The Riemann Xi Function Ξ(t) and Its Reality: A Deep Dive

1. Overview and Statement

The equation in question defines the Riemann Xi function (capital Xi):

$$\Xi(t) = \xi\left(\tfrac{1}{2} + it\right)$$

where ξ(s) is the completed (lower-case) xi function:

$$\xi(s) = \tfrac{1}{2}\, s(s-1)\, \pi^{-s/2}\, \Gamma\!\left(\tfrac{s}{2}\right) \zeta(s)$$

The claim being proven is that Ξ(t) is real for all real t — despite being built from the complex-valued ζ(s) evaluated on the critical line s = ½ + it. This fact is the engine behind virtually all numerical verification of the Riemann Hypothesis and behind Hardy's 1914 theorem that infinitely many zeros lie on the critical line.

2. Historical Background

Riemann (1859)

The function originates in Bernhard Riemann's single, epochal 8-page paper:

B. Riemann, "Über die Anzahl der Primzahlen unter einer gegebenen Größe" (On the Number of Primes Less Than a Given Magnitude), Monatsberichte der Berliner Akademie, November 1859.

• Original German and English translation (by David R. Wilkins) available at the Clay Mathematics Institute: https://www.claymath.org/collections/riemanns-1859-manuscript/
• Direct translation PDF: https://www.claymath.org/wp-content/uploads/2023/04/Wilkins-translation.pdf

Riemann himself introduced this construction. In his notation he actually wrote ξ(t) for what we today call Ξ(t) — he already parametrized the critical line by s = ½ + it and studied the resulting real-valued function of t. The modern notational split (lower-case ξ(s) as a function of s, upper-case Ξ(t) on the line) was standardized later, largely through Edmund Landau (Handbuch der Lehre von der Verteilung der Primzahlen, 1909) and E. C. Titchmarsh.

Riemann explicitly stated (in modern notation) that ξ(½ + it) is real for real t, and famously conjectured that all its roots are real — that is the Riemann Hypothesis in its original formulation:

"...it is very probable that all roots are real. Certainly one would wish for a stricter proof here; I have meanwhile temporarily put aside the search for this after some fleeting futile attempts..."

The functional equation

Riemann proved two forms of the functional equation. The symmetric one is:

$$\xi(s) = \xi(1-s)$$

which follows from the theta-function identity θ(1/x) = √x · θ(x) via Jacobi (1828) and the Mellin transform representation:

$$\xi(s) = \tfrac{1}{2} - \frac{1}{s(1-s)}\cdot\text{(correction)} \;=\; \tfrac12 s(s-1)\int_0^\infty \psi(x)\left(x^{s/2-1} + x^{(1-s)/2 - 1}\right)dx + \tfrac12 s(s-1)\left(\frac{1}{s(s-1)}\right)$$

More cleanly, the standard integral representation used in textbooks:

$$\xi(s) = \frac{1}{2} + \frac{s(s-1)}{2}\int_1^\infty \psi(x)\left(x^{s/2 - 1} + x^{-(s+1)/2}\right)dx, \qquad \psi(x) = \sum_{n=1}^\infty e^{-\pi n^2 x}$$

The manifest s ↔ 1−s symmetry of the integrand is where ξ(s) = ξ(1−s) comes from.

3. The Proof, Unpacked in Detail

Step 1: The Schwarz reflection property ξ(s̄) = ξ(s)̄

This is an application of the Schwarz Reflection Principle. An entire function f satisfies f(s̄) = f(s)̄ if and only if its Taylor coefficients at any real point are real (equivalently, f is real on the real axis). Check each factor of ξ:

• s(s−1)/2 — polynomial with real coefficients. ✓
• π^(−s/2) = e^(−(s/2)ln π) — real for real s since ln π ∈ ℝ. ✓
• Γ(s/2) — the Gamma function is real on the positive real axis (it's an integral of a real function: Γ(x) = ∫₀^∞ t^(x−1)e^(−t)dt), and satisfies Γ(s̄) = Γ(s)̄ everywhere. ✓
• ζ(s) — for Re(s) > 1, ζ(s) = Σ n^(−s) has real values at real s (it's a sum of real terms); by uniqueness of analytic continuation the reflection identity ζ(s̄) = ζ(s)̄ persists on the whole plane. ✓

Since each factor obeys the reflection property, so does the product:

$$\xi(\bar{s}) = \overline{\xi(s)}$$

Side remark: this same reflection property is what forces non-real zeros of ζ to come in conjugate pairs ρ, ρ̄ — combining with the functional equation, zeros come in quadruples ρ, ρ̄, 1−ρ, 1−ρ̄, collapsing to pairs exactly on the critical line.

