Monitoring bromide effect on radiolytic yields using in situ observations of uranyl oxide precipitation in the electron microscope

During electron microscopy observations of uranium-bearing phases and solutions in a liquid cell, the electron beam induced radiolysis causes changes in the chemistry of the system. This could be useful for investigating accelerated alteration of UO2 and can be also used to monitor radiolytic effects. Low concentrations of bromide in aqueous solutions are known to reduce the generation rate of H2O2 during radiolysis and increase H2 production. We deduced the presence of radiolytic H2O2 by monitoring the formation of a uranyl peroxide solid from both solid UO2 and a solution of ammonium uranyl carbonate at neutral pH. Additionally, the effect of bromine on water radiolysis was investigated through chemical modelling and in situ electron microscopy. By measuring the contrast in the electron microscopy images it was possible to monitor H2O2 formation and diffusion from the irradiated zone in agreement with the models.


Introduction
Electron-beam-induced processes that occur during liquid cell in situ electron microscopy can drive pH changes, reduce metal ions, through the generation of reactive radiolytic species. [1][2][3][4] Along with ionization, the interaction of an energetic electron beam with water molecules will generate very short-lived (10 À15 s) electronic excitations that favourably de-excite through intermediate atomic and molecular radicals. The reaction of these radicals with the surrounding aqueous environment occurs on the scale of 10 À9 s, resulting in six dominant species (H 2 O 2 , H 2 , OHc, Hc, e aq À , and H + ). 5 The production rate of these species is described by a 'G-value' that represents the number of species produced per 100 eV deposited in the irradiated solution within a small volume, termed the radiation spur. The Gvalue depends on the linear energy transfer (LET) of the ionizing particle travelling through the medium. As the species form, they will also diffuse rapidly through the uid phase, undergoing other reactions. Oxygen which is not a direct radiolytic product is produced outside the radiation spur; however, its magnitude typically matches the production rate of the other molecular species (H 2 and H 2 O 2 ). The radiolytic processes that occur during liquid cell (LC) electron microscopy (EM) and can lead to many chemical effects. It is important that any observations are also supported by other methods that do not create these reactive species or with an understanding of how radiolysis will affect the LC. Nevertheless, the direct study of electron beam effects can provide intriguing insight into the chemistry of a system and that may help to solve the complexities of mineral dissolution. [6][7][8]

Materials and methods
Experiments were conducted with both solid and dissolved U, and a control study was performed with DI water and solutions containing bromide as described below. For the dissolved U experiments, an approximately 10 mM solution of ammonium uranyl carbonate (AUC), [(NH 4 ) 4 UO 2 (CO 3 ) 3 ] (522.2 g mol À1 ) was prepared by dissolving 0.0529 g of the uranyl phase in 10 mL of DI water. AUC is an important component in the industrial conversion of UF 6 to UO 2 . The bromine solutions were made as follows: 10 À2 M solution of KBr (119.0 g mol À1 ) was produced by dissolving 0.0153 g of KBr in 10 mL of DI water. To make the 10 À4 M KBr solution, 0.1 mL of the stock 10 À2 M solution was diluted to 10 mL. 1 mL of this solution was diluted to 10 mL to yield the 10 À5 M solution, and 1 mL of this solution was diluted to 10 mL to yield the 10 À6 M solution.
For the experiments with solid material, a UO 2 powder was used (<45 mm size sieved material) that was subsequently crushed between two glass slides to create a < 5 mm particles which were suspended in deionised (DI) water.
Experiments were conducted with 6 mL of solution pipetted into a fresh QX-102 WetSEM sample cell (QuantomiX, Rehovot, Israel). These were examined at 15 keV and with a magnication of 100 000Â using an FEI (Thermo Fisher Scientic, Inc., Hillsboro, Oregon, USA) Quanta 250FEG scanning electron microscope. Irradiations were performed by rastering the beam over a 1 mm 2 area and taking backscattered electron (BSE) images every few seconds.
The irradiation of the beam over a specic area could be accomplished in the microscope by selecting a specic area and allowing the beam to raster continuously. Over time, a white area would form, an image would be taken over a larger area encompassing a much larger area than was irradiated. The images revealed bright regions within the irradiated region. To determine the amount of material in these areas, a line histogram was generated as shown in Fig. 1. The integrated intensity yielded a value that was used to represent the amount of material precipitated. The brightness/contrast settings were kept constant during the irradiations. The background contrast level was the same in each experiment and this value was subtracted from the result to yield the integrated intensity from the presence of the peroxide phase.
A beam current of 10.5 pA was measured with Keithley Instruments (Tektronix Inc., Beaverton, Oregon, USA) 485 microvoltmeter. This was obtained with a spot size of 1.5 and the smallest available condenser aperture during both image collection and irradiation. This dose prevented rupture of the liquid cell window. The total dose rate was determined to be 7.28 e À per nm 2 , assuming the irradiated region was 3 mm Â 3 mm. As the electron beam was completely stopped within 5 mm, the estimated dose was 3.5 Â 10 6 Gy s À1 .
The details of the measurement process are described in the ESI † section. Images were analysed with Gatan Digital-Micrograph 3.0 (Gatan Inc., Pleasanton, California, USA). 9 A histogram of intensity was constructed from each image across the forming precipitates and was used as a means to measure H 2 O 2 concentration. Nikon Elements 4.0 (Nikon Instruments, Inc., Melville, New York, USA) was used to determine the area of the features in the images.
Electron scattering effects were modelled using the CASINO program 10 using the cross-section models by Browning et al. 11 that were more applicable to the SEM beam energies.

