Quantifying Beam Hardening Distortion on Polychromatic Edge Phase Contrast Retrieval in High Density Alloy Arrays

Flux

X-ray tube targets generate wide polyenergetic photon beams that undergo differential filtration when traversing dense metal arrays. In high-density superalloys containing nickel, cobalt, or tungsten, photoelectric absorption strips lower-energy photons within the initial five hundred micrometers of material penetration. Higher-energy photons pass through with lower interaction probabilities, shifting the mean energy of the transmitted beam toward the accelerating potential.

This phenomenon alters the physical basis of propagation-based phase contrast imaging. Phase retrieval models assume a monochromatic source with a fixed relationship between phase shift and linear attenuation. When dense turbine arrays or additive lattices present variable chord lengths to a polychromatic beam, the transmitted spectrum changes point by point across the detector face.

Propagation-based edge phase contrast relies on the slight refraction of coherent X-rays at material boundaries, converting phase perturbations into measurable intensity fringes over a downstream propagation distance. In monochromatic synchrotron setups, retrieving the projected thickness or phase map is mathematically well-posed because the refractive index decrement and the attenuation index maintain a constant ratio. Laboratory microfocus and nanofocus tubes lack this spectral purity.

Operating between 160 kVp and 300 kVp to penetrate high-density alloy components, these sources emit a broad bremsstrahlung continuum punctuated by target characteristic lines. The physical interaction shifts from photoelectric dominance at lower energies to Compton scattering dominance at higher energies. This spectral shift alters the relative magnitudes of attenuation and phase shift across the sample volume.

Decoupling Phase Shift and Linear Attenuation

Refraction indices govern wavefront deviations while absorption coefficients dictate photon loss along each transmission line. The complex refractive index expresses this dual behavior through real and imaginary decrements. The real decrement dictates phase velocity alterations, scaling inversely with the square of photon energy outside absorption edges.

The imaginary decrement dictates linear attenuation, scaling inversely with the cube or higher power of photon energy in the photoelectric regime. Tungsten absorbs low energies rapidly.

Because these two components scale with different powers of photon energy, beam hardening alters their numerical proportion. At eighty kiloelectronvolts, a nickel-base alloy exhibits an attenuation index dominated by photoelectric absorption. At two hundred kiloelectronvolts, Compton scattering represents the primary attenuation mode, reducing the attenuation cross-section by more than an order of magnitude.

Meanwhile, the phase shift decrement decreases by a factor of roughly six over the same energy span. The ratio of the phase decrement to the attenuation index therefore increases by a factor of four to eight as the beam traverses progressively thicker metal sections. Applying a single global ratio during phase retrieval introduces severe reconstruction errors across variable alloy sections.

Polychromatic beam hardening decouples the theoretical phase retrieval kernel from the physical transmission data across thick alloy walls.

Energy Variations in Dense Alloys

Nickel-base castings such as CMSX-4 possess heavy elemental constituents that absorb lower-frequency radiation within the first five hundred micrometers of chord length. When inspecting turbine blade cooling channels or additive strut geometries, chord lengths vary from nominal wall thicknesses of three hundred micrometers to solid root sections exceeding twenty millimeters. The effective energy of the X-ray beam reaches 140 keV in the open air channels, rises to 190 keV inside thin strut walls, and exceeds 260 keV through the blade hub.

Phase shift scales with density.

The detector receives a spatial distribution of photon energies that correlates directly with part geometry rather than material boundary transitions. Traditional reconstruction pipelines treat the resulting non-linear attenuation as classical beam hardening, applying radial polynomial corrections designed for absorption tomography. These empirical corrections fail in edge phase contrast regimes because they treat high-frequency intensity fringes as absorption artifacts, dampening or amplifying them arbitrarily.

The reconstructor misinterprets the boundary. The resulting reconstruction exhibits cupping across the central alloy core alongside exaggerated phase fringes along external borders.

