REPORT ON SOME PAPERS RELATED TO THE FUNCTION R(s)

FOUND BY SIEGEL IN RIEMANN’S POSTHUMOUS PAPERS.

J. ARIAS DE REYNA

Abstract. In a letter to Weierstrass Riemann asserted that the number N0 (T ) of zeros

of ζ(s) on the critical line to height T is approximately equal to the total number of
zeros to this height N (T ). Siegel studied some posthumous papers of Riemann trying
to find a proof of this. He found a function R(s) whose zeros are related to the zeros
of the function ζ(s). Siegel concluded that Riemann’s papers contained no ideas for
arXiv:2406.01474v1 [math.NT] 3 Jun 2024

a proof of his assertion, connected the position of the zeros of R(s) with the position
of the zeros of ζ(s) and asked about the position of the zeros of R(s). This paper is
a summary of several papers that we will soon upload to arXiv, in which we try to
answer Siegel’s question about the position of the zeros of R(s). The articles contain
also improvements on Siegel’s results and also other possible ways to prove Riemann’s
assertion, but without achieving this goal.

1. Introduction
Let N (T ) be the number of zeros of ζ(s) in the rectangle (0, 1) × (0, T ] counted with
multiplicity and N0 (T ) the number of those zeros ϱ with 0 < Im ϱ ≤ T and on the critical
line. Riemann claims in a letter to Weierstrass that he has a relatively complicated proof
that N0 (T ) is approximately equal to N (T ). Therefore, we state here the conjecture.
Conjecture (Riemann’s last Theorem). The number of zeros N0 (T ) of ζ(s) on the critical
T T T
line with 0 ≤ Im(ρ) ≤ T is N0 (T ) = 2π log 2π − 2π + o(T ) for T → +∞.
Riemann phrase in his paper [49]: One finds in fact about this many real roots between
these bounds and it is very likely that all of the roots are real, is sufficiently vague for us
to think that he means that he has calculated the first three or four zeros and shown
them to be real. Annotations of these calculations have been found in his posthumous
papers. In a letter to Weierstrass [50, p. 823] he is much more explicit.
I have not yet fully carried out the proof; and about this proof I would
like to note — you may allow me to refer to the papers to be sent to you
very soon — that the two propositions that I only have announced:
T T T
That between 0 and T there are approximately 2π
log 2π − 2π
real roots of
the equation ξ(α) = 0.
and [another one. . . ]
follow from a new development of the function ξ, which I have not yet
sufficiently simplified to announce it. From the correctness of everything
else, although some are only very briefly indicated, you will easily convince
yourself.

2020 Mathematics Subject Classification. Primary 11M06; Secondary 30D99.

Key words and phrases. zeta function, Riemann’s auxiliar function.
1
2 ARIAS DE REYNA

An expansion of the function ξ was found in the papers of Riemann, and Siegel [52]
tried to see what he could get out of this expression. He found the Riemann-Siegel
expansion and studied the function R(s) in connection with the zeros of zeta. By means
of R(s) he obtained that N0 (T ) ≥ CT for some explicit constant C, but concluded that
there was nothing in Riemann’s papers in the direction of proving the above conjecture.
He asked about the position of the zeros of R(s), and some of my work addresses this
question of Siegel.
Riemann’s papers on Number Theory in Göttingen are not the papers that a mathe-
matician would keep from his research; rather, they are the ones he would throw in the
wastebasket after writing down the most important things in his notes. If Riemann had
any ideas for proving his conjecture, they are not contained in those papers (as Siegel
concluded in [52]). What is clear is that Riemann’s papers show that he knew much more
about the zeta function than is contained in his article [49].
Siegel introduced the function
2
x−s eπix
Z
(1) R(s) = πix − e−πix
dx.
0↙1 e

We will call this Riemann’s auxiliary function. The papers we present study this function,
refining Siegel’s results in [52], trying to resolve Siegel’s question about the location of
its zeros, and asking about some issues that we have not been able to resolve.
Apart from this, some of the papers we present try to explore other ways of proving the
Riemann conjecture. In this direction we must cite the work of Levinson [46] who proved
N0 (T ) ≥ 31 N (T ). And the improvements by Conrey [37], and several others, for example
5
Pratt, Robles, Zaharescu and Zeindler [48] prove that N0 (T ) ≥ 12 N (T ) and speculate
about the possibility to get Riemann’s last theorem by refining these techniques.
Siegel (or Riemann) related R(s) to the zeta function by means of the equation
1−s


Γ
(2) ζ(s) = R(s) + χ(s)R(1 − s); χ(s) = π s−1/2 2

, R(s) = R(s).
Γ 2s
In particular, we have the following expression for the Riemann-Siegel function Z(t)1
(3) Z(t) = 2 Re{eiϑ(t) R( 21 + it)}, where ζ( 21 + it) = e−iϑ(t) Z(t) t ∈ R.
The following is a summary of each of the papers, containing our results on R(s).

2. Riemann-Siegel formula
1. High precision computation of Riemann’s zeta function by the Riemann-
Siegel formula I [3]. Siegel obtained the Riemann-Siegel asymptotic expansion without
giving explicit error bounds, Gabcke in his thesis [39] obtained explicit bounds for σ = 21
and after retaining less than ten terms. In this paper, we follow the steps of Gabcke to
obtain realistic explicit bounds after retaining any number of terms at almost any point
in the plane.
In [3, Prop. 6.1] the author needs to prove the inequality
∞ k ∞
X
k+1 r cos kθ
X rk
(4) (−1) ≤ (−1)k+1 ,
k=1
k + 2 k=1
k + 2
1I
am reluctant to call this Hardy’s function, after having seen it in Riemann’s papers at Göttingen,
and in his commentary by Siegel.
 PAPERS RELATED TO RIEMANN’S AUXILIARY FUNCTION 3

for r = 1, r = 8/9 and r = 0.883, and prove it by means of the maximal slope principle.
After [3] was published, J. van de Lune and J. Arias de Reyna [5] proved this inequality
for any r. Later on, Gabcke [40] gave a nice proof of the inequality (4).

