KISMET Tungsten Dispersal Experiment
Copyright 1997, 1998 UC

Kenneth Wohletz, Thomas Kunkle, and Ward Hawkins
(LA-13227 excerpt)


Results of the KISMET tungsten dispersal experiment indicate a relatively small degree of wall-rock contamination caused by this underground explosive experiment. Designed as an add-on to the KISMET test, which was performed in the U-1a.02 drift of the LYNER facility at Nevada Test Site on 1 March 1995, this experiment involved recovery and analysis of wall-rock samples affected by the high-explosive test. The chemical, high-explosive blast drove tungsten powder, placed around the test package as a plutonium analog, into the surrounding wall-rock alluvium. Sample analyses by an analytical digital electron microscope (ADEM) show tungsten dispersed in the rock as tiny (<10 m) particles, agglomerates, and coatings on alluvial clasts. Tungsten concentrations, measured by energy dispersive spectral analysis on the ADEM, indicate penetration depths less than 0.1 m and maximum concentrations of 1.5 wt % in the alluvium.


Underground explosive testing requires some understanding of the dispersion of test materials into the host rock in order to evaluate the potential contaminant migration from the test area. In general, this information can be obtained by drill-back operations that recover samples of the rock adjacent to the test. But for mine-back reentry into the test area, knowledge of the potential range of hazardous material penetration around the test chamber (room) is important for human safety. This potential range is also useful for calculating posttest contaminant migration.

Small-scale underground explosive testing can involve test packages containing materials of potential concern for human safety. The KISMET experiment of 1995 (Kunkle, 1994) involved use of depleted uranium. In order to understand how plutonium might behave in a similar test, we used tungsten as a plutonium analog. With the objective of measuring how far tungsten would be embedded into the alluvium wall rock in the LYNER facility, we placed one kg of tungsten powder into three trays positioned on top and on the left- and right-rib sides of the explosive package.

The following report describes the method for sampling and analysis of the dispersed tungsten in the KISMET experiment (Kunkle, 1994) and results of tungsten concentration measurements in the samples of the LYNER facility alluvium. A first of its kind, this tungsten experiment is limited by the lack of knowledge about the physical behavior of the tungsten powder during the test, how it might interact with and penetrate the wall rock, and the amounts required to create measurable concentration profiles in the alluvium.


Tungsten occurs as very heterogeneously dispersed agglomerated masses, coatings, and small spherical particles. Its most common form is particles mixed into microvesicular agglomerates of quenched iron. In Figure 4, an SEM microphotograph taken from backscattered electrons, bright areas correspond to high Z-number particles composed of iron and tungsten mixtures (the tungsten areas are brightest). Most of these mixed particles have vesicles, which likely formed from gases trapped in the rapidly congealed iron, melted by the blast. Tungsten can also occur as individual spherical particles, but as shown in Figure 5 these are only easily viewed where they have agglomerated into masses generally <10 m in diameter. Some of these tungsten particles are agglomerated into larger masses (Figure 6). Most distinctive are coatings of tungsten on small particles, illustrated in Figure 7. Generally these coatings are less than a few micrometers thick, but because of their brightness in backscattered images, they are readily visible in SEM views.

Figure 4. Example 0.5 x 0.5 mm (0.25 mm2) area analyzed for tungsten
in sample LR-4 (analysis 5). Note the 100-m scale bar and the total width
of the analyzed area (497 m). Tungsten occurs as mixed patches (bright areas
with arrows) in larger vesicular iron globules in this sample. The analysis
of this area showed 2.78 wt % tungsten.

Figure 7. Two SEM microphotographs show tungsten coatings (bright areas) on large particles. Scans for tungsten
L-a x-rays show peaks corresponding to where the scan crosses the tungsten encrustations.

In many samples tungsten particles were difficult to recognize, and many individual analyses found none. For each sample, the entire thin-section area was scanned to find areas likely to show tungsten, and these areas were preferentially analyzed such that the results obtained should represent maximum tungsten concentrations. For samples that were recovered intact and oriented, simple line scans for tungsten were expected to show gradients that decreased from the test room surface inward. Figure 8 illustrates a typical result for such a line scan of sample B-2, an intact sample of the alluvium extending from the surface of the back inward ~5 mm. Smooth concentration gradients were not found, owing to the particulate nature of the embedded tungsten. A general decrease with depth into the sample was only crudely demonstrated.

Figure 8. SEM photomicrograph of sample B-2, an oriented sample
with the left edge at the surface of the KISMET test room alluvium
extending inward to a depth of about 5 mm. The horizontal line shows
the scan trace for tungsten L-a (W LA, plotted below the line) and
L-b (W LB, plotted above the line) concentrations. Both concentration
curves abruptly rise where the scan line crosses over the left edge of
the sample and show only a crude decreasing trend in abundance with
depth into the sample.

In order to test the diffusive character of our data for tungsten concentration, we plot concentration vs depth in Figure 12. The plot shows the fit of diffusive curves to the data using Equation (4). The best fit curve shows a surface concentration of 0.72 and diffusive coefficients of 3.19 x 10-4 m2/s and 1.28 x 10-5 m2/s for short time (t = 0.2 s) and long time (t = 5.0 s) respectively. Although the best fit, this curve does not seem to really predict finite tungsten concentrations deeper than about 0.03 m. To reflect the concentrations observed at greater depths, diffusive coefficients need to be increased to 2.00 x 10-3 m2/s and 8.00 x 10-5 m2/s respectively, which give perhaps more "conservative" tungsten penetration predictions with respect to environmental concerns. However, to really make a conservative prediction that provides an envelope for all data, we show a maximum diffusive curve in Figure 12, which requires diffusive coefficients of 2.50 x 10-2 m2/s and 1.00 x 10-4 m2/s for the short- and long-time diffusive conditions respectively.

Figure 12. Tungsten concentration as a function of sample depth. Error bars are 10% for sample depth and 0.1 wt %
for tungsten concentration. Three exponential curves are fit to the data using the diffusion model expressed in Equation
(1) shown in the figure. The best fit curve shows that concentration averages ~0.72 at the surface and decays with depth
with diffusion coefficients (m
2/s) for short-time (0.2 s) and long-time (5.0 s) models and a correlation coefficient of 0.68.
The eyeball fit curve better approximates deeper concentrations but gives an overall poorer correlation coefficient.
Lastly, a maximum fit curve brackets all concentration measurements and is most conservative for predicting the greatest
penetration of tungsten into the LYNER facility alluvium.


Tungsten used as an analog for plutonium in the KISMET experiment was dispersed by the blast and penetrated the alluvium rock forming the ribs, face, and back of the test room. Analyses of the concentrations of tungsten in the alluvium show its maximum penetration to a depth of ~0.08 m. The surface concentration of tungsten measured in samples representing the outer 1 mm of the alluvium shows a high value of ~1.5 wt %, but averaging ~0.7 wt % for all surface samples. This result supports the hypothesis that initially all of tungsten was uniformly distributed in the outer 0.5 mm of alluvium and later filtered to a maximum depth of ~80 mm. The distribution of measured tungsten concentrations in the alluvium supports a model that the emplacement mechanism (whether it is a filtration process or not) is mathematically diffusive. Maximum diffusive coefficients are 2.5 x 10-2 m2/s and 1.0 x 104 m2/s for the high-pressure phase (t = 0.2 s) and low-pressure phase of the experiment respectively. From this diffusive model, prediction of contaminant penetration for larger explosive experiments (higher P and T) can be achieved by applying this diffusive model and scaling the diffusive coefficients by P and T through a simple Arrhenius relationship.