Part I · Chapter 7 needs review

The zeros and the Riemann Hypothesis

A zero of ζ\zeta is a ss with ζ(s)=0\zeta(s) = 0. This chapter shows what looks like, where zeros can and cannot be, and states the .

What a zero is

ζ(s)\zeta(s) is a complex number, with a and an imaginary part. It is 0 only when both parts are 0 at the same ss. Fix the real part σ\sigma of ss and let the imaginary part tt vary: the real part of ζ\zeta crosses 0 at some values of tt, and the imaginary part of ζ\zeta crosses 0 at others. A zero of ζ\zeta needs both crossings to coincide.

Lab 7.1 the real and imaginary parts of ζ(σ + it), near a chosen height
real part of ζimaginary part of ζ● heights of the known zeros
At σ=12\sigma = \tfrac12 the two curves cross the axis together at each dot. For any other σ\sigma in the range they still each cross the axis, but no longer at the same height.

The ∣ζ∣|\zeta| combines the two parts into one number, the distance of ζ(s)\zeta(s) from 0. A zero is where it touches 0.

Lab 7.2 the size of ζ along a vertical line
|ζ(σ + it)|| ≤ P|● known zeros (σ = ½)
At σ=12\sigma = \tfrac12 the curve touches zero at each dot. Move σ\sigma to the right and every dip lifts off the axis. The orange curve is the product over of chapter 2, cut off at P. For σ>1\sigma \gt 1 it converges to ζ\zeta. For σ≤1\sigma \le 1 it does not converge, but it still dips near the zeros.

A third view: plot the point ζ(σ+it)\zeta(\sigma + it) in the and let tt increase. The point traces a path that loops around. A zero is a moment when the path passes exactly through the origin.

Lab 7.3 the path of ζ(σ + it) as t runs from 1 to T
ζ(σ + it), 1 ≤ t ≤ T● the origin● the point at t = T
Horizontal axis: real part of ζ\zeta. Vertical axis: imaginary part. At σ=12\sigma = \tfrac12 every loop passes through the origin. At σ=0.51\sigma = 0.51 every loop misses it.

Where zeros can be

Three facts, all proved, limit where ζ(s)=0\zeta(s) = 0 is possible.

Together: for σ<0\sigma \lt 0 the functional equation expresses ζ(s)\zeta(s) through ζ(1−s)\zeta(1 - s), where 1−s1 - s has real part greater than 1 and ζ\zeta is not 0. So the only zeros with σ<0\sigma \lt 0 are the trivial ones. Every other zero lies in the band 0≤σ≤10 \le \sigma \le 1, called the . The line σ=12\sigma = \tfrac12 down its middle is the .

Within the strip the zeros have two symmetries. If ρ\rho is a zero, so is its complex conjugate (same real part, opposite height), because ζ\zeta takes real values on the real axis. And so is 1−ρ1 - \rho, by the functional equation. So a zero at β+iγ\beta + i\gamma that is not on the critical line comes with a partner at (1−β)+iγ(1 - \beta) + i\gamma on the other side of it.

The Riemann Hypothesis

The Riemann Hypothesis: every zero in the critical strip has real part exactly 12\tfrac12. Because zeros off the line come in mirror pairs, this is the same as saying that no zero has real part greater than 12\tfrac12.

Riemann computed the first few zeros by hand, found them on the line, and wrote that it is "very likely" that all are. What is known today:

How many zeros there are

The zeros become denser with height. Riemann stated, and proved, a formula for the number N(T)N(T) of zeros in the strip with height between 0 and TT:

N(T)  ≈  T2π ln⁡T2πe  +  78N(T) \;\approx\; \frac{T}{2\pi}\,\ln\frac{T}{2\pi e} \;+\; \frac78 The error grows only like ln⁡T\ln T.

Lab 7.4 counting zeros up to height T
number of zeros with height at most T(T/2π) (T/2πe) + 7/8
The staircase steps up by one at the height of each of the first 30 zeros. The smooth curve is the formula. Compare Lab 1.1, where a staircase of primes is set against a smooth estimate. Here the estimate is far more accurate.

