Inadvertent Production And Explosion Of Liquid Chlorine Dioxide

Alfred P. Sattelberger, Jeanne M. Robinson, and Robert P. Currier
Los Alamos National Laboratory
Los Alamos, N.M.

Chemical & Engineering News (09 December 2002) Vol. 80, No. 49, pp. 4-5

Due in part to its use last year in decontaminating the Hart Senate Office Building, chlorine dioxide (ClO2) has received renewed interest. This inherently unstable compound has been used for decades as a biocide, as a bleaching agent, and in water treatment. Spontaneous gas-phase decomposition to chlorine and oxygen is known to occur at chlorine dioxide concentrations above 10% (liberating approximately 25 kcal per mole) while the condensed forms (liquid and solid) are prone to unpredictable and explosive decomposition (C&EN, Aug. 6, 1951, page 3196 and Oct. 22, 1951, page 4459; "Kirk-Othmer Encyclopedia of Chemical Technology"; "Ullmann's Encyclopedia of Industrial Chemistry"; "Bretherick's Handbook of Reactive Chemical Hazards"). Because of the chemical instability, chlorine dioxide is typically produced at the point of end use.

Researchers at Los Alamos National Laboratory were recently experimenting with ClO2 as a biocide under a wide range of relative humidity. In these experiments, the ClO2 was generated by passing a mixture of nitrogen and chlorine through a packed bed of sodium chlorite flakes (the reactor) at ambient temperature. The reaction between the chlorine and sodium chlorite produced ClO2 (in nitrogen) and solid sodium chloride, liberating approximately 50 kcal per mole ClO2.

The resulting gas mixture was then fed into the sample chamber. Both the reactor and sample chamber were housed within a standard chemical fume hood. Calculations on the decomposition of ClO2 in the vapor phase indicated that the system consisting of reactor, sample chamber, and connecting tubing could withstand the pressure increase, including the dynamic pressure associated with the decomposition "puff." As a matter of practice, the system was equipped with pressure relief valves and an automatic shutoff on the feed gas.

The concentration of ClO2 in the reactor effluent was limited by stoichiometry to twice the concentration in the inlet feed gas. An initial series of experiments ranging from 0.5 to 8% ClO2 indicated that higher concentrations would be needed to meet the experimental objectives, which included measuring biocidal activity at subambient temperatures in the sample chamber (temperatures could be as low as 273 K in the actively chilled chamber).

The literature suggested that higher concentrations could be handled safely, provided the temperatures were kept below 300 K and the risk of photolytic decomposition was minimized [Phys. Chem., 214, 533 (2000)]. The commonly accepted mechanism for vapor-phase decomposition of ClO2 involves a buildup of free-radical intermediates. Measured induction times for vapor-phase decomposition are reported to be relatively long under the experimental conditions of interest (J. Phys. Chem., 72,1849 (1968);J. Chem. Soc. Faraday Trans., 90, 3391 (1994)]. Nevertheless, the increase in stored chemical energy associated with increasing ClO2 concentrations prompted the researchers to scale down the reactor volume by more than a factor of 10 in addition to shielding the entire system from light.

There was also concern over heat generation in the reactor. Calculations using the experimental flow rates (a few hundred standard cubic centimeters per minute) indicated that even with a pure chlorine feed to the reactor, the heat generated in the reactor would be modest (35 W). If this heat were uniformly spread over the entire reactor volume, the temperature rise would be modest. However, if reaction were to occur in a narrow zone within the sodium chlorite bed, the local temperature could rise significantly, resulting in initiation of undesired vapor-phase decompositions.

The researchers thus fitted the scaled-down reactor with thermocouples that could be repositioned within the sodium chlorite bed and began screening experiments to assess the temperature rise and profile in the reactor. Experiments with increasing chlorine feed concentrations, up to and including pure chlorine, indicated that the heat rise was within acceptable limits to avoid initiating a vapor-phase decomposition. The reactor was also given a fresh charge of sodium chlorite flakes as a test of the worst-case scenario (a fresh charge has no NaCl product layer on the flakes to act as a diffusion barrier and moderate the reaction rate).

On the day prior to the accident, the new generator was connected to the sample chamber and the full system performance was verified. Again, modest temperature increases in the reactor were observed. The system was held static overnight, and experiments resumed the next day.

Upon continuing the experiments, a dramatic rise in the reactor temperature was observed. The researchers promptly left the room and activated the automatic shutoff system for the feed that had been engineered into the system. Shortly thereafter, an explosion destroyed the experiment and did considerable damage to half of the fume hood.

Postincident analysis showed that while focusing atten-tion on issues related to the reactor, the researchers failed to appreciate that conditions were created in the sample chamber that permitted ClO2 to condense as liquid. Overnight, the sample chamber had cooled to the point where the interior chamber walls were now at the chiller set point. This, together with the higher partial pressures of ClO2 in use, corresponded to a point right on the gas-liquid phase boundary.

When flow was reestablished, the newly produced ClO2 condensed in the sample chamber, which began to act as a cryo-pump, drawing more feed gas through the reactor to maintain the system pressure. Fortunately, this increase in flow rate through the reactor gave rise to the dramatic increase observed in temperature and provided the experimenters with a signal to evacuate.

Based on the damage assessment, milliliters of liquid ClO2 may have condensed on the walls of the sample chamber in a relatively short time. As an explosive, liquid chlorine dioxide has roughly one-third the power of TNT. Hence, considerable damage occurred and the incident was clearly a "near miss" to a personal injury.

While the experimenters were familiar with the properties of chlorine dioxide and the dangers of the condensed phases, focusing their attention on the dynamics in the reactor section of the system resulted in a failure to recognize the undesirable conditions evolving in the sample chamber. We urge other researchers to keep the precise (quantitative) location of phase boundaries in mind as experimental parameters change during their ClO2 experiments.

In addition to the sources cited above, key properties of chlorine dioxide (vapor pressure, heat capacity, Henry's law, spectra, and so on) can be found on the National Institute of Standards & Technology Chemistry WebBook ( A phase diagram for the ClO2-H2O system can be found in U.S. Patent No. 2,683,651 (1954).

Return to List of Safety Letters
Return to Chemical & Engineering News Home Page

This page last revised June 23, 2004
© Copyright 2004 by the American Chemical Society