This work investigates how knock-on displacements influence fluctuation electron microscopy (FEM) experiments. FEM experiments were conducted on amorphous silicon, formed by self-ion implantation, in a transmission electron microscope at 300 kV and 60 kV at various electron doses, two different binnings and with two different cameras, a CCD and a CMOS one. Furthermore, energy filtering has been utilized in one case. Energy filtering greatly enhances the FEM data by removing the inelastic background intensity, leading to an improved speckle contrast. The CMOS camera yields a slightly larger normalized variance than the CCD at an identical electron dose and appears more prone to noise at low electron counts. Beam-induced atomic displacements affect the 300 kV FEM data, leading to a continuous suppression of the normalized variance with increasing electron dose. Such displacements are considerably reduced for 60 kV experiments since the primary electron's maximum energy transfer to an atom is less than the displacement threshold energy of amorphous silicon. The results show that the variance suppression due to knock-on displacements can be controlled in two ways: Either by minimizing the electron dose to the sample or by conducting the experiment at a lower acceleration voltage.

Variable-resolution fluctuation electron microscopy (VR-FEM) data from measurements on amorphous silicon and PdNiP have been obtained at varying experimental conditions. Measurements have been conducted at identical total electron dose and with an identical electron dose normalized to the respective probe size. STEM probes of different sizes have been created by variation of the semi-convergence angle or by defocus. The results show that defocus yields a reduced normalized variance compared to data from probes created by convergence angle variation. Moreover, the trend of the normalized variance upon probe size variation differs between the two methods. Beam coherence, which affects FEM data, has been analyzed theoretically using geometrical optics on a multi-lens setup and linked to the illumination conditions. Fits to several experimental beam profiles support our geometrical optics theory regarding probe coherence. The normalized variance can be further optimized if one determines the optimal exposure time for the nanobeam diffraction patterns.

We compare experimental fluctuation electron microscopy (FEM) speckle data with electron diffraction simulations for thin amorphous carbon and silicon samples. We find that the experimental speckle intensity variance is generally more than an order of magnitude lower than kinematical scattering theory predicts for spatially coherent illumination.

We hypothesize that decoherence, which randomizes the phase relationship between scattered waves, is responsible for the anomaly. Specifically, displacement decoherence can contribute strongly to speckle suppression, particularly at higher beam energies. Displacement decoherence arises when the local structure is rearranged significantly by interactions with the beam during the exposure. Such motions cause diffraction speckle to twinkle, some of it at observable time scales.

We also find that the continuous random network model of amorphous silicon can explain the experimental variance data if displacement decoherence and multiple scattering is included in the modeling. This may resolve the longstanding discrepancy between X-ray and electron diffraction studies of radial distribution functions, and conclusions reached from previous FEM studies.

Decoherence likely affects all quantitative electron imaging and diffraction studies. It likely contributes to the so-called Stobbs factor, where high-resolution atomic-column image intensities are anomalously lower than predicted by a similar factor to that observed here.

Electron correlation microscopy (ECM) is a new technique that utilizes time-resolved coherent electron nanodiffraction to study dynamic atomic rearrangements in materials. It is the electron scattering equivalent of photon correlation spectroscopy with the added advantage of nanometer-scale spatial resolution. We have applied ECM to a Pd40Ni40P20 metallic glass, heated inside a scanning transmission electron microscope into a supercooled liquid to measure the structural relaxation time τ between the glass transition temperature Tg and the crystallization temperature, Tx. τ determined from the mean diffraction intensity autocorrelation function g2(t) decreases with temperature following an Arrhenius relationship between Tg and Tg+25 K, and then increases as temperature approaches Tx. The distribution of τ determined from the g2(t) of single speckles is broad and changes significantly with temperature.

Fluctuation electron microscopy can reveal the nanoscale order in amorphous materials via the statistical variance in the scattering intensity as a function of position, scattering vector, and resolution. However, several sources of experimental artifacts can seriously affect the magnitude of the variance peaks. The use of a scanning transmission electron microscope for data collection affords a convenient means to check whether artifacts are present. As nanodiffraction patterns are collected in serial, any spatial or temporal dependence of the scattering intensity across the series can easily be detected. We present examples of the major types of artifact and methods to correct the data or to avoid the problem experimentally. We also re-cast the statistical formalism used to identify sources of noise in view of the present results. The present work provides a basis on which to perform fluctuation electron microscopy with a high level of reliability and confidence in the quantitative magnitude of the data.

We examine simulated electron microdiffraction patterns from models of thin polycrystalline silicon. The models are made by a Voronoi tessellation of random points in a box. The Voronoi domains are randomly selected to contain either a randomly-oriented cubic crystalline grain or a region of continuous random network material. The microdiffraction simulations from coherent probes of different widths are computed at the ideal kinematical limit, ignoring inelastic and multiple scattering. By examining the normalized intensity variance that is obtained in fluctuation electron microscopy experiments, we confirm that intensity fluctuations increase monotonically with the percentage of crystalline grains in the material. However, anomalously high variance is observed for models that have 100% crystalline grains with no imperfections. We confirm that the reduced normalized variance, V(k,R) − 1, that is associated with four-body correlations at scattering vector k, varies inversely with specimen thickness. Further, for probe sizes R larger than the mean grain size, we confirm that the reduced normalized variance obeys the predicted form given by Gibson et al. [Ultramicroscopy, 83, 169–178 (2000)] for the kinematical coherent scattering limit.

We report variable resolution fluctuation electron microscopy (VRFEM) measurements on Cu64.5Zr35.5 metallic glass acquired using scanning transmission electron microscopy nanodiffraction using coherent probes 0.8 to 11 nm in diameter. The VRFEM results show that medium range atomic order structure of Cu64.5Zr35.5 bulk metallic glass at the ∼1 nm scale has large fluctuations, but the structure becomes almost completely homogeneous at the 11 nm scale. We show that our experimental VRFEM data are consistent with two different models, the pair persistent model and the amorphous/nanocrystal composite model. We also report a new way to filter VRFEM data to eliminate the effect of specimen thickness gradient using high-angle annular dark field images as references.