20.

Mathieu Functions
BLANCH
GERTRUDE *

Contents
page
Mathematical Properties . . . . . . . . . . . . . . . . . . . . 722
20.1. MathieusEquation . . . . . . . . . . . . . . . . . 722
20.2. Determination of Characteristic Values . . . . . . . . . 722
20.3. FloquetsTheorem and Its Consequences . . . . . . . . 727
20.4. Other Solutions of MathieusEquation . . . . . . . . . 730
20.5. Properties of Orthogonality and Normalization . . . . . . 732
20.6. Solutions of MathieusModified Equation for Integral v . . 732
20.7. Representations by Integrals and Some Integral Equations . 735
20.8. Other Properties . . . . . . . . . . . . . . . . . . . 73 8
20.9. Asymptotic Representations . . . . . . . . . . . . . . 740
20.10. Comparative Notations . . . . . . . . . . . . . . . . 744
References . . . . . . . . . . . . . . . . . . . . . . . . . . 745
Table 20.1. Characteristic Values, Joining Factors, Some Critical
Values (O<q<m) . . . . . . . . . . . . . . . . . . . . 748
Even Solutions

Aeronautical Reaearch Laboratories, Wright-Patterson Air Force Base, Ohio.

721
 20. Mathieu Functions
Mathematical Properties
20.1. MathieusEquation
*
Canonical Form of the Differential Equation
where a is a constant. The above are equivalent
20.1.1 @+(a-2q cos 2v)y=0
dv2 to 20.1.1 and 20.1.2. The constants c and a are
often referred to as separation constants, due to the
MathieusModified Differential Equation
role they play in 20.1.5 and 20.1.6.
20.1.2 dY (a-2q cosh 2u)j=O
- (v=iu, y=f) For some physically important solutions, the
du2- function g must be periodic, of period T or 2s.
Relation Between Mathieu*eEquation and the Wave
It can be shown that there exists a countably
Equation for the Elliptic Cylinder infinite set of characteristic d u e s a,(q)which yield
The wave equation in Cartesian coordinates is even periodic solutions of 20.1.1; there is another
countably infinite sequence of characteristic values
b,(q) which yield odd periodic solutions of 20.1.1.
It is known that there exist periodic solutions of
period kr, where k is any positive integer. In
A solution W is obtainable by separation of vari- what follows, however, the term churucteristic
ables in elliptical coordinates. Thus, let value will be reserved for a value associated with
x=p cosh u cos v; y=p sinh u sin a; z=z; solutions of period r or 2 r only. These character-
istic values are of basic importance to the general
p a positive constant; 20.1.3 becomes theory of the differential equation for arbitrary
parameters a and q.
20.1.4
d2W
* 7+-----
2 An Algebraic Form of Mathiensm a t i o n
bz p2 (cash 2 ~ - C O2S ~ ) 20.1.7
Assuming a solution of the form (1 -t> dZY
-@-t -+(a+ 2q--4qt3y=O
d3/ (mv=t)
dt
w=v(z)f(4g(v) Relation to Spheroidal Wave Equation
and substituting the above into 20.1.4 one obtains, *
after dividing through by W, 20.1.8 (1-t2) d2Y
z2-2(6+l)t d-+(~-4qt)y=O
Y
dt
1 dav Thus, Mathieus equation is a special case of
- --,+Q=O
(Pdz 20.1.8, with E=-fi, e=a+2q.
where 20.2. Determination of Characteristic Values
A solution of 20.1.1 with v replaced by z, having
period T or 2 r is of the form
Since z, u, v are independent variables, it follows
m
that 20.2.1 y = C (A, cos mz+B,,, sin mz)
m-0
20.1.5 @-v
&z+cv=o
where Bo can be taken as zero. If the above is
substituted into 20.1.1 one obtains
where c is a constant.
20.2.2
Again, from the fact that Q=c and that u, v 0
are independent variables, one sets C [(a-m2)Am-q(Am-2+Am+z)I mz
20.1.6
m--2

+m-
- [(U-~~)B,-~(B,~~+B,+~)]
sin mz=O
-1
A-,, B-,=O m>o
722 *See page n.
 MATHIEU FUNCTIONS 723
Equation 20.2.2 can be reduced to one of four 20.2.18 v2=Go4, for odd solutions of period T,
simpler types, given in 20.2.3 and 20.2.4 below along with 20.2.15
m
These three-term recurrence relations among the
20.2.3 yo=C A,,,, cos (2m+p)z, p=O or 1 coefficients indicate that every Gmcan be developed
m=O

m
into two types of continued fractions. Thus
20.2.4 y l = C BZm+p
sin (2m+p)z, p=O or 1 20.2.15 is equivalent to
m-0
20.2.19
If p=O, the solution is of period r; if p=l, the
solution is of period 2r.
Recurrence Relations Among the Coefficients

Even solutions of period r:

20.2.5 aAo-qAz=O
20.2.6 (a-4)A2- q(2Ao+ A,) =0
20.2.7 (a-m2)A,-q(Am-2+Am+2)=0 where
e=d=O; *=2, if Gnr+2=A2JAaa-2
Even solutions of period 2r:
e = d = R = O , if Gm+2=B2JBZ,-~
20.2.8 (a- I)Ai--(Ai +A)=@
e = - l ; R=d=l, if G*i=A2r+1/Aar-1
along with 20.2.7 for m 23. *=d=&=l, if Gm+~=B~a+JB~,-1
Odd solutions of period r: The four choices of the parameters e, e, d
20.2.9 (a-4) B2-qB, =0 correspond to the four types of solutions 20.2.3-
20.2.4. Hereafter, it will be convenient to sep-
20.2.10 (U - mZ)Bm-q(Bm-2+Bm+2)
=O arate the characteristic values a into two major
subsets:
Odd solutions of period 2u:
a=a,, associated with even periodic solutions
20.2.11 (a- I)Bi+q(Bi-Bd =O,
a =b, associated with odd periodic solutions
along with 20.2.10 for m 23.
If 20.2.19 is suitably combined with 20.2.13-20.2.18
Let there result four types of continued fractions, the
20.2.12 Gem=AmJAm-2, GO*=BmIBm-2 ; roots of which yield the required characteristic
values
Gm=Gemor Go,,, when the same operations apply
to both, and no ambiguity is likely to arise. 20.2.21 vo-- 2 -
v2-
1
-
1
v4-
=O Roots: %,
v,- - . .
Further let
20.2.22
20.2.13 Vm=(a-m2)/q. 1 1 -
1
vl-l-- - . . =O Roots:
Equations 20.2.5-20.2.7 are equivalent to v3- v6- v,- , ht+l

2 1 1 1
20.2.14 Ge2= Vo; Ge4= V2-- 20.2.23 Va-- --. . .
Ge2 v4- vu- v,-
20.2.15 Gm=1J(Vm-Gm+2) (m23), 20.2.24
for even solutions of period r.
Similarly
20.2.16 If a is a root of 20.2.21-20.2.24, then the corre-
Vl - 1 = Ge,; for even solutions of period
2u, along with 20.2.15 sponding solution exists and is an entire function
of 2, for general complex values of p.
20.2.17 Vl+1=Go3,for odd solutions of period If p is real, then the Sturmian theory of second
2a, along with 20.2.15 order linear differential equations yields the
*See page n.
724 MATHIEU F " I 0 N S

following: Power Series for Characteristic Vduw

(a) For a fixed real q, characteristic values a, and
20.2.25
b, are real and distinct, if q f O ; ao<bl<al
.j q>O and a,(q)j br(q) ap- h(q)=-f+---
7p' 29$ 68687d
2 128 2304'18874368 + . . .
%<b
2< * *

proach 13 as Q approaches zero.

