EXPLICIT BOUNDS FOR THE RIEMANN ZETA-FUNCTION ON

THE 1-LINE
arXiv:2306.13289v1 [math.NT] 23 Jun 2023

GHAITH A. HIARY, NICOL LEONG AND ANDREW YANG

Abstract. Explicit estimates for the Riemann zeta-function on the 1-line are
derived using various methods, in particular van der Corput lemmas of high
order and a theorem of Borel and Carathéodory.

Let ζ(s) denote the Riemann zeta-function. An open problem in analytic number
theory is the growth rate of ζ(s) on the line s = 1 + it as t → +∞. Assuming the
Riemann hypothesis, Littlewood [Lit26; Lit28] showed that the value of |ζ(1 + it)|
is limited to the range
1
≪ ζ(1 + it) ≪ log log t.
log log t
In comparison, the sharpest estimates requiring no assumption furnish a wider
range of the form
1
(1) ≪ ζ(1 + it) ≪ log2/3 t.
(log t)2/3 (log log t)1/3
The upper bound in the unconditional estimate (1) was made explicit by Ford
[For02b]. He computed numerical constants c and t0 such that 1
ζ(1 + it) ≤ c log2/3 t, (t ≥ t0 ).
The constants c and t0 are often too large for standard applications, including ex-
plicit bounds on S(t) [Tru14b; HSW21] and explicit zero-density estimates [KLN18].
One instead defaults to an asymptotically worse bound of the form ζ(1 + it) ≪ log t
that nevertheless is sharper in a finite range of t of interest.
This motivates our Theorem 1. We establish a new explicit bound on ζ(1 + it)
that is asymptotically sharper than O(log t) but still good at small values of t.
Theorem 1. For t ≥ 3, we have
log t
|ζ(1 + it)| ≤ 1.721 .
log log t
Additionally, using a different method, we prove explicit upper bounds of the
same form, on both 1/|ζ(1 + it)| and |ζ ′ (1 + it)/ζ(1 + it)| .
Theorem 2. For t ≥ 3, we have
1 log t
≤ 430.5 .
|ζ(1 + it)| log log t

2020 Mathematics Subject Classification. Primary: 11Y35, 11M06.

Key words and phrases. Van der Corput estimate, exponential sums, Riemann zeta function.
1The admissible value of c was lowered by Trudgian [Tru14a].

1
2 G. HIARY, N. LEONG AND A. YANG

Furthermore, for t ≥ 500, if

9 log log t
1≤σ ≤1+ ,
250 log t
then
ζ′ log t
(σ + it) ≤ 154 .
ζ log log t
Theorem 1 makes explicit a result due to Weyl [Wey21] (see also [Tit86, §5.16]),
and is sharper than the prior explicit bounds on |ζ(1 + it)| for t in the range
8
exp(3070) ≤ t ≤ exp(3.69 · 108 ) ≈ 101.6·10 .2 As far as we know, the bounds in
Theorem 2 are the first explicit bounds of their kind.

1. Discussion
One can derive an explicit bound on |ζ(1 + it)| of the form log t + α1 using
a version of the Euler–Maclaurin summation. For example, using the version in
Lemma 11, applied with t0 = 30π and h0 = 1/4, together with the elementary
harmonic sum bound3,
X 1 1
(2) ≤ log N + γ + , (N ≥ 1),
n N
1≤n≤N

yields after a small numerical computation, |ζ(1 + it)| ≤ log t − 0.45 for t ≥ 3.
Patel [Pat22] replaced the use of the Euler–Maclaurin summation by an explicit
version of the Riemann–Siegel formula on the 1-line. Consequently, the size of the
main term in the Euler–Maclaurin bound was cut by half. Patel thus obtained
|ζ(1 + it)| ≤ 12 log t + 1.93 for t ≥ 3, a substantial improvement at the expense of a
larger constant term.
In general, we expect that an explicit van der Corput lemma of order k, where
k ≥ 2, will produce a bound of the form
1
(3) |ζ(1 + it)| ≤ log t + αk , (t ≥ 3).
k
One proceeds by approximating ζ(1 + it) by an Euler–Maclaurin sum n≪t n−1−it ,
P
P √ −1−it
or better yet by a Riemann–Siegel sum n≪ t n . In either case, roughly
speaking, the term (1/k) log t arises from bounding the subsum with n ≤ t1/k via
the triangle inequality and the harmonic sum estimate (2), while the constant αk
predominantly arises from bounding the subsum with n > t1/k via van der Corput
lemmas of order ≤ k.
However, the αk obtained using this approach appear to grow fast with k. For
example, in [Pat21], α5 is about 23-times larger than α2 , which already suggests
exponential increase. Despite this, there is still gain in letting k grow with t, pro-
vided the growth is slow enough. We let k grow like log log t, which is permitted
by a recent optimized van der Corput test, uniform in k, derived in [Yan23]. Sub-
sequently, we obtain a bound on |ζ(1 + it)| of the form c0 log t/ log log t with an
explicit number c0 .

2We also note that Theorem 1 is sharper for t ≥ 20400 than the bound |ζ(1 + it)| ≤ 3
log t
4
given in [Tru14a].
3Here, γ = 0.577216 . . . is the Euler constant.
 EXPLICIT BOUNDS FOR THE RIEMANN ZETA-FUNCTION ON THE 1-LINE 3

While we could let k increase faster than log log t, leading to a smaller contri-
bution of the (1/k) log t term, the αk thus obtained would likely take over as the
main term, asymptotically surpassing log t/ log log t in size. This limits the order
of the van der Corput lemmas we employ.
In proving Theorem 1, we encounter an unexpected “gap interval”, not covered
by any of the other explicit estimates, which requires a special treatment. Namely,
the interval I = [exp(16.01), exp(705.64)]. To prove the inequality in Theorem 1
for all t ∈ I, we use the following proposition.
Proposition 3. We have
1

 log t + 4.664, t ≥ exp(16),




 3



1

|ζ(1 + it)| ≤ log t + 4.308, t ≥ exp(82),


 3


8



 log t + 12.53, t ≥ exp(93).

33
To clarify the need for Proposition 3, consider that if one inputs the bounds in
[Pat22], as below, one finds
 
1 1
min log t, log t + 1.93, log t + 44.02
2 5
max ≥ 2.539,
t∈I log t/ log log t
and the maximum of this ratio occurs around t1 = exp(140.3). On the other hand,
the savings we otherwise establish here may materialize only when t is well past t1 .
Therefore, without the improvement enabled by Proposition 3, we cannot do any
better than the bound |ζ(1 + it)| ≤ 2.539 log t/ log log t over t ∈ I. Proposition 3,
which is based on a hybrid of van der Corput lemmas of orders k = 3, 4 and 5,
allows us to circumvent this barrier.
In retrospect, a gap interval like I should be expected. This is because the
estimates in [Pat21] rely, essentially, on van der Corput lemmas of orders k = 2 and
k = 5.4 In particular, bounds that would follow from using van der Corput lemmas
of orders k = 3 and k = 4 are not utilized, which presumably causes the gap.
Lastly, it is reasonable to try to adapt the remarkable approaches in [Sou09, p.
984] and [LLS15, Lemma 2.5] to our context. In these articles, very precise bounds,
conditional on the Riemann hypothesis, are obtained. One can arrange for these
approaches to yield unconditional bounds, and the outcome is similarly constrained
by the width of available explicit zeros-free regions.

2. Proof of Theorem 1
Let t0 be a positive number to be chosen later, subject to t0 ≥ exp(990/7). We
suppose t ≥ t0 throughout this section. (We will use another method for t < t0 .)
Theorem 12 supplies a Riemann–Siegel approximation of ζ(1 + it) in terms of
two sums, plus a remainder term R. We use the triangle inequality to bound the

4The bound obtained via the Riemann–Siegel formula in [Pat21] should correspond to a k = 2
van der Corput lemma.
4 G. HIARY, N. LEONG AND A. YANG

entire second sum in this theorem. This gives

n1
X 1 g(t) p
(4) |ζ(1 + it)| ≤ + √ + R(t), n1 = ⌊ t/(2π)⌋.
n=1
n1+it 2π
Since g(t) and R(t) are well-understood, the bulk of the work is in bounding the
main sum over n ≤ n1 .
We introduce two sequences, {Xm }0≤m≤M and {Yk }3≤k≤r+1 . The sequence of
the Xm will be used to partition the main sum in (4) in a dyadic manner. The
sequence of the Yk will be used to easily bound the partition points Xm from above
and below solely in terms of t (or, what is essentially the same, in terms of n1 ).
To this end, let h1 and h2 be real positive numbers, also to be chosen later,
subject to 1 < h1 ≤ 2 and h2 < 1/ log 2. We use h2 immediately in defining
R := 2r−1 , r := ⌊h2 log log t⌋, r0 := ⌊h2 log log t0 ⌋.
Since t ≥ t0 by supposition, r ≥ r0 . The integer r will correspond to the highest
order van der Corput lemma we use in bounding the tail of the main sum in (4).
Furthermore, let
2K
K := 2k−1 , θk := , (k ≥ 3).
(k − 2)K + 4
In the definition of θk , K is determined by k. Also, θk decreases monotonically to
zero with k ≥ 3, and is bounded from above by θ3 = 1.
We now define the sequences {Xm } and the {Yk }, mentioned earlier. Set
1
X0 := √ t1/2 , Xm := h−m
1 X0 , (m ≥ 1).
2π
So, ⌊X0 ⌋ = n1 and since h1 > 1, Xm decreases monotonically with m. Also set
(5) Yk := X0θk , (k ≥ 3),
so that Yk decreases monotonically with k ≥ 3, starting at Y3 = X0 . Note that if k
is large, then θk ≈ 2/(k − 2), and so Yk ≈ t1/(k−2) for large k. Lastly, define
 
