API PUBL 4456
Aerosol Formation, and Subsequent Transformation and Dispersion, During Accidental Releases of Chemicals
|Publication Date:||1 January 1987|
1. SUMMARY AND CONCLUSIONS
During an accidental chemical release, aerosols can form by the disintegration of a superheated or subcooled jet, particle entrainment from a boiling liquid pool, condensation of released vapors, and condensation of atmospheric water vapor chilled by a released cryogen. Once in the atmosphere, aerosols advected and dispersed downwind can deposit on the ground, evaporate, coagulate into larger particles, disintegrate into smaller particles, and grow as atmospheric vapors condense upon them.
Though many of the basic mechanisms are understood, the calculation of aerosol formation and transformation during accidental chemical releases still lacks a firm basis. During the more than 100 years that aerosols have been studied, researchers have concentrated not on accidental releases but on particle fundamentals and on smoke, processing fumes, automobile exhaust; and aerosols from nuclear processes and war operations. Nonetheless, calculation algorithms in some stage of development are available to deal with every mechanism relevant to an accidental release except entrainment from boiling liquid pools, condensation of atmospheric water vapor, and particle evaporation.
Of the few experimental data relevant to the formation and transformation of aerosols arising from accidental chemical releases, none are from experiments whose primary purpose was to generate such data, many are only qualitative, and some have not been thoroughly analyzed.
There are so few existing data that further analysis would have only limited value. It would probably be more fruitful to generate more field and/or laboratory data. Also, an atmospheric dispersion computer code, capable of handling all aspects of aerosols to the extent that they are currently known, could be developed. Such a code, in addition to being useful for immediate practical calculations, would allow sensitivity studies to determine which aspects of aerosols are most important for final results, and which, therefore, are most important for further experimental effort and theoretical development.
In addition to the general information given in the preceding paragraphs, the principal conclusions of this literature survey are:
• Superheated liquids are often accidentally emitted as flashing sprays, producing aerosol droplets on the order of 10μm (the generally recognized boundary between respirable and non-respirable particles), approximately log-normally distributed, and with drops often located outside the bounds of any visible carrier plume.
• The amount of liquid calculated to flash under adiabatic conditions is the minimum that vaporizes, because there is always some heat transfer from the environment. Flashing, neither completely isentropic nor isenthalpic, is usually assumed to be one or the other.
• TRAUMA is the one model for aerosol formation from flashing liquids sufficiently developed to be encoded in a computer program. A principal conclusion of the writer of TRAUMA, based on the primarily qualitative data that led to its development, is that an elevated jet of flashing liquefied ammonia will remain entirely airborne (both the vapor and liquid droplets) unless the released jet impinges directly on a surface. Calculated results obtained with TRAUMA indicate that flashing is more isenthalpic than isentropic, that flashing within a discharge pipe can change the discharge rate by a factor of 5 for choked flow and by a factor of 20 for unchoked flow, and that, in the absence of liquid pool formation, even diluted liquefied ammonia plumes are denser than air.
• Quantitative data on flashing liquids have been obtained from the Desert Tortoise series of spills of liquefied ammonia. Beta-ray absorption gauges provided a measure of cloud density and liquid aerosol content. The data, not yet analyzed, might provide information regarding aerosol transformation processes over the first 100m downwind, from which information regarding particle-size distribution might be inferred. Unfortunately, little or no information is available from the integrating nephelometers and particle-size analyzer also used during the spills.
• In addition to the aerosols generated during liquefied ammonia experiments, aerosols were generated during fifteen liquefied natural gas and two liquefied propane experiments. The purpose of these experiments was the development and validation of heavier-than-air vapor models, and no quantitative aerosol data were measured. The films and photographs of the dense fogs which occurred probably cannot provide any significant information for aerosol research.
• Aerosol particles resulting from vapor condensation usually have diameters 0.1 to 30μm, finer than particles resulting from liquid breakup. Calculations for such aerosols would be made using a diffusion-limited particle growth law and an initial particle radius of zero.
• Algorithms for particle coagulation, in order of increasing simplicity, assume continuous, discrete, or parametrized size distributions. The discrete distribution has been found generally adequate, and the parametrized distribution may be adequate for many purposes.
• The particle-deposition algorithm included in DuPont's Savannah River Plant model is the most realistic one currently available, treating deposition as a depletion from the bottom surface of the plume. Both wet and dry deposition and both puffs and plumes are treated. To make calculations, empirical deposition velocities, generally not equal to gravitational settling velocities, are needed. Such velocities are obtainable from existing compilations, prediction models, and measurements using non-depositing vapor tracers.