Lecture Notes in Differential Equations - Bruce E. Shapiro

Lecture Notes in Differential Equations - Bruce E. Shapiro Lecture Notes in Differential Equations - Bruce E. Shapiro

bruce.shapiro.com
from bruce.shapiro.com More from this publisher
21.04.2015 Views

294 LESSON 30. THE METHOD OF FROBENIUS By Taylor’s theorem p (k) (t 0 ) = k!p k and S (n−k) (t 0 ) = (n − k)!c n−k , where p k and c k are the Taylor coefficients of p(t) and S(t), respectively, Similarly, the third term in (30.97) is D (n) {[α(α − 1) + αp(t) + q(t)] S} (t 0 ) n∑ ( ) n { = [α(α − 1) + αp(t) + q(t)] (k) (t k 0 )} S (n−k) (t 0 ) (30.109) k=0 Extracting the first (k = 0) term, D (n) {[α(α − 1) + αp(t) + q(t)] S} (t 0 ) = [α(α − 1) + αp 0 + q 0 ] S (n) (t 0 ) n∑ ( ) n [ + αp (k) (t k 0 ) + q (k) (t 0 )] S (n−k) (t 0 ) (30.110) k=1 By the indicial equation, the first term is zero. Substituting the formulas for the Taylor coefficients and simplifying, n∑ ( D (n) n {[α(α − 1) + αp(t) + q(t)] S} (t 0 ) = [αk!p k) k + k!q k ] (n−k)!c n−k k=1 Substituting (30.111) and (30.102) into (30.97), (30.111) 0 =n(n − 1)c n n! + nn!(2α + p 0 )c n n−1 ∑ n∑ + n!(n − k)p k c n−k + n! [αp k + q k ] c n−k (30.112) k=1 Rearranging and simplifying, k=1 n−1 ∑ n(n − 1 + 2α + p 0 )c n + [(n − k + α)p k + q k ]c n−k = 0 (30.113) k=1 Letting r > 0 be the radius of convergence of the Taylor series for p and q, then there is some number M such that |p k | r k ≤ M and |q k | r k ≤ M. Then |n(n − 1 + 2α + p 0 )c n | ≤ M n−1 ∑ k=1 1 r k [|n − k + α| + 1] |c n−k| (30.114) Define the numbers C 0 = |c 0 | , C 1 = |c 1 |, and, for n ≥ 2, |n − 1 + 2α + p 0 | C n = M n n−1 ∑ k=1 1 r k [|n − k + α| + 1] C n−k (30.115)

295 Let j = n − k. Then |n − 1 + 2α + p 0 | C n = M n−1 ∑ nr n r j [|j + α| + 1] C j (30.116) Evaluating (30.116) for n + 1, j=1 |n + 2α + p 0 | C n+1 = M (n+1)r n+1 ∑ n j=1 rj [|j + α| + 1] C j (30.117) Therefore |n + 2α + p 0 | C n+1 r(n + 1) = {n |n − 1 + 2α + p 0 | + M [|n + α| + 1]} C n (30.118) so that C n+1 (t − t 0 ) = n |n − 1 + 2α + p 0| + M [|n + α| + 1] C n |n + 2α + p 0 | (n + 1) Hence = n|n−1+2α+p 0| n 2 |n+2α+p 0|(n+1) lim n→∞ + M |n+α|+1 n (t − t 2 0 ) r n 2 ∣ C n+1 (t − t 0 ) ∣∣∣ ∣ = C n ∣ t − t 0 r (t − t 0 ) r (30.119) (30.120) ∣ < 1 (30.121) Therefore by the ratio test ∑ ∞ k=0 C k(t − t 0 ) k converges; by the comparison test, the sum ∑ ∞ n=0 c n(t − t 0 ) n also converges. Equations (30.96) and (30.113) give formulas for the c n and so that c n = c 1 = − q 1 + αp 1 p 0 + 2α c 0 (30.122) n−1 −1 ∑ [(n − k + α)p k + q k ] c n−k (30.123) n(n − 1 + 2α + p 0 ) k=1 y = (t − t 0 ) α is a solution of (30.70) so long as ∞ ∑ n=0 c n (t − t 0 ) n (30.124) n − 1 + 2α + p 0 ≠ 0. (30.125)

294 LESSON 30. THE METHOD OF FROBENIUS<br />

By Taylor’s theorem p (k) (t 0 ) = k!p k and S (n−k) (t 0 ) = (n − k)!c n−k , where<br />

p k and c k are the Taylor coefficients of p(t) and S(t), respectively,<br />

Similarly, the third term <strong>in</strong> (30.97) is<br />

D (n) {[α(α − 1) + αp(t) + q(t)] S} (t 0 )<br />

n∑<br />

( ) n {<br />

= [α(α − 1) + αp(t) + q(t)] (k) (t<br />

k<br />

0 )}<br />

S (n−k) (t 0 ) (30.109)<br />

k=0<br />

Extract<strong>in</strong>g the first (k = 0) term,<br />

D (n) {[α(α − 1) + αp(t) + q(t)] S} (t 0 )<br />

= [α(α − 1) + αp 0 + q 0 ] S (n) (t 0 )<br />

n∑<br />

( ) n [<br />

+ αp (k) (t<br />

k<br />

0 ) + q (k) (t 0 )]<br />

S (n−k) (t 0 ) (30.110)<br />

k=1<br />

By the <strong>in</strong>dicial equation, the first term is zero. Substitut<strong>in</strong>g the formulas<br />

for the Taylor coefficients and simplify<strong>in</strong>g,<br />

n∑<br />

(<br />

D (n) n<br />

{[α(α − 1) + αp(t) + q(t)] S} (t 0 ) = [αk!p<br />

k)<br />

k + k!q k ] (n−k)!c n−k<br />

k=1<br />

Substitut<strong>in</strong>g (30.111) and (30.102) <strong>in</strong>to (30.97),<br />

(30.111)<br />

0 =n(n − 1)c n n! + nn!(2α + p 0 )c n<br />

n−1<br />

∑<br />

n∑<br />

+ n!(n − k)p k c n−k + n! [αp k + q k ] c n−k (30.112)<br />

k=1<br />

Rearrang<strong>in</strong>g and simplify<strong>in</strong>g,<br />

k=1<br />

n−1<br />

∑<br />

n(n − 1 + 2α + p 0 )c n + [(n − k + α)p k + q k ]c n−k = 0 (30.113)<br />

k=1<br />

Lett<strong>in</strong>g r > 0 be the radius of convergence of the Taylor series for p and<br />

q, then there is some number M such that |p k | r k ≤ M and |q k | r k ≤ M.<br />

Then<br />

|n(n − 1 + 2α + p 0 )c n | ≤ M<br />

n−1<br />

∑<br />

k=1<br />

1<br />

r k [|n − k + α| + 1] |c n−k| (30.114)<br />

Def<strong>in</strong>e the numbers C 0 = |c 0 | , C 1 = |c 1 |, and, for n ≥ 2,<br />

|n − 1 + 2α + p 0 | C n = M n<br />

n−1<br />

∑<br />

k=1<br />

1<br />

r k [|n − k + α| + 1] C n−k (30.115)

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!