Solving the Riemann Hypothesis with Green’s Function

and a Gelfand Triplet

Frederick Moxley

To cite this version:

Frederick Moxley. Solving the Riemann Hypothesis with Green’s Function and a Gelfand Triplet.
2018. <hal-01804653>

HAL Id: hal-01804653

https://hal.archives-ouvertes.fr/hal-01804653
Submitted on 1 Jun 2018

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci-
entific research documents, whether they are pub-
lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
 Solving the Riemann Hypothesis with Green’s Function and a Gelfand Triplet

Frederick Ira Moxley III1

1
Hearne Institute for Theoretical Physics, Department of Physics & Astronomy,
Louisiana State University, Baton Rouge, Louisiana 70803-4001, USA
The Hamiltonian of a quantum mechanical system has an affiliated spectrum. If this spectrum is
the sequence of prime numbers, a connection between quantum mechanics and the nontrivial zeros
of the Riemann zeta function can be made. In this case, the Riemann zeta function is analogous
to chaotic quantum systems, as the harmonic oscillator is for integrable quantum systems. Such
quantum Riemann zeta function analogies have led to the Bender-Brody-Müller (BBM) conjecture,
which involves a non-Hermitian Hamiltonian that maps to the zeros of the Riemann zeta function.
If the BBM Hamiltonian can be shown to be Hermitian, then the Riemann Hypothesis follows.
As such, herein we perform a symmetrization procedure of the BBM Hamiltonian to obtain a
Hermitian Hamiltonian that maps to the nontrivial zeros of the analytic continuation of the Riemann
zeta function, and provide an analytical expression for the eigenvalues of the results using Green’s
functions. A Gelfand triplet is then used to ensure that the eigenvalues are well defined. The
holomorphicity of the resulting eigenvalues is demonstrated, and it is shown that that the expectation
value of the Hamiltonian operator is also zero such that the nontrivial zeros of the Riemann zeta
function are not observable. Moreover, a second quantization of the resulting Schrödinger equation
is performed, and a convergent solution for the nontrivial zeros of the analytic continuation of the
Riemann zeta function is obtained. Finally, from the holomorphicity of the eigenvalues it is shown
that the real part of every nontrivial zero of the Riemann zeta function exists at σ = 1/2, and a
general solution is obtained by performing an invariant similarity transformation.

I. INTRODUCTION

The unification of number theory with quantum mechanics has been the subject of many research investigations
[1–5]. It has been proven that an infinitude of prime numbers exist [6]. In Ref. [7], it was shown that the eigenvalues of
a Bender-Brody-Müller (BBM) Hamiltonian operator correspond to the nontrivial zeros of the Riemann zeta function.
If the Riemann Hypothesis is correct [8], the zeros of the Riemann zeta function can be considered as the spectrum of
ˆ + iĤ, where Ĥ is a self-adjoint Hamiltonian operator [5, 9], and Iˆ is identity. Hilbert proposed
an operator R̂ = I/2
the Riemann Hypothesis as the eighth problem on a list of significant mathematics problems [10]. Although the BBM
Hamiltonian is pseudo-Hermitian [11], it is consistent with the Berry-Keating conjecture [12–14], which states that
when x̂ and p̂ commute, the Hamiltonian reduces to the classical H = 2xp. Berry, Keating and Connes proposed the
classical Hamiltonian in an effort to map the Riemann zeros to the Hamiltonian spectrum. More recently, the classical
Berry-Keating Hamiltonians were quantized, and were found to contain a smooth approximation of the Riemann zeros
[15, 16]. This reformulation was found to be physically equivalent to the Dirac equation in Rindler spacetime [17].
Herein, the eigenvalues of the BBM Hamiltonian are taken to be the imaginary parts of the nontrivial zeroes of the
analytical continuation of the Riemann zeta function
∞
1 X (−1)n−1
ζ(s) = · , (1)
1 − 21−s n=1 ns

where the complex number s = σ + it, and <(s) > 0. The idea that the imaginary parts of the zeros of Eq. (1)
are given by a self-adjoint operator was conjectured by Hilbert and Pólya [18]. Hilbert and Pólya asserted that the
nontrivial zeros of Eq. (1) can be considered as the spectrum of a self-adjoint operator in a suitable Hilbert space.
The Hilbert-Pólya conjecture has also found applications in quantum field theories [19]. The Riemann Hypothesis
states that the zeros of Eq. (1) on 0 ≤ σ < 1 have real part equal to 1/2 [8, 20]. In Ref. [21], Hardy proved that
infinitely many zeros are located at σ = 1/2. According to the Prime Number Theorem [22, 23], no zeros of Eq. (1)
can exist at σ = 1. In this paper we present a Schrödinger equation that maps to the nontrivial zeros of the Riemann
zeta function in Sec. II, and evaluate the convergence of the expression by imposing a normalization constraint on the
density. A self-adjoint Hamiltonian is derived from the BBM Hamiltonian using a similarity transformation [24, 25],
and a second quantization of the resulting Schrödinger equation is then performed to obtain the equations of motion.
In Sec. II F, we study the holomorphic eigenvalues of the Riemann zeta function by taking the expectation values of
the resulting Schrödinger equation to show that the real part of every nontrivial zero of the analytic continuation of
the Riemann zeta function exists at σ = 1/2. Finally we obtain a general solution to the Riemann zeta Schrödinger
equation by performing a similarity transformation in Sec. III, and make concluding remarks in Sec. IV.
 2

A. Preliminaries

Definition 1. The complex valued function (eigenstate) φs (x) = φσ (x) + iφt (x) : X → C is measurable if E
is a measurable subset of the measure space X and for each real number r, the sets {x ∈ E : φσ (x) > r} and
{x ∈ E : φt (x) > r} are measurable for σ, t ∈ R.
Definition 2. Let φs be a complex-valued eigenstate on a measure space X, and φs = φσ + iφt , with φσ and φt real.
Therefore, φs is measurable iff φσ and φt are measurable.
Suppose µ is a measure on the measure space X, and E is a measurable subset of the measure space X, and φs is a
complex-valued eigenstate on X. It follows that φs ∈ (H = L (µ)) on E, and φs is complex square-integrable, if φs
is measurable and
Z
| φs | dµ < +∞. (2)
E

Definition 3. The complex valued function (eigenstate) φs = φσ + iφt defined on the measurable subset E is said to
be integrable if φσ and φt are integrable for σ, t ∈ R, where µ is a measure on the measure space X. The Lebesgue
integral of φs is defined by
Z Z Z
φs dµ = φσ dµ + i φt dµ. (3)
E E E

Definition 4. Let X be a measure space, and E be a measurable subset of X. Given the complex eigenstate φs , then
φs ∈ (H = L 2 (µ)) on E if φs is Lebesgue measurable and if
Z
| φs |2 dµ < +∞, (4)
E

such that φs is square-integrable. For φs ∈ (H = L 2 (µ)) we define the L 2 -norm of φs as

Z 1/2
k φ s k2 = | φs |2 dµ , (5)
E

where µ is the measure on the measure space X.

Definition 5. Let X be a measure space, and E be a measurable subset of X. Given the complex eigenstate φs , then
φs ∈ (H = L p (µ)) on E if φs is Lebesgue measurable and if
Z
| φs |p dµ < +∞, (6)
E

such that φs is p-integrable. For φs ∈ (H = L p (µ)) we define the L p -norm of φs as

Z 1/p
k φs kp = | φs |p dµ , (7)
E

where µ is the measure on the measure space X.

Definition 6. A rigged Hilbert space (i.e., a Gelfand triplet [33]) is a triplet (Φ, H , Φ∗ ), where Φ is a dense subspace
of H and Φ∗ is its continuous dual space.

