A FUNCTIONAL DETERMINANT EXPRESSION FOR THE RIEMANN XI-FUNCTION

Jose Javier Garcia Moreta Graduate student of Physics at the UPV/EHU (University of Basque country) In Solid State Physics Addres: Practicantes Adan y Grijalba 2 5 G P.O 644 48920 Portugalete Vizcaya (Spain) Phone: (00) 34 685 77 16 53 E-mail: josegarc2002@yahoo.es

MSC: 34L05 , 34L15, 65F40 , 35Q40 , 81Q05 , 81Q50

ABSTRACT: We give an interpretation of the Riemann Xi-function .

the quotient of two functional determinants of an Hermitian Hamiltonian

(s) as H=

To get the potential of this Hamiltonian we use the WKB method to

approximate and evaluate the spectral Theta function the Riemann zeros on the critical strip the functional determinant

0 < Re(s) < 1

2 exp(tn ) over n . Using the WKB method

(t) = H

we manage to get the potential inside the Hamiltonian

, also we evaluate

det(H + z 2 )

by means of Zeta regularization, we

discuss the similarity of our method to the method applied to get the Zeros of the Selberg Zeta function. In this paper and for simplicity we use units so

2m = 1 = Keywords:

Riemann Hypothesis, Functional determinant, WKB semiclas-

sical Approximation , Trace formula ,Bolte's law, Quantum chaos.

1. Riemann Zeta function and Selberg Zeta function

Consider a Riemann surface with constant negative curvature. of the modular group this surface can be constructed as the upper complex plane divided by a discrete subgroup

P SL(2, R)

, Selberg [14] studied the problem of the 2-

dimensional Laplacian with the metric

ds2 =

dx2 +dy 2 , and discrete boundary y2

conditions are imposed for the discrete group

P SL(2, R) 1 2 + kn 4 (1)

n (x, y) = y 2
These momenta

2 2 + 2 x2 y

n (x, y) = En n (x, y) En =

kn

are the non-trivial zeros of the Selberg Zeta function, which

can be dened by an Euler product over the Geodesic of the surface in an

analogy with the Riemann Zeta function

1 (s) = = ns n=1

1 (1 ps ) n

Z(s) =
P k=0

1 N (P )(s+k)

(2)

In both cases the Riemann Zeta function and the Selberg Zeta function can be expressed by an innite product Selberg also studied a Trace formula which relates the Zeros (momenta of the Laplacian

) on the critical line Z (D) 4

0

1 2

+ ikn = 0 and the length of the Geodesic ln N (P ) g(ln N (P ) N (P )1/2 N (P )1/2 P p.p.o (3)

of the Surface in the form

h(kn ) =
n

dkkh(k) tanh (k)+

Here, p.p.o means that we are taking the sum over the length of the Geodesic,

h(k)

is a test function and g(k) is the Fourier cosine transform of h(k) g(k) = 1 2 0 dxh(x) cos(kx) (D) is the area of the fundamental domain describing the Riemann surface . In case we had a surface with the length of the Geodesic

ln N (P ) = ln p

for `p' on the second side of the equation a prime number,then

the Selberg Trace is very similar to the Riemann-Weil sum formula [12]

h() = 2h

(n) i 1 g(ln n)+ g(0) ln 2 2 2 n n=1

dsh(s)

1 is + 4 2 (4)

This formula (4) related a sum over the imaginary part of the Riemann zeros to another sum over the primes, here

(n) =
1 Z(s)

integer is the Mangoldt function, in case Selberg and Riemann are related by

ln p n = pk with `k' a positive 0 otherwise ln N (P ) = ln p both zeta function of

=
n=0

(n + s)

and their logarithmic

derivative is quite similar if we set the function

geodesic (P ) =

ln N (P ) 1N (P )1

(s) = (n)ns n=1

Z (s) = Z

geodesic (P )N (P )s
P p.p.o

(5)

In both cases the Riemann and Selberg zeta functions obey a similar functional equation which relates the value at s and 1-s

(1 s) = X(s)(s)

Z(1 s) = exp

(D) 4

s1/2

vtan(v)dv + c Z(s)
0

The constant of integration `c' is determined by setting

2 (2)

s = 1/2

, and

(6) X(s) =

(s) cos

s 2

for the case of the Riemann zeta function.

