This paper describes a protocol to achieve 3D chemical maps combining energy filtered imaging and electron tomography. The chemical distribution of two catalyst supports formed by elements that are difficult to distinguish by other imaging techniques was studied. Each application consists of mapping overlapped chemical elements – respectively spaced-ionization edges.
Energy filtered transmission electron microscopy tomography (EFTEM tomography) can provide three-dimensional (3D) chemical maps of materials at a nanometric scale. EFTEM tomography can separate chemical elements that are very difficult to distinguish using other imaging techniques. The experimental protocol described here shows how to create 3D chemical maps to understand the chemical distribution and morphology of a material. Sample preparation steps for data segmentation are presented. This protocol permits the 3D distribution analysis of chemical elements in a nanometric sample. However, it should be noted that currently, the 3D chemical maps can only be generated for samples that are not beam sensitive, since the recording of filtered images requires long exposure times to an intense electron beam. The protocol was applied to quantify the chemical distribution of the components of two different heterogeneous catalyst supports. In the first study, the chemical distribution of aluminum and titanium in titania-alumina supports was analyzed. The samples were prepared using the swing-pH method. In the second, the chemical distribution of aluminum and silicon in silica-alumina supports that were prepared using the sol-powder and mechanical mixture methods was examined.
The properties of functional materials are dependent on their 3D parameters. To fully comprehend their properties and enhance their functions, it is important to analyze their morphology and chemical distribution in 3D. Electron tomography1 (ET) is one of the best techniques to provide this information at the nanometer scale2,3. It consists of rotating the sample over a large angular range and recording one image at each angular step. The obtained tilt series is used to reconstruct the volume of the sample by using mathematical algorithms based on the Radon transform4,5. Selecting grey levels in the volume helps to model the sample in 3D and quantify 3D parameters like particle localization6 and size distribution7, pore position and size distribution8, etc.
In general, ET is performed with an electron microscope by tilting the sample to the maximum possible angle, preferably more than 70° in either direction. At each tilt angle, a projection of the sample is recorded forming an images tilt series. That tilt series is aligned and used to reconstruct the volume of the sample which will be segmented and quantified. Because the sample cannot be rotated from -90° to +90°, the reconstructed volume has an anisotropic resolution along the orthogonal axis9 due to the blind recording angle.
ET can be performed in different imaging modes. The bright field TEM mode (BF-TEM) is used to study amorphous materials, biological samples, polymers, or catalyst supports with complex shapes. The image analysis is based on the differentiation of the gray levels characterizing the density of the components10 (a dense component will be more dark than a lighter, i.e., less dense component). High-angle annular dark field in scanning TEM mode (HAADF-STEM) is used to analyze crystalline samples. The signal provides chemical information as a function of the atomic number; a heavy component of the sample will appear brighter that a lighter one9. Other modes, like Energy Dispersive X-ray spectroscopy (EDX), which collects the X-ray emitted by the material11, and energy filtered imaging mode (EFTEM)12,13, are also capable of assessing the 3D chemical distribution within the sample.
In EFTEM imaging, the 2D chemical maps can be recorded using a TEM with an electron energy spectrometer. The spectrometer acts as a magnetic prism by dispersing the electrons as a function of their energy. An image is created by the electrons depending on the energy lost from interacting with a specific atom. If the same 2D chemical map is computed at different tilt angles, a tilt series of chemical projections is obtained, which can be used to reconstruct the 3D chemical volume.
Not all the materials can be analyzed by EFTEM tomography. The technique is reserved for samples with weak or disordered materials. Nevertheless, it can be used for analyzing light elements that are very difficult to differentiate when using other imaging techniques. In addition, to obtain reliable 2D chemical maps, the thickness of the material is required to be less than the mean free path of the electrons through the material14. Under this condition, the probability of having a single electron interacting with a single atom is greatest. Two methods are used to calculate a 2D chemical map. The first one, and the most used is the "three-windows method", where two filtered energy windows are recorded before the ionization edge of the element under analysis, and a third after the ionization edge13. The first two images are used to estimate the background, which is extrapolated using a power law at the position of the third window and subtracted from it. The obtained image is the projection of the 3D distribution of the analyzed chemical element in the sample volume. The second method is called the "jump-ratio"; it uses only two energy-filtered images, one before and one after the ionization edge. This method is qualitative, as the final image is calculated only by performing the ratio between those two images, and does not account for background energy variation.