II. RIEMANN ZETA SCHRÖDINGER EQUATION

We consider the eigenvalues of the Hamiltonian

1
Ĥ = (x̂p̂ + p̂x̂)(1 − e−ip̂ ), (8)
1 − e−ip̂
where p̂ = −i~∂x , ~ = 1, and x̂ = x. In Ref. [7], it is conjectured that if the Riemann Hypothesis is correct, the
eigenvalues of Eq. (8) are non-degenerate. Next, we let Ψs (x) be an eigenfunction of Eq. (8) with an eigenvalue
t = i(2s − 1), such that

Ĥ |Ψs (x)i = t |Ψs (x)i , (9)

 3

and x ∈ R+ , s ∈ C. Solutions to Eq. (9) are given by the analytic continuation of the Hurwitz zeta function

|Ψs (x)i = −ζ(s, x + 1)

z s−1 e(x+1)z
I
1
= −Γ(1 − s) dz, (10)
2πi C 1 − ez

on the positive half line x ∈ R+ with eigenvalues i(2s − 1), s ∈ C, <(s) ≤ 1, the contour C is a loop around the
negative real axis, and Γ is the Euler gamma function for <(s) > 0
Z ∞
Γ(s) = xs−1 e−x dx. (11)
0

As − |Ψs (x = 1)i is 1 − ζ(s∗ ), this implies that s belongs to the discrete set of nontrivial zeros of the Riemann zeta
function when s∗ = σ − it and σ = 1/2, and as − |Ψs (x = −1)i is ζ(s), this implies that s belongs to the discrete set
of nontrivial zeros of the Riemann zeta function when s = σ + it and σ = 1/2.
Remark 1. Solutions to Eq. (9) are symmetric about the origin, i.e., x ∈ (−∞, −1] ∪ [1, ∞), and subject to the
singularity at φs (x = 0) = 0 [27].
From inserting Eq. (9) into Eq. (8), we have the relation
1
(x̂p̂ + p̂x̂)(1 − e−ip̂ ) |Ψs (x)i = t |Ψs (x)i . (12)
1 − e−ip̂
Given that Eq. (8) is not Hermitian, it is useful to symmetrize the system. This can be accomplished by letting

|φs (x)i = [1 − exp(−∂x )] |Ψs (x)i ,

= ∆ˆ |Ψs (x)i
= |Ψs (x)i − |Ψs (x − 1)i , (13)

and defining a shift operator

ˆ ≡ 1 − exp(−∂x ).
∆ (14)

For s > 0 the only singularity of ζ(s, x) in the range of 0 ≤ x ≤ 1 is located at x = 0, behaving as x−s . More
specifically,
1
ζ(s, x + 1) = ζ(s, x) − , (15)
xs
with ζ(s, x) finite for x ≥ 1 [27]. As such, it can be seen from Eq. (13) that the Berry-Keating eigenfunction [12, 13]
1
|φs (x)i =
xs  
= exp ln(x)(−σ − it)
 
= exp − σ ln(x) − it ln(x))
  
= exp − σ ln(x) cos(t ln(x)) − i sin(t ln(x))
 
= x−σ cos(t ln(x)) − i sin(t ln(x)) . (16)

Furthermore, the distributional orthonormality relation at x = 1 is satisfied such that [30]

hφs |φs0 i = δss0 . (17)

Upon inserting Eq. (13) into Eq. (12) we obtain

−i[x∂x + ∂x x] |φs (x)i = t |φs (x)i . (18)

 4

Let H be a Hilbert space, and from Eq. (18) we have the Hamiltonian operator
h i
Ĥ = −i~ x∂x + ∂x x
h i
= −i~ 2x∂x + 1 , (19)

for x ∈ R acting in H , such that

hĤf, gi = hf, Ĥgi ∀ f, g ∈ D(Ĥ). (20)

Restricting x ∈ R+ , Eq. (19) is then written

√ √
Ĥ = −2i~ x∂x x, (21)

where s ∈ C, and x ∈ R+ . For the Hamiltonian operator as given by Eq. (21), the Hilbert space is H = L p=2 (1, ∞)
[28–30]. We then impose on Eq. (21) the following minimal requirements, such that its domain is not too artificially
restricted.
i Ĥ is a symmetric (Hermitian) linear operator;

ii Ĥ can be applied on all functions of the form

 x2 
g(x, s) = P (x, s) exp − , (22)
2

where P is a polynomial of x and s. Here, it should be pointed out that Ĥ = T̂ + V̂ , and from Eq. (19), it can be
seen that T̂ = −2i~x∂x , V̂ = −i~. From (ii), V̂ g(x, s) must belong to the Hilbert space H = L 2 defined over the
space x. This is guaranteed as

| −i~ |≤ ~, (23)

where ~ is the reduced Planck constant or Dirac constant. The domain DV̂ of the potential energy V̂ consists of all
φ ∈ H for which V̂ φ ∈ H . As such, V̂ is self-adjoint. It is not necessary to specify the domain of Eq. (21), as it is
only necessary to admit that Eq. (21) is defined on a certain DĤ such that (i) and (ii) are satisfied. If we denote by
D1 the set of all functions in Eq. (22), then (ii) implies that DĤ ⊇ D1 . By letting Ĥ1 be the contraction of Ĥ with
domain D1 , i.e., Ĥ is an extension of Ĥ1 , and letting H̃1 be the closure of Ĥ, it can be seen that H̃1 is self-adjoint.
Since Ĥ is symmetric and Ĥ ⊇ Ĥ1 , i.e., Ĥ is an extension of Ĥ1 , it follows that H̃ = H̃1 and Ĥ is essentially
self-adjoint, where H̃ is the unique self-adjoint extension [31]. Other than eigenfunctions φs (x) in configuration space
as seen in Eq. (16), it is useful to represent eigenfunctions in momentum space Φs (p). The transformation between
configuration space eigenfunctions and momentum space eigenfunctions can be obtained via Plancherel transforms
[32], where the one-to-one correspondence φs (x)
Φs (p) is linear and isometric.

A. Green’s function

In order to obtain eigenstates that are orthonormal when x 6= 1, as seen in Eq. (17), we begin by writing Eq. (21)
as the eigenvalue equation
√ √
−2i~ x∂x xφs (x) = tφs (x). (24)

Dividing by −2i~ on both sides and rearranging the terms, we obtain

1 t 1
φ0s + φs = − φs . (25)
x 2i~ 2x
This can be written as

φ0s + k 2 = Q, (26)
 5

where
r
t
k≡ , (27)
2i~x
and
1
Q≡− φs . (28)
2x
Therefore, we can express Eq. (24) as

(∂x + k 2 )φs = Q. (29)

In order to solve an inhomogeneous differential equation such as Eq. (29), we can find a Green’s function that uses a
delta function source, viz.,

(∂x + k 2 )G(x) = δ(x), (30)

where the delta potential is given by

(
∞ x=0
δ(x) =
6 0
0 x=

with
Z ∞
δ(x)dx = 1. (31)
−∞

It then follows from Eq. (30) that we can express φs as an integral to obtain Q(x), i.e.,
Z
φs (x) = G(x − x0 )Q(x0 )dx0 , (32)

and it must satisfy

Z h i
(∂x + k 2 )φs (x) = (∂x + k 2 )G(x − x0 ) Q(x0 )dx0
Z
= δ(x − x0 )Q(x0 )dx0 = Q(x). (33)

In order to obtain the Green’s function G(x) such that a solution to Eq. (30) can be obtained, we take the Fourier
transform which turns the differential equation into an algebraic one, like
Z
1
G(x) = √ exp(iωx)g(ω)dω, (34)
2π
where g(ω) is the projection, and exp(iωx) is the complete basis set. Upon inserting Eq. (34) into Eq. (30), we obtain
Z
1
(∂x + k 2 )G(x) = √ g(ω)(∂x + k 2 ) exp(iωx)dω = δ(x). (35)
2π
However, since

∂x exp(iωx) = iω exp(iωx), (36)

and
Z
1
δ(x) = √ exp(iωx)dω, (37)
2π
Eq. (30) can be expressed as
Z Z
1 2 1
√ (iω + k ) exp(iωx)g(ω)dω = √ exp(iωx)dω, (38)
2π 2π
 6

where
1
g(ω) = √ . (39)
2π(iω + k 2 )

