Dividing two images cancels what they share. Subtracting keeps it. That one fact decides which map to trust.
An energy filter records whole images at chosen energy losses. Two images before an edge and one after it give an elemental map. Here you record them from a sample you set up, build both kinds of map live, and see what a bend contour, a thickness change, drift, low counts or a badly placed window does to each one.
Precipitates A (left) and B (upper right) are rich in element X, edge at 708 eV (like Fe L2,3). The matrix has none. 128 × 128 px, 1 nm per px.
Counts per pixel in Pre-edge 1, in the matrix (set in GMS by Exposure time and Binning). Bend contour: share of electrons it removes from all images. Slit width is 20 eV for all three images (fixed here). The edge onset is at about 702 eV.
On: the Post-edge image is lined up with Pre-edge 2 by cross-correlation and shifted back by whole pixels.
| Step | What you do | Knob on this page | What goes wrong |
|---|---|---|---|
| 1 Energy loss | Set the Energy loss of Pre-edge 1 and Pre-edge 2 below the edge onset (background only), and of Post-edge just above it (background plus element). Set the Slit width; 10 to 30 eV is common. | Pre-edge 1, Pre-edge 2, Post-edge energy loss | A pre-edge window that touches the edge takes signal away from both maps. |
| 2 Capture | Capture the three images, as fast as the counts allow (Exposure time, Binning). Keep the same objective aperture and focus. | counts per pixel | Long exposures let the specimen drift or bend between images. |
| 3 Drift Correction | Turn on Drift Correction, so the images are lined up (cross-correlation) before any arithmetic. | Drift Correction | A few pixels of drift give bright and dark fringes at every edge. |
| 4 Check | Divide Pre-edge 1 by Pre-edge 2 yourself (it is not a Gatan map). The result should show no sharp features. | check image (4) | Lines or blobs mean the contrast changed between images. Do not trust either map there. |
| 5 Jump-Ratio Map | Post-edge divided by Pre-edge 2. Use it to find where the element is. | map 6 | Not proportional to the amount. Falls in thick areas. |
| 6 Elemental Map | The three-window method: Background model power law, A·E−r, fitted through Pre-edge 1 and 2 pixel by pixel, then subtracted from Post-edge. | map 7, map 8 | Noisier. Keeps diffraction and thickness contrast unless you divide it out. |
Names follow Gatan’s EFTEM software (GMS): Energy loss, Slit width, Exposure time, Binning, Pre-edge 1, Pre-edge 2, Post-edge, Drift Correction, Background model, Jump-Ratio Map, Elemental Map, and Thickness Map (from an Unfiltered and a Zero-loss image).
A bend contour removes electrons by elastic scattering outside the objective aperture. It takes the same share from every image, background and edge alike. A division cancels that share, so the Jump-Ratio Map does not see the contour. A subtraction keeps it, so the Elemental Map shows the contour as a false drop. The trap for the Jump-Ratio Map is contrast that is not shared: a contour that moves between exposures, or drift. The check image catches both.
The scene is made up: two round precipitates of one element in a matrix with none, an edge at 708 eV shaped like an L2,3 edge, and a pure power-law background whose exponent falls a little in thick areas. Plural scattering is a simple extra factor on the background, not a full calculation. The bend contour is a smooth band that removes a fixed share of electrons; real diffraction contrast also depends on energy loss through the objective aperture, which this page does not model. Chromatic blurring, the point-spread of inelastic scattering (delocalization), detector gain and readout noise are left out. Alignment here moves the image by whole pixels only. The areal density in the Elemental Map would need the edge cross-section and the incident intensity to become a number of atoms per nm2.
On this site: Reading the EELS spectrum · EELS core-loss quantification · EELS thickness calculator · Bend contours · EELS fine structure and oxidation state
Usually both, with the check image in the supplement. The Jump-Ratio Map shows where the element is with less noise. The Elemental Map is the one that scales with the amount, so it is the one to quantify, after dividing out shared contrast and stating the thickness.
If the check image (Pre-edge 1 divided by Pre-edge 2) shows the same stripe, no. The contrast changed between exposures, most often a bend contour that moved (lesson 3). Re-record faster or at a different tilt.
The background under the edge is extrapolated from two images. Any noise in them changes the fitted exponent r, and the extrapolation magnifies that change (lesson 6). More counts, wider windows, or pre-edge windows closer to the edge all help.
In the Jump-Ratio Map, plural scattering adds background under the edge, so the ratio falls. In the Elemental Map, more atoms under the beam push it up. Neither is a concentration (lesson 4). Keep t/λ low and record a thickness map.
@misc{tripathy2026eftemelementalmapping,
author = {Tripathy, Manisha},
title = {EFTEM Elemental Mapping},
year = {2026},
howpublished = {\url{https://untetheredatom.com/tem/eftem-elemental-mapping}},
note = {Interactive web tool}
}