Frobenius Series Method
Frobenius Series Method
wheres is a number that may be complex. Our concern in this section is with a
generalization of the class of equations (3.23). An equation that can be written
in the form
P(x)
p(x)=--, q(x) = Q(x) 2 .
X- Xo (x - x0)
Thus Eq. (3.25) has a regular singular point at x0 if and only if the functions
(x - x0)p(x),
are analytic at x0. It turns out that Eq. (3.24) possesses at least one, and some-
times two solutions of the form
0()
where s is a number+ that need not be an integer. The procedure for finding
solutions of the type (3.26) is known as the method of Frobenius. We illustrate
the method with some examples.
3 I+x
p(x) = 2x' q(x) =- 2x2 '
is said to have a regular singular point at xo if b1, b2, ... , b. are analytic at x0 .
t Jt is possible thats may be complex, but this does not happen in the classical equations
in which we are most interested.
3.4 Regular Singular Points 15 1
y'(x) = L (n + s)Anxn+s-1
n=O
and
00
- L Anxn+s - L Anxn+s+I = 0.
n=O n=O
The last sum in this equation begins with a term involving xs+ 1; the remaining
series start with a term involving X 5• In order to combine terms with like
powers of x, we separate out those terms with x' and make a shift of index
(n ---t n - I) in the last series. The result is
(X)
Then
I
Ai =-Ao
I ·5 '
l l
A 2 =-A 1 =---A 0
2· 7 (I · 2)(5 · 7) '
I
A,.=------ A0•
n ! 5 · 7 · · · ( 2n + 3)
Taking A0 = 1, we arrive at the specific" series solution"
y l ( X) = X I I2 [ 1 + f= n ! 5 · 7 · · · (2n + 3)
n 1
X11 ] •
(3.30)
I
A 11 =---A 11 1 n ~ l.
n(2n - 3) - '
I
A ----A
1-l·(-l) o,
I I
A2 = 2 . ( l) A I = l . 2( - 1 )(1) AO'
J l
A3 =-Ai =------Ao,
3 ·3 1 · 2 · 3( - I) · I · 3
I
A,.=-------- A0•
n ! ( - l) I · 3 · · · (2n - 3)
Thus a second series solution of Eq. (3.27) is
00
yi(x)=x-1 [ 1-x-
n
I------· x"
= n ! 1 . 3 ... (211 - 3)
2
] (3.31)
where the power series converges in some interval ( -R, R), is a solution of the
differential equation (3.27) if and only if the exponents and the coefficients A 1
satisfy the relations (3.28) and (3.29). Now the power series in formulas (3.30)
and (3.31) converge everywhere, as can be verified by the use of Theorem 3.1.
Consequently, the functions y1 and Yi are both solutions on the interval
(0, co ). These solutions are linearly independent. For if
C1Y/X) + C2 yi(x) = 0
for x > 0, then, letting x-+ 0, we see that c2 must be zero. This is because
Yi(x)-+ co as x-+ 0. We now have
C1Y1(x) =0
for x > 0. But y1 is not the zero function so c1 = 0 also. Hence y1 and y2 are
linearly independent.
y'(x) = L (n + s)A 11
x"+s- l,
11=0
00
and
= f ( I Qn-k)x"+s · Ak
11=0 k=O
154 Ill Series Solutions
Here we have used Theorem 3.2 to multiply the power series. Upon substitut-
ing the various quantities into the differential equation (3.33) and combining
like powers of x, we obtain the equation
for n ~ 1. By collecting the terms that involve A11, we can write this relation
as
n- 1
[(n + s)(n + s - 1) + (n + s)P0 + Qo]A11 = - L [(k + s)Pn-k + Qn-dAk,
k=O
00
corresponding to the larger exponent s1. It can be shown that the power
series in this formula actually converges, at least for I xi < R, and that the
function y1 is a solution of the differential equation, at least on the interval
(0, R). We refer the reader to Exercise 3, Section 3.8, or to Coddington (1].
Considering now the smaller exponent s2 , we sec that
00
corresponding to the exponent s2• The power series in this formula also
converges for Ix I < R and y2 is a solution, at least on the interval (0, R).