Radiolysis model
Water radiolysis generation models have been developed by Shoesmith and co-workers 12,13 and Wittman et al., 14 for UO 2 dissolution (see Fig. 2); and Schneider et al., 4 for modelling the effects of radiolysis during in situ LC-EM. All models use rate constants from Elliot and McCracken 15 to solve a series of equations with stiff ordinary differential equation (ODE) solvers. The code by Wittman and co-workers 14 uses the FORTRAN ODEPACK solvers developed at the Lawrence Livermore National Laboratory. The Schneider model uses MATLAB (MathWorks, Natick, Massachusetts, USA) solvers (ode15s) that are also based on the ODEPACK routines. The (Curtiss-Hirschfelder) stiff ODE solvers (MATLAB ode15s) are orders of magnitude faster than non-stiff (Dormand-Prince) solvers and MATLAB (ode45), for these types of problems that have orders of magnitude differences in many of the chemical rate equations to be solved. Like the Schneider model, the model used in this study also included the diffusion of species from the area of irradiation, but also included the effects of carbonate and halides. ‡ As the 15 keV electron beam is completely stopped within 3 mm of water (based on CASINO models (see Fig. 3)), the effective LET for 15 keV in this system is comparable to that b-particles from tritium ( 3 H) where the E max ¼ 18.6 keV and the mean energy $ 5.7 keV. Butarbutar and co-workers have suggested that the effective high LET of 3 H b-radiolysis results in G-values that are similar to that of neutrons. 16 In this study, we adopted high LET G-values rather than those used for b/g-radiation. All radiation spurs would occur within this small volume and, hence, molecular products would be favoured over radicals in this environment, as is the case with neutron or a-radiolysis. The G-values used in the model are listed in the ESI. † At higher Fig. 1 Method for determining the production of U-phase during irradiation. Fig. 2 Predicted radiolytic species production rates for both a and b/g radiation. The black lines corresponds to using the rate constants from ref. 15. a-Radiation results in a continuous production rate of H 2 and H 2 O 2 ; whereas, production under b,g-radiation reaches a steady-state condition. ‡ The radiolysis model is described in the ESI section and in several publically available reports referenced in the ESI.
keV there is greater scattering in regions outside of the 1 mm wide beam irradiated zone but this occurs deeper into the solution (see Fig. 3).