(1 − θr+1 ) log X0
M := .
log h1
The integer M will correspond to the number of points we will to partition the
main sum. From the definition, we have
(1 − θr+1 ) log X0 (1 − θr+1 ) log X0
(6) ≤M < + 1.
log h1 log h1
Thus, our choice of M ensures that
θ θ
(7) XM ≤ X0 r+1 = Yr+1 < XM−1 ≤ h1 X0 r+1 .
Additionally, we have
(8) XM ≤ Yr+1 < Yr < · · · < Y4 < Y3 = X0 .
In order for the subsequent analysis to be valid, we will want to suppose r0 is
at least 3, This does not introduce any restriction for us since the van der Corput
lemmas we plan to invoke all have order at least 3. In return, the condition r0 ≥ 3
implies, among other things, that θr+1 < 1 and, hence, M ≥ 1, This will ensure
 EXPLICIT BOUNDS FOR THE RIEMANN ZETA-FUNCTION ON THE 1-LINE 5

that some quantities appearing later (e.g. M − 1) are nonnegative. The condition
also necessitates
3
(9) h2 > .
log log t0
All contingent conditions we impose (including on r0 , h1 , h2 and t0 ) will be
verified at the end, after choosing the values of our free parameters.
By the chains of inequalities (7) and (8), for all m ∈ [1, M − 1], Yr+1 < Xm < Y3 .
Therefore, since the sequence Yr+1 , . . . , Y3 partitions the interval [Yr+1 , Y3 ], for each
integer m ∈ [1, M − 1], there is a unique integer k ∈ [3, r] such that
(10) Xm ∈ (Yk+1 , Yk ].
Note, though, that multiple m’s could correspond to the same k.
With this in mind, and in preparation for applying the high order van der Corput
Lemma 16 to the subsum
X 1 X 1
(11) Sm := = ,
n1+it n1+it
Xm <n≤Xm−1 ⌊Xm ⌋<n≤⌊Xm−1 ⌋

we calculate
⌊Xm−1 ⌋ Xm−1 Xm−1 Xm−1 h1
(12) ≤ = + = h1 + .
⌊Xm ⌋ Xm − 1 Xm Xm (Xm − 1) Xm − 1
Here, we used the elementary inequality x − 1 < ⌊x⌋ ≤ x, valid for any real number
x together with the definition Xm = h−m
1 X0 . So, in view of the inequalities in (7),
and since the Xm are monotonically decreasing, if 1 ≤ m ≤ M − 1, then
⌊Xm−1 ⌋ h1 h1
(13) ≤ h1 + ≤ h1 + θr+1 .
⌊Xm ⌋ XM−1 − 1 X0 −1
Moreover, it follows easily from the definition that
2
(14) θr+1 = .
r − 1 + 22−r
So, since θr+1 decreases continuously with r, and r ≤ h2 log log t, we deduce that
θ
X0 r+1 = exp(θr+1 log X0 ) > κ(h2 , t), where
 
log(t/(2π))
(15) κ(h2 , t) := exp .
h2 log log t − 1 + 22−h2 log log t
It is easily verified that κ(h2 , t) is increasing in t for t ≥ t0 . Thus, in view of the
inequalities (13), we see that
⌊Xm−1 ⌋ h1
(16) ≤ h1 + =: h, say.
⌊Xm ⌋ κ(h2 , t0 ) − 1
Note that h is determined by h1 , h2 and t0 .
We are now ready to apply Lemma 16 to the subsum Sm . In doing so, we
will suppose that h > 1. (This contingent condition will be verified at the end.)
Lemma 16 gives that for each integer m ∈ [1, M − 1] such that Xm ∈ (Yk+1 , Yk ],
and with K = 2k−1 , we have, for any choice of the free parameter η3 > 0,
(17)
−k/(2K−2) 1/(2K−2) k/(2K−2)−2/K −1/(2K−2)
|Sm | ≤ Ck (η3 , h)Xm t + Dk (η3 , h)Xm t .
6 G. HIARY, N. LEONG AND A. YANG

We replace Xm in the second term in (17) with the upper bound Yk . This is
permissible since the exponent of Xm is nonnegative,
k 2
(18) − ≥ 0, (k ≥ 3).
2K − 2 K
Moreover, it is clear we can replace Xm in the first term with the lower bound Yk+1 .
So, we arrive at
−k/(2K−2) 1/(2K−2) k/(2K−2)−2/K −1/(2K−2)
(19) |Sm | ≤ Ck (η3 , h)Yk+1 t + Dk (η3 , h)Yk t .
It is worth clarifying that the right-side depends on m via the requirement (10).
Also, Ck and Dk additionally depend on h1 , h2 and t0 , through h.
Next, we express the bound on Sm in (19) solely in terms of t. Starting with the
first term, we have
−k/(2K−2) 1/(2K−2)
(20) Yk+1 t ≤ (2π)θk+1 k/(4K−4) t−ρ(k) ,
where after a quick calculation,
K −2 1
(21) ρ(k) := · .
K − 1 2(k − 1)K + 4
We estimate the second term in (19) similarly, yielding
k/(2K−2)−2/K −1/(2K−2) K 1
Yk t < (2π)−θk (k/(4K−4)−1/K) t− K−1 · 2(k−2)K+4
(22)
< (2π)−θk (k/(4K−4)−1/K) t−ρ(k) .
The exponent of 2π in the estimate (20) may be bounded for k > 5 via θk+1 ≤ 32/81.
Additionally, recalling the inequality (18), we may bound the contribution of the
2π-term in the estimate (22) simply by 1. This motivates us to define
(23)
k k 1
(
(2π)θk+1 4(K−1) Ck (η3 , h) + (2π)−θk ( 4(K−1) − K ) Dk (η3 , h) 3 ≤ k ≤ 5,
A(η3 , h, k) := 8k
(2π) 81(K−1) Ck (η3 , h) + Dk (η3 , h), 6 ≤ k.
Assembling (19), (20), (22) and (23), we obtain that if 1 ≤ m ≤ M − 1 and k is
determined via the requirement (10), then
(24) |Sm | < A(η3 , h, k) t−ρ(k) .
Let Nk denote the number of sums Sm corresponding to each integer k ≥ 3. In
other words, Nk is the number of m such that Xm ∈ (Yk+1 , Yk ]. We observe that
Xm ∈ (Yk+1 , Yk ] if and only if Mk ≤ m < Mk+1 , where
(1 − θk ) log X0
Mk := .
log h1
Therefore, by the bound on |Sm | in (24),
X 5
X
(25) |Sm | ≤ Nk A(η3 , h, k)t−ρ(k) .
1≤m<M6 k=3

So that, by the triangle inequality, the tail of the main sum in (4) is at most
M−1
X 5
X X
(26) |Sm | ≤ Nk A(η3 , h, k) t−ρ(k) + |Sm |.
m=1 k=3 M6 ≤m≤M−1
 EXPLICIT BOUNDS FOR THE RIEMANN ZETA-FUNCTION ON THE 1-LINE 7

By counting the maximum number of integers in a continuous interval, for k ≥ 3,

(θk − θk+1 ) log X0
(27) Nk ≤ + 1 ≤ φk log t,
log h1
where
θk − θk+1 1
(28) φk = φk (h1 , t0 ) := + .
2 log h1 log t0
Thus, we obtain
5
X
(29) Nk A(η3 , h, k) t−ρ(k) ≤ b0 ,
k=3

where by (27) and on explicitly computing the values of ρ(k) for 3 ≤ k ≤ 5,

log t0 log t0
b0 = b0 (η3 , h1 , h2 , t0 ) := A(η3 , h, 3)φ3 1/30
+ A(η3 , h, 4)φ4 3/182
t0 t0
(30)
log t0
+ A(η3 , h, 5)φ5 7/990
.
t0
Here, we used the supposition t ≥ t0 ≥ exp(990/7) which ensures that log t/tρ(k) is
decreasing for t ≥ t0 , with 3 ≤ k ≤ 5.
To handle the second sum on the right-side in (26), we use a tail argument that
removes the dependence of A on k in the bound (24). Define
(31) A0 = A0 (η3 , h) := sup A(η3 , h, k).
k≥6

We will bound A0 explicitly later, after determining η3 and h. Importantly, under

our assumptions, it will transpire that A0 is finite.
In the meantime, we observe that ρ(k) is continuously decreasing in k ≥ 3
(towards zero), as can be seen from computing the derivative of ρ(k) with respect
to k. So, by simply counting the number of terms in the last sum in (26), and
noting that as m ranges over [M6 , M − 1], k ≤ r, we obtain from (24),
X |b1 − θr+1 |
(32) |Sm | ≤ |M − M6 |A0 t−ρ(r) ≤ A1 log t,
M6 ≤m≤M−1
tρ(r)

where, using the upper bound on M in (6),

1 A0 2 log h1
A1 = A1 (η3 , h1 , h2 , t0 ) := · , b1 = b1 (h1 , t0 ) := θ6 + .
2 log h1 log t0
In indicating the dependencies of A1 , we used that A0 depends on h (as well as η3 ),
and in turn h is determined by h1 , h2 and t0 .
We want to further simplify the bound on the right-side of (32). As in (14),
2
(33) ≤ θr+1 < b1 , (r ≥ 6).
r − 1 + 22−r0
Hence, utilizing the inequality r ≤ h2 log log t as well as log(u) ≤ u − 1, valid for
any real number u > 0, yields that if r ≥ 6, then
b1 − θr+1 1 1
(34) ρ(r)
< exp(b1 − θr+1 − 1 − ρ(r) log t) ≤ ,
t A2 log t
8 G. HIARY, N. LEONG AND A. YANG

where (identifying x with log t below)

(35)  
1 2
A2 = A2 (h1 , h2 , t0 ) := inf exp 1 − b1 + + x ρ(h 2 log x) .
x≥log t0 x h2 log x − 1 + 22−r0
In indicating the dependencies of A2 , we used that r0 is determined by h2 and t0 ,
and b1 further depends on h1 . Now, from the definition of ρ, and using
2h2 log x = xh2 log 2 ,
5
we obtain
xh2 log 2 − 4 1
(36) ρ(h2 log x) = · .
xh2 log 2 − 2 (h2 log x − 1)xh2 log 2 + 4
Since 6 by supposition 0 < h2 < 1/ log 2, the quantity defining A2 tends to +∞ as
x → +∞. As this quantity is continuous and positive for x ≥ log t0 , A2 is finite
and positive for all admissible h2 .
In summary, combining (26), (32) and (34), and provided r ≥ 6, the tail sum
satisfies the inequality
X 1 A1 (η3 , h1 , h2 , t0 )
(37) ≤ + b0 (η3 , h1 , h2 , t0 ).
n1+it A2 (h1 , h2 , t0 )
XM −1 <n≤X0

So, for such r, the tail sum is bounded by a constant independent of t. To ensure
the contingent condition r ≥ 6 holds, we will suppose that r0 ≥ 6.
It remains to bound the initial sum over 1 ≤ n ≤ XM−1 . We use the triangle
inequality followed by the harmonic sum bound (2), which yields
X 1 1
(38) < log XM−1 + γ + .
n1+it XM−1 − 1
1≤n≤XM −1