However, if s1 - s2 = N, where N is a positive integer, then f(s2 + n) is
zero when and only when n = s1 - s2 = N. In this case the recurrence relation
(3.35) becomes, for n = N,
N-1
0 ·AN= L gN(k, s )Ak. 2
k=O
and any value for AN will do. ([n particular we can choose AN = 0.) In this
case we again obtain a series solution of the form (3.37). The case where
s1 - s2 is a positive integer will be treated in a later section. If the exponents
are equal, there is evidently only one solution of the type (3.34). We shall
consider this case in the next section.
156 Ill Series Solutions
13. Show that the solutions (3.36) and (3.37) are linearly independent.
14. Let the function P, Q, and F be analytic at x = 0. Show that the equa-
tion
x2y" + xP(x)y' + Q(x)y = x"F(x)
possesses at least a formal solution of the form
00
y(x) = xa L A"xn
n=O
When the exponents s1 and s2 of the differential equation (3.33) are equal,
we can find only one solution of the form (3.34). In order to get some idea
as to how a second solution can be found, suppose we examine a Cauchy-
Euler equation whose exponents are equal. let the equation be
y(x, s) =X 5
,
we have
(3.38)
aLy(x, s) = L 8y(x, s)
as as
= [2(s - s1) + (s - s1)2 In x]x 5
,
where the first equality involves merely the interchange of the order of dif-
ferentiation betweens and x. Upon settings= s1, we see that
8y(x, s)
L--
I =0
as s=s1
Y2(x)
oy(x. s)
= ---
OS
I s=s1
=X 51
In x.
158 Ill Series Solutions
Writing
rY)
The functions An that are so defined are rational functions of s, and hence
possess derivatives of all orders for Is - s I I < I.
Let us assume that the coefficients Ais) in the series (3.40) have been
chosen in the manner just described. Then from Eq. (3.41) we have
(3.43)
oy(x, s) I = L oy(x, s) I = 0.
OS S'"'S1 OS s=s1
3.5 The Case of Equal Exponents 159
ay(x, s)
Yi (X ) = "
OS
00 <X)
= X
51
L AnCsi)Xn In X + x L A~(si)Xn s1
n=O n=l
or
co
yi(x) = y1(x) In x +X 51
L A~(s )x",1 (3.45)
n=I
Ly = x y" + 3xy' + (I
2
- x)y = 0.
Setting
00
y(x, s) = L A,.(s)xn+s,
n=O
we find that
00
Ao
A1(s) = (s + 2)2'
A s - A1(s) - Ao
i( ) - (s + 3)2 - (s + 2)2 (s + 3)2'
and in general,
Ao
An(s) = i ' n ~ l. (3.46)
(s + 2) (s + 3) 2
• • • (s +n+ l )2
160 Ill Series Solutions
Settings= s1 = - l, we have
Ao Ao
A(-1)=-----
" 12,22 ... n2-(n!)2·
1 oo x"
Y1(x) = x- L-( ')2·
n=O n.
A~(s) [ 1 1 1 ]
AnCs) = -2 s +2+ s + 3 + ··· + s + n + l ·
Then
_
A"_
'(_
-_l) = -2[1 + ~ + ... + ~]
An(-1) 2 n
or
A' ( - 1) = - 2 </>( n) n ~ 1,
n (n !)2 ,
(3.47)
From the general formula (3.45) we see that a second solution of the differen-
tial equation is
oo </>(n)
yi(x) = y1(x) In x - 2x-1 L -- x" 2
11= I (n !)
In Exercises 1-10, verify that the exponents relative to x = 0 are equal, and
find the general solution by using the method described in this section.
y(x) =X 5
L A"x". (3.49)
"=O
(In the special case of a Cauchy-Euler equation, there are always two such
solutions.) In any case, there is always a solution
00
Y1(s) = X
51
L A,.x" (3.50)
"=O
of this type, corresponding to s1•
162 Ill Series Solutions
We now consider the case where there is only one solution of the form
(3.49). Substituting the series
<X>
y(x, s) = xs L A"(s)x"
",- 0
(3.51)
where
and
J(s + n) = (s + n ·- s2 - N)(s + N - si)
when n ~ I. In particular, f(s + N) is zero when s = s2, because
f(s + N) = (s - s 2 )(s + N - s2).
This is the only one of the quantities f(s + n) that vanishes when s = s2.
