# X-RAYS OF THE RIEMANN ZETA AND XI .X-RAYS OF THE RIEMANN ZETA AND XI FUNCTIONS J. M. Hill 1 and R

date post

19-Sep-2018Category

## Documents

view

219download

0

Embed Size (px)

### Transcript of X-RAYS OF THE RIEMANN ZETA AND XI .X-RAYS OF THE RIEMANN ZETA AND XI FUNCTIONS J. M. Hill 1 and R

X-RAYS OF THE RIEMANN ZETA

AND XI FUNCTIONS

by

James M. Hill and Robert K. Wilson,

School of Mathematics and Applied Statistics,

University of Wollongong,

Wollongong, NSW 2522 Australia

email: jhill@uow.edu.au

Contents

Abstract 3

1. Introduction 3

2. Basic Equations 4

3. Symmetries of the Riemann xi functions 5

4. X-rays for (s), (s), (s) and log (s) 7

5. Conclusions 7

2

X-RAYS OF THE RIEMANN ZETA

AND XI FUNCTIONS

J. M. Hill 1 and R. K. Wilson

Abstract. The X-ray of a function of a complex variable are the two families of curves alongwhich the function is either purely real or purely imaginary. Following previous work, a thickerblack line is used to show those curves along which the imaginary part of the function vanishes (i.e.the function is real along such curves), while a thinner grey line designates those curves for whichthe real part of the function vanishes (i.e. the function is purely imaginary along such curves).The X-ray of the Riemann zeta function (s) has been given and fully delineated by a previousauthor. Here we show that a considerably simplified picture emerges for the X-rays of the Riemannxi function (s) and related functions. The pictures so obtained are entirely consistent with thevalidity of the Riemann hypothesis, while the X-ray of the function log (s) produces an unusualsituation.

1. Introduction

Perhaps the most famous outstanding problem of modern mathematics is the Riemann hypothe-sis, formulated by Bernhard Riemann in 1859 (see Edwards [4]). For s = x + iy and x > 1 thezeta-function is defined by

(s) =

n=1

1ns

(1.1)

and the following equation due to Euler

(s) =p

(1 1

ps

)1, (1.2)

where p ranges over all primes (p = 2, 3, 5, 7, 11, . . .), indicates the existence of a deep-lying connec-tion between the theory of the zeta function and prime number theory. The Riemann hypothesisasserts that, all the complex zeros of the function which is the analytical continuation of (1.1) inthe region 0 < 0 all but an infinitesimalproportion of the roots lie within of

In terms of understanding the subtleties of the Riemann zeta function and its analytical con-tinuation, Arias-de-Reyna has proposed the notion of the X-ray of the function [1]. The X-ray ofa function of a complex variable are the two families of curves along which the function is eitherpurely real or purely imaginary. A thicker black line is used to show those curves along whichthe imaginary part of the function vanishes (i.e. the function is real along such curves), while athinner grey line designates those curves for which the real part of the function vanishes (i.e. thefunction is purely imaginary along such curves). Arias-de-Reyna shows that these pictorial devicesprovide an excellent means of showing in a single picture, information which is otherwise scattered.The purpose of this note is to extend this approach, and in particular show that the X-rays ofthe Riemann xi function present an increasingly simplified presentation of this information. Inaddition, we show that the X-ray for the function log (s) gives an unexpected outcome. The com-puter software program Mathematica has been used to draw these implicit function curves. Thisis based on calculating a set of function values, and drawing a curve along the boundary betweenthe positive and negative values.

In the following section we present a brief derivation of the basic equations. In the sectionthereafter we state the basic symmetries of the Riemann xi functions, which together with theCauchy-Riemann equations for an analytic function of a complex variable, constitute the elemen-tary theoretical tools necessary to make credible the X-ray diagrams obtained numerically. InSection 4 we present X-rays for the four functions (s), (s), (s) and log (s), which are alldefined in the following section.

2. Basic equations

We may show that the function (s) which is the appropriate analytical continuation of (s),as defined by either (1.1) or (1.2), may be determined from the Mellin transform expression

(s) =(s/2)(s)

s/2=

0

ts/21(t)dt = (1 s) (2.1)

which is valid for all s, where (t) is the function defined by

(t) =

n=1

en2t 1

2

t, (2.2)

and (s) denotes the usual gamma function, which for 0 is defined by

(s) =

0

etts1dt. (2.3)

For 0 t < , the function (t) is a well behaved monotonically increasing function such that1/2 (t) < 0 (see Figure 1), which is known to satisfy the functional equation

(1/t) =

t(t), (2.4)

which may be expressed in the invariant form

(1/t)1/4(1/t) = t1/4(t). (2.5)

