Biominerals at the nanoscale: transmission electron microscopy methods for studying the special properties of biominerals
Published:January 01, 2013
- PDF LinkChapter PDF
Mihály Pósfai, Takeshi Kasama, Rafal E. Dunin-Borkowski, 2013. "Biominerals at the nanoscale: transmission electron microscopy methods for studying the special properties of biominerals", Minerals at the Nanoscale, F. Nieto, K.J.T. Livi, R. Oberti
Download citation file:
Biominerals have important functions in living organisms: apatite crystals are responsible for the strength of our bones and the hardness of our teeth, calcite and aragonite are used by many organisms for making shells, and magnetite and greigite help bacteria and birds to navigate in magnetic fields. In order to fulfill their roles in organisms, bio-minerals have strictly controlled physical and chemical properties. Transmission electron microscopy (TEM) is ideally suited for the study of the structures, arrangements, compositions, morphologies, crystallographic orientations, crystallographic textures, and magnetic properties of biominerals at the nanoscale. In this chapter, we review the state of the art in the application of TEM techniques to the study of these properties, both in biomineral crystals and at the inorganic-organic interface. Examples are taken primarily from studies of magnetic minerals that form in the cells of magnetotactic bacteria.
Biomineralization is a process by which mineral-like objects form as a result of the functioning of living organisms. The word ‘biomineral’ is used widely, and somewhat inconsistently, in the literature to refer to solid substances that can be crystalline or amorphous (e.g. calcite or amorphous calcium carbonate), inorganic or organic (e.g. oxalates), and can form either inside or outside cells (Mann, 2001). Some organic substances in living organisms have periodic structures (such as two-dimensional protein S-layers), yet are not considered to be biominerals. Despite attempts to narrow down the definition of the term ‘biomineral’ and to introduce the term ‘organomineral’ (Défarge & Trichet, 1995; Perry et al., 2007), it appears that the term ‘biomineral’ continues to be used in the literature to describe all minerals that form as a direct or indirect consequence of life (Weiner & Dove, 2003; Baeuerlein, 2007; Meldrum & Cölfen, 2008).
The study of biomineralization is necessarily a highly interdisciplinary field of science, drawing on knowledge from biology, mineralogy, chemistry, materials science, geology and other disciplines. The significance of biominerals also extends across many fields of science. For example, topics of current interest in biology include the genetic basis of biomineral formation, the functions of biominerals and their history throughout evolution (Evans, 2003; Baeuerlein, 2007). For earth scientists, the formation of biominerals is important because they accumulate in rock-forming quantities, can be used to understand past environmental and climatic changes, and provide an important record of Earth’s history (Knoll, 2003; Banfield et al., 2005; Hazen et al., 2008). Solid-state chemists and materials scientists are interested in the strict control that organisms can exert over the properties of biominerals (Meldrum & Cölfen, 2008), so a new field, which is referred to as biomimetic synthesis, has been developed based on principles derived from studies of biomineralization processes (Mann, 2001; Behrens & Baeuerlein, 2009). Medical research has been carried out on several aspects of biomineralogy, including the health of bones and teeth, the development of biomimetic implants, interactions between minerals and the human body (e.g. nanoparticle toxicity), and mineral-based diagnostic and clinical methods (Pompe et al., 2009). In each of these fields, a thorough knowledge of the physical and chemical properties of biominerals is a prerequisite for understanding the relevant biomineral-forming processes and biomineral functions. As biominerals are typically nanocrystalline, their study requires the development and application of techniques that can be used to provide information at the nanoscale. Among these techniques, transmission electron microscopy (TEM) arguably offers the greatest versatility.
This chapter focuses on the application of TEM techniques to the study of biomineral properties. While most of the methods that are described below are applicable to TEM studies of any type of material, biominerals present some specific challenges for electron microscopy: they typically have small grain sizes; they may be sensitive to the electron beam and/or they are sometimes embedded in organic material. In addition, the questions that need to be answered may differ from those in materials science or in studies of more ‘standard’ mineralogical samples. Therefore, after a brief overview of biomineral types and biomineral formation mechanisms, this chapter addresses how specific properties of biominerals (including morphology, composition, structure, orientation and magnetism) can be studied using a variety of TEM techniques. In each section, there is a progression from techniques that can be termed ‘conventional’ to those that represent the state of the art. A section on specimen preparation and on specialised techniques for imaging mineral/organic interfaces then follows. A brief overview of other techniques that can be combined with TEM concludes the chapter.
Within the limits of a textbook chapter, it is impossible to cover all of the interdisciplinary aspects of nanoscale biomineral research. Instead of attempting to provide a comprehensive review, we select examples that are taken mostly from our own field of interest: biomineralization by magnetotactic bacteria. In addition, selected examples from studies of biominerals from other organisms are shown in order to highlight certain aspects or advantages of a particular TEM technique. For more comprehensive treatments of biomineralization, the reader is referred to several excellent books and compilations that have been published on this topic in recent years (e.g. Mann, 2001; Dove et al., 2003; Banfield et al., 2005; Baeuerlein, 2007; a thematic issue of Chemical Reviews (Estroff, 2008); Behrens & Baeuerlein, 2009).
Primary biomineral-forming processes and biomineral groups
Biomineral-forming processes can be classified as either biologically controlled mineralization (BCM) or biologically induced mineralization (BIM), based on whether the bio-mineral is used by an organism for a biological function (as in BCM), or is the byproduct of an organism’s metabolism (as in BIM) (Lowenstam, 1981; Mann, 2001; Weiner & Dove, 2003). BCM is typically (but not always) associated with the formation of intra-cellular minerals under strict biological control (e.g. Fig. 1a). Organisms use BCM minerals for mechanical functions (protection, support, motion, cutting and grinding), magnetic, gravitational and optical sensing, and storage (Table 1). In order to perform these functions, special properties are needed, such as uniform particle size, unusual or complex crystal morphology, strictly controlled composition and structure (polymorph selection), controlled aggregation, crystallographic orientation and texture (i.e. the organized orientations of assemblages of crystals). It is also common that BCM minerals are assembled into hierarchical structures as, for example, composites of nm-sized apatite crystals and collagen fibres form ordered bundles that are assembled into columnar structures that are, in turn, arranged to form bone (Mann, 2001).
Biologically induced mineralization usually results in the formation of extracellular minerals, although intracellular BIM is also known (Weiner & Dove, 2003; Frankel & Bazylinski, 2003). It should be noted that the term ‘biologically influenced mineralization’, which is distinguished from BIM in some studies to indicate the passive role of an organism in a mineralization process (Dupraz et al., 2009), is treated in this chapter under the heading of BIM. Several metabolic processes can result in the precipitation of minerals at or near a cell wall. In these cases, there is no strict biological control over the properties of the mineral, although it may show a preferred orientation or morphology as a result of templated nucleation if it is precipitated on the cell wall itself. Well known examples of BIM include carbonate precipitation as a result of photosynthesis-induced changes of water chemical composition, and bacterially induced formation of metal oxides (e.g. Fe oxides and oxyhydroxides (Kasama & Murakami, 2001)) and sulfides (e.g. pyrite formation in anoxic marine sediments (Schoonen, 2004)) (Fig. 1b; Table 1). BIM processes can be utilized in practical applications, such as in bioremedia-tion (metal sulfide formation in contaminated environments) or in the enrichment of metals from their ores (Pósfai & Dunin-Borkowski, 2006). Pathological minerals are also produced by BIM, including urinary and gall bladder stones, calcification in arteries, and the formation of metal oxides in the human brain (Baeuerlein, 2007; Collingwood et al., 2008).
The number of mineral species that can form as a direct or indirect consequence of biological activity may be more than 2800 (Hazen et al., 2008), from a total number of mineral species accredited by the International Mineralogical Association of 4834 at the time of writing (January, 2013). Biological mediation in mineral formation may be indirect in some cases, one example being the effect of the availability of oxygen in the atmosphere as a result of oxygenic photosynthesis by the biota on the production of oxide, carbonate, sulfate and other minerals on or near Earth’s surface. Life on Earth has clearly played a major role in shaping the inorganic mineral world as we know it today (Hazen & Ferry, 2010). It is not possible to list all of the mineral species that form as a result of direct or indirect interactions with the biota; instead, a subjective selection is shown in Table 1, listing those groups of minerals that are regarded as being the most important or widespread, alongside typical functions and host organisms for BCM minerals, and typical formation processes for BIM minerals.
Special properties of biominerals and TEM techniques for their study
The study of biomineral properties using TEM is often challenging. Therefore, it is interesting to observe that, as new TEM techniques are developed and made available, they are quickly applied to biomineral research. In some cases, e.g. when applying off-axis electron holography or electron tomography to studies of crystalline nanoparti-cles, specific problems in biomineral research have motivated technique development. In this way, interactions between scientists working on biomaterial research and electron microscopy technique development have resulted in significant progress and innovations over the past ∼ 15 years, notably in specimen preparation, imaging and analysis techniques. The following subsections are devoted to techniques that can be used to study each of the properties listed in Table 2, ranging from standard to cutting edge techniques.
As most of the examples presented in this chapter are taken from studies of magneto-tactic bacteria, an initial brief description of these organisms is warranted. Cells of mag-netotactic bacteria contain intracellular ferrimagnetic nanoparticles of either magnetite (Fe3O4), greigite (Fe3S4), or both (Bazylinski & Frankel, 2004; Blakemore, 1975; Lefèvre et al., 2011a). The magnetic nanoparticles precipitate inside membrane-bound vesicles to create assemblages of membranes and crystals that are termed ‘mag-netosomes’. Nucleation and growth of intracellular magnetic nanocrystals take place under the strict genetic control of proteins that are located in the magnetosome membrane, resulting in sizes, morphologies, and arrangements of magnetite or greigite nanocrystals that are specific to each bacterial strain (Faivre & Schüler, 2008). Each nanocrystal usually contains a single magnetic domain (Dunin-Borkowski et al., 1998). Adjacent magnetosomes are typically arranged in chains so that the cell behaves like a miniature compass needle, with an external magnetic field aligning the cell parallel to an applied magnetic field when it is in an aqueous environment (Fig. 2). Following this passive alignment, the cells swim parallel or antiparallel to the magnetic field direction, i.e. they perform magnetotaxis. This trait of magnetotaxis is beneficial for bacteria that live in chemically stratified environments, because they can find their optimal position faster in a vertical chemical gradient by swimming along a straight line instead of searching for local variations in chemistry in three dimensions (Fig. 2).
