Escolar Documentos
Profissional Documentos
Cultura Documentos
2016年9月19日 下午 06:32
微分方程式
Chapter 6
管傑雄
台大電機系
Chapter 6 Series Solutions of Linear Equations
Second-order differential equation: y” + P(x)y’ + Q(x)y = 0依P(x)及Q(x)是否為解析而分類
為
(1) Ordinary Point: 若P及Q兩者均解析,則y(x)有兩個獨立解析解
(2) Singular Point: 其它條件下,則y(x)可能有二個、一個或無解析解,但仍是保有兩個
獨立解
a series centered at x = a
Convergence
For a specific value of x a power series is a series of constants. If the series equals a
finite real constant for the given x, then the series is said to converge at x. Otherwise it
is said to diverge at x.
Interval of Convergence
Radius of Convergence
A power series converges for | x - a | < R where R is the radius of convergence.
(1) R = 0: A singe point of convergence
(2) R = finite
(3) R = infinite: Converging for all x.
Convergence at an Endpoint
The power series converges at x = R + a or x = -R + a
Absolute Convergence
Ch6 第 1 頁
Absolute Convergence
is convergent.
Finding the Interval of Convergence
Ratio Test of a series
indicates that the series converges absolutely for L < 1. Therefor, the radius of convergence is
If f(x) has a radius of convergence R > 0, then f(x) is continuous, differentiable, and
integrable on the interval of convergence.
, R > 0 for all real number x in the interval of convergence, then cn = 0 for all n.
Analytic at a Point
A function f is analytic at point a if it can be represented by a power series in x - a with
a positive radius of convergence.
Arithmetic of Power Series
Two power series:
then
,
then we substitute it back into the equation and try to find the recursive relation in the
Ch6 第 2 頁
then we substitute it back into the equation and try to find the recursive relation in the
coefficient cn.
.
Note: (1)逐項計算(2)統一代表項,改變足標碼(3)足標碼由最小值討論至代表項
Now setting all the coefficients into zero gives c1 = 0 for n = 0 and the recursive relationship:
for
That is,
2 c2 = 2 c0 c1 = 0
4 c4 = 2 c2 3c3 = 2c1
6 c8 = 2 c4 5c5 = 2c3
................. ................
c2k+1 = 0
The solution is
it is true for all x. This indicates that the radius of convergence R is infinite.
Existence of Solutions
If x = x0 is an ordinary point of the equation, we can always find two linearly
independent analytic solutions in the form of power series centered at x0:
, where R1 is the distance for x0 to the closest singular point (real or complex).
Ch6 第 3 頁
, a power series solution will converge at least for
. The assumption
leads to
For
, the recursive relation in the coefficient cn is given by
n(n - 1)cn + (n + 2)(n + 1)cn+2 + ncn - cn = 0
i.e.,
(n + 2)(n + 1)cn+2 + (n + 1)(n - 1)cn = 0.
For n = 0, and 1, we find that
2c2 = c0
c3 = 0
Hence,
c0 and c1 are arbitrary.
, and c3 = 0.
for
.
Iteration of the last formula gives
c5 = 0
c7 = 0
................
We conclude that for
c2k-1 = 0
and
.
The solution is
Ch6 第 4 頁
With the assumption of y as a series of x:
,
we can find
i.e.,
Hence, with c0 and c1 being arbitrary constants, all other coefficients can be represented by
them:
.................................
The solution is
Forbenius Theorem
If x = x0 is a regualr singular point of the equation, then there exists at least one series
solution of the form
where the number r is a constant which must be determined. The series will converge at least
on some interval 0 < x - x0 < R.
Note: r = 0或正整數,則解變為analytic。
Ch6 第 5 頁
Example: Solve 3xy" + y' -y = 0.
Solution: Since x = 0 is a regular singular point of the equation, we try a solution of the form
i.e.,
For n = -1,
3c0r(r - 1) + c0r = c0r(3r - 2) = 0
For
,
3cn+1(n + 1 + r)(n + r) + cn+1(n + 1 + r) - cn = 0
Since nothing is gained by taking c0 = 0, we must then have
r(3r - 2) = 0
for
The two roots of the r equation gives two independent recursive relationships:
(1) r = 0,
which gives
.
The solution form is
(2) r = 2/3,
which gives
Indicial Equation
The equation of r is called the indicial equation of the problem. For a second-order
differential equation, it is a quadratic equation in r that results from equating the total
coefficient of the lowest power of x to zero.
Ch6 第 6 頁
Now we classify the equation according to the property of the two roots.
Case I Roots Not differing by an Integer
If r1 and r2 are distinct and do not differ by an integer, then there exist two linearly
independent solutions of the form
where
and
. Note that if the difference of r1 and r2 is an integer, the two solution forms may merge into a
single one.
.
The differential equation leads to
That is,
i.e.,
for
.
The solution form is
Ch6 第 7 頁
i.e.,
. If we choose the smaller one, then we may find that c0 = c1 = c2 = .... = cN-1 = 0, i.e., the
solution form for the smaller one merges into the one for the larger root. Then the
independent solution for the smaller root can be found from the formula
Therefore,
Ch6 第 8 頁
Hence, y2 becomes as
i.e.,
c7 is arbitrary (for n = 6)
and for
,
i.e.,
Ch6 第 9 頁
.
The general solution is
,
and the equation leads to
For the case of r = 0, the solution form is exactly the same as above. This indicates that the
Frobenius' method can only find one solution because the difference of the two roots is a
positive integer. The other solution can be found by
(with c2 = 1)
That is,
Ch6 第 10 頁
Hence,
Ch6 第 11 頁
i.e.,
i.e.,
The solution is
解得
Ch6 第 12 頁
where c is an arbitrary constant.
