How Do Snowflakes Form? Why Every One Has Six Sides

Water vapour in a cloud does not simply freeze when the air drops below zero. Pure water droplets can stay liquid to nearly minus forty, because freezing needs a starting point, and in a cloud the starting point is a speck of something else, a grain of clay, a fragment of soot, a bacterium, on which molecules can settle in an orderly way. Once a tiny ice crystal exists, it grows: water vapour in the surrounding air deposits directly onto it as ice, skipping the liquid stage, and the crystal draws in vapour from the supercooled droplets around it, which evaporate to feed it. A snowflake is not a frozen raindrop. It is built from vapour, molecule by molecule, over the fifteen minutes to an hour it spends falling through the cloud.

A water molecule is a bent shape, an oxygen atom with two hydrogens at an angle of about 104 degrees, and when molecules freeze they link hydrogen to oxygen in a lattice whose most stable form is a stack of hexagonal rings, like a honeycomb. The six-fold symmetry of the lattice is invisible in a block of ice but shows in a growing crystal, because the six corners of a hexagonal plate stick out slightly further into the vapour than the flat sides, gather molecules faster, and grow into arms, and each arm's edges do the same at smaller scale, producing the branching that makes a star. Whatever the crystal does, it does it six times at once, because the six corners are in identical surroundings.

The habit of a snow crystal, whether it grows as a flat plate, a branching star, a column or a needle, depends on the temperature of the cloud and on how much excess vapour the air holds, and the dependence was mapped in the 1930s by the Japanese physicist Ukichiro Nakaya, who grew snowflakes on a rabbit hair in a cold laboratory and drew the first diagram of it:

• •Around minus 2 degrees: thin plates
• •Around minus 5: needles and hollow columns
• •Around minus 15: the large, branching stars of Christmas cards, which grow fastest of all and only in a narrow band of temperature
• •Below minus 22: plates and columns again, and at the coldest temperatures small solid prisms
• •More moisture at any temperature: more branching and more delicate structure; less moisture: simpler, solid shapes

A crystal that falls through a cloud passes through changing temperature and humidity, and because all six arms experience the same changes at the same time, they grow in the same way and the flake stays symmetrical. But no two crystals take the same path, and a star that grew plates at one height, branches at another and side-arms at a third has a history that will not be repeated. A large stellar crystal contains around a hundred quintillion water molecules, and the number of ways of arranging its branches is so much larger than the number of snowflakes that have fallen on Earth that two matching ones are, in practice, impossible. Small, simple prisms grown in stable conditions can look identical under a microscope, and researchers have made twin crystals in the laboratory by growing two side by side, but the elaborate flakes of a snowstorm are each unique.

What reaches the ground is rarely a single crystal. Flakes collide and stick, especially near freezing when their surfaces are slightly wet, and the large fluffy snowflakes of a mild snowfall are clusters of dozens or hundreds of crystals. A snowflake can also fall through a layer of supercooled droplets that freeze onto it as it goes, coating it in rime until it becomes a pellet of graupel, or through warm air that melts it to sleet or rain. Fresh snow is around ninety percent air, which is why it insulates the ground, muffles sound and can be squeezed to a tenth of its volume, and why ten centimetres of it is worth about a centimetre of rain. Wilson Bentley, the Vermont farmer, caught his flakes on black velvet and photographed them through a microscope before they could melt, from 1885 until his death in 1931 of pneumonia caught walking home through a blizzard; Nakaya, who put the shapes on a chart, called snow crystals letters sent from the sky.

A single ice crystal is as transparent as glass, yet a field of fresh snow is one of the brightest surfaces in nature. The difference is the number of edges. Every time light crosses a boundary between ice and air, some of it is reflected and the rest is bent in a new direction, and a layer of snow contains countless crystal faces. A ray entering it bounces from face to face until it comes back out, and because ice absorbs all visible colours almost equally, the light that emerges is the same mix that went in, which we see as white.

