HEPA grades explained: what H13, H14 and 0.3 microns really mean

HEPA stands for high-efficiency particulate air. It is not a fabric, not a brand and not a technology. The thing inside the machine is a mat of randomly laid fine fibres — usually borosilicate glass, sometimes spun polypropylene — packed a few millimetres thick and folded into deep pleats so that a large area of it fits into a small box. The word HEPA describes how that mat performed on a test bench, nothing more.

That distinction does real work. Two elements cut from the same roll of media can earn different classes depending on how densely they are packed and how carefully the pleat pack is potted into its frame. And a machine can carry the word HEPA while saying nothing about which class, if any, its element was tested to. Four of the twelve air purifiers here do exactly that.

It is also worth saying up front what the mat cannot do. A fibre filter captures particles. Gases pass through it the way air does, because a gas molecule is a few tenths of a nanometre across — roughly a thousand times smaller than the smallest particle the filter is tested on. Formaldehyde, solvent vapours and cooking smells are gases, and removing them is a different job done by a different material. That is its own guide .

Here is the claim you will read on a hundred forums and in a fair number of product listings: “HEPA cannot catch anything smaller than 0.3 microns.” It is the most common thing said about air purifiers and it is flatly wrong. It is wrong in an interesting way, too, because it mistakes the hardest case for the limit.

Plot a filter’s efficiency against particle size and you do not get a staircase, with everything above some threshold caught and everything below it passing. You get a U-shaped curve — or rather, an upside-down hump in the penetration. Efficiency is very high for large particles, falls to a minimum somewhere in the region of 0.1 to 0.3 microns, and then climbs back to very high again for the smallest particles of all. The bottom of that curve is called the most-penetrating particle size, or MPPS, and for typical HEPA media at typical face velocities it sits at about 0.3 µm.

The convention of quoting efficiency at 0.3 µm exists precisely because that is the worst case. A figure taken at 5 µm would be flattering and useless; a figure taken at the MPPS tells you everything else is better. When an element is rated 99.95 per cent, that is the floor of its performance, not the ceiling.

The practical consequence is large. An element rated 99.95 per cent at the MPPS is typically better than 99.99 per cent against particles of 0.01 to 0.05 µm — the size range that contains fresh combustion soot, cooking nanoparticles and traffic ultrafines. Those are the particles people worry about most, and they are the ones the filter handles most easily. The reason is not luck. It is a different physical mechanism taking over.

The key to the whole curve is that a HEPA mat is not a sieve. The gaps between its fibres are typically tens of microns across — enormous compared with the particles passing through. If filtration worked by straining, a HEPA filter would stop almost nothing under 10 µm and the category would not exist. It works instead by making particles touch a fibre, at which point they stick and stay stuck. There are three ways that happens, and they favour different sizes.

Diffusion is the one people find surprising, so it is worth a sentence on its own. A particle of 0.02 µm has so little mass that the individual air molecules hitting it deliver noticeable kicks. Instead of gliding along a smooth streamline it jitters — Brownian motion — and the smaller it is, the more violently it jitters. A jittering particle sweeps out far more space than its own diameter as it crosses the filter, so its chance of touching a fibre goes up sharply as it gets smaller. Below roughly 0.1 µm, smaller means easier to catch.

Now the U-curve explains itself. Above 1 µm, impaction is doing the work and almost nothing gets through. Below 0.1 µm, diffusion is doing the work and almost nothing gets through. In between there is a band where the particle is too light to leave its streamline by inertia and too heavy to wander off it by diffusion, so it sails along the airflow and its only chance of capture is interception. That band is the MPPS, and that is where the rating is taken.

A fourth mechanism sits alongside these in many modern media: an electrostatic charge baked into the fibres, attracting particles that would otherwise pass. It raises efficiency at a given airflow resistance, which is how charged media manage to be efficient and quiet at once. It also fades — with dust, with oily cooking aerosols and with water. That last is why you must never rinse a HEPA element unless its maker says it is washable, a point that returns in running costs .