2. High precision computation of Riemann’s zeta function by the Riemann-

Siegel formula II [4]. After writing [3] we wanted to implement in Python the com-
putation of R(s) and ζ(s). Given s ̸= 1 and ε, the problem was to compute z ∈ C
such that |ζ(s) − z| < ε. Is this possible? How many terms of the expansion have to
compute? To what precision? We solved this thanks to the power of mpmath a library
for arbitrary-precision arithmetic created by F. Johansson [55]. In mpmath my program
can be used freely. This paper solves these questions. But it is not written in final form.
We worked from scratch. We noticed that the usual implementations of the product of
two complex numbers do not follow the IEEE standard. Given two complex numbers a,
b, we want the approximation c given by the product to satisfy |c − ab| ≤ |ab|2−d when
working with d binary digits. Our calculations would be simpler if this rule were followed.

3. Programs for Riemann’s zeta function [7]. This article deals with the extension
of programs to compute R(s) and ζ(s) in mpmath [55], to the computation of the first
four derivatives of zeta and the n-th zero given n. Computing ϱn starting from n is not
as easy as some pretend, the greatest difficulty lies in precisely isolating the n-th zero
and not any other.

3. Riemann auxiliary function generalities

4. Riemann auxiliary function. Basic results [9]. This paper contains the definition
(1) and some other representation, such as, for example,
e−πis/4 −1+i∞ s/2 ′
Z
−s/2
(5) π Γ(s/2) R(s) = − τ ϑ3 (τ ) dτ.
s −1

In addition, we determine the values of R(n) for any even positive number and for odd
negative integers. The trivial zeros of zeta are also zeros of R.
Siegel wondered where the zeros of the function R(s) were. Some of my works will
provide a partial answer to this question. The X-ray of a function consists of drawing
the lines of the plane where the function takes real values, or purely imaginary values.
Figure 1 shows the X-ray2 of function R(s) in rectangle (−200, 200)2 . The computation
of R(s) for moderate values of s is difficult. To obtain part of this X-ray we have needed
to calculate R(s) with 80 decimal digits. This X-ray shows the distribution of the zeros
of R(s) into three main types. More on this in some of the papers.

5. Statistics of zeros of Riemann’s auxiliary function [10]. We have computed

all zeros β + iγ of R(s) with 0 < γ < 215946.3. A total of 162215 zeros with 25 correct
decimal digits. The zeros with γ > 0 were the most interesting zeros for Siegel. 63.9% of
these zeros satisfies β < 21 . If this were true for all zeros of R(s) with γ > 0, it will follow
that 63.9% of zeros of ζ(s) will be on the critical line. Therefore, it is plausible that the
percentage of zeta zeros on the critical line can be improved by studying R(s). But this
author has not been able to do that.
2In the pdf an X-ray can be magnificated to see the details.
4 ARIAS DE REYNA

There are other interesting things in this paper. For example, studying these zeros, we
may guess that the number of zeros of R(s) with 0 < γ ≤ T satisfies
r
T T T 1 T 3
(6) N (T ) = log − − + + O(log T ).
4π 2π 4π 2 2π 2
In [17] we prove this by substituting the error 23 + O(log T ) by O(T 2/5 log2 T ). We also
observe a periodic structure on the zeros with γ > 0.
Another interesting observation was that zeros with γ > 0 satisfies β < 1. More on
this in [13].

Figure 1. x-ray of R(s) on (−200, 200)2

 PAPERS RELATED TO RIEMANN’S AUXILIARY FUNCTION 5

4. Asymptotic expansion for the auxiliary function

6. Region without zeros for the auxiliary function of Riemann [11]. This paper
focuses on the Siegel result [52] about the left limit of the zeros of R(s) for t > 0. We
extend the asymptotic expansion of Siegel in the second quadrant, extending them to
most of the third quadrant. Giving precise bounds of the error allows us to give an
explicit region free of zeros or with only trivial zeros.
There is an interesting question here. Siegel claims that given ε > 0 there is a t0 > 0
such that there is no zero of R(s) with 1 − σ > ctε and t > t0 . Siegel only proves the
case ε = 3/7. We were unable to extend Siegel’s reasoning beyond 1 − σ > t2/5 log t. This
limits the error, for example, in (6). Therefore, my question:
1−β
Question. Is it true that the zeros β + iγ of R with γ > 0 satisfy lim supγ→+∞ γε
=0
for any ε > 0?

7. Asymptotic expansions of the auxiliary function [12]. The terms of the asymp-
totic expansion in [3] are not analytic functions. This simplifies obtaining the expansion
of the derivatives of R(s) (and consequently those of ζ(s) or Z(t)). We have used this
in our programs in mpmath, but have not written a paper explaining this. By compari-
son, the terms of the expansion obtained by Siegel are analytic functions. Here we give
two related asymptotic expansions for R(s); this is only suggested in Siegel, and we also
extend the range of validity, specifying regions without zeros. For example, we get for
t → +∞

−iϑ(t) 1  t − 14 n πt  πt  12 o
e 1
R( 2 − it) = − √ exp − · i · (1 + O(t−1/2 )),
2 2π 2 2
this is almost purely imaginary and very large. The real part is Z(t), which is relatively
small. Therefore, for t < 0 our expression for Z(t) is almost useless to locate zeros.
In this paper, we show that R(s) is approximated by an adequate zeta sum in the
first and part of the fourth quadrant (see figure 1). There is a narrow zone in the fourth
quadrant limit between the two asymptotic expansions where a line of zeros of R(s) is
found (again, see Figure 1). Finally, we explain the somewhat inexcrutable formulas (84)
and (85) of Siegel in [52].

