EBSD and SEM · X-ray analysis

EDS in the SEM: what it can and cannot tell you

EDS (energy-dispersive X-ray spectroscopy) sorts the X-rays the beam knocks out of your sample by energy. Start by dragging the beam voltage in the first picture.

1 Spatial resolution 2 Detection limit 3 Which elements Takeaways Questions References

What is the spatial resolution of EDS in the SEM?

The beam enters at one spot, but its electrons spread, and X-rays come from the whole red region.

Left: side cut, to scale. Dashed: electron range. Red: where the line is made. Blue: your particle or film; dark red is its share of the signal. Right: both ranges against voltage.
Feature at the surface
Electron range (Kanaya-Okayama)
X-ray range (Anderson-Hasler)
Overvoltage U = E0/Ec
Signal from the feature

Try it: Fe K in iron at 20 kV comes from a region 1.03 µm deep. Drop to 10 kV: it shrinks to 170 nm, but U falls to 1.41 and the line is only weakly excited.

What is the detection limit of EDS?

A peak counts as found when its net counts P beat 3 times the counting noise of the background under it, 3√B.

Top: model spectrum with counting noise, counts per 10 eV. Bottom: zoom on the tested line. Grey dashed: background. Blue dashed: other elements' peaks. Red: the tested element. Shaded: counting window. Peak heights and detection limits come from a simple model: they are illustrative, not measured.
Sample
Counts axis
Line used
Peak width (FWHM)
Net peak counts P
Background B (3√B)
Model detection limit 3√B·C/P
Counts in the spectrum

Accuracy once found: about ±5% relative with standards on a flat polished sample, about ±30% standardless (95% of analyses, Newbury and Ritchie 2015).

Try it: iron with 0.1 wt% Cu at 60 s: not detected, the model's limit is 0.14 wt%. Set the live time to 300 s: the limit drops to 0.062 wt% and the copper is found.

Which elements can EDS detect?

Colour shows which line EDS can use at your voltage. Click an element to see its lines and overlaps.

K line, U ≥ 2L or M line, U ≥ 2weak, U 1.1 to 2no line excitedlight element, hardno X-ray linesMα energy not listedStrip: the chosen line at the detector's peak width; ticks are other elements' lines.
Line to use at this voltage
Overvoltage U
LineEnergy (keV)Needs E0 above (edge, keV)
At5 kV15 kV20 kV

Try it: click S: Mo Lα sits 15 eV away and Pb Mα 38 eV away, both far inside the 90 eV peak width. Then set 5 kV and click Fe: it switches from the K line to the L line.

What to take away

EDS sees about a micrometreFe K X-rays in iron at 20 kV come from a region about 1.0 µm deep and wide (Anderson-Hasler range). A smaller particle mixes its signal with the matrix.
Limits near 0.1 wt% need many countsNewbury and Ritchie (2015) report limits below 0.1 wt% (1000 ppm) in 500 s. The limit falls with the square root of the counts: four times the live time halves it.
Light elements and overlaps go wrongH and He give no characteristic X-ray lines. Li to F give weak, low-energy lines. S Kα and Mo Lα (15 eV apart) merge into one peak. Peak fitting on a spectrum with many counts can still split them (Newbury and Ritchie 2015).
More detail: the equations, the numbers and the limits

Electron range. Kanaya and Okayama: RKO (µm) = 0.0276 A E01.67 / (Z0.89 ρ), E0 in keV, A in g/mol, ρ in g/cm3 (as written by Newbury 2002). A and ρ come from the NIST tables of Hubbell and Seltzer.

X-ray range. Anderson and Hasler: Rx (µm) = 0.064 (E01.68 − Ec1.68) / ρ (Friel and Lyman 2006). Goldstein et al. (2018) and the SEM imaging page use the Kanaya-Okayama form with Ec instead. For Fe K at 20 kV it gives 1.32 µm against 1.03 µm. Both are estimates.

Overvoltage. U = E0/Ec. Peak intensity grows as (U − 1)n, n = 1.3 to 1.7, so it vanishes as U nears 1. U = 2 is the usual choice (Newbury 2002; Goldstein et al. 2018: E0 at least twice Ec). At low voltage U down to 1.1 has to be accepted (Newbury 2002).

Feature fraction. The page assumes X-rays are made evenly in a ball of diameter Rx touching the surface under the beam, and that the feature has the matrix's density. A particle touching the same point then gives (d/Rx)3 of the signal; a film gives the cap above depth t. Real production is not even (Goldstein et al. 2018 use Monte Carlo simulation for the full picture), so these fractions are rough.

