Broadband X-ray Ptychography

• Broadband X-ray ptychography is a technique that employs multi-energy illumination to record diffraction patterns with enhanced photon flux and spectral diversity.
• It utilizes explicit multiplexing and effective convolution models to address coherence challenges and improve resolution in dose-limited imaging.
• The method offers actionable insights into overcoming detector saturation, probe chromaticity, and reconstruction complexities through advanced optics and computational algorithms.

Broadband X-ray ptychography is ptychographic imaging with incident radiation that contains multiple photon energies rather than being strictly monochromatic. In this setting, the detector records either the incoherent sum of diffraction intensities from all spectral components or a stack of energy-binned diffraction datasets from an energy-resolving detector, and the reconstruction task is either to recover a high-quality image despite finite bandwidth or to separate the response at distinct energies from a single mixed diffraction dataset (Cipiccia et al., 2024, Stolp et al., 23 Aug 2025). The topic is driven by a persistent tradeoff: narrow bandwidth gives good temporal coherence and simplifies phase retrieval, but broad bandwidth gives higher flux, which is critically important for resolution, dose efficiency, throughput, and spectroscopic imaging; this is especially attractive for pink beams, laboratory X-ray sources, and single-acquisition XANES imaging (Lin et al., 2023, Molina et al., 2023).

1. Physical basis and problem setting

In standard ptychography, a sample is scanned through an overlapping coherent probe, and the far-field diffraction pattern is recorded at each position. Broadband operation changes the temporal-coherence assumptions of this model. In the hard X-ray regime, the longitudinal coherence length is written as

$L_t = \frac{\lambda E}{\delta E},$

and the classical bandwidth-limited resolution for an object of width $W$ is

$d_S = \frac{W\,\delta E}{E}.$

With detector-limited sampling

$$\Delta x = \frac{\lambda L}{D} = \frac{\lambda}{\theta},$$

the best achievable resolution is approximately $d_{\min} = 2\Delta x$, and temporal coherence does not limit the reconstruction as long as

$\delta E < \frac{d_{\min} E}{W}.$

These relations state the classical reason broadband illumination is viewed as problematic in coherent diffraction imaging: as spectral bandwidth increases, the diffraction pattern is effectively blurred because different wavelengths scatter to different angles (Cipiccia et al., 2024).

This coherence penalty is not merely a detector issue. Conventional ptychography is usually formulated for fully coherent, quasi-monochromatic illumination, whereas many practically attractive sources are broadband. The practical consequence is the loss of sharp speckle contrast and the appearance of blurred diffraction data; under broadband illumination, high-contrast speckles are blurred, and standard monochromatic algorithms can converge to washed-out amplitude and phase reconstructions (Lin et al., 2023).

At the same time, the motivation for broadband operation is explicit across the literature. Broader bandwidth can deliver higher photon flux for a given source, better practicality with laboratory EUV/X-ray sources, compatibility with ultrafast femtosecond or attosecond sources, and potentially shorter acquisition times for high-throughput or dose-limited imaging (Lin et al., 2023). This establishes broadband X-ray ptychography as both a coherence problem and a flux-utilization problem.

2. Forward models and inverse formulations

Two distinct broadband formulations recur in the literature. The first is an explicit multi-energy multiplexing model. In ptychographic imaging multiplexing, temporal partial coherence is represented as a finite set of spectral modes or channels, each incoherent with the others. For scan position $\mathbf{R}_j$ and energy $E_m$, the exit wave is

$$\psi_{j,m}(\mathbf{r}) = P_m(\mathbf{r})\, O_m(\mathbf{r}-\mathbf{R}_j),$$

its far-field diffraction amplitude is

$$\tilde{\psi}_{j,m}(\mathbf{q}) = \mathcal{F}\left\{ \psi_{j,m}(\mathbf{r}) \right\},$$

and the measured intensity is the incoherent sum

$W$0

If the object response is effectively constant over the bandwidth, one may approximate a common object $W$1, but near an absorption edge this approximation fails, and one must allow distinct energy-channel objects $W$2 (Cipiccia et al., 2024).

