X-ray photoelectron spectroscopy (XPS): principle, spectra and analysis

XPS is based on the photoelectric effect. An X-ray photon of energy hν is absorbed by an atom and ejects a core electron; if hν exceeds the electron’s binding energy, the photoelectron leaves with a measurable kinetic energy. The Einstein equation links the three quantities, so measuring the kinetic energy with an electron analyser and knowing the photon energy gives the binding energy: BE = hν − KE (Baio et al. 2020). The most common laboratory source is monochromatic Al Kα at 1487 eV (Baio et al. 2020).

Each element has core levels at characteristic binding energies, so peak positions identify the elements present; all elements except hydrogen and helium are detected (Baio et al. 2020). The exact energy of a peak also depends on the atom’s chemical environment, which is what makes XPS a chemical-state technique.

Why it is surface sensitive. The X-rays penetrate microns into the sample, but the photoelectrons have kinetic energies of a few hundred eV and interact strongly with matter, so only those from near the surface escape without losing energy. The sampling depth is about 2–10 nm, depending on the emission angle and the material (Baio et al. 2020). It is usually taken as three inelastic mean free paths, the depth from which 95% of the photoelectrons come: 3–10 nm for Al Kα, and 8.7 nm, about 26 graphene layers, for graphite (Susi et al. 2015).

XPS spectra are plotted as counts against binding energy, with binding energy increasing from right to left (Baio et al. 2020).

• Survey (wide) scan. A scan over about 0–1100 eV shows which elements are present and how much of each; peak areas, corrected for the probability of creating and detecting each photoelectron, give the composition (Baio et al. 2020). Atomic percentages come from areas divided by relative sensitivity factors, never from raw peak areas (Kuzmyn et al. 2026).
• High-resolution (narrow) scans. A narrow scan of one core level, such as C 1s, resolves chemical species. Bonding carbon to more electronegative atoms raises its binding energy, so C 1s increases in the order C–C < C–O < C=O (Baio et al. 2020). In a poly(2-hydroxypropyl methacrylamide) brush, for example, C–C/C–H, C–N, C–O and C=O components appeared near 285.0, 286.1, 287.0 and 288.0 eV in the ratio expected from the repeat unit (Kuzmyn et al. 2026).
• Background. Inelastically scattered electrons add a stepped background on the high-binding-energy side of each peak; it is removed with a linear, Shirley or Tougaard background before fitting (Susi et al. 2015).
• Line shapes. Peaks are usually fitted with Voigt (Gaussian–Lorentzian) profiles; metallic systems such as graphite give an asymmetric Doniach–Šunjić C 1s line, with the graphite maximum at 284.42 eV (Susi et al. 2015).
• Doublets. Spin–orbit coupling splits levels such as P 2p into two components; for P 2p the predicted splitting is 0.87 eV, with the lower-binding-energy 2p₃/₂ component twice the area of 2p₁/₂ (Susi et al. 2015).

The background shape also hints at depth: atoms at the outermost surface give a background that rises little above the peak, while atoms buried deeper give a steadily rising background (Baio et al. 2020).

• Keep the surface clean. Any solvent exposure, even rinsing, can deposit contaminants and change the surface composition; plan sample handling as part of the experiment (Baio et al. 2020).
• Make it vacuum compatible. XPS runs in ultra-high vacuum, so samples must not outgas, and moving a sample from water into vacuum can rearrange its surface. Frozen-hydrated preparation can lock in a hydrated surface structure (Baio et al. 2020).
• Powders can be pressed into small cups; freeze-dried microbial cells, for example, form powders that are pressed and analysed this way (Wei et al. 2021).
• Control charging. Insulating samples charge positively during analysis, which shifts and broadens peaks; low-energy electron flood guns or dual-mode neutralisers compensate, and their settings should be reported (Kuzmyn et al. 2026).
• Watch for beam damage. Measure sensitive regions early, take several short scans and compare the first and last for changes in area, position or width (Kuzmyn et al. 2026).

Charge referencing to C 1s. On insulators the binding-energy scale is usually corrected by setting the lowest-energy C 1s component, from C–C/C–H carbon, to 285.0 eV; values from 284.8 to 285.0 eV are widely used (Kuzmyn et al. 2026). Treat this as an approximate correction: adventitious carbon is an unreliable reference because its apparent energy varies with substrate, surface condition and charging; a visible conductive substrate peak such as Au 4f or Si 2p can be a better reference; differential charging cannot be fixed by one rigid shift; and forcing C 1s to 285.0 eV can hide real chemical shifts. Report both raw and corrected binding energies when small shifts matter (Kuzmyn et al. 2026). For carbon nanomaterials, the C 1s position must be measured for each sample rather than assumed as a common reference, and the energy scale calibrated with a clean metal (Susi et al. 2015).