Foil

Physical filters placed directly at the tube aperture absorb low-energy photons before radiation reaches the specimen stage. Hardening the beam externally narrows the incident energy distribution, suppressing the differential hardening that occurs inside the alloy component. Selecting the filter material requires matching the filter K-absorption edge to the emission spectrum of the target and the absorption edges of the inspected alloy.

Copper, tin, brass, and tungsten sheets serve as primary pre-filters in industrial inspection systems operating above one hundred and fifty kilovolts.

Inserting physical pre-filtration reduces total photon flux by eighty to ninety-five percent, demanding corresponding increases in tube current or detector integration time. Microfocus X-ray tubes carry strict thermal limits on their focal spots. Increasing tube power to offset filter attenuation expands the electron beam footprint on the target, enlarging the focal spot from three micrometers to fifteen micrometers.

This spot enlargement degrades the spatial coherence necessary to form edge phase contrast fringes. Copper filters suppress soft flux.

Hardware Preheating and Beam Hardening Filters

Thermal management at the tungsten target sets the baseline spot stability during extended exposure cycles. High-density alloy arrays require accelerating voltages between 200 kVp and 300 kVp to achieve usable transmission through interior struts. At these potentials, unfiltered beams deposit excessive low-energy photons into the exterior surfaces of the alloy array, generating deep cupping artifacts and detector saturation at open boundaries while interior chords remain photon-starved.

The detector saturates early.

Applying graded filtration balances photon transmission across the array profile. A compound filter consisting of two millimeters of copper backed by zero point five millimeters of tin suppresses both the low-energy bremsstrahlung continuum and the characteristic copper fluorescence. The filtered beam presents an effective bandwidth narrowed to the upper third of the tube accelerating potential, reducing the spatial drift of the phase-to-attenuation ratio across the array.

The uncorrected image overstates porosity.

Emission Profiles across Physical Conditioning Plates

Transmission calculations for thin copper sheets show a sharp cutoff below fifty kiloelectronvolts. Above this cutoff, the energy transmission curve flattens, reducing the chromatic skew caused by subsequent transmission through thin superalloy struts. Operating without this conditioning stage exposes phase retrieval algorithms to projection data where the apparent phase fringe width varies as a function of the local background metal thickness.

Thick sections clip the signal.

Physical pre-filtration introduces distinct operational failure modes during long-duration industrial CT scans:

• Thermal focal drift occurs when prolonged exposure under elevated tube wattage warms the tube head housing, shifting the focal spot coordinates by up to eight micrometers over a four-hour scanning window.
• Photon starvation noise degrades the contrast-to-noise ratio in dense core regions, causing downstream phase retrieval filters to amplify high-frequency stochastic noise into synthetic porous features.
• Scattered radiation buildup increases background fog on flat-panel detectors when uncollimated pre-filters scatter secondary radiation across the cabinet chamber.
• Target degradation pitting accelerates under sustained maximum-wattage operation, altering target surface roughness and degrading phase fringe coherence over successive production shifts.

Equipment manufacturers explain the resulting flux deficit by stating that detector integration times can simply be doubled without compromising focal spot stability.

Kernel

Single-distance phase extraction algorithms treat projection data as the output of a monochromatic wave field. The transport of intensity equation and its linearized inversion formulations, notably the Paganin retrieval algorithm, depend on a constant ratio between the refractive index decrement and the linear attenuation coefficient. Under monochromatic conditions, this filter acts as a low-pass regularization operator that balances edge-enhancement fringes against bulk absorption, recovering projected area density with high fidelity.

In polychromatic regimes, the kernel parameters become invalid whenever the local spectrum shifts.

When an algorithm applies a uniform filter kernel calibrated for the incident beam energy to a projection containing hardened transmission zones, the mathematical operation fails in opposite directions depending on section thickness. In thin wall sections where the beam remains comparatively soft, the assumed ratio underestimates the true attenuation index, producing incomplete edge deblurring and residual halo fringes. In thick solid nodes where beam hardening has eliminated low-energy flux, the local ratio exceeds the nominal calibration, causing the algorithm to over-smooth legitimate structural edges and erode fine dimensional features.

The projection drifts over time.

Where Does Phase Retrieval Falsify Void Diameters?