Why zeros matter for primes

Chapter 2 ended with −ζ′(s)/ζ(s)=∑Λ(n)/ns-\zeta'(s)/\zeta(s) = \sum \Lambda(n)/n^{s}, the function that carries the prime weights. It has ζ\zeta in the denominator. Wherever ζ(s)=0\zeta(s) = 0, this function is infinite. So the zeros of ζ\zeta are the singular points of the function that encodes the primes.

Things to try

  • In Lab 7.1, center the window on 21.02 and on 25.01. Then set σ=0.6\sigma = 0.6 and find how far apart the two crossings have moved.
  • In Lab 7.2, set σ=12\sigma = \tfrac12 and P = 2, then 3, then 5. A few primes already place the first dip roughly.
  • In Lab 7.3, set T just above 14.13, 21.02 and 25.01 and watch the path cross the origin each time. Then set σ=0.6\sigma = 0.6.
  • In Lab 7.4, read off the formula at T=50T = 50 and count the steps below 50.

How the labs compute ζ\zeta

The defining sum does not converge inside the critical strip, and the formula of chapter 6 converges slowly. The labs use a method due to , with 70 terms, which is a rapidly converging rearrangement of the sum. Checked against published values it is accurate to about six decimal places for heights up to 60, which is why the labs stop there.

Exercises

  1. Suppose ζ\zeta had a zero at 0.3+14i0.3 + 14i. Which other zeros would it force?
    Solution

    The complex conjugate, 0.3−14i0.3 - 14i. The functional equation pairs ss with 1−s1 - s, giving 0.7−14i0.7 - 14i, and its conjugate 0.7+14i0.7 + 14i. So one zero off the line means four, two of them to the right of it.

  2. Use the functional equation to explain why ζ(−2)=0\zeta(-2) = 0.
    Solution

    At s=−2s = -2 the factor sin⁡(πs/2)=sin⁡(−π)=0\sin(\pi s/2) = \sin(-\pi) = 0. The other factors are finite: 2−22^{-2}, π−3\pi^{-3}, Γ(3)=2\Gamma(3) = 2 and ζ(3)≈1.202\zeta(3) \approx 1.202. So the product is 0.

  3. Use the counting formula to estimate how many zeros have height between 0 and 1,000.
    Solution

    10002π⋅ln⁡10002πe+78  ≈  159.15⋅ln⁡58.55+0.875  ≈  159.15⋅4.070+0.875  ≈  648.6.\frac{1000}{2\pi}\cdot\ln\frac{1000}{2\pi e} + \frac78 \;\approx\; 159.15\cdot\ln 58.55 + 0.875 \;\approx\; 159.15\cdot 4.070 + 0.875 \;\approx\; 648.6. The true count is 649.

Check your understanding

  1. In Lab 7.1 at σ=0.6\sigma = 0.6, the real part of ζ\zeta still crosses 0 many times. Why are those not zeros of ζ\zeta?

    • Because 0.6 is not a whole number
    • A zero needs the real and the imaginary part to be 0 at the same point, and there the imaginary part is not 0
    • They are zeros

    One real equation in one unknown usually has solutions. A zero of ζ\zeta is two equations at once. That they are satisfied together at all, and apparently only on one line, is what makes the zeros remarkable.

  2. Why can there be no zeros with real part greater than 1, and none with real part less than 0 except the trivial ones?

    • Nobody has found any
    • Because the sum diverges there
    • For σ>1\sigma \gt 1 the product over primes has no zero factor; for σ<0\sigma \lt 0 the functional equation reduces ζ(s)\zeta(s) to ζ(1−s)\zeta(1 - s), which is in the first region

    Both are theorems. What remains is the strip 0≤σ≤10 \le \sigma \le 1, where neither argument applies.

  3. In Lab 7.2 every dip lifts off the axis as σ\sigma moves from 12\tfrac12 to 0.6. If the Riemann Hypothesis were false, what would be seen somewhere far up the axis?

    • A dip that touches zero at some σ>12\sigma \gt \tfrac12, and not at 12\tfrac12
    • A dip that touches zero at every σ\sigma
    • The curve becoming negative

    A zero at β+it\beta + it off the line means ∣ζ∣=0|\zeta| = 0 on the vertical line σ=β\sigma = \beta, and by the mirror symmetry on σ=1−β\sigma = 1 - \beta, but not at 12\tfrac12. An absolute value is never negative.