Sample preparation
Two different thin/electron beam transparent lamellae (4 × 4 microns size and approximately 100 nm thick) were prepared by means of Focused Ion Beam (FIB). Those slices were FIB cut (using Ga ions during several hours) from larger tesserae fragments having green and white colours (Additional file 1: Fig. S1). All samples were lifted out from FIB on to specific TEM grids for subsequent observation and examination with ASTAR and ADT techniques.
Technique and instrumentation
For FIB specimen preparation the FEI Dual Beam Helios NanoLab600 at LMA Zaragoza (Spain), was used. The FIB protocol for lamella preparation was the standard one (in parenthesis, the acceleration voltage and the FIB current are specified) following the next steps: (1) deposition (via focused electron and ion beam deposition) of some amount of Carbon in the area of interest to protect the top portion of the sample; (2) the milling of the trench/slice in the sample area where the Carbon has been deposited (30 kV–2.5 nA); (3) the polishing of this trench/slice (30 kV–0.23 nA); (4) the undercutting of the trench/slice (30 kV–2.5 nA); (5 and 6) cut-off and lift-out; (7) final thinning (5 kV–68 pA) once the lamella is already welded to the special TEM Cu support. For TEM/STEM (scanning transmission electron microscopy) observations, a TEM Jeol 2100 (LaB6, 200 kV) at the University of Patras (Greece) and a TEM Jeol 2100F (FEG, 200 kV) equipped with “Digistar” precession system (NanoMEGAS SPRL, Belgium) at UPV-Valencia (Spain) and at SiMap (Grenoble, France) were used [9]. The chemical composition from the polished cross thin sections were obtained by means of EPMA (Electron Microprobe Analyzer) using a JEOL JXA-8230 at the Centres Científicsi Tecnològics of the Universitat de Barcelona (Spain). The measurement conditions were 20 kV, at 15 nA probe current, spot size of ca. 2 µm and counting time of 20 s per element. The calibration standards used were: hematite (Fe, LIF, Kα), rutile (Ti, PET, Kα), periclase (Mg, TAP, K Mn, LIF, Kα), rhodonite (Mn, LiF, Ka), Al2O3 (Al, TAP, Kα), metallic antimony (Sb, LIF, Lα), metallic tin (Sn, LIF, Lα), diopside (Si, TAP, Kα), CuO2 (Cu, LIF, Kα), wollastonite (Ca, PET, Kα), metallic silver (Ag, LIF, Lα), metallic cobalt (Co, LIF, Kα), albite (Na, TAP, Kα), orthoclase (K, PET, Lα), galena (Pb, LIF, Lα), AgCl (Cl, PET, Kα) and Celestine (S, PET, Kα). The EMPA point analyses and the elemental maps were performed directly on the carbon-coated surface of the sample on selected spots corresponding to the different regions of interest in order to probe both the glaze and pigments chemical composition.
The STEM EDS (X-Ray analysis) spatial resolution was of range of 3–5 nm (in case of TEM-FEG microscope) more suitable than the EPMA spatial resolution close to 1 micron to study the tesserae pigments. However, the EPMA compared to EDS provides a much better energy resolution for the selected emission lines (5–20 eV vs 130–150 eV for STEM-EDS). Therefore it was very useful in our study to combine high resolution STEM-EDS with high energy resolution EPMA to study crystals composition.
The orientation imaging and phase mapping in ASTAR technique is performed through automated collection of electron diffraction (ED) patterns on an area of several nm while scanning the area of interest with nm beam size; the collected ED patterns were fitted through template matching with pre-calculated theoretical ED patterns (templates) of all possible existing phases and relative orientations (Fig. 1). The resulting coloured crystal orientation map has usually a 1–3 nm spatial resolution (related to the TEM FEG probe size) and each pixel colour corresponds to a particular orientation in the stereographic triangle (see Fig. 1) (orientation resolution is close to 1°). Typically areas of several square microns can be examined (typical step size from 1 to 10 nm) to obtain orientation and phase maps of all known crystals phases within the examined area). It is also possible to obtain crystalline/amorphous map areas at nm scale [1].
The novel TEM based technique ADT (Automated 3D diffraction tomography) allows to analyze nm sized crystal structures using 3D electron diffraction data from single nanocrystals (Fig. 2); it is based on sampling the reciprocal space of the examined crystal in small steps (usually 1° tilt) without any prior information on the structure and orientation of the crystal.
The only essential requirement is that data should be collected from the same crystal, in such a way that large numbers of reflections are typically recorded through a tilt around an arbitrary axis. As a result, the 3D reciprocal volume of the selected crystal is reconstructed where diffraction tomography data contains nearly all reflections present in the covered wedge of reciprocal space. In practice recording of 40–60% of the reciprocal space volume is enough for unit cell and symmetry determination and allows crystal structure solution for most of high symmetry systems (cubic, tetragonal, hexagonal) [4]. It is also possible for complete structure determination to combine datasets taken from various crystals of the same crystal phase, using same reflection intensities as scale factor.
The ADT diffraction tomography can be performed in any TEM using a standard single tilt or tomography holder. An efficient sampling depends on the crystal symmetry; the higher the symmetry the smaller the minimum angular range required, however a tilting range from − 60° to + 60° along the goniometer axis with a tilting step of 1° is an optimal compromise [4]. Therefore, a total tilt wedge of 120° can be recorded, providing 121 diffraction patterns that are usually enough for unit cell and crystal structure determination.
The precession electron diffraction (PED) is an important electron crystallography technique that has been developed the last 15 years as a technique suitable to solve crystal structures of various nanomaterials as renders ED intensities with less dynamical effects [10, 11]. This method is based on the precession of the incident electron beam, which is inclined away from the optical axis of the TEM and precess through a cone surface having the vertex fixed on the sample.
Due to beam precession, (usually applied at 1° semi-angle) reflection intensities are integrated over diffraction conditions that are far from perfect zone axis orientation, therefore dynamical effects in PED patterns are highly reduced. Using PED reflections in combination with ADT-3D tomography is important as symmetry related PED reflections are easily revealed and this enables symmetry (Space Group) determination. Therefore, PED intensity comparison between possibly symmetry related ED intensities enables to distinguish between crystals having similar unit cell (e.g. as close as 1–2%) but different crystal symmetries. In addition, during crystal tilt in ADT (− 60° to + 60°), the use of PED (at 1° semi-angle) helps to recover more reliable “quasi-kinematical” ED reflections intensities within reciprocal space sections taken every 1° tilt step. On the other hand, orientation imaging ASTAR technique also uses PED reflection in comparison with ED theoretical templates as use of PED improves a lot resulting phase and orientation maps [2].