We comment that the final equality of equation (2.1) follows immediately from the substitutiont 1/t and the functional equation (2.4). The above equations can be derived from the wellknown representation of the zeta function (see Edwards [4] page 16 or Whittaker and Watson [8]page 273)

(s) =

1

(t){

ts/2 + t(1s)/2} dt

t 1

s(1 s) , (2.6)

4

which is valid for all s, and (t) denotes simply

(t) =

n=1

en2t, (2.7)

which is known to satisfy the functional equation

(1/t) + 1/2 =

t [(t) + 1/2] . (2.8)

As described in Hill [5], and using the functional equation (2.8), equation (2.6) can be rearranged toyield (2.1), where the final equality follows from the substitution t 1/t and using the functionalequation (2.5). Alternatively, equation (2.1) follows from an immediate application of the Muntzformula (see for example, Titchmarsh [7], page 28). On making the substitution t 1/t, it is clearfrom (2.1) and (2.5) that we have

(s) =

0

t(s1/2)/2t1/4(t)dt

t=

0

t(s1/2)/2t1/4(t)dt

t,

and therefore by addition we obtain

(s) =

0

cosh[12

(s 1

2

)log t

]t1/4(t)

dt

t, (2.9)

which shows that (s) is an even function of s 1/2. Further, with s = x + iy we may deduce

(s) =

0

cosh[12

(x 1

2

)log t

]cos

(y2

log t)

t1/4(t)dt

t

+ i

0

sinh[12

(x 1

2

)log t

]sin

(y2

log t)

t1/4(t)dt

t, (2.10)

which shows in particular that =((s)) vanishes along x = 1/2.

3. Symmetries of the Riemann xi functions

Rather than focussing on (s) defined by (2.1), it is clear from the X-rays given in the followingsection that the xi function used by Riemann himself (see for example Edwards [4], page 301),namely

(s) =12

s (s 1) (s) = 12

s (s 1) (s/2) (s)s/2

, (3.1)

turns out to exhibit the simpler X-ray. However, both xi functions (s) and (s) have the samesymmetries discussed below, but here for definiteness we will discuss the Riemann xi function (s),which we assume has real and imaginary parts u(x, y) and v(x, y), thus

(s) = u(x, y) + iv(x, y), (3.2)

where s = x + iy. The specific analytical forms of u(x, y) and v(x, y) may be determined from(2.10) and (3.1), thus

u(x, y) =12

[x(x 1) y2] u(x, y) y

(x 1

2

)v(x, y),

v(x, y) =12

[x(x 1) y2] v(x, y) + y

(x 1

2

)u(x, y), (3.3)

5

where u(x, y) and v(x, y) denote the integral expressions given in (2.10). From (2.10) and (3.3)it is clear that again =((s)) vanishes along x = 1/2. Further, from (2.1) and (3.1) it is clear that(s) satisfies the functional equation

(s) = (1 s), (3.4)while since (s) is real for real s, we may deduce

(s) = (1 s), (3.5)where as usual the overbar denotes the complex conjugate. Equations (3.4) and (3.5) consti-tute the two basic symmetries of the Riemann xi function, such that if in the critical strip0 <

of their gradients as determined by (3.9) and (3.10) is equal to minus one.

4. X-rays for (s), (s), (s) and log (s)

In order to determine the two families of curves along which either the real or the imaginarypart of a complex function is zero, we employ the following algorithm for f(s) = u(s) + iv(s),where s C:

(i) Calculate values for f(s),

(ii) Replace negative values with symbol -, and positive values with +,

(iii) Draw a smooth curve along the boundary between the positive and negative values.

However, as Israel [6] points out This does not work very well near singularities of the expression,self-intersections of the curve, or points where the Jacobian determinant is zero. More sophisti-cated techniques are outlined in Bruce et al. [2]; which involve the Newton Raphson method andvector calculus.

Figure 2 shows the X-ray for the Riemann zeta function (s), which is in complete accord with thatgiven by Arias-de-Reyna [1] and the reader is referred to this paper for a comprehensive analysisof the curves. Figure 3 shows the X-ray for the Riemann xi function (s) as defined by equation(2.1). Because (s) also satisfies the two basic symmetries (3.4) and (3.5), we see that unlikeFigure 2 the curves here are symmetric, not only about the x-axis, but also about the line x = 1/2,which is a consequence of the functional identity (3.4). The ovals in the Figure arise from the poleof the zeta function (s) at s = 1, and because of the functional equation, this pole is also mirroredat s = 0. From equation (2.6) and the definition (3.1), it is clear that the additional factor s(s1)eliminates these effects, and the X-ray for (s) shown in Figure 4 confirms this fact. In this figurethere are no ovals, and all the curves have a similar shape. Moreover, a grey line cuts a black lineonly along the line x = 1/2, and we may confirm that these points of intersection correspond tothe well k

*View more*