TEM vs SEM for Nanomaterials: Which Microscope Answers Your Question

The question, framed properly

"Which is better, TEM or SEM?" has no answer, in the same way that "which is better, a ruler or a thermometer?" has no answer. They measure different things. The useful question is which one answers the question in front of you, and that is usually decided in one sentence.

Scanning electron microscopy looks at a surface. Transmission electron microscopy looks through a thin specimen. Almost every practical difference — resolution, sample preparation, field of view, cost, turnaround — follows from that.

What each instrument does

SEM

A focused electron beam is scanned across the specimen in a raster. At each point, the beam interacts with the material and generates signals that are collected point by point to build an image. The two principal ones are secondary electrons, which are low-energy, escape only from very near the surface, and give topographic contrast; and backscattered electrons, which come from deeper, and whose yield rises with atomic number, giving compositional contrast. Characteristic X-rays are generated at the same time, which is why energy-dispersive X-ray analysis is a standard SEM accessory.

The consequence of the scanned geometry is a large depth of field. An SEM image of a rough powder looks three-dimensional and is intuitive to read, which is a genuine practical advantage over a projection image.

TEM

A broad, high-energy beam passes through a specimen thin enough to transmit it. Contrast arises from how much the beam is scattered or diffracted on the way through, and the transmitted electrons are focused by electromagnetic lenses into an image or a diffraction pattern. Because the beam interacts with the internal structure rather than the surface, TEM sees inside particles: cores and shells, lattice fringes, internal voids, wall counts, and crystal defects.

The site already carries long-form guides to both — the TEM guide and the SEM guide — and this page is the one that joins them.

Side by side

Property	SEM	TEM
What it images	Surface	Internal structure, in projection
Specimen thickness	Bulk; thickness irrelevant	Must be electron-transparent
Typical resolution	Around a nanometre at best on favourable samples	Sub-Ångström with aberration correction; around 0.2 nm without
Depth of field	Large — images look three-dimensional	Small — a flat projection
Field of view	Millimetres down to nanometres	Micrometres down to atoms
Crystal structure	By electron backscatter diffraction, on prepared flat surfaces	By electron diffraction, routinely
Elemental analysis	EDX, micron-scale interaction volume	EDX and EELS, near-atomic in STEM mode
Sample preparation	Minutes; mounting and often a conductive coating	Hours to days for anything not already a fine dispersion
Throughput	High	Low
Access	Widely available	Usually a shared facility or a service

Resolution, honestly

Quoted resolution figures are the least useful part of any instrument comparison, because they are measured on ideal specimens under ideal conditions. In practice:

• A modern field-emission SEM will resolve features around a nanometre on a conductive, flat, clean sample at high accelerating voltage. On an insulating nanopowder at low voltage — which is what you will actually be looking at — a few nanometres is a more honest expectation.
• A conventional 200 kV TEM reaches roughly 0.2 nm point resolution, which is enough to image lattice fringes in most inorganic materials. Aberration correction pushes below 0.1 nm.
• For a 10 nm particle, the SEM tells you it is there and roughly how big; the TEM tells you its lattice spacing, whether it is a single crystal, and whether it has a shell.

Note also that the resolution limit is often not the instrument. A conductive coating applied to stop charging is itself several nanometres thick and is deposited as grains, so on a 20 nm particle the coating is a substantial fraction of what you are measuring — and it can be misread as a shell.

Which answers which question

• "How big are my particles, and what is the distribution?" Either, for particles above roughly 20 nm. TEM below that, because the SEM's own resolution becomes comparable to the feature.
• "Is my powder agglomerated, and how badly?" SEM. The depth of field and the wide field of view make agglomerate structure legible in a way a projection does not.
• "How thick is the shell on my core-shell particles?" TEM. Nothing else measures it directly.
• "How many walls does this multi-walled nanotube have?" TEM, imaging the walls edge-on.
• "How many layers is this graphene flake?" TEM at a folded edge, or AFM, or Raman. SEM will not tell you.
• "Is my thin film continuous, and how rough?" SEM, in cross-section if you need thickness. AFM for quantitative roughness.
• "What phase is this crystal?" TEM by selected-area electron diffraction, or XRD for a bulk answer over far more material — see the XRD guide.
• "Are these bimetallic particles alloyed or core-shell?" STEM with EDX mapping. An SEM's EDX cannot resolve it, because the X-ray generation volume is far larger than the particle.
• "Where did this contamination come from?" SEM with EDX. Fast, wide field, and adequate for identifying a foreign particle.
• "What is the aspect ratio distribution of my nanorods?" TEM for small rods, SEM for large ones, and count enough of them either way.