Using the inequality (7) to bound XM−1 from above, combined with a similar
calculation as leading to (15) to bound XM−1 from below, the right-side in (38) is
seen to satisfy
θr+1 θr+1 1
(39) ≤ log t − log(2π) + log h1 + γ + .
2 2 κ(h2 , t0 ) − 1
Recalling the lower bound on θr+1 in (33), combined with the supposition r0 ≥ 6,
2 2
≤ θr+1 ≤ .
r − 15/16 r−1
Hence, the main term in (39) satisfies
θr+1 log t log t log log t0 log t
(40) log t ≤ < ≤ ,
2 r−1 h2 log log t − 2 h2 log log t0 − 2 log log t
and also,
θr+1 log(2π)
(41) log(2π) ≥ .
2 h2 log log t0 − 15/16
5By the lower bound on h in (9), we have xh2 log 2 ≥ 8 for for x ≥ log t . So, ρ(h log x) is
2 0 2
well-defined for x ≥ log t0 .
6This step is where a slow enough growth on r is utilized, to ensure that (log t)h2 log 2 = o(log t)
as t → ∞, so that the constant A2 (h2 , t0 ) defined in (35) is positive.
 EXPLICIT BOUNDS FOR THE RIEMANN ZETA-FUNCTION ON THE 1-LINE 9

Putting (38), (39), (40) and (41) together, we obtain

X 1 log t
(42) < A3 ,
n1+it log log t
1≤n≤XM −1

where
log log t0
A3 = A3 (h1 , h2 , t0 ) := +
h log log t0 − 2
 2 
1 log(2π) log log t0
log h1 + γ + − .
κ(h2 , t0 ) − 1 h2 log log t0 − 15/16 log t0
So, combining (37) and (42), we see that the main sum satisfies the inequality

X 1 log t
≤ A4 ,
n1+it log log t
1≤n≤n1

where
 
log log t0
A1
A4 = A4 (η3 , h1 , h2 , t0 ) := A3 + + b0 .
A2 log t0
√
Lastly, the bound on ζ(1+it) in (4) carries an extra term g(t)/ 2π+R(t). Using
the definitions in Theorem 12, this term is easily seen to be bounded by 1 + R(t0 ).
So, letting
log log t0
A5 = A5 (t0 ) := (1 + R(t0 )) ,
log t0
we obtain, overall, that for t ≥ t0 ,
log t
|ζ(1 + it)| ≤ A6 ,
log log t
where
A6 := A6 (η3 , h1 , h2 , t0 ) = A4 + A5 .
We choose the following values for our free parameters, which are suggested by
numerical experimentation.
η3 = 0.8410538348318537,
h1 = 1.0072318915891114,
(43)
h2 = 0.8643140421215492,
t0 = exp(3069).

With these choices, our preconditions on t0 and h2 hold, and the conditions h > 1
and r0 ≥ 6 also hold. More precisely, r0 = 6 and h ≥ 1.0072. Additionally, in §3,
we show that

A0 = 0.0207492006388441 . . .,
A2 = 0.04098649913361486 . . ..
10 G. HIARY, N. LEONG AND A. YANG

Therefore, we can explicitly compute

b0 ≤ 2.116947 · 10−7 ,
A1 ≤ 1.4397435484321421,
A3 ≤ 1.6261617818960195,
A4 ≤ 1.7180616058089129,
A5 ≤ 0.0026161964987356586,
A6 ≤ 1.7206778023076486.
In summary, the assertion of Theorem 1 holds when t ≥ t0 = exp(3069).
For smaller t ≤ t0 we use the bound in Patel [Pat22] and Proposition 3. Specifi-
cally, the bound in Patel [Pat22] yields

1 log t
|ζ(1 + it)| ≤ log t + 44.02 ≤ 1.721 , exp(705.64) ≤ t ≤ exp(3069),
5 log log t
1 log t
|ζ(1 + it)| ≤ log t + 1.93 ≤ 1.721 , 3 ≤ t ≤ exp(16.01).
2 log log t
This leaves the gap interval exp(16.01) ≤ t ≤ exp(705.64) for which we invoke
Proposition 3. Since the gap interval falls in the range of t where the proposition
is applicable, our target bound is verified there as well.

3. Bounding A0 and A2
We derive an upper bound on A0 (η3 , h). For the convenience of the reader, we
recall the definition A0 (η3 , h) := supk≥6 A(η3 , h, k), where for k ≥ 6,
A(η3 , h, k) := (2π)k/(81K−81) Ck (η3 , h) + Dk (η3 , h).
With our choice of free parameter values in (43), we numerically compute the
following list of A(η3 , h, k).
A(η3 , h, 6) = 0.0207492006388441, A(η3 , h, 7) = 0.02070728682845934,
A(η3 , h, 8) = 0.0205218584268324, A(η3 , h, 9) = 0.02034268739049111,
A(η3 , h, 10) = 0.0202077543175717.
The values of Ck and Dk required for the above computation are obtained us-
ing the formulas in Lemma 16. In turn, evaluating Ck (η3 , h) and Dk (η3 , h) relies
on evaluating Ak (η3 , hk ) and Bk (η3 ). When k = 3, Ak and Bk are computed di-
rectly using their definitions in Lemma 14. When k ≥ 4, the recursive formulas in
Lemma 15 are used to compute them. Let us examine these recursive formulas.
First, the following quantity, which appears in the formula for Ak+1 (η3 , ω) in
(93), satisfies
219/12 (K − 1) 219/12 · 511
(44) max p = √ =: µ10 .
k≥10 (2K − 1)(4K − 3) 2092035
Here, K = 2k−1 and the maximum occurs when k = 10. Another quanitity appear-
ing in the formula for Ak+1 (η3 , ω) is δk , which is clearly seen to decrease with k.
Noting that h > 1, hence ω = hk+1 > 1 for any k ≥ 0, we deduce that
 
Ak+1 (η3 , ω) ≤ δ10 1 + µ10 Ak (η3 , ω)1/2 .
 EXPLICIT BOUNDS FOR THE RIEMANN ZETA-FUNCTION ON THE 1-LINE 11

So, we are led to consider the discrete map

 
xn+1 = δ10 1 + µ10 x1/2
n .

By a routine calculation, this map has a single fixed point at

r !2
∗ µ10 δ10 µ210 δ10
2
(45) x = + + δ10 ≤ 2.7600447683620426,
2 4

which is a stable point. Since Ak+1 (η3 , ω) ≤ xk+1 for k ≥ 10 and, by a direct
numerical computation, A10 (η3 , ω) ≤ x∗ , it follows that
(46) Ak (η3 , hk ) ≤ x∗ , (k ≥ 10).
The analysis of Bk+1 for k ≥ 10 proceeds similarly. The recursive formula for
Bk+1 contains the following quantity satisfying
23/2 (K − 1) 23/2 · 511
(47) max p ≤ √ =: ν10 .
k≥10 (2K − 3)(4K − 5) 3 231767
1/2 1/2
So, Bk+1 ≤ δ10 ν10 Bk for k ≥ 10. Consider the discrete map yn+1 = δ10 ν10 yn .
This map has a single fixed point, namely,
y ∗ = (δ10 ν10 )2 < 1.0023472426905147,
which is also stable. Since Bk+1 ≤ yk+1 for k ≥ 10 and, by a direct numerical
computation, B10 ≤ y ∗ , it follows that
(48) Bk (η3 ) ≤ y ∗ , (k ≥ 10).
Furthermore, we numerically verify that for k ≥ 10,
  1
2k/K−k/(2K−2) (k − 1)! 2K−2
h (h − 1) < 0.0073,
2π
 1
 2K−2
k/(2K−2) 1−2/K 2π
h (h − 1) < 0.0001.
(k − 1)!
Combining this with (46) and (48), and recalling the definitions in Lemma 16,
yields the inequalities Ck (η3 , h) ≤ 0.0073x∗ and Dk (η3 , h) ≤ 0.0001y ∗ for k ≥ 10.
Consequently, A(η3 , h, k) < 0.0203 for k ≥ 10.
Furthermore, in view of the values of A(η3 , h, k) listed at the beginning, we also
see that A(η3 , h, k) ≤ A(η3 , h, 6) = 0.02074 . . . for 6 ≤ k ≤ 10. We therefore
conclude A0 (η3 , h) = A(η3 , h, 6).
For the computation related to A2 , we shall use a derivative calculation to show
that the function
2
(49) a2 (u) := 1 − b1 + eu ρ(h2 u) + −u
h2 u − 1 + 22−r0
is increasing in u ≥ u0 := log log t0 . So, on identifying u with log x in the definition
of A2 in (35), the value of A2 is exp(a2 (u0 )).
To this end, we first calculate
 
d u u d d
e ρ(h2 u) = e 1 + log ρ(h2 u) , log ρ(h2 u) = −hã2 (h2 u),
du du du
12 G. HIARY, N. LEONG AND A. YANG

where
(v − 1) log 2 + 1 − 2−v (8(v − 1) log 2 + 6) + 4−v (8(v − 2) log 2 + 8)
ã2 (v) = .
(v − 1 + 22−v )(1 − 21−v )(1 − 22−v )
Note that v is identified with h2 u, so we need only consider v ≥ h2 u0 =: v0 . It is
easy to see that
(v − 1) log 2 + 1
sup ã2 (v) < sup ≤ log 2 + 0.2,
v≥v0 v≥v0 (v − 1 + 22−v0 )(1 − 21−v0 )(1 − 22−v0 )
where the supremum occurs at v = v0 . Hence, the derivative of eu ρ(h2 u) is at least
eu (1 − (log 2 + 0.2)h2 ) for u ≥ u0 . Importantly, 1 − (log 2 + 0.2)h2 is positive. In
comparison, the derivative of the remaining terms in the formula for a2 (u) is
 
d 2 2h2
2−r
−u =− − 1,
du h2 u − 1 + 2 0 (h2 u − 1 + 22−r0 )2
which is at least −h2 − 1 for u ≥ u0 . Since eu (1 − (log 2 + 0.2)h2 ) − h2 − 1 > 0 for
u ≥ u0 , the derivative of a2 (u) is positive for u ≥ u0 , proving our claim.

