Helium-Cooled Cryogenic STEM Imaging and Ptychography for Atomic-Scale Study of Low-Temperature Phases

Abstract

Much of the exotic functionality of prime interest in quantum materials emerges from structural and electronic ground states that can only be accessed at cryogenic temperatures. Understanding device operation therefore requires structural characterization under the same low-temperature conditions at which these functional phases exist, as room-temperature measurements often probe a different structural state. Achieving atomic-resolution in scanning transmission electron microscopy imaging and particularly 4D-STEM electron ptychography at liquid helium temperature has remained extremely challenging because even small amounts of drift, vibration, and thermal instability associated with the cryogen can disrupt the stringent stability requirements of atomic-resolution STEM. In this work we demonstrate atomic-resolution STEM and multislice electron ptychography at temperatures as low as 20 K using a commercial helium cooled holder. We find that rapid scans and a multi-stage registration workflow are critical to reducing artifacts associated with cryogenic instability for atomic-resolution imaging, while for ptychography scan position correction including compensation for coupling between probe aberrations and position refinement is necessary for successful reconstructions. Together these results establish a pathway for reliable atomic-resolution STEM and ptychography at low temperature, enabling direct visualization of structural ground states relevant to quantum technology.

Figures (5)

• Figure 1: Overview of the liquid helium atomic-resolution STEM experiment. Schematics of the (a) microscope optics and detectors and (b) sample holder, transfer line, and helium dewar illustrate the added instabilities in these conditions. The gray shaded region marks the vibration damping holder bellows, separating the holder tip from the cryostat marked with the gradient filled area. Cryostat and approximate tip temperatures for the datasets shown in (c--f) are marked. Registered and averaged images from rapid-frame (c) ADF- and (d) BF-STEM acquisitions of Fe$_3$B$_7$O$_{13}$I with expanded views shown below. (e) The Fourier transform of the ADF-STEM image. (f) A multislice electron ptychography reconstruction of a similar field of view to that in the expanded panels of (c) and (d), with greater sensitivity to the boron and oxygen sublattices.
• Figure 2: Effects of thermal and mechanical instability on STEM imaging. (a) A cartoon visualization of the separation of instability in the system into large unidirectional drift associated with a temperature gradient along the holder rod (left) and the residual instabilities from the cryogen, room environment, etc. (right), not to scale. (b) Plot of the scan times associated with conventional STEM imaging and 4D-STEM acquisitions, with the scan times used in this work marked with arrows. The distortion frequencies higher than the scan rate are shaded below. (c) Schematic illustrating the effect of instability with a frequency lower than the scan resulting in to 1st order rigid shifts between sequential frames (left), and higher frequency instability which causes distortion within individual frames, primarily on the slow scan (horizontal) direction (right). (d) Plot of the cryostat temperature over a day. Showing steps where the temperature was adjusted and plateaus where it was held constant for data collection. (e) Sample positions plotted over various 20 s time periods. Each period was optimized for atomic-resolution imaging, the cryostat temperature was constant and the stage had not been recently moved.
• Figure 3: ADF-STEM imaging performance on a SrTiO$_3$ epitaxial thin film on GdScO$_3$. (a) A single rapid scan acquired at $\sim$40 K shows significant intra-frame distortion. (b) The corresponding FFT shows significant scan artifacts, recognizable as vertical lines and anomalous peaks vertically in line with the Bragg peaks. (c) A registered and averaged image with high signal to noise, largely free from scan distortions. (d) The corresponding FFT, showing sub-ångström information transfer.
• Figure 4: Suitability of registration approaches applied to Fe$_3$B$_7$O$_{13}$I. (a) A single rapid scan showing severe distortions in the ADF-STEM image (top) and corresponding FFT (bottom) from instability. Insets in the FFT expand the area around two peaks: in teal, a strong reflection elongated to a line by the scan distortions, in gold a peak too faint to measure from a single scan. (b) The result of rigid registration with optimized real and reciprocal space filtering and transitivity constrained shift refinement. Scan distortions and artifacts in the FFT are reduced, but not eliminated entirely. (c) The result of non-rigid registration, applied to the rigidly registered result from (b). Information transfer is improved (see: gold inset) however artifacts from scan distortions present in the rigid registration result remain. (d) The result of a flawed non-rigid registration. The image series was not correctly rigidly registered prior to non-rigid optimization. The image and FFT show similar enhanced contrast and information transfer to (c), however, very serious scan artifacts are present in the real space image and FFT.
• Figure 5: Liquid helium multislice electron ptychography on Fe$_3$B$_7$O$_{13}$I. (a) The virtual bright field image from a defocused 64 $\times$ 64 point 4D-STEM scan. Shear in the nominally approximately square lattice of visible iodine sites is clearly visible. (b) A set of CBED patterns from a set of 3 $\times$ 3 points in the larger scan. An atomic column clearly visible in the shadow images formed in the bright disk of the CBED patterns moves with the scan relative to its initial position marked in orange. (c) Corrected scan positions following a global affine transform identified from hyper-parameter tuning with Bayesian optimization, and additional gradient based individual refinement of each scan point during ptychographic reconstruction. (d) The object phase reconstructed from a conventional probe initialized with only defocus, showing a severe shearing and distortion. (e) A single CBED pattern annotated to show the large apparent shear angle between two nominally nearly perpendicular atomic planes. (f) Plots of the probe phase (left) and phase gradient (right) with an aberration function including only defocus (top) and a combination of defocus and astigmatism (bottom). (g) Reconstruction error measured for different reconstruction hyper-parameter sets, plotted as a function of shear angle applied to the scan positions. Results with a probe initialized with defocus only are shown in blue, while results from a probe initialized with defocus and astigmatism are plotted in red. Lines showing the approximate optimal fronts of each setting are shown as guides for the eye. (h) The object phase reconstructed object from a probe initialized with defocus and astigmatism, resulting in a high-quality unsheared reconstruction.