The tau function ,
also called Ramanujan's tau function, is defined for positive
integers
and is related to the divisor function
. It is defined via the Fourier
series of the modular discriminant
for
, where
is the upper half-plane,
by
|
(1)
|
(Apostol 1997, p. 20). The tau function is also given by the Cauchy product
|
(2)
| |||
|
(3)
|
where
is the divisor function (Apostol 1997, pp. 24
and 140),
,
and
.
The tau function has generating function
|
(4)
| |||
|
(5)
| |||
|
(6)
| |||
|
(7)
| |||
|
(8)
|
for , where
is a q-Pochhammer
symbol. The first few values are 1,
, 252,
, 4830, ... (OEIS A000594).
The tau function is given by the Wolfram
Language function RamanujanTau[n].
The series
|
(9)
|
for is known as the tau
Dirichlet series.
Lehmer (1947) conjectured that for all positive integers
, an assertion sometimes known as Lehmer's
conjecture. Lehmer verified the conjecture
for
(Apostol 1997, p. 22).
The following table summarizes historical bounds
for which
holds for all positive
integers
.
| reference | |
| 3316799 | Lehmer (1947) |
| 214928639999 | Lehmer (1949) |
| Serre (1973, p. 98), Serre (1985) | |
| 1213229187071998 | Jennings (1993) |
| 22689242781695999 | Jordan and Kelly (1999) |
| 22798241520242687999 | Bosman (2007, p. 3) |
Ramanujan gave the computationally efficient triangular recurrence relation
|
(10)
|
for , with
and
, where
is the floor function
(Lehmer 1943, Jordan and Kelly 1999). It can be used with the formula
|
(11)
|
for prime and nonnegative integers
(Gandhi 1961, Jordan and Kelly 1999).
Ewell (1999) gave the beautiful formulas
|
(12)
| |
|
(13)
| |
|
(14)
| |
|
(15)
| |
|
(16)
| |
|
(17)
| |
|
(18)
| |
|
(19)
|
where
is a nonnegative integer and
is a positive integer
in these formulas,
,
and
is the exponent
of the exact power of 2 dividing the positive integer
,
is the odd part of
,
is the divisor function of
, and
is the sum of squares
function.
For prime ,
|
(20)
|
for , and
|
(21)
|
for integers and
(Mordell 1917; Apostol 1997, p. 92).
Ramanujan conjectured and Mordell (1917) proved that if , then
|
(22)
|
(Hardy 1999, p. 161). More generally,
|
(23)
|
which reduces to the first form if (Mordell 1917; Apostol 1997, p. 93).
For integers for which the displayed arguments are positive, Ramanujan
(1920) showed that
|
(24)
| |
|
(25)
| |
|
(26)
|
(Darling 1921, Wilton 1930),
|
(27)
|
for or one of the quadratic
nonresidues of 7, i.e., 3, 5, 6, and
|
(28)
|
for one of the quadratic
nonresidues of 23, i.e., 5, 7, 10, 11, 14, 15, 17, 19, 20, 21, 22 (Mordell 1922,
Wilton 1930). Ewell (1999) showed that
|
(29)
|
Ramanujan conjectured and Watson proved that is divisible by 691 for
almost all
,
specifically
|
(30)
|
where
is the divisor function (Wilton 1930; Apostol
1997, pp. 93 and 140; Jordan and Kelly 1999), and 691 is the absolute
value of the numerator of the Bernoulli
number
.
Additional congruences for positive integers include
|
(31)
| |||
|
(32)
| |||
|
(33)
| |||
|
(34)
| |||
|
(35)
| |||
|
(36)
| |||
|
(37)
| |||
|
(38)
| |||
|
(39)
| |||
|
(40)
|
where
is a prime number,
and
are integers,
is the Legendre symbol,
and
is the divisor function (Lehmer 1947, p. 432;
Serre 1968, p. 14-03; Swinnerton-Dyer 1988, Jordan and Kelly 1999). Negative
powers of
denote powers of its modular inverse modulo the
indicated modulus.
is almost always divisible
by
according to Ramanujan. In fact, for every fixed positive
integer
,
is divisible
by
for almost all positive
integers
(Serre 1976).
The summatory function of is defined for positive real
numbers
by
|
(41)
|
Here, the prime indicates that when is an integer, the last term
should be replaced by
.