By combining EFTEM with ET, the analytical tomography of the filtered energy can be obtained. EFTEM tomography and atom probe tomography (APT) are complementary techniques. As compared to APT, EFTEM tomography is a non-destructive characterization analysis that does not need complex sample preparation. It can be used to perform various characterizations on a unique nanoparticle. EFTEM tomography can analyze insulating materials, while APT needs at the very least laser assistance to measure them. APT runs at the atomic scale, while EFTEM tomography performs adequately with a lower resolution. EFTEM tomography is pertinent only for samples that resist beam degradation during the experiment. To record all the filtered images at all the tilted angles, the sample can be exposed to the electron beam for as long as 2 h. Moreover, to record a maximum chemical signal in the 2D maps, longer exposition durations at high beam intensity may be necessary. In such conditions, the beam sensitive samples suffer drastic morphological and chemical changes. Therefore, a precise measurement of the sample sensitivity to the electron beam must be established before the experiment. In addition, EFTEM tomography is the result of recording as many tomograms as necessary to determine the spatial location and nature of the chemical elements that are present in the sample. Nevertheless, EFTEM tomography can provide important information concerning the 3D chemical distribution for samples, such as catalyst supports, to give new insights for modeling their catalytic applications.
Today it is possible to use dedicated software that can select the energy interval, record filtered energy window images, and calculate the chemical maps at different tilt angles. They allow tilting the sample, tracking, focusing, and recording the filtered image in EFTEM mode. The 2D chemical maps can be calculated, and then the tilt series can be aligned, the chemical volume computed using iterative algorithms, and finally the series can be segmented and quantified15,16.
1. Sample Preparation
2. Recording of the Filtered Tilt Series Images
3. Alignment and Reconstruction of the Tilt Series
4. 3D Modeling and Quantification
An example of the application of this protocol is shown in reference13. EFTEM tomography was used for analyzing titania alumina catalyst supports. To enhance the catalytic activity of the active phase of MoS2 nanoparticles, in applications like hydrodesulfurization (HDS), it is important that titania is preponderant at the support surface, and in contact with the active phase. It is known that titania has a smaller specific surface than alumina. The aim of the study is to construct titania supported by alumina (and thus, create an enhanced specific surface), and then to test it as a catalyst support. Here, the EFTEM tomography is used to analyze of titania-alumina heterogeneous catalyst support prepared by the swing pH method. In this study, three samples of different titania concentrations are analyzed. Sample 1 is composed of 50% alumina and 50% titania, Sample 2 is composed of 70% alumina and 30% titania, and Sample 3 is composed of 90% alumina and 10% titania. In Figure 1a–1c, cross sections of the chemical maps parallel to the XY plane are shown. Green represents the spatial chemical distribution of titania, red represents the distribution of alumina, and blue represents the vacuum. The chemical volumes are calculated from the tilt series with the three-windows method. The filtered energy windows were as follows: 10 eV wide, centered at 59, 70, and 81 eV, having an exposition time of 3 s for the Al L23 edge situated at 73 eV and 30 eV wide, centered at 415, 445, and 482 eV, having a recording time of 15 s for the Ti L23 edge situated at 455 eV. The position and the width, as well as the exposure time of the filtered images were chosen in order to obtain a chemical map with a detectable chemical signal. The tilt series were recorded between -71° and +71° with an increment of 4° in the Saxton's mode24 in about 119 min.
It was found that titania at a high concentration is forming clusters that are embedded in the alumina. The models are displayed in Figure 1d (Sample 1), 1e (Sample 2), and 1f (Sample 3). In the models, the titania is displayed in blue and the alumina is displayed in transparent red. These models were quantified using the chemical distribution of titania and alumina on the surface of the samples. It was found that independent of the proportion of titania and alumina in the sample, the surface of the sample is covered with titania in a proportion of 30%. However, the specific surface of the sample is increasing, while the titania proportion is decreasing to reach the specific surface of alumina. For Sample 3 containing only 10% titania, a layer of about 10 nm thick is formed on the surface of the sample. Also, the chemical map is formed by overlapping three volumes: silica in red, titania in green, and zero loss in blue. The mixture between red and green are yellow voxels. The yellow voxels are attributed in the model to the element having the highest intensity in that voxel. This is a limitation in the spatial resolution of 3D chemical maps, which is directly related to the anisotropic resolution in ET and the resolution of the 2D chemical maps provided by EFTEM. The analysis is correlated with other analytical techniques such as X-ray fluorescence, X-ray photoelectron spectroscopy, and N2 porosimetry. It was concluded that the difference between the specific surfaces could play a role in the catalytic applications.