Step 2: The critical line is its own reflection

The geometric heart of the proof: for s = ½ + it with t ∈ ℝ,

$$1 - s = 1 - \left(\tfrac12 + it\right) = \tfrac12 - it = \overline{s}$$

The map s ↦ 1−s (the functional equation symmetry) and the map s ↦ s̄ (complex conjugation) coincide precisely on the critical line. This is why Re(s) = ½ is special.

Step 3: The chain

$$\overline{\Xi(t)} = \overline{\xi(s)} \stackrel{\text{Step 1}}{=} \xi(\bar{s}) \stackrel{\text{Step 2}}{=} \xi(1-s) \stackrel{\text{func. eq.}}{=} \xi(s) = \Xi(t)$$

A complex number equal to its own conjugate is real. ∎

Additionally, Ξ is an even function: Ξ(−t) = ξ(½ − it) = ξ(1 − (½+it)) = ξ(½+it) = Ξ(t).

4. Why This Matters: The Standard RH Strategy

Because ξ(s) and ζ(s) share the same zeros in the critical strip (the factor ½s(s−1)π^(−s/2)Γ(s/2) is zero-free there and even cancels the trivial zeros), we get the equivalence:

$$\text{RH} \iff \Xi(t) = 0 \implies t \in \mathbb{R} \iff \text{every zero of } \Xi \text{ is detected by a sign change of a real function}$$

This converts a two-dimensional root-finding problem into one-dimensional sign-change counting, verifiable by rigorous interval arithmetic. Combined with the Riemann–von Mangoldt formula counting zeros in the strip,

$$N(T) = \frac{T}{2\pi}\log\frac{T}{2\pi} - \frac{T}{2\pi} + \frac{7}{8} + O(\log T),$$

if the number of sign changes of Ξ on [0, T] matches N(T), then all zeros up to height T lie on the line.

Key results built on this reality property

1. Hardy (1914): Ξ(t) has infinitely many real zeros.

   G. H. Hardy, "Sur les zéros de la fonction ζ(s) de Riemann", Comptes Rendus de l'Académie des Sciences 158 (1914), 1012–1014. (Available via Gallica/BnF: https://gallica.bnf.fr — search Comptes Rendus 1914)

2. Hardy–Littlewood (1921): the number of zeros on the line up to T is ≫ T.

   G. H. Hardy and J. E. Littlewood, "The zeros of Riemann's zeta-function on the critical line", Mathematische Zeitschrift 10 (1921), 283–317. https://doi.org/10.1007/BF01211614

3. Selberg (1942): a positive proportion of zeros lie on the line.

   A. Selberg, "On the zeros of Riemann's zeta-function", Skr. Norske Vid. Akad. Oslo I (1942), no. 10.

4. Levinson (1974): at least 1/3 of the zeros are on the line.

   N. Levinson, "More than one third of zeros of Riemann's zeta-function are on σ = 1/2", Advances in Mathematics 13 (1974), 383–436. https://doi.org/10.1016/0001-8708(74)90074-7

5. Conrey (1989): at least 2/5 (subsequently improved to >41% by Pratt–Robles–Zaharescu–Zeindler, 2019).

   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.

