 95% of the area in this spherulite. 25 The lamellar directions of the βOPO-3 and βMC-2 crystals in Fig. 6(a) were parallel to the radial direction of the spherulite at the outer regions. However, the lamellar directions were more or less random in the central region. From analysis of the spherulites of the 25POP:75OPO mixture grown at two temperatures (Fig. 6), we conclude that the rate of crystallization of βOPO-3 crystals was higher than 40 that of βMC-2 crystals in the range of growth temperatures examined. 45 When n-dodecane was added to the binary mixtures of POP and OPO, SR-µ-XRD intensity decreased compared to that of the neat liquid system. Nevertheless, the behavior of the samples and the results obtained were the same as for those grown from neat liquid for the mixtures of both   Fig. 8(a), with 5µm between consecutive positions. The results were the same as those observed in 25POP:75OPO mixture grown from neat liquid: βOPO-3 predominated over βMC-2. Namely, βOPO-3 was always observed, and no βMC-2 75 crystals appeared. The lamellar plane directions of the βOPO-3  Table 3 Long spacing values of βPOP, βOPO and βMC crystals obtained with synchrotron radiation microbeam X-ray-diffraction (SR-µ-XRD) and synchrotron radiation macrobeam X-ray-diffraction (SR-XRD) SR-µ-XRD SR-XRD (neat liquid) 13 Triple  Triple  Double  Triple  Triple  Double  Triple  Triple  Double   5

SAXD and WAXD Patterns
We have presented the SR-µ-XRD results of the spherulites of 75POP:25OPO and 25POP:75OPO mixtures grown from neat 25 liquid and n-dodecane solutions, and we examined them with small-angle X-ray diffraction (SAXD). The long spacing values of the POP, OPO, and MCPOP:OPO crystals examined in the present SR-µ-XRD study are summarized in Table 3, compared with those reported in previous analysis with 30 conventional synchrotron radiation X-ray diffraction (SR-XRD). The chain-length structures of the POP, OPO, and MCPOP:OPO crystals are all the same in the three studies: MCPOP:OPO crystals have a double-chain-length structure, and the other two components have triple-chain-length structures. 35 Although the precise long spacing values differ slightly (by 0.1~0.3nm) among the studies with SR-µ-XRD and SR-XRD, this uncertainty is within experimental error. However, in order to confirm the polymorphic forms of TAG crystals, we observed the wide-angle X-ray diffraction (WAXD) peaks 40 using RINT because WAXD patterns are more sensitive to subcell structures, such as tricilinic parallel (T//) for β polymorph or orthorhombic perpendicular (O⊥) for β' polymorph.
For this purpose, we performed laboratory-scale XRD using 45 a RINT apparatus (laboratory scale X-ray diffraction). As with XRD studies of crystals grown from n-dodecane solutions containing 50% of fat samples, it was quite difficult to obtain clear diffraction patterns with RINT because the 60 diffraction peaks from the solution samples were too weak. However, the mixing behavior of POP:OPO in n-dodecane solution examined with SR-XRD demonstrated that the equilibrium states contained β forms of POP, OPO, and MCPOP:OPO. 14 As equilibration from metastable to most stable 65 forms occurs much faster in the solution system than in neat liquid systems, we conclude that the spherulites of the mixtures of POP:OPO grown from n-dodecane solution are all β forms.

