I started asking myself this question because I was somehow
unsatisfied with the answers to the previous question "Performance of a
thermos bottle relative to contents"
concerning a very specific kind of container. There were assumptions
made in the answers. Even though these assumptions were fair, given
that a thermos bottle is a rather precise and well known object, the
reasonning in the answers did not make them explicit. I tried to avoid
it, by reasonning explicitly about the cork of the bottle, but I still
used properties of the bottle shape without saying so explicitly (I
became aware of it later). And even though my assumption was
close to the actual facts, a thermos bottle is not a cylinder. The
bottom is often somewhat spherical, which could have called for an
extra line of justifications (how high should the bottle be compared
to the radius of the bottom half sphere, unless the top is also
considered a half-sphere ?).
Sometimes, we also make (explicit or implicit) assumptions that are
The other thing that bothered me is that people will often vote for
simple answers they understand quickly (not necessarily the best answer or even
a correct one). At least that is the feeling I get. If warranted, this would
justify making lots of unstated assumptions when answering. Not to
mention the fact that fast answers get a better chance at upvotes,
Then, considering the thermos question, I started wondering about what
could make our statements wrong, and what could be the assumptions
that are often made implicitly, just for that kind of problem (though I
actually made one or two explicit in stating this new problem). Here are
some such assumptions, probably an incomplete list (other ideas are welcome):
role of the cork: can it be ignored as not significant ?
homogeneity of the bottle sides: is it the same kind of material all over ?
shape of the container: is it just a bottle, which we tend to assume ?
heat conductivity of the bottle side: is it isotropic ?
uniformity of liquid cooling: well, that is always wrong, but liquid conductivity is so efficient that it seems a good approximation. Is it?
Then I wondered whether falsifying these assumptions could also
falsify the conclusion.
Once I had satisfied myself that it was the case, I asked the
question, carefully stated so as not to induce any assumption (for
example by always using the word "container" instead of "bottle"). I
was not trying to trap anyone, only experimenting. I unfortunately got
few reactions (thank you to those who did react). I should have started
that more anonymously as some users clearly wondered what I was after
So here is my answer.
The picture is a cross-section image of a container that will cool faster
when it is full than when it is half-full. It consists of a large disk on
top of a sphere, with the same volume so that only the sphere contains
liquid when it is half-full. The opening between them is large so that
heat can flow easily between the two parts when it is full.
Clearly, the disk has a very large surface to volume ratio and will
act effectively as a radiator to cool the liquid it contains, while
the shere has the smallest possible ratio and will not cool fast.
However, when the container is full, heat will flow through the liquid
from the sphere to the disk so that all the liquid content will cool
rather fast, though the disk will get cooler faster than the sphere.
If the container is only half-full, only the sphere contains warm
liquid, with a small surface/volume ratio. Hence it will cool more
If the container is large enough, this should be sufficient.
It can be improved by remarking that a horizontal disk shape is not
very good for convection heat transfer. Replacing the hollow disk with
an inverted bell shape would work better.
You can even reinforce further the effect by using for the side of the
container a material that is more heat conductive transversally than
laterally, so that heat goes out quickly but is not conducted
efficiently from one part of the container side to another. That
avoids heat being transferred to the disk when the container is half
full, and still permits the disk to cool efficiently the liquid it
contains when full. It can be easily produced by using a heat insulating
material with copper nails piercing it at close regular intervals.
The same effect could be achieved with a bottle having a standard
shape, but with isolation only in the bottom part. This is somewhat
close to what I said about the effect of a conductive cork in the
Of course, this is no major discovery in elementary physics. At best a
moderately easy puzzle game. But it may be telling about our
Coming back to the remark about votes. I am wondering whether the initial
downvote for that question (without an explicitly related explanatory
comment) was motivated by such an unstated assumption. Was it really
Now, it is possible that people who are very proficient in a field
will vote more accurately. It is probably more the case, but I think not always (I do have one example
in mind, not from physics). Proficiency is an ill-defined concept,
and schools of thought are often biased, even in hard sciences.