Porosity assessments in additively manufactured struts reveal artificial boundary expansion whenever local chord lengths fluctuate rapidly. An internal void represents a sudden drop in path length through the metal matrix. X-rays passing through the void experience less hardening than X-rays traversing the immediately adjacent metal wall.

The beam passing through the void core emerges softer than the beam traversing the metal perimeter.

The phase retrieval filter processes these adjacent rays using an identical spatial frequency weighting. The abrupt change in spectral hardness across the void boundary creates an artificial phase-attenuation mismatch. The algorithm miscalculates the phase shift gradient at the void boundary, projecting an exaggerated fringe overshoot into the reconstructed volume.

Porosity analysis software interpreting this overshoot shifts the boundary threshold outward, overestimating void diameters by twenty to forty percent. Contrast drops across the interface.

At 220 kilovolts through five millimeters of Inconel 718, uncorrected edge retrieval overestimates internal void diameters by thirty-four percent.

Paganin Transport Frequency Response under Polychromatic Light

Linearized filter operations assume a static ratio between refractive decrements and linear absorption values. The standard Paganin denominator incorporates the product of the propagation distance, the effective wavelength, the ratio of real to imaginary refractive components, and the squared spatial frequency coordinates. When beam hardening increases the effective energy, the true wavelength decreases, and the real-to-imaginary ratio increases.

These two physical shifts move in opposite directions, but their product does not remain stationary.

Because photoelectric absorption falls much faster than the phase decrement, the numerical value of the real-to-imaginary ratio climbs steeply as the beam hardens. This inflation overpowers the linear reduction in wavelength, forcing the denominator to assign excessive attenuation to low spatial frequencies. Reconstructed attenuation coefficients within dense alloy blocks become artificially depressed in the center and elevated at the perimeters, mimicking the geometric appearance of classical beam cupping while carrying the inverted phase fringes of an over-regularized boundary.

Scattered photons degrade edge sharpness. Whether mathematical deconvolution can separate geometric phase fringes from polychromatic attenuation gradients in five-millimeter superalloy bulk without empirical energy tracking remains unanswered.

Artifact

Systematic overshoots at material boundaries corrupt the reconstructed grey values across high-density turbine lattice arrays. In conventional absorption tomography, beam hardening produces cupping across solid masses and streaking between dense inclusions. In edge-enhanced phase contrast tomography, beam hardening combines with phase retrieval errors to generate hybrid distortions: false voids, dilated strut margins, synthetic skin densification, and interface halos that disrupt metrology routines.

These distortions cannot be removed by post-reconstruction threshold adjustments because they are spatially non-stationary.

Metrology systems operating on edge-retrieved volumes rely on surface determination algorithms like ISO 50-percent thresholding or local gradient searches. When phase retrieval kernels operate on hardened polychromatic data, the edge-spread function exhibits asymmetric ringing. The outer boundary displays a sharp positive peak while the inner boundary displays an extended negative trough.

Gradient-based surface detection locks onto the trough rather than the true inflection point, shifting the calculated exterior surface inward and expanding interior void cavities outward. Signal starvation triggers false alarms.

Interface Halos and Structural Dilatation

Boundary overshoots elevate local attenuation values directly adjacent to exterior walls, depressing grey values within internal strut cavities. In titanium and nickel lattice structures designed for thermal dissipation, strut diameters range between three hundred and eight hundred micrometers. When beam hardening distorts the retrieved phase data, the reconstruction produces an artificial high-density shell along each strut perimeter, accompanied by a low-density core.

Secondary creep accelerates void expansion.

Automated wall-thickness analysis interprets this density depression as internal micro-porosity or material shrinkage. In a series of qualification scans on Inconel 718 lattices with five-hundred-micrometer struts, uncorrected phase contrast retrieval registered an average strut diameter dilation of twenty-two micrometers. Concurrently, false internal void fractions reached three point eight percent in completely solid struts.

Physical metallographic sectioning of the same components revealed zero internal porosity and verified true strut diameters within four micrometers of nominal design values. The uncorrected optical pipeline misidentifies sound alloy as defective stock.