With the aid of the Selberg Trace formula (3) , we can evaluate the Eigenvalue staircase for the Laplacian

2 2 = y 2 x + y

E=

1 + p2 4

=
En E

1=
n

(D) 4

dkkh(k) tanh (k)+

0

1 arg Z

1 + ip 2 (7)

E 1 , we can immediatly see that the smooth part of (7) satisfy Here p = 4 (D) Weyl's law in dimension 2 Nsmooth (E) 4 E , the oscillatory part of (7) 1 satisfy Bolte's semiclassical law [4] (page 34, theorem 2.10 ) arg Z 2 + i E 1 with = 1 , the branch of the logarithm inside (7) is chosen, so arg Z 2 = 0 in this case the Selberg Zeta function is the dynamical zeta function of a Quantum
system and the Energies are related to the zeros of

Z(s)

2. A functional determinant for the Riemann Xi function

(s)

From the analogies between the Riemann Zeta function and the Selberg Zeta function, we could ask ourselves if there is a Hamiltonian operator (the simplest second order dierential operator which has a classical and quantum meaning and it is well studied ) in the form

Hn (x) =

d2 n (x) 2 + V (x)n (x) = En n (x) n (0) = 0 = n () En = n dx2 V (x) = f (x) x > 0 x0 (8)

f (x) dened inside the equation (8) for the potential V (x) and H = p2 + V (x) must be evaluated. The idea here is to choose f (x) so the Energies of the Hamiltonian are the square 2 of the imaginary part of the Riemann zeros En = n .,
The function the Hamiltonian In this paper we will prove that this function can be obtained within the WKB approach as

dx 2 the Hamiltonian dened in (8) .

the exact eigenvalue staircase

f 1 (x) = 2

d2

1 1

N (x)

,here

N (x)

is the Eigenvalue staircase of

For the case of this Hamiltonian, which involves the imaginary part of the zeros

N (x)

can be evaluated [9]

N (E) =
n

1 H (E En ) = arg
1 x>0 ,

1 +i E 2
1 4

1 = 1+ arg
T 2

1 ( E) +i E + 2 (9)
T 2

With

H(x) =

x<0

(T ) = arg

+ iT 2

ln

ln

T 2e

1 48T

+ ...

Also we will prove how the Riemann Xi function

(1 s)

is proportional to the functional

s (s) = s(s1) 2 2 (s) = 2 determinant det (H s(1 s)) , and 2 p2 n

we will also show that the density of states can be evaluated from the argument of the Xi-function

E = p2

1 d 2p dp

mlogdet (H + i p) = (E) =
n

As a simple example of how Quantum Mechanics can help to solve problems of nding the roots of functions , let be a particle moving inside an innite potential well , the energy is given by equation [7] in units

E = p2 and the one dimensional Schrdinger = 2m = 1 ( is the reduced Planck's constant with value

= 1.05.1034 J.T 1 H0 un (x) =

d2 un (x) + V (x)un (x) = En un (x) dx2 un (x) = A sin (x) ,

un (0)

= 0 = un () En = n2 (10)

n this case the Euler's product formula for the sine function is the quotient between 2 functional determinants

sin ( x) x = 1 En x n=1

det (H0 x) det (H0 )

We can also compute the density of states to get the Poisson sum formula

1 (E) = (E En ) = 2p n=1

(p n) +
n n

(p n)

1 = e2inp 2p n= (12)

o Zeta regularized determinant for

(s)