Peak width. FWHM (full width at half maximum) is 125 to 129 eV at Mn Kα and about 60 eV at C K (Newbury 2002). The page joins 127 and 60 eV with FWHM2 = a + bE, the form set by counting statistics in the detector.

Spectrum model (illustrative). Background: Kramers' law, Z(E0 − E)/E. Peaks: (U − 1)1.5 times a rough fluorescence yield, Z4/(a + Z4). Line ratios and the detector window are rough guesses, not table values. Absorption in the sample uses a simple Z4/E2.6 formula scaled to the NIST value for Fe Kα in iron, 71 cm2/g (Hubbell and Seltzer). From 1 to 20 keV it stays within about a factor of 2 of the NIST table for these samples. The peak-to-background scale was picked so Cu in iron has a limit just under 0.05 wt% with 107 counts, below the 0.1 wt% in 500 s of Newbury and Ritchie (2015). So every limit and peak height here is illustrative, not measured. The 1/√(counts) trend and the line positions are not model guesses.

Detection test. Found if P > 3√B (Fialin et al. 1999, eq. 2, for a well known background), so CDL = 3√B · C/P. If the background is measured separately for the same time, their eq. 3 uses 3√(2B). The counting window is 1.2 FWHM wide: for a Gaussian peak on a flat background this width gives the largest P/√B. B includes other elements' peaks in the window. Left out: the extra error of separating overlapped peaks, escape and sum peaks, and Mβ lines. Line energies: Bearden (1967). Edges: Bearden and Burr (1967), with some values the X-ray Data Booklet takes from photoemission data.

Questions people ask

Can EDS detect lithium?

Not in routine work. The Li K line is at 54 eV, the lowest K line of any element, and very weak. Hovington et al. (2016) did detect Li K in five lithium compounds with a newly designed EDS system. Their calculations show the emission from these compounds should be at least 12 times weaker than from lithium metal.

Can EDS detect hydrogen?

No. Hydrogen has 1 electron and helium 2, all in the innermost shell. A characteristic X-ray line needs an electron from an outer shell to drop into an inner hole, so these two give no lines. Newbury and Ritchie (2015) list EDS as detecting every element except H and He.

Is EDS quantitative?

Yes, done carefully. For WDS with standards, about 95% of results fall within ±5% relative. Newbury and Ritchie (2015) found SDD-EDS with standards on flat, polished samples well inside that range. Standardless analysis needs about ±30% to cover 95% of results. Rough surfaces can be off by a factor of ten or more (Newbury and Ritchie 2013).

What is the detection limit of EDS?

Newbury and Ritchie (2015) report limits below 0.1 wt% (1000 ppm) in 500 s with an SDD. The limit falls as 1/√(counts), so short spectra give higher limits. In the illustrative model of widget 2, Cu in iron goes from 0.14 wt% at 60 s to 0.048 wt% at 500 s. For a few ppm, WDS is used (Fialin et al. 1999: 6 to 8 ppm).

Why does my EDS not add up to 100 percent?

With standards, a good analysis totals 98 to 102% (Newbury and Ritchie 2015). A low total can mean an element the analysis missed. In their example, leaving out oxygen gives a total of 70 to 80%. Standardless software forces the total to 100%, which hides this.

What kV should I use for EDS?

About twice the edge energy of the highest line you need (U = 2, Newbury 2002). Fe K has its edge at 7.112 keV, so use about 15 kV. At 20 kV every K line up to zinc (edge 9.659 keV) has U above 2. Lower voltage shrinks the X-ray region, but heavier elements then need their L or M lines.

EDS vs WDS: which is better?

EDS records all energies at once and is fast, but its peaks are about 127 eV wide at Mn Kα (Newbury 2002). WDS peaks are 5 to 20 eV wide (Friel and Lyman 2006), so WDS splits S from Mo. With 500 nA and about 15 minutes, WDS reached 6 to 8 ppm for Cr and Ni (Fialin et al. 1999). So EDS is faster; WDS is better for close overlaps and traces.

Why does EDS show carbon on my sample?

Often it is not from the sample. A carbon coat against charging adds C. Newbury (2002) notes carbon is often present as a contaminant. The C K line at 277 eV is also weak and strongly absorbed, so a small carbon number is hard to trust.