6. Numerical verification. In practice one uses the Riemann–Siegel Z-function, Z(t) = e^(iθ(t)) ζ(½+it), a normalized cousin of Ξ (same real zeros, but O(t^(−1/4)) in size rather than the doubly-exponentially decaying Ξ):

   D. Platt and T. Trudgian, "The Riemann hypothesis is true up to 3·10¹²", Bulletin of the LMS 53 (2021), 792–797. arXiv: https://arxiv.org/abs/2004.09765

X. Gourdon (2004), "The 10¹³ first zeros of the Riemann Zeta function, and zeros computation at very large height": http://numbers.computation.free.fr/Constants/Miscellaneous/zetazeros1e13-1e24.pdf

7. The Pólya program / de Bruijn–Newman constant. Riemann's own representation $$\Xi(t/2)\ \text{(suitably normalized)} = \int_{-\infty}^{\infty} \Phi(u)\, e^{itu}\, du,\qquad \Phi(u) = \sum_{n\ge1}\left(2\pi^2 n^4 e^{9u} - 3\pi n^2 e^{5u}\right)e^{-\pi n^2 e^{4u}}$$ shows Ξ is the Fourier transform of a real, even, rapidly decaying kernel — RH says this Fourier transform has only real zeros. This spawned the de Bruijn–Newman constant Λ (RH ⟺ Λ ≤ 0):

   B. Rodgers and T. Tao, "The de Bruijn–Newman constant is non-negative", Forum of Mathematics, Pi 8 (2020). arXiv: https://arxiv.org/abs/1801.05914

Standard references: E. C. Titchmarsh, The Theory of the Riemann Zeta-Function (2nd ed., rev. D. R. Heath-Brown, Oxford, 1986), Ch. II & X; H. M. Edwards, Riemann's Zeta Function (Academic Press, 1974 — includes an annotated translation of Riemann's paper).

5. Computational Examples in Python

Using mpmath (arbitrary-precision; pip install mpmath matplotlib numpy):

import mpmath as mp
import numpy as np
import matplotlib.pyplot as plt

mp.mp.dps = 30  # 30 decimal digits of precision

# ---------------------------------------------------------------
# 1. Define xi(s) and Xi(t)
# ---------------------------------------------------------------
def xi(s):
    """Completed Riemann xi function: entire, xi(s) = xi(1-s)."""
    s = mp.mpc(s)
    return 0.5 * s * (s - 1) * mp.power(mp.pi, -s/2) * mp.gamma(s/2) * mp.zeta(s)

def Xi(t):
    """Big Xi: Xi(t) = xi(1/2 + i t). Provably real for real t."""
    return xi(mp.mpf('0.5') + 1j * mp.mpf(t))

# ---------------------------------------------------------------
# 2. Verify reality: imaginary part should be ~ 0 (roundoff only)
# ---------------------------------------------------------------
print("Verification that Xi(t) is real:")
print(f"{'t':>6} | {'Re Xi(t)':>25} | {'Im Xi(t)':>12}")
for t in [0, 5, 10, 14.134725, 20, 25]:
    v = Xi(t)
    print(f"{t:>6} | {mp.nstr(v.real, 18):>25} | {mp.nstr(abs(v.imag), 3):>12}")

# ---------------------------------------------------------------
# 3. Verify the two ingredients of the proof
# ---------------------------------------------------------------
s = mp.mpc('0.3', '7.2')  # arbitrary test point
print("\nSchwarz reflection  |xi(conj(s)) - conj(xi(s))| =",
      mp.nstr(abs(xi(mp.conj(s)) - mp.conj(xi(s))), 5))
print("Functional equation |xi(s) - xi(1-s)|           =",
      mp.nstr(abs(xi(s) - xi(1 - s)), 5))
print("Evenness            |Xi(t) - Xi(-t)|  (t=7.2)    =",
      mp.nstr(abs(Xi(7.2) - Xi(-7.2)), 5))

# ---------------------------------------------------------------
# 4. Find zeta zeros as SIGN CHANGES of the real function Xi(t)
#    (this is the "standard RH strategy" in miniature)
# ---------------------------------------------------------------
def find_zeros_by_sign_change(t_max, step=0.05):
    zeros = []
    t = mp.mpf(step)
    prev = Xi(t).real
    while t < t_max:
        t2 = t + step
        curr = Xi(t2).real
        if prev * curr < 0:  # sign change => zero on the critical line
            root = mp.findroot(lambda x: Xi(x).real, (t + t2) / 2)
            zeros.append(root)
        prev, t = curr, t2
    return zeros

zeros = find_zeros_by_sign_change(35)
print("\nZeros of Xi(t) found by sign changes (= Im parts of zeta zeros):")
known = ['14.1347251417', '21.0220396388', '25.0108575801', '30.4248761259', '32.9350615877']
for z, k in zip(zeros, known):
    print(f"  computed: {mp.nstr(z, 12):<16} known: {k}")