Given an Operator P with real Eigenvalues larized determinant [6] in the form

{En }

, we can dene its Zeta regu-

det P + k 2 = exp
Here

d P (s, k 2 ) |s=0 ds
s

(13)

is the Spectral Zeta funcn tion of the operator taken over all the Eigenvalues, the relationship between this spectral zeta function and the Theta function

P (s, k 2 ) = T r (P + k 2 )s =

En + k 2

(t) =
n

exp(tEn ) =
n=0 1 (En +k2 )s

0 1 (s)

dN (x)ext

,t>0 always , is given by the Mellin transform

dt tk2 (t)ts1 . If P is a Hamiltonian we can obtain the Theta funct e 0

tion

(t) =
n

exp(tEn )

(approximately) by an integral over the Phase space

[7]

(t) =
n=0

exp (tEn )

1 2

dp
0

dxetp

tf (x)

1 = 2 t

dxetf (x) = W KB (t)

0

(14)
If we compare the semiclassical Theta function (14) and the spectral Theta function

(t) =
n

exp(tEn )

then we nd

(t) =
n=0

exp (tEn ) = s
0

dtN (t)est

1 2

dx
0

dp exp(tp2 tf (x)) (15) 1 = 2 t

1 = 2

1 dq dp exp(tp2 tf (x)) = 2 t

dxetf (x)
0

dretr
0

dV 1 (r) dr

From expressions (14) and (15) and setting

(16) N (0) = 0 (after changes of variable)

s
0

1 dxN (x)esx = 2

d1/2 dxf 1 (x)esx f 1 (x) = 2 1/2 N (x) dx (17)

To prove (16) and (17) we have used the properties of the integral representation for the Laplace inverse transform

D ekt = k ekt

D f (t) =

1 2i

c+i

dsF (s)est s
ci

(18)

L {f (t)} = L {g(t)} im 1 1 arg 2 + i E 1 2 1 (Bolte's semiclassical law in one dimension) so f (x) = d 21 arg 1 + i x 2
nd the fact that if two Laplace transforms are equal then plies that

f (t) = g(t) , for the case of the Riemann Zeros N (E) =

dx 2

, since we want our potential inside (8) to be positive whenever we take the inverse we must choose the POSITIVE branch of the inverse in order to get

f (x) 0 on the interval [0, ) , d 2 f (x) dx 2

1 1

the half derivative and the half integral for any

well behaved function are given in [13]

1 (1/2)

x 0

f (t) dt xt
1

d 2 f (x) dx 2
2 d 21 dx 2
1 1

1 d (1/2) dx

x 0

dtf (t) xt

(19)

We have written implicitly the potential inside (8) , if the function by the functional equation

(x) =

arg

1 2

f (x) is dened 2 H(xn ) +i x = 2 , 2

n xn

then we may evaluate the Spectral Zeta function of the Quantum system given in (8), then

det H + z 2 d d = exp P (s, z 2 ) |s=0 + P (s, 0) |s=0 det (H) ds ds

(z + 1/2) (1/2) (20)

For the potential dened by

f 1 (x) =

2 d 21 dx 2

arg
n

1 2

+i x

, we can evaluate

the Theta kernel using (14) ,(15) and (16)

(t) =

etEn =

1 2 t

dx df dx(x) etx

, for this potential the spectral theta function and its derivative are

1 H (s, z ) = s 2 (n + z 2 ) n=0
2

d H (0, z 2 ) ds

2 ln n + z 2 n=0

1 + in = 0 2 (21)

Taking exponentials we reach to the innite product for the Riemann Xi-function as an spectral determinant (functional determinant over the Eigenvalues of H)

det H + z 2 = det(H)

n=0

2 (n + z 2 )

=
2 n n=0

1+

z2 En

(1/2 + z) (1/2)
1

(22)

n=0
If we choose the positive branch

2 d 21 arg 1 2 dx 2 then the potential will be always positive so the Energies of the Hamiltonian + 2 inside (8) will be all positive En = n R then all the non-trivial zeros of the Riemann Zeta function will be on the critical 1 line Re(s) = 2 1 , with a simple change of variable z = s 2 we obtain

f (x) 0 of the inverse f 1 (x) =

+i x

det H s(1 s) + (s) = (0) det H + 1 4

1 4

(1 s) = (0)

(23)

Equation (22) is the Hadamard product for the Riemann Xi-function in terms of the quotient of 2 functional determinants since the expected value of the Hamiltonian is positive mitian ,with If we set

f (x) 0

then all the Energies are positive

n |H|n 0 and HerEn = s(1 s) R+

Riemann Hypothesis must hold.