(b) A solution of 20.1.1 associated with a, or b, q2 d p4 l l @ +- 49q'
has T zeros in the interval OIz<r, (q real). a*(-p) =1- p---+-----
8 64 1536 36864 589824
(c) The form of 20.2.21 and 20.2.23 shows that if bi(P)
ai, is a root of 20.2.21 and q is different from ---55q' 83d + . . .
zero, then &, cannot be a root of 20.2.23; 9437184 35389440
similarly, no root of 20.2.22 can be a root of
$ Sp'
62(!?)=4---+-- 289p6
20.2.24 if q#O. It may be shown from other 12 13824 79626240
considerations that for a given point (a, q)
there can be at most one periodic solution of 21391ps + . . .
period r or 2r if q f O . This no longer holds +458647142400
for solutions of period sr, 8 2 3 ; for these all 5$ 763q' 1002401q"
solutions are periodic, if one is. 'b2(q)'4+12-m4+ 79626240
-1669068401$ + . . .
458647142400
a3(-p)=9+- P--+-+-
n" 13P' 5pd
16 64 20480 16384
bak)
-196lq" 609q'
23592960+104857600+ . . .
qz 317p' 10049q'
br(q)=16+---
30 864000+2721600000+ . . .

@ -
a&) =IS+-+ 433p'
- 5701q'
30 864000 2721600000+ * * *"

a5(-q)=25+-+---p' 1lq' a"

48 774144 147456
bdd
37 q"
'891813888' ' ' *

@ 187~' 5861633q'
6'(~)=36+%+43904000-92935987200000+ . * *

@ l87q' 6743617 p6
ac(b=36+%~43~4000+92935987200000+ . . .

}
For r27, and Iql not too large, a, is approxi-
mately equal to br, and the following approxima-
tion may be used
20.2.26
ar n" (59+7) q4
b, "+2(rr-+203!2(*-1)3(?-4)
a (9T4+589+29)$ + ...
20.1. Charadcrislic Valuea a,, b,
FXQVBE r=O,l(l)5 T64(T?-l)6(+-4) (9-9)
 The above expansion is not limited to integral
2 - p ~ + Psin62
8ez(z, * ) = s i n 2 Sin42 ( ~ - xsin22)+ . . .
values of T, and it is a very good approximation
for T of the form n+3 where n is an integer. In
case of integral values of T=n, the series holds 20.2.28
only up to terms not involving 9-na in the COS (T+2) 2-p 3
denominator. Subsequent terms must be derived d=cos(T2-P(x/2))-p "I
specially (as shown by Mathieu). Mulholland
cer(2*
d z , a> { [
4(~+1)

1
and Goldstein [20.38] have computed character-
istic values for purely imaginary q and found that -COS [(T--2)2--P(d2)1
a0 and a2have a common real value for 191 in the 4(T- 1)
neighborhood of 1.468; Bouwkamp [20.5] has
computed this number as q o = f i 1.46876852 to 8 COS [(T+~)Z-P(X/~)] COS [(?'-4)Z-p(?r/2)]
decimals. For values of -@>--iq0, a. and a, +@{ 32(~+1)(~+2) + 32(~-.1) (T-2)
are conjugate complex numbers. From equation COS [TZ-P(?r/2)] 2(9+1)
20.2.2.5 it follows that the radius of convergence - 32 [-]}+-*'
for the series defining a. is no greater than Iqo\.
It is shown in [20.36], section 2.25 that the radius with P=o for cer(z, q), P=l for Ser(2, Cr>, 7'23.
of convergence for %,,(q), n 2 2 is greater than 3.
Furthermore
a, -br =o(ff/f-'),T+ a.
yI
726 MATHIEU F u " S

20.3. Odd Periodic Mathieu Functions, Order8 1-6

FIQURE
q= 1.
I- 20.5. Odd Periodic Mathieu Functions, Orders 1-6
FIQURE
q= 10.

For coefficients associated with above functions

20.2.29
At(O) =a+; A:(O) =B:(O) = 1, r>O
A!e=i(-l)'q'/~! S! 22r-1]&+ . . ., s>O

A:+%=[(-l)?! @/4a(T+S)! s!]

&+%
... c:+
rs>O, C:=A: or B:
(r-s-l)!
A:-aa or B:-ae= t?!(r--l)! 4' c+. . .
n'
h p p t o t i c Expansion for Characterisiic Values, q 3 1

Let w=2r+l, q=w4(p,(p real. Then

20.2.30
&+I
4"b r + l & - 2 q + 2 M - ~ - - 271G
(w+3
di 4
_-------- d3 d,
21ZV 217(p3/2 2208 226 612 ' e
cp *

where
34 9
di=i5+-+-
zos w'
33 410 405
FIQURE
20.4. Even Periodic Mathieu Functions, Order8 0-6 d2=-+-+-
q= 10. W d d
 MATHIEU FUNCTIONS
63 1260 2943 486
--+-+-+-
da-d WI we we
527 15617 69001 41607
d 4 = 3 + 7 + 7 + 7
20.2.31 b,+* -a,-247+b~upt7+*e-4fi/?'!, p+==
(given in [20.36] without proof.)
20.3. Floquet's Theorem and Its Consequences
Since the coefficients of Mathieu's equation
20.3.1 y"+ (a-29 cos 2z)y=O
are periodic functions of z, it follows from the
known theory relating to such equations that
there exists a solution of the form
20.3.2 F,(q =efvrP(z),
where v depends on a and q, and P(z) is a periodic
function, of the same period as that of the
coefficients in 20.3.1, namely r. (Floquet's
theorem; see 120.161 or I20.221 for its more general
form.) The constant v is called the charactektic
exponent. Similarly
20.3.3 F,( -z) =e- i'zP(-z)
satisfies 20.3.1 whenever 20.3.2 does. Both F,(z)
and F,(-z) have the property
20.3.4
y(z+kr)=Cky(z), y=F,(z) or F,(-d,
(?=eivr for F,(z), C=e-'- for F,(-z)
Solutions having the property 20.3.4 will hereafler
be termed Floquet solutions. Whenever F,(z)
and F,(-z) are linearly independent, the general
solution of 20.3.1 can be put into the form
20.3.5 y=AF,(z)+BF,(-z)
If ABZO, the above solution will not be a Floquet
solution. It will be seen later, from the method
for determining v when a and q are given, that
there is some ambiguity in the definition of V ;
namely, .v can be replaced by v+2k, where k is an
arbitrary integer. This is as it should be, since
the addition of the factor exp, (2ikz) in 20.3.2 still
leaves a periodic function of period a for the
coefficient of exp ivz.
It turns out that when a belongs to the set of
characteristic values ar and b, of 20.2, then v is
zero or an integer. It is convenient to associate
v=r with a7(p),and v=-r with b,(q); see [20.36].
In the special case when v is an integer, F,(z) is
727
proportional to F, (- z); the second, independent
solution of 20.3.1 then has the form
0
20.3.6 y2=zce7(z,q)+C
&k+p sin (2k+p)z,
k =O
associated with ee,(z, q)
0
20.3.7 YZ=ZSer(Z, p)+cfZk+p cos (Zk+p)z,
k-0
associated with se,(z, g)
The coefficients &k+p and fZk+p depend on the cor-
responding coefficients A, and B,, respectively,
of 20.2, as well as on a and p. See [20.301, section
(7.50)-(7.51) and i20.581, section V, for details.
If v is not an integer, then the Floquet solutions
F,(z) and F,(-z) are linearly independent. It
is clear that 20.3.2 can be written in the form
m
20.3.8 F,(z)= C czkei(s+2t)z.
k=-m
From 20.3.8 it follows that if v is a proper fraction
rn,/m, then every solution of 20.3.1 is periodic,
and of period at most 2 m z . This agrees with
results already noted in 20.2; i.e.,both independent
solutions are periodic, if one is, provided the period
is different from T and 2 ~ .
Method of Generating the Characteristic Exponent
Define two linearly independent solutions of
20.3.1, for fixed a, p by
y1(0) = 1 ; y:(o)=o.
20.3.9
yz(0) =0 ; &(O) = 1.
Then it can be shown that
20.3.10 cos uv--y,(r)=O
Thus v may be obtained from a knowledge of
yl (T) or from a knowledge of both y; 6) 6)and yz
For numerical purposes 20.3.11 may be more
desirable because of the shorter range of integra-
tion, and hence the lesser accumulation of round-
off errors. Either v, -v, or fv+2k (k an arbi-
trary integer) can be taken as the solution of
20.3.11. Once v has been fixed, the coefficients
of 20.3.8 can be determined, except for an arbitrary
multiplier which is independent of z.
The characteristic exponent can also be com-
puted from a continued fraction, in a manner
analogous to developments in 20.2, if a sufficiently
close first approximation to v is available. For
 Corresponding to q=O, yl=cos Gz, y2=sin &iz;
the Floquet solutior,s are exp(kz) and exp(-&).
As a, q vary continuously in the q--a plane,
v describes curves; v is real when (91 a), qZo
lies in the region between -ar(q) and br+l(q) and
FIOIJRE 20.7. Characteristic Exponent i n First Unstable
Region. Differential equation: y+ (a-29 cos 22)y=O.
The Floquet solution y=e+=P(z), where P(z) is a periodic
function of period Z. I n the first unstable region, v = i p ;
p is given for u 2 -5. (Constructed at NBs.)
 MATHIEU FUNCTIONS 729

