Micro-X-ray fluorescence spectroscopy brings elemental XRF analysis to a localized region by focusing or tightly collimating the excitation beam, positioning a small specimen area at the optic’s working distance, and collecting a complete fluorescence spectrum. The method can inspect particles, patterned wafers, solder features, coatings, contamination spots, geological grains, and heterogeneous devices without the electron-beam charging and vacuum requirements of electron microanalysis. Its spatial selectivity is not defined by one catalog spot number: beam energy, optic transmission, specimen depth, incidence geometry, fluorescence escape, and stage motion determine the actual sampling volume.

Micro-XRF is local XRF spectroscopy built around an X-ray optic. A laboratory instrument commonly combines a microfocus tube, polycapillary focusing lens or collimator, motorized sample stage, optical navigation camera, and energy-dispersive detector. Synchrotron systems may use Kirkpatrick–Baez mirrors, Fresnel zone plates, capillaries, or other optics with monochromatic tunable excitation. Confocal micro-XRF adds a collection optic before the detector so excitation and detection foci overlap in a depth-selective probe volume.

The optic trades flux, spot size, working distance, depth of focus, energy band, and alignment tolerance. Polycapillary channels guide X-rays by repeated total external reflection and accept a comparatively large source solid angle, making them effective with divergent laboratory tubes. Their transmission and focal spot are energy dependent because the critical reflection angle decreases as photon energy rises. A nominal spot quoted at one line energy does not describe the polychromatic footprint of every excitation energy.

The focal spot must be measured as a function of photon energy and position. Scan a sharp edge, wire, aperture, or well-characterized particle through the beam and derive an edge- or line-spread function using a declared width metric. Repeat at representative energies or emission-sensitive conditions because a polychromatic polycapillary transmits different energies with different efficiency and focus. Confirm the working-distance peak, depth of focus, beam tails, asymmetry, and stability after optic or tube service.

For a Gaussian approximation, an edge response can be written

Real micro-XRF beams can have halos and energy-dependent non-Gaussian wings, so FWHM alone may understate contamination from adjacent features. Report encircled-energy or tail behavior when measuring isolated pads, narrow lines, or particles. The stage step may oversample the beam but cannot improve the physical resolution beyond the qualified response.

The sampled volume extends below and around the nominal focus. Incident X-rays penetrate according to energy and matrix; fluorescence escapes according to its own energy and takeoff angle. In a bulk specimen, deeper and laterally displaced material can contribute even when the surface beam is micrometer-scale. Surface tilt and topography change working distance, incidence angle, escape length, and detector solid angle. A height error can reduce flux or move the focus onto a neighboring structure.

For thin films, the localized intensity depends on areal mass and layer attenuation. For bulk heterogeneous specimens, grain geometry, surface roughness, particle size, mineralogy, and secondary fluorescence can dominate. Do not apply a bulk calibration to a thin coating or a flat-glass standard to a curved solder bump without demonstrating transfer. Optical height mapping or autofocus can stabilize geometry, but its coordinate system must be registered to the X-ray focus.

A complete energy spectrum is required at every local analysis point. Fit characteristic line families, continuum, coherent and incoherent scatter, detector escape peaks, sum peaks, pileup, and diffraction artifacts. Microbeam illumination can make diffraction spikes more severe because a small number of crystallites satisfy Bragg conditions. Changing sample rotation or incidence can diagnose diffraction; a region-of-interest sum cannot reliably distinguish it from fluorescence.

Detector dead time and pulse pileup depend on local composition and scatter, so a count-rate-qualified acquisition at one location may fail on another. Use live-time correction only within its validated range. Adjust tube current, filters, distance, or detector geometry rather than accepting severe pileup. Store full spectra and live-time metadata, not only elemental intensities.

Element detection follows characteristic energies, but quantitative local composition requires a fundamental-parameter or standards-based model. A compact intensity representation is

where source spectrum (\Phi), optic transmission (T_{\rm opt}), detector solid angle (\Omega), efficiency (\epsilon), local concentration (C_k), production factor (P_k), attenuation/enhancement term (A), and effective volume (V) are coupled. The polycapillary modifies the tube spectrum, so an unfocused-tube calibration cannot simply be reused after the lens.

Quantification needs standards that match scale, matrix, and geometry. A homogeneous reference material can calibrate bulk composition when the beam samples many representative microstructural units. Thin-film standards with certified areal density support coatings and deposits. Microfabricated pads or particles test localization and tails. Standards must be homogeneous at the beam scale; a certificate for a bulk average does not guarantee micrometer uniformity.

Reference-free micro-XRF is possible when source flux spectrum, optic transmission, beam profile, geometry, detector efficiency, solid angle, fundamental parameters, and sample model are independently calibrated. “Standardless” software is not automatically reference-free or traceable. Validate a reference-free result against homogeneous standards and propagate optic and spatial-response uncertainties.