The second formulation is an effective convolution model. Lin and Zhang write the monochromatic far-field intensity at scan position $W$3 as

$W$4

and model broadband diffraction as

$W$5

where $W$6 is described as the Fourier transform of the normalized mutual coherence function. In this picture, the measured broadband diffraction pattern is a blurred version of the monochromatic coherent intensity, and the blur kernel is estimated by blind deconvolution through a Wiener-Lucy tandem deconvolution step. The unknowns are the object $W$7, the probe $W$8, and the blur kernel $W$9, and no knowledge of the illumination spectrum is required a priori (Lin et al., 2023).

These two formulations correspond to different levels of physical explicitness. The multiplexing model resolves separate energy channels and is naturally suited to spectroscopy and absorption-edge imaging. The convolution model absorbs several physical effects into a single reciprocal-space blur kernel. This suggests a methodological distinction within broadband X-ray ptychography: some approaches explicitly reconstruct energy-resolved object and probe functions, whereas others seek broadband tolerance through an effective broadband-aware detector-plane constraint.

3. Broadband tolerance versus spectral separation

A central distinction in the field is between tolerating finite bandwidth and separating closely spaced photon energies. Broadband tolerance can be improved experimentally without changing the reconstruction algorithm. In the curved-wavefront approach, the rule-of-thumb inequality

$d_S = \frac{W\,\delta E}{E}.$0

and its numerical-aperture form

$d_S = \frac{W\,\delta E}{E}.$1

summarize the conventional far-field temporal-coherence constraint. The main result is that a strong wavefront curvature transitions a far-field diffraction geometry to an effectively near-field one, which is less affected by temporal coherence effects. Visible-light experiments reconstructed with the same monochromatic ptychography algorithm showed that resolution worsens with bandwidth for fixed curvature, resolution improves with curvature for fixed bandwidth, and strong curvature can partly compensate the loss from broader bandwidth (Molina et al., 2023).

Energy separation is a stricter objective. In the hard X-ray regime, the energy-resolution analysis of broadband ptychography identifies a detectability principle: spectral diversity must produce a detectable difference either in reciprocal space or in real space. In the achromatic case, separation succeeds if

$d_S = \frac{W\,\delta E}{E}.$2

If a chromatic optic creates energy-dependent probes, the criterion becomes

$d_S = \frac{W\,\delta E}{E}.$3

where $d_S = \frac{W\,\delta E}{E}.$4 is the probe difference at the sample between two energies. Using simulations with Fresnel zone plates, the study reports that $d_S = \frac{W\,\delta E}{E}.$5 eV energy resolution at 10 keV is plausible in hard X-ray ptychography if a sufficiently chromatic optical setup is used; experimentally, ptychographic data of an NMC-622 battery cathode material near the Ni K-edge attained an energy resolution of 5 eV (Cipiccia et al., 2024).

These results define two operational regimes. Curved illumination addresses broadband tolerance and throughput under a standard monochromatic reconstruction. Ptychographic imaging multiplexing addresses true spectral decoding by reconstructing separate object and probe functions per energy channel. The two regimes are related but not identical, and the literature treats them with different forward models and different success criteria (Molina et al., 2023, Cipiccia et al., 2024).

4. Optics, detectors, and beamline design

Broadband X-ray ptychography is also an optics-and-detector problem. A central proposal is to replace the monochromator with a hyperspectral detector that resolves the energy of individual photons. One scan then produces a stack of diffraction datasets,

$d_S = \frac{W\,\delta E}{E}.$6

where $d_S = \frac{W\,\delta E}{E}.$7 indexes scan position and $d_S = \frac{W\,\delta E}{E}.$8 indexes detector energy channel. This permits post hoc choice of bandwidth and, in principle, spectral imaging in a single acquisition. The main bottlenecks identified are detector saturation and chromatic behavior of the pre-sample optics. The SLcam is quoted with a flux limit of roughly $d_S = \frac{W\,\delta E}{E}.$9, far below the flux available at synchrotron beamlines, and conventional Fresnel zone plates are strongly chromatic, creating a probe size problem under broadband illumination (Stolp et al., 23 Aug 2025).

