X-ray ptychography is a coherent diffraction imaging method that scans an illuminating X-ray wavefield across overlapping sample regions and records diffraction intensity patterns. Computational reconstruction uses information shared among those measurements to estimate the sample’s complex transmission and, in many approaches, the illuminating probe.
The gold standard is not simply the sharpest reconstruction. It is a defensible chain from beam, sample position and detector measurements to phase and amplitude estimates, with checks on coherence, overlap, model assumptions, drift, resolution and radiation effects.
eduKateSG · X-ray Ptychography
Find your next learning step
Build the complex-wave forward model, calculate scan overlap and photon wavelength, then test how phase retrieval and instrument limits govern claims.
Full chapter index · Practice and transfer · Education root · How X Works
Full chapter index
Define coherent scanning · Chapters 1–2
Calculate simple examples · Chapters 3–4
Check inversion stability · Chapters 5–7
Interpret reconstructions · Chapters 8–10
Follow X-ray methods · Chapters 11–12
CHAPTER 1 OF 12 · Define coherent scanning
1. Ptychography builds a picture from overlapping measurements
Conventional scanning microscopy forms an image by measuring a signal at a sequence of positions. X-ray ptychography typically records a full coherent diffraction pattern at each probe position, giving much richer information than one scalar intensity value.
The illuminating field is moved relative to the object so that neighbouring illuminated areas overlap. The same sample feature then contributes to several diffraction patterns under different neighbouring illumination contexts.
Because detectors record intensity rather than full complex optical fields, the diffraction phase is missing. Ptychography uses shared constraints across measurements and a forward model to estimate compatible phase and amplitude.
DESY describes reconstructing both the complex sample transmission and the complex illuminating wavefield from the scan. Those are model-based estimates: a reconstruction can be precise-looking even when a wrong position map or partial coherence corrupts it.
An experiment must state sample preparation, radiation energy, optics, scan geometry, measured diffraction patterns and analysis method before a claim of sub-beam resolution can be assessed.
CHAPTER 2 OF 12 · Define coherent scanning
2. The forward model separates object, probe and detector intensity
In a simplified thin-object coherent model, let O(r) be the complex object transmission, P(r − Rj) the illuminating probe shifted to scan position Rj, and ψj(r) = P(r − Rj) O(r) the exit wave.
Under suitable far-field assumptions, the model detector intensity is Ij(q) = |F[ψj(r)]|², where F denotes a Fourier transform and q labels detector spatial frequency. The measured pattern omits the complex Fourier phase.
Writing down the model makes several assumptions visible: a thin projected transmission, a stable probe description, known relative scan positions, correctly calibrated geometry and sufficient coherence for the chosen representation.
Thick or strongly scattering objects may need multislice, dynamical or other extended models. Partial-coherence models may use several probe modes rather than a single coherent field.
Reconstruction alternates or optimizes estimates until simulated diffraction intensities agree with observed patterns within their errors. Agreement alone is not proof of unique material composition or guaranteed spatial resolution.
Take a hypothetical effective probe diameter d = 500 nm and a scan step s = 200 nm along a straight line. Using the simple geometric definition of linear overlap fraction η = (d − s)/d, obtain η = (500 − 200)/500 = 300/500 = 0.60.
The nominal overlap is therefore 60% along the scan direction. This is not necessarily 60% overlapping area for circular beams and not 60% identical complex illumination between two measurements.
The probe may have a structured intensity footprint without a sharp diameter. In such cases the geometric calculation is a convenient planning statistic but not a complete information or stability criterion.
Overlapping measurements provide redundancy that can support phase retrieval, but appropriate overlap depends on probe structure, scan pattern, noise, specimen and algorithm. DESY research discusses overlap choices rather than one universally necessary fraction.
A useful transfer test is to double the step to 400 nm while holding the nominal diameter fixed. Linear overlap falls to (500 − 400)/500 = 20%, reducing shared illuminated area under this simple footprint model.
CHAPTER 4 OF 12 · Calculate simple examples
4. Worked example: count scan positions and compute wavelength
Suppose a hypothetical square scan grid uses positions at 0, 0.2, 0.4, 0.6, 0.8 and 1.0 micrometres along each axis. There are six positions per direction and therefore 6 × 6 = 36 nominal diffraction acquisitions.
These positions specify the location of probe centres, not the total illuminated field of view. The illumination extends around each centre, and the reconstructed area and resolution depend on sampling and boundary assumptions.
Photon energy and wavelength are related by E = hc/λ. A convenient X-ray approximation is λ in nm ≈ 1.23984/(E in keV). For a hypothetical photon energy E = 12.4 keV, λ ≈ 1.23984/12.4 = 0.09999 nm, approximately 0.100 nm.
