A question the internet answers badly

The Window That Never Flowed

You have heard it from a tour guide, a teacher, a documentary: glass is secretly a very slow liquid, and that is why the windows in old cathedrals are thicker at the bottom — centuries of glass quietly oozing downward. It is one of the most repeated facts that is simply not true. Glass is an amorphous solid, sitting hundreds of degrees below its glass transition at room temperature. Below you can calculate a hot pane's characteristic flow time and compare two cited room-temperature estimates: about one second at 1200 °C, billions to trillions of years with those cold inputs. Then we'll find where the old window's uneven thickness actually comes from. The hot-glass model runs live; cold estimates are separately sourced. 20/20 numerical checks green.

1 · So what is glass?

Glass is made by melting sand-and-soda to a runny liquid and then cooling it fast enough that it never gets the chance to crystallize. The atoms freeze in place still arranged in the jumble they had as a liquid — no neat repeating lattice like ice or quartz, just disorder, locked. That frozen-disorder state has a name: an amorphous solid. It is genuinely rigid. As the materials physicist Mark Ediger puts it, "if you grab it, it holds its shape."

The myth takes the one true seed — glass is born from a liquid and keeps a liquid's disordered structure — and grows a false tree from it: that it must therefore still be flowing. Calling glass a "supercooled liquid" is a big part of how the confusion spreads; many physicists now treat that phrase as a misnomer to avoid. The honest one-liner: glass is a solid with a liquid's memory, not a liquid with a solid's patience.

2 · When does glass actually flow?

Whether something "flows" is really a question about viscosity — its resistance to oozing — and viscosity depends ferociously on temperature. The curve below is the Vogel–Fulcher–Tammann predictive model for a soda-lime glass composition (the family used in windows and drinking glasses). Heat it and it pours; cool it and it stiffens by many orders of magnitude. Drag the furnace dial and watch how long a 3 mm pane would take to visibly sag under its own weight.

The point the dial makes by itself: glass really is a liquid you can pour — at 1200 °C a pane slumps in about one second. But cooling it doesn't slow the flow a little. It slows it past every human scale. By the glass transition near 560 °C the same pane needs roughly 550 years in this model. At room temperature we switch to separate estimates: Brill's historical 10²⁰ poise (10¹⁹ Pa·s) and the Physics FAQ's 10²² Pa·s extrapolation. They give about 4.31 billion and 4.31 trillion years, respectively. The lower result is below the 13.8-billion-year age of the universe, but both are far beyond a cathedral's history. These inputs illustrate the scale; they are not bounds for every glass.

3 · Then why is the old window thicker at the bottom?

Because of how it was made. Before the float-glass process was announced in 1959, window glass was made by several methods, including crown and cylinder methods. In the crown process, a glassblower opened and spun hot glass into a disc. Thickness could vary, especially around the central "bull's-eye" and sometimes the rim; much of a disc could be nearly uniform. Panes cut from it could inherit uneven thickness. When glaziers set uneven panes in lead, resting the heavy edge at the bottom would make sense. The illustration below has a fixed cut; choose how to hang its schematic pane.

The orientation experiment makes a limited point: an uneven pane can be hung with its thick edge in any direction. Downward flow would not explain a thick edge at the top. Robert Brill, of the Corning Museum of Glass, relays a secondhand remark attributed to Egon Orowan about panes thick at the top as well as the bottom. His own observation is more direct: thickness can vary noticeably within a single old fragment. He attributes that unevenness to manufacture. This account is not a survey of installation orientations, and it does not document panes hung thick-edge to the side.

4 · "But doesn't it flow a tiny bit?"

This is the honest hedge people retreat to, and it deserves a real number rather than a shrug. In 2017 a team including Ozgur Gulbiten and John Mauro tested a newly made sample matching a typical medieval glass composition used in Westminster Abbey. They used a model, checked against viscosity measurements, to calculate room-temperature flow. Their maximum: about one nanometre, a few atoms' width, over one billion years. Scaling that maximum to 800 years gives 8×10⁻⁷ nm, about 2.7 millionths of a 0.3 nm atomic diameter. "Technically nonzero" and "thick at the bottom you can see" remain very different: an explicit 1 mm comparison is about 12.1 orders of magnitude larger.

There is a twist worth naming, because it cuts against tidiness. That 2017 study found its medieval glass composition has a room-temperature viscosity about 16 orders of magnitude below one previous soda-lime study. Its inferred 10²⁴ to 10²⁵ Pa·s still gives no perceptible flow over a cathedral's history. This comparison is between particular studies, not the lowest viscosity ever cited. (Earlier, Edgar Zanotto's 1998 paper "Do cathedral glasses flow?" had put the relaxation timescale well beyond the age of the universe.)

The strongest evidence needs no model at all. There are glass objects far older than any cathedral — Roman vessels nearly 2,000 years old, ancient Egyptian glass older still — thin, delicate, and showing no sag whatsoever. If glass crept perceptibly under gravity, the oldest, thinnest pieces would have puddled long ago. They haven't. (And recent headlines about "a new liquid phase of glass" from Penn concern engineered nanometre-thin vapour-deposited films, not cathedral windows — they refine glass-transition science; they do not resurrect the flow myth.)