Will a physicist prove the Riemann Hypothesis?

Marek Wolf
Cardinal Stefan Wyszynski University, Faculty of Mathematics and Natural Sciences.
ul. Wóycickiego 1/3, PL-01-938 Warsaw, Poland, e-mail: [email protected]

Abstract
arXiv:1410.1214v3 [math-ph] 1 Dec 2015

In the first part we present the number theoretical properties of the Riemann
zeta function and formulate the Riemann Hypothesis. In the second part we review
some physical problems related to this hypothesis: the Polya–Hilbert conjecture,
the links with Random Matrix Theory, relation with the Lee–Yang theorem on the
zeros of the partition function, random walks, billiards etc.

MSC: 11M01, 11M26, 11M50, 81-02, 82B41

“... upon looking at prime numbers one has the feeling of being
in the presence of one of the inexplicable secrets of creation.”
Don Zagier in [1, p.8, left column]

1 Introduction
There are many links between mathematics and physics. Many branches of mathematics
arose from the need to formalize and clarify the calculations carried out by physicists,
e.g. Hilbert spaces, distribution theory, differential geometry etc. In this article we are
going to describe the opposite situation when the famous open mathematical problem
can be perhaps solved by physical methods. We mean the Riemann Hypothesis (RH), the
over 160 years old problem which solution is of central importance in many branches of
mathematics: there are probably thousands of theorems beginning with: “Assume that
the Riemann Hypothesis is true, then ...”. The RH appeared on the Hilbert’s famous list
of problems for the XX centaury as the first part of the eighth problem [2] (second part
concerned the Goldbach’s conjecture; recently H. A. Helfgott [3] have solved so called
ternary case of Goldbach conjecture). In the year 2000 RH appeared on the list of the
Clay Mathematics Institute problems for the third millennium, this time with 1,000,000
US dollars reward for the solution.
After the announcement of the prize by Clay Mathematics Institute for solving RH
there has been a rash of popular books devoted to this problem: [4], [5], [6], [7], [8].
The classical monographs on RH are: [9], [10], [11], [12], while [13] is a collection
of original papers devoted to RH preceded by extensive introduction to the subject.
We also strongly recommend the web site Number Theory and Physics at the address
http://empslocal.ex.ac.uk/people/staff/mrwatkin/zeta/physics.htm containing a lot in-
formation about links between number theory (in particular about RH) and physics.
In 2011 there appeared the paper “Physics of the Riemann hypothesis” written by
D. Schumayer and D. A.W. Hutchinson [14]. Here we aim to provide complementary
description of the problem which can serve as a starting point for the interested reader.
This review consists of seven Sections and the Concluding Remarks. In the next
Section we present the historical path leading to the formulation of the RH. Next we
briefly discuss possible ways of proving the RH. Next two Sections concern connections
between RH and quantum mechanics and statistical mechanics. In Sect.6 a few other links
between physical problems and RH are presented. In the last Section fractal structure
of the Riemann ζ(s) function is discussed. Because we intend this article to be a guide
we enclose rather exhaustive bibliography containing over 100 references, a lot of these
papers can be downloaded freely from author’s web pages. Let us begin the story.

2 The short history of the Prime Number Theorem

There are infinitely many prime numbers 2, 3, 5, 7, 11, . . . , pn , . . . and the first proof of
this fact was given by Euclid in his Elements around 330 years b.C. His proof was by
contradiction: Assume there is a finite set of primes P = {2, 3, 5, . . . , pn }. Form the
number 2×3×5 . . .×pn +1, then this number divided by primes from P gives the remainder
1, thus it has to be a new prime or it has to factorize into primes not contained in the
set P, hence there must be more primes than n. For example if P = {2, 3, 5, 7, 11, 13},
then 2 · 3 · 5 · 7 · 11 · 13 + 1 = 59 · 509 and 59, 509 ∈/ P. The first direct proof of infinity of
primes was presented by L. Euler around 1740 who has shown that the harmonic sum of
prime numbers pn diverges:
X 1
∼ log log(x).
pn <x
pn
P
Next the problem of determining the function π(x) = n Θ(x−pn ) (Θ(x) is the Heaviside
step function), giving the number of primes up to a given threshold x, has arisen. It is one
of the greatest surprises in the whole mathematics that such an erratic function as π(x)
can be approximated by a simple expression. Namely Carl Friedrich Gauss as a teenager
(different sources put his age between fifteen years and seventeen years) has made in
the end of eighteen centaury the conjecture that π(x) is very well approximated by the
logarithmic integral Li(x):
Z x
du
π(x) ∼ Li(x) := . (1)
2 log(u)

The symbol f (x) ∼ g(x) means here that limx→∞ f (x)/g(x) = 1. Integration by parts
gives the asymptotic expansion which should be cut at the term n0 = blog(x)c, after
which terms begin to increase:
x x 2!x 3!x
Li(x) = + + + + ··· . (2)
log(x) log (x) log (x) log4 (x)
2 3

2
There is a series giving Li(x) for all x > 2 and quickly convergent which has n! in denom-
inator and logn (x) in nominator instead of opposite order in (2) (see [15, Sect. 5.1])
∞
X logn (x)
Li(x) = γ + log log(x) + for x > 1 , (3)
n=1
n · n!
Pn 1

where γ ≈ 0.577216 is the Euler-Mascheroni constant γ = limn→∞ k=1 k − log(n) .
The way of proving (1) was outlined by Bernhard Riemann in a seminal 8-pages long
paper published in 1859 [16]. English translation is available at http://www.maths.tcd.ie/pub/HistMath/
or at CMI web page; it was also included as an appendix in [10]. The handwritten by
Riemann manuscript was saved by his wife and is kept in the Manuscript Department
of the Niedersöchsische Staats und Universitätsbibliothek Göttingen. The scanned pages
are available at http://www.claymath.org/sites/default/files/riemann1859.pdf. In fact in
this paper Riemann has given an exact formula for π(x). The starting point of the Rie-
mann’s reasoning was the mysterious formula discovered by Euler linking the sum of n1ks
with the product over all primes p:
∞ ∞
X 1 Y 1
ζ(s) := s
=  , s = σ + it, <[s] = σ > 1. (4)
n=1
n p=2 1 −
1
ps

 
1
To see that this equality really holds one needs first to recognize in the terms 1/ 1 − ps
the sums of the geometric series ∞ 1
P
k=0 ps . The geometrical series converges absolutely
so the interchange of summation and the product is justified. Finally the fundamental
theorem of arithmetic stating the each positive integer n > 1 can be represented in exactly
one way (up to the order of the factors) as a product of prime powers:
k
Y
n= pα1 1 pα2 2 · · · pαk k = pαi i . (5)
i=1

has to be invoking to obtain the series on lhs of (4). Let us notice that on the rhs
(4) the product cannot start from p = 1 and it explains why the first prime is 2 and
not 1 — physicists often think that 1 is a prime number (before 19–th century P∞ 1 1was
indeed Pconsidered to be a prime). Euler was the first who calculated the sums n=1 n2 =
∞ 1
π 2 /6, 4
n=1 n4 = π /90 and in general ζ(2n). In fact Euler has considered the above
formula only for real exponents s = x and it was Riemann who considered it as a function
of complex argument s = σ + it and thus the function ζ(s) is called the Riemann’s zeta
function. In the context of RH instead of z = x + iy for the complex variable the notation
s = σ + it is traditionally used. The formula (4) is valid only for <[s] = σ > 1 and it
follows from the product of non-zero terms on r.h.s. of (4) that ζ(s) 6= 0 on the right
of the line <[s] = 1. Riemann has generalized ζ(s) to the whole complex plane without
s = 1 where zeta is divergent as an usual harmonic series — the fact established in 14th
century by Nicole Oresme. Riemann did it by analytical continuation (for the proof see
the original Riemann’s paper or e.g. [10, Sect. 1.4]:

Γ(−s) +∞ (−x)s dx
Z
ζ(s) = , (6)
2πi +∞ ex − 1 x

3
 R +∞
where +∞
denotes the integral over the contour
6


--


Appearing in (6) the gamma function Γ(z) has many representations, we present the
Weierstrass product:
∞
e−γz Y  z −1 z/k
Γ(z) = 1+ e . (7)
z k=1 k
From (7) it is seen that Γ(z) is defined for all complex numbers z, except z = −n for
integers n > 0, where are the simple poles
R ∞ of Γ(z). The most popular definition of gamma
−t z−1
function given by the integral Γ(z) = 0 e t dt is valid only for <[z] > 0.
The integral (6) is well defined on the whole complex plane without s = 1, where ζ(s)
has the simple pole, and is equal to (4) on the right of the line <[s] = 1. Recently many
representations of the ζ(s) are known, for review of the integral representations see [17].
The exact formula for π(x) obtained by Riemann involves the function J(x) defined
as
1 1
J(x) = π(x) + π(x1/2 ) + π(x1/3 ) + · · · . (8)
2 3
In words J(x) increases by 1 at each prime number p, by 1/2 at each x being the square of
the prime, in general it jumps by 1/n at argument equal to n-th power of some prime pn .
J(x) is discontinuous at x = pn thus at such arguments is defined as a halfway between
its old value and its new value. Let us mention that J(x) = 0 for 0 ≤ x < 2. Then π(x)
is given (via so called Möbius inversion formula) by the series:
X µ(n)
π(x) = J(x1/n ), (9)
n≥1
n

where the sum is in fact finite because it stops at such N that x1/N > 2 > x1/(N +1) and
µ(n) is the Möbius function:

 1 for n = 1
µ(n) = 0 when n is divisible by the square of some prime p: p2 |n (10)
r
(−1) when n = p1 p2 . . . pr


For example µ(14) = 1, µ(25) = 0, µ(30) = −1. This definition of µ(n) stems from the
formula (4) and the Dirichlet series for the reciprocal of the zeta function:
∞   X∞
1 Y 1 µ(n)
= 1− s = . (11)
ζ(s) p=2 p n=1
ns

The above product over p produces integers n which in the factorization does not contain
square of a prime and those n which factorizes into odd number of primes contribute with
sign −1 while those n which factorizes into even number of primes contribute with sign
+1. We can notice at this point that the Möbius function has the physical interpretation:
namely in [18] it was shown that µ(n) can be interpreted as the operator (−1)F giving the

4
number of fermions in quantum field theory. In this approach the equality µ(n) = 0 for n
divisible by a square of some prime is the manifestation of the Pauli exclusion principle.
The crucial point of the Riemann’s reasoning was the alternative formula for J(x) not
involving primes at all: X
J(x) = Li(x) − Li(xρ ), (12)
ρ

where the sum runs over all zeros of ζ(s), i.e. over such ρ that ζ(ρ) = 0. Let us stress
that the above (12) is an equality, which is remarkable because the left hand side is a
step function, thus somehow at prime powers all of the zeros of zeta cooperate to deform
smooth plot of the first term Li(x) into the stair–like graph with jumps. Then the number
of primes up to x is obtained by combining (9) and (12)
N
!
X µ(n) X
π(x) = Li(x1/n ) − Li(xρ/n ) . (13)
n=1
n ρ

