How vacuum insulated glazing is made
A pane of vacuum insulated glazing looks almost impossibly simple: two thin sheets of glass, a whisker of a gap, and nothing in between. That “nothing” is the clever part — and getting it into a sealed unit that survives decades of weather is one of the harder tricks in the glass industry. This is a look at how a vacuum glazing (VIG) unit is actually made, from raw float glass to the little capped port in the corner, written for the curious rather than the credentialed.
The whole point of the technology is to remove the air from between two panes. Heat crosses an ordinary sealed double-glazed cavity three ways — conduction and convection through the gas, and radiation between the glass surfaces — and the gas routes account for the majority of the loss. Pump the gas out and those routes largely vanish. What is left is radiation, which coatings deal with, and a handful of engineering problems that turn a neat idea into a genuinely difficult product. We will take them in the order a factory does.
First, why removing the air is so hard
Before the steps, it helps to grasp the single fact that shapes every decision that follows. When you evacuate the cavity to a hard vacuum — the working target is around 0.1 Pa, roughly a hundred-thousandth of normal air pressure — the atmosphere outside no longer has anything pushing back from inside. The result is a crushing, permanent load.
That load is about 100 kPa, which works out at roughly 10 tonnes pressing on every square metre of glass. Imagine two family cars parked on a pane the size of a door, trying to squeeze the two sheets together until they touch. If the panes touched, heat would pour straight across the contact and the vacuum would be pointless. So the first structural job is to hold the panes a fraction of a millimetre apart against ten tonnes per square metre, forever, without spoiling the view. That is what the pillar array does, and it is why VIG can never be just “two sheets and a vacuum.”
The gap itself is tiny — about 0.1 to 0.2 mm, a fifth of a millimetre at most. Everything below is engineered around keeping that sliver of vacuum stable and load-bearing.
Step 1: Preparing the two panes
Manufacture starts with two sheets of ordinary soda-lime float glass, the same clear glass used in most windows, usually 3 to 4 mm thick each. Where clarity matters — for a shopfront, or to keep a heritage window looking as colourless as the original single glazing — low-iron glass can be substituted, because standard float has a faint green cast when you look through the edge.
The low-E coating does the work the vacuum leaves behind
A vacuum stops conduction and convection but does nothing about radiation: infrared heat crosses empty space perfectly well. So one internal surface is given a low-emissivity (low-E) coating, a microscopically thin metal-and-oxide stack that reflects infrared back towards the room it came from. Without it, a vacuum unit would perform poorly; with it, radiation is suppressed and the unit finally earns its U-value.
The numbers show why the coating is not optional. Emissivity — the fraction of heat a surface radiates away — runs like this:
- Uncoated glass: about 0.84
- Pyrolytic “hard-coat” low-E: about 0.2
- Sputtered “soft-coat” low-E: about 0.03
Soft-coat reaches the lowest emissivity and so the best performance, but it is delicate — which is no problem here, because it will spend its life sealed inside a vacuum where nothing can touch or tarnish it. That protected cavity is one of the quiet advantages of the format. The coating typically sits on the second surface (the inner face of the outer pane), facing the gap. You can read more about the coatings and the rest of the materials in a VIG unit on the main guide.
Annealed or tempered — a choice that decides everything downstream
Here is the fork in the road that quietly defines the whole product. Glass can be left annealed (ordinary, as-cut) or tempered/toughened (heat-treated so it is far stronger and breaks safely into blunt granules). The catch is that tempering is a heat treatment, and if a later manufacturing step gets hot enough, it simply undoes it — the glass reverts towards annealed and loses its safety rating.
Whether the panes can stay tempered depends entirely on how the edge is sealed, which is Step 3. The original vacuum glazing had to use annealed glass because its seal was fired too hot. Newer low-temperature seals keep the glass tempered. Hold that thought; it is the single most important trade-off in the technology.
Step 2: Laying the support-pillar array
Onto one of the prepared panes goes the grid that will fight those ten tonnes: an array of support pillars (also called microspacers). These are minute discs of high-strength material — commonly stainless steel or a ceramic, and in at least one product a transparent compound — each roughly 0.3 to 0.5 mm across and 0.1 to 0.2 mm tall. They are placed by machine in a regular square grid, and their height sets the width of the vacuum gap.
On a classic annealed unit the grid spacing is about 20 mm, which means a one-square-metre pane carries around 2,600 pillars. That is a lot of tiny dots, but each is smaller than a full stop and, from normal viewing distance, the array reads as a faint, barely-there texture rather than a pattern you notice.
