106 1. Higher order Fourier analysis
Theorem 1.6.14 (Inverse conjecture for the Gowers norms on Z). (See
[GrTaZi2010b].) Let f : [N] → C be such that f
≤1 and f
≥ δ. Then |f, ψ L2([N])|
1 for some degree ≤ s nilsequence of complex-
An extensive heuristic discussion of how this conjecture is proven can
be found in [GrTaZi2010]; for the purposes of this text, we shall simply
accept this theorem as a black box. For a discussion of the history of the
conjecture, including the cases s ≤ 3; see [GrTaZi2009]. An alternate proof
to Theorem 1.6.14 was recently also established in [CaSz2010], [Sz2010b].
These methods are based on the proof of an analogous ergodic-theory result
to Theorem 1.6.14, namely the description of the characteristic factors for
the Gowers-Host-Kra semi-norms in [HoKr2005], which we will not discuss
here, except to say that one of the main ideas is to construct, and then study,
spaces analogous to the Host-Kra groups
and Host-Kra nilmanifolds
associated to an arbitrary function f on a (limit-finite)
interval [N] (or of a function f on an ergodic measure-preserving system).
Exercise 1.6.21 (99% inverse theorem).
(i) (Straightening an approximately linear function) Let ε, κ 0. Let
ξ : [−N, N] → R/Z be a function such that |ξ(a+b)−ξ(a)−ξ(b)| ≤
κ for all but εN
of all a, b ∈ [−N, N] with a + b ∈ [−N, N]. If ε
is suﬃciently small, show that there exists an aﬃne linear function
n → αn + β with α, β ∈ R/Z such that |ξ(n) − αn − β|
all but δ(ε)N values of n ∈ [−N, N], where δ(ε) → 0 as ε → 0.
(Hint: One can take κ to be small. First find a way to lift ξ in a
nice manner from R/Z to R.)
(ii) Let f : [N] → C be such that f
≤ 1 and f
Show that there exists a polynomial P : Z → R/Z of degree ≤ s
such that f − e(P )
≤ δ, where δ = δs(ε) → 0 as ε → 0
(holding s fixed). Hint: Adapt the argument of the analogous
finite field statement. One cannot exploit the discrete nature of
polynomials any longer; and so one must use the preceding part of
the exercise as a substitute.
The inverse conjecture for the Gowers norms, when combined with the
equidistribution theory for nilsequences that we will turn to next, has a
number of consequences, analogous to the consequences for the finite field
analogues of these facts; see [GrTa2010b] for further discussion.
1.6.4. Equidistribution of nilsequences. In the subject of higher order
Fourier analysis and, in particular, in the proof of the inverse conjecture for
the Gowers norms, as well as in several of the applications of this conjecture,