Department Colloquium

Department of Mathematics
Princeton University



November 8th, 2000


Arithmetic progressions of length four


The famous theorem of Szemer\'edi, proving a conjecture of Erd\"os and Tur\'an from 1936, asserts that for every $\delta>0$ and every natural number $k$ there exists an $N$ such that every subset $A\subset\{1,2,\dots,N\}$ of cardinality at least $\delta N$ contains an arithmetic progression of length $k$. When $k=2$ this assertion is trivial. The case $k=3$ was proved by Roth in 1953 using the circle method.
The case $k=4$ is much harder (for reasons that can be made quite explicit) and was not solved until 1969, when Szemer\'edi found a highly ingenious combinatorial argument, which over the next few years he was able to extend to progressions of arbitrary length. In 1977, Furstenberg discovered a completely different and more conceptual argument using ergodic theory, which led to many extensions of the original theorem.
Both these proofs ignored Roth's method, and indeed there are very serious obstacles to extending this method to progressions of length greater than three, as I shall demonstrate. However, it can be done, and this will be the main topic of the talk. One of the main advantages of the new approach to the theorem is that it gives very greatly improved bounds for the dependence of $N$ on $k$ and $\delta$. Another is that the proof is much more closely related to results in additive number theory and may eventually lead to the solution of problems in that area.