If you want a grade that has a definition behind it rather than a marketing department, the one to look for is an EN 1822 class. EN 1822 is the European standard for high-efficiency filter elements, and its defining virtue is exactly the thing this guide has been building to: it specifies that efficiency is measured at the MPPS of the media in question, determined for that filter rather than assumed.

The two figures in each HEPA row are the integral value — the whole element averaged across its face — and the local value, the worst result found when a probe is scanned over that face hunting for a pinhole, a bad seam or a crushed fold. An element earns H13 only by passing both, which is why the class describes one manufactured element and not a type of paper.

North American usage runs on a different convention — “HEPA” there generally means 99.97 per cent at 0.3 µm, which sits between H13 and H14 and, since 0.3 µm is close to the MPPS of most such media, lands in much the same place. Neither convention is a claim about the machine as a whole. Both rate the element.

There is no standard called True HEPA. The phrase was coined to distinguish real filters from the “HEPA-type” and “99 per cent HEPA” media that appeared once the word started selling machines, and it is now printed on boxes as though it were itself a grade. It is not. “HEPA-type” is worse — a phrase built to borrow the word without the test, and it can sit on an E11 element or lower.

You do not have to take that on trust, because this catalogue is the evidence. Twelve air purifiers, and between them eight different ways of writing the same idea.

Count them and the picture is clear. Seven of the twelve machines name H13, in four different spellings. One names H14. The remaining four name no class whatsoever — they say HEPA, or True HEPA, or a trademark. Three separate manufacturers have written what is almost certainly the identical class three different ways, and nobody in the chain thought that odd, because the string is copy rather than data.

A HEPA element resists airflow. That is inherent: the same dense fibre packing that makes particles hit fibres also makes air work to get through. Air, like water and electricity, takes the path of least resistance, and the element is deliberately the most resistant thing in the box. Any gap — a poorly seated gasket, a warped plastic door, a cable grommet, a seam that was never sealed — is a path of far lower resistance, and a disproportionate share of the airflow will find it. A sealed filter path is the specification that says there is no such gap: the element is clamped against a continuous gasket, so the only route from the dirty side to the clean side goes through the media.

The arithmetic is brutal, and it is worth doing once. Suppose a machine carries a genuine H14 element, 99.995 per cent efficient, but one per cent of the airflow slips around it. That one per cent arrives on the clean side untouched. The remaining 99 per cent arrives having lost 99.995 per cent of its particles. Total penetration is 0.01 plus a negligible remainder, so the machine as a whole is about 99 per cent efficient — letting through roughly two hundred times what its element would, and worse than a plain H13 in a housing that does not leak at all.

You cannot measure this at home, but you can look for its signs. Does the filter door close with a definite compression, or merely clip shut? Does the old element come out with a clean rim of dust marking where it sealed? Does dust settle on the clean-air grille after a month of running? Only one of the twelve rows here names sealing at all, in the string “HEPA H13 (sealed)”. The silence of the other eleven is not evidence of leakage — it is evidence that nobody is being asked the question.

Having spent two thousand words on grades, here is the honest conclusion: at home, the grade is the less important number. Single-pass efficiency decides what fraction of the entering particles come out the other side. Room concentration is decided by how much of the room’s air goes through the machine each hour — airflow, not efficiency.

Compare the two levers at the sizes this catalogue actually spans. Moving from H13 to H14 improves a single pass from 99.95 to 99.995 per cent. Moving from the slowest machine here to the fastest goes from 240 m³/h to 600 m³/h, two and a half times the air through the filter every hour. At 99.95 per cent, air leaving the outlet is already effectively clean; the constraint is entirely how quickly the room’s air gets there. That is what CADR and room size is about, and it is the number to fix first.

It also explains what the finder does with the grade. On the allergy path, the air finder credits any machine whose grade string mentions HEPA, H13 or H14 — which, as the table above shows, is all twelve. The grade string separates nothing here, and is not pretending to. The ranking is done on coverage against your room, CADR, sleep-mode noise between 22 dB and 35 dB, whether you want an app, your budget, and the filter cost band.

If you are at the start rather than the end of this decision, how to choose an air purifier puts the grade back in its place among the four or five things that actually decide the purchase — and running costs covers what keeping that element fed will cost you for the next three years.