Seven of the ten gases in a data centre are heavier than air. Most detectors are mounted as if none of them were
4 min read
Article
There is a version of gas detection design that gets signed off regularly and protects almost nothing. Detectors distributed evenly across a floor plan, mounted at a consistent height because consistency looks like rigour, specified from a schedule rather than from the plant. It satisfies a drawing review. It will not necessarily see a leak.
A data centre holds around ten distinct gas hazards, and they do not behave alike. Two of them rise. One sits roughly at breathing height and travels with the air. Seven are heavier than air, and will find the lowest available volume — a plenum, a cable pit, an underfloor void, a basement.
Toxicity tells you what a gas does to a person. Density tells you where to put the detector. Both matter, and the second is where designs most often go wrong.
Three planes, not one
Ceiling zone. Hydrogen is the fastest-moving hazard in the building. It rises quickly, pools at high level, is invisible and odourless, leaks through the smallest gap, and becomes explosive at just 4% by volume. Ammonia and methane behave similarly. A hydrogen detector at working height in a battery room is measuring the wrong part of the room.
Breathing zone. Carbon monoxide sits at roughly the height of a person’s face and travels wherever the air goes, which is what makes generator exhaust re-entry so dangerous — it is odourless and fatal at low concentration. Oxygen monitoring belongs at working height too, since the question being asked is whether a person can safely be in the space.
Floor zone. Refrigerants sink into voids, pits and plenums. CO₂ from suppression systems and cylinder stores sinks. SF₆ is extremely dense and settles into basements and cable trenches. These are the gases most likely to be missed entirely by a high-level detector, and refrigerant represents the largest gas charge on most sites.
Six rooms, six different right answers
Detection designed for a whole building protects none of it properly. Each space has its own gas, its own accumulation behaviour and its own mounting height.
UPS and battery rooms. Primary hazards are hydrogen, lithium-ion off-gas, carbon monoxide and hydrogen fluoride. Detectors belong at or near ceiling level, above the banks, away from supply air so incoming airflow does not dilute the sample before it reaches the sensor.
BESS and energy storage. The same hazard set and the same placement logic, with the added consideration that lithium-ion cells vent flammable and toxic gas within minutes of being driven into thermal runaway – while the surrounding room is full of switchgear, busbars and DC connections. The question is never whether an ignition source is present.
Chiller and mechanical plant. Refrigerants, possibly ammonia, and oxygen depletion. This is where the largest refrigerant charge on site sits, alongside the greatest concentration of joints, valves and compressors. Detectors go low, close to compressors and likely leak points – and the readout goes outside the room, so the atmosphere can be checked before the door is opened.
Generator halls and fuel storage. Carbon monoxide, nitrogen dioxide and fuel vapour. Detection at breathing height near the units, along exhaust routes and at air intakes, with overlapping coverage across bays. Nitrogen dioxide causes respiratory injury at low ppm – the UK eight-hour workplace limit is just 0.5 ppm.
Data halls and underfloor voids. Refrigerants and coolant vapour, now present in the same volume as the compute. Low-level detection, positioned by containment and airflow rather than by a grid on a drawing.
Switch rooms and suppression zones. SF₆, CO₂ and oxygen depletion. Floor level in pits and basements for the dense gases; oxygen at working height in gas stores, where the practical question is whether the room is safe to enter.
Matching the technology, not just the location
Getting the height right is necessary but not sufficient. The sensing technology has to suit the gas and the failure mode you are trying to catch.
For lithium-ion, the earliest available warning is not hydrogen or heat – it is the electrolyte carbonates vented by a stressed cell before there is any fire. A sensor that responds to those vapours buys time that a thermal or smoke-based approach does not.
Infrared suits refrigerants and CO₂. Electrochemical suits CO and NO₂. Flame detection covers generator yards and fuel systems, with hydrogen-specific flame detection where a hydrogen flame would be effectively invisible. And each gas has its own detector performance standard, which is easy to specify incorrectly.
Five decisions that shape the design
Inventory what can actually be released. Battery chemistry and charging regime, refrigerant type and charge per circuit, coolant volume, generator fuel, suppression agent, switchgear medium, water treatment chemicals. Revisit it whenever any of those change.
Place detection by physics. Height follows density. Position follows leak sources and airflow. Readouts go outside the room.
Match the technology to the gas, including the right performance standard for each.
Interlock detection to something that acts. Ventilation, isolation, alarm to a manned position, reporting into BMS and DCIM – then functionally test the whole chain at commissioning and after any change.
Choose for year eight, not day one. A detector needing frequent calibration, in a permit-controlled room, in a facility that resists downtime, will drift. Long-life and low-maintenance sensing is not a convenience feature in a building nobody enters.
Ten gases, one building, and a different right answer in every room.
If you would like a conversation on detector selection and placement across your facility, contact a Crowcon specialist.
INDUSTRY INSIGHTS
Subscribe to get the
latest insights
in your inbox
Read about Crowcon‘s Privacy and Cookie policy here. If you change your mind, you can unsubscribe at any time