Depth profiling. Because the XPS signal decays exponentially with depth, a spectrum is a weighted average over the sampling depth, and extracting a composition profile from it alone needs modelling (Baio et al. 2020). Alternatively, the surface is removed step by step with an ion beam between measurements. For organic layers, argon gas cluster ion beams with about 500–2000 atoms per cluster are preferred, because monatomic Ar⁺ sputtering causes cross-linking, carbonisation and preferential sputtering; X-ray exposure itself can change sputter rates in some polymers (Kuzmyn et al. 2026). Thickness of thin coatings can also be estimated from how much they attenuate the substrate signal (Kuzmyn et al. 2026), and varying the excitation energy, including hard X-rays, probes coatings on core–shell nanoparticles (Radnik et al. 2022).

Doped graphene and carbon. XPS is one of the main tools for the concentration and bonding of nitrogen, boron and phosphorus dopants in graphene and carbon nanotubes (Susi et al. 2015). In N 1s spectra, a component near 398.3 ± 0.3 eV is commonly assigned to pyridinic nitrogen and one near 400.1 ± 0.3 eV to pyrrolic nitrogen, while values from about 400.2 to 401.8 eV have been assigned to substitutional (graphitic) nitrogen; the spread is large, so assignments need care (Susi et al. 2015). For oxidised carbons such as those made from graphene oxide, C 1s components above about 285.1 eV are attributed to carbon bonded to oxygen (Shulga et al. 2021).

Oxide electrodes and coatings. XPS gives the elemental composition and oxidation states of solid-oxide fuel-cell cathodes to a depth of about 5–10 nm and is used to follow cation enrichment at their surfaces; changes in the surface chemical state are linked to performance degradation (Anwar et al. 2022). For composition and thickness of nanoparticle coatings it can be combined with modelling of peak intensities (Radnik et al. 2022).

Thin organic films. XPS sees the outermost ~10 nm of organic films, so for polymer brushes thicker than about 20 nm the elemental ratios reflect the monomer composition (Kuzmyn et al. 2026).

XPS does not tell you the crystal structure: pair it with XRD, which averages over the bulk (Abid et al. 2023). Studies of carbon materials often combine XPS with Raman spectroscopy and electron energy-loss measurements, because the techniques give different estimates of the same quantity (Shulga et al. 2021). See also FTIR and ICP-MS, and the characterisation techniques hub for how they all compare.

• Quantifying from raw areas. Each core level has a different photoionisation cross-section and analyser transmission; compare sensitivity-factor-corrected areas and report the factor library used (Kuzmyn et al. 2026).
• Over-fitting peaks. Assigning binding energies to specific structures is rarely straightforward, and overlapping ranges have been reported for the same configurations (Susi et al. 2015). Symmetric fit functions on an asymmetric C 1s line overestimated the sp³ carbon content of a porous carbon (Shulga et al. 2021).
• Trusting adventitious carbon blindly. Its apparent binding energy depends on the substrate and charging (Kuzmyn et al. 2026).
• Ignoring the background. A careful background subtraction is vital for analysing the primary spectrum (Susi et al. 2015), and the same model should be used for all spectra being compared (Kuzmyn et al. 2026).
• Calling XPS a bulk measurement. It reports the top few nanometres; for thin materials such as graphitic nanomaterials (8.7 nm sampling depth in graphite) it can in effect probe the whole thickness (Susi et al. 2015).
• Overlooking contamination. Some contaminants are easier to see by other methods: ToF-SIMS is far more sensitive to PDMS than XPS (Baio et al. 2020).

A clear XPS methods figure shows the X-ray source, the sample in vacuum, the electron energy analyser and the detector, with the photoelectron path from the surface to the analyser. In LabCanvas the figure editor’s Library has the XPS instrument as an editable object, including a cutaway of the X-ray and photoelectron paths, and an editable data plot for survey and high-resolution spectra. You can reverse the binding-energy axis, label core-level peaks (C 1s, O 1s, N 1s), add fitted components with their assignments and export the figure for a methods section or graphical abstract.