20.8
FIQURE

-4

FIQURE
20.9

-
+ &
Charts of the Characteristic Exponent.
(From 8. J. Zaroodny, An elementary review of the Mathieu-Hill equation of real variable based on numerical solu-
tions, Ballistic Research Laboratory Memo. Rept. 878, Aberdeen Proving Ground, Md., 1965. with permission.)
s=e*r= constant; i n unstable regions
- - - - v=constant; i n stable regions
- . - . - Lines of constant values of -9.
730 MATHIEU FCT1ONS

Expansions for Small q ([!20.36] chapter 2)

If v, p are fixed:
20.3.15
a=3+- (5v2+7)P
$ 1)+ 32(~-
2(v2-- l)a(~2-4)

(9v+58v2+29) + . . . (v#1,2,3).
+64(~~-1)~(v~-4) (~~-9)
For the coefficients Q~ of 20.3.8
20.3.16
c / ---
-4(v+l)
- (v2+4v+7)n3
128(~+1)~(~+2)(v--l)
. . . +

(vfl,2)
c,/e,=$/32(v+1>(v+2)+ - * *

cZt/cO= (- i)*gr r ( v + ~ 1+)+

(v+1) /22*~! . . .
20.3.17
,f(V+Z)Z
---
t, (v-2)2 }]+
F,(z)=c,,
[ e*-
{4(v+1) 4(v-1)
...

(v not an integer)

For small values of a

20.3.18
20.10. Chart of the Characteristic Ezponent.
FIGURE a+ a+
(From 8. J. Zaroodny, An elementm review of the Mathieu-Hill equation (
cos v?r= 1--+-+
2 24
...
)
of real variable based on numerical solutions, Ballistic Research Laboratory
Memo. Rept. 878, Aberdeen Proving Oround, Md., 1955, with permisslon) -@
4 [l+a (l-Z)+ . . .]
__ -constant; i n unstable regions
s=eirr-

- - - - v=constant; i n stable regions +@(e-=+ . .)+. . .

_ .- . - Lines of constant values of -q. 96 256

20.4. Other Solutions of MathieusEquation

all solutions of 20.1.1 for real z are therefore
bounded (stable) ; Y is complex in regions between Following Erdafi [20.1419 [20.151, define
b, and a,; in these regions every solution becomes
20.4.1 Vk(z)=[efr COS (z-b)/cos (Z+b)ltkJk(f)
infinite at least once; hence these regions are
termed unstable regions. The characteristic where
curves a,, b, separate the regions of stability.
For negative q, the stable regions are between 2004.2 f=%COS (z--b) COS (z+-b)l!l
h r + 1 and bzr+zt azr and %,+I; the unstable regions are and Jk(j) is the Bessel function of order k; b is a
between u ~ , + ~ b27+l, az, and bz7.
and fixed, arbitrary complex number. By using the
In Some problems so1utions are required for rea1 recurrence refations for Bessel functions the follow-
values of z only. In such cases a knowledge of ing may be verified:
the characteristic exponent v and the periodic
function P(z) is sufficient for the evaluation of 20.4.3
the required functions. For complex values of z, --
SVk
however, the series defining P(z) converges slowly. 2q(cOs 2z)(Pk+P(Vk-2+&+2) +EVk=O*
dzZ
Other solutions will be determined in the next
section; they a11 have the remarkable property It follows that a fO~ma1Solution Of 20-1-1is given
that they depend on the same coefficients c, by
developed in connection with Floquetstheorem
(except for an arbitrary normalization factor).
20.4,4 c
y= *---oo
(D CZsv22n+v
 MATHIEU F'UNCTIONS 731
where the coefficients cznare those associated with where j satisfies 20.4.2. An examination of the
Floquet's solution. In the above, Y may be ratios $zn+,/$2n+r-z shows that
complex. Except for the special case when v is
an integer, the following holds:

will be a solution provided

/cos(z-b)l>l; IC09 (z+b)(>l.
The above two conditions are necessary even
when v is an integer. Once b is fixed, the regions
in which the solutions converge can be readily
established.
Following [20.36] let
C2n -- C-2n -3 (n-.w)
20.4.7
c-Zn+z 4n2
J,(z)= zp (z); Y,(r) =z;) (5);
~2n-2

Tt follows that 20.4.4 converges absolutely and El); (z)=2:) 'l; (z)=zp (z)
(5); E
uniformly in every closed region where
If z is replaced by -iz in 20.4.5 and 20.4.6
/cos(z-b)l>d,>l. solutions of 20.1.2 are obtained. Thus
20.4.8
There are two such disjoint regions: m
yjj)
(2)= czn(- 1) 'Zk;)+(2 & cosh 2)
n=-m
(1) Y(z-b)>dz>O; (Ices (z-b)l>d,>l)
(Icosh Zl>l)
(11) J(z--b)<--dz<O; (/cos(Z--b)l>dI>l) 20.4.9
m
If v is an integer 20.4.4 converges for all values of (2 & sinh z)
yP (z)=ns--0 cznZif,:,
2. Various representations are found by special-
(Isinh z(>l, j = l , 2, 3, 4)
izing b.
The relation between yjj)(z) and yk')(z)can be
20.4.5 determined from the asymptotic properties of the
Bessel functions for large values of argument. It
can be shown that
((cos zl>l, larg 2 & cos zI Sir) 20.4.10
20.4.6
yjj)(z)/yij)(z)= (9z>O) :
[F.(o)/F,~)I~~~~/z

When v is not an integer, the above solutions

do not vanish identically. See 20.6 for integral
values of v.
If b-.wi, y reduces to a multiple of the solution Solutions Involving Products of Bessel Functions
20.3.8. The fact that 20.3.8, 20.4.5, and 20.4.6
are special cases of 20.4.4 explains why it is that 20.4.11
these apparently dissimilar expansions involve
the same set of coefficients ez,.
yP (z) =-
1
5 cZn(
Czs n=--00
-1) "22;.+s (fie") J, - (die- ')
(j=1, 2,394)
Since 20.4.4 results from the recurrence proper-
ties of Bessel functions, Jk(f) can be replaced by satisfies 20.1.1, where Zi5)(u)is defined in 20.4.7,
Hij)(f), j = 1 , 2, where Hij)
is the Hankel function, the coefficients cznbelong to the Floquet solution,
at least formally. Thus let and s is an arbitrary integer, czaf 0. The solution
converges over the entire complex z-plane if q# 0.
Written with z replaced by -iz, one obtains
solutions of 20.1.2.
 732 MATHIEU FUNCTIONS