To be precise at arguments x equal to prime numbers, when π(x) is not continuous and
jumps by 1, one has to define lhs as lim→∞ {π(x − ) + π(x + )} (the same procedure
was mentioned above for the function J(x)). There is an ambiguity when using defini-
tion of logarithmic integral (1) for Li(xρ ) connected with multivaluedness of logarithm of
complex argument, in particular for complex numbers z1 , z2 the equality log(z1 z2 ) =
log(z1 ) + log(z2 ) does not hold (here are calculations providing the counterexample:
(−z)2 = z 2 , log((−z)2 ) = log(z 2 ), log(−z) + log(−z) = log(z) + log(z), 2 log(−z) =
2 log(z) ⇒ log(−z) = log(z); in particular log(−1) = log(1) = 0 what is not true as
log(−1) = i(2k + 1)π 6= 0). Hence the above logarithmic integral for complex argument
is defined as Li(xρ ) = Li(eρ log(x) ), where for z = u + iv, v 6= 0:
Z u+iv
z ew
Li(e ) = dw, (14)
−∞+iv w
thus Li is in fact defined via the exponential integral. Let us mention, that in Mathematica
to obtain the value of Li(xρk ) the command ExpIntegralEi[ZetaZero[k]*Log[x]] has
to be used.
In equations (12) and (13) we meet the issue of determining zeros ρ of the zeta function:
ζ(ρ) = 0. Riemann has shown that ζ(s) fulfills the functional identity:
 
− 2s
s
− 1−s 1−s
π Γ ζ(s) = π 2 Γ ζ(1 − s), for s ∈ C \ {0, 1}. (15)
2 2
The above form of the functional equation is explicitly symmetrical with respect to the
line <(s) = 1/2: the change s → 12
+ s on both sides of (15) shows that the values of the
s
combination of functions π − 2 Γ 2s ζ(s) are the same at points 12 + s and 12 − s.
Because Γ(z) is singular at all negative integers, thus to fulfill functional identity (15)
ζ(s) has to be zero at all negative even integers:
ζ(−2n) = 0, n = 1, 2, 3, . . .
These zeros are called trivial zeros. The fact that ζ(s) 6= 0 for <(s) > 1 and the shape
of the functional identity entails that nontrivial zeros ρn = βn + iγn are located in the
critical strip:
0 ≤ <[ρn ] = βn ≤ 1.

5
From the symmetry of the functional equation (15) with respect to the line <[s] = 12 it
follows, that if ρn = βn + iγn is a zero, then ρn = βn − iγn and 1 − ρn , 1 − ρn are also zeros:
they are located symmetrically around the straight line <[s] = 21 and the axis t = 0, see
Fig. 1.

Figure 1: The location of zeros of the Riemann ζ(s) function.

The sum over trivial zeros ρ = −2n in (12) can be calculated analytically giving the
explicit (i.e. expressed directly by sum over zeros of ζ(s)) formula for J(x):
Z ∞
X
ρ ρ

du
J(x) = Li(x) − Li(x ) + Li(x ) + 2
− log(2) (16)
x u(u − 1) log(t)
<(ρ)>0
0<=(ρ)<1

and therefore explicit formula for π(x) follows:

 
N Z ∞
X µ(n) 
Li(x n1 ) −
X  ρ ρ
 1 
π(x) = Li(x n ) + Li(x n ) + du − log(2).
n=1
n  x1/n u (u2 − 1) log(u) 
<(ρ)>0
0<=(ρ)<1
(17)
Extending in the first term summation to infinity (it is not an big sin, as terms with large
n tend to zero) gives the function
∞
X µ(n)
R(x) = Li(x1/n ). (18)
n=1
n

6
Figure 2: The plot of π(x) and the r.h.s. of (17) obtained by summing over first
50 nontrivial zeros (blue line) and first 500 nontrivial zeros (red line). We have pro-
duced data for this plots using the equation (22). The animation showing the ef-
fect of adding consecutive zeros one by one to the formula (17) can be seen at
http://empslocal.ex.ac.uk/people/staff/mrwatkin/zeta/encoding1.htm

The function R(x) can be calculated from very quickly converging series
∞
X (log(x))n
R(x) = 1 + . (19)
n=1
nn!ζ(n + 1)

The last sum is called the Gram formula, see [19, p.51] for transformations leading from
(18) to (19). Because the sum over all complex zeros is not absolutely convergent its
value depends on the order of summation. In fact famous (and curious) Riemann’s re-
arrangement theorem, see e.g. [20, Theorem 3.54], asserts that terms of a conditionally
convergent infinite series can be permuted such a way that the new series converges to
any given value ! For (17) Riemann in [16] says that “It may easily be shown, by means
of a more thorough discussion” that the natural order, i.e. the process of pairing together
zeros ρ and ρ in order of increasing imaginary parts of ρ, is the correct one. At the end
of [16] Riemann states about the series in (17) that “when reordered it can take on any
arbitrary real value”.
Again let us point out the curiosity (mystery) of the above equation (17): π(x) on lhs
jumps by 1 at each argument being a prime with constant values (horizontal sections)

7
between two consecutive primes. Thus on the rhs the zeros of zeta have to conspire
to deform smooth plot of the first term Li(x) into the stair–like graph with jumps. It
resembles the Fourier series of smooth sinuses reproducing say the step function on interval
(−π, π). In Fig. 2 we have made plots illustrating these observations.
The formula (17) is less time consuming to obtain π(x) for large x than counting
all primes up to x; the best non–analytical methods of computing π(x) have complex-
ity O(x2/3 / log2 (x)), while involving some variants of the Riemann explicite formula are
O(x1/2 ) in time. For example, the value π(1024 ) = 18, 435, 599, 767, 349, 200, 867, 866 was
obtained by a variant of (17) using 59, 778, 732, 700 nontrivial zeros of ζ(s) [21]. Also
the value π(1025 ) = 176, 846, 309, 399, 143, 769, 411, 680 was announced, see The On-Line
Encyclopedia of Integer Sequence, entry A006880.
Amazingly the horrible looking sum of the integrals in (17) stemming from the trivial
zeros can be brought to the simple closed form:
N Z ∞  N
X µ(n) 1 1 X 1 π
2
du − log(2) = µ(n)+ arctan +(x, N ),
n x 1/n u (u − 1) log(u) 2 log x π log x
n=1 n=1
(20)
PN (x, N ) → 0 as N → ∞, for details see [22]. The special choice of N such that
where
k=1 µ(k) = −2 (e.g. N = 5, 7, 8, 9, 11, 12, . . .) is favoured: the series for arcus tangens
in the vicinity of u = 0 has the form arctan(u) = u − u3 /3 + u5 /5 − u7 /7 + . . . and for
such a special N the first two terms in (20) behave together like (π/ log(x))3 /3 + . . . thus
the contribution from trivial zeros is negligible for large x and hence nontrivial zeros are
prevailing.
So where are the complex zeros of zeta? Riemann has made the assumption, now called
the Riemann Hypothesis, that all nontrivial zeros lie on the critical line <[s] = 21 :

1 1
ρn = + iγn (i.e. βn = for all n). (21)
2 2
Contemporary the above requirement is augmented by the demand that all nontrivial
zeros are simple. Despite many efforts the Riemann Hypothesis remains unproved. In
Fig.1 we illustrate the Riemann Hypothesis and in the Table 1 we give the approximate
values of the first 10 non-trivial zeros of ζ(s). We see that initially γn > n, but at n = 9137
the inequality reverses as γ9137 = 9136.6792 . . . while γ9136 = 9136.1396 . . .. Next zero is
again larger than its index: γ9138 = 9138.10591 . . . but γ9141 = 9140.5783 . . . and up to
n = 10, 000, 000 the inequality γn < n holds and we believe it will be fulfilled for ever.
Incidentally 9137 is a prime, in fact it is a left–truncatable prime: removing successively
left digit gives again prime numbers 9137, 137, 37 and 7.

Table 1
1 1
n 2
+ iγn n 2
+ iγn
1 1
1 2
+ i14.134725142 . . . 6 2
+ i37.586178159 . . .
1 1
2 2
+ i21.022039639 . . . 7 2
+ i40.918719012 . . .
1 1
3 2
+ i25.010857580 . . . 8 2
+ i43.327073281 . . .
1 1
4 2
+ i30.424876126 . . . 9 2
+ i48.005150881 . . .
1 1
5 2
+ i32.935061588 . . . 10 2
+ i49.773832478 . . .

8
 Assuming the RH, i.e. collecting together terms ρn = 12 +iγn and ρn = 1−ρn = 12 −iγn ,
using the Euler identity eiα = cos(α) + i sin(α) we can represent π(x) as a main smooth
trend plus superposition of waves sin(·) and cos(·):
∞
X (log(x))n
π(x) = 1 + −
n=1
nn!ζ(n + 1)
   
( γi log(x) γi log(x)
X 1
x 2n X cos n
+ 2γ i sin n
µ(n) 1 2 (22)
n
log(x) i 4
+ γi
   