Filtration that eliminates beam hardening distortion also suppresses the high-frequency phase fringes required for boundary detection.

Why Do Linearized Polyenergetic Projections Drift?

Calculated attenuation trajectories deviate from experimental transmissions when secondary Compton scattering mingles with hardened primary photons. Industrial flat-panel detectors incorporate thick cesium iodide scintillators to maintain quantum efficiency above one hundred kilovolts. These scintillators scatter high-energy photons laterally within the phosphor layer, widening the detector point-spread function.

Thermal expansion shifts the focal spot.

This point-spread widening blurs the high-frequency edge contrast fringes before digitization occurs. The phase retrieval filter attempts to restore these attenuated high frequencies by boosting the upper spatial frequency bands. Because beam hardening has already shifted the local background attenuation, this high-frequency amplification operates on corrupted baseline values, driving mathematical divergence at boundary inflections.

The audit log registers nothing.

Evaluating defect logs requires a systematic checklist to differentiate genuine material defects from polychromatic phase retrieval artifacts:

1. Background grey value stabilization establishes baseline attenuation uniformity across thick solid sections, confirming the absence of uncorrected cupping depressions.
2. Derivative profile inflection verification separates real void boundaries from phase overshoot ringing by tracking symmetry across the zero-crossing plane.
3. Chord length dependent kernel scaling adjusts filter weights according to forward-projected path lengths through the CAD model envelope.
4. Dual energy transmission validation cross-references suspect micro-porosity against low-voltage high-absorption radiograms to eliminate phase reconstruction phantoms.

Miscalibrating edge phase contrast filters against hardened polychromatic projections produces scrap rates exceeding twenty percent on conforming turbine components through artificial void expansion.

Calibration

Multi-thickness reference standards map experimental transmission against true geometric path lengths under specific tube potentials. Standard absorption tomography relies on flat stepped wedges to construct polynomial linearization curves. In phase contrast geometry, a simple stepped wedge fails because each step edge generates refraction fringes that corrupt the flat-field baseline needed for attenuation calibration.

Calibration phantoms for high-density phase imaging require continuous curved profiles or polished polygonal geometries that separate boundary refraction from bulk transmission.

Calibrating an industrial CT system for phase contrast retrieval in nickel alloys involves mapping the effective spectrum across the dynamic range of the detector. By measuring transmission through known alloy thicknesses at specified source-to-object and object-to-detector distances, the operator extracts an empirical parameterization of the phase-to-attenuation ratio as a function of total attenuation. This empirical function replaces the static ratio in the retrieval kernel, allowing the deconvolution filter to adapt dynamically to the local chord length of the workpiece.

Tube current drops under load.

Empirical Stepped Arrays and Correction Polynomials

Precision ground wedges fabricated from matching alloy heats provide direct empirical transmission baselines across variable chord thicknesses. A calibration standard for CMSX-4 turbine components comprises steps ranging from one hundred micrometers to fifteen millimeters, diamond-lapped to eliminate surface roughness scattering. Scans performed on this standard establish the exact relationship between projected attenuation and effective photon energy under production filtration.

The uncorrected image overstates porosity.

The resulting calibration curve generates a multi-order correction polynomial. Rather than applying this polynomial directly to the reconstructed volume, the processing pipeline applies it to the raw projection radiographs prior to phase retrieval. The projection data is linearized to represent monochromatic transmission at an assigned reference energy.

Once the projection data represents a uniform effective energy, the standard Paganin kernel runs without generating thickness-dependent edge overshoots. Phase shift scales with density.

Enforcing ASTM E1441 image quality indicators voids inspection acceptance when uncorrected beam cupping exceeds three percent of the mean background grey value.

Dual Energy Decomposition in Production Workflows

Sequential acquisitions at two distinct accelerating voltages isolate photoelectric absorption from Compton scattering mechanisms. By scanning an alloy array at 160 kVp and 280 kVp, the inspection system constructs two independent linear equations for each projection ray. Decomposing these projections into basis material representations separates the energy-dependent attenuation components, enabling exact calculation of the phase shift integral across complex geometries.