1 s = 2 +i E

then is it clear that the roots of the functional determinant

det(E H)

are the roots of the function

1 2

+i E

, our claim is that the zeros of (22) and (23)

Xi(s)

lie on the critical line , based on equations

and the fact that the density of states for Our Hamiltonian is just the distributional Riemann-Weil trace formula. In both cases , the functional determinant inside (22) and (23) must be understood in the sense of the Zeta regularization method explained in (13) formula (22) and (23) are not conjectures, if we evaluate the Theta function in the WKB formalism and with

(t)W KB =
1

dp

2 etp tV (x) with

V (x) =
n=0

f (x) x > 0 x0

f 1 (x) =
1 2

2 d 21 dx 2

arg

1 2

+i x

we obtain

(t)W KB =

2 exp(tn )

+ in ) = 0

In the limit

,the smooth part of the Eigenvalue staircase is given by

N (E) log(x)

E 2

log

E 2e

e=
n=0

1 n! , if we use the expression for the logarithm

x 1

as

following holds

0 0

and apply the half derivative expression, then the

fsmooth (x) 4 e

2 2

x + B A( )

1 fsmooth (x)

4 2 e2

/2

A( )x

/2

B (24)

A(

) 3 and B = 2 = 2 , the second (1+ 2 ) asymptotic of f (x) as x , for this potential ,

)=

( 3+ 2

expression inside (24) is the the energies are

W (x) =

(n)n1 n x n! n=1

1 smooth En = f (n) = Nsmooth (E)

4 2 n2 W 2 (ne1 ) (25)

We can also test our formula with the potentials point

1 2 f 1 (x) = 2 d n(x) 1

xn n = 1, 2,

dx 2

(and compare it with (24) )

( an innite potential well is assumed at the

x=0

) these are the cases of the linear potential (bouncer) , Harmonic

oscillator and innite potential well

f 1 (x) = f 1 (x) = x n
1

2 x

f (x) = n

(x)2 4

N (E) =

E 2
1 m 1 m

(26) +1 1 1 E m+2 +3 2 (27) (28)

f (x) = xn x k

1 N (E) = . 4 N (E) = 2E 3/2 3k

f 1 (x) =

f (x) = kx

From expression (23) we can also compute the density of states of our Hamilto-

(p)+(p+) we will also use the Shokhot2p 1 1 sky's formula for the delta function = (x) 0 m x+i
nian with

p=

and

E 2 =

1 d 2 dE arg 1
Here

1 2 1 2p

+i +i E =

E 2 =

p2 2 = (29) + i p)

1 2 1 4

+ ip

ip 2 lim

p 1 1 1 ln 1 1 2 ip 2p 2p + 4 + i 2 4p + i i (p 2 )+ (p+ 2 ) 1 1 d = (E) = 2p dp argdet (H 4p + 2p

2 2xi+2i

= x

i 2

, this factor comes from the loga-

rithmic derivative of

1 2 + ip equation (29) is a distributional version of the Riemann-Weil trace formula ,

s(s 1)

along the critical line

s=

also if we take formally the logarithm of the Euler product for the Riemann

(n) ip ln n e =reg 1 + ip 2 n 1 using the two test functions h(x) and g(x) g(x) = 0 dr cos(rx)h(r) we re (n) g(ln n) cover the oscillatory part of the Riemann-Weil trace formula 2 n n=1 .
Zeta function on the critical line yields to

n=1

Unlike the model of Wu and Sprung, we have considered also the oscillatory part of the Riemann Eigenvalue Staircase semiclassical law , Wu and Sprung [17] have only considered the smooth part of the Eigenvalue staircase in the limit

arg

1 2

+i E

, which satisfy Bolte's

T T 2 ln 2e N (T ) in order to get a Hamiltonian whose Energies are the positive imaginary part

