X-ray fluorescence tomography combines element-sensitive photon measurements with multi-angle scanning to estimate where elements are distributed inside a specimen. The resulting three-dimensional elemental volumes are reconstructions from spectra, geometry and attenuation assumptions—not direct photographs of individual atoms.
The gold standard retains photon counts, energy calibration, scanning geometry, correction algorithms and independent reference tests so elemental concentration and spatial resolution claims can be challenged.
eduKateSG · X-ray Fluorescence Tomography
Find your next learning step
Connect measured X-ray emission to three-dimensional coordinates, work a counting example and test corrections for fluorescence self-absorption.
Full chapter index · Practice and transfer · Education root · How X Works
Full chapter index
Define elemental tomography · Chapters 1–2
Calculate the measurements · Chapters 3–4
Understand reconstruction · Chapters 5–7
Validate three dimensions · Chapters 8–10
Continue X-ray reading · Chapters 11–12
CHAPTER 1 OF 12 · Define elemental tomography
1. Tomography adds depth to elemental fluorescence maps
X-ray fluorescence (XRF) detects characteristic photons emitted when a specimen is excited by incident X-rays. A two-dimensional scan can show where an element-related signal is strongest on a projected region, but several depths along the illuminated path may contribute to the same recorded location.
X-ray fluorescence tomography collects suitably varied views, commonly using translations and rotations, to estimate a three-dimensional elemental distribution. The reconstructed value is derived from measured photon spectra and a physical forward model, not a direct camera photograph of an interior atom.
Research at Argonne’s Advanced Photon Source describes XRF tomography as a way to study trace-metal distributions and highlights the importance of hyperspectral treatment and reconstruction. The method has research applications in biological, geological and materials specimens.
The gold standard names the excitation energy, detector and spectral window, sample, scan geometry, rotation coverage, attenuation model and comparison standards before presenting voxel values as concentrations.
A colorful volume does not by itself certify absolute elemental concentration, phase, oxidation state or biological function. Different techniques can address those distinct questions, and their evidence must be kept separate.
CHAPTER 2 OF 12 · Define elemental tomography
2. What the detector records and what reconstruction infers
The incident beam excites atoms in the specimen. Depending on their element and excitation conditions, some atoms emit characteristic fluorescence photons, which may escape toward an energy-resolving detector.
The observed spectrum includes element-related features, detector response, scattered X-rays and background. Fluorescence counts depend not only on the elemental abundance but also on incident beam intensity, sample geometry, absorption, detector solid angle and acquisition time.
For a thin and weakly attenuating sample, integrating a calibrated fluorescence signal along the beam direction can sometimes be approximated by a simple projected elemental distribution. Thicker and more absorbing samples demand a more complete treatment.
A tomography model connects those line-of-sight signals across multiple rotations to a voxel map. It must also account for the path travelled by excitation and emitted fluorescence, which are generally at different photon energies.
Different elemental emission lines have different attenuation and overlap problems. Reporting a single generic detector count without defining the selected energy line is inadequate for reliable quantification.
CHAPTER 3 OF 12 · Calculate the measurements
3. Worked example: counting uncertainty for a fluorescence peak
Imagine an idealized single spectral feature with 400 accepted signal photons after an appropriately characterized background model. Under a simplified Poisson count approximation, the standard deviation of the raw signal count is approximately √400 = 20 photons.
The corresponding fractional counting uncertainty is 20/400 = 0.05, or 5%. For a second imaginary feature with 100 counts, √100 = 10 and relative uncertainty is 10%, demonstrating the difficulty of weaker signals.
These calculations apply only to the simplified count model. Subtracted background, peak overlap, electronic noise, dead time, pile-up and calibration uncertainty can make actual concentration uncertainty substantially different.
Acquiring more counts can improve counting precision but may increase exposure time or risk sample change. A low-noise display that has been heavily smoothed should not be confused with independently measured fine structure.
A scientific report needs its photon-counting model, spectrum fit, energy calibration and background treatment before discussing detection limits or the significance of a faint voxel.
CHAPTER 4 OF 12 · Calculate the measurements
4. Worked example: rotation and translation sampling workload
Consider a hypothetical pencil-beam experiment taking 100 translation positions at each of 90 rotation angles, with one spectrum collected at each position and angle. The nominal number of spectral acquisitions is 100×90 = 9,000.
If every acquisition used a fictional dwell of 0.10 seconds, ideal photon-collection time would be 9,000×0.10 = 900 seconds, or 15 minutes. Movement, readout, calibration, scan stabilization and repeat frames are excluded.
Doubling angular sampling from 90 to 180 angles under unchanged spatial positions and exposure increases ideal collection time to 30 minutes. This simple count shows why fluorescence tomography can require substantial measurement time.
