The Rogers-Ramanujan continued fraction is a generalized continued fraction defined by
|
(1)
|
(Rogers 1894, Ramanujan 1957, Berndt et al. 1996, 1999, 2000). For , a choice of the fifth root fixes
, with
. It was discovered by Rogers (1894), independently by
Ramanujan around 1913, and again independently by Schur in 1917. Modulo the factor
of
added for convenience, it provides a geometric series q-analog
of the reciprocal of the golden ratio
|
(2)
|
The convergents of
, with
, are given by
|
(3)
| |||
|
(4)
| |||
|
(5)
| |||
|
(6)
| |||
|
(7)
| |||
|
(8)
|
(OEIS A128915 and A127836; Sills 2003, p. 25, identity 3-14).
For ,
the fraction can be expressed in closed form in terms of q-series
by
|
(9)
| |||
|
(10)
| |||
|
(11)
|
and in terms of the Ramanujan theta function
|
(12)
|
by
|
(13)
|
Writing with
in the upper half-plane
and
,
it can also be expressed in terms of the Dedekind
eta function
by
|
(14)
|
where
|
(15)
|
and the square root is chosen by analytic continuation from its positive value for purely imaginary (Trott 2004).
The coefficients of in the Maclaurin series
of
for
,
1, 2, ... are 1,
, 1, 0,
, 1,
, 1, 0,
, 2,
, ... (OEIS A007325).
The fraction converges quickly for points sufficiently far from the unit circle in the complex plane. For , the continued fraction converges, while for
its even and odd convergents have distinct limits.
For the normalized convergents
, these limits are
|
(16)
| |||
|
(17)
|
(Andrews et al. 1992; Sills 2003, pp. 107 and 109; Trott 2004).
Ramanujan asserted that is an algebraic
number for every positive rational
(Duke 2005). Special cases include
|
(18)
| |||
|
(19)
| |||
|
(20)
|
where is the root of
near 0.51142....
can be written as
|
(21)
|
(Yi 2001, Trott 2004). The values of have been computed by Trott for all values of
,
and the algebraic degrees of
,
, ... are 8, 4, 32, 8, 40, 16, 64, 16, 96, 20, ... (OEIS
A082682; Trott 2004).
satisfies the equalities
|
(22)
| |
|
(23)
|
where is a q-Pochhammer
symbol. It also satisfies
|
(24)
| |||
|
(25)
|
(Watson 1929a,b; Berndt 1991, pp. 265-267; Berndt et al. 1996, 2000; Son 1998).
For ,
taking the real fifth root at negative arguments and defining
|
(26)
| |||
|
(27)
| |||
|
(28)
| |||
|
(29)
|
these quantities satisfy the modular equations
|
(30)
| |||
|
(31)
| |||
|
(32)
| |||
|
(33)
| |||
|
(34)
| |||
|
(35)
| |||
|
(36)
| |||
|
(37)
|
(Berndt et al. 1996, 1999, 2000). Trott (2004) gives modular equations of orders 2 to 15 and the primes 17, 19, and 23.
As discussed by Hardy (Ramanujan 1962, pp. xxvii and xxviii), Berndt and Rankin (1995), and Berndt et al. (1996, 2000), Ramanujan also defined the generalized continued fraction
|
(38)
|
Ramanujan also considered the continued fraction
|
(39)
| |||
|
(40)
|
for wherever the denominator is nonzero (Berndt 1991, p. 30;
Berndt et al. 1996, 2000). The special case
is plotted above.
Terminating at a term gives
|
(41)
|
(Berndt et al. 1996, 2000).
The real roots of are 0.576149, 0.815600, 0.882493, 0.913806, 0.931949, 0.943785,
0.952115, ..., the smallest of which was found by Ramanujan (Berndt et al. 1999).
and its smallest positive root are related to the enumeration of coins in a fountain
(Berndt 1991, Berndt et al. 1996, 2000) and the study of birth and death processes
(Berndt et al. 1996, 2000; Parthasarathy et al. 1998). In general,
the least positive root
of
is given as
by
|
(42)
|
(OEIS A050203; Berndt et al. 1996, 2000). Ramanujan gave the approximations and
by omitting the error terms in
|
(43)
| |||
|
(44)
|
respectively. For , these approximations give
|
(45)
| |||
|
(46)
|
For with
in the upper half-plane
and the same choice
, the j-function
is related to
by the icosahedral equation
|
(47)
|
where (Duke 2005). As an example,
has
, so
. Substituting
into the equation, one of its factors will be a quartic
with the root
.
Furthermore, the numerator and the denominator (with a constant) can be combined to form a perfect square,
|
(48)
|
the two sides being polynomial invariants of the icosahedral group.