Condensed Matter Imaging

Imaging of condensed matter seeks to reveal structural, chemical and functional details across scales from the atomic to the mesoscopic. Techniques based on coherent X-ray and electron scattering – including diffraction-based lensless methods – exploit phase-sensitive detection to transcend the limits of conventional optics. Scanning schemes such as ptychography combine overlapping coherent diffraction patterns with iterative phase-retrieval to yield high-resolution, quantitative complex-amplitude maps without lenses. Tomographic variants extend these maps into three dimensions, visualising strain, composition and magnetic textures in bulk materials. Advances in X-ray free-electron lasers have introduced “diffraction-before-destruction” imaging at femtosecond timescales, capturing ultrafast dynamics in quantum materials. Parallel developments in detector technology, high-performance computing and machine-learning-driven inversion now allow streaming reconstruction at kilohertz rates, real-time feedback and imaging of highly radiation-sensitive specimens. Applications range from mapping strain and defects in nanodevices to visualising topological spin textures, offering direct insight into the interplay of structure, function and dynamics in functional materials.

Research from Nature Portfolio

Deep-learning-driven inversion of ptychographic diffraction has been demonstrated directly at the experimental beamline edge, enabling real-time streaming ptychography at up to 2 kHz. By training compact neural networks to approximate phase-retrieval constraints, the workflow removes oversampling requirements and reduces data volume by orders of magnitude, supporting low-dose in situ studies of functional materials and live cells with immediate feedback. A foundational advance in magnetic imaging has come from holographic vector-field electron tomography, which reconstructs all three components of magnetic induction in individual nanostructures with sub-10 nm resolution. Combining off-axis electron holography with tomographic tilt series and novel tensor-algebraic reconstruction, this approach maps 3D magnetisation in bulk nanomagnets and links morphology to local magnetic properties, guiding the design of magnetic devices.

Research from all publishers

A deep-learning framework for three-dimensional ptycho-tomography has achieved a 140× acceleration over conventional methods by training anisotropic convolutional priors to recover nanometre-scale volumes from a severely reduced set of angular projections. By incorporating spatial pyramidal pooling into the reconstruction network, researchers attained (14 nm)³ resolution with only 21 angles versus the nominal 349, demonstrating the potential of learned priors for rapid, large-volume imaging. In the realm of ultrafast diffraction, the first direct observation of single-protein scattering patterns has been reported using femtosecond X-ray free-electron laser pulses. Capturing snapshot diffraction from individual GroEL complexes before their destruction, this work validates “diffraction-before-destruction” at the scale of single macromolecules, paving the way to time-resolved studies of structural dynamics in biological and soft condensed matter.

Condensed Matter Imaging publication trend

The graph below shows the total number of articles in condensed matter imaging across all publications each year (not limited to Nature Index journals).

Technical terms

Coherent diffraction imaging (CDI): Lensless imaging technique using spatially coherent beams to record diffraction patterns, from which phase retrieval algorithms reconstruct real-space structure.

Ptychography: Scanning CDI variant in which an overlapping probe scans the specimen; redundancy in diffraction data enables robust phase-retrieval and quantitative complex-amplitude reconstruction.

Phase retrieval: Computational inversion of measured diffraction intensities to recover lost phase information, allowing the reconstruction of real-space images from far-field data.

Tomography: Three-dimensional imaging obtained by collecting two-dimensional projections over many angular orientations and solving an inverse Radon transform.

Diffraction-before-destruction: Strategy using ultrashort, high-intensity pulses (e.g., from free-electron lasers) to record diffraction data before sample damage occurs.

Deep-learning inversion: Use of trained neural networks to accelerate phase retrieval or tomographic reconstruction, reducing data requirements and enabling real-time processing.

References

1. Deep learning at the edge enables real-time streaming ptychographic imaging. Nature Communications (2023).
2. Holographic vector field electron tomography of three-dimensional nanomagnets. Communications Physics (2019).
3. Three-dimensional nanoscale reduced-angle ptycho-tomographic imaging with deep learning (RAPID). eLight (2023).
4. Observation of a single protein by ultrafast X-ray diffraction. Light: Science & Applications (2024).