The first three experiments represent “normal” ways for me to make ice: one or two trays of tap water left for a day (or so) on either the top or bottom shelf of the freezer.
The baseline experiment is my normal configuration: Two trays filled with tap water, stacked together on the upper shelf of my freezer.
Freezer temperature is measured in FUs: that’s right, freezer units. Time is hours and minutes. The results for each tray are a rough percentage of cubes that were good cubes.
If there’s not enough space on the upper shelf, I might stack the trays on the lower shelf. There’s a weird piece of meat in my freezer. I mean, what animal did that come from?
The bottom tray did much better this time.
If I have lots of space—I mean lots—I might put one tray on the upper shelf and one on the lower.
Before David, Jess, and I formed hypotheses, we performed a few more experiments, varying the the basic variables in fairly non-systematic ways. Below are the summarized results.
There’s a new column rate for modifiers to the rate of freezing (given a certain freezer temperature); legal values are towel-wrap, towel-lid, towel-nest, and fan. We didn’t actually do the fan thing, but we did the towel thing.
‡towel-lid: David says, “Paper towels covering top, sides, and most of bottom.”
The top tray in a normal stack produces good cubes because it freezes (relatively) quickly; the bottom tray in a normal stack produces bad cubes because it freezes (relatively) slowly. A single tray produces good cubes because it freezes quickly.
Quick freezing causes small-grained crystalization. Small-grained crystalization (as with metal alloys) produces a strong crystal without regular fault lines; for ice cubes, this means a cube that will not shatter when you twist the tray.
Under normal conditions, a single tray will freeze quickly enough to produce strong cubes. Typically a tray loses most of its heat through the top, due to evaporation and convection; not much of its heat is lost through the conduction of the (plastic) tray. Therefore, the heat loss of the top tray of a stack is mostly unaffected by the stack, but the heat loss of the bottom tray is slowed because the top tray blocks the air passage above the bottom tray, hindering convection and evaportion.
The key quality of the good ice cube is strength.
The key variables are the ones that affect overall freezing rate.
The top tray in a normal stack produces good cubes because it freezes from the top-down; the bottom tray in a normal stack produces bad cubes because it freezes from the bottom. A single tray produces good cubes because it freezes from the top-down.
A simple explanation for this hypothesis goes something like this: The forming ice crystal nucleates either at the water/air boundary or at the tray walls. Either way, the rest of the freezing water molecules pig-pile on the originals: if the seed crystal started on the sides, then the whole crystal will be stuck to the sides; if the seed crystal formed at the top, then the whole crystal is stuck to nothing.
But surely all cubes, good and bad, freeze from all sides. A better explanation is needed.
We’ve noticed that good cubes, during freezing, often rise above the top of the tray and away from the sides, leaving a small air gap and a thin layer of frost. If the surface froze first, and solidly, subsequent freezing (and expansion) below would raise it like a little elevator platform. This motion—and perhaps the wedge shape of the tray—prevents the cube from bonding to the side of the tray during freezing.
The bad cubes also often rise above the top of the tray, but in a different way: by bulging up in the middle, so that the surface of the cube becomes convex. Perhaps this indicates that expansion during freezing is not causing any movement at the sides of the tray.
The key quality of the good cube is non-stickiness to the tray.
The key variables are the ones that affect freezing gradient.
An outside entry! You can’t say we didn’t research the field: domestic hint scientist Heloise has already documented a hypothesis for shattering ice cubes here. Heloise’s treatment is rather short, so we extrapolate.
Trays produce good cubes when they are free of mineral deposits.
This is an incomplete hypothesis for the purposes of our study, because it does not seem to explain the difference between top and bottom tray positions in a stack. However, note its compatibility with the Top-Down Hypothesis; perhaps having mineral deposits on the tray is either a necessary or sufficient (but not both) condition for the formation of bad cubes starting from the sides and bottom of a tray.
The next page contains experiments designed to prove or disprove certain hypotheses. If you’d like, you can register your vote before reading the results:
Karl is betting on the Top-Down Hypothesis, and has been for a while.
David says 200 quatloos on the newcomer.
Jess refuses to commit. There are so many answers, each right in its own special way.