More angles do not automatically improve resolution if stage drift, photon statistics or spectral saturation dominate. Conversely, too few directions can create directional artefacts and make certain structures difficult to distinguish.
Scan design must match spatial scale, specimen stability and the research question. Argonne has researched hyperspectral reconstruction methods because efficient data acquisition and correct use of full spectra can both matter.
CHAPTER 5 OF 12 · Understand reconstruction
5. Fluorescence tomography is not identical to transmission CT
In transmission computed tomography, the detector measures primary X-rays after they pass through the specimen along a known ray. A logarithmic transformation can approximate attenuation path integrals under suitable beam assumptions.
In XRF tomography, the sample itself emits secondary characteristic photons. They may leave in directions different from the incident beam, and their probability of detection depends on both excitation and escape pathways.
This changes the forward model. Applying an ordinary transmission-CT reconstruction without accounting for fluorescence geometry and absorption can produce false elemental concentrations or biased depths, especially in heterogeneous specimens.
A 2017 quantitative XFCT study explains using iterative correction to address self-absorption in low-atomic-number samples, including absorption arising from material beyond the selected layer.
Some experimental setups combine XRF with transmission or phase-contrast tomography to provide structural context. Matching spatial coordinate systems and measurement times remains necessary to make a joint interpretation.
CHAPTER 6 OF 12 · Understand reconstruction
6. Worked example: how attenuation distorts an emitted signal
For a simplified monochromatic path with linear attenuation coefficient μ and length L, the unscattered transmission fraction is T = exp(−μL). This is an ideal physical relation; actual samples may be heterogeneous and energy dependent.
Imagine excitation travels through 5 mm of material with μin = 0.10 mm⁻¹. Its ideal transmission is exp(−0.10×5) = exp(−0.5) ≈ 0.607.
Suppose a fluorescence photon then travels an additional 5 mm to the detector with an illustrative coefficient μout = 0.10 mm⁻¹. The escape probability in this stylized path model is also approximately 0.607.
The combined multiplicative transmission effect is exp(−0.5)×exp(−0.5) = exp(−1) ≈ 0.368. In this ideal example only about 36.8% of the unattenuated path-weighted signal remains.
This number is not a universal instrument efficiency or detector count fraction. Emission location, photon energy, angle, local matrix and fluorescence generation all matter. The lesson is that identical elemental concentrations at different depths can yield different counts.
Quantitative reconstruction may therefore need an attenuation map, iterative model or calibrated correction rather than treating every detected photon as though it travelled through equal material.
CHAPTER 7 OF 12 · Understand reconstruction
7. Spectral fitting and peak overlap influence every voxel
Energy-dispersive detectors distinguish photon energies with finite resolution. Elemental peaks can overlap or sit on a continuum, and fluorescence generated by more than one element can contribute to a given energy interval.
A raw integration over a chosen window may therefore include unrelated counts. Full-spectrum fitting can help separate contributions when the line shapes, background and detector response are properly characterized.
Argonne’s hyperspectral XRF-tomography work investigates reconstructing rich spectral information rather than discarding it prematurely. The value depends on correct modelling and computational control, not just having a larger data file.
Dead-time effects and pile-up can become important at high count rates, while a beam-stop or geometry change can alter measurement coverage. Preserve quality indicators and avoid interpreting fitted negative or implausible concentrations as direct measurements.
A good result includes calibration and line-identification records as well as spatial positions. It distinguishes measured photon counts, fitted line intensity and reconstructed elemental concentration.
CHAPTER 8 OF 12 · Validate three dimensions
8. Self-absorption correction can be an inverse problem
Self-absorption means fluorescence generated inside the specimen can be attenuated before exiting it. Incoming excitation is also attenuated. The combined effect creates position- and angle-dependent weighting that can masquerade as an uneven concentration distribution.
An iterative approach may start with an approximate elemental map, estimate the resulting attenuation along paths and update the map using differences between predicted and measured spectra. Such algorithms can be useful but depend on material assumptions and calibration.
Research published in AIP Advances on quantitative XFCT demonstrates iterative absorption correction for low-Z specimens. This is evidence for its studied geometry and materials, not an unrestricted guarantee for any thick metal-rich sample.
When fluorescence and attenuation depend on unknown components together, a reconstruction can be nonunique. An improved data fit does not necessarily resolve a missing material phase or a wrong attenuation map.
Test alternative plausible attenuation assumptions, compare with phantom standards and report remaining bias. A visually smooth reconstruction after correction is not by itself proof of quantitative truth.
CHAPTER 9 OF 12 · Validate three dimensions
9. Geometry, drift and the effective spatial resolution
A scanning measurement needs calibrated translation, rotation, beam position, detector orientation and timing. Even small stage offsets between angles can blur or distort a reconstructed feature.