Ice is not perfectly colourless, though. It absorbs slightly more red light than blue, an effect far too weak to notice in a thin crystal but noticeable over long paths. Light that travels deep into a snowbank, or into the dense ice of a glacier, loses more of its red on the way, which is why a hole poked into deep snow glows blue and why glacier crevasses look turquoise. Old, compacted snow and glacial ice have fewer air gaps to turn the light back quickly, so it travels further inside them and comes out bluer.

You do not need a laboratory to see the shapes described here, only a cold day and a little preparation. Leave a piece of dark fabric, such as black felt, a wool glove or a dark card, outside or in a freezer for a while so it is below freezing before any flakes touch it. A warm surface melts crystals instantly. Hold it out during a light snowfall, ideally on a calm day when the air is well below zero, and look closely with a magnifying glass or the macro mode of a phone camera. Breathe away from the cloth, since your warm breath can melt or blur the crystals.

The best conditions are cold, gentle snowfalls, because wet snow near zero tends to arrive as clumped aggregates, and windy weather breaks the arms. You will probably see far fewer perfect stars than you expect. Many crystals are broken, lopsided, rimed with tiny frozen droplets, or simple plates and columns, which is a useful reminder that the textbook flake is the exception. Try to notice how the shapes change as the temperature drops through the day, and compare what you find with Nakaya's chart. It is the same kind of patient observation that Bentley carried out for decades.

The need for a starting speck has been put to deliberate use. In 1946 Vincent Schaefer, working at the General Electric laboratory in New York State, dropped dry ice into a freezer full of supercooled cloud and watched it fill with ice crystals. His colleague Bernard Vonnegut then found that silver iodide could start ice forming at milder temperatures, because its crystals are arranged in a pattern very like the hexagonal lattice of ice, giving water molecules a ready-made template. Silver iodide is still sprayed into clouds today, though how much extra rain or snow it really produces is hard to measure and still argued over.

Ski resorts face a different problem and solve it differently. A snow cannon sprays fine water droplets into cold air, often with compressed air to break them up, and the droplets freeze as they fly. The result is not snowflakes at all but tiny beads of ice, frozen drops rather than crystals grown from vapour. Machine-made snow therefore has none of the delicate arms of natural flakes, packs more densely and feels firmer and icier underfoot, which skiers notice at once. Some resorts add ice-nucleating material, including a protein taken from a bacterium, to help the drops freeze in marginal weather.

People were asking why snowflakes have six sides long before anyone knew what a molecule was. In 1611 the astronomer Johannes Kepler, claiming he had nothing to give his patron in Prague for the New Year, wrote him a short Latin essay instead, On the Six-Cornered Snowflake, playing on the similarity between nix, the Latin for snow, and the German word for nothing. Kepler noticed that honeycombs and the seeds packed inside a pomegranate also form six-sided patterns, and wondered whether snow was built from tiny round particles stacked as tightly as possible. He could not settle the question, and admitted as much.

His side question about stacking became one of the most stubborn problems in mathematics. Kepler asserted that the arrangement greengrocers use for oranges, each layer sitting in the hollows of the one below, is the densest possible way to pack equal spheres, filling about 74 percent of the space. That claim, the Kepler conjecture, resisted proof for nearly four centuries until Thomas Hales produced one in 1998 that relied on heavy computer calculation, confirmed by a computer-checked formal proof in 2014. The real reason for the snowflake's six sides had to wait for X-ray studies of crystals in the twentieth century, which revealed the hexagonal arrangement of molecules in ice.

A snowflake grows from a speck of dust in a cloud as water vapour deposits directly onto it as ice, and it has six-fold symmetry because water molecules freeze into a hexagonal lattice whose corners grow fastest. Temperature and humidity decide whether it becomes a plate, a column, a needle or a branching star, and because each crystal falls through its own sequence of conditions, its arms match each other while no two flakes match.