20.4.12 20.5.4
MXZ,
1
a>=-;- 55
C2s n---
C~,(-l)nZ~r!y+*(~ez>J,-,(~e-? 2&+A:+. . . =A:+&+. . .
. .=@+E+.
=G+Ba,+. . . =l.
It can be verified from 20.4.8 and 20.4.12 that 20.5.5
2r
A2=-lrce28(z,
1 q)dz;A:=IJ ce,(z, p) cosnzdz
20.4.13 2n n . 0

provided cza#O. If cZr=O, the coefficient of l/czs

in 20.4.11 vanishes identically. For details see
120.431, [20.151, [20.36].
If s is chosen so that Icp8(is the largest coefficient
of the set Ic2,1, then rapid convergence of 20.4.12
is obtained, when 9z>O. Even then one must
be on guard against the possible loss of significant
figures in the process of summing the series,
especially so when q is large, and lzl small. (If
jz1, then the phase of the logarithmic terms
occurring in 20.4.12 must be defined, to make the
functions single-valued.)
20.5. Properties of Orthogonality and
Normalization
If a(v+2p, q), a(v+2s, q) are simple roots of
20.3.10 then
20.5.1 s,Fr+zp(~)F,+2r(-~)d~=0,
if pZ8.
Define
1
20.5.2 ce,(z, p)=s [F,(z)+F,(-z)];
1 are always possible. This normalization has in
se,(z, q)=-i [F,(z)-F,(-z)]
ce,(z, q), se,(z, q) are thus even and odd functions
of z, respectively, for all Y (when not identically
zero).
If Y is an integer, then ce,(z, q), se,(z, q) are
either Floquet solutions or identically zero.
The solutions ce,(z, q) are associated with a,;
se,(z, q) are associated with .b,; r an integer.
Normalization for Integral Values of and Real q
Y
For integral values of Y the summation in
20.3.8 reduces to the simpler forma 20.2.3-20.2.4;
on account of 20.5.3, the coefficients Am and Bm
(for all orders r) have the property
For integral values of Y, the functions ce,(z, q)
and se,(z, q) form a complete orthogonal set for
the interval 0 5 2 5 2 ~ . Each of the four systems
C ~ ~ ~ ( Zce2,+,(z),
), sezr(z),se2,+,(2)is complete in the
smaller interval 0 z<_ <&r, and each of the systems
ce,(z), se,(z) is complete in 0 5 z l n . .
If p is not real, there exist multiple roots of
20.3.10; for such special values of a(p),the integrals
in 20.5.3 vanish, and the normalization is therefore
impossible. In applications, the particular nor-
malization adopted is of little importance, except
possibly for obtaining quantitative relations be-
tween solutions of various types. For this reason
the normalization of Fy(z),for arbitrary complex
values of a, q, will not be specified here. It is
worth noting, however, that solutions
ace&, d, @er(z, a)
defined so that
fact been used in [20.59], and also in [20.58],
where the most extensive tabular material is
available. The tabulated entries in [20.58] sup-
ply the conversion factors A=l/a, B=l/p, along
with the coefficients. Thus conversion from one
normalization to another is rather easy.
In a similar vein, no general normalization will
be imposed on the functions defined in 20.4.8.
20.6. Solutions of MathieusModified Equation
20.1.2 for Integral v (Radial Solutions)
Solutions of the first kind
20.6.1
Cew+p(z,a>=Cezr+p(iZ,
-
=GoA:;:; (a) c-h (2k+p)z
associated with a,
 MATHIEU FUNCTIONS 733
0

20.6.2 q)=-isezr+p(iz, $ =EE;:g(q)sinh (2k+p)z, associated with b,

Se2r+p(z,
k-0

writing A:;$g(q)=Azk+, for brevity; similarly for Bzk+,,;

p=O, 1,

20.6.4

20.6.5

See [20.30] for still other forms.

Solutions of the second kind, as well as solutions of the third and fourth kind (analogous to Hankel
functions) are obtainable from 20.4.12.
Q

20.6.7 (2, n) =E(-1) +*A:;

M& k-0
(d[Jk- ,(%)z!!:, +Jk+# (%)zkl#(%)]/e,&:
(u.2)
where eO=2,e,=l, for s=1, 2, . . .; s arbitrary, associated with azr
0

20.6.8 2k + 1 (a)1Jt-I
Mc&~(z,p ) = E (-l)r+LA2r+1 (a) 1 (%)+JL+#+l
Zi!r+ (u1)zit*(%)
k-0
associated with ( ~ l ~ + ~

(-l)k+re;:
0

20*6-10 1(z, n) =E
M&+ :(q)[Jk-#(%)zii# + I(%)-Jl+#+l (%)zi% ]/E:$!
k-0
associated with b2r+l
where
ul=l/Tie-,%=fie, Fd$;, A2S;#O,p=O, 1.
See 20.4.7 for definition of Zg)(x).
Solutions 20.6.7-20.6.10 converge for all values of z, when q#O. If j=2, 3, 4 the logarithmic terms
entering into the Bessel functions Y,,,(uJ must be defined, to make the functions single-valued. This
can be accomplished as follows:
Define (as in [20.58])
20.6.11 In (fie) =In (d?) +z
See [20.15] and [20.36], section 2.75 for derivation.
734 MATHIEU FUNMONS
Other Expressions for the Radial Functions (Valid Over More Limited Regiona)
20.6.12 Mc!!)(z, n)=[ce&, q)I-g
(-l)&E(q)Zi)(24j cash
k-0
2)
OD

+ p) =[cezr +I (0, d I- kC
M c ~:(z, -0
(-- 1)+& Ti (dZii)+i
(24i cash z>
OD

20.6.13 Msi!)(z, q)=[sdr(O,a>]-tanh z C (-1)k++2kEX(q)Z!{)(2fi cosh z)

b-1
OD

Ms!!)+,(z,p)=[se;,+,(O, a)]-tanh z k-0

C (-1)k+r(2k+l)BP,I~:(p)2#+1(24jcosh z)
Valid for 9?z>O, lcosh zl>i; if j = 1 , valid for all z. They agree with 20.6.7-20.6.10 if the Bessel
functions Y,,,(2pt cosh z) are made single-valued in a suitable way. For example, let
2
Ym(u)=; On u>Jnr(4 +w
where 4(u) is single-valued for all finite values of u. With u=2$ cosh z, define
-% r Sarg f(l+e-*)S$ ?r
20.6.14 In (2pt cosh z)=ln 2pt+z+ln i(l+e-)
(If p is not positive, the phase of In 2pmust also be specified, although this specification will not
affect continuity with respect to z. If Ym(u) is defined from some other expression, the definition
must be compatible with 20.6.14.)
&.L
I. % I gMs:2cr,s,

20.13. Radial Mdhieu Function of the Second Kind.

FIGURE
(From J. C. Wfltse and M. J. King, Vduw of the Mathieu functions, The
L -__- . Johns Hop- Univ. Radiatiotl LeboratOW Tech. Rept. AF-63. 1968,
FIGURE20.11. Radial hlathieu Function of the First Kind. with permission)
(From J. C. Wiltse and M. J. King, Values of the Mathieu functions, The
Johns Hopklns Univ. Radhtion Laboratory Tech. Rept. AF-63, 1968,
with permission)

FIQURE 20.12. Derivative 0 thc Radial Mathieu Function F~~~~~ 20.14. ~ a d i * Ma&,, ~ ~ of he
- Third
i Kind.
~
of the d irst Kind.
(From J. C. Wfltse and M. J. King, Derivatives. zeros, and other data per (From J. C. Wiltse snd M. J. ging, Valuer of the Mathieu h&im,The
W i n g to Mathleu functions, The Johns Hopkina Univ. Radiation Lab- Johns Hopldns Unlv. RadlatiOn Lebaatory Tech. RePt. AF-58, 1968,
oratory Tech. Rept. AF-67.1968, with permlssl~n) with permbsioa)
 MATHIEU FUNCTIONS 735
If j= 1, Mcj'!,, and MS$;)+~,p=O, 1 are solutions 20.6.19
of the first kind, proportional to CeZr+pand Se2r+p. MCP' (- 2, q) = -Me?' (2,q)-2fe. ,MCy (2, q)
respectively.
Thus 20.6.20
20.6.15 MsP' (-2, q) =MsS2'(2, q)-2f0. ,Msy (2, a)
cezr Gpp) cear(O, q> where
cezr(z, q)= (- l)r&r
Mca!')(z,q)

ceir+l G i q ) cezr+l(O,
ceh+l(z,a)= (- i)r+l&A;r+I MCP+l(Z,a)
R See [20.58].
sei, (0, q)seir (9 !I) (z,q)
M8~:) In particular the above equations can be used to
Se*(z, d= (-l)rqBp,' extend solutions of 20.6.12-20.6.13 when .%<O.
For although the latter converge for Wz<O,
seir+l(O,q) a e z r + l G , p) provided only Jcosh zl>l, they do not represent
Se2r+l(z, d= (- 1)r f i B r + 1 M88+1 (z the same functions as 20.6.9-20.6.10.
20.7. Representations by Integrals and Some
The Mathieu-Hankel functions are Integral Equations
20.6.16 Let
M$8'(z,q)=M$1'(z,q)++iM:*'(Z, q)
Mp(z, q)=M?)(z,q)-iM$*)(z,q)
20.7.1
f
G(u)= K(u,t)V(t)dt
c
M?=Mc? or Ms!". be defined for u in a domain U and let the contour
From 20.6.7-20.6.11 and the known properties C belong to the region T of the complex t-plane,
of Bessel functions one obtains with t=yo as the starting point of the contour
and t=yl as its end-point. The kernel K(u, t)
20.6.17 and the function V(t) satisfy 20.7.3 and the
Mi?+,(z+inr, q) hypotheses in 20.7.2.
= (
- -
)l
W
"
?+
i
, (
,
z dl
!21+2n+iM4:)+P(z, 20.7.2 K(u, t) and its first two partial derivatives
with respect to u and t are continuous for 1, on C
W)+,(z+inr, q)
dv
=(-l)nwi:'+P(z, a)- 2nMi?+,(z, q)l and u in U; V and - are continuou. in t.
dt
Mi?+p(z+inr,q) 20.7.3
n)+2nM::)+,(z,all
= (--l)"P[W)+,(z,
where M=Mc or M s throughout any of the above
equations.
pg dt Yo
dV
V - g K] 'l=O; -+(a-Zq
dt2 cos 2t)V=0.