Biominerals are typically synthesized in the form of nanoscale particles. For BCM minerals, their strictly controlled sizes are essential for fulfilling their biological functions (Gilbert & Banfield, 2005; Mann, 2001). BIM minerals form in near-surface environments (in water, sediments and soils), where they form fine-grained masses that have large specific surface areas (Banfield & Zhang, 2001). Thus, the measurement of biomineral particle sizes is of great importance for understanding the origins and functions of BCM minerals and the roles of BIM minerals in element cycles in the environment. Specifically, measurements of size distributions provide information about the origins and formation mechanisms of distinct particle populations.
A key function of a transmission electron microscope is imaging; the most obvious piece of information obtainable is the size of a recorded object. Many types of commercially or freely available software are capable of ‘automatically’ locating and measuring the sizes of particles in digitized images. Typically, the image is processed in order to enhance its contrast and to subtract its background intensity. Then, features of interest are identified on the basis of a preset intensity threshold. Finally, the regions of interest are outlined, and their dimensions are measured and listed, as shown in Figure 3. Particles that have irregular outlines can be characterized using parameters derived from area-equivalent circles or ellipses.
In practice, images of biomineral samples rarely lend themselves to simple automated procedures for particle size analysis because of the heterogeneous contrast of the surrounding biological tissue. By using global intensity thresholding, particles of interest often cannot be distinguished from one another or from non-mineral objects that are also present in the same image. In such cases, visual identification and manual outlining of particles, which is extremely labour-intensive, can provide improved information about the physical properties and formation mechanisms of distinct particle populations. Similar measurements can now also be obtained automatically by using a computational approach that makes use of local, rather than global, intensity thresholding (Cervera Gontard et al., 2011).
In the aqueous, typically room-temperature, environments of biosystems, the sizes of crystals are determined by rates of nucleation and growth, which are, in turn, affected by the availability of the crystal’s building units (‘nutrients’) and by the type of system (open or closed) in which crystallization occurs. The shape of the histogram of a crystal-size distribution can be related to certain processes. For example, in an open system in which there is an unlimited supply of nutrients, the rate of crystal growth depends primarily on the surface area that is available for the attachment of ions or ion clusters from solution. The resulting size distribution is lognormal (Eberl et al., 1998). Random processes in an open system (when the amount of material added to the crystal is independent of the crystal’s previous size and is not limited by the availability of nutrients) result in a normal (Gaussian) distribution. Alternatively, Ostwald ripening in a closed system produces a negatively skewed size distribution. In most natural inorganic systems (e.g. the formation of microcline and quartz crystals in a pegmatite (Kile & Eberl, 1999), or aerosol particles by wind erosion (Pósfai & Molnár, 2013)), the sizes of particles follow lognormal distributions. In contrast, biominerals typically have ‘unusual’ size distributions as a result of strict control by the organism over crystal nucleation and growth, particularly for BCM processes. Thus, distinct features of biomineral size distributions can be used as biomarkers.
The identification of objects of biogenic origin in rocks has been a hotly debated topic over the past 20 years (Benzerara & Menguy, 2009). ‘Biomorphs’, i.e. objects with morphologies that resemble biological organisms or parts of them, have been identified in various types of rocks, including meteorites (e.g. the Ivuna carbonaceous chon-drite (Hoover & Rozanov, 2011) and the Martian meteorites ALH84001 (McKay et al., 1996), Nakhla and Shergotty (Gibson et al., 2001)). However, the interpretation of the shapes of nano- or micro-scale objects as biogenic is at best inconclusive or ambiguous. Doubts have even been raised about the biogenic origin of the famous, elongated, segmented features in the 3.5 billion-year-old Warrawoona chert in Australia, which had been widely regarded as containing the oldest fossils on Earth, as a result of inorganic experiments that produced very similar objects by heating carbonates in silica (García-Ruiz et al., 2003).
Instead of the morphologies of objects, their size distributions can be used for identifying their biogenic or abiogenic origins. In the case of magnetofossils, i.e. ferrimagnetic nanoparticles produced by organisms, several physical and chemical features have been suggested for use as biomarkers (Kopp & Kirschvink, 2008). The clearest sign of the biogenicity of magnetic nanoparticles is likely to be a negatively skewed size distribution, which is thought to be unique to magnetotactic bacteria (Fig. 4). Accordingly, the BCM origins of magnetite and greigite particles in sediments and rocks have been identified on the basis of their negatively skewed or normal size distributions (Arató et al., 2005; Pósfai et al., 2001; Vasiliev et al., 2008). This is in contrast to particle populations that produced lognormal distributions and are likely to have resulted from abiogenic or BIM processes (Fig. 4). Similarly, lognormal size distributions of magnetite particles in the ALH84001 Martian meteorite (Fig. 3d) do not support the BCM origin of these nanocrystals. (We note that the carbonate-hosted magnetite shown in Fig. 4 is just one of several types of magnetite/carbonate associations that were described by Thomas-Keprta et al. (2009) from ALH84001.)
It should be remembered that different techniques may produce different size distributions, e.g. X-ray diffraction measures crystallite size, whereas TEM measures particle size, where a single particle may contain many crystallites. It is also noteworthy that particle-size distributions obtained from projected outlines in TEM images are affected by the particle shapes and orientations. For example, even if all of the particles in a sample are uniform regular bodies (suchascubesoroctahedra) that are randomly oriented, they will produce projected shapes that have a variety of diameters within a range that is defined by their longest and shortest physical dimensions (Fig. 5a). The result will be a broadening of the measured size distribution. In contrast, the preferred orientation of particles can lead to a distribution that is characteristic of only a single projection direction. The issue of shape effects on projected particle sizes makes it desirable to use three-dimensional data for particle distributions, leading to the problem of determining the morphologies of nanoscale biomineral grains that is tackled in the following subsection.
The crystallographic habits of BCM minerals are important for some of their functions. For example, apatite platelets, a few nm thick, constitute the inorganic fraction of bones (Fratzl & Gupta, 2007; Pasteris et al., 2008), as randomly shaped crystals would presumably not fit into the spaces between collagen fibres and would destroy the unique and mechanically ideal combination of the organic/inorganic composite material. Similarly, the strength and toughness of some mollusc shells are ensured by the presence of prismatic calcite and plate-like aragonite crystals, respectively (Dalbeck et al., 2006; Espinosa et al., 2009). In magnetotactic bacteria, the elongated habits of some magneto-somes constrain the magnetic field of the ferrimagnetic crystal to be parallel to its elongation axis, and thus impart an overall magnetic moment to the cell in a desired direction (Simpson et al., 2005). A knowledge of the shapes of biominerals is therefore of great importance for understanding many of their functions.
The significance of knowing the shapes of nanoscale biominerals has been highlighted by a lengthy debate over whether or not magnetite nanocrystals observed in the Martian meteorite ALH84001 formed inside the cells of magnetotactic bacteria. In the original publication (McKay et al., 1996), all of the magnetite nanocrystals within carbonate globules in the meteorite were assumed to be biogenic. In later studies (Thomas-Keprta et al., 2000, 2001, 2009), the same authors argued that most of the magnetite crystals had formed inorganically, but that a fraction of them were from magnetotactic bacteria. Their argument for the biogenic origin of the particles was based primarily on the distinct, elongated hexagonal habits of some of the magnetite particles, which resemble the morphologies of nanocrystals formed in the terrestrial magnetotactic bacterial strain MV-1. While the morphological evidence was challenged in several studies (e.g. Buseck et al., 2001; Golden et al., 2004), and most researchers now agree that the meteorite does not hold any remains of ancient life, the nature of the claims is such that the bacterial origin of the magnetite nanoparticles has not been, and is unlikely to ever be, disproved unequivocally. It will always be possible to pinpoint a certain fraction of particles that have shapes similar to those in some strains of terrestrial bacteria. Nevertheless, the case of ALH84001 has stimulated much research aimed at developing techniques for the precise determination of nano-crystal morphologies.
As discussed in the previous section, biominerals are typically nanocrystalline. Therefore, the analysis of their shapes is inherently very challenging. Even though the availability of modern field-emission-gun high-resolution scanning electron microscopes (SEMs) now permits the identification of the morphologies of free-standing particles that are as small as a few tens of nm more easily than was possible 15–20 years ago, biominerals are typically embedded in biological tissue, rendering SEM imaging inappropriate for morphological analysis. For such materials, only TEM can be used to provide data from which individual nanocrystal morphologies can be deduced.
Two-dimensional projections of nanocrystals can be imaged routinely in the TEM and used for the approximate evaluation of particle morphologies. However, the projected shapes of crystals can be deceptive, and the intuitive interpretation of such images can lead to errors. For example, a regular octahedron can have a variety of projected outlines, including a square and a regular hexagon (Fig. 5a). Without obtaining information about the thickness profile of each particle (or if such contrast is not clearly interpretable, as in Fig. 5b), it is difficult to know whether a square-shaped outline represents a -type projection of an octahedron, a cube or a different object. The chain of magnetite magnetosomes shown in Figure 5b can be interpreted as consisting of octahedra that have a range of different sizes and largely random orientations, with the exception of a chain of six particles (boxed in the image), each of which appears to be in a -type orientation. The interpretation presented in Figure 5c is only one of several reasonable possibilities, in this case, inferred from prior knowledge about the likely crystallographic structure and morphology of the particles.
For a more unambiguous identification of nanocrystal morphologies, it is necessary to tilt the specimen in order to obtain images along several projection directions. A series of bright-field TEM images of a chain of pointed, elongated magnetite crystals from a magnetotactic bacterial cell is shown in Figure 6a. The specimen was tilted around an axis approximately parallel to the white arrow, with the tilt angle indicated at the bottom of each panel. At each specimen tilt angle, the crystals exhibit approximately triangular shaped projections at their lower ends, suggesting the presence of straight bounding faces. The upper ends of the crystals (with the exception of the oppositely oriented particle at the bottom) appear to be more highly pointed, with curved bounding contours. Whereas the projected shapes of the more pointed ends do not change significantly with specimen tilt angle, the projected bounding faces of the triangular base vary for each crystal. By using the information from the series of images, and by taking into account constraints on the morphology of each crystal resulting from the known point group of magnetite, the morphologies of the crystals could be interpreted and modelled (Fig. 6b,c). The nanocrystals are inferred to consist of two distinct parts: a base that can be represented by half of an octahedron, and an elongated, pointed part that can only be modelled by a large number of high-index faces (Lefèvre et al., 2011). In reality, this part of the crystal is probably not bounded strictly by crystallographic faces, but takes the form of an irregular, curved surface.