The optics analysis compares conventional FZPs, achromatic optics such as Kirkpatrick–Baez mirrors, pinhole-based approaches, and small-diameter FZPs. For FZPs, the focal length is

$$\Delta x = \frac{\lambda L}{D} = \frac{\lambda}{\theta},$$0

and the ptychographic sampling condition is

$$\Delta x = \frac{\lambda L}{D} = \frac{\lambda}{\theta},$$1

The chromatic mismatch metric

$$\Delta x = \frac{\lambda L}{D} = \frac{\lambda}{\theta},$$2

shows that tolerated relative bandwidth increases as FZP diameter decreases. The explicit conclusion is that a reduced-diameter FZP is the most effective strategy among the optical options investigated, because it retains the detector-friendly far-field spreading of an FZP while reducing probe-size chromaticity (Stolp et al., 23 Aug 2025).

Achromatic mirror systems supply a different route. Total-reflection Advanced Kirkpatrick-Baez mirrors are described as having no chromatic aberrations and being inherently achromatic. At the HXCS beamline of the High Energy Photon Source, AKB-mirror-based hard X-ray ptychography at 12.4 keV achieved about 7.8 nm resolution by Fourier ring correlation, with direct confirmation of sub-8 nm half-pitch resolution in real space by edge profile analysis. The work is a single-energy monochromatic demonstration rather than a broadband dataset, but it explicitly frames achromatic nanofocusing as a route toward broadband, energy-scan 3D spectroscopic imaging with element- or chemical-state specificity (Dong et al., 11 Jun 2026).

5. Reconstruction algorithms and computational infrastructure

Broadband reconstruction amplifies the computational burden of ptychography because it combines overlap redundancy with spectral, modal, and calibration variables. In the hard X-ray multiplexing study, reconstructions were performed using ePIE within PtyREX, with each energy channel assigned its own probe and, when needed, its own object. The paper emphasizes that strong spectral decoding requires a “strong key,” especially a sufficiently chromatic probe, and that initialization is critical; in experiment, probes reconstructed from separate monochromatic datasets were used as initial estimates for the mixed-energy reconstruction (Cipiccia et al., 2024).

Algorithmic treatments of coupled uncertainties are therefore significant even when they are not yet explicit broadband models. Generalized Wirtinger Projections formulates ptychography as a complex-valued optimization problem with a mixed-state forward model,

$$\Delta x = \frac{\lambda L}{D} = \frac{\lambda}{\theta},$$3

a Poisson negative log-likelihood,

$$\Delta x = \frac{\lambda L}{D} = \frac{\lambda}{\theta},$$4

and a mixed-state Wirtinger update

$$\Delta x = \frac{\lambda L}{D} = \frac{\lambda}{\theta},$$5

This framework was developed for partial coherence, noise, and scan-position errors rather than for explicit spectral modes, but its use of incoherent sums over modes is structurally close to broadband channelization (Dong et al., 11 Jun 2026).

Systems-level automation has also become relevant. The federated workflow demonstrated at the 26ID hard X-ray nanoprobe beamline is monochromatic as presented—10.4 keV, 963 scan points, 50 nm step size—but it provides an end-to-end architecture of streaming, remote iterative reconstruction with Tike v0.22 and rPIE, continuous training of a modified PtychoNN v2.0 in PyTorch, and edge deployment through ONNX and TensorRT on an NVIDIA Jetson AGX Xavier. The abstract states that the modified PtychoNN-based phase retrieval surrogate shows two orders of magnitude speedup compared to traditional iterative methods. The paper explicitly states that, at the workflow level, the infrastructure could support broadband experiments, whereas at the physics/model level a polychromatic forward model and broadband-aware labels would be required (Babu et al., 2023).

Residual neural-field ptychography is similar in status. It is validated on monochromatic hard and soft X-ray datasets, but it embeds the physical model as a differentiable layer and makes wavelength, scan positions, propagation distance, and aberration coefficients jointly optimizable. The paper itself states that a broadband extension would replace the monochromatic intensity model with a wavelength-averaged or mode-summed intensity model. This suggests that differentiable calibration and residual complex neural fields are relevant templates for future broadband reconstruction, even though the published implementation is not a broadband X-ray method (Zhao et al., 25 Jan 2026).

