The Universe of Discourse


Wed, 09 Jan 2008

Major screwups in mathematics
I don't remember how I got thinking about this, but for the past week or so I've been trying to think of a major screwup in mathematics. Specifically, I want a statement S such that:

  1. A purported (but erroneous) proof of S was published in the mathematical literature, so that
  2. S was generally accepted as true for a significant period of time, say at least two years, but
  3. S is actually false
I cannot think of an example.

There are many examples of statements that were believed without proof that turned out to be false, such as any number of decidability and completeness (non-)theorems. If it turns out that P=NP, this will be one of those type, but as yet there is no generally accepted proof to the contrary, so it is not an example. Similarly, if would be quite surprising to learn that the Goldbach conjecture was false, but at present mathematicians do not generally believe that it has been proved to be true, so the Goldbach conjecture is not an example of this type, and is unlikely ever to be.

There are a lot of results that could have gone one way or another, such as the three-dimensional kissing number problem. In this case some people believing they could go one way and some the other, and then they found that it was one way, but no proof to the contrary was ever widely accepted.

Then we have results like the independence of the parallel postulate, where people thought for a long time that it should be implied by the rest of Euclidean geometry, and tried to prove it, but couldn't, and eventually it was determined to be independent. But again, there was no generally accepted proof that it was implied by the other postulates. So mathematics got the right answer in this case: the mathematicians tried to prove a false statement, and failed, and then eventually figured it out.

Alfred Kempe is famous for producing an erroneous proof of the four-color map theorem, which was accepted for eleven years before the error was detected. But the four-color map theorem is true. I want an example of a false statement that was believed for years because of an erroneous proof.

If there isn't one, that is an astonishing declaration of success for all of mathematics and for its deductive methods. 2300 years without one major screwup!

It seems too good to be true. Is it?

Glossary for non-mathematicians

  • The "decidability and completeness" results I allude to include the fact that the only systems of mathematical axioms strong enough to prove all true statements of arithmetic, are those that are so strong that they also prove all the false statements of arithmetic. A number of results of this type were big surprises in the early part of the 20th century.

  • If "P=NP" were true, then it would be possible to efficiently find solutions to any problem whose solutions could be efficiently checked for correctness. For example, it is relatively easy to check to see if a proposed conference schedule puts two speakers in the same room at the same time, if it allots the right amount of time for each talk, if it uses no more than the available number of rooms, and so forth. But to generate such schedules seems to be a difficult matter in general. "P=NP" would imply that this problem, and many others that seem equally difficult, was actually easy.

  • The Goldbach conjecture says that every even number is the sum of two prime numbers.

  • The kissing number problem takes a red ping-pong ball and asks how many white ping-pong balls can simultaneously touch it. It is easy to see that there is room for 12 white balls. There is a lot of space left over, and for some time it was an open question whether there was a way to fit in a 13th. The answer turns out to be that there is not.

  • The four-color map theorem asks whether any geographical map (subject to certain restrictions) can be colored with only four colors such that no two adjacent regions are the same color. It is quite easy to see that at least four colors may be necessary (Belgium, France, Germany, and Luxembourg, for example), and not hard to show that five colors are sufficient.

  • Classical Greek geometry contained a number of "postulates", such as "any line can be extended to infinity" and "a circle can be drawn with any radius around any center", but the fifth one, the notorious "parallel postulate", was a complicated and obscure technical matter, which turns out to be equivalent to the statement that, for any line L and point P not on L, there is exactly one line L' through P parallel to L. This in turn is equivalent to the fact that classical geometry is done on a plane, and not on a curved surface.

[ Addendum 20080205: Readers suggested some examples, and I happened upon one myself. For a summary, see this month's addenda. I also wrote a detailed article about a mistake of Kurt Gödel's. ]

[ Addendum 20080206: Another article in this series, asking readers for examples of a different type of screwup. ]


[Other articles in category /math] permanent link