THE ZETA FUNCTION AND ITS RELATION TO THE

PRIME NUMBER THEOREM
BEN RIFFER-REINERT
Abstract. The zeta function is an important function in mathematics. In this paper, I will demonstrate an important fact about
the zeros of the zeta function, and how it relates to the prime
number theorem.

Contents
1. Importance of the Zeta Function
2. Trivial Zeros
3. Important Observations
4. Zeros on Re(z)=1
5. Estimating 1/⇣ and ⇣ 0
6. The Function
7. Acknowledgements
References

1
4
5
7
8
9
12
12

1. Importance of the Zeta Function
The Zeta function is a very important function in mathematics.
While it was not created by Riemann, it is named after him because he
was able to prove an important relationship between its zeros and the
distribution of the prime numbers. His result is critical to the proof of
the prime number theorem.
There are several functions that will be used frequently throughout
this paper. They are defined below.
Definition 1.1. We define the Riemann zeta funtion as
1
X
1
⇣(z) =
nz
n=1

when |z|

1.
1

If 1 Q 1 P n=1 an < 1. with the exception of the number 1. Further. ⇣(z) converges for all z with Re(z) >1. Lemma 1. ⇣(z) = Y p 1 1 p z . for the first time. First we notice that 1 P n=0 1 pnz converges absolutely for all p greater than 1. then the product (1 + an ) converges. In the later proofs we will find a product form of the zeta function more useful. the product converges to 0 if and only n=1 if one if its factors is 0. Proof. We define the gamma function as (z) = Z 1 e t tz 1 dt 0 over the complex plane. Further. This was first proven by Euler showing.2. Suppose {an } is a series. that there is a relationship between the prime numbers and the zeta function. The next lemma will use the fundamental theorem of arithmetic which states that every positive integer. We define the xi function for all z with Re(z) > 1 as ⇠(z) = ⇡ s/2 (s/2)⇣(s). we see that 1 P n=0 1 pnz can be .3. Through a simple manipulation. Definition 1.2 BEN RIFFER-REINERT Definition 1.2.1. Lemma 1. can be written as a unique product of primes.

We can conclude this by observing that the smallest prime is 2 and 2M > N as 2M < 2M since 2M > N. 1 p z p Now it is enough to prove that lim N P N !1 n=1 1 nz  Q⇣ p 1 1 p z ⌘ .THE ZETA FUNCTION AND ITS RELATION TO THE PRIME NUMBER THEOREM 3 rewritten as 1 1 p z . There exists an ✏ > 0 such that. z n 1 p z n=1 p ◆ N X Y✓ 1 1 1 1  1 + z + 2z + · · · + M z needs justification z n p p p n=1 pN ! 1 Y X 1 1 = 1 p z n=M z+1 pnz pN Y✓ 1 ◆  1 p z pN Y✓ 1 ◆  . any prime in this product is repeated less than M times. 1 X 1 1 1 = 1 + + + ··· nz z 2z p p p n=0 1 1 X 1 1 1 1 = z + 2z + 3z + · · · z nz p n=0 p p p p 1 1 X 1 X 1 1 = pz n=0 pnz pnz n=0 ✓ ◆ 1 X 1 1 1 = 1 pnz pz n=0 1 X 1 1 = pnz 1 p n=0 z 1 . 1 P Q⇣ 1 ⌘ 1 We now wish to show that  . This is shown below. Let M and N be positive integers such that M > N. Further. Suppose this is not true. We know that each prime number in this product must be less than or equal to N. . for every integer L. else n > N as all primes in the product are positive integers. Any positive integer n with n  N can be uniquely written as a product of primes by the fundamental theorem of arithmetic.