T >> 1

of the Riemann Zeros, their starting point is the Harmonic oscillator [15] but unlike the normal quantum mechanical oscillator ,whose functional deter-

s 1+ n n=1 the product taken ONLY over the positive imaginary part of the zeros (even if it converges) 1 + s has no meaning n n=0 also the Wu-Sprung model doesn't obey Weyl's law in one dimension mainly Nsmooth (E) = O E d/2 in our case , the Hamiltonian (8) with the Smooth part of the Eigenvalue stair E E case N (E) , satises a Weyl's law with d = 1 + 2 log 2e 2
minant gives the Gamma function

2 (s)

and the spectral determinant (quotient)

(E) (0)

=
n=0

E En 1 2

portional to the Riemann xi function on the critical line

2 E n = n +i E

is pro-

By analogy with the zeros of the Selberg Zeta function, is better to consider the case with the Energies

2 En = n

in this case the Trace of the Resolvent of the Hamiltonian Riemann-Weil trace for the Riemann zeros.

(E + i H)

is the

o Analytic expressions for the potential from Riemann-Weil trace formula :

From the expression for the fractional derivative of powers , we can obtain for the inverse function

dk x dxk

(+1) k (k+1) x

2 d2 f 1 (x) = arg dx 1 2

1 +i x 2

=2
>0

H(x 2 ) x 2

(30)

Using the Riemann-Weil formula we can rewrite (28) as

4 1 (x) = + 2 4x + 1 g(u = ln n, x) =

dr x r2

1 ir + 4 2

ln

J 0 ( u x) x cos(ut) 1 dt = , here the integral can be 0 2 xt2 expressed in terms of the zeroeth order Bessel function.
Here

(n) J0 x ln n n n=1 (31

o Numerical evaluation of the functional determinant :

We need to evaluate the half-derivative inside the inverse of the potential

1 d2 1 dx 2

arg

1 2

+i x

f 1 (x) =

, to do so we can use the Grunwald-Letnikov formula and

[13] with an step

= 0.01
1

q=

1 2

q g(x)
q

d 2 g(x) dx
1 2

1
q

(1)n
n=0 1 2

(q + 1) g (x + (q n) ) (32) (n + 1)(q n + 1)
for big `x' , we have used the Riemann-

In order to evaluate Siegel formula [10]

arg k= x

+i x
U (k)

Z(k) =

1 cos ((k) k ln n) + ik ei(k) = 2 +O 2 n n=1 u(k) =

T 2 k 2 T 2e
,

1 k 1/4

k (33)

The functions inside (A.3) are

[x]

is the oor function and

(T ) = arg

1 4

+ iT 2

T 2

ln

ln

1 48T

+ ...

(34)

For the case of the functional determinant of our Hamiltonian operator with the potential

f 1 (x) =

2 d 21 dx 2

arg

1 2

+i x

dened as

L d2 1 + + f (x) y(x, ) = 0 y(0, ) = 0 = y(L, ) dx2 4 = s(1 s),

(35)

we can evaluate the functional determinant of the operator (35) by the GelfandYaglom method [18] , in this case we need to solve the initial value problem

1 d2 dy(0, ) + + f (x) y(x, ) = 0 y(0, ) = 0 =1 dx2 4 dx

(36)

Unfortunately exact solutions can not be found , in the WKB approximation (36) has the solution

(x) = y(x, )
The 2 constants

f (x) +
x

1 4
x

(37) (t)dt
0

1 (x)1/2

C+ exp
0

(t)dt + C exp

are chosen so (37) solves the initial value problem (36)

