A shut-down plant is not automatically safe: the Catalyst Refiners H2S fatalities

Production had stopped; the hazards had not

On 22 April 2026, a reaction during chemical disposal released hydrogen sulfide at Catalyst Refiners in Institute, West Virginia. Two employees died, four were seriously injured and 18 people were transported for medical evaluation. The CSB investigation remains open, so no final root cause is claimed here.

Silver reclamation had ceased and most process equipment had been removed, but wastewater pretreatment remained operational. Tanks, a pump, chemicals, residual water, energy and decisions were still present. A shut-down plant is not automatically safe.

What happened

Workers transferred roughly 80 gallons of A-50 calcium-chloride coagulant and almost one 275-gallon tote of M-2000A sodium-trithiocarbonate metal precipitant to the receiving tank. They then began adding nitric acid diluted about a week earlier.

A reaction and fog appeared. The operator stopped the flow and later lost consciousness while seeking fresh air, then recovered. A supervisor collapsed. Two employees not involved in the transfer entered to help, were overcome by vapours and later died.

A process island remained alive

The open-top tank was about six feet high, nine feet in diameter and held about 2,227 gallons. Water moved from it to neutralisation, where sodium hydroxide controlled pH, and then to a clarifier separating silver-containing precipitate.

This was a live island inside a facility being dismantled. Every remaining subsystem needs newly defined boundaries, inventory, safeguards and authority.

Chemical incompatibility had to become a decision

The M-2000A SDS described it as incompatible with strong acids; nitric acid is a strong acid. The CSB reports that adding acid generated H2S. The SDS did not predict the exact accident, but its information should have informed disposal and hazard analysis.

Acidification can make sulfur-bearing compounds release toxic gas. The record does not support calculating amount, kinetics or concentration at the victims' locations.

No written procedure for this route

The CSB says there was no written procedure for disposing of A-50, M-2000A and nitric acid through this wastewater system. That specific finding does not prove the company had no procedures at all.

A one-off task needs more deliberate analysis, not less. Safety information controls nothing until it changes the field method.

Why H2S is so dangerous

Hydrogen sulfide is colourless, highly toxic, flammable and heavier than air. Severe exposure may rapidly affect the respiratory and central nervous systems and cause unconsciousness. Dispersion depends on source, ventilation and geometry.

NIOSH lists 100 ppm as IDLH and a 10 ppm ten-minute ceiling recommendation. OSHA lists a 20 ppm ceiling with a specific 50 ppm peak provision. These are US values, not Brazilian legal limits.

If the smell disappears, is the gas gone?

No. Smell rapidly fatigues and cannot reliably warn of continuing H2S. Loss of odour may be failure of human sensing, not removal of the contaminant.

An invisible atmosphere calls for instruments, alarms and predetermined actions for stopping work, evacuation and re-entry.

Personal monitors are one layer

The CSB says Catalyst Refiners did not provide or require personal gas monitors for employees or contractors during decommissioning. A suitable monitor can detect gas, alarm and support early evacuation.

It cannot prevent an incompatible reaction or validate a disposal route. The stronger measure is preventing the hazardous atmosphere from forming.

Does P100 protect against H2S?

No. At least three of four full-face respirators found by the CSB carried 3M 2091 P100 particulate filters. P100 does not protect against gaseous H2S.

Selection depends on contaminant, concentration, oxygen, duration and emergency conditions. Unknown or IDLH conditions may require positive-pressure SCBA or supplied air with auxiliary escape equipment. There is no universal H2S mask.

The second accident inside the first

The two employees who died had not performed the original transfer. They approached to help unconscious colleagues. Their impulse deserves respect; a toxic atmosphere can turn unprotected rescue into a chain of casualties.

Courage cannot replace monitoring, appropriate respiratory protection, communication and a trained rescue plan.

Was this a confined-space incident?

The public record does not support that classification. The release occurred near an open-top tank inside a building, with no reported tank entry. H2S can create a hazardous atmosphere outside classic confined spaces.

Brazilian NR-33 should only be invoked where the workplace actually meets its scope.

Decommissioning is not safe mode

Normal production uses familiar materials, sequences and safeguards. Decommissioning moves residual inventory, opens lines, changes containers, removes equipment and leaves partial systems live while unique tasks replace practised routines.

This does not prove decommissioning is always more dangerous. It means the risk model must change and safeguards must remain until their hazards are gone.

Residual inventory and compatibility

The final inventory should include products, intermediates, waste, cleaning agents, dead legs, partly filled tanks, drains, wastewater, cylinders and temporary containers. Each needs identity, location, condition and a validated removal route.

Teams must also ask what happens if materials meet: heat, gas, pressure, precipitation, corrosion or another dangerous product.

An SDS is process information

An SDS must inform procedures, training, task assessment, storage, transfer, disposal and emergency response. The barrier exists when incompatibility changes planning.

Brazilian NR-26 provides a useful comparison for hazard communication; it is not a legal finding about this US incident.

End of life is management of change

Stopping production changes people, knowledge, inventory, equipment, contractors and emergency arrangements. A sound process asks what leaves, what remains live, which hazards emerge and who holds technical authority.

The CSB is still analysing corporate governance and oversight. Approval of the disposal strategy and removal of safeguards remain open questions.

US duties and Brazilian comparison

OSHA and the UK HSE provide valuable material, but the record does not establish that OSHA PSM 1910.119 covered this facility. Foreign guidance is not Brazilian law.

For Brazil, NR-1 addresses risk management and emergencies, NR-9 exposures and NR-26 hazard communication. Fundacentro's respiratory protection programme guides selection, fit, training, maintenance and management.

Questions before decommissioning

This management checklist does not replace a process-specific study.

• Has every residual chemical been identified?
• Are dead legs, tanks and drains included?
• Are incompatibilities documented?
• Has each disposal route been technically validated?
• Is there a written procedure for non-routine work?
• Has the hazard analysis been revised?
• Are live-system boundaries clear?
• Are stop, evacuation and re-entry criteria defined?
• Are monitoring and personal detectors specified?
• Does respiratory protection match the scenario?
• Is toxic-atmosphere rescue planned and practised?
• Do contractors understand residual hazards?
• Are safeguards retained until the hazard is gone?
• Is closure authority explicit?

Known facts and open questions

A one-mile shelter-in-place order followed the release. On 29 April, WVDEP reported 12 air monitors on-site, at the perimeter and downwind, with no off-site H2S detections to that date. This does not prove community exposure; it shows process safety can extend beyond the fence.

The CSB continues to study chemicals, decommissioning analysis, procedures, PPE, monitoring and governance. A facility finishes operating only when its hazards are in a safe state. Until then, decommissioning is still process.