One of the primary goals of fallout formation models is to predict the activity-size distribution of the resulting fallout. However, these models rely heavily on empirical data from a limited set of testing environments. Furthermore, fallout formation is known to be sensitive to the emplacement environment, particularly in the case when the fireball incorporates larges masses of surrounding material, as in the near-surface burst. The fidelity of these models in new and untested environments is not well constrained. As a result, there is a need to better understand the processes that control fallout formation. In near-surface bursts, the two primary processes that control radionuclide incorporation and the size distribution in fallout are condensation of species from the bomb vapor phase onto melts and agglomeration of melts to form larger melts.
Historically, fallout formation models have ignored agglomeration in predicting both the activity-size and size distribution of fallout. However, attempts at building thermodynamic models of radionuclide fractionation in the fireball show poor agreement with experimental data. Many have hypothesized that this disagreement is due to agglomeration—the collision and accretion of smaller parcels of melt with larger parcels of melt, both of which have incorporated relatively different amounts of radionuclides and when the whole object is dissolved and analyzed, leads to a mixing effect between the two objects. Furthermore, fallout in surface tests is known to form by several different mechanisms, including a small particle population known to form as "primary condensates'', that is, primarily derived from the vapor phase of fireball, which includes high concentrations of vaporized device material and likely some concentration of vaporized soil. Primary condensates have been observed to form up to diameters of 20 μm, within the range of the diameter of some observed agglomerates. As a result, it is also possible that some agglomerates observed in glassy fallout could be derived primarily from the fireball's vapor term, instead of being composed primarily of a mixture of the surrounding soil, as has been previously observed in aerodynamic glassy fallout.
Agglomerates are readily observed adhered to the exteriors of whole samples and agglomerates incorporated into samples are observable in sample cross-sections due to a distinct compositional interface between agglomerates and hosts. To determine the likely origin and role agglomerates play in forming aerodynamic glassy fallout, agglomerates and their host objects (the larger objects agglomerates are attached to) were studied starting with an initial population of 49 glassy fallout samples collected a similar distance from ground zero from a historic U-fueled U.S. nuclear test. After polishing to expose an internal cross-section, host objects and agglomerates in a subset of samples were characterized for major element composition using energy-dispersive X-ray spectroscopy (EDS) and for their U isotope ratios using secondary ion mass spectrometry (SIMS). Using these techniques, two datasets were collected. First, EDS analyses were used to characterize the major element composition across entire host objects (in 37 samples) and agglomerates (53 agglomerates in 15 samples) to compare the compositions between the two populations to determine if agglomerates are likely derived from similar melts relative to host objects. Second, EDS raster analyses were used to characterize the major element compositions within/adjacent to SIMS analysis craters to combine the U isotope ratio and major element composition and determine if agglomerates incorporated distinctly different amounts of radionuclides from the vapor phase in the fireball than hosts. These data sets were analyzed separately using two multivariate techniques: principal components analysis (PCA) and multidimensional scaling (MDS). PCA was used to determine the greatest sources of variance in the dataset and correlations amongst different major elements and the 235U/238U ratio. MDS was used to determine the compositional similarity between agglomerates and their host objects. Finally, quantified EDS maps and NanoSIMS rasters over compositional interfaces were used to determine the major element and U isotopic behavior across the different types of compositional interfaces observed between agglomerates and host objects in glassy fallout from this test.
The size and frequency of agglomerates observed in samples occupies between 0 and 20% of a sample's exposed cross-section, with samples exhibiting as few as 0 and as many as 57 agglomerates, ranging in diameter from 5 to 855 μm. However, the majority of samples have fewer than 20 agglomerates that occupy <10% of the exposed sample's cross-section. Major element compositions of agglomerates are observed to be a subset of the range of major element compositions measured in the host objects, suggesting they were more throughly melted and are well-mixed, in contrast to host objects which often contain partially-melted/unmelted mineral compositions. Both agglomerates and hosts are bounded by compositions measured in unmelted soil, suggesting both the major element composition of both are controlled by the soil incorporated into the fireball. Furthermore, major element homogeneity is strongly correlated with U isotopic homogeneity, indicating bulk mixing is responsible for distributing anthropogenic U throughout the volume of melts and suggesting that agglomerates, which are more compositionally homogeneous than hosts, are also more uniform with respect to their U isotope composition than host objects. Major element compositions from EDS analyses show that agglomerates tend to have similar compositions to hosts, particularly to the hosts they are attached to. Specifically, 81% of agglomerates were measured to have more similar major element compositions to their hosts than 50% of all other characterized agglomerates and hosts. Combined U isotopic and major element compositions show that agglomerates tend to be approximately as similar to their hosts when including U isotopic compositions. Specifically, 83% of agglomerates fall within the 50th percentile in terms of similarity to their hosts when compared to all other characterized agglomerates and hosts. Finally, two types of compositional interfaces are observed: an interface that tends to be enriched in CaO, MgO, FeO, and 235U ("CaMgFe interfaces''), which has been studied previously in the literature, and a compositional interface that tends to only be enriched in SiO2 ("Si interfaces''), which has not been noted in previous studies of fallout.
The variation of the major element and U isotopic composition between samples suggest a highly heterogeneous fireball environment and starting melts. However, the compositional similarity between agglomerates and their hosts suggest individual glassy fallout samples forming and agglomeration occurring in localized regions within the fireball. Because agglomerates are more likely to have similar compositions to their hosts, they most likely formed from similar parcels of melt, which either homogenized locally within the fireball or did not travel far during the fallout formation process, starting once they were swept into the fireball, then incorporated radionuclides from the fireball's vapor term, experienced agglomeration, and then exiting the fireball and quenching. Furthermore, given that agglomerates and host objects incorporated similar amounts of species anthropogenic U from the vapor phase, they likely experienced similar fireball environments (i.e., were swept into the fireball at a similar time and temperature). While some samples are observed to be formed from a relatively large number of agglomerates (occupying at most ≈20% of the exposed cross-section of the sample) the trend towards compositional similarity between agglomerates and their hosts suggests the observed agglomerates do not appreciably alter the overall major element and U isotopic composition of samples. However, there are outliers. Several ∼20 μm scale agglomerates in one sample are observed to be enriched in SiO2 and highly enriched in 235U relative to their host object and the statistical medians of the host population and combined agglomerate and host population. These agglomerates are possible candidates for being formed primarily from the vapor phase. However, their size and frequency are too small to appreciably alter the composition of the sample.