PhaseT3M: 3D Imaging at 1.6 Å Resolution via Electron Cryo-Tomography with Nonlinear Phase Retrieval

Electron cryo-tomography (cryo-ET) enables 3D imaging of complex, radiation-sensitive structures with molecular detail. However, image contrast from the interference of scattered electrons is nonlinear with atomic density and multiple scattering further complicates interpretation. These effects degrade resolution, particularly in conventional reconstruction algorithms, which assume linearity. Particle averaging can reduce such issues but is unsuitable for heterogeneous or dynamic samples ubiquitous in biology, chemistry, and materials sciences. Here, we develop a phase retrieval-based cryo-ET method, PhaseT3M. We experimentally demonstrate its application to a ~7 nm Co3O4 nanoparticle on ~30 nm carbon substrate, achieving a maximum resolution of 1.6 Å, surpassing conventional limits using standard cryo-TEM equipment. PhaseT3M uses a multislice model for multiple scattering and Bayesian optimization for alignment and computational aberration correction, with a positivity constraint to recover ‘missing wedge’ information. Applied directly to biological particles, it enhances resolution and reduces artifacts, establishing a new standard for routine 3D imaging of complex, radiation-sensitive materials.

💡 Research Summary

**
The paper introduces PhaseT3M, a novel cryo‑electron tomography (cryo‑ET) workflow that overcomes the intrinsic non‑linearity and multiple‑scattering effects that limit resolution in conventional CTEM‑based tomography. Traditional reconstruction algorithms assume a linear projection model, which fails for thick or high‑Z specimens where scattered electrons interfere in a complex, nonlinear fashion. PhaseT3M addresses this by employing a multislice forward model that explicitly simulates the propagation of the electron wave through a stack of thin slices, thereby accounting for multiple scattering events.

The authors integrate this physical model with a Bayesian optimization framework that simultaneously refines tilt‑series alignment, tilt‑axis geometry, contrast‑transfer function (CTF) parameters, and other instrumental aberrations. By treating these parameters probabilistically, the method efficiently explores the high‑dimensional parameter space and quantifies uncertainties, enabling robust convergence even under low‑dose conditions.

A key innovation is the enforcement of a positivity constraint on the reconstructed electrostatic potential (the potential must be non‑negative). This physical prior acts as a regularizer that suppresses missing‑wedge artifacts—gaps in Fourier space caused by the limited tilt range—and restores high‑frequency information that would otherwise be lost. In experiments on a ~7 nm Co₃O₄ nanoparticle supported on a ~30 nm amorphous carbon film, PhaseT3M achieved a maximum resolution of 1.6 Å, as evidenced by clear {422} diffraction spots and the recovery of {111} and {220} reflections within the missing‑wedge region.

Dose‑dependence studies showed that while the ultimate resolution degrades with decreasing total electron dose (from 8 460 e⁻/Ų down to 118 e⁻/Ų), structural information up to ~5.8 Å remains detectable even at the lowest dose, demonstrating the method’s suitability for highly radiation‑sensitive specimens. Simulated tilt series, generated under identical experimental conditions and contaminated with Poisson noise, confirmed the experimental findings: the multislice‑based PhaseT3M yields lower reconstruction error and sharper Fourier peaks than a single‑slice approach, and approaches the ground‑truth potential in the infinite‑dose limit.

When compared with a conventional simultaneous iterative reconstruction technique (SIRT), PhaseT3M not only reconstructs the carbon support accurately—something SIRT fails to do—but also eliminates ring‑shaped artifacts arising from nonlinear interference. This underscores that solving the inverse scattering problem, rather than merely correcting for multiple scattering, is essential for high‑fidelity 3D reconstructions.

Beyond inorganic materials, the authors applied PhaseT3M to the publicly available EMPIAR‑10164 dataset of HIV‑1 particles. The PhaseT3M reconstructions displayed 20‑50 % lower R‑factors and higher Fourier Ring Correlation values compared with standard cryo‑ET pipelines, confirming the method’s applicability to biological macromolecules.

In summary, PhaseT3M combines a rigorous multislice scattering model, Bayesian parameter optimization, and a positivity prior to deliver atomic‑scale (1.6 Å) 3D imaging using standard cryo‑TEM hardware without hardware aberration correction. Its ability to handle thick, heterogeneous, and radiation‑sensitive samples positions it as a new benchmark for routine high‑resolution cryo‑ET across materials science, chemistry, geology, and structural biology.