8. Note on the asymptotic part of the auxiliary function [14]. To define explicit
regions without zeros in a first paper, we get an approximation of R(s) of type f (s)(1+U )
with |U | < 1. But this U does not tend to zero when t → +∞. In the present paper we
get an approximation of the form f (s)(1 + o(t)). We clarify here Siegel’s result following
his reasoning. This is essential to get the last Theorems in Siegel’s paper about R(s).
We prove

Theorem 1. There exist constants A and t0 > 1 such that for s in the closed set

Ω = {s ∈ C : t ≥ t0 , 1 − σ ≥ t3/7 },

we have
√
s−1 −πiη 2 2e3πi/8 sin πη 1
(7) R(s) = −χ(s)η e (1 + U ), |U | ≤ At− 21 .
2 cos 2πη
6 ARIAS DE REYNA

5. Zeros of R(s)
9. Riemann auxiliary function. Right limit of zeros [13]. All the zeros β + iγ of
R(s) that we computed satisfies β < 1. Here, we prove that this is true for γ > t0 =
3.93 × 1065 . We use the paper [3] to get a good approximation to R(s) by zeta sums.
This can be a model of how to use [3]. Of course, it will be desirable to prove this for
t0 = 0, but this appears to be difficult.
We need explicit bounds for zeta sums. It has been unfortunate that a resuly by
Landau from 1928 has been widely overlooked, rendering a number of less-than-optimal
results. We clarify this point in the paper [8], and then obtained explicit results for the
van der Corput d-th derivative test. This is the origin of my work in [32].

10. On Siegel results about the zeros of the auxiliary function of Riemann [16].
In this paper, we fill the details on the reasoning at the end of Siegel’s paper [52]. But
we feel free to change some things when we can simplify the reasoning. At some points,
we went a little further. For example, Siegel proved
X T
β = − log 2 + o(T ).
0<γ≤T
4π

we improve this in [16, Prop. 6] giving the error O(T 20/21 ). But we suspect that if the
Siegel claim about the zeros (Question 4) is true, we can get an additional term that will
explain the experimental results in [10].

11. On the number of zeros of R(s) [17]. We prove the formula (6) for the number
of zeros N (T ) for R(s) with 0 < γ ≤ T . This improves Siegel’s result and explains the
experimental results seen in [10].
I take this opportunity to give a reasonable proof of the Backlund Lemma [36]
Lemma 2 (Backlund). Let f be holomorphic in the disc |z − a| ≤ R. Let |f (z)| ≤ M
for |z − a| ≤ R. Let b be a point in the interior of the disc 0 < |b − a| < R. Assume that
f does not vanish on the segment [a, b], then
Z b ′
1 f (z) 1 M  R −1
(8) Re dz ≤ log log .
2πi a f (z) 2 |f (a)| |b − a|
12. Trivial zeros of Riemann auxiliary function [18]. In this paper, it is proved
that the trivial zeros of R(s) are simple. We get an integral expression for the derivative
√ 2
ω πn  in n ∞ (ex2 e−x )n
Z
′
(9) R (−2n) = √ dx,
2 πe 0 sin(ωx πn)
and show that these numbers are not 0.

13. Zeros of R(s) in the fourth quadrant [19]. We show that there exists a sequence
ρ−1 , ρ−2 , . . . of zeros of R(s) in the fourth quadrant. We give an algorithm to compute
the n-th zero on the fourth quadrant, and an asymptotic expansion for it showing that
4π 2 n 4πin
ρ−n ∼ 2 − .
log n log n
This is needed in [20].
 PAPERS RELATED TO RIEMANN’S AUXILIARY FUNCTION 7

14. Infinite product of Riemann auxiliary function [20]. We obtain the product
formula for R(s). We study the phase of R(s) at the critical line, that is, the real
analytic and real function ω(t) such that R( 21 + it) = e−iω(t) g(t) where g is also real and
real analytic. Each zero of cos(ϑ(t) − ω(t)) determine a zero of ζ(s) on the critical line.
We study the relation of ω(t) with the zeros of R(s). For example, for t > 0 we have
ω(t) ≈ 2πNr (t) where Nr (t) is the number of zeros ρ = β + iγ with 0 < γ < T and β > 12 .
Our computation of zeros (see [10]) of R(s) suggests that Nr (t) ≈ ϑ(t) 6π
. If this were
2 2
true ϑ(t) − ω(t) ≈ 3 ϑ(t), and at least 3 of the zeros of ζ(s) will be on the critical line.
The method of Levinson [46], has been refined by several authors [37], [51], [48], to obtain
5
12
zeros of ζ(s) on the critical line. Therefore, the study of the zeros of R(s) has the
potential to improve these results.
The behavior of R(s) on the critical line, similar to that of ζ(s), depends on the sum
X
n−s .
√
n≤ t/2π

We know that the behavior of ζ(s) is not typical until the log log t is really large. There-
fore, assume that Nr (t) ≈ ϑ(t)
6π
from the scarce data we have is risky. Maybe Nr (t) = o(t)
after all, and Riemann had a proof, who knows.

6. Density Theorems
15. Mean Values of the auxiliary function [15]. In this paper, we obtain the main
terms of the mean values
1 T 1 T
Z  t σ Z
2
| R(σ + it)| dt, and | R(σ + it)|2 dt.
T 0 2π T 0
Giving complete proofs of some result of the paper of Siegel on the Riemann Nachlass.
Here, R(s) is the function related to ζ(s) found by Siegel in the Riemann papers.
There are two cryptic passages in Siegel [52, before eq. (64)] where he says
Riemann tried to obtain an expression about the zeros of R(s); from (58)
he found
2 2
x−σ−ti y −σ+ti eπi(x −y )
Z Z
2
(10) | R(σ + ti)| = dx dy,
(eπix − e−πix )(eπiy − e−πiy )
0↙1 0↘1

and brought the complex double integral into a different form by the in-
troduction of new variables, the deformation of the integration area, and
the use of the residue theorem; however, this did not yield a useful result.
then in [52, between eq. (86) and (87)] to compute a mean value he writes

∞ ∞ 2 2
x−σ−ti y −σ+ti eπi(x −y )
Z Z nZ Z o
2 −εt −εt
| R(σ + it)| e dt = e dx dy dt.
0 0 (eπix − e−πix )(eπiy − e−πiy )
0↙1 0↘1

Here, one may transform the right side by deformation of the path of
integration, interchange of the order of integration, and application of the
residue theorem. The calculation yields the statement
Z ∞
1 1
(11) | R(σ + it)|2 e−εt dt ∼ (2πε)σ− 2 Γ( 12 − σ).
0 2ε
8 ARIAS DE REYNA