Thickness and concentration can be correlated. If the film is sufficiently thin that self-absorption is negligible, intensity may scale with elemental areal mass; beyond that regime attenuation creates nonlinear response. Multiple emission lines, excitation conditions, angles, or an independent thickness can improve identifiability. A single local spectrum usually cannot reconstruct an arbitrary multilayer depth profile.

Confocal micro-XRF creates a depth-sensitive voxel rather than an infinitesimal point. A second polycapillary before the detector accepts fluorescence only from its focal region; overlap with the excitation focus defines the probe volume. Both foci and their transmissions depend on energy, and the voxel is often anisotropic. Measure it by translating small particles or sharp interfaces in three dimensions for representative fluorescence energies.

Confocal depth scans still require absorption and enhancement corrections along incident and exit paths. Features smaller than the voxel are diluted, and signals from nearby layers can overlap. Deconvolution is model dependent. Report voxel widths, energy, geometry, scan step, regularization, and boundary response. Tomographic or 3D renderings should not imply isotropic resolution when axial and lateral responses differ.

Define local composition, thin-film areal mass, particle, interface, or depth objective
  -> Select tube/beam energy, focusing optic, detector, atmosphere, and geometry
  -> Predict line overlaps, matrix attenuation, penetration, and feature dimensions
  -> Measure optic transmission, focus, tails, working distance, and stability versus energy
  -> Calibrate detector energy, efficiency, solid angle, live time, stage, and optical registration
  -> Measure blanks plus scale- and matrix-appropriate local standards
  -> Acquire full spectra with height, flux, position, and live-time metadata
  -> Fit spectral families and reject diffraction, pileup, and geometry artifacts
  -> Apply thin-film, bulk, particle, or confocal fundamental-parameter model
  -> Propagate spectral, optic, spatial, matrix, and standard uncertainty
  -> Validate selected locations with XRF maps or complementary microscopy/chemistry
  -> Archive spectra, optic qualification, coordinates, model, and uncertainty

Local heterogeneity changes sampling uncertainty and representativeness. One precise microspot does not establish wafer or batch composition. Define a sampling plan from feature pitch, expected gradients, particle statistics, edge effects, and the decision to be made. For random particles or grains, uncertainty between locations may exceed counting uncertainty. Report distributions, robust summaries, and the sampled area rather than only the most visually interesting point.

Navigation images can misregister with the X-ray focus due to parallax, camera distortion, stage rotation, or height. Verify co-location with an artifact visible optically and by XRF. Recheck after changing working distance or sample holder. When comparing to SEM, Raman, SIMS, or electrical probing, use shared fiducials and include registration uncertainty in any claimed spatial correlation.

Detection limits depend on element, matrix, line overlap, local background, dwell, beam flux, optic throughput, detector, and criterion. Quote the qualified point or feature geometry and distinguish mass, mass fraction, and areal-mass limits. A low absolute mass limit on a particle is not the same as a low bulk concentration limit. Provide blank and reporting limits alongside the estimate.

Radiation dose can damage polymers, battery materials, hydrated phases, organics, and metastable devices. Compare repeat spectra, reverse scan order, vary dwell, and monitor changing ratios. Laboratory instruments and synchrotron microprobes require intact shielding, interlocks, authorized procedures, and trained operation under the institutional radiation-safety program. Never defeat an interlock or improvise an open-beam setup.

Traceability includes the optic because it reshapes both space and spectrum. Store tube or beam energy and current, anode and filters, optic identity, transmission calibration, energy-dependent beam maps, working distance, depth of focus, sample height, incidence/takeoff angles, detector response, live time, stage calibration, optical-to-XRF transform, atmosphere, full spectra, standards and certificates, layer/matrix model, fundamental parameters, residuals, uncertainty, detection limits, dose checks, and software version.

Qualification should include homogeneous bulk and thin-film references, small high-contrast features for beam tails, edge or particle scans versus energy, repeat positioning, height offsets, detector count-rate tests, and multi-day drift checks. For confocal systems, scan a particle through the voxel in three axes and repeat for line energies that bracket the application. Compare local results with conventional XRF, ICP-based chemistry, SEM-EDS, XPS, SIMS, RBS, XRR, or profilometry as appropriate to depth and scale.

The strongest conclusion distinguishes the local measurement from the larger material population. Micro-XRF can establish that a specific pad, particle, region, or voxel differs elementally and can quantify it when the optic and matrix model are calibrated. It cannot turn one spot into a wafer average, one surface coordinate into a sharp depth slice, or one software fundamental-parameter result into traceability without standards and uncertainty.

A defensible Micro-XRF result binds optic response, spectral fit, sampling volume, and specimen geometry. The microbeam makes heterogeneity visible, but it also makes focus drift, surface height, grain diffraction, and standard inhomogeneity consequential. Energy-resolved spatial qualification and matrix-aware calibration convert a small bright spot into reliable local elemental metrology.

The durable way to interpret Micro-XRF is through a source-optic-energy-dependent-focus-working-distance-spectrum-sampling-volume-matrix-standard-confocal-voxel-and-traceability lens.