Jim Peters's idea about crystallographic orientation being important
to the phenomenon is certainly appealing. When I see these tubes grown
outside at large inclinations (a centimeter or three inclined more
than 30 degrees to the vertical) They seem to be dead straight, but
they typically have a "drip membrane" below them. The straightness is
striking, and I've always assumed it was related to a preferred
crystallographic growth direction for that reason.
I live beside a lake, and I have never observed the phenomenon on the
lake surface. Since it occurs in ice cube trays and plastic ice cream
buckets left outside partly filled with water I have always thought
that confinement of the liquid below a frozen surface played a role,
and that can't happen on a lake. I vaguely recall having seen more
than one spike growing from the same surface. I will check more
closely around the lake starting in December. I usually don't look
too closely at the ice there until it gets near to skateable, so I
probably miss seeing it at a good time for spikes. If spikes form on
the lake surface then I'm all wet about the extrusion mechanism.
I've realized that I made a serious error in my first post. While
convection should turn over the water which is cooled from the bottom
it obviously doesn't always do so. One frequently removes ice cubes
prematurely only to find an inclusion of liquid water surrounded by
frozen water. Sorry, Bill, I didn't think! Freezing at the bottom
first is not essential to the phenomenon, however, since I've observed
it in partially-frozen-from-the-top ice cream buckets left outside
overnight. Incidentally those, too often freeze with inclusions of
liquid water. I agree that rigid confinement of the liquid is likely
necessary, however, if the extrusion mechanism is to work, but the
walls of the container should be sufficiently rigid once the upper
surface is frozen over.
I hadn't realized that the upsicle was an unexplained phenomenon. Is
there no reference much earlier than that "Science News" article?