Hence we have poles at

√
k = ± iω. (40)

Now consider the contour integral

Z Z
1 1 exp(izx)
√ f (z)dz = √ dz. (41)
2π C 2π C (iz + k 2 )

Since exp(izx) is an entire function, Eq. (41) has singularities only at the poles, as given in Eq. (40), i.e., z = ik 2 .
As f (z) is

exp(izx) exp(izx) 1
= , (42)
(iz + k 2 ) i (z − ik 2 )

the residue of f (z) at z = ik 2 is

exp(−k 2 x)
Resz=ik2 f (z) = . (43)
i
According to the residue theorem, we then obtain
Z
1 2πi
√ f (z)dz = √ Resz=ik2 f (z)
2π C 2π
√
= 2π exp(−k 2 x) = G(x). (44)

Hence, the most general solution to Eq. (30) is

√ Z  1 
φs (x) = 2π exp(−k 2 x0 ) − φs (x0 ) dx0 . (45)
2x0

From Eq. (16) it can be seen that φs (x0 ) = x−s

0 . As such,

√ Z  x−s−1 
φs (x) = − 2π exp(−k 2 x0 ) 0 dx0
2
exp(−k 2 x0 )
r Z
π
=− dx0
2 xs+1
0
r Z tx0
π exp(− 2i~x )
=− dx0
2 xs+1
0
r Z r Z
π  tx  1
0 π  tx  1
0
=− cos dx0 − i sin dx0 , (46)
2 2~x xs+1
0 2 2~x xs+1
0

Moreover, by using Eq. (16) it can be seen that

Z  tx  1 Z  tx  x−σ  
0 0 0
cos dx0 = cos cos t ln(x 0 ) dx0
2~x xs+1
0 2~x x0
Z  tx  x−σ  
0 0
− i cos sin t ln(x0 ) dx0 , (47)
2~x x0
 7

and
Z  tx  1 Z  tx  x−σ  
0 0 0
sin dx 0 = sin sin t ln(x0 dx0
)
2~x xs+1
0 2~x x0
Z  tx  x−σ  
0 0
+ i sin cos t ln(x0 ) dx0 . (48)
2~x x0
Since φs (x) = φσ (x) + iφt (x), it can be seen that
r Z
π  tx  x−σ   r Z
π  tx  x−σ  
0 0 0 0
φσ (x) = − cos cos t ln(x0 ) dx0 − sin sin t ln(x0 ) dx0
2 2~x x0 2 2~x x0
r Z
π  
=− x−σ−1
0 cos ix0 k 2 − t log(x0 ) dx0
2
r Z  
π    
=− x−σ−1
0 cosh k 2
x 0 cos(t log(x 0 )) + i sinh k 2
x0 sin(t log(x 0 )) dx0 , (49)
2
and
r
π
Z  tx  x−σ   r Z
π  tx  x−σ  
0 0 0 0
φt (x) = cos sin t ln(x0 ) dx0 − sin cos t ln(x0 ) dx0
2 2~x x0 2 2~x x0
rZ
π  
=− x−σ−1
0 sin ix0 k 2 − t log(x0 ) dx0
2
r Z  
π −σ−1

2
 
2

=− x0 − cosh k x0 sin(t log(x0 )) + i sinh k x0 cos(t log(x0 )) dx0 . (50)
2
Here, we can use the identities
  1 1
cos t log(x0 ) = x−it + xit , (51)
2 0 2 0
and
  i i
sin t log(x0 ) = x−it − xit , (52)
2 0
2 0
to rewrite Eqs. (49)-(50) as
∞
r Z
π  
φσ (x) = − x−σ−1
0 cos t log(x 0 ) exp(−k 2 x0 )dx0
2 −∞
r Z ∞
1 π 2
 
=− e−k x0
1 + x2it
0 x0−σ−it−1 dx0 , (53)
2 2 −∞

and
∞
r Z
π  
φt (x) = x−σ−1
0 sin t log(x 0 ) exp(−k 2 x0 )dx0
2 −∞
r Z ∞
1 π 2
 
=− i e−k x0 −1 + x2it
0 x−σ−it−1
0 dx0 . (54)
2 2 −∞

Taking φs (x) = φσ (x) + iφt (x), we arrive at the expression using Eq. (27)
r Z ∞
π 2

φs (x) = −e−k x0 x−σ−it−1
0 dx0
2 −∞
pπ ∗ − 12 π(t+3iσ)


1 (σ+it)
2 (k − ik) e e2πt − e2iπσ −k 4 2 Γ(−it − σ)
= ∗
2k
√ −σ−it− 3 − 1 π(t+3iσ) 2πt
 q   2  12 (σ+it)
2
t − x xt 2 Γ(−it − σ) xt 2
2iπσ


π2 2e 2 e −e
= = 0 ∀ x ∈ R+
≥1 . (55)
t
 8

Hence, the nontrivial zeros of the Riemann zeta function can be considered as the spectrum of an operator R̂ =
ˆ + iĤ, where Ĥ is a self-adjoint Hamiltonian operator [5, 9], and Iˆ is identity, such that
I/2

ˆ + i hĤi
hR̂i = I/2
ˆ
= I/2 (56)

and the eigenvalues hĤi = t are not observable, as seen from Eq. (55).

B. Measure

Theorem 1. The eigenstate φs (x) = x−s : X → C is measurable. That is, φs (x) = φσ (x) + iφt (x) where φσ , φt :
E → (−∞, −1] ∪ [1, ∞) are measurable for s = σ + it and σ, t ∈ R.
Proof. Owing to the one-to-one correspondence obtained from Plancherel transforms between configuration space and
momentum space eigenstates, it can be seen that
Z ∞
1
Φs (p) = √ φs (x) exp(−ipx)dx
2π −∞
1  1 
s−1
= √ exp − iπs (sgn(p) + 1) sin(πs)Γ(1 − s) |p| , 0 < <(s) < 1. (57)
2π 2

and
Z ∞
1
φs (x) = √ Φs (p) exp(ipx)dp. (58)
2π −∞

Since
Z −1 Z ∞ Z −1 Z ∞
k φs k1 ≡ | φs (x) | dx + | φs (x) | dx = | Φs (p) | dp + | Φs (p) | dp ≡k Φs k1 , (59)
−∞ 1 −∞ 1

from which
1 1 q p
k Φs k1 =k φs k1 = − 1/2
exp π=(s) sin2 (πs) Γ(1 − s)2 . (60)
sπ 2
It then follows that φs is complex square-integrable, i.e.,
Z
φs (x) ∈ H ⇐⇒ |φs (x)|dµ < +∞. (61)
E

Theorem 2. Let the complex valued eigenstate φs (x) = φσ (x) + iφt (x) = x−s where s = σ + it, and let the measurable
subset E → [1, ∞). The H = L 2 -norm of the complex-valued eigenstate φs = x−s is ∞, i.e., φs is not p = 2
integrable at σ = 1/2.
Proof. Owing to the one-to-one correspondence obtained from Plancherel transforms between configuration space and
momentum space eigenstates, it can be seen that
Z ∞
1
Φs (p) = √ φs (x) exp(−ipx)dx
2π −∞
1  1 
s−1
= √ exp − iπs (sgn(p) + 1) sin(πs)Γ(1 − s) |p| , 0 < <(s) < 1. (62)
2π 2

and
Z ∞
1
φs (x) = √ Φs (p) exp(ipx)dp, (63)
2π −∞
 9

where
   t 
φσ (x) = (x2 )−σ/2 exp t · arg(x) cos σ · arg(x) + log(x2 ) , (64)
2
and
   t 
φt (x) = −(x2 )−σ/2 exp t · arg(x) sin σ · arg(x) + log(x2 ) (65)
2
for x ∈ R+
≥1 . Since
Z ∞  p1
k φs k p = | φs (x) |p dx , (66)
1