1 2 γi log(x) γi log(x) )
n ( 4
− γi )2 cos n
+ 2γ i sin n
+ 1 1 2 4
,
log(x) 16
+ 2 γi + γi

where we have used two first terms of the expansion Li(x) ≈ x/ log(x) + x/ log2 (x).
Using the above equation with 10000 zeros and second sum over n up to 7 we obtained
25.00267 for π(100), while the numbers of primes up to 100 (without counting 1 as a
prime) is 25. Physicists well know that derivative of the step function is the Dirac delta
function: Θ0 (x) = δ(x), thus the P derivative of π(x) with respect to x is the sum of Dirac
deltas concentrated on primes: pn δ(x − pn ). We have differentiated two first sums
in (22), i.e. skipping terms O(1/γk4 ), summed over first 15000 nontrivial zeros of zeta
and the resulting plot is presented in Fig.3. The animated plot of the delta–like spikes
emerging with increasing number of nontrivial zeros taken into account is available at
http://empslocal.ex.ac.uk/people/staff/mrwatkin/zeta/pianim.htm.
In 1896 J. S. Hadamard (1865 – 1963) and Ch. J. de la Valle Poussin (1866 – 1962)
independently proved that ζ(s) does not have zeros on the line 1 + it, thus |xρ | < x. It
suffices to obtain from (17) the original Gauss’s guess (1), which thus became a theorem
called the Prime Number Theorem (PNT). Indeed: for large x in (17) the first term R(x)
wins over terms with Li(xρ ) and then from (18) we have that R(x) ≈ Li(x).
Already Riemann calculated numerically a few first nontrivial zeros of ζ(s) [10]. Next
in 1903 J.P. Gram [23] calculated that first 15 zeros of ζ(s) are on the critical line; in June
of 1950 A.Turing has used the Mark 1 Electronic Computer at Manchester University to
check that first 1104 zeros are on the critical line. He has done this calculations “in an
optimistic hope that a zero would be found off critical line”, see [24, p. 99]. A few years
ago S. Wedeniwski (2005) was leading the internet project Zetagrid [25] which during four
years determined that 250 × 1012 zeros are on the critical line, i.e. on s = 21 + it up to
t < 29, 538, 618, 432.236. The present record belongs to K. Gourdon(2004) [26]: the first
1013 zeros are on the critical line. A. Odlyzko checked that RH is true in different intervals
around 1020 [27], 1021 [28], 1022 [29], but his aim was not to verify the RH but rather
providing evidence for conjectures that relate nontrivial zeros of ζ(s) to eigenvalues of
random matrices, see Sect. 4. In fact Odlyzko has expressed the view that the hypothetical
10000
zeros off the critical line are unlikely to be encountered for t below 1010 , see [4, p.358].
P Let N (t) denote the function counting the nontrivial zeros up to T , i.e. N (T ) =
n Θ(T − =[ρn ]). In his seminal paper Riemann announced and in 1905 von Mangoldt
proved that:  
T T 7
N (T ) = log + + O(log(T )). (23)
2π 2πe 8
The Fig.4 illustrates how well the above formula predicts N (T ). In 1904 G.H. Hardy

9
Figure 3: The plot of the derivative of two sums on rhs of (22) obtained by summing over first
15000 nontrivial zeros showing delta–like pattern. In the inset fluctuations for 11.5 < x < 12.5
are shown — the high spikes at x = 11 and 13 squeezed them on the original plot.

proved, by considering moments of certain functions related to the zeta function, that
on the critical line there is infinity of zeros of ζ(s) [30]. Levinson (1974) proved that
more than one-third of zeros of Riemanns ζ(s) are on critical line by relating the zeros of
the zeta function to those of its derivative, and Conrey (1989) improved this further to
two-fifths (precisley 40.77 % have <(ρ) = 12 ). The present record seems to belong to S.
Feng, who proved that at least 41.28% of the zeros of the Riemann zeta function are on
the critical line [31].
At the end of this Section we mention, that ζ(s) admits besides the product (4) another
product representation, called the Hadamard product:
∞  
π −1− γ2
Y s s
ζ(s) = s

e e ρk
1− . (24)
(s − 1)Γ 2
−1 k=1
ρk

In contrast to (4) it is valid on the whole complex plane without s = 1. It is an example of

the general Weierstrass factorization theorem: points where function vanishes determine
this function. We also add that the common belief is that the imaginary parts γl of the
nontrivial zeros of ζ(s) are irrational and perhaps even transcendental [32], [33].

10
Figure 4: The plot illustrating the formula (23) for number of nontrivial zeros up to T = 5×106 .
When on vertical axis the interval (0,1) is 4 cm long, then if the horizontal axis would be
plotted on linear scale instead of logarithmic, its length should amount to 200 km. This analogy
in a striking way shows how precise is the formula (23) and that the term 7/8 is important
(mathematicians will not agree with this and skip this constant term).

3 How to prove the Riemann Hypothesis?

Practically nobody is going to prove the RH directly: there are probably well over one
hundred of different facts either equivalent to RH or of whose truth RH will follow (i.e.
sufficient conditions). Hence proving one of these so called criteria for RH will entail the
validity of RH. Below we present a few such criteria for RH.
In 1901 H. von Koch proved [34] that the Riemann Hypothesis is equivalent to the
following error term for the approximation of the prime counting function by logarithmic
integral: √
π(x) = Li(x) + O( x log(x)). (25)
Later the error term was specified explicitly by Schoenfeld [35, Corollary 1] and RH is
equivalent to
1√
|π(x) − Li(x)| ≤ x log(x) for all x ≥ 2657. (26)
8π
The following facts show that the validity of the RH is very delicate and subtle: namely
in some sense RH is valid with accuracy  = 1.14541 × 10−11 (or less, that is the present

11
value of ). Here is the reasoning leading to this conclusion: Let us introduce the function
     
1 2 1 − z2 − 14 z 1 1
ξ(iz) = z − π Γ + ζ z+ . (27)
2 4 2 4 2

We can see from the above formula that: RH is true ⇔ all zeros of ξ(iz) are real. The
point is that ξ(z) can be expressed as the following Fourier transform (for derivation of
this formula see e.g. [9, Sect.10.1]):
Z ∞
1 z 
ξ = Φ(t) cos(zt)dt, (28)
8 2 0

where ∞
2 e4t
X
Φ(t) = (2π 2 n4 e9t − 3πn2 e5t )e−πn . (29)
n=1

The functin ξ(z) can be generalized to the family of functions H(z, λ) parameterized by
λ: Z ∞
2
H(z, λ) = Φ(t)eλt cos(zt)dt. (30)
0

Thus we have H(z, 0) = 81 ξ( 12 z). N. G. De Bruijn proved in 1950 [36] that H(z, λ) has
only real zeros for λ ≥ 21 and if H(z, λ) has only real zeros for some λ0 , then H(z, λ)
has only real zeros for each λ > λ0 . In 1976 Ch. Newman [37] has proved that there
exists parameter λ1 such that H(z, λ1 ) has at least one non-real zero. Thus, there exists
such constant Λ in the interval −∞ < Λ < 21 that H(z, λ) has real zeros ⇔ λ > Λ.
The Riemann Hypothesis is equivalent to Λ ≤ 0. This constant Λ is now called the de
Bruijn–Newman constant. Newman believes that Λ ≥ 0. The computer determination
has provided the numerical estimations of values of de Bruijn–Newman constant; the
current record belongs to Y. Saouter et. al. [38]: Λ > −1.14541 × 10−11 . Because the gap
in which Λ catching the RH is so squeezed, Odlyzko noted in [39], that “...the Riemann
Hypothesis, if true, is just barely true”.
There are also criteria for RH involving integrals. V. V. Volchkov has proved [40] that
following equality is equivalent to RH:
Z ∞ Z ∞
1 − 12t2 3−γ
2 3
log(|ζ(σ + it)|dσdt = π . (31)
0 (1 + 4t ) 12 32

In the paper [41] the above integral was used to express the RH in terms of the Veneziano
amplitude for strings as well as to find some generalizations of the Volchkov’s criterion.
In the paper [42] the equality to zero of the following integral was shown to be equiv-
alent to RH: Z ∞
log(|ζ( 21 + it)|)
Z
log(|ζ(s)|)
|ds| = 1 dt = 0. (32)
<(s)= 21 |s|2 −∞ 4
+ t2
Finally let us mention the elementary Lagarias criterion [44]: the Riemann Hypothesis
is equivalent to the inequalities:
X
σ(n) ≡ d ≤ Hn + eHn log(Hn ) (33)
d|n

12
Figure 5: The plot of σ(n) for 1 < n < 106 . In red are plotted values of σ(n) which approach
the threshold (green line) values closer than 10%. The lower “support” of the graph comes from
n being primes, as prime p is divisible only by 1 and itself, so σ(p) = p + 1. Data for this plot
was obtained with the free package PARI/GP [43].

for each n = 1,P

2, . . ., where σ(n) is the sum of all divisors of n and Hn is the n-th harmonic
number Hn = nj=1 1j . To disprove the RH it suffices to find one n violating the inequality
(33). The Lagarias criterion is not well suited for computer verification (it is not an easy
task to calculate Hn for n ∼ 10100000 with sufficient accuracy) and in [45] K. Briggs has
undertaken instead the verification of the Robin [46] criterion for RH:
X
RH ⇔ d < eγ n log log(n) for n > 5040. (34)
d|n

For some n Briggs obtained for the difference between r.h.s. and l.h.s. of the above
inequality value as small as e−13 ≈ 2.2 × 10−6 , hence again RH is in a danger to be
violated. As A. Ivic has put it “The Riemann Hypothesis is a very delicate mechanism.’,
quoted in [6, p. 123].
Let us notice that the belief in the validity of RH is not common: famous mathemati-
cians J. E. Littlewood, P. Turan and A.M. Turing have believed that the RH is not true,

13
see the paper “On some reasons for doubting the Riemann hypothesis” [47] (reprinted in
[13, p.137]) written by A. Ivić, one of the present day leading expert on RH. When J.
Derbyshire asked A. Odlyzko about his opinion on the validity of RH he replied “Either
it’s true, or else it isn’t” [4, p. 357–358].

4 Quantum Mechanics and RH

The first physical method of proving RH was proposed by George Polya around 1914
during the conversation with Edmund Landau and now is known as the Hilbert-Polya
Conjecture. Landau asked Polya: “Do you know a physical reason that the Riemann
hypothesis should be true?” and his reply was: “This would be the case, I answered, if
the nontrivial zeros of the Ξ-function were so connected with the physical problem that
the Riemann hypothesis would be equivalent to the fact that all the eigenvalues of the
physical problem are real”, 1 see the whole story at the web site [48]. Let us stress that this
talk took place many years before the birth of quantum mechanics and the Schroedinger
equation for energy levels. However in the period 1911–1914 Hermann Weyl published a
few papers on the asymptotic distribution of eigenvalues of the Laplacian in the compact
domain (in particular the eigenfrequencies or natural vibrations of the drums), see e.g.
[49, 50]. Thus presumably Polya was inspired by the Weyl’s papers. If the RH is true
nontrivial zeros lie on critical line and it makes sense to order them according to the
imaginary part and eventually put them into the 1-1 correspondence with the eigenvalues
of some hermitian operator. Therefore the problem is to find a connection between energy
levels En of some quantum system and zeros of ζ(s).
In the autumn of 1971 [8, p.261] H. Montgomery, assuming the RH, has proved theorem
about statistical properties of the spacings between zeta zeros. The formulation of this
theorem is rather complicated and we will not present it here, see his paper [51]. Next
Montgomery made the conjecture that correlation function of the imaginary parts of
nontrivial zeros has the form (here 0 < α < β < ∞ are fixed):
Z β  2 !
X sin πu
1→ 1− du as T → ∞. (35)
0 α πu
0<γ,γ ≤T
2πα 2πβ
log T
≤γ−γ 0 ≤ log T

In the Fig.6 we present a sketchy plot of the both sides of the equation (35). This
result says that the zeros – unlike primes, where it is conjectured that there is infinity
of twins primes, i.e. primes separated by 2, like (3, 5), (5, 7), (11, 13), . . . , (59, 61), . . .)—
would actually repel one another because in the integrand sin(πu)/πu → 1 for u → 0.
Montgomery published this result in 1973 [51], but earlier in 1972 he spoke about it with
F. Dyson in Princeton, see many accounts of this story in the popular books listed in the
Introduction, e.g. [8, p.133–134]. Dyson recognized in (35) the same dependence as in the
behavior of the differences between pairs of eigenvalues of random Hermitian matrices.
The random matrices were introduced into the physics by Eugene Wigner in the fifties of
twenty century to model the energy levels in the nuclei of heavy atoms. Spectra of light
atoms are regular and simple in contrast to the spectra of heavy atomic nuclei, like e.g.
1
appearing here the function Ξ is equal to the lhs of (15) multiplied by s(s − 1)/2, hence it has the
same zeros as ζ(s)

14
238
U, for which hundreds of spectral lines were measured. The hamiltonians of these nuclei
are not known, besides that such many body systems are too complicated for analytical
treatment. Hence the idea to model heavy nuclei by the matrix with random entries
chosen according to the gaussian ensemble and subjected to some symmetry condition
(hermiticity etc.).

