Outdoor air runs a little over 400 parts per million of CO₂. The OSHA limit for an eight hour shift is 5,000. The concentration NIOSH calls immediately dangerous to life and health is 40,000, or 4 percent. At that 4 percent, the oxygen in the room has fallen from 20.9 percent to about 20.1, which is a reading no oxygen deficiency monitor will alarm on.
That gap is the part worth understanding. Carbon dioxide does not hurt people the way nitrogen or argon do, by crowding oxygen out of the way. It acts on the body directly, at concentrations where there is still plenty of oxygen left to breathe.
The Room Fills From the Floor
CO₂ has a molecular weight of 44 against air's 29. It is about one and a half times as dense, and it behaves accordingly. A leak into a below grade room, a walk in cooler, or a pit does not spread evenly through the space. It builds from the floor up. Someone standing in the doorway can read the room as normal while someone bent over a valve at floor level is in a different atmosphere entirely.
It is also colorless and odorless. At high concentration it gives a sharp acidic bite in the nose, from carbonic acid forming on wet tissue, and nobody should treat that as a warning. By the time it registers, the exposure limits are long past.
The white cloud at a liquid leak is not the CO₂. Liquid flashing to atmosphere drops toward 109 degrees below zero and freezes the water in the surrounding air, so what shows up is fog and dry ice snow. The gas doing the damage is invisible, and it is on its way to the lowest point in the room.
Half a Pound Reaches the Limit
The arithmetic is worth running once, because the quantities are smaller than people expect. A pound of liquid CO₂ makes roughly 8.7 cubic feet of gas at room temperature and atmospheric pressure.
Take a storage room 10 feet by 12 feet with an 8 foot ceiling. That is 960 cubic feet. Bringing that room to the OSHA eight hour limit of 5,000 ppm takes about half a pound of CO₂. Bringing it to 4 percent, the immediately dangerous concentration, takes about four and a half pounds.
A restaurant or small plant vessel holds several hundred pounds. A relief valve lifting on a warm afternoon, a fitting that lets go, a transfer hose left connected, any of those puts more than four and a half pounds into a closed room without anyone doing anything unusual. That margin is measured in a few pounds.
What the Code Asks For
The International Fire Code covers this in Section 5307. Where a system holds more than 100 pounds of CO₂ in an enclosed area, the code calls for gas detection, with alarms tied to the 5,000 ppm exposure limit and again near 30,000 ppm. NFPA 55 carries parallel language. Adoption happens at the state and local level, so which edition applies to a given site depends on what that jurisdiction has put in force and when.
That matters because an oxygen deficiency monitor does not answer the requirement and does not protect anyone. Oxygen alarms sit at 19.5 percent. Pushing a room below 19.5 percent oxygen with CO₂ takes close to 7 percent CO₂, which is well past the point where a person stops functioning. The instrument has to measure carbon dioxide.
Placement follows the density. A CO₂ head mounted near the ceiling above a vessel sitting on the floor will pass an inspection and miss the leak.
Purity Has Nothing To Do With It
This is where the grade conversation gets misread. Beverage grade CO₂ runs to 99.9 percent minimum purity, with benzene held to 20 parts per billion and total sulfur under 0.1 parts per million. Those numbers exist because the CO₂ ends up in something a person drinks.
None of them change what the gas does in a closed room. Beverage grade at 4 percent in a walk in is the same hazard as industrial grade at 4 percent. Purity describes what else is riding along in the stream. It says nothing about the behavior of the CO₂ itself.
Our plant in Lewiston, North Carolina, on the Perdue Farms campus, entered continuous commercial beverage grade production in December 2024, using patent-pending cryogenic purification, and the product was certified to that spec. What happens on the receiving end is a separate discipline with its own codes and its own instruments, and a cleaner product does not change any of it.
Where It Actually Goes Wrong
The failures follow a pattern. Below grade rooms with no ventilation and no detection. Fermentation cellars, where the process throws its own CO₂ on top of whatever is stored. Confined space entry into a vessel that held CO₂ and was assumed to be empty. And the ordinary one, a slow leak in a small back room over a long weekend, found by whoever opens the door on Monday.
What those share is a room nobody classified as hazardous, holding a gas nobody classified as hazardous, because it is the same gas that goes in the drink.
Two Questions
Most of this resolves with two questions. Where in the plant is CO₂ stored or piped through a space that is enclosed, below grade, or poorly ventilated. And in each of those spaces, is there a monitor that reads carbon dioxide, mounted low, with a calibration date somebody can produce.
The first list usually exists already, informally, in the heads of the people who work the floor. The second question is the one that goes unasked.