4. Proof of Proposition 3
The calculation is similar to that in [Hia16] on the 1/2-line. We divide the main
sum in (4) into pieces of length ≈ t1/3 , then apply the van der Corput Lemma 13 to
each piece other than an initial segment, which is bounded by the triangle inequality.
So, we shall be brief in some details.
Let t1 be a positive number to be chosen later, subject to t1 ≥ exp(12). We will
suppose t ≥ t1 , unless otherwise is stated. Let P = ⌈t1/3 ⌉, Q = ⌊n1 /P ⌋, and let q0
a positive integer to be chosen later, subject to q0 ≤ Q. Thus, we have
n1 q0X
P −1 Q−1 (q+1)P −1 n1
X 1 1 X X 1 X 1
(50) ≤ + + .
n=1
n1+it n=1
n q=q n1+it n1+it
0 n=qP QP

The harmonic sum bound (2) yields, subject to q0 P > 1,

q0X
P −1
1 1
(51) ≤ log(q0 P − 1) + γ + .
n=1
n q0 P − 1

For q0 ≤ q ≤ Q, let N = qP and let N ′ be an integer such that N ≤ N ′ < N +P .

Partial summation gives
′
N
X 1 1
(52) ≤ max |SN (L)|,
n1+it N 1≤L≤P
n=N

where
N +L−1
X t
SN (L) := exp(2πif (n)) and f (x) = log x.
2π
n=N
′
So that, for N ≤ x ≤ N ,
   
′′′ t t t t
|f (x)| ∈ , ⊆ , .
2πN ′3 2πN 3 2π(q + 1)3 P 3 2πq 3 P 3
 EXPLICIT BOUNDS FOR THE RIEMANN ZETA-FUNCTION ON THE 1-LINE 13

We may therefore apply Lemma 13 to SN (L) with

3
2π(q + 1)3 P 3 (q + 1)3

1
W = , λ= = 1+ > 1.
t q3 q
This gives, for any positive η,
r
1 1 α βW 1/3 αη ηβW 2/3
max |SN (L)| ≤ + + + .
N 1≤L≤P q W 1/3 P P P2
We estimate the (nonnegative) terms under the square-root as follows.

α α βW 1/3 β(2π)1/3 (1 + 1/q0 )q

≤ , ≤ ,
W 1/3 (2π)1/3 (q + 1) P t1/3

αη αη ηβW 2/3 ηβ(2π)2/3 (1 + 1/q0 )2 q 2

≤ 1/3 , ≤ .
P t P2 t2/3
√ √ √
Isolating the first term using the inequality x + y ≤ x+ y, valid for nonnegative
x and y, and using the inequality q ≤ Q ≤ t1/6 /(2π)1/2 to bound q in the remaining
terms, therefore gives
1 α1/2 1 B
(53) max |SN (L)| ≤ 1/6 1/2
+ 1/12 ,
N 1≤L≤P (2π) q(q + 1) qt
where, since t ≥ t1 ,
s
β(1 + 1/q0 ) αη ηβ(1 + 1/q0 )2
B := 1/6
+ 1/6 + 1/6
.
(2π) t1 (2π)1/3 t1
We also note that for all q ≥ q0 we have W ≥ W0 and λ ≤ λ0 , where
 3
3 1
W0 := 2π(q0 + 1) , λ0 := 1 + .
q0
Therefore, for all q under consideration, α ≤ α0 and β ≤ β0 , where
1 32 2λ0 η 2λ0
q
−1/3
α0 := + √ η + W0 + 1/3 + 2/3
η 15 π W W 0 0
64 1 4
β0 := √ √ + 1/3 .
3 π η W
0

It follows that for q ≥ q0 ,

1/2
1 α0 1 B0
(54) max |SN (L)| ≤ 1/6 1/2
+ 1/12 ,
N 1≤L≤P (2π) q(q + 1) qt
where B0 is defined in the same way as B except α and β are replaced by their
maximums α0 and β0 .
Put together, we have
n1 1/2 Q Q
X 1 α0 X 1 B0 X 1
(55) ≤ + .
n1+it (2π)1/6 q=q q(q + 1)1/2 t1/12 q=q q
n=q0 P 0 0
14 G. HIARY, N. LEONG AND A. YANG

Moreover, as shown in [HPY22], and using the bound Q ≤ t1/6 /(2π)1/2 once again,
Q
X 1 2
<p ,
q=q0
q(q + 1)1/2 q0 − 1/2
(56)
Q
X 1 1
≤ log t + log τ1 − log (q0 − 1/2) ,
q=q
q 6
0

where
1 1
τ1 := 1/2
+ 1/6 .
(2π) 2t1
Therefore, observing that 16 tlog t
1/12 is decreasing for t ≥ t1 ≥ exp(12), and combining

this with the initial sum bound (51), we obtain

n1 1/2
X 1 1 α0 2
< log(q0 P − 1) + γ + +
n1+it 1/6
p
n=1
q0 P − 1 (2π) q0 − 1/2
 
B0 1
+ 1/12
log t1 + log τ1 − log (q0 − 1/2) .
t1 6
We now estimate
1 
−1/3

log(q0 P − 1) ≤ log(q0 t1/3 + q0 − 1) ≤ log t + log q0 + (q0 − 1)t1 ,
3
which follows using P < t1/3 + 1 as well as
1 1
≤ 1/3
.
q0 P − 1 q0 t − 1 1
This yields
1 g(t1 )
(57) |ζ(1 + it)| ≤ log t + B1 (q0 ) + √ + R(t1 ),
3 2π
where

−1/3
 1
B1 (q0 ) := γ + log q0 + (q0 − 1)t1 + 1/3
q0 t1 −1
1/2  
α0 2 B0 1
+ + 1/12 log t1 + log τ1 − log (q0 − 1/2) .
(2π)1/6
p
q0 − 1/2 t1 6
Choosing t1 = exp(16), η = 1.304 and q0 = 5, we obtain
1
(58) |ζ(1 + it)| ≤ log t + 4.664, t ≥ exp(16).
3
This is the first part of Proposition 3, and implies Theorem 1 in the range exp(16) ≤
t ≤ exp(82). Similarly, choosing t1 = exp(82), η = 1.028, q0 = 4, we obtain
1
(59) |ζ(1 + it)| ≤ log t + 4.308, t ≥ exp(82).
3
This implies Theorem 1 for exp(82) ≤ t ≤ exp(93).
Consider now the last two parts of Proposition 3. In this critical region, we make
use of a few additional tools to sharpen our estimates. In particular, we employ a
different method of bounding the second sum appearing in Theorem 12, leading to
an estimate of size O(t−1/12 ) instead of O(1). Also, we employ the k-th derivative
 EXPLICIT BOUNDS FOR THE RIEMANN ZETA-FUNCTION ON THE 1-LINE 15

tests used in the proof of Theorem 1, however this time we introduce additional
scaling parameters to further optimise the switching points between derivative tests
of different order.
To this end, let t2 be a real number to be chosen later, subject to t2 ≥ exp(182/3).
Assume t ≥ t2 . Recall from (55) we showed for q0 ≥ 1, P = ⌈t1/3 ⌉ and t ≥ t2 that
the tail sum satisfies
n1 1/2  
X 1 α0 2 B0 1
≤ + log t2 + log τ1 − log (q0 − 1/2) .
n1+it (2π)1/6 q0 − 1/2 t1/12 6
p
n=q0 P 2

As for the initial sum over 1 ≤ n < q0 P , instead of bounding it using the triangle
inequality throughout as in (51), we will refine our estimate by using k-th order
van der Corput lemmas with k = 4 and 5 in the subintervals (Z5 , Z4 ] and (Z6 , Z5 ]
respectively, where
Z4 := q0 P − 1, Z5 := µ5 tθ5 /2 , Z6 := µ6 tθ6 /2 ,
with θ5 , θ6 the same as in the proof of Theorem 1 and µ5 , µ6 real positive numbers
to be chosen later, subject to
θ /2 θ /2 1/3
(60) 1 < µ6 t26 ≤ µ5 t25 ≤ q0 ⌈t2 ⌉ − 1.
We remark that Z5 , Z6 will serve a similar purpose as Yk in the proof of Theorem 1,
whereas the new parameters µ5 , µ6 will allow us to fine-tune our estimates in this
critical region.
Let h′1 be a real number to be chosen, such that h′1 > 1. Let Xm ′
= (h′1 )−m Z4 .
Define the subsum
′
X 1
Sm := 1+it
,
n
′ ′ Xm <n≤Xm−1

as well as $ %
′ log(q0 P − 1) − θ26 log t − log µ6
M := .
log h′1
The definition of M ′ guarantees XM′ ′
′ ≥ Z6 ≥ XM ′ +1 . So that, by a similar calcu-
′
lation as in (12), for each 1 ≤ m ≤ M ,
′
⌊Xm−1 ⌋ ′ h′1 ′ h′1
′ ⌋
≤ h 1 + ′ ≤ h 1 + θ /2
=: h′ , say.
⌊Xm XM ′ − 1 µ6 t26 − 1
θ /2
Here we used the constraint µ6 t26 > 1 to ensure that h′ is well defined.
′
Let k = 4 or 5. Following the argument of Theorem 1, if Zk+1 < Xm ≤ Zk , then
′
|Sm | ≤ Ck (η3 , h′ )Xm
′−k/(2K−2) 1/(2K−2)
t + Dk (η3 , h′ )Xm
′k/(2K−2)−2/K −1/(2K−2)
t
−k/(2K−2) 1/(2K−2) k/(2K−2)−2/K −1/(2K−2)
≤ Ck (η3 , h′ )Zk+1 t + Dk (η3 , h′ )Zk t
≤ A′k t−ρ(k) ,
−1/3
where, letting µ4 := q0 + (q0 − 1)t2 so that Z4 ≤ µ4 t1/3 , as well as noting
k/(2K − 2) − 2/K ≥ 0, we have
(
−2/7 1/28
′ ′ µ5 C4 (η3 , h′ ) + µ4 D4 (η3 , h′ ), k = 4,
Ak (η3 , h ) = −1/6 1/24
µ6 C5 (η3 , h′ ) + µ5 D5 (η3 , h′ ), k = 5.
16 G. HIARY, N. LEONG AND A. YANG

′ ′
Now, the number of Xm satisfying Zk+1 < Xm ≤ Zk , call it Nk′ , clearly satisfies
log Zk − log Zk+1 log t
Nk′ ≤ + 1, therefore Nk′ ≤ Bk′ ,
log h′1 log h′1
where
max{0, log µ4 − log µ5 + log h′1 }

1

 + , k = 4,
 39 log t2


Bk′ :=
28 max{0, log µ5 − log µ6 + log h′1 }


+ , k = 5.