Figure 6: The plot of lhs of (35) calculated from

Figure 7: Histogram of normalized gaps be-
the first 107 zeros compared with the prediction
tween cinsecutive zeros of zeta obtained for
of Montgomery. Points representing correlation
5000000 zeros. Points are spaced with ∆s =
function were calculated from equation on lhs of
0.05.
(35) for 5000000 first zeros of ζ(s) and ∆u =
0.05.

Because the hamiltonian describing interaction inside heavy nuclei is unknown Wigner
proposed to use some matrix of large dimension with random entries selected with the
appropriate distribution probability and subject for example to the hermiticity require-
ment. It means that if M is a square matrix N × N with elements Mij , then probability
P (Mij ∈ (a, b)) that a given matrix element Mij will take value in the interval (a, b) is
given by the integral: Z b
P (Mij ∈ (a, b)) = fij (x)dx,
a
where fij is the density of the probability distribution and matrix elements Mij are mu-
tually statistically independent, what means that the probability for the whole matrix
is the product of above factors for single elements Mij . The requirement of hermitic-
ity (M† = M) and independence with respect to the choice of the base determine the
following form, see [52, Theorem 2.6.3, p. 47]:
2 +btr M+c
P (M) = e−atr M , (36)
where a P
is a positive real number, b i c are real and tr denotes trace of the matrix:
tr M = N i=1 Mii . The value of c is determined by normalization of the probability. For

15
self–adjoint matrix H† = H we have tr H 2 = N
P PN PN PN ?
i=1 k=1 H ik Hki = i=1 k=1 Hik Hik =
PN PN 2
i=1 k=1 |Hik | and all terms in (36) have a Gaussian form. Because exponent of the
sum of terms is the product of the exponents of each factors separately, the right side of the
equation (36) indeed has the form of the product of the density of the normal Gaussian
distributions and such a set of random, Gaussian unitary matrices is called Gaussian
Unitary Ensemble, in short GUE. Eigenvalues of such matrices are not completely random:
“unfolded” gaps s between them are not described by the Poisson distribution e−s , but
for example for GUE by the formula
32 2 −4s2 /π
P (s) = se . (37)
π2
Unfolding means getting rid of constant trend in the spectrum E1 , E2 , . . ., i.e. dividing
dn = En+1 − En by mean gaps between levels d(E) : sn = (En+1 − En ))/d(E). For
zeros of ζ(s) from equation (23) the differences γn+1 − γn are changed into sn = (γn+1 −
γn ) log(γn )/2π. In the Fig.7 we show the comparison of (37) with real gaps for zeros of
ζ(s).
Level-spacing distributions of quantum systems can be grouped into a few universality
classes connected with the symmetry properties of the Hamiltonians: Poisson distribution
for systems with underlying regular classical dynamics, Gaussian orthogonal ensemble
(GOE, also called the Wigner-Dyson distribution) — Hamiltonians invariant under time
reversal, Gaussian unitary ensemble (GUE) — not invariant under time reversal and
Gaussian symplectic ensemble (GSE) for half-spin systems with time reversal symmetry.
There are many reviews on these topics, we cite here [52], [53], [54], we strongly recommend
the review [55]. Dyson and Mehta identified these three types of random matrices with
different intensities of repulsion spacings between consecutive energy levels : GOE with
weakest repulsion between neighboring levels, GUE with medium repulsion and GSE with
strongest repulsion. For quantitative description see [56, Appendix A].
For several years discovered during a brief conversation of Montgomery with Dyson
relationship of nontrivial zeros ζ(s) with the eigenvalues of matrix from the GUE did
not arouse much interest. In the eighties of previous century Andrew Odlyzko performed
over many years computation of zeta zeros in different intervals and calculated their pair-
correlation function numerically. In the first paper [57] he tested Montgomery pair corre-
lation conjecture for first 100,000 zeros and for zeros number 1012 to 1012 + 100, 000. Next
he looked at 1020 -th zero of the Riemann zeta function and 70 million of its neighbors, the
1020 -th zero of the Riemann zeta function and 175 million of its neighbors, last searched
interval was around zero 1022 and involved 109 zeros, see [29]. The reason Odlyzko inves-
tigated zeros further and further is the very slow convergence of various characteristics of
ζ(s) to its asymptotic behavior. The results confirmed the GUE distribution: the gaps
between imaginary parts of consecutive nontrivial zeros of ζ(s) display the same behav-
ior as the differences between pairs of eigenvalues of random Hermitian matrices, see [57,
Fig.1 and Fig.2]. In [58, p.146] Peter Sarnak wrote: “At the phenomenological level this is
perhaps the most striking discovery about the zeta function since Riemann.” In this way
vague hypothesis of Hilbert - Polya has gained credibility and now it is known that a phys-
ical system corresponding to ζ(s) has to break the symmetry with respect to time reversal.
At the conference “Quantum chaos and statistical nuclear physics” held in Cuernavaca,
Mexico, in January 1986 Michael Berry delivered the lecture Riemanns zeta function: a
model for quantum chaos? [59] which became the manifesto of the approach to prove

16
the RH which can be summarized symbolically as ζ( 12 + iĤR ) = 0 with ĤR a hermitian
operator having as eigenvalues imaginary parts of nontrivial zeros γk : ĤR |Ψk i = γk |Ψk i.
The hypothetical quantum system (fictitious element) described by such a hamiltonian
was dubbed by Oriol Bohigas “Riemannium”, see [60, 61]. Additionally to the lack of time
reversal invariance of ĤR Berry in [59] pointed out that ĤR should have a classical limit
with classical orbits which are all chaotic (unstable and bounded trajectories). In fact
the departure of correlation function for zeta zeros from (35) for large spacings (argument
larger than 1 in [57, Fig.1 and Fig.2]) was a manifestation of quantum chaos, as Berry
recognized. Later on M. Berry and J. Keating have argued [62] that ĤR = xp. The main
argument for connection of ĤR = xp with the RH was the fact, that the number of states
of this hamiltonian with energy less than E is given by the formula:
   
E E 7
N (E) = log − 1 + + ...
2π 2π 8

what exactly coincides with (23). In the derivation of above result Berry and Keating
“cheated” using very special Planck cell regularization to avoid infinite phase–space vol-
ume. As a caution we mention here an example of a very special shape billiard for which
the formula for a number of energy levels below E has a leading term exactly the same
as for zeta function (23) but the next terms disagree, see [63, eqs. (34–35)]. We remind
here that two drums can have different shapes but identical eigenvalues of vibrations,
thus the same spectral staircase function. In 2011 S. Endres and F. Steiner [64] showed
that spectrum of ĤR = xp on the positive x axis is purely continuous and thus ĤR = xp
cannot yield the hypothetical Hilbert-Polya operator possessing as eigenvalues the non-
trivial zeros of the ζ(s) function. The choice ĤR = xp for the operator of Riemannium
possesses some additional drawbacks (e.g. it is integrable, and therefore not chaotic) and
some modification of it were proposed, see series of papers by G. Sierra e.g. [65, 66].
In August 1998, during a conference in Seattle devoted to 100–th anniversary of the
PNT, Peter Sarnak offered a bottle of good wine for physicists who will be able to recover
some information from the Montgomery - Odlyzko conjecture that is not formerly known
to mathematicians. Just two years later he had to go to the store to buy promised wine.
At the conference in Vienna in September 1998, Jon Keating delivered a lecture during
which he announced solution (but no proof) of the so called problem of moments of zeta.
These results were published later in a joint work with his PhD student Nina Snaith [67].
For nearly a hundred years mathematicians have tried to calculate moments of the zeta
function on the critical line
1 T
Z
|ζ( 21 + ıt)|2k dt, for T → ∞ (38)
T 0

G.H. Hardy and J.E. Littlewood [68, Theorem. 2.41] calculated the second moment:

1 T
Z
|ζ( 12 + ıt)|2 dt ∼ log(T ), for T → ∞. (39)
T 0

The fourth moment calculated A.E. Ingham in 1926 [69, Th. B]

1 T
Z
|ζ( 12 + ıt)|4 dt ∼ 2π
1
log4 (T ), for T → ∞. (40)
T 0

17
Higher moments, despite many efforts, were not known, but it was supposed for k = 3
[70] that:
Z T ( 4  )
42 Y  1 4 1
|ζ( 12 + it)|6 dt ∼ 1− 1+ + 2 T log9 T for large T, (41)
1 9! p p p p

and even more complex expression for k = 4 [71].