The wavelength conversion defines a physical scale for propagation and diffraction geometry. It does not assert that the reconstructed image has 0.1-nm resolution. Finite detector angles, photon statistics, coherence, object properties and reconstruction dominate that separate claim.
Save the complete scan coordinates, energy calibration and beam geometry. An error in any of them changes the forward model applied to all 36 synthetic patterns.
CHAPTER 5 OF 12 · Check inversion stability
5. Phase information emerges from constraints, not a single detector count
At one detector sample, intensity is proportional to a complex amplitude’s squared magnitude. If two idealized coherent contributions each have unit amplitude and arrive in phase, their combined amplitude is 2 and squared magnitude is 4.
If the relative phase changes to π radians, the same two idealized contributions sum to zero and the intensity vanishes at that detector sample. Those amplitudes are invented to demonstrate interference, not to describe a complete measured diffraction pattern.
Reconstruction exploits the spatial pattern of many such coherent interferences, observed at overlapping scan positions. The intensity at one pixel cannot by itself reveal the relative phases of all contributing object features.
Even an ideal simplified blind-reconstruction model has ambiguities: replacing P by cP and O by O/c for any nonzero constant c leaves their product unchanged. External normalization or suitable conventions are needed to report absolute amplitudes.
Likewise, a global phase shift may leave recorded intensities unchanged. Relative spatial phase and physical interpretation are meaningful only with an explicit reference and suitable model constraints.
CHAPTER 6 OF 12 · Check inversion stability
6. Probe and position errors can imitate sample structure
The illumination function may change with optic drift, beamline fluctuations or alignment. A model that assumes the probe is constant across every scan may respond to such variation by placing artefacts into the inferred object.
Scan-stage errors, mechanical vibration and fly-scan motion alter the nominal positions or exposure trajectories. Small deviations matter when target spatial resolution is much finer than the illuminating spot.
Joint reconstruction algorithms can refine the probe and sometimes scan positions. This can improve consistency but creates additional estimation freedom; an unconstrained fit may explain an instrument error as a false object feature.
Record probe-shape diagnostics, stage metrology, timestamps, beam stability and reference scans. Compare independent scans or alternative geometry calibrations when quantitative phase or dimensions are reported.
Visible convergence of an iterative solver is not sufficient validation: different initializations or regularization may converge to different-looking but data-consistent solutions.
CHAPTER 7 OF 12 · Check inversion stability
7. Photon noise and partial coherence limit the useful information
Photon-counting detectors have counting variability; under an ideal Poisson model, standard deviation of a count N is approximately √N. A hypothetical mean of N = 10,000 photons in one pixel would have a counting standard deviation around 100, or 1% relative noise, before additional detector effects.
Many high-angle diffraction pixels may contain far fewer counts, with much greater fractional shot noise. Those pixels often carry information about finer details, so raw detector range and noise modelling are integral to a resolution claim.
Partial coherence, source bandwidth and mixed illuminating modes can reduce contrast. A single coherent-probe forward model may then produce systematic reconstruction artefacts despite a visually clean image.
Detector dark current, saturation, readout issues, beam stops, cosmic events and dynamic-range limits also shape the recorded patterns. A diffraction pattern with masked or missing regions contains less information than an ideal full-field model assumes.
Radiation exposure can change sensitive samples during scanning. A reconstruction built from evolving specimen states may violate the single fixed-object assumption; dose and repeated-exposure evidence belong in the method.
CHAPTER 8 OF 12 · Interpret reconstructions
8. Overlap and reconstruction depend on the whole acquisition
Overlapping positions create redundancy but do not automatically correct every instrumental problem. A poorly calibrated detector, incomplete positional coverage or unstable sample may break the physical assumptions supporting phase retrieval.
Algorithms can use sequential projection-style updates, simultaneous optimization or model-aware techniques that represent partial coherence, mixed states and position uncertainty. Different methods may have different robustness to noise and scan structure.
DESY publications discuss reducing scan overlap while retaining useful reconstructions under specific initialization methods. Such results do not mean every sparse scan is safe or that overlap no longer matters.
Compare fit residuals, reconstruction repeatability, known reference structures and the effect of changing regularization. A single final phase map is a product of measurements plus a chosen algorithm.
Where a claim depends on a small feature, show a measure of effective resolution tied to independent evidence instead of relying on pixel pitch or an aesthetic sharpness judgment.
CHAPTER 9 OF 12 · Interpret reconstructions
9. Reconstructed pixel size is not the achieved resolution
A real-space output grid is chosen from computational sampling and experimental geometry. A reconstruction can be resampled to smaller pixels without generating new high-angle scattering information.
Spatial resolution depends on measured angular bandwidth, coherence, signal-to-noise ratio, detector dynamic range, position accuracy, specimen stability and the validity of the reconstruction model.
