Skip to content

Does information weigh anything?

Erasing one bit releases a minimum quantity of heat. It was predicted in 1961 and measured in 2012.

Shallowsdepth, meaning: Interesting questions with substantial evidence and real uncertainty remaining.Known. Strongly supported by evidence and broadly established.Inspect the question
00

The question

In the twentieth century, physicists as eminent as von Neumann, Gabor and Brillouin believed that every elementary step of information processing had an unavoidable cost in heat.

Measuring and copying turned out to be free in principle. One operation is not, and it is the humblest: throwing something away.

What has been shown

Known. Strongly supported by evidence and broadly established.

Strongly supported by evidence and broadly established.

The principle

In 1961 Rolf Landauer argued that logically irreversible operations, those that map two different inputs to the same output, must dissipate energy. Erasing a bit reduces the number of states the system can be in, and that reduction has to be paid for in heat released to the environment.

The minimum is small: at room temperature, about three zeptojoules per bit. Today’s computers dissipate far more, so the bound is not what currently limits them.

Where the argument came from

The 1961 paper is about computing hardware rather than about philosophy. It appeared in IBM’s research journal and opens on how fast and how compact computing circuits could get, and what sets the floor.

His answer was that the floor is not in the electronics. It is in the logic: an operation that loses information about its input cannot be run backwards, and irreversibility in the logic forces dissipation in the physics.

It was measured

In 2012 an experiment trapped a single microscopic bead in a double-well potential, used the two wells as the two states of a bit, and measured the heat released while the bit was erased.

The slower the erasure, the closer the dissipated heat came to the predicted minimum, where it levelled off. A fifty-year-old theoretical claim about the relationship between information and thermodynamics became an experimental result.

Experiments since, with a particle in a feedback trap and with nanoscale magnetic bits of the kind used for data storage, have found the same floor.

Why erasing is the expensive operation

Observed
Copying a bit onto a blank one costs nothing in principle.
Inferred
Nothing is lost. The operation can be run backwards.
Observed
Erasing a bit costs at least kT ln 2 of heat.
Inferred
Two possible states become one. The lost distinction has to go somewhere, and it goes into the surroundings.
Observed
The measured heat approaches that floor as erasure slows.
Inferred
The bound is physical rather than a bookkeeping convention.

The usual answer to a famous demon

Maxwell imagined a being that sorts fast molecules from slow ones, creating a temperature difference without work and apparently breaking the second law.

The standard resolution, reached by O. Penrose in 1970 and independently by Charles Bennett in 1982, is that the demon must record which molecules are which, and its memory is finite. Erasing those records costs at least as much as the sorting gained. On this account the second law survives because information processing is part of thermodynamics rather than outside it.

Earman and Norton, writing in 1998 and 1999, do not accept that this is what does the work. If the demon already obeys the second law, no claim about information is needed to save the law, and if it does not, no claim about information will. Bennett grants the point, and defends the principle for what it explains.

The demon fails not because it cannot sort, but because it has to remember.

What is claimed beyond it

Uncertain. Meaningful scientific, historical or philosophical disagreement remains.

Meaningful scientific, historical or philosophical disagreement remains.

How far the identification goes

That erasure has a thermodynamic cost is widely accepted and has been measured. That the universe is in some deeper sense made of information is a much stronger claim, and it is a research programme rather than a result.

Ideas connecting entanglement to the structure of spacetime are among the more concrete versions, and they are mathematical relationships in particular model spacetimes rather than statements about this one.

On a route

The bottom of things

Empty space and information, then the smallest length, black holes and time, where quantum theory and gravity meet and no tested theory joins them, and last, why there is anything at all.

02 of 06This question is stop 2 of 6 on The bottom of things, a route through the Ocean of Reality.
The currents

Where this question leads

See it on the map

Inspecting a passage

Loading the evidence record…