On the reliability of highly magnified micrographs for structural analysis in materials science

One of the great strengths of electrospinning, which has contributed to its success in recent years, is the ability to process a wide variety of polymers with different additives into nanofibers25. Although pure homopolymers usually result in very homogeneous, large-area nonwovens regardless of the type of electrospinning, the addition of further polymers, nanoparticles and other additives leads to increasingly heterogeneous micro and nanostructures26. The occurrence of so-called fiber beads, agglomerates, network structures and membranous regions can be observed27.

Figure 1a,b depict TEM images with different magnifications, taken on a PAN/nickel ferrite nanofiber cross-section. From these two images, it could be concluded that the nickel ferrite nanoparticles are distributed relatively evenly along the cross-section of a nanofiber, with small regions of few 100 nm in-between completely lacking nanoparticles.

Figure 1

(ae) TEM images of individual PAN/nickel ferrite nanofibers; (f) highly magnified HIM image from Fig. 3h for comparison.

Figure 1c, however, provides a completely different perspective. Here it is evident that the nanoparticle-free areas between agglomerates can also have dimensions of several micrometers, and the nanoparticles are distributed very asymmetrically along the fiber. Depending on the image chosen for a paper, completely opposed conclusions on nanoparticle distribution in the fibers can be drawn.

Images such as Fig. 1c–e, which depict the cross-section along the axis of symmetry of the fibers, are rarely encountered in journal articles, as they are less aesthetically pleasing, but often contain more or at least important insights. Ignoring such micrographs may deprive an article of crucial information, but on the other hand, it may facilitate the peer review process, as they would most likely be criticized and possibly rejected by reviewers or editors.

For comparison Fig. 1f shows a similar nanofiber from the outside. Here it can be seen that large particle agglomerates are not completely embedded in the polymer and partially burst out of the fiber. Taking a micrograph 5 µm further along the fiber at the same magnification could easily be used to confirm an even particle distribution inside the fiber as seen in Fig. 1a.

For PAN nanofibers with embedded magnetite nanoparticles, similar conclusions can be drawn, although here agglomerations are less pronounced. Figure 2 shows a comparison of different nanofiber cross-sections, which could be interpreted either as mostly even distribution of nanoparticles or as a mixture of agglomerates with large particle-free regions in-between, depending on the chosen micrograph. During sample preparation by ultramicrotomy, the relatively large nanoparticles are sometimes displaced. The effect could be misinterpreted as voids or air pockets and is more pronounced in some images than in others. It should be mentioned that the term even distribution necessitates a clear definition, e.g. whether the distances between particles inside the fiber are randomly spaced or can be given by a Gaussian distribution with a certain standard deviation; or whether the numbers of particles per matrix volume are in a certain range; or whichever definition is best suited for the respective physical property to be examined.

Figure 2

(ae) TEM images of individual PAN/magnetite nanofibers; (f) highly magnified HIM image from Fig. 4b for comparison.

Again, Fig. 2f shows a HIM image at comparable magnification, indicating that the magnetite nanoparticles can also agglomerate outside the polymeric fiber matrix. Although HIM allows for charge compensation during secondary electron detection by means of an electron flood gun, the samples in this study were sputter-coated with approximately 10 nm of gold, as this is also required for the much more prevalent SEM. The coating thickness is below the average magnetite nanoparticle diameter of 50–100 nm, but still sufficient to allow misinterpretations of their shape (which is rather cubic than round) and to mimic a polymer layer above the nanoparticles. The sputter coating can be clearly distinguished by the icecap-like topology along the sides of the fiber resulting from shadow casting during unidirectional gold deposition.

More HIM images depicted in Fig. 3 show the same PAN/nickel ferrite nonwoven from the bird’s eye view. The overview image in Fig. 3a with a field of view of 500 × 500 µm2 shows that the nanofiber nonwoven is highly inhomogeneous and brimming with numerous big agglomerates and membranous regions. In order to illustrate the heterogeneity of the nonwoven, Fig. 3b–i show magnified subsections of the overview with a commonly used field of view of 50 µm × 50 µm each. Showing only Fig. 3g or i in a manuscript would suggest the conclusion that the sample consists of only straight uniform nanofibers, possibly with small beads, as often seen in electrospun PAN from a DMSO solution28,29. A completely different story, however, is told by Fig. 3e,f,h, clearly showing large agglomerates of polymer and nanoparticles with dimensions of more than 10 µm. The difference between HIM and SEM is particularly noticeable in the outstanding depth of field of the HIM images30,31. Several microns can be seen into the nonwoven through the first fiber layers, while deeper lying fibers are still sharply resolved.

Figure 3

HIM images of a PAN/nickel ferrite nanofiber nonwoven sample: (a) overview; (bi) magnified images taken at different regions, as defined in a, with inserted fiber diameter histograms, normal distributions and mean fiber diameters in nm. Overall mean fiber diameter is (460 ± 245) nm.

These qualitative differences are also reflected in a quantitative analysis, as can be seen from the inserted diameter histograms. Even excluding bigger agglomerates from the diameter evaluation, the average values fluctuate significantly with variable distribution widths. A closer look reveals some differences between typical Gaussian-shaped distributions (Fig. 3c,h) and others, suggesting asymmetric (Fig. 3g,i) or even double-peak distributions (Fig. 3d).

In the same way, HIM images were taken of a PAN/magnetite nonwoven sample, as seen in Fig. 4. Again, some of the magnified images, such as Fig. 4e or h, could unhesitatingly be shown in a manuscript, while others would probably be considered inadequate, such as Fig. 4b,d,f with large membranous regions or Fig. 4g with a disproportionally thick microfiber crossing the image area. As the images show, the samples are so heterogeneous that after a short search with the appropriate magnification, practically any desired hypothesis could be verified. Clearly, TEM micrographs can image a wide range of fiber or membrane cross-sections with any desired nanoparticle arrangement. In this context, it is not even clear whether images like Figs. 1d and 2a show fibers along their longitudinal axis or cross-sections of membranous regions.

Figure 4

HIM images of a PAN/magnetite nanofiber nonwoven sample: (a) overview; (bi) magnified images taken at different regions, as defined in a, with inserted fiber diameter histograms, normal distributions and mean fiber diameters in nm. Overall mean fiber diameter is (547 ± 340) nm.

While there are many applications for which a heterogeneous structure at the nanometer scale is irrelevant, there are certainly some, where it has a significant influence on the macroscopic properties. Here, this is demonstrated using the example of magnetic properties. These are known to depend strongly on the nanoparticle dimensions, where agglomerates can be expected to show completely different shape anisotropies than individual nanoparticles32,33.

Figure 5 shows hysteresis loops that were simulated based on different nanoparticle arrangements corresponding to different TEM images. Additional simulation results of agglomerated/distributed magnetite nanoparticles, elaborating on the influence of distribution and agglomerate size, can be found in Fig. S2 of the supporting information. As expected, the hysteresis loops show significant differences in terms of the slopes of the curves and, in case of magnetite, even in the coercive fields depending on whether the respective nanoparticles are evenly distributed or agglomerated. The simulations underline that a selective presentation of micrographs has a decisive influence on the predictive power of models based on them.

Figure 5

Simulated hysteresis loops for evenly distributed and agglomerated nickel-ferrite and magnetite nanospheres, respectively.

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *