28.5 Second Solutions \mathop{\mathrm{fe}_{{n}}\/}\nolimits, \mathop{\mathrm{ge}_{{n}}\/}\nolimits28.7 Analytic Continuation of Eigenvalues

§28.6 Expansions for Small q

Contents

§28.6(i) Eigenvalues

Leading terms of the power series for \mathop{a_{{m}}\/}\nolimits\!\left(q\right) and \mathop{b_{{m}}\/}\nolimits\!\left(q\right) for m\leq 6 are:

28.6.1\mathop{a_{{0}}\/}\nolimits\!\left(q\right)=-\tfrac{1}{2}q^{2}+\tfrac{7}{128}q^{4}-\tfrac{29}{2304}q^{6}+\tfrac{68687}{188\; 74368}q^{8}+\cdots,
28.6.2\mathop{a_{{1}}\/}\nolimits\!\left(q\right)=1+q-\tfrac{1}{8}q^{2}-\tfrac{1}{64}q^{3}-\tfrac{1}{1536}q^{4}+\tfrac{11}{36864}q^{5}+\tfrac{49}{5\; 89824}q^{6}+\tfrac{55}{94\; 37184}q^{7}-\tfrac{83}{353\; 89440}q^{8}+\cdots,
28.6.3\mathop{b_{{1}}\/}\nolimits\!\left(q\right)=1-q-\tfrac{1}{8}q^{2}+\tfrac{1}{64}q^{3}-\tfrac{1}{1536}q^{4}-\tfrac{11}{36864}q^{5}+\tfrac{49}{5\; 89824}q^{6}-\tfrac{55}{94\; 37184}q^{7}-\tfrac{83}{353\; 89440}q^{8}+\cdots,
28.6.4\mathop{a_{{2}}\/}\nolimits\!\left(q\right)=4+\tfrac{5}{12}q^{2}-\tfrac{763}{13824}q^{4}+\tfrac{10\; 0 2401}{796\; 26240}q^{6}-\tfrac{16690\; 68401}{45\; 86471\; 42400}q^{8}+\cdots,
28.6.5\mathop{b_{{2}}\/}\nolimits\!\left(q\right)=4-\tfrac{1}{12}q^{2}+\tfrac{5}{13824}q^{4}-\tfrac{289}{796\; 26240}q^{6}+\tfrac{21391}{45\; 86471\; 42400}q^{8}+\cdots,
28.6.6\mathop{a_{{3}}\/}\nolimits\!\left(q\right)=9+\tfrac{1}{16}q^{2}+\tfrac{1}{64}q^{3}+\tfrac{13}{20480}q^{4}-\tfrac{5}{16384}q^{5}-\tfrac{1961}{235\; 92960}q^{6}-\tfrac{609}{1048\; 57600}q^{7}+\cdots,
28.6.7\mathop{b_{{3}}\/}\nolimits\!\left(q\right)=9+\tfrac{1}{16}q^{2}-\tfrac{1}{64}q^{3}+\tfrac{13}{20480}q^{4}+\tfrac{5}{16384}q^{5}-\tfrac{1961}{235\; 92960}q^{6}+\tfrac{609}{1048\; 57600}q^{7}+\cdots,
28.6.8\mathop{a_{{4}}\/}\nolimits\!\left(q\right)=16+\tfrac{1}{30}q^{2}+\tfrac{433}{8\; 64000}q^{4}-\tfrac{5701}{27216\; 0 0 0 0 0}q^{6}+\cdots,
28.6.9\mathop{b_{{4}}\/}\nolimits\!\left(q\right)=16+\tfrac{1}{30}q^{2}-\tfrac{317}{8\; 64000}q^{4}+\tfrac{10049}{27216\; 0 0 0 0 0}q^{6}+\cdots,
28.6.10\mathop{a_{{5}}\/}\nolimits\!\left(q\right)=25+\tfrac{1}{48}q^{2}+\tfrac{11}{7\; 74144}q^{4}+\tfrac{1}{1\; 47456}q^{5}+\tfrac{37}{8918\; 13888}q^{6}+\cdots,
28.6.11\mathop{b_{{5}}\/}\nolimits\!\left(q\right)=25+\tfrac{1}{48}q^{2}+\tfrac{11}{7\; 74144}q^{4}-\tfrac{1}{1\; 47456}q^{5}+\tfrac{37}{8918\; 13888}q^{6}+\cdots,
28.6.12\mathop{a_{{6}}\/}\nolimits\!\left(q\right)=36+\tfrac{1}{70}q^{2}+\tfrac{187}{439\; 0 4000}q^{4}+\tfrac{67\; 43617}{9293\; 59872\; 0 0 0 0 0}q^{6}+\cdots,
28.6.13\mathop{b_{{6}}\/}\nolimits\!\left(q\right)=36+\tfrac{1}{70}q^{2}+\tfrac{187}{439\; 0 4000}q^{4}-\tfrac{58\; 61633}{9293\; 59872\; 0 0 0 0 0}q^{6}+\cdots.