Developments in both hardware and software over the past 15 years have made it possible to use electron tomography (ET) for the direct identification of nanocrystal morphologies from large numbers of images acquired as a function of specimen tilt angle, followed by three-dimensional reconstruction and visualization (Midgley & Weyland, 2011). Ideally, in order to reconstruct the morphology of a crystal completely, images need to be acquired over an entire specimen tilt range of 180°. Modern TEM specimen holders that are specifically designed for tomographic experiments can now be tilted routinely to + 80°, with images typically obtained at 2° tilt intervals. A key requirement for acquiring high-quality datasets is the use of either periodic manual realignment or precise computer control of the specimen stage, ensuring that the object of interest remains in the field of view and in focus throughout the experiment. As the acquisition of ∼80 or more images is time consuming, the stability of the specimen under prolonged electron irradiation is also important (Buseck et al., 2001; Kasama et al., 2006).
The contrast mechanism that is used to acquire images for the 3D reconstruction of crystal morphologies depends on the chosen imaging technique. If the specimen is amorphous and consists of a material of uniform density, then the contrast in bright-field (BF) TEM images is related directly to the thickness of the specimen and is suitable for reconstruction of 3D morphology. For example, cell walls and internal cellular components can be studied in this way (see Section 4.1 below). However, crystalline materials can exhibit strong diffraction contrast in BF TEM images, which are then no longer dominated by variations in specimen thickness and density. A solution to this problem is provided by the acquisition of tilt series of images using scanning transmission electron microscopy (STEM). As the electron beam is scanned over the specimen, a high-angle annular dark-field (HAADF) detector is used to collect electrons that have been scattered to angles beyond those typical for Bragg diffraction. The contrast in HAADF images is approximately proportional to the square of the mean atomic number (the technique is referred to as ‘Z-contrast imaging’) and to specimen thickness. Based on the variation of contrast in a series of HAADF images, the 3D shape of a crystal can be reconstructed and crystallographic faces can be identified on crystals that are as small as a few nm in size (Weyland et al., 2001).
HAADF ET has been used for the characterization of the morphologies of magnetite crystals from magnetotactic bacteria (Fig. 7) (Buseck et al., 2001). As the presence of biominerals within an organic matrix is usually associated with a significant change in local density, HAADF image contrast recorded from a composite organic/inorganic object is not only related to specimen thickness. In such cases, when the 3D morphology is reconstructed, the intensity threshold between the mineral and the organic surroundings needs to be chosen carefully (Fig. 8) (especially if the range of accessible specimen tilt angles is limited). It should also be noted that artefacts from non-linear diffraction effects in HAADF images can hinder the precise reconstruction of 3D nanocrystal morphologies (Van den Broek et al., 2012). Nevertheless, the spatial resolution of ET based on STEM HAADF imaging is superior to that of any other technique that has been used for obtaining the morphologies of crystalline biomineral nanoparticles, and is suitable for rendering the 3D relationships between inorganic minerals and cell boundaries, as illustrated in Figure 8 for a dividing magnetotactic bacterial cell that contains both greigite and magnetite crystals. Ongoing progress in ET includes the development of 360° ET using special on-axis TEM specimen holders and needle-shaped samples (Kawase et al., 2007), the use of new algorithms for the processing of experimental datasets (Batenburg et al., 2009), and the demonstration of the feasibility of atomic-resolution ET (Van Aert et al., 2011). However, these cutting-edge techniques have yet to be applied to biominerals.
In contrast to minerals that formed by inorganic, natural processes to produce solid solutions and phases with elemental substitutions, biominerals typically have strictly controlled compositions, especially when they result from BCM processes. It has been suggested that the chemical purity of some biominerals, for example magnetite, can be used as a biomarker (Kopp & Kirschvink, 2008; Thomas-Keprta et al., 2002). Another important consideration is the fact that the physical properties of biominerals may depend on their chemistry. As is well known from mineralogy textbooks, traces of foreign elements can change the colour of a mineral. Solubility can also be affected strongly by composition. For example, the solubility of apatite decreases markedly if F− substitutes for OH−, with consequences for the stability of teeth and bones (Pasteris et al., 2008). The compositions of some biominerals are highly sensitive to environmental changes or seasonal variations in water chemistry that occur during the life of the biomineral-producing organism (e.g. carbonate shells in bivalves (Schöll-Barna et al., 2012), or concentric aragonite layers in fish otoliths (Allemand et al., 2007)). In this way, the compositions of biominerals can provide information about their functions and origins, as well as about the environment of the organism.
Although analytical TEM techniques have been used for decades, they are continually being refined, with some methods recently reaching spatial resolutions and chemical sensitivities that are sufficient for identifying individual atomic columns. An important prerequisite for meaningful analysis, which is also a serious limitation for any analytical technique that involves the use of high-energy electrons, is that the specimen should not decompose or change its composition during electron irradiation (Garvie et al., 2004).
The standard method of choice for elemental analysis in the TEM is energy-dispersive X-ray (EDX) spectroscopy. Since EDX spectroscopy has been an established and relatively user-friendly technique for decades (Williams & Carter, 1996), and its application to the determination of biomineral compositions does not pose any different challenges from the analysis of any other mineral specimen, the details of the aquisition, interpretation and quantification of EDX spectra will not be discussed here (but are covered in another chapter in this volume (Leroux, 2013, this volume)). It should be noted, however, that EDX spectra can contain many artefacts, and the analysis of elements lighter than Na is, in many cases, regarded as being only semi-quantitative as a result of strong absorption of relatively low-energy X-rays within the specimen.
If one is interested in the organic, cellular surroundings of BCM minerals that are composed of C, O and other light elements, then electron energy-loss spectroscopy (EELS) is usually a better choice than EDX spectroscopy. The signals that are acquired using EDX spectroscopy and EELS are of a complementary nature, as they originate from the same physical process, i.e. the excitation of electrons in the atoms of the specimen by the electron beam. While EDX spectroscopy detects X-rays that have been emitted by the excited atoms when their electrons return to their ground states, EELS detects the energy losses that the exciting beam suffers when it interacts with the specimen. Both EDX and EELS measurements can be recorded in the form of point spectra, line scans or maps, depending on the particular problem.
Electrons that have lost energy in inelastic interactions with atoms in a specimen can be used for both spectroscopy and imaging using EELS (e.g. Ferraris & Auchterlonie, 2013, this volume). A major requirement is that specimens that are thicker than ∼100 nm cannot normally be analysed quantitatively using EELS (Williams & Carter, 1996). In addition to providing quantitative information about the composition of the specimen, EEL spectra contain a wealth of other information (e.g. Golla-Schindler & van Aken, 2010). The fine structures of high-energy core-loss edges (E > ∼50 eV) are sensitive to the local bonding environment and can be used to obtain information about the coordination number and oxidation state of the element. For example, several iron oxide biominerals have compositions that are sufficiently close to one another that their chemical identification is difficult even using quantitative EELS analysis. In this case, the near-edge structures of both the Fe L and the O K edges (up to ∼30 eV above the edge) can be used as fingerprints for the identification of the mineral phase (Fig. 9a). By using such an approach, the fossilization process of a microorganism has been studied by distinguishing carbon within calcite from amorphous (probably organic) material using the near-edge structure of the C peak in EEL spectra (Benzerara et al., 2005; Fig. 9b,c). In addition to information contained in core-loss edges, the low-loss regions of EEL spectra (E < ∼50 eV) can be used for obtaining the relative thickness of the specimen (Egerton, 1996).
By selecting only electrons that have lost a specific amount of energy in the specimen, energy-filtered images can be obtained using a post-column or in-column imaging spectrometer. The most common forms of energy-filtered images are ‘three-window elemental maps’, which provide image contrast that is related to the projected concentration of a selected element. Such elemental maps are each constructed from three images. A post-edge image is generated from electrons that have lost energy in a preselected range (usually up to 30 eV) beyond the edge of the element of interest, and two pre-edge images are formed from electrons that have lost different amounts of energy below the edge. After alignment of the three images, a power-law background is calculated at the position of each pixel based on the two pre-edge images and subtracted from the post-edge image to create an element-specific map (Fig. 10a). By combining several such elemental distribution maps, composite colour-coded images can be produced in which different compositional variations can be visualized (Fig. 10b). Such composite images are particularly useful for the analysis of biominerals in a light-element-containing organic environment. As the spatial resolution of EELS can approach the atomic scale, it can be used for mapping intracrystalline compositional heterogeneities in biomineral nanoparticles. This mapping can reveal subtle changes in composition that can either reflect varying conditions during crystal growth or result from alterations in the solid state (Fig. 11).
Recent developments in STEM EELS allow the identification of single atoms (Varela et al., 2004), and even the valence states of transition metals in two dimensions (e.g. Tan et al., 2011). STEM EELS analysis of biominerals is often complicated by the effects of contamination arising from the electron-beam exposure of the associated organic materials. Low-voltage (60 kV) STEM is promising for the analysis of single atoms in beam-sensitive materials (Suenaga et al., 2009). Close-to-atomic-resolution STEM EELS has been used to study the structure and morphology of ferrihydrite cores of fer-ritin, the Fe-storage protein that is ubiquitous in many organisms (Pan et al., 2009). Based on EEL spectra, the absolute number of Fe atoms could be estimated within 2 nm subunits of the ferritin core and, in combination with HAADF images, the arrangements of ferrihydrite subunits within the cavity of the protein could be determined. However, to the authors’ knowledge, truly atomic-resolution EELS has not yet been applied to biominerals.