429 log t2
Put together, we have

X
′ A′k Bk′ −ρ(k)
Sm ≤ t log t.
log h′1
Zk+1 <Xm
′ ≤Z
k

For t ≥ t2 , the function t−ρ(k) log t is decreasing if k = 4 and for k = 5 is bounded

from above by 990/(7e). Therefore,
 
X
′ 1 ′ ′ −ρ(4) 990 ′ ′
(61) Sm ≤ A4 B4 t2 log t2 + A B
log h′1 7e 5 5
Z6 <Xm
′ ≤Z
4

′ ′ ′
In the remaining range 1 ≤ n ≤ XM ′ , we use that h1 Z6 ≥ XM ′ . Combined with

the triangle inequality and (2), this yields

X 1 1
≤ log h′1 Z6 + γ +
n1+it h′1 Z6 −1
(62) 1≤n≤XM ′

θ6 1
log t + log µ6 + log h′1 + γ +
≤ θ6 /2
.
2 ′
h 1 µ6 t 2 − 1
it
P
Next, consider the second sum 1≤n≤n1 n in Theorem 12, which was just
bounded by n1 in our previous calculations. Let N = sP , where s is a positive
integer and, as before, P = ⌈t1/3 ⌉. Also, let s0 a positive integer and η4 a real
positive number, both to be chosen. Using P ≤ t1/3 + 1, the estimate (54) gives for
s ≥ s0 ,
α0 (s0 , η4 , t2 )1/2
 
1/3 1 B0 (s0 , η4 , t2 )
max |SN (L)| ≤ (t + 1) + .
1≤L≤P (2π)1/6 (s + 1)1/2 t1/12
We combine this with the inequality
S Z S−1 √
X 1 dx √
1/2
≤ 1/2
≤ 2 S − 2 s0 ,
s=s
(s + 1) s0 −1 (x + 1)
0

as well as bound the sum over 1 ≤ n < s0 P by the number of terms in the sum,
which gives
n1
" #
1/2 √
X
it 1/3 α0 √ B0
n ≤ s0 P − 1 + (t + 1) (2 S − 2 s0 ) + 1/12 (S − s0 + 1) .
n=1
(2π)1/6 t
 EXPLICIT BOUNDS FOR THE RIEMANN ZETA-FUNCTION ON THE 1-LINE 17

√
Since S ≤ t1/6 / 2π, we obtain for t ≥ t2 ,
n1
X
(63) nit ≤ C t5/12 ,
n=1

where
"
1/2 √ !
−1/3 s0 2α (s0 , η4 , t2 ) 1 s0
C(s0 , η4 , t2 ) := (1 + t2 ) 1/12
+ 0 − 1/12
t2 (2π)1/6 (2π) 1/4
t2
!#
1 s0 − 1
+ B0 (s0 , η4 , t2 ) √ − 1/6 .
2π t2

Therefore, after multiplying by g(t)t−1/2 , we obtain a far sharper bound on the

−1/12
contribution of the second sum in Theorem √ 12. Namely, the bound Cg(t2 ) t ,
which replaces the original estimate g(t2 )/ 2π.
Lastly, combining (63) with (61) and (62), and noting that θ6 = 16/33, we have
for t ≥ t2
8
|ζ(1 + it)| ≤ log t + B3 ,
33
where
 
1 ′ ′ −ρ(4) 990 ′ ′
B3 := A4 B4 t2 log t2 + A B
log h′1 7e 5 5
1 g(t2 )
+ log µ6 + log h′1 + γ + θ6 /2
+ C(s0 , η4 , t2 ) 1/12 + R(t2 ).
′
h 1 µ6 t 2 − 1 t2
We choose

η3 = 1.4942456016768517, η4 = 2.0960121834416348, h′1 = 1.126788460908779,

q0 = 29, s0 = 110, µ5 = 32.114893449766214,
µ6 = 35.523572643294735, t2 = exp(93).

Verifying that (60) holds, we calculate

h′ = 1.106123080171765 . . .,
B3 ≤ 12.52811,

i.e. that
8
|ζ(1 + it)| ≤
log t + 12.53, t ≥ exp(93).
33
This implies Theorem 1 for exp(93) ≤ t ≤ exp(706).

5. Proof of Theorem 2
To help make our notation more intuitive, we will allow reuse of variable names
from previous sections. The reader should keep in mind these variables mean dif-
ferent things in the confines of this section. Since this section is independent from
§2, §3 and §4, this will hopefully cause no confusion.
18 G. HIARY, N. LEONG AND A. YANG

5.1. Some lemmas. Here we give some preliminary lemmas for our proofs of es-
timates on |ζ ′ (1 + it)/ζ(1 + it)| and |1/ζ(1 + it)|.
The first two lemmas, which form the basis of our method, are built upon a
theorem of Borel and Carathéodory. This theorem enables us to deduce an upper
bound for the modulus of a function and its derivatives on a circle, from bounds
for its real part on a larger concentric circle.
Theorem 4 (Borel–Carathéodory). Let s0 be a complex number. Let R be a posi-
tive number, possibly depending on s0 . Suppose that the function f (s) is analytic in
a region containing the disk |s − s0 | ≤ R. Let M denote the maximum of Re f (s) on
the boundary |s−s0 | = R. Then, for any r ∈ (0, R) and any s such that |s−s0 | ≤ r,
2r R+r
|f (s)| ≤ M+ |f (s0 )|.
R−r R−r
If in addition f (s0 ) = 0, then for any r ∈ (0, R) and any s such that |s − s0 | ≤ r,
2R
|f ′ (s)| ≤ M.
(R − r)2
Proof. Proofs for estimating both |f (s)| and |f ′ (s)| are available in [Tit39, §5.5].
However, if we include the restriction f (s0 ) = 0, we can refer to [MV07, Lemma 6.2],
which provides alternative proofs, resulting in a sharper estimate for |f ′ (s)|. 

Lemma 5. Let s0 be a complex number. Let r and α be positive numbers (possibly

depending on s0 ) such that α < 1/2. Suppose that the function f (s) is analytic in
a region containing the disk |s − s0 | ≤ r. Suppose further there is a number A1
independent of s such that,
f (s)
≤ A1 , |s − s0 | ≤ r.
f (s0 )
Then, for any s in the disk |s − s0 | ≤ αr we have

f ′ (s) X 1 4 log A1
− ≤ ,
f (s) ρ
s − ρ r(1 − 2α)2
1
where ρ runs through the zeros of f (s) in the disk |s − s0 | ≤ 2 r, counted with
multiplicity.
Proof. We follow the proof in §3 of [Tit86]. Let g be the function
Y 1
g(s) = f (s) ,
ρ
s−ρ

where ρ in the (finite) product runs through the zeros of f (s), counted with multi-
plicity, that satisfy |ρ − s0 | ≤ 21 r. Since by construction the poles and zeros cancel,
g is analytic in |s− s0 | ≤ r, and does not vanish in the disk |s− s0 | ≤ 21 r. Therefore,
the function
g(s)
h(s) = log ,
g(s0 )
where the logarithm branch is determined by h(s0 ) = 0, is analytic in some region
containing the disk |s − s0 | ≤ 12 r.
 EXPLICIT BOUNDS FOR THE RIEMANN ZETA-FUNCTION ON THE 1-LINE 19

Now, on the circle |s − s0 | = r, we have |s − ρ| ≥ 21 r ≥ |s0 − ρ|. So, on this circle,

 
g(s) f (s) Y s0 − ρ f (s)
= ≤ ≤ A1 .
g(s0 ) f (s0 ) ρ s−ρ f (s0 )

The inequality also holds in the interior of the circle, by the maximum modulus
principle. Hence, Re h(s) ≤ log A1 throughout the disk |s − s0 | ≤ r.
Given this, we apply the Borel–Carathéodory Theorem 4 to h(s) with the con-
centric circles centered at s0 of radii αr and 21 r. This yields throughout the disk
|s − s0 | ≤ αr that
4 log A1
(64) |h′ (s)| ≤ .
r(1 − 2α)2
On the other hand, since g(s0 ) is constant, we have

g ′ (s) f ′ (s) X 1
|h′ (s)| = = − .
g(s) f (s) ρ
s−ρ

Combining this formula with the inequality (64) completes the proof. 

Lemma 6. Let s and s0 complex numbers with real parts σ and σ0 , respectively.
Let r, α, β, A1 and A2 be positive numbers, possibly depending on s0 , such that
α < 1/2 and β < 1. Suppose that the function f (s) satisfies the conditions of
Lemma 5 with r, α and A1 , and that
f ′ (s0 )
≤ A2 .
f (s0 )
Suppose, in addition, that f (s) 6= 0 in for any s in both the disk |s − s0 | ≤ r and
the right half-plane σ ≥ σ0 − αr. Then, for any s in the disk |s − s0 | ≤ αβr,
f ′ (s) 8β log A1 1+β
≤ + A2 .
f (s) r(1 − β)(1 − 2α)2 1−β
Proof. Using the inequality −|z| ≤ −Re z ≤ |z|, valid for any complex number z,
together with Lemma 5 in |s − s0 | ≤ αr, we have throughout this disk,
f ′ (s) 4 log A1 X 1
(65) −Re ≤ 2
− Re ,
f (s) r(1 − 2α) ρ
s−ρ

where ρ runs through the zeros of f (with multiplicity) satisfying |ρ − s0 | ≤ 21 r. By

the assumption on the nonvanishing of f , each ρ in this sum satisfies Re ρ < σ0 −αr.
On the other hand, all s in the disk |s − s0 | ≤ αr clearly satisfy σ ≥ σ0 − αr. Hence,
X 1 X σ − Re ρ
Re = > 0,
ρ
s−ρ ρ
|s − ρ|2

throughout the disk |s − s0 | ≤ αr. We may thus drop the sum in (65) and still
obtain a valid upper bound, call it M1 .
Now, −f ′ (s)/f (s) is analytic in the disk |s − s0 | ≤ αr since by assumption f
does not vanish there. We can therefore apply Theorem 4 to −f ′ (s)/f (s) using
the two concentric circles |s − s0 | ≤ αβr and |s − s0 | ≤ αr, and the upper bound
M1 ≥ M on the circle |s − s0 | ≤ αr, which yields the result. 
20 G. HIARY, N. LEONG AND A. YANG