Z T  9  !
24024 Y 1 9 9 1
|ζ(1/2 + it)|8 dt ∼ 1− 1+ + 2 + 3 T log16 T. (42)
0 16! p p p p p

Keating and Snaith proved the general theorem for moments of random matrices, which
eigenvalues have GUE distribution and if the behavior of ζ(s) is modeled by the determi-
nant of such a matrix, then their result applied to the zeta gives

1 T  1
Z
2k 2
ζ( 2 + ıt) dt ∼ fk a(k)(log T )k , (43)
T 0
where k 2 X
Y ∞  2
1 Γ(m + k)
a(k) = 1− p−m , (44)
p
p m=0
m! Γ(k)

and numbers fk are given by

G2 (k + 1)
fk = .
G(2k + 1)
In the above formula G(·) is the Barnes function satisfying the recurrence G(z + 1) =
Γ(z)G(z) with starting value G(1) = 1, thus for natural arguments this function is a
“factorial over factorials”: G(n) = 1! · 2! · 3! . . . (n − 2)!. Of course, the result of Keating
and Snaith gives formulas (39)–(42), respectively for k = 1, 2, 3, 4.
In [72] P. Crehan has shown that for any sequence of energy levels obeying a certain
growth law (|En | < ean+b , for some a ∈ R+ , b ∈ R), there are infinitely many classically
integrable Hamiltonians for which the corresponding quantum spectrum coincides with
this sequence. Because from PNT it follows, that the n−th prime pn grows like pn ∼
n log(n) the results of Crehan’s paper can be applied and there exist classically integrable
hamiltonians whose spectrum coincides with prime numbers, see also [73]. From (23) it
follows that the imaginary part of the n–th zero of ζ(s) grows like γn ∼ 2πn/ log(n), thus
the theorem of Crehan can be applied and it follows that there exists an infinite family
of classically integrable nonlinear oscillators whose quantum spectrum is given by the
imaginary part of the sequence of zeros on the critical line of the Riemann zeta function.
In the end of XX centaury there were a lot of rumors that Allain Connes has proved
the RH using developed by him noncommutative geometry. Connes [74] constructed a
quantum system that has energy levels corresponding to all the Riemann zeros that lie on
the critical line. To prove RH it has to be shown that there are no zeros outside critical
line, i.e. unaccounted for by his energy levels. The operator he constructed acts on a very
sophisticated geometrical space called the noncommutative space of adele classes. His
approach is very complicated and in fact zeros of the zeta are missing lines (absorption
lines) in the continuous spectra. During the passed time excitement around Conness

18
work has faded and much of the hope that his ideas might lead to the proof of RH has
evaporated. The common opinion now is that he has shifted the problem of proving the
RH to equally difficult problem of the validity of a certain trace formula.
We mention also the paper written by S. Okubo[75] entitled “Lorentz-Invariant Hamil-
tonian and Riemann Hypothesis”. It is not exactly the realization of the idea of Polya and
Hilbert: the appearing in this paper two dimensional differential operator (hamiltonian
H) does not possess as eigenvalues imaginary parts of the nontrivial zeros of the ζ(s).
Instead the special condition for zeros of zeta function is used as the boundary condi-
tion for solutions of the eigenvalue equation H|φi = λ|φi. Unfortunately, the obtained
eigenfunctions are not normalizable.
Let us remark that for trivial zeros −2n of ζ(s) with a constant gap 2 between them it
is possible to construct hamiltonian reproducing these zeros as eigenvalues. Namely, the
eigenvalue problem for the harmonic oscillator in the units m = ~ = ω = 1 has the form:

d2
 
1 2
− 2 + x ψn (x) = (n + 21 )ψn (x) (45)
2 dx

where
x2
ψn (x) = e− 2 · Hn (x), n = 0, 1, 2, . . .
and Hn (x) are Hermite polynomials:

2 dn  −x2 
Hn (x) = (−1)n ex e .
dxn
Multiplying (45) by −2 and rearranging terms we obtain
 2 
d 2
− x + 1 ψn (x) = −2nψn (x).
dx2

Hence the hamiltoniam for trivial zeros of ζ(s) is

d2
Ĥtriv = − x2 + 1.
dx2
Since the advent of quantum computers and the discovery by Peter Shor of the quan-
tum algorithm for integer factorization [76] there is an interest in applying these algorithms
to diverse of problems. Is it possible to devise the quantum computer verifying the RH?
We mean here something more clever than, say, simply mixing the Shor’s algorithm with
Lagarias criterion. Recently there appeared the paper [77], in which authors (assuming
the RH) have built an unitary operator with eigenvalues equal to combination of nontriv-
ial zeros ρj /ρj lying on the unit circle. Next the quantum circuit representing this unitary
matrix is constructed. Recently in [78, p. 4] the quantum computer verifying RH was
proposed, but it seems to us to be artificial and not sufficiently sophisticated: it is based
on the (26) and it counts in a quantum way actual number of prime numbers below x and
looks for departures beyond the bound in (26).
We do not have space here to discuss the use of ζ(s) in the theory of Casimir effect —
devoted to this subject is the extensive review by K. Kirsten in [79], or in string theory
[80, 81].

19
5 Statistical Mechanics and RH
The partition function Z(β) is the basic quantity used in statistical physics, here β =
1/kB T : kB = 1.3806488 . . . × 10−23 [J/K] is the Boltzmann constant and T is the absolute
temperature. All thermodynamical functions can be expressed as derivatives of Z(β).
The phase transitions appear at such temperatures that Z(β) = 0. For the system, which
may be in micro–states with energy En and can exchange heat with environment and
with fixed number of particles, volume and temperature, the partition function is given
by the formula: X
Z(β) = e−βEn . (46)
n

It turns out that for certain systems Z(β) satisfies the relation similar to functional
equation for ζ(s) and positions of zeros of the partition function analytically continued to
the whole complex plane are highly restricted, for example to the circle. These two facts
have become the starting point for attempts to prove HR.
It is very easy to construct the system with the ζ(s) as a partition function. The prob-
lem of construction of a simple one–dimensional Hamiltonian whose spectrum coincides
with the set of primes was considered in [82], [83], [84], see also review [73]. Some mod-
ification should lead to the Hamiltonian H having eigenstates |pi labeled by the prime
numbers p with eigenvalues Ep = E log(p), where E is some constant with dimension of
energy. The n particle state can be decomposed into the statesP |pi using the factorization
theorem (5). The energy of the state |ni is equal to E(n) = E ki=1 αi log(pi ) = E log(n).
Then the partition function Z is given by the Riemann zeta function:
∞   ∞   ∞
X −En X −E log n X 1
Z(T ) = exp = exp = s
= ζ(s), s ≡ E/kB T
n=1
kB T n=1
kB T n=1
n

Such a gas was considered e.g. in [18] and [85] and found applications in the string theory.
The functional equation (15) can be written in non–symmetrical form:
π

2Γ(s) cos s ζ(s) = (2π)s ζ(1 − s).
2
In this form it is analogous to the Kramers–Wannier [86] duality relation for the partition
function Z(J) of the two dimensional Ising model with parameter J expressed in units of
kB T (i.e. equal to interaction constant multiplied by β = 1/kB T )

Z(J) = 2N (cosh(J))2N (tanh(J))N Z(J),

e (47)

where N denotes the number of spins and Je is related to J via e−2J = tanh(J), see e.g.
e

[87]. On the other hand there are so called “Circle theorems” on the zeros of partition
functions of some particular systems. To pursue this analogy one has to express the
partition function by the ζ(s) function. Then one can hope to prove RH by invoking the
Lee–Yang circle theorem on the zeros of the partition function. The Lee–Yang theorem
concerns the phase transitions of some spin systems in external magnetic field and some
other models (for a review see [88]). Let Z(β, z) denote the grand – canonical partition
function, where z = eβH is the fugacity connected with the magnetic field H. Phase
transitions are connected with the singularities of the derivatives of Z(β, z), and they

20
appear when Z(β, z) is zero. The finite sum defining Z(β, z) can not be a zero for real β
or z and the Lee–Yang theorem [89, 90] asserts that in the thermodynamical limit, when
the sum for partition function involves infinite number of terms, all zeros of Z(β, z) for
a class of spin models are imaginary and lie in the complex plane of the magnetic field z
on the unit circle: |z| = 1. The study of zeros of the canonical ensemble in the complex
plane of temperature β was initiated by M.Fisher [91]. He found in the thermodynamic
limit for a special Ising model not immersed in the magnetic field, that the zeros of
the canonical partition function also lie on an unit circles, this time in the plane of the
complex variable v = sinh(2Jβ), where J > 0 is the ferromagnetic coupling constant.
The critical line s = 12 + it can be mapped into the unit circle via the transformation
s → u = s/(1 − s) = ( 21 + it)/( 12 − it) because then |u| = 1. Thus by devising appropriate
spin system with Z(β, z) expressed by the ζ(s) the Lee–Yang theorem can be used to
locate the possible zeros of the latter function and lead to the proof of RH.
In the series of papers A. Knauf [92, 93, 94] has undertaken the above outlined plan to
attack the RH. In these papers he introduced the spin system with the partition function
in the thermodynamical limit expressed by zeta function: Z(s) = ζ(s − 1)/ζ(s) with
s interpreted as the inverse of temperature. However the form of interaction between
spins in his model does not belong to one of the cases for which the circle theorem was
proved. This idea was further developed in paper [95]. The authors of the paper [96] have
shown that RH is equivalent to an inequality satisfied by the Kubo-Martin-Schwinger
states of the Bost and Connes quantum statistical dynamical system in special range of
temperatures. There are many other appearances of the ζ(s) in the statistics of bosons
and fermions, theory of the Bose–Einstein condensate, some special “number theoretical”
gases etc, for introduction see [14, chap. III E].

6 Random walks, billiards, experiments etc.

The Möbius function defined in (10) takes only three values: -1, 0 and 1. The val-
ues µ(n) = 1 and µ(n) = −1 are equiprobable with probabilities 3/π 2 ≈ 0.3039, thus
the probability of value µ(n) = 0 is 1 − 6/π 2 ≈ 0.3921. Using values 1 and -1 of the
Möbius function instead of heads or tails of a coin should hence generate a symmetric
one-dimensional random walk. The total displacement during n steps of suchPa random
walk will be given by the summatory function of the Möbius function: M (x) = n<x µ(n),
which is called the Mertens function. It is well known that the “root mean square” dis-
tance
√ from the starting point of the symmetrical random walk during N steps grows like
N . The resemblance of M (n) to the symmetrical random walk led F. Mertens in the
end of XIX century to make the conjecture that M (n) grows √ not faster than the mean
displacement of the symmetrical random walk, i.e. |M (n)| < n. It is an easy calculation
to show that Mertens conjecture implies the RH (vide (11)):
∞ ∞
X µ(n) X M (n) − M (n − 1) X ∞  
1 1 1
= = = M (n) − =
ζ(s) n=1 ns n=1
ns n=1
ns (n + 1)s

∞ Z n+1 ∞ Z n+1 Z ∞
X sdx X M (x)dx M (x)dx
= M (n) s+1
=s s+1
=s .
n=1 n x n=1 n
x 1 xs+1

21
 √  
 1  |s|
If M (x) < x then the last integral above gives  ζ(s) < σ− 12
, thus to the right of the
line <[s] = 21 the inverse of zeta function is bounded hence there can not be zeros of ζ(s)
in this region and the truth of RH follows. For many √ years mathematicians hoped to
prove the RH by showing the validity of |M (n)| < n. However in 1985 A. Odlyzko and
H. te Riele [97] disproved the Mertens conjecture; in the proof they have used values of
first 2000 zeros of ζ(s) calculated with accuracy 100–105 digits; these calculations took 40
hours on CDC CYBER 750 and 10 hours on Cray-1 supercomputers. Using Mathematica
these computations can be done on the modern laptop in a couple of minutes. Littlewood
proved that the RH is equivalent to slightly modified Mertens conjecture

M (n) = O(n1/2+ ) ⇔ RH is true.