and[40]
Z ∞  p1
k Φs kp = | Φs (p) |p dp , (67)
1

from which
 1  p1
k Φs kp =k φs kp = . (68)
pσ − 1
It then follows that as σ → 1/2,
 1  p1
k Φs kp =k φs kp = p , (69)
2 −1

such that the L p=2 -norm of φs is of indeterminant form. Furthermore, it can be seen from
 1  p1
lim p , (70)
2 −1
p→2

and letting
 1  p1
y= p , (71)
2 −1
then
1  1 
ln(y) = ln p
p 2 −1
1 p 
= ln(1) − ln −1
p 2
1 p 
= − ln −1 , (72)
p 2
and
 1 p 
lim ln(y) = lim − ln −1
p→2 p→2 p 2
= ∞. (73)
Exponentiating both sides, we obtain
h i h  i
exp lim ln(y) = lim exp ln(y)
p→2 p→2
= lim y = exp(∞) = ∞, (74)
p→2

such that we obtain the infinite density [24]

k Φs kp=2 =k φs kp=2 = ∞. (75)
 10

Corollary 1. Let H = L 2 [1, ∞) and consider the Hamiltonian observable given by

√ √
Ĥφs (x) = −2i~ x∂x xφs (x). (76)

Although the action of Ĥ is in principle well-defined for all φs (x) ∈ L 2 , there are functions which are in L 2 , but for
which Ĥφs (x) is no longer an element of L 2 , e.g., when σ = 1/2,

 

et arg(x) cos arg(x)
2 + 1
2 t log x 2
ie t arg(x)
sin arg(x)
2 + 1
2 t log x 2

φ 21 +it (x) = √
4
− √
4
. (77)
x 2 x2

Therefore the domain of Ĥ is given by

Z −1  ∞
√ √ √ √
n 2 Z  2 o
D(Ĥ) = φs (x) ∈ L 2 :  − 2i~ x∂x xφs (x) dx +  − 2i~ x∂x xφs (x) dx < ∞ ⊂ L 2 . (78)
   
−∞ 1

Similarly, the domain of Ĥ 2 is

n Z −1  2 Z ∞  2 o
D(Ĥ 2 ) = φs (x) ∈ L 2 :  − 4~2 x∂x2 xφs (x) dx +  − 4~2 x∂x2 xφs (x) dx < ∞ ⊂ D(Ĥ), (79)
   
−∞ 1

etc. As such, we define

∞
\
Φ≡ D(Ĥ n ), (80)
n=0

such that for every φs (x) ∈ Φ, the solution is well-defined at σ = 1/2.

Eqs. (57) and (58) are two vector representations of the same Hilbert space H = L p=2 (1, ∞). From Eq. (19), it
can be seen that

T̂ = −2i~x∂x , (81)

such that we define a multiplicative operator T̂0 in momentum space (T̂0 Φs )(p) = T̂0 (p)Φs (p), where

T̂0 (p) = 2x̂p̂. (82)

Here, it should be pointed out that as x̂ = i~d/dp, Eq. (82) reduces to

T̂0 (p) = 2i~, (83)

and Eq. (19) is then rewritten in momentum space as Ĥ(p) = i~. The domain D0 of T̂0 is defined as the set of all
functions Φs (p) ∈ H such that T̂0 (p)Φs (p) ∈ H . As such, T̂0 is definitively self-adjoint. From Eq. (22) we have
defined the set D1 of functions in configuration space. From the Plancherel transform [32] of Eq. (22), we obtain the
set D1 of functions in momentum space having the form
 p2 
G(p, s) = P (p, s) exp − , (84)
2
where P is a polynomial of p and s. Eqs. (57) and (58) are true if φs (x) ∈ D1 or Φs (p) ∈ D1 and since Φs (p) ∈ D1 → 0
as p → ∞ then D1 ⊆ D0 . Moreover, for φ ∈ D1 , T̂0 coincides with Eq. (81) [31]. Using Eq. (57) and Ĥ(p) = i~, the
eigenrelation

Ĥ(p) |Φs (p)i = λ |Φs (p)i (85)

is obtained. In order to find the expectation value for Ĥ we take the complex conjugate of Eq. (85), set ~ = 1,
multiply by the eigenfunction Φs (p), and then integrate over p to obtain
Z ∞  − 1 iπs s−1 ∗  − 1 iπs s−1 
e 2 (sgn(p) + 1) sin(πs)Γ(1 − s) |p| e 2 (sgn(p) + 1) sin(πs)Γ(1 − s) |p|
i 1/2 1/2
dp = λ∗ k Φs kp , (86)
−∞ 2π 2π
where λ is the eigenvalue.
 11

Theorem 3. Let the complex valued eigenstate φs (x) = φσ (x) + iφt (x) = x−s where s = σ + it, and let the measurable
subset E → (−∞, −1] ∪ [1, ∞). The following are equivalent for σ, t ∈ R:
1. For each real number r, the set {x ∈ E : φσ (x) > r} is measurable.
2. For each real number r, the set {x ∈ E : φt (x) > r} is measurable.
3. For each real number r, the set {x ∈ E : φσ (x) ≥ r} is measurable.
4. For each real number r, the set {x ∈ E : φt (x) ≥ r} is measurable.
5. For each real number r, the set {x ∈ E : φσ (x) < r} is measurable.
6. For each real number r, the set {x ∈ E : φt (x) < r} is measurable.
7. For each real number r, the set {x ∈ E : φσ (x) ≤ r} is measurable.
8. For each real number r, the set {x ∈ E : φt (x) ≤ r} is measurable.
Proof. Note that the intersection of sets,
∞
\ 1
{x ∈ E : φσ (x) ≥ r} = {x ∈ E : φσ (x) > r − }, (87)
n=1
n

∞
\ 1
{x ∈ E : φt (x) ≥ r} = {x ∈ E : φt (x) > r − }, (88)
n=1
n

∞
\ 1
{x ∈ E : φσ (x) > r} = {x ∈ E : φσ (x) ≥ r + }, (89)
n=1
n

∞
\ 1
{x ∈ E : φt (x) > r} = {x ∈ E : φt (x) ≥ r + }, (90)
n=1
n

where
   t 
φσ (x) = (x2 )−σ/2 exp t · arg(x) cos σ · arg(x) + log(x2 ) , (91)
2
and
   t 
φt (x) = −(x2 )−σ/2 exp t · arg(x) sin σ · arg(x) + log(x2 ) . (92)
2

Theorem 4. Let E → (−∞, −1] ∪ [1, ∞) be a measurable subset of the measure space X. If the complex valued
eigenstate φs (x) = φσ (x) + iφt (x) = x−s where s = σ + it, and φσ (x),and φt are continuous a.e. on E, then φs (x) is
measurable for σ, t ∈ R.
Proof. Let D be the singleton {0} owing to the singularity at x = 0 of φs (x) = x−s . Then µ(D) = 0 and all of its
subsets are measurable. Let r ∈ R and note that

{x ∈ E : φσ (x) > r} = {x ∈ E − D : φσ (x) > r} ∪ {x ∈ D : φσ (x) > r}, (93)

where
   t 
φσ (x) = (x2 )−σ/2 exp t · arg(x) cos σ · arg(x) + log(x2 ) , (94)
2
and
   t 
φt (x) = −(x2 )−σ/2 exp t · arg(x) sin σ · arg(x) + log(x2 ) . (95)
2
 12

Letting
Cσ = {x ∈ E − D : φσ (x) > r}, (96)
for each x ∈ Cσ , as φσ (x) is continuous at x, we can find δx > 0 such that if y ∈ Vδx (x) then φσ (y) > r. It can be
seen that φσ (x) is measurable, since
\
Cσ = (E − D) Vδx (x). (97)
x∈Cσ