We were unable to reproduce any of this. So, we get the result by more standard means.
From (11) Siegel asserts that for all σ < 21

∞ 3
(2ε)− 2
Z  t σ
2 −εt
(12) | R(σ + it)| e dt ∼ .
1 2π 1 − 2σ

He does not give a proof of this implication. There is a statement in Titchmarsh [54,
p. 159] which proves the implication (11) =⇒ (12) for σ < 0. The proof in Titchmarsh
needs hypothesis σ < 0. Siegel only uses (12) for a value of σ = −5.47559 . . . so this will
be sufficient. However, in this paper we give a different and complete proof of (12) for
σ < 21 .
This double integral appears in one of the papers of Riemann Nachlass in Göttingen.
Usually we are able to interpret what Riemann does in any of these papers, but in this
case, they have not aided me in understanding what Siegel says.
Not only for the above considerations, but rather by the transformations that Siegel
writes between equations (36) and (54) of his paper, we wonder from time to time if the
Riemann papers that Siegel consulted returned to their place in the Nachlass. There are
several papers in the actual Nachlass with the Riemann-Siegel expansion, but Siegel does
not appear to be following any of them in his paper. Nevertheless, Siegel says: Just how
strong Riemann’s analytic technique was, it is made particularly clear by his derivation
and manipulation of the asymptotic series for ζ(s).

16. Density theorems for Riemann’s auxiliary function [21]. Iwaniec and Kowal-
ski ([42, p. 2]) say that density theorems about the zeros of L-functions are likely a waste
of time studying the empty set. Rather, these density theorems probably cannot be bet-
ter because they are really theorems about the zeros of R(s), and this function has many
zeros spread over the entire critical strip.
Since
Z(t) = eiϑ(t) ζ( 21 + it) = 2 Re{eiϑ(t) R( 21 + it)}.

ζ(s) have a zero on the critical line at each point s = 12 + it where eiϑ(t) R( 12 + it) is purely
imaginary. This is the same as saying that ζ(s) have a zero at a point s in the critical
line just in case π −s/2 Γ(s/2) R(s) is purely imaginary. Siegel connected the zeros of R(s)
with zeros on the critical line of ζ(s), showing that ζ(s) have at least 2N2 (T ) zeros on
the critical line with ordinates between 0 and T , where N2 (T ) is the number of zeros of
R(s) to the left of the critical line and with ordinates between 0 and T . To understand
the relationship between the zeros, consider the x-ray of eiϑ(t) R( 21 + it) in Figure 2. The
horizontal line is the real axis for t.
Each zero of R(s) to the left of the critical line gives a zero above the real axis in Figure
2. The function Z(t) vanishes at the points where the imaginary lines (blue lines) in the
figure cut the real axis. All imaginary lines (in blue) start at the bottom of the figure 2
so that the imaginary line passing through a zero of eiϑ(t) R( 12 + it) above the real axis
cuts the real axis at two points, producing two zeros of Z(t).
In principle, it would be possible that the imaginary line through a zero below the real
axis in the figure does not cut the real axis. But we see in the examples in the figure that
they are sufficiently close to give also two zeros of Z(t).
 PAPERS RELATED TO RIEMANN’S AUXILIARY FUNCTION 9

Figure 2. x-ray of eiϑ(t) R( 12 + it) in (200 040, 200 060) × (−2, 4)

Figure 3. x-ray of eiϑ(t) G(t) in (200 040, 200 060) × (−2, 4)

Since R(s) can be approximated by Dirichlet polynomials, the methods used to prove
the density theorems for ζ(s) can be applied to R(s), and this is the objective of this
3
paper. Our main Theorem proves that for 1/2 < α ≤ 1, we have N (α, T ) ≪ T 2 −α log3 T ,
where N (α, T ) denotes the number of zeros of R(s) in [α, 1] × [0, T ] counted with multi-
plicity. The proof is standard, but we do not see any advantage in using a mollifier here.
Perhaps that is a point where my result can be easily improved.
These zeros below the real line in our figure correspond to the zeros of R(s) to the right
of the critical line. In this paper, we show that most of the zeros of R(s) to the right of
the critical line are near this line. These density theorems make it almost inevitable that
T T T
each of these zeros produce its two zeros for ζ(s). In this way, the 4π log 2π − 4π zeros of
T T T
R (s) can generate the 2π
log 2π
− 2π
zeros predicted by the Riemann hypothesis.
In this paper, we follow the Theorems in Iwaniec [41], adapting them to R(s). We give
proofs of these Theorems, giving explicit constants.
17. Left density theorems for the auxiliary function [22]. Since ζ(s) − R(s) =
χ(s)R(1 − s) the zeros to the left of the critical line are symmetrical with respect to this
line of the zeros of ζ(s) − R(s). We may approximate this last function by a Dirichlet
polynomial to the right of the critical line. This may be a way to bound the number of
these zeros. This is one of the papers in which my lack of technique is most exposed. We
have not been able to obtain anything reasonable following this line. We mention this in
case anyone can get something.
10 ARIAS DE REYNA

18. On the approximation of the zeta function by Dirichlet polynomials [23].

For its use in [22], we needed an explicit version of Theorem 4.11 of Titchmarsh. There
are many papers dealing with this question, but we found a real chaos in the literature
P Therefore, we prove that for−σs = σ + it
on the range of validity of the various results.
with σ ≥ 0 and 0 < t ≤ x, we have ζ(s) = n≤x n−s + x1−s /(s − 1) + 29 14
Θx , where Θ
is a complex number with |Θ| ≤ 1. This improves Theorem 4.11 of Titchmarsh, making
explicit a result from Iwaniec in [41] to whom we dedicate this work. In the end, we do
not arrive at anything good in [22].

7. Representation and approximation to Riemann-Siegel function Z(t)

There are many functions f (s) satisfying Z(t) = 2 Re{eiϑ(t) f ( 21 + it)}. For example, in
[6] with van de Lune we proved
−ϑ′ (t)Z(t) = Re{eiϑ(t) ζ ′ ( 21 + it)}.
Each of these formulas gives an opportunity to show that many zeros are on the critical
line. Many times we get an approximation to f ( 12 +it) giving us an approximation to Z(t).
The general objective here is to obtain a real function u(t) such that Z(t) = u(t) + o(t−ε ),
T T T
and such that u has 2π log 2π − 2π + o(T ) zeros on [0, T ].