Organisms are often able to select which polymorph of a chemical compound they crystallize, as illustrated by the simultaneous presence of calcite and aragonite shells in the same aqueous environment. In some cases, even though only one polymorph is stable under a given set of thermodynamic variables, an organism can contain two different polymorphs of the same substance (Baeuerlein, 2007; Mann, 2001; Bentov et al., 2010). The different mechanical properties of the distinct polymorphs may then both be utilized by the organism (e.g. in mollusc shells). In other instances, the simultaneous presence of two polymorphs may result from the sluggish transformation of a less stable, but kinetically more easily crystallizable variant, into a more stable one (Gower, 2008). Calcium carbonate is the most prominent mineral that can produce a variety of structures (Rodriguez-Navarro & Ruiz-Agudo, 2013, this volume). Marine mussel shells may contain aragonite in their nacreous layers and calcite in their prismatic layers (Dalbeck et al., 2006), while sea urchins contain amorphous calcium carbonate together with calcite in their spines (Politi et al., 2004). Some magnetotactic bacteria contain both magnetic and non-magnetic Fe sulfide nanoparticles in their cells (greigite and mackinawite, respectively) (Pósfai et al., 1998a). In order to understand the functions of biominerals in organisms, knowledge of their structure is as important as knowledge of their composition. Some biominerals (e.g. Fe oxides) have very similar compositions, which are difficult to identify using an analytical technique such as EDX, especially if they are embedded in a biological material that also contains oxygen. In such cases, identification of their structure can serve as an indirect way of determining their composition.
The crystallographic structures of biominerals are commonly identified using electron diffraction. If the specimen is beam sensitive, then the application of a strong, focused electron beam is to be avoided and selected-area electron diffraction (SAED) can be used. Provided that many particles of the same mineral occur within the field of view, electron diffraction ring patterns can be used for phase identification. In most cases, however, the structures and crystallographic orientations of individual crystallites are of interest. The acquisition of interpretable SAED patterns from biogenic nanocrys-tals can be challenging if several particles occur within the region defined by the smallest selected-area aperture and contribute reflections to the diffraction pattern (Fig. 12). In general, SAED patterns are useful only when they are taken with a major zone axis of the crystal close to the incident electron beam direction. A possible procedure for obtaining such patterns involves tilting a chosen crystal while observing its contrast changing in a bright-field TEM image, with darker contrast corresponding to strongly diffracting (e.g. major zone axis) orientations. By switching to diffraction mode periodically, the effect of tilting on the diffraction pattern can be checked. In addition to structure identification, SAED patterns can also be used to obtain the crystallographic orientations of nanocrystals (Fig. 6b), as has been done in a large number of studies of textured biomineral systems such as mollusc shells (e.g. Suzuki et al., 2010), coccoliths (Mann & Sparks, 1988) and corals (e.g. Vielzeuf et al., 2008). In contrast to SAED, convergent-beam electron diffraction (CBED) has found limited use in studies of biomineralization products (Saruwatari et al., 2008), even though CBED is useful for identifying point and space-group symmetries and crystallographic orientations (Steeds & Morniroli, 1992). The sporadic use of CBED in biomineral studies is probably related to the tendency of many biominerals to suffer structural damage in a focused electron beam, as well as to the nanoscale sizes (and therefore thicknesses) of many biomineral crystals.
For imaging atomic-scale structural details, the technique of high-resolution TEM (HRTEM) is typically used. Contrast in HRTEM images arises from the interference of the electron wave with itself as it passes through the specimen (resulting in ‘phase-contrast imaging’). Contrast in HRTEM images of crystalline specimens is not necessarily intuitively interpretable, as it is usually affected strongly by dynamical diffraction (unless the specimen is extremely thin, in most practical cases less than a few nm), as well as by microscope parameters. For a more detailed treatment of HRTEM, the reader is referred to standard electron microscopy textbooks (e.g. Williams & Carter, 1996). In spite of these difficulties, HRTEM images are often ideally suited for providing information about structural defects, which may affect the mechanical or other physical properties of a mineral, as well as often being associated with local chemical variations (Fig. 13). Even in the case of a crystal that has a perfect crystallographic structure, variations in image contrast can be used to estimate the local specimen thickness, while atomic-scale detail along the edges of a nanoparticle can offer insight into the mechanism of crystal growth (Fig. 14).
In some cases, biomineral crystals are so small or so poorly crystalline that structural identification is performed only from HRTEM images, instead of using SAED patterns. However, the distances between lattice fringes in HRTEM images often cannot be measured accurately or precisely enough for phase identification. Typically, in order to obtain the charateristic d spacings in a crystal, a Fourier transform of an HRTEM image is calculated and used to measure reciprocal lattice spacings (Figs 13–15). However, such measurements can be inaccurate if the diffracting volume (i.e. the particle) is smaller than a few nm in size or not perfectly aligned to a zone-axis orientation (Malm & O’Keefe, 1997). ‘Delocalization’, i.e. the spreading of lattice fringes beyond the boundaries of a crystal, is also inherently present in images obtained using a conventional (i.e. non-aberration-corrected) TEM (Cervera Gontard et al., 2006).
Revolutionary developments in aberration correction for TEM, which can include either or both spherical aberration (Cs) and chromatic aberration (Cc) correction, offer a solution to some of the above imaging artefacts. Although Cs correction does not eliminate delocalization, it allows the use of a single defocus at which delocalization is minimized for all spatial frequencies in an image, resulting in more interpretable image contrast (Fig. 15). The combination of Cc/Cs-corrected TEM using accelerating voltages as low as 20 kV has been demonstrated to be very useful for studying materials that suffer structural damage in the electron beam (Kaiser et al., 2011); however, the specimen then needs to be very thin (<∼4 nm). As the structures of many important groups of biominerals, including silica, phosphates and carbonates (e.g. Rodriguez-Navarro & Ruiz-Agudo, 2013, this volume), are typically very difficult to study using conventional TEM, aberration-corrected TEM is likely to find widespread use in bio-mineral characterization.
‘Electron crystallography’ is another rapidly developing field in TEM (Dorset, 1996). Compared to X-ray crystallography, the main advantages of using electrons for studying atomic-scale structural details are that nm-scale crystals can be studied, and that the phases of crystallographic structure factors can be recovered. Until recently, the applicability of electron crystallography had been limited by dynamical diffraction, which affects diffracted intensities and makes structure determination difficult. However, in the last ten years novel techniques have been developed for both data acquisition and interpretation. Specifically, precession diffraction allows the collection of diffraction data sets that are less affected by dynamic scattering (Vincent & Midgley, 1994), while HRTEM images have also been used directly to obtain the phases of crystallo-graphic structure factors (Zou & Hovmøller, 2008). On the theoretical side, new algorithms have been developed for solving structures using direct methods (e.g. Oszlányi & Sutö, 2008). Among the various data-acquisition techniques for electron crystallography, automated diffraction tomography (ADT) in a STEM is useful for collecting structural data from nanoscale single crystals using low electron doses, thereby minimizing electron-beam damage (Kolb et al., 2007, 2008). ADT has been used for the structural analysis of sponge spicules, with 3D diffraction data collected from nanorods that were only a few nm long (Mugnaioli et al., 2009), showing that precursor phases of the spi-cules have cell parameters that are consistent with a smectite-like structure. The technique has also been used for the ab initio structure determination of vaterite (Mugnaioli et al., 2012), an enigmatic polymorph of CaCO3 that occurs mostly as a biomineral.
The anisotropy of the physical properties of crystals means that BCM minerals typically need to be aligned along particular crystallographic directions in order to fulfill their roles in organisms. It has long been known that the crystallographic directions of prismatic calcite and tablet-shaped aragonite crystals are strictly controlled in sea shells (Weiner & Addadi, 1997), as optimal mechanical strength can be achieved by the ordered arrangement of the calcium carbonate polymorphs. Trilobites (characteristic arthropodes of the Paleozoic) had compound eyes in which the lenses were made of calcite (Clarkson et al., 2006). Because of the strong optical birefringence of calcite, the only direction along which a trilobite would not have seen a double image is along the three-fold axis. All of the calcite crystals in a trilobite eye were aligned with their c axes parallel to one another and towards the light (Clarkson et al., 2006). There are many other examples of strong effects of crystal orientation on magnetic, electric, optical or other physical properties of biominerals.
As discussed in the previous section on crystallographic structure, SAED or CBED patterns and HRTEM images can be used to determine the crystallographic orientation of a mineral grain (e.g. Benzerara et al., 2011a). In the example shown in Figure 16, the orientations of individual magnetite nanocrystals in a double chain of magnetosomes were determined using SAED. One by one, each numbered crystal was tilted to a -type or -type zone-axis orientation, and the rotation angles about the x and y axes of the specimen stage were recorded (shown for crystals 5 and 7 in Fig. 16b). From these values, the relative orientations of the nanocrystals were determined, as shown in the form of a stereogram in Figure 16c. The result indicates that, with the exception of the small crystal numbered 1 at the end of the chain, all of the magneto-somes are aligned with their  axes parallel to one another, and to the chain axis, to within a few degrees. However, the -type axes of the crystals, which are perpendicular to , are not correlated with each other. The chain of magnetosomes can therefore be regarded as ‘beads on a string’, in which the individual beads have a fixed direction along the string, but are free to rotate about it.
This ‘traditional’ approach of obtaining SAED patterns individually from successive nanocrystals is labour-intensive and inefficient for determining crystallographic orientations in a large nanocrystal population. Fortunately, in the last few years, a new, semi-automated technique has emerged that is based on precession electron diffraction (Rauch et al., 2010; Moeck et al., 2011). This technique does not require the individual alignment of each nanocrystal to obtain its orientation. Instead, the electron beam is precessed and scanned over the specimen and electron diffraction patterns are acquired from every crystal while the sample is kept in a single orientation. Then, mapping software is used to identify the relative orientation of each nanocrystal (assuming that the crystal structure is known). Intracrystal changes in orientation can be obtained, as the spatial resolution of the measurements is limited only by the beam diameter. The orientations of prism-shaped, -elongated magnetite crystals in a single chain have been determined in this way (Fig. 17). One of the -type axes in each crystal was found to lie parallel to the chain axis to within a few degrees, whereas the  directions perpendicular to the chain appeared to have two distinct directions. In contrast to the example shown in Figure 16, these results suggest that even the directions that are perpendicular to the chain axis in magneto-somes can be non-randomly distributed. As the hardware that has been developed for controlling the precession and scanning of the electron beam can be attached to practically any TEM (Moeck et al., 2011), the technique of precession electron diffraction mapping can be used conveniently for obtaining nanoscale orientation/structure maps from a wide range of biominerals.