Similarly, noting that

{x ∈ E : φt (x) > r} = {x ∈ E − D : φt (x) > r} ∪ {x ∈ D : φt (x) > r}, (98)
and letting
Ct = {x ∈ E − D : φt (x) > r}, (99)
for each x ∈ Ct , as φt (x) is continuous at x, we can find δx > 0 such that if y ∈ Vδx (x) then φt (y) > r. It can be seen
that φt (x) is measurable since
\
Ct = (E − D) Vδx (x). (100)
x∈Ct

Let {φs } = {φσ } + i{φt } be a sequence of functions defined on the measure space X → C. Denoting
sup φs (x) = sup{φs (x) : s ∈ C} (101)
s

and
 
lim sup φs (x) = lim sup φk (x) , (102)
s s k≥s

it can be seen that

 
lim sup φs (x) = inf sup φk (x) . (103)
s s k≥s

Similarly, from
inf φs (x) = inf{φs (x) : s ∈ C} (104)
s

and
 
lim inf φs (x) = lim inf φk (x) , (105)
s s k≥s

it can be seen that

 
inf φs (x) = − sup − φs (x) , (106)
s s

and
 
lim inf φs (x) = − lim sup − φs (x) . (107)
s s

Theorem 5. Let the sequence of measurable eigenstates {φs } = {φσ } + i{φt } be defined on the measure space X → C.
For the sequence of measurable eigenstates {φσ } : E → (−∞, −1] ∪ [1, ∞)
g(x) = sup φσ (x), (108)
σ

and
h(x) = lim sup φσ (x), (109)
σ

such that g and h are measurable for x ∈ E.

 13

Proof. For any r ∈ R, we obtain

[
{x ∈ E : g(x) > r} = {x ∈ E : φσ (x) > r}. (110)
σ

From Eqs. (103) and (106)-(107), this implies that h is also measurable.
Corollary 2. Let φσ be a sequence of measurable eigenstates defined on the measure space X, and φσ : E →
(−∞, −1] ∪ [1, ∞). Since {φσ } converges pointwise to φσ a.e. on E, then φσ is measurable.
Theorem 6. Let the sequence of measurable eigenstates {φs } = {φσ } + i{φt } be defined on the measure space X → C.
For the sequence of measurable eigenstates {φt } : E → (−∞, −1] ∪ [1, ∞)

g(x) = sup φt (x), (111)

t

and

h(x) = lim sup φt (x), (112)

t

such that g and h are measurable for x ∈ E.

Proof. For any r ∈ R, we obtain
[
{x ∈ E : g(x) > r} = {x ∈ E : φt (x) > r}. (113)
t

From Eqs. (103) and (106)-(107), this implies that h is also measurable.
Corollary 3. Let φt be a sequence of measurable eigenstates defined on the measure space X, and φt : E →
(−∞, −1] ∪ [1, ∞). Since {φt } converges pointwise to φt a.e. on E, then φt is measurable.
Corollary 4. Let φs = φσ + iφt be a sequence of measurable eigenstates defined on the measure space X → C.
Since {φσ } converges pointwise to φσ a.e. on E → (−∞, −1] ∪ [1, ∞), and {φt } converges pointwise to φt a.e. on
E → (−∞, −1] ∪ [1, ∞), then φs is measurable.

C. Expectation Value of the Observable

Definition 7. The Riemann zeta Schrödinger equation is

h i
−~∂s |Ψs (x)i = i ∆ ˆ −1 x̂p̂∆
ˆ +∆
ˆ −1 p̂x̂∆
ˆ |Ψs (x)i , (114)

where ∆ˆ = 1 − exp(−∂x ), x̂ = x, p̂ = −i~∂x , ~ = 1, x ∈ R+ ≥ 1 owing to the difference operator ∆

ˆ |Ψs (x)i, and
s ∈ C.
Upon inserting Eq. (13) into Eq. (114) for x ∈ R+ , we obtain the symmetrized Riemann zeta Schrödinger equation,
i.e.,

∂s |φs (x)i = 1/2(∂σ − i∂t ) |φs (x)i

2√ √
= − x∂x x |φs (x)i . (115)
~
Theorem 7. Let the complex-valued eigenstate

√
 q    12 (σ+it)
3 1 t2 t2
π2−σ−it− 2 e− 2 π(t+3iσ) e2πt − e2iπσ


t−x x2 Γ(−it − σ) x2
φs (x) = , (116)
t
where s = σ + it and σ, t ∈ R, and let the measurable subset of the measure space X be E → (−∞, −1] ∪ [1, ∞), for
√ √
the Hamiltonian operator Ĥ = −2i~ x∂x x.
 14

Proof. Let |φs (x)i be an eigenstate of Ĥ with eigenvalue t, i.e.,

Ĥ |φs (x)i = t |φs (x)i . (117)

In order to find the expectation value of Ĥ we multiply Ĥ by the eigenstate, take the complex conjugate, and then
multiply the result by the eigenstate and integrate over E to obtain
√ √ ∗
Z  Z
2i x∂x xφs (x) φs (x)dx = t∗ φ∗s (x)φs (x)dx
E E
= t∗ k φ k . (118)

Integrating by parts on the LHS then gives

Z −1 Z ∞

∗ d d 
−2i k φ k + φs (x)x φs (x)dx + φ∗s (x)x φs (x)dx = t∗ k φ k . (119)
−∞ dx 1 dx

Carrying out the integration on the LHS we obtain

Z 0 Z −1 Z 2πn Z ∞
∗ d d
φs (x)x φs (x)dxdt = φ∗s (x)x φs (x)dxdt = 0 ∀ n. (120)
−2πn −∞ dx 0 1 dx

Hence it can be seen that

Z 0 Z −1 Z 2πn Z ∞
φ∗s (x)φs (x)dxdt = φ∗s (x)φs (x)dxdt = 0 ∀ n. (121)
−2πn −∞ 0 1

Theorem 8. Let the complex-valued eigenstate φs (x) = φσ (x)+iφt (x) = x−s where s = σ+it and σ, t ∈ R, and let the
√ √
measurable subset of the measure space X be E → (−∞, −1]∪[1, ∞). For the Hamiltonian operator Ĥ = −2i~ x∂x x,
all of the eigenvalues t occur at | σ |= 1/2 with ~ = 1.

Proof. Let |φs (x)i be an eigenstate of Ĥ with eigenvalue t, i.e.,

Ĥ |φs (x)i = t |φs (x)i . (122)

In order to find the expectation value of Ĥ we multiply Ĥ by the eigenstate, take the complex conjugate, and then
multiply the result by the eigenstate and integrate over E to obtain
√ √ ∗
Z  Z
2i x∂x xφs (x) φs (x)dx = t∗ φ∗s (x)φs (x)dx
E E
= t∗ k φ k . (123)

Integrating by parts on the LHS then gives

Z −1 Z ∞

∗ d d 
−2i k φ k + φs (x)x φs (x)dx + φ∗s (x)x φs (x)dx = t∗ k φ k . (124)
−∞ dx 1 dx

Carrying out the integration on the LHS we obtain

 
2i(−1)−2σ (−1)2σ + 1 (σ + it) = (2σ − 1)(t∗ + 2i) k φ k . (125)

Hence it can be seen that

1
| σ |= ∀ t. (126)
2
 15

0.5
2
|φs(x)|

t
0

−0.5

−1

−1 0 1
x
Figure 1: Plot of s = |σ| exp(it) = 1/2 − log(x)/2, Eq. (127). The density is normalized when x cos(t) = 1 (color online).