19. An expression for Riemann Siegel function [24]. Lavrik [43] consider for any
τ ∈ C with Re τ > 0 the function
∞
(πτ )s/2 X Γ( 2s , πn2 τ ) 1
(13) L(τ, s) = − + .
s Γ( 2s ) n=1
Γ( 2s ) ns
This function satisfies
(14) π −s/2 Γ( 2s )ζ(s) = π −s/2 Γ( 2s ) L(τ, s) + π −(1−s)/2 Γ( 1−s
2
) L(1/τ, 1 − s).
We consider specially the case τ = 1 that was considered by Riemann in [49]. We define
L(s) = L(1, s), and we have Z(t) = 2 Re{eiϑ(t) L( 12 + it)}. The non-trivial zeros of L(s)
are contained in the halfplane Re s > 11.25 and are easy to compute.

20. Approximate formula for Z(t)[25]. The function

−iσ+∞
Z
1 π
eiϑ(t+x)−iϑ(t) ζ 1


(15) U (t) = + i(t + x) dx,
2πi 2
2 sinh πx
2
−iσ−∞

where 0 < σ < 12 , satisfies Z(t) = 2 Re{eiϑ(t) U (t)}. Approximating this function, we get
∞
X 1 t
(16) G(t) = 1
n=1 n 2
+it 2πn2 + t
5
satisfies Z(t) = 2 Re{eiϑ(t) G(t)}+ O(t− 6 +ε ). The function G(t) extends to a meromorphic
function on C with poles at (2k − 12 )i and −2πn2 . There are many interesting things
about G(t) for example
∞
t  X 1 t 
(17) G(t) = ζ(2 + 12 + it) − 2+ 12 +it 2πn2 + t
.
2π n=1 n
 PAPERS RELATED TO RIEMANN’S AUXILIARY FUNCTION 11

So that the values of Z(t) appear to depend only on things happening at σ > 2. Also
for small t we observe in Figure 4 that the continuous arg G(t) is increasing very slowly.
Giving the zeros we want for Z(t).

Figure 4. Plot of G(t) in (1000, 1040) × (−10, 10).

21. Integral Representation for Riemann-Siegel Z(t) function [26]. We apply the
solution of the Dirichlet problem in the strip −4 ≤ σ ≤ 5 for adequate boundary values
to get an integral representation of Z(t)
Z ∞
h(x)ζ(4 + ix) Re F (t)
(18) F (t) = x−t dx, Z(t) = 1 1 1 ,
−∞ 7 cosh π 7 ( 4 + t2 ) 2 ( 25
4
+ t2 ) 2
where h(x) is a simple function. After that we get the estimate
 t  74
(19) Z(t) = Re{eiθ(t) H(t)} + O(t−3/4 ).,
2π
with
Z ∞  ix/2  t −7/4 X∞
t ζ(4 + it + ix) 1 2
H(t) = dx = 1
+it 1 + ( t )−7/2
.
−∞ 2π 7 cosh(πx/7) 2π n=1 n 2
2πn2

We arrive at an equation similar to (16). The behaviour of this H is similar to that of G.

8. Some integrals for ζ(s), R(s) and similar functions

22. An entire function defined by Riemann [27]. This is a representation integral
for ζ(s) that we find in one of the papers of Riemann’s Nachlass in Göttingen3. Riemann

3We have a promise to communicate to Göttingen any papers I obtained from the papers of Riemann,
we expect to do it before publishing this.
12 ARIAS DE REYNA

defines a bounded function on R whose Fourier transform vanishes at the real values with
ζ( 12 + iγ) = 0. First, Riemann defines an entire function
eizw
Z
(20) χ(z) := 2πw2 − 1
dw,
L e

where L is a line parallel to the real axis at a height 0 < b < 2−1/2 . It has the power
series expansion
∞
z 1 X ζ(n + 12 )  z 2 n
(21) χ(z) = + √ − .
2 2 n=0 n! 8π
and satisfies the functional equation χ(−z) = χ(z) − z. And is connected with ζ(s) by
Z ∞
(22) χ(x)x2s−1 dx = Γ(2s)ζ(s + 21 ) cos π(2s+1)
4
, Re(s) > 0.
0

23. An integral representation of R(s) due to Gabcke [28]. Gabcke [38] proved
s/2 s/2 πi(s−1)/4
Z 2
e−πiu /2+πiu √
R(s) = 2 π e U (s − 12 , 2πeπi/4 u) du
− 12 ↘ 21 2i cos πu
where U (ν, z) is the usual parabolic cylinder function [1]. Here, we give a new proof and
put it in the form
Z ∞ −πx2 √
s s/2 πis/4 e H−s (x π)
(23) R(s) = −2 π e dx.
−∞ 1 + e−2πωx
where Hν (z) is the generalization of Hermite polynomials [44, S. 10.2].
24. Integral representation of Riemann auxiliary function [29]. We transform
(23) into
2
2s π s eπis/4 ∞ s−1 1 − e−πy +πωy
Z
(24) R (s) = − y dy.
Γ(s) 0 1 − e2πωy
−πy 2 +πωy
What is remarkable here is that the function of the integrand f (y) = 1−e1−e2πωy appearing
in (24) is entire. Zeros√in the denominator are also zeros in the numerator. For y → +∞
we have |f (y)| ∼ e−πy 2 . When we move the integration line, no residues appear. This
may be useful in bounding R(s) without introducing the zeta sum. The function f (y)
has a nice x-ray, shown in the paper.
25. Levinson functions [30]. Levinson [45] find entire functions of s, f (s, τ ) for each
τ ̸= 0 on the closed hiperplane Im τ ≥ 0 (my notations in this paper), such that
h(s)ζ(s) = h(s)f (s, τ ) + h(1 − s)f (1 − s, 1/τ ), h(s) = π −s/2 Γ( 2s ).
This implies that the function
f (s, τ ) + f (s, 1/τ )
Rτ (s) := R(s, τ ) := ,
2
satisfies
(25) Z(t) = 2 Re{eiϑ(t) Rτ ( 12 + it)}.
We also have R−1 (s) = R(s) and Ri (s) = L(s) considered in [24]. The limit cases when τ
is a non-null rational number are especially interesting. The definition given by Levinson
 PAPERS RELATED TO RIEMANN’S AUXILIARY FUNCTION 13

is by integrating a series. In this paper we show that for rational τ , this series can be
simplified to a trigonometric function as in the case of R. To achieve this we prove some
nice summation formulas. Similar to Gaussian sums. For example, we show:
4πi 2
x−s e 3 x  − πi
Z
1 πx