Magnetic minerals occur in many organisms, ranging from bacteria to humans (Pósfai & Dunin-Borkowski, 2009). In a few cases, such as in a specific marine mollusc (Saunders et al., 2009) and the scaly-foot snail (Suzuki et al., 2006), the magnetic mineral (magnetite or greigite) is likely to be used primarily for hardening or protection. In other cases, however, the magnetic crystals are either known or suspected to be involved in magnetic navigation (Kirschvink et al., 2001). As the magnetic moments and switching behaviours of ferrimagnetic minerals depend strongly on the sizes, shapes, chemical compositions, orientations, separations and arrangements of the crystals, TEM techniques that can be used to visualize and quantify magnetic fields within and around nanoscale bio-minerals are extremely useful for understanding the functions of biomin-eral magnetism.
Several TEM techniques were developed in the second half of the 20th century for visualizing information related to magnetic fields in materials. When characterizing magnetic materials, the microscope objective lens is usually turned off because it creates a large (>2 T) magnetic field at the position of the specimen. Instead, a non-immersion Lorentz lens can be used as the imaging lens. This lens allows specimens to be imaged at relatively high magnification in magnetic-field-free conditions. The more traditional techniques of Fresnel (out-of focus) and Foucault (magnetic dark-field) imaging have been used primarily for imaging magnetic domain walls and magnetic domains in thin magnetic films (Chapman, 1984; Williams & Carter, 1996; Kasama et al., 2009).
The technique that is best suited for studying magnetic fields in, >100 nm biomineral crystals is off-axis electron holography (EH), which requires the use of a field-emission electron gun to illuminate the specimen with a coherent electron beam. An electron biprism is used to overlap part of the incident electron wave that passed through the specimen with another part that passed only through vacuum (Fig. 18) or through a clean region of support film of uniform thickness to produce an interference pattern (an electron hologram). The recorded electron holographic interference fringe pattern can be used to recover the amplitude and phase shift of the electron wave that passed through the specimen (Dunin-Borkowski et al. 2004). The phase shift is sensitive to both the mean inner (electrostatic) potential (MIP) of the specimen V0 and the component of the magnetic induction B within and around the specimen that is perpendicular to the incident electron beam direction.
In Equations 1-3, E and E0 are the kinetic and rest-mass energies of the incident electrons, respectively, λ is the (relativistic) electron wavelength, and z and x are directions parallel and perpendicular to the incident electron beam, respectively. The difference between any two points in a magnetic phase image (i.e. in the magnetic contribution to the recorded phase shift) is a measure of the magnetic flux through the region of space bounded by two electron trajectories crossing the sample at the positions of these two points. A phase difference of 2π corresponds to an enclosed magnetic flux of 4.14 × 1015 Tm2.
It should be noted that the separation of electrostatic and magnetic contributions to the phase is almost always mandatory in order to obtain quantitative magnetic information from a phase image. In principle, the most accurate way of achieving the removal of the mean inner potential contribution to the recorded phase, which can be much greater than the magnetic contribution for small (>50nm) magnetic nanocrystals, involves turning the specimen over after acquiring a hologram and then acquiring a second hologram from the same region. The sum and difference of the holograms are then used to determine twice the mean inner potential and twice the magnetic contribution to the phase, respectively. An alternative method, which is often more practical, involves performing a magnetization reversal experiment in the microscope and then selecting pairs of holograms that differ only in the magnetization direction in the specimen. If the magnetization in the specimen does not reverse perfectly, then such reversal measurements may need to be repeated many times so that non-systematic differences between reversed images are averaged out (Dunin-Borkowski et al., 2001). Contours can then be generated from the final magnetic phase image to produce a visual map of the in-plane component of the projected magnetic induction (Fig. 19).
The first EH experiments to study magnetic microstructure in biominerals were performed on magnetotactic bacteria (Dunin-Borkowski et al., 1998; Dunin-Borkowski et al., 2001; McCartney et al., 2001). The magnetic fields in individual magnetite crystals could be correlated with their positions, separations and morphologies. Figure 19 shows the strength and direction of the projected in-plane magnetic induction displayed using colours and contours for two different distributions of magnetite crytals. Although a full interpretation of such magnetic induction maps requires comparisons with simulations, they provide semi-quantitative information about the magnetic state of each crystal. In Figure 19, each magnetite magnetosome is observed to contain a single magnetic domain, with an overall direction of the magnetic moment that is constrained to be parallel to the elongation axis of each prismatic-shaped crystal as a result of the strong effect of crystal shape on magnetization direction.
Figure 19a shows partial chains and scattered crystals that have variable directions of their magnetic moments. In contrast, Figure 19b shows two well ordered pairs of chains of magnetite magnetosomes in a different cell, in which the magnetization directions of the crystals are closely similar to each other and parallel to the chain axes. In such chains, small (>20 nm) crystals, which would be expected to be superparamagnetic if they were isolated, are often found to contain stable single magnetic domains as a result of interactions with adjacent crystals. The physical properties and arrangements of such crystals influence their shape and magnetocrystalline anisotropies to produce an overall magnetic moment that is optimal for magnetotaxis. Specifically: (1) the sizes of most crystals are within the single magnetic domain range for magnetite; (2) the crystals are elongated parallel to the chain axis; (3) the  magnetic easy axis is parallel to the chain in each crystal; (4) because of magnetostatic interactions between neighbouring crystals, both small (>20 nm) and large (<100 nm) crystals can have stable single domain magnetic configurations.
A similar ordered chain of magnetite crystals from a bacterial cell is shown in Figure 20. The BF TEM image shown in Figure 20a suggests that each particle has a bullet-like shape, which is confirmed by the three-dimensional reconstruction of their morphologies using HAADF STEM electron tomography, shown in Figure 20b. The crystallographic direction of elongation of each crystal was determined to be  from SAED patterns and HRTEM images. Thus, shape and magnetocrystalline anisotropy combine to constrain and enhance the magnetic moment of each crystal parallel to the chain axis, to result in the magnetic induction map measured using EH shown in Figure 20c. Individual crystals that are displaced relative to their neighbours in the chain cause kinks in the measured magnetic field direction.
In addition to the information about magnetic fields that can be discerned visually from induction maps, EH can also be used to provide quantitative measurements of magnetic properties such as the coercivity or magnetic moment of a bacterial cell or an individual crystal. The magnetic moment of an individual magnetosome or an entire chain of crystals can be measured from the magnetic contribution to the phase shift (Dunin-Borkowski et al., 2001; Kasama et al., 2006; Simpson et al., 2005; Beleggia et al., 2010). By applying this approach to various types of magnetotactic bacteria, including both magnetite-bearing and greigite-bearing cells, it was found that the magnetic moment per unit length is similar for all chains (Pósfai et al., 2007). This remarkable uniformity of the magnetic moment ensures that all types of magneto-tactic bacteria can navigate efficiently in the weak magnetic field of the Earth. Cells that contain highly ordered chains of magnetic crystals with strictly regulated shapes and orientations (such as those shown in Figs 19b and 16) contain fewer magnetosomes than cells that have more disordered arrangements of magnetosomes (such as those shown in Figs 8 and 10), in which a larger number of magnetosomes compensates for the loss of magnetic moment that arises from the less controlled particle shapes and orientations.
Recently, another technique, energy-loss magnetic chiral dichroism (EMCD), has emerged for measuring the magnetic properties of nanoscale particles using TEM (Schattschneider et al., 2006). As the name implies, the technique is based on EELS, and involves the collection of two EEL spectra, e.g. at the Fe L2,3 ionization edge. As the phase of the transmitted electron beam is angle-dependent, experimental conditions can be set up in such a way that the two electron beams that are used for collecting the two EEL spectra have phases with opposite signs in the diffraction plane. The dichroic signal then corresponds to the difference between the two spectra, and is proportional to the magnetic moment along the electron beam direction. EMCD and EH thus measure the magnetic moment in mutually perpendicular directions. While crystals need to be in specific orientations for EMCD, there is no such experimental constraint for EH. EMCD has been used for studying the magnetization of magnetite particles in a magnetotactic bacterium (Stöger-Pollach et al., 2011).
The mineral–organic interface – issues with sample preparation and imaging
Many scientific questions about the formation, transformation and functions of biomin-erals can be answered optimally by applying TEM to the minerals in their native environment, i.e. within their surrounding organic material. However, the preparation and examination of specimens in this form is difficult because of problems associated with (1) sample preservation during specimen preparation, (2) sample stability in the TEM, and (3) low contrast in images from biological materials. The key issue for the preservation of biological specimens in their original state for TEM studies is to prevent the cells from dehydrating, collapsing and rearranging their contents. Various techniques have evolved over the last few decades to address this problem. It is not possible to cover these techniques exhaustively in this chapter. Instead, after a short, general description of the basis of conventional biological TEM specimen preparation, some recent developments are discussed that have significantly extended the preparation of specimens in which the interface between soft (organic) and hard (mineral) components in organisms is of interest. With regard to sample stability in the TEM, both electron-beam irradiation and the vacuum in the electron microscope can result in damage. The effect of the vacuum can be reduced by using low-vacuum conditions in recently developed environmental TEMs or closed-cell specimen holders, while the effect of the electron beam can be reduced by using aberration-corrected instruments at low accelerating voltage. These new types of electron microscopes are also useful for obtaining images of thick biological specimens with improved contrast and resolution, notably if their use is combined with the application of special imaging techniques, such as the use of Cc correction or phase plates (see below).
Specimen preparation for TEM
The preparation of biological TEM specimens is a scientific discipline in its own right, with an extensive literature (e.g. Glauert & Lewis, 1998). The conventional procedure for specimen preparation typically consists of several steps, including fixation, dehydration, staining for contrast enhancement, embedding, and sectioning to provide thin, electron-transparent slices of the sample. Fixation of a biological sample is typically achieved using a 2.5% glutaraldehyde solution in a buffer. A second fixation step using an osmium tetroxide solution (in buffer) may follow. Then, optionally, a heavy metal salt solution (e.g. uranyl acetate) may be used for staining cellular components, in order to enhance contrast in BF TEM images. Dehydration is achieved by treating the specimen with ethanol, in several steps, at increasing concentrations. The specimen is then infiltrated with resin, again in several steps, and finally embedded in 100% resin. After the resin has polymerized, thin sections can be cut using an ultramicrotome.