The Gelfand-Yaglom theorem, [18] tells us that the functional determinant is related to the solution of the initial value problem (36) and the boundary value problem (35) in the form

= s(1 s) det H + 1 s(1 s) y(L, ) (s) 4 = = = L y(L, 0) (0) det H + 1 4 lim 1

(38) s

The main advantage of the Gelfand-Yaglom method , is that we do not need to evaluate any single eigenvalue in order to obtain the functiona determinant

det H +

1 4

, unfortunately this method is only valid for ordinary dieren-

tial equations

TABLE1 :
(8) with

comparison between the Riemann Zeros (square) from the tables of

1

Odlyzko and the Numerical values of the energies for our Hamiltonian operator

2 d 21 arg 1 + i x , to obtain numerically the potential 2 dx 2 we have used formula (34) to evaluate the fractional derivative and the Riemann1 1 Siegel formula (32) to evaluate S(T ) = arg 2 + iT f 1 (x) =

10

n 0 1 2 3 4 5 6 7 8 9

Zeros (square) 199.7897 441.9244 625.5401 925.6684 1084.7142 1412.7149 1674.3400 1877.2289 2304.4896 6363.8591

Energies 198.7886 441.9240 625.5406 925.6683 1084.7139 1412.7146 1674.3398 1877.2287 2304.4893 6363.8589

o Bessel function

J0 (a x) (x) =

the Riemann zeros

and the density of states for our Hamiltonian and 2 x n : n=0

Let us compare the Riemann-Weyl explicit formula and the denition for the inverse of our potencial function

f 1 (x)

x 2 =

1 1 4 x

1 2

+i x
x

1 4

+i

x 2

1 1 1 ln + 1 2 i x 2 x 2 x 2 x i i ( x 2 )+ ( x+ 2 ) x 1 1 + 4 i 2 4 x 2 x 1 ir + 4 2

= (x)

f 1 (x) =

1 4 + 4x + 1 2

dr x r2

From the equation for our potential half-derivative operator we get

1

dx 2

1 f 1 (x) = 2 d 21 N (x) taking again the dx 2 1 dN (x) d2 1 (x) = 2 dx = (x) , this means that 1 f

(39) (n) J0 ln x ln n n n=1 (40)

perhaps the equations (39) and (40) should be related by a fractional operator

d 2

1 dx 2

, for the smooth part this is almost trivial to nd from the denition of

1 the half-integral

x dt f (t) and the fact that 0 xt

d 2 dx 2
1

g(x) =

dg(x) dx , for

the oscillating part of the potential we must recall the identity

cos (x) (1)n 2n1 = x x (2n)! n=0

J0 (x) =
n=0

(1)n (n!)
2

x 2

2n

cos ( ax) J0 a x = 1 x dx 2 (41) d2

1

equation (41) is a bit harder to prove, we can prove (41) by using the Taylor expansion of the 2 functions and then applying the property of the half derivative
1 (n+1) = n+ 1 xn 2 , so we choose the potential 1 ( 2) dx 2 from the WKB formalism and we also have proved that the density of states of

operator for power series

d 2 xn

our Hamiltonian dened in (8) is just the Riemann-Weil explicit formula in a

11

distributional framework so

f 1 (x) 2

d 2 dx 2
1

(x)

APPENDIX A: An implicit equation for the potential f(x) from the Bohr-Sommerfeld quantization conditions
2 d 21 arg 1 + i x 2 dx 2 Bohr-Sommerfeld quantization conditions [7]
The expression

f 1 (x) =
a

can also be obtained from the

E = f (a)
here

2
0

dx

E f (x) = p(x)
C

pdq = 2 n +

1 2

(A.1)

a is the classical turning point where the momentum is p = 0 and n = N (E)

is the Eigenvalue staircase the rst integral inside (A.1) is a line integral taken over the closed orbit of the classical system, equation (A.1) can be understood as an integral equation for the inverse of the potential in the form