D. Convergence

√ √
Theorem 9. For the symmetrized Riemann zeta Schrödinger equation, i.e., ~∂s |φs (x)i = −2 x∂x x |φs (x)i, the
−s
complex-valued eigenstate |φs (x)i = x where s = |σ| exp(it) and σ, t ∈ R normalizes at x cos(t) = 1, i.e., the density
|φs (x)|2 = 1.
Proof. In order to obtain convergent solutions to the unsymmetric Riemann zeta Schrödinger Eq. (114), it can be
seen that upon inserting Eq. (13) into the symmetric Eq. (115), we obtain

s = |σ| exp(it)
1 log(x)
= − . (127)
2 2
Hence, at x = sec(t),
1  
σ = ± e−it log(sec(t)) − 1 , (128)
2
such that at |σ| = 1/2 in agreement with Eq. (126) for

t = 2πn, (129)
 16

1 1
Im1 - ζ - ⅈ t
2
t 1
20 40 60 80 100 Imζ + ⅈ t
2
-1

-2

-3
Figure 2: Plot of the imaginary components of Eq. (1). Results are compared with Eq. (137) (color online).

where n ∈ Z and t ∈ R. This condition is required such that the density is normalized in agreement with Eq. (75),
i.e.,
XX
k φs k2 = b̂n (s)b̂†m (s) hφm |φn i
m n
X
= |b̂n (s)|2
n
= 1. (130)

Theorem 10. For the Bender-Brody-Müller equation [7], i.e.,

1
(x̂p̂ + p̂x̂)(1 − e−ip̂ ) |Ψs (x)i = t |Ψs (x)i , (131)
1 − e−ip̂
the nontrivial zeros of the RiemannPzeta function can be obtained from the analytic continuation of the Riemann zeta
∞
function, i.e. ζ(s) = (1 − 21−s )−1 n=1 (−1)n−1 n−s at the normalization constraint x = sec(t = 2πn) = 1, such that
|σ| = 1/2 ∀ t ∈ R where s = σ + it and σ, t ∈ R. The nontrivial zeros of the Riemann zeta function are not observable
at |σ| = 1/2, ∀ n ∈ Z.
1
Proof. At x = sec(t = 2πn) = 1, the normalization constraint Eq. (130) is satisfied, σ = 2 − it, and Eq. (10) can be
written
Ψs (x = 1) = −ζ(s = 1/2, 2)
I √ 2z
1 ze
= −Γ(1/2) dz
2πi C 1 − ez
1
= 1 − ζ(σ = − it). (132)
2
where the contour C is about R− . From the analytic continuation relations of Eq. (1)
 
∞ ∞ (−1)n−1 exp − i · t ln(n)
1 X (−1)n−1 1 X
=
1 − 21−s n=1 ns 1 − 21−s n=1 nσ
   
∞ (−1)n−1 cos t · ln(n) ∞ (−1)n−1 sin t · ln(n) i
1 h X X
= −i , (133)
1 − 21−s n=1 nσ n=1
nσ
 17
 
∞ ∞ (−1)n−1 exp i · t ln(n)
1
 X (−1)n−1 ∗ 1 X
1− = 1−
1 − 21−s n=1 ns 1 − 21−s∗ n=1 nσ
 
∞ (−1)n−1 cos t · ln(n)
1 h X
= 1−
1 − 21−s∗ n=1 nσ
 
X∞ (−1)n−1 sin t · ln(n) i
+ i . (134)
n=1
nσ

   
1 X∞
(−1)n−1 X∞
(−1)n−1 −2−σ+1 cos t log(2) cos t ln(n)
= · i2
1 − 21−s n=1 ns nσ
  h 
n=1 2−2σ+2 sin2 t log(2) + 1 − 2−σ+1 cos t log(2)
 
∞ n−1 cos t ln(n)
X (−1)
+ · i2
nσ
  h 
n=1 2−2σ+2 sin2 t log(2) + 1 − 2−σ+1 cos t log(2)
   
−σ+1
X ∞
(−1) n−1 −2 sin t log(2) sin t ln(n)
+ · i2
nσ
  h 
n=1 2−2σ+2 sin2 t log(2) + 1 − 2−σ+1 cos t log(2)
   
−σ+1
X∞
(−1) n−1 −2 sin t log(2) cos t ln(n)
+ i · i2
nσ
  h 
n=1 2−2σ+2 sin2 t log(2) + 1 − 2−σ+1 cos t log(2)
   
∞ −σ+1
X (−1) n−1 2 cos t log(2) sin t ln(n)
+ i · i2
nσ
  h 
n=1 2−2σ+2 sin2 t log(2) + 1 − 2−σ+1 cos t log(2)
 
X∞
(−1)n−1 − sin t ln(n)
+ i · i2 , (135)
nσ
  h 
n=1 2−2σ+2 sin2 t log(2) + 1 − 2−σ+1 cos t log(2)

   
 1 X∞
(−1)n−1 ∗ X∞
(−1)n−1 2−σ+1 cos t log(2) cos t ln(n)
1− = 1 + · i2
1 − 21−s n s nσ
  h 
n=1 n=1 2−2σ+2 sin2 t log(2) + 1 − 2−σ+1 cos t log(2)
 
X ∞
(−1) n−1 − cos t ln(n)
+ · i2
nσ
  h 
n=1 2−2σ+2 sin2 t log(2) + 1 − 2−σ+1 cos t log(2)
   
−σ+1
X ∞
(−1) n−1 −2 sin t log(2) sin t ln(n)
+ · i2
nσ
  h 
n=1 2−2σ+2 sin2 t log(2) + 1 − 2−σ+1 cos t log(2)
   
−σ+1
X∞
(−1) n−1 −2 sin t log(2) cos t ln(n)
+ i · i2
nσ
  h 
n=1 2−2σ+2 sin2 t log(2) + 1 − 2−σ+1 cos t log(2)
   
X∞
(−1)n−1 2−σ+1 cos t log(2) sin t ln(n)
+ i · i2
nσ
  h 
n=1 2−2σ+2 sin2 t log(2) + 1 − 2−σ+1 cos t log(2)
 
X∞
(−1)n−1 − sin t ln(n)
+ i · i2 , (136)
nσ
  h 
n=1 2−2σ+2 sin2 t log(2) + 1 − 2−σ+1 cos t log(2)
 18

such that Owing to the periodicity of t = 2πn at x = sec(t), i.e. Eq. (129), it can be seen that
∞ ∞
h 1 X (−1)n−1 i h  1 X (−1)n−1 ∗ i
= 1−s s
= = 1 − 1−s
. (137)
1−2 n=1
n 1−2 n=1
ns

Owing to Eq. (126), at |σ| = 1/2 we obtain

  √ 

n
(−1)n−1 sin t ln(n) − 2 sin t log
∞
2
h i X
= ζ(s) = i √ · √   . (138)
n=1
n 2 2 cos t log(2) − 3

However, since at |σ| = 1/2 the eigenvalues t are not observable, i.e., hĤi = t = 0, we have
 0
 √  0

n
h i ∞
X (−1) n−1 sin  2πn
:ln(n)

 − 2 sin 2πn
:log


2
= ζ(s) = i √ · √  0
 = 0 ∀ n ∈ Z. (139)
n
n=1 2 2 cos 2πn
:log(2)
 −3

Remark 2. It has been noted that there is a uniquely defined relation between prime numbers and the imaginary parts
of the nontrivial Riemann zeros, independent of their real part [36].