Although this conventional specimen-preparation technique has been perfected for various types of samples (such as animal or plant tissue or single cells), and has provided many important measurements over several decades, it has several drawbacks. First, the procedure is lengthy and complicated, and each of the preparation steps can cause artefacts. Second, poorly fixed specimens disintegrate easily during cutting or in the electron beam. Third, ultramicrotomy of specimens that have uneven hardness can cause the harder parts to be pushed aside by the knife and to become under-represented in the thin section.
The BF TEM image shown in Figure 21 illustrates some of the problems that can result from conventional biological TEM specimen preparation, for a sample that was prepared from cells of a magnetotactic coccus with intracellular magnetite crystals that were arranged in partial chains or scattered throughout the cell. Ultrathin sections were prepared according to the procedure described above, with the goal of visualizing the locations of the magnetosomes relative to the cell wall and possible intracellular organic structures. Some cellular features are visible in Figure 21a, including a cell wall structure and stained material around the magnetite crystals adjacent to the cell wall, suggesting the presence of a magnetosome membrane. On the other hand, the waviness of the cell contour and the empty space in the inner part of the cell (also seen in the BF TEM image shown in Figure 21b) indicate that the fixation step was not entirely successful, and that some of the information about the cell components may have been lost. The energy-filtered elemental maps shown in Figure 21b indicate that iron is concentrated only in the magnetosomes. BF TEM images from untreated, dehydrated cells showed a large number of magnetite crystals in every cell, whereas most of the ultrathin cell sections contained no crystals at all, suggesting that cutting with the ultramicrotome was selective.
The introduction of cryo-fixation (Dubochet et al. 1988) was a major step towards improving biological TEM specimen preparation and advancing ultrastructural studies. This technique is based on the ultra-rapid freezing of a sample to prevent ice crystal formation. When the specimen is cooled at a rate of ∼10000 K/s, the water content freezes to an amorphous state, without ice crystals growing and without destroying the delicate cell structure. In practice, the fast cooling rate is achieved by plunging the specimen into liquid ethane, which is cooled to close to its melting point using liquid nitrogen (Ryan, 1992). The plunge freezing technique is a relatively simple and artefact-free alternative to the complicated procedure of conventional fixation, and even a portable cryo-plunger exists for environmental cryo-sampling (Comolli et al. 2012). However, only thin (∼10 to 500 nm) films can be prepared and imaged in this way, unless high-pressure freezing is used (McDonald, 2009; Miot et al. 2011), which allows the preparation of samples up to ∼200 μm thick. The sample needs to be kept at liquid nitrogen temperature at all times, and the vitreous ice in the specimen is sensitive to radiation damage, requiring low-dose imaging (Kourkoutis et al. 2012).
Cryo-TEM has been extremely successful in revealing the ultrastructures of single-celled organisms, including those of magnetotactic bacteria. The BF TEM image shown in Figure 22a was obtained from a magnetotactic bacterial cell that was prepared by cryo-fixation (Katzmann et al. 2010). Compared to the cell shown in Figure 21, this cell looks more intact, with smooth contours and a clearly visible double cell wall. A chain of magnetosomes, and a few larger amorphous inclusions, are also visible in the image. The inset suggests that a new, empty vesicle (marked by a white arrow) has formed by invagination of the cytoplasmic membrane (marked by black arrows). The advantages of using cryo-fixation are enhanced further if examination of the specimen can be combined with electron tomography, (i.e. cryo-electron tomography). Figure 22b shows a three-dimensional reconstruction of the ultrastructures in a magnetotactic bacterial cell, revealing the positions of vesicles enclosed by magnetosome membranes (yellow), magnetite nanocrystals inside some of the vesicles (red) and a filamentous structure (green), to which the magnetosomes are attached along the length of the cell (Katzmann et al., 2010; Komeili, 2007; Scheffel et al., 2006).
Problems related to the precise sectioning of samples have been partially solved by the use of focused ion beam (FIB) milling, which was introduced for materials-science applications in the 1990s, initially for studies of semiconductors. The applicability of FIB milling for preparing mineral–microbe interfaces has been realized within the last decade (Benzerara et al., 2005; Obst et al., 2005), and the technique now belongs to the standard repertoire of biomineral specimen-preparation methods for TEM (Saunders et al., 2010; Wirth, 2009; Grandfield & Engqvist, 2012). FIB milling typically involves the use of a Ga+ beam (Giannuzzi & Stevie, 1999), which is accelerated to an energy of 1–50 keV. When the ions hit the specimen surface, ions or atoms from the sample are sputtered away. While such an instrument can also be used as an ion microscope, by forming images using the sputtered ions (or secondary electrons that also form in the beam–specimen interaction), the primary interest is the capability of the focused ion beam to mill away precisely controlled parts of a specimen. Thin sections for TEM studies can be cut along preset directions in a sample (Fig. 23) and transferred to a TEM grid using an in situ or ex situ micromanipulator. Despite its great utility, FIB milling also has its drawbacks. Notably, Ga ions are implanted into the specimen, and ion bombardment results in amorphization of the top few tens of nanometers of the specimen surface. FIB milling is especially deleterious for magnetic properties even if the sample appears to be perfect crystallographically. The use of a lower-voltage focused ion beam, or the use of species such as Xe for milling instead of Ga, may alleviate these problems in the future. Alternatively, the damaged surface layer can be removed using a conventional Ar ion milling system after completion of the FIB milling process.
FIB milling of cryogenic specimens facilitates the preparation of TEM specimens of soft materials. In addition, FIB or cryo-FIB milling can be used for visualizing samples in three dimensions. By milling away and then imaging successive slices of a biological specimen, 3D reconstructions of morphologies and compositions are possible (Schaffer et al., 2007). For example, the biogenicity, authenticity and taphonomy of ancient (Precambrian, ∼3400 and ∼1900 million year old) microfossils have been studied by using a combination of FIB-TEM and FIB-SEM nanotomography (Wacey et al., 2012).
Enhancement of image contrast and reduction of radiation damage
As the organic components of biological samples consist primarily of light elements, with very little local variation in scattering, both amplitude and phase images have low contrast. Image contrast is reduced further in the case of cryo-TEM samples, in which the vitrified solution and the biological structures of interest have very similar densities (Evans et al. 2008). In addition, the use of low-dose techniques can result in noisy images. As described above, conventional biological TEM specimen preparation may involve the staining of specimens with heavy metal-containing compounds. In addition to improving contrast, staining typically makes the specimen less sensitive to electron-beam damage. However, the addition of extraneous elements changes the composition of the specimen and can hinder the analysis and understanding of its original structure and composition. In addition, the binding of certain chemicals to cell components can be highly selective (see Figure 21), and the relatively large grain size (∼ 1 to 1.5 nm) of the heavy metal stain limits the achievable resolution (Massover, 2008; Woodcock & Baumeister, 1990).
Several techniques offer useful alternatives to contrast enhancement by staining. For example, energy-filtered TEM can be used to improve image contrast by selecting the energies of electrons that contribute to an image (Angert et al. 2000). Just as for compositional analysis, the energy window can be set to any range in the energy-loss spectrum, the most straightforward choice being the zero-loss peak. As illustrated in Figure 24, the contrast in a zero-loss-filtered BF TEM image can be significantly greater than that in an ulfiltered BF TEM image of an unfixed, unstained biological specimen.
Another approach for enhancing image contrast is to use a lower accelerating voltage. Although biological TEMs already often use voltages of only 80-100 kV, it has been suggested that the use of an extremely low accelerating voltage of 5 kV could result in enhanced contrast (Drummy et al., 2004). However, the specimen would then need to be very thin (< ∼20 nm) and the spatial resolution would be affected significantly by the chromatic aberration of the TEM objective lens, even if spherical aberration correction were used. Spherical aberration correction alone, in combination with low-dose, cryogenic imaging, has been shown to improve spatial resolution in images of radiation-sensitive organic specimens to better than 0.16 nm in two-dimensional, organic crystals (Evans et al., 2008). When Cs correction is combined with electron beam monochromation and/or Cc correction and a low accelerating voltage is used, further improvement can be achieved in the analysis of beam-sensitive specimens (Kabius & Rose, 2008; Kaiser et al., 2011).
Exciting new opportunities for image formation are offered by phase contrast techniques. As the electron wave passes through a TEM specimen, both its phase and its amplitude are modulated. In most conventional TEM imaging techniques, the phase shift is not recorded, and image contrast results primarily from the variation of the amplitude as the electron wave passes through different specimen regions. If the phase of the electron wave can be recorded directly, then images can be reconstructed with contrast that arises from the very small phase shifts introduced into the electron wave by a weak phase object (such as a biological specimen). In this way, neither sample staining nor a large objective defocus are needed to enhance image contrast. One of the possible phase-contrast techniques is electron holography. Although many types of EH are possible (Cowley, 1992), the technique of off-axis EH has gained in popularity in materials science. The application of off-axis EH to the study of magnetic fields in specimens has been discussed above. However, EH also offers great promise for imaging organic materials, and is beginning to be used in biological research for studying the crystal structures of proteins, the ultrastructures of ferritin, bacteriophages and bacterial flagella (Simon et al., 2008). A hardware method for generating phase contrast without the use of EH is referred to as the ‘Zernike phase technique’ (Danev et al., 2010), and involves the use of a phase plate (e.g. a carbon film with a small central hole). While the zero-order, central beam passes through the hole unaltered, the electrons that are scattered to small angles are phase-shifted. Images show enhanced contrast of low spatial frequency features and maintain high contrast near in-focus conditions (Alloyeau et al., 2010). As phase plates can be installed at the position of the objective aperture, the phase-plate technique may become a standard imaging option in the future. Another type of hole-free phase plate involves making use of local charging by the electron beam to generate a phase shift in a continuous film (Malac et al., 2012).
Although some of the above techniques have not yet been used for the study of bio-minerals, they have great potential for imaging hard/soft interfaces in organisms. Depending on the particular problem, a combination of several techniques may produce the best result. For example, the use of zero-loss energy-filtered TEM in a CS-corrected, low-voltage TEM, combined with the phase plate technique, could offer a simultaneous solution to the problems of radiation damage and weak image contrast.