The fact that M (n) behaves like a one dimensional random walk was also pointed out in
[98] and used to show that RH is “true with probability 1”.
In the paper [99] M. Shlesinger has investigated a very special one-dimensional random
walk which can be linked with the RH. The probability of jumping to other sites with
steps having a displacement of ±l sites involves directly the Möbius function:
 
1 1 µ(l)
p(±l) = C ± , β > 0,
2 l 1+β l 1+β−
1
where C = 1
ζ(1+β)+ ζ(1+β)
is a normalization factor, β (to be not confused with β = 1/kB T )
is the fractal dimension of the set of points visited by random walker. He coined the
name Riemann–Möbius for this random walk. Some general properties P ofiklthe ”structure
function” λ(k) being the Fourier of the probabilities p(l): λ(k) = l e p(l), enabled
Shlesinger to locate the complex zeros inside the critical strip, however the result of J.
Hadamard and Ch. J. de la Valle–Poussin that ζ(1 + it) 6= 0 can not be recovered by this
method. What is interesting the existence of off critical line zeros is not in contradiction
with behavior of λ(k) following from the universal laws of probability.
In [100] the stochastic interpretation of the Riemann zeta function was given. There
are much more connections between ζ(s) and random walks as well as Brownian motions
known to mathematicians. The extensive review of obtained results expressing expecta-
tion values of different random variables by ζ(s) or ξ(s) can be found in [101].
In the paper [102] L.A. Bunimovich and C.P. Dettmann considered the point particle
bouncing inside the circular billiard. There is a possibility that the small ball will escape
through a small hole on the reflecting perimeter. Let P1 (t) denote the probability of not
escaping from a circular billiard with one hole till time t. Bunimovich and Dettmann
obtained exact formula for P1 (t) and surprisingly this probability was expressed by ζ(s).
So here again the function of purely number theoretical origin meets the physical reality.
Then they proved that RH is equivalent to

lim lim δ (tP1 (t) − 2/) = 0 (48)

→0 t→∞

be true for every δ > −1/2. Here this value 1/2 is directly connected with the location of
critical line in the formulation of RH. A little bit more complicated condition was obtained
for biliard with two holes. In principle such conditions allow experimental verification of
RH using microwave cavities simulating billiards or optical billiards constructed with
microlasers. Experiments can refute RH if the behavior of tP1 (t) − 2/ in the limit  → 0

22
will be slower than power like dependence 1/2 in the limit of vanishing . To our knowledge
up today no such experiments were performed. In the paper [103] generalization to the
spherical billiard was considered. Again the survival probability in such a 3D biliard is
related to the Riemann hypothesis.
Already in 1947 van der Pol has built the electro–mechanical device verifying the
RH [104]. He has built machine plotting the ζ(1/2 + it) from the following integral
representation: Z ∞
ζ( 21 + it) −x/2 x x/2

−ixt
1 = e be c − e e dx. (49)
2
+ it −∞

Here bxc denotes integer part of x. It has the form of Fourier transform of the function
y(x) = e−x/2 bex c − ex/2 . The plot of integrand is shown in Fig. 8. The shape of this
function was cut precisely with scissors on the edge of a paper disc. The beam of light
was passing between teeth on the perimeter of the disc and detected by the photocell.
The resulting from photoelectric effect current was superimposed with current of varying
frequency to perform analogue Fourier transform. After some additional operations van
der Pol has obtained the plot of modulus |ζ( 21 + it)/ 12 + it| on which the first 29 nontrivial
zeta zeros were located with accuracy better than %1. The authors of [14] have summa-
rized this experiment in the words: “This construction, despite its limited achievement,
deserves to be treated as a gem in the history of the natural sciences.”.
It is well known that two–dimensional electro-
static fields can be found using the functions of com-
plex variables. There arises a question to which elec-
trostatic problem the zeta function can be linked?
In the recent paper [105] A. LeClair has developed
this analogy and he constructed a two–dimensional
→
−
vector field E from the real and imaginary parts of
the zeta function. It allowed him to derive the for-
mula for the n-th zero on the critical line of ζ(s) for
large n expressed as the solution of a simple tran-
scendental equation.
In the written [106] version of his AMS Einstein
Lecture “Birds and frogs” (which was to have been
Figure 8: Plot of the function appear-
given in October 2008 but which unfortunately had
ing in the integral representation (49)
of the zeta function.
to be canceled) Freeman Dyson points to the possi-
bility of proving the RH using the similarity in be-
havior between one–dimensional quasi-crystals and
the zeros of the ζ(s) function. If RH is true then locations of its nontrivial zeros would
define a one–dimensional quasi–crystal but the classification of them is still missing.

7 Zeta is a fractal
In 1975 S.M. Voronin [107] proved remarkable theorem on the universality of the Riemann
ζ(s) function.
Voronin’s theorem: Let 0 < r < 1/4 and f (s) be a complex function continuous for
|s| ≤ r and analytical in the interior of the disk. If f (s) 6= 0, then for every  > 0 there

23
exists real number T = T (, f ) such that:
   
 3 
max f (s) − ζ s + +iT  < . (50)
|s|≤r 4 
Put simply in words it means that the zeta
function approximates locally any smooth
function in a uniform way! By applying this
theorem to itself, i.e. taking as f (s) = ζ(s),
we obtain that ζ(s) is selfsimilar, see S.C.
Woon [108] who has shown that the Riemann
ζ(s) is a fractal. In the paper [109] the
Voronin’s theorem was applied to the phys-
ical problem: to propose a new formulation
of the Feynmans path integral.
Another aspect of fractality of zeta was
found in [110, 84], where the one–dimensional
quantum potential was numerically con-
structed from known zeta zeros which in turn
are reproduced as eigenvalues of this poten-
tial. The fractal dimension of the graph of
this potential was determined to be around
1.5. In [84] even the multifractal nature of
this potential was revealed.
In the late seventies of XX centaury John
Hubbard has analysed the Newton’s method
for finding approximations to the roots of
equation f (x) = 0 to the case of polynomial
z 3 − 1 on the complex plane. In this method
the root x? of f (x? ) = 0 is obtained as a limit
Figure 9: The plot of initial values for (51) x? = limn→∞ xn of the sequence:
−9 < <(z0 ) < 9, −25 < =(z0 ) < 25 showing
in colors the number n of iterations of (51) f (xn )
xn+1 = xn − 0 .
after which |ζ(zn )| < 0.000001. In black are f (xn )
shown regions around zeros of zeta. Clearly If the function f (x) has a few roots the
are visible zeros: -2, -4, -6, -8 and nontrivial limit depends on the choice of the initial x .
1 1 0
2 ± i14.134 . . . , 2 ± i21.022 . . .. Hubbard was interested in the question which
√ starting points z0 ∈ C tend to one of three
3
roots 1, (−1 ± 3i)/2 of z = 1. He obtained one of the first fractal images full of
interwoven corals. Tomoki Kawahira has applied Newton’s method to the Riemann’s zeta
function:
ζ(zn )
zn+1 = zn − 0 . (51)
ζ (zn )
Because ζ(s) has infinitely many roots, instead of looking for basin domains of different
zeta zeros, he looked for the number of iterations of (51) for a given starting point z0
needed to fall into the close vicinity of one of the zeros. Let us mention that such a
modification was also applied to the original problem z 3 = 1. He obtained beautiful
pictures representing the zeros of ζ(s). We present in Fig. 9 the plot obtained by M.
Dukiewicz [111].

24
8 Concluding Remarks
We have given many examples of physical problems connected to the RH. In XIX centaury
all these problems were not known, but it seems that Riemann believed that the questions
of mathematics could be answered with the help of physics and in fact he performed some
physical experiments by himself to check some of his theorems, see [112]. We add here,
that there is a wide spread rumor among the people who are trying to solve the RH that
Fields Medal Laureate Enrico Bombieri believes that RH will be proved by a physicist,
see [8, p. 4].
Some mathematicians enunciate the opinion that RH is not true because long open con-
jectures in analysis tend to be false. In other words nobody has proved RH because simply
it is not true. There are examples from number theory when some conjectures confirmed
by huge “experimental” data finally turned out to be false and possible counterexamples
are so large that never will be accessible to computers. One such common belief was the
inequality Li(x) > π(x) remarked already by Gauss and confirmed by all available data,
now it is about x = 1018 . However, in 1914 J.E. Littlewood has shown [113] that the
difference between the number of primes smaller than x and the logarithmic integral up
to x changes the sign infinitely many times, what was another rather complicated proof of
the infinitude of primes. The smallest value xS such that for the first time π(xS ) ≥ Li(xS )
holds is called Skewes number because in 1933 S. Skewes [114], assuming the truth of the
Riemann hypothesis, argued that it is certain that π(x) − Li(x) changes sign for some
1034
xS < 1010 . In 1955 Skewes [115] has found, without assuming the Riemann hypothe-
3
1010
ses, that π(x) − Li(x) changes sign at some xS < exp exp exp exp(7.705) < 1010 . This
enormous bound for xS was several times lowered and the lowest present day known es-
timation of the Skewes number is around 10316 , see [116] and [117]. The second example
1
is provided by the Mertens conjecture discussed in Sect.6. The inequality |M (x)| < x 2
is confirmed by all available data but finally it is false. Like in the case of the inequality
1
π(x) > Li(x) we can expect first x for which |M (x)| > x 2 at horribly heights. Namely
J. Pintz [118] has shown that the first counterexample appears below exp(3.21 × 1064 ).
This upper bound was later lowered to exp(1.59 × 1040 ) [119]. Such examples show that
.
..
confirmation of some facts up to say 1018 is misleading and somewhere at t = 1010 the
nontrivial zero of ζ(s) with real part different from 12 can be lurking.

Acknowledgement: I thank Magdalena Dukiewicz–Jurka for allowing me to use the

Fig. 9 from her master thesis (performed under my guidance in 2006). I thank Magdalena
Zaluska–Kotur and Jerzy Cislo for comments and remarks.

References
[1] D. Zagier, “The first 50 million prime numbers,” Mathematical Intelligencer, vol. 0,
pp. 7–19, 1977.

[2] D. Hilbert, “Mathematical problems. Lecture Delivered Before The International

Congress Of Mathematicians At Paris In 1900,” Bull. Amer. Math. Soc., vol. 8,
pp. 437–479, 1992.

25
 [3] A. H. Helfgott, “The ternary Goldbach problem,” arXiv: math.NT/1501.05438,
Jan. 2015.

[4] J. Derbyshire, Prime Obsession. Bernhard Riemann and the greatest unsolved prob-
lem in mathematics. Washington: Joseph Henry Press, 2003.

[5] K. Sabbagh, The Riemann hypothesis : the greatest unsolved problem in mathemat-
ics. New York: Farrar, Straus, and Giroux, 2002.

[6] K. Sabbagh, Dr. Riemann’s Zeros: The Search for the $1 million Solution to the
Greatest Problem in Mathematics. Atlantic Books, November 2002.

