How to read an EELS spectrum

Your foil, but as an itemized receipt.

Every electron that crosses your specimen either keeps all of its energy or pays some of it to the atoms on the way through. Electron energy-loss spectroscopy (EELS) lines the transmitted electrons up by exactly how much they paid. Learn to read that ledger and one spectrum will tell you what the material is, how its atoms are bonded, and even how thick the foil is.

1 · One beam in, an itemized receipt out

After the specimen, a magnetic prism bends the transmitted beam. Electrons that lost energy are travelling a touch slower, so the prism bends them a touch harder, spreading the beam into a spectrum: counts versus energy loss. Zero on that axis means "paid nothing." Everything to the right is a specific expense with a specific cause.

Here is the whole receipt for a thin titanium-oxide film. One habit to build immediately: view EELS on a log scale. The zero-loss peak out-counts a core-loss edge by a factor of a thousand or more, so on a linear axis the interesting chemistry is a flat line hugging zero.

2 · The zero-loss peak: electrons that paid nothing

The towering spike at 0 eV collects every electron that made it through elastically (plus those that only excited phonons, a few meV, too small to resolve on most instruments). Two useful readings hide in this "boring" peak. First, its width is your energy resolution: the electron source's energy spread plus spectrometer aberrations. A thermionic LaB₆ gun gives roughly 1 eV, a Schottky field emitter about 0.6–0.8 eV, a cold field emitter 0.3–0.4 eV, and a monochromated instrument can push below 10 meV, sharp enough to see phonons and molecular vibrations.

Second, its size relative to the rest of the spectrum is a thickness gauge: the thicker the foil, the fewer electrons escape unscathed. That single idea powers the log-ratio thickness method in the next section.

3 · Low loss: plasmons, and why thickness is written here

The bumps in the 5–50 eV range are mostly plasmons: an electron zips past and the material's whole valence-electron sea sloshes once in response, at a frequency set by the electron density. The quantum of that slosh is the plasmon energy E_p, about 15.0 eV in aluminum and 16.7 eV in silicon. It is a fingerprint of the electronic density, which is why plasmon shifts can flag alloying or hydride formation before anything else does.

Now the part most people meet first in practice. Scattering is random, so the number of plasmons each electron excites follows Poisson statistics: some electrons excite none, some one, some two. Their proportions depend on one number, the thickness in units of the inelastic mean free path, t/λ. Thicker foil, more double and triple plasmons, smaller zero-loss fraction. Drag the slider and watch the whole low-loss region redistribute:

Look at the last readout. The widget integrates its own zero-loss peak (I₀) and its whole spectrum (I_t), takes one logarithm, and recovers the t/λ you set on the slider. That is the log-ratio thickness method, the standard way to measure specimen thickness from EELS: t/λ = ln(It/I0). Convert λ to nanometers from the beam energy, collection angle, and material, and you have absolute thickness from a two-second acquisition.

One practical rule falls straight out of the Poisson picture: quantitative EELS wants t/λ below about 0.3–0.5. Beyond that, plural scattering buries the single-scattering signal that quantification needs, and beyond t/λ ≈ 2–3 even the zero-loss peak is an endangered species.

4 · Core loss: every element signs its name

Past ~100 eV the spectrum is a falling slope, punctuated by edges: sudden rises where the beam has enough energy to kick an electron out of a specific inner shell of a specific element. The onset energy is that shell's binding energy, so the edge names follow the shells: K for 1s (carbon at 284 eV, oxygen at 532 eV), L_2,3 for 2p (titanium at 456 eV), M_4,5 for 3d (lanthanum at 832 eV). Because binding energies are element-specific, an edge is an elemental signature; because each edge sits on the falling tail of everything before it, quantification always starts by fitting and subtracting a power-law background A·E−r just before the edge.

The edge's shape is a second, deeper signature. The first 30 eV or so, the energy-loss near-edge structure (ELNES), maps the empty electronic states the ejected electron can land in: the sharp "white lines" atop transition-metal L edges sharpen and shift with oxidation state (Ti³⁺ vs Ti⁴⁺, Fe²⁺ vs Fe³⁺), and the carbon K edge tells graphite from diamond from amorphous carbon at a glance. Chemistry, not just composition, from the same acquisition.

Planning an acquisition starts with knowing where your elements' edges sit. Toggle elements below; the lookup flags edges that land close enough to interfere:

5 · EELS or EDS: which one do you point at the problem?

The same inner-shell ionization feeds both spectroscopies. EELS catches the primary event (the beam electron's energy loss, measured with sub-eV precision); EDS catches the aftermath (the X-ray emitted as the atom relaxes, measured by a solid-state detector with roughly 125–130 eV resolution at Mn Kα). That factor of a hundred in energy resolution decides most arguments between them. The classic cautionary tale is titanium nitride: in EDS, the N Kα line (392 eV) and the Ti L lines (~452 eV) melt into one blob, while in EELS the N K edge (402 eV) and Ti L_2,3 edge (456 eV) stand politely apart with fine structure to spare.

To be fair to EDS: it is on every microscope, it shrugs at specimen thickness, it covers high-Z K lines far beyond any spectrometer's range, and for a quick "what is this precipitate" survey it is unbeatable. EELS asks more of you (a thin specimen, ideally t/λ < 0.5, and more careful setup) and pays you back with light-element sensitivity, bonding information, and, as above, a free thickness measurement. On modern instruments they run simultaneously, which is the correct amount of loyalty to either.

@misc{tripathy2026readingtheeelsspectrum,
  author = {Tripathy, Manisha},
  title  = {Reading the EELS Spectrum},
  year   = {2026},
  howpublished = {\url{https://untetheredatom.com/tem/eels-guide}},
  note   = {Interactive teaching resource}
}