A hydrogen explosion occurred in a university biochemistry laboratory. Four persons were taken to the hospital for injuries. Three of these were treated and released shortly thereafter; the fourth was kept overnight and released the following evening. All of the exterior windows in the laboratory were blown out and there was significant damage within the laboratory. One sprinkler was activated that controlled a fire associated with a compressed hydrogen gas cylinder.

First responders from the local community and the university campus were quickly on the scene. Once the injured were attended to and the site secured, response efforts focused first on assessing potential hazards (electrical, fire, hazardous materials, etc). Campus personnel worked into the night to board up windows, isolate utility services, clean up debris, and otherwise secure the affected laboratories. The building reopened to occupants the following morning. There was minor water damage from sprinkler water in the laboratory below the explosion, but no equipment was damaged.

The laboratory works with soil bacteria that cannot survive in the presence of oxygen. As a result, research work is conducted inside a plastic chamber in which the chemical constituents in the air can be controlled. The explosion occurred during the set up of one of these chambers. The chamber is essentially a plastic bag with a volume of approximately two cubic meters. The setup procedure calls for using nitrogen to purge normal atmospheric air out of the bag three times, leaving a very small amount of residual oxygen present. The remaining small amount of oxygen is then removed by reaction with hydrogen in the presence of a palladium catalyst to form water.

Hydrogen was mistakenly introduced into the plastic bag as part of the first purge. As a result, the hydrogen concentration reached an explosive level inside the bag due to the relatively large presence of oxygen. The ignition source was most likely an electrical source inside the chamber or the palladium catalyst becoming too hot. The burn pattern observed after the explosion suggests that the fire ignited at the position of the catalyst, but that doesn't rule out the possibility that a spark was involved. The amount of hydrogen involved could not have exceeded one pound, which is the capacity of the compressed gas cylinder when full.

The investigation uncovered several procedural and design items that contributed to the explosion occurring:

One of the first steps of the procedure is to ensure that the lines leading from the nitrogen and hydrogen compressed gas cylinders are tight and do not contain any leaks. This was done by opening the tanks to pressurize the lines, but leaving the end valve in the "off" position. One then applies soapy water to each connection to see if there are any bubbles that would indicate a leak. When this check was completed, the research technicians forgot to turn the hydrogen tank back off prior to beginning the three purges with nitrogen.
Both compressed gas cylinders were connected to the chamber through a common tube through a "T" connector. A valve on the end had to be held open manually to introduce gas into the plastic bag. Normally, there would be a toggle switch inside the "T" connection that would be used to allow nitrogen or hydrogen, but not both, to be introduced into the chamber. At some point in the past, the right parts were unavailable, so that the "T" connection in use at the time of the explosion did not have a toggle switch. Had the toggle switch been present, the explosion would have been unlikely to have occurred.
One does not need to use pure hydrogen to scavenge for oxygen. One alternative to pure hydrogen would be to purchase a mixture of 95% nitrogen and 5% hydrogen. Since the lower explosive concentration of hydrogen in air is about 4%, the alternative mixture would be less likely to create an explosive atmosphere even if the gas were mistakenly introduced into the chamber prematurely.
The laboratory had previously tried to use hydrogen and oxygen sensors as a precaution to warn of an explosive atmosphere in the chamber. However, the atmosphere proved to be corrosive, which led to an inaccurate sensor readout that resulted in a dangerous gas mixture in the chamber. Had sensors been available while the chamber was brought online, again it is unlikely the explosion would have occurred.

There were a number of factors that mitigated the damage and allowed normal building activities to resume relatively quickly:

The laboratory group practiced good housekeeping, which minimized secondary impacts from the explosion.
Emergency information about hazards in the laboratory was posted outside the door and was helpful to emergency responders.
The Emergency Action Plan for the building had identified exit routes and a plan for evacuating the building in the event of an emergency. Building occupants followed the plan in a timely manner.
During the recent remodeling of the building, utilities were reconfigured for the laboratories such that each laboratory could be isolated. This reconfiguration of utilities allowed service to the affected laboratories to be cut off while service to the remaining laboratories in the building continued to function.
Cooperation between local first responders and campus personnel was excellent. This cooperative approach was built on a history of exercises and coordination meetings to strengthen working relationships.