Data center dust removal: what actually removes dust, and what only moves it

This is what we are best at, and it is worth being precise about. Most of what gets sold as cleaning in a data hall relocates dust rather than removing it. Our general scope of work states its purpose as reducing airborne and surface-level particulates, in that order. Every rule below exists because of the first half of that sentence.

Illustration

Removal versus redistribution

There are four ways to leave a surface looking clean, and only one of them is removal. You can capture the dust and take it out of the room. You can push it onto the next surface along. You can glue it down with residue. Or you can lift it into the air, where it settles over the next few hours onto everything you just finished.

That last one is what fails a hall. It is invisible at the moment it happens, and it undoes work that has already been signed off.

Four rules in our procedures come straight out of that, and you can hold a technician to every one:

• Floors are never swept. Our Floor Surface Cleaning SOP says it in terms: do not sweep the floor, sweeping will put dust into the air. Loose debris comes up with a HEPA vacuum instead. This is the most common failure when a general janitorial contract gets applied to a data hall.
• Cleaner goes on the cloth. It is never sprayed onto the surface and never sprayed above equipment. A cleaner in the air is airborne liquid carrying airborne particulate, released straight above the intakes it is supposed to protect.
• Work runs top-down. We start at the highest point we can safely reach and move down, so anything released above a cleaned surface is captured on the way through instead of landing on finished work.
• Cloths are flipped or replaced every three to six feet, or sooner if they load up. A saturated cloth stops capturing dust and starts distributing it, and it does that silently.

What the HEPA rating is actually specifying

Our bid specifies HEPA filtration rated 99.97% at 0.3 µm. That number describes the vacuum’s exhaust, not its suction. It means 99.97% of particles at that size stay in the filter instead of going back into the room.

A vacuum without it picks up the visible debris and vents the fine fraction straight back into the air at floor level. That fine fraction is the part that matters for airflow, for optics and for the particle count your hall gets measured on. So an unfiltered vacuum is worse than not vacuuming at all.

0.3 µm is a test convention, not a floor. It is the size the North American rating convention tests at, and it was picked because it sits near the hardest size for filter media to catch. Not because anything smaller gets through. Efficiency actually climbs on both sides of that size, so at 0.5, 1 and 5 µm (the sizes a Class 8 count is taken at) a HEPA filter performs better than its rated 99.97%, not equal to it. Anyone who tells you a HEPA filter cannot catch anything below 0.3 µm has the physics the wrong way round.

Filter media have a most penetrating particle size: the diameter where efficiency bottoms out. For HEPA media that usually falls between 0.1 and 0.25 µm. It is a property of the individual filter rather than a fixed number, which is why the European convention (EN 1822 / ISO 29463) rates each filter at its own measured minimum instead of at a nominated size.

Efficiency climbs on both sides of that minimum. Below it, diffusion captures the small particles, and the smaller they are the better it works. Above it, interception and impaction take over, and both improve with size. So a HEPA filter is not less effective below 0.3 µm. It is more effective. At 0.5, 1 and 5 µm (the sizes a Class 8 count is taken at) it performs better than the rated 99.97%, not equal to it. Anyone who tells you a HEPA filter cannot catch anything below 0.3 µm has the physics the wrong way round.

ISO 14644-1 sets Class 8 limits at 0.5, 1 and 5 µm. All three sit below the size at which dust becomes visible on a dark surface under hall lighting, which is why we check with a raking light rather than a glance. What the classification does and does not cover →

Why construction dust is a different problem from occupancy dust

We run two different products because these are two different problems. Which one you have decides which one you should buy.

What a concrete cut actually puts into the air

One published measurement is worth the space. A peer-reviewed study of concrete mixing, drilling and cutting (Azarmi, Kumar and Mulheron, Journal of Hazardous Materials, 2014) found that ultrafine particles below 0.1 µm made up more than 90% of the particles by number and less than 10% of the dust by mass. It measured concrete work specifically, not construction dust in general, and we know of no equivalent measurement taken in a data hall fit-out.

Two things follow. A builder’s clean is judged on mass and on line of sight; a data hall is judged on count. ISO 14644-1 counts particles, it does not weigh them. And the fraction that dominates the count is finer than the test even looks at: particles below 0.1 µm are not counted at all, do not settle on any timescale a construction program cares about, and are untouched by the methods that make a room look finished. They come out by filtration and air change, or they do not come out.

That is reasoning from the physics, not a measurement, and we would rather say so than dress it up. A hall left dirty through a build does not have a dust problem on the floor. It has a dust problem waiting for the first person to walk through it and the first hour the air handlers run.

So the four rules at the top of this section are not housekeeping preferences. Every one of them is a re-entrainment control: four different ways of not putting the reservoir back into the air. The fine fraction is the worst case in both directions: it sticks hard, and once airborne it stays airborne. Only two things work, and we do both: not raising it in the first place, and filtering what is already up there.

Why the sequence is not negotiable

Overhead first, then containment, then walls, then floors. That is the order our SOPs are written in and the order every crew is briefed in. The reason is arithmetic, not tradition. Clean a floor before the cable tray above it and you clean that floor twice, and the second time happens under pressure at the end of a shift. On a construction program, that ordering error is not an inconvenience. It is a day.