[7] D. Rockmore, Stalking the Riemann hypothesis : the quest to find the hidden law of
prime numbers. New York: Pantheon Books, 2005.

[8] M. Sautoy, The music of the primes: searching to solve the greatest mystery in
mathematics. New York: HarperCollins, 2003.

[9] E. C. Titchmarsh, The Theory of the Riemann Zeta-function. New York: The
Clarendon Press Oxford University Press, sec. ed. ed., 1986. Edited and with a
preface by D. R. Heath-Brown.

[10] H. M. Edwards, Riemann’s zeta function. Academic Press, 1974. Pure and Applied
Mathematics, Vol. 58.

[11] A. Ivić, The Riemann zeta-function: the theory of the Riemann zeta-function with
applications. New York: Wiley, 1985.

[12] A. Karatsuba and S. M. Voronin, The Riemann zeta-function. Berlin New York:
Walter de Gruyter, 1992.

[13] P. Borwein, S. Choi, B. Rooney, and A. Weirathmueller, The Riemann Hypothe-

sis: A Resource For The Afficionado And Virtuoso Alike. Springer Verlag, Berlin,
Heidelberg, New York, 2007.

[14] D. Schumayer and D. A. W. Hutchinson, “Physics of the Riemann hypothesis,” Rev.

Mod. Phys., vol. 83, pp. 307–330, Apr 2011.

[15] M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions with For-

mulas, Graphs, and Mathematical Tables. New York: Dover, ninth Dover printing,
tenth GPO printing ed., 1964.

[16] B. Riemann, “Ueber die Anzahl der Primzahlen unter einer gegebenen Grösse,”
Monatsberichte der Kniglich Preußischen Akademie der Wissenschaften zu Berlin.,
November 1859.

[17] M. Milgram, “Integral and Series Representations of Riemann’s Zeta Function and
Dirichlet’s Eta Function and a Medley of Related Results,” Journal of Mathematics,
vol. 2013, p. Article ID 181724, 2013.

[18] D. Spector, “Supersymmetry and the Möbius inversion function,” Communications

in Mathematical Physics, vol. 127, no. 2, pp. 239–252, 1990.

26
[19] H. Riesel, Prime Numbers and Computer Methods for Factorization. Birkhuser
Boston, 1994.

[20] W. Rudin, Principles of mathematical analysis. New York: McGraw-Hill Book Co.,
third ed., 1976. International Series in Pure and Applied Mathematics.

[21] D. J. Platt, “Computing π(x) analytically,” arXiv: math.NT/1203.5712, mar 2012.

[22] H. Riesel and G. Göhl, “Some calculations related to Riemann’s prime number
formula,” Mathematics of Computation, vol. 24, pp. 969–983, 1970.

[23] J. P. Gram, “Note sur les zéros de la fonction de Riemann,” Acta Math., vol. 27,
pp. 289–304, 1903.

[24] A. M. Turing, “Some calculations of the Riemann zeta-function,” Proc. London

Math. Soc. (3), vol. 3, pp. 99–117, 1953.

[25] S. Wedeniwski, “Zetagrid Project.” http://think-automobility.org/

geek-stuff/zetagrid.

[26] X. Gourdon, “The 1013 first zeros of the Riemann Zeta Function, and zeros compu-
tation at very large height,” Oct. 24, 2004. http://numbers.computation.free.
fr/Constants/Miscellaneous/zetazeros1e13-1e24.pdf.

[27] A. M. Odlyzko, “The 1020 -th zero of the Riemann zeta function and 175 million of
its neighbors.” 1992 revision of 1989 manuscript.

[28] A. M. Odlyzko, “The 1021 -st zero of the Riemann zeta function.” Nov. 1998 note
for the informal proceedings of the Sept. 1998 conference on the zeta function at
the Edwin Schroedinger Institute in Vienna.

[29] A. M. Odlyzko, “The 1022 -nd zero of the Riemann zeta function,” in Dynamical,
Spectral, and Arithmetic Zeta Functions (M. van Frankenhuysen and M. L. Lapidus,
eds.), no. 290 in Amer. Math. Soc., Contemporary Math. series, pp. 139–144, 2001.

[30] G. H. Hardy, “Sur les zéros de la fonction ζ(s) de Riemann,” C. R. Acad. Sci. Paris,
vol. 158, pp. 1012–1014, 1914.

[31] S. Feng, “Zeros of the Riemann zeta function on the critical line,” Journal of Number
Theory, vol. 132, pp. 511–542, 2012.

[32] A. M. Odlyzko, “Primes, quantum chaos, and computers,” Number Theory, Na-
tional Research Council, pp. 35–46, 1990.

[33] M. Wolf, “Two arguments that the nontrivial zeros of the Riemann zeta function
are irrational,” arXiv:math.NT/1002.4171, Feb 2010.

[34] H. von Koch, “Sur la distribution des nombres premiers,” Acta Math., vol. 24,
pp. 159–182, 1901.

[35] L. Schoenfeld, “Sharper Bounds for the Chebyshev Functions θ(x) and ψ(x). II,”
Mathematics of Computation, pp. 337–360, 1976.

27
[36] N. de Bruijn, “The roots of trigonometric integrals,” Duke Math. J, vol. 17, no. 3,
pp. 197–226, 1950.

[37] C. M. Newman, “Fourier transforms with only real zeros,” Proc. Amer. Math. Soc.,
vol. 61, no. 2, pp. 245–251, 1976.

[38] Y. Saouter, X. Gourdon, and P. Demichel, “An improved lower bound for the de
Bruijn-Newman constant,” Mathematics of Computation, vol. 80, pp. 2281–2287,
2011.

[39] A. M. Odlyzko, “An improved bound for the de Bruijn–Newman constant,” Numer-
ical Algorithms, vol. 25, no. 1, pp. 293–303, 2000.

[40] V. V. Volchkov, “On an equality equivalent to the Riemann Hypothesis,” Ukraı̈n.

Mat. Zh., vol. 47, no. 3, pp. 422–423, 1995.

[41] Y.-H. He, V. Jejjala, and D. Minic, “From Veneziano to Riemann: A String Theory
Statement of the Riemann Hypothesis,” arXiv/1501.01975, 2015.

[42] M. Balazard, E. Saias, and M. Yor, “Notes sur la fonction ζ de Riemann. II,” Adv.
Math., vol. 143, no. 2, pp. 284–287, 1999.

[43] The PARI Group, Bordeaux, PARI/GP, version 2.3.2, 2008. available from http:
//pari.math.u-bordeaux.fr/.

[44] J. C. Lagarias, “An elementary problem equivalent to the Riemann Hypothesis,”

Amer. Math. Monthly, vol. 109, pp. 534–543, 2002.

[45] K. Briggs, “Abundant Numbers and the Riemann Hypothesis,” Experimental Math-
ematics, vol. 15, Number 2, pp. 251–256, 2006.

[46] G. Robin, “Grandes valeurs de la fonction somme des diviseurs et Hypothèse de

Riemann,” J. Math. Pures Appl. (9), vol. 63, no. 2, pp. 187–213, 1984.

[47] A. Ivic̀, “On some reasons for doubting the Riemann hypothesis,”
arXiv:math/0311162, Nov. 2003.

[48] Correspondence about the origins of the Hilbert-Polya Conjecture at the web site
of A. Odlyzko. http://www.dtc.umn.edu/~odlyzko/polya/index.html.

[49] H. Weyl, “Über die asymptotische Verteilung der Eigenwerte,” Nachrichten der
Königlichen Gesellschaft der Wissenschaften zu Göttingen, vol. 2, pp. 110 –117,
1911.

[50] H. Weyl, “Über die Abhängigkeit der Eigenschwingungen einer Membran und deren
Begrenzung,” Journal für die reine und angewandte Mathematik, vol. 141, pp. 1–11,
1912.

[51] H. L. Montgomery, “The pair correlation of zeros of the zeta function,” in Analytic
Number Theory (Proc. Sympos. Pure Math., Vol. XXIV, St. Louis Univ., St. Louis,
MO., 1972), pp. 181–193, Providence, R.I.: Amer. Math. Soc., 1973.

28
[52] M. L. Mehta, Random Matrices. San Diego, California: Academic Press, second ed.,
1991.

[53] F. Haake, Quantum Signatures of Chaos. Springer Series in Synergetics, Berlin,

Germany: Springer-Verlag, 2nd ed., 2001.

[54] H. A. Weidenmüller and G. E. Mitchell, “Random matrices and chaos in nuclear

physics: Nuclear structure,” Rev. Mod. Phys., vol. 81, pp. 539–589, May 2009.

[55] F. W. K. Firk and S. J. Miller, “Nuclei, primes and the random matrix connection,”
Symmetry, vol. 1, no. 1, pp. 64–105, 2009.

[56] O. Bohigas, “Random matrix theories and chaotic dynamics,” in Chaos and Quan-
tum Physics,Les-Honches Session LII (M. Giannoni and A. Voros, eds.), Lecture
Notes in Physics, pp. 87–199, North-Holland, Amsterdam, 1991.

[57] A. M. Odlyzko, “On the distribution of spacings between zeros of the zeta function,”
Math. Comp., vol. 48, pp. 273–308, 1987.

[58] P. Sarnak, “Quantum Chaos, Symmetry, and Zeta functions, II: Zeta functions,” in
Current Developments in Mathematics (R. Bott, A. Jaffe, D. Jerison, G. Lusztig,
I. Singer, and S.-T. Yau, eds.), vol. 1997, pp. 145–159, International Press, 1997.

[59] M. V. Berry, “Riemanns zeta function: a model for quantum chaos?,” in Quantum
chaos and statistical nuclear physics (T. H. Seligman and H. Nishioka, eds.), vol. 263
of Lecture Notes in Physics, pp. 1–17, Springer Berlin / Heidelberg, 1984.

[60] P. Leboeuf, A. G. Monastra, and O. Bohigas, “The Riemannium,” Regular and

Chaotic Dynamics, vol. 6, pp. 205–210, 2001.

[61] B. Hayes, “The Riemannium,” American Scientist, vol. 91, pp. 296–300, JulyAugust
2003.

[62] M. Berry and J. Keating, “H = xp and the Riemann Zeros.,” in Supersymmetry

and Trace Formulae: Chaos and Disorder, vol. 370 of NATO ASI Series B Physics,
pp. 355–368, Plenum Press, 1999.

[63] F. Steiner and P. Trillenberg, “Refined asymptotic expansion for the partition func-
tion of unbounded quantum billiards.,” J. Math. Phys., vol. 31, no. 7, pp. 1670–1676,
1990.

[64] S. Endres and F. Steiner, “The Berry-Keating operator on L2 (R> , dx) and on com-
pact quantum graphs with general self-adjoint realizations,” Journal of Physics A:
Mathematical and Theoretical, vol. 43, no. 9, p. 095204, 2010.