. The gamma function has an analytic continuation to a meromorphic function on C whose only singularities are simple poles at the negative integers. Theorem 2. these nz 1 p z p n=1 Q 1 two inequalities prove that ⇣(z) = 1 p z . p Letting an = n1z . We will soon turn to the study of the zeros in the critical strip. (z) (1 z) = . as our choice of N was arbitrary. Before this. Since none of the factors are 0 here. and they are easily expressed in a simple formula. this is a contradiction. . . an essential Theorem about the Gamma function.4 BEN RIFFER-REINERT there exists an integer s > L with s P n=1 1 nz > Q⇣ p 1 1 p z ⌘ . For all z 2 C ⇡ sin ⇡z Lemma 2. Finally we must show that the reverse inequality holds. The zeros of the zeta function which lie outside the critical strip are called the trivial zeros. We use the fundamental theorem of arithmetic to see that ◆ N X Y✓ 1 1 1 1 1 + z + 2z + · · · + M z nz p p p n=1 pN ! 1 Y X 1 1 = 1 p z n=M z+1 pnz pN Y✓ 1 ◆ 1 p z pN Y✓ 1 ◆ . 2. Together. and it has zeros nowhere else.1. Theorem 2. But. . 1. 1 p z p 1 P Q⇣ 1 ⌘ 1 As before. We derive this formula in the lemma and theorem below. it is clear that ⇣(z) converges for all z with Re(z) > 1 by Lemma 1. we will state without proof. ⇤ 2. Trivial Zeros The region 0 Re(z)  1 is called the critical strip.2. 1/ (z) is an entire function of z with simple zeros (zeros having a multiplicity of one) at z = 0. we see that ⇣(z) 6= 0 when Re(z) > 1 by lemma1 concluding the proof.1. we see that in the limit.

The only zeros of the zeta function outside of the critical strip are the negative even integers. through a Taylor expansion. z z 1 p 1 p m p p p. ⇤ (z/2) 3. log 1 1 x = for 0  x < 1. m m=1 Combining this fact with lemma 2.2 and a property of the logarithm we see that ✓ ◆ X mz Y X 1 1 p log ⇣(z) = log = log = .1.2 we see that ⇣(1 z) has no zeros when Re(z) < 0. If Re(z) > 1. 2. Lemma 3.1 we know that the simple poles of (1 z) are any z 2 {n|n 2 N}. 1. Theorem 2. First we observe 1 P xm that. We have already seen. Thus 1 is entire with simple zeros in {0.2 that the zeta function has no zeros when Re(z) > 1. By lemma 3. Important Observations Before we can prove a crucial theorem about the zeros of the zeta function on the critical strip.THE ZETA FUNCTION AND ITS RELATION TO THE PRIME NUMBER THEOREM 5 1 Proof.m . 1 has zeros at z = 2. ((1 z)/2) clearly has no zeros. Since.2 allows us to write (z) = (1 z) sin⇡⇡z .3 we see that ⇣(z) = ⇡ z 1 2 ((1 z)/2) ⇣(1 (z/2) z) By lemma 2.}. Theorem 2. Theorem 2. . Proof. First we will consider the case where Im(s) = 0.1.m for some cn ms m = 1 X cn n s n=1 0. in lemma 2.4. ⇠(z) = ⇠(1 z) for all z 2 C. we see that these cancel with the zeros of sin ⇡z. (z) ⇤ Before the proof of the trivial zeros. 4. Proof. . So.3. by theorem 2. Since (z) has no zeros. . then log ⇣(z) = Xp p. we must prove a few more properties about the zeta function. . . Using theorem 3. we will state a theorem known as the fundamental relation. . we will restrict our consideration to z < 0. concluding the proof.

