Macdonald function
modified cylinder function, Bessel function of imaginary argument
A function
where \nu is an arbitrary non-integral real number and
I _ \nu ( z) = \ \sum _ { m= } 0 ^ \infty \frac{\left ( \frac{z}{2} \right ) ^ {\nu + 2 m } }{m ! \Gamma ( \nu + m + 1 ) }
is a cylinder function with pure imaginary argument (cf. Cylinder functions). They have been discussed by H.M. Macdonald [1]. If n is an integer, then
K _ {n} ( z) = \lim\limits _ {\nu \rightarrow n } K _ \nu ( z) .
The Macdonald function K _ \nu ( z) is the solution of the differential equation
\tag{* } z ^ {2} \frac{d ^ {2} y }{d z ^ {2} } + z \frac{d y }{d z } - ( z ^ {2} + \nu ^ {2} ) y = 0
that tends exponentially to zero as z \rightarrow \infty and takes positive values. The functions I _ \nu ( z) and K _ \nu ( z) form a fundamental system of solutions of (*).
For \nu \geq 0 , K _ \nu ( z) has roots only when \mathop{\rm Re} z < 0 . If \pi / 2 < | \mathop{\rm arg} z | < \pi , then the number of roots in these two sectors is equal to the even number nearest to \nu - 1 / 2 , provided that \nu - 1 / 2 is not an integer; in the latter case the number of roots is equal to \nu - 1 / 2 . For \mathop{\rm arg} z = \pm \pi there are no roots if \nu - 1 / 2 is not an integer.
Series and asymptotic representations are:
K _ {n + 1 / 2 } ( z) = \ \left ( \frac \pi {2z} \right ) ^ {1/2} e ^ {-} z \sum _ { r= } 0 ^ { n } \frac{( n + r ) ! }{r ! ( n - r ) ! ( 2 z ) ^ {r} } ,
where n is a non-negative integer;
K _ {0} ( z) = \ - \mathop{\rm ln} \left ( \frac{z}{2} \right ) I _ {0} ( z) + \sum _ { m= } 0 ^ \infty \left ( \frac{z}{2} \right ) ^ {2m} \frac{1}{( m ! ) ^ {2} } \psi ( m + 1 ) ,
\psi ( 1) = - C ,\ \psi ( m + 1 ) = 1 + \frac{1}{2} + \dots + \frac{1}{m} - C ,
where C = 0. 5772157 \dots is the Euler constant;
K _ {n} ( z) = \ \frac{1}{2} \sum _ { m= } 0 ^ { n- } 1 \frac{( - 1 ) ^ {m} ( n - m - 1 ) ! }{m ! ( z / 2 ) ^ {n - 2 m } } +
+ ( - 1 ) ^ {n-} 1 \sum _ { m= } 0 ^ \infty \frac{( z / 2 ) ^ {n + 2 m } }{m ! ( n + m ) ! } \left \{ \mathop{\rm ln} \left ( \frac{z}{2} \right ) - \frac{\psi ( m + 1 ) - \psi ( n + m + 1 ) }{2} \right \} ,
where n \geq 1 is an integer;
K _ {\nu\ } \sim
\sim \ \left ( \frac \pi {2z} \right ) ^ {1/2} e ^ {-} z \left [ 1 + \frac{ 4 \nu ^ {2} - 1 ^ {2} }{1 ! 8 z } + \frac{( 4 \nu ^ {2} - 1 ^ {2} ) ( 4 \nu ^ {2} - 3 ^ {2} ) }{2 ! ( 8 z ) ^ {2} } + \dots \right ] ,
for large z and | \mathop{\rm arg} z | < \pi / 2 .
Recurrence formulas:
K _ {\nu - 1 } ( z) - K _ {\nu + 1 } ( z) = - \frac{2 \nu }{z} K _ \nu ( z) ,
K _ {\nu - 1 } ( z) + K _ {\nu + 1 } ( z) = - 2 \frac{d K _ \nu ( z) }{d z } .
References
[1] | H.M. Macdonald, "Zeroes of the Bessel functions" Proc. London Math. Soc. , 30 (1899) pp. 165–179 |
[2] | G.N. Watson, "A treatise on the theory of Bessel functions" , 1–2 , Cambridge Univ. Press (1952) |
Macdonald function. Encyclopedia of Mathematics. URL: http://encyclopediaofmath.org/index.php?title=Macdonald_function&oldid=47744