Every pillar is doing two jobs at once, and they pull against each other. It has to be strong enough not to be crushed by the atmospheric load, yet each pillar is also a solid bridge for heat to cross the vacuum. More pillars, or fatter ones, mean a stronger unit but more thermal bridging and a slightly worse U-value.
That tension explains why tempered VIG uses a wider, sparser array — because tempered glass is stronger, it can span a bigger gap between supports, so the grid can be stretched out and the number of pillars cut dramatically (to a few hundred per square metre rather than a few thousand). Fewer contact points means fewer thermal bridges and a lower U-value. The support-pillar and edge-seal choices are covered in their own guide, because between them they largely decide how well the finished unit performs.
Why flatness is not negotiable
The pillars only work if the two panes are genuinely, precisely flat. The gap they hold open is a fifth of a millimetre; a pane that is bowed or wavy by even a fraction of that will press unevenly on the array, overloading some pillars and letting others float free. Float glass is naturally very flat, but the tolerances here are tight enough that surface flatness and cleanliness are treated as make-or-break. A single speck of dust trapped under a pillar, or a slight warp, can seed a stress point that cracks the finished unit months later.
Step 3: Sealing the edge
The two panes now have to be joined around their entire perimeter with a hermetic seal — one that will hold a hard vacuum not for months but for decades. This is where the fork from Step 1 finally forces a decision, because there are two families of seal and they demand different temperatures.
The rigid glass-frit seal. The original approach, still used by the first commercial product (Pilkington Spacia), lays a bead of low-melting-point glass “frit” around the edge and fires the whole assembly in a furnace at roughly 350 to 450 °C to fuse it. The result is entirely inorganic, strong and long-lived — glass welded to glass, with no rubber or organic sealant to perish. But two things follow. First, that firing temperature is hot enough to de-temper the glass, which is why frit-sealed units historically had to use annealed panes. Second, a rigid seal is unforgiving: when the inside and outside of the window are at very different temperatures, the two panes want to expand by different amounts, and a stiff seal concentrates that stress rather than absorbing it, raising the risk of thermal-stress cracking.
The flexible low-temperature metal seal. The newer approach bonds the edge with a metal or metal-alloy seal at around 250 °C — cool enough that the glass stays fully tempered. Products such as LandVac and HaanGlas use proprietary low-temperature seals of this kind to deliver a toughened safety unit. A flexible metal seal also flexes with the differential expansion instead of fighting it, which reduces cracking risk. This single change — sealing cool enough to keep the glass tempered — is what unlocked modern tempered safety VIG and the lower U-values that go with a sparser pillar array.
| Rigid glass-frit seal | Flexible low-temperature seal | |
|---|---|---|
| Seal material | Fired glass frit | Metal / metal alloy (e.g. indium-based) |
| Sealing temperature | ~350–450 °C | ~250 °C |
| Effect on glass | De-tempers — panes end up annealed | Glass stays fully tempered (safety-rated) |
| Behaviour under temperature swings | Stiff; concentrates stress | Flexes; accommodates expansion |
| Example products | Pilkington Spacia | LandVac, HaanGlas |
A third, still-emerging route uses a welded metal seal (the pre-commercial US “V-Glass” being the clearest example), and some makers keep their exact seal chemistry proprietary — AGC’s Fineo, notably, discloses neither a solder-glass nor a visible port. The point for a general reader is that the seal is not a detail: it dictates whether the glass can be tempered, how the unit copes with British weather swings, and how long the vacuum is likely to last.
Step 4: Pulling the vacuum
With the edge sealed, the cavity still has ordinary air in it. That air is now removed down to the target of about 0.1 Pa — roughly a thousandth of a Torr, the point at which so few gas molecules remain that conduction and convection through the gap become negligible. There are two ways to do it.
The traditional method leaves a small pump-out port: a tiny tube or hole through one pane, connected to a vacuum pump that draws the air out. This is the origin of the little round feature you can spot in a corner of many vacuum units. The alternative is to assemble and seal the whole unit inside a vacuum chamber, so that when it is closed it is already evacuated and no port is needed — which is why some products (Fineo among them) have no visible port at all.
Getting to 0.1 Pa is not just a matter of running a pump for longer. Glass and seals slowly release trapped gas of their own (outgassing), so the cavity has to be pumped patiently, often with gentle heat to coax the gas out, until the pressure holds steady at the target. Rush it and the vacuum will quietly degrade from within once the unit is closed.
