Lecture Notes in Differential Equations - Bruce E. Shapiro
Lecture Notes in Differential Equations - Bruce E. Shapiro Lecture Notes in Differential Equations - Bruce E. Shapiro
94 LESSON 11. PICARD ITERATION in which the Method of Successive Approximations converges to the unique solution of the initial value problem ⎫ dy = f(t, y) ⎬ dt ⎭ y(t 0 ) = y 0 (11.20) The proof of this theorem is quite involved and will be discussed in the sections 12 and 13. The procedure for using the Method of Successive Approximations is summarized in the following box. 1 Procedure for Picard Iteration To solve y ′ = f(t, y) with initial condition y(t 0 ) = y 0 : 1. Construct the first 3 iterations φ 0 , φ 1 , φ 2 , φ 3 . 2. Attempt to identify a pattern; if one is not obvious you may need to calculate more φ n . 3. Write a formula for the general φ n (t) from the pattern. 4. Prove that when you plug φ n (t) into the right hand side of equation (11.15) you get the same formula for φ n+1 with n replaced by n + 1. 5. Prove that φ(t) = lim n→∞ φ n converges. 6. Verify that φ(t) solve the original differential equation and initial condition. 1 The Method of Successive Approximations is usually referred to as Picard iteration for Charles Emile Picard (1856-1941) who popularized it in a series of textbooks on differential equations and mathematical analysis during the 1890’s. These books became standard references for a generation of mathematicians. Picard attributed the method to Hermann Schwartz, who included it in a Festschrift honoring Karl Weierstrass’ 70’th birthday in 1885. Guisseppe Peano (1887) and Ernst Leonard Lindeloff (1890) also published versions of the method. Since Picard was a faculty member at the Sorbonne when Lindeloff, also at the Sorbonne, published his results, Picard was certainly aware of Lindeloff’s work. A few authors, including Boyce and DiPrima, mention a special case published by Joseph Liouville in 1838 but I haven’t been able to track down the source, and since I can’t read French, I probably won’t be able to answer the question of whether this should be called Liouville iteration anytime soon.
95 Example 11.1. Construct the Picard iterates of the initial value problem } y ′ = −2ty (11.21) y(0) = 1 and determine if they converge to the solution. In terms of equations (11.12) through (11.15), equation (11.21) has f(t, y) = −2ty (11.22) t 0 = 0 (11.23) y 0 = 1 (11.24) Hence from (11.12) φ 0 (t) = y 0 = 1 (11.25) f(s, φ 0 (s)) = −2sφ 0 (s) = −2s (11.26) φ 1 (t) = y 0 + = 1 − 2 ∫ t 0 ∫ t 0 f(s, φ 0 (s))ds (11.27) sds (11.28) = 1 − t 2 (11.29) We then use φ 1 to calculate f(s, φ 1 (s)) and then φ 2 : f(s, φ 1 (s)) = −2sφ 1 (s) = −2s(1 − s 2 ) (11.30) φ 2 (t) = y 0 + = 1 − 2 ∫ t 0 ∫ t 0 f(s, φ 1 (s))ds (11.31) s(1 − s 2 )ds (11.32) = 1 − t 2 + 1 2 t4 (11.33) Continuing as before, use φ 2 to calculate f(s, φ 2 (s)) and then φ 3 : f(s, φ 2 (s)) = −2sφ 2 (s) = −2s (1 − s 2 + 1 ) 2 s4 (11.34) φ 3 (s) = y 0 + = 1 − 2 ∫ t 0 ∫ t 0 f(s, φ 2 (s))ds (11.35) s (1 − s 2 + 1 ) 2 s4 ds (11.36) = 1 − t 2 + 1 2 t4 − 1 6 t6 (11.37)
- Page 51 and 52: Lesson 6 Exponential Relaxation One
- Page 53 and 54: 45 Exponential Runaway First we con
- Page 55 and 56: 47 Figure 6.2: Illustration of the
- Page 57 and 58: 49 This is identical to with Theref
- Page 59 and 60: 51 this becomes a first-order ODE i
- Page 61 and 62: Lesson 7 Autonomous Differential Eq
- Page 63 and 64: 55 Figure 7.1: A plot of the right-
- Page 65 and 66: 57 Figure 7.2: Solutions of the log
- Page 67 and 68: 59 Figure 7.4: Solutions of the thr
- Page 69 and 70: Lesson 8 Homogeneous Equations Defi
- Page 71 and 72: 63 where z = y/t, the differential
- Page 73 and 74: Lesson 9 Exact Equations We can re-
- Page 75 and 76: 67 Now compare equation (9.2) with
- Page 77 and 78: 69 Hence dg dy = 0 =⇒ g = C′ (9
- Page 79 and 80: 71 From the first of equations (9.5
- Page 81 and 82: 73 Differentiating equations (9.81)
- Page 83 and 84: 75 This has the form Mdt + Ndy = 0
- Page 85 and 86: Lesson 10 Integrating Factors Defin
- Page 87 and 88: 79 Differentiating with respect to
- Page 89 and 90: 81 Proof. In each of the five cases
- Page 91 and 92: 83 as required by equation (10.31).