Leading terms of the of the power series for m=7,8,9,\dots are:

28.6.14\rselection{\mathop{a_{{m}}\/}\nolimits\!\left(q\right)\\
\mathop{b_{{m}}\/}\nolimits\!\left(q\right)}=m^{2}+\frac{1}{2(m^{2}-1)}q^{2}+\frac{5m^{2}+7}{32(m^{2}-1)^{3}(m^{2}-4)}q^{4}+\frac{9m^{4}+58m^{2}+29}{64(m^{2}-1)^{5}(m^{2}-4)(m^{2}-9)}q^{6}+\cdots.

The coefficients of the power series of \mathop{a_{{2n}}\/}\nolimits\!\left(q\right), \mathop{b_{{2n}}\/}\nolimits\!\left(q\right) and also \mathop{a_{{2n+1}}\/}\nolimits\!\left(q\right), \mathop{b_{{2n+1}}\/}\nolimits\!\left(q\right) are the same until the terms in q^{{2n-2}} and q^{{2n}}, respectively. Then

28.6.15\mathop{a_{{m}}\/}\nolimits\!\left(q\right)-\mathop{b_{{m}}\/}\nolimits\!\left(q\right)=\frac{2q^{m}}{\left(2^{{m-1}}(m-1)!\right)^{2}}\left(1+\mathop{O\/}\nolimits\!\left(q^{2}\right)\right).

Higher coefficients in the foregoing series can be found by equating coefficients in the following continued-fraction equations:

28.6.16a-(2n)^{2}-\cfrac{q^{2}}{a-(2n-2)^{2}-\cfrac{q^{2}}{a-(2n-4)^{2}-}}\cdots\cfrac{q^{2}}{a-2^{2}-\cfrac{2q^{2}}{a}}=-\cfrac{q^{2}}{(2n+2)^{2}-a-\cfrac{q^{2}}{(2n+4)^{2}-a-\cdots}},a=\mathop{a_{{2n}}\/}\nolimits\!\left(q\right),
28.6.17a-(2n+1)^{2}-\cfrac{q^{2}}{a-(2n-1)^{2}-}\cdots\cfrac{q^{2}}{a-3^{2}-\cfrac{q^{2}}{a-1^{2}-q}}=-\cfrac{q^{2}}{(2n+3)^{2}-a-\cfrac{q^{2}}{(2n+5)^{2}-a-\cdots}},a=\mathop{a_{{2n+1}}\/}\nolimits\!\left(q\right),
28.6.18a-(2n+1)^{2}-\cfrac{q^{2}}{a-(2n-1)^{2}-}\cdots\cfrac{q^{2}}{a-3^{2}-\cfrac{q^{2}}{a-1^{2}+q}}=-\cfrac{q^{2}}{(2n+3)^{2}-a-\cfrac{q^{2}}{(2n+5)^{2}-a-\cdots}},a=\mathop{b_{{2n+1}}\/}\nolimits\!\left(q\right),
28.6.19a-(2n+2)^{2}-\cfrac{q^{2}}{a-(2n)^{2}-\cfrac{q^{2}}{a-(2n-2)^{2}-}}\cdots\cfrac{q^{2}}{a-2^{2}}=-\cfrac{q^{2}}{(2n+4)^{2}-a-\cfrac{q^{2}}{(2n+6)^{2}-a-\cdots}},a=\mathop{b_{{2n+2}}\/}\nolimits\!\left(q\right).

