Abels Theorem on Fourier Series

March 2, 2011

(n) = cn are known and f is Abels theorem allows us to conclude that if the Fourier coecients f piecewise continuous then f is determined. Denition 1. Let 0 < r < 1 and dene
+

Ar f (x) =

cn r|n| einx .

This series converges absolutely and uniformly in x to a continuous function of x for each r < 1. Theorem 1. If f is piecewise continuous 1 lim Ar f (x) = [f (x+ ) + f (x )]. 2 r1 If f is continuous, given an , there is a so that |Ar f (x) f (x)| < for all r such that |r 1| < . We say Ar f converges uniformly in x to f . Proof. Let Pr (t) =
1 2 + |n| int r e .

Let z = reit . Then 1 2 1 = 2 1 = 2 1 z + 1z 1z 2 1 |z | |1 z |2 1 r2 , and 1 + r2 2r cos(t)

Pr (x) =

1 = [1 + 2

2rn cos(nt)]
n=1

Integrating the last series term-by-term with respect to t we get

Pr (t)dt =
0

1 Pr (t)dt = . 2

Now let > 0 and suppose t . By calculus we nd that the minimum of 1 + r2 2r cos(t) on this interval is 1 + r2 2r cos( ). Hence on t 0 < Pr (t) 1 1 r2 . 2 1 + r2 2r cos( ) 1 (1)

Let us change variables and use periodicity, as in Dirichlets theorem to write

Ar f (x0 ) =

f (x0 + t)Pr (t)dt.

+ Fix x0 and choose so that |f (x0 + t) f (x 0 )| if t < 0 and |f (x0 + t) f (x0 )| if 0 < t . Now that has been chosen, pick so that 0 Pr (t) < if 0 < 1 r < when |t| , which we can do by (1). Then

1 Ar f (x0 ) [f (x+ ) + f (x ) = 2 +

[f (x0 + t) f (x 0 )]Pr (t)dt +

[f (x0 + t) f (x 0 )]Pr (t)dt

[f (x0 + t) f (x 0 )]Pr (t)dt +

[f (x0 + t) f (x 0 )]Pr (t)dt

= I + II + III + IV. Well rst estimate III . The estimate on II is similar.

|III |
0

Pr (t)dt
0

Pr (t)dt = . 2

Next we estimate I (IV is similar).

|I | |

[f (x0 + t) f (x 0 )]Pr (t)dt| 2M | | 2M ,

where |f | M . So altogether we get 1 |Ar f (x0 ) [f (x+ ) + f (x )| + 4M , 2 when 0 < 1 r < . This proves the rst statement. The chosen depends on x0 and hence depends on x0 . But if f is continuous on [, ] it is uniformly continuous, so can be chosen independent of x0 and then does not depend on x0 . Ar f (x) is uniformly close to f (x) if r is close enough to 1.