Sample preparation is the real difference

For SEM

Mount on a stub with conductive tape or paint, and provide a conductive path to ground. Non-conductive samples charge, which produces bright drifting artefacts and distortion, and the usual remedies are a sputtered coating of a few nanometres of a metal or carbon, or working at low accelerating voltage with a variable-pressure or environmental mode. Coating costs you fine surface detail and adds the thickness discussed above; low-voltage imaging costs you resolution. Both are trade-offs against the same problem.

For a nanopowder, dispersing it well and depositing a thin, separated layer matters more than anything else. A heaped pile of powder produces an image of a heaped pile of powder.

For TEM

The specimen must be electron-transparent, typically well under a hundred nanometres and thinner still for lattice imaging. For colloidal nanoparticles this is easy: dilute and dry a droplet on a coated grid. For anything else it is work — focused ion beam milling for a site-specific cross-section of a device or an interface, ultramicrotomy for polymer composites and embedded biological material, electropolishing for metals. Beam-sensitive and hydrated samples need cryogenic preparation.

The drop-cast step deserves its own warning, because it produces the most commonly misreported artefact in nanoparticle work: as the droplet dries, capillary forces drag particles together, so clustering on a TEM grid is not evidence of aggregation in suspension. If the question is aggregation state in the liquid, the honest measurements are light scattering or tracking analysis — see the sizing comparison — with microscopy used for the primary particle size.

STEM, where the distinction blurs

Scanning transmission electron microscopy focuses the beam to a fine probe and scans it across a thin specimen, collecting transmitted electrons on annular detectors. It is a scanned technique like SEM and a transmission technique like TEM. High-angle annular dark-field imaging in this mode gives contrast that rises steeply with atomic number, so heavy atoms appear bright against a light support — which is how single heavy atoms on a catalyst support are imaged, and how alloy and core-shell structures are distinguished. Combined with EDX or electron energy-loss spectroscopy it produces elemental maps at close to atomic resolution.

Many modern SEMs also offer a transmission detector, which extends them usefully towards TEM territory for thin specimens without buying a TEM.

Artefacts to expect from both

• Charging — bright streaks, image drift, apparent motion. An SEM problem primarily, and the reason for coating.
• Beam damage — polymers, organics, biological material, some perovskites and halides degrade under the beam. Low dose, lower voltage, or cryogenic conditions.
• Contamination — hydrocarbon builds up under a stationary beam, producing a dark square where you were focusing. Plasma cleaning helps.
• Sampling bias — the imaged field is a vanishing fraction of the sample, and there is a natural tendency to photograph the interesting region. Count from several areas, chosen before looking closely.
• Drying artefacts — as above, and applicable to both instruments whenever the sample arrived as a liquid.

Access and cost

SEM is widely available: many university departments and industrial laboratories have one, sessions are short, and a routine imaging job can be turned around quickly. TEM is a larger, more demanding instrument that generally lives in a shared facility with dedicated staff, and aberration-corrected instruments more so. For most buyers the practical route is a commercial characterisation service or an institutional facility, and the cost that matters is not the instrument's but the preparation time — a drop-cast nanoparticle grid is cheap, a FIB cross-section is not.

What neither of them gives you

Both measure dried specimens in vacuum, over a tiny sampled volume. So neither reports hydrodynamic size in suspension, which is light scattering or tracking analysis; neither reports a bulk-averaged crystallite size or phase fraction, which is X-ray diffraction; neither reports specific surface area, which is gas adsorption; and neither reports surface chemistry, which is XPS. Electron microscopy is the technique that shows you what the material looks like, and it should be read alongside the ones that tell you what it is.

Related guides

• The whole decision, technique by technique: the characterisation workflow.
• The long-form guides this page joins: TEM and SEM.
• Sizing in suspension, which neither microscope does: DLS, NTA or TEM.