Inverse Laplace Transform得
故僅能找到一獨立解。另一解因無Laplace Transform而無法尋得。
Solution: x = 0 is a regular singular point for the equation. We will try to find some r and use
the transformation form y = xru(x) to get an ordinary equation for u(x). The equation for u is
The condition under which the equation becomes ordinary at x = 0 and u(x) has two analytic
solutions is
2r + 1 = 0 and r2 - 1/4 = 0
Therefore, r = -1/2 and the equation is
u" + u = 0
The solution for u(x) is
u = c1cosx + c2sinx
and the solution y(x) is
[Summary]
For the case in which x = 0 is a regular singular point,
The differential equation
y" + P(x)y' + Q(x)y = 0
has the following forms of coefficient functions:
A special case: If q0 = 0, we can conclude that y has at least one analytic solution. We try the
solution form
, and the indicial equation is r(r - 1)c0 + p0rc0 = 0. The roots are r = 0 or r = 1 - p0. The root r =
0 guarantees the solution must be analytic. The sufficient condition that the equation at a
regular singular point has at least one analytic solution is that xQ(x) is analytic.
If we set y = xru(x) and hope that u(x) is an analytic function at x = 0, then
That is,
Ch6 第 13 頁
The condition that u(x) has at least one analytic solution is
r(r - 1) + rp0 + q0 = 0
The condition that u(x) has two analytic soltions is
r(r - 1) + rp0 + q0 = 0 and 2r + p0 = 0 and rp1 + q1 = 0
Legendre's Equation:
(1 - x2)y" - 2xy" + n(n + 1)y = 0
where n is a nonnegative integer.
Solution of Bessel's Equation
Since x = 0 is a regular singular point for Bessel's equation, we assume that the solution
form is
,
and then
When n = 0,
c0r(r - 1) + c0r - 2c0 = 0,
i.e., the indicial equation
r2 - 2 = 0
When n = 1,
c1(r + 1)r + c1(r + 1) - 2c1 = 0
i.e.,
c1[(r + 1)2 - 2] = 0
When
, the iteration equation is
cn[(n + r)2 - 2] + cn-2 = 0
The two roots of the indicial equation is and -. For the case of r1 = , we can conclude
c1= 0
Ch6 第 14 頁
for
.
Hence,
c2n-1 = 0
, n = 1, 2, 3, ....
It is a standard practice to choose c0 to be a specific value -- namely
, n = 0, 1, 2, ....
It follows that one solution is
If
Also, for the second exponet r2 = -, we obtain, in exactly the same manner,
The above two functions are called Bessel's functions of the first kind of order and -.
Ch6 第 15 頁
Therefore, we have (n+1) = n! for a positive integer n. For -1 < x < 0, Gamma function is
defined as
.
The same formula may then be used to the open interval (-2,-1), then to the open interval
(-3,-2), and so on. The graph of Gamma function thus extended is shown in the following
graph.
(2) If is half an odd positive integer, then J(x) and J-(x) are linearly independent. If fact,
J(x) is an elementary function if the order is half an odd integer. For example,
and
Example: Solve
Ch6 第 16 頁
.
Solution: Since = 1/2, the general solution is
y = c1J1/2(x) + c2J-1/2(x)
and the function J(x) are linearly independent. Thus another form of the general solution is y
= c1J(x) + c2Y(x). As
, where m is an integer, Ym(m) exists
and is linearly independent of Jm(x). Hence, for any value of , the general solution can be
written as
y = c1J(x) + c2Y(x)
Y(x) is sometimes called Neumann's function or the Bessel's function of the second kind
of order .
Ch6 第 17 頁
It can be shown that
where
(2)
Note; The above two formulas can be drived directly from the series definition of Bessel
functions.
(3)
(4)
Note: can be derived by carrying out the differentiations in (1) and (2).
Example: J1/2(x)
Solution: The function is
where
Therefore,
Ch6 第 18 頁
Since x = 0 is an ordinary point of Legendre's equation, we assume that the solution
form is
.
Therefore, the equation leads to
When k = 0,
2c2 + n(n + 1)c0 = 0, i.e.,
When k = 1,
6c3 - 2c1 + n(n + 1)c1 = 0, i.e.,
When
Thus, for at least |x| < 1, we obtain two linearly independent power series solutions
and
Notice that if n is an even integer, the first series terminates with xn, whereas y2(x) is an
ifinite series. Similarly when n is an odd integer, y1(x) is an infinite series while y2(x)
terminates with xn. That is, when n is a nonnegative integer, we obtain an nth-degree
polynomial solution of Legendre's equation.
In particular, we choose the specific values of c0 and c1:
c0 = 1 for n = 0
for n = 2, 4, 6, ...
and
c1 = 1 for n = 1
for n = 3, 5, 7, ...
Legendre Polynomials
The specific nth-degree polynomial solutions are called Legendre polynomials and are
denoted by Pn(x). They are
P0(x) = 1 P1(x) = x
Ch6 第 19 頁
P0(x) = 1 P1(x) = x
Properties
(i) Pn(-x) = (-1)nPn(x)
(ii) Pn(1) = 1
(iii) Pn(-1) = (-1)n
(iv) Pn(0) = 0, n = 1, 3, 5, ...
(v)
, n = 0, 2, 4, ....
Recurrence Relation
Using binomial series, we can formally show that.
.
The function on the left is called the generating function for Legendrepolynomials.
Differentiating both sides with respect to t gives
Ch6 第 20 頁
Differentiating both sides with respect to t gives
That is
k = 2, 3, 4, ...
The formula can be used to find Pk+1(x) in terms of Pk(x) and Pk-1(x).
Ch6 第 21 頁