X-ray Absorption Spectroscopy Delivers Powerful Answers Only When Handled Correctly

X-ray absorption spectroscopy has become one of the most widely used tools for probing the inner workings of materials, and a new open-access review in the journal Catal argues that its explosive growth has outpaced many researchers’ understanding of how to use it properly. The review, published on 20 August 2026 by a team led by Hao Zhang, Yalei Fan and Linjuan Zhang with colleagues including Zhiwei Hu and Jian-Qiang Wang, lays out a practical framework for sample preparation, data collection and analysis, warning that errors at any step can lead scientists to draw the wrong conclusions about catalytic mechanisms.

XAS works by tuning the energy of incoming X-rays to the absorption edge of a specific element. When a photon carries enough energy to eject a core-level electron, the absorption coefficient rises sharply, and the fine structure around that edge encodes the element’s oxidation state, spin state and local coordination environment, including bond distances and coordination numbers. Because the technique is element selective and non-destructive, it has become indispensable in condensed-matter physics, chemistry, materials science, biology and environmental science, and it is particularly valuable in catalysis, where identifying active sites and tracking their evolution under real reaction conditions is central to designing materials for clean-energy conversion.

The spread of synchrotron facilities worldwide and the arrival of benchtop laboratory instruments have opened XAS to a much broader research community, but the review’s authors caution that data are sometimes interpreted without sufficient attention to the underlying physics. One frequent and consequential mistake is treating Fourier-transformed R-space extended X-ray absorption fine structure spectra as direct pictures of atomic structure. In reality, the R-space spectrum is a pseudo-radial distribution function: phase effects shift peaks to lower distances, typically by 0.3 to 0.5 angstroms below true bond lengths, and peak amplitudes depend on the backscattering atom, structural disorder and the parameters chosen for the Fourier transform itself. Quantitative bond lengths and coordination numbers can only be obtained by least-squares fitting of the data to the theoretical EXAFS equation, and any analysis that reads values directly off R-space peak positions or heights is fundamentally flawed.

A second overlooked factor concerns catalyst particle size. Because hard X-rays penetrate micrometers to hundreds of micrometers while catalytic reactions often occur within a few nanometers of the surface, the measured spectrum is a volume-weighted average of the active surface shell and the largely unreacted core. The review illustrates the stakes with simple geometry: for a spherical particle 40 nanometers in diameter, a reaction depth of 10 nanometers corresponds to 88.1 percent of the particle volume, whereas a depth of 3 nanometers covers only 39.2 percent. Expressed differently, for a reactive layer 5 nanometers thick, the surface signal contributes 98 percent of the spectrum for a particle with a 10-nanometer radius, but only 58 percent at a 20-nanometer radius and a mere 14 percent at 50 nanometers. If particles are too large, changes at the active surface become nearly invisible, which can undermine inferred reaction mechanisms and explain discrepancies between studies of nominally similar catalysts. The authors recommend characterizing and reporting particle size distributions before experiments and using techniques such as ball milling to reduce particle dimensions.

The review distinguishes carefully between soft and hard XAS regimes. Soft XAS, spanning roughly 100 electronvolts to 2 kiloelectronvolts, accesses the K edges of light elements such as carbon, nitrogen and oxygen and the L edges of 3d transition metals. Its shallow core levels have narrow lifetime broadening, about 0.1 to 0.2 electronvolts for light-element K edges and 0.2 to 0.5 electronvolts for 3d metal L edges, giving it higher energy resolution and strong sensitivity to electronic structure. Its drawbacks are equally clear: air absorbs soft X-rays, so measurements require high or ultra-high vacuum, and the penetration depth is typically less than a micrometer below 1 kiloelectronvolt, making total electron yield detection, which probes only about 2 to 5 nanometers, highly surface sensitive while fluorescence yield reaches beyond 100 nanometers. Hard XAS, from about 2 kiloelectronvolts to more than 100 kiloelectronvolts, covers the K edges of most elements from phosphorus onward, penetrates deeply enough to study bulk properties, and permits measurements in air, liquids and specialized operando reaction cells, though with lower energy resolution.