The ultimate solution for imaging biological objects at (or near) their original conditions is potentially provided by environmental transmission electron microscopy (ETEM). Several different instrument designs with dedicated functions can be grouped under the term ETEM, but common to all instruments is a specimen chamber in which high-vacuum conditions are not retained. The chamber is separated from the rest of the column either by apertures that allow the operation of a differential pumping system, or by ultrathin, electron-transparent windows. ETEM can be used for studying reactions in situ under controlled conditions (for example, by the introduction of reactive gases (Sharma, 2005) or humidity (Wise et al., 2005)), while closed cells allow the study of specimens either in gases (Creemer et al., 2010) or in liquids (de Jonge & Ross, 2011; Grogan et al., 2012). In principle, even the imaging of live, functioning cells is possible (de Jonge et al., 2009). Despite its great potential, ETEM has only been used in a handful of cases for the study of biomineral-related scientific problems. The microbial reduction of Cr(VI) was studied in a liquid-cell environmental TEM (at 100 Torr pressure), combined with EELS to observe changes in the oxidation state of Cr associated with encrusted bacteria (Daulton et al., 2001). The cells of She-wanella oneidensis were unfixed and remained hydrated and intact until the electron beam started to cause radiation damage within minutes of exposure to the beam. The reduction of Fe(III) by the same bacterium species was studied by Kim et al. (2003), who observed a decrease in the basal plane spacing of nontronite in the ETEM when Fe was reduced in the sample.
Combination of TEM with other nanoscale techniques for the study of biominerals
As shown in the above sections, TEM techniques based on imaging, diffraction and spectroscopy can be used to analyse many different biomineral properties. However, it is often advantageous to combine TEM studies with analyses performed using other instrumentation, in order to acquire complementary nanoscale information about the material or to extend the observations to the micro- or mesoscale. The combination of different methods is facilitated by the recent development of tools for the transfer of specimens between different instruments, making it possible to analyse the same sample location using different techniques. In this section, we review several techniques that have been used successfully in combination with TEM for the study of biominerals. As the focus of this chapter is on TEM techniques, the principles and methodologies of the other techniques are not described in detail. Instead, results obtained from applications to biominerals are illustrated in the form of selected examples.
Correlative TEM–optical microscopy and electron backscatter diffraction
Optical microscopy is a natural extension of TEM from the nano- to the microscale, and has always been an indispensable tool in the process of TEM sample preparation. In life sciences, the combined use of optical and electron microscopy has a long history. Fluorescent markers that provide phylogenetic information on the basis of molecular features can be viewed using an optical microscope, and the same areas can be studied at atomic resolution using TEM to provide ultrastructural detail. However, the combined use of TEM and fluorescent optical microscopy faces severe challenges, including the adverse effect of electron-beam damage on fluorescent in situ hybridization (FISH) labelling, and, conversely, damage to cellular ultrastructure by the use of FISH. In order to overcome these problems, new correlative microscopy protocols have been developed for combining FISH with both conventional and cryogenic TEM (Knierim et al., 2012). These methods promise to facilitate the study of biomineralization processes in natural systems of microbial communities by revealing the spatial relationships between microbial community members and biominerals, along with cell-specific phy-logenetic information.
The automated mapping of crystal orientations using electron beam precession and scannning in TEM, as described above in Section 3.5, can be used to determine crystal-lographic directions with a spatial resolution that is limited only by the diameter of the electron beam, which can be <0.1 nm. A widespread and fully automated, albeit lower-resolution, alternative to TEM mapping is electron backscatter diffraction (EBSD) analysis in the SEM. The technique involves the identification of crystal orientations on the basis of Kikuchi lines in backscatter diffraction patterns. EBSD has found popularity, in combination with TEM, in the analysis of textured biomineral systems that include the shells of molluscs (Saruwatari et al., 2009; Okumura et al., 2010), avian eggs (Dalbeck & Cusack, 2006) and corals (Floquet & Vielzeuf, 2011). For example, EBSD analysis has been used to show that the crystallographic directions of calcite crystals in the outer, prismatic layer, and tablet-like aragonite crystals in the inner, nacreous layer of Mytilus edulis shells deviate from one another by <10° (Dalbeck et al., 2006). Controlled orientations of calcite fibres have also been found in brachiopod shells (Cusack et al., 2008), in which each fibre is composed of calcite/organic composite ‘nanogranules’ that share a common crystallographic orientation – the calcite c axis is perpendicular to the fibre axis and to the shell exterior. Such crystallographically organized assemblies of nanocrystals are referred to as ‘mesocrystals’ (Meldrum & Cölfen, 2008). Interestingly, ‘mesotwins’, i.e. micrometer-scale defective systems of oriented nanocrystals, occur in biological systems (e.g. corals), as shown using combined EBDS, SAED and HRTEM techniques (Floquet & Vielzeuf, 2011; Vielzeuf et al., 2010). A special application of EBSD is its combination with FIB milling. Slices of the specimen are milled away using FIB milling and EBDS maps are acquired from each new exposed surface. The maps are then used for the 3D reconstruction of crystal orientations or specimen composition (Wirth, 2009; Schaffer et al., 2007). This approach is commonly called the slice-and-view method.
TEM combined with X-ray-based techniques for compositional analysis
Compositional analyses of biominerals using TEM-based techniques are often complemented by micro-scale or, in the case of homogeneous samples, bulk compositional analyses. A variety of techniques can be used, including X-ray absorption spectroscopy (XAS), micro-X-ray fluorescence spectroscopy (|μXRF), synchrotron X-ray diffraction, atom probe tomography (Miller et al. 2007), Fourier transform infrared (FTIR) and Raman spectroscopy. For example, the authenticity of cellular structures in Precambrian nanofossils (acritarchs) was established using a combination of FIB sample preparation, EELS, atomic force microscopy (AFM) and Raman spectroscopy (Kempe et al., 2005). Another study that made use of HAADF STEM analysis, together with XAS, showed that the cycling of As in the aqueous environment is affected by microorganisms when a ferrous arsenate phase nucleates on cellular exudates of the arsenate-reducing Shewanella sp. ANA-3 (Babechuk et al., 2009). The accumulation of U, Cu and Fe in various parts of a lichen has been traced using high-resolution X-ray mapping in an SEM combined with IR spectroscopy (Purvis et al., 2004). Similarly, a combination of TEM with XANES showed that, in a mixture of kaolinite and bacteria, U accumulated selectively on the bacteria in the form of U6+ (Ohnuki et al., 2005). The oxidation of stoichiometric magnetite produced by magnetotactic bacteria has been studied using a combination of TEM and high-resolution structural analysis using synchrotron X-ray diffraction on bulk samples (Fischer et al., 2011).
For the determination of nanoscale compositional features in biominerals, the synchrotron-based method of scanning transmission X-ray microscopy (STXM) with associated X-ray absorption near-edge structure analysis (XANES, also referred to as ‘near-edge X-ray absorption fine structure’, NEXAFS (Vincze et al., 2010)) offers new opportunities, especially when combined with TEM techniques that provide high-resolution structural information. While the spatial resolution of STXM (typically ∼40 nm) is inferior to that of EELS, its spectral resolution is excellent (<0.1 eV), enabling studies of the speciation of organic compounds and the determination of metal oxidation states. In the last few years, several groups have used STXM XANES with TEM extensively to study biomineralization products, as are illustrated using several examples below.
The roles of polymer exudates of microorganisms in the formation of iron oxyhydr-oxide nanominerals in natural systems (biofilms) have been studied using STXM-based C and Fe XANES spectroscopy, |JLXRF and TEM (Chan et al., 2009; Miot et al., 2009). Polymer-Fe interactions were observed during the course of mineral formation in cultures of Fe-oxidizing microbes. Organic fibrils containing acidic polysaccharides with carboxylic functional groups were found to be correlated spatially with Fe oxy-hydroxide minerals. The fibrils appeared to collect and control the recrystallization of nanoparticles, thereby influencing the phases and shapes of the final mineral products. Other STXM TEM studies have revealed similar interactions between polysac-charide-containing microbial filaments and Fe nanoparticles (Benzerara et al., 2011b), as well as between extracellular polymers associated with cyanobacteria and an amorphous, aragonite-like carbonate (Obst et al., 2009).
Pyritized remains of plant tissue have been studied in a metamorphic rock (Bernard et al., 2010). Sample preparation by FIB milling allowed the analysis of the same specimen area using a variety of TEM techniques (STEM HAADF imaging, EDX mapping, SAED) in combination with STXM and XANES. On the microscopic scale, Raman spectroscopy and compositional maps complemented the analyses. This combination of techniques permitted the comprehensive characterization of the structure of carbonaceous matter in the fossil plant tissue, and the determination of the history of fos-silization (including the distinction between diagenetic and metamorphic effects) from the spatial relationships between the organic matter, iron oxides and iron sulfides. An interesting example of intracellular carbonate precipitation was discovered in cyanobac-teria that were previously thought to induce only extracellular biomineralization (Couradeau et al., 2012). While SAED patterns showed that the inclusions were primarily amorphous, STXM XANES spectra were used to characterize short-range order and to infer compositions. C K edge spectra suggested the presence of a carbonate mineral, while Ca L edge spectra were similar to reference spectra from a Mg-Ca-Sr-Ba carbonate (benstonite), consistent with EDX compositional maps obtained from the inclusions.
A special application of STXM is the mapping of carbonate nanocrystal orientations by using linearly polarized X-rays and measuring the absorbance as a function of polarization direction. The in-plane projected direction of the c axis of aragonite was mapped in this way in Porites coral with 40 nm spatial resolution (Benzerara et al., 2011a), complemented with HRTEM imaging and SAED analysis for smaller structural details. The so-called ‘centre of calcification’ (COC) was shown to consist of crystallographically aligned aragonite nanograins, which formed a ‘mesocrystal’ with single-crystal-like properties over a scale of several micrometers. A similar method was used to analyse the orientations of aragonite platelets in the nacreous layer of abalone (Gilbert et al., 2008), as well as calcite crystals in the prismatic layer of Pinctada fucata, with a spatial resolution of 10 nm (Gilbert et al., 2011).
Additional techniques for the nano-scale characterization of compositions, structures and physical properties in biominerals
In addition to EELS and XANES, the oxidation states and crystallographic site occupancies of Fe in biominerals can be determined if TEM studies are complemented by the application of X-ray magnetic circular dichroism (XMCD) to bulk samples (Staniland et al., 2007). XMCD is a synchrotron-based technique that uses circularly polarized X-rays to measure the magnitude of the projection of the magnetic moment on the propagation direction of the X-rays. The technique, therefore, provides an element- and site-specific probe of magnetism with a spatial resolution that is limited by the beam diameter. The fine structure of the Fe L2,3 edge in an XMCD spectrum can be used to measure the occupancies of the three crystallographic metal positions in the spinel structure (Carvallo et al., 2008). In a combined TEM XMCD study (Coker et al., 2009), the Fe(III)-reducing bacterium Geobacter sulfurreducens was used for the development of an environmentally benign method for the synthesis of cobalt ferrite nanoparticles with enhanced magnetic properties.