70
In the present study, we observed the microstuctures of spherulites of 75POP:25OPO and 25POP:75OPO with SR-µ-XRD. Mixtures having these two ratios were chosen because previous studies indicated that the averaged concentration ratios of POP:MCPOP:OPO and OPO:MCPOP:OPO are 50:50 in the 75 mixtures of 75POP:25OPO and 25POP:75OPO, due to the fact that POP and OPO form MCPOP:OPO crystals at the ratio of 50:50. Therefore, we expected that the microstructures of spherulites might not be homogeneous, due to differences in the rates of crystallization of POP, MCPOP:OPO, and OPO, and 80 Table 4 Occurrence frequency of βOPO and βMC in 25POP:75OPO mixture grown from neat liquid examined by scanning SR-µ-XRD at different positions over many small spherulites at two crystallization temperatures (Tcs) 5 Tc=6 ºC  20 that such heterogeneous structures can be analyzed solely with SR-µ-XRD experiments, as reported in our previous studies of spherulites of a simple TAG system, 20 granular crystals in fat spreads, 39 and oil-in-water emulsion droplets. 37,38 The results obtained by the present experiments are 25 summarized in the following.
(1) For mixtures of 75POP:25OPO, large spherulites grown from both neat liquid and n-dodecane solution were dominated by MCPOP:OPO in the inner regions, whereas the outer region was dominated by POP. From this result, we 30 conclude that the relative crystallization rate of MCPOP:OPO was higher than that of POP because we can reasonably assume that crystals having higher rates of nucleation may first occur in the central part of the spherulite, and successive nucleation and crystal growth having lower rates of nucleation 35 may occur afterward and extend to the outer part of the spherulite at the Tc examined.
(2) For mixtures of 25POP:75OPO, the entire area was homogeneously occupied either with almost exclusively OPO or with coexisting MCPOP:OPO and OPO (Table 4). This result 40 indicates that the relative nucleation rate of OPO is higher than, or at least similar to, that of MCPOP:OPO at the Tc examined.
Based on these results, we may determine relative rates of crystallization of POP and MCPOP:OPO in the 75POP:25OPO 45 mixture, and OPO and MCPOP:OPO in the 25POP:75OPO mixture, based on the following considerations. (a) Spherulite formation starts from the inner regions and continues to the outer regions. (b) As crystallization is performed during rapid cooling, no 50 difference exists between crystallization temperatures at different positions of the spherulite, and the crystallization temperature is quickly reached.
(c) Heterogeneity in the occurrence of different crystal fractions is caused by relative rates of nucleation: the higher 55 the nucleation rate, more dominated is the inner region of the spherulite.
As the nucleation rate of crystals largely increases with increased supercooling (∆T), 40 it is necessary to take into account the effects of ∆T on crystal nucleation when we 60 compare the relative nucleation rates of POP, OPO, and MCPOP:OPO. values at all Tc. However, differences in ∆T between the crystals decrease when Tc increases.
Taking into account the effects of ∆T discussed above, we may reasonably assume the relative rates of nucleation of β forms of POP, OPO, and MCPOP:OPO at different Tc as shown 75 in Fig. 10.
Thus, the nucleation rate of βMC may be higher than that of βPOP at large ∆T values in the mixtures of 75POP:25OPO ( Fig. 10(a)). In this case, Tc may be below the temperature where the nucleation rate of βMC exceeds that of βPOP. In 80 contrast, the nucleation rate of βOPO may be slightly higher than, or similar to, that of βMC in the mixtures of 25POP:75OPO at the crystallization temperatures examined in the experiments (Fig. 10(b)). It is highly significant to observe directly the relative nucleation rates of β forms of POP, OPO, on different polymorphic forms of OPO using a synchrotron radiation X-ray beam. 41

30
Microstructural studies of spherulites of the mixtures of POP+OPO were examined at 75POP:25OPO and 25POP:75OPO in a neat liquid system and with 50% ndodecane by using SR-µ-XRD. This technique enabled the scanning of samples in two dimensions to determine the exact 35 composition at each point, with a difference in distance of 5µm or 10µm. The results indicated that at 75POP:25OPO a βMC is located in the inner part of the spherulite and βPOP appears near the outer part. In contrast, for 25POP:75OPO, βOPO predominates over βMC at the crystallization 40 temperatures examined. This heterogeneous distributions were due to the different crystallization rates of MCPOP:OPO and component POP and OPO crystals, which were also confirmed by polarized microscopy: the morphology of spherulites of 75POP:25OPO grown from 50% n-dodecane and 45 25POP:75OPO grown from neat liquid and 50% n-dodecane were homogeneous, whereas those of 75POP:25OPO grown from neat liquid were heterogeneous (see Fig. 11). Although further study of the relative nucleation rates of these polymorphic forms is required, knowing the compound 50 distribution in the spherulitic system aids in the comprehension and prediction of growth mechanisms.