Related: SEM imaging basics (interaction volume) · EDS quantification in the TEM (thin foils) · TKD at the nanoscale · Reading an EELS spectrum · EBSD sample preparation · Electron beam contamination (where the carbon comes from) · Magnification vs resolution · Atom's research guide

References

Show the 15 references
  1. J. I. Goldstein, D. E. Newbury, J. R. Michael, N. W. M. Ritchie, J. H. J. Scott and D. C. Joy, Scanning Electron Microscopy and X-Ray Microanalysis, 4th ed., Springer, New York (2018). doi:10.1007/978-1-4939-6676-9
  2. K. Kanaya and S. Okayama, Penetration and energy-loss theory of electrons in solid targets, J. Phys. D: Appl. Phys. 5, 43 to 58 (1972). doi:10.1088/0022-3727/5/1/308
  3. C. A. Anderson and M. F. Hasler, Extension of electron microprobe techniques to biochemistry by the use of long wavelength X-rays, in R. Castaing, P. Deschamps and J. Philibert (eds), Proc. 4th Int. Conf. on X-ray Optics and Microanalysis, Hermann, Paris (1966), pp. 310 to 327.
  4. J. J. Friel and C. E. Lyman, Tutorial review: X-ray mapping in electron-beam instruments, Microsc. Microanal. 12, 2 to 25 (2006). doi:10.1017/S1431927606060211.
  5. D. E. Newbury, Barriers to quantitative electron probe X-ray microanalysis for low voltage scanning electron microscopy, J. Res. Natl. Inst. Stand. Technol. 107, 605 to 619 (2002). doi:10.6028/jres.107.049.
  6. D. E. Newbury and N. W. M. Ritchie, Is scanning electron microscopy/energy dispersive X-ray spectrometry (SEM/EDS) quantitative?, Scanning 35, 141 to 168 (2013). doi:10.1002/sca.21041
  7. D. E. Newbury and N. W. M. Ritchie, Performing elemental microanalysis with high accuracy and high precision by scanning electron microscopy/silicon drift detector energy-dispersive X-ray spectrometry (SEM/SDD-EDS), J. Mater. Sci. 50, 493 to 518 (2015). doi:10.1007/s10853-014-8685-2.
  8. J. A. Bearden, X-ray wavelengths, Rev. Mod. Phys. 39, 78 to 124 (1967). doi:10.1103/RevModPhys.39.78
  9. J. A. Bearden and A. F. Burr, Reevaluation of X-ray atomic energy levels, Rev. Mod. Phys. 39, 125 to 142 (1967). doi:10.1103/RevModPhys.39.125
  10. X-Ray Data Booklet, Center for X-ray Optics and Advanced Light Source, Lawrence Berkeley National Laboratory (October 2009): Tables 1-1 (edges, G. P. Williams) and 1-2 (lines, J. B. Kortright and A. C. Thompson). Source of every line and edge energy here. Table 1-1 takes some edges from M. Cardona and L. Ley (eds), Photoemission in Solids I (Springer, 1978) and J. C. Fuggle and N. Mårtensson, J. Electron Spectrosc. Relat. Phenom. 21, 275 (1980).
  11. H. A. Kramers, On the theory of X-ray absorption and of the continuous X-ray spectrum, Phil. Mag. 46, 836 to 871 (1923). doi:10.1080/14786442308565244
  12. M. Fialin, H. Remy, C. Richard and C. Wagner, Trace element analysis with the electron microprobe: new data and perspectives, Am. Mineral. 84, 70 to 77 (1999). doi:10.2138/am-1999-1-207.
  13. P. Hovington, V. Timoshevskii, S. Burgess, H. Demers, P. Statham, R. Gauvin and K. Zaghib, Can we detect Li K X-ray in lithium compounds using energy dispersive spectroscopy?, Scanning 38, 571 to 578 (2016). doi:10.1002/sca.21302
  14. NIST Standard Reference Material 1155a, Stainless Steel (Cr 18-Ni 12-Mo 2), AISI 316 (certificate revised 11 August 2014).
  15. J. H. Hubbell and S. M. Seltzer, X-Ray Mass Attenuation Coefficients, NIST Standard Reference Database 126, NISTIR 5632 (data last updated 2004), Table 1 (Z/A, density) and Table 3 (iron, lead). doi:10.18434/T4D01F
Cite this page: Tripathy, Manisha. “EDS in the SEM.” untethered atom, 2026, https://untetheredatom.com/ebsd/eds-in-the-sem.
BibTeX 
@misc{tripathy2026edsinthesem,
  author = {Tripathy, Manisha},
  title  = {EDS in the SEM},
  year   = {2026},
  howpublished = {\url{https://untetheredatom.com/ebsd/eds-in-the-sem}},
  note   = {Interactive web tool}
}
Last updated 3 November 2026.