Let us choose the coefficient functions A" (where A0 is a fixed nonzero con-
stant) so that they satisfy the recurrence relation
n- l
J(s + n)An(s) = L gik, s)Ais). (3.52)
k=O
Then the functions A1, A2, •.• , A,.._1 are analytic at s = s2 but for n ~ N, the
quantities A"(s) contain the factor s - s2 in the denominator. They may
become infinite ass approaches s2. However, the functions B", where
are analytic at s = s2 and satisfy the recurrence relation (3.52) not only for s
near s2 but also for s =Si. (To see this, multiply through in Eq. (3.52)
by S - Sz .)
Let
y(x, s) = (s - s2)y(x, s)
<X>
= xs L Bn{s)x".
n=O
Because of the occurence of the factor (s - s2)2 in the last expression, it follows
that each of the quantities
oy(x, s)
Y2 ( X) =
OS s= .,2
.Y1(x) = X 52
L B/s2)xn
n=N
eo
= xs2+N L Bn+isi)x"
n=O
00
=X 5
' °" B
L,
n=O
I
n+Ns;s 2 )x" ·
Thus ji1 must simply be a multiple of the solution y1 given in formula (3.50).
In fact,
oy(x, s) = ~ [xs
OS OS n=O
f
B/s)xn]
00 00
yi(x) = y1(x) In x + X 52
L B~(s )x"
2
n=O
or
(3.53)
It can be shown (Exercise 6, Section 3.8) that the power series in this formula
converges, at least for Ix I < R, and that the function Yi is a solution of the
differential equation, at least on the interval (0, R).
164 Ill Series Solutions
The exponents for the equation are s1 =0 and s2 = -1. The recurrence
relation for the coefficients A" is
(n + s)(n + s + l)An(s) = A"_1(s), n ';?:: 1.
Ao A 2 (s) - Ao
Ai(s) = (s + l)(s + 2)' - (s + I )(s + 2)2(s + 2) ,
and in general,
A (s) = Ao
" (s + l)(s + 2)2(s + 3)2 .. · (s + n)2(s +n + l)
for n ';?:: 2. Setting s = s1 = 0 in these formulas, we find that
Ao Ao
A (s1) = -------
2 3 2 = --- , n ';?:: 0.
" 1·2 ·3 .. ·n (n+I) n!(n+l)!
Therefore a solution of the equation which corresponds to the exponent
s1 = 0 is
oo x"
Yi(x) = n~o n!(n + l)!.
The quantity A1 (s)(N = I in this example) becomes infinite ass approaches
s2 = -1, because of the factor (s + 1) in its denominator. Hence the equation
does not possess a second solution of the form (3.49) corresponding to the
exponent s2 . The second solution is therefore logarithmic.
The functions
B0(s) = (s + l)A0,
Ao
B1(s) = (s + I)A1(s) = -- ,
s+ 2
Ao n ';?:: 2,
Bn(s) = (s + I)AnCs) = (s+ 2)2(s+ 3)2 ... (s+n )2(s+n+ I)'
3.6 The Cas a W han the Exponents Differ by an Integer 165
into the differential equation and determining the coefficients en. However,
it is difficult to find a general formula for en using this method.
s(s + 2) = 0,
and that the coefficients Ai must satisfy the conditions
(s + I)(s + 3)A1 = 0,
(s + 2)(s + 4)A2 = 0,
and
+ s)(n + s + 2)An = A,,_3,
(n n ~ 3.
The exponents are s1 = 0 and s2 = -2. For the larger exponent s1, we have
A1 = 0,
A 2 =0,
A,,_3
An= n(n + 2)' n ~ 3.
166 Ill Series Solutions
All the coefficients Ai vanish except those whose subscripts are multiples
of three. We have
I
A3 =--Ao
3 ·5 '
I I
A-
6 A----A
3 0'
- 6. 8 - (3 . 6)(5 . 8)
and in general,
1
AJm = -------------- Ao
(3 · 6 · 9 .. · 3m)[5 · 8 · 11 · .. (3m + 2)]
I
=---------A 0
3mm ! [5 · 8 · 11 · .. (3m + 2)] ·
(n - 2)nAII =An-], n ~ 3.
(Note that A2 is the critical coefficient, since N = 2 in this example.) Here
A2 is arbitrary, and we may choose A2 = 0. A solution that corresponds
to the exponent s2 is found to be
3
oo x m ]
yi(x)=x-2 [ l+ .
m~l 3mm! [] · 4 · 7 ... (3m - 2)]