- Page 93 and 94: 85 Since M y ≠ N t , equation (10
- Page 95 and 96: 87 the revised equation (10.100) is
- Page 97 and 98: 89 Substituting (10.129) into (10.1
- Page 99 and 100: Lesson 11 Method of Successive Appr
- Page 101: 93 because the integral is zero (th
- Page 105 and 106: 97 We can then plug this expression
- Page 107 and 108: Lesson 12 Existence of Solutions* I
- Page 109 and 110: 101 • Interchangeability of Limit
- Page 111 and 112: 103 But on the square −1 ≤ t
- Page 113 and 114: 105 Thus lim φ n = φ 0 + lim n→
- Page 115 and 116: 107 because the right hand side doe
- Page 117 and 118: Lesson 13 Uniqueness of Solutions*
- Page 119 and 120: 111 The proof of theorem (13.1) is
- Page 121 and 122: 113 But δ(t) is an absolute value,
- Page 123 and 124: 115 Substituting (13.66) into (13.6
- Page 125 and 126: Lesson 14 Review of Linear Algebra
- Page 127 and 128: 119 Definition 14.10. An m × n (or
- Page 129 and 130: 121 Definition 14.19. Matrix Multip
- Page 131 and 132: 123 In practical terms, computation
- Page 133 and 134: 125 Simplifying 4x − 2 + 3z = 0 (
- Page 135 and 136: Lesson 15 Linear Operators and Vect
- Page 137 and 138: 129 Example 15.3. By a similar argu
- Page 139 and 140: 131 Therefore ‖y + z‖ 2 ≤ ‖
- Page 141 and 142: 133 Definition 15.5. Two vectors y,
- Page 143 and 144: Lesson 16 Linear Equations With Con
- Page 145 and 146: 137 Hence both r = 1 and r = 3. Thi
- Page 147 and 148: 139 The second order linear initial
- Page 149 and 150: 141 The general solution to is give
- Page 151 and 152: Lesson 17 Some Special Substitution
95<br />
Example 11.1. Construct the Picard iterates of the <strong>in</strong>itial value problem<br />
}<br />
y ′ = −2ty<br />
(11.21)<br />
y(0) = 1<br />
and determ<strong>in</strong>e if they converge to the solution.<br />
In terms of equations (11.12) through (11.15), equation (11.21) has<br />
f(t, y) = −2ty (11.22)<br />
t 0 = 0 (11.23)<br />
y 0 = 1 (11.24)<br />
Hence from (11.12)<br />
φ 0 (t) = y 0 = 1 (11.25)<br />
f(s, φ 0 (s)) = −2sφ 0 (s) = −2s (11.26)<br />
φ 1 (t) = y 0 +<br />
= 1 − 2<br />
∫ t<br />
0<br />
∫ t<br />
0<br />
f(s, φ 0 (s))ds (11.27)<br />
sds (11.28)<br />
= 1 − t 2 (11.29)<br />
We then use φ 1 to calculate f(s, φ 1 (s)) and then φ 2 :<br />
f(s, φ 1 (s)) = −2sφ 1 (s) = −2s(1 − s 2 ) (11.30)<br />
φ 2 (t) = y 0 +<br />
= 1 − 2<br />
∫ t<br />
0<br />
∫ t<br />
0<br />
f(s, φ 1 (s))ds (11.31)<br />
s(1 − s 2 )ds (11.32)<br />
= 1 − t 2 + 1 2 t4 (11.33)<br />
Cont<strong>in</strong>u<strong>in</strong>g as before, use φ 2 to calculate f(s, φ 2 (s)) and then φ 3 :<br />
f(s, φ 2 (s)) = −2sφ 2 (s) = −2s<br />
(1 − s 2 + 1 )<br />
2 s4<br />
(11.34)<br />
φ 3 (s) = y 0 +<br />
= 1 − 2<br />
∫ t<br />
0<br />
∫ t<br />
0<br />
f(s, φ 2 (s))ds (11.35)<br />
s<br />
(1 − s 2 + 1 )<br />
2 s4 ds (11.36)<br />
= 1 − t 2 + 1 2 t4 − 1 6 t6 (11.37)