Riemann zeta function zeros are the zeros of the Riemann zeta function and come in two different types. So-called "trivial zeros" occur at all negative even
integers
,
,
, ..., and "nontrivial
zeros" occur at points
in the "critical
strip"
.
A nontrivial zero is denoted
.
The zeros with positive imaginary part, ordered
by increasing imaginary part and counted with multiplicity,
are commonly denoted
(Brent 1979; Edwards 2001, p. 43).
Wiener (1951) showed that the prime number theorem is literally equivalent to the assertion that has no zeros on
(Hardy 1999, p. 34; Havil 2003, p. 195). The
Riemann hypothesis asserts that the nontrivial
zeros of
all have real part
, a line called the "critical
line."
The quasi-Riemann hypothesis requires only that some fixed real number bound the real part
of every nontrivial zero from above. OpenAI (2026)
announced nonvanishing for
,
with the pole at
excluded. This claim does not place all nontrivial
zeros on the critical line
.
An attractive poster plotting zeros of the Riemann zeta function on the critical line together with annotations for relevant historical information, illustrated above, was created by Wolfram Research (1995). Enlarge the poster (PDF).
The plots above show the real and imaginary parts of plotted in the complex plane together with the complex
modulus of
.
As can be seen, in the right half-plane, the function is fairly flat, but with a
large number of horizontal ridges. It is precisely along these ridges that the nontrivial
zeros of
lie.
The plot above shows the level curves (red) and
(blue) in the complex
plane of
. Along each curve, only the indicated component of
is zero; the other component can vary. The zeros (indicated as black dots) occur
where the curves intersect.
The figures above highlight the zeros in the complex plane by plotting (where the zeros are dips) and
(where the zeros are peaks).
The above plot shows for
between 0 and 60. As can be seen, the first few nontrivial
zeros occur at the values given in the following table (Wagon 1991, pp. 361-362
and 367-368; Havil 2003, p. 196; Odlyzko), where the corresponding negative
values are also roots. The integers closest to these values are 14, 21, 25, 30, 33,
38, 41, 43, 48, 50, ... (OEIS A002410). The
numbers of nontrivial zeros with positive imaginary
part less than 10,
,
, ..., counted with multiplicity,
are 0, 29, 649, 10142, 138069, 1747146, ... (OEIS A072080;
Odlyzko).
| OEIS | ||
| 1 | A058303 | 14.134725 |
| 2 | 21.022040 | |
| 3 | 25.010858 | |
| 4 | 30.424876 | |
| 5 | 32.935062 | |
| 6 | 37.586178 |
The so-called xi-function defined by Riemann has precisely the same zeros as the
nontrivial zeros of
with the additional benefit that
is entire and
is purely real, making its zeros simpler to locate.
Brent et al. (1982) found exactly simple zeros on the critical
line with
and verified that this range contains
no other zeros with positive imaginary
part. ZetaGrid was a distributed computing project for calculating zeros.
It reported 1029.9 billion zeros as of Feb. 18, 2005.
The 900 billion zeros reported in 2004 correspond to a height
of about
(Platt and Trudgian 2021). The project reported different totals over time, so a
single cutoff height should not be inferred from those reports. Gourdon (2004) used
an algorithm of Odlyzko and Schönhage to calculate the first
nontrivial zeros,
reaching a height of approximately
(Pegg 2004, Pegg and Weisstein 2004, Platt
and Trudgian 2021).
Platt and Trudgian (2021) used interval arithmetic to verify that all nontrivial zeros
with positive imaginary part up to
lie on the critical
line. The table lists historical benchmarks by count, including both reported
computations and rigorous verifications (Gourdon 2004, Platt and Trudgian 2021).
| year | author | |
| 1903 | 15 | J. P. Gram |
| 1914 | 79 | R. J. Backlund |
| 1925 | 138 | J. I. Hutchinson |
| 1935 | E. C. Titchmarsh | |
| 1953 | A. M. Turing | |
| 1956 | D. H. Lehmer | |
| 1956 | D. H. Lehmer | |
| 1958 | N. A. Meller | |
| 1966 | R. S. Lehman | |
| 1968 | J. B. Rosser, J. M. Yohe, L. Schoenfeld | |
| 1977 | R. P. Brent | |
| 1979 | R. P. Brent | |
| 1982 | R. P. Brent, J. van de Lune, H. J. J. te Riele, D. T. Winter | |
| 1983 | J. van de Lune, H. J. J. te Riele | |
| 1986 | J. van de Lune, H. J. J. te Riele, D. T. Winter | |
| 2001 | J. van de Lune (unpublished) | |
| 2004 | S. Wedeniwski | |
| 2004 | X. Gourdon and P. Demichel | |
| 2021 | D. Platt and T. Trudgian |
It is not known whether all values of corresponding to nontrivial
zeros are irrational. Wolf (2020) obtained
numerical evidence supporting this conjecture by studying
continued fractions of high-precision approximations
to the first 40000 such values, but this does not prove that these values are irrational.