otherh~pertiesof Characteristic Functiorm, q R e d If K satisfies

(Ac~socintedWith a, and b,)
Consider
20.6.18
X,=-Mc!*)(z,q)+McS"(-z, q ) ; then Q(u) is a solution of Mathieu's modified
x.=Msp'(z,q)-MsS')(-z, q) equation 20.1.2.
If K(u, t) satisfies
Since XIis an even solution it must be proportional
to Mc!')(z, q); for 20.1.2 admits of only one even
solution (aside from an arbitrary constant factor).
Similarly, X2is proportional to MsS')(z, q). The
proportionalityfactors can be found by considering then Q(u) is a solution of Mathieu's equation
values of the functions at z=O. Define, therefore, 20.1.1, with u replaciig o.
736 MATRIEU FUNCTIONS

Kernels &(z, t) and Kz(s, t) where Fv(t)is the Floquet solution 20.3.8. The
20.7.6 t)]-v'2J
K1(z,t)=z;')(u)[M(~, (gz>O) path C is chosen so that G(z, t, a) exists, and
20.7.2, 20.7.3 are satisfied. Then it may be
where verified that K3(z,t, a), considered as a function
of z and t, satisfies 20.7.4; also, considered as a
20.7.7 U= J2q(cosh ~ Z + C O S 2t)
function of a and t, K3 satisfies 20.7.5. Conse-
20.7.8 M(z, t)=cosh (z+d)/cosh (2-3) quently G(z, q, a)=Y(z, q)y(a, q), where Y and y
satisfy 20.1.2 and 20.1.1, respectively.
To make M-4, single-valued, define Choice of Path C.Three paths will be defined:
20.7.9
20.7.16
cosh (z+ir)=e" cosh z
Path C3:from -dl+ia to d2-ia , d,, d2 real
cosh (z-i?~)=eWk cosh z
M(z,O)=l -dl<arg [&{cosh (z+ia) f l } l < ~ - d ~
[M(z,~ ) ] - f ~ = e - ~ ~ M0)( z , -d2<arg [&{cosh (z-k) &l}l<~-d2
Let
20.7.17
20.7.10
Path C,: from d2-ia to 2?r+ia-d,
where F,(t) is defined in 20.3.8. It may be verified
(same dl,d2 as in 20.7.16)
that KIFvsatisfies 20.7.3, K satisfies 20.7.2 and
20.7.4. Hence is a solution of 20.1.2 (with. z 20.7.18
replacing u). It can be shown that Kl may be re-
placed by the more general function
20.7.11
&(z, t)=Z'?)+2s(~)[M(z,t)l-trt.,
s any integer.
where M?(z,q) is also given by 20.4.12.
See 20.4.7 for definition of Z(?+z8(u).
20.7.19 Path C,: from -d,+ioJ to 2r-d1+ia
From the known expansions for Z(?'+,,(u) when
W z is large and positive it may be verified that
20.7.12
My (2,q)=
See 120.361, section 2.68.
If is an integer the paths can be simplified;
Y

for in that case F,(t) is periodic and the integrals

exist when the path is taken from 0 to 2u. Still
(9Z>0,9j?Y+*)>O)
further simplifications are possible, if z is also real.
where M? (z, q) is given by 20.4.12, s=O, 1, . . ., The following are among the more important
cz,#O, and Fv(t)is the Floquet solution, 20.3.8. integral representations for the periodic functions
Kernel K&. t, a) cer(z, q), se,(z, q) and for the associated radial
solutions.
20.7.13 Ka(z,t, a)=eaAw
where Let r=2s+p, p=O or 1

20.7.14 W=cOsh z COS u COS t+sinh 2 Sin Q sin t 20.7.20

20.7.15 G(z, q, a)=$ $ e2'Gw;F,(t)dt

 MATHIEU FUNCTIONS 737
20.7.21 ce,(z, q ) = c r r 2cash (2fi sin z sin t)[(l-p)+p COS z COS t]cer(t, q)dt

20.7.22 se,(z, q ) = p , r 2 sin (2& cos z cos t+p sin z sin t se,(t, q)dt

20.7.23 se,(z, q ) = u r r sinh (2fi sin z sin t)[(l-p) cos z cos t+p]se,(t, q)dt
where
2
2007.24 pr=-ceza
" 6, q) /&'(q); /dijAY+'(q)if p=l, for functions ce,(z, q)

2 4
u,=- ce2,(0, q)/&(q) if p=O; ur=- ce2,+l(0,q)/A:'+'(q),if p = l ; associated with functions ce,(z, q)
n- "
4 2
ur=; sela(0,q)/fi@(q), if p=O; ur=; seL,+l(O,q)/fiB;l'+'(q), if p=l; associated with ser(z, q)
Integrals Involving Bessel Function Kernels
Let
20.7.25 u=d2q(cosh 2z+cos 2t), (B'cosh 22>1; if j=1, valid also when z=O)
20.7.26
w

20.7.27 M&)(Z,$=
(-1)'+'8q sinh 22 f Zjj)(u) sin 2t Selr(t, p)dt
rP27 U2

Msj!)+l(z,a)=
(-1),8fi sinh z f Z:j)(u) sin t seg,+l(t,q)dt
U@r+l U

In the above the j-convention of 20.4.7 applies and the functions Mc, M s are defined in 20.5.1-
20.5.4. (These solutions are normalized so that they approach the corresponding Bessel-Hankel
functions as B'z+ Q).)
Other Integrals for M&)(z, q) and M#)(a, q)

20.7.28

20.7.29 Mc!"(z, q)=Tr

1;[(1-p)+p cosh z COS t] COS (2& sinh z sin t)ce,(t, q)dt

20.7.30

20.7.31

20.7.32

20.7.33
 Integrals With Infinite Limits

r=2s+p
In 20.7.38-20.7.41 below, z and p are positive.

20.7.38 Mcj)(z,q)=rrLm
sin (2 fi cosh z cosh t+p
/T&, if p=o -yr=2ce;,+l (;, q) /& TA?+~, if p=l

20.7.39 q)=rr
MsJ)(z, sinh z sinh t 2& cosh z cosh t-p i)] Msj(t,p)dt
yr=-4se;, (;, p) /fi?r~s, if p=o yr=-4se2s+l (;, p) /lr~;ll+l,if p = l
20.7.40 MC~(Z,
q)=.rlm COS (2fi cash 2 cash t-p 2> Mcj(t, n)dt
yr= -2cezs(3x, p)/lrAa,,
if p=O yr=2ce;8+1(3r,~)/T&A?++,
if p = l

20.7.41 Msj2(zlp)=rr sin (2& cosh z cosh t+p sinh z sinh t Msil)(t,q)dt

77=-4se;,(3~, a)/& TBa2,if p=O

Additional forms in [20.30], [20.36], [20.15]. 20.8.2
20.8. Other Properties a(-v, q)=a(v, -q)=a(v, q), Y not an integer

Relations Between Soltitions for Parameters q and -q c;m(-a)=p(-l)mC~,(p), Y not an integer

Replacing z by 4,-z in 20.1.1 one obtains (cZmdefined in 20.3.8) and p depending on the
20.8.1 y+ (a+29 cos 2z)y=O normalization ;
Hence if u(z) is a solution of 20.1.1 then u(3~-z)
satisfies 20.8.1. It can be shown that FV(z, -q)=pe-iY*~2Fv
 MATHIEU FUNCTIONS 739
0