Lemma 7. Let C > 0 and t0 ≥ ee be constants. Suppose t ≥ t0 . Define

C(log log t)2
σt := 1 − .
log t
Additionally, suppose that C satisfies
C(log log te )2 1 2
(66) ≤ , where te := ee .
log te 2
For each t ≥ t0 , if σt ≤ σ ≤ 2, then
2 71.2 C 3/2
(67) |ζ(σ + it)| ≤ A logB t where A = 76.2, B= + .
3 e2
Proof. From Ford [For02a], if 1/2 ≤ σ ≤ 1, then
3/2
(68) |ζ(σ + it)| ≤ 76.2 t4.45(1−σ) log2/3 t.
As σt ≥ 1/2, a consequence of the assumption (66) and considering that (log log t)2 / log t
reaches its maximum at t = te , we may use (68) for any σ ∈ [σt , 1].
In view of this, since (1 − σ)3/2 is monotonically decreasing with σ, if σt ≤ σ ≤ 1,
then (68) gives
 
3/2 2
(69) |ζ(σ + it)| ≤ 76.2 exp 4.45(1 − σt ) log t + log log t .
3
Or, written differently,
 
(70) |ζ(σ + it)| ≤ A exp B̃ log log t ,

where A is defined as in (67) and

2 (log log t)2
B̃ := + 4.45 C 3/2 √ .
3 log t
√ 4
As (log log t)2 / log t reaches its maximum at t = ee , attaining the value 16/e2,
we obtain after a simple computation that B̃ ≤ B, with B defined as in (67). The
desired result therefore follows for σt ≤ σ ≤ 1 on replacing B̃ with B in (70).
To show that the result holds for 1 ≤ σ ≤ 2 as well, we use the Phragmén–
Lindelöf Principle on the holomorphic function
f (s) = (s − 1)ζ(s).
To this end, on the 1-line we have
|f (1 + it)| = |t| |ζ(1 + it)| ≤ 62.6 |2 + it| log2/3 |2 + it|.
This inequality is verified numerically for |t| < 3 and is a consequence of the bound
|ζ(1 + it)| ≤ 62.6 log2/3 t for |t| ≥ 3, given in [Tru14a]. Also, on the 2-line, plainly,
|f (2 + it)| ≤ |1 + it| ζ(2) < 62.6|3 + it| log2/3 |3 + it|,
for all real t. By [Tru14b, Lemma 3], we thus obtain7
(71) |f (s)| ≤ 62.6|1 + s| log2/3 |1 + s|, (1 ≤ σ ≤ 2).

7We apply [Tru14b, Lemma 3] with the following parameter values (in the notation in the cited
paper): a = 1, b = 2, Q = 1, α1 = β1 = 1, α2 = β2 = 2/3 and A = B = 62.6. Note that the A
and B in [Tru14b, Lemma 3] are different from the A and B in the statement of our lemma.
 EXPLICIT BOUNDS FOR THE RIEMANN ZETA-FUNCTION ON THE 1-LINE 21

Moreover, for 1 ≤ σ ≤ 2 we have the easy estimates,

√
|1 + s| 9 + t2
q
≤ ≤ 1 + 9t−2
0 ,
|s − 1| t
p q
log |1 + s| ≤ log 9 + t2 ≤ log t + log 1 + 9t−2
0 .

So that, for 1 ≤ σ ≤ 2 and t ≥ t0 , the inequality (71) implies that

 q 
q log 1 + 9t−2
0
|ζ(σ + it)| ≤ 62.6 1 + 9t−2
0
1 +  log2/3 t < A logB t,
log t0

where A and B are defined as in (67), as desired. 

Remark. Following Titchmarsh’s argument [Tit86, Theorem 5.17], we could make
use of [Yan23, Theorem 1.1]. For simplicity, however, we used the Richert-type
bound (68) due to Ford [For02a] instead.
Lemma 8. Let t0 ≥ ee and d1 > 0 be constants. Suppose t ≥ t0 . We have
 
log log t log t
ζ 1 + d1 + it ≤ C ′ (d1 , t0 ) ,
log t log log t
where γ is Euler’s constant, and
 
′ 1 log log t0
(72) C (d1 , t0 ) = exp γd1 .
d1 log t0
Proof. We use the following uniform bound due to Bastien and Rogalski [BR02]:
eγ(σ−1)
|ζ(σ + it)| ≤ ζ(σ) ≤ , (σ > 1).
σ−1
Substituting σ = 1 + d1 log log t/ log t therefore gives for t ≥ t0 ,
eγd1 log log t/ log t
 
1 γd1 log log t0 log t
ζ(σ) ≤ ≤ exp ,
d1 log log t/ log t d1 log t0 log log t
as claimed. In the last inequality, we used that log log t/ log t reaches its maximum
at t = ee and is monotonically decreasing after that. 
5.2. Bounding |ζ ′ (1 + it)/ζ(1 + it)|. We follow the argument of [Tru15], which
gives an explicit version of results in [Tit86, §3]. The method relies on the Borel–
Carathédory Theorem 4.
We will therefore construct concentric disks, centred just to the right of the line
σ = 1 + it, and extend their radius slightly to the left and into the critical strip.
Ultimately, we want to apply Lemmas 5 and 6 with f (s) = ζ(s), which will give us
the desired results.
The process is as follows. Let t0 ≥ ee and suppose t′ ≥ t0 . Let d be a real
positive constant, which will be chosen later. Denote the center of the concentric
disks to be constructed by s′ = σ ′ + it′ , where
d log log t′
(73) σ ′ = 1 + δt′ , with δt′ := .
log t′
Notice that δt′ is decreasing with t′ and is at most d/e. Let r > 0 denote the radius
of our largest concentric disk. In the sequel, r will be determined as function of s′ .
22 G. HIARY, N. LEONG AND A. YANG

In addition, suppose σ ′ + r ≤ 1 + ǫ where ǫ ≤ 1 is another positive constant to

be chosen later. This implies, for instance, that r ≤ ǫ ≤ 1. In order to enforce this
inequality on σ ′ and r, it is enough to demand that
d
(74) + r ≤ ǫ.
e
Let s = σ + it be a complex number. Aiming to apply Lemma 5 in the disk
|s − s′ | ≤ r, we first seek a valid A1 in that disk. Similarly, for Lemma 6, we seek
a valid A2 at the disk center s = s′ . In applying Lemma 6, we moreover need to
ensure that the non-vanishing condition on f (s) = ζ(s) is fulfilled, which we do by
way of a zero-free region of zeta.
Let us first determine a valid A1 , with the aid of Lemma 7. To this end, let
C > 0 be a parameter restricted by the inequality in (66), so that (in the notation
of Lemma 7) σt ≥ 1/2 for t ≥ e. Note that σt is monotonically decreasing for
e ≤ t ≤ te and monotonically increasing thereafter. So, the maximum of σt over
the interval t′ − r ≤ t ≤ t′ + r occurs at one of the end-points. For example, if
t′ − r ≥ te , then the maximum occurs at t = t′ + r. And if t′ + r ≤ te , then the
maximum occurs at t = t′ − r. With this in mind, we calculate
C0 (log log t′ )2
(75) max σt = max{σt′ −r , σt′ +r } < 1 − ,
|t−t |≤r
′ log t′
where, since r ≤ ǫ,
( 2 )
log log(t′ − ǫ) log t′
C0 := C max , .
log log t′ log(t′ + ǫ)
Clearly, C0 = C0 (t′ ) increases monotonically with t′ (towards C). We therefore set
C1 (log log t′ )2
(76) C1 := C0 (t0 ) and σ1,t′ := 1 − .
log t′
So that, by (75), no matter t′ , r and ǫ, so long as they are subject to our conditions,
σ1,t′ ≥ max σt .
|t−t |≤r
′

Consequently, Lemma 7 gives that for each t ∈ [t′ − r, t′ + r] and any σ ∈ [σ1,t′ , 2],
and on recalling r ≤ ǫ,
B
log(1 + ǫ/t′ )

B ′ B ′
(77) |ζ(s)| ≤ A log (t + r) ≤ A3 log t , A3 = A 1 + .
log t′
We would like the inequality (77) to hold throughout the disk |s − s′ | ≤ r.
This will be certainly the case if this disk lies entirely in the rectangle specified by
σ ∈ [σ1,t′ , 2] and t ∈ [t′ − r, t′ + r]. This follows, in turn, on requiring σ ′ − r ≥ σ1,t′
or, equivalently, that
C1 (log log t′ )2
r≤ + δt′ .
log t′
Therefore, subject to the constraint (74), the following choice of r works:
(log log t′ )2
 
d
(78) r = rt′ := C1 + ′
.
log log t log t′
To ensure the constraint (74) is met for all t′ ≥ t0 , it suffices that
d 4
(79) + (C1 + d) 2 ≤ ǫ ≤ 1.
e e
 EXPLICIT BOUNDS FOR THE RIEMANN ZETA-FUNCTION ON THE 1-LINE 23

Here, we used that (log log t′ )2 / log t′ reaches its maximum of 4/e2 at t′ = te .
Overall, combining (77) and (78) with the trivial bound |ζ(s′ )| ≥ ζ(2σ ′ )/ζ(σ ′ ),
easily seen on considering the Euler product of zeta, we obtain that throughout the
disk |s − s′ | ≤ r,
ζ(s) ζ(σ ′ ) B ′ X(σ ′ )
≤ A3 log t = A3 logB t′ ,
ζ(s′ ) ζ(2σ ′ ) σ′ − 1
where
ζ(σ ′ )(σ ′ − 1)
X(σ ′ ) = .
ζ(2σ ′ )
Since σ ′ = σ ′ (t′ ) decreases to 1 with t′ ≥ t0 , X(σ ′ ) decreases to 1/ζ(2) with t′ .
Thus, recalling from (73) that σ ′ = 1 + δt′ and observing A3 is also decreasing in
t′ ≥ t0 , we obtain that throughout the disk |s − s′ | ≤ r,
ζ(s) Amax Xmax logB+1 t′
(80) ≤ ,
ζ(s′ ) d log log t′
where Amax = A3 (t0 ) and Xmax = X(σ ′ (t0 )).
In summary, when applying Lemma 5 in the disk |s − s′ | ≤ r, with r as in (78)
and subject to (79), we may take
 