R−4/3 (s) = √ 2e 6 cos 3 + i cos(πx) dx
2 3 0↙1 eπix − e−πix
(26) 3πi 2
x−s e 4 x  πi
Z
1 πx − 3πi

+ e + 2 cos 2 + e
4 4 cos(πx)
4 0↙1 eπix − e−πix
2
x−s e2πix cos πx
Z
1
R−2 (s) = πix − e−πix
dx
2 0↙1 e
(27) πi 2
x−s e 2 x  πi
Z
1 − πi

+ √ e 4 + e 4 cos(πx) dx.
2 2 0↙1 eπix − e−πix
This must be joined with (25). In the paper we give general similar expressions for any
rational number not equal to 0.
The above expressions hint that we may transform the approximate functional equation
in Theorem 4.13 of Titchmarsh into asymptotic expansions.

9. A naive integral
26. A naive integral [33]. In [9] two real functions g(x, t) and f (x, t) are defined, so
that the Riemann-Siegel Z function is given as
n u(t)e πi8 Z ∞ o
Z(t) = Re 1 g(x, t)eif (x,t) dt ,
2
+ it 0
where u(t) is a real function of order t−1/4 when t → +∞. The function g(x, t) is
indefinitely differentiable and tends to 0 as well as all its derivatives when x → 0+ or
x → +∞. Since, furthermore, for t → +∞ the function f (x, t) tends to +∞ we may
expect that the integral depends essentially on the behavior of g(x, t) at the extremes.
As Polya in an analogous situation [47] we consider the substitution of ψ(x) by a
simpler similar function. A simple function with this behavior is
π
ψ0 (x) := 2π(1 + 41 x−5/2 )e−πx− 4x .
Therefore, we define J0 (t) replacing in the definition of J(t) the function ψ(x) by the
simpler ψ0 (x).
Z ∞
5 π 1 1
(28) J0 (t) = 2π (1 + 14 x− 2 )e−πx− 4x (1 − ix) 2 ( 2 +it) dx.
0
The resulting Z0 (t) disappoints us
n 2 n t t t π o 2  r t o
Z0 (t) ≍ Re √ exp i log − − + exp πi , t → +∞.
π 2 2π 2 8 (2πt)1/4 2π
However, the integral J0 (t) is interesting as a technical challenge. And still we have the
possibility to get a better result improving ψ0 (x).
This paper needs a strong revision, but we will upload it to arXiv without this revi-
sion. It is an interesting integral and can be a good challenge for those interested in the
asymptotic development of integrals.
14 ARIAS DE REYNA

10. Some other related results

27. Simple bounds for the auxiliary function of Riemann [31]. Given s ̸= 1 in C
and ε > 0, in [4] we want to compute z such that |ζ(s) − z| < ε. To this end we need to
compute many products ab with a prescribed error. To achieve this we need some rough
bounds of the quantities a and b. This is the origin of this paper. Its object is to give
bounds of ζ(s), ϑ(s), R(s), Z(t), for use in the numerical computations. It is good to
have reliable, not necessarily very good, but easy to compute bounds. The objective were
not to publish this, but only serve me for reference. Since we are not going to publish
these papers and we have many references, we prefer to upload it as it currently stands.

28. Explicit van der Corput estimates [32]. In [13] it was required to bound a zeta
sum. We required explicit bounds, so the book of Titchmarsh was not sufficient. We
noticed that many papers overlooked a result by Landau. This made us upload [8] to
arXiv, which is nothing more than a translation of the paper in German by Landau,
where he proves what is the best constant for an inequality which some pretended to
improve upon.
Then, we searched for theorems to bound the zeta sum that can be turned explicit.
The obvious election was van der Corput’s d-th estimates. We searched for the original
papers and found, besides the usual references, three papers of van der Corput giving
explicit estimates. It was surprising that no one referred to these papers. It is a very
elaborate construction, so we started studying the papers. We found an error in the van
der Corput reasoning. This explains why nobody refers to these papers. We found that
the error can be rectified fairly easily. This paper is the exposition of van der Corput
reasoning on these papers with adequate correction. We prove
Theorem 3. Let X, and Y ∈ R be such that ⌊Y ⌋ > d where d ≥ 3 is a natural number.
Let f : (X, X + Y ] → R be a real function with continuous derivatives up to the order d.
Assume that 0 < λ ≤ f (d) (x) ≤ Λ for X < x ≤ X + Y . Denote by D = 2d . Then
1 X n  Λ 2/D  Λ2 1/(D−2) o
(29) e(f (n)) ≤ max Ad , Bd , Cd (λY d )−2/D ,
Y X<n≤X+Y λY λ

where Ad , Bd , and Cd are explicit constants. They depend on d but for d ≥ 2 for example
Ad < 7.5, Bd < 5.8 and Cd < 10.9.
We apply this theorem to zeta sums, giving the best choice of d in each case. Also, we
prove that our Theorem implies Titchmarsh’s Theorem 5.13.

29. Levinson-Conrey method following the exposition of Iwaniec [34]. Contains

a detailed exposition of the Levinson-Conrey method. This is the result of my study of
the book by Iwaniec [41]. Contains several changes to the book and improvements, some
of them due to Iwaniec. I appreciate his answers to my many questions. My aim was to
apply this method directly to the R(s) function, but I have not reached any useful result.
This is almost a booklet in which I have worked much,but it is as much my work as it is
Iwaniec’s.