so it converges whenever Re(z) > 1 giving us another analytic function.2. Lemma 3. + it and observe it Re(n z ) = Re(n ) ( it) log n = Re(e ) log n it log n = Re(e e ) log n = Re(e (cos(t log n) + i sin( t log n)) log n =e cos(t log n) = n cos(t log n). P p mz To show that converges absolutely we apply the ratio test. m m p (m+1)z m m = lim p z mz m!1 m + 1 p m!1 m + 1 =p z  1 as p 1 and z 1. P p mz By analytic continuation. It follows by calculation that this is positive.1. It now follows that log |⇣ 3 ( ) ⇣ 4 ( + it) ⇣ ( + 2it) | = 3 log |⇣( )| + 4 log |⇣( + it)| + log |⇣( + 2it)| = 3Re(log ⇣( )) + 4Re(log ⇣( + it)) + Re(log ⇣( + 2it)) 1 X = cn n (3 + 4 cos(t log n) + cos 2t log n) n=1 By lemma 3.6 BEN RIFFER-REINERT We will now show that the above double sum converges absolutely. Now. we now have log ⇣(z) = whenever m p. If 1 m m if n = p and cn = 0 otherwise. P p mz To show that converges absolutely we observe that the quanm lim p tity mz is greater than one since m is at least one and z is greater than one. if we let cn = have completed the proof. p 1. Since the double sum converges absolutely the order of the summation does not matter. We can let z = 0.m Re(z) > 1. we ⇤ > 1 and Im(t) = 0 then log |⇣ 3 ( ) ⇣ 4 ( + it) ⇣ ( + 2it) | Proof. The double sum converges absolutely. P p mz P p mz It suffices to show that and both converge absolutely m m p m when m. ⇤ .

then 1/f is holomorphic at this point.1.2. The zeta function has no zeros on the line Re(z) = 1. Two important concepts used in this proof are those of zeros and poles. ⇣ is homomorphic at z0 .3 as the logarithm of a number less than one is negative concluding the proof. a non-vanishing holomorphic function g on U . Then there exists a neighborhood U ⇢ ⌦ of z0 . Theorem 4.3. Theorem 4. there exists a C > 0 such that |⇣( + it0 )|4 lim  1. by theorem 4. ⇤ . so. Since ⇣ is holomorphic at + 2it0 for all . We will now state two important theorems about zeros and poles. |⇣( + 2it0 ) stays bounded as approaches 1 by continuity.1. and f (z) 6= 0 for all z 2 ⌦. Let f be holomorphic in a connected open set ⌦ such that it has a zero at z0 2 ⌦. Theorem 4. Thus lim |⇣ 3 ( )⇣ 4 ( + it)⇣( + 2it)| = 0. if 1/f is defined to be zero. A pole of a functionf is a point at which. by theorem 4.2 (as z=1 is a simple pole). Zeros on Re(z)=1 We are almost ready to prove an important theorem about the zeros of the zeta function in the critical strip. we know there exists a C 0 > 0 such that |⇣( )|3 !1 C 0 ( 1) lim 3  1.THE ZETA FUNCTION AND ITS RELATION TO THE PRIME NUMBER THEOREM 7 4. We are now ready to prove our theorem. Suppose that there exists a point z0 = 1 + it0 with t0 6= 0 such that ⇣(z0 ) = 0. and a unique positive integer n such that f (z) = (z z0 )n g(z) for all z 2 U . A zero of a function is a point at which that function vanishes. Proof. !1 But this contradicts lemma 4. If f has a pole at z0 2 ⌦ then in a neighborhood of that point there exist a non-vanishing holomorphic function h and a unique positive integer n such that f (z) = (z z0 ) n h(z). !1 C( 1)4 Similarly.