E. Second Quantization

Theorem 11. By representing the complex-valued eigenstate |φs (x)i = |φσ (x)i + i |φt (x)i = x−s where s = |σ|
√exp(it)
√
and σ, t ∈ R as a linear combination of basis states, then the eigenspectrum of the Hamiltonian operator −2i~ x∂x x
is not observable, i.e. zero, on the measure space E → (−∞, −1] ∪ [1, ∞) when |σ| = 1/2 and ~ = 1.
Proof. A standard way to introduce topology into the algebra of observables is to make them operators on a Hilbert
space. In order to perform a second quantization [35], we can express the complex-valued eigenstate as a linear
combination of basis states
X
|φs (x)i = b̂n (s) |φn (x)i , (140)
n∈Z

where s = |σ| exp(it) ∈ C, and σ, t ∈ R. As such, using Eq. (16) we can rewrite Eq. (140) as
X
|φs (x)i = b̂n (s)x−n . (141)
n∈Z

From using this second quantization in Eq. (115), we find

d
~ b̂n (s) = −tn b̂n (s). (142)
ds
We now find a Hamiltonian that yields Eq. (142) as the equation of motion, hence, we take
Z ∞ √ √
Z −1 √ √
hφs0 (x)| Ĥ |φs (x)i = −2 hφs0 (x)| x∂x x |φs (x)i dx − 2 hφs0 (x)| x∂x x |φs (x)i dx, (143)
1 −∞

as the expectation value. Upon substituting Eq. (141) into Eq. (143), we obtain the harmonic oscillator
X XZ ∞ √ √ X XZ −1 √ √
1 1 1 1
hφm (x)| Ĥ |φn (x)i = −2 1 x∂x x 1 dx − 2 1 x∂x x 1 dx
1 x 2 −im x 2 +in −∞ x 2 −im x 2 +in
m∈Z n∈Z m∈Z n∈Z
XX  2n(exp(π(n − m)) − 1) 
= b̂†m (s)b̂n (s) hm| |ni , (144)
m−n
m∈Z n∈Z
 19

for |mi , |ni = 1, 2, 3, . . . , ∞. Hence at m = n, hn|ni = δnn = 1 and

X  
hφn (x)| Ĥ |φn (x)i = |b̂n (s)|2 − 2πn . (145)
n∈Z

In accordance with Eq. (126) and Eq. (130), at |σ| = 1/2 and the zero periodicity of the eigenvalues t,

hφn (x)| Ĥ |φn (x)i = 0. (146)

Taking b̂n (s) as an operator, and b̂†n (s) as the adjoint, we obtain the usual properties:

[b̂n (s), b̂m (s)] = [b̂†n (s), b̂†m (s)] = 0,

[b̂n (s), b̂†m (s)] = δnm . (147)

From the analogous Heisenberg equations of motion,

d X
−~ b̂n (s) = [b̂n (s), Ĥ]−
ds
n∈Z
X  
= Em b̂n (s)b̂†m (s)b̂m (s) − b̂†m (s)b̂m (s)b̂n (s)
m∈Z
X  
= Em δnm b̂m (s) − b̂†m (s)b̂n (s)b̂m (s) − b̂†m (s)b̂m (s)b̂n (s)
m∈Z
X  
= Em δnm b̂m (s) + b̂†m (s)b̂m (s)b̂n (s) − b̂†m (s)b̂m (s)b̂n (s)
m∈Z
X
= b̂†m (s)b̂n (s)tn . (148)
n∈Z

The eigenvalues of Ĥ are then unobservable, i.e.,

hφn (x)| Ĥ |φn (x)i = 0. (149)

From Eq. (148) it can be seen that

d
−~ b̂n = 0,
ds
d
−~ b̂†m = −0. (150)
ds
Remark 3. Theorem 11 implies the Riemann hypothesis, as the spectrum of a Hermitian operator consists of real
numbers as seen in Theorem 7, and 0 is a real number.

F. Holomorphicity

√ √
Theorem 12. The densely defined Hamiltonian operator Ĥ = −2 x∂x x on the Hilbert space H = L2 [1, ∞) is
symmetric (Hermitian) [34], for the complex-valued eigenstate |φs (x)i = |φσ (x)i+i |φt (x)i = x−s where s = |σ| exp(it)
and σ, t ∈ R when |σ| = 1/2 and ~ = 1.
Proof. By expressing the complex-valued eigenstate as a linear combination of basis states such that
X
|φs (x)i = b̂n (s) |φn (x)i , (151)
n∈Z

where s = |σ| exp(it) ∈ C, and σ, t ∈ R, it can be seen that by using Eq. (16) we can rewrite Eq. (151) as
X
|φs (x)i = b̂n (s)x−n . (152)
n∈Z
 20

By taking the inner product

X XZ ∞ √ √ X XZ −1 √ √
1 1 1 1
(Ĥφ∗n , φm ) = −2 1 x∂x x 1
−in
dx − 2 1 x∂x x 1 dx
1 x 2 +im x2 −∞ x 2 +im x 2 −in
m∈Z n∈Z m∈Z n∈Z
XX  2n(exp(π(n − m)) − 1) 
= b̂†m (s)b̂n (s) hm| |ni , (153)
m−n
m∈Z n∈Z

for |mi , |ni = 1, 2, 3, . . . , ∞. Hence at m = n, hn|ni = δnn = 1 and

X  
hφn (x)| Ĥ |φn (x)i = |b̂n (s)|2 2πn . (154)
n∈Z

In accordance with Eq. (126) and Eq. (130), at |σ| = 1/2 and the zero periodicity of the eignenvalues t,

hφn (x)| Ĥ |φn (x)i = 0. (155)

Furthermore, by taking the inner product

X XZ ∞ 1 √ √ 1 X X Z −1 1 √ √ 1
(φ∗m , Ĥφn ) = −2 1
−im
x∂ x x 1
+in
dx − 2 1
−im
x∂x x 1 +in dx
m∈Z n∈Z 1
x 2 x 2
m∈Z n∈Z −∞
x 2 x 2

XX  2n(exp(π(n − m)) − 1) 
= b̂†m (s)b̂n (s) hm| |ni , (156)
m−n
m∈Z n∈Z

for |mi , |ni = 1, 2, 3, . . . , ∞. Hence at m = n, hn|ni = δnn = 1 and

X  
hφn (x)| Ĥ |φn (x)i = |b̂n (s)|2 − 2πn . (157)
n∈Z

In accordance with Eq. (126) and Eq. (130), at |σ| = 1/2,

hφn (x)| Ĥ |φn (x)i = 0. (158)

Finally,

(Ĥφ∗n , φm ) = (φ∗m , Ĥφn ) = 2πn = 0 ∀ n ∈ Z. (159)

Remark 4. The Riemann Hypothesis states that the real part of all of the nontrivial zeros of the Riemann zeta
function are located at σ = 1/2 [8].

III. SIMILARITY SOLUTIONS

Since Eq. (115), the Riemann zeta Schrödinger equation (RZSE) possesses symmetry about the origin x = 0, we
then seek a similarity solution [38] of the form:

φs (x) = xα f (η), (160)

where η = s/xβ , and the RZSE becomes an ordinary differential equation (ODE) for f . As such, we consider Eq.
(115), and introduce the transformation ξ = a x, and τ = b s, so that

w(ξ, τ ) = c φ(−a ξ, −b τ ), (161)

where  ∈ R, and τ ∈ C.
 21

From performing this change of variable we obtain

∂ ∂w ∂τ
φ = −c
∂s ∂τ ∂s
b−c ∂w
=  , (162)
∂τ
and
√
√ ∂ √ √ ∂ x √ ∂φ 
−2 x xφ = −2 x φ+ x
∂x ∂x ∂x
√ 1 √ √ ∂φ
= −2 x √ φ − 2 x x
2 x ∂x
∂φ
= −φ − 2x , (163)
∂x
where
∂φ ∂w ∂ξ
= −c
∂x ∂ξ ∂x
∂w
= a−c . (164)
∂ξ
By using Eqs. (162)-(164) in Eq. (115), the RZSE is then written
h ∂w ∂w i
−c b + w + 2ξ = 0, (165)
∂τ ∂ξ

and is invariant under the transformation ∀  if b = 2, i.e.,

h b  ∂w ∂w  ∂w i
−c −i + w + 2ξ = 0, (166)
2 ∂τ< ∂τ= ∂ξ
and
log(2) + 2iπn
b = , ∀ n ∈ Z. (167)
log()

Therefore, it can be seen that since φ solves the RZSE for x and s, then w = −c φ solves the RZSE at x = −a ξ, and
s = −b τ . We now construct a group of independent variables such that
ξ a x
=
τ a/b (b s)a/b
x
= a/b
s
= η(x, s), (168)

and the similarity variable is then

a log()
η(x, s) = xs− log(2)+2iπn . (169)