Finite beam size, detector acceptance, photon counts and reconstruction regularization all contribute to effective spatial resolution. The displayed voxel spacing is simply the reconstruction sampling grid, not the smallest feature reliably resolved.
Long scans risk specimen movement, dehydration or radiation-induced change. A volume reconstructed as if the sample were static may contain inconsistency if it evolved during the acquisition.
Co-registering XRF tomography with absorption or phase tomography can help identify geometry, but registration uncertainty must be reported. Overlaying two images on a common display is not equivalent to proving that their voxels sample exactly the same material.
A scientifically useful resolution statement is linked to an independent reference feature or reproducible criterion, with stated field of view, contrast, source conditions and uncertainty.
CHAPTER 10 OF 12 · Validate three dimensions
10. From relative signal to credible elemental concentration
An image showing twice as many characteristic photon counts at one point need not imply twice the elemental mass concentration. The incident flux, sampling depth, self-absorption, matrix and detector response must be sufficiently comparable.
Absolute concentration claims require suitable standards or a validated physical model. Reference specimens should represent relevant geometries and compositions rather than being chosen solely for convenience.
A 2023 Scientific Reports study of self-absorption correction for XRF maps illustrates that absorption effects can distort spatial elemental quantification even in two-dimensional mapping, before tomography introduces further geometrical complexity.
Comparing reconstructed and known concentrations in independently prepared phantoms can reveal systematic bias. Repeated scans and held-out samples help assess robustness beyond one demonstration.
Keep statistical uncertainty separate from systematic uncertainty. Averaging more photons reduces some random noise while leaving a wrong path model or spectral identification unresolved.
CHAPTER 11 OF 12 · Continue X-ray reading
11. Practice, FAQs and adjoining X-ray techniques
Practice A: Calculate √400 = 20 and 20/400 = 5% for an ideal photon-count signal. Why might background-subtracted intensity have greater uncertainty?
Practice B: For 90 angles and 100 translations, confirm 9,000 spectra. At 0.10 seconds each, ideal dwell totals 900 seconds; why will real acquisition take longer?
Practice C: Two ideal independent path transmissions of exp(−0.5) combine to exp(−1) ≈ 0.368. Explain why actual self-absorption depends on sample position and emitted photon energy.
FAQ — Is XRF tomography the same as X-ray absorption CT? No. XRF analyses element-related secondary emission, while conventional CT reconstructs primary attenuation.
FAQ — Does an XRF voxel value identify oxidation state? Not generally. Energy-dependent XANES or other chemical-state evidence may be needed.
FAQ — Can smaller reconstruction pixels create more measured detail? No. Resolution is limited by information in the acquisition and the method.
Follow the connected roots to distinguish elemental mapping, structural tomography, diffraction, chemistry and spatial uncertainty.
Verified eduKateSG reading routes covering related measurement methods:
- The Gold Standard of X-ray Fluorescence Mapping — energy-resolved two-dimensional analysis
- The Gold Standard of X-ray Fluorescence — emission physics and characteristic lines
- The Gold Standard of X-ray Computed Tomography — primary attenuation reconstruction
- The Gold Standard of X-ray Phase-Contrast Imaging — distinct phase-sensitive structural information
- The Gold Standard of X-ray Absorption Near-Edge Structure — element-specific local electronic environment
- The Gold Standard of Spectroscopy — spectral fitting and interpretation
- The Gold Standard of Statistics — measurement uncertainty and independent model tests
CHAPTER 12 OF 12 · Continue X-ray reading
12. Primary sources and the next learning step
Argonne’s XRF-tomography work and peer-reviewed methods for attenuation correction explain why three-dimensional elemental reconstruction is more complicated than simply stacking two-dimensional colour images.
Every scan count, exposure time, attenuation coefficient, photon number and reconstructed interpretation in this article is an invented teaching case. It is not an instrument specification or elemental analysis of any real specimen.
Build an evidence ledger for a chosen research figure: beam energy, detector, spectral fits, scan angles, coordinate calibration, self-absorption model, sample stability, phantom comparison, spatial resolution and uncertainty.
Institutional and peer-reviewed methods for XRF tomography and attenuation:
- Argonne: Hyperspectral XRF Tomography Reconstruction — data acquisition and reconstruction
- AIP Advances: Quantitative XFCT with Iterative Absorption Correction — self-absorption modelling
- Scientific Reports: Self-Absorption in XRF Maps — matrix corrections in elemental mapping
- PMC: Experimental XFCT Imaging Geometries — 3D elemental projection approaches
- PMC: 3D Topography and Fluorescence Absorption — geometry-dependent correction of emission
- Australian Synchrotron: High-Definition XRF Tomography — experimental 3D elemental imaging
The immutable Clementi Secondary 1 Mathematics reference illustrates the pedagogical structure, not certified X-ray imaging procedures.