30. Programs for computing the zeros [35]. It is difficult to get all the zeros
ρ = β + iγ of R(s) with 0 < γ ≤ T . Here, we explain the procedure we have followed.
But the paper is incomplete, and consequently, I will not upload it for the moment.
 PAPERS RELATED TO RIEMANN’S AUXILIARY FUNCTION 15

31. X-ray’s of holomorphic functions. Still, only a project. It will give definitions,
some interesting properties, and how to read an x-ray. It will contain some history, for
example that Gauss published the first x-ray, as an illustration of his first proof of the
fundamental theorem of algebra. The complex plane and the x-rays appeared at the same
time.

11. Some Open Problems

An outside observer will see here nothing but methods already known, when a new idea
is surely needed to advance the proof of Riemann’s last theorem. Actually, this is just
the result of new ideas not getting me anywhere, as, for example, in [22], but I always
thought that everything we knew about R(s) could be useful to get to the difficult proof
that Riemann claimed to have. I have not even finished expressing the already known
methods; for example, I could write a paper explaining the relation of R(s) and the
Lindelöf hypothesis. The Lindelöf hypothesis is equivalent to saying that R( 21 + it) ≪ tε
for all ε > 0. Therefore, Riemann’s efforts to bound R( 12 + it) in (10), could lead him to
prove Lindelöf’s hypothesis.
I also want to point out here the difference between the density theorems for R(s) and
ζ(s). Putting the zeros of ζ(s) close to the critical line does not seem to move in the
direction of proving Riemann’s last theorem, but doing the same with R(s) is different.
Every zero of R(s) close to the line is likely to place two zeros of ζ(s) right on the line.
Here we collect some open problems related to R(s).
Problem 1. Can we get some useful bound from the expression of | R(σ + it)|2 as a
double integral.
I mention here that the very similar double integrals of Riemann in his Nachlass are
not equal to the ones given by Siegel.
Problem 2. Let R(ρ) = 0 with Im(γ) > 0. Is it true that Re(β) < 1?
This is proved in [13] for Im(γ) > 3.93 × 1065 , and is true for Im(γ) < 215946.
Another important problem is the left limit of the zeros. Siegel [52] claim to have
proved it. But we have not been able to prove it, so we consider it an open problem.
Problem 3. Let ε > 0. Is it true that there is no zero of R(s) in the region defined by
1 − σ > tε and t > t0 , for some t0 > 0?
Problem 4. Let NR (T ) be the number of zeros of R(s) with 0 < Im(ρ) ≤ T , and NR (β <
1/2, T ), the number of those zeros with Re(ρ) < 1/2, both counted with multiplicities.
Does
NR (β < 1/2, T )
lim
T →+∞ NR (T )
exist? Is its value equal to 1?
An affirmative answer to the second question in Problem 4 implies the Riemann last
theorem.
One possibility of proving the Riemann last theorem is that it would be a very general
property.
16 ARIAS DE REYNA

Problem 5. Let f be a holomorphic function for Re s > 0 with |f (s)−1| < 1/2 for σ > 2
and t > t0 . Assume also that
Z T
|f ( 21 + it)|2 dt = O(T log T ),
0

and that there is some constant a such that f (s) = O(ta ) for 0 < σ < 2 and t → +∞.
T
Is it true that the real function Re{eiϑ(t) f ( 12 + it)} has ∼ 2π T
log 2π zeros on the segment
(0, T )?
Here we have to be open about the exact conditions for f .

References
[1] M. Abramowitz, I. A. Stegun, Handbook of Mathematical Functions with Formulas, Graphs
and Mathematical Tables, National Bureau of Standards Appl. Series, vol 55, U. S. Government
Printing Office, Washington, DC, 1964.
[2] J. Arias de Reyna, Riemann’s fragment on limit values of elliptic modular functions, Ramanujan
J. 8 (2004), 57–123.
[3] J. Arias de Reyna, High precision computation of Riemann’s zeta function by the Riemann-Siegel
formula, I , Math. Comp. 80 (2011), no. 274, 995–1009.
[4] J. Arias de Reyna, High precision computation of Riemann’s zeta function by the Riemann-Siegel
Formula II, 2022, arXiv:2201.00342.
[5] J. Arias de Reyna and J. van de Lune, A proof of a trigonometric inequality. A glimpse inside
the mathematical kitchen, J. Math. Inequal., 5 (2011) 341–353.
[6] J. Arias de Reyna and J. van de Lune, On the exact location of the non-trivial zeros of
Riemann’s zeta function, Acta Arithmetica, 163 (2014) 215–245.
[7] J. Arias de Reyna, Programs for Riemann’s zeta function, in J van de Lune (ed.), Herman J. J.
te Riele Liber Amicorum, Centrum voor Wiskunde en Informatica, Amsterdam, pp. 102–112.
[8] J. Arias de Reyna, On Kuzmin-Landau Lemma, 2020, arXiv:2002.05982 .
[9] J. Arias de Reyna, Riemann’s auxiliary function: Basic Results, preprint (166).
[10] J. Arias de Reyna, Statistics of zeros of the auxiliary function, preprint (172).
[11] J. Arias de Reyna, Region without zeros for the auxiliary function of Riemann, preprint (98).
[12] J. Arias de Reyna, Asymptotic expansions of the auxiliary function, preprint (100).
[13] J. Arias de Reyna, Riemann’s auxiliary function. Right Limit of zeros, preprint (173).
[14] J. Arias de Reyna, Note on the asymptotic of the auxiliary function, preprint (193).
[15] J. Arias de Reyna, Mean values of the auxiliary function, preprint (101).
[16] J. Arias de Reyna, On Siegel results about the zeros of the auxiliary function of Riemann,
preprint (102).
[17] J. Arias de Reyna, On the number of zeros of R(s), preprint (185).
[18] J. Arias de Reyna, Trivial zeros of Riemann auxiliary function, preprint (186).
[19] J. Arias de Reyna, Zeros of R(s) on the fourth quadrant, preprint (108).
[20] J. Arias de Reyna, Infinite product of the Riemann auxiliary function, preprint (66).
[21] J. Arias de Reyna, Density Theorems for Riemann’s auxiliary function, preprint (174).
[22] J. Arias de Reyna, Left density theorems for auxiliary function, preprint (191).
[23] J. Arias de Reyna, On the approximation of the zeta function by Dirichlet polynomials, preprint
(192).
[24] J. Arias de Reyna, An expression for Riemann Siegel function, preprint (75).
[25] J. Arias de Reyna, Approximate formula for Z(t), preprint (60).
[26] J. Arias de Reyna, Integral representation for Riemann-Siegel Z(t), preprint (162).
[27] J. Arias de Reyna, An entire function defined by Riemann, preprint (152).
[28] J. Arias de Reyna, An integral representation of R(s) due to Gabcke, preprint (187).
[29] J. Arias de Reyna, Integral representation of Riemann auxiliary function, preprint (188).
[30] J. Arias de Reyna, Levinson functions, preprint (197).
[31] J. Arias de Reyna, Simple bounds for the auxiliary function of Riemann, preprint (92).
[32] J. Arias de Reyna, Explicit van der Corput’s d-th estimates, preprint (181).
 PAPERS RELATED TO RIEMANN’S AUXILIARY FUNCTION 17