As a second example, the study detailed in17 is shown. In this study, we analyzed a series of silica alumina catalyst supports. These catalyst supports have an acidity provided by the mixture between the alumina and silica, forming an aluminosilicate. The goal of the study was to quantify the mixture between the two components. The experimental challenge lied in the fact that the L23 edges of Al and Si are very close, at 73 eV and 99 eV respectively, and the ionization edge of Al overlaps with the ionization edge of Si. Under these conditions, the three-window method is less accurate for extracting the chemical signal. In order to differentiate the two signals of Al and Si, the "R-ratio" method was developed, detailed in reference12. The filtered images tilt series were recorded by tilting the sample form -71° to +71° with an increment step of 4° in the Saxton's mode in about 83-104 min. Three filtered images were recorded to isolate the signal of the L23 ionization edge of Si. The images were centered at 59 eV, 70 eV, and 81 eV, were 10 eV wide, and exposed for 5 s. For the signal corresponding to the L23 edge of Al, only two filtered images were recorded, centered at 99 eV and 110 eV, 10 eV wide, and exposed for 12 s.
In this study, we analyzed a series of four samples of Al and Si prepared by different methods. Figure 2a is a cross section of the chemical map parallel to the XY plane and the model of the sample prepared by the sol-powder method. This sample was thermally treated under steam, yielding the second sample, whose chemical map and model are displayed in Figure 2b. Figure 2c shows the chemical maps of the sample prepared by mechanical mixture. From this sample, after a thermal treatment under hot steam, we obtained the fourth sample, shown in Figure 2d. The chemical maps and models for alumina are red and for silica are green, while the enhanced blue represents the boundary at the surface between the silica and alumina. The acid catalytic activity is given by the mixture between alumina and silica on the surface of the sample. It is found that independent of the preparation method, the silica covers only 30% of the sample surface. After the thermal treatment, the chemical distribution is more homogeneous, and the surface is covered with 50% silica and 50% alumina. The sol-powder method provides samples with a high homogeneity between the components compared with the mechanical mixture. Small domains of silica embedded in alumina are present in the sample. For the sample prepared by mechanical mixture, silica forms the core of the sample and alumina is present as a shell. As a general characteristic of both samples that are not thermally treated, the silica is in the center, and the alumina is on the surface.
The acid site density provided by the aluminosilicate phase formed by the intimate mixture between silica and alumina, creating Brønsted acid sites at its surface, is measured in arbitrary units (a.u.)/m2 by CO adsorption. The quantization of the boundary between silica and alumina was performed in m/g or m/m2, which are known physical units. Of course, the interface between silica and alumina could be thicker, but the spatial resolution reached did not allow the calculation of an exact value of the corresponding width. However, this study opens the way towards a deeper understanding of the interface between silica and alumina.
Figure 1: Cross sections and reconstructed models of the titania and alumina samples. Cross sections through the chemical map parallel to the XY plane, where the chemical distributions are shown for titania (in green), alumina (in red), and the vacuum (in blue): (a) Sample 1, 50% alumina and 50% titania, (b) Sample 2, 70% alumina and 30% titania, and (c) Sample 3, 90% alumina and 10% titania. (d) (e) and (f) display the models of Sample 1, Sample 2, and Sample 3, respectively, with titania in blue and alumina in transparent red. In Sample 1 and Sample 2, the alumina embeds the titania. In Sample 3, a thin layer of 10 nm of titania is formed on the surface of the sample. This figure has been modified from Roiban et al.13 Please click here to view a larger version of this figure.
Figure 2: Cross sections and reconstructed models of the silica and alumina samples. On the left are cross sections parallel to the XY plane from the chemical volumes, on the right are the reconstructed models. Alumina is shown in red, silica in green, and the boundary between the surfaces of the silica and alumina are indicated in blue. In those models, the boundary is artificially dilated by a 4-voxel sphere to make it visible. (a) The sample prepared by the sol-powder method, (b) the sample prepared by the sol-powder method and thermally treated, (c) the sample prepared by the mechanical mixture method, and (d) the sample prepared by the mechanical mixture method and thermally treated. This figure has been modified from Roiban et al.17 Please click here to view a larger version of this figure.