1 + n(E) 2
1

a=a(E)

=2
0

E V (x)dx = 2
0

Ex

1 1 + arg 1 + i x in this case this result is completely equivalent 1 2 dx 2 2 to the one we got by Zeta regularization and by the WKB approximation of the 1 Theta function dxetf (x) = W KB (t) . 2 t 0
d2
From equation

If we take the half derivative on both sides of (A.2) we

1 df 1 = Dx 2 f (x) dx (A.2) 1 would get f (x) =

f 1 (x) = 2

d2
1 dx 2

1 2

arg
1 d2

1 2
1

+i x

the density of states

could be evaluated approximately as

1 2

(x)

dx 2

= (x) =
n

2 x n

References

[1] Abramowitz, M. and Stegun, I. A. (Eds.).

Other Sums of Reciprocal Powers." 23.2 in  Handbook of Mathematical Func-

"Riemann

Zeta Function and

tions. New York: Dover, pp. 807-808, 1972.

[2] Apostol Tom  Introduction to Analytic Number theory ED: Springuer , (1976) [3] Berry M. Keating J.P  The Riemann Zeros and Eigenvalue Asymptotics  Siam Review archive Vol. 41 Issue 2, June 1999 pages: 236-266

12

chaos Open Sys. & Information Dyn.6 (1999) 167-226 , avaliable at:http://arxiv.org/abs/chao-dyn/9702003
[5] Cartier P . and Voros A.  Une nouvelle interpretation de la formule des traces de Selberg. In P. Cartier, L. Illusie, N. M. Katz, G. Laumon, Y. Manin, and K. A. Ribet, editors, The Grothendieck Festschrift (vol.II), volume 87 of Progress in Mathematics, pages 1-67. Birkhauser, 1990. [6] Elizalde E. , Odintsov S. ,Romeo A. , Bytsenko A. , Zerbini S Zeta Regularization Techniques With Applications ISBN-10: 9810214413 Ed: Scientic pub (1994) [7] Griths, David J. (2004).  Introduction to Quantum Mechanics Prentice Hall. ISBN 0-13-111892-7. [8] Grosche  Path Integrals , Hyperbolic Spaces and Selberg trace formulae (1996) ISBN-10: 9810224311 [9] Gutzwiller, M.  Chaos in Classical and Quantum Mechanics , (1990) SpringerVerlag, New York ISBN=0-387-97173-4. [10] Hejhal D.  The Selberg Trace formula and the Riemann Zeta function  Duke Mathematical Journal 43 (1976) [11] Hurt N. Quantum Chaos and Mesoscopic Systems: Mathematical Methods World

[4] Bolte J: Semiclassical trace formulae and eigenvalue statistics in quantum

in the Quantum Signatures of Chaos (2010) ISBN-13: 978-9048148110

[12] Marklof J.  Selberg Trace formula: http://arxiv.org/abs/math/0407288 [13] Nishimoto, K.  Fractional Calculus. New Haven, CT: University of New Haven Press, 1989.

An introduction' PDF avaliable at

[14] Odlyzko A.  Tables of Zeros of Riemann Zeta functions ,webpage http://www.dtc.umn.edu/~odlyzko/zeta_ [15] Selberg, Atle (1956),

symmetric Riemannian spaces with applications to Dirichlet series", J. Indian

"Harmonic analysis and discontinuous groups in weakly

Math. Soc. (N.S.)

20:

47-87, MR0088511

[16] Voros A.  Exercises in exact quantization e-print avaliable at http://arxiv.org/pdf/math-ph/0005029v2.pd [17] Wu H. and D. W. L. Sprung Riemann zeros and a fractal potential Phys. Rev. E 48, 2595-2598 (1993) [18] Yaglom, A.M Gelfand I .M Integration in functional spaces and its applications in quantum physics, J. Math. Phys. 1, 48-69 _1960_.

13