Scanning electron microscopy (SEM): how it works, SE vs BSE images and EDS

An SEM focuses a beam of electrons into a small spot and scans it line by line over the specimen. At each point the beam generates several signals, and a detector turns one of them into the brightness of the corresponding pixel (Erlich et al. 2020):

• Secondary electrons (SE) come from the surface and give topographic information: edges, steps and pores (Erlich et al. 2020). How much of the surface they report depends on the accelerating voltage (Bijelić et al. 2023).
• Backscattered electrons (BSE) are beam electrons scattered back out of the sample. Their yield depends on the atomic number of the material and on the beam energy, so heavier phases look brighter (Rau & Zaitsev 2025; Erlich et al. 2020).
• Characteristic X-rays identify the elements under the beam; an energy-dispersive spectrometer (EDS or EDX) records the whole spectrum at once and detects every element except H and He (Newbury & Ritchie 2015).

Raising the beam energy drives electrons deeper into the sample and changes both the depth that is sampled and the image contrast (Bromley et al. 2026). For the inside of nanoparticles, or for atomic structure, you need transmission electron microscopy (TEM) instead (Bijelić et al. 2023).

SE images carry the surface topography (Erlich et al. 2020): use them to judge particle shape, size and morphology. Read the scale bar, and for a size distribution count many non-overlapping particles from several regions rather than one field of view (Bijelić et al. 2023).

BSE images are composition maps in shades of grey: brighter regions contain heavier elements on average (Erlich et al. 2020). In a polymer blend, for example, BSE atomic-number contrast is enough to show which phase is dispersed and which is continuous (Zhao et al. 2022).

EDS spectra show peaks at the characteristic X-ray energies of each element on a continuous background. An EDS map shows where each element sits. Check every automatically labelled peak by hand: commercial software occasionally misidentifies even major peaks, and overlapping peaks (as in PbS, MoS₂ or BaTiO₃) need peak fitting (Newbury & Ritchie 2015).

• Conductivity. Insulating samples charge under the beam. Coat them with a thin conductive layer (carbon, Au, Pt or Cr); coatings that are too thick hide fine surface detail (Belz & Auchterlonie 1995).
• Low-vacuum SEM. If you cannot coat, a low-vacuum (variable-pressure or environmental) SEM lets positive ions in the residual gas neutralise the negative charge that builds up on non-conductive samples (Sawaguchi 2026).
• For quantitative EDS the ideal specimen is flat and highly polished, with a carbon coat under 10 nm thick. Rough surfaces, fracture faces and loose particles change the X-ray intensities and can shift apparent concentrations a great deal (Newbury & Ritchie 2015).
• Cross-sections and 3D. A focused ion beam (FIB) in a dual-beam SEM mills away material with nanometre precision to expose subsurface structure or build 3D volumes slice by slice (Faber et al. 2024).

• MOFs and COFs: crystal size and shape of ZIF-8, HKUST-1 or UiO-66 powders. Pair the images with XRD for phase and BET for porosity.
• 2D materials: large-area coverage, wrinkles and folds. In CVD graphene on a Cu/Ni(111) foil, SEM showed folds 10–300 nm wide separated by roughly 20 µm (Huang et al. 2018); AFM gives their height.
• Battery materials: particle morphology in NMC or LiFePO₄ electrodes, and element distribution by EDS (see the battery materials hub).
• Composites and multiphase solids: BSE shows phases by atomic number (Zhao et al. 2022), and EDS maps show which elements are where (Newbury & Ritchie 2015).

For surface chemistry and oxidation states, use XPS; for bulk elemental composition at trace level, ICP-MS. All of these are compared on the characterisation techniques page.

• Charging artefacts (bright streaks, drifting images) on uncoated insulators; coat thinly or use low vacuum (Belz & Auchterlonie 1995; Sawaguchi 2026).
• Over-thick coatings that hide the fine structure you want to see (Belz & Auchterlonie 1995).
• Treating standardless EDS as exact. Standardless results are normalised to 100%, and about ±30% relative error is needed to capture 95% of results; by the authors’ informal survey probably more than 98% of reported quantitative EDS results are standardless (Newbury & Ritchie 2015).
• Quantifying EDS on rough particles. Shape and surface tilt alter X-ray absorption paths, so apparent Mg and Fe concentrations in one homogeneous glass spanned almost an order of magnitude on irregular specimens (Newbury & Ritchie 2015).
• Reading height from SEM. SEM gives limited height information; measure thickness and roughness with AFM (Joo et al. 2025).
• Small sample statistics. An image covers a tiny fraction of the sample; check with bulk methods such as XRD (Bijelić et al. 2023).

For a methods figure, draw the electron column above the sample with the SE and BSE detectors and the EDS detector at an angle, then show an SE image, a BSE image and an EDS spectrum side by side. In LabCanvas the figure editor Library has the SEM as an editable object, including a cutaway of the electron beam path, and an editable data plot for the EDS spectrum. You can label the element peaks, add a scale bar, and export the figure for a methods section or a graphical abstract.