Results of computational modelling
The coupled kinetics/diffusion rate equations for H 2 O 2 on discrete special zones (n) can be expressed in terms concentrations [H 2 O 2 ] n , uxes J n and dose rate d according to: Assuming nonzero reaction kinetics and dose-rate only in the radiation zone (x R ) with diffusion out to the boundary (x B ), the steady-state solution to eqn (1) aer inserting Fick's law uxes containing diffusion constant D and boundary concentration [H 2 O 2 ] B can be written: Eqn (2) serves as a working denition of a "conditional" Gvalue (G c ) for [H 2 O 2 ] which describes H 2 O 2 generation conditional on the local water chemistry. As compared with the radiolysis model with all approximately 100 reactions to fully describe the system, 15  The effect of even small concentrations of Br À can be seen in Fig. 4 where there is a decrease in the concentration of H 2 O 2 . As the H 2 O 2 concentration decreases, H 2 will increase. One question is whether this increasing H 2 affects subsequent radiolysis. If the initial H 2 concentration is high in the in situ LC, H 2 would compete with Br À for OHc. The model indicates that even at an initial concentration of 100-400 mM of Br À , the presence of H 2 should bring G c from 0.1 to nearly 1 molecule/100 eV. In the experiments, we did not see evidence for this type of process and the results t models where the initial concentration of H 2 is low. During radiolysis, the secondary species, O 2 , is also generated. In these experiments, the solutions were in equilibrium with atmospheric O 2 . In the ESI, † the effect of initial concentrations of O 2 is described. It is always important to consider the initial conditions. However, the models suggest that if the system was initially devoid of O 2 , there would be no difference in the H 2 and H 2 O 2 concentrations at steady state.
The radiation source is a surface or point, and allows for the radiolytic species to diffuse away from the irradiation region until steady-state conditions are reached distant from the source. 19 This creates a different environment to a homogeneous eld. In the electron microscope, the radiation source is also spatially isolated which may lead to different results compared to uniform irradiation.
The main reason for the effect is reaction 94 in Table 1. By competing for the $OHc, Br À disables the mechanism for

Monitoring bromide effect in the liquid cell
In the in situ LC, the formation of a uranyl peroxide exhibited strong bright BSE contrast that was observed within seconds in the DI water case. Fig. 5 is a series of SEM-BSE images that show the development of a uranyl peroxide precipitate in the central region where the irradiation was performed.
As the irradiation proceeded, material tended to precipitate far from the region of irradiation. This could be due to multiple scattering of the electrons leading to events outside the region irradiated directly or it could be from the diffusion of the longer-lived H 2 O 2 . We performed modelling of the electron scattering using the CASINO program 10 to show the effects of different beam energies on electron scattering in the water. At 15 keV, the electrons are completely stopped within 3-5 mm.
Resulting in an average LET close to 10 keV mm À1 , similar to that of neutrons or tritium. As the beam energy is raised to 30 keV, the electrons penetrate much deeper and there is more scattering. The higher energy beam would be expected to have a lower overall LET, and would have radiolytic production distributions that were more similar to b/g-sources. The effective of LET of electron beam sources is discussed further in the ESI. † Similar time-resolved irradiations were performed with different concentrations of bromide (10 À4 M, 10 À5 M, and 10 À6 M) with ammonium uranyl carbonate and monitored with BSE images. The integrated BSE contrast across the collected images    was summed and plotted in Fig. 7. These results clearly show the dramatic effect of bromide on H 2 O 2 production. As the bromide concentration was increased to 10 À4 M, the bright contrast from the uranyl peroxide precipitate reduced in intensity signicantly. Even at low bromide concentrations (10 À5 M and 10 À6 M) there was a slow-down in the production of H 2 O 2 based on the reduction in the contrast from the U-peroxide phase.
In Fig. 7, colourised/heat SEM-BSE images of the solutions with the different bromide solutions are all shown at approximately the same time of about 7 minutes of irradiation. The intensity of the image has been related to the concentration of U-peroxide. These images show, perhaps more clearly than the plot in Fig. 6, how the bromide inuenced the precipitation phenomena. As Br À increased, there was a reduction in the Uperoxide phase concentration. This is in agreement with the model where the steady state concentration of H 2 O 2 is impacted by Br À .
The addition of milli-molar amounts of Br À is commonly used to protect H 2 from scavenging by OH radicals in reaction. The overall impact of Br À , as well as other halides (I À and Cl À ), is reaction with cOH radical (see below). The rate constants for this forward reaction with Br À are faster than those for Cl À , at 1.1 Â 10 10 M À1 s À1 against 4.3 Â 10 9 M À1 s À1 . However, the reverse reactions, are considerably slower for cBrOH À (3.3 Â 10 7 M À1 s À1 ) than for cClOH À (4.3 Â 10 9 M À1 s À1 ). 20 cOH + Br À / cBrOH À cBrOH À / cBr + OH À cBrOH À + Br À / cBr 2 À + OH À Additional reactions of the cBr 2À species, results in the formation of Br 3 À . Furthermore, additional Br radical species reactions in the radiolytic system tend to regenerate Br À . Altogether, these reactions result in Br À becoming a much more potent scavenger of cOH radical even at much lower effective concentrations than Cl À . The e aq À and cH will also react with Br 3 À and cBr 2 À , resulting in the formation of H + and Br À .
Knowledge of the bromide concentration in contacting solutions is very important for understanding and predicting its potential impact on the radiolytic system. It is also possible that radioiodine from nuclear ssion may be present in the wastes debris at $10 À4 M. Experiments by Bauhn et al. 238 Pu-laced solutions containing solid UO 2 , do not agree with these results and models. 21 Their results showed that 1 mM bromide-bearing solutions had no effect on H 2 O 2 production. According to our radiolytic model, ‡ as the Br À content in the solution increased, H 2 O 2 production decreased and H 2 production increased (see Fig. 8). In this instance, the radiation source is modelled as emerging from a surface, and the concentrations represent those distant from the radiation source, such that the solution is no longer directly exposed to aparticles.
This might imply that the building H 2 concentration should affect further radiolysis and then prevent destruction of H 2 O 2 despite the presence of bromide. In the in situ LC experiments, we do not see any delay production in H 2 O 2 . This is because there is no H 2 at the start of the irradiation. If H 2 had been present initially, this reversing effect, should have occurred.