If the inequality does not hold. Now choosing A = 2cc 00 we get c0 ( 0 1)( 3/4)|t| ✏/4 = 2c00 ( Thus. Now We look at two cases. With this theorem and the ones preceding it we will finally be able to analyse the theorems which eventually lead to the prime number theorem. Now we use the mean value theorem with theorem 6. This now implies |⇣( + it)| |⇣( 0 + it)| 0 |⇣( + it) ⇣( + it)|.1 to get |⇣( + it) ⇣( + it)|  c00 | using (1) and setting = c0 ( ⇣( + it)| 0 4 0 0 ||t|✏  c00 | 0 1||t|✏ .2. 0  Now we prove a theorem dealing with the growth of 1/⇣.1 we get |⇣ 4 ( + it)| c|⇣ 3 ( )||t| c0 ( ✏ 1)3 |t|✏ for all 1 and |t| 1. Applying theorem 6. The first is the case where 1 A|t| 5✏ holds. For every ✏ > 0. then we take 0 > with 0 1 = A|t| 5✏ . we have 1/|⇣(z)|  c✏ |t|✏ when z = + it. if 0  and |t| 1 (2) |⇣ 0 (z)|  c✏ |t|1 0 +✏ . we have |⇣( + it)| A00 |t| 4✏ completing the proof. We will now prove theorems that will deal with this. (1) now gives us |⇣( + it)| A0 |t| 4✏ .3 we know that |⇣ 3 ( )⇣ 4 ( +it)⇣( + 2it)| 1 when 1 else we end up with a contradiction. t 2 R. Theorem 5. Estimating 1/⇣ and ⇣ 0 The proof of the prime number requires knowledge of the rates of growth of 1/⇣ and ⇣ 0 . and every ✏ > 0. Theorem 5. We will choose a constant A later in the proof. 0 1)|t|✏ . we see that 0 1)3/4 |t| ✏/4 c00 ( 1)|t|✏ . and |t| 1.1.[1] handling the growth of ⇣ 0 . there exists a constant c✏ .8 BEN RIFFER-REINERT 5. ⇤ . 1. By taking the a fourth root we arrive at the equation (1) c0 ( |⇣( + it)| 1)3/4 |t|✏/4 when 1 and |t| 1. Then for each 0  1. Suppose s = + it with . From the poof of Theorem 5. so that (1) |⇣(z)|  c✏ |t|1 0 +✏ . Proof. if 1  and |t| 1 0. First we will state a theorem given in Complex Analysis.

We define ⇤ = log p if n = pm for some prime p and m 1. we have x lim (x) = 1. Theorem 6. If (x) ⇠ x as x ! 1.THE ZETA FUNCTION AND ITS RELATION TO THE PRIME NUMBER THEOREM 9 6. log p log p px px Now we divide by x to find (x)  ⇡(x)xlog x . p is a prime number and m is a positive integer. We can now create two other definition.now we have X (x) = ⇤(n) 1nx and (x) = X  log x px log p log p where [x] is the greatest integer function. X  log x X log x (x) = log p  log p = ⇡(x) log x. So. x↵ <px . We will now prove just how closely related the function is to the distribution of prime numbers. The Function Finally we will look at a function which imitates the distribution of the primes. x!1 x P P Now. Since (x) ⇠ x. x!1 x!1 x x Now notice the following bound on the function which comes from the definition of the greatest integer function. Thus. and 0 otherwise. and notice that (x) log p log p px (⇡(x) ↵ ↵ ⇡(x )) log x . We will prove this by showing that the following two inequalities hold log x log x 1  lim inf ⇡(x) and lim sup ⇡(x)  1. pm x In this notation. the first inequality is clearly true. It was created by Tchebychev.1. now (x) + ↵⇡(x↵ ) log x ↵⇡(x) log x. We define X (x) = log p. In the following theorems ⇡(x) denotes the number of primes less than or equal to x. then ⇡(x) ⇠ x/ log x as x ! 1. Proof. fix 0 < ↵ < 1.