Numerical values of the radii of convergence \rho _{n}^{{(j)}} of the power series (28.6.1)–(28.6.14) for n=0,1,\dots,9 are given in Table 28.6.1. Here j=1 for \mathop{a_{{2n}}\/}\nolimits\!\left(q\right), j=2 for \mathop{b_{{2n+2}}\/}\nolimits\!\left(q\right), and j=3 for \mathop{a_{{2n+1}}\/}\nolimits\!\left(q\right) and \mathop{b_{{2n+1}}\/}\nolimits\!\left(q\right). (Table 28.6.1 is reproduced from Meixner et al. (1980, §2.4).)

Table 28.6.1: Radii of convergence for power-series expansions of eigenvalues of Mathieu’s equation.
n \rho^{{(1)}}_{n} \rho^{{(2)}}_{n} \rho^{{(3)}}_{n}
0 or 1 1.46876 86138 6.92895 47588 3.76995 74940
2 7.26814 68935 16.80308 98254 11.27098 52655
3 16.47116 58923 30.09677 28376 22.85524 71216
4 30.42738 20960 48.13638 18593 38.52292 50099
5 47.80596 57026 69.59879 32769 58.27413 84472
6 69.92930 51764 95.80595 67052 82.10894 36067
7 95.47527 27072 125.43541 1314 110.02736 9210
8 125.76627 89677 159.81025 4642 142.02943 1279
9 159.47921 26694 197.60667 8692 178.11513 940

It is conjectured that for large n, the radii increase in proportion to the square of the eigenvalue number n; see Meixner et al. (1980, §2.4). It is known that

28.6.20\liminf _{{n\to\infty}}\frac{\rho _{n}^{{(j)}}}{n^{2}}\geq kk^{{\prime}}(\mathop{K\/}\nolimits\!\left(k\right))^{2}=2.04183\; 4\dots,

where k is the unique root of the equation 2\!\mathop{E\/}\nolimits\!\left(k\right)=\mathop{K\/}\nolimits\!\left(k\right) in the interval (0,1), and k^{{\prime}}=\sqrt{1-k^{2}}. For \mathop{E\/}\nolimits\!\left(k\right) and \mathop{K\/}\nolimits\!\left(k\right) see §19.2(ii).

§28.6(ii) Functions \mathop{\mathrm{ce}_{{n}}\/}\nolimits and \mathop{\mathrm{se}_{{n}}\/}\nolimits

Leading terms of the power series for the normalized functions are:

For m=3,4,5,\dots,

28.6.26\mathop{\mathrm{ce}_{{m}}\/}\nolimits\!\left(z,q\right)=\mathop{\cos\/}\nolimits mz-\frac{q}{4}\left(\frac{1}{m+1}\mathop{\cos\/}\nolimits(m+2)z-\frac{1}{m-1}\mathop{\cos\/}\nolimits(m-2)z\right)+\frac{q^{2}}{32}\left(\frac{1}{(m+1)(m+2)}\mathop{\cos\/}\nolimits(m+4)z+\frac{1}{(m-1)(m-2)}\mathop{\cos\/}\nolimits(m-4)z-\frac{2(m^{2}+1)}{(m^{2}-1)^{2}}\mathop{\cos\/}\nolimits mz\right)+\cdots.

For the corresponding expansions of \mathop{\mathrm{se}_{{m}}\/}\nolimits\!\left(z,q\right) for m=3,4,5,\dots change \mathop{\cos\/}\nolimits to \mathop{\sin\/}\nolimits everywhere in (28.6.26).

The radii of convergence of the series (28.6.21)–(28.6.26) are the same as the radii of the corresponding series for \mathop{a_{{n}}\/}\nolimits\!\left(q\right) and \mathop{b_{{n}}\/}\nolimits\!\left(q\right); compare Table 28.6.1 and (28.6.20).