In situ oxidative corrosion of UO 2
Dissolution of particles of UO 2 was initially investigated by irradiating the UO 2 particles in solution to establish the conditions for both preserving the cells and obtaining suitable images that could be analysed. Similar work has been performed by Traboulsi et al. using He + ions 22 and with direct addition of H 2 O 2 . 23 These experiments resulted in the formation or meta-studtite from UO 2 corrosion. If H 2 is present in the system initially, the scavenging effect of Br À diminishes. In Fig. 9, UO 2 particles in DI water were irradiated continuously over a xed area (indicated by the red square in Fig. 9b). The production of H 2 O 2 from irradiation resulted in the dissolution of a portion of the UO 2 and the precipitation of a uranyl oxide phase outside of the irradiated region over the course of several minutes. No signicant changes were noted in the size of the initial UO 2 particles because of their relatively large mass. Although it was not possible to identify directly the phase of the newly-formed uranium-bearing particles, it is likely (based on particle morphology and knowledge of H 2 O 2 reaction with U) that these were the uranyl oxide peroxide phase studtite. The location of the studtite phase outside of the irradiated region points to the diffusion of H 2 O 2 and the importance of considering this effect during irradiation of solutions in the electron microscope cells. The CASINO electron scattering model for 15 keV electrons shows that very little electron scattering occurs >3 mm outside the irradiated region (see Fig. 3), supporting the contention that the material found outside the irradiated region is from the diffusion of H 2 O 2 and resulting reaction with U in solution.
Nuclear materials, when contacted with water, can induce radiolysis in the surrounding solution from their radioactive emissions. Uraninite (UO 2 ) is thermodynamically unstable under oxidizing conditions and combined with the alpha (a)radiation eld drive oxidative dissolution. [24][25][26][27] The oxidation process is controlled by the concentration of H 2 O 2 ; where the high oxidation potential will lead to the formation of soluble UO 2(aq)

2+
. 28 In the presence of high concentrations of H 2 O 2 , a secondary phase studtite [(UO 2 )(O 2 )(H 2 O) 2 $2H 2 O] may precipitate. [29][30][31][32] Indeed, studtite has been identied on the surface of spent UO 2 nuclear fuel. [33][34][35][36] Studtite is known to be stable in H 2 O 2 containing environments. 37 The recent studies by Bauhn et al. that show 1 mM bromide did not impact H 2 O 2 production in an alpha-dominated environment 21 may be due the spatial location of the radiation eld. We show that studtite was formed outside of the irradiated region in agreement with the ndings of Traboulsi et al. 22 for He + ions. It is unknown if direct irradiation would similarly allow the crystallisation of the secondary phase.

Conclusions
In situ LC electron microscopy was used to examine effect of different bromide concentrations on the production of H 2 O 2 by monitoring precipitation of a U-peroxide phases during the irradiation. We were able to corroborate a radiolysis model for H 2 O 2 generation by observing a reduction in H 2 O 2 concentration with added bromide; although, this is in disagreement with the work of Bauhn et al. 21 The study also points to the role of radiolysis in the electron microscope that can easily overwhelm a chemical system. However, with careful selection of beam current, beam energy, and dose, it is possible to conduct useful experiments with these types of liquid cell. We have shown that although it is possible to investigate radiolytic phenomena with the in situ LC, models need to include the diffusion of species to account for the nature of species migration in the cell and not just the steady state conditions under homogeneous irradiation.

Conflicts of interest
There are no conicts to declare.