The aim of this paper is to describe how to obtain 3D chemical maps using EFTEM tomography. This protocol is completely original and was developed by the authors.
EFTEM tomography as described here has several drawbacks: (i) Only samples that are electron beam resistant can be analyzed, due to the long exposure time needed for obtaining filtered images. (ii) EFTEM tomography is sensitive to the diffraction contrast. (iii) Many of the alignments were performed manually. To obtain the 3D chemical map, the zero-loss volume and the chemical volumes need to be in a single system of coordinates. This necessitates that all the tilt series be aligned perfectly in the same coordinate system. This represents a long work period of at least two weeks per sample. Despite being time consuming, this protocol allows the computation of 3D chemical maps at a nanometric resolution. In addition, combined with other spectroscopic and analytic techniques such as, X-ray fluorescence, X-ray photoelectron spectroscopy,FTIR spectroscopy, or magic-angle-spinning (MAS) NMR spectroscopy, a full description of a functional material can be created.
The electron beam intensity is generally controlled by condensing the electron beam. The width of the energy window through which the filtered images are recorded, and their exposure time will influence the chemical signal intensity recorded in each chemical projection, which will be used as tilted projections to reconstruct the chemical volume. The exposure time of the filtered images will influence the total exposure time of the sample under an intense electron beam during the recording of the tilt series. If a sample remains for too much time under the beam, it can suffer drastic modifications. The width of the energy windows influences the background approximation using a power law to extract the chemical signal from the post-edge filtered image using the three-windows method.
Since it is a challenging technique, and time consuming, EFTEM tomography is not practical for wide scale implementation. However, new technical improvements such as the development of more sensitive spectrometers25 and fast recording cameras26,27 (the cameras references list is a partial list) will reduce the total record time of the tilt series and will enhance the energetic resolution of the chemical maps. As mentioned before, many of the alignments are manual, from signal extraction and computation of the chemical projections to the alignment of all projections on the same reference. The development of automatic procedures will create a more general use of EFTEM tomography in routine analysis.
The authors have nothing to disclose.
We are grateful to the French Ministry of Higher Education and Research, Conventions Industrielles de Formation par la Recherche (CIFRE) and IFP Energies Nouvelles for their financial support.
JEOL 2100f | JEOL | Electron microscope | |
Tridiem Gatan Imaging Filter (GIF) | Gatan | Post colum energy filter | |
Digital micrograph | Gatan | Software | |
Gatan EFTEM tomography plugin | Gatan | Dedicated software to record filtered tilt series for EFTEM tomograohy | |
Tomoj | Imagej plugin http://www.cmib.fr/en/download/softwares/ | Free software developed by Currie Institute in Paris, France for electron tomography | |
EFTEM-Tomoj | Imagej plugin http://www.cmib.fr/en/download/softwares/ | Free software developed by Currie Institute in Paris, France , for EFTEM imaging | |
Imod | http://bio3d.colorado.edu/imod/ | Free software developed by University of Colorado, USA for electron tomography | |
Imagej | https://imagej.nih.gov/ij/ | Free software developed by National Institute of Mental Health, Bethesda, Maryland, USA for images treatment | |
Merge channels | https://imagej.net/Color_Image_Processing | Fonction in Imagej allowing to give different colors to volumes while they are overlapped | |
3D Slicer | https://www.slicer.org/ | Free software developed by a large consortium lead by Ron Kikinis , Harvard Medical School, Boston, MA, SUA | |
Chimera | https://www.cgl.ucsf.edu/chimera/ | Free software developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco,for data segmentation, cuatification and visualisation of 3D models | |
silica alumina support of catalyst | IFPEN | sample prepared for eleboration of this protocol | |
titania alumina support of catalyst | IFPEN | sample prepared for eleboration of this protocol | |
alcohol | |||
water | |||
Au nanoparticles of 5 nm | BBI Solutions | ||
Holey carbn film 200 mesh microscopy grid | Agar | ||
EDX sepctrometer | Oxford Instruments |