Abstract:For more than 100 years, X-ray crystallography has provided a unique view on the three-dimensional structure of atoms and molecules in crystals. However, to determine even the simplest structures now and a hundred years ago, one needs to overcome a mathematical hurdle for which the solution is not known even to this day. To reconstruct the 3-dimensional electron density map, from which the structure can be inferred, the complex structure factors F = |F| exp(iφ) of a sufficiently large number of diffracted reflections must be known. In a conventional diffraction experiment, only the amplitudes |F| are obtained, while the phases φ are lost. This is the crystallographic phase problem. Seventy years of research has established successful ab initio phasing methods such as direct methods and charge flipping. However, these methods are limited to atomic- resolution data, complicating structure determination from weakly-scattering crystals. Here, we show that a neural network can solve the crystallographic phase problem at a resolution of only 2 Å. We have developed an approach to generate millions of artificial structures and respective diffraction data for training of a neural network. We demonstrate that ab initio phasing based on this neural network is possible using 10 % to 20 % of the data needed for present-day methods, breaking the paradigm that atomic resolution is necessary for ab initio structure solution. The current neural network works in common centrosymmetric space groups and for modest unit cell dimensions, and suggests that neural networks can be used to solve the phase problem in the general case. This approach will enable structure solution for weakly-scattering crystals such as metal-organic frameworks or nanometer-sized crystals investigated using electron diffraction.

What problem does this paper attempt to address?

This paper attempts to solve an important problem in crystallography: the **Crystallographic Phase Problem**. Specifically, although X - ray crystallography has been able to provide a unique perspective on the three - dimensional structures of atoms and molecules in crystals, in order to determine these structures, a mathematical problem must be overcome - recovering phase information from experimentally measured diffraction data. ### Problem Background In traditional diffraction experiments, only the amplitude \( |F| \) of the diffraction reflection can be obtained, while the phase \( \phi \) is lost. This results in the inability to directly reconstruct the three - dimensional electron density map through Fourier transform and then infer the crystal structure. This problem is known as the **Crystallographic Phase Problem**. ### Limitations of Existing Methods Existing methods for solving the phase problem mainly include Direct Methods, Charge Flipping, etc. However, these methods usually require data with atomic resolution (i.e., \( d_{\text{min}}=1.2 \, \text{\AA} \) or better), which is a great limitation for weakly scattering crystals (such as metal - organic frameworks or nano - sized crystals). ### Innovations of the Paper The paper proposes a new method based on deep learning - **PhAI** - for solving the crystallographic phase problem. Specific contributions include: 1. **Effective use of low - resolution data**: PhAI can successfully solve the phase problem at a resolution of only \( 2.0 \, \text{\AA} \), and the amount of diffraction data required is only 10% - 20% of that of existing methods. 2. **Breaking the limit of atomic resolution**: Traditional methods consider atomic resolution as a necessary condition, while PhAI breaks this paradigm, indicating that structure analysis can be carried out even at lower resolutions. 3. **High - precision electron density map**: The electron density map generated by PhAI has extremely high accuracy, suggesting that a more robust structure analysis process can be developed. 4. **Applicable to multiple space groups**: Although the training mainly focuses on common centrosymmetric space groups (such as P21/c), PhAI also shows the potential for application in other common space groups (such as C2/c, Pbca, Pnma, Pbcn, C2/m). 5. **Application of experimental data**: In addition to simulated data, PhAI has also been tested on actual X - ray diffraction experimental data and has shown excellent performance. ### Summary By introducing deep - learning technology, this paper provides new ideas and tools for solving the crystallographic phase problem, especially for the structure analysis of low - resolution and weakly scattering crystals. This method not only improves the efficiency of structure analysis but also broadens the application range of crystallography research.

PhAI: A deep learning approach to solve the crystallographic phase problem

Automating crystal-structure phase mapping by combining deep learning with constraint reasoning

Towards end-to-end structure determination from x-ray diffraction data using deep learning

AlphaCrystal: Contact map based crystal structure prediction using deep learning

Artificial intelligence to solve the X-ray crystallography phase problem: a case study report

CrystalX: Ultra-Precision Crystal Structure Resolution and Error Correction Using Deep Learning

Ab Initio Structure Solutions from Nanocrystalline Powder Diffraction Data

Solving the Crystallographic Phase Problem using Dynamical Scattering in Electron Diffraction

Using Phases Retrieved from Two-Dimensional Projections to Facilitate Structure Solution from X-Ray Powder Diffraction Data

Neural networks trained on synthetically generated crystals can extract structural information from ICSD powder X-ray diffractograms

End-to-End Crystal Structure Prediction from Powder X-Ray Diffraction

Paradigm shift in electron-based crystallography via machine learning

Neural Network-based Classification of Crystal Symmetries from X-Ray Diffraction Patterns

DiffraGAN: a conditional generative adversarial network for phasing single molecule diffraction data to atomic resolution

Direct recognition of crystal structures via three-dimensional convolutional neural networks with high accuracy and tolerance to random displacements and missing atoms

Machine learning assisted crystallographic reconstruction from atom probe tomographic images

Deep residual networks for crystallography trained on synthetic data

Crystallography companion agent for high-throughput materials discovery

A modified phase-retrieval algorithm to facilitate automatic de novo macromolecular structure determination in single-wavelength anomalous diffraction

Phase quantification using deep neural network processing of XRD patterns

AI-enabled high-resolution scanning coherent diffraction imaging