Fluorescence detection brings its own trap: self-absorption. In concentrated or optically thick samples, the coupled attenuation of incident and emitted photons breaks the linear relationship between fluorescence signal and true absorption, compressing edge steps and white lines and attenuating EXAFS amplitudes. The severity depends on composition, concentration, density, thickness and geometry, and, notably, light elements are not immune; sulfur K-edge measurements in sulfide materials pose a major concern. Mathematical corrections pioneered by Tröger and refined by Booth and Bridges, now implemented in analysis packages such as Athena, can reduce distortion when sample parameters are precisely known, but the authors stress that experimental minimization through dilution, thinner samples or geometry changes remains the preferred strategy, with the inverse partial fluorescence yield technique offering a powerful alternative at 3d transition-metal L edges.

Valence and spin states demand particular care. At soft X-ray L edges, increasing a metal’s oxidation state by one unit generally shifts the spectrum to higher energy by one or more electronvolts, but the shift is not linear with valence because metal-ligand covalency screens higher oxidation states, making the change from trivalent to tetravalent smaller than from divalent to trivalent. O K-edge spectra shift the opposite way, toward lower energy, with pre-edge weight growing as valence rises. For hard X-ray K edges, the review recommends defining the edge energy at 70 to 80 percent of the normalized edge step rather than at the first-derivative maximum or main peak, because those lower-energy features are entangled with pre-edge transitions and local symmetry; at the 0.8 intensity point, manganese and cobalt edge energies shift systematically by more than 2 electronvolts per unit change in oxidation state. Spin states are best quantified through the branching ratio of L3 to L2 edge intensities, which runs roughly 2.2 to 3.5 for high-spin states and 1.2 to 2.0 for low-spin ones; in the cobalt example highlighted, low-spin Co3+ yields a ratio of 1.8 while high-spin Co3+ reaches 2.6. The authors also warn against conflating spin state transitions, in which total spin changes at constant oxidation state, with valence changes, a confusion that has led to incorrect mechanistic interpretations in the literature.

On the quantitative side, the review details the pitfalls of EXAFS fitting. Coordination number uncertainties of 10 to 20 percent are typical even for high-quality data because coordination number couples strongly with the Debye-Waller disorder factor and with the inverse-square distance decay of peak amplitude. The threshold energy E0 should be treated as a refinable parameter with a single shared energy shift rather than fixed to a spectral feature, since an incorrect E0 shifts the phase of the oscillations and biases bond distances. Overfitting is bounded by information theory: the number of independent variables is limited to approximately twice the product of the k-space and R-space ranges divided by pi, plus two. Collinear atomic arrangements produce focusing effects that amplify multiple-scattering signals, anharmonic vibrations at high temperatures require cumulant corrections, and wavelet transform analysis, which resolved nickel and molybdenum backscatterers at identical radial distances in a NiMo alloy, is a diagnostic tool that still cannot replace rigorous fitting for extracting coordination numbers or bond lengths. For massive time-resolved datasets, principal component analysis and multivariate curve resolution can isolate transient intermediate states that bulk measurements obscure.

The review closes with laboratory XAS, which has matured to the point that a copper foil K-edge spectrum acquired in 60 minutes on a commercial instrument was virtually indistinguishable from synchrotron data, and the three Ti K-edge pre-edge features of TiO2 were clearly resolved after an hour of acquisition. Laboratory instruments trade photon flux for resolution, requiring longer dwell times, and their larger, more divergent beams place stringent demands on sample homogeneity. Detector dead time, following either paralyzable or non-paralyzable models, can distort white-line intensities and edge steps when count rates exceed the validated linear range. The overarching message is that sample homogeneity, detection mode, energy calibration, normalization, self-absorption control, dead-time management and statistically sound fitting must be treated as one integrated workflow; neglecting any link can introduce substantial artifacts and lead the entire mechanistic story astray.

Subject of Research: Best practices and common pitfalls in the experimental design and interpretation of synchrotron and laboratory X-ray absorption spectroscopy for catalysis research

Article Title: Precise analysis and proper application of X-ray absorption spectroscopy

Article References: Zhang, H., Fan, Y., Zhu, J., Yu, H., Wang, J.-Q., Hu, Z., & Zhang, L. (2026). Precise analysis and proper application of X-ray absorption spectroscopy. Catal, 2(1), Article 18. https://doi.org/10.1007/s44422-026-00030-0

Image Credits: AI Generated

DOI: 10.1007/s44422-026-00030-0

Keywords: X-ray absorption spectroscopy, XANES, EXAFS, catalysis, synchrotron radiation, oxidation state, spin state, self-absorption, operando spectroscopy, particle size effects, laboratory XAS, wavelet transform