# Theorem of Ince (1922)

If a nontrivial solution of Mathieu’s equation with $q\neq 0$ has period $\pi$ or $2\pi$, then any linearly independent solution cannot have either period.

Second solutions of (28.2.1) are given by

 28.5.1 $\mathop{\mathrm{fe}_{n}\/}\nolimits\!\left(z,q\right)=C_{n}(q)\left(z\mathop{% \mathrm{ce}_{n}\/}\nolimits\!\left(z,q\right)+f_{n}(z,q)\right),$ Defines: $\mathop{\mathrm{fe}_{n}\/}\nolimits\!\left(z,q\right)$: second solution, Mathieu’s equation Symbols: $\mathop{\mathrm{ce}_{n}\/}\nolimits\!\left(z,q\right)$: Mathieu function, $q=h^{2}$: parameter, $n$: integer, $z$: complex variable, $f_{n}(z,q)$: functions and $C_{n}(q)$: factor A&S Ref: 20.3.6 (in different form) Referenced by: §28.5(i), §28.5(i) Permalink: http://dlmf.nist.gov/28.5.E1 Encodings: TeX, pMML, png

when $a=\mathop{a_{n}\/}\nolimits\!\left(q\right)$, $n=0,1,2,\dots$, and by

 28.5.2 $\mathop{\mathrm{ge}_{n}\/}\nolimits\!\left(z,q\right)=S_{n}(q)\left(z\mathop{% \mathrm{se}_{n}\/}\nolimits\!\left(z,q\right)+g_{n}(z,q)\right),$ Defines: $\mathop{\mathrm{ge}_{n}\/}\nolimits\!\left(z,q\right)$: second solution, Mathieu’s equation Symbols: $\mathop{\mathrm{se}_{n}\/}\nolimits\!\left(z,q\right)$: Mathieu function, $q=h^{2}$: parameter, $n$: integer, $z$: complex variable, $g_{n}(z,q)$: functions and $S_{n}(q)$: factor A&S Ref: 20.3.7 (in different form) Referenced by: §28.5(i), §28.5(i) Permalink: http://dlmf.nist.gov/28.5.E2 Encodings: TeX, pMML, png

when $a=\mathop{b_{n}\/}\nolimits\!\left(q\right)$, $n=1,2,3,\dots$. For $m=0,1,2,\dots$, we have

 28.5.3 $\begin{array}[]{ll}f_{2m}(z,q)&\mbox{\pi-periodic, odd},\\ f_{2m+1}(z,q)&\mbox{\pi-antiperiodic, odd},\end{array}$ Symbols: $m$: integer, $q=h^{2}$: parameter, $n$: integer, $z$: complex variable and $f_{n}(z,q)$: functions Permalink: http://dlmf.nist.gov/28.5.E3 Encodings: TeX, pMML, png

and

 28.5.4 $\begin{array}[]{ll}g_{2m+1}(z,q)&\mbox{\pi-antiperiodic, even},\\ g_{2m+2}(z,q)&\mbox{\pi-periodic, even};\end{array}$ Symbols: $m$: integer, $q=h^{2}$: parameter, $n$: integer, $z$: complex variable and $g_{n}(z,q)$: functions Permalink: http://dlmf.nist.gov/28.5.E4 Encodings: TeX, pMML, png

compare §28.2(vi). The functions $f_{n}(z,q)$, $g_{n}(z,q)$ are unique.

The factors $C_{n}(q)$ and $S_{n}(q)$ in (28.5.1) and (28.5.2) are normalized so that

 28.5.5 $(C_{n}(q))^{2}\int_{0}^{2\pi}(f_{n}(x,q))^{2}dx=(S_{n}(q))^{2}\int_{0}^{2\pi}(% g_{n}(x,q))^{2}dx=\pi.$

As $q\to 0$ with $n\neq 0$, $C_{n}(q)\to 0$, $S_{n}(q)\to 0$, $C_{n}(q)f_{n}(z,q)\to\mathop{\sin\/}\nolimits nz$, and $S_{n}(q)g_{n}(z,q)\to\mathop{\cos\/}\nolimits nz$. This determines the signs of $C_{n}(q)$ and $S_{n}(q)$. (Other normalizations for $C_{n}(q)$ and $S_{n}(q)$ can be found in the literature, but most formulas—including connection formulas—are unaffected since $\mathop{\mathrm{fe}_{n}\/}\nolimits\!\left(z,q\right)/C_{n}(q)$ and $\mathop{\mathrm{ge}_{n}\/}\nolimits\!\left(z,q\right)/S_{n}(q)$ are invariant.)

 28.5.6 $\displaystyle C_{2m}(-q)$ $\displaystyle=C_{2m}(q),$ $\displaystyle C_{2m+1}(-q)$ $\displaystyle=S_{2m+1}(q),$ $\displaystyle S_{2m+2}(-q)$ $\displaystyle=S_{2m+2}(q).$

For $q=0$,

 28.5.7 $\displaystyle\mathop{\mathrm{fe}_{0}\/}\nolimits\!\left(z,0\right)$ $\displaystyle=z,$ $\displaystyle\mathop{\mathrm{fe}_{n}\/}\nolimits\!\left(z,0\right)$ $\displaystyle=\mathop{\sin\/}\nolimits nz,$ $\displaystyle\mathop{\mathrm{ge}_{n}\/}\nolimits\!\left(z,0\right)$ $\displaystyle=\mathop{\cos\/}\nolimits nz$, $n=1,2,3,\dots$;

compare (28.2.29).

As a consequence of the factor $z$ on the right-hand sides of (28.5.1), (28.5.2), all solutions of Mathieu’s equation that are linearly independent of the periodic solutions are unbounded as $z\to\pm\infty$ on $\Real$.

# Wronskians

 28.5.8 $\displaystyle\mathop{\mathscr{W}\/}\nolimits\left\{\mathop{\mathrm{ce}_{n}\/}% \nolimits,\mathop{\mathrm{fe}_{n}\/}\nolimits\right\}$ $\displaystyle=\mathop{\mathrm{ce}_{n}\/}\nolimits\!\left(0,q\right){\mathop{% \mathrm{fe}_{n}\/}\nolimits^{\prime}}\!\left(0,q\right),$ 28.5.9 $\displaystyle\mathop{\mathscr{W}\/}\nolimits\left\{\mathop{\mathrm{se}_{n}\/}% \nolimits,\mathop{\mathrm{ge}_{n}\/}\nolimits\right\}$ $\displaystyle=-{\mathop{\mathrm{se}_{n}\/}\nolimits^{\prime}}\!\left(0,q\right% )\mathop{\mathrm{ge}_{n}\/}\nolimits\!\left(0,q\right).$

See (28.22.12) for ${\mathop{\mathrm{fe}_{n}\/}\nolimits^{\prime}}\!\left(0,q\right)$ and $\mathop{\mathrm{ge}_{n}\/}\nolimits\!\left(0,q\right)$.

For further information on $C_{n}(q)$, $S_{n}(q)$, and expansions of $f_{n}(z,q)$, $g_{n}(z,q)$ in Fourier series or in series of $\mathop{\mathrm{ce}_{n}\/}\nolimits$, $\mathop{\mathrm{se}_{n}\/}\nolimits$ functions, see McLachlan (1947, Chapter VII) or Meixner and Schäfke (1954, §2.72).