3. Lemma 6. We define Z x (u)du. It is actually easier to work with the 1 function.1. Theorem 6.2.And so we have Thus (x) x  (x)  1/2( + 1) x!1 x Thus. . If 1 (x) ⇠ x2 /2 as x ! 1. x!1 x ⇤ This concludes the proof.[1]. By theorem 6.10 BEN RIFFER-REINERT dividing by x gives (x) ↵⇡(x↵ ) log x + x x ↵⇡(x↵ ) log x 1+ x 1 1 ↵⇡(x) log x x ↵⇡(x) log x since (x) ⇠ x x logx(⇡(x) ⇡(xc antarell↵ ↵ x log x ↵ lim sup ⇡(x) as ↵ < 1. For all c > 1 ✓ 0 ◆ Z c+i1 1 xs+1 ⇣ (s) ds 1 (x) = 2⇡i c i1 s(s + 1) ⇣(s) With this theorem we are now ready to prove the fact about the zeta function which lead to the prime number theorem. 1 (x) = 1 We will now show that the same asymptotic properties applies to this function. 1 ⇠ x2 2 as x ! 1. Similarly. ). ⇤ We now state a lemma from Complex Analysis. (1 ↵)x ↵x x( 1) x We now immediately have (x)  ( 1 ( 1 ( x) 1)x ( x)2 1 (x) x2 ( 1 ( 1( 1)x x) 1 (x)). Theorem 6. we have proved that lim sup (x)/x  1. Proof.1 we only need to prove that (x) ⇠ x as x ! 1. then we have the following as is an increasing function Z x Z x 1 1 (u)du  (x)  (u)du. we can show lim sup x!1 lim inf (x)/x x!1 1 concluding the proof. then and therefore ⇡(x) ⇠ x/logx as x ! 1 (x) ⇠ x as x ! 1. If ↵ < 1 < .

) s(s + 1) R If we let s 2 1 . we can pick a greater than zero such that the zeta function has no zeros when |Im(s)|  T and 1  Re(s)  1 as the zeta function has no zeros on Re(s) = 1. 2⇡i (T ) 2 2⇡i (T. t2 T 1 R Since this converges we can find T large enough so that | 1 F (s)ds|  R ✏ 2 x and | 5 F (s)ds|  2✏ x2 . ). For now we assume that x 2.THE ZETA FUNCTION AND ITS RELATION TO THE PRIME NUMBER THEOREM 11 R c+i1 xs+1 ⇣ ⇣ 0 (s) ⌘ 1 Proof. By theorem 5. when c > 1. For any T. ✓ 0 ◆ xs+1 ⇣ (s) Let F (s) = . We fix a T to satisfy this and pick a 2 small enough as well.2 we can find a C such that Z Z 1 1/2 |t| 2 | F (s)ds|  Cx dt. the residue of F (s) at s = 1 is x2 . . ) 1 + i1 5 s q= 1 qs = 1 4 3 s= q 1 2 1 c i1 1 i1 1 i1 An application of Cauchy’s theorem shows that Z c+i1 Z 1 1 F (s)ds = F (s)ds. we see that |x1+s | 1 F (s)ds|  2✏ x2 | = x2 . Thus we find Z Z 1 x2 1 xs+1 F (s)ds = + F (s)ds. s(s + 1) ⇣(s) We define the paths below using T to determine the size of the “box” containing the point s. 2 By calculation. We will now transform this to integrate over Re(s) = 1. In ⇣(s) c i1 s(s+1) this integral. By lemma 6. F (s) has a simple pole at s = 1. we integrate on the line Re(s) = c. Re(s) = 1 c + i1 (T) 1 + i1 (T. 2⇡i c i1 2⇡i (T ) Now we need to show the equivalence of the integral over (T.1 we know 1 (x) = 2⇡i ds.

Stein. Rami Shakarchi. 2003. we can find a C 0 such R on 3 . log x Thus. log x 1 2 Thus. . We can also approximate 2 and 4 by Z Z 1 x2 00 | F (s)ds|  C x1+ d  C 00 . References [1] Elias M. Princeton University Press. New Jersey. we have | 2 1 (x) x2 1|  4✏ ⇤ 7. we have 0 2 that | 3 F (s)ds|  C x . Acknowledgements I would like to thank my mentor Mohammad Abbas-Rezaei for helping me select my topic and guiding me along the way.12 BEN RIFFER-REINERT Since. we now have | 1 (x) x2 |  ✏x2 + C 0 x2 2 + C 00 x2 . Complex Analysis. |x1+s | = x1+1 = x2 .