′ ′ Amax Xmax
log A1 = (B + 1) log log t − log log log t + log .
d
Next, in preparation for applying Lemma 6, we want to ensure that the inter-
section of the disk |s − s′ | ≤ r and the right half-plane σ ≥ σ ′ − αr lies entirely in
a zero-free region for zeta. By Lemma 17, we have the following zero-free region,
valid for t ≥ 3.
c0 log log t 1
ζ(σ + it) 6= 0 for σ > 1 − , where c0 = .
log t 21.432
Since 1 − c0 log log t/ log t is monotonically increasing for t ≥ t0 and since t′ ≥ t0
and r ≤ ǫ, it suffices for our purposes to require that
c0 log log(t′ + ǫ)
σ ′ − αr ≥ 1 − .
log(t′ + ǫ)
In other words, it suffices that
c0 log log(t′ + ǫ)
 
1
α≤ δt′ + .
r log(t′ + ǫ)
Therefore, subject to α < 1/2 as demanded by Lemma 6, the following choice of α
works.
1 (d + c1 ) log log t′ d + c1
(81) α= · ′
= ,
r log t d + C1 log log t′
where (using monotonicity to deduce the inequality below)
c0 log t0
(82) c1 := , so that c1 ≤ c0 .
log(t0 + ǫ)
In particular, the constraint α < 1/2 is fulfilled if
d + c1 1
(83) < .
d + C1 log log t0 2
24 G. HIARY, N. LEONG AND A. YANG

Now, to determine A2 , we utilise [HT88, Lemma 70.1], which asserts that if

σ > 1, then
ζ ′ (s) ζ ′ (σ) 1
≤− < .
ζ(s) ζ(σ) σ−1
Taking s = s′ in this inequality leads to
ζ ′ (s′ ) 1 log t′
′
< = .
ζ(s ) δt′ d log log t′
So, we may take A2 in Lemma 6 to be
log t′
A2 = .
d log log t′
Lastly, in applying Lemma 6, we need to be able to reach the 1-line from our
position at s′ . So, we set
δt′ d
β= = .
αr c1 + d
Clearly, β < 1, as required by Lemma 6. The conclusion of the lemma therefore
holds throughout the disk |s − s′ | ≤ αβr = δt′ . This includes, in particular, the
horizontal line segment s = σ + it′ , with 1 ≤ σ ≤ 1 + 2δt′ .
Putting it all together, and making the substitution t′ → t, Lemma 6 hence
furnishes the bound
ζ ′ (s) 8β log A1 1+β
(84) ≤ + A2 ,
ζ(s) (1 − β)(1 − 2α)2 r 1−β
valid in the region
2d log log t
1≤σ ≤1+ , t ≥ t0 .
log t
We now input the values we determined for A1 , A2 , r, α and β into (84). We
also note
 
1+β 2d 1+β 1 2 log t
= 1+ , so that A2 = + ,
1−β c1 1−β d c1 log log t
as well as
 2
8β 8d 1 C1 log log t + d
= , = .
1−β c1 (1 − 2α)2 C1 log log t − d − 2c1
Hence, on observing that 1/(1 − 2α)2 is decreasing in t, we obtain the following
lemma.
Lemma 9. Let t0 ≥ ee and suppose that t ≥ t0 . Let C, d and ǫ be any positive
constants satisfying the constraints (66), (79) and (83). If
2d log log t
1≤σ ≤1+ ,
log t
then
ζ ′ (σ + it) log t
≤ Q1 (C, d, ǫ, t0 ) ,
ζ(σ + it) log log t
where
" !#
log log log t0 1 Amax Xmax
Q1 (C, d, ǫ, t0 ) := λ1 B+1− + log + λ2 .
log log t0 log log t0 d
 EXPLICIT BOUNDS FOR THE RIEMANN ZETA-FUNCTION ON THE 1-LINE 25

Here,
 2
1 8d C1 log log t0 + d 1 2
λ1 := · · , λ2 := + .
C1 c1 C1 log log t0 − d − 2c1 d c1
The number B is defined as in Lemma 7 and depends on C. The numbers Amax
and Xmax are defined as in (80) and depend on ǫ and t0 . The number C1 is defined
in (76), and depends on C, ǫ and t0 . The number c1 is defined in (82) and depends
on ǫ, t0 and c0 = 1/21.432.
5.3. Bounding |1/ζ(1+it)|. Moving from a bound on the logarithmic derivative of
ζ to one on the reciprocal of ζ is done in the usual way (see [Tit86, Theorem 3.11]),
with some improvements using the trigonometric polynomial (see [Car+22, Propo-
sition A.2]).
Lemma 10. Let t0 ≥ ee and d > 0 be constants. Suppose that for each t ≥ t0 ,
2d log log t ζ ′ (σ + it) log t
if 1≤σ ≤1+ , then ≤ R1 .
log t ζ(σ + it) log log t
Then for any t ≥ t0 and any real number d1 such that 0 < d1 ≤ 2d, we have
1 log t
≤ Q2 (d1 , t0 ) ,
ζ(1 + it) log log t
where
 3/4   1/4
log log t0 1 ′ log 2
Q2 (d1 , t0 ) = exp(R1 d1 ) + C (d1 , 2t0 ) +1 .
log t0 d1 log t0
The number C ′ is defined as in (72).
Proof. Suppose t ≥ t0 . Let d1 is a real positive parameter such that d1 ≤ 2d. Let
δ1 = d1 log log t/ log t. It is easy to see that
1
log = −Re log ζ(1 + it)
ζ(1 + it)
1+δ1
ζ′
Z  
= −Re log ζ (1 + δ1 + it) + Re (σ + it) dσ.
1 ζ
So, using our suppositions, we have
1
(85) log ≤ − log |ζ (1 + δ1 + it)| + R1 d1 .
ζ(1 + it)
On the other hand, for σ > 1, and using the classical nonnegativity argument
involving the trigonometric polynomial 3 + 4 cos θ + cos 2θ, gives that
ζ 3 (σ)|ζ 4 (σ + it)ζ(σ + 2it)| ≥ 1.
(See e.g. [Tit86, § 3.3].) Combining this with the inequality ζ(σ) ≤ σ/(σ − 1) and
Lemma 8, we therefore obtain, on taking σ = 1 + δ1 ,
1
≤ |ζ(1 + δ1 )|3/4 |ζ(1 + δ1 + 2it)|1/4
ζ(1 + δ1 + it)
(86)  3/4  1/4
1 + δ1 ′ log 2t
≤ C (d1 , 2t0 ) .
δ1 log log 2t
26 G. HIARY, N. LEONG AND A. YANG

We now observe
 
1 + δ1 log log t0 1 log t
≤ + ,
δ1 log t0 d1 log log t
(87)  
log 2t log 2 log t
≤ +1 .
log log 2t log t0 log log t
So that, combining (87) and (86), and noting by (85) we have
1 exp(R1 d1 )
≤ ,
ζ(1 + it) |ζ(1 + δ1 + it)|
the lemma follows. 

5.4. Numeric calculations. We define the following upper bounds:

Q1 (C, d, ǫ, t0 ) ≤ R1 and Q2 (d1 , t0 ) ≤ R2 .
The numeric calculations for R1 and R2 proceed as follows. First fix t0 ≥ ee
and a positive ǫ ≤ 1. For the purposes of obtaining an upper bound R1 , we then
optimise Q1 (C, d, ǫ, t0 ) in Lemma 9 over C and d, all the while ensuring C, d, and
ǫ are subject to our constraints (66), (79) and (83)
Thereafter, with these values of d and R1 , the relationship d1 ≤ 2d allows us to
obtain R2 by optimising Q2 (d1 , t0 ) in Lemma 10 over d1 .
Our choices of
t0 = 500, ǫ = 0.52,

e2
C= , d = 0.018, d1 = 0.0065,
8
give us
r ≤ 0.502, α ≤ 0.038, β = 0.278 . . . ,
and the upper bounds
R1 = 154, R2 = 430.5.
In summary, we have for t ≥ 500,
1 log t
(88) ≤ 430.5 .
|ζ(1 + it)| log log t
Furthermore, in the region
9 log log t
1≤σ ≤1+ ,
250 log t
we have for t ≥ 500,
ζ′ log t
(89) (σ + it) ≤ 154 .
ζ log log t
Finally, to cover the range from 3 ≤ t ≤ 500 in (88), we refer to the proof of
Proposition A.2 in [Car+22], where via interval arithmetic, they computed that for
2 ≤ t ≤ 500,
1
≤ 2.079 log t.
|ζ(1 + it)|
 EXPLICIT BOUNDS FOR THE RIEMANN ZETA-FUNCTION ON THE 1-LINE 27

Comparing this with the estimate in (88), we see after a simple verification that
the bound
1 log t
≤ 430.5
|ζ(1 + it)| log log t
holds for all t ≥ 3.
This concludes our proof of Theorem 2. These bounds can certainly be improved.
Some possible avenues of attack would be to treat the sum over zeros in the proof
of Lemma 6 more carefully, or to use a larger zero-free region.
Remark. It is possible with our methods to directly get a bound like (89) for
t < 500, especially if one were not concerned with the size of R2 . For instance,
choosing t0 = 16, ǫ = 0.52, C = e2 /8, d = 0.0147 and d1 = 0.0056, would return
for t ≥ 16 that if
147 log log t
1≤σ ≤1+ ,
5000 log t
then
ζ′ log t
(σ + it) ≤ 178.4 .
ζ log log t
While this would cause a modest jump in R1 , the size of R2 would increase sig-
nificantly to R2 = 513.6. This is because R2 increases exponentially with R1 , as
evident from Lemma (10).