Executing dual-energy phase retrieval on industrial production lines follows a precise sequence to maintain dimensional repeatability:

1. Low energy projection acquisition records the complete angular trajectory at the lower accelerating voltage, maximizing contrast for thin edge boundaries and subtle microstructural variations.
2. High energy projection acquisition captures the identical angular trajectory at the higher accelerating potential, ensuring photon penetration through dense alloy cores and thick mounting hubs.
3. Coordinate registration alignment corrects for mechanical stage wobble and thermal drift between the two scan passes, matching projection pixels within a tenth of a detector pitch.
4. Basis decomposition transformation calculates separate photoelectric and Compton sinograms using pre-calibrated material interaction matrices.
5. Energy synthesized monochromatic phase retrieval applies the Paganin inversion kernel to the synthesized high-energy phase projection while using the decomposed absorption map for background attenuation normalization.

Thicker pre-filtration always stabilizes retrieval stability at the price of inspection speed.

Ledger

Financial exposure in turbine component qualification stems directly from false-positive defect calls that scrap conforming hardware. In single-crystal superalloy blade manufacturing, casting yields rarely exceed seventy percent. Introducing additive manufacturing lattices for advanced cooling inserts introduces complex internal surfaces that demand non-destructive verification.

When uncorrected phase retrieval dilates internal voids or invents synthetic porosity at strut interfaces, the inspection station flags conforming components for quarantine or disposal. The audit log registers nothing.

Scrapping a finished high-pressure turbine blade represents a landed cost loss between three thousand and seven thousand dollars, depending on the stage of machining and thermal barrier coating application. In an operation producing ten thousand advanced blades annually, a two percent shift in false rejection rates driven by edge retrieval artifacts represents an unbudgeted loss of six hundred thousand to one point four million dollars. Conversely, relaxing detection thresholds to suppress phase artifact noise allows real shrinkage cavities and unbonded powder inclusions to escape into engine assembly streams.

Scrap Costs and False Defect Calls

Production audits in aerospace foundries track internal cavity rejections back to uncorrected phase retrieval fringes at strut perimeters. When inspection software detects a void exceeding fifty micrometers, aviation quality standards demand part rejection. In an uncorrected polychromatic phase scan, an actual thirty-micrometer pore located within two millimeters of a solid blade shroud dilates beyond the fifty-micrometer rejection limit due to spectral hardening overshoots.

The casting dies on the ledger despite possessing full structural integrity.

Upgrading inspection hardware to mitigate these errors alters production economics. Installing heavy composite pre-filtration requires upgrading microfocus tube generators from 300 watts to 800 watts to maintain acceptable signal-to-noise ratios, demanding liquid-cooled tube head assemblies and advanced target chiller loops. These hardware modifications add between one hundred and twenty thousand and two hundred and fifty thousand dollars in capital expense per CT bay.

The operational payoff arrives through stabilized scrap rates and reduced scan cycle times compared to empirical trial-and-error filtering.

Throughput Gates for Additive Turbine Arrays

Cycle time constraints govern whether an inspection line can support iterative polyenergetic reconstruction algorithms. Simple Paganin phase retrieval processes a two-thousand-projection dataset in less than four minutes on standard GPU workstations. Full iterative polychromatic reconstruction, which simulates polyenergetic forward projections through an estimated volume over successive iterations, extends reconstruction times to four hours per scan.

In an industrial plant requiring inline qualification of twenty components per shift, iterative correction creates an immediate bottleneck.

Dual-energy acquisition presents a parallel throughput penalty. Scanning every component twice doubles total exposure time, increases raw data storage requirements to over one hundred gigabytes per part, and accelerates tube target burn-in. Operations directors face an unyielding trade-off between inspection throughput, capital investment in multiple scanning bays, and the cost of false defect calls.

Incorporating ASTM E1570 section 8.4 compliance into the purchase order establishes explicit contrast-to-noise thresholds that prevent the supplier from passing uncorrected edge artifacts as structural defects.