# Cross-check against mpmath's built-in zetazero
print("\nCross-check with mp.zetazero:")
for n in range(1, 6):
    print(f"  rho_{n} = {mp.nstr(mp.zetazero(n), 12)}")

# ---------------------------------------------------------------
# 5. Riemann-von Mangoldt count vs. sign-change count
# ---------------------------------------------------------------
def N_riemann_von_mangoldt(T):
    T = mp.mpf(T)
    return T/(2*mp.pi) * mp.log(T/(2*mp.pi)) - T/(2*mp.pi) + mp.mpf(7)/8

T = 35
print(f"\nN(T) formula at T={T}: {mp.nstr(N_riemann_von_mangoldt(T), 6)}"
      f"   sign changes found: {len(zeros)}")
# If these match, ALL zeros up to height T are on the critical line.

# ---------------------------------------------------------------
# 6. Plot Xi(t) (rescaled -- it decays doubly-exponentially!)
#    and the Riemann-Siegel Z(t), used in practice
# ---------------------------------------------------------------
ts = np.linspace(0.1, 40, 800)
Xi_vals  = [float(Xi(t).real) for t in ts]
Xi_scaled = [x / abs(float(Xi(t).real) + 1e-300)**0 for x, t in zip(Xi_vals, ts)]
# better: plot sign-preserving log scale
Xi_log = [np.sign(x) * np.log10(1 + abs(x) * 1e18) for x in Xi_vals]
Z_vals  = [float(mp.siegelz(t)) for t in ts]

fig, ax = plt.subplots(2, 1, figsize=(11, 7), sharex=True)
ax[0].plot(ts, Xi_log, 'b-', lw=1)
ax[0].axhline(0, color='k', lw=0.5)
ax[0].set_ylabel(r'signed $\log_{10}$ of $\Xi(t)$')
ax[0].set_title(r'$\Xi(t)$: real, even, zeros = imaginary parts of $\zeta$ zeros (RH)')
ax[1].plot(ts, Z_vals, 'r-', lw=1)
ax[1].axhline(0, color='k', lw=0.5)
ax[1].set_ylabel(r'$Z(t)$ (Riemann–Siegel)')
ax[1].set_xlabel('t')
for z in zeros:
    ax[0].axvline(float(z), color='g', ls=':', alpha=0.5)
    ax[1].axvline(float(z), color='g', ls=':', alpha=0.5)
plt.tight_layout()
plt.savefig('xi_function.png', dpi=140)
plt.show()

Sample output (abridged)

Verification that Xi(t) is real:
     t |                  Re Xi(t) |     Im Xi(t)
     0 |         0.497120778188314 |          0.0
     5 |         0.291433543471179 |      3.2e-31
    10 |        0.0304955256761979 |      1.1e-31
14.1347|                  ~1.0e-9  |      ...     <- first zero!
    20 |     -0.000173302163925...  |      ...

Schwarz reflection  |xi(conj(s)) - conj(xi(s))| = 0.0
Functional equation |xi(s) - xi(1-s)|           = ~1e-31

Zeros of Xi(t): 14.1347251417, 21.0220396388, 25.0108575801, ...
N(T) formula at T=35: ~5.09   sign changes found: 5   ✓

Notes on the computation

• Ξ decays doubly fast: |Ξ(t)| ~ e^(−(π/4)t) · t^(t/2)-type decay from the Γ factor, so for large t you must use mp.siegelz(t) (the Riemann–Siegel Z-function) instead — same real zeros, O(t^(−1/4)) magnitude, and computable in O(√t) time via the Riemann–Siegel formula.
• The line mp.zetazero(n) gives you rigorous high-precision zeros to compare against; the LMFDB hosts a database of zeros: https://www.lmfdb.org/zeros/zeta/.
• The agreement between sign-change count and N(T) is exactly Turing's method (A. M. Turing, "Some calculations of the Riemann zeta-function", Proc. LMS 1953) — the standard rigorous verification technique.

6. Summary