Non-classical crystallization of CeO2 by means of in situ electron microscopy

During in situ liquid-phase electron microscopy (LP-EM) observations, the application of different irradiation dose rates may considerably alter the chemistry of the studied solution and influence processes, in particular growth pathways. While many processes have been studied using LP-EM in the last decade, the extent of the influence of the electron beam is not always understood and comparisons with corresponding bulk experiments are lacking. Here, we employ the radiolytic oxidation of Ce3+ in aqueous solution as a model reaction for the in situ LP-EM study of the formation of CeO2 particles. We compare our findings to the results from our previous study where a larger volume of Ce3+ precursor solution was subjected to γ-irradiation. We systematically analyze the effects of the applied irradiation dose rates and the induced diffusion of Ce ions on the growth mechanisms and the morphology of ceria particles. Our results show that an eight orders of magnitude higher dose rate applied during homogeneous electron-radiation in LP-EM compared to the dose rate using gamma-radiation does not affect the CeO2 particle growth pathway despite the significant higher Ce3+ to Ce4+ oxidation rate. Moreover, in both cases highly ordered structures (mesocrystals) are formed. This finding is explained by the stepwise formation of ceria particles via an intermediate phase, a signature of non-classical crystallization. Furthermore, when irradiation is applied locally using LP scanning transmission electron microscopy (LP-STEM), the higher conversion rate induces Ce-ion concentration gradients affecting the CeO2 growth. The appearance of branched morphologies is associated with the change to diffusion limited growth.


DOSE RATE ESTIMATION
The nominal dose rate in liquid phase transmission electron microscopy (LP-TEM) (cylindrical irradiated volume approximated, in experiment no perfect parallel beam, but slightly divergent leading to a capped cone) was estimated following Eq. 7.8 in [1]: (1) using a total stopping power (density normalized, collisional + radiative) taken from [2] ("water liquid") with an acceleration voltage of the electrons of 200 kV (thus, 200 keV energy) S = 2.789 MeV · cm 2 /g.
The beam current applied in the experiment was measured using of the calibrated fluorescence screen, when no sample was inserted, to: For that given beam area, the dose rate is estimated to: D TEM = 16.9 · 10 6 Gy/s or in electrons per area per time (electron flux instead of dose rate)

D TEM = 3.737 electrons/A 2 /s
The dose rate in LP-TEM was set by changing the "C3 condenser lens" current and thereby changing the convergence angle of the electron beam with respect to the sample surface (slightly more parallel for higher dose rate settings), thus, changing effectively the exposed liquid volume while the electron current is kept constant. Even though the approximation of the exposed volume as a cylindrical volume might not hold for a slightly converged beam, it is believed to be a good approximation when the sample is thin and the applied convergence little in comparison to the beam diameter.
In LP-scanning (S)TEM, the electron current of the beam was kept constant as well (gun and condenser settings) and measured on the calibrated (calibrated on both, electron counts and spatially) pixelated camera I STEM,Beam = 1.7 · 10 7 electrons/s = 2.72· 10 -12 C/s. A scanning electron beam induces an inhomogeneous electron current density within the scanned region, because the actual irradiation is local (beam on a scanning spot/step or not) and therefore time dependent. That is why the estimated dose rate can be defined in different ways. Either, the dose rate represents an average value over the entire scanned region (represented as an image, each scan spot/step refers to one pixel in the image where the intensity in a specific scattering range was acquired for a dwell time of the beam on those scan spots/steps), similar to the estimation in LP-TEM, here neglecting inhomogeneity caused by the scanning. This value is significantly higher than the dose rate calculated for applied LP-TEM or the dose rate averaged over the entire scanned region. However, it needs to be considered that this dose rate locally applied only for the dwell time of a single scan step (t dwell = 32 µs or less).

QUANTIFICATION OF PROJECTED GROWTH OBSERVATIONS IN IN SITU LP-TEM
For a comparable quantification of projected growth, measured objects (primary particles or aggregates) identified in the image have been manually assigned as ovals in Gatan's Microscopy Suite. The two main axes of the ovals have been measured automatically and the average is taken as the determined average diameter, which thereby underlies a coarse approximation as a sphere.
Uncertainties are estimated from the standard deviation of the measurements in an image and represent intrinsically the distribution of sizes to some extent.
The higher spatial resolution in the second liquid-cell (LC) (figure 2) enables the determination of primary particle size evolution (orange curve figure S1). This may be possible thanks to a significant thinner liquid sample. A thinner liquid sample would be in agreement with a lower density of formed aggregates (only 3 after more than 4 min and 30 s in figure 2 versus more than