Kozr+t(z,q>=C
k=O
[It-s (ut)Kt+s+l
(UZ)

+Ik+s+l(~dKt--s(uz)l/Azs+t
where I,,,(x), K,(z) are the modified Bessel func-
tions, ut, uz are defined below 20.6.10. Super-
scripts are omitted, es=2,if s=O, e,=1 if sZ0.
Then for functions of first kind:
20.8.10
Mcif'(2,-a> = (-1IrIezr(2, 9)
Msif)(2, -a> = (- 1 )rlozr (2, q>
M&)+~(Z, -a> = (- l)riIezr+t(Z, a)
1(z, -a> = -1 )'iIoZr +t (2, q>
Msij)+
For the Mathieu-Hankel function of first kind:
20.8.11
2
Mc:? (z, -q) = (- I)r+li- Kezr(z, q)
A

2
M&) (Z, -a) - Kozr(Z,
= (- l)'+'i
A
a)
2
M&)+1 (Z, -a) = (- - Kezr+1( Z,
A
a)
2
M&)+1 (Z, -q) = ( - - Kozr+t(z,
A
a>
For M?)(z, -q), j=2, 4, one may use the defini-
tions
My)=-i(MS3)-MMI')); Mr=Mcror Ms,
also
M:)(z, -q)=2M:)(z, -q)-M$3)(~, -q)

M=Mc or Ms; for real z, q, MI"(z, -q)

are in general complex if j= 2,4.
Zeros of the FunctionsXor Real Values of q.
See [20.36], section 2.8 for further results.
Zeros of ce45, q) and se,(5, q), Mcl"(5, q), Ms!"(5, q ) .
In OIz<,~,ce,(z, q) and ser(z, q) have T real *
zeros.
There are complex zeros if q>O.
o any zero of cer(z, q), ser(z, q) in
If ~ = s o + i y is

are also zeros, k an integer.

'See page n.
740 MATHIEU FUNCTIONB

7r 17
In the strip --<z,<-t 2 2 the imaginary zeros of
ce,(z, q), se,(z, q) are the real zeros of Ce,(z, q),
Se,(z, q), hence also the real zeros of Mcp)(z, q)
and Ms:)(z, q), respectively.
For small q, the large zeros of Cer(z,q), Se,(z, q)
approach the zeros of Jr(2&cosh 2).
Tabulation of Zeros

Ince [20.56] tabulates the first non-trivial

2 >
zero (i.e. different from 0, IJ?r or ce,(z), se,(z),
r=2(1)5 and for se6(z)to within OlO-, for q=O(l)
10(2)40. He also gives the turning points
(zeros of the derivative) and also expansions for The functionsf,,,, go,r,jc.r,
gC.,are tabulated in
them for small q. Wiltse and King [20.61,2] tabulate [20.58] for q225.
the &st two (non-trivial) zeros of A4cs(z, q) and
Msy)(z, q) and of their derivatives r=O, 1, 2 for 20.9. Asymptotic Representations
6 or 7 values of p between .25 and 10. The graphs
reproduced here indicate their location. The representations given below are applicable
Between two real zeros of Mcp)(z, q), Msj(z,q) to the characteristic solutions, for real values of q,
there is a zero of Mc;~)(z, q), Msy(z,p), respec- unless otherwise noted. The Floquet exponent v
tively. No tabulation of such zeros exists yet. is defined below, as in [20.36] to be BS follows:
Available tables are described in the References. In solutions associated with a,: V=T
The most comprehensive tabulation of the char- In solutions associated with b,: v=-r.
acteristic values a,, b, (in a somewhat different
notation) and of the coefficients proportional to For the functions defined in 20.6.7-20.6.10:
A, and Bmas defined in 20.5.4 and 20.5.5 can be
found in [20.58]. In addition, the table contains 20.9.1
certain important joining factors, with the aid Mc?(zJq)
of which it is possible to obtain values of (-l),MSp(z, q)
Mc$j)(z,q) and Ms;j)(z, q).as well as their deriva-
tives, at s=O. Values of the functions ce,(z, q)
and se,(z, q) for orders up to five or six can be
found in [20.56]. Tabulations of less extensive
character, but important in some aspects, are out- where D-l=D-z=O; Do=l, and the coefficients
lined in the other references cited. In this Dmare obtainable from the following recurrence
chapter only representative values of the various formula :
functions are given, along with several graphs. 20.9.2
Special Values for Arguments 0 and
2
20.8.12
[16q(l-u2)-8i&j mlD,-l
+4q (2m-3) (2m--l),(l -U2)Dm--2=0
20.9.3
McP(z, a)
(-l)~Msy(z, q)
-f 2 qcoshr-lrr-~r]
N e [ J !a 4 dm
?r+q(coshz-a)+ $0 [4i&j(cosh z-a)]
d-1=d-2=O; do=l, and
20.9.4

In the above
-2r<arg
+4q(2m-3)(2m-1)(1

fi cosh z<r
lcosh z-cJ>lcfll, Wz>O,
but u is otherwise arbitrary. If u2=1, 20.9.2 and
20.9.4 become three-term recurrence relations.
Formulas 20.9.1 and 20.9.3 are valid for arbi-
trary a, p, provided Y is also known; they give
multiples of 20.4.12, normalized so as to approach
the corresponding Hankel functions HL1)(&ez),
HF)(Ge'), as z--.

ratio of two successive terms in the expansion is

essentially

If a, p, z, Y are real, the real and imaginary

components of Mc13)(z,q) are Mcl"(z, p) and
Mc?)(z, a), respectively; similarly for the com-
ponents of Ms13)(z,a). If the parameters are
complex
20.9.5 Mc;')(z, p)=4 [Mci3)(z,p)+Mcf)(z, q)]

20.9.6 Mc$"(z,p)=-2
a

sponding relations among Ms$"(z,q ) .

MATHIEU FUNCTIONS

-U~)G?,_~=O.

See [20.36], section 2.63.

The formula is especially useful if lcosh zlis large
and p is not too large ; thus if u= -1, the absolute

p)-Mcj"(~,a)]
[Mc$~)(z,

Replacing c by s in the above will yield corre-

Formulas in vhich the parameter a does not

enter explicitly :
Goldstein's Expamiom
20.9.9

20.9.10

20.9.11

cer(5, a)

20.9.12
ser+1(2,
1
'2048p

+16384$12
-(2wb+124d+1122w)
COS^^ z

+
&%
[
[

{
w4+8SzoZ+105-w4+22d+57
cosh' z
-(wb+14d+33w)

+ (we+505w4+12139w2+10395)

-ce,(O,
FO(O)
12 cosh' z

n) -se:+I(O,
p)2'-*
cosh2z
3wb+290d+1627w

(d-47 ~ ' 4667d-I-2835)

-
12 cosh2 z

-}I+.
Goshs z

5W4+34w2+9

See [20.18] for details and an added term in

q-b12;a correction to the latter is noted in [20.58].
The expansions 20.9.7 are especially useful when
p is large and z is bounded away from zero. The
order of magnitude of Mc:(O, p) cannot be obtained
from the expansion. The expansion can also be
used, with some success, for z=ix, when p is large,
if /coszl>>O;
real, one obtains
cosh2 z

they fail at x=A2 r. Thus, if p, x are

W P O ( 4 -P1(2)1

+W2[P0(4+P1(411

!?>.l+l W 1 [ ~ 0 ( ~ > - ~ 1 ( 4 1

-W2[Po(4 +PI
741

I+...
1

. .

