SEM vs TEM: what is the difference?
Both instruments use electrons instead of light to see features far below the optical diffraction limit, and both are essential to semiconductor manufacturing. They differ in how the beam interacts with the sample, and that difference decides what each one can show, what the sample must look like and where in the fab it is used.
The short answer
A scanning electron microscope (SEM) scans a focused electron beam across a sample and builds an image from the electrons that come back out of the surface, so it can look at bulk samples and even whole wafers at roughly nanometre resolution. A transmission electron microscope (TEM) sends a high-energy beam through a lamella thinned to around 100 nm or less and images the transmitted electrons, which resolves lattice planes and, with aberration correction, individual atomic columns. In a fab, SEMs measure and review on the line; TEMs are used off-line for process development and failure analysis, at the cost of destroying the sample.
Side-by-side comparison
| Aspect | SEM | TEM |
|---|---|---|
| Imaging principle | A focused beam is rastered over the surface; secondary and backscattered electrons are collected point by point | A broad or focused beam passes through a thin specimen; transmitted and diffracted electrons are imaged by lenses |
| Information obtained | Surface topography, cross-section shape, compositional contrast | Internal structure projected through the thickness, lattice images, interfaces, crystal defects |
| Typical resolution | Around 1 nm, depending on instrument and conditions | Sub-nanometre lattice resolution; below 0.1 nm with aberration correction |
| Accelerating voltage | From a few hundred volts to about 30 kV | Typically 80-300 kV |
| Sample thickness | No limit; whole wafers can be loaded in in-line tools | Must be electron-transparent, usually 100 nm or less |
| Sample preparation | Often none for top-down imaging; cleaving or FIB milling for cross-sections | Thinning by FIB or other methods is mandatory and takes hours |
| Destructive? | Low-voltage in-line SEMs return the wafer to the process flow | Yes; the area examined is cut out of the wafer |
| Field of view | Can zoom continuously from wide areas to high magnification | Limited to the small region captured in the lamella |
| Main fab uses | CD-SEM linewidth metrology, defect review, cross-section imaging | Gate stack thickness and interfaces, GAA nanosheet checks, failure analysis |
| Common add-on analysis | EDX for elemental composition | STEM imaging, EDX and EELS for elemental and chemical mapping |
Surface signals versus transmitted electrons
In an SEM, the beam is focused to a spot a few nanometres across and scanned in a raster pattern. Where it lands, secondary electrons and backscattered electrons leave the sample, and the detector signal at each point becomes one pixel. Secondary electrons escape only from a very shallow depth, so the image is dominated by surface topography; backscattered electrons add contrast from atomic number.
In a TEM, the specimen is made thin enough for electrons to pass through. Electrons scattered and diffracted by atoms inside the specimen are focused by electromagnetic lenses, so the image shows the internal structure projected through the thickness rather than the surface. Under suitable conditions, rows of atoms appear directly as lattice fringes.
Why the production line uses SEM
The electron microscopes found in high-volume fabs are SEM-based. A CD-SEM measures the width of resist and etched patterns from above without breaking the wafer, usually at a low landing energy of around 1 kV or less to limit charging and beam damage to resist. A defect review SEM drives to the coordinates reported by an optical or e-beam inspection tool, images each defect at high magnification and classifies it.
After either measurement the wafer continues through the process. A TEM cannot play this role, because preparing a TEM sample means cutting a piece out of the wafer.
When only TEM will do
Once gate dielectrics and GAA nanosheets are only a few nanometres thick, the edges of each layer blur in an SEM image and their thickness cannot be measured reliably. The thickness of a high-k film and its interfacial layer, the fill of a metal gate, or stacking faults in an epitaxial layer all need atomic-scale resolution.
Scanning TEM (STEM) mode, where a fine probe is scanned across the lamella, can be combined with EDX and EELS so that structure and elemental distribution are mapped at the same location. For new process development, and for pinning down the physical cause of an electrical failure, TEM is usually the final confirmation step.
Sample preparation is the bottleneck
In practice, the limiting factor for TEM is not the microscope but the specimen. Cutting a lamella that contains a specific transistor with a focused ion beam (FIB) and thinning it to a few tens of nanometres takes a skilled operator several hours, and the ion beam can leave a damaged surface layer that distorts the result.
Labs therefore work in sequence: survey a wide area in the SEM, narrow down the anomaly, and then prepare a TEM lamella at the exact site. A dual-beam FIB-SEM combines both steps in one chamber, so the operator can watch the cut with the electron beam while milling with the ion beam.
FAQ
Can an SEM image cross-sections?
Yes. A wafer can be cleaved, or a cross-section can be milled with a FIB, and the exposed face imaged in the SEM. For films only a few nanometres thick, or for the atomic arrangement at an interface, the resolution of a TEM is still needed.
Is STEM closer to SEM or to TEM?
STEM uses the same thin lamella as TEM and detects transmitted electrons, but it forms the image by scanning a focused probe, as an SEM does. Its sample requirements and resolution are TEM-like, and point-by-point scanning makes it well suited to EDX and EELS mapping.
How thin must a TEM sample be?
Generally 100 nm or less, and a few tens of nanometres for high-resolution lattice imaging. Higher accelerating voltages penetrate thicker specimens, but thinner samples give sharper images with less overlap of features along the beam direction.
Which Japanese companies make electron microscopes?
Hitachi High-Tech and JEOL are the main Japanese makers of both SEMs and TEMs, and Hitachi High-Tech also supplies CD-SEMs for semiconductor lines. Thermo Fisher Scientific and Carl Zeiss are major non-Japanese suppliers.
Terms compared
- Scanning Electron Microscope (SEM)A microscope that scans a focused electron beam across a sample and builds an image from the emitted electrons, resolving features near 1 nm.
- Failure Analysis (FA)The work of finding the physical root cause of a failed chip, moving step by step from electrical tests to cross-sections under the microscope.
- CD-SEM (Critical Dimension SEM)A dedicated scanning electron microscope that measures the width and shape of circuit patterns on wafers with nanometre precision.
- Focused Ion Beam (FIB)A finely focused beam of ions, usually gallium, that mills, images or deposits material at chosen spots with nanometre precision.
Japanese companies involved
- JEOL · Electron beam lithography systems, electron microscopes and surface analysis
- Hitachi High-Tech · CD-SEM metrology, defect inspection and etch systems
- Hamamatsu Photonics · Failure analysis, process monitoring and static elimination equipment
- Holon · Photomask defect review, repair and CD-SEM metrology
More inspection and metrology comparisons
Sources
Semiconductor comparisons
- All semiconductor comparisons
- XPS vs EDX: what is the difference?
- Czochralski vs float zone: what is the difference?
- SOI vs bulk silicon: what is the difference?
- Epitaxial wafer vs polished wafer: what is the difference?
- Fabless vs foundry: what is the difference?
- Tokyo Electron vs Applied Materials: what is the difference?
- EUV vs DUV lithography: what is the difference?
Last updated: November 2, 2026