Step 5: The getter, the port and the cap
A vacuum, once made, does not stay perfect on its own. Over years, minute amounts of gas keep seeping from the glass and seal, or permeate very slowly from outside. Left alone, that trickle would gradually soften the vacuum and erode the insulation. The fix is a getter — a small piece of chemically reactive material sealed inside the cavity that “mops up” stray gas molecules by bonding with them, holding the pressure low for the long haul. Most VIG uses a non-evaporable getter, sometimes a visible pellet, sometimes built discreetly into the edge seal. It is the component most responsible for the long vacuum lifespans manufacturers advertise — though it is worth noting those 25-to-60-year figures are manufacturer claims; real-world longevity at scale is not yet independently proven.
Finally, on any unit made with a pump-out port, the tube is sealed off once evacuation is complete — melted or crimped shut — and a small protection cap is fitted over it to guard the delicate sealed tip. That cap is the little “button” you can find in a corner of, say, a Pilkington Spacia unit. On chamber-evacuated products there is nothing to cap, which is why they look seamless.
At this point the unit is, in principle, finished: two coated panes, a pillar array, a hermetic edge, a stable vacuum and a getter keeping it that way. Many units then go on to a second life — laminated to a further pane for safety and acoustics, or bonded into a conventional sealed unit to make a “hybrid” that pushes the U-value lower still.
The process at a glance
| Step | What happens | Why it matters |
|---|---|---|
| 1. Prepare the panes | Cut and clean two float (or low-iron) panes; apply a low-E coating to one inner face; choose annealed or tempered | The coating handles radiation; the temper choice is dictated by the seal in Step 3 |
| 2. Lay the pillar array | Place ~0.3–0.5 mm pillars, ~0.1–0.2 mm tall, on a ~20 mm grid (fewer on tempered units) | Holds the panes apart against ~10 tonnes/m²; flatness is critical |
| 3. Seal the edge | Fire a rigid glass frit (~350–450 °C) or bond a flexible metal seal (~250 °C) | Rigid de-tempers the glass; low-temperature keeps it tempered |
| 4. Evacuate | Pump the cavity to ~0.1 Pa via a port, or seal inside a vacuum chamber | Removes gas conduction and convection — the whole point |
| 5. Getter, port, cap | Activate the getter; seal the pump-out tube; fit the protection cap | Keeps the vacuum stable for decades and protects the sealed port |
Why this is genuinely hard to do at scale
Read down that table and every step is individually plausible. The difficulty is that they all have to succeed together, on every unit, at production speed, for a product expected to last as long as the building.
The pillar array must be placed to sub-millimetre accuracy across a whole pane, thousands of times per unit, with no misplaced or missing pillars. The edge seal has to be perfectly hermetic around the entire perimeter — a single microscopic leak anywhere and the vacuum is lost, slowly and invisibly, sometimes only becoming obvious after installation. The evacuation has to reach and hold a pressure far below anything a home ever experiences, while the glass itself is quietly trying to outgas. And the getter has to keep winning that battle for decades.
On top of all that sits the mechanical reality that the finished unit lives its whole life under ten tonnes per square metre of squeeze, flexing every time the sun comes out or a cold night falls. That is why the seal and pillar choices matter so much, why tempered low-temperature-sealed units have been such a step forward, and why vacuum glazing costs meaningfully more than ordinary double glazing — roughly two to four times as much, on indicative market figures. It is also why the field is still relatively small: the tolerances are punishing, and the equipment to hit them reliably is specialised. For how those manufacturing choices translate into real thermal numbers, see our explainer on vacuum glazing U-values, and the underlying research is collected on our science page.
It is worth keeping one accuracy point straight, because marketing muddies it. The classic frit-sealed unit lands at about 1.1 W/m²K at the centre of the glass — as good as a modern double-glazed unit, but in single-glazing thickness. The headline figures of 0.4 to 0.7 W/m²K belong to the tempered, low-temperature-sealed products and to hybrid units, not to every vacuum pane. The manufacturing route is what separates them.
The bottom line
Vacuum insulated glazing is made by coating and preparing two thin panes, scattering a precise array of microscopic pillars to hold them apart, sealing the edge either hot-and-rigid or cool-and-flexible, pumping the cavity down to a hard vacuum, and letting a getter guard that vacuum for the long term. None of the steps is exotic on its own. What makes VIG hard is doing all of them perfectly, together, on a component that then spends decades under ten tonnes of atmospheric load per square metre. Understand that, and you understand both why the glass can be barely thicker than a single pane — and why it is not cheap. The seal you choose and the pillars you lay are not manufacturing footnotes; they are the whole story of how the finished window performs.