Reporting ‘5 nm pixels’ does not establish that adjacent 5 nm structures were separately resolved. Resolution estimates may use reference targets, independent data splits, Fourier shell or ring correlation under appropriate assumptions, or other validated tests.
Another distinction is chemical specificity: an X-ray phase map constrains complex transmission under the measured energy and model. An exact elemental concentration typically requires additional spectroscopy, fluorescence or independent calibration.
Argonne’s Advanced Photon Source describes ptychography, X-ray fluorescence microscopy and tomography as complementary capabilities. A complex image can be part of a multimodal measurement but should not be mistaken for all modalities at once.
CHAPTER 10 OF 12 · Interpret reconstructions
10. Connect ptychography with phase contrast, tomography and fluorescence
Ptychography can map projected phase and amplitude. When repeated over suitable rotation angles and analysed with appropriate tomography models, it can contribute to three-dimensional structural estimates. That extension introduces alignment, dose, incomplete-angle and reconstruction problems.
X-ray phase-contrast imaging is a broader family that includes propagation-based and interferometric approaches. Ptychography is one coherent diffraction-based method; not every phase-contrast X-ray image is ptychographic.
X-ray fluorescence microscopy reports energy-dependent characteristic emissions associated with elements. Ptychography and fluorescence may be acquired together, but their spatial responses and scan requirements differ.
X-ray diffraction often addresses crystal order through angular distributions and indexing, while ptychography reconstructs a spatially varying complex response. The same sample may motivate both, but each supports different claims.
Keep a modality ledger: measured detector signal, reconstructed quantity, spatial and spectral resolution, target depth, dose and what independent comparison would falsify the interpretation.
CHAPTER 11 OF 12 · Follow X-ray methods
11. Practice, FAQs and changed-condition transfer
Practice A: For d = 800 nm and s = 200 nm in the same simple geometric model, overlap fraction is (800 − 200)/800 = 0.75, or 75%. Explain why this is a linear, not exact area, fraction.
Practice B: If each of 36 positions uses a hypothetical 0.1-second acquisition, pure exposure time totals 3.6 seconds. Real collection takes longer when scan motion, readout, alignment and additional exposures are included.
FAQ — Does ptychography measure phase directly? No. The detector measures diffraction intensity; phase is reconstructed through constraints and modelling.
FAQ — Is finer pixel spacing proof of higher resolution? No. Effective spatial resolution depends on measured information and independent tests.
FAQ — Can I reduce overlap without consequences? Possibly for a specific validated strategy; reconstruction stability and information content depend on more than a single percentage.
Transfer from a stable microelectronic specimen to a dose-sensitive biological sample. Revisit sample evolution, exposure planning, partial coherence, scan coverage and confidence in phase retrieval.
These connected eduKateSG learning routes address different stages of the observation, physics, modelling and verification chain. Each destination has been checked as an existing subject root.
- The Gold Standard of X-ray Optics — beam coherence and X-ray optical geometry
- The Gold Standard of X-ray Diffraction — diffraction measurements and crystal-structure questions
- The Gold Standard of X-ray Phase-Contrast Imaging — other phase-sensitive imaging models
- The Gold Standard of X-ray Computed Tomography — three-dimensional reconstruction and dimensional evidence
- The Gold Standard of Wave Propagation — wave phase, interference and propagation
- The Gold Standard of X-ray Fluorescence — element-sensitive complementary microscopy
- The Gold Standard of Statistics — model uncertainty and independent image validation
CHAPTER 12 OF 12 · Follow X-ray methods
12. Connected X-ray optics and reconstruction routes
Ptychography links coherent wave propagation to inverse reconstruction. A learner should understand wave phase and diffraction before treating phase maps as independently measured object properties.
Use the adjacent subject roots to distinguish instrument observables, materials interpretation and the statistical validation of a claimed improvement.
Primary, institutional and research references to test these explanations:
- DESY, Ptychography — coherent scanning and object–probe reconstruction
- Argonne Advanced Photon Source, PtychoProbe Beamline — imaging capabilities and complementary fluorescence techniques
- DESY, Object Initialization for Ptychographic Scans with Reduced Overlap — scan overlap and reconstruction initialization
- Ptychography with a Virtually Enlarged Illumination — overlapping diffraction patterns and probe–object estimation
- Efficient Ptychographic Phase Retrieval via Matrix-Free Optimization — inverse algorithms and modelling constraints
- Dual-Energy X-ray Ptycho-Fluorescence Imaging — overlap considerations and multimodal imaging
The thin-object model, photon statistics and nominal geometric overlap are explicitly illustrative. More advanced multislice, partial-coherence, beamline-specific or dose-limited methods require their own validated forward models.
Transfer requires a stated model, calculation, limitations and changed-context check. The immutable Clementi Secondary 1 Mathematics teaching reference exemplifies this instructional approach but is not a specialist source for X-ray Ptychography.