(41 1
the above, Po(z) and P1(x) are obtainable
20.9.7
MC:3) (2, p) - ~ M.5":s (Z, q)
= [Fo(z)-iFl(z)]ei+/lr*q*(cosh z)*
I n

from Fo(z), Fl(x) in 20.9.9-20.9.10 by replacing

cosh z with cos x and sinh z with sin x. Thus
Po(z)=F,(ix); P1(z)=--iF1(iZ):
where
20.9.13
20.9.8
r
1 Wl=e 2fisin [cos ($X+f?r)]2'+1/(cos Z)'+1
+=24sinh z-- (2r+ 1) arctan sinh z,
2
Bz>O, q>>l, w=2r+l
742 MATMEU F"CTION8

20.9.14 20.9.20

8,-@'q%/(r % [1-w
2r+ 1

See 20.9.23-20.9.24 for expressions relating to

ce,(O, q) and se:(O, q ) . When lcos zl>-/qt,
+r44-29- 1219- 122~-84
2048q 12
+-16384qS' . . .I-"
20.9.11-20.9.12 are useful. The approximations j2=22+5r4- 4169-6299- 1162r-476
become poorer as T increases.
Expansions in Terms of Parabolic Cylinder Functions
It should be noted that 20.9.15 is also valid as an
approximation for se,+l(z,q), but 20.9.16 may give
(Good for angles close to $x, for large values of slightly better results. See [20.4.]
p, especially when lcos sl<2t/@.) Due to Sips
[20.44-20.46 1. Explicit Expansions for Ordem 0, 1, to Terms in q-"'
(q
2009.15 Ce,(z, a)- C,[&(a) +%(a)] 20.9.21 For r=O:
20.9.16
se,+l(z,q)-S,[&(a)-&(a)] sin z, a=2$ cos z.
2 dk
Ifit D,=Dk(a)=(-l)keta - e-+a2.
cw
20.9.17 21+
- -16&
-02 1
16q (--+-)9Dz
16
D
64e

20.9.22 For r = l :

+ 315
-A81
D,-* I+. ..

20.9.18

+-k+
1 D,+( (9--25~-36)
16!? 64 D?+2

+
r(r-1) (-9-27~+10)
64 D,-2-- ()
45 r
4 6
D,-n+ . . .]
Formulaa Involving ce,(O,,b) and se,(O, q)
20.9.19
20.9.23

r4+29+263?+262r+ 108 jl . . .I-"