Also,
w c φ
=
τ c/b (b s)c/b
φ
= c/b
s
= ν(η), (170)

suggesting that we seek a solution of the RZSE with the form

c log()
φs (x) = s log(2)+2iπn ν(η). (171)
 22

Since the RZSE is invariant under the transformation, it is to be expected that the solution will also be invariant
under the variable transformation. Taking a = c = log−1 (), the partial derivatives transform like

∂ ∂  log(2)+2iπn
1
  1
 ∂η
φs (x) = s ν(η) + s log(2)+2iπn ν 0 (η)
∂s ∂s ∂s
1
−1+ log(2)+2iπn
s h i
= ν(η) − ν 0 (η) , (172)
log(2) + 2iπn

and
∂  1
 ∂η
φs (x) = s log(2)+2iπn ν 0 (η)
∂x ∂x
= ν 0 (η), (173)

where
∂η s−1
= − , (174)
∂s 2iπn + log(2)

and
∂η 1
= s− 2iπn+log(2) . (175)
∂x
The RZSE then reduces to the ODE
h i h i
s−1 + log(2) + 2iπn ν(η) + − s−1 + 2 log(2)η + 4iπnη ν 0 (η) = 0, ∀ n ∈ Z. (176)

A. General Solution

The homogenous linear differential Eq. (176) is separable [39], viz.,

dν 2iπn + s−1 + log(2)

= −1 dη. (177)
ν s − 4iπnη − η log(4)

Integrating on both sides, we obtain

   
2iπn + s−1 + log(2) log s−1 − 4iπnη − η log(4)
ln |ν| = c1 − . (178)
4iπn + log(4)

Exponentiating both sides,

−1 +log(2)
 − 2iπn+s
4iπn+log(4)
|ν| = exp(c1 ) s−1 − 4iπnη − η log(4) . (179)

Renaming the constant exp(c1 ) = C and dropping the absolute value recovers the lost solution ν(η) = 0, giving the
general solution to Eq. (176)
−1 +log(2)
 − 2iπn+s
4iπn+log(4)
νn (η) = C s−1 − 4iπnη − η log(4) , ∀ n ∈ Z, ∀ C ∈ R. (180)

By setting C = 1, and using Eqs. (169) and (171) in Eq. (180), we obtain the general solution to the RZSE Eq.
(115), written
 − 2πns+is log(2)+i
1 1 1
  4πns−is log(4)
φs (x) = s log(2)+2iπn + s− log(2)+2iπn − x log(4) − 4iπnx , ∀ n ∈ Z. (181)
s
 23

IV. CONCLUSION

In this study, we have discussed the convergence of the real part of every nontrivial zero of the analytic continuation
of the Riemann zeta function. This was accomplished by developing a Riemann zeta Schrödinger equation and compar-
ing it with the Bender-Brody-Müller conjecture in both configuration space and momentum space. A symmetrization
procedure was implemented to study the convergence of the system, and the expectation values were calculated from
the resulting system to study the nontrivial zeros of the analytic continuation of the Riemann zeta function. It was
found using Green’s functions that the expectation value of the Hamiltonian operator for the eigenstates along the
critical line σ = 1/2 is also zero such that the nontrivial zeros of the Riemann zeta function are not observable. A
Gelfand triplet was implemented to ensure that the eigenvalues are well defined. Moreover, a second quantization pro-
cedure was performed for the Riemann zeta Schrödinger equation to obtain the equations of motion and an analytical
expression for the eigenvalues. It was also demonstrated that the eigenvalues are holomorphic across the measurable
subspace of the measure space. A normalized convergent expression for the analytic continuation of the nontrivial
zeros of the Riemann zeta function was obtained, and a convergence test for the expression was performed demon-
strating that the real part of every nontrivial zero of the Riemann zeta function exists at σ = 1/2. Finally, a general
solution to the Riemann zeta Schrödinger equation was found from performing an invariant similarity transformation.
 24

[1] H.C. Rosu, Modern Physics Letters A 18(18), 1205-1213 (2003).

[2] D. Schumayer, B.P. van Zyl, and D.A. Hutchinson, Physical Review E 78(5), 056215 (2008).
[3] B.L. Julia, Physica A: Statistical Mechanics and its Applications 203(3 − 4), 425-436 (1994).
[4] D. Spector, Journal of Mathematical Physics 39(4), 1919-1927 (1998).
[5] D. Schumayer and D.A. Hutchinson, Reviews of Modern Physics 83(2), 307 (2011).
[6] P. Ribenboim, The little Book of Big Primes, (1991).
[7] C.M. Bender, D.C. Brody and M.P. Müller, Physical Review Letters 118(13) 130201 (2017).
[8] B. Riemann, On the Number of Prime Numbers less than a Given Quantity.
[9] W.G. Faris and R.B. Lavine, Communications in Mathematical Physics 35(1), 39-48 (1974).
[10] D. Hilbert, Bulletin of the American Mathematical Society 8(10), 437-479 (1902).
[11] C.M. Bender, S. Boettcher and P.N. Meisinger, Journal of Mathematical Physics 40(5) 2201-2229 (1999).
[12] M.V. Berry and J.P. Keating, In Supersymmetry and Trace Formulae, Springer US,355-367 (1999).
[13] M.V. Berry and J.P. Keating, SIAM review 41(2), 236-266 (1999).
[14] A. Connes, Selecta Mathematica, New Series 5(1) 29-106 (1999).
[15] G. Sierra and J. Rodriguez-Laguna, Physical review letters 106(20) 200201 (2011).
[16] M.V. Berry and J.P. Keating, Journal of Physics A: Mathematical and Theoretical 44(28) 285203 (2011).
[17] G. Sierra, Journal of Physics A: Mathematical and Theoretical 47(32) 325204 (2014).
[18] A.M. Odlyzko, Contemporary Mathematics 290 139 (2001).
[19] J.C. Andrade, International Journal of Modern Physics A 28 50072 (2013).
[20] E.C. Titchmarsh and D.R. Heath-Brown, Oxford University Press (1986).
[21] G.H. Hardy, CR Acad. Sci. Paris 158, 1012-1014 (1914).
[22] J. Hadamard, Bull. Soc. Math. France 24 199 (1896).
[23] C. J. de la Vallée-Poussin, Ann. Soc. Sci. Bruxelles 20, 183 (1896).
[24] B.F. Samsonov, Phil. Trans. R. Soc. A, 371(1989), 20120044 (2013).
[25] A. Mostafazadeh, Journal of Mathematical Physics, 43(8), 3944-3951 (2002).
[26] Whittaker, E.T. and Watson, G.N., 1996. A Course of Modern Analysis. Cambridge University Press.
[27] O. Espinosa and V.H. Moll, The Ramanujan Journal 6(2), 159 (2002).
[28] G. Sierra, Nuclear Physics B 776(3) 327-364 (2007).
[29] J. Twamley and G.J.Milburn, New Journal of Physics 8(12), 328 (2006).
[30] S. Endres and F. Steiner, Journal of Physics A: Mathematical and Theoretical 43(9) 095204 (2010).
[31] T. Kato, Transactions of the American Mathematical Society 70(2) 195 (1951).
[32] M. Plancherel and M. Leffler, Rendiconti del Circolo Matematico di Palermo 30(1) 289 (1884).
[33] R. de la Madrid, Journal of Physics A: Mathematical and General 35(2) 319 (2002).
[34] Reed, M. and Simon, B., Methods of Modern Mathematical Physics: Functional analysis (1) (1980).
[35] J.M. Cook, Transactions of the American Mathematical Society 74(2) 222 (1953).
[36] S. Choi, J.W. Chung, and Kim, K.S., Journal of Mathematical Physics, 53(12), 122108 (2012).
[37] M. Abramowitz and I.A. Stegun, Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables,
9th printing. New York: Dover 358 (1972).
[38] M. Pakdemirli and M. Yurusoy, SIAM review, 40(1) 96 (1998).
[39] J.L. Massera and J.J. Schaffer, Annals of mathematics 517 (1958).
[40] Here, the reader is cautioned not to confuse the L p -norm with the momentum p.