[65] G. Sierra and J. Rodriguez-Laguna, “The H = xp model revisited and the Riemann
zeros,” Physical Review Letters, vol. 106, p. 200201, 2011.

[66] G. Sierra, “General covariant xp models and the Riemann zeros,” Journal of Physics
A: Mathematical and Theoretical, vol. 45, no. 5, p. 055209, 2012.

29
[67] J. Keating and N. Snaith, “Random matrix theory and ζ(1/2+ıt),” Commun. Math.
Phys, vol. 214, pp. 57 – 89, 2000.

[68] G. H. Hardy and J. E. Littlewood, “Contributions to the theory of the Riemann

zeta function and the theory of prime distribution,” Acta Mathematica, vol. 41,
pp. 119–196, 1918.

[69] A. Ingham, “Mean-Value Theorems in the Theory of the Riemann Zeta-Function,”

Proc. London Math. Soc., vol. s2-27, pp. pp. 273–300, 01 1928.

[70] J. B. Conrey and A. Ghosh, “A conjecture for the sixth power moment of the
Riemann zeta-function,” Int. Math. Res. Notices, vol. 1998, pp. 775–780, 1998.

[71] J. B. Conrey and S. M. Gonek, “High moments of the Riemann zeta-function,”

Duke Math. J., vol. 107, pp. 577–604, 04 2001.

[72] P. Crehan, “Chaotic spectra of classically integrable systems,” J. Phys. A: Math.

Gen., vol. 28, pp. 6389–6394, 1995.

[73] H. C. Rosu, “Quantum hamiltonians and prime numbers,” Modern Physics Letters
A, vol. 18, pp. 1205–1213, 2003.

[74] A. Connes, “Trace formula in noncommutative geometry and the zeros of the Rie-
mann zeta function,” Selecta Math. (N.S.), vol. 5, no. 1, pp. 29–106, 1999.

[75] S. Okubo, “Lorentz-invariant Hamiltonian and Riemann hypothesis,” Journal of

Physics A: Mathematical and General, vol. 31, no. 3, pp. 1049–1057, 1998.

[76] P. W. Shor, “Polynomial-Time Algorithms for Prime Factorization and Discrete

Logarithms on a Quantum Computer,” arXiv:quant-ph/9508027, Aug. 1995.

[77] R. V. Ramos and F. V. Mendes, “Riemannian Quantum Circuit,” Physics Letters

A, vol. 378, pp. 1346–1349, May 2014.

[78] J. I. Latorre and G. Sierra, “There is entanglement in the primes,” Quantum Infor-
mation & Computation, vol. 15, no. 7&8, pp. 0622–0659, 2015.

[79] K. Kirsten, “Basic zeta functions and some applications in physics,” in A Window
into Zeta and Modular Physics (J. Dehesa, J. Gomez, and A. Polls, eds.), vol. 57 of
MSRI Publications, pp. 101–143, Cambridge University Press, 2010.

[80] S. Cacciatori and M. Cardella, “Equidistribution rates, closed string amplitudes,

and the Riemann hypothesis,” Journal of High Energy Physics, vol. 2010, no. 12,
2010.

[81] C. Angelantonj, M. Cardella, S. Elitzur, and E. Rabinovici, “Vacuum stability,

string density of states and the Riemann zeta function,” Journal of High Energy
Physics, vol. 2011, no. 2, pp. 1–27, 2011.

[82] G. Mussardo, “The Quantum Mechanical Potential for the Prime Numbers,”
arXiv:cond-mat/9712010, 1997.

30
[83] S. K. Sekatskii, “On the Hamiltonian whose spectrum coincides with the set of
primes,” ArXiv e-prints, 2007.

[84] D. Schumayer, B. P. van Zyl, and D. A. W. Hutchinson, “Quantum mechanical

potentials related to the prime numbers and Riemann zeros,” Phys. Rev. E, vol. 78,
p. 056215, Nov 2008.

[85] I. Bakas and M. J. Bowick, “Curiosities of arithmetic gases,” Journal of Mathemat-

ical Physics, vol. 32, no. 7, pp. 1881–1884, 1991.

[86] H. A. Kramers and G. H. Wannier, “Statistics of the Two-Dimensional Ferromagnet.

Part I,” Phys. Rev., vol. 60, pp. 252–262, Aug 1941.

[87] R. P. Feynman, Statistical mechanics: a set of lectures by R. P. Feynman. Frontiers

in physics, New York, NY, USA: W. A. Benjamin, Inc., 1972. Notes taken by R.
Kikuchi and H. A. Feiveson. Edited by Jacob Shaham.

[88] I. Bena, M. Droz, and A. Lipowski, “Statistical Mechanics Of Equilibrium And

Nonequilibrium Phase Transitions: The Yang-Lee Formalism,” International Jour-
nal of Modern Physics B, vol. 19, no. 29, pp. 4269–4329, 2005.

[89] C. N. Yang and T. D. Lee, “Statistical Theory of Equations of State and Phase
Transitions. I. Theory of Condensation,” Phys. Rev., vol. 87, pp. 404–409, Aug
1952.

[90] T. D. Lee and C. N. Yang, “Statistical Theory of Equations of State and Phase
Transitions. II. Lattice Gas and Ising Model,” Phys. Rev., vol. 87, pp. 410–419, Aug
1952.

[91] M. E. Fisher, “The nature of critical points,” in Lectures in Theoretical Physics

(W. Brittin, ed.), vol. VIIC, p. p.1, University of Colorado Press, Boulder, Colorado,
1965.

[92] A. Knauf, “On a ferromagnetic spin chain,” Communications in Mathematical

Physics, vol. 153, no. 1, pp. 77–115, 1993.

[93] A. Knauf, “On a ferromagnetic spin chain. II. Thermodynamic limit,” Journal of
Mathematical Physics, vol. 35, no. 1, pp. 228–236, 1994.

[94] A. Knauf, “The Number-Theoretical Spin Chain and the Riemann Zeroes,” Com-
munications in Mathematical Physics, vol. 196, no. 3, pp. 703–731, 1998.

[95] J. Fiala and P. Kleban, “Generalized Number Theoretic Spin Chain – Connec-
tions to Dynamical Systems and Expectation Values,” Journal of Statistical Physics,
vol. 121, no. 3-4, pp. 553–577, 2005.

[96] M. Planat, P. Solé, and S. Omar, “Riemann hypothesis and quantum mechanics,”
Journal of Physics A Mathematical General, vol. 44, p. 145203, Apr. 2011.

[97] A. M. Odlyzko and H. J. J. te Riele, “Disproof of the Mertens Conjecture,” J. Reine

Angew. Math., vol. 357, pp. 138–160, 1985.

31
 [98] I. J. Good and R. F. Churchhouse, “The Riemann hypothesis and pseudorandom
features of the Möbius sequence,” Mathematics of Computation, vol. 22, pp. 857–
861, 1968.

[99] M. F. Shlesinger, “On the Riemann hypothesis: a fractal random walk approach,”
Physica. A, vol. 138, no. 1–2, pp. 310–319, 1986.

[100] K. S. Alexander, K. Baclawski, and G. C. Rota, “A stochastic interpretation of the

Riemann zeta function,” Proceedings of the National Academy of Sciences, vol. 90,
no. 2, pp. 697–699, 1993.

[101] P. Biane, J. Pitman, and M. Yor, “Probability laws related to the Jacobi theta and
Riemann zeta functions, and Brownian excursions,” Bull. Amer. Math. Soc., vol. 38,
no. 4, pp. 435–465, 2001.

[102] L. Bunimovich and C. Dettmann, “Open Circular Billiards and the Riemann Hy-
pothesis,” Physical Review Letters, vol. 94, no. 10, p. 100201, 2005.

[103] C. P. Dettmann and M. Rahman, “Survival probability for open spherical billiards,”
Chaos, vol. 24, no. 043130, 2014.

[104] B. Van der Pol, “An Electro-Mechanical Investigation Of The Riemann Zeta Func-
tion In The Critical Strip,” Bull. Amer. Math. Soc., vol. 53, pp. 976–981, 1947.

[105] A. Leclair, “An Electrostatic Depiction of the Validity of the Riemann Hypothesis
and a Formula for the N –th Zero at Large N ,” International Journal of Modern
Physics A, vol. 28, p. 50151, Dec. 2013.

[106] F. Dyson, “Birds and frogs,” Notices of the AMS, vol. 56, pp. 212–223, February
2009.

[107] S. M. Voronin, “Theorem on the Universality of the Riemann Zeta Function,” Izv.
Akad. Nauk SSSR, Ser. Matem., vol. 39, pp. 475–486, 1975. Reprinted in Math.
USSR Izv. 9, 443-445, 1975.

[108] S. C. Woon, “Riemann zeta function is a fractal,” arXiv:chao-dyn/9406003, Jun

1994.

[109] K. Bitar, N. N. Khuri, and H. C. Ren, “Path integrals as discrete sums,” Phys. Rev.
Lett., vol. 67, pp. 781–784, Aug 1991.

[110] H. Wu and D. W. L. Sprung, “Riemann zeros and a fractal potential,” Phys. Rev.
E, vol. 48, pp. 2595–2598, Oct 1993.

[111] M. Dukiewicz, “Zastosowanie Metody Newtona Do Wyznaczania Zer Funkcji Rie-

manna,” Master’s thesis, Department of Physics and Astronomy, University of Wro-
claw, (in Polish), 2006.

[112] E. Elizalde, “Bernhard Riemann, a(rche)typical mathematical-physicist?,” Frontiers

in Physics, vol. 1, no. 11, 2013.

32
[113] J. Littlewood, “Sur la distribution des nombres premieres,” Comptes Rendus,
vol. 158, pp. 1869–1872, 1914.

[114] S. Skewes, “On the difference π(x) − Li(x),” J. London Math. Soc., vol. 8, pp. 277–
283, 1934.

[115] S. Skewes, “On the difference π(x) − Li(x) II,” Proc. London Math. Soc., vol. 5,
pp. 48–70, 1955.

[116] C. Bays and R. Hudson, “A new bound for the smallest x with π(x) − Li(x),”
Mathematics of Computation, vol. 69, pp. 1285–1296, 2000.

[117] Y. Saouter and P. Demichel, “A sharp region where π(x) − li(x) is positive.,” Math.
Comput., vol. 79, no. 272, pp. 2395–2405, 2010.

[118] J. Pintz, “An effective disproof of the Mertens conjecture,” Asterisque, vol. 147-148,
pp. 325–333, 1987.

[119] T. Kotnik and H. te Riele, “The Mertens conjecture revisited,” in 7-th ANTS,
vol. 4076 of Lecture Notes in Computer Science, pp. 156–167, 2006.

33