SEM / TEM
Turn a micrograph into particle statistics automatically.
Electron microscopy turns morphology into numbers: SEM with EDS gives particle statistics and composition maps at the micron scale; TEM goes further down, where lattice fringes and diffraction patterns identify the phases themselves.
What it measures
The two instruments answer at different scales, and most of the value is in making their images quantitative:
- SEM imaging: surface morphology from nanometers to millimeters. Segmenting the image converts qualitative “looks cracked” into counted statistics: particle sizes, shape factors, cracked fraction.
- EDS: characteristic X-rays identify elements; quantification requires ZAF correction (atomic number, absorption, fluorescence) before peak intensities become composition, and overlapping lines need explicit handling.
- TEM / SAED: transmitted electrons resolve lattice fringes, and selected-area diffraction gives d-spacings through the camera equation d = λL/R, identifying crystalline phases from ring or spot patterns. FFT of lattice-fringe images recovers the same spacings locally; nanoparticle populations fit Gaussian or log-normal size distributions.
How to read the output
Read distributions, not anecdotes. A single striking micrograph is a hypothesis; a segmented population of several hundred particles is evidence. For EDS, check that quantification used ZAF correction and that no reported element rests on an overlapped peak. For SAED, match the full set of d-spacings to a candidate phase: one matching ring is coincidence, four are an identification. And always carry the magnification context: a defect rate measured on a 50 µm field says little about a meter-wide electrode unless fields were sampled systematically.
A real use case
After a calendering pressure increase to chase higher electrode density, the team suspects NMC secondary particles are cracking. Cross-section SEM at matched magnification, before and after the change, goes through segmentation: the cracked particle fraction jumps from 4% to 19%, concentrated in the largest particles. EDS mapping on the same sections confirms the Ni:Mn:Co ratio is unchanged, it is a mechanical problem, not a chemistry problem. The pressure setpoint comes back down and the density target moves to a particle-size blend instead; the counted fraction, not an impression from two images, is what justified reversing a process change.
Common mistakes
- Drawing conclusions from one or two images. Sampling bias dominates microscopy; statistics come from systematically sampled fields and counted populations.
- Reading raw EDS peak heights as composition, without ZAF correction and overlap handling the numbers are wrong in proportion to how interesting they look.
- Quantifying light elements (Li, sometimes O) by EDS at all; the physics barely permits it and battery work constantly tempts it.
- Calibrating SAED carelessly: an off camera constant scales every d-spacing and can point phase identification at the wrong structure.
- Comparing images taken at different accelerating voltages or detectors and calling the contrast difference a material difference.
Micrographs in, populations out
Niobia makes the microscope quantitative. SEM images go through segmentation to particle-size distributions, histograms with mean and standard deviation annotated, and EDS spectra get peak identification with explicit overlap warnings, ZAF-corrected quantification, elemental ratios, and mapping analysis with composition bar charts. On the TEM side it measures SAED ring and spot patterns through the camera constant, matches d-spacings for phase identification, analyzes lattice fringes by FFT, identifies zone axes, and fits nanoparticle size distributions as Gaussian or log-normal, including aspect-ratio characterization for anisotropic particles. The cracked-fraction comparison above is exactly this pipeline run on two sample sets.
Frequently asked
How many particles make a size distribution trustworthy?
Hundreds, not dozens, and sampled from multiple fields. The mean stabilizes early; the tails, which usually carry the engineering consequence, need population before they mean anything.
Can EDS see lithium?
Effectively no, lithium's X-ray yield is too low for practical EDS quantification. Lithium accounting belongs to XPS (surface), ICP (bulk), or electrochemistry; EDS handles the transition metals and heavier elements.
When do I need TEM instead of SEM?
When the question is crystallographic or below ~10 nm: phase identity of a coating layer, lattice spacing changes, primary-particle structure. If the question is morphology, cracking, or micron-scale composition, SEM with EDS answers it at far lower cost.