6. Background lemmas
Lemma 11 ([Sim20] Corollary 2). Let s = σ+it where 1/2 ≤ σ ≤ 1 and t ≥ t0 > 0.
If h0 > (2π)−1 , then
X 1   
1 1 1 1
q
−2
(90) ζ(s) − ≤ h + + 3 1 + t0 1− cot .
ns (ht0 )σ 2 2h 2h
1≤n≤h0 t
p
Theorem 12 ([Pat22] Theorem 3.9). For t > 0 and n1 = ⌊ t/(2π)⌋ we have
n1 n1
X 1 g(t) X 1
|ζ(1 + it)| ≤ + +R
n=1
n1+it t1/2 n=1 n−it
where
√
 
5 π
g(t) = 2π exp + ,
3t2 6t
and
r  r !
1 π g(t) 1 π g(t)
R = R(t) := 1/2 + + 9 +p
t 2 2 t 2 π(3 − 2 log 2)
968π 3/2 + g(t)242π
 
1
+ 3/2 .
t 700
Lemma 13 (Explicit third derivative test). Let f (x) be a real-valued function with
three continuous derivatives on [N + 1, N + L]. Suppose there are W > 0 and λ ≥ 1
such that 1/W ≤ |f ′′′ (x)| ≤ λ/W for N + 1 ≤ x ≤ N + L. If η > 0, then
N +L 2
X
2πif (n)
(91) e ≤ (LW −1/3 + η)(αL + βW 2/3 ),
n=N +1
28 G. HIARY, N. LEONG AND A. YANG

where
1 32 p 2λη 2λ
α := + √ η + W −1/3 + 1/3 + 2/3 ,
η 15 π W W
(92)
64 1 4
β := √ √ + 1/3 .
3 π η W
Remark. If W ≤ 1, say, then α will be greater than 1, and so αL2 ≥ L2 . Hence,
for small W , it is better to use the triangle inequality than Lemma 13.
Lemma 14 (Another explicit third derivative test). Let a and N be integers such
that N ≥ 1. Suppose f (x) has three continuous derivatives and f ′′′ (x) is monotonic
over the interval (a, a + N ]. Suppose further that there are numbers λ3 > 0 and
ω > 1 such that λ3 ≤ |f ′′′ (x)| ≤ ωλ3 for all x ∈ (a, a + N ]. Then for any η3 > 0,
a+N
1/6 −1/6
X
e2πif (n) ≤ A3 ω 1/2 N λ3 + B3 N 1/2 λ3
n=a+1

where
s
1 32 1
q 
1/3 1/3 1/3
A3 = A3 (η3 , ω) := + √ η3 + λ0 + η3 + λ0 λ0 δ3 ,
η3 ω 15 π 3
√
32
B3 = B3 (η3 ) := √ δ ,
1/4 3
1/4
3π η3
and λ0 and δ3 are defined by
!−3
1/2
1 32η3 ω
λ0 = λ0 (η3 , ω) := + √ ,
η3 15 π
s r
1 1 3 3/2
δ3 = δ3 (η3 ) := + 1 + π 1/2 η3 .
2 2 8
Remark. The reason we state this additional third derivative test is because A3
and B3 comprise the boundary cases for the recursive formulas in Lemma 15.
Lemma 15 (General explicit k-th derivative test). Let a, N and k be integers such
that N ≥ 1 and k ≥ 4. Suppose f (x) is equipped with k continuous derivatives and
f (k) (x)is monotonic over the interval (a, a + N ]. Suppose further that there are
numbers λk > 0 and ω > 1 such that λk ≤ |f (k) (x)| ≤ ωλk for all x ∈ (a, a + N ].
Then for any η3 > 0,
a+N
1/(2K−2) −1/(2K−2)
X
e2πif (n) ≤ Ak ω 2/K N λk + Bk N 1−2/K λk ,
n=a+1

where K = 2k−1 , Ak and Bk are defined recursively via the formulas

!
−1/K 219/12 (K − 1)
(93) Ak+1 (η3 , ω) := δk ω +p Ak (η3 , ω)1/2 ,
(2K − 1)(4K − 3)

23/2 (K − 1)
(94) Bk+1 (η3 ) := δk p Bk (η3 )1/2 ,
(2K − 3)(4K − 5)
 REFERENCES 29

with A3 and B3 as in Lemma 14, and δk defined by

s
 1/K
2 9π
(95) δk = δk (η3 ) := 1 + η3 .
23371−2/K 1024
Lemma 16. Let a, b and k be integers such that a > 0, b > a and k ≥ 3. Let
t > 0. Let h > 1 be a number such that b ≤ ha. Then for any η3 > 0,
X 1
≤ Ck a−k/(2K−2) t1/(2K−2) + Dk ak/(2K−2)−2/K t−1/(2K−2) ,
n1+it
a<n≤b

where
  1
(k − 1)! 2K−2
Ck = Ck (η3 , h) := Ak (η3 , hk )h2k/K−k/(2K−2) (h − 1) ,
2π
 1
 2K−2
2π
Dk = Dk (η3 , h) := Bk (η3 )hk/(2K−2) (h − 1)1−2/K .
(k − 1)!
and Ak (η3 , hk ) and Bk (η3 ) defined in accordance with Lemma 15.
Proof. This is a special case of Lemma 15 and Lemma 14 with the phase function
t t (k − 1)!
f (x) = − log x, and hence f (k) (x) = (−1)k .
2π 2π xk
Since a < b ≤ ha by assumption, λk ≤ |f (k) (x)| ≤ ωλk for a < x ≤ b, where
t (k − 1)!
λk = , ω = hk .
2π hk ak
Suppose that k ≥ 4. Then, we may apply Lemma 15 with these values of λk and
ω, and with N = b − a. This furnishes a bound that increases with N . Noting that
N ≤ (h − 1)a, we thus obtain
X 1  1/(2K−2)
k k 2/K t (k − 1)!
≤ A (η
k 3 , h )(h ) (h − 1)a
nit 2π hk ak
a<n≤b
 −1/(2K−2)
t (k − 1)!
+ Bk (η3 )((h − 1)a)1−2/K
2π hk ak
= Ck a1−k/(2K−2) t1/(2K−2) + Dk a1−2/K+k/(2K−2) t−1/(2K−2) .
The desired result hence follows from partial summation in this case. If k = 3, then
the result follows from Lemma 14 using a completely similar calculation. 
Lemma 17. [Yan23, Corollary 1.2] There are no zeroes of ζ(σ + it) in the region
log log |t|
σ >1− , |t| ≥ 3.
21.432 log |t|
References
[BR02] G. Bastien and M. Rogalski. “Convexité, complête monotonie et inégalités
sur les fonctions zêta et gamma, sur les fonctions des opérateurs de
Baskakov et sur des fonctions arithmétiques”. In: Canadian Journal of
Mathematics 54.5 (2002), pp. 916–944.
30 REFERENCES

[Car+22] E. Carneiro, A. Chirre, H. A. Helfgott, and J. Mejı́a-Cordero. “Op-

timality for the two-parameter quadratic sieve”. In: Acta Arithmetica
203 (2022), pp. 195–226.
[For02a] K. Ford. “Vinogradov’s Integral and Bounds for the Riemann Zeta Func-
tion”. In: Proceedings of the London Mathematical Society 85.3 (2002),
pp. 565–633.
[For02b] K. Ford. “Zero-free regions for the Riemann zeta function”. In: Number
theory for the millennium, II (Urbana, IL, 2000). A K Peters, Natick,
MA, 2002, pp. 25–56.
[HT88] R. R. Hall and G. Tenenbaum. Divisors. Vol. 90. Cambridge Tracts in
Math. Cambridge University Press, Cambridge, 1988.
[HSW21] E. Hasanalizade, Q. Shen, and P.-J. Wong. “Counting zeros of the Rie-
mann zeta function”. In: (2021).
[Hia16] G. A. Hiary. “An explicit van der Corput estimate for ζ(1/2 + it)”. In:
Indagationes Mathematicae 27.2 (Mar. 2016), pp. 524–533.
[HPY22] G. A. Hiary, D. Patel, and A. Yang. “An improved explicit estimate for
ζ(1/2 + it)”. In: (2022).
[KLN18] H. Kadiri, A. Lumley, and N. Ng. “Explicit zero density for the Riemann
zeta function”. In: Journal of Mathematical Analysis and Applications
465.1 (2018), pp. 22–46.
[LLS15] Y. Lamzouri, X. Li, and K. Soundararajan. “Conditional bounds for
the least quadratic non-residue and related problems”. In: Math. Comp.
84.295 (2015), pp. 2391–2412.
[Lit26] J. E. Littlewood. “On the Riemann Zeta-Function”. In: Proceedings of
the London Mathematical Society s2-24.1 (1926), pp. 175–201.
[Lit28] J. E. Littlewood. “Mathematical Notes (5): On the Function 1/ζ(1 +
ti)”. In: Proceedings of the London Mathematical Society s2-27.1 (1928),
pp. 349–357.
[MV07] H. L. Montgomery and R. C. Vaughan. Multiplicative number theory I:
Classical theory. 97. Cambridge university press, 2007.
[Pat21] D. Patel. “Explicit sub-Weyl bound for the Riemann Zeta function”.
PhD Thesis. The Ohio State University, 2021.
[Pat22] D. Patel. “An explicit upper bound for |ζ(1 + it)|”. In: Indagationes
Mathematicae 33.5 (2022), pp. 1012–1032.
[Sim20] A. Simonič. “Explicit zero density estimate for the Riemann zeta-function
near the critical line”. In: Journal of Mathematical Analysis and Appli-
cations 491.1, 124303, 41pp. (Nov. 2020).
[Sou09] K. Soundararajan. “Moments of the Riemann zeta function”. In: Ann.
of Math. (2) 170.2 (2009), pp. 981–993.
[Tit39] E. C. Titchmarsh. The theory of functions. Oxford university press,
1939.
[Tit86] E. C. Titchmarsh. The Theory of the Riemann Zeta-function. Oxford:
Oxford Science Publications, 1986.
[Tru14a] T. Trudgian. “A new upper bound for |ζ(1 + it)|”. In: Bulletin of the
Australian Mathematical Society 89.2 (2014).
[Tru15] T. Trudgian. “Explicit bounds on the logarithmic derivative and the re-
ciprocal of the Riemann zeta-function”. In: Functiones et Approximatio
Commentarii Mathematici 52.2 (2015), pp. 253–261.
 REFERENCES 31

[Tru14b] T. S. Trudgian. “An improved upper bound for the argument of the
Riemann zeta-function on the critical line II”. In: Journal of Number
Theory 134 (2014), pp. 280–292.
[Wey21] H. Weyl. “Zur Abschätzung von ζ(1+ti)”. de. In: Mathematische Zeitschrift
10.1-2 (Apr. 1921), pp. 88–101.
[Yan23] A. Yang. “Explicit bounds on ζ(s) in the critical strip and a zero-free
region”. In: (2023). Preprint available at arXiv:2301.03165.
32 REFERENCES

GH: Department of Mathematics, The Ohio State University, 231 West 18th Ave,
Columbus, OH 43210, USA
Email address: [email protected]

NL: School of Science, University of New South Wales (Canberra), Northcott

Drive, Campbell, ACT 2600, Australia
Email address: [email protected]

AY: School of Science, University of New South Wales (Canberra), Northcott

Drive, Campbell, ACT 2600, Australia
Email address: [email protected]