Coherent Diffractive Imaging Iterative Phase Retrieval (ePIE)

X-ray Ptychography & STXM Nanotomography Lab

Model coherent X-ray probe formation, scanning overlap redundancy constraints, far-field Fraunhofer diffraction patterns, iterative phase retrieval (ePIE/rPIE), quantitative complex refractive index mapping \(n = 1 - \delta + i\beta\), sub-2nm 3D tomography, and 20-year cleanroom capital budgeting DCF.

Imaging Modality
PXCT (Ptychographic Nanotomography)
Spatial Resolution: Sub-2 nm (<2.0 nm) Non-Destructive
Industrial Metrology Presets:
Achieved Resolution
1.85 nm
Gain: 16.5× Over Lens
Ptycho Overlap
70.5%
Redundancy verified (>60%)
Probe Size (FZP)
30.5 nm
Fresnel zone plate focus
ePIE Fourier Error
2.45%
Phase retrieval converged
Tomogram Scan Time
4.2 hrs
360 angle projections
20-Yr DCF NPV
$88.20M
IRR: 120% | 1.1 yrs

X-ray Optics & Scan Controls

7.2 keV
Penetration depth vs resonant absorption edge tuning.
25 nm
Governs physical focused probe spot size \(d_{\text{probe}} \approx 1.22 \Delta r_N\).
9 nm
Must satisfy overlap condition (\(\ge 60\%\)) for ePIE phase recovery.
1.85 m
Sets maximum numerical aperture and detector pixel sampling.
120 iters
360 proj

Cleanroom Fab Economics & DCF

50000 /mo
38 bps
Basis points (bps) of yield saved via non-destructive 3D void & strain inspection.

Optics, Focused Probe & Overlap Scheme

Coherent Wave Propagation
Fresnel zone plate focuses coherent X-rays onto the specimen. Overlapping raster positions (\(\ge 60\%\) overlap) generate redundant far-field Fraunhofer speckle patterns on the 2D pixel detector.

Diffraction Pattern & Reconstructed Phase Map

Diffraction speckles contain Fourier magnitude \(|F|\); ePIE phase retrieval reconstructs sub-2nm quantitative phase shift \(\Delta\phi(x,y)\).

Ptychographic Physical Parameters

Physical Metric Calculated Value

Cleanroom DCF Cash Flow Projection

Period Gross Benefit Opex Net Cash Flow Discounted Cumulative

X-ray Ptychography & STXM Nanotomography Physics Guide

1. Coherent Illumination & Abbe Limit Inversion

In standard transmission X-ray microscopy (TXM), image resolution is governed by the Abbe diffraction limit of the objective lens (\(\delta_{\text{lens}} \approx 1.22 \Delta r_N\)). High-aspect-ratio lithography restricts outermost zone widths \(\Delta r_N\) to 15 to 30 nm. Ptychography circumvents this lens barrier by recording coherent diffraction patterns \(I_j(\mathbf{k})\) directly on a framing detector without an objective lens:

\[ I_j(\mathbf{k}) = \left| \mathcal{F}\left\{ O(\mathbf{r} - \mathbf{R}_j) P(\mathbf{r}) \right\} \right|^2 \]

where \(O(\mathbf{r})\) is the complex specimen transmission function and \(P(\mathbf{r})\) is the coherent illumination probe. Because the phase is recovered numerically, the achievable resolution is limited only by the maximum recorded scattering vector: \(\delta_{\text{res}} = \lambda / (2 \sin \theta_{\text{max}})\), routinely reaching sub-2 nm.

2. Overlap Constraint & ePIE Phase Retrieval

To invert the phase problem without prior specimen support bounds, adjacent scan positions \(\mathbf{R}_j\) must overlap by at least 60% (typically 65% to 85% area overlap). The extended Ptychographical Iterative Engine (ePIE) iteratively updates both the object \(O(\mathbf{r})\) and probe \(P(\mathbf{r})\):

\[ O_{n+1}(\mathbf{r}) = O_n(\mathbf{r}) + \alpha \frac{P_n^*(\mathbf{r})}{|P_n(\mathbf{r})|_{\text{max}}^2} \left[ \psi_{\text{new}}(\mathbf{r}) - \psi_n(\mathbf{r}) \right] \]

where \(\psi_{\text{new}}(\mathbf{r}) = \mathcal{F}^{-1}\{ \sqrt{I_j(\mathbf{k})} \exp(i \arg[\mathcal{F}\{\psi_n\}]) \}\). This dual object-probe update corrects for real-time beam drift, zone plate aberrations, and partial coherence.