The combination of XMCD with STXM provides new opportunities for studying magnetic properties and compositions at the nanoscale. Although the spatial resolution of the technique (∼25 nm) is inferior to that of similar TEM-based methods, XMCD signals from individual magnetite particles within chains of magnetosomes in the cells of the magnetotactic bacterial strain MV-1 were measured successfully using STXM (Lam et al., 2010; Kalirai et al., 2012). As the fine structures of the C and O edges also contain information about the organic components, the combination of STXM XANES with XMCD provides a sensitive probe of both biochemistry and magnetization (Kalirai et al., 2012), and is, therefore, a promising approach for the study of magnetic biominerals.
An important extension to compositional analysis is provided by nanoSIMS (nanoscale secondary ion mass spectrometry), a technique that can be used for the study of iso-topic compositions (Boxer et al., 2009) with a spatial resolution approaching ∼50 nm. NanoSIMS has been used in combination with a cryo-tomography TEM study of intracellular features in a magnetotactic bacterium, Desulfovibrio magneticus (Byrne et al., 2010). Two types of Fe-bearing inclusions, membrane-bound amorphous Fe-P granules and magnetite nanocrystals, were found in the cells. The development of Fe-P granules and magnetite crystals was followed by isotopically labelling the Fe that was given to Fe-starved cells. In the initial 3 hours and in the rest of the experiment, 56Fe and 57Fe-bearing compounds were given to the cells, respectively. The isotopic composition of Fe in the amorphous granules and in the magnetite crystals suggested that two different cellular mechanisms were responsible for the formation of the two types of solid Fe-bearing intracellular particles, and that the amorphous granules were unlikely to serve as precursors for the formation of the magnetite nanocrystals. In another study on the biosequestration of uranium by Geobacter sulfurreducens (Fayek et al., 2005), combined HRTEM and nanoSIMS imaging indicated that the cells incorporated acetic acid and reduced U6+ to U4+, resulting in the precipitation of uraninite on the cell surfaces.
In order to measure the mechanical properties of biominerals on the nanoscale, various versions of scanning probe microscopy can be combined with the versatility of TEM. AFM has been used extensively for the study of coccoliths, diatoms, and bone and teeth material (Baeuerlein, 2007), while magnetic force microscopy (MFM) has been used to study magnetic nanoparticles in bacteria (Proksch et al., 1995) and fish (Diebel et al. 2000). In addition to the benefits of 3D imaging capabilities and ambient operating conditions, AFM is also useful for the micromanipulation of biological material, for example the dissection of cells (Yamamoto et al., 2010). Nanoindentation measurements on various biominerals have provided interesting data about their mechanical properties. For example, sharks and humans have teeth with comparable hardness (Enax et al., 2012); more surprisingly, the hardness of crayfish mandible is also similar to that of mammalian teeth, despite their different mineralogical compositions (amorphous calcium carbonate and phosphate minerals in crayfish, and crystalline apatite in mammals) (Bentov et al., 2010). A study that combined TEM with Raman spectroscopy and nanoindentation measurements found that the mineralized radular teeth of the chiton, a marine mollusc, are three times harder than human enamel or the carbonate shells of molluscs, and represent the hardest biominerals that have been reported to date (Weaver et al., 2010).
In the last 15 years, a wide range of TEM techniques has been developed for the characterization of biominerals with much greater spatial resolution and greater versatility than was possible before. Because of their nanometric sizes and interesting physical properties, biominerals have been used as a ‘testing ground’ for some of the new techniques that have been discussed in this chapter, such as electron holography and cryo-tomography. Some other cutting-edge techniques, such as atomic-scale mapping of compositions, have not yet been applied fully to the study of biomineralization products. As methods for biological specimen preparation and TEM imaging are continually being improved, new avenues will open for the atomic-scale characterization of electron beam-sensitive samples, as well as for the correlative application of molecular-scale biological microscopy with techniques used for the physical and chemical characterization of specimens. Given the multidisciplinary nature of the scientific problems and the research methodologies that are associated with and used for the study of biominerals, a combination of a wide range of techniques will probably become even more common than today. However, TEM is likely to remain at the core of biomineral studies.
We thank Dennis Bazylinski, Peter Buseck, Richard Frankel and Dirk Schüler for many years of joint work on magnetotactic bacteria. Experimental measurements, analyses, sample preparation and other scientific contributions by Balázs Arató, Ryan Chong, István Dódony, Zoltán Kristóf, Christopher Lefèvre, Anna Lohße, Ilona Nyirö-Kósa, Nicolas Menguy, Péter Pekker, Oliver Raschdorf, Aleksander Rečnik, Ed Simpson, Éva Tompa and René Uebe were used for the writing of this chapter and are gratefully acknowledged. Access to the transmission electron microscope facility at the Institute for Technical Physics and Materials Sciences of the Hungarian Academy of Sciences is acknowledged. Reviews by Ken Livi and Karim Benzerara improved the manuscript and are gratefully acknowledged. This work is based on research that was supported by EU FP7 (BIO2MAN4MRI) and joint EU–Hungarian (TAMOP 4.2.2/B–10/1/2010– 0025/KKDI) grants and a European Research Council Advanced Grant (RDB).
Figures & Tables
Minerals at the Nanoscale
The first transmission electron microscope (TEM) was built in the 1930s for which the inventor Ernst Ruska received a Nobel prize for Physics in 1986. Since those early years, Mineralogy has played an important role in the development of TEM methods and technology. In the beginning, early specimens included natural samples of biomineralization (diatoms in 1937), and later, some of the larger unit cells and microstructures of minerals were accessible to the low-resolution microscopes of that time (see chapter 5, for example). One of the main drives for the development of the electron microscope was its application in biology and medicine. Mineralogy would rejoin with biology at a time closer to the new millennium as interest in biogeochemistry became more prevalent (chapters 10 and 11). Through the following decades, the limitations of specimen thickness, contamination, beam damage and resolution were progressively overcome. In the 1970s, there were a few major influences on the application of TEM to minerals: the increase of resolution to the point of direct observation of mineral structures (see chapter 4 for minerals about which important discoveries were made at this time), the development of the ion mill to thin crystals to electron transparency, and the increase in funding for these new expensive machines to tackle problems such as lunar and extraterrestrial petrology (chapter 2). This decade also saw the increase in high-voltage technology that decreased the wavelength of electrons, and thus, increased their resolving power. We also saw the principles developed in 'Materials Science' being applied to minerals. Students of microscopy would probably be trained by both mineral crystallographers (to understand mineral structures) and Materials Science/Physics microscopists (to understand microstructures) (see chapters 3 and 8 for modern studies of defect microstructures and nonstoichiometery). Developing in parallel with the TEM was the dedicated scanning TEM (STEM). This technology would have a slower trajectory before establishing its importance in Materials Science, but has yet to be fully applied to mineralogy (see chapter 1 for a few examples of STEM applications). In the 1980s, universities were purchasing high-end TEMs for both Materials Science and Mineralogy. In several universities, Earth Science departments obtained scopes that were either dedicated to their own research or used for the university as a whole. Thus, the study of geological materials became more common. This was aided by the improvement of goniometric stages allowing orientation of samples, removing the need for the operator to pre-select a grain with adequate orientation. Two major analytical tools became available at this time: energy-dispersive X-ray spectroscopy (EDX) and electron energy-loss spectroscopy (EELS) (see chapter 1). Due to simpler implementation and interpretation, EDX rapidly became an integral tool in mineralogical investigations, whereas EELS, with its more stringent thickness requirements, more complicated interpretation, and more expensive instrumentation, would increase slowly in terms of its mineralogical applications. EDX allowed mineralogists to extend the acquisition of quantitative analyses below the micrometer threshold of the electron microprobe. This opened up the field of investigations into fine-grained materials such as sedimentary and low-temperature metamorphic rocks (chapters 6 and 7). The 1980s were a good time to buy a microscope, because the technology had stabilized and improvements in the 1990s were mainly in the area of computerization of the microscopy interface. However, the Schottky Field Emission gun became increasingly common during the 1990s, which did improve the STEM imaging capability and the spatial resolution of chemical analyses of conventional TEMs. In this decade, we began to see an increase in applications of TEM to environmental materials, biominerals, and the use of EELS to determine valence states of transition elements and produce energy-filtered compositional images. An important development in sample preparation and characterization came to fruition at this time through the methods of focused ion beam and dual beam instrumentation (these methods are not covered in this book, however). These tools enabled researchers to target specific areas for extraction and preparation into thin foils for TEM observation. They have also become important characterization instruments in their own right as nanoscale 3D reconstruction investigations of solids have proven valuable. The new millennium brought a breakthrough in imaging technology through the realization of aberration correctors on both conventional TEMs and dedicated STEMs which were now being remarketed. The resolution of electron microscopes, which had improved only slowly over the previous two decades, suddenly broke the 1Ångstrom mark — we truly entered the picometer era. This came at the time when interest in nanoparticles in Materials Science, Chemistry and Physics was exploding. Parallel to that was the understanding of the effects of nanosized particles on the environment (chapter 9) and in health (for good and bad reasons). In addition, tomography techniques were developed to bring out the third dimension texturally and chemically (see chapter 11). Added to this, environmental TEMs enabled the observation of in situ reactions under changing environments. So now in the 2010s, we are witnessing the push for the improvement of detectors for imaging (direct detection of electrons) and X-rays (silicon drift detectors and annular high solid-angle of collection detectors), the development of new support materials (e.g. graphene) and liquid cells for TEMs. Most of these new technologies have not yet been applied to mineralogical problems but we hope they will be in the near future. A new application using older technology is found in scanned precession electron diffraction (SPED). This technique can: (1) decrease the dynamical behaviour of electron diffraction, and thus, allow for the use of diffraction intensities for ab initio structure determination of nanocrystals; and (2) automatically produce orientation maps at the nanoscale. SPED is a recent development that has seen immediate applications to mineralogy.