[33] J. Arias de Reyna, A naive integral, preprint (123).

[34] J. Arias de Reyna, Levinson-Conrey method following the exposition of Iwaniec, preprint (194).
[35] J. Arias de Reyna, Programs to compute the zeros, preprint (180).
[36] R. J. Backlund, Sur les zéros de la fonction ζ(s) de Riemann, Comptes Rendues de l’Académie
des Sciences 158 (1914) 1979–1981.
[37] J. B. Conrey, More than two fifths of the zeros of the Riemann zeta function are on the critical
line, J. Reine Angew. Math. 399 (1989) 1–26.
[38] W. Gabcke, A Parabolic Cylinder Function in the Riemann-Siegel Integral Formula, (2015),
arXiv:1512.01186.
[39] W. Gabcke, Neue Herleitung und explizite Restabschätzung der Riemann-Siegel Formel , Ph.D.
thesis, Georg-August-Universität zu Göttingen, 1979.
[40] W. Gabcke, On an Inequality Related to a Certain Fourier Cosine Series, 2015. arXiv:1510.06282.
[41] H. Iwaniec, Lectures on the Riemann’s zeta function, Amer. Math. Soc., Providence, Rhode Island,
2014.
[42] H. Iwaniec & E. Kowalski, Analytic Number Theory, Amer. Math. Soc., Providence, Rhode
Island, 2004.
[43] A. F. Lavrik An approximate functional equation for the Dirichlet L-function, Trans. Moscow
Math. Soc. 18 (1968), 101–115.
[44] N. N. Lebedev, Special functions and their applications, Revised ed., translated from the Russian
and ed. by Richard A. Silverman. Dover Publ. Inc., New York 1972.
[45] N. Levinson, Remarks on a Formula of Riemann for his Zeta-Function, J. Math. Anal. Appl. 41
(1973) 345–351.
[46] N. Levinson, More than one third of zeros of Riemann’s zeta-function are on σ = 12 , Adv. Math.
13 (1974) 383–436.
[47] G. Pólya, Bemerkung über die Integraldarstellung der Riemannschen ξ-Function, Acta Math. 48
(1946) 305–317.
[48] K. Pratt, N. Robles, A. Zaharescu, & D. Zeindler, More than five-twelfths of the zeros of
ζ are on the critical line, Res. Math. Sci. 7 (2020), no. 2, Paper No. 2, 74 pp.
[49] B. Riemann, Über die Anzahl der Primzahlen unter einer gegebenen Grösse, Monatsber. Akad.
Berlin (1859) 671–680.
[50] B. Riemann, Gesamemelte Mathematische Werke, Wissenschaftlicher Nachlass und Nachträge—
Collected Papers, Ed. R. Narasimhan, Springer, 1990.
[51] Shaoji Feng, Zeros of the Riemann zeta function on the critical line, J. Number Theory, 132,
(2012), 511–542.
[52] C. L. Siegel, Über Riemann Nachlaß zur analytischen Zahlentheorie, Quellen und Studien zur
Geschichte der Mathematik Astronomie und Physik 2 (1932) 45–80. (Reprinted in [53], 1, 275–310.)
English version.
[53] C. L. Siegel, Carl Ludwig Siegel’s Gesammelte Abhandlungen, (edited by K. Chandrasekharan
and H. Maaß), Springer-Verlag, Berlin, 1966.
[54] E. C. Titchmarsh The Theory of the Riemann Zeta-function, Second edition. Edited and with
a preface by D. R. Heath-Brown. The Clarendon Press, Oxford University Press, New York, 1986.
[55] The mpmath development team, mpmath: a Python library for arbitrary-precision floating-
point arithmetic (version 1.3.0), 2023.

Afterword
These twenty-odd articles represent my unsuccessful attempts to prove Riemann’s last
theorem. As long as I did not achieve my goal, I did not publish my progress. This has
always been my way of doing mathematics. But at 76 years of age, I realised that I did
not want these small advances to be definitively lost with my death. So a few years ago
I started to write them more carefully. With the idea of uploading them to arXiv.
I don’t achieve my goal, but I bring some knowledge about the function R(s), and I
pose several problems that I think are interesting (see section 11). The main reason for
publishing them is that I still think that there is a real way to prove the last Riemann
18 ARIAS DE REYNA

theorem by means of the R(s) function. Sure some mathematician will be able to achieve
what I have not been able to.
I have reviewed these papers as if I were the referee. In such a large volume I’m sure
some error has escaped me, but I hope there are no serious problems. I can continue
reviewing, but there will be no end to it. There is one article that goes nowhere [22] and
another one [33] that needs revision and I’m not going to do that because it’s just a side
issue. That’s why I’m deciding to upload them now.
I would be appreciate any comments that could improve a possible new version.
Universidad de Sevilla, Facultad de Matemáticas, c/Tarfia, sn, 41012-Sevilla, Spain.
Email address: arias@us.es, ariasdereyna1947@gmail.com