No zeros with order greater than one are known. While the existence of such zeros would not disprove the Riemann hypothesis, it would cause serious problems for many current computational techniques (Derbyshire 2004, p. 385).
Knausgård (2026) reported unconditional asymptotic lower bounds for distinct and simple nontrivial zeros. Let count zeros with imaginary part in
, including multiplicity, and let
,
,
and
count distinct zeros, simple zeros, and simple zeros on the critical
line, respectively. The bounds are
|
(1)
| |||
|
(2)
| |||
|
(3)
|
Knausgård (2026) describes the work as AI-assisted. The Lean development uses three additional axioms recording external computation results, so those finite checks are not verified by Lean's kernel.
A different bound concerns the union of simple zeros and zeros on the critical line. Let count nontrivial zeros with imaginary
part in
that satisfy at least one of these properties, counted with multiplicity.
Cobeli and Zaharescu (2026) proved unconditionally that
|
(4)
|
Their argument gives a lower bound approximately 0.88893. This is not a separate bound of
for either simple zeros or zeros on the critical line.
Some nontrivial zeros lie extremely close together, a property known as Lehmer's phenomenon.
The Riemann zeta function can be factored over its nontrivial zeros as the Hadamard product
|
(5)
|
(Titchmarsh 1987, Voros 1987).
For a positive integer , define the sum over all nontrivial zeros
, counted with multiplicity,
by
|
(6)
|
(Lehmer 1988, Keiper 1992, Finch 2003, p. 168), sometimes also denoted
(e.g., Finch 2003, p. 168). For
,
the sum is interpreted by symmetric height limits
as
.
For
, it is absolutely
convergent. The functional equation pairs
each zero
of positive imaginary part with
of negative imaginary part, including multiplicity. Thus,
with
running over the zeros with
,
|
(7)
|
This paired sum is absolutely convergent for every
and does not require the Riemann hypothesis.
Such sums can be computed analytically, and the first few are
|
(8)
| |||||
|
(9)
| |||||
|
(10)
| |||||
|
(11)
| |||||
|
(12)
| |||||
|
(13)
|
where
is the Euler-Mascheroni constant,
are Stieltjes constants,
is the Riemann zeta
function, and
is Apéry's constant.
These values can also be written in terms of the Li constants (Bombieri and Lagarias
1999).
The case
|
(14)
|
(OEIS A074760; Edwards 2001, p. 160) is classical and was known to Riemann, who used it in his computation of the roots of
(Davenport 1980, pp. 83-84; Edwards 2001, pp. 67 and 159). It is also equal
to the constant
from Li's criterion.
Assuming the truth of the Riemann hypothesis (so that ), equation (7) can be written
for the first few values of
in the simple forms
|
(15)
| |||
|
(16)
| |||
|
(17)
| |||
|
(18)
|
and so on.
Zeros of the Riemann zeta function on the critical line are implemented in the Wolfram
Language as ZetaZero[k],
where a positive integer selects the
th zero with positive imaginary
part.
Assuming the Riemann hypothesis, Bondarenko and Heap (2026) proved
|
(19)
|
Here the positive ordinates are counted with multiplicity.
The result gives normalized gaps at most
in the infimum limit; it
does not assert a strict inequality.
Durkan and Pearce-Crump (2026) extended the half-spacing infimum limit bound unconditionally to every fixed primitive number
theoretic character. Under the Riemann hypothesis,
they also proved that, for every ,
a positive proportion of consecutive Riemann
zeta function zero gaps have normalized size less than
. Under the corresponding hypothesis for a
fixed primitive Dirichlet L-series,
they obtained
|
(20)
|
Here
are its positive nontrivial zero ordinates counted with multiplicity.
The Riemann zeta function is included as
the conductor-one case. Durkan and Pearce-Crump (2026) report using ChatGPT to run
the large-gap optimizations.