6. Boundary cases, limitations, and open directions

Several nearby methods clarify what broadband X-ray ptychography is not. Translative lens-based full-field coherent X-ray imaging is a monochromatic coherent imaging approach implemented in a classical Galilean X-ray microscope; its relevance is indirect, as a lens-translation synthetic-aperture and pupil-diversity comparison point rather than a broadband ptychography method (Detlefs et al., 2019). Ptychographic X-ray Speckle Tracking is a ptychography-adjacent wavefront metrology and projection-imaging method that is robust to low-coherence X-ray sources and highly divergent wavefields, but it does not solve the inversion of spectrally mixed diffraction data into wavelength-dependent sample and probe fields (Morgan et al., 2020). Quantitative 3D Bragg ptychography of distorted micro-crystals is explicitly monochromatic, using a Si(111) double crystal monochromator at 11.8 keV; its relevance lies in showing how overlap redundancy stabilizes inversion in a Bragg-sensitive geometry that would need to be generalized for finite bandwidth (Li et al., 12 Mar 2026).

The limitations of current broadband formulations are equally explicit. The blind-deconvolution approach is an effective rather than fully physical model: it does not explicitly include a discrete set of spectral modes, wavelength-dependent probe $$\Delta x = \frac{\lambda L}{D} = \frac{\lambda}{\theta},$$6, wavelength-dependent object transmission $$\Delta x = \frac{\lambda L}{D} = \frac{\lambda}{\theta},$$7, detector spectral response, or an explicit integral over source spectrum $$\Delta x = \frac{\lambda L}{D} = \frac{\lambda}{\theta},$$8 (Lin et al., 2023). The optics literature states that broadband ptychography is not simply “use more energy spread”; detector count-rate limits and optic chromaticity remain primary practical bottlenecks, and even within a detector bin the wavefield is not exactly monochromatic because bins must be finite width and detector energy resolution causes photons from outside the nominal bin to contribute (Stolp et al., 23 Aug 2025).

There are also domain-specific caveats for X-rays. The X-ray transfer of optical proof-of-principle methods raises additional issues: detector quantum efficiency can vary strongly with energy; absorption and phase shift are strongly energy dependent, especially near absorption edges; the probe in X-ray ptychography may vary substantially with wavelength because of chromatic optics; thick or heterogeneous samples may require multislice treatment; and a single effective blur kernel may not capture all coherence structure in real beamlines (Lin et al., 2023). In the hard X-ray multiplexing study, the spectral components are effectively treated as known discrete energies, and the discussion explicitly states that whether quantitative energy resolution holds in the continuum case still needs verification (Cipiccia et al., 2024).

A recurring misconception is that broadband X-ray ptychography is a single technique. The literature instead shows at least three distinct agendas: broadband tolerance under a standard monochromatic reconstruction, broadband-aware inversion through an effective blur or deconvolution model, and true energy-resolved multiplexed reconstruction. Another misconception is that an energy-resolving detector alone solves the problem. The detector–optics design analysis shows that detector saturation, probe-size chromaticity, order sorting, and detector-plane illumination homogeneity are central constraints, and that the best optic is not the one that captures the most source photons in absolute terms but the one that maximizes useful recorded photons within detector constraints while preserving broadband ptychographic compatibility (Stolp et al., 23 Aug 2025).

Taken together, the field defines broadband X-ray ptychography as a coupled problem of temporal coherence, spectral encoding, pre-sample optics, detector count-rate, and inversion architecture. The published work already spans single-acquisition spectral channelization, blind deconvolution without spectrum knowledge, chromatic-probe energy separation down to electronvolt scales, curved-illumination tolerance strategies, achromatic mirror nanofocusing, and workflow and differentiable-physics frameworks that are not broadband themselves but are technically aligned with broadband extensions (Lin et al., 2023, Cipiccia et al., 2024, Stolp et al., 23 Aug 2025, Dong et al., 11 Jun 2026).