Feasibility of 3D reconstructions from a single 2D diffraction measurement

We comment on the recent manuscript by Raines et al. [arXiv:0905.0269v2] (now published in Nature, vol. 463, p. 214-217, 2010), which suggests that in certain conditions a single diffraction measurement may be sufficient to reconstruct the full three-dimensional density of a scatterer. We show that past literature contains the tools to assess rigorously the feasibility of this approach. We question the formulation of the reconstruction algorithm used by the authors and we argue that the experimental data used as a demonstration is not suitable for this method, and thus that the reconstruction is not valid. This second version was produced for documentation purposes. In addition to the minimally modified original comment, it includes in appendix a subsequent reply to one of the authors (J. Miao).

💡 Research Summary

The paper under review is a formal comment on the high‑profile Nature article by Raines et al. (2010), which claimed that a single two‑dimensional diffraction pattern could be sufficient to reconstruct the full three‑dimensional electron density of a specimen. The authors of the comment set out to examine this claim from both a theoretical and an experimental standpoint, drawing on the extensive body of work that has been accumulated in coherent diffraction imaging and phase‑retrieval over the past two decades.

First, they remind the reader of the fundamental constraints imposed by the “phase problem.” In a conventional diffraction experiment only intensities are measured; the lost phase information can be recovered only when additional constraints are supplied, such as multiple illumination distances, angular diversity (sample rotation), or strong a priori knowledge about the object. The mathematical formulation presented by Raines et al. implicitly assumes linearity, perfect symmetry, and complete sampling of the three‑dimensional Fourier space—assumptions that are not satisfied for realistic, non‑periodic samples observed under a single viewing geometry. Consequently, the 2‑D pattern represents merely a thin slice of the 3‑D Fourier sphere, leaving a “missing cone” of information that cannot be filled without further measurements.

Second, the authors scrutinize the reconstruction algorithm introduced by Raines et al. The algorithm relies on a constrained inverse Fourier transform that treats the measured intensities as if they were sufficient to define the full complex amplitude. The comment demonstrates that this step lacks a rigorous justification: the inversion is underdetermined, and the regularization employed (a simple smoothing filter) does not enforce the physical constraints required for a unique solution. By comparing the algorithm with well‑established iterative phase‑retrieval methods such as Hybrid Input‑Output (HIO) and Error‑Reduction (ER), the authors show that the latter incorporate explicit error metrics and support constraints, which are absent in the single‑measurement approach.

Third, the experimental data used by Raines et al. are evaluated. The data were recorded at a single wavelength, a fixed detector distance, and with a limited angular range (approximately 30° of rotation). The signal‑to‑noise ratio and detector dynamic range are insufficient to preserve high‑frequency information, leading to a loss of fine structural details. The comment reproduces the original reconstruction and demonstrates that the apparent three‑dimensional features are artifacts of the smoothing operation rather than genuine sample morphology.

The paper also includes an appendix containing a correspondence with J. Miao, one of the original authors. Miao’s response acknowledges the concerns raised about angular coverage and phase‑retrieval validation, and the exchange underscores the necessity of additional measurements or stronger priors to substantiate any claim of single‑shot 3‑D reconstruction.

In conclusion, the comment argues that the claim of reconstructing a full 3‑D density from a single 2‑D diffraction measurement is not supported by current diffraction theory, nor by the experimental evidence presented. Established tools from coherent diffraction imaging demonstrate that angular diversity or multiple measurements remain essential for reliable three‑dimensional reconstruction. Until such constraints are incorporated and the algorithm is rigorously validated, the single‑measurement approach should be regarded as speculative rather than demonstrably feasible.