3. Quantitative Complex Refractive Index

X-rays interact with matter through the complex refractive index \(n(\mathbf{r}) = 1 - \delta(\mathbf{r}) + i \beta(\mathbf{r})\). The reconstructed complex transmission function yields independent quantitative channels for phase shift \(\Delta\phi\) and linear absorption \(\mu\):

\[ \Delta\phi(\mathbf{r}) = -\frac{2\pi}{\lambda} \int \delta(\mathbf{r}, z) dz, \quad \mu(\mathbf{r}) = \frac{4\pi}{\lambda} \int \beta(\mathbf{r}, z) dz \]

In hard X-rays (7 to 15 keV), \(\delta \gg \beta\) (phase shift is 100 to 1,000 times stronger than absorption), enabling high-contrast imaging of low-Z materials (photoresists, carbon nanotubes, electrolytes) and atomic channel strain in thick silicon wafers.

4. 3D Ptychographic Computed Tomography (PXCT)

By rotating the specimen through 180° or 360° and acquiring 2D ptychographic phase projections at each angle, filtered backprojection or iterative SIRT algorithms reconstruct a 3D isotropic voxel map of the electron density. According to the Crowther criterion, the angular sampling required is:

\[ N_{\text{angles}} \ge \frac{\pi \cdot D}{\delta_{\text{res}}} \]

PXCT provides non-destructive 3D inspection of packaged microchips, resolving sub-nanometer voids in Cu-Cu hybrid bonding pads without mechanical micro-cleaving.

Frequently Asked Questions

X-ray ptychography is a coherent diffractive imaging (CDI) technique that overcomes the manufacturing limits of X-ray optics (such as Fresnel zone plates, which are limited to 15 to 30 nm resolution by outermost zone fabrication). A finite coherent probe illuminates adjacent overlapping regions of a specimen (typically with >60% spatial overlap). At each raster point, a 2D framing detector records the far-field Fraunhofer diffraction pattern. Because the phase problem is solved computationally using iterative phase retrieval algorithms (such as ePIE or rPIE), the achievable spatial resolution is limited only by the maximum scattering angle recorded on the detector (delta_res = lambda / (2*sin(theta_max))), achieving sub-2 nm resolution non-destructively.

In conventional coherent diffractive imaging, isolating the complex exit wave from a single diffraction pattern requires strict specimen support constraints. Ptychography replaces artificial support constraints with experimental overlap redundancy: each specimen patch is illuminated multiple times by different portions of the shifted probe function P(r - R_j). This shared spatial information provides massive mathematical redundancy, enabling the extended Ptychographical Iterative Engine (ePIE) to reconstruct both the specimen transmission function O(r) and the complex illuminating probe P(r) simultaneously without prior knowledge of probe aberrations.

The reconstructed 2D complex transmission function is O(r) = exp(-2*pi*beta*t / lambda + i * 2*pi*delta*t / lambda), where n = 1 - delta + i*beta is the complex refractive index. The phase shift delta reflects electron density and mass density, providing extreme sensitivity to light elements (carbon, oxygen, silicon) even in thick specimens where absorption contrast beta is negligible. Meanwhile, tuning the incident X-ray photon energy across elemental absorption edges (such as Ni K-edge or Fe L3-edge) provides resonant spectroscopic chemical contrast.

3D Ptychographic Computed Tomography (PXCT) combines 2D ptychographic projections acquired over hundreds of sample tilt angles with tomographic backprojection reconstruction. Unlike transmission electron microscopy (TEM), which requires cutting the chip into fragile sub-50 nm lamellae with focused ion beams (destroying packaging strain and causing warping), hard X-ray PXCT penetrates through intact multi-micron semiconductor dies, non-destructively imaging 3D Gate-All-Around nanosheets, copper hybrid bonding micro-voids, and through-silicon vias (TSVs) with isotropic sub-5 nm 3D voxels.

A high-throughput laboratory or beamline ptychography inspection station (equipped with coherent high-brightness X-ray source, high-speed hybrid pixel detector like EIGER or JUNGFRAU, interferometric laser piezo stage, and multi-GPU reconstruction cluster) requires $4,800,000 in capital investment with $240,000 in annual operating costs. In advanced 2nm logic and 3D chiplet fabs processing 50,000 wafer starts per month, non-destructive root-cause defect elimination accelerates node yield ramps by 38 basis points, avoiding $18M to $35M in wafer scrap, delivering discounted payback in 1.1 to 1.3 years and a 20-year net present value exceeding $80M.

Related Advanced Metrology & Semiconductor Simulations

LEEM & PEEM Lab

Model cathode objective immersion retarding lenses, monoatomic step phase contrast, and 2D Fabry-Pérot LEEM-IV layer counting.

XSW Atomic Site Lab

Triangulate dopant and impurity positions within crystal lattices via X-ray dynamical standing wave fields.

Submicron X-ray CT Lab

Model cone-beam micro-CT non-destructive volumetric packaging inspection and void tomography.

Neutron Depth Profiling Lab

Model non-destructive nuclear reaction analysis, stopping power, operando lithium dendrites, and gigafactory DCF.