+ 2048q +is@+
fi=67.6+15r4+12809+19Q59+1778r+572
MATHIEU FUNCTIONS 743

For higher orders, these ratios are increasingly

more difficult to obtain. One method of estimat-
ing values at the origin is to evaluate both 20.9.11
and 20.9.15 for some z where both expansions are
satisfactory, and so to use 20.9.11 as a means to
solve for ce,(O, q); similarly for sei@, q).
Other asymptotic expansions, valid over various
regions of the complex z-plane, for real values of
(I, p, have been given by Langer [20.25]. It is not
always easy, however, to determine the linear
combinations of Langer's solutions which coincide
with those defined here.
744 MATHIEU FUNCTIONS
3 2 r( N d
t t d
!?
8,
V 3 3 i3 d
..
:
L
-
0
N'
N
'
V N'
Y v
4 ."
u d
N'
Y
* * J
ss -
0
-
i
Y
i N ' w-
Y -.
P
Y Y
" f r( sj
 MATHIEU FUNCTIONS
References
Texts
[20.1] W. G. Bickley, The tabulation of Mathieu func-
tions, Math. Tables Aids Comp. 1, 40CF419
(1945).
(20.21 W. G. Bickley and N. W. McLachlan, Mathieu
functions of integral order and their tabulation,
Math. Tables Aids Comp. 2, 1-11 (1946).
(20.31 G. Blanch, On the computation of Mathieu func-
tions, J. Math. Phys. 25, 1-20 (1946).
[20.4] G. Blanch, The asymptotic expansions for the odd
periodic Mathieu functions, Trans. Amer. Math.
SOC.97, 2, 357-366 (1960).
[20.5] C. J. Bouwkamp, A note on Mathieu functions,
Kon. Nederl. Akad. Wetensch. Proc. 51, 891-893
(1948).
(20.61 C. J. Bouwkamp, On spheroidal wave functions of
order zero. J. Math. Phys. 26, 79-92 (1947).
[20.7] M. R. Campbell, Sur les solutions de p6riode 2 ST
de ldquationde Mathieu associde, C.R. Acad.
Sci., Paris, 223, 123-125 (1946).
[20.8] M. R. Campbell, Sur une catdgorie remarquable de
solutions de ldquation de Mathieu associde,
C.R. Acad. Sci., Paris, 226, 2114-2116 (1948).
[20.9] T. M. Cherry, Uniform asymptotic formulae for
functions with transition points, Trans. Amer.
Math. SOC.68, 224-257 (1950).
[20.10] S. C. Dhar, Mathieu functions (Calcutta Univ.
Press, Calcutta, India, 1928).
(20.111 J. Dougall, The solution of Mathieusdifferential
equation, Proc. Edinburgh Math. SOC.34, 176-
196 (1916).
[20.12] J. Dougall, On the solutions of Mathieusdifferen-
tial equation, and their asymptotic expansions,
Proc. Edinburgh Math. SOC.41, 26-48 (1923).
[20.13] A. Erddlyi, Uber die Integration der Mathieuschen
Differentialgleichung durch Laplacesche Inte-
grale, Math. Z. 41, 653-664 (1936).
[20.14] A. Erddlyi, On certain expansions of the solutions
of Mathieusdifferential equation, Proc. Cam-
bridge Philos. SOC.38, 28-33 (1942).
[20.15] A. Erddlyi et al.,Higher transcendental functions,
vol. 3 (McGraw-Hill Book Co., Inc., New York,
N.Y., 1955).
120.161 G. Floquet, Sur les dquations diffdrentielles
lindaires A coefficients pdriodiques, Ann. Ecole
Norm. Sup. 12, 47 (1883).
[20.17] S. Goldstein, The second solution of Mathieus
differential equation, Proc. Cambridge Philos.
SOC.24, 223-230 (1928).
(20.181 S. Goldstein, Mathieu functions, Trans. Cambridge
Philos. SOC. 23, 303-336 (1927).
(20.191 G. W. Hill, On the path of motion of the lunar
perigee, Acta Math. 8, 1 (1886).
[20.20] E. Hille, On the zeros of the Mathieu functions,
Proc. London Math. SOC.23, 185-237 (1924).
[20.21] E. L. Ince, A proof of the impossibility of the co-
existence of two Mathieu functions, Proc.
Cambridge Philos. SOC.21, 117-120 (1922).
745
120.221 E. L. Ince, Ordinary differential equations (Long-
mans, Green & Co., 1927, reprinted by Dover
Publications, Inc., New York, N.Y., 1944).
[20.23] H. Jeffreys, On the modified Mathieusequation,
Proc. London Math. Soc. 23, 449-454 (1924).
[20.24] V. D. Kupradze, Fundamental problems in the
mathematical theory of diffraction (1935).
Translated from the Russian by Curtis D.
Benster, NBS Report 200 (Oct. 1952).
[20.25] R. E. Langer, The solutions of the Mathieu equa-
tion with a complex variable and at least one
parameter large, Trans. Amer. Math. SOC.36,
637-695 (1934).
[20.26] S. Lubkin and J. J. Stoker, Stability of columns and
strings under periodically varying forces, Quart.
Appl. Math. 1, 215-236 (1943).
(20.271 E. Mathieu, Mdmoire sur le mouvement vibratoire
dunemembrane de forme elliptique, J. Math.
Pures Appl. 13, 137-203 (1868).
(20.281 N. W. McLachlan, Mathieu functions and their
classification, J. Math. Phys. 25,209-240 (1946).
[20.29] N. W. McLachlanJMathieu functions of fractional
order, J. Math. Phys. 26, 29-41 (1947).
[20.30] N. W. McLachlan, Theory and application of
Mathieu functions (Clarendon Press, Oxford,
England, 1947).
[20.31] N. W. McLachlan, Application of Mathieus
equation to stability of non-linear oscillator,
Math. Gal. 35, 105-107 (1951).
[20.32] J. Meixner, Uber das ssymptotische Verhalten von
Funktionen, die durch Reihen nach Zylinder-
funktionen dargestellt werden konnen, Math.
Nachr. 3, 9-13, Reihenentwicklungen von
Produkten zweier Mathieuschen Funktionen
nach Produkten von Zylinder und Exponential-
funktionen, 14-19 (1949).
(20.331 J. Meixner, Integralbeziehungen zwischen
Mathieuschen Funktionen, Math. Nachr. 5,
371-378 (1951).
[20.34] J. Meixner, Reihenentwicklungen vom Siegerschen
Typus fur die Spharoid Funktionen, Arch. Math.
Oberwolfach 1. 432-440 (1949).
(20.351 J. Meixner, Asymptotische Entwicklung der
Eigenwerte und Eigenfunktionen der Differen-
tialgleichungen der Sphilroidfunktionen und der
Mathieuschen Funktionen, Z. Angew. Math.
Mech. 28, 304-310 (1948).
(20.361 J. Meixner and F. W. Schafke, Mathieusche
Funktionen und Sphiiroidfunktionen (Springer-
Verlag, Berlin, Germany, 1954).
(20.371 P. M. Morse and P. J. Rubinstein, The diffraction
of waves by ribbons and by slits, Phys. Rev. 54,
895-898 (1938).
i20.381 H. P. Mulholland and S. Goldstein, The charscter-
istic numbers of the Mathieu equation with
purely imaginary parameters, Phil. Mag. 8,
834-840 (1929).
(20.391 L. Onsager, Solutions of the Mathieu equation of
period 4~ and certain related functions (Yale
Univ. Dissertation, New Haven, Conn., 1935).
 746 MATHIEU FUNCTIONS
[20.401 F. W. Schiifke, Uber die Stabilitatskarte der.
Mathieuschen Differentialgleichung, Math.
Nachr. 4. 175-183 (1950).
[%I411 F. W. Schiifke, Das Additions theorem der
Mathieuschen Funktionen, Math. Z. 58,436-447
(1953).
[20.421 F. W. Schiicke, Eine Methode aur Berechnung des
charakteristischen Exponenten einer Hillschen
Differentialgleichung, Z. Angew. Math. Mech.
33. 279-280 (1953).
[20.43] B. Sieger, Die Beugung einer ebenen elektrischen
Welle an einem Schirm von elliptischem
Querschnitt, Ann. Physik. 4, 27, 626-664 (1908).
120.441 R. Sips, Reprhsentation asymptotique des fonctions
de Mathieu et des fonctions dondesphhroidales,
Trans. Amer. Math. SOC.66, 93-134 (1949).
120.451 R. Sips, Reprbentation asymptotique des fonctions
de Mathieu et des fonctions sph6roidales 11,
Trans. Amer. Math. SOC.90, 2, 340-368 (1959).
(20.461 R. Sips, Recherches sur les fonctions de Mathieu,
Bull. Soc. Roy. Sci. Liege 22, 341-355, 374-387,
444-455, 530-540 (1953); 23, 37-47, 90-103
(1954).
120.471 M. J. 0. Strutt, Die Hillsche Differentialgleichung
im komplexen Gebiet, Nieuw. Arch. Wisk. 18
31-55 (1935).
(20.481 M. J. 0. Strutt, Lam&che, Mathieusche und
verwandte Funktionen in Physik und Technik,
Ergeb. Math. Grenageb. 1, 199-323 (1932).
120.491 M. J. 0. Strutt, On Hillsproblems with complex
parameters and a real periodic function, Proc.
Roy. SOC.Edinburgh Sect. A 62, 278-296 (1948).
[20.50] E. T. Whittaker, On functiorw associated with
elliptic cylinders in harmonic analysis, Proc.
Intl. Congr. Math. Cambr. 1, 366 (1912).
(20.511 E. T. Whittaker, On the general solution o f
Mathieusequation, Proc. Edinburgh Math. SOC.
32, 75-80 (1914).
(20.521 E. T. Whittaker and G. N. Watson, A course of
modern analysis, 4th ed. (Cambridge Univ. Press,
Cambridge, England, 1952).
Tables
[20.53] G. Blanch and I. Rhodes, Table of characteristic
values of Mathieusequation for large values of
the parameter, J. Washington Aced. Sci. 45, 6,
166-196 (1955). Be&) =a,(q) +2q-2(27+ 1) f i
Bo&) = b,(q) + 2q-2(2r- 1) at= 1/2fi r =0(1)
15, O l t 5 . 1 , with PI a*; 8D (about); inter-
polable.
[20.54] J. G. Brainerd, H. J. Gray, and R. Merwin, Solu
tion of the Mathieu equation, Am. Inst. Elec.
Engrs. 67 (1948). Characteristic exponent over
a wide range. p, M foi c=1(1)10; k=.l(.l)l,
5D; g(t), h(t) for t=0(.1)3.1, z, 5D; e=l(l)lO,
k=.l(.l)l, where g(t), h(t) are solutions of
y+c(l+k COS t)y=O, with g(O)=h(O)=l,
g(O)=h(O)=O, cos 2*p=2g(*)h(r)-11 M=
[- 8(*) !?(TI/h(4h ( 4111.
[20.55] J. G. Brainerd and C. N. Weygandt, Solutions of
Mathieus equation, Phil. Mag. 30, 45-77
(1940).
(20.561 E. L. Ince, Tables of the elliptic cylinder functions,
Proc. Roy. SOC.Edinburgh 52, 355-423; Zeros
and turning points, 52,424-433 (1932). Charac-
teristic values aa, al, . . ., as, bl, bz, . . ., bs,
and coefficients for O=O(l) 10(2)20(4)40; 7D.
Also ce,(z, e), se,(z, e), e=o(l)lo, ~ = 0 ~ ( 1 ~ ) 9 0 ~ ;
5D, corresponding to characteristic values in
the tables. a,=be,-Zq; b,=bo,-2q; @=q.
[20.571 E. T. Kirkpatrick, Tables of values of the modified
Mathieu function, Math. Comp. 14, 70 (1960).
CeAu, q), r=0(1)5, Ser(u, q), r=1(1)6;
u= .1(.1) 1; q = 1 (1)20.
i20.581 National Bureau of Standards, Tables relating to
Mathieu functions (Columbia Univ. Press, New
York, N.Y., 1951). Characteristic values be,(s),
bo&) for 0 5 s 5100, along with P*, interpolable
to 8D; coefficients Dek(s), Do&) and conversion
factors for ce,(q), sel(q), same range, without
differences but interpolable to 9D with Lagran-
gian formulas of order 7. Joining factors
sirgg,.,, s*rgo,v,sf..,, sf.,. along with P*; inter-.
polable to 85.
[20.591 J. A. Stratton, P. M. Morse, L. J. Chu and R. A.
Hutner, Elliptic cylinder and spheroidal wave
functions (John Wiley & Sons, Inc., New York,
N.Y., 1941). Theory and tables for bo, bl, bz,
bt, bd, b;, b;, bj, b;, and coefficients for Ser(s, z)
and So,(s, 2) for c=0(.2)4.4 and .5(1)4.5; mostly
55; c-aqt, b,=a,+2q, b:=b,+2q.
[20.60] T. Tamir, Characteristic exponents of Mathieu *
equatisns, Math. Comp. 16, 77 (1962). The
Floquet exponent v, of the first three stable
regions; namely r=O, 1, 2; q=.1(.1)2.5; a=r
(.l)r+l, 5D.
[20.61] J. C. Wiltse and M. J. King, Values of the Mathieu
functions, The Johns Hopkins Univ. Radiation
Laboratory Technical Report AF-53, Baltimore,
Md. (1958). (Notation of [20.581 used:
ce,(v, q)/A,sen(o,q)/B for 12 values of q betwe2
.25 and 10 and from 8 to 14 values of u; dr/2
M&(u, q), Mg(u, q), j = 1 , 2 for 6 to 8
values of q between .25 a n d 1 0 and about 20
values of u, r=O, 1, 2; dr/2 M C S Y - I U ~ , q),
dT/2 M s ! ~ ( - ~ uq),~ ,r=O, 1, 2 for about 9 values
of u and q, 2 to 4 D in all.
[20.62] J. C. Wiltse and M. J. King, Derivatives, zeros,
and other data pertaining to Mathieu functions,
The Johns Hopkins Univ. Radiation Laboratory
Technical Report AF-57, Baltimore, Md.
(1958).
(20.631 S. J. Zaroodny, An elementary review of the
Mathieu-Hill equation of real variable based on
numerical solutions, Ballistic Research Labora-
tory Memorandum Report 878, Aberdeen
Proving Ground, Md. (1955). Chart of the
characteristic exponent.
See also [20.18]. It contains, among other
tabulations, values of a,, b, and coefficients for
ce.(z, q), se& q), q=40(20)100(50)200; 5D,
r 52.
*See page 11.