Sie sind auf Seite 1von 2

almost a proof of the Riemann hypothesis

R.H.
March 16, 2016

1 proof
consider the set of completely multiplicative functions

f (nm) = f (n)f (m)

such that
f (p) = ±1 if p is prime
call this set F .
consider the corresponding Dirichlet series :

X
F(s) = f (n)n−s
n=1

in this article we prove that consider the value of f (p) as a fair coin toss
random variable, i.e.
1
P(f (p) = 1) = P(f (p) = −1) =
2
then
E[f (n)] = |λ(n)|
where λ(n) is the Liouville function, i.e |λ(n2 )| = 1 and |λ(n)| = 0 if n is not a
perfect square.
hence
 P 
−σ 2 = E PN f (n)n−σ PN f (m)m−σ
h  i
E Nn=1 f (n)n n=1 m=1

 N N  N N N X
N N
 X X  X X X X X
= E  f (nm)(nm)−σ  =
 (nm)−σ E[f (nm)] = |λ(nm)|(nm)−σ = k −2σ 1
n=1 m=1 n=1 m=1 n=1 m=1 k=1 d|k,d≤N

and if σ > 1, letting N → ∞ we get :

1
h i N
X X N
X X
E |F(σ )|2 = lim k −2σ 1= k −2σ 1 = ζ 2 (2σ )
N →∞
k=1 d|k,d≤N k=1 d|k
h i
hence, E |F(σ )|2 converges for σ > 1/2 !
this means that for almost every choice of f (p), we get that |F(σ )|2 is con-
vergent for σ > 1/2, hence F(σ ) is also convergent and the Riemann hypothesis
is true for that choice of f (p) !
in the same way, with
∞ ∞
F(s) X 1 X
= f (n)|µ(n)|n−s , = f (n)µ(n)n−s
F(2s) F(s)
n=1 n=1

we get
F(σ ) 2 1 2
" # " #
E = E = ζ(2σ )
F(2σ ) F(σ )
both converging for σ > 1/2
now, instead of considering any ±1 sequence for f (p), we consider only
periodic sequences. this means thatf (n) still is completely multiplicative, and
c(n) is a Q periodic Q sequence
Q which is now the considered random variable,
and f (p) = c(p), f ( i pi ) = c(pi ).
PQ−1
we also restrict c(0) = 0, Q to be prime, and n=1 c(n) = 0. hence, if c(n) is
multiplicative then f (n) = c(n) is a Dirichlet character. in the general case, c(n)
is not multiplicative but
X
ln F(s) = C(χ) ln L(s, χ)
χ (mod Q),χ,χ0

for some coefficients C(χ).


we prove again that :

E[f (n)] = |λ(n)|


if now the random variables are the C(χ), then :
P
k C(k) ln(s,χk )
Y 1
F(s) = e = P −s
p
1−( k C(k)χ k (p)) p

given that they are independent :


Y YZ b Y e2b<(ln(s,χk )) − e2a<(ln(s,χk ))
2 2Re(C(k) ln(s,χk ))
E[|F(s)| ] = E[e ]= e2x<(ln(s,χk )) dx =
a <(ln(s, χk ))
k k k

Das könnte Ihnen auch gefallen